poly_int: vectorizable_conversion
[official-gcc.git] / gcc / ChangeLog
blob87c70af60f2eba7a142326cdfe2da1947550b5c8
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_conversion): Treat the number
6         of units as polynomial.  Choose between WIDE and NARROW based
7         on multiple_p.
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 (simd_clone_subparts): New function.
14         (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
16 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
17             Alan Hayward  <alan.hayward@arm.com>
18             David Sherwood  <david.sherwood@arm.com>
20         * tree-vect-stmts.c (vectorizable_call): Treat the number of
21         vectors as polynomial.  Use build_index_vector for
22         IFN_GOMP_SIMD_LANE.
24 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
25             Alan Hayward  <alan.hayward@arm.com>
26             David Sherwood  <david.sherwood@arm.com>
28         * tree-vect-stmts.c (get_load_store_type): Treat the number of
29         units as polynomial.  Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
30         for variable-length vectors.
31         (vectorizable_mask_load_store): Treat the number of units as
32         polynomial, asserting that it is constant if the condition has
33         already been enforced.
34         (vectorizable_store, vectorizable_load): Likewise.
36 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
37             Alan Hayward  <alan.hayward@arm.com>
38             David Sherwood  <david.sherwood@arm.com>
40         * tree-vect-loop.c (vectorizable_live_operation): Treat the number
41         of units as polynomial.  Punt if we can't tell at compile time
42         which vector contains the final result.
44 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
45             Alan Hayward  <alan.hayward@arm.com>
46             David Sherwood  <david.sherwood@arm.com>
48         * tree-vect-loop.c (vectorizable_induction): Treat the number
49         of units as polynomial.  Punt on SLP inductions.  Use an integer
50         VEC_SERIES_EXPR for variable-length integer reductions.  Use a
51         cast of such a series for variable-length floating-point
52         reductions.
54 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
55             Alan Hayward  <alan.hayward@arm.com>
56             David Sherwood  <david.sherwood@arm.com>
58         * tree.h (build_index_vector): Declare.
59         * tree.c (build_index_vector): New function.
60         * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
61         of units as polynomial, forcibly converting it to a constant if
62         vectorizable_reduction has already enforced the condition.
63         (vect_create_epilog_for_reduction): Likewise.  Use build_index_vector
64         to create a {1,2,3,...} vector.
65         (vectorizable_reduction): Treat the number of units as polynomial.
66         Choose vectype_in based on the largest scalar element size rather
67         than the smallest number of units.  Enforce the restrictions
68         relied on above.
70 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
71             Alan Hayward  <alan.hayward@arm.com>
72             David Sherwood  <david.sherwood@arm.com>
74         * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
75         number of units as polynomial.
77 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
78             Alan Hayward  <alan.hayward@arm.com>
79             David Sherwood  <david.sherwood@arm.com>
81         * target.h (vector_sizes, auto_vector_sizes): New typedefs.
82         * target.def (autovectorize_vector_sizes): Return the vector sizes
83         by pointer, using vector_sizes rather than a bitmask.
84         * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
85         * targhooks.c (default_autovectorize_vector_sizes): Likewise.
86         * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
87         Likewise.
88         * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
89         * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
90         * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
91         * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
92         * omp-general.c (omp_max_vf): Likewise.
93         * omp-low.c (omp_clause_aligned_alignment): Likewise.
94         * optabs-query.c (can_vec_mask_load_store_p): Likewise.
95         * tree-vect-loop.c (vect_analyze_loop): Likewise.
96         * tree-vect-slp.c (vect_slp_bb): Likewise.
97         * doc/tm.texi: Regenerate.
98         * tree-vectorizer.h (current_vector_size): Change from an unsigned int
99         to a poly_uint64.
100         * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
101         the vector size as a poly_uint64 rather than an unsigned int.
102         (current_vector_size): Change from an unsigned int to a poly_uint64.
103         (get_vectype_for_scalar_type): Update accordingly.
104         * tree.h (build_truth_vector_type): Take the size and number of
105         units as a poly_uint64 rather than an unsigned int.
106         (build_vector_type): Add a temporary overload that takes
107         the number of units as a poly_uint64 rather than an unsigned int.
108         * tree.c (make_vector_type): Likewise.
109         (build_truth_vector_type): Take the number of units as a poly_uint64
110         rather than an unsigned int.
112 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
113             Alan Hayward  <alan.hayward@arm.com>
114             David Sherwood  <david.sherwood@arm.com>
116         * target.def (get_mask_mode): Take the number of units and length
117         as poly_uint64s rather than unsigned ints.
118         * targhooks.h (default_get_mask_mode): Update accordingly.
119         * targhooks.c (default_get_mask_mode): Likewise.
120         * config/i386/i386.c (ix86_get_mask_mode): Likewise.
121         * doc/tm.texi: Regenerate.
123 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
124             Alan Hayward  <alan.hayward@arm.com>
125             David Sherwood  <david.sherwood@arm.com>
127         * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
128         * omp-general.c (omp_max_vf): Likewise.
129         * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
130         (expand_omp_simd): Handle polynomial safelen.
131         * omp-low.c (omplow_simd_context): Add a default constructor.
132         (omplow_simd_context::max_vf): Change from int to poly_uint64.
133         (lower_rec_simd_input_clauses): Update accordingly.
134         (lower_rec_input_clauses): Likewise.
136 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
137             Alan Hayward  <alan.hayward@arm.com>
138             David Sherwood  <david.sherwood@arm.com>
140         * tree-vectorizer.h (vect_nunits_for_cost): New function.
141         * tree-vect-loop.c (vect_model_reduction_cost): Use it.
142         * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
143         (vect_analyze_slp_cost): Likewise.
144         * tree-vect-stmts.c (vect_model_store_cost): Likewise.
145         (vect_model_load_cost): Likewise.
147 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
148             Alan Hayward  <alan.hayward@arm.com>
149             David Sherwood  <david.sherwood@arm.com>
151         * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
152         (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
153         from an unsigned int * to a poly_uint64_pod *.
154         (calculate_unrolling_factor): New function.
155         (vect_analyze_slp_instance): Use it.  Track polynomial max_nunits.
157 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
158             Alan Hayward  <alan.hayward@arm.com>
159             David Sherwood  <david.sherwood@arm.com>
161         * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
162         from an unsigned int to a poly_uint64.
163         (_loop_vec_info::slp_unrolling_factor): Likewise.
164         (_loop_vec_info::vectorization_factor): Change from an int
165         to a poly_uint64.
166         (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
167         (vect_get_num_vectors): New function.
168         (vect_update_max_nunits, vect_vf_for_cost): Likewise.
169         (vect_get_num_copies): Use vect_get_num_vectors.
170         (vect_analyze_data_ref_dependences): Change max_vf from an int *
171         to an unsigned int *.
172         (vect_analyze_data_refs): Change min_vf from an int * to a
173         poly_uint64 *.
174         (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
175         than an unsigned HOST_WIDE_INT.
176         * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
177         (vect_analyze_data_ref_dependence): Change max_vf from an int *
178         to an unsigned int *.
179         (vect_analyze_data_ref_dependences): Likewise.
180         (vect_compute_data_ref_alignment): Handle polynomial vf.
181         (vect_enhance_data_refs_alignment): Likewise.
182         (vect_prune_runtime_alias_test_list): Likewise.
183         (vect_shift_permute_load_chain): Likewise.
184         (vect_supportable_dr_alignment): Likewise.
185         (dependence_distance_ge_vf): Take the vectorization factor as a
186         poly_uint64 rather than an unsigned HOST_WIDE_INT.
187         (vect_analyze_data_refs): Change min_vf from an int * to a
188         poly_uint64 *.
189         * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
190         vfm1 as a poly_uint64 rather than an int.  Make the same change
191         for the returned bound_scalar.
192         (vect_gen_vector_loop_niters): Handle polynomial vf.
193         (vect_do_peeling): Likewise.  Update call to
194         vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
195         (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
196         be constant.
197         * tree-vect-loop.c (vect_determine_vectorization_factor)
198         (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
199         (vect_get_known_peeling_cost): Likewise.
200         (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
201         (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
202         (vect_transform_loop): Likewise.  Use the lowest possible VF when
203         updating the upper bounds of the loop.
204         (vect_min_worthwhile_factor): Make static.  Return an unsigned int
205         rather than an int.
206         * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
207         polynomial unroll factors.
208         (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
209         (vect_make_slp_decision): Likewise.
210         (vect_supported_load_permutation_p): Likewise, and polynomial
211         vf too.
212         (vect_analyze_slp_cost): Handle polynomial vf.
213         (vect_slp_analyze_node_operations): Likewise.
214         (vect_slp_analyze_bb_1): Likewise.
215         (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
216         than an unsigned HOST_WIDE_INT.
217         * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
218         (vectorizable_load): Handle polynomial vf.
219         * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
220         a poly_uint64.
221         (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
223 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
224             Alan Hayward  <alan.hayward@arm.com>
225             David Sherwood  <david.sherwood@arm.com>
227         * match.pd: Handle bit operations involving three constants
228         and try to fold one pair.
230 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
232         * tree-vect-loop-manip.c: Include gimple-fold.h.
233         (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
234         niters_maybe_zero parameters.  Handle other cases besides a step of 1.
235         (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
236         Add a path that uses a step of VF instead of 1, but disable it
237         for now.
238         (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
239         and niters_no_overflow parameters.  Update calls to
240         slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
241         Create a new SSA name if the latter choses to use a ste other
242         than zero, and return it via niters_vector_mult_vf_var.
243         * tree-vect-loop.c (vect_transform_loop): Update calls to
244         vect_do_peeling, vect_gen_vector_loop_niters and
245         slpeel_make_loop_iterate_ntimes.
246         * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
247         (vect_gen_vector_loop_niters): Update declarations after above changes.
249 2018-01-02  Michael Meissner  <meissner@linux.vnet.ibm.com>
251         * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
252         128-bit round to integer instructions.
253         (ceil<mode>2): Likewise.
254         (btrunc<mode>2): Likewise.
255         (round<mode>2): Likewise.
257 2018-01-02  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
259         * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
260         unaligned VSX load/store on P8/P9.
261         (expand_block_clear): Allow the use of unaligned VSX
262         load/store on P8/P9.
264 2018-01-02  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
266         * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
267         New function.
268         (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
269         swap associated with both a load and a store.
271 2018-01-02  Andrew Waterman  <andrew@sifive.com>
273         * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
274         * config/riscv/riscv.md (clear_cache): Use it.
276 2018-01-02  Artyom Skrobov  <tyomitch@gmail.com>
278         * web.c: Remove out-of-date comment.
280 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
282         * expr.c (fixup_args_size_notes): Check that any existing
283         REG_ARGS_SIZE notes are correct, and don't try to re-add them.
284         (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
285         (emit_single_push_insn): ...here.
287 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
289         * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
290         (const_vector_encoded_nelts): New function.
291         (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
292         (const_vector_int_elt, const_vector_elt): Declare.
293         * emit-rtl.c (const_vector_int_elt_1): New function.
294         (const_vector_elt): Likewise.
295         * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
296         of CONST_VECTOR_ELT.
298 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
300         * expr.c: Include rtx-vector-builder.h.
301         (const_vector_mask_from_tree): Use rtx_vector_builder and operate
302         directly on the tree encoding.
303         (const_vector_from_tree): Likewise.
304         * optabs.c: Include rtx-vector-builder.h.
305         (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
306         sequence of "u" values.
307         * vec-perm-indices.c: Include rtx-vector-builder.h.
308         (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
309         directly on the vec_perm_indices encoding.
311 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
313         * doc/rtl.texi (const_vector): Describe new encoding scheme.
314         * Makefile.in (OBJS): Add rtx-vector-builder.o.
315         * rtx-vector-builder.h: New file.
316         * rtx-vector-builder.c: Likewise.
317         * rtl.h (rtx_def::u2): Add a const_vector field.
318         (CONST_VECTOR_NPATTERNS): New macro.
319         (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
320         (CONST_VECTOR_DUPLICATE_P): Likewise.
321         (CONST_VECTOR_STEPPED_P): Likewise.
322         (CONST_VECTOR_ENCODED_ELT): Likewise.
323         (const_vec_duplicate_p): Check for a duplicated vector encoding.
324         (unwrap_const_vec_duplicate): Likewise.
325         (const_vec_series_p): Check for a non-duplicated vector encoding.
326         Say that the function only returns true for integer vectors.
327         * emit-rtl.c: Include rtx-vector-builder.h.
328         (gen_const_vec_duplicate_1): Delete.
329         (gen_const_vector): Call gen_const_vec_duplicate instead of
330         gen_const_vec_duplicate_1.
331         (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
332         (gen_const_vec_duplicate): Use rtx_vector_builder.
333         (gen_const_vec_series): Likewise.
334         (gen_rtx_CONST_VECTOR): Likewise.
335         * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
336         (swap_const_vector_halves): Take an rtx pointer rather than rtx.
337         Build a new vector rather than modifying a CONST_VECTOR in-place.
338         (handle_special_swappables): Update call accordingly.
339         * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
340         (swap_const_vector_halves): Take an rtx pointer rather than rtx.
341         Build a new vector rather than modifying a CONST_VECTOR in-place.
342         (handle_special_swappables): Update call accordingly.
344 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
346         * simplify-rtx.c (simplify_const_binary_operation): Use
347         CONST_VECTOR_ELT instead of XVECEXP.
349 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
351         * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
352         the selector elements to be different from the data elements
353         if the selector is a VECTOR_CST.
354         * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
355         ssizetype for the selector.
357 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
359         * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
360         before testing each element individually.
361         * tree-vect-generic.c (lower_vec_perm): Likewise.
363 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
365         * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
366         * selftest-run-tests.c (selftest::run_tests): Call it.
367         * vector-builder.h (vector_builder::operator ==): New function.
368         (vector_builder::operator !=): Likewise.
369         * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
370         (vec_perm_indices::all_from_input_p): New function.
371         * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
372         (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
373         * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
374         instead of reading the VECTOR_CST directly.  Detect whether both
375         vector inputs are the same before constructing the vec_perm_indices,
376         and update the number of inputs argument accordingly.  Use the
377         utility functions added above.  Only construct sel2 if we need to.
379 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
381         * optabs.c (expand_vec_perm_var): Use an explicit encoding for
382         the broadcast of the low byte.
383         (expand_mult_highpart): Use an explicit encoding for the permutes.
384         * optabs-query.c (can_mult_highpart_p): Likewise.
385         * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
386         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
387         (vectorizable_bswap): Likewise.
388         * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
389         explicit encoding for the power-of-2 permutes.
390         (vect_permute_store_chain): Likewise.
391         (vect_grouped_load_supported): Likewise.
392         (vect_permute_load_chain): Likewise.
394 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
396         * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
397         * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
398         * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
399         * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
400         * tree-vect-stmts.c (vectorizable_bswap): Likewise.
401         (vect_gen_perm_mask_any): Likewise.
403 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
405         * int-vector-builder.h: New file.
406         * vec-perm-indices.h: Include int-vector-builder.h.
407         (vec_perm_indices): Redefine as an int_vector_builder.
408         (auto_vec_perm_indices): Delete.
409         (vec_perm_builder): Redefine as a stand-alone class.
410         (vec_perm_indices::vec_perm_indices): New function.
411         (vec_perm_indices::clamp): Likewise.
412         * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
413         (vec_perm_indices::new_vector): New function.
414         (vec_perm_indices::new_expanded_vector): Update for new
415         vec_perm_indices class.
416         (vec_perm_indices::rotate_inputs): New function.
417         (vec_perm_indices::all_in_range_p): Operate directly on the
418         encoded form, without computing elided elements.
419         (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
420         encoding.  Update for new vec_perm_indices class.
421         * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
422         the given vec_perm_builder.
423         (expand_vec_perm_var): Update vec_perm_builder constructor.
424         (expand_mult_highpart): Use vec_perm_builder instead of
425         auto_vec_perm_indices.
426         * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
427         vec_perm_indices instead of auto_vec_perm_indices.  Use a single
428         or double series encoding as appropriate.
429         * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
430         vec_perm_indices instead of auto_vec_perm_indices.
431         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
432         * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
433         (vect_permute_store_chain): Likewise.
434         (vect_grouped_load_supported): Likewise.
435         (vect_permute_load_chain): Likewise.
436         (vect_shift_permute_load_chain): Likewise.
437         * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
438         (vect_transform_slp_perm_load): Likewise.
439         (vect_schedule_slp_instance): Likewise.
440         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
441         (vectorizable_mask_load_store): Likewise.
442         (vectorizable_bswap): Likewise.
443         (vectorizable_store): Likewise.
444         (vectorizable_load): Likewise.
445         * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
446         vec_perm_indices instead of auto_vec_perm_indices.  Use
447         tree_to_vec_perm_builder to read the vector from a tree.
448         * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
449         vec_perm_builder instead of a vec_perm_indices.
450         (have_whole_vector_shift): Use vec_perm_builder and
451         vec_perm_indices instead of auto_vec_perm_indices.  Leave the
452         truncation to calc_vec_perm_mask_for_shift.
453         (vect_create_epilog_for_reduction): Likewise.
454         * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
455         from auto_vec_perm_indices to vec_perm_indices.
456         (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
457         instead of changing individual elements.
458         (aarch64_vectorize_vec_perm_const): Use new_vector to install
459         the vector in d.perm.
460         * config/arm/arm.c (expand_vec_perm_d::perm): Change
461         from auto_vec_perm_indices to vec_perm_indices.
462         (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
463         instead of changing individual elements.
464         (arm_vectorize_vec_perm_const): Use new_vector to install
465         the vector in d.perm.
466         * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
467         Update vec_perm_builder constructor.
468         (rs6000_expand_interleave): Likewise.
469         * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
470         (rs6000_expand_interleave): Likewise.
472 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
474         * optabs-query.c (can_vec_perm_var_p): Check whether lowering
475         to qimode could truncate the indices.
476         * optabs.c (expand_vec_perm_var): Likewise.
478 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
480         * Makefile.in (OBJS): Add vec-perm-indices.o.
481         * vec-perm-indices.h: New file.
482         * vec-perm-indices.c: Likewise.
483         * target.h (vec_perm_indices): Replace with a forward class
484         declaration.
485         (auto_vec_perm_indices): Move to vec-perm-indices.h.
486         * optabs.h: Include vec-perm-indices.h.
487         (expand_vec_perm): Delete.
488         (selector_fits_mode_p, expand_vec_perm_var): Declare.
489         (expand_vec_perm_const): Declare.
490         * target.def (vec_perm_const_ok): Replace with...
491         (vec_perm_const): ...this new hook.
492         * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
493         (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
494         * doc/tm.texi: Regenerate.
495         * optabs.def (vec_perm_const): Delete.
496         * doc/md.texi (vec_perm_const): Likewise.
497         (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
498         * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
499         expand_vec_perm for constant permutation vectors.  Assert that
500         the mode of variable permutation vectors is the integer equivalent
501         of the mode that is being permuted.
502         * optabs-query.h (selector_fits_mode_p): Declare.
503         * optabs-query.c: Include vec-perm-indices.h.
504         (selector_fits_mode_p): New function.
505         (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
506         is defined, instead of checking whether the vec_perm_const_optab
507         exists.  Use targetm.vectorize.vec_perm_const instead of
508         targetm.vectorize.vec_perm_const_ok.  Check whether the indices
509         fit in the vector mode before using a variable permute.
510         * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
511         vec_perm_indices instead of an rtx.
512         (expand_vec_perm): Replace with...
513         (expand_vec_perm_const): ...this new function.  Take the selector
514         as a vec_perm_indices rather than an rtx.  Also take the mode of
515         the selector.  Update call to shift_amt_for_vec_perm_mask.
516         Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
517         Use vec_perm_indices::new_expanded_vector to expand the original
518         selector into bytes.  Check whether the indices fit in the vector
519         mode before using a variable permute.
520         (expand_vec_perm_var): Make global.
521         (expand_mult_highpart): Use expand_vec_perm_const.
522         * fold-const.c: Includes vec-perm-indices.h.
523         * tree-ssa-forwprop.c: Likewise.
524         * tree-vect-data-refs.c: Likewise.
525         * tree-vect-generic.c: Likewise.
526         * tree-vect-loop.c: Likewise.
527         * tree-vect-slp.c: Likewise.
528         * tree-vect-stmts.c: Likewise.
529         * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
530         Delete.
531         * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
532         * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
533         (aarch64_vectorize_vec_perm_const_ok): Fuse into...
534         (aarch64_vectorize_vec_perm_const): ...this new function.
535         (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
536         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
537         * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
538         * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
539         * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
540         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
541         (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
542         into...
543         (arm_vectorize_vec_perm_const): ...this new function.  Explicitly
544         check for NEON modes.
545         * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
546         * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
547         * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
548         (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
549         into...
550         (ix86_vectorize_vec_perm_const): ...this new function.  Incorporate
551         the old VEC_PERM_CONST conditions.
552         * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
553         * config/ia64/vect.md (vec_perm_const<mode>): Delete.
554         * config/ia64/ia64.c (ia64_expand_vec_perm_const)
555         (ia64_vectorize_vec_perm_const_ok): Merge into...
556         (ia64_vectorize_vec_perm_const): ...this new function.
557         * config/mips/loongson.md (vec_perm_const<mode>): Delete.
558         * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
559         * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
560         * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
561         * config/mips/mips.c (mips_expand_vec_perm_const)
562         (mips_vectorize_vec_perm_const_ok): Merge into...
563         (mips_vectorize_vec_perm_const): ...this new function.
564         * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
565         * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
566         * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
567         * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
568         * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
569         (rs6000_expand_vec_perm_const): Delete.
570         * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
571         Delete.
572         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
573         (altivec_expand_vec_perm_const_le): Take each operand individually.
574         Operate on constant selectors rather than rtxes.
575         (altivec_expand_vec_perm_const): Likewise.  Update call to
576         altivec_expand_vec_perm_const_le.
577         (rs6000_expand_vec_perm_const): Delete.
578         (rs6000_vectorize_vec_perm_const_ok): Delete.
579         (rs6000_vectorize_vec_perm_const): New function.
580         (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
581         an element count and rtx array.
582         (rs6000_expand_extract_even): Update call accordingly.
583         (rs6000_expand_interleave): Likewise.
584         * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
585         * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
586         * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
587         * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
588         (rs6000_expand_vec_perm_const): Delete.
589         * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
590         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
591         (altivec_expand_vec_perm_const_le): Take each operand individually.
592         Operate on constant selectors rather than rtxes.
593         (altivec_expand_vec_perm_const): Likewise.  Update call to
594         altivec_expand_vec_perm_const_le.
595         (rs6000_expand_vec_perm_const): Delete.
596         (rs6000_vectorize_vec_perm_const_ok): Delete.
597         (rs6000_vectorize_vec_perm_const): New function.  Remove stray
598         reference to the SPE evmerge intructions.
599         (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
600         an element count and rtx array.
601         (rs6000_expand_extract_even): Update call accordingly.
602         (rs6000_expand_interleave): Likewise.
603         * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
604         * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
605         new function.
606         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
608 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
610         * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
611         vector mode and that that mode matches the mode of the data
612         being permuted.
613         (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
614         out into expand_vec_perm_var.  Do all CONST_VECTOR handling here,
615         directly using expand_vec_perm_1 when forcing selectors into
616         registers.
617         (expand_vec_perm_var): New function, split out from expand_vec_perm.
619 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
621         * optabs-query.h (can_vec_perm_p): Delete.
622         (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
623         * optabs-query.c (can_vec_perm_p): Split into...
624         (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
625         (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
626         particular selector is valid.
627         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
628         * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
629         (vect_grouped_load_supported): Likewise.
630         (vect_shift_permute_load_chain): Likewise.
631         * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
632         (vect_transform_slp_perm_load): Likewise.
633         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
634         (vectorizable_bswap): Likewise.
635         (vect_gen_perm_mask_checked): Likewise.
636         * fold-const.c (fold_ternary_loc): Likewise.  Don't take
637         implementations of variable permutation vectors into account
638         when deciding which selector to use.
639         * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
640         vec_perm_const_optab is supported; instead use can_vec_perm_const_p
641         with a false third argument.
642         * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
643         to test whether the constant selector is valid and can_vec_perm_var_p
644         to test whether a variable selector is valid.
646 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
648         * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
649         * optabs-query.c (can_vec_perm_p): Likewise.
650         * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
651         instead of vec_perm_indices.
652         * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
653         (vect_gen_perm_mask_checked): Likewise,
654         * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
655         (vect_gen_perm_mask_checked): Likewise,
657 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
659         * optabs-query.h (qimode_for_vec_perm): Declare.
660         * optabs-query.c (can_vec_perm_p): Split out qimode search to...
661         (qimode_for_vec_perm): ...this new function.
662         * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
664 2018-01-02  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
666         * rtlanal.c (canonicalize_condition): Return 0 if final rtx
667         does not have a conditional at the top.
669 2018-01-02  Richard Biener  <rguenther@suse.de>
671         * ipa-inline.c (big_speedup_p): Fix expression.
673 2018-01-02  Jan Hubicka  <hubicka@ucw.cz>
675         PR target/81616
676         * config/i386/x86-tune-costs.h: Increase cost of integer load costs
677         for generic 4->6.
679 2018-01-02  Jan Hubicka  <hubicka@ucw.cz>
681         PR target/81616
682         Generic tuning.
683         * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
684         cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
685         and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
686         cond_taken_branch_cost 3->4.
688 2018-01-01  Jakub Jelinek  <jakub@redhat.com>
690         PR tree-optimization/83581
691         * tree-loop-distribution.c (pass_loop_distribution::execute): Return
692         TODO_cleanup_cfg if any changes have been made.
694         PR middle-end/83608
695         * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
696         convert_modes if target mode has the right side, but different mode
697         class.
699         PR middle-end/83609
700         * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
701         last argument when extracting from CONCAT.  If either from_real or
702         from_imag is NULL, use expansion through memory.  If result is not
703         a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
704         the parts directly to inner mode, if even that fails, use expansion
705         through memory.
707         PR middle-end/83623
708         * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
709         check for bswap in mode rather than HImode and use that in expand_unop
710         too.
712 2017-12-31  Jakub Jelinek  <jakub@redhat.com>
714         PR target/83536
715         * config/i386/i386.c (ix86_attribute_table): Remove excess
716         initializer for "shared" attribute.
718 2017-12-30  Tom de Vries  <tom@codesourcery.com>
720         PR libgomp/83046
721         * omp-expand.c (expand_omp_target): If in_lto_p, mark offload_funcs with
722         DECL_PRESERVE_P.
723         * lto-streamer-out.c (prune_offload_funcs): New function.  Remove
724         offload_funcs entries that no longer have a corresponding cgraph_node.
725         Mark the remaining ones as DECL_PRESERVE_P.
726         (output_lto): Call prune_offload_funcs.
728 2017-12-30  Jakub Jelinek  <jakub@redhat.com>
730         * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
731         vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>,
732         vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
733         vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
734         vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
735         vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1, vpdpbusd_<mode>,
736         vpdpbusd_<mode>_mask, vpdpbusd_<mode>_maskz, vpdpbusd_<mode>_maskz_1,
737         vpdpbusds_<mode>, vpdpbusds_<mode>_mask, vpdpbusds_<mode>_maskz,
738         vpdpbusds_<mode>_maskz_1, vpdpwssd_<mode>, vpdpwssd_<mode>_mask,
739         vpdpwssd_<mode>_maskz, vpdpwssd_<mode>_maskz_1, vpdpwssds_<mode>,
740         vpdpwssds_<mode>_mask, vpdpwssds_<mode>_maskz,
741         vpdpwssds_<mode>_maskz_1, vaesdec_<mode>, vaesdeclast_<mode>,
742         vaesenc_<mode>, vpclmulqdq_<mode>,
743         avx512vl_vpshufbitqmb<mode><mask_scalar_merge_name>): Formatting fixes.
745 2017-12-28  Michael Meissner  <meissner@linux.vnet.ibm.com>
747         * builtins.def: (_Float<N> and _Float<N>X BUILT_IN_CEIL): Add
748         _Float<N> and _Float<N>X variants for rounding built-in
749         functions.
750         (_Float<N> and _Float<N>X BUILT_IN_FLOOR): Likewise.
751         (_Float<N> and _Float<N>X BUILT_IN_NEARBYINT): Likewise.
752         (_Float<N> and _Float<N>X BUILT_IN_RINT): Likewise.
753         (_Float<N> and _Float<N>X BUILT_IN_ROUND): Likewise.
754         (_Float<N> and _Float<N>X BUILT_IN_TRUNC): Likewise.
755         * builtins.c (mathfn_built_in_2): Likewise.
756         * internal-fn.def (CEIL): Likewise.
757         (FLOOR): Likewise.
758         (NEARBYINT): Likewise.
759         (RINT): Likewise.
760         (ROUND): Likewise.
761         (TRUNC): Likewise.
762         * convert.c (convert_to_integer_1): Likewise.
763         * fold-const.c (tree_call_nonnegative_warnv_p): Likewise.
764         (integer_valued_real_call_p): Likewise.
765         * fold-const-call.c (fold_const_call_ss): Likewise.
766         * gencfn-macros.c (print_case_cfn): Change CFN and operator
767         printers to take a const char * suffix instead of a bool.
768         (print_define_operator_list): Likewise.
769         (fltall_suffixes): New list of suffixes, that include the
770         traditional suffixes as well as all of the _Float<N> and
771         _Float<N>X suffixes.
772         (main): For _Float<N> and _Float<N>X functions, emit both
773         <name>_FN and <name>_ALL variants.  The <macro>_FN variant only
774         has the _Float<N> and _Float<N>X case names or operators.  The
775         <name>_ALL variant has both the traditional and the
776         _Float<N>/_Float<N>X case names or operators.
777         * match.pd (COPYSIGN optimizations): Provide optimizations for
778         _Float<N> and _Float<N>X types where possible.
779         (MIN/MAX optimizations): Likewise.
780         (sqrt optimizations): Likewise.
781         (rounding optimizations): Likewise.
783 2017-12-28  Richard Sandiford  <richard.sandiford@linaro.org>
785         * config/powerpcspe/powerpcspe.md (ieee_128bit_negative_zero): Use
786         gen_int_mode rather than GEN_INT.
787         * config/rs6000/rs6000.md (ieee_128bit_negative_zero): Likewise.
789 2017-12-28  Richard Sandiford  <richard.sandiford@linaro.org>
791         * emit-rtl.c (gen_const_vec_series): Use valid_for_const_vector_p
792         instead of CONSTANT_P.
793         (gen_vec_series): Likewise.
794         * simplify-rtx.c (simplify_binary_operation_1): Likewise.
796 2017-12-28  Andreas Schwab  <schwab@linux-m68k.org>
798         * config/m68k/m68k.md (ashrdi3_const1, lshrdi3_const1): Add
799         CC_STATUS_INIT.
801 2017-12-27  Kugan Vivekanandarajah  <kuganv@linaro.org>
803         * config/aarch64/aarch64-simd.md (aarch64_ld1x2<VQ:mode>): New.
804         (aarch64_ld1x2<VDC:mode>): Likewise.
805         (aarch64_simd_ld1<mode>_x2): Likewise.
806         (aarch64_simd_ld1<mode>_x2): Likewise.
807         * config/aarch64/arm_neon.h (vld1_u8_x2): New.
808         (vld1_s8_x2): Likewise.
809         (vld1_u16_x2): Likewise.
810         (vld1_s16_x2): Likewise.
811         (vld1_u32_x2): Likewise.
812         (vld1_s32_x2): Likewise.
813         (vld1_u64_x2): Likewise.
814         (vld1_s64_x2): Likewise.
815         (vld1_f16_x2): Likewise.
816         (vld1_f32_x2): Likewise.
817         (vld1_f64_x2): Likewise.
818         (vld1_p8_x2): Likewise.
819         (vld1_p16_x2): Likewise.
820         (vld1_p64_x2): Likewise.
821         (vld1q_u8_x2): Likewise.
822         (vld1q_s8_x2): Likewise.
823         (vld1q_u16_x2): Likewise.
824         (vld1q_s16_x2): Likewise.
825         (vld1q_u32_x2): Likewise.
826         (vld1q_s32_x2): Likewise.
827         (vld1q_u64_x2): Likewise.
828         (vld1q_s64_x2): Likewise.
829         (vld1q_f16_x2): Likewise.
830         (vld1q_f32_x2): Likewise.
831         (vld1q_f64_x2): Likewise.
832         (vld1q_p8_x2): Likewise.
833         (vld1q_p16_x2): Likewise.
834         (vld1q_p64_x2): Likewise.
836 2017-12-27  Martin Liska  <mliska@suse.cz>
838         PR tree-optimization/83552
839         * tree-ssa-strlen.c (fold_strstr_to_strncmp): Assign result
840         of get_string_lenth to a SSA_NAME if not a GIMPLE value.
842 2017-12-27  Tom de Vries  <tom@codesourcery.com>
844         * config/nvptx/nvptx.c (nvptx_option_override): Disable
845         -gstatement-frontiers.
847 2017-12-26  Alexander Monakov  <amonakov@ispras.ru>
849         PR rtl-optimization/83513
850         * sel-sched.c (sel_rank_for_schedule): Order by non-zero usefulness
851         before priority comparison.
853 2017-12-25  Jakub Jelinek  <jakub@redhat.com>
855         PR target/83488
856         * config/i386/i386.opt (-mavx512vpopcntdq, -mmavx512bitalg): Move from
857         ix86_isa_flags2 to ix86_isa_flags.
858         * config/i386/i386-c.c (ix86_target_macros_internal): Test
859         OPTION_MASK_ISA_AVX512BITALG and OPTION_MASK_ISA_AVX512VPOPCNTDQ in
860         isa_flags rather than isa_flags2.
861         * config/i386/i386.c (ix86_target_string): Move -mavx512vpopcntdq
862         and -mavx512bitalg from isa2_opts to isa_opts.
863         (ix86_option_override_internal): Test OPTION_MASK_ISA_AVX512VPOPCNTDQ
864         in x_ix86_isa_flags_explicit rather than x_ix86_isa_flags2_explicit
865         and set it in x_ix86_isa_flags rather than x_ix86_isa_flags2.
866         Formatting fixes.
867         (def_builtin): Treat OPTION_MASK_ISA_AVX512BW or
868         OPTION_MASK_ISA_AVX512F ored with another option similarly to
869         OPTION_MASK_ISA_AVX512VL.  Even for OPTION_MASK_ISA_AVX512VL don't
870         clear it if mask is just OPTION_MASK_ISA_AVX512VL itself.
871         (ix86_expand_builtin): Don't handle OPTION_MASK_ISA_GFNI and
872         OPTION_MASK_ISA_VPCLMULQDQ specially, instead handle
873         OPTION_MASK_ISA_AVX512BW and OPTION_MASK_ISA_AVX512F that way.
874         * config/i386/i386-builtin.def: Move AVX512VPOPCNTDQ and AVX512BITALG
875         builtins from bdesc_args2 to bdesc_args section.
876         (__builtin_ia32_compressstoreuqi512_mask,
877         __builtin_ia32_compressstoreuhi512_mask,
878         __builtin_ia32_compressstoreuqi256_mask,
879         __builtin_ia32_expandloadqi512_mask,
880         __builtin_ia32_expandloadqi512_maskz,
881         __builtin_ia32_expandloadhi512_mask,
882         __builtin_ia32_expandloadhi512_maskz,
883         __builtin_ia32_compressqi512_mask, __builtin_ia32_compresshi512_mask,
884         __builtin_ia32_compressqi256_mask, __builtin_ia32_expandqi512_mask,
885         __builtin_ia32_expandqi512_maskz, __builtin_ia32_expandhi512_mask,
886         __builtin_ia32_expandhi512_maskz, __builtin_ia32_expandqi256_mask,
887         __builtin_ia32_expandqi256_maskz, __builtin_ia32_vpshrd_v32hi_mask,
888         __builtin_ia32_vpshld_v32hi_mask, __builtin_ia32_vpshrdv_v32hi_mask,
889         __builtin_ia32_vpshrdv_v32hi_maskz, __builtin_ia32_vpshldv_v32hi_mask,
890         __builtin_ia32_vpshldv_v32hi_maskz,
891         __builtin_ia32_vpopcountb_v64qi_mask,
892         __builtin_ia32_vpopcountw_v32hi_mask,
893         __builtin_ia32_vpshufbitqmb512_mask,
894         __builtin_ia32_vpshufbitqmb256_mask): Add
895         " | OPTION_MASK_ISA_AVX512BW".
896         (__builtin_ia32_expandloadqi256_mask,
897         __builtin_ia32_expandloadqi256_maskz,
898         __builtin_ia32_vpopcountb_v32qi_mask): Add
899         " | OPTION_MASK_ISA_AVX512VL | OPTION_MASK_ISA_AVX512BW".
900         (__builtin_ia32_expandloadhi256_mask,
901         __builtin_ia32_expandloadhi256_maskz,
902         __builtin_ia32_expandloadqi128_mask,
903         __builtin_ia32_expandloadqi128_maskz,
904         __builtin_ia32_expandloadhi128_mask,
905         __builtin_ia32_expandloadhi128_maskz,
906         __builtin_ia32_vpshrd_v16hi, __builtin_ia32_vpshrd_v16hi_mask,
907         __builtin_ia32_vpshrd_v8hi, __builtin_ia32_vpshrd_v8hi_mask,
908         __builtin_ia32_vpshrd_v8si, __builtin_ia32_vpshrd_v8si_mask,
909         __builtin_ia32_vpshrd_v4si, __builtin_ia32_vpshrd_v4si_mask,
910         __builtin_ia32_vpshrd_v4di, __builtin_ia32_vpshrd_v4di_mask,
911         __builtin_ia32_vpshrd_v2di, __builtin_ia32_vpshrd_v2di_mask,
912         __builtin_ia32_vpshld_v16hi, __builtin_ia32_vpshld_v16hi_mask,
913         __builtin_ia32_vpshld_v8hi, __builtin_ia32_vpshld_v8hi_mask,
914         __builtin_ia32_vpshld_v8si, __builtin_ia32_vpshld_v8si_mask,
915         __builtin_ia32_vpshld_v4si, __builtin_ia32_vpshld_v4si_mask,
916         __builtin_ia32_vpshld_v4di, __builtin_ia32_vpshld_v4di_mask,
917         __builtin_ia32_vpshld_v2di, __builtin_ia32_vpshld_v2di_mask,
918         __builtin_ia32_vpshrdv_v16hi, __builtin_ia32_vpshrdv_v16hi_mask,
919         __builtin_ia32_vpshrdv_v16hi_maskz, __builtin_ia32_vpshrdv_v8hi,
920         __builtin_ia32_vpshrdv_v8hi_mask, __builtin_ia32_vpshrdv_v8hi_maskz,
921         __builtin_ia32_vpshrdv_v8si, __builtin_ia32_vpshrdv_v8si_mask,
922         __builtin_ia32_vpshrdv_v8si_maskz, __builtin_ia32_vpshrdv_v4si,
923         __builtin_ia32_vpshrdv_v4si_mask, __builtin_ia32_vpshrdv_v4si_maskz,
924         __builtin_ia32_vpshrdv_v4di, __builtin_ia32_vpshrdv_v4di_mask,
925         __builtin_ia32_vpshrdv_v4di_maskz, __builtin_ia32_vpshrdv_v2di,
926         __builtin_ia32_vpshrdv_v2di_mask, __builtin_ia32_vpshrdv_v2di_maskz,
927         __builtin_ia32_vpshldv_v16hi, __builtin_ia32_vpshldv_v16hi_mask,
928         __builtin_ia32_vpshldv_v16hi_maskz, __builtin_ia32_vpshldv_v8hi,
929         __builtin_ia32_vpshldv_v8hi_mask, __builtin_ia32_vpshldv_v8hi_maskz,
930         __builtin_ia32_vpshldv_v8si, __builtin_ia32_vpshldv_v8si_mask,
931         __builtin_ia32_vpshldv_v8si_maskz, __builtin_ia32_vpshldv_v4si,
932         __builtin_ia32_vpshldv_v4si_mask, __builtin_ia32_vpshldv_v4si_maskz,
933         __builtin_ia32_vpshldv_v4di, __builtin_ia32_vpshldv_v4di_mask,
934         __builtin_ia32_vpshldv_v4di_maskz, __builtin_ia32_vpshldv_v2di,
935         __builtin_ia32_vpshldv_v2di_mask, __builtin_ia32_vpshldv_v2di_maskz,
936         __builtin_ia32_vpopcountb_v32qi, __builtin_ia32_vpopcountb_v16qi,
937         __builtin_ia32_vpopcountb_v16qi_mask, __builtin_ia32_vpopcountw_v16hi,
938         __builtin_ia32_vpopcountw_v16hi_mask, __builtin_ia32_vpopcountw_v8hi,
939         __builtin_ia32_vpopcountw_v8hi_mask): Add
940         " | OPTION_MASK_ISA_AVX512VL".
941         * config/i386/avx512vbmi2intrin.h (_mm512_shrdi_epi16,
942         _mm512_shrdi_epi32, _mm512_mask_shrdi_epi32, _mm512_maskz_shrdi_epi32,
943         _mm512_shrdi_epi64, _mm512_mask_shrdi_epi64, _mm512_maskz_shrdi_epi64,
944         _mm512_shldi_epi16, _mm512_shldi_epi32, _mm512_mask_shldi_epi32,
945         _mm512_maskz_shldi_epi32, _mm512_shldi_epi64, _mm512_mask_shldi_epi64,
946         _mm512_maskz_shldi_epi64, _mm512_shrdv_epi16, _mm512_shrdv_epi32,
947         _mm512_mask_shrdv_epi32, _mm512_maskz_shrdv_epi32, _mm512_shrdv_epi64,
948         _mm512_mask_shrdv_epi64, _mm512_maskz_shrdv_epi64, _mm512_shldv_epi16,
949         _mm512_shldv_epi32, _mm512_mask_shldv_epi32, _mm512_maskz_shldv_epi32,
950         _mm512_shldv_epi64, _mm512_mask_shldv_epi64,
951         _mm512_maskz_shldv_epi64): Don't require avx512bw for these intrinsics.
952         * config/i386/avx512bitalgintrin.h (_mm_bitshuffle_epi64_mask,
953         _mm_mask_bitshuffle_epi64_mask): Likewise.
954         * common/config/i386/i386-common.c
955         (OPTION_MASK_ISA_AVX512VPOPCNTDQ_SET,
956         OPTION_MASK_ISA_AVX512BITALG_SET): Or in OPTION_MASK_ISA_AVX512F_SET.
957         (OPTION_MASK_ISA_AVX512F_UNSET): Or in
958         OPTION_MASK_ISA_AVX512VPOPCNTDQ_UNSET and
959         OPTION_MASK_ISA_AVX512BITALG_UNSET.
960         (OPTION_MASK_ISA2_AVX512F_UNSET,
961         OPTION_MASK_ISA2_GENERAL_REGS_ONLY_UNSET): Define.
962         (ix86_handle_option): For -mno-general-regs-only, clear from
963         ix86_isa_flags2 OPTION_MASK_ISA2_GENERAL_REGS_ONLY_UNSET rather than
964         just OPTION_MASK_ISA_MPX.  For -mno-sse{,2,3,4,4.1,4.2,avx,avx2} and
965         -mno-ssse3 clear OPTION_MASK_ISA2_AVX512F_UNSET bits from
966         ix86_isa_flags2.  For -mno-avx512f likewise, instead of masking
967         individually listed ISAs.  For -m{,no-}avx512{vpopcntdq,bitalg} adjust
968         for moving from ix86_isa_flags2 to ix86_isa_flags.
970 2017-12-23  Jakub Jelinek  <jakub@redhat.com>
972         PR c++/83553
973         * fold-const.c (struct contains_label_data): New type.
974         (contains_label_1): Return non-NULL even for CASE_LABEL_EXPR, unless
975         inside of a SWITCH_BODY seen during the walk.
976         (contains_label_p): Use walk_tree instead of
977         walk_tree_without_duplicates, prepare data for contains_label_1 and
978         provide own pset.
980 2017-12-22  Jakub Jelinek  <jakub@redhat.com>
982         PR debug/83547
983         * tree-iterator.c (alloc_stmt_list): Start with cleared
984         TREE_SIDE_EFFECTS regardless whether a new STATEMENT_LIST is allocated
985         or old one reused.
987         PR target/83488
988         * config/i386/avx512vnniintrin.h: Don't check for __AVX512F__ nor
989         enable avx512f explicitly in #pragma GCC target.
990         * config/i386/i386-builtin.def (__builtin_ia32_vpdpbusd_v8si,
991         __builtin_ia32_vpdpbusd_v8si_mask, __builtin_ia32_vpdpbusd_v8si_maskz,
992         __builtin_ia32_vpdpbusd_v4si, __builtin_ia32_vpdpbusd_v4si_mask,
993         __builtin_ia32_vpdpbusd_v4si_maskz, __builtin_ia32_vpdpbusds_v8si,
994         __builtin_ia32_vpdpbusds_v8si_mask,
995         __builtin_ia32_vpdpbusds_v8si_maskz, __builtin_ia32_vpdpbusds_v4si,
996         __builtin_ia32_vpdpbusds_v4si_mask,
997         __builtin_ia32_vpdpbusds_v4si_maskz, __builtin_ia32_vpdpwssd_v8si,
998         __builtin_ia32_vpdpwssd_v8si_mask, __builtin_ia32_vpdpwssd_v8si_maskz,
999         __builtin_ia32_vpdpwssd_v4si, __builtin_ia32_vpdpwssd_v4si_mask,
1000         __builtin_ia32_vpdpwssd_v4si_maskz, __builtin_ia32_vpdpwssds_v8si,
1001         __builtin_ia32_vpdpwssds_v8si_mask,
1002         __builtin_ia32_vpdpwssds_v8si_maskz, __builtin_ia32_vpdpwssds_v4si,
1003         __builtin_ia32_vpdpwssds_v4si_mask,
1004         __builtin_ia32_vpdpwssds_v4si_maskz): Use
1005         OPTION_MASK_ISA_AVX512VNNI | OPTION_MASK_ISA_AVX512VL instead of
1006         just OPTION_MASK_ISA_AVX512VNNI.
1008 2017-12-22  Martin Jambor  <mjambor@suse.cz>
1010         PR lto/82027
1011         * lto-cgraph.c (output_cgraph_opt_summary_p): Also check former
1012         clones.
1014 2017-12-22  Julia Koval  <julia.koval@intel.com>
1015             Sebastian Peryt  <sebastian.peryt@intel.com>
1017         * common/config/i386/i386-common.c (OPTION_MASK_ISA_AVX512BITALG_SET,
1018         OPTION_MASK_ISA_AVX512BITALG_UNSET): New.
1019         (ix86_handle_option): Handle -mavx512bitalg, fix 4VNNIW formatting.
1020         * config.gcc: Add avx512vpopcntdqvlintrin.h and avx512bitalgintrin.h.
1021         * config/i386/avx512bitalgintrin.h (_mm512_popcnt_epi8, _mm512_popcnt_epi16,
1022         _mm512_mask_popcnt_epi8, _mm512_maskz_popcnt_epi8, _mm512_mask_popcnt_epi16,
1023         _mm512_maskz_popcnt_epi16, _mm512_bitshuffle_epi64_mask, _mm256_popcnt_epi8,
1024         _mm512_mask_bitshuffle_epi64_mask, _mm256_mask_popcnt_epi8, _mm_popcnt_epi8,
1025         _mm256_maskz_popcnt_epi8, _mm_bitshuffle_epi64_mask, _mm256_popcnt_epi16,
1026         _mm_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
1027         _mm256_mask_bitshuffle_epi64_mask, _mm_popcnt_epi16, _mm_maskz_popcnt_epi8,
1028         _mm256_mask_popcnt_epi16, _mm256_maskz_popcnt_epi16, _mm_mask_popcnt_epi8,
1029         _mm_mask_popcnt_epi16, _mm_maskz_popcnt_epi16): New intrinsics.
1030         * config/i386/avx512vpopcntdqvlintrin.h (_mm_popcnt_epi32, _mm_popcnt_epi64,
1031         _mm_mask_popcnt_epi32, _mm_maskz_popcnt_epi32, _mm256_popcnt_epi32,
1032         _mm256_mask_popcnt_epi32, _mm256_maskz_popcnt_epi32, _mm_mask_popcnt_epi64,
1033         _mm_maskz_popcnt_epi64, _mm256_popcnt_epi64, _mm256_mask_popcnt_epi64,
1034         _mm256_maskz_popcnt_epi64): New intrinsics.
1035         * config/i386/cpuid.h (bit_AVX512BITALG): New bit.
1036         * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mavx512bitalg.
1037         * config/i386/i386-builtin-types.def (V64QI_FTYPE_V64QI, V64QI_FTYPE_V64QI,
1038         V4DI_FTYPE_V4DI, UHI_FTYPE_V2DI_V2DI_UHI, USI_FTYPE_V4DI_V4DI_USI,
1039         V4SI_FTYPE_V4SI_V4SI_UHI, V8SI_FTYPE_V8SI_V8SI_UHI): New types.
1040         * config/i386/i386-builtin.def (__builtin_ia32_vpopcountq_v4di,
1041         __builtin_ia32_vpopcountq_v4di_mask, __builtin_ia32_vpopcountq_v2di,
1042         __builtin_ia32_vpopcountq_v2di_mask, __builtin_ia32_vpopcountd_v4si,
1043         __builtin_ia32_vpopcountd_v4si_mask, __builtin_ia32_vpopcountd_v8si,
1044         __builtin_ia32_vpopcountd_v8si_mask, __builtin_ia32_vpopcountb_v64qi,
1045         __builtin_ia32_vpopcountb_v64qi_mask, __builtin_ia32_vpopcountb_v32qi,
1046         __builtin_ia32_vpopcountb_v32qi_mask, __builtin_ia32_vpopcountb_v16qi,
1047         __builtin_ia32_vpopcountb_v16qi_mask, __builtin_ia32_vpopcountw_v32hi,
1048         __builtin_ia32_vpopcountw_v32hi_mask, __builtin_ia32_vpopcountw_v16hi,
1049         __builtin_ia32_vpopcountw_v16hi_mask, __builtin_ia32_vpopcountw_v8hi,
1050         __builtin_ia32_vpopcountw_v8hi_mask, __builtin_ia32_vpshufbitqmb128_mask,
1051         __builtin_ia32_vpshufbitqmb256_mask,
1052         __builtin_ia32_vpshufbitqmb512_mask): New builtins.
1053         * config/i386/i386-c.c (__AVX512BITALG__): New.
1054         * config/i386/i386.c (isa2_opts): Add -mavx512bitalg.
1055         (ix86_valid_target_attribute_inner_p): Ditto.
1056         (ix86_expand_args_builtin): Handle new types.
1057         * config/i386/i386.h (TARGET_AVX512BITALG, TARGET_AVX512BITALG_P): New.
1058         * config/i386/i386.opt: Add -mavx512bitalg.
1059         * config/i386/immintrin.h: Add avx512vpopcntdqvlintrin.h and
1060         avx512bitalgintrin.h.
1061         * config/i386/sse.md (VI48_AVX512VLBW): New iterator.
1062         (vpopcount<mode><mask_name>): Add more types.
1063         (avx512vl_vpshufbitqmb<mode><mask_scalar_merge_name>): New.
1064         * doc/invoke.texi: Add -mavx512bitalg and -mavx512vpopcntdq.
1066 2017-12-22  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
1068         * common/config/i386/i386-common.c (OPTION_MASK_ISA_AVX512VNNI_SET):
1069         Or in OPTION_MASK_ISA_AVX512F_SET.
1070         (OPTION_MASK_ISA_AVX512F_UNSET): Or in
1071         OPTION_MASK_ISA_AVX512VNNI_UNSET.
1072         (ix86_handle_option): Adjust for
1073         OPTION_MASK_ISA_AVX512VNNI_*SET being in ix86_isa_flags.
1074         * config/i386/i386-builtin.def: Move VNNI builtins from ARGS2
1075         section to ARGS.
1076         * config/i386/i386-c.c: Check for OPTION_MASK_ISA_AVX512VNNI in
1077         isa_flag instead of isa_flag2.
1078         * config/i386/i386.c (ix86_target_string): Move -mavx512vnni from
1079         isa_opts2 to isa_opts.
1080         * config/i386/i386.opt (mavx512vnni): Move from ix86_isa_flags2
1081         to ix86_isa_flags.
1083 2017-12-22  Mike Stump  <mikestump@comcast.net>
1084             Eric Botcazou  <ebotcazou@adacore.com>
1086         * doc/extend.texi (Loop-Specific Pragmas): Document pragma GCC unroll.
1088 2017-12-21  Jakub Jelinek  <jakub@redhat.com>
1090         PR middle-end/83487
1091         * config/i386/i386.c (ix86_function_arg_boundary): Return
1092         PARM_BOUNDARY for TYPE_EMPTY_P types.
1094         PR c/83448
1095         * gimple-ssa-sprintf.c (maybe_warn): Don't call set_caret_index
1096         if navail is >= dir.len.
1098 2017-12-21  Steve Ellcey  <sellcey@cavium.com>
1100         * config/aarch64/t-aarch64-linux (MULTILIB_OSDIRNAMES): Fix
1101         triplet for ilp32.
1103 2017-12-21  Jakub Jelinek  <jakub@redhat.com>
1105         PR rtl-optimization/80747
1106         PR rtl-optimization/83512
1107         * cfgrtl.c (force_nonfallthru_and_redirect): When splitting
1108         succ edge from ENTRY, copy partition from e->dest to the newly
1109         created bb.
1110         * bb-reorder.c (reorder_basic_blocks_simple): If last_tail is
1111         ENTRY, use BB_PARTITION of its successor block as current_partition.
1112         Don't copy partition when splitting succ edge from ENTRY.
1114         PR tree-optimization/83523
1115         * tree-ssa-math-opts.c (is_widening_mult_p): Return false if
1116         for INTEGER_TYPE TYPE_OVERFLOW_TRAPS.
1117         (convert_mult_to_fma): Likewise.
1119         PR tree-optimization/83521
1120         * tree-ssa-phiopt.c (factor_out_conditional_conversion): Use
1121         gimple_build_assign without code on result of
1122         fold_build1 (VIEW_CONVERT_EXPR, ...), as it might not create
1123         a VIEW_CONVERT_EXPR.
1125 2017-12-21  Andrew Pinski  <apinski@cavium.com>
1126             Steve Ellcey  <sellcey@cavium.com>
1128         * config/aarch64/t-aarch64-linux (MULTILIB_OSDIRNAMES): Handle
1129         multi-arch for ilp32.
1131 2017-12-21  Uros Bizjak  <ubizjak@gmail.com>
1133         PR target/83467
1134         * config/i386/i386.md (*ashl<mode>3_mask): Add operand
1135         constraints to operand 2.
1136         (*ashl<mode>3_mask_1): Ditto.
1137         (*<shift_insn><mode>3_mask): Ditto.
1138         (*<shift_insn><mode>3_mask_1): Ditto.
1139         (*<rotate_insn><mode>3_mask): Ditto.
1140         (*<rotate_insn><mode>3_mask_1): Ditto.
1142 2017-12-21  Alexandre Oliva  <aoliva@redhat.com>
1144         * reorg.c (make_return_insns): Reemit each insn with its own location.
1146 2017-12-21  James Greenhalgh  <james.greenhalgh@arm.com>
1148         * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
1149         generation for cases where splatting a value is not useful.
1150         * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
1151         across a vec_duplicate and a paradoxical subreg forming a vector
1152         mode to a vec_concat.
1154 2017-12-21  James Greenhalgh  <james.greenhalgh@arm.com>
1156         * combine.c (simplify_set): Do not transform subregs to zero_extends
1157         if the destination is not a scalar int mode.
1159 2017-12-21  Eric Botcazou  <ebotcazou@adacore.com>
1161         PR c++/82872
1162         * convert.c (convert_to_integer_1) <POINTER_TYPE>: Do not return the
1163         shared zero if the input has overflowed.
1165 2017-12-21  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
1167         * config/arm/driver-arm.c (arm_cpu_table): Specify dotprod
1168         support for Cortex-A55 and Cortex-A75.
1170 2017-12-21  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
1172         * common/config/arm/arm-common.c (compare_opt_names): Add function
1173         comment.  Use strcmp instead of manual loop.
1175 2017-12-21  Martin Liska  <mliska@suse.cz>
1177         PR gcov-profile/83509
1178         * gcov-dump.c (dump_gcov_file): Do not read info about
1179         support_unexecuted_blocks for gcda files.
1181 2017-12-21  Jakub Jelinek  <jakub@redhat.com>
1183         PR rtl-optimization/82973
1184         * emit-rtl.h (valid_for_const_vec_duplicate_p): Rename to ...
1185         (valid_for_const_vector_p): ... this.
1186         * emit-rtl.c (valid_for_const_vec_duplicate_p): Rename to ...
1187         (valid_for_const_vector_p): ... this.  Adjust function comment.
1188         (gen_vec_duplicate): Adjust caller.
1189         * optabs.c (expand_vector_broadcast): Likewise.
1190         * simplify-rtx.c (simplify_const_unary_operation): Don't optimize into
1191         CONST_VECTOR if some element isn't simplified valid_for_const_vector_p
1192         constant.
1193         (simplify_const_binary_operation): Likewise.  Use CONST_FIXED_P macro
1194         instead of GET_CODE == CONST_FIXED.
1195         (simplify_subreg): Use CONST_FIXED_P macro instead of
1196         GET_CODE == CONST_FIXED.
1198         PR target/83488
1199         * config/i386/i386.c (ix86_target_string): Move -mavx512vbmi2 and
1200         -mshstk entries from isa_opts2 to isa_opts and -mhle, -mmovbe,
1201         -mclzero and -mmwaitx entries from isa_opts to isa_opts2.
1202         (ix86_option_override_internal): Adjust for
1203         OPTION_MASK_ISA_{HLE,MOVBE,CLZERO,MWAITX} moving to ix86_isa_flags2
1204         and OPTION_MASK_ISA_SHSTK moving to ix86_isa_flags.
1205         (BDESC_VERIFYS): Remove SPECIAL_ARGS2 related checks.
1206         (ix86_init_mmx_sse_builtins): Remove bdesc_special_args2 handling.
1207         Use def_builtin2 instead of def_builtin for OPTION_MASK_ISA_MWAITX
1208         and OPTION_MASK_ISA_CLZERO builtins.  Use def_builtin instead of
1209         def_builtin2 for CET builtins.
1210         (ix86_expand_builtin): Remove bdesc_special_args2 handling.  Fix
1211         up formatting in IX86_BUILTIN_RDPID code.
1212         * config/i386/i386-builtin.def: Move VBMI2 builtins from SPECIAL_ARGS2
1213         section to SPECIAL_ARGS and from ARGS2 section to ARGS.
1214         * config/i386/i386.opt (mavx512vbmi2, mshstk): Move from
1215         ix86_isa_flags2 to ix86_isa_flags.
1216         (mhle, mmovbe, mclzero, mmwaitx): Move from ix86_isa_flags to
1217         ix86_isa_flags2.
1218         * config/i386/i386-c.c (ix86_target_macros_internal): Check for
1219         OPTION_MASK_ISA_{CLZERO,MWAITX} in isa_flag2 instead of isa_flag.
1220         Check for OPTION_MASK_ISA_{SHSTK,AVX512VBMI2} in isa_flag instead
1221         of isa_flag2.
1222         * common/config/i386/i386-common.c (OPTION_MASK_ISA_AVX512VBMI2_SET):
1223         Or in OPTION_MASK_ISA_AVX512F_SET.
1224         (OPTION_MASK_ISA_AVX512F_UNSET): Or in
1225         OPTION_MASK_ISA_AVX512VBMI2_UNSET.
1226         (ix86_handle_option): Adjust for
1227         OPTION_MASK_ISA_{SHSTK,AVX512VBMI2}_*SET being in ix86_isa_flags
1228         and OPTION_MASK_ISA_{MOVBE,MWAITX,CLZERO}_*SET in ix86_isa_flags2.
1230 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1231             Alan Hayward  <alan.hayward@arm.com>
1232             David Sherwood  <david.sherwood@arm.com>
1234         * tree-data-ref.h (prune_runtime_alias_test_list): Take the
1235         factor as a poly_uint64 rather than an unsigned HOST_WIDE_INT.
1236         * tree-data-ref.c (prune_runtime_alias_test_list): Likewise.
1237         Track polynomial offsets.
1239 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1240             Alan Hayward  <alan.hayward@arm.com>
1241             David Sherwood  <david.sherwood@arm.com>
1243         * tree-vect-data-refs.c (vect_compute_data_ref_alignment):
1244         Treat drb->init as a poly_int.  Fail if its misalignment wrt
1245         vector_alignment isn't known.
1247 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1248             Alan Hayward  <alan.hayward@arm.com>
1249             David Sherwood  <david.sherwood@arm.com>
1251         * tree-vectorizer.h (_loop_vec_info): Add a versioning_threshold
1252         field.
1253         (LOOP_VINFO_VERSIONING_THRESHOLD): New macro
1254         (vect_loop_versioning): Take the loop versioning threshold as a
1255         separate parameter.
1256         * tree-vect-loop-manip.c (vect_loop_versioning): Likewise.
1257         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
1258         versioning_threshold.
1259         (vect_analyze_loop_2): Compute the loop versioning threshold
1260         whenever loop versioning is needed, and store it in the new
1261         field rather than combining it with the cost model threshold.
1262         (vect_transform_loop): Update call to vect_loop_versioning.
1263         Try to combine the loop versioning and cost thresholds here.
1265 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1266             Alan Hayward  <alan.hayward@arm.com>
1267             David Sherwood  <david.sherwood@arm.com>
1269         * tree-ssa-loop-ivopts.h (strip_offset): Return the offset as
1270         poly_uint64_pod rather than an unsigned HOST_WIDE_INT.
1271         * tree-loop-distribution.c (classify_builtin_st): Update accordingly.
1272         * tree-ssa-loop-ivopts.c (iv_use::addr_offset): Change from
1273         an unsigned HOST_WIDE_INT to a poly_uint64_pod.
1274         (group_compare_offset): Update accordingly.
1275         (split_small_address_groups_p): Likewise.
1276         (record_use): Take addr_offset as a poly_uint64 rather than
1277         an unsigned HOST_WIDE_INT.
1278         (strip_offset): Return the offset as a poly_uint64 rather than
1279         an unsigned HOST_WIDE_INT.
1280         (record_group_use, split_address_groups): Track polynomial offsets.
1281         (add_iv_candidate_for_use): Likewise.
1282         (addr_offset_valid_p): Take the offset as a poly_int64 rather
1283         than a HOST_WIDE_INT.
1284         (strip_offset_1): Return the offset as a poly_int64 rather than
1285         a HOST_WIDE_INT.
1287 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1288             Alan Hayward  <alan.hayward@arm.com>
1289             David Sherwood  <david.sherwood@arm.com>
1291         * tree.h (get_binfo_at_offset): Take the offset as a poly_int64
1292         rather than a HOST_WIDE_INT.
1293         * tree.c (get_binfo_at_offset): Likewise.
1295 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1296             Alan Hayward  <alan.hayward@arm.com>
1297             David Sherwood  <david.sherwood@arm.com>
1299         * ipa-prop.h (build_ref_for_offset): Take the offset as a poly_int64
1300         rather than a HOST_WIDE_INT.
1301         * tree-sra.c (build_ref_for_offset): Likewise.
1303 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1304             Alan Hayward  <alan.hayward@arm.com>
1305             David Sherwood  <david.sherwood@arm.com>
1307         * fold-const.h (mem_ref_offset): Return a poly_offset_int rather
1308         than an offset_int.
1309         * tree.c (mem_ref_offset): Likewise.
1310         (build_simple_mem_ref_loc): Treat MEM_REF offsets as poly_ints.
1311         * builtins.c (get_object_alignment_2): Likewise.
1312         * expr.c (get_inner_reference, expand_expr_real_1): Likewise.
1313         * gimple-fold.c (get_base_constructor): Likewise.
1314         * gimple-ssa-strength-reduction.c (restructure_reference): Likewise.
1315         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref):
1316         Likewise.
1317         * ipa-polymorphic-call.c
1318         (ipa_polymorphic_call_context::ipa_polymorphic_call_context): Likewise.
1319         * ipa-prop.c (compute_complex_assign_jump_func): Likewise.
1320         (get_ancestor_addr_info): Likewise.
1321         * ipa-param-manipulation.c (ipa_get_adjustment_candidate): Likewise.
1322         * match.pd: Likewise.
1323         * tree-data-ref.c (dr_analyze_innermost): Likewise.
1324         * tree-dfa.c (get_addr_base_and_unit_offset_1): Likewise.
1325         * tree-eh.c (tree_could_trap_p): Likewise.
1326         * tree-object-size.c (addr_object_size): Likewise.
1327         * tree-ssa-address.c (copy_ref_info): Likewise.
1328         * tree-ssa-alias.c (indirect_ref_may_alias_decl_p): Likewise.
1329         (indirect_refs_may_alias_p): Likewise.
1330         * tree-ssa-sccvn.c (copy_reference_ops_from_ref): Likewise.
1331         * tree-ssa.c (maybe_rewrite_mem_ref_base): Likewise.
1332         (non_rewritable_mem_ref_base): Likewise.
1333         * tree-vect-data-refs.c (vect_check_gather_scatter): Likewise.
1334         * tree-vrp.c (vrp_prop::check_array_ref): Likewise.
1335         * varasm.c (decode_addr_const): 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         * gimple-ssa-stor-merging.c (find_bswap_or_nop_load): Track polynomial
1342         offsets for MEM_REFs.
1344 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1345             Alan Hayward  <alan.hayward@arm.com>
1346             David Sherwood  <david.sherwood@arm.com>
1348         * tree-ssanames.h (adjust_ptr_info_misalignment): Take the increment
1349         as a poly_uint64 rather than an unsigned int.
1350         * tree-ssanames.c (adjust_ptr_info_misalignment): Likewise.
1352 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1353             Alan Hayward  <alan.hayward@arm.com>
1354             David Sherwood  <david.sherwood@arm.com>
1356         * varasm.c (addr_const::offset): Change from HOST_WIDE_INT
1357         to poly_int64.
1358         (decode_addr_const): Update accordingly.
1360 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1361             Alan Hayward  <alan.hayward@arm.com>
1362             David Sherwood  <david.sherwood@arm.com>
1364         * tree.h (bit_field_size, bit_field_offset): New functions.
1365         * hsa-gen.c (gen_hsa_addr): Use them.
1366         * tree-ssa-forwprop.c (simplify_bitfield_ref): Likewise.
1367         (simplify_vector_constructor): Likewise.
1368         * tree-ssa-sccvn.c (copy_reference_ops_from_ref): Likewise.
1369         * tree-cfg.c (verify_expr): Require the sizes and offsets of a
1370         BIT_FIELD_REF to be poly_uint64s rather than uhwis.
1371         * fold-const.c (fold_ternary_loc): Protect tree_to_uhwi with
1372         tree_fits_uhwi_p.
1374 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1375             Alan Hayward  <alan.hayward@arm.com>
1376             David Sherwood  <david.sherwood@arm.com>
1378         * expr.h (emit_group_load, emit_group_load_into_temps)
1379         (emit_group_store): Take the size as a poly_int64 rather than an int.
1380         * expr.c (emit_group_load_1, emit_group_load): Likewise.
1381         (emit_group_load_into_temp, emit_group_store): Likewise.
1383 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1384             Alan Hayward  <alan.hayward@arm.com>
1385             David Sherwood  <david.sherwood@arm.com>
1387         * ira-int.h (ira_spilled_reg_stack_slot::width): Change from
1388         an unsigned int to a poly_uint64.
1389         * ira.h (ira_reuse_stack_slot, ira_mark_new_stack_slot): Take the
1390         sizes as poly_uint64s rather than unsigned ints.
1391         * ira-color.c (ira_reuse_stack_slot, ira_mark_new_stack_slot):
1392         Likewise.
1394 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1395             Alan Hayward  <alan.hayward@arm.com>
1396             David Sherwood  <david.sherwood@arm.com>
1398         * lra-constraints.c (emit_inc): Change inc_amount from an int
1399         to a poly_int64.
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         * cfgexpand.c (stack_var::size): Change from a HOST_WIDE_INT
1406         to a poly_uint64.
1407         (add_stack_var, stack_var_cmp, partition_stack_vars)
1408         (dump_stack_var_partition): Update accordingly.
1409         (alloc_stack_frame_space): Take the size as a poly_int64 rather
1410         than a HOST_WIDE_INT.
1411         (expand_stack_vars, expand_one_stack_var_1): Handle polynomial sizes.
1412         (defer_stack_allocation, estimated_stack_frame_size): Likewise.
1413         (account_stack_vars, expand_one_var): Likewise.  Return a poly_uint64
1414         rather than a HOST_WIDE_INT.
1416 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1417             Alan Hayward  <alan.hayward@arm.com>
1418             David Sherwood  <david.sherwood@arm.com>
1420         * target.def (return_pops_args): Treat both the input and output
1421         sizes as poly_int64s rather than HOST_WIDE_INTS.
1422         * targhooks.h (default_return_pops_args): Update accordingly.
1423         * targhooks.c (default_return_pops_args): Likewise.
1424         * doc/tm.texi: Regenerate.
1425         * emit-rtl.h (incoming_args): Change pops_args, size and
1426         outgoing_args_size from int to poly_int64_pod.
1427         * function.h (expr_status): Change x_pending_stack_adjust and
1428         x_stack_pointer_delta from int to poly_int64.
1429         (args_size::constant): Change from HOST_WIDE_INT to poly_int64.
1430         (ARGS_SIZE_RTX): Update accordingly.
1431         * calls.c (highest_outgoing_arg_in_use): Change from int to
1432         unsigned int.
1433         (stack_usage_watermark, stored_args_watermark): New variables.
1434         (stack_region_maybe_used_p, mark_stack_region_used): New functions.
1435         (emit_call_1): Change the stack_size and rounded_stack_size
1436         parameters from HOST_WIDE_INT to poly_int64.  Track n_popped
1437         as a poly_int64.
1438         (save_fixed_argument_area): Check stack_usage_watermark.
1439         (initialize_argument_information): Change old_pending_adj from
1440         a HOST_WIDE_INT * to a poly_int64_pod *.
1441         (compute_argument_block_size): Return the size as a poly_int64
1442         rather than an int.
1443         (finalize_must_preallocate): Track polynomial argument sizes.
1444         (compute_argument_addresses): Likewise.
1445         (internal_arg_pointer_based_exp): Track polynomial offsets.
1446         (mem_overlaps_already_clobbered_arg_p): Rename to...
1447         (mem_might_overlap_already_clobbered_arg_p): ...this and take the
1448         size as a poly_uint64 rather than an unsigned HOST_WIDE_INT.
1449         Check stored_args_used_watermark.
1450         (load_register_parameters): Update accordingly.
1451         (check_sibcall_argument_overlap_1): Likewise.
1452         (combine_pending_stack_adjustment_and_call): Take the unadjusted
1453         args size as a poly_int64 rather than an int.  Return a bool
1454         indicating whether the optimization was possible and return
1455         the new adjustment by reference.
1456         (check_sibcall_argument_overlap): Track polynomail argument sizes.
1457         Update stored_args_watermark.
1458         (can_implement_as_sibling_call_p): Handle polynomial argument sizes.
1459         (expand_call): Likewise.  Maintain stack_usage_watermark and
1460         stored_args_watermark.  Update calls to
1461         combine_pending_stack_adjustment_and_call.
1462         (emit_library_call_value_1): Handle polynomial argument sizes.
1463         Call stack_region_maybe_used_p and mark_stack_region_used.
1464         Maintain stack_usage_watermark.
1465         (store_one_arg): Likewise.  Update call to
1466         mem_overlaps_already_clobbered_arg_p.
1467         * config/arm/arm.c (arm_output_function_prologue): Add a cast to
1468         HOST_WIDE_INT.
1469         * config/avr/avr.c (avr_outgoing_args_size): Likewise.
1470         * config/microblaze/microblaze.c (microblaze_function_prologue):
1471         Likewise.
1472         * config/cr16/cr16.c (cr16_return_pops_args): Update for new
1473         TARGET_RETURN_POPS_ARGS interface.
1474         (cr16_compute_frame, cr16_initial_elimination_offset): Add casts
1475         to HOST_WIDE_INT.
1476         * config/ft32/ft32.c (ft32_compute_frame): Likewise.
1477         * config/i386/i386.c (ix86_return_pops_args): Update for new
1478         TARGET_RETURN_POPS_ARGS interface.
1479         (ix86_expand_split_stack_prologue): Add a cast to HOST_WIDE_INT.
1480         * config/moxie/moxie.c (moxie_compute_frame): Likewise.
1481         * config/m68k/m68k.c (m68k_return_pops_args): Update for new
1482         TARGET_RETURN_POPS_ARGS interface.
1483         * config/vax/vax.c (vax_return_pops_args): Likewise.
1484         * config/pa/pa.h (STACK_POINTER_OFFSET): Add a cast to poly_int64.
1485         (EXIT_IGNORE_STACK): Update reference to crtl->outgoing_args_size.
1486         * config/arm/arm.h (CALLER_INTERWORKING_SLOT_SIZE): Likewise.
1487         * config/powerpcspe/aix.h (STACK_DYNAMIC_OFFSET): Likewise.
1488         * config/powerpcspe/darwin.h (STACK_DYNAMIC_OFFSET): Likewise.
1489         * config/powerpcspe/powerpcspe.h (STACK_DYNAMIC_OFFSET): Likewise.
1490         * config/rs6000/aix.h (STACK_DYNAMIC_OFFSET): Likewise.
1491         * config/rs6000/darwin.h (STACK_DYNAMIC_OFFSET): Likewise.
1492         * config/rs6000/rs6000.h (STACK_DYNAMIC_OFFSET): Likewise.
1493         * dojump.h (saved_pending_stack_adjust): Change x_pending_stack_adjust
1494         and x_stack_pointer_delta from int to poly_int64.
1495         * dojump.c (do_pending_stack_adjust): Update accordingly.
1496         * explow.c (allocate_dynamic_stack_space): Handle polynomial
1497         stack_pointer_deltas.
1498         * function.c (STACK_DYNAMIC_OFFSET): Add a cast to poly_int64.
1499         (pad_to_arg_alignment): Track polynomial offsets.
1500         (assign_parm_find_stack_rtl): Likewise.
1501         (assign_parms, locate_and_pad_parm): Handle polynomial argument sizes.
1502         * toplev.c (output_stack_usage): Update reference to
1503         current_function_pushed_stack_size.
1505 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1506             Alan Hayward  <alan.hayward@arm.com>
1507             David Sherwood  <david.sherwood@arm.com>
1509         * function.c (in_arg_offset, var_offset, dynamic_offset)
1510         (out_arg_offset, cfa_offset): Change from int to poly_int64.
1511         (instantiate_new_reg): Return the new offset as a poly_int64_pod
1512         rather than a HOST_WIDE_INT.
1513         (instantiate_virtual_regs_in_rtx): Track polynomial offsets.
1514         (instantiate_virtual_regs_in_insn): Likewise.
1516 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1517             Alan Hayward  <alan.hayward@arm.com>
1518             David Sherwood  <david.sherwood@arm.com>
1520         * rtl.h (get_args_size, add_args_size_note): New functions.
1521         (find_args_size_adjust): Return a poly_int64 rather than a
1522         HOST_WIDE_INT.
1523         (fixup_args_size_notes): Likewise.  Make the same change to the
1524         end_args_size parameter.
1525         * rtlanal.c (get_args_size, add_args_size_note): New functions.
1526         * builtins.c (expand_builtin_trap): Use add_args_size_note.
1527         * calls.c (emit_call_1): Likewise.
1528         * explow.c (adjust_stack_1): Likewise.
1529         * cfgcleanup.c (old_insns_match_p): Update use of
1530         find_args_size_adjust.
1531         * combine.c (distribute_notes): Track polynomial arg sizes.
1532         * dwarf2cfi.c (dw_trace_info): Change beg_true_args_size,
1533         end_true_args_size, beg_delay_args_size and end_delay_args_size
1534         from HOST_WIDE_INT to poly_int64.
1535         (add_cfi_args_size): Take the args_size as a poly_int64 rather
1536         than a HOST_WIDE_INT.
1537         (notice_args_size, notice_eh_throw, maybe_record_trace_start)
1538         (maybe_record_trace_start_abnormal, scan_trace, connect_traces): Track
1539         polynomial arg sizes.
1540         * emit-rtl.c (try_split): Use get_args_size.
1541         * recog.c (peep2_attempt): Likewise.
1542         * reload1.c (reload_as_needed): Likewise.
1543         * expr.c (find_args_size_adjust): Return the adjustment as a
1544         poly_int64 rather than a HOST_WIDE_INT.
1545         (fixup_args_size_notes): Change end_args_size from a HOST_WIDE_INT
1546         to a poly_int64 and change the return type in the same way.
1547         (emit_single_push_insn): Track polynomial arg sizes.
1549 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1550             Alan Hayward  <alan.hayward@arm.com>
1551             David Sherwood  <david.sherwood@arm.com>
1553         * expr.h (push_block, emit_push_insn): Change the "extra" parameter
1554         from HOST_WIDE_INT to poly_int64.
1555         * expr.c (push_block, emit_push_insn): Likewise.
1557 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1558             Alan Hayward  <alan.hayward@arm.com>
1559             David Sherwood  <david.sherwood@arm.com>
1561         * function.h (frame_space): Change start and length from HOST_WIDE_INT
1562         to poly_int64.
1563         (get_frame_size): Return the size as a poly_int64 rather than a
1564         HOST_WIDE_INT.
1565         (frame_offset_overflow): Take the offset as a poly_int64 rather
1566         than a HOST_WIDE_INT.
1567         (assign_stack_local_1, assign_stack_local, assign_stack_temp_for_type)
1568         (assign_stack_temp): Likewise for the size.
1569         * function.c (get_frame_size): Return a poly_int64 rather than
1570         a HOST_WIDE_INT.
1571         (frame_offset_overflow): Take the offset as a poly_int64 rather
1572         than a HOST_WIDE_INT.
1573         (try_fit_stack_local): Take the start, length and size as poly_int64s
1574         rather than HOST_WIDE_INTs.  Return the offset as a poly_int64_pod
1575         rather than a HOST_WIDE_INT.
1576         (add_frame_space): Take the start and end as poly_int64s rather than
1577         HOST_WIDE_INTs.
1578         (assign_stack_local_1, assign_stack_local, assign_stack_temp_for_type)
1579         (assign_stack_temp): Likewise for the size.
1580         (temp_slot): Change size, base_offset and full_size from HOST_WIDE_INT
1581         to poly_int64.
1582         (find_temp_slot_from_address): Handle polynomial offsets.
1583         (combine_temp_slots): Likewise.
1584         * emit-rtl.h (rtl_data::x_frame_offset): Change from HOST_WIDE_INT
1585         to poly_int64.
1586         * cfgexpand.c (alloc_stack_frame_space): Return the offset as a
1587         poly_int64 rather than a HOST_WIDE_INT.
1588         (expand_one_stack_var_at): Take the offset as a poly_int64 rather
1589         than a HOST_WIDE_INT.
1590         (expand_stack_vars, expand_one_stack_var_1, expand_used_vars): Handle
1591         polynomial frame offsets.
1592         * config/m32r/m32r-protos.h (m32r_compute_frame_size): Take the size
1593         as a poly_int64 rather than an int.
1594         * config/m32r/m32r.c (m32r_compute_frame_size): Likewise.
1595         * config/v850/v850-protos.h (compute_frame_size): Likewise.
1596         * config/v850/v850.c (compute_frame_size): Likewise.
1597         * config/xtensa/xtensa-protos.h (compute_frame_size): Likewise.
1598         * config/xtensa/xtensa.c (compute_frame_size): Likewise.
1599         * config/pa/pa-protos.h (pa_compute_frame_size): Likewise.
1600         * config/pa/pa.c (pa_compute_frame_size): Likewise.
1601         * explow.h (get_dynamic_stack_base): Take the offset as a poly_int64
1602         rather than a HOST_WIDE_INT.
1603         * explow.c (get_dynamic_stack_base): Likewise.
1604         * final.c (final_start_function): Use the constant lower bound
1605         of the frame size for -Wframe-larger-than.
1606         * ira.c (do_reload): Adjust for new get_frame_size return type.
1607         * lra.c (lra): Likewise.
1608         * reload1.c (reload): Likewise.
1609         * config/avr/avr.c (avr_asm_function_end_prologue): Likewise.
1610         * config/pa/pa.h (EXIT_IGNORE_STACK): Likewise.
1611         * rtlanal.c (get_initial_register_offset): Return the offset as
1612         a poly_int64 rather than a HOST_WIDE_INT.
1614 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1615             Alan Hayward  <alan.hayward@arm.com>
1616             David Sherwood  <david.sherwood@arm.com>
1618         * reload1.c (elim_table): Change initial_offset, offset and
1619         previous_offset from HOST_WIDE_INT to poly_int64_pod.
1620         (offsets_at): Change the target array's element type from
1621         HOST_WIDE_INT to poly_int64_pod.
1622         (set_label_offsets, eliminate_regs_1, eliminate_regs_in_insn)
1623         (elimination_costs_in_insn, update_eliminable_offsets)
1624         (verify_initial_elim_offsets, set_offsets_for_label)
1625         (init_eliminable_invariants): Update after above changes.
1627 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1628             Alan Hayward  <alan.hayward@arm.com>
1629             David Sherwood  <david.sherwood@arm.com>
1631         * reload.h (reload::inc): Change from an int to a poly_int64_pod.
1632         * reload.c (combine_reloads, debug_reload_to_stream): Likewise.
1633         (decomposition): Change start and end from HOST_WIDE_INT
1634         to poly_int64_pod.
1635         (decompose, immune_p): Update accordingly.
1636         (find_inc_amount): Return a poly_int64 rather than an int.
1637         * reload1.c (inc_for_reload): Take the inc_amount as a poly_int64
1638         rather than an int.
1640 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1641             Alan Hayward  <alan.hayward@arm.com>
1642             David Sherwood  <david.sherwood@arm.com>
1644         * tree.h (get_inner_reference): Return the bitsize and bitpos
1645         as poly_int64_pods rather than HOST_WIDE_INT.
1646         * fold-const.h (ptr_difference_const): Return the pointer difference
1647         as a poly_int64_pod rather than a HOST_WIDE_INT.
1648         * expr.c (get_inner_reference): Return the bitsize and bitpos
1649         as poly_int64_pods rather than HOST_WIDE_INT.
1650         (expand_expr_addr_expr_1, expand_expr_real_1): Track polynomial
1651         offsets and sizes.
1652         * fold-const.c (make_bit_field_ref): Take the bitpos as a poly_int64
1653         rather than a HOST_WIDE_INT.  Update call to get_inner_reference.
1654         (optimize_bit_field_compare): Update call to get_inner_reference.
1655         (decode_field_reference): Likewise.
1656         (fold_unary_loc): Track polynomial offsets and sizes.
1657         (split_address_to_core_and_offset): Return the bitpos as a
1658         poly_int64_pod rather than a HOST_WIDE_INT.
1659         (ptr_difference_const): Likewise for the pointer difference.
1660         * asan.c (instrument_derefs): Track polynomial offsets and sizes.
1661         * config/mips/mips.c (r10k_safe_mem_expr_p): Likewise.
1662         * dbxout.c (dbxout_expand_expr): Likewise.
1663         * dwarf2out.c (loc_list_for_address_of_addr_expr_of_indirect_ref)
1664         (loc_list_from_tree_1, fortran_common): Likewise.
1665         * gimple-laddress.c (pass_laddress::execute): Likewise.
1666         * gimple-ssa-store-merging.c (find_bswap_or_nop_load): Likewise.
1667         * gimplify.c (gimplify_scan_omp_clauses): Likewise.
1668         * simplify-rtx.c (delegitimize_mem_from_attrs): Likewise.
1669         * tree-affine.c (tree_to_aff_combination): Likewise.
1670         (get_inner_reference_aff): Likewise.
1671         * tree-data-ref.c (split_constant_offset_1): Likewise.
1672         (dr_analyze_innermost): Likewise.
1673         * tree-scalar-evolution.c (interpret_rhs_expr): Likewise.
1674         * tree-sra.c (ipa_sra_check_caller): Likewise.
1675         * tree-vect-data-refs.c (vect_check_gather_scatter): Likewise.
1676         * ubsan.c (maybe_instrument_pointer_overflow): Likewise.
1677         (instrument_bool_enum_load, instrument_object_size): Likewise.
1678         * gimple-ssa-strength-reduction.c (slsr_process_ref): Update call
1679         to get_inner_reference.
1680         * hsa-gen.c (gen_hsa_addr): Likewise.
1681         * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Likewise.
1682         * tsan.c (instrument_expr): Likewise.
1683         * match.pd: Update call to ptr_difference_const.
1685 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1686             Alan Hayward  <alan.hayward@arm.com>
1687             David Sherwood  <david.sherwood@arm.com>
1689         * fold-const.c (fold_comparison): Track sizes and offsets as
1690         poly_int64s rather than HOST_WIDE_INTs when folding address
1691         comparisons.
1693 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1694             Alan Hayward  <alan.hayward@arm.com>
1695             David Sherwood  <david.sherwood@arm.com>
1697         * expr.h (get_bit_range): Return the bitstart and bitend as
1698         poly_uint64s rather than unsigned HOST_WIDE_INTs.  Return the bitpos
1699         as a poly_int64 rather than a HOST_WIDE_INT.
1700         * expr.c (get_bit_range): Likewise.
1701         (expand_assignment): Update call accordingly.
1702         * fold-const.c (optimize_bit_field_compare): Likewise.
1704 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1705             Alan Hayward  <alan.hayward@arm.com>
1706             David Sherwood  <david.sherwood@arm.com>
1708         * config/aarch64/aarch64-protos.h (aarch64_addr_query_type): New enum.
1709         (aarch64_legitimate_address_p): Use it instead of an rtx code,
1710         as an optional final parameter.
1711         * config/aarch64/aarch64.c (aarch64_classify_address): Likewise.
1712         (aarch64_legitimate_address_p): Likewise.
1713         (aarch64_print_address_internal): Take an aarch64_addr_query_type
1714         instead of an rtx code.
1715         (aarch64_address_valid_for_prefetch_p): Update calls accordingly.
1716         (aarch64_legitimate_address_hook_p): Likewise.
1717         (aarch64_print_ldpstp_address): Likewise.
1718         (aarch64_print_operand_address): Likewise.
1719         (aarch64_address_cost): Likewise.
1720         * config/aarch64/constraints.md (Uml, Umq, Ump, Utq): Likewise.
1721         * config/aarch64/predicates.md (aarch64_mem_pair_operand): Likewise.
1722         (aarch64_mem_pair_lanes_operand): Likewise.
1724 2017-12-20  Richard Biener  <rguenther@suse.de>
1726         * tree-ssa-dom.c (dom_opt_dom_walker::optimize_stmt): Call
1727         update_stmt_if_modified.
1729 2017-12-20  Wilco Dijkstra  <wdijkstr@arm.com>
1731         PR tree-optimization/83491
1732         * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Check for SSA_NAME
1733         before walking uses.  Improve coding style and comments.
1735 2017-12-20  Tom de Vries  <tom@codesourcery.com>
1737         * gimple-fold.c (fold_internal_goacc_dim): Simplify.
1739 2017-12-20  Jakub Jelinek  <jakub@redhat.com>
1741         PR ipa/83506
1742         * ipa-fnsummary.c (pass_data_ipa_free_fn_summary): Use 0 for
1743         todo_flags_finish.
1744         (pass_ipa_free_fn_summary): Add small_p private data member,
1745         initialize to false in the ctor.
1746         (pass_ipa_free_fn_summary::clone,
1747         pass_ipa_free_fn_summary::set_pass_param,
1748         pass_ipa_free_fn_summary::gate): New methods.
1749         (pass_ipa_free_fn_summary::execute): Return TODO_remove_functions
1750         | TODO_dump_symtab if small_p.
1751         * passes.def: Add true parm for the existing pass_ipa_free_fn_summary
1752         entry and add another instance of the pass with false parm after
1753         ipa-pure-const.
1754         * ipa-pure-const.c (pass_ipa_pure_const): Don't call
1755         ipa_free_fn_summary here.
1757 2017-12-20  Paolo Carlini  <paolo.carlini@oracle.com>
1759         * gimplify.c (gimplify_return_expr): Remove dead error_mark_node check.
1761 2017-12-20  Martin Sebor  <msebor@redhat.com>
1763         PR testsuite/83131
1764         * builtins.c (expand_builtin_strlen): Use get_callee_fndecl.
1765         (expand_builtin_strcmp): Call maybe_warn_nonstring_arg.
1766         (expand_builtin_strncmp): Same.
1768 2017-12-20  Alexandre Oliva  <aoliva@redhat.com>
1770         PR bootstrap/83396
1771         * cfgexpand.c (label_rtx_for_bb): Revert SFN changes that
1772         allowed debug stmts before labels.
1773         (expand_gimple_basic_block): Likewise.
1774         * gimple-iterator.c (gimple_find_edge_insert_loc): Likewise.
1775         * gimple-iterator.h (gsi_after_labels): Likewise.
1776         * tree-cfgcleanup (remove_forwarder_block): Likewise, but
1777         rename reused variable, and simplify using gsi_move_before.
1778         * tree-ssa-tail-merge.c (find_duplicate): Likewise.
1779         * tree-cfg.c (make_edges, cleanup_dead_labels): Likewise.
1780         (gimple_can_merge_blocks_p, verify_gimple_in_cfg): Likewise.
1781         (gimple_verify_flow_info, gimple_block_label): Likewise.
1782         (make_blocks): Move debug markers after adjacent labels.
1783         * cfgrtl.c (skip_insns_after_block): Revert SFN changes that
1784         allowed debug insns outside blocks.
1785         * df-scan.c (df_insn_delete): Likewise.
1786         * lra-constraints.c (update_ebb_live_info): Likewise.
1787         * var-tracking.c (get_first_insn, vt_emit_notes): Likewise.
1788         (vt_initialize, delete_vta_debug_insns): Likewise.
1789         (reemit_marker_as_note): Drop BB parm.  Adjust callers.
1791 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
1792             Alan Hayward  <alan.hayward@arm.com>
1793             David Sherwood  <david.sherwood@arm.com>
1795         * poly-int-types.h (round_down_to_byte_boundary): New macro.
1796         (round_up_to_byte_boundary): Likewise.
1797         * expr.h (get_bit_range): Add temporary shim.
1798         * gimple-ssa-store-merging.c (store_operand_info): Change the
1799         bitsize, bitpos, bitregion_start and bitregion_end fields from
1800         unsigned HOST_WIDE_INT to poly_uint64.
1801         (merged_store_group): Likewise load_align_base.
1802         (compatible_load_p, compatible_load_p): Update accordingly.
1803         (imm_store_chain_info::coalesce_immediate_stores): Likewise.
1804         (split_group, imm_store_chain_info::output_merged_store): Likewise.
1805         (mem_valid_for_store_merging): Return the bitsize, bitpos,
1806         bitregion_start and bitregion_end as poly_uint64s rather than
1807         unsigned HOST_WIDE_INTs.  Track polynomial offsets internally.
1808         (handled_load): Take the bitsize, bitpos,
1809         bitregion_start and bitregion_end as poly_uint64s rather than
1810         unsigned HOST_WIDE_INTs.
1811         (pass_store_merging::process_store): Update call to
1812         mem_valid_for_store_merging.
1814 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
1815             Alan Hayward  <alan.hayward@arm.com>
1816             David Sherwood  <david.sherwood@arm.com>
1818         * builtins.c (get_object_alignment_2): Track polynomial offsets
1819         and sizes.  Update the alignment handling.
1821 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
1822             Alan Hayward  <alan.hayward@arm.com>
1823             David Sherwood  <david.sherwood@arm.com>
1825         * tree.h (get_inner_reference): Add a version that returns the
1826         offset and size as poly_int64_pods rather than HOST_WIDE_INTs.
1827         * cfgexpand.c (expand_debug_expr): Track polynomial offsets.  Simply
1828         the case in which bitpos is not associated with the first byte.
1830 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
1831             Alan Hayward  <alan.hayward@arm.com>
1832             David Sherwood  <david.sherwood@arm.com>
1834         * tree-affine.h (get_inner_reference_aff): Return the size as a
1835         poly_widest_int.
1836         * tree-affine.c (get_inner_reference_aff): Likewise.
1837         * tree-data-ref.c (dr_may_alias_p): Update accordingly.
1838         * tree-ssa-loop-im.c (mem_refs_may_alias_p): Likewise.
1840 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
1841             Alan Hayward  <alan.hayward@arm.com>
1842             David Sherwood  <david.sherwood@arm.com>
1844         * fold-const.c (pointer_may_wrap_p): Take the offset as a
1845         HOST_WIDE_INT rather than a poly_int64.
1847 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
1848             Alan Hayward  <alan.hayward@arm.com>
1849             David Sherwood  <david.sherwood@arm.com>
1851         * gimple-ssa-store-merging.c (symbolic_number::bytepos): Change from
1852         HOST_WIDE_INT to poly_int64_pod.
1853         (perform_symbolic_merge): Update accordingly.
1854         (bswap_replace): Likewise.
1856 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
1857             Alan Hayward  <alan.hayward@arm.com>
1858             David Sherwood  <david.sherwood@arm.com>
1860         * tree-affine.h (aff_tree::offset): Change from widest_int
1861         to poly_widest_int.
1862         (wide_int_ext_for_comb): Delete.
1863         (aff_combination_const, aff_comb_cannot_overlap_p): Take the
1864         constants as poly_widest_int rather than widest_int.
1865         (aff_combination_constant_multiple_p): Return the multiplier
1866         as a poly_widest_int.
1867         (aff_combination_zero_p, aff_combination_singleton_var_p): Handle
1868         polynomial offsets.
1869         * tree-affine.c (wide_int_ext_for_comb): Make original widest_int
1870         version static and add an overload for poly_widest_int.
1871         (aff_combination_const, aff_combination_add_cst)
1872         (wide_int_constant_multiple_p, aff_comb_cannot_overlap_p): Take
1873         the constants as poly_widest_int rather than widest_int.
1874         (tree_to_aff_combination): Generalize INTEGER_CST case to
1875         poly_int_tree_p.
1876         (aff_combination_to_tree): Track offsets as poly_widest_ints.
1877         (aff_combination_add_product, aff_combination_mult): Handle
1878         polynomial offsets.
1879         (aff_combination_constant_multiple_p): Return the multiplier
1880         as a poly_widest_int.
1881         * tree-predcom.c (determine_offset): Return the offset as a
1882         poly_widest_int.
1883         (split_data_refs_to_components, suitable_component_p): Update
1884         accordingly.
1885         (valid_initializer_p): Update call to
1886         aff_combination_constant_multiple_p.
1887         * tree-ssa-address.c (addr_to_parts): Handle polynomial offsets.
1888         * tree-ssa-loop-ivopts.c (get_address_cost_ainc): Take the step
1889         as a poly_int64 rather than a HOST_WIDE_INT.
1890         (get_address_cost): Handle polynomial offsets.
1891         (iv_elimination_compare_lt): Likewise.
1892         (rewrite_use_nonlinear_expr): Likewise.
1894 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
1895             Alan Hayward  <alan.hayward@arm.com>
1896             David Sherwood  <david.sherwood@arm.com>
1898         * tree-dfa.h (get_addr_base_and_unit_offset_1): Return the offset
1899         as a poly_int64_pod rather than a HOST_WIDE_INT.
1900         (get_addr_base_and_unit_offset): Likewise.
1901         * tree-dfa.c (get_addr_base_and_unit_offset_1): Likewise.
1902         (get_addr_base_and_unit_offset): Likewise.
1903         * doc/match-and-simplify.texi: Change off from HOST_WIDE_INT
1904         to poly_int64 in example.
1905         * fold-const.c (fold_binary_loc): Update call to
1906         get_addr_base_and_unit_offset.
1907         * gimple-fold.c (gimple_fold_builtin_memory_op): Likewise.
1908         (maybe_canonicalize_mem_ref_addr): Likewise.
1909         (gimple_fold_stmt_to_constant_1): Likewise.
1910         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref):
1911         Likewise.
1912         * ipa-param-manipulation.c (ipa_modify_call_arguments): Likewise.
1913         * match.pd: Likewise.
1914         * omp-low.c (lower_omp_target): Likewise.
1915         * tree-sra.c (build_ref_for_offset): Likewise.
1916         (build_debug_ref_for_model): Likewise.
1917         * tree-ssa-address.c (maybe_fold_tmr): Likewise.
1918         * tree-ssa-alias.c (ao_ref_init_from_ptr_and_size): Likewise.
1919         * tree-ssa-ccp.c (optimize_memcpy): Likewise.
1920         * tree-ssa-forwprop.c (forward_propagate_addr_expr_1): Likewise.
1921         (constant_pointer_difference): Likewise.
1922         * tree-ssa-loop-niter.c (expand_simple_operations): Likewise.
1923         * tree-ssa-phiopt.c (jump_function_from_stmt): Likewise.
1924         * tree-ssa-pre.c (create_component_ref_by_pieces_1): Likewise.
1925         * tree-ssa-sccvn.c (vn_reference_fold_indirect): Likewise.
1926         (vn_reference_maybe_forwprop_address, vn_reference_lookup_3): Likewise.
1927         (set_ssa_val_to): Likewise.
1928         * tree-ssa-strlen.c (get_addr_stridx, addr_stridxptr)
1929         (maybe_diag_stxncpy_trunc): Likewise.
1930         * tree-vrp.c (vrp_prop::check_array_ref): Likewise.
1931         * tree.c (build_simple_mem_ref_loc): Likewise.
1932         (array_at_struct_end_p): Likewise.
1934 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
1935             Alan Hayward  <alan.hayward@arm.com>
1936             David Sherwood  <david.sherwood@arm.com>
1938         * tree-dfa.h (get_ref_base_and_extent): Return the base, size and
1939         max_size as poly_int64_pods rather than HOST_WIDE_INTs.
1940         (get_ref_base_and_extent_hwi): Declare.
1941         * tree-dfa.c (get_ref_base_and_extent): Return the base, size and
1942         max_size as poly_int64_pods rather than HOST_WIDE_INTs.
1943         (get_ref_base_and_extent_hwi): New function.
1944         * cfgexpand.c (expand_debug_expr): Update call to
1945         get_ref_base_and_extent.
1946         * dwarf2out.c (add_var_loc_to_decl): Likewise.
1947         * gimple-fold.c (get_base_constructor): Return the offset as a
1948         poly_int64_pod rather than a HOST_WIDE_INT.
1949         (fold_const_aggregate_ref_1): Track polynomial sizes and offsets.
1950         * ipa-polymorphic-call.c
1951         (ipa_polymorphic_call_context::set_by_invariant)
1952         (extr_type_from_vtbl_ptr_store): Track polynomial offsets.
1953         (ipa_polymorphic_call_context::ipa_polymorphic_call_context)
1954         (check_stmt_for_type_change): Use get_ref_base_and_extent_hwi
1955         rather than get_ref_base_and_extent.
1956         (ipa_polymorphic_call_context::get_dynamic_type): Likewise.
1957         * ipa-prop.c (ipa_load_from_parm_agg, compute_complex_assign_jump_func)
1958         (get_ancestor_addr_info, determine_locally_known_aggregate_parts):
1959         Likewise.
1960         * ipa-param-manipulation.c (ipa_get_adjustment_candidate): Update
1961         call to get_ref_base_and_extent.
1962         * tree-sra.c (create_access, get_access_for_expr): Likewise.
1963         * tree-ssa-alias.c (ao_ref_base, aliasing_component_refs_p)
1964         (stmt_kills_ref_p): Likewise.
1965         * tree-ssa-dce.c (mark_aliased_reaching_defs_necessary_1): Likewise.
1966         * tree-ssa-scopedtables.c (avail_expr_hash, equal_mem_array_ref_p):
1967         Likewise.
1968         * tree-ssa-sccvn.c (vn_reference_lookup_3): Likewise.
1969         Use get_ref_base_and_extent_hwi rather than get_ref_base_and_extent
1970         when calling native_encode_expr.
1971         * tree-ssa-structalias.c (get_constraint_for_component_ref): Update
1972         call to get_ref_base_and_extent.
1973         (do_structure_copy): Use get_ref_base_and_extent_hwi rather than
1974         get_ref_base_and_extent.
1975         * var-tracking.c (track_expr_p): Likewise.
1977 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
1978             Alan Hayward  <alan.hayward@arm.com>
1979             David Sherwood  <david.sherwood@arm.com>
1981         * ipa-param-manipulation.h (ipa_parm_adjustment::offset): Change from
1982         HOST_WIDE_INT to poly_int64_pod.
1983         * ipa-param-manipulation.c (ipa_modify_call_arguments): Track
1984         polynomail parameter offsets.
1986 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
1987             Alan Hayward  <alan.hayward@arm.com>
1988             David Sherwood  <david.sherwood@arm.com>
1990         * gengtype.c (main): Handle poly_int64_pod.
1991         * dwarf2out.h (dw_cfi_oprnd_cfa_loc): New dw_cfi_oprnd_type.
1992         (dw_cfi_oprnd::dw_cfi_cfa_loc): New field.
1993         (dw_cfa_location::offset, dw_cfa_location::base_offset): Change
1994         from HOST_WIDE_INT to poly_int64_pod.
1995         * dwarf2cfi.c (queued_reg_save::cfa_offset): Likewise.
1996         (copy_cfa): New function.
1997         (lookup_cfa_1): Use the cached dw_cfi_cfa_loc, if it exists.
1998         (cfi_oprnd_equal_p): Handle dw_cfi_oprnd_cfa_loc.
1999         (cfa_equal_p, dwarf2out_frame_debug_adjust_cfa)
2000         (dwarf2out_frame_debug_cfa_offset, dwarf2out_frame_debug_expr)
2001         (initial_return_save): Treat offsets as poly_ints.
2002         (def_cfa_0): Likewise.  Cache the CFA in dw_cfi_cfa_loc if either
2003         offset is nonconstant.
2004         (reg_save): Take the offset as a poly_int64.  Fall back to
2005         DW_CFA_expression for nonconstant offsets.
2006         (queue_reg_save): Take the offset as a poly_int64.
2007         * dwarf2out.c (dw_cfi_oprnd2_desc): Handle DW_CFA_def_cfa_expression.
2009 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2010             Alan Hayward  <alan.hayward@arm.com>
2011             David Sherwood  <david.sherwood@arm.com>
2013         * rtl.h (operand_subword, operand_subword_force): Take the offset
2014         as a poly_uint64 an unsigned int.
2015         * emit-rtl.c (operand_subword, operand_subword_force): Likewise.
2017 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2018             Alan Hayward  <alan.hayward@arm.com>
2019             David Sherwood  <david.sherwood@arm.com>
2021         * doc/rtl.texi: Update documentation of SUBREG_BYTE.  Document the
2022         'p' format code.  Use INT_LIST rather than SUBREG as the example of
2023         a code with an XINT and an XEXP.  Remove the implication that
2024         accessing an rtx field using XINT is expected to work.
2025         * rtl.def (SUBREG): Change format from "ei" to "ep".
2026         * rtl.h (rtunion::rt_subreg): New field.
2027         (XCSUBREG): New macro.
2028         (SUBREG_BYTE): Use it.
2029         (subreg_shape): Change offset from an unsigned int to a poly_uint16.
2030         Update constructor accordingly.
2031         (subreg_shape::operator ==): Update accordingly.
2032         (subreg_shape::unique_id): Return an unsigned HOST_WIDE_INT rather
2033         than an unsigned int.
2034         (subreg_lsb, subreg_lowpart_offset, subreg_highpart_offset): Return
2035         a poly_uint64 rather than an unsigned int.
2036         (subreg_lsb_1): Likewise.  Take the offset as a poly_uint64 rather
2037         than an unsigned int.
2038         (subreg_size_offset_from_lsb, subreg_size_lowpart_offset)
2039         (subreg_size_highpart_offset): Return a poly_uint64 rather than
2040         an unsigned int.  Take the sizes as poly_uint64s.
2041         (subreg_offset_from_lsb): Return a poly_uint64 rather than
2042         an unsigned int.  Take the shift as a poly_uint64 rather than
2043         an unsigned int.
2044         (subreg_regno_offset, subreg_offset_representable_p): Take the offset
2045         as a poly_uint64 rather than an unsigned int.
2046         (simplify_subreg_regno): Likewise.
2047         (byte_lowpart_offset): Return the memory offset as a poly_int64
2048         rather than an int.
2049         (subreg_memory_offset): Likewise.  Take the subreg offset as a
2050         poly_uint64 rather than an unsigned int.
2051         (simplify_subreg, simplify_gen_subreg, subreg_get_info)
2052         (gen_rtx_SUBREG, validate_subreg): Take the subreg offset as a
2053         poly_uint64 rather than an unsigned int.
2054         * rtl.c (rtx_format): Describe 'p' in comment.
2055         (copy_rtx, rtx_equal_p_cb, rtx_equal_p): Handle 'p'.
2056         * emit-rtl.c (validate_subreg, gen_rtx_SUBREG): Take the subreg
2057         offset as a poly_uint64 rather than an unsigned int.
2058         (byte_lowpart_offset): Return the memory offset as a poly_int64
2059         rather than an int.
2060         (subreg_memory_offset): Likewise.  Take the subreg offset as a
2061         poly_uint64 rather than an unsigned int.
2062         (subreg_size_lowpart_offset, subreg_size_highpart_offset): Take the
2063         mode sizes as poly_uint64s rather than unsigned ints.  Return a
2064         poly_uint64 rather than an unsigned int.
2065         (subreg_lowpart_p): Treat subreg offsets as poly_ints.
2066         (copy_insn_1): Handle 'p'.
2067         * rtlanal.c (set_noop_p): Treat subregs offsets as poly_uint64s.
2068         (subreg_lsb_1): Take the subreg offset as a poly_uint64 rather than
2069         an unsigned int.  Return the shift in the same way.
2070         (subreg_lsb): Return the shift as a poly_uint64 rather than an
2071         unsigned int.
2072         (subreg_size_offset_from_lsb): Take the sizes and shift as
2073         poly_uint64s rather than unsigned ints.  Return the offset as
2074         a poly_uint64.
2075         (subreg_get_info, subreg_regno_offset, subreg_offset_representable_p)
2076         (simplify_subreg_regno): Take the offset as a poly_uint64 rather than
2077         an unsigned int.
2078         * rtlhash.c (add_rtx): Handle 'p'.
2079         * genemit.c (gen_exp): Likewise.
2080         * gengenrtl.c (type_from_format, gendef): Likewise.
2081         * gensupport.c (subst_pattern_match, get_alternatives_number)
2082         (collect_insn_data, alter_predicate_for_insn, alter_constraints)
2083         (subst_dup): Likewise.
2084         * gengtype.c (adjust_field_rtx_def): Likewise.
2085         * genrecog.c (find_operand, find_matching_operand, validate_pattern)
2086         (match_pattern_2): Likewise.
2087         (rtx_test::SUBREG_FIELD): New rtx_test::kind_enum.
2088         (rtx_test::subreg_field): New function.
2089         (operator ==, safe_to_hoist_p, transition_parameter_type)
2090         (print_nonbool_test, print_test): Handle SUBREG_FIELD.
2091         * genattrtab.c (attr_rtx_1): Say that 'p' is deliberately not handled.
2092         * genpeep.c (match_rtx): Likewise.
2093         * print-rtl.c (print_poly_int): Include if GENERATOR_FILE too.
2094         (rtx_writer::print_rtx_operand): Handle 'p'.
2095         (print_value): Handle SUBREG.
2096         * read-rtl.c (apply_int_iterator): Likewise.
2097         (rtx_reader::read_rtx_operand): Handle 'p'.
2098         * alias.c (rtx_equal_for_memref_p): Likewise.
2099         * cselib.c (rtx_equal_for_cselib_1, cselib_hash_rtx): Likewise.
2100         * caller-save.c (replace_reg_with_saved_mem): Treat subreg offsets
2101         as poly_ints.
2102         * calls.c (expand_call): Likewise.
2103         * combine.c (combine_simplify_rtx, expand_field_assignment): Likewise.
2104         (make_extraction, gen_lowpart_for_combine): Likewise.
2105         * loop-invariant.c (hash_invariant_expr_1, invariant_expr_equal_p):
2106         Likewise.
2107         * cse.c (remove_invalid_subreg_refs): Take the offset as a poly_uint64
2108         rather than an unsigned int.  Treat subreg offsets as poly_ints.
2109         (exp_equiv_p): Handle 'p'.
2110         (hash_rtx_cb): Likewise.  Treat subreg offsets as poly_ints.
2111         (equiv_constant, cse_insn): Treat subreg offsets as poly_ints.
2112         * dse.c (find_shift_sequence): Likewise.
2113         * dwarf2out.c (rtl_for_decl_location): Likewise.
2114         * expmed.c (extract_low_bits): Likewise.
2115         * expr.c (emit_group_store, undefined_operand_subword_p): Likewise.
2116         (expand_expr_real_2): Likewise.
2117         * final.c (alter_subreg): Likewise.
2118         (leaf_renumber_regs_insn): Handle 'p'.
2119         * function.c (assign_parm_find_stack_rtl, assign_parm_setup_stack):
2120         Treat subreg offsets as poly_ints.
2121         * fwprop.c (forward_propagate_and_simplify): Likewise.
2122         * ifcvt.c (noce_emit_move_insn, noce_emit_cmove): Likewise.
2123         * ira.c (get_subreg_tracking_sizes): Likewise.
2124         * ira-conflicts.c (go_through_subreg): Likewise.
2125         * ira-lives.c (process_single_reg_class_operands): Likewise.
2126         * jump.c (rtx_renumbered_equal_p): Likewise.  Handle 'p'.
2127         * lower-subreg.c (simplify_subreg_concatn): Take the subreg offset
2128         as a poly_uint64 rather than an unsigned int.
2129         (simplify_gen_subreg_concatn, resolve_simple_move): Treat
2130         subreg offsets as poly_ints.
2131         * lra-constraints.c (operands_match_p): Handle 'p'.
2132         (match_reload, curr_insn_transform): Treat subreg offsets as poly_ints.
2133         * lra-spills.c (assign_mem_slot): Likewise.
2134         * postreload.c (move2add_valid_value_p): Likewise.
2135         * recog.c (general_operand, indirect_operand): Likewise.
2136         * regcprop.c (copy_value, maybe_mode_change): Likewise.
2137         (copyprop_hardreg_forward_1): Likewise.
2138         * reginfo.c (simplifiable_subregs_hasher::hash, simplifiable_subregs)
2139         (record_subregs_of_mode): Likewise.
2140         * rtlhooks.c (gen_lowpart_general, gen_lowpart_if_possible): Likewise.
2141         * reload.c (operands_match_p): Handle 'p'.
2142         (find_reloads_subreg_address): Treat subreg offsets as poly_ints.
2143         * reload1.c (alter_reg, choose_reload_regs): Likewise.
2144         (compute_reload_subreg_offset): Likewise, and return an poly_int64.
2145         * simplify-rtx.c (simplify_truncation, simplify_binary_operation_1):
2146         (test_vector_ops_duplicate): Treat subreg offsets as poly_ints.
2147         (simplify_const_poly_int_tests<N>::run): Likewise.
2148         (simplify_subreg, simplify_gen_subreg): Take the subreg offset as
2149         a poly_uint64 rather than an unsigned int.
2150         * valtrack.c (debug_lowpart_subreg): Likewise.
2151         * var-tracking.c (var_lowpart): Likewise.
2152         (loc_cmp): Handle 'p'.
2154 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2155             Alan Hayward  <alan.hayward@arm.com>
2156             David Sherwood  <david.sherwood@arm.com>
2158         * ira.c (get_subreg_tracking_sizes): New function.
2159         (init_live_subregs): Take an integer size rather than a register.
2160         (build_insn_chain): Use get_subreg_tracking_sizes.  Update calls
2161         to init_live_subregs.
2163 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2164             Alan Hayward  <alan.hayward@arm.com>
2165             David Sherwood  <david.sherwood@arm.com>
2167         * expr.c (store_constructor_field): Change bitsize from a
2168         unsigned HOST_WIDE_INT to a poly_uint64 and bitpos from a
2169         HOST_WIDE_INT to a poly_int64.
2170         (store_constructor): Change size from a HOST_WIDE_INT to
2171         a poly_int64.
2172         (store_field): Likewise bitsize and bitpos.
2174 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2175             Alan Hayward  <alan.hayward@arm.com>
2176             David Sherwood  <david.sherwood@arm.com>
2178         * expmed.h (store_bit_field): Change bitregion_start and
2179         bitregion_end from unsigned HOST_WIDE_INT to poly_uint64.
2180         * expmed.c (adjust_bit_field_mem_for_reg, strict_volatile_bitfield_p)
2181         (store_bit_field_1, store_integral_bit_field, store_bit_field)
2182         (store_fixed_bit_field, store_split_bit_field): Likewise.
2183         * expr.c (store_constructor_field, store_field): Likewise.
2184         (optimize_bitfield_assignment_op): Likewise.  Make the same change
2185         to bitsize and bitpos.
2186         * machmode.h (bit_field_mode_iterator): Change m_bitregion_start
2187         and m_bitregion_end from HOST_WIDE_INT to poly_int64.  Make the
2188         same change in the constructor arguments.
2189         (get_best_mode): Change bitregion_start and bitregion_end from
2190         unsigned HOST_WIDE_INT to poly_uint64.
2191         * stor-layout.c (bit_field_mode_iterator::bit_field_mode_iterator):
2192         Change bitregion_start and bitregion_end from HOST_WIDE_INT to
2193         poly_int64.
2194         (bit_field_mode_iterator::next_mode): Update for new types
2195         of m_bitregion_start and m_bitregion_end.
2196         (get_best_mode): Change bitregion_start and bitregion_end from
2197         unsigned HOST_WIDE_INT to poly_uint64.
2199 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2200             Alan Hayward  <alan.hayward@arm.com>
2201             David Sherwood  <david.sherwood@arm.com>
2203         * rtl.h (simplify_gen_subreg): Add a temporary overload that
2204         accepts poly_uint64 offsets.
2205         * expmed.h (extract_bit_field): Take bitsize and bitnum as
2206         poly_uint64s rather than unsigned HOST_WIDE_INTs.
2207         * expmed.c (lowpart_bit_field_p): Likewise.
2208         (extract_bit_field_as_subreg): New function, split out from...
2209         (extract_bit_field_1): ...here.  Take bitsize and bitnum as
2210         poly_uint64s rather than unsigned HOST_WIDE_INTs.  For vector
2211         extractions, check that BITSIZE matches the size of the extracted
2212         value and that BITNUM is an exact multiple of that size.
2213         If all else fails, try forcing the value into memory if
2214         BITNUM is variable, and adjusting the address so that the
2215         offset is constant.  Split the part that can only handle constant
2216         bitsize and bitnum out into...
2217         (extract_integral_bit_field): ...this new function.
2218         (extract_bit_field): Take bitsize and bitnum as poly_uint64s
2219         rather than unsigned HOST_WIDE_INTs.
2221 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2222             Alan Hayward  <alan.hayward@arm.com>
2223             David Sherwood  <david.sherwood@arm.com>
2225         * expmed.h (store_bit_field): Take bitsize and bitnum as
2226         poly_uint64s rather than unsigned HOST_WIDE_INTs.
2227         * expmed.c (simple_mem_bitfield_p): Likewise.  Add a parameter
2228         that returns the byte size.
2229         (store_bit_field_1): Take bitsize and bitnum as
2230         poly_uint64s rather than unsigned HOST_WIDE_INTs.  Update call
2231         to simple_mem_bitfield_p.  Split the part that can only handle
2232         constant bitsize and bitnum out into...
2233         (store_integral_bit_field): ...this new function.
2234         (store_bit_field): Take bitsize and bitnum as poly_uint64s rather
2235         than unsigned HOST_WIDE_INTs.
2236         (extract_bit_field_1): Update call to simple_mem_bitfield_p.
2238 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2239             Alan Hayward  <alan.hayward@arm.com>
2240             David Sherwood  <david.sherwood@arm.com>
2242         * lra-int.h (lra_reg): Change offset from int to poly_int64.
2243         (lra_insn_recog_data): Change sp_offset from HOST_WIDE_INT
2244         to poly_int64.
2245         (lra_eliminate_regs_1, eliminate_regs_in_insn): Change
2246         update_sp_offset from a HOST_WIDE_INT to a poly_int64.
2247         (lra_update_reg_val_offset, lra_reg_val_equal_p): Take the
2248         offset as a poly_int64 rather than an int.
2249         * lra-assigns.c (find_hard_regno_for_1): Handle poly_int64 offsets.
2250         (setup_live_pseudos_and_spill_after_risky_transforms): Likewise.
2251         * lra-constraints.c (equiv_address_substitution): Track offsets
2252         as poly_int64s.
2253         (emit_inc): Check poly_int_rtx_p instead of CONST_INT_P.
2254         (curr_insn_transform): Handle the new form of sp_offset.
2255         * lra-eliminations.c (lra_elim_table): Change previous_offset
2256         and offset from HOST_WIDE_INT to poly_int64.
2257         (print_elim_table, update_reg_eliminate): Update accordingly.
2258         (self_elim_offsets): Change from HOST_WIDE_INT to poly_int64_pod.
2259         (get_elimination): Update accordingly.
2260         (form_sum): Check poly_int_rtx_p instead of CONST_INT_P.
2261         (lra_eliminate_regs_1, eliminate_regs_in_insn): Change
2262         update_sp_offset from a HOST_WIDE_INT to a poly_int64.  Handle
2263         poly_int64 offsets generally.
2264         (curr_sp_change): Change from HOST_WIDE_INT to poly_int64.
2265         (mark_not_eliminable, init_elimination): Update accordingly.
2266         (remove_reg_equal_offset_note): Return a bool and pass the new
2267         offset back by pointer as a poly_int64.
2268         * lra-remat.c (change_sp_offset): Take sp_offset as a poly_int64
2269         rather than a HOST_WIDE_INT.
2270         (do_remat): Track offsets poly_int64s.
2271         * lra.c (lra_update_insn_recog_data, setup_sp_offset): Likewise.
2273 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2274             Alan Hayward  <alan.hayward@arm.com>
2275             David Sherwood  <david.sherwood@arm.com>
2277         * rtl.h (mem_attrs): Add a default constructor.  Change size and
2278         offset from HOST_WIDE_INT to poly_int64.
2279         * emit-rtl.h (set_mem_offset, set_mem_size, adjust_address_1)
2280         (adjust_automodify_address_1, set_mem_attributes_minus_bitpos)
2281         (widen_memory_access): Take the sizes and offsets as poly_int64s
2282         rather than HOST_WIDE_INTs.
2283         * alias.c (ao_ref_from_mem): Handle the new form of MEM_OFFSET.
2284         (offset_overlap_p): Take poly_int64s rather than HOST_WIDE_INTs
2285         and ints.
2286         (adjust_offset_for_component_ref): Change the offset from a
2287         HOST_WIDE_INT to a poly_int64.
2288         (nonoverlapping_memrefs_p): Track polynomial offsets and sizes.
2289         * cfgcleanup.c (merge_memattrs): Update after mem_attrs changes.
2290         * dce.c (find_call_stack_args): Likewise.
2291         * dse.c (record_store): Likewise.
2292         * dwarf2out.c (tls_mem_loc_descriptor, dw_sra_loc_expr): Likewise.
2293         * print-rtl.c (rtx_writer::print_rtx): Likewise.
2294         * read-rtl-function.c (test_loading_mem): Likewise.
2295         * rtlanal.c (may_trap_p_1): Likewise.
2296         * simplify-rtx.c (delegitimize_mem_from_attrs): Likewise.
2297         * var-tracking.c (int_mem_offset, track_expr_p): Likewise.
2298         * emit-rtl.c (mem_attrs_eq_p, get_mem_align_offset): Likewise.
2299         (mem_attrs::mem_attrs): New function.
2300         (set_mem_attributes_minus_bitpos): Change bitpos from a
2301         HOST_WIDE_INT to poly_int64.
2302         (set_mem_alias_set, set_mem_addr_space, set_mem_align, set_mem_expr)
2303         (clear_mem_offset, clear_mem_size, change_address)
2304         (get_spill_slot_decl, set_mem_attrs_for_spill): Directly
2305         initialize mem_attrs.
2306         (set_mem_offset, set_mem_size, adjust_address_1)
2307         (adjust_automodify_address_1, offset_address, widen_memory_access):
2308         Likewise.  Take poly_int64s rather than HOST_WIDE_INT.
2310 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2311             Alan Hayward  <alan.hayward@arm.com>
2312             David Sherwood  <david.sherwood@arm.com>
2314         * rtlanal.c (rtx_addr_can_trap_p_1): Take the offset and size
2315         as poly_int64s rather than HOST_WIDE_INTs.  Use a size of -1
2316         rather than 0 to represent an unknown size.  Assert that the size
2317         is known when the mode isn't BLKmode.
2318         (may_trap_p_1): Use -1 for unknown sizes.
2319         (rtx_addr_can_trap_p): Likewise.  Pass BLKmode rather than VOIDmode.
2321 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2322             Alan Hayward  <alan.hayward@arm.com>
2323             David Sherwood  <david.sherwood@arm.com>
2325         * dse.c (store_info): Change offset and width from HOST_WIDE_INT
2326         to poly_int64.  Update commentary for positions_needed.large.
2327         (read_info_type): Change offset and width from HOST_WIDE_INT
2328         to poly_int64.
2329         (set_usage_bits): Likewise.
2330         (canon_address): Return the offset as a poly_int64 rather than
2331         a HOST_WIDE_INT.  Use strip_offset_and_add.
2332         (set_all_positions_unneeded, any_positions_needed_p): Use
2333         positions_needed.large to track stores with non-constant widths.
2334         (all_positions_needed_p): Likewise.  Take the offset and width
2335         as poly_int64s rather than ints.  Assert that rhs is nonnull.
2336         (record_store): Cope with non-constant offsets and widths.
2337         Nullify the rhs of an earlier store if we can't tell which bytes
2338         of it are needed.
2339         (find_shift_sequence): Take the access_size and shift as poly_int64s
2340         rather than ints.
2341         (get_stored_val): Take the read_offset and read_width as poly_int64s
2342         rather than HOST_WIDE_INTs.
2343         (check_mem_read_rtx, scan_stores, scan_reads, dse_step5): Handle
2344         non-constant offsets and widths.
2346 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2347             Alan Hayward  <alan.hayward@arm.com>
2348             David Sherwood  <david.sherwood@arm.com>
2350         * inchash.h (inchash::hash::add_poly_int): New function.
2351         * tree-ssa-alias.h (ao_ref::offset, ao_ref::size, ao_ref::max_size):
2352         Use poly_int64 rather than HOST_WIDE_INT.
2353         (ao_ref::max_size_known_p): New function.
2354         * tree-ssa-sccvn.h (vn_reference_op_struct::off): Use poly_int64_pod
2355         rather than HOST_WIDE_INT.
2356         * tree-ssa-alias.c (ao_ref_base): Apply get_ref_base_and_extent
2357         to temporaries until its interface is adjusted to match.
2358         (ao_ref_init_from_ptr_and_size): Handle polynomial offsets and sizes.
2359         (aliasing_component_refs_p, decl_refs_may_alias_p)
2360         (indirect_ref_may_alias_decl_p, indirect_refs_may_alias_p): Take
2361         the offsets and max_sizes as poly_int64s instead of HOST_WIDE_INTs.
2362         (refs_may_alias_p_1, stmt_kills_ref_p): Adjust for changes to
2363         ao_ref fields.
2364         * alias.c (ao_ref_from_mem): Likewise.
2365         * tree-ssa-dce.c (mark_aliased_reaching_defs_necessary_1): Likewise.
2366         * tree-ssa-dse.c (valid_ao_ref_for_dse, normalize_ref)
2367         (clear_bytes_written_by, setup_live_bytes_from_ref, compute_trims)
2368         (maybe_trim_complex_store, maybe_trim_constructor_store)
2369         (live_bytes_read, dse_classify_store): Likewise.
2370         * tree-ssa-sccvn.c (vn_reference_compute_hash, vn_reference_eq):
2371         (copy_reference_ops_from_ref, ao_ref_init_from_vn_reference)
2372         (fully_constant_vn_reference_p, valueize_refs_1): Likewise.
2373         (vn_reference_lookup_3): Likewise.
2374         * tree-ssa-uninit.c (warn_uninitialized_vars): Likewise.
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         * tree-ssa-alias.c (indirect_ref_may_alias_decl_p)
2381         (indirect_refs_may_alias_p): Use ranges_may_overlap_p
2382         instead of ranges_overlap_p.
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         * tree-ssa-alias.c (same_addr_size_stores_p): Take the offsets and
2389         sizes as poly_int64s rather than HOST_WIDE_INTs.
2391 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2392             Alan Hayward  <alan.hayward@arm.com>
2393             David Sherwood  <david.sherwood@arm.com>
2395         * gimple-fold.h (fold_ctor_reference): Take the offset and size
2396         as poly_uint64 rather than unsigned HOST_WIDE_INT.
2397         * gimple-fold.c (fold_ctor_reference): Likewise.
2399 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2400             Alan Hayward  <alan.hayward@arm.com>
2401             David Sherwood  <david.sherwood@arm.com>
2403         * target.def (dwarf_poly_indeterminate_value): New hook.
2404         * targhooks.h (default_dwarf_poly_indeterminate_value): Declare.
2405         * targhooks.c (default_dwarf_poly_indeterminate_value): New function.
2406         * doc/tm.texi.in (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Document.
2407         * doc/tm.texi: Regenerate.
2408         * dwarf2out.h (build_cfa_loc, build_cfa_aligned_loc): Take the
2409         offset as a poly_int64.
2410         * dwarf2out.c (new_reg_loc_descr): Move later in file.  Take the
2411         offset as a poly_int64.
2412         (loc_descr_plus_const, loc_list_plus_const, build_cfa_aligned_loc):
2413         Take the offset as a poly_int64.
2414         (build_cfa_loc): Likewise.  Use loc_descr_plus_const.
2415         (frame_pointer_fb_offset): Change to a poly_int64.
2416         (int_loc_descriptor): Take the offset as a poly_int64.  Use
2417         targetm.dwarf_poly_indeterminate_value for polynomial offsets.
2418         (based_loc_descr): Take the offset as a poly_int64.
2419         Use strip_offset_and_add to handle (plus X (const)).
2420         Use new_reg_loc_descr instead of an open-coded version of the
2421         previous implementation.
2422         (mem_loc_descriptor): Handle CONST_POLY_INT.
2423         (compute_frame_pointer_to_fb_displacement): Take the offset as a
2424         poly_int64.  Use strip_offset_and_add to handle (plus X (const)).
2426 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2427             Alan Hayward  <alan.hayward@arm.com>
2428             David Sherwood  <david.sherwood@arm.com>
2430         * rtl.h (reg_attrs::offset): Change from HOST_WIDE_INT to poly_int64.
2431         (gen_rtx_REG_offset): Take the offset as a poly_int64.
2432         * inchash.h (inchash::hash::add_poly_hwi): New function.
2433         * gengtype.c (main): Register poly_int64.
2434         * emit-rtl.c (reg_attr_hasher::hash): Use inchash.  Treat the
2435         offset as a poly_int.
2436         (reg_attr_hasher::equal): Use must_eq to compare offsets.
2437         (get_reg_attrs, update_reg_offset, gen_rtx_REG_offset): Take the
2438         offset as a poly_int64.
2439         (set_reg_attrs_from_value): Treat the offset as a poly_int64.
2440         * print-rtl.c (print_poly_int): New function.
2441         (rtx_writer::print_rtx_operand_code_r): Treat REG_OFFSET as
2442         a poly_int.
2443         * var-tracking.c (track_offset_p, get_tracked_reg_offset): New
2444         functions.
2445         (var_reg_set, var_reg_delete_and_set, var_reg_delete): Use them.
2446         (same_variable_part_p, track_loc_p): Take the offset as a poly_int64.
2447         (vt_get_decl_and_offset): Return the offset as a poly_int64.
2448         Enforce track_offset_p for parts of a PARALLEL.
2449         (vt_add_function_parameter): Use const_offset for the final
2450         offset to track.  Use get_tracked_reg_offset for the parts
2451         of a PARALLEL.
2453 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2454             Alan Hayward  <alan.hayward@arm.com>
2455             David Sherwood  <david.sherwood@arm.com>
2457         * target.def (truly_noop_truncation): Take poly_uint64s instead of
2458         unsigned ints.  Change default to hook_bool_puint64_puint64_true.
2459         * doc/tm.texi: Regenerate.
2460         * hooks.h (hook_bool_uint_uint_true): Delete.
2461         (hook_bool_puint64_puint64_true): Declare.
2462         * hooks.c (hook_bool_uint_uint_true): Delete.
2463         (hook_bool_puint64_puint64_true): New function.
2464         * config/mips/mips.c (mips_truly_noop_truncation): Take poly_uint64s
2465         instead of unsigned ints.
2466         * config/spu/spu.c (spu_truly_noop_truncation): Likewise.
2467         * config/tilegx/tilegx.c (tilegx_truly_noop_truncation): Likewise.
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         * optabs.h (expand_operand): Add an int_value field.
2474         (create_expand_operand): Add an int_value parameter and use it
2475         to initialize the new expand_operand field.
2476         (create_integer_operand): Replace with a declaration of a function
2477         that accepts poly_int64s.  Move the implementation to...
2478         * optabs.c (create_integer_operand): ...here.
2479         (maybe_legitimize_operand): For EXPAND_INTEGER, check whether
2480         the mode preserves the value of int_value, instead of calling
2481         const_int_operand on the rtx.  Use gen_int_mode to generate
2482         the new rtx.
2484 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2485             Alan Hayward  <alan.hayward@arm.com>
2486             David Sherwood  <david.sherwood@arm.com>
2488         * dumpfile.h (dump_dec): Declare.
2489         * dumpfile.c (dump_dec): New function.
2490         * pretty-print.h (pp_wide_integer): Turn into a function and
2491         declare a poly_int version.
2492         * pretty-print.c (pp_wide_integer): New function for poly_ints.
2494 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2495             Alan Hayward  <alan.hayward@arm.com>
2496             David Sherwood  <david.sherwood@arm.com>
2498         * doc/generic.texi (POLY_INT_CST): Document.
2499         * tree.def (POLY_INT_CST): New tree code.
2500         * treestruct.def (TS_POLY_INT_CST): New tree layout.
2501         * tree-core.h (tree_poly_int_cst): New struct.
2502         (tree_node): Add a poly_int_cst field.
2503         * tree.h (POLY_INT_CST_P, POLY_INT_CST_COEFF): New macros.
2504         (wide_int_to_tree, force_fit_type): Take a poly_wide_int_ref
2505         instead of a wide_int_ref.
2506         (build_int_cst, build_int_cst_type): Take a poly_int64 instead
2507         of a HOST_WIDE_INT.
2508         (build_int_cstu, build_array_type_nelts): Take a poly_uint64
2509         instead of an unsigned HOST_WIDE_INT.
2510         (build_poly_int_cst, tree_fits_poly_int64_p, tree_fits_poly_uint64_p)
2511         (ptrdiff_tree_p): Declare.
2512         (tree_to_poly_int64, tree_to_poly_uint64): Likewise.  Provide
2513         extern inline implementations if the target doesn't use POLY_INT_CST.
2514         (poly_int_tree_p): New function.
2515         (wi::unextended_tree): New class.
2516         (wi::int_traits <unextended_tree>): New override.
2517         (wi::extended_tree): Add a default constructor.
2518         (wi::extended_tree::get_tree): New function.
2519         (wi::widest_extended_tree, wi::offset_extended_tree): New typedefs.
2520         (wi::tree_to_widest_ref, wi::tree_to_offset_ref): Use them.
2521         (wi::tree_to_poly_widest_ref, wi::tree_to_poly_offset_ref)
2522         (wi::tree_to_poly_wide_ref): New typedefs.
2523         (wi::ints_for): Provide overloads for extended_tree and
2524         unextended_tree.
2525         (poly_int_cst_value, wi::to_poly_widest, wi::to_poly_offset)
2526         (wi::to_wide): New functions.
2527         (wi::fits_to_boolean_p, wi::fits_to_tree_p): Handle poly_ints.
2528         * tree.c (poly_int_cst_hasher): New struct.
2529         (poly_int_cst_hash_table): New variable.
2530         (tree_node_structure_for_code, tree_code_size, simple_cst_equal)
2531         (valid_constant_size_p, add_expr, drop_tree_overflow): Handle
2532         POLY_INT_CST.
2533         (initialize_tree_contains_struct): Handle TS_POLY_INT_CST.
2534         (init_ttree): Initialize poly_int_cst_hash_table.
2535         (build_int_cst, build_int_cst_type, build_invariant_address): Take
2536         a poly_int64 instead of a HOST_WIDE_INT.
2537         (build_int_cstu, build_array_type_nelts): Take a poly_uint64
2538         instead of an unsigned HOST_WIDE_INT.
2539         (wide_int_to_tree): Rename to...
2540         (wide_int_to_tree_1): ...this.
2541         (build_new_poly_int_cst, build_poly_int_cst): New functions.
2542         (force_fit_type): Take a poly_wide_int_ref instead of a wide_int_ref.
2543         (wide_int_to_tree): New function that takes a poly_wide_int_ref.
2544         (ptrdiff_tree_p, tree_to_poly_int64, tree_to_poly_uint64)
2545         (tree_fits_poly_int64_p, tree_fits_poly_uint64_p): New functions.
2546         * lto-streamer-out.c (DFS::DFS_write_tree_body, hash_tree): Handle
2547         TS_POLY_INT_CST.
2548         * tree-streamer-in.c (lto_input_ts_poly_tree_pointers): Likewise.
2549         (streamer_read_tree_body): Likewise.
2550         * tree-streamer-out.c (write_ts_poly_tree_pointers): Likewise.
2551         (streamer_write_tree_body): Likewise.
2552         * tree-streamer.c (streamer_check_handled_ts_structures): Likewise.
2553         * asan.c (asan_protect_global): Require the size to be an INTEGER_CST.
2554         * cfgexpand.c (expand_debug_expr): Handle POLY_INT_CST.
2555         * expr.c (expand_expr_real_1, const_vector_from_tree): Likewise.
2556         * gimple-expr.h (is_gimple_constant): Likewise.
2557         * gimplify.c (maybe_with_size_expr): Likewise.
2558         * print-tree.c (print_node): Likewise.
2559         * tree-data-ref.c (data_ref_compare_tree): Likewise.
2560         * tree-pretty-print.c (dump_generic_node): Likewise.
2561         * tree-ssa-address.c (addr_for_mem_ref): Likewise.
2562         * tree-vect-data-refs.c (dr_group_sort_cmp): Likewise.
2563         * tree-vrp.c (compare_values_warnv): Likewise.
2564         * tree-ssa-loop-ivopts.c (determine_base_object, constant_multiple_of)
2565         (get_loop_invariant_expr, add_candidate_1, get_computation_aff_1)
2566         (force_expr_to_var_cost): Likewise.
2567         * tree-ssa-loop.c (for_each_index): Likewise.
2568         * fold-const.h (build_invariant_address, size_int_kind): Take a
2569         poly_int64 instead of a HOST_WIDE_INT.
2570         * fold-const.c (fold_negate_expr_1, const_binop, const_unop)
2571         (fold_convert_const, multiple_of_p, fold_negate_const): Handle
2572         POLY_INT_CST.
2573         (size_binop_loc): Likewise.  Allow int_const_binop_1 to fail.
2574         (int_const_binop_2): New function, split out from...
2575         (int_const_binop_1): ...here.  Handle POLY_INT_CST.
2576         (size_int_kind): Take a poly_int64 instead of a HOST_WIDE_INT.
2577         * expmed.c (make_tree): Handle CONST_POLY_INT_P.
2578         * gimple-ssa-strength-reduction.c (slsr_process_add)
2579         (slsr_process_mul): Check for INTEGER_CSTs before using them
2580         as candidates.
2581         * stor-layout.c (bits_from_bytes): New function.
2582         (bit_from_pos): Use it.
2583         (layout_type): Likewise.  For vectors, multiply the TYPE_SIZE_UNIT
2584         by BITS_PER_UNIT to get the TYPE_SIZE.
2585         * tree-cfg.c (verify_expr, verify_types_in_gimple_reference): Allow
2586         MEM_REF and TARGET_MEM_REF offsets to be a POLY_INT_CST.
2588 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2589             Alan Hayward  <alan.hayward@arm.com>
2590             David Sherwood  <david.sherwood@arm.com>
2592         * doc/rtl.texi (const_poly_int): Document.  Also document the
2593         rtl sharing behavior.
2594         * gengenrtl.c (excluded_rtx): Return true for CONST_POLY_INT.
2595         * rtl.h (const_poly_int_def): New struct.
2596         (rtx_def::u): Add a cpi field.
2597         (CASE_CONST_UNIQUE, CASE_CONST_ANY): Add CONST_POLY_INT.
2598         (CONST_POLY_INT_P, CONST_POLY_INT_COEFFS): New macros.
2599         (wi::rtx_to_poly_wide_ref): New typedef
2600         (const_poly_int_value, wi::to_poly_wide, rtx_to_poly_int64)
2601         (poly_int_rtx_p): New functions.
2602         (trunc_int_for_mode): Declare a poly_int64 version.
2603         (plus_constant): Take a poly_int64 instead of a HOST_WIDE_INT.
2604         (immed_wide_int_const): Take a poly_wide_int_ref rather than
2605         a wide_int_ref.
2606         (strip_offset): Declare.
2607         (strip_offset_and_add): New function.
2608         * rtl.def (CONST_POLY_INT): New rtx code.
2609         * rtl.c (rtx_size): Handle CONST_POLY_INT.
2610         (shared_const_p): Use poly_int_rtx_p.
2611         * emit-rtl.h (gen_int_mode): Take a poly_int64 instead of a
2612         HOST_WIDE_INT.
2613         (gen_int_shift_amount): Likewise.
2614         * emit-rtl.c (const_poly_int_hasher): New class.
2615         (const_poly_int_htab): New variable.
2616         (init_emit_once): Initialize it when NUM_POLY_INT_COEFFS > 1.
2617         (const_poly_int_hasher::hash): New function.
2618         (const_poly_int_hasher::equal): Likewise.
2619         (gen_int_mode): Take a poly_int64 instead of a HOST_WIDE_INT.
2620         (immed_wide_int_const): Rename to...
2621         (immed_wide_int_const_1): ...this and make static.
2622         (immed_wide_int_const): New function, taking a poly_wide_int_ref
2623         instead of a wide_int_ref.
2624         (gen_int_shift_amount): Take a poly_int64 instead of a HOST_WIDE_INT.
2625         (gen_lowpart_common): Handle CONST_POLY_INT.
2626         * cse.c (hash_rtx_cb, equiv_constant): Likewise.
2627         * cselib.c (cselib_hash_rtx): Likewise.
2628         * dwarf2out.c (const_ok_for_output_1): Likewise.
2629         * expr.c (convert_modes): Likewise.
2630         * print-rtl.c (rtx_writer::print_rtx, print_value): Likewise.
2631         * rtlhash.c (add_rtx): Likewise.
2632         * explow.c (trunc_int_for_mode): Add a poly_int64 version.
2633         (plus_constant): Take a poly_int64 instead of a HOST_WIDE_INT.
2634         Handle existing CONST_POLY_INT rtxes.
2635         * expmed.h (expand_shift): Take a poly_int64 instead of a
2636         HOST_WIDE_INT.
2637         * expmed.c (expand_shift): Likewise.
2638         * rtlanal.c (strip_offset): New function.
2639         (commutative_operand_precedence): Give CONST_POLY_INT the same
2640         precedence as CONST_DOUBLE and put CONST_WIDE_INT between that
2641         and CONST_INT.
2642         * rtl-tests.c (const_poly_int_tests): New struct.
2643         (rtl_tests_c_tests): Use it.
2644         * simplify-rtx.c (simplify_const_unary_operation): Handle
2645         CONST_POLY_INT.
2646         (simplify_const_binary_operation): Likewise.
2647         (simplify_binary_operation_1): Fold additions of symbolic constants
2648         and CONST_POLY_INTs.
2649         (simplify_subreg): Handle extensions and truncations of
2650         CONST_POLY_INTs.
2651         (simplify_const_poly_int_tests): New struct.
2652         (simplify_rtx_c_tests): Use it.
2653         * wide-int.h (storage_ref): Add default constructor.
2654         (wide_int_ref_storage): Likewise.
2655         (trailing_wide_ints): Use GTY((user)).
2656         (trailing_wide_ints::operator[]): Add a const version.
2657         (trailing_wide_ints::get_precision): New function.
2658         (trailing_wide_ints::extra_size): Likewise.
2660 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2661             Alan Hayward  <alan.hayward@arm.com>
2662             David Sherwood  <david.sherwood@arm.com>
2664         * emit-rtl.h (gen_int_shift_amount): Declare.
2665         * emit-rtl.c (gen_int_shift_amount): New function.
2666         * asan.c (asan_emit_stack_protection): Use gen_int_shift_amount
2667         instead of GEN_INT.
2668         * calls.c (shift_return_value): Likewise.
2669         * cse.c (fold_rtx): Likewise.
2670         * dse.c (find_shift_sequence): Likewise.
2671         * expmed.c (init_expmed_one_mode, store_bit_field_1, expand_shift_1)
2672         (expand_shift, expand_smod_pow2): Likewise.
2673         * lower-subreg.c (shift_cost): Likewise.
2674         * optabs.c (expand_superword_shift, expand_doubleword_mult)
2675         (expand_unop, expand_binop, shift_amt_for_vec_perm_mask)
2676         (expand_vec_perm_var): Likewise.
2677         * simplify-rtx.c (simplify_unary_operation_1): Likewise.
2678         (simplify_binary_operation_1): Likewise.
2679         * combine.c (try_combine, find_split_point, force_int_to_mode)
2680         (simplify_shift_const_1, simplify_shift_const): Likewise.
2681         (change_zero_ext): Likewise.  Use simplify_gen_binary.
2683 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2685         * poly-int.h (multiple_p): Fix handling of two non-poly_ints.
2687 2017-12-20  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
2689         * doc/invoke.texi (ARM Options): Document accepted extension options
2690         for -march=armv8.3-a.
2692 2017-12-20  Richard Earnshaw  <rearnsha@arm.com>
2694         PR target/83105
2695         * config.gcc (arm*-*-linux*): When configured with --with-float=hard
2696         or --with-float=softfp, set the default CPU to arm10e.
2698 2017-12-20  Eric Botcazou  <ebotcazou@adacore.com>
2700         * config/visium/constraints.md (J, K, L): Use IN_RANGE macro.
2701         * config/visium/predicates.md (const_shift_operand): Likewise.
2702         * config/visium/visium.c (visium_legitimize_address): Fix oversight.
2703         (visium_legitimize_reload_address): Likewise.
2705 2017-12-20  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
2707         PR target/82975
2708         * config/arm/arm.h (TEST_REGNO): Adjust comment as expected in r255830.
2710 2017-12-20  Jakub Jelinek  <jakub@redhat.com>
2712         PR c++/83490
2713         * calls.c (compute_argument_addresses): Ignore TYPE_EMPTY_P arguments.
2715 2017-12-20  Julia Koval  <julia.koval@intel.com>
2717         * common/config/i386/i386-common.c (OPTION_MASK_ISA_VPCLMULQDQ_SET,
2718         OPTION_MASK_ISA_VPCLMULQDQ_UNSET): New.
2719         (ix86_handle_option): Handle -mvpclmulqdq, move cx6 to flags2.
2720         * config.gcc: Include vpclmulqdqintrin.h.
2721         * config/i386/cpuid.h: Handle bit_VPCLMULQDQ.
2722         * config/i386/driver-i386.c (host_detect_local_cpu): Handle
2723         -mvpclmulqdq.
2724         * config/i386/i386-builtin.def (__builtin_ia32_vpclmulqdq_v2di,
2725         __builtin_ia32_vpclmulqdq_v4di, __builtin_ia32_vpclmulqdq_v8di): New.
2726         * config/i386/i386-c.c (__VPCLMULQDQ__): New.
2727         * config/i386/i386.c (isa2_opts): Add -mcx16.
2728         (isa_opts): Add -mpclmulqdq, remove -mcx16.
2729         (ix86_option_override_internal): Move mcx16 to flags2.
2730         (ix86_valid_target_attribute_inner_p): Add vpclmulqdq.
2731         (ix86_expand_builtin): Handle OPTION_MASK_ISA_VPCLMULQDQ.
2732         * config/i386/i386.h (TARGET_VPCLMULQDQ, TARGET_VPCLMULQDQ_P): New.
2733         * config/i386/i386.opt: Add mvpclmulqdq, move mcx16 to flags2.
2734         * config/i386/immintrin.h: Include vpclmulqdqintrin.h.
2735         * config/i386/sse.md (vpclmulqdq_<mode>): New pattern.
2736         * config/i386/vpclmulqdqintrin.h (_mm512_clmulepi64_epi128,
2737         _mm_clmulepi64_epi128, _mm256_clmulepi64_epi128): New intrinsics.
2738         * doc/invoke.texi: Add -mvpclmulqdq.
2740 2017-12-20  Tom de Vries  <tom@codesourcery.com>
2742         PR middle-end/83423
2743         * config/i386/i386.c (ix86_static_chain): Move
2744         DECL_STATIC_CHAIN test ...
2745         * calls.c (rtx_for_static_chain): ... here.  New function.
2746         * calls.h (rtx_for_static_chain): Declare.
2747         * builtins.c (expand_builtin_setjmp_receiver): Use rtx_for_static_chain
2748         instead of targetm.calls.static_chain.
2749         * df-scan.c (df_get_entry_block_def_set): Same.
2751 2017-12-19  Tom de Vries  <tom@codesourcery.com>
2753         PR tree-optimization/83493
2754         * graphite-isl-ast-to-gimple.c (translate_isl_ast_node_for): Unshare ub
2755         and lb.
2757 2017-12-19  Jakub Jelinek  <jakub@redhat.com>
2759         * gimple-ssa-sprintf.c (format_directive): Use inform_n instead of
2760         inform with hardcoded english plural handling.
2762 2017-12-18  Jeff Law  <law@redhat.com>
2764         PR tree-optimization/83477
2765         * tree-ssa-threadedge.c (record_temporary_equivalences_from_phis): For
2766         a non-virtual PHI, always push a new range.
2768 2017-12-19  Martin Sebor  <msebor@redhat.com>
2770         PR middle-end/77608
2771         * builtins.c (compute_objsize): Handle non-constant offsets.
2773 2017-12-19  Jakub Jelinek  <jakub@redhat.com>
2775         PR tree-optimization/83444
2776         * tree-ssa-strlen.c (strlen_check_and_optimize_stmt): For the
2777         character load case, if get_stridx on MEM_REF's operand doesn't
2778         look usable, retry with get_addr_stridx.
2780 2017-12-19  Alexandre Oliva  <aoliva@redhat.com>
2782         PR debug/83422
2783         * var-tracking.c (vt_debug_insns_local): Do not drop markers.
2784         (variable_tracking_main_1): Keep markers even when VTA fails.
2786         PR bootstrap/83396
2787         * cfgexpand.c (expand_gimple_basic_block): Expand label first,
2788         even if there are markers before it.
2789         * cfgrtl.c (rtl_verify_bb_layout): Reject DEBUG_INSNs outside BBs.
2791 2017-12-19  Jakub Jelinek  <jakub@redhat.com>
2793         * read-rtl.c (parse_reg_note_name): Replace Yoda conditions with
2794         typical order conditions.
2795         * sel-sched.c (extract_new_fences_from): Likewise.
2796         * config/visium/constraints.md (J, K, L): Likewise.
2797         * config/visium/predicates.md (const_shift_operand): Likewise.
2798         * config/visium/visium.c (visium_legitimize_address,
2799         visium_legitimize_reload_address): Likewise.
2800         * config/m68k/m68k.c (output_reg_adjust, emit_reg_adjust): Likewise.
2801         * config/arm/arm.c (arm_block_move_unaligned_straight): Likewise.
2802         * config/avr/constraints.md (Y01, Ym1, Y02, Ym2): Likewise.
2803         * config/avr/avr-log.c (avr_vdump, avr_log_set_avr_log,
2804         SET_DUMP_DETAIL): Likewise.
2805         * config/avr/predicates.md (const_8_16_24_operand): Likewise.
2806         * config/avr/avr.c (STR_PREFIX_P, avr_popcount_each_byte,
2807         avr_is_casesi_sequence, avr_casei_sequence_check_operands,
2808         avr_set_core_architecture, avr_set_current_function,
2809         avr_legitimize_reload_address, avr_asm_len, avr_print_operand,
2810         output_movqi, output_movsisf, avr_out_plus, avr_out_bitop,
2811         avr_out_fract, avr_adjust_insn_length, avr_encode_section_info,
2812         avr_2word_insn_p, output_reload_in_const, avr_has_nibble_0xf,
2813         avr_map_decompose, avr_fold_builtin): Likewise.
2814         * config/avr/driver-avr.c (avr_devicespecs_file): Likewise.
2815         * config/avr/gen-avr-mmcu-specs.c (str_prefix_p, print_mcu): Likewise.
2816         * config/i386/i386.c (ix86_parse_stringop_strategy_string): Likewise.
2817         * config/m32c/m32c-pragma.c (m32c_pragma_memregs): Likewise.
2818         * config/m32c/m32c.c (m32c_conditional_register_usage,
2819         m32c_address_cost): Likewise.
2820         * config/m32c/predicates.md (shiftcount_operand,
2821         longshiftcount_operand): Likewise.
2822         * config/iq2000/iq2000.c (iq2000_expand_prologue): Likewise.
2823         * config/nios2/nios2.c (nios2_handle_custom_fpu_insn_option,
2824         can_use_cdx_ldstw): Likewise.
2825         * config/nios2/nios2.h (CDX_REG_P): Likewise.
2826         * config/cr16/cr16.h (RETURN_ADDR_RTX, REGNO_MODE_OK_FOR_BASE_P):
2827         Likewise.
2828         * config/cr16/cr16.md (*mov<mode>_double): Likewise.
2829         * config/cr16/cr16.c (cr16_create_dwarf_for_multi_push): Likewise.
2830         * config/h8300/h8300.c (h8300_rtx_costs, get_shift_alg): Likewise.
2831         * config/vax/constraints.md (U06, U08, U16, CN6, S08, S16): Likewise.
2832         * config/vax/vax.c (adjacent_operands_p): Likewise.
2833         * config/ft32/constraints.md (L, b, KA): Likewise.
2834         * config/ft32/ft32.c (ft32_load_immediate, ft32_expand_prologue):
2835         Likewise.
2836         * cfgexpand.c (expand_stack_alignment): Likewise.
2837         * gcse.c (insert_expr_in_table): Likewise.
2838         * print-rtl.c (rtx_writer::print_rtx_operand_codes_E_and_V): Likewise.
2839         * cgraphunit.c (cgraph_node::expand): Likewise.
2840         * ira-build.c (setup_min_max_allocno_live_range_point): Likewise.
2841         * emit-rtl.c (add_insn): Likewise.
2842         * input.c (dump_location_info): Likewise.
2843         * passes.c (NEXT_PASS): Likewise.
2844         * read-rtl-function.c (parse_note_insn_name,
2845         function_reader::read_rtx_operand_r, function_reader::parse_mem_expr):
2846         Likewise.
2847         * sched-rgn.c (sched_rgn_init): Likewise.
2848         * diagnostic-show-locus.c (layout::show_ruler): Likewise.
2849         * combine.c (find_split_point, simplify_if_then_else, force_to_mode,
2850         if_then_else_cond, simplify_shift_const_1, simplify_comparison):
2851         Likewise.
2852         * explow.c (eliminate_constant_term): Likewise.
2853         * final.c (leaf_renumber_regs_insn): Likewise.
2854         * cfgrtl.c (print_rtl_with_bb): Likewise.
2855         * genhooks.c (emit_init_macros): Likewise.
2856         * poly-int.h (maybe_ne, maybe_le, maybe_lt): Likewise.
2857         * tree-data-ref.c (conflict_fn): Likewise.
2858         * selftest.c (assert_streq): Likewise.
2859         * expr.c (store_constructor_field, expand_expr_real_1): Likewise.
2860         * fold-const.c (fold_range_test, extract_muldiv_1, fold_truth_andor,
2861         fold_binary_loc, multiple_of_p): Likewise.
2862         * reload.c (push_reload, find_equiv_reg): Likewise.
2863         * et-forest.c (et_nca, et_below): Likewise.
2864         * dbxout.c (dbxout_symbol_location): Likewise.
2865         * reorg.c (relax_delay_slots): Likewise.
2866         * dojump.c (do_compare_rtx_and_jump): Likewise.
2867         * gengtype-parse.c (type): Likewise.
2868         * simplify-rtx.c (simplify_gen_ternary, simplify_gen_relational,
2869         simplify_const_relational_operation): Likewise.
2870         * reload1.c (do_output_reload): Likewise.
2871         * dumpfile.c (get_dump_file_info_by_switch): Likewise.
2872         * gengtype.c (type_for_name): Likewise.
2873         * gimple-ssa-sprintf.c (format_directive): Likewise.
2875 2017-12-19  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
2877         PR target/82975
2878         * config/arm/arm.h (TEST_REGNO): Check reg_renumber is set before
2879         accessing it.  Adjust comment.
2881 2017-12-19  Jakub Jelinek  <jakub@redhat.com>
2883         PR middle-end/81914
2884         * predict.c (zero_one_minusone): New function.
2885         (apply_return_prediction): Avoid return prediction for functions
2886         returning only -1, 0 and 1 values, unless they only return -1 and 0
2887         or 0 and 1.
2889 2017-12-19  Claudiu Zissulescu  <claziss@synopsys.com>
2891         * config/arc/arc.c (legitimate_scaled_address_p): Clean
2892         fall-through warning.
2893         (arc_compute_frame_size): Remove unused variables.
2894         (arc_print_operand): Fix fprintif format.
2895         (arc_can_follow_jump): Clean fall-through warning.
2897 2017-12-19  Marek Polacek  <polacek@redhat.com>
2899         PR c++/83489
2900         * config/i386/i386.c (init_cumulative_args): Don't check TYPE_EMPTY_P
2901         on an error node.
2903 2017-12-19  Claudiu Zissulescu  <claziss@synopsys.com>
2905         * config/arc/arc.c (overriderregs): New variable.
2906         (arc_override_options): Track fixed/call saved/call options.
2907         (arc_conditional_register_usage): Check against overrideregs
2908         variable whenever we change register properties.
2910 2017-12-19  Nathan Sidwell  <nathan@acm.org>
2912         * opts.c (finish_options): Don't prefix dump_base_name if it
2913         already contains directories.
2915 2017-12-19  Martin Liska  <mliska@suse.cz>
2917         PR rtl-optimization/82675
2918         * loop-unroll.c (unroll_loop_constant_iterations): Allocate one
2919         more element in sbitmap.
2921 2017-12-19  Paolo Carlini  <paolo.carlini@oracle.com>
2923         * gimplify.c (gimplify_expr): Use error_operand_p.
2925 2017-12-19  Sebastian Huber  <sebastian.huber@embedded-brains.de>
2927         PR target/83387
2928         * config/rs6000/t-rtems (MULTILIB_REQUIRED): Remove 64-bit soft-float
2929         multilib.
2931 2017-12-19  Daniel Cederman  <cederman@gaisler.com>
2933         * config/sparc/sparc.c (sparc_do_work_around_errata): Make sure
2934         the jump is to a label.
2936 2017-12-19  Jakub Jelinek  <jakub@redhat.com>
2938         PR tree-optimization/83444
2939         * tree-ssa-strlen.c (strlen_check_and_optimize_stmt): Optimize
2940         character loads.
2942         PR ipa/82801
2943         PR ipa/83346
2944         * ipa-inline.c (flatten_remove_node_hook): New function.
2945         (ipa_inline): Keep only nodes with flatten attribute at the end of
2946         the array in the order from ipa_reverse_postorder, only walk that
2947         portion of array for flattening, if there is more than one such
2948         node, temporarily register a removal hook and ignore removed nodes.
2950         PR tree-optimization/80631
2951         * tree-vect-loop.c (vect_create_epilog_for_reduction): Compare
2952         induc_code against MAX_EXPR or MIN_EXPR instead of reduc_fn against
2953         IFN_REDUC_MAX or IFN_REDUC_MIN.
2955 2017-12-18  Jeff Law  <law@redhat.com>
2957         * tree-ssa-dom.c (record_equivalences_from_phis): Do not
2958         record symbolic equivalences from backedges in the CFG.
2960         Revert
2961         2017-11-19  Jeff Law  <law@redhat.com>
2963         * tree-ssa-dom.c (record_equivalences_from_phis): Fix handling
2964         of degenerates resulting from ignoring an edge.
2966 2017-12-18  Martin Sebor  <msebor@redhat.com>
2968         PR middle-end/83373
2969         PR tree-optimization/78450
2970         * tree-ssa-strlen.c (maybe_set_strlen_range): New function.
2971         (handle_builtin_strlen): Call it.
2973 2017-12-18  Segher Boessenkool  <segher@kernel.crashing.org>
2975         PR rtl-optimization/83424
2976         * rtlanal.c (dead_or_set_regno_p): Handle CLOBBER just like SET.
2978 2017-12-18  Marek Polacek  <polacek@redhat.com>
2980         PR middle-end/83463
2981         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref):
2982         Check if TYPE is INTEGRAL_TYPE_P before accessing its min/max
2983         values.
2985 2017-12-18  Claudiu Zissulescu  <claziss@synopsys.com>
2987         * config/arc/arc.md (maddsidi4, maddsidi4_split): Update pattern.
2988         (umaddsidi4, umaddsidi_split): Likewise.
2990 2017-12-18  Claudiu Zissulescu  <claziss@synopsys.com>
2992         * config/arc/arc.c (arc_legitimate_constant_p): Always check all
2993         constants.
2995 2017-12-18  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
2997         PR target/83420
2998         * config/s390/s390.c (s390_option_override): Avoid strncpy.
3000 2017-12-18  Richard Biener  <rguenther@suse.de>
3002         PR tree-optimization/81877
3003         * tree-ssa-loop-im.c (ref_indep_loop_p): Remove safelen parameters.
3004         (outermost_indep_loop): Adjust.
3005         (ref_indep_loop_p_1): Likewise.  Remove safelen handling again.
3006         (can_sm_ref_p): Adjust.
3008 2017-12-18  Richard Biener  <rguenther@suse.de>
3010         PR middle-end/77291
3011         * tree.c (array_at_struct_end_p): Return true if the underlying
3012         object has space for at least one element in excess of what
3013         the array domain specifies.
3015 2017-12-17  Sandra Loosemore  <sandra@codesourcery.com>
3017         * doc/extend.texi (x86 Function Attributes): Reformat nocf_check
3018         example to avoid overfull hbox.
3019         * doc/invoke.texi (Option Summary): Add missing @gol.
3020         (C++ Dialect Options): Reformat -Wnoexcept-type example to avoid
3021         overfull hbox.
3023 2017-12-17  Sandra Loosemore  <sandra@codesourcery.com>
3024             Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
3026         * doc/invoke.texi (Option Summary): Add -mverbose-cost-dump
3027         to AArch64 and ARM lists, plus missing -mflip-thumb for ARM.
3028         (AArch64 Options): Document -mverbose-cost-dump.
3029         (ARM Options): Likewise, plus -mflip-thumb.
3031 2017-12-17  Martin Sebor  <msebor@redhat.com>
3033         PR bootstrap/83446
3034         * gimple-ssa-warn-restrict.c
3035         (builtin_memref::offset_out_of_bounds): Correct the handling of
3036         anti-ranges.
3038 2017-12-17  Sandra Loosemore  <sandra@codesourcery.com>
3040         * doc/invoke.texi (ARC Options): Add missing -mlra entry.
3042 2017-12-17  John David Anglin  <danglin@gcc.gnu.org>
3044         * config/pa/pa.c (pa_som_asm_init_sections): Fix comment.
3046 2017-12-17  Markus Trippelsdorf  <markus@trippelsdorf.de>
3048         * x86-tune-costs.h (skylake_cost, core_cost): Decrease r64 multiply
3049         latencies.
3051 2017-12-16  Sandra Loosemore  <sandra@codesourcery.com>
3053         * doc/invoke.texi: Fix some typos.
3055 2017-12-16  Martin Sebor  <msebor@redhat.com>
3057         PR tree-optimization/78918
3058         * Makefile.in (OBJS): Add gimple-ssa-warn-restrict.o.
3059         * builtins.c (check_sizes): Rename...
3060         (check_access): ...to this.  Rename function arguments for clarity.
3061         (check_memop_sizes): Adjust names.
3062         (expand_builtin_memchr, expand_builtin_memcpy): Same.
3063         (expand_builtin_memmove, expand_builtin_mempcpy): Same.
3064         (expand_builtin_strcat, expand_builtin_stpncpy): Same.
3065         (check_strncat_sizes, expand_builtin_strncat): Same.
3066         (expand_builtin_strncpy, expand_builtin_memset): Same.
3067         (expand_builtin_bzero, expand_builtin_memcmp): Same.
3068         (expand_builtin_memory_chk, maybe_emit_chk_warning): Same.
3069         (maybe_emit_sprintf_chk_warning): Same.
3070         (expand_builtin_strcpy): Adjust.
3071         (expand_builtin_stpcpy): Same.
3072         (expand_builtin_with_bounds): Detect out-of-bounds accesses
3073         in pointer-checking forms of memcpy, memmove, and mempcpy.
3074         (gcall_to_tree_minimal, max_object_size): Define new functions.
3075         * builtins.h (max_object_size): Declare.
3076         * calls.c (alloc_max_size): Call max_object_size instead of
3077         hardcoding ssizetype limit.
3078         (get_size_range): Handle new argument.
3079         * calls.h (get_size_range): Add a new argument.
3080         * cfgexpand.c (expand_call_stmt): Propagate no-warning bit.
3081         * doc/invoke.texi (-Wrestrict): Adjust, add example.
3082         * gimple-fold.c (gimple_fold_builtin_memory_op): Detect overlapping
3083         operations.
3084         (gimple_fold_builtin_memory_chk): Same.
3085         (gimple_fold_builtin_stxcpy_chk): New function.
3086         * gimple-ssa-warn-restrict.c: New source.
3087         * gimple-ssa-warn-restrict.h: New header.
3088         * gimple.c (gimple_build_call_from_tree): Propagate location.
3089         * passes.def (pass_warn_restrict): Add new pass.
3090         * tree-pass.h (make_pass_warn_restrict): Declare.
3091         * tree-ssa-strlen.c (handle_builtin_strcpy): Detect overlapping
3092         operations.
3093         (handle_builtin_strcat): Same.
3094         (strlen_optimize_stmt): Rename...
3095         (strlen_check_and_optimize_stmt): ...to this.  Handle strncat,
3096         stpncpy, strncpy, and their checking forms.
3098 2017-12-16  Jan Hubicka  <hubicka@ucw.cz>
3100         PR rtl-optimization/82849
3101         * modulo-sched.c (sms_schedule): Use get_estimated_loop_iterations_int
3102         and get_max_loop_iterations_int.
3104 2017-12-16  Richard Sandiford  <richard.sandiford@linaro.org>
3105             Alan Hayward  <alan.hayward@arm.com>
3106             David Sherwood  <david.sherwood@arm.com>
3108         * machmode.h (mode_for_size, int_mode_for_size, float_mode_for_size)
3109         (smallest_mode_for_size, smallest_int_mode_for_size): Take the mode
3110         size as a poly_uint64.
3111         (mode_for_vector, mode_for_int_vector): Take the number of vector
3112         elements as a poly_uint64.
3113         * stor-layout.c (mode_for_size, smallest_mode_for_size): Take the mode
3114         size as a poly_uint64.
3115         (mode_for_vector, mode_for_int_vector): Take the number of vector
3116         elements as a poly_uint64.
3118 2017-12-16  Richard Sandiford  <richard.sandiford@linaro.org>
3119             Alan Hayward  <alan.hayward@arm.com>
3120             David Sherwood  <david.sherwood@arm.com>
3122         * machmode.h (MACRO_MODE): New macro.
3123         * addresses.h (base_reg_class, ok_for_base_p_1): Use it.
3124         * targhooks.c (default_libcall_value, default_secondary_reload)
3125         (default_memory_move_cost, default_register_move_cost)
3126         (default_class_max_nregs): Likewise.
3128 2017-12-16  Richard Sandiford  <richard.sandiford@linaro.org>
3129             Alan Hayward  <alan.hayward@arm.com>
3130             David Sherwood  <david.sherwood@arm.com>
3132         * doc/sourcebuild.texi: Document IN_TARGET_CODE.
3133         * genattrtab.c (write_header): Define IN_TARGET_CODE to 1 in the
3134         target C file.
3135         * genautomata.c (main): Likewise.
3136         * genconditions.c (write_header): Likewise.
3137         * genemit.c (main): Likewise.
3138         * genextract.c (print_header): Likewise.
3139         * genopinit.c (main): Likewise.
3140         * genoutput.c (output_prologue): Likewise.
3141         * genpeep.c (main): Likewise.
3142         * genpreds.c (write_insn_preds_c): Likewise.
3143         * genrecog.c (writer_header): Likewise.
3144         * config/aarch64/aarch64-builtins.c (IN_TARGET_CODE): Define.
3145         * config/aarch64/aarch64-c.c (IN_TARGET_CODE): Likewise.
3146         * config/aarch64/aarch64.c (IN_TARGET_CODE): Likewise.
3147         * config/aarch64/cortex-a57-fma-steering.c (IN_TARGET_CODE): Likewise.
3148         * config/aarch64/driver-aarch64.c (IN_TARGET_CODE): Likewise.
3149         * config/alpha/alpha.c (IN_TARGET_CODE): Likewise.
3150         * config/alpha/driver-alpha.c (IN_TARGET_CODE): Likewise.
3151         * config/arc/arc-c.c (IN_TARGET_CODE): Likewise.
3152         * config/arc/arc.c (IN_TARGET_CODE): Likewise.
3153         * config/arc/driver-arc.c (IN_TARGET_CODE): Likewise.
3154         * config/arm/aarch-common.c (IN_TARGET_CODE): Likewise.
3155         * config/arm/arm-builtins.c (IN_TARGET_CODE): Likewise.
3156         * config/arm/arm-c.c (IN_TARGET_CODE): Likewise.
3157         * config/arm/arm.c (IN_TARGET_CODE): Likewise.
3158         * config/arm/driver-arm.c (IN_TARGET_CODE): Likewise.
3159         * config/avr/avr-c.c (IN_TARGET_CODE): Likewise.
3160         * config/avr/avr-devices.c (IN_TARGET_CODE): Likewise.
3161         * config/avr/avr-log.c (IN_TARGET_CODE): Likewise.
3162         * config/avr/avr.c (IN_TARGET_CODE): Likewise.
3163         * config/avr/driver-avr.c (IN_TARGET_CODE): Likewise.
3164         * config/avr/gen-avr-mmcu-specs.c (IN_TARGET_CODE): Likewise.
3165         * config/bfin/bfin.c (IN_TARGET_CODE): Likewise.
3166         * config/c6x/c6x.c (IN_TARGET_CODE): Likewise.
3167         * config/cr16/cr16.c (IN_TARGET_CODE): Likewise.
3168         * config/cris/cris.c (IN_TARGET_CODE): Likewise.
3169         * config/darwin.c (IN_TARGET_CODE): Likewise.
3170         * config/epiphany/epiphany.c (IN_TARGET_CODE): Likewise.
3171         * config/epiphany/mode-switch-use.c (IN_TARGET_CODE): Likewise.
3172         * config/epiphany/resolve-sw-modes.c (IN_TARGET_CODE): Likewise.
3173         * config/fr30/fr30.c (IN_TARGET_CODE): Likewise.
3174         * config/frv/frv.c (IN_TARGET_CODE): Likewise.
3175         * config/ft32/ft32.c (IN_TARGET_CODE): Likewise.
3176         * config/h8300/h8300.c (IN_TARGET_CODE): Likewise.
3177         * config/i386/djgpp.c (IN_TARGET_CODE): Likewise.
3178         * config/i386/driver-i386.c (IN_TARGET_CODE): Likewise.
3179         * config/i386/driver-mingw32.c (IN_TARGET_CODE): Likewise.
3180         * config/i386/host-cygwin.c (IN_TARGET_CODE): Likewise.
3181         * config/i386/host-i386-darwin.c (IN_TARGET_CODE): Likewise.
3182         * config/i386/host-mingw32.c (IN_TARGET_CODE): Likewise.
3183         * config/i386/i386-c.c (IN_TARGET_CODE): Likewise.
3184         * config/i386/i386.c (IN_TARGET_CODE): Likewise.
3185         * config/i386/intelmic-mkoffload.c (IN_TARGET_CODE): Likewise.
3186         * config/i386/msformat-c.c (IN_TARGET_CODE): Likewise.
3187         * config/i386/winnt-cxx.c (IN_TARGET_CODE): Likewise.
3188         * config/i386/winnt-stubs.c (IN_TARGET_CODE): Likewise.
3189         * config/i386/winnt.c (IN_TARGET_CODE): Likewise.
3190         * config/i386/x86-tune-sched-atom.c (IN_TARGET_CODE): Likewise.
3191         * config/i386/x86-tune-sched-bd.c (IN_TARGET_CODE): Likewise.
3192         * config/i386/x86-tune-sched-core.c (IN_TARGET_CODE): Likewise.
3193         * config/i386/x86-tune-sched.c (IN_TARGET_CODE): Likewise.
3194         * config/ia64/ia64-c.c (IN_TARGET_CODE): Likewise.
3195         * config/ia64/ia64.c (IN_TARGET_CODE): Likewise.
3196         * config/iq2000/iq2000.c (IN_TARGET_CODE): Likewise.
3197         * config/lm32/lm32.c (IN_TARGET_CODE): Likewise.
3198         * config/m32c/m32c-pragma.c (IN_TARGET_CODE): Likewise.
3199         * config/m32c/m32c.c (IN_TARGET_CODE): Likewise.
3200         * config/m32r/m32r.c (IN_TARGET_CODE): Likewise.
3201         * config/m68k/m68k.c (IN_TARGET_CODE): Likewise.
3202         * config/mcore/mcore.c (IN_TARGET_CODE): Likewise.
3203         * config/microblaze/microblaze-c.c (IN_TARGET_CODE): Likewise.
3204         * config/microblaze/microblaze.c (IN_TARGET_CODE): Likewise.
3205         * config/mips/driver-native.c (IN_TARGET_CODE): Likewise.
3206         * config/mips/frame-header-opt.c (IN_TARGET_CODE): Likewise.
3207         * config/mips/mips.c (IN_TARGET_CODE): Likewise.
3208         * config/mmix/mmix.c (IN_TARGET_CODE): Likewise.
3209         * config/mn10300/mn10300.c (IN_TARGET_CODE): Likewise.
3210         * config/moxie/moxie.c (IN_TARGET_CODE): Likewise.
3211         * config/msp430/driver-msp430.c (IN_TARGET_CODE): Likewise.
3212         * config/msp430/msp430-c.c (IN_TARGET_CODE): Likewise.
3213         * config/msp430/msp430.c (IN_TARGET_CODE): Likewise.
3214         * config/nds32/nds32-cost.c (IN_TARGET_CODE): Likewise.
3215         * config/nds32/nds32-fp-as-gp.c (IN_TARGET_CODE): Likewise.
3216         * config/nds32/nds32-intrinsic.c (IN_TARGET_CODE): Likewise.
3217         * config/nds32/nds32-isr.c (IN_TARGET_CODE): Likewise.
3218         * config/nds32/nds32-md-auxiliary.c (IN_TARGET_CODE): Likewise.
3219         * config/nds32/nds32-memory-manipulation.c (IN_TARGET_CODE): Likewise.
3220         * config/nds32/nds32-pipelines-auxiliary.c (IN_TARGET_CODE): Likewise.
3221         * config/nds32/nds32-predicates.c (IN_TARGET_CODE): Likewise.
3222         * config/nds32/nds32.c (IN_TARGET_CODE): Likewise.
3223         * config/nios2/nios2.c (IN_TARGET_CODE): Likewise.
3224         * config/nvptx/mkoffload.c (IN_TARGET_CODE): Likewise.
3225         * config/nvptx/nvptx.c (IN_TARGET_CODE): Likewise.
3226         * config/pa/pa.c (IN_TARGET_CODE): Likewise.
3227         * config/pdp11/pdp11.c (IN_TARGET_CODE): Likewise.
3228         * config/powerpcspe/driver-powerpcspe.c (IN_TARGET_CODE): Likewise.
3229         * config/powerpcspe/host-darwin.c (IN_TARGET_CODE): Likewise.
3230         * config/powerpcspe/host-ppc64-darwin.c (IN_TARGET_CODE): Likewise.
3231         * config/powerpcspe/powerpcspe-c.c (IN_TARGET_CODE): Likewise.
3232         * config/powerpcspe/powerpcspe-linux.c (IN_TARGET_CODE): Likewise.
3233         * config/powerpcspe/powerpcspe.c (IN_TARGET_CODE): Likewise.
3234         * config/riscv/riscv-builtins.c (IN_TARGET_CODE): Likewise.
3235         * config/riscv/riscv-c.c (IN_TARGET_CODE): Likewise.
3236         * config/riscv/riscv.c (IN_TARGET_CODE): Likewise.
3237         * config/rl78/rl78-c.c (IN_TARGET_CODE): Likewise.
3238         * config/rl78/rl78.c (IN_TARGET_CODE): Likewise.
3239         * config/rs6000/driver-rs6000.c (IN_TARGET_CODE): Likewise.
3240         * config/rs6000/host-darwin.c (IN_TARGET_CODE): Likewise.
3241         * config/rs6000/host-ppc64-darwin.c (IN_TARGET_CODE): Likewise.
3242         * config/rs6000/rs6000-c.c (IN_TARGET_CODE): Likewise.
3243         * config/rs6000/rs6000-linux.c (IN_TARGET_CODE): Likewise.
3244         * config/rs6000/rs6000-p8swap.c (IN_TARGET_CODE): Likewise.
3245         * config/rs6000/rs6000-string.c (IN_TARGET_CODE): Likewise.
3246         * config/rs6000/rs6000.c (IN_TARGET_CODE): Likewise.
3247         * config/rx/rx.c (IN_TARGET_CODE): Likewise.
3248         * config/s390/driver-native.c (IN_TARGET_CODE): Likewise.
3249         * config/s390/s390-c.c (IN_TARGET_CODE): Likewise.
3250         * config/s390/s390.c (IN_TARGET_CODE): Likewise.
3251         * config/sh/sh-c.c (IN_TARGET_CODE): Likewise.
3252         * config/sh/sh-mem.cc (IN_TARGET_CODE): Likewise.
3253         * config/sh/sh.c (IN_TARGET_CODE): Likewise.
3254         * config/sh/sh_optimize_sett_clrt.cc (IN_TARGET_CODE): Likewise.
3255         * config/sh/sh_treg_combine.cc (IN_TARGET_CODE): Likewise.
3256         * config/sparc/driver-sparc.c (IN_TARGET_CODE): Likewise.
3257         * config/sparc/sparc-c.c (IN_TARGET_CODE): Likewise.
3258         * config/sparc/sparc.c (IN_TARGET_CODE): Likewise.
3259         * config/spu/spu-c.c (IN_TARGET_CODE): Likewise.
3260         * config/spu/spu.c (IN_TARGET_CODE): Likewise.
3261         * config/stormy16/stormy16.c (IN_TARGET_CODE): Likewise.
3262         * config/tilegx/mul-tables.c (IN_TARGET_CODE): Likewise.
3263         * config/tilegx/tilegx-c.c (IN_TARGET_CODE): Likewise.
3264         * config/tilegx/tilegx.c (IN_TARGET_CODE): Likewise.
3265         * config/tilepro/mul-tables.c (IN_TARGET_CODE): Likewise.
3266         * config/tilepro/tilepro-c.c (IN_TARGET_CODE): Likewise.
3267         * config/tilepro/tilepro.c (IN_TARGET_CODE): Likewise.
3268         * config/v850/v850-c.c (IN_TARGET_CODE): Likewise.
3269         * config/v850/v850.c (IN_TARGET_CODE): Likewise.
3270         * config/vax/vax.c (IN_TARGET_CODE): Likewise.
3271         * config/visium/visium.c (IN_TARGET_CODE): Likewise.
3272         * config/vms/vms-c.c (IN_TARGET_CODE): Likewise.
3273         * config/vms/vms-f.c (IN_TARGET_CODE): Likewise.
3274         * config/vms/vms.c (IN_TARGET_CODE): Likewise.
3275         * config/xtensa/xtensa.c (IN_TARGET_CODE): Likewise.
3277 2017-12-16  Richard Sandiford  <richard.sandiford@linaro.org>
3278             Alan Hayward  <alan.hayward@arm.com>
3279             David Sherwood  <david.sherwood@arm.com>
3281         * tree-ssa-sccvn.c (vn_reference_lookup_3): Avoid repeated
3282         checks for MEM_REF.
3284 2017-12-16  Richard Sandiford  <richard.sandiford@linaro.org>
3285             Alan Hayward  <alan.hayward@arm.com>
3286             David Sherwood  <david.sherwood@arm.com>
3288         * doc/generic.texi (VEC_SERIES_EXPR): Document.
3289         * doc/md.texi (vec_series@var{m}): Document.
3290         * tree.def (VEC_SERIES_EXPR): New tree code.
3291         * tree.h (build_vec_series): Declare.
3292         * tree.c (build_vec_series): New function.
3293         * cfgexpand.c (expand_debug_expr): Handle VEC_SERIES_EXPR.
3294         * tree-pretty-print.c (dump_generic_node): Likewise.
3295         * gimple-pretty-print.c (dump_binary_rhs): Likewise.
3296         * tree-inline.c (estimate_operator_cost): Likewise.
3297         * expr.c (expand_expr_real_2): Likewise.
3298         * optabs-tree.c (optab_for_tree_code): Likewise.
3299         * tree-cfg.c (verify_gimple_assign_binary): Likewise.
3300         * fold-const.c (const_binop): Fold VEC_SERIES_EXPRs of constants.
3301         * expmed.c (make_tree): Handle VEC_SERIES.
3302         * optabs.def (vec_series_optab): New optab.
3303         * optabs.h (expand_vec_series_expr): Declare.
3304         * optabs.c (expand_vec_series_expr): New function.
3305         * tree-vect-generic.c (expand_vector_operations_1): Check that
3306         the operands also have vector type.
3308 2017-12-16  Richard Sandiford  <richard.sandiford@linaro.org>
3309             Alan Hayward  <alan.hawyard@arm.com>
3310             David Sherwood  <david.sherwood@arm.com>
3312         * doc/generic.texi (VEC_DUPLICATE_EXPR): Document.
3313         (VEC_COND_EXPR): Add missing @tindex.
3314         * doc/md.texi (vec_duplicate@var{m}): Document.
3315         * tree.def (VEC_DUPLICATE_EXPR): New tree codes.
3316         * tree.c (build_vector_from_val): Add stubbed-out handling of
3317         variable-length vectors, using VEC_DUPLICATE_EXPR.
3318         (uniform_vector_p): Handle VEC_DUPLICATE_EXPR.
3319         * cfgexpand.c (expand_debug_expr): Likewise.
3320         * tree-cfg.c (verify_gimple_assign_unary): Likewise.
3321         * tree-inline.c (estimate_operator_cost): Likewise.
3322         * tree-pretty-print.c (dump_generic_node): Likewise.
3323         * tree-vect-generic.c (ssa_uniform_vector_p): Likewise.
3324         * fold-const.c (const_unop): Fold VEC_DUPLICATE_EXPRs of a constant.
3325         (test_vec_duplicate_folding): New function.
3326         (fold_const_c_tests): Call it.
3327         * optabs.def (vec_duplicate_optab): New optab.
3328         * optabs-tree.c (optab_for_tree_code): Handle VEC_DUPLICATE_EXPR.
3329         * optabs.h (expand_vector_broadcast): Declare.
3330         * optabs.c (expand_vector_broadcast): Make non-static.  Try using
3331         vec_duplicate_optab.
3332         * expr.c (store_constructor): Try using vec_duplicate_optab for
3333         uniform vectors.
3334         (expand_expr_real_2): Handle VEC_DUPLICATE_EXPR.
3336 2017-12-15  Markus Trippelsdorf  <markus@trippelsdorf.de>
3338         PR target/83358
3339         * config/i386/x86-tune-costs.h (skylake_cost, core_cost): Increase
3340         div/mod latencies a bit.
3342 2017-12-15  Jeff Law  <law@redhat.com>
3344         PR tree-optimization/36550
3345         * tree-ssa-threadupdate.c (count_stmts_and_phis_in_block): New.
3346         (mark_threaded_blocks): Rewrite code to avoid block copying when
3347         optimizing for size.  Don't pessimize blocks which will be
3348         copied, but all the statements will be dead.
3350 2017-12-15  Alexandre Oliva  <aoliva@redhat.com>
3352         PR tree-optimization/81165
3353         * tree-ssa-threadupdate.c (uses_in_bb): New.
3354         (estimate_threading_killed_stmts): New.
3355         * tree-ssa-threadupdate.h (estimate_threading_killed_stmts): Prototype.
3356         * tree-ssa-threadedge.c
3357         (record_temporary_equivalences_from_stmts_at_dest): Expand limit
3358         when its hit.
3360 2017-12-15  Jeff Law  <law@redhat.com>
3362         PR tree-optimization/83410
3363         * tree-ssa-threadupdate.c (thread_block_1): Avoid certain jump
3364         threads when parallelizing loops.
3366 2017-12-15  Jakub Jelinek  <jakub@redhat.com>
3368         * tree-core.h (struct attribute_spec): Swap affects_type_identity and
3369         handler fields.
3370         * config/alpha/alpha.c (vms_attribute_table): Swap
3371         affects_type_identity and handler fields, adjust comments.
3372         * config/mips/mips.c (mips_attribute_table): Likewise.
3373         * config/visium/visium.c (visium_attribute_table): Likewise.
3374         * config/epiphany/epiphany.c (epiphany_attribute_table): Likewise.
3375         * config/microblaze/microblaze.c (microblaze_attribute_table):
3376         Likewise.
3377         * config/spu/spu.c (spu_attribute_table): Likewise.
3378         * config/mcore/mcore.c (mcore_attribute_table): Likewise.
3379         * config/arc/arc.c (arc_attribute_table): Likewise.
3380         * config/m68k/m68k.c (m68k_attribute_table): Likewise.
3381         * config/v850/v850.c (v850_handle_interrupt_attribute,
3382         v850_handle_data_area_attribute): Formatting fixes.
3383         (v850_attribute_table): Swap affects_type_identity and handler
3384         fields, adjust comments.
3385         * config/m32r/m32r.c (m32r_attribute_table): Likewise.
3386         * config/arm/arm.c (arm_attribute_table): Likewise.
3387         * config/avr/avr.c (avr_attribute_table): Likewise.
3388         * config/s390/s390.c (s390_attribute_table): Likewise.
3389         * config/sh/sh.c (sh_attribute_table): Likewise.
3390         * config/i386/i386.c (ix86_handle_cconv_attribute,
3391         ix86_handle_callee_pop_aggregate_return): Formatting fixes.
3392         (ix86_attribute_table): Swap affects_type_identity and handler
3393         fields, adjust comments.
3394         * config/i386/cygming.h (SUBTARGET_ATTRIBUTE_TABLE): Likewise.
3395         * config/sparc/sparc.c (sparc_attribute_table): Likewise.
3396         * config/m32c/m32c.c (m32c_attribute_table): Likewise.
3397         * config/sol2.h (SOLARIS_ATTRIBUTE_TABLE): Likewise.
3398         * config/ia64/ia64.c (ia64_attribute_table): Likewise.
3399         * config/msp430/msp430.c (msp430_attribute_table): Likewise.
3400         * config/rx/rx.c (rx_attribute_table): Likewise.
3401         * config/cr16/cr16.c (cr16_attribute_table): Likewise.
3402         * config/h8300/h8300.c (h8300_attribute_table): Likewise.
3403         * config/nvptx/nvptx.c (nvptx_attribute_table): Likewise.
3404         * config/powerpcspe/powerpcspe.c (rs6000_attribute_table): Likewise.
3405         * config/darwin.h (SUBTARGET_ATTRIBUTE_TABLE): Likewise.
3406         * config/stormy16/stormy16.c (xstormy16_attribute_table): Likewise.
3407         * config/bfin/bfin.c (bfin_attribute_table): Likewise.
3408         * config/rs6000/rs6000.c (rs6000_attribute_table): Likewise.
3409         * config/rl78/rl78.c (rl78_attribute_table): Likewise.
3410         * config/nds32/nds32.c (nds32_attribute_table): Likewise.
3411         * doc/plugins.texi (user_attr): Likewise.  Add NULL for
3412         exclude.
3413         * attribs.c (empty_attribute_table): Swap affects_type_identity and
3414         handler fields.
3415         (register_scoped_attributes, decl_attributes): Formatting fixes.
3417         PR tree-optimization/83269
3418         * fold-const.c (fold_binary_loc): Perform (-A) - B -> (-B) - A
3419         subtraction in arg0's type if type is signed and arg0 is unsigned.
3420         Formatting fix.
3422         PR sanitizer/81281
3423         * match.pd ((T)(P + A) - (T)P -> (T) A): Use @@0 instead of @0 and
3424         convert? on @0 instead of convert.  Check type of @1, not @0.
3425         ((T)P - (T)(P + A) -> -(T) A): Use @@0 instead of @0 and
3426         convert? on @0 instead of convert.  Check type of @1, not @0.
3427         ((T)(P + A) - (T)(P + B) -> (T)A - (T)B): Use @@0 instead of @0,
3428         only optimize if either both @1 and @2 types are narrower
3429         precision, or both are wider or equal precision, and in the former
3430         case only if both have undefined overflow.
3432 2017-12-15  Richard Biener  <rguenther@suse.de>
3434         PR lto/83388
3435         * internal-fn.def (IFN_NOP): Add.
3436         * internal-fn.c (expand_NOP): Do nothing.
3437         * lto-streamer-in.c (input_function): Instead of removing
3438         sanitizer calls replace them with IFN_NOP calls.
3440 2017-12-15  Richard Sandiford  <richard.sandiford@linaro.org>
3441             Alan Hayward  <alan.hayward@arm.com>
3442             David Sherwood  <david.sherwood@arm.com>
3444         * dse.c (store_info, read_info_type): Replace begin and end with
3445         offset and width.
3446         (print_range): New function.
3447         (set_all_positions_unneeded, any_positions_needed_p)
3448         (check_mem_read_rtx, scan_stores, scan_reads, dse_step5): Update
3449         accordingly.
3450         (record_store): Likewise.  Optimize the case in which all positions
3451         are unneeded.
3452         (get_stored_val): Replace read_begin and read_end with read_offset
3453         and read_width.
3454         (replace_read): Update call accordingly.
3456 2017-12-15  Bin Cheng  <bin.cheng@arm.com>
3458         * gimple-loop-interchange.cc (STMT_COST_RATIO): New macro.
3459         (loop_cand::m_num_stmts, loop_cand::m_const_init_reduc): New members.
3460         (loop_cand::loop_cand): Initialize above members.
3461         (loop_cand::supported_operations): Delete.
3462         (loop_cand::can_interchange_p): Inline above function.
3463         (loop_cand::classify_simple_reduction): Record number of constant
3464         initialized simple reductions.
3465         (should_interchange_loops): New parameters.  Check stmt cost of loops
3466         to be interchange.
3467         (tree_loop_interchange::interchange): Prepare stmt cost of outer loop.
3468         Update call to should_interchange_loops.
3469         (should_interchange_loop_nest): Update call to
3470         should_interchange_loops.
3472 2017-12-15  Eric Botcazou  <ebotcazou@adacore.com>
3474         PR target/66488
3475         * ggc-page.c (HOST_BITS_PER_PTR): Do not define here...
3476         * hwint.h (HOST_BITS_PER_PTR): ...but here instead.
3477         * config/i386/xm-mingw32.h (HOST_BITS_PER_PTR): Delete.
3479 2017-12-15  Julia Koval  <julia.koval@intel.com>
3481         * config/i386/i386-builtin.def (__builtin_ia32_vaesenclast_v16qi,
3482         __builtin_ia32_vaesenclast_v32qi, __builtin_ia32_vaesenclast_v64qi):
3483         New.
3484         * config/i386/sse.md (vaesenclast_<mode>): New pattern.
3485         * config/i386/vaesintrin.h (_mm256_aesenclast_epi128,
3486         _mm512_aesenclast_epi128, _mm_aesenclast_epi128): New intrinsics.
3488 2017-12-15  Julia Koval  <julia.koval@intel.com>
3490         * config/i386/i386-builtin.def (__builtin_ia32_vaesenc_v16qi,
3491         __builtin_ia32_vaesenc_v32qi, __builtin_ia32_vaesenc_v64qi): New.
3492         * config/i386/sse.md (vaesenc_<mode>): New pattern.
3493         * config/i386/vaesintrin.h (_mm256_aesenc_epi128, _mm512_aesenc_epi128,
3494         _mm_aesenc_epi128): New intrinsics.
3496 2017-12-15  Julia Koval  <julia.koval@intel.com>
3498         * config/i386/i386-builtin.def (__builtin_ia32_vaesdeclast_v16qi,
3499         __builtin_ia32_vaesdeclast_v32qi, __builtin_ia32_vaesdeclast_v64qi):
3500         New.
3501         * config/i386/sse.md (vaesdeclast_<mode>): New pattern.
3502         * config/i386/vaesintrin.h (_mm256_aesdeclast_epi128,
3503         _mm512_aesdeclast_epi128, _mm_aesdeclast_epi128): New intrinsics.
3505 2017-12-14  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
3507         * gimple-ssa-strength-reduction.c (analyze_increments):
3508         Distinguish replacement costs for constant strides from those for
3509         unknown strides.
3511 2017-12-14  Jakub Jelinek  <jakub@redhat.com>
3513         * var-tracking.c (variable_tracking_main_1): Formatting fix.
3515 2017-12-14  Bernd Edlinger  <bernd.edlinger@hotmail.de>
3517         * doc/invoke.texi: Document -Wcast-function-type.
3518         * recog.h (stored_funcptr): Change signature.
3519         * tree-dump.c (dump_node): Avoid warning.
3520         * typed-splay-tree.h (typed_splay_tree): Avoid warning.
3522 2017-12-14  Qing Zhao  <qing.zhao@oracle.com>
3524         PR middle_end/79538
3525         * gimple-fold.c (get_range_strlen): Add the handling of non-member 
3526         array.
3528 2017-12-14  David Malcolm  <dmalcolm@redhat.com>
3530         PR tree-optimization/83312
3531         * domwalk.h (dom_walker::dom_walker): Fix typo in comment.
3532         * tree-cfg.c (find_taken_edge): Update to handle NULL_TREE for
3533         "val" param, and to cope with arbitrary basic blocks.
3534         (find_taken_edge_cond_expr): Add "cond_stmt" param and use it to
3535         handle NULL_TREE for "val", dropping "bb" param.
3536         (find_taken_edge_switch_expr): Make "switch_stmt" param const and
3537         drop "bb" param.  Handle NULL_TREE for "val".
3538         (find_case_label_for_value): Make "switch_stmt" param const.
3539         * tree-vrp.c (class check_array_bounds_dom_walker): New subclass
3540         of dom_walker.
3541         (vrp_prop::check_all_array_refs): Reimplement as...
3542         (check_array_bounds_dom_walker::before_dom_children): ...this new
3543         vfunc.  Replace linear search through BB block list, excluding
3544         those with non-executable in-edges via dominator walk.
3546 2017-12-14  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
3548         * config/arm/arm.opt (mverbose-cost-dump): New option.
3549         * config/arm/arm.c (arm_rtx_costs): Use it.
3551 2017-12-14  Andreas Schwab  <schwab@linux-m68k.org>
3553         PR bootstrap/83396
3554         * reload1.c (emit_input_reload_insns): Skip debug markers.
3556 2017-12-14  Alexandre Oliva  <aoliva@redhat.com>
3558         * config/i386/i386.c (rest_of_insert_endbranch): Use call loc
3559         for its nop_endbr.
3561         PR bootstrap/83396
3562         * config/arc/arc.c (hwloop_optimize): Skip debug insns.
3563         * config/sh/sh-protos.h (sh_find_set_of_reg): Adjust.
3564         * config/sh/sh.c: Skip debug insns besides notes.
3565         * config/sh/sh.md: Likewise.
3566         * config/sh/sh_treg_combine.cc: Likewise.
3567         * config/sh/sync.md: Likewise.
3569 2017-12-14  Tom de Vries  <tom@codesourcery.com>
3571         * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
3572         Add item for weak_undefined.
3574 2017-12-14  Richard Biener  <rguenther@suse.de>
3576         PR tree-optimization/67842
3577         * sese.h (bb_in_region): Remove #if 0'ed code.
3579 2017-12-14  Richard Biener  <rguenther@suse.de>
3581         PR tree-optimization/83326
3582         * tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Add
3583         may_be_zero parameter and handle it by not marking the first
3584         peeled copy as not exiting the loop.
3585         (try_peel_loop): Likewise.
3586         (canonicalize_loop_induction_variables): Use number_of_iterations_exit
3587         to handle the case of constant or zero iterations and perform
3588         loop header copying on-the-fly.
3590 2017-12-14  Richard Biener  <rguenther@suse.de>
3592         PR tree-optimization/83418
3593         * vr-values.c (vr_values::extract_range_for_var_from_comparison_expr):
3594         Instead of asserting we don't get unfolded comparisons deal with them.
3596 2017-12-14  Jakub Jelinek  <jakub@redhat.com>
3598         PR bootstrap/83396
3599         * var-tracking.c (vt_initialize): Ignore non-DEBUG_INSNs outside of
3600         basic blocks.  Assert debug bind insns don't appear outside of bbs,
3601         don't reset them.  Assert insns without BLOCK_FOR_INSN are outside of
3602         bb.  Simplify.
3604         PR tree-optimization/83198
3605         * gimple-ssa-sprintf.c (format_floating): Set type solely based on
3606         dir.modifier, regardless of TREE_TYPE (arg).  Assume non-REAL_CST
3607         value if arg is a REAL_CST with incompatible type.
3609 2017-12-14  Sudakshina Das  <sudi.das@arm.com>
3610             Bin Cheng  <bin.cheng@arm.com>
3612         PR target/81228
3613         * config/aarch64/aarch64.c (aarch64_select_cc_mode): Move LTGT
3614         to CCFPEmode.
3615         * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Add
3616         LTGT.
3618 2017-12-14  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
3620         * config/aarch64/aarch64-cores.def (cortex-a55, cortex-a75,
3621         cortex-a75.cortex-a55): Specify AARCH64_FL_F16 in the arch features.
3623 2017-12-14  Richard Sandiford  <richard.sandiford@linaro.org>
3624             Alan Hayward  <alan.hayward@arm.com>
3625             David Sherwood  <david.sherwood@arm.com>
3627         * poly-int.h: New file.
3628         * poly-int-types.h: Likewise.
3629         * coretypes.h: Include them.
3630         (POLY_INT_CONVERSION): Define.
3631         * target.def (estimated_poly_value): New hook.
3632         * doc/tm.texi.in (TARGET_ESTIMATED_POLY_VALUE): New hook.
3633         * doc/tm.texi: Regenerate.
3634         * doc/poly-int.texi: New file.
3635         * doc/gccint.texi: Include it.
3636         * doc/rtl.texi: Describe restrictions on subreg modes.
3637         * Makefile.in (TEXI_GCCINT_FILES): Add poly-int.texi.
3638         * genmodes.c (NUM_POLY_INT_COEFFS): Provide a default definition.
3639         (emit_insn_modes_h): Emit a definition of NUM_POLY_INT_COEFFS.
3640         * targhooks.h (default_estimated_poly_value): Declare.
3641         * targhooks.c (default_estimated_poly_value): New function.
3642         * target.h (estimated_poly_value): Likewise.
3643         * wide-int.h (WI_UNARY_RESULT): Use wi::binary_traits.
3644         (wi::unary_traits): Delete.
3645         (wi::binary_traits::signed_shift_result_type): Define for
3646         offset_int << HOST_WIDE_INT, etc.
3647         (generic_wide_int::operator <<=): Define for all types and use
3648         wi::lshift instead of <<.
3649         (wi::hwi_with_prec): Add a default constructor.
3650         (wi::ints_for): New class.
3651         (operator <<): Define for all wide-int types.
3652         (operator /): New function.
3653         (operator %): Likewise.
3654         * selftest.h (ASSERT_KNOWN_EQ, ASSERT_KNOWN_EQ_AT, ASSERT_MAYBE_NE)
3655         (ASSERT_MAYBE_NE_AT): New macros.
3657 2017-12-13  Eric Botcazou  <ebotcazou@adacore.com>
3658             Dominik Vogt  <vogt@linux.vnet.ibm.com>
3660         PR middle-end/78468
3661         * emit-rtl.c (init_emit): Remove ??? comment.
3662         * explow.c (get_dynamic_stack_size): Take known alignment of stack
3663         pointer + STACK_DYNAMIC_OFFSET into account in lieu of STACK_BOUNDARY.
3664         * config/sparc/sparc.h (INIT_EXPANDERS): In 32-bit mode, lower the
3665         alignment of 3 virtual registers to BITS_PER_WORD.
3667         * config/sparc/sparc.c (sparc_compute_frame_size): Simplify.
3669 2017-12-13  Peter Bergner  <bergner@vnet.ibm.com>
3671         * config/rs6000/ppc-auxv.h (PPC_FEATURE2_HTM_NO_SUSPEND): New define.
3672         * config/rs6000/rs6000.c (cpu_supports_info): Use it.
3674 2017-12-13  Alexandre Oliva  <aoliva@redhat.com>
3676         PR bootstrap/83396
3677         * reload1.c (eliminate_regs_in_insn): Skip debug markers.
3679 2017-12-13  Jakub Jelinek  <jakub@redhat.com>
3681         * tree-cfg.c (verify_gimple_in_cfg): Verify no non-label stmts
3682         with the exception of debug begin stmt markers appear before
3683         labels.
3685         PR bootstrap/83396
3686         * final.c (rest_of_handle_final): Call variable_tracking_main only
3687         if !flag_var_tracking.
3689 2017-12-13  Alexandre Oliva  <aoliva@redhat.com>
3690             Jakub Jelinek  <jakub@redhat.com>
3692         PR bootstrap/83396
3693         PR debug/83391
3694         * tree-cfgcleanup.c (remove_forwarder_block): Keep after
3695         labels debug stmts that can only appear after labels.
3697 2017-12-13  Alexander Monakov  <amonakov@ispras.ru>
3699         PR rtl-optimization/82398
3700         * sel-sched.c (sel_rank_for_schedule): Fix check for zero
3701         EXPR_USEFULNESS in priority comparison.
3703 2017-12-13  Segher Boessenkool  <segher@kernel.crashing.org>
3705         PR rtl-optimization/83393
3706         * combine.c (move_deaths): If reg_stat points to a too new insn in
3707         last_death, do not use it: find the proper insn instead.
3709 2017-12-12  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
3711         PR target/82190
3712         * config/rs6000/rs6000-string.c (expand_block_compare,
3713         expand_strn_compare): Fix set_mem_size() calls.
3715 2017-12-12  Jeff Law  <law@redhat.com>
3717         PR tree-optimization/83298
3718         PR tree-optimization/83362
3719         PR tree-optimization/83383
3720         * gimple-ssa-evrp-analyze.h (class evrp_range_analyzer): Make
3721         push_value_range a public interface.  Add new argument to
3722         record_ranges_from_stmt.
3723         * gimple-ssa-evrp-analyze.c
3724         (evrp_range_analyzer::record_ranges_from_stmt): Add new argument.
3725         Update comments.  Handle recording temporary equivalences.
3726         * tree-ssa-dom.c (dom_opt_opt_walker::before_dom_children): Add
3727         new argument to call to evrp_range_analyzer::record_ranges_from_stmt.
3728         * gimple-ssa-evrp.c (evrp_dom_walker::before_dom_children): Likewise.
3729         * tree-ssa-threadedge.c: Include alloc-pool.h, vr-values.h and
3730         gimple-ssa-evrp-analyze.h.
3731         (record_temporary_equivalences_from_phis): Add new argument.  When
3732         the PHI arg is an SSA_NAME, set the result's range to the range
3733         of the PHI arg.
3734         (record_temporary_equivalences_from_stmts_at_dest): Record ranges
3735         from statements too.
3736         (thread_through_normal_block): Accept new argument, evrp_range_analyzer.
3737         Pass it down to children as needed.
3738         (thread_outgoing_edges): Likewise.
3739         (thread_across_edge): Likewise.   Push/pop range state as needed.
3740         * tree-ssa-threadedge.h (thread_outgoing_edges): Update prototype.
3742 2017-12-12  Julia Koval  <julia.koval@intel.com>
3744         * config/i386/i386.c (PTA_SKYLAKE_AVX512): Add PTA_CLWB.
3745         (PTA_CANNONLAKE): Remove PTA_CLWB.
3747 2017-12-12  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
3749         PR target/83332
3750         * config/rs6000/vector.md (vcondv2dfv2di): New define_expand.
3751         (vcondv2div2df): Likewise.
3752         (vconduv2dfv2di): Likewise.
3754 2017-12-12  Jakub Jelinek  <jakub@redhat.com>
3756         * builtins.def (BUILT_IN_NEXTAFTER, BUILT_IN_NEXTAFTERF,
3757         BUILT_IN_NEXTAFTERL, BUILT_IN_NEXTTOWARD, BUILT_IN_NEXTTOWARDF,
3758         BUILT_IN_NEXTTOWARDL): Use ATTR_MATHFN_ERRNO instead of
3759         ATTR_MATHFN_FPROUNDING_ERRNO.
3761 2017-12-12  Richard Biener  <rguenther@suse.de>
3763         PR tree-optimization/83385
3764         * graphite-scop-detection.c (get_order, order): Remove.
3765         (bb_to_rpo): New global.
3766         (cmp_pbbs): Adjust.
3767         (build_scops): Sort pbbs in RPO order.
3769 2017-12-12  James Greenhalgh  <james.greenhalgh@arm.com>
3771         * combine.c (simplify_set): Do not transform subregs to zero_extends
3772         if the destination mode is a vector mode.
3774 2017-12-12  Jakub Jelinek  <jakub@redhat.com>
3776         PR tree-optimization/83359
3777         * tree-cfg.h (fold_loop_internal_call): Declare.
3778         * tree-vectorizer.c (fold_loop_internal_call): Moved to ...
3779         * tree-cfg.c (fold_loop_internal_call): ... here.  No longer static.
3780         (find_loop_dist_alias): New function.
3781         (move_sese_region_to_fn): If any dloop->orig_loop_num value is
3782         updated, also adjust any corresponding LOOP_DIST_ALIAS internal
3783         calls.
3785         PR tree-optimization/80631
3786         * tree-vect-loop.c (get_initial_def_for_reduction): Fix comment typo.
3787         (vect_create_epilog_for_reduction): Add INDUC_VAL and INDUC_CODE
3788         arguments, for INTEGER_INDUC_COND_REDUCTION use INDUC_VAL instead of
3789         hardcoding zero as the value if COND_EXPR is never true.  For
3790         INTEGER_INDUC_COND_REDUCTION don't emit the final COND_EXPR if
3791         INDUC_VAL is equal to INITIAL_DEF, and use INDUC_CODE instead of
3792         hardcoding MAX_EXPR as the reduction operation.
3793         (is_nonwrapping_integer_induction): Allow negative step.
3794         (vectorizable_reduction): Compute INDUC_VAL and INDUC_CODE for
3795         vect_create_epilog_for_reduction, if no value is suitable, don't
3796         use INTEGER_INDUC_COND_REDUCTION for now.  Formatting fixes.
3798 2017-12-12  Richard Biener  <rguenther@suse.de>
3800         PR tree-optimization/81889
3801         * tree-ssa-loop-niter.c (infer_loop_bounds_from_signedness): Use
3802         range info from the non-wrapping IV instead of just the range
3803         of the type.
3805 2017-12-12  Julia Koval  <julia.koval@intel.com>
3807         * config.gcc: Add vaesintrin.h.
3808         * config/i386/i386-builtin-types.def (V64QI_FTYPE_V64QI_V64QI):
3809         New type.
3810         * config/i386/i386-builtin.def (__builtin_ia32_vaesdec_v16qi,
3811         __builtin_ia32_vaesdec_v32qi, __builtin_ia32_vaesdec_v64qi):
3812         New builtins.
3813         * config/i386/i386.c (ix86_expand_args_builtin): Handle new type.
3814         * config/i386/immintrin.h: Include vaesintrin.h.
3815         * config/i386/sse.md (vaesdec_<mode>): New pattern.
3816         * config/i386/vaesintrin.h (_mm256_aesdec_epi128, _mm512_aesdec_epi128,
3817         _mm_aesdec_epi128): New intrinsics.
3819 2017-12-12  Julia Koval  <julia.koval@intel.com>
3821         * common/config/i386/i386-common.c (OPTION_MASK_ISA_VAES_SET,
3822         OPTION_MASK_ISA_VAES_UNSET): New.
3823         (ix86_handle_option): Handle -mvaes.
3824         * config/i386/cpuid.h: Define bit_VAES.
3825         * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mvaes.
3826         * config/i386/i386-c.c (__VAES__): New.
3827         * config/i386/i386.c (ix86_target_string): Add -mvaes.
3828         (ix86_valid_target_attribute_inner_p): Ditto.
3829         * config/i386/i386.h (TARGET_VAES, TARGET_VAES_P): New.
3830         * config/i386/i386.opt: Add -mvaes.
3831         * doc/invoke.texi: Ditto.
3833 2017-12-12  Alexandre Oliva  <aoliva@redhat.com>
3835         * debug.h (gcc_debug_hooks): Add inline_entry.
3836         * dbxout.c (dbx_debug_hooks, xcoff_debug_hooks): Likewise.
3837         * debug.c (do_nothing_debug_hooks): Likewise.
3838         * vmsdbgout.c (vmsdbg_debug_hooks): Likewise.
3839         * dwarf2out.c (dwarf2_debug_hooks): Likewise.
3840         (dwarf2_lineno_debug_hooks): Likewise.
3842         * common.opt (gstatement-frontiers): New, setting
3843         debug_nonbind_markers_p.
3844         * rtl.h (MAY_HAVE_DEBUG_MARKER_INSNS): Activate.
3845         * toplev.c (process_options): Autodetect value for debug statement
3846         frontiers option.
3847         * tree.h (MAY_HAVE_DEBUG_MARKER_STMTS): Activate.
3848         * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers): New.
3850         * cfgexpand.c (expand_gimple_basic_block): Handle begin stmt
3851         markers.  Integrate source bind into debug stmt expand loop.
3852         (pass_expand::execute): Check debug marker limit.  Avoid deep
3853         TER and expand debug locations for debug bind insns only.
3854         * cse.c (insn_live_p): Keep nonbind markers and debug bindings
3855         followed by them.
3856         * df-scan.c (df_insn_delete): Accept out-of-block debug insn.
3857         * final.c (reemit_insn_block_notes): Take current block from
3858         nonbind markers.  Declare note where it's first set.
3859         (final_scan_insn): Handle begin stmt notes.  Emit is_stmt according to
3860         begin stmt markers if enabled.
3861         (notice_source_line): Handle nonbind markers.  Fail if their
3862         location is unknown or that of builtins.
3863         (rest_of_handle_final): Convert begin stmt markers to notes if
3864         var-tracking didn't run.
3865         (rest_of_clean_state): Skip begin stmt markers.
3866         * gimple-pretty-print.c (dump_gimple_debug): Handle begin stmt
3867         markers.
3868         * function.c (allocate_struct_function): Set begin_stmt_markers.
3869         * function.h (struct function): Add debug_marker_count counter
3870         and debug_nonbind_markers flag.
3871         * gimple-iterator.c (gsi_remove): Adjust debug_marker_count.
3872         * gimple-low.c (lower_function_body): Adjust
3873         debug_nonbind_markers.
3874         (lower_stmt): Drop or skip gimple debug stmts.
3875         (lower_try_catch): Skip debug stmts.
3876         * gimple.c (gimple_build_debug_begin_stmt): New.
3877         (gimple_copy): Increment debug_marker_count if copying one.
3878         * gimple.h (gimple_build_debug_begin_stmt): Declare.
3879         * gimplify.c (rexpr_location): New.
3880         (rexpr_has_location): New.
3881         (warn_switch_unreachable_r): Handle gimple debug stmts.
3882         (shortcut_cond_r): Call expr_location.
3883         (find_goto): New.
3884         (find_goto_label): New.
3885         (shortcut_cond_expr): Call expr_has_location, expr_location, and
3886         find_goto_label.
3887         (gimplify_cond_expr): Call find_goto_label, expr_has_location, and
3888         expr_location.
3889         (gimplify_expr): Handle begin stmt markers.  Reject debug expr decls.
3890         * langhooks-def.h (LANG_HOOKS_EMITS_BEGIN_STMT): New.  Add to...
3891         (LANG_HOOKS_INITIALIZER): ... this.
3892         * langhooks.h (struct lang_hooks): Add emits_begin_stmt.
3893         * lra-contraints.c (inherit_reload_reg): Tolerate between-blocks
3894         debug insns.
3895         (update_ebb_live_info): Skip debug insn markers.
3896         * lra.c (debug_insn_static_data): Rename to...
3897         (debug_bind_static_data): ... this.
3898         (debug_marker_static_data): New.
3899         (lra_set_insn_recog_data): Select one of the above depending
3900         on debug insn kind.
3901         (lra_update_isn_regno_info): Don't assume debug insns have
3902         freqs.
3903         (push_insns): Skip debug insns.
3904         * lto-streamer-in.c (input_function): Drop debug stmts
3905         depending on active options.  Adjust debug_nonbind_markers.
3906         * params.def (PARAM_MAX_DEBUG_MARKER_COUNT): New.
3907         * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
3908         begin stmt marker notes.
3909         (print_insn): Likewise.
3910         * recog.c (extract_insn): Recognize rtl for debug markers.
3911         * rtl.def (DEBUG_MARKER): New.
3912         * tree-inline.c: Include params.h.
3913         (remap_gimple_stmt): Handle nonbind markers.
3914         (maybe_move_debug_stmts_to_successors): Likewise.
3915         (copy_debug_stmt): Likewise.
3916         * tree-iterator.c (append_to_statement_list_1): Append begin stmt
3917         markers regardless of no side effects.
3918         (tsi_link_before): Don't update container's side effects when adding
3919         a begin stmt marker.
3920         (tsi_link_after): Likewise.
3921         (expr_first): Skip begin stmt markers.
3922         (expr_last): Likewise.
3923         * tree-pretty-print (dump_generic_node): Handle begin stmt markers.
3924         * tree-ssa-threadedge.c (propagate_threaded_block_debug_info):
3925         Disregard nonbind markers.
3926         * tree.c (make_node_stat): Don't set side effects for begin stmt
3927         markers.
3928         (build1_stat): Likewise.
3929         * tree.def (DEBUG_BEGIN_STMT): New.
3930         * tree.h (GOTO_DESTINATION): Require a GOTO_EXPR.
3931         * var-tracking.c (delete_debug_insns): Renamed to...
3932         (delete_vta_debug_insns): ... this.
3933         (reemit_marker_as_note): New.
3934         (vt_initialize): Reemit markers.
3935         (delete_vta_debug_insns): Likewise.
3936         (vt_debug_insns_local): Reemit or delete markers.
3937         (variable_tracking_main_1): Likewise.
3938         * doc/generic.texi (DEBUG_BEGIN_STMT): Document.
3939         * doc/gimple.texi (gimple_debug_begin_stmt_p): New.
3940         (gimple_debug_nonbind_marker_p): New.
3941         (gimple_build_debug_bind): Adjust.
3942         (gimple_build_debug_begin_stmt): New.
3943         * doc/invoke.texi (max-debug-marker-count): New param.
3944         * doc/rtl.texi (debug_implicit_ptr, entry_value): New.
3945         (debug_parameter_ref, debug_marker): New.
3946         (NOTE_INSN_BEGIN_STMT): New.
3947         (DEBUG_INSN): Describe begin stmt markers.
3949         * cfgbuild.c (find_bb_boundaries): Don't purge dead edges if,
3950         without debug insns, we wouldn't, but clean up debug insns
3951         after a control flow insn nevertheless.
3953         * cfgcleanup.c (delete_unreachable_blocks): Use alternate
3954         block removal order if MAY_HAVE_DEBUG_BIND_INSNS.
3955         * cfgexpand.c (label_rtx_for_bb): Skip debug insns.
3956         * cfgrtl.c (try_redirect_by_replacing_jump): Skip debug insns.
3957         (rtl_tidy_fallthru_edge): Likewise.
3958         (rtl_verify_fallthru): Likewise.
3959         (rtl_verify_bb_layout): Likewise.
3960         (skip_insns_after_block): Likewise.
3961         (duplicate_insn_chain): Use DEBUG_BIND_INSN_P.
3962         * dwarf2out.c: Include print-rtl.h.
3963         (dwarf2out_next_real_insn): New.
3964         (dwarf2out_var_location): Call it.  Disregard begin stmt markers.
3965         Dump debug binds in asm comments.
3966         * gimple-iterator.c (gimple_find_edge_insert_loc): Skip debug stmts.
3967         * gimple-iterator.h (gsi_start_bb_nondebug): Remove; adjust
3968         callers to use gsi_start_nondebug_bb instead.
3969         (gsi_after_labels): Skip gimple debug stmts.
3970         (gsi_start_nondebug): New.
3971         * gimple-loop-interchange.c (find_deps_in_bb_for_stmt): Adjust.
3972         (proper_loop_form_for_interchange): Adjust.
3973         * gimple-low.c (gimple_seq_may_fallthru): Take last nondebug stmt.
3974         * gimple.h (gimple_seq_last_nondebug_stmt): New.
3975         * gimplify.c (last_stmt_in_scope): Skip debug stmts.
3976         (collect_fallthrough_labels): Likewise.
3977         (should_warn_for_implicit_fallthrough): Likewise.
3978         (warn_implicit_fallthrough_r): Likewise.
3979         (expand_FALLTHROUGH_r): Likewise.
3980         * graphite-isl-ast-to-gimple.c (gsi_insert_earliest): Adjust.
3981         (graphite_copy_stmts_from_block): Skip nonbind markers.
3982         * haifa-sched.c (sched_extend_bb): Skip debug insns.
3983         * ipa-icf-gimple.c (func_checker::compare_bb): Adjust.
3984         * jump.c (clean_barriers): Skip debug insns.
3985         * omp-expand.c (expand_parallel_call): Skip debug insns.
3986         (expand_task_call): Likewise.
3987         (remove_exit_barrier): Likewise.
3988         (expand_omp_taskreg): Likewise.
3989         (expand_omp_for_init_counts): Likewise.
3990         (expand_omp_for_generic): Likewise.
3991         (expand_omp_for_static_nochunk): Likewise.
3992         (expand_omp_for_static_chunk): Likewise.
3993         (expand_omp_simd): Likewise.
3994         (expand_omp_taskloop_for_outer): Likewise.
3995         (expand_omp_taskloop_for_inner): Likewise.
3996         (expand_oacc_for): Likewise.
3997         (expand_omp_sections): Likewise.
3998         (expand_omp_single): Likewise.
3999         (expand_omp_synch): Likewise.
4000         (expand_omp_atomic_load): Likewise.
4001         (expand_omp_atomic_store): Likewise.
4002         (expand_omp_atomic_fetch_op): Likewise.
4003         (expand_omp_atomic_pipeline): Likewise.
4004         (expand_omp_atomic_mutex): Likewise.
4005         (expand_omp_target): Likewise.
4006         (grid_expand_omp_for_loop): Likewise.
4007         (grid_expand_target_grid_body): Likewise.
4008         (build_omp_regions_1): Likewise.
4009         * omp-low.c (check_combined_parallel): Skip debug stmts.
4010         * postreload.c (fixup_debug_insns): Skip nonbind debug insns.
4011         * regcprop.c (find_oldest_value_reg): Ensure REGNO is not a pseudo.
4012         * sese.c (sese_trivially_empty_bb_p): Call is_gimple_debug in
4013         test.
4014         * tree-cfg.c (make_blobs_1): Skip debug stmts.
4015         (make_edges): Likewise.
4016         (cleanup_dead_labels): Likewise.
4017         (gimple_can_merge_blocks_p): Likewise.
4018         (stmt_starts_bb_p): Likewise.
4019         (gimple_block_label): Likewise.
4020         (gimple_redirect_edge_and_branch): Likewise.
4021         * tree-cfgcleanup.c (remove_forwarder_block): Rearrange skipping
4022         of debug stmts.
4023         (execute_cleanup_cfg_post_optimizing): Dump enumerated decls with
4024         TDF_SLIM.
4025         * tree-pretty-print (print_declaration): Omit initializer in slim
4026         dumps.
4027         * tree-ssa-dce.c (mark_stmt_if_obviously_necessary): Mark begin stmt
4028         markers.
4029         (eliminate_unnecessary_stmts): Stabilize block removal order.
4030         * tree-ssa-tail-merge.c (find_duplicate): Skip debug stmts.
4031         * var-tracking.c (get_first_insn): New.
4032         (vt_emit_notes): Call it.
4033         (vt_initialize): Walk any insns before the first BB.
4034         (delete_debug_insns): Likewise.
4036         * gimple.h (enum gimple_debug_subcode): Add
4037         GIMPLE_DEBUG_BEGIN_STMT.
4038         (gimple_debug_begin_stmt_p): New.
4039         (gimple_debug_nonbind_marker_p): New.
4040         * tree.h (MAY_HAVE_DEBUG_MARKER_STMTS): New.
4041         (MAY_HAVE_DEBUG_BIND_STMTS): Renamed from....
4042         (MAY_HAVE_DEBUG_STMTS): ... this.  Check both.
4043         * insn-notes.def (BEGIN_STMT): New.
4044         * rtl.h (MAY_HAVE_DEBUG_MARKER_INSNS): New.
4045         (MAY_HAVE_DEBUG_BIND_INSNS): Renamed from....
4046         (MAY_HAVE_DEBUG_INSNS): ... this.  Check both.
4047         (NOTE_MARKER_LOCATION, NOTE_MARKER_P): New.
4048         (DEBUG_BIND_INSN_P, DEBUG_MARKER_INSN_P): New.
4049         (INSN_DEBUG_MARKER_KIND): New.
4050         (GEN_RTX_DEBUG_MARKER_BEGIN_STMT_PAT): New.
4051         (INSN_VAR_LOCATION): Check for VAR_LOCATION.
4052         (INSN_VAR_LOCATION_PTR): New.
4053         * cfgexpand.c (expand_debug_locations): Handle debug bind insns
4054         only.
4055         (expand_gimple_basic_block): Likewise.  Emit debug temps for TER
4056         deps only if debug bind insns are enabled.
4057         (pass_expand::execute): Avoid deep TER and expand
4058         debug locations for debug bind insns only.
4059         * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Narrow
4060         debug stmts special handling down to debug bind stmts.
4061         * combine.c (try_combine): Narrow debug insns special handling
4062         down to debug bind insns.
4063         * cse.c (delete_trivially_dead_insns): Handle debug bindings.
4064         Narrow debug insns preexisting special handling down to debug
4065         bind insns.
4066         * dce.c (rest_of_handle_ud_dce): Narrow debug insns special
4067         handling down to debug bind insns.
4068         * function.c (instantiate_virtual_regs): Skip debug markers,
4069         adjust handling of debug binds.
4070         * gimple-ssa-backprop.c (backprop::prepare_change): Try debug
4071         temp insertion iff MAY_HAVE_DEBUG_BIND_STMTS.
4072         * haifa-sched.c (schedule_insn): Narrow special handling of debug
4073         insns to debug bind insns.
4074         * ipa-param-manipulation.c (ipa_modify_call_arguments): Narrow
4075         special handling of debug stmts to debug bind stmts.
4076         * ipa-split.c (split_function): Likewise.
4077         * ira.c (combine_and_move_insns): Adjust debug bind insns only.
4078         * loop-unroll.c (apply_opt_in_copies): Adjust tests on bind
4079         debug insns.
4080         * reg-stack.c (convert_regs_1): Use DEBUG_BIND_INSN_P.
4081         * regrename.c (build_def_use): Likewise.
4082         * regcprop.c (copyprop_hardreg_forward_1): Likewise.
4083         (pass_cprop_hardreg): Narrow special casing of debug insns to
4084         debug bind insns.
4085         * regstat.c (regstat_init_n_sets_and_refs): Likewise.
4086         * reload1.c (reload): Likewise.
4087         * sese.c (sese_insert_phis_for_liveouts): Narrow special
4088         casing of debug stmts to debug bind stmts.
4089         * shrink-wrap.c (move_insn_for_shrink_wrap): Likewise.
4090         * ssa-iterators.h (num_imm_uses): Likewise.
4091         * tree-cfg.c (gimple_merge_blocks): Narrow special casing of
4092         debug stmts to debug bind stmts.
4093         * tree-inline.c (tree_function_versioning): Narrow special casing
4094         of debug stmts to debug bind stmts.
4095         * tree-loop-distribution.c (generate_loops_for_partition):
4096         Narrow special casing of debug stmts to debug bind stmts.
4097         * tree-sra.c (analyze_access_subtree): Narrow special casing
4098         of debug stmts to debug bind stmts.
4099         * tree-ssa-dce.c (remove_dead_stmt): Narrow special casing of debug
4100         stmts to debug bind stmts.
4101         * tree-ssa-loop-ivopt.c (remove_unused_ivs): Narrow special
4102         casing of debug stmts to debug bind stmts.
4103         * tree-ssa-reassoc.c (reassoc_remove_stmt): Likewise.
4104         * tree-ssa-tail-merge.c (tail_merge_optimize): Narrow special
4105         casing of debug stmts to debug bind stmts.
4106         * tree-ssa-threadedge.c (propagate_threaded_block_debug_info):
4107         Likewise.
4108         * tree-ssa.c (flush_pending_stmts): Narrow special casing of
4109         debug stmts to debug bind stmts.
4110         (gimple_replace_ssa_lhs): Likewise.
4111         (insert_debug_temp_for_var_def): Likewise.
4112         (insert_debug_temps_for_defs): Likewise.
4113         (reset_debug_uses): Likewise.
4114         * tree-ssanames.c (release_ssa_name_fn): Likewise.
4115         * tree-vect-loop-manip.c (adjust_debug_stmts_now): Likewise.
4116         (adjust_debug_stmts): Likewise.
4117         (adjust_phi_and_debug_stmts): Likewise.
4118         (vect_do_peeling): Likewise.
4119         * tree-vect-loop.c (vect_transform_loop): Likewise.
4120         * valtrack.c (propagate_for_debug): Use BIND_DEBUG_INSN_P.
4121         * var-tracking.c (adjust_mems): Narrow special casing of debug
4122         insns to debug bind insns.
4123         (dv_onepart_p, dataflow_set_clar_at_call, use_type): Likewise.
4124         (compute_bb_dataflow, vt_find_locations): Likewise.
4125         (vt_expand_loc, emit_notes_for_changes): Likewise.
4126         (vt_init_cfa_base): Likewise.
4127         (vt_emit_notes): Likewise.
4128         (vt_initialize): Likewise.
4129         (vt_finalize): Likewise.
4131         * emit-rtl.c (next_nondebug_insn, prev_nondebug_insn): Reorder.
4132         (next_nonnote_nondebug_insn, prev_nonnote_nondebug_insn): Reorder.
4133         (next_nonnote_nondebug_insn_bb): New.
4134         (prev_nonnote_nondebug_insn_bb): New.
4135         (prev_nonnote_insn_bb, next_nonnote_insn_bb): Remove.
4136         * rtl.h (prev_nonnote_insn_bb, next_nonnote_insn_bb): Remove decls.
4137         (prev_nonnote_nondebug_insn_bb): Declare.
4138         (next_nonnote_nondebug_insn_bb): Declare.
4139         * cfgbuild.c (find_bb_boundaries): Adjust to skip debug insns.
4140         * cfgrtl.c (get_last_bb_insn): Likewise.
4141         * lra.c (push_insns): Likewise.
4143 2017-12-11  David Malcolm  <dmalcolm@redhat.com>
4145         PR c/82050
4146         * selftest-run-tests.c (selftest::run_tests): Move start/finish code
4147         to...
4148         * selftest.c (selftest::test_runner::test_runner): New ctor.
4149         (selftest::test_runner::~test_runner): New dtor.
4150         * selftest.h (class selftest::test_runner): New class.
4152 2017-12-11  Carl Love  <cel@us.ibm.com>
4154         * config/rs6000/altivec.h (vec_extract_fp32_from_shorth,
4155         vec_extract_fp32_from_shortl]): Add #defines.
4156         * config/rs6000/rs6000-builtin.def (VSLDOI_2DI): Add macro expansion.
4157         * config/rs6000/rs6000-c.c (ALTIVEC_BUILTIN_VEC_UNPACKH,
4158         ALTIVEC_BUILTIN_VEC_UNPACKL, ALTIVEC_BUILTIN_VEC_AND,
4159         ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SRL,
4160         ALTIVEC_BUILTIN_VEC_SRO, ALTIVEC_BUILTIN_VEC_SLD,
4161         ALTIVEC_BUILTIN_VEC_SLL): Add expansions.
4162         * doc/extend.texi: Add documentation for the added builtins.
4164 2017-12-11  Alexandre Oliva  <aoliva@redhat.com>
4166         PR rtl-optimization/80693
4167         PR rtl-optimization/81019
4168         PR rtl-optimization/81020
4169         * combine.c (distribute_notes): Reset any REG_UNUSED REGs that
4170         are not mentioned in i3.  Place the REG_UNUSED note on i2,
4171         possibly modified to REG_DEAD, if it did not originate in i3.
4173 2017-12-11  Jakub Jelinek  <jakub@redhat.com>
4175         * recog.c (store_data_bypass_p_1): New function.
4176         (store_data_bypass_p): Handle USE in a PARALLEL like CLOBBER.  Use
4177         store_data_bypass_p_1 to avoid code duplication.  Formatting fixes.
4179 2017-12-11  Segher Boessenkool  <segher@kernel.crashing.org>
4181         PR rtl-optimization/83361
4182         * ifcvt.c (if_convert): Call fixup_partitions.
4184 2017-12-11  Will Schmidt  <will_schmidt@vnet.ibm.com>
4186         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add support for
4187         early folding of splat_u{8,16,32}.
4189 2017-12-11  Jakub Jelinek  <jakub@redhat.com>
4191         * config/aarch64/aarch64.c (aarch64_print_operand): Don't start
4192         output_operand_lossage first argument with capital letter.
4193         (aarch64_override_options): Don't start error and sorry first argument
4194         with capital letter.
4196 2017-12-11  Andi Kleen  <ak@linux.intel.com>
4198         PR gcov-profile/83355
4199         * auto-profile.c (string_table::get_index_by_decl): Don't
4200         recurse when abstract origin points to itself.
4202 2017-12-11  Bin Cheng  <bin.cheng@arm.com>
4204         PR tree-optimization/83320
4205         * gimple-loop-interchange.cc (free_data_refs_with_aux): Use delete.
4206         (prune_datarefs_not_in_loop): Ditto.
4208 2017-12-10  Gerald Pfeifer  <gerald@pfeifer.com>
4210         * doc/install.texi (Specific): Tweak link to mkssoftware.com.
4212 2017-12-10  Jakub Jelinek  <jakub@redhat.com>
4214         PR tree-optimization/83337
4215         * gimple-loop-interchange.cc (compute_access_stride): Handle
4216         bitfield DRs properly.
4218 2017-12-09  Jakub Jelinek  <jakub@redhat.com>
4220         PR tree-optimization/83338
4221         * tree-vect-stmts.c (vectorizable_operation): Handle POINTER_DIFF_EXPR
4222         vectorization as MINUS_EXPR with a subsequent VIEW_CONVERT_EXPR from
4223         vector of unsigned integers to vector of signed integers.
4225 2017-12-08  Vladimir Makarov  <vmakarov@redhat.com>
4227         PR rtl-optimization/83317
4228         * lra-constraints.c (process_address_1): Add insn code check.
4230 2017-12-08  Michael Matz  <matz@suse.de>
4232         Fix PR tree-optimization/83323
4233         * gimple-loop-jam (unroll_jam_possible_p): Correct test for
4234         head-controlled loops and loop BBs.
4235         * common.opt (funroll-and-jam): Remove, instead ...
4236         (floop-unroll-and-jam): ... reuse this option.
4237         * opts.c (default_options_table): Use OPT_floop_unroll_and_jam.
4238         * doc/invoke.texi (-funroll-and-jam): Move docu to ...
4239         (-floop-unroll-and-jam): ... this option.
4241 2017-12-08  Jakub Jelinek  <jakub@redhat.com>
4243         * ipa-polymorphic-call.c (noncall_stmt_may_be_vtbl_ptr_store): Fix
4244         a comment typo, get_base_ref_and_offset -> get_ref_base_and_extent.
4245         * ipa-prop.c (stmt_may_be_vtbl_ptr_store): Likewise.
4247 2017-12-08  Richard Biener  <rguenther@suse.de>
4249         PR middle-end/81782
4250         * tree-ssa-uninit.c (warn_uninitialized_vars): Properly
4251         handle accesses outside of zero-sized vars.
4253 2017-12-08  Martin Jambor  <mjambor@suse.cz>
4255         PR tree-optimization/83141
4256         * tree-sra.c (contains_vce_or_bfcref_p): Move up in the file, also
4257         test for MEM_REFs implicitely changing types with padding.  Remove
4258         inline keyword.
4259         (build_accesses_from_assign): Added contains_vce_or_bfcref_p checks.
4261 2017-12-08  Jakub Jelinek  <jakub@redhat.com>
4263         * config/arc/arc.c (arc_attribute_table): Add exclusions to
4264         the comment.
4265         * config/avr/avr.c (avr_attribute_table): Likewise.
4266         * config/msp430/msp430.c (msp430_attribute_table): Likewise.
4267         * config/rl78/rl78.c (rl78_attribute_table): Likewise.
4268         * config/nds32/nds32.c (nds32_attribute_table): Likewise.
4269         * config/darwin.h (SUBTARGET_ATTRIBUTE_TABLE): Initialize new member
4270         of struct attribute_spec.
4271         * config/i386/cygming.h (SUBTARGET_ATTRIBUTE_TABLE): Likewise.
4273 2017-12-08  Ulrich Weigand  <uweigand@de.ibm.com>
4275         PR target/82960
4276         * config/spu/spu.c (pad_bb): Only check INSN_CODE when INSN_P is true.
4278 2017-12-08  Jan Hubicka  <hubicka@ucw.cz>
4280         PR middle-end/83609
4281         * profile-count.c (profile_count::from_gcov_type): Move from
4282         profile-count.h; handle overflow.
4283         * profile-count.h (profile_count::from_gcov_type): Move offline.
4285 2017-12-08  Segher Boessenkool  <segher@kernel.crashing.org>
4287         PR rtl-optimization/83304
4288         * combine.c (move_deaths): If we do not know where a register died,
4289         search for it.
4291 2017-12-08  Richard Biener  <rguenther@suse.de>
4293         * gimple-loop-interchange.cc (tree_loop_interchange::interchange):
4294         Provide -fopt-info-loop feedback when we interchange in a nest.
4296 2017-06-08  Richard Earnshaw  <rearnsha@arm.com>
4298         * config/arm/driver-arm.c (arm_cpu_table): Use fp-capable product names
4299         for armv6 ARM CPU IDs.
4301 2017-06-08  Richard Earnshaw  <rearnsha@arm.com>
4303         * common/config/arm/arm-common.c: Include <algorithm>.
4304         (INCLUDE_VECTOR): Define.
4305         (compare_opt_names): New function.
4306         (arm_rewrite_selected_arch): Only strip out extensions that can be
4307         expressed through -mfpu.  Sort the remaining extensions
4308         alphabetically.
4310 2017-06-08  Richard Earnshaw  <rearnsha@arm.com>
4312         * config/arm/arm.h (arm_asm_auto_mfpu): Declare.
4313         (ASM_CPU_SPEC_FUNCTIONS): Add new rule asm_auto_mfpu.
4314         (ASM_CPU_SPEC): Use it if -mfpu is set to auto.
4315         * common/config/arm/arm-common.c (arm_asm_auto_mfpu): New function.
4317 2017-06-08  Tristan Gingold  <gindold@adacore.com>
4319         PR ada/81470
4320         * dwarf2out.c (dwarf2out_do_cfi_startproc): Only emit
4321         .cfi_personality or .cfi_lsda if the eh data format is dwarf2.
4323 2017-12-08  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
4325         * config/sol2.h (SOLARIS_ATTRIBUTE_TABLE): Initialize new member
4326         of struct attribute_spec.
4328 2017-12-08  Julia Koval  <julia.koval@intel.com>
4330         * config/i386/avx512vnniintrin.h (_mm512_dpwssds_epi32,
4331         _mm512_mask_dpwssds_epi32, _mm512_maskz_dpwssds_epi32): New intrinsics.
4332         * config/i386/avx512vnnivlintrin.h (_mm256_dpwssds_epi32,
4333         _mm256_mask_dpwssds_epi32, _mm256_maskz_dpwssds_epi32,
4334         _mm_dpwssds_epi32, _mm_mask_dpwssds_epi32,
4335         _mm_maskz_dpwssds_epi32): Ditto.
4337 2017-12-08  Richard Biener  <rguenther@suse.de>
4339         PR tree-optimization/81303
4340         * tree-vect-stmts.c (vect_is_simple_cond): For invariant
4341         conditions try to create a comparison vector type matching
4342         the data vector type.
4343         (vectorizable_condition): Adjust.
4344         * tree-vect-patterns.c (vect_recog_mask_conversion_pattern):
4345         Leave invariant conditions alone in case we can vectorize those.
4347 2017-12-08  Julia Koval  <julia.koval@intel.com>
4349         * config/i386/avx512vnniintrin.h (_mm512_dpwssd_epi32,
4350         _mm512_mask_dpwssd_epi32, _mm512_maskz_dpwssd_epi32): New intrinsics.
4351         * config/i386/avx512vnnivlintrin.h (_mm256_dpwssd_epi32,
4352         _mm256_mask_dpwssd_epi32, _mm256_maskz_dpwssd_epi32, _mm_dpwssd_epi32,
4353         _mm_mask_dpwssd_epi32, _mm_maskz_dpwssd_epi32): Ditto.
4355 2017-12-08  Julia Koval  <julia.koval@intel.com>
4357         * config/i386/avx512vnniintrin.h (_mm512_dpbusds_epi32,
4358         _mm512_mask_dpbusds_epi32, _mm512_maskz_dpbusds_epi32): New.
4359         * config/i386/avx512vnnivlintrin.h (_mm256_dpbusds_epi32,
4360         _mm256_mask_dpbusds_epi32, _mm256_maskz_dpbusds_epi32,
4361         _mm_dpbusds_epi32, _mm_mask_dpbusds_epi32,
4362         _mm_maskz_dpbusds_epi32): New intrinsics.
4364 2017-12-07  Sandra Loosemore  <sandra@codesourcery.com>
4366         * config/nios2/nios2.md (ld<bhw_uns>io): Add splitter for memory
4367         operand.
4368         (ld<bh>io_signed): Likewise.
4369         (st<bhw>io): Likewise.
4370         * config/nios2/predicates.md (ldstio_memory_operand): Allow
4371         SMALL_INT12 constant integer operand.
4373 2017-12-07  Sandra Loosemore  <sandra@codesourcery.com>
4375         * config/nios2/nios2.c (nios2_symbolic_constant_allowed):
4376         Rename to...
4377         (nios2_large_constant_allowed): ...this.  Adjust uses.
4378         (nios2_plus_symbolic_constant_p): Rename to...
4379         (nios2_plus_large_constant_p): ...this.  Adjust uses.
4380         (nios2_legitimate_address_p): Correct CONST_INT handling.
4381         (nios2_symbolic_memory_operand_p): Rename to...
4382         (nios2_large_constant_memory_operand_p): ...this.  Adjust uses.
4383         (nios2_large_constant_p): Check for large constant integers too.
4384         (nios2_split_large_constant): Handle constant integers.
4385         (nios2_split_symbolic_memory_operand): Rename to...
4386         (nios2_split_large_constant_memory_operand): ...this.  Adjust uses.
4387         (nios2_legitimize_constant_address): Handle constant integers.
4388         (r0rel_constant_p): Handle small constant integers.
4389         (nios2_print_operand_address): Handle r0-relative integer addresses.
4390         * config/nios2/nios2-protos.h: Adjust for renamed functions.
4391         * config/nios2/nios2.md: Adjust for renamed functions.
4393 2017-12-07  Andrew Waterman  <andrew@sifive.com>
4395         * config/riscv/riscv.c (TARGET_ASM_SELECT_SECTION): New define.
4396         (TARGET_HAVE_SRODATA_SECTION): New define.
4397         (riscv_select_section): New function.
4399 2017-12-08  Joseph Myers  <joseph@codesourcery.com>
4400             Alexander Monakov  <amonakov@ispras.ru>
4401             Jakub Jelinek  <jakub@redhat.com>
4403         PR target/81906
4404         * config/i386/i386.c (ix86_expand_rint): Handle flag_rounding_math.
4406 2017-12-07  Richard Sandiford  <richard.sandiford@linaro.org>
4408         * config/aarch64/aarch64.c (aarch64_print_address_internal): Return
4409         a bool success value.  Don't call output_operand_lossage here.
4410         (aarch64_print_ldpstp_address): Return a bool success value.
4411         (aarch64_print_operand_address): Call output_addr_const if
4412         aarch64_print_address_internal fails.
4413         (aarch64_print_operand): Don't assert that the mode is 16 bytes for
4414         'y'; call output_operand_lossage instead.  Call output_operand_lossage
4415         if aarch64_print_ldpstp_address fails.
4417 2017-12-07  Richard Sandiford  <richard.sandiford@linaro.org>
4419         * tree-vector-builder.h
4420         (tree_vector_builder::binary_encoded_nelts): Declare.
4421         * tree-vector-builder.c
4422         (tree_vector_builder::binary_encoded_nelts): New function.
4423         * fold-const.c (negate_expr_p): Likewise.
4424         (operand_equal_p, fold_checksum_tree): Likewise.
4425         * tree-loop-distribution.c (const_with_all_bytes_same): Likewise.
4426         * tree.c (integer_zerop, integer_onep, integer_all_onesp, real_zerop)
4427         (real_onep, real_minus_onep, add_expr, initializer_zerop): Likewise.
4428         (uniform_vector_p): Likewise.
4429         * varasm.c (const_hash_1, compare_constant): Likewise.
4430         * tree-ssa-ccp.c: Include tree-vector-builder.h.
4431         (valid_lattice_transition): Operate directly on the VECTOR_CST
4432         encoding.
4433         * ipa-icf.c: Include tree-vector-builder.h.
4434         (sem_variable::equals): Operate directly on the VECTOR_CST encoding.
4435         * print-tree.c (print_node): Print encoding of VECTOR_CSTs.
4437 2017-12-07  Richard Sandiford  <richard.sandiford@linaro.org>
4439         * tree.c (build_vector): Delete.
4440         * tree.h (build_vector): Make static and move into the self-testing
4441         block.
4443 2017-12-07  Richard Sandiford  <richard.sandiford@linaro.org>
4445         * vector-builder.h (vector_builder::derived): New const overload.
4446         (vector_builder::elt): New function.
4447         * tree-vector-builder.h (tree_vector_builder::type): New function.
4448         (tree_vector_builder::apply_step): Declare.
4449         * tree-vector-builder.c (tree_vector_builder::apply_step): New
4450         function.
4451         * gimple-fold.h (tree_vector_builder): Declare.
4452         (gimple_build_vector): Take a tree_vector_builder instead of a
4453         type and vector of elements.
4454         * gimple-fold.c (gimple_build_vector): Likewise.
4455         * tree-vect-loop.c (get_initial_def_for_reduction): Update call
4456         accordingly.
4457         (get_initial_defs_for_reduction): Likewise.
4458         (vectorizable_induction): Likewise.
4460 2017-12-07  Richard Sandiford  <richard.sandiford@linaro.org>
4462         * tree-vector-builder.h
4463         (tree_vector_builder::new_binary_operation): Declare.
4464         * tree-vector-builder.c
4465         (tree_vector_builder::new_binary_operation): New function.
4466         * fold-const.c (fold_relational_const): Use it.
4467         (const_binop): Likewise.  Check that both input vectors have
4468         the same number of elements, thus excluding things like WIDEN_SUM.
4469         Check whether it is possible to operate directly on the encodings
4470         of stepped inputs.
4472 2017-12-07  Richard Sandiford  <richard.sandiford@linaro.org>
4474         * fold-const.c (fold_negate_expr_1): Use tree_vector_builder and
4475         new_unary_operation, operating only on the encoded elements.
4476         (const_unop): Likewise.
4477         (exact_inverse): Likewise.
4478         (distributes_over_addition_p): New function.
4479         (const_binop): Use tree_vector_builder and new_unary_operation
4480         for combinations of VECTOR_CST and INTEGER_CST.  Operate only
4481         on the encoded elements unless the encoding is strided and the
4482         operation does not distribute over addition.
4483         (fold_convert_const):  Use tree_vector_builder and
4484         new_unary_operation.  Operate only on the encoded elements
4485         for truncating integer conversions, or for non-stepped encodings.
4487 2017-12-07  Richard Sandiford  <richard.sandiford@linaro.org>
4489         * config/sparc/sparc.c: Include tree-vector-builder.h.
4490         (sparc_fold_builtin): Use tree_vector_builder instead of build_vector.
4491         * expmed.c: Include tree-vector-builder.h.
4492         (make_tree): Use tree_vector_builder instead of build_vector.
4493         * fold-const.c: Include tree-vector-builder.h.
4494         (const_binop): Use tree_vector_builder instead of build_vector.
4495         (const_unop): Likewise.
4496         (native_interpret_vector): Likewise.
4497         (fold_vec_perm): Likewise.
4498         (fold_ternary_loc): Likewise.
4499         * gimple-fold.c: Include tree-vector-builder.h.
4500         (gimple_fold_stmt_to_constant_1): Use tree_vector_builder instead
4501         of build_vector.
4502         * tree-ssa-forwprop.c: Include tree-vector-builder.h.
4503         (simplify_vector_constructor): Use tree_vector_builder instead
4504         of build_vector.
4505         * tree-vect-generic.c: Include tree-vector-builder.h.
4506         (add_rshift): Use tree_vector_builder instead of build_vector.
4507         (expand_vector_divmod): Likewise.
4508         (optimize_vector_constructor): Likewise.
4509         * tree-vect-loop.c: Include tree-vector-builder.h.
4510         (vect_create_epilog_for_reduction): Use tree_vector_builder instead
4511         of build_vector.  Explicitly use a stepped encoding for
4512         { 1, 2, 3, ... }.
4513         * tree-vect-slp.c: Include tree-vector-builder.h.
4514         (vect_get_constant_vectors): Use tree_vector_builder instead
4515         of build_vector.
4516         (vect_transform_slp_perm_load): Likewise.
4517         (vect_schedule_slp_instance): Likewise.
4518         * tree-vect-stmts.c: Include tree-vector-builder.h.
4519         (vectorizable_bswap): Use tree_vector_builder instead of build_vector.
4520         (vect_gen_perm_mask_any): Likewise.
4521         (vectorizable_call): Likewise.  Explicitly use a stepped encoding.
4522         * tree.c: (build_vector_from_ctor): Use tree_vector_builder instead
4523         of build_vector.
4524         (build_vector_from_val): Likewise.  Explicitly use a duplicate
4525         encoding.
4527 2017-12-07  Richard Sandiford  <richard.sandiford@arm.com>
4529         * doc/generic.texi (VECTOR_CST): Describe new representation of
4530         vector constants.
4531         * vector-builder.h: New file.
4532         * tree-vector-builder.h: Likewise.
4533         * tree-vector-builder.c: Likewise.
4534         * Makefile.in (OBJS): Add tree-vector-builder.o.
4535         * tree.def (VECTOR_CST): Update comment to refer to generic.texi.
4536         * tree-core.h (tree_base): Add a vector_cst field to the u union.
4537         (tree_vector): Change the number of elements to
4538         vector_cst_encoded_nelts.
4539         * tree.h (VECTOR_CST_NELTS): Redefine using TYPE_VECTOR_SUBPARTS.
4540         (VECTOR_CST_ELTS): Delete.
4541         (VECTOR_CST_ELT): Redefine using vector_cst_elt.
4542         (VECTOR_CST_LOG2_NPATTERNS, VECTOR_CST_NPATTERNS): New macros.
4543         (VECTOR_CST_NELTS_PER_PATTERN, VECTOR_CST_DUPLICATE_P): Likewise.
4544         (VECTOR_CST_STEPPED_P, VECTOR_CST_ENCODED_ELTS): Likewise.
4545         (VECTOR_CST_ENCODED_ELT): Likewise.
4546         (vector_cst_encoded_nelts): New function.
4547         (make_vector): Take the values of VECTOR_CST_LOG2_NPATTERNS and
4548         VECTOR_CST_NELTS_PER_PATTERN as arguments.
4549         (vector_cst_int_elt, vector_cst_elt): Declare.
4550         * tree.c: Include tree-vector-builder.h.
4551         (tree_code_size): Abort if passed VECTOR_CST.
4552         (tree_size): Update for new VECTOR_CST layout.
4553         (make_vector): Take the values of VECTOR_CST_LOG2_NPATTERNS and
4554         VECTOR_CST_NELTS_PER_PATTERN as arguments.
4555         (build_vector): Use tree_vector_builder.
4556         (vector_cst_int_elt, vector_cst_elt): New functions.
4557         (drop_tree_overflow): For VECTOR_CST, drop the TREE_OVERFLOW from the
4558         encoded elements and then create the vector in the canonical form.
4559         (check_vector_cst, check_vector_cst_duplicate, check_vector_cst_fill)
4560         (check_vector_cst_stepped, test_vector_cst_patterns): New functions.
4561         (tree_c_tests): Call test_vector_cst_patterns.
4562         * lto-streamer-out.c (DFS::DFS_write_tree_body): Handle the new
4563         VECTOR_CST fields.
4564         (hash_tree): Likewise.
4565         * tree-streamer-out.c (write_ts_vector_tree_pointers): Likewise.
4566         (streamer_write_tree_header): Likewise.
4567         * tree-streamer-in.c (lto_input_ts_vector_tree_pointers): Likewise.
4568         (streamer_alloc_tree): Likewise.  Update call to make_vector.
4569         * fold-const.c (fold_ternary_loc): Avoid using VECTOR_CST_ELTS.
4571 2017-12-07  Richard Sandiford  <richard.sandiford@linaro.org>
4573         * selftest.h (ASSERT_TRUE_AT, ASSERT_FALSE_AT, ASSERT_EQ_AT)
4574         (ASSERT_NE, ASSERT_PRED1): Add underscores to local variable names
4575         * selftest-rtl.h (ASSERT_RTX_EQ, ASSERT_RTX_PTR_EQ): Likewise.
4577 2017-12-07  Bin Cheng  <bin.cheng@arm.com>
4578             Richard Biener  <rguenther@suse.de>
4580         PR tree-optimization/81303
4581         * Makefile.in (gimple-loop-interchange.o): New object file.
4582         * common.opt (floop-interchange): Reuse the option from graphite.
4583         * doc/invoke.texi (-floop-interchange): Ditto.  New document for
4584         -floop-interchange and mention it for -O3.
4585         * opts.c (default_options_table): Enable -floop-interchange at -O3.
4586         * gimple-loop-interchange.cc: New file.
4587         * params.def (PARAM_LOOP_INTERCHANGE_MAX_NUM_STMTS): New parameter.
4588         (PARAM_LOOP_INTERCHANGE_STRIDE_RATIO): New parameter.
4589         * passes.def (pass_linterchange): New pass.
4590         * timevar.def (TV_LINTERCHANGE): New time var.
4591         * tree-pass.h (make_pass_linterchange): New declaration.
4592         * tree-ssa-loop-ivcanon.c (create_canonical_iv): Change to external
4593         interchange.  Record IV before/after increment in new parameters.
4594         * tree-ssa-loop-ivopts.h (create_canonical_iv): New declaration.
4595         * tree-vect-loop.c (vect_is_simple_reduction): Factor out reduction
4596         path check into...
4597         (check_reduction_path): ...New function here.
4598         * tree-vectorizer.h (check_reduction_path): New declaration.
4600 2017-12-07  Vladimir Makarov  <vmakarov@redhat.com>
4602         PR target/83252
4603         PR rtl-optimization/80818
4604         * lra.c (add_regs_to_insn_regno_info): Make a hard reg in CLOBBER
4605         always early clobbered.
4606         * lra-lives.c (process_bb_lives): Check input hard regs for early
4607         clobbered non-operand hard reg.
4609 2017-12-07  Jakub Jelinek  <jakub@redhat.com>
4611         PR middle-end/83164
4612         * tree-cfg.c (verify_gimple_assign_binary): Don't require
4613         types_compatible_p, just that TYPE_MODE is the same.
4615 2017-12-07  Martin Sebor  <msebor@redhat.com>
4617         PR c/81544
4618         * attribs.c (empty_attribute_table): Initialize new member of
4619         struct attribute_spec.
4620         (decl_attributes): Add argument.  Handle mutually exclusive
4621         combinations of attributes.
4622         (selftests::test_attribute_exclusions): New function.
4623         (selftests::attribute_c_tests): Ditto.
4624         * attribs.h (decl_attributes): Add default argument.
4625         * selftest.h (attribute_c_tests): Declare.
4626         * selftest-run-tests.c (selftest::run_tests): Call attribute_c_tests.
4627         * tree-core.h (attribute_spec::exclusions, exclude): New type and
4628         member.
4629         * doc/extend.texi (Common Function Attributes): Update const and pure.
4630         * config/alpha/alpha.c (vms_attribute_table): Initialize new member
4631         of struct attribute_spec.
4632         * config/arc/arc.c (arc_attribute_table): Same.
4633         * config/arm/arm.c (arm_attribute_table): Same.
4634         * config/avr/avr.c ( avr_attribute_table): Same.
4635         * config/bfin/bfin.c (bfin_attribute_table): Same.
4636         * config/cr16/cr16.c (cr16_attribute_table): Same.
4637         * config/epiphany/epiphany.c (epiphany_attribute_table): Same.
4638         * config/h8300/h8300.c (h8300_attribute_table): Same.
4639         * config/i386/i386.c (ix86_attribute_table): Same.
4640         * config/ia64/ia64.c (ia64_attribute_table): Same.
4641         * config/m32c/m32c.c (m32c_attribute_table): Same.
4642         * config/m32r/m32r.c (m32r_attribute_table): Same.
4643         * config/m68k/m68k.c (m68k_attribute_table): Same.
4644         * config/mcore/mcore.c (mcore_attribute_table): Same.
4645         * config/microblaze/microblaze.c (microblaze_attribute_table): Same.
4646         * config/mips/mips.c (mips_attribute_table): Same.
4647         * config/msp430/msp430.c (msp430_attribute_table): Same.
4648         * config/nds32/nds32.c (nds32_attribute_table): Same.
4649         * config/nvptx/nvptx.c (nvptx_attribute_table): Same.
4650         * config/powerpcspe/powerpcspe.c (rs6000_attribute_table): Same.
4651         * config/rl78/rl78.c (rl78__attribute_table): Same.
4652         * config/rs6000/rs6000.c (rs6000_attribute_table): Same.
4653         * onfig/rx/rx.c (rx_attribute_table): Same.
4654         * config/s390/s390.c (s390_handle_vectorbool_attribute): Same.
4655         * config/sh/sh.c (sh_attribute_table): Same.
4656         * config/sparc/sparc.c (sparc_attribute_table): Same.
4657         * config/spu/spu.c (spu_attribute_table): Same.
4658         * config/stormy16/stormy16.c (xstormy16_attribute_table): Same.
4659         * config/v850/v850.c (v850_attribute_table): Same.
4660         * config/visium/visium.c (visium_attribute_table): Same.
4662 2017-12-07  Tamar Christina  <tamar.christina@arm.com>
4664         PR target/82641
4665         * config/arm/arm.c (INCLUDE_STRING): Define.
4666         (arm_last_printed_arch_string, arm_last_printed_fpu_string): New.
4667         (arm_declare_function_name): Conservatively emit .arch, .arch_extensions
4668         and .fpu.
4670 2017-12-07  Michael Matz  <matz@suse.de>
4672         Add unroll and jam pass
4674         * gimple-loop-jam.c: New file.
4675         * Makefile.in (OBJS): Add gimple-loop-jam.o.
4676         * common.opt (funroll-and-jam): New option.
4677         * opts.c (default_options_table): Add unroll-and-jam at -O3.
4678         * params.def (PARAM_UNROLL_JAM_MIN_PERCENT): New param.
4679         (PARAM_UNROLL_JAM_MAX_UNROLL): Ditto.
4680         * passes.def: Add pass_loop_jam.
4681         * timevar.def (TV_LOOP_JAM): Add.
4682         * tree-pass.h (make_pass_loop_jam): Declare.
4683         * cfgloop.c (flow_loop_tree_node_add): Add AT argument.
4684         * cfgloop.h (flow_loop_tree_node_add): Adjust declaration.
4685         * cfgloopmanip.c (duplicate_loop): Add AT argument, adjust call
4686         to flow_loop_tree_node_add.
4687         (duplicate_subloops, copy_loops_to): Append to sibling list.
4688         * cfgloopmanip.h: (duplicate_loop): Adjust declaration.
4689         * doc/invoke.texi (-funroll-and-jam): Document new option.
4690         (unroll-jam-min-percent, unroll-jam-max-unroll): Document new params.
4692 2017-12-07  Richard Biener  <rguenther@suse.de>
4694         PR tree-optimization/83296
4695         PR tree-optimization/67769
4696         * tree-ssa-phiopt.c (conditional_replacement): Do not reset
4697         flow sensitive info in an unrelated BB.
4698         (value_replacement): Use reset_flow_sensitive_info.
4699         (minmax_replacement): Reset flow sensitive info on the def
4700         we move.  Do not reset flow sensitive info in the whole BB
4701         we move the stmt to.
4702         (abs_replacement): Likewise.
4704 2017-12-07  Segher Boessenkool  <segher@kernel.crashing.org>
4706         PR target/43871
4707         * config/rs6000/rs6000.c (rs6000_option_override_internal): Set
4708         rs6000_cpu to the given -mcpu=, or to the default processor.
4710 2017-12-07  Segher Boessenkool  <segher@kernel.crashing.org>
4712         * config/rs6000/rs6000.h (rs6000_cpu_attr): Delete.
4713         * config/rs6000/rs6000.c (rs6000_variable_issue_1): Use rs6000_tune
4714         instead of rs6000_cpu_attr.
4715         (rs6000_adjust_cost): Ditto.
4716         (is_microcoded_insn): Ditto.
4717         (rs6000_adjust_priority): Ditto.
4718         (rs6000_issue_rate): Ditto.
4719         (rs6000_use_sched_lookahead): Ditto.
4720         (rs6000_use_sched_lookahead_guard): Ditto.
4721         (rs6000_sched_reorder): Ditto.
4722         (force_new_group): Ditto.
4723         * config/rs6000/rs6000.md (cpu attribute): Ditto.
4724         (group_ending_nop): Ditto.
4726 2017-12-07  Segher Boessenkool  <segher@kernel.crashing.org>
4728         * config/rs6000/rs6000.opt (rs6000_tune): New variable.
4729         * config/rs6000/rs6000.c (rs6000_option_override_internal): Also set
4730         rs6000_tune.  Use rs6000_tune instead of rs6000_cpu where appropriate.
4731         (rs6000_loop_align): Use rs6000_tune instead of rs6000_cpu where
4732         appropriate.
4733         (rs6000_reassociation_width): Ditto.
4734         (rs6000_emit_epilogue): Ditto.
4735         (rs6000_adjust_cost): Ditto.
4736         (is_microcoded_insn): Ditto.
4737         (is_cracked_insn): Ditto.
4738         (rs6000_adjust_priority): Ditto.
4739         (rs6000_sched_reorder): Ditto.
4740         (rs6000_sched_reorder2): Ditto.
4741         (insn_must_be_first_in_group): Ditto.
4742         (insn_must_be_last_in_group): Ditto.
4743         (rs6000_register_move_cost): Ditto.
4744         * config/rs6000/rs6000.h (rs6000_cpu_attr): Use rs6000_tune instead of
4745         rs6000_cpu.
4747 2017-12-07  Julia Koval  <julia.koval@intel.com>
4749         * config.gcc: Add vaesintrin.h.
4750         * config/i386/i386-builtin-types.def (V64QI_FTYPE_V64QI_V64QI):
4751         New type.
4752         * config/i386/i386-builtin.def (__builtin_ia32_vaesdec_v16qi,
4753         __builtin_ia32_vaesdec_v32qi, __builtin_ia32_vaesdec_v64qi):
4754         New builtins.
4755         * config/i386/i386.c (ix86_expand_args_builtin): Handle new type.
4756         * config/i386/immintrin.h: Include vaesintrin.h.
4757         * config/i386/sse.md (vaesdec_<mode>): New pattern.
4758         * config/i386/vaesintrin.h (_mm256_aesdec_epi128, _mm512_aesdec_epi128,
4759         _mm_aesdec_epi128): New intrinsics.
4761 2017-12-06  David Malcolm  <dmalcolm@redhat.com>
4763         * Makefile.in (C_COMMON_OBJS): Add c-family/c-spellcheck.o.
4764         * spellcheck-tree.c (find_closest_macro_cpp_cb): Move to
4765         c-family/c-spellcheck.cc.
4766         (best_macro_match::best_macro_match): Likewise.
4767         * spellcheck-tree.h
4768         (struct edit_distance_traits<cpp_hashnode *>): Move to
4769         c-family/c-spellcheck.h.
4770         (class best_macro_match): Likewise.
4772 2017-12-06  Jakub Jelinek  <jakub@redhat.com>
4774         PR tree-optimization/83293
4775         * gimple-ssa-strength-reduction.c (insert_initializers): Use
4776         GSI_NEW_STMT instead of GSI_SAME_STMT in gsi_insert_after that
4777         might insert into empty bb.
4779         PR sanitizer/81281
4780         * match.pd ((T)(P + A) - (T)P -> (T) A): Split into separate
4781         simplify for plus with :c added, and pointer_plus without that.
4782         ((T)P - (T)(P + A) -> -(T) A): Likewise.  If type is integral
4783         with undefined overflow and the conversion is not widening,
4784         perform negation in utype and only convert to type afterwards.
4785         ((T)(P + A) - (T)(P + B) -> (T)A - (T)B): Split into separate
4786         simplify for plus with :c added, and pointer_plus without that.
4787         If type is integral with undefined overflow and the conversion is
4788         not widening, perform minus in utype and only convert to type
4789         afterwards.  Move the last pointer_diff_expr simplify into the
4790         two outermost ifs.
4792 2017-12-06  Martin Sebor  <msebor@redhat.com>
4794         PR tree-optimization/82646
4795         * builtins.c (maybe_emit_chk_warning): Use size as the bound for
4796         strncpy, not maxlen.
4798 2017-12-06  Martin Sebor  <msebor@redhat.com>
4800         * doc/invoke.texi (-Wstringop-truncation): Mention attribute
4801         nonstring.
4803         PR tree-optimization/83075
4804         * tree-ssa-strlen.c (handle_builtin_stxncpy): Avoid assuming
4805         strncat/strncpy don't change length of source string.
4807 2017-12-06  Eric Botcazou  <ebotcazou@adacore.com>
4809         Revert
4810         2017-11-29  Martin Aberg  <maberg@gaisler.com>
4812         * config/sparc/sparc.md (divdf3_fix): Add NOP and adjust length
4813         to prevent b2bst errata sequence.
4814         (sqrtdf2_fix): Likewise.
4816 2017-12-06  Jakub Jelinek  <jakub@redhat.com>
4818         PR tree-optimization/81945
4819         * cfgloop.h (FOR_EACH_LOOP_FN): Use FN instead of hardcoding fn.
4820         * tree-cfg.c (move_sese_region_to_fn): If any of the loops moved
4821         to dest_cfun has orig_loop_num set, either remap it to the new
4822         loop number if the loop got moved too, or clear it.
4824 2017-12-05  Steve Ellcey  <sellcey@cavium.com>
4826         * config/aarch64/thunderx2-t99.md (thunderx2t99_branch): Add trap
4827         to reservation.
4828         (thunderx2t99_nothing): New insn reservation.
4829         (thunderx2t99_mrs): New insn reservation.
4830         (thunderx2t99_multiple): New insn reservation.
4831         (thunderx2t99_alu_basi): Add bfx to reservation.
4832         (thunderx2t99_fp_cmp): Add fccmps and fccmpd to reservation.
4834 2017-12-05  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
4836         PR target/82248
4837         * config/arm/arm.md (probe_stack) : Use the 'o' constraint.
4839 2017-12-05  Bin Cheng  <bin.cheng@arm.com>
4841         * tree-ssa-dce.c (simple_dce_from_worklist): Move and rename from
4842         tree-ssa-pre.c::remove_dead_inserted_code.
4843         * tree-ssa-dce.h: New file.
4844         * tree-ssa-pre.c (tree-ssa-dce.h): Include new header file.
4845         (remove_dead_inserted_code): Move and rename to function
4846         tree-ssa-dce.c::simple_dce_from_worklist.
4847         (pass_pre::execute): Update use.
4849 2017-12-05  Richard Biener  <rguenther@suse.de>
4851         PR tree-optimization/83277
4852         * graphite-isl-ast-to-gimple.c (should_copy_to_new_region): Make sure
4853         to code-gen liveout vars.
4855 2017-12-05  Richard Sandiford  <richard.sandiford@linaro.org>
4857         * config/aarch64/aarch64-simd.md (aarch64_simd_bsldi_internal)
4858         (aarch64_simd_bsldi_alt): Check REG_P before GP_REGNUM_P.
4859         (aarch64_cm<optab>di, aarch64_cmtstdi): Add leading "&&" to
4860         split condition.
4862 2017-12-05  Max Filippov  <jcmvbkbc@gmail.com>
4864         * config/xtensa/xtensa.c (xtensa_asan_shadow_offset): New
4865         function.
4866         (TARGET_ASAN_SHADOW_OFFSET): New macro definition.
4867         * config/xtensa/xtensa.h (FRAME_GROWS_DOWNWARD): Set to 1 if
4868         ASAN is enabled.
4870 2017-12-05  Richard Biener  <rguenther@suse.de>
4872         * timevar.def (TV_TREE_RECIP, TV_TREE_SINCOS, TV_TREE_WIDEN_MUL): Add.
4873         * tree-ssa-math-opts.c (pass_data_cse_reciprocal): Use TV_TREE_RECIP.
4874         (pass_data_cse_sincos): Use TV_TREE_SINCOS.
4875         (pass_data_optimize_widening_mul): Use TV_TREE_WIDEN_MUL.
4877 2017-12-05  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
4879         * dbxout.c (dbxout_block): Grow buf to 30 bytes.
4881 2017-12-05  Martin Liska  <mliska@suse.cz>
4882             Jakub Jelinek  <jakub@redhat.com>
4884         * doc/invoke.texi: Document the options.
4885         * flag-types.h (enum sanitize_code): Add
4886         SANITIZE_POINTER_COMPARE and SANITIZE_POINTER_SUBTRACT.
4887         * ipa-inline.c (sanitize_attrs_match_for_inline_p): Add handling
4888         of SANITIZE_POINTER_COMPARE and SANITIZE_POINTER_SUBTRACT.
4889         * opts.c: Define new sanitizer options.
4890         * sanitizer.def (BUILT_IN_ASAN_POINTER_COMPARE): Likewise.
4891         (BUILT_IN_ASAN_POINTER_SUBTRACT): Likewise.
4893 2017-12-05  Julia Koval  <julia.koval@intel.com>
4895         * common/config/i386/i386-common.c (OPTION_MASK_ISA_AVX512VNNI_SET,
4896         OPTION_MASK_ISA_AVX512VNNI_UNSET): New.
4897         (ix86_handle_option): Handle -mavx512vnni.
4898         * config/i386/cpuid.h (bit_AVX512VNNI): New bit.
4899         * config/i386/driver-i386.c (host_detect_local_cpu): Handle new bit.
4900         * config/i386/i386-c (__AVX512VNNI__): New.
4901         * config/i386/i386.c (ix86_target_string): Handle new option.
4902         (ix86_valid_target_attribute_inner_p): Handle new option.
4903         * config/i386/i386.h (TARGET_AVX512VNNI, TARGET_AVX512VNNI_P): New.
4904         * config/i386/i386.opt (mavx512vnni): New option.
4906 2017-12-01  Jan Hubicka  <hubicka@ucw.cz>
4908         PR target/81616
4909         * athlon.md: Disable for generic.
4910         * haswell.md: Enable for generic.
4911         * i386.c (ix86_sched_init_global): Add core hooks for generic.
4912         * x86-tune-sched.c (ix86_issue_rate): Increase issue rate for generic
4913         to 4.
4914         (ix86_adjust_cost): Move generic to haswell path.
4916 2017-12-04  Eric Botcazou  <ebotcazou@adacore.com>
4918         * config/sparc/sparc.c (sparc_do_work_around_errata): Use mem_ref
4919         instead of MEM_P in a couple more places.  Fix formatting issues.
4921 2017-12-04  Jim Wilson  <jimw@sifive.com>
4923         * config/riscv/riscv.c (riscv_for_each_saved_reg): Use GP_REG_LAST
4924         instead of GP_REG_LAST-1.
4925         (riscv_adjust_libcall_cfi_prologue): Likewise.
4926         (riscv_adjust_libcall_cri_epilogue): Likewise.
4927         * config/riscv/riscv.h (CALL_USED_REGISTERS): Change a6 to t6 in
4928         comment.
4930 2017-12-04  Luis Machado  <luis.machado@linaro.org>
4932         * ipa-pure-const.c (check_decl): Add missing newline.
4933         (state_from_flags): Likewise.
4935 2017-12-04  Jeff Law  <law@redhat.com>
4937         PR tree-optimizatin/78496
4938         * gimple-ssa-evrp-analyze.h
4939         (evrp_range_analyzer::get_vr_values): Simplify.
4940         * gimple-ssa-evrp-analyze.c: Corresponding changes.
4941         * tree-ssa-dom.c: Include alloc-pool.h, tree-vrp.h, vr-values.h
4942         and gimple-ssa-evrp-analyze.h.
4943         (dom_opt_dom_walker class): Add evrp_range_analyzer member.
4944         (simplify_stmt_for_jump_threading): Copy a blob of code from
4945         tree-vrp.c to use ranges to simplify statements.
4946         (dom_opt_dom_walker::before_dom_children): Call
4947         evrp_range_analyzer::{enter,record_ranges_from_stmt} methods.
4948         (dom_opt_dom_walker::after_dom_children): Similarly for
4949         evrp_range_analyzer::leave.
4950         (dom_opt_dom_walker::optimize_stmt): Use EVRP ranges to optimize
4951         conditionals.
4953         * gimple-ssa-evrp-analyze.c
4954         (evrp_range_analyzer::extract_range_from_stmt):  Always use
4955         vr_values::update_value_range so preexisting range info is
4956         medged with new range info, even if the new range is VR_VARYING.
4958 2017-12-04  Segher Boessenkool  <segher@kernel.crashing.org>
4960         * combine.c: Adjust comment.
4961         (use_crosses_set_p): Delete.
4962         (can_combine_p): Use modified_between_p instead of use_crosses_set_p.
4963         (try_combine): Ditto.
4965 2017-12-04  Richard Biener  <rguenther@suse.de>
4967         PR tree-optimization/83255
4968         * graphite-isl-ast-to-gimple.c (translate_isl_ast_node_for):
4969         Re-add zero-iteration check.
4971 2017-12-04  Segher Boessenkool  <segher@kernel.crashing.org>
4973         PR rtl-optimization/83245
4974         * lra.c (collect_non_operand_hard_regs): Treat clobbers of non-operand
4975         hard registers as earlyclobber, also if not in an asm.
4977 2017-12-04  Segher Boessenkool  <segher@kernel.crashing.org>
4979         PR bootstrap/83265
4980         Revert
4981         2017-12-01  Segher Boessenkool  <segher@kernel.crashing.org>
4983         PR target/43871
4984         * config/rs6000/rs6000.c (rs6000_option_override_internal): Set
4985         rs6000_cpu based on cpu_index, not tune_index.
4987 2017-12-04  Richard Biener  <rguenther@suse.de>
4989         PR tree-optimization/83238
4990         * graphite-scop-detection.c (scop_detection::merge_sese): Make
4991         code match comment, rejecting invalid SESE regions.
4993 2017-12-03  John David Anglin  <danglin@gcc.gnu.org>
4995         * config/pa/pa.c (pa_legitimate_address_p): For scaled indexing,
4996         require base operand is a REG_POINTER prior to reload on targets
4997         with non-equivalent space registers.
4999 2017-12-01  Jan Hubicka  <hubicka@ucw.cz>
5001         * ipa-cp.c (ipcp_lattice<valtype>::print): Update dumping.
5002         (update_specialized_profile): Fix updating of counts.
5003         (perhaps_add_new_callers): Likewise.
5005 2017-12-01  Jan Hubicka  <hubicka@ucw.cz>
5007         PR target/81616
5008         * x86-tune.def: Remove obsolette FIXMEs.
5009         (X86_TUNE_PARTIAL_FLAG_REG_STALL): Disable for generic
5010         (X86_TUNE_FUSE_CMP_AND_BRANCH_32, X86_TUNE_FUSE_CMP_AND_BRANCH_64,
5011         X86_TUNE_FUSE_CMP_AND_BRANCH_SOFLAGS, X86_TUNE_FUSE_ALU_AND_BRANCH):
5012         Enable for generic.
5013         (X86_TUNE_PAD_RETURNS): Disable for generic.
5015 2017-12-02  Jakub Jelinek  <jakub@redhat.com>
5017         PR tree-optimization/83170
5018         PR tree-optimization/83241
5019         * gimple-ssa-store-merging.c
5020         (imm_store_chain_info::try_coalesce_bswap): Update vuse field from
5021         gimple_vuse (ins_stmt) in case it has changed.
5022         (imm_store_chain_info::output_merged_store): Likewise.
5024         * tree-chkp.c (chkp_compute_bounds_for_assignment): Handle
5025         POINTER_DIFF_EXPR.
5027         PR c++/81212
5028         * tree-cfg.c (pass_warn_function_return::execute): Handle
5029         __builtin_ubsan_handle_missing_return like __builtin_unreachable
5030         with BUILTINS_LOCATION.
5032         PR target/78643
5033         PR target/80583
5034         * expr.c (get_inner_reference): If DECL_MODE of a non-bitfield
5035         is BLKmode for vector field with vector raw mode, use TYPE_MODE
5036         instead of DECL_MODE.
5038         * config/i386/i386-protos.h (standard_sse_constant_opcode): Change
5039         last argument to rtx pointer.
5040         * config/i386/i386.c (standard_sse_constant_opcode): Replace X argument
5041         with OPERANDS.  For AVX+ 128-bit VEX encoded instructions over 256-bit
5042         or 512-bit.  If setting EXT_REX_SSE_REG_P, use EVEX encoded insn
5043         depending on the chosen ISAs.
5044         * config/i386/i386.md (*movxi_internal_avx512f, *movoi_internal_avx,
5045         *movti_internal, *movdi_internal, *movsi_internal, *movtf_internal,
5046         *movdf_internal, *movsf_internal): Adjust standard_sse_constant_opcode
5047         callers.
5048         * config/i386/sse.md (mov<mode>_internal): Likewise.
5049         * config/i386/mmx.md (*mov<mode>_internal): Likewise.
5051 2017-12-01  Segher Boessenkool  <segher@kernel.crashing.org>
5053         * doc/invoke.texi (-dp): Say that instruction cost is printed as well.
5055 2017-12-01  Segher Boessenkool  <segher@kernel.crashing.org>
5057         * config/rs6000/rs6000-protos.h (emit_fusion_addis): Remove last two
5058         parameters from prototype.
5059         * config/rs6000/rs6000.c (emit_fusion_addis): Remove last two
5060         parameters.  Don't print a comment.
5061         (emit_fusion_gpr_load): Adjust.
5062         (emit_fusion_load_store): Adjust.
5063         * config/rs6000/rs6000.md (*fusion_p9_<mode>_constant): Adjust.
5064         * config/rs6000/vsx.md (two peepholes): Print the "vector load fusion"
5065         comment on the second line.
5067 2017-12-01  Segher Boessenkool  <segher@kernel.crashing.org>
5069         PR target/43871
5070         * config/rs6000/rs6000.c (rs6000_option_override_internal): Set
5071         rs6000_cpu based on cpu_index, not tune_index.
5073 2017-12-01  Segher Boessenkool  <segher@kernel.crashing.org>
5075         * final.c (output_asm_name): Print insn_cost.  Shorten output.  Print
5076         which_alternative instead of which_alternative + 1.
5077         (output_asm_insn): Print an extra tab if the template is short.
5079 2017-12-01  Jim Wilson  <jimw@sifive.com>
5081         * common.opt (use_gnu_debug_info_extensions): Delete DWARF_DEBUG from
5082         comment.
5083         * config/vx-common.h (DWARF_DEBUGGING_INFO): Delete undef.
5084         * doc/tm.texi.in (PREFERRED_DEBUGGING_TYPE): Delete DWARF_DEBUG
5085         reference.
5086         * doc/tm.texi: Regenerate.
5088 2017-12-01  Michael Meissner  <meissner@linux.vnet.ibm.com>
5090         PR target/81959
5091         * config/rs6000/rs6000.c (rs6000_address_for_fpconvert): Check for
5092         whether we can allocate pseudos before trying to fix an address.
5093         * config/rs6000/rs6000.md (float_<mode>si2_hw): Make sure the
5094         memory address is indexed or indirect.
5095         (floatuns_<mode>si2_hw2): Likewise.
5097 2017-12-01  Jason Merrill  <jason@redhat.com>
5099         * Makefile.in (TAGS): Add c-family/*.cc.
5101 2017-12-01  Wilco Dijkstra  <wdijkstr@arm.com>
5103         * config/aarch64/aarch64.md (call_insn): Use %c rather than %a.
5104         (call_value_insn): Likewise.
5105         (sibcall_insn): Likewise.
5106         (sibcall_value_insn): Likewise.
5107         (movsi_aarch64): Likewise.
5108         (movdi_aarch64): Likewise.
5109         (add_losym_): Likewise.
5110         (ldr_got_small_): Likewise.
5111         (ldr_got_small_sidi): Likewise.
5112         (ldr_got_small_28k_): Likewise.
5113         (ldr_got_small_28k_sidi): Likewise.
5114         * config/aarch64/aarch64.c (aarch64_print_address_internal):
5115         Move output_addr_const to symbolic case. Add error check.
5117 2017-12-01  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
5119         * config/s390/predicates.md (plus16_Q_operand): New predicate.
5120         * config/s390/s390.md: Disable MVC merging peephole if it would
5121         disable operand forwarding.
5122         (new peephole2): Split MVCs if it would turn them into up to 2
5123         forwardable MVCs.
5125 2017-12-01  Richard Biener  <rguenther@suse.de>
5127         PR tree-optimization/83232
5128         * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Fix
5129         detection of same access. Instead of breaking the group here
5130         do not consider the duplicate.  Add comment explaining real fix.
5132 2017-12-01  Jonathan Wakely  <jwakely@redhat.com>
5134         * doc/md.texi (Insn Splitting): Fix "central flowgraph" typo.
5136 2017-12-01  Sudakshina Das  <sudi.das@arm.com>
5138         * config/arm/vfp.md (*movhf_vfp_fp16): Add conds attribute.
5140 2017-12-01  Jakub Jelinek  <jakub@redhat.com>
5142         * function.h (struct function): Remove cilk_frame_decl,
5143         is_cilk_function and calls_cilk_spawn fields.
5144         * tree-inline.h (struct copy_body_data): Remove remap_var_for_cilk
5145         field.
5146         * omp-simd-clone.c (simd_clone_clauses_extract): Don't clear
5147         cilk_elemental field.
5148         * cgraph.h (struct cgraph_simd_clone): Remove cilk_elemental field.
5149         * target.def: Adjust comment.
5150         * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
5151         Don't test cilk_elemental.
5153         PR tree-optimization/83233
5154         * gimple-ssa-store-merging.c (nop_stats, bswap_stats): Use
5155         bswap_stat name for the struct.
5157         PR c/79153
5158         * tree.h (SWITCH_BREAK_LABEL_P): Define.
5159         * gimplify.c (collect_fallthrough_labels): Handle GIMPLE_BIND
5160         starting with a GIMPLE_SWITCH and ending with GIMPLE_LABEL with
5161         SWITCH_BREAK_LABEL_P set on the label.
5162         (gimplify_switch_expr): Set SWITCH_BREAK_LABEL_P on the label
5163         added for default case if it was missing and not all cases covered.
5164         Wrap GIMPLE_SWITCH and the switch_body_seq into a GIMPLE_BIND if
5165         switch_body_seq ends with a GIMPLE_LABEL with SWITCH_BREAK_LABEL_P
5166         set on the label.
5167         * tree-chrec.c (evolution_function_is_univariate_p): Add return true;
5168         to avoid -Wimplicit-fallthrough warning.
5169         * config/i386/i386.c (ix86_expand_special_args_builtin): Add
5170         FALLTHRU comment to avoid -Wimplicit-fallthrough warning.
5172         PR tree-optimization/83221
5173         * tree-ssa-reassoc.c (sort_by_operand_rank): Shift bb_rank
5174         down by 16.
5175         (init_reassoc): Formatting fix.
5177         PR sanitizer/81275
5178         * tree-cfg.c (group_case_labels_stmt): Don't optimize away
5179         C++ FE implicitly added __builtin_unreachable () until -Wreturn-type
5180         is diagnosed.
5182         PR sanitizer/83219
5183         * tree-cfg.c: Include asan.h.
5184         (gimple_seq_unreachable_p): Return false for -fsanitize=unreachable.
5186 2017-12-01  Sergey Shalnov  <Sergey.Shalnov@intel.com>
5188         * config/i386/i386.md: Fix AVX512 register width in AVX512 instruction.
5190 2017-12-01  Sergey Shalnov  <Sergey.Shalnov@intel.com>
5192         * config/i386/i386.c (standard_sse_constant_opcode): Fix wrong form for
5193         vpcmpeqd instruction.
5195 2017-12-01  Sergey Shalnov  <Sergey.Shalnov@intel.com>
5197         * config/i386/i386.c (standard_sse_constant_opcode): Fix
5198         registers type for 128bit mode.
5200 2017-11-30  Jakub Jelinek  <jakub@redhat.com>
5202         * spellcheck-tree.c (test_find_closest_identifier): Use ; instead
5203         of ;;.
5204         * gengtype-state.c (read_state_pair): Likewise.
5205         * gimple-fold.c (gimple_fold_builtin_string_compare): Likewise.
5206         * sel-sched-dump.c (dump_insn_rtx_1): Likewise.
5207         * ipa-cp.c (intersect_aggregates_with_edge): Likewise.
5208         * ifcvt.c (noce_try_store_flag_constants): Likewise.
5209         * tree-ssa-ccp.c (ccp_finalize): Likewise.
5210         * omp-grid.c (grid_process_kernel_body_copy): Likewise.
5211         * builtins.c (fold_builtin_3): Likewise.
5212         * graphite-scop-detection.c
5213         (scop_detection::stmt_has_simple_data_refs_p): Likewise.
5214         * hsa-gen.c (hsa_function_representation::hsa_function_representation):
5215         Likewise.
5217 2017-12-01  Maxim Ostapenko  <m.ostapenko@samsung.com>
5219         PR sanitizer/81697
5220         * asan.c (asan_protect_global): Add new ignore_decl_rtl_set_p
5221         parameter. Return true if ignore_decl_rtl_set_p is true and other
5222         conditions are satisfied.
5223         * asan.h (asan_protect_global): Add new parameter.
5224         * varasm.c (categorize_decl_for_section): Pass true as second parameter
5225         to asan_protect_global calls.
5227 2017-11-30  Jim Wilson  <jimw@sifive.com>
5229         * doc/invoke.texi (RISC-V Options): Delete nonexistent -mmemcpy and
5230         -mno-memcpy options.  For -mplt, -mfdiv, -mdiv, -msave-restore, and
5231         -mstrict-align, add info on default value.  Delete redundant lines for
5232         -mabi.  Add missing -mexplicit-relocs docs.
5234 2017-11-30  Claudiu Zissulescu  <claziss@synopsys.com>
5236         * config/arc/arc.md (trap): New pattern.
5238 2017-11-30  Claudiu Zissulescu  <claziss@synopsys.com>
5240         * config/arc/arc.c (hwloop_optimize): Prevent the last ZOL
5241         instruction to end into a delay slot.
5242         * config/arc/arc.md (cond_delay_insn): Check if the instruction
5243         can be placed into a delay slot against reg_note.
5245 2017-11-30  Claudiu Zissulescu  <claziss@synopsys.com>
5247         * config/arc/arc.c (hwloop_optimize): Update hw-loop's end/start
5248         labels number of usages.
5250 2017-11-30  Claudiu Zissulescu  <claziss@synopsys.com>
5252         * config/arc/arc.c (arc_cannot_substitue_mem_equiv_p): New
5253         function.
5254         (TARGET_CANNOT_SUBSTITUTE_MEM_EQUIV_P): Define.
5256 2017-11-30  Jakub Jelinek  <jakub@redhat.com>
5258         PR target/83210
5259         * internal-fn.c (expand_mul_overflow): Optimize unsigned
5260         multiplication by power of 2 constant into two shifts + comparison.
5262 2017-11-30  Jan Hubicka  <hubicka@ucw.cz>
5264         PR target/81616
5265         * config/i386/x86-tune-costs.h (generic_cost): Revise for modern CPUs.
5267 2017-11-30  Richard Biener  <rguenther@suse.de>
5269         PR tree-optimization/83202
5270         * tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Add
5271         allow_peel argument and guard peeling.
5272         (canonicalize_loop_induction_variables): Likewise.
5273         (canonicalize_induction_variables): Pass false.
5274         (tree_unroll_loops_completely_1): Pass unroll_outer to disallow
5275         peeling from cunrolli.
5277 2017-11-29  Segher Boessenkool  <segher@kernel.crashing.org>
5279         * combine.c (try_combine): Print a message to dump file whenever
5280         I0, I1, or I2 cannot be combined into I3.
5282 2017-11-29  Segher Boessenkool  <segher@kernel.crashing.org>
5284         PR rtl-optimization/83156
5285         PR rtl-optimization/82621
5286         * combine.c (try_combine): Don't split an I2 if one of the dests is
5287         set again before I3.  Allow unused dests.
5289 2017-11-29  Segher Boessenkool  <segher@kernel.crashing.org>
5291         * config/rs6000/rs6000.md (*add<mode>3_carry_in_internal2): New.
5293 2017-11-29  Vladimir Makarov  <vmakarov@redhat.com>
5295         PR rtl-optimization/80818
5296         * lra.c (collect_non_operand_hard_regs): New arg insn.  Pass it
5297         recursively.  Use insn code for clobber.
5298         (lra_set_insn_recog_data): Pass the new arg to
5299         collect_non_operand_hard_regs.
5300         (add_regs_to_insn_regno_info): Pass insn instead of uid.  Use insn
5301         code for clobber.
5302         (lra_update_insn_regno_info): Pass insn to
5303         add_regs_to_insn_regno_info.
5305 2017-11-29  Jim Wilson  <jimw@sifive.com>
5306             Andrew Waterman  <andrew@sifive.com>
5308         * config/riscv/riscv.c (SINGLE_SHIFT_COST): New.
5309         (riscv_rtx_costs): Case ZERO_EXTRACT, match new pattern, and return
5310         SINGLE_SHIFT_COST.  Case LT and ZERO_EXTEND, likewise.  Case ASHIFT,
5311         use SINGLE_SHIFT_COST.
5312         * config/riscv/riscv.md (lshrsi3_zero_extend_1): New.
5313         (lshrsi3_zero_extend_2, lshrsi3_zero_extend_3): New.
5315 2017-11-29  Julia Koval  <julia.koval@intel.com>
5317         * config/i386/avx512vbmi2intrin.h (_mm512_shldv_epi16,
5318         _mm512_mask_shldv_epi16, _mm512_maskz_shldv_epi16, _mm512_shldv_epi32,
5319         _mm512_mask_shldv_epi32, _mm512_maskz_shldv_epi32, _mm512_shldv_epi64,
5320         _mm512_mask_shldv_epi64, _mm512_maskz_shldv_epi64): New intrinsics.
5321         * config/i386/avx512vbmi2vlintrin.h (_mm256_shldv_epi16,
5322         _mm256_mask_shldv_epi16, _mm256_maskz_shldv_epi16, _mm256_shldv_epi32,
5323         _mm256_mask_shldv_epi32, _mm256_maskz_shldv_epi32, _mm256_shldv_epi64,
5324         _mm256_mask_shldv_epi64, _mm256_maskz_shldv_epi64, _mm_shldv_epi16,
5325         _mm_mask_shldv_epi16, _mm_maskz_shldv_epi16, _mm_shldv_epi32,
5326         _mm_mask_shldv_epi32, _mm_maskz_shldv_epi32, _mm_shldv_epi64,
5327         _mm_mask_shldv_epi64, _mm_maskz_shldv_epi64): Ditto.
5328         * config/i386/i386-builtin.def (__builtin_ia32_vpshldv_v32hi,
5329         __builtin_ia32_vpshldv_v32hi_mask, __builtin_ia32_vpshldv_v32hi_maskz,
5330         __builtin_ia32_vpshldv_v16hi, __builtin_ia32_vpshldv_v16hi_mask,
5331         __builtin_ia32_vpshldv_v16hi_maskz, __builtin_ia32_vpshldv_v8hi,
5332         __builtin_ia32_vpshldv_v8hi_mask, __builtin_ia32_vpshldv_v8hi_maskz,
5333         __builtin_ia32_vpshldv_v16si, __builtin_ia32_vpshldv_v16si_mask,
5334         __builtin_ia32_vpshldv_v16si_maskz, __builtin_ia32_vpshldv_v8si,
5335         __builtin_ia32_vpshldv_v8si_mask, __builtin_ia32_vpshldv_v8si_maskz,
5336         __builtin_ia32_vpshldv_v4si, __builtin_ia32_vpshldv_v4si_mask,
5337         __builtin_ia32_vpshldv_v4si_maskz, __builtin_ia32_vpshldv_v8di,
5338         __builtin_ia32_vpshldv_v8di_mask, __builtin_ia32_vpshldv_v8di_maskz,
5339         __builtin_ia32_vpshldv_v4di, __builtin_ia32_vpshldv_v4di_mask,
5340         __builtin_ia32_vpshldv_v4di_maskz, __builtin_ia32_vpshldv_v2di,
5341         __builtin_ia32_vpshldv_v2di_mask,
5342         __builtin_ia32_vpshldv_v2di_maskz): New builtins.
5343         * config/i386/sse.md (vpshldv_<mode>, vpshldv_<mode>_mask,
5344         vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): New patterns.
5346 2017-11-29  Julia Koval  <julia.koval@intel.com>
5348         * config/i386/avx512vbmi2intrin.h (_mm512_shrdv_epi16,
5349         _mm512_mask_shrdv_epi16, _mm512_maskz_shrdv_epi16, _mm512_shrdv_epi32,
5350         _mm512_mask_shrdv_epi32, _mm512_maskz_shrdv_epi32, _mm512_shrdv_epi64,
5351         _mm512_mask_shrdv_epi64, _mm512_maskz_shrdv_epi64): New intrinsics.
5352         * config/i386/avx512vbmi2vlintrin.h (_mm256_shrdv_epi16,
5353         _mm256_mask_shrdv_epi16, _mm256_maskz_shrdv_epi16, _mm256_shrdv_epi32,
5354         _mm256_mask_shrdv_epi32, _mm256_maskz_shrdv_epi32, _mm256_shrdv_epi64,
5355         _mm256_mask_shrdv_epi64, _mm256_maskz_shrdv_epi64, _mm_shrdv_epi16,
5356         _mm_mask_shrdv_epi16, _mm_maskz_shrdv_epi16, _mm_shrdv_epi32,
5357         _mm_mask_shrdv_epi32, _mm_maskz_shrdv_epi32, _mm_shrdv_epi64,
5358         _mm_mask_shrdv_epi64, _mm_maskz_shrdv_epi64): Ditto.
5359         * config/i386/i386-builtin-types.def (V32HI_FTYPE_V32HI_V32HI_V32HI,
5360         V32HI_FTYPE_V32HI_V32HI_V32HI_INT, V16HI_FTYPE_V16HI_V16HI_V16HI_INT,
5361         V8HI_FTYPE_V8HI_V8HI_V8HI_INT, V8SI_FTYPE_V8SI_V8SI_V8SI_INT,
5362         V4SI_FTYPE_V4SI_V4SI_V4SI_INT, V8DI_FTYPE_V8DI_V8DI_V8DI,
5363         V8DI_FTYPE_V8DI_V8DI_V8DI_INT, V4DI_FTYPE_V4DI_V4DI_V4DI_INT,
5364         V16SI_FTYPE_V16SI_V16SI_V16SI, V16SI_FTYPE_V16SI_V16SI_V16SI_INT,
5365         V2DI_FTYPE_V2DI_V2DI_V2DI_INT): New types.
5366         * config/i386/i386.c (ix86_expand_args_builtin): Handle new types.
5367         * config/i386/sse.md (vpshrdv_<mode>, vpshrdv_<mode>_mask,
5368         vpshrdv_<mode>_maskz, vpshrdv_<mode>_maskz_1): New pattern.
5370 2017-11-29  Daniel Cederman  <cederman@gaisler.com>
5372         * config/sparc/sparc.c (sparc_do_work_around_errata): Treat the
5373         movsi_pic_gotdata_op instruction as a load for the UT699 errata
5374         workaround.
5376 2017-11-29  Martin Aberg  <maberg@gaisler.com>
5378         * config/sparc/sparc.md (divdf3_fix): Add NOP and adjust length
5379         to prevent b2bst errata sequence.
5380         (sqrtdf2_fix): Likewise.
5382 2017-11-29  Daniel Cederman  <cederman@gaisler.com>
5384         * config/sparc/sparc.c (fpop_reg_depend_p): New function.
5385         (div_sqrt_insn_p): New function.
5386         (sparc_do_work_around_errata): Insert NOP instructions to
5387         prevent sequences that could trigger the TN-0013 errata for
5388         certain LEON3 processors.
5389         (pass_work_around_errata::gate): Also test sparc_fix_lost_divsqrt.
5390         (sparc_option_override): Set sparc_fix_lost_divsqrt appropriately.
5391         * config/sparc/sparc.md (fix_lost_divsqrt): New attribute.
5392         (in_branch_delay): Prevent div and sqrt in delay slot if
5393         fix_lost_divsqrt.
5394         * config/sparc/sparc.opt (sparc_fix_lost_divsqrt): New variable.
5396 2017-11-29  Daniel Cederman  <cederman@gaisler.com>
5398         * config/sparc/sparc.c (atomic_insn_p): New function.
5399         (sparc_do_work_around_errata): Insert NOP instructions to
5400         prevent sequences that could trigger the TN-0010 errata for
5401         UT700.
5402         * config/sparc/sync.md (atomic_compare_and_swap_leon3_1): Make
5403         instruction referable in atomic_insns_p.
5405 2017-11-29  Daniel Cederman  <cederman@gaisler.com>
5407         * config/sparc/sync.md (swapsi): 16-byte align if sparc_fix_gr712rc.
5408         (atomic_compare_and_swap_leon3_1): Likewise.
5409         (ldstub): Likewise.
5411 2017-11-29  Daniel Cederman  <cederman@gaisler.com>
5413         * config/sparc/sparc.c (fpop_insn_p): New function.
5414         (sparc_do_work_around_errata): Insert NOP instructions to
5415         prevent sequences that could trigger the TN-0012 errata for
5416         GR712RC.
5417         (pass_work_around_errata::gate): Also test sparc_fix_gr712rc.
5418         * config/sparc/sparc.md (fix_gr712rc): New attribute.
5419         (in_branch_annul_delay): Prevent floating-point instructions
5420         in delay slot of annulled integer branch.
5422 2017-11-29  Richard Biener  <rguenther@suse.de>
5424         PR tree-optimization/83202
5425         * tree-vect-slp.c (scalar_stmts_set_t): New typedef.
5426         (bst_fail): Use it.
5427         (vect_analyze_slp_cost_1): Add visited set, do not account SLP
5428         nodes vectorized to the same stmts multiple times.
5429         (vect_analyze_slp_cost): Allocate a visited set and pass it down.
5430         (vect_analyze_slp_instance): Adjust.
5431         (scalar_stmts_to_slp_tree_map_t): New typedef.
5432         (vect_schedule_slp_instance): Add a map recording the SLP node
5433         representing the vectorized stmts for a set of scalar stmts.
5434         Avoid code-generating redundancies.
5435         (vect_schedule_slp): Allocate map and pass it down.
5437 2017-11-29  Nathan Sidwell  <nathan@acm.org>
5439         PR c++/83187
5440         * tree.c (build_complex_type): Fix canonicalization.  Only fill in
5441         type if it is new.
5443 2017-11-29  Wilco Dijkstra  <wdijkstr@arm.com>
5445         * config/aarch64/aarch64.c (aarch64_print_operand): Add new
5446         cases for printing LDP/STP memory addresses.
5447         (aarch64_print_address_internal): Renamed from
5448         aarch64_print_operand_address, added parameter, add Pmode check.
5449         (aarch64_print_ldpstp_address): New function for LDP/STP addresses.
5450         (aarch64_print_operand_address): Indirect to
5451         aarch64_print_address_internal.
5452         * config/aarch64/aarch64-simd.md (store_pair_lanes): Use new
5453         'y' operand output specifier.
5455 2017-11-29  Jakub Jelinek  <jakub@redhat.com>
5457         PR middle-end/83185
5458         * tree.c (build_simple_mem_ref_loc): Handle
5459         get_addr_base_and_unit_offset returning a MEM_REF.
5461         PR middle-end/80929
5462         * rtlanal.c (seq_cost): For non-single_set insns try to use insn_cost.
5464         PR target/80819
5465         * config/i386/sse.md (vec_concatv2di): Remove * from (=Yr,0,*rm)
5466         alternative.
5468 2017-11-28  David Malcolm  <dmalcolm@redhat.com>
5470         * diagnostic-show-locus.c (layout::print_trailing_fixits): Handle
5471         m_x_offset.
5472         (layout::move_to_column): Likewise.
5474 2017-11-28  Jakub Jelinek  <jakub@redhat.com>
5476         PR sanitizer/81275
5477         * tree.c (block_may_fallthru): Return false if SWITCH_ALL_CASES_P
5478         is set on SWITCH_EXPR and !block_may_fallthru (SWITCH_BODY ()).
5480 2017-11-28  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
5481             Martin Jambor  <mjambor@suse.cz>
5483         PR ipa/82808
5484         * tree.h (expr_type_first_operand_type_p): Declare
5485         * tree.c (expr_type_first_operand_type_p): New function.
5486         * ipa-prop.h (ipa_get_type): Allow i to be out of bounds.
5487         (ipa_value_from_jfunc): Adjust declaration.
5488         * ipa-cp.c (ipa_get_jf_pass_through_result): New parameter RES_TYPE.
5489         Use it as result type for arithmetics, unless it is NULL in which case
5490         be more conservative.
5491         (ipa_value_from_jfunc): New parameter PARM_TYPE, pass it to
5492         ipa_get_jf_pass_through_result.
5493         (propagate_vals_across_pass_through): Likewise.
5494         (propagate_scalar_across_jump_function): New parameter PARM_TYPE, pass
5495         is to propagate_vals_across_pass_through.
5496         (propagate_constants_across_call): Pass PARM_TYPE to
5497         propagate_scalar_across_jump_function.
5498         (find_more_scalar_values_for_callers_subset): Pass parameter type to
5499         ipa_value_from_jfunc.
5500         (cgraph_edge_brings_all_scalars_for_node): Likewise.
5501         * ipa-fnsummary.c (evaluate_properties_for_edge): Renamed parms_info
5502         to caller_parms_info, pass parameter type to ipa_value_from_jfunc.
5503         * ipa-prop.c (try_make_edge_direct_simple_call): New parameter
5504         target_type, pass it to ipa_value_from_jfunc.
5505         (update_indirect_edges_after_inlining): Pass parameter type to
5506         try_make_edge_direct_simple_call.
5508 2017-11-28  Jeff Law  <law@redhat.com>
5510         * gimple-ssa-evrp-analyze.c
5511         (evrp_range_analyzer::record_ranges_from_phis): Only use SCEV to
5512         refine ranges if scev_initialized_p returns true.
5513         * vr-values.c (vr_values::extract_range_from_phi_node): Likewise.
5515 2017-11-28  Julia Koval  <julia.koval@intel.com>
5517         * config/i386/avx512vbmi2intrin.h (_mm512_shrdi_epi16,
5518         _mm512_mask_shrdi_epi16, _mm512_maskz_shrdi_epi16, _mm512_shrdi_epi32,
5519         _mm512_mask_shrdi_epi32, _mm512_maskz_shrdi_epi32, _mm512_shrdi_epi64,
5520         _mm512_mask_shrdi_epi64, _mm512_maskz_shrdi_epi64): New intrinsics.
5521         * config/i386/avx512vbmi2vlintrin.h (_mm256_shrdi_epi16,
5522         _mm256_mask_shrdi_epi16, _mm256_maskz_shrdi_epi16,
5523         _mm256_mask_shrdi_epi32, _mm256_maskz_shrdi_epi32, _mm256_shrdi_epi32,
5524         _mm256_mask_shrdi_epi64, _mm256_maskz_shrdi_epi64, _mm256_shrdi_epi64,
5525         _mm_mask_shrdi_epi16, _mm_maskz_shrdi_epi16, _mm_shrdi_epi16,
5526         _mm_mask_shrdi_epi32, _mm_maskz_shrdi_epi32, _mm_shrdi_epi32,
5527         _mm_mask_shrdi_epi64, _mm_maskz_shrdi_epi64, _mm_shrdi_epi64): Ditto.
5528         * config/i386/i386-builtin.def (__builtin_ia32_vpshrd_v32hi,
5529         __builtin_ia32_vpshrd_v32hi_mask, __builtin_ia32_vpshrd_v16hi,
5530         __builtin_ia32_vpshrd_v16hi_mask, __builtin_ia32_vpshrd_v8hi,
5531         __builtin_ia32_vpshrd_v8hi_mask, __builtin_ia32_vpshrd_v16si,
5532         __builtin_ia32_vpshrd_v16si_mask, __builtin_ia32_vpshrd_v8si,
5533         __builtin_ia32_vpshrd_v8si_mask, __builtin_ia32_vpshrd_v4si,
5534         __builtin_ia32_vpshrd_v4si_mask, __builtin_ia32_vpshrd_v8di,
5535         __builtin_ia32_vpshrd_v8di_mask, __builtin_ia32_vpshrd_v4di,
5536         __builtin_ia32_vpshrd_v4di_mask, __builtin_ia32_vpshrd_v2di,
5537         __builtin_ia32_vpshrd_v2di_mask): New builtins.
5538         * config/i386/sse.md (vpshrd_<mode><mask_name>): New pattern.
5540 2017-11-28  Julia Koval  <julia.koval@intel.com>
5542         * config/i386/avx512vbmi2intrin.h (_mm512_shldi_epi16,
5543         _mm512_mask_shldi_epi16, _mm512_maskz_shldi_epi16, _mm512_shldi_epi32,
5544         _mm512_mask_shldi_epi32, _mm512_maskz_shldi_epi32, _mm512_shldi_epi64,
5545         _mm512_mask_shldi_epi64, _mm512_maskz_shldi_epi64): New intrinsics.
5546         * config/i386/avx512vbmi2vlintrin.h (_mm256_shldi_epi16,
5547         _mm256_mask_shldi_epi16, _mm256_maskz_shldi_epi16,
5548         _mm256_mask_shldi_epi32, _mm256_maskz_shldi_epi32, _mm256_shldi_epi32,
5549         _mm256_mask_shldi_epi64, _mm256_maskz_shldi_epi64, _mm256_shldi_epi64,
5550         _mm_mask_shldi_epi16, _mm_maskz_shldi_epi16, _mm_shldi_epi16,
5551         _mm_mask_shldi_epi32, _mm_maskz_shldi_epi32, _mm_shldi_epi32,
5552         _mm_mask_shldi_epi64, _mm_maskz_shldi_epi64, _mm_shldi_epi64): Ditto.
5553         * config/i386/i386-builtin-types.def (V32HI_FTYPE_V32HI_V32HI_INT,
5554         V32HI_FTYPE_V32HI_V32HI_INT_V32HI_INT, V16SI_FTYPE_V16SI_V16SI_INT,
5555         V16SI_FTYPE_V16SI_V16SI_INT_V16SI_INT,
5556         V8DI_FTYPE_V8DI_V8DI_INT_V8DI_INT, V8SI_FTYPE_V8SI_V8SI_INT_V8SI_INT,
5557         V16HI_FTYPE_V16HI_V16HI_INT_V16HI_INT,
5558         V4DI_FTYPE_V4DI_V4DI_INT_V4DI_INT,
5559         V8HI_FTYPE_V8HI_V8HI_INT_V8HI_INT,
5560         V4SI_FTYPE_V4SI_V4SI_INT_V4SI_INT,
5561         V2DI_FTYPE_V2DI_V2DI_INT_V2DI_INT): New types.
5562         * config/i386/i386-builtin.def (__builtin_ia32_vpshld_v32hi,
5563         __builtin_ia32_vpshld_v32hi_mask, __builtin_ia32_vpshld_v16hi,
5564         __builtin_ia32_vpshld_v16hi_mask, __builtin_ia32_vpshld_v8hi,
5565         __builtin_ia32_vpshld_v8hi_mask, __builtin_ia32_vpshld_v16si,
5566         __builtin_ia32_vpshld_v16si_mask, __builtin_ia32_vpshld_v8si,
5567         __builtin_ia32_vpshld_v8si_mask, __builtin_ia32_vpshld_v4si,
5568         __builtin_ia32_vpshld_v4si_mask, __builtin_ia32_vpshld_v8di,
5569         __builtin_ia32_vpshld_v8di_mask, __builtin_ia32_vpshld_v4di,
5570         __builtin_ia32_vpshld_v4di_mask, __builtin_ia32_vpshld_v2di,
5571         __builtin_ia32_vpshld_v2di_mask): New builtins.
5572         * config/i386/i386.c (ix86_expand_args_builtin): Handle new types.
5573         * config/i386/sse.md (vpshld_<mode><mask_name>): New pattern.
5575 2017-11-28  Richard Biener  <rguenther@suse.de>
5577         PR tree-optimization/80776
5578         * gimple-ssa-evrp-analyze.h (evrp_range_analyzer::set_ssa_range_info):
5579         Declare.
5580         * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::set_ssa_range_info):
5581         New function.
5582         (evrp_range_analyzer::record_ranges_from_incoming_edges):
5583         If the incoming edge is an effective fallthru because the other
5584         edge only reaches a __builtin_unreachable () then record ranges
5585         derived from the controlling condition in SSA info.
5586         (evrp_range_analyzer::record_ranges_from_phis): Use set_ssa_range_info.
5587         (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
5589 2017-11-28  Olivier Hainque  <hainque@adacore.com>
5591         * Makefile.in (SELFTEST_FLAGS): Use nul instead of /dev/null
5592         on mingw build hosts.
5594 2017-11-28  Jakub Jelinek  <jakub@redhat.com>
5596         PR debug/81307
5597         * dbxout.c (dbx_block_with_cold_children): Fix function comment.
5599 2017-11-28  Richard Biener  <rguenther@suse.de>
5601         PR middle-end/83141
5602         * gimple-fold.c (gimple_fold_builtin_memory_op): For aggregate
5603         copies generated from memcpy use a character array as reference
5604         type.
5606 2017-11-28  Julia Koval  <julia.koval@intel.com>
5607             Sebastian Peryt  <sebastian.peryt@intel.com>
5609         * Makefile.in (cilkplus.def, cilk-builtins.def, c-family/cilk.o,
5610         c-family/c-cilkplus.o, c-family/array-notation-common.o,
5611         cilk-common.o, cilk.h, cilk-common.c): Remove.
5612         * builtin-types.def
5613         (BT_FN_INT_PTR_PTR_PTR_FTYPE_BT_INT_BT_PTR_BT_PTR_BT_PTR): Remove.
5614         * builtins.c (is_builtin_name): Remove cilkplus condition.
5615         (BUILT_IN_CILK_DETACH, BUILT_IN_CILK_POP_FRAME): Remove.
5616         * builtins.def (DEF_CILK_BUILTIN_STUB, DEF_CILKPLUS_BUILTIN,
5617         cilk-builtins.def, cilkplus.def): Remove.
5618         * cif-code.def (CILK_SPAWN): Remove.
5619         * cilk-builtins.def: Delete.
5620         * cilk-common.c: Ditto.
5621         * cilk.h: Ditto.
5622         * cilkplus.def: Ditto.
5623         * config/darwin.h (fcilkplus): Delete.
5624         * cppbuiltin.c: Ditto.
5625         * doc/extend.texi: Remove cilkplus doc.
5626         * doc/generic.texi: Ditto.
5627         * doc/invoke.texi: Ditto.
5628         * doc/passes.texi: Ditto.
5629         * gcc.c (fcilkplus): Remove.
5630         * gengtype.c (cilk.h): Remove.
5631         * gimple-pretty-print.c (dump_gimple_omp_for): Remove cilkplus
5632         support.
5633         * gimple.h (GF_OMP_FOR_KIND_CILKFOR, GF_OMP_FOR_KIND_CILKSIMD):
5634         Remove.
5635         * gimplify.c (gimplify_return_expr, maybe_fold_stmt,
5636         gimplify_call_expr,
5637         is_gimple_stmt, gimplify_modify_expr, gimplify_scan_omp_clauses,
5638         gimplify_adjust_omp_clauses, gimplify_omp_for, gimplify_expr): Remove
5639         cilkplus conditions.
5640         * ipa-fnsummary.c (ipa_dump_fn_summary, compute_fn_summary,
5641         inline_read_section): Ditto.
5642         * ipa-inline-analysis.c (cilk.h): Remove.
5643         * ira.c (ira_setup_eliminable_regset): Remove cilkplus support.
5644         * lto-wrapper.c (merge_and_complain, append_compiler_options,
5645         append_linker_options): Remove condition for fcilkplus.
5646         * lto/lto-lang.c (cilk.h): Remove.
5647         (lto_init): Remove condition for fcilkplus.
5648         * omp-expand.c (expand_cilk_for_call): Delete.
5649         (expand_omp_taskreg, expand_omp_for_static_chunk,
5650         expand_omp_for): Remove cilkplus
5651         conditions.
5652         (expand_cilk_for): Delete.
5653         * omp-general.c (omp_extract_for_data): Remove cilkplus support.
5654         * omp-low.c (scan_sharing_clauses, create_omp_child_function,
5655         execute_lower_omp, diagnose_sb_0): Ditto.
5656         * omp-simd-clone.c (simd_clone_clauses_extract): Ditto.
5657         * tree-core.h (OMP_CLAUSE__CILK_FOR_COUNT_): Delete.
5658         * tree-nested.c: Ditto.
5659         * tree-pretty-print.c (dump_omp_clause): Remove cilkplus support.
5660         (dump_generic_node): Ditto.
5661         * tree.c (OMP_CLAUSE__CILK_FOR_COUNT_): Delete.
5662         * tree.def (cilk_simd, cilk_for, cilk_spawn_stmt, cilk_sync_stmt):
5663         Delete.
5664         * tree.h (CILK_SPAWN_FN, EXPR_CILK_SPAWN): Delete.
5666 2017-11-28  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
5668         * config/aarch64/aarch64.md (div<mode>3): Change check to TARGET_FLOAT.
5669         * config/aarch64/aarch64.c (aarch64_emit_approx_div): Add early exit
5670         for vector mode and !TARGET_SIMD.
5672 2017-11-28  Jakub Jelinek  <jakub@redhat.com>
5674         * tree.def (SWITCH_EXPR): Change from 3 operand to 2 operand tree.
5675         Adjust comment.
5676         * tree.h (SWITCH_LABELS): Remove.
5677         * gimplify.c (gimplify_switch_expr): Don't test SWITCH_LABELS,
5678         assert SWITCH_BODY is non-NULL.
5679         * tree-pretty-print.c (dump_generic_node): Remove SWITCH_LABELS
5680         handling.
5681         * tree.c (block_may_fallthru): Always return true; for SWITCH_EXPR.
5683         PR tree-optimization/80788
5684         * match.pd (X +- C1 CMP C2 -> X CMP C2 -+ C1): If res
5685         has TREE_OVERFLOW set, call drop_tree_overflow.
5687 2017-11-28  Richard Biener  <rguenther@suse.de>
5689         PR tree-optimization/83158
5690         * tree-vrp.c (intersect_ranges): Prefer ~[0, 0] in a few more cases.
5692 2017-11-28  Segher Boessenkool  <segher@kernel.crashing.org>
5694         PR 81288/target
5695         * config/rs6000/rs6000.c (rs6000_rtx_costs): Do not handle
5696         TARGET_ISEL && !TARGET_MFCRF differently.  Simplify code.
5698 2017-11-27  Segher Boessenkool  <segher@kernel.crashing.org>
5700         * config/rs6000/rs6000.md (<code><GPR:mode><GPR2:mode>2_isel): Change
5701         LT/GT/LTU/GTU to LE/GE/LEU/GEU where possible.
5703 2017-11-27  Michael Meissner  <meissner@linux.vnet.ibm.com>
5705         PR middle_end/82333
5706         * varasm.c (compare_constant): Take the mode of the constants into
5707         account when comparing floating point constants.
5709 2017-11-27  Gerald Pfeifer  <gerald@pfeifer.com>
5711         * hash-set.h (DEFINE_DEBUG_HASH_SET): Remove static qualifier
5712         from explicit instantiation of debug_helper.
5713         * vec.h (DEFINE_DEBUG_VEC): Ditto.
5715 2017-11-27  Richard Biener  <rguenther@suse.de>
5717         * gimple-fold.c (gimple_fold_builtin_memory_op): Remove dead code,
5718         refactor a bit.
5720 2017-11-27  Richard Biener  <rguenther@suse.de>
5722         * tree.c (wide_int_to_tree): Free discarded INTEGER_CST.
5723         (type_hash_canon): Also clear int_cst_hash_table entry for
5724         TYPE_MIN/MAX_VALUE.
5725         (build_nonstandard_integer_type): Hash all TYPE_MAX_VALUEs.
5727 2017-11-27  Tamar Christina  <tamar.christina@arm.com>
5729         * doc/extend.texi: Add -A suffix (ARMv8*-A, ARMv7-A).
5730         * doc/invoke.texi: Add -A suffix (ARMv8*-A, ARMv7-A).
5731         * doc/sourcebuild.texi: Add -A suffix (ARMv8*-A, ARMv7-A).
5733 2017-11-27  Markus Trippelsdorf  <markus@trippelsdorf.de>
5735         * hash-map.h (gt_cleare_cache): Avoid UB.
5737 2017-11-27  Eric Botcazou  <ebotcazou@adacore.com>
5739         * cfgloop.h (struct loop): Document usage of USHRT_MAX for unroll.
5740         * loop-unroll.c (decide_unroll_constant_iterations): Implement it.
5741         (decide_unroll_runtime_iterations): Likewise.
5742         (decide_unroll_stupid): Likewise.
5744 2017-11-27  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
5746         PR target/83109
5747         * config/i386/i386.md: Add a loop with incssp.
5749 2017-11-27  Martin Jambor  <mjambor@suse.cz>
5751         PR tree-optimization/81248
5752         * tree-sra.c (splice_param_accesses): Remove size check.
5753         (decide_one_param_reduction): Fix size check.
5754         * gimple-pretty-print.c (dump_profile): Silence warning.
5755         * params.def (PARAM_IPA_SRA_PTR_GROWTH_FACTOR): Adjust description.
5757 2017-11-27  Jakub Jelinek  <jakub@redhat.com>
5759         PR debug/81307
5760         * dbxout.c (lastlineno): New variable.
5761         (dbx_debug_hooks): Use dbxout_switch_text_section as
5762         switch_text_section debug hook.
5763         (dbxout_function_end): Switch to current_function_section
5764         rather than function_section.  If crtl->has_bb_partition,
5765         output just one N_FUN, depending on in_cold_section_p.
5766         (dbxout_source_line): Remember last lineno in lastlineno.
5767         (dbxout_switch_text_section): New function.
5768         (dbxout_function_decl): Adjust dbxout_block caller.
5769         (dbx_block_with_cold_children): New function.
5770         (dbxout_block): Return true if any LBRAC/RBRAC have been
5771         emitted.  Use dbx_block_with_cold_children at depth == 0
5772         in second partition.  Add PARENT_BLOCKNUM argument, pass
5773         it optionally adjusted to children.  Output LBRAC/RBRAC
5774         around recursive call only if the block is in the current
5775         partition, if not and anything was output, emit empty
5776         range LBRAC/RBRAC.
5777         * final.c (final_scan_insn): Compute cold_function_name
5778         before calling switch_text_section debug hook.  Call
5779         that hook even if dwarf2out_do_frame if not emitting
5780         dwarf debug info.
5782         PR target/83100
5783         * varasm.c (bss_initializer_p): Return true for DECL_COMMON
5784         TREE_READONLY decls.
5786 2017-11-27  Markus Trippelsdorf  <markus@trippelsdorf.de>
5788         PR rtl-optimization/82488
5789         * expr.c (fixup_args_size_notes): Avoid signed integer overflow.
5791 2017-11-26  Julia Koval  <julia.koval@intel.com>
5793         * config/i386/i386.c (processor_target_table): Add skylake_cost for
5794         skylake-avx512.
5795         * config/i386/x86-tune-costs.h (skylake_memcpy, skylake_memset,
5796         skylake_cost): New.
5798 2017-11-26  Julia Koval  <julia.koval@intel.com>
5800         * config/i386/driver-i386.c (host_detect_local_cpu):
5801         Detect skylake-avx512.
5803 2017-11-26  Julia Koval  <julia.koval@intel.com>
5805         * config.gcc: Add -march=cannonlake.
5806         * config/i386/driver-i386.c (host_detect_local_cpu): Detect cannonlake.
5807         * config/i386/i386-c.c (ix86_target_macros_internal): Handle cannonlake.
5808         * config/i386/i386.c (processor_costs): Add m_CANNONLAKE.
5809         (PTA_CANNONLAKE): New.
5810         (processor_target_table): Add cannonlake.
5811         (ix86_option_override_internal): Ditto.
5812         (fold_builtin_cpu): Ditto.
5813         (get_builtin_code_for_version): Handle cannonlake.
5814         (M_INTEL_COREI7_CANNONLAKE): New.
5815         * config/i386/i386.h (TARGET_CANNONLAKE, PROCESSOR_CANNONLAKE): New.
5816         * doc/invoke.texi: Add -march=cannonlake.
5818 2017-11-14  Boris Kolpackov  <boris@codesynthesis.com>
5820         * plugin.c (add_new_plugin): Use platform-specific library extensions.
5821         (try_init_one_plugin): Alternative implementation for MinGW.
5822         * Makefile.in (plugin_implib): New.
5823         (gengtype-lex.c): Fix broken AIX workaround.
5824         * configure: Regenerate.
5825         * doc/plugins.texi: Document support for MinGW.
5827 2017-11-25  Jakub Jelinek  <jakub@redhat.com>
5829         PR rtl-optimization/81553
5830         * combine.c (simplify_if_then_else): In (if_then_else COND (OP Z C1) Z)
5831         to (OP Z (mult COND (C1 * STORE_FLAG_VALUE))) optimization, if OP
5832         is a shift where C1 has different mode than the whole shift, use C1's
5833         mode for MULT rather than the shift's mode.
5835         PR target/82848
5836         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Don't fold
5837         builtins not enabled in the currently selected ISA.
5839 2017-11-24  Jackson Woodruff  <jackson.woodruff@arm.com>
5841         PR tree-optimization/71026
5842         * tree-ssa-math-opts (is_division_by_square, is_square_of): New.
5843         (insert_reciprocals): Change to insert reciprocals before a division
5844         by a square and to insert the square of a reciprocal.
5845         (execute_cse_reciprocals_1): Change to consider division by a square.
5846         (register_division_in): Add importance parameter.
5848 2017-11-24  Richard Biener  <rguenther@suse.de>
5850         PR tree-optimization/82402
5851         * tree-vect-loop-manip.c (create_lcssa_for_virtual_phi): Properly
5852         set SSA_NAME_OCCURS_IN_ABNORMAL_PHI.
5854 2017-11-24  Marc Glisse  <marc.glisse@inria.fr>
5856         * match.pd (0-ptr): New transformation.
5858 2017-11-24  Jan Hubicka  <hubicka@ucw.cz>
5860         PR bootstrap/83015
5861         * ipa-inline.c (inline_small_functions): Set current badnes correctly
5862         when skipping checking.
5864 2017-11-24  Richard Biener  <rguenther@suse.de>
5866         PR tree-optimization/83128
5867         * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Handle STRING_CSTs.
5868         (vn_reference_lookup_3): Likewise.
5870 2017-11-24  Jakub Jelinek  <jakub@redhat.com>
5872         PR sanitizer/83014
5873         * ubsan.c (ubsan_type_descriptor): Use pp_unsigned_wide_integer
5874         instead of pp_printf with HOST_WIDE_INT_PRINT_DEC.  Avoid calling
5875         tree_to_uhwi twice.
5877         * tree-object-size.c (pass_through_call): Use gimple_call_return_flags
5878         ERF_RETURN*ARG* for builtins other than BUILT_IN_ASSUME_ALIGNED,
5879         check for the latter with gimple_call_builtin_p.  Do not handle
5880         BUILT_IN_STPNCPY_CHK which is not a pass through call.
5882 2017-11-24  Christophe Lyon  <christophe.lyon@linaro.org>
5884         * config/arm/arm_neon.h: Fix pragma GCC push_options before
5885         vdot_u32.
5887 2017-11-23  Julia Koval  <julia.koval@intel.com>
5889         * config/i386/avx512vbmi2intrin.h (_mm512_mask_expand_epi8,
5890         _mm512_maskz_expand_epi8, _mm512_mask_expandloadu_epi8,
5891         _mm512_maskz_expandloadu_epi8, _mm512_mask_expand_epi16,
5892         _mm512_maskz_expand_epi16, _mm512_mask_expandloadu_epi16,
5893         _mm512_maskz_expandloadu_epi16): New intrinsics.
5894         * config/i386/avx512vbmi2vlintrin.h (_mm_mask_expand_epi8,
5895         _mm_maskz_expand_epi8, _mm_mask_expandloadu_epi8,
5896         _mm_maskz_expandloadu_epi8, _mm_mask_expand_epi16,
5897         _mm_maskz_expand_epi16, _mm_mask_expandloadu_epi16,
5898         _mm_maskz_expandloadu_epi16, _mm256_mask_expand_epi16,
5899         _mm256_maskz_expand_epi16, _mm256_mask_expandloadu_epi16,
5900         _mm256_maskz_expandloadu_epi16, _mm256_mask_expand_epi8,
5901         _mm256_maskz_expand_epi8, _mm256_mask_expandloadu_epi8,
5902         _mm256_maskz_expandloadu_epi8): New intrinsics.
5903         * config/i386/i386-builtin-types.def (V64QI_FTYPE_PCV64QI_V64QI_UDI,
5904         V32HI_FTYPE_PCV32HI_V32HI_USI, V32QI_FTYPE_PCV32QI_V32QI_USI,
5905         V16HI_FTYPE_PCV16HI_V16HI_UHI, V16QI_FTYPE_PCV16QI_V16QI_UHI,
5906         V8HI_FTYPE_PCV8HI_V8HI_UQI): New types.
5907         * config/i386/i386.c (ix86_expand_special_args_builtin): Use new types.
5908         * config/i386/sse.md (VI248_VLBW): New iterator.
5909         (expand<mode>_mask, expand<mode>_maskz): New patterns.
5911 2017-11-23  Julia Koval  <julia.koval@intel.com>
5913         * config.gcc (avx512vbmi2intrin.h, avx512vbmi2vlintrin): New headers.
5914         * config/i386/avx512vbmi2intrin.h (_mm512_mask_compress_epi8,
5915         _mm512_maskz_compress_epi8, _mm512_mask_compressstoreu_epi8,
5916         _mm512_mask_compress_epi16, _mm512_maskz_compress_epi16,
5917         _mm512_mask_compressstoreu_epi16): New.
5918         * config/i386/avx512vbmi2vlintrin.h (_mm_mask_compress_epi8,
5919         _mm_maskz_compress_epi8, _mm256_mask_compressstoreu_epi16,
5920         _mm_mask_compress_epi16, _mm_maskz_compress_epi16,
5921         _mm256_mask_compress_epi16, _mm256_maskz_compress_epi16,
5922         _mm_mask_compressstoreu_epi8, _mm_mask_compressstoreu_epi16,
5923         _mm256_mask_compress_epi8, _mm256_maskz_compress_epi8,
5924         _mm256_mask_compressstoreu_epi8): New.
5925         * config/i386/i386-builtin-types.def (VOID_FTYPE_PV64QI_V64QI_UDI,
5926         VOID_FTYPE_PV32HI_V32HI_USI, VOID_FTYPE_PV32QI_V32QI_USI,
5927         VOID_FTYPE_PV16QI_V16QI_UHI, VOID_FTYPE_PV16HI_V16HI_UHI,
5928         VOID_FTYPE_PV8HI_V8HI_UQI): New types.
5929         * config/i386/i386-builtin.def (__builtin_ia32_compressqi512_mask,
5930         __builtin_ia32_compresshi512_mask, __builtin_ia32_compressqi256_mask,
5931         __builtin_ia32_compressqi128_mask, __builtin_ia32_compresshi256_mask,
5932         __builtin_ia32_compresshi128_mask,
5933         __builtin_ia32_compressstoreuqi512_mask,
5934         __builtin_ia32_compressstoreuhi512_mask,
5935         __builtin_ia32_compressstoreuqi256_mask,
5936         __builtin_ia32_compressstoreuqi128_mask,
5937         __builtin_ia32_compressstoreuhi256_mask,
5938         __builtin_ia32_compressstoreuhi128_mask): New builtins.
5939         * config/i386/i386.c (ix86_init_mmx_sse_builtins): Create special args
5940         array for flags2.
5941         (ix86_expand_special_args_builtin): Handle new types.
5942         (s4fma_expand): Handle new builtin array.
5943         * config/i386/immintrin.h: Include new headers.
5944         * config/i386/sse.md (VI12_AVX512VLBW): New iterator.
5945         (compress<mode>_mask, compressstore<mode>_mask): New patterns.
5947 2017-11-23  Jakub Jelinek  <jakub@redhat.com>
5949         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Formatting
5950         fixes.  Declare temp and g variables at the top in order to avoid
5951         {} in most of the cases.
5953 2017-11-23  Marc Glisse  <marc.glisse@inria.fr>
5955         * match.pd (ptr-0): New transformation.
5957 2017-11-23  Charles Baylis  <charles.baylis@linaro.org>
5959         * config/arm/arm-protos.h (enum arm_addr_mode_op): New.
5960         (struct addr_mode_cost_table): New.
5961         (struct tune_params): Add field addr_mode_costs.
5962         * config/arm/arm.c (generic_addr_mode_costs): New.
5963         (arm_slowmul_tune): Initialise addr_mode_costs field.
5964         (arm_fastmul_tune): Likewise.
5965         (arm_strongarm_tune): Likewise.
5966         (arm_xscale_tune): Likewise.
5967         (arm_9e_tune): Likewise.
5968         (arm_marvell_pj4_tune): Likewise.
5969         (arm_v6t2_tune): Likewise.
5970         (arm_cortex_tune): Likewise.
5971         (arm_cortex_a8_tune): Likewise.
5972         (arm_cortex_a7_tune): Likewise.
5973         (arm_cortex_a15_tune): Likewise.
5974         (arm_cortex_a35_tune): Likewise.
5975         (arm_cortex_a53_tune): Likewise.
5976         (arm_cortex_a57_tune): Likewise.
5977         (arm_exynosm1_tune): Likewise.
5978         (arm_xgene1_tune): Likewise.
5979         (arm_cortex_a5_tune): Likewise.
5980         (arm_cortex_a9_tune): Likewise.
5981         (arm_cortex_a12_tune): Likewise.
5982         (arm_cortex_a73_tune): Likewise.
5983         (arm_v7m_tune): Likewise.
5984         (arm_cortex_m7_tune): Likewise.
5985         (arm_v6m_tune): Likewise.
5986         (arm_fa726te_tune): Likewise.
5987         (arm_mem_costs): Use table lookup to calculate cost of addressing mode.
5989 2017-11-23  Charles Baylis  <charles.baylis@linaro.org>
5991         * config/arm/arm.c (arm_mem_costs): New function.
5992         (arm_rtx_costs_internal): Use arm_mem_costs.
5994 2017-11-23  Mark Wielaard  <mark@klomp.org>
5996         * dwarf2out.c (init_sections_and_labels): Use generation to create
5997         unique ranges_section_label and ranges_base_label. Return generation.
5998         (output_rnglists): Add generation argument. Use generation to create
5999         unique ranges labels.
6000         (dwarf2out_finish): Get generation from init_sections_and_labels
6001         and pass generation to output_rnglists.
6003 2017-11-23  Mike Stump  <mikestump@comcast.net>
6004             Eric Botcazou  <ebotcazou@adacore.com>
6006         * doc/generic.texi (ANNOTATE_EXPR): Document 3rd operand.
6007         * cfgloop.h (struct loop): Add unroll field.
6008         * function.h (struct function): Add has_unroll bitfield.
6009         * gimplify.c (gimple_boolify) <ANNOTATE_EXPR>: Deal with unroll kind.
6010         (gimplify_expr) <ANNOTATE_EXPR>: Propagate 3rd operand.
6011         * loop-init.c (pass_loop2::gate): Return true if cfun->has_unroll.
6012         (pass_rtl_unroll_loops::gate): Likewise.
6013         * loop-unroll.c (decide_unrolling): Tweak note message.  Skip loops
6014         for which loop->unroll==1.
6015         (decide_unroll_constant_iterations): Use note for consistency and
6016         take loop->unroll into account.  Return early if loop->unroll is set.
6017         Fix thinko in existing test.
6018         (decide_unroll_runtime_iterations): Use note for consistency and
6019         take loop->unroll into account.
6020         (decide_unroll_stupid): Likewise.
6021         * lto-streamer-in.c (input_cfg): Read loop->unroll.
6022         * lto-streamer-out.c (output_cfg): Write loop->unroll.
6023         * tree-cfg.c (replace_loop_annotate_in_block) <annot_expr_unroll_kind>:
6024         New case.
6025         (replace_loop_annotate) <annot_expr_unroll_kind>: Likewise.
6026         (print_loop): Print loop->unroll if set.
6027         * tree-core.h (enum annot_expr_kind): Add annot_expr_unroll_kind.
6028         * tree-inline.c (copy_loops): Copy unroll and set cfun->has_unroll.
6029         * tree-pretty-print.c (dump_generic_node) <annot_expr_unroll_kind>:
6030         New case.
6031         * tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Bail out if
6032         loop->unroll is set and smaller than the trip count.  Otherwise bypass
6033         entirely the heuristics if loop->unroll is set.  Remove dead note.
6034         Fix off-by-one bug in other note.
6035         (try_peel_loop): Bail out if loop->unroll is set.  Fix formatting.
6036         (tree_unroll_loops_completely_1): Force unrolling if loop->unroll
6037         is greater than 1.
6038         (tree_unroll_loops_completely): Make static.
6039         (pass_complete_unroll::execute): Use correct type for variable.
6040         (pass_complete_unrolli::execute): Fix formatting.
6041         * tree.def (ANNOTATE_EXPR): Add 3rd operand.
6043 2017-11-23  Sergey Shalnov  <Sergey.Shalnov@intel.com>
6045         * config/i386/i386.h (TARGET_PREFER_AVX256): Also
6046         enable when TARGET_PREFER_AVX128 is set.
6048 2017-11-23  Jan Hubicka  <hubicka@ucw.cz>
6050         * ipa-profile.c (ipa_propagate_frequency_1): Use count instead of
6051         frequency.
6052         * cgraph.c (cgraph_edge::dump_edge_flags): Dump sreal frequencies.
6053         (cgraph_edge::maybe_hot_p): Use sreal frequencies.
6055 2017-11-23  Jan Hubicka  <hubicka@ucw.cz>
6057         * ipa-fnsummary.c (record_modified_bb_info): Use sreal
6058         frequencies. Fix estimation of aggregate parameters.
6060 2017-11-23  Jan Hubicka  <hubicka@ucw.cz>
6062         * cgraphclones.c (cgraph_node::create_clone): Fix updating of profile
6063         when inlining.
6065 2017-11-23  Tom de Vries  <tom@codesourcery.com>
6067         * config/sh/sh.h (ASM_OUTPUT_ADDR_VEC_ELT): Wrap in "do {} while (0)".
6069 2017-11-23  Tom de Vries  <tom@codesourcery.com>
6071         * config/ft32/ft32.h (ASM_OUTPUT_ADDR_VEC_ELT): Remove semicolon after
6072         macro.
6074 2017-11-23  Oleg Endo  <olegendo@gcc.gnu.org>
6076         PR target/83111
6077         * config/sh/sh.md (udivsi3, divsi3, sibcall_value_pcrel,
6078         sibcall_value_pcrel_fdpic): Use local variable instead of
6079         operands[3].
6080         (calli_tbr_rel): Add missing operand 2.
6081         (call_valuei_tbr_rel): Add missing operand 3.
6083 2017-11-23  Jakub Jelinek  <jakub@redhat.com>
6085         PR middle-end/82253
6086         * expr.c (expand_assignment): For CONCAT to_rtx, complex type from and
6087         bitpos/bitsize covering the whole destination, use store_expr only if
6088         the complex mode is the same.  Otherwise, use expand_normal and if
6089         it returns CONCAT, subreg each part separately instead of trying to
6090         subreg the whole result.
6092 2017-11-23  Richard Biener  <rguenther@suse.de>
6094         PR tree-optimization/23094
6095         * tree-ssa-sccvn.c (vuse_ssa_val): Handle VN_TOP when we
6096         come here from walking over backedges in the first iteration.
6097         (vn_reference_lookup_3): Skip clobbers that store the same value.
6099 2017-11-23  Richard Biener  <rguenther@suse.de>
6101         PR tree-optimization/81403
6102         * tree-ssa-pre.c (get_representative_for): Add parameter specifying
6103         a block we need a leader relative to.
6104         (phi_translate_1): For nary processing require a leader from
6105         get_representative_for given we run expression simplification
6106         using match-and-simplify.  Remove previous fix.
6108 2017-11-22  Jeff Law  <law@redhat.com>
6110         * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::try_find_new_range):
6111         Use new method allocate_value_range rather than accessing the
6112         vrp_value_range_pool data member directly.
6113         * tree-vrp.c (simplify_stmt_for_jump_threading): Tweak slightly
6114         to use extract_range_from_stmt method to avoid need for
6115         extract_range_from_assignment method.
6116         (vrp_prop::vrp_finalize): Use set_lattice_propagation_complete
6117         method rather than setting values_propgated data member directly.
6118         * vr-values.h (class vr_values): Privatize vrp_value_range_pool,
6119         and values propagated data members and extract_range_from_assignment
6120         method.  Reorder private data members to conform to standards.
6121         Add new methods set_lattice_propagation_complete and
6122         allocate_value_range.
6124 2017-11-22  Eric Botcazou  <ebotcazou@adacore.com>
6126         PR rtl-optimization/83030
6127         * doc/rtl.texi (Flags in an RTL Expression): Alphabetize, add entry
6128         for CROSSING_JUMP_P and mention usage of 'jump' for JUMP_INSNs.
6129         (Insns): Delete entry for REG_CROSSING_JUMP in register notes.
6130         * bb-reorder.c (update_crossing_jump_flags): Do not test whether the
6131         CROSSING_JUMP_P flag is already set before setting it.
6132         * cfgrtl.c (fixup_partition_crossing): Likewise.
6133         * reorg.c (relax_delay_slots): Do not consider a CROSSING_JUMP_P
6134         insn as useless.
6136 2017-11-22  Jakub Jelinek  <jakub@redhat.com>
6138         * simplify-rtx.c (simplify_binary_operation_1) <case VEC_SERIES>:
6139         Handle the case where both arguments are using gen_const_vec_series.
6141 2017-11-22  David Malcolm  <dmalcolm@redhat.com>
6143         PR c++/62170
6144         * pretty-print.c (pp_format): Move quoting implementation to
6145         pp_begin_quote and pp_end_quote.  Update pp_format_decoder call
6146         to pass address of "quote" local.
6147         (pp_begin_quote): New function.
6148         (pp_end_quote): New function.
6149         * pretty-print.h (printer_fn): Convert penultimate param from bool
6150         to bool *.
6151         (pp_begin_quote): New decl.
6152         (pp_end_quote): New decl.
6153         * tree-diagnostic.c (default_tree_printer): Convert penultimate
6154         param from bool to bool *.
6155         * tree-diagnostic.h (default_tree_printer): Likewise.
6157 2017-11-22  Jeff Law  <law@redhat.com>
6159         * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::evrp_range_analyzer)
6160         Initialize vr_values.
6161         (evrp_range_analyzer::try_find_new_range): Call methods attached to
6162         vr_values via vr_values class instance rather than delegators.
6163         (evrp_range_analyzer::record_ranges_from_phis): Likewise.
6164         (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
6165         (evrp_range_analyzer::push_value_range): Likewise.
6166         (evrp_range_analyzer::pop_value_range): Likewise.
6167         * gimple-ssa-evrp-analyze.h (class evrp_range_analyzer): Remove
6168         most delegators.  Those remaining are exposed as public interfaces.
6169         Make vr_values a pointer and private.
6170         (evrp_range_analyzer::~evrp_range_analyzer): Delete the attached
6171         vr_values.
6172         (evrp_range_analyzer::get_vr_value): New method.
6173         * gimple-ssa-evrp.c (class evrp_folder): Use DISABLE_COPY_AND_ASSIGN.
6174         (evrp_folder::evrp_folder): New ctor to initialize vr_values.
6175         (class evrp_dom_walker): Attach evrp_folder class, initialize
6176         it in the ctor.  Remove temporary delegators.
6177         (evrp_dom_walker::before_dom_children): Call methods in attached
6178         evrp_range_analyzer class via class instance pointer.  Use
6179         free value_range_constant_singleton to remove need for
6180         op_with_constant_singleton_value delegator method.  Do not
6181         create a vrp_prop class instance for every call!  Narrow
6182         scope of a couple variables.
6183         (evrp_dom_walker::cleanup): Call methods in attached
6184         evrp_range_analyzer class via class instance pointer.
6185         * vr-values.h (class vr_values): Privatize many methods and
6186         data members.
6188 2017-11-22  Richard Sandiford  <richard.sandiford@linaro.org>
6190         * tree.def (REDUC_MAX_EXPR, REDUC_MIN_EXPR, REDUC_PLUS_EXPR): Delete.
6191         * cfgexpand.c (expand_debug_expr): Remove handling for them.
6192         * expr.c (expand_expr_real_2): Likewise.
6193         * fold-const.c (const_unop): Likewise.
6194         * optabs-tree.c (optab_for_tree_code): Likewise.
6195         * tree-cfg.c (verify_gimple_assign_unary): Likewise.
6196         * tree-inline.c (estimate_operator_cost): Likewise.
6197         * tree-pretty-print.c (dump_generic_node): Likewise.
6198         (op_code_prio): Likewise.
6199         (op_symbol_code): Likewise.
6200         * internal-fn.def (DEF_INTERNAL_SIGNED_OPTAB_FN): Define.
6201         (IFN_REDUC_PLUS, IFN_REDUC_MAX, IFN_REDUC_MIN): New internal functions.
6202         * internal-fn.c (direct_internal_fn_optab): New function.
6203         (direct_internal_fn_array, direct_internal_fn_supported_p
6204         (internal_fn_expanders): Handle DEF_INTERNAL_SIGNED_OPTAB_FN.
6205         * fold-const-call.c (fold_const_reduction): New function.
6206         (fold_const_call): Handle CFN_REDUC_PLUS, CFN_REDUC_MAX and
6207         CFN_REDUC_MIN.
6208         * tree-vect-loop.c: Include internal-fn.h.
6209         (reduction_code_for_scalar_code): Rename to...
6210         (reduction_fn_for_scalar_code): ...this and return an internal
6211         function.
6212         (vect_model_reduction_cost): Take an internal_fn rather than
6213         a tree_code.
6214         (vect_create_epilog_for_reduction): Likewise.  Build calls rather
6215         than assignments.
6216         (vectorizable_reduction): Use internal functions rather than tree
6217         codes for the reduction operation.  Update calls to the functions
6218         above.
6219         * config/aarch64/aarch64-builtins.c (aarch64_gimple_fold_builtin):
6220         Use calls to internal functions rather than REDUC tree codes.
6221         * config/aarch64/aarch64-simd.md: Update comment accordingly.
6223 2017-11-22  Olivier Hainque  <hainque@adacore.com>
6225         * config/vxworks.c (vxworks_override_options): Pick default
6226         dwarf version from macro value, VXWORKS_DWARF_VERSION_DEFAULT.
6227         * config/vxworks.h: Define VXWORKS_DWARF_VERSION_DEFAULT and
6228         DWARF_GNAT_ENCODINGS_DEFAULT.
6229         * config/vxworksae.h: Likewise.
6231 2017-11-22  Marc Glisse  <marc.glisse@inria.fr>
6233         PR tree-optimization/83104
6234         * vr-values.c (simplify_stmt_using_ranges): Check integral argument,
6235         not result.
6237 2017-11-22  Marek Polacek  <polacek@redhat.com>
6238             H.J. Lu  <hongjiu.lu@intel.com>
6239             Jason Merrill  <jason@redhat.com>
6241         PR c++/60336
6242         PR middle-end/67239
6243         PR target/68355
6244         * calls.c (initialize_argument_information): Call
6245         warn_parameter_passing_abi target hook.
6246         (store_one_arg): Use 0 for empty record size.  Don't push 0 size
6247         argument onto stack.
6248         (must_pass_in_stack_var_size_or_pad): Return false for empty types.
6249         * common.opt: Update -fabi-version description.
6250         * config/i386/i386.c (init_cumulative_args): Set cum->warn_empty.
6251         (ix86_gimplify_va_arg): Call arg_int_size_in_bytes instead of
6252         int_size_in_bytes.
6253         (ix86_is_empty_record): New function.
6254         (ix86_warn_parameter_passing_abi): New function.
6255         (TARGET_EMPTY_RECORD_P): Redefine.
6256         (TARGET_WARN_PARAMETER_PASSING_ABI): Redefine.
6257         * config/i386/i386.h (CUMULATIVE_ARGS): Add warn_empty.
6258         * doc/tm.texi: Regenerated.
6259         * doc/tm.texi.in (TARGET_EMPTY_RECORD_P,
6260         TARGET_WARN_PARAMETER_PASSING_ABI): Add.
6261         * dwarf2out.c (get_ultimate_context): Move to tree.c.
6262         * explow.c (hard_function_value): Call arg_int_size_in_bytes
6263         instead of int_size_in_bytes.
6264         * expr.c (copy_blkmode_to_reg): Likewise.
6265         * function.c (aggregate_value_p): Return 0 for empty types.
6266         (assign_parm_find_entry_rtl): Call warn_parameter_passing_abi
6267         target hook.
6268         (locate_and_pad_parm): Call arg size_in_bytes instead
6269         size_in_bytes.
6270         * lto-streamer-out.c (hash_tree): Hash TYPE_EMPTY_P and DECL_PADDING_P.
6271         * stor-layout.c (finalize_type_size): Set TYPE_EMPTY_P.
6272         * target.def (empty_record_p, warn_parameter_passing_abi): New target
6273         hooks.
6274         * targhooks.c (hook_void_CUMULATIVE_ARGS_tree): New hook.
6275         (std_gimplify_va_arg_expr): Skip empty records.  Call
6276         arg_size_in_bytes instead size_in_bytes.
6277         * targhooks.h (hook_void_CUMULATIVE_ARGS_tree): Declare.
6278         * tree-core.h (tree_type_common): Add empty_flag.
6279         (tree_decl_common): Update comments.
6280         * tree-streamer-in.c (unpack_ts_decl_common_value_fields): Stream
6281         DECL_PADDING_P.
6282         (unpack_ts_type_common_value_fields): Stream TYPE_EMPTY_P.
6283         * tree-streamer-out.c (pack_ts_decl_common_value_fields): Stream
6284         DECL_PADDING_P.
6285         (pack_ts_type_common_value_fields): Stream TYPE_EMPTY_P.
6286         * tree.c (default_is_empty_type): New function.
6287         (default_is_empty_record): New function.
6288         (arg_int_size_in_bytes): New function.
6289         (arg_size_in_bytes): New function.
6290         (get_ultimate_context): New function.
6291         * tree.h: Define TYPE_EMPTY_P, DECL_PADDING_P and
6292         TRANSLATION_UNIT_WARN_EMPTY_P.
6293         (default_is_empty_record, arg_int_size_in_bytes,
6294         arg_size_in_bytes, get_ultimate_context): Declare.
6296 2017-11-22  Thomas Preud'homme  <thomas.preudhomme@arm.com>
6298         * config/arm/arm.c (cmse_clear_registers): New function.
6299         (cmse_nonsecure_call_clear_caller_saved): Replace register clearing
6300         code by call to cmse_clear_registers.
6301         (cmse_nonsecure_entry_clear_before_return): Likewise.
6303 2017-11-22  Tamar Christina  <tamar.christina@arm.com>
6305         * config/arm/arm_neon.h (vdot_u32, vdotq_u32)
6306         (vdot_s32, vdotq_s32): New.
6307         (vdot_lane_u32, vdotq_lane_u32): New.
6308         (vdot_lane_s32, vdotq_lane_s32): New.
6311 2017-11-22  Richard Sandiford  <richard.sandiford@linaro.org>
6313         PR middle-end/82547
6314         * wide-int.cc (wi::add_large, wi::sub_large): Fix overflow detection
6315         for unsigned values with fewer HWIs than the precision.
6316         (test_overflow): New function.
6317         (wide_int_cc_tests): Call it.
6319 2017-11-22  Richard Sandiford  <richard.sandiford@linaro.org>
6321         * emit-rtl.c (init_derived_machine_modes): Make sure ptr_mode
6322         has the same mode class as Pmode.
6324 2017-11-22  Thomas Preud'homme  <thomas.preudhomme@arm.com>
6326         * config/arm/arm.c (cmse_nonsecure_call_clear_caller_saved): Get rid of
6327         padding_bits_to_clear_ptr.
6328         (cmse_nonsecure_entry_clear_before_return): Likewise.
6330 2017-11-22  Thomas Preud'homme  <thomas.preudhomme@arm.com>
6332         * config/arm/arm.c (cmse_nonsecure_call_clear_caller_saved): Use
6333         auto_sbitap instead of integer bitfield to control register needing
6334         clearing.
6336 2017-11-22  Jakub Jelinek  <jakub@redhat.com>
6338         PR tree-optimization/83044
6339         * tree-vrp.c (vrp_prop::check_array_ref): If eltsize is not
6340         INTEGER_CST or is 0, clear up_bound{,_p1} and later ignore tests
6341         that need the upper bound.  Subtract offset from
6342         get_addr_base_and_unit_offset only if positive and subtract it
6343         before division by eltsize rather than after it.
6345         PR debug/83084
6346         * valtrack.c (propagate_for_debug_subst, propagate_for_debug): Reset
6347         debug insns if they would contain UNSPEC_VOLATILE or volatile asm.
6348         (dead_debug_insert_temp): Likewise, but also ignore even non-volatile
6349         asm.
6351         PR middle-end/82875
6352         * optabs.c (expand_doubleword_mult, expand_binop): Before calling
6353         expand_binop with *mul_widen_optab, make sure at least one of the
6354         operands doesn't have VOIDmode.
6356         PR debug/83034
6357         * dwarf2out.c (mem_loc_descriptor): Handle VEC_SERIES.
6359         PR rtl-optimization/82044
6360         PR tree-optimization/82042
6361         * dse.c (record_store): Check for overflow.
6362         (check_mem_read_rtx): Properly check for overflow if width == -1, call
6363         add_wild_read instead of clear_rhs_from_active_local_stores on
6364         overflow and log it into dump_file.
6366 2017-11-22  Richard Biener  <rguenther@suse.de>
6368         * gimple-iterator.c (gimple_find_edge_insert_loc): Ignore
6369         fake edges to exit when looking for a place to insert.
6370         * tree-ssa-pre.c (clear_expression_ids): Inline into callers
6371         and remove.
6372         (insert_into_preds_of_block): Commit edge insertion immediately,
6373         assert that doesn't require new BBs.
6374         (fini_pre): Release expressions.
6375         (pass_pre::execute): Shuffle things around a bit, if the fn
6376         is too large do not compute AVAIL either as this is really the
6377         quadratic bit.
6379 2017-11-22  Richard Biener  <rguenther@suse.de>
6381         PR tree-optimization/83089
6382         * tree-if-conv.c (pass_if_conversion::execute): If anything
6383         changed reset SCEV and free the number of iteration estimates.
6385 2017-11-21  Martin Sebor  <msebor@redhat.com>
6387         PR tree-optimization/82945
6388         * calls.h (warn_nonstring_bound): Remove unused function.
6390 2017-11-21  Martin Sebor  <msebor@redhat.com>
6392         PR tree-optimization/82945
6393         * builtins.c (expand_builtin_strlen): Call maybe_warn_nonstring_arg.
6394         * calls.h (maybe_warn_nonstring_arg): Declare new function.
6395         * calls.c (get_attr_nonstring_decl, maybe_warn_nonstring_arg): New
6396         functions.
6397         (initialize_argument_information): Call maybe_warn_nonstring_arg.
6398         * calls.h (get_attr_nonstring_decl): Declare new function.
6399         * doc/extend.texi (attribute nonstring): Update.
6400         * gimple-fold.c (gimple_fold_builtin_strncpy): Call
6401         get_attr_nonstring_decl and handle it.
6402         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Same.  Improve
6403         detection of nul-termination.
6404         (strlen_to_stridx): Change to a pointer.
6405         (handle_builtin_strlen, handle_builtin_stxncpy): Adjust.
6406         (pass_strlen::execute): Same.
6408 2017-11-21  Sergey Shalnov  <Sergey.Shalnov@intel.com>
6410         * config/i386/i386-opts.h (enum prefer_vector_width): Added new enum
6411         for the new option -mprefer-vector-width=[none|128|256|512].
6412         * config/i386/i386.c (ix86_target_string): remove old style options
6413         -mprefer-avx256 and make -mprefer-avx128 as alias.
6414         (ix86_option_override_internal):  Apply defaults for the
6415         -mprefer-vector-width=[128|256] option.
6416         * config/i386/i386.h (TARGET_PREFER_AVX128, TARGET_PREFER_AVX256):
6417         Implement macros to work with -mprefer-vector-width=.
6418         * config/i386/i386.opt: Implemented option
6419         -mprefer-vector-width=[none|128|256|512].
6420         * doc/invoke.texi: Documentation for
6421         -mprefer-vector-width=[none|128|256|512].
6423 2017-11-21  Pat Haugen  <pthaugen@us.ibm.com>
6425         * config/rs6000/ppc-asm.h (f50, vs50): Fix values.
6427 2017-11-21  Jonathan Wakely  <jwakely@redhat.com>
6429         * doc/invoke.texi (-Wmaybe-uninitialized): Rephrase for clarity.
6431 2017-11-21  Marc Glisse  <marc.glisse@inria.fr>
6433         * doc/generic.texi: Document POINTER_DIFF_EXPR, update
6434         POINTER_PLUS_EXPR.
6435         * cfgexpand.c (expand_debug_expr): Handle POINTER_DIFF_EXPR.
6436         * expr.c (expand_expr_real_2): Likewise.
6437         * fold-const.c (const_binop, fold_addr_of_array_ref_difference,
6438         fold_binary_loc): Likewise.
6439         * match.pd (X-X, P+(Q-P), &D-P, (P+N)-P, P-(P+N), (P+M)-(P+N),
6440         P-Q==0, -(A-B), X-Z<Y-Z, (X-Z)-(Y-Z), Z-X<Z-Y, (Z-X)-(Z-Y),
6441         (A-B)+(C-A)): New transformations for POINTER_DIFF_EXPR, based on
6442         MINUS_EXPR transformations.
6443         * optabs-tree.c (optab_for_tree_code): Handle POINTER_DIFF_EXPR.
6444         * tree-cfg.c (verify_expr, verify_gimple_assign_binary): Likewise.
6445         * tree-inline.c (estimate_operator_cost): Likewise.
6446         * tree-pretty-print.c (dump_generic_node, op_code_prio,
6447         op_symbol_code): Likewise.
6448         * tree-vect-stmts.c (vectorizable_operation): Likewise.
6449         * vr-values.c (extract_range_from_binary_expr): Likewise.
6450         * varasm.c (initializer_constant_valid_p_1): Likewise.
6451         * tree.def: New tree code POINTER_DIFF_EXPR.
6453 2017-11-21  Uros Bizjak  <ubizjak@gmail.com>
6455         * config/i386/i386.md (*bswap<mode>2_movbe): Add
6456         integer suffix to movbe mnemonic.
6457         (*bswaphi2_movbe): Ditto.
6458         (bswaphi_lowpart): Merge with *bswaphi_lowpart_1.
6460 2017-11-21  Jakub Jelinek  <jakub@redhat.com>
6462         PR c++/83045
6463         * tree-cfg.c (pass_warn_function_return::execute): Formatting fix.
6464         Also warn if seen __builtin_unreachable () call with BUILTINS_LOCATION.
6465         Use LOCATION_LOCUS when comparing against UNKNOWN_LOCATION.
6467 2017-11-21  Martin Liska  <mliska@suse.cz>
6469         * tree-inline.c (expand_call_inline): Remove not needed
6470         xstrdup_for_dump.
6472 2017-11-21  James Cowgill  <James.Cowgill@imgtec.com>
6473             Jakub Jelinek  <jakub@redhat.com>
6475         PR target/82880
6476         * config/mips/frame-header-opt.c (mips_register_frame_header_opt):
6477         Remove static keyword from f variable.
6479 2017-11-21  Jakub Jelinek  <jakub@redhat.com>
6481         PR tree-optimization/83086
6482         * gimple-ssa-store-merging.c
6483         (imm_store_chain_info::try_coalesce_bswap): Test this_n.base_addr
6484         rather than n.base_addr.
6486 2017-11-21  Martin Liska  <mliska@suse.cz>
6488         PR rtl-optimization/82044
6489         PR tree-optimization/82042
6490         * dse.c (check_mem_read_rtx): Check for overflow.
6492 2017-11-21  Aldy Hernandez  <aldyh@redhat.com>
6494         * tree-ssa-threadbackward.c (find_jump_threads_backwards): Fix
6495         typo in comment.
6497 2017-11-21  Claudiu Zissulescu  <claziss@synopsys.com>
6499         * cfgrtl.c (force_nonfallthru_and_redirect): Don't split a call
6500         and its corresponding call arg location note.
6502 2017-11-21  Claudiu Zissulescu  <claziss@synopsys.com>
6503             Andrew Burgess  <andrew.burgess@embecosm.com>
6505         * config/arc/arc-protos.h (arc_compute_frame_size): Delete
6506         declaration.
6507         (arc_return_slot_offset): Likewise.
6508         (arc_eh_return_address_location): New declaration.
6509         * config/arc/arc.c (TARGET_BUILTIN_SETJMP_FRAME_VALUE): Define.
6510         (MUST_SAVE_REGISTER): Add exception handler case.
6511         (MUST_SAVE_RETURN_ADDR): Likewise.
6512         (arc_frame_pointer_required): Likewise.
6513         (arc_frame_pointer_needed): New function.
6514         (arc_compute_frame_size): Changed.
6515         (arc_expand_prologue): Likewise.
6516         (arc_expand_epilogue): Likewise.
6517         (arc_initial_elimination_offset): Likewise.
6518         (arc_return_slot_offset): Delete.
6519         (arc_eh_return_address_location): New function.
6520         (arc_builtin_setjmp_frame_value): Likewise.
6521         * config/arc/arc.h (EH_RETURN_DATA_REGNO): Use 2 registers.
6522         (EH_RETURN_STACKADJ_RTX): Define.
6523         (EH_RETURN_HANDLER_RTX): Likewise.
6524         * config/arc/arc.md (eh_return): Delete.
6526 2017-11-21  Aldy Hernandez  <aldyh@redhat.com>
6528         * print-tree.h (debug_vec_tree): Remove prototype.
6529         * gdbinit.in (pvt): Remove macro.
6531 2017-11-21  Jakub Jelinek  <jakub@redhat.com>
6533         PR tree-optimization/83047
6534         * gimple-ssa-store-merging.c
6535         (imm_store_chain_info::output_merged_store): If the loads with the
6536         same vuse are in different basic blocks, for load_gsi pick a load
6537         location that is dominated by the other loads.
6539         PR c++/83059
6540         * config/i386/i386.c (ix86_memmodel_check): Start
6541         -Winvalid-memory-model diagnostics with lowercase letter.
6543         PR debug/82718
6544         * dwarf2out.c (dw_loc_list): If crtl->has_bb_partition, temporarily
6545         set in_cold_section_p to the partition containing loc_list->first.
6546         When seeing loc_list->last_before_switch node, update secname and
6547         perform range_across_switch second partition handling only after that.
6549         PR debug/82933
6550         * run-rtl-passes.c: Include debug.h.
6551         (run_rtl_passes): Call debug_hooks->assembly_start.
6552         * dwarf2out.c (dwarf2out_assembly_start): Return early if invoked
6553         multiple times.
6555         PR target/82981
6556         * internal-fn.c (expand_mul_overflow): Use OPTAB_WIDEN instead of
6557         OPTAB_DIRECT in calls to expand_simple_binop.
6559 2017-11-20  David Malcolm  <dmalcolm@redhat.com>
6561         PR c/81404
6562         * Makefile.in (C_COMMON_OBJS): Add c-family/known-headers.o.
6564 2017-11-20  David Malcolm  <dmalcolm@redhat.com>
6566         PR c++/72786
6567         * spellcheck.h (best_match::blithely_get_best_candidate): New
6568         accessor.
6570 2017-11-20  Jakub Jelinek  <jakub@redhat.com>
6572         * config/i386/i386.c (parse_mtune_ctrl_str): Start diagnostics
6573         with lower case letter.
6575 2017-11-20  Uros Bizjak  <ubizjak@gmail.com>
6577         * config/i386/i386.md (bswaphi2): New expander.
6578         (*bswaphi2_movbe): New insn pattern.
6579         (bswaphi -> rorhi pepehole2): New peephole pattern.
6581 2017-11-20  Jeff Law  <law@redhat.com>
6583         * Makefile.in (OBJS): Add gimple-ssa-evrp-analyze.o.
6584         * gimple-ssa-evrp-analyze.c: New file pulled from gimple-ssa-evrp.c.
6585         * gimple-ssa-evrp-analyze.h: New file pulled from gimple-ssa-evrp.c.
6586         * gimple-ssa-evrp.c: Remove bits moved into new files.  Include
6587         gimple-ssa-evrp-analyze.h.
6589         * gimple-ssa-evrp.c (evrp_dom_walker::before_dom_children): Do not
6590         set BB_VISITED here.
6591         (evrp_range_analyzer::enter): Set BB_VISITED here instead.
6593 2017-11-20  Bin Cheng  <bin.cheng@arm.com>
6595         * tree-predcom.c: Add general comment on Store-Store chains.
6596         (split_data_refs_to_components): Postpone clearing eliminate_store_p
6597         flag in component.
6598         (get_chain_last_ref_at): Rename into...
6599         (get_chain_last_write_at): ...this.
6600         (get_chain_last_write_before_load): New function.
6601         (add_ref_to_chain): Promote type of chain from CT_STORE_LOAD to
6602         CT_STORE_STORE when write reference is added.
6603         (determine_roots_comp): Support load ref in CT_STORE_STORE chains.
6604         (is_inv_store_elimination_chain): Update get_chain_last_write_at call.
6605         (initialize_root_vars_store_elim_1): Ditto.
6606         (initialize_root_vars_store_elim_2): Ditto.  Replace rhs once default
6607         definition is created.
6608         (execute_pred_commoning_chain): Support load ref in CT_STORE_STORE
6609         chain by replacing it with dominant stored value.
6611 2017-11-20  Bin Cheng  <bin.cheng@arm.com>
6613         * tree-predcom.c (add_ref_to_chain): Remove check on distance.
6615 2017-11-20  Marc Glisse  <marc.glisse@inria.fr>
6617         * vr-values.c (extract_range_from_binary_expr): Use a full range
6618         for VR_VARYING.
6620 2017-11-20  Thomas Preud'homme  <thomas.preudhomme@arm.com>
6622         * config/arm/arm.md (R4_REGNUM): Define constant.
6623         (nonsecure_call_internal): Remove r4 clobber.
6624         (nonsecure_call_value_internal): Likewise.
6625         * config/arm/thumb1.md (nonsecure_call_reg_thumb1_v5): Remove second
6626         clobber and resequence match_operands.
6627         (nonsecure_call_value_reg_thumb1_v5): Likewise.
6628         * config/arm/thumb2.md (nonsecure_call_reg_thumb2): Likewise.
6629         (nonsecure_call_value_reg_thumb2): Likewise.
6631 2017-11-20  Jakub Jelinek  <jakub@redhat.com>
6633         PR tree-optimization/78821
6634         * gimple-ssa-store-merging.c (find_bswap_or_nop_load): Give up
6635         if base is TARGET_MEM_REF.  If base is not MEM_REF, set base_addr
6636         to the address of the base rather than the base itself.
6637         (find_bswap_or_nop_1): Just use pointer comparison for vuse check.
6638         (find_bswap_or_nop_finalize): New function.
6639         (find_bswap_or_nop): Use it.
6640         (bswap_replace): Return a tree rather than bool, change first
6641         argument from gimple * to gimple_stmt_iterator, allow inserting
6642         into an empty sequence, allow ins_stmt to be NULL - then emit
6643         all stmts into gsi.  Fix up MEM_REF address gimplification.
6644         (pass_optimize_bswap::execute): Adjust bswap_replace caller.
6645         (struct store_immediate_info): Add N and INS_STMT non-static
6646         data members.
6647         (store_immediate_info::store_immediate_info): Initialize them
6648         from newly added ctor args.
6649         (merged_store_group::apply_stores): Formatting fixes.  Sort by
6650         bitpos at the end.
6651         (stmts_may_clobber_ref_p): For stores call also
6652         refs_anti_dependent_p.
6653         (gather_bswap_load_refs): New function.
6654         (imm_store_chain_info::try_coalesce_bswap): New method.
6655         (imm_store_chain_info::coalesce_immediate_stores): Use it.
6656         (split_group): Handle LROTATE_EXPR and NOP_EXPR rhs_code specially.
6657         (imm_store_chain_info::output_merged_store): Fail if number of
6658         new estimated stmts is bigger or equal than old.  Handle LROTATE_EXPR
6659         and NOP_EXPR rhs_code.
6660         (pass_store_merging::process_store): Compute n and ins_stmt, if
6661         ins_stmt is non-NULL and the store rhs is otherwise invalid, use
6662         LROTATE_EXPR rhs_code.  Pass n and ins_stmt to store_immediate_info
6663         ctor.
6664         (pass_store_merging::execute): Calculate dominators.
6666         * tree-ssa-math-opts.c (nop_stats, bswap_stats, struct symbolic_number,
6667         BITS_PER_MARKER, MARKER_MASK, MARKER_BYTE_UNKNOWN, HEAD_MARKER, CMPNOP,
6668         CMPXCHG, do_shift_rotate, verify_symbolic_number_p,
6669         init_symbolic_number, find_bswap_or_nop_load, perform_symbolic_merge,
6670         find_bswap_or_nop_1, find_bswap_or_nop, pass_data_optimize_bswap,
6671         class pass_optimize_bswap, bswap_replace,
6672         pass_optimize_bswap::execute): Moved to ...
6673         * gimple-ssa-store-merging.c: ... this file.
6674         Include optabs-tree.h.
6675         (nop_stats, bswap_stats, do_shift_rotate, verify_symbolic_number_p,
6676         init_symbolic_number, find_bswap_or_nop_load, perform_symbolic_merge,
6677         find_bswap_or_nop_1, find_bswap_or_nop, bswap_replace): Put into
6678         anonymous namespace, remove static keywords.
6679         (pass_optimize_bswap::gate): Test BITS_PER_UNIT == 8 here...
6680         (pass_optimize_bswap::execute): ... rather than here.  Formatting fix.
6682 2017-11-20  Jan Hubicka  <hubicka@ucw.cz>
6684         PR bootstrap/83062
6685         * ipa-inline.c (can_inline_edge_p): Fix typo in previous patch.
6687 2017-11-20  Aldy Hernandez  <aldyh@redhat.com>
6689         * vec.h (debug_helper): New function.
6690         (DEFINE_DEBUG_VEC): New macro.
6691         * hash-set.h (debug_helper): New function.
6692         (DEFINE_DEBUG_HASH_SET): New macro.
6693         * cfg.c (debug_slim (edge)): New function.
6694         Call DEFINE_DEBUG_VEC for edges.
6695         Call DEFINE_DEBUG_HASH_SET for edges.
6696         * cfghooks.c (debug_slim (basic_block)): New function.
6697         Call DEFINE_DEBUG_VEC for basic blocks.
6698         Call DEFINE_DEBUG_HASH_SET for basic blocks.
6699         * print-tree.c (debug_slim): New function to handle trees.
6700         Call DEFINE_DEBUG_VEC for trees.
6701         Call DEFINE_DEBUG_HASH_SET for trees.
6702         (debug (vec<tree, va_gc>) &): Remove.
6703         (debug (<vec<tree, va_gc>) *): Remove.
6704         * print-rtl.c (debug_slim): New function to handle const_rtx.
6705         Call DEFINE_DEBUG_VEC for rtx_def.
6706         Call DEFINE_DEBUG_VEC for rtx_insn.
6707         Call DEFINE_DEBUG_HASH_SET for rtx_def.
6708         Call DEFINE_DEBUG_HASH_SET for rtx_insn.
6709         * sel-sched-dump.c (debug (vec<rtx_insn *> &): Remove.
6710         (debug (vec<rtx_insn *> *ptr): Remove.
6711         (debug_insn_vector): Remove.
6712         * stor-layout.c (debug_rli): Call debug() instead of debug_vec_tree.
6714 2017-11-20  Tom de Vries  <tom@codesourcery.com>
6716         PR rtl-optimization/82020
6717         * simplify-rtx.c (simplify_ternary_operation): Fix comparison mode of
6718         IF_THEN_ELSE condition.
6720 2017-11-19  Jeff Law  <law@redhat.com>
6722         * tree-ssa-dom.c (record_equivalences_from_phis): Fix handling
6723         of degenerates resulting from ignoring an edge.
6725 2017-11-19  Jan Hubicka  <hubicka@ucw.cz>
6727         PR ipa/81360
6728         * ipa-inline.c (can_inline_edge_p): Also check that caller is optimized
6730 2017-11-19  Jan Hubicka  <hubicka@ucw.cz>
6732         PR ipa/83001
6733         * profile-count.c (profile_count::to_sreal_scale): Fix return value
6734         for uninitialied counts.
6736 2017-11-19  Jan Hubicka  <hubicka@ucw.cz>
6738         PR ipa/60243
6739         * tree-inline.c (estimate_num_insns): Set to 1 at least.
6741 2017-11-19  Jan Hubicka  <hubicka@ucw.cz>
6743         PR target/82713
6744         * i386.c (ix86_builtin_vectorization_cost): Be ready for insane types.
6746 2017-11-19  Tom de Vries  <tom@codesourcery.com>
6748         * config/arc/arc.h (FUNCTION_PROFILER): Remove semicolon after
6749         "do while (0)".
6751 2017-11-19  Tom de Vries  <tom@codesourcery.com>
6753         * config/phoenix.h (TARGET_OS_CPP_BUILTINS): Remove semicolon after
6754         "do {} while (0)".
6756 2017-11-19  Tom de Vries  <tom@codesourcery.com>
6758         * config/visium/visium.h (ASM_OUTPUT_CASE_END): Remove semicolon after
6759         macro body.
6761 2017-11-19  Tom de Vries  <tom@codesourcery.com>
6763         * config/ft32/ft32.h (REGISTER_TARGET_PRAGMAS): Remove semicolon after
6764         "do {} while (0)".
6765         * config/spu/spu.h (REGISTER_TARGET_PRAGMAS): Same.
6767 2017-11-19  Tom de Vries  <tom@codesourcery.com>
6769         * config/mcore/mcore-elf.h (MCORE_EXPORT_NAME): Remove semicolon after
6770         "do {} while (0)".
6771         * config/mcore/mcore.h (ASM_OUTPUT_ALIGNED_COMMON): After missing
6772         semicolon after MCORE_EXPORT_NAME call.
6774 2017-11-19  Tom de Vries  <tom@codesourcery.com>
6776         PR target/82961
6777         * vmsdbgout.c (vmsdbgout_early_finish): New function.
6778         (vmsdbg_debug_hooks): Set early_finish field to vmsdbgout_early_finish.
6780 2017-11-18  Jan Hubicka  <hubicka@ucw.cz>
6782         * cgraphclones.c (cgraph_edge::clone): Rename gcov_count to prof_count.
6783         (cgraph_edge::clone): Cleanup updating of profile.
6784         * ipa-cp.c (update_profiling_info): Likewise.
6785         * ipa-inline-transform.c (inline_transform): Likewise.
6786         * ipa-inline.c (inline_small_functions): Add missing space to dump.
6787         * ipa-split.c (execute_split_functions): Do not split when function
6788         is cold.
6789         * predict.c (estimate_bb_frequencies): Cleanup updating of profile.
6790         * profile-count.c (profile_count::dump): Add global0.
6791         (profile_count::to_cgraph_frequency): Do not ICE when entry is
6792         undefined.
6793         (profile_count::to_sreal_scale): Likewise.
6794         (profile_count::adjust_for_ipa_scaling): Fix typo in comment.
6795         (profile_count::combine_with_ipa_count): New function.
6796         * profile-count.h (profile_guessed_global0adjusted): New.
6797         (profile_count::adjusted_zero): New.
6798         (profile_count::global0adjusted): New.
6799         (profile_count::combine_with_ipa_count): New.
6800         * tree-inline.c (copy_edges_for_bb): Add NUM/DEN arugment;
6801         correct profile of return block of split functions.
6802         (copy_cfg_body): Remove unused profile_count.
6803         (copy_body): Likewise.
6804         (expand_call_inline): Update.
6805         (tree_function_versioning): Update.
6807 2017-11-18  Aldy Hernandez  <aldyh@redhat.com>
6809         * hash-set.h (hash_set::empty): New.
6810         * tree-ssa-threadbackward.h: Delete.
6811         * tree-ssa-threadbackward.c (class thread_jumps): New.
6812         Move max_threaded_paths into class.
6813         (fsm_find_thread_path): Remove arguments that are now in class.
6814         (profitable_jump_thread_path): Rename to...
6815         (thread_jumps::profitable_jump_thread_path): ...this.
6816         (convert_and_register_jump_thread_path): Rename to...
6817         (thread_jumps::convert_and_register_current_path): ...this.
6818         (check_subpath_and_update_thread_path): Rename to...
6819         (thread_jumps::check_subpath_and_update_thread_path): ...this.
6820         (register_jump_thread_path_if_profitable): Rename to...
6821         (thread_jumps::register_jump_thread_path_if_profitable): ...this.
6822         (handle_phi): Rename to...
6823         (thread_jumps::handle_phi): ...this.
6824         (handle_assignment): Rename to...
6825         (thread_jumps::handle_assignment): ...this.
6826         (fsm_find_control_statement_thread_paths): Rename to...
6827         (thread_jumps::fsm_find_control_statement_thread_paths): ...this.
6828         (find_jump_threads_backwards): Rename to...
6829         (thread_jumps::find_jump_threads_backwards): ...this.
6830         Initialize path local data.
6831         (pass_thread_jumps::execute): Call find_jump_threads_backwards
6832         from within thread_jumps class.
6833         (pass_early_thread_jumps::execute): Same.
6835 2017-11-17  Jan Hubicka  <hubicka@ucw.cz>
6837         * cgraph.c (cgraph_node::dump): Do IPA sanity checking on IPA counts.
6839 2017-11-17  Steve Ellcey  <sellcey@cavium.com>
6841         * config/aarch64/aarch64-simd.md (fnma<mode>4): Move neg operator
6842         to canonical location.
6844 2017-11-17  Steve Ellcey  <sellcey@cavium.com>
6846         PR target/81356
6847         * config/aarch64/aarch64.c (aarch64_use_by_pieces_infrastructure_p):
6848         Remove.
6849         (TARGET_USE_BY_PIECES_INFRASTRUCTURE_P): Remove define.
6851 2017-11-17  H.J. Lu  <hongjiu.lu@intel.com>
6853         * config/gnu-user.h (GNU_USER_TARGET_STARTFILE_SPEC): Use
6854         rcrt1.o%s/grcrt1.o%s for -static-pie.
6856 2017-11-17  Jan Hubicka  <hubicka@ucw.cz>
6858         * i386.c (ix86_multiplication_cost, ix86_division_cost,
6859         ix86_shift_rotate_cost): Break out from ...
6860         (ix86_rtx_costs): ... here.
6861         (ix86_add_stmt_cost): Use rtx cost machinery to compute cost of
6862         vector operations.
6864 2017-11-17  Jan Hubicka  <hubicka@ucw.cz>
6866         * predict.c (determine_unlikely_bbs): Set cgraph node count to 0
6867         when entry block was promoted unlikely.
6868         (estimate_bb_frequencies): Increase frequency scale.
6869         * profile-count.h (profile_count): Export precision info.
6871 2017-11-17  Jan Hubicka  <hubicka@ucw.cz>
6873         * tree-tailcall.c (eliminate_tail_call): Be more careful about not
6874         disturbin profile of entry block.
6876 2017-11-17  Jan Hubicka  <hubicka@ucw.cz>
6878         * ipa-fnsummary.c (estimate_node_size_and_time): Be more tolerant for
6879         roundoff errors.
6881 2017-11-17  Jan Hubicka  <hubicka@ucw.cz>
6883         * ipa-cp.c (update_profiling_info): Handle conversion to local profile.
6884         * tree-cfg.c (execute_fixup_cfg): Do fixup same way as inliner does.
6886 2017-11-17  Jeff Law  <law@redhat.com>
6888         * gimple-ssa-evrp.c (class evrp_range_analyzer): New class extracted
6889         from evrp_dom_walker class.  Various methods moved into new class.
6890         (evrp_range_analyzer::evrp_range_analyzer): Constructor for new class.
6891         (evrp_range_analyzer::enter): New method.
6892         (evrp_range_analyzer::leave): New method.
6893         (evrp_dom_walker): Remove delegators no longer needed by this class.
6894         Replace vr_values data member with evrp_range_analyzer
6896         * gimple-ssa-evrp.c (evrp_dom_walker::record_ranges_from_phis): New
6897         method extracted from evrp_dom_walker::before_dom_children.
6898         (evrp_dom_walker::record_ranges_from_stmt): Likewise.
6899         (evrp_dom_walker::record_ranges_from_incoming_edge): Likewise.
6901         * gimple-ssa-evrp.c (evrp_dom_walker): Add cleanup method.
6902         Add private copy constructor and move assignment operators.
6903         Privatize methods and class data where trivially possible.
6904         (evrp_dom_walker::cleanup): New function, extracted from
6905         execute_early_vrp.  Simplify access to class data.
6907         * vr-values.h (get_output_for_vrp): Prototype.
6908         * vr-values.c (get_output_for_vrp): New function extracted from
6909         vrp_visit_assignment_or_call and extract_range_from_stmt.
6910         (vrp_visit_assignment_or_call): Use get_output_for_vrp.  Simplify.
6912 2017-11-17  Luis Machado  <luis.machado@linaro.org>
6914         * config/aarch64/aarch64.c
6915         (qdf24xx_prefetch_tune) <default_opt_level>: Set to -1.
6916         (qdf24xx_tunings) <autoprefetcher_model>: Set to
6917         tune_params::AUTOPREFETCHER_WEAK.
6919 2017-11-17  Tamar Christina  <tamar.christina@arm.com>
6921         PR target/82641
6922         * config/arm/arm.c (arm_valid_target_attribute_rec):
6923         Parse "arch=" and "+<ext>".
6924         (arm_valid_target_attribute_tree): Re-init global options.
6925         (arm_option_override): Make non-static.
6926         (arm_options_perform_arch_sanity_checks): Make errors fatal.
6927         * gcc/config/arm/arm-c.c (__ARM_FEATURE_CMSE): Support undef.
6928         (__ARM_FEATURE_CRC32): Support undef.
6929         * config/arm/arm_acle.h (__ARM_FEATURE_CRC32): Replace with pragma.
6930         * doc/extend.texi (ARM Function Attributes): Add pragma and target.
6932 2017-11-17  David Malcolm  <dmalcolm@redhat.com>
6934         * gdbinit.in (break-on-diagnostic): New command.
6936 2017-11-17  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
6938         * config/i386/i386.c (ix86_expand_epilogue): Change simple
6939         return to indirect jump for EH return if control-flow
6940         protection is enabled. Change explicit 'false' argument in
6941         pro_epilogue_adjust_stack with a value of
6942         flag_cf_protection.
6943         * config/i386/i386.md (simple_return_indirect_internal):
6944         Remove SImode restriction to support 64-bit.
6946 2017-11-17  Segher Boessenkool  <segher@kernel.crashing.org>
6948         * combine.c (added_notes_insn): New.
6949         (try_combine): Handle added_notes_insn like added_links_insn.
6950         Rewrite return value code.
6951         (distribute_notes): Set added_notes_insn to the earliest insn we added
6952         a note to.
6954 2017-11-17  Segher Boessenkool  <segher@kernel.crashing.org>
6956         PR rtl-optimization/82621
6957         * combine.c (try_combine): Do not split PARALLELs of two SETs if the
6958         dest of one of those SETs is unused.
6960 2017-11-17  Richard Biener  <rguenther@suse.de>
6962         PR fortran/83017
6963         * tree-core.h (enum annot_expr_kind): Add annot_expr_parallel_kind.
6964         * tree-pretty-print.c (dump_generic_node): Handle
6965         annot_expr_parallel_kind.
6966         * tree-cfg.c (replace_loop_annotate_in_block): Likewise.
6967         * gimplify.c (gimple_boolify): Likewise.
6969 2017-11-17  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
6971         * config.gcc (extra_headers): Add cet.h for x86 targets.
6972         * config/i386/cet.h: New file.
6973         * doc/install.texi: Add --enable-cet/--disable-cet.
6975 2017-11-17  Richard Biener  <rguenther@suse.de>
6977         PR tree-optimization/83017
6978         * tree-parloops.c (MIN_PER_THREAD): Use --param parloops-min-per-thread.
6979         (gen_parallel_loop): Properly count iterations.
6980         (parallelize_loops): Handle loop->can_be_parallel independent
6981         of flag_loop_parallelize_all.  Make static profitability test match
6982         the runtime one.
6983         * params.def (PARAM_PARLOOPS_MIN_PER_THREAD): New.
6984         * invoke.texi (parloops-min-per-thread): Document.
6986 2017-11-17  Vineet Gupta  <vgupta@synopsys.com>
6988         * config/arc/linux.h: GLIBC_DYNAMIC_LINKER update per glibc
6989         upstreaming review comments.
6991 2017-11-17  Tamar Christina  <tamar.christina@arm.com>
6993         * expr.c (copy_blkmode_to_reg): Fix bitsize for targets
6994         with fast unaligned access.
6995         * doc/sourcebuild.texi (word_mode_no_slow_unalign): New.
6997 2017-11-17  Thomas Preud'homme  <thomas.preudhomme@arm.com>
6999         * config/arm/arm.c (cmse_nonsecure_entry_clear_before_return): Allocate
7000         to_clear_arg_regs_bitmap to the same size as to_clear_bitmap.
7002 2017-11-17  Richard Biener  <rguenther@suse.de>
7004         * tree-ssa-pre.c (phi_translate_1): Remove redundant constant
7005         folding of references.
7007 2017-11-17  Qing Zhao  <qing.zhao@oracle.com>
7009         PR middle-end/78809
7010         * gimple-fold.c (gimple_fold_builtin_string_compare): Add handling
7011         of replacing call to strncmp with corresponding call to strcmp when
7012         meeting conditions.
7014 2017-11-17  Sergey Shalnov  <Sergey.Shalnov@intel.com>
7016         * config/i386/x86-tune.def (X86_TUNE_AVX256_OPTIMAL): Add tuning
7017         option prefer-avx256 for skylake-avx512 configuration.
7018         * config/i386/i386.c (ix86_option_override_internal): Ditto.
7019         (get_builtin_code_for_version): Ditto.
7021 2017-11-17  Chung-Ju Wu  <jasonwucj@gmail.com>
7022             Monk Chiang  <sh.chiang04@gmail.com>
7024         * config/nds32/nds32.h (FIRST_PSEUDO_REGISTER): Modify.
7025         (FIXED_REGISTERS): Reserve more register numbers.
7026         (CALL_USED_REGISTERS): Likewise.
7027         (REG_ALLOC_ORDER): Likewise.
7028         (REG_CLASS_CONTENTS): Likewise.
7029         (REGISTER_NAMES): Likewise.
7031 2017-11-17  Chung-Ju Wu  <jasonwucj@gmail.com>
7032             Kito Cheng  <kito.cheng@gmail.com>
7034         * config/nds32/nds32-modes.def: Add vector mode V4QI V2HI V8QI V4HI
7035         V2SI.
7036         * config/nds32/iterators.md: Add vector mode iterators and attributes.
7038 2017-11-16  Steven Munroe  <munroesj@gcc.gnu.org>
7040         * config/rs6000/mmintrin.h (_mm_add_pi32[_ARCH_PWR]): Correct
7041         parameter list for vec_splats.
7043 2017-11-16  Joseph Myers  <joseph@codesourcery.com>
7045         * doc/invoke.texi (-std=c17): Refer to 2018 expected publication
7046         date of C17.
7047         (-std=c18, -std=iso9899:2018, -std=gnu18): Document option aliases.
7049 2017-11-16  Thomas Koenig  <tkoenig@gcc.gnu.org>
7051         PR bootstrap/82856
7052         * doc/install.texi: Document incompatibility of Perl >=5.6.26
7053         with the required version of automake 1.11.6.
7055 2017-11-16  Pat Haugen  <pthaugen@us.ibm.com>
7057         * rs6000/power9.md (power9fpdiv): New automaton and cpu_unit defined
7058         for it.
7059         (DU_C2_3_power9): Correct reservation combinations.
7060         (FP_DIV_power9, VEC_DIV_power9): New.
7061         (power9-alu): Split out rotate/shift...
7062         (power9-rot): ...to here, correct dispatch resource.
7063         (power9-cracked-alu, power9-mul, power9-mul-compare): Correct dispatch
7064         resource.
7065         (power9-fp): Correct latency.
7066         (power9-sdiv): Add div/sqrt resource.
7067         (power9-ddiv): Correct latency, add div/sqrt resource.
7068         (power9-sqrt, power9-dsqrt): Add div/sqrt resource.
7069         (power9-vecfdiv, power9-vecdiv): Correct latency, add div/sqrt
7070         resource.
7071         (power9-qpdiv, power9-qpmul): Adjust resource usage.
7073 2017-11-15  Michael Meissner  <meissner@linux.vnet.ibm.com>
7075         * config/rs6000/rs6000.c (rs6000_expand_builtin): Do not do the
7076         switch statement mapping KF built-ins to TF built-ins if we don't
7077         have the proper ISA 3.0 assembler support.
7079 2017-11-16  Jan Hubicka  <hubicka@ucw.cz>
7081         * tree-emutls.c (lower_emutls_data): Remove unused bb_freq.
7082         (lower_emutls_function_body): Do not compute it.
7084 2017-11-16  Jan Hubicka  <hubicka@ucw.cz>
7086         * ipa-split.c (split_bb_info): Turn time to sreal.
7087         (split_point): Likewise.
7088         (dump_split_point): Likewise.
7089         (fine_split_points): Likewise.
7090         (execute_split_functions): Only zero split_bbs; turn time to sreals.
7092 2017-11-16  Jan Hubicka  <hubicka@ucw.cz>
7094         * ipa-fnsummary.c (analyze_function_body): Accumulate time consistently
7095         in sreal.
7097 2017-11-16  Jan Hubicka  <hubicka@ucw.cz>
7099         * predict.c (combine_predictions_for_bb): Preserve zero predicted
7100         edges.
7101         (expensive_function_p): Remove useless assert.
7102         (determine_unlikely_bbs): Propagate also forward; determine cold blocks
7104 2017-11-16  Martin Sebor  <msebor@redhat.com>
7106         PR tree-optimization/82588
7107         PR tree-optimization/82583
7108         * tree-vrp.c (check_array_ref): Handle flexible array members,
7109         string literals, and inner indices.
7110         (search_for_addr_array): Add detail to diagnostics.
7112 2017-11-16  Nathan Sidwell  <nathan@acm.org>
7114         PR c++/82836
7115         PR c++/82737
7116         * tree.h (COPY_DECL_RTL): Rename parms for clarity.
7117         (SET_DECL_ASSEMBLER_NAME): Forward to
7118         overwrite_decl_assembler_name.
7119         (COPY_DECL_ASSEMBLER_NAME): Rename parms for clarity.
7120         (overwrite_decl_assembler_name): Declare.
7121         * tree.c (overwrite_decl_assembler_name): New.
7122         * langhooks-def.h (lhd_overwrite_decl_assembler_name): Declare.
7123         (LANG_HOOKS_OVERWRITE_DECL_ASSEMBLER_NAME): Provide default.
7124         (LANG_HOOKS_INITIALIZER): Add it.
7125         * langhooks.h (struct lang_hooks): Add overwrite_decl_assembler_name.
7126         * langhooks.c (lhd_set_decl_assembler_name): Use
7127         SET_DECL_ASSEMBLER_NAME.
7128         (lhd_overwrite_decl_assembler_name): Default implementation.
7130 2017-11-16  Wilco Dijkstra  <wdijkstr@arm.com>
7131             Jackson Woodruff  <jackson.woodruff@arm.com>
7133         PR tree-optimization/71026
7134         * match.pd: Canonicalize constant multiplies in division.
7136 2017-11-16  Wilco Dijkstra  <wdijkstr@arm.com>
7138         * opts.c (default_options_table): Add OPT_fomit_frame_pointer entry.
7139         * common/config/alpha/alpha-common.c (TARGET_OPTION_OPTIMIZATION_TABLE):
7140         Remove OPT_fomit_frame_pointer entry.
7141         * common/config/arc/arc-common.c: Likewise.
7142         * common/config/arm/arm-common.c: Likewise.
7143         * common/config/avr/avr-common.c: Likewise.
7144         * common/config/c6x/c6x-common.c: Likewise.
7145         * common/config/cr16/cr16-common.c: Likewise.
7146         * common/config/cris/cris-common.c: Likewise.
7147         * common/config/epiphany/epiphany-common.c: Likewise.
7148         * common/config/fr30/fr30-common.c: Likewise.
7149         * common/config/frv/frv-common.c: Likewise.
7150         * common/config/ia64/ia64-common.c: Likewise.
7151         * common/config/iq2000/iq2000-common.c: Likewise.
7152         * common/config/lm32/lm32-common.c: Likewise.
7153         * common/config/m32r/m32r-common.c: Likewise.
7154         * common/config/mcore/mcore-common.c: Likewise.
7155         * common/config/microblaze/microblaze-common.c: Likewise.
7156         * common/config/mips/mips-common.c: Likewise.
7157         * common/config/mmix/mmix-common.c: Likewise.
7158         * common/config/mn10300/mn10300-common.c: Likewise.
7159         * common/config/nios2/nios2-common.c: Likewise.
7160         * common/config/pa/pa-common.c: Likewise.
7161         * common/config/pdp11/pdp11-common.c: Likewise.
7162         * common/config/powerpcspe/powerpcspe-common.c: Likewise.
7163         * common/config/riscv/riscv-common.c: Likewise.
7164         * common/config/rs6000/rs6000-common.c: Likewise.
7165         * common/config/rx/rx-common.c: Likewise.
7166         * common/config/s390/s390-common.c: Likewise.
7167         * common/config/sh/sh-common.c: Likewise.
7168         * common/config/sparc/sparc-common.c: Likewise.
7169         * common/config/tilegx/tilegx-common.c: Likewise.
7170         * common/config/tilepro/tilepro-common.c: Likewise.
7171         * common/config/v850/v850-common.c: Likewise.
7172         * common/config/visium/visium-common.c: Likewise.
7173         * common/config/xstormy16/xstormy16-common.c: Likewise.
7174         * common/config/xtensa/xtensa-common.c: Likewise.
7175         * invoke.texi (-fomit-frame-pointer): Update documentation.
7177 2017-11-16  Jan Hubicka  <hubicka@ucw.cz>
7179         * tree-cfg.c (gimple_find_sub_bbs): Do not compute freq.
7181 2017-11-14  Jan Hubicka  <hubicka@ucw.cz>
7183         * cfg.c (scale_bbs_frequencies_int,
7184         cale_bbs_frequencies_gcov_type): Remove.
7185         * cfg.h (scale_bbs_frequencies_int,
7186         cale_bbs_frequencies_gcov_type): Remove.
7188 2017-11-14  Jan Hubicka  <hubicka@ucw.cz>
7190         * tree-ssa-loop-manip.c
7191         (scale_dominated_blocks_in_loop): Update to profile counts.
7192         (tree_transform_and_unroll_loop): Likewise.
7194 2017-11-14  Jan Hubicka  <hubicka@ucw.cz>
7196         * tree-vect-loop-manip.c (vect_do_peeling): Do not use
7197         scale_bbs_frequencies_int.
7199 2017-11-14  Jan Hubicka  <hubicka@ucw.cz>
7201         * final.c (compute_alignments): Use counts rather than frequencies.
7203 2017-11-14  Jan Hubicka  <hubicka@ucw.cz>
7205         * cfgloopanal.c: Include sreal.h
7206         (average_num_loop_insns): Use counts and sreal for accounting.
7208 2017-11-14  Jan Hubicka  <hubicka@ucw.cz>
7210         * cfgloopmanip.c (duplicate_loop_to_header_edge): Cleanup profile
7211         manipulation.
7213 2017-11-16  Chung-Ju Wu  <jasonwucj@gmail.com>
7214             Kito Cheng  <kito.cheng@gmail.com>
7216         * config/nds32/constraints.md: Provide more constraints.
7217         * config/nds32/nds32.h (enum reg_class, REG_CLASS_NAMES,
7218         REG_CLASS_CONTENTS): Define R5_REG, R8_REG, and FRAME_POINTER_REG to
7219         support constraints usage.
7221 2017-11-16  Chung-Ju Wu  <jasonwucj@gmail.com>
7223         * config/nds32/constants.md (UNSPEC_VOLATILE_FUNC_RETURN): Remove.
7225 2017-11-16  Chung-Ju Wu  <jasonwucj@gmail.com>
7226             Kito Cheng  <kito.cheng@gmail.com>
7228         * config/nds32/nds32.opt: Add mext-perf, mext-perf2, mext-string.
7229         * config/nds32/nds32.opt: Refine the layout.
7230         * config/nds32/nds32.c (TARGET_EXT_PERF, TARGET_EXT_PERF2,
7231         TARGET_EXT_STRING): Support new options.
7232         * config/nds32/nds32.h: Likewise.
7233         * config/nds32/nds32.md: Likewise.
7234         * config/nds32/nds32-predicates.c: Likewise.
7235         * config/nds32/constraints.md: Likewise.
7236         * common/config/nds32/nds32-common.c: Likewise.
7238 2017-11-16  Julia Koval  <julia.koval@intel.com>
7240         PR target/82983
7241         * config/i386/gfniintrin.h: Add sse check.
7242         * config/i386/i386.c (ix86_expand_builtin): Fix gfni check.
7244 2017-11-16  Julia Koval  <julia.koval@intel.com>
7246         * common/config/i386/i386-common.c (OPTION_MASK_ISA_AVX512VBMI2_SET,
7247         OPTION_MASK_ISA_AVX512VBMI2_UNSET): New.
7248         (ix86_handle_option): Handle -mavx512vbmi2.
7249         * config/i386/cpuid.h: Add bit_AVX512VBMI2.
7250         * config/i386/driver-i386.c (host_detect_local_cpu): Handle new bit.
7251         * config/i386/i386-c.c (__AVX512VBMI2__): New.
7252         * config/i386/i386.c (ix86_target_string): Handle -mavx512vbmi2.
7253         (ix86_valid_target_attribute_inner_p): Ditto.
7254         * config/i386/i386.h (TARGET_AVX512VBMI2, TARGET_AVX512VBMI2_P): New.
7255         * config/i386/i386.opt (mavx512vbmi2): New option.
7256         * doc/invoke.texi: Add new option.
7258 2017-11-16  Julia Koval  <julia.koval@intel.com>
7260         * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8, _mm256_gf2p8mul_epi8,
7261         _mm_mask_gf2p8mul_epi8, _mm_maskz_gf2p8mul_epi8,
7262         _mm256_mask_gf2p8mul_epi8, _mm256_maskz_gf2p8mul_epi8,
7263         _mm512_mask_gf2p8mul_epi8, _mm512_maskz_gf2p8mul_epi8,
7264         _mm512_gf2p8mul_epi8): New intrinsics.
7265         * config/i386/i386-builtin-types.def
7266         (V64QI_FTYPE_V64QI_V64QI): New type.
7267         * config/i386/i386-builtin.def (__builtin_ia32_vgf2p8mulb_v64qi,
7268         __builtin_ia32_vgf2p8mulb_v64qi_mask, __builtin_ia32_vgf2p8mulb_v32qi,
7269         __builtin_ia32_vgf2p8mulb_v32qi_mask, __builtin_ia32_vgf2p8mulb_v16qi,
7270         __builtin_ia32_vgf2p8mulb_v16qi_mask): New builtins.
7271         * config/i386/sse.md (vgf2p8mulb_*): New pattern.
7272         * config/i386/i386.c (ix86_expand_args_builtin): Handle new type.
7274 2017-11-15  Uros Bizjak  <ubizjak@gmail.com>
7276         * config/i386/i386.c (x86_print_call_or_nop): Emit 5 byte nop
7277         explicitly as a stream of bytes.
7279 2017-11-15  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
7281         * config/rs6000/altivec.h (vec_xst_be): New #define.
7282         * config/rs6000/altivec.md (altivec_vperm_<mode>_direct): Rename
7283         and externalize from *altivec_vperm_<mode>_internal.
7284         * config/rs6000/rs6000-builtin.def (XL_BE_V16QI): Remove macro
7285         instantiation.
7286         (XL_BE_V8HI): Likewise.
7287         (XL_BE_V4SI): Likewise.
7288         (XL_BE_V4SI): Likewise.
7289         (XL_BE_V2DI): Likewise.
7290         (XL_BE_V4SF): Likewise.
7291         (XL_BE_V2DF): Likewise.
7292         (XST_BE): Add BU_VSX_OVERLOAD_X macro instantiation.
7293         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Correct
7294         all array entries with these keys: VSX_BUILTIN_VEC_XL,
7295         VSX_BUILTIN_VEC_XL_BE, VSX_BUILTIN_VEC_XST.  Add entries for key
7296         VSX_BUILTIN_VEC_XST_BE.
7297         * config/rs6000/rs6000.c (altivec_expand_xl_be_builtin): Remove.
7298         (altivec_expand_builtin): Remove handling for VSX_BUILTIN_XL_BE_*
7299         built-ins.
7300         (altivec_init_builtins): Replace conditional calls to def_builtin
7301         for __builtin_vsx_ld_elemrev_{v8hi,v16qi} and
7302         __builtin_vsx_st_elemrev_{v8hi,v16qi} based on TARGET_P9_VECTOR
7303         with unconditional calls.  Remove calls to def_builtin for
7304         __builtin_vsx_le_be_<mode>.  Add a call to def_builtin for
7305         __builtin_vec_xst_be.
7306         * config/rs6000/vsx.md (vsx_ld_elemrev_v8hi): Convert define_insn
7307         to define_expand, and add alternate RTL generation for P8.
7308         (*vsx_ld_elemrev_v8hi_internal): New define_insn based on
7309         vsx_ld_elemrev_v8hi.
7310         (vsx_ld_elemrev_v16qi): Convert define_insn to define_expand, and
7311         add alternate RTL generation for P8.
7312         (*vsx_ld_elemrev_v16qi_internal): New define_insn based on
7313         vsx_ld_elemrev_v16qi.
7314         (vsx_st_elemrev_v8hi): Convert define_insn
7315         to define_expand, and add alternate RTL generation for P8.
7316         (*vsx_st_elemrev_v8hi_internal): New define_insn based on
7317         vsx_st_elemrev_v8hi.
7318         (vsx_st_elemrev_v16qi): Convert define_insn to define_expand, and
7319         add alternate RTL generation for P8.
7320         (*vsx_st_elemrev_v16qi_internal): New define_insn based on
7321         vsx_st_elemrev_v16qi.
7323 2017-11-15  H.J. Lu  <hongjiu.lu@intel.com>
7325         PR target/82990
7326         * config/i386/i386.c (pass_insert_vzeroupper::gate): Remove
7327         TARGET_AVX512ER check.
7328         (ix86_option_override_internal): Set MASK_VZEROUPPER if
7329         neither -mzeroupper nor -mno-zeroupper is used and
7330         TARGET_EMIT_VZEROUPPER is set.
7331         * config/i386/i386.h (TARGET_EMIT_VZEROUPPER): New.
7332         * config/i386/x86-tune.def: Add X86_TUNE_EMIT_VZEROUPPER.
7334 2017-11-15  Will Schmidt  <will_schmidt@vnet.ibm.com>
7336         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add support for
7337         folding of vector compares.
7338         (fold_build_vec_cmp): New helper function.
7339         (fold_compare_helper): New helper function.
7340         (builtin_function_type): Add compare builtins to the list of functions
7341         having unsigned arguments.  Cosmetic updates to comment indentation.
7342         * config/rs6000/vsx.md (vcmpneb, vcmpneh, vcmpnew): Update to specify
7343         the not+eq combination.
7345 2017-11-15  Bin Cheng  <bin.cheng@arm.com>
7347         PR tree-optimization/82726
7348         PR tree-optimization/70754
7349         * tree-predcom.c (order_drefs_by_pos): New function.
7350         (combine_chains): Move code setting has_max_use_after to...
7351         (try_combine_chains): ...here.  New parameter.  Sort combined chains
7352         according to position information.
7353         (tree_predictive_commoning_loop): Update call to above function.
7354         (update_pos_for_combined_chains, pcom_stmt_dominates_stmt_p): New.
7356 2017-11-15  Bin Cheng  <bin.cheng@arm.com>
7358         PR tree-optimization/82726
7359         Revert
7360         2017-01-23  Bin Cheng  <bin.cheng@arm.com>
7362         PR tree-optimization/70754
7363         * tree-predcom.c (stmt_combining_refs): New parameter INSERT_BEFORE.
7364         (reassociate_to_the_same_stmt): New parameter INSERT_BEFORE.  Insert
7365         combined stmt before it if not NULL.
7366         (combine_chains): Process refs reversely and compute dominance point
7367         for root ref.
7369         Revert
7370         2017-02-23  Bin Cheng  <bin.cheng@arm.com>
7372         PR tree-optimization/79663
7373         * tree-predcom.c (combine_chains): Process refs in reverse order
7374         only for ZERO length chains, and add explaining comment.
7376 2017-11-15  Tamar Christina  <tamar.christina@arm.com>
7378         * config/arm/arm-cpus.in (armv8_3, ARMv8_3a, armv8.3-a): New
7379         * config/arm/arm-tables.opt (armv8.3-a): Regenerated.
7380         * doc/invoke.texi (ARM Options): Add armv8.3-a.
7382 2017-11-15  Tamar Christina  <tamar.christina@arm.com>
7384         * config/arm/arm.h (TARGET_DOTPROD): Add arm_arch8_2.
7386 2017-11-15  Martin Liska  <mliska@suse.cz>
7388         * tree-cfg.c (pass_warn_function_return::execute):
7389         Compare warn_return_type for greater than zero.
7391 2017-11-15  Sebastian Peryt  <sebastian.peryt@intel.com>
7393         PR target/82941
7394         PR target/82942
7395         * config/i386/i386.c (pass_insert_vzeroupper): Modify gate condition
7396         to return true on Xeon and not on Xeon Phi.
7397         (ix86_check_avx256_register): Changed to ...
7398         (ix86_check_avx_upper_register): ... this. Add extra check for
7399         VALID_AVX512F_REG_OR_XI_MODE.
7400         (ix86_avx_u128_mode_needed): Changed
7401         ix86_check_avx256_register to ix86_check_avx_upper_register.
7402         (ix86_check_avx256_stores): Changed to ...
7403         (ix86_check_avx_upper_stores): ... this. Changed
7404         ix86_check_avx256_register to ix86_check_avx_upper_register.
7405         (ix86_avx_u128_mode_after): Changed
7406         avx_reg256_found to avx_upper_reg_found. Changed
7407         ix86_check_avx256_stores to ix86_check_avx_upper_stores.
7408         (ix86_avx_u128_mode_entry): Changed
7409         ix86_check_avx256_register to ix86_check_avx_upper_register.
7410         (ix86_avx_u128_mode_exit): Ditto.
7411         * config/i386/i386.h: (host_detect_local_cpu): New define.
7413 2017-11-15  Dominik Infuehr  <dominik.infuehr@theobroma-systems.com>
7415         * config/arm/xgene1.md (xgene1): Split into automatons
7416         xgene1_main, xgene1_decoder, xgene1_div, xgene1_simd.
7417         (xgene1_f_load): Adjust reservations and/or types.
7418         (xgene1_f_store): Likewise.
7419         (xgene1_load_pair): Likewise.
7420         (xgene1_store_pair): Likewise.
7421         (xgene1_fp_load1): Likewise.
7422         (xgene1_load1): Likewise.
7423         (xgene1_store1): Likewise.
7424         (xgene1_move): Likewise.
7425         (xgene1_alu): Likewise.
7426         (xgene1_simd): Likewise.
7427         (xgene1_bfm): Likewise.
7428         (xgene1_neon_load1): Likewise.
7429         (xgene1_neon_store1): Likewise.
7430         (xgene1_neon_logic): Likewise.
7431         (xgene1_neon_st1): Likewise.
7432         (xgene1_neon_ld1r): Likewise.
7433         (xgene1_alu_cond): Added.
7434         (xgene1_shift_reg): Likwise.
7435         (xgene1_bfx): Likewise.
7436         (xgene1_mul): Split into xgene1_mul32, xgene1_mul64.
7438 2017-11-15  Jakub Jelinek  <jakub@redhat.com>
7440         PR target/82981
7441         * internal-fn.c: Include gimple-ssa.h, tree-phinodes.h and
7442         ssa-iterators.h.
7443         (can_widen_mult_without_libcall): New function.
7444         (expand_mul_overflow): If only checking unsigned mul overflow,
7445         not result, and can do efficiently MULT_HIGHPART_EXPR, emit that.
7446         Don't use WIDEN_MULT_EXPR if it would involve a libcall, unless
7447         no other way works.  Add MULT_HIGHPART_EXPR + MULT_EXPR support.
7448         (expand_DIVMOD): Formatting fix.
7449         * expmed.h (expand_mult): Add NO_LIBCALL argument.
7450         * expmed.c (expand_mult): Likewise.  Use OPTAB_WIDEN rather
7451         than OPTAB_LIB_WIDEN if NO_LIBCALL is true, and allow it to fail.
7453         PR tree-optimization/82977
7454         * tree-ssa-strlen.c (strlen_optimize_stmt): Pass a reference to a copy
7455         constructed temporary to strlen_to_stridx.put.
7457 2017-11-15  Martin Liska  <mliska@suse.cz>
7459         * configure.ac: Remove -fkeep-inline-functions from coverage_flags.
7460         * configure: Regenerate.
7462 2017-11-15  Martin Liska  <mliska@suse.cz>
7464         PR target/82927
7465         * config/sh/sh-mem.cc: Use proper probability for
7466         REG_BR_PROB_NOTE.
7468 2017-11-14  Jeff Law  <law@redhat.com>
7470         * explow.c (anti_adjust_stack_and_probe_stack_clash): Avoid probing
7471         the red zone for stack_clash_protection_final_dynamic_probe targets
7472         when the total dynamic stack size is zero bytes.
7474         * tree-ssa-threadupdate.c (thread_through_all_blocks): Thread
7475         blocks is post order.
7477 2017-11-15  Alexandre Oliva  <aoliva@redhat.com>
7479         * dumpfile.h (TDF_COMPARE_DEBUG): New.
7480         * final.c (rest_of_clean_state): Set it for the
7481         -fcompare-debug dump.
7482         * tree-pretty-print.c (dump_generic_node): Omit OBJ_TYPE_REF
7483         class when TDF_COMPARE_DEBUG is set.
7485         * dwarf2out.c (gen_producer_string): Discard
7486         OPT_fcompare_debug.
7488 2017-11-15  Joseph Myers  <joseph@codesourcery.com>
7490         PR c/81156
7491         * doc/extend.texi (Other Builtins): Document __builtin_tgmath.
7492         * ginclude/tgmath.h (__tg_cplx, __tg_ldbl, __tg_dbl, __tg_choose)
7493         (__tg_choose_2, __tg_choose_3, __TGMATH_REAL_1_2)
7494         (__TGMATH_REAL_2_3): Remove macros.
7495         (__TGMATH_CPLX, __TGMATH_CPLX_2, __TGMATH_REAL, __TGMATH_REAL_2)
7496         (__TGMATH_REAL_3, __TGMATH_CPLX_ONLY): Define using
7497         __builtin_tgmath.
7498         (frexp, ldexp, nexttoward, scalbn, scalbln): Define using
7499         __TGMATH_REAL_2.
7500         (remquo): Define using __TGMATH_REAL_3.
7502 2017-11-14  Jeff Law  <law@redhat.com>
7504         * vr-values.c: New file with contents extracted from tree-vrp.c.
7505         * Makefile.in (OBJS): Add vr-values.o
7506         * tree-vrp.h (set_value_range_to_nonnull): Prototype.
7507         (set_value_range, set_and_canonicalize_value_range): Likewise.
7508         (vrp_bitmap_equal_p, range_is_nonnull): Likewise.
7509         (value_range_constant_singleton, symbolic_range_p): Likewise.
7510         (compare_values, compare_values_warnv, vrp_val_is_min): Likewise.
7511         (vrp_val_is_max, copy_value_range, set_value_range_to_value): Likewise.
7512         (extract_range_from_binary_expr_1, vrp_val_min, vrp_val_max): Likewise.
7513         (set_value_range_to_null, range_int_cst_p, opreand_less_p): Likewise.
7514         (find_case_label_range, find_case_label_index): Likewise.
7515         (zero_nonzero_bits_from_vr, overflow_comparison_p): Likewise.
7516         (range_int_cst_singleton_p, value_inside_range): Likewise.
7517         (get_single_symbol): Likewise.
7518         (switch_update): Move structure definition here.
7519         (to_remove_edges, to_update_switch_stmts): Provide externs.
7520         * tree-vrp.c: Move all methods for vr-values class to vr-values.c
7521         (vrp_val_max, vrp_val_min, vrp_val_is_max): Make externally visible.
7522         (vrp_val_is_min, set_value_range): Likewise.
7523         (set_and_canonicalize_value_range, copy_value_range): Likewise.
7524         (set_value_range_to_value, set_value_range_to_nonnull): Likewise.
7525         (set_value_range_to_null, vrp_bitmap_equal_p): Likewise.
7526         (range_is_nonnull, range_int_cst_p): Likewwise.
7527         (range_int_cst_singleton_p, symbolic_range_p): Likewise.
7528         (get_single_symbol, operand_less_p): Likewise
7529         (compare_values_warnv, compare_values): Likewise.
7530         (value_inside_range, value_range_constant_singleton): Likewise.
7531         (zero_nonzero_bitgs_from_vr): Likewise.
7532         (extract_range_from_binary_expr_1): Likewise.
7533         (overflow_comparison_p): Likewise.
7534         (to_remove_edges, to_update_switch_stmts): Likewise.
7535         (find_case_label-index, find_case_label_range): Likewise.
7536         (switch_update, set_value_range_to_nonnegative): Remove.
7537         (set_value_range_to_truthvalue): Likewise.
7538         (symbolic_range_based_on_p, gimple_assign_nonzero_p): Likewise.
7539         (gimple_stmt_nonzero_p, compare_ranges): Likewise.
7540         (compare_range_with_value, vrp_valueize, vrp_valueize_1): Likewise.
7541         (find_case_label_ranges, test_for_singularity): Likewise.
7542         (range_fits_type_p, simplify_conversion_using_ranges): LIkewise.
7543         (x_vr_values): Move to its remaining use site.
7545 2017-11-10  Jeff Law  <law@redhat.com>
7547         * vr-values.h (VR_INITIALIZER): Move #define here.
7548         * gimple-ssa-evrp.c: New file with contents extracted from tree-vrp.c
7549         * Makefile.in (OBJS): Add tree-evrp.o
7550         * tree-vrp.h (assert_info): Move structure definition here.
7551         (set_value_range_to_varying): Prototype.
7552         (vrp_operand_equal_p, range_includes_zero_p): Likewise.
7553         (infer_value_range, register_edge_assert_for): Likewise.
7554         (stmt_interesting_for_vrp): Likewise.
7555         * tree-vrp.c: Move all methods for evrp class into gimple-ssa-evrp.c.
7556         (set_value_range_to_varying): No longer static.
7557         (vrp_operand_equal_p, range_includes_zero_p): Likewise.
7558         (infer_value_range, register_edge_assert_for): Likewise.
7560 2017-11-14  Michael Meissner  <meissner@linux.vnet.ibm.com>
7562         * config/rs6000/rs6000.md (bswapdi2): On 32-bit ISA 3.0, don't
7563         generate the XXBRD instruction.
7565         * config/rs6000/rs6000-c.c (is_float128_p): New helper function.
7566         (rs6000_builtin_type_compatible): Treat _Float128 and long double
7567         as being compatible if -mabi=ieeelongdouble.
7568         * config/rs6000/rs6000-builtin.def (BU_FLOAT128_HW_1): New macros
7569         to setup float128 built-ins with hardware support.
7570         (BU_FLOAT128_HW_2): Likewise.
7571         (BU_FLOAT128_HW_3): Likewise.
7572         (BU_FLOAT128_HW_VSX_1): Likewise.
7573         (BU_FLOAT128_HW_VSX_2): Likewise.
7574         (scalar_extract_expq): Change float128 built-in functions to
7575         accommodate having both KFmode and TFmode functions.  Use the
7576         KFmode variant as the default.
7577         (scalar_extract_sigq): Likewise.
7578         (scalar_test_neg_qp): Likewise.
7579         (scalar_insert_exp_q): Likewise.
7580         (scalar_insert_exp_qp): Likewise.
7581         (scalar_test_data_class_qp): Likewise.
7582         (sqrtf128_round_to_odd): Delete processing the round to odd
7583         built-in functions as special built-in functions, and define them
7584         as float128 built-ins.  Use the KFmode variant as the default.
7585         (truncf128_round_to_odd): Likewise.
7586         (addf128_round_to_odd): Likewise.
7587         (subf128_round_to_odd): Likewise.
7588         (mulf128_round_to_odd): Likewise.
7589         (divf128_round_to_odd): Likewise.
7590         (fmaf128_round_to_odd): Likewise.
7591         * config/rs6000/rs6000.c (rs6000_expand_binop_builtin): Add
7592         support for KFmode and TFmode xststdcqp calls.
7593         (rs6000_expand_builtin): If long double is IEEE 128-bit floating
7594         point, switch the built-in handlers for the get/set float128
7595         exponent, get float128 mantissa, float128 test built-ins, and the
7596         float128 round to odd built-in functions.  Eliminate creating the
7597         float128 round to odd built-in functions as special built-ins.
7598         (rs6000_init_builtins): Eliminate special creation of the float128
7599         round to odd built-in functions.
7600         * config/rs6000/vsx.md (xsxexpqp_<mode>): Change float128 built-in
7601         function insns to support both TFmode and KFmode variants.
7602         (xsxsigqp_<mode>): Likewise.
7603         (xsiexpqpf_<mode>): Likewise.
7604         (xsiexpqp_<mode>): Likewise.
7605         (xststdcqp_<mode>): Likewise.
7606         (xststdcnegqp_<mode>): Likewise.
7607         (xststdcqp_<mode>): Likewise.
7609 2017-11-14  Jan Hubicka  <hubicka@ucw.cz>
7611         * tree-ssa-threadupdate.c (compute_path_counts): Remove
7612         unused path_in_freq_ptr parameter.
7613         (ssa_fix_duplicate_block_edges): Do not pass around path_in_freq
7615 2017-11-14  Jan Hubicka  <hubicka@ucw.cz>
7617         * ipa-inline.c (edge_badness): Dump sreal frequency.
7618         (compute_inlined_call_time): Match natural implementaiton ...
7619         * ipa-fnsummary.c (estimate_edge_size_and_time): ... here; remove
7620         forgotten division by CGRAPH_FREQ_BASE.
7622 2017-11-14  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
7624         * config.gcc (*-*-solaris2*): Enable default_use_cxa_atexit since
7625         Solaris 11.  Update comment.
7626         * configure.ac (gcc_cv_ld_pid): Adapt comment for Solaris 12
7627         renaming.
7628         * config/sol2.h (STARTFILE_SPEC): Likewise.
7629         * configure: Regenerate.
7631 2017-11-14  Carl Love  <cel@us.ibm.com>
7633         * config/rs6000/rs6000.c (swap_endian_selector_for_mode): Remove
7634         le_ and be_ prefixes to swap* variables.  Remove
7635         if (VECTOR_ELT_ORDER_BIG) statement. Remove E_V16QImode case
7636         statements.
7638 2017-11-14  Jason Merrill  <jason@redhat.com>
7640         Support GTY((cache)) on hash_map.
7641         * hash-traits.h (ggc_remove): Add ggc_maybe_mx member function.
7642         (ggc_cache_remove): Override it instead of ggc_mx.
7643         * hash-table.h (gt_ggc_mx): Call it instead of ggc_mx.
7644         (gt_cleare_cache): Call ggc_mx instead of gt_ggc_mx.
7645         * hash-map-traits.h (simple_hashmap_traits): Add maybe_mx member.
7646         (simple_cache_map_traits): Override maybe_mx.
7647         * hash-map.h (hash_entry): Add ggc_maybe_mx and keep_cache_entry.
7648         (hash_map): Friend gt_cleare_cache.
7649         (gt_cleare_cache): New.
7650         * tree.h (tree_cache_traits): New hash_map traits class.
7651         (tree_cache_map): New typedef.
7653 2017-11-14  Richard Biener  <rguenther@suse.de>
7655         * tree-cfgcleanup.c (cleanup_control_expr_graph): Remove first_p
7656         paramter and handling.
7657         (cleanup_control_flow_bb): Likewise.
7658         (cleanup_control_flow_pre): New helper performing a DFS walk
7659         to call cleanup_control_flow_bb in PRE order.
7660         (cleanup_tree_cfg_1): Do the first phase of cleanup_control_flow_bb
7661         via cleanup_control_flow_pre.
7663 2017-11-14  James Greenhalgh  <james.greenhalgh@arm.com>
7665         * config/aarch64/aarch64-simd.md
7666         (aarch64_simd_bsl<mode>_internal): Remove DImode.
7667         (*aarch64_simd_bsl<mode>_alt): Likewise.
7668         (aarch64_simd_bsldi_internal): New.
7669         (aarch64_simd_bsldi_alt): Likewise.
7671 2017-11-13  Jan Hubicka  <hubicka@ucw.cz>
7673         * tracer.c (better_p): Do not compare frequencies.
7674         * reg-stack.c (better_edge): Likewise.
7675         * shrink-wrap.c (try_shrink_wrapping): Do not convert to gcov counts
7676         and back.
7678 2017-11-13  Jan Hubicka  <hubicka@ucw.cz>
7680         * auto-profile.c (afdo_annotate_cfg): Use update_max_bb_count.
7681         * cgraphunit.c (cgraph_node::expand_thunk): Use update_max_bb_count.
7682         * ipa-utils.c (ipa_merge_profiles): Use update_max_bb_count.
7683         * lto-streamer-in.c (input_function): Use update_max_bb_count.
7684         * omp-expand.c (expand_omp_taskreg): Use update_max_bb_count.
7685         * predict.c (maybe_hot_frequency_p): Inline to ...
7686         (maybe_hot_count_p): ... here; rewrite to counts.
7687         (counts_to_freqs): Rename to ...
7688         (update_max_bb_count): ... this one.
7689         (expensive_function_p): Use counts.
7690         (estimate_bb_frequencies): Update.
7691         (rebuild_frequencies): Update.
7692         * predict.h (counts_to_freqs): Rename to ...
7693         (update_max_bb_count): ... this one.
7694         * profile.c (compute_branch_probabilities): Add debug info
7695         * tree-inline.c (expand_call_inline): Update debug info.
7696         (optimize_inline_calls): Use update_max_bb_count..
7697         (tree_function_versioning): Use update_max_bb_count..
7698         * value-prof.c (gimple_value_profile_transformations):
7699         Do not use update_max_bb_count.
7701 2017-11-13  Jan Hubicka  <hubicka@ucw.cz>
7703         * ipa-inline.c (compute_uninlined_call_time, compute_inlined_call_time):
7704         always use frequencies.
7706 2017-11-13  Jan Hubicka  <hubicka@ucw.cz>
7708         * bb-reorder.c: Remove frequencies from comments.
7709         (better_edge_p): Use profile counts.
7710         (find_traces): Dump profile counts.
7711         (rotate_loop): Use profile counts.
7712         (find_traces_1_round): Likewise.
7713         (connect_better_edge_p): Use counts instead of probabilities for
7714         reverse walk.
7715         (copy_bb_p): Drop early check for non-0 frequency.
7716         (sanitize_hot_paths): Update comments.
7718 2017-11-13  Jan Hubicka  <hubicka@ucw.cz>
7720         * ipa-split.c (struct split_point): Add count.
7721         (consider_split): Do not compute incoming frequency; compute incoming
7722         count and store it to split_point.
7723         (split_function): Set count of the call to split part correctly.
7725 2017-11-13  Carl Love  <cel@us.ibm.com>
7727         * config/rs6000/altivec.md (altivec_vsumsws_be): Add define_expand.
7729 2017-11-13  Tom Tromey  <tom@tromey.com>
7731         * doc/cpp.texi (Variadic Macros): Document __VA_OPT__.
7733 2017-11-13  Carl Love  <cel@us.ibm.com>
7735         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins):
7736         Add support for builtins:
7737         unsigned int vec_first_{,miss}_match_{,or_eos}index,
7738         vector {un,}signed {char,int,short},
7739         vector {un,}signed {char,int,short}) arguments.
7740         * config/rs6000/rs6000-builtin.def (VFIRSTMATCHINDEX,
7741         VFIRSTMATCHOREOSINDEX, VFIRSTMISMATCHINDEX, VFIRSTMISMATCHOREOSINDEX):
7742         Add BU_P9V_AV_2 expansions for the builtins.
7743         * config/rs6000/altivec.h (vec_first_match_index,
7744         vec_first_mismatch_index, vec_first_match_or_eos_index,
7745         vec_first_mismatch_or_eos_index): Add #defines for the builtins.
7746         * config/rs6000/rs6000-protos.h (bytes_in_mode): Add
7747         new extern declaration.
7748         * config/rs6000/rs6000.c (bytes_in_mode): Add new function.
7749         * config/rs6000/vsx.md (first_match_index_<mode>,
7750         first_match_or_eos_index_<mode>, first_mismatch_index_<mode>,
7751         first_mismatch_or_eos_index_<mode>): Add define expand.
7752         (vctzlsbb_<mode>): Add mode field to define_insn for vctzlsbb.
7753         * doc/extend.texi: Update the built-in documenation file for the new
7754         built-in functions.
7756 2017-11-13  Michael Meissner  <meissner@linux.vnet.ibm.com>
7758         * match.pd: Convert fminf<N>, fminf<N>x, fmax<N>, and fmax<N>x
7759         into the min/max operations for _Float<N> and _Float<N>X types.
7761 2017-11-13  Eric Botcazou  <ebotcazou@adacore.com>
7763         PR lto/81351
7764         * dwarf2out.c (do_eh_frame): New static variable.
7765         (dwarf2out_begin_prologue): Set it.
7766         (dwarf2out_frame_finish): Test it instead of dwarf2out_do_eh_frame.
7768 2017-11-13  Jan Hubicka  <hubicka@ucw.cz>
7770         * tree-ssa-coalesce.c (coalesce_cost): Fix formating.
7772         * tree-ssa-sink.c (select_best_block): Do not use frequencies.
7774 2017-11-13  Eric Botcazou  <ebotcazou@adacore.com>
7776         PR lto/81351
7777         * debug.h (dwarf2out_do_eh_frame): Declare.
7778         * dwarf2cfi.c (dwarf2out_do_eh_frame): New predicate.
7779         (dwarf2out_do_frame): Use it.
7780         (dwarf2out_do_cfi_asm): Likewise.
7781         * dwarf2out.c (dwarf2out_frame_finish): Likewise.
7782         (dwarf2out_assembly_start): Likewise.
7783         (dwarf2out_begin_prologue): Fix comment.
7784         * toplev.c (compile_file): Always call dwarf2out_frame_finish
7785         if the target needs either debug or unwind DWARF2 info.
7786         * lto-opts.c (lto_write_options): Do not save -fexceptions,
7787         -fnon-call-exceptions, -ffp-contract, -fmath-errno, -fsigned-zeros,
7788         -ftrapping-math, -ftrapv and -fwrapv.
7790 2017-11-13  Jan Hubicka  <hubicka@ucw.cz>
7792         * cgraph.c (cgraph_edge::sreal_frequency): New function.
7793         * cgraph.h (cgraph_edge::sreal_frequency): Declare.
7794         * ipa-fnsummary.c (dump_ipa_call_summary): Use sreal_frequency.
7795         (estimate_edge_size_and_time): Likewise.
7796         (ipa_merge_fn_summary_after_inlining): Likewise.
7797         * ipa-inline.c (cgraph_freq_base_rec): Remove.
7798         (compute_uninlined_call_time): Use sreal_frequency.
7799         (compute_inlined_call_time): Likewise.
7800         (ipa_inline): Do not initialize cgraph_freq_base_rec.
7801         * profile-count.c: Include sreal.h.
7802         (profile_count::to_sreal_scale): New.
7803         * profile-count.h: Forward declare sreal.
7804         (profile_count::to_sreal_scale): Declare.
7806 2017-11-13  Nathan Sidwell  <nathan@acm.org>
7808         * diagnostic.c (maybe_line_and_column): New.
7809         (diagnostic_get_location_text): Use it.
7810         (diagnostic_report_current_module): Likewise.
7811         (test_diagnostic_get_location_text): Add tests.
7813 2017-11-13  Luis Machado  <luis.machado@linaro.org>
7815         * doc/md.texi (Specifying processor pipeline description): Fix
7816         incorrect latency for the div instruction example.
7818 2017-11-13  Jakub Jelinek  <jakub@redhat.com>
7820         PR tree-optimization/78821
7821         * gimple-ssa-store-merging.c (compatible_load_p): Don't require
7822         that bit_not_p is the same.
7823         (imm_store_chain_info::coalesce_immediate_stores): Likewise.
7824         (split_group): Count precisely bit_not_p bits in each statement.
7825         (invert_op): New function.
7826         (imm_store_chain_info::output_merged_store): Use invert_op to
7827         emit BIT_XOR_EXPR with a xor_mask instead of BIT_NOT_EXPR if some
7828         but not all orig_stores have BIT_NOT_EXPR in the corresponding spots.
7830 2017-11-13  Martin Liska  <mliska@suse.cz>
7832         * gcov.c (struct coverage_info): Remove typedef of coverage_t.
7833         (struct source_info): Likewise.
7834         (add_branch_counts): Likewise.
7835         (add_line_counts): Likewise.
7836         (function_summary): Likewise.
7837         (output_intermediate_line): Likewise.
7838         (generate_results): Likewise.
7840 2017-11-13  Martin Liska  <mliska@suse.cz>
7842         * gcov.c (struct block_info): Remove typedef for block_t.
7843         (struct line_info): Likewise.
7844         (line_info::has_block): Likewise.
7845         (EXIT_BLOCK): Likewise.
7846         (unblock): Likewise.
7847         (circuit): Likewise.
7848         (get_cycles_count): Likewise.
7849         (process_file): Likewise.
7850         (read_graph_file): Likewise.
7851         (solve_flow_graph): Likewise.
7852         (find_exception_blocks): Likewise.
7853         (add_line_counts): Likewise.
7854         (accumulate_line_info): Likewise.
7855         (output_line_details): Likewise.
7857 2017-11-13  Martin Liska  <mliska@suse.cz>
7859         * gcov.c (struct arc_info): Remove typedef for arc_t.
7860         (struct line_info): Likewise.
7861         (add_branch_counts): Likewise.
7862         (output_branch_count): Likewise.
7863         (function_info::~function_info): Likewise.
7864         (circuit): Likewise.
7865         (output_intermediate_line): Likewise.
7866         (read_graph_file): Likewise.
7867         (solve_flow_graph): Likewise.
7868         (find_exception_blocks): Likewise.
7869         (add_line_counts): Likewise.
7870         (accumulate_line_info): Likewise.
7871         (output_line_details): Likewise.
7872         (output_function_details): Likewise.
7874 2017-11-13  Martin Liska  <mliska@suse.cz>
7876         * gcov.c (struct function_info): Remove typedef for function_t.
7877         (struct source_info): Likewise.
7878         (source_info::get_functions_at_location): Likewise.
7879         (solve_flow_graph): Likewise.
7880         (find_exception_blocks): Likewise.
7881         (add_line_counts): Likewise.
7882         (output_intermediate_file): Likewise.
7883         (process_file): Likewise.
7884         (generate_results): Likewise.
7885         (release_structures): Likewise.
7886         (read_graph_file): Likewise.
7887         (read_count_file): Likewise.
7888         (accumulate_line_counts): Likewise.
7889         (output_lines): Likewise.
7891 2017-11-13  Martin Liska  <mliska@suse.cz>
7893         * gcov.c (function_info::function_info): Remove num_counts
7894         and add vector<gcov_type>.
7895         (function_info::~function_info): Use the vector.
7896         (process_file): Likewise.
7897         (read_graph_file): Likewise.
7898         (read_count_file): Likewise.
7899         (solve_flow_graph): Likewise.
7901 2017-11-13  Martin Liska  <mliska@suse.cz>
7903         * gcov.c (function_info::is_artificial): New function.
7904         (process_file): Erase all artificial early.
7905         (generate_results): Skip as all artificial are already
7906         removed.
7908 2017-11-13  Martin Liska  <mliska@suse.cz>
7910         * gcov.c (read_graph_file): Store to global vector of functions.
7911         (read_count_file): Iterate the vector.
7912         (process_file): Likewise.
7913         (generate_results): Likewise.
7914         (release_structures): Likewise.
7916 2017-11-13  Jakub Jelinek  <jakub@redhat.com>
7918         PR tree-optimization/82954
7919         * gimple-ssa-store-merging.c
7920         (imm_store_chain_info::coalesce_immediate_stores): If
7921         !infof->ops[N].base_addr, split group if info->ops[N].base_addr.
7923 2017-11-13  Richard Sandiford  <richard.sandiford@linaro.org>
7925         * config/aarch64/aarch64-simd.md (aarch64_store_lane0<mode>):
7926         Upddate call to ENDIAN_LANE_N.
7927         (aarch64_<sur>dot_lane<vsi2qi>): Use aarch64_endian_lane_rtx.
7928         (aarch64_<sur>dot_laneq<vsi2qi>): Likewise.
7929         (*aarch64_simd_vec_copy_lane<mode>): Update calls to ENDIAN_LANE_N
7930         and use aarch64_endian_lane_rtx.
7931         (*aarch64_simd_vec_copy_lane_<vswap_width_name><mode>): Likewise.
7933 2017-11-12  Tom de Vries  <tom@codesourcery.com>
7935         * config/riscv/riscv.h (ASM_OUTPUT_LABELREF): Wrap in do {} while (0).
7937 2017-11-12  Tom de Vries  <tom@codesourcery.com>
7939         * config/elfos.h (ASM_OUTPUT_ASCII): Remove semicolon after macro body.
7941 2017-11-12  Tom de Vries  <tom@codesourcery.com>
7943         * config/cr16/cr16.h (ASM_OUTPUT_LABELREF): Remove semicolon after macro
7944         body.
7945         * config/powerpcspe/xcoff.h (ASM_OUTPUT_LABELREF): Same.
7946         * config/rs6000/xcoff.h (ASM_OUTPUT_LABELREF): Same.
7947         * defaults.h (ASM_OUTPUT_LABELREF): Same.
7949 2017-11-11  Martin Sebor  <msebor@redhat.com>
7951         PR c/81117
7952         * doc/extend.texi (attribute nonstring): Remove spurious argument.
7954         PR bootstrap/82948
7955         * prefic.c (translate_name): Replace strncpy with memcpy to
7956         avoid -Wstringop-truncation.
7958 2017-11-10  Jan Hubicka  <hubicka@ucw.cz>
7960         * tree-ssa-loop-im.c (execute_sm_if_changed): Do not compute freq_sum.
7962 2017-11-10  Jan Hubicka  <hubicka@ucw.cz>
7964         * predict.c (maybe_hot_frequency_p): Do not use cfun.
7966 2017-11-10  Jan Hubicka  <hubicka@ucw.cz>
7968         * tree-ssa-tail-merge.c (replace_block_by): Fix and re-enable profile
7969         merging.
7971 2017-11-10  Michael Meissner  <meissner@linux.vnet.ibm.com>
7973         * config/rs6000/rs6000.md (bswaphi2_reg): On ISA 3.0 systems,
7974         enable generating XXBRH if the value is in a vector register.
7975         (bswapsi2_reg): On ISA 3.0 systems, enable generating XXBRW if the
7976         value is in a vector register.
7977         (bswapdi2_reg): On ISA 3.0 systems, always use XXBRD to do
7978         register to register bswap64's instead of doing the GPR sequence
7979         used on previous machines.
7980         (bswapdi2_xxbrd): New insn.
7981         (bswapdi2_reg): Disallow on ISA 3.0.
7982         (register to register bswap64 splitter): Do not split the insn on
7983         ISA 3.0 systems that use XXBRD.
7985 2017-11-10  Martin Sebor  <msebor@redhat.com>
7987         PR c/81117
7988         * config/darwin-c.c (framework_construct_pathname): Replace strncpy
7989         with memcpy.
7990         (find_subframework_file): Same.
7992 2017-11-10  Jan Hubicka  <hubicka@ucw.cz>
7994         * auto-profile.c (afdo_indirect_call): Drop frequency.
7995         * cgraph.c (symbol_table::create_edge): Drop frequency argument.
7996         (cgraph_node::create_edge): Drop frequency argument.
7997         (cgraph_node::create_indirect_edge): Drop frequency argument.
7998         (cgraph_edge::make_speculative): Drop frequency arguments.
7999         (cgraph_edge::resolve_speculation): Do not update frequencies
8000         (cgraph_edge::dump_edge_flags): Do not dump frequency.
8001         (cgraph_node::dump): Check consistency in IPA mode.
8002         (cgraph_edge::maybe_hot_p): Use IPA counter.
8003         (cgraph_edge::verify_count_and_frequency): Rename to ...
8004         (cgraph_edge::verify_count): ... this one; drop frequency checking.
8005         (cgraph_node::verify_node): Update.
8006         * cgraph.h (struct cgraph_edge): Drop frequency.
8007         (cgraph_edge::frequency): New function.
8008         * cgraphbuild.c (pass_build_cgraph_edges::execute): Donot pass
8009         frequencies.
8010         (cgraph_edge::rebuild_edges): Likewise.
8011         * cgraphclones.c (cgraph_edge::clone): Scale only counts.
8012         (duplicate_thunk_for_node): Do not pass frequency.
8013         (cgraph_node::create_clone): Scale only counts.
8014         (cgraph_node::create_virtual_clone): Do not pass frequency.
8015         (cgraph_node::create_edge_including_clones): Do not pass frequency.
8016         (cgraph_node::create_version_clone): Do not pass frequency.
8017         * cgraphunit.c (cgraph_node::analyze): Do not pass frequency.
8018         (cgraph_node::expand_thunk): Do not pass frequency.
8019         (cgraph_node::create_wrapper): Do not pass frequency.
8020         * gimple-iterator.c (update_call_edge_frequencies): Do not pass
8021         frequency.
8022         * gimple-streamer-in.c (input_bb): Scale only IPA counts.
8023         * ipa-chkp.c (chkp_produce_thunks): Do not pass frequency.
8024         * ipa-cp.c (ipcp_lattice::print): Use frequency function.
8025         (gather_caller_stats): Use frequency function.
8026         (ipcp_cloning_candidate_p): Use frequency function.
8027         (ipcp_propagate_stage): Use frequency function.
8028         (get_info_about_necessary_edges): Use frequency function.
8029         (update_profiling_info): Update only IPA profile.
8030         (update_specialized_profile): Use frequency functoin.
8031         (perhaps_add_new_callers): Update only IPA profile.
8032         * ipa-devirt.c (ipa_devirt): Use IPA profile.
8033         * ipa-fnsummary.c (redirect_to_unreachable): Do not set frequrency.
8034         (dump_ipa_call_summary): Use frequency function.
8035         (estimate_edge_size_and_time): Use frequency function.
8036         (ipa_merge_fn_summary_after_inlining): Use frequency function.
8037         * ipa-inline-analysis.c (do_estimate_edge_time): Use IPA profile.
8038         * ipa-inline-transform.c (update_noncloned_frequencies): Rename to ..
8039         (update_noncloned_counts): ... ths one; scale counts only.
8040         (clone_inlined_nodes): Do not scale frequency.
8041         (inline_call): Do not pass frequency.
8042         * ipa-inline.c (compute_uninlined_call_time): Use IPA profile.
8043         (compute_inlined_call_time): Use IPA profile.
8044         (want_inline_small_function_p): Use IPA profile.
8045         (want_inline_self_recursive_call_p): Use IPA profile.
8046         (edge_badness): Use IPA profile.
8047         (lookup_recursive_calls): Use IPA profile.
8048         (recursive_inlining): Do not pass frequency.
8049         (resolve_noninline_speculation): Do not update frequency.
8050         (inline_small_functions): Collect max of IPA profile.
8051         (dump_overall_stats): Dump IPA porfile.
8052         (dump_inline_stats): Dump IPA porfile.
8053         (ipa_inline): Collect IPA stats.
8054         * ipa-inline.h (clone_inlined_nodes): Update prototype.
8055         * ipa-profile.c (ipa_propagate_frequency_1): Use frequency function.
8056         (ipa_propagate_frequency): Use frequency function.
8057         (ipa_profile): Cleanup.
8058         * ipa-prop.c (ipa_make_edge_direct_to_target): Do not pass frequency
8059         * ipa-utils.c (ipa_merge_profiles): Merge all profiles.
8060         * lto-cgraph.c (lto_output_edge): Do not stream frequency.
8061         (input_node): Do not stream frequency.
8062         (input_edge): Do not stream frequency.
8063         (merge_profile_summaries): Scale only IPA profiles.
8064         * omp-simd-clone.c (simd_clone_adjust): Do not pass frequency.
8065         * predict.c (drop_profile): Do not recompute frequency.
8066         * trans-mem.c (ipa_tm_insert_irr_call): Do not pass frequency.
8067         (ipa_tm_insert_gettmclone_call): Do not pass frequency.
8068         * tree-cfg.c (execute_fixup_cfg): Drop profile to global0 if needed.
8069         * tree-chkp.c (chkp_copy_bounds_for_assign): Do not pass frequency.
8070         * tree-emutls.c (gen_emutls_addr): Do not pass frequency.
8071         * tree-inline.c (copy_bb): Do not scale frequency.
8072         (expand_call_inline): Do not scale frequency.
8073         (tree_function_versioning): Do not scale frequency.
8074         * ubsan.c (ubsan_create_edge): Do not pass frequency.
8076 2017-11-10  Julia Koval  <julia.koval@intel.com>
8078         * config/i386/gfniintrin.h (_mm_gf2p8affine_epi64_epi8)
8079         (_mm256_gf2p8affine_epi64_epi8, _mm_mask_gf2p8affine_epi64_epi8)
8080         (_mm_maskz_gf2p8affine_epi64_epi8, _mm256_mask_gf2p8affine_epi64_epi8)
8081         (_mm256_maskz_gf2p8affine_epi64_epi8)
8082         (_mm512_mask_gf2p8affine_epi64_epi8, _mm512_gf2p8affine_epi64_epi8)
8083         (_mm512_maskz_gf2p8affine_epi64_epi8): New intrinsics.
8084         * config/i386/i386-builtin.def (__builtin_ia32_vgf2p8affineqb_v64qi)
8085         (__builtin_ia32_vgf2p8affineqb_v32qi)
8086         (__builtin_ia32_vgf2p8affineqb_v16qi): New builtins.
8087         * config/i386/sse.md (vgf2p8affineqb_<mode><mask_name>): New pattern.
8089 2017-11-10  Tamar Christina  <tamar.christina@arm.com>
8091         PR target/82641
8092         * config/arm/arm.c
8093         (arm_option_override): Refactor.
8094         (arm_option_reconfigure_globals): New.
8095         (arm_options_perform_arch_sanity_checks): New.
8096         * config/arm/arm-protos.h (arm_option_reconfigure_globals):
8097         New prototype.
8098         (arm_options_perform_arch_sanity_checks): Likewise
8100 2017-11-10  Pat Haugen  <pthaugen@us.ibm.com>
8102         * rs6000/power9.md (power9-qpdiv): Correct DFU pipe usage.
8103         (power9-qpmul): New.
8104         * rs6000/rs6000.md ("type" attr): Add qmul.
8105         (mul<mode>3, fma<mode>4_hw, *fms<mode>4_hw, *nfma<mode>4_hw,
8106         *nfms<mode>4_hw, mul<mode>3_odd, fma<mode>4_odd, *fms<mode>4_odd,
8107         *nfma<mode>4_odd, *nfms<mode>4_odd): Change type to qmul.
8109 2017-11-10  Martin Sebor  <msebor@redhat.com>
8111         PR c/81117
8112         * builtins.c (compute_objsize): Handle arrays that
8113         compute_builtin_object_size likes to fail for.  Make extern.
8114         * builtins.h (compute_objsize): Declare.
8115         (check_strncpy_sizes): New function.
8116         (expand_builtin_strncpy): Call check_strncpy_sizes.
8117         * gimple-fold.c (gimple_fold_builtin_strncpy): Implement
8118         -Wstringop-truncation.
8119         (gimple_fold_builtin_strncat): Same.
8120         * gimple.c (gimple_build_call_from_tree): Set call location.
8121         * tree-ssa-strlen.c (strlen_to_stridx): New global variable.
8122         (maybe_diag_bound_equal_length, is_strlen_related_p): New functions.
8123         (handle_builtin_stxncpy, handle_builtin_strncat): Same.
8124         (handle_builtin_strlen): Use strlen_to_stridx.
8125         (strlen_optimize_stmt): Handle flavors of strncat, strncpy, and
8126         stpncpy.
8127         Use strlen_to_stridx.
8128         (pass_strlen::execute): Release strlen_to_stridx.
8129         * doc/invoke.texi (-Wsizeof-pointer-memaccess): Document enhancement.
8130         (-Wstringop-truncation): Document new option.
8132 2017-11-10  Martin Liska  <mliska@suse.cz>
8134         PR gcov-profile/82702
8135         * gcov.c (main): Handle intermediate files in a different
8136         way.
8137         (get_gcov_intermediate_filename): New function.
8138         (output_gcov_file): Remove support of intermediate files.
8139         (generate_results): Allocate intermediate file.
8140         (release_structures): Clean-up properly fn_end.
8141         (output_intermediate_file): Start iterating with line 1.
8143 2017-11-10  Jakub Jelinek  <jakub@redhat.com>
8145         PR tree-optimization/82929
8146         * gimple-ssa-store-merging.c (struct store_immediate_info): Add
8147         ops_swapped_p non-static data member.
8148         (store_immediate_info::store_immediate_info): Clear it.
8149         (imm_store_chain_info::coalesce_immediate_stores): If swapping
8150         ops set ops_swapped_p.
8151         (count_multiple_uses): Handle ops_swapped_p.
8153 2017-11-10  Martin Liska  <mliska@suse.cz>
8155         * coverage.c (coverage_init): Stream information about
8156         support of has_unexecuted_blocks.
8157         * doc/gcov.texi: Document that.
8158         * gcov-dump.c (dump_gcov_file): Support it in gcov_dump tool.
8159         * gcov.c (read_graph_file): Likewise.
8160         (output_line_beginning): Fix a small issue with
8161         color output.
8163 2017-11-10  Bin Cheng  <bin.cheng@arm.com>
8165         * tree-predcom.c (determine_roots_comp): Avoid memory leak by freeing
8166         reference of trivial component.
8168 2017-11-10  Jakub Jelinek  <jakub@redhat.com>
8170         PR bootstrap/82916
8171         * gimple-ssa-store-merging.c
8172         (pass_store_merging::terminate_all_aliasing_chains): For
8173         gimple_store_p stmts also call refs_output_dependent_p.
8175         PR rtl-optimization/82913
8176         * compare-elim.c (try_merge_compare): Punt if def_insn is not
8177         single set.
8179 2017-11-09  Jeff Law  <law@redhat.com>
8181         * vr-values.h: New file with vr_values class.
8182         * tree-vrp.c: Include vr-values.h
8183         (vrp_value_range_pool, vrp_equiv_obstack, num_vr_values): Move static
8184         data objects into the vr_values class.
8185         (vr_value, values_propagated, vr_phi_edge_counts): Likewise.
8186         (get_value_range): Make it a member function within vr_values class.
8187         (set_defs_to_varying, update_value_range, add_equivalence): Likewise.
8188         (vrp_stmt_computes_nonzero_p, op_with_boolean_value_range_p): Likewise.
8189         (op_with_constant_singleton_value_range): Likewise.
8190         (extract_range_for_var_from_comparison_expr): Likewise.
8191         (extract_range_from_assert, extract_range_from_ssa_name): Likewise.
8192         (extract_range_from_binary_expr): Likewise.
8193         (extract_range_from_unary_expr): Likewise.
8194         (extract_range_from_cond_expr, extrat_range_from_comparison): Likewise.
8195         (check_for_binary_op_overflow, extract_range_basic): Likewise.
8196         (extract_range_from_assignment, adjust_range_with_scev): Likewise.
8197         (dump_all_value_ranges, get_vr_for_comparison): Likewise.
8198         (compare_name_with_value, compare_names): Likewise.
8199         (vrp_evaluate_conditional_warnv_with_ops_using_ranges): Likewise.
8200         (vrp_evaluate_conditional_warnv_with_ops): Likewise.  Remove prototype.
8201         (vrp_evaluate_conditional, vrp_visit_cond_stmt): Likewise.
8202         (vrp_visit_switch_stmt, extract_range_from_stmt): Likewise.
8203         (extract_range_from_phi_node): Likewise.
8204         (simplify_truth_ops_using_ranges): Likewise.
8205         (simplify_div_or_mod_using_ranges): Likewise.
8206         (simplify_min_or_max_using_ranges, simplify_abs_using_ranges): Likewise.
8207         (simplify_bit_ops_using_ranges, simplify_cond_using_ranges_1): Likewise.
8208         (simplify_cond_using_ranges_2, simplify_switch_using_ranges): Likewise.
8209         (simplify_float_conversion_using_ranges): Likewise.
8210         (simplify_internal_call_using_ranges): Likewise.
8211         (two_valued_val_range_p, simplify_stmt_using_ranges): Likewise.
8212         (vrp_visit_assignment_or_call): Likewise.  Smuggle class instance
8213         poitner via x_vr_values for calls into gimple folder.
8214         (vrp_initialize_lattice): Make this the vr_values ctor.
8215         (vrp_free_lattice): Make this the vr_values dtor.
8216         (set_vr_value): New function.
8217         (class vrp_prop): Add vr_values data member.  Add various member
8218         functions as well as member functions that delegate to vr_values.
8219         (check_array_ref): Make a member function within vrp_prop class.
8220         (search_for_addr_array, vrp_initialize): Likewise.
8221         (vrp_finalize): Likewise.  Revamp to avoid direct access to
8222         vr_value, values_propagated, etc.
8223         (check_array_bounds): Extract vrp_prop class instance pointer from
8224         walk info structure.  Use it to call member functions.
8225         (check_all_array_refs): Make a member function within vrp_prop class.
8226         Smuggle class instance pointer via walk info structure.
8227         (x_vr_values): New local static.
8228         (vrp_valueize): Use x_vr_values to get class instance.
8229         (vr_valueize_1): Likewise.
8230         (class vrp_folder): Add vr_values data member.  Add various member
8231         functions as well as member functions that delegate to vr_values.
8232         (fold_predicate_in): Make a mber fucntion within vrp_folder class.
8233         (simplify_stmt_for_jump_threading): Extract smuggled vr_values
8234         class instance from vr_values.  Use it to call member functions.
8235         (vrp_dom_walker): Add vr_values data member.
8236         (vrp_dom_walker::after_dom_children): Smuggle vr_values class
8237         instance via x_vr_values.
8238         (identify_jump_threads): Accept vr_values as argument.  Store
8239         it into the walker structure.
8240         (evrp_dom_walker): Add vr_values class data member.  Add various
8241         delegators.
8242         (evrp_dom_walker::try_find_new_range): Use vr_values data
8243         member to access the memory allocator.
8244         (evrp_dom_walker::before_dom_children): Store vr_values class
8245         instance into the vrp_folder class.
8246         (evrp_dom_walker::push_value_range): Rework to avoid direct
8247         access to num_vr_values and vr_value.
8248         (evrp_dom_walker::pop_value_range): Likewise.
8249         (execute_early_vrp): Remove call to vrp_initialize_lattice.
8250         Use vr_values to get to dump_all_value_ranges member function.
8251         Remove call to vrp_free_lattice.  Call vrp_initialize, vrp_finalize,
8252         and simplify_cond_using_ranges_2 via vrp_prop class instance.
8253         Pass vr_values class instance down to identify_jump_threads.
8254         Remove call to vrp_free_lattice.
8255         (debug_all_value_ranges): Remove.
8257         * tree-vrp.c (vrp_prop): Move class to earlier point in the file.
8258         (vrp_folder): Likewise.
8260         * tree-vrp.c (set_value_range): Do not reference vrp_equiv_obstack.
8261         Get it from the existing bitmap instead.
8262         (vrp_intersect_ranges_1): Likewise.
8264 2017-11-09  Jakub Jelinek  <jakub@redhat.com>
8266         * gimple-ssa-store-merging.c (struct store_immediate_info): Add
8267         bit_not_p field.
8268         (store_immediate_info::store_immediate_info): Add bitnotp argument,
8269         set bit_not_p to it.
8270         (imm_store_chain_info::coalesce_immediate_stores): Break group
8271         if bit_not_p is different.
8272         (count_multiple_uses, split_group,
8273         imm_store_chain_info::output_merged_store): Handle info->bit_not_p.
8274         (handled_load): Avoid multiple chained BIT_NOT_EXPRs.
8275         (pass_store_merging::process_store): Handle BIT_{AND,IOR,XOR}_EXPR
8276         result inverted using BIT_NOT_EXPR, compute bit_not_p, pass it
8277         to store_immediate_info ctor.
8279 2017-11-09  Jim Wilson  <jimw@sifive.com>
8281         * collect2.c (OBJECT_FORMAT_COFF): Remove EXTENDED_COFF support.
8282         (scan_prog_file): Likewise.
8284 2017-11-09  Jan Hubicka  <hubicka@ucw.cz>
8286         * bb-reorder.c (max_entry_frequency): Remove.
8287         (find_traces, rotate_loop, mark_bb_visited, connect_better_edge_p,
8288         connect_traces, push_to_next_round_p): Remove prototypes.
8289         (find_traces_1_round): Use counts only.
8290         (push_to_next_round_p): Likewise.
8291         (find_traces): Likewise.
8292         (rotate_loop): Likewise.
8293         (find_traces_1_round): Likewise.
8294         (connect_traces): Likewise.
8295         (edge_order): Likewise.
8297 2017-11-09  Thomas Preud'homme  <thomas.preudhomme@arm.com>
8299         * config/arm/arm.c (output_return_instruction): Add comments to
8300         indicate requirement for cmse_nonsecure_entry return to account
8301         for the size of clearing instruction output here.
8302         (thumb_exit): Likewise.
8303         * config/arm/thumb2.md (thumb2_cmse_entry_return): Fix length for
8304         return in hardfloat mode.
8306 2017-11-09  Segher Boessenkool  <segher@kernel.crashing.org>
8308         * config/rs6000/rs6000.c (machine_function): Add a bool,
8309         "toc_is_wrapped_separately".
8310         (rs6000_option_override_internal): Enable OPTION_MASK_SAVE_TOC_INDIRECT
8311         if it wasn't explicitly set or unset, we are optimizing for speed, and
8312         doing separate shrink-wrapping.
8313         (rs6000_get_separate_components): Enable the TOC component if
8314         saving the TOC register in the prologue.
8315         (rs6000_components_for_bb): Handle the TOC component.
8316         (rs6000_emit_prologue_components): Store the TOC register where needed.
8317         (rs6000_set_handled_components): Mark TOC as handled, if handled.
8318         (rs6000_emit_prologue): Don't save the TOC if that is already done.
8320 2017-11-09  Martin Jambor  <mjambor@suse.cz>
8322         * ipa-param-manipulation.c: New file.
8323         * ipa-param-manipulation.h: Likewise.
8324         * Makefile.in (OBJS): Add ipa-param-manipulation.o.
8325         (PLUGIN_HEADERS): Addded ipa-param-manipulation.h
8326         * ipa-param.h (ipa_parm_op): Moved to ipa-param-manipulation.h.
8327         (ipa_parm_adjustment): Likewise.
8328         (ipa_parm_adjustment_vec): Likewise.
8329         (ipa_get_vector_of_formal_parms): Moved declaration to
8330         ipa-param-manipulation.h.
8331         (ipa_get_vector_of_formal_parm_types): Likewise.
8332         (ipa_modify_formal_parameters): Likewise.
8333         (ipa_modify_call_arguments): Likewise.
8334         (ipa_combine_adjustments): Likewise.
8335         (ipa_dump_param_adjustments): Likewise.
8336         (ipa_modify_expr): Likewise.
8337         (ipa_get_adjustment_candidate): Likewise.
8338         * ipa-prop.c (ipa_get_vector_of_formal_parms): Moved to
8339         ipa-param-manipulation.c.
8340         (ipa_get_vector_of_formal_parm_types): Likewise.
8341         (ipa_modify_formal_parameters): Likewise.
8342         (ipa_modify_call_arguments): Likewise.
8343         (ipa_modify_expr): Likewise.
8344         (get_ssa_base_param): Likewise.
8345         (ipa_get_adjustment_candidate): Likewise.
8346         (index_in_adjustments_multiple_times_p): Likewise.
8347         (ipa_combine_adjustments): Likewise.
8348         (ipa_dump_param_adjustments): Likewise.
8349         * tree-sra.c: Also include ipa-param-manipulation.h
8350         * omp-simd-clone.c: Include ipa-param-manipulation.h instead of
8351         ipa-param.h.
8353 2017-11-09  Richard Sandiford  <richard.sandiford@linaro.org>
8354             Alan Hayward  <alan.hayward@arm.com>
8355             David Sherwood  <david.sherwood@arm.com>
8357         * doc/sourcebuild.texi (vect_masked_store): Document.
8359 2017-11-09  Richard Sandiford  <richard.sandiford@linaro.org>
8360             Alan Hayward  <alan.hayward@arm.com>
8361             David Sherwood  <david.sherwood@arm.com>
8363         * doc/sourcebuild.texi (vect_align_stack_vars): Document.
8365 2017-11-09  Richard Sandiford  <richard.sandiford@linaro.org>
8366             Alan Hayward  <alan.hayward@arm.com>
8367             David Sherwood  <david.sherwood@arm.com>
8369         * doc/sourcebuild.texi (vect_variable_length): Document.
8371 2017-11-09  Richard Sandiford  <richard.sandiford@linaro.org>
8372             Alan Hayward  <alan.hayward@arm.com>
8373             David Sherwood  <david.sherwood@arm.com>
8375         * doc/sourcebuild.texi (vect_unaligned_possible): Document.
8377 2017-11-09  Richard Sandiford  <richard.sandiford@linaro.org>
8378             Alan Hayward  <alan.hayward@arm.com>
8379             David Sherwood  <david.sherwood@arm.com>
8381         * doc/sourcebuild.texi (vect_element_align_preferred): Document.
8383 2017-11-09  Richard Sandiford  <richard.sandiford@linaro.org>
8384             Alan Hayward  <alan.hayward@arm.com>
8385             David Sherwood  <david.sherwood@arm.com>
8387         * doc/sourcebuild.texi (vect_perm_short, vect_perm_byte): Document
8388         previously undocumented selectors.
8389         (vect_perm3_byte, vect_perm3_short, vect_perm3_int): Document.
8391 2017-11-09  Richard Sandiford  <richard.sandiford@linaro.org>
8393         * doc/rtl.texi (const_vector): Say that elements can be
8394         const_wide_ints too.
8395         * emit-rtl.h (valid_for_const_vec_duplicate_p): Declare.
8396         * emit-rtl.c (valid_for_const_vec_duplicate_p): New function.
8397         (gen_vec_duplicate): Use it instead of CONSTANT_P.
8398         * optabs.c (expand_vector_broadcast): Likewise.
8400 2017-11-09  Richard Sandiford  <richard.sandiford@linaro.org>
8401             Alan Hayward  <alan.hayward@arm.com>
8402             David Sherwood  <david.sherwood@arm.com>
8404         * tree-ssa-loop-ivopts.c (get_address_cost): Try using a
8405         scaled index even if the unscaled address was invalid.
8406         Don't increase the complexity of using a scale in that case.
8408 2017-11-09  Richard Sandiford  <richard.sandiford@linaro.org>
8409             Alan Hayward  <alan.hayward@arm.com>
8410             David Sherwood  <david.sherwood@arm.com>
8412         * doc/rtl.texi: Rewrite the subreg rules so that they partition
8413         the inner register into REGMODE_NATURAL_SIZE bytes rather than
8414         UNITS_PER_WORD bytes.
8415         * emit-rtl.c (validate_subreg): Divide subregs into blocks
8416         based on REGMODE_NATURAL_SIZE of the inner mode.
8417         (gen_lowpart_common): Split the SCALAR_FLOAT_MODE_P and
8418         !SCALAR_FLOAT_MODE_P cases.  Use REGMODE_NATURAL_SIZE for the latter.
8419         * expmed.c (lowpart_bit_field_p): Divide the value up into
8420         chunks of REGMODE_NATURAL_SIZE rather than UNITS_PER_WORD.
8421         * expr.c (store_constructor): Use REGMODE_NATURAL_SIZE to test
8422         whether something is likely to occupy more than one register.
8424 2017-11-09  Jan Hubicka  <hubicka@ucw.cz>
8426         PR ipa/82879
8427         * ipa-inline-transform.c (update_noncloned_frequencies): Use
8428         profile_count::adjust_for_ipa_scaling.
8429         * tree-inline.c (copy_bb, copy_cfg_body): Likewise.
8430         * profile-count.c (profile_count::adjust_for_ipa_scaling): New member
8431         function.
8432         * profile-count.h (profile_count::adjust_for_ipa_scaling): Declare.
8434 2017-11-09  Jakub Jelinek  <jakub@redhat.com>
8436         * gimple-ssa-store-merging.c (count_multiple_uses): New function.
8437         (split_group): Add total_orig and total_new arguments, estimate the
8438         number of statements related to the store group without store merging
8439         and with store merging.
8440         (imm_store_chain_info::output_merged_store): Adjust split_group
8441         callers, punt if estimated number of statements with store merging
8442         is not smaller than estimated number of statements without it.
8443         Formatting fix.
8444         (handled_load): Remove has_single_use checks.
8445         (pass_store_merging::process_store): Likewise.
8447 2017-11-09  Richard Biener  <rguenther@suse.de>
8449         PR tree-optimization/82902
8450         * tree-ssa-phiprop.c (propagate_with_phi): Test proper type.
8452 2017-11-09  Martin Liska  <mliska@suse.cz>
8454         PR target/82863
8455         * emit-rtl.c (init_emit_regs): Initialize split_branch_probability to
8456         uninitialized.
8458 2017-11-09  Martin Liska  <mliska@suse.cz>
8460         PR tree-optimization/82669
8461         * sbitmap.h (bmp_iter_set_init): Remove non needed check.
8463 2017-11-09  Martin Liska  <mliska@suse.cz>
8465         PR gcov-profile/48463
8466         * coverage.c (coverage_begin_function): Output also end locus
8467         of a function and information whether the function is
8468         artificial.
8469         * gcov-dump.c (tag_function): Parse and print the information.
8470         * gcov.c (INCLUDE_MAP): Add include.
8471         (INCLUDE_SET): Likewise.
8472         (struct line_info): Move earlier in the source file because
8473         of vector<line_info> in function_info structure.
8474         (line_info::line_info): Likewise.
8475         (line_info::has_block): Likewise.
8476         (struct source_info): Add new member index.
8477         (source_info::get_functions_at_location): New function.
8478         (function_info::group_line_p): New function.
8479         (output_intermediate_line): New function.
8480         (output_intermediate_file): Use the mentioned function.
8481         (struct function_start): New.
8482         (struct function_start_pair_hash): Likewise.
8483         (process_file): Add code that identifies group functions.
8484         Assign lines either to global or function scope.
8485         (generate_results): Skip artificial functions.
8486         (find_source): Assign index for each source file.
8487         (read_graph_file): Read new flag artificial and end_line.
8488         (add_line_counts): Assign it either to global of function scope.
8489         (accumulate_line_counts): Isolate core of the function to
8490         accumulate_line_info and call it for both function and global
8491         scope lines.
8492         (accumulate_line_info): New function.
8493         (output_line_beginning): Fix GNU coding style.
8494         (print_source_line): New function.
8495         (output_line_details): Likewise.
8496         (output_function_details): Likewise.
8497         (output_lines): Iterate both source (global) scope and function
8498         scope.
8499         (struct function_line_start_cmp): New class.
8500         * doc/gcov.texi: Reflect changes in documentation.
8502 2017-11-09  Jakub Jelinek  <jakub@redhat.com>
8504         PR debug/82837
8505         * dwarf2out.c (const_ok_for_output_1): Reject NEG in addition to NOT.
8506         (mem_loc_descriptor): Handle (const (neg (...))) as (neg (const (...)))
8507         and similarly for not instead of neg.
8509 2017-11-08  Andi Kleen  <ak@linux.intel.com>
8511         * config/i386/i386.opt: Add -mforce-indirect-call.
8512         * config/i386/predicates.md: Check for flag_force_indirect_call.
8513         * doc/invoke.texi: Document -mforce-indirect-call
8515 2017-11-08  Kito Cheng  <kito.cheng@gmail.com>
8517         * config/riscv/riscv-protos.h (riscv_slow_unaligned_access_p):
8518         New extern.
8519         (MOVE_RATIO): Use riscv_slow_unaligned_access_p.
8520         config/riscv/riscv.c (predict.h): New include.
8521         (riscv_slow_unaligned_access_p): No longer static.
8522         (riscv_block_move_straight): Add require.
8523         * config/riscv/riscv-protos.h (riscv_hard_regno_nregs): Delete.
8525 2017-11-08  Jakub Jelinek  <jakub@redhat.com>
8527         PR target/82855
8528         * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>,
8529         <avx512>_eq<mode>3<mask_scalar_merge_name>_1): Use
8530         nonimmediate_operand predicate for operand 1 instead of
8531         register_operand.
8533 2017-11-08  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
8535         * config/aarch64/aarch64-simd.md (store_pair_lanes<mode>):
8536         New pattern.
8537         * config/aarch64/constraints.md (Uml): New constraint.
8538         * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): New
8539         predicate.
8541 2017-11-08  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
8543         * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
8544         of two vec_duplicates into a vec_concat.
8546 2017-11-08  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
8548         * simplify-rtx.c (simplify_ternary_operation, VEC_MERGE):
8549         Simplify vec_merge of vec_duplicate and vec_concat.
8550         * config/aarch64/constraints.md (Utq): New constraint.
8551         * config/aarch64/aarch64-simd.md (load_pair_lanes<mode>): New
8552         define_insn.
8554 2017-11-08  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
8556         * simplify-rtx.c (simplify_ternary_operation, VEC_MERGE):
8557         Simplify vec_merge of vec_duplicate and const_vector.
8558         * config/aarch64/predicates.md (aarch64_simd_or_scalar_imm_zero):
8559         New predicate.
8560         * config/aarch64/aarch64-simd.md (*aarch64_combinez<mode>): Use VDC
8561         mode iterator.  Update predicate on operand 1 to
8562         handle non-const_vec constants.  Delete constraints.
8563         (*aarch64_combinez_be<mode>): Likewise for operand 2.
8565 2017-11-08  Jakub Jelinek  <jakub@redhat.com>
8567         PR tree-optimization/78821
8568         * gimple-ssa-store-merging.c (struct store_operand_info): Add bit_not_p
8569         data member.
8570         (store_operand_info::store_operand_info): Initialize it to false.
8571         (pass_store_merging::terminate_all_aliasing_chains): Rewritten to use
8572         ref_maybe_used_by_stmt_p and stmt_may_clobber_ref_p on lhs of each
8573         store in the group, and if chain_info is non-NULL, to ignore altogether
8574         that chain.
8575         (compatible_load_p): Fail if bit_not_p does not match.
8576         (imm_store_chain_info::output_merged_store): Handle bit_not_p loads.
8577         (handled_load): Fill in bit_not_p.  Handle BIT_NOT_EXPR.
8578         (pass_store_merging::process_store): Adjust
8579         terminate_all_aliasing_chains calls to pass NULL in all current spots,
8580         call terminate_all_aliasing_chains newly when adding a store into
8581         a chain with non-NULL chain_info.
8583 2017-11-08  Wilco Dijkstra  <wdijkstr@arm.com>
8585         * config/aarch64/aarch64.c (aarch64_can_eliminate): Simplify logic.
8587 2017-11-08  Wilco Dijkstra  <wdijkstr@arm.com>
8589         * config/aarch64/aarch64.c (aarch64_frame_pointer_required)
8590         Remove.
8591         (aarch64_layout_frame): Initialise emit_frame_chain.
8592         (aarch64_can_eliminate): Remove omit leaf frame pointer code.
8593         (TARGET_FRAME_POINTER_REQUIRED): Remove define.
8595 2017-11-08  Martin Liska  <mliska@suse.cz>
8597         * gimplify.c (expand_FALLTHROUGH_r): Simplify usage
8598         of gimple_call_internal_p.
8600 2017-11-07  Tom de Vries  <tom@codesourcery.com>
8602         * config/mips/mips.h (ASM_OUTPUT_LABELREF): Wrap in "do {} while (0)".
8604 2017-11-07  Tom de Vries  <tom@codesourcery.com>
8606         * config/mips/mips.h (ASM_OUTPUT_CASE_END): Remove semicolon after
8607         "do {} while (0)".
8609 2017-11-08  Martin Liska  <mliska@suse.cz>
8611         PR sanitizer/82792
8612         * gimplify.c (expand_FALLTHROUGH_r): Skip IFN_ASAN_MARK.
8614 2017-11-07  Eric Botcazou  <ebotcazou@adacore.com>
8616         * gimple-pretty-print.c (dump_profile): Return "" instead of NULL.
8618 2017-11-07  Jakub Jelinek  <jakub@redhat.com>
8620         PR target/82855
8621         * config/i386/i386.md (SWI1248_AVX512BWDQ2_64): New mode iterator.
8622         (*cmp<mode>_ccz_1): New insn with $k alternative.
8624         PR target/82855
8625         * config/i386/i386.c (ix86_swap_binary_operands_p): Treat
8626         RTX_COMM_COMPARE as commutative as well.
8627         (ix86_binary_operator_ok): Formatting fix.
8628         * config/i386/sse.md (*mul<mode>3<mask_name><round_name>,
8629         *<code><mode>3<mask_name><round_saeonly_name>,
8630         *<code><mode>3<mask_name>, *<code>tf3, *mul<mode>3<mask_name>,
8631         *<s>mul<mode>3_highpart<mask_name>,
8632         *vec_widen_umult_even_v16si<mask_name>,
8633         *vec_widen_umult_even_v8si<mask_name>,
8634         *vec_widen_umult_even_v4si<mask_name>,
8635         *vec_widen_smult_even_v16si<mask_name>,
8636         *vec_widen_smult_even_v8si<mask_name>, *sse4_1_mulv2siv2di3<mask_name>,
8637         *avx2_pmaddwd, *sse2_pmaddwd, *<sse4_1_avx2>_mul<mode>3<mask_name>,
8638         *avx2_<code><mode>3, *avx512f_<code><mode>3<mask_name>,
8639         *sse4_1_<code><mode>3<mask_name>, *<code>v8hi3,
8640         *sse4_1_<code><mode>3<mask_name>, *<code>v16qi3, *avx2_eq<mode>3,
8641         <avx512>_eq<mode>3<mask_scalar_merge_name>_1, *sse4_1_eqv2di3,
8642         *sse2_eq<mode>3, <mask_codefor><code><mode>3<mask_name>,
8643         *<code><mode>3, *<sse2_avx2>_uavg<mode>3<mask_name>,
8644         *<ssse3_avx2>_pmulhrsw<mode>3<mask_name>, *ssse3_pmulhrswv4hi3): Use
8645         !(MEM_P (operands[1]) && MEM_P (operands[2])) condition instead of
8646         ix86_binary_operator_ok.  Formatting fixes.
8647         (*<plusminus_insn><mode>3<mask_name><round_name>,
8648         *<plusminus_insn><mode>3, *<plusminus_insn><mode>3_m): Formatting
8649         fixes.
8651 2017-11-07  Segher Boessenkool  <segher@kernel.crashing.org>
8653         * config/rs6000/rs6000.md (GPR2): New mode_iterator.
8654         ("cstore<mode>4"): Don't always expand with rs6000_emit_int_cmove for
8655         eq and ne if TARGET_ISEL.
8656         (cmp): New code_iterator.
8657         (UNS, UNSU_, UNSIK): New code_attrs.
8658         (<code><GPR:mode><GPR2:mode>2_isel): New define_insn_and_split.
8659         ("eq<mode>3"): New define_expand, rename the define_insn_and_split
8660         to...
8661         ("eq<mode>3"): ... this.
8662         ("ne<mode>3"): New define_expand, rename the define_insn_and_split
8663         to...
8664         ("ne<mode>3"): ... this.
8666 2017-11-07  Julia Koval  <julia.koval@intel.com>
8668         PR target/82812
8669         * common/config/i386/i386-common.c
8670         (OPTION_MASK_ISA_GENERAL_REGS_ONLY_UNSET): Remove MPX from flag.
8671         (ix86_handle_option): Move MPX to isa_flags2 and GFNI to isa_flags.
8672         * config/i386/i386-c.c (ix86_target_macros_internal): Ditto.
8673         * config/i386/i386.opt: Ditto.
8674         * config/i386/i386.c (ix86_target_string): Ditto.
8675         (ix86_option_override_internal): Ditto.
8676         (ix86_init_mpx_builtins): Move MPX to args2.
8677         (ix86_expand_builtin): Special handling for OPTION_MASK_ISA_GFNI.
8678         * config/i386/i386-builtin.def (__builtin_ia32_vgf2p8affineinvqb_v64qi,
8679         __builtin_ia32_vgf2p8affineinvqb_v64qi_mask,
8680         __builtin_ia32_vgf2p8affineinvqb_v32qi,
8681         __builtin_ia32_vgf2p8affineinvqb_v32qi_mask,
8682         __builtin_ia32_vgf2p8affineinvqb_v16qi,
8683         __builtin_ia32_vgf2p8affineinvqb_v16qi_mask): Move to ARGS array.
8685 2017-11-07  Uros Bizjak  <ubizjak@gmail.com>
8687         PR target/80425
8688         * config/i386.i386.md (*zero_extendsidi2): Change (?r,*Yj), (?*Yi,r)
8689         and (*x,m) to ($r,Yj), ($Yi,r) and ($x,m).
8690         (zero-extendsidi peephole2): Remove peephole.
8692 2017-11-07  Eric Botcazou  <ebotcazou@adacore.com>
8694         PR c/53037
8695         * stor-layout.c: Include attribs.h.
8696         (handle_warn_if_not_align): Replace test on TYPE_USER_ALIGN with
8697         explicit lookup of "aligned" attribute.
8699 2017-11-07  Andrew Waterman  <andrew@sifive.com>
8701         * config/riscv/riscv-protos.h (riscv_hard_regno_nregs): New prototype.
8702         (riscv_expand_block_move): Likewise.
8703         * config/riscv/riscv.h (MOVE_RATIO): Tune cost to movmemsi
8704         implementation.
8705         (RISCV_MAX_MOVE_BYTES_PER_LOOP_ITER): New define.
8706         (RISCV_MAX_MOVE_BYTES_STRAIGHT): New define.
8707         * config/riscv/riscv.c (riscv_block_move_straight): New function.
8708         (riscv_adjust_block_mem): Likewise.
8709         (riscv_block_move_loop): Likewise.
8710         (riscv_expand_block_move): Likewise.
8711         * config/riscv/riscv.md (movmemsi): New pattern.
8713 2017-11-07  Michael Clark  <michaeljclark@mac.com>
8715         * config/riscv/linux.h (MUSL_ABI_SUFFIX): New define.
8716         (MUSL_DYNAMIC_LINKER): Likewise.
8718 2017-11-07  Richard Sandiford  <richard.sandiford@linaro.org>
8720         * config/aarch64/aarch64.md (ashl<mode>3, ashr<mode>3, lshr<mode>3)
8721         (rotr<mode>3, rotl<mode>3): Use aarch64_reg_or_imm instead of
8722         nonmmory_operand.
8724 2017-11-07  Richard Biener  <rguenther@suse.de>
8726         * match.pd: Fix build.
8728 2017-11-07  Wilco Dijkstra  <wdijkstr@arm.com>
8729             Jackson Woodruff  <jackson.woodruff@arm.com>
8731         PR tree-optimization/71026
8732         * match.pd: Canonicalize negate in division.
8734 2017-11-07  Sudakshina Das  <sudi.das@arm.com>
8736         PR middle-end/80131
8737         * match.pd: Simplify 1 << (C - x) where C = precision (x) - 1.
8739 2017-11-07  Marc Glisse  <marc.glisse@inria.fr>
8741         * match.pd ((a&~b)|(a^b),(a&~b)^~a,(a|b)&~(a^b),a|~(a^b),
8742         (a|b)|(a&^b),(a&b)|~(a^b),~(~a&b),~X^Y): New transformations.
8744 2017-11-07  Marc Glisse  <marc.glisse@inria.fr>
8746         * fold-const.c (negate_expr_p) [PLUS_EXPR, MINUS_EXPR]: Handle
8747         non-scalar integral types.
8748         * match.pd (negate_expr_p): Handle MINUS_EXPR.
8749         (-(A-B), -(~A)): New transformations.
8751 2017-11-07  Tom de Vries  <tom@codesourcery.com>
8753         * config/powerpcspe/aix43.h (SUBTARGET_OVERRIDE_OPTIONS): Remove
8754         semicolon after "do {} while (0)".
8755         * config/powerpcspe/aix51.h (SUBTARGET_OVERRIDE_OPTIONS): Same.
8756         * config/powerpcspe/aix52.h (SUBTARGET_OVERRIDE_OPTIONS): Same.
8757         * config/powerpcspe/aix53.h (SUBTARGET_OVERRIDE_OPTIONS): Same.
8758         * config/powerpcspe/aix61.h (SUBTARGET_OVERRIDE_OPTIONS): Same.
8759         * config/powerpcspe/aix71.h (SUBTARGET_OVERRIDE_OPTIONS): Same.
8761 2017-11-07  Tom de Vries  <tom@codesourcery.com>
8763         * config/rs6000/aix43.h (SUBTARGET_OVERRIDE_OPTIONS): Remove semicolon
8764         after "do {} while (0)".
8765         * config/rs6000/aix51.h (SUBTARGET_OVERRIDE_OPTIONS): Same.
8766         * config/rs6000/aix52.h (SUBTARGET_OVERRIDE_OPTIONS): Same.
8767         * config/rs6000/aix53.h (SUBTARGET_OVERRIDE_OPTIONS): Same.
8768         * config/rs6000/aix61.h (SUBTARGET_OVERRIDE_OPTIONS): Same.
8769         * config/rs6000/aix71.h (SUBTARGET_OVERRIDE_OPTIONS): Same.
8771 2017-11-07  Tom de Vries  <tom@codesourcery.com>
8773         PR other/82784
8774         * config/arm/arm.c (HANDLE_NARROW_SHIFT_ARITH): Remove semicolon after
8775         "while {} do (0)".
8776         (arm_rtx_costs_internal): Add missing semicolon after
8777         HANDLE_NARROW_SHIFT_ARITH call.
8779 2017-11-06  Segher Boessenkool  <segher@kernel.crashing.org>
8781         * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
8782         disable isel if it was not set explicitly.
8784 2017-11-06  James Bowman  <james.bowman@ftdichip.com>
8786         * gcc/dwarf2out.c (modified_type_die): Retain ADDR_SPACE qualifiers.
8787         (add_type_attribute) likewise.
8789 2017-11-06  H.J. Lu  <hongjiu.lu@intel.com>
8791         * config/i386/i386.c (ix86_can_use_return_insn_p): Use reference
8792         of struct ix86_frame.
8793         (ix86_initial_elimination_offset): Likewise.
8794         (ix86_expand_split_stack_prologue): Likewise.
8796 2017-11-06  Marc Glisse  <marc.glisse@inria.fr>
8798         * tree-vrp.h (enum value_range_type): Update stale comment.
8800 2017-11-06  Richard Sandiford  <richard.sandiford@linaro.org>
8801             Alan Hayward  <alan.hayward@arm.com>
8802             David Sherwood  <david.sherwood@arm.com>
8804         * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm)
8805         (aarch64_expand_vec_perm_const): Take the number of units too.
8806         * config/aarch64/aarch64.c (aarch64_expand_vec_perm)
8807         (aarch64_expand_vec_perm_const): Likewise.
8808         * config/aarch64/aarch64-simd.md (vec_perm_const<mode>)
8809         (vec_perm<mode>): Update accordingly.
8811 2017-11-06  Richard Sandiford  <richard.sandiford@linaro.org>
8812             Alan Hayward  <alan.hayward@arm.com>
8813             David Sherwood  <david.sherwood@arm.com>
8815         * config/aarch64/aarch64-protos.h (aarch64_simd_vect_par_cnst_half):
8816         Take the number of units too.
8817         * config/aarch64/aarch64.c (aarch64_simd_vect_par_cnst_half): Likewise.
8818         (aarch64_simd_check_vect_par_cnst_half): Update call accordingly,
8819         but check for a vector mode before rather than after the call.
8820         * config/aarch64/aarch64-simd.md (aarch64_split_simd_mov<mode>)
8821         (move_hi_quad_<mode>, vec_unpack<su>_hi_<mode>)
8822         (vec_unpack<su>_lo_<mode, vec_widen_<su>mult_lo_<mode>)
8823         (vec_widen_<su>mult_hi_<mode>, vec_unpacks_lo_<mode>)
8824         (vec_unpacks_hi_<mode>, aarch64_saddl2<mode>, aarch64_uaddl2<mode>)
8825         (aarch64_ssubl2<mode>, aarch64_usubl2<mode>, widen_ssum<mode>3)
8826         (widen_usum<mode>3, aarch64_saddw2<mode>, aarch64_uaddw2<mode>)
8827         (aarch64_ssubw2<mode>, aarch64_usubw2<mode>, aarch64_sqdmlal2<mode>)
8828         (aarch64_sqdmlsl2<mode>, aarch64_sqdmlal2_lane<mode>)
8829         (aarch64_sqdmlal2_laneq<mode>, aarch64_sqdmlsl2_lane<mode>)
8830         (aarch64_sqdmlsl2_laneq<mode>, aarch64_sqdmlal2_n<mode>)
8831         (aarch64_sqdmlsl2_n<mode>, aarch64_sqdmull2<mode>)
8832         (aarch64_sqdmull2_lane<mode>, aarch64_sqdmull2_laneq<mode>)
8833         (aarch64_sqdmull2_n<mode>): Update accordingly.
8835 2017-11-06  Richard Sandiford  <richard.sandiford@linaro.org>
8836             Alan Hayward  <alan.hayward@arm.com>
8837             David Sherwood  <david.sherwood@arm.com>
8839         * config/aarch64/aarch64-protos.h (aarch64_reverse_mask): Take
8840         the number of units too.
8841         * config/aarch64/aarch64.c (aarch64_reverse_mask): Likewise.
8842         * config/aarch64/aarch64-simd.md (vec_load_lanesoi<mode>)
8843         (vec_store_lanesoi<mode>, vec_load_lanesci<mode>)
8844         (vec_store_lanesci<mode>, vec_load_lanesxi<mode>)
8845         (vec_store_lanesxi<mode>): Update accordingly.
8847 2017-11-06  Richard Sandiford  <richard.sandiford@linaro.org>
8848             Alan Hayward  <alan.hayward@arm.com>
8849             David Sherwood  <david.sherwood@arm.com>
8851         * config/aarch64/aarch64-protos.h (aarch64_endian_lane_rtx): Declare.
8852         * config/aarch64/aarch64.c (aarch64_endian_lane_rtx): New function.
8853         * config/aarch64/aarch64.h (ENDIAN_LANE_N): Take the number
8854         of units rather than the mode.
8855         * config/aarch64/iterators.md (nunits): New mode attribute.
8856         * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args):
8857         Use aarch64_endian_lane_rtx instead of GEN_INT (ENDIAN_LANE_N ...).
8858         * config/aarch64/aarch64-simd.md (aarch64_dup_lane<mode>)
8859         (aarch64_dup_lane_<vswap_width_name><mode>, *aarch64_mul3_elt<mode>)
8860         (*aarch64_mul3_elt_<vswap_width_name><mode>): Likewise.
8861         (*aarch64_mul3_elt_to_64v2df, *aarch64_mla_elt<mode>): Likewise.
8862         (*aarch64_mla_elt_<vswap_width_name><mode>, *aarch64_mls_elt<mode>)
8863         (*aarch64_mls_elt_<vswap_width_name><mode>, *aarch64_fma4_elt<mode>)
8864         (*aarch64_fma4_elt_<vswap_width_name><mode>):: Likewise.
8865         (*aarch64_fma4_elt_to_64v2df, *aarch64_fnma4_elt<mode>): Likewise.
8866         (*aarch64_fnma4_elt_<vswap_width_name><mode>): Likewise.
8867         (*aarch64_fnma4_elt_to_64v2df, reduc_plus_scal_<mode>): Likewise.
8868         (reduc_plus_scal_v4sf, reduc_<maxmin_uns>_scal_<mode>): Likewise.
8869         (reduc_<maxmin_uns>_scal_<mode>): Likewise.
8870         (*aarch64_get_lane_extend<GPI:mode><VDQQH:mode>): Likewise.
8871         (*aarch64_get_lane_zero_extendsi<mode>): Likewise.
8872         (aarch64_get_lane<mode>, *aarch64_mulx_elt_<vswap_width_name><mode>)
8873         (*aarch64_mulx_elt<mode>, *aarch64_vgetfmulx<mode>): Likewise.
8874         (aarch64_sq<r>dmulh_lane<mode>, aarch64_sq<r>dmulh_laneq<mode>)
8875         (aarch64_sqrdml<SQRDMLH_AS:rdma_as>h_lane<mode>): Likewise.
8876         (aarch64_sqrdml<SQRDMLH_AS:rdma_as>h_laneq<mode>): Likewise.
8877         (aarch64_sqdml<SBINQOPS:as>l_lane<mode>): Likewise.
8878         (aarch64_sqdml<SBINQOPS:as>l_laneq<mode>): Likewise.
8879         (aarch64_sqdml<SBINQOPS:as>l2_lane<mode>_internal): Likewise.
8880         (aarch64_sqdml<SBINQOPS:as>l2_laneq<mode>_internal): Likewise.
8881         (aarch64_sqdmull_lane<mode>, aarch64_sqdmull_laneq<mode>): Likewise.
8882         (aarch64_sqdmull2_lane<mode>_internal): Likewise.
8883         (aarch64_sqdmull2_laneq<mode>_internal): Likewise.
8884         (aarch64_vec_load_lanesoi_lane<mode>): Likewise.
8885         (aarch64_vec_store_lanesoi_lane<mode>): Likewise.
8886         (aarch64_vec_load_lanesci_lane<mode>): Likewise.
8887         (aarch64_vec_store_lanesci_lane<mode>): Likewise.
8888         (aarch64_vec_load_lanesxi_lane<mode>): Likewise.
8889         (aarch64_vec_store_lanesxi_lane<mode>): Likewise.
8890         (aarch64_simd_vec_set<mode>): Update use of ENDIAN_LANE_N.
8891         (aarch64_simd_vec_setv2di): Likewise.
8893 2017-11-06  Carl Love  <cel@us.ibm.com>
8895         * config/rs6000/rs6000-c.c (P8V_BUILTIN_VEC_REVB): Add power 8
8896         definitions.
8897         (P9V_BUILTIN_VEC_REVB): Remove the power 9 instance definitions.
8898         * config/rs6000/altivec.h (vec_revb): Change the #define from power 9
8899         to power 8.
8900         * config/rs6000/r6000-protos.h (swap_endian_selector_for_mode): Add new
8901         extern declaration.
8902         * config/rs6000/rs6000.c (swap_endian_selector_for_mode): Add function.
8903         * config/rs6000/rs6000-builtin.def (BU_P8V_VSX_1, BU_P8V_OVERLOAD_1):
8904         Add power 8 macro expansions.
8905         (BU_P9V_OVERLOAD_1): Remove power 9 overload expansion.
8906         * config/rs6000/vsx.md (revb_<mode>): Add define_expand to generate
8907         power 8 instructions.  (VSX_XXBR): Add iterator.
8909 2017-11-06  Wilco Dijkstra  <wdijkstr@arm.com>
8911         * config/arm/arm.md (predicable_short_it): Change default to "no",
8912         improve documentation, remove uses that are identical to the default.
8913         (enabled_for_depr_it): Rename to enabled_for_short_it.
8914         * gcc/config/arm/arm-fixed.md (predicable_short_it):
8915         Remove default uses.
8916         * gcc/config/arm/ldmstm.md (predicable_short_it): Likewise.
8917         * gcc/config/arm/sync.md (predicable_short_it): Likewise.
8918         * gcc/config/arm/thumb2.md (predicable_short_it): Likewise.
8919         * gcc/config/arm/vfp.md (predicable_short_it): Likewise.
8921 2017-11-06  Michael Meissner  <meissner@linux.vnet.ibm.com>
8923         PR target/82748
8924         * config/rs6000/rs6000-builtin.def (BU_FLOAT128_1): Delete
8925         float128 helper macros, which are no longer used after deleting
8926         the old 'q' built-in functions, and moving the round to odd
8927         built-in functions to being special built-in functions.
8928         (BU_FLOAT128_2): Likewise.
8929         (BU_FLOAT128_1_HW): Likewise.
8930         (BU_FLOAT128_2_HW): Likewise.
8931         (BU_FLOAT128_3_HW): Likewise.
8932         (FABSQ): Delete old 'q' built-in functions.
8933         (COPYSIGNQ): Likewise.
8934         (SQRTF128_ODD): Move round to odd built-in functions to be
8935         special built-in functions, so that we can handle
8936         -mabi=ieeelongdouble.
8937         (TRUNCF128_ODD): Likewise.
8938         (ADDF128_ODD): Likewise.
8939         (SUBF128_ODD): Likewise.
8940         (MULF128_ODD): Likewise.
8941         (DIVF128_ODD): Likewise.
8942         (FMAF128_ODD): Likewise.
8943         * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Map old 'q'
8944         built-in names to 'f128'.
8945         * config/rs6000/rs6000.c (rs6000_fold_builtin): Remove folding the
8946         old 'q' built-in functions, as the machine independent code for
8947         'f128' built-in functions handles this.
8948         (rs6000_expand_builtin): Add expansion for float128 round to odd
8949         functions, keying off on -mabi=ieeelongdouble of whether to use
8950         the KFmode or TFmode variant.
8951         (rs6000_init_builtins): Initialize the _Float128 round to odd
8952         built-in functions.
8953         * doc/extend.texi (PowerPC Built-in Functions): Document the old
8954         _Float128 'q' built-in functions are now mapped into the new
8955         'f128' built-in functions.
8957 2017-11-06  David Edelsohn  <dje.gcc@gmail.com>
8959         * collect2.c (add_lto_object): Compile for OBJECT_COFF.
8960         (scan_prog_file): Don't skip PASS_LTOINFO. Scan for LTO objects.
8962 2017-11-06  David Malcolm  <dmalcolm@redhat.com>
8964         PR jit/82826
8965         * ipa-fnsummary.c (ipa_fnsummary_c_finalize): New function.
8966         * ipa-fnsummary.h (ipa_fnsummary_c_finalize): New decl.
8967         * toplev.c: Include "ipa-fnsummary.h".
8968         (toplev::finalize): Call ipa_fnsummary_c_finalize.
8970 2017-11-06  Jakub Jelinek  <jakub@redhat.com>
8972         PR tree-optimization/82838
8973         * gimple-ssa-store-merging.c
8974         (imm_store_chain_info::output_merged_store): Call force_gimple_operand_1
8975         on a separate gimple_seq which is then appended to seq.
8977 2017-11-06  Jeff Law  <law@redhat.com>
8979         PR target/82788
8980         * config/i386/i386.c (PROBE_INTERVAL): Remove.
8981         (get_probe_interval): New functions.
8982         (ix86_adjust_stack_and_probe_stack_clash): Use get_probe_interval.
8983         (ix86_adjust_stack_and_probe): Likewise.
8984         (output_adjust_stack_and_probe): Likewise.
8985         (ix86_emit_probe_stack_range): Likewise.
8986         (ix86_expand_prologue): Likewise.
8988 2017-11-06  Richard Sandiford  <richard.sandiford@linaro.org>
8990         PR tree-optimization/82816
8991         * tree-ssa-math-opts.c (convert_mult_to_widen): Return false
8992         if the modes of the two types are the same.
8993         (convert_plusminus_to_widen): Likewise.
8995 2017-11-06  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
8997         * config/rs6000/altivec.md (*p9_vadu<mode>3) Rename to
8998         p9_vadu<mode>3.
8999         (usadv16qi): New define_expand.
9000         (usadv8hi): New define_expand.
9002 2017-11-06  Jan Hubicka  <hubicka@ucw.cz>
9004         PR bootstrap/82832
9005         * ipa-inline-transform.c (update_noncloned_frequencies): Always
9006         scale.
9007         (inline_transform): Likewise.
9008         * predict.c (counts_to_freqs): Remove useless conditional.
9009         * profile-count.h (profile_count::apply_scale): Move sanity check.
9010         * tree-inline.c (copy_bb): Always scale.
9011         (copy_cfg_body): Likewise.
9013 2017-11-06  Christophe Lyon  <christophe.lyon@linaro.org>
9015         PR target/67591
9016         * config/arm/arm.md (*sub_shiftsi): Add predicable_short_it
9017         attribute.
9018         (*cmp_ite0): Add enabled_for_depr_it attribute.
9019         (*cmp_ite1): Likewise.
9021 2017-11-06  Segher Boessenkool  <segher@kernel.crashing.org>
9023         * config/rs6000/rs6000.c (rs6000_insn_cost): Handle TYPE_MFCR and
9024         TYPE_MFCRF.
9026 2017-11-06  Richard Sandiford  <richard.sandiford@linaro.org>
9028         * tree-vrp.c (vrp_int_const_binop): Return true on success and
9029         return the value by pointer.
9030         (extract_range_from_multiplicative_op_1): Update accordingly.
9031         Return as soon as an operation fails.
9033 2017-11-05  Tom de Vries  <tom@codesourcery.com>
9035         PR other/82784
9036         * asan.c (DEF_SANITIZER_BUILTIN_1): Factor out of ...
9037         (DEF_SANITIZER_BUILTIN): ... here.
9038         (initialize_sanitizer_builtins): Use DEF_SANITIZER_BUILTIN_1 instead of
9039         DEF_SANITIZER_BUILTIN in if stmt.  Add missing semicolon.
9041 2017-11-05  Tom de Vries  <tom@codesourcery.com>
9043         PR other/82784
9044         * config/elfos.h (ASM_OUTPUT_BEFORE_CASE_LABEL): Remove semicolon after
9045         macro body.
9046         (ASM_OUTPUT_CASE_LABEL): Add semicolon after
9047         ASM_OUTPUT_BEFORE_CASE_LABEL call.
9048         * config/arc/arc.h (ASM_OUTPUT_BEFORE_CASE_LABEL): Remove semicolon
9049         after macro body.
9050         * config/m68k/m68kelf.h (ASM_OUTPUT_BEFORE_CASE_LABEL): Same.
9051         * config/mips/mips.h (ASM_OUTPUT_BEFORE_CASE_LABEL): Same.
9052         * config/v850/v850.h (ASM_OUTPUT_BEFORE_CASE_LABEL): Same.
9054 2017-11-05  Tom de Vries  <tom@codesourcery.com>
9056         PR other/82784
9057         * graphite-scop-detection.c (DEBUG_PRINT): Remove semicolon after
9058         "do {} while (0)".
9060 2017-11-04  Michael Clark  <michaeljclark@mac.com>
9062         * config/riscv/riscv.c (riscv_print_operand): Add a 'i' format.
9063         config/riscv/riscv.md (addsi3): Use 'i' for immediates.
9064         (adddi3): Likewise.
9065         (*addsi3_extended): Likewise.
9066         (*addsi3_extended2): Likewise.
9067         (<optab>si3): Likewise.
9068         (<optab>di3): Likewise.
9069         (<optab><mode>3): Likewise.
9070         (<*optabe>si3_internal): Likewise.
9071         (zero_extendqi<SUPERQI:mode>2): Likewise.
9072         (*add<mode>hi3): Likewise.
9073         (*xor<mode>hi3): Likewise.
9074         (<optab>di3): Likewise.
9075         (*<optab>si3_extend): Likewise.
9076         (*sge<u>_<X:mode><GPR:mode>): Likewise.
9077         (*slt<u>_<X:mode><GPR:mode>): Likewise.
9078         (*sle<u>_<X:mode><GPR:mode>): Likewise.
9080 2017-11-04  Andrew Waterman  <andrew@sifive.com>
9082         * config/riscv/riscv.c (riscv_option_override): Conditionally set
9083         TARGET_STRICT_ALIGN based upon -mtune argument.
9085 2017-11-04  Andrew Waterman  <andrew@sifive.com>
9087         * config/riscv/riscv.h (SLOW_BYTE_ACCESS): Change to 1.
9089 2017-11-04  Daniel Santos  <daniel.santos@pobox.com>
9091         * config/i386/i386.c (choose_basereg): Use optional scratch
9092         register and add assertion.
9093         (x86_emit_outlined_ms2sysv_save): Use scratch register when
9094         needed, and don't allocate stack.
9095         (ix86_expand_prologue): Rearrange where SSE saves/stub call is
9096         emitted, correct wrong allocation with -mcall-ms2sysv-xlogues.
9097         (ix86_emit_outlined_ms2sysv_restore): Fix non-immediate offsets.
9099 2017-11-03  Jeff Law  <law@redhat.com>
9101         * config/i386/i386.c (ix86_emit_restore_reg_using_pop): Prototype.
9102         (ix86_adjust_stack_and_probe_stack_clash): Use a push/pop sequence
9103         to probe at the start of a noreturn function.
9105 2017-11-03  Jakub Jelinek  <jakub@redhat.com>
9107         PR tree-optimization/78821
9108         * gimple-ssa-store-merging.c: Update the file comment.
9109         (MAX_STORE_ALIAS_CHECKS): Define.
9110         (struct store_operand_info): New type.
9111         (store_operand_info::store_operand_info): New constructor.
9112         (struct store_immediate_info): Add rhs_code and ops data members.
9113         (store_immediate_info::store_immediate_info): Add rhscode, op0r
9114         and op1r arguments to the ctor, initialize corresponding data members.
9115         (struct merged_store_group): Add load_align_base and load_align
9116         data members.
9117         (merged_store_group::merged_store_group): Initialize them.
9118         (merged_store_group::do_merge): Update them.
9119         (merged_store_group::apply_stores): Pick the constant for
9120         encode_tree_to_bitpos from one of the two operands, or skip
9121         encode_tree_to_bitpos if neither operand is a constant.
9122         (class pass_store_merging): Add process_store method decl.  Remove
9123         bool argument from terminate_all_aliasing_chains method decl.
9124         (pass_store_merging::terminate_all_aliasing_chains): Remove
9125         var_offset_p argument and corresponding handling.
9126         (stmts_may_clobber_ref_p): New function.
9127         (compatible_load_p): New function.
9128         (imm_store_chain_info::coalesce_immediate_stores): Terminate group
9129         if there is overlap and rhs_code is not INTEGER_CST.  For
9130         non-overlapping stores terminate group if rhs is not mergeable.
9131         (get_alias_type_for_stmts): Change first argument from
9132         auto_vec<gimple *> & to vec<gimple *> &.  Add IS_LOAD, CLIQUEP and
9133         BASEP arguments.  If IS_LOAD is true, look at rhs1 of the stmts
9134         instead of lhs.  Compute *CLIQUEP and *BASEP in addition to the
9135         alias type.
9136         (get_location_for_stmts): Change first argument from
9137         auto_vec<gimple *> & to vec<gimple *> &.
9138         (struct split_store): Remove orig_stmts data member, add orig_stores.
9139         (split_store::split_store): Create orig_stores rather than orig_stmts.
9140         (find_constituent_stmts): Renamed to ...
9141         (find_constituent_stores): ... this.  Change second argument from
9142         vec<gimple *> * to vec<store_immediate_info *> *, push pointers
9143         to info structures rather than the statements.
9144         (split_group): Rename ALLOW_UNALIGNED argument to
9145         ALLOW_UNALIGNED_STORE, add ALLOW_UNALIGNED_LOAD argument and handle
9146         it.  Adjust find_constituent_stores caller.
9147         (imm_store_chain_info::output_merged_store): Handle rhs_code other
9148         than INTEGER_CST, adjust split_group, get_alias_type_for_stmts and
9149         get_location_for_stmts callers.  Set MR_DEPENDENCE_CLIQUE and
9150         MR_DEPENDENCE_BASE on the MEM_REFs if they are the same in all stores.
9151         (mem_valid_for_store_merging): New function.
9152         (handled_load): New function.
9153         (pass_store_merging::process_store): New method.
9154         (pass_store_merging::execute): Use process_store method.  Adjust
9155         terminate_all_aliasing_chains caller.
9157 2017-11-03  Wilco Dijkstra  <wdijkstr@arm.com>
9159         * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
9160         Return true for more constants, symbols and label references.
9161         (aarch64_valid_floating_const): Remove unused function.
9163 2017-11-03  Wilco Dijkstra  <wdijkstr@arm.com>
9165         PR target/82786
9166         * config/aarch64/aarch64.c (aarch64_layout_frame):
9167         Undo forcing of LR at bottom of frame.
9169 2017-11-03  Jeff Law  <law@redhat.com>
9171         PR target/82823
9172         * config/i386/i386.c (ix86_expand_prologue): Tighten assert
9173         for int_registers_saved.
9175         * cfganal.c (single_pred_edge_ignoring_loop_edges): New function
9176         extracted from tree-ssa-dom.c.
9177         * cfganal.h (single_pred_edge_ignoring_loop_edges): Prototype.
9178         * tree-ssa-dom.c (single_incoming_edge_ignoring_loop_edges): Remove.
9179         (record_equivalences_from_incoming_edge): Add additional argument
9180         to single_pred_edge_ignoring_loop_edges call.
9181         * tree-ssa-uncprop.c (single_incoming_edge_ignoring_loop_edges): Remove.
9182         (uncprop_dom_walker::before_dom_children): Add additional argument
9183         to single_pred_edge_ignoring_loop_edges call.
9184         * tree-ssa-sccvn.c (sccvn_dom_walker::before_dom_children): Use
9185         single_pred_edge_ignoring_loop_edges rather than open coding.
9186         * tree-vrp.c (evrp_dom_walker::before_dom_children): Similarly.
9188 2017-11-03  Marc Glisse  <marc.glisse@inria.fr>
9190         * match.pd (-(-A)): Rewrite.
9192 2017-11-03  Segher Boessenkool  <segher@kernel.crashing.org>
9194         * config/rs6000/rs60000-protos.h (rs6000_emit_sISEL): Delete.
9195         (rs6000_emit_int_cmove): New declaration.
9196         * config/rs6000/rs6000.c (rs6000_emit_int_cmove): Delete declaration.
9197         (rs6000_emit_sISEL): Delete.
9198         (rs6000_emit_int_cmove): Make non-static.
9199         * config/rs6000/rs6000.md (cstore<mode>4): Use rs6000_emit_int_cmove
9200         instead of rs6000_emit_sISEL.
9202 2017-11-03  Jan Hubicka  <hubicka@ucw.cz>
9204         * asan.c (create_cond_insert_point): Maintain profile.
9205         * ipa-utils.c (ipa_merge_profiles): Be sure only IPA profiles are
9206         merged.
9207         * basic-block.h (struct basic_block_def): Remove frequency.
9208         (EDGE_FREQUENCY): Use to_frequency
9209         * bb-reorder.c (push_to_next_round_p): Use only IPA counts for global
9210         heuristics.
9211         (find_traces): Update to use to_frequency.
9212         (find_traces_1_round): Likewise; use only IPA counts.
9213         (bb_to_key): Likewise.
9214         (connect_traces): Use IPA counts only.
9215         (copy_bb_p): Update to use to_frequency.
9216         (fix_up_crossing_landing_pad): Likewise.
9217         (sanitize_hot_paths): Likewise.
9218         * bt-load.c (basic_block_freq): Likewise.
9219         * cfg.c (init_flow): Set count_max to uninitialized.
9220         (check_bb_profile): Remove frequencies; check counts.
9221         (dump_bb_info): Do not dump frequencies.
9222         (update_bb_profile_for_threading): Update counts only.
9223         (scale_bbs_frequencies_int): Likewise.
9224         (MAX_SAFE_MULTIPLIER): Remove.
9225         (scale_bbs_frequencies_gcov_type): Update counts only.
9226         (scale_bbs_frequencies_profile_count): Update counts only.
9227         (scale_bbs_frequencies): Update counts only.
9228         * cfg.h (struct control_flow_graph): Add count-max.
9229         (update_bb_profile_for_threading): Update prototype.
9230         * cfgbuild.c (find_bb_boundaries): Do not update frequencies.
9231         (find_many_sub_basic_blocks): Likewise.
9232         * cfgcleanup.c (try_forward_edges): Likewise.
9233         (try_crossjump_to_edge): Likewise.
9234         * cfgexpand.c (expand_gimple_cond): Likewise.
9235         (expand_gimple_tailcall): Likewise.
9236         (construct_init_block): Likewise.
9237         (construct_exit_block): Likewise.
9238         * cfghooks.c (verify_flow_info): Check consistency of counts.
9239         (dump_bb_for_graph): Do not dump frequencies.
9240         (split_block_1): Do not update frequencies.
9241         (split_edge): Do not update frequencies.
9242         (make_forwarder_block): Do not update frequencies.
9243         (duplicate_block): Do not update frequencies.
9244         (account_profile_record): Do not update frequencies.
9245         * cfgloop.c (find_subloop_latch_edge_by_profile): Use IPA counts
9246         for global heuristics.
9247         * cfgloopanal.c (average_num_loop_insns): Update to use to_frequency.
9248         (expected_loop_iterations_unbounded): Use counts only.
9249         * cfgloopmanip.c (scale_loop_profile): Simplify.
9250         (create_empty_loop_on_edge): Simplify
9251         (loopify): Simplify
9252         (duplicate_loop_to_header_edge): Simplify
9253         * cfgrtl.c (force_nonfallthru_and_redirect): Update profile.
9254         (update_br_prob_note): Take care of removing note when profile
9255         becomes undefined.
9256         (relink_block_chain): Do not dump frequency.
9257         (rtl_account_profile_record): Use to_frequency.
9258         * cgraph.c (symbol_table::create_edge): Convert count to ipa count.
9259         (cgraph_edge::redirect_call_stmt_to_calle): Conver tcount to ipa count.
9260         (cgraph_update_edges_for_call_stmt_node): Likewise.
9261         (cgraph_edge::verify_count_and_frequency): Update.
9262         (cgraph_node::verify_node): Temporarily disable frequency verification.
9263         * cgraphbuild.c (compute_call_stmt_bb_frequency): Use
9264         to_cgraph_frequency.
9265         (cgraph_edge::rebuild_edges): Convert to ipa counts.
9266         * cgraphunit.c (init_lowered_empty_function): Do not initialize
9267         frequencies.
9268         (cgraph_node::expand_thunk): Update profile.
9269         * except.c (dw2_build_landing_pads): Do not update frequency.
9270         * final.c (compute_alignments): Use to_frequency.
9271         (dump_basic_block_info): Do not dump frequency.
9272         * gimple-pretty-print.c (dump_profile): Do not dump frequency.
9273         (dump_gimple_bb_header): Do not dump frequency.
9274         * gimple-ssa-isolate-paths.c (isolate_path): Do not update frequency;
9275         do update count.
9276         * gimple-streamer-in.c (input_bb): Do not stream frequency.
9277         * gimple-streamer-out.c (output_bb): Do not stream frequency.
9278         * haifa-sched.c (sched_pressure_start_bb): Use to_freuqency.
9279         (init_before_recovery): Do not update frequency.
9280         (sched_create_recovery_edges): Do not update frequency.
9281         * hsa-gen.c (convert_switch_statements): Do not update frequency.
9282         * ipa-cp.c (ipcp_propagate_stage): Update search for max_count.
9283         (ipa_cp_c_finalize): Set max_count to uninitialized.
9284         * ipa-fnsummary.c (get_minimal_bb): Use counts.
9285         (param_change_prob): Use counts.
9286         * ipa-profile.c (ipa_profile_generate_summary): Do not summarize
9287         local profiles.
9288         * ipa-split.c (consider_split): Use to_frequency.
9289         (split_function): Use to_frequency.
9290         * ira-build.c (loop_compare_func): Likewise.
9291         (mark_loops_for_removal): Likewise.
9292         (mark_all_loops_for_removal): Likewise.
9293         * loop-doloop.c (doloop_modify): Do not update frequency.
9294         * loop-unroll.c (unroll_loop_runtime_iterations): Do not update
9295         frequency.
9296         * lto-streamer-in.c (input_function): Update count_max.
9297         * omp-expand.c (expand_omp_taskreg): Update count_max.
9298         * omp-simd-clone.c (simd_clone_adjust): Update profile.
9299         * predict.c (maybe_hot_frequency_p): Use to_frequency.
9300         (maybe_hot_count_p): Use ipa counts only.
9301         (maybe_hot_bb_p): Simplify.
9302         (maybe_hot_edge_p): Simplify.
9303         (probably_never_executed): Do not take frequency argument.
9304         (probably_never_executed_bb_p): Do not pass frequency.
9305         (probably_never_executed_edge_p): Likewise.
9306         (combine_predictions_for_bb): Check that profile is nonzero.
9307         (propagate_freq): Do not set frequency.
9308         (drop_profile): Simplify.
9309         (counts_to_freqs): Simplify.
9310         (expensive_function_p): Use to_frequency.
9311         (propagate_unlikely_bbs_forward): Simplify.
9312         (determine_unlikely_bbs): Simplify.
9313         (estimate_bb_frequencies): Add hack to silence graphite issues.
9314         (compute_function_frequency): Use ipa counts.
9315         (pass_profile::execute): Update.
9316         (rebuild_frequencies): Use counts only.
9317         (force_edge_cold): Use counts only.
9318         * profile-count.c (profile_count::dump): Dump new count types.
9319         (profile_count::differs_from_p): Check compatiblity.
9320         (profile_count::to_frequency): New function.
9321         (profile_count::to_cgraph_frequency): New function.
9322         * profile-count.h (struct function): Declare.
9323         (enum profile_quality): Add profile_guessed_local and
9324         profile_guessed_global0.
9325         (class profile_proability): Decrease number of bits to 29;
9326         update from_reg_br_prob_note and to_reg_br_prob_note.
9327         (class profile_count: Update comment; decrease number of bits
9328         to 61. Check compatibility.
9329         (profile_count::compatible_p): New private member function.
9330         (profile_count::ipa_p): New member function.
9331         (profile_count::operator<): Handle global zero correctly.
9332         (profile_count::operator>): Handle global zero correctly.
9333         (profile_count::operator<=): Handle global zero correctly.
9334         (profile_count::operator>=): Handle global zero correctly.
9335         (profile_count::nonzero_p): New member function.
9336         (profile_count::force_nonzero): New member function.
9337         (profile_count::max): New member function.
9338         (profile_count::apply_scale): Handle IPA scalling.
9339         (profile_count::guessed_local): New member function.
9340         (profile_count::global0): New member function.
9341         (profile_count::ipa): New member function.
9342         (profile_count::to_frequency): Declare.
9343         (profile_count::to_cgraph_frequency): Declare.
9344         * profile.c (OVERLAP_BASE): Delete.
9345         (compute_frequency_overlap): Delete.
9346         (compute_branch_probabilities): Do not use compute_frequency_overlap.
9347         * regs.h (REG_FREQ_FROM_BB): Use to_frequency.
9348         * sched-ebb.c (rank): Use counts only.
9349         * shrink-wrap.c (handle_simple_exit): Use counts only.
9350         (try_shrink_wrapping): Use counts only.
9351         (place_prologue_for_one_component): Use counts only.
9352         * tracer.c (find_best_predecessor): Use to_frequency.
9353         (find_trace): Use to_frequency.
9354         (tail_duplicate): Use to_frequency.
9355         * trans-mem.c (expand_transaction): Do not update frequency.
9356         * tree-call-cdce.c: Do not update frequency.
9357         * tree-cfg.c (gimple_find_sub_bbs): Likewise.
9358         (gimple_merge_blocks): Likewise.
9359         (gimple_split_edge): Likewise.
9360         (gimple_duplicate_sese_region): Likewise.
9361         (gimple_duplicate_sese_tail): Likewise.
9362         (move_sese_region_to_fn): Likewise.
9363         (gimple_account_profile_record): Likewise.
9364         (insert_cond_bb): Likewise.
9365         * tree-complex.c (expand_complex_div_wide): Likewise.
9366         * tree-eh.c (lower_resx): Update profile.
9367         * tree-inline.c (copy_bb): Simplify count scaling; do not scale
9368         frequencies.
9369         (initialize_cfun): Do not initialize frequencies
9370         (freqs_to_counts): Delete.
9371         (copy_cfg_body): Ignore count parameter.
9372         (copy_body): Update.
9373         (expand_call_inline): Update count_max.
9374         (optimize_inline_calls): Update count_max.
9375         (tree_function_versioning): Update count_max.
9376         * tree-ssa-coalesce.c (coalesce_cost_bb): Use to_frequency.
9377         * tree-ssa-ifcombine.c (update_profile_after_ifcombine): Do not update
9378         frequency.
9379         * tree-ssa-loop-im.c (execute_sm_if_changed): Use counts only.
9380         * tree-ssa-loop-ivcanon.c (unloop_loops): Do not update freuqency.
9381         (try_peel_loop): Likewise.
9382         * tree-ssa-loop-ivopts.c (get_scaled_computation_cost_at): Use
9383         to_frequency.
9384         * tree-ssa-loop-manip.c (niter_for_unrolled_loop): Pass -1.
9385         (tree_transform_and_unroll_loop): Do not use frequencies
9386         * tree-ssa-loop-niter.c (estimate_numbers_of_iterations):
9387         Use reliable prediction only.
9388         * tree-ssa-loop-unswitch.c (hoist_guard): Do not use frequencies.
9389         * tree-ssa-sink.c (select_best_block): Use to_frequency.
9390         * tree-ssa-tail-merge.c (replace_block_by): Temporarily disable
9391         probability scaling.
9392         * tree-ssa-threadupdate.c (create_block_for_threading): Do
9393         not update frequency
9394         (any_remaining_duplicated_blocks): Likewise.
9395         (update_profile): Likewise.
9396         (estimated_freqs_path): Delete.
9397         (freqs_to_counts_path): Delete.
9398         (clear_counts_path): Delete.
9399         (ssa_fix_duplicate_block_edges): Likewise.
9400         (duplicate_thread_path): Likewise.
9401         * tree-switch-conversion.c (gen_inbound_check): Use counts.
9402         * tree-tailcall.c (decrease_profile): Do not update frequency.
9403         (eliminate_tail_call): Likewise.
9404         * tree-vect-loop-manip.c (vect_do_peeling): Likewise.
9405         * tree-vect-loop.c (scale_profile_for_vect_loop): Likewise.
9406         (optimize_mask_stores): Likewise.
9407         * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
9408         * ubsan.c (ubsan_expand_null_ifn): Update profile.
9409         (ubsan_expand_ptr_ifn): Update profile.
9410         * value-prof.c (gimple_ic): Simplify.
9411         * value-prof.h (gimple_ic): Update prototype.
9412         * ipa-inline-transform.c (inline_transform): Fix scaling conditoins.
9413         * ipa-inline.c (compute_uninlined_call_time): Be sure that
9414         counts are nonzero.
9415         (want_inline_self_recursive_call_p): Likewise.
9416         (resolve_noninline_speculation): Only cummulate defined counts.
9417         (inline_small_functions): Use nonzero_p.
9418         (ipa_inline): Do not access freed node.
9420 2017-11-03  Wilco Dijkstra  <wdijkstr@arm.com>
9422         * config/aarch64/aarch64.c (aarch64_override_options_internal):
9423         Set PARAM_SCHED_PRESSURE_ALGORITHM to SCHED_PRESSURE_MODEL.
9425 2017-11-03  Kito Cheng  <kito.cheng@gmail.com>
9427         * config/riscv/riscv.c (riscv_legitimize_move): Handle
9428         non-legitimate address.
9430 2017-11-03  Segher Boessenkool  <segher@kernel.crashing.org>
9432         * config/rs6000/rs6000.md (*lt0_disi): Delete.
9433         (*lt0_<mode>di, *lt0_<mode>si): New.
9435 2017-11-03  Segher Boessenkool  <segher@kernel.crashing.org>
9437         * config/rs6000/rs6000.md (move_from_CR_ov_bit): Change condition to
9438         TARGET_PAIRED_FLOAT.
9440 2017-11-03  Siddhesh Poyarekar  <siddhesh.poyarekar@linaro.org>
9441             Jim Wilson  <jim.wilson@linaro.org>
9443         * config/aarch64/aarch64-cores.def (saphira): New CPU.
9444         * config/aarch64/aarch64-tune.md: Regenerated.
9445         * doc/invoke.texi (AArch64 Options/-mtune): Add "saphira".
9446         * gcc/config/aarch64/aarch64.c (saphira_tunings): New tuning table.
9448 2017-11-03  Cupertino Miranda  <cmiranda@synopsys.com>
9450         * config/arc/arc.c (arc_save_restore): Corrected CFA note.
9451         (arc_expand_prologue): Restore blink for millicode.
9452         * config/arc/linux.h (LINK_EH_SPEC): Defined.
9454 2017-11-03  Richard Sandiford  <richard.sandiford@linaro.org>
9456         PR target/82809
9457         * config/i386/i386.c (ix86_vector_duplicate_value): Use
9458         gen_vec_duplicate after forcing the scalar into a register.
9460 2017-11-02  Segher Boessenkool  <segher@kernel.crashing.org>
9462         * combine (try_combine): Print the insns input to try_combine to the
9463         dump file.
9465 2017-11-02  Steve Ellcey  <sellcey@cavium.com>
9467         PR target/79868
9468         * config/aarch64/aarch64-c.c (aarch64_pragma_target_parse):
9469         Remove second argument from aarch64_process_target_attr call.
9470         * config/aarch64/aarch64-protos.h (aarch64_process_target_attr):
9471         Ditto.
9472         * config/aarch64/aarch64.c (aarch64_attribute_info): Change
9473         field type.
9474         (aarch64_handle_attr_arch): Remove second argument.
9475         (aarch64_handle_attr_cpu): Ditto.
9476         (aarch64_handle_attr_tune): Ditto.
9477         (aarch64_handle_attr_isa_flags): Ditto.
9478         (aarch64_process_one_target_attr): Ditto.
9479         (aarch64_process_target_attr): Ditto.
9480         (aarch64_option_valid_attribute_p): Remove second argument.
9481         on aarch64_process_target_attr call.
9483 2017-11-02  David Malcolm  <dmalcolm@redhat.com>
9485         * diagnostic.c: Include "selftest-diagnostic.h".
9486         (selftest::assert_location_text): New function.
9487         (selftest::test_diagnostic_get_location_text): New function.
9488         (selftest::diagnostic_c_tests): Call it.
9490 2017-11-02  David Malcolm  <dmalcolm@redhat.com>
9492         * Makefile.in (OBJS-libcommon): Add selftest-diagnostic.o.
9493         * diagnostic-show-locus.c: Include "selftest-diagnostic.h".
9494         (class selftest::test_diagnostic_context): Move to...
9495         * selftest-diagnostic.c: New file.
9496         * selftest-diagnostic.h: New file.
9498 2017-11-02  James Bowman  <james.bowman@ftdichip.com>
9500         * config/ft32/ft32.c (ft32_addr_space_legitimate_address_p): increase
9501         offset range for FT32B.
9502         * config/ft32/ft32.h: option "mcompress" enables relaxation.
9503         * config/ft32/ft32.md: Add TARGET_NOPM.
9504         * config/ft32/ft32.opt: Add mft32b, mcompress, mnopm.
9505         * gcc/doc/invoke.texi: Add mft32b, mcompress, mnopm.
9507 2017-11-02  Wilco Dijkstra  <wdijkstr@arm.com>
9509         * config/aarch64/aarch64.h (MALLOC_ABI_ALIGNMENT): New define.
9511 2017-11-02  Jeff Law  <law@redhat.com>
9513         * gimple-ssa-sprintf.c (sprintf_dom_walker): Remove
9514         virtual keyword on FINAL OVERRIDE members.
9516         * tree-ssa-propagate.h (ssa_propagation_engine): Group
9517         virtuals together.  Add virtual destructor.
9518         (substitute_and_fold_engine): Similarly.
9520 2017-11-02  Jan Hubicka  <hubicka@ucw.cz>
9522         * x86-tune.def (X86_TUNE_USE_INCDEC): Enable for Haswell+.
9524 2017-11-02  Richard Biener  <rguenther@suse.de>
9526         PR tree-optimization/82795
9527         * tree-if-conv.c (predicate_mem_writes): Remove bogus assert.
9529 2017-11-02  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
9531         * acinclude.m4 (gcc_AC_INITFINI_ARRAY): Don't require
9532         gcc_SUN_LD_VERSION.
9533         (gcc_GAS_CHECK_FEATURE): Remove.
9534         * configure.ac (ld_vers) <*-*-solaris2*>: Move comments from
9535         gcc_AC_INITFINI_ARRAY here.  Update for Solaris 11.4 changes.
9536         * configure: Regenerate.
9538 2017-11-02  Claudiu Zissulescu <claziss@synopsys.com>
9540         * config/arc/arc.c (hwloop_optimize): Account for empty
9541         body loops.
9543 2017-11-02  Richard Biener  <rguenther@suse.de>
9545         PR middle-end/82765
9546         * varasm.c (decode_addr_const): Make offset HOST_WIDE_INT.
9547         Truncate ARRAY_REF index and element size.
9549 2017-11-01  Palmer Dabbelt  <palmer@dabbelt.com>
9551         * doc/invoke.texi (RISC-V Options): Use "@minus{}2 GB", not "-2 GB".
9553 2017-11-01  Jeff Law  <law@redhat.com>
9555         * tree-ssa-ccp.c (ccp_folder): New class derived from
9556         substitute_and_fold_engine.
9557         (ccp_folder::get_value): New member function.
9558         (ccp_folder::fold_stmt): Renamed from ccp_fold_stmt.
9559         (ccp_fold_stmt): Remove prototype.
9560         (ccp_finalize): Call substitute_and_fold from the ccp_class.
9561         * tree-ssa-copy.c (copy_folder): New class derived from
9562         substitute_and_fold_engine.
9563         (copy_folder::get_value): Renamed from get_value.
9564         (fini_copy_prop): Call substitute_and_fold from copy_folder class.
9565         * tree-vrp.c (vrp_folder): New class derived from
9566         substitute_and_fold_engine.
9567         (vrp_folder::fold_stmt): Renamed from vrp_fold_stmt.
9568         (vrp_folder::get_value): New member function.
9569         (vrp_finalize): Call substitute_and_fold from vrp_folder class.
9570         (evrp_dom_walker::before_dom_children): Similarly for replace_uses_in.
9571         * tree-ssa-propagate.h (substitute_and_fold_engine): New class to
9572         provide a class interface to folder/substitute routines.
9573         (ssa_prop_fold_stmt_fn): Remove typedef.
9574         (ssa_prop_get_value_fn): Likewise.
9575         (subsitute_and_fold): Remove prototype.
9576         (replace_uses_in): Likewise.
9577         * tree-ssa-propagate.c (substitute_and_fold_engine::replace_uses_in):
9578         Renamed from replace_uses_in.  Call the virtual member function
9579         (substitute_and_fold_engine::replace_phi_args_in): Similarly.
9580         (substitute_and_fold_dom_walker): Remove initialization of
9581         data member entries for calbacks.  Add substitute_and_fold_engine
9582         member and initialize it.
9583         (substitute_and_fold_dom_walker::before_dom_children0: Use the
9584         member functions for get_value, replace_phi_args_in c
9585         replace_uses_in, and fold_stmt calls.
9586         (substitute_and_fold_engine::substitute_and_fold): Renamed from
9587         substitute_and_fold.  Remove assert.   Update ctor call.
9589         * tree-ssa-propagate.h (ssa_prop_visit_stmt_fn): Remove typedef.
9590         (ssa_prop_visit_phi_fn): Likewise.
9591         (class ssa_propagation_engine): New class to provide an interface
9592         into ssa_propagate.
9593         * tree-ssa-propagate.c (ssa_prop_visit_stmt): Remove file scoped
9594         variable.
9595         (ssa_prop_visit_phi): Likewise.
9596         (ssa_propagation_engine::simulate_stmt): Moved into class.
9597         Call visit_phi/visit_stmt from the class rather than via
9598         file scoped static variables.
9599         (ssa_propagation_engine::simulate_block): Moved into class.
9600         (ssa_propagation_engine::process_ssa_edge_worklist): Similarly.
9601         (ssa_propagation_engine::ssa_propagate): Similarly.  No longer
9602         set file scoped statics for the visit_stmt/visit_phi callbacks.
9603         * tree-complex.c (complex_propagate): New class derived from
9604         ssa_propagation_engine.
9605         (complex_propagate::visit_stmt): Renamed from complex_visit_stmt.
9606         (complex_propagate::visit_phi): Renamed from complex_visit_phi.
9607         (tree_lower_complex): Call ssa_propagate via the complex_propagate
9608         class.
9609         * tree-ssa-ccp.c: (ccp_propagate): New class derived from
9610         ssa_propagation_engine.
9611         (ccp_propagate::visit_phi): Renamed from ccp_visit_phi_node.
9612         (ccp_propagate::visit_stmt): Renamed from ccp_visit_stmt.
9613         (do_ssa_ccp): Call ssa_propagate from the ccp_propagate class.
9614         * tree-ssa-copy.c (copy_prop): New class derived from
9615         ssa_propagation_engine.
9616         (copy_prop::visit_stmt): Renamed from copy_prop_visit_stmt.
9617         (copy_prop::visit_phi): Renamed from copy_prop_visit_phi_node.
9618         (execute_copy_prop): Call ssa_propagate from the copy_prop class.
9619         * tree-vrp.c (vrp_prop): New class derived from ssa_propagation_engine.
9620         (vrp_prop::visit_stmt): Renamed from vrp_visit_stmt.
9621         (vrp_prop::visit_phi): Renamed from vrp_visit_phi_node.
9622         (execute_vrp): Call ssa_propagate from the vrp_prop class.
9624 2017-11-01  Jakub Jelinek  <jakub@redhat.com>
9626         PR rtl-optimization/82778
9627         PR rtl-optimization/82597
9628         * compare-elim.c (struct comparison): Add in_a_setter field.
9629         (find_comparison_dom_walker::before_dom_children): Remove killed
9630         bitmap and df_simulate_find_defs call, instead walk the defs.
9631         Compute last_setter and initialize in_a_setter.  Merge definitions
9632         with first initialization for a few variables.
9633         (try_validate_parallel): Use insn_invalid_p instead of
9634         recog_memoized.  Return insn rather than just the pattern.
9635         (try_merge_compare): Fix up comment.  Don't uselessly test if
9636         in_a is a REG_P.  Use cmp->in_a_setter instead of walking UD
9637         chains.
9638         (execute_compare_elim_after_reload): Remove df_chain_add_problem
9639         call.
9641 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9642             Alan Hayward  <alan.hayward@arm.com>
9643             David Sherwood  <david.sherwood@arm.com>
9645         * config/aarch64/aarch64.c (aarch64_rtx_costs): Use
9646         aarch64_hard_regno_nregs to get the number of registers
9647         in a mode.
9649 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9650             Alan Hayward  <alan.hayward@arm.com>
9651             David Sherwood  <david.sherwood@arm.com>
9653         * config/aarch64/constraints.md (Upl): Rename to...
9654         (Uaa): ...this.
9655         * config/aarch64/aarch64.md
9656         (*zero_extend<SHORT:mode><GPI:mode>2_aarch64, *addsi3_aarch64_uxtw):
9657         Update accordingly.
9659 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9660             Alan Hayward  <alan.hayward@arm.com>
9661             David Sherwood  <david.sherwood@arm.com>
9663         * config/aarch64/aarch64.c (aarch64_add_constant_internal)
9664         (aarch64_add_constant, aarch64_add_sp, aarch64_sub_sp): Move
9665         earlier in file.
9667 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9668             Alan Hayward  <alan.hayward@arm.com>
9669             David Sherwood  <david.sherwood@arm.com>
9671         * config/aarch64/aarch64.c (aarch64_evpc_trn, aarch64_evpc_uzp)
9672         (aarch64_evpc_zip, aarch64_evpc_ext, aarch64_evpc_rev)
9673         (aarch64_evpc_dup): Generate rtl direcly, rather than using
9674         named expanders.
9675         (aarch64_expand_vec_perm_const_1): Explicitly check for permutes
9676         of a single element.
9677         * config/aarch64/iterators.md: Add a comment above the permute
9678         unspecs to say that they are generated directly by
9679         aarch64_expand_vec_perm_const.
9680         * config/aarch64/aarch64-simd.md: Likewise the permute instructions.
9682 2017-11-01  Nathan Sidwell  <nathan@acm.org>
9684         * tree-dump.c (dequeue_and_dump): Use HAS_DECL_ASSEMBLER_NAME_P.
9686 2017-11-01  Palmer Dabbelt  <palmer@dabbelt.com>
9688         * doc/invoke.texi (RISC-V Options): Explicitly name the medlow
9689         and medany code models, and describe what they do.
9691 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9693         Revert accidental duplicate:
9695         * combine.c (can_change_dest_mode): Reject changes in
9696         REGMODE_NATURAL_SIZE.
9698 2017-11-01  Segher Boessenkool  <segher@kernel.crashing.org>
9700         PR rtl-optimization/64682
9701         PR rtl-optimization/69567
9702         PR rtl-optimization/69737
9703         PR rtl-optimization/82683
9704         * combine.c (distribute_notes) <REG_DEAD>: If the new I2 sets the same
9705         register mentioned in the note, drop the note, unless it came from I3,
9706         in which case it should go to I3 again.
9708 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9710         * tree-ssa-dse.c (normalize_ref): Check whether the ranges overlap
9711         and return false if not.
9712         (clear_bytes_written_by, live_bytes_read): Update accordingly.
9714 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9716         * tree-ssa-alias.h (ranges_overlap_p): Return false if either
9717         range is known to be empty.
9719 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9720             Alan Hayward  <alan.hayward@arm.com>
9721             David Sherwood  <david.sherwood@arm.com>
9723         * simplify-rtx.c (simplify_const_unary_operation): Use GET_MODE_NUNITS
9724         and CONST_VECTOR_NUNITS instead of computing the number of units from
9725         the byte sizes of the vector and element.
9726         (simplify_binary_operation_1): Likewise.
9727         (simplify_const_binary_operation): Likewise.
9728         (simplify_ternary_operation): Likewise.
9730 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9731             Alan Hayward  <alan.hayward@arm.com>
9732             David Sherwood  <david.sherwood@arm.com>
9734         * var-tracking.c (INT_MEM_OFFSET): Replace with...
9735         (int_mem_offset): ...this new function.
9736         (var_mem_set, var_mem_delete_and_set, var_mem_delete)
9737         (find_mem_expr_in_1pdv, dataflow_set_preserve_mem_locs)
9738         (same_variable_part_p, use_type, add_stores, vt_get_decl_and_offset):
9739         Update accordingly.
9741 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9742             Alan Hayward  <alan.hayward@arm.com>
9743             David Sherwood  <david.sherwood@arm.com>
9745         * lower-subreg.c (interesting_mode_p): New function.
9746         (compute_costs, find_decomposable_subregs, decompose_register)
9747         (simplify_subreg_concatn, can_decompose_p, resolve_simple_move)
9748         (resolve_clobber, dump_choices): Use it.
9750 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9751             Alan Hayward  <alan.hayward@arm.com>
9752             David Sherwood  <david.sherwood@arm.com>
9754         * rtlhash.c (add_rtx): Use add_hwi for 'w' and add_int for 'i'.
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         * alias.c (find_base_value, find_base_term): Only process integer
9761         truncations.  Check the precision rather than the size.
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         * machmode.h (is_narrower_int_mode): New function
9768         * optabs.c (expand_float, expand_fix): Use it.
9769         * dwarf2out.c (rotate_loc_descriptor): Likewise.
9771 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9772             Alan Hayward  <alan.hayward@arm.com>
9773             David Sherwood  <david.sherwood@arm.com>
9775         * rtl.h (narrower_subreg_mode): New function.
9776         * ira-color.c (update_costs_from_allocno): Use it.
9778 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9779             Alan Hayward  <alan.hayward@arm.com>
9780             David Sherwood  <david.sherwood@arm.com>
9782         * optabs-query.h (convert_optab_p): New function, split out from...
9783         (convert_optab_handler): ...here.
9784         (widening_optab_handler): Delete.
9785         (find_widening_optab_handler): Remove permit_non_widening parameter.
9786         (find_widening_optab_handler_and_mode): Likewise.  Provide an
9787         override that operates on mode class wrappers.
9788         * optabs-query.c (widening_optab_handler): Delete.
9789         (find_widening_optab_handler_and_mode): Remove permit_non_widening
9790         parameter.  Assert that the two modes are the same class and that
9791         the "from" mode is narrower than the "to" mode.  Use
9792         convert_optab_handler instead of widening_optab_handler.
9793         * expmed.c (expmed_mult_highpart_optab): Use convert_optab_handler
9794         instead of widening_optab_handler.
9795         * expr.c (expand_expr_real_2): Update calls to
9796         find_widening_optab_handler.
9797         * optabs.c (expand_widen_pattern_expr): Likewise.
9798         (expand_binop_directly): Take the insn_code as a parameter.
9799         (expand_binop): Only call find_widening_optab_handler for
9800         conversion optabs; use optab_handler otherwise.  Update calls
9801         to find_widening_optab_handler and expand_binop_directly.
9802         Use convert_optab_handler instead of widening_optab_handler.
9803         * tree-ssa-math-opts.c (convert_mult_to_widen): Update calls to
9804         find_widening_optab_handler and use scalar_mode rather than
9805         machine_mode.
9806         (convert_plusminus_to_widen): Likewise.
9808 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9809             Alan Hayward  <alan.hayward@arm.com>
9810             David Sherwood  <david.sherwood@arm.com>
9812         * machmode.h (fixed_size_mode): New class.
9813         * rtl.h (get_pool_mode): Return fixed_size_mode.
9814         * gengtype.c (main): Add fixed_size_mode.
9815         * target.def (get_raw_result_mode): Return a fixed_size_mode.
9816         (get_raw_arg_mode): Likewise.
9817         * doc/tm.texi: Regenerate.
9818         * targhooks.h (default_get_reg_raw_mode): Return a fixed_size_mode.
9819         * targhooks.c (default_get_reg_raw_mode): Likewise.
9820         * config/ia64/ia64.c (ia64_get_reg_raw_mode): Likewise.
9821         * config/mips/mips.c (mips_get_reg_raw_mode): Likewise.
9822         * config/msp430/msp430.c (msp430_get_raw_arg_mode): Likewise.
9823         (msp430_get_raw_result_mode): Likewise.
9824         * config/avr/avr-protos.h (regmask): Use as_a <fixed_side_mode>
9825         * dbxout.c (dbxout_parms): Require fixed-size modes.
9826         * expr.c (copy_blkmode_from_reg, copy_blkmode_to_reg): Likewise.
9827         * gimple-ssa-store-merging.c (encode_tree_to_bitpos): Likewise.
9828         * omp-low.c (lower_oacc_reductions): Likewise.
9829         * simplify-rtx.c (simplify_immed_subreg): Take fixed_size_modes.
9830         (simplify_subreg): Update accordingly.
9831         * varasm.c (constant_descriptor_rtx::mode): Change to fixed_size_mode.
9832         (force_const_mem): Update accordingly.  Return NULL_RTX for modes
9833         that aren't fixed-size.
9834         (get_pool_mode): Return a fixed_size_mode.
9835         (output_constant_pool_2): Take a fixed_size_mode.
9837 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9838             Alan Hayward  <alan.hayward@arm.com>
9839             David Sherwood  <david.sherwood@arm.com>
9841         * doc/rtl.texi (vec_series): Document.
9842         (const): Say that the operand can be a vec_series.
9843         * rtl.def (VEC_SERIES): New rtx code.
9844         * rtl.h (const_vec_series_p_1): Declare.
9845         (const_vec_series_p): New function.
9846         * emit-rtl.h (gen_const_vec_series): Declare.
9847         (gen_vec_series): Likewise.
9848         * emit-rtl.c (const_vec_series_p_1, gen_const_vec_series)
9849         (gen_vec_series): Likewise.
9850         * optabs.c (expand_mult_highpart): Use gen_const_vec_series.
9851         * simplify-rtx.c (simplify_unary_operation): Handle negations
9852         of vector series.
9853         (simplify_binary_operation_series): New function.
9854         (simplify_binary_operation_1): Use it.  Handle VEC_SERIES.
9855         (test_vector_ops_series): New function.
9856         (test_vector_ops): Call it.
9857         * config/powerpcspe/altivec.md (altivec_lvsl): Use
9858         gen_const_vec_series.
9859         (altivec_lvsr): Likewise.
9860         * config/rs6000/altivec.md (altivec_lvsl, altivec_lvsr): Likewise.
9862 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9863             Alan Hayward  <alan.hayward@arm.com>
9864             David Sherwood  <david.sherwood@arm.com>
9866         * doc/rtl.texi (const): Update description of address constants.
9867         Say that vector constants are allowed too.
9868         * common.md (E, F): Use CONSTANT_P instead of checking for
9869         CONST_VECTOR.
9870         * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
9871         checking for CONST_VECTOR.
9872         * expmed.c (make_tree): Use build_vector_from_val for a CONST
9873         VEC_DUPLICATE.
9874         * expr.c (expand_expr_real_2): Check for vector modes instead
9875         of checking for CONST_VECTOR.
9876         * rtl.h (const_vec_p): New function.
9877         (const_vec_duplicate_p): Check for a CONST VEC_DUPLICATE.
9878         (unwrap_const_vec_duplicate): Handle them here too.
9880 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9881             David Malcolm  <dmalcolm@redhat.com>
9882             Alan Hayward  <alan.hayward@arm.com>
9883             David Sherwood  <david.sherwood@arm.com>
9885         * rtl.h (vec_duplicate_p): New function.
9886         * selftest-rtl.c (assert_rtx_eq_at): New function.
9887         * selftest-rtl.h (ASSERT_RTX_EQ): New macro.
9888         (assert_rtx_eq_at): Declare.
9889         * selftest.h (selftest::simplify_rtx_c_tests): Declare.
9890         * selftest-run-tests.c (selftest::run_tests): Call it.
9891         * simplify-rtx.c: Include selftest.h and selftest-rtl.h.
9892         (simplify_unary_operation_1): Recursively handle vector duplicates.
9893         (simplify_binary_operation_1): Likewise.  Handle VEC_SELECTs of
9894         vector duplicates.
9895         (simplify_subreg): Handle subregs of vector duplicates.
9896         (make_test_reg, test_vector_ops_duplicate, test_vector_ops)
9897         (selftest::simplify_rtx_c_tests): New functions.
9899 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9900             Alan Hayward  <alan.hayward@arm.com>
9901             David Sherwood  <david.sherwood@arm.com>
9903         * emit-rtl.h (gen_const_vec_duplicate): Declare.
9904         (gen_vec_duplicate): Likewise.
9905         * emit-rtl.c (gen_const_vec_duplicate_1): New function, split
9906         out from...
9907         (gen_const_vector): ...here.
9908         (gen_const_vec_duplicate, gen_vec_duplicate): New functions.
9909         (gen_rtx_CONST_VECTOR): Use gen_const_vec_duplicate for constants
9910         whose elements are all equal.
9911         * optabs.c (expand_vector_broadcast): Use gen_const_vec_duplicate.
9912         * simplify-rtx.c (simplify_const_unary_operation): Likewise.
9913         (simplify_relational_operation): Likewise.
9914         * config/aarch64/aarch64.c (aarch64_simd_gen_const_vector_dup):
9915         Likewise.
9916         (aarch64_simd_dup_constant): Use gen_vec_duplicate.
9917         (aarch64_expand_vector_init): Likewise.
9918         * config/arm/arm.c (neon_vdup_constant): Likewise.
9919         (neon_expand_vector_init): Likewise.
9920         (arm_expand_vec_perm): Use gen_const_vec_duplicate.
9921         (arm_block_set_unaligned_vect): Likewise.
9922         (arm_block_set_aligned_vect): Likewise.
9923         * config/arm/neon.md (neon_copysignf<mode>): Likewise.
9924         * config/i386/i386.c (ix86_expand_vec_perm): Likewise.
9925         (expand_vec_perm_even_odd_pack): Likewise.
9926         (ix86_vector_duplicate_value): Use gen_vec_duplicate.
9927         * config/i386/sse.md (one_cmpl<mode>2): Use CONSTM1_RTX.
9928         * config/ia64/ia64.c (ia64_expand_vecint_compare): Use
9929         gen_const_vec_duplicate.
9930         * config/ia64/vect.md (addv2sf3, subv2sf3): Use CONST1_RTX.
9931         * config/mips/mips.c (mips_gen_const_int_vector): Use
9932         gen_const_vec_duplicate.
9933         (mips_expand_vector_init): Use CONST0_RTX.
9934         * config/powerpcspe/altivec.md (abs<mode>2, nabs<mode>2): Likewise.
9935         (define_split): Use gen_const_vec_duplicate.
9936         * config/rs6000/altivec.md (abs<mode>2, nabs<mode>2): Use CONST0_RTX.
9937         (define_split): Use gen_const_vec_duplicate.
9938         * config/s390/vx-builtins.md (vec_genmask<mode>): Likewise.
9939         (vec_ctd_s64, vec_ctd_u64, vec_ctsl, vec_ctul): Likewise.
9940         * config/spu/spu.c (spu_const): Likewise.
9942 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9943             Alan Hayward  <alan.hayward@arm.com>
9944             David Sherwood  <david.sherwood@arm.com>
9946         * combine.c (can_change_dest_mode): Reject changes in
9947         REGMODE_NATURAL_SIZE.
9949 2017-10-31  Sandra Loosemore  <sandra@codesourcery.com>
9951         * configure.ac (--enable-libssp): New.
9952         (gcc_cv_libc_provides_ssp): Check for explicit setting before
9953         trying to determine target-specific default.  Adjust indentation.
9954         * configure: Regenerated.
9955         * doc/install.texi (Configuration): Expand --disable-libssp
9956         documentation.
9958 2017-10-31  Daniel Santos  <daniel.santos@pobox.com>
9960         config/i386/i386.c (ix86_expand_epilogue): Correct stack
9961         calculation.
9963 2017-10-31  Martin Jambor  <mjambor@suse.cz>
9965         PR c++/81702
9966         * gimple-fold.c (gimple_get_virt_method_for_vtable): Remove assert.
9968 2017-10-31  David Malcolm  <dmalcolm@redhat.com>
9970         * auto-profile.c (autofdo_source_profile::read): Use
9971         UNKNOWN_LOCATION rather than 0.
9972         * diagnostic-core.h (warning_at_rich_loc): Rename to...
9973         (warning_at): ...this overload.
9974         (warning_at_rich_loc_n): Rename to...
9975         (warning_n): ...this overload.
9976         (error_at_rich_loc): Rename to...
9977         (error_at): ...this overload.
9978         (pedwarn_at_rich_loc): Rename to...
9979         (pedwarn): ...this overload.
9980         (permerror_at_rich_loc): Rename to...
9981         (permerror): ...this overload.
9982         (inform_at_rich_loc): Rename to...
9983         (inform): ...this overload.
9984         * diagnostic.c: (diagnostic_n_impl): Delete location_t-based decl.
9985         (diagnostic_n_impl_richloc): Rename to...
9986         (diagnostic_n_impl): ...this rich_location *-based decl.
9987         (inform_at_rich_loc): Rename to...
9988         (inform): ...this, and add an assertion.
9989         (inform_n): Update for removal of location_t-based diagnostic_n_impl.
9990         (warning_at_rich_loc): Rename to...
9991         (warning_at): ...this, and add an assertion.
9992         (warning_at_rich_loc_n): Rename to...
9993         (warning_n): ...this, and add an assertion.
9994         (warning_n): Update location_t-based implementation for removal of
9995         location_t-based diagnostic_n_impl.
9996         (pedwarn_at_rich_loc): Rename to...
9997         (pedwarn): ...this, and add an assertion.
9998         (permerror_at_rich_loc): Rename to...
9999         (permerror): ...this, and add an assertion.
10000         (error_n): Update for removal of location_t-based diagnostic_n_impl.
10001         (error_at_rich_loc): Rename to...
10002         (error_at): ...this, and add an assertion.
10003         * gcc.c (do_spec_1): Use UNKNOWN_LOCATION rather than 0.
10004         (driver::do_spec_on_infiles): Likewise.
10005         * substring-locations.c (format_warning_va): Update for renaming
10006         of inform_at_rich_loc.
10008 2017-10-31  Michael Meissner  <meissner@linux.vnet.ibm.com>
10010         * builtins.def (DEF_FLOATN_BUILTIN): Change most _Float<N> and
10011         _Float<N>X built-in functions so that the variant without the
10012         "__builtin_" prefix is only enabled for the GNU C and Objective C
10013         languages when they are in non-strict ANSI/ISO mode.
10014         (DEF_EXT_LIB_FLOATN_NX_BUILTINS): Likewise.
10015         * target.def (floatn_builtin_p): Add a target hook to control
10016         whether _Float<N> and _Float<N>X built-in functions without the
10017         "__builtin_" prefix are enabled, and return true for C and
10018         Objective C in the default hook.  Include langhooks.h in
10019         targhooks.c.
10020         * targhooks.h (default_floatn_builtin_p): Likewise.
10021         * targhooks.c (default_floatn_builtin_p): Likewise.
10022         * doc/tm.texi.in (TARGET_FLOATN_BUILTIN_P): Document the
10023         floatn_builtin_p target hook.
10024         * doc/tm.texi (TARGET_FLOATN_BUILTIN_P): Likewise.
10026 2017-10-31  Matthew Fortune  <matthew.fortune@imgtec.com>
10027             Eric Botcazou  <ebotcazou@adacore.com>
10029         PR rtl-optimization/81803
10030         * lra-constraints.c (curr_insn_transform): Also reload the whole
10031         register for a strict subreg no wider than a word if this is for
10032         a WORD_REGISTER_OPERATIONS target.
10034 2017-10-31  Jason Merrill  <jason@redhat.com>
10036         * gdbinit.in: Skip over inlines from timevar.h.
10038 2017-10-31  Martin Liska  <mliska@suse.cz>
10040         * doc/gcov.texi: Document new option.
10041         * gcov.c (print_usage): Likewise print it.
10042         (process_args): Support the argument.
10043         (format_count): New function.
10044         (format_gcov): Use the function.
10046 2017-10-31  Martin Liska  <mliska@suse.cz>
10048         * gcov.c (struct name_map): do not use typedef.
10049         Define operator== and operator<.
10050         (name_search): Remove.
10051         (name_sort): Remove.
10052         (main): Do not allocate names.
10053         (process_file): Add vertical space.
10054         (generate_results): Use std::find.
10055         (release_structures): Do not release memory.
10056         (find_source): Use std::find.
10058 2017-10-31  Martin Liska  <mliska@suse.cz>
10060         * gcov.c (struct line_info): Remove it's typedef.
10061         (line_info::line_info): Add proper ctor.
10062         (line_info::has_block): Do not use a typedef.
10063         (struct source_info): Do not use typedef.
10064         (circuit): Likewise.
10065         (get_cycles_count): Likewise.
10066         (output_intermediate_file): Iterate via vector iterator.
10067         (add_line_counts): Use std::vector methods.
10068         (accumulate_line_counts): Likewise.
10069         (output_lines): Likewise.
10071 2017-10-31  Martin Liska  <mliska@suse.cz>
10073         * gcov.c (struct source_info): Remove typedef.
10074         (source_info::source_info): Add proper ctor.
10075         (accumulate_line_counts): Use struct, not it's typedef.
10076         (output_gcov_file): Likewise.
10077         (output_lines): Likewise.
10078         (main): Do not allocate an array.
10079         (output_intermediate_file): Use size of vector container.
10080         (process_file): Resize the vector.
10081         (generate_results): Do not preallocate, use newly added vector
10082         lines.
10083         (release_structures): Do not release sources.
10084         (find_source): Use vector methods.
10085         (add_line_counts): Do not use typedef.
10087 2017-10-31  Martin Liska  <mliska@suse.cz>
10089         * doc/gcov.texi: Document that.
10090         * gcov.c (add_line_counts): Mark lines with a non-executed
10091         statement.
10092         (output_line_beginning): Handle such lines.
10093         (output_lines): Pass new argument.
10094         (output_intermediate_file): Print it in intermediate format.
10096 2017-10-31  Martin Liska  <mliska@suse.cz>
10098         * color-macros.h: New file.
10099         * diagnostic-color.c: Factor out color related to macros to
10100         color-macros.h.
10101         * doc/gcov.texi: Document -k option.
10102         * gcov.c (INCLUDE_STRING): Include string.h.
10103         (print_usage): Add -k option.
10104         (process_args): Parse it.
10105         (pad_count_string): New function.
10106         (output_line_beginning): Likewise.
10107         (DEFAULT_LINE_START): New macro.
10108         (output_lines): Support color output.
10110 2017-10-31  Martin Liska  <mliska@suse.cz>
10112         PR gcov-profile/82633
10113         * doc/gcov.texi: Document -fkeep-{static,inline}-functions and
10114         their interaction with GCOV infrastructure.
10115         * configure.ac: Add -fkeep-{inline,static}-functions to
10116         coverage_flags.
10117         * configure: Regenerate.
10119 2017-10-31  Uros Bizjak  <ubizjak@gmail.com>
10121         PR target/82772
10122         * config/alpha/sync.md (fetchop_constr) <and>: Change to "rINM".
10124 2017-10-31  Segher Boessenkool  <segher@kernel.crashing.org>
10126         PR target/82674
10127         * config/rs6000/rs6000.md (allocate_stack): Force update interval
10128         into a register if it does not fit into an immediate offset field.
10130 2017-10-31  Olivier Hainque  <hainque@adacore.com>
10132         * gcc/Makefile.in (FLAGS_TO_PASS): Pass libsubdir as well.
10134 2017-10-31  Julia Koval  <julia.koval@intel.com>
10136         * config.gcc: Add gfniintrin.h.
10137         * config/i386/gfniintrin.h: New.
10138         * config/i386/i386-builtin-types.def
10139         (__builtin_ia32_vgf2p8affineinvqb_v64qi,
10140         __builtin_ia32_vgf2p8affineinvqb_v64qi_mask,
10141         __builtin_ia32_vgf2p8affineinvqb_v32qi,
10142         __builtin_ia32_vgf2p8affineinvqb_v32qi_mask,
10143         __builtin_ia32_vgf2p8affineinvqb_v16qi,
10144         __builtin_ia32_vgf2p8affineinvqb_v16qi_mask): New builtins.
10145         * config/i386/i386-builtin.def (V64QI_FTYPE_V64QI_V64QI_INT_V64QI_UDI,
10146         V32QI_FTYPE_V32QI_V32QI_INT_V32QI_USI,
10147         V16QI_FTYPE_V16QI_V16QI_INT_V16QI_UHI,
10148         V64QI_FTYPE_V64QI_V64QI_INT): New types.
10149         * config/i386/i386.c (ix86_expand_args_builtin): Handle new types.
10150         * config/i386/immintrin.h: Include gfniintrin.h.
10151         * config/i386/sse.md (vgf2p8affineinvqb_*) New pattern.
10153 2017-10-30  Eric Botcazou  <ebotcazou@adacore.com>
10155         * gcc.c (HAVE_TARGET_EXECUTABLE_SUFFIX): Remove old kludge.
10157 2017-10-30  Wilco Dijkstra  <wdijkstr@arm.com>
10159         * config/arm/arm.md (ashldi3): Remove shift by 1 expansion.
10160         (arm_ashldi3_1bit): Remove pattern.
10161         (ashrdi3): Remove shift by 1 expansion.
10162         (arm_ashrdi3_1bit): Remove pattern.
10163         (lshrdi3): Remove shift by 1 expansion.
10164         (arm_lshrdi3_1bit): Remove pattern.
10165         * config/arm/arm.c (arm_rtx_costs_internal): Slightly increase
10166         cost of ashldi3 by 1.
10167         * config/arm/neon.md (ashldi3_neon): Remove shift by 1 expansion.
10168         (<shift>di3_neon): Likewise.
10170 2017-10-30  Dominik Infuehr  <dominik.infuehr@theobroma-systems.com>
10172         * config/aarch64/aarch64-simd.md (*aarch64_simd_mov): Rename
10173         both identically named patterns to (*aarch64_simd_mov<VD:mode>)
10174         and (*aarch64_simd_mov<VQ:mode>).
10175         (*aarch64_simd_mov<VD:mode>): Change type attribute to match
10176         pattern alternative.
10177         (*aarch64_simd_mov<VQ:mode>): Re-order and change type
10178         attributes to match pattern alternative.
10180 2017-10-30  Steven Munroe  <munroesj@gcc.gnu.org>
10182         * config.gcc (powerpc*-*-*): Add emmintrin.h.
10183         * config/rs6000/emmintrin.h: New file.
10184         * config/rs6000/x86intrin.h [__ALTIVEC__]: Include emmintrin.h.
10186 2017-10-30  Wilco Dijkstra  <wdijkstr@arm.com>
10188         * config/arm/vfp.md (movdi_vfp): Merge changes from movdi_vfp_cortexa8.
10189         * (movdi_vfp_cortexa8): Remove pattern.
10191 2017-10-30  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
10193         * doc/install.texi (Specific, alpha*-*-*): Remove DEC OSF/1
10194         etc. reference.
10195         (Specific, alpha*-dec-osf5.1): Remove.
10196         (Specific, mips-sgi-irix5): Remove.
10197         (Specific, mips-sgi-irix6): Remove.
10199 2017-10-30  Jakub Jelinek  <jakub@redhat.com>
10201         PR middle-end/22141
10202         * gimple-ssa-store-merging.c (merged_store_group::apply_stores): Fix
10203         arguments to clear_bit_region_be.
10205 2017-10-30  Jim Wilson  <wilson@tuliptree.org>
10207         * gimplify.c: Include memmodel.h.
10209 2017-10-30  Martin Jambor  <mjambor@suse.cz>
10211         * omp-grid.c (grid_attempt_target_gridification): Also insert a
10212         condition whether loop should be executed at all.
10214 2017-10-30  Will Schmidt  <will_schmidt@vnet.ibm.com>
10216         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add support for
10217         gimple folding of vec_madd() intrinsics.
10218         * config/rs6000/altivec.md (mulv8hi3): Rename altivec_vmladduhm to
10219         fmav8hi4.  (altivec_vmladduhm): Rename to fmav8hi4.
10220         * config/rs6000/rs6000-builtin.def: Rename vmladduhm to fmav8hi4
10222 2017-10-30  Richard Biener  <rguenther@suse.de>
10224         PR tree-optimization/82762
10225         Revert
10226         2017-10-23  Richard Biener  <rguenther@suse.de>
10228         PR tree-optimization/82129
10229         Revert
10230         2017-08-01  Richard Biener  <rguenther@suse.de>
10232         PR tree-optimization/81181
10233         * tree-ssa-pre.c (compute_antic_aux): Defer clean() to ...
10234         (compute_antic): ... end of iteration here.
10236 2017-10-30  Joseph Myers  <joseph@codesourcery.com>
10238         * doc/invoke.texi (C Dialect Options): Document -std=c17,
10239         -std=iso9899:2017 and -std=gnu17.
10240         * doc/standards.texi (C Language): Document C17 support.
10241         * doc/cpp.texi (Overview): Mention -std=c17.
10242         (Standard Predefined Macros): Document C11 and C17 values of
10243         __STDC_VERSION__.  Do not refer to C99 support as incomplete.
10244         * doc/extend.texi (Inline): Do not list individual options for
10245         standards newer than C99.
10246         * dwarf2out.c (highest_c_language, gen_compile_unit_die): Handle
10247         "GNU C17".
10248         * config/rl78/rl78.c (rl78_option_override): Handle "GNU C17"
10249         language name.
10251 2017-10-30  Maxim Ostapenko  <m.ostapenko@samsung.com>
10253         * asan.c (asan_finish_file): Align asan globals array by shadow
10254         granularity.
10256 2017-10-30  Jakub Jelinek  <jakub@redhat.com>
10258         PR middle-end/22141
10259         * gimple-ssa-store-merging.c: Include rtl.h and expr.h.
10260         (struct store_immediate_info): Add bitregion_start and bitregion_end
10261         fields.
10262         (store_immediate_info::store_immediate_info): Add brs and bre
10263         arguments and initialize bitregion_{start,end} from those.
10264         (struct merged_store_group): Add bitregion_start, bitregion_end,
10265         align_base and mask fields.  Drop unnecessary struct keyword from
10266         struct store_immediate_info.  Add do_merge method.
10267         (clear_bit_region_be): Use memset instead of loop storing zeros.
10268         (merged_store_group::do_merge): New method.
10269         (merged_store_group::merge_into): Use do_merge.  Allow gaps in between
10270         stores as long as the surrounding bitregions have no gaps.
10271         (merged_store_group::merge_overlapping): Use do_merge.
10272         (merged_store_group::apply_stores): Test that bitregion_{start,end}
10273         is byte aligned, rather than requiring that start and width are
10274         byte aligned.  Drop unnecessary struct keyword from
10275         struct store_immediate_info.  Allocate and populate also mask array.
10276         Make start of the arrays relative to bitregion_start rather than
10277         start and size them according to bitregion_{end,start} difference.
10278         (struct imm_store_chain_info): Drop unnecessary struct keyword from
10279         struct store_immediate_info.
10280         (pass_store_merging::gate): Punt if BITS_PER_UNIT or CHAR_BIT is not 8.
10281         (pass_store_merging::terminate_all_aliasing_chains): Drop unnecessary
10282         struct keyword from struct store_immediate_info.
10283         (imm_store_chain_info::coalesce_immediate_stores): Allow gaps in
10284         between stores as long as the surrounding bitregions have no gaps.
10285         Formatting fixes.
10286         (struct split_store): Add orig non-static data member.
10287         (split_store::split_store): Initialize orig to false.
10288         (find_constituent_stmts): Return store_immediate_info *, non-NULL
10289         if there is exactly a single original stmt.  Change stmts argument
10290         to pointer from reference, if NULL, don't push anything to it.  Add
10291         first argument, use it to optimize skipping over orig stmts that
10292         are known to be before bitpos already.  Simplify.
10293         (split_group): Return unsigned int count how many stores are or
10294         would be needed rather than a bool.  Add allow_unaligned argument.
10295         Change split_stores argument from reference to pointer, if NULL,
10296         only do a dry run computing how many stores would be produced.
10297         Rewritten algorithm to use both alignment and misalign if
10298         !allow_unaligned and handle bitfield stores with gaps.
10299         (imm_store_chain_info::output_merged_store): Set start_byte_pos
10300         from bitregion_start instead of start.  Compute allow_unaligned
10301         here, if true, do 2 split_group dry runs to compute which one
10302         produces fewer stores and prefer aligned if equal.  Punt if
10303         new count is bigger or equal than original before emitting any
10304         statements, rather than during that.  Remove no longer needed
10305         new_ssa_names tracking.  Replace num_stmts with
10306         split_stores.length ().  Use 32-bit stack allocated entries
10307         in split_stores auto_vec.  Try to reuse original store lhs/rhs1
10308         if possible.  Handle bitfields with gaps.
10309         (pass_store_merging::execute): Ignore bitsize == 0 stores.
10310         Compute bitregion_{start,end} for the stores and construct
10311         store_immediate_info with that.  Formatting fixes.
10313 2017-10-30  Uros Bizjak  <ubizjak@gmail.com>
10315         PR target/82725
10316         * config/i386/i386.c (legitimate_pic_address_disp_p): Allow
10317         UNSPEC_DTPOFF and UNSPEC_NTPOFF with SImode immediate offset.
10319 2017-10-29  Jim Wilson  <wilson@tuliptree.org>
10321         * gimplify.c: Include tm_p.h.
10323         * common.opt (gcoff): Re-add as ignored option.
10324         (gcoff1, gcoff2, gcoff3): Likewise.
10326         * Makefile.in (OBJS): Delete sdbout.o.
10327         (GTFILES): Delete $(srcdir)/sdbout.c.
10328         * debug.h: Delete sdb_debug_hooks.
10329         * final.c: Delete sdbout.h include.
10330         (final_scan_insn): Delete SDB_DEBUG check.
10331         (rest_of_clean_state): Likewise.
10332         * output.h: Delete sdb_begin_function_line.
10333         * sdbout.c: Delete.
10334         * sdbout.h: Delete.
10335         * toplev.c: Delete sdbout.h include.
10336         (process_options): Delete SDB_DEBUG check.
10337         * tree-core.h (tree_type_common): Delete pointer field of
10338         tree_type_symtab.
10339         * tree.c (copy_node): Clear TYPE_SYMTAB_DIE instead of
10340         TYPE_SYMTAB_POINTER.
10341         * tree.h (TYPE_SYMTAB_POINTER): Delete.
10342         (TYPE_SYMTAB_IS_POINTER): Delete.
10343         (TYPE_SYMTAB_IS_DIE): Renumber.
10344         * xcoffout.c: Refer to former sdbout.c file.
10345         (xcoffout_begin_prologue): Use past tense for sdbout.c reference.
10347         * doc/install.texi (--with-stabs): Delete COFF and ECOFF info.
10348         * doc/invoke.texi (SEEALSO): Delete adb and sdb references.
10349         (Debugging Options): Delete -gcoff.
10350         (-gstabs): Delete SDB reference.
10351         (-gcoff): Delete.
10352         (-gcoff@var{level}): Delete.
10353         * doc/passes.texi (Debugging information output): Delete SDB and
10354         sdbout.c references.
10355         * doc/tm.texi: Regenerate.
10356         * doc/tm.texi.in (DWARF_CIE_DATA_ALIGNMENT): Delete SDB from xref.
10357         (SDB and DWARF): Change node name to DWARF and delete SDB and COFF
10358         references.
10359         (DEBUGGER_AUTO_OFFSET): Delete COFF and SDB references.
10360         (PREFERRED_DEBUGGING_TYPE): Delete SDB_DEBUG and -gcoff references.
10361         (SDB_DEBUGGING_INFO): Delete.
10362         (PUT_SDB_@dots{}, SDB_DELIM, SDB_ALLOW_UNKNOWN_REFERENCES)
10363         SDB_ALLOW_FORWARD_REFERENCES, SDB_OUTPUT_SOURCE_LINE): Delete.
10364         * target.def (output_source_filename): Delete COFF reference.
10366         * common.opt (gcoff): Delete.
10367         (gxcoff+): Update Negative chain.
10368         * defaults.h: Delete all references to SDB_DEBUGGING_INFO and
10369         SDB_DEBUG.
10370         * dwarf2out.c (gen_array_type_die): Change SDB to debuggers.
10371         * flag-types.h (enum debug_info_type): Delete SDB_DEBUG.
10372         * function.c (number_blocks): Delete SDB_DEBUGGING_INFO, SDB_DEBUG,
10373         and SDB references.
10374         (expand_function_start): Change sdb reference to past tense.
10375         (expand_function_end): Change sdb reference to past tense.
10376         * gcc.c (cpp_unique_options): Delete gcoff3 reference.
10377         * opts.c (debug_type_names): Delete coff entry.
10378         (common_handle_option): Delete OPT_gcoff case.
10379         * system.h (SDB_DEBUG, SDB_DEBUGGING_INFO): Poison.
10381         * config/dbxcoff.h (PREFERRED_DEBUGGING_TYPE): Set to DBX_DEBUG.
10382         * config/cris/cris.h: Delete SDB reference in comment.
10383         * config/i386/cygming.h: Don't define SDB_DEBUGGING_INFO.
10384         (ASM_DECLARE_FUNCTION_NAME): Delete SDB reference from comment.
10385         * config/i386/gas.h: Don't define SDB_DEBUGGING_INFO.
10386         * config/i386/i386.c (svr4_dbx_register_map): Change SDB references
10387         to past tense.
10388         (ix86_expand_prologue): Likewise.
10389         * config/i386/winnt.c (i386_pe_start_function): Don't check SDB_DEBUG.
10390         * config/ia64/ia64.h: Likewise.
10391         * config/m68k/m68kelf.h (DBX_REGISTER_NUMBER): Delete SDB reference.
10392         * config/mips/mips.h (SUBTARGET_ASM_DEBUGGING_SPEC): Delete gcoff*
10393         support.
10394         * config/mmix/mmix.h: Likewise.
10395         * config/nds32/nds32.c: Likewise.
10396         * config/stormy/storym16.h: Likewise.
10397         * config/visium/visium.h: Likewise.
10398         * config/vx-common.h (SDB_DEBUGGING_INFO): Delete undef.
10400 2017-10-28  Sandra Loosemore  <sandra@codesourcery.com>
10402         * config/nios2/nios2.h (FRAME_GROWS_DOWNWARD): Define to 1.
10403         * config/nios2/nios2.c (nios2_initial_elimination_offset):  Make
10404         FRAME_POINTER_REGNUM point at high end of local var area.
10406 2017-10-27  Eric Botcazou  <ebotcazou@adacore.com>
10408         * bb-reorder.c (find_traces_1_round): Fix off-by-one index.
10409         Move comment around.  Do not reset best_edge for a copiable
10410         destination if the copy would cause a partition change.
10411         (better_edge_p): Remove redundant check.
10413 2017-10-27  Uros Bizjak  <ubizjak@gmail.com>
10415         * config/i386/i386-protos.h (ix86_fp_compare_mode): Remove prototype.
10417 2017-10-27  Michael Meissner  <meissner@linux.vnet.ibm.com>
10419         * builtins.c (CASE_MATHFN_FLOATN): New helper macro to add cases
10420         for math functions that have _Float<N> and _Float<N>X variants.
10421         (mathfn_built_in_2): Add support for math functions that have
10422         _Float<N> and _Float<N>X variants.
10423         (DEF_INTERNAL_FLT_FLOATN_FN): New helper macro.
10424         (expand_builtin_mathfn_ternary): Add support for fma with
10425         _Float<N> and _Float<N>X variants.
10426         (expand_builtin): Likewise.
10427         (fold_builtin_3): Likewise.
10428         * builtins.def (DEF_EXT_LIB_FLOATN_NX_BUILTINS): New macro to
10429         create math function _Float<N> and _Float<N>X variants as external
10430         library builtins.
10431         (BUILT_IN_COPYSIGN _Float<N> and _Float<N>X variants) Use
10432         DEF_EXT_LIB_FLOATN_NX_BUILTINS to make built-in functions using
10433         the __builtin_ prefix and if not strict ansi, without the prefix.
10434         (BUILT_IN_FABS _Float<N> and _Float<N>X variants): Likewise.
10435         (BUILT_IN_FMA _Float<N> and _Float<N>X variants): Likewise.
10436         (BUILT_IN_FMAX _Float<N> and _Float<N>X variants): Likewise.
10437         (BUILT_IN_FMIN _Float<N> and _Float<N>X variants): Likewise.
10438         (BUILT_IN_NAN _Float<N> and _Float<N>X variants): Likewise.
10439         (BUILT_IN_SQRT _Float<N> and _Float<N>X variants): Likewise.
10440         * builtin-types.def (BT_FN_FLOAT16_FLOAT16_FLOAT16_FLOAT16): New
10441         function signatures for fma _Float<N> and _Float<N>X variants.
10442         (BT_FN_FLOAT32_FLOAT32_FLOAT32_FLOAT32): Likewise.
10443         (BT_FN_FLOAT64_FLOAT64_FLOAT64_FLOAT64): Likewise.
10444         (BT_FN_FLOAT128_FLOAT128_FLOAT128_FLOAT128): Likewise.
10445         (BT_FN_FLOAT32X_FLOAT32X_FLOAT32X_FLOAT32X): Likewise.
10446         (BT_FN_FLOAT64X_FLOAT64X_FLOAT64X_FLOAT64X): Likewise.
10447         (BT_FN_FLOAT128X_FLOAT128X_FLOAT128X_FLOAT128X): Likewise.
10448         * gencfn-macros.c (print_case_cfn): Add support for math functions
10449         that have _Float<N> and _Float<N>X variants.
10450         (print_define_operator_list): Likewise.
10451         (fltfn_suffixes): Likewise.
10452         (main): Likewise.
10453         * internal-fn.def (DEF_INTERNAL_FLT_FLOATN_FN): New helper macro
10454         for math functions that have _Float<N> and _Float<N>X variants.
10455         (SQRT): Add support for sqrt, copysign, fmin and fmax _Float<N>
10456         and _Float<N>X variants.
10457         (COPYSIGN): Likewise.
10458         (FMIN): Likewise.
10459         (FMAX): Likewise.
10460         * fold-const.c (tree_call_nonnegative_warnv_p): Add support for
10461         copysign, fma, fmax, fmin, and sqrt _Float<N> and _Float<N>X
10462         variants.
10463         (integer_valued_read_call_p): Likewise.
10464         * fold-const-call.c (fold_const_call_ss): Likewise.
10465         (fold_const_call_sss): Add support for copysign, fmin, and fmax
10466         _Float<N> and _Float<N>X variants.
10467         (fold_const_call_ssss): Add support for fma _Float<N> and
10468         _Float<N>X variants.
10469         * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Add
10470         support for copysign and fma _Float<N> and _Float<N>X variants.
10471         (backprop::process_builtin_call_use): Likewise.
10472         * tree-call-cdce.c (can_test_argument_range); Add support for
10473         sqrt _Float<N> and _Float<N>X variants.
10474         (edom_only_function): Likewise.
10475         (get_no_error_domain): Likewise.
10476         * tree-ssa-math-opts.c (internal_fn_reciprocal): Likewise.
10477         * tree-ssa-reassoc.c (attempt_builtin_copysign): Add support for
10478         copysign _Float<N> and _Float<N>X variants.
10479         * config/rs6000/rs6000-builtin.def (SQRTF128): Delete, this is now
10480         handled by machine independent code.
10481         (FMAF128): Likewise.
10482         * doc/cpp.texi (Common Predefined Macros): Document defining
10483         __FP_FAST_FMAF<N> and __FP_FAST_FMAF<N>X if the backend supports
10484         fma _Float<N> and _Float<N>X variants.
10486 2017-10-27  Uros Bizjak  <ubizjak@gmail.com>
10488         PR target/82692
10489         * config/i386/i386-modes.def (CCFPU): Remove definition.
10490         * config/i386/i386.c (put_condition_mode): Remove CCFPU mode handling.
10491         (ix86_cc_modes_compatible): Ditto.
10492         (ix86_expand_carry_flag_compare): Ditto.
10493         (ix86_expand_int_movcc): Ditto.
10494         (ix86_expand_int_addcc): Ditto.
10495         (ix86_reverse_condition): Ditto.
10496         (ix86_unordered_fp_compare): Rename from ix86_fp_compare_mode.
10497         Return true/false for unordered/ordered fp comparisons.
10498         (ix86_cc_mode): Always return CCFPmode for float mode comparisons.
10499         (ix86_prepare_fp_compare_args): Update for rename.
10500         (ix86_expand_fp_compare): Update for rename.  Generate unordered
10501         compare RTXes wrapped with UNSPEC_NOTRAP unspec.
10502         (ix86_expand_sse_compare_and_jump): Ditto.
10503         * config/i386/predicates.md (fcmov_comparison_operator):
10504         Remove CCFPU mode handling.
10505         (ix86_comparison_operator): Ditto.
10506         (ix86_carry_flag_operator): Ditto.
10507         * config/i386/i386.md (UNSPEC_NOTRAP): New unspec.
10508         (*cmpu<mode>_i387): Wrap compare RTX with UNSPEC_NOTRAP unspec.
10509         (*cmpu<mode>_cc_i387): Ditto.
10510         (FPCMP): Remove mode iterator.
10511         (unord): Remove mode attribute.
10512         (unord_subst): New define_subst transformation
10513         (unord): New define_subst attribute.
10514         (unordered): Ditto.
10515         (*cmpi<unord><MODEF:mode>): Rewrite using unord_subst transformation.
10516         (*cmpi<unord>xf_i387): Ditto.
10517         * config/i386/sse.md (<sse>_<unord>comi<round_saeonly_name>): Merge
10518         from <sse>_comi<round_saeonly_name> and <sse>_ucomi<round_saeonly_name>
10519         using unord_subst transformation.
10520         * config/i386/subst.md (SUBST_A): Remove CCFP and CCFPU modes.
10521         (round_saeonly): Also handle CCFP mode.
10522         * reg-stack.c (subst_stack_regs_pat): Handle UNSPEC_NOTRAP unspec.
10523         Remove UNSPEC_SAHF unspec handling.
10525 2017-10-27  Jan Hubicka  <hubicka@ucw.cz>
10527         * x86-tune.def (X86_TUNE_INTER_UNIT_MOVES_TO_VEC): Disable for Zen.
10529 2017-10-27  Jeff Law  <law@redhat.com>
10531         * gimple-ssa-sprintf.c: Include domwalk.h.
10532         (class sprintf_dom_walker): New class, derived from dom_walker.
10533         (sprintf_dom_walker::before_dom_children): New function.
10534         (struct call_info): Moved into sprintf_dom_walker class
10535         (compute_formath_length, handle_gimple_call): Likewise.
10536         (sprintf_length::execute): Call the dominator walker rather
10537         than walking the statements.
10539         * tree-vrp.c (check_all_array_refs): Do not use wi->info to smuggle
10540         gimple statement locations.
10541         (check_array_bounds): Corresponding changes.  Get the statement's
10542         location directly from wi->stmt.
10544 2017-10-27  Palmer Dabbelt  <palmer@dabbelt.com>
10546         PR target/82717
10547         * doc/invoke.texi (RISC-V) <-mabi>: Correct and improve.
10549 2017-10-27  Jan Hubicka  <hubicka@ucw.cz>
10551         * config/i386/x86-tune.def (X86_TUNE_PARTIAL_REG_DEPENDENCY,
10552         X86_TUNE_MOVX): Disable for Haswell and newer CPUs.
10554 2017-10-27  Jakub Jelinek  <jakub@redhat.com>
10556         PR target/82703
10557         * config/i386/i386-protos.h (maybe_get_pool_constant): Removed.
10558         * config/i386/i386.c (maybe_get_pool_constant): Removed.
10559         (ix86_split_to_parts): Use avoid_constant_pool_reference instead of
10560         maybe_get_pool_constant.
10561         * config/i386/predicates.md (zero_extended_scalar_load_operand):
10562         Likewise.
10564 2017-10-27  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
10566         * doc/install.texi (Specific, i?86-*-solaris2.10): Simplify gas
10567         2.26 caveat.  Update gas and gld versions.
10568         (Specific, *-*-solaris2*): Update binutils version.  Remove caveat
10569         reference.
10571 2017-10-27  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
10573         * cgraph.h (set_malloc_flag): Declare.
10574         * cgraph.c (set_malloc_flag_1): New function.
10575         (set_malloc_flag): Likewise.
10576         * ipa-fnsummary.h (ipa_call_summary): Add new field is_return_callee.
10577         * ipa-fnsummary.c (ipa_call_summary::reset): Set is_return_callee to
10578         false.
10579         (read_ipa_call_summary): Add support for reading is_return_callee.
10580         (write_ipa_call_summary): Stream is_return_callee.
10581         * ipa-inline.c (ipa_inline): Remove call to ipa_free_fn_summary.
10582         * ipa-pure-const.c: Add headers ssa.h, alloc-pool.h, symbol-summary.h,
10583         ipa-prop.h, ipa-fnsummary.h.
10584         (pure_const_names): Change to static.
10585         (malloc_state_e): Define.
10586         (malloc_state_names): Define.
10587         (funct_state_d): Add field malloc_state.
10588         (varying_state): Set malloc_state to STATE_MALLOC_BOTTOM.
10589         (check_retval_uses): New function.
10590         (malloc_candidate_p): Likewise.
10591         (analyze_function): Add support for malloc attribute.
10592         (pure_const_write_summary): Stream malloc_state.
10593         (pure_const_read_summary): Add support for reading malloc_state.
10594         (dump_malloc_lattice): New function.
10595         (propagate_malloc): New function.
10596         (warn_function_malloc): New function.
10597         (ipa_pure_const::execute): Call propagate_malloc and
10598         ipa_free_fn_summary.
10599         (pass_local_pure_const::execute): Add support for malloc attribute.
10600         * ssa-iterators.h (RETURN_FROM_IMM_USE_STMT): New macro.
10601         * doc/invoke.texi: Document Wsuggest-attribute=malloc.
10603 2017-10-27  Martin Liska  <mliska@suse.cz>
10605         PR gcov-profile/82457
10606         * doc/invoke.texi: Document that one needs a non-strict ISO mode
10607         for fork-like functions to be properly instrumented.
10609 2017-10-27  Richard Biener  <rguenther@suse.de>
10611         PR middle-end/81659
10612         * tree-eh.c (pass_lower_eh_dispatch::execute): Free dominator
10613         info when we redirected EH.
10615 2017-10-26  Michael Collison  <michael.collison@arm.com>
10617         * config/aarch64/aarch64.md(<optab>_trunc><vf><GPI:mode>2):
10618         New pattern.
10619         (<optab>_trunchf<GPI:mode>2: New pattern.
10620         (<optab>_trunc<vgp><GPI:mode>2: New pattern.
10621         * config/aarch64/iterators.md (wv): New mode attribute.
10622         (vf, VF): New mode attributes.
10623         (vgp, VGP): New mode attributes.
10624         (s): Update attribute with SImode and DImode prefixes.
10626 2017-10-26  Sandra Loosemore  <sandra@codesourcery.com>
10628         * config/nios2/constraints.md ("S"): Match r0rel_constant_p too.
10629         * config/nios2/nios2-protos.h (r0rel_constant_p): Declare.
10630         * config/nios2/nios2.c: (nios2_r0rel_sec_regex): New.
10631         (nios2_option_overide): Initialize it.  Don't allow R0-relative
10632         addressing with PIC.
10633         (nios2_rtx_costs): Handle r0rel_constant_p like gprel_constant_p.
10634         (nios2_symbolic_constant_p): Likewise.
10635         (nios2_legitimate_address_p): Likewise.
10636         (nios2_r0rel_section_name_p): New.
10637         (nios2_symbol_ref_in_r0rel_data_p): New.
10638         (nios2_emit_move_sequence): Handle r0rel_constant_p.
10639         (r0rel_constant_p): New.
10640         (nios2_print_operand_address): Handle r0rel_constant_p.
10641         (nios2_cdx_narrow_form_p): Likewise.
10642         * config/nios2/nios2.opt (mr0rel-sec=): New option.
10643         * doc/invoke.texi (Option Summary): Add -mr0rel-sec.
10644         (Nios II Options): Document -mr0rel-sec.
10646 2017-10-26  Sandra Loosemore  <sandra@codesourcery.com>
10648         * config/nios2/nios2.c: Include xregex.h.
10649         (nios2_gprel_sec_regex): New.
10650         (nios2_option_overide): Initialize it.  Don't allow GP-relative
10651         addressing with PIC.
10652         (nios2_small_section_name_p): Check for regex match.
10653         * config/nios2/nios2.opt (mgprel-sec=): New option.
10654         * doc/invoke.texi (Option Summary): Add -mgprel-sec.
10655         (Nios II Options): Document -mgprel-sec.
10657 2017-10-26  Jim Wilson  <wilson@tuliptree.org>
10659         * doc/invoke.texi (-fdebug-prefix-map): Expand documentation.
10661 2017-10-26  Tom de Vries  <tom@codesourcery.com>
10663         PR tree-optimization/82707
10664         * gimple.c (gimple_copy): Fix unsharing of
10665         GIMPLE_OMP_{SINGLE,TARGET,TEAMS}.
10667 2017-10-26  Olga Makhotina  <olga.makhotina@intel.com>
10669         * config/i386/avx512fintrin.h (_mm512_cmpeq_pd_mask,
10670         _mm512_cmple_pd_mask, _mm512_cmplt_pd_mask,
10671         _mm512_cmpneq_pd_mask, _mm512_cmpnle_pd_mask,
10672         _mm512_cmpnlt_pd_mask, _mm512_cmpord_pd_mask,
10673         _mm512_cmpunord_pd_mask, _mm512_mask_cmpeq_pd_mask,
10674         _mm512_mask_cmple_pd_mask, _mm512_mask_cmplt_pd_mask,
10675         _mm512_mask_cmpneq_pd_mask, _mm512_mask_cmpnle_pd_mask,
10676         _mm512_mask_cmpnlt_pd_mask, _mm512_mask_cmpord_pd_mask,
10677         _mm512_mask_cmpunord_pd_mask, _mm512_cmpeq_ps_mask,
10678         _mm512_cmple_ps_mask, _mm512_cmplt_ps_mask,
10679         _mm512_cmpneq_ps_mask, _mm512_cmpnle_ps_mask,
10680         _mm512_cmpnlt_ps_mask, _mm512_cmpord_ps_mask,
10681         _mm512_cmpunord_ps_mask, _mm512_mask_cmpeq_ps_mask,
10682         _mm512_mask_cmple_ps_mask, _mm512_mask_cmplt_ps_mask,
10683         _mm512_mask_cmpneq_ps_mask, _mm512_mask_cmpnle_ps_mask,
10684         _mm512_mask_cmpnlt_ps_mask, _mm512_mask_cmpord_ps_mask,
10685         _mm512_mask_cmpunord_ps_mask): New intrinsics.
10687 2017-10-26  Michael Meissner  <meissner@linux.vnet.ibm.com>
10689         * config/rs6000/aix.h (TARGET_IEEEQUAD_DEFAULT): Set long double
10690         default to IBM.
10691         * config/rs6000/darwin.h (TARGET_IEEEQUAD_DEFAULT): Likewise.
10692         * config/rs6000/rs6000.opt (-mabi=ieeelongdouble): Move the
10693         warning to rs6000.c.  Remove the Undocumented flag, since it has
10694         been documented.
10695         (-mabi=ibmlongdouble): Likewise.
10696         * config/rs6000/rs6000.c (TARGET_IEEEQUAD_DEFAULT): If it is not
10697         already set, set the default format for long double.
10698         (rs6000_debug_reg_global): Print whether long double is IBM or
10699         IEEE.
10700         (rs6000_option_override_internal): Rework setting long double
10701         format.  Only warn if the user is changing the long double default
10702         and they did not use -Wno-psabi.
10703         * doc/invoke.texi (PowerPC options): Update the documentation for
10704         -mabi=ieeelongdouble and -mabi=ibmlongdouble.
10706 2017-10-26  Richard Sandiford  <richard.sandiford@linaro.org>
10707             Alan Hayward  <alan.hayward@arm.com>
10708             David Sherwood  <david.sherwood@arm.com>
10710         * rtl.h (wider_subreg_mode): New function.
10711         * ira.h (ira_sort_regnos_for_alter_reg): Take a machine_mode *
10712         rather than an unsigned int *.
10713         * ira-color.c (regno_max_ref_width): Replace with...
10714         (regno_max_ref_mode): ...this new variable.
10715         (coalesced_pseudo_reg_slot_compare): Update accordingly.
10716         Use wider_subreg_mode.
10717         (ira_sort_regnos_for_alter_reg): Likewise.  Take a machine_mode *
10718         rather than an unsigned int *.
10719         * lra-constraints.c (uses_hard_regs_p): Use wider_subreg_mode.
10720         (process_alt_operands): Likewise.
10721         (invariant_p): Likewise.
10722         * lra-spills.c (assign_mem_slot): Likewise.
10723         (add_pseudo_to_slot): Likewise.
10724         * lra.c (collect_non_operand_hard_regs): Likewise.
10725         (add_regs_to_insn_regno_info): Likewise.
10726         * reload1.c (regno_max_ref_width): Replace with...
10727         (regno_max_ref_mode): ...this new variable.
10728         (reload): Update accordingly.  Update call to
10729         ira_sort_regnos_for_alter_reg.
10730         (alter_reg): Update to use regno_max_ref_mode.  Call wider_subreg_mode.
10731         (init_eliminable_invariants): Update to use regno_max_ref_mode.
10732         (scan_paradoxical_subregs): Likewise.
10734 2017-10-26  Wilco Dijkstra  <wdijkstr@arm.com>
10736         * config/aarch64/aarch64.h (EXIT_IGNORE_STACK): Set if alloca is used.
10737         (aarch64_frame): Add emit_frame_chain boolean.
10738         * config/aarch64/aarch64.c (aarch64_frame_pointer_required)
10739         Move eh_return case to aarch64_layout_frame.
10740         (aarch64_layout_frame): Initialize emit_frame_chain.
10741         (aarch64_expand_prologue): Use emit_frame_chain.
10743 2017-10-26  Wilco Dijkstra  <wdijkstr@arm.com>
10745         * config/aarch64/aarch64.c (aarch64_layout_frame):
10746         Ensure LR is always stored at the bottom of the callee-saves.
10747         Remove rarely used frame layout which saves callee-saves at top of
10748         frame, so the store of LR can be used as a valid probe in all cases.
10750 2017-10-26  Wilco Dijkstra  <wdijkstr@arm.com>
10752         * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
10753         Improve unaligned TImode/TFmode base/offset split.
10755 2017-10-26  Richard Sandiford  <richard.sandiford@linaro.org>
10756             Alan Hayward  <alan.hayward@arm.com>
10757             David Sherwood  <david.sherwood@arm.com>
10759         * caller-save.c (mark_referenced_regs):  Use read_modify_subreg_p.
10760         * combine.c (find_single_use_1): Likewise.
10761         (expand_field_assignment): Likewise.
10762         (move_deaths): Likewise.
10763         * lra-constraints.c (simplify_operand_subreg): Likewise.
10764         (curr_insn_transform): Likewise.
10765         * lra.c (collect_non_operand_hard_regs): Likewise.
10766         (add_regs_to_insn_regno_info): Likewise.
10767         * rtlanal.c (reg_referenced_p): Likewise.
10768         (covers_regno_no_parallel_p): Likewise.
10770 2017-10-26  Richard Sandiford  <richard.sandiford@linaro.org>
10772         * wide-int-print.cc (print_hex): Loop based on extract_uhwi.
10773         Don't print any bits outside the precision of the value.
10774         * wide-int.cc (test_printing): Add some new tests.
10776 2017-10-26  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
10778         * configure.ac (gcc_cv_as_ix86_xbrace_comment): Check if assembler
10779         supports -xbrace_comment option.
10780         * configure: Regenerate.
10781         * config.in: Regenerate.
10782         * config/i386/sol2.h (ASM_XBRACE_COMMENT_SPEC): Define.
10783         (ASM_CPU_SPEC): Use it.
10785 2017-10-26  Richard Sandiford  <richard.sandiford@linaro.org>
10787         * target.def (static_rtx_alignment): New hook.
10788         * targhooks.h (default_static_rtx_alignment): Declare.
10789         * targhooks.c (default_static_rtx_alignment): New function.
10790         * doc/tm.texi.in (TARGET_STATIC_RTX_ALIGNMENT): New hook.
10791         * doc/tm.texi: Regenerate.
10792         * varasm.c (force_const_mem): Use targetm.static_rtx_alignment
10793         instead of targetm.constant_alignment.  Remove call to
10794         set_mem_attributes.
10795         * config/cris/cris.c (TARGET_STATIC_RTX_ALIGNMENT): Redefine.
10796         (cris_preferred_mininum_alignment): New function, split out from...
10797         (cris_constant_alignment): ...here.
10798         (cris_static_rtx_alignment): New function.
10799         * config/i386/i386.c (ix86_static_rtx_alignment): New function,
10800         split out from...
10801         (ix86_constant_alignment): ...here.
10802         (TARGET_STATIC_RTX_ALIGNMENT): Redefine.
10803         * config/mmix/mmix.c (TARGET_STATIC_RTX_ALIGNMENT): Redefine.
10804         (mmix_static_rtx_alignment): New function.
10805         * config/spu/spu.c (spu_static_rtx_alignment): New function.
10806         (TARGET_STATIC_RTX_ALIGNMENT): Redefine.
10808 2017-10-26  Tamar Christina  <tamar.christina@arm.com>
10810         PR target/81800
10811         * config/aarch64/aarch64.md (lrint<GPF:mode><GPI:mode>2):
10812         Add flag_trapping_math and flag_fp_int_builtin_inexact.
10814 2017-10-25  Palmer Dabbelt  <palmer@dabbelt.com>
10816         * config/riscv/riscv.md (ZERO_EXTEND_LOAD): Define.
10817         * config/riscv/pic.md (local_pic_load): Rename to local_pic_load_s,
10818         mark as a sign-extending load.
10819         (local_pic_load_u): Define.
10821 2017-10-25  Eric Botcazou  <ebotcazou@adacore.com>
10823         PR middle-end/82062
10824         * fold-const.c (operand_equal_for_comparison_p): Also return true
10825         if ARG0 is a simple variant of ARG1 with narrower precision.
10826         (fold_ternary_loc): Always pass unstripped operands to the predicate.
10828 2017-10-25  Jan Hubicka  <hubicka@ucw.cz>
10830         * i386.c (ix86_builtin_vectorization_cost): Compute scatter/gather
10831         cost correctly.
10832         * i386.h (processor_costs): Add gather_static, gather_per_elt,
10833         scatter_static, scatter_per_elt.
10834         * x86-tune-costs.h: Add new cost entries.
10836 2017-10-25  Richard Biener  <rguenther@suse.de>
10838         * tree-ssa-sccvn.h (vn_eliminate): Declare.
10839         * tree-ssa-pre.c (class eliminate_dom_walker, eliminate,
10840         class pass_fre): Move to ...
10841         * tree-ssa-sccvn.c (class eliminate_dom_walker, vn_eliminate,
10842         class pass_fre): ... here and adjust for statistics.
10844 2017-10-25  Jakub Jelinek  <jakub@redhat.com>
10846         PR libstdc++/81706
10847         * attribs.c (attribute_value_equal): Use omp_declare_simd_clauses_equal
10848         for comparison of OMP_CLAUSEs regardless of flag_openmp{,_simd}.
10849         (duplicate_one_attribute, copy_attributes_to_builtin): New functions.
10850         * attribs.h (duplicate_one_attribute, copy_attributes_to_builtin): New
10851         declarations.
10853 2017-10-25  Richard Biener  <rguenther@suse.de>
10855         * tree-ssa-pre.c (need_eh_cleanup, need_ab_cleanup, el_to_remove,
10856         el_to_fixup, el_todo, el_avail, el_avail_stack, eliminate_avail,
10857         eliminate_push_avail, eliminate_insert): Move inside...
10858         (class eliminate_dom_walker): ... this class in preparation
10859         of move.
10860         (fini_eliminate): Remove by merging with ...
10861         (eliminate): ... this function.  Adjust for class changes.
10862         (pass_pre::execute): Remove fini_eliminate call.
10863         (pass_fre::execute): Likewise.
10865 2017-10-24  Jakub Jelinek  <jakub@redhat.com>
10867         PR target/82460
10868         * config/i386/sse.md (UNSPEC_VPERMI2, UNSPEC_VPERMI2_MASK): Remove.
10869         (VPERMI2, VPERMI2I): New mode iterators.
10870         (<avx512>_vpermi2var<mode>3_maskz): Remove 3 define_expand patterns.
10871         (<avx512>_vpermi2var<mode>3<sd_maskz_name>): Remove 3 define_insn
10872         patterns.
10873         (<avx512>_vpermi2var<mode>3_mask): New define_expand using VPERMI2
10874         mode iterator.  Remove 3 old define_insn patterns.
10875         (*<avx512>_vpermi2var<mode>3_mask): 2 new define_insn patterns.
10876         (<avx512>_vpermt2var<mode>3_maskz): Adjust 1 define_expand to use
10877         VPERMI2 mode iterator, remove the other two expanders.
10878         (<avx512>_vpermt2var<mode>3<sd_maskz_name>): Adjust 1 define_insn
10879         to use VPERMI2 mode iterator, add another alternative for vpermi2*
10880         instructions, remove the other two patterns.
10881         (<avx512>_vpermt2var<mode>3_mask): Adjust 1 define_insn to use VPERMI2
10882         mode iterator, remove the other two patterns.
10883         * config/i386/i386.c (ix86_expand_vec_perm_vpermi2): Renamed to ...
10884         (ix86_expand_vec_perm_vpermt2): ... this.  Swap mask and op0
10885         arguments, use gen_*vpermt2* expanders instead of gen_*vpermi2*
10886         and adjust argument order accordingly.
10887         (ix86_expand_vec_perm): Adjust caller.
10888         (expand_vec_perm_1): Likewise.
10889         (expand_vec_perm_vpermi2_vpshub2): Rename to ...
10890         (expand_vec_perm_vpermt2_vpshub2): ... this.
10891         (ix86_expand_vec_perm_const_1): Adjust caller.
10892         (ix86_vectorize_vec_perm_const_ok): Adjust comments.
10894         PR target/82370
10895         * config/i386/sse.md (VIMAX_AVX2): Remove V4TImode.
10896         (VIMAX_AVX2_AVX512BW, VIMAX_AVX512VL): New mode iterators.
10897         (vec_shl_<mode>): Remove unused expander.
10898         (avx512bw_<shift_insn><mode>3): New define_insn.
10899         (<sse2_avx2>_ashl<mode>3, <sse2_avx2>_lshr<mode>3): Replaced by ...
10900         (<sse2_avx2>_<shift_insn><mode>3): ... this.  New define_insn.
10902 2017-10-24  Paolo Carlini  <paolo.carlini@oracle.com>
10904         PR c++/82466
10905         * doc/invoke.texi ([Wbuiltin-declaration-mismatch]): Extend
10906         description.
10908 2017-10-24  Wilco Dijkstra  <wdijkstr@arm.com>
10910         PR rtl-optimization/82396
10911         * gcc/haifa-sched.c (ready_sort_real): Remove qsort workaround.
10912         (autopref_multipass_init): Simplify initialization.
10913         (autopref_rank_data): Simplify sort order.
10914         * gcc/sched-int.h (autopref_multipass_data_): Remove
10915         multi_mem_insn_p, min_offset and max_offset.
10917 2017-10-24  Wilco Dijkstra  <wdijkstr@arm.com>
10919         PR middle-end/60580
10920         * config/aarch64/aarch64.c (aarch64_frame_pointer_required)
10921         Check special value of flag_omit_frame_pointer.
10922         (aarch64_can_eliminate): Likewise.
10923         (aarch64_override_options_after_change_1): Simplify handling of
10924         -fomit-frame-pointer and -fomit-leaf-frame-pointer.
10926 2017-10-24  Richard Biener  <rguenther@suse.de>
10928         PR tree-optimization/82697
10929         * tree-ssa-phiopt.c (cond_store_replacement): Use alias-set
10930         zero for conditional load and unconditional store.
10932 2017-10-24  H.J. Lu  <hongjiu.lu@intel.com>
10934         * doc/install.texi: Document bootstrap-cet.
10936 2017-10-24  H.J. Lu  <hongjiu.lu@intel.com>
10938         PR target/82659
10939         * config/i386/i386.c (rest_of_insert_endbranch): Don't insert
10940         ENDBR instruction at function entrance if function is only
10941         called directly.
10943 2017-10-24  Jakub Jelinek  <jakub@redhat.com>
10945         PR target/82628
10946         * config/i386/i386.md (addcarry<mode>, subborrow<mode>): Change
10947         patterns to better describe from which operation the CF is computed.
10948         (addcarry<mode>_0, subborrow<mode>_0): New patterns.
10949         * config/i386/i386.c (ix86_expand_builtin) <case handlecarry>: Pass
10950         one LTU with [DT]Imode and another one with [SD]Imode.  If arg0
10951         is 0, use _0 suffixed expanders instead of emitting a comparison
10952         before it.
10954 2017-10-06  Sergey Shalnov  <Sergey.Shalnov@intel.com>
10956         * config/i386/i386.md(*movsf_internal, *movdf_internal):
10957         Avoid 512-bit AVX modes for TARGET_PREFER_AVX256.
10959 2017-10-24  Eric Botcazou  <ebotcazou@adacore.com>
10961         PR middle-end/82569
10962         * tree-outof-ssa.h (always_initialized_rtx_for_ssa_name_p): Delete.
10963         * expr.c (expand_expr_real_1) <expand_decl_rtl>: Revert latest change.
10964         * loop-iv.c (iv_get_reaching_def): Likewise.
10965         * cfgexpand.c (expand_one_ssa_partition): Initialize the RTX if the
10966         variable is promoted and the partition contains undefined values.
10968 2017-10-23  Sandra Loosemore  <sandra@codesourcery.com>
10970         * config/nios2/nios2.c (nios2_rtx_costs): Make costs better
10971         reflect reality.
10972         (nios2_address_cost): Define.
10973         (nios2_legitimize_address): Recognize (exp + constant) directly.
10974         (TARGET_ADDRESS_COST): Define.
10976 2017-10-23  Sandra Loosemore  <sandra@codesourcery.com>
10978         * config/nios2/nios2-protos.h (nios2_large_constant_p): Declare.
10979         (nios2_symbolic_memory_operand_p): Declare.
10980         (nios2_split_large_constant): Declare.
10981         (nios2_split_symbolic_memory_operand): Declare.
10982         * config/nios2/nios2.c: Adjust includes.
10983         (nios2_symbolic_constant_allowed): New.
10984         (nios2_symbolic_constant_p): New.
10985         (nios2_plus_symbolic_constant_p): New.
10986         (nios2_valid_addr_expr_p): Recognize addresses involving
10987         symbolic constants.
10988         (nios2_legitimate_address_p): Likewise, also LO_SUM.
10989         (nios2_symbolic_memory_operand_p): New.
10990         (nios2_large_constant_p): New.
10991         (nios2_split_large_constant): New.
10992         (nios2_split_plus_large_constant): New.
10993         (nios2_split_symbolic_memory_operand): New.
10994         (nios2_legitimize_address): Code refactoring.  Handle addresses
10995         involving symbolic constants.
10996         (nios2_emit_move_sequence): Likewise.
10997         (nios2_print_operand): Improve error output.
10998         (nios2_print_operand_address): Handle LO_SUM.
10999         (nios2_cdx_narrow_form_p): Likewise.
11000         * config/nios2/nios2.md (movqi_internal): Add splitter for memory
11001         operands involving symbolic constants.
11002         (movhi_internal, movsi_internal): Likewise.
11003         (zero_extendhisi2, zero_extendqi<mode>2): Likewise.
11004         (extendhisi2, extendqi<mode>2): Likewise.
11006 2017-10-23  Sandra Loosemore  <sandra@codesourcery.com>
11008         * tree-pass.h (PROP_rtl_split_insns): Define.
11009         * recog.c (pass_data_split_all_insns): Provide PROP_rtl_split_insns.
11011 2017-10-23  Sandra Loosemore  <sandra@codesourcery.com>
11013         * config/nios2/nios2.c (TARGET_LRA_P): Don't override.
11015 2017-10-23  Jakub Jelinek  <jakub@redhat.com>
11017         PR debug/82630
11018         * target.def (const_not_ok_for_debug_p): Default to
11019         default_const_not_ok_for_debug_p instead of hook_bool_rtx_false.
11020         * targhooks.h (default_const_not_ok_for_debug_p): New declaration.
11021         * targhooks.c (default_const_not_ok_for_debug_p): New function.
11022         * dwarf2out.c (const_ok_for_output_1): Only reject UNSPECs for
11023         which targetm.const_not_ok_for_debug_p returned true.
11024         * config/arm/arm.c (arm_const_not_ok_for_debug_p): Return true
11025         for UNSPECs.
11026         * config/powerpcspe/powerpcspe.c (rs6000_const_not_ok_for_debug_p):
11027         Likewise.
11028         * config/rs6000/rs6000.c (rs6000_const_not_ok_for_debug_p): Likewise.
11029         * config/i386/i386.c (ix86_delegitimize_address_1): Don't delegitimize
11030         UNSPEC_GOTOFF with addend into addend - _GLOBAL_OFFSET_TABLE_ + symbol
11031         if !base_term_p.
11032         (ix86_const_not_ok_for_debug_p): New function.
11033         (i386_asm_output_addr_const_extra): Handle UNSPEC_GOTOFF.
11034         (TARGET_CONST_NOT_OK_FOR_DEBUG_P): Redefine.
11036 2017-10-23  David Malcolm  <dmalcolm@redhat.com>
11038         PR bootstrap/82610
11039         * system.h: Conditionally include "unique-ptr.h" if
11040         INCLUDE_UNIQUE_PTR is defined.
11041         * unique-ptr-tests.cc: Remove include of "unique-ptr.h" in favor
11042         of defining INCLUDE_UNIQUE_PTR before including "system.h".
11044 2017-10-23  Sebastian Perta  <sebastian.perta@renesas.com>
11046         * config/rl78/rl78.md: New define_expand "subdi3".
11048 2017-10-23  H.J. Lu  <hongjiu.lu@intel.com>
11050         PR target/82673
11051         * config/i386/i386.c (ix86_finalize_stack_frame_flags): Skip
11052         DF_REF_INSN if DF_REF_INSN_INFO is false.
11054 2017-10-23  Jan Hubicka  <hubicka@ucw.cz>
11056         * i386.c (dimode_scalar_chain::compute_convert_gain): Use
11057         xmm_move instead of sse_move.
11058         (sse_store_index): New function.
11059         (ix86_register_move_cost): Be more sensible about mismatch stall;
11060         model AVX moves correctly; make difference between sse->integer and
11061         integer->sse.
11062         (ix86_builtin_vectorization_cost): Model correctly aligned and unaligned
11063         moves; make difference between SSE and AVX.
11064         * i386.h (processor_costs): Remove sse_move; add xmm_move, ymm_move
11065         and zmm_move. Increase size of sse load and store tables;
11066         add unaligned load and store tables; add ssemmx_to_integer.
11067         * x86-tune-costs.h: Update all entries according to real
11068         move latencies from Agner Fog's manual and chip documentation.
11070 2017-10-23  Jakub Jelinek  <jakub@redhat.com>
11072         PR target/82628
11073         * config/i386/predicates.md (x86_64_dwzext_immediate_operand): New.
11074         * config/i386/constraints.md (Wf): New constraint.
11075         * config/i386/i386.md (UNSPEC_SBB): New unspec.
11076         (cmp<dwi>_doubleword): Removed.
11077         (sub<mode>3_carry_ccc, *sub<mode>3_carry_ccc_1): New patterns.
11078         (sub<mode>3_carry_ccgz): Use unspec instead of compare.
11079         * config/i386/i386.c (ix86_expand_branch) <case E_TImode>: Don't
11080         expand with cmp<dwi>_doubleword.  For LTU and GEU use
11081         sub<mode>3_carry_ccc instead of sub<mode>3_carry_ccgz and use CCCmode.
11083         * common.opt (gcolumn-info): Enable by default.
11084         * doc/invoke.texi (gcolumn-info): Document new default.
11086 2017-10-23  Richard Biener  <rguenther@suse.de>
11088         PR tree-optimization/82672
11089         * graphite-isl-ast-to-gimple.c (graphite_copy_stmts_from_block):
11090         Fold the stmt if we propagated into it.
11092 2017-10-23  Richard Biener  <rguenther@suse.de>
11094         * tree-ssa-pre.c (bitmap_remove_from_set): Rename to...
11095         (bitmap_remove_expr_from_set): ... this.  All callers call this
11096         for non-constant values.
11097         (bitmap_set_subtract): Rename to...
11098         (bitmap_set_subtract_expressions): ... this.  Adjust and
11099         optimize.
11100         (bitmap_set_contains_value): Remove superfluous check.
11101         (bitmap_set_replace_value): Inline into single caller ...
11102         (bitmap_value_replace_in_set): ... here and simplify.
11103         (dependent_clean): Merge into ...
11104         (clean): ... this using an overload.  Adjust.
11105         (prune_clobbered_mems): Adjust.
11106         (compute_antic_aux): Likewise.
11107         (compute_partial_antic_aux): Likewise.
11109 2017-10-23  Richard Biener  <rguenther@suse.de>
11111         PR tree-optimization/82129
11112         Revert
11113         2017-08-01  Richard Biener  <rguenther@suse.de>
11115         PR tree-optimization/81181
11116         * tree-ssa-pre.c (compute_antic_aux): Defer clean() to ...
11117         (compute_antic): ... end of iteration here.
11119 2017-10-23  Richard Sandiford  <richard.sandiford@linaro.org>
11121         * target.def (starting_frame_offset): New hook.
11122         * doc/tm.texi (STARTING_FRAME_OFFSET): Remove in favor of...
11123         (TARGET_STARTING_FRAME_OFFSET): ...this new hook.
11124         * doc/tm.texi.in: Regenerate.
11125         * hooks.h (hook_hwi_void_0): Declare.
11126         * hooks.c (hook_hwi_void_0): New function.
11127         * doc/rtl.texi: Refer to TARGET_STARTING_FRAME_OFFSET instead of
11128         STARTING_FRAME_OFFSET.
11129         * builtins.c (expand_builtin_setjmp_receiver): Likewise.
11130         * reload1.c (reload): Likewise.
11131         * cfgexpand.c (expand_used_vars): Use targetm.starting_frame_offset
11132         instead of STARTING_FRAME_OFFSET.
11133         * function.c (try_fit_stack_local): Likewise.
11134         (assign_stack_local_1): Likewise
11135         (instantiate_virtual_regs): Likewise.
11136         * rtlanal.c (rtx_addr_can_trap_p_1): Likewise.
11137         * config/avr/avr.md (nonlocal_goto_receiver): Likewise.
11138         * config/aarch64/aarch64.h (STARTING_FRAME_OFFSET): Delete.
11139         * config/alpha/alpha.h (STARTING_FRAME_OFFSET): Likewise.
11140         * config/arc/arc.h (STARTING_FRAME_OFFSET): Likewise.
11141         * config/arm/arm.h (STARTING_FRAME_OFFSET): Likewise.
11142         * config/bfin/bfin.h (STARTING_FRAME_OFFSET): Likewise.
11143         * config/c6x/c6x.h (STARTING_FRAME_OFFSET): Likewise.
11144         * config/cr16/cr16.h (STARTING_FRAME_OFFSET): Likewise.
11145         * config/cris/cris.h (STARTING_FRAME_OFFSET): Likewise.
11146         * config/fr30/fr30.h (STARTING_FRAME_OFFSET): Likewise.
11147         * config/frv/frv.h (STARTING_FRAME_OFFSET): Likewise.
11148         * config/ft32/ft32.h (STARTING_FRAME_OFFSET): Likewise.
11149         * config/h8300/h8300.h (STARTING_FRAME_OFFSET): Likewise.
11150         * config/i386/i386.h (STARTING_FRAME_OFFSET): Likewise.
11151         * config/ia64/ia64.h (STARTING_FRAME_OFFSET): Likewise.
11152         * config/m32c/m32c.h (STARTING_FRAME_OFFSET): Likewise.
11153         * config/m68k/m68k.h (STARTING_FRAME_OFFSET): Likewise.
11154         * config/mcore/mcore.h (STARTING_FRAME_OFFSET): Likewise.
11155         * config/mn10300/mn10300.h (STARTING_FRAME_OFFSET): Likewise.
11156         * config/moxie/moxie.h (STARTING_FRAME_OFFSET): Likewise.
11157         * config/msp430/msp430.h (STARTING_FRAME_OFFSET): Likewise.
11158         * config/nds32/nds32.h (STARTING_FRAME_OFFSET): Likewise.
11159         * config/nios2/nios2.h (STARTING_FRAME_OFFSET): Likewise.
11160         * config/nvptx/nvptx.h (STARTING_FRAME_OFFSET): Likewise.
11161         * config/pdp11/pdp11.h (STARTING_FRAME_OFFSET): Likewise.
11162         * config/riscv/riscv.h (STARTING_FRAME_OFFSET): Likewise.
11163         * config/rl78/rl78.h (STARTING_FRAME_OFFSET): Likewise.
11164         * config/rx/rx.h (STARTING_FRAME_OFFSET): Likewise.
11165         * config/s390/s390.h (STARTING_FRAME_OFFSET): Likewise.
11166         * config/sh/sh.h (STARTING_FRAME_OFFSET): Likewise.
11167         * config/sparc/sparc.c (sparc_compute_frame_size): Likewise.
11168         * config/sparc/sparc.h (STARTING_FRAME_OFFSET): Likewise.
11169         * config/spu/spu.h (STARTING_FRAME_OFFSET): Likewise.
11170         * config/stormy16/stormy16.h (STARTING_FRAME_OFFSET): Likewise.
11171         * config/tilegx/tilegx.h (STARTING_FRAME_OFFSET): Likewise.
11172         * config/tilepro/tilepro.h (STARTING_FRAME_OFFSET): Likewise.
11173         * config/v850/v850.h (STARTING_FRAME_OFFSET): Likewise.
11174         * config/visium/visium.h (STARTING_FRAME_OFFSET): Likewise.
11175         * config/avr/avr.h (STARTING_FRAME_OFFSET): Likewise.
11176         * config/avr/avr-protos.h (avr_starting_frame_offset): Likewise.
11177         * config/avr/avr.c (avr_starting_frame_offset): Make static and
11178         return a HOST_WIDE_INT.
11179         (avr_builtin_setjmp_frame_value): Use it instead of
11180         STARTING_FRAME_OFFSET.
11181         (TARGET_STARTING_FRAME_OFFSET): Redefine.
11182         * config/epiphany/epiphany.h (STARTING_FRAME_OFFSET): Delete.
11183         * config/epiphany/epiphany.c (epiphany_starting_frame_offset):
11184         New function.
11185         (TARGET_STARTING_FRAME_OFFSET): Redefine.
11186         * config/iq2000/iq2000.h (STARTING_FRAME_OFFSET): Delete.
11187         * config/iq2000/iq2000.c (iq2000_starting_frame_offset): New function.
11188         (TARGET_CONSTANT_ALIGNMENT): Redefine.
11189         * config/lm32/lm32.h (STARTING_FRAME_OFFSET): Delete.
11190         * config/lm32/lm32.c (lm32_starting_frame_offset): New function.
11191         (TARGET_STARTING_FRAME_OFFSET): Redefine.
11192         * config/m32r/m32r.h (STARTING_FRAME_OFFSET): Delete.
11193         * config/m32r/m32r.c (m32r_starting_frame_offset): New function.
11194         (TARGET_STARTING_FRAME_OFFSET): Redefine.
11195         * config/microblaze/microblaze.h (STARTING_FRAME_OFFSET): Delete.
11196         * config/microblaze/microblaze.c (microblaze_starting_frame_offset):
11197         New function.
11198         (TARGET_STARTING_FRAME_OFFSET): Redefine.
11199         * config/mips/mips.h (STARTING_FRAME_OFFSET): Delete.
11200         * config/mips/mips.c (mips_compute_frame_info): Refer to
11201         TARGET_STARTING_FRAME_OFFSET instead of STARTING_FRAME_OFFSET.
11202         (mips_starting_frame_offset): New function.
11203         (TARGET_STARTING_FRAME_OFFSET): Redefine.
11204         * config/mmix/mmix.h (STARTING_FRAME_OFFSET): Delete.
11205         * config/mmix/mmix-protos.h (mmix_starting_frame_offset): Delete.
11206         * config/mmix/mmix.c (mmix_starting_frame_offset): Make static
11207         and return a HOST_WIDE_INT.
11208         (TARGET_STARTING_FRAME_OFFSET): Redefine.
11209         (mmix_initial_elimination_offset): Refer to
11210         TARGET_STARTING_FRAME_OFFSET instead of STARTING_FRAME_OFFSET.
11211         * config/pa/pa.h (STARTING_FRAME_OFFSET): Delete.
11212         * config/pa/pa.c (pa_starting_frame_offset): New function.
11213         (pa_compute_frame_size): Use it instead of STARTING_FRAME_OFFSET.
11214         (pa_expand_prologue): Likewise.
11215         (TARGET_STARTING_FRAME_OFFSET): Redefine.
11216         * config/powerpcspe/aix.h (STARTING_FRAME_OFFSET): Split out
11217         !FRAME_GROWS_DOWNWARD handling to...
11218         (RS6000_STARTING_FRAME_OFFSET): ...this new macro.
11219         * config/powerpcspe/darwin.h (STARTING_FRAME_OFFSET): Split out
11220         !FRAME_GROWS_DOWNWARD handling to...
11221         (RS6000_STARTING_FRAME_OFFSET): ...this new macro.
11222         * config/powerpcspe/powerpcspe.h (STARTING_FRAME_OFFSET): Split out
11223         !FRAME_GROWS_DOWNWARD handling to...
11224         (RS6000_STARTING_FRAME_OFFSET): ...this new macro.
11225         * config/powerpcspe/powerpcspe.c (TARGET_STARTING_FRAME_OFFSET):
11226         Redefine.
11227         (rs6000_starting_frame_offset): New function.
11228         * config/rs6000/aix.h (STARTING_FRAME_OFFSET): Split out
11229         !FRAME_GROWS_DOWNWARD handling to...
11230         (RS6000_STARTING_FRAME_OFFSET): ...this new macro.
11231         * config/rs6000/darwin.h (STARTING_FRAME_OFFSET): Split out
11232         !FRAME_GROWS_DOWNWARD handling to...
11233         (RS6000_STARTING_FRAME_OFFSET): ...this new macro.
11234         * config/rs6000/rs6000.h (STARTING_FRAME_OFFSET): Split out
11235         !FRAME_GROWS_DOWNWARD handling to...
11236         (RS6000_STARTING_FRAME_OFFSET): ...this new macro.
11237         * config/rs6000/rs6000.c (TARGET_STARTING_FRAME_OFFSET): Refine.
11238         (rs6000_starting_frame_offset): New function.
11239         * config/vax/elf.h (STARTING_FRAME_OFFSET): Delete.
11240         * config/vax/vax.h (STARTING_FRAME_OFFSET): Delete.
11241         * config/vax/vax.c (vax_starting_frame_offset): New function.
11242         (vax_expand_prologue): Use it instead of STARTING_FRAME_OFFSET.
11243         (TARGET_STARTING_FRAME_OFFSET): Redefine.
11244         * config/xtensa/xtensa.h (STARTING_FRAME_OFFSET): Delete.
11245         * config/xtensa/xtensa.c (xtensa_starting_frame_offset): New function.
11246         (TARGET_STARTING_FRAME_OFFSET): Redefine.
11247         * system.h (STARTING_FRAME_OFFSET): Poison.
11249 2017-10-23  Richard Sandiford  <richard.sandiford@linaro.org>
11251         * tree-vect-loop.c (vect_create_epilog_for_reduction): Use
11252         SCALAR_TYPE_MODE instead of TYPE_MODE.
11254 2017-10-23  Richard Sandiford  <richard.sandiford@linaro.org>
11255             Alan Hayward  <alan.hayward@arm.com>
11256             David Sherwood  <david.sherwood@arm.com>
11258         * dwarf2out.c (loc_list_from_tree_1): Use SCALAR_INT_TYPE_MODE
11260 2017-10-23  Richard Sandiford  <richard.sandiford@linaro.org>
11261             Alan Hayward  <alan.hayward@arm.com>
11262             David Sherwood  <david.sherwood@arm.com>
11264         * expmed.c (expand_shift_1): Use scalar_mode for scalar_mode.
11266 2017-10-23  Richard Biener  <rguenther@suse.de>
11268         PR tree-optimization/82129
11269         * tree-ssa-pre.c (bitmap_set_and): Remove.
11270         (compute_antic_aux): Compute ANTIC_OUT intersection in a way
11271         canonicalizing expressions in the set to those with lowest
11272         ID rather than taking that from the first edge.
11274 2017-10-23  Richard Sandiford  <richard.sandiford@linaro.org>
11276         * combine.c (rtx_equal_for_field_assignment_p): Use
11277         byte_lowpart_offset.
11279 2017-10-22  Richard Sandiford  <richard.sandiford@linaro.org>
11280             Alan Hayward  <alan.hayward@arm.com>
11281             David Sherwood  <david.sherwood@arm.com>
11283         * internal-fn.c (expand_direct_optab_fn): Don't assign directly
11284         to a SUBREG_PROMOTED_VAR.
11286 2017-10-22  Richard Sandiford  <richard.sandiford@linaro.org>
11287             Alan Hayward  <alan.hayward@arm.com>
11288             David Sherwood  <david.sherwood@arm.com>
11290         * cfgexpand.c (expand_debug_expr): Use GET_MODE_UNIT_PRECISION.
11291         (expand_debug_source_expr): Likewise.
11292         * combine.c (combine_simplify_rtx): Likewise.
11293         * cse.c (fold_rtx): Likewise.
11294         * optabs.c (expand_float): Likewise.
11295         * simplify-rtx.c (simplify_unary_operation_1): Likewise.
11296         (simplify_binary_operation_1): Likewise.
11298 2017-10-22  Richard Sandiford  <richard.sandiford@linaro.org>
11299             Alan Hayward  <alan.hayward@arm.com>
11300             David Sherwood  <david.sherwood@arm.com>
11302         * combine.c (simplify_comparison): Use HWI_COMPUTABLE_MODE_P.
11303         (record_promoted_value): Likewise.
11304         * expr.c (expand_expr_real_2): Likewise.
11305         * ree.c (update_reg_equal_equiv_notes): Likewise.
11306         (combine_set_extension): Likewise.
11307         * rtlanal.c (low_bitmask_len): Likewise.
11308         * simplify-rtx.c (neg_const_int): Likewise.
11309         (simplify_binary_operation_1): Likewise.
11311 2017-10-22  Richard Sandiford  <richard.sandiford@linaro.org>
11312             Alan Hayward  <alan.hayward@arm.com>
11313             David Sherwood  <david.sherwood@arm.com>
11315         * lra-spills.c (assign_mem_slot): Use subreg_size_lowpart_offset.
11316         * regcprop.c (maybe_mode_change): Likewise.
11317         * reload1.c (alter_reg): Likewise.
11319 2017-10-22  Richard Sandiford  <richard.sandiford@linaro.org>
11321         * inchash.h (inchash::hash::add_wide_int): New function.
11322         * lto-streamer-out.c (hash_tree): Use it.
11324 2017-10-22  Richard Sandiford  <richard.sandiford@linaro.org>
11326         * inchash.h (inchash::hash::add_wide_int): Rename to...
11327         (inchash::hash::add_hwi): ...this.
11328         * ipa-devirt.c (hash_odr_vtable): Update accordingly.
11329         (polymorphic_call_target_hasher::hash): Likewise.
11330         * ipa-icf.c (sem_function::get_hash, sem_function::init): Likewise.
11331         (sem_item::add_expr, sem_item::add_type, sem_variable::get_hash)
11332         (sem_item_optimizer::update_hash_by_addr_refs): Likewise.
11333         * lto-streamer-out.c (hash_tree): Likewise.
11334         * optc-save-gen.awk: Likewise.
11335         * tree.c (add_expr): Likewise.
11337 2017-10-22  Uros Bizjak  <ubizjak@gmail.com>
11339         PR target/52451
11340         * config/i386/i386.c (ix86_fp_compare_mode): Return CCFPmode
11341         for ordered inequality comparisons even with TARGET_IEEE_FP.
11343 2017-10-22  Uros Bizjak  <ubizjak@gmail.com>
11345         PR target/82628
11346         * config/i386/i386.md (cmp<dwi>_doubleword): New pattern.
11347         * config/i386/i386.c (ix86_expand_branch) <case E_TImode>:
11348         Expand with cmp<dwi>_doubleword.
11350 2017-10-21  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
11352         * extend.texi: Add x86 specific to 'nocf_check' attribute.
11353         List CET intrinsics.
11354         * invoke.texi: Add -mcet, -mibt, -mshstk options.  Add x86
11355         specific to -fcf-protection option.
11357 2017-10-21  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
11359         * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET): New.
11360         (OPTION_MASK_ISA_SHSTK_SET): Likewise.
11361         (OPTION_MASK_ISA_IBT_UNSET): Likewise.
11362         (OPTION_MASK_ISA_SHSTK_UNSET): Likewise.
11363         (ix86_handle_option): Add -mibt, -mshstk, -mcet handling.
11364         * config.gcc (extra_headers): Add cetintrin.h for x86 targets.
11365         (extra_objs): Add cet.o for Linux/x86 targets.
11366         (tmake_file): Add i386/t-cet for Linux/x86 targets.
11367         * config/i386/cet.c: New file.
11368         * config/i386/cetintrin.h: Likewise.
11369         * config/i386/t-cet: Likewise.
11370         * config/i386/cpuid.h (bit_SHSTK): New.
11371         (bit_IBT): Likewise.
11372         * config/i386/driver-i386.c (host_detect_local_cpu): Detect and
11373         pass IBT and SHSTK bits.
11374         * config/i386/i386-builtin-types.def
11375         (VOID_FTYPE_UNSIGNED_PVOID): New.
11376         (VOID_FTYPE_UINT64_PVOID): Likewise.
11377         * config/i386/i386-builtin.def: Add CET intrinsics.
11378         * config/i386/i386-c.c (ix86_target_macros_internal): Add
11379         OPTION_MASK_ISA_IBT, OPTION_MASK_ISA_SHSTK handling.
11380         * config/i386/i386-passes.def: Add pass_insert_endbranch pass.
11381         * config/i386/i386-protos.h (make_pass_insert_endbranch): New
11382         prototype.
11383         * config/i386/i386.c (rest_of_insert_endbranch): New.
11384         (pass_data_insert_endbranch): Likewise.
11385         (pass_insert_endbranch): Likewise.
11386         (make_pass_insert_endbranch): Likewise.
11387         (ix86_notrack_prefixed_insn_p): Likewise.
11388         (ix86_target_string): Add -mibt, -mshstk flags.
11389         (ix86_option_override_internal): Add flag_cf_protection
11390         processing.
11391         (ix86_valid_target_attribute_inner_p): Set OPT_mibt, OPT_mshstk.
11392         (ix86_print_operand): Add 'notrack' prefix output.
11393         (ix86_init_mmx_sse_builtins): Add CET intrinsics.
11394         (ix86_expand_builtin): Expand CET intrinsics.
11395         (x86_output_mi_thunk): Add 'endbranch' instruction.
11396         * config/i386/i386.h (TARGET_IBT): New.
11397         (TARGET_IBT_P): Likewise.
11398         (TARGET_SHSTK): Likewise.
11399         (TARGET_SHSTK_P): Likewise.
11400         * config/i386/i386.md (unspecv): Add UNSPECV_NOP_RDSSP,
11401         UNSPECV_INCSSP, UNSPECV_SAVEPREVSSP, UNSPECV_RSTORSSP,
11402         UNSPECV_WRSS, UNSPECV_WRUSS, UNSPECV_SETSSBSY, UNSPECV_CLRSSBSY.
11403         (builtin_setjmp_setup): New pattern.
11404         (builtin_longjmp): Likewise.
11405         (rdssp<mode>): Likewise.
11406         (incssp<mode>): Likewise.
11407         (saveprevssp): Likewise.
11408         (rstorssp): Likewise.
11409         (wrss<mode>): Likewise.
11410         (wruss<mode>): Likewise.
11411         (setssbsy): Likewise.
11412         (clrssbsy): Likewise.
11413         (nop_endbr): Likewise.
11414         * config/i386/i386.opt: Add -mcet, -mibt, -mshstk and -mcet-switch
11415         options.
11416         * config/i386/immintrin.h: Include <cetintrin.h>.
11417         * config/i386/linux-common.h
11418         (file_end_indicate_exec_stack_and_cet): New prototype.
11419         (TARGET_ASM_FILE_END): New.
11421 2017-10-20  Jan Hubicka  <hubicka@ucw.cz>
11423         * i386.c (ix86_builtin_vectorization_cost): Use existing rtx_cost
11424         latencies instead of having separate table; make difference between
11425         integer and float costs.
11426         * i386.h (processor_costs): Remove scalar_stmt_cost,
11427         scalar_load_cost, scalar_store_cost, vec_stmt_cost, vec_to_scalar_cost,
11428         scalar_to_vec_cost, vec_align_load_cost, vec_unalign_load_cost,
11429         vec_store_cost.
11430         * x86-tune-costs.h: Remove entries which has been removed in
11431         procesor_costs from all tables; make cond_taken_branch_cost
11432         and cond_not_taken_branch_cost COST_N_INSNS based.
11434 2017-10-20  Jan Hubicka  <hubicka@ucw.cz>
11436         * x86-tune-costs.h (intel_cost, generic_cost): Fix move costs.
11438 2017-10-20  Jakub Jelinek  <jakub@redhat.com>
11440         * config/i386/i386.md (isa): Remove fma_avx512f.
11441         * config/i386/sse.md (<avx512>_fmadd_<mode>_mask<round_name>,
11442         <avx512>_fmadd_<mode>_mask3<round_name>,
11443         <avx512>_fmsub_<mode>_mask<round_name>,
11444         <avx512>_fmsub_<mode>_mask3<round_name>,
11445         <avx512>_fnmadd_<mode>_mask<round_name>,
11446         <avx512>_fnmadd_<mode>_mask3<round_name>,
11447         <avx512>_fnmsub_<mode>_mask<round_name>,
11448         <avx512>_fnmsub_<mode>_mask3<round_name>,
11449         <avx512>_fmaddsub_<mode>_mask<round_name>,
11450         <avx512>_fmaddsub_<mode>_mask3<round_name>,
11451         <avx512>_fmsubadd_<mode>_mask<round_name>,
11452         <avx512>_fmsubadd_<mode>_mask3<round_name>): Remove isa attribute.
11453         (*vec_widen_umult_even_v16si<mask_name>,
11454         *vec_widen_smult_even_v16si<mask_name>): Likewise.
11455         (<mask_codefor>avx512bw_dbpsadbw<mode><mask_name>): Likewise.
11457 2017-10-20  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
11459         * extend.texi: Add 'nocf_check' documentation.
11460         * gimple.texi: Add second parameter to
11461         gimple_build_call_from_tree.
11462         * invoke.texi: Add -fcf-protection documentation.
11463         * rtl.texi: Add REG_CALL_NOTRACK documenation.
11465 2017-10-20  Richard Biener  <rguenther@suse.de>
11467         PR tree-optimization/82473
11468         * tree-vect-loop.c (vectorizable_reduction): Properly get at
11469         the largest input type.
11471 2017-10-20  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
11473         * c-attribs.c (handle_nocf_check_attribute): New function.
11474         (c_common_attribute_table): Add 'nocf_check' handling.
11475         * gimple-parser.c: Add second argument NULL to
11476         gimple_build_call_from_tree.
11477         * attrib.c (comp_type_attributes): Check nocf_check attribute.
11478         * cfgexpand.c (expand_call_stmt): Set REG_CALL_NOCF_CHECK for
11479         call insn.
11480         * combine.c (distribute_notes): Add REG_CALL_NOCF_CHECK handling.
11481         * common.opt: Add fcf-protection flag.
11482         * emit-rtl.c (try_split): Add REG_CALL_NOCF_CHECK handling.
11483         * flag-types.h: Add enum cf_protection_level.
11484         * gimple.c (gimple_build_call_from_tree): Add second parameter.
11485         Add 'nocf_check' attribute propagation to gimple call.
11486         * gimple.h (gf_mask): Add GF_CALL_NOCF_CHECK.
11487         (gimple_build_call_from_tree): Update prototype.
11488         (gimple_call_nocf_check_p): New function.
11489         (gimple_call_set_nocf_check): Likewise.
11490         * gimplify.c: Add second argument to gimple_build_call_from_tree.
11491         * ipa-icf.c: Add nocf_check attribute in statement hash.
11492         * recog.c (peep2_attempt): Add REG_CALL_NOCF_CHECK handling.
11493         * reg-notes.def: Add REG_NOTE (CALL_NOCF_CHECK).
11494         * toplev.c (process_options): Add flag_cf_protection handling.
11496 2017-10-19  Jan Hubicka  <hubicka@ucw.cz>
11498         * x86-tune-costs.h (core_cost): Fix div, move and sqrt latencies.
11500 2017-10-20  Richard Biener  <rguenther@suse.de>
11502         PR tree-optimization/82603
11503         * tree-if-conv.c (predicate_mem_writes): Make sure to only
11504         remove false predicated stores.
11506 2017-10-20  Richard Biener  <rguenther@suse.de>
11508         * graphite-isl-ast-to-gimple.c
11509         (translate_isl_ast_to_gimple::graphite_copy_stmts_from_block):
11510         Remove return value and simplify, dump copied stmt after lhs
11511         adjustment.
11512         (translate_isl_ast_to_gimple::translate_isl_ast_node_user):
11513         Reduce dump verbosity.
11514         (gsi_insert_earliest): Likewise.
11515         (translate_isl_ast_to_gimple::copy_bb_and_scalar_dependences): Adjust.
11516         * graphite.c (print_global_statistics): Adjust dumping.
11517         (print_graphite_scop_statistics): Likewise.
11518         (print_graphite_statistics): Do not dump loops here.
11519         (graphite_transform_loops): But here.
11521 2017-10-20  Nicolas Roche  <roche@adacore.com>
11523         * configure.ac (ACX_PROG_GNAT): Append "libgnat" to include search dir.
11524         * configure: Regenerate.
11526 2017-10-20  Jakub Jelinek  <jakub@redhat.com>
11528         PR target/82158
11529         * tree-cfg.c (pass_warn_function_return::execute): In noreturn
11530         functions when optimizing replace GIMPLE_RETURN stmts with
11531         calls to __builtin_unreachable ().
11533         PR sanitizer/82595
11534         * config/gnu-user.h (LIBTSAN_EARLY_SPEC): Add libtsan_preinit.o
11535         for -fsanitize=thread link of executables.
11536         (LIBLSAN_EARLY_SPEC): Add liblsan_preinit.o for -fsanitize=leak
11537         link of executables.
11539         PR target/82370
11540         * config/i386/sse.md (VI248_AVX2, VI248_AVX512BW, VI248_AVX512BW_2):
11541         New mode iterators.
11542         (<shift_insn><mode>3<mask_name>): Change the last of the 3
11543         define_insns for logical vector shifts to use VI248_AVX512BW
11544         iterator instead of VI48_AVX512, remove <mask_mode512bit_condition>
11545         condition, useless isa and prefix attributes.  Change the first
11546         2 of these define_insns to ...
11547         (<mask_codefor><shift_insn><mode>3<mask_name>): ... this, new
11548         define_insn for avx512vl.
11549         (<shift_insn><mode>3): ... and this, new define_insn without
11550         masking for non-avx512vl.
11552         PR target/82370
11553         * config/i386/sse.md (*andnot<mode>3,
11554         <mask_codefor><code><mode>3<mask_name>, *<code><mode>3): Split
11555         (=v,v,vm) alternative into (=x,x,xm) and (=v,v,vm), for 128-bit
11556         and 256-bit vectors, the (=x,x,xm) alternative and when mask is
11557         not applied use empty suffix even for TARGET_AVX512VL.
11558         * config/i386/subst.md (mask_prefix3, mask_prefix4): When mask
11559         is applied, supply evex,evex or evex,evex,evex instead of just
11560         evex.
11562 2017-10-20  Julia Koval  <julia.koval@intel.com>
11564         * common/config/i386/i386-common.c (OPTION_MASK_ISA_GFNI_SET,
11565         (OPTION_MASK_ISA_GFNI_UNSET): New.
11566         (ix86_handle_option): Handle OPT_mgfni.
11567         * config/i386/cpuid.h (bit_GFNI): New.
11568         * config/i386/driver-i386.c (host_detect_local_cpu): Detect gfni.
11569         * config/i386/i386-c.c (ix86_target_macros_internal): Define __GFNI__.
11570         * config/i386/i386.c (ix86_target_string): Add -mgfni.
11571         (ix86_valid_target_attribute_inner_p): Add OPT_mgfni.
11572         * config/i386/i386.h (TARGET_GFNI, TARGET_GFNI_P): New.
11573         * config/i386/i386.opt: Add mgfni.
11575 2017-10-20  Orlando Arias  <oarias@knights.ucf.edu>
11577         * config/msp430/msp430.c (msp430_option_override): Disable
11578         -fdelete-null-pointer-checks.
11579         * doc/invoke.text (-fdelete-null-pointer-checks): Document that.
11581 2017-10-19  Jan Hubicka  <hubicka@ucw.cz>
11583         * x86-tune-costs.h (generic_cost, core_cost): Correct costs
11584         of x87 and SSE instructions.
11586 2017-10-19  Jan Hubicka  <hubicka@ucw.cz>
11588         * asan.c (create_cond_insert_point): Do not update edge count.
11589         * auto-profile.c (afdo_propagate_edge): Update for edge count removal.
11590         (afdo_propagate_circuit): Likewise.
11591         (afdo_calculate_branch_prob): Likewise.
11592         (afdo_annotate_cfg): Likewise.
11593         * basic-block.h (struct edge_def): Remove count.
11594         (edge_def::count): New accessor.
11595         * bb-reorder.c (rotate_loop): Update.
11596         (find_traces_1_round): Update.
11597         (connect_traces): Update.
11598         (sanitize_hot_paths): Update.
11599         * cfg.c (unchecked_make_edge): Update.
11600         (make_single_succ_edge): Update.
11601         (check_bb_profile): Update.
11602         (dump_edge_info): Update.
11603         (update_bb_profile_for_threading): Update.
11604         (scale_bbs_frequencies_int): Update.
11605         (scale_bbs_frequencies_gcov_type): Update.
11606         (scale_bbs_frequencies_profile_count): Update.
11607         (scale_bbs_frequencies): Update.
11608         * cfganal.c (connect_infinite_loops_to_exit): Update.
11609         * cfgbuild.c (compute_outgoing_frequencies): Update.
11610         (find_many_sub_basic_blocks): Update.
11611         * cfgcleanup.c (try_forward_edges): Update.
11612         (try_crossjump_to_edge): Update
11613         * cfgexpand.c (expand_gimple_cond): Update
11614         (expand_gimple_tailcall): Update
11615         (construct_exit_block): Update
11616         * cfghooks.c (verify_flow_info): Update
11617         (redirect_edge_succ_nodup): Update
11618         (split_edge): Update
11619         (make_forwarder_block): Update
11620         (duplicate_block): Update
11621         (account_profile_record): Update
11622         * cfgloop.c (find_subloop_latch_edge_by_profile): Update.
11623         * cfgloopanal.c (expected_loop_iterations_unbounded): Update.
11624         * cfgloopmanip.c (scale_loop_profile): Update.
11625         (loopify): Update.
11626         (lv_adjust_loop_entry_edge): Update.
11627         * cfgrtl.c (try_redirect_by_replacing_jump): Update.
11628         (force_nonfallthru_and_redirect): Update.
11629         (purge_dead_edges): Update.
11630         (rtl_flow_call_edges_add): Update.
11631         * cgraphunit.c (init_lowered_empty_function): Update.
11632         (cgraph_node::expand_thunk): Update.
11633         * gimple-pretty-print.c (dump_probability): Update.
11634         (dump_edge_probability): Update.
11635         * gimple-ssa-isolate-paths.c (isolate_path): Update.
11636         * haifa-sched.c (sched_create_recovery_edges): Update.
11637         * hsa-gen.c (convert_switch_statements): Update.
11638         * ifcvt.c (dead_or_predicable): Update.
11639         * ipa-inline-transform.c (inline_transform): Update.
11640         * ipa-split.c (split_function): Update.
11641         * ipa-utils.c (ipa_merge_profiles): Update.
11642         * loop-doloop.c (add_test): Update.
11643         * loop-unroll.c (unroll_loop_runtime_iterations): Update.
11644         * lto-streamer-in.c (input_cfg): Update.
11645         (input_function): Update.
11646         * lto-streamer-out.c (output_cfg): Update.
11647         * modulo-sched.c (sms_schedule): Update.
11648         * postreload-gcse.c (eliminate_partially_redundant_load): Update.
11649         * predict.c (maybe_hot_edge_p): Update.
11650         (unlikely_executed_edge_p): Update.
11651         (probably_never_executed_edge_p): Update.
11652         (dump_prediction): Update.
11653         (drop_profile): Update.
11654         (propagate_unlikely_bbs_forward): Update.
11655         (determine_unlikely_bbs): Update.
11656         (force_edge_cold): Update.
11657         * profile.c (compute_branch_probabilities): Update.
11658         * reg-stack.c (better_edge): Update.
11659         * shrink-wrap.c (handle_simple_exit): Update.
11660         * tracer.c (better_p): Update.
11661         * trans-mem.c (expand_transaction): Update.
11662         (split_bb_make_tm_edge): Update.
11663         * tree-call-cdce.c: Update.
11664         * tree-cfg.c (gimple_find_sub_bbs): Update.
11665         (gimple_split_edge): Update.
11666         (gimple_duplicate_sese_region): Update.
11667         (gimple_duplicate_sese_tail): Update.
11668         (gimple_flow_call_edges_add): Update.
11669         (insert_cond_bb): Update.
11670         (execute_fixup_cfg): Update.
11671         * tree-cfgcleanup.c (cleanup_control_expr_graph): Update.
11672         * tree-complex.c (expand_complex_div_wide): Update.
11673         * tree-eh.c (lower_resx): Update.
11674         (unsplit_eh): Update.
11675         (cleanup_empty_eh_move_lp): Update.
11676         * tree-inline.c (copy_edges_for_bb): Update.
11677         (freqs_to_counts): Update.
11678         (copy_cfg_body): Update.
11679         * tree-ssa-dce.c (remove_dead_stmt): Update.
11680         * tree-ssa-ifcombine.c (update_profile_after_ifcombine): Update.
11681         * tree-ssa-loop-im.c (execute_sm_if_changed): Update.
11682         * tree-ssa-loop-ivcanon.c (remove_exits_and_undefined_stmts): Update.
11683         (unloop_loops): Update.
11684         * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Update.
11685         * tree-ssa-loop-split.c (connect_loops): Update.
11686         (split_loop): Update.
11687         * tree-ssa-loop-unswitch.c (hoist_guard): Update.
11688         * tree-ssa-phionlycprop.c (propagate_rhs_into_lhs): Update.
11689         * tree-ssa-phiopt.c (replace_phi_edge_with_variable): Update.
11690         * tree-ssa-reassoc.c (branch_fixup): Update.
11691         * tree-ssa-tail-merge.c (replace_block_by): Update.
11692         * tree-ssa-threadupdate.c (remove_ctrl_stmt_and_useless_edges): Update.
11693         (compute_path_counts): Update.
11694         (update_profile): Update.
11695         (recompute_probabilities): Update.
11696         (update_joiner_offpath_counts): Update.
11697         (estimated_freqs_path): Update.
11698         (freqs_to_counts_path): Update.
11699         (clear_counts_path): Update.
11700         (ssa_fix_duplicate_block_edges): Update.
11701         (duplicate_thread_path): Update.
11702         * tree-switch-conversion.c (hoist_edge_and_branch_if_true): Update.
11703         (case_bit_test_cmp): Update.
11704         (collect_switch_conv_info): Update.
11705         (gen_inbound_check): Update.
11706         (do_jump_if_equal): Update.
11707         (emit_cmp_and_jump_insns): Update.
11708         * tree-tailcall.c (decrease_profile): Update.
11709         (eliminate_tail_call): Update.
11710         * tree-vect-loop-manip.c (slpeel_add_loop_guard): Update.
11711         (vect_do_peeling): Update.
11712         * tree-vect-loop.c (scale_profile_for_vect_loop): Update.
11713         * ubsan.c (ubsan_expand_null_ifn): Update.
11714         (ubsan_expand_ptr_ifn): Update.
11715         * value-prof.c (gimple_divmod_fixed_value): Update.
11716         (gimple_mod_pow2): Update.
11717         (gimple_mod_subtract): Update.
11718         (gimple_ic): Update.
11719         (gimple_stringop_fixed_value): Update.
11721 2017-10-19  Uros Bizjak  <ubizjak@gmail.com>
11723         PR target/82618
11724         * config/i386/i386.md (sub to cmp): New peephole2 pattern.
11726 2017-10-19  Alexander Monakov  <amonakov@ispras.ru>
11728         PR rtl-optimization/82395
11729         * ira-color.c (allocno_priority_compare_func): Fix comparison step
11730         based on non_spilled_static_chain_regno_p.
11732 2017-10-19  Uros Bizjak  <ubizjak@gmail.com>
11734         * config/i386/i386.c (output_387_binary_op): Rewrite SSE part.
11735         (ix86_emit_mode_set): Rewrite insn mnemonic construction.
11736         (ix86_prepare_fp_compare_args): Redefine is_sse as bool.
11738 2017-10-19  Martin Sebor  <msebor@redhat.com>
11740         PR tree-optimization/82596
11741         * tree.c (array_at_struct_end_p): Handle STRING_CST.
11743 2017-10-19  Eric Botcazou  <ebotcazou@adacore.com>
11745         * asan.c (handle_builtin_alloca): Deal with all alloca variants.
11746         (get_mem_refs_of_builtin_call): Likewise.
11747         * builtins.c (expand_builtin_apply): Adjust call to
11748         allocate_dynamic_stack_space.
11749         (expand_builtin_alloca): For __builtin_alloca_with_align_and_max, pass
11750         the third argument to allocate_dynamic_stack_space, otherwise -1.
11751         (expand_builtin): Deal with all alloca variants.
11752         (is_inexpensive_builtin): Likewise.
11753         * builtins.def (BUILT_IN_ALLOCA_WITH_ALIGN_AND_MAX): New.
11754         * calls.c (special_function_p): Deal with all alloca variants.
11755         (initialize_argument_information): Adjust call to
11756         allocate_dynamic_stack_space.
11757         (expand_call): Likewise.
11758         * cfgexpand.c (expand_call_stmt): Deal with all alloca variants.
11759         * doc/extend.texi (Built-ins): Add __builtin_alloca_with_align_and_max
11760         * explow.c (allocate_dynamic_stack_space): Add MAX_SIZE parameter and
11761         use it for the stack usage computation.
11762         * explow.h (allocate_dynamic_stack_space): Adjust prototype.
11763         * function.c (gimplify_parameters): Call build_alloca_call_expr.
11764         * gimple-ssa-warn-alloca.c (alloca_call_type): Simplify control flow.
11765         Take into account 3rd argument of __builtin_alloca_with_align_and_max.
11766         (in_loop_p): Remove first argument and useless check.
11767         (pass_walloca::execute): Remove useless test and adjust call to above.
11768         * gimple.c (gimple_build_call_from_tree): Deal with all alloc variants
11769         * gimplify.c (gimplify_vla_decl): Call build_alloca_call_expr.
11770         (gimplify_call_expr): Deal with all alloca variants.
11771         * hsa-gen.c (gen_hsa_alloca): Likewise.
11772         (gen_hsa_insns_for_call): Likewise.
11773         * ipa-pure-const.c (special_builtin_state): Likewise.
11774         * tree-chkp.c (chkp_build_returned_bound): Likewise.
11775         * tree-object-size.c (alloc_object_size): Likewise.
11776         * tree-ssa-alias.c (ref_maybe_used_by_call_p_1): Likewise.
11777         (call_may_clobber_ref_p_1): Likewise.
11778         * tree-ssa-ccp.c (evaluate_stmt): Likewise.
11779         (ccp_fold_stmt): Likewise.
11780         (optimize_stack_restore): Likewise.
11781         * tree-ssa-dce.c (mark_stmt_if_obviously_necessary): Likewise.
11782         (mark_all_reaching_defs_necessary_1): Likewise.
11783         (propagate_necessity): Likewise.
11784         (eliminate_unnecessary_stmts): Likewise.
11785         * tree.c (build_common_builtin_nodes): Build
11786         BUILT_IN_ALLOCA_WITH_ALIGN_AND_MAX.
11787         (build_alloca_call_expr): New function.
11788         * tree.h (ALLOCA_FUNCTION_CODE_P): New macro.
11789         (CASE_BUILT_IN_ALLOCA): Likewise.
11790         (build_alloca_call_expr): Declare.
11791         * varasm.c (incorporeal_function_p): Deal with all alloca variants.
11793 2017-10-19  Eric Botcazou  <ebotcazou@adacore.com>
11795         PR debug/82509
11796         * dwarf2out.c (new_die_raw): New static inline function.
11797         (new_die): Use it to create the DIE.
11798         (add_AT_external_die_ref): Likewise.
11799         (clone_die): Likewise.
11800         (clone_as_declaration): Likewise.
11801         (dwarf2out_vms_debug_main_pointer): Likewise.
11802         (base_type_die): Likewise.  Remove early return for corner cases.
11803         Do not call add_pubtype on the DIE here.
11804         (is_base_type): Remove ERROR_MARK and return 0 for VOID_TYPE.
11805         (modified_type_die): Adjust the lookup for reverse order DIEs.  Skip
11806         typedefs for base types with DW_AT_endianity.  Make sure a DIE with
11807         native order exists for base types, attach the DIE manually and call
11808         add_pubtype on it.  Do not equate a reverse order DIE to the type.
11810 2017-10-19  Richard Earnshaw  <rearnsha@arm.com>
11812         * config/arm/arm.c (align_ok_ldrd_strd): New function.
11813         (mem_ok_for_ldrd_strd): New parameter align.  Extract the alignment of
11814         the mem into it.
11815         (gen_operands_ldrd_strd): Validate the alignment of the accesses.
11817 2017-10-19  Jakub Jelinek  <jakub@redhat.com>
11819         * flag-types.h (enum sanitize_code): Add SANITIZE_BUILTIN.  Or
11820         SANITIZE_BUILTIN into SANITIZE_UNDEFINED.
11821         * sanitizer.def (BUILT_IN_UBSAN_HANDLE_INVALID_BUILTIN,
11822         BUILT_IN_UBSAN_HANDLE_INVALID_BUILTIN_ABORT): New builtins.
11823         * opts.c (sanitizer_opts): Add builtin.
11824         * ubsan.c (instrument_builtin): New function.
11825         (pass_ubsan::execute): Call it.
11826         (pass_ubsan::gate): Enable even for SANITIZE_BUILTIN.
11827         * doc/invoke.texi: Document -fsanitize=builtin.
11829         * ubsan.c (ubsan_expand_null_ifn): Use _v1 suffixed type mismatch
11830         builtins, store max (log2 (align), 0) into uchar field instead of
11831         align into uptr field.
11832         (ubsan_expand_objsize_ifn): Use _v1 suffixed type mismatch builtins,
11833         store uchar 0 field instead of uptr 0 field.
11834         (instrument_nonnull_return): Use _v1 suffixed nonnull return builtin,
11835         instead of passing one address of struct with 2 locations pass
11836         two addresses of structs with 1 location each.
11837         * sanitizer.def (BUILT_IN_UBSAN_HANDLE_TYPE_MISMATCH,
11838         BUILT_IN_UBSAN_HANDLE_TYPE_MISMATCH_ABORT,
11839         BUILT_IN_UBSAN_HANDLE_NONNULL_RETURN,
11840         BUILT_IN_UBSAN_HANDLE_NONNULL_RETURN_ABORT): Removed.
11841         (BUILT_IN_UBSAN_HANDLE_TYPE_MISMATCH_V1,
11842         BUILT_IN_UBSAN_HANDLE_TYPE_MISMATCH_V1_ABORT,
11843         BUILT_IN_UBSAN_HANDLE_NONNULL_RETURN_V1,
11844         BUILT_IN_UBSAN_HANDLE_NONNULL_RETURN_V1_ABORT): New builtins.
11846 2017-10-19  Martin Liska  <mliska@suse.cz>
11848         PR driver/81829
11849         * file-find.c (remove_prefix): Remove.
11850         * file-find.h (remove_prefix): Likewise.
11851         * gcc-ar.c: Remove smartness of lookup.
11853 2017-10-19  Segher Boessenkool  <segher@kernel.crashing.org>
11855         * config/rs6000/rs6000.md (*call_indirect_aix<mode>,
11856         *call_value_indirect_aix<mode>, *call_indirect_elfv2<mode>,
11857         *call_value_indirect_elfv2<mode>): Add correct mode to the unspec.
11859 2017-10-19  Jakub Jelinek  <jakub@redhat.com>
11861         PR target/82580
11862         * config/i386/i386.md (setcc + movzbl to xor + setcc): New peephole2.
11863         (setcc + and to xor + setcc): New peephole2.
11865 2017-10-19  Tom de Vries  <tom@codesourcery.com>
11867         * doc/sourcebuild.texi (Test Directives, Variants of
11868         dg-require-support): Add dg-require-stack-size.
11870 2017-10-19  Martin Liska  <mliska@suse.cz>
11872         PR sanitizer/82517
11873         * gimplify.c (gimplify_decl_expr): Do not instrument variables
11874         that have a large alignment.
11875         (gimplify_target_expr): Likewise.
11877 2017-10-18  Segher Boessenkool  <segher@kernel.crashing.org>
11879         PR rtl-optimization/82602
11880         * ira.c (rtx_moveable_p): Return false for volatile asm.
11882 2017-10-18  Uros Bizjak  <ubizjak@gmail.com>
11884         PR target/82580
11885         * config/i386/i386-modes.def (CCGZ): New CC mode.
11886         * config/i386/i386.md (sub<mode>3_carry_ccgz): New insn pattern.
11887         * config/i386/predicates.md (ix86_comparison_operator):
11888         Handle CCGZmode.
11889         * config/i386/i386.c (ix86_expand_branch) <case E_TImode>:
11890         Emulate LE, LEU, GT, GTU, LT, LTU, GE and GEU double-word comparisons
11891         with double-word subtraction.
11892         (put_condition_code): Handle CCGZmode.
11894 2017-10-18  Aldy Hernandez  <aldyh@redhat.com>
11896         * wide-int.cc (debug (const wide_int &)): New.
11897         (debug (const wide_int *)): New.
11898         (debug (const widest_int &)): New.
11899         (debug (const widest_int *)): New.
11901 2017-10-18  Vladimir Makarov  <vmakarov@redhat.com>
11903         PR middle-end/82556
11904         * lra-constraints.c (curr_insn_transform): Use non-input operand
11905         instead of output one for matched reload.
11907 2017-10-18  Bin Cheng  <bin.cheng@arm.com>
11909         * tree-loop-distribution.c (INCLUDE_ALGORITHM): New header file.
11910         (tree-ssa-loop-ivopts.h): New header file.
11911         (struct builtin_info): New fields.
11912         (classify_builtin_1): Compute and record base and offset parts for
11913         memset builtin partition by calling strip_offset.
11914         (offset_cmp, fuse_memset_builtins): New functions.
11915         (finalize_partitions): Fuse adjacent memset partitions by calling
11916         above function.
11917         * tree-ssa-loop-ivopts.c (strip_offset): Delete static declaration.
11918         Expose the interface.
11919         * tree-ssa-loop-ivopts.h (strip_offset): New declaration.
11921 2017-10-18  Bin Cheng  <bin.cheng@arm.com>
11923         PR tree-optimization/82574
11924         * tree-loop-distribution.c (find_single_drs): New parameter.  Check
11925         that data reference must be executed exactly once per iteration
11926         against the outermost loop in nest.
11927         (classify_partition): Update call to above function.
11929 2017-10-18  Richard Biener  <rguenther@suse.de>
11931         PR tree-optimization/82591
11932         * graphite.c (graphite_transform_loops): Move code gen message
11933         printing ...
11934         * graphite-isl-ast-to-gimple.c (graphite_regenerate_ast_isl):
11935         Here.  Handle scop_to_isl_ast failing.
11936         (scop_to_isl_ast): Limit the number of ISL operations.
11938 2017-10-18  Richard Biener  <rguenther@suse.de>
11940         * graphite-isl-ast-to-gimple.c
11941         (translate_isl_ast_to_gimple::set_rename): Simplify.
11942         (translate_isl_ast_to_gimple::set_rename_for_each_def): Inline...
11943         (graphite_copy_stmts_from_block): ... here.
11944         (copy_bb_and_scalar_dependences): Simplify.
11945         (add_parameters_to_ivs_params): Canonicalize.
11946         (generate_entry_out_of_ssa_copies): Simplify.
11947         * graphite-sese-to-poly.c (extract_affine_name): Simplify
11948         by passing in ISL dimension.
11949         (parameter_index_in_region_1): Rename to ...
11950         (parameter_index_in_region): ... this.
11951         (extract_affine): Adjust assert, pass down parameter index.
11952         (add_param_constraints): Use range-info when available.
11953         (build_scop_context): Adjust.
11954         * sese.c (new_sese_info): Adjust.
11955         (free_sese_info): Likewise.
11956         * sese.h (bb_map_t, rename_map_t, phi_rename, init_back_edge_pair_t):
11957         Remove unused typedefs.
11958         (struct sese_info_t): Simplify rename_map, remove incomplete_phis.
11960 2017-10-18  Martin Liska  <mliska@suse.cz>
11962         * combine.c (simplify_compare_const): Add gcc_fallthrough.
11964 2017-10-18  Robin Dapp  <rdapp@linux.vnet.ibm.com>
11966         * config/s390/s390.c (s390_bb_fallthru_entry_likely): New function.
11967         (s390_sched_init): Do not reset s390_sched_state if we entered the
11968         current basic block via a fallthru edge and all others are unlikely.
11970 2017-10-18  Robin Dapp  <rdapp@linux.vnet.ibm.com>
11972         * config/s390/s390.c (NUM_SIDES): New variable.
11973         (LONGRUNNING_THRESHOLD): New variable.
11974         (LATENCY_FACTOR): New variable.
11975         (s390_sched_score): Decrease score for long-running instructions on
11976         wrong side.
11977         (s390_sched_variable_issue): Perform bookkeeping for long-running
11978         instructions.
11980 2017-10-18  Richard Biener  <rguenther@suse.de>
11982         * graphite-isl-ast-to-gimple.c (gcc_expression_from_isl_ast_expr_id):
11983         Simplify with removal of the parameter rename map.
11984         (set_rename): Likewise.
11985         (should_copy_to_new_region): Likewise.
11986         (graphite_copy_stmts_from_block): Likewise.
11987         (copy_bb_and_scalar_dependences): Remove initialization of
11988         unused copied_bb_map.
11989         (copy_def): Remove.
11990         (copy_internal_parameters): Likewise.
11991         (graphite_regenerate_ast_isl): Do not call copy_internal_parameters.
11992         * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
11993         Use INTEGRAL_TYPE_P.
11994         (parameter_index_in_region_1): Rename to ...
11995         (assign_parameter_index_in_region): ... this.  Assert we have
11996         a parameter we handle.
11997         (scan_tree_for_params): Adjust.
11998         * sese.h (parameter_rename_map_t): Remove.
11999         (struct sese_info_t): Remove unused parameter_rename_map and
12000         copied_bb_map members.
12001         * sese.c (new_sese_info): Adjust.
12002         (free_sese_info): Likewise.
12004 2017-10-18  Martin Liska  <mliska@suse.cz>
12006         PR sanitizer/82545
12007         * asan.c (asan_expand_poison_ifn): Do not put gimple stmt
12008         on an abnormal edge.
12010 2017-10-18  Sebastian Huber  <sebastian.huber@embedded-brains.de>
12012         * doc/invoke.texi (ffunction-sections and fdata-sections):
12013         Update.
12015 2017-10-17  Eric Botcazou  <ebotcazou@adacore.com>
12017         * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Bail out only if
12018         the use statement can throw internally.
12020 2017-10-17  Eric Botcazou  <ebotcazou@adacore.com>
12022         * config/visium/visium.c (visium_select_cc_mode): Return CCmode for
12023         any RTX present on the RHS of a SET.
12024         * compare-elim.c (try_eliminate_compare): Restore comment.
12026 2017-10-17  Jakub Jelinek  <jakub@redhat.com>
12028         * langhooks.h (struct lang_hooks): Document that tree_size langhook
12029         may be also called on tcc_type nodes.
12030         * langhooks.c (lhd_tree_size): Likewise.
12032 2017-10-17  David Malcolm  <dmalcolm@redhat.com>
12034         * gimple-ssa-sprintf.c (fmtwarn): Update for changed signature of
12035         format_warning_at_substring.
12036         (maybe_warn): Convert source_range * param to a location_t.  Pass
12037         UNKNOWN_LOCATION rather than NULL to fmtwarn.
12038         (format_directive): Remove code to extract source_ranges and
12039         source_range * in favor of just a location_t.
12040         (parse_directive): Pass UNKNOWN_LOCATION rather than NULL to
12041         fmtwarn.
12042         * substring-locations.c (format_warning_va): Convert
12043         source_range * param to a location_t.
12044         (format_warning_at_substring): Likewise.
12045         * substring-locations.h (format_warning_va): Likewise.
12046         (format_warning_at_substring): Likewise.
12048 2017-10-17  Jan Hubicka  <hubicka@ucw.cz>
12050         * target.h (enum vect_cost_for_stmt): Add vec_gather_load and
12051         vec_scatter_store
12052         * tree-vect-stmts.c (record_stmt_cost): Make difference between normal
12053         and scatter/gather ops.
12055         * aarch64/aarch64.c (aarch64_builtin_vectorization_cost): Add
12056         vec_gather_load and vec_scatter_store.
12057         * arm/arm.c (arm_builtin_vectorization_cost): Likewise.
12058         * powerpcspe/powerpcspe.c (rs6000_builtin_vectorization_cost): Likewise.
12059         * rs6000/rs6000.c (rs6000_builtin_vectorization_cost): Likewise.
12060         * s390/s390.c (s390_builtin_vectorization_cost): Likewise.
12061         * spu/spu.c (spu_builtin_vectorization_cost): Likewise.
12062         * i386/i386.c (x86_builtin_vectorization_cost): Likewise.
12064 2017-10-17  Uros Bizjak  <ubizjak@gmail.com>
12066         * reg-stack.c (compare_for_stack_reg): Add bool argument.
12067         Detect FTST instruction and handle its register pops.  Only pop
12068         second operand if can_pop_second_op is true.
12069         (subst_stack_regs_pat) <case COMPARE>: Detect FCOMI instruction to
12070         set can_pop_second_op to false in the compare_for_stack_reg call.
12072         * config/i386/i386.md (*cmpi<FPCMP:unord><MODEF:mode>): Only call
12073         output_fp_compare for stack register operands.
12074         * config/i386/i386.c (output_fp_compare): Do not output SSE compare
12075         instructions here.  Do not emit stack register pops here.  Assert
12076         that FCOMPP pops next to top stack register.  Rewrite function.
12078 2017-10-17  Nathan Sidwell  <nathan@acm.org>
12080         PR middle-end/82577
12081         * alias.c (compare_base_decls): Check HAS_DECL_ASSEMBLER_NAME_P,
12082         use DECL_ASSEMBLER_NAME_RAW.
12084         PR middle-end/82546
12085         * tree.c (tree_code_size): Reformat.  Punt to lang hook for unknown
12086         TYPE nodes.
12088 2017-10-17  Qing Zhao <qing.zhao@oracle.com>
12089             Wilco Dijkstra <wilco.dijkstra@arm.com>
12091         * builtins.c (expand_builtin_update_setjmp_buf): Add a
12092         converstion to Pmode from the buf_addr.
12094 2017-10-17  Richard Biener  <rguenther@suse.de>
12096         * graphite-dependences.c (scop_get_reads_and_writes): Change
12097         output parameters to references.
12099 2017-10-17  Jackson Woodruff  <jackson.woodruff@arm.com>
12101         PR 71026/tree-optimization
12102         * fold-const.c (distribute_real_division): Removed.
12103         (fold_binary_loc): Remove calls to distribute_real_divison.
12105 2017-10-17  Richard Biener  <rguenther@suse.de>
12107         * graphite-scop-detection.c
12108         (scop_detection::stmt_has_simple_data_refs_p): Always use
12109         the full nest as region.
12110         (try_generate_gimple_bb): Likewise.
12111         * sese.c (scalar_evolution_in_region): Simplify now that
12112         SCEV can handle instantiation in regions.
12113         * tree-scalar-evolution.c (instantiate_scev_name): Also instantiate
12114         in the non-loop part of a function if requested.
12116 2017-10-17  Richard Biener  <rguenther@suse.de>
12118         PR tree-optimization/82563
12119         * graphite-isl-ast-to-gimple.c (generate_entry_out_of_ssa_copies):
12120         New function.
12121         (graphite_regenerate_ast_isl): Call it.
12122         * graphite-scop-detection.c (build_scops): Remove entry edge split.
12124 2017-10-17  Jakub Jelinek  <jakub@redhat.com>
12126         PR tree-optimization/82549
12127         * fold-const.c (optimize_bit_field_compare, fold_truth_andor_1):
12128         Formatting fixes.  Instead of calling make_bit_field_ref with negative
12129         bitpos return 0.
12131 2017-10-17  Olga Makhotina  <olga.makhotina@intel.com>
12133         * config/i386/avx512dqintrin.h (_mm_mask_reduce_sd,
12134         _mm_maskz_reduce_sd, _mm_mask_reduce_ss,=20
12135         _mm_maskz_reduce_ss): New.
12136         * config/i386/i386-builtin.def (__builtin_ia32_reducesd_mask,
12137         __builtin_ia32_reducess_mask): Ditto..
12138         (__builtin_ia32_reducesd, __builtin_ia32_reducess): Remove.
12139         * config/i386/sse.md (reduces<mode>): Renamed to ...
12140         (reduces<mode><mask_scalar_name>): ... this.
12141         (vreduce<ssescalarmodesuffix>\t{%3, %2, %1, %0|%0, %1, %2, %3}):
12142         Changed to ...
12143         (vreduce<ssescalarmodesuffix>\t{%3, %2, %1, %0<mask_scalar_operand4>|
12144         %0<mask_scalar_operand4>, %1, %2, %3}): ... this.
12146 2017-10-16  David Malcolm  <dmalcolm@redhat.com>
12148         * Makefile.in (OBJS): Add unique-ptr-tests.o.
12149         * selftest-run-tests.c (selftest::run_tests): Call
12150         selftest::unique_ptr_tests_cc_tests.
12151         * selftest.h (selftest::unique_ptr_tests_cc_tests): New decl.
12152         * unique-ptr-tests.cc: New file.
12154 2017-10-16  Vladimir Makarov  <vmakarov@redhat.com>
12156         PR sanitizer/82353
12157         * lra.c (collect_non_operand_hard_regs): Don't ignore operator
12158         locations.
12159         * lra-lives.c (bb_killed_pseudos, bb_gen_pseudos): Move up.
12160         (make_hard_regno_born, make_hard_regno_dead): Update
12161         bb_killed_pseudos and bb_gen_pseudos for fixed regs.
12163 2017-10-16  Jeff Law  <law@redhat.com>
12165         * tree-ssa-dse.c (live_bytes_read): Fix thinko.
12167 2017-10-16  Jan Hubicka  <hubicka@ucw.cz>
12169         * x86-tune-costs.h (znver1_cost): Fix move cost tables.
12171 2017-10-16  Olivier Hainque  <hainque@adacore.com>
12173         * gcc/config.gcc (powerpc*-*-*spe*): Pick 8548 as the default
12174         with_cpu if we were configured for an e500v2 target cpu name.
12176 2017-10-16  Thomas Preud'homme  <thomas.preudhomme@arm.com>
12178         * config/arm/arm-cpus.in (cortex-m33): Add nodsp option.
12179         * doc/invoke.texi: Document +nodsp as a valid extension for
12180         -mcpu=cortex-m33.
12182 2017-10-16  Martin Liska  <mliska@suse.cz>
12184         * sbitmap.c (bitmap_bit_in_range_p_checking): New function.
12185         (test_set_range): Likewise.
12186         (test_range_functions): Rename to ...
12187         (test_bit_in_range): ... this.
12188         (sbitmap_c_tests): Add new test.
12190 2017-10-16  Tamar Christina  <tamar.christina@arm.com>
12192         * config/aarch64/arm_neon.h (vdot_u32, vdotq_u32, vdot_s32, vdotq_s32):
12193         New.
12194         (vdot_lane_u32, vdot_laneq_u32, vdotq_lane_u32, vdotq_laneq_u32): New.
12195         (vdot_lane_s32, vdot_laneq_s32, vdotq_lane_s32, vdotq_laneq_s32): New.
12197 2017-10-16  Tamar Christina  <tamar.christina@arm.com>
12199         * config/aarch64/aarch64-builtins.c
12200         (aarch64_types_quadopu_lane_qualifiers): New.
12201         (TYPES_QUADOPU_LANE): New.
12202         * config/aarch64/aarch64-simd.md (aarch64_<sur>dot<vsi2qi>): New.
12203         (<sur>dot_prod<vsi2qi>, aarch64_<sur>dot_lane<vsi2qi>): New.
12204         (aarch64_<sur>dot_laneq<vsi2qi>): New.
12205         * config/aarch64/aarch64-simd-builtins.def (sdot, udot): New.
12206         (sdot_lane, udot_lane, sdot_laneq, udot_laneq): New.
12207         * config/aarch64/iterators.md (sur): Add UNSPEC_SDOT, UNSPEC_UDOT.
12208         (Vdottype, DOTPROD): New.
12209         (sur): Add SDOT and UDOT.
12211 2017-10-16  Tamar Christina  <tamar.christina@arm.com>
12213         * config/aarch64/aarch64.h (AARCH64_FL_DOTPROD): New.
12214         (AARCH64_ISA_DOTPROD, TARGET_DOTPROD): New.
12215         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
12216         Add TARGET_DOTPROD.
12217         * config/aarch64/aarch64-option-extensions.def (dotprod): New.
12218         * config/aarch64/aarch64-cores.def (cortex-a55, cortex-a75):
12219         Enable TARGET_DOTPROD.
12220         (cortex-a75.cortex-a55): Likewise.
12221         * doc/invoke.texi (aarch64-feature-modifiers): Document dotprod.
12223 2017-10-16  Tamar Christina  <tamar.christina@arm.com>
12225         * config/arm/arm-builtins.c (arm_unsigned_uternop_qualifiers): New.
12226         (UTERNOP_QUALIFIERS, arm_umac_lane_qualifiers, UMAC_LANE_QUALIFIERS):
12227         New.
12228         * config/arm/arm_neon_builtins.def (sdot, udot, sdot_lane, udot_lane):
12229         New.
12230         * config/arm/iterators.md (DOTPROD, VSI2QI, vsi2qi): New.
12231         (UNSPEC_DOT_S, UNSPEC_DOT_U, opsuffix): New.
12232         * config/arm/neon.md (neon_<sup>dot<vsi2qi>): New.
12233         (neon_<sup>dot_lane<vsi2qi>, <sup>dot_prod<vsi2qi>): New.
12234         * config/arm/types.md (neon_dot, neon_dot_q): New.
12235         * config/arm/unspecs.md (sup): Add UNSPEC_DOT_S, UNSPEC_DOT_U.
12237 2017-10-16  Tamar Christina  <tamar.christina@arm.com>
12239         * config/arm/arm.h (TARGET_DOTPROD): New.
12240         * config/arm/arm.c (arm_arch_dotprod): New.
12241         (arm_option_reconfigure_globals): Add arm_arch_dotprod.
12242         * config/arm/arm-c.c (__ARM_FEATURE_DOTPROD): New.
12243         * config/arm/arm-cpus.in (armv8.2-a): Enabled +dotprod.
12244         (feature dotprod, group dotprod, ALL_SIMD_INTERNAL): New.
12245         (ALL_FPU_INTERNAL): Use ALL_SIMD_INTERNAL.
12246         * config/arm/t-multilib (v8_2_a_simd_variants): Add dotprod.
12247         * doc/invoke.texi (armv8.2-a): Document dotprod
12249 2017-10-14  Jan Hubicka  <hubicka@ucw.cz>
12251         * i386.c (ix86_vec_cost): New function.
12252         (ix86_rtx_costs): Handle vector operations better.
12253         * i386.h (struct processor_costs): Add sse_op, fmasd, fmass.
12254         * x86-tune-costs.h: Add new costs to all tables.
12256 2017-10-14  Jan Hubicka  <hubicka@ucw.cz>
12258         * i386.c (ix86_rtx_costs): Make difference between x87 and SSE
12259         operations.
12260         * i386.h (struct processor_costs): Add addss, mulss, mulsd, divss,
12261         divsd, sqrtss and sqrtsd
12262         * x86-tune-costs.h: Add new entries to all costs.
12263         (znver1_cost): Fix to match real instruction latencies.
12265 2017-10-14  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
12266             Michael Collison <michael.collison@arm.com>
12268         * compare-elim.c: Include emit-rtl.h.
12269         (can_merge_compare_into_arith): New function.
12270         (try_validate_parallel): Likewise.
12271         (try_merge_compare): Likewise.
12272         (try_eliminate_compare): Call the above when no previous clobber
12273         is available.
12274         (execute_compare_elim_after_reload): Add DF_UD_CHAIN and DF_DU_CHAIN
12275         dataflow problems.
12277 2017-10-14  Jakub Jelinek  <jakub@redhat.com>
12279         PR middle-end/62263
12280         PR middle-end/82498
12281         * tree-ssa-phiopt.c (value_replacement): Comment fix.  Handle
12282         up to 2 preparation statements for ASSIGN in MIDDLE_BB.
12284         PR middle-end/62263
12285         PR middle-end/82498
12286         * tree-ssa-forwprop.c (simplify_rotate): Allow def_arg1[N]
12287         to be any operand_equal_p operands.  For & (B - 1) require
12288         B to be power of 2.  Recognize
12289         (X << (Y & (B - 1))) | (X >> ((-Y) & (B - 1))) and similar patterns.
12291 2017-10-14  Uros Bizjak  <ubizjak@gmail.com>
12293         PR bootstrap/82553
12294         * optabs.c (expand_memory_blockage): Fix call of
12295         targetm.have_memory_blockage.
12297 2017-10-14  Jakub Jelinek  <jakub@redhat.com>
12299         PR bootstrap/82548
12300         * config.gcc (*-*-solaris2*, i[34567]86-*-cygwin*,
12301         x86_64-*-cygwin*, i[34567]86-*-mingw* | x86_64-*-mingw*): Append
12302         objects to extra_objs instead of overwriting it.
12304 2017-10-14  Uros Bizjak  <ubizjak@gmail.com>
12306         * config/i386/sync.md (FILD_ATOMIC/FIST_ATOMIC FP load peephole2):
12307         Use any_fp_register_operand as operand[3] predicate.  Simplify
12308         equality test for operands[2] and operands[4] memory location.
12309         (LDX_ATOMIC/STX_ATOMIC FP load peephole2): Ditto.
12310         (FILD_ATOMIC/FIST_ATOMIC FP load peephole2 with mem blockage): New.
12311         (LDX_ATOMIC/LDX_ATOMIC FP load peephole2 with mem blockage): Ditto.
12312         (FILD_ATOMIC/FIST_ATOMIC FP store peephole2): Use
12313         any_fp_register_operand as operand[1] predicate.  Simplify
12314         equality test for operands[0] and operands[3] memory location.
12315         (LDX_ATOMIC/STX_ATOMIC FP store peephole2): Ditto.
12316         (FILD_ATOMIC/FIST_ATOMIC FP store peephole2 with mem blockage): New.
12317         (LDX_ATOMIC/LDX_ATOMIC FP storepeephole2 with mem blockage): Ditto.
12319 2017-10-14  Uros Bizjak  <ubizjak@gmail.com>
12321         * target-insns.def: Add memory_blockage.
12322         * optabs.c (expand_memory_blockage): New function.
12323         (expand_asm_memory_barrier): Rename ...
12324         (expand_asm_memory_blockage): ... to this.
12325         (expand_mem_thread_fence): Call expand_memory_blockage
12326         instead of expand_asm_memory_barrier.
12327         (expand_mem_singnal_fence): Ditto.
12328         (expand_atomic_load): Ditto.
12329         (expand_atomic_store): Ditto.
12330         * doc/md.texi (Standard Pattern Names For Generation):
12331         Document memory_blockage instruction pattern.
12333 2017-10-13  Sebastian Perta  <sebastian.perta@renesas.com>
12335         * config/rl78/rl78.c (rl78_emit_libcall): New function.
12336         * config/rl78/rl78-protos.h (rl78_emit_libcall): New function.
12337         * config/rl78/rl78.md: New define_expand "adddi3".
12339 2017-10-13  Jan Hubicka  <hubicka@ucw.cz>
12341         * cfghooks.c (verify_flow_info): Disable check that all probabilities
12342         are set correctly.
12344 2017-10-13  Jeff Law  <law@redhat.com>
12346         * tree-ssa-reassoc.c (reassociate_bb): Clarify code slighly.
12348 2017-10-13  Jakub Jelinek  <jakub@redhat.com>
12350         PR target/82274
12351         * internal-fn.c (expand_mul_overflow): If both operands have
12352         the same highpart of -1 or 0 and the topmost bit of lowpart
12353         is different, overflow is if res <= 0 rather than res < 0.
12355 2017-10-13  Pat Haugen  <pthaugen@us.ibm.com>
12357         * config/rs6000/rs6000.c (rs6000_builtin_vectorization_cost): Remove
12358         TARGET_P9_VECTOR code for unaligned_load case.
12360 2017-10-13  Jan Hubicka  <hubicka@ucw.cz>
12362         * cfghooks.c (verify_flow_info): Check that edge probabilities are set.
12364 2017-10-13  Nathan Sidwell  <nathan@acm.org>
12366         * tree-core.h (tree_contains_struct): Make bool.
12367         * tree.c (tree_contains_struct): Likewise.
12368         * tree.h (MARK_TS_BASE): Remove do ... while (0) idiom.
12369         (MARK_TS_TYPED, MARK_TS_COMMON, MARK_TS_TYPE_COMMON,
12370         MARK_TS_TYPE_WITH_LANG_SPECIFIC, MARK_TS_DECL_MINIMAL,
12371         MARK_TS_DECL_COMMON, MARK_TS_DECL_WRTL, MARK_TS_DECL_WITH_VIS,
12372         MARK_TS_DECL_NON_COMMON): Likewise, use comma operator.
12374 2017-10-13  Richard Biener  <rguenther@suse.de>
12376         * graphite-isl-ast-to-gimple.c
12377         (translate_isl_ast_to_gimple::get_rename_from_scev): Remove unused
12378         parameters and dominance check.
12379         (translate_isl_ast_to_gimple::graphite_copy_stmts_from_block): Adjust.
12380         (translate_isl_ast_to_gimple::copy_bb_and_scalar_dependences): Likewise.
12381         (translate_isl_ast_to_gimple::graphite_regenerate_ast_isl):
12382         Do not update SSA form here or do intermediate IL verification.
12383         * graphite.c: Include tree-ssa.h and tree-into-ssa.h.
12384         (graphite_initialize): Remove check on the number of loops in
12385         the function and inline into graphite_transform_loops.
12386         (graphite_finalize): Inline into graphite_transform_loops.
12387         (graphite_transform_loops): Perform SSA update and IL verification
12388         here.
12389         * params.def (PARAM_GRAPHITE_MIN_LOOPS_PER_FUNCTION): Remove.
12391 2017-10-13  Richard Biener  <rguenther@suse.de>
12393         * graphite-isl-ast-to-gimple.c (max_mode_int_precision,
12394         graphite_expression_type_precision): Avoid global constructor
12395         by moving ...
12396         (translate_isl_ast_to_gimple::translate_isl_ast_to_gimple): Here.
12397         (translate_isl_ast_to_gimple::graphite_expr_type): Add type member.
12398         (translate_isl_ast_to_gimple::translate_isl_ast_node_for): Use it.
12399         (translate_isl_ast_to_gimple::build_iv_mapping): Likewise.
12400         (translate_isl_ast_to_gimple::graphite_create_new_guard): Likewise.
12401         * graphite-sese-to-poly.c (build_original_schedule): Return nothing.
12403 2017-10-13  H.J. Lu  <hongjiu.lu@intel.com>
12405         PR target/82499
12406         * config/i386/i386.h (ix86_red_zone_size): New.
12407         * config/i386/i386.md (push peephole2s): Replace
12408         "!ix86_using_red_zone ()" with "ix86_red_zone_size == 0".
12410 2017-10-13  Richard Sandiford  <richard.sandiford@linaro.org>
12411             Alan Hayward  <alan.hayward@arm.com>
12412             David Sherwood  <david.sherwood@arm.com>
12414         * combine.c (can_change_dest_mode): Reject changes in
12415         REGMODE_NATURAL_SIZE.
12417 2017-10-13  Richard Sandiford  <richard.sandiford@linaro.org>
12418             Alan Hayward  <alan.hayward@arm.com>
12419             David Sherwood  <david.sherwood@arm.com>
12421         * cfgexpand.c (expand_debug_expr): Use GET_MODE_UNIT_BITSIZE.
12422         (expand_debug_source_expr): Likewise.
12423         * combine.c (combine_simplify_rtx): Likewise.
12424         * cse.c (fold_rtx): Likewise.
12425         * fwprop.c (canonicalize_address): Likewise.
12426         * targhooks.c (default_shift_truncation_mask): Likewise.
12428 2017-10-13  Richard Sandiford  <richard.sandiford@linaro.org>
12429             Alan Hayward  <alan.hayward@arm.com>
12430             David Sherwood  <david.sherwood@arm.com>
12432         * optabs.c (add_equal_note): Use GET_MODE_UNIT_SIZE.
12433         (widened_mode): Likewise.
12434         (expand_unop): Likewise.
12435         * ree.c (transform_ifelse): Likewise.
12436         (merge_def_and_ext): Likewise.
12437         (combine_reaching_defs): Likewise.
12438         * simplify-rtx.c (simplify_unary_operation_1): Likewise.
12440 2017-10-13  Richard Sandiford  <richard.sandiford@linaro.org>
12441             Alan Hayward  <alan.hayward@arm.com>
12442             David Sherwood  <david.sherwood@arm.com>
12444         * caller-save.c (replace_reg_with_saved_mem): Use byte_lowpart_offset.
12445         * combine.c (gen_lowpart_for_combine): Likewise.
12446         * dwarf2out.c (rtl_for_decl_location): Likewise.
12447         * final.c (alter_subreg): Likewise.
12448         * rtlhooks.c (gen_lowpart_general): Likewise.
12449         (gen_lowpart_if_possible): Likewise.
12451 2017-10-13  Richard Sandiford  <richard.sandiford@linaro.org>
12452             Alan Hayward  <alan.hayward@arm.com>
12453             David Sherwood  <david.sherwood@arm.com>
12455         * calls.c (expand_call): Use subreg_lowpart_offset.
12456         * cse.c (cse_insn): Likewise.
12457         * regcprop.c (copy_value): Likewise.
12458         (copyprop_hardreg_forward_1): Likewise.
12460 2017-10-13  Jakub Jelinek  <jakub@redhat.com>
12462         PR target/82524
12463         * config/i386/i386.md (addqi_ext_1, andqi_ext_1,
12464         *andqi_ext_1_cc, *<code>qi_ext_1, *xorqi_ext_1_cc): Change
12465         =Q constraints to +Q and into insn condition add check
12466         that operands[0] and operands[1] are equal.
12467         (*addqi_ext_2, *andqi_ext_2, *<code>qi_ext_2): Change
12468         =Q constraints to +Q and into insn condition add check
12469         that operands[0] is equal to either operands[1] or operands[2].
12471         PR target/82498
12472         * fold-const.c (fold_binary_loc) <bit_rotate>: Code cleanups,
12473         instead of handling MINUS_EXPR twice (once for each argument),
12474         canonicalize operand order and handle just once, use rtype where
12475         possible.  Handle (A << B) | (A >> (-B & (Z - 1))).
12477         PR target/82498
12478         * config/i386/ia32intrin.h (__rold, __rord, __rolq, __rorq): Allow
12479         any values of __C while still being pattern recognizable as a simple
12480         rotate instruction.
12482 2017-10-13  Richard Biener  <rguenther@suse.de>
12484         PR tree-optimization/82451
12485         Revert
12486         2017-10-02  Richard Biener  <rguenther@suse.de>
12488         PR tree-optimization/82355
12489         * graphite-isl-ast-to-gimple.c (build_iv_mapping): Also build
12490         a mapping for the enclosing loop but avoid generating one for
12491         the loop tree root.
12492         (copy_bb_and_scalar_dependences): Remove premature codegen
12493         error on PHIs in blocks duplicated into multiple places.
12494         * graphite-scop-detection.c
12495         (scop_detection::stmt_has_simple_data_refs_p): For a loop not
12496         in the region use it as loop and nest to analyze the DR in.
12497         (try_generate_gimple_bb): Likewise.
12498         * graphite-sese-to-poly.c (extract_affine_chrec): Adjust.
12499         (add_loop_constraints): For blocks in a loop not in the region
12500         create a dimension with a single iteration.
12501         * sese.h (gbb_loop_at_index): Remove assert.
12503         * cfgloop.c (loop_preheader_edge): For the loop tree root
12504         return the single successor of the entry block.
12505         * graphite-isl-ast-to-gimple.c (graphite_regenerate_ast_isl):
12506         Reset the SCEV hashtable and niters.
12507         * graphite-scop-detection.c
12508         (scop_detection::graphite_can_represent_scev): Add SCOP parameter,
12509         assert that we only have POLYNOMIAL_CHREC that vary in loops
12510         contained in the region.
12511         (scop_detection::graphite_can_represent_expr): Adjust.
12512         (scop_detection::stmt_has_simple_data_refs_p): For loops
12513         not in the region set loop to NULL.  The nest is now the
12514         entry edge to the region.
12515         (try_generate_gimple_bb): Likewise.
12516         * sese.c (scalar_evolution_in_region): Adjust for
12517         instantiate_scev change.
12518         * tree-data-ref.h (graphite_find_data_references_in_stmt):
12519         Make nest parameter the edge into the region.
12520         (create_data_ref): Likewise.
12521         * tree-data-ref.c (dr_analyze_indices): Make nest parameter an
12522         entry edge into a region and adjust instantiate_scev calls.
12523         (create_data_ref): Likewise.
12524         (graphite_find_data_references_in_stmt): Likewise.
12525         (find_data_references_in_stmt): Pass the loop preheader edge
12526         from the nest argument.
12527         * tree-scalar-evolution.h (instantiate_scev): Make instantiate_below
12528         parameter the edge into the region.
12529         (instantiate_parameters): Use the loop preheader edge as entry.
12530         * tree-scalar-evolution.c (analyze_scalar_evolution): Handle
12531         NULL loop.
12532         (get_instantiated_value_entry): Make instantiate_below parameter
12533         the edge into the region.
12534         (instantiate_scev_name): Likewise.  Adjust dominance checks,
12535         when we cannot use loop-based instantiation instantiate by
12536         walking use-def chains.
12537         (instantiate_scev_poly): Adjust.
12538         (instantiate_scev_binary): Likewise.
12539         (instantiate_scev_convert): Likewise.
12540         (instantiate_scev_not): Likewise.
12541         (instantiate_array_ref): Remove.
12542         (instantiate_scev_3): Likewise.
12543         (instantiate_scev_2): Likewise.
12544         (instantiate_scev_1): Likewise.
12545         (instantiate_scev_r): Do not blindly handle N-operand trees.
12546         Do not instantiate array-refs.  Handle all constants and invariants.
12547         (instantiate_scev): Make instantiate_below parameter
12548         the edge into the region.
12549         (resolve_mixers): Use the loop preheader edge for the region
12550         parameter to instantiate_scev_r.
12551         * tree-ssa-loop-prefetch.c (determine_loop_nest_reuse): Adjust.
12553 2017-10-13  Richard Biener  <rguenther@suse.de>
12555         PR tree-optimization/82525
12556         * graphite-isl-ast-to-gimple.c
12557         (translate_isl_ast_to_gimple::widest_int_from_isl_expr_int): Split
12558         out from ...
12559         (translate_isl_ast_to_gimple::gcc_expression_from_isl_expr_int): Here.
12560         Fail code generation when we cannot represent the isl integer.
12561         (binary_op_to_tree): Elide modulo operations that are no-ops
12562         in the type we code generate.  Remove now superfluous code
12563         generation errors.
12565 2017-10-13  Richard Biener  <rguenther@suse.de>
12567         * graphite-scop-detection.c (loop_ivs_can_be_represented): Remove.
12568         (scop_detection::harmful_loop_in_region): Remove premature
12569         IV type restriction.
12570         (scop_detection::graphite_can_represent_scev): We can handle
12571         pointer IVs just fine.
12573 2017-10-13  Alan Modra  <amodra@gmail.com>
12575         * doc/extend.texi (Extended Asm <Clobbers>): Rename to
12576         "Clobbers and Scratch Registers".  Add paragraph on
12577         alternative to clobbers for scratch registers and OpenBLAS
12578         example.
12580 2017-10-13  Alan Modra  <amodra@gmail.com>
12582         * doc/extend.texi (Clobbers): Correct vax example.  Delete old
12583         example of a memory input for a string of known length.  Move
12584         commentary out of table.  Add a number of new examples
12585         covering array memory inputs.
12587 2017-10-12  Martin Liska  <mliska@suse.cz>
12589         PR tree-optimization/82493
12590         * sbitmap.c (bitmap_bit_in_range_p): Fix the implementation.
12591         (test_range_functions): New function.
12592         (sbitmap_c_tests): Likewise.
12593         * selftest-run-tests.c (selftest::run_tests): Run new tests.
12594         * selftest.h (sbitmap_c_tests): New function.
12596         * tree-ssa-dse.c (live_bytes_read): Fix thinko.
12598 2017-10-12  Michael Meissner  <meissner@linux.vnet.ibm.com>
12600         * config/rs6000/amo.h: Fix spacing issue.
12602 2017-10-12  Jakub Jelinek  <jakub@redhat.com>
12604         PR target/82498
12605         * config/i386/i386.md (*ashl<mode>3_mask_1,
12606         *<shift_insn><mode>3_mask_1, *<rotate_insn><mode>3_mask_1,
12607         *<btsc><mode>_mask_1, *btr<mode>_mask_1): New define_insn_and_split
12608         patterns.
12610 2017-10-12  Jan Hubicka  <hubicka@ucw.cz>
12612         * profile-count.h (safe_scale_64bit): Fix GCC4.x path.
12613         (profile_probability): Set max_probability
12614         to (uint32_t) 1 << (n_bits - 2) and update accessors to avoid overlfows
12615         in temporaries.
12616         * profile-count.c (profile_probability::differs_from_p): Do not
12617         rely on max_probaiblity == 10000
12619 2017-10-12  Jeff Law  <law@redhat.com>
12621         * tree-ssa-dse.c (valid_ao_ref_for_dse): Reject ao_refs with
12622         negative offsets.
12624 2017-10-12  Martin Sebor  <msebor@redhat.com>
12626         PR other/82301
12627         PR c/82435
12628         * cgraphunit.c (maybe_diag_incompatible_alias): New function.
12629         (handle_alias_pairs): Call it.
12630         * common.opt (-Wattribute-alias): New option.
12631         * doc/extend.texi (ifunc attribute): Discuss C++ specifics.
12632         * doc/invoke.texi (-Wattribute-alias): Document.
12634 2017-10-12  Vladimir Makarov  <vmakarov@redhat.com>
12636         Revert
12637         2017-10-11  Vladimir Makarov  <vmakarov@redhat.com>
12638         PR sanitizer/82353
12639         * lra.c (collect_non_operand_hard_regs): Don't ignore operator
12640         locations.
12641         * lra-lives.c (bb_killed_pseudos, bb_gen_pseudos): Move up.
12642         (make_hard_regno_born, make_hard_regno_dead): Update
12643         bb_killed_pseudos and bb_gen_pseudos.
12645 2017-10-12  Jan Hubicka  <hubicka@ucw.cz>
12647         * config/i386/x86-tune-sched.c (ix86_adjust_cost): Fix Zen support.
12649 2017-10-12  Uros Bizjak  <ubizjak@gmail.com>
12651         * config/alpha/alpha.c (alpha_split_conditional_move):
12652         Use std::swap instead of manually swapping.
12653         (alpha_stdarg_optimize_hook): Ditto.
12654         (alpha_canonicalize_comparison): Ditto.
12656 2017-10-12  Bin Cheng  <bin.cheng@arm.com>
12658         * tree-loop-distribution.c (struct builtin_info): New struct.
12659         (struct partition): Refactor fields into struct builtin_info.
12660         (partition_free): Free struct builtin_info.
12661         (build_size_arg_loc, build_addr_arg_loc): Delete.
12662         (generate_memset_builtin, generate_memcpy_builtin): Get memory range
12663         information from struct builtin_info.
12664         (find_single_drs): New function refactored from classify_partition.
12665         Also moved builtin validity checks to this function.
12666         (compute_access_range, alloc_builtin): New functions.
12667         (classify_builtin_st, classify_builtin_ldst): New functions.
12668         (classify_partition): Refactor code into functions find_single_drs,
12669         classify_builtin_st and classify_builtin_ldst.
12670         (distribute_loop): Don't do runtime alias check when distributing
12671         loop nest.
12672         (find_seed_stmts_for_distribution): New function.
12673         (pass_loop_distribution::execute): Refactor code finding seed
12674         stmts into above function.  Support distribution for the innermost
12675         two-level loop nest.  Adjust dump information.
12677 2017-10-12  Bin Cheng  <bin.cheng@arm.com>
12679         * tree-loop-distribution.c: Adjust the general comment.
12680         (NUM_PARTITION_THRESHOLD): New macro.
12681         (ssa_name_has_uses_outside_loop_p): Support loop nest distribution.
12682         (classify_partition): Skip builtin pattern of loop nest's inner loop.
12683         (merge_dep_scc_partitions): New parameter ignore_alias_p and use it
12684         in call to build_partition_graph.
12685         (finalize_partitions): New parameter.  Make loop distribution more
12686         conservative by fusing more partitions.
12687         (distribute_loop): Don't do runtime alias check in case of loop nest
12688         distribution.
12689         (find_seed_stmts_for_distribution): New function.
12690         (prepare_perfect_loop_nest): New function.
12691         (pass_loop_distribution::execute): Refactor code finding seed stmts
12692         and loop nest into above functions.  Support loop nest distribution.
12693         Adjust dump information accordingly.
12695 2017-10-12  Bin Cheng  <bin.cheng@arm.com>
12697         * tree-loop-distribution.c (break_alias_scc_partitions): Add comment
12698         and set PTYPE_SEQUENTIAL for merged partition.
12700 2017-10-12  Richard Biener  <rguenther@suse.de>
12702         PR tree-optimization/69728
12703         Revert
12704         2017-09-19  Richard Biener  <rguenther@suse.de>
12706         PR tree-optimization/69728
12707         * graphite-sese-to-poly.c (schedule_error): New global.
12708         (add_loop_schedule): Handle empty domain by failing the
12709         schedule.
12710         (build_original_schedule): Handle schedule_error.
12712         * graphite-sese-to-poly.c (add_loop_schedule): Handle empty
12713         domain by returning an unchanged schedule.
12715 2017-10-12  Jakub Jelinek  <jakub@redhat.com>
12717         * genrecog.c (validate_pattern): For VEC_SELECT verify that
12718         CONST_INT selectors are 0 to GET_MODE_NUNITS (imode) - 1.
12720 2017-10-12  Aldy Hernandez  <aldyh@redhat.com>
12722         * Makefile.in (TAGS): Merge all the *.def files into one pattern.
12723         Handle params.def.
12725 2017-10-12  Jakub Jelinek  <jakub@redhat.com>
12727         PR c++/82159
12728         * expr.c (store_field): Don't optimize away bitsize == 0 store
12729         from CALL_EXPR with addressable return type.
12731 2017-10-11  Segher Boessenkool  <segher@kernel.crashing.org>
12733         * config/rs6000/rs6000.h (TARGET_ISEL64): Delete.
12734         * config/rs6000/rs6000.md (sel): Delete mode attribute.
12735         (mov<mode>cc, isel_signed_<mode>, isel_unsigned_<mode>,
12736         *isel_reversed_signed_<mode>, *isel_reversed_unsigned_<mode>): Use
12737         TARGET_ISEL instead of TARGET_ISEL<sel>.
12739 2017-10-11  David Edelsohn  <dje.gcc@gmail.com>
12741         * config/rs6000/rs6000.c
12742         (rs6000_xcoff_asm_output_aligned_decl_common): Test for NULL decl.
12744 2017-10-11  Segher Boessenkool  <segher@kernel.crashing.org>
12746         * config/rs6000/predicates.md (zero_constant, all_ones_constant):
12747         Move up in file.
12748         (reg_or_cint_operand): Fix comment.
12749         (reg_or_zero_operand): New predicate.
12750         * config/rs6000/rs6000-protos.h (output_isel): Delete.
12751         * config/rs6000/rs6000.c (output_isel): Delete.
12752         * config/rs6000/rs6000.md (isel_signed_<mode>): Use reg_or_zero_operand
12753         instead of reg_or_cint_operand.  Output instruction directly (not via
12754         output_isel).
12755         (isel_unsigned_<mode>): Ditto.
12756         (*isel_reversed_signed_<mode>): Use reg_or_zero_operand instead of
12757         gpc_reg_operand.  Add an instruction alternative for this.  Output
12758         instruction directly.
12759         (*isel_reversed_unsigned_<mode>): Ditto.
12761 2017-10-11  Uros Bizjak  <ubizjak@gmail.com>
12763         * config/i386/i386.c (ix86_canonicalize_comparison): New function.
12764         (TARGET_CANONICALIZE_COMPARISON): Define.
12766 2017-10-11  Qing Zhao  <qing.zhao@oracle.com>
12768         PR target/81422
12769         * config/aarch64/aarch64.c (aarch64_load_symref_appropriately):
12770         Check whether the dest is REG before adding REG_EQUIV note.
12772 2017-10-11  Vladimir Makarov  <vmakarov@redhat.com>
12774         PR sanitizer/82353
12775         * lra.c (collect_non_operand_hard_regs): Don't ignore operator
12776         locations.
12777         * lra-lives.c (bb_killed_pseudos, bb_gen_pseudos): Move up.
12778         (make_hard_regno_born, make_hard_regno_dead): Update
12779         bb_killed_pseudos and bb_gen_pseudos.
12781 2017-10-11  Nathan Sidwell  <nathan@acm.org>
12783         * incpath.h (enum incpath_kind): Name enum, prefix values.
12784         (add_path, add_cpp_dir_path, get_added_cpp_dirs): Use incpath_kind.
12785         * incpath.c (heads, tails): Use INC_MAX.
12786         (add_env_var_paths, add_standard_paths): Use incpath_kind.
12787         (merge_include_chains, split_quote_chain,
12788         register_include_chains): Update incpath_kind names.
12789         (add_cpp_dir_path, add_path, get_added_cpp_dirs): Use incpath_kind.
12790         * config/darwin-c.c (add_system_framework_path): Update incpath_kind
12791         names.
12792         (add_framework_path, darwin_register_objc_includes): Likewise.
12793         * config/vms/vms-c.c (vms_c_register_includes): Likewise.
12795 2017-10-11  Uros Bizjak  <ubizjak@gmail.com>
12797         * config/i386/i386.md (*cmp<X87MODEF:mode>_<SWI24:mode>_i387):
12798         Do not use float_operator operator predicate.
12799         (*cmp<X87MODEF:mode>_<SWI24:mode>_cc_i387): Ditto.
12800         * config/i386/predicates.md (float_operator): Remove predicate.
12802 2017-10-11  Uros Bizjak  <ubizjak@gmail.com>
12804         * config/i386/i386.md (*jcc<mode>_0_i387): Remove insn pattern.
12805         (*jccxf_i387): Ditto.
12806         (*jcc<mode>_i387): Ditto.
12807         (*jccu<mode>_i387): Ditto.
12808         (*jcc<X87MODEF:mode>_<SWI24:mode>_i387): Ditto.
12809         (*jcc_*_i387 splitters): Remove.
12810         * config/i386/i386-protos.h (ix86_split_fp_branch): Remove prototype.
12811         * config/i386/i386.c (ix86_split_fp_branch): Remove.
12812         * config/i386/predicates.md (ix86_swapped_fp_comparison_operator):
12813         Remove predicate.
12815 2017-10-11  Jan Hubicka  <hubicka@ucw.cz>
12817         * profile-count.h (slow_safe_scale_64bit): New function.
12818         (safe_scale_64bit): New inline.
12819         (profile_count::max_safe_multiplier): Remove; use safe_scale_64bit.
12820         * profile-count.c: Include wide-int.h
12821         (slow_safe_scale_64bit): New.
12823 2017-10-11  Nathan Sidwell  <nathan@acm.org>
12825         * tree.h (DECL_ASSEMBLER_NAME_SET_P): Don't check
12826         HAS_DECL_ASSEMBLER_NAME_P.
12827         * gimple-expr.c (gimple_decl_printable_name: Check
12828         HAS_DECL_ASSEMBLER_NAME_P too.
12829         * ipa-utils.h (type_in_anonymous_namespace_p): Check
12830         DECL_ASSEMBLER_NAME_SET_P of TYPE_NAME.
12831         (odr_type_p): No need to assert TYPE_NAME is a TYPE_DECL.
12832         * passes.c (rest_of_decl_compilation): Check
12833         HAS_DECL_ASSEMBLER_NAME_P too.
12834         * recog.c (verify_changes): Likewise.
12835         * tree-pretty-print.c (dump_decl_name): Likewise.
12836         * tree-ssa-structalias.c (alias_get_name): Likewise.  Reimplement.
12838         * tree.h (DECL_ASSEMBLER_NAME_RAW): New.
12839         (SET_DECL_ASSEMBLER_NAME): Use it.
12840         (DECL_ASSEMBLER_NAME_SET_P): Likewise.
12841         (COPY_DECL_ASSEMBLER_NAME): Likewise.
12842         * tree.c (decl_assembler_name): Use DECL_ASSEMBLER_NAME_RAW.
12844 2017-10-11  Jan Hubicka  <hubicka@ucw.cz>
12846         * config.gcc (i386, x86_64): Add extra objects.
12847         * config/i386/i386-protos.h (ix86_rip_relative_addr_p): Declare.
12848         (ix86_min_insn_size): Declare.
12849         (ix86_issue_rate): Declare.
12850         (ix86_adjust_cost): Declare.
12851         (ia32_multipass_dfa_lookahead): Declare.
12852         (ix86_macro_fusion_p): Declare.
12853         (ix86_macro_fusion_pair_p): Declare.
12854         (ix86_bd_has_dispatch): Declare.
12855         (ix86_bd_do_dispatch): Declare.
12856         (ix86_core2i7_init_hooks): Declare.
12857         (ix86_atom_sched_reorder): Declare.
12858         * config/i386/i386.c Move all CPU cost tables to x86-tune-costs.h.
12859         (COSTS_N_BYTES): Move to x86-tune-costs.h.
12860         (DUMMY_STRINGOP_ALGS):Move to x86-tune-costs.h.
12861         (rip_relative_addr_p): Rename to ...
12862         (ix86_rip_relative_addr_p): ... this one; export.
12863         (memory_address_length): Update.
12864         (ix86_issue_rate): Move to x86-tune-sched.c.
12865         (ix86_flags_dependent): Move to x86-tune-sched.c.
12866         (ix86_agi_dependent): Move to x86-tune-sched.c.
12867         (exact_dependency_1): Move to x86-tune-sched.c.
12868         (exact_store_load_dependency): Move to x86-tune-sched.c.
12869         (ix86_adjust_cost): Move to x86-tune-sched.c.
12870         (ia32_multipass_dfa_lookahead): Move to x86-tune-sched.c.
12871         (ix86_macro_fusion_p): Move to x86-tune-sched.c.
12872         (ix86_macro_fusion_pair_p): Move to x86-tune-sched.c.
12873         (do_reorder_for_imul): Move to x86-tune-sched-atom.c.
12874         (swap_top_of_ready_list): Move to x86-tune-sched-atom.c.
12875         (ix86_sched_reorder): Move to x86-tune-sched-atom.c.
12876         (core2i7_first_cycle_multipass_init): Move to x86-tune-sched-core.c.
12877         (core2i7_dfa_post_advance_cycle): Move to x86-tune-sched-core.c.
12878         (min_insn_size): Rename to ...
12879         (ix86_min_insn_size): ... this one; export.
12880         (core2i7_first_cycle_multipass_begin): Move to x86-tune-sched-core.c.
12881         (core2i7_first_cycle_multipass_issue): Move to x86-tune-sched-core.c.
12882         (core2i7_first_cycle_multipass_backtrack): Move to
12883         x86-tune-sched-core.c.
12884         (core2i7_first_cycle_multipass_end): Move to x86-tune-sched-core.c.
12885         (core2i7_first_cycle_multipass_fini): Move to x86-tune-sched-core.c.
12886         (ix86_sched_init_global): Break up logic to ix86_core2i7_init_hooks.
12887         (ix86_avoid_jump_mispredicts): Update.
12888         (TARGET_SCHED_DISPATCH): Move to ix86-tune-sched-bd.c.
12889         (TARGET_SCHED_DISPATCH_DO): Move to ix86-tune-sched-bd.c.
12890         (TARGET_SCHED_REORDER): Move to ix86-tune-sched-bd.c.
12891         (DISPATCH_WINDOW_SIZE): Move to ix86-tune-sched-bd.c.
12892         (MAX_DISPATCH_WINDOWS): Move to ix86-tune-sched-bd.c.
12893         (MAX_INSN): Move to ix86-tune-sched-bd.c.
12894         (MAX_IMM): Move to ix86-tune-sched-bd.c.
12895         (MAX_IMM_SIZE): Move to ix86-tune-sched-bd.c.
12896         (MAX_IMM_32): Move to ix86-tune-sched-bd.c.
12897         (MAX_IMM_64): Move to ix86-tune-sched-bd.c.
12898         (MAX_LOAD): Move to ix86-tune-sched-bd.c.
12899         (MAX_STORE): Move to ix86-tune-sched-bd.c.
12900         (BIG): Move to ix86-tune-sched-bd.c.
12901         (enum dispatch_group): Move to ix86-tune-sched-bd.c.
12902         (enum insn_path): Move to ix86-tune-sched-bd.c.
12903         (get_mem_group): Move to ix86-tune-sched-bd.c.
12904         (is_cmp): Move to ix86-tune-sched-bd.c.
12905         (dispatch_violation): Move to ix86-tune-sched-bd.c.
12906         (is_branch): Move to ix86-tune-sched-bd.c.
12907         (is_prefetch): Move to ix86-tune-sched-bd.c.
12908         (init_window): Move to ix86-tune-sched-bd.c.
12909         (allocate_window): Move to ix86-tune-sched-bd.c.
12910         (init_dispatch_sched): Move to ix86-tune-sched-bd.c.
12911         (is_end_basic_block): Move to ix86-tune-sched-bd.c.
12912         (process_end_window): Move to ix86-tune-sched-bd.c.
12913         (allocate_next_window): Move to ix86-tune-sched-bd.c.
12914         (find_constant): Move to ix86-tune-sched-bd.c.
12915         (get_num_immediates): Move to ix86-tune-sched-bd.c.
12916         (has_immediate): Move to ix86-tune-sched-bd.c.
12917         (get_insn_path): Move to ix86-tune-sched-bd.c.
12918         (get_insn_group): Move to ix86-tune-sched-bd.c.
12919         (count_num_restricted): Move to ix86-tune-sched-bd.c.
12920         (fits_dispatch_window): Move to ix86-tune-sched-bd.c.
12921         (add_insn_window): Move to ix86-tune-sched-bd.c.
12922         (add_to_dispatch_window): Move to ix86-tune-sched-bd.c.
12923         (debug_dispatch_window_file): Move to ix86-tune-sched-bd.c.
12924         (debug_dispatch_window): Move to ix86-tune-sched-bd.c.
12925         (debug_insn_dispatch_info_file): Move to ix86-tune-sched-bd.c.
12926         (debug_ready_dispatch): Move to ix86-tune-sched-bd.c.
12927         (do_dispatch): Move to ix86-tune-sched-bd.c.
12928         (has_dispatch): Move to ix86-tune-sched-bd.c.
12929         * config/i386/t-i386: Add new object files.
12930         * config/i386/x86-tune-costs.h: New file.
12931         * config/i386/x86-tune-sched-atom.c: New file.
12932         * config/i386/x86-tune-sched-bd.c: New file.
12933         * config/i386/x86-tune-sched-core.c: New file.
12934         * config/i386/x86-tune-sched.c: New file.
12936 2017-10-11  Liu Hao  <lh_mouse@126.com>
12938         * pretty-print.c [_WIN32] (colorize_init): Remove.  Use
12939         the generic version below instead.
12940         (should_colorize): Recognize Windows consoles as terminals
12941         for MinGW targets.
12942         * pretty-print.c [__MINGW32__] (write_all): New function.
12943         [__MINGW32__] (find_esc_head): Likewise.
12944         [__MINGW32__] (find_esc_terminator): Likewise.
12945         [__MINGW32__] (eat_esc_sequence): Likewise.
12946         [__MINGW32__] (mingw_ansi_fputs): New function that handles
12947         ANSI escape codes.
12948         (pp_write_text_to_stream): Use mingw_ansi_fputs instead of fputs
12949         for MinGW targets.
12951 2017-10-11  Richard Biener  <rguenther@suse.de>
12953         * tree-ssa-loop-niter.c (infer_loop_bounds_from_pointer_arith):
12954         Properly call analyze_scalar_evolution with the loop of the stmt.
12956 2017-10-11  Richard Biener  <rguenther@suse.de>
12958         * tree.def (POLYNOMIAL_CHREC): Remove CHREC_VARIABLE tree operand.
12959         * tree-core.h (tree_base): Add chrec_var union member.
12960         * tree.h (CHREC_VAR): Remove.
12961         (CHREC_LEFT, CHREC_RIGHT, CHREC_VARIABLE): Adjust.
12962         * tree-chrec.h (build_polynomial_chrec): Adjust.
12963         * tree-chrec.c (reset_evolution_in_loop): Use build_polynomial_chrec.
12964         * tree-pretty-print.c (dump_generic_node): Use CHREC_VARIABLE.
12966 2017-10-11  Marc Glisse  <marc.glisse@inria.fr>
12968         * fold-const.c (fold_binary_loc) [X +- Y CMP X]: Move ...
12969         * match.pd: ... here.
12970         ((T) X == (T) Y): Relax condition.
12972 2017-10-11  Bin Cheng  <bin.cheng@arm.com>
12974         PR tree-optimization/82472
12975         * tree-loop-distribution.c (sort_partitions_by_post_order): Refine
12976         comment.
12977         (break_alias_scc_partitions): Update postorder number.
12979 2017-10-11  Martin Liska  <mliska@suse.cz>
12981         PR sanitizer/82490
12982         * opts.c (parse_no_sanitize_attribute): Do not use error_value
12983         variable.
12984         * opts.h (parse_no_sanitize_attribute): Remove last argument.
12986 2017-10-11  Martin Liska  <mliska@suse.cz>
12988         * print-rtl.c (print_insn): Move declaration of idbuf
12989         to same scope as name.
12991 2017-10-11  Martin Liska  <mliska@suse.cz>
12993         Revert r253637:
12995         PR sanitizer/82484
12996         * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
12997         volatile arguments.
12999 2017-10-11  Martin Liska  <mliska@suse.cz>
13001         PR sanitizer/82484
13002         * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
13003         volatile arguments.
13005 2017-10-11  Adhemerval Zanella  <adhemerval.zanella@linaro.org>
13007         * config.gcc (default_gnu_indirect_function): Default to yes for
13008         arm*-*-linux* with glibc.
13010 2017-10-11  Richard Biener  <rguenther@suse.de>
13012         * tree-scalar-evolution.c (get_scalar_evolution): Handle
13013         default-defs and types we do not want to analyze.
13014         (interpret_loop_phi): Replace unreachable code with an assert.
13015         (compute_scalar_evolution_in_loop): Remove and inline ...
13016         (analyze_scalar_evolution_1): ... here, replacing condition with
13017         what makes the intent clearer.  Remove handling of cases
13018         get_scalar_evolution now handles.
13020 2017-10-10  Jim Wilson  <wilson@tuliptree.org>
13022         PR rtl-optimization/81434
13023         * haifa-sched.c (prune_ready_list): Init min_cost_group to 0.  Update
13024         comment for main loop.  In sched_group_found if, also add checks for
13025         pass and min_cost_group.
13027 2017-10-10  Segher Boessenkool  <segher@kernel.crashing.org>
13029         * config/rs6000/rs6000.c (TARGET_INSN_COST): New.
13030         (rs6000_insn_cost): New function.
13031         * config/rs6000/rs6000.md (cost): New attribute.
13033 2017-10-10  Jakub Jelinek  <jakub@redhat.com>
13034             H.J. Lu  <hongjiu.lu@intel.com>
13036         PR target/79565
13037         PR target/82483
13038         * config/i386/i386.c (ix86_init_mmx_sse_builtins): Add
13039         OPTION_MASK_ISA_MMX for __builtin_ia32_maskmovq,
13040         __builtin_ia32_vec_ext_v4hi and __builtin_ia32_vec_set_v4hi.
13041         (ix86_expand_builtin): Treat OPTION_MASK_ISA_MMX similarly
13042         to OPTION_MASK_ISA_AVX512VL - builtins that have both
13043         OPTION_MASK_ISA_MMX and some other bit set require both
13044         mmx and the ISAs without the mmx bit.
13045         * config/i386/i386-builtin.def (__builtin_ia32_cvtps2pi,
13046         __builtin_ia32_cvttps2pi, __builtin_ia32_cvtpi2ps,
13047         __builtin_ia32_pavgb, __builtin_ia32_pavgw, __builtin_ia32_pmulhuw,
13048         __builtin_ia32_pmaxub, __builtin_ia32_pmaxsw, __builtin_ia32_pminub,
13049         __builtin_ia32_pminsw, __builtin_ia32_psadbw, __builtin_ia32_pmovmskb,
13050         __builtin_ia32_pshufw, __builtin_ia32_cvtpd2pi,
13051         __builtin_ia32_cvttpd2pi, __builtin_ia32_cvtpi2pd,
13052         __builtin_ia32_pmuludq, __builtin_ia32_pabsb, __builtin_ia32_pabsw,
13053         __builtin_ia32_pabsd, __builtin_ia32_phaddw, __builtin_ia32_phaddd,
13054         __builtin_ia32_phaddsw, __builtin_ia32_phsubw, __builtin_ia32_phsubd,
13055         __builtin_ia32_phsubsw, __builtin_ia32_pmaddubsw,
13056         __builtin_ia32_pmulhrsw, __builtin_ia32_pshufb, __builtin_ia32_psignb,
13057         __builtin_ia32_psignw, __builtin_ia32_psignd, __builtin_ia32_movntq,
13058         __builtin_ia32_paddq, __builtin_ia32_psubq, __builtin_ia32_palignr):
13059         Add OPTION_MASK_ISA_MMX.
13061 2017-10-10  Andreas Tobler  <andreast@gcc.gnu.org>
13063         * config.gcc (armv7*-*-freebsd*): New target.
13064         (armv6*-*-freebsd*): Remove obsolete TARGET_FREEBSD_ARMv6 define.
13066 2017-10-10  Jan Hubicka  <hubicka@ucw.cz>
13068         * x86-tune.def (X86_TUNE_AVOID_FALSE_DEP_FOR_BMI,
13069         X86_TUNE_ADJUST_UNROLL, X86_TUNE_ONE_IF_CONV_INSN): Move to right
13070         spot in the file.
13072 2017-10-10  Richard Sandiford  <richard.sandiford@linaro.org>
13074         * wide-int.h (wide_int_ref_storage): Make host_dependent_precision
13075         a template parameter.
13076         (WIDE_INT_REF_FOR): Update accordingly.
13077         * tree.h (wi::int_traits <const_tree>): Delete.
13078         (wi::tree_to_widest_ref, wi::tree_to_offset_ref): New typedefs.
13079         (wi::to_widest, wi::to_offset): Use them.  Expand commentary.
13080         (wi::tree_to_wide_ref): New typedef.
13081         (wi::to_wide): New function.
13082         * calls.c (get_size_range): Use wi::to_wide when operating on
13083         trees as wide_ints.
13084         * cgraph.c (cgraph_node::create_thunk): Likewise.
13085         * config/i386/i386.c (ix86_data_alignment): Likewise.
13086         (ix86_local_alignment): Likewise.
13087         * dbxout.c (stabstr_O): Likewise.
13088         * dwarf2out.c (add_scalar_info, gen_enumeration_type_die): Likewise.
13089         * expr.c (const_vector_from_tree): Likewise.
13090         * fold-const-call.c (host_size_t_cst_p, fold_const_call_1): Likewise.
13091         * fold-const.c (may_negate_without_overflow_p, negate_expr_p)
13092         (fold_negate_expr_1, int_const_binop_1, const_binop)
13093         (fold_convert_const_int_from_real, optimize_bit_field_compare)
13094         (all_ones_mask_p, sign_bit_p, unextend, extract_muldiv_1)
13095         (fold_div_compare, fold_single_bit_test, fold_plusminus_mult_expr)
13096         (pointer_may_wrap_p, expr_not_equal_to, fold_binary_loc)
13097         (fold_ternary_loc, multiple_of_p, fold_negate_const, fold_abs_const)
13098         (fold_not_const, round_up_loc): Likewise.
13099         * gimple-fold.c (gimple_fold_indirect_ref): Likewise.
13100         * gimple-ssa-warn-alloca.c (alloca_call_type_by_arg): Likewise.
13101         (alloca_call_type): Likewise.
13102         * gimple.c (preprocess_case_label_vec_for_gimple): Likewise.
13103         * godump.c (go_output_typedef): Likewise.
13104         * graphite-sese-to-poly.c (tree_int_to_gmp): Likewise.
13105         * internal-fn.c (get_min_precision): Likewise.
13106         * ipa-cp.c (ipcp_store_vr_results): Likewise.
13107         * ipa-polymorphic-call.c
13108         (ipa_polymorphic_call_context::ipa_polymorphic_call_context): Likewise.
13109         * ipa-prop.c (ipa_print_node_jump_functions_for_edge): Likewise.
13110         (ipa_modify_call_arguments): Likewise.
13111         * match.pd: Likewise.
13112         * omp-low.c (scan_omp_1_op, lower_omp_ordered_clauses): Likewise.
13113         * print-tree.c (print_node_brief, print_node): Likewise.
13114         * stmt.c (expand_case): Likewise.
13115         * stor-layout.c (layout_type): Likewise.
13116         * tree-affine.c (tree_to_aff_combination): Likewise.
13117         * tree-cfg.c (group_case_labels_stmt): Likewise.
13118         * tree-data-ref.c (dr_analyze_indices): Likewise.
13119         (prune_runtime_alias_test_list): Likewise.
13120         * tree-dump.c (dequeue_and_dump): Likewise.
13121         * tree-inline.c (remap_gimple_op_r, copy_tree_body_r): Likewise.
13122         * tree-predcom.c (is_inv_store_elimination_chain): Likewise.
13123         * tree-pretty-print.c (dump_generic_node): Likewise.
13124         * tree-scalar-evolution.c (iv_can_overflow_p): Likewise.
13125         (simple_iv_with_niters): Likewise.
13126         * tree-ssa-address.c (addr_for_mem_ref): Likewise.
13127         * tree-ssa-ccp.c (ccp_finalize, evaluate_stmt): Likewise.
13128         * tree-ssa-loop-ivopts.c (constant_multiple_of): Likewise.
13129         * tree-ssa-loop-niter.c (split_to_var_and_offset)
13130         (refine_value_range_using_guard, number_of_iterations_ne_max)
13131         (number_of_iterations_lt_to_ne, number_of_iterations_lt)
13132         (get_cst_init_from_scev, record_nonwrapping_iv)
13133         (scev_var_range_cant_overflow): Likewise.
13134         * tree-ssa-phiopt.c (minmax_replacement): Likewise.
13135         * tree-ssa-pre.c (compute_avail): Likewise.
13136         * tree-ssa-sccvn.c (vn_reference_fold_indirect): Likewise.
13137         (vn_reference_maybe_forwprop_address, valueized_wider_op): Likewise.
13138         * tree-ssa-structalias.c (get_constraint_for_ptr_offset): Likewise.
13139         * tree-ssa-uninit.c (is_pred_expr_subset_of): Likewise.
13140         * tree-ssanames.c (set_nonzero_bits, get_nonzero_bits): Likewise.
13141         * tree-switch-conversion.c (collect_switch_conv_info, array_value_type)
13142         (dump_case_nodes, try_switch_expansion): Likewise.
13143         * tree-vect-loop-manip.c (vect_gen_vector_loop_niters): Likewise.
13144         (vect_do_peeling): Likewise.
13145         * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
13146         * tree-vect-stmts.c (vectorizable_load): Likewise.
13147         * tree-vrp.c (compare_values_warnv, vrp_int_const_binop): Likewise.
13148         (zero_nonzero_bits_from_vr, ranges_from_anti_range): Likewise.
13149         (extract_range_from_binary_expr_1, adjust_range_with_scev): Likewise.
13150         (overflow_comparison_p_1, register_edge_assert_for_2): Likewise.
13151         (is_masked_range_test, find_switch_asserts, maybe_set_nonzero_bits)
13152         (vrp_evaluate_conditional_warnv_with_ops, intersect_ranges): Likewise.
13153         (range_fits_type_p, two_valued_val_range_p, vrp_finalize): Likewise.
13154         (evrp_dom_walker::before_dom_children): Likewise.
13155         * tree.c (cache_integer_cst, real_value_from_int_cst, integer_zerop)
13156         (integer_all_onesp, integer_pow2p, integer_nonzerop, tree_log2)
13157         (tree_floor_log2, tree_ctz, mem_ref_offset, tree_int_cst_sign_bit)
13158         (tree_int_cst_sgn, get_unwidened, int_fits_type_p): Likewise.
13159         (get_type_static_bounds, num_ending_zeros, drop_tree_overflow)
13160         (get_range_pos_neg): Likewise.
13161         * ubsan.c (ubsan_expand_ptr_ifn): Likewise.
13162         * config/darwin.c (darwin_mergeable_constant_section): Likewise.
13163         * config/aarch64/aarch64.c (aapcs_vfp_sub_candidate): Likewise.
13164         * config/arm/arm.c (aapcs_vfp_sub_candidate): Likewise.
13165         * config/avr/avr.c (avr_fold_builtin): Likewise.
13166         * config/bfin/bfin.c (bfin_local_alignment): Likewise.
13167         * config/msp430/msp430.c (msp430_attr): Likewise.
13168         * config/nds32/nds32.c (nds32_insert_attributes): Likewise.
13169         * config/powerpcspe/powerpcspe-c.c
13170         (altivec_resolve_overloaded_builtin): Likewise.
13171         * config/powerpcspe/powerpcspe.c (rs6000_aggregate_candidate)
13172         (rs6000_expand_ternop_builtin): Likewise.
13173         * config/rs6000/rs6000-c.c
13174         (altivec_resolve_overloaded_builtin): Likewise.
13175         * config/rs6000/rs6000.c (rs6000_aggregate_candidate): Likewise.
13176         (rs6000_expand_ternop_builtin): Likewise.
13177         * config/s390/s390.c (s390_handle_hotpatch_attribute): Likewise.
13179 2017-10-10  Bin Cheng  <bin.cheng@arm.com>
13181         * tree-vect-loop-manip.c (rename_variables_in_bb): Rename PHI nodes
13182         when copying loop nest with only one inner loop.
13184 2017-10-10  Richard Biener  <rguenther@suse.de>
13186         * tree-cfgcleanup.c (cleanup_tree_cfg_noloop): Avoid compacting
13187         blocks if SCEV is active.
13188         * tree-scalar-evolution.c (analyze_scalar_evolution_1): Remove
13189         dead code.
13190         (analyze_scalar_evolution): Handle cached evolutions the obvious way.
13191         (scev_initialize): Assert we are not yet initialized.
13193 2017-10-10  Bin Cheng  <bin.cheng@arm.com>
13195         * tree-loop-distribution.c (generate_loops_for_partition): Remove
13196         inner loop's exit stmt by making it always exit the loop, otherwise
13197         we would generate an infinite empty loop.
13199 2017-10-10  Bin Cheng  <bin.cheng@arm.com>
13201         * tree-vect-loop-manip.c (slpeel_tree_duplicate_loop_to_edge_cfg): Skip
13202         renaming variables in new preheader if it's deleted.
13204 2017-10-10  Bin Cheng  <bin.cheng@arm.com>
13206         * tree-loop-distribution.c (struct partition): Remove unused field
13207         loops of the structure.
13208         (partition_alloc, partition_free): Ditto.
13209         (build_rdg_partition_for_vertex): Ditto.
13211 2017-10-09  Jeff Law  <law@redhat.com>
13213         * targhooks.c (default_stack_clash_protection_final_dynamic_probe): Fix
13214         return type to match prototype and documentation.
13216 2010-10-09  Segher Boessenkool  <segher@kernel.crashing.org>
13218         * config/rs6000/rs6000.c (processor_costs): Move to ...
13219         * config/rs6000/rs6000.h: ... here.
13220         (rs6000_cost): Declare.
13222 2017-10-09  Eric Botcazou  <ebotcazou@adacore.com>
13224         * except.c (setjmp_fn): New global variable.
13225         (init_eh): Initialize it if DONT_USE_BUILTIN_SETJMP is defined.
13226         (sjlj_emit_function_enter): Call it instead of BUILTIN_SETJMP
13227         if DONT_USE_BUILTIN_SETJMP is defined.
13229 2017-10-09  Segher Boessenkool  <segher@kernel.crashing.org>
13231         * target.def (insn_cost): New hook.
13232         * doc/tm.texi.in (TARGET_INSN_COST): New hook.
13233         * doc/tm.texi: Regenerate.
13234         * rtlanal.c (insn_cost): Use the new hook.
13236 2017-10-09  Segher Boessenkool  <segher@kernel.crashing.org>
13238         * combine.c (combine_validate_cost): Compute the new insn_cost,
13239         not just pattern_cost.
13240         (try_combine): Adjust comment.
13242 2017-10-09  Segher Boessenkool  <segher@kernel.crashing.org>
13244         * cfgrtl.c (rtl_account_profile_record): Replace insn_rtx_cost with
13245         insn_cost.
13246         * combine.c (uid_insn_cost): Adjust comment.
13247         (combine_validate_cost): Adjust comment.  Use pattern_cost instead
13248         of insn_rtx_cost
13249         (combine_instructions): Use insn_cost instead of insn_rtx_cost.
13250         * dse.c (find_shift_sequence): Ditto.
13251         * ifcvt.c (cheap_bb_rtx_cost_p): Ditto.
13252         (bb_valid_for_noce_process_p): Use pattern_cost.
13253         * rtl.h (insn_rtx_cost): Delete.
13254         (pattern_cost): New prototype.
13255         (insn_cost): New prototype.
13256         * rtlanal.c (insn_rtx_cost): Rename to...
13257         (pattern_cost): ... this.
13258         (insn_cost): New.
13260 2017-10-09  Uros Bizjak  <ubizjak@gmail.com>
13262         * config/i386/i386.md (*jcc_2): Remove insn pattern.
13263         (*jcc<mode>_0_r_i387): Ditto.
13264         (*jccxf_r_i387): Ditto.
13265         (*jcc<mode>_r_i387): Ditto.
13266         (*jccu<mode>_r_i387): Ditto.
13267         (*jcc<X87MODEF:mode>_<SWI24:mode>_r_i387): Ditto.
13268         (*jcc): Rename from *jcc_1.
13270 2017-10-09  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
13272         * config/rs6000/rs6000-p8swap.c (rs6000_analyze_swaps): Process
13273         deferred rescans after the lvx/stvx recombination pre-pass.
13275 2017-10-09  Michael Meissner  <meissner@linux.vnet.ibm.com>
13277         * config/rs6000/amo.h: New include file to provide ISA 3.0 atomic
13278         memory operation instruction support.
13279         * config.gcc (powerpc*-*-*): Include amo.h as an extra header.
13280         (rs6000-ibm-aix[789]*): Likewise.
13281         * doc/extend.texi (PowerPC Atomic Memory Operation Functions):
13282         Document new functions.
13284 2017-10-09  Richard Biener  <rguenther@suse.de>
13286         PR tree-optimization/82397
13287         * tree-data-ref.c (data_ref_compare_tree): Make sure to return
13288         equality only for semantically equal trees.
13290 2017-10-09  Richard Biener  <rguenther@suse.de>
13292         PR tree-optimization/82449
13293         * sese.c (scev_analyzable_p): Check whether the SCEV is linear.
13294         * tree-chrec.h (evolution_function_is_constant_p): Adjust to
13295         allow constant addresses.
13296         * tree-chrec.c (scev_is_linear_expression): Constant evolutions
13297         are linear.
13299 2017-10-09  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
13301         * config/s390/s390-builtins.def (vec_nabs, vec_vfi): Fix builtin
13302         flags.
13304 2017-10-09  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
13306         PR target/82463
13307         * config/s390/vecintrin.h (vec_madd, vec_msub): Fix macro
13308         definitions.
13310 2017-10-09  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
13312         PR target/82465
13313         * config/s390/s390-builtins.def (vec_sqrt): Fix builtin flags.
13315 2017-10-09  Jakub Jelinek  <jakub@redhat.com>
13317         PR target/82464
13318         * config/s390/s390-builtins.def (s390_vec_xor_flt_a,
13319         s390_vec_xor_flt_b, s390_vec_xor_flt_c): New.
13321 2017-10-09  Richard Sandiford  <richard.sandiford@linaro.org>
13323         * wide-int.h (WI_BINARY_OPERATOR_RESULT): New macro.
13324         (WI_BINARY_PREDICATE_RESULT): Likewise.
13325         (wi::binary_traits::operator_result): New type.
13326         (wi::binary_traits::predicate_result): Likewise.
13327         (generic_wide_int::operator~, unary generic_wide_int::operator-)
13328         (generic_wide_int::operator==, generic_wide_int::operator!=)
13329         (generic_wide_int::operator&, generic_wide_int::and_not)
13330         (generic_wide_int::operator|, generic_wide_int::or_not)
13331         (generic_wide_int::operator^, generic_wide_int::operator+
13332         (binary generic_wide_int::operator-, generic_wide_int::operator*):
13333         Delete.
13334         (operator~, unary operator-, operator==, operator!=, operator&)
13335         (operator|, operator^, operator+, binary operator-, operator*): New
13336         functions.
13337         * expr.c (get_inner_reference): Use wi::bit_and_not.
13338         * fold-const.c (fold_binary_loc): Likewise.
13339         * ipa-prop.c (ipa_compute_jump_functions_for_edge): Likewise.
13340         * tree-ssa-ccp.c (get_value_from_alignment): Likewise.
13341         (bit_value_binop): Likewise.
13342         * tree-ssa-math-opts.c (find_bswap_or_nop_load): Likewise.
13343         * tree-vrp.c (zero_nonzero_bits_from_vr): Likewise.
13344         (extract_range_from_binary_expr_1): Likewise.
13345         (masked_increment): Likewise.
13346         (simplify_bit_ops_using_ranges): Likewise.
13348 2017-10-09  Martin Jambor  <mjambor@suse.cz>
13350         PR hsa/82416
13351         * hsa-common.h (hsa_op_with_type): New method extend_int_to_32bit.
13352         * hsa-gen.c (hsa_extend_inttype_to_32bit): New function.
13353         (hsa_type_for_scalar_tree_type): Use it.  Always force min32int for
13354         COMPLEX types.
13355         (hsa_fixup_mov_insn_type): New function.
13356         (hsa_op_with_type::get_in_type): Use it.
13357         (hsa_build_append_simple_mov): Likewise.  Allow sub-32bit
13358         immediates in an assert.
13359         (hsa_op_with_type::extend_int_to_32bit): New method.
13360         (gen_hsa_insns_for_bitfield): Fixup instruction and intermediary
13361         types.  Convert to dest type if necessary.
13362         (gen_hsa_insns_for_bitfield_load): Fixup load type if necessary.
13363         (reg_for_gimple_ssa): Pass false as min32int to
13364         hsa_type_for_scalar_tree_type.
13365         (gen_hsa_addr): Fixup type when creating addresable temporary.
13366         (gen_hsa_cmp_insn_from_gimple): Extend operands if necessary.
13367         (gen_hsa_unary_operation): Extend operands and convert to dest type if
13368         necessary.  Call hsa_fixup_mov_insn_type.
13369         (gen_hsa_binary_operation): Changed operand types to hsa_op_with_type,
13370         extend operands and convert to dest type if necessary.
13371         (gen_hsa_insns_for_operation_assignment): Extend operands and convert
13372         to dest type if necessary.
13373         (set_output_in_type): Call hsa_fixup_mov_insn_type.  Just ude dest
13374         if conversion nt necessary and size matches.
13375         (gen_hsa_insns_for_load): Call hsa_fixup_mov_insn_type, convert
13376         to dest type if necessary.
13377         (gen_hsa_insns_for_store): Call hsa_fixup_mov_insn_type.
13378         (gen_hsa_insns_for_switch_stmt): Likewise. Also extend operands if
13379         necessary.
13380         (gen_hsa_clrsb): Likewise.
13381         (gen_hsa_ffs): Likewise.
13382         (gen_hsa_divmod): Extend operands and convert to dest type if
13383         necessary.
13384         (gen_hsa_atomic_for_builtin): Change type of op to hsa_op_with_type.
13386 2017-10-08  Segher Boessenkool  <segher@kernel.crashing.org>
13388         * config/rs6000/rs6000.md (conditional branch): Clean up formatting.
13389         Remove empty default arguments.  Use a brace block as output
13390         statement.
13391         (conditional return): Ditto.
13392         (jump): Ditto.
13393         (indirect_jump): Ditto.  Use b%T0 instead of bctr/blr.
13394         (tablejump, tablejumpsi, tablejumpdi, *tablejump<mode>_internal1):
13395         Ditto.
13396         (group_ending_nop): Ditto.
13397         (doloop_end): Ditto.
13398         (ctr<mode>, ctr<mode>_internal1, ctr<mode>_internal2): Ditto.
13399         (splitters for those): Ditto.
13401 2017-10-08  Segher Boessenkool  <segher@kernel.crashing.org>
13403         * config/rs6000/rs6000-string.c (expand_strncmp_align_check): Invert
13404         a conditional jump (and the compare for it) so that pc_rtx is the
13405         last operand.
13406         * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Adjust
13407         for the deleted and renamed ctr<mode>_internal[234] patterns.
13408         * config/rs6000/rs6000.md: Delete second conditional branch pattern.
13409         Delete second conditional return pattern.
13410         (ctr<mode>_internal2): Delete this second bdnz pattern.
13411         (ctr<mode>_internal3): Rename to ctr<mode>_internal2.
13412         (ctr<mode>_internal4): Delete this second bdz pattern.
13414 2017-10-08  Eric Botcazou  <ebotcazou@adacore.com>
13416         * tree-outof-ssa.h (ssaexpand): Add partitions_for_undefined_values.
13417         (always_initialized_rtx_for_ssa_name_p): New predicate.
13418         * tree-outof-ssa.c (remove_ssa_form): Initialize new field of SA.
13419         (finish_out_of_ssa): Free new field of SA.
13420         * tree-ssa-coalesce.h (get_undefined_value_partitions): Declare.
13421         * tree-ssa-coalesce.c: Include tree-ssa.h.
13422         (get_parm_default_def_partitions): Remove extern keyword.
13423         (get_undefined_value_partitions): New function.
13424         * expr.c (expand_expr_real_1) <expand_decl_rtl>: For a SSA_NAME, do
13425         not set SUBREG_PROMOTED_VAR_P on the sub-register if it may contain
13426         uninitialized bits.
13427         * loop-iv.c (iv_get_reaching_def): Disqualify all subregs.
13429 2017-10-08  Eric Botcazou  <ebotcazou@adacore.com>
13431         * builtins.def (BUILT_IN_SETJMP): Revert latest change.
13433 2017-10-08  Jan Hubicka  <hubicka@ucw.cz>
13435         * config/i386/i386.c (ix86_expand_set_or_movmem): Disable 512bit loops
13436         for targets that preffer 128bit.
13438 2017-10-08  Jan Hubicka  <hubicka@ucw.cz>
13440         * config/i386/i386.c (has_dispatch): Disable for Ryzen.
13442 2017-10-08  Olivier Hainque  <hainque@adacore.com>
13444         * config/arm/arm.c (arm_set_return_address): Use MEM_VOLATILE_P
13445         on the target mem instead of RTX_FRAME_RELATED_P on the insn to
13446         prevent DSE.
13447         (thumb_set_return_address): Likewise.
13449 2017-10-08  Olivier Hainque  <hainque@adacore.com>
13451         * common/config/arm/arm-common.c (arm_except_unwind_info):
13452         Handle DWARF2_UNWIND_INFO.
13454 2017-10-07  Michael Collison <michael.collison@arm.com>
13456         * config/aarch64/aarch64.md (*aarch64_reg_<optab>_minus<mode>3):
13457         New pattern.
13459 2017-10-07  Eric Botcazou  <ebotcazou@adacore.com>
13461         * builtins.def (BUILT_IN_SETJMP): Declare as library builtin instead
13462         of GCC builtin if DONT_USE_BUILTIN_SETJMP is defined.
13463         * except.c (sjlj_emit_function_enter): If DONT_USE_BUILTIN_SETJMP is
13464         defined, force the creation of a new block for a dispatch label.
13466 2017-10-07  Jan Hubicka  <hubicka@ucw.cz>
13468         * invoke.texi (Wsuggest-attribute=cold): Document.
13469         * common.opt (Wsuggest-attribute=cold): New
13470         * ipa-pure-const.c (warn_function_cold): New function.
13471         * predict.c (compute_function_frequency): Use it.
13472         * predict.h (warn_function_cold): Declare.
13474 2017-10-06  Jan Hubicka  <hubicka@ucw.cz>
13476         * tree-switch-conversion.c (do_jump_if_equal, emit_cmp_and_jump_insns):
13477         Update profile.
13479 2017-10-06  Martin Liska  <mliska@suse.cz>
13481         * sanopt.c (struct sanopt_tree_triplet_hash): Remove inline
13482         keyword for member functions.
13483         (struct sanopt_tree_couple): New struct.
13484         (struct sanopt_tree_couple_hash): New function.
13485         (struct sanopt_ctx): Add new hash_map.
13486         (has_dominating_ubsan_ptr_check): New function.
13487         (record_ubsan_ptr_check_stmt): Likewise.
13488         (maybe_optimize_ubsan_ptr_ifn): Likewise.
13489         (sanopt_optimize_walker): Handle IFN_UBSAN_PTR.
13490         (pass_sanopt::execute): Handle also SANITIZE_POINTER_OVERFLOW.
13492 2017-10-06  Sudakshina Das  <sudi.das@arm.com>
13494         PR target/82440
13495         * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Only call
13496         aarch64_simd_valid_immediate on CONST_VECTORs.
13497         (aarch64_reg_or_bic_imm): Likewise.
13499 2017-10-06  Wilco Dijkstra  <wdijkstr@arm.com>
13501         PR rtl-optimization/82396
13502         * haifa-sched.c (ready_sort_real): Disable qsort checking.
13504 2017-10-06  Sebastian Pop  <sebpop@gmail.com>
13506         * graphite-dependences.c (scop_get_reads): Move code to...
13507         (scop_get_must_writes): Move code to...
13508         (scop_get_may_writes): Move code to...
13509         (scop_get_reads_and_writes): ... here.
13510         (scop_get_dependences): Call scop_get_reads_and_writes.
13512 2017-10-06  Jakub Jelinek  <jakub@redhat.com>
13514         PR tree-optimization/82434
13515         * fold-const.h (can_native_encode_type_p,
13516         can_native_encode_string_p): Remove.
13517         * fold-const.c (native_encode_int): Formatting fixes.  If ptr is NULL,
13518         don't encode anything, just return what would be otherwise returned.
13519         (native_encode_fixed, native_encode_complex, native_encode_vector):
13520         Likewise.
13521         (native_encode_string): Likewise.  Inline by hand
13522         can_native_encode_string_p.
13523         (can_native_encode_type_p): Remove.
13524         (can_native_encode_string_p): Remove.
13525         * tree-vect-stmts.c (vectorizable_store): Instead of testing just
13526         STRING_CSTs using can_native_encode_string_p, test all
13527         CONSTANT_CLASS_P values using native_encode_expr with NULL ptr.
13528         * gimple-ssa-store-merging.c (encode_tree_to_bitpos): Remove last
13529         argument from native_encode_expr.
13530         (rhs_valid_for_store_merging_p): Use native_encode_expr with NULL ptr.
13531         (pass_store_merging::execute): Don't unnecessarily look for 3 stmts,
13532         but just 2.
13534 2017-10-06  Richard Biener  <rguenther@suse.de>
13536         PR tree-optimization/82397
13537         * tree-vect-data-refs.c (dr_group_sort_cmp): Do not use
13538         operand_equal_p but rely on data_ref_compare_tree for detecting
13539         equalities.
13540         (vect_analyze_data_ref_accesses): Use data_ref_compare_tree
13541         to match up with dr_group_sort_cmp.
13543 2017-10-06  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
13545         PR target/82322
13546         * config/s390/s390-builtins.def (s390_vfi): Define new overloaded
13547         builtin.
13548         * config/s390/s390-builtin-types.def: Regenerate.
13550 2017-10-06  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
13552         PR target/82317
13553         * config/s390/s390-builtin-types.def: Regenerate.
13554         * config/s390/s390-builtins.def (s390_vfmaxdb_4, s390_vfmindb_4):
13555         Change flag from B_VXE to B_VX.
13556         (s390_vec_min_dbl): Remove B_VXE flag.
13558 2017-10-06  Richard Biener  <rguenther@suse.de>
13560         * graphite-isl-ast-to-gimple.c: Include ssa.h and tree-ssa.h.
13561         (translate_isl_ast_to_gimple::translate_pending_phi_nodes,
13562         translate_isl_ast_to_gimple::is_valid_rename,
13563         translate_isl_ast_to_gimple::get_rename,
13564         translate_isl_ast_to_gimple::get_def_bb_for_const,
13565         translate_isl_ast_to_gimple::get_new_name,
13566         translate_isl_ast_to_gimple::collect_all_ssa_names,
13567         translate_isl_ast_to_gimple::copy_loop_phi_args,
13568         translate_isl_ast_to_gimple::collect_all_ssa_names,
13569         translate_isl_ast_to_gimple::copy_loop_phi_args,
13570         translate_isl_ast_to_gimple::copy_loop_phi_nodes,
13571         translate_isl_ast_to_gimple::add_close_phis_to_merge_points,
13572         translate_isl_ast_to_gimple::add_close_phis_to_outer_loops,
13573         translate_isl_ast_to_gimple::copy_loop_close_phi_args,
13574         translate_isl_ast_to_gimple::copy_loop_close_phi_nodes,
13575         translate_isl_ast_to_gimple::copy_cond_phi_args,
13576         translate_isl_ast_to_gimple::copy_cond_phi_nodes,
13577         translate_isl_ast_to_gimple::edge_for_new_close_phis,
13578         translate_isl_ast_to_gimple::add_phi_arg_for_new_expr,
13579         translate_isl_ast_to_gimple::rename_uses,
13580         translate_isl_ast_to_gimple::rename_all_uses): Remove.
13581         (translate_isl_ast_to_gimple::get_rename_from_scev): Simplify.
13582         (set_rename_for_each_def): Likewise.
13583         (graphite_copy_stmts_from_block): Handle debug stmt resetting
13584         here.  Handle rewriting SCEV analyzable uses here.
13585         (copy_bb_and_scalar_dependences): Generate code for PHI
13586         copy-in/outs.
13587         (graphite_regenerate_ast_isl): Adjust.
13588         * graphite-scop-detection.c (trivially_empty_bb_p): Move to sese.[ch].
13589         (add_write, add_read): New functions.
13590         (build_cross_bb_scalars_def): Use it and simplify.
13591         (build_cross_bb_scalars_use): Likewise.
13592         (graphite_find_cross_bb_scalar_vars): Inline into...
13593         (try_generate_gimple_bb): ...here.  Add dependences for PHIs,
13594         simulating out-of-SSA.  Compute liveout and add dependencies.
13595         (build_scops): Force an empty entry block.
13596         * sese.h (sese_info_t::liveout, sese_info_t::debug_liveout): New
13597         members.
13598         (sese_build_liveouts): Declare.
13599         (sese_trivially_empty_bb_p): Likewise.
13600         * sese.c (sese_build_liveouts_bb): Properly handle PHIs,
13601         compute liveout and debug_liveout.
13602         (sese_bad_liveouts_use): Remove.
13603         (sese_reset_debug_liveouts_bb): Likewise.
13604         (sese_reset_debug_liveouts): Rewrite in terms of debug_liveout.
13605         (sese_build_liveouts): Build liveout and debug_liveout and store
13606         it in region.
13607         (new_sese_info): Adjust.
13608         (free_sese_info): Likewise.
13609         (sese_insert_phis_for_liveouts): Reset debug stmts from here,
13610         do not build liveout here.
13611         (move_sese_in_condition): Adjust region entry.
13612         (scev_analyzable_p): Match up with chrec_apply requirements.
13613         (sese_trivially_empty_bb_p): New.
13614         * tree-into-ssa.c (get_reaching_def): Properly support generating
13615         default-defs for incremental rewrite of anonymous names.
13617 2017-10-06  Richard Biener  <rguenther@suse.de>
13619         * graphite-sese-to-poly.c (extract_affine): For casts increasing
13620         precision do not perform modulo reduction.
13622 2017-10-06  Richard Biener  <rguenther@suse.de>
13624         PR tree-optimization/82436
13625         * tree-vect-slp.c (vect_supported_load_permutation_p): More
13626         conservatively choose the vectorization factor when checking
13627         whether we can perform the required load permutation.
13628         (vect_transform_slp_perm_load): Assert when we may not fail.
13630 2017-10-05  Segher Boessenkool  <segher@kernel.crashing.org>
13632         * config/rs6000/sysv4.h (SUBTARGET_OVERRIDE_OPTIONS): Correct error
13633         message for incompatible -msdata=* and -mcall-* options.
13635 2017-10-05  Jan Hubicka <hubicka@ucw.cz>
13637         * config/i386/i386.c (ia32_multipass_dfa_lookahead): Default to issue
13638         rate for post-reload scheduling.
13640 2017-10-05  Tamar Christina  <tamar.christina@arm.com>
13642         * doc/sourcebuild.texi (vect_sizes_16B_8B, vect_sizes_32B_16B): New.
13644 2017-10-05  Jan Hubicka <hubicka@ucw.cz>
13646         * config/i386/i386.c (znver1_cost): Set branch_cost to 3 (instead of 2)
13647         to improve monte carlo in scimark.
13649 2017-10-05  Jan Hubicka <hubicka@ucw.cz>
13651         * config/i386/i386.c (ix86_size_cost, i386_cost, i486_cost,
13652         pentium_cost, lakemont_cost, pentiumpro_cost, geode_cost, k6_cost,
13653         athlon_cost, k8_cost, amdfam10_cost, btver1_cost, btver2_cost,
13654         pentium4_cost, nocona_cost): Set reassociation width to 1.
13655         (bdver1_cost, bdver2_cost, bdver3_cost, bdver4_cost): Set reassociation
13656         width to 2 for fp operations and 1 otherwise.
13657         (znver1_cost): Set scalar reassoc width to 4 and vector to 3 and 6
13658         for int and fp.
13659         (atom_cost): Set reassociation width to 2.
13660         (slm_cost, generic_cost): Set fp reassociation width
13661         to 2 and 1 otherwise.
13662         (intel_cost): Set fp reassociation width to 4 and 1 otherwise.
13663         (core_cost): Set fp reassociation width to 4 and vector to 2.
13664         (ix86_reassociation_width): Rewrite using cost table; special case
13665         plus/minus on Zen; honor X86_TUNE_SSE_SPLIT_REGS
13666         and TARGET_AVX128_OPTIMAL.
13667         * config/i386/i386.h (processor_costs): Add
13668         reassoc_int, reassoc_fp, reassoc_vec_int, reassoc_vec_fp.
13669         (TARGET_VECTOR_PARALLEL_EXECUTION, TARGET_REASSOC_INT_TO_PARALLEL,
13670         TARGET_REASSOC_FP_TO_PARALLEL): Remove.
13671         * x86-tune.def (X86_TUNE_REASSOC_INT_TO_PARALLEL): Remove.
13672         (X86_TUNE_REASSOC_FP_TO_PARALLEL): Remove.
13673         (X86_TUNE_VECTOR_PARALLEL_EXECUTION):  Remove.
13675 2017-10-05  Nathan Sidwell  <nathan@acm.org>
13677         * doc/invoke.texi (Wparentheses): Document C++ MVP behaviour.
13679 2017-10-05  Tamar Christina  <tamar.christina@arm.com>
13681         * config/arm/arm.c (arm_test_fpu_data): New.
13682         (arm_run_selftests): Call arm_test_fpu_data.
13684 2017-10-04  Nathan Sidwell  <nathan@acm.org>
13686         * toplev.c (toplev::main): Remove excess parens on pretty_printer
13687         decl.
13688         * caller-save.c (insert_save): Remove excess parens on TO_SAVE parm.
13690 2017-10-04  Sudakshina Das  <sudi.das@arm.com>
13692         * config/aarch64/aarch64-protos.h (enum simd_immediate_check): New
13693         check type for aarch64_simd_valid_immediate.
13694         (aarch64_output_simd_mov_immediate): Update prototype.
13695         (aarch64_simd_valid_immediate): Update prototype.
13696         * config/aarch64/aarch64-simd.md (orr<mode>3): modified pattern to add
13697         support for ORR-immediate.
13698         (and<mode>3): modified pattern to add support for BIC-immediate.
13699         * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Function
13700         now checks for valid immediate for BIC and ORR based on new enum
13701         argument.
13702         (aarch64_output_simd_mov_immediate): Function now used to output
13703         BIC/ORR imm as well based on new enum argument.
13704         * config/aarch64/constraints.md (Do): New vector immediate constraint.
13705         (Db) : Likewise.
13706         * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): New predicate.
13707         (aarch64_reg_or_bic_imm): Likewise.
13709 2017-10-04  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
13711         * config/s390/vx-builtins.md ("vec_mergeh<mode>")
13712         ("vec_mergel<mode>"): Change mode iterator to V_128_NOSINGLE.
13714 2017-10-04  Wilco Dijkstra  <wdijkstr@arm.com>
13716         Revert r253399:
13718         PR rtl-optimization/82396
13719         * haifa-sched.c (autopref_multipass_init): Simplify
13720         initialization.
13721         (autopref_rank_data): Simplify sort order.
13722         * sched-int.h (autopref_multipass_data_): Remove
13723         multi_mem_insn_p, min_offset and max_offset.
13725 2017-10-04  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
13727         * doc/sourcebuild.texi: Document vect_peeling_profitable.
13729 2017-10-04  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
13731         * doc/sourcebuild.texi: Document vect_intdouble_cvt and
13732         vect_doubleint_cvt.
13734 2017-10-04  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
13736         * doc/sourcebuild.texi: Document vect_long_mult.
13738 2017-10-04  Richard Sandiford  <richard.sandiford@linaro.org>
13740         PR tree-optimization/82413
13741         * fold-const.c (build_range_check): Use widest_int when comparing
13742         the maximum ETYPE value with HIGH.
13744 2017-10-04  Wilco Dijkstra  <wdijkstr@arm.com>
13746         PR rtl-optimization/82396
13747         * haifa-sched.c (autopref_multipass_init): Simplify
13748         initialization.
13749         (autopref_rank_data): Simplify sort order.
13750         * sched-int.h (autopref_multipass_data_): Remove
13751         multi_mem_insn_p, min_offset and max_offset.
13753 2017-10-04  Jakub Jelinek  <jakub@redhat.com>
13755         PR tree-optimization/82381
13756         * tree-ssa-reassoc.c (sort_by_operand_rank): Check for different
13757         oeN->rank first.  Return 1 or -1 if one op is SSA_NAME and the other
13758         is not.
13760         PR tree-optimization/82374
13761         * omp-low.c (create_omp_child_function): Copy DECL_ATTRIBUTES,
13762         DECL_FUNCTION_SPECIFIC_OPTIMIZATION,
13763         DECL_FUNCTION_SPECIFIC_TARGET and DECL_FUNCTION_VERSIONED from
13764         current_function_decl to the new decl.
13766 2017-10-03  Michael Meissner  <meissner@linux.vnet.ibm.com>
13768         * config/rs6000/rs6000-builtin.def (BU_FLOAT128_2_HW): Define new
13769         helper macro for IEEE float128 hardware built-in functions.
13770         (SQRTF128_ODD): Add built-in functions with the round-to-odd
13771         semantics.
13772         (TRUNCF128_ODD): Likewise.
13773         (ADDF128_ODD): Likewise.
13774         (SUBF128_ODD): Likewise.
13775         (MULF128_ODD): Likewise.
13776         (DIVF128_ODD): Likewise.
13777         (FMAF128_ODD): Likewise.
13778         * config/rs6000/rs6000.md (UNSPEC_ROUND_TO_ODD): Rename to
13779         UNSPEC_TRUNC_ROUND_TO_ODD.
13780         (UNSPEC_TRUNC_ROUND_TO_ODD): Likewise.
13781         (UNSPEC_ADD_ROUND_TO_ODD): New unspec codes for the IEEE 128-bit
13782         floating point round to odd instructions.
13783         (UNSPEC_SUB_ROUND_TO_ODD): Likewise.
13784         (UNSPEC_MUL_ROUND_TO_ODD): Likewise.
13785         (UNSPEC_DIV_ROUND_TO_ODD): Likewise.
13786         (UNSPEC_FMA_ROUND_TO_ODD): Likewise.
13787         (UNSPEC_SQRT_ROUND_TO_ODD): Likewise.
13788         (trunc<mode>sf2_hw): Change the truncate with round to odd
13789         expansion to use UNSPEC_TRUNC_ROUND_TO_ODD.
13790         (add<mode>3_odd): Add insns for IEEE 128-bit floating point round
13791         to odd hardware instructions.
13792         (sub<mode>3_odd): Likewise.
13793         (mul<mode>3_odd): Likewise.
13794         (div<mode>3_odd): Likewise.
13795         (sqrt<mode>2_odd): Likewise.
13796         (fma<mode>4_odd): Likewise.
13797         (fms<mode>4_odd): Likewise.
13798         (nfma<mode>4_odd): Likewise.
13799         (nfms<mode>4_odd): Likewise.
13800         (trunc<mode>df2_odd): Change the truncate with round to odd
13801         expansion to use UNSPEC_TRUNC_ROUND_TO_ODD.  Add a generator
13802         function.
13803         * doc/extend.texi (PowerPC built-in functions): Update documentation
13804         for existing IEEE float128-bit built-in functions.  Add built-in
13805         functions that generate the IEEE 128-bit floating point round to
13806         odd instructions.
13808 2017-10-03  Segher Boessenkool  <segher@kernel.crashing.org>
13810         PR rtl-optimization/77729
13811         * simplify-rtx.c (simplify_binary_operation_1): Delete the (X&C1)|C2
13812         to (X&(C1&~C2))|C2 transformations.
13814 2017-10-03  Martin Jambor  <mjambor@suse.cz>
13816         PR tree-optimization/82363
13817         * tree-sra.c (propagate_subaccesses_across_link): In unrecoverable
13818         mismatch, mark lacc written regardless of racc.
13820 2017-10-03  Jakub Jelinek  <jakub@redhat.com>
13822         PR tree-optimization/82381
13823         * tree-ssa-reassoc.c (sort_by_operand_rank): Don't check
13824         stmt_to_insert nor wheather SSA_NAMEs are default defs.
13825         Return 1 or -1 if one of bba and bbb is NULL. If bb_rank is equal,
13826         fallthrough into reassoc_stmt_dominates_stmt_p.
13828         PR target/82386
13829         * combine.c (combine_instructions): Don't combine in unreachable
13830         basic blocks.
13832 2017-10-02  Peter Bergner  <bergner@vnet.ibm.com>
13834         PR target/80210
13835         * config/rs6000/rs6000.c (rs6000_option_override_internal): Rewrite
13836         function to not use the have_cpu variable.  Do not set cpu_index,
13837         rs6000_cpu_index or rs6000_tune_index if we end up using TARGET_DEFAULT
13838         or the default cpu.
13839         (rs6000_valid_attribute_p): Remove duplicate initializations of
13840         old_optimize and func_optimize.
13841         (rs6000_pragma_target_parse): Call rs6000_activate_target_options ().
13842         (rs6000_activate_target_options): Make global.
13843         * config/rs6000/rs6000-protos.h (rs6000_activate_target_options): Add
13844         prototype.
13846 2017-10-02  Jakub Jelinek  <jakub@redhat.com>
13848         * tree-dfa.c (get_ref_base_and_extent): Set *pmax_size to -1
13849         if *poffset + *pmax_size overflows in HOST_WIDE_INT.
13850         Set *poffset to 0 and *psize and *pmax_size to -1 if
13851         *poffset + *psize overflows in HOST_WIDE_INT.
13853         PR tree-optimization/82387
13854         PR tree-optimization/82388
13855         PR tree-optimization/82389
13856         * tree-ssa-dse.c (dse_classify_store): Test byte_tracking_enabled
13857         instead of live_bytes non-NULL.
13859 2017-10-02  Georg-Johann Lay  <avr@gjlay.de>
13861         PR target/41076
13862         * confg/avr/avr.md (*iorhi3.ashift8-ext.zerox): Add "r,r,0"
13863         alternative.
13865 2017-10-02  Richard Biener  <rguenther@suse.de>
13867         * graphite-isl-ast-to-gimple.c (set_codegen_error): With
13868         -fchecking and --param graphite-allow-codegen-errors=0 ICE.
13869         * params.def (PARAM_GRAPHITE_ALLOW_CODEGEN_ERRORS): New param.
13871 2017-10-02  Richard Sandiford  <richard.sandiford@linaro.org>
13873         * tree.h (wi::int_traits <const_tree>::decompose): Assert that the
13874         requested precision matches the type's.
13875         * calls.c (alloc_max_size): Calculate the new candidate size as
13876         a widest_int and use wi::to_widest when comparing it with the
13877         current candidate size.
13878         * gimple-ssa-warn-alloca.c (pass_walloca::execute): Compare with
13879         zero rather than integer_zero_node.
13880         * match.pd: Check for a no-op conversion before using wi::add
13881         rather than after.  Use tree_to_uhwi when summing small shift
13882         counts into an unsigned int.
13884 2017-10-02  Richard Sandiford  <richard.sandiford@linaro.org>
13885             Alan Hayward  <alan.hayward@arm.com>
13886             David Sherwood  <david.sherwood@arm.com>
13888         PR target/71307
13889         * config/aarch64/aarch64.h (POINTER_AND_FP_REGS): New reg class.
13890         (REG_CLASS_NAMES, REG_CLASS_CONTENTS): Update accordingly.
13891         * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
13892         POINTER_AND_FP_REGS.
13894 2017-10-02  Richard Biener  <rguenther@suse.de>
13896         PR tree-optimization/82355
13897         * graphite-isl-ast-to-gimple.c (build_iv_mapping): Also build
13898         a mapping for the enclosing loop but avoid generating one for
13899         the loop tree root.
13900         (copy_bb_and_scalar_dependences): Remove premature codegen
13901         error on PHIs in blocks duplicated into multiple places.
13902         * graphite-scop-detection.c
13903         (scop_detection::stmt_has_simple_data_refs_p): For a loop not
13904         in the region use it as loop and nest to analyze the DR in.
13905         (try_generate_gimple_bb): Likewise.
13906         * graphite-sese-to-poly.c (extract_affine_chrec): Adjust.
13907         (add_loop_constraints): For blocks in a loop not in the region
13908         create a dimension with a single iteration.
13909         * sese.h (gbb_loop_at_index): Remove assert.
13911 2017-10-01  Kevin Buettner  <kevinb@redhat.com>
13913         * omp-expand.c (adjust_context_scope): New function.
13914         (expand_parallel_call): Call adjust_context_scope.
13916 2017-10-01  Jeff Law  <law@redhat.com>
13918         * tree-ssa-dom.c (optimize_stmt): Make this a method within the
13919         dom_opt_dom_walker class with direct access to private members.
13920         Add comments.  Call test_for_singularity.
13921         (dom_opt_dom_walker::before_dom_children): Corresponding changes.
13922         (dom_opt_dom_walker::after_dom_children): Do not lazily initialize
13923         m_dummy_cond anymore.
13924         (class dom_opt_dom_walker): Initialize m_dummy_cond member in the
13925         class ctor.
13926         (pass_dominator:execute): Build the dummy_cond here and pass it
13927         to the dom_opt_dom_walker ctor.
13928         (test_for_singularity): New function.
13930 2017-09-30  Krister Walfridsson  <krister.walfridsson@gmail.com>
13931             Maya Rashish  <coypu@sdf.org>
13933         * config.gcc (*-*-netbsd*): New variable nbsd_tm_file containing
13934         netbsd.h, netbsd-stdint.h, and netbsd-elf.h.
13935         (alpha*-*-netbsd*) Use nbsd_tm_file.
13936         (arm*-*-netbsdelf*) Likewise.
13937         (i[34567]86-*-netbsdelf*) Likewise.
13938         (x86_64-*-netbsd*) Likewise.
13939         (mips*-*-netbsd*) Likewise.
13940         (powerpc-*-netbsd*) Likewise.
13941         (sh*-*-netbsd*) Likewise.
13942         (sparc-*-netbsdelf*) Likewise.
13943         (sparc64-*-netbsd*) Likewise.
13944         (m68k*-*-netbsdelf*) Use nbsd_tm_file and add CHAR_FAST8/SHORT_FAST16
13945         to tm_defines.
13946         (vax-*-netbsdelf*) Likewise.
13947         * config/netbsd-stdint.h (INT_FAST8_TYPE): Check CHAR_FAST8.
13948         (UINT_FAST8_TYPE) Likewise.
13949         (INT_FAST16_TYPE) Check CHAR_FAST16.
13950         (UINT_FAST16_TYPE) Likewise.
13952 2017-09-30  Jakub Jelinek  <jakub@redhat.com>
13954         PR target/82361
13955         * config/i386/i386.md
13956         (TARGET_USE_8BIT_IDIV zext divmodsi4 splitter): New define_split.
13957         (divmodsi4_zext_1, divmodsi4_zext_2, *divmodsi4_zext_1,
13958         *divmodsi4_zext_2): New define_insn_and_split.
13959         (*divmodsi4_noext_zext_1, *divmodsi4_noext_zext_2): New define_insn.
13960         (TARGET_USE_8BIT_IDIV zext udivmodsi4 splitter): New define_split.
13961         (udivmodsi4_zext_1, udivmodsi4_zext_2, *udivmodsi4_zext_1,
13962         *udivmodsi4_zext_2, *udivmodsi4_pow2_zext_1, *udivmodsi4_pow2_zext_2):
13963         New define_insn_and_split.
13964         (*udivmodsi4_noext_zext_1, *udivmodsi4_noext_zext_2): New define_insn.
13965         * config/i386/i386.c (ix86_split_idivmod): Handle operands[0] or
13966         operands[1] having DImode when mode is SImode.
13968         * config/i386/i386.c (ix86_split_idivmod): Use mode instead of
13969         always SImode for DIV and MOD in REG_EQUAL notes.
13971 2017-09-29  Yury Gribov  <tetra2005@gmail.com>
13973         PR middle-end/82319
13974         * match.pd: Fix handling of NaNs in pattern.
13976 2017-09-29  Jeff Law  <law@redhat.com>
13978         * sbitmap.c (bitmap_bit_in_range_p): New function.
13979         * sbitmap.h (bitmap_bit_in_range_p): Prototype.
13980         * tree-ssa-dse.c (live_bytes_read): New function.
13981         (dse_classify_store): Ignore reads of dead bytes.
13983         * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Fix
13984         typos and whitespace errors.
13985         * config/i386/predicates.md (address_no_seg_operand): Likewise.
13986         * config/s390/s390.c (s390_emit_prologue): Likewise.
13988 2017-09-29  Vladimir Makarov  <vmakarov@redhat.com>
13990         PR target/81481
13991         * ira-costs.c (scan_one_insn): Don't take into account PIC equiv
13992         with a symbol for LRA.
13994 2017-09-29  Vladimir Makarov  <vmakarov@redhat.com>
13996         PR rtl-optimization/82338
13997         * lra-constraints.c (inherit_in_ebb): Check usage_insns check.
13999 2017-09-29  Alexander Monakov  <amonakov@ispras.ru>
14001         * genmodes.c (calc_wider_mode): Suppress qsort macro.
14002         * system.h [CHECKING_P] (qsort): Redirect to qsort_chk.
14003         (qsort_chk): Declare.
14004         * vec.c [CHECKING_P] (qsort_chk_error): New static function.
14005         (qsort_chk): New function.
14007 2017-09-29  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
14009         PR tree-optimization/82337
14010         * gimple-ssa-strength-reduction.c (find_phi_def): Don't record a
14011         phi definition if the PHI result appears in an abnormal PHI.
14012         (find_basis_for_base_expr): Don't record a basis if the LHS of the
14013         basis appears in an abnormal PHI.
14015 2017-09-29  Richard Biener  <rguenther@suse.de>
14017         * graphite-isl-ast-to-gimple.c
14018         (translate_isl_ast_to_gimple::set_codegen_error): New function.
14019         (binary_op_to_tree): Use it.
14020         (get_rename_from_scev): Likewise.
14021         (copy_loop_phi_nodes): Likewise.
14022         (copy_bb_and_scalar_dependences): Likewise.
14023         (translate_pending_phi_nodes): Likewise.
14025 2017-09-29  Jakub Jelinek  <jakub@redhat.com>
14027         PR target/82339
14028         * config/i386/i386.md (*movdi_internal peephole2): New -Os peephole
14029         for movabsq $(i32 << shift), r64.
14031 2017-09-28  Uros Bizjak  <ubizjak@gmail.com>
14033         * config/i386/i386.c (ix86_print_operand_address_as): Do not check
14034         index when encoding %esp as %rsp to avoid 0x67 prefix.
14036 2017-09-28  Sergey Shalnov  <Sergey.Shalnov@intel.com>
14038         * config/i386/i386.md (*movsf_internal, *movdf_internal):
14039         Return 256-bit AVX modes for TARGET_PREFER_AVX256.
14041 2017-09-28  Thomas Preud'homme  <thomas.preudhomme@arm.com>
14043         * config/arm/arm.c (arm_option_override): Forbid ARMv8-M Security
14044         Extensions with more than 16 double VFP registers.
14045         (cmse_nonsecure_entry_clear_before_return): Remove second entry of
14046         to_clear_mask and all code related to it.  Replace the remaining
14047         entry by a sbitmap and adapt code accordingly.
14049 2017-09-28  Henry Linjamäki  <henry.linjamaki@parmance.com>
14051         * brig-builtins.def: Change pure attributes to const.
14053 2017-09-28  Joseph Myers  <joseph@codesourcery.com>
14055         * config.gcc (default_gnu_indirect_function): Default to yes for
14056         sparc*-*-linux* with glibc.
14058 2017-09-28  Joseph Myers  <joseph@codesourcery.com>
14060         * config/aarch64/aarch64.c (aarch64_elf_asm_constructor)
14061         (aarch64_elf_asm_destructor): Pass SECTION_NOTYPE to get_section
14062         when creating .init_array and .fini_array sections with priority
14063         specified.
14065 2017-09-27  Christophe Lyon  <christophe.lyon@linaro.org>
14067         PR target/71727
14068         * config/aarch64/aarch64.c
14069         (aarch64_builtin_support_vector_misalignment): Always return false
14070         when misalignment is unknown.
14072 2017-09-27  Kelvin Nilsen  <kelvin@gcc.gnu.org>
14074         * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Revise
14075         this function to return false if the definition used by the swap
14076         instruction is artificial, or if the memory address from which the
14077         constant value is loaded is not represented by a base address held
14078         in a register or if the base address register is a frame or stack
14079         pointer.  Additionally, return false if the base address of the
14080         loaded constant is a SYMBOL_REF but is not considered to be a
14081         constant.
14082         (replace_swapped_load_constant): New function.
14083         (rs6000_analyze_swaps): Add a new pass to replace a swap of a
14084         loaded constant vector with a load of a swapped constant vector.
14086 2017-09-27  Carl Love  <cel@us.ibm.com>
14088         * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
14089         (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
14090         * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
14091         fctiw instruction.
14093 2017-09-27  Alexander Monakov  <amonakov@ispras.ru>
14095         * haifa-sched.c (autopref_rank_for_schedule): Order 'irrelevant' insns
14096         first, always call autopref_rank_data otherwise.
14098 2017-09-27  Richard Biener  <rguenther@suse.de>
14100         * graphite-scop-detection.c (find_scop_parameters): Move
14101         loop bound handling ...
14102         (gather_bbs::before_dom_children): ... here, avoiding the need
14103         to build scop_info->loop_nest.
14104         (record_loop_in_sese): Remove.
14105         * sese.h (sese_info_t::loop_nest): Remove.
14106         * sese.c (new_sese_info): Do not allocate loop_nest.
14107         (free_sese_info): Do not free loop_nest.
14109 2017-09-27  Jakub Jelinek  <jakub@redhat.com>
14111         PR c++/82159
14112         * gimplify.c (gimplify_modify_expr): Don't optimize away zero sized
14113         lhs from calls if the lhs has addressable type.
14115 2017-09-27  Richard Biener  <rguenther@suse.de>
14117         * graphite.h (scop::max_alias_set): New member.
14118         * graphite-scop-detection.c: Remove references to non-existing
14119         --param in comments.
14120         (build_alias_sets): Record the maximum alias set used for drs.
14121         (build_scops): Support zero as unlimited for
14122         --param graphite-max-arrays-per-scop.
14123         * graphite-sese-to-poly.c (add_scalar_version_numbers): Remove
14124         and inline into ...
14125         (build_poly_sr_1): ... here.  Compute alias set based on the
14126         maximum alias set used for drs rather than
14127         PARAM_GRAPHITE_MAX_ARRAYS_PER_SCOP
14129 2017-09-27  Richard Biener  <rguenther@suse.de>
14131         * graphite-optimize-isl.c (get_schedule_for_node_st): Allow
14132         --param loop-block-tile-size=0 to disable tiling.
14134 2017-09-27  Richard Biener  <rguenther@suse.de>
14136         * doc/invoke.texi (graphite-max-bbs-per-function): Remove.
14137         (graphite-max-nb-scop-params): Document special value zero.
14138         * domwalk.h (dom_walker::STOP): New symbolical constant.
14139         (dom_walker::dom_walker): Add optional parameter for bb to
14140         RPO mapping.
14141         (dom_walker::~dom_walker): Declare.
14142         (dom_walker::before_dom_children): Document STOP return value.
14143         (dom_walker::m_user_bb_to_rpo): New member.
14144         (dom_walker::m_bb_to_rpo): Likewise.
14145         * domwalk.c (dom_walker::dom_walker): Compute bb to RPO
14146         mapping here if not provided by the user.
14147         (dom_walker::~dom_walker): Free bb to RPO mapping if not
14148         provided by the user.
14149         (dom_walker::STOP): Define.
14150         (dom_walker::walk): Do not compute bb to RPO mapping here.
14151         Support STOP return value from before_dom_children to stop
14152         walking.
14153         * graphite-optimize-isl.c (optimize_isl): If the schedule
14154         is the same still generate code if -fgraphite-identity
14155         or -floop-parallelize-all are given.
14156         * graphite-scop-detection.c: Include cfganal.h.
14157         (gather_bbs::gather_bbs): Get and pass through bb to RPO
14158         mapping.
14159         (gather_bbs::before_dom_children): Return STOP for BBs
14160         not in the region.
14161         (build_scops): Compute bb to RPO mapping and pass it to
14162         the domwalk.  Treat --param graphite-max-nb-scop-params=0
14163         as not limiting the number of params.
14164         * graphite.c (graphite_initialize): Remove limit on the
14165         number of basic-blocks in a function.
14166         * params.def (PARAM_GRAPHITE_MAX_BBS_PER_FUNCTION): Remove.
14167         (PARAM_GRAPHITE_MAX_NB_SCOP_PARAMS): Adjust to documented
14168         default value of 10.
14170 2017-09-26  Michael Meissner  <meissner@linux.vnet.ibm.com>
14172         * config/rs6000/vsx.md (peephole for optimizing move SF to GPR):
14173         Adjust code to eliminate needing to do the shift right 32-bits
14174         operation after XSCVDPSPN.
14176 2017-09-26  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
14178         * match.pd ((X / Y) == 0 -> X < Y): New pattern.
14179         ((X / Y) != 0 -> X >= Y): Likewise.
14181 2017-09-26  Carl Love  <cel@us.ibm.com>
14183         * config/rs6000/rs6000-c.c (P9V_BUILTIN_VEC_XL_LEN_R,
14184         P9V_BUILTIN_VEC_XST_LEN_R): Add support for builtins
14185         vector unsigned char vec_xl_len_r (unsigned char *, size_t);
14186         void vec_xst_len_r (vector unsigned char, unsigned char *, size_t);
14187         * config/rs6000/altivec.h (vec_xl_len_r, vec_xst_len_r): Add defines.
14188         * config/rs6000/rs6000-builtin.def (XL_LEN_R, XST_LEN_R): Add
14189         definitions and overloading.
14190         * config/rs6000/rs6000.c (altivec_expand_builtin): Add case
14191         statement for P9V_BUILTIN_XST_LEN_R.
14192         (altivec_init_builtins): Add def_builtin for P9V_BUILTIN_STXVLL.
14193         * config/rs6000/vsx.md (lxvll, stxvll, xl_len_r, xst_len_r): Add
14194         define_expand and define_insn for the instructions and builtins.
14195         * doc/extend.texi: Update the built-in documentation file for the new
14196         built-in functions.
14197         * config/rs6000/altivec.md (altivec_lvsl_reg, altivec_lvsr_reg): Add
14198         define_insn for the instructions
14200 2017-09-26  Krister Walfridsson  <krister.walfridsson@gmail.com>
14202         PR target/39570
14203         * gcc/config/netbsd-protos.h: New file.
14204         * gcc/config/netbsd.c: New file.
14205         * gcc/config/netbsd.h (SUBTARGET_INIT_BUILTINS): Define.
14206         * gcc/config/t-netbsd: New file.
14207         * gcc/config.gcc (tm_p_file): Add netbsd-protos.h.
14208         (tmake_file) Add t-netbsd.
14209         (extra_objs) Add netbsd.o.
14211 2017-09-26  Janus Weil  <janus@gcc.gnu.org>
14213         PR fortran/82143
14214         PR fortran/82324
14215         * doc/sourcebuild.texi: Document fortran_real_10 and fortran_real_16.
14217 2017-09-26  Michael Meissner  <meissner@linux.vnet.ibm.com>
14219         * config/rs6000/rs6000.md (extendsi<mode>2): Add a splitter to do
14220         sign extension from a vector register to a GPR by doing a 32-bit
14221         direct move and then an EXTSW.
14222         (extendsi<mode>2 splitter): Likewise.
14223         (movsi_from_sf): Adjust code to eliminate doing a 32-bit shift
14224         right or vector extract after doing XSCVDPSPN.  Use
14225         zero_extendsidi2 instead of p8_mfvsrd_4_disf to move the value to
14226         the GPRs.
14227         (movdi_from_sf_zero_ext): Likewise.
14228         (reload_gpr_from_vsxsf): Likewise.
14229         (p8_mfvsrd_4_disf): Delete, no longer used.
14230         (movsi_from_df): Optimize converting a DFmode to a SFmode, and
14231         then needing to move the SFmode to a GPR to use the XSCVDPSP
14232         instruction instead of FRSP and XSCVDPSPN.
14233         * config/rs6000/vsx.md (vsx_xscvspdp_scalar2): Move insn so that
14234         it is adjacent to the other XSCVSPDP insns.
14235         (vsx_xscvdpsp_scalar): Use "ww" constraint instead of "f" to allow
14236         SFmode to be in traditional Altivec registers.
14237         (vsx_xscvdpspn): Eliminate useless alternative constraint.
14238         (vsx_xscvspdpn): Likewise.
14239         (vsx_xscvspdpn_scalar): Likewise.
14241 2017-09-26  Martin Jambor  <mjambor@suse.cz>
14243         * tree-sra.c (compare_access_positions): Put integral types first,
14244         stabilize sorting of integral types, remove conditions putting
14245         non-full-precision integers last.
14246         (sort_and_splice_var_accesses): Disable scalarization if a
14247         non-integert would be represented by a non-full-precision integer.
14249 2017-09-26  Joseph Myers  <joseph@codesourcery.com>
14251         * config/microblaze/linux.h (TARGET_ASM_FILE_END): Likewise.
14252         * config/pa/pa.h (NEED_INDICATE_EXEC_STACK): Likewise.
14253         * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Likewise.
14254         * config/pa/pa.c (pa_hpux_file_end): Rename to pa_file_end.
14255         Define unconditionally, with [ASM_OUTPUT_EXTERNAL_REAL]
14256         conditionals inside the function instead of around it.  Call
14257         file_end_indicate_exec_stack if NEED_INDICATE_EXEC_STACK.
14258         (TARGET_ASM_FILE_END): Define unconditionally to pa_file_end.
14260 2017-09-26  Richard Biener  <rguenther@suse.de>
14262         * graphite-scop-detection.c (scop_detection::build_scop_depth): Rewrite,
14263         fold in ...
14264         (scop_detection::build_scop_breadth): ... this.  Removed.
14265         (scop_detection::loop_is_valid_in_scop): Fold into single caller.
14266         (scop_detection::harmful_stmt_in_bb): Likewise.
14267         (scop_detection::graphite_can_represent_stmt): Likewise.
14268         (scop_detection::loop_body_is_valid_scop): Likewise.  Remove recursion.
14269         (scop_detection::can_represent_loop): Remove recursion, fold in ...
14270         (scop_detection::can_represent_loop_1): ... this.  Removed.
14271         (scop_detection::harmful_loop_in_region): Simplify after inlining
14272         the above and remove more quadraticness.
14273         (build_scops): Adjust.
14274         * tree-data-ref.c (loop_nest_has_data_refs): Remove pointless
14275         quadraticness.
14277 2017-09-26  Jakub Jelinek  <jakub@redhat.com>
14279         PR target/82267
14280         * config/i386/i386.c (ix86_print_operand_address_as): Only test
14281         REGNO (base) == SP_REG if base is a REG.
14283         PR middle-end/35691
14284         * tree-ssa-reassoc.c (update_range_test): Dump r->exp each time
14285         if it is different SSA_NAME.
14286         (optimize_range_tests_cmp_bitwise): New function.
14287         (optimize_range_tests): Call it.
14289 2017-09-26  Richard Biener  <rguenther@suse.de>
14291         PR tree-optimization/82321
14292         * graphite.c (canonicalize_loop_closed_ssa): Properly check
14293         for the def being inside the loop.
14295 2017-09-26  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
14297         * config/s390/vx-builtins.md ("vmslg"): Add missing operand in
14298         assembler output.
14299         * config/s390/s390-builtins.def: Fix constraint on op4.
14301 2017-09-26  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
14303         * config/s390/s390.c (s390_expand_vec_compare): Use the new mode
14304         independent expanders.
14305         * config/s390/vector.md ("vec_cmpuneq", "vec_cmpltgt")
14306         ("vec_ordered", "vec_unordered"): New expanders.
14308 2017-09-26  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
14310         * config/s390/s390.c (s390_preferred_simd_mode): Return V4SFmode
14311         for SFmode.
14313 2017-09-26  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
14315         * config/s390/vector.md ("vec_unpacks_low_v16qi"): Rename to
14316         vec_unpacks_lo_v16qi.
14317         ("vec_unpacku_low_v16qi"): Rename to vec_unpacku_lo_v16qi.
14319 2017-09-26  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
14321         * config/s390/vector.md ("vec_unpacks_lo_v4sf")
14322         ("vec_unpacks_hi_v4sf", "vec_unpacks_lo_v2df")
14323         ("vec_unpacks_hi_v2df", "vec_pack_trunc_v2df"): New expanders.
14325 2017-09-26  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
14327         * config/s390/predicates.md ("const_shift_by_byte_operand"): New
14328         predicate.
14329         * config/s390/vector.md ("*vec_srb<mode>"): Change modes to V_128
14330         and V16QI.
14331         ("*vec_slb<mode>"): New insn pattern.
14332         ("vec_shr_<mode>"): New expander.
14333         * config/s390/vx-builtins.md ("vec_slb<mode>"): Turn into expander
14334         and force the shift count operand to V16QImode.
14335         ("vec_srb<mode>"): Set shift count mode to V16QI.
14337 2017-09-26  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
14339         * config/s390/vector.md ("vec_widen_umult_lo_<mode>")
14340         ("vec_widen_umult_hi_<mode>", "vec_widen_smult_lo_<mode>")
14341         ("vec_widen_smult_hi_<mode>"): New expander definitions.
14343 2017-09-26  Richard Earnshaw  <rearnsha@arm.com>
14345         PR target/82175
14346         * config/arm/arm.h (DRIVER_SELF_SPECS): Separate sub-rules with commas.
14348 2017-09-26  Richard Biener  <rguenther@suse.de>
14350         PR tree-optimization/82320
14351         * tree-ssa-sccvn.c (set_ssa_val_to): Changing undef to undef
14352         isn't a change.
14354 2017-09-25  Jeff Law  <law@redhat.com>
14356         * config/rs6000/rs6000-protos.h (output_probe_stack_range): Update
14357         prototype for new argument.
14358         * config/rs6000/rs6000.c (rs6000_emit_allocate_stack_1): New function,
14359         mostly extracted from rs6000_emit_allocate_stack.
14360         (rs6000_emit_probe_stack_range_stack_clash): New function.
14361         (rs6000_emit_allocate_stack): Call
14362         rs6000_emit_probe_stack_range_stack_clash as needed.
14363         (rs6000_emit_probe_stack_range): Add additional argument
14364         to call to gen_probe_stack_range{si,di}.
14365         (output_probe_stack_range): New.
14366         (output_probe_stack_range_1): Renamed from output_probe_stack_range.
14367         (output_probe_stack_range_stack_clash): New.
14368         (rs6000_emit_prologue): Emit notes into dump file as requested.
14369         * rs6000.md (allocate_stack): Handle -fstack-clash-protection.
14370         (probe_stack_range<P:mode>): Operand 0 is now early-clobbered.
14371         Add additional operand and pass it to output_probe_stack_range.
14373 2017-09-25  Bin Cheng  <bin.cheng@arm.com>
14375         PR tree-optimization/82163
14376         * tree-ssa-loop-manip.h (verify_loop_closed_ssa): New parameter.
14377         (checking_verify_loop_closed_ssa): New parameter.
14378         * tree-ssa-loop-manip.c (check_loop_closed_ssa_use): Delete.
14379         (check_loop_closed_ssa_stmt): Delete.
14380         (check_loop_closed_ssa_def, check_loop_closed_ssa_bb): New functions.
14381         (verify_loop_closed_ssa): Check loop closed ssa form for LOOP.
14382         (tree_transform_and_unroll_loop): Check loop closed ssa form only for
14383         changed loops.
14385 2017-09-25  Pekka Jaaskelainen <pekka@parmance.com>
14387         * brig-builtins.def: Treat HSAIL barrier builtins as
14388         setjmp/longjump style functions.
14390 2017-09-25  Richard Sandiford  <richard.sandiford@linaro.org>
14392         * target.def (constant_alignment): New hook.
14393         * defaults.h (CONSTANT_ALIGNMENT): Delete.
14394         * doc/tm.texi.in (CONSTANT_ALIGNMENT): Replace with...
14395         (TARGET_CONSTANT_ALIGNMENT): ...this new hook.
14396         * doc/tm.texi: Regenerate.
14397         * targhooks.h (default_constant_alignment): Declare.
14398         (constant_alignment_word_strings): Likewise.
14399         * targhooks.c (default_constant_alignment): New function.
14400         (constant_alignment_word_strings): Likewise.
14401         * builtins.c (get_object_alignment_2): Use targetm.constant_alignment
14402         instead of CONSTANT_ALIGNMENT.
14403         * varasm.c (align_variable, get_variable_align, build_constant_desc)
14404         (force_const_mem): Likewise.
14405         * config/aarch64/aarch64.h (CONSTANT_ALIGNMENT): Delete.
14406         * config/aarch64/aarch64.c (aarch64_constant_alignment): New function.
14407         (aarch64_classify_address): Call it instead of CONSTANT_ALIGNMENT.
14408         (TARGET_CONSTANT_ALIGNMENT): Redefine.
14409         * config/alpha/alpha.h (CONSTANT_ALIGNMENT): Delete commented-out
14410         definition.
14411         * config/arc/arc.h (CONSTANT_ALIGNMENT): Delete.
14412         * config/arc/arc.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14413         constant_alignment_word_strings.
14414         * config/arm/arm.h (CONSTANT_ALIGNMENT_FACTOR): Delete.
14415         (CONSTANT_ALIGNMENT): Likewise.
14416         * config/arm/arm.c (TARGET_CONSTANT_ALIGNMENT): Redefine.
14417         (arm_constant_alignment): New function.
14418         * config/bfin/bfin.h (CONSTANT_ALIGNMENT): Delete.
14419         * config/bfin/bfin.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14420         constant_alignment_word_strings.
14421         * config/cr16/cr16.h (CONSTANT_ALIGNMENT): Delete.
14422         * config/cr16/cr16.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14423         constant_alignment_word_strings.
14424         * config/cris/cris.h (CONSTANT_ALIGNMENT): Delete.
14425         * config/cris/cris.c (TARGET_CONSTANT_ALIGNMENT): Redefine.
14426         (cris_constant_alignment): New function.
14427         * config/epiphany/epiphany.h (CONSTANT_ALIGNMENT): Delete.
14428         * config/epiphany/epiphany.c (TARGET_CONSTANT_ALIGNMENT): Redefine.
14429         (epiphany_constant_alignment): New function.
14430         * config/fr30/fr30.h (CONSTANT_ALIGNMENT): Delete.
14431         * config/fr30/fr30.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14432         constant_alignment_word_strings.
14433         * config/frv/frv.h (CONSTANT_ALIGNMENT): Delete.
14434         * config/frv/frv.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14435         constant_alignment_word_strings.
14436         * config/ft32/ft32.h (CONSTANT_ALIGNMENT): Delete.
14437         * config/ft32/ft32.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14438         constant_alignment_word_strings.
14439         * config/i386/i386.h (CONSTANT_ALIGNMENT): Delete.
14440         * config/i386/i386-protos.h (ix86_constant_alignment): Delete.
14441         * config/i386/i386.c (ix86_constant_alignment): Make static.
14442         Use the same interface as the target hook.
14443         (TARGET_CONSTANT_ALIGNMENT): Redefine.
14444         * config/ia64/ia64.h (CONSTANT_ALIGNMENT): Delete.
14445         * config/ia64/ia64.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14446         constant_alignment_word_strings.
14447         * config/iq2000/iq2000.h (CONSTANT_ALIGNMENT): Delete.
14448         * config/iq2000/iq2000.c (iq2000_constant_alignment): New function.
14449         (TARGET_CONSTANT_ALIGNMENT): Redefine.
14450         * config/lm32/lm32.h (CONSTANT_ALIGNMENT): Delete.
14451         * config/lm32/lm32.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14452         constant_alignment_word_strings.
14453         * config/m32r/m32r.h (CONSTANT_ALIGNMENT): Delete.
14454         * config/m32r/m32r.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14455         constant_alignment_word_strings.
14456         * config/mcore/mcore.h (CONSTANT_ALIGNMENT): Delete.
14457         * config/mcore/mcore.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14458         constant_alignment_word_strings.
14459         * config/microblaze/microblaze.h (CONSTANT_ALIGNMENT): Delete.
14460         * config/microblaze/microblaze.c (microblaze_constant_alignment):
14461         New function.
14462         (TARGET_CONSTANT_ALIGNMENT): Redefine.
14463         * config/mips/mips.h (CONSTANT_ALIGNMENT): Delete.
14464         * config/mips/mips.c (mips_constant_alignment): New function.
14465         (TARGET_CONSTANT_ALIGNMENT): Redefine.
14466         * config/mmix/mmix.h (CONSTANT_ALIGNMENT): Delete.
14467         * config/mmix/mmix-protos.h (mmix_constant_alignment): Delete.
14468         * config/mmix/mmix.c (TARGET_CONSTANT_ALIGNMENT): Redefine.
14469         (mmix_constant_alignment): Make static.  Use the same interface
14470         as the target hook.
14471         * config/moxie/moxie.h (CONSTANT_ALIGNMENT): Delete.
14472         * config/moxie/moxie.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14473         constant_alignment_word_strings.
14474         * config/nios2/nios2.h (CONSTANT_ALIGNMENT): Delete.
14475         * config/nios2/nios2.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14476         constant_alignment_word_strings.
14477         * config/pa/pa.h (CONSTANT_ALIGNMENT): Delete.
14478         * config/pa/pa.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14479         constant_alignment_word_strings.
14480         * config/powerpcspe/powerpcspe.h (CONSTANT_ALIGNMENT): Delete.
14481         * config/powerpcspe/powerpcspe.c (TARGET_CONSTANT_ALIGNMENT): Redefine.
14482         (rs6000_constant_alignment): New function.
14483         * config/riscv/riscv.h (CONSTANT_ALIGNMENT): Delete.
14484         * config/riscv/riscv.c (riscv_constant_alignment): New function.
14485         (TARGET_CONSTANT_ALIGNMENT): Redefine.
14486         * config/rs6000/rs6000.h (CONSTANT_ALIGNMENT): Delete.
14487         * config/rs6000/rs6000.c (TARGET_CONSTANT_ALIGNMENT): Redefine.
14488         (rs6000_constant_alignment): New function.
14489         * config/s390/s390.h (CONSTANT_ALIGNMENT): Delete.
14490         * config/s390/s390.c (s390_constant_alignment): New function.
14491         (TARGET_CONSTANT_ALIGNMENT): Redefine.
14492         * config/sh/sh.h (CONSTANT_ALIGNMENT): Delete.
14493         * config/sh/sh.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14494         constant_alignment_word_strings.
14495         * config/sparc/sparc.h (CONSTANT_ALIGNMENT): Delete.
14496         * config/sparc/sparc.c (TARGET_CONSTANT_ALIGNMENT): Redefine.
14497         (sparc_constant_alignment): New function.
14498         * config/spu/spu.h (CONSTANT_ALIGNMENT): Delete.
14499         * config/spu/spu.c (spu_constant_alignment): New function.
14500         (TARGET_CONSTANT_ALIGNMENT): Redefine.
14501         * config/stormy16/stormy16.h (CONSTANT_ALIGNMENT): Delete.
14502         * config/stormy16/stormy16.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14503         constant_alignment_word_strings.
14504         * config/tilegx/tilegx.h (CONSTANT_ALIGNMENT): Delete.
14505         * config/tilegx/tilegx.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14506         constant_alignment_word_strings.
14507         * config/tilepro/tilepro.h (CONSTANT_ALIGNMENT): Delete.
14508         * config/tilepro/tilepro.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14509         constant_alignment_word_strings.
14510         * config/visium/visium.h (CONSTANT_ALIGNMENT): Delete.
14511         * config/visium/visium.c (TARGET_CONSTANT_ALIGNMENT): Redefine.
14512         (visium_constant_alignment): New function.
14513         * config/xtensa/xtensa.h (CONSTANT_ALIGNMENT): Delete.
14514         * config/xtensa/xtensa.c (TARGET_CONSTANT_ALIGNMENT): Redefine.
14515         (xtensa_constant_alignment): New function.
14516         * system.h (CONSTANT_ALIGNMENT): Poison.
14518 2017-09-25  Will Schmidt  <will_schmidt@vnet.ibm.com>
14520         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add handling
14521         for early folding of vector stores (ALTIVEC_BUILTIN_ST_*).
14522         (rs6000_builtin_valid_without_lhs): New helper function.
14523         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
14524         Remove obsoleted code for handling ALTIVEC_BUILTIN_VEC_ST.
14526 2017-09-25  Richard Sandiford  <richard.sandiford@linaro.org>
14528         * target.h (vec_perm_indices): Use unsigned short rather than
14529         unsigned char.
14530         (auto_vec_perm_indices): Likewise.
14531         * config/aarch64/aarch64.c (aarch64_vectorize_vec_perm_const_ok):
14532         Use unsigned int rather than unsigned char.
14533         * config/arm/arm.c (arm_vectorize_vec_perm_const_ok): Likewise.
14535 2017-09-25  Richard Biener  <rguenther@suse.de>
14537         * cfgloop.h (sort_sibling_loops): Declare.
14538         * cfgloop.c (sort_sibling_loops_cmp): New helper.
14539         (sort_sibling_loops): New function sorting the sibling loop list
14540         in RPO order.
14541         * graphite.c (graphite_transform_loops): Sort sibling loops.
14543 2017-09-25  Richard Sandiford  <richard.sandifird@linaro.org>
14545         * target.def (vec_perm_const_ok): Change sel parameter to
14546         vec_perm_indices.
14547         * optabs-query.c (can_vec_perm_p): Update accordingly.
14548         * doc/tm.texi: Regenerate.
14549         * config/aarch64/aarch64.c (expand_vec_perm_d): Change perm to
14550         auto_vec_perm_indices and remove separate nelt field.
14551         (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
14552         (aarch64_evpc_ext, aarch64_evpc_rev, aarch64_evpc_dup)
14553         (aarch64_evpc_tbl, aarch64_expand_vec_perm_const_1)
14554         (aarch64_expand_vec_perm_const): Update accordingly.
14555         (aarch64_vectorize_vec_perm_const_ok): Likewise.  Change sel
14556         to vec_perm_indices.
14557         * config/arm/arm.c (expand_vec_perm_d): Change perm to
14558         auto_vec_perm_indices and remove separate nelt field.
14559         (arm_evpc_neon_vuzp, arm_evpc_neon_vzip, arm_evpc_neon_vrev)
14560         (arm_evpc_neon_vtrn, arm_evpc_neon_vext, arm_evpc_neon_vtbl)
14561         (arm_expand_vec_perm_const_1, arm_expand_vec_perm_const): Update
14562         accordingly.
14563         (arm_vectorize_vec_perm_const_ok): Likewise.  Change sel
14564         to vec_perm_indices.
14565         * config/i386/i386.c (ix86_vectorize_vec_perm_const_ok): Change
14566         sel to vec_perm_indices.
14567         * config/ia64/ia64.c (ia64_vectorize_vec_perm_const_ok): Likewise.
14568         * config/mips/mips.c (mips_vectorize_vec_perm_const_ok): Likewise.
14569         * config/powerpcspe/powerpcspe.c (rs6000_vectorize_vec_perm_const_ok):
14570         Likewise.
14571         * config/rs6000/rs6000.c (rs6000_vectorize_vec_perm_const_ok):
14572         Likewise.
14574 2017-09-25  Pierre-Marie de Rodat  <derodat@adacore.com>
14576         PR debug/82155
14577         * dwarf2out.c (dwarf2out_early_global_decl): Call dwarf2out_decl
14578         on the FUNCTION_DECL function context if it has a DIE that is a
14579         declaration.
14581 2017-09-25  Richard Biener  <rguenther@suse.de>
14583         PR tree-optimization/82285
14584         * tree-vect-patterns.c (vect_recog_bool_pattern): Also handle
14585         enumeral types.
14587 2017-09-25  Tom de Vries  <tom@codesourcery.com>
14589         PR target/80035
14590         PR target/81069
14591         * config/nvptx/nvptx.c (nvptx_output_call_insn): Add exit after call to
14592         noreturn function.
14594 2017-09-25  Richard Biener  <rguenther@suse.de>
14596         * graphite-optimize-isl.c (optimize_isl): Fail and dump if
14597         ISL errors other than isl_error_quota happen.  Dump if the
14598         schedule is the same.
14599         * graphite-sese-to-poly.c (build_poly_scop): Fail on ISL
14600         errors instead of aborting inside ISL.
14602 2017-09-25  Iain Sandoe  <iain@codesourcery.com>
14604         PR target/80556
14605         * config/i386/darwin.h (REAL_LIB_SPEC): New; put libSystem ahead
14606         of libgcc_eh for m64.
14607         * config/i386/darwin64.h: Likewise.
14609 2017-09-25  Richard Biener  <rguenther@suse.de>
14611         PR middle-end/82144
14612         * dwarf2out.c (gen_enumeration_type_die): Do not add alignment
14613         attribute for incomplete types nor twice for complete ones.
14615 2017-09-24  Uros Bizjak  <ubizjak@gmail.com>
14617         PR target/82267
14618         * config/i386/i386.c (ix86_print_operand_address_as): Encode
14619         %esp as %rsp to avoid 0x67 prefix if there is no index or base
14620         register.
14622 2017-09-23  Uros Bizjak  <ubizjak@gmail.com>
14624         PR bootstrap/82306
14625         * config/i386/i386.opt (mprefer-avx256): Use
14626         ix86_target_flags variable.
14627         * config/i386/i386.c (ix86_target_string): Move
14628         -mprefer-avx256 to flag2_opts.
14630 2017-09-22  Jakub Jelinek  <jakub@redhat.com>
14632         PR middle-end/35691
14633         * match.pd: Simplify x == -1 & y == -1 into (x & y) == -1
14634         and x != -1 | y != -1 into (x & y) != -1.
14636 2017-09-22  Steve Ellcey  <sellcey@cavium.com>
14638         * config.gcc: Add new case statement to set
14639         default_gnu_indirect_function.  Remove it from x86_64-*-linux*,
14640         i[34567]86-*, powerpc*-*-linux*spe*, powerpc*-*-linux*, s390-*-linux*,
14641         s390x-*-linux* case statements.   Added aarch64 to the list of
14642         supported architectures.
14644 2017-09-22  Richard Sandiford  <richard.sandiford@linaro.org>
14646         PR tree-optimization/82289
14647         * tree-vect-data-refs.c (vect_get_peeling_costs_all_drs): Check
14648         STMT_VINFO_RELEVANT_P.
14650 2017-09-22  Richard Sandiford  <richard.sandiford@linaro.org>
14651             Alan Hayward  <alan.hayward@arm.com>
14652             David Sherwood  <david.sherwood@arm.com>
14654         * tree-vrp.c (extract_range_from_multiplicative_op_1): Assert
14655         for VR_RANGE only; don't allow VR_ANTI_RANGE.
14656         (extract_range_from_binary_expr_1): Don't call
14657         extract_range_from_multiplicative_op_1 if !range_int_cst_p.
14659 2017-09-22  Richard Sandiford  <richard.sandiford@linaro.org>
14660             Alan Hayward  <alan.hayward@arm.com>
14661             David Sherwood  <david.sherwood@arm.com>
14663         * target.def (preferred_vector_alignment): New hook.
14664         * doc/tm.texi.in (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): New
14665         hook.
14666         * doc/tm.texi: Regenerate.
14667         * targhooks.h (default_preferred_vector_alignment): Declare.
14668         * targhooks.c (default_preferred_vector_alignment): New function.
14669         * tree-vectorizer.h (dataref_aux): Add a target_alignment field.
14670         Expand commentary.
14671         (DR_TARGET_ALIGNMENT): New macro.
14672         (aligned_access_p): Update commentary.
14673         (vect_known_alignment_in_bytes): New function.
14674         * tree-vect-data-refs.c (vect_calculate_required_alignment): New
14675         function.
14676         (vect_compute_data_ref_alignment): Set DR_TARGET_ALIGNMENT.
14677         Calculate the misalignment based on the target alignment rather than
14678         the vector size.
14679         (vect_update_misalignment_for_peel): Use DR_TARGET_ALIGMENT
14680         rather than TYPE_ALIGN / BITS_PER_UNIT to update the misalignment.
14681         (vect_enhance_data_refs_alignment): Mask the byte misalignment with
14682         the target alignment, rather than masking the element misalignment
14683         with the number of elements in a vector.  Also use the target
14684         alignment when calculating the maximum number of peels.
14685         (vect_find_same_alignment_drs): Use vect_calculate_required_alignment
14686         instead of TYPE_ALIGN_UNIT.
14687         (vect_duplicate_ssa_name_ptr_info): Remove stmt_info parameter.
14688         Measure DR_MISALIGNMENT relative to DR_TARGET_ALIGNMENT.
14689         (vect_create_addr_base_for_vector_ref): Update call accordingly.
14690         (vect_create_data_ref_ptr): Likewise.
14691         (vect_setup_realignment): Realign by ANDing with
14692         -DR_TARGET_MISALIGNMENT.
14693         * tree-vect-loop-manip.c (vect_gen_prolog_loop_niters): Calculate
14694         the number of peels based on DR_TARGET_ALIGNMENT.
14695         * tree-vect-stmts.c (get_group_load_store_type): Compare the gap
14696         with the guaranteed alignment boundary when deciding whether
14697         overrun is OK.
14698         (vectorizable_mask_load_store): Interpret DR_MISALIGNMENT
14699         relative to DR_TARGET_ALIGNMENT instead of TYPE_ALIGN_UNIT.
14700         (ensure_base_align): Remove stmt_info parameter.  Get the
14701         target base alignment from DR_TARGET_ALIGNMENT.
14702         (vectorizable_store): Update call accordingly.   Interpret
14703         DR_MISALIGNMENT relative to DR_TARGET_ALIGNMENT instead of
14704         TYPE_ALIGN_UNIT.
14705         (vectorizable_load): Likewise.
14707 2017-09-22  Richard Sandiford  <richard.sandiford@linaro.org>
14708             Alan Hayward  <alan.hayward@arm.com>
14709             David Sherwood  <david.sherwood@arm.com>
14711         * tree-vectorizer.h (vect_get_scalar_dr_size): New function.
14712         * tree-vect-data-refs.c (vect_update_misalignment_for_peel): Use it.
14713         (vect_enhance_data_refs_alignment): Likewise.
14715 2017-09-22  Richard Earnshaw  <richard.earnshaw@arm.com>
14717         * config/arm/parsecpu.awk (fatal): Note that we've encountered an
14718         error.  Only quit immediately if parsing is complete.
14719         (BEGIN): Initialize fatal_err and parse_done.
14720         (begin fpu, end fpu): Check number of arguments.
14721         (begin arch, end arch): Likewise.
14722         (begin cpu, end cpu): Likewise.
14723         (cname, tune for, tune flags, architecture, fpu, option): Likewise.
14724         (optalias): Likewise.
14726 2017-09-22  Richard Earnshaw  <richard.earnshaw@arm.com>
14728         * config.gcc (arm*-*-*): Don't add arm-isa.h to tm_p_file.
14729         * config/arm/arm-isa.h: Delete.  Move definitions to ...
14730         * arm-cpus.in: ... here.  Use new feature and fgroup values.
14731         * config/arm/arm.c (arm_option_override): Use lower case for feature
14732         bit names.
14733         * config/arm/arm.h (TARGET_HARD_FLOAT): Likewise.
14734         (TARGET_VFP3, TARGET_VFP5, TARGET_FMA): Likewise.
14735         * config/arm/parsecpu.awk (END): Add new command 'isa'.
14736         (isa_pfx): Delete.
14737         (print_isa_bits_for): New function.
14738         (gen_isa): New function.
14739         (gen_comm_data): Use print_isa_bits_for.
14740         (define feature): New keyword.
14741         (define fgroup): New keyword.
14742         * config/arm/t-arm (TM_H): Remove.
14743         (GTM_H): Add arm-isa.h.
14744         (arm-isa.h): Add rule to generate file.
14745         * common/config/arm/arm-common.c: (arm_canon_arch_option): Use lower
14746         case for feature bit names.
14748 2017-09-22  Richard Biener  <rguenther@suse.de>
14750         * graphite-isl-ast-to-gimple.c (graphite_verify): Inline into
14751         single caller.
14752         (graphite_regenerate_ast_isl): Do not reset SCEV.  Move debug
14753         print of no dependency loops ...
14754         * graphite.c (graphite_transform_loops): ... here.
14755         (canonicalize_loop_closed_ssa_form): Work from inner to outer
14756         loops.
14757         (same_close_phi_node, remove_duplicate_close_phi,
14758         make_close_phi_nodes_unique, defined_in_loop_p): Fold into ...
14759         (canonicalize_loop_closed_ssa): ... here and simplify.
14760         * graphite-optimize-isl.c: Include tree-vectorizer.h.
14761         (optimize_isl): Use dump_printf_loc to tell when we stopped
14762         optimizing because of an ISL timeout.
14764 2017-09-22  Richard Biener  <rguenther@suse.de>
14766         PR tree-optimization/82291
14767         * tree-if-conv.c (predicate_mem_writes): Make sure to
14768         remove writes in blocks predicated with false.
14770 2017-09-22  Richard Biener  <rguenther@suse.de>
14772         * sese.c: Include cfganal.h.
14773         (if_region_set_false_region): Remove.
14774         (create_if_region_on_edge): Likewise.
14775         (move_sese_in_condition): Re-implement without destroying
14776         dominators.
14778 2017-09-22  Richard Biener  <rguenther@suse.de>
14780         * graphite-isl-ast-to-gimple.c (translate_pending_phi_nodes):
14781         Verify both BBs contain loop PHI nodes before dispatching to
14782         copy_loop_phi_args.
14783         (graphite_regenerate_ast_isl): Do not recompute dominators,
14784         do not verify three times.  Restructure for clarity.
14785         * graphite-scop-detection.c (same_close_phi_node,
14786         remove_duplicate_close_phi, make_close_phi_nodes_unique,
14787         defined_in_loop_p, canonicalize_loop_closed_ssa,
14788         canonicalize_loop_closed_ssa_form): Simplify, remove excess
14789         checking and SSA rewrite, move to ...
14790         * graphite.c: ... here.  Include ssa.h and tree-ssa-loop-manip.h.
14791         (graphite_initialize): Do not pass in ctx, do not reset the
14792         SCEV cache, compute only dominators.
14793         (graphite_transform_loops): Allocate ISL ctx after
14794         graphite_initialize.  Call canonicalize_loop_closed_ssa_form.
14795         Maintain post-dominators only around build_scops.
14796         * sese.c (if_region_set_false_region): Make static.  Free
14797         and recompute dominators.
14798         (move_sese_in_condition): Assert we don't get called with
14799         post-dominators computed.
14800         * sese.h (if_region_set_false_region): Remove.
14802 2017-09-22  Sergey Shalnov  <sergey.shalnov@intel.com>
14804         * config/i386/sse.md ("mov<mode>_internal"): Use <sseinsnmode>
14805         mode attribute for TARGET_AVX512VL.
14807 2017-09-21  Sergey Shalnov  <sergey.shalnov@intel.com>
14809         * config/i386/i386.opt (mprefer-avx256): New option.
14810         * config/i386/i386.c (ix86_target_string): Add -mprefer-avx256
14811         to flag_opts.
14812         (ix86_preferred_simd_mode): Return 256-bit AVX modes
14813         for TARGET_PREFER_AVX256.
14814         * doc/invoke.texi (x86 Options): Document -mprefer-avx256.
14816 2017-09-21  Jeff Law  <law@redhat.com>
14818         * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash):
14819         Fix dump output if the only stack space is for pushed registers.
14821 2017-09-21  Richard Sandiford  <richard.sandiford@linaro.org>
14823         * config/spu/spu.c (spu_sched_adjust_cost): Update after renaming
14824         of insn_cost.
14826 2017-09-21  Martin Sebor  <msebor@redhat.com>
14828         PR c/81882
14829         * doc/extend.texi (attribute ifunc): Avoid relying on ill-formed
14830         code (in C++) or code that triggers warnings.
14832 2017-09-21  Eric Botcazou  <ebotcazou@adacore.com>
14834         * stor-layout.c (bit_from_pos): Do not distribute the conversion.
14836 2017-09-21  Segher Boessenkool  <segher@kernel.crashing.org>
14838         * haifa-sched.c: Rename insn_cost to insn_sched_cost.
14839         * sched-rgn.c: Ditto.
14840         * sel-sched-ir.c: Ditto.
14842 2017-09-21  Alexander Monakov  <amonakov@ispras.ru>
14844         * toplev.h (set_random_seed): Adjust return type.
14845         * toplev.c (init_local_tick): Move eager initialization of random_seed
14846         to get_random_seed.  Adjust comment.
14847         (init_random_seed): Inline to get_random_seed, delete.
14848         (get_random_seed): Initialize random_seed lazily.
14849         (set_random_seed): Do not return previous value.
14850         (print_switch_value): Do not call get_random_seed.
14852 2017-09-21  Evgeny Kudryashov  <kudryashov@ispras.ru>
14854         * cgraph.c (delete_function_version): New, broken out from...
14855         (cgraph_node::delete_function_version): ...here.  Rename to
14856         cgraph_node::delete_function_version_by_decl.  Update all uses.
14857         (cgraph_node::remove): Call delete_function_version.
14859 2017-09-21  Jakub Jelinek  <jakub@redhat.com>
14861         PR sanitizer/81715
14862         * tree-inline.c (expand_call_inline): Emit clobber stmts for
14863         VAR_DECLs to which addressable non-volatile parameters are mapped
14864         and for id->retvar after the return value assignment.  Clear
14865         id->retval and id->retbnd after inlining.
14867 2017-09-21  Richard Biener  <rguenther@suse.de>
14869         PR tree-optimization/82276
14870         PR tree-optimization/82244
14871         * tree-vrp.c (build_assert_expr_for): Set
14872         SSA_NAME_OCCURS_IN_ABNORMAL_PHI if the variable we assert on
14873         has it set.
14874         (remove_range_assertions): Revert earlier change.
14876 2017-09-21  Wilco Dijkstra  <wdijkstr@arm.com>
14878         PR target/71951
14879         * config/aarch64/aarch64.h (LIBGCC2_UNWIND_ATTRIBUTE): Define.
14881 2017-09-21  Richard Biener  <rguenther@suse.de>
14883         * graphite-isl-ast-to-gimple.c (graphite_regenerate_ast_isl):
14884         Restore valid IL after code generation errors.
14885         * graphite.c (graphite_transform_loops): Diagnose code
14886         generation issues as MSG_MISSED_OPTIMIZATION and continue
14887         with processing SCOPs.
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         * calls.c (compute_argument_addresses): Use simplify_gen_binary
14894         rather than choosing between plus_constant and gen_rtx_<CODE>.
14895         * expr.c (emit_push_insn): Likewise.
14896         (expand_expr_real_2): Likewise.
14898 2017-09-21  Richard Sandiford  <richard.sandiford@linaro.org>
14899             Alan Hayward  <alan.hayward@arm.com>
14900             David Sherwood  <david.sherwood@arm.com>
14902         * loop-unroll.c (split_iv): Call copy_rtx on the step.
14904 2017-09-21  Richard Sandiford  <richard.sandiford@linaro.org>
14905             Alan Hayward  <alan.hayward@arm.com>
14906             David Sherwood  <david.sherwood@arm.com>
14908         * tree.c (find_atomic_core_type): Check tree_fits_uhwi_p before
14909         calling tree_to_uhwi.
14911 2017-09-21  Richard Sandiford  <richard.sandiford@linaro.org>
14912             Alan Hayward  <alan.hayward@arm.com>
14913             David Sherwood  <david.sherwood@arm.com>
14915         * tree-ssa-ccp.c (get_value_for_expr): Use a positive test for
14916         INTEGER_CST rather than a negative test for ADDR_EXPR.
14918 2017-09-21  Richard Sandiford  <richard.sandiford@linaro.org>
14919             Alan Hayward  <alan.hayward@arm.com>
14920             David Sherwood  <david.sherwood@arm.com>
14922         * tree-vrp.c (extract_range_from_binary_expr_1): Check
14923         int_cst_rangeN before calling value_range_constant_singleton (&vrN).
14925 2017-09-21  Richard Biener  <rguenther@suse.de>
14927         PR tree-optimization/71351
14928         * graphite-isl-ast-to-gimple.c (translate_isl_ast_to_gimple::
14929         graphite_create_new_loop_guard): Remove, fold remaining parts
14930         into caller ...
14931         (translate_isl_ast_node_for): ... here and simplify.
14933 2017-09-21  Jakub Jelinek  <jakub@redhat.com>
14935         PR target/82260
14936         * config/i386/i386.md (*movqi_internal): Replace (=q,q) alternative
14937         with (=Q,Q), (=R,R) and (=r,r) alternatives, only enable the
14938         latter two for 64-bit, renumber alternatives, for -Os imov (=q,n)
14939         alternative always use QI mode, for -Os imov (=R,R) alternative
14940         always use SI mode, for imov (=Q,Q) or (=r,r) alternatives
14941         ignore -Os.
14943 2017-09-20  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
14944             Jeff Law  <law@redhat.com>
14946         * config/s390/s390.c (MIN_UNROLL_PROBES): Define.
14947         (allocate_stack_space): New function, partially extracted from
14948         s390_emit_prologue.
14949         (s390_emit_prologue): Track offset to most recent stack probe.
14950         Code to allocate space moved into allocate_stack_space.
14951         Dump actions when no stack is allocated.
14952         (s390_prologue_plus_offset): New function.
14953         (s390_emit_stack_probe): Likewise.
14955 2017-09-20  Alexandre Oliva  <aoliva@redhat.com>
14957         * common.opt (Wa, Wl, Wp, g, gz=): Add
14958         RejectNegative.
14959         (gno-column-info): Remove.
14960         (gcolumn-info): Drop RejectNegative.
14961         (gno-): New prefix.
14962         (gno-record-gcc-switches): Remove.
14963         (grecord-gcc-switches): Drop RejectNegative.
14964         (gno-split-dwarf): Remove.
14965         (gsplit-dwarf): Drop RejectNegative.
14966         (gno-strict-dwarf): Remove.
14967         (gstrict-dwarf): Drop RejectNegative.
14968         * config/darwin.opt (gfull, gused): Add RejectNegative.
14969         * dwarf2out.c (gen_producer_string): Drop
14970         gno-record-gcc-switches handler.
14971         * optc-gen.awk: Add g to prefixes with negative forms.
14972         * opts-common.c (remapping_prefix_p): New.
14973         (find_opt): Check it.
14974         (generate_canonical_option): Test g prefix.
14975         (option_map): Add -gno- mapping.
14976         (add_misspelling_candidates): Check remapping_prefix_p.
14978 2017-09-20  Jeff Law  <law@redhat.com>
14980         * config/powerpcspe/powerpcspe.c (rs6000_expand_prologue): Fix
14981         thinko in stack clash protection support.
14983         * explow.c (compute_stack_clash_protection_loop_data): Use
14984         CONST_INT_P instead of explicit test.  Verify object is a
14985         CONST_INT_P before looking at INTVAL.
14986         (anti_adjust_stack_and_probe_stack_clash): Use CONST_INT_P
14987         instead of explicit test.
14989 2017-09-20  Segher Boessenkool  <segher@kernel.crashing.org>
14991         PR target/77687
14992         * config/rs6000/rs6000.md (stack_restore_tie): Store to a scratch
14993         address instead of to r1 and r11.
14995 2017-09-20  Sebastian Peryt  <sebastian.peryt@intel.com>
14997         * config.gcc: Support "knm".
14998         * config/i386/driver-i386.c (host_detect_local_cpu): Detect "knm".
14999         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
15000         PROCESSOR_KNM.
15001         * config/i386/i386.c (m_KNM): Define.
15002         (processor_target_table): Add "knm".
15003         (PTA_KNM): Define.
15004         (ix86_option_override_internal): Add "knm".
15005         (ix86_issue_rate): Add PROCESSOR_KNM.
15006         (ix86_adjust_cost): Ditto.
15007         (ia32_multipass_dfa_lookahead): Ditto.
15008         (get_builtin_code_for_version): Handle PROCESSOR_KNM.
15009         (fold_builtin_cpu): Add M_INTEL_KNM.
15010         * config/i386/i386.h (processor_costs): Define TARGET_KNM.
15011         (processor_type): Add PROCESSOR_KNM.
15012         * config/i386/x86-tune.def: Add m_KNM.
15013         * doc/invoke.texi: Add knm as x86 -march=/-mtune= CPU type.
15015 2017-09-20  Richard Biener  <rguenther@suse.de>
15017         PR tree-optimization/80213
15018         * graphite-scop-detection.c (trivially_empty_bb_p): Labels
15019         are allowed in empty BBs as well.
15020         (canonicalize_loop_closed_ssa): Also look for other complex
15021         edges.
15022         (scop_detection::get_sese): Include the loop-closed PHI block
15023         in loop SESEs.
15024         (scop_detection::merge_sese): Remove code adding extra blocks.
15025         (scop_detection::region_has_one_loop): Adjust for get_sese changes.
15026         (build_scops): Assert the final returned scop is invalid.
15028 2017-09-20  Richard Biener  <rguenther@suse.de>
15030         PR tree-optimization/82264
15031         * tree-ssa-sccvn.c (vn_phi_eq): Use safe_dyn_cast to check
15032         for GIMPLE_CONDs.
15033         (vn_phi_lookup): Likewise.
15034         (vn_phi_insert): Likewise.
15036 2017-09-20  Jakub Jelinek  <jakub@redhat.com>
15038         * dwarf2out.c (tree_add_const_value_attribute): For INTEGER_CST
15039         that fits into uhwi or shwi, add DW_AT_const_value regardless
15040         of early_dwarf without going through RTL, using add_AT_unsigned
15041         or add_AT_int.
15043         * dwarf2out.c (DEBUG_LTO_DWO_INFO_SECTION): Reorder defines.
15044         (DEBUG_LTO_ABBREV_SECTION): Likewise.
15045         (DEBUG_LTO_MACINFO_SECTION): Likewise.
15046         (DEBUG_MACRO_SECTION): Likewise.
15047         (DEBUG_LTO_MACRO_SECTION): Likewise.
15048         (DEBUG_STR_DWO_SECTION): Likewise.
15049         (DEBUG_LTO_STR_DWO_SECTION): Likewise.
15050         (DEBUG_LTO_LINE_SECTION): Drop .dwo suffix from the name.
15051         (DEBUG_LTO_DWO_LINE_SECTION): Define.
15052         (DEBUG_LTO_LINE_STR_SECTION): Define.
15053         (init_sections_and_labels): Initialize debug_line_str_section
15054         variable.  Initialize debug_loc_section for -gdwarf-5 to
15055         DEBUG_LOCLISTS_SECTION.  Formatting fixes.
15057 2017-09-20  Richard Biener  <rguenther@suse.de>
15059         * graphite-sese-to-poly.c (extract_affine): Properly handle
15060         POINTER_PLUS_EXPR, BIT_NOT_EXPR and conversion to signed.
15062 2017-09-20  Richard Biener  <rguenther@suse.de>
15064         PR tree-optimization/81373
15065         * graphite-scop-detection.c (build_cross_bb_scalars_def):
15066         Force SESE live-out defs to be handled even if they are
15067         scev_analyzable_p.
15069 2017-09-19  Jeff Law  <law@redhat.com>
15071         * combine-stack-adj.c (combine_stack_adjustments_for_block): Do
15072         nothing for stack adjustments with REG_STACK_CHECK.
15073         * sched-deps.c (parse_add_or_inc): Reject insns with
15074         REG_STACK_CHECK from dependency breaking.
15075         * config/i386/i386.c (pro_epilogue_adjust_stack): Return insn.
15076         (ix86_adjust_satck_and_probe_stack_clash): Add REG_STACK_NOTEs.
15077         * reg-notes.def (STACK_CHECK): New note.
15079         * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): New.
15080         (ix86_expand_prologue): Dump stack clash info as needed.
15081         Call ix86_adjust_stack_and_probe_stack_clash as needed.
15083         * function.c (dump_stack_clash_frame_info): New function.
15084         * function.h (dump_stack_clash_frame_info): Prototype.
15085         (enum stack_clash_probes): New enum.
15087         * config/alpha/alpha.c (alpha_expand_prologue): Also check
15088         flag_stack_clash_protection.
15089         * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Likewise.
15090         (arm_expand_prologue, thumb1_expand_prologue): Likewise.
15091         (arm_frame_pointer_required): Likewise.
15092         * config/ia64/ia64.c (ia64_compute_frame_size): Likewise.
15093         (ia64_expand_prologue): Likewise.
15094         * config/mips/mips.c (mips_expand_prologue): Likewise.
15095         * config/powerpcspe/powerpcspe.c (rs6000_expand_prologue): Likewise.
15096         * config/sparc/sparc.c (sparc_expand_prologue): Likewise.
15097         (sparc_flat_expand_prologue): Likewise.
15098         * config/spu/spu.c (spu_expand_prologue): Likewise.
15100         * explow.c: Include "params.h".
15101         (anti_adjust_stack_and_probe_stack_clash): New function.
15102         (get_stack_check_protect): Likewise.
15103         (compute_stack_clash_protection_loop_data): Likewise.
15104         (emit_stack_clash_protection_loop_start): Likewise.
15105         (emit_stack_clash_protection_loop_end): Likewise.
15106         (allocate_dynamic_stack_space): Use get_stack_check_protect.
15107         Use anti_adjust_stack_and_probe_stack_clash.
15108         * explow.h (compute_stack_clash_protection_loop_data): Prototype.
15109         (emit_stack_clash_protection_loop_start): Likewise.
15110         (emit_stack_clash_protection_loop_end): Likewise.
15111         * rtl.h (get_stack_check_protect): Prototype.
15112         * target.def (stack_clash_protection_final_dynamic_probe): New hook.
15113         * targhooks.c (default_stack_clash_protection_final_dynamic_probe): New.
15114         * targhooks.h (default_stack_clash_protection_final_dynamic_probe):
15115         Prototype.
15116         * doc/tm.texi.in (TARGET_STACK_CLASH_PROTECTION_FINAL_DYNAMIC_PROBE):
15117         Add @hook.
15118         * doc/tm.texi: Rebuilt.
15119         * config/aarch64/aarch64.c (aarch64_expand_prologue): Use
15120         get_stack_check_protect.
15121         * config/alpha/alpha.c (alpha_expand_prologue): Likewise.
15122         * config/arm/arm.c (arm_expand_prologue): Likewise.
15123         (arm_frame_pointer_required): Likewise.
15124         * config/i386/i386.c (ix86_expand_prologue): Likewise.
15125         * config/ia64/ia64.c (ia64_expand_prologue): Likewise.
15126         * config/mips/mips.c (mips_expand_prologue): Likewise.
15127         * config/powerpcspe/powerpcspe.c (rs6000_emit_prologue): Likewise.
15128         * config/rs6000/rs6000.c (rs6000_emit_prologue): Likewise.
15129         * config/sparc/sparc.c (sparc_expand_prologue): Likewise.
15130         (sparc_flat_expand_prologue): Likewise.
15132         * common.opt (-fstack-clash-protection): New option.
15133         * flag-types.h (enum stack_check_type): Note difference between
15134         -fstack-check= and -fstack-clash-protection.
15135         * params.def (PARAM_STACK_CLASH_PROTECTION_GUARD_SIZE): New PARAM.
15136         (PARAM_STACK_CLASH_PROTECTION_PROBE_INTERVAL): Likewise.
15137         * toplev.c (process_options): Issue warnings/errors for cases
15138         not handled with -fstack-clash-protection.
15139         * doc/invoke.texi (-fstack-clash-protection): Document new option.
15140         (-fstack-check): Note additional problem with -fstack-check=generic.
15141         Note that -fstack-check is primarily for Ada and refer users
15142         to -fstack-clash-protection for stack-clash-protection.
15143         Document new params for stack clash protection.
15145 2017-09-19  Uros Bizjak  <ubizjak@gmail.com>
15147         * config/i386/i386.c (ix86_split_long_move): Do not handle
15148         address used for LEA in a special way.
15150 2017-09-19  Segher Boessenkool  <segher@kernel.crashing.org>
15152         * simplify-rtx.c (simplify_binary_operation_1): Fix typo in comment.
15154 2017-09-19  Martin Sebor  <msebor@redhat.com>
15156         PR c/81854
15157         * cgraphunit.c (handle_alias_pairs): Reject aliases between functions
15158         of incompatible types.
15160 2017-09-19  Will Schmidt  <will_schmidt@vnet.ibm.com>
15162         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add handling
15163         for early folding of vector loads (ALTIVEC_BUILTIN_LVX_*).
15164         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
15165         Remove obsoleted code for handling ALTIVEC_BUILTIN_VEC_LD.
15167 2017-09-19  Richard Biener  <rguenther@suse.de>
15169         PR tree-optimization/82244
15170         * tree-vrp.c (remove_range_assertions): Do not propagate
15171         a constant to abnormals but replace the assert with a copy.
15173 2017-09-19  Alexander Monakov  <amonakov@ispras.ru>
15175         PR rtl-optimization/57878
15176         PR rtl-optimization/68988
15177         * lra-assigns.c (reload_pseudo_compare_func): Remove fragmentation
15178         avoidance test involving non_reload_pseudos.  Move frequency test
15179         below the general fragmentation avoidance test.
15181 2017-09-19  Richard Biener  <rguenther@suse.de>
15183         PR tree-optimization/69728
15184         * graphite-sese-to-poly.c (schedule_error): New global.
15185         (add_loop_schedule): Handle empty domain by failing the
15186         schedule.
15187         (build_original_schedule): Handle schedule_error.
15189 2017-09-19  Richard Biener  <rguenther@suse.de>
15191         * graphite-scop-detection.c (scop_detection::can_represent_loop):
15192         Do not iterate to sibling loops but only to siblings of inner
15193         loops.
15195 2017-09-18  Andreas Schwab  <schwab@linux-m68k.org>
15197         PR target/81613
15198         * config/m68k/m68k.md (moveq feeding equality comparison): Check
15199         that the registers are different.
15201 2017-09-18  Uros Bizjak  <ubizjak@gmail.com>
15203         * config/i386/i386.c (fold_builtin_cpu): Add M_AMDFAM17H
15204         to processor_model and "amdfam17h" to arch_names_table.
15205         * doc/extend.texi (__builtin_cpu_is): Document amdfam17h CPU name.
15207 2017-09-18  Jakub Jelinek  <jakub@redhat.com>
15209         PR c/82234
15210         * doc/extend.texi: Add @findex entry for __builtin_shuffle.
15212 2017-09-18  Richard Sandiford  <richard.sandiford@linaro.org>
15213             Alan Hayward  <alan.hayward@arm.com>
15214             David Sherwood  <david.sherwood@arm.com>
15216         * tree-vectorizer.h (vect_slp_analyze_operations): Replace parameters
15217         with a vec_info *.
15218         * tree-vect-loop.c (vect_analyze_loop_operations): Update call
15219         accordingly.
15220         * tree-vect-slp.c (vect_slp_analyze_node_operations): Add vec_info *
15221         parameter.  Set SLP_TREE_NUMBER_OF_VEC_STMTS here rather than in
15222         vect_schedule_slp_instance.
15223         (vect_slp_analyze_operations): Replace parameters with a vec_info *.
15224         Update call to vect_slp_analyze_node_operations.  Simplify return
15225         value.
15226         (vect_slp_analyze_bb_1): Update call accordingly.
15227         (vect_schedule_slp_instance): Remove vectorization_factor parameter.
15228         Don't calculate SLP_TREE_NUMBER_OF_VEC_STMTS here.
15229         (vect_schedule_slp): Update call accordingly.
15231 2017-09-18  Richard Sandiford  <richard.sandiford@linaro.org>
15232             Alan Hayward  <alan.hayward@arm.com>
15233             David Sherwood  <david.sherwood@arm.com>
15235         * tree-vect-data-refs.c (vect_get_smallest_scalar_type): Cope
15236         with types that aren't in fact scalar.
15238 2017-09-18  Richard Sandiford  <richard.sandiford@linaro.org>
15240         * tree-vect-slp.c (vect_record_max_nunits): New function,
15241         split out from...
15242         (vect_build_slp_tree_1): ...here.
15243         (vect_build_slp_tree_2): Call it for phis too.
15245 2017-09-18  Richard Sandiford  <richard.sandiford@linaro.org>
15247         * tree-vect-stmts.c (vectorizable_mask_load_store): Pass mask_vectype
15248         to vect_get_vec_def_for_operand when getting the mask operand.
15250 2017-09-18  Richard Sandiford  <richard.sandiford@linaro.org>
15251             Alan Hayward  <alan.hayward@arm.com>
15252             David Sherwood  <david.sherwood@arm.com>
15254         * tree-vect-loop.c (vectorizable_live_operation): Fix type of
15255         bitstart.
15257 2017-09-18  Richard Sandiford  <richard.sandiford@linaro.org>
15258             Alan Hayward  <alan.hayward@arm.com>
15259             David Sherwood  <david.sherwood@arm.com>
15261         * tree-vect-loop.c (vectorizable_live_operation): Fix element size
15262         calculation for vector booleans.
15264 2017-09-18  Richard Sandiford  <richard.sandiford@linaro.org>
15265             Alan Hayward  <alan.hayward@arm.com>
15266             David Sherwood  <david.sherwood@arm.com>
15268         * tree-vect-stmts.c (can_vectorize_live_stmts): New function,
15269         split out from...
15270         (vect_transform_stmt): ...here.
15271         (vect_analyze_stmt): Use it instead of calling
15272         vectorizable_live_operation directly.
15274 2017-09-18  Cesar Philippidis  <cesar@codesourcery.com>
15276         * omp-offload.c (oacc_xform_loop): Enable SIMD vectorization on
15277         non-SIMT targets in acc vector loops.
15279 2017-09-18  Claudiu Zissulescu  <claziss@synopsys.com>
15281         * configure.ac: Add arc and check if assembler supports gdwarf2.
15282         * configure: Regenerate.
15284 2017-09-18  Richard Biener  <rguenther@suse.de>
15286         PR tree-optimization/82220
15287         * tree-vect-loop.c (vect_estimate_min_profitable_iters): Exclude
15288         epilogue niters from the min_profitable_iters compute.
15290 2017-09-18  Jakub Jelinek  <jakub@redhat.com>
15292         PR target/82145
15293         * config/i386/i386.c (ix86_init_large_pic_reg): Revert 2017-09-01
15294         changes.  Turn CODE_LABEL into NOTE_INSN_DELETED_LABEL immediately.
15295         (ix86_init_pic_reg): Revert 2017-09-01 changes.
15297 2017-09-18  Eric Botcazou  <ebotcazou@adacore.com>
15299         PR target/81361
15300         * dwarf2cfi.c (add_cfis_to_fde): Do not generate DW_CFA_set_loc after
15301         switching to a new text section.
15303 2017-09-18  Richard Biener  <rguenther@suse.de>
15305         * graphite-scop-detection.c (scop_detection::stmt_has_simple_data_ref):
15306         Simplify.
15307         (build_alias_set): Reject aliases with no access function.
15309 2017-09-18  Richard Biener  <rguenther@suse.de>
15311         PR tree-optimization/79622
15312         * graphite-scop-detection.c (build_cross_bb_scalars_def): Properly
15313         handle PHIs.
15314         (build_cross_bb_scalars_use): Likewise.
15316 2017-09-18  Pierre-Marie de Rodat  <derodat@adacore.com>
15318         * cgraph.h (cgraph_thunk_info): Fix a typo in a comment.
15320 2017-09-18  Alan Modra  <amodra@gmail.com>
15322         PR target/81996
15323         * gcc/config/rs6000/rs6000.c (rs6000_return_addr): Use
15324         stack_pointer_rtx for count 0.  Update comments.  Break up
15325         large rtl expression.
15327 2017-09-17  Daniel Santos  <daniel.santos@pobox.com>
15329         * config/i386/i386.c (xlogue_layout::STUB_NAME_MAX_LEN):
15330         Increase to 20 bytes.
15331         (xlogue_layout::s_stub_names): Add an additional size-2 diminsion.
15332         (xlogue_layout::get_stub_name): Modify to select the appropairate sse
15333         or avx version of the stub.
15335 2017-09-17  H.J. Lu  <hongjiu.lu@intel.com>
15337         PR target/82166
15338         * config/i386/i386.c (ix86_finalize_stack_frame_flags): Properly
15339         compute the minimum stack alignment.  Also update preferred stack
15340         boundary for leaf functions.
15342 2017-09-16  Richard Sandiford  <richard.sandiford@linaro.org>
15344         PR tree-optimization/82228
15345         * tree-vect-loop.c (vectorizable_live_operation): Move initialization
15346         of ncopies.
15348 2017-09-16  Chung-Ju Wu  <jasonwucj@gmail.com>
15350         * common/config/nds32/nds32-common.c
15351         (nds32_option_optimization_table): Refine formatting.
15352         (nds32_option_optimization_table): Use -fsched-pressure and
15353         -fomit-frame-pointer for specific optimization level.
15355 2017-09-16  Chung-Ju Wu  <jasonwucj@gmail.com>
15357         * config/nds32/nds32.c: Refine formatting and comments.
15358         * config/nds32/nds32.h: Likewise.
15359         * config/nds32/nds32.md: Likewise.
15360         * config/nds32/nds32-cost.c: Likewise.
15361         * config/nds32/nds32-isr.c: Likewise.
15362         * config/nds32/nds32-md-auxiliary.c: Likewise.
15363         * config/nds32/nds32-multiple.md: Likewise.
15364         * config/nds32/nds32-predicates.c: Likewise.
15366 2017-09-15  Andrew Sutton  <andrew.n.sutton@gmail.com>
15367             Jakub Jelinek  <jakub@redhat.com>
15369         Add support for -std=c++2a.
15370         * doc/cpp.texi (__cplusplus): Document value for -std=c++2a
15371         or -std=gnu+2a.
15372         * doc/invoke.texi: Document -std=c++2a and -std=gnu++2a.
15374 2017-09-15  Steve Ellcey  <sellcey@cavium.com>
15376         PR target/82066
15377         * doc/extend.texi (Common Function Attributes): Add 
15378         references to ARM, AArch64, and S/390 specific attributes.
15379         (Function Specific Option Pragmas): Add AArch64 and S/390
15380         to list of back ends that support the target pragma.
15382 2017-09-15  Nathan Sidwell  <nathan@acm.org>
15384         * doc/standards.texi: Fix C++17 description.  Update URLs for
15385         C++11 & 14.
15387 2017-09-15  Bernd Edlinger  <bernd.edlinger@hotmail.de>
15389         * common.opt (Wcast-align=strict): New warning option.
15390         * doc/invoke.texi: Document -Wcast-align=strict. 
15392 2017-09-15  Pierre-Marie de Rodat  <derodat@adacore.com>
15394         * cgraph.h (cgraph_thunk_info): Add comments.
15395         * cgraph.c (cgraph_node::create_thunk): Adjust comment, make
15396         assert for VIRTUAL_* arguments stricter.
15398 2017-09-15  Jackson Woodruff  <jackson.woodruff@arm.com>
15400         PR tree-optimization/71026
15401         * match.pd: Move RDIV patterns from fold-const.c
15402         * fold-const.c (distribute_real_division): Removed.
15403         (fold_binary_loc): Remove calls to distribute_real_divison.
15405 2017-09-15  Jakub Jelinek  <jakub@redhat.com>
15407         * doc/invoke.texi: Document -std=c++17 and -std=gnu++17 and document
15408         c++1z and gnu++1z as deprecated.  Change other references to
15409         -std=c++1z to -std=c++17 and -std=gnu++1z to -std=gnu++17.
15410         Change -Wc++1z-compat to -Wc++17-compat.
15411         * doc/cpp.texi: Document -std=c++17 defines __cplusplus 201703L.
15412         * dwarf2out.c (highest_c_language): Handle C++17.
15413         (gen_compile_unit_die): Likewise.
15415 2017-09-15  Jakub Jelinek  <jakub@redhat.com>
15417         PR rtl-optimization/82192
15418         * combine.c (make_extraction): Don't look through non-paradoxical
15419         SUBREGs or TRUNCATE if pos + len is or might be bigger than
15420         inner's mode.
15422 2017-09-15  Richard Sandiford  <richard.sandiford@linaro.org>
15423             Alan Hayward  <alan.hayward@arm.com>
15424             David Sherwood  <david.sherwood@arm.com>
15426         * target.def (function_arg_offset): New hook.
15427         * targhooks.h (default_function_arg_offset): Declare.
15428         * targhooks.c (default_function_arg_offset): New function.
15429         * function.c (locate_and_pad_parm): Use
15430         targetm.calls.function_arg_offset instead of FUNCTION_ARG_OFFSET.
15431         * doc/tm.texi.in (FUNCTION_ARG_OFFSET): Replace with...
15432         (TARGET_FUNCTION_ARG_OFFSET): ...this.
15433         * doc/tm.texi: Regenerate.
15434         * config/spu/spu.h (FUNCTION_ARG_OFFSET): Delete.
15435         * config/spu/spu.c (spu_function_arg_offset): New function.
15436         (TARGET_FUNCTION_ARG_OFFSET): Redefine.
15437         * system.h (FUNCTION_ARG_OFFSET): Poison.
15439 2017-09-15  Richard Sandiford  <richard.sandiford@linaro.org>
15440             Alan Hayard  <alan.hayward@arm.com>
15441             David Sherwood  <david.sherwood@arm.com>
15443         * target.def (truly_noop_truncation): New hook.
15444         (mode_rep_extended): Refer to TARGET_TRULY_NOOP_TRUNCATION rather
15445         than TRULY_NOOP_TRUNCATION.
15446         * hooks.h (hook_bool_uint_uint_true): Declare.
15447         * hooks.c (hook_bool_uint_uint_true): New function.
15448         * doc/tm.texi.in (TRULY_NOOP_TRUNCATION): Replace with...
15449         (TARGET_TRULY_NOOP_TRUNCATION): ...this.
15450         * doc/tm.texi: Regenerate.
15451         * combine.c (make_extraction): Refer to TARGET_TRULY_NOOP_TRUNCATION
15452         rather than TRULY_NOOP_TRUNCATION in comments.
15453         (simplify_comparison): Likewise.
15454         (record_truncated_value): Likewise.
15455         * expmed.c (extract_bit_field_1): Likewise.
15456         (extract_split_bit_field): Likewise.
15457         * convert.c (convert_to_integer_1): Use targetm.truly_noop_truncation
15458         instead of TRULY_NOOP_TRUNCATION.
15459         * function.c (assign_parm_setup_block): Likewise.
15460         * machmode.h (TRULY_NOOP_TRUNCATION_MODES_P): Likewise.
15461         * rtlhooks.c: Include target.h.
15462         * config/aarch64/aarch64.h (TRULY_NOOP_TRUNCATION): Delete.
15463         * config/alpha/alpha.h (TRULY_NOOP_TRUNCATION): Delete.
15464         * config/arc/arc.h (TRULY_NOOP_TRUNCATION): Delete.
15465         * config/arm/arm.h (TRULY_NOOP_TRUNCATION): Delete.
15466         * config/avr/avr.h (TRULY_NOOP_TRUNCATION): Delete.
15467         * config/bfin/bfin.h (TRULY_NOOP_TRUNCATION): Delete.
15468         * config/c6x/c6x.h (TRULY_NOOP_TRUNCATION): Delete.
15469         * config/cr16/cr16.h (TRULY_NOOP_TRUNCATION): Delete.
15470         * config/cris/cris.h (TRULY_NOOP_TRUNCATION): Delete.
15471         * config/epiphany/epiphany.h (TRULY_NOOP_TRUNCATION): Delete.
15472         * config/fr30/fr30.h (TRULY_NOOP_TRUNCATION): Delete.
15473         * config/frv/frv.h (TRULY_NOOP_TRUNCATION): Delete.
15474         * config/ft32/ft32.h (TRULY_NOOP_TRUNCATION): Delete.
15475         * config/h8300/h8300.h (TRULY_NOOP_TRUNCATION): Delete.
15476         * config/i386/i386.h (TRULY_NOOP_TRUNCATION): Delete.
15477         * config/ia64/ia64.h (TRULY_NOOP_TRUNCATION): Delete.
15478         * config/iq2000/iq2000.h (TRULY_NOOP_TRUNCATION): Delete.
15479         * config/lm32/lm32.h (TRULY_NOOP_TRUNCATION): Delete.
15480         * config/m32c/m32c.h (TRULY_NOOP_TRUNCATION): Delete.
15481         * config/m32r/m32r.h (TRULY_NOOP_TRUNCATION): Delete.
15482         * config/m68k/m68k.h (TRULY_NOOP_TRUNCATION): Delete.
15483         * config/mcore/mcore.h (TRULY_NOOP_TRUNCATION): Delete.
15484         * config/microblaze/microblaze.h (TRULY_NOOP_TRUNCATION): Delete.
15485         * config/mips/mips.h (TRULY_NOOP_TRUNCATION): Delete.
15486         * config/mips/mips.c (mips_truly_noop_truncation): New function.
15487         (TARGET_TRULY_NOOP_TRUNCATION): Redefine.
15488         * config/mips/mips.md: Refer to TARGET_TRULY_NOOP_TRUNCATION
15489         rather than TRULY_NOOP_TRUNCATION in comments.
15490         * config/mmix/mmix.h (TRULY_NOOP_TRUNCATION): Delete.
15491         * config/mn10300/mn10300.h (TRULY_NOOP_TRUNCATION): Delete.
15492         * config/moxie/moxie.h (TRULY_NOOP_TRUNCATION): Delete.
15493         * config/msp430/msp430.h (TRULY_NOOP_TRUNCATION): Delete.
15494         * config/nds32/nds32.h (TRULY_NOOP_TRUNCATION): Delete.
15495         * config/nios2/nios2.h (TRULY_NOOP_TRUNCATION): Delete.
15496         * config/nvptx/nvptx.h (TRULY_NOOP_TRUNCATION): Delete.
15497         * config/pa/pa.h (TRULY_NOOP_TRUNCATION): Delete.
15498         * config/pdp11/pdp11.h (TRULY_NOOP_TRUNCATION): Delete.
15499         * config/powerpcspe/powerpcspe.h (TRULY_NOOP_TRUNCATION): Delete.
15500         * config/riscv/riscv.h (TRULY_NOOP_TRUNCATION): Delete.
15501         * config/riscv/riscv.md: Refer to TARGET_TRULY_NOOP_TRUNCATION
15502         rather than TRULY_NOOP_TRUNCATION in comments.
15503         * config/rl78/rl78.h (TRULY_NOOP_TRUNCATION): Delete.
15504         * config/rs6000/rs6000.h (TRULY_NOOP_TRUNCATION): Delete.
15505         * config/rx/rx.h (TRULY_NOOP_TRUNCATION): Delete.
15506         * config/s390/s390.h (TRULY_NOOP_TRUNCATION): Delete.
15507         * config/sh/sh.h (MAYBE_BASE_REGISTER_RTX_P): Remove
15508         TRULY_NOOP_TRUNCATION condition.
15509         (MAYBE_INDEX_REGISTER_RTX_P): Likewise.
15510         (TRULY_NOOP_TRUNCATION): Delete.
15511         * config/sparc/sparc.h (TRULY_NOOP_TRUNCATION): Delete.
15512         * config/spu/spu.h (TRULY_NOOP_TRUNCATION): Delete.
15513         * config/spu/spu.c (spu_truly_noop_truncation): New function.
15514         (TARGET_TRULY_NOOP_TRUNCATION): Redefine.
15515         * config/stormy16/stormy16.h (TRULY_NOOP_TRUNCATION): Delete.
15516         * config/tilegx/tilegx.h (TRULY_NOOP_TRUNCATION): Delete.
15517         * config/tilegx/tilegx.c (tilegx_truly_noop_truncation): New fuction.
15518         (TARGET_TRULY_NOOP_TRUNCATION): Redefine.
15519         * config/tilegx/tilegx.md: Refer to TARGET_TRULY_NOOP_TRUNCATION
15520         rather than TRULY_NOOP_TRUNCATION in comments.
15521         * config/tilepro/tilepro.h (TRULY_NOOP_TRUNCATION): Delete.
15522         * config/v850/v850.h (TRULY_NOOP_TRUNCATION): Delete.
15523         * config/vax/vax.h (TRULY_NOOP_TRUNCATION): Delete.
15524         * config/visium/visium.h (TRULY_NOOP_TRUNCATION): Delete.
15525         * config/xtensa/xtensa.h (TRULY_NOOP_TRUNCATION): Delete.
15526         * system.h (TRULY_NOOP_TRUNCATION): Poison.
15528 2017-09-15  Christophe Lyon  <christophe.lyon@linaro.org>
15530         PR target/67591
15531         * config/arm/arm.md (*cmp_and): Add enabled_for_depr_it attribute.
15532         (*cmp_ior): Likewise.
15533         (*ior_scc_scc): Add alternative for enabled_for_depr_it attribute.
15534         (*ior_scc_scc_cmp): Likewise.
15535         (*and_scc_scc): Likewise.
15536         (*and_scc_scc_cmp): Likewise.
15538 2017-09-15  Richard Sandiford  <richard.sandiford@linaro.org>
15539             Alan Hayard  <alan.hayward@arm.com>
15540             David Sherwood  <david.sherwood@arm.com>
15542         * target.def (can_change_mode_class): New hook.
15543         (mode_rep_extended): Refer to it instead of CANNOT_CHANGE_MODE_CLASS.
15544         (hard_regno_nregs): Likewise.
15545         * hooks.h (hook_bool_mode_mode_reg_class_t_true): Declare.
15546         * hooks.c (hook_bool_mode_mode_reg_class_t_true): New function.
15547         * doc/tm.texi.in (CANNOT_CHANGE_MODE_CLASS): Replace with...
15548         (TARGET_CAN_CHANGE_MODE_CLASS): ...this.
15549         (LOAD_EXTEND_OP): Update accordingly.
15550         * doc/tm.texi: Regenerate.
15551         * doc/rtl.texi: Refer to TARGET_CAN_CHANGE_MODE_CLASS instead of
15552         CANNOT_CHANGE_MODE_CLASS.
15553         * hard-reg-set.h (REG_CANNOT_CHANGE_MODE_P): Replace with...
15554         (REG_CAN_CHANGE_MODE_P): ...this new macro.
15555         * combine.c (simplify_set): Update accordingly.
15556         * emit-rtl.c (validate_subreg): Likewise.
15557         * recog.c (general_operand): Likewise.
15558         * regcprop.c (mode_change_ok): Likewise.
15559         * reload1.c (choose_reload_regs): Likewise.
15560         (inherit_piecemeal_p): Likewise.
15561         * rtlanal.c (simplify_subreg_regno): Likewise.
15562         * postreload.c (reload_cse_simplify_set): Use REG_CAN_CHANGE_MODE_P
15563         instead of CANNOT_CHANGE_MODE_CLASS.
15564         (reload_cse_simplify_operands): Likewise.
15565         * reload.c (push_reload): Use targetm.can_change_mode_class
15566         instead of CANNOT_CHANGE_MODE_CLASS.
15567         (push_reload): Likewise.  Also use REG_CAN_CHANGE_MODE_P instead of
15568         REG_CANNOT_CHANGE_MODE_P.
15569         * config/alpha/alpha.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15570         * config/alpha/alpha.c (alpha_can_change_mode_class): New function.
15571         (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15572         * config/arm/arm.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15573         * config/arm/arm.c (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15574         (arm_can_change_mode_class): New function.
15575         * config/arm/neon.md: Refer to TARGET_CAN_CHANGE_MODE_CLASS rather
15576         than CANNOT_CHANGE_MODE_CLASS in comments.
15577         * config/i386/i386.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15578         * config/i386/i386-protos.h (ix86_cannot_change_mode_class): Delete.
15579         * config/i386/i386.c (ix86_cannot_change_mode_class): Replace with...
15580         (ix86_can_change_mode_class): ...this new function, inverting the
15581         sense of the return value.
15582         (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15583         * config/ia64/ia64.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15584         * config/ia64/ia64.c (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15585         (ia64_can_change_mode_class): New function.
15586         * config/m32c/m32c.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15587         * config/m32c/m32c-protos.h (m32c_cannot_change_mode_class): Delete.
15588         * config/m32c/m32c.c (m32c_cannot_change_mode_class): Replace with...
15589         (m32c_can_change_mode_class): ...this new function, inverting the
15590         sense of the return value.
15591         (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15592         * config/mips/mips.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15593         * config/mips/mips-protos.h (mips_cannot_change_mode_class): Delete.
15594         * config/mips/mips.c (mips_cannot_change_mode_class): Replace with...
15595         (mips_can_change_mode_class): ...this new function, inverting the
15596         sense of the return value.
15597         (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15598         * config/msp430/msp430.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15599         * config/msp430/msp430.c (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15600         (msp430_can_change_mode_class): New function.
15601         * config/nvptx/nvptx.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15602         * config/nvptx/nvptx.c (nvptx_can_change_mode_class): New function.
15603         (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15604         * config/pa/pa32-regs.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15605         * config/pa/pa64-regs.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15606         * config/pa/pa-protos.h (pa_cannot_change_mode_class): Delete.
15607         * config/pa/pa.c (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15608         (pa_cannot_change_mode_class): Replace with...
15609         (pa_can_change_mode_class): ...this new function, inverting the
15610         sense of the return value.
15611         (pa_modes_tieable_p): Refer to TARGET_CAN_CHANGE_MODE_CLASS rather
15612         than CANNOT_CHANGE_MODE_CLASS in comments.
15613         * config/pdp11/pdp11.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15614         * config/pdp11/pdp11-protos.h (pdp11_cannot_change_mode_class): Delete.
15615         * config/pdp11/pdp11.c (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15616         (pdp11_cannot_change_mode_class): Replace with...
15617         (pdp11_can_change_mode_class): ...this new function, inverting the
15618         sense of the return value.
15619         * config/powerpcspe/powerpcspe.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15620         * config/powerpcspe/powerpcspe-protos.h
15621         (rs6000_cannot_change_mode_class_ptr): Delete.
15622         * config/powerpcspe/powerpcspe.c
15623         (rs6000_cannot_change_mode_class_ptr): Delete.
15624         (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15625         (rs6000_option_override_internal): Assign to
15626         targetm.can_change_mode_class instead of
15627         rs6000_cannot_change_mode_class_ptr.
15628         (rs6000_cannot_change_mode_class): Replace with...
15629         (rs6000_can_change_mode_class): ...this new function, inverting the
15630         sense of the return value.
15631         (rs6000_debug_cannot_change_mode_class): Replace with...
15632         (rs6000_debug_can_change_mode_class): ...this new function.
15633         * config/riscv/riscv.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15634         * config/riscv/riscv.c (riscv_can_change_mode_class): New function.
15635         (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15636         * config/rs6000/rs6000.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15637         * config/rs6000/rs6000-protos.h (rs6000_cannot_change_mode_class_ptr):
15638         Delete.
15639         * config/rs6000/rs6000.c (rs6000_cannot_change_mode_class_ptr): Delete.
15640         (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15641         (rs6000_option_override_internal): Assign to
15642         targetm.can_change_mode_class instead of
15643         rs6000_cannot_change_mode_class_ptr.
15644         (rs6000_cannot_change_mode_class): Replace with...
15645         (rs6000_can_change_mode_class): ...this new function, inverting the
15646         sense of the return value.
15647         (rs6000_debug_cannot_change_mode_class): Replace with...
15648         (rs6000_debug_can_change_mode_class): ...this new function.
15649         * config/s390/s390.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15650         * config/s390/s390-protos.h (s390_cannot_change_mode_class): Delete.
15651         * config/s390/s390.c (s390_cannot_change_mode_class): Replace with...
15652         (s390_can_change_mode_class): ...this new function, inverting the
15653         sense of the return value.
15654         (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15655         * config/sh/sh.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15656         * config/sh/sh-protos.h (sh_cannot_change_mode_class): Delete.
15657         * config/sh/sh.c (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15658         (sh_cannot_change_mode_class): Replace with...
15659         (sh_can_change_mode_class): ...this new function, inverting the
15660         sense of the return value.
15661         * config/sparc/sparc.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15662         * config/sparc/sparc.c (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15663         (sparc_can_change_mode_class): New function.
15664         * config/spu/spu.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15665         * config/spu/spu.c (spu_can_change_mode_class): New function.
15666         (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15667         * config/visium/visium.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15668         * config/visium/visium.c (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15669         (visium_can_change_mode_class): New function.
15670         * system.h (CANNOT_CHANGE_MODE_CLASS): Poison.
15672 2017-09-15  Richard Biener  <rguenther@suse.de>
15674         PR tree-optimization/82217
15675         * tree-ssa-sccvn.c (visit_phi): Properly handle all VN_TOP
15676         but not undefined case.
15678 2017-09-15  Jakub Jelinek  <jakub@redhat.com>
15680         PR target/82145
15681         * postreload.c (reload_cse_simplify_operands): Skip
15682         NOTE_INSN_DELETED_LABEL similarly to skipping CODE_LABEL.
15684 2017-09-15  Richard Biener  <rguenther@suse.de>
15686         PR tree-optimization/68823
15687         * graphite-scop-detection.c (build_alias_set): If we have a
15688         possible dependence check whether we can handle them by just
15689         looking at the DRs DR_ACCESS_FNs.
15690         (build_scops): If build_alias_set fails, fail the SCOP.
15692 2017-09-14  Michael Meissner  <meissner@linux.vnet.ibm.com>
15694         * config/rs6000/rs6000-builtin.def (BU_FLOAT128_1_HW): New macros
15695         to support float128 built-in functions that require the ISA 3.0
15696         hardware.
15697         (BU_FLOAT128_3_HW): Likewise.
15698         (SQRTF128): Add support for the IEEE 128-bit square root and fma
15699         built-in functions.
15700         (FMAF128): Likewise.
15701         (FMAQ): Likewise.
15702         * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add
15703         support for built-in functions that need the ISA 3.0 IEEE 128-bit
15704         floating point instructions.
15705         (rs6000_invalid_builtin): Likewise.
15706         (rs6000_builtin_mask_names): Likewise.
15707         * config/rs6000/rs6000.h (MASK_FLOAT128_HW): Likewise.
15708         (RS6000_BTM_FLOAT128_HW): Likewise.
15709         (RS6000_BTM_COMMON): Likewise.
15710         * config/rs6000/rs6000.md (fma<mode>4_hw): Add a generator
15711         function.
15712         * doc/extend.texi (RS/6000 built-in functions): Document the
15713         IEEE 128-bit floating point square root and fused multiply-add
15714         built-in functions.
15716 2017-09-14  Pat Haugen  <pthaugen@us.ibm.com>
15718         * config/rs6000/rs6000.c (rs6000_set_up_by_prologue): Make sure the TOC
15719         reg (r2) isn't in the set of registers defined in the prologue.
15721 2017-09-14  Richard Sandiford  <richard.sandiford@linaro.org>
15722             Alan Hayward  <alan.hayward@arm.com>
15723             David Sherwood  <david.sherwood@arm.com>
15725         * tree-vectorizer.h (_loop_vec_info): Add max_vectorization_factor.
15726         (LOOP_VINFO_MAX_VECT_FACTOR): New macro.
15727         (LOOP_VINFO_ORIG_VECT_FACTOR): Replace with...
15728         (LOOP_VINFO_ORIG_MAX_VECT_FACTOR): ...this new macro.
15729         * tree-vect-data-refs.c (vect_analyze_data_ref_dependences): Update
15730         accordingly.
15731         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
15732         max_vectorization_factor.
15733         (vect_analyze_loop_2): Set LOOP_VINFO_MAX_VECT_FACTOR.
15735 2017-09-14  Richard Sandiford  <richard.sandiford@linaro.org>
15736             Alan Hayward  <alan.hayward@arm.com>
15737             David Sherwood  <david.sherwood@arm.com>
15739         * tree-vectorizer.h (vect_min_worthwhile_factor): Delete.
15740         (vect_worthwhile_without_simd_p): Declare.
15741         * tree-vect-loop.c (vect_worthwhile_without_simd_p): New function.
15742         (vectorizable_reduction): Use it.
15743         * tree-vect-stmts.c (vectorizable_shift): Likewise.
15744         (vectorizable_operation): Likewise.
15746 2017-09-14  Richard Sandiford  <richard.sandiford@linaro.org>
15747             Alan Hayward  <alan.hayward@arm.com>
15748             David Sherwood  <david.sherwood@arm.com>
15750         * tree-vectorizer.h (vect_get_num_copies): New function.
15751         * tree-vect-data-refs.c (vect_get_data_access_cost): Use it.
15752         * tree-vect-loop.c (vectorizable_reduction): Likewise.
15753         (vectorizable_induction): Likewise.
15754         (vectorizable_live_operation): Likewise.
15755         * tree-vect-stmts.c (vectorizable_mask_load_store): Likewise.
15756         (vectorizable_bswap): Likewise.
15757         (vectorizable_call): Likewise.
15758         (vectorizable_conversion): Likewise.
15759         (vectorizable_assignment): Likewise.
15760         (vectorizable_shift): Likewise.
15761         (vectorizable_operation): Likewise.
15762         (vectorizable_store): Likewise.
15763         (vectorizable_load): Likewise.
15764         (vectorizable_condition): Likewise.
15765         (vectorizable_comparison): Likewise.
15766         (vect_analyze_stmt): Pass the slp node to vectorizable_live_operation.
15768 2017-09-14  Richard Sandiford  <richard.sandiford@linaro.org>
15769             Alan Hayward  <alan.hayward@arm.com>
15770             David Sherwood  <david.sherwood@arm.com>
15772         * tree-vect-loop.c (vectorizable_induction): Use gimple_build instead
15773         of vect_init_vector.
15775 2017-09-14  Richard Sandiford  <richard.sandiford@linaro.org>
15776             Alan Hayward  <alan.hayward@arm.com>
15777             David Sherwood  <david.sherwood@arm.com>
15779         * gimple-fold.h (gimple_build_vector_from_val): Declare, and provide
15780         an inline wrapper that provides a location.
15781         (gimple_build_vector): Likewise.
15782         * gimple-fold.c (gimple_build_vector_from_val): New function.
15783         (gimple_build_vector): Likewise.
15784         * tree-vect-loop.c (get_initial_def_for_reduction): Use the new
15785         functions to build the initial value.  Always return a gimple value.
15786         (get_initial_defs_for_reduction): Likewise.  Only compute
15787         neutral_vec once.
15788         (vect_create_epilog_for_reduction): Don't call force_gimple_operand or
15789         vect_init_vector on the results from get_initial_def(s)_for_reduction.
15790         (vectorizable_induction): Use gimple_build_vector rather than
15791         vect_init_vector.
15793 2017-09-14  Richard Sandiford  <richard.sandiford@linaro.org>
15794             Alan Hayward  <alan.hayward@arm.com>
15795             David Sherwood  <david.sherwood@arm.com>
15797         * target.h (vec_perm_indices): New typedef.
15798         (auto_vec_perm_indices): Likewise.
15799         * optabs-query.h: Include target.h
15800         (can_vec_perm_p): Take a vec_perm_indices *.
15801         * optabs-query.c (can_vec_perm_p): Likewise.
15802         (can_mult_highpart_p): Update accordingly.  Use auto_vec_perm_indices.
15803         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
15804         * tree-vect-generic.c (lower_vec_perm): Likewise.
15805         * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
15806         (vect_grouped_load_supported): Likewise.
15807         (vect_shift_permute_load_chain): Likewise.
15808         (vect_permute_store_chain): Use auto_vec_perm_indices.
15809         (vect_permute_load_chain): Likewise.
15810         * fold-const.c (fold_vec_perm): Take vec_perm_indices.
15811         (fold_ternary_loc): Update accordingly.  Use auto_vec_perm_indices.
15812         Update uses of can_vec_perm_p.
15813         * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Replace the
15814         mode with a number of elements.  Take a vec_perm_indices *.
15815         (vect_create_epilog_for_reduction): Update accordingly.
15816         Use auto_vec_perm_indices.
15817         (have_whole_vector_shift): Likewise.  Update call to can_vec_perm_p.
15818         * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
15819         (vect_transform_slp_perm_load): Likewise.
15820         (vect_schedule_slp_instance): Use auto_vec_perm_indices.
15821         * tree-vectorizer.h (vect_gen_perm_mask_any): Take a vec_perm_indices.
15822         (vect_gen_perm_mask_checked): Likewise.
15823         * tree-vect-stmts.c (vect_gen_perm_mask_any): Take a vec_perm_indices.
15824         (vect_gen_perm_mask_checked): Likewise.
15825         (vectorizable_mask_load_store): Use auto_vec_perm_indices.
15826         (vectorizable_store): Likewise.
15827         (vectorizable_load): Likewise.
15828         (perm_mask_for_reverse): Likewise.  Update call to can_vec_perm_p.
15829         (vectorizable_bswap): Likewise.
15831 2017-09-14  Richard Sandiford  <richard.sandiford@linaro.org>
15832             Alan Hayward  <alan.hayward@arm.com>
15833             David Sherwood  <david.sherwood@arm.com>
15835         * tree.h (build_vector): Take a vec<tree> instead of a tree *.
15836         * tree.c (build_vector): Likewise.
15837         (build_vector_from_ctor): Update accordingly.
15838         (build_vector_from_val): Likewise.
15839         * gimple-fold.c (gimple_fold_stmt_to_constant_1): Likewise.
15840         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
15841         * tree-vect-generic.c (add_rshift): Likewise.
15842         (expand_vector_divmod): Likewise.
15843         (optimize_vector_constructor): Likewise.
15844         * tree-vect-slp.c (vect_get_constant_vectors): Likewise.
15845         (vect_transform_slp_perm_load): Likewise.
15846         (vect_schedule_slp_instance): Likewise.
15847         * tree-vect-stmts.c (vectorizable_bswap): Likewise.
15848         (vectorizable_call): Likewise.
15849         (vect_gen_perm_mask_any): Likewise.  Add elements in order.
15850         * expmed.c (make_tree): Likewise.
15851         * fold-const.c (fold_negate_expr_1): Use auto_vec<tree> when building
15852         a vector passed to build_vector.
15853         (fold_convert_const): Likewise.
15854         (exact_inverse): Likewise.
15855         (fold_ternary_loc): Likewise.
15856         (fold_relational_const): Likewise.
15857         (const_binop): Likewise.  Use VECTOR_CST_ELT directly when operating
15858         on VECTOR_CSTs, rather than going through vec_cst_ctor_to_array.
15859         (const_unop): Likewise.  Store the reduction accumulator in a
15860         variable rather than an array.
15861         (vec_cst_ctor_to_array): Take the number of elements as a parameter.
15862         (fold_vec_perm): Update calls accordingly.  Use auto_vec<tree> for
15863         the new vector, rather than constructing it after the input arrays.
15864         (native_interpret_vector): Use auto_vec<tree> when building
15865         a vector passed to build_vector.  Add elements in order.
15866         * tree-vect-loop.c (get_initial_defs_for_reduction): Use
15867         auto_vec<tree> when building a vector passed to build_vector.
15868         (vect_create_epilog_for_reduction): Likewise.
15869         (vectorizable_induction): Likewise.
15870         (get_initial_def_for_reduction): Likewise.  Fix indentation of
15871         case statements.
15872         * config/sparc/sparc.c (sparc_handle_vis_mul8x16): Change n_elts
15873         to a vec<tree> *.
15874         (sparc_fold_builtin): Use auto_vec<tree> when building a vector
15875         passed to build_vector.
15877 2017-09-14  Richard Sandiford  <richard.sandiford@linaro.org>
15878             Alan Hayward  <alan.hayward@arm.com>
15879             David Sherwood  <david.sherwood@arm.com>
15881         * tree-core.h (tree_base::u): Add an "nelts" field.
15882         (tree_vector): Use VECTOR_CST_NELTS as the length.
15883         * tree.c (tree_size): Likewise.
15884         (make_vector): Initialize VECTOR_CST_NELTS.
15885         * tree.h (VECTOR_CST_NELTS): Use the u.nelts field.
15886         * cfgexpand.c (expand_debug_expr): Use VECTOR_CST_NELTS instead of
15887         TYPE_VECTOR_SUBPARTS.
15888         * expr.c (const_vector_mask_from_tree): Consistently use "units"
15889         as the number of units, setting it from VECTOR_CST_NELTS.
15890         (const_vector_from_tree): Likewise.
15891         * fold-const.c (negate_expr_p): Use VECTOR_CST_NELTS instead of
15892         TYPE_VECTOR_SUBPARTS for the number of elements in a VECTOR_CST.
15893         (fold_negate_expr_1): Likewise.
15894         (fold_convert_const): Likewise.
15895         (const_binop): Likewise.  Differentiate the number of output and
15896         input elements.
15897         (const_unop): Likewise.
15898         (fold_ternary_loc): Use VECTOR_CST_NELTS for the number of elements
15899         in a VECTOR_CST, asserting that it is the same as TYPE_VECTOR_SUBPARTS
15900         in cases that did the opposite.
15902 2017-09-14  Richard Biener  <rguenther@suse.de>
15904         * tree-ssa-sccvn.c (visit_phi): Merge undefined values similar
15905         to VN_TOP.
15907 2017-09-14  Eric Botcazou  <ebotcazou@adacore.com>
15909         * dwarf2out.c (dwarf2out_source_line): Remove superfluous test.
15911 2017-09-14  Jakub Jelinek  <jakub@redhat.com>
15913         PR target/81325
15914         * cfgbuild.c (find_bb_boundaries): Ignore debug insns in decisions
15915         if and where to split a bb, except for splitting before debug insn
15916         sequences followed by non-label real insn.  Delete debug insns
15917         in between basic blocks.
15919         * combine.c (make_compound_operation_int): Formatting fixes.
15921         * config/alpha/elf.h (LINK_EH_SPEC): Add -static-pie support.
15922         * config/alpha/linux.h (LINK_GCC_C_SEQUENCE_SPEC): Likewise.
15923         * config/netbsd.h (LINK_EH_SPEC): Likewise.
15924         * config/sol2.h (LINK_EH_SPEC): Likewise.
15925         * config/arm/uclinux-elf.h (LINK_GCC_C_SEQUENCE_SPEC): Likewise.
15926         * config/s390/linux.h (LINK_SPEC): Likewise.
15927         * config/freebsd.h (LINK_EH_SPEC): Likewise.
15928         * config/openbsd.h (LINK_EH_SPEC): Likewise.
15929         * config/lm32/uclinux-elf.h (LINK_GCC_C_SEQUENCE_SPEC): Likewise.
15930         * config/aarch64/aarch64-linux.h (LINUX_TARGET_LINK_SPEC): Likewise.
15931         * config/powerpcspe/sysv4.h (LINK_EH_SPEC): Likewise.
15932         * config/bfin/linux.h (LINK_GCC_C_SEQUENCE_SPEC): Likewise.
15933         * config/i386/gnu-user64.h (GNU_USER_TARGET_LINK_SPEC): Fix a typo.
15934         * config/i386/gnu-user.h (GNU_USER_TARGET_LINK_SPEC): Formatting fix.
15936 2017-09-13  Jakub Jelinek  <jakub@redhat.com>
15938         * config/rs6000/sysv4.h (STARTFILE_LINUX_SPEC): Add -static-pie
15939         support.
15940         (ENDFILE_LINUX_SPEC): Likewise.
15941         (LINK_EH_SPEC): Likewise.
15942         * config/rs6000/linux64.h (LINK_SHLIB_SPEC): Likewise.
15943         (LINK_OS_LINUX_SPEC32): Likewise.
15944         (LINK_OS_LINUX_SPEC64): Likewise.
15945         * config/rs6000/linux.h (LINK_SHLIB_SPEC): Likewise.
15946         (LINK_OS_LINUX_SPEC): Likewise.
15948 2017-09-13  Martin Liska  <mliska@suse.cz>
15950         PR middle-end/82154
15951         * stmt.c (expand_sjlj_dispatch_table): Use CASE_LOW when
15952         CASE_HIGH is NULL_TREE.
15954 2017-09-13  Richard Sandiford  <richard.sandiford@linaro.org>
15955             Alan Hayward  <alan.hayward@arm.com>
15956             David Sherwood  <david.sherwood@arm.com>
15958         * target.def (secondary_memory_needed): New hook.
15959         (secondary_reload): Refer to TARGET_SECONDARY_MEMORY_NEEDED
15960         instead of SECONDARY_MEMORY_NEEDED.
15961         (secondary_memory_needed_mode): Likewise.
15962         * hooks.h (hook_bool_mode_reg_class_t_reg_class_t_false): Declare.
15963         * hooks.c (hook_bool_mode_reg_class_t_reg_class_t_false): New function.
15964         * doc/tm.texi.in (SECONDARY_MEMORY_NEEDED): Replace with...
15965         (TARGET_SECONDARY_MEMORY_NEEDED): ...this.
15966         (SECONDARY_MEMORY_NEEDED_RTX): Update reference accordingly.
15967         * doc/tm.texi: Regenerate.
15968         * config/alpha/alpha.h (SECONDARY_MEMORY_NEEDED): Delete.
15969         * config/alpha/alpha.c (alpha_secondary_memory_needed): New function.
15970         (TARGET_SECONDARY_MEMORY_NEEDED): Redefine.
15971         * config/i386/i386.h (SECONDARY_MEMORY_NEEDED): Delete.
15972         * config/i386/i386-protos.h (ix86_secondary_memory_needed): Delete.
15973         * config/i386/i386.c (inline_secondary_memory_needed): Put the
15974         mode argument first and change the reg_class arguments to reg_class_t.
15975         (ix86_secondary_memory_needed): Likewise.  Remove the strict parameter.
15976         Make static.  Update the call to inline_secondary_memory_needed.
15977         (ix86_register_move_cost): Update the call to
15978         inline_secondary_memory_needed.
15979         (TARGET_SECONDARY_MEMORY_NEEDED): Redefine.
15980         * config/ia64/ia64.h (SECONDARY_MEMORY_NEEDED): Delete commented-out
15981         definition.
15982         * config/ia64/ia64.c (spill_xfmode_rfmode_operand): Refer to
15983         TARGET_SECONDARY_MEMORY_NEEDED rather than SECONDARY_MEMORY_NEEDED
15984         in comment.
15985         * config/mips/mips.h (SECONDARY_MEMORY_NEEDED): Delete.
15986         * config/mips/mips-protos.h (mips_secondary_memory_needed): Delete.
15987         * config/mips/mips.c (mips_secondary_memory_needed): Make static
15988         and match hook interface.  Add comment from mips.h.
15989         (TARGET_SECONDARY_MEMORY_NEEDED): Redefine.
15990         * config/mmix/mmix.md (truncdfsf2): Refer to
15991         TARGET_SECONDARY_MEMORY_NEEDED rather than SECONDARY_MEMORY_NEEDED
15992         in comment.
15993         * config/pa/pa-64.h (SECONDARY_MEMORY_NEEDED): Rename to...
15994         (PA_SECONDARY_MEMORY_NEEDED): ...this, and put the mode argument first.
15995         * config/pa/pa.c (TARGET_SECONDARY_MEMORY_NEEDED): Redefine.
15996         (pa_secondary_memory_needed): New function.
15997         * config/pdp11/pdp11.h (SECONDARY_MEMORY_NEEDED): Delete.
15998         * config/pdp11/pdp11-protos.h (pdp11_secondary_memory_needed): Delete.
15999         * config/pdp11/pdp11.c (TARGET_SECONDARY_MEMORY_NEEDED): Redefine.
16000         (pdp11_secondary_memory_needed): Make static and match hook interface.
16001         * config/powerpcspe/powerpcspe.h (SECONDARY_MEMORY_NEEDED): Delete.
16002         * config/powerpcspe/powerpcspe-protos.h
16003         (rs6000_secondary_memory_needed_ptr): Delete.
16004         * config/powerpcspe/powerpcspe.c (rs6000_secondary_memory_needed_ptr):
16005         Delete.
16006         (TARGET_SECONDARY_MEMORY_NEEDED): Redefine.
16007         (rs6000_option_override_internal): Assign to
16008         targetm.secondary_memory_needed rather than
16009         rs6000_secondary_memory_needed_ptr.
16010         (rs6000_secondary_memory_needed): Match hook interface.
16011         (rs6000_debug_secondary_memory_needed): Likewise.
16012         * config/riscv/riscv.h (SECONDARY_MEMORY_NEEDED): Delete.
16013         * config/riscv/riscv.c (riscv_secondary_memory_needed): New function.
16014         (riscv_register_move_cost): Use it instead of SECONDARY_MEMORY_NEEDED.
16015         (TARGET_SECONDARY_MEMORY_NEEDED): Redefine.
16016         * config/rs6000/rs6000.h (SECONDARY_MEMORY_NEEDED): Delete.
16017         * config/rs6000/rs6000-protos.h (rs6000_secondary_memory_needed_ptr):
16018         Delete.
16019         * config/rs6000/rs6000.c (rs6000_secondary_memory_needed_ptr): Delete.
16020         (TARGET_SECONDARY_MEMORY_NEEDED): Redefine.
16021         (rs6000_option_override_internal): Assign to
16022         targetm.secondary_memory_needed rather than
16023         rs6000_secondary_memory_needed_ptr.
16024         (rs6000_secondary_memory_needed): Match hook interface.
16025         (rs6000_debug_secondary_memory_needed): Likewise.
16026         * config/s390/s390.h (SECONDARY_MEMORY_NEEDED): Delete.
16027         * config/s390/s390.c (s390_secondary_memory_needed): New function.
16028         (TARGET_SECONDARY_MEMORY_NEEDED): Redefine.
16029         * config/sparc/sparc.h (SECONDARY_MEMORY_NEEDED): Delete.
16030         * config/sparc/sparc.c (TARGET_SECONDARY_MEMORY_NEEDED): Redefine.
16031         (sparc_secondary_memory_needed): New function.
16032         * lra-constraints.c (check_and_process_move): Refer to
16033         TARGET_SECONDARY_MEMORY_NEEDED rather than SECONDARY_MEMORY_NEEDED
16034         in comment.
16035         (curr_insn_transform): Likewise.
16036         (process_alt_operands): Use targetm.secondary_memory_needed
16037         instead of TARGET_SECONDARY_MEMORY_NEEDED.
16038         (check_secondary_memory_needed_p): Likewise.
16039         (choose_split_class): Likewise.
16040         * reload.c: Unconditionally include code that was previously
16041         conditional on SECONDARY_MEMORY_NEEDED.
16042         (push_secondary_reload): Use targetm.secondary_memory_needed
16043         instead of TARGET_SECONDARY_MEMORY_NEEDED.
16044         (push_reload): Likewise.
16045         * reload1.c: Unconditionally include code that was previously
16046         conditional on SECONDARY_MEMORY_NEEDED.
16047         (choose_reload_regs): Use targetm.secondary_memory_needed
16048         instead of TARGET_SECONDARY_MEMORY_NEEDED.
16049         (gen_reload): Likewise.
16050         * system.h (SECONDARY_MEMORY_NEEDED): Poison.
16052 2017-09-13  Richard Sandiford  <richard.sandiford@linaro.org>
16053             Alan Hayward  <alan.hayward@arm.com>
16054             David Sherwood  <david.sherwood@arm.com>
16056         * target.def (secondary_memory_needed_mode): New hook:
16057         * targhooks.c (default_secondary_memory_needed_mode): Declare.
16058         * targhooks.h (default_secondary_memory_needed_mode): New function.
16059         * doc/tm.texi.in (SECONDARY_MEMORY_NEEDED_MODE): Replace with...
16060         (TARGET_SECONDARY_MEMORY_NEEDED_MODE): ...this.
16061         * doc/tm.texi: Regenerate.
16062         * lra-constraints.c (check_and_process_move): Use
16063         targetm.secondary_memory_needed_mode instead of
16064         TARGET_SECONDARY_MEMORY_NEEDED_MODE.
16065         (curr_insn_transform): Likewise.
16066         * reload.c (get_secondary_mem): Likewise.
16067         * config/alpha/alpha.h (SECONDARY_MEMORY_NEEDED_MODE): Delete.
16068         * config/alpha/alpha.c (alpha_secondary_memory_needed_mode): New
16069         function.
16070         (TARGET_SECONDARY_MEMORY_NEEDED_MODE): Redefine.
16071         * config/i386/i386.h (SECONDARY_MEMORY_NEEDED_MODE): Delete.
16072         * config/i386/i386.c (ix86_secondary_memory_needed_mode): New function.
16073         (TARGET_SECONDARY_MEMORY_NEEDED_MODE): Redefine.
16074         * config/powerpcspe/powerpcspe.h (SECONDARY_MEMORY_NEEDED_MODE):
16075         Delete.
16076         * config/powerpcspe/powerpcspe-protos.h
16077         (rs6000_secondary_memory_needed_mode): Delete.
16078         * config/powerpcspe/powerpcspe.c
16079         (TARGET_SECONDARY_MEMORY_NEEDED_MODE): Redefine.
16080         (rs6000_secondary_memory_needed_mode): Make static.
16081         * config/rs6000/rs6000.h (SECONDARY_MEMORY_NEEDED_MODE): Delete.
16082         * config/rs6000/rs6000-protos.h (rs6000_secondary_memory_needed_mode):
16083         Delete.
16084         * config/rs6000/rs6000.c (TARGET_SECONDARY_MEMORY_NEEDED_MODE):
16085         Redefine.
16086         (rs6000_secondary_memory_needed_mode): Make static.
16087         * config/s390/s390.h (SECONDARY_MEMORY_NEEDED_MODE): Delete.
16088         * config/s390/s390.c (s390_secondary_memory_needed_mode): New function.
16089         (TARGET_SECONDARY_MEMORY_NEEDED_MODE): Redefine.
16090         * config/sparc/sparc.h (SECONDARY_MEMORY_NEEDED_MODE): Delete.
16091         * config/sparc/sparc.c (TARGET_SECONDARY_MEMORY_NEEDED_MODE):
16092         Redefine.
16093         (sparc_secondary_memory_needed_mode): New function.
16094         * system.h (TARGET_SECONDARY_MEMORY_NEEDED_MODE): Poison.
16096 2017-09-13  Jackson Woodruff  <jackson.woodruff@arm.com>
16098         * config/aarch64/constraints.md (Umq): New constraint.
16099         * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>):
16100         Change to use Umq.
16101         (mov<mode>): Update condition.
16103 2017-09-13  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
16105         * gimple-ssa-store-merging.c (sort_by_bitpos): Compare store order
16106         when bitposition is the same.
16108 2017-09-13  Richard Biener  <rguenther@suse.de>
16110         * dwarf2out.c (output_die_symbol): Remove.
16111         (output_die): Do not output a DIEs symbol.
16113 2017-09-13  Richard Biener  <rguenther@suse.de>
16115         PR middle-end/82128
16116         * gimple-fold.c (gimple_fold_call): Update SSA name in-place to
16117         default-def to avoid breaking iterator update with the weird
16118         interaction with cgraph_update_edges_for_call_stmt_node.
16120 2017-09-13  Richard Biener  <rguenther@suse.de>
16122         * tree-cfg.c (verify_gimple_assign_binary): Add verification
16123         for WIDEN_SUM_EXPR, VEC_WIDEN_MULT_{HI,LO,EVEN,ODD}_EXPR,
16124         VEC_PACK_{TRUNC,SAT,FIX_TRUNC}_EXPR.
16125         (verify_gimple_assign_ternary): Add verification for DOT_PROD_EXPR.
16127 2017-09-13  Kugan Vivekanandarajah  <kuganv@linaro.org>
16129         * config/aarch64/aarch64.c (aarch64_override_options_after_change_1):
16130         Disable pc relative literal load irrespective of
16131         TARGET_FIX_ERR_A53_84341 for default.
16133 2017-09-12  Eric Botcazou  <ebotcazou@adacore.com>
16135         * config/sparc/sparc.c (output_return): Output the source location of
16136         the insn in the delay slot, if any.
16137         (output_sibcall): Likewise.
16139 2017-09-12  H.J. Lu  <hongjiu.lu@intel.com>
16141         PR driver/81498
16142         * common.opt (-static-pie): New alias.
16143         (shared): Negate static-pie.
16144         (-no-pie): Update help text.
16145         (-pie): Likewise.
16146         (static-pie): New option.
16147         * config/gnu-user.h (GNU_USER_TARGET_STARTFILE_SPEC): Add
16148         -static-pie support.
16149         (GNU_USER_TARGET_ENDFILE_SPEC): Likewise.
16150         (LINK_EH_SPEC): Likewise.
16151         (LINK_GCC_C_SEQUENCE_SPEC): Likewise.
16152         * config/i386/gnu-user.h (GNU_USER_TARGET_LINK_SPEC): Likewise.
16153         * config/i386/gnu-user64.h (GNU_USER_TARGET_LINK_SPEC): Likewise.
16154         * gcc.c (LINK_COMMAND_SPEC): Likewise.
16155         (init_gcc_specs): Likewise.
16156         (init_spec): Likewise.
16157         (display_help): Update help message for -pie.
16158         * doc/invoke.texi: Update -pie, -no-pie and -static.  Document
16159         -static-pie.
16161 2017-09-12  Wilco Dijkstra  <wdijkstr@arm.com>
16163         * config/aarch64/aarch64.md (movsi_aarch64): Remove all '*'.
16164         (movdi_aarch64): Likewise.
16165         (movti_aarch64): Likewise.
16167 2017-09-12 Simon Wright <simon@pushface.org>
16169         PR target/80204
16170         * config/darwin-driver.c (darwin_find_version_from_kernel): Eliminate
16171         calculation of the minor version, always output as 0.
16173 2017-09-12  Jakub Jelinek  <jakub@redhat.com>
16175         PR target/82112
16176         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): For
16177         ALTIVEC_BUILTIN_VEC_LD if arg1 has array type call default_conversion
16178         on it early, rather than manual conversion late.  For
16179         ALTIVEC_BUILTIN_VEC_ST if arg2 has array type call default_conversion
16180         instead of performing manual conversion.
16182 2017-09-12  Carl Love  <cel@us.ibm.com>
16184         * config/rs6000/altivec.md (vec_widen_umult_even_v4si,
16185         vec_widen_smult_even_v4si): Add define expands for vmuleuw, vmulesw,
16186         vmulouw, vmulosw.
16187         * config/rs6000/rs6000-builtin.def (VMLEUW, VMULESW, VMULOUW,
16188         VMULOSW): Add definitions.
16189         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
16190         ALTIVEC_BUILTIN_VMULESW, ALTIVEC_BUILTIN_VMULEUW,
16191         ALTIVEC_BUILTIN_VMULOSW, ALTIVEC_BUILTIN_VMULOUW entries.
16192         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin,
16193         builtin_function_type): Add ALTIVEC_BUILTIN_* case statements.
16195 2017-09-12  James Greenhalgh  <james.greenhalgh@arm.com>
16197         * config/aarch64/aarch64.md (movdi_aarch64): Set load/store
16198         types correctly.
16199         (movti_aarch64): Likewise.
16200         (movdf_aarch64): Likewise.
16201         (movtf_aarch64): Likewise.
16202         (load_pairdi): Likewise.
16203         (store_pairdi): Likewise.
16204         (load_pairdf): Likewise.
16205         (store_pairdf): Likewise.
16206         (loadwb_pair<GPI:mode>_<P:mode>): Likewise.
16207         (storewb_pair<GPI:mode>_<P:mode>): Likewise.
16208         (ldr_got_small_<mode>): Likewise.
16209         (ldr_got_small_28k_<mode>): Likewise.
16210         (ldr_got_tiny): Likewise.
16211         * config/aarch64/iterators.md (ldst_sz): New.
16212         (ldpstp_sz): Likewise.
16213         * config/aarch64/thunderx.md (thunderx_storepair): Split store_8
16214         to store_16.
16215         (thunderx_load): Split load_8 to load_16.
16216         * config/aarch64/thunderx2t99.md (thunderx2t99_loadpair): Split
16217         load_8 to load_16.
16218         (thunderx2t99_storepair_basic): Split store_8 to store_16.
16219         * config/arm/xgene1.md (xgene1_load_pair): Split load_8 to load_16.
16220         (xgene1_store_pair): Split store_8 to store_16.
16221         * config/aarch64/falkor.md (falkor_ld_3_ld): Split load_8 to load_16.
16222         (falkor_st_0_st_sd): Split store_8 to store_16.
16224 2017-09-12  James Greenhalgh  <james.greenhalgh@arm.com>
16226         * config/arm/types.md (type): Rename load1/2/3/4 to load_4/8/12/16
16227         and store1/2/3/4 to store_4/8/12/16.
16228         * config/aarch64/aarch64.md: Update for rename.
16229         * config/arm/arm.md: Likewise.: Likewise.
16230         * config/arm/arm.c: Likewise.
16231         * config/arm/thumb1.md: Likewise.
16232         * config/arm/thumb2.md: Likewise.
16233         * config/arm/vfp.md: Likewise.
16234         * config/arm/arm-generic.md: Likewise.
16235         * config/arm/arm1020e.md: Likewise.
16236         * config/arm/arm1026ejs.md: Likewise.
16237         * config/arm/arm1136jfs.md: Likewise.
16238         * config/arm/arm926ejs.md: Likewise.
16239         * config/arm/cortex-a15.md: Likewise.
16240         * config/arm/cortex-a17.md: Likewise.
16241         * config/arm/cortex-a5.md: Likewise.
16242         * config/arm/cortex-a53.md: Likewise.
16243         * config/arm/cortex-a57.md: Likewise.
16244         * config/arm/cortex-a7.md: Likewise.
16245         * config/arm/cortex-a8.md: Likewise.
16246         * config/arm/cortex-a9.md: Likewise.
16247         * config/arm/cortex-m4.md: Likewise.
16248         * config/arm/cortex-m7.md: Likewise.
16249         * config/arm/cortex-r4.md: Likewise.
16250         * config/arm/exynos-m1.md: Likewise.
16251         * config/arm/fa526.md: Likewise.
16252         * config/arm/fa606te.md: Likewise.
16253         * config/arm/fa626te.md: Likewise.
16254         * config/arm/fa726te.md: Likewise.
16255         * config/arm/fmp626.md: Likewise.
16256         * config/arm/iwmmxt.md: Likewise.
16257         * config/arm/ldmstm.md: Likewise.
16258         * config/arm/marvell-pj4.md: Likewise.
16259         * config/arm/xgene1.md: Likewise.
16260         * config/aarch64/thunderx.md: Likewise.
16261         * config/aarch64/thunderx2t99.md: Likewise.
16262         * config/aarch64/falkor.md: Likewise.
16264 2017-09-12  Martin Liska  <mliska@suse.cz>
16266         * attribs.c (private_lookup_attribute): New function.
16267         * attribs.h (private_lookup_attribute): Declared here.
16268         (lookup_attribute): Called from this place.
16270 2017-09-12  Richard Biener  <rguenther@suse.de>
16272         PR tree-optimization/82157
16273         * tree-ssa-pre.c (remove_dead_inserted_code): Do not remove
16274         stmts with side-effects.
16276 2017-09-12  Richard Sandiford  <richard.sandiford@linaro.org>
16277             Alan Hayward  <alan.hayward@arm.com>
16278             David Sherwood <david.sherwood@arm.com>
16280         * target.def (hard_regno_nregs): New hook.
16281         (class_max_nregs): Refer to it instead of HARD_REGNO_NREGS.
16282         * targhooks.h (default_hard_regno_nregs): Declare.
16283         * targhooks.c (default_hard_regno_nregs): New function.
16284         * doc/tm.texi.in (HARD_REGNO_NREGS): Replace with...
16285         (TARGET_HARD_REGNO_NREGS): ...this hook.
16286         (HARD_REGNO_NREGS_HAS_PADDING): Update accordingly.
16287         (CLASS_MAX_NREGS): Likewise.
16288         * doc/tm.texi: Regenerate.
16289         * reginfo.c (init_reg_modes_target): Use targetm.hard_regno_nregs
16290         instead of HARD_REGNO_NREGS.
16291         * rtl.h (REG_NREGS): Refer to TARGET_HARD_REGNO_NREGS rather than
16292         HARD_REGNO_NREGS in the comment.
16293         * config/aarch64/aarch64.h (HARD_REGNO_NREGS): Delete.
16294         * config/aarch64/aarch64-protos.h (aarch64_hard_regno_nregs): Delete.
16295         * config/aarch64/aarch64.c (aarch64_hard_regno_nregs): Make static.
16296         Return an unsigned int.
16297         (TARGET_HARD_REGNO_NREGS): Redefine.
16298         * config/alpha/alpha.h (HARD_REGNO_NREGS): Delete.
16299         * config/arc/arc.h (HARD_REGNO_NREGS): Delete.
16300         * config/arc/arc.c (TARGET_HARD_REGNO_NREGS): Redefine.
16301         (arc_hard_regno_nregs): New function.
16302         * config/arm/arm.h (HARD_REGNO_NREGS): Delete.
16303         * config/arm/arm.c (TARGET_HARD_REGNO_NREGS): Redefine.
16304         (arm_hard_regno_nregs): New function.
16305         * config/avr/avr.h (HARD_REGNO_NREGS): Delete.
16306         * config/bfin/bfin.h (HARD_REGNO_NREGS): Delete.
16307         * config/bfin/bfin.c (bfin_hard_regno_nregs): New function.
16308         (TARGET_HARD_REGNO_NREGS): Redefine.
16309         * config/c6x/c6x.h (HARD_REGNO_NREGS): Delete.
16310         * config/cr16/cr16.h (LONG_REG_P): Use targetm.hard_regno_nregs.
16311         (HARD_REGNO_NREGS): Delete.
16312         * config/cr16/cr16.c (TARGET_HARD_REGNO_NREGS): Redefine.
16313         (cr16_hard_regno_nregs): New function.
16314         (cr16_memory_move_cost): Use it instead of HARD_REGNO_NREGS.
16315         * config/cris/cris.h (HARD_REGNO_NREGS): Delete.
16316         * config/cris/cris.c (TARGET_HARD_REGNO_NREGS): Redefine.
16317         (cris_hard_regno_nregs): New function.
16318         * config/epiphany/epiphany.h (HARD_REGNO_NREGS): Delete.
16319         * config/fr30/fr30.h (HARD_REGNO_NREGS): Delete.
16320         (CLASS_MAX_NREGS): Use targetm.hard_regno_nregs.
16321         * config/frv/frv.h (HARD_REGNO_NREGS): Delete.
16322         (CLASS_MAX_NREGS): Remove outdated copy of documentation.
16323         * config/frv/frv-protos.h (frv_hard_regno_nregs): Delete.
16324         * config/frv/frv.c (TARGET_HARD_REGNO_NREGS): Redefine.
16325         (frv_hard_regno_nregs): Make static.  Take and return an
16326         unsigned int.
16327         (frv_class_max_nregs): Remove outdated copy of documentation.
16328         * config/ft32/ft32.h (HARD_REGNO_NREGS): Delete.
16329         * config/h8300/h8300.h (HARD_REGNO_NREGS): Delete.
16330         * config/h8300/h8300-protos.h (h8300_hard_regno_nregs): Delete.
16331         * config/h8300/h8300.c (h8300_hard_regno_nregs): Delete.
16332         * config/i386/i386.h (HARD_REGNO_NREGS): Delete.
16333         * config/i386/i386.c (ix86_hard_regno_nregs): New function.
16334         (TARGET_HARD_REGNO_NREGS): Redefine.
16335         * config/ia64/ia64.h (HARD_REGNO_NREGS): Delete.
16336         (CLASS_MAX_NREGS): Update comment.
16337         * config/ia64/ia64.c (TARGET_HARD_REGNO_NREGS): Redefine.
16338         (ia64_hard_regno_nregs): New function.
16339         * config/iq2000/iq2000.h (HARD_REGNO_NREGS): Delete.
16340         * config/lm32/lm32.h (HARD_REGNO_NREGS): Delete.
16341         * config/m32c/m32c.h (HARD_REGNO_NREGS): Delete.
16342         * config/m32c/m32c-protos.h (m32c_hard_regno_nregs): Delete.
16343         * config/m32c/m32c.c (m32c_hard_regno_nregs_1): Take and return
16344         an unsigned int.
16345         (m32c_hard_regno_nregs): Likewise.  Make static.
16346         (TARGET_HARD_REGNO_NREGS): Redefine.
16347         * config/m32r/m32r.h (HARD_REGNO_NREGS): Delete.
16348         * config/m68k/m68k.h (HARD_REGNO_NREGS): Delete.
16349         * config/m68k/m68k.c (TARGET_HARD_REGNO_NREGS): Redefine.
16350         (m68k_hard_regno_nregs): New function.
16351         * config/mcore/mcore.h (HARD_REGNO_NREGS): Delete.
16352         * config/microblaze/microblaze.h (HARD_REGNO_NREGS): Delete.
16353         * config/mips/mips.h (HARD_REGNO_NREGS): Delete.
16354         * config/mips/mips-protos.h (mips_hard_regno_nregs): Delete.
16355         * config/mips/mips.c (mips_hard_regno_nregs): Make static.
16356         Take and return an unsigned int.
16357         (TARGET_HARD_REGNO_NREGS): Redefine.
16358         * config/mmix/mmix.h (HARD_REGNO_NREGS): Delete.
16359         (CLASS_MAX_NREGS): Use targetm.hard_regno_nregs.
16360         * config/mn10300/mn10300.h (HARD_REGNO_NREGS): Delete.
16361         * config/moxie/moxie.h (HARD_REGNO_NREGS): Delete.
16362         * config/msp430/msp430.h (HARD_REGNO_NREGS): Delete.
16363         * config/msp430/msp430-protos.h (msp430_hard_regno_nregs): Delete.
16364         * config/msp430/msp430.c (TARGET_HARD_REGNO_NREGS): Redefine.
16365         (msp430_hard_regno_nregs): Make static.  Take and return an
16366         unsigned int.
16367         * config/nds32/nds32.h (HARD_REGNO_NREGS): Delete.
16368         * config/nds32/nds32-protos.h (nds32_hard_regno_nregs): Delete.
16369         * config/nds32/nds32.c (nds32_hard_regno_nregs): Delete.
16370         (nds32_hard_regno_mode_ok): Use targetm.hard_regno_nregs.
16371         * config/nios2/nios2.h (HARD_REGNO_NREGS): Delete.
16372         * config/nvptx/nvptx.h (HARD_REGNO_NREGS): Delete.
16373         * config/nvptx/nvptx.c (nvptx_hard_regno_nregs): New function.
16374         (TARGET_HARD_REGNO_NREGS): Redefine.
16375         * config/pa/pa32-regs.h (HARD_REGNO_NREGS): Rename to...
16376         (PA_HARD_REGNO_NREGS): ...this.
16377         * config/pa/pa64-regs.h (HARD_REGNO_NREGS): Rename to...
16378         (PA_HARD_REGNO_NREGS): ...this.
16379         * config/pa/pa.c (TARGET_HARD_REGNO_NREGS): Redefine.
16380         (pa_hard_regno_nregs): New function.
16381         * config/pdp11/pdp11.h (HARD_REGNO_NREGS): Delete.
16382         * config/pdp11/pdp11.c (TARGET_HARD_REGNO_NREGS): Redefine.
16383         (pdp11_hard_regno_nregs): New function.
16384         * config/powerpcspe/powerpcspe.h (HARD_REGNO_NREGS): Delete.
16385         * config/powerpcspe/powerpcspe.c (TARGET_HARD_REGNO_NREGS): Redefine.
16386         (rs6000_hard_regno_nregs_hook): New function.
16387         * config/riscv/riscv.h (HARD_REGNO_NREGS): Delete.
16388         * config/riscv/riscv-protos.h (riscv_hard_regno_nregs): Delete.
16389         * config/riscv/riscv.c (riscv_hard_regno_nregs): Make static.
16390         Take and return an unsigned int.  Move earlier in file.
16391         (TARGET_HARD_REGNO_NREGS): Redefine.
16392         * config/rl78/rl78.h (HARD_REGNO_NREGS): Delete.
16393         * config/rl78/rl78-protos.h (rl78_hard_regno_nregs): Delete.
16394         * config/rl78/rl78.c (TARGET_HARD_REGNO_NREGS): Reefine.
16395         (rl78_hard_regno_nregs): Make static.  Take and return an
16396         unsigned int.
16397         * config/rs6000/rs6000.h (HARD_REGNO_NREGS): Delete.
16398         * config/rs6000/rs6000.c (TARGET_HARD_REGNO_NREGS): Redefine.
16399         (rs6000_hard_regno_nregs_hook): New function.
16400         * config/rx/rx.h (HARD_REGNO_NREGS): Delete.
16401         * config/rx/rx.c (rx_hard_regno_nregs): New function.
16402         (TARGET_HARD_REGNO_NREGS): Redefine.
16403         * config/s390/s390.h (HARD_REGNO_NREGS): Delete.
16404         * config/s390/s390.c (REGNO_PAIR_OK): Use s390_hard_regno_nregs
16405         instead of HARD_REGNO_NREGS.
16406         (s390_hard_regno_nregs): New function.
16407         (s390_hard_regno_mode_ok): Add comment from s390.h.
16408         (TARGET_HARD_REGNO_NREGS): Redefine.
16409         * config/sh/sh.h (HARD_REGNO_NREGS): Delete.
16410         * config/sh/sh.c (TARGET_HARD_REGNO_NREGS): Redefine.
16411         (sh_hard_regno_nregs): New function.
16412         (sh_pass_in_reg_p): Use it.
16413         * config/sparc/sparc.h (HARD_REGNO_NREGS): Delete.
16414         * config/sparc/sparc.c (TARGET_HARD_REGNO_NREGS): Redefine.
16415         (sparc_hard_regno_nregs): New function.
16416         * config/spu/spu.h (HARD_REGNO_NREGS): Delete.
16417         * config/spu/spu.c (spu_hard_regno_nregs): New function.
16418         (spu_function_arg_advance): Use it, supplying a valid register number.
16419         (TARGET_HARD_REGNO_NREGS): Redefine.
16420         * config/stormy16/stormy16.h (HARD_REGNO_NREGS): Delete.
16421         * config/tilegx/tilegx.h (HARD_REGNO_NREGS): Delete.
16422         * config/tilepro/tilepro.h (HARD_REGNO_NREGS): Delete.
16423         * config/v850/v850.h (HARD_REGNO_NREGS): Delete.
16424         * config/vax/vax.h (HARD_REGNO_NREGS): Delete.
16425         * config/visium/visium.h (HARD_REGNO_NREGS): Delete.
16426         (CLASS_MAX_NREGS): Remove copy of old documentation.
16427         * config/visium/visium.c (TARGET_HARD_REGNO_NREGS): Redefine.
16428         (visium_hard_regno_nregs): New function.
16429         (visium_hard_regno_mode_ok): Use it instead of HARD_REGNO_NREGS.
16430         * config/xtensa/xtensa.h (HARD_REGNO_NREGS): Delete.
16431         * config/xtensa/xtensa.c (TARGET_HARD_REGNO_NREGS): Redefine.
16432         xtensa_hard_regno_nregs): New function.
16433         * system.h (HARD_REGNO_NREGS): Poison.
16435 2017-09-12  Richard Sandiford  <richard.sandiford@linaro.org>
16437         * config/arm/arm.h (THUMB_SECONDARY_INPUT_RELOAD_CLASS): Use
16438         hard_regno_nregs instead of HARD_REGNO_NREGS.
16439         (THUMB_SECONDARY_OUTPUT_RELOAD_CLASS): Likewise.
16440         * config/c6x/c6x.c (c6x_expand_prologue): Likewise.
16441         (c6x_expand_epilogue): Likewise.
16442         * config/frv/frv.c (frv_alloc_temp_reg): Likewise.
16443         (frv_read_iacc_argument): Likewise.
16444         * config/sh/sh.c: Include regs.h.
16445         (sh_print_operand): Use hard_regno_nregs instead of HARD_REGNO_NREGS.
16446         (regs_used): Likewise.
16447         (output_stack_adjust): Likewise.
16448         * config/xtensa/xtensa.c (xtensa_copy_incoming_a7): Likewise.
16449         * expmed.c: Include regs.h.
16450         (store_bit_field_1): Use hard_regno_nregs instead of HARD_REGNO_NREGS.
16451         * ree.c: Include regs.h.
16452         (combine_reaching_defs): Use hard_regno_nregs instead of
16453         HARD_REGNO_NREGS.
16454         (add_removable_extension): Likewise.
16456 2017-09-12  Richard Sandiford  <richard.sandiford@linaro.org>
16458         * regs.h (hard_regno_nregs): Turn into a function.
16459         (end_hard_regno): Update accordingly.
16460         * caller-save.c (setup_save_areas): Likewise.
16461         (save_call_clobbered_regs): Likewise.
16462         (replace_reg_with_saved_mem): Likewise.
16463         (insert_restore): Likewise.
16464         (insert_save): Likewise.
16465         * combine.c (can_change_dest_mode): Likewise.
16466         (move_deaths): Likewise.
16467         (distribute_notes): Likewise.
16468         * config/mips/mips.c (mips_hard_regno_call_part_clobbered): Likewise.
16469         * config/powerpcspe/powerpcspe.c (rs6000_cannot_change_mode_class)
16470         (rs6000_split_multireg_move): Likewise.
16471         (rs6000_register_move_cost): Likewise.
16472         (rs6000_memory_move_cost): Likewise.
16473         * config/rs6000/rs6000.c (rs6000_cannot_change_mode_class): Likewise.
16474         (rs6000_split_multireg_move): Likewise.
16475         (rs6000_register_move_cost): Likewise.
16476         (rs6000_memory_move_cost): Likewise.
16477         * cselib.c (cselib_reset_table): Likewise.
16478         (cselib_lookup_1): Likewise.
16479         * emit-rtl.c (set_mode_and_regno): Likewise.
16480         * function.c (aggregate_value_p): Likewise.
16481         * ira-color.c (setup_profitable_hard_regs): Likewise.
16482         (check_hard_reg_p): Likewise.
16483         (calculate_saved_nregs): Likewise.
16484         (assign_hard_reg): Likewise.
16485         (improve_allocation): Likewise.
16486         (calculate_spill_cost): Likewise.
16487         * ira-emit.c (modify_move_list): Likewise.
16488         * ira-int.h (ira_hard_reg_set_intersection_p): Likewise.
16489         (ira_hard_reg_in_set_p): Likewise.
16490         * ira.c (setup_reg_mode_hard_regset): Likewise.
16491         (clarify_prohibited_class_mode_regs): Likewise.
16492         (check_allocation): Likewise.
16493         * lra-assigns.c (find_hard_regno_for_1): Likewise.
16494         (lra_setup_reg_renumber): Likewise.
16495         (setup_try_hard_regno_pseudos): Likewise.
16496         (spill_for): Likewise.
16497         (assign_hard_regno): Likewise.
16498         (setup_live_pseudos_and_spill_after_risky_transforms): Likewise.
16499         * lra-constraints.c (in_class_p): Likewise.
16500         (lra_constraint_offset): Likewise.
16501         (simplify_operand_subreg): Likewise.
16502         (lra_constraints): Likewise.
16503         (split_reg): Likewise.
16504         (split_if_necessary): Likewise.
16505         (invariant_p): Likewise.
16506         (inherit_in_ebb): Likewise.
16507         * lra-lives.c (process_bb_lives): Likewise.
16508         * lra-remat.c (reg_overlap_for_remat_p): Likewise.
16509         (get_hard_regs): Likewise.
16510         (do_remat): Likewise.
16511         * lra-spills.c (assign_spill_hard_regs): Likewise.
16512         * mode-switching.c (create_pre_exit): Likewise.
16513         * postreload.c (reload_combine_recognize_pattern): Likewise.
16514         * recog.c (peep2_find_free_register): Likewise.
16515         * regcprop.c (kill_value_regno): Likewise.
16516         (set_value_regno): Likewise.
16517         (copy_value): Likewise.
16518         (maybe_mode_change): Likewise.
16519         (find_oldest_value_reg): Likewise.
16520         (copyprop_hardreg_forward_1): Likewise.
16521         * regrename.c (check_new_reg_p): Likewise.
16522         (regrename_do_replace): Likewise.
16523         * reload.c (push_reload): Likewise.
16524         (combine_reloads): Likewise.
16525         (find_dummy_reload): Likewise.
16526         (operands_match_p): Likewise.
16527         (find_reloads): Likewise.
16528         (find_equiv_reg): Likewise.
16529         (reload_adjust_reg_for_mode): Likewise.
16530         * reload1.c (count_pseudo): Likewise.
16531         (count_spilled_pseudo): Likewise.
16532         (find_reg): Likewise.
16533         (clear_reload_reg_in_use): Likewise.
16534         (free_for_value_p): Likewise.
16535         (allocate_reload_reg): Likewise.
16536         (choose_reload_regs): Likewise.
16537         (reload_adjust_reg_for_temp): Likewise.
16538         (emit_reload_insns): Likewise.
16539         (delete_output_reload): Likewise.
16540         * rtlanal.c (subreg_get_info): Likewise.
16541         * sched-deps.c (sched_analyze_reg): Likewise.
16542         * sel-sched.c (init_regs_for_mode): Likewise.
16543         (mark_unavailable_hard_regs): Likewise.
16544         (choose_best_reg_1): Likewise.
16545         (verify_target_availability): Likewise.
16546         * valtrack.c (dead_debug_insert_temp): Likewise.
16547         * var-tracking.c (track_loc_p): Likewise.
16548         (emit_note_insn_var_location): Likewise.
16549         * varasm.c (make_decl_rtl): Likewise.
16550         * reginfo.c (choose_hard_reg_mode): Likewise.
16551         (init_reg_modes_target): Refer directly to
16552         this_target_regs->x_hard_regno_nregs.
16554 2017-09-12  Richard Sandiford  <richard.sandiford@linaro.org>
16556         * ira-costs.c (record_operand_costs): Use in_hard_reg_set_p
16557         instead of hard_regno_nregs.
16559 2017-09-12  Richard Sandiford  <richard.sandiford@linaro.org>
16561         * config/aarch64/aarch64.c (aarch64_hard_regno_mode_ok): Use
16562         end_hard_regno instead of hard_regno_nregs.
16563         * config/s390/s390.c (s390_reg_clobbered_rtx): Likewise.
16564         * config/sparc/sparc.h (ASM_DECLARE_REGISTER_GLOBAL): Likewise.
16565         * config/visium/visium.c (visium_hard_regno_mode_ok): Likewise.
16566         * ira-color.c (improve_allocation): Likewise.
16567         * lra-assigns.c (find_hard_regno_for_1): Likewise.
16568         * lra-lives.c (mark_regno_live): Likewise.
16569         (mark_regno_dead): Likewise.
16570         * lra-remat.c (operand_to_remat): Likewise.
16571         * lra.c (collect_non_operand_hard_regs): Likewise.
16572         * postreload.c (reload_combine_note_store): Likewise.
16573         (move2add_valid_value_p): Likewise.
16574         * reload.c (regno_clobbered_p): Likewise.
16576 2017-09-12  Richard Sandiford  <richard.sandiford@linaro.org>
16578         * config/frv/frv.c (FOR_EACH_REGNO): Use END_REGNO instead of
16579         hard_regno_nregs.
16580         * config/v850/v850.c (v850_reorg): Likewise.
16581         * reload.c (refers_to_regno_for_reload_p): Likewise.
16582         (find_equiv_reg): Likewise.
16583         * reload1.c (reload_reg_reaches_end_p): Likewise.
16585 2017-09-12  Richard Sandiford  <richard.sandiford@linaro.org>
16587         * caller-save.c (add_used_regs): Use REG_NREGS instead of
16588         hard_regno_nregs.
16589         * config/aarch64/aarch64.c (aarch64_split_combinev16qi): Likewise.
16590         * config/arm/arm.c (output_move_neon): Likewise.
16591         (arm_attr_length_move_neon): Likewise.
16592         (neon_split_vcombine): Likewise.
16593         * config/c6x/c6x.c (c6x_mark_reg_read): Likewise.
16594         (c6x_mark_reg_written): Likewise.
16595         (c6x_dwarf_register_span): Likewise.
16596         * config/i386/i386.c (ix86_save_reg): Likewise.
16597         * config/ia64/ia64.c (mark_reg_gr_used_mask): Likewise.
16598         (rws_access_reg): Likewise.
16599         * config/s390/s390.c (s390_call_saved_register_used): Likewise.
16600         * mode-switching.c (create_pre_exit): Likewise.
16601         * ree.c (combine_reaching_defs): Likewise.
16602         (add_removable_extension): Likewise.
16603         * regcprop.c (find_oldest_value_reg): Likewise.
16604         (copyprop_hardreg_forward_1): Likewise.
16605         * reload.c (reload_inner_reg_of_subreg): Likewise.
16606         (push_reload): Likewise.
16607         (combine_reloads): Likewise.
16608         (find_dummy_reload): Likewise.
16609         (reload_adjust_reg_for_mode): Likewise.
16610         * reload1.c (find_reload_regs): Likewise.
16611         (forget_old_reloads_1): Likewise.
16612         (reload_reg_free_for_value_p): Likewise.
16613         (reload_adjust_reg_for_temp): Likewise.
16614         (emit_reload_insns): Likewise.
16615         (delete_output_reload): Likewise.
16616         * sel-sched.c (choose_best_reg_1): Likewise.
16617         (choose_best_pseudo_reg): Likewise.
16619 2017-09-12  Richard Sandiford  <richard.sandiford@linaro.org>
16620             Alan Hayward  <alan.hayward@arm.com>
16621             David Sherwood <david.sherwood@arm.com>
16623         * defaults.h (SLOW_UNALIGNED_ACCESS): Delete.
16624         * target.def (slow_unaligned_access): New hook.
16625         * targhooks.h (default_slow_unaligned_access): Declare.
16626         * targhooks.c (default_slow_unaligned_access): New function.
16627         * doc/tm.texi.in (SLOW_UNALIGNED_ACCESS): Replace with...
16628         (TARGET_SLOW_UNALIGNED_ACCESS): ...this.
16629         * doc/tm.texi: Regenerate.
16630         * config/alpha/alpha.h (SLOW_UNALIGNED_ACCESS): Delete.
16631         * config/arm/arm.h (SLOW_UNALIGNED_ACCESS): Delete.
16632         * config/i386/i386.h (SLOW_UNALIGNED_ACCESS): Delete commented-out
16633         definition.
16634         * config/powerpcspe/powerpcspe.h (SLOW_UNALIGNED_ACCESS): Delete.
16635         * config/powerpcspe/powerpcspe.c (TARGET_SLOW_UNALIGNED_ACCESS):
16636         Redefine.
16637         (rs6000_slow_unaligned_access): New function.
16638         (rs6000_emit_move): Use it instead of SLOW_UNALIGNED_ACCESS.
16639         (expand_block_compare): Likewise.
16640         (expand_strn_compare): Likewise.
16641         (rs6000_rtx_costs): Likewise.
16642         * config/riscv/riscv.h (SLOW_UNALIGNED_ACCESS): Delete.
16643         (riscv_slow_unaligned_access): Likewise.
16644         * config/riscv/riscv.c (riscv_slow_unaligned_access): Rename to...
16645         (riscv_slow_unaligned_access_p): ...this and make static.
16646         (riscv_option_override): Update accordingly.
16647         (riscv_slow_unaligned_access): New function.
16648         (TARGET_SLOW_UNALIGNED_ACCESS): Redefine.
16649         * config/rs6000/rs6000.h (SLOW_UNALIGNED_ACCESS): Delete.
16650         * config/rs6000/rs6000.c (TARGET_SLOW_UNALIGNED_ACCESS): Redefine.
16651         (rs6000_slow_unaligned_access): New function.
16652         (rs6000_emit_move): Use it instead of SLOW_UNALIGNED_ACCESS.
16653         (rs6000_rtx_costs): Likewise.
16654         * config/rs6000/rs6000-string.c (expand_block_compare)
16655         (expand_strn_compare): Use targetm.slow_unaligned_access instead
16656         of SLOW_UNALIGNED_ACCESS.
16657         * config/tilegx/tilegx.h (SLOW_UNALIGNED_ACCESS): Delete.
16658         * config/tilepro/tilepro.h (SLOW_UNALIGNED_ACCESS): Delete.
16659         * calls.c (expand_call): Use targetm.slow_unaligned_access instead
16660         of SLOW_UNALIGNED_ACCESS.
16661         * expmed.c (simple_mem_bitfield_p): Likewise.
16662         * expr.c (alignment_for_piecewise_move): Likewise.
16663         (emit_group_load_1): Likewise.
16664         (emit_group_store): Likewise.
16665         (copy_blkmode_from_reg): Likewise.
16666         (emit_push_insn): Likewise.
16667         (expand_assignment): Likewise.
16668         (store_field): Likewise.
16669         (expand_expr_real_1): Likewise.
16670         * gimple-fold.c (gimple_fold_builtin_memory_op): Likewise.
16671         * lra-constraints.c (simplify_operand_subreg): Likewise.
16672         * stor-layout.c (bit_field_mode_iterator::next_mode): Likewise.
16673         * gimple-ssa-store-merging.c: Likewise in block comment at start
16674         of file.
16675         * tree-ssa-strlen.c: Include target.h.
16676         (handle_builtin_memcmp): Use targetm.slow_unaligned_access instead
16677         of SLOW_UNALIGNED_ACCESS.
16678         * system.h (SLOW_UNALIGNED_ACCESS): Poison.
16680 2017-09-12  Richard Sandiford  <richard.sandiford@linaro.org>
16682         PR rtl-optimization/82185
16683         * expmed.c (emit_store_flag_int): Only test tem if it has been
16684         initialized.
16686 2017-09-12  Richard Biener  <rguenther@suse.de>
16688         PR middle-end/82149
16689         * match.pd ((FTYPE) N CMP CST): Fix typo.
16691 2017-09-12  Simon Atanasyan  <simon.atanasyan@imgtec.com>
16693         * config/mips/mips.c (mips_attribute_table): Add 'short_call'
16694         attribute.
16695         (mips_near_type_p): Add 'short_call' attribute as a synonym
16696         for 'near'.
16697         * doc/extend.texi (short_call): Document new function attribute.
16699 2017-09-12  Jakub Jelinek  <jakub@redhat.com>
16701         PR target/82112
16702         * c-common.c (sync_resolve_size): Instead of c_dialect_cxx ()
16703         assertion check that in the condition.
16704         (get_atomic_generic_size): Likewise.  Before testing if parameter
16705         has pointer type, if it has array type, call for C++
16706         default_conversion to perform array-to-pointer conversion.
16708 2017-09-12  Richard Biener  <rguenther@suse.de>
16710         * tree-vect-generic.c (expand_vector_operations_1): Do nothing
16711         for operations we cannot scalarize.
16713 2017-09-12  Aldy Hernandez  <aldyh@redhat.com>
16715         * tree-ssa-threadbackward.c (fsm_find_thread_path): Make GC
16716         vectors heap vectors.  Clean up comments.
16717         Make visited_bbs a reference.
16718         (profitable_jump_thread_path): Make GC
16719         vectors heap vectors.  Clean up comments.
16720         Misc cleanups.
16721         (convert_and_register_jump_thread_path): Make GC vectors heap
16722         vectors.
16723         (check_subpath_and_update_thread_path): Same.  Clean up comments.
16724         Make visited_bbs a reference.
16725         (handle_phi): Abstract common code to to
16726         register_jump_thread_path_if_profitable.
16727         Rename VAR_BB to DEF_BB.
16728         Update comments.
16729         Make GC vectors heap vectors.
16730         Make visited_bbs a reference.
16731         (handle_assignment): Same.
16732         (register_jump_thread_path_if_profitable): New.
16733         (fsm_find_control_statement_thread_paths): Rename VAR_BB to
16734         DEF_BB.
16735         Make GC vectors heap vectors.  Clean up comments.
16736         Make visited_bbs a reference.
16737         (find_jump_threads_backwards): Make visited_bbs live in the stack.
16738         * tree-ssa-threadupdate.c (delete_jump_thread_path): Fix typo in
16739         comment.
16741 2017-09-11  Max Filippov  <jcmvbkbc@gmail.com>
16743         PR target/82181
16744         * config/xtensa/xtensa.c (xtensa_mem_offset): Check that both
16745         words of E_DImode object are reachable by xtensa_uimm8x4 access.
16747 2017-09-11  Vidya Praveen  <vidyapraveen@arm.com>
16749         Revert r251800 and r251799.
16751 2017-09-11  Martin Jambor  <mjambor@suse.cz>
16753         PR hsa/82119
16754         * hsa-gen.c (gen_hsa_phi_from_gimple_phi): Process ADDR_EXPRs in
16755         arguments in advance.
16756         * hsa-regalloc.c (naive_process_phi): New parameter predecessors,
16757         use it to find predecessor edges.
16758         (naive_outof_ssa): Collect vector of predecessors.
16760 2017-09-08  Jason Merrill  <jason@redhat.com>
16762         PR c++/70029 - ICE with ref-qualifier and -flto
16763         * langhooks.h (struct lang_hooks_for_types): Add
16764         copy_lang_qualifiers.
16765         * attribs.c (build_type_attribute_qual_variant): Use it.
16766         * langhooks-def.h (LANG_HOOKS_COPY_LANG_QUALIFIERS): Default to
16767         NULL.
16768         (LANG_HOOKS_FOR_TYPES_INITIALIZER): Use it.
16769         * tree.c (verify_type): Re-enable TYPE_CANONICAL main variant check.
16771 2017-09-08  Eric Botcazou  <ebotcazou@adacore.com>
16773         PR target/81988
16774         * config/sparc/sparc.md (mulsi3): Rename into *mulsi3_sp32.
16775         (*mulsi3_sp64): New instruction.
16776         (mulsi3): New expander.
16778 2017-09-08  Uros Bizjak  <ubizjak@gmail.com>
16780         * config/alpha/alpha.c (alpha_print_operand) <case 'S'>: Remove.
16782 2017-09-08  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
16784         * sancov.c: Include memmodel.h.
16786 2017-09-07  Eric Botcazou  <ebotcazou@adacore.com>
16788         PR target/80897
16789         * config/sparc/sparc.c (sparc_emit_set_symbolic_const64): Deal with too
16790         large offsets.
16792 2017-09-07  Carl Love  <cel@us.ibm.com>
16794         * config/rs6000/vsx.md (define_insn "*stxvl"): Add missing argument to
16795         the sldi instruction.
16797 2017-09-07  David Edelsohn  <dje.gcc@gmail.com>
16799         * sancov.c: Include tm_p.h.
16801 2017-09-07  Jakub Jelinek  <jakub@redhat.com>
16803         PR target/81979
16804         * output.h (switch_to_other_text_partition): New declaration.
16805         * varasm.c (switch_to_other_text_partition): New function.
16806         * config/rs6000/rs6000.c (uses_TOC): Return 2 if
16807         NOTE_INSN_SWITCH_TEXT_SECTIONS is seen before finding load_toc_* insn.
16808         (rs6000_elf_declare_function_name): If uses_TOC returned 2, switch
16809         to the other text partition before emitting LCL label and switch back
16810         after emitting the word after it.
16812 2017-09-07  Richard Biener  <rguenther@suse.de>
16814         * passes.def (pass_split_crit_edges): Remove instance before PRE.
16815         * tree-ssa-pre.c (pass_pre::execute): Instead manually split
16816         critical edges here, after loop init.
16817         (pass_data_pre): Remove PROP_no_crit_edges flags.
16818         * tree-ssa-sccvn.c (vn_reference_lookup_3): Use vn_valueize
16819         for valueization of call args to avoid leaking VN_TOP.
16820         (visit_use): Assert we do not visit default defs.
16821         (init_scc_vn): Use build_decl for VN_TOP to make name nicer.
16822         Use error_mark_node to more easily detect leaking VN_TOP.
16823         All default-defs are varying, not VN_TOP.  Mark them visited.
16824         (run_scc_vn): Make code match comment.
16826 2017-09-07  Michael Meissner  <meissner@linux.vnet.ibm.com>
16828         * config/rs6000/rs6000-cpus.def (OTHER_VSX_VECTOR_MASKS): Delete
16829         OPTION_MASK_FLOAT128_KEYWORD.
16830         (POWERPC_MASKS): Likewise.
16831         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Delete
16832         support for the -mfloat128-type option, and make -mfloat128
16833         default on PowerPC Linux systems.  Define or undefine
16834         __FLOAT128__ and  __FLOAT128_HARDWARE__ for the current options.
16835         Define __float128 to be __ieee128 if IEEE 128-bit support is
16836         enabled, or undefine it.
16837         (rs6000_cpu_cpp_builtins): Delete defining __FLOAT128__ here.
16838         Delete defining __FLOAT128_TYPE__.
16839         * config/rs6000/rs6000.opt (x_TARGET_FLOAT128_TYPE): Delete the
16840         -mfloat128-type option and make -mfloat128 default on PowerPC
16841         Linux systems.
16842         (TARGET_FLOAT128_TYPE): Likewise.
16843         (-mfloat128-type): Likewise.
16844         * config/rs6000/rs6000.c (rs6000_option_override_internal):
16845         Delete the -mfloat128-type option and make -mfloat128 default on
16846         PowerPC Linux systems.  Always use __ieee128 to be the keyword for
16847         the IEEE 128-bit type, and map __float128 to __ieee128 if IEEE
16848         128-bit floating point is enabled.  Change tests from using
16849         -mfloat128-type to -mfloat128.
16850         (rs6000_mangle_type): Use the correct mangling for the __float128
16851         type even if normal long double is restricted to 64-bits.
16852         (floatn_mode): Enable the _Float128 type by default on VSX Linux
16853         systems.
16854         * config/rs6000/rs6000.h (MASK_FLOAT128_TYPE): Delete.
16855         (MASK_FLOAT128_KEYWORD): Define new shortcut macro.
16856         (RS6000BTM_FLOAT128): Define in terms of -mfloat128, not
16857         -mfloat128-type.
16858         * doc/invoke.texi (RS/6000 and PowerPC Options): Update
16859         documentation for -mfloat128.
16861 2017-09-06  Olivier Hainque  <hainque@adacore.com>
16863         * config.gcc (powerpc-wrs-vxworksspe): Now match as vxworks*spe.
16865 2017-09-06  Wish Wu  <wishwu007@gmail.com>
16866             Jakub Jelinek  <jakub@redhat.com>
16868         * asan.c (initialize_sanitizer_builtins): Add
16869         BT_FN_VOID_UINT8_UINT8, BT_FN_VOID_UINT16_UINT16,
16870         BT_FN_VOID_UINT32_UINT32, BT_FN_VOID_UINT64_UINT64,
16871         BT_FN_VOID_FLOAT_FLOAT, BT_FN_VOID_DOUBLE_DOUBLE and
16872         BT_FN_VOID_UINT64_PTR variables.
16873         * builtin-types.def (BT_FN_VOID_UINT8_UINT8): New fn type.
16874         (BT_FN_VOID_UINT16_UINT16): Likewise.
16875         (BT_FN_VOID_UINT32_UINT32): Likewise.
16876         (BT_FN_VOID_FLOAT_FLOAT): Likewise.
16877         (BT_FN_VOID_DOUBLE_DOUBLE): Likewise.
16878         (BT_FN_VOID_UINT64_PTR): Likewise.
16879         * common.opt (flag_sanitize_coverage): New variable.
16880         (fsanitize-coverage=trace-pc): Remove.
16881         (fsanitize-coverage=): Add.
16882         * flag-types.h (enum sanitize_coverage_code): New enum.
16883         * fold-const.c (fold_range_test): Disable non-short-circuit
16884         optimization if flag_sanitize_coverage.
16885         (fold_truth_andor): Likewise.
16886         * tree-ssa-ifcombine.c (ifcombine_ifandif): Likewise.
16887         * opts.c (COVERAGE_SANITIZER_OPT): Define.
16888         (coverage_sanitizer_opts): New array.
16889         (get_closest_sanitizer_option): Add OPTS argument, handle also
16890         OPT_fsanitize_coverage_.
16891         (parse_sanitizer_options): Adjusted to also handle
16892         OPT_fsanitize_coverage_.
16893         (common_handle_option): Add OPT_fsanitize_coverage_.
16894         * sancov.c (instrument_comparison, instrument_switch): New function.
16895         (sancov_pass): Add trace-cmp support.
16896         * sanitizer.def (BUILT_IN_SANITIZER_COV_TRACE_CMP1,
16897         BUILT_IN_SANITIZER_COV_TRACE_CMP2, BUILT_IN_SANITIZER_COV_TRACE_CMP4,
16898         BUILT_IN_SANITIZER_COV_TRACE_CMP8,
16899         BUILT_IN_SANITIZER_COV_TRACE_CONST_CMP1,
16900         BUILT_IN_SANITIZER_COV_TRACE_CONST_CMP2,
16901         BUILT_IN_SANITIZER_COV_TRACE_CONST_CMP4,
16902         BUILT_IN_SANITIZER_COV_TRACE_CONST_CMP8,
16903         BUILT_IN_SANITIZER_COV_TRACE_CMPF, BUILT_IN_SANITIZER_COV_TRACE_CMPD,
16904         BUILT_IN_SANITIZER_COV_TRACE_SWITCH): New builtins.
16905         * doc/invoke.texi: Document -fsanitize-coverage=trace-cmp.
16907 2017-09-06  Richard Earnshaw  <rearnsha@arm.com>
16909         * config/arm/parsecpu.awk (fatal): Note that we've encountered an
16910         error.  Only quit immediately if parsing is complete.
16911         (BEGIN): Initialize fatal_err and parse_done.
16912         (begin fpu, end fpu): Check number of arguments.
16913         (begin arch, end arch): Likewise.
16914         (begin cpu, end cpu): Likewise.
16915         (cname, tune for, tune flags, architecture, fpu, option): Likewise.
16916         (optalias): Likewise.
16918 2017-09-06  Richard Earnshaw  <rearnsha@arm.com>
16920         * config.gcc (arm*-*-*): Don't add arm-isa.h to tm_p_file.
16921         * config/arm/arm-isa.h: Delete.  Move definitions to ...
16922         * arm-cpus.in: ... here.  Use new feature and fgroup values.
16923         * config/arm/arm.c (arm_option_override): Use lower case for feature
16924         bit names.
16925         * config/arm/arm.h (TARGET_HARD_FLOAT): Likewise.
16926         (TARGET_VFP3, TARGET_VFP5, TARGET_FMA): Likewise.
16927         * config/arm/parsecpu.awk (END): Add new command 'isa'.
16928         (isa_pfx): Delete.
16929         (print_isa_bits_for): New function.
16930         (gen_isa): New function.
16931         (gen_comm_data): Use print_isa_bits_for.
16932         (define feature): New keyword.
16933         (define fgroup): New keyword.
16934         * config/arm/t-arm (OPTIONS_H_EXTRA): Add arm-isa.h
16935         (arm-isa.h): Add rule to generate file.
16936         * common/config/arm/arm-common.c: (arm_canon_arch_option): Use lower
16937         case for feature bit names.
16939 2017-09-06  Richard Biener  <rguenther@suse.de>
16941         * tree-ssa-pre.c (NECESSARY): Remove.
16942         (create_expression_by_pieces): Do not touch pass-local flags.
16943         (insert_into_preds_of_block): Likewise.
16944         (do_pre_regular_insertion): Likewise.
16945         (eliminate_insert): Likewise.
16946         (eliminate_dom_walker::before_dom_children): Likewise.
16947         (fini_eliminate): Do not look at inserted_exprs.
16948         (mark_operand_necessary): Remove.
16949         (remove_dead_inserted_code): Replace with simple work-list
16950         algorithm based on inserted_exprs and SSA uses.
16951         (pass_pre::execute): Re-order fini_eliminate and
16952         remove_dead_inserted_code.
16954 2017-09-06  Olivier Hainque  <hainque@adacore.com>
16956         * config/powerpcspe/vxworks.h (VXCPU_FOR_8548): Correct definition
16957         for VxWorks 7.  Adjust surrounding comments.
16959 2017-09-06  Richard Biener  <rguenther@suse.de>
16961         * gimple-ssa-strength-reduction.c
16962         (find_candidates_dom_walker::before_dom_children): Also allow
16963         pointer types.
16965 2017-09-06  Richard Biener  <rguenther@suse.de>
16967         PR tree-optimization/82108
16968         * tree-vect-stmts.c (vectorizable_load): Fix pointer adjustment
16969         for gap in the non-permutation SLP case.
16971 2017-09-06  Martin Jambor  <mjambor@suse.cz>
16973         PR tree-optimization/82078
16974         * tree-sra.c (sort_and_splice_var_accesses): Move call to
16975         add_access_to_work_queue...
16976         (build_accesses_from_assign): ...here.
16977         (propagate_all_subaccesses): Make sure racc is the group
16978         representative, if there is one.
16980 2017-09-06  Jakub Jelinek  <jakub@redhat.com>
16982         PR middle-end/82095
16983         * varasm.c (categorize_decl_for_section): Use SECCAT_TBSS for TLS vars with
16984         NULL DECL_INITIAL.
16986 2017-09-06  Richard Biener  <rguenther@suse.de>
16988         * gimple-ssa-strength-reduction.c
16989         (find_candidates_dom_walker::before_doom_children): Use a
16990         type and not a mode check.
16992 2017-09-06  Bernd Edlinger  <bernd.edlinger@hotmail.de>
16994         PR target/77308
16995         * config/arm/predicates.md (arm_general_adddi_operand): Create new
16996         non-vfp predicate.
16997         * config/arm/arm.md (*arm_adddi3, *arm_subdi3): Use new predicates.
16999 2017-09-05  Jeff Law  <law@redhat.com>
17001         PR tree-optimization/64910
17002         * tree-ssa-reassoc.c (reassociate_bb): Restrict last change to
17003         cases where we have 3 or more operands.
17005 2017-09-05  Jakub Jelinek  <jakub@redhat.com>
17007         PR middle-end/81768
17008         * omp-low.c (lower_omp_for): Recompute tree invariant if
17009         gimple_omp_for_initial/final is ADDR_EXPR.
17011         PR middle-end/81768
17012         * omp-expand.c (expand_omp_simd): Force second operands of COND_EXPR
17013         into gimple val before gimplification fo the COND_EXPR.
17015 2017-09-05  Aldy Hernandez  <aldyh@redhat.com>
17017         * tree-ssa-threadupdate.c (duplicate_thread_path): Remove unused
17018         REGION_COPY argument.
17019         (thread_through_all_blocks): Remove unused argument to
17020         duplicate_thread_path.
17022 2017-09-05  Richard Sandiford  <richard.sandiford@linaro.org>
17023             Alan Hayward  <alan.hayward@arm.com>
17024             David Sherwood  <david.sherwood@arm.com>
17026         * config/aarch64/aarch64-protos.h (aarch64_gen_adjusted_ldpstp):
17027         Take a scalar_mode rather than a machine_mode.
17028         (aarch64_operands_adjust_ok_for_ldpstp): Likewise.
17029         * config/aarch64/aarch64.c (aarch64_simd_container_mode): Likewise.
17030         (aarch64_operands_adjust_ok_for_ldpstp): Likewise.
17031         (aarch64_gen_adjusted_ldpstp): Likewise.
17032         (aarch64_expand_vector_init): Use scalar_mode instead of machine_mode.
17034 2017-09-05  Richard Sandiford  <richard.sandiford@linaro.org>
17035             Alan Hayward  <alan.hayward@arm.com>
17036             David Sherwood  <david.sherwood@arm.com>
17038         * config/aarch64/aarch64-protos.h (aarch64_is_extend_from_extract):
17039         Take a scalar_int_mode instead of a machine_mode.
17040         (aarch64_mask_and_shift_for_ubfiz_p): Likewise.
17041         (aarch64_output_scalar_simd_mov_immediate): Likewise.
17042         (aarch64_simd_scalar_immediate_valid_for_move): Likewise.
17043         (aarch64_simd_attr_length_rglist): Delete.
17044         * config/aarch64/aarch64.c (aarch64_is_extend_from_extract): Take
17045         a scalar_int_mode instead of a machine_mode.
17046         (aarch64_add_offset): Likewise.
17047         (aarch64_internal_mov_immediate): Likewise
17048         (aarch64_add_constant_internal): Likewise.
17049         (aarch64_add_constant): Likewise.
17050         (aarch64_movw_imm): Likewise.
17051         (aarch64_rtx_arith_op_extract_p): Likewise.
17052         (aarch64_mask_and_shift_for_ubfiz_p): Likewise.
17053         (aarch64_simd_scalar_immediate_valid_for_move): Likewise.
17054         Remove assert that the mode isn't a vector.
17055         (aarch64_output_scalar_simd_mov_immediate): Likewise.
17056         (aarch64_expand_mov_immediate): Update calls after above changes.
17057         (aarch64_output_casesi): Use as_a <scalar_int_mode>.
17058         (aarch64_and_bitmask_imm): Check for scalar integer modes.
17059         (aarch64_move_imm): Likewise.
17060         (aarch64_can_const_movi_rtx_p): Likewise.
17061         (aarch64_strip_extend): Likewise.
17062         (aarch64_extr_rtx_p): Likewise.
17063         (aarch64_rtx_costs): Likewise, using wode_mode as the mode of
17064         a CONST_INT when the mode parameter is VOIDmode.
17065         (aarch64_float_const_rtx_p): Use scalar_int_mode for a temporary.
17067 2017-09-05  Richard Sandiford  <richard.sandiford@linaro.org>
17069         * machmode.h (bitwise_mode_for_mode): Return opt_mode.
17070         * stor-layout.c (bitwise_mode_for_mode): Likewise.
17071         (bitwise_type_for_mode): Update accordingly.
17073 2017-09-05  Richard Sandiford  <richard.sandiford@linaro.org>
17075         * stor-layout.h (mode_for_size_tree): Return an opt_mode.
17076         * stor-layout.c (mode_for_size_tree): Likewise.
17077         (mode_for_array): Update accordingly.
17078         (layout_decl): Likewise.
17079         (compute_record_mode): Likewise.  Only set the mode once.
17081 2017-09-05  Richard Sandiford  <richard.sandiford@linaro.org>
17083         * target.def (get_mask_mode): Change return type to opt_mode.
17084         Expand commentary.
17085         * doc/tm.texi: Regenerate.
17086         * targhooks.h (default_get_mask_mode): Return an opt_mode.
17087         * targhooks.c (default_get_mask_mode): Likewise.
17088         * config/i386/i386.c (ix86_get_mask_mode): Likewise.
17089         * optabs-query.c (can_vec_mask_load_store_p): Update use of
17090         targetm.get_mask_mode.
17091         * tree.c (build_truth_vector_type): Likewise.
17093 2017-09-05  Richard Sandiford  <richard.sandiford@linaro.org>
17095         * machmode.h (mode_for_vector): Return an opt_mode.
17096         * stor-layout.c (mode_for_vector): Likewise.
17097         (mode_for_int_vector): Update accordingly.
17098         (layout_type): Likewise.
17099         * config/i386/i386.c (emit_memmov): Likewise.
17100         (ix86_expand_set_or_movmem): Likewise.
17101         (ix86_expand_vector_init): Likewise.
17102         (ix86_get_mask_mode): Likewise.
17103         * config/powerpcspe/powerpcspe.c (rs6000_expand_vec_perm_const_1):
17104         Likewise.
17105         * config/rs6000/rs6000.c (rs6000_expand_vec_perm_const_1): Likewise.
17106         * expmed.c (extract_bit_field_1): Likewise.
17107         * expr.c (expand_expr_real_2): Likewise.
17108         * optabs-query.c (can_vec_perm_p): Likewise.
17109         (can_vec_mask_load_store_p): Likewise.
17110         * optabs.c (expand_vec_perm): Likewise.
17111         * targhooks.c (default_get_mask_mode): Likewise.
17112         * tree-vect-stmts.c (vectorizable_store): Likewise.
17113         (vectorizable_load): Likewise.
17114         (get_vectype_for_scalar_type_and_size): Likewise.
17116 2017-09-05  Richard Sandiford  <richard.sandiford@linaro.org>
17118         * machmode.h (mode_for_int_vector): New function.
17119         * stor-layout.c (mode_for_int_vector): Likewise.
17120         * config/aarch64/aarch64.c (aarch64_emit_approx_sqrt): Use it.
17121         * config/powerpcspe/powerpcspe.c (rs6000_do_expand_vec_perm): Likewise.
17122         * config/rs6000/rs6000.c (rs6000_do_expand_vec_perm): Likewise.
17123         * config/s390/s390.c (s390_expand_vec_compare_cc): Likewise.
17124         (s390_expand_vcond): Likewise.
17126 2017-09-05  Richard Sandiford  <richard.sandiford@linaro.org>
17128         * machmode.h (opt_machine_mode): New type.
17129         (opt_mode<T>): Allow construction from anything that can be
17130         converted to a T.
17131         (is_a, as_a, dyn_cast): Add overloads for opt_mode.
17132         (mode_for_size): Return an opt_machine_mode.
17133         * stor-layout.c (mode_for_size): Likewise.
17134         (mode_for_size_tree): Update call accordingly.
17135         (bitwise_mode_for_mode): Likewise.
17136         (make_fract_type): Likewise.
17137         (make_accum_type): Likewise.
17138         * caller-save.c (replace_reg_with_saved_mem): Update call
17139         accordingly.
17140         * config/alpha/alpha.h (SECONDARY_MEMORY_NEEDED_MODE): Likewise.
17141         * config/i386/i386.h (SECONDARY_MEMORY_NEEDED_MODE): Likewise.
17142         * config/s390/s390.h (SECONDARY_MEMORY_NEEDED_MODE): Likewise.
17143         * config/sparc/sparc.h (SECONDARY_MEMORY_NEEDED_MODE): Likewise.
17144         * expmed.c (extract_bit_field_1): Likewise.
17145         * reload.c (get_secondary_mem): Likewise.
17146         * varasm.c (assemble_integer): Likewise.
17147         * lower-subreg.c (simplify_subreg_concatn): Likewise.  Move
17148         early-out.
17150 2017-09-05  Richard Sandiford  <richard.sandiford@linaro.org>
17152         * machmode.h (decimal_float_mode_for_size): New function.
17153         * real.h (REAL_VALUE_TO_TARGET_LONG_DOUBLE): Use float_mode_for_size.
17154         (REAL_VALUE_TO_TARGET_DOUBLE): Likewise.
17155         (REAL_VALUE_TO_TARGET_SINGLE): Likewise.
17156         (REAL_VALUE_TO_TARGET_DECIMAL128): Use decimal_float_mode_for_size.
17157         (REAL_VALUE_TO_TARGET_DECIMAL64): Likewise.
17158         (REAL_VALUE_TO_TARGET_DECIMAL32): Likewise.
17160 2017-09-05  Richard Sandiford  <richard.sandiford@linaro.org>
17162         * builtins.c (expand_builtin_powi): Use int_mode_for_size.
17163         (get_builtin_sync_mode): Likewise.
17164         (expand_ifn_atomic_compare_exchange): Likewise.
17165         (expand_builtin_atomic_clear): Likewise.
17166         (expand_builtin_atomic_test_and_set): Likewise.
17167         (fold_builtin_atomic_always_lock_free): Likewise.
17168         * calls.c (compute_argument_addresses): Likewise.
17169         (emit_library_call_value_1): Likewise.
17170         (store_one_arg): Likewise.
17171         * combine.c (combine_instructions): Likewise.
17172         * config/aarch64/aarch64.c (aarch64_function_value): Likewise.
17173         * config/arm/arm.c (arm_function_value): Likewise.
17174         (aapcs_allocate_return_reg): Likewise.
17175         * config/c6x/c6x.c (c6x_expand_movmem): Likewise.
17176         * config/i386/i386.c (construct_container): Likewise.
17177         (ix86_gimplify_va_arg): Likewise.
17178         (ix86_expand_sse_cmp): Likewise.
17179         (emit_memmov): Likewise.
17180         (emit_memset): Likewise.
17181         (expand_small_movmem_or_setmem): Likewise.
17182         (ix86_expand_pextr): Likewise.
17183         (ix86_expand_pinsr): Likewise.
17184         * config/lm32/lm32.c (lm32_block_move_inline): Likewise.
17185         * config/microblaze/microblaze.c (microblaze_block_move_straight):
17186         Likewise.
17187         * config/mips/mips.c (mips_function_value_1) Likewise.
17188         (mips_block_move_straight): Likewise.
17189         (mips_expand_ins_as_unaligned_store): Likewise.
17190         * config/powerpcspe/powerpcspe.c
17191         (rs6000_darwin64_record_arg_advance_flush): Likewise.
17192         (rs6000_darwin64_record_arg_flush): Likewise.
17193         * config/rs6000/rs6000.c
17194         (rs6000_darwin64_record_arg_advance_flush): Likewise.
17195         (rs6000_darwin64_record_arg_flush): Likewise.
17196         * config/sparc/sparc.c (sparc_function_arg_1): Likewise.
17197         (sparc_function_value_1): Likewise.
17198         * config/spu/spu.c (adjust_operand): Likewise.
17199         (spu_emit_branch_or_set): Likewise.
17200         (arith_immediate_p): Likewise.
17201         * emit-rtl.c (gen_lowpart_common): Likewise.
17202         * expr.c (expand_expr_real_1): Likewise.
17203         * function.c (assign_parm_setup_block): Likewise.
17204         * gimple-ssa-store-merging.c (encode_tree_to_bitpos): Likewise.
17205         * reload1.c (alter_reg): Likewise.
17206         * stor-layout.c (mode_for_vector): Likewise.
17207         (layout_type): Likewise.
17209 2017-09-05  Richard Sandiford  <richard.sandiford@linaro.org>
17211         * config/spu/spu.c (exp2_immediate_p): Use int_mode_for_mode.
17212         (spu_convert_move): Likewise.
17213         * lower-subreg.c (resolve_simple_move): Likewise.
17215 2017-09-05  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
17217         PR target/81833
17218         * config/rs6000/altivec.md (altivec_vsum2sws): Convert from a
17219         define_insn to a define_expand.
17220         (altivec_vsum2sws_direct): New define_insn.
17221         (altivec_vsumsws): Convert from a define_insn to a define_expand.
17223 2017-09-05  Wilco Dijkstra  <wdijkstr@arm.com>
17225         * config/arm/arm.c (arm_option_params_internal): Improve setting of
17226         max_insns_skipped.
17228 2017-09-05  H.J. Lu  <hongjiu.lu@intel.com>
17230         PR target/59501
17231         PR target/81624
17232         PR target/81769
17233         * config/i386/i386.c (ix86_finalize_stack_frame_flags): Don't
17234         realign stack if stack alignment needed is less than incoming
17235         stack boundary.
17237 2017-09-05  Marek Polacek  <polacek@redhat.com>
17239         PR sanitizer/82072
17240         * convert.c (convert_to_integer_1) <case NEGATE_EXPR>: Move the ubsan
17241         check earlier.
17243 2017-09-05  Wilco Dijkstra  <wdijkstr@arm.com>
17245         * explow.c (get_dynamic_stack_size): Improve dynamic alignment.
17247 2017-09-05  Richard Biener  <rguenther@suse.de>
17249         PR tree-optimization/82084
17250         * fold-const.c (can_native_encode_string_p): Handle wide characters.
17252 2017-09-05  Richard Biener  <rguenther@suse.de>
17254         PR tree-optimization/82102
17255         * tree-ssa-pre.c (fini_eliminate): Check if lhs is NULL.
17257 2017-09-05  Martin Liska  <mliska@suse.cz>
17259         PR tree-optimization/82032
17260         * tree-cfg.c (generate_range_test): New function.
17261         * tree-cfg.h (generate_range_test): Declared here.
17262         * tree-cfgcleanup.c (convert_single_case_switch): New function.
17263         (cleanup_control_expr_graph): Use it.
17264         * tree-switch-conversion.c (try_switch_expansion): Remove
17265         assert.
17266         (emit_case_nodes): Use generate_range_test.
17268 2017-09-04  Uros Bizjak  <ubizjak@gmail.com>
17270         PR target/82098
17271         * config/i386/i386.md (*<btsc><mode>_mask): Add
17272         TARGET_USE_BT to insn constraint.
17273         (*btr<mode>_mask): Ditto.
17275 2017-09-04  Wilco Dijkstra  <wdijkstr@arm.com>
17277         * config/arm/arm.c (arm_legitimate_index_p): Add comment.
17278         (thumb2_legitimate_index_p): Use correct range for DI/DF mode.
17280 2017-09-04  Bernd Edlinger  <bernd.edlinger@hotmail.de>
17282         PR target/77308
17283         * config/arm/arm.md (*arm_adddi3, *arm_subdi3): Split early except for
17284         TARGET_NEON and TARGET_IWMMXT.
17285         (anddi3, iordi3, xordi3, one_cmpldi2): Split while expanding except for
17286         TARGET_NEON and TARGET_IWMMXT.
17287         (*one_cmpldi2_insn): Moved the body of one_cmpldi2 here.
17289 2017-09-04  Uros Bizjak  <ubizjak@gmail.com>
17291         * config/i386/i386-protos.h (ix86_tls_address_pattern_p) New prototype.
17292         (ix86_rewrite_tls_address): Ditto.
17293         * config/i386/i386.c (ix86_tls_address_pattern_p) New function.
17294         (ix86_rewrite_tls_address_1): Ditto.
17295         (ix86_rewrite_tls_address): Ditto.
17296         * config/i386/predicates.md (tls_address_pattern): New predicate.
17297         * config/i386/i386.md (TLS address splitter): New splitter.
17299 2017-09-04  Richard Biener  <rguenther@suse.de>
17301         PR tree-optimization/82084
17302         * fold-const.h (can_native_encode_string_p): Declare.
17303         * fold-const.c (can_native_encode_string_p): Factor out from ...
17304         (native_encode_string): ... here.
17305         * tree-vect-stmts.c (vectorizable_store): Call it to avoid
17306         vectorizing stores from constants we later cannot handle.
17308 2017-09-04  Marek Polacek  <polacek@redhat.com>
17310         PR c/81783
17311         * doc/invoke.texi: Update -Wtautological-compare documentation.
17313 2017-09-04  Jeff Law  <law@redhat.com>
17315         PR tree-optimization/64910
17316         * tree-ssa-reassoc.c (reassociate_bb): For bitwise binary ops,
17317         swap the first and last operand if the last is a constant.
17319 2017-09-04  Marek Polacek  <polacek@redhat.com>
17321         PR sanitizer/82072
17322         * convert.c (do_narrow): When sanitizing signed integer overflows,
17323         bail out for signed types.
17324         (convert_to_integer_1) <case NEGATE_EXPR>: Likewise.
17326 2017-09-04  Richard Biener  <rguenther@suse.de>
17328         PR tree-optimization/82060
17329         * tree-ssa-pre.c (eliminate_dom_walker::before_dom_children):
17330         Move devirtualization after stmt folding and before EH/AB/noreturn
17331         cleanup to get the stmt refs canonicalized.  Use a bool instead
17332         of gimple_modified_p since that doesn't work for NOPs.  Schedule
17333         NOPs generated by folding for removal.
17335 2017-09-04  Richard Sandiford  <richard.sandiford@linaro.org>
17336             Alan Hayward  <alan.hayward@arm.com>
17337             David Sherwood  <david.sherwood@arm.com>
17339         * coretypes.h (pad_direction): New enum.
17340         * defaults.h (DEFAULT_FUNCTION_ARG_PADDING): Delete.
17341         (FUNCTION_ARG_PADDING): Likewise.
17342         * target.def (function_arg_padding): New hook.
17343         * targhooks.h (default_function_arg_padding): Declare.
17344         * targhooks.c (default_function_arg_padding): New function.
17345         * doc/tm.texi.in (FUNCTION_ARG_PADDING): Replace with...
17346         (TARGET_FUNCTION_ARG_PADDING): ...this.
17347         * doc/tm.texi: Regenerate.
17348         * calls.c (store_unaligned_arguments_into_pseudos): Use pad_direction
17349         instead of direction.
17350         (compute_argument_addresses): Likewise.
17351         (load_register_parameters): Likewise.
17352         (emit_library_call_value_1): Likewise.
17353         (store_one_arg): Use targetm.calls.function_arg_padding instead
17354         of FUNCTION_ARG_PADDING.
17355         (must_pass_in_stack_var_size_or_pad): Likewise.
17356         * expr.c (emit_group_load_1): Use pad_direction instead of direction.
17357         (emit_group_store): Likewise.
17358         (emit_single_push_insn_1): Use targetm.calls.function_arg_padding
17359         instead of FUNCTION_ARG_PADDING.
17360         (emit_push_insn): Likewise, and propagate enum change throughout
17361         function.
17362         * function.h (direction): Delete.
17363         (locate_and_pad_arg_data::where_pad): Use pad_direction instead
17364         of direction.
17365         * function.c (assign_parm_find_stack_rtl): Likewise.
17366         (assign_parm_setup_block_p): Likewise.
17367         (assign_parm_setup_block): Likewise.
17368         (gimplify_parameters): Likewise.
17369         (locate_and_pad_parm): Use targetm.calls.function_arg_padding
17370         instead of FUNCTION_ARG_PADDING, and propagate enum change throughout
17371         function.
17372         * config/aarch64/aarch64.h (FUNCTION_ARG_PADDING): Delete.
17373         (BLOCK_REG_PADDING): Use pad_direction instead of direction.
17374         * config/aarch64/aarch64-protos.h (aarch64_pad_arg_upward): Delete.
17375         * config/aarch64/aarch64.c (aarch64_pad_arg_upward): Replace with...
17376         (aarch64_function_arg_padding): ...this new function.
17377         (aarch64_gimplify_va_arg_expr): Use pad_direction instead of direction.
17378         (TARGET_FUNCTION_ARG_PADDING): Redefine.
17379         * config/arm/arm.h (FUNCTION_ARG_PADDING): Delete.
17380         (BLOCK_REG_PADDING): Use pad_direction instead of direction.
17381         * config/arm/arm-protos.h (arm_pad_arg_upward): Delete.
17382         * config/arm/arm.c (TARGET_FUNCTION_ARG_PADDING): Redefine.
17383         (arm_pad_arg_upward): Replace with...
17384         (arm_function_arg_padding): ...this new function.
17385         * config/c6x/c6x.h (BLOCK_REG_PADDING): Use pad_direction instead
17386         of direction.
17387         * config/ia64/hpux.h (FUNCTION_ARG_PADDING): Delete.
17388         * config/ia64/ia64-protos.h (ia64_hpux_function_arg_padding): Delete.
17389         * config/ia64/ia64.c (TARGET_FUNCTION_ARG_PADDING): Redefine.
17390         (ia64_hpux_function_arg_padding): Replace with...
17391         (ia64_function_arg_padding): ...this new function.  Use pad_direction
17392         instead of direction.  Check for TARGET_HPUX.
17393         * config/iq2000/iq2000.h (FUNCTION_ARG_PADDING): Delete.
17394         * config/iq2000/iq2000.c (TARGET_FUNCTION_ARG_PADDING): Redefine.
17395         (iq2000_function_arg_padding): New function.
17396         * config/mips/mips-protos.h (mips_pad_arg_upward): Delete.
17397         * config/mips/mips.c (mips_pad_arg_upward): Replace with...
17398         (mips_function_arg_padding): ...this new function.
17399         (mips_pad_reg_upward): Update accordingly.
17400         (TARGET_FUNCTION_ARG_PADDING): Redefine.
17401         * config/mips/mips.h (PAD_VARARGS_DOWN): Use
17402         targetm.calls.function_arg_padding.
17403         (FUNCTION_ARG_PADDING): Delete.
17404         (BLOCK_REG_PADDING): Use pad_direction instead of direction.
17405         * config/nios2/nios2.h (FUNCTION_ARG_PADDING): Delete.
17406         (PAD_VARARGS_DOWN): Use targetm.calls.function_arg_padding.
17407         * config/nios2/nios2-protos.h (nios2_function_arg_padding): Delete.
17408         (nios2_block_reg_padding): Return pad_direction instead of direction.
17409         * config/nios2/nios2.c (nios2_block_reg_padding): Return pad_direction
17410         instead of direction.
17411         (nios2_function_arg_padding): Likewise.  Make static.
17412         (TARGET_FUNCTION_ARG_PADDING): Redefine.
17413         * config/pa/pa.h (FUNCTION_ARG_PADDING): Delete.
17414         (BLOCK_REG_PADDING): Use targetm.calls.function_arg_padding.
17415         * config/pa/pa-protos.h (pa_function_arg_padding): Delete.
17416         * config/pa/pa.c (TARGET_FUNCTION_ARG_PADDING): Redefine.
17417         (pa_function_arg_padding): Make static.  Return pad_direction instead
17418         of direction.
17419         * config/powerpcspe/powerpcspe.h (FUNCTION_ARG_PADDING): Delete.
17420         (PAD_VARARGS_DOWN): Use targetm.calls.function_arg_padding.
17421         * config/powerpcspe/aix.h (BLOCK_REG_PADDING): Use pad_direction
17422         instead of direction.  Use targetm.calls.function_arg_padding.
17423         * config/powerpcspe/darwin.h (BLOCK_REG_PADDING): Likewise.
17424         * config/powerpcspe/freebsd64.h (BLOCK_REG_PADDING): Likewise.
17425         * config/powerpcspe/linux64.h (BLOCK_REG_PADDING): Likewise.
17426         * config/powerpcspe/powerpcspe-protos.h (function_arg_padding): Delete.
17427         * config/powerpcspe/powerpcspe.c (TARGET_FUNCTION_ARG_PADDING):
17428         Redefine.
17429         (function_arg_padding): Rename to...
17430         (rs6000_function_arg_padding): ...this.  Make static.  Return
17431         pad_direction instead of direction.
17432         (rs6000_return_in_msb): Use rs6000_function_arg_padding.
17433         * config/rs6000/rs6000.h (FUNCTION_ARG_PADDING): Delete.
17434         (PAD_VARARGS_DOWN): Use targetm.calls.function_arg_padding.
17435         * config/rs6000/aix.h (BLOCK_REG_PADDING): Use pad_direction
17436         instead of direction.  Use targetm.calls.function_arg_padding.
17437         * config/rs6000/darwin.h (BLOCK_REG_PADDING): Likewise.
17438         * config/rs6000/freebsd64.h (BLOCK_REG_PADDING): Likewise.
17439         * config/rs6000/linux64.h (BLOCK_REG_PADDING): Likewise.
17440         * config/rs6000/rs6000-protos.h (function_arg_padding): Delete.
17441         * config/rs6000/rs6000.c (TARGET_FUNCTION_ARG_PADDING): Redefine.
17442         (function_arg_padding): Rename to...
17443         (rs6000_function_arg_padding): ...this.  Make static.  Return
17444         pad_direction instead of direction.
17445         (rs6000_return_in_msb): Use rs6000_function_arg_padding.
17446         * config/s390/s390.h (FUNCTION_ARG_PADDING): Delete.
17447         * config/s390/s390.c (s390_function_arg_padding): New function.
17448         (TARGET_FUNCTION_ARG_PADDING): Redefine.
17449         * config/sparc/sparc.h (FUNCTION_ARG_PADDING): Delete.
17450         * config/sparc/sparc-protos.h (function_arg_padding): Delete.
17451         * config/sparc/sparc.c (TARGET_FUNCTION_ARG_PADDING): Redefine.
17452         (function_arg_padding): Rename to...
17453         (sparc_function_arg_padding): ...this.  Make static.  Return
17454         pad_direction instead of direction.
17455         * config/spu/spu.h (FUNCTION_ARG_PADDING): Delete.
17456         * config/spu/spu.c (spu_function_arg_padding): New function.
17457         (TARGET_FUNCTION_ARG_PADDING): Redefine.
17458         * system.h (FUNCTION_ARG_PADDING): Poison.
17460 2017-09-04  Richard Sandiford  <richard.sandiford@linaro.org>
17461             Alan Hayward  <alan.hayward@arm.com>
17462             David Sherwood  <david.sherwood@arm.com>
17464         * target.def (modes_tieable_p): New hook.
17465         * doc/tm.texi (MODES_TIEABLE_P): Replace with...
17466         (TARGET_MODES_TIEABLE_P): ...this.
17467         * doc/tm.texi.in: Regenerate.
17468         * hooks.h (hook_bool_mode_mode_true): Declare.
17469         * hooks.c (hook_bool_mode_mode_true): New function.
17470         * combine.c (subst): Use targetm.modes_tieable_p instead of
17471         MODES_TIEABLE_P.
17472         * dse.c (find_shift_sequence): Likewise.
17473         * expmed.c (extract_low_bits): Likewise.
17474         * lower-subreg.c: Include target.h.
17475         (find_decomposable_subregs): Use targetm.modes_tieable_p instead of
17476         MODES_TIEABLE_P.
17477         * rtlanal.c (rtx_cost): Likewise.
17478         * config/aarch64/aarch64.h (MODES_TIEABLE_P): Delete.
17479         * config/aarch64/aarch64-protos.h (aarch64_modes_tieable_p): Delete.
17480         * config/aarch64/aarch64.c (aarch64_modes_tieable_p): Make static.
17481         (TARGET_MODES_TIEABLE_P): Redefine.
17482         * config/alpha/alpha.h (MODES_TIEABLE_P): Delete.
17483         * config/alpha/alpha.c (alpha_modes_tieable_p): New function.
17484         (TARGET_MODES_TIEABLE_P): Redefine.
17485         * config/arc/arc.h (MODES_TIEABLE_P): Delete.
17486         * config/arc/arc.c (TARGET_MODES_TIEABLE_P): Redefine.
17487         (arc_modes_tieable_p): New function.
17488         * config/arm/arm.h (MODES_TIEABLE_P): Delete.
17489         * config/arm/arm-protos.h (arm_modes_tieable_p): Delete.
17490         * config/arm/arm.c (TARGET_MODES_TIEABLE_P): Redefine.
17491         (arm_modes_tieable_p): Make static.
17492         * config/avr/avr.h (MODES_TIEABLE_P): Delete.
17493         * config/bfin/bfin.h (MODES_TIEABLE_P): Delete.
17494         * config/bfin/bfin.c (bfin_modes_tieable_p): New function.
17495         (TARGET_MODES_TIEABLE_P): Redefine.
17496         * config/c6x/c6x.h (MODES_TIEABLE_P): Delete.
17497         * config/c6x/c6x.c (c6x_modes_tieable_p): New function.
17498         (TARGET_MODES_TIEABLE_P): Redefine.
17499         * config/cr16/cr16.h (MODES_TIEABLE_P): Delete.
17500         * config/cr16/cr16.c (TARGET_MODES_TIEABLE_P): Redefine.
17501         (cr16_modes_tieable_p): New function.
17502         * config/cris/cris.h (MODES_TIEABLE_P): Delete.
17503         * config/epiphany/epiphany.h (MODES_TIEABLE_P): Delete.
17504         * config/fr30/fr30.h (MODES_TIEABLE_P): Delete.
17505         (TRULY_NOOP_TRUNCATION): Update comment.
17506         * config/frv/frv.h (MODES_TIEABLE_P): Delete.
17507         (TRULY_NOOP_TRUNCATION): Update comment.
17508         * config/frv/frv.c (TARGET_MODES_TIEABLE_P): Redefine.
17509         (frv_modes_tieable_p): New function.
17510         * config/ft32/ft32.h (MODES_TIEABLE_P): Delete.
17511         * config/h8300/h8300.h (MODES_TIEABLE_P): Delete.
17512         * config/h8300/h8300.c (h8300_modes_tieable_p): New function.
17513         (TARGET_MODES_TIEABLE_P): Redefine.
17514         * config/i386/i386.h (MODES_TIEABLE_P): Delete.
17515         * config/i386/i386-protos.h (ix86_modes_tieable_p): Delete.
17516         * config/i386/i386.c (ix86_modes_tieable_p): Make static.
17517         (TARGET_MODES_TIEABLE_P): Redefine.
17518         * config/ia64/ia64.h (MODES_TIEABLE_P): Delete.
17519         * config/ia64/ia64.c (TARGET_MODES_TIEABLE_P): Redefine.
17520         (ia64_modes_tieable_p): New function.
17521         * config/iq2000/iq2000.h (MODES_TIEABLE_P): Delete.
17522         * config/iq2000/iq2000.c (TARGET_MODES_TIEABLE_P): Redefine.
17523         (iq2000_modes_tieable_p): New function.
17524         * config/lm32/lm32.h (MODES_TIEABLE_P): Delete.
17525         * config/lm32/lm32.c (TARGET_MODES_TIEABLE_P): Redefine.
17526         (lm32_modes_tieable_p): New function.
17527         * config/m32c/m32c.h (MODES_TIEABLE_P): Delete.
17528         * config/m32c/m32c-protos.h (m32c_modes_tieable_p): Delete.
17529         * config/m32c/m32c.c (m32c_modes_tieable_p): Make static.
17530         (TARGET_MODES_TIEABLE_P): Redefine.
17531         * config/m32r/m32r.h (MODES_TIEABLE_P): Delete.
17532         * config/m32r/m32r.c (TARGET_MODES_TIEABLE_P): Redefine.
17533         (m32r_modes_tieable_p): New function.
17534         * config/m68k/m68k.h (MODES_TIEABLE_P): Delete.
17535         * config/m68k/m68k.c (TARGET_MODES_TIEABLE_P): Redefine.
17536         (m68k_modes_tieable_p): New function.
17537         * config/mcore/mcore.h (MODES_TIEABLE_P): Delete.
17538         * config/mcore/mcore.c (TARGET_MODES_TIEABLE_P): Redefine.
17539         (mcore_modes_tieable_p): New function.
17540         * config/microblaze/microblaze.h (MODES_TIEABLE_P): Delete.
17541         * config/microblaze/microblaze.c (microblaze_modes_tieable_p): New
17542         function.
17543         (TARGET_MODES_TIEABLE_P): Redefine.
17544         * config/mips/mips.h (MODES_TIEABLE_P): Delete.
17545         * config/mips/mips-protos.h (mips_modes_tieable_p): Delete.
17546         * config/mips/mips.c (mips_modes_tieable_p): Make static.
17547         (TARGET_MODES_TIEABLE_P): Redefine.
17548         * config/mmix/mmix.h (MODES_TIEABLE_P): Delete.
17549         * config/mn10300/mn10300.h (MODES_TIEABLE_P): Delete.
17550         * config/mn10300/mn10300-protos.h (mn10300_modes_tieable): Delete.
17551         * config/mn10300/mn10300.c (mn10300_modes_tieable): Rename to...
17552         (mn10300_modes_tieable_p): ...this and make static.
17553         (TARGET_MODES_TIEABLE_P): Redefine.
17554         * config/moxie/moxie.h (MODES_TIEABLE_P): Delete.
17555         * config/msp430/msp430.h (MODES_TIEABLE_P): Delete.
17556         * config/msp430/msp430-protos.h (msp430_modes_tieable_p): Delete.
17557         * config/msp430/msp430.c (TARGET_MODES_TIEABLE_P): Redefine.
17558         (msp430_modes_tieable_p): Make static.
17559         * config/nds32/nds32.h (MODES_TIEABLE_P): Delete.
17560         * config/nds32/nds32.c (nds32_modes_tieable_p): New function.
17561         (TARGET_MODES_TIEABLE_P): Redefine.
17562         * config/nios2/nios2.h (MODES_TIEABLE_P): Delete.
17563         * config/nvptx/nvptx.h (MODES_TIEABLE_P): Delete.
17564         * config/nvptx/nvptx.c (nvptx_modes_tieable_p): New function.
17565         (TARGET_MODES_TIEABLE_P): Redefine.
17566         * config/pa/pa.h (MODES_TIEABLE_P): Delete.
17567         * config/pa/pa-protos.h (pa_modes_tieable_p): Delete.
17568         * config/pa/pa.c (pa_modes_tieable_p): Make static.
17569         (TARGET_MODES_TIEABLE_P): Redefine.
17570         * config/pdp11/pdp11.h (MODES_TIEABLE_P): Delete.
17571         * config/pdp11/pdp11.c (TARGET_MODES_TIEABLE_P): Redefine.
17572         (pdp11_modes_tieable_p): New function.
17573         * config/powerpcspe/powerpcspe.h (MODES_TIEABLE_P): Delete.
17574         * config/powerpcspe/powerpcspe.c (TARGET_MODES_TIEABLE_P): Redefine.
17575         (rs6000_modes_tieable_p): New function.
17576         (rs6000_debug_reg_global): Use it instead of MODES_TIEABLE_P.
17577         * config/powerpcspe/powerpcspe.md: Update comment.
17578         * config/riscv/riscv.h (MODES_TIEABLE_P): Delete.
17579         * config/riscv/riscv.c (riscv_modes_tieable_p): New function.
17580         (TARGET_MODES_TIEABLE_P): Redefine.
17581         * config/rl78/rl78.h (MODES_TIEABLE_P): Delete.
17582         * config/rl78/rl78.c (TARGET_MODES_TIEABLE_P): Redefine.
17583         (rl78_modes_tieable_p): New function.
17584         * config/rs6000/rs6000.h (MODES_TIEABLE_P): Delete.
17585         * config/rs6000/rs6000.c (TARGET_MODES_TIEABLE_P): Redefine.
17586         (rs6000_modes_tieable_p): New function.
17587         (rs6000_debug_reg_global): Use it instead of MODES_TIEABLE_P.
17588         * config/rs6000/rs6000.md: Update comment.
17589         * config/rx/rx.h (MODES_TIEABLE_P): Delete.
17590         * config/rx/rx.c (rx_modes_tieable_p): New function.
17591         (TARGET_MODES_TIEABLE_P): Redefine.
17592         * config/s390/s390.h (MODES_TIEABLE_P): Delete.
17593         * config/s390/s390.c (s390_modes_tieable_p): New function.
17594         (TARGET_MODES_TIEABLE_P): Redefine.
17595         * config/sh/sh.h (MODES_TIEABLE_P): Delete.
17596         * config/sh/sh.c (TARGET_MODES_TIEABLE_P): Redefine.
17597         (sh_modes_tieable_p): New function.
17598         * config/sparc/sparc.h (MODES_TIEABLE_P): Delete.
17599         * config/sparc/sparc-protos.h (sparc_modes_tieable_p): Delete.
17600         * config/sparc/sparc.c (TARGET_MODES_TIEABLE_P): Redefine.
17601         (sparc_modes_tieable_p): Make static.
17602         * config/spu/spu.h (MODES_TIEABLE_P): Delete.
17603         * config/spu/spu.c (spu_modes_tieable_p): New function.
17604         (TARGET_MODES_TIEABLE_P): Redefine.
17605         * config/stormy16/stormy16.h (MODES_TIEABLE_P): Delete.
17606         * config/stormy16/stormy16.c (xstormy16_modes_tieable_p): New function.
17607         (TARGET_MODES_TIEABLE_P): Redefine.
17608         * config/tilegx/tilegx.h (MODES_TIEABLE_P): Delete.
17609         * config/tilepro/tilepro.h (MODES_TIEABLE_P): Delete.
17610         * config/v850/v850.h (MODES_TIEABLE_P): Delete.
17611         * config/v850/v850.c (v850_modes_tieable_p): New function.
17612         (TARGET_MODES_TIEABLE_P): Redefine.
17613         * config/vax/vax.h (MODES_TIEABLE_P): Delete.
17614         * config/visium/visium.h (MODES_TIEABLE_P): Delete.
17615         * config/visium/visium.c (TARGET_MODES_TIEABLE_P): Redefine.
17616         (visium_modes_tieable_p): New function.
17617         * config/xtensa/xtensa.h (MODES_TIEABLE_P): Delete.
17618         * config/xtensa/xtensa.c (TARGET_MODES_TIEABLE_P): Redefine.
17619         (xtensa_modes_tieable_p): New function.
17620         * system.h (MODES_TIEABLE_P): Poison.
17622 2017-09-04  Richard Sandiford  <richard.sandiford@linaro.org>
17623             Alan Hayward  <alan.hayward@arm.com>
17624             David Sherwood  <david.sherwood@arm.com>
17626         * target.def (hard_regno_mode_ok): New hook.
17627         * doc/tm.texi (HARD_REGNO_MODE_OK): Replace with...
17628         (TARGET_HARD_REGNO_MODE_OK): ...this.
17629         * doc/tm.texi.in: Regenerate.
17630         * hooks.h (hook_bool_uint_mode_true): Declare.
17631         * hooks.c (hook_bool_uint_mode_true): New function.
17632         * doc/md.texi: Refer to targetm.hard_regno_mode_ok instead of
17633         HARD_REGNO_MODE_OK.
17634         * genpreds.c (write_insn_preds_c): Add an include of target.h.
17635         * alias.c (init_alias_target): Use targetm.hard_regno_mode_ok
17636         instead of HARD_REGNO_MODE_OK.
17637         * caller-save.c: Include target.h.
17638         (reg_save_code): Use targetm.hard_regno_mode_ok instead of
17639         HARD_REGNO_MODE_OK.
17640         * combine.c (can_combine_p): Likewise.
17641         (combinable_i3pat): Likewise.
17642         (can_change_dest_mode): Likewise.
17643         * expr.c (init_expr_target): Likewise.
17644         (convert_move): Likewise.
17645         (convert_modes): Likewise.
17646         * ira.c (setup_prohibited_class_mode_regs): Likewise.
17647         (setup_prohibited_mode_move_regs): Likewise.
17648         * ira.h (target_ira): Likewise.
17649         * lra-assigns.c (find_hard_regno_for_1): Likewise.
17650         * lra-constraints.c (process_alt_operands): Likewise.
17651         (split_reg): Likewise.
17652         * recog.c (peep2_find_free_register): Likewise.
17653         * ree.c (combine_reaching_defs): Likewise.
17654         * regcprop.c (maybe_mode_change): Likewise.
17655         * reginfo.c (init_reg_sets_1): Likewise.
17656         (choose_hard_reg_mode): Likewise.
17657         (simplifiable_subregs): Likewise.
17658         * regrename.c (check_new_reg_p): Likewise.
17659         * reload.c (find_valid_class): Likewise.
17660         (find_valid_class_1): Likewise.
17661         (reload_inner_reg_of_subreg): Likewise.
17662         (push_reload): Likewise.
17663         (combine_reloads): Likewise.
17664         (find_dummy_reload): Likewise.
17665         (find_reloads): Likewise.
17666         * reload1.c (find_reg): Likewise.
17667         (set_reload_reg): Likewise.
17668         (allocate_reload_reg): Likewise.
17669         (choose_reload_regs): Likewise.
17670         (reload_adjust_reg_for_temp): Likewise.
17671         * rtlanal.c (subreg_size_offset_from_lsb): Likewise.
17672         (simplify_subreg_regno): Likewise.
17673         * sel-sched.c (init_regs_for_mode): Likewise.
17674         * varasm.c (make_decl_rtl): Likewise.
17675         * config/aarch64/aarch64.h (HARD_REGNO_MODE_OK): Delete.
17676         (MODES_TIEABLE_P): Use targetm.hard_regno_mode_ok instead of
17677         HARD_REGNO_MODE_OK.
17678         * config/aarch64/aarch64-protos.h (aarch64_hard_regno_mode_ok): Delete.
17679         * config/aarch64/aarch64.c (aarch64_hard_regno_mode_ok): Make static.
17680         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17681         * config/alpha/alpha.h (HARD_REGNO_MODE_OK): Delete.
17682         * config/alpha/alpha.c (alpha_hard_regno_mode_ok): New function.
17683         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17684         * config/arc/arc.h (arc_hard_regno_mode_ok): Delete.
17685         (arc_mode_class): Delete.
17686         (HARD_REGNO_MODE_OK): Delete.
17687         * config/arc/arc.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17688         (arc_hard_regno_mode_ok): Rename old array to...
17689         (arc_hard_regno_mode_ok_modes): ...this.
17690         (arc_conditional_register_usage): Update accordingly.
17691         (arc_mode_class): Make static.
17692         (arc_hard_regno_mode_ok): New function.
17693         * config/arm/arm.h (HARD_REGNO_MODE_OK): Delete.
17694         * config/arm/arm-protos.h (arm_hard_regno_mode_ok): Delete.
17695         * config/arm/arm.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17696         (arm_hard_regno_mode_ok): Make static.
17697         * config/arm/arm.md (movdi): Use targetm.hard_regno_mode_ok instead of
17698         HARD_REGNO_MODE_OK.
17699         * config/avr/avr-protos.h (avr_hard_regno_mode_ok): Delete.
17700         * config/avr/avr.h (HARD_REGNO_MODE_OK): Delete.
17701         * config/avr/avr.c (avr_hard_regno_mode_ok): Make static and
17702         return a bool.
17703         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17704         * config/bfin/bfin-protos.h (hard_regno_mode_ok): Delete.
17705         * config/bfin/bfin.h (HARD_REGNO_MODE_OK): Delete.
17706         * config/bfin/bfin.c (hard_regno_mode_ok): Rename to...
17707         (bfin_hard_regno_mode_ok): ...this.  Make static and return a bool.
17708         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17709         * config/bfin/predicates.md (valid_reg_operand): Use
17710         targetm.hard_regno_mode_ok instead of HARD_REGNO_MODE_OK.
17711         * config/c6x/c6x.h (HARD_REGNO_MODE_OK): Delete.
17712         * config/c6x/c6x.c (c6x_hard_regno_mode_ok): New function.
17713         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17714         * config/cr16/cr16.h (HARD_REGNO_MODE_OK): Delete.
17715         * config/cr16/cr16-protos.h (cr16_hard_regno_mode_ok): Delete.
17716         * config/cr16/cr16.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17717         (cr16_hard_regno_mode_ok): Make static and return a bool.
17718         * config/cris/cris.h (HARD_REGNO_MODE_OK): Delete.
17719         * config/cris/cris.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17720         (cris_hard_regno_mode_ok): New function.
17721         * config/epiphany/epiphany.h (epiphany_hard_regno_mode_ok): Delete.
17722         (epiphany_mode_class): Delete.
17723         (HARD_REGNO_MODE_OK): Delete.
17724         * config/epiphany/epiphany-protos.h (hard_regno_mode_ok): Delete.
17725         * config/epiphany/epiphany.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17726         (hard_regno_mode_ok): Rename to...
17727         (epiphany_hard_regno_mode_ok): ...this.  Make static and return a bool.
17728         * config/fr30/fr30.h (HARD_REGNO_MODE_OK): Delete.
17729         * config/fr30/fr30.md: Refer to targetm.hard_regno_mode_ok instead of
17730         HARD_REGNO_MODE_OK.
17731         * config/frv/frv.h (HARD_REGNO_MODE_OK): Delete.
17732         * config/frv/frv-protos.h (frv_hard_regno_mode_ok): Delete.
17733         * config/frv/frv.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17734         (frv_hard_regno_mode_ok): Make static and return a bool.
17735         * config/frv/frv.md: Refer to targetm.hard_regno_mode_ok instead of
17736         HARD_REGNO_MODE_OK.
17737         * config/ft32/ft32.h (HARD_REGNO_MODE_OK): Delete.
17738         * config/h8300/h8300.h (HARD_REGNO_MODE_OK): Delete.
17739         * config/h8300/h8300-protos.h (h8300_hard_regno_mode_ok): Delete.
17740         * config/h8300/h8300.c (h8300_hard_regno_mode_ok): Make static
17741         and return a bool.
17742         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17743         * config/i386/i386.h (HARD_REGNO_MODE_OK): Delete.
17744         * config/i386/i386-protos.h (ix86_hard_regno_mode_ok): Delete.
17745         * config/i386/i386.c (ix86_hard_regno_mode_ok): Make static and
17746         return a bool.
17747         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17748         * config/ia64/ia64.h (HARD_REGNO_MODE_OK): Delete.
17749         * config/ia64/ia64.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17750         (ia64_hard_regno_mode_ok): New function.
17751         * config/iq2000/iq2000.h (HARD_REGNO_MODE_OK): Delete.
17752         * config/iq2000/iq2000.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17753         (iq2000_hard_regno_mode_ok): New function.
17754         * config/lm32/lm32.h (HARD_REGNO_MODE_OK): Delete.
17755         * config/lm32/lm32.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17756         (lm32_hard_regno_mode_ok): New function.
17757         * config/m32c/m32c.h (HARD_REGNO_MODE_OK): Delete.
17758         * config/m32c/m32c-protos.h (m32c_hard_regno_ok): Delete.
17759         * config/m32c/m32c.c (class_can_hold_mode): Use m32c_hard_regno_mode_ok
17760         instead of HARD_REGNO_MODE_OK.
17761         (m32c_hard_regno_ok): Rename to...
17762         (m32c_hard_regno_mode_ok): ...this.  Make static and return a bool.
17763         (m32c_cannot_change_mode_class): Update accordingly.
17764         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17765         * config/m32r/m32r.h (m32r_hard_regno_mode_ok): Delete.
17766         (m32r_mode_class): Delete.
17767         (HARD_REGNO_MODE_OK): Delete.
17768         * config/m32r/m32r.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17769         (m32r_hard_regno_mode_ok): Rename to...
17770         (m32r_hard_regno_modes): ...this.
17771         (m32r_mode_class): Make static.
17772         (m32r_hard_regno_mode_ok): New function.
17773         * config/m68k/m68k.h (HARD_REGNO_MODE_OK): Delete.
17774         * config/m68k/m68k-protos.h (m68k_regno_mode_ok): Delete.
17775         * config/m68k/m68k.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17776         (m68k_hard_regno_mode_ok): Make static.
17777         * config/mcore/mcore.h (HARD_REGNO_MODE_OK): Delete.
17778         * config/mcore/mcore.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17779         (mcore_hard_regno_mode_ok): New function.
17780         * config/microblaze/microblaze.h (microblaze_hard_regno_mode_ok)
17781         (HARD_REGNO_MODE_OK): Delete.
17782         * config/microblaze/microblaze.c (microblaze_hard_regno_mode_ok):
17783         Rename to...
17784         (microblaze_hard_regno_mode_ok_p): ...this and make static.
17785         (microblaze_hard_regno_mode_ok): New function.
17786         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17787         * config/mips/mips.h (HARD_REGNO_MODE_OK): Delete.
17788         (mips_hard_regno_mode_ok): Delete.
17789         * config/mips/mips.c (mips_hard_regno_mode_ok): Rename to...
17790         (mips_hard_regno_mode_ok_p): ...this and make static.
17791         (mips_hard_regno_mode_ok_p): Rename to...
17792         (mips_hard_regno_mode_ok_uncached): ...this.
17793         (mips_hard_regno_mode_ok): New function.
17794         (mips_class_max_nregs): Use mips_hard_regno_mode_ok instead
17795         of HARD_REGNO_MODE_OK.
17796         (mips_option_override): Update after above name changes.
17797         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17798         * config/mmix/mmix.h (HARD_REGNO_MODE_OK): Delete.
17799         * config/mn10300/mn10300.h (HARD_REGNO_MODE_OK): Delete.
17800         * config/mn10300/mn10300-protos.h (mn10300_hard_regno_mode_ok): Delete.
17801         * config/mn10300/mn10300.c (mn10300_hard_regno_mode_ok): Make static.
17802         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17803         * config/moxie/moxie.h (HARD_REGNO_MODE_OK): Delete.
17804         * config/msp430/msp430.h (HARD_REGNO_MODE_OK): Delete.
17805         * config/msp430/msp430-protos.h (msp430_hard_regno_mode_ok): Delete.
17806         * config/msp430/msp430.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17807         (msp430_hard_regno_mode_ok): Make static and return a bool.
17808         * config/nds32/nds32.h (HARD_REGNO_MODE_OK): Delete.
17809         * config/nds32/nds32-protos.h (nds32_hard_regno_mode_ok): Delete.
17810         * config/nds32/nds32.c (nds32_hard_regno_mode_ok): Make static
17811         and return a bool.
17812         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17813         * config/nios2/nios2.h (HARD_REGNO_MODE_OK): Delete.
17814         * config/nvptx/nvptx.h (HARD_REGNO_MODE_OK): Delete.
17815         * config/pa/pa.h (MODES_TIEABLE_P): Update commentary.
17816         * config/pa/pa32-regs.h (HARD_REGNO_MODE_OK): Rename to...
17817         (PA_HARD_REGNO_MODE_OK): ...this
17818         * config/pa/pa64-regs.h (HARD_REGNO_MODE_OK): Rename to...
17819         (PA_HARD_REGNO_MODE_OK): ...this.
17820         * config/pa/pa.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17821         (pa_hard_regno_mode_ok): New function.
17822         * config/pdp11/pdp11.h (HARD_REGNO_MODE_OK): Delete.
17823         * config/pdp11/pdp11.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17824         (pdp11_hard_regno_mode_ok): New function.
17825         * config/powerpcspe/powerpcspe.h (HARD_REGNO_MODE_OK): Delete.
17826         * config/powerpcspe/powerpcspe-protos.h (rs6000_hard_regno_mode_ok_p):
17827         Delete.
17828         * config/powerpcspe/powerpcspe.c (rs6000_hard_regno_mode_ok_p):
17829         Make static.
17830         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17831         (rs6000_hard_regno_mode_ok): Rename to...
17832         (rs6000_hard_regno_mode_ok_uncached): ...this.
17833         (rs6000_init_hard_regno_mode_ok): Update accordingly.
17834         (rs6000_hard_regno_mode_ok): New function.
17835         * config/riscv/riscv.h (HARD_REGNO_MODE_OK): Delete.
17836         * config/riscv/riscv-protos.h (riscv_hard_regno_mode_ok_p): Delete.
17837         * config/riscv/riscv.c (riscv_hard_regno_mode_ok_p): Rename to...
17838         (riscv_hard_regno_mode_ok): ...this and make static.
17839         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17840         * config/rl78/rl78.h (HARD_REGNO_MODE_OK): Delete.
17841         * config/rl78/rl78-protos.h (rl78_hard_regno_mode_ok): Delete.
17842         * config/rl78/rl78.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17843         (rl78_hard_regno_mode_ok): Make static and return bool.
17844         * config/rs6000/rs6000.h (HARD_REGNO_MODE_OK): Delete.
17845         * config/rs6000/rs6000-protos.h (rs6000_hard_regno_mode_ok_p):
17846         Delete.
17847         * config/rs6000/rs6000.c (rs6000_hard_regno_mode_ok_p): Make static.
17848         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17849         (rs6000_hard_regno_mode_ok): Rename to...
17850         (rs6000_hard_regno_mode_ok_uncached): ...this.
17851         (rs6000_init_hard_regno_mode_ok): Update accordingly.
17852         (rs6000_hard_regno_mode_ok): New function.
17853         * config/rx/rx.h (HARD_REGNO_MODE_OK): Delete.
17854         * config/rx/rx.c (rx_hard_regno_mode_ok): New function.
17855         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17856         * config/s390/s390.h (HARD_REGNO_MODE_OK): Delete.
17857         * config/s390/s390-protos.h (s390_hard_regno_mode_ok): Delete.
17858         * config/s390/s390.c (s390_hard_regno_mode_ok): Make static.
17859         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17860         * config/sh/sh.h (HARD_REGNO_MODE_OK): Delete.
17861         * config/sh/sh-protos.h (sh_hard_regno_mode_ok): Delete.
17862         * config/sh/sh.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17863         (sh_hard_regno_mode_ok): Make static.
17864         * config/sparc/constraints.md: Refer to targetm.hard_regno_mode_ok
17865         instead of HARD_REGNO_MODE_OK.
17866         * config/sparc/sparc.h (hard_regno_mode_classes): Delete.
17867         (sparc_mode_class): Delete.
17868         (HARD_REGNO_MODE_OK): Delete.
17869         * config/sparc/sparc.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17870         (hard_regno_mode_classes): Make static.
17871         (sparc_mode_class): Likewise.
17872         (sparc_hard_regno_mode_ok): New function.
17873         * config/spu/spu.h (HARD_REGNO_MODE_OK): Delete.
17874         * config/stormy16/stormy16.h (HARD_REGNO_MODE_OK): Delete.
17875         * config/stormy16/stormy16.c (xstormy16_hard_regno_mode_ok): New
17876         function.
17877         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17878         * config/tilegx/tilegx.h (HARD_REGNO_MODE_OK): Delete.
17879         * config/tilepro/tilepro.h (HARD_REGNO_MODE_OK): Delete.
17880         * config/v850/v850.h (HARD_REGNO_MODE_OK): Delete.
17881         * config/v850/v850.c (v850_hard_regno_mode_ok): New function.
17882         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17883         * config/vax/vax.h (HARD_REGNO_MODE_OK): Delete.
17884         * config/visium/visium.h (HARD_REGNO_MODE_OK): Delete.
17885         * config/visium/visium.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17886         (visium_hard_regno_mode_ok): New function.
17887         * config/visium/visium.md: Refer to targetm.hard_regno_mode_ok
17888         instead of HARD_REGNO_MODE_OK.
17889         * config/xtensa/xtensa.h (xtensa_hard_regno_mode_ok): Delete.
17890         (HARD_REGNO_MODE_OK): Delete.
17891         * config/xtensa/xtensa.c (xtensa_hard_regno_mode_ok): Rename to...
17892         (xtensa_hard_regno_mode_ok_p): ...this and make static.
17893         (xtensa_option_override): Update accordingly.
17894         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17895         (xtensa_hard_regno_mode_ok): New function.
17896         * system.h (HARD_REGNO_MODE_OK): Poison.
17898 2017-09-04  Richard Sandiford  <richard.sandiford@linaro.org>
17899             Alan Hayward  <alan.hayward@arm.com>
17900             David Sherwood  <david.sherwood@arm.com>
17902         * target.def (hard_regno_call_part_clobbered): New hook.
17903         * doc/tm.texi.in (HARD_REGNO_CALL_PART_CLOBBERED): Replace with...
17904         (TARGET_HARD_REGNO_CALL_PART_CLOBBERED): ...this hook.
17905         * doc/tm.texi: Regenerate.
17906         * hooks.h (hook_bool_uint_mode_false): Declare.
17907         * hooks.c (hook_bool_uint_mode_false): New function.
17908         * regs.h (HARD_REGNO_CALL_PART_CLOBBERED): Delete.
17909         * cselib.c (cselib_process_insn): Use
17910         targetm.hard_regno_call_part_clobbered instead of
17911         HARD_REGNO_CALL_PART_CLOBBERED.
17912         * ira-conflicts.c (ira_build_conflicts): Likewise.
17913         * ira-costs.c (ira_tune_allocno_costs): Likewise.
17914         * lra-constraints.c (need_for_call_save_p): Likewise.
17915         * lra-lives.c: Include target.h.
17916         (check_pseudos_live_through_calls): Use
17917         targetm.hard_regno_call_part_clobbered instead of
17918         HARD_REGNO_CALL_PART_CLOBBERED.
17919         * regcprop.c: Include target.h.
17920         (copyprop_hardreg_forward_1): Use
17921         targetm.hard_regno_call_part_clobbered instead of
17922         HARD_REGNO_CALL_PART_CLOBBERED.
17923         * reginfo.c (choose_hard_reg_mode): Likewise.
17924         * regrename.c (check_new_reg_p): Likewise.
17925         * reload.c (find_equiv_reg): Likewise.
17926         * reload1.c (emit_reload_insns): Likewise.
17927         * sched-deps.c (deps_analyze_insn): Likewise.
17928         * sel-sched.c (init_regs_for_mode): Likewise.
17929         (mark_unavailable_hard_regs): Likewise.
17930         * targhooks.c (default_dwarf_frame_reg_mode): Likewise.
17931         * config/aarch64/aarch64.h (HARD_REGNO_CALL_PART_CLOBBERED): Delete.
17932         * config/aarch64/aarch64.c (aarch64_hard_regno_call_part_clobbered):
17933         New function.
17934         (TARGET_HARD_REGNO_CALL_PART_CLOBBERED): Redefine.
17935         * config/avr/avr.h (HARD_REGNO_CALL_PART_CLOBBERED): Delete.
17936         * config/avr/avr-protos.h (avr_hard_regno_call_part_clobbered):
17937         Delete.
17938         * config/avr/avr.c (avr_hard_regno_call_part_clobbered): Make static
17939         and return a bool.
17940         (TARGET_HARD_REGNO_CALL_PART_CLOBBERED): Redefine.
17941         * config/i386/i386.h (HARD_REGNO_CALL_PART_CLOBBERED): Delete.
17942         * config/i386/i386.c (ix86_hard_regno_call_part_clobbered): New
17943         function.
17944         (TARGET_HARD_REGNO_CALL_PART_CLOBBERED): Redefine.
17945         * config/mips/mips.h (HARD_REGNO_CALL_PART_CLOBBERED): Delete.
17946         * config/mips/mips.c (mips_hard_regno_call_part_clobbered): New
17947         function.
17948         (TARGET_HARD_REGNO_CALL_PART_CLOBBERED): Redefine.
17949         * config/powerpcspe/powerpcspe.h (HARD_REGNO_CALL_PART_CLOBBERED):
17950         Delete.
17951         * config/powerpcspe/powerpcspe.c
17952         (rs6000_hard_regno_call_part_clobbered): New function.
17953         (TARGET_HARD_REGNO_CALL_PART_CLOBBERED): Redefine.
17954         * config/rs6000/rs6000.h (HARD_REGNO_CALL_PART_CLOBBERED): Delete.
17955         * config/rs6000/rs6000.c (rs6000_hard_regno_call_part_clobbered):
17956         New function.
17957         (TARGET_HARD_REGNO_CALL_PART_CLOBBERED): Redefine.
17958         * config/s390/s390.h (HARD_REGNO_CALL_PART_CLOBBERED): Delete.
17959         * config/s390/s390.c (s390_hard_regno_call_part_clobbered): New
17960         function.
17961         (TARGET_HARD_REGNO_CALL_PART_CLOBBERED): Redefine.
17962         * config/sh/sh.h (HARD_REGNO_CALL_PART_CLOBBERED): Delete.
17963         * system.h (HARD_REGNO_CALL_PART_CLOBBERED): Poison.
17965 2017-09-04  Richard Sandiford  <richard.sandiford@linaro.org>
17966             Alan Hayward  <alan.hayward@arm.com>
17967             David Sherwood  <david.sherwood@arm.com>
17969         * rtl.h (subreg_memory_offset): Declare.
17970         * emit-rtl.c (subreg_memory_offset): New function.
17971         * expmed.c (store_bit_field_1): Use it.
17972         * expr.c (undefined_operand_subword_p): Likewise.
17973         * simplify-rtx.c (simplify_subreg): Likewise.
17975 2017-09-04  Alexander Monakov  <amonakov@ispras.ru>
17977         PR rtl-optimization/57448
17978         PR target/67458
17979         PR target/81316
17980         * optabs.c (expand_atomic_load): Place compiler memory barriers if
17981         using atomic_load pattern.
17982         (expand_atomic_store): Likewise.
17984 2017-09-04  Jakub Jelinek  <jakub@redhat.com>
17986         PR sanitizer/81981
17987         * gimple-fold.c (gimple_fold_call): Optimize away useless UBSAN_PTR
17988         and UBSAN_BOUNDS internal calls.  Clean up IFN_UBSAN_OBJECT_SIZE
17989         handling.  Use replace_call_with_value with NULL instead of
17990         gsi_replace, unlink_stmt_vdef and release_defs.
17992         * gdbhooks.py (OptMachineModePrinter.to_string): Use 8 spaces
17993         instead of tab.
17995         * lra-remat.c (reg_overlap_for_remat_p): Fix a pasto.
17997 2017-09-04  Richard Sandiford  <richard.sandiford@linaro.org>
17999         PR bootstrap/82045
18000         * rtl.h (emit_library_call_value_1): Declare.
18001         (emit_library_call): Replace declaration with a series of overloads.
18002         Remove the parameter count argument.
18003         (emit_library_call_value): Likewise.
18004         * calls.c (emit_library_call_value_1): Make global.  Replace varargs
18005         with an "rtx_mode_t *".
18006         (emit_library_call_value): Delete.
18007         (emit_library_call): Likewise.
18008         * asan.c (asan_emit_stack_protection): Update calls accordingly.
18009         (asan_emit_allocas_unpoison): Likewise.
18010         * builtins.c (expand_builtin_powi): Likewise.
18011         (expand_asan_emit_allocas_unpoison): Likewise.
18012         * cfgexpand.c (expand_main_function): Likewise.
18013         * config/aarch64/aarch64.c (aarch64_trampoline_init): Likewise.
18014         * config/aarch64/aarch64.h (PROFILE_HOOK): Likewise.
18015         * config/alpha/alpha.c (alpha_trampoline_init): Likewise.
18016         * config/arm/arm.c (arm_trampoline_init): Likewise.
18017         (arm_call_tls_get_addr): Likewise.
18018         (arm_expand_divmod_libfunc): Likewise.
18019         * config/bfin/bfin.md (umulsi3_highpart): Likewise.
18020         (smulsi3_highpart): Likewise.
18021         * config/c6x/c6x.c (c6x_initialize_trampoline): Likewise.
18022         (c6x_expand_compare): Likewise.
18023         (c6x_expand_movmem): Likewise.
18024         * config/frv/frv.c (frv_trampoline_init): Likewise.
18025         * config/i386/i386.c (ix86_trampoline_init): Likewise.
18026         (ix86_expand_divmod_libfunc): Likewise.
18027         * config/ia64/ia64.c (ia64_expand_tls_address): Likewise.
18028         (ia64_expand_compare): Likewise.
18029         (ia64_profile_hook): Likewise.
18030         * config/ia64/ia64.md (save_stack_nonlocal): Likewise.
18031         (nonlocal_goto): Likewise.
18032         (restore_stack_nonlocal): Likewise.
18033         * config/m32r/m32r.c (block_move_call): Likewise.
18034         (m32r_trampoline_init): Likewise.
18035         * config/m68k/linux.h (FINALIZE_TRAMPOLINE): Likewise.
18036         * config/m68k/m68k.c (m68k_call_tls_get_addr): Likewise.
18037         (m68k_call_m68k_read_tp): Likewise.
18038         * config/microblaze/microblaze.c (microblaze_call_tls_get_addr)
18039         (microblaze_expand_divide): Likewise.
18040         * config/mips/mips.h (mips_args): Likewise.
18041         * config/mips/sdemtk.h (mips_sync_icache): Likewise.
18042         (MIPS_ICACHE_SYNC): Likewise.
18043         * config/nios2/nios2.c (nios2_emit_expensive_div): Likewise.
18044         (nios2_trampoline_init): Likewise.
18045         * config/pa/pa.c (hppa_tls_call): Likewise.
18046         (pa_trampoline_init): Likewise.
18047         * config/pa/pa.md (canonicalize_funcptr_for_compare): Likewise.
18048         * config/powerpcspe/powerpcspe.c (rs6000_legitimize_tls_address)
18049         (expand_strn_compare): Likewise.
18050         (rs6000_generate_compare): Likewise.
18051         (rs6000_expand_float128_convert): Likewise.
18052         (output_profile_hook): Likewise.
18053         (rs6000_trampoline_init): Likewise.
18054         * config/powerpcspe/powerpcspe.md (neg<mode>2): Likewise.
18055         * config/riscv/riscv.h (PROFILE_HOOK): Likewise.
18056         * config/rs6000/rs6000-string.c (expand_strn_compare): Likewise.
18057         * config/rs6000/rs6000.c (rs6000_legitimize_tls_address): Likewise.
18058         (rs6000_generate_compare): Likewise.
18059         (rs6000_expand_float128_convert): Likewise.
18060         (output_profile_hook): Likewise.
18061         (rs6000_trampoline_init): Likewise.
18062         * config/rs6000/rs6000.md (neg<mode>2): Likewise.
18063         * config/sh/sh.c (sh_trampoline_init): Likewise.
18064         * config/sparc/sparc.c (emit_soft_tfmode_libcall): Likewise.
18065         (sparc_emit_float_lib_cmp): Likewise.
18066         (sparc32_initialize_trampoline): Likewise.
18067         (sparc64_initialize_trampoline): Likewise.
18068         (sparc_profile_hook): Likewise.
18069         * config/spu/spu.c (ea_load_store): Likewise.
18070         * config/spu/spu.md (floatunssidf2): Likewise.
18071         * config/tilegx/tilegx.c (tilegx_trampoline_init): Likewise.
18072         * config/tilepro/tilepro.c (tilepro_trampoline_init): Likewise.
18073         * config/visium/visium.c (expand_block_move_4): Likewise.
18074         (expand_block_move_2): Likewise.
18075         (expand_block_move_1): Likewise.
18076         (expand_block_set_4): Likewise.
18077         (expand_block_set_2): Likewise.
18078         (expand_block_set_1): Likewise.
18079         (visium_trampoline_init): Likewise.
18080         (visium_profile_hook): Likewise.
18081         * config/xtensa/xtensa.c (xtensa_expand_nonlocal_goto): Likewise.
18082         (xtensa_setup_frame_addresses): Likewise.
18083         (xtensa_trampoline_init): Likewise.
18084         * except.c (sjlj_emit_function_enter): Likewise.
18085         (sjlj_emit_function_exit): Likewise.
18086         * explow.c (allocate_dynamic_stack_space): Likewise.
18087         (probe_stack_range): Likewise.
18088         * expr.c (convert_mode_scalar): Likewise.
18089         * optabs.c (expand_binop): Likewise.
18090         (expand_twoval_binop_libfunc): Likewise.
18091         (expand_unop): Likewise.
18092         (prepare_cmp_insn): Likewise.
18093         (prepare_float_lib_cmp): Likewise.
18094         (expand_float): Likewise.
18095         (expand_fix): Likewise.
18096         (expand_fixed_convert): Likewise.
18097         (maybe_emit_sync_lock_test_and_set): Likewise.
18098         (expand_atomic_compare_and_swap): Likewise.
18099         (expand_mem_thread_fence): Likewise.
18100         (expand_atomic_fetch_op): Likewise.
18102 2017-09-03  Gerald Pfeifer  <gerald@pfeifer.com>
18104         * doc/generic.texi (OpenACC): Adjust URL.
18105         * doc/invoke.texi (C Dialect Options): Ditto.
18107 2017-09-03  Uros Bizjak  <ubizjak@gmail.com>
18109         * config/i386/i386.md (*bt<mode>): Use nonimmediate_operand
18110         predicate for operand 1.  Add (m,<S>) constraint.
18111         (*jcc_bt<mode>): Use nonimmediate_operand predicate for operand 1.
18112         Prevent memory operand 1 with register operand 2.
18114 2017-09-01  Segher Boessenkool  <segher@kernel.crashing.org>
18116         PR rtl-optimization/82024
18117         * combine.c (try_combine): If the combination result is a PARALLEL,
18118         and we only need to retain the SET in there that would be placed
18119         at I2, check that we can place that at I3 instead, before doing so.
18121 2017-09-01  Jakub Jelinek  <jakub@redhat.com>
18123         PR target/81766
18124         * config/i386/i386.c (ix86_init_large_pic_reg): Return label
18125         instead of void.
18126         (ix86_init_pic_reg): Remember label from ix86_init_large_pic_reg,
18127         if non-NULL and preceded by NOTE_INSN_BASIC_BLOCK, swap the note
18128         and label.
18130 2017-09-01  Joerg Sonnenberger  <joerg@bec.de>
18131             Jeff Law  <law@redhat.com>
18133         * varasm.c (bss_initializer_p): Do not put constants into .bss
18134         (categorize_decl_for_section): Handle bss_initializer_p returning
18135         false when DECL_INITIAL is NULL.
18137 2017-09-01  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
18139         PR target/82012
18140         * config/s390/s390.c (s390_can_inline_p): New function.
18142 2017-09-01  Jeff Law  <law@redhat.com>
18144         PR tree-optimization/82052
18145         * tree-ssa-scopedtables.c (avail_exprs_stack::lookup_avail_expr):
18146         Always initialize the returned slot after a hash table miss
18147         when INSERT is true.
18149 2017-09-01  Alexander Monakov  <amonakov@ispras.ru>
18151         * config/s390/s390.md (mem_signal_fence): Remove.
18152         * doc/md.texi (mem_signal_fence): Remove.
18153         * optabs.c (expand_mem_signal_fence): Remove uses of mem_signal_fence.
18154         Update comments.
18155         * target-insns.def (mem_signal_fence): Remove.
18157 2017-09-01  Jakub Jelinek  <jakub@redhat.com>
18159         PR sanitizer/81902
18160         * doc/invoke.texi: Document -fsanitize=pointer-overflow.
18162         PR sanitizer/81923
18163         * asan.c (create_odr_indicator): Strip name encoding from assembler
18164         name before appending it after __odr_asan_.
18166 2017-09-01  Martin Liska  <mliska@suse.cz>
18168         PR tree-optimization/82059
18169         * gimple-ssa-isolate-paths.c (isolate_path): Add profile and
18170         frequency only when an edge is redirected.
18172 2017-09-01  Claudiu Zissulescu  <claziss@synopsys.com>
18174         * config/arc/arc-c.c (__ARC_LPC_WIDTH__): Add builtin define.
18175         * config/arc/arc.c (ARC_MAX_LOOP_LENGTH): Define.
18176         (arc_conditional_register_usage): Remove ARC600 lp_count
18177         exception.
18178         (arc_file_start): Emit Tag_ARC_CPU_variation.
18179         (arc_can_use_doloop_p): New conditions to use ZOLs.
18180         (hwloop_fail): New function.
18181         (hwloop_optimize): Likewise.
18182         (hwloop_pattern_reg): Likewise.
18183         (arc_doloop_hooks): New struct, to be used with reorg_loops.
18184         (arc_reorg_loops): New function, calls reorg_loops.
18185         (arc_reorg): Call arc_reorg_loops.  Remove old ZOL handling.
18186         (arc600_corereg_hazard): Remove ZOL checking, case handled by
18187         hwloop_optimize.
18188         (arc_loop_hazard): Remove function, functionality moved into
18189         hwloop_optimize.
18190         (arc_hazard): Remove arc_loop_hazard call.
18191         (arc_adjust_insn_length): Remove ZOL handling, functionality moved
18192         into hwloop_optimize.
18193         (arc_label_align): Remove ZOL handling.
18194         * config/arc/arc.h (LOOP_ALIGN): Changed to 0.
18195         * config/arc/arc.md (doloop_begin): Remove pattern.
18196         (doloop_begin_i): Likewise.
18197         (doloop_end_i): Likewise.
18198         (doloop_fallback): Likewise.
18199         (doloop_fallback_m): Likewise.
18200         (doloop_end): Reimplement expand.
18201         (arc_lp): New pattern for LP instruction.
18202         (loop_end): New pattern.
18203         (loop_fail): Likewise.
18204         (decrement_and_branch_until_zero): Likewise.
18205         * config/arc/arc.opt (mlpc-width): New option.
18206         * doc/invoke.texi (mlpc-width): Document option.
18208 2017-09-01  Claudiu Zissulescu  <claziss@synopsys.com>
18210         * config/arc/arc.c (arc_ifcvt): Remove use of merge_blocks call.
18211         (arc_ccfsm_advance): Fix checking for delay slots.
18212         (arc_reorg): Add rtl dump after each call to arc_ifcvt.
18214 2017-09-01  Claudiu Zissulescu  <claziss@synopsys.com>
18216         * config/arc/arc.md (movqi_insn): Add stores to save constant long
18217         immediates.
18218         (movhi_insn): Update store instruction constraint which are saving
18219         6-bit short immediates.
18220         (movsi_insn): Consider also short scaled load operations.
18221         (zero_extendhisi2_i): Use Usd constraint instead of T.
18222         (extendhisi2_i): Add q constraint.
18223         (arc_clzsi2): Add type and length attributes.
18224         (arc_ctzsi2): Likewise.
18225         * config/arc/constraints.md (Usc): Update constraint, the
18226         assembler can parse two relocations for a single instruction.
18228 2017-09-01  Claudiu Zissulescu  <claziss@synopsys.com>
18230         * config/arc/arc.c (arc_use_anchors_for_symbol_p): New function.
18231         (TARGET_USE_ANCHORS_FOR_SYMBOL_P): Define.
18233 2017-08-31  Olivier Hainque  <hainque@adacore.com>
18235         * config.gcc (powerpc-wrs-vxworks|vxworksae|vxworksmils): Now
18236         match as powerpc-wrs-vxworks*.
18238 2017-08-31  James Greenhalgh  <james.greenhalgh@arm.com>
18240         * config/aarch64/aarch64-simd.md (aarch64_mla_elt_merge<mode>): Fix
18241         register constraint for by-element operand.
18242         (aarch64_mls_elt_merge<mode>): Likewise.
18244 2017-08-31  Claudiu Zissulescu  <claziss@synopsys.com>
18246         * config/arc/arc.c (arc_can_follow_jump): Check for short
18247         branches.
18249 2017-08-31  Claudiu Zissulescu  <claziss@synopsys.com>
18251         * config.gcc: Use g.opt for arc.
18252         * config/arc/arc.c (LEGITIMATE_SCALED_ADDRESS_P): Deleted,
18253         functionality moved to ...
18254         (legitimate_scaled_address_p): New function, ...here.
18255         (LEGITIMATE_SMALL_DATA_OFFSET_P): New define.
18256         (LEGITIMATE_SMALL_DATA_ADDRESS_P): Use the above define.
18257         (legitimate_offset_address_p): Delete TARGET_NO_SDATA_SET
18258         condition.
18259         (arc_override_options): Handle G option.
18260         (arc_output_pic_addr_const): Correct function definition.
18261         (arc_legitimate_address_p): Use legitimate_scaled_address_p.
18262         (arc_decl_anon_ns_mem_p): Delete.
18263         (arc_in_small_data_p): Overhaul this function to take into
18264         consideration the value given via G option.
18265         (arc_rewrite_small_data_1): Renamed and corrected old
18266         arc_rewrite_small_data function.
18267         (arc_rewrite_small_data): New function.
18268         (small_data_pattern): Don't use pic_offset_table_rtx.
18269         * config/arc/arc.h (CC1_SPEC): Recognize G option.
18270         * config/arc/simdext.md (movmisalignv2hi): Use
18271         prepare_move_operands function.
18272         (mov*): Likewise.
18273         (movmisalign*): Likewise.
18274         * doc/invoke.texi (ARC options): Document -G option.
18276 2017-08-31  Claudiu Zissulescu  <claziss@synopsys.com>
18278         * config/arc/arc-protos.h (compact_sda_memory_operand): Update
18279         prototype.
18280         * config/arc/arc.c (arc_print_operand): Output scalled address for
18281         sdata whenever is possible.
18282         (arc_in_small_data_p): Allow sdata for 64bit datum when double
18283         load/stores are available.
18284         (compact_sda_memory_operand): Check for the alignment required by
18285         code density instructions.
18286         * config/arc/arc.md (movsi_insn): Use newly introduced Us0
18287         constraint.
18288         * config/arc/constraints.md (Usd): Update constraint.
18289         (Us0): New constraint.
18290         (Usc): Update constraint.
18292 2017-08-31  Richard Biener  <rguenther@suse.de>
18294         PR middle-end/82054
18295         * dwarf2out.c (dwarf2out_early_global_decl): Process each
18296         function only once.
18298 2017-08-31  Tamar Christina  <tamar.christina@arm.com>
18300         * config/aarch64/aarch64-builtins.c (aarch64_init_simd_builtins):
18301         Resize type_signature.
18303 2017-08-31  Richard Sandiford  <richard.sandiford@linaro.org>
18304             Alan Hayward  <alan.hayward@arm.com>
18305             David Sherwood  <david.sherwood@arm.com>
18307         * config/aarch64/aarch64.c (aarch64_base_register_rtx_p): Only allow
18308         subregs whose inner modes can be stored in GPRs.
18309         (aarch64_classify_index): Likewise.
18311 2017-08-31  Richard Sandiford  <richard.sandiford@linaro.org>
18312             Alan Hayward  <alan.hayward@arm.com>
18313             David Sherwood  <david.sherwood@arm.com>
18315         * config/aarch64/iterators.md (V_cmp_result): Rename to...
18316         (V_INT_EQUIV): ...this.
18317         (v_cmp_result): Rename to...
18318         (v_int_equiv): ...this.
18319         * config/aarch64/aarch64.md (xorsign<mode>3): Update accordingly.
18320         * config/aarch64/aarch64-simd.md (xorsign<mode>3): Likewise.
18321         (copysign<mode>3): Likewise.
18322         (aarch64_simd_bsl<mode>_internal): Likewise.
18323         (aarch64_simd_bsl<mode>): Likewise.
18324         (vec_cmp<mode><mode>): Likewise.
18325         (vcond<mode><mode>): Likewise.
18326         (vcond<v_cmp_mixed><mode>): Likewise.
18327         (vcondu<mode><v_cmp_mixed>): Likewise.
18328         (aarch64_cm<optab><mode>): Likewise.
18329         (aarch64_cmtst<mode>): Likewise.
18330         (aarch64_fac<optab><mode>): Likewise.
18331         (vec_perm_const<mode>): Likewise.
18332         (vcond_mask_<mode><v_cmp_result>): Rename to...
18333         (vcond_mask_<mode><v_int_equiv>): ...this.
18334         (vec_cmp<mode><v_cmp_result>): Rename to...
18335         (vec_cmp<mode><v_int_equiv>): ...this.
18337 2017-08-31  Richard Sandiford  <richard.sandiford@linaro.org>
18338             Alan Hayward  <alan.hayward@arm.com>
18339             David Sherwood  <david.sherwood@arm.com>
18341         * config/aarch64/aarch64-modes.def: Remove 32-, 48- and 64-byte
18342         vector modes.
18343         * config/aarch64/iterators.md (VRL2, VRL3, VRL4): Delete.
18344         * config/aarch64/aarch64.md (UNSPEC_LD2_DREG, UNSPEC_LD3_DREG)
18345         (UNSPEC_LD4_DREG): New unspecs.
18346         * config/aarch64/aarch64-simd.md (aarch64_ld2<mode>_dreg_le)
18347         (aarch64_ld2<mode>_dreg_be): Replace with...
18348         (aarch64_ld2<mode>_dreg): ...this pattern and use the new DREG
18349         unspec.
18350         (aarch64_ld3<mode>_dreg_le)
18351         (aarch64_ld3<mode>_dreg_be): Replace with...
18352         (aarch64_ld3<mode>_dreg): ...this pattern and use the new DREG
18353         unspec.
18354         (aarch64_ld4<mode>_dreg_le)
18355         (aarch64_ld4<mode>_dreg_be): Replace with...
18356         (aarch64_ld4<mode>_dreg): ...this pattern and use the new DREG
18357         unspec.
18359 2017-08-30  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
18361         PR tree-optimization/81987
18362         * gimple-ssa-strength-reduction.c (insert_initializers): Don't
18363         insert an initializer in a location not dominated by the stride
18364         definition.
18366 2017-08-30  Eric Botcazou  <ebotcazou@adacore.com>
18368         * tree-eh.c (lower_try_finally_switch): Set the location of the finally
18369         on the entire header of the finally block in the fallthru case.
18371 2017-08-30  Eric Botcazou  <ebotcazou@adacore.com>
18373         * varasm.c (decode_addr_const): Deal with INDIRECT_REF <INTEGER_CST>.
18375 2017-08-30  Pat Haugen  <pthaugen@us.ibm.com>
18377         * config/rs6000/rs6000.c (rs6000_emit_prologue_move_from_cr): Rename from
18378         rs6000_emit_move_from_cr and call renamed function.
18379         (rs6000_emit_prologue): Call renamed functions.
18380         * config/rs6000/rs6000.md (prologue_movesi_from_cr): Rename from
18381         movesi_from_cr, remove volatile CRs.
18383 2017-08-30  Jon Beniston  <jon@beniston.com>
18384             Richard Biener  <rguenther@suse.de>
18386         * tree-vect-patterns.c (vect_pattern_recog_1): Use VECTOR_TYPE_P instead
18387         of VECTOR_MODE_P check.
18388         * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Allow single
18389         element vector types.
18391 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18393         * df.h (df_read_modify_subreg_p): Remove in favor of...
18394         * rtl.h (read_modify_subreg_p): ...this new function.  Take a
18395         const_rtx instead of an rtx.
18396         * cprop.c (local_cprop_find_used_regs): Update accordingly.
18397         * df-problems.c (df_word_lr_mark_ref): Likewise.
18398         * ira-lives.c (mark_pseudo_reg_live): Likewise.
18399         (mark_pseudo_reg_dead): Likewise.
18400         (mark_ref_dead): Likewise.
18401         * reginfo.c (init_subregs_of_mode): Likewise.
18402         * sched-deps.c (sched_analyze_1): Likewise.
18403         * df-scan.c (df_def_record_1): Likewise.
18404         (df_uses_record): Likewise.
18405         (df_read_modify_subreg_p): Remove in favor of...
18406         * rtlanal.c (read_modify_subreg_p): ...this new function.  Take a
18407         const_rtx instead of an rtx.
18409 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18410             Alan Hayward  <alan.hayward@arm.com>
18411             David Sherwood  <david.sherwood@arm.com>
18413         * rtl.h (partial_subreg_p): New function.
18414         * caller-save.c (save_call_clobbered_regs): Use it.
18415         * calls.c (expand_call): Likewise.
18416         * combine.c (combinable_i3pat): Likewise.
18417         (simplify_set): Likewise.
18418         (make_extraction): Likewise.
18419         (make_compound_operation_int): Likewise.
18420         (gen_lowpart_or_truncate): Likewise.
18421         (force_to_mode): Likewise.
18422         (make_field_assignment): Likewise.
18423         (reg_truncated_to_mode): Likewise.
18424         (record_truncated_value): Likewise.
18425         (move_deaths): Likewise.
18426         * cse.c (record_jump_cond): Likewise.
18427         (cse_insn): Likewise.
18428         * cselib.c (cselib_lookup_1): Likewise.
18429         * expmed.c (extract_bit_field_using_extv): Likewise.
18430         * function.c (assign_parm_setup_reg): Likewise.
18431         * ifcvt.c (noce_convert_multiple_sets): Likewise.
18432         * ira-build.c (create_insn_allocnos): Likewise.
18433         * lra-coalesce.c (merge_pseudos): Likewise.
18434         * lra-constraints.c (match_reload): Likewise.
18435         (simplify_operand_subreg): Likewise.
18436         (curr_insn_transform): Likewise.
18437         * lra-lives.c (process_bb_lives): Likewise.
18438         * lra.c (new_insn_reg): Likewise.
18439         (lra_substitute_pseudo): Likewise.
18440         * regcprop.c (mode_change_ok): Likewise.
18441         (maybe_mode_change): Likewise.
18442         (copyprop_hardreg_forward_1): Likewise.
18443         * reload.c (push_reload): Likewise.
18444         (find_reloads): Likewise.
18445         (find_reloads_subreg_address): Likewise.
18446         * reload1.c (alter_reg): Likewise.
18447         (eliminate_regs_1): Likewise.
18448         * simplify-rtx.c (simplify_unary_operation_1): Likewise.
18450 2017-08-30  David Edelsohn  <dje.gcc@gmail.com>
18452         * config/rs6000/rs6000.c (rs6000_expand_binop_builtin): Revert
18453         back to if statements, including unpack.
18455 2017-08-30  Martin Liska  <mliska@suse.cz>
18457         PR inline-asm/82001
18458         * ipa-icf-gimple.c (func_checker::compare_tree_list_operand):
18459         Rename to ...
18460         (func_checker::compare_asm_inputs_outputs): ... this function.
18461         (func_checker::compare_gimple_asm): Use the function to compare
18462         also ASM constrains.
18463         * ipa-icf-gimple.h: Rename the function.
18465 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18466             Alan Hayward  <alan.hayward@arm.com>
18467             David Sherwood  <david.sherwood@arm.com>
18469         * coretypes.h (complex_mode): New type.
18470         * gdbhooks.py (build_pretty_printer): Handle it.
18471         * machmode.h (complex_mode): New class.
18472         (complex_mode::includes_p): New function.
18473         (is_complex_int_mode): Likewise.
18474         (is_complex_float_mode): Likewise.
18475         * genmodes.c (get_mode_class): Handle complex mode classes.
18476         * function.c (expand_function_end): Use is_complex_int_mode.
18478 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18479             Alan Hayward  <alan.hayward@arm.com>
18480             David Sherwood  <david.sherwood@arm.com>
18482         * coretypes.h (scalar_mode_pod): New typedef.
18483         * gdbhooks.py (build_pretty_printer): Handle it.
18484         * machmode.h (gt_ggc_mx, gt_pch_nx): New functions.
18485         * fixed-value.h (fixed_value::mode): Change type to scalar_mode_pod.
18486         * fold-const.c (fold_convert_const_int_from_fixed): Use scalar_mode.
18487         * tree-streamer-in.c (unpack_ts_fixed_cst_value_fields): Use
18488         as_a <scalar_mode>.
18490 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18491             Alan Hayward  <alan.hayward@arm.com>
18492             David Sherwood  <david.sherwood@arm.com>
18494         * machmode.h (mode_for_vector): Take a scalar_mode instead
18495         of a machine_mode.
18496         * stor-layout.c (mode_for_vector): Likewise.
18497         * explow.c (promote_mode): Use as_a <scalar_mode>.
18498         * sdbout.c (sdbout_parms): Use is_a <scalar_mode>.
18500 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18501             Alan Hayward  <alan.hayward@arm.com>
18502             David Sherwood  <david.sherwood@arm.com>
18504         * target.def (preferred_simd_mode): Take a scalar_mode
18505         instead of a machine_mode.
18506         * targhooks.h (default_preferred_simd_mode): Likewise.
18507         * targhooks.c (default_preferred_simd_mode): Likewise.
18508         * config/arc/arc.c (arc_preferred_simd_mode): Likewise.
18509         * config/arm/arm.c (arm_preferred_simd_mode): Likewise.
18510         * config/c6x/c6x.c (c6x_preferred_simd_mode): Likewise.
18511         * config/epiphany/epiphany.c (epiphany_preferred_simd_mode): Likewise.
18512         * config/i386/i386.c (ix86_preferred_simd_mode): Likewise.
18513         * config/mips/mips.c (mips_preferred_simd_mode): Likewise.
18514         * config/nvptx/nvptx.c (nvptx_preferred_simd_mode): Likewise.
18515         * config/powerpcspe/powerpcspe.c (rs6000_preferred_simd_mode):
18516         Likewise.
18517         * config/rs6000/rs6000.c (rs6000_preferred_simd_mode): Likewise.
18518         * config/s390/s390.c (s390_preferred_simd_mode): Likewise.
18519         * config/sparc/sparc.c (sparc_preferred_simd_mode): Likewise.
18520         * config/aarch64/aarch64.c (aarch64_preferred_simd_mode): Likewise.
18521         (aarch64_simd_scalar_immediate_valid_for_move): Update accordingly.
18522         * doc/tm.texi: Regenerate.
18523         * optabs-query.c (can_vec_mask_load_store_p): Return false for
18524         non-scalar modes.
18526 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18527             Alan Hayward  <alan.hayward@arm.com>
18528             David Sherwood  <david.sherwood@arm.com>
18530         * target.def (scalar_mode_supported_p): Take a scalar_mode
18531         instead of a machine_mode.
18532         * targhooks.h (default_scalar_mode_supported_p): Likewise.
18533         * targhooks.c (default_scalar_mode_supported_p): Likewise.
18534         * config/aarch64/aarch64.c (aarch64_scalar_mode_supported_p): Likewise.
18535         * config/alpha/alpha.c (alpha_scalar_mode_supported_p): Likewise.
18536         * config/arm/arm.c (arm_scalar_mode_supported_p): Likewise.
18537         * config/avr/avr.c (avr_scalar_mode_supported_p): Likewise.
18538         * config/c6x/c6x.c (c6x_scalar_mode_supported_p): Likewise.
18539         * config/i386/i386.c (ix86_scalar_mode_supported_p): Likewise.
18540         * config/ia64/ia64.c (ia64_scalar_mode_supported_p): Likewise.
18541         * config/mips/mips.c (mips_scalar_mode_supported_p): Likewise.
18542         * config/msp430/msp430.c (msp430_scalar_mode_supported_p): Likewise.
18543         * config/pa/pa.c (pa_scalar_mode_supported_p): Likewise.
18544         * config/pdp11/pdp11.c (pdp11_scalar_mode_supported_p): Likewise.
18545         * config/powerpcspe/powerpcspe.c (rs6000_scalar_mode_supported_p):
18546         Likewise.
18547         * config/rs6000/rs6000.c (rs6000_scalar_mode_supported_p): Likewise.
18548         * config/s390/s390.c (s390_scalar_mode_supported_p): Likewise.
18549         * config/spu/spu.c (spu_scalar_mode_supported_p): Likewise.
18550         * config/tilegx/tilegx.c (tilegx_scalar_mode_supported_p): Likewise.
18551         * config/tilepro/tilepro.c (tilepro_scalar_mode_supported_p):
18552         Likewise.
18553         * doc/tm.texi: Regenerate.
18555 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18556             Alan Hayward  <alan.hayward@arm.com>
18557             David Sherwood  <david.sherwood@arm.com>
18559         * coretypes.h (opt_scalar_mode): New typedef.
18560         * gdbhooks.py (build_pretty_printers): Handle it.
18561         * machmode.h (mode_iterator::get_2xwider): Add overload for
18562         opt_mode<T>.
18563         * emit-rtl.c (init_emit_once): Use opt_scalar_mode when iterating
18564         over scalar modes.
18565         * expr.c (convert_mode_scalar): Likewise.
18566         * omp-low.c (omp_clause_aligned_alignment): Likewise.
18567         * optabs.c (expand_float): Likewise.
18568         (expand_fix): Likewise.
18569         * tree-vect-stmts.c (vectorizable_conversion): Likewise.
18571 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18572             Alan Hayward  <alan.hayward@arm.com>
18573             David Sherwood  <david.sherwood@arm.com>
18575         * optabs.c (expand_float): Explicitly check for scalars before
18576         using a branching expansion.
18577         (expand_fix): Likewise.
18579 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18580             Alan Hayward  <alan.hayward@arm.com>
18581             David Sherwood  <david.sherwood@arm.com>
18583         * expr.c (convert_mode): Split scalar handling out into...
18584         (convert_mode_scalar): ...this new function.  Treat the modes
18585         as scalar_modes.
18587 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18588             Alan Hayward  <alan.hayward@arm.com>
18589             David Sherwood  <david.sherwood@arm.com>
18591         * omp-expand.c (expand_omp_atomic): Use is_int_mode, is_float_mode
18592         and scalar_mode.
18593         * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Likewise.
18595 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18596             Alan Hayward  <alan.hayward@arm.com>
18597             David Sherwood  <david.sherwood@arm.com>
18599         * fixed-value.h (fixed_from_double_int): Take a scalar_mode
18600         rather than a machine_mode.
18601         (fixed_from_string): Likewise.
18602         (fixed_convert): Likewise.
18603         (fixed_convert_from_int): Likewise.
18604         (fixed_convert_from_real): Likewise.
18605         (real_convert_from_fixed): Likewise.
18606         * fixed-value.c (fixed_from_double_int): Likewise.
18607         (fixed_from_string): Likewise.
18608         (fixed_convert): Likewise.
18609         (fixed_convert_from_int): Likewise.
18610         (fixed_convert_from_real): Likewise.
18611         (real_convert_from_fixed): Likewise.
18612         * config/avr/avr.c (avr_out_round): Use as_a <scalar_mode>.
18614 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18615             Alan Hayward  <alan.hayward@arm.com>
18616             David Sherwood  <david.sherwood@arm.com>
18618         * emit-rtl.c (immed_double_const): Use is_a <scalar_mode> instead
18619         of separate mode class checks.  Do not allow vector modes here.
18620         (immed_wide_int_const): Use as_a <scalar_mode>.
18621         * explow.c (trunc_int_for_mode): Likewise.
18622         * rtl.h (wi::int_traits<rtx_mode_t>::get_precision): Likewise.
18623         (wi::shwi): Likewise.
18624         (wi::min_value): Likewise.
18625         (wi::max_value): Likewise.
18626         * dwarf2out.c (loc_descriptor): Likewise.
18627         * simplify-rtx.c (simplify_immed_subreg): Fix rtx_mode_t argument
18628         for CONST_WIDE_INT.
18630 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18631             Alan Hayward  <alan.hayward@arm.com>
18632             David Sherwood  <david.sherwood@arm.com>
18634         * tree.h (SCALAR_TYPE_MODE): New macro.
18635         * expr.c (expand_expr_addr_expr_1): Use it.
18636         (expand_expr_real_2): Likewise.
18637         * fold-const.c (fold_convert_const_fixed_from_fixed): Likeise.
18638         (fold_convert_const_fixed_from_int): Likewise.
18639         (fold_convert_const_fixed_from_real): Likewise.
18640         (native_encode_fixed): Likewise
18641         (native_encode_complex): Likewise
18642         (native_encode_vector): Likewise.
18643         (native_interpret_fixed): Likewise.
18644         (native_interpret_real): Likewise.
18645         (native_interpret_complex): Likewise.
18646         (native_interpret_vector): Likewise.
18647         * omp-simd-clone.c (simd_clone_adjust_return_type): Likewise.
18648         (simd_clone_adjust_argument_types): Likewise.
18649         (simd_clone_init_simd_arrays): Likewise.
18650         (simd_clone_adjust): Likewise.
18651         * stor-layout.c (layout_type): Likewise.
18652         * tree.c (build_minus_one_cst): Likewise.
18653         * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
18654         * tree-inline.c (estimate_move_cost): Likewise.
18655         * tree-ssa-math-opts.c (convert_plusminus_to_widen): Likewise.
18656         * tree-vect-loop.c (vect_create_epilog_for_reduction): Likewise.
18657         (vectorizable_reduction): Likewise.
18658         * tree-vect-patterns.c (vect_recog_widen_mult_pattern): Likewise.
18659         (vect_recog_mixed_size_cond_pattern): Likewise.
18660         (check_bool_pattern): Likewise.
18661         (adjust_bool_pattern): Likewise.
18662         (search_type_for_mask_1): Likewise.
18663         * tree-vect-slp.c (vect_schedule_slp_instance): Likewise.
18664         * tree-vect-stmts.c (vectorizable_conversion): Likewise.
18665         (vectorizable_load): Likewise.
18666         (vectorizable_store): Likewise.
18667         * ubsan.c (ubsan_encode_value): Likewise.
18668         * varasm.c (output_constant): Likewise.
18670 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18671             Alan Hayward  <alan.hayward@arm.com>
18672             David Sherwood  <david.sherwood@arm.com>
18674         * coretypes.h (scalar_mode): New class.
18675         * machmode.h (scalar_mode): Likewise.
18676         (scalar_mode::includes_p): New function.
18677         (mode_to_inner): Return a scalar_mode rather than a machine_mode.
18678         * gdbhooks.py (build_pretty_printers): Handle scalar_mode.
18679         * genmodes.c (get_mode_class): Handle remaining scalar modes.
18680         * cfgexpand.c (expand_debug_expr): Use scalar_mode.
18681         * expmed.c (store_bit_field_1): Likewise.
18682         (extract_bit_field_1): Likewise.
18683         * expr.c (write_complex_part): Likewise.
18684         (read_complex_part): Likewise.
18685         (emit_move_complex_push): Likewise.
18686         (expand_expr_real_2): Likewise.
18687         * function.c (assign_parm_setup_reg): Likewise.
18688         (assign_parms_unsplit_complex): Likewise.
18689         * optabs.c (expand_binop): Likewise.
18690         * rtlanal.c (subreg_get_info): Likewise.
18691         * simplify-rtx.c (simplify_immed_subreg): Likewise.
18692         * varasm.c (output_constant_pool_2): Likewise.
18694 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18695             Alan Hayward  <alan.hayward@arm.com>
18696             David Sherwood  <david.sherwood@arm.com>
18698         * expmed.c (extract_high_half): Use scalar_int_mode and remove
18699         assertion.
18700         (expmed_mult_highpart_optab): Likewise.
18701         (expmed_mult_highpart): Likewise.
18703 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18704             Alan Hayward  <alan.hayward@arm.com>
18705             David Sherwood  <david.sherwood@arm.com>
18707         * builtins.h (builtin_strncpy_read_str): Take a scalar_int_mode
18708         instead of a machine_mode.
18709         (builtin_memset_read_str): Likewise.
18710         * builtins.c (c_readstr): Likewise.
18711         (builtin_memcpy_read_str): Likewise.
18712         (builtin_strncpy_read_str): Likewise.
18713         (builtin_memset_read_str): Likewise.
18714         (builtin_memset_gen_str): Likewise.
18715         (expand_builtin_signbit): Use scalar_int_mode for local variables.
18716         * cfgexpand.c (convert_debug_memory_address): Take a scalar_int_mode
18717         instead of a machine_mode.
18718         * combine.c (simplify_if_then_else): Use scalar_int_mode for local
18719         variables.
18720         (make_extraction): Likewise.
18721         (try_widen_shift_mode): Take and return scalar_int_modes instead
18722         of machine_modes.
18723         * config/aarch64/aarch64.c (aarch64_libgcc_cmp_return_mode): Return
18724         a scalar_int_mode instead of a machine_mode.
18725         * config/avr/avr.c (avr_addr_space_address_mode): Likewise.
18726         (avr_addr_space_pointer_mode): Likewise.
18727         * config/cr16/cr16.c (cr16_unwind_word_mode): Likewise.
18728         * config/msp430/msp430.c (msp430_addr_space_pointer_mode): Likewise.
18729         (msp430_unwind_word_mode): Likewise.
18730         * config/spu/spu.c (spu_unwind_word_mode): Likewise.
18731         (spu_addr_space_pointer_mode): Likewise.
18732         (spu_addr_space_address_mode): Likewise.
18733         (spu_libgcc_cmp_return_mode): Likewise.
18734         (spu_libgcc_shift_count_mode): Likewise.
18735         * config/rl78/rl78.c (rl78_addr_space_address_mode): Likewise.
18736         (rl78_addr_space_pointer_mode): Likewise.
18737         (fl78_unwind_word_mode): Likewise.
18738         (rl78_valid_pointer_mode): Take a scalar_int_mode instead of a
18739         machine_mode.
18740         * config/alpha/alpha.c (vms_valid_pointer_mode): Likewise.
18741         * config/ia64/ia64.c (ia64_vms_valid_pointer_mode): Likewise.
18742         * config/mips/mips.c (mips_mode_rep_extended): Likewise.
18743         (mips_valid_pointer_mode): Likewise.
18744         * config/tilegx/tilegx.c (tilegx_mode_rep_extended): Likewise.
18745         * config/ft32/ft32.c (ft32_valid_pointer_mode): Likewise.
18746         (ft32_addr_space_pointer_mode): Return a scalar_int_mode instead
18747         of a machine_mode.
18748         (ft32_addr_space_address_mode): Likewise.
18749         * config/m32c/m32c.c (m32c_valid_pointer_mode): Take a
18750         scalar_int_mode instead of a machine_mode.
18751         (m32c_addr_space_pointer_mode): Return a scalar_int_mode instead
18752         of a machine_mode.
18753         (m32c_addr_space_address_mode): Likewise.
18754         * config/powerpcspe/powerpcspe.c (rs6000_abi_word_mode): Likewise.
18755         (rs6000_eh_return_filter_mode): Likewise.
18756         * config/rs6000/rs6000.c (rs6000_abi_word_mode): Likewise.
18757         (rs6000_eh_return_filter_mode): Likewise.
18758         * config/s390/s390.c (s390_libgcc_cmp_return_mode): Likewise.
18759         (s390_libgcc_shift_count_mode): Likewise.
18760         (s390_unwind_word_mode): Likewise.
18761         (s390_valid_pointer_mode): Take a scalar_int_mode rather than a
18762         machine_mode.
18763         * target.def (mode_rep_extended): Likewise.
18764         (valid_pointer_mode): Likewise.
18765         (addr_space.valid_pointer_mode): Likewise.
18766         (eh_return_filter_mode): Return a scalar_int_mode rather than
18767         a machine_mode.
18768         (libgcc_cmp_return_mode): Likewise.
18769         (libgcc_shift_count_mode): Likewise.
18770         (unwind_word_mode): Likewise.
18771         (addr_space.pointer_mode): Likewise.
18772         (addr_space.address_mode): Likewise.
18773         * doc/tm.texi: Regenerate.
18774         * dojump.c (prefer_and_bit_test): Take a scalar_int_mode rather than
18775         a machine_mode.
18776         (do_jump): Use scalar_int_mode for local variables.
18777         * dwarf2cfi.c (init_return_column_size): Take a scalar_int_mode
18778         rather than a machine_mode.
18779         * dwarf2out.c (convert_descriptor_to_mode): Likewise.
18780         (scompare_loc_descriptor_wide): Likewise.
18781         (scompare_loc_descriptor_narrow): Likewise.
18782         * emit-rtl.c (adjust_address_1): Use scalar_int_mode for local
18783         variables.
18784         * except.c (sjlj_emit_dispatch_table): Likewise.
18785         (expand_builtin_eh_copy_values): Likewise.
18786         * explow.c (convert_memory_address_addr_space_1): Likewise.
18787         Take a scalar_int_mode rather than a machine_mode.
18788         (convert_memory_address_addr_space): Take a scalar_int_mode rather
18789         than a machine_mode.
18790         (memory_address_addr_space): Use scalar_int_mode for local variables.
18791         * expmed.h (expand_mult_highpart_adjust): Take a scalar_int_mode
18792         rather than a machine_mode.
18793         * expmed.c (mask_rtx): Likewise.
18794         (init_expmed_one_conv): Likewise.
18795         (expand_mult_highpart_adjust): Likewise.
18796         (extract_high_half): Likewise.
18797         (expmed_mult_highpart_optab): Likewise.
18798         (expmed_mult_highpart): Likewise.
18799         (expand_smod_pow2): Likewise.
18800         (expand_sdiv_pow2): Likewise.
18801         (emit_store_flag_int): Likewise.
18802         (adjust_bit_field_mem_for_reg): Use scalar_int_mode for local
18803         variables.
18804         (extract_low_bits): Likewise.
18805         * expr.h (by_pieces_constfn): Take a scalar_int_mode rather than
18806         a machine_mode.
18807         * expr.c (pieces_addr::adjust):  Likewise.
18808         (can_store_by_pieces): Likewise.
18809         (store_by_pieces): Likewise.
18810         (clear_by_pieces_1): Likewise.
18811         (expand_expr_addr_expr_1): Likewise.
18812         (expand_expr_addr_expr): Use scalar_int_mode for local variables.
18813         (expand_expr_real_1): Likewise.
18814         (try_casesi): Likewise.
18815         * final.c (shorten_branches): Likewise.
18816         * fold-const.c (fold_convert_const_int_from_fixed): Change the
18817         type of "mode" to machine_mode.
18818         * internal-fn.c (expand_arith_overflow_result_store): Take a
18819         scalar_int_mode rather than a machine_mode.
18820         (expand_mul_overflow): Use scalar_int_mode for local variables.
18821         * loop-doloop.c (doloop_modify): Likewise.
18822         (doloop_optimize): Likewise.
18823         * optabs.c (expand_subword_shift): Take a scalar_int_mode rather
18824         than a machine_mode.
18825         (expand_doubleword_shift_condmove): Likewise.
18826         (expand_doubleword_shift): Likewise.
18827         (expand_doubleword_clz): Likewise.
18828         (expand_doubleword_popcount): Likewise.
18829         (expand_doubleword_parity): Likewise.
18830         (expand_absneg_bit): Use scalar_int_mode for local variables.
18831         (prepare_float_lib_cmp): Likewise.
18832         * rtl.h (convert_memory_address_addr_space_1): Take a scalar_int_mode
18833         rather than a machine_mode.
18834         (convert_memory_address_addr_space): Likewise.
18835         (get_mode_bounds): Likewise.
18836         (get_address_mode): Return a scalar_int_mode rather than a
18837         machine_mode.
18838         * rtlanal.c (get_address_mode): Likewise.
18839         * stor-layout.c (get_mode_bounds): Take a scalar_int_mode rather
18840         than a machine_mode.
18841         * targhooks.c (default_mode_rep_extended): Likewise.
18842         (default_valid_pointer_mode): Likewise.
18843         (default_addr_space_valid_pointer_mode): Likewise.
18844         (default_eh_return_filter_mode): Return a scalar_int_mode rather
18845         than a machine_mode.
18846         (default_libgcc_cmp_return_mode): Likewise.
18847         (default_libgcc_shift_count_mode): Likewise.
18848         (default_unwind_word_mode): Likewise.
18849         (default_addr_space_pointer_mode): Likewise.
18850         (default_addr_space_address_mode): Likewise.
18851         * targhooks.h (default_eh_return_filter_mode): Likewise.
18852         (default_libgcc_cmp_return_mode): Likewise.
18853         (default_libgcc_shift_count_mode): Likewise.
18854         (default_unwind_word_mode): Likewise.
18855         (default_addr_space_pointer_mode): Likewise.
18856         (default_addr_space_address_mode): Likewise.
18857         (default_mode_rep_extended): Take a scalar_int_mode rather than
18858         a machine_mode.
18859         (default_valid_pointer_mode): Likewise.
18860         (default_addr_space_valid_pointer_mode): Likewise.
18861         * tree-ssa-address.c (addr_for_mem_ref): Use scalar_int_mode for
18862         local variables.
18863         * tree-ssa-loop-ivopts.c (get_shiftadd_cost): Take a scalar_int_mode
18864         rather than a machine_mode.
18865         * tree-switch-conversion.c (array_value_type): Use scalar_int_mode
18866         for local variables.
18867         * tree-vrp.c (simplify_float_conversion_using_ranges): Likewise.
18868         * var-tracking.c (use_narrower_mode): Take a scalar_int_mode rather
18869         than a machine_mode.
18871 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18872             Alan Hayward  <alan.hayward@arm.com>
18873             David Sherwood  <david.sherwood@arm.com>
18875         * dojump.c (do_jump_by_parts_greater_rtx): Change the type of
18876         the mode argument to scalar_int_mode.
18877         (do_jump_by_parts_zero_rtx): Likewise.
18878         (do_jump_by_parts_equality_rtx): Likewise.
18879         (do_jump_by_parts_greater): Take a mode argument.
18880         (do_jump_by_parts_equality): Likewise.
18881         (do_jump_1): Update calls accordingly.
18883 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18884             Alan Hayward  <alan.hayward@arm.com>
18885             David Sherwood  <david.sherwood@arm.com>
18887         * is-a.h (safe_dyn_cast): New function.
18888         * rtl.h (rtx_jump_table_data::get_data_mode): New function.
18889         (jump_table_for_label): Likewise.
18890         * final.c (final_addr_vec_align): Take an rtx_jump_table_data *
18891         instead of an rtx_insn *.
18892         (shorten_branches): Use dyn_cast instead of LABEL_P and
18893         JUMP_TABLE_DATA_P.  Use jump_table_for_label and
18894         rtx_jump_table_data::get_data_mode.
18895         (final_scan_insn): Likewise.
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         * combine.c (try_combine): Use is_a <scalar_int_mode> when
18902         trying to combine a full-register integer set with a subreg
18903         integer set.
18905 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18906             Alan Hayward  <alan.hayward@arm.com>
18907             David Sherwood  <david.sherwood@arm.com>
18909         * expr.c (expand_expr_addr_expr): Add a new_tmode local variable
18910         that is always either address_mode or pointer_mode.
18912 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18913             Alan Hayward  <alan.hayward@arm.com>
18914             David Sherwood  <david.sherwood@arm.com>
18916         * expr.c (expand_expr_real_2): Use word_mode instead of innermode
18917         when the two are known to be equal.
18919 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18920             Alan Hayward  <alan.hayward@arm.com>
18921             David Sherwood  <david.sherwood@arm.com>
18923         * simplify-rtx.c (simplify_const_unary_operation): Use
18924         is_a <scalar_int_mode> instead of checking for a nonzero
18925         precision.  Forcibly convert op_mode to a scalar_int_mode
18926         in that case.  More clearly differentiate the operand and
18927         result modes and use the former when deciding what the value
18928         of a count-bits operation should be.  Use is_int_mode instead
18929         of checking for a MODE_INT.  Remove redundant check for whether
18930         this mode has a zero precision.
18932 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18933             Alan Hayward  <alan.hayward@arm.com>
18934             David Sherwood  <david.sherwood@arm.com>
18936         * optabs.c (widen_leading): Change the type of the mode argument
18937         to scalar_int_mode.  Use opt_scalar_int_mode for the mode iterator.
18938         (widen_bswap): Likewise.
18939         (expand_parity): Likewise.
18940         (expand_ctz): Change the type of the mode argument to scalar_int_mode.
18941         (expand_ffs): Likewise.
18942         (epand_unop): Check for scalar integer modes before calling the
18943         above routines.
18945 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18946             Alan Hayward  <alan.hayward@arm.com>
18947             David Sherwood  <david.sherwood@arm.com>
18949         * expr.c (const_scalar_mask_from_tree): Add a mode argument.
18950         Expand commentary.
18951         (expand_expr_real_1): Update call accordingly.
18953 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18954             Alan Hayward  <alan.hayward@arm.com>
18955             David Sherwood  <david.sherwood@arm.com>
18957         * expmed.c (store_bit_field_using_insv): Add op0_mode and
18958         value_mode arguments.  Use scalar_int_mode internally.
18959         (store_bit_field_1): Rename the new integer mode from imode
18960         to op0_mode and use it instead of GET_MODE (op0).  Update calls
18961         to store_split_bit_field, store_bit_field_using_insv and
18962         store_fixed_bit_field.
18963         (store_fixed_bit_field): Add op0_mode and value_mode arguments.
18964         Use scalar_int_mode internally.  Use a bit count rather than a mode
18965         when calculating the largest bit size for get_best_mode.
18966         Update calls to store_split_bit_field and store_fixed_bit_field_1.
18967         (store_fixed_bit_field_1): Add mode and value_mode arguments.
18968         Remove assertion that OP0 has a scalar integer mode.
18969         (store_split_bit_field): Add op0_mode and value_mode arguments.
18970         Update calls to extract_fixed_bit_field.
18971         (extract_bit_field_using_extv): Add an op0_mode argument.
18972         Use scalar_int_mode internally.
18973         (extract_bit_field_1): Rename the new integer mode from imode to
18974         op0_mode and use it instead of GET_MODE (op0).  Update calls to
18975         extract_split_bit_field, extract_bit_field_using_extv and
18976         extract_fixed_bit_field.
18977         (extract_fixed_bit_field): Add an op0_mode argument.  Update calls
18978         to extract_split_bit_field and extract_fixed_bit_field_1.
18979         (extract_fixed_bit_field_1): Add a mode argument.  Remove assertion
18980         that OP0 has a scalar integer mode.  Use as_a <scalar_int_mode>
18981         on the target mode.
18982         (extract_split_bit_field): Add an op0_mode argument.  Update call
18983         to extract_fixed_bit_field.
18985 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18986             Alan Hayward  <alan.hayward@arm.com>
18987             David Sherwood  <david.sherwood@arm.com>
18989         * cse.c (cse_insn): Use opt_scalar_int_mode for the mode iterator.
18990         * explow.c (hard_function_value): Likewise.
18991         * expmed.c (extract_fixed_bit_field_1): Likewise.  Move the
18992         convert_to_mode call outside the loop.
18993         * expr.c (alignment_for_piecewise_move): Use opt_scalar_int_mode
18994         for the mode iterator.  Require the mode specified by max_pieces
18995         to exist.
18996         (emit_block_move_via_movmem): Use opt_scalar_int_mode for the
18997         mode iterator.
18998         (copy_blkmode_to_reg): Likewise.
18999         (set_storage_via_setmem): Likewise.
19000         * optabs.c (prepare_cmp_insn): Likewise.
19001         * rtlanal.c (init_num_sign_bit_copies_in_rep): Likewise.
19002         * stor-layout.c (finish_bitfield_representative): Likewise.
19004 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19005             Alan Hayward  <alan.hayward@arm.com>
19006             David Sherwood  <david.sherwood@arm.com>
19008         * rtl.h (subreg_unpromoted_mode, subreg_promoted_mode): New functions.
19009         * expr.c (convert_move): Use them.
19010         (convert_modes): Likewise.
19011         (store_expr_with_bounds): Likewise.
19013 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19014             Alan Hayward  <alan.hayward@arm.com>
19015             David Sherwood  <david.sherwood@arm.com>
19017         * rtl.h (rtl_hooks::reg_nonzero_bits): Add a scalar_int_mode
19018         parameter for the mode of "x".  Remove the "known_x", "known_mode"
19019         and "known_ret" arguments.  Change the type of the mode argument
19020         to scalar_int_mode.
19021         (rtl_hooks:reg_num_sign_bit_copies): Likewise.
19022         * combine.c (reg_nonzero_bits_for_combine): Update accordingly.
19023         (reg_num_sign_bit_copies_for_combine): Likewise.
19024         * rtlanal.c (nonzero_bits1): Likewise.
19025         (num_sign_bit_copies1): Likewise.
19026         * rtlhooks-def.h (reg_nonzero_bits_general): Likewise.
19027         (reg_num_sign_bit_copies_general): Likewise.
19028         * rtlhooks.c (reg_num_sign_bit_copies_general): Likewise.
19029         (reg_nonzero_bits_general): Likewise.
19031 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19032             Alan Hayward  <alan.hayward@arm.com>
19033             David Sherwood  <david.sherwood@arm.com>
19035         * rtlanal.c (num_sign_bit_copies): Handle VOIDmode here rather
19036         than in subroutines.  Return 1 for non-integer modes.
19037         (cached_num_sign_bit_copies): Change the type of the mode parameter
19038         to scalar_int_mode.
19039         (num_sign_bit_copies1): Likewise.  Remove early exit for other mode
19040         classes.  Handle CONST_INT_P first and then check whether X also
19041         has a scalar integer mode.  Check the same thing for inner registers
19042         of a SUBREG and for values that are being extended or truncated.
19044 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19045             Alan Hayward  <alan.hayward@arm.com>
19046             David Sherwood  <david.sherwood@arm.com>
19048         * rtlanal.c (nonzero_bits): Handle VOIDmode here rather than
19049         in subroutines.  Return the mode mask for non-integer modes.
19050         (cached_nonzero_bits): Change the type of the mode parameter
19051         to scalar_int_mode.
19052         (nonzero_bits1): Likewise.  Remove early exit for other mode
19053         classes.  Handle CONST_INT_P first and then check whether X
19054         also has a scalar integer mode.
19056 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19057             Alan Hayward  <alan.hayward@arm.com>
19058             David Sherwood  <david.sherwood@arm.com>
19060         * expr.c (widest_int_mode_for_size): Make the comment match the code.
19061         Return a scalar_int_mode and assert that the size is greater than
19062         one byte.
19063         (by_pieces_ninsns): Update accordingly and remove VOIDmode handling.
19064         (op_by_pieces_d::op_by_pieces_d): Likewise.
19065         (op_by_pieces_d::run): Likewise.
19066         (can_store_by_pieces): Likewise.
19068 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19069             Alan Hayward  <alan.hayward@arm.com>
19070             David Sherwood  <david.sherwood@arm.com>
19072         * combine.c (extract_left_shift): Add a mode argument and update
19073         recursive calls.
19074         (make_compound_operation_int): Change the type of the mode parameter
19075         to scalar_int_mode and update the call to extract_left_shift.
19077 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19078             Alan Hayward  <alan.hayward@arm.com>
19079             David Sherwood  <david.sherwood@arm.com>
19081         * combine.c (simplify_and_const_int): Change the type of the mode
19082         parameter to scalar_int_mode.
19083         (simplify_and_const_int_1): Likewise.  Update recursive call.
19085 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19086             Alan Hayward  <alan.hayward@arm.com>
19087             David Sherwood  <david.sherwood@arm.com>
19089         * combine.c (simplify_compare_const): Check that the mode is a
19090         scalar_int_mode (rather than VOIDmode) before testing its
19091         precision.
19092         (simplify_comparison): Move COMPARISON_P handling out of the
19093         loop and restrict the latter part of the loop to scalar_int_modes.
19094         Check is_a <scalar_int_mode> before calling HWI_COMPUTABLE_MODE_P
19095         and when considering SUBREG_REGs.  Use is_int_mode instead of
19096         checking GET_MODE_CLASS against MODE_INT.
19098 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19099             Alan Hayward  <alan.hayward@arm.com>
19100             David Sherwood  <david.sherwood@arm.com>
19102         * combine.c (try_widen_shift_mode): Move check for equal modes to...
19103         (simplify_shift_const_1): ...here.  Use scalar_int_mode for
19104         shift_unit_mode and for modes involved in scalar shifts.
19106 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19107             Alan Hayward  <alan.hayward@arm.com>
19108             David Sherwood  <david.sherwood@arm.com>
19110         * combine.c (force_int_to_mode): New function, split out from...
19111         (force_to_mode): ...here.  Keep xmode up-to-date and use it
19112         instead of GET_MODE (x).
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         * optabs-query.h (extraction_insn::struct_mode): Change type to
19119         opt_scalar_int_mode and update comment.
19120         (extraction_insn::field_mode): Change type to scalar_int_mode.
19121         (extraction_insn::pos_mode): Likewise.
19122         * combine.c (make_extraction): Update accordingly.
19123         * optabs-query.c (get_traditional_extraction_insn): Likewise.
19124         (get_optab_extraction_insn): Likewise.
19125         * recog.c (simplify_while_replacing): Likewise.
19126         * expmed.c (narrow_bit_field_mem): Change the type of the mode
19127         parameter to opt_scalar_int_mode.
19129 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19130             Alan Hayward  <alan.hayward@arm.com>
19131             David Sherwood  <david.sherwood@arm.com>
19133         * machmode.h (bit_field_mode_iterator::next_mode): Take a pointer
19134         to a scalar_int_mode instead of a machine_mode.
19135         (bit_field_mode_iterator::m_mode): Change type to opt_scalar_int_mode.
19136         (get_best_mode): Return a boolean and use a pointer argument to store
19137         the selected mode.  Replace the limit mode parameter with a bit limit.
19138         * expmed.c (adjust_bit_field_mem_for_reg): Use scalar_int_mode
19139         for the values returned by bit_field_mode_iterator::next_mode.
19140         (store_bit_field): Update call to get_best_mode.
19141         (store_fixed_bit_field): Likewise.
19142         (extract_fixed_bit_field): Likewise.
19143         * expr.c (optimize_bitfield_assignment_op): Likewise.
19144         * fold-const.c (optimize_bit_field_compare): Likewise.
19145         (fold_truth_andor_1): Likewise.
19146         * stor-layout.c (bit_field_mode_iterator::next_mode): As above.
19147         Update for new type of m_mode.
19148         (get_best_mode): As above.
19150 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19151             Alan Hayward  <alan.hayward@arm.com>
19152             David Sherwood  <david.sherwood@arm.com>
19154         * expmed.c (strict_volatile_bitfield_p): Change the type of fieldmode
19155         to scalar_int_mode.  Remove check for SCALAR_INT_MODE_P.
19156         (store_bit_field): Check is_a <scalar_int_mode> before calling
19157         strict_volatile_bitfield_p.
19158         (extract_bit_field): Likewise.
19160 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19161             Alan Hayward  <alan.hayward@arm.com>
19162             David Sherwood  <david.sherwood@arm.com>
19164         * target.def (cstore_mode): Return a scalar_int_mode.
19165         * doc/tm.texi: Regenerate.
19166         * config/sparc/sparc.c (sparc_cstore_mode): Return a scalar_int_mode.
19167         * targhooks.h (default_cstore_mode): Likewise.
19168         * targhooks.c (default_cstore_mode): Likewise, using a forced
19169         conversion.
19170         * expmed.c (emit_cstore): Expect the target of the cstore to be
19171         a scalar_int_mode.
19173 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19174             Alan Hayward  <alan.hayward@arm.com>
19175             David Sherwood  <david.sherwood@arm.com>
19177         * cfgloop.h (rtx_iv): Change type of extend_mode and mode to
19178         scalar_int_mode.
19179         (niter_desc): Likewise mode.
19180         (iv_analyze): Add a mode parameter.
19181         (biv_p): Likewise.
19182         (iv_analyze_expr): Pass the mode paraeter before the rtx it describes
19183         and change its type to scalar_int_mode.
19184         * loop-iv.c: Update commentary at head of file.
19185         (iv_constant): Pass the mode paraeter before the rtx it describes
19186         and change its type to scalar_int_mode.  Remove VOIDmode handling.
19187         (iv_subreg): Change the type of the mode parameter to scalar_int_mode.
19188         (iv_extend): Likewise.
19189         (shorten_into_mode): Likewise.
19190         (iv_add): Use scalar_int_mode.
19191         (iv_mult): Likewise.
19192         (iv_shift): Likewise.
19193         (canonicalize_iv_subregs): Likewise.
19194         (get_biv_step_1): Pass the outer_mode parameter before the rtx
19195         it describes and change its mode to scalar_int_mode.   Also change
19196         the type of the returned inner_mode to scalar_int_mode.
19197         (get_biv_step): Likewise, turning outer_mode from a pointer
19198         into a direct parameter.  Update call to get_biv_step_1.
19199         (iv_analyze_biv): Add an outer_mode parameter.  Update calls to
19200         iv_constant and get_biv_step.
19201         (iv_analyze_expr): Pass the mode parameter before the rtx it describes
19202         and change its type to scalar_int_mode.  Don't initialise iv->mode
19203         to VOIDmode and remove later checks for its still being VOIDmode.
19204         Update calls to iv_analyze_op and iv_analyze_expr.  Check
19205         is_a <scalar_int_mode> when changing the mode under consideration.
19206         (iv_analyze_def): Ignore registers that don't have a scalar_int_mode.
19207         Update call to iv_analyze_expr.
19208         (iv_analyze_op): Add a mode parameter.  Reject subregs whose
19209         inner register is not also a scalar_int_mode.  Update call to
19210         iv_analyze_biv.
19211         (iv_analyze): Add a mode parameter.  Update call to iv_analyze_op.
19212         (biv_p): Add a mode parameter.  Update call to iv_analyze_biv.
19213         (iv_number_of_iterations): Use is_a <scalar_int_mode> instead of
19214         separate mode class checks.  Update calls to iv_analyze.  Remove
19215         fix-up of VOIDmodes after iv_analyze_biv.
19216         * loop-unroll.c (analyze_iv_to_split_insn): Reject registers that
19217         don't have a scalar_int_mode.  Update call to biv_p.
19219 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19220             Alan Hayward  <alan.hayward@arm.com>
19221             David Sherwood  <david.sherwood@arm.com>
19223         * cfgexpand.c (convert_debug_memory_address): Use
19224         as_a <scalar_int_mode>.
19225         * combine.c (expand_compound_operation): Likewise.
19226         (make_extraction): Likewise.
19227         (change_zero_ext): Likewise.
19228         (simplify_comparison): Likewise.
19229         * cse.c (cse_insn): Likewise.
19230         * dwarf2out.c (minmax_loc_descriptor): Likewise.
19231         (mem_loc_descriptor): Likewise.
19232         (loc_descriptor): Likewise.
19233         * expmed.c (init_expmed_one_mode): Likewise.
19234         (synth_mult): Likewise.
19235         (emit_store_flag_1): Likewise.
19236         (expand_divmod): Likewise.  Use HWI_COMPUTABLE_MODE_P instead
19237         of a comparison with size.
19238         * expr.c (expand_assignment): Use as_a <scalar_int_mode>.
19239         (reduce_to_bit_field_precision): Likewise.
19240         * function.c (expand_function_end): Likewise.
19241         * internal-fn.c (expand_arith_overflow_result_store): Likewise.
19242         * loop-doloop.c (doloop_modify): Likewise.
19243         * optabs.c (expand_binop): Likewise.
19244         (expand_unop): Likewise.
19245         (expand_copysign_absneg): Likewise.
19246         (prepare_cmp_insn): Likewise.
19247         (maybe_legitimize_operand): Likewise.
19248         * recog.c (const_scalar_int_operand): Likewise.
19249         * rtlanal.c (get_address_mode): Likewise.
19250         * simplify-rtx.c (simplify_unary_operation_1): Likewise.
19251         (simplify_cond_clz_ctz): Likewise.
19252         * tree-nested.c (get_nl_goto_field): Likewise.
19253         * tree.c (build_vector_type_for_mode): Likewise.
19254         * var-tracking.c (use_narrower_mode): Likewise.
19256 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19257             Alan Hayward  <alan.hayward@arm.com>
19258             David Sherwood  <david.sherwood@arm.com>
19260         * tree.h (SCALAR_INT_TYPE_MODE): New macro.
19261         * builtins.c (expand_builtin_signbit): Use it.
19262         * cfgexpand.c (expand_debug_expr): Likewise.
19263         * dojump.c (do_jump): Likewise.
19264         (do_compare_and_jump): Likewise.
19265         * dwarf2cfi.c (expand_builtin_init_dwarf_reg_sizes): Likewise.
19266         * expmed.c (make_tree): Likewise.
19267         * expr.c (expand_expr_real_2): Likewise.
19268         (expand_expr_real_1): Likewise.
19269         (try_casesi): Likewise.
19270         * fold-const-call.c (fold_const_call_ss): Likewise.
19271         * fold-const.c (unextend): Likewise.
19272         (extract_muldiv_1): Likewise.
19273         (fold_single_bit_test): Likewise.
19274         (native_encode_int): Likewise.
19275         (native_encode_string): Likewise.
19276         (native_interpret_int): Likewise.
19277         * gimple-fold.c (gimple_fold_builtin_memset): Likewise.
19278         * internal-fn.c (expand_addsub_overflow): Likewise.
19279         (expand_neg_overflow): Likewise.
19280         (expand_mul_overflow): Likewise.
19281         (expand_arith_overflow): Likewise.
19282         * match.pd: Likewise.
19283         * stor-layout.c (layout_type): Likewise.
19284         * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
19285         * tree-ssa-math-opts.c (convert_mult_to_widen): Likewise.
19286         * tree-ssanames.c (get_range_info): Likewise.
19287         * tree-switch-conversion.c (array_value_type) Likewise.
19288         * tree-vect-patterns.c (vect_recog_rotate_pattern): Likewise.
19289         (vect_recog_divmod_pattern): Likewise.
19290         (vect_recog_mixed_size_cond_pattern): Likewise.
19291         * tree-vrp.c (extract_range_basic): Likewise.
19292         (simplify_float_conversion_using_ranges): Likewise.
19293         * tree.c (int_fits_type_p): Likewise.
19294         * ubsan.c (instrument_bool_enum_load): Likewise.
19295         * varasm.c (mergeable_string_section): Likewise.
19296         (narrowing_initializer_constant_valid_p): Likewise.
19297         (output_constant): Likewise.
19299 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19300             Alan Hayward  <alan.hayward@arm.com>
19301             David Sherwood  <david.sherwood@arm.com>
19303         * machmode.h (NARROWEST_INT_MODE): New macro.
19304         * expr.c (alignment_for_piecewise_move): Use it instead of
19305         GET_CLASS_NARROWEST_MODE (MODE_INT).
19306         (push_block): Likewise.
19307         * stor-layout.c (bit_field_mode_iterator::bit_field_mode_iterator):
19308         Likewise.
19309         * tree-vrp.c (simplify_float_conversion_using_ranges): Likewise.
19311 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19312             Alan Hayward  <alan.hayward@arm.com>
19313             David Sherwood  <david.sherwood@arm.com>
19315         * postreload.c (move2add_valid_value_p): Change the type of the
19316         mode parameter to scalar_int_mode.
19317         (move2add_use_add2_insn): Add a mode parameter and use it instead
19318         of GET_MODE (reg).
19319         (move2add_use_add3_insn): Likewise.
19320         (reload_cse_move2add): Update accordingly.
19322 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19323             Alan Hayward  <alan.hayward@arm.com>
19324             David Sherwood  <david.sherwood@arm.com>
19326         * expr.c (expand_expr_real_2): Use scalar_int_mode for the
19327         double-word mode.
19328         * lower-subreg.c (resolve_shift_zext): Use is_a <scalar_int_mode>.
19329         * optabs.c (expand_unop): Likewise.
19331 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19332             Alan Hayward  <alan.hayward@arm.com>
19333             David Sherwood  <david.sherwood@arm.com>
19335         * dwarf2out.c (typed_binop): Change mode parameter to scalar_int_mode.
19336         (clz_loc_descriptor): Likewise.  Remove SCALAR_INT_MODE_P check.
19337         (popcount_loc_descriptor): Likewise.
19338         (bswap_loc_descriptor): Likewise.
19339         (rotate_loc_descriptor): Likewise.
19340         (mem_loc_descriptor): Add is_a <scalar_int_mode> checks before
19341         calling the functions above.
19343 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19344             Alan Hayward  <alan.hayward@arm.com>
19345             David Sherwood  <david.sherwood@arm.com>
19347         * combine.c (sign_extend_short_imm): Add is_a <scalar_int_mode>
19348         checks.
19349         (try_combine): Likewise.
19350         (simplify_if_then_else): Likewise.
19351         * cse.c (cse_insn): Likewise.
19352         * dwarf2out.c (mem_loc_descriptor): Likewise.
19353         * emit-rtl.c (gen_lowpart_common): Likewise.
19354         * simplify-rtx.c (simplify_truncation): Likewise.
19355         (simplify_binary_operation_1): Likewise.
19356         (simplify_const_relational_operation): Likewise.
19357         (simplify_ternary_operation): Likewise.
19358         * tree-ssa-loop-ivopts.c (force_expr_to_var_cost): Likewise.
19360 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19361             Alan Hayward  <alan.hayward@arm.com>
19362             David Sherwood  <david.sherwood@arm.com>
19364         * cse.c (cse_insn): Add is_a <scalar_int_mode> checks.
19365         * reload.c (push_reload): Likewise.
19366         (find_reloads): Likewise.
19368 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19369             Alan Hayward  <alan.hayward@arm.com>
19370             David Sherwood  <david.sherwood@arm.com>
19372         * combine.c (find_split_point): Add is_a <scalar_int_mode> checks.
19373         (make_compound_operation_int): Likewise.
19374         (change_zero_ext): Likewise.
19375         * expr.c (convert_move): Likewise.
19376         (convert_modes): Likewise.
19377         * fwprop.c (forward_propagate_subreg): Likewise.
19378         * loop-iv.c (get_biv_step_1): Likewise.
19379         * optabs.c (widen_operand): Likewise.
19380         * postreload.c (move2add_valid_value_p): Likewise.
19381         * recog.c (simplify_while_replacing): Likewise.
19382         * simplify-rtx.c (simplify_unary_operation_1): Likewise.
19383         (simplify_binary_operation_1): Likewise.  Remove redundant
19384         mode equality check.
19386 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19387             Alan Hayward  <alan.hayward@arm.com>
19388             David Sherwood  <david.sherwood@arm.com>
19390         * combine.c (combine_simplify_rtx): Add checks for
19391         is_a <scalar_int_mode>.
19392         (simplify_if_then_else): Likewise.
19393         (make_field_assignment): Likewise.
19394         (simplify_comparison): Likewise.
19395         * ifcvt.c (noce_try_bitop): Likewise.
19396         * loop-invariant.c (canonicalize_address_mult): Likewise.
19397         * simplify-rtx.c (simplify_unary_operation_1): Likewise.
19399 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19400             Alan Hayward  <alan.hayward@arm.com>
19401             David Sherwood  <david.sherwood@arm.com>
19403         * gimple-fold.c (gimple_fold_builtin_memory_op): Use
19404         is_a <scalar_int_mode> instead of != BLKmode.
19406 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19407             Alan Hayward  <alan.hayward@arm.com>
19408             David Sherwood  <david.sherwood@arm.com>
19410         * cfgexpand.c (expand_debug_expr): Use is_a <scalar_int_mode>
19411         instead of != VOIDmode.
19412         * combine.c (if_then_else_cond): Likewise.
19413         (change_zero_ext): Likewise.
19414         * dwarf2out.c (mem_loc_descriptor): Likewise.
19415         (loc_descriptor): Likewise.
19416         * rtlanal.c (canonicalize_condition): Likewise.
19417         * simplify-rtx.c (simplify_relational_operation_1): Likewise.
19419 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19420             Alan Hayward  <alan.hayward@arm.com>
19421             David Sherwood  <david.sherwood@arm.com>
19423         * simplify-rtx.c (simplify_binary_operation_1): Use
19424         is_a <scalar_int_mode> instead of !VECTOR_MODE_P.
19426 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19427             Alan Hayward  <alan.hayward@arm.com>
19428             David Sherwood  <david.sherwood@arm.com>
19430         * wide-int.h (int_traits<unsigned char>) New class.
19431         (int_traits<unsigned short>) Likewise.
19432         * cfgexpand.c (expand_debug_expr): Use is_a <scalar_int_mode>.
19433         Use GET_MODE_UNIT_PRECISION and remove redundant test for
19434         SCALAR_INT_MODE_P.
19435         * combine.c (set_nonzero_bits_and_sign_copies): Use
19436         is_a <scalar_int_mode>.
19437         (find_split_point): Likewise.
19438         (combine_simplify_rtx): Likewise.
19439         (simplify_logical): Likewise.
19440         (expand_compound_operation): Likewise.
19441         (expand_field_assignment): Likewise.
19442         (make_compound_operation): Likewise.
19443         (extended_count): Likewise.
19444         (change_zero_ext): Likewise.
19445         (simplify_comparison): Likewise.
19446         * dwarf2out.c (scompare_loc_descriptor): Likewise.
19447         (ucompare_loc_descriptor): Likewise.
19448         (minmax_loc_descriptor): Likewise.
19449         (mem_loc_descriptor): Likewise.
19450         (loc_descriptor): Likewise.
19451         * expmed.c (init_expmed_one_mode): Likewise.
19452         * lra-constraints.c (lra_constraint_offset): Likewise.
19453         * optabs.c (prepare_libcall_arg): Likewise.
19454         * postreload.c (move2add_note_store): Likewise.
19455         * reload.c (operands_match_p): Likewise.
19456         * rtl.h (load_extend_op): Likewise.
19457         * rtlhooks.c (gen_lowpart_general): Likewise.
19458         * simplify-rtx.c (simplify_truncation): Likewise.
19459         (simplify_unary_operation_1): Likewise.
19460         (simplify_binary_operation_1): Likewise.
19461         (simplify_const_binary_operation): Likewise.
19462         (simplify_const_relational_operation): Likewise.
19463         (simplify_subreg): Likewise.
19464         * stor-layout.c (bitwise_mode_for_mode): Likewise.
19465         * var-tracking.c (adjust_mems): Likewise.
19466         (prepare_call_arguments): Likewise.
19468 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19469             Alan Hayward  <alan.hayward@arm.com>
19470             David Sherwood  <david.sherwood@arm.com>
19472         * machmode.h (is_int_mode): New fuction.
19473         * combine.c (find_split_point): Use it.
19474         (combine_simplify_rtx): Likewise.
19475         (simplify_if_then_else): Likewise.
19476         (simplify_set): Likewise.
19477         (simplify_shift_const_1): Likewise.
19478         (simplify_comparison): Likewise.
19479         * config/aarch64/aarch64.c (aarch64_rtx_costs): Likewise.
19480         * cse.c (notreg_cost): Likewise.
19481         (cse_insn): Likewise.
19482         * cselib.c (cselib_lookup_1): Likewise.
19483         * dojump.c (do_jump_1): Likewise.
19484         (do_compare_rtx_and_jump): Likewise.
19485         * dse.c (get_call_args): Likewise.
19486         * dwarf2out.c (rtl_for_decl_init): Likewise.
19487         (native_encode_initializer): Likewise.
19488         * expmed.c (emit_store_flag_1): Likewise.
19489         (emit_store_flag): Likewise.
19490         * expr.c (convert_modes): Likewise.
19491         (store_field): Likewise.
19492         (expand_expr_real_1): Likewise.
19493         * fold-const.c (fold_read_from_constant_string): Likewise.
19494         * gimple-ssa-sprintf.c (get_format_string): Likewise.
19495         * optabs-libfuncs.c (gen_int_libfunc): Likewise.
19496         * optabs.c (expand_binop): Likewise.
19497         (expand_unop): Likewise.
19498         (expand_abs_nojump): Likewise.
19499         (expand_one_cmpl_abs_nojump): Likewise.
19500         * simplify-rtx.c (mode_signbit_p): Likewise.
19501         (val_signbit_p): Likewise.
19502         (val_signbit_known_set_p): Likewise.
19503         (val_signbit_known_clear_p): Likewise.
19504         (simplify_relational_operation_1): Likewise.
19505         * tree.c (vector_type_mode): Likewise.
19507 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19508             Alan Hayward  <alan.hayward@arm.com>
19509             David Sherwood  <david.sherwood@arm.com>
19511         * machmode.h (smallest_mode_for_size): Fix formatting.
19512         (smallest_int_mode_for_size): New function.
19513         * cfgexpand.c (expand_debug_expr): Use smallest_int_mode_for_size
19514         instead of smallest_mode_for_size.
19515         * combine.c (make_extraction): Likewise.
19516         * config/arc/arc.c (arc_expand_movmem): Likewise.
19517         * config/arm/arm.c (arm_expand_divmod_libfunc): Likewise.
19518         * config/i386/i386.c (ix86_get_mask_mode): Likewise.
19519         * config/s390/s390.c (s390_expand_insv): Likewise.
19520         * config/sparc/sparc.c (assign_int_registers): Likewise.
19521         * config/spu/spu.c (spu_function_value): Likewise.
19522         (spu_function_arg): Likewise.
19523         * coverage.c (get_gcov_type): Likewise.
19524         (get_gcov_unsigned_t): Likewise.
19525         * dse.c (find_shift_sequence): Likewise.
19526         * expmed.c (store_bit_field_1): Likewise.
19527         * expr.c (convert_move): Likewise.
19528         (store_field): Likewise.
19529         * internal-fn.c (expand_arith_overflow): Likewise.
19530         * optabs-query.c (get_best_extraction_insn): Likewise.
19531         * optabs.c (expand_twoval_binop_libfunc): Likewise.
19532         * stor-layout.c (layout_type): Likewise.
19533         (initialize_sizetypes): Likewise.
19534         * targhooks.c (default_get_mask_mode): Likewise.
19535         * tree-ssa-loop-manip.c (canonicalize_loop_ivs): Likewise.
19537 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19538             Alan Hayward  <alan.hayward@arm.com>
19539             David Sherwood  <david.sherwood@arm.com>
19541         * machmode.h (opt_mode::else_blk): New function.
19542         (int_mode_for_mode): Declare.
19543         * stor-layout.c (int_mode_for_mode): Return an opt_scalar_int_mode.
19544         * builtins.c (expand_builtin_signbit): Adjust for new int_mode_for_mode
19545         return type.
19546         * cfgexpand.c (expand_debug_expr): Likewise.
19547         * combine.c (gen_lowpart_or_truncate): Likewise.
19548         (gen_lowpart_for_combine): Likewise.
19549         * config/aarch64/aarch64.c (aarch64_emit_approx_sqrt): Likewise.
19550         * config/avr/avr.c (avr_to_int_mode): Likewise.
19551         (avr_out_plus_1): Likewise.
19552         (avr_out_plus): Likewise.
19553         (avr_out_round): Likewise.
19554         * config/i386/i386.c (ix86_split_to_parts): Likewise.
19555         * config/s390/s390.c (s390_expand_vec_compare_cc): Likewise.
19556         (s390_expand_vcond): Likewise.
19557         * config/spu/spu.c (spu_split_immediate): Likewise.
19558         (spu_expand_mov): Likewise.
19559         * dse.c (get_stored_val): Likewise.
19560         * expmed.c (store_bit_field_1): Likewise.
19561         (convert_extracted_bit_field): Use int_mode_for_mode instead of
19562         int_mode_for_size.
19563         (extract_bit_field_1): Adjust for new int_mode_for_mode return type.
19564         (extract_low_bits): Likewise.
19565         * expr.c (emit_group_load_1): Likewise.  Separate out the BLKmode
19566         handling rather than repeating the check.
19567         (emit_group_store): Likewise.
19568         (emit_move_via_integer): Adjust for new int_mode_for_mode return type.
19569         * optabs.c (expand_absneg_bit): Likewise.
19570         (expand_copysign_absneg): Likewise.
19571         (expand_copysign_bit): Likewise.
19572         * tree-if-conv.c (ifcvt_can_use_mask_load_store): Likewise.
19573         * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
19574         * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
19575         * var-tracking.c (prepare_call_arguments):  Likewise.
19576         * config/powerpcspe/powerpcspe.c (rs6000_do_expand_vec_perm): Use
19577         int_mode_for_mode instead of mode_for_size.
19578         * config/rs6000/rs6000.c (rs6000_do_expand_vec_perm): Likewise.
19580 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19581             Alan Hayward  <alan.hayward@arm.com>
19582             David Sherwood  <david.sherwood@arm.com>
19584         * machmode.h (int_mode_for_size): New function.
19585         * builtins.c (set_builtin_user_assembler_name): Use int_mode_for_size
19586         instead of mode_for_size.
19587         * calls.c (save_fixed_argument_area): Likewise.  Make use of BLKmode
19588         explicit.
19589         * combine.c (expand_field_assignment): Use int_mode_for_size
19590         instead of mode_for_size.
19591         (make_extraction): Likewise.
19592         (simplify_shift_const_1): Likewise.
19593         (simplify_comparison): Likewise.
19594         * dojump.c (do_jump): Likewise.
19595         * dwarf2out.c (mem_loc_descriptor): Likewise.
19596         * emit-rtl.c (init_derived_machine_modes): Likewise.
19597         * expmed.c (flip_storage_order): Likewise.
19598         (convert_extracted_bit_field): Likewise.
19599         * expr.c (copy_blkmode_from_reg): Likewise.
19600         * graphite-isl-ast-to-gimple.c (max_mode_int_precision): Likewise.
19601         * internal-fn.c (expand_mul_overflow): Likewise.
19602         * lower-subreg.c (simple_move): Likewise.
19603         * optabs-libfuncs.c (init_optabs): Likewise.
19604         * simplify-rtx.c (simplify_unary_operation_1): Likewise.
19605         * tree.c (vector_type_mode): Likewise.
19606         * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
19607         * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Likewise.
19608         * tree-vect-generic.c (expand_vector_parallel): Likewise.
19609         * tree-vect-stmts.c (vectorizable_load): Likewise.
19610         (vectorizable_store): Likewise.
19612 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19613             Alan Hayward  <alan.hayward@arm.com>
19614             David Sherwood  <david.sherwood@arm.com>
19616         * coretypes.h (pod_mode): New type.
19617         (scalar_int_mode_pod): New typedef.
19618         * machmode.h (pod_mode): New class.
19619         (int_n_data_t::m): Change type to scalar_int_mode_pod.
19620         * genmodes.c (emit_mode_int_n): Update accordingly.
19621         * lower-subreg.h (target_lower_subreg): Change type to
19622         scalar_int_mode_pod.
19623         * gdbhooks.py (build_pretty_printer): Handle pod_mode and
19624         scalar_int_mode_pod.
19626 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19627             Alan Hayward  <alan.hayward@arm.com>
19628             David Sherwood  <david.sherwood@arm.com>
19630         * config/powerpcspe/powerpcspe.h (rs6000_pmode): Change type from
19631         machine_mode to scalar_int_mode.
19632         * config/powerpcspe/powerpcspe.c (rs6000_pmode): Likewise.
19633         (rs6000_option_override_internal): Remove cast to int.
19634         * config/rs6000/rs6000.h (rs6000_pmode): Change type from
19635         machine_mode to scalar_int_mode.
19636         * config/rs6000/rs6000.c (rs6000_pmode): Likewise.
19637         (rs6000_option_override_internal): Remove cast to int.
19638         * config/s390/s390.h (Pmode): Remove cast to machine_mode.
19639         * config/epiphany/epiphany.h (RTX_OK_FOR_OFFSET_P): Add cast
19640         to machine_mode.
19641         * config/s390/s390.c (s390_expand_builtin): Likewise.
19642         * coretypes.h (scalar_int_mode): New type.
19643         (opt_scalar_int_mode): New typedef.
19644         * machmode.h (scalar_int_mode): New class.
19645         (scalar_int_mode::includes_p): New function.
19646         (byte_mode): Change type to scalar_int_mode.
19647         (word_mode): Likewise.
19648         (ptr_mode): Likewise.
19649         * emit-rtl.c (byte_mode): Likewise.
19650         (word_mode): Likewise.
19651         (ptr_mode): Likewise.
19652         (init_derived_machine_modes): Update accordingly.
19653         * genmodes.c (get_mode_class): Return scalar_int_mode for MODE_INT
19654         and MODE_PARTIAL_INT.
19655         * gdbhooks.py (build_pretty_printer): Handle scalar_int_mode and
19656         opt_scalar_int_mode.
19658 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19659             Alan Hayward  <alan.hayward@arm.com>
19660             David Sherwood  <david.sherwood@arm.com>
19662         * target.def (libgcc_floating_mode_supported_p): Take a
19663         scalar_float_mode.
19664         * doc/tm.texi: Regenerate.
19665         * targhooks.h (default_libgcc_floating_mode_supported_p): Take a
19666         scalar_float_mode.
19667         * targhooks.c (default_libgcc_floating_mode_supported_p): Likewise.
19668         * config/aarch64/aarch64.c (aarch64_libgcc_floating_mode_supported_p):
19669         Likewise.
19671 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19672             Alan Hayward  <alan.hayward@arm.com>
19673             David Sherwood  <david.sherwood@arm.com>
19675         * target.def (default_floatn_mode): Return an opt_scalar_float_mode.
19676         * doc/tm.texi: Regenerate.
19677         * config/arm/arm.c (arm_floatn_mode): Return an opt_scalar_float_mode.
19678         * config/powerpcspe/powerpcspe.c (rs6000_floatn_mode): Likewise.
19679         * config/rs6000/rs6000.c (rs6000_floatn_mode): Likewise.
19680         * targhooks.h (default_floatn_mode): Likewise.
19681         * targhooks.c (default_floatn_mode): Likewise.
19682         * tree.c (build_common_tree_nodes): Update accordingly.
19684 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19685             Alan Hayward  <alan.hayward@arm.com>
19686             David Sherwood  <david.sherwood@arm.com>
19688         * machmode.h (mode_iterator::start): Provide overload for opt_modes.
19689         (mode_iterator::iterate_p): Likewise.
19690         (mode_iterator::get_wider): Likewise.
19691         * expr.c (init_expr_target): Use opt_scalar_float_mode.
19693 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19694             Alan Hayward  <alan.hayward@arm.com>
19695             David Sherwood  <david.sherwood@arm.com>
19697         * coretypes.h (opt_scalar_float_mode): New typedef.
19698         * machmode.h (float_mode_for_size): New function.
19699         * emit-rtl.c (double_mode): Delete.
19700         (init_emit_once): Use float_mode_for_size.
19701         * stor-layout.c (layout_type): Likewise.
19702         * gdbhooks.py (build_pretty_printer): Handle opt_scalar_float_mode.
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         * output.h (assemble_real): Take a scalar_float_mode.
19709         * config/arm/arm.c (arm_assemble_integer): Update accordingly.
19710         * config/arm/arm.md (consttable_4): Likewise.
19711         (consttable_8): Likewise.
19712         (consttable_16): Likewise.
19713         * config/mips/mips.md (consttable_float): Likewise.
19714         * config/s390/s390.c (s390_output_pool_entry): Likewise.
19715         * varasm.c (assemble_real): Take a scalar_float_mode.
19716         (output_constant_pool_2): Update accordingly.
19717         (output_constant): Likewise.
19719 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19720             Alan Hayward  <alan.hayward@arm.com>
19721             David Sherwood  <david.sherwood@arm.com>
19723         * tree.h (SCALAR_FLOAT_TYPE_MODE): New macro.
19724         * builtins.c (expand_builtin_signbit): Use it instead of TYPE_MODE.
19725         * fold-const.c (fold_convert_const_real_from_fixed): Likewise.
19726         (native_encode_real): Likewise.
19727         (native_interpret_real): Likewise.
19728         * hsa-brig.c (emit_immediate_scalar_to_buffer): Likewise.
19729         * tree-vrp.c (simplify_float_conversion_using_ranges): Likewise.
19731 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19732             Alan Hayward  <alan.hayward@arm.com>
19733             David Sherwood  <david.sherwood@arm.com>
19735         * optabs-libfuncs.c (gen_trunc_conv_libfunc): Use is_a
19736         <scalar_float_mode>.  Simplify.
19737         (gen_extend_conv_libfunc): Likewise.
19739 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19740             Alan Hayward  <alan.hayward@arm.com>
19741             David Sherwood  <david.sherwood@arm.com>
19743         * coretypes.h (scalar_float_mode): New type.
19744         * machmode.h (mode_traits::from_int): Use machine_mode if
19745         USE_ENUM_MODES is defined.
19746         (is_a): New function.
19747         (as_a): Likewise.
19748         (dyn_cast): Likewise.
19749         (scalar_float_mode): New class.
19750         (scalar_float_mode::includes_p): New function.
19751         (is_float_mode): Likewise.
19752         * gdbhooks.py (MachineModePrinter): New class.
19753         (build_pretty_printer): Use it for scalar_float_mode.
19754         * real.h (FLOAT_MODE_FORMAT): Use as_a <scalar_float_mode>.
19755         (format_helper::format_helper): Turn into a template.
19756         * genmodes.c (get_mode_class): New function.
19757         (emit_insn_modes_h): Give modes the class returned by get_mode_class,
19758         or machine_mode if none.
19759         * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Use
19760         as_a <scalar_float_mode>.
19761         * dwarf2out.c (mem_loc_descriptor): Likewise.
19762         (insert_float): Likewise.
19763         (add_const_value_attribute): Likewise.
19764         * simplify-rtx.c (simplify_immed_subreg): Likewise.
19765         * optabs.c (expand_absneg_bit): Take a scalar_float_mode.
19766         (expand_unop): Update accordingly.
19767         (expand_abs_nojump): Likewise.
19768         (expand_copysign_absneg): Take a scalar_float_mode.
19769         (expand_copysign_bit): Likewise.
19770         (expand_copysign): Update accordingly.
19772 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19773             Alan Hayward  <alan.hayward@arm.com>
19774             David Sherwood  <david.sherwood@arm.com>
19776         * coretypes.h (opt_mode): New class.
19777         * machmode.h (opt_mode): Likewise.
19778         (opt_mode::else_void): New function.
19779         (opt_mode::require): Likewise.
19780         (opt_mode::exists): Likewise.
19781         (GET_MODE_WIDER_MODE): Turn into a function and return an opt_mode.
19782         (GET_MODE_2XWIDER_MODE): Likewise.
19783         (mode_iterator::get_wider): Update accordingly.
19784         (mode_iterator::get_2xwider): Likewise.
19785         (mode_iterator::get_known_wider): Likewise, turning into a template.
19786         * combine.c (make_extraction): Update use of GET_MODE_WIDER_MODE,
19787         forcing a wider mode to exist.
19788         * config/cr16/cr16.h (LONG_REG_P): Likewise.
19789         * rtlanal.c (init_num_sign_bit_copies_in_rep): Likewise.
19790         * config/c6x/c6x.c (c6x_rtx_costs): Update use of
19791         GET_MODE_2XWIDER_MODE, forcing a wider mode to exist.
19792         * lower-subreg.c (init_lower_subreg): Likewise.
19793         * optabs-libfuncs.c (init_sync_libfuncs_1): Likewise, but not
19794         on the final iteration.
19795         * config/i386/i386.c (ix86_expand_set_or_movmem): Check whether
19796         a wider mode exists before asking for a move pattern.
19797         (get_mode_wider_vector): Update use of GET_MODE_WIDER_MODE,
19798         forcing a wider mode to exist.
19799         (expand_vselect_vconcat): Update use of GET_MODE_2XWIDER_MODE,
19800         returning false if no such mode exists.
19801         * config/ia64/ia64.c (expand_vselect_vconcat): Likewise.
19802         * config/mips/mips.c (mips_expand_vselect_vconcat): Likewise.
19803         * expmed.c (init_expmed_one_mode): Update use of GET_MODE_WIDER_MODE.
19804         Avoid checking for a MODE_INT if we already know the mode is not a
19805         SCALAR_INT_MODE_P.
19806         (extract_high_half): Update use of GET_MODE_WIDER_MODE,
19807         forcing a wider mode to exist.
19808         (expmed_mult_highpart_optab): Likewise.
19809         (expmed_mult_highpart): Likewise.
19810         * expr.c (expand_expr_real_2): Update use of GET_MODE_WIDER_MODE,
19811         using else_void.
19812         * lto-streamer-in.c (lto_input_mode_table): Likewise.
19813         * optabs-query.c (find_widening_optab_handler_and_mode): Likewise.
19814         * stor-layout.c (bit_field_mode_iterator::next_mode): Likewise.
19815         * internal-fn.c (expand_mul_overflow): Update use of
19816         GET_MODE_2XWIDER_MODE.
19817         * omp-low.c (omp_clause_aligned_alignment): Likewise.
19818         * tree-ssa-math-opts.c (convert_mult_to_widen): Update use of
19819         GET_MODE_WIDER_MODE.
19820         (convert_plusminus_to_widen): Likewise.
19821         * tree-switch-conversion.c (array_value_type): Likewise.
19822         * var-tracking.c (emit_note_insn_var_location): Likewise.
19823         * tree-vrp.c (simplify_float_conversion_using_ranges): Likewise.
19824         Return false inside rather than outside the loop if no wider mode
19825         exists
19826         * optabs.c (expand_binop): Update use of GET_MODE_WIDER_MODE
19827         and GET_MODE_2XWIDER_MODE
19828         (can_compare_p): Use else_void.
19829         * gdbhooks.py (OptMachineModePrinter): New class.
19830         (build_pretty_printer): Use it for opt_mode.
19832 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19833             Alan Hayward  <alan.hayward@arm.com>
19834             David Sherwood  <david.sherwood@arm.com>
19836         * tree-switch-conversion.c (array_value_type): Only read TYPE_MODE
19837         once.  Use get_narrowest_mode instead of GET_CLASS_NARROWEST_MODE.
19839 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19840             Alan Hayward  <alan.hayward@arm.com>
19841             David Sherwood  <david.sherwood@arm.com>
19843         * machmode.h (mode_traits): New structure.
19844         (get_narrowest_mode): New function.
19845         (mode_iterator::start): Likewise.
19846         (mode_iterator::iterate_p): Likewise.
19847         (mode_iterator::get_wider): Likewise.
19848         (mode_iterator::get_known_wider): Likewise.
19849         (mode_iterator::get_2xwider): Likewise.
19850         (FOR_EACH_MODE_IN_CLASS): New mode iterator.
19851         (FOR_EACH_MODE): Likewise.
19852         (FOR_EACH_MODE_FROM): Likewise.
19853         (FOR_EACH_MODE_UNTIL): Likewise.
19854         (FOR_EACH_WIDER_MODE): Likewise.
19855         (FOR_EACH_2XWIDER_MODE): Likewise.
19856         * builtins.c (expand_builtin_strlen): Use new mode iterators.
19857         * combine.c (simplify_comparison): Likewise
19858         * config/i386/i386.c (type_natural_mode): Likewise.
19859         * cse.c (cse_insn): Likewise.
19860         * dse.c (find_shift_sequence): Likewise.
19861         * emit-rtl.c (init_derived_machine_modes): Likewise.
19862         (init_emit_once): Likewise.
19863         * explow.c (hard_function_value): Likewise.
19864         * expmed.c (extract_fixed_bit_field_1): Likewise.
19865         (extract_bit_field_1): Likewise.
19866         (expand_divmod): Likewise.
19867         (emit_store_flag_1): Likewise.
19868         * expr.c (init_expr_target): Likewise.
19869         (convert_move): Likewise.
19870         (alignment_for_piecewise_move): Likewise.
19871         (widest_int_mode_for_size): Likewise.
19872         (emit_block_move_via_movmem): Likewise.
19873         (copy_blkmode_to_reg): Likewise.
19874         (set_storage_via_setmem): Likewise.
19875         (compress_float_constant): Likewise.
19876         * omp-low.c (omp_clause_aligned_alignment): Likewise.
19877         * optabs-query.c (get_best_extraction_insn): Likewise.
19878         * optabs.c (expand_binop): Likewise.
19879         (expand_twoval_unop): Likewise.
19880         (expand_twoval_binop): Likewise.
19881         (widen_leading): Likewise.
19882         (widen_bswap): Likewise.
19883         (expand_parity): Likewise.
19884         (expand_unop): Likewise.
19885         (prepare_cmp_insn): Likewise.
19886         (prepare_float_lib_cmp): Likewise.
19887         (expand_float): Likewise.
19888         (expand_fix): Likewise.
19889         (expand_sfix_optab): Likewise.
19890         * postreload.c (move2add_use_add2_insn): Likewise.
19891         * reg-stack.c (reg_to_stack): Likewise.
19892         * reginfo.c (choose_hard_reg_mode): Likewise.
19893         * rtlanal.c (init_num_sign_bit_copies_in_rep): Likewise.
19894         * stor-layout.c (mode_for_size): Likewise.
19895         (smallest_mode_for_size): Likewise.
19896         (mode_for_vector): Likewise.
19897         (finish_bitfield_representative): Likewise.
19898         * tree-ssa-math-opts.c (target_supports_divmod_p): Likewise.
19899         * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
19900         * tree-vect-stmts.c (vectorizable_conversion): Likewise.
19901         * var-tracking.c (prepare_call_arguments): Likewise.
19903 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19904             Alan Hayward  <alan.hayward@arm.com>
19905             David Sherwood  <david.sherwood@arm.com>
19907         * genconditions.c (write_header): Add a "#define USE_ENUM_MODES".
19908         * genmodes.c (emit_insn_modes_h): Define FOOmode to E_FOOmode if
19909         USE_ENUM_MODES is defined and to ((void) 0, E_FOOmode) otherwise.
19910         * machmode.h (mode_size): Move earlier in file.
19911         (mode_precision): Likewise.
19912         (mode_inner): Likewise.
19913         (mode_nunits): Likewise.
19914         (mode_unit_size): Likewise.
19915         (unit_unit_precision): Likewise.
19916         (mode_wider): Likewise.
19917         (mode_2xwider): Likewise.
19918         (machine_mode): New class.
19919         (mode_to_bytes): New function.
19920         (mode_to_bits): Likewise.
19921         (mode_to_precision): Likewise.
19922         (mode_to_inner): Likewise.
19923         (mode_to_unit_size): Likewise.
19924         (mode_to_unit_precision): Likewise.
19925         (mode_to_nunits): Likewise.
19926         (GET_MODE_SIZE): Use mode_to_bytes.
19927         (GET_MODE_BITSIZE): Use mode_to_bits.
19928         (GET_MODE_PRECISION): Use mode_to_precision.
19929         (GET_MODE_INNER): Use mode_to_inner.
19930         (GET_MODE_UNIT_SIZE): Use mode_to_unit_size.
19931         (GET_MODE_UNIT_PRECISION): Use mode_to_unit_precision.
19932         (GET_MODE_NUNITS): Use mode_to_nunits.
19933         * system.h (ALWAYS_INLINE): New macro.
19934         * config/powerpcspe/powerpcspe-c.c
19935         (altivec_resolve_overloaded_builtin): Use machine_mode instead of
19936         int for arg1_mode and arg2_mode.
19938 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19939             Alan Hayward  <alan.hayward@arm.com>
19940             David Sherwood  <david.sherwood@arm.com>
19942         * config/aarch64/aarch64-builtins.c (aarch64_simd_builtin_std_type):
19943         Prefix mode names with E_ in case statements.
19944         * config/aarch64/aarch64-elf.h (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
19945         * config/aarch64/aarch64.c (aarch64_split_simd_combine): Likewise.
19946         (aarch64_split_simd_move): Likewise.
19947         (aarch64_gen_storewb_pair): Likewise.
19948         (aarch64_gen_loadwb_pair): Likewise.
19949         (aarch64_gen_store_pair): Likewise.
19950         (aarch64_gen_load_pair): Likewise.
19951         (aarch64_get_condition_code_1): Likewise.
19952         (aarch64_constant_pool_reload_icode): Likewise.
19953         (get_rsqrte_type): Likewise.
19954         (get_rsqrts_type): Likewise.
19955         (get_recpe_type): Likewise.
19956         (get_recps_type): Likewise.
19957         (aarch64_gimplify_va_arg_expr): Likewise.
19958         (aarch64_simd_container_mode): Likewise.
19959         (aarch64_emit_load_exclusive): Likewise.
19960         (aarch64_emit_store_exclusive): Likewise.
19961         (aarch64_expand_compare_and_swap): Likewise.
19962         (aarch64_gen_atomic_cas): Likewise.
19963         (aarch64_emit_bic): Likewise.
19964         (aarch64_emit_atomic_swap): Likewise.
19965         (aarch64_emit_atomic_load_op): Likewise.
19966         (aarch64_evpc_trn): Likewise.
19967         (aarch64_evpc_uzp): Likewise.
19968         (aarch64_evpc_zip): Likewise.
19969         (aarch64_evpc_ext): Likewise.
19970         (aarch64_evpc_rev): Likewise.
19971         (aarch64_evpc_dup): Likewise.
19972         (aarch64_gen_ccmp_first): Likewise.
19973         (aarch64_gen_ccmp_next): Likewise.
19974         * config/alpha/alpha.c (alpha_scalar_mode_supported_p): Likewise.
19975         (alpha_emit_xfloating_libcall): Likewise.
19976         (emit_insxl): Likewise.
19977         (alpha_arg_type): Likewise.
19978         * config/arc/arc.c (arc_vector_mode_supported_p): Likewise.
19979         (arc_preferred_simd_mode): Likewise.
19980         (arc_secondary_reload): Likewise.
19981         (get_arc_condition_code): Likewise.
19982         (arc_print_operand): Likewise.
19983         (arc_legitimate_constant_p): Likewise.
19984         * config/arc/arc.h (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
19985         * config/arc/arc.md (casesi_load): Likewise.
19986         (casesi_compact_jump): Likewise.
19987         * config/arc/predicates.md (proper_comparison_operator): Likewise.
19988         (cc_use_register): Likewise.
19989         * config/arm/aout.h (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
19990         * config/arm/arm-builtins.c (arm_simd_builtin_std_type): Likewise.
19991         (arm_init_iwmmxt_builtins): Likewise.
19992         * config/arm/arm.c (thumb1_size_rtx_costs): Likewise.
19993         (neon_expand_vector_init): Likewise.
19994         (arm_attr_length_move_neon): Likewise.
19995         (maybe_get_arm_condition_code): Likewise.
19996         (arm_emit_vector_const): Likewise.
19997         (arm_preferred_simd_mode): Likewise.
19998         (arm_output_iwmmxt_tinsr): Likewise.
19999         (thumb1_output_casesi): Likewise.
20000         (thumb2_output_casesi): Likewise.
20001         (arm_emit_load_exclusive): Likewise.
20002         (arm_emit_store_exclusive): Likewise.
20003         (arm_expand_compare_and_swap): Likewise.
20004         (arm_evpc_neon_vuzp): Likewise.
20005         (arm_evpc_neon_vzip): Likewise.
20006         (arm_evpc_neon_vrev): Likewise.
20007         (arm_evpc_neon_vtrn): Likewise.
20008         (arm_evpc_neon_vext): Likewise.
20009         (arm_validize_comparison): Likewise.
20010         * config/arm/neon.md (neon_vc<cmp_op><mode>): Likewise.
20011         * config/avr/avr-c.c (avr_resolve_overloaded_builtin): Likewise.
20012         * config/avr/avr.c (avr_rtx_costs_1): Likewise.
20013         * config/c6x/c6x.c (c6x_vector_mode_supported_p): Likewise.
20014         (c6x_preferred_simd_mode): Likewise.
20015         * config/epiphany/epiphany.c (get_epiphany_condition_code): Likewise.
20016         (epiphany_rtx_costs): Likewise.
20017         * config/epiphany/predicates.md (proper_comparison_operator):
20018         Likewise.
20019         * config/frv/frv.c (condexec_memory_operand): Likewise.
20020         (frv_emit_move): Likewise.
20021         (output_move_single): Likewise.
20022         (output_condmove_single): Likewise.
20023         (frv_hard_regno_mode_ok): Likewise.
20024         (frv_matching_accg_mode): Likewise.
20025         * config/h8300/h8300.c (split_adds_subs): Likewise.
20026         (h8300_rtx_costs): Likewise.
20027         (h8300_print_operand): Likewise.
20028         (compute_mov_length): Likewise.
20029         (output_logical_op): Likewise.
20030         (compute_logical_op_length): Likewise.
20031         (compute_logical_op_cc): Likewise.
20032         (h8300_shift_needs_scratch_p): Likewise.
20033         (output_a_shift): Likewise.
20034         (compute_a_shift_length): Likewise.
20035         (compute_a_shift_cc): Likewise.
20036         (expand_a_rotate): Likewise.
20037         (output_a_rotate): Likewise.
20038         * config/i386/i386.c (classify_argument): Likewise.
20039         (function_arg_advance_32): Likewise.
20040         (function_arg_32): Likewise.
20041         (function_arg_64): Likewise.
20042         (function_value_64): Likewise.
20043         (ix86_gimplify_va_arg): Likewise.
20044         (ix86_legitimate_constant_p): Likewise.
20045         (put_condition_code): Likewise.
20046         (split_double_mode): Likewise.
20047         (ix86_avx256_split_vector_move_misalign): Likewise.
20048         (ix86_expand_vector_logical_operator): Likewise.
20049         (ix86_split_idivmod): Likewise.
20050         (ix86_expand_adjust_ufix_to_sfix_si): Likewise.
20051         (ix86_build_const_vector): Likewise.
20052         (ix86_build_signbit_mask): Likewise.
20053         (ix86_match_ccmode): Likewise.
20054         (ix86_cc_modes_compatible): Likewise.
20055         (ix86_expand_branch): Likewise.
20056         (ix86_expand_sse_cmp): Likewise.
20057         (ix86_expand_sse_movcc): Likewise.
20058         (ix86_expand_int_sse_cmp): Likewise.
20059         (ix86_expand_vec_perm_vpermi2): Likewise.
20060         (ix86_expand_vec_perm): Likewise.
20061         (ix86_expand_sse_unpack): Likewise.
20062         (ix86_expand_int_addcc): Likewise.
20063         (ix86_split_to_parts): Likewise.
20064         (ix86_vectorize_builtin_gather): Likewise.
20065         (ix86_vectorize_builtin_scatter): Likewise.
20066         (avx_vpermilp_parallel): Likewise.
20067         (inline_memory_move_cost): Likewise.
20068         (ix86_tieable_integer_mode_p): Likewise.
20069         (x86_maybe_negate_const_int): Likewise.
20070         (ix86_expand_vector_init_duplicate): Likewise.
20071         (ix86_expand_vector_init_one_nonzero): Likewise.
20072         (ix86_expand_vector_init_one_var): Likewise.
20073         (ix86_expand_vector_init_concat): Likewise.
20074         (ix86_expand_vector_init_interleave): Likewise.
20075         (ix86_expand_vector_init_general): Likewise.
20076         (ix86_expand_vector_set): Likewise.
20077         (ix86_expand_vector_extract): Likewise.
20078         (emit_reduc_half): Likewise.
20079         (ix86_emit_i387_round): Likewise.
20080         (ix86_mangle_type): Likewise.
20081         (ix86_expand_round_sse4): Likewise.
20082         (expand_vec_perm_blend): Likewise.
20083         (canonicalize_vector_int_perm): Likewise.
20084         (ix86_expand_vec_one_operand_perm_avx512): Likewise.
20085         (expand_vec_perm_1): Likewise.
20086         (expand_vec_perm_interleave3): Likewise.
20087         (expand_vec_perm_even_odd_pack): Likewise.
20088         (expand_vec_perm_even_odd_1): Likewise.
20089         (expand_vec_perm_broadcast_1): Likewise.
20090         (ix86_vectorize_vec_perm_const_ok): Likewise.
20091         (ix86_expand_vecop_qihi): Likewise.
20092         (ix86_expand_mul_widen_hilo): Likewise.
20093         (ix86_expand_sse2_abs): Likewise.
20094         (ix86_expand_pextr): Likewise.
20095         (ix86_expand_pinsr): Likewise.
20096         (ix86_preferred_simd_mode): Likewise.
20097         (ix86_simd_clone_compute_vecsize_and_simdlen): Likewise.
20098         * config/i386/sse.md (*andnot<mode>3): Likewise.
20099         (<mask_codefor><code><mode>3<mask_name>): Likewise.
20100         (*<code><mode>3): Likewise.
20101         * config/ia64/ia64.c (ia64_expand_vecint_compare): Likewise.
20102         (ia64_expand_atomic_op): Likewise.
20103         (ia64_arg_type): Likewise.
20104         (ia64_mode_to_int): Likewise.
20105         (ia64_scalar_mode_supported_p): Likewise.
20106         (ia64_vector_mode_supported_p): Likewise.
20107         (expand_vec_perm_broadcast): Likewise.
20108         * config/iq2000/iq2000.c (iq2000_move_1word): Likewise.
20109         (iq2000_function_arg_advance): Likewise.
20110         (iq2000_function_arg): Likewise.
20111         * config/m32c/m32c.c (m32c_preferred_reload_class): Likewise.
20112         * config/m68k/m68k.c (output_dbcc_and_branch): Likewise.
20113         (m68k_libcall_value): Likewise.
20114         (m68k_function_value): Likewise.
20115         (sched_attr_op_type): Likewise.
20116         * config/mcore/mcore.c (mcore_output_move): Likewise.
20117         * config/microblaze/microblaze.c (microblaze_function_arg_advance):
20118         Likewise.
20119         (microblaze_function_arg): Likewise.
20120         * config/mips/mips.c (mips16_build_call_stub): Likewise.
20121         (mips_print_operand): Likewise.
20122         (mips_mode_ok_for_mov_fmt_p): Likewise.
20123         (mips_vector_mode_supported_p): Likewise.
20124         (mips_preferred_simd_mode): Likewise.
20125         (mips_expand_vpc_loongson_even_odd): Likewise.
20126         (mips_expand_vec_unpack): Likewise.
20127         (mips_expand_vi_broadcast): Likewise.
20128         (mips_expand_vector_init): Likewise.
20129         (mips_expand_vec_reduc): Likewise.
20130         (mips_expand_msa_cmp): Likewise.
20131         * config/mips/mips.md (casesi_internal_mips16_<mode>): Likewise.
20132         * config/mn10300/mn10300.c (mn10300_print_operand): Likewise.
20133         (cc_flags_for_mode): Likewise.
20134         * config/msp430/msp430.c (msp430_print_operand): Likewise.
20135         * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Likewise.
20136         (nds32_output_casesi_pc_relative): Likewise.
20137         * config/nds32/nds32.h (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
20138         * config/nvptx/nvptx.c (nvptx_ptx_type_from_mode): Likewise.
20139         (nvptx_gen_unpack): Likewise.
20140         (nvptx_gen_pack): Likewise.
20141         (nvptx_gen_shuffle): Likewise.
20142         (nvptx_gen_wcast): Likewise.
20143         (nvptx_preferred_simd_mode): Likewise.
20144         * config/pa/pa.c (pa_secondary_reload): Likewise.
20145         * config/pa/predicates.md (base14_operand): Likewise.
20146         * config/powerpcspe/powerpcspe-c.c
20147         (altivec_resolve_overloaded_builtin): Likewise.
20148         * config/powerpcspe/powerpcspe.c (rs6000_setup_reg_addr_masks):
20149         Likewise.
20150         (rs6000_preferred_simd_mode): Likewise.
20151         (output_vec_const_move): Likewise.
20152         (rs6000_expand_vector_extract): Likewise.
20153         (rs6000_split_vec_extract_var): Likewise.
20154         (reg_offset_addressing_ok_p): Likewise.
20155         (rs6000_legitimate_offset_address_p): Likewise.
20156         (rs6000_legitimize_address): Likewise.
20157         (rs6000_emit_set_const): Likewise.
20158         (rs6000_const_vec): Likewise.
20159         (rs6000_emit_move): Likewise.
20160         (spe_build_register_parallel): Likewise.
20161         (rs6000_darwin64_record_arg_recurse): Likewise.
20162         (swap_selector_for_mode): Likewise.
20163         (spe_init_builtins): Likewise.
20164         (paired_init_builtins): Likewise.
20165         (altivec_init_builtins): Likewise.
20166         (do_load_for_compare): Likewise.
20167         (rs6000_generate_compare): Likewise.
20168         (rs6000_expand_float128_convert): Likewise.
20169         (emit_load_locked): Likewise.
20170         (emit_store_conditional): Likewise.
20171         (rs6000_output_function_epilogue): Likewise.
20172         (rs6000_handle_altivec_attribute): Likewise.
20173         (rs6000_function_value): Likewise.
20174         (emit_fusion_gpr_load): Likewise.
20175         (emit_fusion_p9_load): Likewise.
20176         (emit_fusion_p9_store): Likewise.
20177         * config/powerpcspe/predicates.md (easy_fp_constant): Likewise.
20178         (fusion_gpr_mem_load): Likewise.
20179         (fusion_addis_mem_combo_load): Likewise.
20180         (fusion_addis_mem_combo_store): Likewise.
20181         * config/rs6000/predicates.md (easy_fp_constant): Likewise.
20182         (fusion_gpr_mem_load): Likewise.
20183         (fusion_addis_mem_combo_load): Likewise.
20184         (fusion_addis_mem_combo_store): Likewise.
20185         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
20186         Likewise.
20187         * config/rs6000/rs6000-string.c (do_load_for_compare): Likewise.
20188         * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): Likewise.
20189         (rs6000_preferred_simd_mode): Likewise.
20190         (output_vec_const_move): Likewise.
20191         (rs6000_expand_vector_extract): Likewise.
20192         (rs6000_split_vec_extract_var): Likewise.
20193         (reg_offset_addressing_ok_p): Likewise.
20194         (rs6000_legitimate_offset_address_p): Likewise.
20195         (rs6000_legitimize_address): Likewise.
20196         (rs6000_emit_set_const): Likewise.
20197         (rs6000_const_vec): Likewise.
20198         (rs6000_emit_move): Likewise.
20199         (rs6000_darwin64_record_arg_recurse): Likewise.
20200         (swap_selector_for_mode): Likewise.
20201         (paired_init_builtins): Likewise.
20202         (altivec_init_builtins): Likewise.
20203         (rs6000_expand_float128_convert): Likewise.
20204         (emit_load_locked): Likewise.
20205         (emit_store_conditional): Likewise.
20206         (rs6000_output_function_epilogue): Likewise.
20207         (rs6000_handle_altivec_attribute): Likewise.
20208         (rs6000_function_value): Likewise.
20209         (emit_fusion_gpr_load): Likewise.
20210         (emit_fusion_p9_load): Likewise.
20211         (emit_fusion_p9_store): Likewise.
20212         * config/rx/rx.c (rx_gen_move_template): Likewise.
20213         (flags_from_mode): Likewise.
20214         * config/s390/predicates.md (s390_alc_comparison): Likewise.
20215         (s390_slb_comparison): Likewise.
20216         * config/s390/s390.c (s390_handle_vectorbool_attribute): Likewise.
20217         (s390_vector_mode_supported_p): Likewise.
20218         (s390_cc_modes_compatible): Likewise.
20219         (s390_match_ccmode_set): Likewise.
20220         (s390_canonicalize_comparison): Likewise.
20221         (s390_emit_compare_and_swap): Likewise.
20222         (s390_branch_condition_mask): Likewise.
20223         (s390_rtx_costs): Likewise.
20224         (s390_secondary_reload): Likewise.
20225         (__SECONDARY_RELOAD_CASE): Likewise.
20226         (s390_expand_cs): Likewise.
20227         (s390_preferred_simd_mode): Likewise.
20228         * config/s390/vx-builtins.md (vec_packsu_u<mode>): Likewise.
20229         * config/sh/sh.c (sh_print_operand): Likewise.
20230         (dump_table): Likewise.
20231         (sh_secondary_reload): Likewise.
20232         * config/sh/sh.h (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
20233         * config/sh/sh.md (casesi_worker_1): Likewise.
20234         (casesi_worker_2): Likewise.
20235         * config/sparc/predicates.md (icc_comparison_operator): Likewise.
20236         (fcc_comparison_operator): Likewise.
20237         * config/sparc/sparc.c (sparc_expand_move): Likewise.
20238         (emit_soft_tfmode_cvt): Likewise.
20239         (sparc_preferred_simd_mode): Likewise.
20240         (output_cbranch): Likewise.
20241         (sparc_print_operand): Likewise.
20242         (sparc_expand_vec_perm_bmask): Likewise.
20243         (vector_init_bshuffle): Likewise.
20244         * config/spu/spu.c (spu_scalar_mode_supported_p): Likewise.
20245         (spu_vector_mode_supported_p): Likewise.
20246         (spu_expand_insv): Likewise.
20247         (spu_emit_branch_or_set): Likewise.
20248         (spu_handle_vector_attribute): Likewise.
20249         (spu_builtin_splats): Likewise.
20250         (spu_builtin_extract): Likewise.
20251         (spu_builtin_promote): Likewise.
20252         (spu_expand_sign_extend): Likewise.
20253         * config/tilegx/tilegx.c (tilegx_scalar_mode_supported_p): Likewise.
20254         (tilegx_simd_int): Likewise.
20255         * config/tilepro/tilepro.c (tilepro_scalar_mode_supported_p): Likewise.
20256         (tilepro_simd_int): Likewise.
20257         * config/v850/v850.c (const_double_split): Likewise.
20258         (v850_print_operand): Likewise.
20259         (ep_memory_offset): Likewise.
20260         * config/vax/vax.c (vax_rtx_costs): Likewise.
20261         (vax_output_int_move): Likewise.
20262         (vax_output_int_add): Likewise.
20263         (vax_output_int_subtract): Likewise.
20264         * config/visium/predicates.md (visium_branch_operator): Likewise.
20265         * config/visium/visium.c (rtx_ok_for_offset_p): Likewise.
20266         (visium_print_operand_address): Likewise.
20267         * config/visium/visium.h (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
20268         * config/xtensa/xtensa.c (xtensa_mem_offset): Likewise.
20269         (xtensa_expand_conditional_branch): Likewise.
20270         (xtensa_copy_incoming_a7): Likewise.
20271         (xtensa_output_literal): Likewise.
20272         * dfp.c (decimal_real_maxval): Likewise.
20273         * targhooks.c (default_libgcc_floating_mode_supported_p): Likewise.
20275 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
20276             Alan Hayward  <alan.hayward@arm.com>
20277             David Sherwood  <david.sherwood@arm.com>
20279         * genmodes.c (mode_size_inline): Add an E_ prefix to mode names.
20280         (mode_nunits_inline): Likewise.
20281         (mode_inner_inline): Likewise.
20282         (mode_unit_size_inline): Likewise.
20283         (mode_unit_precision_inline): Likewise.
20284         (emit_insn_modes_h): Likewise.  Also emit a #define of the
20285         unprefixed name.
20286         (emit_mode_wider): Add an E_ prefix to mode names.
20287         (emit_mode_complex): Likewise.
20288         (emit_mode_inner): Likewise.
20289         (emit_mode_adjustments): Likewise.
20290         (emit_mode_int_n): Likewise.
20291         * config/aarch64/aarch64-builtins.c (v8qi_UP, v4hi_UP, v4hf_UP)
20292         (v2si_UP, v2sf_UP, v1df_UP, di_UP, df_UP, v16qi_UP, v8hi_UP, v8hf_UP)
20293         (v4si_UP, v4sf_UP, v2di_UP, v2df_UP, ti_UP, oi_UP, ci_UP, xi_UP)
20294         (si_UP, sf_UP, hi_UP, hf_UP, qi_UP): Likewise.
20295         (CRC32_BUILTIN, ENTRY): Likewise.
20296         * config/aarch64/aarch64.c (aarch64_push_regs): Likewise.
20297         (aarch64_pop_regs): Likewise.
20298         (aarch64_process_components): Likewise.
20299         * config/alpha/alpha.c (alpha_emit_conditional_move): Likewise.
20300         * config/arm/arm-builtins.c (v8qi_UP, v4hi_UP, v4hf_UP, v2si_UP)
20301         (v2sf_UP, di_UP, v16qi_UP, v8hi_UP, v8hf_UP, v4si_UP, v4sf_UP)
20302         (v2di_UP, ti_UP, ei_UP, oi_UP, hf_UP, si_UP, void_UP): Likewise.
20303         * config/arm/arm.c (arm_init_libfuncs): Likewise.
20304         * config/i386/i386-builtin-types.awk (ix86_builtin_type_vect_mode):
20305         Likewise.
20306         * config/i386/i386-builtin.def (pcmpestr): Likewise.
20307         (pcmpistr): Likewise.
20308         * config/microblaze/microblaze.c (double_memory_operand): Likewise.
20309         * config/mmix/mmix.c (mmix_output_condition): Likewise.
20310         * config/powerpcspe/powerpcspe.c (rs6000_init_hard_regno_mode_ok):
20311         Likewise.
20312         * config/rl78/rl78.c (mduc_regs): Likewise.
20313         * config/rs6000/rs6000.c (rs6000_init_hard_regno_mode_ok): Likewise.
20314         (htm_expand_builtin): Likewise.
20315         * config/sh/sh.h (REGISTER_NATURAL_MODE): Likewise.
20316         * config/sparc/sparc.c (emit_save_or_restore_regs): Likewise.
20317         * config/xtensa/xtensa.c (print_operand): Likewise.
20318         * expmed.h (NUM_MODE_PARTIAL_INT): Likewise.
20319         (NUM_MODE_VECTOR_INT): Likewise.
20320         * genoutput.c (null_operand): Likewise.
20321         (output_operand_data): Likewise.
20322         * genrecog.c (print_parameter_value): Likewise.
20323         * lra.c (debug_operand_data): Likewise.
20325 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
20327         * dwarf2out.c (scompare_loc_descriptor_wide)
20328         (scompare_loc_descriptor_narrow): New functions, split out from...
20329         (scompare_loc_descriptor): ...here.
20330         * expmed.c (emit_store_flag_int): New function, split out from...
20331         (emit_store_flag): ...here.
20333 2017-08-30  Richard Biener  <rguenther@suse.de>
20335         * dwarf2out.c (dwarf2out_finish): Remove setting AT_pubnames.
20336         (dwarf2out_early_finish): Move setting of AT_pubnames from
20337         early debug output to early finish.
20339 2017-08-30  Jozef Lawrynowicz  <jozef.l@somniumtech.com>
20341         * gcc/config/msp430/msp430.h: Pass -mcode/data-region to the linker
20342         and -mdata-region to the assembler.
20344 2017-08-30  Richard Biener  <rguenther@suse.de>
20346         * dwarf2out.c (add_dwarf_attr): Check we don't add duplicate
20347         attributes.
20348         (gen_subprogram_die): Add DW_AT_object_pointer only early.
20349         (dwarf2out_early_global_decl): Only generate a DIE for the
20350         abstract origin if it doesn't already exist or is a declaration DIE.
20351         (resolve_addr): Do not add the linkage name twice when
20352         generating a stub DIE for the DW_TAG_GNU_call_site target.
20354 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
20356         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
20357         Use machine_mode rather than int for arg1_mode.
20359 2017-08-29  Michael Meissner  <meissner@linux.vnet.ibm.com>
20361         PR target/82015
20362         * config/rs6000/rs6000.c (rs6000_expand_binop_builtin): Insure
20363         that the second argument of the built-in functions to unpack
20364         128-bit scalar types to 64-bit values is 0 or 1.  Change to use a
20365         switch statement instead a lot of if statements.
20366         * config/rs6000/rs6000.md (unpack<mode>, FMOVE128_VSX iterator):
20367         Allow 64-bit values to be in Altivec registers as well as
20368         traditional floating point registers.
20369         (pack<mode>, FMOVE128_VSX iterator): Likewise.
20371 2017-08-29  Alexander Monakov  <amonakov@ispras.ru>
20373         * ira-costs.c (record_address_regs): Handle both operands of PLUS for
20374         MAX_REGS_PER_ADDRESS == 1.
20376 2017-08-29  Uros Bizjak  <ubizjak@gmail.com>
20378         * config/i386/i386.opt (flag_fentry): Do not init to -1.
20379         * config/i386/i386.c (ix86_option_override_internal): Simplify
20380         setting of opts->x_flag_entry.
20382 2017-08-29  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
20383             Jakub Jelinek  <jakub@redhat.com>
20384             Richard Biener  <rguenther@suse.de>
20386         PR tree-optimization/81503
20387         * gimple-ssa-strength-reduction.c (replace_mult_candidate): Ensure
20388         folded constant fits in the target type; reorder tests for clarity.
20390 2017-08-29  Martin Liska  <mliska@suse.cz>
20392         * passes.def: Include pass_lower_switch.
20393         * stmt.c (dump_case_nodes): Remove and move to
20394         tree-switch-conversion.
20395         (case_values_threshold): Likewise.
20396         (expand_switch_as_decision_tree_p): Likewise.
20397         (emit_case_decision_tree): Likewise.
20398         (expand_case): Likewise.
20399         (balance_case_nodes): Likewise.
20400         (node_has_low_bound): Likewise.
20401         (node_has_high_bound): Likewise.
20402         (node_is_bounded): Likewise.
20403         (emit_case_nodes): Likewise.
20404         (struct simple_case_node): New struct.
20405         (add_case_node): Remove.
20406         (emit_case_dispatch_table): Use vector instead of case_list.
20407         (reset_out_edges_aux): Remove.
20408         (compute_cases_per_edge): Likewise.
20409         (expand_case): Build list of simple_case_node.
20410         (expand_sjlj_dispatch_table): Use it.
20411         * tree-switch-conversion.c (struct case_node): Moved from
20412         stmt.c and adjusted.
20413         (emit_case_nodes): Likewise.
20414         (node_has_low_bound): Likewise.
20415         (node_has_high_bound): Likewise.
20416         (node_is_bounded): Likewise.
20417         (case_values_threshold): Likewise.
20418         (reset_out_edges_aux): Likewise.
20419         (compute_cases_per_edge): Likewise.
20420         (add_case_node): Likewise.
20421         (dump_case_nodes): Likewise.
20422         (balance_case_nodes): Likewise.
20423         (expand_switch_as_decision_tree_p): Likewise.
20424         (emit_jump): Likewise.
20425         (emit_case_decision_tree): Likewise.
20426         (try_switch_expansion): Likewise.
20427         (do_jump_if_equal): Likewise.
20428         (emit_cmp_and_jump_insns): Likewise.
20429         (fix_phi_operands_for_edge): New function.
20430         (record_phi_operand_mapping): Likewise.
20431         (class pass_lower_switch): New pass.
20432         (pass_lower_switch::execute): New function.
20433         (make_pass_lower_switch): Likewise.
20434         (conditional_probability):
20435         * timevar.def: Add TV_TREE_SWITCH_LOWERING.
20436         * tree-pass.h: Add make_pass_lower_switch.
20438 2017-08-29  Jozef Lawrynowicz  <jozef.l@somniumtech.com>
20440         PR target/80993
20441         * gcc/config/msp430/msp430.c (msp430_attr): Mark interrupt
20442         handlers as used.
20444 2017-08-29  Richard Biener  <rguenther@suse.de>
20446         * dwarf2out.c (add_dwarf_attr): When checking is enabled verify
20447         we do not add a DW_AT_inline attribute twice.
20448         (gen_subprogram_die): Remove code setting DW_AT_inline on
20449         DECL_ABSTRACT_P nodes.
20451 2017-08-29  Richard Sandiford  <richard.sandiford@linaro.org>
20453         * gimplify.c (gimplify_call_expr): Copy the nothrow flag to
20454         calls to internal functions.
20455         (gimplify_modify_expr): Likewise.
20456         * tree-call-cdce.c (use_internal_fn): Likewise.
20457         * tree-ssa-math-opts.c (pass_cse_reciprocals::execute): Likewise.
20458         (convert_to_divmod): Set the nothrow flag.
20459         * tree-if-conv.c (predicate_mem_writes):  Likewise.
20460         * tree-vect-stmts.c (vectorizable_mask_load_store): Likewise.
20461         (vectorizable_call): Likewise.
20462         (vectorizable_store): Likewise.
20463         (vectorizable_load): Likewise.
20464         * tree-vect-patterns.c (vect_recog_pow_pattern): Likewise.
20465         (vect_recog_mask_conversion_pattern): Likewise.
20467 2017-08-29  Martin Liska  <mliska@suse.cz>
20469         PR other/39851
20470         * gcc.c (driver_handle_option): Add new argument.
20471         * opts-common.c (handle_option): Pass
20472         target_option_override_hook.
20473         * opts-global.c (lang_handle_option): Add new option.
20474         (set_default_handlers):  Add new argument.
20475         (decode_options): Likewise.
20476         * opts.c (target_handle_option): Likewise.
20477         (common_handle_option): Call target_option_override_hook.
20478         * opts.h (struct cl_option_handler_func): Add hook for
20479         target option override.
20480         (struct cl_option_handlers): Likewise.
20481         (set_default_handlers): Add new argument.
20482         (decode_options): Likewise.
20483         (common_handle_option): Likewise.
20484         (target_handle_option): Likewise.
20485         * toplev.c (toplev::main): Pass targetm.target_option.override
20486         hook.
20488 2017-08-29  Richard Biener  <rguenther@suse.de>
20489         Dominik Infuehr <dominik.infuehr@theobroma-systems.com>
20491         * tree-vect-slp.c (vect_bb_slp_scalar_cost): Properly confine
20492         life to the active subtree.
20494 2017-08-28  Jeff Law  <law@redhat.com>
20496         * tree-ssa-dom.c (edge_info::record_simple_equiv): Call
20497         derive_equivalences.
20498         (derive_equivalences_from_bit_ior, record_temporary_equivalences):
20499         Code moved into....
20500         (edge_info::derive_equivalences): New private member function
20502         * tree-ssa-dom.c (class edge_info): Changed from a struct
20503         to a class.  Add ctor/dtor, methods and data members.
20504         (edge_info::edge_info): Renamed from allocate_edge_info.
20505         Initialize additional members.
20506         (edge_info::~edge_info): New.
20507         (free_dom_edge_info): Delete the edge info.
20508         (record_edge_info): Use new class & associated member functions.
20509         Tighten forms for testing for edge equivalences.
20510         (record_temporary_equivalences): Iterate over the simple
20511         equivalences rather than assuming there's only one per edge.
20512         (cprop_into_successor_phis): Iterate over the simple
20513         equivalences rather than assuming there's only one per edge.
20514         (optimize_stmt): Use operand_equal_p rather than pointer
20515         equality for mini-DSE code.
20517 2017-08-28  Nathan Sidwell  <nathan@acm.org>
20519         * gcc.c (execute): Fold SIGPIPE handling into switch
20520         statement.  Adjust internal error message.
20522 2017-08-28  Richard Biener  <rguenther@suse.de>
20524         PR debug/81993
20525         * dwarf2out.c (gen_remaining_tmpl_value_param_die_attributes):
20526         Do nothing for removed DIEs.
20528 2017-08-28  Richard Biener  <rguenther@suse.de>
20530         PR tree-optimization/81977
20531         * tree-ssa-sccvn.c (vn_reference_lookup_3): Fix look through
20532         memcpy.
20534 2017-08-28  Alexander Monakov  <amonakov@ispras.ru>
20536         PR target/80640
20537         * doc/md.texi (mem_thread_fence): Remove mention of mode.  Rewrite.
20538         * optabs.c (expand_mem_thread_fence): Emit a compiler barrier when
20539         using targetm.gen_mem_thread_fence.
20541 2017-08-27  Uros Bizjak  <ubizjak@gmail.com>
20543         PR target/81995
20544         * config/i386/i386.md (*<btsc><mode>): Change operand 2
20545         predicate to register_operand.  Reorder operands.
20546         (*btr<mode>): Ditto.
20547         (*<btsc><mode>_mask): Change operand 3 predicate to register_operand.
20548         (*btr<mode>_mask): Ditto.
20550 2017-08-25  Steven Munroe  <munroesj@gcc.gnu.org>
20552         * config.gcc (powerpc*-*-*): Add xmmintrin.h and mm_malloc.h.
20553         * config/rs6000/xmmintrin.h: New file.
20554         * config/rs6000/x86intrin.h [__ALTIVEC__]: Include xmmintrin.h.
20556 2017-08-25  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
20558         PR target/81504
20559         * config/rs6000/rs6000-p8swap.c (find_alignment_op): Add reference
20560         parameter and_insn and return it.
20561         (recombine_lvx_pattern): Insert a copy to ensure availability of
20562         the base register of the copied masking operation at the point of
20563         the instruction replacement.
20564         (recombine_stvx_pattern): Likewise.
20566 2017-08-25  Michael Meissner  <meissner@linux.vnet.ibm.com>
20568         * config/rs6000/rs6000.opt (-mpower9-dform-scalar): Delete
20569         undocumented switches.
20570         (-mpower9-dform-vector): Likewise.
20571         (-mpower9-dform): Likewise.
20572         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
20573         comments to delete references to -mpower9-dform* switches.
20574         * config/rs6000/predicates.md (vsx_quad_dform_memory_operand):
20575         Delete reference to -mpower9-dform* switches, test for
20576         -mpower9-vector instead.
20577         * config/rs6000/rs6000-cpus.def (ISA_3_0_MASKS_SERVER): Likewise.
20578         (OTHER_P9_VECTOR_MASKS): Likewise.
20579         (POWERPC_MASKS): Likewise.
20580         * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): Change
20581         tests against -mpower9-dform* to -mpower9-vector.  Delete code
20582         that checked for -mpower9-dform* consistancy with other options.
20583         Add test for -mpower9-misc to enable other power9 switches.
20584         (rs6000_init_hard_regno_mode_ok): Likewise.
20585         (rs6000_option_override_internal): Likewise.
20586         (rs6000_emit_prologue): Likewise.
20587         (rs6000_emit_epilogue): Likewise.
20588         (rs6000_opt_masks): Delete -mpower9-dform-{scalar,vector}.
20589         (rs6000_disable_incompatiable_switches): Delete -mpower9-dform.
20590         (emit_fusion_p9_load): Change tests for -mpower9-dform-scalar
20591         -mpower9-vector.
20592         (emit_fusion_p9_store): Likewise.
20593         * config/rs6000/rs6000.h (TARGET_P9_DFORM_SCALAR): Delete
20594         resetting these macros if the assembler does not support ISA 3.0
20595         instructions.
20596         (TARGET_P9_DFORM_VECTOR): Likewise.
20597         * config/rs6000/rs6000.md (peepholes to optimize altivec memory):
20598         Change to use -mpower9-vector instead of -mpower9-dform-scalar.
20600 2017-08-25  Alan Modra  <amodra@gmail.com>
20602         PR rtl-optimization/81747
20603         * cse.c (cse_extended_basic_block): Don't attempt to record
20604         equivalences for degenerate conditional branches that branch
20605         to their fall-through.
20607 2017-08-24  Martin Sebor  <msebor@redhat.com>
20609         PR middle-end/81908
20610         * gimple-fold.c (size_must_be_zero_p): New function.
20611         (gimple_fold_builtin_memory_op): Call it.
20613 2017-08-24  Steven Munroe  <munroesj@gcc.gnu.org>
20615         * config/rs6000/mm_malloc.h: New file.
20617 2017-08-24  Bin Cheng  <bin.cheng@arm.com>
20619         PR tree-optimization/81913
20620         * tree-ssa-loop-niter.c (number_of_iterations_cond): Skip niter
20621         analysis when either IVs in condition can wrap.
20623 2017-08-24  Uros Bizjak  <ubizjak@gmail.com>
20625         * dwarf2out.c (MAX_ARTIFICIAL_LABEL_BYTES): Increase to 40.
20626         * dwarf2cfi.c (MAX_ARTIFICIAL_LABEL_BYTES): Remove.
20628 2017-08-24  Richard Biener  <rguenther@suse.de>
20630         PR target/81921
20631         * targhooks.c (default_target_can_inline_p): Properly
20632         use target_option_default_node when no DECL_FUNCTION_SPECIFIC_TARGET
20633         is present and always compare.
20634         * config/i386/i386.c (ix86_valid_target_attribute_tree): Do not
20635         infer -mfpmath=sse from TARGET_SSE_P.
20636         (ix86_can_inline_p): Properly use target_option_default_node when
20637         no DECL_FUNCTION_SPECIFIC_TARGET is present and always compare.
20639 2017-08-24  Richard Biener  <rguenther@suse.de>
20641         PR debug/81936
20642         * dwarf2out.c (output_die): Handle flag_generate_offload like
20643         flag_generate_lto.
20644         (output_comp_unit): Likewise.
20645         (gen_array_type_die): Likewise.
20646         (dwarf2out_early_finish): Likewise.
20647         (note_variable_value_in_expr): Likewise.
20648         (dwarf2out_finish): Likewise.  Adjust assert.
20649         * cgraphunit.c (symbol_table::compile): Move setting of
20650         flag_generate_offload earlier ...
20651         (symbol_table::finalize_compilation_unit): ... here, before
20652         early debug finalization.
20654 2017-08-24  Richard Biener  <rguenther@suse.de>
20656         * config/i386/i386.c: Include symbol-summary.h, ipa-prop.h
20657         and ipa-fnsummary.h.
20658         (ix86_can_inline_p): When ix86_fpmath flags do not match
20659         check whether the callee uses FP math at all.
20661 2017-08-24  Aldy Hernandez  <aldyh@redhat.com>
20663         PR middle-end/81931
20664         * tree-ssanames.c (get_nonzero_bits): Use element_precision
20665         instead of TYPE_PRECISION.
20667 2017-08-24  Richard Sandiford  <richard.sandiford@linaro.org>
20668             Alan Hayward  <alan.hayward@arm.com>
20669             David Sherwood  <david.sherwood@arm.com>
20671         * combine.c (make_extraction): Use subreg_offset_from_lsb.
20673 2017-08-23  Daniel Santos  <daniel.santos@pobox.com>
20675         * config/i386/i386.h (ix86_frame::stack_realign_allocate_offset):
20676         Remove field.
20677         (ix86_frame::stack_realign_allocate): New field.
20678         (struct machine_frame_state): Modify comments.
20679         (machine_frame_state::sp_realigned_fp_end): New field.
20680         * config/i386/i386.c (ix86_compute_frame_layout): Rework stack frame
20681         layout calculation.
20682         (sp_valid_at): Add assertion to assure no attempt to access invalid
20683         offset of a realigned stack.
20684         (fp_valid_at): Likewise.
20685         (choose_baseaddr): Modify comments.
20686         (ix86_emit_outlined_ms2sysv_save): Adjust to changes in
20687         ix86_expand_prologue.
20688         (ix86_expand_prologue): Modify stack realignment and allocation.
20689         (ix86_expand_epilogue): Modify comments.
20690         * doc/sourcebuild.texi: Add documentation for target selectors avx2,
20691         avx2_runtime, avx512f, and avx512f_runtime.
20693 2017-08-23  Uros Bizjak  <ubizjak@gmail.com>
20695         * config/i386/i386.opt: Remove unneeded Init(0) initializations.
20696         (mstackrealign): Do not init to -1.
20697         * config/i386/i386.c (ix86_option_override_internal):
20698         Check opts_set, not opts when setting default value of
20699         opts->x_ix86_force_align_arg_pointer.
20701 2017-08-23  Richard Biener  <rguenther@suse.de>
20703         * function.c (fndecl_name): Use verbosity 1 (no arguments) for
20704         lang_hooks.decl_printable_name.
20705         * print-rtl-function.c (print_rtx_function): Likewise.
20706         * tree-pretty-print.c (dump_function_header): Likewise.
20708 2017-08-23  Richard Biener  <rguenther@suse.de>
20710         PR lto/81940
20711         * dwarf2out.c (dwarf2out_abstract_function): Handle LTO with
20712         -g0 at compile-time.
20714 2017-08-23  Tamar Christina  <tamar.christina@arm.com>
20716         PR middle-end/19706
20717         * doc/sourcebuild.texi (Other hardware attributes):
20718         Document xorsign.
20720 2017-08-23  Tamar Christina  <tamar.christina@arm.com>
20722         PR middle-end/19706
20723         * tree-ssa-math-opts.c (convert_expand_mult_copysign):
20724         Fix single-use check.
20726 2017-08-23  Thomas Preud'homme  <thomas.preudhomme@arm.com>
20728         * gcc.c (execute): Only test for SIGKILL and SIGQUIT if available.
20730 2017-08-22  Daniel Santos  <daniel.santos@pobox.com>
20732         * doc/install.texi: Modify to add more details on running selected
20733         tests.
20735 2017-08-22  Daniel Santos  <daniel.santos@pobox.com>
20737         * config/i386/i386.c (ix86_option_override_internal): Error when -mx32
20738         is combined with -mabi=ms.
20739         (ix86_function_type_abi): Limit errors for mixing -mx32 with attribute
20740         ms_abi.
20742 2017-08-22  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
20744         PR tree-optimization/81488
20745         * gimple-ssa-strength-reduction (struct slsr_cand_d): Add visited
20746         and cached_basis fields.
20747         (MAX_SPREAD): New constant.
20748         (alloc_cand_and_find_basis): Initialize new fields.
20749         (clear_visited): New function.
20750         (create_phi_basis_1): Rename from create_phi_basis, set visited
20751         and cached_basis fields.
20752         (create_phi_basis): New wrapper function.
20753         (phi_add_costs_1): Rename from phi_add_costs, add spread
20754         parameter, set visited field, short-circuit when limits reached.
20755         (phi_add_costs): New wrapper function.
20756         (record_phi_increments_1): Rename from record_phi_increments, set
20757         visited field.
20758         (record_phi_increments): New wrapper function.
20759         (phi_incr_cost_1): Rename from phi_incr_cost, set visited field.
20760         (phi_incr_cost): New wrapper function.
20761         (all_phi_incrs_profitable_1): Rename from
20762         all_phi_incrs_profitable, set visited field.
20763         (all_phi_incrs_profitable): New wrapper function.
20765 2017-08-22  Richard Sandiford  <richard.sandiford@linaro.org>
20766             Alan Hayward  <alan.hayward@arm.com>
20767             David Sherwood  <david.sherwood@arm.com>
20769         * rtl.h (paradoxical_subreg_p): Define inline, and add a version
20770         that takes the outer and inner modes.
20771         * doc/rtl.texi: Use paradoxical_subreg_p instead of a GET_MODE_SIZE
20772         comparison as the canonical test for a paradoxical subreg.
20773         * combine.c (simplify_set): Use paradoxical_subreg_p.
20774         (make_extraction): Likewise.
20775         (force_to_mode): Likewise.
20776         (rtx_equal_for_field_assignment_p): Likewise.
20777         (gen_lowpart_for_combine): Likewise.
20778         (simplify_comparison): Likewise.
20779         * cse.c (equiv_constant): Likewise.
20780         * expmed.c (store_bit_field_1): Likewise.
20781         * final.c (alter_subreg): Likewise.
20782         * fwprop.c (propagate_rtx): Likewise.
20783         (forward_propagate_subreg): Likewise.
20784         * ira-conflicts.c (ira_build_conflicts): Likewise.
20785         * lower-subreg.c (simplify_gen_subreg_concatn): Likewise.
20786         * lra-constraints.c (curr_insn_transform): Likewise.
20787         (split_reg): Likewise.
20788         * lra-eliminations.c (move_plus_up): Likewise.
20789         (lra_eliminate_regs_1): Likewise.
20790         * recog.c (general_operand): Likewise.
20791         * ree.c (combine_reaching_defs): Likewise.
20792         * reload.c (push_reload): Likewise.
20793         (find_reloads): Likewise.
20794         * reload1.c (elimination_effects): Likewise.
20795         (compute_reload_subreg_offset): Likewise.
20796         (choose_reload_regs): Likewise.
20797         * rtlanal.c (subreg_lsb_1): Likewise.
20798         * simplify-rtx.c (simplify_unary_operation_1): Likewise.
20799         (simplify_subreg): Likewise.
20800         * var-tracking.c (track_loc_p): Likewise.
20801         * emit-rtl.c (byte_lowpart_offset): Likewise.
20802         (paradoxical_subreg_p): Delete out-of-line definition.
20804 2017-08-22  Jeff Law  <law@redhat.com>
20806         PR tree-optimization/81741
20807         PR tree-optimization/71947
20808         * tree-ssa-dom.c: Include tree-inline.h.
20809         (record_temporary_equivalences): Only record SSA_NAME = SSA_NAME
20810         equivalences if one is more expensive to compute than the other.
20811         * tree-ssa-scopedtables.h (class const_or_copies): Make
20812         record_const_or_copy_raw method private.
20813         (class avail_exprs_stack): New method simplify_binary_operation.
20814         * tree-ssa-scopedtables.c (avail_exprs_stack::lookup_avail_expr): Call
20815         avail_exprs_stack::simplify_binary_operation as needed.
20816         (avail_exprs_stack::simplify_binary_operation): New function.
20818 2017-08-22  Sebastian Huber  <sebastian.huber@embedded-brains.de>
20820         * config.gcc (powerpc-*-rtems*): Add rs6000/linux64.opt.
20821         * config/rs6000/rtems.h (ASM_PREFERRED_EH_DATA_FORMAT): New define.
20822         (DOT_SYMBOLS): Likewise.
20823         (MINIMAL_TOC_SECTION_ASM_OP): Likewise.
20824         (RELOCATABLE_NEEDS_FIXUP): Likewise.
20825         (RS6000_ABI_NAME): Likewise.
20826         (TARGET_CMODEL): Likewise.
20827         (TOC_SECTION_ASM_OP): Likewise.
20828         (SET_CMODEL): New macro.
20829         (SUBSUBTARGET_OVERRIDE_OPTIONS): Evaluate cmodel options.
20831 2017-08-22  Richard Biener  <rguenther@suse.de>
20833         * tree-inline.c (remap_type_1): Change asserts on TYPE_SIZE[_UNIT]
20834         to allow for free-lang-data replacements similar to verify_type_variant.
20836 2017-08-22  Richard Sandiford  <richard.sandiford@linaro.org>
20837             Alan Hayward  <alan.hayward@arm.com>
20838             David Sherwood  <david.sherwood@arm.com>
20840         * config/aarch64/aarch64.md (casesi): Use DImode rather than
20841         VOIDmode for the LABEL_REF.
20843 2017-08-22  Richard Biener  <rguenther@suse.de>
20845         * tree-cfg.c (gimple_split_edge): Avoid reallocating target PHI nodes.
20847 2017-08-22  Richard Biener  <rguenther@suse.de>
20849         * common.opt (feliminate-dwarf2-dups): Ignore.
20850         * doc/invoke.texi (feliminate-dwarf2-dups): Remove documentation.
20851         * dwarf2out.c (push_new_compile_unit, pop_compile_unit,
20852         same_die_p_wrap, compute_section_prefix,
20853         is_symbol_die, assign_symbol_names, break_out_includes): Remove.
20854         (comdat_symbol_id, comdat_symbol_number): Likewise.
20855         (cu_hash_table_entry, cu_hash_table_entry_hasher, cu_hash_type):
20856         Likewise.
20857         (check_duplicate_cu, record_comdat_symbol_number): Likewise.
20858         (output_die): Mark unreachable path unreachable.
20859         (dwarf2out_start_source_file): Do not create DW_TAG_GNU_BINCL.
20860         (dwarf2out_end_source_file): Do not create DW_TAG_GNU_EINCL.
20861         (dwarf2out_init): Remove code handling flag_eliminate_dwarf2_dups.
20862         (dwarf2out_early_finish): Likewise.
20864 2017-08-22  Aldy Hernandez  <aldyh@redhat.com>
20866         * wide-int.h (hwi_with_prec::hwi_with_prec): Sign extend.
20868 2017-08-22  Georg-Johann Lay  <avr@gjlay.de>
20870         PR target/81910
20871         * config/avr/avr.c (avr_handle_addr_attribute): Early return if
20872         not VAR_P. Filter attribute warnings with OPT_Wattributes.
20873         (avr_attribute_table) <io, io_low, address>: Initialize
20874         .decl_required with true.
20876 2017-08-21  Michael Meissner  <meissner@linux.vnet.ibm.com>
20878         * config/rs6000/rs6000-cpus.def (-mvsx-scalar-float): Delete
20879         undocumented debugging options.
20880         (-mvsx-scalar-double): Likewise.
20881         (-mallow-df-permute): Likewise.
20882         (-mvectorize-builtins): Likewise.
20883         * config/rs6000/rs6000.c (rs6000_init_hard_regno_mode_ok): Likewise.
20884         (rs6000_builtin_vectorized_function): Likewise.
20885         (rs6000_builtin_md_vectorized_function): Likewise.
20886         (rs6000_opt_vars): Likewise.
20888 2017-08-21  Uros Bizjak  <ubizjak@gmail.com>
20890         PR target/46091
20891         * config/i386/i386.md (*btsq_imm): Rename from *btsq.
20892         (*btrq_imm): Rename from *btrq.
20893         (*btcq_imm): Rename from *btcq.
20894         (btsc): New code attribute.
20895         (*<btsc><mode>): New insn pattern.
20896         (*btr<mode>): Ditto.
20897         (*<btsc><mode>_mask): New insn_and_split pattern.
20898         (*btr<mode>_mask): Ditto.
20900 2017-08-21  Richard Sandiford  <richard.sandiford@linaro.org>
20901             Alan Hayward  <alan.hayward@arm.com>
20902             David Sherwood  <david.sherwood@arm.com>
20904         * function.c (pad_below): Simplify padding calculation.
20906 2017-08-21  Richard Sandiford  <richard.sandiford@linaro.org>
20907             Alan Hayward  <alan.hayward@arm.com>
20908             David Sherwood  <david.sherwood@arm.com>
20910         * target.def (function_prologue): Remove frame size argument.
20911         (function_epilogue): Likewise.
20912         * doc/tm.texi: Regenerate.
20913         * final.c (final_start_function): Update call to function_prologue.
20914         (final_end_function): Update call to function_epilogue.
20915         (default_function_pro_epilogue): Remove frame size argument.
20916         * output.h (default_function_pro_epilogue): Likewise.
20917         * config/arm/arm.c (arm_output_function_epilogue): Likewise.
20918         (arm_output_function_prologue): Likewise.
20919         * config/frv/frv.c (frv_function_prologue): Likewise.
20920         (frv_function_epilogue): Likewise.
20921         * config/i386/i386.c (ix86_output_function_epilogue): Likewise.
20922         * config/ia64/ia64.c (ia64_output_function_prologue): Likewise.
20923         (ia64_output_function_epilogue): Likewise.
20924         * config/m32r/m32r.c (m32r_output_function_prologue): Likewise.
20925         (m32r_output_function_epilogue): Likewise.
20926         * config/microblaze/microblaze.c (microblaze_function_prologue)
20927         (microblaze_function_epilogue): Likewise.
20928         * config/mips/mips.c (mips_output_function_prologue): Likewise.
20929         (mips_output_function_epilogue): Likewise.
20930         * config/mmix/mmix.c (mmix_target_asm_function_prologue): Likewise.
20931         (mmix_target_asm_function_epilogue): Likewise.
20932         * config/msp430/msp430.c (msp430_start_function): Likewise.
20933         * config/nds32/nds32.c (nds32_asm_function_prologue): Likewise.
20934         (nds32_asm_function_epilogue): Likewise.
20935         * config/nios2/nios2.c (nios2_asm_function_prologue): Likewise.
20936         * config/pa/pa.c (pa_output_function_prologue): Likewise.
20937         (pa_output_function_epilogue): Likewise.
20938         * config/powerpcspe/powerpcspe.c (rs6000_output_function_prologue)
20939         (rs6000_output_function_epilogue): Likewise.
20940         * config/rl78/rl78.c (rl78_start_function): Likewise.
20941         * config/rs6000/rs6000.c (rs6000_output_function_prologue): Likewise.
20942         (rs6000_output_function_epilogue): Likewise.
20943         * config/rx/rx.c (rx_output_function_prologue): Likewise.
20944         * config/sh/sh.c (sh_output_function_epilogue): Likewise.
20945         * config/sparc/sparc.c (sparc_asm_function_prologue): Likewise.
20946         (sparc_asm_function_epilogue): Likewise.
20948 2017-08-21  Richard Sandiford  <richard.sandiford@linaro.org>
20950         * tree.h (type_has_mode_precision_p): New function.
20951         * convert.c (convert_to_integer_1): Use it.
20952         * expr.c (expand_expr_real_2): Likewise.
20953         (expand_expr_real_1): Likewise.
20954         * fold-const.c (fold_single_bit_test_into_sign_test): Likewise.
20955         * match.pd: Likewise.
20956         * tree-ssa-forwprop.c (simplify_rotate): Likewise.
20957         * tree-ssa-math-opts.c (convert_mult_to_fma): Likewise.
20958         * tree-tailcall.c (process_assignment): Likewise.
20959         * tree-vect-loop.c (vectorizable_reduction): Likewise.
20960         * tree-vect-patterns.c (vect_recog_vector_vector_shift_pattern)
20961         (vect_recog_mult_pattern, vect_recog_divmod_pattern): Likewise.
20962         * tree-vect-stmts.c (vectorizable_conversion): Likewise.
20963         (vectorizable_assignment): Likewise.
20964         (vectorizable_shift): Likewise.
20965         (vectorizable_operation): Likewise.
20966         * tree-vrp.c (register_edge_assert_for_2): Likewise.
20968 2017-08-21  Wilco Dijkstra  <wdijkstr@arm.com>
20970         * match.pd: Add pow (C, x) simplification.
20972 2017-08-21  Richard Biener  <rguenther@suse.de>
20974         PR tree-optimization/81900
20975         * tree-ssa-pre.c (compute_antic_aux): Properly compute changed
20976         for blocks with abnormal predecessors.
20977         (compute_antic): Do not set visited flag prematurely.
20979 2017-08-21  Georg-Johann Lay  <avr@gjlay.de>
20981         PR target/79883
20982         * config/avr/avr.c (avr_set_current_function): Typo in diagnostic.
20984 2017-08-21  Richard Sandiford  <richard.sandiford@linaro.org>
20986         * stor-layout.h (vector_type_mode): Move to...
20987         * tree.h (vector_type_mode): ...here.
20988         * stor-layout.c (vector_type_mode): Move to...
20989         * tree.c (vector_type_mode): ...here.  Include rtl.h and regs.h.
20991 2017-08-21  Richard Biener  <rguenther@suse.de>
20993         * debug.h (struct gcc_debug_hooks): Add die_ref_for_decl and
20994         register_external_die hooks.
20995         (debug_false_tree_charstarstar_uhwistar): Declare.
20996         (debug_nothing_tree_charstar_uhwi): Likewise.
20997         * debug.c (do_nothing_debug_hooks): Adjust.
20998         (debug_false_tree_charstarstar_uhwistar): New do nothing.
20999         (debug_nothing_tree_charstar_uhwi): Likewise.
21000         * dbxout.c (dbx_debug_hooks): Adjust.
21001         (xcoff_debug_hooks): Likewise.
21002         * sdbout.c (sdb_debug_hooks): Likewise.
21003         * vmsdbgout.c (vmsdbg_debug_hooks): Likewise.
21004         * dwarf2out.c (macinfo_label_base): New global.
21005         (dwarf2out_register_external_die): New function for the
21006         register_external_die hook.
21007         (dwarf2out_die_ref_for_decl): Likewise for die_ref_for_decl.
21008         (dwarf2_debug_hooks): Use them.
21009         (dwarf2_lineno_debug_hooks): Adjust.
21010         (struct die_struct): Add with_offset flag.
21011         (DEBUG_LTO_DWO_INFO_SECTION, DEBUG_LTO_INFO_SECTION,
21012         DEBUG_LTO_DWO_ABBREV_SECTION, DEBUG_LTO_ABBREV_SECTION,
21013         DEBUG_LTO_DWO_MACINFO_SECTION, DEBUG_LTO_MACINFO_SECTION,
21014         DEBUG_LTO_DWO_MACRO_SECTION, DEBUG_LTO_MACRO_SECTION,
21015         DEBUG_LTO_LINE_SECTION, DEBUG_LTO_DWO_STR_OFFSETS_SECTION,
21016         DEBUG_LTO_STR_DWO_SECTION, DEBUG_STR_LTO_SECTION): New macros
21017         defining section names for the early LTO debug variants.
21018         (reset_indirect_string): New helper.
21019         (add_AT_external_die_ref): Helper for dwarf2out_register_external_die.
21020         (print_dw_val): Add support for offsetted symbol references.
21021         (get_ultimate_context): Split out from is_cxx.
21022         (is_cxx): Use get_ultimate_context.
21023         (is_fortran): Add decl overload.
21024         (compute_comp_unit_symbol): Split out worker from
21025         compute_section_prefix.
21026         (compute_section_prefix): Call compute_comp_unit_symbol and
21027         set comdat_type_p here.
21028         (output_die): Skip DIE symbol output for the LTO added one.
21029         Handle DIE symbol references with offset.
21030         (output_comp_unit): Guard section name mangling properly.
21031         For LTO debug sections emit a symbol at the section beginning
21032         which we use to refer to its DIEs.
21033         (add_abstract_origin_attribute): For DIEs registered via
21034         dwarf2out_register_external_die directly refer to the early
21035         DIE rather than indirectly through the shadow one we created.
21036         Remove obsolete call to dwarf2out_abstract_function for
21037         non-function/block origins.
21038         (gen_array_type_die): When generating early LTO debug do
21039         not emit DW_AT_string_length.
21040         (gen_formal_parameter_die): Do not re-create DIEs for PARM_DECLs
21041         late when in LTO.  As suggested place a gcc_unreachable for
21042         the DECL_ABSTRACT_P case.
21043         (gen_subprogram_die): Avoid another specification DIE
21044         for early built declarations/definitions for the late LTO case.
21045         (gen_variable_die): Add type references for late duplicated VLA dies
21046         when in late LTO.
21047         (gen_inlined_subroutine_die): Do not call dwarf2out_abstract_function,
21048         we have the abstract instance already.
21049         (process_scope_var): Adjust decl DIE contexts in LTO which
21050         first puts them in limbo.
21051         (gen_decl_die): Do not generate type DIEs late apart from
21052         types for VLAs or for decls we do not yet have a DIE.  Do not
21053         call dwarf2out_abstract_function late.
21054         (dwarf2out_early_global_decl): Make sure to create DIEs
21055         for abstract instances of a decl first.
21056         (dwarf2out_late_global_decl): Adjust comment.
21057         (output_macinfo_op): With multiple macro sections use
21058         macinfo_label_base to distinguish labels.
21059         (output_macinfo): Likewise.  Update macinfo_label_base.
21060         Pass in the line info label.
21061         (note_variable_value_in_expr): When generating LTO resolve
21062         all variable values here by generating DIEs as needed.
21063         (init_sections_and_labels): Add early LTO debug flag parameter
21064         and generate different sections and names if set.  Add generation
21065         counter for the labels so we can have multiple of them.
21066         (reset_dies): Helper to allow DIEs to be output multiple times.
21067         (dwarf2out_finish): When outputting DIEs to the fat part of an
21068         LTO object first reset DIEs.
21069         (dwarf2out_early_finish): Output early DIEs when generating LTO.
21070         (modified_type_die): Check for decl_ultimate_origin being self
21071         before recursing.
21072         (gen_type_die_with_usage): Likewise.
21073         (gen_typedef_die): Allow decl_ultimate_origin being self.
21074         (set_decl_abstract_flags): Remove.
21075         (set_block_abstract_flags): Likewise.
21076         (dwarf2out_abstract_function): Treat the early generated DIEs
21077         as the abstract copy and only add DW_AT_inline and
21078         DW_AT_artificial here and call set_decl_origin_self.
21079         If the DIE has an abstract origin don't do anything.
21080         * tree.c (free_lang_data): Build a dummy TRANSLATION_UNIT_DECL
21081         if we have none yet (Go fails to build one, PR78628).
21082         (variably_modified_type_p): Prevent endless recursion for Ada
21083         cyclic pointer types.
21084         * lto-streamer-in.c: Include debug.h.
21085         (dref_queue): New global.
21086         (lto_read_tree_1): Stream in DIE references.
21087         (lto_input_tree): Register DIE references.
21088         (input_function): Stream DECL_DEBUG_ARGS.
21089         * lto-streamer-out.c: Include debug.h.
21090         (lto_write_tree_1): Output DIE references.
21091         (DFS::DFS_write_tree_body): Follow DECL_ABSTRACT_ORIGIN.
21092         Force a TRANSLATION_UNIT_DECL DECL_CONTEXT for file-scope decls.
21093         (output_function): Stream DECL_DEBUG_ARGS.
21094         * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
21095         Stream DECL_ABSTRACT_ORIGIN.
21096         * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
21097         (write_ts_decl_minimal_tree_pointers): Force a TRANSLATION_UNIT_DECL
21098         DECL_CONTEXT for file-scope decls.
21099         * lto-streamer.h (struct dref_entry): Declare.
21100         (dref_queue): Likewise.
21101         * cfgexpand.c (pass_expand::execute): Do not call the
21102         outlining_inline_function hook here.
21103         * lto-wrapper.c (debug_obj): New global.
21104         (tool_cleanup): Unlink it if required.
21105         (debug_objcopy): New function.
21106         (run_gcc): Handle early debug sections in the IL files by
21107         extracting them to separate files, partially linkin them and
21108         feeding the result back as result to the linker.
21109         * config/darwin.h (DEBUG_LTO_INFO_SECTION, DEBUG_LTO_ABBREV_SECTION,
21110         DEBUG_LTO_MACINFO_SECTION, DEBUG_LTO_LINE_SECTION,
21111         DEBUG_STR_LTO_SECTION, DEBUG_LTO_MACRO_SECTION): Put early debug
21112         sections into a separate segment.
21113         * config/darwin.c (darwin_asm_named_section): Handle __GNU_DWARF_LTO
21114         segments.
21115         (darwin_asm_dwarf_section): Likewise.
21116         (darwin_asm_output_dwarf_offset): Likewise.
21117         * config/i386/i386.c (make_resolver_func): Set DECL_IGNORED_P.
21119 2017-08-21  Richard Sandiford  <richard.sandiford@linaro.org>
21120             Alan Hayward  <alan.hayward@arm.com>
21121             David Sherwood  <david.sherwood@arm.com>
21123         * read-md.h (md_reader::record_potential_iterator_use): Replace
21124         pointer argument with an rtx and an index.
21125         * read-rtl.c (iterator_group::apply_iterator): Likewise.
21126         (apply_mode_iterator): Likewise.
21127         (apply_code_iterator): Likewise.
21128         (apply_int_iterator): Likewise.
21129         (apply_subst_iterator): Likewise.
21130         (record_iterator_use): Likewise.
21131         (record_attribute_use): Likewise.
21132         (md_reader::record_potential_iterator_use): Likewise.  Update calls
21133         to record_iterator_use and apply_iterator.
21134         (iterator_use): Replace ptr with x and index.
21135         (attribute_use): Likewise.
21136         (apply_attribute_uses): Update calls to apply_iterator.
21137         (apply_iterators): Likewise.  Update initialization of iterator_use.
21138         (rtx_reader::read_rtx_code): Update calls to record_iterator_use
21139         and record_potential_iterator_use.
21140         (rtx_reader::read_rtx_operand): Likewise.
21142 2017-08-21  Richard Sandiford  <richard.sandiford@linaro.org>
21143             Alan Hayward  <alan.hayward@arm.com>
21144             David Sherwood  <david.sherwood@arm.com>
21146         * varasm.c (const_rtx_hash_1): Don't hash in the mode of a
21147         CONST_WIDE_INT.
21149 2017-08-21  Richard Biener  <rguenther@suse.de>
21151         PR middle-end/81884
21152         * tree-ssa-alias.c (stmt_kills_ref_p): Handle array accesses
21153         at struct end conservatively when comparing common bases.
21155 2017-08-21  Richard Biener  <rguenther@suse.de>
21157         * tree-ssa-loop-im.c (struct lim_aux_data): Add ref index member.
21158         (mem_ref_in_stmt): Remove.
21159         (determine_max_movement): Use ref index to get at the reference.
21160         (invariantness_dom_walker::before_dom_children): Deal with
21161         lim data already initialized.
21162         (gather_mem_refs_stmt): Initialize lim data and record ref index.
21164 2017-08-19  Uros Bizjak  <ubizjak@gmail.com>
21166         * config/i386/i386.h (OPTION_MASK_ISA_ROUND): Remove.
21167         (TARGET_ISA_ROUND): Ditto.
21168         (TARGET_ROUND): Ditto.
21169         * config/i386/i386.c: Substitute TARGET_ROUND with TARGET_SSE4_1.
21170         * config/i386/i386.md: Ditto.
21171         * config/i386/sse.md: Ditto.
21172         * config/i386/i386-builtin.def: Substitute OPTION_MASK_ISA_ROUND
21173         with OPTION_MASK_ISA_SSE4_1.
21175 2017-08-19  Uros Bizjak  <ubizjak@gmail.com>
21177         PR target/81894
21178         * doc/extend.texi (x86 Built-in Functions): Correct the name of
21179         __builtin_ia32_lzcnt_u16.
21181 2017-08-18  Peter Bergner  <bergner@vnet.ibm.com>
21183         PR target/80210
21184         * config/rs6000/rs6000.c (rs6000_activate_target_options): New function.
21185         (rs6000_set_current_function): Rewrite function to use it.
21187 2017-08-18  H.J. Lu  <hongjiu.lu@intel.com>
21189         PR c/53037
21190         * print-tree.c (print_node): Support DECL_WARN_IF_NOT_ALIGN
21191         and TYPE_WARN_IF_NOT_ALIGN.
21192         * stor-layout.c (do_type_align): Merge DECL_WARN_IF_NOT_ALIGN.
21193         (handle_warn_if_not_align): New.
21194         (place_union_field): Call handle_warn_if_not_align.
21195         (place_field): Call handle_warn_if_not_align.
21196         Copy TYPE_WARN_IF_NOT_ALIGN.
21197         (finish_builtin_struct): Copy TYPE_WARN_IF_NOT_ALIGN.
21198         (layout_type): Likewise.
21199         * tree-core.h (tree_type_common): Add warn_if_not_align.  Set
21200         spare to 18.
21201         (tree_decl_common): Add warn_if_not_align.
21202         * tree.c (build_range_type_1): Copy TYPE_WARN_IF_NOT_ALIGN.
21203         * tree.h (TYPE_WARN_IF_NOT_ALIGN): New.
21204         (SET_TYPE_WARN_IF_NOT_ALIGN): Likewise.
21205         (DECL_WARN_IF_NOT_ALIGN): Likewise.
21206         (SET_DECL_WARN_IF_NOT_ALIGN): Likewise.
21207         * doc/extend.texi: Document warn_if_not_aligned attribute.
21208         * doc/invoke.texi: Document -Wif-not-aligned and -Wpacked-not-aligned.
21210 2017-08-17  Martin Liska  <mliska@suse.cz>
21212         PR bootstrap/81864
21213         * tree-loop-distribution.c (ddrs_table): Change type to pointer type.
21214         (get_data_dependence): Use it as pointer type.
21215         (distribute_loop): Likewise.
21217 2017-08-17  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
21219         * config/rs6000/altivec.md (UNSPEC_VMRGOW_DIRECT): New constant.
21220         (p8_vmrgew_v4sf_direct): Generalize to p8_vmrgew_<mode>_direct.
21221         (p8_vmrgow_<mode>_direct): New define_insn.
21222         * config/rs6000/rs6000.c (altivec_expand_vec_perm_const): Properly
21223         handle endianness for vmrgew and vmrgow permute patterns.
21225 2017-08-17  Peter Bergner  <bergner@vnet.ibm.com>
21227         * config/rs6000/altivec.md (VParity): Remove TARGET_VSX_TIMODE.
21228         * config/rs6000/rs6000-cpus.def: Remove comment.
21229         (ISA_2_7_MASKS_SERVER): Delete OPTION_MASK_VSX_TIMODE;
21230         (POWERPC_MASKS): Likewise.
21231         * config/rs6000/rs6000.c (rs6000_hard_regno_mode_ok): Remove unneeded
21232         use of TARGET_VSX_TIMODE.
21233         (rs6000_setup_reg_addr_masks): Change TARGET_VSX_TIMODE to TARGET_VSX.
21234         (rs6000_init_hard_regno_mode_ok): Remove unneeded uses of
21235         TARGET_VSX_TIMODE.  Change use of TARGET_VSX_TIMODE to TARGET_VSX.
21236         (rs6000_option_override_internal): Remove dead code.
21237         (rs6000_legitimize_address): Change TARGET_VSX_TIMODE to TARGET_VSX.
21238         (rs6000_legitimize_reload_address): Likewise.
21239         (rs6000_legitimate_address_p): Likewise.
21240         (rs6000_opt_masks): Delete "vsx-timode".
21241         (rs6000_disable_incompatible_switches): Remove mention of -mvsx-timode
21242         from function comment.
21243         * config/rs6000/rs6000.h (MASK_VSX_TIMODE): Delete.
21244         * config/rs6000/rs6000.md (FMOVE128_GPR): Remove TARGET_VSX_TIMODE.
21245         (V16QI, V8HI, V4SI, V4SF, V2DI, V2DF, V1TI): Remove useless empty
21246         condition.
21247         * config/rs6000/rs6000.opt (mvsx-timode): Replace with stub.
21248         * config/rs6000/vector.md (VEC_IP): Remove TARGET_VSX_TIMODE.
21249         * config/rs6000/vsx.md (VSX_LE_128): Likewise.
21250         (VSX_TI): Likewise.
21251         (VSX_M): Likewise.
21252         (define_peephole2): Likewise.
21254 2017-08-17  Martin Sebor  <msebor@redhat.com>
21256         PR c/81859
21257         * pretty-print.c (pp_format): Use strnlen in %.*s to avoid reading
21258         past the end of an array.
21259         (test_pp_format): Add test cases.
21261 2017-08-17  Richard Sandiford  <richard.sandiford@linaro.org>
21263         * internal-fn.def (CLRSB, CLZ, CTZ, FFS, PARITY, POPCOUNT): Add
21264         missing ECF_NOTHROW flags.
21266 2017-08-17  Peter Bergner  <bergner@vnet.ibm.com>
21268         PR target/72804
21269         * config/rs6000/vsx.md (*vsx_le_permute_<mode>): Add support for
21270         operands residing in integer registers.
21271         (*vsx_le_perm_load_<mode>): Likewise.
21272         (*vsx_le_perm_store_<mode>): Likewise.
21273         (define_peephole2): Add peepholes to optimize the above.
21275 2017-08-17  Marek Polacek  <polacek@redhat.com>
21277         PR middle-end/81814
21278         * fold-const.c (operand_equal_for_comparison_p): Remove code that used
21279         to mimic what shorten_compare did.  Change the return type to bool.
21280         (fold_cond_expr_with_comparison): Update call to
21281         operand_equal_for_comparison_p.
21282         (fold_ternary_loc): Likewise.
21284 2017-08-17  Jackson Woodruff  <jackson.woodruff@arm.com>
21286         * aarch64-simd.md (mov<mode>): No longer force zero immediate into
21287         register.
21288         (*aarch64_simd_mov<mode>): Add new case for stp using zero immediate.
21290 2017-08-17  Richard Biener  <rguenther@suse.de>
21292         * tree-ssa-structalias.c (solve_graph): When propagating
21293         to successors update the graphs succ edges and avoid duplicate work.
21295 2017-08-17  Maxim Ostapenko  <m.ostapenko@samsung.com>
21297         PR target/81861
21298         * config/i386/i386.c (ix86_option_override_internal): Save target
21299         specific options after ix86_stack_protector_guard_reg was changed.
21301 2017-08-17  Richard Biener  <rguenther@suse.de>
21303         PR tree-optimization/81827
21304         * tree-ssa-structalias.c (struct variable_info): Add is_reg_var flag.
21305         (new_var_info): Initialize it conservatively.
21306         (get_call_vi): Mark register vars.
21307         (new_scalar_tmp_constraint_exp): Likewise.
21308         (handle_rhs_call): Likewise.
21309         (handle_const_call): Likewise.
21310         (create_function_info_for): Likewise.
21311         (solve_constraints): Sort varinfos to separate register from
21312         non-register vars to pack points-to solution bitmaps during
21313         iteration.
21315 2017-08-17  Marek Polacek  <polacek@redhat.com>
21317         * gimplify.c (gimplify_adjust_omp_clauses): Compare with 0 instead of 1.
21319 2017-08-17  Richard Biener  <rguenther@suse.de>
21321         * tree-vrp.c (vrp_int_const_binop): Do not set *overflow_p
21322         to true when overflow is undefined and we saturated the result.
21324 2017-08-17  Alan Modra  <amodra@gmail.com>
21326         PR target/80938
21327         * config/rs6000/rs6000.c (rs6000_savres_strategy): Revert 2017-08-09.
21328         Don't use store multiple if only one reg needs saving.
21329         (interesting_frame_related_regno): New function.
21330         (rs6000_frame_related): Don't emit frame info for regs that
21331         don't need saving.
21332         (rs6000_emit_epilogue): Likewise.
21334 2017-08-16  Nathan Sidwell  <nathan@acm.org>
21336         * tree-core.h (tree_type_non_common): Rename binfo to lang_1.
21337         * tree.h (TYPE_BINFO): Use type_non_common.maxval.
21338         (TYPE_LANG_SLOT_1): Use type_non_common.lang_1, for any type.
21339         * tree.c (free_lang_data_in_type): Use else-if chain.  Always
21340         clear TYPE_LANG_1.  Remove obsolete member-function stripping.
21341         (find_decls_types_r): Comment about TYPE_MAX_VALUES_RAW.
21342         (verify_type): Adjust for TYPE_BINFO move.
21343         * lto-streamer-out.c (DFS::DFS_write_tree_body): No need to
21344         process TYPE_BINFO directly.
21345         (hash_tree): Likewise.
21346         * tree-streamer-in.c (lto_input_ts_type_non_common_tree_pointers):
21347         Likewise.
21348         * tree-streamer-out.c (write_ts_type_non_common_tree_pointers):
21349         Likewise.
21351 2017-08-16  David Malcolm  <dmalcolm@redhat.com>
21353         * diagnostic-show-locus.c (colorizer::m_caret): Remove unused field.
21355 2017-08-16  Uros Bizjak  <ubizjak@gmail.com>
21357         PR target/46091
21358         * config/i386/i386.md (*anddi_1_btr): Change predicates of
21359         operand 0 and operand 1 to nomimmediate_operand. Add "m" constraint.
21360         Add ix86_binary_operator_ok to insn constraint.
21361         (*iordi_1_bts): Ditto.
21362         (*xordi_1_btc): Ditto.
21363         (*btsq): Change predicate of operand 0 to nonimmediate_operand.
21364         Update corresponding peephole2 pattern.
21365         (*btrq): Ditto.
21366         (*btcq): Ditto.
21368 2017-08-16  Bin Cheng  <bin.cheng@arm.com>
21370         PR tree-optimization/81832
21371         * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): Don't
21372         copy loop header which has IFN_LOOP_DIST_ALIAS call.
21374 2017-08-16  Marek Polacek  <polacek@redhat.com>
21376         PR middle/81695
21377         * fold-const.c (fold_indirect_ref_1): Restore original behavior
21378         regarding size_zero_node.
21380 2017-08-16  Martin Liska  <mliska@suse.cz>
21382         PR target/81753
21383         * config.gcc: Respect previously set extra_objs in case
21384         of darwin target.
21386 2017-08-16  Richard Sandiford  <richard.sandiford@linaro.org>
21388         PR tree-optimization/81835
21389         * tree-vect-loop.c (vect_is_simple_reduction): Simply checks for
21390         the phi SSA_NAME.  Check that the condition in a COND_EXPR does
21391         not depend on the phi.
21393 2017-08-16  Alan Modra  <amodra@gmail.com>
21395         * config/rs6000/rs6000.c (rs6000_init_hard_regno_mode_ok): Delete
21396         dead code.
21398 2017-08-16  Alan Modra  <amodra@gmail.com>
21400         * config/rs6000/rs6000.c (rs6000_reg_live_or_pic_offset_p): Merge..
21401         (save_reg_p): ..into this.  Update all callers.
21402         (first_reg_to_save): Simplify.
21404 2017-08-16  Alan Modra  <amodra@gmail.com>
21406         * config/rs6000/rs6000.c (rs6000_savres_strategy): Don't restore
21407         fixed regs.
21409 2017-08-15  Joseph Myers  <joseph@codesourcery.com>
21411         PR target/78460
21412         PR target/67712
21413         * config/sh/sh-mem.cc (sh_expand_cmpnstr): Only unroll for
21414         constant count if that count is less than 32.
21416 2017-08-15  Nathan Sidwell  <nathan@acm.org>
21418         * gcc.c (execute): Emit friendlier message if inferior is killed
21419         by an external cause.
21421 2017-08-15  Richard Biener  <rguenther@suse.de>
21423         PR tree-optimization/81790
21424         * tree-ssa-sccvn.c (vn_lookup_simplify_result): Handle both
21425         CONSTRUCTORs from simplifying and VN.
21427 2017-08-14  Martin Sebor  <msebor@redhat.com>
21429         * builtin-attrs.def: Add comments.
21431 2017-08-14  Martin Sebor  <msebor@redhat.com>
21433         PR c/81117
21434         * doc/extend.texi (attribute nonstring): Document new attribute.
21436 2017-08-14  Martin Sebor  <msebor@redhat.com>
21438         PR c/81117
21439         * tree-diagnostic.c (default_tree_printer): Handle %G.
21440         * gimple-pretty-print.h (percent_G_format): Declare new function.
21441         * gimple-pretty-print.c (percent_G_format): Define.
21442         * tree-pretty-print.c (percent_K_format): Add argument.
21444 2017-08-14  Martin Sebor  <msebor@redhat.com>
21446         PR translation/79998
21447         * gimple-ssa-sprintf.c (pass_sprintf_length::handle_gimple_call):
21448         Remove a stray space.
21450 2017-08-14  Uros Bizjak  <ubizjak@gmail.com>
21452         PR target/46091
21453         * config/i386/i386.md (*anddi_1_btr): New insn_and_split pattern.
21454         (*iordi_1_bts): Ditto.
21455         (*xordi_1_btc): Ditto.
21457 2017-08-14  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
21459         PR target/79845
21460         * config/rs6000/linux64.h (INVALID_64BIT): Use quoted strings.
21461         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
21462         Likewise.
21463         * config/rs6000/rs6000.c (rs6000_init_hard_regno_mode_ok): Use
21464         quoted strings, and make more translator-friendly.
21465         (darwin_rs6000_override_options): Likewise.
21466         (rs6000_option_override_internal): Likewise.
21467         (rs6000_return_in_memory): Fix overlong line.
21468         (init_cmulative_args): Use quoted strings, and make more
21469         translator-friendly.
21470         (rs6000_pass_by_reference): Fix overlong line.
21471         (def_builtin): Use quoted strings.
21472         (altivec_expand_predicate_builtin): Use quoted strings, and make
21473         more translator-friendly.
21474         (htm_expand_builtin): Use quoted strings.
21475         (cpu_expand_builtin): Use quoted strings, and make more
21476         translator-friendly.
21477         (altivec_expand_builtin): Likewise.
21478         (paired_expand_predicate_builtin): Likewise.
21479         (rs6000_invalid_builtin): Likewise.
21480         (builtin_function_type): Use quoted strings.
21481         (rs6000_expand_split_stack_prologue): Use quoted strings, and make
21482         more translator-friendly.
21483         (rs6000_trampoline_init): Likewise.
21484         (rs6000_handle_altivec_attribute): Likewise.
21485         (rs6000_inner_target_options): Use quoted strings.
21486         (rs6000_disable_incompatible_switches): Likewise.
21487         * config/rs6000/sysv4.h (SUBTARGET_OVERRIDE_OPTIONS): Use quoted
21488         strings, and make more translator-friendly.
21489         (SUBSUBTARGET_OVERRIDE_OPTIONS): Use quoted strings.
21491 2017-08-14  Bin Cheng  <bin.cheng@arm.com>
21493         PR tree-optimization/81799
21494         * tree-loop-distribution.c (version_loop_by_alias_check): Force
21495         cond_expr to simple gimple operand.
21497 2017-08-14  Wilco Dijkstra  <wdijkstr@arm.com>
21499         PR middle-end/46932
21500         * doc/sourcebuild.texi (autoincdec): Add autoincdec description.
21502 2017-08-14  Georg-Johann Lay  <avr@gjlay.de>
21504         PR target/81754
21505         PR target/81268
21506         * config/avr/avr.opt (mgas-isr-prologues): New Var avr_gasisr_prologues.
21507         * config/avr/avr.md (gasisr, *gasisr): Use it instead of
21508         TARGET_GASISR_PROLOGUES.
21509         * config/avr/avr.c (avr_option_override): Same.
21510         (avr_pass_pre_proep::execute): Same.
21512 2017-08-13  H.J. Lu  <hongjiu.lu@intel.com>
21514         PR target/81820
21515         * config/i386/i386.c (ix86_finalize_stack_frame_flags): Replace
21516         frame pointer with stack pointer - UNITS_PER_WORD in debug insns.
21518 2017-08-13  Uros Bizjak  <ubizjak@gmail.com>
21520         * config/i386/i386.md (*load_tp_<mode>): Redefine as
21521         define_insn_and_split.  Split to a memory load from 0 in
21522         DEFAULT_TLS_SEG_REG address space.  Merge with *load_tp_x32
21523         using PTR mode iterator.
21524         (*load_tp_x32_zext"): Redefine as define_insn_and_split.
21525         Split to a memory load from 0 in DEFAULT_TLS_SEG_REG address space.
21526         (*add_tp_<mode>): Redefine as define_insn_and_split.
21527         Split to an add with a memory load from 0 in DEFAULT_TLS_SEG_REG
21528         address space.  Merge with *add_tp_x32 using PTR mode iterator.
21529         (*add_tp_x32_zext"): Redefine as define_insn_and_split.
21530         Split to an add with a  memory load from 0 in
21531         DEFAULT_TLS_SEG_REG address space.
21533 2017-08-12  Andrew Pinski  <apinski@cavium.com>
21535         * config/aarch64/aarch64-option-extensions.def (rdma):
21536         Fix feature string to what Linux prints out in /proc/cpuinfo.
21538 2017-08-12  Pierre-Marie de Rodat  <derodat@adacore.com>
21540         PR ada/79542
21541         * dwarf2out.c (modified_type_die): For C typedef types that have
21542         an ultimate origin, process the ultimate origin instead of the
21543         input type.
21544         (gen_typedef_die): Assert that input DECLs have no ultimate
21545         origin.
21546         (gen_type_die_with_usage): For typedef variants that have an
21547         ultimate origin, just call gen_decl_die on the original DECL.
21548         (process_scope_var): Avoid creating DIEs for local typedefs and
21549         concrete static variables.
21551 2017-08-12  Alan Modra  <amodra@gmail.com>
21553         PR target/81170
21554         PR target/81295
21555         * config/rs6000/sysv4.h (STARTFILE_LINUX_SPEC): Upgrade to
21556         match gnu-user.h startfile.
21557         (ENDFILE_LINUX_SPEC): Similarly.
21559 2017-08-11  Thomas Schwinge  <thomas@codesourcery.com>
21561         PR lto/81430
21562         * config/nvptx/nvptx.c (nvptx_override_options_after_change):
21563         Remove function.
21564         (TARGET_OVERRIDE_OPTIONS_AFTER_CHANGE): Remove definition.
21566 2017-08-11  Tamar Christina  <tamar.christina@arm.com>
21567         * config/aarch64/aarch64.md (mov<mode>): Change.
21568         (*movhf_aarch64, *movsf_aarch64, *movdf_aarch64):
21569         aarch64_reg_or_fp_float into aarch64_reg_or_fp_zero.
21570         * config/aarch64/predicates.md (aarch64_reg_or_fp_float): Removed.
21572 2017-08-11  Eric Botcazou  <ebotcazou@adacore.com>
21574         * tree-sra.c (build_access_from_expr_1): Use more precise diagnostics
21575         for storage order barriers.
21577 2017-08-11  Martin Liska  <mliska@suse.cz>
21579         PR tree-opt/79987
21580         * tree-chkp.c (chkp_get_bounds_for_decl_addr): Do not instrument
21581         variables of void type.
21583 2017-08-11  Martin Liska  <mliska@suse.cz>
21585         * asan.c (asan_protect_global): Replace ASM_OUTPUT_DEF with
21586         TARGET_SUPPORTS_ALIASES.
21587         * cgraph.c (cgraph_node::create_same_body_alias): Likewise.
21588         * ipa-visibility.c (can_replace_by_local_alias): Likewise.
21589         (optimize_weakref): Likewise.
21590         * symtab.c (symtab_node::noninterposable_alias): Likewise.
21591         * varpool.c (varpool_node::create_extra_name_alias): Likewise.
21592         * defaults.h: Introduce TARGET_SUPPORTS_ALIASES.
21594 2017-08-11  Martin Liska  <mliska@suse.cz>
21596         PR ipa/81213
21597         * config/i386/i386.c (make_resolver_func): Do complete
21598         refactoring of the function.
21600 2017-08-10  Uros Bizjak  <ubizjak@gmail.com>
21602         PR target/81708
21603         * config/i386/i386.opt (mstack-protector-guard-symbol=): New option
21604         * config/i386/i386.c (ix86_stack_protect_guard): Use
21605         ix86_stack_protect_guard_symbol_str to generate varible declaration.
21606         * doc/invoke.texi (x86 Options): Document
21607         -mstack-protector-guard-symbol= option.
21609 2017-08-10  Uros Bizjak  <ubizjak@gmail.com>
21611         * config/i386/i386-protos.h (ix86_split_stack_guard): New prototype.
21612         * config/i386/i386.c (ix86_split_stack_guard): New function.
21613         (ix86_xpand_split_stack_prologue): Call ix86_split_stack_guard.
21614         (ix86_legitimate_address_p) <case UNSPEC_STACK_CHECK>: Remove.
21615         (i386_asm_output_addr_const_extra) <case UNSPEC_STACK_CHECK>: Ditto.
21616         (optput_pic_addr_const): Remove UNSPEC_STACK_CHECK handling.
21617         * config/i386/i386.md (unspec): Remove UNSPEC_STACK_CHECK.
21618         (split_stack_space_check): Call ix86_split_stack_guard.
21620 2017-08-10  Martin Sebor  <msebor@redhat.com>
21622         * print-tree.c (print_node): Print location using the established
21623         format %s:%i%i.
21624         Replace spaces with colons.
21625         (debug_raw, debug): Ditto.
21627 2017-08-10  Martin Sebor  <msebor@redhat.com>
21629         PR c++/81586
21630         * pretty-print.c (pp_format): Correct the handling of %s precision.
21632 2017-08-10  H.J. Lu  <hongjiu.lu@intel.com>
21634         PR target/81736
21635         * config/i386/i386.c (ix86_finalize_stack_realign_flags): Renamed
21636         to ...
21637         (ix86_finalize_stack_frame_flags): This.  Also clear
21638         frame_pointer_needed if -fno-omit-frame-pointer is used without
21639         stack access.
21640         (ix86_expand_prologue): Replace ix86_finalize_stack_realign_flags
21641         with ix86_finalize_stack_frame_flags.
21642         (ix86_expand_epilogue): Likewise.
21643         (ix86_expand_split_stack_prologue): Likewise.
21644         * doc/invoke.texi: Add a note for -fno-omit-frame-pointer.
21646 2017-08-10  Martin Liska  <mliska@suse.cz>
21648         PR c++/81355
21649         * c-attribs.c (handle_target_attribute):
21650         Report warning for an empty string argument of target attribute.
21652 2017-08-09  Jakub Jelinek  <jakub@redhat.com>
21654         PR c/81687
21655         * omp-low.c (omp_copy_decl): Don't remap FORCED_LABEL or DECL_NONLOCAL
21656         LABEL_DECLs.
21657         * tree-cfg.c (move_stmt_op): Don't adjust DECL_CONTEXT of FORCED_LABEL
21658         or DECL_NONLOCAL labels.
21659         (move_stmt_r) <case GIMPLE_LABEL>: Adjust DECL_CONTEXT of FORCED_LABEL
21660         or DECL_NONLOCAL labels here.
21662 2017-08-09  Will Schmidt  <will_schmidt@vnet.ibm.com>
21664         * config/rs6000/rs6000.c (rs6000_option_override_internal): Add blurb
21665         to indicate when early gimple folding has been disabled.
21666         (rs6000_gimple_fold_builtin): Add debug content.
21667         (rs6000_invalid_builtin): Fix whitespace.
21668         (rs6000_expand_builtin): Fix whitespace.
21669         * config/rs6000/rs6000.opt: Add option for -mfold-gimple.
21671 2017-08-09  Segher Boessenkool  <segher@kernel.crashing.org>
21673         PR target/80938
21674         * config/rs6000/rs6000.c (rs6000_savres_strategy): Don't use
21675         SAVE_MULTIPLE if not all the registers that saves, should be saved.
21677 2017-08-09  Jim Wilson  <jim.wilson@linaro.org>
21679         * config/aarch64/aarch64-cores.def (falkor): Use falkor pipeline.
21680         (qdf24xx): Likewise.
21681         * config/aarch64/aarch64.md: Include falkor.md.
21682         * config/aarch64/falkor.md: New.
21684 2017-08-09  Marek Polacek  <polacek@redhat.com>
21686         PR c/81233
21687         * diagnostic-core.h (emit_diagnostic_valist): Add declaration.
21688         * diagnostic.c (emit_diagnostic): Add a comment.
21689         (emit_diagnostic_valist): New function.
21691 2017-08-09  Marek Polacek  <polacek@redhat.com>
21693         PR c/81417
21694         * input.c (make_location): New overload.
21695         * input.h (make_location): Declare.
21697 2017-08-08  Alan Modra  <amodra@gmail.com>
21698             H.J. Lu  <hongjiu.lu@intel.com>
21700         PR driver/81523
21701         * gcc.c (NO_PIE_SPEC): Delete.
21702         (PIE_SPEC): Define as !no-pie/pie.  Move static|shared|r
21703         exclusion..
21704         (LINK_PIE_SPEC): ..to here.
21705         (LINK_COMMAND_SPEC): Support -no-pie.
21706         * config/gnu-user.h (GNU_USER_TARGET_STARTFILE_SPEC): Correct
21707         chain of crtbegin*.o selection, update for PIE_SPEC changes and
21708         format.
21709         (GNU_USER_TARGET_ENDFILE_SPEC): Similarly.
21710         * config/sol2.h (STARTFILE_CRTBEGIN_SPEC): Similarly.
21711         (ENDFILE_CRTEND_SPEC): Similarly.
21713 2017-08-08  Uros Bizjak  <ubizjak@gmail.com>
21715         PR target/81708
21716         * config/i386/i386.opt (mstack-protector-guard-reg=): New option
21717         (mstack-protector-guard-offset=): Ditto.
21718         * config/i386/i386.c (ix86_option_override): Handle
21719         -mstack-protector-guard-reg= and -mstack-protector-guard-offset=
21720         options.
21721         (ix86_stack_protect_guard): Use ix86_stack_protect_guard_reg and
21722         ix86_stack_protect_guard_offset variables.
21723         (TARGET_STACK_PROTECT_GUARD): Always define.
21724         * doc/invoke.texi (x86 Options): Document -mstack-protector-guard-reg=
21725         and -mstack-protector-guard-offset= options.
21727 2017-08-08  Bin Cheng  <bin.cheng@arm.com>
21729         * tree-ssa-loop-ivopts.c (relate_compare_use_with_all_cands): Handle
21730         boundary case for the last candidate.
21732 2017-08-08  Bin Cheng  <bin.cheng@arm.com>
21734         * doc/invoke.texi: Document -ftree-loop-distribution for O3.
21735         * opts.c (default_options_table): Add OPT_ftree_loop_distribution.
21737 2017-08-08  Tamar Christina  <tamar.christina@arm.com>
21739         PR middle-end/19706
21740         * config/aarch64/aarch64.md (xorsign<mode>3): New optabs.
21741         * config/aarch64/aarch64-builtins.c
21742         (aarch64_builtin_vectorized_function): Added CASE_CFN_XORSIGN.
21743         * config/aarch64/aarch64-simd-builtins.def: Added xorsign BINOP.
21744         * config/aarch64/aarch64-simd.md: Added xorsign<mode>3
21746 2017-08-08  Tamar Christina  <tamar.christina@arm.com>
21747             Andrew Pinski <pinskia@gmail.com>
21749         PR middle-end/19706
21750         * internal-fn.def (XORSIGN): New.
21751         * optabs.def (xorsign_optab): New.
21752         * tree-ssa-math-opts.c (is_copysign_call_with_1): New.
21753         (convert_expand_mult_copysign): New.
21754         (pass_optimize_widening_mul::execute): Call
21755         convert_expand_mult_copysign.
21757 2017-08-08  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
21759         PR tree-optimization/81354
21760         * gimple-ssa-strength-reduction.c (create_add_on_incoming_edge):
21761         Insert on edges rather than explicitly creating landing pads.
21762         (analyze_candidates_and_replace): Commit edge inserts.
21764 2017-08-08  Richard Biener  <rguenther@suse.de>
21766         PR middle-end/81719
21767         * tree-ssa-loop-niter.c: Include tree-dfa.h.
21768         (expand_simple_operations): Also look through ADDR_EXPRs with
21769         MEM_REF bases treating them as POINTER_PLUS_EXPR.
21771 2017-08-08  Richard Biener  <rguenther@suse.de>
21773         PR tree-optimization/81723
21774         * tree-vect-slp.c (struct bst_traits): New hash traits.
21775         (bst_fail): New global.
21776         (vect_build_slp_tree_2): New worker, split out from ...
21777         (vect_build_slp_tree): ... this now wrapping it with using
21778         bst_fail set to cache SLP tree build fails.  Properly handle
21779         max_tree_size.
21780         (vect_analyze_slp_instance): Allocate and free bst_fail.
21782 2017-08-08  Martin Liska  <mliska@suse.cz>
21784         PR tree-opt/81696
21785         * ipa-icf-gimple.c (func_checker::compare_cst_or_decl): Consider
21786         LABEL_DECLs that can be from a different function.
21788 2017-08-08  Bin Cheng  <bin.cheng@arm.com>
21790         PR tree-optimization/81744
21791         * tree-predcom.c (prepare_finalizers_chain): Deep copy expr of
21792         loop's number of iterations.
21794 2017-08-08  Martin Liska  <mliska@suse.cz>
21796         * asan.c: Include header files.
21797         * attribs.c (build_decl_attribute_variant): New function moved
21798         from tree.[ch].
21799         (build_type_attribute_qual_variant): Likewise.
21800         (cmp_attrib_identifiers): Likewise.
21801         (simple_cst_list_equal): Likewise.
21802         (omp_declare_simd_clauses_equal): Likewise.
21803         (attribute_value_equal): Likewise.
21804         (comp_type_attributes): Likewise.
21805         (build_type_attribute_variant): Likewise.
21806         (lookup_ident_attribute): Likewise.
21807         (remove_attribute): Likewise.
21808         (merge_attributes): Likewise.
21809         (merge_type_attributes): Likewise.
21810         (merge_decl_attributes): Likewise.
21811         (merge_dllimport_decl_attributes): Likewise.
21812         (handle_dll_attribute): Likewise.
21813         (attribute_list_equal): Likewise.
21814         (attribute_list_contained): Likewise.
21815         * attribs.h (lookup_attribute): New function moved from tree.[ch].
21816         (lookup_attribute_by_prefix): Likewise.
21817         * bb-reorder.c: Include header files.
21818         * builtins.c: Likewise.
21819         * calls.c: Likewise.
21820         * cfgexpand.c: Likewise.
21821         * cgraph.c: Likewise.
21822         * cgraphunit.c: Likewise.
21823         * convert.c: Likewise.
21824         * dwarf2out.c: Likewise.
21825         * final.c: Likewise.
21826         * fold-const.c: Likewise.
21827         * function.c: Likewise.
21828         * gimple-expr.c: Likewise.
21829         * gimple-fold.c: Likewise.
21830         * gimple-pretty-print.c: Likewise.
21831         * gimple.c: Likewise.
21832         * gimplify.c: Likewise.
21833         * hsa-common.c: Likewise.
21834         * hsa-gen.c: Likewise.
21835         * internal-fn.c: Likewise.
21836         * ipa-chkp.c: Likewise.
21837         * ipa-cp.c: Likewise.
21838         * ipa-devirt.c: Likewise.
21839         * ipa-fnsummary.c: Likewise.
21840         * ipa-inline.c: Likewise.
21841         * ipa-visibility.c: Likewise.
21842         * ipa.c: Likewise.
21843         * lto-cgraph.c: Likewise.
21844         * omp-expand.c: Likewise.
21845         * omp-general.c: Likewise.
21846         * omp-low.c: Likewise.
21847         * omp-offload.c: Likewise.
21848         * omp-simd-clone.c: Likewise.
21849         * opts-global.c: Likewise.
21850         * passes.c: Likewise.
21851         * predict.c: Likewise.
21852         * sancov.c: Likewise.
21853         * sanopt.c: Likewise.
21854         * symtab.c: Likewise.
21855         * toplev.c: Likewise.
21856         * trans-mem.c: Likewise.
21857         * tree-chkp.c: Likewise.
21858         * tree-eh.c: Likewise.
21859         * tree-into-ssa.c: Likewise.
21860         * tree-object-size.c: Likewise.
21861         * tree-parloops.c: Likewise.
21862         * tree-profile.c: Likewise.
21863         * tree-ssa-ccp.c: Likewise.
21864         * tree-ssa-live.c: Likewise.
21865         * tree-ssa-loop.c: Likewise.
21866         * tree-ssa-sccvn.c: Likewise.
21867         * tree-ssa-structalias.c: Likewise.
21868         * tree-ssa.c: Likewise.
21869         * tree-streamer-in.c: Likewise.
21870         * tree-vectorizer.c: Likewise.
21871         * tree-vrp.c: Likewise.
21872         * tsan.c: Likewise.
21873         * ubsan.c: Likewise.
21874         * varasm.c: Likewise.
21875         * varpool.c: Likewise.
21876         * tree.c: Remove functions moved to attribs.[ch].
21877         * tree.h: Likewise.
21878         * config/aarch64/aarch64.c: Add attrs.h header file.
21879         * config/alpha/alpha.c: Likewise.
21880         * config/arc/arc.c: Likewise.
21881         * config/arm/arm.c: Likewise.
21882         * config/avr/avr.c: Likewise.
21883         * config/bfin/bfin.c: Likewise.
21884         * config/c6x/c6x.c: Likewise.
21885         * config/cr16/cr16.c: Likewise.
21886         * config/cris/cris.c: Likewise.
21887         * config/darwin.c: Likewise.
21888         * config/epiphany/epiphany.c: Likewise.
21889         * config/fr30/fr30.c: Likewise.
21890         * config/frv/frv.c: Likewise.
21891         * config/ft32/ft32.c: Likewise.
21892         * config/h8300/h8300.c: Likewise.
21893         * config/i386/winnt.c: Likewise.
21894         * config/ia64/ia64.c: Likewise.
21895         * config/iq2000/iq2000.c: Likewise.
21896         * config/lm32/lm32.c: Likewise.
21897         * config/m32c/m32c.c: Likewise.
21898         * config/m32r/m32r.c: Likewise.
21899         * config/m68k/m68k.c: Likewise.
21900         * config/mcore/mcore.c: Likewise.
21901         * config/microblaze/microblaze.c: Likewise.
21902         * config/mips/mips.c: Likewise.
21903         * config/mmix/mmix.c: Likewise.
21904         * config/mn10300/mn10300.c: Likewise.
21905         * config/moxie/moxie.c: Likewise.
21906         * config/msp430/msp430.c: Likewise.
21907         * config/nds32/nds32-isr.c: Likewise.
21908         * config/nds32/nds32.c: Likewise.
21909         * config/nios2/nios2.c: Likewise.
21910         * config/nvptx/nvptx.c: Likewise.
21911         * config/pa/pa.c: Likewise.
21912         * config/pdp11/pdp11.c: Likewise.
21913         * config/powerpcspe/powerpcspe.c: Likewise.
21914         * config/riscv/riscv.c: Likewise.
21915         * config/rl78/rl78.c: Likewise.
21916         * config/rx/rx.c: Likewise.
21917         * config/s390/s390.c: Likewise.
21918         * config/sh/sh.c: Likewise.
21919         * config/sol2.c: Likewise.
21920         * config/sparc/sparc.c: Likewise.
21921         * config/spu/spu.c: Likewise.
21922         * config/stormy16/stormy16.c: Likewise.
21923         * config/tilegx/tilegx.c: Likewise.
21924         * config/tilepro/tilepro.c: Likewise.
21925         * config/v850/v850.c: Likewise.
21926         * config/vax/vax.c: Likewise.
21927         * config/visium/visium.c: Likewise.
21928         * config/xtensa/xtensa.c: Likewise.
21930 2017-08-07  Michael Meissner  <meissner@linux.vnet.ibm.com>
21932         PR target/81593
21933         * config/rs6000/vsx.md (vsx_concat_<mode>, VSX_D): Cleanup
21934         constraints since the -mupper-regs-* switches have been
21935         eliminated.
21936         (vsx_concat_<mode>_1): New combiner insns to recognize inserting
21937         into a vector from a double word element that was extracted from
21938         another vector, and eliminate extra XXPERMDI instructions.
21939         (vsx_concat_<mode>_2): Likewise.
21940         (vsx_concat_<mode>_3): Likewise.
21941         (vsx_set_<mode>, VSX_D): Rewrite vector set in terms of vector
21942         concat to allow optimizing inserts from previous extracts.
21944 2017-08-07  Uros Bizjak  <ubizjak@gmail.com>
21946         * config/i386/i386.c (ix86_stack_protect_guard): Generate
21947         memory reference to a SSP offset in TLS address space.
21948         (ix86_print_operand) <case '@'>: Remove.
21949         (ix86_print_operand_punct_valid_p): Remove '@' code.
21950         * config/i386/i386.md (unspec): Remove UNSPEC_SP_TLS_SET and
21951         UNSPEC_SP_TLS_TEST.
21952         (stack_tls_protect_set_<mode>): Remove.
21953         (stack_protect_set): Do not call gen_stack_tls_protect_set_<mode>.
21954         (stack_tls_protect_test_<mode>): Remove.
21955         (stack_protect_test): Do not call gen_stack_tls_protect_test_<mode>.
21957 2017-08-07  Olivier Hainque  <hainque@adacore.com>
21959         PR target/81755
21960         * config/vxworksae.h (VXWORKS_HAVE_TLS): Define.
21962 2017-08-07  Douglas Rupp  <rupp@adacore.com>
21964         * Makefile.in (install-mkheaders): Fix typo, where the multi_dir
21965         variable was referenced as multidir in command.
21967 2017-08-07  Jakub Jelinek  <jakub@redhat.com>
21969         PR c/69389
21970         * gimplify.c (goa_stabilize_expr): Handle BIT_INSERT_EXPR and
21971         BIT_FIELD_REF.
21973 2017-08-07  Martin Liska  <mliska@suse.cz>
21975         * config/m32c/m32c.c: Add include of stringpool.h and attribs.h.
21976         * config/rl78/rl78.c: Add include of attribs.h.
21977         * config/sh/sh.c: Likewise.
21978         * config/v850/v850.c: Likewise.
21980 2017-08-07  Tom de Vries  <tom@codesourcery.com>
21982         PR middle-end/78266
21983         * omp-expand.c (expand_oacc_for): Ensure diff_type is large enough.
21985 2017-08-07  Martin Liska  <mliska@suse.cz>
21987         * config/mips/mips.c: Include attribs.h.
21989 2017-08-07  Thomas Koenig  <tkoenig@gcc.gnu.org>
21991         PR fortran/68829
21992         * doc/invoke.texi: Document change in behvaior for -Ofast for
21993         Fortran.
21995 2017-08-07  Wilco Dijkstra  <wdijkstr@arm.com>
21997         * config/aarch64/aarch64.c (aarch64_pushwb_single_reg):
21998         Use gen_frame_mem.
21999         (aarch64_pop_regs): Likewise.
22000         (aarch64_gen_load_pair): Likewise.
22001         (aarch64_save_callee_saves): Likewise.
22002         (aarch64_restore_callee_saves): Likewise.
22004 2017-08-07  H.J. Lu  <hongjiu.lu@intel.com>
22006         * config/i386/i386.c: Revert the last change.
22008 2017-08-07  H.J. Lu  <hongjiu.lu@intel.com>
22010         PR target/81736
22011         * config/i386/i386.c (ix86_finalize_stack_realign_flags): Renamed
22012         to ...
22013         (ix86_finalize_stack_frame_flags): This.  Also clear
22014         frame_pointer_needed if -fno-omit-frame-pointer is used without
22015         stack access.
22016         (ix86_expand_prologue): Replace ix86_finalize_stack_realign_flags
22017         with ix86_finalize_stack_frame_flags.
22018         (ix86_expand_epilogue): Likewise.
22019         (ix86_expand_split_stack_prologue): Likewise.
22021 2017-08-07  H.J. Lu  <hongjiu.lu@intel.com>
22023         PR target/81743
22024         * config/i386/i386.c (get_builtin_code_for_version): Set priority
22025         to P_AES for Westmere.
22027 2017-08-07  Jonathan Yong  <10walls@gmail.com>
22029         * config/i386/mingw.opt (fset-stack-executable): Removed.
22030         * config/i386/cygming.opt (fset-stack-executable): Moved
22031         from mingw.opt.
22032         * config/i386/cygwin.h: Define CHECK_EXECUTE_STACK_ENABLED.
22034 2017-08-07  Segher Boessenkool  <segher@kernel.crashing.org>
22036         * print-rtl.c (print_exp): Print NOT as "~" instead of as "!".
22038 2017-08-07  Marek Polacek  <polacek@redhat.com>
22040         PR middle-end/81737
22041         * fold-const.c (fold_indirect_ref_1): Check type_domain.
22043 2017-08-07  Martin Liska  <mliska@suse.cz>
22045         * attribs.h (canonicalize_attr_name): New function.
22046         (cmp_attribs): Move from c-format.c and adjusted.
22047         (is_attribute_p): Moved from tree.h.
22048         * tree-inline.c: Add new includes.
22049         * tree.c (cmp_attrib_identifiers): Use cmp_attribs.
22050         (private_is_attribute_p): Remove.
22051         (private_lookup_attribute): Likewise.
22052         (private_lookup_attribute_by_prefix): Simplify.
22053         (remove_attribute): Use is_attribute_p.
22054         * tree.h: Remove removed declarations.
22056 2017-08-07  Jakub Jelinek  <jakub@redhat.com>
22058         PR middle-end/81698
22059         * stmt.c (emit_case_dispatch_table): Add DEFAULT_EDGE argument,
22060         instead of computing it in the function.  Formatting fix.
22061         (expand_case): Don't rely on default_edge being the first edge,
22062         clear it if removing it, pass default_edge to
22063         emit_case_dispatch_table.
22064         (expand_sjlj_dispatch_table): Pass NULL as DEFAULT_EDGE, formatting
22065         fix.
22067 2017-08-06  Uros Bizjak  <ubizjak@gmail.com>
22069         * config/alpha/alpha.c (alpha_reorg): If trap is the last active
22070         insn in the function, emit NOP after the insn.
22072 2017-08-06  Tom de Vries  <tom@codesourcery.com>
22074         * omp-expand.c (expand_oacc_for): Add missing edge probability for tile
22075         and element loops.
22077 2017-08-06  Tom de Vries  <tom@codesourcery.com>
22079         * omp-expand.c (expand_oacc_for): Add missing edge probability for chunk
22080         loop.
22082 2017-08-04  Yury Gribov  <tetra2005@gmail.com>
22084         PR tree-optimization/57371
22085         * match.pd: New pattern.
22087 2017-08-04  Marek Polacek  <polacek@redhat.com>
22089         PR middle-end/81695
22090         * fold-const.c (fold_indirect_ref_1): For ((int *)&a + 4 -> a[1],
22091         perform the computation in offset_int.
22093 2017-08-04  Richard Sandiford  <richard.sandiford@linaro.org>
22095         PR tree-optimization/81136
22096         * tree-vectorizer.h: Include tree-hash-traits.h.
22097         (vec_base_alignments): New typedef.
22098         (vec_info): Add a base_alignments field.
22099         (vect_record_base_alignments): Declare.
22100         * tree-data-ref.h (data_reference): Add an is_conditional_in_stmt
22101         field.
22102         (DR_IS_CONDITIONAL_IN_STMT): New macro.
22103         (create_data_ref): Add an is_conditional_in_stmt argument.
22104         * tree-data-ref.c (create_data_ref): Likewise.  Use it to initialize
22105         the is_conditional_in_stmt field.
22106         (data_ref_loc): Add an is_conditional_in_stmt field.
22107         (get_references_in_stmt): Set the is_conditional_in_stmt field.
22108         (find_data_references_in_stmt): Update call to create_data_ref.
22109         (graphite_find_data_references_in_stmt): Likewise.
22110         * tree-ssa-loop-prefetch.c (determine_loop_nest_reuse): Likewise.
22111         * tree-vect-data-refs.c (vect_analyze_data_refs): Likewise.
22112         (vect_record_base_alignment): New function.
22113         (vect_record_base_alignments): Likewise.
22114         (vect_compute_data_ref_alignment): Adjust base_addr and aligned_to
22115         for nested statements even if we fail to compute a misalignment.
22116         Use pooled base alignments for unconditional references.
22117         (vect_find_same_alignment_drs): Compare base addresses instead
22118         of base objects.
22119         (vect_analyze_data_refs_alignment): Call vect_record_base_alignments.
22120         * tree-vect-slp.c (vect_slp_analyze_bb_1): Likewise.
22122 2017-08-04  Richard Sandiford  <richard.sandiford@linaro.org>
22124         * tree-vectorizer.h (vec_info): Add a constructor and destructor.
22125         Add an explicit name for the enum.  Use auto_vec for slp_instances
22126         and grouped_stores.
22127         (_loop_vec_info): Add a constructor and destructor.  Use auto_vec
22128         for all vectors.
22129         (_bb_vec_info): Add a constructor and destructor.
22130         (vinfo_for_stmt): Return NULL for uids of -1 as well.
22131         (destroy_loop_vec_info): Delete.
22132         (vect_destroy_datarefs): Likewise.
22133         * tree-vectorizer.c (vect_destroy_datarefs): Delete.
22134         (vec_info::vec_info): New function.
22135         (vec_info::~vec_info): Likewise.
22136         (vectorize_loops): Use delete instead of destroy_loop_vec_info.
22137         * tree-parloops.c (gather_scalar_reductions): Use delete instead of
22138         destroy_loop_vec_info.
22139         * tree-vect-loop.c (new_loop_vec_info): Replace with...
22140         (_loop_vec_info::_loop_vec_info): ...this.
22141         (destroy_loop_vec_info): Replace with...
22142         (_loop_vec_info::~_loop_vec_info): ...this.  Unconditionally delete
22143         the stmt_vec_infos.  Leave handling of vec_info information to its
22144         destructor.  Remove explicit vector releases.
22145         (vect_analyze_loop_form): Use new instead of new_loop_vec_info.
22146         (vect_analyze_loop): Use delete instead of destroy_loop_vec_info.
22147         * tree-vect-slp.c (new_bb_vec_info): Replace with...
22148         (_bb_vec_info::_bb_vec_info): ...this.  Don't reserve space in
22149         BB_VINFO_GROUPED_STORES or BB_VINFO_SLP_INSTANCES.
22150         (destroy_bb_vec_info): Replace with...
22151         (_bb_vec_info::~_bb_vec_info): ...this.  Leave handling of vec_info
22152         information to its destructor.
22153         (vect_slp_analyze_bb_1): Use new and delete instead of
22154         new_bb_vec_info and destroy_bb_vec_info.
22155         (vect_slp_bb): Replace 2 calls to destroy_bb_vec_info with a
22156         single delete.
22158 2017-08-04  Richard Sandiford  <richard.sandiford@linaro.org>
22160         * tree-data-ref.h (subscript): Add access_fn field.
22161         (data_dependence_relation): Add could_be_independent_p.
22162         (SUB_ACCESS_FN, DDR_COULD_BE_INDEPENDENT_P): New macros.
22163         (same_access_functions): Move to tree-data-ref.c.
22164         * tree-data-ref.c (ref_contains_union_access_p): New function.
22165         (access_fn_component_p): Likewise.
22166         (access_fn_components_comparable_p): Likewise.
22167         (dr_analyze_indices): Add a reference to access_fn_component_p.
22168         (dump_data_dependence_relation): Use SUB_ACCESS_FN instead of
22169         DR_ACCESS_FN.
22170         (constant_access_functions): Likewise.
22171         (add_other_self_distances): Likewise.
22172         (same_access_functions): Likewise.  (Moved from tree-data-ref.h.)
22173         (initialize_data_dependence_relation): Use XCNEW and remove
22174         explicit zeroing of DDR_REVERSED_P.  Look for a subsequence
22175         of access functions that have the same type.  Allow the
22176         subsequence to end with different bases in some circumstances.
22177         Record the chosen access functions in SUB_ACCESS_FN.
22178         (build_classic_dist_vector_1): Replace ddr_a and ddr_b with
22179         a_index and b_index.  Use SUB_ACCESS_FN instead of DR_ACCESS_FN.
22180         (subscript_dependence_tester_1): Likewise dra and drb.
22181         (build_classic_dist_vector): Update calls accordingly.
22182         (subscript_dependence_tester): Likewise.
22183         * tree-ssa-loop-prefetch.c (determine_loop_nest_reuse): Check
22184         DDR_COULD_BE_INDEPENDENT_P.
22185         * tree-vectorizer.h (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Test
22186         comp_alias_ddrs instead of may_alias_ddrs.
22187         * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr):
22188         New function.
22189         (vect_analyze_data_ref_dependence): Use it if
22190         DDR_COULD_BE_INDEPENDENT_P, but fall back to using the recorded
22191         distance vectors if that fails.
22192         (dependence_distance_ge_vf): New function.
22193         (vect_prune_runtime_alias_test_list): Use it.  Don't clear
22194         LOOP_VINFO_MAY_ALIAS_DDRS.
22196 2017-08-04  Richard Biener  <rguenther@suse.de>
22198         PR middle-end/81705
22199         * fold-const.c (fold_binary_loc): Properly restrict
22200         minus_var0 && minus_var1 case when associating undefined overflow
22201         entities.
22203 2017-08-04  Tom de Vries  <tom@codesourcery.com>
22205         * omp-simd-clone.c (simd_clone_adjust): Add missing edge probability.
22207 2017-08-03  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
22209         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
22210         Don't start diagnostic messages with a capital letter.
22211         * config/rs6000/rs6000.c (rs6000_option_override_internal):
22212         Likewise.
22213         (rs6000_invalid_builtin): Likewise.
22214         (rs6000_trampoline_init): Likewise.
22216 2017-08-03  Jakub Jelinek  <jakub@redhat.com>
22218         PR target/81621
22219         * bb-reorder.c (pass_partition_blocks::execute): Return TODO_df_finish
22220         after setting changeable df flags.
22222 2017-08-03  Richard Biener  <rguenther@suse.de>
22224         * tree-ssa-reassoc.c (should_break_up_subtract): Also break
22225         up if the use is in USE - X.
22227 2017-08-03  Alexander Monakov  <amonakov@ispras.ru>
22229         * toplev.c (dumpfile.h): New include.
22230         (internal_error_reentered): New static function.  Use it...
22231         (internal_error_function): ...here to handle reentered internal_error.
22233 2017-08-03  Richard Biener  <rguenther@suse.de>
22235         PR middle-end/81148
22236         * fold-const.c (split_tree): Add minus_var and minus_con
22237         arguments, remove unused loc arg.  Never generate NEGATE_EXPRs
22238         here but always use minus_*.
22239         (associate_trees): Assert we never associate with MINUS_EXPR
22240         and NULL first operand.  Do not recurse for PLUS_EXPR operands
22241         when associating as MINUS_EXPR either.
22242         (fold_binary_loc): Track minus_var and minus_con.
22244 2017-08-03  Tom de Vries  <tom@codesourcery.com>
22246         PR lto/81430
22247         * tree-streamer-in.c (lto_input_ts_function_decl_tree_pointers): If
22248         ACCEL_COMPILER, apply finish_options on
22249         DECL_FUNCTION_SPECIFIC_OPTIMIZATION.
22251 2017-08-03  Tom de Vries  <tom@codesourcery.com>
22253         PR target/81662
22254         * config/nvptx/nvptx.c (nvptx_option_override): Emit sorry if
22255         function_entry_patch_area_size > 0.
22257 2017-08-03  Jakub Jelinek  <jakub@redhat.com>
22259         PR driver/81650
22260         * calls.c (alloc_max_size): Use HOST_WIDE_INT_UC (10??)
22261         instead of 10??LU, perform unit multiplication in wide_int,
22262         don't change alloc_object_size_limit if the limit is larger
22263         than SSIZE_MAX.
22265         PR tree-optimization/81655
22266         PR tree-optimization/81588
22267         * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Handle also
22268         the case when ranges[i].low and high are 1 for unsigned type with
22269         precision 1.
22271         PR middle-end/81052
22272         * omp-low.c (diagnose_sb_0): Handle flag_openmp_simd like flag_openmp.
22273         (pass_diagnose_omp_blocks::gate): Enable also for flag_openmp_simd.
22275 2017-08-03  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
22277         * tree-vrp.h: Add include guard.
22279 2017-08-02  Uros Bizjak  <ubizjak@gmail.com>
22281         PR target/81644
22282         * config/i386/i386.md (unspecv): Add UNSPECV_UD2.
22283         (ud2): New insn pattern.
22284         * config/i386/i386.c (ix86_expand_epilogue):
22285         For naked functions, generate ud2 instead of trap insn.
22287 2017-08-02  Marek Polacek  <polacek@redhat.com>
22289         PR other/81667
22290         * alloc-pool.h (base_pool_allocator): Initialize m_elt_size.
22292 2017-08-02  Tom de Vries  <tom@codesourcery.com>
22293             Cesar Philippidis  <cesar@codesourcery.com>
22295         * config/nvptx/nvptx.c (nvptx_lockless_update, nvptx_lockfull_update):
22296         Add missing edge probabilities.
22298 2017-08-02  Tamar Christina  <tamar.christina@arm.com>
22300         * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
22301         Correct endianness.
22303 2017-08-02  Jakub Jelinek  <jakub@redhat.com>
22305         PR middle-end/79499
22306         * function.c (thread_prologue_and_epilogue_insns): Determine blocks
22307         for find_many_sub_basic_blocks bitmap by looking up BLOCK_FOR_INSN
22308         of first NONDEBUG_INSN_P in each of the split_prologue_seq and
22309         prologue_seq sequences - if any.
22311 2017-08-02  Richard Biener  <rguenther@suse.de>
22313         * tree-vect-stmts.c (vectorizable_store): Perform vector extracts
22314         via vectors if supported, integer extracts via punning if supported
22315         or otherwise vector extracts.
22317 2017-08-02  Richard Biener  <rguenther@suse.de>
22319         * tree-ssa-pre.c (bitmap_insert_into_set_1): Remove and inline
22320         into ...
22321         (bitmap_insert_into_set): ... this.
22323 2017-08-02  Richard Biener  <rguenther@suse.de>
22325         PR tree-optimization/81633
22326         Revert
22327         2015-08-17  Alan Hayward  <alan.hayward@arm.com>
22329         PR tree-optimization/71752
22330         * tree-vect-slp.c (vect_get_slp_defs): Handle null operands.
22332 2017-08-01  Daniel Santos  <daniel.santos@pobox.com>
22334         * config/i386/i386.h (ix86_frame::outlined_save_offset): Remove field.
22335         (machine_function::call_ms2sysv_pad_out): Remove field.
22336         * config/i386/i386.c (xlogue_layout::get_stack_space_used): Modify.
22337         (ix86_compute_frame_layout): Likewise.
22339 2017-08-01  H.J. Lu  <hongjiu.lu@intel.com>
22341         PR target/81654
22342         * config/i386/i386.c (ix86_set_func_type): Disallow naked
22343         attribute with interrupt attribute.
22345 2017-08-01  Andrew Pinski  <apinski@cavium.com>
22347         * tree-ssa-scopedtables.c (hashable_expr_equal_p): Check
22348         BIT_INSERT_EXPR's operand 1
22349         to see if the types precision matches.
22351 2017-08-01  Martin Liska  <mliska@suse.cz>
22353         PR middle-end/70140
22354         * builtins.c (expand_builtin_memcpy_args): Remove.
22355         (expand_builtin_memcpy): Call newly added function
22356         expand_builtin_memory_copy_args.
22357         (expand_builtin_memcpy_with_bounds): Likewise.
22358         (expand_builtin_mempcpy): Remove last argument.
22359         (expand_builtin_mempcpy_with_bounds): Likewise.
22360         (expand_builtin_memory_copy_args): New function created from
22361         expand_builtin_mempcpy_args with small modifications.
22362         (expand_builtin_mempcpy_args): Remove.
22363         (expand_builtin_stpcpy): Remove unused argument.
22364         (expand_builtin): Likewise.
22365         (expand_builtin_with_bounds): Likewise.
22367 2017-08-01  Martin Liska  <mliska@suse.cz>
22369         Revert r250771
22370         Make mempcpy more optimal (PR middle-end/70140).
22372 2017-08-01  Jakub Jelinek  <jakub@redhat.com>
22374         PR target/81622
22375         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): For
22376         __builtin_vec_cmpne verify both arguments are compatible vectors
22377         before looking at TYPE_MODE on the element type.  For __builtin_vec_ld
22378         verify arg1_type is a pointer or array type.  For __builtin_vec_st,
22379         move computation of aligned to after checking the argument types.
22380         Formatting fixes.
22382         PR target/80846
22383         * config/rs6000/vsx.md (vextract_fp_from_shorth,
22384         vextract_fp_from_shortl): Add element mode after mode in gen_vec_init*
22385         calls.
22387 2017-08-01  Jerome Lambourg  <lambourg@adacore.com>
22388             Doug Rupp  <rupp@adacore.com>
22389             Olivier Hainque  <hainque@adacore.com>
22391         * config.gcc (arm-wrs-vxworks*): Rework to handle arm-wrs-vxworks7 as
22392         well as arm-wrs-vxworks. Update target_cpu_name from arm6 (arch v3) to
22393         arm8 (arch v4).
22394         * config/arm/vxworks.h (MAYBE_TARGET_BPABI_CPP_BUILTINS): New, helper
22395         for TARGET_OS_CPP_BUILTIN.
22396         (TARGET_OS_CPP_BUILTIN): Invoke MAYBE_TARGET_BPABI_CPP_BUILTINS(),
22397         refine CPU definitions for arm_arch5 and add those for arm_arch6 and
22398         arm_arch7.
22399         (MAYBE_ASM_ABI_SPEC): New, helper for SUBTARGET_EXTRA_ASM_SPEC,
22400         passing required abi options to the assembler for EABI configurations.
22401         (EXTRA_CC1_SPEC): New macro, to help prevent the implicit production
22402         of .text.hot and .text.unlikely sections for kernel modules when
22403         using ARM style exceptions.
22404         (CC1_SPEC): Remove obsolete attempt at mimicking Diab toolchain
22405         options. Add EXTRA_CC1_SPEC.
22406         (VXWORKS_ENDIAN_SPEC): Adjust comment and remove handling of Diab
22407         toolchain options.
22408         (DWARF2_UNWIND_INFO): Redefine to handle the pre/post VxWorks 7
22409         transition.
22410         (ARM_TARGET2_DWARF_FORMAT): Define.
22411         * config/arm/t-vxworks: Adjust multilib control to removal of the
22412         Diab command line options.
22414 2017-08-01  Martin Liska  <mliska@suse.cz>
22416         PR gcov-profile/81561
22417         * gcov.c (unblock): Make unblocking safe as we need to preserve
22418         index correspondence of blocks and block_lists.
22420 2017-08-01  Richard Biener  <rguenther@suse.de>
22422         PR tree-optimization/81181
22423         * tree-ssa-pre.c (compute_antic_aux): Defer clean() to ...
22424         (compute_antic): ... end of iteration here.
22426 2017-08-01  James Greenhalgh  <james.greenhalgh@arm.com>
22428         * common.opt (ftree-vectorize): No longer set flag_tree_vectorize.
22429         (ftree-loop-vectorize): Set as EnabledBy ftree-vectorize.
22430         (ftree-slp-vectorize): Likewise.
22431         * omp-expand (expand_omp_simd): Remove flag_tree_vectorize, as it
22432         can no longer be set independent of flag_tree_loop_vectorize.
22433         * omp-general.c (emp_max_vf): Likewise.
22434         * opts.c (enable_fdo_optimizations): Remove references to
22435         flag_tree_vectorize, these are now implicit.
22436         (common_handle_option): Remove handling for OPT_ftree_vectorize,
22437         and leave it for the options machinery.
22439 2017-08-01  Martin Liska  <mliska@suse.cz>
22441         PR middle-end/70140
22442         * builtins.c (expand_builtin_memcpy_args): Remove.
22443         (expand_builtin_memcpy): Call newly added function
22444         expand_builtin_memory_copy_args.
22445         (expand_builtin_memcpy_with_bounds): Likewise.
22446         (expand_builtin_mempcpy): Remove last argument.
22447         (expand_builtin_mempcpy_with_bounds): Likewise.
22448         (expand_builtin_memory_copy_args): New function created from
22449         expand_builtin_mempcpy_args with small modifications.
22450         (expand_builtin_mempcpy_args): Remove.
22451         (expand_builtin_stpcpy): Remove unused argument.
22452         (expand_builtin): Likewise.
22453         (expand_builtin_with_bounds): Likewise.
22455 2017-08-01  Uros Bizjak  <ubizjak@gmail.com>
22457         PR target/81641
22458         * config/i386/i386.c (ix86_print_operand_address_as): For -masm=intel
22459         print "ds:" only for immediates in generic address space.
22461 2017-08-01  Uros Bizjak  <ubizjak@gmail.com>
22463         PR target/81639
22464         * config/i386/i386.c (ix86_funciton_naked): New prototype.
22465         (ix86_function_ok_for_sibcall): Return false for naked functions.
22467 2017-08-01  Richard Biener  <rguenther@suse.de>
22469         * tree-ssa-pre.c (print_pre_expr): Handle NULL expr.
22470         (compute_antic): Seed worklist with exit block predecessors.
22471         * cfganal.c (dfs_find_deadend): For a cycle return the source
22472         of the edge closing it.
22474 2017-08-01  Tamar Christina  <tamar.christina@arm.com>
22476         * config/aarch64/aarch64.c
22477         (aarch64_can_const_movi_rtx_p): Move 0 check.
22479 2017-08-01  Bin Cheng  <bin.cheng@arm.com>
22481         * tree.h (POINTER_TYPE_OVERFLOW_UNDEFINED): Delete.
22482         * fold-const.c (fold_comparison, fold_binary_loc): Delete use of
22483         above macro.
22484         * match.pd: Ditto in address comparison pattern.
22486 2017-08-01  Bin Cheng  <bin.cheng@arm.com>
22488         PR tree-optimization/81627
22489         * tree-predcom.c (prepare_finalizers): Always rewrite into loop
22490         closed ssa form for store-store chain.
22492 2017-08-01  Bin Cheng  <bin.cheng@arm.com>
22494         PR tree-optimization/81620
22495         * tree-predcom.c (add_ref_to_chain): Don't set has_max_use_after
22496         for store-store chain.
22498 2017-08-01  Jakub Jelinek  <jakub@redhat.com>
22500         PR tree-optimization/81588
22501         * tree-ssa-reassoc.c (optimize_range_tests_var_bound): If
22502         ranges[i].in_p, invert comparison code ccode.  For >/>=,
22503         swap rhs1 and rhs2 and comparison code unconditionally,
22504         for </<= don't do that.  Don't swap rhs1/rhs2 again if
22505         ranges[i].in_p, instead invert comparison code ccode if
22506         opcode or oe->rank is BIT_IOR_EXPR.
22508         PR target/80846
22509         * optabs.def (vec_extract_optab, vec_init_optab): Change from
22510         a direct optab to conversion optab.
22511         * optabs.c (expand_vector_broadcast): Use convert_optab_handler
22512         with GET_MODE_INNER as last argument instead of optab_handler.
22513         * expmed.c (extract_bit_field_1): Likewise.  Use vector from
22514         vector extraction if possible and optab is available.
22515         * expr.c (store_constructor): Use convert_optab_handler instead
22516         of optab_handler.  Use vector initialization from smaller
22517         vectors if possible and optab is available.
22518         * tree-vect-stmts.c (vectorizable_load): Likewise.
22519         * doc/md.texi (vec_extract, vec_init): Document that the optabs
22520         now have two modes.
22521         * config/i386/i386.c (ix86_expand_vector_init): Handle expansion
22522         of vec_init from half-sized vectors with the same element mode.
22523         * config/i386/sse.md (ssehalfvecmode): Add V4TI case.
22524         (ssehalfvecmodelower, ssescalarmodelower): New mode attributes.
22525         (reduc_plus_scal_v8df, reduc_plus_scal_v4df, reduc_plus_scal_v2df,
22526         reduc_plus_scal_v16sf, reduc_plus_scal_v8sf, reduc_plus_scal_v4sf,
22527         reduc_<code>_scal_<mode>, reduc_umin_scal_v8hi): Add element mode
22528         after mode in gen_vec_extract* calls.
22529         (vec_extract<mode>): Renamed to ...
22530         (vec_extract<mode><ssescalarmodelower>): ... this.
22531         (vec_extract<mode><ssehalfvecmodelower>): New expander.
22532         (rotl<mode>3, rotr<mode>3, <shift_insn><mode>3, ashrv2di3): Add
22533         element mode after mode in gen_vec_init* calls.
22534         (VEC_INIT_HALF_MODE): New mode iterator.
22535         (vec_init<mode>): Renamed to ...
22536         (vec_init<mode><ssescalarmodelower>): ... this.
22537         (vec_init<mode><ssehalfvecmodelower>): New expander.
22538         * config/i386/mmx.md (vec_extractv2sf): Renamed to ...
22539         (vec_extractv2sfsf): ... this.
22540         (vec_initv2sf): Renamed to ...
22541         (vec_initv2sfsf): ... this.
22542         (vec_extractv2si): Renamed to ...
22543         (vec_extractv2sisi): ... this.
22544         (vec_initv2si): Renamed to ...
22545         (vec_initv2sisi): ... this.
22546         (vec_extractv4hi): Renamed to ...
22547         (vec_extractv4hihi): ... this.
22548         (vec_initv4hi): Renamed to ...
22549         (vec_initv4hihi): ... this.
22550         (vec_extractv8qi): Renamed to ...
22551         (vec_extractv8qiqi): ... this.
22552         (vec_initv8qi): Renamed to ...
22553         (vec_initv8qiqi): ... this.
22554         * config/rs6000/vector.md (VEC_base_l): New mode attribute.
22555         (vec_init<mode>): Renamed to ...
22556         (vec_init<mode><VEC_base_l>): ... this.
22557         (vec_extract<mode>): Renamed to ...
22558         (vec_extract<mode><VEC_base_l>): ... this.
22559         * config/rs6000/paired.md (vec_initv2sf): Renamed to ...
22560         (vec_initv2sfsf): ... this.
22561         * config/rs6000/altivec.md (splitter, altivec_copysign_v4sf3,
22562         vec_unpacku_hi_v16qi, vec_unpacku_hi_v8hi, vec_unpacku_lo_v16qi,
22563         vec_unpacku_lo_v8hi, mulv16qi3, altivec_vreve<mode>2): Add
22564         element mode after mode in gen_vec_init* calls.
22565         * config/aarch64/aarch64-simd.md (vec_init<mode>): Renamed to ...
22566         (vec_init<mode><Vel>): ... this.
22567         (vec_extract<mode>): Renamed to ...
22568         (vec_extract<mode><Vel>): ... this.
22569         * config/aarch64/iterators.md (Vel): New mode attribute.
22570         * config/s390/s390.c (s390_expand_vec_strlen, s390_expand_vec_movstr):
22571         Add element mode after mode in gen_vec_extract* calls.
22572         * config/s390/vector.md (non_vec_l): New mode attribute.
22573         (vec_extract<mode>): Renamed to ...
22574         (vec_extract<mode><non_vec_l>): ... this.
22575         (vec_init<mode>): Renamed to ...
22576         (vec_init<mode><non_vec_l>): ... this.
22577         * config/s390/s390-builtins.def (s390_vlgvb, s390_vlgvh, s390_vlgvf,
22578         s390_vlgvf_flt, s390_vlgvg, s390_vlgvg_dbl): Add element mode after
22579         vec_extract mode.
22580         * config/arm/iterators.md (V_elem_l): New mode attribute.
22581         * config/arm/neon.md (vec_extract<mode>): Renamed to ...
22582         (vec_extract<mode><V_elem_l>): ... this.
22583         (vec_extractv2di): Renamed to ...
22584         (vec_extractv2didi): ... this.
22585         (vec_init<mode>): Renamed to ...
22586         (vec_init<mode><V_elem_l>): ... this.
22587         (reduc_plus_scal_<mode>, reduc_plus_scal_v2di, reduc_smin_scal_<mode>,
22588         reduc_smax_scal_<mode>, reduc_umin_scal_<mode>,
22589         reduc_umax_scal_<mode>, neon_vget_lane<mode>, neon_vget_laneu<mode>):
22590         Add element mode after gen_vec_extract* calls.
22591         * config/mips/mips-msa.md (vec_init<mode>): Renamed to ...
22592         (vec_init<mode><unitmode>): ... this.
22593         (vec_extract<mode>): Renamed to ...
22594         (vec_extract<mode><unitmode>): ... this.
22595         * config/mips/loongson.md (vec_init<mode>): Renamed to ...
22596         (vec_init<mode><unitmode>): ... this.
22597         * config/mips/mips-ps-3d.md (vec_initv2sf): Renamed to ...
22598         (vec_initv2sfsf): ... this.
22599         (vec_extractv2sf): Renamed to ...
22600         (vec_extractv2sfsf): ... this.
22601         (reduc_plus_scal_v2sf, reduc_smin_scal_v2sf, reduc_smax_scal_v2sf):
22602         Add element mode after gen_vec_extract* calls.
22603         * config/mips/mips.md (unitmode): New mode iterator.
22604         * config/spu/spu.c (spu_expand_prologue, spu_allocate_stack,
22605         spu_builtin_extract): Add element mode after gen_vec_extract* calls.
22606         * config/spu/spu.md (inner_l): New mode attribute.
22607         (vec_init<mode>): Renamed to ...
22608         (vec_init<mode><inner_l>): ... this.
22609         (vec_extract<mode>): Renamed to ...
22610         (vec_extract<mode><inner_l>): ... this.
22611         * config/sparc/sparc.md (veltmode): New mode iterator.
22612         (vec_init<VMALL:mode>): Renamed to ...
22613         (vec_init<VMALL:mode><VMALL:veltmode>): ... this.
22614         * config/ia64/vect.md (vec_initv2si): Renamed to ...
22615         (vec_initv2sisi): ... this.
22616         (vec_initv2sf): Renamed to ...
22617         (vec_initv2sfsf): ... this.
22618         (vec_extractv2sf): Renamed to ...
22619         (vec_extractv2sfsf): ... this.
22620         * config/powerpcspe/vector.md (VEC_base_l): New mode attribute.
22621         (vec_init<mode>): Renamed to ...
22622         (vec_init<mode><VEC_base_l>): ... this.
22623         (vec_extract<mode>): Renamed to ...
22624         (vec_extract<mode><VEC_base_l>): ... this.
22625         * config/powerpcspe/paired.md (vec_initv2sf): Renamed to ...
22626         (vec_initv2sfsf): ... this.
22627         * config/powerpcspe/altivec.md (splitter, altivec_copysign_v4sf3,
22628         vec_unpacku_hi_v16qi, vec_unpacku_hi_v8hi, vec_unpacku_lo_v16qi,
22629         vec_unpacku_lo_v8hi, mulv16qi3): Add element mode after mode in
22630         gen_vec_init* calls.
22632 2017-08-01  Richard Biener  <rguenther@suse.de>
22634         PR tree-optimization/81297
22635         * tree-vrp.c (get_single_symbol): Remove assert, instead drop
22636         TREE_OVERFLOW from INTEGER_CSTs.
22638 2017-07-31  Segher Boessenkool  <segher@kernel.crashing.org>
22640         * config/rs6000/rs6000.c (enum rs6000_reg_type): Delete trailing comma.
22642 2017-07-31  Carl Love  <cel@us.ibm.com>
22644         * config/rs6000/rs6000-c: Add support for built-in functions
22645         vector signed char vec_xl_be (signed long long, signed char *);
22646         vector unsigned char vec_xl_be (signed long long, unsigned char *);
22647         vector signed int vec_xl_be (signed long long, signed int *);
22648         vector unsigned int vec_xl_be (signed long long, unsigned int *);
22649         vector signed long long vec_xl_be (signed long long, signed long long *);
22650         vector unsigned long long vec_xl_be (signed long long, unsigned long long *);
22651         vector signed short vec_xl_be (signed long long, signed short *);
22652         vector unsigned short vec_xl_be (signed long long, unsigned short *);
22653         vector double vec_xl_be (signed long long, double *);
22654         vector float vec_xl_be (signed long long, float *);
22655         * config/rs6000/altivec.h (vec_xl_be): Add #define.
22656         * config/rs6000/rs6000-builtin.def (XL_BE_V16QI, XL_BE_V8HI, XL_BE_V4SI)
22657         XL_BE_V2DI, XL_BE_V4SF, XL_BE_V2DF, XL_BE): Add definitions
22658         for the builtins.
22659         * config/rs6000/rs6000.c (altivec_expand_xl_be_builtin): Add function.
22660         (altivec_expand_builtin): Add switch statement to call
22661         altivec_expand_xl_be for each builtin.
22662         (altivec_init_builtins): Add def_builtin for _builtin_vsx_le_be_v8hi,
22663         __builtin_vsx_le_be_v4si, __builtin_vsx_le_be_v2di,
22664         __builtin_vsx_le_be_v4sf, __builtin_vsx_le_be_v2df,
22665         __builtin_vsx_le_be_v16qi.
22666         * doc/extend.texi: Update the built-in documentation file for the
22667         new built-in functions.
22669 2017-07-31  Uros Bizjak  <ubizjak@gmail.com>
22671         PR target/25967
22672         * config/i386/i386.c (ix86_allocate_stack_slots_for_args):
22673         New function.
22674         (TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS): Define.
22676 2017-07-31  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
22678         * config.gcc: Add z14.
22679         * config/s390/driver-native.c (s390_host_detect_local_cpu): Add
22680         CPU model numbers for z13s and z14.
22681         * config/s390/s390-c.c (s390_resolve_overloaded_builtin): Replace
22682         arch12 with z14.
22683         * config/s390/s390-opts.h (enum processor_type): Rename
22684         PROCESSOR_ARCH12 to PROCESSOR_3906_Z14.
22685         * config/s390/s390.c (processor_table): Add field for CPU name to
22686         be passed to Binutils.
22687         (s390_asm_output_machine_for_arch): Use the new field in
22688         processor_table for Binutils.
22689         (s390_expand_builtin): Replace arch12 with z14.
22690         (s390_issue_rate): Rename PROCESSOR_ARCH12 to PROCESSOR_3906_Z14.
22691         (s390_get_sched_attrmask): Likewise.
22692         (s390_get_unit_mask): Likewise.
22693         * config/s390/s390.opt: Add z14 to processor_type enum.
22695 2017-07-31  Martin Jambor  <mjambor@suse.cz>
22697         PR hsa/81477
22698         * ipa-fnsummary.c (ipa_fn_summary_generate): Set versionable
22699         regardless of optimization level.
22701 2017-07-31  Jan Hubicka <hubicka@ucw.cz>
22702             Martin Liska  <mliska@suse.cz>
22704         * predict.def: Remove old comment and adjust probability.
22705         * gimplify.c (should_warn_for_implicit_fallthrough): Ignore
22706         PREDICT statements.
22708 2017-07-31  Uros Bizjak  <ubizjak@gmail.com>
22710         PR target/25967
22711         * config/i386/i386.c (ix86_function_naked): New function.
22712         (ix86_can_use_return_insn_p): Return false for naked functions.
22713         (ix86_expand_prologue): Skip prologue for naked functions.
22714         (ix86_expand_epilogue): Skip epilogue for naked functions
22715         and emit trap instruction.
22716         (ix86_warn_func_return): New function.
22717         (ix86_attribute_table): Add "naked" attribute specification.
22718         (TARGET_WARN_FUNC_RETURN): Define.
22719         * doc/extend.texi (x86 Function Attributes) <naked>: Document it.
22721 2017-07-31  Martin Liska  <mliska@suse.cz>
22723         * gimple-pretty-print.c (dump_gimple_label): Never dump BB info.
22724         (dump_gimple_bb_header): Always dump BB info.
22725         (pp_cfg_jump): Do not append info about BB when dumping a jump.
22727 2017-07-31  Martin Liska  <mliska@suse.cz>
22729         PR sanitize/81530
22730         * convert.c (convert_to_integer_1): Guard condition with flag_sanitize_p
22731         also with current_function_decl non-null equality.
22733 2017-07-31  Jakub Jelinek  <jakub@redhat.com>
22735         PR sanitizer/81604
22736         * ubsan.c (ubsan_type_descriptor): For UBSAN_PRINT_ARRAY don't
22737         change type to the element type, instead add eltype variable and
22738         use it where we are interested in the element type.
22740         PR tree-optimization/81603
22741         * ipa-polymorphic-call.c
22742         (ipa_polymorphic_call_context::ipa_polymorphic_call_context): Perform
22743         offset arithmetic in offset_int, bail out if the resulting bit offset
22744         doesn't fit into shwi.
22746 2017-07-31  Martin Liska  <mliska@suse.cz>
22748         * gimplify.c (mostly_copy_tree_r): Remove Java specific hunk.
22749         (gimplify_save_expr): Fix comment.
22751 2017-07-30  H.J. Lu  <hongjiu.lu@intel.com>
22753         PR target/79793
22754         * config/i386/i386.c (ix86_function_arg): Update arguments for
22755         exception handler.
22756         (ix86_compute_frame_layout): Set the initial stack offset to
22757         INCOMING_FRAME_SP_OFFSET.  Update red-zone offset with
22758         INCOMING_FRAME_SP_OFFSET.
22759         (ix86_expand_epilogue): Don't pop the 'ERROR_CODE' off the
22760         stack before exception handler returns.
22761         * config/i386/i386.h (INCOMING_FRAME_SP_OFFSET): Add the
22762         the 'ERROR_CODE' for exception handler.
22764 2017-07-30  Uros Bizjak  <ubizjak@gmail.com>
22766         * config/i386/i386.h (ASM_PRINTF_EXTENSIONS): New macro.
22767         (ASM_OUTPUT_REG_PUSH): Rewrite with new operand modifiers.
22768         (ASM_OUTPUT_REG_POP): Ditto.
22769         * config/i386/i386.c (ix86_asm_output_function_label): Use fputs
22770         instead of asm_fprintf to output pure string.
22772 2017-07-29  Jakub Jelinek  <jakub@redhat.com>
22774         * debug.h (struct gcc_debug_hooks): Add IMPLICIT argument
22775         to imported_module_or_decl hook.
22776         (debug_nothing_tree_tree_tree_bool): Remove.
22777         (debug_nothing_tree_tree_tree_bool_bool): New declaration.
22778         * debug.c (do_nothing_debug_hooks): Use
22779         debug_nothing_tree_tree_tree_bool_bool instead of
22780         debug_nothing_tree_tree_tree_bool.
22781         * vmsdbgout.c (vmsdbg_debug_hooks): Likewise.
22782         * dbxout.c (dbx_debug_hooks, xcoff_debug_hooks): Likewise.
22783         * sdbout.c (sdb_debug_hooks): Likewise.
22784         * dwarf2out.c (dwarf2_lineno_debug_hooks): Likewise.
22785         (gen_namespace_die): Add DW_AT_export_symbols attribute if
22786         langhook wants it.
22787         (dwarf2out_imported_module_or_decl): Add IMPLICIT argument,
22788         if true, -gdwarf-5 and decl will have DW_AT_export_symbols
22789         attribute, don't add anything.
22791 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22793         * fold-const.c (fold_build1_stat_loc): Remove _stat from name.
22794         (fold_build2_stat_loc): Likewise.
22795         (fold_build3_stat_loc): Likewise.
22796         * fold-const.h (fold_build1, fold_build2, fold_build3): Adjust.
22797         (fold_build1_loc): Remove macro.
22798         (fold_build2_loc): Likewise.
22799         (fold_build3_loc): Likewise.
22801 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22803         * gimple.c (gimple_build_debug_bind_stat): Remove _stat from name.
22804         (gimple_build_debug_bind_source_stat): Likewise.
22805         * gimple.h (gimple_build_debug_bind): Remove macro.
22806         (gimple_build_debug_bind_source): Likewise.
22808 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22810         * bitmap.c (bitmap_alloc): Adjust.
22811         (bitmap_gc_alloc): Likewise.
22812         * bitmap.h (bitmap_initialize_stat): Remove _stat from name.
22814 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22816         * bitmap.c (bitmap_obstack_alloc_stat): Rename to bitmap_alloc.
22817         (bitmap_gc_alloc_stat): Rename to bitmap_gc_alloc.
22818         * bitmap.h (bitmap_obstack_alloc_stat): Adjust prototype.
22819         (bitmap_gc_alloc_stat): Likewise.
22820         (BITMAP_ALLOC, BITMAP_GGC_ALLOC): Adjust.
22822 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22824         * rtl.c (shallow_copy_rtx_stat): Remove _stat from name.
22825         * rtl.h (shallow_copy_rtx): Remove macro.
22827 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22829         * emit-rtl.c (gen_raw_REG): Adjust.
22830         * gengenrtl.c (gendef): Likewise.
22831         * rtl.c (rtx_alloc_stat): Remove _stat from name.
22832         * rtl.h (rtx_alloc): Remove macro.
22834 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22836         * tree.c (build_tree_list_vec_stat): Remove _stat from name.
22837         (build_tree_list_stat): Likewise.
22838         * tree.h (build_tree_list): Remove macro.
22839         (build_tree_list_vec): Likewise.
22841 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22843         * tree.c (make_vector_stat): Remove _stat from name.
22844         (build_vector_stat): Likewise.
22845         * tree.h (make_vector_stat): Remove macro.
22846         (build_vector_stat): Likewise.
22848 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22850         * tree.h (build_var_debug_value): Remove prototype.
22852 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22854         * tree.c (tree_cons_stat): Remove _stat from name.
22855         * tree.h (tree_cons): Remove macro.
22857 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22859         * tree.c (build_vl_exp_stat): Remove _stat from name.
22860         * tree.h (build_vl_exp): Remove macro.
22862 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22864         * tree.c (build_decl_stat): Remove _stat from name.
22865         * tree.h (build_decl): Remove macro.
22867 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22869         * gimple.c (gimple_build_with_ops_stat): Adjust.
22870         (gimple_alloc_stat): Remove _stat from name.
22871         * gimple.h (gimple_alloc): Remove macro.
22873 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22875         * tree.c (make_tree_vec_stat): Remove _stat from name.
22876         (grow_tree_vec_stat): Likewise.
22877         * tree.h (make_tree_vec_stat): Adjust prototype.
22878         (grow_tree_vec_stat): Likewise.
22879         (make_tree_vec): Remove macro.
22880         (grow_tree_vec): Likewise.
22882 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22884         * fold-const.c (fold_build1_stat_loc): Adjust.
22885         (fold_build2_stat_loc): Likewise.
22886         (fold_build3_stat_loc): Likewise.
22887         * tree.c (build0_stat): Remove _stat from name.
22888         (build1_stat): Likewise.
22889         (build2_stat): Likewise.
22890         (build3_stat): Likewise.
22891         (build4_stat): Likewise.
22892         (build5_stat): Likewise.
22893         * tree.h (build1_loc): Remove macro, and rename _stat function
22894         to this.
22895         (build2_loc): Likewise.
22896         (build3_loc): Likewise.
22897         (build4_loc): Likewise.
22898         (build5_loc): Likewise.
22900 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22902         * tree.c (make_int_cst_stat): Remove _stat from name.
22903         * tree.h (make_int_cst_stat): Adjust prototype.
22904         (make_int_cst): Remove macro.
22906 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22908         * tree.c (make_tre_binfo_stat): Remove _stat from name.
22909         * tree.h (make_tree_binfo_stat): Adjust prototype.
22910         (make_tree_binfo): Remove.
22912 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22914         * tree.c (copy_node_stat): Rename to copy_node.
22915         (build_distinct_type_copy): Adjust.
22916         * tree.h (copy_node_stat): Adjust prototype.
22917         (copy_node): Remove macro.
22919 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22921         * tree.c (make_node_stat): rename to make_node.
22922         (build_tree_list_stat): Adjust.
22923         (build0_stat): Likewise.
22924         (build2_stat): Likewise.
22925         (build3_stat): Likewise.
22926         (build4_stat): Likewise.
22927         (build5_stat): Likewise.
22928         (build_decl_stat): Likewise.
22929         * tree.h (make_node_stat): Adjust prototype.
22930         (make_node): remove macro.
22932 2017-07-28  Peter Bergner  <bergner@vnet.ibm.com>
22934         * config/rs6000/ppc-auxv.h (PPC_FEATURE2_DARN): New define.
22935         (PPC_FEATURE2_SCV): Likewise.
22936         * config/rs6000/rs6000.c (cpu_supports_info): Use them.
22938 2017-07-28  Tamar Christina  <tamar.christina@arm.com>
22940         * config/aarch64/aarch64.c
22941         (aarch64_internal_mov_immediate): Add new special pattern.
22942         * config/aarch64/aarch64.md (*movdi_aarch64):
22943         Add reg/32bit const mov case.
22945 2017-07-28  Tamar Christina  <tamar.christina@arm.com>
22946             Richard Sandiford <richard.sandiford@linaro.org>
22948         * config/aarch64/aarch64.md (mov<mode>): Generalize.
22949         (*movhf_aarch64, *movsf_aarch64, *movdf_aarch64):
22950         Add integer and movi cases.
22951         (movi-split-hf-df-sf split, fp16): New.
22952         (enabled): Added TARGET_FP_F16INST.
22953         * config/aarch64/iterators.md (GPF_HF): New.
22954         * config/aarch64/predicates.md (aarch64_reg_or_fp_float): New.
22956 2017-07-28  Tamar Christina  <tamar.christina@arm.com>
22958         * config/aarch64/aarch64.c
22959         (aarch64_simd_container_mode): Add prototype.
22960         (aarch64_expand_mov_immediate): Add HI support.
22961         (aarch64_reinterpret_float_as_int, aarch64_float_const_rtx_p: New.
22962         (aarch64_can_const_movi_rtx_p): New.
22963         (aarch64_preferred_reload_class):
22964         Remove restrictions of using FP registers for certain SIMD operations.
22965         (aarch64_rtx_costs): Added new cost for CONST_DOUBLE moves.
22966         (aarch64_valid_floating_const): Add integer move validation.
22967         (aarch64_simd_imm_scalar_p): Remove.
22968         (aarch64_output_scalar_simd_mov_immediate): Generalize function.
22969         (aarch64_legitimate_constant_p): Expand list of supported cases.
22970         * config/aarch64/aarch64-protos.h
22971         (aarch64_float_const_rtx_p, aarch64_can_const_movi_rtx_p): New.
22972         (aarch64_reinterpret_float_as_int): New.
22973         (aarch64_simd_imm_scalar_p): Remove.
22974         * config/aarch64/constraints.md (Uvi): New.
22975         (Dd): Split into Ds and new Dd.
22976         * config/aarch64/aarch64.md (*movsi_aarch64):
22977         Add SIMD mov case.
22978         (*movdi_aarch64): Add SIMD mov case.
22980 2017-07-28  Bin Cheng  <bin.cheng@arm.com>
22982         * tree-predcom.c: (struct chain): Handle store-store chain in which
22983         stores for elimination only store loop invariant values.
22984         (execute_pred_commoning_chain): Ditto.
22985         (prepare_initializers_chain_store_elim): Ditto.
22986         (prepare_finalizers): Ditto.
22987         (is_inv_store_elimination_chain): New function.
22988         (initialize_root_vars_store_elim_1): New function.
22990 2017-07-28  Bin Cheng  <bin.cheng@arm.com>
22992         * tree-predcom.c: Revise general description of the pass.
22993         (enum chain_type): New enum type for store elimination.
22994         (struct chain): New field supporting store elimination.
22995         (struct component): Ditto.
22996         (dump_chain): Dump store-stores chain.
22997         (release_chain): Release resources.
22998         (split_data_refs_to_components): Compute and create component
22999         contains only stores for elimination.
23000         (get_chain_last_ref_at): New function.
23001         (make_invariant_chain): Initialization.
23002         (make_rooted_chain): Specify chain type in parameter and record it.
23003         (add_looparound_copies): Skip for store-stores chain.
23004         (determine_roots_comp): Compute type of chain and pass it to
23005         make_rooted_chain.
23006         (initialize_root_vars_store_elim_2): New function.
23007         (finalize_eliminated_stores): New function.
23008         (remove_stmt): Handle store for elimination.
23009         (execute_pred_commoning_chain): Execute predictive commoning on
23010         store-store chains.
23011         (determine_unroll_factor): Skip unroll for store-stores chain.
23012         (prepare_initializers_chain_store_elim): New function.
23013         (prepare_initializers_chain): Hanlde store-store chain.
23014         (prepare_finalizers_chain, prepare_finalizers): New function.
23015         (tree_predictive_commoning_loop): Return integer value indicating
23016         if loop is unrolled or lcssa form is corrupted.
23017         (tree_predictive_commoning): Rewrite for lcssa form if necessary.
23019 2017-07-28  Bin Cheng  <bin.cheng@arm.com>
23021         * tree-predcom.c (initialize_root): Delete.
23022         (execute_pred_commoning_chain): Initialize root vars and replace
23023         reference of non-combined chain directly, rather than call above
23024         function.
23026 2017-07-28  Bin Cheng  <bin.cheng@arm.com>
23028         * tree-predcom.c (ref_at_iteration): Add parameter NITERS.  Compute
23029         memory reference to DR at (NITERS + ITERS)-th iteration of loop.
23031 2017-07-28  Bin Cheng  <bin.cheng@arm.com>
23033         * tree-predcom.c (struct chain): New field init_seq.
23034         (release_chain): Release init_seq.
23035         (prepare_initializers_chain): Record intialization stmts in above
23036         field.
23037         (insert_init_seqs): New function.
23038         (tree_predictive_commoning_loop): Call insert_init_seqs.
23040 2017-07-28  Bin Cheng  <bin.cheng@arm.com>
23042         * tree-predcom.c (determine_roots_comp): Skip trivial components.
23044 2017-07-28  Richard Biener  <rguenther@suse.de>
23046         * match.pd: Remove superfluous :c.
23047         * genmatch.c (simplify::id): Add member.
23048         (lower_commutative, lower_opt_convert, lower_cond, lower_for):
23049         Copy id.
23050         (current_id): New global.
23051         (dt_node::parent): Move from ...
23052         (dt_operand::parent): ... here.  Add for_id member.
23053         (is_a_helper <dt_operand *>::test): DT_TRUE is also a dt_operand.
23054         (decision_tree::find_node): Relax order requirement when
23055         merging DT_TRUE nodes to ones inbetween the current simplify
23056         and the one we try to merge with.  Add diagnostic whenever
23057         we need to enforce pattern order by not merging.
23058         (decision_tree::insert): Set current_id.
23059         (decision_tree::print_node): Dump parent node and for_id.
23060         (parser::last_id): Add member.
23061         (parser::push_simplify): Assign unique id.
23062         (parser::parser): Initialize last_id.
23064 2017-07-28  Martin Liska  <mliska@suse.cz>
23066         PR sanitizer/81340
23067         * sanopt.c (sanitize_rewrite_addressable_params): Set VALUE_EXPR after
23068         gimple_build_debug_bind.
23070 2017-07-28  Richard Biener  <rguenther@suse.de>
23072         PR tree-optimization/81502
23073         * match.pd: Add pattern combining BIT_INSERT_EXPR with
23074         BIT_FIELD_REF.
23075         * tree-cfg.c (verify_expr): Verify types of BIT_FIELD_REF
23076         size/pos operands.
23077         (verify_gimple_assign_ternary): Likewise for BIT_INSERT_EXPR pos.
23078         * gimple-fold.c (maybe_canonicalize_mem_ref_addr): Use bitsizetype
23079         for BIT_FIELD_REF args.
23080         * fold-const.c (make_bit_field_ref): Likewise.
23081         * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
23083 2017-07-28  Jakub Jelinek  <jakub@redhat.com>
23085         PR sanitizer/80998
23086         * sanopt.c (pass_sanopt::execute): Handle IFN_UBSAN_PTR.
23087         * tree-ssa-alias.c (call_may_clobber_ref_p_1): Likewise.
23088         * flag-types.h (enum sanitize_code): Add SANITIZER_POINTER_OVERFLOW.
23089         Or it into SANITIZER_UNDEFINED.
23090         * ubsan.c: Include gimple-fold.h and varasm.h.
23091         (ubsan_expand_ptr_ifn): New function.
23092         (instrument_pointer_overflow): New function.
23093         (maybe_instrument_pointer_overflow): New function.
23094         (instrument_object_size): Formatting fix.
23095         (pass_ubsan::execute): Call instrument_pointer_overflow
23096         and maybe_instrument_pointer_overflow.
23097         * internal-fn.c (expand_UBSAN_PTR): New function.
23098         * ubsan.h (ubsan_expand_ptr_ifn): Declare.
23099         * sanitizer.def (__ubsan_handle_pointer_overflow,
23100         __ubsan_handle_pointer_overflow_abort): New builtins.
23101         * tree-ssa-tail-merge.c (merge_stmts_p): Handle IFN_UBSAN_PTR.
23102         * internal-fn.def (UBSAN_PTR): New internal function.
23103         * opts.c (sanitizer_opts): Add pointer-overflow.
23104         * lto-streamer-in.c (input_function): Handle IFN_UBSAN_PTR.
23105         * fold-const.c (build_range_check): Compute pointer range check in
23106         integral type if pointer arithmetics would be needed.  Formatting
23107         fixes.
23109 2017-07-28  Martin Liska  <mliska@suse.cz>
23111         PR sanitizer/81460
23112         * sanopt.c (sanitize_rewrite_addressable_params): Do not rewrite
23113         parameters that are of a variable-length.
23115 2017-07-28  Sebastian Huber  <sebastian.huber@embedded-brains.de>
23117         * config.gcc (powerpc-*-rtems*): Remove rs6000/eabi.h.  Add
23118         rs6000/biarch64.h.
23119         * config/rs6000/rtems.h (ASM_DECLARE_FUNCTION_SIZE): New macro.
23120         (ASM_OUTPUT_SPECIAL_POOL_ENTRY_P): Likewise.
23121         (CRT_CALL_STATIC_FUNCTION): Likewise.
23122         (ASM_DEFAULT_SPEC): New define.
23123         (ASM_SPEC32): Likewise.
23124         (ASM_SPEC64): Likewise.
23125         (ASM_SPEC_COMMON): Likewise.
23126         (ASM_SPEC): Likewise.
23127         (INVALID_64BIT): Likewise.
23128         (LINK_OS_DEFAULT_SPEC): Likewise.
23129         (LINK_OS_SPEC32): Likewise.
23130         (LINK_OS_SPEC64): Likewise.
23131         (POWERPC_LINUX): Likewise.
23132         (PTRDIFF_TYPE): Likewise.
23133         (RESTORE_FP_PREFIX): Likewise.
23134         (RESTORE_FP_SUFFIX): Likewise.
23135         (SAVE_FP_PREFIX): Likewise.
23136         (SAVE_FP_SUFFIX): Likewise.
23137         (SIZE_TYPE): Likewise.
23138         (SUBSUBTARGET_OVERRIDE_OPTIONS): Likewise.
23139         (TARGET_64BIT): Likewise.
23140         (TARGET_64BIT): Likewise.
23141         (TARGET_AIX): Likewise.
23142         (WCHAR_TYPE_SIZE): Likewise.
23143         (WCHAR_TYPE): Undefine.
23144         (TARGET_OS_CPP_BUILTINS): Add 64-bit PowerPC defines.
23145         (CPP_OS_DEFAULT_SPEC): Use previous CPP_OS_RTEMS_SPEC.
23146         (CPP_OS_RTEMS_SPEC): Delete.
23147         (SUBSUBTARGET_EXTRA_SPECS): Remove cpp_os_rtems.  Add
23148         asm_spec_common, asm_spec32, asm_spec64, link_os_spec32, and
23149         link_os_spec64.
23150         * config/rs6000/t-rtems: Add mcpu=e6500/m64 multilibs.
23152 2017-07-28  Jakub Jelinek  <jakub@redhat.com>
23154         PR tree-optimization/81578
23155         * tree-parloops.c (build_new_reduction): Bail out if
23156         reduction_code isn't one of the standard OpenMP reductions.
23157         Move the details printing after that decision.
23159 2017-07-27  Peter Bergner  <bergner@vnet.ibm.com>
23161         * config/rs6000/predicates.md (volatile_mem_operand): Remove code
23162         related to reload_in_progress.
23163         (splat_input_operand): Likewise.
23164         * config/rs6000/rs6000-protos.h (rs6000_secondary_memory_needed_rtx):
23165         Delete prototype.
23166         * config/rs6000/rs6000.c (machine_function): Remove sdmode_stack_slot
23167         field.
23168         (TARGET_EXPAND_TO_RTL_HOOK): Delete.
23169         (TARGET_INSTANTIATE_DECLS): Likewise.
23170         (legitimate_indexed_address_p): Delete reload_in_progress code.
23171         (rs6000_debug_legitimate_address_p): Likewise.
23172         (rs6000_eliminate_indexed_memrefs): Likewise.
23173         (rs6000_emit_le_vsx_store): Likewise.
23174         (rs6000_emit_move_si_sf_subreg): Likewise.
23175         (rs6000_emit_move): Likewise.
23176         (register_to_reg_type): Likewise.
23177         (rs6000_pre_atomic_barrier): Likewise.
23178         (rs6000_machopic_legitimize_pic_address): Likewise.
23179         (rs6000_allocate_stack_temp): Likewise.
23180         (rs6000_address_for_fpconvert): Likewise.
23181         (rs6000_address_for_altivec): Likewise.
23182         (rs6000_secondary_memory_needed_rtx): Delete function.
23183         (rs6000_check_sdmode): Likewise.
23184         (rs6000_alloc_sdmode_stack_slot): Likewise.
23185         (rs6000_instantiate_decls): Likewise.
23186         * config/rs6000/rs6000.h (SECONDARY_MEMORY_NEEDED_RTX): Delete.
23187         * config/rs6000/rs6000.md (splitter for *movsi_got_internal):
23188         Delete reload_in_progress.
23189         (*vec_reload_and_plus_<mptrsize>): Likewise.
23190         * config/rs6000/vsx.md (vsx_mul_v2di): Likewise.
23191         (vsx_div_v2di): Likewise.
23192         (vsx_udiv_v2di): Likewise.
23194 2017-07-27  Peter Bergner  <bergner@vnet.ibm.com>
23196         * config/rs6000/rs6000.opt (mlra): Replace with stub.
23197         * config/rs6000/rs6000-cpus.def (POWERPC_MASKS): Delete OPTION_MASK_LRA.
23198         * config/rs6000/rs6000.c (TARGET_LRA_P): Delete.
23199         (rs6000_debug_reg_global): Delete print of LRA status.
23200         (rs6000_option_override_internal): Delete dead LRA related code.
23201         (rs6000_lra_p): Delete function.
23202         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mlra.
23204 2017-07-27  Sebastian Huber  <sebastian.huber@embedded-brains.de>
23206         * config.gcc (riscv*-*-elf*): Add (riscv*-*-rtems*).
23207         * config/riscv/rtems.h: New file.
23209 2017-07-27  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
23210             Sudakshina Das  <sudi.das@arm.com>
23212         * config/aarch64/aarch64.md
23213         (define_split for and<mode>3nr_compare): Move
23214         non aarch64_logical_operand to a register.
23215         (define_split for and_<SHIFT:optab><mode>3nr_compare0): Move non
23216         register immediate operand to a register.
23217         * config/aarch64/predicates.md (aarch64_mov_imm_operand): New.
23219 2017-07-27  Peter Bergner  <bergner@vnet.ibm.com>
23221         PR middle-end/81564
23222         * tree-cfg.c (group_case_labels_stmt): Handle already deleted blocks.
23224 2017-07-27  Richard Biener  <rguenther@suse.de>
23226         PR tree-optimization/81573
23227         PR tree-optimization/81494
23228         * tree-vect-loop.c (vect_create_epilog_for_reduction): Handle
23229         multi defuse cycle case.
23231 2017-07-27  Richard Biener  <rguenther@suse.de>
23233         PR tree-optimization/81571
23234         * tree-vect-slp.c (vect_build_slp_tree): Properly verify reduction
23235         PHIs.
23237 2017-07-27  Eric Botcazou  <ebotcazou@adacore.com>
23239         * config/sparc/sparc.c (sparc_option_override): Set MASK_FSMULD flag
23240         earlier and only if MASK_FPU is set.  Adjust formatting.
23242 2017-07-27  Martin Liska  <mliska@suse.cz>
23244         * opt-functions.awk: Add validation of value of Init.
23245         * optc-gen.awk: Pass new argument.
23247 2017-07-27  Martin Liska  <mliska@suse.cz>
23249         * auto-profile.c (autofdo_source_profile::update_inlined_ind_target):
23250         Fix wrong condition.
23252 2017-07-27  Martin Liska  <mliska@suse.cz>
23254         * auto-profile.c (afdo_annotate_cfg): Assign zero counts to
23255         BBs and edges seen by autoFDO.
23257 2017-07-27  Richard Biener  <rguenther@suse.de>
23259         PR tree-optimization/81502
23260         * tree-ssa.c (non_rewritable_lvalue_p): Handle BIT_INSERT_EXPR
23261         with incompatible but same sized type.
23262         (execute_update_addresses_taken): Likewise.
23264 2017-07-27  James Greenhalgh  <james.greenhalgh@arm.com>
23266         * tree-ssa-loop-ch.c (pass_ch::process_loop_p): Guard on
23267         flag_tree_loop_vectorize rather than flag_tree_vectorize.
23269 2017-07-27  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
23271         PR target/81534
23272         * config/s390/s390.md ("*atomic_compare_and_swap<mode>_1")
23273         ("*atomic_compare_and_swapdi_2", "*atomic_compare_and_swapsi_3"):
23274         Change s_operand to memory_operand.
23276 2017-07-27  Richard Sandiford  <richard.sandiford@linaro.org>
23278         * config/rs6000/rs6000-protos.h (rs6000_emit_le_vsx_permute): Declare.
23279         * config/rs6000/rs6000.c (rs6000_gen_le_vsx_permute): Replace with...
23280         (rs6000_emit_le_vsx_permute): ...this.  Take the destination as input.
23281         Emit instructions rather than returning an expression.  Handle TFmode
23282         and KFmode by casting to TImode.
23283         (rs6000_emit_le_vsx_load): Update to use rs6000_emit_le_vsx_permute.
23284         (rs6000_emit_le_vsx_store): Likewise.
23285         * config/rs6000/vsx.md (VSX_TI): New iterator.
23286         (*vsx_le_permute_<mode>): Use it instead of VSX_LE_128.
23287         (*vsx_le_undo_permute_<mode>): Likewise.
23288         (*vsx_le_perm_load_<mode>): Use rs6000_emit_le_vsx_permute to
23289         emit the split sequence.
23290         (*vsx_le_perm_store_<mode>): Likewise.
23292 2017-07-27  Jakub Jelinek  <jakub@redhat.com>
23294         PR tree-optimization/81555
23295         PR tree-optimization/81556
23296         * tree-ssa-reassoc.c (rewrite_expr_tree): Add NEXT_CHANGED argument,
23297         if true, force CHANGED for the recursive invocation.
23298         (reassociate_bb): Remember original length of ops array, pass
23299         len != orig_len as NEXT_CHANGED in rewrite_expr_tree call.
23301         * attribs.c (decl_attributes): Imply noinline, noclone and no_icf
23302         attributes for noipa attribute.  For naked attribute use
23303         lookup_attribute first before lookup_attribute_spec.
23304         * final.c (rest_of_handle_final): Disable IPA RA for functions with
23305         noipa attribute.
23306         * ipa-visibility.c (non_local_p): Fix comment typos.  Return true
23307         for functions with noipa attribute.
23308         (cgraph_externally_visible_p): Return true for functions with noipa
23309         attribute.
23310         * cgraph.c (cgraph_node::get_availability): Return AVAIL_INTERPOSABLE
23311         for functions with noipa attribute.
23312         * doc/extend.texi: Document noipa function attribute.
23313         * tree-ssa-structalias.c (refered_from_nonlocal_fn): Set *nonlocal_p
23314         also for functions with noipa attribute.
23315         (ipa_pta_execute): Set nonlocal_p also for nodes with noipa attribute.
23317 2017-07-26  Andrew Pinski  <apinski@cavium.com>
23319         * config/aarch64/aarch64.c (thunderx_vector_cost): Decrease cost of
23320         vec_unalign_load_cost and vec_unalign_store_cost.
23322 2017-07-26  Michael Meissner  <meissner@linux.vnet.ibm.com>
23324         * config/rs6000/rs6000-cpus.def (ISA_2_7_MASKS_SERVER): Delete
23325         -mvsx-small-integer option.
23326         (ISA_3_0_MASKS_IEEE): Likewise.
23327         (OTHER_VSX_VECTOR_MASKS): Likewise.
23328         (POWERPC_MASKS): Likewise.
23329         * config/rs6000/rs6000.opt (-mvsx-small-integer): Likewise.
23330         * config/rs6000/rs6000.c (rs6000_hard_regno_mode_ok): Simplify
23331         code, only testing for DImode being allowed in non-VSX floating
23332         point registers.
23333         (rs6000_init_hard_regno_mode_ok): Change TARGET_VSX_SMALL_INTEGER
23334         to TARGET_P8_VECTOR test.  Remove redundant VSX test inside of
23335         another VSX test.
23336         (rs6000_option_override_internal): Delete -mvsx-small-integer.
23337         (rs6000_expand_vector_set): Change TARGET_VSX_SMALL_INTEGER to
23338         TARGET_P8_VECTOR test.
23339         (rs6000_secondary_reload_simple_move): Likewise.
23340         (rs6000_preferred_reload_class): Delete TARGET_VSX_SMALL_INTEGER,
23341         since TARGET_P9_VECTOR was already tested.
23342         (rs6000_opt_masks): Remove -mvsx-small-integer.
23343         * config/rs6000/vsx.md (vsx_extract_<mode>): Delete
23344         TARGET_VSX_SMALL_INTEGER, since a test for TARGET_P9_VECTOR was
23345         used.
23346         (vsx_extract_<mode>_p9): Delete TARGET_VSX_SMALL_INTEGER, since a
23347         test for TARGET_VEXTRACTUB was used, and that uses
23348         TARGET_P9_VECTOR.
23349         (p9 extract splitter): Likewise.
23350         (vsx_extract_<mode>_di_p9): Likewise.
23351         (vsx_extract_<mode>_store_p9): Likewise.
23352         (vsx_extract_si): Delete TARGET_VSX_SMALL_INTEGER, since a test
23353         for TARGET_P9_VECTOR was used.  Delete code that is now dead with
23354         the elimination of TARGET_VSX_SMALL_INTEGER.
23355         (vsx_extract_<mode>_p8): Likewise.
23356         (vsx_ext_<VSX_EXTRACT_I:VS_scalar>_fl_<FL_CONV:mode>): Likewise.
23357         (vsx_ext_<VSX_EXTRACT_I:VS_scalar>_ufl_<FL_CONV:mode>): Likewise.
23358         (vsx_set_<mode>_p9): Likewise.
23359         (vsx_set_v4sf_p9): Likewise.
23360         (vsx_set_v4sf_p9_zero): Likewise.
23361         (vsx_insert_extract_v4sf_p9): Likewise.
23362         (vsx_insert_extract_v4sf_p9_2): Likewise.
23363         * config/rs6000/rs6000.md (sign extend splitter): Change
23364         TARGET_VSX_SMALL_INTEGER to TARGET_P8_VECTOR test.
23365         (floatsi<mode>2_lfiwax_mem): Likewise.
23366         (floatunssi<mode>2_lfiwzx_mem): Likewise.
23367         (float<QHI:mode><FP_ISA3:mode>2): Delete TARGET_VSX_SMALL_INTEGER,
23368         since a test for TARGET_P9_VECTOR was used.
23369         (float<QHI:mode><FP_ISA3:mode>2_internal): Likewise.
23370         (floatuns<QHI:mode><FP_ISA3:mode>2): Likewise.
23371         (floatuns<QHI:mode><FP_ISA3:mode>2_internal): Likewise.
23372         (fix_trunc<mode>si2): Change TARGET_VSX_SMALL_INTEGER to
23373         TARGET_P8_VECTOR test.
23374         (fix_trunc<mode>si2_stfiwx): Likewise.
23375         (fix_trunc<mode>si2_internal): Likewise.
23376         (fix_trunc<SFDF:mode><QHI:mode>2): Delete
23377         TARGET_VSX_SMALL_INTEGER, since a test for TARGET_P9_VECTOR was
23378         used.
23379         (fix_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
23380         (fixuns_trunc<mode>si2): Change TARGET_VSX_SMALL_INTEGER to
23381         TARGET_P8_VECTOR test.
23382         (fixuns_trunc<mode>si2_stfiwx): Likewise.
23383         (fixuns_trunc<SFDF:mode><QHI:mode>2): Delete
23384         TARGET_VSX_SMALL_INTEGER, since a test for TARGET_P9_VECTOR was
23385         used.
23386         (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
23387         (fctiw<u>z_<mode>_smallint): Delete TARGET_VSX_SMALL_INTEGER,
23388         since a test for TARGET_P9_VECTOR was used.
23389         (splitter for loading small constants): Likewise.
23391 2017-07-26  Andrew Pinski  <apinski@cavium.com>
23393         * config/aarch64/aarch64.c (thunderx_vector_cost): Fix
23394         vec_fp_stmt_cost.
23396 2017-07-26  H.J. Lu  <hongjiu.lu@intel.com>
23398         PR target/81563
23399         * config/i386/i386.c (sp_valid_at): Properly check CFA offset.
23400         (fp_valid_at): Likewise.
23402 2017-07-26  James Greenhalgh  <james.greenhalgh@arm.com>
23404         * config/aarch64/aarch64.c (cortexa57_addrcost_table): Remove.
23405         (qdf24xx_addrcost_table): Likewise.
23406         (cortexa57_tunings): Update to use generic_branch_cost.
23407         (cortexa72_tunings): Likewise.
23408         (cortexa73_tunings): Likewise.
23409         (qdf24xx_tunings): Likewise.
23411 2017-07-26  James Greenhalgh  <james.greenhalgh@arm.com>
23413         * config/aarch64/aarch64.c (cortexa57_branch_cost): Remove.
23414         (thunderx2t99_branch_cost): Likewise.
23415         (cortexa35_tunings): Update to use generic_branch_cost.
23416         (cortexa53_tunings): Likewise.
23417         (cortexa57_tunings): Likewise.
23418         (cortexa72_tunings): Likewise.
23419         (cortexa73_tunings): Likewise.
23420         (thunderx2t99_tunings): Likewise.
23422 2017-07-26  Sebastian Huber  <sebastian.huber@embedded-brains.de>
23424         * config/sparc/sparc.c (dump_target_flag_bits): Dump MASK_FSMULD.
23425         (sparc_option_override): Honour MASK_FSMULD.
23426         * config/sparc/sparc.h (MASK_FEATURES): Add MASK_FSMULD.
23427         * config/sparc/sparc.md (muldf3_extend): Use TARGET_FSMULD.
23428         * config/sparc/sparc.opt (mfsmuld): New option.
23429         * doc/invoke.texi (mfsmuld): Document option.
23431 2017-07-26  Marek Polacek  <polacek@redhat.com>
23433         PR middle-end/70992
23434         * tree.c (build2_stat): Don't set TREE_CONSTANT on divisions by zero.
23436 2017-07-26  Richard Biener  <rguenther@suse.de>
23438         * gimple-match-head.c (do_valueize): Return OP if valueize
23439         returns NULL_TREE.
23440         (get_def): New helper to get at the def stmt of a SSA name
23441         if valueize allows.
23442         * genmatch.c (dt_node::gen_kids_1): Use get_def instead of
23443         do_valueize to get at the def stmt.
23444         (dt_operand::gen_gimple_expr): Simplify do_valueize calls.
23446 2017-07-26  Wilco Dijkstra  <wdijkstr@arm.com>
23448         PR middle-end/46932
23449         * auto-inc-dec.c (parse_add_or_inc): Block autoinc on sfp.
23451 2017-07-26  Martin Liska  <mliska@suse.cz>
23453         PR sanitize/81186
23454         * function.c (expand_function_start): Make expansion of
23455         nonlocal_goto_save_area after parm_birth_insn.
23457 2017-07-26  Sebastian Huber  <sebastian.huber@embedded-brains.de>
23459         * config/sparc/sparc.c (sparc_option_override): Remove MASK_FPU
23460         from all CPU target flags enable members.
23462 2017-07-26  Richard Biener  <rguenther@suse.de>
23464         * genmatch.c (dt_simplify::gen): Make iterator vars const.
23465         (decision_tree::gen): Make 'type' const.
23466         (write_predicate): Likewise.
23468 2017-07-24  Michael Meissner  <meissner@linux.vnet.ibm.com>
23470         * config/rs6000/rs6000.c (rs6000_init_hard_regno_mode_ok):
23471         Eliminate TARGET_UPPER_REGS_{DF,DI,SF} usage.
23472         (rs6000_option_override_internal): Likewise.
23473         (rs6000_expand_vector_set): Likewise.
23474         * config/rs6000/rs6000.h (TARGET_UPPER_REGS_DF): Delete.
23475         (TARGET_UPPER_REGS_SF): Likewise.
23476         (TARGET_UPPER_REGS_DI): Likewise.
23477         (TARGET_VEXTRACTUB): Eliminate TARGET_UPPER_REGS_{DF,DI,SF}.
23478         (TARGET_DIRECT_MOVE_64BIT): Likewise.
23479         * config/rs6000/rs6000.md (ALTIVEC_DFORM): Likewise.
23480         (float<QHI:mode><FP_ISA3:mode>2_internal): Likewise.
23481         (Splitters for DI constants in Altivec registers): Likewise.
23482         * config/rs6000/vsx.md (vsx_set_<mode>_p9): Likewise.
23483         (vsx_set_v4sf_p9): Likewise.
23484         (vsx_set_v4sf_p9_zero): Likewise.
23485         (vsx_insert_extract_v4sf_p9): Likewise.
23486         (vsx_insert_extract_v4sf_p9_2): Likewise.
23488 2017-07-25  Carl Love  <cel@us.ibm.com>
23490         * doc/extend.texi: Update the built-in documentation file for the
23491         existing built-in functions
23492         vector signed char vec_cnttz (vector signed char);
23493         vector unsigned char vec_cnttz (vector unsigned char);
23494         vector signed short vec_cnttz (vector signed short);
23495         vector unsigned short vec_cnttz (vector unsigned short);
23496         vector signed int vec_cnttz (vector signed int);
23497         vector unsigned int vec_cnttz (vector unsigned int);
23498         vector signed long long vec_cnttz (vector signed long long);
23499         vector unsigned long long vec_cnttz (vector unsigned long long);
23501 2017-07-25  Andrew Pinski  <apinski@cavium.com>
23503         * tree-ssa-uninit.c (warn_uninitialized_vars): Don't warn about memory
23504         accesses where the use is for the first operand of a BIT_INSERT.
23506 2017-07-25  Jim Wilson  <jim.wilson@linaro.org>
23508         PR bootstrap/81521
23509         * config/i386/winnt-cxx.c (i386_pe_adjust_class_at_definition): Look
23510         for FUNCTION_DECLs in TYPE_FIELDS rather than TYPE_METHODS.
23512 2017-07-25  Jim Wilson  <jim.wilson@linaro.org>
23514         * config/i386/gstabs.h: Delete.
23515         * config/i386/openbsd.h, config/i386/t-openbsd: Likewise.
23517 2017-07-25  Uros Bizjak  <ubizjak@gmail.com>
23519         * config/i386/i386.c (ix86_decompose_address): Do not check for
23520         register RTX when looking at index_reg or base_reg.
23521         * config/i386/i386.h (INCOMING_RETURN_ADDR_RTX): Use stack_pointer_rtx.
23523 2017-07-25  Eric Botcazou  <ebotcazou@adacore.com>
23525         * gimple.c (gimple_assign_set_rhs_with_ops): Do not ask gsi_replace
23526         to update EH info here.
23528 2017-07-25  Alexander Monakov  <amonakov@ispras.ru>
23530         * match.pd ((X * CST1) * CST2): Simplify to X * (CST1 * CST2).
23532 2017-07-25  Alexander Monakov  <amonakov@ispras.ru>
23534         * match.pd ((X * CST) * Y): Reassociate to (X * Y) * CST.
23536 2017-07-25  Torsten Duwe  <duwe@suse.de>
23538         * common.opt: Introduce -fpatchable-function-entry
23539         command line option, and its variables function_entry_patch_area_size
23540         and function_entry_patch_area_start.
23541         * opts.c (common_handle_option): Add -fpatchable_function_entry_ case,
23542         including a two-value parser.
23543         * target.def (print_patchable_function_entry): New target hook.
23544         * targhooks.h (default_print_patchable_function_entry): New function.
23545         * targhooks.c (default_print_patchable_function_entry): Likewise.
23546         * toplev.c (process_options): Switch off IPA-RA if
23547         patchable function entries are being generated.
23548         * varasm.c (assemble_start_function): Look at the
23549         patchable-function-entry command line switch and current
23550         function attributes and maybe generate NOP instructions by
23551         calling the print_patchable_function_entry hook.
23552         * doc/extend.texi: Document patchable_function_entry attribute.
23553         * doc/invoke.texi: Document -fpatchable_function_entry
23554         command line option.
23555         * doc/tm.texi.in (TARGET_ASM_PRINT_PATCHABLE_FUNCTION_ENTRY):
23556         New target hook.
23557         * doc/tm.texi: Re-generate.
23559 2017-07-25  Jakub Jelinek  <jakub@redhat.com>
23561         PR target/81532
23562         * config/i386/constraints.md (Yd, Ye): Use ALL_SSE_REGS for
23563         TARGET_AVX512DQ rather than TARGET_AVX512BW.
23565 2017-07-25  Tamar Christina  <tamar.christina@arm.com>
23567         * config/arm/parsecpu.awk (all_cores): Remove duplicates.
23569 2017-07-25  Richard Biener  <rguenther@suse.de>
23571         PR tree-optimization/81455
23572         * tree-ssa-loop-unswitch.c (find_loop_guard): Make sure to
23573         not walk in cycles when looking for guards.
23575 2017-07-25  Richard Biener  <rguenther@suse.de>
23577         PR tree-optimization/81529
23578         * tree-vect-stmts.c (process_use): Disregard live induction PHIs
23579         when optimizing backedge uses.
23581 2017-07-25  David Edelsohn  <dje.gcc@gmail.com>
23583         * dwarf2asm.c (dw2_asm_output_nstring): Encode double quote
23584         character for AIX.
23585         * dwarf2out.c (output_macinfo): Copy debug_line_section_label
23586         to dl_section_ref.  On AIX, append an expression to subtract
23587         the size of the section length to dl_section_ref.
23589 2017-07-25  Segher Boessenkool  <segher@kernel.crashing.org>
23591         * configure.ac: If any of the config.* scripts fail, exit 1.
23592         * configure: Regenerate.
23594 2017-07-25  Richard Biener  <rguenther@suse.de>
23596         PR middle-end/81546
23597         * tree-ssa-operands.c (verify_imm_links): Remove cap on number
23598         of immediate uses, be more verbose on errors.
23600 2017-07-25  Richard Biener  <rguenther@suse.de>
23602         PR tree-optimization/81510
23603         * tree-vect-loop.c (vect_is_simple_reduction): When the
23604         reduction stmt is not inside the loop bail out.
23606 2017-07-25  Richard Biener  <rguenther@suse.de>
23608         PR tree-optimization/81303
23609         * tree-vect-loop-manip.c (vect_loop_versioning): Build
23610         profitability check against LOOP_VINFO_NITERSM1.
23612 2017-07-25  Alexander Monakov  <amonakov@ispras.ru>
23614         * domwalk.c (cmp_bb_postorder): Simplify.
23615         (sort_bbs_postorder): New function.  Use it...
23616         (dom_walker::walk): ...here to optimize common cases.
23618 2017-07-25  Martin Liska  <mliska@suse.cz>
23620         PR ipa/81520
23621         * ipa-visibility.c (function_and_variable_visibility): Make the
23622         redirection just on target that supports aliasing.
23623         Fix GNU coding style.
23625 2017-07-25  Sebastian Huber  <sebastian.huber@embedded-brains.de>
23627         PR libgcc/61152
23628         * config/aarch64/rtems.h: Add GCC Runtime Library Exception.
23629         Format changes.
23630         * config/arm/rtems.h: Likewise.
23631         * config/bfin/rtems.h: Likewise.
23632         * config/i386/rtemself.h: Likewise.
23633         * config/lm32/rtems.h: Likewise.
23634         * config/m32c/rtems.h: Likewise.
23635         * config/m68k/rtemself.h: Likewise.
23636         * config/microblaze/rtems.h: Likewise.
23637         * config/mips/rtems.h: Likewise.
23638         * config/moxie/rtems.h: Likewise.
23639         * config/nios2/rtems.h: Likewise.
23640         * config/powerpcspe/rtems.h: Likewise.
23641         * config/rs6000/rtems.h: Likewise.
23642         * config/rtems.h: Likewise.
23643         * config/sh/rtems.h: Likewise.
23644         * config/sh/rtemself.h: Likewise.
23645         * config/sparc/rtemself.h: Likewise.
23647 2017-07-25  Georg-Johann Lay  <avr@gjlay.de>
23649         PR 81487
23650         * hsa-brig.c (brig_init): Use xasprintf instead of asprintf.
23651         * gimple-pretty-print.c (dump_profile, dump_probability): Same.
23652         * tree-ssa-structalias.c (alias_get_name): Same.
23654 2017-07-25  Bin Cheng  <bin.cheng@arm.com>
23656         PR target/81414
23657         * config/aarch64/cortex-a57-fma-steering.c (analyze): Skip fmul/fmac
23658         instructions if no du chain is found.
23660 2017-07-25  Georg-Johann Lay  <avr@gjlay.de>
23662         * config/avr/avr-log.c (avr_log_vadump) ['T']: Print NULL-TREE.
23664 2017-07-25  Richard Biener  <rguenther@suse.de>
23666         PR middle-end/81505
23667         * fold-const.c (fold_negate_const): TREE_OVERFLOW should be
23668         sticky.
23670 2017-07-24  Michael Meissner  <meissner@linux.vnet.ibm.com>
23672         * config/rs6000/rs6000-cpus.def (ISA_2_6_MASKS_SERVER): Delete
23673         upper-regs options.
23674         (ISA_2_7_MASKS_SERVER): Likewise.
23675         (ISA_3_0_MASKS_IEEE): Likewise.
23676         (OTHER_P8_VECTOR_MASKS): Likewise.
23677         (OTHER_VSX_VECTOR_MASKS): Likewise.
23678         (POWERPC_MASKS): Likewise.
23679         (power7 cpu): Use ISA_2_6_MASKS_SERVER instead of using a
23680         duplicate list of options.
23681         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove
23682         explicit -mupper-regs options.
23683         * config/rs6000/rs6000.opt (-mvsx-scalar-memory): Delete
23684         -mupper-regs* options.  Delete -mvsx-scalar-memory, which was an
23685         alias for -mupper-regs-df.
23686         * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): Likewise.
23687         (rs6000_init_hard_regno_mode_ok): Likewise.
23688         (rs6000_option_override_internal): Likewise.
23689         (rs6000_opt_masks): Likewise.
23690         * config/rs6000/rs6000.h (TARGET_UPPER_REGS_DF): Define upper regs
23691         options in terms of whether -mvsx or -mpower8-vector was used.
23692         (TARGET_UPPER_REGS_DI): Likewise.
23693         (TARGET_UPPER_REGS_SF): Likewise.
23694         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete the
23695         -mupper-regs-* options.
23697 2017-07-24  Segher Boessenkool  <segher@kernel.crashing.org>
23699         * passes.c (emergency_dump_function): Print some empty lines and a
23700         header before the RTL dump.
23702 2017-07-24  Segher Boessenkool  <segher@kernel.crashing.org>
23704         * cfgrtl.c (rtl_dump_bb): Don't call NEXT_INSN on NULL.
23706 2017-07-24  Wilco Dijkstra  <wdijkstr@arm.com>
23708         PR target/79041
23709         * config/aarch64/aarch64.c (aarch64_classify_symbol):
23710         Avoid SYMBOL_SMALL_ABSOLUTE for literals with pc-relative literals.
23712 2017-07-24  Carl Love  <cel@us.ibm.com>
23714         * config/rs6000/rs6000-c.c: Add support for built-in functions
23715         vector float vec_extract_fp32_from_shorth (vector unsigned short);
23716         vector float vec_extract_fp32_from_shortl (vector unsigned short);
23717         * config/rs6000/altivec.h (vec_extract_fp_from_shorth,
23718         vec_extract_fp_from_shortl): Add defines for the two builtins.
23719         * config/rs6000/rs6000-builtin.def (VEXTRACT_FP_FROM_SHORTH,
23720         VEXTRACT_FP_FROM_SHORTL): Add BU_P9V_OVERLOAD_1 and BU_P9V_VSX_1
23721         new builtins.
23722         * config/rs6000/vsx.md vsx_xvcvhpsp): Add define_insn.
23723         (vextract_fp_from_shorth, vextract_fp_from_shortl): Add define_expands.
23724         * doc/extend.texi: Update the built-in documentation file for the
23725         new built-in function.
23727 2017-07-24  Jakub Jelinek  <jakub@redhat.com>
23729         PR bootstrap/81521
23730         * tree.def: Remove TYPE_METHODS documentation, adjust TYPE_FIELDS
23731         documentation.
23732         * doc/generic.texi: Likewise.
23733         * config/i386/winnt-cxx.c (i386_pe_adjust_class_at_definition): Look
23734         for FUNCTION_DECLs in TYPE_FIELDS rather than TYPE_METHODS.
23736 2017-07-24  Jackson Woodruff  <jackson.woodruff@arm.com>
23738         * config/aarch64/aarch64-simd.md (aarch64_mla_elt_merge<mode>): New.
23739         (aarch64_mls_elt_merge<mode>): Likewise.
23741 2017-07-23  Krister Walfridsson  <krister.walfridsson@gmail.com>
23743         * config.gcc (*-*-netbsd*): Remove check for NetBSD versions not
23744         having __cxa_atexit.
23746 2017-07-23  Michael Collison  <michael.collison@arm.com>
23748         * config/arm/arm.c (arm_option_override): Deprecate
23749         use of -mstructure-size-boundary.
23750         * config/arm/arm.opt: Deprecate -mstructure-size-boundary.
23751         * doc/invoke.texi: Deprecate -mstructure-size-boundary.
23753 2017-07-23  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
23755         PR target/80695
23756         * config/rs6000/rs6000.c (rs6000_builtin_vectorization_cost):
23757         Reduce cost estimate for direct moves.
23759 2017-07-23  Uros Bizjak  <ubizjak@gmail.com>
23761         PR target/80569
23762         * config/i386/i386.c (ix86_option_override_internal): Disable
23763         BMI, BMI2 and TBM instructions for -m16.
23765 2017-07-21  Carl Love  <cel@us.ibm.com>
23767         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
23768         ALTIVEC_BUILTIN_VMULESW, ALTIVEC_BUILTIN_VMULEUW,
23769         ALTIVEC_BUILTIN_VMULOSW, ALTIVEC_BUILTIN_VMULOUW entries.
23770         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin,
23771         builtin_function_type): Add ALTIVEC_BUILTIN_* case statements.
23772         * config/rs6000/altivec.md (MVULEUW, VMULESW, VMULOUW,
23773         VMULOSW): New enum "unspec" values.
23774         (altivec_vmuleuw, altivec_vmulesw, altivec_vmulouw,
23775         altivec_vmulosw): New patterns.
23776         * config/rs6000/rs6000-builtin.def (VMLEUW, VMULESW, VMULOUW,
23777         VMULOSW): Add definitions.
23779 2017-07-21  Jim Wilson  <jim.wilson@linaro.org>
23781         * config/aarch64/aarch64-cores.def (falkor): Add AARCH64_FL_RDMA.
23782         (qdf24xx): Likewise.
23783         * config/aarch64/aarch64-options-extensions.def (rdma); New.
23784         * config/aarch64/aarch64.h (AARCH64_FL_RDMA): New.
23785         (AARCH64_FL_V8_1): Renumber.
23786         (AARCH64_FL_FOR_ARCH8_1): Add AARCH64_FL_RDMA.
23787         (AARCH64_ISA_RDMA): Use AARCH64_FL_RDMA.
23788         * config/aarch64/arm_neon.h: Use +rdma instead of arch=armv8.1-a.
23789         * doc/invoke.texi (AArch64 Options): Mention +rmda in -march docs.  Add
23790         rdma to feature modifiers list.
23792 2017-07-21  Yury Gribov  <tetra2005@gmail.com>
23794         PR middle-end/56727
23795         * ipa-visibility (function_and_variable_visibility): Convert
23796         recursive PLT call to direct call if appropriate.
23798 2017-07-21  Andrew Pinski  <apinski@cavium.com>
23800         * tree-ssa-sccvn.c (vn_nary_op_eq): Check BIT_INSERT_EXPR's
23801         operand 1 to see if the types precision matches.
23802         * fold-const.c (operand_equal_p): Likewise.
23804 2017-07-21  Richard Biener  <rguenther@suse.de>
23806         PR tree-optimization/81303
23807         * tree-vect-data-refs.c (vect_get_peeling_costs_all_drs): Pass
23808         in datarefs vector.  Allow NULL dr0 for no peeling cost estimate.
23809         (vect_peeling_hash_get_lowest_cost): Adjust.
23810         (vect_enhance_data_refs_alignment): Likewise.  Use
23811         vect_get_peeling_costs_all_drs to compute the penalty for no
23812         peeling to match up costs.
23814 2017-07-21  Richard Biener  <rguenther@suse.de>
23816         PR tree-optimization/81500
23817         * tree-vect-loop.c (vect_is_simple_reduction): Properly fail if
23818         we didn't identify a reduction path.
23820 2017-07-21  Tom de Vries  <tom@codesourcery.com>
23821             Cesar Philippidis  <cesar@codesourcery.com>
23823         PR gcov-profile/81442
23824         * config/nvptx/nvptx.c (nvptx_goacc_reduction_init): Add missing edge
23825         probabilities.
23827 2017-07-21  Tom de Vries  <tom@codesourcery.com>
23829         PR lto/81430
23830         * config/nvptx/nvptx.c (nvptx_override_options_after_change): New
23831         function.
23832         (TARGET_OVERRIDE_OPTIONS_AFTER_CHANGE): Define to
23833         nvptx_override_options_after_change.
23835 2017-07-21  Ulrich Drepper  <drepper@redhat.com>
23837         * dwarf2out.c (output_file_names): Avoid double testing for
23838         dwarf_version >= 5.
23840 2017-07-21  Georg-Johann Lay  <avr@gjlay.de>
23842         * doc/invoke.texi (AVR Built-in Functions): Re-layout section.
23844 2016-07-21  Jan Hubicka  <hubicka@ucw.cz>
23846         * cfgcleanup.c (flow_find_cross_jump): Do not crossjump across
23847         hot/cold regions.
23848         (try_crossjump_to_edge): Do not punt on partitioned functions.
23850 2016-07-21  Jan Hubicka  <hubicka@ucw.cz>
23852         * bb-reorder.c (find_rarely_executed_basic_blocks_and_crossing_edges):
23853         Put all BBs reachable only via paths crossing cold region to cold
23854         region.
23855         * cfgrtl.c (find_bbs_reachable_by_hot_paths): New function.
23857 2016-07-21  Richard Biener  <rguenther@suse.de>
23859         PR tree-optimization/81303
23860         * tree-vect-loop.c (vect_estimate_min_profitable_iters): Take
23861         into account prologue and epilogue iterations when raising
23862         min_profitable_iters to sth at least covering one vector iteration.
23864 2017-07-21  Tamar Christina  <tamar.christina@arm.com>
23866         * config/arm/arm.c (arm_test_cpu_arch_dat):
23867         Check for overlap.
23869 2017-07-20  Nathan Sidwell  <nathan@acm.org>
23871         Remove TYPE_METHODS.
23872         * tree.h (TYPE_METHODS): Delete.
23873         * dwarf2out.c (gen_member_die): Member fns are on TYPE_FIELDS.
23874         * dbxout.c (dbxout_type_fields): Ignore FUNCTION_DECLs.
23875         (dbxout_type_methods): Scan TYPE_FIELDS.
23876         (dbxout_type): Don't check TYPE_METHODS here.
23877         * function.c (use_register_for_decl): Always ignore register for
23878         class types when not optimizing.
23879         * ipa-devirt.c (odr_types_equivalent_p): Delete TYPE_METHODS scan.
23880         * tree.c (free_lang_data_in_type): Stitch out member functions and
23881         templates from TYPE_FIELDS.
23882         (build_distinct_type_copy, verify_type_variant,
23883         verify_type): Member fns are on TYPE_FIELDS.
23884         * tree-dump.c (dequeue_and_dump): No TYPE_METHODS.
23885         * tree-pretty-print.c (dump_generic_node): Likewise.
23887 2017-07-20  Jakub Jelinek  <jakub@redhat.com>
23889         PR target/80846
23890         * config/i386/i386.c (ix86_expand_vector_init_general): Handle
23891         V2TImode and V4TImode.
23892         (ix86_expand_vector_extract): Likewise.
23893         * config/i386/sse.md (VMOVE): Enable V4TImode even for just
23894         TARGET_AVX512F, instead of only for TARGET_AVX512BW.
23895         (ssescalarmode): Handle V4TImode and V2TImode.
23896         (VEC_EXTRACT_MODE): Add V4TImode and V2TImode.
23897         (*vec_extractv2ti, *vec_extractv4ti): New insns.
23898         (VEXTRACTI128_MODE): New mode iterator.
23899         (splitter for *vec_extractv?ti first element): New.
23900         (VEC_INIT_MODE): New mode iterator.
23901         (vec_init<mode>): Consolidate 3 expanders into one using
23902         VEC_INIT_MODE mode iterator.
23904 2017-07-20  Alexander Monakov  <amonakov@ispras.ru>
23906         * lra-assigns.c (pseudo_compare_func): Fix comparison step based on
23907         non_spilled_static_chain_regno_p.
23909 2017-07-20  Alexander Monakov  <amonakov@ispras.ru>
23911         * gimple-ssa-store-merging.c (sort_by_bitpos): Return 0 on equal bitpos.
23913 2017-07-20  Jan Hubicka  <hubicka@ucw.cz>
23915         * bb-reorder.c (connect_traces): Allow copying of blocks within
23916         single partition.
23918 2017-07-20  Richard Biener  <rguenther@suse.de>
23920         * gimple.h (gimple_phi_result): Add gphi * overload.
23921         (gimple_phi_result_ptr): Likewise.
23922         (gimple_phi_arg): Likewise.  Adjust index assert to only
23923         allow actual argument accesses rather than all slots available
23924         by capacity.
23925         (gimple_phi_arg_def): Add gphi * overload.
23926         * tree-phinodes.c (make_phi_node): Initialize only actual
23927         arguments.
23928         (resize_phi_node): Clear memory not covered by old node,
23929         do not initialize excess argument slots.
23930         (reserve_phi_args_for_new_edge): Initialize new argument slot
23931         completely.
23933 2017-07-20  Bin Cheng  <bin.cheng@arm.com>
23935         PR tree-optimization/81388
23936         Revert r238585:
23937         2016-07-21  Bin Cheng  <bin.cheng@arm.com>
23939         * tree-ssa-loop-niter.c (number_of_iterations_lt_to_ne): Clean up
23940         by removing computation of may_be_zero.
23942 2017-07-18  Jan Hubicka  <hubicka@ucw.cz>
23943             Tom de Vries  <tom@codesourcery.com>
23945         PR middle-end/81030
23946         * cfgbuild.c (find_many_sub_basic_blocks): Update REG_BR_PROB note
23947         when gimple level profile disagrees with what RTL expander did.
23949 2017-07-20  Richard Biener  <rguenther@suse.de>
23951         PR tree-optimization/61171
23952         * tree-vectorizer.h (slp_instance): Add reduc_phis member.
23953         (vect_analyze_stmt): Add slp instance parameter.
23954         (vectorizable_reduction): Likewise.
23955         * tree-vect-loop.c (vect_analyze_loop_operations): Adjust.
23956         (vect_is_simple_reduction): Deal with chains not detected
23957         as SLP reduction chain, specifically not properly associated
23958         chains containing a mix of plus/minus.
23959         (get_reduction_op): Remove.
23960         (get_initial_defs_for_reduction): Simplify, pass in whether
23961         this is a reduction chain, pass in the SLP node for the PHIs.
23962         (vect_create_epilog_for_reduction): Get the SLP instance as
23963         arg and adjust.
23964         (vectorizable_reduction): Get the SLP instance as arg.
23965         During analysis remember the SLP node with the PHIs in the
23966         instance.  Simplify getting at the vectorized reduction PHIs.
23967         * tree-vect-slp.c (vect_slp_analyze_node_operations): Pass
23968         through SLP instance.
23969         (vect_slp_analyze_operations): Likewise.
23970         * tree-vect-stms.c (vect_analyze_stmt): Likewise.
23971         (vect_transform_stmt): Likewise.
23973 2017-07-20  Tom de Vries  <tom@codesourcery.com>
23975         PR tree-optimization/81489
23976         * gimple-ssa-isolate-paths.c (find_implicit_erroneous_behavior): Move
23977         read of phi arg location to before loop that modifies phi.
23979 2017-07-20  Naveen H.S  <Naveen.Hurugalawadi@cavium.com>
23981         * match.pd (((m1 >/</>=/<= m2) * d -> (m1 >/</>=/<= m2) ? d : 0):
23982         New pattern.
23984 2017-07-19  Jan Hubicka  <hubicka@ucw.cz>
23986         PR middle-end/81331
23987         * except.c (execute): Fix ordering issue.
23989 2018-07-19  Segher Boessenkool  <segher@kernel.crashing.org>
23991         PR rtl-optimization/81423
23992         * combine.c (make_compound_operation_int): Don't try to optimize
23993         the AND of a SUBREG of an LSHIFTRT if that SUBREG is paradoxical.
23995 2017-07-19  Segher Boessenkool  <segher@kernel.crashing.org>
23997         PR rtl-optimization/81423
23998         * simplify-rtx.c (simplify_truncation): Handle truncating an IOR
23999         with a constant that is -1 in the truncated to mode.
24001 2017-07-19  Jan Hubicka  <hubicka@ucw.cz>
24003         * predict.c (propagate_unlikely_bbs_forward): Break out from ...
24004         (determine_unlikely_bbs): ... here.
24005         * predict.h (propagate_unlikely_bbs_forward): Declare.
24006         * cfgexpand.c (pass_expand::execute): Use it.
24007         * bb-reorder.c (sanitize_hot_paths): Do not consider known to be
24008         unlikely edges.
24009         (find_rarely_executed_basic_blocks_and_crossing_edges): Use
24010         propagate_unlikely_bbs_forward.
24012 2017-07-19  Jan Hubicka  <hubicka@ucw.cz>
24014         PR middle-end/81331
24015         * except.c (maybe_add_nop_after_section_switch): New function.
24016         (execute): Use it.
24018 2017-07-19  Tom de Vries  <tom@codesourcery.com>
24020         * gimple.h (gimple_phi_set_arg): Make assert more strict.
24022 2017-07-19  Tom de Vries  <tom@codesourcery.com>
24024         * gimple.h (gimple_phi_arg): Make assert more strict.
24026 2017-07-19  Steven Munroe  <munroesj@gcc.gnu.org>
24028         * config.gcc (powerpc*-*-*): Add mmintrin.h.
24029         * config/rs6000/mmintrin.h: New file.
24030         * config/rs6000/x86intrin.h [__ALTIVEC__]: Include mmintrin.h.
24032 2017-07-19  Jakub Jelinek  <jakub@redhat.com>
24034         PR tree-optimization/81346
24035         * match.pd: Optimize (X - 1U) <= INT_MAX-1U into (int) X > 0.
24037 2017-07-19  Tom de Vries  <tom@codesourcery.com>
24039         * config/nvptx/nvptx.md (VECIM): Add V2DI.
24041 2017-07-19  Tom de Vries  <tom@codesourcery.com>
24043         * config/nvptx/nvptx-modes.def: Add V2DImode.
24044         * config/nvptx/nvptx-protos.h (nvptx_data_alignment): Declare.
24045         * config/nvptx/nvptx.c (nvptx_ptx_type_from_mode): Handle V2DImode.
24046         (nvptx_output_mov_insn): Handle lack of mov.b128.
24047         (nvptx_print_operand): Handle 'H' and 'L' codes.
24048         (nvptx_vector_mode_supported): Allow V2DImode.
24049         (nvptx_preferred_simd_mode): New function.
24050         (nvptx_data_alignment): New function.
24051         (TARGET_VECTORIZE_PREFERRED_SIMD_MODE): Redefine to
24052         nvptx_preferred_simd_mode.
24053         * config/nvptx/nvptx.h (STACK_BOUNDARY, BIGGEST_ALIGNMENT): Change from
24054         64 to 128 bits.
24055         (DATA_ALIGNMENT): Define.  Set to nvptx_data_alignment.
24057 2017-07-19  Tom de Vries  <tom@codesourcery.com>
24059         * config/nvptx/nvptx-modes.def: New file.  Add V2SImode.
24060         * config/nvptx/nvptx.c (nvptx_ptx_type_from_mode): Handle V2SImode.
24061         (nvptx_vector_mode_supported): New function.  Allow V2SImode.
24062         (TARGET_VECTOR_MODE_SUPPORTED_P): Redefine to nvptx_vector_mode_supported.
24063         * config/nvptx/nvptx.md (VECIM): New mode iterator. Add V2SI.
24064         (mov<VECIM>_insn): New define_insn.
24065         (define_expand "mov<VECIM>): New define_expand.
24067 2017-07-19  Tom de Vries  <tom@codesourcery.com>
24069         * config/nvptx/nvptx.c (nvptx_print_operand): Handle v2 vector mode.
24071 2017-07-19  Jakub Jelinek  <jakub@redhat.com>
24073         PR tree-optimization/81346
24074         * fold-const.h (fold_div_compare, range_check_type): Declare.
24075         * fold-const.c (range_check_type): New function.
24076         (build_range_check): Use range_check_type.
24077         (fold_div_compare): No longer static, rewritten into
24078         a match.pd helper function.
24079         (fold_comparison): Don't call fold_div_compare here.
24080         * match.pd (X / C1 op C2): New optimization using fold_div_compare
24081         as helper function.
24083 2017-07-19  Nathan Sidwell  <nathan@acm.org>
24085         * tree.h (TYPE_MINVAL, TYPE_MAXVAL): Rename to ...
24086         (TYPE_MIN_VALUE_RAW, TYPE_MAX_VALUE_RAW): ... these.
24087         * tree.c (find_decls_types_r, verify_type): Use
24088         TYPE_{MIN,MAX}_VALUE_RAW.
24089         * lto-streamer-out.c (DFS::DFS_write_tree_body): Likewise.
24090         (hash_tree): Likewise.
24091         * tree-streamer-in.c (lto_input_ts_type_non_common_tree_pointers):
24092         Likewise.
24093         * tree-streamer-out.c (write_ts_type_non_common_tree_pointers):
24094         Likewise.
24096 2017-07-18  Tom de Vries  <tom@codesourcery.com>
24098         PR middle-end/81464
24099         * omp-expand.c (expand_omp_for_static_chunk): Handle
24100         equal-argument loop exit phi.
24102 2017-07-18  Uros Bizjak  <ubizjak@gmail.com>
24104         PR target/81471
24105         * config/i386/i386.md (rorx_immediate_operand): New mode attribute.
24106         (*bmi2_rorx<mode>3_1): Use rorx_immediate_operand as
24107         operand 2 predicate.
24108         (*bmi2_rorxsi3_1_zext): Use const_0_to_31_operand as
24109         operand 2 predicate.
24110         (ror,rol -> rorx splitters): Use const_int_operand as
24111         operand 2 predicate.
24113 2017-06-18  Richard Biener  <rguenther@suse.de>
24115         PR tree-optimization/81410
24116         * tree-vect-stmts.c (vectorizable_load): Properly adjust for
24117         the gap in the ! slp_perm SLP case after each group.
24119 2017-07-18  Jan Hubicka  <hubicka@ucw.cz>
24121         PR middle-end/81463
24122         * cfgloopmanip.c (scale_loop_profile): Watch out for zero frequency
24123         again.
24125 2017-07-18  Jan Hubicka  <hubicka@ucw.cz>
24127         PR middle-end/81462
24128         * predict.c (set_even_probabilities): Cleanup; do not affect
24129         probabilities that are already known.
24130         (combine_predictions_for_bb): Call even when count is set.
24132 2017-07-18  Nathan Sidwell  <nathan@acm.org>
24134         * tree-parloops.c (try_transform_to_exit_first_loop_alt): Use
24135         TYPE_MAX_VALUE.
24137 2017-07-18  Bin Cheng  <bin.cheng@arm.com>
24139         PR target/81408
24140         * tree-ssa-loop-niter.c (number_of_iterations_exit): Dump missed
24141         optimization for loop niter analysis.
24143 2017-07-18  Georg-Johann Lay  <avr@gjlay.de>
24145         PR target/81473
24146         * config/avr/avr.c (avr_optimize_casesi): Don't use
24147         INT8_MIN, INT8_MAX, UINT8_MAX, INT16_MIN, INT16_MAX, UINT16_MAX.
24149 2017-07-18  Robin Dapp  <rdapp@linux.vnet.ibm.com>
24151         * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Remove
24152         body_cost_vec from _vect_peel_extended_info.
24153         (vect_peeling_hash_get_lowest_cost): Do not set body_cost_vec.
24154         (vect_peeling_hash_choose_best_peeling): Remove body_cost_vec and
24155         npeel.
24157 2017-07-18  Bin Cheng  <bin.cheng@arm.com>
24159         * config/arm/arm.c (emit_unlikely_jump): Remove unused var.
24161 2017-07-18  Richard Biener  <rguenther@suse.de>
24163         PR tree-optimization/80620
24164         PR tree-optimization/81403
24165         * tree-ssa-pre.c (phi_translate_1): Clear range and points-to
24166         info when re-using a VN table entry.
24168 2017-07-18  Richard Biener  <rguenther@suse.de>
24170         PR tree-optimization/81418
24171         * tree-vect-loop.c (vectorizable_reduction): Properly compute
24172         vectype_in.  Verify that with lane-reducing reduction operations
24173         we have a single def-use cycle.
24175 2017-07-17  Carl Love  <cel@us.ibm.com>
24177         Revert commit r249424   2017-06-20  Carl Love  <cel@us.ibm.com>
24179         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
24180         ALTIVEC_BUILTIN_VMULESW, ALTIVEC_BUILTIN_VMULEUW,
24181         ALTIVEC_BUILTIN_VMULOSW, ALTIVEC_BUILTIN_VMULOUW entries.
24182         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin,
24183         builtin_function_type): Add ALTIVEC_BUILTIN_* case statements.
24184         * config/rs6000/altivec.md (MVULEUW, VMULESW, VMULOUW,
24185         VMULOSW): New enum "unspec" values.
24186         (vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
24187         vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si,
24188         altivec_vmuleuw, altivec_vmulesw, altivec_vmulouw,
24189         altivec_vmulosw): New patterns.
24190         * config/rs6000/rs6000-builtin.def (VMLEUW, VMULESW, VMULOUW,
24191         VMULOSW): Add definitions.
24193 2017-07-17  Uros Bizjak  <ubizjak@gmail.com>
24195         * config/alpha/alpha.c: Include predict.h.
24197 2017-07-17  Yury Gribov  <tetra2005@gmail.com>
24199         * tree-vrp.c (compare_assert_loc): Fix comparison function
24200         to return predictable results.
24202 2017-07-17  Claudiu Zissulescu  <claziss@synopsys.com>
24204         * config/arc/arc.md (adddi3): Remove support for mexpand-adddi
24205         option.
24206         (subdi3): Likewise.
24207         * config/arc/arc.opt (mexpand-adddi): Deprecate it.
24208         * doc/invoke.texi (mexpand-adddi): Update text.
24210 2017-07-17  Claudiu Zissulescu  <claziss@synopsys.com>
24212         * config/arc/arc.md (clzsi2): Expand to an arc_clzsi2 instruction
24213         that also clobbers the CC register. The old expand code is moved
24214         to ...
24215         (*arc_clzsi2): ... here.
24216         (ctzsi2): Expand to an arc_ctzsi2 instruction that also clobbers
24217         the CC register. The old expand code is moved to ...
24218         (arc_ctzsi2): ... here.
24220 2017-07-17  Claudiu Zissulescu  <claziss@synopsys.com>
24222         * config/arc/arc.opt (mindexed-loads): Use initial value
24223         TARGET_INDEXED_LOADS_DEFAULT.
24224         (mauto-modify-reg): Use initial value
24225         TARGET_AUTO_MODIFY_REG_DEFAULT.
24226         * config/arc/elf.h (TARGET_INDEXED_LOADS_DEFAULT): Define.
24227         (TARGET_AUTO_MODIFY_REG_DEFAULT): Likewise.
24228         * config/arc/linux.h (TARGET_INDEXED_LOADS_DEFAULT): Define.
24229         (TARGET_AUTO_MODIFY_REG_DEFAULT): Likewise.
24231 2017-07-17  Martin Liska  <mliska@suse.cz>
24233         PR sanitizer/81302
24234         * opts.c (finish_options): Do not allow -fgnu-tm
24235         w/ -fsanitize={kernel-,}address.  Say sorry.
24237 2017-07-17  Bin Cheng  <bin.cheng@arm.com>
24239         PR target/81369
24240         * tree-loop-distribution.c (classify_partition): Only assert on
24241         numer of iterations.
24242         (merge_dep_scc_partitions): Delete prameter.  Update function call.
24243         (distribute_loop): Remove code handling loop with unknown niters.
24244         (pass_loop_distribution::execute): Skip loop with unknown niters.
24246 2017-07-17  Bin Cheng  <bin.cheng@arm.com>
24248         PR target/81369
24249         * tree-loop-distribution.c (merge_dep_scc_partitions): Sink call to
24250         function sort_partitions_by_post_order.
24252 2017-07-17  Bin Cheng  <bin.cheng@arm.com>
24254         PR tree-optimization/81374
24255         * tree-loop-distribution.c (pass_loop_distribution::execute): Record
24256         the max index of basic blocks, rather than number of basic blocks.
24258 2017-07-17  Claudiu Zissulescu  <claziss@synopsys.com>
24260         * config/arc/arc-protos.h (arc_legitimate_pc_offset_p): Remove
24261         proto.
24262         (arc_legitimate_pic_operand_p): Likewise.
24263         * config/arc/arc.c (arc_legitimate_pic_operand_p): Remove
24264         function.
24265         (arc_needs_pcl_p): Likewise.
24266         (arc_legitimate_pc_offset_p): Likewise.
24267         (arc_legitimate_pic_addr_p): Remove LABEL_REF case, as this
24268         function is also used in constrains.md.
24269         (arc_legitimate_constant_p): Use arc_legitimate_pic_addr_p to
24270         validate pic constants. Handle CONST_INT, CONST_DOUBLE, MINUS and
24271         PLUS.  Only return true/false in known cases, otherwise assert.
24272         (arc_legitimate_address_p): Remove arc_legitimate_pic_addr_p as it
24273         is already called in arc_legitimate_constant_p.
24274         * config/arc/arc.h (CONSTANT_ADDRESS_P): Consider also LABEL for
24275         pic addresses.
24276         (LEGITIMATE_PIC_OPERAND_P): Use
24277         arc_raw_symbolic_reference_mentioned_p function.
24278         * config/arc/constraints.md (Cpc): Use arc_legitimate_pic_addr_p
24279         function.
24280         (Cal): Likewise.
24281         (C32): Likewise.
24283 2017-07-17  Claudiu Zissulescu  <claziss@synopsys.com>
24284         Andrew Burgess  <andrew.burgess@embecosm.com>
24286         * config/arc/arc-protos.h (arc_compute_function_type): Change prototype.
24287         (arc_return_address_register): New function.
24288         * config/arc/arc.c (arc_handle_fndecl_attribute): New function.
24289         (arc_handle_fndecl_attribute): Add naked attribute.
24290         (TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS): Define.
24291         (TARGET_WARN_FUNC_RETURN): Likewise.
24292         (arc_allocate_stack_slots_for_args): New function.
24293         (arc_warn_func_return): Likewise.
24294         (machine_function): Change type fn_type.
24295         (arc_compute_function_type): Consider new naked function type,
24296         change function return type.
24297         (arc_must_save_register): Adapt to handle new
24298         arc_compute_function_type's return type.
24299         (arc_expand_prologue): Likewise.
24300         (arc_expand_epilogue): Likewise.
24301         (arc_return_address_regs): Delete.
24302         (arc_return_address_register): New function.
24303         (arc_epilogue_uses): Use above function.
24304         * config/arc/arc.h (arc_return_address_regs): Delete prototype.
24305         (arc_function_type): Change encoding, add naked type.
24306         (ARC_INTERRUPT_P): Change to handle the new encoding.
24307         (ARC_FAST_INTERRUPT_P): Likewise.
24308         (ARC_NORMAL_P): Define.
24309         (ARC_NAKED_P): Likewise.
24310         (arc_compute_function_type): Delete prototype.
24311         * config/arc/arc.md (in_ret_delay_slot): Use
24312         arc_return_address_register function.
24313         (simple_return): Likewise.
24314         (p_return_i): Likewise.
24316 2017-07-17  Jakub Jelinek  <jakub@redhat.com>
24318         PR tree-optimization/81428
24319         * match.pd (X / X -> one): Don't optimize _Fract divisions, as 1
24320         can't be built for those types.
24322 2017-07-17  Georg-Johann Lay  <avr@gjlay.de>
24324         Remove stuff dead since r239246.
24326         * config/avr/avr-arch.h (avr_inform_devices): Remove dead proto.
24327         * config/avr/avr-devices.c (mcu_name, comparator, avr_mcus_str)
24328         (avr_inform_devices): Remove dead stuff.
24330 2017-07-17  Tamar Christina  <tamar.christina@arm.com>
24332         * config/arm/arm_neon.h: Fix softp typo.
24334 2017-07-17  Jakub Jelinek  <jakub@redhat.com>
24336         PR tree-optimization/81365
24337         * tree-ssa-phiprop.c (propagate_with_phi): When considering hoisting
24338         aggregate moves onto bb predecessor edges, make sure there are no
24339         loads that could alias the lhs in between the start of bb and the
24340         loads from *phi.
24342 2017-07-17  Georg-Johann Lay  <avr@gjlay.de>
24344         PR 80929
24345         * config/avr/avr.c (avr_mul_highpart_cost): New static function.
24346         (avr_rtx_costs_1) [TRUNCATE]: Use it to compute mul_highpart cost.
24347         [LSHIFTRT, outer_code = TRUNCATE]: Same.
24349 2017-07-17  Jakub Jelinek  <jakub@redhat.com>
24351         PR tree-optimization/81396
24352         * tree-ssa-math-opts.c (struct symbolic_number): Add n_ops field.
24353         (init_symbolic_number): Initialize it to 1.
24354         (perform_symbolic_merge): Add n_ops from both operands into the new
24355         n_ops.
24356         (find_bswap_or_nop): Don't consider n->n == cmpnop computations
24357         without base_addr as useless if they need more than one operation.
24358         (bswap_replace): Handle !bswap case for NULL base_addr.
24360 2017-07-17  Tom de Vries  <tom@codesourcery.com>
24362         PR target/81069
24363         * config/nvptx/nvptx.c (nvptx_single): Insert diverging branch as late
24364         as possible.
24366 2017-07-17  Sebastian Huber  <sebastian.huber@embedded-brains.de>
24368         * config/sparc/rtemself.h (TARGET_OS_CPP_BUILTINS): Add
24369         conditional builtin define __FIX_LEON3FT_B2BST.
24371 2017-07-17  Daniel Cederman  <cederman@gaisler.com>
24373         * config/sparc/t-rtems: Add mfix-gr712rc multilibs. Replace
24374         MULTILIB_EXCEPTIONS with MULTILIB_REQUIRED. Match -mfix-gr712rc
24375         with -mfix-ut700.
24377 2017-07-16  Eric Botcazou  <ebotcazou@adacore.com>
24379         PR rtl-optimization/81424
24380         * optabs.c (prepare_cmp_insn): Use copy_to_reg instead of force_reg
24381         to remove potential trapping from operands if -fnon-call-exceptions.
24383 2017-07-16  Jan Hubicka  <hubicka@ucw.cz>
24385         * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Use
24386         profile_proability for scalling.
24387         * scale_profile_for_vect_loop.c (scale_profile_for_vect_loop): Likewise.
24389 2017-07-16  Jan Hubicka  <hubicka@ucw.cz>
24391         * cgraph.c (cgraph_edge::redirect_call_stmt_to_caller): Cleanup.
24393 2017-07-16  Jan Hubicka  <hubicka@ucw.cz>
24395         * cfgloopmanip.c (scale_loop_profile): Avoid use of REG_BR_PROB_BASE
24396         fixpoint arithmetics.
24398 2017-07-16  Jan Hubicka  <hubicka@ucw.cz>
24400         * tree-ssa-loop-unswitch.c (hoist_guard): Avoid use of REG_BR_PROB_BASE
24401         fixpoint arithmetics.
24403 2017-07-16  Jan Hubicka  <hubicka@ucw.cz>
24405         * asan.c (create_cond_insert_point): Avoid use of REG_BR_PROB_BASE
24406         fixpoint arithmetics.
24408 2017-07-16  Jan Hubicka  <hubicka@ucw.cz>
24410         * profile-count.h (profile_probability::from_reg_br_prob_note,
24411         profile_probability::to_reg_br_prob_note): New functions.
24412         * doc/rtl.texi (REG_BR_PROB_NOTE): Update documentation.
24413         * reg-notes.h (REG_BR_PROB, REG_BR_PRED): Update docs.
24414         * predict.c (probability_reliable_p): Update.
24415         (edge_probability_reliable_p): Update.
24416         (br_prob_note_reliable_p): Update.
24417         (invert_br_probabilities): Update.
24418         (add_reg_br_prob_note): New function.
24419         (combine_predictions_for_insn): Update.
24420         * asan.c (asan_clear_shadow): Update.
24421         * cfgbuild.c (compute_outgoing_frequencies): Update.
24422         * cfgrtl.c (force_nonfallthru_and_redirect): Update.
24423         (update_br_prob_note): Update.
24424         (rtl_verify_edges): Update.
24425         (purge_dead_edges): Update.
24426         (fixup_reorder_chain): Update.
24427         * emit-rtl.c (try_split): Update.
24428         * ifcvt.c (cond_exec_process_insns): Update.
24429         (cond_exec_process_if_block): Update.
24430         (dead_or_predicable): Update.
24431         * internal-fn.c (expand_addsub_overflow): Update.
24432         (expand_neg_overflow): Update.
24433         (expand_mul_overflow): Update.
24434         * loop-doloop.c (doloop_modify): Update.
24435         * loop-unroll.c (compare_and_jump_seq): Update.
24436         * optabs.c (emit_cmp_and_jump_insn_1): Update.
24437         * predict.h: Update.
24438         * reorg.c (mostly_true_jump): Update.
24439         * rtl.h: Update.
24440         * config/aarch64/aarch64.c (aarch64_emit_unlikely_jump): Update.
24441         * config/alpha/alpha.c (emit_unlikely_jump): Update.
24442         * config/arc/arc.c: (emit_unlikely_jump): Update.
24443         * config/arm/arm.c: (emit_unlikely_jump): Update.
24444         * config/bfin/bfin.c (cbranch_predicted_taken_p): Update.
24445         * config/frv/frv.c (frv_print_operand_jump_hint): Update.
24446         * config/i386/i386.c (ix86_expand_split_stack_prologue): Update.
24447         (ix86_print_operand): Update.
24448         (ix86_split_fp_branch): Update.
24449         (predict_jump): Update.
24450         * config/ia64/ia64.c (ia64_print_operand): Update.
24451         * config/mmix/mmix.c (mmix_print_operand): Update.
24452         * config/powerpcspe/powerpcspe.c (output_cbranch): Update.
24453         (rs6000_expand_split_stack_prologue): Update.
24454         * config/rs6000/rs6000.c: Update.
24455         * config/s390/s390.c (s390_expand_vec_strlen): Update.
24456         (s390_expand_vec_movstr): Update.
24457         (s390_expand_cs_tdsi): Update.
24458         (s390_expand_split_stack_prologue): Update.
24459         * config/sh/sh.c (sh_print_operand): Update.
24460         (expand_cbranchsi4): Update.
24461         (expand_cbranchdi4): Update.
24462         * config/sparc/sparc.c (output_v9branch): Update.
24463         * config/spu/spu.c (get_branch_target): Update.
24464         (ea_load_store_inline): Update.
24465         * config/tilegx/tilegx.c (cbranch_predicted_p): Update.
24466         * config/tilepro/tilepro.c: Update.
24468 2017-07-16  Eric Botcazou  <ebotcazou@adacore.com>
24470         * gimplify.c (mostly_copy_tree_r): Revert latest change.
24471         (gimplify_save_expr): Likewise.
24473 2017-07-07  Jan Hubicka  <hubicka@ucw.cz>
24475         * ipa-visibility.c (function_and_variable_visibility): Fix pasto.
24477 2017-07-07  Jan Hubicka  <hubicka@ucw.cz>
24479         * ipa-fnsummary.c (pass_data_ipa_fn_summary): Use
24480         TV_IPA_FNSUMMARY.
24481         * timevar.def (TV_IPA_FNSUMMARY): Define.
24483 2017-07-16  Daniel Cederman  <cederman@gaisler.com>
24485         * config/sparc/sparc.md (divdf3_fix): Add NOP to prevent back
24486         to back store errata sensitive sequence from being generated.
24487         (sqrtdf2_fix): Likewise.
24489 2017-07-07  Jan Hubicka  <hubicka@ucw.cz>
24491         * tree-ssa-threadupdate.c (compute_path_counts,
24492         update_joiner_offpath_counts): Use profile_probability.
24494 2017-07-15  Thomas Preud'homme  <thomas.preudhomme@arm.com>
24496         Revert:
24497         2017-07-14  Thomas Preud'homme  <thomas.preudhomme@arm.com>
24499         * config/arm/arm-c.c (arm_cpu_builtins): Define
24500         __ARM_FEATURE_NUMERIC_MAXMIN solely based on TARGET_VFP5.
24502 2017-07-14  Kelvin Nilsen  <kelvin@gcc.gnu.org>
24504         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
24505         array entries to represent __ieee128 versions of the
24506         scalar_test_data_class, scalar_test_neg, scalar_extract_exp,
24507         scalar_extract_sig, and scalar_insert_exp built-in functions.
24508         (altivec_resolve_overloaded_builtin): Add special case handling
24509         for the __builtin_scalar_insert_exp function, as represented by
24510         the P9V_BUILTIN_VEC_VSIEDP constant.
24511         * config/rs6000/rs6000-builtin.def (VSEEQP): Add scalar extract
24512         exponent support for __ieee128 argument.
24513         (VSESQP): Add scalar extract signature support for __ieee128
24514         argument.
24515         (VSTDCNQP): Add scalar test negative support for __ieee128
24516         argument.
24517         (VSIEQP): Add scalar insert exponent support for __int128 argument
24518         with __ieee128 result.
24519         (VSIEQPF): Add scalar insert exponent support for __ieee128
24520         argument with __ieee128 result.
24521         (VSTDCQP): Add scalar test data class support for __ieee128
24522         argument.
24523         (VSTDCNQP): Add overload support for scalar test negative with
24524         __ieee128 argument.
24525         (VSTDCQP): Add overload support for scalar test data class
24526         __ieee128 argument.
24527         * config/rs6000/vsx.md (UNSPEC_VSX_SXSIG) Replace
24528         UNSPEC_VSX_SXSIGDP.
24529         (UNSPEC_VSX_SIEXPQP): New constant.
24530         (xsxexpqp): New insn for VSX scalar extract exponent quad
24531         precision.
24532         (xsxsigqp): New insn for VSX scalar extract significand quad
24533         precision.
24534         (xsiexpqpf): New insn for VSX scalar insert exponent quad
24535         precision with floating point argument.
24536         (xststdcqp): New expand for VSX scalar test data class quad
24537         precision.
24538         (xststdcnegqp): New expand for VSX scalar test negative quad
24539         precision.
24540         (xststdcqp): New insn to match expansions for VSX scalar test data
24541         class quad precision and VSX scalar test negative quad precision.
24542         * config/rs6000/rs6000.c (rs6000_expand_binop_builtin): Add
24543         special case operand checking to enforce that second operand of
24544         VSX scalar test data class with quad precision argument is a 7-bit
24545         unsigned literal.
24546         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Add
24547         prototypes and descriptions of __ieee128 versions of
24548         scalar_extract_exp, scalar_extract_sig, scalar_insert_exp,
24549         scalar_test_data_class, and scalar_test_neg built-in functions.
24551 2016-07-14  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
24553         PR tree-optimization/81162
24554         * gimple-ssa-strength-reduction.c (replace_mult_candidate): Don't
24555         replace a negate with an add.
24557 2017-07-14  James Greenhalgh  <james.greenhalgh@arm.com>
24559         * doc/invoke.texi (arm/-mcpu): Document +crypto.
24561 2017-07-14  Thomas Preud'homme  <thomas.preudhomme@arm.com>
24563         * config/arm/arm-c.c (arm_cpu_builtins): Define
24564         __ARM_FEATURE_NUMERIC_MAXMIN solely based on TARGET_VFP5.
24566 2017-07-14  Thomas Preud'homme  <thomas.preudhomme@arm.com>
24568         * config/arm/arm-cpus.in (cortex-r52): Add new entry.
24569         (armv8-r): Set ARM Cortex-R52 as default CPU.
24570         * config/arm/arm-tables.opt: Regenerate.
24571         * config/arm/arm-tune.md: Regenerate.
24572         * config/arm/driver-arm.c (arm_cpu_table): Add entry for ARM
24573         Cortex-R52.
24574         * doc/invoke.texi: Mention -mtune=cortex-r52 and availability of fp.dp
24575         extension for -mcpu=cortex-r52.
24577 2017-07-14  Thomas Preud'homme  <thomas.preudhomme@arm.com>
24579         * config/arm/arm-isa.h (isa_bit_FP_ARMv8): Delete enumerator.
24580         (ISA_FP_ARMv8): Define as ISA_FPv5 and ISA_FP_D32.
24581         * config/arm/arm-cpus.in (armv8-r): Define fp.sp as enabling FPv5.
24582         (fp-armv8): Define it as FP_ARMv8 only.
24583         config/arm/arm.h (TARGET_FPU_ARMV8): Delete.
24584         (TARGET_VFP_FP16INST): Define using TARGET_VFP5 rather than
24585         TARGET_FPU_ARMV8.
24586         config/arm/arm.c (arm_rtx_costs_internal): Replace checks against
24587         TARGET_FPU_ARMV8 by checks against TARGET_VFP5.
24588         * config/arm/arm-builtins.c (arm_builtin_vectorized_function): Define
24589         first ARM_CHECK_BUILTIN_MODE definition using TARGET_VFP5 rather
24590         than TARGET_FPU_ARMV8.
24591         * config/arm/arm-c.c (arm_cpu_builtins): Likewise for
24592         __ARM_FEATURE_NUMERIC_MAXMIN macro definition.
24593         * config/arm/arm.md (cmov<mode>): Condition on TARGET_VFP5 rather than
24594         TARGET_FPU_ARMV8.
24595         * config/arm/neon.md (neon_vrint): Likewise.
24596         (neon_vcvt): Likewise.
24597         (neon_<fmaxmin_op><mode>): Likewise.
24598         (<fmaxmin><mode>3): Likewise.
24599         * config/arm/vfp.md (l<vrint_pattern><su_optab><mode>si2): Likewise.
24600         * config/arm/predicates.md (arm_cond_move_operator): Check against
24601         TARGET_VFP5 rather than TARGET_FPU_ARMV8 and fix spacing.
24603 2017-07-14  Jackson Woodruff  <jackson.woodruff@arm.com>
24605         * config/aarch64/aarch64.c (aarch64_print_operand): Move comments
24606         to top of function.
24608 2017-07-14  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
24610         * gimple-ssa-store-merging.c (clear_bit_region): Replace reference to
24611         loop in comment with memset.
24613 2017-07-14  Martin Liska  <mliska@suse.cz>
24615         * cfgexpand.c (expand_gimple_basic_block): Remove dead comment.
24616         * dwarf2out.c (is_java): Remove the function.
24617         (output_pubname): Remove usage of the function.
24618         (lower_bound_default): Remove usage of DW_LANG_Java.
24619         (gen_compile_unit_die): Likewise.
24620         * gcc.c: Remove compiler defaults for .java and .zip files.
24621         * gimple-expr.c (remove_suffix): Change as there's no longer
24622         extension than 4-letter one.
24623         * gimplify.c (mostly_copy_tree_r): Remove Java-special part.
24624         (gimplify_save_expr): Likewise.
24625         * ipa-utils.h (polymorphic_type_binfo_p): Remove the comment
24626         as it's possible even for other languages than Java.
24627         * langhooks.h (struct lang_hooks): Remove Java from a comment.
24628         * lto-opts.c (lto_write_options): Remove reference to Java.
24629         * opts.c (strip_off_ending): Update file extension handling.
24630         * tree-cfg.c (verify_gimple_call): Remove comment with Java.
24631         * tree-eh.c (lower_resx): Likewise.
24632         * tree.c (free_lang_data_in_type): Remove dead code.
24633         (find_decls_types_r): Likewise.
24634         (build_common_builtin_nodes): Remove Java from a comment.
24635         (verify_type): Remove dead code.
24636         * varasm.c (assemble_external): Remove Java from a comment.
24638 2017-07-14  Martin Liska  <mliska@suse.cz>
24640         * opts.c (finish_options): Add quotes.
24641         (common_handle_option): Likewise.
24643 2017-07-14  Martin Liska  <mliska@suse.cz>
24645         * dbxout.c (get_lang_number): Do not handle GNU Pascal.
24646         * dbxout.h (extern void dbxout_stab_value_internal_label_diff):
24647         Remove N_SO_PASCAL.
24648         * dwarf2out.c (lower_bound_default): Do not handle
24649         DW_LANG_Pascal83.
24650         (gen_compile_unit_die): Likewise.
24651         * gcc.c: Remove default extension binding for GNU Pascal.
24652         * stmt.c: Remove Pascal language from a comment.
24653         * xcoffout.c: Likewise.
24655 2017-07-13  David Malcolm  <dmalcolm@redhat.com>
24657         PR c/81405
24658         * diagnostic-show-locus.c (fixit_cmp): New function.
24659         (layout::layout): Sort m_fixit_hints.
24660         (column_range::column_range): Assert that the values are valid.
24661         (struct char_span): New struct.
24662         (correction::overwrite): New method.
24663         (struct source_line): New struct.
24664         (line_corrections::add_hint): Add assertions.  Reimplement memcpy
24665         calls in terms of classes source_line and char_span, and
24666         correction::overwrite.
24667         (selftest::test_overlapped_fixit_printing_2): New function.
24668         (selftest::diagnostic_show_locus_c_tests): Call it.
24670 2017-07-13  Will Schmidt  <will_schmidt@vnet.ibm.com>
24672         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Return
24673         early if there is no lhs.
24675 2017-07-13  Martin Liska  <mliska@suse.cz>
24677         * dwarf2out.c (gen_pointer_type_die): Remove dead code.
24678         (gen_reference_type_die): Likewise.
24679         * stor-layout.c: Remove Pascal-related comment.
24681 2017-07-13  Martin Liska  <mliska@suse.cz>
24683         * opts.c (finish_options): Add quotes to error messages.
24684         (parse_sanitizer_options): Likewise.
24686 2017-07-13  Thomas Preud'homme  <thomas.preudhomme@arm.com>
24688         * doc/invoke.texi (armv8-r): Document +fp.sp ARMv8-R extension.
24690 2017-07-13  Richard Earnshaw  <rearnsha@arm.com>
24692         * config/arm/vxworks.h (TARGET_ENDIAN_DEFAULT): Define.
24694 2017-07-13  Maxim Ostapenko  <m.ostapenko@samsung.com>
24696         * asan.c (asan_emit_allocas_unpoison): Use ptr_mode for arguments
24697         during expansion.
24698         * builtins.c (expand_asan_emit_allocas_unpoison): Likewise.
24700 2017-07-12  Michael Meissner  <meissner@linux.vnet.ibm.com>
24702         PR target/81193
24703         * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If GLIBC
24704         provides the hardware capability bits, define the macro
24705         __BUILTIN_CPU_SUPPORTS__.
24706         * config/rs6000/rs6000.c (cpu_expand_builtin): Generate a warning
24707         if GLIBC does not provide the hardware capability bits.  Add a
24708         gcc_unreachable call if the built-in cpu function is neither
24709         __builtin_cpu_is nor __builtin_cpu_supports.
24710         (rs6000_get_function_versions_dispatcher): Change the warning
24711         that an old GLIBC is used which does not export the capability
24712         bits to be an error.
24713         * doc/extend.texi (target_clones attribute): Document the
24714         restriction that GLIBC 2.23 or newer is needed on the PowerPC.
24715         (PowerPC built-in functions): Document that GLIBC 2.23 or newer is
24716         needed by __builtin_cpu_is and __builtin_cpu_supports.  Document
24717         the macros defined by GCC if the newer GLIBC is available.
24719 2017-07-12  Jeff Law  <law@redhat.com>
24721         * config/riscv/riscv.c: Remove unnecessary includes.  Reorder
24722         remaining includes slightly.
24723         * config/riscv/riscv-builtins.c: Include profile-count.h.
24725 2017-07-12  Georg-Johann Lay  <avr@gjlay.de>
24727         PR target/79883
24728         * config/avr/avr.c (avr_set_current_function): In diagnostic
24729         messages: Quote keywords and (parts of) identifiers.
24730         [WITH_AVRLIBC]: Warn for functions named "ISR", "SIGNAL" or
24731         "INTERUPT".
24733 2017-07-12  Carl Love  <cel@us.ibm.com>
24735         * config/rs6000/rs6000-c.c: Add support for built-in functions
24736         vector bool char vec_revb (vector bool char);
24737         vector bool short vec_revb (vector short char);
24738         vector bool int vec_revb (vector bool int);
24739         vector bool long long vec_revb (vector bool long long);
24740         * doc/extend.texi: Update the built-in documentation file for the
24741         new built-in functions.
24743 2017-07-12  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
24745         * config/s390/s390.md: Remove movcc splitter.
24747 2017-07-12  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
24749         * config/s390/s390.c (s390_rtx_costs): Return proper costs for
24750         load/store on condition.
24752 2017-07-12  Georg-Johann Lay  <avr@gjlay.de>
24754         PR target/81407
24755         * config/avr/avr.c (avr_encode_section_info)
24756         [progmem && !TREE_READONLY]: Error if progmem object needs
24757         constructing.
24759 2017-07-11  Michael Collison  <michael.collison@arm.com>
24761         * config/aarch64/aarch64-simd.md (aarch64_sub<mode>_compare0):
24762         New pattern.
24764 2017-07-11  Carl Love  <cel@us.ibm.com>
24766         * config/rs6000/rs6000-c.c: Add support for builtins
24767         vector unsigned int vec_parity_lsbb (vector signed int);
24768         vector unsigned int vec_parity_lsbb (vector unsigned int);
24769         vector unsigned __int128 vec_parity_lsbb (vector signed __int128);
24770         vector unsigned __int128 vec_parity_lsbb (vector unsigned __int128);
24771         vector unsigned long long vec_parity_lsbb (vector signed long long);
24772         vector unsigned long long vec_parity_lsbb (vector unsigned long long);
24773         * config/rs6000/rs6000-builtin.def (VPARITY_LSBB): Add BU_P9V_OVERLOAD1.
24774         * config/rs6000/altivec.h (vec_parity_lsbb): Add define.
24775         * doc/extend.texi: Update the built-in documentation file for the
24776         new built-in functions.
24778 2017-07-11  David Malcolm  <dmalcolm@redhat.com>
24780         * diagnostic-show-locus.c: Include "gcc-rich-location.h".
24781         (layout::m_primary_loc): New field.
24782         (layout::layout): Initialize new field.  Move location filtering
24783         logic from here to...
24784         (layout::maybe_add_location_range): ...this new method.  Add
24785         support for filtering to just the lines already specified by other
24786         locations.
24787         (layout::will_show_line_p): New method.
24788         (gcc_rich_location::add_location_if_nearby): New method.
24789         (selftest::test_add_location_if_nearby): New test function.
24790         (selftest::diagnostic_show_locus_c_tests): Call it.
24791         * gcc-rich-location.h (gcc_rich_location::add_location_if_nearby):
24792         New method.
24794 2017-07-11  Tom de Vries  <tom@codesourcery.com>
24796         * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG): New macro.
24797         (bb_first_real_insn): New function.
24798         (nvptx_single): Add extra initialization of broadcasted condition
24799         variables.
24801 2017-07-11  Nathan Sidwell  <nathan@acm.org>
24803         * dwarf2out.c (gen_member_die): Remove useless check for anon ctors.
24805 2017-07-11  Georg-Johann Lay  <avr@gjlay.de>
24807         * doc/extend.texi (AVR Function Attributes): Remove weblink to
24808         Binutils doc as TEXI will mess them up.
24809         * doc/invoke.texi (AVR Options): Same here.
24811 2017-07-11  Daniel Cederman  <cederman@gaisler.com>
24813         * config/sparc/sparc.opt (mfix-ut700): New option.
24814         (mfix-gr712rc): Likewise.
24815         (sparc_fix_b2bst): New variable.
24816         * doc/invoke.texi (SPARC options): Document them.
24817         (ARM options): Fix warnings.
24818         * config/sparc/sparc.c (sparc_do_work_around_errata): Insert NOP
24819         instructions to prevent sequences that can trigger the store-store
24820         errata for certain LEON3FT processors.
24821         (pass_work_around_errata::gate): Also test sparc_fix_b2bst.
24822         (sparc_option_override): Set sparc_fix_b2bst appropriately.
24823         * config/sparc/sparc.md (fix_b2bst): New attribute.
24824         (in_branch_delay): Prevent stores in delay slot if fix_b2bst.
24826 2017-07-10  Uros Bizjak  <ubizjak@gmail.com>
24828         PR target/81375
24829         * config/i386/i386.md (divsf3): Add TARGET_SSE to TARGET_SSE_MATH.
24830         (rcpps): Ditto.
24831         (*rsqrtsf2_sse): Ditto.
24832         (rsqrtsf2): Ditto.
24833         (div<mode>3): Macroize insn from divdf3 and divsf3
24834         using MODEF mode iterator.
24836 2017-07-10  Martin Sebor  <msebor@redhat.com>
24838         PR tree-optimization/80397
24839         * gimple-ssa-sprintf.c (format_integer): Use INTEGRAL_TYPE_P()
24840         instead of testing for equality to INTEGER_TYPE.
24842 2017-07-10  Vineet Gupta <vgupta@synopsys.com>
24844         * config.gcc: Remove uclibc from arc target spec.
24846 2017-07-10  Claudiu Zissulescu  <claziss@synopsys.com>
24848         * config/arc/arc.h (ADDITIONAL_REGISTER_NAMES): Define.
24850 2017-07-07  Jan Hubicka  <hubicka@ucw.cz>
24852         PR lto/80838
24853         * lto-wrapper.c (remove_option): New function.
24854         (merge_and_complain): Merge PIC/PIE options more realistically.
24856 2017-07-10  Georg-Johann Lay  <avr@gjlay.de>
24858         Better ISR prologues by supporting GASes __gcc_isr pseudo insn.
24860         PR target/20296
24861         PR target/81268
24862         * configure.ac [target=avr]: Add GAS check for -mgcc-isr.
24863         (HAVE_AS_AVR_MGCCISR_OPTION):  If so, AC_DEFINE it.
24864         * config.in: Regenerate.
24865         * configure: Regenerate.
24866         * doc/extend.texi (AVR Function Attributes) <no_gccisr>: Document it.
24867         * doc/invoke.texi (AVR Options) <-mgas-isr-prologues>: Document it.
24868         * config/avr/avr.opt (-mgas-isr-prologues): New option and...
24869         (TARGET_GASISR_PROLOGUES): ...target mask.
24870         * common/config/avr/avr-common.c
24871         (avr_option_optimization_table) [OPT_LEVELS_1_PLUS_NOT_DEBUG]:
24872         Set -mgas-isr-prologues.
24873         * config/avr/avr-passes.def (avr_pass_pre_proep): Add
24874         INSERT_PASS_BEFORE for it.
24875         * config/avr/avr-protos.h (make_avr_pass_pre_proep): New proto.
24876         * config/avr/avr.c (avr_option_override)
24877         [!HAVE_AS_AVR_MGCCISR_OPTION]: Unset TARGET_GASISR_PROLOGUES.
24878         (avr_no_gccisr_function_p, avr_hregs_split_reg): New static functions.
24879         (avr_attribute_table) <no_gccisr>: Add new function attribute.
24880         (avr_set_current_function) <is_no_gccisr>: Init machine field.
24881         (avr_pass_data_pre_proep, avr_pass_pre_proep): New pass data
24882         and rtl_opt_pass.
24883         (make_avr_pass_pre_proep): New function.
24884         (emit_push_sfr) <treg>: Add argument to function and use it
24885         instead of TMP_REG.
24886         (avr_expand_prologue) [machine->gasisr.maybe]: Emit gasisr insn
24887         and set machine->gasisr.yes.
24888         (avr_expand_epilogue) [machine->gasisr.yes]: Similar.
24889         (avr_asm_function_end_prologue) [machine->gasisr.yes]: Add
24890         __gcc_isr.n_pushed to .L__stack_usage.
24891         (TARGET_ASM_FINAL_POSTSCAN_INSN): Define to...
24892         (avr_asm_final_postscan_insn): ...this new static function.
24893         * config/avr/avr.h (machine_function)
24894         <is_no_gccisr, use_L__stack_usage>: New fields.
24895         <gasisr, gasisr.yes, gasisr.maybe, gasisr.regno>: New fields.
24896         * config/avr/avr.md (UNSPECV_GASISR): Add unspecv enum.
24897         (GASISR_Prologue, GASISR_Epilogue, GASISR_Done): New define_constants.
24898         (gasisr, *gasisr): New expander and insn.
24899         * config/avr/gen-avr-mmcu-specs.c (print_mcu)
24900         [HAVE_AS_AVR_MGCCISR_OPTION]: Print asm_gccisr spec.
24901         * config/avr/specs.h (ASM_SPEC) <asm_gccisr>: Add sub spec.
24903 2017-07-10  Richard Earnshaw  <rearnsha@arm.com>
24905         * config/arm/parsecpu.awk (gen_comm_data): Do not escape single quotes
24906         in quoted strings.
24908 2017-07-10  Georg-Johann Lay  <avr@gjlay.de>
24910         Move jump-tables out of .text again.
24912         PR target/81075
24913         * config/avr/avr.c (ASM_OUTPUT_ADDR_VEC_ELT): Remove function.
24914         (ASM_OUTPUT_ADDR_VEC): New function.
24915         (avr_adjust_insn_length) [JUMP_TABLE_DATA_P]: Return 0.
24916         (avr_final_prescan_insn) [avr_log.insn_addresses]: Dump
24917         INSN_ADDRESSes as asm comment.
24918         * config/avr/avr.h (JUMP_TABLES_IN_TEXT_SECTION): Adjust comment.
24919         (ASM_OUTPUT_ADDR_VEC_ELT): Remove define.
24920         (ASM_OUTPUT_ADDR_VEC): Define to avr_output_addr_vec.
24921         * config/avr/avr.md (*tablejump): Adjust comment.
24922         * config/avr/elf.h (ASM_OUTPUT_BEFORE_CASE_LABEL): Remove.
24923         * config/avr/avr-log.c (avr_log_set_avr_log) <insn_addresses>:
24924         New detail.
24925         * config/avr/avr-protos.h (avr_output_addr_vec_elt): Remove proto.
24926         (avr_output_addr_vec): New proto.
24927         (avr_log_t) <insn_addresses>: New field.
24929 2017-07-09  H.J. Lu  <hongjiu.lu@intel.com>
24931         PR target/81313
24932         * config/i386/i386.c (ix86_function_arg_advance): Set
24933         outgoing_args_on_stack to true if there are outgoing arguments
24934         on stack.
24935         (ix86_function_arg): Likewise.
24936         (ix86_get_drap_rtx): Use DRAP only if there are outgoing
24937         arguments on stack and ACCUMULATE_OUTGOING_ARGS is false.
24938         * config/i386/i386.h (machine_function): Add
24939         outgoing_args_on_stack.
24941 2017-07-09  Krister Walfridsson  <krister.walfridsson@gmail.com>
24943         * config.gcc (*-*-netbsd*): Remove check for NetBSD versions not
24944         supporting pthreds.
24945         * config/netbsd.h (NETBSD_LIBGCC_SPEC): Always enable pthreads.
24947 2017-07-08  Richard Sandiford  <richard.sandiford@linaro.org>
24949         * Makefile.in (HOOKS_H, RTL_BASE_H, FUNCTION_H, EXPR_H, REGS_H)
24950         (REAL_H): Remove $(MACHMODE_H).
24951         (FIXED_VALUE_H, TREE_CORE_H, CFGLOOP_H): Remove $(MACHMODE_H) and
24952         double-int.h.
24953         (CORETYPES_H): Add signop.h, wide-int.h, wide-int-print.h,
24954         $(MACHMODE_H) and double-int.h.
24955         (build/min-insn-modes.o): Depend on $(CORETYPES_H) rather than
24956         $(MACHMODE_H).
24957         (gengtype-state.o, gengtype.o, build/gengtype.o): Don't depend on
24958         double-int.h.
24960 2017-07-07  Andrew Pinski  <apinski@cavium.com>
24962         * config/aarch64/aarch64.c (aarch_macro_fusion_pair_p): Check
24963         prev_set and curr_set for AARCH64_FUSE_ALU_BRANCH.
24965 2017-07-07  Michael Meissner  <meissner@linux.vnet.ibm.com>
24967         * config/rs6000/rs6000.c (rs6000_get_function_versions_dispatcher):
24968         Add warning if GCC was not configured to link against a GLIBC that
24969         exports the hardware capability bits.
24970         (make_resolver_func): Make resolver function private and not a
24971         COMDAT function.  Create the name with clone_function_name instead
24972         of make_unique_name.
24974         PR target/81348
24975         * config/rs6000/rs6000.md (HI sign_extend splitter): Use the
24976         correct operand in doing the split.
24978 2017-07-07 Carl Love  <cel@us.ibm.com>
24980         * config/rs6000/rs6000-c: Add support for built-in function
24981         vector unsigned short vec_pack_to_short_fp32 (vector float,
24982                                                       vector float).
24983         * config/rs6000/rs6000-builtin.def (CONVERT_4F32_8I16): Add
24984         BU_P9V_AV_2 and BU_P9V_OVERLOAD_2 definitions.
24985         * config/rs6000/altivec.h (vec_pack_to_short_fp32): Add define.
24986         * config/rs6000/altivec.md(UNSPEC_CONVERT_4F32_8I16): Add UNSPEC.
24987         (convert_4f32_8i16): Add define_expand.
24988         * doc/extend.texi: Update the built-in documentation file for the
24989         new built-in function.
24991 2017-07-07  Jose E. Marchesi  <jose.marchesi@oracle.com>
24993         * config/sparc/m8.md: New file.
24994         * config/sparc/sparc.md: Include m8.md.
24996 2017-07-07  Jose E. Marchesi  <jose.marchesi@oracle.com>
24998         * config/sparc/sparc.opt: New option -mvis4b.
24999         * config/sparc/sparc.c (dump_target_flag_bits): Handle MASK_VIS4B.
25000         (sparc_option_override): Handle VIS4B.
25001         (enum sparc_builtins): Define
25002         SPARC_BUILTIN_DICTUNPACK{8,16,32},
25003         SPARC_BUILTIN_FPCMP{LE,GT,EQ,NE}{8,16,32}SHL,
25004         SPARC_BUILTIN_FPCMPU{LE,GT}{8,16,32}SHL,
25005         SPARC_BUILTIN_FPCMPDE{8,16,32}SHL and
25006         SPARC_BUILTIN_FPCMPUR{8,16,32}SHL.
25007         (check_constant_argument): New function.
25008         (sparc_vis_init_builtins): Define builtins
25009         __builtin_vis_dictunpack{8,16,32},
25010         __builtin_vis_fpcmp{le,gt,eq,ne}{8,16,32}shl,
25011         __builtin_vis_fpcmpu{le,gt}{8,16,32}shl,
25012         __builtin_vis_fpcmpde{8,16,32}shl and
25013         __builtin_vis_fpcmpur{8,16,32}shl.
25014         (sparc_expand_builtin): Check that the constant operands to
25015         __builtin_vis_fpcmp*shl and _builtin_vis_dictunpack* are indeed
25016         constant and in range.
25017         * config/sparc/sparc-c.c (sparc_target_macros): Handle
25018         TARGET_VIS4B.
25019         * config/sparc/sparc.h (SPARC_IMM2_P): Define.
25020         (SPARC_IMM5_P): Likewise.
25021         * config/sparc/sparc.md (cpu_feature): Add new feagure "vis4b".
25022         (enabled): Handle vis4b.
25023         (UNSPEC_DICTUNPACK): New unspec.
25024         (UNSPEC_FPCMPSHL): Likewise.
25025         (UNSPEC_FPUCMPSHL): Likewise.
25026         (UNSPEC_FPCMPDESHL): Likewise.
25027         (UNSPEC_FPCMPURSHL): Likewise.
25028         (cpu_feature): New CPU feature `vis4b'.
25029         (dictunpack{8,16,32}): New insns.
25030         (FPCSMODE): New mode iterator.
25031         (fpcscond): New code iterator.
25032         (fpcsucond): Likewise.
25033         (fpcmp{le,gt,eq,ne}{8,16,32}{si,di}shl): New insns.
25034         (fpcmpu{le,gt}{8,16,32}{si,di}shl): Likewise.
25035         (fpcmpde{8,16,32}{si,di}shl): Likewise.
25036         (fpcmpur{8,16,32}{si,di}shl): Likewise.
25037         * config/sparc/constraints.md: Define constraints `q' for unsigned
25038         2-bit integer constants and `t' for unsigned 5-bit integer
25039         constants.
25040         * config/sparc/predicates.md (imm5_operand_dictunpack8): New
25041         predicate.
25042         (imm5_operand_dictunpack16): Likewise.
25043         (imm5_operand_dictunpack32): Likewise.
25044         (imm2_operand): Likewise.
25045         * doc/invoke.texi (SPARC Options): Document -mvis4b.
25046         * doc/extend.texi (SPARC VIS Built-in Functions): Document the
25047         ditunpack* and fpcmp*shl builtins.
25049 2017-07-07  Jose E. Marchesi  <jose.marchesi@oracle.com>
25051         * config.gcc: Handle m8 in --with-{cpu,tune} options.
25052         * config.in: Add HAVE_AS_SPARC6 define.
25053         * config/sparc/driver-sparc.c (cpu_names): Add entry for the SPARC
25054         M8.
25055         * config/sparc/sol2.h (CPP_CPU64_DEFAULT_SPEC): Define for
25056         TARGET_CPU_m8.
25057         (ASM_CPU32_DEFAUILT_SPEC): Likewise.
25058         (CPP_CPU_SPEC): Handle m8.
25059         (ASM_CPU_SPEC): Likewise.
25060         * config/sparc/sparc-opts.h (enum processor_type): Add
25061         PROCESSOR_M8.
25062         * config/sparc/sparc.c (m8_costs): New struct.
25063         (sparc_option_override): Handle TARGET_CPU_m8.
25064         (sparc32_initialize_trampoline): Likewise.
25065         (sparc64_initialize_trampoline): Likewise.
25066         (sparc_issue_rate): Likewise.
25067         (sparc_register_move_cost): Likewise.
25068         * config/sparc/sparc.h (TARGET_CPU_m8): Define.
25069         (CPP_CPU64_DEFAULT_SPEC): Define for M8.
25070         (ASM_CPU64_DEFAULT_SPEC): Likewise.
25071         (CPP_CPU_SPEC): Handle M8.
25072         (ASM_CPU_SPEC): Likewise.
25073         (AS_M8_FLAG): Define.
25074         * config/sparc/sparc.md: Add m8 to the cpu attribute.
25075         * config/sparc/sparc.opt: New option -mcpu=m8 for sparc targets.
25076         * configure.ac (HAVE_AS_SPARC6): Check for assembler support for
25077         M8 instructions.
25078         * configure: Regenerate.
25079         * doc/invoke.texi (SPARC Options): Document -mcpu=m8 and
25080         -mtune=m8.
25082 2017-07-07  Jose E. Marchesi  <jose.marchesi@oracle.com>
25084         * config/sparc/niagara7.md: Rework the DFA scheduler to use insn
25085         subtypes.
25086         * config/sparc/sparc.md: Remove the `v3pipe' insn attribute.
25087         ("*movdi_insn_sp32"): Do not set v3pipe.
25088         ("*movsi_insn"): Likewise.
25089         ("*movdi_insn_sp64"): Likewise.
25090         ("*movsf_insn"): Likewise.
25091         ("*movdf_insn_sp32"): Likewise.
25092         ("*movdf_insn_sp64"): Likewise.
25093         ("*zero_extendsidi2_insn_sp64"): Likewise.
25094         ("*sign_extendsidi2_insn"): Likewise.
25095         ("*mov<VM32:mode>_insn"): Likewise.
25096         ("*mov<VM64:mode>_insn_sp64"): Likewise.
25097         ("*mov<VM64:mode>_insn_sp32"): Likewise.
25098         ("<plusminus_insn><VADDSUB:mode>3"): Likewise.
25099         ("<vlop:code><VL:mode>3"): Likewise.
25100         ("*not_<vlop:code><VL:mode>3"): Likewise.
25101         ("*nand<VL:mode>_vis"): Likewise.
25102         ("*<vlnotop:code>_not1<VL:mode>_vis"): Likewise.
25103         ("*<vlnotop:code>_not2<VL:mode>_vis"): Likewise.
25104         ("one_cmpl<VL:mode>2"): Likewise.
25105         ("faligndata<VM64:mode>_vis"): Likewise.
25106         ("alignaddrsi_vis"): Likewise.
25107         ("alignaddrdi_vis"): Likweise.
25108         ("alignaddrlsi_vis"): Likewise.
25109         ("alignaddrldi_vis"): Likewise.
25110         ("fcmp<gcond:code><GCM:gcm_name><P:mode>_vis"): Likewise.
25111         ("bmaskdi_vis"): Likewise.
25112         ("bmasksi_vis"): Likewise.
25113         ("bshuffle<VM64:mode>_vis"): Likewise.
25114         ("cmask8<P:mode>_vis"): Likewise.
25115         ("cmask16<P:mode>_vis"): Likewise.
25116         ("cmask32<P:mode>_vis"): Likewise.
25117         ("pdistn<P:mode>_vis"): Likewise.
25118         ("<vis3_addsub_ss_patname><VASS:mode>3"): Likewise.
25120 2017-07-07  Jose E. Marchesi  <jose.marchesi@oracle.com>
25122         * config/sparc/sparc.md ("subtype"): New insn attribute.
25123         ("*wrgsr_sp64"): Set insn subtype.
25124         ("*rdgsr_sp64"): Likewise.
25125         ("alignaddrsi_vis"): Likewise.
25126         ("alignaddrdi_vis"): Likewise.
25127         ("alignaddrlsi_vis"): Likewise.
25128         ("alignaddrldi_vis"): Likewise.
25129         ("<plusminus_insn><VADDSUB:mode>3"): Likewise.
25130         ("fexpand_vis"): Likewise.
25131         ("fpmerge_vis"): Likewise.
25132         ("faligndata<VM64:mode>_vis"): Likewise.
25133         ("bshuffle<VM64:mode>_vis"): Likewise.
25134         ("cmask8<P:mode>_vis"): Likewise.
25135         ("cmask16<P:mode>_vis"): Likewise.
25136         ("cmask32<P:mode>_vis"): Likewise.
25137         ("fchksm16_vis"): Likewise.
25138         ("v<vis3_shift_patname><GCM:mode>3"): Likewise.
25139         ("fmean16_vis"): Likewise.
25140         ("fp<plusminus_insn>64_vis"): Likewise.
25141         ("<plusminus_insn>v8qi3"): Likewise.
25142         ("<vis3_addsub_ss_patname><VASS:mode>3"): Likewise.
25143         ("<vis4_minmax_patname><VMMAX:mode>3"): Likewise.
25144         ("<vis4_uminmax_patname><VMMAX:mode>3"): Likewise.
25145         ("<vis3_addsub_ss_patname>v8qi3"): Likewise.
25146         ("<vis4_addsub_us_patname><VAUS:mode>3"): Likewise.
25147         ("*movqi_insn"): Likewise.
25148         ("*movhi_insn"): Likewise.
25149         ("*movsi_insn"): Likewise.
25150         ("movsi_pic_gotdata_op"): Likewise.
25151         ("*movdi_insn_sp32"): Likewise.
25152         ("*movdi_insn_sp64"): Likewise.
25153         ("movdi_pic_gotdata_op"): Likewise.
25154         ("*movsf_insn"): Likewise.
25155         ("*movdf_insn_sp32"): Likewise.
25156         ("*movdf_insn_sp64"): Likewise.
25157         ("*zero_extendhisi2_insn"): Likewise.
25158         ("*zero_extendqihi2_insn"): Likewise.
25159         ("*zero_extendqisi2_insn"): Likewise.
25160         ("*zero_extendqidi2_insn"): Likewise.
25161         ("*zero_extendhidi2_insn"): Likewise.
25162         ("*zero_extendsidi2_insn_sp64"): Likewise.
25163         ("ldfsr"): Likewise.
25164         ("prefetch_64"): Likewise.
25165         ("prefetch_32"): Likewise.
25166         ("tie_ld32"): Likewise.
25167         ("tie_ld64"): Likewise.
25168         ("*tldo_ldub_sp32"): Likewise.
25169         ("*tldo_ldub1_sp32"): Likewise.
25170         ("*tldo_ldub2_sp32"): Likewise.
25171         ("*tldo_ldub_sp64"): Likewise.
25172         ("*tldo_ldub1_sp64"): Likewise.
25173         ("*tldo_ldub2_sp64"): Likewise.
25174         ("*tldo_ldub3_sp64"): Likewise.
25175         ("*tldo_lduh_sp32"): Likewise.
25176         ("*tldo_lduh1_sp32"): Likewise.
25177         ("*tldo_lduh_sp64"): Likewise.
25178         ("*tldo_lduh1_sp64"): Likewise.
25179         ("*tldo_lduh2_sp64"): Likewise.
25180         ("*tldo_lduw_sp32"): Likewise.
25181         ("*tldo_lduw_sp64"): Likewise.
25182         ("*tldo_lduw1_sp64"): Likewise.
25183         ("*tldo_ldx_sp64"): Likewise.
25184         ("*mov<VM32:mode>_insn"): Likewise.
25185         ("*mov<VM64:mode>_insn_sp64"): Likewise.
25186         ("*mov<VM64:mode>_insn_sp32"): Likewise.
25188 2017-07-07  Jose E. Marchesi  <jose.marchesi@oracle.com>
25190         * config/sparc/sparc.md ("type"): New insn type viscmp.
25191         ("fcmp<gcond:code><GCM:gcm_name><P:mode>_vis"): Set insn type to
25192         viscmp.
25193         ("fpcmp<gcond:code>8<P:mode>_vis"): Likewise.
25194         ("fucmp<gcond:code>8<P:mode>_vis"): Likewise.
25195         ("fpcmpu<gcond:code><GCM:gcm_name><P:mode>_vis"): Likewise.
25196         * config/sparc/niagara7.md ("n7_vis_logical_v3pipe"): Handle
25197         viscmp.
25198         ("n7_vis_logical_11cycle"): Likewise.
25199         * config/sparc/niagara4.md ("n4_vis_logical"): Likewise.
25200         * config/sparc/niagara2.md ("niag3_vis": Likewise.
25201         * config/sparc/niagara.md ("niag_vis"): Likewise.
25202         * config/sparc/ultra3.md ("us3_fga"): Likewise.
25203         * config/sparc/ultra1_2.md ("us1_fga_double"): Likewise.
25205 2017-07-07  Jose E. Marchesi  <jose.marchesi@oracle.com>
25207         * config/sparc/sparc.md: New instruction type `bmask'.
25208         (bmaskdi_vis): Use the `bmask' type.
25209         (bmasksi_vis): Likewise.
25210         * config/sparc/ultra3.md (us3_array): Likewise.
25211         * config/sparc/niagara7.md (n7_array): Likewise.
25212         * config/sparc/niagara4.md (n4_array): Likewise.
25213         * config/sparc/niagara2.md (niag2_vis): Likewise.
25214         (niag3_vis): Likewise.
25215         * config/sparc/niagara.md (niag_vis): Likewise.
25217 2017-07-06  Jan Hubicka  <hubicka@ucw.cz>
25219         * ipa-comdats.c: Remove optimize check from gate.
25220         * ipa-fnsummary.c (ipa_fn_summary_generate): do not generate summary
25221         for functions not optimized.
25222         (ipa_fn_summary_read): Skip optimize check.
25223         (ipa_fn_summary_write): Likewise.
25224         * ipa-inline-analysis.c (do_estimate_growth_1): Check that caller
25225         is optimized.
25226         * ipa-inline.c (can_inline_edge_p): Not optimized functions are
25227         uninlinable.
25228         (can_inline_edge_p): Check flag_pcc_struct_return for match.
25229         (check_callers): Give up on caller which is not optimized.
25230         (inline_small_functions): Likewise.
25231         (ipa_inline): Do not give up when not optimizing.
25232         * ipa-visbility.c (function_and_variable_visibility): Do not optimize
25233         away unoptimizes cdtors.
25234         (whole_program_function_and_variable_visibility): Do
25235         ipa_discover_readonly_nonaddressable_vars in LTO mode.
25236         * ipa.c (process_references): Do not check optimize.
25237         (symbol_table::remove_unreachable_nodes): Update optimize check.
25238         (set_writeonly_bit): Update optimize check.
25239         (pass_ipa_cdtor_merge::gate): Do not check optimize.
25240         (pass_ipa_single_use::gate): Remove.
25242 2017-07-06  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
25244         * config/rs6000/rs6000.c (union_defs, union_uses, insn_is_load_p,
25245         insn_is_store_p, insn_is_swap_p, const_load_sequence_p, v2df_reduction_p,
25246         rtx_is_swappable_p, insn_is_swappable_p, chain_contains_only_swaps,
25247         mark_swaps_for_removal, swap_const_vector_halves, adjust_subreg_index,
25248         permute_load, permute_store, adjust_extract, adjust_splat,
25249         adjust_xxpermdi, adjust_concat, adjust_vperm, handle_special_swappables,
25250         replace_swap_with_copy, dump_swap_insn_table,
25251         alignment_with_canonical_addr, alignment_mask, find_alignment_op,
25252         recombine_lvx_pattern, recombine_stvx_pattern,
25253         recombine_lvx_stvx_patterns, rs6000_analyze_swaps,
25254         make_pass_analyze_swaps): Move all code related to p8 swap optimizations
25255         to file rs6000-p8swap.c.
25256         * config/rs6000/rs6000-p8swap.c: New file.
25257         * config/rs6000/t-rs6000: Add rule to build rs6000-p8swap.o.
25258         * config.gcc: Add rs6000-p8swap.o to extra_objs for powerpc*-*-*
25259         and rs6000*-*-* targets.
25261 2017-07-06  David Malcolm  <dmalcolm@redhat.com>
25263         * Makefile.in (selftest): Remove dependency on s-selftest-c++.
25265 2017-07-06  Jan Hubicka  <hubicka@ucw.cz>
25267         * lto-wrapper.c (merge_and_complain): Do not merge
25268         fexceptions, fnon_call_exceptions, ftrapv, ffp_contract_, fmath_errno,
25269         fsigned_zeros, ftrapping_math, fwrapv.
25270         (append_compiler_options): Do not track these options.
25271         (append_linker_options): Likewie
25273 2017-07-06  Jan Hubicka  <hubicka@ucw.cz>
25275         * cgraphunit.c (cgraph_node::finalize_function): When
25276         !flag_toplevel_reorde set no_reorder flag.
25277         (varpool_node::finalize_decl): Likewise.
25278         (symbol_table::compile): Drop no toplevel reorder path.
25280 2017-07-06  Jan Hubicka  <hubicka@ucw.cz>
25282         * bb-reorder.c (better_edge_p): Do not build traces across abnormal/eh
25283         edges; zero probability is not better than uninitialized.
25285 2017-07-06  Maxim Ostapenko  <m.ostapenko@samsung.com>
25287         * asan.h (asan_sanitize_allocas_p): Declare.
25288         * asan.c (asan_sanitize_allocas_p): New function.
25289         (handle_builtin_stack_restore): Bail out if !asan_sanitize_allocas_p.
25290         (handle_builtin_alloca): Likewise.
25291         * cfgexpand.c (expand_used_vars): Do not add allocas unpoisoning stuff
25292         if !asan_sanitize_allocas_p.
25293         * params.def (asan-instrument-allocas): Add new option.
25294         * params.h (ASAN_PROTECT_ALLOCAS): Define.
25295         * opts.c (common_handle_option): Disable allocas sanitization for
25296         KASan by default.
25298 2017-07-06  Maxim Ostapenko  <m.ostapenko@samsung.com>
25300         * asan.c: Include gimple-fold.h.
25301         (get_last_alloca_addr): New function.
25302         (handle_builtin_stackrestore): Likewise.
25303         (handle_builtin_alloca): Likewise.
25304         (asan_emit_allocas_unpoison): Likewise.
25305         (get_mem_refs_of_builtin_call): Add new parameter, remove const
25306         quallifier from first paramerer. Handle BUILT_IN_ALLOCA,
25307         BUILT_IN_ALLOCA_WITH_ALIGN and BUILT_IN_STACK_RESTORE builtins.
25308         (instrument_builtin_call): Pass gimple iterator to
25309         get_mem_refs_of_builtin_call.
25310         (last_alloca_addr): New global.
25311         * asan.h (asan_emit_allocas_unpoison): Declare.
25312         * builtins.c (expand_asan_emit_allocas_unpoison): New function.
25313         (expand_builtin): Handle BUILT_IN_ASAN_ALLOCAS_UNPOISON.
25314         * cfgexpand.c (expand_used_vars): Call asan_emit_allocas_unpoison
25315         if function calls alloca.
25316         * gimple-fold.c (replace_call_with_value): Remove static keyword.
25317         * gimple-fold.h (replace_call_with_value): Declare.
25318         * internal-fn.c: Include asan.h.
25319         * sanitizer.def (BUILT_IN_ASAN_ALLOCA_POISON,
25320         BUILT_IN_ASAN_ALLOCAS_UNPOISON): New builtins.
25322 2017-07-06  David Malcolm  <dmalcolm@redhat.com>
25324         * Makefile.in (SELFTEST_FLAGS): Drop "-x c", moving it to...
25325         (C_SELFTEST_FLAGS): New.
25326         (CPP_SELFTEST_FLAGS): New.
25327         (SELFTEST_DEPS): New, from deps of s-selftest.
25328         (C_SELFTEST_DEPS): New, from deps of s-selftest.
25329         (CPP_SELFTEST_DEPS): New.
25330         (selftest): Add dependency on s-selftest-c++.
25331         (s-selftest): Rename to...
25332         (s-selftest-c): ...this, moving deps to SELFTEST_DEPS
25333         and C_SELFTEST_DEPS, and using C_SELFTEST_FLAGS rather
25334         than SELFTEST_FLAGS.
25335         (selftest-gdb): Rename to...
25336         (selftest-c-gdb): ...this, using C_SELFTEST_DEPS and
25337         C_SELFTEST_FLAGS.
25338         (selftest-gdb): Reintroduce as an alias for selftest-c-gdb.
25339         (selftest-valgrind): Rename to...
25340         (selftest-c-valgrind): ...this, using C_SELFTEST_DEPS and
25341         C_SELFTEST_FLAGS.
25342         (selftest-valgrind): Reintroduce as an alias for
25343         selftest-c-valgrind.
25344         (s-selftest-c++): New.
25345         (selftest-c++-gdb): New.
25346         (selftest-c++-valgrind): New.
25348 2017-07-06  Olivier Hainque  <hainque@adacore.com>
25350         * gcc.c (process_command): When deciding if undefined variables
25351         should be ignored when processing specs, accept "gcc -v" as well.
25353 2017-07-06  Jan Hubicka  <hubicka@ucw.cz>
25355         * auto-profile.c (afdo_set_bb_count, afdo_propagate_edge,
25356         afdo_annotate_cfg): Set counts/probabilities as determined by afdo.
25358 2017-07-06  Thomas Preud'homme  <thomas.preudhomme@arm.com>
25360         * config/arm/arm-cpus.in (armv8-r): Add new entry.
25361         * config/arm/arm-isa.h (ISA_ARMv8r): Define macro.
25362         * config/arm/arm-tables.opt: Regenerate.
25363         * config/arm/arm.h (enum base_architecture): Add BASE_ARCH_8R
25364         enumerator.
25365         * doc/invoke.texi: Mention -march=armv8-r and its extensions.
25367 2017-07-06  Carl Love  <cel@us.ibm.com>
25369         * ChangeLog: Clean up from mid air collision
25371 2017-07-06  Carl Love  <cel@us.ibm.com>
25373         * config/rs6000/rs6000-c.c: Add support for built-in functions
25374         vector signed int vec_subc (vector signed int, vector signed int);
25375         vector signed __int128 vec_subc (vector signed __int128,
25376                                          vector signed __int128);
25377         vector unsigned __int128 vec_subc (vector unsigned __int128,
25378                                            vector unsigned __int128);
25379         vector signed int vec_sube (vector signed int, vector signed int,
25380                                     vector signed int);
25381         vector unsigned int vec_sube (vector unsigned int,
25382                                       vector unsigned int,
25383                                       vector unsigned int);
25384         vector signed __int128 vec_sube (vector signed __int128,
25385                                          vector signed __int128,
25386                                          vector signed__int128);
25387         vector unsigned __int128 vec_sube (vector unsigned __int128,
25388                                            vector unsigned __int128,
25389                                            vector unsigned __int128);
25390         vector signed int vec_subec (vector signed int, vector signed int,
25391                                      vector signed int);
25392         vector unsigned int vec_subec (vector unsigned int,
25393                                        vector unsigned int,
25394                                        vector unsigned int);
25395         vector signed __int128 vec_subec (vector signed __int128,
25396                                           vector signed __int128,
25397                                           vector signed__int128);
25398         vector unsigned __int128 vec_subec (vector unsigned __int128,
25399                                             vector unsigned __int128,
25400                                             vector unsigned __int128);
25401         * config/rs6000/rs6000.c (ALTIVEC_BUILTIN_VEC_SUBE,
25402         ALTIVEC_BUILTIN_VEC_SUBEC): Add ef_builtins.
25403         * config/rs6000/rs6000-builtin.def (SUBE, SUBEC): Add
25404         BU_ALTIVEC_OVERLOAD_X definitions.
25405         * config/rs6000/altivec.h (vec_sube, vec_subec): Add builtin defines.
25406         * doc/extend.texi: Update the built-in documentation file for the new
25407         built-in functions.
25409 2017-07-06  David Malcolm  <dmalcolm@redhat.com>
25411         PR c++/79300
25412         * diagnostic-show-locus.c (layout::layout): Use start and finish
25413         spelling location for the start and finish of each range.
25414         * genmatch.c (linemap_client_expand_location_to_spelling_point):
25415         Add unused aspect param.
25416         * input.c (expand_location_1): Add "aspect" param, and use it
25417         to access the correct part of the location.
25418         (expand_location): Pass LOCATION_ASPECT_CARET to new param of
25419         expand_location_1.
25420         (expand_location_to_spelling_point): Likewise.
25421         (linemap_client_expand_location_to_spelling_point): Add "aspect"
25422         param, and pass it to expand_location_1.
25424 2017-07-06  Sebastian Peryt  <sebastian.peryt@intel.com>
25426         * config/i386/avx512fintrin.h (_mm_mask_getexp_round_ss,
25427         _mm_maskz_getexp_round_ss, _mm_mask_getexp_round_sd,
25428         _mm_maskz_getexp_round_sd, _mm_mask_getmant_round_sd,
25429         _mm_maskz_getmant_round_sd, _mm_mask_getmant_round_ss,
25430         _mm_maskz_getmant_round_ss, _mm_mask_getexp_ss, _mm_maskz_getexp_ss,
25431         _mm_mask_getexp_sd, _mm_maskz_getexp_sd, _mm_mask_getmant_sd,
25432         _mm_maskz_getmant_sd, _mm_mask_getmant_ss,
25433         _mm_maskz_getmant_ss): New intrinsics.
25434         (__builtin_ia32_getexpss128_mask): Changed to ...
25435         __builtin_ia32_getexpss128_round ... this.
25436         (__builtin_ia32_getexpsd128_mask): Changed to ...
25437         __builtin_ia32_getexpsd128_round ... this.
25438         * config/i386/i386-builtin-types.def
25439         ((V2DF, V2DF, V2DF, INT, V2DF, UQI, INT),
25440         (V4SF, V4SF, V4SF, INT, V4SF, UQI, INT)): New function type aliases.
25441         * config/i386/i386-builtin.def (__builtin_ia32_getexpsd_mask_round,
25442         __builtin_ia32_getexpss_mask_round,     __builtin_ia32_getmantsd_mask_round,
25443         __builtin_ia32_getmantss_mask_round): New builtins.
25444         * config/i386/i386.c (V2DF_FTYPE_V2DF_V2DF_INT_V2DF_UQI_INT,
25445         V4SF_FTYPE_V4SF_V4SF_INT_V4SF_UQI_INT): Handle new types.
25446         (CODE_FOR_avx512f_vgetmantv2df_mask_round,
25447         CODE_FOR_avx512f_vgetmantv4sf_mask_round): New cases.
25448         * config/i386/sse.md
25449         (avx512f_sgetexp<mode><round_saeonly_name>): Changed to ...
25450         avx512f_sgetexp<mode><mask_scalar_name>
25451         <round_saeonly_scalar_name> ... this.
25452         (vgetexp<ssescalarmodesuffix>\t{<round_saeonly_op3>%2, %1, %0|
25453         %0, %1, %2<round_saeonly_op3>}): Changed to ...
25454         vgetexp<ssescalarmodesuffix>
25455         \t{<round_saeonly_scalar_mask_op3>%2, %1, %0<mask_scalar_operand3>|
25456         %0<mask_scalar_operand3>, %1, %2<round_saeonly_scalar_mask_op3>} ... this.
25457         (avx512f_vgetmant<mode><round_saeonly_name>): Changed to ...
25458         avx512f_vgetmant<mode><mask_scalar_name>
25459         <round_saeonly_scalar_name> ... this.
25460         (vgetmant<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
25461         %0, %1, %2<round_saeonly_op4>, %3}): Changed to ...
25462         vgetmant<ssescalarmodesuffix>
25463         \t{%3, <round_saeonly_scalar_mask_op4>%2, %1, %0<mask_scalar_operand4>|
25464         %0<mask_scalar_operand4>, %1, %2
25465         <round_saeonly_scalar_mask_op4>, %3} ... this.
25466         * config/i386/subst.md (mask_scalar_operand4,
25467         round_saeonly_scalar_mask_operand4,     round_saeonly_scalar_mask_op4,
25468         round_saeonly_scalar_nimm_predicate): New subst attributes.
25470 2017-07-06  Julia Koval  <julia.koval@intel.com>
25472         * config/i386/i386.c (ix86_erase_embedded_rounding):
25473         Remove code for old rounding pattern.
25475 2017-07-06  Richard Earnshaw  <rearnsha@arm.com>
25477         * config/arm/t-arm (GTM_H): Add arm-cpu.h.
25479 2017-07-06  Christophe Lyon  <christophe.lyon@linaro.org>
25481         * doc/sourcebuild.texi (Test Directives, Variants of
25482         dg-require-support): Add documentation for dg-require-stack-check.
25484 2017-07-05  Sebastian Peryt  <sebastian.peryt@intel.com>
25486         * config/i386/subst.md (mask_scalar, round_scalar,
25487         round_saeonly_scalar): New meta-templates.
25488         (mask_scalar_name, mask_scalar_operand3, round_scalar_name,
25489         round_scalar_mask_operand3, round_scalar_mask_op3,
25490         round_scalar_constraint, round_scalar_prefix, round_saeonly_scalar_name,
25491         round_saeonly_scalar_mask_operand3, round_saeonly_scalar_mask_op3,
25492         round_saeonly_scalar_constraint,
25493         round_saeonly_scalar_prefix): New subst attribute.
25494         * config/i386/sse.md
25495         (<sse>_vm<plusminus_insn><mode>3<mask_name><round_name>): Renamed to ...
25496         <sse>_vm<plusminus_insn><mode>3<mask_scalar_name>
25497         <round_scalar_name> ... this.
25498         (<sse>_vm<multdiv_mnemonic><mode>3<mask_name><round_name>): Renamed to ...
25499         <sse>_vm<multdiv_mnemonic><mode>3<mask_scalar_name>
25500         <round_scalar_name> ... this.
25501         (<sse>_vm<code><mode>3<mask_name><round_saeonly_name>): Renamed to ...
25502         <sse>_vm<code><mode>3<mask_scalar_name>
25503         <round_saeonly_scalar_name> ... this.
25504         (v<plusminus_mnemonic><ssescalarmodesuffix>
25505         \t{<round_mask_op3>%2, %1, %0<mask_operand3>|
25506         %0<mask_operand3>, %1, %<iptr>2<round_mask_op3>}): Changed to ...
25507         v<plusminus_mnemonic><ssescalarmodesuffix>
25508         \t{<round_scalar_mask_op3>%2, %1, %0<mask_scalar_operand3>|
25509         %0<mask_scalar_operand3>, %1, %<iptr>2<round_scalar_mask_op3>} ... this.
25510         (v<multdiv_mnemonic><ssescalarmodesuffix>
25511         \t{<round_mask_op3>%2, %1, %0<mask_operand3>|
25512         %0<mask_operand3>, %1, %<iptr>2<round_mask_op3>}): Changed to ...
25513         v<multdiv_mnemonic><ssescalarmodesuffix>
25514         \t{<round_scalar_mask_op3>%2, %1, %0<mask_scalar_operand3>|
25515         %0<mask_scalar_operand3>, %1, %<iptr>2<round_scalar_mask_op3>} ... this.
25516         (v<maxmin_float><ssescalarmodesuffix>
25517         \t{<round_saeonly_mask_op3>%2, %1, %0<mask_operand3>|
25518         %0<mask_operand3>, %1, %<iptr>2<round_saeonly_mask_op3>}): Changed to ...
25519         v<maxmin_float><ssescalarmodesuffix>
25520         \t{<round_saeonly_scalar_mask_op3>%2, %1, %0<mask_scalar_operand3>|
25521         %0<mask_scalar_operand3>, %1, %<iptr>2
25522         <round_saeonly_scalar_mask_op3>} ... this.
25524 2017-07-05  Richard Earnshaw  <rearnsha@arm.com>
25526         * config/arm/arm.c (arm_fixed_condition_code_regs): New function.
25527         (TARGET_FIXED_CONDITION_CODE_REGS): Redefine.
25529 2017-07-05  Richard Sandiford  <richard.sandiford@linaro.org>
25530             Alan Hayward  <alan.hayward@arm.com>
25531             David Sherwood  <david.sherwood@arm.com>
25533         * combine.c (simplify_if_then_else): Remove "enum" before
25534         "machine_mode".
25535         * compare-elim.c (can_eliminate_compare): Likewise.
25536         * config/aarch64/aarch64-builtins.c (aarch64_simd_builtin_std_type):
25537         Likewise.
25538         (aarch64_lookup_simd_builtin_type): Likewise.
25539         (aarch64_simd_builtin_type): Likewise.
25540         (aarch64_init_simd_builtin_types): Likewise.
25541         (aarch64_simd_expand_args): Likewise.
25542         * config/aarch64/aarch64-protos.h (aarch64_simd_attr_length_rglist):
25543         Likewise.
25544         (aarch64_reverse_mask): Likewise.
25545         (aarch64_simd_emit_reg_reg_move): Likewise.
25546         (aarch64_gen_adjusted_ldpstp): Likewise.
25547         (aarch64_ccmp_mode_to_code): Likewise.
25548         (aarch64_operands_ok_for_ldpstp): Likewise.
25549         (aarch64_operands_adjust_ok_for_ldpstp): Likewise.
25550         * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
25551         Likewise.
25552         (aarch64_min_divisions_for_recip_mul): Likewise.
25553         (aarch64_reassociation_width): Likewise.
25554         (aarch64_get_condition_code_1): Likewise.
25555         (aarch64_simd_emit_reg_reg_move): Likewise.
25556         (aarch64_simd_attr_length_rglist): Likewise.
25557         (aarch64_reverse_mask): Likewise.
25558         (aarch64_operands_ok_for_ldpstp): Likewise.
25559         (aarch64_operands_adjust_ok_for_ldpstp): Likewise.
25560         (aarch64_gen_adjusted_ldpstp): Likewise.
25561         * config/aarch64/cortex-a57-fma-steering.c (fma_node::rename):
25562         Likewise.
25563         * config/arc/arc.c (legitimate_offset_address_p): Likewise.
25564         * config/arm/arm-builtins.c (arm_simd_builtin_std_type): Likewise.
25565         (arm_lookup_simd_builtin_type): Likewise.
25566         (arm_simd_builtin_type): Likewise.
25567         (arm_init_simd_builtin_types): Likewise.
25568         (arm_expand_builtin_args): Likewise.
25569         * config/arm/arm-protos.h (arm_expand_builtin): Likewise.
25570         * config/ft32/ft32.c (ft32_libcall_value): Likewise.
25571         (ft32_setup_incoming_varargs): Likewise.
25572         (ft32_function_arg): Likewise.
25573         (ft32_function_arg_advance): Likewise.
25574         (ft32_pass_by_reference): Likewise.
25575         (ft32_arg_partial_bytes): Likewise.
25576         (ft32_valid_pointer_mode): Likewise.
25577         (ft32_addr_space_pointer_mode): Likewise.
25578         (ft32_addr_space_legitimate_address_p): Likewise.
25579         * config/i386/i386-protos.h (ix86_operands_ok_for_move_multiple):
25580         Likewise.
25581         * config/i386/i386.c (ix86_setup_incoming_vararg_bounds): Likewise.
25582         (ix86_emit_outlined_ms2sysv_restore): Likewise.
25583         (iamcu_alignment): Likewise.
25584         (canonicalize_vector_int_perm): Likewise.
25585         (ix86_noce_conversion_profitable_p): Likewise.
25586         (ix86_mpx_bound_mode): Likewise.
25587         (ix86_operands_ok_for_move_multiple): Likewise.
25588         * config/microblaze/microblaze-protos.h
25589         (microblaze_expand_conditional_branch_reg): Likewise.
25590         * config/microblaze/microblaze.c
25591         (microblaze_expand_conditional_branch_reg): Likewise.
25592         * config/powerpcspe/powerpcspe.c (rs6000_init_hard_regno_mode_ok):
25593         Likewise.
25594         (rs6000_reassociation_width): Likewise.
25595         (rs6000_invalid_binary_op): Likewise.
25596         (fusion_p9_p): Likewise.
25597         (emit_fusion_p9_load): Likewise.
25598         (emit_fusion_p9_store): Likewise.
25599         * config/riscv/riscv-protos.h (riscv_regno_mode_ok_for_base_p):
25600         Likewise.
25601         (riscv_hard_regno_mode_ok_p): Likewise.
25602         (riscv_address_insns): Likewise.
25603         (riscv_split_symbol): Likewise.
25604         (riscv_legitimize_move): Likewise.
25605         (riscv_function_value): Likewise.
25606         (riscv_hard_regno_nregs): Likewise.
25607         (riscv_expand_builtin): Likewise.
25608         * config/riscv/riscv.c (riscv_build_integer_1): Likewise.
25609         (riscv_build_integer): Likewise.
25610         (riscv_split_integer): Likewise.
25611         (riscv_legitimate_constant_p): Likewise.
25612         (riscv_cannot_force_const_mem): Likewise.
25613         (riscv_regno_mode_ok_for_base_p): Likewise.
25614         (riscv_valid_base_register_p): Likewise.
25615         (riscv_valid_offset_p): Likewise.
25616         (riscv_valid_lo_sum_p): Likewise.
25617         (riscv_classify_address): Likewise.
25618         (riscv_legitimate_address_p): Likewise.
25619         (riscv_address_insns): Likewise.
25620         (riscv_load_store_insns): Likewise.
25621         (riscv_force_binary): Likewise.
25622         (riscv_split_symbol): Likewise.
25623         (riscv_force_address): Likewise.
25624         (riscv_legitimize_address): Likewise.
25625         (riscv_move_integer): Likewise.
25626         (riscv_legitimize_const_move): Likewise.
25627         (riscv_legitimize_move): Likewise.
25628         (riscv_address_cost): Likewise.
25629         (riscv_subword): Likewise.
25630         (riscv_output_move): Likewise.
25631         (riscv_canonicalize_int_order_test): Likewise.
25632         (riscv_emit_int_order_test): Likewise.
25633         (riscv_function_arg_boundary): Likewise.
25634         (riscv_pass_mode_in_fpr_p): Likewise.
25635         (riscv_pass_fpr_single): Likewise.
25636         (riscv_pass_fpr_pair): Likewise.
25637         (riscv_get_arg_info): Likewise.
25638         (riscv_function_arg): Likewise.
25639         (riscv_function_arg_advance): Likewise.
25640         (riscv_arg_partial_bytes): Likewise.
25641         (riscv_function_value): Likewise.
25642         (riscv_pass_by_reference): Likewise.
25643         (riscv_setup_incoming_varargs): Likewise.
25644         (riscv_print_operand): Likewise.
25645         (riscv_elf_select_rtx_section): Likewise.
25646         (riscv_save_restore_reg): Likewise.
25647         (riscv_for_each_saved_reg): Likewise.
25648         (riscv_register_move_cost): Likewise.
25649         (riscv_hard_regno_mode_ok_p): Likewise.
25650         (riscv_hard_regno_nregs): Likewise.
25651         (riscv_class_max_nregs): Likewise.
25652         (riscv_memory_move_cost): Likewise.
25653         * config/rl78/rl78-protos.h (rl78_split_movsi): Likewise.
25654         * config/rl78/rl78.c (rl78_split_movsi): Likewise.
25655         (rl78_addr_space_address_mode): Likewise.
25656         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
25657         Likewise.
25658         * config/rs6000/rs6000.c (rs6000_init_hard_regno_mode_ok): Likewise.
25659         (rs6000_reassociation_width): Likewise.
25660         (rs6000_invalid_binary_op): Likewise.
25661         (fusion_p9_p): Likewise.
25662         (emit_fusion_p9_load): Likewise.
25663         (emit_fusion_p9_store): Likewise.
25664         * config/visium/visium-protos.h (prepare_move_operands): Likewise.
25665         (ok_for_simple_move_operands): Likewise.
25666         (ok_for_simple_move_strict_operands): Likewise.
25667         (ok_for_simple_arith_logic_operands): Likewise.
25668         (visium_legitimize_reload_address): Likewise.
25669         (visium_select_cc_mode): Likewise.
25670         (output_cbranch): Likewise.
25671         (visium_split_double_move): Likewise.
25672         (visium_expand_copysign): Likewise.
25673         (visium_expand_int_cstore): Likewise.
25674         (visium_expand_fp_cstore): Likewise.
25675         * config/visium/visium.c (visium_pass_by_reference): Likewise.
25676         (visium_function_arg): Likewise.
25677         (visium_function_arg_advance): Likewise.
25678         (visium_libcall_value): Likewise.
25679         (visium_setup_incoming_varargs): Likewise.
25680         (visium_legitimate_constant_p): Likewise.
25681         (visium_legitimate_address_p): Likewise.
25682         (visium_legitimize_address): Likewise.
25683         (visium_secondary_reload): Likewise.
25684         (visium_register_move_cost): Likewise.
25685         (visium_memory_move_cost): Likewise.
25686         (prepare_move_operands): Likewise.
25687         (ok_for_simple_move_operands): Likewise.
25688         (ok_for_simple_move_strict_operands): Likewise.
25689         (ok_for_simple_arith_logic_operands): Likewise.
25690         (visium_function_value_1): Likewise.
25691         (rtx_ok_for_offset_p): Likewise.
25692         (visium_legitimize_reload_address): Likewise.
25693         (visium_split_double_move): Likewise.
25694         (visium_expand_copysign): Likewise.
25695         (visium_expand_int_cstore): Likewise.
25696         (visium_expand_fp_cstore): Likewise.
25697         (visium_split_cstore): Likewise.
25698         (visium_select_cc_mode): Likewise.
25699         (visium_split_cbranch): Likewise.
25700         (output_cbranch): Likewise.
25701         (visium_print_operand_address): Likewise.
25702         * expmed.c (flip_storage_order): Likewise.
25703         * expmed.h (emit_cstore): Likewise.
25704         (flip_storage_order): Likewise.
25705         * genrecog.c (validate_pattern): Likewise.
25706         * hsa-gen.c (gen_hsa_addr): Likewise.
25707         * internal-fn.c (expand_arith_overflow): Likewise.
25708         * ira-color.c (allocno_copy_cost_saving): Likewise.
25709         * lra-assigns.c (find_hard_regno_for_1): Likewise.
25710         * lra-constraints.c (prohibited_class_reg_set_mode_p): Likewise.
25711         (process_invariant_for_inheritance): Likewise.
25712         * lra-eliminations.c (move_plus_up): Likewise.
25713         * omp-low.c (lower_oacc_reductions): Likewise.
25714         * simplify-rtx.c (simplify_subreg): Likewise.
25715         * target.def (TARGET_SETUP_INCOMING_VARARG_BOUNDS): Likewise.
25716         (TARGET_CHKP_BOUND_MODE): Likewise..
25717         * targhooks.c (default_chkp_bound_mode): Likewise.
25718         (default_setup_incoming_vararg_bounds): Likewise.
25719         * targhooks.h (default_chkp_bound_mode): Likewise.
25720         (default_setup_incoming_vararg_bounds): Likewise.
25721         * tree-ssa-math-opts.c (divmod_candidate_p): Likewise.
25722         * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
25723         (have_whole_vector_shift): Likewise.
25724         * tree-vect-stmts.c (vectorizable_load): Likewise.
25725         * doc/tm.texi: Regenerate.
25727 2017-07-05  Georg-Johann Lay  <avr@gjlay.de>
25729         Graceful degrade if Binutils PR21472 is not available.
25731         PR target/81072
25732         * configure.ac [target=avr]: WARN instead of ERROR if avrxmega3
25733         .rodata in flash test fails.
25734         (HAVE_LD_AVR_AVRXMEGA3_RODATA_IN_FLASH): Define it if test passes.
25735         * confgure: Regenerate.
25736         * config.in: Regenerate.
25737         * config/avr/avr.c (avr_asm_named_section)
25738         [HAVE_LD_AVR_AVRXMEGA3_RODATA_IN_FLASH]: Only trigger
25739         __do_copy_data for stuff in .rodata if flash_pm_offset = 0.
25740         (avr_asm_init_sections): Same.
25742 2017-07-05  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
25744         * config/arm/neon.md (fma<VCVTF:mode>4): Remove %?.
25745         (fma<VH:mode>4_intrinsic): Likewise.
25746         (*fmsub<VCVTF:mode>4): Likewise.
25747         (*fmsub<VH:mode>4_intrinsic): Likewise.
25749 2017-07-05  Georg-Johann Lay  <avr@gjlay.de>
25751         PR target/81305
25752         * config/avr/avr.c (avr_out_movhi_mr_r_xmega) [CONSTANT_ADDRESS_P]:
25753         Don't depend on "optimize > 0".
25754         (out_movhi_r_mr, out_movqi_mr_r): Same.
25755         (out_movhi_mr_r, out_movqi_r_mr): Same.
25756         (avr_address_cost) [CONSTANT_ADDRESS_P]: Don't depend cost for
25757         io_address_operand on "optimize > 0".
25759 2017-07-05  Bin Cheng  <bin.cheng@arm.com>
25761         * tree-loop-distribution.c: Add general explanantion on the pass.
25762         (generate_loops_for_partition): Mark distributed loop.
25763         (pg_add_dependence_edges): New parameter.  Handle alias data
25764         dependence specially and record it in the parameter if asked.
25765         (struct pg_vdata, pg_edata, pg_edge_callback_data): New structs.
25766         (init_partition_graph_vertices, add_partition_graph_edge): New.
25767         (pg_skip_alias_edge, free_partition_graph_edata_cb): New.
25768         (free_partition_graph_vdata, build_partition_graph): New.
25769         (sort_partitions_by_post_order, merge_dep_scc_partitions): New.
25770         (pg_collect_alias_ddrs, break_alias_scc_partitions): New.
25771         (data_ref_segment_size, latch_dominated_by_data_ref): New.
25772         (compute_alias_check_pairs, version_loop_by_alias_check): New.
25773         (version_for_distribution_p, finalize_partitions): New.
25774         (distribute_loop): Handle alias data dependence specially.  Factor
25775         out loop fusion code as functions and call these functions.
25777 2017-07-05  Bin Cheng  <bin.cheng@arm.com>
25779         * tree-loop-distribution.c (classify_partition): New parameter and
25780         better handle reduction statement.
25781         (rdg_build_partitions): Revise comment.
25782         (distribute_loop): Compute statements in all partitions and pass it
25783         to classify_partition.
25785 2017-07-05  Bin Cheng  <bin.cheng@arm.com>
25787         * tree-loop-distribution.c (enum partition_type): New.
25788         (struct partition): New field type.
25789         (partition_merge_into): Add parameter.  Update partition type.
25790         (data_dep_in_cycle_p, update_type_for_merge): New functions.
25791         (build_rdg_partition_for_vertex): Compute partition type.
25792         (rdg_build_partitions): Dump partition type.
25793         (distribute_loop): Update calls to partition_merge_into.
25795 2017-07-05  Bin Cheng  <bin.cheng@arm.com>
25797         * tree-loop-distribution.c (struct ddr_hasher): New.
25798         (ddr_hasher::hash, ::equal, get_data_dependence): New function.
25799         (ddrs_table): New.
25800         (classify_partition): Call get_data_dependence.
25801         (pg_add_dependence_edges): Ditto.
25802         (distribute_loop): Release data dependence hash table.
25804 2017-07-05  Bin Cheng  <bin.cheng@arm.com>
25806         * tree-loop-distribution.c (ref_base_address): Delete.
25807         (similar_memory_accesses): Rename ...
25808         (share_memory_accesses): ... to this.  Check if partitions access
25809         the same memory reference.
25810         (distribute_loop): Call share_memory_accesses.
25812 2017-07-05  Bin Cheng  <bin.cheng@arm.com>
25814         * tree-loop-distribution.c (struct partition): New field recording
25815         its data reference.
25816         (partition_alloc, partition_free): Init and release data refs.
25817         (partition_merge_into): Merge data refs.
25818         (build_rdg_partition_for_vertex): Collect data refs for partition.
25819         (pg_add_dependence_edges): Change parameters from vector to bitmap.
25820         Update uses.
25821         (distribute_loop): Remve data refs from vertice data of partition
25822         graph.
25824 2017-07-05  Bin Cheng  <bin.cheng@arm.com>
25826         * tree-loop-distribution.c (params.h): Include header file.
25827         (MAX_DATAREFS_NUM, DR_INDEX): New macro.
25828         (datarefs_vec): New global var.
25829         (create_rdg_vertices): Use datarefs_vec directly.
25830         (free_rdg): Don't free data references.
25831         (build_rdg): Update use.  Don't free data references.
25832         (distribute_loop): Compute global variable for data references.
25833         Bail out if there are too many data references.
25835 2017-07-05  Bin Cheng  <bin.cheng@arm.com>
25837         * tree-loop-distribution.c (loop_nest): New global var.
25838         (build_rdg): Use loop directly, rather than loop nest.
25839         (pg_add_dependence_edges): Remove loop nest parameter.  Use global
25840         variable directly.
25841         (distribute_loop): Compute global variable loop nest.  Update use.
25843 2017-07-05  Bin Cheng  <bin.cheng@arm.com>
25845         * tree-loop-distribution.c (enum fuse_type, fuse_message): New.
25846         (partition_merge_into): New parameter.  Dump reason for fusion.
25847         (distribute_loop): Update use of partition_merge_into.
25849 2017-07-05  Bin Cheng  <bin.cheng@arm.com>
25851         * tree-loop-distribution.c (bb_top_order_index): New.
25852         (bb_top_order_index_size, bb_top_order_cmp): New.
25853         (stmts_from_loop): Use topological order.
25854         (pass_loop_distribution::execute): Compute and release topological
25855         order for basic blocks.
25857 2017-07-05  Bin Cheng  <bin.cheng@arm.com>
25859         * tree-loop-distribution.c (pass_loop_distribution::execute): Skip
25860         if no loops.
25862 2017-07-05  Bin Cheng  <bin.cheng@arm.com>
25864         * cfgloop.h (struct loop): Add comment.  New field orig_loop_num.
25865         * cfgloopmanip.c (lv_adjust_loop_entry_edge): Comment change.
25866         * internal-fn.c (expand_LOOP_DIST_ALIAS): New function.
25867         * internal-fn.def (LOOP_DIST_ALIAS): New.
25868         * tree-vectorizer.c (fold_loop_vectorized_call): Rename to ...
25869         (fold_loop_internal_call): ... this.
25870         (vect_loop_dist_alias_call): New function.
25871         (set_uid_loop_bbs): Call fold_loop_internal_call.
25872         (vectorize_loops): Fold IFN_LOOP_VECTORIZED and IFN_LOOP_DIST_ALIAS
25873         internal calls.
25875 2017-07-04  Uros Bizjak  <ubizjak@gmail.com>
25877         PR target/81300
25878         * config/i386/i386.md (setcc + movzbl/and to xor + setcc peepholes):
25879         Require dead FLAGS_REG at the beginning of a peephole.
25881 2017-07-04  Uros Bizjak  <ubizjak@gmail.com>
25883         PR target/81294
25884         * config/i386/adxintrin.h (_subborrow_u32): Swap _X and _Y
25885         arguments in the call to __builtin_ia32_sbb_u32.
25886         (_subborrow_u64): Swap _X and _Y arguments in the call to
25887         __builtin_ia32_sbb_u64.
25889 2017-07-04  Jakub Jelinek  <jakub@redhat.com>
25891         PR debug/81278
25892         * tree-vrp.c (compare_assert_loc): Turn into a function template
25893         with stable template parameter.  Only test if a->e is NULL,
25894         !a->e == !b->e has been verified already.  Use e == NULL or
25895         e != NULL instead of e or ! e tests.  If stable is true, don't use
25896         iterative_hash_expr, on the other side allow a or b or both NULL
25897         and sort the NULLs last.
25898         (process_assert_insertions): Sort using compare_assert_loc<false>
25899         instead of compare_assert_loc, later sort using
25900         compare_assert_loc<true> before calling process_assert_insertions_for
25901         in a loop.  Use break instead of continue once seen NULL pointer.
25903 2017-07-04  Thomas Preud'homme  <thomas.preudhomme@arm.com>
25905         * config/arm/driver-arm.c (arm_cpu_table): Add entry for ARM
25906         Cortex-R7 and Cortex-R8 processors.
25908 2017-07-04  Jan Hubicka  <hubicka@ucw.cz>
25910         * ipa-utils.c (ipa_merge_profiles): Fix merging when dst is
25911         uninitialized while src is not.
25913 2017-07-04  Richard Earnshaw  <rearnsha@arm.com>
25915         * common/config/arm/arm-common.c: Adjust include path for
25916         arm-cpu-cdata.h
25917         * config/arm/t-arm (TM_H): Adjust path for arm-cpu.h.
25918         (arm-cpu.h): Create in build directory.  Adjust dependency rules.
25919         (arm-cpu-data.h): Likewise.
25920         (arm-cpu-cdata.h): Likewise.
25921         * config/arm/arm-cpu.h: Delete.
25922         * config/arm/arm-cpu-cdata.h: Delete.
25923         * config/arm/arm-cpu-data.h: Delete.
25925 2017-07-04  James Greenhalgh  <james.greenhalgh@arm.com>
25927         * config/arm/arm-cpus.in (cortex-a55): New.
25928         (cortex-a75): Likewise.
25929         (cortex-a75.cortex-a55): Likewise.
25930         * config/arm/driver-arm.c (arm_cpu_table): Add cortex-a55 and
25931         cortex-a75.
25932         * doc/invoke.texi (-mcpu): Document cortex-a55 and cortex-a75.
25933         * config/arm/arm-cpu-cdata.h: Regenerate.
25934         * config/arm/arm-cpu-data.h: Regenerate.
25935         * config/arm/arm-cpu.h: Regenerate.
25936         * config/arm/arm-tables.opt: Regenerate.
25937         * config/arm/arm-tune.md: Regenerate.
25939 2017-07-04  Jan Hubicka  <hubicka@ucw.cz>
25941         * haifa-sched.c (sched_create_recovery_edges): Update profile.
25943 2017-07-04  Jan Hubicka  <hubicka@ucw.cz>
25945         * bb-reorder.c (better_edge_p): Fix handling of uninitialized
25946         probability.
25948 2017-07-04  Richard Sandiford  <richard.sandiford@linaro.org>
25950         PR tree-optimization/81292
25951         * tree-ssa-strlen.c (handle_builtin_strlen): When setting
25952         full_string_p, also call adjust_related_strinfos if the adjustment
25953         is simple, otherwise invalidate related strinfos.
25955 2017-07-04  Martin Liska  <mliska@suse.cz>
25957         PR sanitizer/81040
25958         * sanopt.c (sanitize_rewrite_addressable_params): Mark the
25959         newly created variable as DECL_IGNORED_P.
25961 2017-07-04  Martin Liska  <mliska@suse.cz>
25963         PR ipa/81293
25964         * ipa-inline.c (inline_small_functions):
25965         Use xstrdup_for_dump.
25967 2017-07-04  Tom de Vries  <tom@codesourcery.com>
25969         * graph.c (draw_cfg_edges): Save and restore EDGE_DFS_BACK.
25971 2017-07-03  Dominique d'Humieres  <dominiq@lps.ens.fr>
25973         PR target/81033
25974         * config/darwin.c (darwin_function_switched_text_sections):
25975         Fix spaces.
25977 2017-07-03  Jan Hubicka  <hubicka@ucw.cz>
25979         * tree-vect-loop-manip.c (vect_do_peeling): Fix scaling up.
25981 2017-07-03  Richard Earnshaw  <rearnsha@arm.com>
25983         * doc/invoke.texi (ARM Options): Add -mbe8 and -mbe32 to option summary.
25985 2017-07-03  Richard Sandiford  <richard.sandiford@linaro.org>
25987         * tree-vect-loop.c (vect_analyze_loop_2): Treat min_scalar_loop_bound,
25988         min_profitable_iters, and th as inclusive lower bounds.
25989         Fix LOOP_VINFO_PEELING_FOR_GAPS condition.
25990         (vect_estimate_min_profitable_iters): Return inclusive lower bounds
25991         for min_profitable_iters and min_profitable_estimate.
25992         (vect_transform_loop): Treat th as an inclusive lower bound.
25993         * tree-vect-loop-manip.c (vect_loop_versioning): Likewise.
25995 2017-07-03  Dominique d'Humieres  <dominiq@lps.ens.fr>
25997         PR target/81033
25998         * config/darwin.c (darwin_function_switched_text_sections):
25999         Replace DECL_NAME with DECL_ASSEMBLER_NAME, split assemble_name_raw
26000         in two pieces, and suppress the use of buf.
26002 2017-07-03  Nathan Sidwell  <nathan@acm.org>
26004         * hash-table.h (hash_table_mod1): Fix indentation.
26006 2017-07-02  Jan Hubicka  <hubicka@ucw.cz>
26008         PR middle-end/81290
26009         * predict.c (force_edge_cold): Be more careful about propagation
26010         backward.
26011         * profile-count.h (profile_probability::guessed,
26012         profile_probability::fdo, profile_count::guessed, profile_count::fdo):
26013         New.
26014         * tree-ssa-threadupdate.c (recompute_probabilities): Result is guessed.
26016 2017-07-03  James Greenhalgh  <james.greenhalgh@arm.com>
26018         * doc/invoke.texi (rcpc architecture extension): Document it.
26020 2017-07-03  Richard Biener  <rguenther@suse.de>
26022         PR tree-optimization/60510
26023         * tree-vect-loop.c (vect_create_epilog_for_reduction): Pass in
26024         the scalar reduction PHI and use it.
26025         (vectorizable_reduction): Properly guard the single_defuse_cycle
26026         path for non-SLP reduction chains where we cannot use it.
26027         Rework reduc_def/index and vector type deduction.  Rework
26028         vector operand gathering during reduction op code-gen.
26029         * tree-vect-slp.c (vect_analyze_slp): For failed SLP reduction
26030         chains dissolve the chain and leave it to non-SLP reduction
26031         handling.
26033 2017-07-03  Richard Sandiford  <richard.sandiford@linaro.org>
26035         * tree-data-ref.h (dr_alignment): Declare.
26036         * tree-data-ref.c (dr_alignment): New function.
26037         * tree-vectorizer.h (dataref_aux): Remove base_element_aligned.
26038         * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
26039         set it.
26040         * tree-vect-stmts.c (vectorizable_store): Use dr_alignment.
26042 2017-07-03  Richard Sandiford  <richard.sandiford@linaro.org>
26044         * tree-data-ref.h (innermost_loop_behavior): Add base_alignment
26045         and base_misalignment fields.
26046         (DR_BASE_ALIGNMENT, DR_BASE_MISALIGNMENT): New macros.
26047         * tree-data-ref.c: Include builtins.h.
26048         (dr_analyze_innermost): Set up the new innmost_loop_behavior fields.
26049         * tree-vectorizer.h (STMT_VINFO_DR_BASE_ALIGNMENT): New macro.
26050         (STMT_VINFO_DR_BASE_MISALIGNMENT): Likewise.
26051         * tree-vect-data-refs.c: Include tree-cfg.h.
26052         (vect_compute_data_ref_alignment): Use the new innermost_loop_behavior
26053         fields instead of calculating an alignment here.
26054         (vect_analyze_data_refs): Use dr_analyze_innermost.  Dump the new
26055         innermost_loop_behavior fields.
26057 2017-07-03  Richard Sandiford  <richard.sandiford@linaro.org>
26059         * tree-data-ref.h (innermost_loop_behavior): Add a step_alignment
26060         field.
26061         (DR_STEP_ALIGNMENT): New macro.
26062         * tree-vectorizer.h (STMT_VINFO_DR_STEP_ALIGNMENT): Likewise.
26063         * tree-data-ref.c (dr_analyze_innermost): Initalize step_alignment.
26064         (create_data_ref): Print it.
26065         * tree-vect-stmts.c (vectorizable_load): Use the step alignment
26066         to tell whether the step preserves vector (mis)alignment.
26067         * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Likewise.
26068         Move the check for an integer step and generalise to all INTEGER_CST.
26069         (vect_analyze_data_refs): Set DR_STEP_ALIGNMENT when setting DR_STEP.
26070         Print the outer step alignment.
26072 2017-07-03  Richard Sandiford  <richard.sandiford@linaro.org>
26074         * tree-data-ref.h (innermost_loop_behavior): Replace aligned_to
26075         with offset_alignment.
26076         (DR_ALIGNED_TO): Delete.
26077         (DR_OFFSET_ALIGNMENT): New macro.
26078         * tree-vectorizer.h (STMT_VINFO_DR_ALIGNED_TO): Delete.
26079         (STMT_VINFO_DR_OFFSET_ALIGNMENT): New macro.
26080         * tree-data-ref.c (dr_analyze_innermost): Update after above changes.
26081         (create_data_ref): Likewise.
26082         * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Likewise.
26083         (vect_analyze_data_refs): Likewise.
26084         * tree-if-conv.c (if_convertible_loop_p_1): Use memset before
26085         creating dummy innermost behavior.
26087 2017-07-03  Richard Sandiford  <richard.sandiford@linaro.org>
26089         * tree-data-ref.h (dr_analyze_innermost): Replace the dr argument
26090         with a "innermost_loop_behavior *" and refeence tree.
26091         * tree-data-ref.c (dr_analyze_innermost): Likewise.
26092         (create_data_ref): Update call accordingly.
26093         * tree-predcom.c (find_looparound_phi): Likewise.
26095 2017-07-03  Richard Sandiford  <richard.sandiford@linaro.org>
26097         * tree-vectorizer.h (_stmt_vec_info): Replace individual dr_*
26098         fields with dr_wrt_vec_loop.
26099         (STMT_VINFO_DR_BASE_ADDRESS, STMT_VINFO_DR_INIT, STMT_VINFO_DR_OFFSET)
26100         (STMT_VINFO_DR_STEP, STMT_VINFO_DR_ALIGNED_TO): Update accordingly.
26101         (STMT_VINFO_DR_WRT_VEC_LOOP): New macro.
26102         (vect_dr_behavior): New function.
26103         (vect_create_addr_base_for_vector_ref): Remove loop parameter.
26104         * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
26105         vect_dr_behavior.  Use a step_preserves_misalignment_p boolean to
26106         track whether the step preserves the misalignment.
26107         (vect_create_addr_base_for_vector_ref): Remove loop parameter.
26108         Use vect_dr_behavior.
26109         (vect_setup_realignment): Update call accordingly.
26110         (vect_create_data_ref_ptr): Likewise.  Use vect_dr_behavior.
26111         * tree-vect-loop-manip.c (vect_gen_prolog_loop_niters): Update
26112         call to vect_create_addr_base_for_vector_ref.
26113         (vect_create_cond_for_align_checks): Likewise.
26114         * tree-vect-patterns.c (vect_recog_bool_pattern): Copy
26115         STMT_VINFO_DR_WRT_VEC_LOOP as a block.
26116         (vect_recog_mask_conversion_pattern): Likewise.
26117         * tree-vect-stmts.c (compare_step_with_zero): Use vect_dr_behavior.
26118         (new_stmt_vec_info): Remove redundant zeroing.
26120 2017-07-03  Richard Earnshaw  <rearnsha@arm.com>
26122         * common/config/arm/arm-common.c (arm_be8_option): New function.
26123         * config/arm/arm-isa.h (isa_feature): Add new feature bit isa_bit_be8.
26124         (ISA_ARMv6): Add isa_bit_be8.
26125         * config/arm/arm.h (arm_be8_option): Add prototype.
26126         (BE8_SPEC_FUNCTION): New define.
26127         (EXTRA_SPEC_FUNCTIONS): Add BE8_SPEC_FUNCTION.
26128         * config/arm/arm.opt (mbig-endian): Mark as Negative of mlittle-endian.
26129         (mlittle-endian): Similarly.
26130         (mbe8, mbe32): New options.
26131         * config/arm/bpabi.h (BE8_LINK_SPEC): Call arm_be8_option.
26132         * doc/invoke.texi (ARM Options): Document -mbe8 and -mbe32.
26134 2017-07-02  Jan Hubicka  <hubicka@ucw.cz>
26136         * tree-cfg.c (gimple_find_sub_bbs): Fix profile updating.
26138 2017-07-02  Jan Hubicka  <hubicka@ucw.cz>
26140         * tree-cfgcleanup.c (want_merge_blocks_p): New function.
26141         (cleanup_tree_cfg_bb): Use it.
26142         * profile-count.h (profile_count::of_for_merging, profile_count::merge):
26143         New functions.
26144         * tree-cfg.c (gimple_merge_blocks): Use profile_count::merge.
26146 2017-07-02  Jan Hubicka  <hubicka@ucw.cz>
26148         PR bootstrap/81285
26149         * loop-doloop.c (add_test): Update profile.
26151 2017-07-03  Martin Liska  <mliska@suse.cz>
26153         PR sanitize/81040
26154         * sanopt.c (rewrite_usage_of_param): New function.
26155         (sanitize_rewrite_addressable_params): Likewise.
26156         (pass_sanopt::execute): Call rewrite_usage_of_param.
26158 2017-07-03  Richard Biener  <rguenther@suse.de>
26160         * tree-vect-loop.c (vect_create_epilog_for_reduction): Revert
26161         back to using VIEW_CONVERT_EXPR.
26163 2017-07-03  Martin Liska  <mliska@suse.cz>
26165         PR other/78366
26166         * doc/extend.texi: Document when a resolver function is
26167         generated for target_clones.
26169 2017-07-03  Martin Liska  <mliska@suse.cz>
26171         * asan.c (asan_emit_stack_protection): Unpoison just red zones
26172         and shadow memory of auto variables which are subject of
26173         use-after-scope sanitization.
26174         (asan_expand_mark_ifn): Add do set only when is_poison.
26176 2016-07-03  Richard Biener  <rguenther@suse.de>
26178         * tree-vect-loop.c (vect_analyze_loop_operations): Also analyze
26179         reduction PHIs.
26180         (vect_force_simple_reduction): Record reduction def -> phi mapping.
26181         (vectorizable_reduction): Perform reduction PHI creation when
26182         visiting a reduction PHI and adjust and simplify code generation
26183         phase of the reduction op.  Cache dts, use fold_binary, not fold_build2.
26184         (vect_transform_loop): Visit reduction PHIs.
26185         * tree-vect-slp.c (vect_get_and_check_slp_defs): Record reduction
26186         defs into the SLP tree.
26187         (vect_build_slp_tree): Reduction defs terminate the recursion.
26188         * tree-vect-stmts.c (vect_get_vec_def_for_operand_1): Allow lookup
26189         of reduction defs.
26190         (vect_get_vec_defs_for_stmt_copy): Export.
26191         (vect_get_vec_defs): Likewise.
26192         * tree-vectorizer.h (struct _stmt_vec_info): Amend reduc_def
26193         purpose.
26194         (vect_get_vec_defs_for_stmt_copy): Declare.
26195         (vect_get_vec_defs): Likewise.
26197 2017-07-03  Richard Sandiford  <richard.sandiford@linaro.org>
26199         * tree-data-ref.c (dr_analyze_innermost): Replace the "nest"
26200         parameter with a "loop" parameter and use it instead of the
26201         loop containing DR_STMT.  Don't check simple_iv when doing
26202         BB analysis.  Describe the two analysis modes in the comment.
26204 2017-07-03  Tom de Vries  <tom@codesourcery.com>
26206         PR tree-optimization/69468
26207         * tree-ssa-tail-merge.c (ignore_edge_flags): New constant.
26208         (find_same_succ_bb): Handle ignore_edge_flags.
26210 2017-07-03  Tom de Vries  <tom@codesourcery.com>
26212         PR tree-optimization/81192
26213         * tree-ssa-tail-merge.c (same_succ_hash): Use bb->loop_father->num in
26214         hash.
26215         (same_succ::equal): Don't find bbs to be equal if bb->loop_father
26216         differs.
26217         (find_same_succ_bb): Remove obsolete test on bb->loop_father->latch.
26219 2017-07-03  Tom de Vries  <tom@codesourcery.com>
26221         PR tree-optimization/81192
26222         * tree-ssa-tail-merge.c (same_succ_flush_bb): Handle
26223         BB_SAME_SUCC (bb) == NULL.
26225 2017-07-02  Jan Hubicka  <hubicka@ucw.cz>
26227         * cfgrtl.c (rtl_verify_edges): Enable checking of profile_probability
26228         consistency.
26230 2017-07-02  Jan Hubicka  <hubicka@ucw.cz>
26232         * dumpfile.c: Include profile-count.h
26233         * tree-cfg.c (gimple_duplicate_sese_tail): Drop UNUSED attributes;
26234         update profile.
26235         (insert_cond_bb): Update profile.
26236         * tree-cfg.h (insert_cond_bb): Update prototype.
26237         * tree-chkp-opt.c (chkp_optimize_string_function_calls): Update.
26238         * tree-dump.c: Do not include tree-cfg.
26240 2017-07-02  Jan Hubicka  <hubicka@ucw.cz>
26242         * bb-reorder.c (fix_up_crossing_landing_pad): Update profile.
26244 2017-07-02  Jan Hubicka  <hubicka@ucw.cz>
26246         * expect.c (dw2_build_landing_pads): Update profile of the landing pad
26247         bb.
26249 2017-07-02  Jan Hubicka  <hubicka@ucw.cz>
26251         * tree-complex.c (expand_complex_div_wide): update profile.
26253 2017-07-02  Richard Sandiford  <richard.sandiford@linaro.org>
26254             Alan Hayward  <alan.hayward@arm.com>
26255             David Sherwood  <david.sherwood@arm.com>
26257         * Makefile.in (MACHMODE_H): Remove insn-modes.h
26258         (CORETYPES_H): New define.
26259         (MOSTLYCLEANFILES): Add insn-modes-inline.h.
26260         (insn-modes-inline.h, s-modes-inline-h): New rules.
26261         (generated_files): Add insn-modes-inline.h.
26262         (RTL_BASE_H, TREE_CORE_H): Use CORETYPES_H instead of coretypes.h.
26263         (build/gensupport.o, build/ggc-none.o, build/print-rtl.o): Likewise.
26264         (build/read-md.o, build/read-rtl.o, build/rtl.o): Likewise.
26265         (build/vec.o, build/hash-table.o, build/inchash.o): Likewise.
26266         (build/gencondmd.o, build/genattr.o, build/genattr-common.o): Likewise.
26267         (build/genattrtab.o, build/genautomata.o, build/gencheck.o): Likewise.
26268         (build/gencodes.o, build/genconditions.o): Likewise.
26269         (build/genconfig.o, build/genconstants.o, build/genemit.o): Likewise.
26270         (build/genenums.o, build/genextract.o, build/genflags.o): Likewise.
26271         (build/gentarget-def.o, build/genmddeps.o, build/genopinit.o)
26272         (build/genoutput.o, build/genpeep.o, build/genpreds.o): Likewise.
26273         (build/genrecog.o, build/genmddump.o, build/genmatch.o): Likewise.
26274         (build/gencfn-macros.o, build/gcov-iov.o): Likewise.
26275         * coretypes.h: Include everything up to real.h for generators.
26276         Include insn-modes.h first.  Include wide-int-print.h after
26277         wide-int.h.  Include insn-modes-inline.h and then machmode.h.
26278         * machmode.h: Don't include insn-modes.h here.
26279         * function-tests.c: Remove includes of signop.h, machmode.h,
26280         double-int.h and wide-int.h.
26281         * rtl.h: Likewise.
26282         * gcc-rich-location.c: Remove includes of machmode.h, double-int.h
26283         and wide-int.h.
26284         * optc-save-gen.awk: Likewise.
26285         * gencheck.c (BITS_PER_UNIT): Delete dummy definition.
26286         * godump.c: Remove include of wide-int-print.h.
26287         * pretty-print.h: Likewise.
26288         * wide-int-print.cc: Likewise.
26289         * wide-int.cc: Likewise.
26290         * hash-map-tests.c: Remove include of signop.h.
26291         * hash-set-tests.c: Likewise.
26292         * rtl-tests.c: Likewise.
26293         * mkconfig.sh: Remove include of machmode.h.
26294         * genmodes.c (emit_insn_modes_h): Split emission of inline functions
26295         into...
26296         (emit_insn_modes_inline_h): ...this new function.  Emit the code
26297         into an insn-modes-inline.h header file, adding appropriate
26298         include guards and end comments.
26299         (emit_insn_modes_c_header): Remove include of machmode.h.
26300         (emit_min_insn_modes_c_header): Include coretypes.h rather than
26301         machmode.h.
26302         (main): Handle -i flag and call emit_insn_modes_inline_h when
26303         it is passed.
26305 2017-07-02  Richard Sandiford  <richard.sandiford@linaro.org>
26307         * tree-ssa-strlen.c (strinfo): Rename the length field to
26308         nonzero_chars.  Add a full_string_p field.
26309         (compare_nonzero_chars, zero_length_string_p): New functions.
26310         (get_addr_stridx): Add an offset_out parameter.
26311         Use compare_nonzero_chars.
26312         (get_stridx): Update accordingly.  Use compare_nonzero_chars.
26313         (new_strinfo): Update after above changes to strinfo.
26314         (set_endptr_and_length): Set full_string_p.
26315         (get_string_length): Update after above changes to strinfo.
26316         (unshare_strinfo): Update call to new_strinfo.
26317         (maybe_invalidate): Likewise.
26318         (get_stridx_plus_constant): Change off to unsigned HOST_WIDE_INT.
26319         Use compare_nonzero_chars and zero_string_p.  Treat nonzero_chars
26320         as a uhwi instead of an shwi.  Update after above changes to
26321         strinfo and new_strinfo.
26322         (zero_length_string): Assert that chainsi contains full strings.
26323         Use zero_length_string_p.  Update call to new_strinfo.
26324         (adjust_related_strinfos): Update after above changes to strinfo.
26325         Copy full_string_p from origsi.
26326         (adjust_last_stmt): Use zero_length_string_p.
26327         (handle_builtin_strlen): Update after above changes to strinfo and
26328         new_strinfo.  Install the lhs as the string length if the previous
26329         entry didn't describe a full string.
26330         (handle_builtin_strchr): Update after above changes to strinfo
26331         and new_strinfo.
26332         (handle_builtin_strcpy): Likewise.
26333         (handle_builtin_strcat): Likewise.
26334         (handle_builtin_malloc): Likewise.
26335         (handle_pointer_plus): Likewise.
26336         (handle_builtin_memcpy): Likewise.  Track nonzero characters
26337         that aren't necessarily followed by a nul terminator.
26338         (handle_char_store): Likewise.
26340 2017-07-02  Richard Sandiford  <richard.sandiford@linaro.org>
26342         PR tree-optimization/80769
26343         * tree-ssa-strlen.c (strinfo): Document that "stmt" is also used
26344         for malloc and calloc.  Document the new invariant that all related
26345         strinfos have delayed lengths or none do.
26346         (verify_related_strinfos): Move earlier in file.
26347         (set_endptr_and_length): New function, split out from...
26348         (get_string_length): ...here.  Also set the lengths of related
26349         strinfos.
26350         (zero_length_string): Assert that chainsi has known (rather than
26351         delayed) lengths.
26352         (adjust_related_strinfos): Likewise.
26354 2017-07-02  Richard Sandiford  <richard.sandiford@linaro.org>
26356         PR tree-optimization/81136
26357         * tree-vect-data-refs.c (vect_update_misalignment_for_peel): Only
26358         assert that two references with the same misalignment have the same
26359         compile-time misalignment if those compile-time misalignments
26360         are known.
26362 2017-07-01  Andi Kleen  <ak@linux.intel.com>
26364         * print-tree.c (print_node): Print all attributes.
26366 2017-07-01  Jan Hubicka  <hubicka@ucw.cz>
26368         * cfg.c (scale_bbs_frequencies): New function.
26369         * cfg.h (scale_bbs_frequencies): Declare it.
26370         * cfgloopanal.c (single_likely_exit): Cleanup.
26371         * cfgloopmanip.c (scale_loop_frequencies): Take profile_probability
26372         as parameter.
26373         (scale_loop_profile): Likewise.
26374         (loop_version): Likewise.
26375         (create_empty_loop_on_edge): Update.
26376         * cfgloopmanip.h (scale_loop_frequencies, scale_loop_profile,
26377         scale_loop_frequencies, scale_loop_profile, loopify,
26378         loop_version): Update prototypes.
26379         * modulo-sched.c (sms_schedule): Update.
26380         * predict.c (unlikely_executed_edge_p): Also check probability.
26381         (probably_never_executed_edge_p): Fix typo.
26382         * tree-if-conv.c (version_loop_for_if_conversion): Update.
26383         * tree-parloops.c (gen_parallel_loop): Update.
26384         * tree-ssa-loop-ivcanon.c (try_peel_loop): Update.
26385         * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Update.
26386         * tree-ssa-loop-split.c (split_loop): Update.
26387         * tree-ssa-loop-unswitch.c (tree_unswitch_loop): Update.
26388         * tree-vect-loop-manip.c (vect_do_peeling): Update.
26389         (vect_loop_versioning): Update.
26390         * tree-vect-loop.c (scale_profile_for_vect_loop): Update.
26392 2017-07-01  Jan Hubicka  <hubicka@ucw.cz>
26394         * trans-mem.c (split_bb_make_tm_edge): Update profile.
26396 2017-07-01  Jan Hubicka  <hubicka@ucw.cz>
26398         * tree-if-conv.c (combine_blocks): Use make_single_succ_edge
26399         to keep profile consistent.
26401 2017-07-01  Jan Hubicka  <hubicka@ucw.cz>
26403         * cfgrtl.c (rtl_flow_call_edges_add): Update profile.
26404         * tree-cfg.c (gimple_flow_call_edges_add): Likewise.
26405         * profile-count.h (max_safe_multiplier): Make unsigned.
26406         (profile_count::guessed_zero): New.
26408 2017-07-01  Jan Hubicka  <hubicka@ucw.cz>
26410         * bb-reorder.c (fix_up_crossing_landing_pad,
26411         fix_crossing_conditional_branches): Use make_single_succ_edge
26412         to keep profile consistent.
26414 2017-07-01  Jan Hubicka  <hubicka@ucw.cz>
26416         * tree-vect-loop.c (optimize_mask_stores): Use make_single_succ_edge
26417         to update profile.
26419 2017-07-01  Jakub Jelinek  <jakub@redhat.com>
26421         PR sanitizer/81262
26422         * bb-reorder.c (fix_up_fall_thru_edges): Move variable declarations to
26423         the right scopes, make sure cond_jump isn't preserved between multiple
26424         iterations.  Search for fallthru edge whenever there are 3+ edges and
26425         use find_fallthru_edge for it.
26427 2017-06-29  Jan Hubicka  <hubicka@ucw.cz>
26429         Patch by Alexander Monakov <amonakov@ispras.ru>
26430         * sel-sched-ir.c (compute_succs_info): Handle uninitialized
26431         probabilities consistently.
26433 2017-06-29  Jan Hubicka  <hubicka@ucw.cz>
26435         * pa.c (pa_expand_compare_and_swap_loop): Update call of
26436         emit_cmp_and_jump_insns.
26438 2017-06-29  Jan Hubicka  <hubicka@ucw.cz>
26440         PR ipa/81261
26441         * tree-inline.c (expand_call_inline): Combine profile statuses.
26443 2017-06-30  Andrew Pinski  <apinski@cavium.com>
26445         * tree-if-conv.c (predicate_scalar_phi): Update new_stmt if
26446         fold_stmt returned true.
26448 2017-06-30  Nathan Sidwell  <nathan@acm.org>
26450         * ggc.h (empty_string): Delete.
26451         * cfgexpand.c (expand_asm_stmt): Use plain "".
26452         * optabs.c (expand_asm_memory_barrier): Likewise.
26453         * stringpool.c (empty_string): Delete.
26454         (digit_vector, digit_string): Delete.
26455         (ggc_alloc_string): Use plain "", don't optimize single digit
26456         strings.  Use ggc_alloc_atomic.
26458 2017-06-30  Richard Earnshaw  <rearnsha@arm.com>
26460         * rtlanal.c (insn_rtx_cost): If a parallel contains exactly one
26461         comparison set and one other set, use the cost of the non-comparison
26462         set.
26464 2017-06-30  Nathan Sidwell  <nathan@acm.org>
26466         * ggc.h: Replace all 'static inline' with plain 'inline'.  Fix
26467         some formatting.
26469 2017-06-30  Peter Bergner  <bergner@vnet.ibm.com>
26471         * tree-cfg.c (group_case_labels_stmt): Merge scanning and compressing
26472         loops.  Remove now unneeded calls to gimple_switch_set_label() that
26473         just set removed labels to NULL_TREE.
26475 2017-06-30  Aldy Hernandez  <aldyh@redhat.com>
26477         * tree-ssanames.c (set_range_info_raw): Abstract from ...
26478         (set_range_info): ...here.  Only call set_range_info_raw if domain
26479         is useful.
26480         (set_nonzero_bits): Call set_range_info_raw.
26481         * tree-ssanames.h (set_range_info_raw): New.
26483 2017-06-30  Jakub Jelinek  <jakub@redhat.com>
26485         PR target/81225
26486         * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): For V8FI,
26487         V16FI and VI8F_256 iterators, use <store_mask_predicate> instead
26488         of nonimmediate_operand and <store_mask_constraint> instead of m
26489         for the input operand.  For V8FI iterator, always split if input
26490         is a MEM.  For V16FI and V8SF_256 iterators, don't test if both
26491         operands are MEM if <mask_applied>.  For VI4F_256 iterator, use
26492         <store_mask_predicate> instead of register_operand and
26493         <store_mask_constraint> instead of v for the input operand.  Make
26494         sure both operands aren't MEMs for if not <mask_applied>.
26496 2017-06-30  Sylvestre Ledru  <sylvestre@debian.org>
26498         * lto-wrapper.c (copy_file) Close both file descriptors before
26499         exiting normally.
26501 2017-06-30  Martin Liska  <mliska@suse.cz>
26503         PR ipa/81214
26504         * multiple_target.c (create_dispatcher_calls): Make ifunc
26505         also for function that don't have calls or are not referenced.
26507 2017-06-30  Richard Biener  <rguenther@suse.de>
26509         * tree-vect-slp.c (vect_slp_analyze_node_operations): Only
26510         analyze the first scalar stmt.  Move vector type computation
26511         for the BB case here from ...
26512         * tree-vect-stmts.c (vect_analyze_stmt): ... here.  Guard
26513         live operation processing in the SLP case properly.
26515 2017-06-30  Richard Biener  <rguenther@suse.de>
26517         * graph.c (draw_cfg_node_succ_edges): Fix broken dot syntax.
26519 2017-06-30  Martin Liska  <mliska@suse.cz>
26521         PR sanitizer/81021
26522         * tree-eh.c (lower_resx): Call BUILT_IN_ASAN_HANDLE_NO_RETURN
26523         before BUILT_IN_UNWIND_RESUME when ASAN is used.
26525 2017-06-30  Yvan Roux  <yvan.roux@linaro.org>
26527         * doc/invoke.texi (AArch64): Add missing options and remove redundant
26528         ones.
26530 2017-06-30  Richard Biener  <rguenther@suse.de>
26532         PR tree-optimization/81249
26533         * tree-vect-loop.c (vect_create_epilog_for_reduction): Convert
26534         condition reduction result to original scalar type.
26536 2017-06-30  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
26538         * profile-count.h (enum profile_quality): Fix typos and whitespace
26539         issues.
26541 2017-06-30  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
26543         * config/s390/s390.c (s390_expand_setmem): Adjust to the new data
26544         type for branch probabilities.
26546 2017-06-29  Julian Brown  <julian@codesourcery.com>
26547             Naveen H.S  <Naveen.Hurugalawadi@cavium.com>
26549         * config/aarch64/aarch64-fusion-pairs.def: Add ALU_BRANCH entry.
26550         * config/aarch64/aarch64.c (AARCH64_FUSE_ALU_BRANCH): New fusion type.
26551         (thunderx2t99_tunings): Set AARCH64_FUSE_ALU_BRANCH flag.
26552         (aarch_macro_fusion_pair_p): Add support for AARCH64_FUSE_ALU_BRANCH.
26554 2017-06-29  Naveen H.S  <Naveen.Hurugalawadi@cavium.com>
26556         * config/aarch64/aarch64.c (aarch_macro_fusion_pair_p): Push the
26557         check for CC usage into AARCH64_FUSE_CMP_BRANCH.
26558         * config/i386/i386.c (ix86_macro_fusion_pair_p): Push the check for
26559         CC usage from generic code to here.
26560         * sched-deps.c (sched_macro_fuse_insns): Move the condition for
26561         CC usage into the target macros.
26563 2017-06-29  Maya Rashish  <coypu@sdf.org>
26565         * config/netbsd.h (NETBSD_LIB_SPEC): Add -lc when creating shared
26566         objects.
26568 2017-06-29  Jan Hubicka  <hubicka@ucw.cz>
26570         * arm/arm-builtins.c: Include profile-count.h
26571         * except.c (sjlj_emit_function_enter): Use
26572         profile_probability::unlikely.
26574 2017-06-29  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
26576         * config/rs6000/rs6000.c (toc_relative_expr_p): Make tocrel_base
26577         and tocrel_offset be pointer args rather than implicitly using
26578         static versions.
26579         (legitimate_constant_pool_address_p, rs6000_emit_move,
26580         const_load_sequence_p, adjust_vperm): Add local tocrel_base and
26581         tocrel_offset and use in toc_relative_expr_p call.
26582         (print_operand, print_operand_address): Use static tocrel_base_oac
26583         and tocrel_offset_oac.
26584         (rs6000_output_addr_const_extra): Use static tocrel_base_oac and
26585         tocrel_offset_oac.
26587 2017-06-29  Maya Rashish  <coypu@sdf.org>
26589         * config/vax/builtins.md (ffssi2_internal): Correct constraint.
26591 2017-06-29  Eric Botcazou  <ebotcazou@adacore.com>
26593         * expr.c (expand_expr) <normal_inner_ref>: When testing for unaligned
26594         objects, take into account only the alignment of 'op0' and 'mode1' if
26595         'op0' is a MEM.
26597 2017-06-29  Steve Ellcey  <sellcey@cavium.com>
26599         * ccmp.c (ccmp_tree_comparison_p): New function.
26600         (ccmp_candidate_p): Update to use above function.
26601         (get_compare_parts): New function.
26602         (expand_ccmp_next): Update to use new functions.
26603         (expand_ccmp_expr_1): Take tree arg instead of gimple, update to use
26604         new functions.
26605         (expand_ccmp_expr): Pass tree instead of gimple to expand_ccmp_expr_1,
26606         take mode as argument.
26607         * ccmp.h (expand_ccmp_expr): Add mode as argument.
26608         * expr.c (expand_expr_real_1): Pass mode as argument.
26610 2017-06-29  Segher Boessenkool  <segher@kernel.crashing.org>
26612         * combine.c (combine_instructions): Print insns to dump_file, together
26613         with their costs.
26615 2017-06-29  Jan Hubicka  <hubicka@ucw.cz>
26617         * asan.c (asan_emit_stack_protection): Update.
26618         (create_cond_insert_point): Update.
26619         * auto-profile.c (afdo_propagate_circuit): Update.
26620         * basic-block.h (struct edge_def): Turn probability to
26621         profile_probability.
26622         (EDGE_FREQUENCY): Update.
26623         * bb-reorder.c (find_traces_1_round): Update.
26624         (better_edge_p): Update.
26625         (sanitize_hot_paths): Update.
26626         * cfg.c (unchecked_make_edge): Initialize probability to uninitialized.
26627         (make_single_succ_edge): Update.
26628         (check_bb_profile): Update.
26629         (dump_edge_info): Update.
26630         (update_bb_profile_for_threading): Update.
26631         * cfganal.c (connect_infinite_loops_to_exit): Initialize new edge
26632         probabilitycount to 0.
26633         * cfgbuild.c (compute_outgoing_frequencies): Update.
26634         * cfgcleanup.c (try_forward_edges): Update.
26635         (outgoing_edges_match): Update.
26636         (try_crossjump_to_edge): Update.
26637         * cfgexpand.c (expand_gimple_cond): Update make_single_succ_edge.
26638         (expand_gimple_tailcall): Update.
26639         (construct_init_block): Use make_single_succ_edge.
26640         (construct_exit_block): Use make_single_succ_edge.
26641         * cfghooks.c (verify_flow_info): Update.
26642         (redirect_edge_succ_nodup): Update.
26643         (split_edge): Update.
26644         (account_profile_record): Update.
26645         * cfgloopanal.c (single_likely_exit): Update.
26646         * cfgloopmanip.c (scale_loop_profile): Update.
26647         (set_zero_probability): Remove.
26648         (duplicate_loop_to_header_edge): Update.
26649         * cfgloopmanip.h (loop_version): Update prototype.
26650         * cfgrtl.c (try_redirect_by_replacing_jump): Update.
26651         (force_nonfallthru_and_redirect): Update.
26652         (update_br_prob_note): Update.
26653         (rtl_verify_edges): Update.
26654         (purge_dead_edges): Update.
26655         (rtl_lv_add_condition_to_bb): Update.
26656         * cgraph.c: (cgraph_edge::redirect_call_stmt_to_calle): Update.
26657         * cgraphunit.c (init_lowered_empty_function): Update.
26658         (cgraph_node::expand_thunk): Update.
26659         * cilk-common.c: Include profile-count.h
26660         * dojump.c (inv): Remove.
26661         (jumpifnot): Update.
26662         (jumpifnot_1): Update.
26663         (do_jump_1): Update.
26664         (do_jump): Update.
26665         (do_jump_by_parts_greater_rtx): Update.
26666         (do_compare_rtx_and_jump): Update.
26667         * dojump.h (jumpifnot, jumpifnot_1, jumpif_1, jumpif, do_jump,
26668         do_jump_1. do_compare_rtx_and_jump): Update prototype.
26669         * dwarf2cfi.c: Include profile-count.h
26670         * except.c (dw2_build_landing_pads): Use make_single_succ_edge.
26671         (sjlj_emit_dispatch_table): Likewise.
26672         * explow.c: Include profile-count.h
26673         * expmed.c (emit_store_flag_force): Update.
26674         (do_cmp_and_jump): Update.
26675         * expr.c (compare_by_pieces_d::generate): Update.
26676         (compare_by_pieces_d::finish_mode): Update.
26677         (emit_block_move_via_loop): Update.
26678         (store_expr_with_bounds): Update.
26679         (store_constructor): Update.
26680         (expand_expr_real_2): Update.
26681         (expand_expr_real_1): Update.
26682         * expr.h (try_casesi, try_tablejump): Update prototypes.
26683         * gimple-pretty-print.c (dump_probability): Update.
26684         (dump_profile): New.
26685         (dump_gimple_label): Update.
26686         (dump_gimple_bb_header): Update.
26687         * graph.c (draw_cfg_node_succ_edges): Update.
26688         * hsa-gen.c (convert_switch_statements): Update.
26689         * ifcvt.c (cheap_bb_rtx_cost_p): Update.
26690         (find_if_case_1): Update.
26691         (find_if_case_2): Update.
26692         * internal-fn.c (expand_arith_overflow_result_store): Update.
26693         (expand_addsub_overflow): Update.
26694         (expand_neg_overflow): Update.
26695         (expand_mul_overflow): Update.
26696         (expand_vector_ubsan_overflow): Update.
26697         * ipa-cp.c (good_cloning_opportunity_p): Update.
26698         * ipa-split.c (split_function): Use make_single_succ_edge.
26699         * ipa-utils.c (ipa_merge_profiles): Update.
26700         * loop-doloop.c (add_test): Update.
26701         (doloop_modify): Update.
26702         * loop-unroll.c (compare_and_jump_seq): Update.
26703         (unroll_loop_runtime_iterations): Update.
26704         * lra-constraints.c (lra_inheritance): Update.
26705         * lto-streamer-in.c (input_cfg): Update.
26706         * lto-streamer-out.c (output_cfg): Update.
26707         * mcf.c (adjust_cfg_counts): Update.
26708         * modulo-sched.c (sms_schedule): Update.
26709         * omp-expand.c (expand_omp_for_init_counts): Update.
26710         (extract_omp_for_update_vars): Update.
26711         (expand_omp_ordered_sink): Update.
26712         (expand_omp_for_ordered_loops): Update.
26713         (expand_omp_for_generic): Update.
26714         (expand_omp_for_static_nochunk): Update.
26715         (expand_omp_for_static_chunk): Update.
26716         (expand_cilk_for): Update.
26717         (expand_omp_simd): Update.
26718         (expand_omp_taskloop_for_outer): Update.
26719         (expand_omp_taskloop_for_inner): Update.
26720         * omp-simd-clone.c (simd_clone_adjust): Update.
26721         * optabs.c (expand_doubleword_shift): Update.
26722         (expand_abs): Update.
26723         (emit_cmp_and_jump_insn_1): Update.
26724         (expand_compare_and_swap_loop): Update.
26725         * optabs.h (emit_cmp_and_jump_insns): Update prototype.
26726         * predict.c (predictable_edge_p): Update.
26727         (edge_probability_reliable_p): Update.
26728         (set_even_probabilities): Update.
26729         (combine_predictions_for_insn): Update.
26730         (combine_predictions_for_bb): Update.
26731         (propagate_freq): Update.
26732         (estimate_bb_frequencies): Update.
26733         (force_edge_cold): Update.
26734         * profile-count.c (profile_count::dump): Add missing space into dump.
26735         (profile_count::debug): Add newline.
26736         (profile_count::differs_from_p): Explicitly convert to unsigned.
26737         (profile_count::stream_in): Update.
26738         (profile_probability::dump): New member function.
26739         (profile_probability::debug): New member function.
26740         (profile_probability::differs_from_p): New member function.
26741         (profile_probability::differs_lot_from_p): New member function.
26742         (profile_probability::stream_in): New member function.
26743         (profile_probability::stream_out): New member function.
26744         * profile-count.h (profile_count_quality): Rename to ...
26745         (profile_quality): ... this one.
26746         (profile_probability): New.
26747         (profile_count): Update.
26748         * profile.c (compute_branch_probabilities): Update.
26749         * recog.c (peep2_attempt): Update.
26750         * sched-ebb.c (schedule_ebbs): Update.
26751         * sched-rgn.c (find_single_block_region): Update.
26752         (compute_dom_prob_ps): Update.
26753         (schedule_region): Update.
26754         * sel-sched-ir.c (compute_succs_info): Update.
26755         * stmt.c (struct case_node): Update.
26756         (do_jump_if_equal): Update.
26757         (get_outgoing_edge_probs): Update.
26758         (conditional_probability): Update.
26759         (emit_case_dispatch_table): Update.
26760         (expand_case): Update.
26761         (expand_sjlj_dispatch_table): Update.
26762         (emit_case_nodes): Update.
26763         * targhooks.c: Update.
26764         * tracer.c (better_p): Update.
26765         (find_best_successor): Update.
26766         * trans-mem.c (expand_transaction): Update.
26767         * tree-call-cdce.c: Update.
26768         * tree-cfg.c (gimple_split_edge): Upate.
26769         (move_sese_region_to_fn): Upate.
26770         * tree-cfgcleanup.c (cleanup_control_expr_graph): Upate.
26771         * tree-eh.c (lower_resx): Upate.
26772         (cleanup_empty_eh_move_lp): Upate.
26773         * tree-if-conv.c (version_loop_for_if_conversion): Update.
26774         * tree-inline.c (copy_edges_for_bb): Update.
26775         (copy_cfg_body): Update.
26776         * tree-parloops.c (gen_parallel_loop): Update.
26777         * tree-profile.c (gimple_gen_ic_func_profiler): Update.
26778         (gimple_gen_time_profiler): Update.
26779         * tree-ssa-dce.c (remove_dead_stmt): Update.
26780         * tree-ssa-ifcombine.c (update_profile_after_ifcombine): Update.
26781         * tree-ssa-loop-im.c (execute_sm_if_changed): Update.
26782         * tree-ssa-loop-ivcanon.c (remove_exits_and_undefined_stmts): Update.
26783         (unloop_loops): Update.
26784         (try_peel_loop): Update.
26785         * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Update.
26786         * tree-ssa-loop-split.c (connect_loops): Update.
26787         (split_loop): Update.
26788         * tree-ssa-loop-unswitch.c (tree_unswitch_loop): Update.
26789         (hoist_guard): Update.
26790         * tree-ssa-phionlycprop.c (propagate_rhs_into_lhs): Update.
26791         * tree-ssa-phiopt.c (replace_phi_edge_with_variable): Update.
26792         (value_replacement): Update.
26793         * tree-ssa-reassoc.c (branch_fixup): Update.
26794         * tree-ssa-tail-merge.c (replace_block_by): Update.
26795         * tree-ssa-threadupdate.c (remove_ctrl_stmt_and_useless_edges): Update.
26796         (create_edge_and_update_destination_phis): Update.
26797         (compute_path_counts): Update.
26798         (recompute_probabilities): Update.
26799         (update_joiner_offpath_counts): Update.
26800         (freqs_to_counts_path): Update.
26801         (duplicate_thread_path): Update.
26802         * tree-switch-conversion.c (hoist_edge_and_branch_if_true): Update.
26803         (struct switch_conv_info): Update.
26804         (gen_inbound_check): Update.
26805         * tree-vect-loop-manip.c (slpeel_add_loop_guard): Update.
26806         (vect_do_peeling): Update.
26807         (vect_loop_versioning): Update.
26808         * tree-vect-loop.c (scale_profile_for_vect_loop): Update.
26809         (optimize_mask_stores): Update.
26810         * ubsan.c (ubsan_expand_null_ifn): Update.
26811         * value-prof.c (gimple_divmod_fixed_value): Update.
26812         (gimple_divmod_fixed_value_transform): Update.
26813         (gimple_mod_pow2): Update.
26814         (gimple_mod_pow2_value_transform): Update.
26815         (gimple_mod_subtract): Update.
26816         (gimple_mod_subtract_transform): Update.
26817         (gimple_ic): Update.
26818         (gimple_stringop_fixed_value): Update.
26819         (gimple_stringops_transform): Update.
26820         * value-prof.h: Update.
26822 2017-06-29  Carl Love  <cel@us.ibm.com>
26824         * config/rs6000/rs6000-c.c: Add support for built-in functions
26825         vector signed int vec_signed (vector float);
26826         vector signed long long vec_signed (vector double);
26827         vector signed int vec_signed2 (vector double, vector double);
26828         vector signed int vec_signede (vector double);
26829         vector signed int vec_signedo (vector double);
26830         * config/rs6000/rs6000.c (rs6000_generate_vsigned2_code): Add
26831         instruction generator.
26832         * config/rs6000/vsx.md (UNSPEC_VSX_XVCVSPSXWS, UNSPEC_VSX_XVCVSPSXDS,
26833         UNSPEC_VSX_VSIGNED2): Add UNSPECS.
26834         (vsx_xvcvspsxws, vsx_xvcvdpuxds_scale, vsx_xvcvspuxws, vsigned2_v2df):
26835         Add define_insn.
26836         (vsignedo_v2df, vsignede_v2df, vunsigned2_v2df, vunsignedo_v2df,
26837         vunsignede_v2df): Add define_expands.
26838         * config/rs6000/rs6000-builtin.def (VEC_SIGNED, VEC_UNSIGNED,
26839         VEC_SIGNED2, VEC_UNSIGNED2, VEC_SIGNEDE, VEC_UNSIGNEDE, VEC_SIGNEDO,
26840         VEC_UNSIGNEDO): Add definitions.
26841         * config/vsx.md (UNSPEC_VSX_XVCVSPSXWS, UNSPEC_VSX_XVCVSPSXDS,
26842         UNSPEC_VSX_VSIGNED2): Add UNSPECs.
26843         (vsx_xvcvspsxws, vsx_xvcvspuxws): Add define_insn.
26844         (vsigned2_v2df, vsigendo_v2df, vsignede_v2df,
26845         vunsigned2_v2df, vunsignedo_v2df, vunsignede_v2df): Add define_expands.
26846         * config/rs6000/altivec.h (vec_signed, vec_signed2,
26847         vec_signede and vec_signedo, vec_unsigned, vec_unsigned2,
26848         vec_unsignede, vec_unsignedo): Add builtin defines.
26849         * config/rs6000-protos.h (rs6000_generate_vsigned2_code): Add extern
26850         declaration.
26851         * doc/extend.texi: Update the built-in documentation file for the
26852         new built-in functions.
26854 2017-06-29  Richard Biener  <rguenther@suse.de>
26856         * tree-vect-loop.c (vect_analyze_scalar_cycles_1): Do not add
26857         reduction chains to LOOP_VINFO_REDUCTIONS.
26858         * tree-vect-slp.c (vect_analyze_slp): Continue looking for
26859         SLP reductions after processing reduction chains.
26861 2017-06-29  Nathan Sidwell  <nathan@acm.org>
26863         * builtins.c (fold_builtin_FUNCTION): Use
26864         lang_hooks.decl_printable_name.
26866 2017-06-29  Peter Bergner  <bergner@vnet.ibm.com>
26868         PR middle-end/81194
26869         * cfgexpand.c (expand_gimple_stmt_1): Handle switch statements
26870         with only one label.
26871         * stmt.c (expand_case): Assert NCASES is greater than one.
26873 2017-06-29  Richard Biener  <rguenther@suse.de>
26875         * tree-cfg.c (group_case_labels_stmt): Return whether we changed
26876         anything.
26877         (group_case_labels): Likewise.
26878         (find_taken_edge): Push sanity checking on val to workers...
26879         (find_taken_edge_cond_expr): ... here
26880         (find_taken_edge_switch_expr): ... and here, handle cases
26881         with just a default label.
26882         * tree-cfg.h (group_case_labels_stmt): Adjust prototype.
26883         (group_case_labels): Likewise.
26884         * tree-cfgcleanup.c (execute_cleanup_cfg_post_optimizing): When
26885         group_case_labels does anything cleanup the CFG again.
26887 2017-06-29  Bin Cheng  <bin.cheng@arm.com>
26889         PR tree-optimization/81196
26890         * tree-ssa-loop-niter.c (number_of_iterations_cond): Handle loop
26891         exit condition comparing two IVs.
26893 2017-06-29  Richard Earnshaw  <rearnsha@arm.com>
26895         * config/arm/parsecpu.awk (gen_comm_data): Add initializer for
26896         profile to the dummy entry at the end of the list of architectures.
26897         * config/arm/arm-cpu-cdata.h: Regenerated.
26899 2017-06-29  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
26900             Michael Collison <michael.collison@arm.com>
26902         PR target/70119
26903         * config/aarch64/aarch64.md (*aarch64_<optab>_reg_<mode>3_mask1):
26904         New pattern.
26905         (*aarch64_reg_<mode>3_neg_mask2): New pattern.
26906         (*aarch64_reg_<mode>3_minus_mask): New pattern.
26907         (*aarch64_<optab>_reg_di3_mask2): New pattern.
26908         * config/aarch64/aarch64.c (aarch64_rtx_costs): Account for cost
26909         of shift when the shift amount is masked with constant equal to
26910         the size of the mode.
26911         * config/aarch64/predicates.md (subreg_lowpart_operator): New
26912         predicate.
26914 2017-06-29  Martin Liska  <mliska@suse.cz>
26916         * config/i386/i386.opt: Change range from [1,5] to [0,5].
26918 2017-06-29  Yury Gribov  <tetra2005@gmail.com>
26920         PR bootstrap/80565
26921         * ipa-cp.c (allocate_and_init_ipcp_value): Add initialization
26922         code.
26923         * ipa-inline.h
26924         (edge_growth_cache_entry::edge_growth_cache_entry): New
26925         function.
26926         (reset_edge_growth_cache): Update to use constructor.
26928 2017-06-28  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
26930         * config/aarch64/aarch64.h (AARCH64_EXPAND_ALIGNMENT): New.
26931         (DATA_ALIGNMENT): Update to use AARCH64_EXPAND_ALIGNMENT.
26932         (LOCAL_ALIGNMENT): Update to use AARCH64_EXPAND_ALIGNMENT.
26934 2017-06-28  Sebastian Peryt  <sebastian.peryt@intel.com>
26936         * config/i386/avx512vlintrin.h (_mm256_permutexvar_epi64)
26937         (_mm256_permutexvar_epi32, _mm256_permutex_epi64): New intrinsics.
26939 2017-06-28  Szabolcs Nagy  <szabolcs.nagy@arm.com>
26941         * config.gcc (*-linux-musl*): Add t-musl tmake_file.
26942         (*-linux-uclibc*): Add t-uclibc tmake_file.
26943         * config/t-musl: New.
26944         * config/t-uclibc: New.
26946 2017-06-28  Richard Earnshaw  <rearnsha@arm.com>
26948         * config/arm/parsecpu.awk (profile): Parse new keyword in an arch
26949         context.
26950         (gen_comm_data): Emit architectural setting of arch_prof.
26951         * config/arm/arm-cpus.in (armv6-m, armv6s-m, armv7-a, armv7ve): Set the
26952         profile.
26953         (armv7-r, armv7-m, armv7e-m, armv8-a, armv8.1-a, armv8.2-a): Likewise.
26954         (armv8-m.base, armv8-m.main): Likewise.
26955         * arm-protos.h (arm_build_target): Add profile field.
26956         (arch_option): Likewise.
26957         * config/arm/arm.c (arm_configure_build_target): Copy the profile to
26958         the active target.
26959         * config/arm/arm.h (TARGET_ARM_ARCH_PROFILE): Use
26960         arm_active_target.profile.
26962 2017-06-28  Richard Biener  <rguenther@suse.de>
26964         PR middle-end/81227
26965         * fold-const.c (negate_expr_p): Use TYPE_UNSIGNED, not
26966         TYPE_OVERFLOW_WRAPS.
26967         * match.pd (negate_expr_p): Likewise.
26968         * tree-ssa-reassoc.c (optimize_range_tests_diff): Use
26969         fold_build2, not fold_binary.
26971 2017-06-28  Wilco Dijkstra  <wdijkstr@arm.com>
26973         * config/aarch64/aarch64 (aarch64_expand_mov_immediate):
26974         Convert memory address to Pmode.
26975         (aarch64_print_operand): Assert MEM operands are always Pmode.
26977 2017-06-28  Wilco Dijkstra  <wdijkstr@arm.com>
26979         PR target/79665
26980         * config/arm/aarch-common.c (arm_no_early_alu_shift_dep):
26981         Remove redundant if.
26982         (aarch_forward_to_shift_is_not_shifted_reg): Remove.
26983         * config/arm/aarch-common-protos.h
26984         (aarch_forward_to_shift_is_not_shifted_re): Remove.
26985         * config/arm/cortex-a53.md: Use arm_no_early_alu_shift_dep in bypass.
26987 2017-06-28  Michael Meissner  <meissner@linux.vnet.ibm.com>
26989         PR ipa/81238
26990         * multiple_target.c (create_dispatcher_calls): Set the default
26991         clone to be static, not public.
26993 2017-06-28  Richard Biener  <rguenther@suse.de>
26995         * tree-vect-loop.c (vectorizable_reduction): Move special
26996         cond reduction IV var creation ...
26997         (vect_create_epilog_for_reduction): ... here.  Remove induction_index
26998         parameter.  Use STMT_VINFO_VECTYPE.
26999         * tree-vect-slp.c (vect_get_constant_vectors): Properly reset
27000         constant_p.
27002 2017-06-28  Martin Liska  <mliska@suse.cz>
27004         PR ipa/81128
27005         * ipa-visibility.c (non_local_p): Handle visibility.
27007 2017-06-28  Martin Liska  <mliska@suse.cz>
27009         PR driver/79659
27010         * common.opt: Add IntegerRange to various options.
27011         * opt-functions.awk (integer_range_info): New function.
27012         * optc-gen.awk: Add integer_range_info to cl_options struct.
27013         * opts-common.c (decode_cmdline_option): Handle
27014         CL_ERR_INT_RANGE_ARG.
27015         (cmdline_handle_error): Likewise.
27016         * opts.c (print_filtered_help): Show valid interval in
27017         when --help is provided.
27018         * opts.h (struct cl_option): Add range_min and range_max fields.
27019         * config/i386/i386.opt: Add IntegerRange for -mbranch-cost.
27021 2017-06-28  Marc Glisse  <marc.glisse@inria.fr>
27023         * match.pd ((X & ~Y) | (~X & Y)): Generalize to + and ^.
27024         (x * C EQ/NE y * C): New transformation.
27026 2017-06-28  Christophe Lyon  <christophe.lyon@linaro.org>
27028         * genmultilib (combination_space): Accept '+' in option names.
27030 2017-06-28  Martin Liska  <mliska@suse.cz>
27032         PR sanitizer/81224
27033         * asan.c (instrument_derefs): Bail out inner references
27034         that are hard register variables.
27036 2017-06-28  Jakub Jelinek  <jakub@redhat.com>
27038         PR target/81175
27039         * config/i386/i386.c (ix86_init_mmx_sse_builtins): Use def_builtin
27040         rather than def_builtin_pure for __builtin_ia32_gatherpf*.
27042 2017-06-28  Richard Biener  <rguenther@suse.de>
27044         * tree-vectorizer.h (vect_get_vec_defs): Remove.
27045         (vect_get_slp_defs): Adjust.
27046         * tree-vect-loop.c (get_initial_defs_for_reduction): Split
27047         out from ...
27048         * tree-vect-slp.c (vect_get_constant_vectors): ... here and
27049         simplify.
27050         * tree-vect-loop.c (vect_create_epilog_for_reduction): Use
27051         get_initial_defs_for_reduction instead of vect_get_vec_defs.
27052         (vectorizable_reduction): Adjust.
27053         * tree-vect-slp.c (vect_get_constant_vectors): Remove reduction
27054         handling.
27055         (vect_get_slp_defs): Likewise.
27056         * tree-vect-stmts.c (vect_get_vec_defs): Make static and adjust.
27057         (vectorizable_bswap): Adjust.
27058         (vectorizable_call): Likewise.
27059         (vectorizable_conversion): Likewise.
27060         (vectorizable_assignment): Likewise.
27061         (vectorizable_shift): Likewise.
27062         (vectorizable_operation): Likewise.
27063         (vectorizable_store): Likewise.
27064         (vectorizable_condition): Likewise.
27065         (vectorizable_comparison): Likewise.
27067 2017-06-28  Michael Collison  <michael.collison@arm.com>
27069         PR target/68535
27070         * config/arm/arm.c (gen_ldm_seq): Remove last unnecessary
27071         set of base_reg
27072         (arm_gen_movmemqi): Removed unused variable 'i'.
27073         Convert 'for' loop into 'while' loop.
27074         (arm_expand_prologue): Remove last unnecessary set of insn.
27075         (thumb_pop): Remove unused variable 'pushed_words'.
27076         (thumb_exit): Remove last unnecessary set of regs_to_pop.
27078 2017-06-28  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
27080         * config/s390/predicates.md: Use s390_rel_address_ok_p.
27081         * config/s390/s390-protos.h: Add prototype of
27082         s390_rel_address_ok_p.
27083         * config/s390/s390.c (s390_got_symbol): New function.
27084         (s390_rel_address_ok_p): New function.
27085         (legitimize_pic_address): Use s390_rel_address_ok_p.
27086         (s390_load_got): Use s390_got_symbol.
27087         (s390_option_override): Issue error if
27088         -mno-pic-data-is-text-relative is used without -fpic/-fPIC.
27089         * config/s390/s390.h (TARGET_DEFAULT_PIC_DATA_IS_TEXT_RELATIVE):
27090         New macro.
27091         * config/s390/s390.opt: New option mpic-data-is-text-relative.
27093 2017-06-27  Andrew Pinski  <apinski@cavium.com>
27095         * match.pd (X >/>=/</<= 0 ? 1.0 : -1.0): New patterns.
27096         (X * copysign (1.0, X)): New pattern.
27097         (X * copysign (1.0, -X)): New pattern.
27098         (copysign (-1.0, CST)): New pattern.
27100 2017-06-27  Joseph Myers  <joseph@codesourcery.com>
27102         * genmultilib (combination_space): Remove variable.
27103         Validate reuse rules against regular expression for any sequence
27104         of multilib options in any order.
27106 2017-06-27  Michael Collison  <michael.collison@arm.com>
27108         * config/aarch64/aarch64-simd.md (aarch64_combine<mode>): Directly
27109         call aarch64_split_simd_combine.
27110         * (aarch64_combine_internal<mode>): Delete pattern.
27111         * config/aarch64/aarch64.c (aarch64_split_simd_combine):
27112         Allow register and subreg operands.
27114 2017-06-27  Jerome Lambourg  <lambourg@adacore.com>
27116         * config/i386/vxworks.h (ASM_SPEC): Remove definition. No target
27117         specific need, just fallback on defaults.
27118         (ASM_OUTPUT_ALIGNED_BSS): Add #undef before #define.
27120 2017-06-27  Jerome Lambourg  <lambourg@adacore.com>
27121             Olivier Hainque  <hainque@adacore.com>
27123         * config/i386/vxworks.h (DBX_REGISTER_NUMBER): Pick distinct
27124         map for 64bits.
27125         (TARGET_OS_CPP_BUILTINS): builtin_define CPU to X86_64 for 64bit
27126         targets. Pick a default if no particular attempt applied.
27127         (STACK_CHECK_PROTECT): Double for 64bit targets, which have
27128         larger contexts.
27130 2017-06-27  Jerome Lambourg  <lambourg@adacore.com>
27132         * config.gcc (i*86-wrs-vxworks7): Handle new acceptable triplet.
27133         (x86_64-wrs-vxworks7): Likewise.
27135 2017-06-27  Marek Polacek  <polacek@redhat.com>
27137         PR sanitizer/81223
27138         * ubsan.c (instrument_null): Check get_base_address's result for null.
27140 2017-06-27  Marc Glisse  <marc.glisse@inria.fr>
27142         * match.pd ((A+-B)+(C-A), (A+B)-(A-C)): New transformations.
27144 2017-06-27  Marc Glisse  <marc.glisse@inria.fr>
27146         * builtin-types.def (BT_FENV_T_PTR, BT_CONST_FENV_T_PTR,
27147         BT_FEXCEPT_T_PTR, BT_CONST_FEXCEPT_T_PTR): New primitive types.
27148         (BT_FN_INT_FENV_T_PTR, BT_FN_INT_CONST_FENV_T_PTR,
27149         BT_FN_INT_FEXCEPT_T_PTR_INT, BT_FN_INT_CONST_FEXCEPT_T_PTR_INT):
27150         New function types.
27151         * builtins.def (BUILT_IN_FECLEAREXCEPT, BUILT_IN_FEGETENV,
27152         BUILT_IN_FEGETEXCEPTFLAG, BUILT_IN_FEGETROUND,
27153         BUILT_IN_FEHOLDEXCEPT, BUILT_IN_FERAISEEXCEPT,
27154         BUILT_IN_FESETENV, BUILT_IN_FESETEXCEPTFLAG,
27155         BUILT_IN_FESETROUND, BUILT_IN_FETESTEXCEPT,
27156         BUILT_IN_FEUPDATEENV): New builtins.
27157         * tree-core.h (TI_FENV_T_PTR_TYPE, TI_CONST_FENV_T_PTR_TYPE,
27158         TI_FEXCEPT_T_PTR_TYPE, TI_CONST_FEXCEPT_T_PTR_TYPE): New entries.
27159         * tree.h (fenv_t_ptr_type_node, const_fenv_t_ptr_type_node,
27160         fexcept_t_ptr_type_node, const_fexcept_t_ptr_type_node): New
27161         macros.
27162         (builtin_structptr_types): Adjust size.
27163         * tree.c (builtin_structptr_types): Add four entries.
27165 2017-06-27  Jerome Lambourg  <lambourg@adacore.com>
27166             Olivier Hainque  <hainque@adacore.com>
27168         * config/vxworks.h (VXWORKS_LIB_SPEC): Incorporate ...
27169         (TLS_SYM): New local macro, forcing reference to __tls__ on
27170         link command lines for VxWorks 7 RTPs, triggering initialization
27171         of tlsLib.
27172         (VXWORKS_HAVE_TLS): New macro. State whether the target VxWorks
27173         OS features TLS support, true for RTPs on VxWorks 7.
27174         * config/vxworks.c (vxworks_override_options): Setup emutls
27175         accordingly.
27177 2017-06-27  Jakub Jelinek  <jakub@redhat.com>
27179         * predict.c (test_prediction_value_range): Use -1U instead of -1
27180         to avoid narrowing conversion warning.
27181         * dumpfile.c (dump_options): Wrap all value into dump_flags_t cast
27182         to avoid narrowing conversion warning.
27183         * opt-functions.awk (var_ref): Return (unsigned short) -1 instead of
27184         -1.
27185         * optc-gen.awk (END): Expect (unsigned short) -1 instead of -1.
27187 2017-06-27  Jerome Lambourg  <lambourg@adacore.com>
27189         * config/vxworks.h (VXWORKS_LIBS_RTP): Alternative definition for
27190         64bit configurations.
27191         (PTR_DIFF_TYPE): Alternative definition for TARGET_LP64.
27192         (SIZE_TYPE): Likewise.
27193         * config/vxworks.c (vxworks_emutls_var_fields): Use
27194         long_unsigned_type_node instead of unsigned_type_node as the offset
27195         field type, which is "pointer" mode in emutls.c.
27197 2017-06-27  Jakub Jelinek  <jakub@redhat.com>
27199         PR sanitizer/81209
27200         * ubsan.c (ubsan_encode_value): Initialize DECL_CONTEXT on var.
27202         PR middle-end/81207
27203         * gimple-fold.c (replace_call_with_call_and_fold): Handle
27204         gimple_vuse copying separately from gimple_vdef copying.
27206 2017-06-27  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
27208         * value-prof.c (free_hist): Remove call to memset and the enclosing if
27209         condition.
27211 2017-06-26  Jerome Lambourg  <lambourg@adacore.com>
27212             Olivier Hainque  <hainque@adacore.com>
27214         * config.gcc (*-*-vxworks*): Add TARGET_VXWORKS7=1 to tm_defines
27215         for all vxworks7 targets.
27216         * config/vxworks.h (TARGET_VXWORKS7): If not defined, define to 0.
27217         (VXWORKS_ADDITIONAL_CPP_SPEC): Alternative definition for VXWORKS7.
27218         (VXWORKS_LIBS_RTP, VXWORKS_LIBS_RTP_DIR): New macros, allowing
27219         variations for VX6/VX7 and 32/64bits later on in ...
27220         (VXWORKS_LIB_SPEC): Leverage new macros.
27221         (VXWORKS_OS_CPP_BUILTINS): Define _VSB_CONFIG_FILE for VXWORKS7,
27222         as well as _ALLOW_KEYWORD_MACROS when "inline" is not a keyword.
27224 2017-06-26  Jerome Lambourg  <lambourg@adacore.com>
27226         * config/vxworks.h (VXWORKS_OS_CPP_BUILTINS): builtin_define
27227         _VX_TOOL_FAMILY and _VX_TOOL to gnu.
27229 2017-06-26  Carl Love  <cel@us.ibm.com>
27231         * config/rs6000/rs6000-c.c: Add support for built-in functions
27232         vector bool char vec_reve (vector bool char);
27233         vector signed char vec_reve (vector signed char);
27234         vector unsigned char vec_reve (vector unsigned char);
27235         vector bool int vec_reve (vector bool int);
27236         vector signed int vec_reve (vector signed int);
27237         vector unsigned int vec_reve (vector unsigned int);
27238         vector bool long long vec_reve (vector bool long long);
27239         vector signed long long vec_reve (vector signed long long);
27240         vector unsigned long long vec_reve (vector unsigned long long);
27241         vector bool short vec_reve (vector bool short);
27242         vector signed short vec_reve (vector signed short);
27243         vector double vec_reve (vector double);
27244         vector float vec_reve (vector float);
27245         * config/rs6000/rs6000-builtin.def (VREVE_V2DI, VREVE_V4SI,
27246         VREVE_V8HI, VREVE_V16QI, VREVE_V2DF, VREVE_V4SF, VREVE): New builtin.
27247         * config/rs6000/altivec.md (UNSPEC_VREVEV): New UNSPEC.
27248         (altivec_vreve): New pattern.
27249         * config/rs6000/altivec.h (vec_reve): New define.
27250         * doc/extend.texi (vec_rev): Update the built-in documentation file
27251         for the new built-in functions.
27253 2016-06-26  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
27255         PR tree-optimization/71815
27256         * gimple-ssa-strength-reduction.c (uses_consumed_by_stmt): New
27257         function.
27258         (find_basis_for_candidate): Call uses_consumed_by_stmt rather than
27259         has_single_use.
27260         (slsr_process_phi): Likewise.
27261         (replace_uncond_cands_and_profitable_phis): Don't replace a
27262         multiply candidate with a stride of 1 (copy or cast).
27263         (phi_incr_cost): Call uses_consumed_by_stmt rather than
27264         has_single_use.
27265         (lowest_cost_path): Likewise.
27266         (total_savings): Likewise.
27268 2017-06-26  Richard Biener  <rguenther@suse.de>
27270         PR target/81175
27271         * config/i386/i386.c (ix86_init_mmx_sse_builtins):
27272         Use def_builtin_pure for all gather builtins.
27274 2017-06-26  Richard Biener  <rguenther@suse.de>
27276         PR tree-optimization/81203
27277         * tree-tailcall.c (find_tail_calls): Do not move stmts into
27278         non-dominating BBs.
27280 2017-06-26  Marek Polacek  <polacek@redhat.com>
27282         PR c/80116
27283         * doc/invoke.texi: Document -Wmultistatement-macros.
27285 2017-06-26  Christophe Lyon  <christophe.lyon@linaro.org>
27287         * doc/sourcebuild.texi (ARM-specific attributes): Document new
27288         arm_neon_ok_no_float_abi effective target.
27290 2017-06-26  Richard Biener  <rguenther@suse.de>
27292         PR tree-optimization/80928
27293         * cfghooks.c (duplicate_block): Do not copy BB_DUPLICATED flag.
27294         (copy_bbs): Set BB_DUPLICATED flag early.
27295         (execute_on_growing_pred): Do not execute for BB_DUPLICATED
27296         marked blocks.
27297         (execute_on_shrinking_pred): Likewise.
27298         * tree-ssa.c (ssa_redirect_edge): Do not look for PHI args in
27299         BB_DUPLICATED blocks.
27300         * tree-ssa-phionlycoprop.c (eliminate_degenerate_phis_1): Properly
27301         iterate over all PHIs considering removal of *gsi.
27303 2017-06-23  Jim Wilson  <jim.wilson@linaro.org>
27305         * doc/invoke.texi (AArch64 Options, -mtune): Re-add falkor and
27306         qdf24xx.
27308 2017-06-23  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
27310         * config/rs6000/rs6000-string.c: (expand_block_clear,
27311         do_load_for_compare, select_block_compare_mode,
27312         compute_current_alignment, expand_block_compare,
27313         expand_strncmp_align_check, expand_strn_compare,
27314         expand_block_move, rs6000_output_load_multiple)
27315         Move functions related to string/block move/compare
27316         to a separate file.
27317         * config/rs6000/rs6000.c: Move above functions to rs6000-string.c.
27318         * config/rs6000/rs6000-protos.h (rs6000_emit_dot_insn): Add prototype
27319         for this function which is now used in two files.
27320         * config/rs6000/t-rs6000: Add rule to compile rs6000-string.o.
27321         * config.gcc: Add rs6000-string.o to extra_objs for
27322         targets powerpc*-*-* and rs6000*-*-*.
27324 2017-06-23  Michael Meissner  <meissner@linux.vnet.ibm.com>
27326         PR target/80510
27327         * config/rs6000/rs6000.md (ALTIVEC_DFORM): Do not allow DImode in
27328         32-bit, since indexed is not valid for DImode.
27329         (mov<mode>_hardfloat32): Reorder ISA 2.07 load/stores before ISA
27330         3.0 d-form load/stores to be the same as mov<mode>_hardfloat64.
27331         (define_peephole2 for Altivec d-form load): Add 32-bit support.
27332         (define_peephole2 for Altivec d-form store): Likewise.
27334         PR ipa/81185
27335         * multiple_target.c (create_dispatcher_calls): Only create the
27336         dispatcher call if the function is the default clone of a
27337         versioned function.
27339 2017-06-23  Segher Boessenkool  <segher@kernel.crashing.org>
27341         PR middle-end/80902
27342         * builtins.c (expand_builtin_atomic_fetch_op): If emitting code after
27343         a call, force the call to not be a tail call.
27345 2017-06-23  Jeff Law  <law@redhat.com>
27347         * doc/contrib.texi: Add entry for Steven Pemberton's work on
27348         enquire.
27350 2017-06-23  Will Schmidt  <will_schmidt@vnet.ibm.com>
27352         * config/rs6000/rs6000.c: Add include of ssa-propagate.h for
27353         update_call_from_tree().  (rs6000_gimple_fold_builtin): Add
27354         handling for early expansion of vector shifts (sl,sr,sra,rl).
27355         (builtin_function_type): Add vector shift right instructions
27356         to the unsigned argument list.
27358 2017-06-23  Bernd Edlinger  <bernd.edlinger@hotmail.de>
27360         rtl-optimizatoin/79286
27361         * ira.c (update_equiv_regs): Revert to using may_trap_or_fault_p again.
27362         * rtlanal.c (rtx_addr_can_trap_p_1): SYMBOL_REF_FUNCTION_P can never
27363         trap.  PIC register plus a const unspec without offset can never trap.
27365 2017-06-23  Marc Glisse  <marc.glisse@inria.fr>
27367         * tree.h (builtin_structptr_type): New type.
27368         (builtin_structptr_types): Declare new array.
27369         * tree.c (builtin_structptr_types): New array.
27370         (free_lang_data, build_common_tree_nodes): Use it.
27372 2017-06-23  Jonathan Wakely  <jwakely@redhat.com>
27374         PR c++/81187
27375         * doc/invoke.texi (-Wnoexcept-type): Fix name of option, from
27376         -Wnoexcept.
27378 2017-06-22  Matt Turner  <mattst88@gmail.com>
27380         * config/i386/driver-i386.c (host_detect_local_cpu): Add Kaby
27381         Lake models to skylake case.  Assume skylake for unknown
27382         models with clflushopt.
27384 2017-06-22  Jeff Law  <law@redhat.com>
27386         * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Handle
27387         frame sizes that do not satisfy aarch64_uimm12_shift.
27389 2017-06-22  Jan Hubicka <hubicka@ucw.cz>
27391         * profile-count.h (apply_probability,
27392         apply_scale, probability_in): Fix checks for zero.
27394 2017-06-22  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
27396         * incpath.c (add_sysroot_to_chain): Allow for $SYSROOT prefix.
27397         * doc/cppdiropts.texi (-I @var{dir}): Document it.
27399 2016-06-22  Richard Biener  <rguenther@suse.de>
27401         * tree-vect-loop.c (vect_model_reduction_cost): Handle
27402         COND_REDUCTION and INTEGER_INDUC_COND_REDUCTION without
27403         REDUC_MAX_EXPR support.
27404         (vectorizable_reduction): Likewise.
27405         (vect_create_epilog_for_reduction): Likewise.
27407 2017-06-22  James Greenhalgh  <james.greenhalgh@arm.com>
27409         * match.pd (A / (1 << B) -> A >> B): New.
27410         * generic-match-head.c: Include optabs-tree.h.
27411         * gimple-match-head.c: Likewise.
27412         * optabs-tree.h (target_supports_op_p): New.
27413         * optabs-tree.c (target_supports_op_p): New.
27415 2017-06-22  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
27417         * configure.ac (gcc_cv_ld_static_dynamic): Also check stderr for
27418         $gcc_cv_ld --help output.
27419         (gcc_cv_ld_demangle): Likewise.
27420         (gcc_cv_ld_eh_frame_hdr): Likewise.
27421         (gcc_cv_ld_pie): Likewise.
27422         (gcc_cv_ld_as_needed): Likewise.  Prefer native forms unless $gnu_ld.
27423         (gcc_cv_ld_buildid): Likewise.
27424         (gcc_cv_ld_sysroot): Likewise.
27425         (ld_bndplt_support): Likewise.
27426         (ld_pushpopstate_support): Likewise.
27427         * configure: Regenerate.
27428         * config/sol2.h [!USE_GLD] (SYSROOT_SPEC): Define.
27430 2017-06-21  Jakub Jelinek  <jakub@redhat.com>
27432         PR target/81151
27433         * config/i386/sse.md (round<mode>2): Renumber match_dup and
27434         operands indexes to avoid gap between operands and match_dups.
27436 2017-06-21  Andrew Pinski  <apinski@cavium.com>
27438         * config/aarch64/aarch64-cost-tables.h (thunderx_extra_costs):
27439         Increment Arith_shift and Arith_shift_reg by 1.
27440         * config/aarch64/aarch64-tuning-flags.def (cheap_shift_extend):
27441         New tuning flag.
27442         * config/aarch64/aarch64.c (thunderx_tunings): Enable
27443         AARCH64_EXTRA_TUNE_CHEAP_SHIFT_EXTEND.
27444         (aarch64_strip_extend): Add new argument and test for it.
27445         (aarch64_cheap_mult_shift_p): New function.
27446         (aarch64_rtx_mult_cost): Call aarch64_cheap_mult_shift_p and don't
27447         add a cost if it is true.
27448         Update calls to aarch64_strip_extend.
27449         (aarch64_rtx_costs): Update calls to aarch64_strip_extend.
27451 2017-06-21  Andrew Pinski  <apinski@cavium.com>
27453         * config/aarch64/aarch64-cores.def (thunderxt88p1): Use thunderxt88
27454         tunings.
27455         (thunderxt88): Likewise.
27456         * config/aarch64/aarch64.c (thunderxt88_prefetch_tune): New variable.
27457         (thunderx_prefetch_tune): New variable.
27458         (thunderx2t99_prefetch_tune): Update for the correct values.
27459         (thunderxt88_tunings): New variable.
27460         (thunderx_tunings): Use thunderx_prefetch_tune instead of
27461         generic_prefetch_tune.
27462         (thunderx2t99_tunings): Use AUTOPREFETCHER_WEAK.
27464 2017-06-21  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
27466         * config/aarch64/atomics.md (aarch64_compare_and_swap<mode>_lse,
27467         SHORT): Relax operand 3 to aarch64_reg_or_zero and constraint to Z.
27468         (aarch64_compare_and_swap<mode>_lse, GPI): Likewise.
27469         (aarch64_atomic_cas<mode>, SHORT): Likewise for operand 2.
27470         (aarch64_atomic_cas<mode>, GPI): Likewise.
27472 2017-06-21  Martin Liska  <mliska@suse.cz>
27474         * gimplify.c (gimplify_label_expr): Insert GIMPLE_PREDICT
27475         statements on cold and hot labels.
27476         * predict.c (tree_estimate_probability_bb): Remove the
27477         prediction from this place.
27479 2017-06-21  Martin Liska  <mliska@suse.cz>
27481         PR tree-optimization/79489
27482         * gimplify.c (maybe_add_early_return_predict_stmt): New
27483         function.
27484         (gimplify_return_expr): Call the function.
27485         * predict.c (tree_estimate_probability_bb): Remove handling
27486         of early return.
27487         * predict.def: Update comment about early return predictor.
27488         * gimple-predict.h (is_gimple_predict): New function.
27489         * predict.def: Change default value of early return to 66.
27490         * tree-tailcall.c (find_tail_calls): Skip GIMPLE_PREDICT
27491         statements.
27492         * passes.def: Put pass_strip_predict_hints to the beginning of
27493         IPA passes.
27495 2017-06-21  Pierre-Marie de Rodat  <derodat@adacore.com>
27497         * dwarf2out.c (gen_decl_die): Remove the guard to skip file-scope
27498         FUNCTION_DECL declarations.
27499         (dwarf2out_early_global_decl): Remove the guard to skip FUNCTION_DECL
27500         declarations.
27501         (dwaf2out_decl): Likewise.
27502         * godump.c (go_early_global_decl): Skip call to the real debug hook
27503         for FUNCTION_DECL declarations.
27504         * passes.c (rest_of_decl_compilation): Skip call to the
27505         early_global_decl debug hook for FUNCTION_DECL declarations, unless
27506         -fdump-go-spec is passed.
27508 2017-06-21  Marc Glisse  <marc.glisse@inria.fr>
27510         * config/i386/i386.c (struct builtin_isa): New field pure_p.
27511         Reorder for compactness.
27512         (def_builtin, def_builtin2, ix86_add_new_builtins): Handle pure_p.
27513         (def_builtin_pure, def_builtin_pure2): New functions.
27514         (ix86_init_mmx_sse_builtins) [__builtin_ia32_stmxcsr]: Mark as pure.
27516 2017-06-21  Marc Glisse  <marc.glisse@inria.fr>
27518         * match.pd (nop_convert): New predicate.
27519         ((A +- CST1) +- CST2): Allow some NOP conversions.
27521 2017-06-21  Jakub Jelinek  <jakub@redhat.com>
27523         PR c++/81130
27524         * gimplify.c (omp_add_variable): Don't force GOVD_SEEN for types
27525         with ctors/dtors if GOVD_SHARED is set.
27527 2017-06-21  Wilco Dijkstra  <wdijkstr@arm.com>
27529         * config/aarch64/aarch64.md (movti_aarch64):
27530         Emit mov rather than orr.
27531         (movtf_aarch64): Likewise.
27532         * config/aarch64/aarch64-simd.md (aarch64_simd_mov):
27533         Emit mov rather than orr.
27535 2017-06-21  Wilco Dijkstra  <wdijkstr@arm.com>
27537         * config/aarch64/aarch64-simd.md (aarch64_simd_dup):
27538         Swap alternatives, make integer dup more expensive.
27540 2017-06-21  Wilco Dijkstra  <wdijkstr@arm.com>
27542         * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
27543         Return true for non-tls symbols.
27545 2017-06-21  James Greenhalgh  <james.greenhalgh@arm.com>
27547         * config/aarch64/aarch64-cores.def (cortex-a55): New.
27548         (cortex-a75): Likewise.
27549         (cortex-a75.cortex-a55): Likewise.
27550         * config/aarch64/aarch64-tune.md: Regenerate.
27551         * doc/invoke.texi (-mtune): Document new values for -mtune.
27553 2017-06-21  Tom de Vries  <tom@codesourcery.com>
27555         * doc/sourcebuild.texi (Add Options, Features for dg-add-options): Add
27556         stack_size feature.
27557         (Effective-Target Keywords, Other attributes): Suggest using
27558         dg-add-options stack_size feature to get stack limit in stack_size
27559         effective target documentation.
27561 2017-06-21  Julian Brown  <julian@codesourcery.com>
27562             Naveen H.S  <Naveen.Hurugalawadi@cavium.com>
27564         * config/aarch64/aarch64-simd.md (aarch64_crypto_pmulldi)
27565         (aarch64_crypto_pmullv2di): Change type attribute to crypto_pmull.
27566         * config/aarch64/thunderx2t99.md (thunderx2t99_pmull): New
27567         reservation.
27568         * config/arm/cortex-a53.md (cortex_a53_advsimd_type): Add crypto_pmull to
27569         attribute type list for neon_multiply.
27570         * config/arm/cortex-a57.md (cortex_a57_neon_type): Add crypto_pmull to
27571         attribute type list for neon_multiply.
27572         * config/arm/crypto.md (crypto_vmullp64): Change type to crypto_pmull.
27573         * config/arm/exynos-m1.md (exynos_m1_neon_type): Add crypto_pmull to
27574         attribute type list for neon_multiply.
27575         * config/arm/types.md (crypto_pmull): Add.
27576         * config/arm/xgene1.md (xgene1_neon_pmull): Add crypto_pmull to
27577         attribute type list.
27579 2017-06-20  Andreas Tobler  <andreast@gcc.gnu.org>
27581         * config.gcc (armv6*-*-freebsd*): Change the target_cpu_cname to
27582         arm1176jzf-s.
27584 2017-06-20  Jakub Jelinek  <jakub@redhat.com>
27586         * ira-costs.c (find_costs_and_classes): Initialize cost_classes later
27587         to make sure not to dereference a NULL cost_classes_ptr pointer.
27589 2017-06-20  Carl Love  <cel@us.ibm.com>
27591         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
27592         ALTIVEC_BUILTIN_VMULESW, ALTIVEC_BUILTIN_VMULEUW,
27593         ALTIVEC_BUILTIN_VMULOSW, ALTIVEC_BUILTIN_VMULOUW entries.
27594         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin,
27595         builtin_function_type): Add ALTIVEC_BUILTIN_* case statements.
27596         * config/rs6000/altivec.md (MVULEUW, VMULESW, VMULOUW,
27597         VMULOSW): New enum "unspec" values.
27598         (vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
27599         vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si,
27600         altivec_vmuleuw, altivec_vmulesw, altivec_vmulouw,
27601         altivec_vmulosw): New patterns.
27602         * config/rs6000/rs6000-builtin.def (VMLEUW, VMULESW, VMULOUW,
27603         VMULOSW): Add definitions.
27605 2017-06-20  Julia Koval  <julia.koval@intel.com>
27607         * config/i386/i386.c: Fix rounding expand for new pattern.
27608         * config/i386/subst.md: Fix pattern (parallel -> unspec).
27610 2017-06-20  James Greenhalgh  <james.greenhalgh@arm.com>
27612         * config/aarch64/aarch64-option-extensions.def (rcpc): New.
27613         * config/aarch64/aarch64.h (AARCH64_FL_RCPC): New.
27615 2017-06-20  James Greenhalgh  <james.greenhalgh@arm.com>
27617         * config/aarch64/aarch64-option-extensions.def (fp16): Fix expected
27618         feature string.
27620 2017-06-20  James Greenhalgh  <james.greenhalgh@arm.com>
27622         * config/aarch64/aarch64-cores.def: Rearrange to sort by
27623         architecture, then by implementer ID.
27624         * config/aarch64/aarch64-tune.md: Regenerate.
27626 2017-06-20  Richard Biener  <rguenther@suse.de>
27628         PR middle-end/81097
27629         * fold-const.c (split_tree): Fold to type before negating.
27631 2017-06-20  David Malcolm  <dmalcolm@redhat.com>
27633         * diagnostic-show-locus.c
27634         (selftest::test_fixit_deletion_affecting_newline): New function.
27635         (selftest::diagnostic_show_locus_c_tests): Call it.
27637 2017-06-20  Andreas Schwab  <schwab@suse.de>
27639         PR target/80970
27640         * config/m68k/m68k.md (bsetdreg, bchgdreg, bclrdreg): Use "=d"
27641         instead of "+d".
27643 2017-06-20  Prakhar Bahuguna  <prakhar.bahuguna@arm.com>
27645         * config/arm/arm-c.c (arm_cpu_builtins): New block to define
27646         __ARM_FEATURE_COPROC according to support.
27648 2017-06-20  Jakub Jelinek  <jakub@redhat.com>
27650         * tree-chkp.c (chkp_get_hard_register_var_fake_base_address):
27651         Rewritten to avoid overflow for > 32-bit pointers.
27653         PR sanitizer/81125
27654         * ubsan.h (ubsan_encode_value): Workaround buggy clang++ parser
27655         by removing enum keyword.
27656         (ubsan_type_descriptor): Likewise.  Formatting fix.
27658         PR target/81121
27659         * config/i386/i386.md (TARGET_USE_VECTOR_CONVERTS float si->{sf,df}
27660         splitter): Require TARGET_SSE2 in the condition.
27662 2017-06-20  Michael Meissner  <meissner@linux.vnet.ibm.com>
27664         PR target/79799
27665         * config/rs6000/rs6000.c (rs6000_expand_vector_init): Add support
27666         for doing vector set of SFmode on ISA 3.0.
27667         * config/rs6000/vsx.md (vsx_set_v4sf_p9): Likewise.
27668         (vsx_set_v4sf_p9_zero): Special case setting 0.0f to a V4SF
27669         element.
27670         (vsx_insert_extract_v4sf_p9): Add an optimization for inserting a
27671         SFmode value into a V4SF variable that was extracted from another
27672         V4SF variable without converting the element to double precision
27673         and back to single precision vector format.
27674         (vsx_insert_extract_v4sf_p9_2): Likewise.
27676 2017-06-19  Jakub Jelinek  <jakub@redhat.com>
27678         * tree-ssa-structalias.c (get_constraint_for_ptr_offset): Multiply
27679         in UWHI to avoid undefined overflow.
27681         PR sanitizer/81125
27682         * ubsan.h (enum ubsan_encode_value_phase): New.
27683         (ubsan_encode_value): Change second argument to
27684         enum ubsan_encode_value_phase with default value of
27685         UBSAN_ENCODE_VALUE_GENERIC.
27686         * ubsan.c (ubsan_encode_value): Change second argument to
27687         enum ubsan_encode_value_phase PHASE from bool IN_EXPAND_P,
27688         adjust uses, for UBSAN_ENCODE_VALUE_GENERIC use just
27689         create_tmp_var_raw instead of create_tmp_var and use a
27690         TARGET_EXPR.
27691         (ubsan_expand_bounds_ifn, ubsan_build_overflow_builtin,
27692         instrument_bool_enum_load, ubsan_instrument_float_cast): Adjust
27693         ubsan_encode_value callers.
27695         PR sanitizer/81111
27696         * ubsan.c (ubsan_encode_value): If current_function_decl is NULL,
27697         use create_tmp_var_raw instead of create_tmp_var, mark it addressable
27698         just by setting TREE_ADDRESSABLE on the result and use a TARGET_EXPR.
27700 2017-06-19  Richard Biener  <rguenther@suse.de>
27702         PR middle-end/81118
27703         * tree-cfgcleanup.c (cleanup_tree_cfg_noloop): Clear niter
27704         estimates if we changed anything.
27706 2017-06-19  Richard Biener  <rguenther@suse.de>
27708         PR tree-optimization/80887
27709         * tree-ssa-sccvn.c (mprts_hook_cnt): New global.
27710         (vn_lookup_simplify_result): Allow only mprts_hook_cnt succesful
27711         simplified lookups, then reset mprts_hook.
27712         (vn_nary_build_or_lookup_1): Set mprts_hook_cnt to 9 before
27713         simplifying.
27714         (try_to_simplify): Likewise.
27716 2017-06-19  Martin Liska  <mliska@suse.cz>
27718         PR sanitizer/80879
27719         * gimplify.c (gimplify_switch_expr):
27720         Initialize live_switch_vars for SWITCH_BODY == STATEMENT_LIST.
27722 2017-06-19  Martin Liska  <mliska@suse.cz>
27724         * doc/install.texi: Document that PGO runs in 4 stages.
27726 2017-06-19  Martin Liska  <mliska@suse.cz>
27728         PR ipa/80732
27729         * attribs.c (make_dispatcher_decl): Do not append '.ifunc'
27730         to dispatcher function name.
27731         * multiple_target.c (replace_function_decl): New function.
27732         (create_dispatcher_calls): Redirect both edges and references.
27734 2017-06-19  Jan Hubicka <hubicka@ucw.cz>
27736         * profile-count.c (profile_count::dump): Dump quality.
27737         (profile_count::differs_from_p): Update for unsigned val.
27738         * profile-count.h (profile_count_quality): New enum.
27739         (profile_count): Turn m_val to 62bit unsigned, add quality tracking.
27741 2017-06-19  Richard Biener  <rguenther@suse.de>
27743         * tree-ssa-loop-niter.h (estimate_numbers_of_iterations): Take
27744         struct function as arg.
27745         (estimate_numbers_of_iterations): Export overload with loop arg.
27746         (free_numbers_of_iterations_estimates_loop): Use an overload of
27747         free_numbers_of_iterations_estimates instead.
27748         * tree-cfg.c (remove_bb): Adjust.
27749         * tree-cfgcleanup.c (remove_forwarder_block_with_phi): Likewise.
27750         * tree-parloops.c (gen_parallel_loop): Likewise.
27751         * tree-ssa-loop-ivcanon.c (canonicalize_induction_variables):
27752         Likewise.
27753         (tree_unroll_loops_completely): Likewise.
27754         * tree-ssa-loop-niter.c (estimate_numbers_of_iterations_loop):
27755         Use an overload instead and export.
27756         (estimated_loop_iterations): Adjust.
27757         (max_loop_iterations): Likewise.
27758         (likely_max_loop_iterations): Likewise.
27759         (estimate_numbers_of_iterations): Take struct function as arg
27760         and adjust.
27761         (loop_exits_before_overflow): Adjust.
27762         (free_numbers_of_iterations_estimates_loop): Use an overload.
27763         * tree-vect-loop.c (vect_analyze_loop_form): Adjust.
27764         * tree-vectorizer.c (vect_free_loop_info_assumptions): Likewise.
27766 2017-06-19  Richard Biener  <rguenther@suse.de>
27768         PR ipa/81112
27769         * ipa-prop.c (find_constructor_constant_at_offset): Handle
27770         RANGE_EXPR conservatively.
27772 2017-06-16  Carl Love  <cel@us.ibm.com>
27774         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
27775         definitions for vec_float, vec_float2, vec_floato,
27776         vec_floate built-ins.
27777         * config/rs6000/vsx.md (define_c_enum "unspec"): Add RTL code
27778         for instructions vsx_xvcvsxws vsx_xvcvuxwsp, float2, floato and
27779         floate.
27780         * config/rs6000/rs6000-builtin.def (FLOAT2_V2DI, FLOATE_V2DF,
27781         FLOATE_2DI, FLOATO_V2DF, FLOATEE_V2DI, XVCVSXWSP_V4SF,
27782         UNS_FLOATO_V2DI, UNS_FLOATE_V2DI): Add definitions.
27783         * config/altivec.md (define_insn "p8_vmrgew_<mode>",
27784         define_mode_attr VF_sxddp): Add V4SF type to p8_vmrgew.
27785         * config/rs6000/altivec.h (vec_float, vec_float2, vec_floate,
27786         vec_floato): Add builtin defines.
27787         * doc/extend.texi (vec_float, vec_float2, vec_floate, vec_floato):
27788         Update the built-in documentation file for the new built-in
27789         functions.
27791 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27793         * config/arm/arm.opt (marm): Mark as the negative of of -mthumb.
27794         (mthumb): Mark as the negative of -marm.
27796 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27798         * doc/invoke.texi (ARM Options, -mcpu): Document supported
27799         extension options.
27800         (ARM Options, -mtune): Document that this accepts the same
27801         extension options as -mcpu.
27802         (ARM Options, -mfpu): Document addition of -mfpu=auto.
27804 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27806         * doc/invoke.texi (ARM Options, -march=): Document new syntax and
27807         permitted extensions.
27809 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27811         * config/arm/arm-cpus.in (armv7): Add extension +nofp.
27812         (armv7-r): Add aliases vfpv3xd and vfpv3-d16.
27813         (armv8-m.main): Add option +nodsp.
27814         * config/arm/arm-cpu-cdata.h: Regenerated.
27816 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27818         * config/arm/t-fuchsia: New file.
27819         * config.gcc (arm*-*-fuchsia*): Use it.
27821 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27823         * config/arm/t-symbian: Rewrite for new option infrastructure.
27825 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27827         * config/arm/t-phoenix (MULTILIB_REUSE): Clear variable.
27828         (MULTILIB_REQUIRED): Likewise.
27830 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27832         * config/arm/t-linux-eabi (MULTILIB_EXCEPTIONS): Set to empty.
27833         (MULTILIB_RESUE): Likewise.
27834         (MULTILIB_MATCHES): Likewise.
27835         (MULTLIB_REQUIRED): Likewise.
27837 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27839         * config/arm/t-rtems: Rewrite for new option framework.
27841 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27843         * config/arm/t-aprofile (v7_a_nosimd_variants, v7_a_simd_variants)
27844         (v7ve_nosimd_variatns, v7ve_vfpv3_simd_variants)
27845         (v7ve_vfpv4_simd_variants, v8_a_nosimd_variants, v8_a_simd_variants)
27846         (v8_1_a_simd_variants, v8_2_a_simd_variants): Move to ...
27847         * config/arm/t-multilib: ... here.
27848         (MULTILIB_OPTIONS): Add armv7 and armv7+fp architectures.
27849         (MULTILIB_MATCHES): Use armv7 libraries for armv7-r.  Also use for
27850         armv7-a and armv8*-a when A-profile libraries have not been built.
27851         * config/arm/t-rmprofile: Rewrite.
27853 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27855         * genmultilib (multilib_reuse): Allow an explicit period to be escaped
27856         with a backslash.  Remove the backslash after substituting unescaped
27857         periods.
27858         * doc/fragments.texi (MULTILIB_REUSE): Document it.
27860 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27862         * config.gcc: (arm*-*-*): When building a-profile libraries, force
27863         the driver to pass through the default setting of -mfloat-abi.
27864         * common/config/arm/arm-common.c (arm_target_thumb_only): Return -marm
27865         rather than NULL.
27866         * config/arm/t-multilib (MULTILIB_REUSE): Initialize to empty.
27867         (all_feat_combs): New rule.
27868         (MULTILIB_OPTIONS): Use explicit ARM and Thumb directories.  Rework
27869         default libraries.
27870         * config/arm/t-aprofile: Rewrite.
27872 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27874         * config/arm/arm.h (FPUTYPE_AUTO): Define.
27875         * config/arm/arm.c (arm_option_override): Use FPUTYPE_AUTO if the
27876         fpu is not specified by the user/command-line.
27877         * config/arm/bpabi.h (FPUTYPE_DEFAULT): Delete.
27878         * config/arm/netbsd-elf.h (FPUTYPE_DEFAULT): Delete.
27879         * config/arm/linux-elf.h (FPUTYPE_DEFAULT): Delete.
27880         * config/arm/vxworks.h (FPUTYPE_DEFAULT): Delete.
27881         * common/config/arm/arm-common.c (arm_canon_arch_option): Use
27882         FPUTYPE_AUTO insted of FPUTYPE_DEFAULT.
27884 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27886         * config/arm/elf.h (MULTILIB_DEFAULTS): Delete.
27887         * config/arm/t-arm-elf: Rewritten.
27889 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27891         * config/arm/arm.h (TARGET_HARD_FLOAT): Also check that we
27892         have some floating-point instructions.
27893         (TARGET_SOFT_FLOAT): Define as inverse of TARGET_HARD_FLOAT.
27894         (TARGET_MAYBE_HARD_FLOAT): New macro.
27895         * config/arm/arm-builtins.c (arm_init_builtins): Use
27896         TARGET_MAYBE_HARD_FLOAT.
27897         * config/arm/arm.c (arm_option_override): Use TARGET_HARD_FLOAT_ABI.
27899 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27901         * common/config/arm/arm-common.c: Define INCLUDE_LIST.
27902         (configargs.h): Include it.
27903         (arm_print_hint_for_fpu_option): New function.
27904         (arm_parse_fpu_option): New function.
27905         (candidate_extension): New class.
27906         (arm_canon_for_multilib): New function.
27907         * config/arm/arm.h (CANON_ARCH_SPEC_FUNCTION): New macro.
27908         (EXTRA_SPEC_FUNCTIONS): Add CANON_ARCH_SPEC_FUNCTION.
27909         (ARCH_CANONICAL_SPECS): New macro.
27910         (DRIVER_SELF_SPECS): Add ARCH_CANONICAL_SPECS.
27912 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27914         * config.gcc (arm*-*-*): Ensure both target_cpu_cname and with_cpu
27915         are set after handling multilib fragments.  Set target_cpu_default2
27916         from with_cpu.
27918 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27920         * config.gcc (arm*-*-fucshia*): Set target_cpu_cname to the real
27921         cpu name.
27922         (arm*-*-*): Set target_cpu_default2 to a quoted string.
27923         * config/arm/parsecpu.awk (check_cpu): Validate any extension
27924         options.
27925         (check_arch): Likewise.
27926         * config/arm/arm.c (arm_configure_build_target): Handle
27927         TARGET_CPU_DEFAULT being a string constant.  Scan any feature
27928         options in the default.
27930 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27932         * config/arm/arm-protos.h (cpu_arch_extension): Add field to record
27933         when an option is an alias of another.
27934         * config/arm/parsecpu.awk (optalias): New parser token.
27935         (gen_comm_data): Mark non-alias options as such.  Emit entries
27936         for extension aliases.
27937         * config/arm/arm-cpus.in (armv5e): Make vfpv2 an alias.
27938         (armv5te, armv5tej, armv6, armv6j, armv6k, armv6z): Likewise.
27939         (armv6kz, armv6zk, armv6t2): Likewise.
27940         (armv7): Make vfpv3-d16 an alias.
27941         (armv7-a): Make vfpv3-d16, neon and neon-vfpv3 aliases.  Sort in
27942         canonical order.
27943         (armv7ve): Make vfpv4-d16, neon-vfpv3 and neon-vfpv4 aliases.
27944         Sort in canonical order.
27945         (armv8-a): Sort in canonical order.
27946         (armv8.1-a, armv8.2-a):  Likewise.
27947         (generic-armv7-a): Make neon and neon-vfpv3 aliases.  Sort in
27948         canonical order.
27949         (cortex-a9): Sort in canonical order.
27950         * config/arm/arm.c (selftests.h): Include it.
27951         (arm_test_cpu_arch_data): New function.
27952         (arm_run_self_tests): New function.
27953         (TARGET_RUN_TARGET_SELFTESTS): Redefine.
27954         (targetm): Move declaration to the end of the file.
27955         * arm-cpu-cdata.h: Regenerated.
27957 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27959         * config/arm/arm.h (TARGET_MODE_SPECS): Add additional parameter to
27960         call to target_mode_check describing the type of option passed.
27961         * common/config/arm/arm-common.c (arm_arch_core_flag): Delete.
27962         (arm_target_thumb_only): Use arm_parse_arch_option_name or
27963         arm_parse_cpu_option_name to match parameters against list of
27964         available targets.
27965         * config/arm/parsecpu.awk (gen_comm_data): Don't generate
27966         arm_arch_core_flags data structure.
27967         * config/arm/arm-cpu_cdata.h: Regenerated.
27969 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27971         * common/config/arm/arm-common.c (arm_initialize_isa): Moved here from
27972         config/arm/arm.c.
27973         (arm_print_hint_for_cpu_option): Likewise.
27974         (arm_print_hint_for_arch_option): Likewise.
27975         (arm_parse_cpu_option_name): Likewise.
27976         (arm_parse_arch_option_name): Likewise.
27977         * config/arm/arm.c (arm_identify_fpu_from_isa): Use the computed number
27978         of entries in the all_fpus list.
27979         * config/arm/arm-protos.h (all_architectures, all_cores): Declare.
27980         (arm_parse_cpu_option_name): Declare.
27981         (arm_parse_arch_option_name): Declare.
27982         (arm_parse_option_features): Declare.
27983         (arm_intialize_isa): Declare.
27984         * config/arm/parsecpu.awk (gen_data): Move CPU and architecture
27985         data tables to ...
27986         (gen_comm_data): ... here.  Make definitions non-static.
27987         * config/arm/arm-cpu-data.h: Regenerated.
27988         * config/arm/arm-cpu-cdata.h: Regenerated.
27990 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27992         * config/arm/arm-protos.h (arm_build_target): Remove arch_core.
27993         (cpu_arch_extension): New structure.
27994         (cpu_arch_option, arch_option, cpu_option): New structures.
27995         * config/arm/parsecpu.awk (gen_headers): Build an enumeration of
27996         architecture types.
27997         (gen_data): Generate new format data tables.
27998         * config/arm/arm.c (cpu_tune): New structure.
27999         (cpu_option, processors): Delete.
28000         (arm_print_hint_for_core_or_arch): Delete.  Replace with ...
28001         (arm_print_hint_for_cpu_option): ... this and ...
28002         (arm_print_hint_for_arch_option): ... this.
28003         (arm_parse_arch_cpu_name): Delete.  Replace with ...
28004         (arm_parse_cpu_option_name): ... this and ...
28005         (arm_parse_arch_option_name): ... this.
28006         (arm_unrecognized_feature): Change type of target parameter to
28007         cpu_arch_option.
28008         (arm_parse_arch_cpu_features): Delete.  Replace with ...
28009         (arm_parse_option_features): ... this.
28010         (arm_configure_build_target): Rework to use new configuration data
28011         tables.
28012         (arm_print_tune_info): Rework for new configuration data tables.
28013         * config/arm/arm-cpu-data.h: Regenerated.
28014         * config/arm/arm-cpu.h: Regenerated.
28016 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
28018         * Makefile.in (OBJS): Move sbitmap.o from here ...
28019         (OBJS-libcommon): ... to here.
28021 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
28023         * config/arm/arm-isa.h (ISA_ALL_FPU_INTERNAL): Renamed from ISA_ALL_FPU.
28024         (ISA_ALL_CRYPTO): New macro.
28025         (ISA_ALL_SIMD): New macro
28026         (ISA_ALL_FP): New macro.
28027         * config/arm/arm.c (fpu_bitlist): Update initializer.
28028         * config/arm/arm-cpus.in: Use new ISA_ALL macros to disable crypto,
28029         simd or fp.
28030         (arm9e): Add fpu.  Add option for nofp
28031         (arm946e-s, arm966e-s, arm968e-s, arm10e, arm1020e, arm1022e): Likewise.
28032         (arm926ej-s, arm1026ej-s): Likewise.
28033         (generic-armv7-a): Add fpu.  Add options for simd, vfpv3, vfpv3-d16,
28034         vfpv3-fp16, vfpv3-d16-fp16, vfpv4, vfpv4-d16, neon, neon-vfp3,
28035         neon-fp16, neon-vfpv4, nofp and nosimd.
28036         (cortex-a5, cortex-a7): Add fpu.  Add options for nosimd and nofp.
28037         (cortex-a8): Add fpu.  Add option for nofp.
28038         (cortex-a9): Add fpu.  Add options for nosimd and nofp.
28039         (cortex-a12, cortex-a15, cortex-a17): Add fpu.  Add option for nofp.
28040         (cortex-r4f): Add fpu.
28041         (cortex-r5): Add fpu.  Add options for nofp.dp and nofp.
28042         (cortex-r7): Use idiv option from architecture.  Add fpu.  Add option
28043         for nofp.
28044         (cortex-r8): Likewise.
28045         (cortex-m4): Add fpu.  Add option for nofp.
28046         (cortex-a15.cortex-a7): Add fpu.  Add option for nofp.
28047         (cortex-a17.cortex-a7): Likewise.
28048         (cortex-a32): Add fpu.  Add options for crypto and nofp.
28049         (cortex-a35, cortex-a53): Likewise.
28050         (cortex-a57): Add fpu.  Add option for crypto.
28051         (cortex-a72, cortex-a73): Likewise.
28052         (exynos-m1): Likewise.
28053         (cortex-a57.cortex-a53, cortex-a72.cortex-a53): Likewise.
28054         (cortex-a73.cortex-a35, cortex-a73.cortex-a53): Likewise.
28055         (cortex-m33): Add fpu.  Add option for nofp.
28056         * config/arm/arm-cpu-cdata.h: Regenerated
28057         * config/arm/arm-cpu-data.h: Regenerated.
28059 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
28061         * arm-cpus.in (armv5e): Add options fp, vfpv2 and nofp.
28062         (armv5te, armv5tej): Likewise.
28063         (armv6, armv6j, armv6k, armv6z, armv6kz, armv6zk, armv6t2): Likewise.
28064         (armv7): Add options fp and vfpv3-d16.
28065         (armv7-a): Add options fp, simd, vfpv3, vfpv3-d16, vfpv3-d16-fp16,
28066         vfpv3-fp16, vfpv4, vfpv4-d16, neon, neon-vfpv3, neon-fp16, neon-vfpv4,
28067         nofp and nosimd.
28068         (armv7ve): Likewise.
28069         (armv7-r): Add options fp, fp.sp, idiv, nofp and noidiv.
28070         (armv7e-m): Add options fp, fpv5, fp.dp and nofp.
28071         (armv8-a): Add nocrypto option.
28072         (armv8.1-a, armv8.2-a): Likewise.
28073         (armv8-m.main): add options fp, fp.dp and nofp.
28075 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
28077         * config/arm/arm-cpus.in (armv8-a): Add options crc, simd crypto and
28078         nofp.
28079         (armv8-a+crc): Delete.
28080         (armv8.1-a): Add options simd, crypto and nofp.
28081         (armv8.2-a): Add options fp16, simd, crypto and nofp.
28082         (armv8.2-a+fp16): Delete.
28083         (armv8-m.main): Add option dsp.
28084         (armv8-m.main+dsp): Delete.
28085         (cortex-a8): Add fpu.  Add nofp option.
28086         (cortex-a9): Add fpu.  Add nofp and nosimd options.
28087         * config/arm/parsecpu.awk (gen_data): Generate option tables and
28088         link to main cpu and architecture data structures.
28089         (gen_comm_data): Only put isa attributes from the main architecture
28090         in common tables.
28091         (option): New statement for architecture and CPU entries.
28092         * arm.c (struct cpu_option): New structure.
28093         (struct processors): Add entry for options.
28094         (arm_unrecognized_feature): New function.
28095         (arm_parse_arch_cpu_name): Ignore any characters after the first
28096         '+' character.
28097         (arm_parse_arch_cpu_feature): New function.
28098         (arm_configure_build_target): Separate out any CPU and architecture
28099         features and parse separately.  Don't error out if -mfpu=auto is
28100         used with only an architecture string.
28101         (arm_print_asm_arch_directives): New function.
28102         (arm_file_start): Call it.
28103         * config/arm/arm-cpu-cdata.h: Regenerated.
28104         * config/arm/arm-cpu-data.h: Likewise.
28105         * config/arm/arm-tables.opt: Likewise.
28107 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
28109         * config/arm/elf.h (ASM_SPEC): Only pass -mfpu through to the
28110         assembler when it is not -mfpu=auto.
28112 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
28114         * config/arm/arm.h (BIG_LITTLE_SPEC): Delete macro.
28115         (ASM_REWRITE_SPEC_FUNCTIONS): New macro.
28116         (BIG_LITTLE_CPU_SPEC_FUNCTIONS): Delete macro.
28117         (ASM_CPU_SPEC): Rewrite.
28118         (MCPU_MTUNE_NATIVE_FUNCTIONS): New macro.
28119         (EXTRA_SPEC_FUNCTIONS): Move outside of ifdef.  Use
28120         MCPU_MTUNE_NATIVE_FUNCTIONS and ASM_REWRITE_SPEC_FUNCTIONS.  Remove
28121         reference to BIG_LITTLE_CPU_SPEC_FUNCTIONS.
28122         * common/config/arm/arm-common.c (arm_rewrite_selected_cpu): Ensure
28123         copied string is NUL-terminated.  Also strip any characters prefixed
28124         by '+'.
28125         (arm_rewrite_selected_arch): New function.
28126         (arm_rewrite_march): New function.
28128 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
28130         * config/arm/arm.opt (x_arm_arch_string): New TargetSave option.
28131         (x_arm_cpu_string, x_arm_tune_string): Likewise.
28132         (march, mcpu, mtune): Convert to string-based options.
28133         * config/arm/arm.c (arm_print_hint_for_core_or_arch): New function.
28134         (arm_parse_arch_cpu_name): New function.
28135         (arm_configure_build_target): Use arm_parse_arch_cpu_name to
28136         identify selected architecture or CPU.
28137         (arm_option_save): New function.
28138         (TARGET_OPTION_SAVE): Redefine.
28139         (arm_option_restore): Restore string options.
28140         (arm_option_print): Print string options.
28142 2017-06-16  Martin Sebor  <msebor@redhat.com>
28144         PR tree-optimization/80933
28145         PR tree-optimization/80934
28146         * builtins.c (fold_builtin_3): Do not handle bcmp here.
28147         * gimple-fold.c (gimple_fold_builtin_bcmp): New function.
28148         (gimple_fold_builtin_bcopy, gimple_fold_builtin_bzero): Likewise.
28149         (gimple_fold_builtin): Call them.
28151 2017-06-16  Jan Hubicka  <hubicka@ucw.cz>
28153         * gimple-ssa-isolate-paths.c (isolate_path): Set edge leading to path
28154         as unlikely; update profile.
28156 2017-06-16  Jan Hubicka  <hubicka@ucw.cz>
28158         * predict.c (force_edge_cold): Handle declaring edges impossible
28159         more aggresively.
28161 2017-06-16  Jan Hubicka  <hubicka@ucw.cz>
28163         * tree-ssa-loop-ivcanon.c (remove_exits_and_undefined_stmts): Update
28164         profile.
28165         (try_unroll_loop_completely): Fix reporting.
28167 2017-06-16  Jan Hubicka  <hubicka@ucw.cz>
28169         * tree-ssa-tail-merge.c (replace_block_by): Fix profile updating.
28171 2017-06-16  James Greenhalgh  <james.greenhalgh@arm.com>
28173         PR target/71778
28174         * config/arm/arm-builtins.c (arm_expand_builtin_args): Return TARGET
28175         if given a non-constant argument for an intrinsic which requires a
28176         constant.
28178 2017-06-16  Jan Hubicka  <hubicka@ucw.cz>
28180         * profile.c (compare_freqs): New function.
28181         (branch_prob): Sort edge list.
28182         (find_spanning_tree): Assume that the list is priority sorted.
28184 2017-06-16  Richard Biener  <rguenther@suse.de>
28186         PR tree-optimization/81090
28187         * passes.def (pass_record_bounds): Remove.
28188         * tree-pass.h (make_pass_record_bounds): Likewise.
28189         * tree-ssa-loop.c (pass_data_record_bounds, pass_record_bounds,
28190         make_pass_record_bounds): Likewise.
28191         * tree-ssa-loop-ivcanon.c (canonicalize_induction_variables): Do
28192         not free niter estimates at the beginning but at the end.
28193         * tree-scalar-evolution.c (scev_finalize): Free niter estimates.
28195 2017-06-16  Richard Biener  <rguenther@suse.de>
28197         * tree-switch-conversion.c (emit_case_bit_tests): Adjust
28198         initializer to workaround ICE in host GCC 4.8.
28200 2017-06-16  Jan Hubicka  <hubicka@ucw.cz>
28202         * ipa-inline-transform.c (update_noncloned_frequencies): Update also
28203         counts.
28204         (clone_inlined_nodes): Update.
28206 2017-06-16  Maxim Kuvyrkov  <maxim.kuvyrkov@linaro.org>
28208         * config/aarch64/aarch64.c (qdf24xx_prefetch_tune): Update
28209         prefetch settings, and enable prefetching by default at -O3.
28211 2017-06-16  Maxim Kuvyrkov  <maxim.kuvyrkov@linaro.org>
28213         * config/aarch64/aarch64.c (aarch64_override_options_internal):
28214         Set flag_prefetch_loop_arrays according to tuning data.
28216 2017-06-16  Maxim Kuvyrkov  <maxim.kuvyrkov@linaro.org>
28218         * config/aarch64/aarch64-protos.h (struct cpu_prefetch_tune):
28219         New tune structure.
28220         (struct tune_params): Use cpu_prefetch_tune instead of cache_line_size.
28221         [Unrelated to main purpose of the patch] Place the pointer field last
28222         to enable type checking errors when tune structure are wrongly merged.
28223         * config/aarch64/aarch64.c (generic_prefetch_tune,)
28224         (exynosm1_prefetch_tune, qdf24xx_prefetch_tune,)
28225         (thunderx2t99_prefetch_tune): New tune constants.
28226         (tune_params *_tunings): Update all tunings (no functional change).
28227         (aarch64_override_options_internal): Set PARAM_SIMULTANEOUS_PREFETCHES,
28228         PARAM_L1_CACHE_SIZE, PARAM_L1_CACHE_LINE_SIZE, and PARAM_L2_CACHE_SIZE
28229         from tunings structures.
28231 2017-06-16  Jakub Jelinek  <jakub@redhat.com>
28233         PR sanitizer/81094
28234         * ubsan.c (instrument_null): Add T argument, use it instead
28235         of computing it based on IS_LHS.
28236         (instrument_object_size): Likewise.
28237         (pass_ubsan::execute): Adjust instrument_null and
28238         instrument_object_size callers to pass gimple_get_lhs or
28239         gimple_assign_rhs1 result to it.  Use instrument_null instead of
28240         calling get_base_address and instrument_mem_ref.  Handle
28241         aggregate call arguments for object-size sanitization.
28243 2017-06-16  Yury Gribov  <tetra2005@gmail.com>
28245         PR tree-optimization/81089
28246         * tree-vrp.c (is_masked_range_test): Validate operands of
28247         subexpression.
28249 2017-06-15  Martin Sebor  <msebor@redhat.com>
28251         PR c++/80560
28252         * dumpfile.c (dump_register): Avoid calling memset to initialize
28253         a class with a default ctor.
28254         * gcc.c (struct compiler): Remove const qualification.
28255         * genattrtab.c (gen_insn_reserv): Replace memset with initialization.
28256         * hash-table.h: Ditto.
28257         * ipa-cp.c (allocate_and_init_ipcp_value): Replace memset with
28258           assignment.
28259         * ipa-prop.c (ipa_free_edge_args_substructures): Ditto.
28260         * omp-low.c (lower_omp_ordered_clauses): Replace memset with
28261         default ctor.
28262         * params.h (struct param_info): Make struct members non-const.
28263         * tree-switch-conversion.c (emit_case_bit_tests): Replace memset
28264         with default initialization.
28265         * vec.h (vec_copy_construct, vec_default_construct): New helper
28266         functions.
28267         (vec<T>::copy, vec<T>::splice, vec<T>::reserve): Replace memcpy
28268         with vec_copy_construct.
28269         (vect<T>::quick_grow_cleared): Replace memset with default ctor.
28270         (vect<T>::vec_safe_grow_cleared, vec_safe_grow_cleared): Same.
28271         * doc/invoke.texi (-Wclass-memaccess): Document.
28273 2017-06-15  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
28275         * emit-rtl.h (is_leaf): Update comment about local
28276         register allocator.
28278 2017-06-15  Jozef Lawrynowicz   <jozef.l@somniumtech.com>
28280         PR target/78818
28281         * config/msp430/msp430.c (msp430_data_attr): Check that it's possible
28282         for a variable to have a section before checking if the section has a
28283         name.
28284         Set section to.persistent if persistent attribute is set.
28285         Warn if .persistent attribute is used on an automatic variable.
28287 2017-06-15  Eric Botcazou  <ebotcazou@adacore.com>
28289         PR rtl-optimization/80474
28290         * reorg.c (update_block): Do not ignore instructions in a delay slot.
28292 2017-06-15  Segher Boessenkool  <segher@kernel.crashing.org>
28294         * config/rs6000/rs6000.md (add<mode>3): Use reg_or_subregno instead
28295         of REGNO.
28297 2017-06-14  Maciej W. Rozycki  <macro@imgtec.com>
28299         * config/mips/mips.md (MIPS16_T_REGNUM): Remove constant.
28300         (casesi): Emit bounds checking as RTL.
28301         (casesi_internal_mips16_<mode>): Remove bounds checking.
28303 2017-06-14  Max Filippov  <jcmvbkbc@gmail.com>
28305         * config/xtensa/xtensa.c (xtensa_option_override): Append
28306         MASK_CONST16 to target_flags in the absence of TARGET_L32R.
28307         (hwloop_optimize, hwloop_fail, hwloop_pattern_reg,
28308          xtensa_doloop_hooks): Define unconditionally.
28309         (xtensa_reorg_loops): Only call reorg_loops in the presence of
28310         TARGET_LOOPS.
28311         * config/xtensa/xtensa.h (TARGET_L32R): New definition.
28312         (TARGET_DEFAULT): Remove XCHAL_HAVE_L32R condition and account
28313         for it in xtensa_option_override.
28314         (HARD_FRAME_POINTER_IS_FRAME_POINTER,
28315          HARD_FRAME_POINTER_IS_ARG_POINTER): New definitions.
28317 2017-06-14  Boris Kolpackov  <boris@codesynthesis.com>
28319         * doc/cppopts.texi: Document '-' special value to -MF.
28321 2017-06-14  Wilco Dijkstra  <wdijkstr@arm.com>
28323         * config/arm/cortex-a53.md (cortex_a53_fpalu) Adjust latency.
28324         (cortex_a53_fconst): Likewise.
28325         (cortex_a53_fpmul): Likewise.
28326         (cortex_a53_f_load_64): Likewise.
28327         (cortex_a53_f_load_many): Likewise.
28328         (cortex_a53_advsimd_alu): Likewise.
28329         (cortex_a53_advsimd_alu_q): Likewise.
28330         (cortex_a53_advsimd_mul): Likewise.
28331         (cortex_a53_advsimd_mul_q): Likewise.
28332         (fpmac bypass): Add new bypass for fpmac-fpmac case.
28333         Add missing fmul, r2f_cvt and fconst cases.
28335 2017-06-14  Richard Biener  <rguenther@suse.de>
28337         PR middle-end/81088
28338         * fold-const.c (split_tree): Drop TREE_OVERFLOW flag from
28339         literal constants.
28340         (fold_binary_loc): When associating do not treat pre-existing
28341         TREE_OVERFLOW on literal constants as a reason to allow
28342         TREE_OVERFLOW on associated literal constants.
28344 2017-06-14  Eric Botcazou  <ebotcazou@adacore.com>
28346         * config/sparc/sparc.h (MASK_ISA): Add MASK_LEON and MASK_LEON3.
28347         (MASK_FEATURES): New macro.
28348         * config/sparc/sparc.c (sparc_option_override): Remove the special
28349         handling of -mfpu and generalize it to all MASK_FEATURES switches.
28351 2017-06-14  Eric Botcazou  <ebotcazou@adacore.com>
28353         * simplify-rtx.c (simplify_binary_operation_1) <UDIV>: Do not simplify
28354         a division of 0 if non-call exceptions are enabled.
28356 2017-06-14  Andrew Pinski  <apinski@cavium.com>
28357             Naveen H.S  <Naveen.Hurugalawadi@cavium.com>
28359         PR target/71663
28360         * config/aarch64/aarch64.c (aarch64_expand_vector_init):
28361         Improve vector initialization code gen for only variable case.
28363 2017-06-14  Eric Botcazou  <ebotcazou@adacore.com>
28365         * config/sparc/driver-sparc.c (cpu_names): Add SPARC-T5 entry.
28367 2017-06-14  Richard Biener  <rguenther@suse.de>
28369         PR tree-optimization/81083
28370         * tree-ssa-sccvn.c (vn_reference_lookup_3): Do not use abnormals
28371         as values.
28373 2017-06-13  Segher Boessenkool  <segher@kernel.crashing.org>
28375         * config/rs6000/rs6000.c: Update all comments that mentioned SPE.
28376         (rs6000_expand_builtin): Remove RS6000_BTC_EVSEL.
28377         * config/rs6000/rs6000.h (RS6000_BTC_EVSEL): Delete.
28378         * config/rs6000/vxworks.h (VXCPU_FOR_8548): Delete.  Adjust former use.
28379         * config/rs6000/vxworksae.h (VXCPU_FOR_8548): Delete.
28380         * config/rs6000/vxworksmils.h (VXCPU_FOR_8548): Delete.
28382 2017-06-13  Segher Boessenkool  <segher@kernel.crashing.org>
28384         * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete VECTOR_SPE.
28385         * config/rs6000/rs6000.c (rs6000_debug_vector_unit): Delete VECTOR_SPE.
28387 2017-06-13  Segher Boessenkool  <segher@kernel.crashing.org>
28389         * config/rs6000/rs6000.h (FIXED_SCRATCH): Delete.
28391 2017-06-13  Segher Boessenkool  <segher@kernel.crashing.org>
28393         * config/rs6000/t-rtems: Don't handle SPE.
28395 2017-06-13  Segher Boessenkool  <segher@kernel.crashing.org>
28397         * config/rs6000/t-linux: Don't handle SPE.
28399 2017-06-13  Segher Boessenkool  <segher@kernel.crashing.org>
28401         * config/rs6000/eabispe.h: Delete file.
28403 2017-06-13  Segher Boessenkool  <segher@kernel.crashing.org>
28405         * config/rs6000/t-spe: Delete file.
28407 2017-06-13  Segher Boessenkool  <segher@kernel.crashing.org>
28409         * config/rs6000/rs6000.c (SPE_CONST_OFFSET_OK): Delete.
28410         (rs6000_legitimate_offset_address_p): Return false for anything in
28411         V2SImode or V2SFmode.
28413 2017-06-13  Segher Boessenkool  <segher@kernel.crashing.org>
28415         * config/rs6000/rs6000-modes.def: Remove all 8-byte vector modes
28416         except V2SF and V2SI.  Rearrange the vector modes, and add comments.
28417         * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove V8QImode
28418         and V4HImode.
28419         (reg_offset_addressing_ok_p): Remove V4HImode and V1DImode.
28420         (rs6000_legitimate_offset_address_p): Ditto.
28421         (rs6000_emit_move): Ditto.
28422         (rs6000_init_builtins): Remove V4HI_type_node.
28424 2017-06-13  Martin Liska  <mliska@suse.cz>
28426         PR sanitize/78204
28427         * asan.c (asan_sanitize_stack_p): Use sanitize_flags_p.
28428         (gate_asan): Likewise.
28429         * asan.h (asan_no_sanitize_address_p): Remove the function.
28430         (sanitize_flags_p): New function.
28431         * builtins.def: Fix coding style.
28432         * common.opt: Use renamed enum value.
28433         * convert.c (convert_to_integer_1): Use sanitize_flags_p.
28434         * doc/extend.texi: Document no_sanitize attribute.
28435         * flag-types.h (enum sanitize_code): Rename SANITIZE_NONDEFAULT
28436         to SANITIZE_UNDEFINED_NONDEFAULT.
28437         * gcc.c (sanitize_spec_function): Use the renamed enum value.
28438         * gimple-fold.c (optimize_atomic_compare_exchange_p):
28439         Use sanitize_flags_p.
28440         * gimplify.c (gimplify_function_tree): Likewise.
28441         * ipa-inline.c (sanitize_attrs_match_for_inline_p): Likewise.
28442         * opts.c (parse_no_sanitize_attribute): New function.
28443         (common_handle_option): Use renamed enum value.
28444         * opts.h (parse_no_sanitize_attribute): Declare.
28445         * tree.c (sanitize_flags_p): New function.
28446         * tree.h: Declared here.
28447         * tsan.c: Use sanitize_flags_p.
28448         * ubsan.c (ubsan_expand_null_ifn): Likewise.
28449         (instrument_mem_ref): Likewise.
28450         (instrument_bool_enum_load): Likewise.
28451         (do_ubsan_in_current_function): Remove the function.
28452         (pass_ubsan::execute): Use sanitize_flags_p.
28453         * ubsan.h: Remove do_ubsan_in_current_function
28454         * tree-cfg.c (print_no_sanitize_attr_value): New function.
28455         (dump_function_to_file): Use it here.
28457 2017-06-13  Martin Jambor  <mjambor@suse.cz>
28459         PR tree-optimization/80803
28460         PR tree-optimization/81063
28461         * tree-sra.c (subtree_mark_written_and_enqueue): Move up in the file.
28462         (propagate_subaccesses_across_link): Enqueue subtree whenever
28463         necessary instead of relying on the caller.
28465 2017-06-13  Martin Jambor  <mjambor@suse.cz>
28467         * tree-sra.c (add_access_to_work_queue): Only enqueue accesses
28468         that have a first_link.
28469         (sort_and_splice_var_accesses): Do not check first_link before
28470         enquing.
28471         (subtree_mark_written_and_enqueue): Likewise.
28472         (propagate_all_subaccesses): Likewise and do not stop at first
28473         parent with a first_link.
28475 2017-06-13  Martin Jambor  <mjambor@suse.cz>
28477         * tree-sra.c (dump_access_tree_1): Fix accidental dumping to stderr
28478         instead of f.
28480 2017-06-13  Yury Gribov  <tetra2005@gmail.com>
28482         * match.pd: New pattern.
28484 2017-06-13  Yury Gribov  <tetra2005@gmail.com>
28486         * tree-vrp.c (is_masked_range_test): New function.
28487         (register_edge_assert_for): Determine ranges for
28488         some bit tests.
28490 2017-06-13  Yury Gribov  <tetra2005@gmail.com>
28492         PR tree-optimization/67328
28493         * fold-const.c (maskable_range_p): New function.
28494         (build_range_check): Generate bittests if possible.
28496 2017-06-13  Martin Liska  <mliska@suse.cz>
28498         * gimple-pretty-print.c (dump_probability): Add new argument.
28499         (dump_edge_probability): Dump both probability and count.
28500         (dump_gimple_label): Likewise.
28501         (dump_gimple_bb_header): Likewise.
28503 2017-06-13  Georg-Johann Lay  <avr@gjlay.de>
28505         PR target/81072
28506         * config/avr/avr-devices.c: Fix indentation.
28507         * config/avr/gen-avr-mmcu-specs.c: Dito.
28509 2017-06-13  Richard Biener  <rguenther@suse.de>
28511         * tree-vect-loop.c (vect_model_reduction_cost): Do not fail,
28512         instead get vector type from stmt_info.
28513         (vectorizable_reduction): Adjust.  Remove dead code.
28515 2017-06-13  Richard Biener  <rguenther@suse.de>
28517         PR middle-end/81065
28518         * fold-const.c (extract_muldiv_1): Remove bogus distribution
28519         case of C * (x * C2 + C3).
28520         (fold_addr_of_array_ref_difference): Properly fold index difference.
28522 2017-06-12  David S. Miller  <davem@davemloft.net>
28524         PR target/80968
28525         * config/sparc/sparc.md (return expander): Emit frame blockage if
28526         function uses alloca.
28528 2017-06-12  Richard Sandiford  <richard.sandiford@linaro.org>
28530         * combine.c (make_field_assignment): Check len rather than the mode
28531         precision when calling force_to_mode.
28533 2017-06-12  Georg-Johann Lay  <avr@gjlay.de>
28535         Support multilibs and devices that see flash in RAM address range.
28537         PR target/81072
28538         * config/avr/avr-arch.h (avr_arch_id) <ARCH_AVRXMEGA3>: New enum.
28539         (avr_mcu_t) <flash_pm_offset>: New field.
28540         (avr_device_specific_features) <AVR_ISA_RCALL>: New enum.
28541         * config/avr/avr.h (AVR_SHORT_CALLS): New define.
28542         (AVR_HAVE_JMP_CALL): Don't set if AVR_SHORT_CALLS.
28543         (AVR_TINY_PM_OFFSET): Remove macro.
28544         * config/avr/avr.opt (-mshort-calls): New option.
28545         * config/avr/gen-avr-mmcu-specs.c (print_mcu)
28546         [*self_spec]: Add / remove -mshort-calls depending on AVR_ISA_RCALL.
28547         * config/avr/avr-c.c (avr_cpu_cpp_builtins)
28548         <__AVR_SHORT_CALLS__>: Built-in define if AVR_SHORT_CALLS.
28549         <__AVR_HAVE_JMP_CALL__>: Use AVR_HAVE_JMP_CALL as condition
28550         instead of avr_arch->have_jmp_call.
28551         <__AVR_PM_BASE_ADDRESS__>: Built-in define if avr_arch->flash_pm_offset.
28552         [AVR_TINY] <__AVR_TINY_PM_BASE_ADDRESS__>: Use
28553         avr_arch->flash_pm_offset to define.
28554         * config/avr/avr-devices.c (avr_arch_types): Add initializers for
28555         new field flash_pm_offset.  Add entry for avrxmega3.
28556         (avr_texinfo): Add entry for avrxmega3.
28557         * config/avr/avr-mcus.def: Add entries for: avrxmega3,
28558         attiny212, attiny214,
28559         attiny412, attiny414, attiny416, attiny417,
28560         attiny814, attiny816, attiny817,
28561         attiny1614, attiny1616, attiny1617,
28562         attiny3214, attiny3216, attiny3217.
28563         * config/avr/avr.c (avr_assemble_integer)[AVR_TINY]: Use
28564         avr_arch->flash_pm_offset instead of AVR_TINY_PM_OFFSET.
28565         (avr_print_operand_address) [AVR_TINY]: Same.
28566         (avr_asm_init_sections) <readonly_data_section>: Only patch
28567         callback if avr_arch->flash_pm_offset = 0.
28568         (avr_asm_named_section) <avr_need_copy_data_p>: Skip setting it
28569         for rodata if avr_arch->flash_pm_offset != 0.
28570         (avr_encode_section_info) [AVR_TINY]: Adjust comment.
28571         * config/avr/genmultilib.awk (dir_rcall, opt_rcall): New vars.
28572         (opts) [AVR_ISA_RCALL]: Append opt_rcall.
28573         (m_options): Append opt_rcall.
28574         (m_dirnames): Append dir_rcall.
28575         * config/avr/t-multilib: Regenerate.
28577         * configure.ac [target=avr]: Check whether avrxmega3 default
28578         linker description file works as needed.
28579         * configure: Regenerate.
28580         * doc/avr-mmcu.texi: Regenerate.
28581         * doc/invoke.texi (AVR Options) <-mshort-calls>: Document it.
28582         <__AVR_ARCH__>: Document avrxmega3 and 103.
28583         <__AVR_HAVE_JMP_CALL__>: Adjust documentation.
28584         <__AVR_SHORT_CALLS__>: Document it.
28585         <__AVR_PM_BASE_ADDRESS__>: Document it.
28586         * doc/extend.texi (AVR Options) <-mshort-calls>: Document it.
28587         (AVR Variable Attributes) <progmem>: Document this is
28588         not needed for avrxmega3.
28589         (AVR Named Address Spaces) <__flash>: Dito.
28591 2017-06-12  Jan Hubicka  <hubicka@ucw.cz>
28593         * cgraph.c (cgraph_node::dump): Complain about profile insanities.
28595 2017-06-12  Doug Rupp  <rupp@adacore.com>
28597         * config.gcc (*-*-vxworks*): Set use_gcc_stdint to "provide".
28598         Append vxworks-stdint.h to the tm_file list.
28599         * config/vxworks-stdint.h: New file.
28601 2017-06-12  Martin Liska  <mliska@suse.cz>
28603         PR tree-optimization/81041
28604         * tree-profile.c (gimple_gen_ic_func_profiler):
28605         Create an extra BB in profile-generate
28606         (gimple_gen_time_profiler): Likewise.
28608 2017-06-12  Jakub Jelinek  <jakub@redhat.com>
28610         PR tree-optimization/81003
28611         * tree-ssa-reassoc.c (force_into_ssa_name): New function.
28612         (update_range_test): Use it instead of force_gimple_operand_gsi.
28614 2017-06-12  Richard Biener  <rguenther@suse.de>
28616         PR tree-optimization/81053
28617         * tree-vect-loop.c (vect_is_simple_reduction): Handle PHI
28618         with backedge value not defined in loop.  Simplify def stmt
28619         compute.
28621 2017-06-11  Tom de Vries  <tom@codesourcery.com>
28623         PR target/79939
28624         * config/nvptx/nvptx.c (nvptx_cannot_force_const_mem): New function.
28625         Return true.
28626         (TARGET_CANNOT_FORCE_CONST_MEM): Redefine to
28627         nvptx_cannot_force_const_mem.
28629 2017-06-10  Jan Hubicka  <hubicka@ucw.cz>
28631         * opts.c (finish_options): Move test for flag_split_stack after
28632         it has been initialized.
28634 2017-06-11  Jason Merrill  <jason@redhat.com>
28636         * tree.h (id_equal): New.
28637         * dwarf2out.c, hsa-gen.c, ipa-devirt.c, omp-expand.c,
28638         omp-simd-clone.c, read-rtl-function.c, tree-chkp.c, tree.c: Use it
28639         instead of strcmp of IDENTIFIER_POINTER.
28641 2017-06-10  Jan Hubicka  <hubicka@ucw.cz>
28643         * ipa-inline-transform.c: Include function.h, cfg.h and basic-block.h
28644         (mark_all_inlined_calls_cdtor): Fix formating.
28645         (inline_transform): Rescale profile before inlining.
28647 2017-06-10  Jan Hubicka  <hubicka@ucw.cz>
28649         * cgraph.h (cgraph_edge::clone): Update prototype.
28650         * cgraphclones.c (cgraph_edge::clone): Update profile scaling.
28651         (cgraph_node::create_clone): Update.
28652         (cgraph_node::create_version_clone): Update.
28653         * tree-inline.c (copy_bb): Update.
28654         (expand_call_inline): Update.
28656 2017-06-10  Segher Boessenkool  <segher@kernel.crashing.org>
28658         * config/rs6000/rs6000.c (emit_vrsave_prologue): New function,
28659         factored out from ...
28660         (rs6000_emit_prologue): ... here.
28662 2017-06-10  Segher Boessenkool  <segher@kernel.crashing.org>
28664         * config/rs6000/rs6000.c (emit_split_stack_prologue): New function,
28665         factored out from ...
28666         (rs6000_emit_prologue): ... here.
28668 2017-06-10  Jan Hubicka  <hubicka@ucw.cz>
28670         * predict.c (drop_profile): Also drop individual bb/edge and cgraph
28671         edge counts.
28672         (handle_missing_profiles): Fix computation of tp_first_run.
28673         (counts_to_freqs): Do not touch freqs when count is 0.
28675 2017-06-10  Jan Hubicka  <hubicka@ucw.cz>
28677         * cgraphbuild.c (cgraph_edge::rebuild_references): Do not touch
28678         profile.
28680 2017-06-10  Tom de Vries  <tom@codesourcery.com>
28682         * doc/sourcebuild.texi (Effective-Target Keywords, Environment
28683         attributes): Document signal effective target.
28685 2017-06-10  Tom de Vries  <tom@codesourcery.com>
28687         * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
28688         Document effective target stack_size.
28690 2017-06-09  David Malcolm  <dmalcolm@redhat.com>
28692         * diagnostic.c (diagnostic_report_diagnostic): Only add fixits
28693         to the edit_context if they can be auto-applied.
28695 2017-06-9  Ian Lance Taylor  <iant@golang.org>
28697         * opts.c (finish_options): If -fsplit-stack, disable implicit
28698         -forder-blocks-and-partition.
28699         * doc/invoke.texi (Optimize Options): Document that when using
28700         -fsplit-stack -forder-blocks-and-partition is not implicitly
28701         enabled.
28703 2017-06-09  Jan Hubicka  <hubicka@ucw.cz>
28705         * builtin-attrs.def (ATTR_NORETURN_NOTHROW_LEAF_COLD_LIST,
28706         ATTR_CONST_NORETURN_NOTHROW_LEAF_COLD_LIST,
28707         ATTR_TMPURE_NORETURN_NOTHROW_LEAF_COLD_LIST): New.
28708         * builtins.def (abort, trap, unreachable): Declare cold.
28709         * calls.c (flags_from_decl_or_type): Lookup ECF_COLD.
28710         * tree-core.h (ECF_COLD): New.
28711         * tree.c (set_call_expr_flags): Handle ECF_COLD.
28712         (build_common_builtin_nodes): Mark unreachable and abort as cold.
28714 2017-06-09  Jan Hubicka  <hubicka@ucw.cz>
28716         * predict.c (unlikely_executed_stmt_p): Cleanup.
28718 2017-06-09  Richard Biener  <rguenther@suse.de>
28720         * tree-ssa-loop-im.c (execute_sm): Do not force multi-threaded
28721         model if the ref is always written to.
28723 2017-06-09  Tamar Christina  <tamar.christina@arm.com>
28725         * config/aarch64/aarch64.md (lrint<GPF:mode><GPI:mode>2): New.
28727 2017-06-09  Tamar Christina  <tamar.christina@arm.com>
28729         * config/arm/arm.c (arm_rtx_costs_internal): Make sdiv more expensive
28730         than udiv.
28732 2017-06-09  Tom de Vries  <tom@codesourcery.com>
28734         PR target/80855
28735         * config/nvptx/nvptx.md (define_expand "mov<QHSDISDFM>"): Error out with
28736         "target cannot support label values" when encountering LABEL_REF.
28738 2017-06-09  Martin Liska  <mliska@suse.cz>
28740         * tree-profile.c (gimple_gen_ic_profiler): Update comment.
28741         (gimple_gen_ic_func_profiler): Emit direct comparison
28742         of __gcov_indirect_call_callee with NULL.
28743         (gimple_gen_time_profiler): Change probability from
28744         PROB_VERY_UNLIKELY to PROB_UNLIKELY.
28746 2017-06-09  Jan Hubicka  <hubicka@ucw.cz>
28748         * profile.c (edge_gcov_counts): Turn to pointer.
28749         (compute_branch_probabilities, compute_branch_probabilities): Update.
28750         (branch_prob): Do not clear edge_gcov_count.
28751         * profile.h (edge_gcov_counts): Turn to pointer.
28752         (edge_gcov_count): Update.
28754 2017-06-09  Jan Hubicka  <hubicka@ucw.cz>
28756         * gimple.h (gimple_check_failed): Mark cold.
28758 2017-06-09  Richard Biener  <rguenther@suse.de>
28760         PR tree-optimization/66623
28761         * tree-vect-loop.c (vect_is_simple_reduction): Cleanup,
28762         refactor check_reduction into two parts, properly computing
28763         whether we have to check reduction validity for outer loop
28764         vectorization.
28766 2017-06-09  Richard Biener  <rguenther@suse.de>
28768         PR tree-optimization/79483
28769         * graphite-scop-detection.c (order): New global.
28770         (get_order): Compute bb to order mapping that satisfies code
28771         generation constraints.
28772         (cmp_pbbs): New helper.
28773         (build_scops): Start domwalk at entry block, sort generated
28774         pbbs.
28776 2017-06-09  Richard Biener  <rguenther@suse.de>
28778         PR middle-end/81007
28779         * ipa-polymorphic-call.c
28780         (ipa_polymorphic_call_context::restrict_to_inner_class):
28781         Skip FIELD_DECLs with error_mark_node type.
28782         * passes.def (all_lowering_passes): Run pass_build_cgraph_edges
28783         last again.
28785 2017-06-09  Martin Liska  <mliska@suse.cz>
28787         * predict.c (struct branch_predictor): New struct.
28788         (test_prediction_value_range): New test.
28789         (predict_c_tests): New function.
28790         * selftest-run-tests.c (selftest::run_tests): Run the function.
28791         * selftest.h: Declare new tests.
28793 2017-06-09  Segher Boessenkool  <segher@kernel.crashing.org>
28795         PR target/80966
28796         * config/rs6000/rs6000.c (rs6000_emit_allocate_stack): Assert that
28797         gen_add3_insn did not fail.
28798         * config/rs6000/rs6000.md (add<mode>3): If asked to add a constant to
28799         r0, construct that number in a temporary reg and add that reg to r0.
28800         If asked to put the result in r0 as well, fail.
28802 2017-06-08  Will Schmidt  <will_schmidt@vnet.ibm.com>
28804         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add handling
28805         for early expansion of vec_eqv.
28807 2017-06-08  Jakub Jelinek  <jakub@redhat.com>
28809         PR middle-end/81005
28810         * ubsan.c (instrument_null): Avoid pointless code temporary.
28811         (pass_ubsan::execute): Instrument aggregate arguments of calls.
28813 2017-06-08  Uros Bizjak  <ubizjak@gmail.com>
28815         PR target/81015
28816         Revert:
28817         2016-12-14  Uros Bizjak  <ubizjak@gmail.com>
28819         PR target/59874
28820         * config/i386/i386.md (*ctzhi2): New insn_and_split pattern.
28821         (*clzhi2): Ditto.
28823 2017-06-08  Jan Hubicka  <hubicka@ucw.cz>
28825         * predict.c (unlikely_executed_edge_p): Move ahead.
28826         (probably_never_executed_edge_p): Use it.
28828 2017-06-08  Alexander Ivchenko  <alexander.ivchenko@intel.com>
28830         PR middle-end/79988
28831         * tree-chkp.c (chkp_gimple_call_builtin_p): Remove
28832         gimple_call_builtin_p call.
28834 2017-06-08  Jan Hubicka  <hubicka@ucw.cz>
28836         * system.h (fancy_abort): Annotate by ATTRIBUTE_COLD.
28837         * rtl.h (rtl_check_failed_bounds, rtl_check_failed_type1,
28838         rtl_check_failed_type2, rtl_check_failed_code1,
28839         rtl_check_failed_code2, rtl_check_failed_code_mode,
28840         rtl_check_failed_block_symbol, cwi_check_failed_bounds,
28841         rtvec_check_failed_bounds, rtl_check_failed_flag,
28842         _fatal_insn_not_found, _fatal_insn): Likewise.
28843         * tree.h (tree_contains_struct_check_failed,
28844         tree_check_failed, tree_not_check_failed,
28845         tree_class_check_failed, tree_range_check_failed,
28846         tree_not_class_check_failed, tree_int_cst_elt_check_failed,
28847         tree_vec_elt_check_failed, phi_node_elt_check_failed,
28848         tree_operand_check_failed, omp_clause_check_failed,
28849         omp_clause_operand_check_failed, omp_clause_range_check_failed):
28850         Likewise.
28852 2017-06-08  Jan Hubicka  <hubicka@ucw.cz>
28854         * cgraph.c (cgraph_edge::maybe_hot_p): Do not check
28855         flag_branch_probabilities.
28856         * ipa-inline.c (edge_badness): Likewise.
28857         * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
28858         * postreload-gcse.c (eliminate_partially_redundant_load): Likewise.
28859         * predict.c (maybe_hot_frequency_p): Likewise.
28860         (probably_never_executed): Likewise.
28861         * sched-ebb.c (schedule_ebbs): Likewise.
28862         * sched-rgn.c (find_single_block_region): Likewise.
28863         * tracer.c (tail_duplicate): Likewise.
28865 2017-06-08  Jan Hubicka  <hubicka@ucw.cz>
28867         * opts.c (finish_options): x_flag_reorder_blocks_and_partition no
28868         longer requires x_flag_profile_use.
28870 2017-06-08  Jan Hubicka  <hubicka@ucw.cz>
28872         * cfgrtl.c (cfg_layout_initialize): Check crtl->has_bb_partition
28873         instead of flag_reorder_blocks_and_partition.
28874         * dbxout.c (dbxout_function_end): Likewise.
28875         * dwarf2out.c (gen_subprogram_die): Likewise.
28876         * haifa-sched.c (sched_create_recovery_edges): Likewise.
28877         * hw-doloop.c (reorg_loops): Likewise.
28878         * varasm.c (assemble_start_function,
28879         assemble_end_function): Likewise.
28880         (decide_function_section): Do not check for
28881         flag_reorder_blocks_and_partition.
28883 2017-06-08  Alexander Ivchenko  <alexander.ivchenko@intel.com>
28885         * tree-chkp.c (chkp_get_hard_register_var_fake_base_address):
28886         New function.
28887         (chkp_get_hard_register_fake_addr_expr): Ditto.
28888         (chkp_build_addr_expr): Add check for hard reg case.
28889         (chkp_parse_array_and_component_ref): Ditto.
28890         (chkp_find_bounds_1): Ditto.
28891         (chkp_process_stmt): Don't generate bounds store for
28892         hard reg case.
28894 2017-06-08  Jan Hubicka  <hubicka@ucw.cz>
28896         * predict.c (maybe_hot_bb_p): Do not check profile status.
28897         (maybe_hot_edge_p): Likewise.
28898         (probably_never_executed): Check for zero counts even if profile
28899         is not read.
28900         (unlikely_executed_edge_p): New function.
28901         (unlikely_executed_stmt_p): New function.
28902         (unlikely_executed_bb_p): New function.
28903         (set_even_probabilities): Use unlikely predicates.
28904         (combine_predictions_for_bb): Likewise.
28905         (predict_paths_for_bb): Likewise.
28906         (predict_paths_leading_to_edge): Likewise.
28907         (determine_unlikely_bbs): New function.
28908         (estimate_bb_frequencies): Use it.
28909         (compute_function_frequency): Use zero counts even if profile is
28910         not read.
28911         * profile-count.h: Fix typo.
28913 2017-08-08  Julia Koval  <julia.koval@intel.com>
28915         * config/i386/avx512bwintrin.h (_mm512_mask_cvtepi16_storeu_epi8,
28916         _mm512_mask_cvtsepi16_storeu_epi8,
28917         _mm512_mask_cvtusepi16_storeu_epi8): New intrinsics.
28918         * config/i386/avx512vlbwintrin.h (_mm256_mask_cvtepi16_storeu_epi8,
28919         _mm_mask_cvtsepi16_storeu_epi8, _mm256_mask_cvtsepi16_storeu_epi8,
28920         _mm_mask_cvtusepi16_storeu_epi8, _mm256_mask_cvtusepi16_storeu_epi8,
28921         _mm_mask_cvtepi16_storeu_epi8): New intrinsics.
28922         * config/i386/i386-builtin-types.def (PV8Q, V8QI): New pointer type.
28923         (VOID_FTYPE_PV32QI_V32HI_USI, VOID_FTYPE_PV8QI_V8HI_UQI,
28924         VOID_FTYPE_PV16QI_V16HI_UHI): New function types.
28925         * config/i386/i386-builtin.def (__builtin_ia32_pmovwb128mem_mask,
28926         __builtin_ia32_pmovwb256mem_mask, __builtin_ia32_pmovswb128mem_mask,
28927         __builtin_ia32_pmovswb256mem_mask, __builtin_ia32_pmovuswb128mem_mask,
28928         __builtin_ia32_pmovuswb256mem_mask,
28929         __builtin_ia32_pmovuswb512mem_mask, __builtin_ia32_pmovswb512mem_mask)
28930         __builtin_ia32_pmovwb512mem_mask): New builtins.
28932 2017-08-08  Julia Koval  <julia.koval@intel.com>
28934         PR target/73350,80862
28935         * config/i386/subst.md (round): Fix round pattern.
28936         * config/i386/i386.c (ix86_erase_embedded_rounding):
28937         Fix erasing rounding for the fixed pattern.
28939 2017-06-08  Jan Hubicka  <hubicka@ucw.cz>
28941         * cfgbuild.c (find_many_sub_basic_blocks): Fix thinko.
28943 2017-06-08  Martin Liska  <mliska@suse.cz>
28945         PR gcov-profile/80911
28946         * gcov.c (block_info::block_info): New constructor.
28948 2017-06-07  Carl Love  <cel@us.ibm.com>
28950         * config/rs6000/rs6000-c: The return type of the following
28951         built-in functions was implemented as int not long long.  Fix sign
28952         of return value for the unsigned version of vec_mulo and vec_mule.
28953         vector unsigned long long vec_bperm (vector unsigned long long,
28954                                              vector unsigned char)
28955         vector signed long long vec_mule (vector signed int,
28956                                           vector signed int)
28957         vector unsigned long long vec_mule (vector unsigned int,
28958                                             vector unsigned int)
28959         vector signed long long vec_mulo (vector signed int,
28960                                           vector signed int)
28961         vector unsigned long long vec_mulo (vector unsigned int,
28962                                             vector unsigned int)
28963         * doc/extend.texi: Fix the documentation for the built-in
28964         functions.
28966 2017-06-07  Carl Love  <cel@us.ibm.com>
28968         PR target/80982
28969         * config/rs6000/altivec.md (double<mode>2): Fix the implementation of
28970         for BE.
28972 2017-06-07  Carl Love  <cel@us.ibm.com>
28974         * config/rs6000/altivec.md: Fix argument swizzle in vec_doublel
28975         support, Generate       doublehv for signed int/float for BE case only.
28977 2017-06-07  Alexander Monakov  <amonakov@ispras.ru>
28979         * doc/invoke.texi (mcx16): Rewrite.
28981 2017-06-07  Segher Boessenkool  <segher@kernel.crashing.org>
28983         * config/rs6000/predicates.md (rs6000_nonimmediate_operand): Delete.
28984         * config/rs6000/rs6000.md (*movsi_internal1, movsi_from_sf,
28985         *mov<mode>_softfloat, and an anonymous splitter): Use
28986         nonimmediate_operand instead of rs6000_nonimmediate_operand.
28988 2017-06-07  Segher Boessenkool  <segher@kernel.crashing.org>
28990         * config/rs6000/darwin.h (REGISTER_NAMES): Delete the SPE_ACC and
28991         SPEFSCR registers.
28992         * config/rs6000/rs6000.c (rs6000_reg_names, alt_reg_names): Ditto.
28993         (enum rs6000_reg_type): Delete SPE_ACC_TYPE and SPEFSCR_REG_TYPE.
28994         (rs6000_debug_reg_global): Adjust.
28995         (rs6000_init_hard_regno_mode_ok): Adjust.
28996         (rs6000_dbx_register_number): Adjust.
28997         * config/rs6000/rs6000.h (FIRST_PSEUDO_REGISTER): Change to 115.
28998         (FIXED_REGISTERS, CALL_USED_REGISTERS, CALL_REALLY_USED_REGISTERS):
28999         Remove SPE_ACC and SPEFSCR.
29000         (REG_ALLOC_ORDER): Ditto.
29001         (FRAME_POINTER_REGNUM): Change to 111.
29002         (enum reg_class): Remove the SPE_ACC and SPEFSCR registers.
29003         (REG_CLASS_NAMES): Ditto.
29004         (REG_CLASS_CONTENTS): Delete the SPE_ACC and SPEFSCR registers.
29005         (REGISTER_NAMES): Ditto.
29006         (ADDITIONAL_REG_NAMES): Ditto.
29007         (rs6000_reg_names): Ditto.
29008         * config/rs6000/rs6000.md: Renumber some register number
29009         define_constants.
29011 2017-06-07  Segher Boessenkool  <segher@kernel.crashing.org>
29013         * config/rs6000/darwin.h (REGISTER_NAMES): Delete the SPE high
29014         registers.
29015         * config/rs6000/rs6000.c (rs6000_reg_names, alt_reg_names): Ditto.
29016         * config/rs6000/rs6000.h (FIRST_PSEUDO_REGISTER): Change from 149
29017         to 117.
29018         (DWARF_REG_TO_UNWIND_COLUMN): Do not define.
29019         (FIXED_REGISTERS, CALL_USED_REGISTERS, CALL_REALLY_USED_REGISTERS):
29020         Delete the SPE high registers.
29021         (REG_ALLOC_ORDER): Ditto.
29022         (enum reg_class): Remove SPE_HIGH_REGS.
29023         (REG_CLASS_NAMES): Ditto.
29024         (REG_CLASS_CONTENTS): Delete the SPE high registers.
29025         (REGISTER_NAMES): Ditto.
29026         (rs6000_reg_names): Ditto.
29027         * doc/tm.texi.in: Remove SPE as example.
29028         * doc/tm.texi: Regenerate.
29030 2017-06-07  Segher Boessenkool  <segher@kernel.crashing.org>
29032         * config/rs6000/8540.md (ppc8540_brinc): Delete.
29033         * config/rs6000/e500mc.md (e500mc_brinc): Delete.
29034         * config/rs6000/e500mc64.md (e500mc64_brinc): Delete.
29035         * config/rs6000/rs6000.md (type): Remove "brinc".
29037 2017-06-07  Segher Boessenkool  <segher@kernel.crashing.org>
29039         * config.gcc (powerpc*-*-*): Don't add spe.h to extra_headers.
29040         (powerpc*-linux*spe*): Use ${cpu_type} instead of rs6000.
29041         * config/rs6000/linuxspe.h: Delete file.
29042         * config/rs6000/rs6000.md: Don't include spe.md.
29043         * config/rs6000/spe.h: Delete file.
29044         * config/rs6000/spe.md: Delete file.
29045         * config/rs6000/t-rs6000: Remove spe.md.
29047 2017-06-07  Segher Boessenkool  <segher@kernel.crashing.org>
29049         * config/rs6000/predicates.md (reg_or_mem_operand): Reformat.
29050         (reg_or_none500mem_operand): Delete.
29051         * config/rs6000/rs6000.md (extendsfdf2): Use reg_or_mem_operand
29052         instead of reg_or_none500mem_operand.
29054 2017-06-07  Segher Boessenkool  <segher@kernel.crashing.org>
29056         * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
29057         handling of SPE flags.
29058         * config/rs6000/rs6000.opt (-mspe, -mspe=no, -mspe=yes): Delete.
29060 2017-06-07  Segher Boessenkool  <segher@kernel.crashing.org>
29062         * config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
29063         SPE ABI handling.
29064         * config/rs6000/paired.md (paired_negv2sf2): Rename to negv2sf2.
29065         (paired_absv2sf2, paired_addv2sf3, paired_subv2sf3, paired_mulv2sf3,
29066         paired_divv2sf3): Similar.
29067         * config/rs6000/predicates.md: Replace TARGET_SPE, TARGET_SPE_ABI,
29068         SPE_VECTOR_MODE and SPE_HIGH_REGNO_P by 0; simplify.
29069         * config/rs6000/rs6000-builtin.def: Delete RS6000_BUILTIN_E and
29070         RS6000_BUILTIN_S.
29071         Delete BU_SPE_1, BU_SPE_2, BU_SPE_3, BU_SPE_E, BU_SPE_P, and BU_SPE_X.
29072         Rename the paired_* instruction patterns.
29073         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Do not
29074         define __SPE__.
29075         * config/rs6000/rs6000-protos.h (invalid_e500_subreg): Delete.
29076         * config/rs6000/rs6000.c: Delete RS6000_BUILTIN_E and RS6000_BUILTIN_S.
29077         (struct rs6000_stack): Delete fields spe_gp_save_offset, spe_gp_size,
29078         spe_padding_size, and spe_64bit_regs_used.  Replace TARGET_SPE and
29079         TARGET_SPE_ABI with 0, simplify.  Replace SPE_VECTOR_MODE with
29080         PAIRED_VECTOR_MODE.
29081         (struct machine_function): Delete field spe_insn_chain_scanned_p.
29082         (spe_func_has_64bit_regs_p): Delete.
29083         (spe_expand_predicate_builtin): Delete.
29084         (spe_expand_evsel_builtin): Delete.
29085         (TARGET_DWARF_REGISTER_SPAN): Do not define.
29086         (TARGET_MEMBER_TYPE_FORCES_BLK): Do not define.
29087         (invalid_e500_subreg): Delete.
29088         (rs6000_legitimize_address): Always force_reg op2 as well, for
29089         paired single memory accesses.
29090         (rs6000_member_type_forces_blk): Delete.
29091         (rs6000_spe_function_arg): Delete.
29092         (rs6000_expand_unop_builtin): Delete SPE handling.
29093         (rs6000_expand_binop_builtin): Ditto.
29094         (spe_expand_stv_builtin): Delete.
29095         (bdesc_2arg_spe): Delete.
29096         (spe_expand_builtin): Delete.
29097         (spe_expand_predicate_builtin): Delete.
29098         (spe_expand_evsel_builtin): Delete.
29099         (rs6000_invalid_builtin): Remove RS6000_BTM_SPE handling.
29100         (spe_init_builtins): Delete.
29101         (spe_func_has_64bit_regs_p): Delete.
29102         (savres_routine_name): Delete "info" parameter.  Adjust callers.
29103         (rs6000_emit_stack_reset): Ditto.
29104         (rs6000_dwarf_register_span): Delete.
29105         * config/rs6000/rs6000.h (TARGET_SPE_ABI, TARGET_SPE,
29106         UNITS_PER_SPE_WORD, SPE_HIGH_REGNO_P, SPE_SIMD_REGNO_P,
29107         SPE_VECTOR_MODE, RS6000_BTM_SPE, RS6000_BUILTIN_E, RS6000_BUILTIN_S):
29108         Delete.
29109         * config/rs6000/rs6000.md (FIRST_SPE_HIGH_REGNO, LAST_SPE_HIGH_REGNO):
29110         Delete.
29111         * config/rs6000/rs6000.opt (-mabi=spe, -mabi=no-spe): Delete.
29112         * config/rs6000/spe.md: Delete every pattern that uses TARGET_SPE.
29113         * config/rs6000/vector.md (absv2sf2, negv2sf2, addv2sf3, subv2sf3,
29114         mulv2sf3, divv2sf3): Delete expanders.
29116 2017-06-07  Segher Boessenkool  <segher@kernel.crashing.org>
29118         config/rs6000/rs6000.md (UNSPEC_MV_CR_GT): Delete.
29120 2017-06-07  Segher Boessenkool  <segher@kernel.crashing.org>
29122         * config/rs6000/rs6000-protos.h (output_e500_flip_gt_bit): Delete.
29123         * config/rs6000/rs6000.c: Ditto.
29125 2017-06-07  Segher Boessenkool  <segher@kernel.crashing.org>
29127         * config/rs6000/predicated.md (rs6000_cbranch_operator): Delete.
29128         * config/rs6000/rs6000.md: Replace rs6000_cbranch_operator by
29129         comparison_operator.
29131 2017-06-07  Segher Boessenkool  <segher@kernel.crashing.org>
29133         * config/rs6000/rs6000.c: Remove everything related to -mfloat-gprs.
29134         * config/rs6000/rs6000.opt: Ditto.
29135         * config/rs6000/t-rtems: Ditto.
29137 2017-06-07  Segher Boessenkool  <segher@kernel.crashing.org>
29139         * config/rs6000/predicates.md: Replace TARGET_E500_DOUBLE and
29140         TARGET_E500_SINGLE by 0, simplify.
29141         * config/rs6000/rs6000.c: Ditto.
29142         (rs6000_option_override_internal): Delete CHECK_E500_OPTIONS.
29143         (spe_build_register_parallel): Delete.
29144         * config/rs6000/rs6000.h: Delete TARGET_E500_SINGLE,
29145         TARGET_E500_DOUBLE, and CHECK_E500_OPTIONS.
29146         * config/rs6000/rs6000.md: Replace TARGET_E500_DOUBLE,
29147         TARGET_E500_SINGLE, and <E500_CONVERT> by 0, simplify.
29148         (E500_CONVERT): Delete.
29149         * config/rs6000/spe.md: Remove many patterns and all define_constants.
29151 2017-06-07  Segher Boessenkool  <segher@kernel.crashing.org>
29153         * config/rs6000/darwin.md: Replace TARGET_FPRS by 1 and simplify.
29154         * config/rs6000/dfp.md: Ditto.
29155         (negdd2, *negdd2_fpr): Merge.
29156         (absdd2, *absdd2_fpr): Merge.
29157         (negtd2, *negtd2_fpr): Merge.
29158         (abstd2, *abstd2_fpr): Merge.
29159         * config/rs6000/e500.h: Delete file.
29160         * config/rs6000/predicates.md (rs6000_cbranch_operator): Replace
29161         TARGET_FPRS by 1 and simplify.
29162         * config/rs6000/rs6000-c.c: Ditto.
29163         * config/rs6000/rs6000.c: Ditto.  Also replace TARGET_SF_SPE and
29164         TARGET_DF_SPE by 0.
29165         * config/rs6000/rs6000.h: Ditto.  Delete TARGET_SF_SPE and
29166         TARGET_DF_SPE.
29167         * config/rs6000/rs6000.md: Ditto.
29168         (floatdidf2, *floatdidf2_fpr): Merge.
29169         (move_from_CR_gt_bit): Delete.
29170         * config/rs6000/spe.md: Replace TARGET_FPRS by 1 and simplify.
29171         (E500_CR_IOR_COMPARE): Delete.
29172         (All patterns that require !TARGET_FPRS): Delete.
29173         * config/rs6000/vsx.md: Replace TARGET_FPRS by 1 and simplify.
29175 2017-06-07  Bin Cheng  <bin.cheng@arm.com>
29177         * passes.def (pass_iv_canon): Move before pass_loop_distribution.
29179 2017-06-07  Bin Cheng  <bin.cheng@arm.com>
29181         * graphds.c (add_edge): Intitialize edge's attached data.
29182         (foll_in_subgraph, dfs_fst_edge, dfs_next_edge): New function
29183         pointer parameter.  Call pointed function on each edge during
29184         graph traversing.  Skip traversing the edge when the function
29185         returns true.
29186         (graphds_dfs, graphds_scc): Ditto.
29187         (for_each_edge): New parameter.  Pass the new parameter to callback
29188         function.
29189         * graphds.h (skip_edge_callback): New function pointer type.
29190         (graphds_dfs, graphds_scc): New function pointer parameter.
29191         (graphds_edge_callback, for_each_edge): New parameter.
29193 2017-06-07  Bin Cheng  <bin.cheng@arm.com>
29195         * tree-vect-data-refs.c (vect_mark_for_runtime_alias_test): Factor
29196         out code checking if runtime alias check is possible to below ...
29197         Call the new function.
29198         * tree-data-ref.c (runtime_alias_check_p): ... to new function.
29199         * tree-data-ref.h (runtime_alias_check_p): New decalaration.
29201 2017-06-07  Marek Polacek  <polacek@redhat.com>
29203         PR sanitizer/80932
29204         * fold-const.c (extract_muldiv_1) <case MINUS_EXPR>: Add
29205         TYPE_OVERFLOW_WRAPS check.
29207 2017-06-07  Bin Cheng  <bin.cheng@arm.com>
29209         * tree-vect-loop-manip.c (vect_do_peeling): Don't skip vector loop
29210         if versioning is required.
29211         * tree-vect-loop.c (vect_analyze_loop_2): Merge niter check for loop
29212         peeling with the check for versioning.
29214 2017-06-07  Bin Cheng  <bin.cheng@arm.com>
29216         * tree-vectorizer.h (vect_build_loop_niters): New parameter.
29217         * tree-vect-loop-manip.c (vect_build_loop_niters): New parameter.
29218         Set true to new parameter if new ssa variable is defined.
29219         (vect_gen_vector_loop_niters): Refactor.  Set range information
29220         for the new vector loop bound variable.
29221         (vect_do_peeling): Ditto.
29223 2017-06-07  Bin Cheng  <bin.cheng@arm.com>
29225         * tree-affine.c (ssa.h): Include header file.
29226         (tree_to_aff_combination): Handle (T1)(X - CST) when inner type
29227         has wrapping overflow behavior.
29229 2017-06-07  Bin Cheng  <bin.cheng@arm.com>
29231         * tree-affine.c (tree_to_aff_combination): Handle (T1)(X + X).
29233 2017-06-07  Bin Cheng  <bin.cheng@arm.com>
29235         (aff_combination_expand): Move (T1)(X *+- CST) simplification to ...
29236         (tree_to_aff_combination): ... here.
29238 2017-06-07  Bin Cheng  <bin.cheng@arm.com>
29240         * tree-ssa-loop-ivopts.c (ivopts_estimate_reg_pressure): New
29241         reg_pressure model function.
29242         (ivopts_global_cost_for_size): Delete.
29243         (determine_set_costs, iv_ca_recount_cost): Call new model function
29244         ivopts_estimate_reg_pressure.
29246 2017-06-07  Tamar Christina  <tamar.christina@arm.com>
29248         * config/aarch64/aarch64.c (aarch64_rtx_costs): Make sdiv more
29249         expensive than udiv.  Remove floating point cases from mod.
29251 2017-06-07  Tamar Christina  <tamar.christina@arm.com>
29253         * config/arm/aarch-cost-tables.h (cortexa53_extra_cost):
29254         Increase idiv cost.
29256 2017-06-07  Tamar Christina  <tamar.christina@arm.com>
29258         * config/aarch64/aarch64.md
29259         (copysignsf3): Fix mask generation.
29261 2017-06-07  Jakub Jelinek  <jakub@redhat.com>
29263         * dumpfile.h (enum tree_dump_index): Rename TDI_generic to
29264         TDI_gimple.
29265         (class dump_manager): Add register_dumps method.
29266         * dumpfile.c: Include langhooks.h.
29267         (dump_files): Use 0 instead of 3/4/5 for TDI_{original,gimple,nested}.
29268         (FIRST_AUTO_NUMBERED_DUMP): Decrease to 1.
29269         (FIRST_ME_AUTO_NUMBERED_DUMP): Define.
29270         (dump_manager::dump_register): Start with 512 entries instead of 32.
29271         (dump_manager::register_dumps): New method.
29272         * toplev.c (general_init): Instead of invoking register_dumps
29273         langhook, invoke register_dumps method on the dump manager.
29274         * gimplify.c (gimplify_function_tree): Use TDI_gimple instead of
29275         TDI_generic.
29277 2017-06-07  Richard Sandiford  <richard.sandiford@linaro.org>
29279         * doc/md.texi: Clarify the restrictions on a define_insn condition.
29280         Say that # requires an associated define_split to exist, and that
29281         the define_split must be suitable for use after register allocation.
29283 2017-06-06  Jan Hubicka  <hubicka@ucw.cz>
29285         * cfgbuild.c (find_bb_boundaries): Initialize profile of split blocks.
29286         (compute_outgoing_frequencies): Also initialize zero counts.
29287         (find_many_sub_basic_blocks): Do not produce uninitialized profile
29288         around loops; preserve more of profile when nothing changes.
29290 2017-06-06  Jim Wilson  <jim.wilson@linaro.org>
29292         * config/aarch64/aarch64-cost-tables.h (qdf24xx_extra_costs): Move to
29293         here.
29294         * config/arm/aarch-cost-tables.h (qdf24xx_extra_costs): From here.
29295         * config/arm/arm-cpu-cdata.h: Regenerate.
29296         * config/arm/arm-cpu-data.h, config/arm/arm-cpu.h: Likewise.
29297         * config/arm/arm-tables.opt, config/arm/arm-tune.md: Likewise.
29298         * config/arm/arm-cpus.in: Delete falkor and qdf24xx entries.
29299         * config/arm/arm.c (arm_qdf24xx_tune): Delete.
29300         * config/arm/bpabi.h (BE8_LINK_SPEC): Delete falkor and qdf24xx
29301         support.
29302         * config/arm/t-aprofile (MULTILIB_MATCHES): Delete falkor and qdf24xx
29303         support.
29304         * config/arm/t-rmprofile: Likewise.
29305         * doc/invoke.texi (ARM Options): Drop falkor and qdf24xx support.
29307 2017-06-06  David S. Miller  <davem@davemloft.net>
29309         PR target/80968
29310         * config/sparc/sparc.c (sparc_expand_prologue): Emit frame
29311         blockage if function uses alloca.
29313 2017-06-06  Maxim Kuvyrkov  <maxim.kuvyrkov@linaro.org>
29315         * tree-ssa-loop-prefetch.c (struct mem_ref_group, struct mem_ref):
29316         New "uid" fields to hold pretty-print IDs of group and ref.
29317         Memory references are now identified as <group_id>:<ref_id>
29318         instead of using [random] addresses.
29319         (dump_mem_details): Simplify, no functional change.
29320         (dump_mem_ref): Simplify and make output more concise.
29321         Replace couple of fprintf's throughout code with calls to dump_mem_ref.
29322         (find_or_create_group): Initialize group uid.
29323         (record_ref): Initialize ref uid.  Improve debug output.
29324         (prune_group_by_reuse, should_issue_prefetch_p,)
29325         (should_issue_prefetch_p, schedule_prefetches, issue_prefetch_ref,)
29326         (mark_nontemporal_store, determine_loop_nest_reuse):
29327         Improve debug output.
29329 2017-06-06  Maxim Kuvyrkov  <maxim.kuvyrkov@linaro.org>
29331         * dbgcnt.def (prefetch): New debug counter.
29332         * tree-ssa-loop-prefetch.c (dbgcnt.h): New include.
29333         (schedule_prefetches): Stop issueing prefetches if debug counter
29334         tripped.
29336 2017-06-06  Tom de Vries  <tom@codesourcery.com>
29338         * doc/sourcebuild.texi (Testsuites, C Language Testsuites,
29339         gcc.c-torture/compile): Remove mention of NO_LABEL_VALUES in fixme.
29341 2017-06-06  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
29343         * config/aarch64/atomics.md (atomic_compare_and_swap<mode> expander):
29344         Use aarch64_reg_or_zero predicate for operand 4.
29345         (aarch64_compare_and_swap<mode> define_insn_and_split):
29346         Use aarch64_reg_or_zero predicate for operand 3.  Add 'Z' constraint.
29347         (aarch64_store_exclusive<mode>): Likewise for operand 2.
29349 2017-06-06  Thomas Preud'homme  <thomas.preudhomme@arm.com>
29351         * config/arm/arm.c (arm_compute_save_reg_mask): Rename into ...
29352         (arm_compute_save_core_reg_mask): This.
29353         (thumb1_compute_save_reg_mask): Rename into ...
29354         (thumb1_compute_save_core_reg_mask): This.
29355         (arm_compute_save_reg0_reg12_mask): Adapt comment.
29356         (arm_compute_frame_layout): Likewise.
29358 2017-06-06  Richard Biener  <rguenther@suse.de>
29360         PR tree-optimization/80974
29361         * tree-ssa-sccvn.c (set_ssa_val_to): Do not change but only
29362         keep or clear leaders SSA info.
29364 2017-06-06  Tom de Vries  <tom@codesourcery.com>
29366         * config/nvptx/nvptx.c (split_mode_p): New function.
29367         (nvptx_declare_function_name, nvptx_print_operand): Use split_mode_p.
29369 2017-06-06  Tom de Vries  <tom@codesourcery.com>
29371         * config/nvptx/nvptx.c (nvptx_print_operand): Use maybe_split_mode.
29373 2017-06-06  Jan Hubicka  <hubicka@ucw.cz>
29375         PR bootstrap/80978
29376         * tree-cfg.c (execute_fixup_cfg): Fix condition on when to rescale
29377         profile.
29379 2017-06-06  Jan Hubicka  <hubicka@ucw.cz>
29381         * shrink-wrap.c (handle_simple_exit): Update profile.
29382         (try_shrink_wrapping): Upate profile.
29384 2017-06-06  Jan Hubicka  <hubicka@ucw.cz>
29386         * predict.c (tree_estimate_probability_bb): Add LOCAL_ONLY.
29387         (tree_guess_outgoing_edge_probabilities): New.
29388         * predict.h (tree_guess_outgoing_edge_probabilities): Declare.
29389         * tree-cfg.c (gimple_find_sub_bbs): Propagate profile.
29391 2017-06-06  Jan Hubicka  <hubicka@ucw.cz>
29393         * ipa-split.c (split_function): Initialize return bb profile.
29395 2017-06-06  Jan Hubicka  <hubicka@ucw.cz>
29397         * profile.c (compute_branch_probabilities): Also initialize
29398         EXIT_BLOCK profile.
29400 2017-06-06  Richard Biener  <rguenther@suse.de>
29402         PR tree-optimization/80928
29403         * tree-vect-loop.c (vect_update_vf_for_slp): Amend dumps.
29404         (vect_analyze_loop_operations): Properly guard analysis for
29405         pure SLP case.
29406         (vect_transform_loop): Likewise.
29407         (vect_analyze_loop_2): Also reset SLP type on PHIs.
29408         (vect_model_induction_cost): Do not cost for pure SLP.
29409         (vectorizable_induction): Pass in SLP node, implement SLP vectorization
29410         of induction in inner loop vectorization.
29411         * tree-vect-slp.c (vect_create_new_slp_node): Handle PHIs.
29412         (vect_get_and_check_slp_defs): Handle vect_induction_def.
29413         (vect_build_slp_tree): Likewise.  Handle PHIs as terminating the
29414         recursion.
29415         (vect_analyze_slp_cost_1): Cost induction.
29416         (vect_detect_hybrid_slp_stmts): Handle PHIs.
29417         (vect_get_slp_vect_defs): Likewise.
29418         * tree-vect-stmts.c (vect_analyze_stmt): Handle induction.
29419         (vect_transform_stmt): Handle SLP reductions.
29420         * tree-vectorizer.h (vectorizable_induction): Adjust.
29422 2017-06-05  Michael Meissner  <meissner@linux.vnet.ibm.com>
29424         * config/rs6000/rs6000.c (make_resolver_func): Update
29425         init_lowered_empty_function call.
29427 2017-06-05  Bernd Edlinger  <bernd.edlinger@hotmail.de>
29429         * doc/invoke.texi: Document the -fprofile-abs-path option.
29430         * common.opt (fprofile-abs-path): New option.
29431         * gcov-io.h (gcov_write_filename): Declare.
29432         * gcov-io.c (gcov_write_filename): New function.
29433         * coverage.c (coverage_begin_function): Use gcov_write_filename.
29434         * profile.c (output_location): Likewise.
29436 2017-06-05  Jan Hubicka  <hubicka@ucw.cz>
29438         * shring-wrap.c: Revert accidental commit.
29440 2017-06-05  Volker Reichelt  <v.reichelt@netcologne.de>
29442         * doc/invoke.texi (-Wduplicated-branches): Add to warning list.
29444 2017-06-05  Jan Hubicka  <hubicka@ucw.cz>
29446         * cfgexpand.c (expand_gimple_tailcall): Initialize profile of
29447         new edge.
29448         * ipa-inline.c (want_inline_self_recursive_call_p): Watch for missing
29449         profile in callgraph edge.
29450         * profile-count.h (apply_probability): If THIS is 0, then result is 0
29451         (apply_scale): Likewise.
29452         * tree-inline.c (copy_bb, copy_edges_for_bb, copy_cfg_body):
29453         Also scale profile when inlining function with zero profile.
29454         (initialize_cfun): Update exit block profile even when it is zero.
29455         * tree-ssa-threadupdate.c (clear_counts_path): Handle correctly case
29456         when profile is read.
29458 2017-06-05  Michael Meissner  <meissner@linux.vnet.ibm.com>
29460         * config/rs6000/rs6000.c (toplevel): Include attribs.h.
29461         (CLONE_*): New constants to define the processors we can generate
29462         code for with the target_clone attribute.
29463         (rs6000_clone_map): New array to identify which clone processors
29464         the current program is running on.
29465         (TARGET_COMPARE_VERSION_PRIORITY): Define to enable the
29466         target_clone attribute.
29467         (TARGET_GENERATE_VERSION_DISPATCHER_BODY): Likewise.
29468         (TARGET_GET_FUNCTION_VERSIONS_DISPATCHER): Likewise.
29469         (TARGET_OPTION_FUNCTION_VERSIONS): Likewise.
29470         (cpu_expand_builtin): Add support for target_clone attribute.
29471         (rs6000_valid_attribute_p): Allow "default" attribute.
29472         (get_decl_name): New debug function to simplify printing the
29473         current function name in debugging statements.
29474         (rs6000_clone_priority): New functions to support the target_clone
29475         attribute, and be able to generate code to switch between ISA 2.05
29476         through ISA 3.0 (power6 through power9).
29477         (rs6000_compare_version_priority): Likewise.
29478         (rs6000_get_function_versions_dispatcher): Likewise.
29479         (make_resolver_func): Likewise.
29480         (add_condition_to_bb): Likewise.
29481         (dispatch_function_versions): Likewise.
29482         (rs6000_generate_version_dispatcher_body): Likewise.
29483         (rs6000_can_inline_p): Call get_decl_name for debugging usage.
29484         (fusion_gpr_load_p): Fix a spacing issue.
29485         * doc/extend.texi (Common Function Attributes): Document that the
29486         PowerPC supports the target_clone attribute.
29488 2017-06-05  Thomas Preud'homme  <thomas.preudhomme@arm.com>
29490         * config/arm/arm.h: explain F symbol found in description of ARM
29491         register allocation in its legend.
29493 2017-06-05  Jan Hubicka  <hubicka@ucw.cz>
29495         * config/mips/frame-header-opt.c: Include profile-count.h.
29496         * config/riscv/riscv.c: Include profile-count.h
29498 2017-06-05  Jan Hubicka  <hubicka@ucw.cz>
29500         * tree-ssa-loop-im.c (execute_sm_if_changed): Add FLAG_BBS parameter;
29501         update profile.
29502         (sm_set_flag_if_changed): Add bbs field.
29503         (execute_sm_if_changed_flag_set): Pass BBS.
29504         (execute_sm): Update.
29506 2017-06-05  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
29508         * config/aarch64/aarch64-simd.md (aarch64_store_lane0<mode>):
29509         New pattern.
29511 2017-06-05  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
29513         * config/aarch64/aarch64.md (sub<mode>3_compare1_imm): New define_insn.
29514         (peephole2): New peephole2 to emit the above.
29515         * config/aarch64/predicates.md (aarch64_sub_immediate): New predicate.
29517 2017-06-05  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
29519         * config/aarch64/aarch64.c (define_peephole2 above
29520         *sub_<shift>_<mode>): New peephole.
29522 2017-05-23  Jan Hubicka  <hubicka@ucw.cz>
29524         * config/i386/i386.c (make_resolver_func): Update.
29525         * Makefile.in: Add profile-count.h and profile-count.o
29526         * auto-profile.c (afdo_indirect_call): Update to new API.
29527         (afdo_set_bb_count): Update.
29528         (afdo_propagate_edge): Update.
29529         (afdo_propagate_circuit): Update.
29530         (afdo_calculate_branch_prob): Update.
29531         (afdo_annotate_cfg): Update.
29532         * basic-block.h: Include profile-count.h
29533         (struct edge_def): Turn count to profile_count.
29534         (struct basic_block_def): Likewie.
29535         (REG_BR_PROB_BASE): Move to profile-count.h
29536         (RDIV): Move to profile-count.h
29537         * bb-reorder.c (max_entry_count): Turn to profile_count.
29538         (find_traces): Update.
29539         (rotate_loop):Update.
29540         (connect_traces):Update.
29541         (sanitize_hot_paths):Update.
29542         * bt-load.c (migrate_btr_defs): Update.
29543         * cfg.c (RDIV): Remove.
29544         (init_flow): Use alloc_block.
29545         (alloc_block): Uninitialize count.
29546         (unchecked_make_edge): Uninitialize count.
29547         (check_bb_profile): Update.
29548         (dump_edge_info): Update.
29549         (dump_bb_info): Update.
29550         (update_bb_profile_for_threading): Update.
29551         (scale_bbs_frequencies_int): Update.
29552         (scale_bbs_frequencies_gcov_type): Update.
29553         (scale_bbs_frequencies_profile_count): New.
29554         * cfg.h (update_bb_profile_for_threading): Update.
29555         (scale_bbs_frequencies_profile_count): Declare.
29556         * cfgbuild.c (compute_outgoing_frequencies): Update.
29557         (find_many_sub_basic_blocks): Update.
29558         * cfgcleanup.c (try_forward_edges): Update.
29559         (try_crossjump_to_edge): Update.
29560         * cfgexpand.c (expand_gimple_tailcall): Update.
29561         (construct_exit_block): Update.
29562         * cfghooks.c (verify_flow_info): Update.
29563         (dump_bb_for_graph): Update.
29564         (split_edge): Update.
29565         (make_forwarder_block): Update.
29566         (duplicate_block): Update.
29567         (account_profile_record): Update.
29568         * cfgloop.c (find_subloop_latch_edge_by_profile): Update.
29569         (get_estimated_loop_iterations): Update.
29570         * cfgloopanal.c (expected_loop_iterations_unbounded): Update.
29571         (single_likely_exit): Update.
29572         * cfgloopmanip.c (scale_loop_profile): Update.
29573         (loopify): Update.
29574         (set_zero_probability): Update.
29575         (lv_adjust_loop_entry_edge): Update.
29576         * cfgrtl.c (force_nonfallthru_and_redirect): Update.
29577         (purge_dead_edges): Update.
29578         (rtl_account_profile_record): Update.
29579         * cgraph.c (cgraph_node::create): Uninitialize count.
29580         (symbol_table::create_edge): Uninitialize count.
29581         (cgraph_update_edges_for_call_stmt_node): Update.
29582         (cgraph_edge::dump_edge_flags): Update.
29583         (cgraph_node::dump): Update.
29584         (cgraph_edge::maybe_hot_p): Update.
29585         * cgraph.h: Include profile-count.h
29586         (create_clone), create_edge, create_indirect_edge): Update.
29587         (cgraph_node): Turn count to profile_count.
29588         (cgraph_edge0: Likewise.
29589         (make_speculative, clone): Update.
29590         (create_edge): Update.
29591         (init_lowered_empty_function): Update.
29592         * cgraphclones.c (cgraph_edge::clone): Update.
29593         (duplicate_thunk_for_node): Update.
29594         (cgraph_node::create_clone): Update.
29595         * cgraphunit.c (cgraph_node::analyze): Update.
29596         (cgraph_node::expand_thunk): Update.
29597         * final.c (dump_basic_block_info): Update.
29598         * gimple-streamer-in.c (input_bb): Update.
29599         * gimple-streamer-out.c (output_bb): Update.
29600         * graphite.c (print_global_statistics): Update.
29601         (print_graphite_scop_statistics): Update.
29602         * hsa-brig.c: Include basic-block.h.
29603         * hsa-dump.c: Include basic-block.h.
29604         * hsa-gen.c (T sum_slice): Update.
29605         (convert_switch_statements):Update.
29606         * hsa-regalloc.c: Include basic-block.h.
29607         * ipa-chkp.c (chkp_produce_thunks): Update.
29608         * ipa-cp.c (struct caller_statistics): Update.
29609         (init_caller_stats): Update.
29610         (gather_caller_stats): Update.
29611         (ipcp_cloning_candidate_p): Update.
29612         (good_cloning_opportunity_p): Update.
29613         (get_info_about_necessary_edges): Update.
29614         (dump_profile_updates): Update.
29615         (update_profiling_info): Update.
29616         (update_specialized_profile): Update.
29617         (perhaps_add_new_callers): Update.
29618         (decide_about_value): Update.
29619         (ipa_cp_c_finalize): Update.
29620         * ipa-devirt.c (struct odr_type_warn_count): Update.
29621         (struct decl_warn_count): Update.
29622         (struct final_warning_record): Update.
29623         (possible_polymorphic_call_targets): Update.
29624         (ipa_devirt): Update.
29625         * ipa-fnsummary.c (redirect_to_unreachable): Update.
29626         * ipa-icf.c (sem_function::merge): Update.
29627         * ipa-inline-analysis.c (do_estimate_edge_time): Update.
29628         * ipa-inline.c (compute_uninlined_call_time): Update.
29629         (compute_inlined_call_time): Update.
29630         (want_inline_small_function_p): Update.
29631         (want_inline_self_recursive_call_p): Update.
29632         (edge_badness): Update.
29633         (lookup_recursive_calls): Update.
29634         (recursive_inlining): Update.
29635         (inline_small_functions): Update.
29636         (dump_overall_stats): Update.
29637         (dump_inline_stats): Update.
29638         * ipa-profile.c (ipa_profile_generate_summary): Update.
29639         (ipa_propagate_frequency): Update.
29640         (ipa_profile): Update.
29641         * ipa-prop.c (ipa_make_edge_direct_to_target): Update.
29642         * ipa-utils.c (ipa_merge_profiles): Update.
29643         * loop-doloop.c (doloop_modify): Update.
29644         * loop-unroll.c (report_unroll): Update.
29645         (unroll_loop_runtime_iterations): Update.
29646         * lto-cgraph.c (lto_output_edge): Update.
29647         (lto_output_node): Update.
29648         (input_node): Update.
29649         (input_edge): Update.
29650         (merge_profile_summaries): Update.
29651         * lto-streamer-in.c (input_cfg): Update.
29652         * lto-streamer-out.c (output_cfg): Update.
29653         * mcf.c (create_fixup_graph): Update.
29654         (adjust_cfg_counts): Update.
29655         (sum_edge_counts): Update.
29656         * modulo-sched.c (sms_schedule): Update.
29657         * postreload-gcse.c (eliminate_partially_redundant_load): Update.
29658         * predict.c (maybe_hot_count_p): Update.
29659         (probably_never_executed): Update.
29660         (dump_prediction): Update.
29661         (combine_predictions_for_bb): Update.
29662         (propagate_freq): Update.
29663         (handle_missing_profiles): Update.
29664         (counts_to_freqs): Update.
29665         (rebuild_frequencies): Update.
29666         (force_edge_cold): Update.
29667         * predict.h: Include profile-count.h
29668         (maybe_hot_count_p, counts_to_freqs): UPdate.
29669         * print-rtl-function.c: Do not include cfg.h
29670         * print-rtl.c: Include basic-block.h
29671         * profile-count.c: New file.
29672         * profile-count.h: New file.
29673         * profile.c (is_edge_inconsistent): Update.
29674         (correct_negative_edge_counts): Update.
29675         (is_inconsistent): Update.
29676         (set_bb_counts): Update.
29677         (read_profile_edge_counts): Update.
29678         (compute_frequency_overlap): Update.
29679         (compute_branch_probabilities): Update; Initialize and deinitialize
29680         gcov_count tables.
29681         (branch_prob): Update.
29682         * profile.h (bb_gcov_counts, edge_gcov_counts): New.
29683         (edge_gcov_count): New.
29684         (bb_gcov_count): New.
29685         * shrink-wrap.c (try_shrink_wrapping): Update.
29686         * tracer.c (better_p): Update.
29687         * trans-mem.c (expand_transaction): Update.
29688         (ipa_tm_insert_irr_call): Update.
29689         (ipa_tm_insert_gettmclone_call): Update.
29690         * tree-call-cdce.c: Update.
29691         * tree-cfg.c (gimple_duplicate_sese_region): Update.
29692         (gimple_duplicate_sese_tail): Update.
29693         (gimple_account_profile_record): Update.
29694         (execute_fixup_cfg): Update.
29695         * tree-inline.c (copy_bb): Update.
29696         (copy_edges_for_bb): Update.
29697         (initialize_cfun): Update.
29698         (freqs_to_counts): Update.
29699         (copy_cfg_body): Update.
29700         (expand_call_inline): Update.
29701         * tree-ssa-ifcombine.c (update_profile_after_ifcombine): Update.
29702         * tree-ssa-loop-ivcanon.c (unloop_loops): Update.
29703         (try_unroll_loop_completely): Update.
29704         (try_peel_loop): Update.
29705         * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Update.
29706         * tree-ssa-loop-niter.c (estimate_numbers_of_iterations_loop): Update.
29707         * tree-ssa-loop-split.c (connect_loops): Update.
29708         * tree-ssa-loop-unswitch.c (hoist_guard): Update.
29709         * tree-ssa-reassoc.c (branch_fixup): Update.
29710         * tree-ssa-tail-merge.c (replace_block_by): Update.
29711         * tree-ssa-threadupdate.c (create_block_for_threading): Update.
29712         (compute_path_counts): Update.
29713         (update_profile): Update.
29714         (recompute_probabilities): Update.
29715         (update_joiner_offpath_counts): Update.
29716         (estimated_freqs_path): Update.
29717         (freqs_to_counts_path): Update.
29718         (clear_counts_path): Update.
29719         (ssa_fix_duplicate_block_edges): Update.
29720         (duplicate_thread_path): Update.
29721         * tree-switch-conversion.c (case_bit_test_cmp): Update.
29722         (struct switch_conv_info): Update.
29723         * tree-tailcall.c (decrease_profile): Update.
29724         * tree-vect-loop-manip.c (slpeel_add_loop_guard): Update.
29725         * tree-vect-loop.c (scale_profile_for_vect_loop): Update.
29726         * value-prof.c (check_counter): Update.
29727         (gimple_divmod_fixed_value): Update.
29728         (gimple_mod_pow2): Update.
29729         (gimple_mod_subtract): Update.
29730         (gimple_ic_transform): Update.
29731         (gimple_stringop_fixed_value): Update.
29732         * value-prof.h (gimple_ic): Update.
29734 2017-06-02  Carl Love  <cel@us.ibm.com>
29736         * config/rs6000/rs6000-c: Add support for built-in functions
29737         vector double vec_doublee (vector signed int);
29738         vector double vec_doublee (vector unsigned int);
29739         vector double vec_doublee (vector float);
29740         vector double vec_doubleh (vector signed int);
29741         vector double vec_doubleh (vector unsigned int);
29742         vector double vec_doubleh (vector float);
29743         vector double vec_doublel (vector signed int);
29744         vector double vec_doublel (vector unsigned int);
29745         vector double vec_doublel (vector float);
29746         vector double vec_doubleo (vector signed int);
29747         vector double vec_doubleo (vector unsigned int);
29748         vector double vec_doubleo (vector float);.
29749         * config/rs6000/rs6000-builtin.def: Add definitions for DOUBLEE,
29750         DOUBLEO, DOUBLEH, DOUBLEL, UNS_DOUBLEO, UNS_DOUBLEE, UNS_DOUBLEH,
29751         UNS_DOUBLEL.
29752         * config/rs6000/altivec.md: Add code generator for doublee<mode>2,
29753         unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2, doubleh<mode>2,
29754         unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2, add mode attribute
29755         VS_sxwsp.
29756         * config/rs6000/altivec.h: Add define for vec_doublee, vec_doubleo,
29757         vec_doublel, vec_doubleh.
29758         * doc/extend.texi: Update the built-in documentation file for the
29759         new built-in functions.
29761 2017-06-02  David Malcolm  <dmalcolm@redhat.com>
29763         PR jit/80954
29764         * ipa-inline-analysis.c (free_growth_caches): Set
29765         edge_removal_hook_holder to NULL after removing it.
29767 2017-06-02  Sudakshina Das  <sudi.das@arm.com>
29769         * config/aarch64/aarch64.c (aarch64_select_cc_mode): Return CC_SWP for
29770         comparision with zero.
29772 2017-06-02  Will Schmidt  <will_schmidt@vnet.ibm.com>
29773         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add handling
29774         for early expansion of vec_min and vec_max builtins.
29775         (builtin_function_type): Add min/max unsigned variants to those
29776         identified as having unsigned arguments.
29778 2017-06-02  Olivier Hainque  <hainque@adacore.com>
29780         * config/vx-common.h (DWARF_UNWIND_INFO): Switch #define to 1.
29782 2017-06-02  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
29784         * config/aarch64/aarch64-simd.md (*aarch64_simd_vec_copy_lane<mode>):
29785         Use VALL_F16 iterator rather than VALL.
29787 2017-06-02  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
29789         * config/aarch64/aarch64.c (aarch64_split_compare_and_swap):
29790         Emit CBNZ inside loop when doing a strong exchange and comparing
29791         against zero.  Generate the CC flags after the loop.
29793 2017-06-02  David Edelsohn  <dje.gcc@gmail.com>
29795         * dwarf2out.c (DWARF_INITIAL_LENGTH_SIZE_STR): New.
29796         (dl_section_ref): New.
29797         (dwarf2out_finish): Copy debug_line_section_label to dl_section_ref.
29798         On AIX, append an expression to subtract the size of the
29799         section length to dl_section_ref.
29801 2017-06-02  Will Schmidt  <will_schmidt@vnet.ibm.com>
29803         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add handling
29804         for early expansion of vector absolute builtins.
29806 2017-06-02  Richard Biener  <rguenther@suse.de>
29808         * tree-vect-slp.c (vect_detect_hybrid_slp_2): Match up
29809         what we consider a relevant use stmt with vect_detect_hybrid_slp_stmts.
29811 2017-06-02  Richard Biener  <rguenther@suse.de>
29813         PR tree-optimization/80948
29814         * tree-tailcall.c (find_tail_calls): Track stmts to move in
29815         stmt order as well.
29817 2017-06-02  Richard Biener  <rguenther@suse.de>
29819         * tree-vect-loop.c (vect_analyze_loop_operations): Not relevant
29820         PHIs are ok.
29821         * tree-vect-stmts.c (process_use): Do not mark backedge defs
29822         for inductions as relevant.
29824 2017-06-02  Richard Biener  <rguenther@suse.de>
29826         * tree-vect-loop.c (get_initial_def_for_induction): Inline into ...
29827         (vectorizable_induction): ... this.  Remove dead code.
29829 2017-06-02  Eric Botcazou  <ebotcazou@adacore.com>
29831         * builtins. (expand_builtin_alloca): Remove second parameter and
29832         infer its value from the first parameter instead.
29833         (expand_builtin) <BUILT_IN_ALLOCA>: Adjust call to above.
29835 2017-06-02  Jakub Jelinek  <jakub@redhat.com>
29837         PR rtl-optimization/80903
29838         * loop-doloop.c (add_test): Unshare sequence.
29840 2017-06-02  Bernd Edlinger  <bernd.edlinger@hotmail.de>
29842         * doc/invoke.texi: Document the -Wsizeof-pointer-div warning.
29844 2017-06-01  Bernd Edlinger  <bernd.edlinger@hotmail.de>
29846         * config/i386/i386.c (x86_64_ms_sysv_extra_clobbered_registers): Make
29847         static.
29848         (xlogue_layout::get_stack_space_used, xlogue_layout::s_instances,
29849         xlogue_layout::get_instance, logue_layout::xlogue_layout,
29850         sp_valid_at, fp_valid_at, choose_basereg): Formatting.
29851         (xlogue_layout::get_stub_rtx): Make static.
29852         (xlogue_layout::get_stub_name): Avoid const-cast, make static.
29853         (xlogue_layout::compute_stub_managed_regs): Rename to...
29854         (xlogue_layout::count_stub_managed_regs): ...this.
29855         (xlogue_layout::is_stub_managed_reg): New function.
29856         (xlogue_layout::m_stub_names): Rename to...
29857         (xlogue_layout::s_stub_names): ...this, make static.
29858         (xlogue_layout::STUB_INDEX_OFFSET, xlogue_layout::MIN_REGS,
29859         xlogue_layout::MAX_REGS, xlogue_layout::MAX_EXTRA_REGS,
29860         xlogue_layout::VARIANT_COUNT, xlogue_layout::STUB_NAME_MAX_LEN,
29861         xlogue_layout::s_stub_names): Instantiate statics.
29862         (stub_managed_regs): Remove.
29863         (ix86_save_reg): Use xlogue_layout::compute_stub_managed_regs.
29864         (disable_call_ms2sysv_xlogues): Rename to...
29865         (warn_once_call_ms2sysv_xlogues): ...this, and warn only once.
29866         (ix86_initial_elimination_offset, ix86_expand_call): Fix call_ms2sysv
29867         warning logic.
29868         (ix86_static_chain): Make sure that ix86_static_chain_on_stack can't
29869         change after reload_completed.
29870         (ix86_can_use_return_insn_p): Use the ix86_frame data structure
29871         directly.
29872         (ix86_expand_prologue): Likewise.
29873         (ix86_expand_epilogue): Likewise.
29874         (ix86_expand_split_stack_prologue): Likewise.
29875         (ix86_compute_frame_layout): Remove frame parameter ...
29876         (TARGET_COMPUTE_FRAME_LAYOUT): ... and export it as a target hook.
29877         (ix86_finalize_stack_realign_flags): Call ix86_compute_frame_layout
29878         only if necessary.
29879         (ix86_init_machine_status): Don't set use_fast_prologue_epilogue_nregs.
29880         (ix86_frame): Move from here ...
29881         * config/i386/i386.h (ix86_frame): ... to here.
29882         (machine_function): Remove use_fast_prologue_epilogue_nregs, cache the
29883         complete ix86_frame data structure instead.  Remove some_ld_name.
29885 2017-06-01  Pierre-Marie de Rodat  <derodat@adacore.com>
29887         * dwarf2out.c (dwarf2out_late_global_decl): Add locations for
29888         symbols that hold a DECL_VALUE_EXPR.
29890 2017-06-01  Martin Jambor  <mjambor@suse.cz>
29892         PR tree-optimization/80898
29893         * tree-sra.c (process_subtree_disqualification): Removed.
29894         (disqualify_candidate): Do not acll
29895         process_subtree_disqualification.
29896         (subtree_mark_written_and_enqueue): New function.
29897         (propagate_all_subaccesses): Set grp_write of LHS subtree if the
29898         RHS has been disqualified and re-queue LHS if necessary.  Apart
29899         from that, ignore disqualified RHS.
29901 2017-06-01  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
29903         * config/s390/s390.c (s390_emit_epilogue): Disable early return
29904         address fetch for z10 or later.
29906 2017-06-01  Claudiu Zissulescu  <claziss@synopsys.com>
29908         * config/arc/arc.md (tst_movb): Add guard when splitting.
29910 2017-06-01  Claudiu Zissulescu  <claziss@synopsys.com>
29912         * config/arc/arc.c (arc_can_eliminate): Test against
29913         arc_frame_pointer_needed.
29915 2017-06-01  Claudiu Zissulescu  <claziss@synopsys.com>
29917         * config/arc/arc.c (arc_expand_prologue): Emit a special barrier
29918         to prevent store reordering.
29919         * config/arc/arc.md (UNSPEC_ARC_STKTIE): Define.
29920         (type): Add block type.
29921         (stack_tie): Define special instruction to be used in
29922         expand_prologue.
29924 2017-06-01  Claudiu Zissulescu  <claziss@synopsys.com>
29926         * config/arc/arc.md (commutative_binary_comparison): Remove 'I'
29927         constraint. It is not valid for the pattern.
29928         (noncommutative_binary_comparison): Likewise.
29930 2017-06-01  Claudiu Zissulescu  <claziss@synopsys.com>
29932         * config/arc/simdext.md (movv2hi_insn): Change predicate to avoid
29933         scaled addresses.
29935 2017-06-01  Claudiu Zissulescu  <claziss@synopsys.com>
29937         * config/arc/arc.c (arc_conditional_register_usage): Allow r30 to
29938         be used by the reg-alloc.
29940 2017-06-01  Claudiu Zissulescu  <claziss@synopsys.com>
29942         * config/arc/arc.md (mulsi3): Avoid use of hard registers before
29943         reg-alloc when having mul64 or mul32x16 instructions.
29944         (mulsidi3): Likewise.
29945         (umulsidi3): Likewise.
29946         (mulsi32x16): New pattern.
29947         (mulsi64): Likewise.
29948         (mulsidi64): Likewise.
29949         (umulsidi64): Likewise.
29950         (MUL32x16_REG): Define.
29951         (mul64_600): Use MUL32x16_REG.
29952         (mac64_600): Likewise.
29953         (umul64_600): Likewise.
29954         (umac64_600): Likewise.
29956 2017-06-01  Claudiu Zissulescu  <claziss@synopsys.com>
29958         * config/arc/arc.md (mulsi3_700): Make it commutative.
29960 2017-06-01  Jose E. Marchesi  <jose.marchesi@oracle.com>
29962         * config/sparc/sparc.md (*zero_extendsidi2_insn_sp64): Set insn
29963         type for movstouw.
29964         (*sign_extendsidi2_insn): Likewise for movstosw.
29966 2017-06-01  Pierre-Marie de Rodat  <derodat@adacore.com>
29968         * dwarf2out.c (get_discr_value): Call the get_debug_type hook on
29969         the type of the input discriminant value.  Convert the
29970         discriminant value of signedness vary.
29972 2017-06-01  Volker Reichelt  <v.reichelt@netcologne.de>
29974         * doc/invoke.texi (-Wcatch-value): Document new shortcut.
29975         Add to -Wall section.
29977 2017-06-01  Richard Biener  <rguenther@suse.de>
29979         PR middle-end/66313
29980         * fold-const.c (fold_plusminus_mult_expr): If the factored
29981         factor may be zero use a wrapping type for the inner operation.
29982         * tree-tailcall.c (independent_of_stmt_p): Pass in to_move bitmap
29983         and handle moved defs.
29984         (process_assignment): Properly guard the unary op case.  Return a
29985         tri-state indicating that moving the stmt before the call may allow
29986         to continue.  Pass through to_move.
29987         (find_tail_calls): Handle moving unrelated defs before
29988         the call.
29990 2017-05-31  Segher Boessenkool  <segher@kernel.crashing.org>
29992         PR target/80618
29993         * config/rs6000/vector.md (*vector_uneq<mode>): Write the nor in the
29994         splitter result in the canonical way.
29996 2017-05-31  Uros Bizjak  <ubizjak@gmail.com>
29998         * config/i386/i386.md (*zero_extendsidi2): Enable alternative (?r, *Yj)
29999         also for 32bit target.  Update insn attributes.
30000         (zero-extendsidi2 splitter): Allow all registers for operand 1.
30002 2017-05-31  Sebastian Peryt  <sebastian.peryt@intel.com>
30004         * config/i386/avx512fintrin.h (_mm_mask_max_sd)
30005         (_mm_maskz_max_sd, _mm_mask_max_ss, _mm_maskz_max_ss)
30006         (_mm_mask_min_sd, _mm_maskz_min_sd, _mm_mask_min_ss)
30007         (_mm_maskz_min_ss): New intrinsics.
30009 2017-05-31  Martin Liska  <mliska@suse.cz>
30011         * tree-vect-loop.c (vect_create_epilog_for_reduction):
30012         Change comment style to one we normally use.
30013         (vectorizable_reduction): Likewise.
30014         (vectorizable_induction): Likewise.
30015         * tree-vect-stmts.c (vectorizable_mask_load_store): Likewise.
30016         (vectorizable_call): Likewise.
30017         (vectorizable_simd_clone_call): Likewise.
30018         (vectorizable_conversion): Likewise.
30019         (vectorizable_assignment): Likewise.
30020         (vectorizable_shift): Likewise.
30021         (vectorizable_operation): Likewise.
30022         (vectorizable_store): Likewise.
30023         (vectorizable_load): Likewise.
30024         * tree-vectorizer.h: Likewise.
30026 2017-05-31  Alexander Monakov  <amonakov@ispras.ru>
30028         * passes.c (emergency_dump_function): New.
30029         * tree-pass.h (emergency_dump_function): Declare.
30030         * plugin.c (plugins_internal_error_function): Remove.
30031         * plugin.h (plugins_internal_error_function): Remove declaration.
30032         * toplev.c (internal_error_function): New static function.  Use it...
30033         (general_init): ...here.
30035 2017-05-31  Graham Markall  <graham.markall@embecosm.com>
30037         * config/arc/arc.c (arc_print_operand): Handle constant operands.
30038         (arc_rtx_costs): Add costs for new patterns.
30039         * config/arc/arc.md: Additional *add_n and *sub_n patterns.
30040         * config/arc/predicates.md: Add _1_2_3_operand predicate.
30042 2017-05-31  Richard Sandiford  <richard.sandiford@linaro.org>
30044         * tree-ssa-strlen.c (get_next_strinfo): New function.
30045         (get_stridx_plus_constant): Use it.
30046         (zero_length_string): Likewise.
30047         (adjust_related_strinfos): Likewise.
30048         (adjust_last_stmt): Likewise.
30050 2017-05-31  Richard Biener  <rguenther@suse.de>
30052         PR target/80880
30053         * config/i386/i386.c (ix86_expand_builtin): Remove assert
30054         for arg being an SSA name when expanding IX86_BUILTIN_BNDRET.
30056 2017-05-31  Richard Sandiford  <richard.sandiford@linaro.org>
30058         * tree-vect-data-refs.c (vect_find_same_alignment_drs): Remove
30059         loop_vinfo argument and use of dependence distance vectors.
30060         Check instead whether the two references differ only in their
30061         initial value and assume that they have the same alignment if the
30062         difference is a multiple of the vector alignment.
30063         (vect_analyze_data_refs_alignment): Update call accordingly.
30065 2017-05-31  Martin Liska  <mliska@suse.cz>
30067         PR target/79155
30068         * config/i386/cpuid.h: Fix typo in a comment in cpuid.h.
30070 2017-05-31  Bin Cheng  <bin.cheng@arm.com>
30072         * tree-vect-loop-manip.c (create_intersect_range_checks_index)
30073         (create_intersect_range_checks): Move from ...
30074         * tree-data-ref.c (create_intersect_range_checks_index)
30075         (create_intersect_range_checks): ... to here.
30076         (create_runtime_alias_checks): New function factored from ...
30077         * tree-vect-loop-manip.c (vect_create_cond_for_alias_checks): ...
30078         here.  Call above function.
30079         * tree-data-ref.h (create_runtime_alias_checks): New function.
30081 2017-05-31  Bin Cheng  <bin.cheng@arm.com>
30083         * tree-data-ref.c (prune_runtime_alias_test_list): Relax minimal
30084         segment length for dr_b and compute it in wide_int.
30086 2017-05-31  Richard Biener  <rguenther@suse.de>
30088         PR tree-optimization/80906
30089         * graphite-isl-ast-to-gimple.c (copy_loop_close_phi_nodes): Get
30090         and pass through iv_map.
30091         (copy_bb_and_scalar_dependences): Adjust.
30092         (translate_pending_phi_nodes): Likewise.
30093         (copy_loop_close_phi_args): Handle code-generating IVs instead
30094         of ICEing.
30096 2017-05-30  David Malcolm  <dmalcolm@redhat.com>
30098         * diagnostic-color.c (color_dict): Add "type-diff".
30099         (parse_gcc_colors): Update comment.
30100         * doc/invoke.texi (Diagnostic Message Formatting Options): Add
30101         -fdiagnostics-show-template-tree and -fno-elide-type.
30102         (GCC_COLORS): Add type-diff to example.
30103         (type-diff=): New.
30104         (-fdiagnostics-show-template-tree): New.
30105         (-fno-elide-type): New.
30106         * pretty-print.c (pp_format): Pass quote and formatters[argno] to
30107         the pp_format_decoder callback.  Call any m_format_postprocessor's
30108         "handle" method.
30109         (pretty_printer::pretty_printer): Initialize
30110         m_format_postprocessor.
30111         (pretty_printer::~pretty_printer): Delete any
30112         m_format_postprocessor.
30113         * pretty-print.h (printer_fn): Add bool and const char ** parameters.
30114         (class format_postprocessor): New class.
30115         (struct pretty_printer::format_decoder): Document the new parameters.
30116         (struct pretty_printer::m_format_postprocessor): New field.
30117         * tree-diagnostic.c (default_tree_printer): Update for new
30118         bool and const char ** params.
30119         * tree-diagnostic.h (default_tree_printer): Likewise.
30121 2017-05-30  Segher Boessenkool  <segher@kernel.crashing.org>
30123         * config/rs6000/predicates.md (cc_reg_not_micro_cr0_operand): Delete.
30124         (lwa_operand): Delete rs6000_gen_cell_microcode test.
30125         * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
30126         rs6000_gen_cell_microcode code.
30127         (rs6000_final_prescan_insn): Delete.
30128         (rs6000_opt_vars): Delete the "gen-cell-microcode" and
30129         "warn-cell-microcode" entries.
30130         * config/rs6000/rs6000.h (FINAL_PRESCAN_INSN): Delete.
30131         * config/rs6000/rs6000.md: Delete rs6000_gen_cell_microcode tests
30132         throughout.  Change cc_reg_not_micro_cr0_operand to
30133         cc_reg_not_cr0_operand throughout.
30134         (*extendhi<mode>2_noload): Delete.
30135         * config/rs6000/rs6000.opt (mgen-cell-microcode): Replace by stub.
30136         (mwarn-cell-microcode): Delete.
30137         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete
30138         -mgen-cell-microcode and -mwarn-cell-microcode.
30140 2017-05-30  Uros Bizjak  <ubizjak@gmail.com>
30142         PR target/80833
30143         * config/i386/constraints.md (Yd): New constraint.
30144         (Ye): Ditto.
30145         * config/i386/i386.md (*movti_internal): Add (?r, Ye)
30146         and (?Yd, r) alternatives.  Update insn attributes.
30147         * config/i386/i386.md (*movti_internal): Add (?r, *Ye)
30148         and (?*Yd, r) alternatives.  Update insn attributes.
30149         (double-mode inter-unit splitters): Add new GR<->XMM splitters.
30151 2017-05-30  Pierre-Marie de Rodat  <derodat@adacore.com>
30153         * gimplify.c (gimplify_modify_expr): Don't create a
30154         DECL_DEBUG_EXPR link if *FROM_P does not belong to the current
30155         function.
30157 2017-05-30  Wilco Dijkstra  <wdijkstr@arm.com>
30159         * config/arm/arm-builtins.c (arm_expand_builtin): Remove const.
30161 2017-05-30  Richard Biener  <rguenther@suse.de>
30163         * tree-vectorizer.h (struct _stmt_vec_info): Add reduc_type
30164         and reduc_def fields.
30165         (STMT_VINFO_REDUC_TYPE): New define.
30166         (STMT_VINFO_REDUC_DEF): Likewise.
30167         (vect_force_simple_reduction): Adjust prototype.
30168         * tree-vect-loop.c (vect_analyze_scalar_cycles_1): Adjust.
30169         (vect_is_simple_reduction): Remove check_reduction argument.
30170         (vect_force_simple_reduction): Adjust and set
30171         STMT_VINFO_REDUC_TYPE and STMT_VINFO_REDUC_DEF.
30172         (vectorizable_reduction): Do not re-do reduction analysis
30173         but use STMT_VINFO_REDUC_TYPE and STMT_VINFO_REDUC_DEF.
30174         * tree-parloops.c (gather_scalar_reductions): Adjust.
30176 2017-05-30  Richard Biener  <rguenther@suse.de>
30178         PR middle-end/80901
30179         * cfgexpand.c (expand_gimple_cond): Match up loop fixup with
30180         split_edge code.
30182 2017-05-24  Robin Dapp  <rdapp@linux.vnet.ibm.com>
30184         * tree-vect-data-refs.c (vect_get_peeling_costs_all_drs):
30185         Introduce unknown_misalignment parameter and remove vf.
30186         (vect_peeling_hash_get_lowest_cost):
30187         Pass unknown_misalignment parameter.
30188         (vect_enhance_data_refs_alignment):
30189         Fix unsupportable data ref treatment.
30191 2017-05-30  Robin Dapp  <rdapp@linux.vnet.ibm.com>
30193         * tree-vect-data-refs.c (vect_get_data_access_cost):
30194         Workaround for SLP handling.
30195         (vect_enhance_data_refs_alignment):
30196         Compute costs for doing no peeling at all, compare to the best
30197         peeling costs so far and avoid peeling if cheaper.
30199 2017-05-30  Robin Dapp  <rdapp@linux.vnet.ibm.com>
30201         * tree-vect-data-refs.c (vect_peeling_hash_choose_best_peeling):
30202         Return peeling info and set costs to zero for unlimited cost
30203         model.
30204         (vect_enhance_data_refs_alignment): Also inspect all datarefs
30205         with unknown misalignment. Compute and costs for unknown
30206         misalignment, compare them to the costs for known misalignment
30207         and choose the cheapest for peeling.
30209 2017-05-30  Robin Dapp  <rdapp@linux.vnet.ibm.com>
30211         * tree-vect-data-refs.c (vect_update_misalignment_for_peel): Rename.
30212         (vect_get_peeling_costs_all_drs): Create function.
30213         (vect_peeling_hash_get_lowest_cost):
30214         Use vect_get_peeling_costs_all_drs.
30215         (vect_peeling_supportable): Create function.
30216         (vect_enhance_data_refs_alignment): Use vect_peeling_supportable.
30218 2017-05-30  Robin Dapp  <rdapp@linux.vnet.ibm.com>
30220         * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Create
30221         DR_HAS_NEGATIVE_STEP.
30222         (vect_update_misalignment_for_peel): Define DR_MISALIGNMENT.
30223         (vect_enhance_data_refs_alignment): Use.
30224         (vect_duplicate_ssa_name_ptr_info): Use.
30225         * tree-vectorizer.h (dr_misalignment): Use.
30226         (known_alignment_for_access_p): Use.
30228 2017-05-30  Jozef Lawrynowicz  <jozef.l@somniumtech.com>
30230         PR target/78838
30231         * config/msp430/msp430.c (gen_prefix): Return NULL when section name is
30232         .lowtext.
30233         (has_section_name): New function.
30235 2017-05-30  Martin Liska  <mliska@suse.cz>
30237         PR other/80909
30238         * auto-profile.c (get_function_decl_from_block): Fix
30239         parenthesis.
30241 2017-05-30  Richard Biener  <rguenther@suse.de>
30243         PR middle-end/80876
30244         * cfgexpand.c (expand_gimple_cond): Fixup preserving loops again.
30246 2017-05-30  Martin Liska  <mliska@suse.cz>
30248         * dumpfile.c: Use newly added macro DUMP_FILE_INFO.
30249         * dumpfile.h (struct dump_file_info): Remove ctors.
30251 2017-05-30  Martin Liska  <mliska@suse.cz>
30253         * predict.def: Fix GNU coding style.
30255 2017-05-29  Max Filippov  <jcmvbkbc@gmail.com>
30257         * config/xtensa/xtensa.c (xtensa_initial_elimination_offset):
30258         Mark 'to' argument with ATTRIBUTE_UNUSED.
30260 2017-05-29  Max Filippov  <jcmvbkbc@gmail.com>
30262         * config/xtensa/xtensa.c (xtensa_emit_call): Use
30263         HOST_WIDE_INT_PRINT_HEX instead of 0x%lx format string.
30264         (print_operand): Use HOST_WIDE_INT_PRINT_DEC instead of %ld
30265         format string.
30267 2017-05-29  Eric Botcazou  <ebotcazou@adacore.com>
30269         * doc/install.texi (Options specification): Restore entry of
30270         --enable-sjlj-exceptions.
30272 2017-05-27  Michael Eager  <eager@eagercon.com>
30274         Revert:
30275         2016-01-21  Ajit Agarwal  <ajitkum@xilinx.com>
30277         See https://gcc.gnu.org/ml/gcc/2017-05/msg00221.html.
30279         * config/microblaze/microblaze.h
30280         (FIXED_REGISTERS): Update in macro.
30281         (CALL_USED_REGISTERS): Update in macro.
30283 2017-05-27  François-Xavier Coudett  <fxcoudert@gcc.gnu.org>
30285         * doc/install.texi: Add links to macOS binary distributions.
30287 2017-05-27  Jakub Jelinek  <jakub@redhat.com>
30289         PR bootstrap/80887
30290         Revert:
30291         2017-05-25  Marc Glisse  <marc.glisse@inria.fr>
30293         * match.pd ((A +- CST1) +- CST2): Allow some conversions.
30295 2017-05-26  Martin Liska  <mliska@suse.cz>
30297         * dumpfile.h (enum dump_kind): Renumber TDF_* flags to be contiguous.
30299 2017-05-26  Martin Liska  <mliska@suse.cz>
30301         * cfg.c (check_bb_profile): Do not use TDF_COMMENT and print
30302         always leading ';; '.
30303         (dump_bb_info): Likewise.
30304         (brief_dump_cfg): Likewise.
30305         * cfgrtl.c (print_rtl_with_bb): Do not use TDF_COMMENT.
30306         * dumpfile.c: Remove usage of TDF_VERBOSE.
30307         * dumpfile.h (enum dump_kind): Likewise.
30308         (dump_gimple_bb_header): Do not use TDF_COMMENT.
30309         * print-tree.c (debug_verbose): Remove.
30310         * tree-cfg.c (gimple_dump_cfg): Do not use TDF_COMMENT.
30311         (dump_function_to_file): Remove dumps guarded with TDF_VERBOSE.
30312         * tree-diagnostic.c (default_tree_printer): Replace
30313         TDF_DIAGNOSTIC with TDF_SLIM.
30315 2017-05-26  Bin Cheng  <bin.cheng@arm.com>
30317         * tree-vect-loop-manip.c (create_intersect_range_checks_index): Pass
30318         in parameter loop, rather than loop_vinfo.
30319         (create_intersect_range_checks): Ditto.
30320         (vect_create_cond_for_alias_checks): Update call to above functions.
30322 2017-05-26  Bin Cheng  <bin.cheng@arm.com>
30324         PR tree-optimization/80815
30325         * tree-data-ref.c (prune_runtime_alias_test_list): Simplify condition
30326         for merging runtime alias checks.  Handle negative DR_STEPs.
30328 2017-05-26  Bin Cheng  <bin.cheng@arm.com>
30330         * tree-vect-data-refs.c (Operator==, comp_dr_with_seg_len_pair):
30331         Move from ...
30332         * tree-data-ref.c (Operator==, comp_dr_with_seg_len_pair): To here.
30333         * tree-vect-data-refs.c (vect_prune_runtime_alias_test_list): Factor
30334         out code pruning runtime alias checks.
30335         * tree-data-ref.c (prune_runtime_alias_test_list): New function
30336         factored out from above.
30337         * tree-vectorizer.h (struct dr_with_seg_len, dr_with_seg_len_pair_t):
30338         Move from ...
30339         * tree-data-ref.h (struct dr_with_seg_len, dr_with_seg_len_pair_t):
30340         ... to here.
30341         (prune_runtime_alias_test_list): New decalaration.
30343 2017-05-26  Bin Cheng  <bin.cheng@arm.com>
30345         * tree-vect-data-refs.c (compare_tree): Rename and move ...
30346         * tree-data-ref.c (data_ref_compare_tree): ... to here.
30347         * tree-data-ref.h (data_ref_compare_tree): New decalaration.
30348         * tree-vect-data-refs.c (dr_group_sort_cmp): Update uses.
30349         (operator==, comp_dr_with_seg_len_pair): Ditto.
30350         (vect_prune_runtime_alias_test_list): Ditto.
30352 2017-05-26  Martin Liska  <mliska@suse.cz>
30354         PR ipa/80663
30355         * params.def: Bound partial-inlining-entry-probability param.
30357 2017-05-26  Marek Polacek  <polacek@redhat.com>
30359         PR sanitizer/80875
30360         * fold-const.c (fold_binary_loc) <case MULT_EXPR>: Check if OP1
30361         can be negated.
30363 2017-05-26  Richard Biener  <rguenther@suse.de>
30365         PR tree-optimization/80842
30366         * tree-ssa-ccp.c (set_lattice_value): Always meet with the old
30367         value.
30369 2017-05-26  Richard Biener  <rguenther@suse.de>
30371         PR tree-optimization/80844
30372         * tree-vectorizer.c (adjust_simduid_builtins): Propagate results.
30374 2017-05-25  Sebastian Peryt  <sebastian.peryt@intel.com>
30376         * doc/md.texi (Machine Constraints): Update x86 family
30377         machine constraints section to match 'config/i386/constraints.md'.
30379 2017-05-25  Volker Reichelt  <v.reichelt@netcologne.de>
30381         * doc/invoke.texi (-Wcatch-value=): Document new warning option.
30383 2017-05-25  Nathan Sidwell  <nathan@acm.org>
30385         * doc/invoke.texi (--enable-languages): Update documentation.
30387 2017-05-25  Martin Liska  <mliska@suse.cz>
30389         * dumpfile.c: Add TDF_FOLDING.
30390         * dumpfile.h (enum dump_kind): Likewise.
30391         * genmatch.c (dt_simplify::gen_1): Use it.
30393 2017-05-25  Marc Glisse  <marc.glisse@inria.fr>
30395         * match.pd (view_convert (convert@0 @1)): Handle zero-extension.
30397 2017-05-25  Marc Glisse  <marc.glisse@inria.fr>
30399         * match.pd ((A +- CST1) +- CST2): Allow some conversions.
30400         * tree.c (drop_tree_overflow): Handle COMPLEX_CST and VECTOR_CST.
30402 2017-05-25  Marc Glisse  <marc.glisse@inria.fr>
30404         * fold-const.c (fold_binary_loc) [(A & C) == D]: Remove transformation.
30405         * match.pd (X == C): Rewrite it here.
30406         (with_possible_nonzero_bits, with_possible_nonzero_bits2,
30407         with_certain_nonzero_bits2): New predicates.
30408         * tree-ssanames.c (get_nonzero_bits): Handle INTEGER_CST.
30410 2017-05-24  Nathan Sidwell  <nathan@acm.org>
30412         * lto-streamer-in.c (lto_input_data_block): Adjust T const cast to
30413         avoid warning.
30415         * auto-profile.c (afdo_propagate): Adjust T const cast to avoid
30416         warning.
30418 2017-05-24  Segher Boessenkool  <segher@kernel.crashing.org>
30420         * config/powerpcspe: New port.  Files are copied from the rs6000
30421         port, with "rs6000" in filenames replaced by "powerpcspe".
30423 2017-05-24  Wilco Dijkstra  <wdijkstr@arm.com>
30425         PR rtl-optimization/80754
30426         * lra-remat.c (do_remat): Add overlap checks for dst_regno.
30428 2017-05-24  Sheldon Lobo  <smlobo@sheldon.us.oracle.com>
30430         * config/sparc/sparc.md (length): Return the correct value for -mflat
30431         sibcalls to match output_sibcall.
30433 2017-05-24  Segher Boessenkool  <segher@kernel.crashing.org>
30435         PR bootstrap/80860
30436         PR bootstrap/80843
30437         * config/rs6000/rs6000.c (struct machine_function): Add new field
30438         n_components.
30439         (rs6000_get_separate_components): Init that field, use it.
30440         (rs6000_components_for_bb): Use the field.
30442 2017-05-24  Thomas Preud'homme  <thomas.preudhomme@arm.com>
30444         * config/arm/arm.c (arm_expand_prologue): Fix typo in comment.
30446 2017-05-24  Peter Bergner  <bergner@vnet.ibm.com>
30448         PR middle-end/80823
30449         * tree-cfg.c (group_case_labels_stmt): Delete increment of "i";
30451 2017-05-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
30453         PR target/80725
30454         * config/s390/s390.c (s390_check_qrst_address): Check incoming
30455         address against address_operand predicate.
30456         * config/s390/s390.md ("*indirect_jump"): Swap alternatives.
30458 2017-05-24  Eric Botcazou  <ebotcazou@adacore.com>
30460         * var-tracking.c (track_expr_p): Do not return 0 for tracked record
30461         parameters passed indirectly.
30463 2017-05-23  Uros Bizjak  <ubizjak@gmail.com>
30465         * config/i386/i386.md (*movdi_internal): Remove SSE4
30466         alternative 18 (?r, *v).  Update insn attributes.
30467         (*movsi_internal): Remove SSE4 alternative 13 (?r, *v).
30468         Update insn attributes.
30469         (*zero_extendsidi2): Remove SSE4 alternative (?r, *x).
30470         Update insn attributes.
30471         * config/i386/sse.md (vec_extract<ssevecmodelower>_0): Remove SSE4
30472         alternative 1 (r, v). Remove isa attribute.
30473         * config/i386/i386.c (dimode_scalar_chain::make_vector_copies):
30474         Always move value through stack for !TARGET_INTER_UNIT_MOVES_TO_VEC
30475         and !TARGET_INTER_UNIT_MOVES_TO_VEC targets.
30477 2017-05-23  Tom de Vries  <tom@codesourcery.com>
30479         * doc/sourcebuild.texi (Directives, Verify compiler message): Document
30480         dg-line directive.
30482 2017-05-23  Jan Hubicka  <hubicka@ucw.cz>
30484         * cgraphunit.c (symbol_table::process_new_functions): Update.
30485         * ipa-fnsummary.c (pass_data_inline_parameters): Remove.
30486         (inline_generate_summary): Rename to ...
30487         (ipa_fn_summary_generate): ... this one.
30488         (inline_read_summary): Rename to ...
30489         (ipa_fn_summary_read): ... this one.
30490         (inline_write_summary): Rename to ...
30491         (ipa_fn_summary_write): ... this one.
30492         (inline_free_summary): Rename to ...
30493         (ipa_free_fn_summary): ... this one.
30494         (pass_data_local_fn_summary, pass_local_fn_summary,
30495         make_pass_local_fn_summary, pass_data_ipa_free_fn_summary,
30496         pass_ipa_free_fn_summary, make_pass_ipa_free_fn_summary,
30497         pass_data_ipa_fn_summary, pass_ipa_fn_summary,
30498         make_pass_ipa_fn_summary): New.
30499         * ipa-fnsummary.h (inline_generate_summary, inline_read_summary,
30500         inline_write_summary, inline_free_summary): Remove.
30501         (ipa_free_fn_summary) : New.
30502         * ipa-inline.c (ipa_inline): Update.
30503         (pass_ipa_inline): Do not generate summaries.
30504         * ipa.c (pass_data_ipa_free_fn_summary, pass_ipa_free_fn_summary):
30505         Remove.
30506         * passes.def: Replace pass_inline_parameters by pass_local_fn_summary
30507         and add pass_ipa_fn_summary.
30508         * tree-pass.h (make_pass_ipa_fn_summary, make_pass_local_fn_summary):
30509         New.
30510         (make_pass_inline_parameters): Remove.
30512 2017-05-23  Thomas Schwinge  <thomas@codesourcery.com>
30514         * omp-low.c (struct omp_context): Remove "default_kind" member.
30515         Adjust all users.
30517         * omp-offload.c (execute_oacc_device_lower): Remove the
30518         parallelism dimensions function attributes for unparallelized
30519         OpenACC kernels constructs.
30521 2017-05-23  Martin Liska  <mliska@suse.cz>
30523         * cgraph.c (cgraph_node::get_create): Use symtab_node::dump_{asm_,}name
30524         functions.
30525         (cgraph_edge::make_speculative): Likewise.
30526         (cgraph_edge::resolve_speculation): Likewise.
30527         (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
30528         (cgraph_node::dump): Likewise.
30529         * cgraph.h: Likewise.
30530         * cgraphunit.c (analyze_functions): Likewise.
30531         (symbol_table::compile): Likewise.
30532         * ipa-cp.c (print_all_lattices): Likewise.
30533         (determine_versionability): Likewise.
30534         (initialize_node_lattices): Likewise.
30535         (ipcp_verify_propagated_values): Likewise.
30536         (estimate_local_effects): Likewise.
30537         (update_profiling_info): Likewise.
30538         (create_specialized_node): Likewise.
30539         (perhaps_add_new_callers): Likewise.
30540         (decide_about_value): Likewise.
30541         (decide_whether_version_node): Likewise.
30542         (identify_dead_nodes): Likewise.
30543         (ipcp_store_bits_results): Likewise.
30544         * ipa-devirt.c (dump_targets): Likewise.
30545         (ipa_devirt): Likewise.
30546         * ipa-icf.c (sem_item::dump): Likewise.
30547         (sem_function::equals): Likewise.
30548         (sem_variable::equals): Likewise.
30549         (sem_item_optimizer::read_section): Likewise.
30550         (sem_item_optimizer::execute): Likewise.
30551         (congruence_class::dump): Likewise.
30552         * ipa-inline-analysis.c (dump_ipa_call_summary): Likewise.
30553         (dump_inline_summary): Likewise.
30554         (estimate_node_size_and_time): Likewise.
30555         (inline_analyze_function): Likewise.
30556         * ipa-inline-transform.c (inline_call): Likewise.
30557         * ipa-inline.c (report_inline_failed_reason): Likewise.
30558         (want_early_inline_function_p): Likewise.
30559         (edge_badness): Likewise.
30560         (update_edge_key): Likewise.
30561         (inline_small_functions): Likewise.
30562         * ipa-profile.c (ipa_profile): Likewise.
30563         * ipa-prop.c (ipa_print_node_jump_functions): Likewise.
30564         (ipa_make_edge_direct_to_target): Likewise.
30565         (remove_described_reference): Likewise.
30566         (ipa_impossible_devirt_target): Likewise.
30567         (propagate_controlled_uses): Likewise.
30568         (ipa_print_node_params): Likewise.
30569         (ipcp_transform_function): Likewise.
30570         * ipa-pure-const.c (pure_const_read_summary): Likewise.
30571         (propagate_pure_const): Likewise.
30572         * ipa-reference.c (generate_summary): Likewise.
30573         (read_write_all_from_decl): Likewise.
30574         (propagate): Likewise.
30575         (ipa_reference_read_optimization_summary): Likewise.
30576         * ipa-utils.c (ipa_merge_profiles): Likewise.
30577         * ipa.c (walk_polymorphic_call_targets): Likewise.
30578         (symbol_table::remove_unreachable_nodes): Likewise.
30579         (ipa_single_use): Likewise.
30580         * passes.c (execute_todo): Likewise.
30581         * predict.c (drop_profile): Likewise.
30582         * symtab.c (symtab_node::get_dump_name): New function.
30583         (symtab_node::dump_name): Likewise.
30584         (symtab_node::dump_asm_name): Likewise.
30585         (symtab_node::dump_references): Likewise.
30586         (symtab_node::dump_referring): Likewise.
30587         (symtab_node::dump_base): Likewise.
30588         (symtab_node::debug_symtab): Likewise.
30589         * tree-sra.c (convert_callers_for_node): Likewise.
30590         * tree-ssa-structalias.c (ipa_pta_execute): Likewise.
30591         * value-prof.c (init_node_map): Likewise.
30593 2017-05-23  Martin Liska  <mliska@suse.cz>
30595         * cgraph.h: Move symtab_node::dump_table to symbol_table::dump
30596         and symtab_node::debug_symtab to symbol_table::debug.
30597         * cgraphunit.c (analyze_functions): Use the renamed function.
30598         (symbol_table::compile): Likewise.
30599         * ipa-cp.c (ipcp_verify_propagated_values): Likewise.
30600         * ipa-icf.c (sem_item_optimizer::execute): Likewise.
30601         * passes.c (execute_todo): Likewise.
30602         * symtab.c (symbol_table::dump): New function.
30603         * tree-ssa-structalias.c (ipa_pta_execute): Use the renamed function.
30605 2017-05-23  Jan Hubicka  <hubicka@ucw.cz>
30607         * ipa-fnsummary.c (estimate_node_size_and_time): Do not sanity check
30608         that nonconst implies exec.
30610 2017-05-23  Jan Hubicka  <hubicka@ucw.cz>
30612         * ipa-inline-analysis.c (cgraph_2edge_hook_list, cgraph_edge_hook_list,
30613         inline_edge_removal_hook, inline_edge_duplication_hook): Remove.
30614         (inline_edge_summary_vec): Turn into ...
30615         (ipa_call_summaries): ... this one.
30616         (redirect_to_unreachable, edge_set_predicate,
30617         evaluate_properties_for_edge, inline_summary_alloc,
30618         reset_ipa_call_summary, reset_inline_summary,
30619         inline_summary_t::duplicate): Update.
30620         (inline_edge_duplication_hook): Turn to ...
30621         (ipa_call_summary_t::duplicate): ... this one.
30622         (inline_edge_removal_hook): Turn to ...
30623         (ipa_call_summary_t::remove): ... this one.
30624         (dump_inline_edge_summary): Turn to ...
30625         (dump_ipa_call_summary): ... this one.
30626         (estimate_function_body_sizes): Update.
30627         (inline_update_callee_summaries): Update.
30628         (remap_edge_change_prob): Update.
30629         (remap_edge_summaries): Update.
30630         (inline_merge_summary): Update.
30631         (do_estimate_edge_time): Update.
30632         (inline_generate_summary): Update.
30633         (inline_read_section): Update.
30634         (inline_read_summary): Update.
30635         (inline_free_summary): Update.
30636         * ipa-inline.c (can_inline_edge_p): Update.
30637         (compute_inlined_call_time): Update.
30638         (want_inline_small_function_p): Update.
30639         (edge_badness): Update.
30640         (early_inliner): Update.
30641         * ipa-inline.h (inline_edge_summary): Turn to ...
30642         (ipa_call_summary): ... this one.
30643         (ipa_call_summary_t): New class.
30644         (inline_edge_summary_t, inline_edge_summary_vec): Remove.
30645         (ipa_call_summaries): New.
30646         (inline_edge_summary): Remove.
30647         (estimate_edge_growth): Update.
30648         * ipa-profile.c (ipa_propagate_frequency_1): Update.
30649         * ipa-prop.c (ipa_make_edge_direct_to_target): Update.
30650         * ipa-split.c (execute_split_functions): Update.
30651         * ipa.c (symbol_table::remove_unreachable_nodes): Update.
30653 2017-05-23  Tom de Vries  <tom@codesourcery.com>
30655         * doc/sourcebuild.texi (Effective-Target Keywords, Other hardware
30656         attributes): Document rdrand effective target.
30658 2017-05-23  Tom de Vries  <tom@codesourcery.com>
30660         * doc/sourcebuild.texi (Effective-Target Keywords, Other hardware
30661         attributes): Sort alphabetically.
30663 2017-05-23  Georg-Johann Lay  <avr@gjlay.de>
30665         * config/avr/genmultilib.awk: Use gsub instead of gensub.
30667 2017-05-22  Michael Meissner  <meissner@linux.vnet.ibm.com>
30669         PR target/80718
30670         * config/rs6000/vsx.md (vsx_splat_<mode>, VSX_D iterator): Split
30671         V2DF/V2DI splat into two separate patterns, one that handles
30672         registers, and the other that only handles memory.  Drop support
30673         for splatting from a GPR on ISA 2.07 and then splitting the
30674         splat into direct move and splat.
30675         (vsx_splat_<mode>_reg): Likewise.
30676         (vsx_splat_<mode>_mem): Likewise.
30678 2017-05-22  Segher Boessenkool  <segher@kernel.crashing.org>
30680         * cfgcleanup.c (bb_is_just_return): Allow CLOBBERs.
30682 2017-05-22  Jakub Jelinek  <jakub@redhat.com>
30684         PR middle-end/80809
30685         * omp-low.c (finish_taskreg_remap): New function.
30686         (finish_taskreg_scan): If unit size of ctx->record_type
30687         is non-constant, unshare the size expression and replace
30688         decls in it with possible outer var refs.
30690         PR middle-end/80809
30691         * gimplify.c (omp_add_variable): For GOVD_DEBUG_PRIVATE use
30692         GOVD_SHARED rather than GOVD_PRIVATE with it.
30693         (gimplify_adjust_omp_clauses_1, gimplify_adjust_omp_clauses): Expect
30694         GOVD_SHARED rather than GOVD_PRIVATE with GOVD_DEBUG_PRIVATE.
30696         PR middle-end/80853
30697         * omp-low.c (lower_reduction_clauses): Pass OMP_CLAUSE_PRIVATE
30698         as last argument to build_outer_var_ref for pointer bases of array
30699         section reductions.
30701 2017-05-19  Martin Sebor  <msebor@redhat.com>
30703         * print-tree.c (print_node): Print DECL_READ_P flag.
30705 2017-05-22  Jan Hubicka  <hubicka@ucw.cz>
30707         * Makefile.in: Add ipa-fnsummary.o and ipa-fnsummary.h
30708         * auto-profile.c: Replace ipa-inline.h by ipa-fnsummary.h
30709         * cgraph.c: Likewise.
30710         * cgraphunit.c: Likewise.
30711         * gengtype.c: Likewise.
30712         * ipa-cp.c: Likewise.
30713         * ipa-devirt.c: Likewise.
30714         * ipa-icf.c: Likewise.
30715         * ipa-predicate.c: Likewise.
30716         * ipa-profile.c: Likewise.
30717         * ipa-prop.c: Likewise.
30718         * ipa-split.c: Likewise.
30719         * ipa.c: Likewise.
30720         * ipa-inline-analysis.c (inline_summaries, ipa_call_summaries,
30721         edge_predicate_pool, dump_inline_hints,
30722         inline_summary::account_size_time, redirect_to_unreachable,
30723         edge_set_predicate, set_hint_predicate,
30724         evaluate_conditions_for_known_args, evaluate_properties_for_edge,
30725         inline_summary_alloc, ipa_call_summary::reset, inline_summary::reset,
30726         inline_summary_t::remove, remap_hint_predicate_after_duplication,
30727         inline_summary_t::duplicate, ipa_call_summary_t::duplicate,
30728         ipa_call_summary_t::remove, initialize_growth_caches,
30729         free_growth_caches, dump_ipa_call_summary, dump_inline_summary,
30730         debug_inline_summary, dump_inline_summaries, initialize_inline_failed,
30731         mark_modified, unmodified_parm_1, unmodified_parm,
30732         unmodified_parm_or_parm_agg_item, eliminated_by_inlining_prob,
30733         set_cond_stmt_execution_predicate, set_switch_stmt_execution_predicate,
30734         compute_bb_predicates, will_be_nonconstant_expr_predicate,
30735         will_be_nonconstant_predicate, record_modified_bb_info,
30736         get_minimal_bb, record_modified, param_change_prob,
30737         phi_result_unknown_predicate, predicate_for_phi_result,
30738         array_index_predicate, clobber_only_eh_bb_p, fp_expression_p,
30739         estimate_function_body_sizes, compute_inline_parameters,
30740         compute_inline_parameters_for_curren, pass_data_inline_parameters,
30741         estimate_node_size_and_time, estimate_ipcp_clone_size_and_time,
30742         inline_update_callee_summaries, remap_edge_change_prob,
30743         remap_edge_summaries, remap_hint_predicate, inline_merge_summary,
30744         inline_update_overall_summary, inline_indirect_intraprocedural_analysis,
30745         inline_analyze_function, inline_summary_t::insert,
30746         inline_generate_summary, read_ipa_call_summary, inline_read_section,
30747         inline_read_summary, write_ipa_call_summary, inline_write_summary,
30748         inline_free_summary): Move to ipa-fnsummary.h
30749         (predicate_t): Remove.
30750         * ipa-fnsummary.c: New file.
30751         * ipa-inline.h:  Do not include sreal.h and ipa-predicate.h
30752         (enum inline_hints_vals, inline_hints, agg_position_info,
30753         INLINE_SIZE_SCALE, size_time_entry, inline_summary, inline_summary_t,
30754         inline_summaries, ipa_call_summary, ipa_call_summary_t,
30755         ipa_call_summaries, debug_inline_summary, dump_inline_summaries,
30756         dump_inline_summary, dump_inline_hints, inline_generate_summary,
30757         inline_read_summary, inline_write_summary, inline_free_summary,
30758         inline_analyze_function, initialize_inline_failed,
30759         inline_merge_summary, inline_update_overall_summary,
30760         compute_inline_parameters): Move to ipa-fnsummary.h
30761         * ipa-fnsummary.h: New file.
30762         * ipa-inline-transform.h: Include ipa-inline.h.
30763         * ipa-inline.c: LIkewise.
30765 2017-05-22  Jan Hubicka  <hubicka@ucw.cz>
30767         * ipa-inline.c (edge_badness): Use inlined_time instead of
30768         inline_summaries->get.
30770 2017-05-22  Jan Hubicka  <hubicka@ucw.cz>
30772         * ipa-inline.c (edge_badness): Use estimate_size_after_inlining.
30774 2017-05-22  Nathan Sidwell  <nathan@acm.org>
30776         * doc/invoke.texi (fdump-translation-unit): Delete documentation.
30777         (fdump-lang): Document 'raw' option.
30778         * dumpfile.h (TDI_tu): Delete.
30779         * dumpfile.c (dump_files): Remove translation-unit.
30780         (FIRST_AUTO_NUMBERED_DUMP): Decrement.
30782 2017-05-22  Georg-Johann Lay  <avr@gjlay.de>
30784         * config/avr/t-avr (AWK) [t-multilib]: Remove "-v FORMAT=Makefile"
30785         command option from $(AWK) call.
30786         * config/avr/genmultilib.awk: Simplify and rewrite so that it
30787         generates MULTILIB_REQUIRED instead of MULTILIB_EXCEPTIONS.
30788         [FORMAT]: Remove handling of variable.
30789         * config/avr/t-multilib: Regenerate.
30791 2017-05-22  Jan Hubicka  <hubicka@ucw.cz>
30793         * ipa-inline-analysis.c (inline_summary::reset): Do not reset
30794         self_time.
30795         (dump_inline_summary): Do not print self_time.
30796         (estimate_function_body_sizes): Do not set self_time.
30797         (compute_inline_parameters): Likewise.
30798         (inline_read_section, inline_write_summary): Do not stream self_time.
30799         * ipa-inline.h (inline_summary): Drop self_time.
30801 2017-05-22  Jan Hubicka  <hubicka@ucw.cz>
30803         * ipa-inline-analysis.c (account_size_time): Rename to ...
30804         (inline_summary::account_size_time): ... this one.
30805         (reset_ipa_call_summary): Turn to ...
30806         (ipa_call_summary::reset): ... this one.
30807         (reset_inline_summary): Turn to ...
30808         (inline_summary::reset): ... this one.
30809         (inline_summary_t::remove): Update.
30810         (inline_summary_t::duplicate): Update.
30811         (ipa_call_summary_t::remove): Update.
30812         (dump_inline_summary): Update.
30813         (estimate_function_body_sizes): Update.
30814         (compute_inline_parameters): Update.
30815         (estimate_node_size_and_time): Update.
30816         (inline_merge_summary): Update.
30817         (inline_update_overall_summary): Update.
30818         (inline_read_section): Update.
30819         (inline_write_summary): Update.
30820         * ipa-inline.h (inline_summary): Rename entry to size_time_table;
30821         add account_size_time and reset member functions.
30822         (ipa_call_summary): Add reset function.
30823         * ipa-predicate.h (predicate::operator &): Constify.
30825 2017-05-22  Richard Biener  <rguenther@suse.de>
30827         * df-scan.c (df_insn_refs_verify): Speedup when not verifying.
30829 2017-05-19  Jason Merrill  <jason@redhat.com>
30831         * tree.c (make_tree_vec_stat, grow_tree_vec_stat): Use size_t.
30833 2017-05-19  Marek Polacek  <polacek@redhat.com>
30835         PR sanitizer/80800
30836         * fold-const.c (extract_muldiv_1) <case TRUNC_DIV_EXPR>: Add
30837         TYPE_OVERFLOW_WRAPS checks.
30839 2017-05-19  Thomas Schwinge  <thomas@codesourcery.com>
30841         * tree-core.h (enum omp_clause_default_kind): Add
30842         "OMP_CLAUSE_DEFAULT_PRESENT".
30843         * tree-pretty-print.c (dump_omp_clause): Handle it.
30844         * gimplify.c (enum gimplify_omp_var_data): Add
30845         "GOVD_MAP_FORCE_PRESENT".
30846         (gimplify_adjust_omp_clauses_1): Map it to
30847         "GOMP_MAP_FORCE_PRESENT".
30848         (oacc_default_clause): Handle "OMP_CLAUSE_DEFAULT_PRESENT".
30850         * gimplify.c (oacc_default_clause): Clarify.
30852 2017-05-19  Nathan Sidwell  <nathan@acm.org>
30854         LANG_HOOK_REGISTER_DUMPS
30855         * toplev.c (general_init): Call register dump lang hook.
30856         * doc/invoke.texi: Document -fdump-lang option family.
30857         * dumpfile.c (dump_files): Remove class dump here.
30858         (FIRST_AUTO_NUMBERED_DUMP): Adjust.
30859         * dumpfile.h (tree_dump_index): Remove TDI_class.
30860         * langhooks-def.h (lhd_register_dumps): Declare.
30861         (LANG_HOOKS_REGISTER_DUMPS): Define.
30862         (LANG_HOOKS_INITIALIZER): Add it.
30863         * langhooks.c (lhd_register_dumps): Define.
30864         * langhooks.h (struct lang_hooks): Add register_dumps.
30866 2017-05-19  Nathan Sidwell  <nathan@acm.org>
30868         * context.h (context::set_passes): New.
30869         * context.c (context::context): Do not create pass manager.
30870         * toplev.c (general_init): Create pass manager here.
30872 2017-05-19  Segher Boessenkool  <segher@kernel.crashing.org>
30874         * config/rs6000/rs6000.md (splitter to load of -1 and mask): Don't
30875         use this splitter if two add or or instructions would also work for
30876         the constant we want to generate.
30878 2017-05-19  Richard Biener  <rguenther@suse.de>
30880         PR build/80821
30881         * genmatch.c (dt_node::gen_kids_1): Add missing scope around
30882         predicate evaluation.
30884 2017-05-19  Jan Hubicka  <hubicka@ucw.cz>
30886         * ipa-inline.h (ipa_call_summary): Turn sizes into signed;
30887         add ctor.
30888         * ipa-inline.c (want_inline_small_function_p): Do not cast to
30889         unsigned.
30891 2017-05-19  Jan Hubicka  <hubicka@ucw.cz>
30893         * ipa-inline-analysis.c (cgraph_2edge_hook_list, cgraph_edge_hook_list,
30894         inline_edge_removal_hook, inline_edge_duplication_hook): Remove.
30895         (inline_edge_summary_vec): Turn into ...
30896         (ipa_call_summaries): ... this one.
30897         (redirect_to_unreachable, edge_set_predicate,
30898         evaluate_properties_for_edge, inline_summary_alloc,
30899         reset_ipa_call_summary, reset_inline_summary,
30900         inline_summary_t::duplicate): Update.
30901         (inline_edge_duplication_hook): Turn to ...
30902         (ipa_call_summary_t::duplicate): ... this one.
30903         (inline_edge_removal_hook): Turn to ...
30904         (ipa_call_summary_t::remove): ... this one.
30905         (dump_inline_edge_summary): Turn to ...
30906         (dump_ipa_call_summary): ... this one.
30907         (estimate_function_body_sizes): Update.
30908         (inline_update_callee_summaries): Update.
30909         (remap_edge_change_prob): Update.
30910         (remap_edge_summaries): Update.
30911         (inline_merge_summary): Update.
30912         (do_estimate_edge_time): Update.
30913         (inline_generate_summary): Update.
30914         (inline_read_section): Update.
30915         (inline_read_summary): Update.
30916         (inline_free_summary): Update.
30917         * ipa-inline.c (can_inline_edge_p): Update.
30918         (compute_inlined_call_time): Update.
30919         (want_inline_small_function_p): Update.
30920         (edge_badness): Update.
30921         (early_inliner): Update.
30922         * ipa-inline.h (inline_edge_summary): Turn to ...
30923         (ipa_call_summary): ... this one.
30924         (ipa_call_summary_t): New class.
30925         (inline_edge_summary_t, inline_edge_summary_vec): Remove.
30926         (ipa_call_summaries): New.
30927         (inline_edge_summary): Remove.
30928         (estimate_edge_growth): Update.
30929         * ipa-profile.c (ipa_propagate_frequency_1): Update.
30930         * ipa-prop.c (ipa_make_edge_direct_to_target): Update.
30931         * ipa-split.c (execute_split_functions): Update.
30932         * ipa.c (symbol_table::remove_unreachable_nodes): Update.
30934 2017-05-19  Richard Biener  <rguenther@suse.de>
30936         PR middle-end/80764
30937         * cfgexpand.c (expand_gimple_cond): Fix loop fixup.
30939 2017-05-18  Segher Boessenkool  <segher@kernel.crashing.org>
30941         * config/rs6000/rs6000.c (struct machine_function): Add field
30942         fpr_is_wrapped_separately.
30943         (rs6000_get_separate_components): Use 64 components.  Handle the
30944         new FPR components.
30945         (rs6000_components_for_bb): Handle the FPR components.
30946         (rs6000_emit_prologue_components): Handle the FPR components.
30947         (rs6000_emit_epilogue_components): Handle the FPR components.
30948         (rs6000_set_handled_components): Handle the FPR components.
30949         (rs6000_emit_prologue): Don't output prologue code for those FPRs
30950         that are already separately shrink-wrapped.
30951         (rs6000_emit_epilogue): Don't output epilogue code for those FPRs
30952         that are already separately shrink-wrapped.
30954 2017-05-18  Michael Meissner  <meissner@linux.vnet.ibm.com>
30956         PR target/80510
30957         * config/rs6000/predicates.md (simple_offsettable_mem_operand):
30958         New predicate.
30960         * config/rs6000/rs6000.md (ALTIVEC_DFORM): New iterator.
30961         (define_peephole2 for Altivec d-form load): Add peepholes to catch
30962         cases where the register allocator uses a move and an offsettable
30963         memory operation to/from a FPR register on ISA 2.06/2.07.
30964         (define_peephole2 for Altivec d-form store): Likewise.
30966 2017-05-18  Uros Bizjak  <ubizjak@gmail.com>
30968         PR target/80799
30969         * config/i386/mmx.md (*mov<mode>_internal): Enable
30970         alternatives 11, 12, 13 and 14 also for 32bit targets.
30971         Remove alternatives 15, 16, 17 and 18.
30972         * config/i386/sse.md (vec_concatv2di): Change
30973         alternative (!x, *y) to (x, ?!*Yn).
30975 2017-05-18  Paolo Carlini  <paolo.carlini@oracle.com>
30977         * dumpfile.h (enum dump_kind): Remove stray comma.
30979 2017-05-18  Jan Hubicka  <hubicka@ucw.cz>
30981         * Makefile.in: Add ipa-predicate.o and ipa-predicate.h
30982         * ipa-inline-analysis.c (NUM_CONDITIONS): turn into
30983         predicate::num_conditions
30984         (IS_NOT_CONSTANT): turn into predicate::is_not_constant.
30985         (CHANGED): turn into predicate::changed.
30986         (agg_position_info): Move to ipa-predicate.h
30987         (add_condition, predicate::add_clause, predicate::operator &=,
30988         predicate::or_with, predicate::evaluate, predicate::probability,
30989         dump_condition, dump_clause, predicate::dump,
30990         predicate::remap_after_duplication, predicate::remap_after_inlining,
30991         predicate::stream_in, predicate::stream_out): Move to ipa-predicate.c
30992         (evaluate_conditions_for_known_args): Update.
30993         (set_cond_stmt_execution_predicate): Update.
30994         * ipa-inline.h: Include ipa-predicate.h
30995         (condition, inline_param_summary, conditions, agg_position_info,
30996         predicate): Move to ipa-predicate.h
30997         * ipa-predicate.c: New file.
30998         * ipa-predicate.h: New file.
31000 2017-05-18  Wilco Dijkstra  <wdijkstr@arm.com>
31002         * final.c (leaf_function_p): Check we are not in a sequence.
31004 2017-05-18  Martin Liska  <mliska@suse.cz>
31006         * cfgrtl.c (rtl_verify_edges): Remove usage of TDF_RTL.
31007         * dumpfile.c (dump_register): Use new enum dump_kind.
31008         (get_dump_file_name): Likewise.
31009         (dump_enable_all): Likewise.
31010         (dump_switch_p_1): Likewise.
31011         (enable_rtl_dump_file): Remove usage of TDF_RTL.
31012         * dumpfile.h (enum dump_kind): New enum type.
31013         (struct dump_file_info): Create constructor and
31014         format fields and comments.
31015         * passes.c (pass_manager::register_one_dump_file):
31016         Use num dump_kind.
31017         * statistics.c (statistics_early_init): Likewise.
31018         * tree-ssa-loop-prefetch.c (dump_mem_details): Replace
31019         TDF_TREE with TDF_SLIM.
31020         (gather_memory_references_ref): Likewise.
31022 2017-05-18  Martin Liska  <mliska@suse.cz>
31024         * vec.h (struct vnull): Use it.
31026 2017-05-18  Jan Hubicka  <hubicka@ucw.cz>
31028         * ipa-inline-analysis.c (predicate_conditions): Move to ipa-inline.h
31029         (true_predicate, false_predicate, true_predicate_p,
31030         false_predicate_p): Remove.
31031         (single_cond_predicate, not_inlined_predicate): Turn to member function
31032         in ipa-inline.h
31033         (add_condition): Update.
31034         (add_clause): Turn to...
31035         (predicate::add_clause): ... this one; update; allow passing NULL
31036         as parameter.
31037         (and_predicates): Turn to ...
31038         (predicate::operator &=): ... this one.
31039         (predicates_equal_p): Move to predicate::operator == in ipa-inline.h
31040         (or_predicates): Turn to ...
31041         (predicate::or_with): ... this one.
31042         (evaluate_predicate): Turn to ...
31043         (predicate::evaluate): ... this one.
31044         (predicate_probability): Turn to ...
31045         (predicate::probability): ... this one.
31046         (dump_condition): Update.
31047         (dump_predicate): Turn to ...
31048         (predicate::dump): ... this one.
31049         (account_size_time): Update.
31050         (edge_set_predicate): Update.
31051         (set_hint_predicate): UPdate.
31052         (evaluate_conditions_for_known_args): Update.
31053         (evaluate_properties_for_edge): Update.
31054         (remap_predicate_after_duplication): Turn to...
31055         (predicate::remap_after_duplication): ... this one.
31056         (remap_hint_predicate_after_duplication): Update.
31057         (inline_summary_t::duplicate): UPdate.
31058         (dump_inline_edge_summary): Update.
31059         (dump_inline_summary): Update.
31060         (set_cond_stmt_execution_predicate): Update.
31061         (set_switch_stmt_execution_predicate): Update.
31062         (compute_bb_predicates): Update.
31063         (will_be_nonconstant_expr_predicate): Update.
31064         (will_be_nonconstant_predicate): Update.
31065         (phi_result_unknown_predicate): Update.
31066         (predicate_for_phi_result): Update.
31067         (array_index_predicate): Update.
31068         (estimate_function_body_sizes): Update.
31069         (estimate_node_size_and_time): Update.
31070         (estimate_ipcp_clone_size_and_time): Update.
31071         (remap_predicate): Rename to ...
31072         (predicate::remap_after_inlining): ... this one.
31073         (remap_hint_predicate): Update.
31074         (inline_merge_summary): Update.
31075         (inline_update_overall_summary): Update.
31076         (estimate_size_after_inlining): Update.
31077         (read_predicate): Rename to ...
31078         (predicate::stream_in): ... this one.
31079         (read_inline_edge_summary): Update.
31080         (write_predicate): Rename to ...
31081         (predicate::stream_out): ... this one.
31082         (write_inline_edge_summary): Update.
31083         * ipa-inline.h (MAX_CLAUSES): Turn to predicate::max_clauses.
31084         (clause_t): Turn to uint32_t
31085         (predicate): Turn to class; implement constructor and operators
31086         ==, !=, &
31087         (size_time_entry): Update.
31088         (inline_summary): Update.
31089         (inline_edge_summary): Update.
31091 2017-05-18  Marc Glisse  <marc.glisse@inria.fr>
31093         * fold-const.c (fold_binary_loc): Move transformation...
31094         * match.pd (C - X CMP X): ... here.
31096 2017-05-18  Sheldon Lobo  <sheldon.lobo@oracle.com>
31098         * config/sparc/sparc.c (sparc_option_override): Set function
31099         alignment for -mcpu=niagara7 to 64 to match the I$ line.
31100         * config/sparc/sparc.h (BRANCH_COST): Set the SPARC M7 branch
31101         latency to 1.
31102         * config/sparc/sparc.h (BRANCH_COST): Set the SPARC T4 branch
31103         latency to 2.
31104         * config/sparc/sol2.h: Fix a ASM_CPU32_DEFAULT_SPEC typo.
31106 2017-05-18  Marek Polacek  <polacek@redhat.com>
31108         PR sanitizer/80797
31109         * ubsan.c (instrument_null): Unwrap ADDR_EXPRs.
31110         (pass_ubsan::execute): Call gimple_assign_single_p instead of
31111         gimple_assign_load_p.
31113 2017-05-17  Segher Boessenkool  <segher@kernel.crashing.org>
31115         PR middle-end/80692
31116         * real.c (do_compare): Give decimal_do_compare preference over
31117         comparing just the signs.
31119 2017-05-17  Uros Bizjak  <ubizjak@gmail.com>
31121         * doc/md.texi (Canonicalization of Instructions): Describe the
31122         canonical form of instructions that inherently set a condition
31123         code register.
31125 2017-05-17  Peter Bergner  <bergner@vnet.ibm.com>
31127         PR middle-end/80775
31128         * tree-cfg.c: Move deletion of unreachable case statements to after
31129         the merging of consecutive case labels.
31131 2017-05-17  Thomas Preud'homme  <thomas.preudhomme@arm.com>
31133         * config/arm/arm.c (cmse_nonsecure_call_clear_caller_saved): Refer
31134         readers to __gnu_cmse_nonsecure_call libcall for saving, clearing and
31135         restoring of callee-saved registers.
31137 2017-05-17  Eric Botcazou  <ebotcazou@adacore.com>
31139         * compare-elim.c (equivalent_reg_at_start): Adjust outdated comment.
31140         * config/visium/visium.c (single_set_and_flags): Likewise.
31141         * config/visium/visium.md (Substitutions): Likewise.
31143 2017-05-17  Martin Liska  <mliska@suse.cz>
31145         * cfg.c: Introduce dump_flags_t type and
31146         use it instead of int type.
31147         * cfg.h: Likewise.
31148         * cfghooks.c: Likewise.
31149         * cfghooks.h (struct cfg_hooks): Likewise.
31150         * cfgrtl.c: Likewise.
31151         * cfgrtl.h: Likewise.
31152         * cgraph.c (cgraph_node::get_body): Likewise.
31153         * coretypes.h: Likewise.
31154         * domwalk.c: Likewise.
31155         * domwalk.h: Likewise.
31156         * dumpfile.c (struct dump_option_value_info): Likewise.
31157         (dump_enable_all): Likewise.
31158         (dump_switch_p_1): Likewise.
31159         (opt_info_switch_p): Likewise.
31160         * dumpfile.h (enum tree_dump_index): Likewise.
31161         (struct dump_file_info): Likewise.
31162         * genemit.c: Likewise.
31163         * generic-match-head.c: Likewise.
31164         * gengtype.c (open_base_files): Likewise.
31165         * gimple-pretty-print.c: Likewise.
31166         * gimple-pretty-print.h: Likewise.
31167         * graph.c (print_graph_cfg): Likewise.
31168         * graphite-scop-detection.c (dot_all_sese): Likewise.
31169         * ipa-devirt.c (build_type_inheritance_graph): Likewise.
31170         * loop-unroll.c (report_unroll): Likewise.
31171         * passes.c (pass_manager::register_one_dump_file): Likewise.
31172         * print-tree.c: Likewise.
31173         * statistics.c: Likewise.
31174         * tree-cfg.c: Likewise.
31175         * tree-cfg.h: Likewise.
31176         * tree-dfa.c: Likewise.
31177         * tree-dfa.h: Likewise.
31178         * tree-dump.c (dump_function): Likewise.
31179         * tree-dump.h (struct dump_info): Likewise.
31180         * tree-pretty-print.c: Likewise.
31181         * tree-pretty-print.h: Likewise.
31182         * tree-ssa-live.c: Likewise.
31183         * tree-ssa-live.h: Likewise.
31184         * tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Likewise.
31185         * tree-vect-loop.c: Likewise.
31186         * tree-vect-slp.c: Likewise.
31188 2017-05-16  James Greenhalgh  <james.greenhalgh@arm.com>
31189             Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
31191         PR tree-optimization/80457
31192         * tree-vect-stmts.c (vect_model_simple_cost): Model the cost
31193         of all arguments to a statement as scalar_to_vec operations.
31194         (vectorizable_call): Adjust call to vect_model_simple_cost for
31195         new parameter.
31196         (vectorizable_conversion): Likewise.
31197         (vectorizable_assignment): Likewise.
31198         (vectorizable_shift): Likewise.
31199         (vectorizable_operation): Likewise.
31200         (vectorizable_comparison): Likewise.
31201         (vect_is_simple_cond): Record the def types for operands.
31202         (vectorizable_condition): Likewise, call vect_model_simple_cost.
31203         * tree-vectorizer.h (vect_model_simple_cost): Add new parameter
31204         for statement argument count.
31206 2017-05-16  Carl Love  <cel@us.ibm.com>
31208         * config/rs6000/rs6000-c: Add support for built-in functions
31209         vector unsigned long long vec_bperm (vector unsigned long long,
31210                                              vector unsigned char)
31211         vector signed long long vec_mule (vector signed int,
31212                                           vector signed int)
31213         vector unsigned long long vec_mule (vector unsigned int,
31214                                             vector unsigned int)
31215         vector signed long long vec_mulo (vector signed int,
31216                                           vector signed int)
31217         vector unsigned long long vec_mulo (vector unsigned int,
31218                                             vector unsigned int)
31219         vector signed char vec_sldw (vector signed char,
31220                                      vector signed char,
31221                                      const int)
31222         vector unsigned char vec_sldw (vector unsigned char,
31223                                        vector unsigned char,
31224                                        const int)
31225         vector signed short vec_sldw (vector signed short,
31226                                       vector signed short,
31227                                       const int)
31228         vector unsigned short vec_sldw (vector unsigned short,
31229                                         vector unsigned short,
31230                                         const int)
31231         vector signed int vec_sldw (vector signed int,
31232                                     vector signed int,
31233                                     const int)
31234         vector unsigned int vec_sldw (vector unsigned int,
31235                                       vector unsigned int,
31236                                       const int)
31237         vector signed long long vec_sldw (vector signed long long,
31238                                           vector signed long long,
31239                                           const int)
31240         vector unsigned long long vec_sldw (vector unsigned long long,
31241                                             vector unsigned long long,
31242                                             const int)
31243         * config/rs6000/rs6000-c: Add support for built-in functions
31244         * config/rs6000/rs6000-builtin.def: Add definition for SLDW.
31245         * config/rs6000/altivec.h: Add defintion for vec_sldw.
31246         * doc/extend.texi: Update the built-in documentation for the
31247         new built-in functions.
31249 2017-05-16  Marek Polacek  <polacek@redhat.com>
31251         PR sanitizer/80536
31252         PR sanitizer/80386
31253         * tree.c (save_expr): Don't fold the expression.
31255 2017-05-16  Uros Bizjak  <ubizjak@gmail.com>
31257         * config/i386/i386.md (*movsi_internal): Split (?rm,*y) alternative
31258         to (?r,*Yn) and (?m,*y) alternatives, and (?*y,rm) to (?*Ym,r)
31259         and (?*y,m).  Update insn attributes.
31261 2017-05-16  Martin Liska  <mliska@suse.cz>
31263         * cgraph.c (cgraph_edge::resolve_speculation): Add default value for
31264         flags argument of print_gimple_stmt, print_gimple_expr,
31265         print_generic_stmt and print_generic_expr.
31266         * cgraphclones.c (symbol_table::materialize_all_clones): Likewise.
31267         * coretypes.h: Likewise.
31268         * except.c (dump_eh_tree): Likewise.
31269         * gimple-fold.c (gimple_fold_stmt_to_constant_1): Likewise.
31270         * gimple-pretty-print.h: Likewise.
31271         * gimple-ssa-backprop.c (dump_usage_prefix): Likewise.
31272         (backprop::push_to_worklist): Likewise.
31273         (backprop::pop_from_worklist): Likewise.
31274         (backprop::process_use): Likewise.
31275         (backprop::intersect_uses): Likewise.
31276         (note_replacement): Likewise.
31277         * gimple-ssa-store-merging.c
31278         (pass_store_merging::terminate_all_aliasing_chains): Likewise.
31279         (imm_store_chain_info::coalesce_immediate_stores): Likewise.
31280         (pass_store_merging::execute): Likewise.
31281         * gimple-ssa-strength-reduction.c (dump_candidate): Likewise.
31282         (ssa_base_cand_dump_callback): Likewise.
31283         (dump_incr_vec): Likewise.
31284         (replace_refs): Likewise.
31285         (replace_mult_candidate): Likewise.
31286         (create_add_on_incoming_edge): Likewise.
31287         (create_phi_basis): Likewise.
31288         (insert_initializers): Likewise.
31289         (all_phi_incrs_profitable): Likewise.
31290         (introduce_cast_before_cand): Likewise.
31291         (replace_one_candidate): Likewise.
31292         * gimplify.c (gimplify_expr): Likewise.
31293         * graphite-isl-ast-to-gimple.c (is_valid_rename): Likewise.
31294         (set_rename): Likewise.
31295         (rename_uses): Likewise.
31296         (copy_loop_phi_nodes): Likewise.
31297         (add_close_phis_to_merge_points): Likewise.
31298         (copy_loop_close_phi_args): Likewise.
31299         (copy_cond_phi_args): Likewise.
31300         (graphite_copy_stmts_from_block): Likewise.
31301         (translate_pending_phi_nodes): Likewise.
31302         * graphite-poly.c (print_pdr): Likewise.
31303         (dump_gbb_cases): Likewise.
31304         (dump_gbb_conditions): Likewise.
31305         (print_scop_params): Likewise.
31306         * graphite-scop-detection.c (build_cross_bb_scalars_def): Likewise.
31307         (build_cross_bb_scalars_use): Likewise.
31308         (gather_bbs::before_dom_children): Likewise.
31309         * hsa-dump.c (dump_hsa_immed): Likewise.
31310         * ipa-cp.c (print_ipcp_constant_value): Likewise.
31311         (get_replacement_map): Likewise.
31312         * ipa-inline-analysis.c (dump_condition): Likewise.
31313         (estimate_function_body_sizes): Likewise.
31314         * ipa-polymorphic-call.c (check_stmt_for_type_change): Likewise.
31315         (ipa_polymorphic_call_context::get_dynamic_type): Likewise.
31316         * ipa-prop.c (ipa_dump_param): Likewise.
31317         (ipa_print_node_jump_functions_for_edge): Likewise.
31318         (ipa_modify_call_arguments): Likewise.
31319         (ipa_modify_expr): Likewise.
31320         (ipa_dump_param_adjustments): Likewise.
31321         (ipa_dump_agg_replacement_values): Likewise.
31322         (ipcp_modif_dom_walker::before_dom_children): Likewise.
31323         * ipa-pure-const.c (check_stmt): Likewise.
31324         (pass_nothrow::execute): Likewise.
31325         * ipa-split.c (execute_split_functions): Likewise.
31326         * omp-offload.c (dump_oacc_loop_part): Likewise.
31327         (dump_oacc_loop): Likewise.
31328         * trans-mem.c (tm_log_emit): Likewise.
31329         (tm_memopt_accumulate_memops): Likewise.
31330         (dump_tm_memopt_set): Likewise.
31331         (dump_tm_memopt_transform): Likewise.
31332         * tree-cfg.c (gimple_verify_flow_info): Likewise.
31333         (print_loop): Likewise.
31334         * tree-chkp-opt.c (chkp_print_addr): Likewise.
31335         (chkp_gather_checks_info): Likewise.
31336         (chkp_get_check_result): Likewise.
31337         (chkp_remove_check_if_pass): Likewise.
31338         (chkp_use_outer_bounds_if_possible): Likewise.
31339         (chkp_reduce_bounds_lifetime): Likewise.
31340         * tree-chkp.c (chkp_register_addr_bounds): Likewise.
31341         (chkp_mark_completed_bounds): Likewise.
31342         (chkp_register_incomplete_bounds): Likewise.
31343         (chkp_mark_invalid_bounds): Likewise.
31344         (chkp_maybe_copy_and_register_bounds): Likewise.
31345         (chkp_build_returned_bound): Likewise.
31346         (chkp_get_bound_for_parm): Likewise.
31347         (chkp_build_bndldx): Likewise.
31348         (chkp_get_bounds_by_definition): Likewise.
31349         (chkp_generate_extern_var_bounds): Likewise.
31350         (chkp_get_bounds_for_decl_addr): Likewise.
31351         * tree-chrec.c (chrec_apply): Likewise.
31352         * tree-data-ref.c (dump_data_reference): Likewise.
31353         (dump_subscript): Likewise.
31354         (dump_data_dependence_relation): Likewise.
31355         (analyze_overlapping_iterations): Likewise.
31356         * tree-inline.c (expand_call_inline): Likewise.
31357         (tree_function_versioning): Likewise.
31358         * tree-into-ssa.c (dump_defs_stack): Likewise.
31359         (dump_currdefs): Likewise.
31360         (dump_names_replaced_by): Likewise.
31361         (dump_update_ssa): Likewise.
31362         (update_ssa): Likewise.
31363         * tree-object-size.c (pass_object_sizes::execute): Likewise.
31364         * tree-parloops.c (build_new_reduction): Likewise.
31365         (try_create_reduction_list): Likewise.
31366         (ref_conflicts_with_region): Likewise.
31367         (oacc_entry_exit_ok_1): Likewise.
31368         (oacc_entry_exit_single_gang): Likewise.
31369         * tree-pretty-print.h: Likewise.
31370         * tree-scalar-evolution.c (set_scalar_evolution): Likewise.
31371         (get_scalar_evolution): Likewise.
31372         (add_to_evolution): Likewise.
31373         (get_loop_exit_condition): Likewise.
31374         (analyze_evolution_in_loop): Likewise.
31375         (analyze_initial_condition): Likewise.
31376         (analyze_scalar_evolution): Likewise.
31377         (instantiate_scev): Likewise.
31378         (number_of_latch_executions): Likewise.
31379         (gather_chrec_stats): Likewise.
31380         (final_value_replacement_loop): Likewise.
31381         (scev_const_prop): Likewise.
31382         * tree-sra.c (dump_access): Likewise.
31383         (disqualify_candidate): Likewise.
31384         (create_access): Likewise.
31385         (reject): Likewise.
31386         (maybe_add_sra_candidate): Likewise.
31387         (create_access_replacement): Likewise.
31388         (analyze_access_subtree): Likewise.
31389         (analyze_all_variable_accesses): Likewise.
31390         (sra_modify_assign): Likewise.
31391         (initialize_constant_pool_replacements): Likewise.
31392         (find_param_candidates): Likewise.
31393         (decide_one_param_reduction): Likewise.
31394         (replace_removed_params_ssa_names): Likewise.
31395         * tree-ssa-ccp.c (ccp_fold_stmt): Likewise.
31396         * tree-ssa-copy.c (dump_copy_of): Likewise.
31397         (copy_prop_visit_cond_stmt): Likewise.
31398         * tree-ssa-dce.c (mark_operand_necessary): Likewise.
31399         * tree-ssa-dom.c (pass_dominator::execute): Likewise.
31400         (record_equivalences_from_stmt): Likewise.
31401         * tree-ssa-dse.c (compute_trims): Likewise.
31402         (delete_dead_call): Likewise.
31403         (delete_dead_assignment): Likewise.
31404         * tree-ssa-forwprop.c (forward_propagate_into_gimple_cond): Likewise.
31405         (forward_propagate_into_cond): Likewise.
31406         (pass_forwprop::execute): Likewise.
31407         * tree-ssa-ifcombine.c (ifcombine_ifandif): Likewise.
31408         * tree-ssa-loop-im.c (invariantness_dom_walker::before_dom_children):
31409         Likewise.
31410         (move_computations_worker): Likewise.
31411         (execute_sm): Likewise.
31412         * tree-ssa-loop-ivcanon.c (tree_estimate_loop_size): Likewise.
31413         (remove_exits_and_undefined_stmts): Likewise.
31414         (remove_redundant_iv_tests): Likewise.
31415         * tree-ssa-loop-ivopts.c (dump_use): Likewise.
31416         (adjust_iv_update_pos): Likewise.
31417         * tree-ssa-math-opts.c (bswap_replace): Likewise.
31418         * tree-ssa-phiopt.c (factor_out_conditional_conversion): Likewise.
31419         (value_replacement): Likewise.
31420         * tree-ssa-phiprop.c (phiprop_insert_phi): Likewise.
31421         * tree-ssa-pre.c (print_pre_expr): Likewise.
31422         (get_representative_for): Likewise.
31423         (create_expression_by_pieces): Likewise.
31424         (insert_into_preds_of_block): Likewise.
31425         (eliminate_insert): Likewise.
31426         (eliminate_dom_walker::before_dom_children): Likewise.
31427         (eliminate): Likewise.
31428         (remove_dead_inserted_code): Likewise.
31429         * tree-ssa-propagate.c (substitute_and_fold): Likewise.
31430         * tree-ssa-reassoc.c (get_rank): Likewise.
31431         (eliminate_duplicate_pair): Likewise.
31432         (eliminate_plus_minus_pair): Likewise.
31433         (eliminate_not_pairs): Likewise.
31434         (undistribute_ops_list): Likewise.
31435         (eliminate_redundant_comparison): Likewise.
31436         (update_range_test): Likewise.
31437         (optimize_range_tests_var_bound): Likewise.
31438         (optimize_vec_cond_expr): Likewise.
31439         (rewrite_expr_tree): Likewise.
31440         (rewrite_expr_tree_parallel): Likewise.
31441         (linearize_expr): Likewise.
31442         (break_up_subtract): Likewise.
31443         (linearize_expr_tree): Likewise.
31444         (attempt_builtin_powi): Likewise.
31445         (attempt_builtin_copysign): Likewise.
31446         (transform_stmt_to_copy): Likewise.
31447         (transform_stmt_to_multiply): Likewise.
31448         (dump_ops_vector): Likewise.
31449         * tree-ssa-sccvn.c (vn_nary_build_or_lookup_1): Likewise.
31450         (print_scc): Likewise.
31451         (set_ssa_val_to): Likewise.
31452         (visit_reference_op_store): Likewise.
31453         (visit_use): Likewise.
31454         (sccvn_dom_walker::before_dom_children): Likewise.
31455         (run_scc_vn): Likewise.
31456         * tree-ssa-scopedtables.c (avail_exprs_stack::lookup_avail_expr):
31457         Likewise.
31458         (expr_hash_elt::print): Likewise.
31459         (const_and_copies::pop_to_marker): Likewise.
31460         (const_and_copies::record_const_or_copy_raw): Likewise.
31461         * tree-ssa-structalias.c (compute_dependence_clique): Likewise.
31462         * tree-ssa-uninit.c (collect_phi_def_edges): Likewise.
31463         (dump_predicates): Likewise.
31464         (find_uninit_use): Likewise.
31465         (warn_uninitialized_phi): Likewise.
31466         (pass_late_warn_uninitialized::execute): Likewise.
31467         * tree-ssa.c (verify_vssa): Likewise.
31468         (verify_ssa): Likewise.
31469         (maybe_optimize_var): Likewise.
31470         * tree-vrp.c (dump_value_range): Likewise.
31471         (dump_all_value_ranges): Likewise.
31472         (dump_asserts_for): Likewise.
31473         (register_edge_assert_for_2): Likewise.
31474         (vrp_visit_cond_stmt): Likewise.
31475         (vrp_visit_switch_stmt): Likewise.
31476         (vrp_visit_stmt): Likewise.
31477         (vrp_visit_phi_node): Likewise.
31478         (simplify_cond_using_ranges_1): Likewise.
31479         (fold_predicate_in): Likewise.
31480         (evrp_dom_walker::before_dom_children): Likewise.
31481         (evrp_dom_walker::push_value_range): Likewise.
31482         (evrp_dom_walker::pop_value_range): Likewise.
31483         (execute_early_vrp): Likewise.
31485 2017-05-16  Richard Biener  <rguenther@suse.de>
31487         * dwarf2out.c (loc_list_from_tree_1): Do not create
31488         DW_OP_GNU_variable_value for DECL_IGNORED_P decls.
31490 2017-05-16  Richard Biener  <rguenther@suse.de>
31492         * dwarf2out.c (resolve_variable_value_in_expr): Lookup DIE
31493         just generated.
31494         (note_variable_value_in_expr): If we resolved the decl ref
31495         do not push to the stack.
31497 2017-05-16  Matthew Wahab  <matthew.wahab@arm.com>
31499         * config/arm/arm_neon.h (vadd_f16): Use standard arithmetic
31500         operations in fast-math mode.
31501         (vaddq_f16): Likewise.
31502         (vmul_f16): Likewise.
31503         (vmulq_f16): Likewise.
31504         (vsub_f16): Likewise.
31505         (vsubq_f16): Likewise.
31506         * config/arm/neon.md (add<mode>3): New.
31507         (sub<mode>3): New.
31508         (fma:<VH:mode>3): New.  Also remove outdated comment.
31509         (mul<mode>3): New.
31511 2017-05-16  Martin Liska  <mliska@suse.cz>
31513         PR ipa/79849.
31514         PR ipa/79850.
31515         * ipa-devirt.c (warn_types_mismatch): Fix typo.
31516         (odr_types_equivalent_p): Likewise.
31518 2017-05-15  Sylvestre Ledru  <sylvestre@debian.org>
31520         * plugin.c (try_init_one_plugin): Fix ressource leaks (CID 726637).
31522 2017-05-15  Uros Bizjak  <ubizjak@gmail.com>
31524         PR target/80425
31525         * config/i386.i386.md (*zero_extendsidi2): Do not penalize
31526         non-interunit SSE move alternatives with '?'.
31527         (zero-extendsidi peephole2): New peephole to skip intermediate
31528         general register in SSE zero-extend sequence.
31530 2017-05-15  Jeff Law  <law@redhat.com>
31532         * reorg.c (relax_delay_slots): Create a new variable to hold
31533         the temporary target rather than clobbering TARGET_LABEL.
31535         * config/tilegx/tilegx.c (tilegx_expand_unaligned_load): Add
31536         missing argument to extract_bit_field call.
31537         * config/tilepro/tilepro.c (tilepro_expand_unaligned_load): Likewise.
31539 2017-05-15  Martin Liska  <mliska@suse.cz>
31541         PR driver/31468
31542         * gcc.c (process_command): Do not allow empty argument of -o option.
31544 2017-05-15  Renlin Li  <renlin.li@arm.com>
31546         * config/aarch64/aarch64-protos.h (aarch64_expand_call): Declare.
31547         * config/aarch64/aarch64.c (aarch64_expand_call): Define.
31548         * config/aarch64/constraints.md (Usf): Add long call check.
31549         * config/aarch64/aarch64.md (call): Use aarch64_expand_call.
31550         (call_value): Likewise.
31551         (sibcall): Likewise.
31552         (sibcall_value): Likewise.
31553         (call_insn): New.
31554         (call_value_insn): New.
31555         (sibcall_insn): Update rtx pattern.
31556         (sibcall_value_insn): Likewise.
31557         (call_internal): Remove.
31558         (call_value_internal): Likewise.
31559         (sibcall_internal): Likewise.
31560         (sibcall_value_internal): Likewise.
31561         (call_reg): Likewise.
31562         (call_symbol): Likewise.
31563         (call_value_reg): Likewise.
31564         (call_value_symbol): Likewise.
31566 2017-05-14  Krister Walfridsson  <krister.walfridsson@gmail.com>
31568         PR target/80600
31569         * config/netbsd.h (NETBSD_LIBGCC_SPEC): Always add -lgcc.
31571 2017-05-14  Uros Bizjak  <ubizjak@gmail.com>
31573         * config/i386.i386.c (ix86_cc_modes_compatible): CCNOmode is
31574         compatible with CCGOCmode and with CCZmode.
31576 2017-05-14  Martin Sebor  <msebor@redhat.com>
31578         PR middle-end/77671
31579         * gimple-fold.c (gimple_fold_builtin_sprintf): Make extern.
31580         (gimple_fold_builtin_snprintf): Same.
31581         * gimple-fold.h (gimple_fold_builtin_sprintf): Declare.
31582         (gimple_fold_builtin_snprintf): Same.
31583         * gimple-ssa-sprintf.c (get_format_string): Correct the detection
31584         of character types.
31585         (is_call_safe): New function.
31586         (try_substitute_return_value): Call it.
31587         (try_simplify_call): New function.
31588         (pass_sprintf_length::handle_gimple_call): Call it.
31590 2017-05-14  Martin Sebor  <msebor@redhat.com>
31592         PR middle-end/80669
31593         * builtins.c (expand_builtin_stpncpy): Simplify.
31595 2017-05-14  Daniel Santos  <daniel.santos@pobox.com>
31597         * config/i386/i386.opt: Add option -mcall-ms2sysv-xlogues.
31598         * config/i386/i386.h
31599         (x86_64_ms_sysv_extra_clobbered_registers): Change type to unsigned.
31600         (NUM_X86_64_MS_CLOBBERED_REGS): New macro.
31601         (struct machine_function): Add new members call_ms2sysv,
31602         call_ms2sysv_pad_in, call_ms2sysv_pad_out and call_ms2sysv_extra_regs.
31603         (struct machine_frame_state): New fields sp_realigned and
31604         sp_realigned_offset.
31605         * config/i386/i386.c
31606         (enum xlogue_stub): New enum.
31607         (enum xlogue_stub_sets): New enum.
31608         (class xlogue_layout): New class.
31609         (struct ix86_frame): New fields stack_realign_allocate_offset,
31610         stack_realign_offset and outlined_save_offset.  Modify comments to
31611         detail stack layout when using out-of-line stubs.
31612         (ix86_target_string): Add -mcall-ms2sysv-xlogues option.
31613         (ix86_option_override_internal): Add sorry() for TARGET_SEH and
31614         -mcall-ms2sysv-xlogues.
31615         (stub_managed_regs): New static variable.
31616         (ix86_save_reg): Add new parameter ignore_outlined to optionally omit
31617         registers managed by out-of-line stub.
31618         (disable_call_ms2sysv_xlogues): New function.
31619         (ix86_compute_frame_layout): Modify re-alignment calculations, disable
31620         m->call_ms2sysv when appropriate and compute frame layout for
31621         out-of-line stubs.
31622         (sp_valid_at, fp_valid_at): New inline functions.
31623         (choose_basereg): New function.
31624         (choose_baseaddr): Add align parameter, use choose_basereg and modify
31625         all callers.
31626         (ix86_emit_save_reg_using_mov, ix86_emit_restore_sse_regs_using_mov):
31627         Use align parameter of choose_baseaddr to generated aligned SSE movs
31628         when possible.
31629         (pro_epilogue_adjust_stack): Modify to track
31630         machine_frame_state::sp_realigned.
31631         (ix86_nsaved_regs): Modify to accommodate changes to ix86_save_reg.
31632         (ix86_nsaved_sseregs): Likewise.
31633         (ix86_emit_save_regs): Likewise.
31634         (ix86_emit_save_regs_using_mov): Likewise.
31635         (ix86_emit_save_sse_regs_using_mov): Likewise.
31636         (get_scratch_register_on_entry): Likewise.
31637         (gen_frame_set): New function.
31638         (gen_frame_load): Likewise.
31639         (gen_frame_store): Likewise.
31640         (emit_outlined_ms2sysv_save): Likewise.
31641         (emit_outlined_ms2sysv_restore): Likewise.
31642         (ix86_expand_prologue): Modify stack re-alignment code and call
31643         emit_outlined_ms2sysv_save when appropriate.
31644         (ix86_emit_leave): Clear machine_frame_state::sp_realigned.  Add
31645         parameter rtx_insn *insn, which allows the function to be used to only
31646         generate the notes.
31647         (ix86_expand_epilogue): Modify validity checks of frame and stack
31648         pointers, and call emit_outlined_ms2sysv_restore when appropriate.
31649         (ix86_expand_call): Modify to enable m->call_ms2sysv when appropriate.
31650         * config/i386/predicates.md
31651         (save_multiple): New predicate.
31652         (restore_multiple): Likewise.
31653         * config/i386/sse.md
31654         (save_multiple<mode>): New pattern.
31655         (save_multiple_realign<mode>): Likewise.
31656         (restore_multiple<mode>): Likewise.
31657         (restore_multiple_and_return<mode>): Likewise.
31658         (restore_multiple_leave_return<mode>): Likewise.
31659         * Makefile.in: Export HOSTCXX and HOSTCXXFLAGS to site.exp
31661 2017-05-14  Julia Koval  <julia.koval@intel.com>
31663         * config/i386/i386-builtin-types.def (VOID_FTYPE_INT_INT64): New type.
31664         * config/i386/i386-builtin.def (__builtin_ia32_xgetbv)
31665         (__builtin_ia32_xsetbv): New builtins.
31666         * config/i386/i386.c (ix86_expand_special_args_builtin):
31667         Process new types.
31668         (ix86_expand_builtin): Special expand for new intrinsics.
31669         * config/i386/i386.md (UNSPECV_XGETBV, UNSPECV_XSETBV): New.
31670         (xsetbv, xsetbv_rex64, xgetbv, xgetbv_rex64): New insn patterns.
31671         * config/i386/xsaveintrin.h (_xsetbv, _getbv): New intrinsics.
31673 2017-05-13  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
31675         * cfganal.c (inverted_post_order_compute): Change argument type
31676         to vec *.
31677         * cfganal.h (inverted_post_order_compute): Adjust prototype.
31678         * df-core.c (rest_of_handle_df_initialize): Adjust.
31679         (rest_of_handle_df_finish): Likewise.
31680         (df_analyze_1): Likewise.
31681         (df_analyze): Likewise.
31682         (loop_inverted_post_order_compute): Change argument to be a vec *.
31683         (df_analyze_loop): Adjust.
31684         (df_get_n_blocks): Likewise.
31685         (df_get_postorder): Likewise.
31686         * df.h (struct df_d): Change field to be a vec.
31687         * lcm.c (compute_laterin): Adjust.
31688         (compute_available): Likewise.
31689         * lra-lives.c (lra_create_live_ranges_1): Likewise.
31690         * tree-ssa-dce.c (remove_dead_stmt): Likewise.
31691         * tree-ssa-pre.c (compute_antic): Likewise.
31693 2017-05-13  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
31695         * cfganal.c (connect_infinite_loops_to_exit): Adjust.
31696         (depth_first_search::depth_first_search): Change structure init
31697         function to this constructor.
31698         (depth_first_search::add_bb): Rename function to this member.
31699         (depth_first_search::execute): Likewise.
31700         (flow_dfs_compute_reverse_finish): Adjust.
31702 2017-05-13  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
31704         * ddg.c (find_nodes_on_paths): Use auto_sbitmap.
31705         (longest_simple_path): Likewise.
31706         * shrink-wrap.c (spread_components): Likewise.
31707         (disqualify_problematic_components): Likewise.
31708         (emit_common_heads_for_components): Likewise.
31709         (emit_common_tails_for_components): Likewise.
31710         (insert_prologue_epilogue_for_components): Likewise.
31712 2017-05-13  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
31714         * tree-ssa-dse.c (dse_dom_walker): Make m_live_byes a
31715         auto_sbitmap.
31717 2017-05-13  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
31719         * df-core.c (df_set_blocks): Start using auto_bitmap.
31720         (df_compact_blocks): Likewise.
31721         * df-problems.c (df_rd_confluence_n): Likewise.
31722         * df-scan.c (df_insn_rescan_all): Likewise.
31723         (df_process_deferred_rescans): Likewise.
31724         (df_update_entry_block_defs): Likewise.
31725         (df_update_exit_block_uses): Likewise.
31726         (df_entry_block_bitmap_verify): Likewise.
31727         (df_exit_block_bitmap_verify): Likewise.
31728         (df_scan_verify): Likewise.
31729         * lra-constraints.c (lra_constraints): Likewise.
31730         (undo_optional_reloads): Likewise.
31731         (lra_undo_inheritance): Likewise.
31732         * lra-remat.c (calculate_gen_cands): Likewise.
31733         (do_remat): Likewise.
31734         * lra-spills.c (assign_spill_hard_regs): Likewise.
31735         (spill_pseudos): Likewise.
31736         * tree-ssa-pre.c (bitmap_set_and): Likewise.
31737         (bitmap_set_subtract_values): Likewise.
31739 2017-05-13  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
31741         * haifa-sched.c (estimate_shadow_tick): Replace manual bitmap
31742         management with auto_bitmap.
31743         (fix_inter_tick): Likewise.
31744         (fix_recovery_deps): Likewise.
31745         * ira.c (add_store_equivs): Likewise.
31746         (find_moveable_pseudos): Likewise.
31747         (split_live_ranges_for_shrink_wrap): Likewise.
31748         * print-rtl.c (rtx_reuse_manager::rtx_reuse_manager): Likewise.
31749         (rtx_reuse_manager::seen_def_p): Likewise.
31750         (rtx_reuse_manager::set_seen_def): Likewise.
31751         * print-rtl.h (class rtx_reuse_manager): Likewise.
31753 2017-05-13  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
31755         * bt-load.c (combine_btr_defs): Use auto_bitmap to manage bitmap
31756         lifetime.
31757         (migrate_btr_def): Likewise.
31758         * cfgloop.c (get_loop_body_in_bfs_order): Likewise.
31759         * df-core.c (loop_post_order_compute): Likewise.
31760         (loop_inverted_post_order_compute): Likewise.
31761         * hsa-common.h: Likewise.
31762         * hsa-gen.c (hsa_bb::~hsa_bb): Likewise.
31763         * init-regs.c (initialize_uninitialized_regs): Likewise.
31764         * ipa-inline.c (resolve_noninline_speculation): Likewise.
31765         (inline_small_functions): Likewise.
31766         * ipa-reference.c (ipa_reference_write_optimization_summary): Likewise.
31767         * ira.c (combine_and_move_insns): Likewise.
31768         (build_insn_chain): Likewise.
31769         * loop-invariant.c (find_invariants): Likewise.
31770         * lower-subreg.c (propagate_pseudo_copies): Likewise.
31771         * predict.c (tree_predict_by_opcode): Likewise.
31772         (predict_paths_leading_to): Likewise.
31773         (predict_paths_leading_to_edge): Likewise.
31774         (estimate_loops_at_level): Likewise.
31775         (estimate_loops): Likewise.
31776         * shrink-wrap.c (try_shrink_wrapping): Likewise.
31777         (spread_components): Likewise.
31778         * tree-cfg.c (remove_edge_and_dominated_blocks): Likewise.
31779         * tree-loop-distribution.c (rdg_build_partitions): Likewise.
31780         * tree-predcom.c (tree_predictive_commoning_loop): Likewise.
31781         * tree-ssa-coalesce.c (coalesce_ssa_name): Likewise.
31782         * tree-ssa-phionlycprop.c (pass_phi_only_cprop::execute): Likewise.
31783         * tree-ssa-pre.c (remove_dead_inserted_code): Likewise.
31784         * tree-ssa-sink.c (nearest_common_dominator_of_uses): Likewise.
31785         * tree-ssa-threadupdate.c (compute_path_counts): Likewise.
31786         (mark_threaded_blocks): Likewise.
31787         (thread_through_all_blocks): Likewise.
31788         * tree-ssa.c (verify_ssa): Likewise.
31789         (execute_update_addresses_taken): Likewise.
31790         * tree-ssanames.c (verify_ssaname_freelists): Likewise.
31792 2017-05-13  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
31794         * cfganal.c (mark_dfs_back_edges): Replace manual stack with
31795         auto_vec.
31796         (post_order_compute): Likewise.
31797         (inverted_post_order_compute): Likewise.
31798         (pre_and_rev_post_order_compute_fn): Likewise.
31800 2017-05-13  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
31802         * genrecog.c (int_set::int_set): Explicitly construct our
31803         auto_vec base class.
31804         * vec.h (auto_vec::auto_vec): New constructor.
31806 2017-05-13  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
31808         * bitmap.h (class auto_bitmap): New constructor taking
31809         bitmap_obstack * argument.
31811 2017-05-13  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
31813         * bitmap.h (class auto_bitmap): Change type of m_bits to
31814         bitmap_head, and adjust ctor / dtor and member operators.
31816 2017-05-13  Uros Bizjak  <ubizjak@gmail.com>
31818         * compare-elim.c (equivalent_reg_at_start): Return NULL_RTX
31819         when returned register mode doesn't match original mode.
31821 2017-05-12  Jeff Law  <law@redhat.com>
31822             Jakub Jelinek  <jakub@redhat.com>
31824         * config/mn10300/mn10300.c (mn10300_match_ccmode): Fix where
31825         we look for cc setter after the compare-elim changes.
31826         * config/mn10300/mn10300.md (addsi3_flags): Fix order of patterns
31827         within the vector to match what compare-elim now expects.
31828         (subsi3_flags, andsi3_flags, iorsi3_flags): Likewise.
31829         (xorsi3_flags, one_cmplsi2_flags): Likewise.
31831         * config/rx/rx.c (rx_match_ccmode): Fix where we look cc setter
31832         after the compare-elim changes.
31833         * config/rx/rx.md (abssi2_flags): Fix order of patterns within
31834         the vector to match what compare-elim now expects.
31835         (addsi3_flags, adc_flags, addsi3_flags peepholes): Likewise.
31836         (andsi3_flags, negsi2_flags, one_cmplsi2_flags): Likewise.
31837         (iorsi3_flags, rotlsi3_flags, rotrsi3_flags): Likewise.
31838         (ashrsi3_flags, lshrsi3_flags, ashlsi3_flags): Likewise.
31839         (ssaddsi3, subsi3_flags, sbb_flags, xorsi3_flags): Likewise.
31841         * config/visium/visium.c (single_set_and_flags): Fix where
31842         we look for cc setter after the compare-elim changes.
31843         * config/visium/visium.md (flags_subst_logic): Fix order of patterns
31844         with the vector to match what compare-elim now expects.
31845         (flags_subst_arith, add<mode>3_insn_set_carry): Likewise.
31846         (add<mode>3_insn_set_overflow, addsi3_insn_set_carry): Likewise.
31847         (addsi3_insn_set_overflow, sub<mode>3_insn_set_carry): Likewise.
31848         (sub<mode>3_insn_set_overflow, subsi3_insn_set_carry): Likewise.
31849         (subsi3_insn_set_overflow, negsi2_insn_set_carry): Likewise.
31850         (neg<mode>2_insn_set_overflow): Likewise.
31852 2017-05-12  Jim Wilson  <jim.wilson@linaro.org>
31854         PR middle-end/79794
31855         * expmed.c (extract_bit_field_1): Add alt_rtl argument.  Before
31856         maybe_expand_insn call, set ops[0].target.  If still set after call,
31857         set alt_rtl.  Add extra arg to recursive calls.
31858         (extract_bit_field): Add alt_rtl argument.  Pass to
31859         extract_bit_field.
31860         * expmed.h (extract_bit_field): Fix prototype.
31861         * expr.c (emit_group_load_1, copy_blkmode_from_reg)
31862         (copy_blkmode_to_reg, read_complex_part, store_field): Pass extra NULL
31863         to extract_bit_field_calls.
31864         (expand_expr_real_1): Pass alt_rtl to expand_expr_real instead of 0.
31865         Pass alt_rtl to extract_bit_field calls.
31866         * calls.c (store_unaligned_arguments_into_psuedos)
31867         load_register_parameters): Pass extra NULL to extract_bit_field calls.
31868         * optabs.c (maybe_legitimize_operand): Clear op->target when call
31869         gen_reg_rtx.
31870         * optabs.h (struct expand_operand): Add target bitfield.
31872 2017-05-12  Uros Bizjak  <ubizjak@gmail.com>
31874         * compare-elim.c (try_eliminate_compare): Canonicalize
31875         operation with embedded compare to
31876         [(set (reg:CCM) (compare:CCM (operation) (immediate)))
31877          (set (reg) (operation)].
31879         * config/i386/i386.c (TARGET_FLAGS_REGNUM): New define.
31881 2017-05-12  Uros Bizjak  <ubizjak@gmail.com>
31883         PR target/80723
31884         * config/i386/i386.c (ix86_rtx_cost) [case PLUS]: Ignore the
31885         cost of adding a carry flag for ADC instruction.
31886         [case MINUS]: Ignore the cost of subtracting a carry flag
31887         for SBB instruction.
31889 2017-05-12  Steven Munroe  <munroesj@gcc.gnu.org>
31891         * config.gcc (powerpc*-*-*): Add bmi2intrin.h, bmiintrin.h,
31892         and x86intrin.h
31893         * config/rs6000/bmiintrin.h: New file.
31894         * config/rs6000/bmi2intrin.h: New file.
31895         * config/rs6000/x86intrin.h: New file.
31897 2017-05-12  Jeff Law  <law@redhat.com>
31899         * tree-vrp.c (vrp_dom_walker::before_dom_children): Push unwinding
31900         markers.
31902 2017-05-12  Peter Bergner  <bergner@vnet.ibm.com>
31904         PR middle-end/80707
31905         * tree-cfg.c: Remove cfg edges of unreachable case statements.
31907 2017-05-12  Will Schmidt  <will_schmidt@vnet.ibm.com>
31909         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add handling for
31910         early expansion of vector divide builtins.
31911         (builtin_function_type): Add VSX_BUILTIN_UDIV_V2DI to the list of
31912         builtins identified as having unsigned arguments.
31914 2017-05-12  Will Schmidt  <will_schmidt@vnet.ibm.com>
31916         * config/rs6000/rs6000.c (gimple-fold.h): New #include.
31917         (rs6000_gimple_fold_builtin): Add handling for early GIMPLE
31918         expansion of vector logical operations (and, andc, or, xor,
31919         nor, orc, nand).
31921 2017-05-12  Will Schmidt  <will_schmidt@vnet.ibm.com>
31923         * gimple-fold.c (create_tmp_reg_or_ssa_name): Remove static declaration.
31924         * gimple-fold.h (create_tmp_reg_or_ssa_name): New prototype.
31926 2017-05-12  Will Schmidt  <will_schmidt@vnet.ibm.com>
31928         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add handling for
31929         early GIMPLE expansion of vector multiplies.
31931 2017-05-12  Prakhar Bahuguna  <prakhar.bahuguna@arm.com>
31933         * config/arm/arm.md (movsi): Add TARGET_32BIT in addition to the
31934         TARGET_HAVE_MOVT conditional.
31935         (movt splitter): Likewise.
31937 2017-05-12  Richard Biener  <rguenther@suse.de>
31939         * tree-ssa-sccvn.h (has_VN_INFO): Declare.
31940         * tree-ssa-pre.c (eliminate_dom_walker::before_dom_children):
31941         Fold all stmts not inplace.
31943 2017-05-12  Richard Biener  <rguenther@suse.de>
31945         PR tree-optimization/80713
31946         * tree-ssa-pre.c (remove_dead_inserted_code): Clear
31947         inserted_exprs bit for not removed stmts.
31949 2017-05-12  Thomas Schwinge  <thomas@codesourcery.com>
31951         PR middle-end/69921
31952         * tree-parloops.c (create_parallel_loop): Set "oacc kernels
31953         parallelized" attribute for parallelized OpenACC kernels.
31954         * omp-offload.c (execute_oacc_device_lower): Use it.
31956         * omp-expand.c (expand_omp_target) <GF_OMP_TARGET_KIND_OACC_KERNELS>:
31957         Set "oacc kernels" attribute.
31958         * omp-general.c (oacc_set_fn_attrib): Remove is_kernel formal
31959         parameter.  Adjust all users.
31960         (oacc_fn_attrib_kernels_p): Remove function.
31961         * omp-offload.c (execute_oacc_device_lower): Look for "oacc
31962         kernels" attribute instead of calling oacc_fn_attrib_kernels_p.
31963         * tree-ssa-loop.c (gate_oacc_kernels): Likewise.
31964         * tree-parloops.c (create_parallel_loop): If oacc_kernels_p,
31965         assert "oacc kernels" attribute is set.
31967 2017-05-11  Carl Love  <cel@us.ibm.com>
31969         * config/rs6000/rs6000-c: Add support for built-in functions
31970         vector unsigned char vec_popcnt (vector signed char)
31971         vector unsigned char vec_popcnt (vector unsigned char)
31972         vector unsigned short vec_popcnt (vector signed short)
31973         vector unsigned short vec_popcnt (vector unsigned short)
31974         vector unsigned int vec_popcnt (vector signed int)
31975         vector unsigned int vec_popcnt (vector unsigned int)
31976         vector unsigned long long vec_popcnt (vector signed long long)
31977         vector unsigned long long vec_popcnt (vector unsigned long long)
31978         vector signed long long vec_slo (vector signed long long,
31979                                          vector signed char)
31980         vector signed long long vec_slo (vector signed long long,
31981                                          vector unsigned char)
31982         vector unsigned long long vec_slo (vector unsigned long long,
31983                                            vector signed char)
31984         vector unsigned long long vec_slo (vector unsigned long long,
31985                                            vector unsigned char)
31986         * config/rs6000/rs6000-builtin.def: Add definitions for VPOPCNTUB,
31987         VPOPCNTUH, VPOPCNTUW, and VPOPCNTUD overloads.
31988         * config/rs6000/altivec.h: Add define for vec_popcnt, vec_popcntb,
31989         vec_popcnth, vec_popcntw and vec_popcntd built-in functions.
31990         * doc/extend.texi: Update the built-in documentation file for the
31991         new built-in functions.
31993 2017-05-11  Michael Meissner  <meissner@linux.vnet.ibm.com>
31995         * attribs.h (sorted_attr_string): Move machine independent
31996         functions for target clone support from the i386 port to common
31997         code.  Rename ix86_function_versions to common_function_versions.
31998         Rename make_name to make_unique_name.
31999         (common_function_versions): Likewise.
32000         (make_unique_name): Likewise.
32001         (make_dispatcher_decl): Likewise.
32002         (is_function_default_version): Likewise.
32003         * attribs.c (attr_strcmp): Likewise.
32004         (sorted_attr_string): Likewise.
32005         (common_function_versions): Likewise.
32006         (make_unique_name): Likewise.
32007         (make_dispatcher_decl): Likewise.
32008         (is_function_default_version): Likewise.
32009         * config/i386/i386.c (attr_strcmp): Likewise.
32010         (sorted_attr_string): Likewise.
32011         (ix86_function_versions): Likewise.
32012         (make_name): Likewise.
32013         (make_dispatcher_decl): Likewise.
32014         (is_function_default_version): Likewise.
32015         (TARGET_OPTION_FUNCTION_VERSIONS): Update target function hook.
32017 2017-05-11  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
32019         PR target/80695
32020         * config/rs6000/rs6000.c (rs6000_builtin_vectorization_cost):
32021         Account for direct move costs for vec_construct of integer
32022         vectors.
32024 2017-05-11  Uros Bizjak  <ubizjak@gmail.com>
32026         PR target/80706
32027         * config/i386/sync.md (UNSPEC_LDX_ATOMIC): New unspec.
32028         (UNSPEC_STX_ATOMIC): Ditto.
32029         (loaddi_via_sse): New insn.
32030         (storedi_via_sse): Ditto.
32031         (atomic_loaddi_fpu): Emit loaddi_via_sse and storedi_via_sse.
32032         Update corresponding peephole2 patterns.
32033         (atomic_storedi_fpu): Ditto.
32035 2017-05-11  Julia Koval  <julia.koval@intel.com>
32037         * config/i386/avx512fintrin.h (_mm_mask_rsqrt14_sd)
32038         (_mm_maskz_rsqrt14_sd, _mm_mask_rsqrt14_ss, _mm_maskz_rsqrt14_ss):
32039         New intrinsics.
32040         * config/i386/i386-builtin.def (__builtin_ia32_rsqrt14sd_mask)
32041         (__builtin_ia32_rsqrt14ss_mask): New builtins.
32042         * config/i386/sse.md (rsqrt14_<mode>_mask): New pattern.
32044 2017-05-11  Nathan Sidwell  <nathan@acm.org>
32046         * graphite-poly.c: Include dumpfile.h.
32048         * dumpfle.h (dump_function): Declare here ...
32049         * tree-dump.h (dump_function): ... not here.
32050         * dumpfile.c: #include tree-cfg.h.
32051         (dump_function): Move here from ...
32052         * tree-dump.c (dump_function): ... here.
32053         * gimplify.c: #include splay-tree.h, not tree-dump.h.
32054         * graphite-poly.c: Don't include tree-dump.h.
32055         * cgraphclones.c: Include dumpfile.h not tree-dump.h.
32056         * print-tree.c: Likewise.
32057         * stor-layout.c: Likewise.
32058         * tree-nested.c: Likewise.
32060         * dumpfile.c (dump_start): Use TDF_FLAGS.
32061         (dump_enable_all): Fix TDF_KIND check thinko.
32063 2017-05-11  Kelvin Nilsen  <kelvin@gcc.gnu.org>
32065         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
32066         array entries to represent two legal parameterizations of the
32067         overloaded __builtin_cmpb function, as represented by the
32068         P6_OV_BUILTIN_CMPB constant.
32069         (altivec_resolve_overloaded_builtin): Add special case handling
32070         for the __builtin_cmpb function, as represented by the
32071         P6_OV_BUILTIN_CMPB constant.
32072         * config/rs6000/rs6000-builtin.def (BU_P6_2): New macro.
32073         (BU_P6_64BIT_2): New macro.
32074         (BU_P6_OVERLOAD_2): New macro
32075         (CMPB_32): Add 32-bit compare-bytes support for 32-bit only targets.
32076         (CMPB): Add 64-bit compare-bytes support for 32-bit and 64-bit targets.
32077         (CMPB): Add overload support to represent both 32-bit and 64-bit
32078         compare-bytes function.
32079         * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add
32080         support for TARGET_CMPB.
32081         * config/rs6000/rs6000.h: Add support for RS6000_BTM_CMPB.
32082         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Add
32083         documentation of the __builtin_cmpb overloaded built-in function.
32085 2017-05-11  Richard Biener  <rguenther@suse.de>
32087         PR tree-optimization/80705
32088         * tree-vect-data-refs.c (vect_analyze_data_refs): DECL_NONALIASED
32089         bases are not vectorizable.
32091 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32093         * tree-ssa-loop-ivopts.c (determine_set_costs): Skip non-interger
32094         when counting register pressure.
32096 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32098         * tree-ssa-loop-ivopts.c (compare_cost_pair): New.
32099         (iv_ca_more_deps): Renamed to ...
32100         (iv_ca_compare_deps): ... this.
32101         (iv_ca_extend): Extend iv_ca if NEW_CP is cheaper than OLD_CP.
32103 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32105         * tree-ssa-loop-ivopts.c (find_interesting_uses): Move inv vars dump
32106         to ...
32107         (determine_group_iv_costs): ... here.
32108         (find_inv_vars_cb): Record inv var if it's not recorded before.
32110 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32112         * tree-ssa-loop-ivopts.c (allow_ip_end_pos_p): Refine comments.
32113         (get_shiftadd_cost): Ditto.
32115 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32117         * tree-ssa-address.c: Include header file.
32118         (move_hint_to_base): Return TRUE if BASE_HINT is moved to memory
32119         address.
32120         (add_to_parts): Refactor.
32121         (addr_to_parts): New parameter.  Update use of move_hint_to_base.
32122         (create_mem_ref): Update use of addr_to_parts.  Re-associate addr
32123         in new order.
32125 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32127         PR tree-optimization/53090
32128         * tree-ssa-loop-ivopts.c (enum comp_iv_rewrite): New enum value
32129         COMP_IV_EXPR_2.
32130         (extract_cond_operands): Detect condition with IV on both sides
32131         and return COMP_IV_EXPR_2.
32132         (find_interesting_uses_cond): Add iv_use for both IVs in condition.
32133         (rewrite_use_compare): Simplify by removing call to function
32134         extract_cond_operands.
32136 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32138         * tree-ssa-loop-ivopts.c (enum comp_iv_rewrite): New.
32139         (extract_cond_operands): Detect condition comparing against non-
32140         invariant bound and return appropriate enum value.
32141         (find_interesting_uses_cond): Update use of extract_cond_operands.
32142         Handle its return value accordingly.
32143         (determine_group_iv_cost_cond, rewrite_use_compare): Ditto.
32145 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32147         * tree-ssa-loop-ivopts.c (rewrite_use_nonlinear_expr): Re-associate
32148         nonlinear iv_use computation in loop invariant sensitive way.
32150 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32152         * tree-ssa-loop-ivopts.c (relate_compare_use_with_all_cands): New.
32153         (find_iv_candidates): Call relate_compare_use_with_all_cands.
32155 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32157         * tree-ssa-loop-ivopts.c (struct iv_cand): New field inv_exprs.
32158         (dump_cand): Support iv_cand.inv_exprs.
32159         (add_candidate_1): Record invariant exprs in iv_cand.inv_exprs
32160         for candidates.
32161         (iv_ca_set_no_cp, iv_ca_set_cp, free_loop_data): Support
32162         iv_cand.inv_exprs.
32164 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32166         * tree-ssa-loop-ivopts.c (multiplier_allowed_in_address_p): Move
32167         from ...
32168         * tree-ssa-address.c (multiplier_allowed_in_address_p): ... to here
32169         as local function.  Include necessary header files.
32170         * tree-ssa-loop-ivopts.h (multiplier_allowed_in_address_p): Delete.
32172 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32174         * tree-ssa-loop-ivopts.c (autoinc_possible_for_pair): Simplify.
32176 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32178         * tree-ssa-loop-ivopts.c (force_expr_to_var_cost): Handle more
32179         operators: TRUNC_DIV_EXPR, BIT_AND_EXPR, BIT_IOR_EXPR, LSHIFT_EXPR,
32180         RSHIFT_EXPR and BIT_NOT_EXPR.
32182 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32184         * tree-ssa-loop-ivopts.c (get_loop_invariant_expr): Simplify.
32185         (adjust_setup_cost): New parameter supporting round up adjustment.
32186         (struct address_cost_data): Delete.
32187         (force_expr_to_var_cost): Don't bound cost with spill_cost.
32188         (split_address_cost, ptr_difference_cost): Delete.
32189         (difference_cost, compare_aff_trees, record_inv_expr): Delete.
32190         (struct ainc_cost_data): New struct.
32191         (get_address_cost_ainc): New function.
32192         (get_address_cost, get_computation_cost): Reimplement.
32193         (determine_group_iv_cost_address): Record inv_expr for all uses of
32194         a group.
32195         (determine_group_iv_cost_cond): Call get_loop_invariant_expr.
32196         (iv_ca_has_deps): Reimplemented to ...
32197         (iv_ca_more_deps): ... this.  Check if NEW_CP introduces more deps
32198         than OLD_CP.
32199         (iv_ca_extend): Call iv_ca_more_deps.
32201 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32203         * tree-ssa-address.c (struct mem_address): Move to header file.
32204         (valid_mem_ref_p, move_fixed_address_to_symbol): Make it global.
32205         * tree-ssa-address.h (struct mem_address): Move from C file.
32206         (valid_mem_ref_p, move_fixed_address_to_symbol): Declare.
32208 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32210         * tree-affine.h (aff_combination_type): New interface.
32211         (aff_combination_zero_p): Remove static.
32212         (aff_combination_const_p): New interface.
32213         (aff_combination_singleton_var_p): New interfaces.
32215 2017-05-11  Richard Biener  <rguenther@suse.de>
32217         * tree-ssa-pre.c (eliminate_dom_walker::before_dom_children):
32218         Skip unreachable blocks and destinations.
32219         (eliminate): Move stmt removal and fixup ...
32220         (fini_eliminate): ... here.  Skip inserted exprs.
32221         (pass_pre::execute): Move fini_pre after fini_eliminate.
32222         * tree-ssa-tailmerge.c: Include tree-cfgcleanup.h.
32223         (tail_merge_optimize): Run cleanup_tree_cfg if requested by
32224         PRE to get rid of dead code that has invalid SSA form and
32225         split critical edges again.
32227 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32229         * rtlanal.c (rtx_cost): Handle TRUNCATE between tieable modes.
32231 2017-05-11  Richard Biener  <rguenther@suse.de>
32233         * passes.c (execute_function_todo): Verify loops if they are
32234         said to be up-to-date.
32235         * cfgexpand.c (pass_expand::execute): Discard loops for -dx.
32236         * trans-mem.c (pass_tm_edges::execute): Mark loops for fixup.
32238 2017-05-10  John David Anglin  <danglin@gcc.gnu.org>
32240         PR target/80090
32241         * config/pa/pa.c (pa_assemble_integer): When outputting a SYMBOL_REF,
32242         handle calling assemble_external ourself.
32244         PR target/79027
32245         * config/pa/pa.c (pa_cannot_change_mode_class): Reject changes to/from
32246         modes with zero size.  Enhance comment.
32248 2017-05-10  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
32250         * config/rs6000/rs6000.c (altivec_init_builtins): Define POWER8
32251         built-ins for vec_xl and vec_xst with short and char pointer
32252         arguments.
32254 2017-05-10  Sebastian Peryt  <sebastian.peryt@intel.com>
32256         * config/i386/avx512fintrin.h (_mm_mask_max_round_sd)
32257         (_mm_maskz_max_round_sd, _mm_mask_max_round_ss)
32258         (_mm_maskz_max_round_ss, _mm_mask_min_round_sd)
32259         (_mm_maskz_min_round_sd, _mm_mask_min_round_ss)
32260         (_mm_maskz_min_round_ss): New intrinsics.
32261         * config/i386/i386-builtin-types.def (V2DF, V2DF, V2DF, V2DF, UQI, INT)
32262         (V4SF, V4SF, V4SF, V4SF, UQI, INT): New function type aliases.
32263         * config/i386/i386-builtin.def (__builtin_ia32_maxsd_mask_round)
32264         (__builtin_ia32_maxss_mask_round, __builtin_ia32_minsd_mask_round)
32265         (__builtin_ia32_minss_mask_round): New builtins.
32266         * config/i386/i386.c (V2DF_FTYPE_V2DF_V2DF_V2DF_UQI_INT)
32267         (V4SF_FTYPE_V4SF_V4SF_V4SF_UQI_INT): Handle new types.
32268         * config/i386/sse.md (<sse>_vm<code><mode>3<round_saeonly_name>):
32269         Rename to ...
32270         (<sse>_vm<code><mode>3<mask_name><round_saeonly_name>): ... this.
32271         (v<maxmin_float><ssescalarmodesuffix>\t{<round_saeonly_op3>%2, %1, %0|%0, %1, %<iptr>2<round_saeonly_op3>}):
32272         Change to ...
32273         (v<maxmin_float><ssescalarmodesuffix>\t{<round_saeonly_mask_op3>%2, %1, %0<mask_operand3>|%0<mask_operand3>, %1, %<iptr>2<round_saeonly_mask_op3>}):
32274         ... this.
32276 2017-05-10  Sebastian Peryt  <sebastian.peryt@intel.com>
32278         * config/i386/avx512fintrin.h (_mm_mask_mul_round_sd)
32279         (_mm_maskz_mul_round_sd, _mm_mask_mul_round_ss)
32280         (_mm_maskz_mul_round_ss, _mm_mask_div_round_sd)
32281         (_mm_maskz_div_round_sd, _mm_mask_div_round_ss)
32282         (_mm_maskz_div_round_ss, _mm_mask_mul_sd, _mm_maskz_mul_sd)
32283         (_mm_mask_mul_ss, _mm_maskz_mul_ss, _mm_mask_div_sd)
32284         (_mm_maskz_div_sd, _mm_mask_div_ss, _mm_maskz_div_ss): New intrinsics.
32285         * config/i386/i386-builtin-types.def (V2DF_FTYPE_V2DF_V2DF_V2DF_UQI_INT)
32286         (V4SF_FTYPE_V4SF_V4SF_V4SF_UQI_INT): New function type aliases.
32287         * config/i386/i386-builtin.def (__builtin_ia32_divsd_mask_round)
32288         (__builtin_ia32_divss_mask_round, __builtin_ia32_mulsd_mask_round)
32289         (__builtin_ia32_mulss_mask_round): New builtins.
32290         * config/i386/i386.c (V2DF_FTYPE_V2DF_V2DF_V2DF_UQI_INT)
32291         (V4SF_FTYPE_V4SF_V4SF_V4SF_UQI_INT): Handle new types.
32292         * config/i386/sse.md (<sse>_vm<multdiv_mnemonic><mode>3<round_name>):
32293         Rename to ...
32294         (<sse>_vm<multdiv_mnemonic><mode>3<mask_name><round_name>): ... this.
32295         (v<multdiv_mnemonic><ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|%0, %1, %<iptr>2<round_op3>}):
32296         Change to ...
32297         (v<multdiv_mnemonic><ssescalarmodesuffix>\t{<round_mask_op3>%2, %1, %0<mask_operand3>|%0<mask_operand3>, %1, %<iptr>2<round_mask_op3>}):
32298         ... this.
32300 2017-05-10  Julia Koval  <julia.koval@intel.com>
32302         * config/i386/avxintrin.h (_mm256_set_m128, _mm256_set_m128d)
32303         (_mm256_set_m128i, _mm256_setr_m128, _mm256_setr_m128d)
32304         (_mm256_setr_m128i): New intrinsics.
32306 2017-05-10  Julia Koval  <julia.koval@intel.com>
32308         * config/i386/avx512fintrin.h (_mm_mask_rcp14_sd)
32309         (_mm_maskz_rcp14_sd, _mm_mask_rcp14_ss)
32310         (_mm_maskz_rcp14_ss): New intrinsics.
32311         * config/i386/i386-builtin.def (__builtin_ia32_rcp14sd_mask)
32312         (__builtin_ia32_rcp14ss_mask): New builtins.
32313         * config/i386/sse.md (srcp14<mode>_mask): New pattern.
32315 2017-05-10  Peter Bergner  <bergner@vnet.ibm.com>
32317         PR tree-optimization/51513
32318         * tree-cfg.c (gimple_seq_unreachable_p): New function.
32319         (assert_unreachable_fallthru_edge_p): Use it.
32320         (group_case_labels_stmt): Likewise.
32321         * tree-cfg.h: Prototype it.
32322         * stmt.c: Include cfghooks.h and tree-cfg.h.
32323         (emit_case_dispatch_table) <gap_label>: New local variable.
32324         Use it to fill dispatch table gaps.
32325         Test for default_label before updating probabilities.
32326         (expand_case) <default_label>: Remove unneeded initialization.
32327         Test for unreachable default case statement and remove its edge.
32328         Set default_label accordingly.
32329         * tree-ssa-ccp.c (optimize_unreachable): Update comment.
32331 2017-05-10  Carl Love  <cel@us.ibm.com>
32333         * config/rs6000/rs6000-c: Add support for built-in functions
32334         vector signed char      vec_neg (vector signed char)
32335         vector signed short int vec_neg (vector short int)
32336         vector signed int       vec_neg (vector signed int)
32337         vector signed long long vec_neg (vector signed long long)
32338         vector float            vec_neg (vector float)
32339         vector double           vec_neg (vector double)
32340         * config/rs6000/rs6000-builtin.def: Add definitions for NEG function
32341         overload.
32342         * config/rs6000/altivec.h: Add define for vec_neg
32343         * doc/extend.texi: Update the built-in documentation for the
32344         new built-in functions.
32346 2017-05-10  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
32348         PR tree-optimization/77644
32349         * match.pd (sqrt(x) cmp sqrt(y) -> x cmp y): New pattern.
32351 2017-05-10  Nathan Sidwell  <nathan@acm.org>
32353         * dumpfile.h (TDI_lang_all): New.
32354         (TDF_KIND): New. Renumber others
32355         (TDF_LANG, TDF_TREE, TDF_RTL, TDF_IPA): Enumerate value, rather
32356         than bits.
32357         * dumpfile.c (dump_files): Mark language dumps as TDF_LANG.  add
32358         lang-all.
32359         (get_dump_file_name): Adjust suffix generation.
32360         (dump_enable_all): Use TDF_KIND.
32361         * doc/invoke.texi (-fdump-lang-all): Document.
32363         * dumpfile.h: Tabify.
32365 2017-05-10  Wilco Dijkstra  <wdijkstr@arm.com>
32367         PR target/80671
32368         * config/aarch64/cortex-a57-fma-steering.c (merge_forest):
32369         Move member access before delete.
32371 2017-05-10  Alexandre Oliva  <aoliva@redhat.com>
32373         * tree-inline.c (expand_call_inline): Split block at stmt
32374         before the call.
32376 2017-05-09  Michael Meissner  <meissner@linux.vnet.ibm.com>
32378         PR target/68163
32379         * config/rs6000/rs6000.md (f32_lr): Delete mode attributes that
32380         are now unused after splitting mov{sf,sd}_hardfloat.
32381         (f32_lr2): Likewise.
32382         (f32_lm): Likewise.
32383         (f32_lm2): Likewise.
32384         (f32_li): Likewise.
32385         (f32_li2): Likewise.
32386         (f32_lv): Likewise.
32387         (f32_sr): Likewise.
32388         (f32_sr2): Likewise.
32389         (f32_sm): Likewise.
32390         (f32_sm2): Likewise.
32391         (f32_si): Likewise.
32392         (f32_si2): Likewise.
32393         (f32_sv): Likewise.
32394         (f32_dm): Likewise.
32395         (f32_vsx): Likewise.
32396         (f32_av): Likewise.
32397         (mov<mode>_hardfloat): Split into separate movsf and movsd pieces.
32398         For movsf, order stores so the VSX stores occur before the GPR
32399         store which encourages the register allocator to use a traditional
32400         FPR instead of a GPR.  For movsd, order the stores so that the GPR
32401         store comes before the VSX stores to allow the power6 to work.
32402         This is due to the power6 not having a 32-bit integer store
32403         instruction from a FPR.
32404         (movsf_hardfloat): Likewise.
32405         (movsd_hardfloat): Likewise.
32407 2017-05-09  Martin Sebor  <msebor@redhat.com>
32409         PR translation/80280
32410         * config/sol2-c.c (cmn_err_flag_specs): Initialize new data member
32411         added in r247778.
32413         PR translation/80280
32414         * config/i386/msformat-c.c (ms_printf_flag_specs): Initialize new
32415         data member added in r247778.
32416         (ms_scanf_flag_specs, ms_strftime_flag_specs): Same.
32418 2017-05-09  Nathan Sidwell  <nathan@acm.org>
32420         * tree.h (tree_fits_shwi_p, tree_fits_uhwi_p): Unconditionally pure.
32422         * ipa-devirt.c (default_hash_traits<type_pair>): Skip struct-scope
32423         typedefs.
32425 2017-05-09  Marek Polacek  <polacek@redhat.com>
32427         * doc/invoke.texi: Fix typo.
32429 2017-05-09  Richard Biener  <rguenther@suse.de>
32431         * tree-vrp.c (vrp_val_is_max): Adjust comment.
32432         (vrp_val_is_min): Likewise.
32433         (set_value_range_to_value): Likewise.
32434         (set_value_range_to_nonnegative): Likewise.
32435         (gimple_assign_nonzero_p): Likewise.
32436         (gimple_stmt_nonzero_p): Likewise.
32437         (vrp_int_const_binop): Likewise.  Remove unreachable case.
32438         (adjust_range_with_scev): Adjust comments.
32439         (compare_range_with_value): Likewise.
32440         (extract_range_from_phi_node): Likewise.
32441         (test_for_singularity): Likewise.
32443 2017-05-09  Richard Biener  <rguenther@suse.de>
32445         * tree-vrp.c (get_single_symbol): Add assert that we don't
32446         get overflowed constants as invariant part.
32447         (compare_values_warnv): Add comment before the TREE_NO_WARNING
32448         checks.  Use wi::cmp instead of recursing for integer constants.
32449         (compare_values): Just ignore whether we assumed undefined
32450         overflow instead of failing the compare.
32451         (extract_range_for_var_from_comparison_expr): Add comment before the
32452         TREE_NO_WARNING sets.
32453         (test_for_singularity): Likewise.
32454         (extract_range_from_comparison): Do not disable optimization
32455         when we assumed undefined overflow.
32456         (extract_range_basic): Remove init of unused var.
32458 2017-05-09  Richard Biener  <rguenther@suse.de>
32460         * tree-vrp.c (vrp_int_const_binop): Use wide-ints and simplify.
32461         (extract_range_from_multiplicative_op_1): Adjust.
32462         (extract_range_from_binary_expr_1): Use int_const_binop.
32464 2017-05-08  Kelvin Nilsen  <kelvin@gcc.gnu.org>
32466         PR target/80101
32467         * config/rs6000/power6.md: Replace store_data_bypass_p calls with
32468         rs6000_store_data_bypass_p in seven define_bypass directives and
32469         in several comments.
32470         * config/rs6000/rs6000-protos.h: Add prototype for
32471         rs6000_store_data_bypass_p function.
32472         * config/rs6000/rs6000.c (rs6000_store_data_bypass_p): New
32473         function implements slightly different (rs6000-specific) semantics
32474         than store_data_bypass_p, returning false rather than aborting
32475         with assertion error when arguments do not satisfy the
32476         requirements of store data bypass.
32477         (rs6000_adjust_cost): Replace six calls of store_data_bypass_p with
32478         rs6000_store_data_bypass_p.
32480 2017-05-08  Max Filippov  <jcmvbkbc@gmail.com>
32482         * config/xtensa/xtensa-protos.h
32483         (xtensa_initial_elimination_offset): New declaration.
32484         * config/xtensa/xtensa.c (xtensa_initial_elimination_offset):
32485         New function. Move its body from the INITIAL_ELIMINATION_OFFSET
32486         macro definition, add case for FRAME_POINTER_REGNUM when
32487         FRAME_GROWS_DOWNWARD.
32488         * config/xtensa/xtensa.h (FRAME_GROWS_DOWNWARD): New macro definition.
32489         (INITIAL_ELIMINATION_OFFSET): Replace body with call to
32490         xtensa_initial_elimination_offset.
32492 2017-05-08  Nathan Sidwell  <nathan@acm.org>
32494         * doc/invoke.texi: Alphabetize -fdump options.
32496 2017-05-08  Martin Sebor  <msebor@redhat.com>
32498         PR translation/80280
32499         * config/sol2-c.c (solaris_pragma_align): Correct quoting.
32501 2017-05-08  Bernd Edlinger  <bernd.edlinger@hotmail.de>
32503         * target.def (compute_frame_layout): New optional target hook.
32504         * doc/tm.texi.in (TARGET_COMPUTE_FRAME_LAYOUT): Add hook.
32505         * doc/tm.texi (TARGET_COMPUTE_FRAME_LAYOUT): Add documentation.
32506         * lra-eliminations.c (update_reg_eliminate): Call compute_frame_layout
32507         target hook.
32508         * reload1.c (verify_initial_elim_offsets): Likewise.
32509         * config/arm/arm.c (TARGET_COMPUTE_FRAME_LAYOUT): Define.
32510         (use_simple_return_p): Call arm_compute_frame_layout if needed.
32511         (arm_get_frame_offsets): Split up into this ...
32512         (arm_compute_frame_layout): ... and this function.
32514 2017-05-08  Richard Sandiford  <richard.sandiford@arm.com>
32516         * config/aarch64/constraints.md (Usa): New constraint.
32517         * config/aarch64/aarch64.md (*movsi_aarch64, *movdi_aarch64): Use it.
32519 2017-05-08  Thomas Preud'homme  <thomas.preudhomme@arm.com>
32521         * config.gcc (arm*-*-*): Set TM_MULTILIB_CONFIG from
32522         with_multilib_list after it has been checked.
32524 2017-05-08  Richard Biener  <rguenther@suse.de>
32526         * tree-ssa-pre.c (bitmap_set_and): Avoid bitmap copy.
32527         (bitmap_set_subtract_values): Likewise.
32529 2017-05-08  Richard Biener  <rguenther@suse.de>
32531         * tree-vrp.c (gimple_assign_nonzero_warnv_p): Rename to ...
32532         (gimple_assign_nonzero): ... this and remove strict_overflow_p
32533         argument.
32534         (gimple_stmt_nonzero_warnv_p): Rename to ...
32535         (gimple_stmt_nonzero_p): ... this and remove strict_overflow_p
32536         argument.
32537         (vrp_stmt_computes_nonzero): Remove strict_overflow_p argument.
32538         (extract_range_basic): Adjust, do not disable propagation on
32539         strict overflow sensitive simplification.
32540         (vrp_visit_cond_stmt): Likewise.
32542 2017-05-05  Jan Hubicka  <hubicka@ucw.cz>
32544         * ipa-inline-analysis.c (estimate_function_body_sizes): Recompute
32545         body size unconditionally.
32547 2017-05-07  Jeff Law  <law@redhat.com>
32549         Revert:
32550         2017-05-06  Jeff Law  <law@redhat.com>
32551         PR tree-optimization/78496
32552         * tree-vrp.c (simplify_assert_expr_using_ranges): Remove debugging
32553         code.
32555         PR tree-optimization/78496
32556         * tree-vrp.c (simplify_assert_expr_using_ranges): New function.
32557         (simplify_stmt_using_ranges): Call it.
32558         (vrp_dom_walker::before_dom_children): Extract equivalences
32559         from an ASSERT_EXPR with an equality comparison against a
32560         constant.
32562 2017-05-06  Jeff Law  <law@redhat.com>
32564         PR tree-optimization/78496
32565         * tree-vrp.c (simplify_assert_expr_using_ranges): Remove debugging
32566         code.
32568         PR tree-optimization/78496
32569         * tree-vrp.c (simplify_assert_expr_using_ranges): New function.
32570         (simplify_stmt_using_ranges): Call it.
32571         (vrp_dom_walker::before_dom_children): Extract equivalences
32572         from an ASSERT_EXPR with an equality comparison against a
32573         constant.
32575 2017-05-06  Richard Sandiford  <richard.sandiford@linaro.org>
32577         * lra-constraints.c (lra_copy_reg_equiv): New function.
32578         (split_reg): Use it to copy equivalence information from the
32579         original register to the spill register.
32581 2017-05-06  Richard Sandiford  <richard.sandiford@linaro.org>
32583         PR rtl-optimization/75964
32584         * simplify-rtx.c (simplify_const_relational_operation): Remove
32585         invalid handling of comparisons of integer ABS.
32587 2017-05-06  Uros Bizjak  <ubizjak@gmail.com>
32589         * config/i386/i386.c (ext_80387_constant_init): Do not explicitly
32590         initialize to zero.
32591         (init_regs): Remove declaration.
32592         (function_arg_advance_32): Initialize error_p as boolean variable.
32594 2017-05-05  Nathan Sidwell  <nathan@acm.org>
32596         * store-motion.c (remove_reachable_equiv_notes): Reformat long
32597         lines.  Use for (;;).
32599 2017-05-05  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
32601         * config/rs6000/rs6000.c (rs6000_vect_nonmem): New static var.
32602         (rs6000_init_cost): Initialize rs6000_vect_nonmem.
32603         (rs6000_add_stmt_cost): Update rs6000_vect_nonmem.
32604         (rs6000_finish_cost): Avoid vectorizing simple copy loops with
32605         VF=2 that require versioning.
32607 2017-05-05  David Malcolm  <dmalcolm@redhat.com>
32609         * diagnostic.h (CARET_LINE_MARGIN): Convert from macro to const
32610         int.
32612 2017-05-05  David Malcolm  <dmalcolm@redhat.com>
32614         * diagnostic.h (diagnostic_override_option_index): Convert from
32615         macro to inline function.
32617 2017-05-05  David Malcolm  <dmalcolm@redhat.com>
32619         * diagnostic.c (last_module_changed_p): New function.
32620         (set_last_module): New function.
32621         (diagnostic_report_current_module): Convert macro usage to
32622         the above functions.
32623         * diagnostic.h (diagnostic_context::last_module): Strengthen
32624         from const line_map * to const line_map_ordinary *.
32625         (diagnostic_last_module_changed): Delete macro.
32626         (diagnostic_set_last_module): Delete macro.
32628 2017-05-05  David Malcolm  <dmalcolm@redhat.com>
32630         * diagnostic.c (diagnostic_impl): Replace report_diagnostic
32631         with diagnostic_report_diagnostic.
32632         (diagnostic_n_impl_richloc): Likewise.
32633         * diagnostic.h (report_diagnostic): Delete macro.
32634         * rtl-error.c (diagnostic_for_asm): Replace report_diagnostic
32635         with diagnostic_report_diagnostic.
32636         * substring-locations.c (format_warning_va): Likewise.
32638 2017-05-05  David Malcolm  <dmalcolm@redhat.com>
32640         * diagnostic.c (diagnostic_report_diagnostic): Eliminate
32641         save/restor of format_spec.  Move option-printing code to...
32642         (print_option_information): ...this new function, and
32643         reimplement by simply printing to the pretty_printer,
32644         rather than appending to the format string.
32646 2017-05-05  David Malcolm  <dmalcolm@redhat.com>
32648         * diagnostic.c (diagnostic_report_diagnostic): Split out pragma
32649         handling logic into...
32650         (update_effective_level_from_pragmas): ...this new function.
32652 2017-05-04  Andrew Waterman  <andrew@sifive.com>
32654         * config/riscv/riscv.opt (mstrict-align): New option.
32655         * config/riscv/riscv.h (STRICT_ALIGNMENT): Use it.  Update comment.
32656         (SLOW_UNALIGNED_ACCESS): Define.
32657         (riscv_slow_unaligned_access): Declare.
32658         * config/riscv/riscv.c (riscv_tune_info): Add slow_unaligned_access
32659         field.
32660         (riscv_slow_unaligned_access): New variable.
32661         (rocket_tune_info): Set slow_unaligned_access to true.
32662         (optimize_size_tune_info): Set slow_unaligned_access to false.
32663         (riscv_cpu_info_table): Add entry for optimize_size_tune_info.
32664         (riscv_valid_lo_sum_p): Use TARGET_STRICT_ALIGN.
32665         (riscv_option_override): Set riscv_slow_unaligned_access.
32666         * doc/invoke.texi: Add -mstrict-align to RISC-V.
32668 2017-05-04  Kito Cheng  <kito.cheng@gmail.com>
32670         * config/riscv/riscv.md: Unify indentation.
32672 2017-05-05  Michael Meissner  <meissner@linux.vnet.ibm.com>
32674         PR target/79038
32675         PR target/79202
32676         PR target/79203
32677         * config/rs6000/rs6000.md (u code attribute): Add FIX and
32678         UNSIGNED_FIX.
32679         (extendsi<mode>2): Add support for doing sign extension via
32680         VUPKHSW and XXPERMDI if the value is in Altivec registers and we
32681         don't have ISA 3.0 instructions.
32682         (extendsi<mode>2 splitter): Likewise.
32683         (fix_trunc<mode>si2): If we are at ISA 2.07 (VSX small integer),
32684         generate the normal insns since SImode can now go in vector
32685         registers.  Disallow the special UNSPECs needed for previous
32686         machines to hide SImode being used.  Add new insns
32687         fctiw{,w}_<mode>_smallint if SImode can go in vector registers.
32688         (fix_trunc<mode>si2_stfiwx): Likewise.
32689         (fix_trunc<mode>si2_internal): Likewise.
32690         (fixuns_trunc<mode>si2): Likewise.
32691         (fixuns_trunc<mode>si2_stfiwx): Likewise.
32692         (fctiw<u>z_<mode>_smallint): Likewise.
32693         (fctiw<u>z_<mode>_mem): New combiner pattern to prevent conversion
32694         of floating point to 32-bit integer from doing a direct move to
32695         the GPR registers to do a store.
32696         (fctiwz_<mode>): Break long line.
32698 2017-05-05  Bin Cheng  <bin.cheng@arm.com>
32700         * Makefile.in (GTFILES): Add tree-ssa-loop-ivopts.c.
32701         * tree-ssa-loop-ivopts.c (compute_max_addr_offset): Delete.
32702         (addr_list, addr_offset_valid_p): New.
32703         (split_address_groups): Check offset validity with above function.
32704         (gt-tree-ssa-loop-ivopts.h): Include header file.
32706 2017-05-05  Nathan Sidwell  <nathan@acm.org>
32708         * config.gcc (arm*-*-*): Add missing 'fi'.
32710 2017-05-05  Steve Ellcey  <sellcey@cavium.com>
32712         * doc/invoke.texi (-fopt-info): Explicitly say order of options
32713         included in -fopt-info does not matter.
32714         * doc/optinfo.texi (-fopt-info): Fix description of default
32715         behavour. Explicitly say order of options included in -fopt-info
32716         does not matter.
32718 2017-05-05  Thomas Preud'homme  <thomas.preudhomme@arm.com>
32720         * config.gcc: Allow combinations of aprofile and rmprofile values for
32721         --with-multilib-list.
32722         * config/arm/t-multilib: New file.
32723         * config/arm/t-aprofile: Remove initialization of MULTILIB_*
32724         variables.  Remove setting of ISA and floating-point ABI in
32725         MULTILIB_OPTIONS and MULTILIB_DIRNAMES.  Set architecture and FPU in
32726         MULTI_ARCH_OPTS_A and MULTI_ARCH_DIRS_A rather than MULTILIB_OPTIONS
32727         and MULTILIB_DIRNAMES respectively.  Add comment to introduce all
32728         matches.  Add architecture matches for marvel-pj4 and generic-armv7-a
32729         CPU options.
32730         * config/arm/t-rmprofile: Likewise except for the matches changes.
32731         * doc/install.texi (--with-multilib-list): Document the combination of
32732         aprofile and rmprofile values and warn about pitfalls in doing that.
32734 2017-05-05  Wilco Dijkstra  <wdijkstr@arm.com>
32736         * config/aarch64/aarch64.md (movsi_aarch64): Remove '*' from r=w.
32737         (movdi_aarch64): Likewise.
32739 2017-05-05  Jakub Jelinek  <jakub@redhat.com>
32741         PR tree-optimization/80632
32742         * tree-switch-conversion.c (struct switch_conv_info): Add target_vop
32743         field.
32744         (build_arrays): Initialize it for virtual phis.
32745         (fix_phi_nodes): Use it for virtual phis.
32747         PR tree-optimization/80558
32748         * tree-vrp.c (extract_range_from_binary_expr_1): Optimize
32749         [x, y] op z into [x op, y op z] for op & or | if conditions
32750         are met.
32752 2017-05-05  Andre Vieira  <andre.simoesdiasvieira@arm.com>
32753             Prakhar Bahuguna  <prakhar.bahuguna@arm.com>
32755         PR target/71607
32756         * config/arm/arm.md (use_literal_pool): Remove.
32757         (64-bit immediate split): No longer takes cost into consideration
32758         if arm_disable_literal_pool is enabled.
32759         * config/arm/arm.c (arm_tls_referenced_p): Add diagnostic if TLS is
32760         used when arm_disable_literal_pool is enabled.
32761         (arm_max_const_double_inline_cost): Remove use of
32762         arm_disable_literal_pool.
32763         (push_minipool_fix): Add assert.
32764         (arm_reorg): Add return if arm_disable_literal_pool is enabled.
32765         * config/arm/vfp.md (no_literal_pool_df_immediate): New.
32766         (no_literal_pool_sf_immediate): New.
32768 2017-05-05  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
32770         PR tree-optimization/80613
32771         * tree-ssa-dce.c (propagate_necessity): Remove cases for
32772         BUILT_IN_STRDUP and BUILT_IN_STRNDUP.
32774 2017-05-05  Richard Biener  <rguenther@suse.de>
32776         * tree-ssa-pre.c (get_or_alloc_expr_for): Simplify.
32778 2017-05-05  Georg-Johann Lay  <avr@gjlay.de>
32780         * config/avr/avr.md [flag_strict_overflow]: Remove any occurence
32781         of this flag from insn conditions due to removal from r247495.
32783 2017-05-05  Wilco Dijkstra  <wdijkstr@arm.com>
32785         * config/arm/aarch-common.c (arm_early_load_addr_dep_ptr):
32786         New function.
32787         (arm_early_store_addr_dep_ptr): Likewise.
32788         * config/arm/aarch-common-protos.h
32789         (arm_early_load_addr_dep_ptr): Add prototype.
32790         (arm_early_store_addr_dep_ptr): Likewise.
32791         * config/arm/cortex-a53.md: Add new bypasses.
32793 2017-05-05  Jakub Jelinek  <jakub@redhat.com>
32795         * tree.c (next_type_uid): Change type to unsigned.
32796         (type_hash_canon): Decrement back next_type_uid if
32797         freeing a type node with the highest TYPE_UID.  For INTEGER_TYPEs
32798         also ggc_free TYPE_MIN_VALUE, TYPE_MAX_VALUE and TYPE_CACHED_VALUES
32799         if possible.
32801 2017-05-04  Martin Sebor  <msebor@redhat.com>
32803         * builtins.c: Fix a trivial typo in a comment.
32805         PR middle-end/79234
32806         * builtins.c (check_sizes): Adjust to handle reading past the end.
32807         Avoid printing excessive upper bound of ranges.  Use %E to print
32808         tree nodes instead of converting them to %wu.
32809         (expand_builtin_memchr): New function.
32810         (compute_dest_size): Rename...
32811         (compute_objsize): ...to this.
32812         (expand_builtin_memcpy): Adjust.
32813         (expand_builtin_mempcpy): Adjust.
32814         (expand_builtin_strcat): Adjust.
32815         (expand_builtin_strcpy): Adjust.
32816         (check_strncat_sizes): Adjust.
32817         (expand_builtin_strncat): Adjust.
32818         (expand_builtin_strncpy): Adjust and simplify.
32819         (expand_builtin_memset): Adjust.
32820         (expand_builtin_bzero): Adjust.
32821         (expand_builtin_memcmp): Adjust.
32822         (expand_builtin): Handle memcmp.
32823         (maybe_emit_chk_warning): Check strncat just once.
32825 2017-05-04  Martin Sebor  <msebor@redhat.com>
32827         PR preprocessor/79214
32828         PR middle-end/79222
32829         PR middle-end/79223
32830         * builtins.c (check_sizes): Add inlining context and issue
32831         warnings even when -Wno-system-headers is set.
32832         (check_strncat_sizes): Same.
32833         (expand_builtin_strncat): Same.
32834         (expand_builtin_memmove): New function.
32835         (expand_builtin_stpncpy): Same.
32836         (expand_builtin): Handle memmove and stpncpy.
32838 2017-05-04  Bin Cheng  <bin.cheng@arm.com>
32840         * tree-ssa-loop-ivopts.c (struct cost_pair): Remove field inv_expr
32841         which is not used any more.
32843 2017-05-04  Wilco Dijkstra  <wdijkstr@arm.com>
32845         * config/aarch64/aarch64.c (generic_tunings): Update prefetch model.
32847 2017-05-04  Wilco Dijkstra  <wdijkstr@arm.com>
32849         * config/aarch64/aarch64.c (cortexa35_tunings): Set jump alignment to 4.
32850         (cortexa53_tunings): Likewise.
32851         (cortexa57_tunings): Likewise.
32852         (cortexa72_tunings): Likewise.
32853         (cortexa73_tunings): Likewise.
32855 2017-05-04  Wilco Dijkstra  <wdijkstr@arm.com>
32857         * config/aarch64/aarch64.c (generic_tunings): Set jump alignment to 4.
32858         Set loop alignment to 8.
32860 2017-05-04  Martin Sebor  <msebor@redhat.com>
32862         PR translation/80280
32863         * builtins.c (expand_builtin_object_size): Add missing quoting to
32864         %D and like directives.
32865         * hsa-gen.c (hsa_type_for_scalar_tree_type): Same.
32866         (hsa_type_for_tree_type): Same.
32867         (verify_function_arguments): Same.
32868         * symtab.c (symbol_table::change_decl_assembler_name): Same.
32869         * varasm.c (get_section): Same.
32870         (mark_weak): Same.
32872 2017-05-04  Martin Sebor  <msebor@redhat.com>
32874         PR translation/80280
32875         * config/i386/i386.c (ix86_function_versions): Quote a %D directive.
32877 2017-05-04  Wilco Dijkstra  <wdijkstr@arm.com>
32879         * config/aarch64/aarch64.c (generic_addrcost_table):
32880         Change HI/TI mode setting.
32882 2017-05-04  Martin Jambor  <mjambor@suse.cz>
32884         PR tree-optimization/80622
32885         * tree-sra.c (comes_initialized_p): New function.
32886         (build_accesses_from_assign): Only set write lazily when
32887         comes_initialized_p is false.
32888         (analyze_access_subtree): Use comes_initialized_p.
32889         (propagate_subaccesses_across_link): Assert !comes_initialized_p
32890         instead of testing for PARM_DECL.
32892 2017-05-04  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
32894         * config/aarch64/aarch64.md (prefetch); Adjust predicate and
32895         constraint on operand 0 to allow more general addressing modes.
32896         Adjust output template.
32897         * config/aarch64/aarch64.c (aarch64_address_valid_for_prefetch_p):
32898         New function.
32899         * config/aarch64/aarch64-protos.h
32900         (aarch64_address_valid_for_prefetch_p): Declare prototype.
32901         * config/aarch64/constraints.md (Dp): New address constraint.
32902         * config/aarch64/predicates.md (aarch64_prefetch_operand): New
32903         predicate.
32905 2017-05-04  Jan Hubicka  <hubicka@ucw.cz>
32907         * ipa-cp.c (perform_estimation_of_a_value): Drop base_time parameter;
32908         update use of estimate_ipcp_clone_size_and_time.
32909         (estimate_local_effects): Update use of
32910         estimate_ipcp_clone_size_and_time and perform_estimation_of_a_value.
32911         * ipa-inline.h (estimate_ipcp_clone_size_and_time): Update prototype.
32912         * ipa-inline-analysis.c (estimate_ipcp_clone_size_and_time):
32913         Return nonspecialized time.
32915 2017-05-04  Richard Biener  <rguenther@suse.de>
32917         * tree-ssa-alias.c (get_continuation_for_phi): Improve looking
32918         for the last VUSE which def dominates the PHI.  Directly call
32919         maybe_skip_until.
32920         (get_continuation_for_phi_1): Remove.
32922 2017-05-04  Richard Sandiford  <richard.sandiford@linaro.org>
32924         * tree-ssa-loop-manip.c (niter_for_unrolled_loop): Add commentary
32925         to explain the use of truncating division.  Cap the number of
32926         iterations to the maximum given by nb_iterations_upper_bound,
32927         if defined.
32929 2017-05-04  Thomas Preud'homme  <thomas.preudhomme@arm.com>
32931         * configure.ac (--enable-mingw-wildcard): Add new configurable feature.
32932         * configure: Regenerate.
32933         * config.in: Regenerate.
32934         * config/i386/driver-mingw32.c: new file.
32935         * config/i386/x-mingw32: Add rule to build driver-mingw32.o.
32936         * config.host: Link driver-mingw32.o on MinGW host.
32937         * doc/install.texi: Document new --enable-mingw-wildcard configure
32938         option.
32940 2017-05-04  Marek Polacek  <polacek@redhat.com>
32942         PR tree-optimization/80612
32943         * calls.c (get_size_range): Check for INTEGRAL_TYPE_P.
32945 2017-05-04  Prakhar Bahuguna  <prakhar.bahuguna@arm.com>
32946             Andre Simoes Dias Vieira  <andre.simoesdiasvieira@arm.com>
32948         * config/arm/arm.md (movsi): Change TARGET_32BIT to TARGET_HAVE_MOVT.
32949         (movt splitter): Likewise.
32950         * config/arm/arm.c (arm_option_check_internal): Change arm_arch_thumb2
32951         to TARGET_HAVE_MOVT, and merge with -mslow-flash-data check.
32952         (const_ok_for_arm): Change else to else if (TARGET_THUMB2) and add else
32953         block for Thumb-1 with MOVT.
32954         (thumb2_legitimate_address_p): Move code block ...
32955         (can_avoid_literal_pool_for_label_p): ... into this new function.
32956         (thumb1_legitimate_address_p): Add check for TARGET_HAVE_MOVT and
32957         literal pool.
32958         (thumb_legitimate_constant_p): Add conditional on TARGET_HAVE_MOVT
32959         * doc/invoke.texi (-mpure-code): Change "ARMv7-M targets" for
32960         "M-profile targets with the MOVT instruction".
32962 2017-05-04  Prakhar Bahuguna  <prakhar.bahuguna@arm.com>
32964         * config/arm/arm-builtins.c (arm_init_builtins): Rename
32965         __builtin_arm_ldfscr to __builtin_arm_get_fpscr, and rename
32966         __builtin_arm_stfscr to __builtin_arm_set_fpscr.
32968 2017-05-04  Martin Liska  <mliska@suse.cz>
32970         * tree-vrp.c (simplify_cond_using_ranges_2): Remove unused
32971         variable cond_code.
32973 2017-05-04  Richard Biener  <rguenther@suse.de>
32975         * tree.c (array_at_struct_end_p): Handle arrays at struct
32976         end with flexarrays more conservatively.  Refactor and treat
32977         arrays of arrays or aggregates more strict.  Fix
32978         VIEW_CONVERT_EXPR handling.  Remove allow_compref argument.
32979         * tree.c (array_at_struct_end_p): Adjust prototype.
32980         * emit-rtl.c (set_mem_attributes_minus_bitpos): Adjust.
32981         * gimple-fold.c (get_range_strlen): Likewise.
32982         * tree-chkp.c (chkp_may_narrow_to_field): Likewise.
32984 2017-05-04  Richard Biener  <rguenther@suse.de>
32986         PR tree-optimization/31130
32987         * tree-vrp.c (needs_overflow_infinity): Remove as always returning
32988         false.
32989         (supports_overflow_infinity): Likewise.
32990         (is_negative_overflow_infinity): Likewise.
32991         (is_positive_overflow_infinity): Likewise.
32992         (is_overflow_infinity): Likewise.
32993         (stmt_overflow_infinity): Likewise.
32994         (overflow_infinity_range_p): Likewise.
32995         (usable_range_p): Remove as always returning true.
32996         (make_overflow_infinity): Remove.
32997         (negative_overflow_infinity): Likewise.
32998         (positive_overflow_infinity): Likewise.
32999         (avoid_overflow_infinity): Likewise.
33000         (set_value_range): Adjust accordingly.
33001         (set_value_range_to_nonnegative): Likewise, remove now unused
33002         overflow_infinity arg.
33003         (vrp_operand_equal_p): Adjust.
33004         (update_value_range): Likewise.
33005         (range_int_cst_singleton_p): Likewise.
33006         (operand_less_p): Likewise.
33007         (compare_values_warnv): Likewise.
33008         (extract_range_for_var_from_comparison_expr): Likewise.
33009         (vrp_int_const_binop): Likewise.
33010         (zero_nonzero_bits_from_vr): Likewise.
33011         (extract_range_from_multiplicative_op_1): Likewise.
33012         (extract_range_from_binary_expr_1): Likewise.
33013         (extract_range_from_unary_expr): Likewise.
33014         (extract_range_from_comparison): Likewise.
33015         (extract_range_basic): Likewise.
33016         (adjust_range_with_scev): Likewise.
33017         (compare_ranges): Likewise.
33018         (compare_range_with_value): Likewise.
33019         (dump_value_range): Likewise.
33020         (test_for_singularity): Likewise, remove strict_overflow_p parameter
33021         never used.
33022         (simplify_cond_using_ranges): Adjust.
33024 2017-05-04  Pekka Jääskeläinen  <pekka.jaaskelainen@parmance.com>
33026         * brig-builtins.def: Added a builtin for class_f64.
33027         * builtin-types.def: Added a builtin type needed by class_f64.
33029 2017-05-03  Jason Merrill  <jason@redhat.com>
33031         * timevar.def: Add TV_CONSTEXPR.
33033 2017-05-03  David Malcolm  <dmalcolm@redhat.com>
33035         * common.opt (fdiagnostics-parseable-fixits): Fix typo.
33037 2017-05-03  Martin Jambor  <mjambor@suse.cz>
33039         * ipa-prop.c (ipa_update_after_lto_read): Removed.
33040         * ipa-prop.h (ipa_update_after_lto_read): Remove declaration.
33041         * ipa-cp.c (ipcp_propagate_stage): Do not call
33042         ipa_update_after_lto_read.
33043         * ipa-inline.c (ipa_inline): Likewise.
33045 2017-05-03  Martin Jambor  <mjambor@suse.cz>
33047         * ipa-prop.h (ipa_edge_args): Make a class.  Mark with for_user GTY
33048         tag.  Added a default constructor and a destructor.
33049         (ipa_edge_args_sum_t): New class;
33050         (ipa_edge_args_sum): Declare.
33051         (ipa_edge_args_vector): Remove declaration.
33052         (IPA_EDGE_REF): Use ipa_edge_args_sum.
33053         (ipa_free_edge_args_substructures): Remove declaration.
33054         (ipa_check_create_edge_args): Use ipa_edge_args_sum.
33055         (ipa_edge_args_info_available_for_edge_p): Likewise.
33056         * ipa-prop.c (ipa_edge_args_vector): Removed.
33057         (edge_removal_hook_holder): Likewise.
33058         (edge_duplication_hook_holder): Likewise.
33059         (ipa_edge_args_sum): New variable.
33060         (ipa_propagate_indirect_call_infos): Test ipa_edge_args_sum instead of
33061         ipa_edge_args_vector.
33062         (ipa_free_edge_args_substructures): Likewise.
33063         (ipa_free_all_edge_args): Free ipa_edge_args_sum instead of
33064         ipa_edge_args_vector.
33065         (ipa_edge_removal_hook): Turned into method
33066         ipa_edge_args_sum_t::remove.
33067         (ipa_edge_duplication_hook): Turned into method
33068         ipa_edge_args_sum_t::duplicate.
33069         (ipa_register_cgraph_hooks): Create ipa_edge_args_sum instead of
33070         registering edge hooks.
33071         (ipa_unregister_cgraph_hooks): Do not unregister edge hooks.
33072         * ipa-inline-analysis.c (estimate_function_body_sizes): Test
33073         ipa_edge_args_sum instead of ipa_edge_args_vector.
33074         * ipa-profile.c (ipa_profile): Likewise.
33076 2017-05-03  Martin Jambor  <mjambor@suse.cz>
33078         * symbol-summary.h (function_summary): New method exists.
33079         (function_summary::symtab_removal): Deallocate through release.
33080         (call_summary): New class.
33081         (gt_ggc_mx): New overload.
33082         (gt_pch_nx): Likewise.
33083         (gt_pch_nx): Likewise.
33085 2017-05-03  Jeff Law  <law@redhat.com>
33087         PR tree-optimization/78496
33088         * tree-vrp.c (simplify_cond_using_ranges_1): Renamed
33089         from simplify_cond_using_ranges.  Split off code to walk
33090         backwards through casts into ...
33091         (simplify_cond_using_ranges_2): New function.
33092         (simplify_stmt_using_ranges): Call simplify_cond_using_ranges_1.
33093         (execute_vrp): After identifying jump threads, call
33094         simplify_cond_using_ranges_2.
33096 2017-05-03  Jan Hubicka  <hubicka@ucw.cz>
33098         PR bootstrap/80609
33099         * ipa-inline.h (inline_summary): Add ctor.
33100         (create_ggc): Do not use ggc_cleared_alloc.
33102 2017-05-03  Jeff Downs  <heydowns@somuchpressure.net>
33103             Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
33105         * gcc.c (handle_braces): Support escaping in switch matching
33106         text.
33107         * doc/invoke.texi (Spec Files): Document it.
33108         Remove superfluous @code markup in items.
33110 2017-05-03  David Malcolm  <dmalcolm@redhat.com>
33112         * diagnostic-show-locus.c (struct column_range): New struct.
33113         (get_affected_columns): New function.
33114         (get_printed_columns): New function.
33115         (struct correction): New struct.
33116         (correction::ensure_capacity): New function.
33117         (correction::ensure_terminated): New function.
33118         (struct line_corrections): New struct.
33119         (line_corrections::~line_corrections): New dtor.
33120         (line_corrections::add_hint): New function.
33121         (layout::print_trailing_fixits): Reimplement in terms of the new
33122         classes.
33123         (selftest::test_overlapped_fixit_printing): New function.
33124         (selftest::diagnostic_show_locus_c_tests): Call it.
33126 2017-05-03  Nathan Sidwell  <nathan@acm.org>
33128         Canonicalize canonical type hashing
33129         * tree.h (type_hash_canon_hash): Declare.
33130         * tree.c (type_hash_list, attribute_hash_list): Move into
33131         type_hash_canon_hash.
33132         (build_type_attribute_qual_variant): Break out hash code calc into
33133         type_hash_canon_hash.
33134         (type_hash_canon_hash): New.  Generic type hash computation.
33135         (build_range_type_1, build_array_type_1, build_function_type,
33136         build_method_type_directly, build_offset_type, build_complex_type,
33137         make_vector_type): Call it.
33139 2017-05-03  Richard Biener  <rguenther@suse.de>
33141         * tree-vect-data-refs.c (vect_enhance_data_refs_alignment):
33142         When all DRs have unknown misaligned do not always peel
33143         when there is a store but apply the same costing model as if
33144         there were only loads.
33146 2017-05-03  Richard Biener  <rguenther@suse.de>
33148         Revert
33149         PR tree-optimization/80492
33150         * tree-ssa-alias.c (decl_refs_may_alias_p): Handle
33151         compare_base_decls returning dont-know properly.
33153 2017-05-03  Thomas Preud'homme  <thomas.preudhomme@arm.com>
33155         * config/arm/iterators.md (CCSI): New mode iterator.
33156         (arch): New mode attribute.
33157         * config/arm/sync.md (atomic_compare_and_swap<mode>_1): Rename into ...
33158         (atomic_compare_and_swap<CCSI:arch><NARROW:mode>_1): This and ...
33159         (atomic_compare_and_swap<CCSI:arch><SIDI:mode>_1): This.  Use CCSI
33160         code iterator for success result mode.
33161         * config/arm/arm.c (arm_expand_compare_and_swap): Adapt code to use
33162         the corresponding new insn generators.
33164 2017-05-03  Bin Cheng  <bin.cheng@arm.com>
33166         Revert r247509
33167         2017-05-02  Bin Cheng  <bin.cheng@arm.com>
33168         * rtlanal.c (rtx_cost): Handle TRUNCATE between tieable modes.
33170 2017-05-03  Richard Sandiford  <richard.sandiford@linaro.org>
33172         * tree-data-ref.h (SUB_CONFLICTS_IN_A): Wrap SUB argument in brackets.
33173         (SUB_CONFLICTS_IN_B, SUB_LAST_CONFLICT, SUB_DISTANCE): Likewise.
33174         (DDR_A): Wrap DDR argument in brackets.
33175         (DDR_B, DDR_AFFINE_P, DDR_ARE_DEPENDENT, DDR_SUBSCRIPTS): Likewise.
33176         (DDR_LOOP_NEST, DDR_INNER_LOOP, DDR_SELF_REFERENCE): Likewise.
33177         (DDR_REVERSED_P): Likewise.
33179 2017-05-03  Jakub Jelinek  <jakub@redhat.com>
33181         PR tree-optimization/79472
33182         * tree-switch-conversion.c (struct switch_conv_info): Add
33183         contiguous_range and default_case_nonstandard fields.
33184         (collect_switch_conv_info): Compute contiguous_range and
33185         default_case_nonstandard fields, don't clear final_bb if
33186         contiguous_range and only the default case doesn't have the required
33187         structure.
33188         (check_all_empty_except_final): Set default_case_nonstandard instead
33189         of failing if contiguous_range and the default case doesn't have empty
33190         block.
33191         (check_final_bb): Add SWTCH argument, don't fail if contiguous_range
33192         and only the default case doesn't have the required constants.  Skip
33193         virtual phis.
33194         (gather_default_values): Skip virtual phis.  Allow non-NULL CASE_LOW
33195         if default_case_nonstandard.
33196         (build_constructors): Build constant 1 just once.  Assert that default
33197         values aren't inserted in between cases if contiguous_range.  Skip
33198         virtual phis.
33199         (build_arrays): Skip virtual phis.
33200         (prune_bbs): Add DEFAULT_BB argument, don't remove that bb.
33201         (fix_phi_nodes): Don't add e2f phi arg if default_case_nonstandard.
33202         Handle virtual phis.
33203         (gen_inbound_check): Handle default_case_nonstandard case.
33204         (process_switch): Adjust check_final_bb caller.  Call
33205         gather_default_values with the first non-default case instead of
33206         default case if default_case_nonstandard.
33208 2017-05-02  Nathan Sidwell  <nathan@acm.org>
33210         * ggc-page.c (move_ptes_to_front): Replace unsigned >0 with i--
33211         check.  Fix formatting.
33213 2017-05-02  Jan Hubicka  <hubicka@ucw.cz>
33215         * ipa-inline-analysis.c (estimate_node_size_and_time): Allow roundoff
33216         errors when comparing specialized and unspecialized times.
33218 2017-05-02  David Malcolm  <dmalcolm@redhat.com>
33220         * diagnostic-show-locus.c
33221         (layout::should_print_annotation_line_p): Make private.
33222         (layout::print_annotation_line): Make private.
33223         (layout::annotation_line_showed_range_p): Make private.
33224         (layout::show_ruler): Make private.
33225         (layout::print_source_line): Make private.  Pass in line and
33226         line_width, rather than calling location_get_source_line.  Drop
33227         returned value.
33228         (layout::print_leading_fixits): New method.
33229         (layout::print_any_fixits): Rename to...
33230         (layout::print_trailing_fixits): ...this, and make private.
33231         Don't print newline fixits.
33232         (diagnostic_show_locus): Move logic for printing one row into...
33233         (layout::print_line): ...this new function.  Move the
33234         location_get_source_line call and error-handling from
33235         print_source_line to here.  Call print_leading_fixits, and rename
33236         print_any_fixits to print_trailing_fixits.
33237         (selftest::test_fixit_insert_containing_newline): Update now that
33238         newlines are partially supported.
33239         (selftest::test_fixit_insert_containing_newline_2): New test.
33240         (selftest::test_fixit_replace_containing_newline): Update comments.
33241         (selftest::diagnostic_show_locus_c_tests): Call the new test.
33242         * edit-context.c (class added_line): New class.
33243         (class edited_line): Describe newline handling in comment.
33244         (edited_line::actually_edited_p): New method.
33245         (edited_line::print_content): Delete redundant decl.
33246         (edited_line::m_predecessors): New field.
33247         (edited_file::print_content): Call edited_line::print_content.
33248         (edited_file::print_diff): Update to support newlines.
33249         (edited_file::print_diff_hunk): Likewise.
33250         (edited_file::print_run_of_changed_lines): New function.
33251         (edited_file::print_diff_line): Convert to...
33252         (print_diff_line): ...this.
33253         (edited_file::get_effective_line_count): New function.
33254         (edited_line::edited_line): Initialize new field m_predecessors.
33255         (edited_line::~edited_line): Clean up m_predecessors.
33256         (edited_line::apply_fixit): Handle newlines.
33257         (edited_line::get_effective_line_count): New function.
33258         (edited_line::print_content): New function.
33259         (edited_line::print_diff_lines): New function.
33260         (selftest::test_applying_fixits_insert_containing_newline): New
33261         test.
33262         (selftest::test_applying_fixits_replace_containing_newline): New
33263         test.
33264         (selftest::insert_line): New function.
33265         (selftest::test_applying_fixits_multiple_lines): Add example of
33266         inserting a line.
33267         (selftest::edit_context_c_tests): Call the new tests.
33269 2017-05-02  Bin Cheng  <bin.cheng@arm.com>
33271         * tree-ssa-loop-ivopts.c (get_scaled_computation_cost_at): Delete
33272         parameter cand.  Update dump information.
33273         (get_computation_cost): Update uses.
33275 2017-05-02  Bin Cheng  <bin.cheng@arm.com>
33277         * tree-ssa-loop-ivopts.c (get_computation_aff_1): New.
33278         (get_computation_aff): Reorder parameters.  Use get_computation_aff_1.
33279         (get_computation_at, rewrite_use_address): Update use of
33280         get_computation_aff.
33282 2017-05-02  Bin Cheng  <bin.cheng@arm.com>
33284         * tree-ssa-loop-ivopts.c (get_computation_at): Reorder parameters.
33285         (get_computation): Delete.
33286         (get_computation_cost): Implement like get_computation_cost_at.
33287         Use get_computation_at.
33288         (get_computation_cost_at): Delete.
33289         (rewrite_use_nonlinear_expr): Use get_computation_at.
33290         (rewrite_use_compare, remove_unused_ivs): Ditto.
33292 2017-05-02  Bin Cheng  <bin.cheng@arm.com>
33294         * tree-ssa-loop-ivopts.c (rewrite_use_address): Simple refactor.
33296 2017-05-02  Bin Cheng  <bin.cheng@arm.com>
33298         * tree-ssa-loop-ivopts.c (struct iv_ca): Rename n_regs to n_invs.
33299         (ivopts_global_cost_for_size): Rename parameter and update uses.
33300         (iv_ca_recount_cost): Update uses.
33301         (iv_ca_set_remove_invs, iv_ca_set_no_cp): Record invariants and
33302         candidates seperately in n_invs and n_cands.
33303         (iv_ca_set_add_invs, iv_ca_set_cp, iv_ca_new): Ditto.
33305 2017-05-02  Bin Cheng  <bin.cheng@arm.com>
33307         * tree-ssa-loop-ivopts.c (struct walk_tree_data): New.
33308         (find_inv_vars_cb): New.
33309         (find_depends): Renamed to ...
33310         (find_inv_vars): ... this.
33311         (add_candidate_1, force_var_cost): Call find_inv_vars.
33312         (split_address_cost, determine_group_iv_cost_cond): Ditto.
33314 2017-05-02  Bin Cheng  <bin.cheng@arm.com>
33316         * tree-ssa-loop-ivopts.c (struct cost_pair): Rename depends_on to
33317         inv_vars.  Add inv_exprs.
33318         (struct iv_cand): Rename depends_on to inv_vars.
33319         (struct ivopts_data): Rename max_inv_id/n_invariant_uses to
33320         max_inv_var_id/n_inv_var_uses.  Move max_inv_expr_id around.
33321         Refactor field used_inv_exprs from has_map to array n_inv_expr_uses.
33322         (dump_cand): Dump inv_vars.
33323         (tree_ssa_iv_optimize_init): Support inv_vars and inv_exprs.
33324         (record_invariant, find_depends, add_candidate_1): Ditto.
33325         (set_group_iv_cost, force_var_cost): Ditto.
33326         (split_address_cost, ptr_difference_cost, difference_cost): Ditto.
33327         (get_computation_cost_at, get_computation_cost): Ditto.
33328         (determine_group_iv_cost_generic): Ditto.
33329         (determine_group_iv_cost_address): Ditto.
33330         (determine_group_iv_cost_cond, autoinc_possible_for_pair): Ditto.
33331         (determine_group_iv_costs): Ditto.
33332         (iv_ca_recount_cost): Update call to ivopts_global_cost_for_size.
33333         (iv_ca_set_remove_invariants): Renamed to ...
33334         (iv_ca_set_remove_invs): ... this.  Support inv_vars and inv_exprs.
33335         (iv_ca_set_no_cp): Use iv_ca_set_remove_invs.
33336         (iv_ca_set_add_invariants):  Renamed to ...
33337         (iv_ca_set_add_invs): ... this.  Support inv_vars and inv_exprs.
33338         (iv_ca_set_cp): Use iv_ca_set_add_invs.
33339         (iv_ca_has_deps): Support inv_vars and inv_exprs.
33340         (iv_ca_new, iv_ca_free, iv_ca_dump, free_loop_data): Ditto.
33341         (create_new_ivs): Remove useless dump.
33343 2017-05-02  Bin Cheng  <bin.cheng@arm.com>
33345         * tree-ssa-loop-ivopts.c (get_computation_cost_at): Remove pseudo
33346         iv_cand code.
33347         (determine_group_iv_cost_cond, determine_iv_cost): Ditto.
33348         (iv_ca_set_no_cp, create_new_iv): Ditto.
33350 2017-05-02  Bin Cheng  <bin.cheng@arm.com>
33352         * rtlanal.c (rtx_cost): Handle TRUNCATE between tieable modes.
33354 2017-05-02  Xi Ruoyao  <ryxi@stu.xidian.edu.cn>
33356         * tree.h (EXPR_CILK_SPAWN): Use macro TREE_CHECK2 instead of
33357         function tree_check2.
33359 2017-05-02  Martin Liska  <mliska@suse.cz>
33361         * doc/gcov.texi: Add missing preposition.
33362         * gcov.c (function_info::function_info): Properly fill up
33363         all member variables.
33365 2017-05-02  Tamar Christina  <tamar.christina@arm.com>
33367         * expr.c (expand_expr_real_2): Re-cost if previous costs are the same.
33369 2017-05-02  Tamar Christina  <tamar.christina@arm.com>
33371         * simplify-rtx.c (simplify_binary_operation_1): Add LSHIFTRT case.
33373 2017-05-02  Martin Liska  <mliska@suse.cz>
33375         PR lto/77954.
33376         * lto-streamer-in.c (lto_read_tree_1): Remove
33377         LTO_STREAMER_DEBUG.
33378         * lto-streamer.c (struct tree_hash_entry): Likewise.
33379         (struct tree_entry_hasher): Likewise.
33380         (tree_entry_hasher::hash): Likewise.
33381         (tree_entry_hasher::equal): Likewise.
33382         (lto_streamer_init): Likewise.
33383         (lto_orig_address_map): Likewise.
33384         (lto_orig_address_get): Likewise.
33385         (lto_orig_address_remove): Likewise.
33386         * lto-streamer.h: Likewise.
33387         * tree-streamer-in.c (streamer_alloc_tree): Likewise.
33388         * tree-streamer-out.c (streamer_write_tree_header): Likewise.
33390 2017-05-02  Sebastian Peryt  <sebastian.peryt@intel.com>
33392         * config/i386/avx512fintrin.h (_mm_mask_add_round_sd)
33393         (_mm_maskz_add_round_sd, _mm_mask_add_round_ss)
33394         (mm_maskz_add_round_ss, _mm_mask_sub_round_sd)
33395         (mm_maskz_sub_round_sd, _mm_mask_sub_round_ss)
33396         (mm_maskz_sub_round_ss, _mm_mask_add_sd)
33397         (mm_maskz_add_sd, _mm_mask_add_ss, _mm_maskz_add_ss)
33398         (mm_mask_sub_sd, _mm_maskz_sub_sd, _mm_mask_sub_ss)
33399         (mm_maskz_sub_ss): New intrinsics.
33400         * config/i386/i386-builtin-types.def (V2DF_FTYPE_V2DF_V2DF_V2DF_UQI_INT)
33401         (V4SF_FTYPE_V4SF_V4SF_V4SF_UQI_INT): New function type aliases.
33402         * config/i386/i386-builtin.def (__builtin_ia32_addsd_mask_round)
33403         (__builtin_ia32_addss_mask_round, __builtin_ia32_subsd_mask_round)
33404         (__builtin_ia32_subss_mask_round): New builtins.
33405         * config/i386/i386.c (V2DF_FTYPE_V2DF_V2DF_V2DF_UQI_INT)
33406         (V4SF_FTYPE_V4SF_V4SF_V4SF_UQI_INT): Handle new types.
33407         * config/i386/sse.md (<sse>_vm<plusminus_insn><mode>3<round_name>):
33408         Renamed to ...
33409         (<sse>_vm<plusminus_insn><mode>3<mask_name><round_name>): ... this.
33410         (v<plusminus_mnemonic><ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|%0, %1, %<iptr>2<round_op3>}):
33411         Changed to ...
33412         (v<plusminus_mnemonic><ssescalarmodesuffix>\t{<round_mask_op3>%2, %1, %0<mask_operand3>|%0<mask_operand3>, %1, %<iptr>2<round_mask_op3>}):
33413         ... this.
33415 2017-05-02  Martin Jambor  <mjambor@suse.cz>
33417         PR tree-optimization/78687
33418         * tree-sra.c (access): New field parent.
33419         (process_subtree_disqualification): New function.
33420         (disqualify_candidate): Call it.
33421         (build_accesses_from_assign): Reset write flag if creating an
33422         assighnment link.
33423         (build_access_subtree): Fill in parent field and also prpagate
33424         down grp_write flag.
33425         (create_artificial_child_access): New parameter set_grp_write, set
33426         grp_write to its value.
33427         (propagate_subaccesses_across_link): Also propagate grp_write flag
33428         values.
33429         (propagate_all_subaccesses): Push the closest parent back to work
33430         queue if add_access_to_work_queue returned true.
33432 2017-05-02  Richard Biener  <rguenther@suse.de>
33434         * common.opt (fstrict-overflow): Alias negative to fwrapv.
33435         * doc/invoke.texi (fstrict-overflow): Remove all traces of
33436         -fstrict-overflow documentation.
33437         * tree.h (TYPE_OVERFLOW_UNDEFINED): Do not test flag_strict_overflow.
33438         (POINTER_TYPE_OVERFLOW_UNDEFINED): Test !flag_wrapv instead of
33439         flag_strict_overflow.
33440         * ipa-inline.c (can_inline_edge_p): Do not test flag_strict_overflow.
33441         * lto-opts.c (lto_write_options): Do not stream it.
33442         * lto-wrapper.c (merge_and_complain): Do not handle it.
33443         * opts.c (default_options_table): Do not set -fstrict-overflow.
33444         (finish_options): Likewise do not clear it when sanitizing.
33445         * simplify-rtx.c (simplify_const_relational_operation): Do not
33446         test flag_strict_overflow.
33448 2017-05-02  Uros Bizjak  <ubizjak@gmail.com>
33450         * config/alpha/alpha.md (*add<mode>3_ieee): Merge to add<mode>3
33451         using enabled attribute.
33452         (*sub<mode>3_ieee): Merge to sub<mode>3 using enabled attribute.
33453         (*mul<mode>3_ieee): Merge to mul<mode>3 using enabled attribute.
33454         (*div<mode>3_ieee): Merge to div<mode>3 using enabled attribute.
33455         (*sqrt<mode>2_ieee): Merge to sqrt<mode>2 using enabled attribute.
33456         (*fix_truncdfdi_ieee): Merge to *fix_truncdfdi2 using enabled attribute.
33457         (*fix_truncsfdi_ieee): Merge to *fix_truncsfdi2 using enabled attribute.
33458         (*floatdisf_ieee): Merge to floatdisf2 using enabled attribute.
33459         (*floatdidf_ieee): Merge to floatdidf2 using enabled attribute.
33460         (*truncdfsf2_ieee): Merge to truncdfsf2 using enabled attribute.
33461         (*cmpdf_ieee): Merge to *cmpdf_internal using enabled attribute.
33463 2017-05-02  Uros Bizjak  <ubizjak@gmail.com>
33465         * config/i386/i386.c (ix86_code_end): Use {FIRST,LAST}_INT_REG.
33467 2017-05-02  Richard Biener  <rguenther@suse.de>
33469         PR tree-optimization/80591
33470         Revert
33471         2017-04-10  Richard Biener  <rguenther@suse.de>
33473         * tree-ssa-structalias.c (find_func_aliases): Properly handle
33474         asm inputs.
33476 2017-05-02  Richard Biener  <rguenther@suse.de>
33478         PR tree-optimization/80549
33479         * tree-cfgcleanup.c (mfb_keep_latches): New helper.
33480         (cleanup_tree_cfg_noloop): Create forwarders to known loop
33481         headers if they do not have a preheader.
33483 2017-05-02  Martin Liska  <mliska@suse.cz>
33485         PR other/80589
33486         * common.opt: Fix typo.
33487         * doc/invoke.texi: Likewise.
33489 2017-05-01  Jan Beulich  <jbeulich@suse.com>
33491         * config/i386/sse.md (xop_vpermil2<mode>3): Do not allow operand
33492         swapping, add (x,x,m,x,n) alternative.
33494 2017-05-01  Nathan Sidwell  <nathan@acm.org>
33496         * calls.c (combine_pending_stack_adjustment_and_call): Remove
33497         unnecessary unadjusted_alignment check.
33499 2017-05-01  Xi Ruoyao  <ryxi@stu.xidian.edu.cn>
33501         PR c++/80038
33502         * cilk_common.c (expand_builtin_cilk_detach): Move pedigree
33503         operations here.
33504         * gimplify.c (gimplify_cilk_detach): New function.
33505         (gimplify_call_expr, gimplify_modify_expr): Call it as needed.
33506         * tree-core.h: Document EXPR_CILK_SPAWN.
33507         * tree.h (EXPR_CILK_SPAWN): Define.
33509 2017-05-01  David Malcolm  <dmalcolm@redhat.com>
33511         * diagnostic-show-locus.c (layout::get_expanded_location): Rewrite
33512         to use new fixit_hint representation, using the "replace" logic.
33513         (get_line_span_for_fixit_hint): Likewise.
33514         (layout::print_any_fixits): Likewise.
33515         (selftest::test_one_liner_many_fixits): Rename to...
33516         (selftest::test_one_liner_many_fixits_1): ...this, and update
33517         comment and expected output to reflect that the multiple fix-it
33518         hints are now consolidated into one insertion.
33519         (selftest::test_one_liner_many_fixits_2): New test.
33520         (selftest::test_diagnostic_show_locus_one_liner): Update for
33521         above.
33522         (selftest::test_fixit_consolidation): Update for fix-it API
33523         change.
33524         * diagnostic.c (print_parseable_fixits): Likewise.
33525         * edit-context.c (edited_line::m_line_events): Convert from
33526         auto_vec <line_event *> to auto_vec <line_event>.
33527         (class line_event): Convert from abstract base class to a concrete
33528         class, taking over the role of replace_event.
33529         (class insert_event): Delete.
33530         (class replace_event): Rename to class line_event.  Convert to
33531         half-open range.
33532         (edit_context::add_fixits): Reimplement.
33533         (edit_context::apply_insert): Delete.
33534         (edit_context::apply_replace): Rename to...
33535         (edit_context::apply_fixit): ...this.  Convert to half-open range.
33536         (edited_file::apply_insert): Delete.
33537         (edited_file::apply_replace): Rename to...
33538         (edited_file::apply_fixit): ...this.
33539         (edited_line::~edited_line): Drop deletion of events.
33540         (edited_line::apply_insert): Delete.
33541         (edited_line::apply_replace): Rename to...
33542         (edited_line::apply_fixit): ...this.  Convert to half-open range.
33543         Update for change to type of m_line_events.
33544         * edit-context.h (edit_context::apply_insert): Delete.
33545         (edit_context::apply_replace): Rename to...
33546         (edit_context::apply_fixit): ...this.
33548 2017-05-01  Martin Sebor  <msebor@redhat.com>
33550         * gimple-ssa-sprintf.c (format_integer): Set knownrange when it's
33551         known.
33553 2017-05-01  Uros Bizjak  <ubizjak@gmail.com>
33555         PR target/68491
33556         * config/i386/cpuid.h (__get_cpuid): Always return 0 when
33557         __get_cpuid_max returns 0.
33558         (__get_cpuid_count): Ditto.
33560 2017-05-01  Eric Botcazou  <ebotcazou@adacore.com>
33562         * tree.c (substitute_in_expr) <tcc_vl_exp>: Also inline a call if the
33563         replacement expression is another instance of one of its arguments.
33565 2017-05-01  Jakub Jelinek  <jakub@redhat.com>
33567         PR target/79430
33568         * rtlanal.c (reg_set_p): If reg is a stack_pointer_rtx, also
33569         check for stack push/pop autoinc.
33570         * config/i386/i386.c (ix86_agi_dependent): Return false
33571         if the only reason why modified_in_p returned true is that
33572         addr is SP based and set_insn is a push or pop.
33574 2017-04-29  Jan Hubicka  <hubicka@ucw.cz>
33576         * ipa-inline.c (compute_inlined_call_time): Remove now unnecesary
33577         overflow check.
33579 2017-04-29  Jan Hubicka  <hubicka@ucw.cz>
33581         PR ipa/79224
33582         * ipa-inline-analysis.c (dump_predicate): Add optional parameter NL.
33583         (account_size_time): Use two predicates - exec_pred and
33584         nonconst_pred_ptr.
33585         (evaluate_conditions_for_known_args): Compute both clause and
33586         nonspec_clause.
33587         (evaluate_properties_for_edge): Evaulate both clause and nonspec_clause.
33588         (inline_summary_t::duplicate): Update.
33589         (estimate_function_body_sizes): Caluculate exec and nonconst predicates
33590         separately.
33591         (compute_inline_parameters): Likewise.
33592         (estimate_edge_size_and_time): Update caluclation of time.
33593         (estimate_node_size_and_time): Compute both time and nonspecialized
33594         time.
33595         (estimate_ipcp_clone_size_and_time): Update.
33596         (inline_merge_summary): Update.
33597         (do_estimate_edge_time): Update.
33598         (do_estimate_edge_size): Update.
33599         (do_estimate_edge_hints): Update.
33600         (inline_read_section, inline_write_summary): Stream both new predicates.
33601         * ipa-inline.c (compute_uninlined_call_time): Take uninlined_call_time
33602         as argument.
33603         (compute_inlined_call_time): Cleanup.
33604         (big_speedup_p): Update.
33605         (edge_badness): Update.
33606         * ipa-inline.h (INLINE_TIME_SCALE): Remove.
33607         (size_time_entry): Replace predicate by exec_predicate and
33608         nonconst_predicate.
33609         (edge_growth_cache_entry): Cache both time nad nonspecialized time.
33610         (estimate_edge_time): Return also nonspec_time.
33611         (reset_edge_growth_cache): Update.
33613 2017-04-29  Jakub Jelinek  <jakub@redhat.com>
33615         PR rtl-optimization/80491
33616         * ifcvt.c (noce_process_if_block): When looking for x setter
33617         with missing else_bb, don't check only the insn right before
33618         cond_earliest, but look for the last insn that x is modified in
33619         within the same bb.
33621         PR rtl-optimization/80491
33622         * alias.c (memory_modified_in_insn_p): Return true for CALL_INSNs.
33624 2017-04-29  Marc Glisse  <marc.glisse@inria.fr>
33626         PR tree-optimization/80487
33627         * tree-ssa-alias.c (stmt_kills_ref_p): Handle stpncpy and strncpy.
33629 2017-04-29  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
33631         PR tree-optimization/79697
33632         * tree-ssa-dce.c (mark_stmt_if_obviously_necessary): Check if callee
33633         is BUILT_IN_STRDUP, BUILT_IN_STRNDUP, BUILT_IN_REALLOC.
33634         (propagate_necessity): Check if def_callee is BUILT_IN_STRDUP or
33635         BUILT_IN_STRNDUP.
33636         * gimple-fold.c (gimple_fold_builtin_realloc): New function.
33637         (gimple_fold_builtin): Call gimple_fold_builtin_realloc.
33639 2017-04-28  Martin Sebor  <msebor@redhat.com>
33641         PR tree-optimization/80523
33642         * gimple-ssa-sprintf.c (target_to_host_charmap): New global variable.
33643         (init_target_to_host_charmap, target_to_host, target_strtol10): New
33644         functions.
33645         (maybe_warn, format_directive, parse_directive): Use new functions.
33646         (pass_sprintf_length::execute): Call init_target_to_host_charmap.
33648 2017-04-28  Marc Glisse  <marc.glisse@inria.fr>
33650         * match.pd (X+Z OP Y+Z, X-Z OP Y-Z, Z-X OP Z-Y): New transformations.
33652 2017-04-28  Bernd Edlinger  <bernd.edlinger@hotmail.de>
33654         * configure.ac (SYSTEM_HEADER_DIR, BUILD_SYSTEM_HEADER_DIR,
33655         target_header_dir): Set correctly.
33656         * configure: Regenerated.
33657         * Makefile.in (BUILD_SYSTEM_HEADER_DIR): New make variabe.
33658         (LIMITS_H_TEST, if_multiarch, stmp-fixinc): Use BUILD_SYSTEM_HEADER_DIR
33659         instead of SYSTEM_HEADER_DIR.
33661 2017-04-28  Jan Hubicka  <hubicka@ucw.cz>
33663         * ipa-cp.c (perform_estimation_of_a_value): Turn time to sreal.
33664         (estimate_local_effects): Likewise.
33665         * ipa-inline.c (compute_inlined_call_time, want_inline_small_function_p,
33666         edge_badness, inline_small_functions, dump_overall_stats): LIkewise.
33667         * ipa-inline.h (edge_growth_cache_entry, estimate_time_after_inlining,
33668         estimate_ipcp_clone_size_and_time, do_estimate_edge_time,
33669         do_estimate_edge_time, estimate_edge_time): Likewise.
33670         * ipa-inline-analysis.c (estimate_node_size_and_time,
33671         estimate_ipcp_clone_size_and_time, do_estimate_edge_time): Likewise.
33672         (estimate_time_after_inlining): Remove.
33674 2017-04-28  Martin Liska  <mliska@suse.cz>
33676         * doc/gcov.texi: Enhance documentation of gcov.
33678 2017-04-28  Martin Liska  <mliska@suse.cz>
33680         * doc/gcov.texi: Sort options in alphabetic order.
33681         * doc/gcov-dump.texi: Likewise.
33682         * doc/gcov-tool.texi: Likewise.
33683         * gcov.c (print_usage): Likewise.
33684         * gcov-dump.c (print_usage): Likewise.
33685         * gcov-tool.c (print_merge_usage_message): Likewise.
33686         (print_rewrite_usage_message): Likewise.
33687         (print_overlap_usage_message): Likewise.
33689 2017-04-28  Martin Liska  <mliska@suse.cz>
33691         PR gcov-profile/53915
33692         * gcov.c (format_gcov): Print 'NAN %' when top > bottom.
33694 2017-04-28  Martin Liska  <mliska@suse.cz>
33696         PR gcov-profile/79891
33697         * gcov.c (add_line_counts): Assign BBs to lines just if the BB
33698         is marked by compiler as living on a line.
33699         (get_cycles_count): Remove usage of the union.
33700         (output_intermediate_file): Likewise.
33701         (find_source): Fix GNU coding style.
33702         (accumulate_line_counts): Remove old non-all block mode.
33703         (output_lines): Remove usage of the union.
33704         * profile.c (output_location): Include all BBs, even if
33705         belonging to a same line (and file) as a previous BB.
33707 2017-04-28  Martin Liska  <mliska@suse.cz>
33709         * gcov.c (process_args): Handle new argument 'w'.
33710         (read_graph_file): Assign ID to BBs.
33711         (output_branch_count): Display BB # if verbose flag is set.
33712         (output_lines): Likewise for arcs.
33713         (print_usage): Add '--verbose' option help.
33714         * doc/gcov.texi: Document --verbose (-w) option.
33716 2017-04-28  Martin Liska  <mliska@suse.cz>
33718         * gcov.c (struct block_location_info): New struct.
33719         (process_file): Fill up the new structure.
33720         (read_graph_file): Replace usage of encoding by the newly added
33721         struct.
33722         (add_line_counts): Likewise.
33723         (accumulate_line_counts): Remove usage of the union.
33724         (function_info::function_info): New function.
33725         (function_info::~function_info): Likewise.
33726         (process_file): Call delete instead of release_function.
33727         (release_function): Release the function.
33728         (release_structures): Call delete instead of release_function.
33729         (solve_flow_graph): Replace usage of num_blocks.
33730         (find_exception_blocks): Likewise.
33731         (output_lines): Fix GNU coding style.
33733 2017-04-28  Martin Liska  <mliska@suse.cz>
33735         PR driver/56469
33736         * coverage.c (coverage_remove_note_file): New function.
33737         * coverage.h: Declare the function.
33738         * toplev.c (finalize): Clean if an error has been seen.
33740 2017-04-28  Martin Liska  <mliska@suse.cz>
33742         PR gcov-profile/80031
33743         * gcov-dump.c (tag_blocks): Just print number of basic blocks.
33744         * gcov-io.h (GCOV_TAG_BLOCKS_NUM): Remove unused macro.
33745         * gcov.c (read_graph_file): Read just number of blocks.
33746         * profile.c (branch_prob): Do not stream 0 flags per a basic
33747         block.
33749 2017-04-28  Martin Liska  <mliska@suse.cz>
33751         * gcov-dump.c (tag_*): Add new argument to declarations.
33752         (dump_gcov_file): Likewise.
33753         (tag_blocks): Add and use new argument depth.
33754         (tag_arcs): Likewise.
33755         (tag_lines): Likewise.
33756         (tag_counters): Likewise.
33757         (tag_summary): Likewise.
33758         (dump_working_sets): Use depth to do a proper indentation.
33760 2017-04-28  Jakub Jelinek  <jakub@redhat.com>
33762         PR bootstrap/80531
33763         * cgraph.h (symtab_node::debug_symtab): No longer inline.
33764         * symtab.c (symtab_node::debug_symtab): Move definition here.
33766 2017-04-28  Richard Biener  <rguenther@suse.de>
33768         * lto-streamer.h (LTO_major_version): Bump to 7.
33770 2017-04-28  Richard Biener  <rguenther@suse.de>
33772         * tree-vrp.c (assert_info): New struct.
33773         (add_assert_info): New helper.
33774         (register_edge_assert_for_2): Refactor to add asserts to a vector
33775         of assert_info.
33776         (register_edge_assert_for_1): Likewise.
33777         (register_edge_assert_for): Likewise.
33778         (finish_register_edge_assert_for): New helper actually registering
33779         asserts where live on edge.
33780         (find_conditional_asserts): Adjust.
33781         (find_switch_asserts): Likewise.
33782         (evrp_dom_walker::try_find_new_range): Generalize.
33783         (evrp_dom_walker::before_dom_children): Use register_edge_assert_for.
33785 2017-04-27  Marek Polacek  <polacek@redhat.com>
33787         PR sanitizer/80349
33788         * fold-const.c (fold_binary_loc) <case EQ_EXPR, NE_EXPR>: Convert
33789         arg10 and arg11 to itype.
33791 2017-04-27  Jonathan Wakely  <jwakely@redhat.com>
33793         * doc/extend.texi (Object Size Checking): Improve grammar.
33795 2017-04-27  Richard Earnshaw  <rearnsha@arm.com>
33797         PR target/80530
33798         * config/aarch64/aarch64.c (aarch64_emit_approx_sqrt): Ensure
33799         that the logic for permitting reciprocal estimates matches that
33800         in use_rsqrt_p.
33802 2017-04-27  Jakub Jelinek  <jakub@redhat.com>
33804         PR c++/80534
33805         * tree.c (type_cache_hasher::equal): Only compare
33806         TYPE_TYPELESS_STORAGE flag on non-aggregate element types.
33807         (build_array_type_1): Only hash TYPE_TYPELESS_STORAGE flag on
33808         non-aggregate element types.
33809         * tree.h (TYPE_TYPELESS_STORAGE): Fix comment typo, add more details
33810         about the flag on ARRAY_TYPEs in the comment, formatting fix.
33812 2017-04-27  Richard Biener  <rguenther@suse.de>
33814         PR middle-end/80533
33815         * emit-rtl.c (set_mem_attributes_minus_bitpos): When
33816         stripping ARRAY_REFs from MEM_EXPR make sure we're not
33817         keeping a reference to a trailing array.
33819 2017-04-27  Richard Biener  <rguenther@suse.de>
33821         PR middle-end/80539
33822         * tree-chrec.c (chrec_fold_plus_poly_poly): Deal with not
33823         being in loop-closed SSA form conservatively.
33824         (chrec_fold_multiply_poly_poly): Likewise.
33826 2017-04-27  Tamar Christina  <tamar.christina@arm.com>
33828         PR middle-end/79665
33829         * expr.c (expand_expr_real_2): Move TRUNC_MOD_EXPR, FLOOR_MOD_EXPR,
33830         CEIL_MOD_EXPR, ROUND_MOD_EXPR cases.
33832 2017-04-27  Jakub Jelinek  <jakub@redhat.com>
33834         PR target/77728
33835         * config/aarch64/aarch64.c (struct aarch64_fn_arg_alignment): Remove.
33836         (aarch64_function_arg_alignment): Return unsigned int again, but still
33837         ignore TYPE_FIELDS chain decls other than FIELD_DECLs.
33838         (aarch64_layout_arg): Adjust aarch64_function_arg_alignment caller.
33839         Don't emit -Wpsabi note.
33840         (aarch64_function_arg_boundary): Likewise.
33841         (aarch64_gimplify_va_arg_expr): Adjust aarch64_function_arg_alignment
33842         caller.
33844 2017-04-26  Nathan Sidwell  <nathan@acm.org>
33846         * tree.h (crc32_unsigned_n): Declare.
33847         (crc32_unsigned, crc32_unsigned): Make inline.
33848         * tree.c (crc32_unsigned_bits): Replace with ...
33849         (crc32_unsigned_n): ... this.
33850         (crc32_unsigned, crc32_byte): Remove.
33851         (crc32_string): Remove unnecessary braces.
33853 2017-04-25  Jan Hubicka  <hubicka@ucw.cz>
33855         * ipa-cp.c (estimate_local_effects): Convert sreal to int.
33856         * ipa-inline-analysis.c (MAX_TIME): Remove.
33857         (account_size_time): Use sreal for time.
33858         (dump_inline_summary): Update.
33859         (estimate_function_body_sizes): Update.
33860         (estimate_edge_size_and_time): Update.
33861         (estimate_calls_size_and_time): Update.
33862         (estimate_node_size_and_time): Update.
33863         (inline_merge_summary): Update.
33864         (inline_update_overall_summary): Update.
33865         (estimate_time_after_inlining): Update.
33866         (inline_read_section): Update.
33867         (inline_write_summary): Update.
33868         * ipa-inline.c (compute_uninlined_call_time): Update.
33869         (compute_inlined_call_time): Update.
33870         (recursive_inlining): Update.
33871         (inline_small_functions): Update.
33872         (dump_overall_stats): Update.
33873         * ipa-inline.h: Include sreal.h.
33874         (size_time_entry): Turn time to sreal.
33875         (inline_summary): Turn self_time nad time to sreal.
33877 2017-04-25  Jan Hubicka  <hubicka@ucw.cz>
33879         * sreal.c: Include backend.h, tree.h, gimple.h, cgraph.h and
33880         data-streamer.h
33881         (sreal::stream_out, sreal::stream_in): New.
33882         * sreal.h (sreal::stream_out, sreal::stream_in): Declare.
33884 2017-04-25  Jakub Jelinek  <jakub@redhat.com>
33886         * Makefile.in (s-options): Invoke opt-gather.awk with LC_ALL=C in the
33887         environment.
33889 2017-04-25  Uros Bizjak  <ubizjak@gmail.com>
33891         PR target/70799
33892         * config/i386/i386.c (dimode_scalar_to_vector_candidate_p):
33893         Handle ASHIFTRT.
33894         (dimode_scalar_chain::compute_convert_gain): Ditto.
33895         (dimode_scalar_chain::make_vector_copies): Ditto.
33896         (dimode_scalar_chain::convert_reg): Ditto.
33897         (dimode_scalar_chain::convert_insn): Ditto.
33898         * config/i386/sse.md (VI24_AVX512BW_1): Remove mode iterator.
33899         (VI248_AVX512BW_1): New mode iterator.
33900         (<mask_codefor>ashr<mode>3<mask_name>): Merge insn pattern with
33901         <mask_codefor>ashrv2di3<mask_name> insn using VI248_AVX512BW_1
33902         mode iterator.
33904 2017-04-25  Martin Sebor  <msebor@redhat.com>
33906         PR tree-optimization/80497
33907         * gimple-ssa-sprintf.c (get_int_range): Avoid assuming all integer
33908         constants are representable in HOST_WIDE_INT.
33909         (parse_directive): Ditto.
33911 2017-04-25  Martin Sebor  <msebor@redhat.com>
33913         PR bootstrap/80486
33914         * dominance.c (dom_info::m_n_basic_blocks): Change type to unsigned.
33915         (new_zero_array): Adjust signature.
33916         (dom_info::dom_init): Used unsigned rather that size_t.
33917         (dom_info::dom_info): Same.
33919 2017-04-25  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
33920             Jakub Jelinek  <jakub@redhat.com>
33922         PR target/77728
33923         * config/arm/arm.c: Include gimple.h.
33924         (aapcs_layout_arg): Emit -Wpsabi note if arm_needs_doubleword_align
33925         returns negative, increment ncrn only if it returned positive.
33926         (arm_needs_doubleword_align): Return int instead of bool,
33927         ignore DECL_ALIGN of non-FIELD_DECL TYPE_FIELDS chain
33928         members, but if there is any such non-FIELD_DECL
33929         > PARM_BOUNDARY aligned decl, return -1 instead of false.
33930         (arm_function_arg): Emit -Wpsabi note if arm_needs_doubleword_align
33931         returns negative, increment nregs only if it returned positive.
33932         (arm_setup_incoming_varargs): Likewise.
33933         (arm_function_arg_boundary): Emit -Wpsabi note if
33934         arm_needs_doubleword_align returns negative, return
33935         DOUBLEWORD_ALIGNMENT only if it returned positive.
33937 2017-04-25  Marek Polacek  <polacek@redhat.com>
33939         PR sanitizer/80349
33940         * fold-const.c (fold_binary_loc) <case BIT_IOR_EXPR>: Convert arg0's
33941         first argument to type.
33943 2017-04-25  Bill Seurer  <seurer@linux.vnet.ibm.com>
33945         PR target/80482
33946         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Change
33947         type checks to test for compatibility instead of equality.
33949 2017-04-25  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
33950             Jakub Jelinek  <jakub@redhat.com>
33952         PR target/77728
33953         * config/aarch64/aarch64.c (struct aarch64_fn_arg_alignment): New
33954         type.
33955         (aarch64_function_arg_alignment): Return aarch64_fn_arg_alignment
33956         struct.  Ignore DECL_ALIGN of decls other than FIELD_DECL for
33957         the alignment computation, but return their maximum in warn_alignment.
33958         (aarch64_layout_arg): Adjust aarch64_function_arg_alignment caller.
33959         Emit a -Wpsabi note if warn_alignment is 16 bytes, but alignment
33960         is smaller.
33961         (aarch64_function_arg_boundary): Likewise.  Simplify using MIN/MAX.
33962         (aarch64_gimplify_va_arg_expr): Adjust aarch64_function_arg_alignment
33963         caller.
33965 2017-04-25  Claudiu Zissulescu  <claziss@synopsys.com>
33967         * config/arc/simdext.md (dmpyh): Fix typo.
33969 2017-04-25  Richard Biener  <rguenther@suse.de>
33971         PR tree-optimization/80492
33972         * alias.c (compare_base_decls): Handle registers with asm
33973         specification conservatively.
33974         * tree-ssa-alias.c (decl_refs_may_alias_p): Handle
33975         compare_base_decls returning dont-know properly.
33977 2017-04-25  Claudiu Zissulescu  <claziss@synopsys.com>
33979         * config/arc/arc.c (LEGITIMATE_OFFSET_ADDRESS_P): Delete macro.
33980         (legitimate_offset_address_p): New function.
33981         (arc_legitimate_address_p): Use above function.
33983 2017-04-25  Claudiu Zissulescu  <claziss@synopsys.com>
33985         * config/arc/arc.c (arc_output_mi_thunk): Emit PIC calls.
33987 2017-04-25  Claudiu Zissulescu  <claziss@synopsys.com>
33989         * config/arc/arc.c (arc_conditional_register_usage): Use ACCL,
33990         ACCH registers whenever they are available.
33992 2017-04-25  Claudiu Zissulescu  <claziss@synopsys.com>
33994         * config/arc/arc.c (arc_conditional_register_usage): Make D0, D1
33995         double regs fix when not used.
33997 2017-04-25  Claudiu Zissulescu  <claziss@synopsys.com>
33999         * config/arc/arc.h (REGNO_OK_FOR_BASE_P): Consider also extension
34000         core registers.
34001         (REG_OK_FOR_INDEX_P_NONSTRICT): Likewise.
34002         (REG_OK_FOR_BASE_P_NONSTRICT): Likewise.
34004 2017-04-25  Claudiu Zissulescu  <claziss@synopsys.com>
34006         * config/arc/arc.c (arc_output_addsi): Check for h-register class
34007         when emitting short ADD instructions.
34009 2017-04-25  Claudiu Zissulescu  <claziss@synopsys.com>
34011         * config/arc/arc.md (cmpsi_cc_insn_mixed): Use 'h' register
34012         constraint.
34013         (cmpsi_cc_c_insn): Likewise.
34014         (cbranchsi4_scratch): Compute proper instruction length using
34015         compact_hreg_operand.
34016         * config/arc/predicates.md (compact_hreg_operand): New predicate.
34018 2017-04-25  Richard Biener  <rguenther@suse.de>
34020         PR middle-end/80509
34021         * passes.c (pass_manager::pass_manager): Initialize
34022         m_name_to_pass_map.
34024 2017-04-25  Richard Biener  <rguenther@suse.de>
34026         PR tree-optimization/79201
34027         * tree-ssa-sink.c (statement_sink_location): Handle calls.
34029 2017-04-25  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
34031         PR target/80464
34032         * config/s390/vector.md: Split MEM->GPR vector moves for
34033         non-s_operand addresses.
34035 2017-04-25  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
34037         PR target/79895
34038         * config/s390/predicates.md (reload_const_wide_int_operand): New
34039         predicate.
34040         * config/s390/s390.md ("movti"): Remove d/P alternative.
34041         ("movti_bigconst"): New pattern definition.
34043 2017-04-25  Dominik Vogt  <vogt@linux.vnet.ibm.com>
34045         PR target/80080
34046         * s390-protos.h (s390_expand_cs_hqi): Removed.
34047         (s390_expand_cs, s390_expand_atomic_exchange_tdsi): New prototypes.
34048         * config/s390/s390.c (s390_emit_compare_and_swap): Handle all integer
34049         modes as well as CCZ1mode and CCZmode.
34050         (s390_expand_atomic_exchange_tdsi, s390_expand_atomic): Adapt to new
34051         signature of s390_emit_compare_and_swap.
34052         (s390_expand_cs_hqi): Likewise, make static.
34053         (s390_expand_cs_tdsi): Generate an explicit compare before trying
34054         compare-and-swap, in some cases.
34055         (s390_expand_cs): Wrapper function.
34056         (s390_expand_atomic_exchange_tdsi): New backend specific expander for
34057         atomic_exchange.
34058         (s390_match_ccmode_set): Allow CCZmode <-> CCZ1 mode.
34059         * config/s390/s390.md ("atomic_compare_and_swap<mode>"): Merge the
34060         patterns for small and large integers.  Forbid symref memory operands.
34061         Move expander to s390.c.  Require cc register.
34062         ("atomic_compare_and_swap<DGPR:mode><CCZZ1:mode>_internal")
34063         ("*atomic_compare_and_swap<TDI:mode><CCZZ1:mode>_1")
34064         ("*atomic_compare_and_swapdi<CCZZ1:mode>_2")
34065         ("*atomic_compare_and_swapsi<CCZZ1:mode>_3"): Use s_operand to forbid
34066         symref memory operands.  Remove CC mode and call s390_match_ccmode
34067         instead.
34068         ("atomic_exchange<mode>"): Allow and implement all integer modes.
34070 2017-04-25  Dominik Vogt  <vogt@linux.vnet.ibm.com>
34072         * config/s390/s390.md (define_peephole2): New peephole to help
34073         combining the load-and-test pattern with volatile memory.
34075 2017-04-25  Dominik Vogt  <vogt@linux.vnet.ibm.com>
34077         * config/s390/s390.md ("cstorecc4"): Use load-on-condition and deal
34078         with CCZmode for TARGET_Z196.
34080 2017-04-25  Jakub Jelinek  <jakub@redhat.com>
34082         PR rtl-optimization/80501
34083         * combine.c (make_compound_operation_int): Set subreg_code to SET
34084         even for AND with mask of the sign bit of mode.
34086         PR rtl-optimization/80500
34087         * loop-unroll.c (combine_var_copies_in_loop_exit): Call copy_rtx on
34088         sum's initial value.
34090 2017-04-25  Julian Brown  <julian@codesourcery.com>
34091             Naveen H.S  <Naveen.Hurugalawadi@cavium.com>
34093         * config/aarch64/thunderx2t99.md (thunderx2t99_crc): New Reservation.
34095 2017-04-25  Marc Glisse  <marc.glisse@inria.fr>
34097         * fold-const.c (tree_single_nonzero_warnv_p): Handle SSA_NAME.
34099 2017-04-25  Julian Brown  <julian@codesourcery.com>
34100             Naveen H.S  <Naveen.Hurugalawadi@cavium.com>
34102         * config/aarch64/thunderx2t99.md (thunderx2t99_aes): New Reservation.
34103         (thunderx2t99_sha): New Reservation.
34105 2017-04-25  Julian Brown  <julian@codesourcery.com>
34106             Naveen H.S  <Naveen.Hurugalawadi@cavium.com>
34108         * config/aarch64/aarch64-simd.md (aarch64_simd_vec_set<mode>): Fix
34109         type for 1-element load.
34111 2017-04-24  Marc Glisse  <marc.glisse@inria.fr>
34113         * match.pd (X/[ex]C CMP Y/[ex]C): New transformation.
34115 2017-04-24  Martin Jambor  <mjambor@suse.cz>
34117         PR tree-optimization/80293
34118         * tree-sra.c (scalarizable_type_p): New parameter const_decl, make
34119         char arrays not totally scalarizable if it is false.
34120         (analyze_all_variable_accesses): Pass correct value in the new
34121         parameter.  Add a statistics counter.
34123 2017-04-24  Jan Hubicka  <hubicka@ucw.cz>
34125         PR middle-end/79931
34126         * ipa-devirt.c (dump_possible_polymorphic_call_targets): Fix ICE.
34128 2017-04-24  Richard Biener  <rguenther@suse.de>
34130         PR tree-optimization/80494
34131         * tree-scalar-evolution.c (analyze_scalar_evolution_1): Bail
34132         out for complex types.
34134 2017-04-24  Richard Biener  <rguenther@suse.de>
34136         * tree-ssa-sccvn.h (run_scc_vn): Adjust prototype.
34137         * tree-ssa-sccvn.c (print_scc): Print SCC size.
34138         (extract_and_process_scc_for_name): Never fail but drop SCC to varying.
34139         (DFS): Adjust and never fail.
34140         (sccvn_dom_walker::fail): Remove.
34141         (sccvn_dom_walker::before_dom_children): Adjust.
34142         (run_scc_vn): Likewise and never fail.
34143         * tree-ssa-pre.c (pass_pre::execute): Adjust.
34144         (pass_fre::execute): Likewise.
34146 2017-04-24  Richard Biener  <rguenther@suse.de>
34148         PR tree-optimization/79725
34149         * tree-ssa-sink.c (statement_sink_location): Return whether
34150         failure reason was zero uses.  Move that check later.
34151         (sink_code_in_bb): Deal with zero uses by removing the stmt
34152         if possible.
34154 2017-04-24  Richard Biener  <rguenther@suse.de>
34156         PR c++/2972
34157         * tree-ssa-uninit.c (warn_uninitialized_vars): Handle some
34158         pointer-based references.
34160 2017-04-24  Richard Biener  <rguenther@suse.de>
34162         PR bootstrap/79814
34163         * pass_manager.h (pass_manager::operator new): Remove.
34164         (pass_manager::operator delete): Likewise.
34165         * passes.c (pass_manager::operator new): Remove.
34166         (pass_manager::operator delete): Likewise.
34167         (pass_manager::pass_manager): Zero individual pass members.
34169 2017-04-23  Uros Bizjak  <ubizjak@gmail.com>
34171         PR target/70799
34172         * config/i386/i386.c (dimode_scalar_to_vector_candidate_p)
34173         <case ASHIFT, case LSHIFTRT>: Also consider variable shifts.
34174         Check "XEXP (src, 1)" operand here.
34175         <case PLUS, case MINUS, case IOR, case XOR, case AND>:
34176         Check "XEXP (src, 1)" operand here.
34177         (dimode_scalar_chain::make_vector_copies): Detect count register
34178         of a shift instruction.  Zero extend count register from QImode
34179         to DImode to satisfy vector shift pattern count operand predicate.
34180         Substitute vector shift count operand with a DImode copy.
34181         (dimode_scalar_chain::convert_reg): Ditto, zero-extend from
34182         vector register.
34184 2017-04-21  Uros Bizjak  <ubizjak@gmail.com>
34186         * config/i386/i386.md (*extzvqi_mem_rex64): Move above *extzv<mode>.
34187         Remove UNSPEC_NOREX_MEM tag.  Update corresponding peephole2 pattern.
34188         (*insvqi_1_mem_rex64): Move above insv<mode>_1.  Remove
34189         UNSPEC_NOREX_MEM tag.  Update corresponding peephole2 pattern.
34190         (UNSPEC_NOREX_MEM): Remove definition.
34192 2017-04-21  Richard Biener  <rguenther@suse.de>
34194         PR tree-optimization/79547
34195         * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
34196         Handle strlen, strcmp, strncmp, strcasecmp, strncasecmp, memcmp,
34197         bcmp, strspn, strcspn, __builtin_object_size and __builtin_constant_p
34198         without any constraints.
34200 2017-04-21  Richard Biener  <rguenther@suse.de>
34202         PR tree-optimization/78847
34203         * fold-const.c (split_tree): Handle POINTER_PLUS_EXPR.
34205 2017-04-21  Richard Biener  <rguenther@suse.de>
34207         * tree.h (build_qualified_type): Annotate with CXX_MEM_STAT_INFO.
34208         (build_distinct_type_copy): Likewise.
34209         (build_variant_type_copy): Likewise.
34210         * tree.c (build_qualified_type): Pass down mem-stat info.
34211         (build_distinct_type_copy): Likewise.
34212         (build_variant_type_copy): Likewise.
34214 2017-04-21  Richard Biener  <rguenther@suse.de>
34216         PR tree-optimization/80237
34217         * tree-ssa-pre.c (find_leader_in_sets): Add third set argument,
34218         defaulted to NULL.
34219         (phi_translate_1): Also allow a leader in AVAIL_OUT of pred
34220         for a simplified result.
34222 2016-04-21  Richard Biener  <rguenther@suse.de>
34224         * tree-ssa-loop-ivcanon.c (constant_after_peeling): Do not require
34225         sth as strict as a simple_iv but a chrec without symbols and an
34226         operand defined in the loop we are peeling (and not some subloop).
34227         (propagate_constants_for_unrolling): Propagate all constants.
34229 2017-04-20  Uros Bizjak  <ubizjak@gmail.com>
34231         PR target/79804
34232         * config/i386/i386.c (print_reg): Remove assert for disalowed
34233         regno values, call output_operand_lossage instead.
34235 2017-04-20  Uros Bizjak  <ubizjak@gmail.com>
34237         PR target/78090
34238         * config/i386/constraints.md (Yc): New register constraint.
34239         * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
34240         Use Yc constraint for alternative 2 of operand 0.  Remove
34241         preferred_for_speed attribute.
34243 2017-04-20  Alexander Monakov  <amonakov@ispras.ru>
34245         * omp-low.c (lower_lastprivate_clauses): Correct handling of linear and
34246         lastprivate clauses in SIMT case.
34248 2017-04-20  Volker Reichelt  <v.reichelt@netcologne.de>
34250         * doc/invoke.texi (-Wextra-semi): Document new warning option.
34252 2017-04-20  Richard Biener  <rguenther@suse.de>
34254         PR tree-optimization/57796
34255         * tree-vect-stmts.c (vect_model_store_cost): Cost scatters
34256         as N scalar stores.
34257         (vect_model_load_cost): Cost gathers as N scalar loads.
34259 2017-04-20  Richard Biener  <rguenther@suse.de>
34261         * ggc-page.c (ggc_allocated_p): Rename to ...
34262         (safe_lookup_page_table_entry): ... this and return the lookup
34263         result.
34264         (gt_ggc_m_S): Use safe_lookup_page_table_entry.
34266 2017-04-20  Richard Biener  <rguenther@suse.de>
34268         PR tree-optimization/80453
34269         * tree-ssa-sccvn.h (struct vn_phi_s): Add cclhs and ccrhs members.
34270         * tree-ssa-sccvn.c (cond_stmts_equal_p): Use recorded lhs and rhs
34271         from the conditions.
34272         (vn_phi_eq): Pass them down.
34273         (vn_phi_lookup): Record them.
34274         (vn_phi_insert): Likewise.
34276 2017-04-20  Matthew Fortune  <matthew.fortune@imgtec.com>
34278         * config/mips/mips.c (mips_expand_vec_perm_const): Re-fix
34279         uninitialized variable warning to avoid buffer overrun.
34281 2017-04-20  Alexander Monakov  <amonakov@ispras.ru>
34283         PR other/71250
34284         * doc/invoke.texi (-Wmissing-field-initializers): Mention that warning
34285         is suppressed for '{ 0 }' in C.
34287 2017-04-20  Jakub Jelinek  <jakub@redhat.com>
34289         * BASE-VER: Set to 8.0.0.
34291 2017-04-20  Thomas Preud'homme  <thomas.preudhomme@arm.com>
34293         * config/arm/arm.c (arm_elf_asm_cdtor): Create non-default
34294         priority .init_array and .fini_array section with SECTION_NOTYPE
34295         flag.
34297 2017-04-20  Jakub Jelinek  <jakub@redhat.com>
34299         PR middle-end/80423
34300         * tree.h (build_array_type): Add typeless_storage default argument.
34301         * tree.c (type_cache_hasher::equal): Also compare
34302         TYPE_TYPELESS_STORAGE flag for ARRAY_TYPEs.
34303         (build_array_type): Add typeless_storage argument, set
34304         TYPE_TYPELESS_STORAGE to it, if shared also hash it, and pass to
34305         recursive call.
34306         (build_nonshared_array_type): Adjust build_array_type_1 caller.
34307         (build_array_type): Likewise.  Add typeless_storage argument.
34309 2017-04-19  Eric Botcazou  <ebotcazou@adacore.com>
34310             Jakub Jelinek  <jakub@redhat.com>
34312         PR tree-optimization/80426
34313         * tree-vrp.c (extract_range_from_binary_expr_1): For an additive
34314         operation on symbolic operands, also compute the overflow for the
34315         invariant part when the operation degenerates into a negation.
34317 2017-04-19  Jakub Jelinek  <jakub@redhat.com>
34319         PR debug/80461
34320         * dwarf2out.c (modified_type_die, gen_type_die_with_usage):
34321         Check for t with zero TYPE_QUALS_NO_ADDR_SPACE.
34323         PR debug/80436
34324         * tree-ssa-loop-manip.c (find_uses_to_rename_def): Ignore debug uses.
34326 2017-04-19  Georg-Johann Lay  <avr@gjlay.de>
34328         PR target/80462
34329         * config/avr/avr.c (tree.h): Include it.
34330         (cgraph.h): Include it.
34331         (avr_encode_section_info): Don't warn for uninitialized progmem
34332         variable if it's just an alias.
34334 2017-04-19  Richard Biener  <rguenther@suse.de>
34336         PR ipa/65972
34337         * auto-profile.c (afdo_vpt_for_early_inline): Update SSA
34338         when needed by AutoPGO.
34340 2017-04-19  Paulo J. Matos  <paulo@matos-sorge.com>
34342         PR lto/50345
34343         * doc/lto.texi: Remove an extra 'that'.
34345 2017-04-19  Segher Boessenkool  <segher@kernel.crashing.org>
34347         PR rtl-optimization/80429
34348         * ira.c (split_live_ranges_for_shrink_wrap): Don't split regs that
34349         are only used in debug insns.
34351 2017-04-19  Eric Botcazou  <ebotcazou@adacore.com>
34352             Vladimir Makarov  <vmakarov@redhat.com>
34354         * config/sparc/predicates.md (input_operand): Add comment.  Return
34355         true for any memory operand when LRA is in progress.
34356         * config/sparc/sparc.c (sparc_expand_move): Minor formatting fix.
34358 2017-04-18  Jeff Law  <law@redhat.com>
34360         PR target/74563
34361         * mips.md ({return,simple_return}_internal): Do not overwrite
34362         operands[0].
34364 2017-04-18  Jakub Jelinek  <jakub@redhat.com>
34366         PR tree-optimization/80443
34367         * tree-vrp.c (intersect_ranges): For signed 1-bit precision type,
34368         instead of adding 1, subtract -1 and similarly instead of subtracting
34369         1 add -1.
34371 2017-04-18  Richard Sandiford  <richard.sandiford@arm.com>
34373         PR rtl-optimization/80357
34374         * haifa-sched.c (tmp_bitmap): New variable.
34375         (model_recompute): Handle duplicate use records.
34376         (alloc_global_sched_pressure_data): Initialize tmp_bitmap.
34377         (free_global_sched_pressure_data): Free it.
34379 2017-04-18  Bernd Edlinger  <bernd.edlinger@hotmail.de>
34381         Revert:
34382         2017-02-20  Bernd Edlinger  <bernd.edlinger@hotmail.de>
34383         * Makefile.in (BUILD_SYSTEM_HEADER_DIR): New make variabe.
34384         (LIMITS_H_TEST, if_multiarch, stmp-fixinc): Use BUILD_SYSTEM_HEADER_DIR
34385         instead of SYSTEM_HEADER_DIR.
34387 2017-04-18  Jeff Law  <law@redhat.com>
34389         PR middle-end/80422
34390         * cfgcleanup.c (try_crossjump_to_edge): Verify SRC1 and SRC2 have
34391         predecessors after walking up the insn chain.
34393 2017-04-18  Jakub Jelinek  <jakub@redhat.com>
34395         PR debug/80263
34396         * dwarf2out.c (modified_type_die): Try harder not to emit internal
34397         sizetype type into debug info.
34399 2017-04-18  Michael Meissner  <meissner@linux.vnet.ibm.com>
34401         PR target/80099
34402         * config/rs6000/rs6000.c (rs6000_expand_vector_extract): Eliminate
34403         unneeded test for TARGET_UPPER_REGS_SF.
34404         * config/rs6000/vsx.md (vsx_extract_v4sf_var): Likewise.
34406 2017-04-18  Jakub Jelinek  <jakub@redhat.com>
34408         PR sanitizer/80444
34409         * sancov.c (sancov_pass): Use gsi_start_nondebug_after_labels_bb
34410         instead of gsi_after_labels.
34412 2017-04-18  Jeff Law  <law@redhat.com>
34414         * regcprop.c (maybe_mode_change): Avoid creating copies of the
34415         stack pointer.
34417         Revert:
34418         2017-04-13  Jeff Law  <law@redhat.com>
34419         * config/mips.mips.md (zero_extendsidi2): Do not allow SP to appear
34420         in operands[1] if it is a MEM and TARGET_MIPS16 is active.
34422 2017-04-18  Georg-Johann Lay  <avr@gjlay.de>
34424         PR target/79453
34425         * config/avr/avr.c (intl.h): Include it.
34426         (avr_pgm_check_var_decl) [reason]: Wrap diagnostic snippets into _().
34428 2017-04-18  Martin Liska  <mliska@suse.cz>
34430         PR gcov-profile/78783
34431         * gcov-tool.c (gcov_output_files): Validate that destination
34432         file is either removed by the tool or by a user.
34434 2017-04-14  Andrew Burgess  <andrew.burgess@embecosm.com>
34435             Guy Benyei  <guybe@mellanox.com>
34437         * config/arc/arc.c (arc_reorg): Move loop_end_id into a more local
34438         block, and do not negate it, the stored id is already negative.
34440 2017-04-14  Andrew Burgess  <andrew.burgess@embecosm.com>
34442         * config/arc/arc.md (doloop_begin_i): Use @pcl assembler syntax.
34444 2017-04-14  Michael Meissner  <meissner@linux.vnet.ibm.com>
34446         PR target/80098
34447         * config/rs6000/rs6000-cpus.def (OTHER_P9_VECTOR_MASKS): Define
34448         masks of options that should be turned off if the VSX vector
34449         options are turned off.
34450         (OTHER_P8_VECTOR_MASKS): Likewise.
34451         (OTHER_VSX_VECTOR_MASKS): Likewise.
34452         * config/rs6000/rs6000.c (rs6000_option_override_internal): Call
34453         rs6000_disable_incompatible_switches to validate no type switches
34454         like -mvsx.
34455         (rs6000_incompatible_switch): New function to disallow turning on
34456         other vector options if -mno-vsx, -mno-power8-vector, or
34457         -mno-power9-vector are specified.
34459 2017-04-14  Claudiu Zissulescu  <claziss@synopsys.com>
34461         * config/arc/arc.h (CRT_CALL_STATIC_FUNCTION): Use long calls.
34463 2017-04-14  Claudiu Zissulescu  <claziss@synopsys.com>
34465         * config/arc/arc-protos.h (arc_decl_pretend_args): Remove.
34466         * config/arc/arc.c (arc_decl_pretend_args): Likewise.
34467         * config/arc/arc.h (CFA_FRAME_BASE_OFFSET): Likewise.
34468         (ARG_POINTER_CFA_OFFSET): Likewise.
34470 2017-04-14  Claudiu Zissulescu  <claziss@synopsys.com>
34472         * config/arc/arc.c (arc_mode_dependent_address_p): Relax
34473         conditions to take advantage of various optimizations.
34475 2017-04-13  Jeff Law  <law@redhat.com>
34477         * config/mips.mips.md (zero_extendsidi2): Do not allow SP to appear
34478         in operands[1] if it is a MEM and TARGET_MIPS16 is active.
34479         (zero_extendsidi2_dext): Likewise.
34481 2017-04-13  Jakub Jelinek  <jakub@redhat.com>
34483         PR sanitizer/80403
34484         * fold-const.c (fold_ternary_loc): Revert
34485         use op0 instead of fold_convert_loc (loc, type, arg0) part of
34486         2017-04-12 change.
34488 2017-04-13  Vladimir Makarov  <vmakarov@redhat.com>
34490         PR rtl-optimization/80343
34491         * lra-remat.c (update_scratch_ops): Assign original hard reg to
34492         new scratch pseudo.
34494 2017-04-13  Denis Khalikov <d.khalikov@partner.samsung.com>
34496         PR sanitizer/80414
34497         * ubsan.c (ubsan_expand_bounds_ifn): Pass original index
34498         to ubsan_encode_value.
34500 2017-04-13  Jeff Law  <law@redhat.com>
34502         * reload1.c (eliminate_regs_1): Call gen_rtx_raw_SUBREG for SUBREGs
34503         appearing in DEBUG_INSNs.
34505 2017-04-13  Martin Liska  <mliska@suse.cz>
34507         PR gcov-profile/80413
34508         * gcov-io.c (gcov_write_string): Copy to buffer just when
34509         allocated size is greater than zero.
34511 2017-04-13  Jakub Jelinek  <jakub@redhat.com>
34513         PR debug/80321
34514         * dwarf2out.c (decls_for_scope): Ignore declarations of
34515         current_function_decl in BLOCK_NONLOCALIZED_VARS.
34517 2017-04-12  Jan Hubicka  <hubicka@ucw.cz>
34519         PR lto/69953
34520         * ipa-visibility.c (non_local_p): Fix typos.
34521         (localize_node): When localizing symbol in same comdat group,
34522         dissolve the group only when we know external symbols are going
34523         to be privatized.
34524         (function_and_variable_visibility): Do not localize DECL_EXTERNAL.
34526 2017-04-12  Jakub Jelinek  <jakub@redhat.com>
34528         PR tree-optimization/79390
34529         * optabs.c (emit_conditional_move): If the preferred op2/op3 operand
34530         order does not result in usable sequence, retry with reversed operand
34531         order.
34533         PR sanitizer/80403
34534         PR sanitizer/80404
34535         PR sanitizer/80405
34536         * fold-const.c (fold_ternary_loc): Use op1 instead of arg1 as argument
34537         to fold_build2_loc.  Convert TREE_OPERAND (tem, 0) to type.  Use
34538         op0 instead of fold_convert_loc (loc, type, arg0).
34540 2017-04-12  Jeff Law  <law@redhat.com>
34542         * genattrtab.c (write_eligible_delay): Verify DELAY_INSN still
34543         has a delay slot in the generated code.
34545         * config/cris/cris.md (cris_preferred_reload_class): Return
34546         GENNONACR_REGS rather than GENERAL_REGS.
34548 2017-04-12  Jakub Jelinek  <jakub@redhat.com>
34550         PR c/80163
34551         * expr.c <CASE_CONVERT>: For EXPAND_INITIALIZER determine SIGN_EXTEND
34552         vs. ZERO_EXTEND based on signedness of treeop0's type rather than
34553         signedness of the result type.
34555 2017-04-12  Richard Biener  <rguenther@suse.de>
34556             Jeff Law  <law@redhat.com>
34558         PR tree-optimization/80359
34559         * tree-ssa-dse.c (maybe_trim_partially_dead_store): Do not
34560         trim stores to TARGET_MEM_REFs.
34562 2017-04-12  Richard Biener  <rguenther@suse.de>
34564         PR tree-optimization/79390
34565         * gimple-ssa-split-paths.c (is_feasible_trace): Restrict
34566         threading case even more.
34568 2017-04-12  Segher Boessenkool  <segher@kernel.crashing.org>
34570         PR target/80382
34571         * config/rs6000/sync.md (atomic_load<mode>, atomic_store<mode): Test
34572         for quad_address_p for TImode, instead of just not indexed_address.
34574 2017-04-12  Richard Biener  <rguenther@suse.de>
34575             Bernd Edlinger  <bernd.edlinger@hotmail.de>
34577         PR middle-end/79671
34578         * alias.c (component_uses_parent_alias_set_from): Handle
34579         TYPE_TYPELESS_STORAGE.
34580         (get_alias_set): Likewise.
34581         * tree-core.h (tree_type_common): Add typeless_storage flag.
34582         * tree.h (TYPE_TYPELESS_STORAGE): New macro.
34583         * stor-layout.c (place_union_field): Set TYPE_TYPELESS_STORAGE
34584         for types containing members with TYPE_TYPELESS_STORAGE.
34585         (place_field): Likewise.
34586         (layout_type): Likewise for ARRAY_TYPE.
34587         * lto-streamer-out.c (hash_tree): Hash TYPE_TYPELESS_STORAGE.
34588         * tree-streamer-in.c (unpack_ts_type_common_value_fields): Stream
34589         TYPE_TYPELESS_STORAGE.
34590         * tree-streamer-out.c (pack_ts_type_common_value_fields): Likewise.
34592 2017-04-12  Jakub Jelinek  <jakub@redhat.com>
34594         PR sanitizer/80349
34595         * fold-const.c (fold_binary_loc) <case BIT_IOR_EXPR>: Convert arg0's
34596         first argument to type.
34598 2017-04-11  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
34600         PR target/80376
34601         PR target/80315
34602         * config/rs6000/rs6000.c (rs6000_expand_unop_builtin): Return
34603         CONST0_RTX (mode) rather than const0_rtx where appropriate.
34604         (rs6000_expand_binop_builtin): Likewise.
34605         (rs6000_expand_ternop_builtin): Likewise; also add missing
34606         vsx_xxpermdi_* variants; also fix typo (arg1 => arg2) for
34607         vshasigma built-ins.
34608         * doc/extend.texi: Document that vec_xxpermdi's third argument
34609         must be a constant.
34611 2017-04-11  Uros Bizjak  <ubizjak@gmail.com>
34613         * config/i386/i386.c (dimode_scalar_chain::compute_convert_gain):
34614         Use shift_const cost parameter when calculating gain of STV shifts.
34616 2017-04-11  Vladimir Makarov  <vmakarov@redhat.com>
34618         PR rtl-optimization/70478
34619         * lra-constraints.c (process_alt_operands): Check memory for
34620         disfavoring memory insn operand.
34622 2017-04-11  Jakub Jelinek  <jakub@redhat.com>
34624         PR middle-end/80100
34625         * simplify-rtx.c (simplify_binary_operation_1) <case IOR>: Perform
34626         left shift in unsigned HOST_WIDE_INT type.
34628         PR rtl-optimization/80385
34629         * simplify-rtx.c (simplify_unary_operation_1): Don't transform
34630         (not (neg X)) into (plus X -1) for complex or non-integral modes.
34632         PR libgomp/80394
34633         * omp-low.c (scan_omp_task): Don't optimize away empty tasks
34634         if they have any depend clauses.
34636 2017-04-11  Martin Liska  <mliska@suse.cz>
34638         PR ipa/80212
34639         * cgraph.c (cgraph_node::dump): Dump calls_comdat_local.
34640         * ipa-split.c (split_function): Create a local comdat symbol
34641         if caller is in a comdat group.
34643 2017-04-11  Martin Liska  <mliska@suse.cz>
34645         PR ipa/80212
34646         * ipa-cp.c (determine_versionability): Handle calls_comdat_local
34647         flags.
34649 2017-04-11  Martin Sebor  <msebor@redhat.com>
34651         PR middle-end/80364
34652         * gimple-ssa-sprintf.c (get_int_range): Remove second argument and
34653         always use the int type.  Use INTEGRAL_TYPE_P() rather than testing
34654         for INTEGER_TYPE.
34655         (directive::set_width, directive::set_precision, format_character):
34656         Adjust.
34657         (parse_directive): Use INTEGRAL_TYPE_P() rather than testing for
34658         INTEGER_TYPE.
34660 2017-04-11  Richard Earnshaw  <rearnsha@arm.com>
34662         PR target/80389
34663         * config/arm/arm.c (arm_configure_build_target): When -mcpu and -arch
34664         conflict, set target->arch_name instead of target->cpu_name.
34666 2017-04-11  Richard Biener  <rguenther@suse.de>
34668         PR tree-optimization/80374
34669         * tree-ssa-dom.c (derive_equivalences_from_bit_ior): Use
34670         build_zero_cst, remove fold_convertible_p check again.
34672 2017-04-11  Martin Liska  <mliska@suse.cz>
34674         PR sanitizer/70878
34675         * ubsan.c (instrument_object_size): Do not instrument register
34676         variables.
34678 2017-04-11  Jakub Jelinek  <jakub@redhat.com>
34680         PR target/80381
34681         * config/i386/i386-builtin-types.def
34682         (V16HI_FTYPE_V16HI_INT_V16HI_UHI_COUNT,
34683         V16HI_FTYPE_V16HI_V8HI_V16HI_UHI_COUNT,
34684         V16SI_FTYPE_V16SI_INT_V16SI_UHI_COUNT,
34685         V16SI_FTYPE_V16SI_V4SI_V16SI_UHI_COUNT,
34686         V2DI_FTYPE_V2DI_INT_V2DI_UQI_COUNT,
34687         V2DI_FTYPE_V2DI_V2DI_V2DI_UQI_COUNT,
34688         V32HI_FTYPE_V32HI_INT_V32HI_USI_COUNT,
34689         V32HI_FTYPE_V32HI_V8HI_V32HI_USI_COUNT,
34690         V4DI_FTYPE_V4DI_INT_V4DI_UQI_COUNT,
34691         V4DI_FTYPE_V4DI_V2DI_V4DI_UQI_COUNT,
34692         V4SI_FTYPE_V4SI_INT_V4SI_UQI_COUNT,
34693         V4SI_FTYPE_V4SI_V4SI_V4SI_UQI_COUNT,
34694         V8DI_FTYPE_V8DI_INT_V8DI_UQI_COUNT,
34695         V8DI_FTYPE_V8DI_V2DI_V8DI_UQI_COUNT,
34696         V8HI_FTYPE_V8HI_INT_V8HI_UQI_COUNT,
34697         V8HI_FTYPE_V8HI_V8HI_V8HI_UQI_COUNT,
34698         V8SI_FTYPE_V8SI_INT_V8SI_UQI_COUNT,
34699         V8SI_FTYPE_V8SI_V4SI_V8SI_UQI_COUNT): New function type aliases.
34700         * config/i386/i386-builtin.def (__builtin_ia32_pslld512_mask,
34701         __builtin_ia32_pslldi512_mask, __builtin_ia32_psllq512_mask,
34702         __builtin_ia32_psllqi512_mask, __builtin_ia32_psrad512_mask,
34703         __builtin_ia32_psradi512_mask, __builtin_ia32_psraq512_mask,
34704         __builtin_ia32_psraqi512_mask, __builtin_ia32_psrld512_mask,
34705         __builtin_ia32_psrldi512_mask, __builtin_ia32_psrlq512_mask,
34706         __builtin_ia32_psrlqi512_mask, __builtin_ia32_psllwi128_mask,
34707         __builtin_ia32_pslldi128_mask, __builtin_ia32_psllqi128_mask,
34708         __builtin_ia32_psllw128_mask, __builtin_ia32_pslld128_mask,
34709         __builtin_ia32_psllq128_mask, __builtin_ia32_psllwi256_mask,
34710         __builtin_ia32_psllw256_mask, __builtin_ia32_pslldi256_mask,
34711         __builtin_ia32_pslld256_mask, __builtin_ia32_psllqi256_mask,
34712         __builtin_ia32_psllq256_mask, __builtin_ia32_psradi128_mask,
34713         __builtin_ia32_psrad128_mask, __builtin_ia32_psradi256_mask,
34714         __builtin_ia32_psrad256_mask, __builtin_ia32_psraqi128_mask,
34715         __builtin_ia32_psraq128_mask, __builtin_ia32_psraqi256_mask,
34716         __builtin_ia32_psraq256_mask, __builtin_ia32_psrldi128_mask,
34717         __builtin_ia32_psrld128_mask, __builtin_ia32_psrldi256_mask,
34718         __builtin_ia32_psrld256_mask, __builtin_ia32_psrlqi128_mask,
34719         __builtin_ia32_psrlq128_mask, __builtin_ia32_psrlqi256_mask,
34720         __builtin_ia32_psrlq256_mask, __builtin_ia32_psrawi256_mask,
34721         __builtin_ia32_psraw256_mask, __builtin_ia32_psrawi128_mask,
34722         __builtin_ia32_psraw128_mask, __builtin_ia32_psrlwi256_mask,
34723         __builtin_ia32_psrlw256_mask, __builtin_ia32_psrlwi128_mask,
34724         __builtin_ia32_psrlw128_mask, __builtin_ia32_psllwi512_mask,
34725         __builtin_ia32_psllw512_mask, __builtin_ia32_psrawi512_mask,
34726         __builtin_ia32_psraw512_mask, __builtin_ia32_psrlwi512_mask,
34727         __builtin_ia32_psrlw512_mask): Use _COUNT suffixed function type
34728         aliases.
34729         * config/i386/i386.c (ix86_expand_args_builtin): Rename last_arg_count
34730         flag to second_arg_count, handle 4 argument function type _COUNT
34731         aliases, handle second_arg_count on second argument rather than last.
34733 2017-04-10  Jeff Law  <law@redhat.com>
34735         PR tree-optimization/80374
34736         * tree-ssa-dom.c (derive_equivalences_from_bit_ior): Do not try to
34737         record anything if we can not convert integer_zero_node to the
34738         desired type.
34740 2017-04-10  Kelvin Nilsen  <kelvin@gcc.gnu.org>
34742         PR target/80108
34743         * config/rs6000/rs6000.c (rs6000_option_override_internal):
34744         Enhance special handling given to the TARGET_P9_MINMAX option in
34745         relation to certain other options.
34747 2017-04-10  Bin Cheng  <bin.cheng@arm.com>
34749         PR tree-optimization/80153
34750         * tree-ssa-loop-ivopts.c (add_iv_candidate_for_use): Check and
34751         remove POINTER_PLUS_EXPR's base part directly, rather than through
34752         aff_tree.
34754 2017-04-10  Richard Biener  <rguenther@suse.de>
34755             Bin Cheng  <bin.cheng@arm.com>
34757         PR tree-optimization/80153
34758         * tree-affine.c (aff_combination_to_tree): Get base pointer from
34759         the first element of pointer type aff_tree.  Build result expr in
34760         aff_tree's type.
34761         (add_elt_to_tree): Convert to type unconditionally.  Remove other
34762         fold_convert calls.
34763         * tree-ssa-loop-ivopts.c (alloc_iv): Pass in consistent types.
34764         (rewrite_use_nonlinear_expr): Check invariant using iv information.
34766 2017-04-10  Richard Biener  <rguenther@suse.de>
34768         * tree-ssa-structalias.c (find_func_aliases): Properly handle
34769         asm inputs.
34771 2017-04-10  Vladimir Makarov  <vmakarov@redhat.com>
34773         PR rtl-optimization/70478
34774         * lra-constraints.c (curr_small_class_check): New.
34775         (update_and_check_small_class_inputs): New.
34776         (process_alt_operands): Update curr_small_class_check.  Disfavor
34777         alternative insn memory operands.  Check available regs for small
34778         class operands.
34780 2017-03-31  Matthew Fortune  <matthew.fortune@imgtec.com>
34782         PR target/80057
34783         * config/mips/mips.opt (-mvirt): Update description.
34784         * doc/invoke.texi (-mvirt): Likewise.
34786 2017-04-10  Richard Biener  <rguenther@suse.de>
34788         PR middle-end/80362
34789         * fold-const.c (fold_binary_loc): Look at unstripped ops when
34790         looking for NEGATE_EXPR in -A / -B to A / B folding.
34792 2017-04-10  Martin Liska  <mliska@suse.cz>
34794         PR gcov-profile/80224
34795         * gcov.c (print_usage): Fix usage string.
34796         (get_gcov_intermediate_filename): Remove.
34797         (output_gcov_file): Use both for normal and intermediate format.
34798         (generate_results): Do not initialize special file for
34799         intermediate format.
34801 2017-04-10  Richard Biener  <rguenther@suse.de>
34803         PR tree-optimization/80304
34804         * tree-ssa-loop-im.c (ref_indep_loop_p_1): Also recurse
34805         for safelen.
34807 2017-04-10  Nathan Sidwell  <nathan@acm.org>
34809         PR target/79905
34810         * config/rs6000/rs6000.c (rs6000_vector_type): New.
34811         (rs6000_init_builtins): Use it.
34813 2016-04-10  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
34815         * config/arm/arm.md (<mrc>): Add mode to SET source.
34816         (<mrrc>): Likewise.
34818 2017-04-10  Richard Biener  <rguenther@suse.de>
34820         PR middle-end/80344
34821         * gimplify.c (is_gimple_mem_rhs_or_call): Allow CLOBBERs.
34823 2017-04-10  Jakub Jelinek  <jakub@redhat.com>
34825         PR target/80324
34826         * config/i386/avx512fintrin.h (_mm512_reduce_add_epi32,
34827         _mm512_reduce_mul_epi32, _mm512_reduce_and_epi32,
34828         _mm512_reduce_or_epi32, _mm512_mask_reduce_add_epi32,
34829         _mm512_mask_reduce_mul_epi32, _mm512_mask_reduce_and_epi32,
34830         _mm512_mask_reduce_or_epi32, _mm512_reduce_min_epi32,
34831         _mm512_reduce_max_epi32, _mm512_reduce_min_epu32,
34832         _mm512_reduce_max_epu32, _mm512_mask_reduce_min_epi32,
34833         _mm512_mask_reduce_max_epi32, _mm512_mask_reduce_min_epu32,
34834         _mm512_mask_reduce_max_epu32, _mm512_reduce_add_ps,
34835         _mm512_reduce_mul_ps, _mm512_mask_reduce_add_ps,
34836         _mm512_mask_reduce_mul_ps, _mm512_reduce_min_ps, _mm512_reduce_max_ps,
34837         _mm512_mask_reduce_min_ps, _mm512_mask_reduce_max_ps,
34838         _mm512_reduce_add_epi64, _mm512_reduce_mul_epi64,
34839         _mm512_reduce_and_epi64, _mm512_reduce_or_epi64,
34840         _mm512_mask_reduce_add_epi64, _mm512_mask_reduce_mul_epi64,
34841         _mm512_mask_reduce_and_epi64, _mm512_mask_reduce_or_epi64,
34842         _mm512_reduce_min_epi64, _mm512_reduce_max_epi64,
34843         _mm512_mask_reduce_min_epi64, _mm512_mask_reduce_max_epi64,
34844         _mm512_reduce_min_epu64, _mm512_reduce_max_epu64,
34845         _mm512_mask_reduce_min_epu64, _mm512_mask_reduce_max_epu64,
34846         _mm512_reduce_add_pd, _mm512_reduce_mul_pd, _mm512_mask_reduce_add_pd,
34847         _mm512_mask_reduce_mul_pd, _mm512_reduce_min_pd, _mm512_reduce_max_pd,
34848         _mm512_mask_reduce_min_pd, _mm512_mask_reduce_max_pd): New intrinsics.
34850 2017-04-08  Vladimir Makarov  <vmakarov@redhat.com>
34852         PR rtl-optimization/70478
34853         * lra-constraints.c: Reverse the last patch.
34855 2017-04-08  Andreas Tobler  <andreast@gcc.gnu.org>
34857         * config/aarch64/aarch64-freebsd.h: Define MCOUNT_NAME.
34858         Add comment for WCHAR_T.
34860 2017-04-08  Martin Liska  <mliska@suse.cz>
34862         Revert:
34863         2017-04-07  Martin Liska  <mliska@suse.cz>
34865         PR ipa/80212
34866         * ipa-split.c (split_function): Add function part to a same comdat
34867         group.
34869 2017-04-08  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
34871         PR target/80358
34872         * config/rs6000/rs6000.c (expand_block_compare): Fix boundary check.
34874 2017-04-07  Pat Haugen  <pthaugen@us.ibm.com>
34876         * rs6000/rs6000.c (vec_load_pendulum): Rename...
34877         (vec_pairing): ...to this.
34878         (power9_sched_reorder2): Rewrite code for pairing vector/vecload insns.
34879         (rs6000_sched_init): Adjust for name change.
34880         (struct rs6000_sched_context): Likewise.
34881         (rs6000_init_sched_context): Likewise.
34882         (rs6000_set_sched_context): Likewise.
34884 2017-04-07  Jakub Jelinek  <jakub@redhat.com>
34886         PR target/80322
34887         PR target/80323
34888         PR target/80325
34889         PR target/80326
34890         * config/i386/avxintrin.h (_mm256_cvtsd_f64, _mm256_cvtss_f32): New
34891         intrinsics.
34892         * config/i386/avx512fintrin.h (_mm512_int2mask, _mm512_mask2int,
34893         _mm512_abs_ps, _mm512_mask_abs_ps, _mm512_abs_pd, _mm512_mask_abs_pd,
34894         _mm512_cvtsd_f64, _mm512_cvtss_f32): Likewise.
34896 2017-04-07  Andreas Tobler  <andreast@gcc.gnu.org>
34898         * config/aarch64/aarch64-freebsd.h: Define WCHAR_TYPE.
34900 2017-04-07  Vladimir Makarov  <vmakarov@redhat.com>
34902         PR rtl-optimization/70703
34903         * ira-color.c (update_conflict_hard_regno_costs): Use
34904         int64_t instead of HOST_WIDE_INT.
34906 2017-04-07  Vladimir Makarov  <vmakarov@redhat.com>
34908         PR rtl-optimization/70478
34909         * lra-constraints.c (process_alt_operands): Disfavor alternative
34910         insn memory operands.
34912 2017-04-07  Jeff Law  <law@redhat.com>
34914         * config/iq2000/iq2000.c (final_prescan_insn): Do not separate a
34915         CALL and NOTE_INSN_CALL_ARG_LOCATION.
34917 2017-04-07  Martin Liska  <mliska@suse.cz>
34919         PR target/79889
34920         * config/aarch64/aarch64.c (aarch64_process_target_attr):
34921         Show error message instead of an ICE.
34923 2017-04-07  Martin Liska  <mliska@suse.cz>
34925         PR ipa/80212
34926         * ipa-split.c (split_function): Add function part to a same comdat
34927         group.
34929 2017-04-07  Richard Biener  <rguenther@suse.de>
34931         PR middle-end/80341
34932         * tree.c (get_unwidened): Also handle ! for_type case for
34933         INTEGER_CSTs.
34934         * convert.c (do_narrow): Split out from ...
34935         (convert_to_integer_1): ... here.  Do not pass final truncation
34936         type to get_unwidened for TRUNC_DIV_EXPR.
34938 2017-04-07  Richard Biener  <rguenther@suse.de>
34940         * tree-affine.c (wide_int_ext_for_comb): Take type rather
34941         than aff_tree.
34942         (aff_combination_const): Adjust.
34943         (aff_combination_scale): Likewise.
34944         (aff_combination_add_elt): Likewise.
34945         (aff_combination_add_cst): Likewise.
34946         (aff_combination_convert): Likewise.
34947         (add_elt_to_tree): Likewise.  Remove unused argument.
34948         (aff_combination_to_tree): Adjust calls to add_elt_to_tree.
34950 2017-04-07  Sebastian Huber  <sebastian.huber@embedded-brains.de>
34952         * config/arm/arm.h (ARM_DEFAULT_SHORT_ENUMS): Provide default
34953         definition.
34954         * config/arm/arm.c (arm_default_short_enums): Use
34955         ARM_DEFAULT_SHORT_ENUMS.
34956         * config/arm/rtems.h (ARM_DEFAULT_SHORT_ENUMS): Define.
34958 2017-04-06  Jakub Jelinek  <jakub@redhat.com>
34960         PR debug/80234
34961         * dwarf2out.c (gen_member_die): Handle C++17 inline static data
34962         members with redundant out-of-class redeclaration.
34964 2017-04-06  Uros Bizjak  <ubizjak@gmail.com>
34966         PR target/80286
34967         * config/i386/sse.md (*vec_extractv4si_0_zext_sse4): New pattern.
34968         * config/i386/i386.md (*zero_extendsidi2):
34969         Add (?*x,*x) and (?*v,*v) alternatives.
34971 2017-04-06  Uros Bizjak  <ubizjak@gmail.com>
34973         PR target/79733
34974         * config/i386/i386.c (ix86_expand_builtin)
34975         <case IX86_BUILTIN_K{,OR}TEST{C,Z}{8,16,32,64}>: Determine insn operand
34976         mode from insn data. Convert operands to insn operand mode.
34977         Copy operands that don't satisfy insn predicate to a register.
34979 2017-04-06  Sam Thursfield  <sam.thursfield@codethink.co.uk>
34981         * config/rs6000/x-aix: Increase memory limit for genautomata on AIX.
34982         Update comments.
34984 2017-04-06  Richard Biener  <rguenther@suse.de>
34986         PR tree-optimization/80334
34987         * tree-ssa-loop-ivopts.c (rewrite_use_address): Properly
34988         preserve alignment of accesses.
34990 2017-04-06  Richard Biener  <rguenther@suse.de>
34992         PR tree-optimization/80262
34993         * tree-sra.c (build_ref_for_offset): Preserve address-space
34994         information.
34995         * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address):
34996         Drop useless address-space information on MEM_REF offsets.
34998 2017-04-05  Andreas Schwab  <schwab@linux-m68k.org>
35000         * builtins.def (BUILT_IN_UPDATE_SETJMP_BUF): Fix type.
35002 2017-04-05  Vladimir Makarov  <vmakarov@redhat.com>
35004         PR rtl-optimization/70703
35005         * ira-color.c (update_conflict_hard_regno_costs): Use
35006         HOST_WIDE_INT instead of long.
35008 2017-04-05  Uros Bizjak  <ubizjak@gmail.com>
35010         PR target/80298
35011         * config/i386/mmintrin.h: Add -msse target option when __SSE__ is
35012         not defined for x86_64 target.  Add -mmmx target option when __SSE2__
35013         is not defined.
35014         * config/i386/mm3dnow.h: Add -msse target when __SSE__ is not defined
35015         for x86_64 target.  Handle -m3dnowa option.
35017 2017-04-05  Vladimir Makarov  <vmakarov@redhat.com>
35019         PR rtl-optimization/70703
35020         * ira-color.c (update_costs_from_allocno): Use the smallest mode.
35021         (update_conflict_hard_regno_costs): Use long instead of unsigned
35022         arithmetic for cost calculation.
35024 2017-04-05  Jakub Jelinek  <jakub@redhat.com>
35025             Bernd Edlinger  <bernd.edlinger@hotmail.de>
35027         PR sanitizer/80308
35028         * asan.c (asan_store_shadow_bytes): Fix location of last_chunk_value
35029         for big endian.
35031 2017-04-05  Eric Botcazou  <ebotcazou@adacore.com>
35033         PR target/78002
35034         * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Replace
35035         ptr_mode with Pmode throughout.
35036         * config/aarch64/aarch64.md (probe_stack_range_<PTR:mode): Rename
35037         into probe_stack_range and use DImode.
35039 2017-04-05  Dominik Vogt  <vogt@linux.vnet.ibm.com>
35041         PR target/79890
35042         * config/s390/s390.c (s390_register_info_gprtofpr): Return if
35043         call_eh_return is true.
35045 2017-04-05  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35047         * config/s390/s390-c.c (s390_resolve_overloaded_builtin):
35048         Initialize last_match_fntype_index.
35050 2017-04-05  Jakub Jelinek  <jakub@redhat.com>
35052         PR target/80310
35053         * tree-nvr.c: Include internal-fn.h.
35054         (pass_return_slot::execute): Ignore internal calls without
35055         direct optab.
35057 2017-04-04  Jakub Jelinek  <jakub@redhat.com>
35058             Richard Biener  <rguenther@suse.de>
35060         PR c++/80297
35061         * genmatch.c (capture::gen_transform): For GENERIC unshare_expr
35062         captures used multiple times, except for the last use.
35063         * generic-match-head.c: Include gimplify.h.
35065 2017-04-04  Jakub Jelinek  <jakub@redhat.com>
35067         PR tree-optimization/79390
35068         * target.h (struct noce_if_info): Declare.
35069         * targhooks.h (default_noce_conversion_profitable_p): Declare.
35070         * target.def (noce_conversion_profitable_p): New target hook.
35071         * ifcvt.h (struct noce_if_info): New type, moved from ...
35072         * ifcvt.c (struct noce_if_info): ... here.
35073         (noce_conversion_profitable_p): Renamed to ...
35074         (default_noce_conversion_profitable_p): ... this.  No longer
35075         static nor inline.
35076         (noce_try_store_flag_constants, noce_try_addcc,
35077         noce_try_store_flag_mask, noce_try_cmove, noce_try_cmove_arith,
35078         noce_convert_multiple_sets): Use targetm.noce_conversion_profitable_p
35079         instead of noce_conversion_profitable_p.
35080         * config/i386/i386.c: Include ifcvt.h.
35081         (ix86_option_override_internal): Don't override
35082         PARAM_MAX_RTL_IF_CONVERSION_INSNS default.
35083         (ix86_noce_conversion_profitable_p): New function.
35084         (TARGET_NOCE_CONVERSION_PROFITABLE_P): Redefine.
35085         * config/i386/x86-tune.def (X86_TUNE_ONE_IF_CONV_INSN): Adjust comment.
35086         * doc/tm.texi.in (TARGET_NOCE_CONVERSION_PROFITABLE_P): Add.
35087         * doc/tm.texi: Regenerated.
35089 2017-04-04  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
35091         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Grammar
35092         correction.
35094 2017-04-04  Thomas Preud'homme  <thomas.preudhomme@arm.com>
35096         PR target/80307
35097         * config/arm/arm.c (thumb1_rtx_costs): Give a cost of 32
35098         instructions for small multiply cores.
35100 2017-04-04  Jeff Law  <law@redhat.com>
35102         * config/mips/mips.c (mips_multi_add): Zero initialize the newly
35103         added member.
35104         (mips_expand_vec_perm_const): Initialize elements in orig_perm
35105         that are not set by the loop over the elements.
35107 2017-04-04  Jakub Jelinek  <jakub@redhat.com>
35109         PR target/80286
35110         * config/i386/i386.c (ix86_expand_args_builtin): If op has scalar
35111         int mode, convert_modes it to mode as unsigned, otherwise use
35112         lowpart_subreg to mode rather than SImode.
35113         * config/i386/sse.md (<mask_codefor>ashr<mode>3<mask_name>,
35114         ashr<mode>3, ashr<mode>3<mask_name>, <shift_insn><mode>3<mask_name>):
35115         Use DImode instead of SImode for the shift count operand.
35116         * config/i386/mmx.md (mmx_ashr<mode>3, mmx_<shift_insn><mode>3):
35117         Likewise.
35119 2017-04-04  Richard Biener  <rguenther@suse.de>
35121         PR middle-end/80281
35122         * match.pd (A + (-B) -> A - B): Make sure to preserve unsigned
35123         arithmetic done for the negate or the plus.  Simplify.
35124         (A - (-B) -> A + B): Likewise.
35125         * fold-const.c (split_tree): Make sure to not negate pointers.
35127 2017-04-04  Segher Boessenkool  <segher@kernel.crashing.org>
35129         PR rtl-optimization/60818
35130         * simplify-rtx.c (simplify_binary_operation_1): Do not replace
35131         a compare of comparisons with the thing compared if this results
35132         in a different machine mode.
35134 2017-04-03  Jonathan Wakely  <jwakely@redhat.com>
35136         * alias.c (base_alias_check): Fix typo in comment.
35137         * cgraph.h (class ipa_polymorphic_call_context): Likewise.
35138         * cgraphunit.c (symbol_table::compile): Likewise.
35139         * collect2.c (maybe_run_lto_and_relink): Likewise.
35140         * config/arm/arm.c (arm_thumb1_mi_thunk): Likewise.
35141         * config/avr/avr-arch.h (avr_arch_info_t): Likewise.
35142         * config/avr/avr.c (avr_map_op_t): Likewise.
35143         * config/cr16/cr16.h (DATA_ALIGNMENT): Likewise.
35144         * config/epiphany/epiphany.c (TARGET_ARG_PARTIAL_BYTES): Likewise.
35145         * config/epiphany/epiphany.md (movcc): Likewise.
35146         * config/i386/i386.c (legitimize_pe_coff_extern_decl): Likewise.
35147         * config/m68k/m68k.c (struct _sched_ib, m68k_sched_variable_issue):
35148         Likewise.
35149         * config/mips/mips.c (mips_save_restore_reg): Likewise.
35150         * config/rx/rx.c (rx_is_restricted_memory_address): Likewise.
35151         * config/s390/s390.c (Z10_EARLYLOAD_DISTANCE): Likewise.
35152         * config/sh/sh.c (sh_rtx_costs): Likewise.
35153         * fold-const.c (fold_truth_andor): Likewise.
35154         * genautomata.c (collapse_flag): Likewise.
35155         * gengtype.h (struct type::u::s): Likewise.
35156         * gensupport.c (has_subst_attribute, add_mnemonic_string): Likewise.
35157         * input.c (FORMAT_AMOUNT): Likewise.
35158         * ipa-cp.c (class ipcp_lattice, agg_replacements_to_vector)
35159         (known_aggs_to_agg_replacement_list): Likewise.
35160         * ipa-inline-analysis.c: Likewise.
35161         * ipa-inline.h (estimate_edge_time, estimate_edge_hints): Likewise.
35162         * ipa-polymorphic-call.c
35163         (ipa_polymorphic_call_context::restrict_to_inner_class): Likewise.
35164         * loop-unroll.c (analyze_insn_to_expand_var): Likewise.
35165         * lra.c (lra_optional_reload_pseudos, lra_subreg_reload_pseudos):
35166         Likewise.
35167         * modulo-sched.c (apply_reg_moves): Likewise.
35168         * omp-expand.c (build_omp_regions_1): Likewise.
35169         * trans-mem.c (struct tm_wrapper_hasher): Likewise.
35170         * tree-ssa-loop-ivopts.c (may_eliminate_iv): Likewise.
35171         * tree-ssa-loop-niter.c (maybe_lower_iteration_bound): Likewise.
35172         * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
35173         * value-prof.c: Likewise.
35174         * var-tracking.c (val_reset): Likewise.
35176 2017-04-03  Richard Biener  <rguenther@suse.de>
35178         PR tree-optimization/80275
35179         * fold-const.c (split_address_to_core_and_offset): Handle
35180         POINTER_PLUS_EXPR.
35182 2017-04-03  Eric Botcazou  <ebotcazou@adacore.com>
35184         * tree-nested.c (get_descriptor_type): Make sure that the alignment of
35185         descriptors is at least equal to that of functions.
35187 2017-04-02  Uros Bizjak  <ubizjak@gmail.com>
35189         * config/i386/sse.md (movdi_to_sse): Add missing DONE.
35191 2017-04-02  Uros Bizjak  <ubizjak@gmail.com>
35193         PR target/80250
35194         * config/i386/sse.md (mov<IMOD4:mode>): Remove insn pattern.
35195         (mov<IMOD4:mode>): New expander.
35196         (*mov<IMOD4:mode>_internal): New insn and split pattern.
35198 2017-03-31  Segher Boessenkool  <segher@kernel.crashing.org>
35200         PR rtl-optimization/79405
35201         * fwprop.c (propagations_left): New variable.
35202         (forward_propagate_into): Decrement it.
35203         (fwprop_init): Initialize it.
35204         (fw_prop): If the variable has reached zero, stop propagating.
35205         (fwprop_addr): Ditto.
35207 2017-03-31  Jakub Jelinek  <jakub@redhat.com>
35209         PR debug/79255
35210         * dwarf2out.c (decls_for_scope): If BLOCK_NONLOCALIZED_VAR is
35211         a FUNCTION_DECL, pass it as decl instead of origin to
35212         process_scope_var.
35214 2017-03-31  Alexander Monakov  <amonakov@ispras.ru>
35216         * config/nvptx/nvptx.c (nvptx_output_softstack_switch): Correct format
35217         string.
35219 2017-03-31  Pat Haugen  <pthaugen@us.ibm.com>
35221         PR target/80107
35222         * config/rs6000/rs6000.md (extendhi<mode>2): Add test for
35223         TARGET_VSX_SMALL_INTEGER.
35225 2017-03-31  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
35227         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Add
35228         reference to the OpenPOWER 64-Bit ELF V2 ABI Specification.
35230 2017-03-31  Matthew Fortune  <matthew.fortune@imgtec.com>
35232         * config/mips/mips-msa.md (msa_vec_extract_<msafmt_f>): Update
35233         extraction from odd-numbered MSA register.
35235 2017-03-31  Jakub Jelinek  <jakub@redhat.com>
35237         PR middle-end/80173
35238         * expmed.c (store_bit_field_1): Don't attempt to create
35239         a word subreg out of hard registers wider than word if they
35240         have HARD_REGNO_NREGS of 1 for their mode.
35242         PR middle-end/80163
35243         * varasm.c (initializer_constant_valid_p_1): Disallow sign-extending
35244         conversions to integer types wider than word and pointer.
35246         PR debug/80025
35247         * cselib.h (rtx_equal_for_cselib_1): Add depth argument.
35248         (rtx_equal_for_cselib_p): Pass 0 to it.
35249         * cselib.c (cselib_hasher::equal): Likewise.
35250         (rtx_equal_for_cselib_1): Add depth argument.  If depth
35251         is 128, don't look up VALUE locs and punt.  Increment
35252         depth in recursive calls when walking VALUE locs.
35254 2017-03-31  Bernd Edlinger  <bernd.edlinger@hotmail.de>
35256         * gcov.c (md5sum_to_hex): Fix output of MD5 hex bytes.
35257         (make_gcov_file_name): Use the canonical path name for generating
35258         the MD5 value.
35259         (read_line): Fix handling of files with ascii null bytes.
35261 2017-03-30  Matthew Fortune  <matthew.fortune@imgtec.com>
35263         * config/mips/mips.c (mips_expand_vector_init): Create a const_vector
35264         to initialise a vector register instead
35265         of using a const_int.
35267 2017-03-30  Jakub Jelinek  <jakub@redhat.com>
35269         PR translation/80189
35270         * gimplify.c (omp_default_clause): Use %qs instead of %s in
35271         diagnostic messages.
35273 2017-03-30  Peter Bergner  <bergner@vnet.ibm.com>
35275         PR target/80246
35276         * config/rs6000/dfp.md (dfp_dxex_<mode>): Update mode of operand 0.
35277         (dfp_diex_<mode>): Update mode of operand 1.
35278         * doc/extend.texi (dxex, dxexq): Document change to return type.
35279         (diex, diexq): Document change to argument type.
35281 2017-03-30  Martin Jambor  <mjambor@suse.cz>
35283         PR ipa/77333
35284         * cgraph.h (cgraph_build_function_type_skip_args): Declare.
35285         * cgraph.c (redirect_call_stmt_to_callee): Set gimple fntype so that
35286         it reflects the signature changes performed at the callee side.
35287         * cgraphclones.c (build_function_type_skip_args): Make public, renamed
35288         to cgraph_build_function_type_skip_args.
35289         (build_function_decl_skip_args): Adjust call to the above function.
35291 2017-03-30  Jakub Jelinek  <jakub@redhat.com>
35293         PR target/80206
35294         * config/i386/sse.md
35295         (<extract_type>_vextract<shuffletype><extract_suf>_mask): Use
35296         register as dest whenever it is a MEM not rtx_equal_p to the
35297         corresponding dup operand, and when forcing into reg move the
35298         reg into the memory afterwards.
35299         (<extract_type_2>_vextract<shuffletype><extract_suf_2>_mask):
35300         Likewise.  Use <ssehalfvecmode> instead of <ssequartermode>
35301         for the force_reg mode.
35302         (avx512vl_vextractf128<mode>): Use register as dest either
35303         always when a MEM, or when it is a MEM not rtx_equal_p to the
35304         corresponding dup operand, or even not when it is a CONST_VECTOR
35305         depending on the mode and lo vs. hi.
35306         (avx512dq_vextract<shuffletype>64x2_1_maskm): Remove extraneous
35307         parens.
35308         (avx512f_vextract<shuffletype>32x4_1_maskm): Likewise.
35309         (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name>):
35310         Likewise.  Require that operands[2] is even.
35311         (<mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name>):
35312         Remove extraneous parens.  Require that operands[2] is a multiple
35313         of 4.
35314         (vec_extract_lo_<mode><mask_name>): Don't bother testing if
35315         operands[0] is a MEM if <mask_applied>, the predicates/constraints
35316         disallow memory then.
35318 2017-03-30  Richard Biener  <rguenther@suse.de>
35320         PR tree-optimization/77498
35321         * tree-ssa-pre.c (phi_translate_1): Do not allow simplifications
35322         to non-constants over backedges.
35324 2017-03-29  Segher Boessenkool  <segher@kernel.crashing.org>
35326         PR rtl-optimization/80233
35327         * combine.c (combine_instructions): Only take NONDEBUG_INSN_P insns
35328         as last_combined_insn.  Do not test for BARRIER_P separately.
35330 2017-03-29  Andreas Schwab  <schwab@suse.de>
35332         PR ada/80146
35333         * calls.c (prepare_call_address): Convert funexp to Pmode before
35334         copying to temp reg.
35336 2017-03-29  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
35338         PR tree-optimization/80158
35339         * gimple-ssa-strength-reduction.c (replace_mult_candidate):
35340         Handle possible future case of more than one alternate
35341         interpretation.
35342         (replace_rhs_if_not_dup): Likewise.
35343         (replace_one_candidate): Likewise.
35345 2017-03-28  Vladimir Makarov  <vmakarov@redhat.com>
35347         PR rtl-optimization/80193
35348         * ira.c (ira): Do not check allocation for LRA.
35350 2017-03-28  Alexander Monakov  <amonakov@ispras.ru>
35352         * config/nvptx/nvptx-protos.h (nvptx_output_simt_enter): Declare.
35353         (nvptx_output_simt_exit): Declare.
35354         * config/nvptx/nvptx.c (nvptx_init_unisimt_predicate): Use
35355         cfun->machine->unisimt_location.  Handle NULL unisimt_predicate.
35356         (init_softstack_frame): Move initialization of crtl->is_leaf to...
35357         (nvptx_declare_function_name): ...here.  Emit declaration of local
35358         memory space buffer for omp_simt_enter insn.
35359         (nvptx_output_unisimt_switch): New.
35360         (nvptx_output_softstack_switch): New.
35361         (nvptx_output_simt_enter): New.
35362         (nvptx_output_simt_exit): New.
35363         * config/nvptx/nvptx.h (struct machine_function): New fields
35364         has_simtreg, unisimt_location, simt_stack_size, simt_stack_align.
35365         * config/nvptx/nvptx.md (UNSPECV_SIMT_ENTER): New unspec.
35366         (UNSPECV_SIMT_EXIT): Ditto.
35367         (omp_simt_enter_insn): New insn.
35368         (omp_simt_enter): New expansion.
35369         (omp_simt_exit): New insn.
35370         * config/nvptx/nvptx.opt (msoft-stack-reserve-local): New option.
35372         * internal-fn.c (expand_GOMP_SIMT_ENTER): New.
35373         (expand_GOMP_SIMT_ENTER_ALLOC): New.
35374         (expand_GOMP_SIMT_EXIT): New.
35375         * internal-fn.def (GOMP_SIMT_ENTER): New internal function.
35376         (GOMP_SIMT_ENTER_ALLOC): Ditto.
35377         (GOMP_SIMT_EXIT): Ditto.
35378         * target-insns.def (omp_simt_enter): New insn.
35379         (omp_simt_exit): Ditto.
35380         * omp-low.c (struct omplow_simd_context): New fields simt_eargs,
35381         simt_dlist.
35382         (lower_rec_simd_input_clauses): Implement SIMT privatization.
35383         (lower_rec_input_clauses): Likewise.
35384         (lower_lastprivate_clauses): Handle SIMT privatization.
35386         * omp-offload.c: Include langhooks.h, tree-nested.h, stor-layout.h.
35387         (ompdevlow_adjust_simt_enter): New.
35388         (find_simtpriv_var_op): New.
35389         (execute_omp_device_lower): Handle IFN_GOMP_SIMT_ENTER,
35390         IFN_GOMP_SIMT_ENTER_ALLOC, IFN_GOMP_SIMT_EXIT.
35392         * tree-inline.h (struct copy_body_data): New field dst_simt_vars.
35393         * tree-inline.c (expand_call_inline): Handle SIMT privatization.
35394         (copy_decl_for_dup_finish): Ditto.
35396         * tree-ssa.c (execute_update_addresses_taken): Handle GOMP_SIMT_ENTER.
35398 2017-03-28  Uros Bizjak  <ubizjak@gmail.com>
35400         PR target/53383
35401         * config/i386/i386.c (ix86_option_override_internal): Always
35402         allow -mpreferred-stack-boundary=3 for 64-bit targets.
35404 2017-03-28  Bin Cheng  <bin.cheng@arm.com>
35406         * tree-vect-loop.c (optimize_mask_stores): Add bb to the right loop.
35408 2017-03-28  Bin Cheng  <bin.cheng@arm.com>
35410         * tree-vect-loop-manip.c (slpeel_add_loop_guard): New param and
35411         mark new edge's irreducible flag accordign to it.
35412         (vect_do_peeling): Check loop preheader edge's irreducible flag
35413         and pass it to function slpeel_add_loop_guard.
35415 2017-03-28  Richard Sandiford  <richard.sandiford@arm.com>
35417         PR tree-optimization/80218
35418         * tree-call-cdce.c (shrink_wrap_one_built_in_call_with_conds):
35419         Update block frequencies and counts.
35421 2017-03-28  Richard Biener  <rguenther@suse.de>
35423         PR tree-optimization/78644
35424         * tree-ssa-ccp.c (evaluate_stmt): When we may not use the value
35425         of a simplification result we may not use it at all.
35427 2017-03-28  Richard Biener  <rguenther@suse.de>
35429         PR ipa/80205
35430         * tree-inline.c (copy_phis_for_bb): Do not create PHI node
35431         without arguments, generate default definition of a SSA name.
35433 2017-03-28  Richard Biener  <rguenther@suse.de>
35435         PR middle-end/80222
35436         * gimple-fold.c (gimple_fold_indirect_ref): Do not touch
35437         TYPE_REF_CAN_ALIAS_ALL references.
35438         * fold-const.c (fold_indirect_ref_1): Likewise.
35440 2017-03-28  Martin Liska  <mliska@suse.cz>
35442         PR ipa/80104
35443         * cgraphunit.c (cgraph_node::expand_thunk): Mark argument of a
35444         thunk call as DECL_GIMPLE_REG_P when vector or complex type.
35446 2017-03-28  Claudiu Zissulescu  <claziss@synopsys.com>
35447             Thomas Petazzoni  <thomas.petazzoni@free-electrons.com>
35449         * config/arc/arc.h (CPP_SPEC): Add subtarget_cpp_spec.
35450         (EXTRA_SPECS): Define.
35451         (SUBTARGET_EXTRA_SPECS): Likewise.
35452         (SUBTARGET_CPP_SPEC): Likewise.
35453         * config/arc/elf.h (EXTRA_SPECS): Renamed to
35454         SUBTARGET_EXTRA_SPECS.
35455         * config/arc/linux.h (SUBTARGET_CPP_SPEC): Define.
35457 2017-03-28  Claudiu Zissulescu  <claziss@synopsys.com>
35459         * config/arc/simdext.md (vst64_insn): Update pattern.
35460         (vld32wh_insn): Likewise.
35461         (vld32wl_insn): Likewise.
35462         (vld64_insn): Likewise.
35463         (vld32_insn): Likewise.
35465 2017-03-28  Marek Polacek  <polacek@redhat.com>
35467         PR sanitizer/80067
35468         * fold-const.c (fold_comparison): Use protected_set_expr_location
35469         instead of SET_EXPR_LOCATION.
35471 2017-03-28  Markus Trippelsdorf  <markus@trippelsdorf.de>
35473         * tree.c (add_expr): Avoid name lookup warning.
35475 2017-03-27  Jeff Law  <law@redhat.com>
35477         PR tree-optimization/80216
35478         * tree-ssa-dom.c (derive_equivalences_from_bit_ior): Fix typo in
35479         function name.  Limit recursion depth.
35480         (record_temporary_equivalences): Corresponding changes.
35482 2017-03-27  Jonathan Wakely  <jwakely@redhat.com>
35484         * doc/invoke.texi (-Wno-narrowing): Reorder so default behavior is
35485         covered first.
35487 2017-03-27  Jakub Jelinek  <jakub@redhat.com>
35489         PR target/80102
35490         * reg-notes.def (REG_CFA_NOTE): Define.  Use it for CFA related
35491         notes.
35492         * cfgcleanup.c (reg_note_cfa_p): New array.
35493         (insns_have_identical_cfa_notes): New function.
35494         (old_insns_match_p): Don't cross-jump in between /f
35495         and non-/f instructions.  If both i1 and i2 are frame related,
35496         verify all CFA notes, their order and content.
35498 2017-03-27  Michael Meissner  <meissner@linux.vnet.ibm.com>
35500         PR target/78543
35501         * config/rs6000/rs6000.md (bswaphi2_extenddi): Combine bswap
35502         HImode and SImode with zero extend to DImode to one insn.
35503         (bswap<mode>2_extenddi): Likewise.
35504         (bswapsi2_extenddi): Likewise.
35505         (bswaphi2_extendsi): Likewise.
35506         (bswaphi2): Combine bswap HImode and SImode into one insn.
35507         Separate memory insns from swapping register.
35508         (bswapsi2): Likewise.
35509         (bswap<mode>2): Likewise.
35510         (bswaphi2_internal): Delete, no longer used.
35511         (bswapsi2_internal): Likewise.
35512         (bswap<mode>2_load): Split bswap HImode/SImode into separate load,
35513         store, and gpr<-gpr swap insns.
35514         (bswap<mode>2_store): Likewise.
35515         (bswaphi2_reg): Register only splitter, combine with the splitter.
35516         (bswaphi2 splitter): Likewise.
35517         (bswapsi2_reg): Likewise.
35518         (bswapsi2 splitter): Likewise.
35519         (bswapdi2): If we have the LDBRX and STDBRX instructions, split
35520         the insns into load, store, and register/register insns.
35521         (bswapdi2_ldbrx): Likewise.
35522         (bswapdi2_load): Likewise.
35523         (bswapdi2_store): Likewise.
35524         (bswapdi2_reg): Likewise.
35526 2017-03-27  Gunther Nikl  <gnikl@users.sourceforge.net>
35528         * system.h (HAVE_DESIGNATED_INITIALIZERS): Fix non C++ case.
35529         (HAVE_DESIGNATED_UNION_INITIALIZERS): Likewise.
35531 2017-03-27  Kelvin Nilsen  <kelvin@gcc.gnu.org>
35533         PR target/80103
35534         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Edit and
35535         add comments.
35536         * config/rs6000/rs6000.c (rs6000_option_override_internal): Add
35537         special handling for target option conflicts between dform
35538         options (-mpower9-dform, -mpower9-dform-vector,
35539         -mpower9-dform-scalar) and -mno-direct-move.
35541 2017-03-27  Richard Biener  <rguenther@suse.de>
35543         PR tree-optimization/80181
35544         * tree-ssa-ccp.c (likely_value): UNDEFINED ^ X is UNDEFINED.
35546 2017-03-27  Claudiu Zissulescu  <claziss@synopsys.com>
35548         * config/arc/predicates.md (move_double_src_operand): Replace the
35549         call to move_double_src_operand with a call to address_operand.
35551 2017-03-27  Claudiu Zissulescu  <claziss@synopsys.com>
35553         * config/arc/elf.h (ARGET_ARC_TP_REGNO_DEFAULT): Define.
35554         * config/arc/linux.h (ARGET_ARC_TP_REGNO_DEFAULT): Likewise.
35555         * config/arc/arc.opt (mtp-regno): Use ARGET_ARC_TP_REGNO_DEFAULT.
35557 2017-03-27  Claudiu Zissulescu  <claziss@synopsys.com>
35559         * config/arc/predicates.md (long_immediate_loadstore_operand):
35560         Consider scaled addresses cases.
35562 2017-03-27  Claudiu Zissulescu  <claziss@synopsys.com>
35564         * config/arc/arc.c (arc_epilogue_uses): BLINK should be also
35565         restored when in interrupt.
35566         * config/arc/arc.md (simple_return): ARCv2 rtie instruction
35567         doesn't have delay slot.
35569 2017-03-27  Richard Biener  <rguenther@suse.de>
35571         PR ipa/79776
35572         * tree-ssa-structalias.c (associate_varinfo_to_alias): Skip
35573         inlined thunk clones.
35575 2017-03-27  Jakub Jelinek  <jakub@redhat.com>
35577         PR sanitizer/80168
35578         * asan.c (instrument_derefs): Copy over last operand from
35579         original COMPONENT_REF to the new COMPONENT_REF with
35580         DECL_BIT_FIELD_REPRESENTATIVE.
35581         * ubsan.c (instrument_object_size): Likewise.
35583 2017-03-27  Richard Biener  <rguenther@suse.de>
35585         PR tree-optimization/80170
35586         * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Make
35587         sure DR/SCEV didnt fold in constants we do not see when looking
35588         at the reference base alignment.
35590 2017-03-27  Richard Biener  <rguenther@suse.de>
35592         PR middle-end/80171
35593         * gimple-fold.c (fold_ctor_reference): Properly guard against
35594         NULL return value from canonicalize_constructor_val.
35596 2017-03-25  Uros Bizjak  <ubizjak@gmail.com>
35598         PR target/80180
35599         * config/i386/i386.c (ix86_expand_builtin)
35600         <IX86_BUILTIN_RDSEED{16,32,64}_STEP>: Do not expand arg0 between
35601         flags reg setting and flags reg using instructions.
35602         <IX86_BUILTIN_RDRAND{16,32,64}_STEP>: Ditto.  Use non-flags reg
35603         clobbering instructions to zero extend op2.
35605 2017-03-25  Gerald Pfeifer  <gerald@pfeifer.com>
35607         * doc/install.texi (Configuration) <--with-aix-soname>:
35608         Update link to AIX ld.
35610 2017-03-25  Bernd Schmidt  <bschmidt@redhat.com>
35612         PR rtl-optimization/80160
35613         PR rtl-optimization/80159
35614         * lra-assigns.c (must_not_spill_p): Tighten new test to also take
35615         reg_alternate_class into account.
35617 2017-03-24  Vladimir Makarov  <vmakarov@redhat.com>
35619         PR target/80148
35620         * lra-assigns.c (assign_by_spills): Add spilled non-reload pseudos
35621         to consider in curr_insn_transform.
35623 2017-03-24  Jakub Jelinek  <jakub@redhat.com>
35625         * genrecog.c (validate_pattern): Add VEC_SELECT validation.
35626         * genmodes.c (emit_min_insn_modes_c): Call emit_mode_nunits
35627         and emit_mode_inner.
35629 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35631         * config/s390/s390-builtins.def: Add VXE builtins.  Add a flags
35632         argument to the overloaded builtin variants.  Use the new flag to
35633         deprecate certain builtin variants.
35634         * config/s390/s390-builtin-types.def: Add new builtin types.
35635         * config/s390/s390-builtins.h: Support new flags field for
35636         overloaded builtins.
35637         * config/s390/s390-c.c (OB_DEF_VAR): New flags field.
35638         (s390_macro_to_expand): Enable vector float data type.
35639         (s390_cpu_cpp_builtins_internal): Indicate support of the new
35640         builtins by incrementing the __VEC__ version number.
35641         (s390_expand_overloaded_builtin): Support expansion of vec_xl and
35642         vec_xst.
35643         (s390_resolve_overloaded_builtin): Emit error messages depending
35644         on the builtin flags.
35645         * config/s390/s390.c (s390_expand_builtin): Support additional
35646         flags argument.  Change error message to match the messages
35647         emitted in s390-c.c.
35648         * config/s390/s390.md: New UNSPEC_* constants.
35649         (op_type): Add new instruction types.
35650         * config/s390/vecintrin.h: Add new builtins and test data class
35651         constants.
35652         * config/s390/vx-builtins.md (V_HW_32_64): Add V4SF.
35653         (V_HW_4, VEC_HW, VECF_HW): New mode iterators.
35654         (VEC_INEXACT, VEC_NOINEXACT): New constants.
35655         ("vec_splats<mode>", "vec_insert<mode>", "vec_promote<mode>")
35656         ("vec_insert_and_zero<mode>", "vec_mergeh<mode>")
35657         ("vec_mergel<mode>"): V_HW -> VEC_HW.
35659         ("vlrlrv16qi", "vstrlrv16qi", "vbpermv16qi", "vec_msumv2di")
35660         ("vmslg", "*vftci<mode>_cconly", "vftci<mode>_intcconly")
35661         ("*vftci<mode>", "vftci<mode>_intcc", "vec_double_s64")
35662         ("vec_double_u64", "vfmin<mode>", "vfmax<mode>"): New definition.
35664         ("and_av2df3", "and_cv2df3", "vec_andc_av2df3")
35665         ("vec_andc_cv2df3", "xor_av2df3", "xor_cv2df3", "vec_nor_av2df3")
35666         ("vec_nor_cv2df3", "ior_av2df3", "ior_cv2df3", "vec_nabs")
35667         ("*vftcidb", "*vftcidb_cconly", "vftcidb"): Remove definition.
35669         ("vec_all_<fpcmpcc:code>v2df", "vec_any_<fpcmpcc:code>v2df")
35670         ("vec_scatter_elementv4si_DI", "vec_cmp<fpcmp:code>v2df")
35671         ("vec_di_to_df_s64", "vec_di_to_df_u64", "vec_df_to_di_u64")
35672         ("vfidb", "*vldeb", "*vledb", "*vec_cmp<insn_cmp>v2df_cconly")
35673         ("vec_cmpeqv2df_cc", "vec_cmpeqv2df_cc", "vec_cmphv2df_cc")
35674         ("vec_cmphev2df_cc", "*vec_cmpeqv2df_cc")
35675         ("*vec_cmphv2df_cc", "*vec_cmphev2df_cc"): Enable new modes as ...
35677         ("vec_all_<fpcmpcc:code><mode>", "vec_any_<fpcmpcc:code><mode>")
35678         ("vec_scatter_element<V_HW_4:mode>_DI")
35679         ("vec_cmp<fpcmp:code><mode>", "vcdgb", "vcdlgb", "vclgdb")
35680         ("vec_fpint<mode>", "vflls")
35681         ("vflrd", "*vec_cmp<insn_cmp><mode>_cconly", "vec_cmpeq<mode>_cc")
35682         ("vec_cmpeq<mode>_cc", "vec_cmph<mode>_cc", "vec_cmphe<mode>_cc")
35683         ("*vec_cmpeq<mode>_cc", "*vec_cmph<mode>_cc")
35684         ("*vec_cmphe<mode>_cc"): ... these.
35686         ("vec_ctd_s64", "vec_ctsl", "vec_ctul", "vec_st2f"): Use rounding
35687         mode constant instead of magic value.
35689 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35691         * config/s390/s390.c (s390_expand_vec_compare): Support other
35692         vector floating point modes than just V2DF.
35693         (s390_expand_vcond): Likewise.
35694         (s390_hard_regno_mode_ok): Allow SFmode values in VRs.
35695         (s390_cannot_change_mode_class): Prevent mode changes between TF
35696         and V1TF in vector registers.
35697         * config/s390/s390.md (DF, SF): New mode attributes.
35698         ("*cmp<mode>_ccs", "add<mode>3", "sub<mode>3", "mul<mode>3")
35699         ("fma<mode>4", "fms<mode>4", "div<mode>3", "*neg<mode>2"): Add
35700         SFmode support for VRs.
35701         * config/s390/vector.md (V_HW, V_HW2, VT_HW, ti*, nonvec): Add new
35702         vector fp modes.
35703         (VFT, VF_HW): New mode iterators.
35704         (vw, sdx): New mode attributes.
35705         ("addv2df3", "subv2df3", "mulv2df3", "divv2df3", "sqrtv2df2")
35706         ("fmav2df4","fmsv2df4", "negv2df2", "absv2df2", "*negabsv2df2")
35707         ("smaxv2df3", "sminv2df3", "*vec_cmp<VFCMP_HW_OP:code>v2df_nocc")
35708         ("vec_cmpuneqv2df", "vec_cmpltgtv2df", "vec_orderedv2df")
35709         ("vec_unorderedv2df"): Adjust the v2df only patterns to support
35710         also the new vector floating point modes.  Renaming to ...
35712         ("add<mode>3", "sub<mode>3", "mul<mode>3", "div<mode>3")
35713         ("sqrt<mode>2", "fma<mode>4", "fms<mode>4", "neg<mode>2")
35714         ("abs<mode>2", "negabs<mode>2", "smax<mode>3")
35715         ("smin<mode>3", "*vec_cmp<VFCMP_HW_OP:code><mode>_nocc")
35716         ("vec_cmpuneq<mode>", "vec_cmpltgt<mode>", "vec_ordered<mode>")
35717         ("vec_unordered<mode>"): ... these.
35719         ("neg_fma<mode>4", "neg_fms<mode>4", "*smax<mode>3_vxe")
35720         ("*smin<mode>3_vxe", "*sminv2df3_vx", "*vec_extendv4sf")
35721         ("*vec_extendv2df"): New insn definitions.
35723 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35725         * config/s390/s390.md ("*adddi3_sign", "*subdi3_sign", "mulditi3")
35726         ("mulditi3_2", "*muldi3_sign"): New patterns.
35727         ("muldi3", "*muldi3", "mulsi3", "*mulsi3"): Add an expander and
35728         rename the pattern definition.
35730 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35732         * config/s390/s390.md ("indirect_jump"): Turn insn definition into
35733         expander.
35734         ("*indirect_jump", "*indirect2_jump"): New pattern definitions.
35736 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35738         * config/s390/s390.c (s390_expand_vec_init): Use vllezl
35739         instruction if possible.
35740         * config/s390/vector.md (vec_halfnumelts): New mode
35741         attribute.
35742         ("*vec_vllezlf<mode>"): New pattern.
35744 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35746         * config/s390/vector.md ("popcountv16qi2", "popcountv8hi2")
35747         ("popcountv4si2", "popcountv2di2"): Rename to ...
35748         ("popcount<mode>2", "popcountv8hi2_vx", "popcountv4si2_vx")
35749         ("popcountv2di2_vx"): ... these and add !TARGET_VXE to the
35750         condition.
35751         ("popcount<mode>2_vxe"): New pattern.
35753 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35755         * common/config/s390/s390-common.c (processor_flags_table): Add
35756         arch12.
35757         * config.gcc: Add arch12.
35758         * config/s390/driver-native.c (s390_host_detect_local_cpu):
35759         Default to arch12 for unknown CPU model numbers.
35760         * config/s390/s390-builtins.def: Add B_VXE builtin flag.
35761         * config/s390/s390-c.c (s390_cpu_cpp_builtins_internal): Adjust
35762         PROCESSOR_max sanity check.
35763         * config/s390/s390-opts.h (enum processor_type): Add
35764         PROCESSOR_ARCH12.
35765         * config/s390/s390.c (processor_table): Add arch12.
35766         (s390_expand_builtin): Add check for B_VXE flag.
35767         (s390_issue_rate): Add PROCESSOR_ARCH12.
35768         (s390_get_sched_attrmask): Likewise.
35769         (s390_get_unit_mask): Likewise.
35770         (s390_sched_score): Enable z13 scheduling for arch12.
35771         (s390_sched_reorder): Likewise.
35772         (s390_sched_variable_issue): Likewise.
35773         * config/s390/s390.h (enum processor_flags): Add PF_ARCH12 and
35774         PF_VXE.
35775         (s390_tune_attr): Use z13 scheduling also for arch12.
35776         (TARGET_CPU_ARCH12, TARGET_CPU_ARCH12_P, TARGET_CPU_VXE)
35777         (TARGET_CPU_VXE_P, TARGET_ARCH12, TARGET_ARCH12_P, TARGET_VXE)
35778         (TARGET_VXE_P): New macros.
35779         * config/s390/s390.md: Add arch12 to cpu attribute.  Add arch12
35780         and vxe to cpu_facility.  Add arch12 and vxe to enabled attribute.
35781         * config/s390/s390.opt: Add arch12 as processor_type.
35783 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35785         * config/s390/s390.md
35786         ("fixuns_truncdddi2", "fixuns_trunctddi2")
35787         ("fixuns_trunc<BFP:mode><GPR:mode>2"): Merge into ...
35788         ("fixuns_trunc<FP:mode><GPR:mode>2"): New expander.
35790         ("fixuns_trunc<BFP:mode><GPR:mode>2", "fixuns_trunc<mode>si2"):
35791         Rename expanders to ...
35793         ("fixuns_trunc<BFP:mode><GPR:mode>2_emu")
35794         ("fixuns_truncdddi2_emu"): ... these.
35796         ("fixuns_trunc<mode>si2_emu"): New expander.
35798         ("*fixuns_truncdfdi2_z13"): Rename to ...
35799         ("*fixuns_truncdfdi2_vx"): ... this.
35801 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35803         * config/s390/2964.md: Remove the single element vector compare
35804         instructions which are no longer used.
35805         * config/s390/s390.c (s390_select_ccmode): Remove handling of
35806         vector CCmodes.
35807         (s390_canonicalize_comparison): Remove handling of DFmode
35808         compares.
35809         (s390_expand_vec_compare_scalar): Remove function.
35810         (s390_emit_compare): Don't call s390_expand_vec_compare_scalar.
35811         * config/s390/s390.md ("*vec_cmp<insn_cmp>df_cconly"): Remove
35812         pattern.
35813         ("*cmp<mode>_ccs"): Add wfcdb instruction.
35815 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35817         * config/s390/s390.md ("mov<mode>_64dfp" DD_DF): Use vleig for loading a
35818         FP zero.
35819         ("*mov<mode>_64" DD_DF): Remove the vector instructions. These
35820         will anyway by matched by mov<mode>_64dfp.
35822 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35824         * config/s390/s390.md ("mov<mode>" SD_SF): Change vleg/vsteg to
35825         vlef/vstef.  Add missing operand to vleif.
35827 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35829         * config/s390/s390.c (s390_expand_vec_init): Enable vector load
35830         pair for all vector types with 64 bit elements.
35831         * config/s390/vx-builtins.md (V_HW_64): Move mode iterator to ...
35832         * config/s390/vector.md (V_HW_64): ... here.
35833         (V_128_NOSINGLE): New mode iterator.
35834         ("vec_init<V_HW:mode>"): Use V_128 as mode iterator.
35835         ("*vec_splat<mode>"): Use V_128_NOSINGLE mode iterator.
35836         ("*vec_tf_to_v1tf", "*vec_ti_to_v1ti"): New pattern definitions.
35837         ("*vec_load_pairv2di"): Change to ...
35838         ("*vec_load_pair<mode>"): ... this one.
35840 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35842         * config/s390/constraints.md: Add comments.
35843         (jKK): Reject element sizes > 8 bytes.
35844         * config/s390/s390.c (s390_split_ok_p): Enable splitting also for
35845         s_operands.
35846         * config/s390/s390.md: Add the s_operand checks formerly in
35847         s390_split_ok_p to various splitters where they are still
35848         required.
35849         * config/s390/vector.md ("mov<mode>" V_128): Add GPR alternatives
35850         for 128 bit vectors.  Plus two splitters.
35852 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35854         * config/s390/s390.md: Rename the cpu facilty vec to vx throughout
35855         the file.
35857 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35859         PR target/79893
35860         * config/s390/s390-c.c (s390_adjust_builtin_arglist): Issue an
35861         error if the boundary argument is not constant.
35863 2017-03-24  Jakub Jelinek  <jakub@redhat.com>
35865         PR rtl-optimization/80112
35866         * loop-doloop.c (doloop_condition_get): Don't check condition
35867         if cmp isn't SET with IF_THEN_ELSE src.
35869 2017-03-24  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
35871         PR tree-optimization/80158
35872         * gimple-ssa-strength-reduction.c (replace_mult_candidate): When
35873         replacing a candidate statement, also replace it for the
35874         candidate's alternate interpretation.
35875         (replace_rhs_if_not_dup): Likewise.
35876         (replace_one_candidate): Likewise.
35878 2017-03-24  Richard Biener  <rguenther@suse.de>
35880         PR tree-optimization/80167
35881         * graphite-isl-ast-to-gimple.c
35882         (translate_isl_ast_to_gimple::is_valid_rename): Handle default-defs
35883         properly.
35884         (translate_isl_ast_to_gimple::get_rename): Likewise.
35886 2017-03-23  Kelvin Nilsen  <kelvin@gcc.gnu.org>
35888         * config/rs6000/rs6000.c (rs6000_option_override_internal): Change
35889         handling of certain combinations of target options, including the
35890         combinations -mpower8-vector vs. -mno-vsx, -mpower9-vector vs.
35891         -mno-power8-vector, and -mpower9_dform vs. -mno-power9-vector.
35893 2017-03-23  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
35895         PR target/71436
35896         * config/arm/arm.md (*load_multiple): Add reload_completed to
35897         matching condition.
35899 2017-03-23  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
35900             Richard Biener  <rguenth@suse.de>
35902         PR tree-optimization/79908
35903         PR tree-optimization/80136
35904         * tree-stdarg.c (expand_ifn_va_arg_1): For a VA_ARG whose LHS has
35905         been cast away, gimplify_and_add suffices.
35907 2017-03-23  Markus Trippelsdorf  <markus@trippelsdorf.de>
35909         * tree-vrp.c (identify_jump_threads): Delete avail_exprs.
35911 2017-03-23  Richard Biener  <rguenther@suse.de>
35913         PR tree-optimization/80032
35914         * gimplify.c (gimple_push_cleanup): Forced unconditional
35915         cleanups still have to go to the conditional_cleanups
35916         sequence.
35918 2017-03-22  Jakub Jelinek  <jakub@redhat.com>
35920         PR tree-optimization/80072
35921         * tree-ssa-reassoc.c (struct operand_entry): Change id field type
35922         to unsigned int.
35923         (next_operand_entry_id): Change type to unsigned int.
35924         (sort_by_operand_rank): Make sure to return the right return value
35925         even if unsigned fields are bigger than INT_MAX.
35926         (struct oecount): Change cnt and id type to unsigned int.
35927         (oecount_hasher::equal): Formatting fix.
35928         (oecount_cmp): Make sure to return the right return value
35929         even if unsigned fields are bigger than INT_MAX.
35930         (undistribute_ops_list): Change next_oecount_id type to unsigned int.
35932         PR c++/80129
35933         * gimplify.c (gimplify_modify_expr_rhs) <case COND_EXPR>: Clear
35934         TREE_READONLY on result if writing it more than once.
35936         PR sanitizer/80110
35937         * doc/invoke.texi (-fsanitize=thread): Document that with
35938         -fnon-call-exceptions atomics are not able to throw
35939         exceptions.
35941         PR sanitizer/80110
35942         * tsan.c: Include tree-eh.h.
35943         (instrument_builtin_call): Call maybe_clean_eh_stmt or
35944         maybe_clean_or_replace_eh_stmt where needed.
35945         (instrument_memory_accesses): Add cfg_changed argument.
35946         Call gimple_purge_dead_eh_edges on each block and set *cfg_changed
35947         if it returned true.
35948         (tsan_pass): Adjust caller.  Return TODO_cleanup_cfg if cfg_changed.
35950         PR rtl-optimization/63191
35951         * config/i386/i386.c (ix86_delegitimize_address): Turn into small
35952         wrapper function, moved the whole old content into ...
35953         (ix86_delegitimize_address_1): ... this.  New inline function.
35954         (ix86_find_base_term): Use ix86_delegitimize_address_1 with
35955         true as last argument instead of ix86_delegitimize_address.
35957 2017-03-22  Wilco Dijkstra  <wdijkstr@arm.com>
35959         * config/aarch64/aarch64.c (generic_branch_cost): Copy
35960         cortexa57_branch_cost.
35962 2017-03-22  Wilco Dijkstra  <wdijkstr@arm.com>
35964         * config/aarch64/aarch64.c (generic_tunings): Add AES fusion.
35966 2017-03-21  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
35968         PR target/80123
35969         * doc/md.texi (Constraints): Document wA constraint.
35970         * config/rs6000/constraints.md (wA): New.
35971         * config/rs6000/rs6000.c (rs6000_debug_reg_global): Add wA reg_class.
35972         (rs6000_init_hard_regno_mode_ok): Init wA constraint.
35973         * config/rs6000/rs6000.h (RS6000_CONSTRAINT_wA): New.
35974         * config/rs6000/vsx.md (vsx_splat_<mode>): Use wA constraint.
35976 2017-03-22  Cesar Philippidis  <cesar@codesourcery.com>
35978         PR c++/80029
35979         * gimplify.c (is_oacc_declared): New function.
35980         (oacc_default_clause): Use it to set default flags for acc declared
35981         variables inside parallel regions.
35982         (gimplify_scan_omp_clauses): Strip firstprivate pointers for acc
35983         declared variables.
35984         (gimplify_oacc_declare): Gimplify the declare clauses.  Add the
35985         declare attribute to any decl as necessary.
35987 2017-03-22  Thomas Preud'homme  <thomas.preudhomme@arm.com>
35989         PR target/80082
35990         * config/arm/arm-isa.h (isa_bit_lpae): New feature bit.
35991         (ISA_ARMv7ve): Add isa_bit_lpae to the definition.
35992         * config/arm/arm-protos.h (arm_arch7ve): Rename into ...
35993         (arm_arch_lpae): This.
35994         * config/arm/arm.c (arm_arch7ve): Rename into ...
35995         (arm_arch_lpae): This.  Define it in term of isa_bit_lpae.
35996         * config/arm/arm.h (TARGET_HAVE_LPAE): Redefine in term of
35997         arm_arch_lpae.
35999 2017-03-22  Martin Liska  <mliska@suse.cz>
36001         PR target/79906
36002         * config/rs6000/rs6000.c (rs6000_inner_target_options): Show
36003         error message instead of an ICE.
36005 2017-03-21  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
36007         * doc/extend.texi (6.11 Additional Floating Types): Revise.
36009 2017-03-21  Kelvin Nilsen  <kelvin@gcc.gnu.org>
36011         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Add
36012         comments.
36013         * config/rs6000/rs6000.c (rs6000_option_override_internal): Add
36014         comments.
36016 2017-03-21  Martin Sebor  <msebor@redhat.com>
36018         * doc/extend.texi: Use "cannot" instead of "can't."
36019         * doc/hostconfig.texi: Same.
36020         * doc/install.texi: Same.
36021         * doc/invoke.texi: Same.
36022         * doc/loop.texi: Same.
36023         * doc/md.texi: Same.
36024         * doc/objc.texi: Same.
36025         * doc/rtl.texi: Same.
36026         * doc/tm.texi: Same.
36027         * doc/tm.texi.in: Same.
36028         * doc/trouble.texi: Same.
36030 2017-03-21  Alexandre Oliva  <aoliva@redhat.com>
36032         PR debug/63238
36033         * dwarf2out.c (struct checksum_attributes): Add at_alignment.
36034         (collect_checksum_attributes): Set it.
36035         (die_checksum_ordered): Use it.
36037 2017-03-21  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
36039         PR tree-optimization/79908
36040         * tree-stdarg.c (expand_ifn_va_arg_1): Revert the following
36041         change: For a VA_ARG whose LHS has been cast away, use
36042         force_gimple_operand to construct the side effects.
36044 2017-03-21  David Malcolm  <dmalcolm@redhat.com>
36046         PR translation/80001
36047         * omp-offload.c (oacc_loop_fixed_partitions): Make diagnostics
36048         more amenable to translation.
36049         (oacc_loop_auto_partitions): Likewise.
36051 2017-03-21  Marek Polacek  <polacek@redhat.com>
36052             Martin Sebor  <msebor@redhat.com>
36054         PR tree-optimization/80109
36055         * gimple-ssa-warn-alloca.c (alloca_call_type): Only call get_range_info
36056         on INTEGRAL_TYPE_P.
36058 2017-03-21  Jakub Jelinek  <jakub@redhat.com>
36059             Segher Boessenkool  <segher@kernel.crashing.org>
36061         PR target/80125
36062         * combine.c (can_combine_p): Revert the 2017-03-20 change, only
36063         check reg_used_between_p between insn and one of succ or succ2
36064         depending on if succ is artificial insn not inserted into insn
36065         stream.
36067 2017-03-21  Martin Liska  <mliska@suse.cz>
36069         PR gcov-profile/80081
36070         * Makefile.in: Add gcov-dump and fix installation of gcov-tool.
36071         * doc/gcc.texi: Include gcov-dump stuff.
36072         * doc/gcov-dump.texi: New file.
36074 2017-03-21  Toma Tabacu  <toma.tabacu@imgtec.com>
36076         PR rtl-optimization/79150
36077         * config/mips/mips.c (mips_block_move_loop): Emit a NOP after the
36078         conditional jump, if the jump is the last insn of the loop.
36080 2017-03-21  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
36081             Richard Biener  <rguenth@suse.de>
36083         PR tree-optimization/79908
36084         * tree-stdarg.c (expand_ifn_va_arg_1): For a VA_ARG whose LHS has
36085         been cast away, use force_gimple_operand to construct the side
36086         effects.
36088 2017-03-21  Martin Liska  <mliska@suse.cz>
36090         PR libfortran/79956
36091         * simplify-rtx.c (simplify_immed_subreg): Initialize a variable
36092         to NULL.
36094 2017-03-21  Brad Spengler <spender@grsecurity.net>
36096         PR plugins/80094
36097         * plugin.c (htab_hash_plugin): New function.
36098         (add_new_plugin): Use it and adjust.
36099         (parse_plugin_arg_opt): Adjust.
36100         (init_one_plugin): Likewise.
36102 2017-03-21  Richard Biener  <rguenther@suse.de>
36104         PR tree-optimization/80032
36105         * gimplify.c (gimple_push_cleanup): Add force_uncond parameter,
36106         if set force the cleanup to happen unconditionally.
36107         (gimplify_target_expr): Push inserted clobbers with force_uncond
36108         to avoid them being removed by control-dependent DCE.
36110 2017-03-21  Richard Biener  <rguenther@suse.de>
36112         PR tree-optimization/80122
36113         * tree-inline.c (copy_bb): Do not expans va-arg packs or
36114         va_arg_pack_len when the inlined call stmt requires pack
36115         expansion itself.
36116         * tree-inline.h (struct copy_body_data): Make call_stmt a gcall *.
36118 2017-03-21  Jakub Jelinek  <jakub@redhat.com>
36120         PR sanitizer/78158
36121         * tsan.c (instrument_builtin_call): If the memory model argument
36122         is not a constant, assume it is valid.
36124         PR c/67338
36125         * fold-const.c (round_up_loc): Negate divisor in unsigned type to
36126         avoid UB.
36128 2017-03-20  Segher Boessenkool  <segher@kernel.crashing.org>
36130         PR rtl-optimization/79910
36131         * combine.c (can_combine_p): Do not allow combining an I0 or I1
36132         if its dest is used by an insn before I2 (other than the combined
36133         insns themselves, which are properly handled already).
36135 2017-03-20  Segher Boessenkool  <segher@kernel.crashing.org>
36137         Revert:
36138         2017-03-17  Bernd Schmidt  <bschmidt@redhat.com>
36140         * combine.c (record_used_regs): New static function.
36141         (try_combine): Handle situations where there is an additional
36142         instruction between I2 and I3 which needs to have a LOG_LINK
36143         updated.
36145         Revert:
36146         2017-03-17  Jim Wilson  <jim.wilson@linaro.org>
36148         * combine.c (try_combine): Delete redundant i1 test.  Call
36149         prev_nonnote_nondebug_insn instead of prev_nonnote_insn.
36151 2017-03-20  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
36153         PR target/80083
36154         * config/rs6000/rs6000.md (*movsi_internal1): Fix constraints for
36155         alternatives 13/14.
36157 2017-03-20  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
36159         PR tree-optimization/80054
36160         * gimple-ssa-strength-reduction.c (all_phi_incrs_profitable): Fail
36161         the optimization if a PHI or any of its arguments is not dominated
36162         by the candidate's basis.  Use gphi* rather than gimple* as
36163         appropriate.
36164         (replace_profitable_candidates): Clean up a gimple* variable that
36165         should be a gphi* variable.
36167 2017-03-20  Martin Sebor  <msebor@redhat.com>
36169         PR c++/52477
36170         * doc/extend.texi (attribute constructor): Document present limitation.
36172 2017-03-20  Kelvin Nilsen  <kelvin@gcc.gnu.org>
36174         PR target/79963
36175         * config/rs6000/altivec.h (vec_all_ne): Under __cplusplus__ and
36176         __POWER9_VECTOR__ #ifdef control, change template definition to
36177         use Power9-specific built-in function.
36178         (vec_any_eq): Likewise.
36179         * config/rs6000/vector.md (vector_ae_v2di_p): Change the flag used
36180         to control outcomes from this test.
36181         (vector_ae_<mode>p): For VEC_F modes, likewise.
36183 2017-03-20  Ian Lance Taylor  <iant@google.com>
36185         * config/i386/i386.c (ix86_function_regparm): Save an extra
36186         register for -fsplit-stack with DECL_STATIC_CHAIN.
36188 2017-03-17  Palmer Dabbelt  <palmer@dabbelt.com>
36190         PR target/79912
36191         * config/riscv/riscv.c (riscv_preferred_reload_class): Remove.
36192         (TARGET_PREFERRED_RELOAD_CLASS): Likewise.
36194 2017-03-17  Palmer Dabbelt  <palmer@dabbelt.com>
36196         * config/riscv/riscv.c (riscv_print_operand): Use "fence
36197         iorw,ow".
36198         * config/riscv/sync.mc (mem_thread_fence_1): Use "fence
36199         iorw,iorw".
36201 2017-03-20  Marek Polacek  <polacek@redhat.com>
36203         PR sanitizer/80063
36204         * asan.c (DEF_SANITIZER_BUILTIN): Use do { } while (0).
36206 2017-03-20  Richard Biener  <rguenther@suse.de>
36208         PR tree-optimization/80113
36209         * graphite-isl-ast-to-gimple.c (copy_loop_phi_nodes): Do not
36210         allocate extra SSA name for PHI def.
36211         (add_close_phis_to_outer_loops): Likewise.
36212         (add_close_phis_to_merge_points): Likewise.
36213         (copy_loop_close_phi_args): Likewise.
36214         (copy_cond_phi_nodes): Likewise.
36216 2017-03-20  Martin Liska  <mliska@suse.cz>
36218         PR middle-end/79753
36219         * tree-chkp.c (chkp_build_returned_bound): Do not build
36220         returned bounds for a LHS that's not a BOUNDED_P type.
36222 2017-03-20  Martin Liska  <mliska@suse.cz>
36224         PR target/79769
36225         PR target/79770
36226         * tree-chkp.c (chkp_find_bounds_1): Handle REAL_CST,
36227         COMPLEX_CST and VECTOR_CST.
36229 2017-03-20  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
36231         PR target/78857
36232         * config/s390/s390.md ("cmp<mode>_ccs_0"): Add a clobber of the
36233         target operand.  A new splitter adds the clobber statement in case
36234         the target operand is dead anyway.
36236 2017-03-19  Gerald Pfeifer  <gerald@pfeifer.com>
36238         * doc/install.texi (Specific) <sparc-*-linux*>: No longer refer
36239         to age-old versions of binutils and glibc.
36241 2017-03-18  Segher Boessenkool  <segher@kernel.crashing.org>
36243         * doc/contrib.texi (Contributors): Remove duplicate entry for myself.
36245 2017-03-18  Gerald Pfeifer  <gerald@pfeifer.com>
36247         * doc/contrib.texi (Contributors): Add Segher Boessenkool.
36249 2017-03-18  Gerald Pfeifer  <gerald@pfeifer.com>
36251         * doc/install.texi (Specific) <arm-*-eabi>: Remove old
36252         requirement for binutils 2.13.
36254 2017-03-17  Jim Wilson  <jim.wilson@linaro.org>
36256         * combine.c (try_combine): Delete redundant i1 test.  Call
36257         prev_nonnote_nondebug_insn instead of prev_nonnote_insn.
36259 2017-03-17  Palmer Dabbelt  <palmer@dabbelt.com
36261         * doc/install.texi (Specific) <riscv32-*-elf>: Add riscv32-*-elf,
36262         riscv32-*-linux, riscv64-*-elf, riscv64-*-linux to the table of
36263         contents.
36264         <riscv64-*-elf>: Re-arrange section
36265         <riscv32-*-elf>: Add a note about requiring binutils 2.28.
36266         <riscv32-*-linux>: Likewise.
36267         <riscv64-*-elf>: Likewise
36268         <riscv64-*-linux>: Likewise.
36270 2017-03-17  Richard Earnshaw  <rearnsha@arm.com>
36272         PR target/80052
36273         * aarch64.opt(verbose-cost-dump): Fix typo.
36275 2017-03-17  Pat Haugen  <pthaugen@us.ibm.com>
36277         PR target/79951
36278         * config/rs6000/rs6000.md (copysign<mode>3_fcpsgn): Test
36279         for VECTOR_UNIT_VSX_P (<MODE>mode) too.
36281 2017-03-17  Bernd Schmidt  <bschmidt@redhat.com>
36283         * reload.c (find_reloads): When reloading a nonoffsettable address,
36284         use RELOAD_OTHER for it and its address reloads.
36286         PR rtl-optimization/79910
36287         * combine.c (record_used_regs): New static function.
36288         (try_combine): Handle situations where there is an additional
36289         instruction between I2 and I3 which needs to have a LOG_LINK
36290         updated.
36292 2017-03-17  Jeff Law  <law@redhat.com>
36294         PR tree-optimization/71437
36295         * tree-vrp.c (simplify_stmt_for_jump_threading): Lookup the
36296         conditional in the hash table first.
36297         (vrp_dom_walker::before_dom_children): Extract condition from
36298         ASSERT_EXPR.  Record condition, its inverion and any implied
36299         conditions as well.
36301 2017-03-17  Marek Polacek  <polacek@redhat.com>
36302             Markus Trippelsdorf  <markus@trippelsdorf.de>
36304         PR tree-optimization/80079
36305         * gimple-ssa-store-merging.c (class pass_store_merging): Initialize
36306         m_stores_head.
36308 2017-03-17  Richard Biener  <rguenther@suse.de>
36310         PR middle-end/80075
36311         * tree-eh.c (stmt_could_throw_1_p): Only handle gimple assigns.
36312         Properly verify the LHS before the RHS possibly claims to be
36313         handled.
36314         (stmt_could_throw_p): Hande gimple conds fully here.  Clobbers
36315         do not throw.
36317 2017-03-17  Martin Jambor  <mjambor@suse.cz>
36319         * doc/invoke.texi (Option Options): Include -fipa-vrp in the list.
36320         (List of -O2 options): Likewise.
36321         (-fipa-bit-cp): Replace "ipa" with "interprocedural."
36322         (-fipa-vrp) New.
36324 2017-03-17  Tom de Vries  <tom@codesourcery.com>
36326         * gcov-dump.c (print_usage): Print bug_report_url.
36328 2017-03-17  Richard Biener  <rguenther@suse.de>
36330         PR middle-end/80050
36331         * genmatch.c (parser::next): Remove pointless check for CPP_EOF.
36332         (parser::peek): Likewise.
36334 2017-03-17  Richard Biener  <rguenther@suse.de>
36336         PR tree-optimization/80048
36337         * sese.c (free_sese_info): Properly release rename_map and
36338         copied_bb_map elements.
36340 2017-03-16  Alexandre Oliva  <aoliva@redhat.com>
36342         * gimple-ssa-store-merging.c (struct imm_store_chain_info):
36343         Add linked-list forward and backlinks.  Insert on
36344         construction, remove on destruction.
36345         (class pass_store_merging): Add m_stores_head field.
36346         (pass_store_merging::terminate_and_process_all_chains):
36347         Iterate over m_stores_head list.
36348         (pass_store_merging::terminate_all_aliasing_chains):
36349         Likewise.
36350         (pass_store_merging::execute): Check for debug stmts first.
36351         Push new chains onto the m_stores_head stack.
36353 2017-03-16  Michael Meissner  <meissner@linux.vnet.ibm.com>
36355         PR target/71294
36356         * config/rs6000/vsx.md (vsx_splat_<mode>, VSX_D iterator): Allow a
36357         SPLAT operation on ISA 2.07 64-bit systems that have direct move,
36358         but no MTVSRDD support, by doing MTVSRD and XXPERMDI.
36360 2017-03-16  Jeff Law  <law@redhat.com>
36362         PR tree-optimization/71437
36363         * tree-ssa-dom.c (dom_opt_dom_walker): Remove thread_across_edge
36364         member function.  Implementation moved into after_dom_children
36365         member function and into the threader's thread_outgoing_edges
36366         function.
36367         (dom_opt_dom_walker::after_dom_children): Simplify by moving
36368         some code into new thread_outgoing_edges.
36369         * tree-ssa-threadedge.c (thread_across_edge): Make static and simplify
36370         definition.  Simplify marker handling (do it here).   Assume we always
36371         have the available expression and the const/copies tables.
36372         (thread_outgoing_edges): New function extracted from tree-ssa-dom.c
36373         and tree-vrp.c
36374         * tree-ssa-threadedge.h (thread_outgoing_edges): Declare.
36375         * tree-vrp.c (equiv_stack): No longer file scoped.
36376         (vrp_dom_walker): New class.
36377         (vrp_dom_walker::before_dom_children): New member function.
36378         (vrp_dom_walker::after_dom_children): Likewise.
36379         (identify_jump_threads):  Setup domwalker.  Use it rather than
36380         walking edges in a random order by hand.  Simplify setup/finalization.
36381         (finalize_jump_threads): Remove.
36382         (vrp_finalize): Do not call identify_jump_threads here.
36383         (execute_vrp): Do it here instead and call thread_through_all_blocks
36384         here too.
36386         PR tree-optimization/71437
36387         * tree-ssa-dom.c (pfn_simplify): Add basic_block argument.  All
36388         callers changed.
36389         (simplify_stmt_for_jump_threading): Add basic_block argument.  All
36390         callers changed.
36391         (lhs_of_dominating_assert): Moved from here into tree-vrp.c.
36392         (dom_opt_dom_walker::thread_across_edge): Remove
36393         handle_dominating_asserts argument.  All callers changed.
36394         (record_temporary_equivalences_from_stmts_at_dest): Corresponding
36395         changes.  Remove calls to lhs_of_dominating_assert.  Other
36396         uses of handle_dominating_asserts turn into unconditional code
36397         (simplify_control_stmt_condition_1): Likewise.
36398         (simplify_control_stmt_condition): Likewise.
36399         (thread_through_normal_block, thread_across_edge): Likewise.
36400         * tree-ssa-threadedge.h (thread_across_edge): Corresponding changes.
36401         * tree-vrp.c (lhs_of_dominating_assert): Move here.  Return original
36402         object if it is not an SSA_NAME.
36403         (simplify_stmt_for_jump_threading): Call lhs_of_dominating_assert
36404         before calling into the VRP specific simplifiers.
36405         (identify_jump_threads): Remove handle_dominating_asserts
36406         argument.
36408 2017-03-16  Jakub Jelinek  <jakub@redhat.com>
36410         PR fortran/79886
36411         * tree-diagnostic.c (default_tree_printer): No longer static.
36412         * tree-diagnostic.h (default_tree_printer): New prototype.
36414 2017-03-16  Tamar Christina  <tamar.christina@arm.com>
36416         * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>)
36417         Change ins into fmov.
36419 2017-03-16  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
36421         * config/aarch64/iterators.md (h_con): Return "x" for V4HF and V8HF.
36422         * config/aarch64/aarch64-simd.md (*aarch64_fma4_elt_from_dup<mode>):
36423         Use h_con constraint for operand 1.
36424         (*aarch64_fnma4_elt_from_dup<mode>): Likewise.
36425         (*aarch64_mulx_elt_from_dup<mode>): Likewise for operand 2.
36427 2017-03-15  Jeff Law  <law@redhat.com>
36429         PR tree-optimization/71437
36430         * tree-ssa-dom.c (derive_equivalences_from_bit_ior): New function.
36431         (record_temporary_equivalences): Use it.
36433         PR tree-optimization/71437
36434         * tree-ssa-dom.c (struct cond_equivalence): Moved from here into
36435         tree-ssa-scopedtables.
36436         (lookup_avail_expr, build_and_record_new_cond): Likewise.
36437         (record_conditions, record_cond, vuse_eq): Likewise.
36438         (record_edge_info): Adjust to API tweak of record_conditions.
36439         (simplify_stmt_for_jump_threading): Similarly for lookup_avail_expr.
36440         (record_temporary_equivalences, optimize_stmt): Likewise.
36441         (eliminate_redundant_computations): Likewise.
36442         (record_equivalences_from_stmt): Likewise.
36443         * tree-ssa-scopedtables.c: Include options.h and params.h.
36444         (vuse_eq): New function, moved from tree-ssa-dom.c
36445         (build_and_record_new_cond): Likewise.
36446         (record_conditions): Likewise.  Accept vector of conditions rather
36447         than edge_equivalence structure for first argument.
36448         for the first argument.
36449         (avail_exprs_stack::lookup_avail_expr): New member function, moved
36450         from tree-ssa-dom.c.
36451         (avail_exprs_stack::record_cond): Likewise.
36452         * tree-ssa-scopedtables.h (struct cond_equivalence): Moved here
36453         from tree-ssa-dom.c.
36454         (avail_exprs_stack): Add new member functions lookup_avail_expr
36455         and record_cond.
36456         (record_conditions): Declare.
36458 2017-03-15  Vladimir Makarov  <vmakarov@redhat.com>
36460         PR target/80017
36461         * lra-constraints.c (process_alt_operands): Increase reject for
36462         reloading an input/output operand.
36464 2017-03-15  Michael Meissner  <meissner@linux.vnet.ibm.com>
36466         PR target/79038
36467         * config/rs6000/rs6000.md (float<QHI:mode><IEEE128:mode>2): Define
36468         insns to convert from signed/unsigned char/short to IEEE 128-bit
36469         floating point.
36470         (floatuns<QHI:mode><IEEE128:mode>2): Likewise.
36472 2017-03-15  Uros Bizjak  <ubizjak@gmail.com>
36474         PR target/80019
36475         * config/i386/i386.c (ix86_vector_duplicate_value): Create
36476         subreg of inner mode for values already in registers.
36478 2017-03-15  Bernd Schmidt  <bschmidt@redhat.com>
36480         * config/c6x/c6x.c (hwloop_optimize): Handle case where the old
36481         iteration reg is used after the loop.
36483 2017-03-14  Martin Sebor  <msebor@redhat.com>
36485         PR tree-optimization/79800
36486         * gimple-ssa-sprintf.c (format_floating: Add argument.  Handle
36487         precision in negative-positive range.
36488         (format_floating): Call non-const overload with adjusted precision.
36490 2017-03-14  Michael Meissner  <meissner@linux.vnet.ibm.com>
36492         PR target/79947
36493         * config/rs6000/rs6000.h (TARGET_FRSQRTES): Add check for
36494         -mpowerpc-gfxopt.
36496 2017-03-14  Martin Sebor  <msebor@redhat.com>
36498         PR middle-end/80020
36499         * builtin-attrs.def (ATTR_ALLOC_SIZE_2_NOTHROW_LIST): New macro.
36500         * builtins.def (aligned_alloc): Use it.
36502         PR c/79936
36503         * Makefile.in (GTFILES): Add calls.c.
36504         * calls.c: Include "gt-calls.h".
36506 2017-03-14  Bernd Schmidt  <bschmidt@redhat.com>
36508         PR rtl-optimization/79728
36509         * regs.h (struct target_regs): New field
36510         x_contains_allocatable_regs_of_mode.
36511         (contains_allocatable_regs_of_mode): New macro.
36512         * reginfo.c (init_reg_sets_1): Initialize it, and change
36513         contains_reg_of_mode so it includes global regs as well.
36514         * reload.c (push_reload): Use contains_allocatable_regs_of_mode
36515         rather than contains_regs_of_mode.
36517 2017-03-14  Martin Liska  <mliska@suse.cz>
36519         * doc/invoke.texi: Document options that can't be combined with
36520         -fcheck-pointer-bounds.
36522 2017-03-14  Martin Liska  <mliska@suse.cz>
36524         PR middle-end/79831
36525         * doc/invoke.texi (-Wchkp): Document the option.
36527 2017-03-14  Martin Liska  <mliska@suse.cz>
36529         * Makefile.in: Install gcov-dump.
36531 2017-03-14  Martin Liska  <mliska@suse.cz>
36533         * multiple_target.c (expand_target_clones): Bail out for
36534         an invalid attribute.
36536 2017-03-14  Richard Biener  <rguenther@suse.de>
36538         * alias.c (struct alias_set_entry): Pack properly.
36539         * cfgloop.h (struct loop): Likewise.
36540         * cse.c (struct set): Likewise.
36541         * ipa-utils.c (struct searchc_env): Likewise.
36542         * loop-invariant.c (struct invariant): Likewise.
36543         * lra-remat.c (struct cand): Likewise.
36544         * recog.c (struct change_t): Likewise.
36545         * rtl.h (struct address_info): Likewise.
36546         * symbol-summary.h (function_summary): Likewise.
36547         * tree-loop-distribution.c (struct partition): Likewise.
36548         * tree-object-size.c (struct object_size_info): Likewise.
36549         * tree-ssa-loop-ivopts.c (struct cost_pair): Likewise.
36550         * tree-ssa-threadupdate.c (struct ssa_local_info_t): Likewise.
36551         * tree-vect-data-refs.c (struct _vect_peel_info): Likewise.
36552         * tree-vect-slp.c (struct _slp_oprnd_info): Likewise.
36553         * tree-vect-stmts.c (struct simd_call_arg_info): Likewise.
36554         * tree-vectorizer.h (struct _loop_vec_info): Likewise.
36555         (struct _stmt_vec_info): Likewise.
36557 2017-03-14  Martin Liska  <mliska@suse.cz>
36559         PR target/79892
36560         * multiple_target.c (create_dispatcher_calls): Check that
36561         a target can create a function dispatcher.
36563 2017-03-14  Martin Liska  <mliska@suse.cz>
36565         PR lto/66295
36566         * multiple_target.c (expand_target_clones): Drop local.local
36567         flag for default implementation.
36569 2017-03-14  Richard Biener  <rguenther@suse.de>
36571         PR tree-optimization/80030
36572         * tree-vect-stmts.c (vectorizable_store): Plug memleak.
36574 2017-03-13  Kito Cheng  <kito.cheng@gmail.com>
36576         * config/riscv/riscv.c (riscv_emit_float_compare>: Use
36577         gcc_fallthrough() instead of __attribute__((fallthrough));
36579 2017-03-13  Gerald Pfeifer  <gerald@pfeifer.com>
36581         * doc/gcc.texi: Remove "up" link to (DIR).
36582         * doc/gccint.texi: Ditto.
36584 2017-03-13  Gerald Pfeifer  <gerald@pfeifer.com>
36586         * doc/install.texi (Specific) <avr>: Remove reference to
36587         binutils 2.13.
36589 2017-03-13  Jeff Law  <law@redhat.com>
36591         * config/riscv/riscv.c (riscv_emit_float_compare): Use fallthru
36592         attribute rather than comments.
36594         * config/pdp11/pdp11.md (movmemhi): Adjust operand numbers to
36595         match_scratch operand is highest.
36597 2017-03-13  Martin Liska  <mliska@suse.cz>
36599         PR middle-end/78339
36600         * ipa-pure-const.c (warn_function_noreturn): If the declarations
36601         is a CHKP clone, use original declaration.
36603 2017-03-13  Claudiu Zissulescu  <claziss@synopsys.com>
36605         * config/arc/arc.c (arc_init): Use multiplier whenever we have it.
36606         (arc_conditional_register_usage): Use a different allocation order
36607         when optimizing for size.
36608         * common/config/arc/arc-common.c (arc_option_optimization_table):
36609         Section anchors default on when optimizing for size.
36611 2017-03-13  Claudiu Zissulescu  <claziss@synopsys.com>
36613         * config/arc/arc.md (*tst_bitfield_tst): Fix pattern.
36615 2017-03-13  Claudiu Zissulescu  <claziss@synopsys.com>
36617         * config/arc/arc.c (arc_output_addsi): Emit code density adds.
36618         * config/arc/arc.md (cpu_facility): Add cd variant.
36619         (*movqi_insn): Add code density variant.
36620         (*movhi_insn): Likewise.
36621         (*movqi_insn): Likewise.
36622         (*addsi3_mixed): Likewise.
36623         (subsi3_insn): Likewise.
36625 2017-03-13  Claudiu Zissulescu  <claziss@synopsys.com>
36627         * config/arc/arc.md (movsi_cond_exec): Update constraint.
36629 2017-03-13  Claudiu Zissulescu  <claziss@synopsys.com>
36631         * config/arc/arc.c (arc_legitimize_pic_address): Handle PIC
36632         expressions with MINUS and UNARY ops.
36634 2017-03-13  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
36636         PR target/79911
36637         * config/arm/neon.md (vec_sel_widen_ssum_lo<VQI:mode><VW:mode>3):
36638         Rename to...
36639         (vec_sel_widen_ssum_lo<mode><V_half>3): ... This. Avoid mismatch
36640         between vec_select and vector argument.
36641         (vec_sel_widen_ssum_hi<VQI:mode><VW:mode>3): Rename to...
36642         (vec_sel_widen_ssum_hi<mode><V_half>3): ... This. Likewise.
36643         (vec_sel_widen_usum_lo<VQI:mode><VW:mode>3): Rename to...
36644         (vec_sel_widen_usum_lo<mode><V_half>3): ... This.
36645         (vec_sel_widen_usum_hi<VQI:mode><VW:mode>3): Rename to...
36646         (vec_sel_widen_usum_hi<mode><V_half>3): ... This.
36648 2017-03-13  Richard Biener  <rguenther@suse.de>
36650         PR other/79991
36651         * params.def (vect-max-peeling-for-alignment): Fix typo.
36653 2017-03-12  Gerald Pfeifer  <gerald@pfeifer.com>
36655         * doc/install.texi (Specific) <mips-*-*>: Remove description of
36656         issue that only occurred with binutils below 2.18.
36658 2017-03-12  Gerald Pfeifer  <gerald@pfeifer.com>
36660         * doc/install.texi (Specific) <cris-axis-elf>: No longer
36661         refer to binutils 2.11/2.12 minimum.
36663 2017-03-12  Gerald Pfeifer  <gerald@pfeifer.com>
36665         * doc/install.texi (Specific) <powerpc-*-*>: Remove link to
36666         ftp.kernel.org and simplify binutils requirement.
36668 2017-03-11  Gerald Pfeifer  <gerald@pfeifer.com>
36670         * doc/invoke.texi (Warning Options): Fix spelling of link-time
36671         optimization.
36672         (Optimize Options): Ditto.  Also remove redundancy.
36674 2017-03-10  David Malcolm  <dmalcolm@redhat.com>
36676         PR translation/79848
36677         * ipa-devirt.c (warn_types_mismatch): Simplify uses of "%<%s%>" to
36678         "%qs".
36679         * ipa-pure-const.c (suggest_attribute): Likewise.  Convert _
36680         to G_ to avoid double translation.
36682 2017-03-10  David Malcolm  <dmalcolm@redhat.com>
36684         PR translation/79923
36685         * auto-profile.c (get_combined_location): Convert leading
36686         character of diagnostics to lower case and remove trailing period.
36687         (read_profile): Likewise for various diagnostics.
36688         * config/arm/arm.c (arm_option_override): Remove trailing period
36689         from various diagnostics.
36690         * config/msp430/msp430.c (msp430_expand_delay_cycles): Likewise.
36691         (msp430_expand_delay_cycles): Likewise.
36693 2017-03-10  David Malcolm  <dmalcolm@redhat.com>
36695         PR target/79925
36696         * config/aarch64/aarch64.c (aarch64_validate_mcpu): Quote the
36697         full command-line argument, rather than just "str".
36698         (aarch64_validate_march): Likewise.
36699         (aarch64_validate_mtune): Likewise.
36701 2017-03-10  Bernd Schmidt  <bschmidt@redhat.com>
36703         PR rtl-optimization/78911
36704         * lra-assigns.c (must_not_spill_p): New function.
36705         (spill_for): Use it.
36707 2017-03-10  Jakub Jelinek  <jakub@redhat.com>
36709         PR tree-optimization/79981
36710         * tree-vrp.c (extract_range_basic): Handle IMAGPART_EXPR of
36711         ATOMIC_COMPARE_EXCHANGE ifn result.
36712         (stmt_interesting_for_vrp, vrp_visit_stmt): Handle
36713         IFN_ATOMIC_COMPARE_EXCHANGE.
36715 2017-03-10  David Malcolm  <dmalcolm@redhat.com>
36717         PR driver/79875
36718         * opts.c (parse_sanitizer_options): Add missing question mark to
36719         "did you mean" message.
36721 2017-03-10  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
36723         * config/rs6000/rs6000-builtin.def (VMULEUB_UNS): Remove orphaned
36724         built-in.
36725         (VMULEUH_UNS): Likewise.
36726         (VMULOUB_UNS): Likewise.
36727         (VMULOUH_UNS): Likewise.
36728         * config/rs6000/rs6000.c (builtin_function_type): Remove
36729         references to ALTIVEC_BUILTIN_VMUL[EO]U[BH]_UNS.
36731 2017-03-10  David Malcolm  <dmalcolm@redhat.com>
36733         PR bootstrap/79952
36734         * read-rtl-function.c (function_reader::read_rtx_operand): Update
36735         x with result of extra_parsing_for_operand_code_0.
36736         (function_reader::extra_parsing_for_operand_code_0): Convert
36737         return type from void to rtx, returning x.  When reading
36738         SYMBOL_REF with SYMBOL_FLAG_HAS_BLOCK_INFO, reallocate x to the
36739         larger size containing struct block_symbol.
36741 2017-03-10  Segher Boessenkool  <segher@kernel.crashing.org>
36743         * config/rs6000/rs6000.c (rs6000_option_override_internal): Disallow
36744         -mfloat128-hardware without -m64.
36746 2017-03-10  Will Schmidt <will_schmidt@vnet.ibm.com>
36748         PR target/79941
36749         * config/rs6000/rs6000.c (builtin_function_type): Add VMUL*U[HB]
36750         entries to the case statement that marks unsigned arguments to
36751         overloaded functions.
36753 2017-03-10  Kelvin Nilsen  <kelvin@gcc.gnu.org>
36755         * config/rs6000/rs6000.c (rs6000_option_override_internal): Fix
36756         two typographic errors in the handling of TARGET_UPPER_REGS_DI.
36758 2017-03-10  Pat Haugen  <pthaugen@us.ibm.com>
36760         PR target/79907
36761         * config/rs6000/rs6000.c (rs6000_init_hard_regno_mode_ok): Test
36762         TARGET_UPPER_REGS_DI when setting 'wi' constraint regclass.
36764 2017-03-10  Martin Liska  <mliska@suse.cz>
36766         PR target/65705
36767         PR target/69804
36768         * toplev.c (process_options): Enable MPX with LSAN and UBSAN.
36769         * tree-chkp.c (chkp_walk_pointer_assignments): Verify that
36770         FIELD != NULL.
36772 2017-03-10  Olivier Hainque  <hainque@adacore.com>
36774         * tree-switch-conversion (array_value_type): Start by resetting
36775         candidate type to it's main variant.
36777 2017-03-10  Jakub Jelinek  <jakub@redhat.com>
36779         PR rtl-optimization/79909
36780         * combine.c (try_combine): Use simplify_replace_rtx on individual
36781         CALL_INSN_FUNCTION_USAGE elements instead of replace_rtx on copy_rtx
36782         of the whole CALL_INSN_FUNCTION_USAGE.
36784         PR tree-optimization/79972
36785         * gimple-ssa-warn-alloca.c (alloca_call_type): Only call
36786         get_range_info on SSA_NAMEs.  Formatting fixes.
36788 2017-03-10  Richard Biener  <rguenther@suse.de>
36789             Jakub Jelinek  <jakub@redhat.com>
36791         PR tree-optimization/77975
36792         * tree-ssa-loop-niter.c (get_base_for): Allow phi argument from latch
36793         edge to be constant.
36794         (get_val_for): For constant x return it.  Formatting fix.
36795         (loop_niter_by_eval): Avoid pointless looping if the next iteration
36796         would use the same bases as the current one.
36798 2017-03-09  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
36800         * config/rs6000/rs6000.c (rs6000_gen_le_vsx_permute): Use rotate
36801         instead of vec_select for V1TImode.
36802         * conifg/rs6000/vsx.md (VSX_LE): Remove mode iterator that is no
36803         longer needed.
36804         (VSX_LE_128): Add V1TI to this mode iterator.
36805         (*vsx_le_perm_load_<mode>): Change to use VSX_D mode iterator.
36806         (*vsx_le_perm_store_<mode>): Likewise.
36807         (pre-reload splitter for VSX stores): Likewise.
36808         (post-reload splitter for VSX stores): Likewise.
36809         (*vsx_xxpermdi2_le_<mode>): Likewise.
36810         (*vsx_lxvd2x2_le_<mode>): Likewise.
36811         (*vsx_stxvd2x2_le_<mode>): Likewise.
36813 2017-03-09  Michael Eager  <eager@eagercon.com>
36815         Correct failures with --enable-checking=yes,rtl.
36817         * config/microblaze/microblaze.c (microblaze_expand_shift):
36818         Replace GET_CODE test with CONST_INT_P and INTVAL test with
36819         test for const0_rtx.
36820         * config/microblaze/microblaze.md (ashlsi3_byone, ashrsi3_byone,
36821         lshrsi3_byone): Replace INTVAL with test for const1_rtx.
36823 2017-03-09  Richard Biener  <rguenther@suse.de>
36825         PR tree-optimization/79977
36826         * graphite-scop-detection.c (scop_detection::merge_sese):
36827         Handle the case of extra exits to blocks dominating the entry.
36829 2017-03-09  Toma Tabacu  <toma.tabacu@imgtec.com>
36831         * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
36832         Document rdynamic.
36834 2017-03-09  Vladimir Makarov  <vmakarov@redhat.com>
36836         PR rtl-optimization/79949
36837         * lra-constraints.c (process_alt_operands): Check memory when
36838         trying to predict a cycle.  Print about the overall increase.
36840 2017-03-09  Richard Biener  <rguenther@suse.de>
36842         PR middle-end/79971
36843         * gimple-expr.c (useless_type_conversion_p): Preserve
36844         TYPE_SATURATING for fixed-point types.
36846 2017-03-09  Richard Biener  <rguenther@suse.de>
36848         PR ipa/79970
36849         * ipa-prop.c (ipa_modify_formal_parameters): Avoid changing
36850         alignment of BLKmode params.
36852 2017-03-09  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
36854         PR target/79913
36855         * config/aarch64/iterators.md (VALL_F16_NO_V2Q): New mode iterator.
36856         (VALL_NO_V2Q): Likewise.
36857         (VDQF_DF): Delete.
36858         * config/aarch64/aarch64-simd.md
36859         (aarch64_dup_lane_<vswap_width_name><mode>): Use VALL_F16_NO_V2Q
36860         iterator.
36861         (*aarch64_simd_vec_copy_lane_<vswap_width_name><mode>): Use
36862         VALL_NO_V2Q mode iterator.
36863         (*aarch64_vgetfmulx<mode>): Use VDQF iterator.
36865 2017-03-09  Martin Liska  <mliska@suse.cz>
36867         PR tree-optimization/79631
36868         * tree-chkp-opt.c (chkp_is_constant_addr): Call
36869         tree_int_cst_sign_bit just for INTEGER constants.
36871 2017-03-09  Martin Liska  <mliska@suse.cz>
36873         PR target/65705
36874         PR target/69804
36875         * toplev.c (process_options): Disable -fcheck-pointer-bounds with
36876         sanitizers.
36878 2017-03-09  Marek Polacek  <polacek@redhat.com>
36880         PR c++/79672
36881         * tree.c (inchash::add_expr): Handle TREE_VEC.
36883 2017-03-09  Martin Liska  <mliska@suse.cz>
36885         PR ipa/79764
36886         (chkp_narrow_size_and_offset): New function.
36887         (chkp_parse_array_and_component_ref): Support BIT_FIELD_REF.
36888         (void chkp_parse_bit_field_ref): New function.
36889         (chkp_make_addressed_object_bounds): Add case for BIT_FIELD_REF.
36890         (chkp_process_stmt): Use chkp_parse_bit_field_ref.
36892 2017-03-09  Martin Liska  <mliska@suse.cz>
36894         PR ipa/79761
36895         * tree-chkp.c (chkp_get_bound_for_parm): Get bounds for a param.
36896         (chkp_find_bounds_1): Remove gcc_unreachable.
36898 2017-03-09  Jakub Jelinek  <jakub@redhat.com>
36900         PR sanitizer/79944
36901         * asan.c (get_mem_refs_of_builtin_call): For BUILT_IN_ATOMIC* and
36902         BUILT_IN_SYNC*, determine the access type from the size suffix and
36903         always build a MEM_REF with that type.  Handle forgotten
36904         BUILT_IN_SYNC_FETCH_AND_NAND_16 and BUILT_IN_SYNC_NAND_AND_FETCH_16.
36906         PR target/79932
36907         * config/i386/avx512vlintrin.h (_mm256_cmpge_epi32_mask,
36908         _mm256_cmpge_epi64_mask, _mm256_cmpge_epu32_mask,
36909         _mm256_cmpge_epu64_mask, _mm256_cmple_epi32_mask,
36910         _mm256_cmple_epi64_mask, _mm256_cmple_epu32_mask,
36911         _mm256_cmple_epu64_mask, _mm256_cmplt_epi32_mask,
36912         _mm256_cmplt_epi64_mask, _mm256_cmplt_epu32_mask,
36913         _mm256_cmplt_epu64_mask, _mm256_cmpneq_epi32_mask,
36914         _mm256_cmpneq_epi64_mask, _mm256_cmpneq_epu32_mask,
36915         _mm256_cmpneq_epu64_mask, _mm256_mask_cmpge_epi32_mask,
36916         _mm256_mask_cmpge_epi64_mask, _mm256_mask_cmpge_epu32_mask,
36917         _mm256_mask_cmpge_epu64_mask, _mm256_mask_cmple_epi32_mask,
36918         _mm256_mask_cmple_epi64_mask, _mm256_mask_cmple_epu32_mask,
36919         _mm256_mask_cmple_epu64_mask, _mm256_mask_cmplt_epi32_mask,
36920         _mm256_mask_cmplt_epi64_mask, _mm256_mask_cmplt_epu32_mask,
36921         _mm256_mask_cmplt_epu64_mask, _mm256_mask_cmpneq_epi32_mask,
36922         _mm256_mask_cmpneq_epi64_mask, _mm256_mask_cmpneq_epu32_mask,
36923         _mm256_mask_cmpneq_epu64_mask, _mm_cmpge_epi32_mask,
36924         _mm_cmpge_epi64_mask, _mm_cmpge_epu32_mask, _mm_cmpge_epu64_mask,
36925         _mm_cmple_epi32_mask, _mm_cmple_epi64_mask, _mm_cmple_epu32_mask,
36926         _mm_cmple_epu64_mask, _mm_cmplt_epi32_mask, _mm_cmplt_epi64_mask,
36927         _mm_cmplt_epu32_mask, _mm_cmplt_epu64_mask, _mm_cmpneq_epi32_mask,
36928         _mm_cmpneq_epi64_mask, _mm_cmpneq_epu32_mask, _mm_cmpneq_epu64_mask,
36929         _mm_mask_cmpge_epi32_mask, _mm_mask_cmpge_epi64_mask,
36930         _mm_mask_cmpge_epu32_mask, _mm_mask_cmpge_epu64_mask,
36931         _mm_mask_cmple_epi32_mask, _mm_mask_cmple_epi64_mask,
36932         _mm_mask_cmple_epu32_mask, _mm_mask_cmple_epu64_mask,
36933         _mm_mask_cmplt_epi32_mask, _mm_mask_cmplt_epi64_mask,
36934         _mm_mask_cmplt_epu32_mask, _mm_mask_cmplt_epu64_mask,
36935         _mm_mask_cmpneq_epi32_mask, _mm_mask_cmpneq_epi64_mask,
36936         _mm_mask_cmpneq_epu32_mask, _mm_mask_cmpneq_epu64_mask): Move
36937         definitions outside of __OPTIMIZE__ guarded section.
36939         PR target/79932
36940         * config/i386/avx512bwintrin.h (_mm512_packs_epi32,
36941         _mm512_maskz_packs_epi32, _mm512_mask_packs_epi32,
36942         _mm512_packus_epi32, _mm512_maskz_packus_epi32,
36943         _mm512_mask_packus_epi32): Move definitions outside of __OPTIMIZE__
36944         guarded section.
36946 2017-03-09  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
36948         * config/s390/vx-builtins.md ("vfee<mode>", "vfeez<mode>")
36949         ("vfenez<mode>"): Add missing constraints.
36951 2017-03-08  Martin Sebor  <msebor@redhat.com>
36953         PR target/79928
36954         * config/nds32/nds32.c (nds32_option_override):
36955         Fix misspelled diagnostic.
36957 2017-03-08  Jakub Jelinek  <jakub@redhat.com>
36959         PR c/79940
36960         * gimplify.c (gimplify_omp_for): Replace index var in outer
36961         taskloop statement with an artificial variable and add
36962         OMP_CLAUSE_PRIVATE clause for it.
36964 2017-03-08  Richard Biener  <rguenther@suse.de>
36966         PR tree-optimization/79955
36967         * tree-ssa-uninit.c (warn_uninitialized_vars): Do not warn
36968         for accesses that are completely outside of the variable.
36970 2017-03-08  Andrew Haley  <aph@redhat.com>
36972         PR tree-optimization/79943
36973         * tree-ssa-loop-split.c (compute_new_first_bound): When
36974         calculating the new upper bound, (END-BEG) should be added, not
36975         subtracted.
36977 2017-03-08  Jakub Jelinek  <jakub@redhat.com>
36979         * config/avr/avr.md (setmemhi): Make sure match_dup
36980         operand number comes before match_scratch.
36982 2017-03-08  Richard Biener  <rguenther@suse.de>
36984         PR tree-optimization/79920
36985         * tree-vect-slp.c (vect_create_mask_and_perm): Remove and inline
36986         with ncopies == 1 to ...
36987         (vect_transform_slp_perm_load): ... here.  Properly compute
36988         all element loads by iterating VF times over the group.  Do
36989         not handle ncopies (computed in a broken way) in
36990         vect_create_mask_and_perm.
36992 2017-03-08  Jakub Jelinek  <jakub@redhat.com>
36994         PR sanitizer/79904
36995         * internal-fn.c (expand_vector_ubsan_overflow): If arg0 or arg1
36996         is a uniform vector, use uniform_vector_p return value instead of
36997         building ARRAY_REF on folded VIEW_CONVERT_EXPR to array type.
36999 2017-03-07  Marek Polacek  <polacek@redhat.com>
37001         PR middle-end/79809
37002         * gimple-ssa-warn-alloca.c (pass_walloca::gate): Use HOST_WIDE_INT.
37003         (alloca_call_type): Likewise.
37005 2017-03-07  Martin Liska  <mliska@suse.cz>
37007         * gcov.c (process_args): Put comment to correct location.
37009 2017-03-07  Martin Liska  <mliska@suse.cz>
37011         PR middle-end/68270
37012         * tree-chkp.c (chkp_may_narrow_to_field): Add new argument ref.
37013         Use array_at_struct_end_p instead of DECL_CHAIN (field).
37014         (chkp_narrow_bounds_for_field): Likewise.
37015         (chkp_parse_array_and_component_ref): Pass one more argument to
37016         call.
37018 2017-03-07  Richard Biener  <rguenther@suse.de>
37020         * tree-vect-loop-manip.c (slpeel_add_loop_guard): Preserve
37021         preheaders.
37023 2017-03-07  Segher Boessenkool  <segher@kernel.crashing.org>
37025         * config/i386/i386.c (ix86_local_alignment): Align most aggregates
37026         of 16 bytes and more to 16 bytes, not those of 16 bits and more.
37028 2017-03-07  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
37030         PR c/79855
37031         * params.def (PARAM_STORE_MERGING_ALLOW_UNALIGNED): Add full stop
37032         to end of description.
37033         (PARAM_MAX_STORES_TO_MERGE): Likewise.
37035 2017-03-07  Jakub Jelinek  <jakub@redhat.com>
37037         PR rtl-optimization/79901
37038         * config/i386/sse.md (*avx512bw_<code><mode>3<mask_name>): Renamed to
37039         ...
37040         (*avx512f_<code><mode>3<mask_name>): ... this.
37041         (<code><mode>3 with maxmin code iterator): Use VI8_AVX2_AVX512F
37042         iterator instead of VI8_AVX2_AVX512BW.
37044         PR rtl-optimization/79901
37045         * expr.c (expand_expr_real_2): For vector MIN/MAX, if there is no
37046         min/max expander, expand it using expand_vec_cond_expr.
37048         PR sanitizer/79897
37049         * ubsan.c (ubsan_encode_value): Call mark_addressable on the
37050         temporary.
37052 2017-03-06  Jakub Jelinek  <jakub@redhat.com>
37054         PR c++/79821
37055         * dwarf2out.h (dw_vec_const): Change array type from unsigned char *
37056         to void * for PCH reasons.
37057         * dwarf2out.c (output_loc_operands, output_die): Cast
37058         v.val_vec.array to unsigned char *.
37060 2017-03-06  John David Anglin  <danglin@gcc.gnu.org>
37062         PR target/77850
37063         * config/pa/pa-64.h (PAD_VARARGS_DOWN): Don't pad down complex and
37064         vector types.
37066 2017-03-06  Vladimir Makarov  <vmakarov@redhat.com>
37068         PR rtl-optimization/79571
37069         * lra-constraints.c (process_alt_operands): Calculate static
37070         reject and subtract it from overall when only addresses will be
37071         reloaded.
37073 2017-03-06  Julia Koval  <julia.koval@intel.com>
37075         PR target/79793
37076         * config/i386/i386.c (ix86_minimum_incoming_stack_boundary): Set
37077         incoming stack boundary to 128 for 64-bit targets.
37079 2017-03-06  Richard Biener  <rguenther@suse.de>
37081         PR tree-optimization/79894
37082         * tree-vectorizer.c (vectorize_loops): Set loop_vectorized_call
37083         to NULL after folding it.
37085 2017-03-06  Richard Biener  <rguenther@suse.de>
37087         PR tree-optimization/79824
37088         * tree-vect-stmts.c (get_group_load_store_type): Fix alignment
37089         check disabling peeling for gaps.
37091 2017-03-06  Toma Tabacu  <toma.tabacu@imgtec.com>
37093         * doc/sourcebuild.texi (Effective-Target Keywords, Environment
37094         attributes): Document gettimeofday.
37096 2017-03-06  Robin Dapp  <rdapp@linux.vnet.ibm.com>
37098         * config/s390/s390.c (s390_option_override_internal): Set
37099         PARAM_MIN_VECT_LOOP_BOUND
37101 2017-03-06  Robin Dapp  <rdapp@linux.vnet.ibm.com>
37103         * config/s390/s390.c (s390_asm_output_function_label): Use nopr %r0.
37104         * config/s390/s390.md: Likewise.
37106 2017-03-06  Jakub Jelinek  <jakub@redhat.com>
37108         PR target/79812
37109         * config/i386/sse.md (VI8F_256_512): Remove mode iterator.
37110         (<avx2_avx512>_perm<mode>): Rename to ...
37111         (avx2_perm<mode>): ... this.  Use VI8F_256 iterator instead
37112         of VI8F_256_512.
37113         (<avx512>_perm<mode>_mask): Rename to ...
37114         (avx512vl_perm<mode>_mask): ... this.  Use VI8F_256 iterator instead
37115         of VI8F_256_512.
37116         (<avx2_avx512>_perm<mode>_1<mask_name>): Rename to ...
37117         (avx2_perm<mode>_1<mask_name): ... this.  Use VI8F_256 iterator
37118         instead of VI8F_256_512.
37119         (avx512f_perm<mode>): New define_expand.
37120         (avx512f_perm<mode>_mask): Likewise.
37121         (avx512f_perm<mode>_1<mask_name>): New define_insn.
37122         (<avx512>_vec_dup<mode>_1): Fix up vec_select mode.
37124 2017-03-06  Prachi Godbole  <prachi.godbole@imgtec.com>
37126         * config/mips/mips-msa.md (msa_fmax_a_<msafmt>, msa_fmin_a_<msafmt>,
37127         msa_max_a_<msafmt>, msa_min_a_<msafmt>): Introduce mode interator for
37128         if_then_else.
37129         (smin<mode>3, smax<mode>3): Change operand print code from 'B' to 'E'.
37131 2017-03-06  Martin Liska  <mliska@suse.cz>
37133         PR sanitize/79783
37134         * asan.c (asan_expand_poison_ifn): Do not expand ASAN_POISON
37135         when having a SSA NAME w/o VAR_DECL assigned to it.
37137 2017-03-06  Prachi Godbole  <prachi.godbole@imgtec.com>
37139         * config/mips/mips-msa.md (msa_dotp_<su>_d, msa_dpadd_<su>_d,
37140         msa_dpsub_<su>_d): Fix MODE for vec_select.
37142 2017-03-06  Prachi Godbole  <prachi.godbole@imgtec.com>
37144         * config/mips/mips.c (mips_gen_const_int_vector): Change type of last
37145         argument.
37146         * config/mips/mips-protos.h (mips_gen_const_int_vector): Likewise.
37148 2017-03-06  Richard Biener  <rguenther@suse.de>
37150         * lto-streamer.c (lto_check_version): Use %qs in diagnostics.
37151         * plugin.c (register_plugin_info): Likewise.
37152         * tree-chkp.c (chkp_make_static_const_bounds): Likewise.
37154 2017-03-05  Jakub Jelinek  <jakub@redhat.com>
37156         * config/i386/sse.md (sse_storehps, sse_storelps,
37157         avx_<castmode><avxsizesuffix>_<castmode>,
37158         avx512f_<castmode><avxsizesuffix>_<castmode>,
37159         avx512f_<castmode><avxsizesuffix>_256<castmode>): Require
37160         in condition that at least one operand is not a MEM.
37162 2017-03-03  Jakub Jelinek  <jakub@redhat.com>
37164         PR middle-end/79805
37165         * internal-fn.def (ATOMIC_BIT_TEST_AND_SET, ATOMIC_BIT_TEST_AND_RESET,
37166         ATOMIC_BIT_TEST_AND_COMPLEMENT, ATOMIC_COMPARE_EXCHANGE): Remove
37167         ECF_NOTHROW.
37168         * gimple-fold.c (fold_builtin_atomic_compare_exchange): Set
37169         gimple_call_nothrow_p flag based on whether original builtin can throw.
37170         If it can, emit following stmts on the fallthrough edge.
37171         * tree-ssa-ccp.c (optimize_atomic_bit_test_and): Similarly, except
37172         don't create new bb if inserting just debug stmts on the edge, try to
37173         insert them on the fallthru bb or just reset debug stmts.
37175 2017-03-03  Segher Boesssenkool  <segher@kernel.crashing.org>
37177         PR target/43763
37178         * config/rs6000/rs6000.c (rs6000_final_prescan_insn): Save and
37179         restore recog_data (including the operand rtxes inside it) around
37180         the call to get_insn_template.
37182 2017-03-03  Martin Sebor  <msebor@redhat.com>
37184         PR tree-optimization/79699
37185         * context.c (context::~context): Free MPFR caches to avoid
37186         a memory leak on program exit.
37188 2017-03-03  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
37190         * config/aarch64/aarch64.c (aarch64_float_const_representable_p):
37191         Use wide_int::ulow () instead of .elt (0).
37193 2017-03-03  Uros Bizjak  <ubizjak@gmail.com>
37195         * config/i386/i386.md (*pushtf): Change *roF constraint to *roC.
37196         (*pushxf): Limit oF constraint to 32bit targets and add oC
37197         constraint for 64bit targets.
37198         (pushxf splitter): Use PUSH_ROUNDING to calculate stack adjustment.
37199         (*pushdf): Change rmF constraint to rmC.
37201 2017-03-03  Martin Liska  <mliska@suse.cz>
37203         * tree-ssa-loop-prefetch.c (pass_loop_prefetch::execute):
37204         Remove unused variable.
37206 2017-03-03  Jakub Jelinek  <jakub@redhat.com>
37208         PR target/79807
37209         * config/i386/i386.c (ix86_expand_multi_arg_builtin): If target
37210         is a memory operand, increase num_memory.
37211         (ix86_expand_args_builtin): Likewise.
37213 2017-03-03  Jan Hubicka  <jh@suse.cz>
37215         PR lto/79760
37216         * ipa-devirt.c (maybe_record_node): Properly handle
37217         __cxa_pure_virtual visibility.
37219 2017-03-03  Martin Liska  <mliska@suse.cz>
37221         PR tree-optimization/79803
37222         * tree-ssa-loop-prefetch.c (tree_ssa_prefetch_arrays): Remove
37223         assert.
37224         (pass_loop_prefetch::execute): Disabled optimization if an
37225         assumption about L1 cache size is not met.
37227 2017-03-03  Martin Liska  <mliska@suse.cz>
37229         PR rtl-optimization/79574
37230         * gcse.c (struct gcse_expr): Use HOST_WIDE_INT instead of int.
37231         (hash_scan_set): Likewise.
37232         (dump_hash_table): Likewise.
37233         (hoist_code): Likewise.
37235 2017-03-03  Richard Biener  <rguenther@suse.de>
37237         * fixed-value.c (fixed_from_string): Restore use of elt (1)
37238         in place of uhigh ().
37239         (fixed_convert_from_real): Likewise.
37241 2017-03-03  Uros Bizjak  <ubizjak@gmail.com>
37243         PR target/79514
37244         * config/i386/i386.md (*pushxf_rounded): Use Pmode instead of DImode.
37246 2017-03-03  Richard Biener  <rguenther@suse.de>
37248         PR middle-end/79818
37249         * match.pd ( X +- C1 CMP C2 -> X CMP C2 -+ C1): Add missing
37250         TYPE_OVERFLOW_UNDEFINED check.
37252 2017-03-02  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
37254         * config/rs6000/vector.md (vector_ne_<mode>_p): Correct operand
37255         numbers.
37256         (vector_ae_<mode>_p): Likewise.
37257         (vector_nez_<mode>_p): Likewise.
37258         (vector_ne_v2di_p): Likewise.
37259         (vector_ae_v2di_p): Likewise.
37260         (vector_ne_<mode>_p): Likewise.
37261         * config/rs6000/vsx.md (vsx_tsqrt<mode>2_fg): Correct operand
37262         numbers.
37263         (vsx_tsqrt<mode>2_fe): Likewise.
37265 2017-03-02  Uros Bizjak  <ubizjak@gmail.com>
37267         PR target/79514
37268         * config/i386/i386.md (*pushxf_rounded): New insn_and_split pattern.
37270 2017-03-02  Jakub Jelinek  <jakub@redhat.com>
37272         PR rtl-optimization/79780
37273         * cprop.c (one_cprop_pass): When second and further conditional trap
37274         in a single basic block is turned into an unconditional trap, turn it
37275         into a deleted note to avoid RTL verification failures.
37277 2017-03-02  Richard Biener  <rguenther@suse.de>
37279         * fold-const.c (const_binop): Use ulow () instead of elt (0).
37281 2017-03-02  Richard Biener  <rguenther@suse.de>
37283         PR tree-optimization/79345
37284         PR c++/42000
37285         * tree-ssa-alias.c (walk_aliased_vdefs_1): Take a limit
37286         param and abort the walk, returning -1 if it is hit.
37287         (walk_aliased_vdefs): Take a limit param and pass it on.
37288         * tree-ssa-alias.h (walk_aliased_vdefs): Add a limit param,
37289         defaulting to 0 and return a signed int.
37290         * tree-ssa-uninit.c (struct check_defs_data): New struct.
37291         (check_defs): New helper.
37292         (warn_uninitialized_vars): Use walk_aliased_vdefs to warn
37293         about uninitialized memory.
37294         * fixed-value.c (fixed_from_string): Use ulow/uhigh to avoid
37295         bogus uninitialized warning.
37296         (fixed_convert_from_real): Likewise.
37298 2017-03-02  Bin Cheng  <bin.cheng@arm.com>
37300         PR tree-optimization/66768
37301         * tree-ssa-loop-ivopts.c (find_interesting_uses_address): Skip addr
37302         iv_use if base object can't be determined.
37304 2017-03-02  Jakub Jelinek  <jakub@redhat.com>
37306         PR tree-optimization/79345
37307         * gensupport.h (struct pattern_stats): Add min_scratch_opno field.
37308         * gensupport.c (get_pattern_stats_1) <case MATCH_SCRATCH>: Update it.
37309         (get_pattern_stats): Initialize it.
37310         * genemit.c (gen_expand): Verify match_scratch numbers come after
37311         match_operand/match_dup numbers.
37312         * config/i386/i386.md (<s>mul<mode>3_highpart): Swap match_dup and
37313         match_scratch numbers.
37314         * config/i386/sse.md (avx2_gathersi<mode>, avx2_gatherdi<mode>):
37315         Likewise.
37316         * config/s390/s390.md (trunctdsd2): Likewise.
37318 2017-03-02  Richard Biener  <rguenther@suse.de>
37320         * wide-int.h (wide_int_storage::operator=): Implement in terms
37321         of wi::copy.
37323 2017-03-02  Richard Biener  <rguenther@suse.de>
37325         PR tree-optimization/79777
37326         * tree-ssa-pre.c (eliminate_insert): Give up if we simplify
37327         the to insert expression to sth existing.
37329 2017-03-01  Martin Sebor  <msebor@redhat.com>
37331         PR middle-end/79692
37332         * gimple-ssa-sprintf.c
37333         (directive::known_width_and_precision): New function.
37334         (format_integer): Use it.
37335         (get_mpfr_format_length): Consider the full range of precision
37336         when computing %g output with the # flag.  Set the likely byte
37337         count to 3 rather than 1 when precision is indeterminate.
37338         (format_floating): Correct the lower bound of precision.
37340 2017-03-01  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
37342         * doc/invoke.texi: Document default code model for 64-bit Linux.
37344 2017-03-01  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
37346         PR target/79752
37347         * config/rs6000/rs6000.md (peephole2 for udiv/umod): Should emit
37348         udiv rather than div since input pattern is unsigned.
37350 2017-03-01  Uros Bizjak  <ubizjak@gmail.com>
37352         * config/i386/i386.c (print_reg): Warn for values of
37353         unsupported size in integer register.
37355 2017-03-01  Michael Meissner  <meissner@linux.vnet.ibm.com>
37357         PR target/79439
37358         * config/rs6000/predicates.md (current_file_function_operand): Do
37359         not allow self calls to be local if the function is replaceable.
37361 2017-03-01  Kelvin Nilsen  <kelvin@gcc.gnu.org>
37363         PR target/79395
37364         * config/rs6000/altivec.h (vec_ctz and others): Change the
37365         preprocessor macro that controls conditional compilation from
37366         _ARCH_PWR9 to __POWER9_VECTOR__.
37367         (vec_all_ne): Change parameterization of __altivec_scalar_pred
37368         macro expansion under preprocessor #ifdef __POWER9_VECTOR__
37369         control (instead of _ARCH_PWR9 control) so that template
37370         definition uses power9-specific function.
37371         (vec_any_eq): Likewise.
37372         (vec_all_ne): Change macro definition to use a power9-specific
37373         expansion under #ifdef __POWER9_VECTOR__ control (instead of
37374         _ARCH_PWR9 control).
37375         (vec_any_eq) Likewise.
37376         * config/rs6000/rs6000-builtin.def (CMPNEF): Remove BU_P9V_AV_2
37377         expansion for CMPNEF to remove support for xvcmpnesp instruction.
37378         (CMPNED): Remove BU_P9V_AV2 expansion for CMPNED to remove
37379         support for xvcmpnedp instruction.
37380         (VCMPNEB_P): Replace BU_P9V_AV_P macro expansion with BU_P9V_AV_2
37381         macro expansion so that Power9 implementation of vec_all_ne does
37382         not use the AltiVec predicate framework.
37383         (VCMPNEH_P): Likewise.
37384         (VCMPNEW_P): Likewise.
37385         (VCMPNED_P): Likewise.
37386         (VCMPNEFP_P): Likewise.
37387         (VCMPNEDP_P): Likewise.
37388         (VCMPAEB_P): Add BU_P9V_AV_2 macro expansion to change
37389         implementation of vec_any_eq to not use AltiVec predicate
37390         framework.
37391         (VCMPAEH_P): Likewise.
37392         (VCMPAEW_P): Likewise.
37393         (VCMPAED_P): Likewise.
37394         (VCMPAEFP_P): Likewise.
37395         (VCMPAEDP_P): Likewise.
37396         (VCMPNE_P): Replace BU_P9V_OVERLOAD_P macro expansion with
37397         BU_P9V_OVERLOAD_2 so that Power9 implementation of vec_all_ne does
37398         not use the AltiVec predicate framework.
37399         (VCMPAE_P): Add BU_P9V_OVERLOAD_2 macro to change implementation
37400         of vec_any_eq to not use AltiVec predicate framework.
37401         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Add
37402         support for predefined __POWER9_VECTOR__ macro to indicate that
37403         Power9 instruction selection is enabled.
37404         (altivec_overloaded_builtins): Remove extraneous
37405         ALTIVEC_BUILTIN_VEC_CMPNE entry for overloaded
37406         function argument types RS6000_BTI_bool_V16QI and
37407         RS6000_BTI_bool_V16QI.  Remove erroneous ALTIVEC_BUILTIN_VEC_CMPNE
37408         entry for overloaded function argument types RS6000_BTI_bool_V4SI
37409         andRS6000_BTI_bool_V4SI, mapping to P9V_BUILTIN_CMPNEB.  Remove
37410         two entries mapping to P9V_BUITIN_CMPNED and one entry mapping to
37411         P9V_BUILTIN_CMPNEF to force use of instructions not specific to
37412         Power9 for implementations of vec_cmpne.  Change the signature for
37413         all definitions of the overloaded P9V_BUILTIN_VEC_CMPNE_P function
37414         (representing vec_all_ne) to remove the previously described first
37415         argument of type RS6000_BTI_INTSI, as this was an artifact of
37416         reliance on the AltiVec predicate framework, which is no longer
37417         used in the implementation of these functions.  Add
37418         P9V_BUILTIN_VEC_VCMPAE_P entries (representing the vec_anyeq
37419         function) to match all of the P9V_BUILTIN_VEC_VCMNE_P entries
37420         since, unlike the AltiVec predicate framework implementation, we
37421         do not share function descriptors between vec_alle and vec_anyeq.
37422         (altivec_resolve_overloaded_builtin): Add SFmode and DFmode to the
37423         set of modes that receive special treatment even when
37424         TARGET_P9_VECTOR is true.  The special treatment emits code that
37425         does not depend on Power9 instructions.
37426         * config/rs6000/vector.md (vector_ne_<mode>_p): Change this
37427         define_expand to not rely on AltiVec predicate framework.
37428         (vector_ae_<mode>p): New define_expand to represent vec_any_eq
37429         function.
37430         (vector_ne_v2di_p): Change this define_expand to not rely on
37431         AltiVec predicate framework.
37432         (vector_ae_v2di_p): New define_expand to represent vec_any_eq
37433         function.
37434         (vector_ne_<mode>_p): Change this define_expand to not rely on
37435         AltiVec predicate framework.
37436         (vector_ae_<mode>p): New define_expand to represent vec_any_eq
37437         function.
37438         * config/rs6000/vsx.md (*vsx_ne_<mode>_p): For modes VSX_EXTRACT_I
37439         (V16QI, V8HI, V4SI), correct a typo in the code emitted for this
37440         define_insn pattern.
37441         (*vsx_ne_<mode>_p): For modes VSX_F (V4SF and V2DF), remove this
37442         define_insn pattern because the xvcmpne<VSs>. instruction is not
37443         supported.
37444         (vcmpne<VSs>): Remove this define_insn because xvcmpne<VSs>
37445         instruction is not supported.
37447 2017-03-01  Jakub Jelinek  <jakub@redhat.com>
37449         * config/nvptx/nvptx.c: Include intl.h.
37451 2017-03-01  Martin Jambor  <mjambor@suse.cz>
37453         PR lto/78140
37454         * ipa-prop.h (ipa_bits): Removed field known.
37455         (ipa_jump_func): Removed field vr_known.  Changed fields bits and m_vr
37456         to pointers.  Adjusted their comments to warn about their sharing.
37457         (ipcp_transformation_summary): Change bits to a vector of pointers.
37458         (ipa_check_create_edge_args): Moved to ipa-prop.c, declare.
37459         (ipa_get_ipa_bits_for_value): Declare.
37460         * tree-vrp.h (value_range): Mark as GTY((for_user)).
37461         * ipa-prop.c (ipa_bit_ggc_hash_traits): New.
37462         (ipa_bits_hash_table): Likewise.
37463         (ipa_vr_ggc_hash_traits): Likewise.
37464         (ipa_vr_hash_table): Likewise.
37465         (ipa_print_node_jump_functions_for_edge): Adjust for bits and m_vr
37466         being pointers and vr_known being removed.
37467         (ipa_set_jf_unknown): Likewise.
37468         (ipa_get_ipa_bits_for_value): New function.
37469         (ipa_set_jfunc_bits): Likewise.
37470         (ipa_get_value_range): New overloaded functions.
37471         (ipa_set_jfunc_vr): Likewise.
37472         (ipa_compute_jump_functions_for_edge): Use the above functions to
37473         construct bits and vr parts of jump functions.
37474         (ipa_check_create_edge_args): Move here from ipa-prop.h, also allocate
37475         ipa_bits_hash_table and ipa_vr_hash_table if they do not already
37476         exist.
37477         (ipcp_grow_transformations_if_necessary): Also allocate
37478         ipa_bits_hash_table and ipa_vr_hash_table if they do not already
37479         exist.
37480         (ipa_node_params_t::duplicate): Do not copy bits, just pointers to
37481         them.  Fix too long lines.
37482         (ipa_write_jump_function): Adjust for bits and m_vr being pointers and
37483         vr_known being removed.
37484         (ipa_read_jump_function): Use new setter functions to construct bits
37485         and vr parts of jump functions or set them to NULL.
37486         (write_ipcp_transformation_info): Adjust for bits being pointers.
37487         (read_ipcp_transformation_info): Likewise.
37488         (ipcp_update_bits): Likewise.  Fix excessively long lines a trailing
37489         space.
37490         Include gt-ipa-prop.h.
37491         * ipa-cp.c (propagate_bits_across_jump_function): Adjust for bits
37492         being pointers.
37493         (ipcp_store_bits_results): Likewise.
37494         (propagate_vr_across_jump_function): Adjust for m_vr being a pointer.
37495         Do not write to existing jump functions but use a temporary instead.
37497 2017-03-01  Jakub Jelinek  <jakub@redhat.com>
37499         PR c++/79681
37500         * fold-const.c (make_bit_field_ref): If orig_inner is COMPONENT_REF,
37501         attempt to use its first operand as BIT_FIELD_REF base.
37503 2017-03-01  Richard Biener  <rguenther@suse.de>
37505         PR middle-end/79721
37506         * tree-chrec.c (chrec_evaluate): Perform computation of Newtons
37507         interpolating formula in wrapping arithmetic.
37508         (chrec_apply): Convert chrec_evaluate return value to wanted type.
37510 2017-03-01  Jakub Jelinek  <jakub@redhat.com>
37512         PR tree-optimization/79734
37513         * tree-vect-generic.c (expand_vector_condition): Optimize
37514         AVX512 vector boolean VEC_COND_EXPRs into bitwise operations.
37515         Handle VEC_COND_EXPR where comparison has different inner width from
37516         type's inner width.
37518 2017-02-28  Sandra Loosemore  <sandra@codesourcery.com>
37520         * doc/invoke.texi (ARC Options): Copy-edit to fix punctuation,
37521         markup, and similar issues.  Remove @opindex entries for things
37522         that aren't options.  Add missing -mmpy-option entries.
37524 2017-02-28  Jakub Jelinek  <jakub@redhat.com>
37526         PR tree-optimization/79737
37527         * gimple-ssa-store-merging.c (encode_tree_to_bitpos): If bitlen is
37528         a multiple of BITS_PER_UNIT and !BYTES_BIG_ENDIAN, clear
37529         tmpbuf[byte_size - 1].  Call natice_encode_expr with byte_size - 1
37530         instead of byte_size.  Formatting fix.
37531         (shift_bytes_in_array_right): Formatting fix.
37533 2017-02-28  Eric Botcazou  <ebotcazou@adacore.com>
37535         PR target/79749
37536         * config/sparc/sparc.c (sparc_frame_pointer_required): Add missing
37537         condition on optimize for the leaf function test.
37539 2017-02-28  Martin Liska  <mliska@suse.cz>
37541         PR lto/79625
37542         * read-rtl-function.c (function_reader::handle_unknown_directive):
37543         Bail out when one uses -flto.
37545 2017-02-28  Martin Liska  <mliska@suse.cz>
37547         * common.opt: Replace space with tabular for options of <number>
37548         type.
37549         * config/i386/i386.opt: Show <number> value for
37550         -mlarge-data-threshold.
37551         * opts.c (print_filtered_help): Do not display number in hexadecimal
37552         format.
37554 2017-02-28  Martin Liska  <mliska@suse.cz>
37556         * common.opt: Fix --help=option -Q for options which are of
37557         an enum type.
37559 2017-02-28  Uros Bizjak  <ubizjak@gmail.com>
37561         * config/i386/i386.c (print_reg): Error out for values
37562         of 8-bit size in invalid integer register.
37564 2017-02-28  Martin Sebor  <msebor@redhat.com>
37566         PR tree-optimization/79691
37567         * passes.def (pass_all_optimizations_g): Enable pass_sprintf_length.
37569 2017-02-28  Jakub Jelinek  <jakub@redhat.com>
37571         PR target/79729
37572         * config/i386/i386.c (ix86_print_operand) <case 'R'>: Replace
37573         gcc_unreachable with output_operand_lossage.
37575 2017-02-28  Richard Biener  <rguenther@suse.de>
37577         PR tree-optimization/79740
37578         * tree-ssa-sccvn.c (vn_nary_op_insert_into): Allow redundant
37579         inserts.
37580         (visit_nary_op): Insert the nary into the hashtable if we
37581         pattern-matched sth.
37582         * tree-ssa-pre.c (eliminate_insert): Robustify.
37584 2017-02-28  Richard Biener  <rguenther@suse.de>
37586         PR middle-end/79731
37587         * fold-const.c (decode_field_reference): Reject out-of-bound
37588         accesses.
37590 2017-02-28  Jakub Jelinek  <jakub@redhat.com>
37592         * config/i386/i386.c: Include intl.h.
37593         (ix86_option_override_internal): Use cond ? G_("...") : G_("...")
37594         instead of just cond ? "..." : "...".
37595         * config/nvptx/nvptx.c (nvptx_goacc_validate_dims): Likewise.
37596         * coverage.c (read_counts_file): Likewise.
37597         * omp-offload.c: Include intl.h.
37598         (oacc_loop_fixed_partitions): Use cond ? G_("...") : G_("...") instead
37599         of just cond ? "..." : "...".
37600         * gcov.c (read_count_file): Use cond ? N_("...") : N_("...") instead
37601         of just cond ? "..." : "...".
37603 2017-02-28  Richard Earnshaw  <rearnsha@arm.com>
37605         PR target/79742
37606         * config/arm/parsecpu.awk (gen_data): Set tuning target to 'tune for'
37607         entry, if present.
37608         * config/arm/arm-cpus.in (cortex-m0plus.small-multiply): Correct
37609         'tune for' CPU name.
37610         * config/arm/arm-cpu-data.h: Regenerated.
37612 2017-02-28  Richard Biener  <rguenther@suse.de>
37614         PR tree-optimization/79732
37615         * tree-inline.c (expand_call_inline): Do not shadow var.
37617 2017-02-28  Richard Biener  <rguenther@suse.de>
37619         PR tree-optimization/79723
37620         * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Preserve
37621         address-space properly.
37623 2017-02-28  Thomas Schwinge  <thomas@codesourcery.com>
37625         * doc/optinfo.texi (Optimization groups): Fix option used for
37626         OPTGROUP_ALL.
37627         * doc/invoke.texi (-fopt-info): Document "omp".
37628         * dumpfile.h: Sort OPTGROUP_OMP before OPTGROUP_VEC.
37629         (OPTGROUP_ALL): Add OPTGROUP_OMP.
37630         * hsa-gen.c (pass_data_gen_hsail): Use OPTGROUP_OMP.
37631         * ipa-hsa.c (pass_data_ipa_hsa): Likewise.
37632         * omp-simd-clone.c (pass_data_omp_simd_clone): Likewise.
37634         * dumpfile.h (OPTGROUP_OPENMP): Rename to OPTGROUP_OMP.  Adjust
37635         all users.
37636         * dumpfile.c (optgroup_options): Instead of "openmp", associate
37637         OPTGROUP_OMP with "omp".
37639 2017-02-27  Pat Haugen  <pthaugen@us.ibm.com>
37641         PR target/79544
37642         * config/rs6000/rs6000-c.c (struct altivec_builtin_types): Use VSRAD
37643         for arithmetic shift of unsigned V2DI.
37645 2017-02-27  Claudiu Zissulescu  <claziss@synopsys.com>
37647         * config.gcc (arc*-): Clean up, use arc/big.h, arc/elf.h, and
37648         arc/linux.h headers.
37649         * config/arc/arc.h (TARGET_OS_CPP_BUILTINS): Remove.
37650         (LINK_SPEC): Likewise.
37651         (ARC_TLS_EXTRA_START_SPEC): Likewise.
37652         (EXTRA_SPECS): Likewise.
37653         (STARTFILE_SPEC): Likewise.
37654         (ENDFILE_SPEC): Likewise.
37655         (LIB_SPEC): Likewise.
37656         (TARGET_SDATA_DEFAULT): Likewise.
37657         (TARGET_MMEDIUM_CALLS_DEFAULT): Likewise.
37658         (MULTILIB_DEFAULTS): Likewise.
37659         (DWARF2_UNWIND_INFO): Likewise.
37660         * config/arc/big.h: New file.
37661         * config/arc/elf.h: Likewise.
37662         * config/arc/linux.h: Likewise.
37663         * config/arc/t-uClibc: Remove.
37665 2017-02-27  Bin Cheng  <bin.cheng@arm.com>
37667         PR tree-optimization/77536
37668         * tree-ssa-loop-manip.c (niter_for_unrolled_loop): New function.
37669         (tree_transform_and_unroll_loop): Use above function to compute the
37670         estimated niter of unrolled loop and use it when scaling profile.
37671         Also use count info rather than frequency if it's non-zero.
37672         * tree-ssa-loop-manip.h niter_for_unrolled_loop(): New declaration.
37673         * tree-vect-loop.c (scale_profile_for_vect_loop): New function.
37674         (vect_transform_loop): Call above function.
37676 2017-02-27  Richard Biener  <rguenther@suse.de>
37678         PR tree-optimization/45397
37679         * tree-ssa-pre.c (eliminate_insert): Handle BIT_AND_EXPR.
37680         * tree-ssa-sccvn.c (valueized_wider_op): New helper.
37681         (visit_nary_op): Add pattern matching for CSEing sign-changed
37682         or truncated operations with wider ones.
37684 2017-02-27  Richard Biener  <rguenther@suse.de>
37686         PR tree-optimization/79690
37687         * tree-vect-stmts.c (vectorizable_store): Use vector type
37688         built from the DR with address-space.
37690 2017-02-26  Gerald Pfeifer  <gerald@pfeifer.com>
37692         * doc/invoke.texi (Optimize Options): Refine the description
37693         of asan-use-after-return.
37695 2017-02-25  Alan Modra  <amodra@gmail.com>
37697         PR rtl-optimization/79584
37698         * lra-constraints.c (base_to_reg): Reload ad->base, the entire
37699         base, not ad->base_term, the reg within base.  Remove assertion
37700         that ad->base == ad->base_term.  Replace gen_int_mode using
37701         bogus mode with const0_rtx.
37703 2017-02-25  Jakub Jelinek  <jakub@redhat.com>
37705         PR middle-end/79396
37706         * tree-eh.c (operation_could_trap_p, stmt_could_throw_1_p): Handle
37707         FMA_EXPR like tcc_binary or tcc_unary.
37709         * tree-ssa-loop-niter.c (number_of_iterations_exit): Simplify warning.
37711         PR debug/77589
37712         * dwarf2out.c (struct dw_loc_list_struct): Add noted_variable_value
37713         bitfield.
37714         (size_of_loc_descr): Handle DW_OP_GNU_variable_value.
37715         (output_loc_operands): Handle DW_OP_call_ref and
37716         DW_OP_GNU_variable_value.
37717         (struct variable_value_struct): New type.
37718         (struct variable_value_hasher): Likewise.
37719         (variable_value_hash): New variable.
37720         (string_types): Remove.
37721         (copy_loc_descr): New function.
37722         (add_loc_descr_to_each): Clarify comment.  Use copy_loc_descr.
37723         (prepend_loc_descr_to_each): New function.
37724         (add_loc_list): Fix comment typo.  Use prepend_loc_descr_to_each
37725         instead of add_loc_descr_to_each if the first argument is single
37726         location list and the second has multiple.
37727         (resolve_args_picking_1): Handle DW_OP_GNU_variable_value.
37728         (loc_list_from_tree_1): For early_dwarf, emit DW_OP_GNU_variable_value
37729         when looking for variable value which doesn't have other location info.
37730         (loc_list_from_tree): Formatting fix.
37731         (gen_array_type_die): Simplify DW_AT_string_length handling.
37732         (adjust_string_types): Remove.
37733         (gen_subprogram_die): Don't call adjust_string_types nor test/set
37734         string_types.  Call resolve_variable_values.
37735         (prune_unused_types_walk_loc_descr): Handle DW_OP_GNU_variable_value.
37736         (resolve_addr_in_expr): Likewise.  Add A argument.
37737         (copy_deref_exprloc): Remove deref argument.  Adjust for the
37738         original expression being DW_OP_GNU_variable_value with optionally
37739         DW_OP_stack_value after it instead of DW_OP_call4 with DW_OP_deref
37740         optionally after it.
37741         (optimize_string_length): Rework for DW_OP_GNU_variable_value.
37742         (resolve_addr): Adjust optimize_string_length and resolve_addr_in_expr
37743         callers.  Set remove_AT_byte_size if removing DW_AT_string_length.
37744         (variable_value_hasher::hash, variable_value_hasher::equal): New
37745         methods.
37746         (resolve_variable_value_in_expr, resolve_variable_value,
37747         resolve_variable_values, note_variable_value_in_expr,
37748         note_variable_value): New functions.
37749         (dwarf2out_early_finish): Call note_variable_value on all toplevel
37750         DIEs.
37752 2017-02-24  Jakub Jelinek  <jakub@redhat.com>
37754         PR c/79677
37755         * opts.h (handle_generated_option): Add GENERATED_P argument.
37756         * opts-common.c (handle_option): Adjust function comment.
37757         (handle_generated_option): Add GENERATED_P argument, pass it to
37758         handle_option.
37759         (control_warning_option): Pass false to handle_generated_option
37760         GENERATED_P.
37761         * opts.c (maybe_default_option): Pass true to handle_generated_option
37762         GENERATED_P.
37763         * optc-gen.awk: Likewise.
37765 2017-02-24  Segher Boessenkool  <segher@kernel.crashing.org>
37767         * config/sh/sh.md (tstsi_t): If operands[0] is a SUBREG instead of
37768         a REG, look at the REG it is a SUBREG of.
37769         (splitter for cmpeqsi_t): Ditto.
37771 2017-02-24  Segher Boessenkool  <segher@kernel.crashing.org>
37773         * config/pa/pa.c (pa_combine_instructions): Do not share RTL.  Make
37774         the special USEs with the pattern of the insn, not the insn itself.
37776 2017-02-24  Matthew Fortune  <matthew.fortune@imgtec.com>
37778         PR target/79473
37779         * doc/invoke.texi: Document -mload-store-pairs.
37781 2017-02-24  Segher Boessenkool  <segher@kernel.crashing.org>
37782             Sandra Loosemore  <sandra@codesourcery.com>
37784         * config/nios2/nios2.c (nios2_simple_const_p): Returns false if the
37785         argument isn't a CONST_INT.
37786         (nios2_alternate_compare_const): Assert op is a CONST_INT.
37787         (nios2_valid_compare_const_p): Assert op is a CONST_INT.
37788         (nios2_validate_compare): Bypass alternate compare logic if *op2
37789         is not a CONST_INT.
37790         (ldstwm_operation_p): Return false if first_base is not a REG or
37791         if first_offset is not a CONST_INT.
37793 2017-02-24  Segher Boessenkool  <segher@kernel.crashing.org>
37795         * config/cris/cris.md: Use correct operand in a define_peephole2.
37797 2017-02-24  Segher Boessenkool  <segher@kernel.crashing.org>
37799         * config/c6x/c6x.c (predicate_insn): Do not incorrectly share RTL.
37801 2017-02-24  Segher Boessenkool  <segher@kernel.crashing.org>
37803         * config/arc/arc.c (arc_ccfsm_advance): Only take the PATTERN of
37804         this_insn if it is an INSN or JUMP_INSN.
37805         (force_offsettable): Look at base, not at addr.
37806         * config/arc/predicates.md (brcc_nolimm_operator): Don't call INTVAL
37807         on things that aren't necessarily CONST_INTs.
37809 2017-02-24  Uros Bizjak  <ubizjak@gmail.com>
37811         * doc/invoke.texi (x86 Options, -mfpmath=sse): Mention that
37812         -mfpmath=sse is the default also for x86-32 targets with SSE2
37813         instruction set when @option{-ffast-math} is enabled
37815 2017-02-24  Jeff Law  <law@redhat.com>
37817         PR rtl-optimizatoin/79286
37818         * ira.c (update_equiv_regs): Drop may_trap_p exception to
37819         dominance test.
37821 2017-02-24  Richard Biener  <rguenther@suse.de>
37823         PR tree-optimization/79389
37824         * gimple-ssa-split-paths.c (is_feasible_trace): Properly skip
37825         debug insns.
37827 2017-02-24  Aldy Hernandez  <aldyh@redhat.com>
37829         * tree-ssa-loop-niter.c (number_of_iterations_exit): Update
37830         function comment to reflect reality.
37831         (loop_exits_before_overflow): Fix typo in function description.
37833 2017-02-24  Richard Biener  <rguenther@suse.de>
37835         PR tree-optimization/79389
37836         * gimple-ssa-split-paths.c (is_feasible_trace): Verify more
37837         properly that a threading opportunity exists.  Detect conditional
37838         copy/constant propagation opportunities.
37840 2017-02-23  Eric Botcazou  <ebotcazou@adacore.com>
37842         * config/visium/visium.md (type): Add trap.
37843         (b): New mode attribute.
37844         (*btst): Rename into...
37845         (*btst<mode>): ...this and adjust.
37846         (*cbranchsi4_btst_insn): Rename into...
37847         (*cbranch<mode>4_btst_insn): ...this and adjust.
37848         (trap): New define_insn.
37850 2017-02-23  Jakub Jelinek  <jakub@redhat.com>
37852         PR tree-optimization/79389
37853         * ifcvt.c (struct noce_if_info): Add rev_cond field.
37854         (noce_reversed_cond_code): New function.
37855         (noce_emit_store_flag): Use rev_cond if non-NULL instead of
37856         reversed_comparison_code.  Formatting fix.
37857         (noce_try_store_flag): Test rev_cond != NULL in addition to
37858         reversed_comparison_code.
37859         (noce_try_store_flag_constants): Likewise.
37860         (noce_try_store_flag_mask): Likewise.
37861         (noce_try_addcc): Use rev_cond if non-NULL instead of
37862         reversed_comparison_code.
37863         (noce_try_cmove_arith): Likewise.  Formatting fixes.
37864         (noce_try_minmax, noce_try_abs): Clear rev_cond.
37865         (noce_find_if_block): Initialize rev_cond.
37866         (find_cond_trap): Call noce_get_condition with then_bb == trap_bb
37867         instead of false as last argument never attempt to reverse it
37868         afterwards.
37870 2017-02-23  Bin Cheng  <bin.cheng@arm.com>
37872         PR tree-optimization/79663
37873         * tree-predcom.c (combine_chains): Process refs in reverse order
37874         only for ZERO length chains, and add explaining comment.
37876 2017-02-23  Jeff Law  <law@redhat.com>
37878         PR tree-optimization/79578
37879         * tree-ssa-dse.c (clear_bytes_written_by): Use OEP_ADDRESS_OF
37880         in call to operand_equal_p.
37882 2017-01-23  Dominique d'Humieres  <dominiq@lps.ens.fr>
37884         PR target/71017
37885         * config/i386/cpuid.h: Fix another undefined behavior.
37887 2017-02-23  Richard Biener  <rguenther@suse.de>
37889         PR tree-optimization/79683
37890         * tree-vect-stmts.c (vect_analyze_stmt): Do not overwrite
37891         vector types for data-refs.
37893 2017-02-23  Martin Liska  <mliska@suse.cz>
37895         * params.def (PARAM_MIN_NONDEBUG_INSN_UID): Change default to 0.
37897 2017-02-23  Jakub Jelinek  <jakub@redhat.com>
37899         PR middle-end/79665
37900         * internal-fn.c (get_range_pos_neg): Moved to ...
37901         * tree.c (get_range_pos_neg): ... here.  No longer static.
37902         * tree.h (get_range_pos_neg): New prototype.
37903         * expr.c (expand_expr_real_2) <case TRUNC_DIV_EXPR>: If both arguments
37904         are known to be in between 0 and signed maximum inclusive, try to
37905         expand both unsigned and signed divmod and use the cheaper one from
37906         those.
37908 2017-02-22  Jeff Law  <law@redhat.com>
37910         PR tree-optimization/79578
37911         * tree-ssa-dse.c (clear_bytes_written_by): Use operand_equal_p
37912         to compare base operands.
37914 2017-02-22  Segher Boessenkool  <segher@kernel.crashing.org>
37916         PR target/79211
37917         * config/rs6000/rs6000.md (*fsel<SFDF:mode><SFDF2:mode>4): Use
37918         gpc_reg_operand instead of fpr_reg_operand.
37920 2017-02-22  Sameera Deshpande  <sameera.deshpande@imgtec.com>
37922         * config/mips/mips.c (mips_return_in_memory): Force FP
37923         vector types to be returned in memory for o32 ABI.
37925 2017-02-22  Jakub Jelinek  <jakub@redhat.com>
37927         * dwarf2out.c (gen_variable_die): For -gdwarf-5, use DW_TAG_variable
37928         instead of DW_TAG_member for static data member declarations and don't
37929         set no_linkage_name for static inline data members.
37930         (gen_member_die): For -gdwarf-5 don't change DW_TAG_variable
37931         to DW_TAG_member.
37933 2017-02-22  Martin Liska  <mliska@suse.cz>
37935         * doc/invoke.texi: Replace inequality signs with square brackets
37936         for -Wnormalized.
37938 2017-02-22  Matthew Fortune  <matthew.fortune@imgtec.com>
37940         PR target/78660
37941         * lra-constraints.c (simplify_operand_subreg): Handle
37942         WORD_REGISTER_OPERATIONS targets.
37944 2017-02-22  Jakub Jelinek  <jakub@redhat.com>
37946         PR target/70465
37947         * reg-stack.c (emit_swap_insn): Treat (float_extend:?F (mem:?F))
37948         and (const_double:?F) like (mem:?F) for the purpose of fxch %st(1)
37949         elimination by swapping fld*.
37951 2017-02-22  Richard Biener  <rguenther@suse.de>
37953         PR tree-optimization/79673
37954         * tree-ssa-pre.c (compute_avail): Use wide_int_to_tree to
37955         convert the [TARGET_]MEM_REF offset INTEGER_CST, scrapping off
37956         irrelevant address-space qualifiers and avoiding a
37957         ADDR_SPACE_CONVERT_EXPR from fold_convert.
37959 2017-02-22  Richard Biener  <rguenther@suse.de>
37961         PR tree-optimization/79666
37962         * tree-vrp.c (extract_range_from_binary_expr_1): Make sure
37963         to not symbolically negate if that may introduce undefined
37964         overflow.
37966 2017-02-22  Martin Liska  <mliska@suse.cz>
37968         PR lto/79587
37969         * data-streamer-in.c (streamer_read_gcov_count): Remove assert.
37970         * data-streamer-out.c (streamer_write_gcov_count_stream):
37971         Likewise.
37972         * value-prof.c (stream_out_histogram_value): Make assert more
37973         precise based on type of counter.
37975 2017-02-21  Uros Bizjak  <ubizjak@gmail.com>
37977         PR target/79593
37978         * config/i386/i386.md (standard_x87sse_constant_load splitter):
37979         Use nonimmediate_operand instead of memory_operand for operand 1.
37980         (float-extend standard_x87sse_constant_load splitter): Ditto.
37982 2017-02-21  Jeff Law  <law@redhat.com>
37984         PR tree-optimization/79621
37985         * gimple-ssa-isolate-paths.c (find_implicit_erroneous_behavior): Ignore
37986         blocks with edges to themselves.
37988 2017-02-21  Jakub Jelinek  <jakub@redhat.com>
37990         PR target/79633
37991         * tree-chkp-opt.c (chkp_optimize_string_function_calls): Use
37992         is_gimple_call instead of comparing gimple_code with GIMPLE_CALL.
37993         Use gimple_call_builtin_p.
37995         PR target/79570
37996         * sel-sched.c (moveup_expr_cached): Don't call sel_bb_head
37997         on temporarily removed DEBUG_INSNs.
37999         PR tree-optimization/79649
38000         * tree-loop-distribution.c (classify_partition): Give up on
38001         non-generic address space loads/stores.
38003 2017-02-21  Aldy Hernandez  <aldyh@redhat.com>
38005         * doc/loop.texi (Loop manipulation): Remove nonexistent
38006         tree_ssa_loop_version from the documentation.
38007         * cfgloopmanip.c (loop_version): Document CONDITION_BB argument.
38009 2017-02-21  Jakub Jelinek  <jakub@redhat.com>
38011         PR target/79494
38012         * config/i386/i386.c (ix86_expand_split_stack_prologue): Call
38013         make_reg_eh_region_note_nothrow_nononlocal on call_insn.
38014         * config/rs6000/rs6000.c: Include except.h.
38015         (rs6000_expand_split_stack_prologue): Call
38016         make_reg_eh_region_note_nothrow_nononlocal on the call insn.
38018 2017-02-21  Martin Jambor  <mjambor@suse.cz>
38020         PR lto/79579
38021         * ipa-prop.c (ipa_prop_write_jump_functions): Bail out if no edges
38022         have been analyzed.
38024 2017-02-21  Martin Jambor  <mjambor@suse.cz>
38026         * common.opt (-fipa-cp-alignment): Mark as ignored and preserved
38027         for backward compatibility only.
38028         * doc/invoke.texi (Option Summary): Remove all references to
38029         -fipa-cp-alignment.
38031 2017-02-21  Matthew Fortune  <matthew.fortune@imgtec.com>
38033         PR target/78660
38034         Revert:
38035         2017-02-20  Matthew Fortune  <matthew.fortune@imgtec.com>
38037         * lra-constraints.c (curr_insn_transform): Handle
38038         WORD_REGISTER_OPERATIONS requirements when reloading SUBREGs.
38040 2017-02-21  Martin Liska  <mliska@suse.cz>
38042         * config/i386/i386.opt: Replace -masm-dialect with -masm.
38044 2017-02-21  Thomas Schwinge  <thomas@codesourcery.com>
38046         PR translation/79638
38047         * config/nvptx/nvptx.c (ENTRY_TEMPLATE): Single out "%ntid.y".
38049 2017-02-21  Eric Botcazou  <ebotcazou@adacore.com>
38051         PR ada/67205
38052         * config/arm/arm.c (TARGET_CUSTOM_FUNCTION_DESCRIPTORS): Define.
38053         (arm_function_ok_for_sibcall): Return false for an indirect call by
38054         descriptor if all the argument registers are used.
38055         (arm_relayout_function): Use FUNCTION_ALIGNMENT macro to adjust the
38056         alignment of the function.
38058 2017-02-21  Jakub Jelinek  <jakub@redhat.com>
38060         PR tree-optimization/61441
38061         * simplify-rtx.c (simplify_const_unary_operation): For
38062         -fsignaling-nans and sNaN operand, return NULL_RTX rather than
38063         the sNaN unmodified.
38065 2017-02-20  Bernd Edlinger  <bernd.edlinger@hotmail.de>
38067         * Makefile.in (BUILD_SYSTEM_HEADER_DIR): New make variabe.
38068         (LIMITS_H_TEST, if_multiarch, stmp-fixinc): Use BUILD_SYSTEM_HEADER_DIR
38069         instead of SYSTEM_HEADER_DIR.
38071 2017-02-20  Gerald Pfeifer  <gerald@pfeifer.com>
38072             Martin LiÅ¡ka  <mliska@suse.cz>
38074         * doc/invoke.texi (use-after-scope-direct-emission-threshold):
38075         Fix typos and grammar, use active voice, and clarify.
38077 2017-02-20  Marek Polacek  <polacek@redhat.com>
38079         PR middle-end/79537
38080         * gimplify.c (gimplify_expr): Handle unused *&&L;.
38082         PR sanitizer/79558
38083         * ubsan.c (ubsan_type_descriptor): Check if TYPE_MAX_VALUE is null.
38085 2017-02-20  Jakub Jelinek  <jakub@redhat.com>
38087         PR target/79568
38088         * config/i386/i386.c (ix86_expand_builtin): Handle
38089         OPTION_MASK_ISA_AVX512VL and OPTION_MASK_ISA_64BIT in
38090         ix86_builtins_isa[fcode].isa as a requirement of those
38091         flags and any other flag in the bitmask.
38092         (ix86_init_mmx_sse_builtins): Use 0 instead of
38093         ~OPTION_MASK_ISA_64BIT as mask.
38094         * config/i386/i386-builtin.def (__builtin_ia32_rdtsc,
38095         __builtin_ia32_rdtscp, __builtin_ia32_pause, __builtin_ia32_bsrsi,
38096         __builtin_ia32_rdpmc, __builtin_ia32_rolqi, __builtin_ia32_rolhi,
38097         __builtin_ia32_rorqi, __builtin_ia32_rorhi): Likewise.
38099 2017-02-20  Matthew Fortune  <matthew.fortune@imgtec.com>
38101         PR target/78012
38102         * lra-constraints.c (split_reg): Check requested split mode
38103         is supported by the register.
38105 2017-02-20  Matthew Fortune  <matthew.fortune@imgtec.com>
38107         * lra-constraints.c (simplify_operand_subreg): Remove early
38108         return false.
38110 2017-02-20  Matthew Fortune  <matthew.fortune@imgtec.com>
38112         PR target/78660
38113         * lra-constraints.c (curr_insn_transform): Tighten condition
38114         for converting SUBREG reloads from OP_OUT to OP_INOUT.
38116 2017-02-20  Matthew Fortune  <matthew.fortune@imgtec.com>
38118         PR target/78660
38119         * lra-constraints.c (curr_insn_transform): Handle
38120         WORD_REGISTER_OPERATIONS requirements when reloading SUBREGs.
38122 2017-02-19  Uros Bizjak  <ubizjak@gmail.com>
38124         Revert:
38125         2016-05-30  Uros Bizjak  <ubizjak@gmail.com>
38127         * config/i386/sync.md (mfence_nosse): Use "lock orl $0, -4(%esp)".
38129 2017-02-19  Jonathan Wakely  <jwakely@redhat.com>
38131         PR c++/69523
38132         * doc/invoke.texi (C++ Dialect Options) [-Wliteral-suffix]: Update
38133         description.
38135 2017-02-19  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
38137         * gimple-pretty-print.c (dump_ternary_rhs): Adjust gimple dump format
38138         for FMA_EXPR.
38140 2017-02-18  Jakub Jelinek  <jakub@redhat.com>
38142         * final.c (last_columnnum, override_columnnum): New variables.
38143         (final_start_function): Set last_columnnum, pass it to begin_prologue
38144         hook and pass 0 to dwarf2out_begin_prologue.
38145         (final_scan_insn): Update override_columnnum.  Pass last_columnnum
38146         to source_line debug hook.
38147         (notice_source_line): Compute last_columnnum and for debug_column_info
38148         return true on column changes.
38149         * debug.h (struct gcc_debug_hooks): Add column argument to
38150         source_line and begin_prologue hooks.
38151         (debug_nothing_int_charstar_int_bool): Remove prototype.
38152         (debug_nothing_int_int_charstar,
38153         debug_nothing_int_int_charstar_int_bool): New prototypes.
38154         (dwarf2out_begin_prologue): Add column argument.
38155         * debug.c (do_nothing_debug_hooks): Adjust source_line and
38156         begin_prologue hooks.
38157         (debug_nothing_int_charstar_int_bool): Remove.
38158         (debug_nothing_int_int_charstar,
38159         debug_nothing_int_int_charstar_int_bool): New functions.
38160         * dwarf2out.c (dwarf2out_begin_prologue): Add column argument, pass it
38161         through to dwarf2out_source_line.
38162         (dwarf2_lineno_debug_hooks): Adjust begin_prologue hook.
38163         (dwarf2out_source_line): Add column argument, emit it if requested.
38164         * sdbout.c (sdbout_source_line, sdbout_begin_prologue): Add column
38165         arguments.
38166         * xcoffout.h (xcoffout_begin_prologue, xcoffout_source_line): Likewise.
38167         * xcoffout.c (xcoffout_begin_prologue, xcoffout_source_line): Likewise.
38168         * vmsdbgout.c (vmsdbgout_begin_prologue): Add column argument, pass it
38169         through to dwarf2out_begin_prologue.
38170         (vmsdbgout_source_line): Add column argument, pass it through to
38171         dwarf2out_source_line.
38172         * dbxout.c (dbxout_begin_prologue): Add column argument, adjust
38173         dbxout_source_line caller.
38174         (dbxout_source_line): Add column argument.
38176         * common.opt (gno-column-info, gcolumn-info): New options.
38177         * dwarf2out.c (dwarf2_lineno_debug_hooks): Formatting fix.
38178         (check_die): Also test for multiple DW_AT_decl_column attributes.
38179         (add_src_coords_attributes, dwarf2out_imported_module_or_decl_1): Add
38180         DW_AT_decl_column if requested.
38181         (gen_subprogram_die): Compare and/or add also DW_AT_decl_column
38182         if requested.
38183         (gen_variable_die): Likewise.
38184         (add_call_src_coords_attributes): Add DW_AT_call_column if requested.
38185         * doc/invoke.texi (-gcolumn-info, -gno-column-info): Document.
38187         PR target/79569
38188         * config/i386/i386.opt (m3dnowa): Replace Undocumented with Report.
38189         * common/config/i386/i386-common.c (OPTION_MASK_ISA_3DNOW_A_SET): Define.
38190         (ix86_handle_option): Handle OPT_m3dnowa.
38191         * doc/invoke.texi (-m3dnowa): Document.
38192         * doc/extend.texi (__builtin_ia32_pmulhuw, __builtin_ia32_pf2iw): Use
38193         -m3dnowa instead of -m3dnow -march=athlon.
38195         PR target/79559
38196         * config/i386/i386.c (ix86_print_operand): Use output_operand_lossage
38197         instead of gcc_assert for K, r and R code checks.  Formatting fixes.
38199 2017-02-17  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
38201         PR target/79261
38202         * config/rs6000/rs6000.c (rs6000_expand_ternop_builtin): Add
38203         support for CODE_FOR_vsx_xxpermdi_v2d[fi]_be.
38204         * config/rs6000/rs6000.md (reload_gpr_from_vsx<mode>): Call
38205         generator for vsx_xxpermdi_<mode>_be.
38206         * config/rs6000/vsx.md (vsx_xxpermdi_<mode>): Remove logic to
38207         force big-endian semantics.
38208         (vsx_xxpermdi_<mode>_be): New define_expand with same
38209         implementation as previous version of vsx_xxpermdi_<mode>.
38211 2017-02-17  Jakub Jelinek  <jakub@redhat.com>
38213         PR tree-optimization/79327
38214         * gimple-ssa-sprintf.c (format_integer): Remove likely_adjust
38215         variable, its initialization and use.
38217 2017-02-17  Julia Koval  <julia.koval@intel.com>
38219         * common/config/i386/i386-common.c (OPTION_MASK_ISA_RDPID_SET): New.
38220         (OPTION_MASK_ISA_PKU_UNSET): New.
38221         (ix86_handle_option): Handle -mrdpid.
38222         * config/i386/cpuid.h (bit_RDPID): New.
38223         * config/i386/driver-i386.c (host_detect_local_cpu):
38224         Detect RDPID feature.
38225         * config/i386/i386-builtin.def (__builtin_ia32_rdpid): New.
38226         * config/i386/i386-c.c (ix86_target_macros_internal):
38227         Handle RDPID flag.
38228         * config/i386/i386.c (ix86_target_string): Add -mrdpid to isa2_opts.
38229         (ix86_valid_target_attribute_inner_p): Add "rdpid".
38230         (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
38231         * config/i386/i386.h (TARGET_RDPID, TARGET_RDPID_P): New.
38232         * config/i386/i386.md (define_insn "rdpid"): New.
38233         * config/i386/i386.opt Add -mrdpid.
38234         * config/i386/immintrin.h (_rdpid_u32): New.
38236 2017-02-17  Vladimir Makarov  <vmakarov@redhat.com>
38238         PR rtl-optimization/79541
38239         * lra-constraints.c (curr_insn_transform): Remove wrong asm insn
38240         instead of transforming it into USE.
38242 2017-02-17  Segher Boessenkool  <segher@kernel.crashing.org>
38244         * config/rs6000/rs6000.md (extendsfdf2): Remove default arguments.
38245         If HONOR_SNANS (SFmode) force the input to a register.
38246         (*extendsfdf2_fpr): Add !HONOR_SNANS (SFmode) condition.
38247         (*extendsfdf2_snan): New pattern, used when using SNaNs; it generates
38248         an frsp or similar insn.
38250 2017-02-17  Martin Liska  <mliska@suse.cz>
38252         PR rtl-optimization/79577
38253         * params.def (selsched-max-sched-times): Increase minimum to 1.
38255 2017-02-17  Martin Liska  <mliska@suse.cz>
38257         PR rtl-optimization/79574
38258         * gcse.c (want_to_gcse_p): Prevent integer overflow.
38260 2017-02-17  Martin Liska  <mliska@suse.cz>
38262         PR tree-optimization/79529
38263         * tree-ssa-loop-unswitch.c (is_maybe_undefined): Use
38264         ssa_defined_default_def_p to handle cases which are implicitly
38265         defined.
38266         * tree-ssa.c (ssa_defined_default_def_p): New function.
38267         (ssa_undefined_value_p): Use ssa_defined_default_def_p to handle cases
38268         which are implicitly defined.
38269         * tree-ssa.h (ssa_defined_default_def_p): Declare.
38271 2017-02-17  Richard Biener  <rguenther@suse.de>
38273         PR middle-end/79576
38274         * params.def (max-ssa-name-query-depth): Limit to 10.
38276 2017-02-17  Richard Biener  <rguenther@suse.de>
38278         PR tree-optimization/79552
38279         * tree-ssa-structalias.c (visit_loadstore): Properly verify
38280         default defs.
38282 2017-02-17  Richard Biener  <rguenther@suse.de>
38284         PR bootstrap/79567
38285         * genmatch.c (output_line_directive): Handle DIR_SEPARATOR_2.
38287 2017-02-17  Marek Polacek  <polacek@redhat.com>
38289         PR middle-end/79536
38290         * fold-const.c (fold_negate_expr_1): Renamed from fold_negate_expr.
38291         (fold_negate_expr): New wrapper.
38293 2017-02-16  Sandra Loosemore  <sandra@codesourcery.com>
38295         * doc/invoke.texi (C++ Dialect Options) [-Wno-non-template-friend]: 
38296         Correct terminology and de-emphasize pre-standard behavior.
38298 2017-02-16  Alan Modra  <amodra@gmail.com>
38300         PR rtl-optimization/79286
38301         * ira.c (def_dominates_uses): New function.
38302         (update_equiv_regs): Don't create an equivalence for insns that
38303         may trap where the register def does not dominate the use.
38305 2017-02-16  Vladimir Makarov  <vmakarov@redhat.com>
38307         PR rtl-optimization/78127
38308         * lra.c (lra): Call lra_eliminate before finish the loop after
38309         lra_constraint.
38311 2017-02-16  Richard Biener  <rguenther@suse.de>
38313         * graphite.h: Do not include isl/isl_val_gmp.h, instead include
38314         isl/isl_val.h.
38315         * graphite-isl-ast-to-gimple.c (gmp_cst_to_tree): Remove.
38316         (gcc_expression_from_isl_expr_int): Use generic isl_val interface.
38317         * graphite-sese-to-poly.c: Do not include isl/isl_val_gmp.h.
38318         (isl_val_int_from_wi): New function.
38319         (extract_affine_gmp): Rename to ...
38320         (extract_affine_wi): ... this, take a widest_int.
38321         (extract_affine_int): Just wrap extract_affine_wi.
38322         (add_param_constraints): Use isl_val_int_from_wi.
38323         (add_loop_constraints): Likewise, and extract_affine_wi.
38325 2017-02-15  Jeff Law  <law@redhat.com>
38327         PR middle-end/79521
38328         * ira-costs.c (scan_one_insn): Check have_regs_of_mode before calling
38329         ira_init_register_move_cost_if_necessary.
38331 2017-02-15  Martin Sebor  <msebor@redhat.com>
38333         PR middle-end/32003
38334         * doc/invoke.texi (-fdump-final-insns): Replace option accidentally
38335         removed in a prior commit.
38337 2017-02-15  Bin Cheng  <bin.cheng@arm.com>
38339         PR tree-optimization/79347
38340         * tree-vect-loop-manip.c (vect_do_peeling): Maintain profile
38341         counters during peeling.
38343 2017-02-15  Thomas Schwinge  <thomas@codesourcery.com>
38345         * Makefile.in (site.exp): Remove "set ISLVER".
38347 2017-02-15  Jakub Jelinek  <jakub@redhat.com>
38349         PR target/79487
38350         * real.c (real_from_integer): Call real_convert even for decimal.
38352 2017-02-15  Dominik Vogt  <vogt@linux.vnet.ibm.com>
38354         PR target/79421
38355         * config/s390/s390.c: define TARGET_CUSTOM_FUNCTION_DESCRIPTORS.
38357 2017-02-14  Andrew Pinski  <apinski@cavium.com>
38359         * config/aarch64/aarch64-cores.def (thunderx2t99): Move to under 'C"
38360         cores and change the partno/implementer to be correct.
38361         (thunderx2t99p1): New core which replaces thunderx2t99 and still has
38362         the 'B" as the implementer.
38363         * config/aarch64/aarch64-tune.md: Regenerate.
38365 2017-02-14  Carl Love  <cel@us.ibm.com>
38367         * config/rs6000/rs6000.c: Add case statement entry to make the
38368         xvcvuxdsp built-in argument unsigned.
38369         * config/rs6000/vsx.md: Fix the source and return operand types so they
38370         match the instruction definitions from the ISA document.  Fix typo
38371         in the instruction generation for the (define_insn "vsx_xvcvuxdsp"
38372         statement.
38374 2017-02-14  Vladimir Makarov  <vmakarov@redhat.com>
38376         PR target/79282
38377         * lra-int.h (struct lra_operand_data, struct lra_insn_reg): Add
38378         member early_clobber_alts.
38379         * lra-lives.c (reg_early_clobber_p): New.
38380         (process_bb_lives): Use it.
38381         * lra.c (new_insn_reg): New arg early_clobber_alts.  Use it.
38382         (debug_operand_data): Initialize early_clobber_alts.
38383         (setup_operand_alternative): Set up early_clobber_alts.
38384         (collect_non_operand_hard_regs): Ditto.  Pass early clobber
38385         alternatives to new_insn_reg.
38386         (add_regs_to_insn_regno_info): Add arg early_clobber_alts.  Use
38387         it.
38388         (lra_update_insn_regno_info): Pass the new arg.
38390 2017-02-14  Jakub Jelinek  <jakub@redhat.com>
38392         PR middle-end/79505
38393         * omp-offload.c (free_oacc_loop): Release loop->ifns vector.
38394         (new_oacc_loop_raw): Don't clear already cleared fields.
38396         PR target/79481
38397         * config/i386/avx512pfintrin.h (_mm512_prefetch_i32gather_pd,
38398         _mm512_prefetch_i32gather_ps, _mm512_prefetch_i64gather_pd,
38399         _mm512_prefetch_i64gather_ps): New inline functions and macros.
38401 2017-02-14  Uros Bizjak  <ubizjak@gmail.com>
38403         PR target/79495
38404         * config/i386/i386.md (*movxf_internal): Add (o,rC) alternative.
38406 2017-02-14  H.J. Lu  <hongjiu.lu@intel.com>
38408         PR target/79498
38409         * config/i386/i386.c (timode_scalar_chain::convert_insn): Insert
38410         the extra instruction to the right place to store 128-bit constant
38411         when needed.
38413 2017-02-14  Martin Sebor  <msebor@redhat.com>
38415         PR middle-end/79448
38416         * gimple-ssa-sprintf.c (format_directive): Avoid issuing INT_MAX
38417           warning for strings of unknown length.
38419 2017-02-13  Segher Boessenkool  <segher@kernel.crashing.org>
38421         * config.gcc (supported_defaults) [powerpc*-*-*]: Update.
38423 2017-02-14  Jeff Law  <law@redhat.com>
38425         PR target/79404
38426         * ira-costs.c (scan_one_insn): Initialize register move costs
38427         for pseudos seen in USE/CLOBBER insns.
38429         PR tree-optimization/79095
38430         * tree-vrp.c (extract_range_from_binary_expr_1): For EXACT_DIV_EXPR,
38431         if the numerator has the range ~[0,0] make the resultant range ~[0,0].
38432         (extract_range_from_binary_expr): For MINUS_EXPR with no derived range,
38433         if the operands are known to be not equal, then the resulting range
38434         is ~[0,0].
38435         (intersect_ranges): If the new range is ~[0,0] and the old range is
38436         wide, then prefer ~[0,0].
38437         * tree-vrp.c (overflow_comparison_p_1): New function.
38438         (overflow_comparison_p): New function.
38439         * tree-vrp.c (register_edge_assert_for_2): Register additional asserts
38440         if NAME is used in an overflow test.
38441         (vrp_evaluate_conditional_warnv_with_ops): If the ops represent an
38442         overflow check that can be expressed as an equality test, then adjust
38443         ops to be that equality test.
38445 2017-02-14  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
38447         * config/s390/s390-builtin-types.def: Remove flags argument.
38448         * config/s390/s390.c (s390_init_builtins): Likewise.
38450 2017-02-14  Martin Liska  <mliska@suse.cz>
38452         * tree-ssa-loop-unswitch.c (hoist_guard): Release get_loop_body
38453         vector.  Fix trailing white spaces.
38455 2017-02-14  James Greenhalgh  <james.greenhalgh@arm.com>
38457         * config/aarch64/aarch64.c (aarch64_simd_container_mode): Handle
38458         HFmode.
38460 2017-02-14  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
38462         PR rtl-optimization/68664
38463         * config/arm/arm.c (arm_sched_can_speculate_insn):
38464         New function.  Declare prototype.
38465         (TARGET_SCHED_CAN_SPECULATE_INSN): Define.
38467 2017-02-14  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
38469         PR rtl-optimization/68664
38470         * config/aarch64/aarch64.c (aarch64_sched_can_speculate_insn):
38471         New function.
38472         (TARGET_SCHED_CAN_SPECULATE_INSN): Define.
38474 2017-02-14  Amit Pawar  <amit.pawar@amd.com>
38476         * config/i386/i386.c (znver1_cost): Fix the alignment for function and
38477         max skip bytes for function, loop and jump.
38479 2017-02-14  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
38481         * gimple-pretty-print.c (dump_unary_rhs): Adjust dump format for
38482         ABS_EXPR for gimple dump.
38484 2017-02-14  Jakub Jelinek  <jakub@redhat.com>
38486         PR target/79462
38487         * config/sh/sh.c (expand_cbranchdi4): Don't clear operands[4].
38489         PR tree-optimization/79408
38490         * tree-vrp.c (simplify_div_or_mod_using_ranges): Handle also the
38491         case when on TRUNC_MOD_EXPR op0 is INTEGER_CST.
38492         (simplify_stmt_using_ranges): Call simplify_div_or_mod_using_ranges
38493         also if rhs1 is INTEGER_CST.
38495 2017-02-14  Richard Biener  <rguenther@suse.de>
38497         PR middle-end/79432
38498         * tree-into-ssa.c (insert_phi_nodes): When the function can
38499         have abnormal edges rewrite SSA names with broken use-def
38500         dominance out of SSA and register them for PHI insertion.
38502 2017-02-13  Martin Sebor  <msebor@redhat.com>
38504         PR middle-end/79496
38505         * gimple-ssa-sprintf.c (pass_sprintf_length::handle_gimple_call): Avoid
38506         clearing info.nowrite flag when snprintf size argument is a range.
38508 2017-02-13  Jakub Jelinek  <jakub@redhat.com>
38510         * cprop.c (cprop_jump): Add missing space in string literal.
38511         * tree-ssa-structalias.c (rewrite_constraints): Likewise.
38512         (get_constraint_for_component_ref): Likewise.
38513         * df-core.c (df_worklist_dataflow_doublequeue): Likewise.
38514         * tree-outof-ssa.c (insert_partition_copy_on_edge): Likewise.
38515         * lra-constraints.c (process_alt_operands): Likewise.
38516         * ipa-inline.c (inline_small_functions): Likewise.
38517         * tree-ssa-sccvn.c (visit_reference_op_store): Likewise.
38518         * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
38519         * trans-mem.c (diagnose_tm_1_op): Likewise.
38520         * omp-grid.c (grid_find_single_omp_among_assignments): Likewise.
38521         (grid_parallel_clauses_gridifiable): Likewise.
38523         * config/nvptx/mkoffload.c (process): Add space in between
38524         , and %d.
38526         * config/i386/i386.h (REG_CLASS_NAMES): Add , in between
38527         "MOD4_SSE_REGS" and "ALL_REGS".
38529         * spellcheck.c (test_data): Add , in between "foo" and "food".
38531 2017-02-13  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
38533         PR target/79449
38534         * config/rs6000/rs6000.c (expand_block_compare): Make sure runtime
38535         boundary crossing check and subsequent code generation agree.
38537 2017-02-13  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
38539         * config/aarch64/aarch64.c (has_memory_op): Delete.
38540         (aarch64_madd_needs_nop): Use contains_mem_rtx_p instead of
38541         has_memory_op.
38543 2017-02-13  Jakub Jelinek  <jakub@redhat.com>
38545         PR rtl-optimization/79388
38546         PR rtl-optimization/79450
38547         * combine.c (distribute_notes): When removing TEM_INSN for which
38548         corresponding dest has last value recorded, invalidate that last
38549         value.
38551 2017-02-13  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
38553         * config/arm/arm.c (arm_print_tune_info): Use ASM_COMMENT_START instead
38554         of explicit '@'.  Add missing assembly comment marker on branch costs
38555         printout.
38557 2017-02-13  Nathan Sidwell  <nathan@acm.org>
38559         * gengtype-lex.l (<in_struct>): Add '/'.
38561 2017-02-13  Martin Liska  <mliska@suse.cz>
38563         PR c/79471
38564         * calls.c (expand_call): Replace XALLOCAVEC with XCNEWVEC.
38566 2017-02-13  Richard Biener  <rguenther@suse.de>
38568         * configure.ac (HAVE_ISL_OPTIONS_SET_SCHEDULE_SERIALIZE_SCCS):
38569         Remove.
38570         * configure: Re-generate.
38571         * config.in: Likewise.
38572         * graphite-dependences.c: Simplify as if
38573         HAVE_ISL_OPTIONS_SET_SCHEDULE_SERIALIZE_SCCS was defined.
38574         * graphite-isl-ast-to-gimple.c: Likewise.
38575         * graphite-optimize-isl.c: Likewise.
38576         * graphite-poly.c: Likewise.
38577         * graphite-sese-to-poly.c: Likewise.
38578         * graphite.h: Likewise.
38579         * toplev.c: Include isl/version.h and use isl_version () for
38580         printing the ISL version.
38581         * doc/install.texi: Update ISL requirement.
38583 2017-02-12  Gerald Pfeifer  <gerald@pfeifer.com>
38585         * doc/standards.texi (Standards): Update reference to
38586         Objective-C 2.0.
38588 2017-02-12  Gerald Pfeifer  <gerald@pfeifer.com>
38590         * doc/extend.texi (Named Address Spaces): sourceware.org now
38591         defaults to https.
38592         * doc/install.texi (Binaries): Ditto.
38593         (Specific): Ditto.
38595 2017-02-11  Sandra Loosemore  <sandra@codesourcery.com>
38597         * doc/cpp.texi: Replace "stringify"/"stringification" with C
38598         standard terminology "stringize"/"stringizing" throughout.
38599         * doc/cppinternals.texi: Likewise.
38601 2017-02-11  Sandra Loosemore  <sandra@codesourcery.com>
38603         * doc/extend.texi: Fix some spelling mistakes and typos.
38604         * doc/invoke.texi: Likewise.
38606 2017-02-11  Jan Hubicka  <hubicka@ucw.cz>
38608         PR ipa/79224
38609         * params.def (inline-min-speedup) Change from 10 to 8.
38611 2017-02-11  Jakub Jelinek  <jakub@redhat.com>
38613         * doc/invoke.texi (fopenmp): Bump OpenMP version from 4.0 to
38614         4.5.
38616 2017-02-11  Jan Hubicka  <hubicka@ucw.cz>
38618         PR ipa/79224
38619         * ipa-inline-analysis.c (get_minimal_bb): New function.
38620         (record_modified): Use it.
38621         (remap_edge_change_prob): Handle also ancestor functions.
38623 2017-02-11  Gerald Pfeifer  <gerald@pfeifer.com>
38625         * doc/contrib.texi (Contributors): Remove broken link into
38626         the Mauve CVS repository.
38628 2017-02-11  Jakub Jelinek  <jakub@redhat.com>
38630         PR middle-end/79454
38631         * internal-fn.c (expand_vector_ubsan_overflow): Use piece-wise
38632         result computation whenever lhs doesn't have vector mode, not
38633         just when it has BLKmode.
38635 2017-02-10  Gerald Pfeifer  <gerald@pfeifer.com>
38637         * doc/makefile.texi (profiledbootstrap): Refer to the
38638         installation instructions only in textual form.
38640 2017-02-10  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
38642         PR target/79295
38643         * config/rs6000/altivec.md (bcd<bcd_add_sub>): Fix constraints.
38645 2017-02-10  Gerald Pfeifer  <gerald@pfeifer.com>
38647         * doc/install.texi (Specific): Use https for blackfin.uclinux.org.
38648         (Specific): Update mingw-w64 reference.
38649         (Binaries): Ditto.
38650         (Specific): Remove broken link to Renesas RX processor.
38652 2017-02-10  Richard Biener  <rguenther@suse.de>
38654         * toplev.c (process_options): Do not mention obsolete graphite
38655         options when printing sorry message about missing graphite support.
38656         Mention -floop-nest-optimize.
38658 2017-02-10  Christophe Lyon  <christophe.lyon@linaro.org>
38660         * config/aarch64/arm_neon.h (vtst_p8): Rewrite without asm.
38661         (vtst_p16): Likewise.
38662         (vtstq_p8): Likewise.
38663         (vtstq_p16): Likewise.
38664         (vtst_p64): New.
38665         (vtstq_p64): Likewise.
38666         * config/arm/arm_neon.h (vgetq_lane_p64): New.
38667         (vset_lane_p64): New.
38668         (vsetq_lane_p64): New.
38670 2017-02-10  Jakub Jelinek  <jakub@redhat.com>
38672         PR tree-optimization/79411
38673         * tree-ssa-reassoc.c (is_reassociable_op): Return false if
38674         stmt operands are SSA_NAMEs used in abnormal phis.
38675         (can_reassociate_p): Return false if op is SSA_NAME used in abnormal
38676         phis.
38678 2017-02-09  Jan Hubicka  <hubicka@ucw.cz>
38680         PR ipa/70795
38681         * cgraphunit.c (cgraph_node::add_new_function): Set externally_visible
38682         flag if needed.
38684 2017-02-09  Jan Hubicka  <hubicka@ucw.cz>
38686         * tree-ssa-loop-unswitch.c (hoist_guard): Update profile.
38688 2017-02-09  Jakub Jelinek  <jakub@redhat.com>
38690         * omp-offload.c (oacc_loop_auto_partitions): Use || instead of |
38691         to avoid warning.
38693         PR c/79413
38694         * gimplify.h (is_gimple_sizepos): Only test for INTEGER_CST constants,
38695         not arbitrary TREE_CONSTANT.
38697         PR c/79431
38698         * gimplify.c (gimplify_adjust_omp_clauses): Ignore
38699         "omp declare target link" attribute unless is_global_var.
38700         * omp-offload.c (find_link_var_op): Likewise.
38702 2017-02-09  Nathan Sidwell  <nathan@codesourcery.com>
38703             Chung-Lin Tang  <cltang@codesourcery.com>
38705         * gimplify.c (gimplify_scan_omp_clauses): No special handling for
38706         OMP_CLAUSE_TILE.
38707         (gimplify_adjust_omp_clauses): Don't delete TILE.
38708         (gimplify_omp_for): Deal with TILE.
38709         * internal-fn.c (expand_GOACC_TILE): New function.
38710         * internal-fn.def (GOACC_DIM_POS): Comment may be overly conservative.
38711         (GOACC_TILE): New.
38712         * omp-expand.c (struct oacc_collapse): Add tile and outer fields.
38713         (expand_oacc_collapse_init): Add LOC paramter.  Initialize tile
38714         element fields.
38715         (expand_oacc_collapse_vars): Add INNER parm, adjust for tiling,
38716         avoid DIV for outermost collapse var.
38717         (expand_oacc_for): Insert tile element loop as needed.  Adjust.
38718         Remove out of date comments, fix whitespace.
38719         * omp-general.c (omp_extract_for_data): Deal with tiling.
38720         * omp-general.h (enum oacc_loop_flags): Add OLF_TILE flag,
38721         adjust OLF_DIM_BASE value.
38722         (struct omp_for_data): Add tiling field.
38723         * omp-low.c (scan_sharing_clauses): Allow OMP_CLAUSE_TILE.
38724         (lower_oacc_head_mark): Add OLF_TILE as appropriate.  Ensure 2 levels
38725         for auto loops.  Remove default auto determining, moved to
38726         oacc_loop_fixed_partitions.
38727         * omp-offload.c (struct oacc_loop): Change 'ifns' to vector of call
38728         stmts, add e_mask field.
38729         (oacc_dim_call): New function, abstracted out from oacc_thread_numbers.
38730         (oacc_thread_numbers): Use oacc_dim_call.
38731         (oacc_xform_tile): New.
38732         (new_oacc_loop_raw): Initialize e_mask, adjust for ifns vector.
38733         (finish_oacc_loop): Adjust for ifns vector.
38734         (oacc_loop_discover_walk): Append loop abstraction sites to list,
38735         add case for GOACC_TILE fns.
38736         (oacc_loop_xform_loop): Delete.
38737         (oacc_loop_process): Iterate over call list directly, and add
38738         handling for GOACC_TILE fns.
38739         (oacc_loop_fixed_partitions): Determine default auto, deal with TILE,
38740         dump partitioning.
38741         (oacc_loop_auto_partitions): Add outer_assign parm. Assign all but
38742         vector partitioning to outer loops.  Assign 2 partitions to loops
38743         when available. Add TILE handling.
38744         (oacc_loop_partition): Adjust oacc_loop_auto_partitions call.
38745         (execite_oacc_device_lower): Process GOACC_TILE fns,
38746         ignore unknown specs.
38747         * tree-nested.c (convert_nonlocal_omp_clauses): Allow OMP_CLAUSE_TILE.
38748         * tree.c (omp_clause_num_ops): Adjust TILE ops.
38749         * tree.h (OMP_CLAUSE_TILE_ITERVAR, OMP_CLAUSE_TILE_COUNT): New.
38751 2017-02-09  Gerald Pfeifer  <gerald@pfeifer.com>
38753         * configure.ac (ACX_BUGURL): Update.
38754         * configure: Regenerate.
38756 2017-02-09  Richard Biener  <rguenther@suse.de>
38758         PR tree-optimization/69823
38759         * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
38760         Properly enumerate all BBs in the region.  Use auto_vec/auto_bitmap.
38762 2017-02-09  Andrew Burgess  <andrew.burgess@embecosm.com>
38764         * config/arc/arc-c.def: Add __NPS400__ definition.
38765         * config/arc/arc.h (CPP_SPEC): Don't define __NPS400__ here.
38766         (TARGET_NPS400): Define.
38768 2017-02-09  Andrew Burgess  <andrew.burgess@embecosm.com>
38770         * config/arc/arc-arch.h (arc_arch_t): Move unchanged to earlier in
38771         file.
38772         (arc_cpu_t): Change base_architecture field, arch, to a arc_arc_t
38773         pointer, arch_info.
38774         (arc_cpu_types): Fill the arch_info field with a pointer into the
38775         arc_arch_types table.
38776         (arc_selected_cpu): Declare.
38777         * config/arc/arc.c (arc_selected_cpu): Make global.
38778         (arc_selected_arch): Delete.
38779         (arc_base_cpu): Delete.
38780         (arc_override_options): Remove references to deleted variables,
38781         update access to arch information.
38782         (ARC_OPT): Update access to arch information.
38783         (ARC_OPTX): Likewise.
38784         * config/arc/arc.h (arc_base_cpu): Remove declaration.
38785         (TARGET_ARC600): Update access to arch information.
38786         (TARGET_ARC601): Likewise.
38787         (TARGET_ARC700): Likewise.
38788         (TARGET_EM): Likewise.
38789         (TARGET_HS): Likewise.
38790         * config/arc/driver-arc.c (arc_cpu_to_as): Update access to arch
38791         information.
38793 2017-02-08  Pat Haugen  <pthaugen@us.ibm.com>
38795         PR target/78604
38796         * config/rs6000/rs6000.c (rs6000_emit_vector_cond_expr): Invert
38797         condition/operands for integer GE/LE/GEU/LEU operations.
38799 2017-02-08  Segher Boessenkool  <segher@kernel.crashing.org>
38801         PR translation/79397
38802         * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Fix spelling
38803         of AltiVec.
38805 2017-02-08  Martin Jambor  <mjambor@suse.cz>
38807         PR ipa/79375
38808         * ipa-prop.c (ipa_alloc_node_params): Make static, return bool
38809         whether allocation happened.
38810         (ipa_initialize_node_params): Do not call ipa_alloc_node_params if
38811         nothing was allocated.
38813 2017-02-08  Jakub Jelinek  <jakub@redhat.com>
38815         PR tree-optimization/79408
38816         * tree-vrp.c (simplify_div_or_mod_using_ranges): If op1 is not
38817         constant, but SSA_NAME with a known integer range, use the minimum
38818         of that range instead of op1 to determine if modulo can be replaced
38819         with its first operand.
38821 2016-02-08  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
38823         * config/riscv/riscv.c (riscv_build_integer_1): Avoid use of INT16_MAX.
38825 2017-02-08  Richard Biener  <rguenther@suse.de>
38827         PR tree-optimization/71824
38828         * graphite-scop-detection.c (scop_detection::build_scop_breadth):
38829         Check all loops contained in the merged region.
38831 2017-02-07  Andrew Pinski  <apinski@cavium.com>
38833         * config/aarch64/aarch64.md (popcount<mode>2): New pattern.
38835 2017-02-07  Andrew Pinski  <apinski@cavium.com>
38837         * config/aarch64/aarch64-cores.def (thunderx): Disable LSE.
38838         (thunderxt88): Likewise.
38839         (thunderxt81): Disable LSE and change v8.1 to v8.
38840         (thunderxt83): Likewise.
38842 2017-02-07  Jakub Jelinek  <jakub@redhat.com>
38843             Richard Biener  <rguenther@suse.de>
38845         PR middle-end/79399
38846         * ira-int.h (struct target_ira_int): Change x_max_struct_costs_size
38847         type from int to size_t.
38848         * ira-costs.c (struct_costs_size): Change type from int to size_t.
38850 2017-02-07  Jakub Jelinek  <jakub@redhat.com>
38852         PR rtl-optimization/79386
38853         * cprop.c (bypass_conditional_jumps): Initialize
38854         bypass_last_basic_block already before splitting bbs after
38855         unconditional traps...
38856         (bypass_conditional_jumps): ... rather than here.
38858         PR target/79299
38859         * config/i386/sse.md (xtg_mode, gatherq_mode): New mode attrs.
38860         (*avx512f_gathersi<mode>, *avx512f_gathersi<mode>_2,
38861         *avx512f_gatherdi<mode>, *avx512f_gatherdi<mode>_2): Use them,
38862         fix -masm=intel patterns.
38864 2017-02-07  Richard Biener  <rguenther@suse.de>
38866         PR tree-optimization/79256
38867         PR middle-end/79278
38868         * builtins.c (get_object_alignment_2): Use min_align_of_type
38869         to extract alignment for MEM_REFs to honor BIGGEST_FIELD_ALIGNMENT
38870         and ADJUST_FIELD_ALIGN.
38872         * doc/tm.texi.in (ADJUST_FIELD_ALIGN): Adjust to take additional
38873         type parameter.
38874         * doc/tm.texi: Regenerate.
38875         * stor-layout.c (layout_decl): Adjust.
38876         (update_alignment_for_field): Likewise.
38877         (place_field): Likewise.
38878         (min_align_of_type): Likewise.
38879         * config/arc/arc.h (ADJUST_FIELD_ALIGN): Adjust.
38880         * config/epiphany/epiphany.h (ADJUST_FIELD_ALIGN): Likewise.
38881         * config/epiphany/epiphany.c (epiphany_adjust_field_align): Likewise.
38882         * config/frv/frv.h (ADJUST_FIELD_ALIGN): Likewise.
38883         * config/frv/frv.c (frv_adjust_field_align): Likewise.
38884         * config/i386/i386.h (ADJUST_FIELD_ALIGN): Likewise.
38885         * config/i386/i386.c (x86_field_alignment): Likewise.
38886         * config/rs6000/aix.h (ADJUST_FIELD_ALIGN): Likewise.
38887         * config/rs6000/darwin.h (ADJUST_FIELD_ALIGN): Likewise.
38888         * config/rs6000/freebsd64.h (ADJUST_FIELD_ALIGN): Likewise.
38889         * config/rs6000/linux64.h (ADJUST_FIELD_ALIGN): Likewise.
38890         * config/rs6000/sysv4.h (ADJUST_FIELD_ALIGN): Likewise.
38891         * config/rs6000/rs6000.c (rs6000_special_adjust_field_align_p):
38892          Likewise.
38894         Revert
38895         2017-01-30  Richard Biener  <rguenther@suse.de>
38897         PR tree-optimization/79256
38898         * targhooks.c (default_builtin_vector_alignment_reachable): Honor
38899         BIGGEST_FIELD_ALIGNMENT and ADJUST_FIELD_ALIGN to fix up bogus
38900         alignment on TYPE.
38902 2017-02-07  Toma Tabacu  <toma.tabacu@imgtec.com>
38904         * config/mips/mips.c (mips_expand_builtin_insn): Convert the QImode
38905         argument of the pshufh, psllh, psllw, psrah, psraw, psrlh, psrlw
38906         builtins to SImode and emit a zero-extend, if necessary.
38908 2017-02-06  Palmer Dabbelt  <palmer@dabbelt.com>
38910         * docs/invoke.texi (RISC-V Options): Alphabetize.
38912 2017-02-06  Palmer Dabbelt  <palmer@dabbelt.com>
38914         * doc/invoke.texi (RISC-V Options): Use two spaces to separate
38915         options.
38917 2017-02-06  Palmer Dabbelt  <palmer@dabbelt.com>
38919         * config/riscv/riscv.c: New file.
38920         * common/config/riscv/riscv-common.c: Likewise.
38921         * config.gcc: Likewise.
38922         * config/riscv/constraints.md: Likewise.
38923         * config/riscv/elf.h: Likewise.
38924         * config/riscv/generic.md: Likewise.
38925         * config/riscv/linux.h: Likewise.
38926         * config/riscv/multilib-generator: Likewise.
38927         * config/riscv/peephole.md: Likewise.
38928         * config/riscv/pic.md: Likewise.
38929         * config/riscv/predicates.md: Likewise.
38930         * config/riscv/riscv-builtins.c: Likewise.
38931         * config/riscv/riscv-c.c: Likewise.
38932         * config/riscv/riscv-ftypes.def: Likewise.
38933         * config/riscv/riscv-modes.def: Likewise.
38934         * config/riscv/riscv-opts.h: Likewise.
38935         * config/riscv/riscv-protos.h: Likewise.
38936         * config/riscv/riscv.h: Likewise.
38937         * config/riscv/riscv.md: Likewise.
38938         * config/riscv/riscv.opt: Likewise.
38939         * config/riscv/sync.md: Likewise.
38940         * config/riscv/t-elf-multilib: Likewise.
38941         * config/riscv/t-linux: Likewise.
38942         * config/riscv/t-linux-multilib: Likewise.
38943         * config/riscv/t-riscv: Likewise.
38944         * configure.ac: Likewise.
38945         * doc/contrib.texi: Add Kito Cheng, Palmer Dabbelt, and Andrew
38946         Waterman as RISC-V maintainers.
38947         * doc/install.texi: Add RISC-V entries.
38948         * doc/invoke.texi: Add RISC-V options section.
38949         * doc/md.texi: Add RISC-V constraints section.
38950         * configure: Regenerated.
38952 2017-02-06  Michael Meissner  <meissner@linux.vnet.ibm.com>
38954         PR target/66144
38955         * config/rs6000/vector.md (vcond<mode><mode>): Allow the true and
38956         false values to be constant vectors with all 0 or all 1 bits set.
38957         (vcondu<mode><mode>): Likewise.
38958         * config/rs6000/predicates.md (vector_int_reg_or_same_bit): New
38959         predicate.
38960         (fpmask_comparison_operator): Update comment.
38961         (vecint_comparison_operator): New predicate.
38962         * config/rs6000/rs6000.c (rs6000_emit_vector_cond_expr): Optimize
38963         vector conditionals when the true and false values are constant
38964         vectors with all 0 bits or all 1 bits set.
38966 2017-02-06  Martin Sebor  <msebor@redhat.com>
38968         PR  tree-optimization/79376
38969         * gimple-fold.c (get_range_strlen): Set the minimum length to zero.
38971 2017-02-06  Uros Bizjak  <ubizjak@gmail.com>
38973         * config/i386/sse.md (vector modes -> vec_extract* splitter): Use
38974         explicit subreg RTX with operand 1.  Use VECTOR_MODE_P predicate
38975         to simplify split condition.
38977 2017-02-06  Jakub Jelinek  <jakub@redhat.com>
38979         * omp-expand.c (oxpand_omp_atomic_fetch_op,
38980         expand_omp_atomic_pipeline): Return false if can_atomic_load_p is
38981         false.
38983 2017-02-06  Segher Boessenkool  <segher@kernel.crashing.org>
38985         PR rtl-optimization/68664
38986         * target.def (can_speculate_insn): New hook.
38987         * doc/tm.texi.in (TARGET_SCHED_CAN_SPECULATE_INSN): New hook.
38988         * doc/tm.texi: Regenerate.
38989         * sched-rgn.c (can_schedule_ready_p): Use the new hook.
38990         * config/rs6000/rs6000.c (TARGET_SCHED_CAN_SPECULATE_INSN): New macro.
38991         (rs6000_sched_can_speculate_insn): New function.
38993 2017-02-06  Jakub Jelinek  <jakub@redhat.com>
38995         PR tree-optimization/79284
38996         * tree-vectorizer.h (VECT_SCALAR_BOOLEAN_TYPE_P): Define.
38997         * tree-vect-stmts.c (vect_get_vec_def_for_operand,
38998         vectorizable_mask_load_store, vectorizable_operation,
38999         vect_is_simple_cond, get_same_sized_vectype): Use it instead
39000         of comparing TREE_CODE of a type against BOOLEAN_TYPE.
39001         * tree-vect-patterns.c (check_bool_pattern, search_type_for_mask_1,
39002         vect_recog_bool_pattern, vect_recog_mask_conversion_pattern): Likewise.
39003         * tree-vect-slp.c (vect_get_constant_vectors): Likewise.
39004         * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
39005         Remove redundant gimple_code (stmt) == GIMPLE_ASSIGN test after
39006         is_gimple_assign (stmt).  Replace another such test with
39007         is_gimple_assign (stmt).
39009 2017-02-06  Georg-Johann Lay  <avr@gjlay.de>
39011         PR target/78883
39012         * config/avr/avr.c (rtl-iter.h): Include it.
39013         (TARGET_LEGITIMATE_COMBINED_INSN): New hook define...
39014         (avr_legitimate_combined_insn): ...and implementation.
39016 2017-02-06  Dominik Vogt  <vogt@linux.vnet.ibm.com>
39018         * config/s390/predicates.md ("larl_operand"): Use macros from hwint.h.
39019         * config/s390/s390.c (s390_const_operand_ok)
39020         (s390_canonicalize_comparison, s390_extract_part)
39021         (s390_single_part, s390_contiguous_bitmask_nowrap_p)
39022         (s390_contiguous_bitmask_p, s390_rtx_costs)
39023         (legitimize_pic_address): Likewise.
39024         * config/s390/s390.md ("clzdi2", "clztidi2"): Likewise.
39025         * config/s390/vx-builtins.md ("vec_genbytemaskv16qi")
39026         ("vec_permi<mode>", "vfae<mode>", "*vfaes<mode>", "vstrc<mode>")
39027         ("*vstrcs<mode>"): Use UINTVAL() to set unsigned HOST_WIDE_INT.
39028         * config/s390/vector.md ("vec_vfenes<mode>"): Likewise.
39030 2017-02-06  Georg-Johann Lay  <avr@gjlay.de>
39032         * config/avr/avr.md (*addhi3_zero_extend): Add alternative where
39033         REGNO($0) == REGNO($1).
39035 2017-02-06  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
39037         * config/s390/linux.h(SIZE_TYPE): Add comment.
39039 2017-02-06  Julian Brown  <julian@codesourcery.com>
39040             Naveen H.S  <Naveen.Hurugalawadi@cavium.com>
39041             Virendra Pathak  <virendra.pathak@broadcom.com>
39043         * config/aarch64/aarch64-cores.def: Change the scheduler
39044         to Thunderx2t99.
39045         * config/aarch64/aarch64.md: Include thunderx2t99.md.
39046         * config/aarch64/thunderx2t99.md: New file.
39048 2017-02-05  Gerald Pfeifer  <gerald@pfeifer.com>
39050         * doc/standards.texi (Go Language): Update link to language
39051         standard.
39053 2017-02-05  Jan Hubicka  <hubicka@ucw.cz>
39055         * tree-eh.c (lower_resx): Sanitize profile.
39056         (cleanup_empty_eh_move_lp): Likewise.
39058 2017-02-05  Jan Hubicka  <hubicka@ucw.cz>
39060         PR tree-ssa/79347
39061         * cfgloopmanip.c (lv_adjust_loop_entry_edge, loop_version): Add
39062         ELSE_PROB.
39063         * cfgloopmanip.h (loop_version): Update prototype.
39064         * modulo-sched.c (sms_schedule): Update call of loop_version.
39065         * tree-if-conv.c(version_loop_for_if_conversion): Likewise.
39066         * tree-parloops.c (gen_parallel_loop): Likewise.
39067         * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Likewise.
39068         * tree-ssa-loop-split.c (split_loop): Likewise.
39069         * tree-ssa-loop-unswitch.c (tree_unswitch_loop): Likewise.
39070         * tree-vect-loop-manip.c (vect_loop_versioning): Likewise.
39072 2017-02-05  Martin Liska  <mliska@suse.cz>
39074         PR bootstrap/78985
39075         * config/s390/s390.c (s390_gimplify_va_arg): Initialize local
39076         variable to NULL.
39077         (print_operand_address): Initialize a struct to zero.
39079 2017-02-05  Gerald Pfeifer  <gerald@pfeifer.com>
39081         * doc/contrib.texi (Contributors): Refer to Hans Boehm's
39082         garbage collector only in textual form.
39084 2017-02-05  Gerald Pfeifer  <gerald@pfeifer.com>
39086         * doc/extend.texi (x86 specific memory model extensions for
39087         transactional memory): Simplify a phrase.
39089 2017-02-05  Eric Botcazou  <ebotcazou@adacore.com>
39091         PR target/79353
39092         * config/sparc/sync.md (atomic_loaddi_1): Replace 'U' constraint with
39093         'r', 'm' constraint with 'T' and !TARGET_ARCH64 with TARGET_ARCH32.
39094         (atomic_storedi_1): Likewise.
39096 2017-02-04  Jakub Jelinek  <jakub@redhat.com>
39098         PR tree-optimization/79338
39099         * tree-parloops.c (gather_scalar_reductions): Don't call
39100         vect_analyze_loop_form for loop->inner before destroying loop's
39101         loop_vinfo.
39103 2017-02-03  Martin Sebor  <msebor@redhat.com>
39105         PR tree-optimization/79327
39106         * gimple-ssa-sprintf.c (tree_digits): Avoid adding the base prefix
39107         when precision has resulted in leading zeros.
39108         (format_integer): Adjust the likely counter to assume an unknown
39109         argument that may be zero is non-zero.
39111 2017-02-03  Jason Merrill  <jason@redhat.com>
39113         PR c++/78689
39114         * tree-inline.c (copy_tree_body_r) [COND_EXPR]: Revert change to
39115         avoid copying non-taken branch.
39117 2017-02-03  Jakub Jelinek  <jakub@redhat.com>
39119         PR tree-optimization/79340
39120         * tree-vect-loop.c (vectorizable_reduction): Release
39121         vec_defs elements after safe_splicing them into other vectors.
39122         Formatting fixes.
39124         PR tree-optimization/79327
39125         * gimple-ssa-sprintf.c (adjust_range_for_overflow): If returning
39126         true, always set *argmin and *argmax to TYPE_{MIN,MAX}_VALUE of
39127         dirtype.
39128         (format_integer): Use wide_int_to_tree instead of build_int_cst
39129         + to_?hwi.  If argmin is NULL, just set argmin and argmax to
39130         TYPE_{MIN,MAX}_VALUE of argtype.  Simplify and fix computation
39131         of shortest and longest sequence.
39133 2017-02-03  Uros Bizjak  <ubizjak@gmail.com>
39135         * config/i386/i386.c (dimode_scalar_chain::convert_reg):
39136         Use pextrd for TARGET_SSE4_1 when creating scalar copy.
39138 2017-02-03  Walter Lee  <walt@tilera.com>
39140         PR target/78862
39141         * config/tilegx/tilegx.md (tilegx_expand_prologue): Add blockage
39142         after initial stackframe link reg save.
39143         * config/tilepro/tilepro.md (tilepro_expand_prologue): Likewise.
39145 2017-02-03  Jakub Jelinek  <jakub@redhat.com>
39147         PR target/79354
39148         * config/rs6000/rs6000.md (movsi_from_sf): Use wb constraint instead of
39149         wu for stxssp alternative.
39151 2017-02-03  Martin Sebor  <msebor@redhat.com>
39153         PR tree-optimization/79352
39154         * gimple-fold.c (get_range_strlen): Add argument.
39155         (get_range_strlen): Change return type to bool.
39156         (get_maxval_strlen): Pass in a dummy argument.
39157         * gimple-fold.h (get_range_strlen): Change return type to bool.
39158         * gimple-ssa-sprintf.c (get_string_length): Set unlikely counter.
39159         * tree.h (array_at_struct_end_p): Add argument.
39160         * tree.c (array_at_struct_end_p): Handle it.
39162 2017-02-03  Martin Liska  <mliska@suse.cz>
39164         PR lto/66295
39165         * multiple_target.c (create_dispatcher_calls): Redirect edge
39166         from a caller of a dispatcher.
39167         (expand_target_clones): Make the clones local.
39168         (ipa_target_clone): Do both target clones and resolvers.
39169         (ipa_dispatcher_calls): Remove the pass.
39170         (pass_dispatcher_calls::gate): Likewise.
39171         (make_pass_dispatcher_calls): Likewise.
39172         * passes.def (pass_target_clone): Put as very first IPA early
39173         pass.
39175 2017-02-03  Martin Liska  <mliska@suse.cz>
39177         * symtab.c (symtab_node::binds_to_current_def_p): Bail out
39178         in case of a function with ifunc attribute.
39180 2017-02-03  Martin Liska  <mliska@suse.cz>
39182         * cgraph.c (cgraph_node::dump): Dump function version info.
39183         * symtab.c (symtab_node::dump_base): Add missing new line.
39185 2017-02-02  Jan Hubicka  <hubicka@ucw.cz>
39187         * tree-ssa-ifcombine.c (update_profile_after_ifcombine): New function.
39188         (ifcombine_ifandif): Use it.
39190 2017-02-03  Martin Liska  <mliska@suse.cz>
39192         * doc/invoke.texi: Document default value for
39193         use-after-scope-direct-emission-threshold.
39195 2017-02-03  Martin Liska  <mliska@suse.cz>
39197         PR tree-optimization/79339
39198         * gimple-ssa-sprintf.c (format_floating_max): Call mpfr_clear.
39199         (format_floating): Likewise.
39201 2017-02-03  Martin Liska  <mliska@suse.cz>
39203         PR ipa/79337
39204         * ipa-prop.c (ipa_node_params_t::insert): Remove current
39205         implementation.
39206         (ipa_node_params_t::remove): Likewise.
39207         * ipa-prop.h (ipa_node_params::ipa_node_params): Make default
39208         initialization from removed ipa_node_params_t::insert.
39209         (ipa_node_params::~ipa_node_params): Move from removed
39210         ipa_node_params_t::release.
39211         * symbol-summary.h (symbol_summary::m_released): New member.
39212         Do not release a summary twice.  Do not allow to call finalizer
39213         for types of a summary that live in GGC memory.
39215 2017-02-02  Naveen H.S  <Naveen.Hurugalawadi@cavium.com>
39217         * config/aarch64/aarch64.c (thunderx2t99_tunings): Enable AES and
39218         cmp_branch fusion.
39220 2017-02-02  Martin Sebor  <msebor@redhat.com>
39222         PR middle-end/79275
39223         * gimple-ssa-sprintf.c (get_string_length): Set lower bound to zero.
39224         (format_string): Tighten up the range of output for non-constant
39225         strings and correct the expected range for wide non-constant strings.
39227 2017-02-02  Martin Sebor  <msebor@redhat.com>
39229         * doc/invoke.texi (-maccumulate-args): Fix bad grammar.
39231         PR middle-end/32003
39232         * doc/invoke.texi (-fdump-tree-): Remove pass-specific options from
39233         index.
39234         (-fdump-tree-@var): Add to index and document how to come up
39235         with pass-specific option and dump file names.
39236         (-fdump-passes): Clarify where to look for output.
39238 2017-02-02  Jan Hubicka  <hubicka@ucw.cz>
39240         PR middle-end/77445
39241         * tree-ssa-threadbackward.c (profitable_jump_thread_path): Dump
39242         statistics of the analyzed path; allow threading for speed when
39243         any of BBs along the path are optimized for speed.
39245 2017-02-02  Eric Botcazou  <ebotcazou@adacore.com>
39247         PR middle-end/78468
39248         * emit-rtl.c (init_emit): Add ??? comment for problematic alignment
39249         settings of the virtual registers.
39251         Revert again
39252         2016-08-23  Dominik Vogt  <vogt@linux.vnet.ibm.com>
39254         * explow.c (get_dynamic_stack_size): Take known alignment of stack
39255         pointer + STACK_DYNAMIC_OFFSET into account when calculating the size
39256         needed.
39258 2017-02-02  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
39260         * config/s390/vx-builtins.md ("vec_ceil", "vec_floor")
39261         ("vec_trunc", "vec_roundc", "vec_round"): Remove expanders.
39263 2017-02-02  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
39265         * config/s390/s390.md: Add missing comments with the expanded
39266         mnemonics.
39267         * config/s390/vector.md: Likewise.
39268         * config/s390/vx-builtins.md: Likewise.
39270 2017-02-02  Jakub Jelinek  <jakub@redhat.com>
39272         PR target/79197
39273         * config/rs6000/rs6000.md (*fixuns_trunc<mode>di2_fctiduz): Rename to ...
39274         (fixuns_trunc<mode>di2): ... this, remove previous expander.  Put all
39275         conditions on a single line.
39277 2017-02-02  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
39279         * config/s390/s390-c.c (s390_cpu_cpp_builtins_internal): Rename
39280         __S390_VX__ to __VX__.
39282 2017-02-01  Andrew Pinski  <apinski@cavium.com>
39284         * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost): Pass
39285         stmt_info to record_stmt_cost.
39286         (vect_get_known_peeling_cost): Pass stmt_info if known to
39287         record_stmt_cost.
39288         * config/aarch64/aarch64-protos.h (cpu_vector_cost): Split
39289         cpu_vector_cost field into
39290         scalar_int_stmt_cost and scalar_fp_stmt_cost.  Split vec_stmt_cost
39291         field into vec_int_stmt_cost and vec_fp_stmt_cost.
39292         * config/aarch64/aarch64.c (generic_vector_cost): Update for the
39293         splitting of scalar_stmt_cost and vec_stmt_cost.
39294         (thunderx_vector_cost): Likewise.
39295         (cortexa57_vector_cost): LIkewise.
39296         (exynosm1_vector_cost): Likewise.
39297         (xgene1_vector_cost): Likewise.
39298         (thunderx2t99_vector_cost): Improve after the splitting of the two
39299         fields.
39300         (aarch64_builtin_vectorization_cost): Update for the splitting of
39301         scalar_stmt_cost and vec_stmt_cost.
39303 2017-02-01  Torvald Riegel  <triegel@redhat.com>
39304             Richard Henderson  <rth@redhat.com>
39306         * builtins.c (fold_builtin_atomic_always_lock_free): Make "lock-free"
39307         conditional on existance of a fast atomic load.
39308         * optabs-query.c (can_atomic_load_p): New function.
39309         * optabs-query.h (can_atomic_load_p): Declare it.
39310         * optabs.c (expand_atomic_exchange): Always delegate to libatomic if
39311         no fast atomic load is available for the particular size of access.
39312         (expand_atomic_compare_and_swap): Likewise.
39313         (expand_atomic_load): Likewise.
39314         (expand_atomic_store): Likewise.
39315         (expand_atomic_fetch_op): Likewise.
39316         * testsuite/lib/target-supports.exp
39317         (check_effective_target_sync_int_128): Remove x86 because it provides
39318         no fast atomic load.
39319         (check_effective_target_sync_int_128_runtime): Likewise.
39321 2017-02-01  Richard Biener  <rguenther@suse.de>
39323         * graphite.c: Include tree-vectorizer.h for find_loop_location.
39324         (graphite_transform_loops): Provide opt-info for optimized nests.
39325         * tree-parloop.c (parallelize_loops): Provide opt-info for
39326         parallelized loops.
39328 2017-02-01  Richard Biener  <rguenther@suse.de>
39330         PR middle-end/79315
39331         * tree-cfg.c (move_stmt_op): Never set TREE_BLOCK when it
39332         was not set before.
39334 2017-02-01  Richard Biener  <rguenther@suse.de>
39336         PR tree-optimization/71824
39337         * graphite-scop-detection.c (scop_detection::build_scop_breadth):
39338         Verify the loops are valid in the merged SESE region.
39339         (scop_detection::can_represent_loop_1): Check analyzing the
39340         evolution of the number of iterations in the region succeeds.
39342 2017-01-31  Ian Lance Taylor  <iant@golang.org>
39344         * config/i386/i386.c (ix86_expand_split_stack_prologue): Add
39345         REG_ARGS_SIZE note to 32-bit push insns and call insn.
39347 2017-01-31  David Malcolm  <dmalcolm@redhat.com>
39349         PR preprocessor/79210
39350         * input.c (get_substring_ranges_for_loc): Replace line_width
39351         assertion with error-handling.
39353 2017-01-31  Richard Biener  <rguenther@suse.de>
39355         PR tree-optimization/77318
39356         * graphite-sese-to-poly.c (extract_affine): Fix assert.
39357         (create_pw_aff_from_tree): Take loop parameter.
39358         (add_condition_to_pbb): Pass loop of the condition to
39359         create_pw_aff_from_tree.
39361 2017-01-31  Jakub Jelinek  <jakub@redhat.com>
39363         * config/s390/s390.c (s390_asan_shadow_offset): New function.
39364         (TARGET_ASAN_SHADOW_OFFSET): Redefine.
39366 2017-01-31  Michael Meissner  <meissner@linux.vnet.ibm.com>
39368         PR target/78597
39369         PR target/79038
39370         * config/rs6000/rs6000-protos.h (convert_float128_to_int): Delete,
39371         no longer used.
39372         (convert_int_to_float128): Likewise.
39373         * config/rs6000/rs6000.c (convert_float128_to_int): Likewise.
39374         (convert_int_to_float128): Likewise.
39375         * config/rs6000/rs6000.md (UNSPEC_IEEE128_MOVE): Likewise.
39376         (UNSPEC_IEEE128_CONVERT): Likewise.
39377         (floatsi<mode>2, FLOAT128 iterator): Bypass calling
39378         rs6000_expand_float128_convert if we have IEEE 128-bit hardware.
39379         Use local variables for IBM extended format.
39380         (fix_trunc<mode>si2, FLOAT128 iterator): Likewise.
39381         (fix_trunc<mode>si2_fprs): Likewise.
39382         (fixuns_trunc<IEEE128:mode><SDI:mode>2): Likewise.
39383         (floatuns<IEEE128:mode>2, IEEE128 iterator): Likewise.
39384         (fix<uns>_<mode>si2_hw): Rework the IEEE 128-bt hardware support
39385         to know that we can now have integers of all sizes in vector
39386         registers.
39387         (fix<uns>_<mode>di2_hw): Likewise.
39388         (float<uns>_<mode>si2_hw): Likewise.
39389         (fix_<mode>si2_hw): Likewise.
39390         (fixuns_<mode>si2_hw): Likewise.
39391         (float<uns>_<mode>di2_hw): Likewise.
39392         (float_<mode>di2_hw): Likewise.
39393         (float_<mode>si2_hw): Likewise.
39394         (floatuns_<mode>di2_hw): Likewise.
39395         (floatuns_<mode>si2_hw): Likewise.
39396         (xscvqp<su>wz_<mode>): Delete, no longer used.
39397         (xscvqp<su>dz_<mode>): Likewise.
39398         (xscv<su>dqp_<mode>): Likewise.
39399         (ieee128_mfvsrd_64bit): Likewise.
39400         (ieee128_mfvsrd_32bit): Likewise.
39401         (ieee128_mfvsrwz): Likewise.
39402         (ieee128_mtvsrw): Likewise.
39403         (ieee128_mtvsrd_64bit): Likewise.
39404         (ieee128_mtvsrd_32bit): Likewise.
39406 2017-01-31  Martin Liska  <mliska@suse.cz>
39408         PR ipa/79285
39409         * ipa-prop.c (ipa_free_all_node_params): Call release method
39410         instead of ~sumbol_summary to not to trigger double times
39411         dtor of hash_map.
39413 2017-01-31  Aldy Hernandez  <aldyh@redhat.com>
39415         PR tree-optimization/71691
39416         * bitmap.h (class auto_bitmap): New.
39417         * tree-ssa-loop-unswitch.c (tree_may_unswitch_on): Call
39418         is_maybe_undefined instead of ssa_undefined_value_p.
39420 2017-01-31  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
39422         * config/s390/s390-c.c (s390_cpu_cpp_builtins_internal): Rename
39423         __S390_ARCH_LEVEL__ to __ARCH__.
39425 2017-01-31  Jakub Jelinek  <jakub@redhat.com>
39427         PR tree-optimization/79267
39428         * value-prof.c (gimple_ic): Only drop lhs for noreturn calls
39429         if should_remove_lhs_p is true.
39431 2017-01-30  Alexandre Oliva  <aoliva@redhat.com>
39433         PR debug/63238
39434         * dwarf2out.c (clone_as_declaration): Drop DW_AT_alignment.
39435         (add_alignment_attribute): New.
39436         (base_type_die): Add alignment attribute.
39437         (subrange_type_die): Likewise.
39438         (modified_type_die): Likewise.
39439         (gen_array_type_die): Likewise.
39440         (gen_descr_array_type_die: Likewise.
39441         (gen_enumeration_type_die): Likewise.
39442         (gen_subprogram_die): Likewise.
39443         (gen_variable_die): Likewise.
39444         (gen_field_die): Likewise.
39445         (gen_ptr_to_mbr_type_die): Likewise.
39446         (gen_struct_or_union_type_die): Likewise.
39447         (gen_subroutine_type_die): Likewise.
39448         (gen_typedef_die): Likewise.
39449         (base_type_cmp): Compare alignment attribute.
39451 2017-01-30  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
39453         PR target/79170
39454         * config/rs6000/altivec.md (*setb_internal): Rename to setb_signed.
39455         (setb_unsigned) New pattern for setb with CCUNS.
39456         * config/rs6000/rs6000.c (expand_block_compare): Use a different
39457         subfc./subfe sequence to avoid overflow problems.  Generate a
39458         shorter sequence with cmpld/setb for power9.
39459         * config/rs6000/rs6000.md (subf<mode>3_carry_dot2): Add a new pattern
39460         for generating subfc. instruction.
39461         (cmpstrsi): Add TARGET_POPCNTD predicate as the generate sequence
39462         now uses this instruction.
39464 2017-01-30  Ian Lance Taylor  <iant@google.com>
39466         PR debug/79289
39467         * dwarf2out.c (gen_type_die_with_usage): When picking a variant
39468         for FUNCTION_TYPE/METHOD_TYPE, use the first matching one.
39470 2017-01-30  Martin Sebor  <msebor@redhat.com>
39472         * gimple-ssa-sprintf.c (fmtresult::adjust_for_width_or_precision):
39473         Move constant to the right of a relational operator.
39474         (get_mpfr_format_length, format_character, format_string): Ditto.
39475         (should_warn_p, maybe_warn): Same.
39477         * doc/invoke.texi (-Wformat-truncation=1): Fix typo.
39479 2017-01-30  Maxim Ostapenko  <m.ostapenko@samsung.com>
39481         PR lto/79061
39482         * asan.c (get_translation_unit_decl): Remove function.
39483         (asan_add_global): Force has_dynamic_init to zero in LTO mode.
39485 2017-01-30  Martin Liska  <mliska@suse.cz>
39487         PR gcov-profile/79259
39488         * opts.c (common_handle_option): Enable flag_ipa_bit_cp w/
39489         -fprofile-generate.
39491 2017-01-30  Martin Liska  <mliska@suse.cz>
39493         PR bootstrap/78985
39494         * config/aarch64/cortex-a57-fma-steering.c (func_fma_steering::analyze):
39495         Initialize variables with NULL value.
39497 2017-01-30  Richard Earnshaw  <rearnsh@arm.com>
39499         PR target/79260
39500         * config.gcc (arm*-*-*): Add arm/arm-flags.h and arm/arm-isa.h to
39501         tm_p_file.
39502         * arm/arm-protos.h: Don't directly include arm-flags.h and arm-isa.h.
39504 2017-01-30  Richard Biener  <rguenther@suse.de>
39506         PR tree-optimization/79276
39507         * tree-vrp.c (process_assert_insertions): Properly adjust common
39508         when removing a duplicate.
39510 2017-01-30  Richard Biener  <rguenther@suse.de>
39512         PR tree-optimization/79256
39513         * targhooks.c (default_builtin_vector_alignment_reachable): Honor
39514         BIGGEST_FIELD_ALIGNMENT and ADJUST_FIELD_ALIGN to fix up bogus
39515         alignment on TYPE.
39516         * tree.c (build_aligned_type): Set TYPE_USER_ALIGN.
39518 2017-01-30  Dominik Vogt  <vogt@linux.vnet.ibm.com>
39520         PR target/79240
39521         * config/s390/s390.md ("*r<noxa>sbg_<mode>_srl_bitmask")
39522         ("*r<noxa>sbg_<mode>_sll_bitmask")
39523         ("*extzv_<mode>_srl<clobbercc_or_nocc>")
39524         ("*extzv_<mode>_sll<clobbercc_or_nocc>"):
39525         Use contiguous_bitmask_nowrap_operand.
39527 2017-01-29  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
39529         PR target/79268
39530         * config/rs6000/altivec.h (vec_xl): Revise #define.
39531         (vec_xst): Likewise.
39533 2017-01-27  Uros Bizjak  <ubizjak@gmail.com>
39535         * config/i386/i386.c (print_reg): Use REGNO instead of true_regnum.
39537 2017-01-27  Bernd Schmidt  <bschmidt@redhat.com>
39539         PR rtl-optimization/79194
39540         * cprop.c (one_cprop_pass): Move deletion of code after unconditional
39541         traps before call to bypass_conditional_jumps.
39543 2017-01-27  Vladimir Makarov  <vmakarov@redhat.com>
39545         PR tree-optimization/71374
39546         * lra-constraints.c (check_conflict_input_operands): New.
39547         (match_reload): Use it.
39549 2017-01-27  Vladimir Makarov  <vmakarov@redhat.com>
39551         PR target/79131
39552         * lra-assigns.c (find_hard_regno_for_1): Take endianess for into
39553         account to calculate conflict_set.
39555 2017-01-27  Bin Cheng  <bin.cheng@arm.com>
39557         PR rtl-optimization/78559
39558         * combine.c (try_combine): Discard REG_EQUAL and REG_EQUIV for
39559         other_insn in combine.
39561 2017-01-27  Pekka Jääskeläinen  <pekka.jaaskelainen@parmance.com>
39563         * builtin-types.def: Use unsigned_char_type_node for BT_UINT8.  Use
39564         uint16_type_node for BT_UINT16.
39566 2017-01-27  David Malcolm  <dmalcolm@redhat.com>
39568         * doc/sourcebuild.texi (Testsuites): Add "GIMPLE Tests" and
39569         "RTL Tests" to menu.
39570         (GIMPLE Tests): New node.
39571         (RTL Tests): New node.
39573 2017-01-27  Richard Biener  <rguenther@suse.de>
39575         PR tree-optimization/79245
39576         * tree-loop-distribution.c (distribute_loop): Apply cost
39577         modeling also to detected patterns.
39579 2017-01-27  Richard Biener  <rguenther@suse.de>
39581         PR tree-optimization/71433
39582         * tree-vrp.c (register_new_assert_for): Revert earlier changes.
39583         (compare_assert_loc): New function.
39584         (process_assert_insertions): Sort and optimize assert locations
39585         to remove duplicates and push down identical assertions on
39586         edges to their destination block.
39588 2017-01-27  Richard Biener  <rguenther@suse.de>
39590         PR tree-optimization/79244
39591         * tree-vrp.c (remove_range_assertions): Forcefully propagate
39592         out SSA names even if abnormal.
39594 2017-01-27  Jakub Jelinek  <jakub@redhat.com>
39596         * realmpfr.h: Poison MPFR_RND{N,Z,U,D}.
39597         * gimple-ssa-sprintf.c (format_floating_max): Use GMP_RNDN
39598         instead of MPFR_RNDN.
39600 2017-01-27  Richard Earnshaw  <rearnsha@arm.com>
39602         PR target/79239
39603         * arm.c (arm_option_override): Don't call build_target_option_node
39604         until after doing all option overrides.
39605         (arm_valid_target_attribute_tree): Likewise.
39607 2017-01-27  Martin Liska  <mliska@suse.cz>
39609         * doc/invoke.texi (-fprofile-arcs): Document profiling support
39610         for {cd}tors and C++ {cd}tors.
39612 2017-01-27  Dominik Vogt  <vogt@linux.vnet.ibm.com>
39614         * config/s390/s390.md ("*setmem_long_and")
39615         ("*setmem_long_and_31z"): Use zero_extend instead of and.
39617 2017-01-26  Martin Sebor  <msebor@redhat.com>
39619         * gimple-ssa-sprintf.c (format_floating): Simplify the computation
39620         of precision.
39622 2017-01-26  Martin Sebor  <msebor@redhat.com>
39624         * gimple-ssa-sprintf.c (format_floating): Test HAVE_XFmode and
39625         HAVE_DFmode before using XFmode or DFmode.
39626         (parse_directive): Avoid using the z length modifier to avoid
39627         the ISO C++98 does not support the â€˜z’ gnu_printf length modifier.
39629         PR middle-end/78703
39630         * gimple-ssa-sprintf.c (adjust_for_width_or_precision): Change
39631         to accept adjustment as an array.
39632         (get_int_range): New function.
39633         (struct directive): Make width and prec arrays.
39634         (directive::set_width, directive::set_precision): Call get_int_range.
39635         (format_integer, format_floating): Handle width and precision ranges.
39636         (format_string, parse_directive): Same.
39638 2017-01-26  Jakub Jelinek  <jakub@redhat.com>
39640         PR debug/79129
39641         * dwarf2out.c (generate_skeleton_bottom_up): For children with
39642         comdat_type_p set, just clone them, but keep the children in the
39643         original DIE.
39645         PR debug/78835
39646         * dwarf2out.c (prune_unused_types): Mark all functions with DIEs
39647         which have direct callers with -fvar-tracking-assignments enabled
39648         in the current TU.
39649         (resolve_addr): Avoid adding skeleton DIEs for DW_AT_call_origin
39650         inside of type units.
39652 2017-01-26  Martin Sebor  <msebor@redhat.com>
39654         PR middle-end/78703
39655         * gimple-ssa-sprintf.c (struct result_range): Add likely and
39656         unlikely counters.
39657         (struct format_result): Replace number_chars, number_chars_min,
39658         and number_chars_max with a single member of struct result_range.
39659         Remove bounded.
39660         (format_result::operator+=): Adjust.
39661         (struct fmtresult): Remove bounded.  Handle likely and unlikely
39662         counters.
39663         (fmtresult::adjust_for_width_or_precision): New function.
39664         (fmtresult:type_max_digits): New function.
39665         (bytes_remaining): Handle likely and unlikely counters.
39666         (min_bytes_remaining): Remove.
39667         (format_percent): Simplify.
39668         (format_integer, format_floating): Set likely and unlikely counters.
39669         (get_string_length, format_character, format_string): Same.
39670         (format_plain, should_warn_p): New function.
39671         (maybe_warn): Call should_warn_p.  Update diagnostic messages
39672         and handle those for all directives, including plain strings.
39673         (format_directive): Handle likely and unlikely counters.
39674         Remove unnecessary quoting from diagnostics.  Add an informational
39675         note.
39676         (add_bytes): Remove.
39677         (pass_sprintf_length::compute_format_length): Simplify.
39678         (try_substitute_return_value): Handle likely and unlikely counters.
39680 2017-01-26  Carl Love  <cel@us.ibm.com>
39682         * config/rs6000/rs6000-c (altivec_overloaded_builtins): Remove
39683         bogus entries for the P8V_BUILTIN_VEC_VGBBD built-ins
39685 2017-01-26  Vladimir Makarov  <vmakarov@redhat.com>
39687         PR target/79131
39688         * lra-assigns.c (setup_live_pseudos_and_spill_after_risky): Take
39689         endianess for subregs into account.
39690         * lra-constraints.c (lra_constraints): Do risky transformations
39691         always on the first iteration.
39692         * lra-lives.c (check_pseudos_live_through_calls): Add arg
39693         last_call_used_reg_set.
39694         (process_bb_lives): Define and use last_call_used_reg_set.
39695         * lra.c (lra): Always continue after lra_constraints on the first
39696         iteration.
39698 2017-01-26  Kirill Yukhin  <kirill.yukhin@gmail.com>
39700         * gcc.target/i386/avx512bw-kshiftlq-2.c: Use unsigned long long
39701         constant.
39702         * gcc.target/i386/avx512bw-kshiftrq-2.c: Ditto.
39704 2017-01-26  Jakub Jelinek  <jakub@redhat.com>
39706         * config/i386/avx512fintrin.h (_ktest_mask16_u8,
39707         _ktestz_mask16_u8, _ktestc_mask16_u8, _kadd_mask16): Move to ...
39708         * config/i386/avx512dqintrin.h (_ktest_mask16_u8,
39709         _ktestz_mask16_u8, _ktestc_mask16_u8, _kadd_mask16): ... here.
39710         * config/i386/i386-builtin.def (__builtin_ia32_ktestchi,
39711         __builtin_ia32_ktestzhi, __builtin_ia32_kaddhi): Use
39712         OPTION_MASK_ISA_AVX512DQ instead of OPTION_MASK_ISA_AVX512F.
39713         * config/i386/sse.md (SWI1248_AVX512BWDQ2): New mode iterator.
39714         (kadd<mode>, ktest<mode>): Use it instead of SWI1248_AVX512BWDQ.
39716 2017-01-26  Marek Polacek  <polacek@redhat.com>
39718         PR c/79199
39719         * fold-const.c (operand_equal_p) [COND_EXPR]: Use OP_SAME_WITH_NULL
39720         for the third operand.
39722 2017-01-26  Jakub Jelinek  <jakub@redhat.com>
39724         PR middle-end/79236
39725         * omp-low.c (struct omp_context): Add simt_stmt field.
39726         (scan_omp_for): Return omp_context *.
39727         (scan_omp_simd): Set simt_stmt on the non-_simt_ SIMD
39728         context to the _simt_ SIMD stmt.
39729         (lower_omp_for): For combined SIMD with sibling _simt_
39730         SIMD, make sure to use the same decls in _looptemp_
39731         clauses as in the sibling.
39733 2017-01-26  David Sherwood  <david.sherwood@arm.com>
39735         PR middle-end/79212
39736         * gimplify.c (omp_notice_variable): Add GOVD_SEEN flag to variables in
39737         all contexts.
39739 2017-01-26  Jakub Jelinek  <jakub@redhat.com>
39741         PR target/70465
39742         * reg-stack.c (emit_swap_insn): Instead of fld a; fld b; fxchg %st(1);
39743         emit fld b; fld a; if possible.
39745         * brig-builtins.def: Update copyright years.
39746         * config/arm/arm_acle_builtins.def: Update copyright years.
39748 2017-01-25  Michael Meissner  <meissner@linux.vnet.ibm.com>
39750         PR target/79179
39751         * config/rs6000/vsx.md (vsx_extract_<mode>_store): Use wY
39752         constraint instead of o for the stxsd instruction.
39754 2017-01-25  Carl Love  <cel@us.ibm.com>
39756         * config/rs6000/rs6000-c (altivec_overloaded_builtins): Fix order
39757         of entries for ALTIVEC_BUILTIN_VEC_PACKS and P8V_BUILTIN_VEC_VGBBD.
39759 2017-01-25  Jonathan Wakely  <jwakely@redhat.com>
39761         * doc/invoke.texi (C++ Dialect Options): Fix typo.
39763 2017-01-25  Richard Biener  <rguenther@suse.de>
39765         PR tree-optimization/69264
39766         * target.def (vector_alignment_reachable): Improve documentation.
39767         * doc/tm.texi: Regenerate.
39768         * targhooks.c (default_builtin_vector_alignment_reachable): Simplify
39769         and add a comment.
39770         * tree-vect-data-refs.c (vect_supportable_dr_alignment): Revert
39771         earlier changes with respect to TYPE_USER_ALIGN.
39772         (vector_alignment_reachable_p): Likewise.  Improve dumping.
39774 2016-01-25  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
39776         PR target/79145
39777         * config/arm/arm.md (xordi3): Force constant operand into a register
39778         for TARGET_IWMMXT.
39780 2016-01-25  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
39782         * doc/invoke.texi (-fstore-merging): Correct default optimization
39783         levels at which it is enabled.
39784         (-O): Move -fstore-merging from list to...
39785         (-O2): ... Here.
39787 2017-01-25  Richard Biener  <rguenther@suse.de>
39789         PR debug/78363
39790         * omp-expand.c: Include debug.h.
39791         (expand_omp_taskreg): Make sure to generate early debug before
39792         outlining anything from a function.
39793         (expand_omp_target): Likewise.
39794         (grid_expand_target_grid_body): Likewise.
39796 2017-01-25  Maxim Ostapenko  <m.ostapenko@samsung.com>
39798         PR lto/79061
39799         * asan.c (get_translation_unit_decl): New function.
39800         (asan_add_global): Extract modules file name from globals
39801         TRANSLATION_UNIT_DECL name.
39803 2017-01-24  Eric Botcazou  <ebotcazou@adacore.com>
39805         PR target/77439
39806         * config/arm/arm.c (arm_function_ok_for_sibcall): Add back restriction
39807         for long calls with APCS frame and VFP.
39809 2017-01-24  David Malcolm  <dmalcolm@redhat.com>
39811         * cfg.c (original_copy_tables_initialized_p): New function.
39812         * cfg.h (original_copy_tables_initialized_p): New decl.
39813         * cfgrtl.c (relink_block_chain): Guard the call to
39814         free_original_copy_tables with a call to
39815         original_copy_tables_initialized_p.
39816         * cgraph.h (symtab_node::native_rtl_p): New decl.
39817         * cgraphunit.c (symtab_node::native_rtl_p): New function.
39818         (symtab_node::needed_p): Don't assert for early assembly output
39819         for __RTL functions.
39820         (cgraph_node::finalize_function): Set "force_output" for __RTL
39821         functions.
39822         (cgraph_node::analyze): Bail out early for __RTL functions.
39823         (analyze_functions): Update assertion to support __RTL functions.
39824         (cgraph_node::expand): Bail out early for __RTL functions.
39825         * final.c (rest_of_clean_state): Don't call delete_tree_ssa for
39826         __RTL functions.
39827         * function.h (struct function): Update comment for field
39828         "pass_startwith".
39829         * gimple-expr.c: Include "tree-pass.h".
39830         (gimple_has_body_p): Return false for __RTL functions.
39831         * Makefile.in (OBJS): Add run-rtl-passes.o.
39832         * pass_manager.h (gcc::pass_manager::get_rest_of_compilation): New
39833         accessor.
39834         (gcc::pass_manager::get_clean_slate): New accessor.
39835         * passes.c: Include "insn-addr.h".
39836         (should_skip_pass_p): Add logging.  Update logic for running
39837         "expand" to be compatible with both __GIMPLE and __RTL.  Guard
39838         property-provider override so it is only done for gimple passes.
39839         Don't skip dfinit.
39840         (skip_pass): New function.
39841         (execute_one_pass): Call skip_pass when skipping passes.
39842         * read-md.c (md_reader::read_char): Support filtering
39843         the input to a subset of line numbers.
39844         (md_reader::md_reader): Initialize fields
39845         m_first_line and m_last_line.
39846         (md_reader::read_file_fragment): New function.
39847         * read-md.h (md_reader::read_file_fragment): New decl.
39848         (md_reader::m_first_line): New field.
39849         (md_reader::m_last_line): New field.
39850         * read-rtl-function.c (function_reader::create_function): Only
39851         create cfun if it doesn't already exist.  Set PROP_rtl on cfun's
39852         curr_properties.  Set DECL_INITIAL to a dummy block.
39853         (read_rtl_function_body_from_file_range): New function.
39854         * read-rtl-function.h (read_rtl_function_body_from_file_range):
39855         New decl.
39856         * run-rtl-passes.c: New file.
39857         * run-rtl-passes.h: New file.
39859 2017-01-24  Jeff Law  <law@redhat.com>
39861         * config/microblaze/microblaze.h (ASM_FORMAT_PRIVATE_NAME): Increase
39862         buffer size.
39864 2017-01-24  Bin Cheng  <bin.cheng@arm.com>
39866         PR tree-optimization/79159
39867         * tree-ssa-loop-niter.c (get_cst_init_from_scev): New function.
39868         (record_nonwrapping_iv): Improve boundary using above function if no
39869         value range information.
39871 2017-01-24  Pekka Jääskeläinen  <pekka@parmance.com>
39872             Martin Jambor  <mjambor@suse.cz>
39874         * brig-builtins.def: New file.
39875         * builtins.def (DEF_HSAIL_BUILTIN): New macro.
39876         (DEF_HSAIL_ATOMIC_BUILTIN): Likewise.
39877         (DEF_HSAIL_SAT_BUILTIN): Likewise.
39878         (DEF_HSAIL_INTR_BUILTIN): Likewise.
39879         (DEF_HSAIL_CVT_ZEROI_SAT_BUILTIN): Likewise.
39880         * builtin-types.def (BT_INT8): New.
39881         (BT_INT16): Likewise.
39882         (BT_UINT8): Likewise.
39883         (BT_UINT16): Likewise.
39884         (BT_FN_ULONG): Likewise.
39885         (BT_FN_UINT_INT): Likewise.
39886         (BT_FN_UINT_ULONG): Likewise.
39887         (BT_FN_UINT_LONG): Likewise.
39888         (BT_FN_UINT_PTR): Likewise.
39889         (BT_FN_ULONG_PTR): Likewise.
39890         (BT_FN_INT8_FLOAT): Likewise.
39891         (BT_FN_INT16_FLOAT): Likewise.
39892         (BT_FN_UINT32_FLOAT): Likewise.
39893         (BT_FN_UINT16_FLOAT): Likewise.
39894         (BT_FN_UINT8_FLOAT): Likewise.
39895         (BT_FN_UINT64_FLOAT): Likewise.
39896         (BT_FN_UINT16_UINT32): Likewise.
39897         (BT_FN_UINT32_UINT16): Likewise.
39898         (BT_FN_UINT16_UINT16_UINT16): Likewise.
39899         (BT_FN_INT_PTR_INT): Likewise.
39900         (BT_FN_UINT_PTR_UINT): Likewise.
39901         (BT_FN_LONG_PTR_LONG): Likewise.
39902         (BT_FN_ULONG_PTR_ULONG): Likewise.
39903         (BT_FN_VOID_UINT64_UINT64): Likewise.
39904         (BT_FN_UINT8_UINT8_UINT8): Likewise.
39905         (BT_FN_INT8_INT8_INT8): Likewise.
39906         (BT_FN_INT16_INT16_INT16): Likewise.
39907         (BT_FN_INT_INT_INT): Likewise.
39908         (BT_FN_UINT_FLOAT_UINT): Likewise.
39909         (BT_FN_FLOAT_UINT_UINT): Likewise.
39910         (BT_FN_ULONG_UINT_UINT): Likewise.
39911         (BT_FN_ULONG_UINT_PTR): Likewise.
39912         (BT_FN_ULONG_ULONG_ULONG): Likewise.
39913         (BT_FN_UINT_UINT_UINT): Likewise.
39914         (BT_FN_VOID_UINT_PTR): Likewise.
39915         (BT_FN_UINT_UINT_PTR: Likewise.
39916         (BT_FN_UINT32_UINT64_PTR): Likewise.
39917         (BT_FN_INT_INT_UINT_UINT): Likewise.
39918         (BT_FN_UINT_UINT_UINT_UINT): Likewise.
39919         (BT_FN_UINT_UINT_UINT_PTR): Likewise.
39920         (BT_FN_UINT_ULONG_ULONG_UINT): Likewise.
39921         (BT_FN_ULONG_ULONG_ULONG_ULONG): Likewise.
39922         (BT_FN_LONG_LONG_UINT_UINT): Likewise.
39923         (BT_FN_ULONG_ULONG_UINT_UINT): Likewise.
39924         (BT_FN_VOID_UINT32_UINT64_PTR): Likewise.
39925         (BT_FN_VOID_UINT32_UINT32_PTR): Likewise.
39926         (BT_FN_UINT_UINT_UINT_UINT_UINT): Likewise.
39927         (BT_FN_UINT_FLOAT_FLOAT_FLOAT_FLOAT): Likewise.
39928         (BT_FN_ULONG_ULONG_ULONG_UINT_UINT): Likewise.
39929         * doc/frontends.texi: List BRIG FE.
39930         * doc/install.texi (Testing): Add BRIG tesring requirements.
39931         * doc/invoke.texi (Overall Options): Mention BRIG.
39932         * doc/standards.texi (Standards): Doucment BRIG HSA version.
39934 2017-01-24  Richard Biener  <rguenther@suse.de>
39936         PR translation/79208
39937         * ipa-devirt.c (odr_types_equivalent_p): Fix typo in diagnostic.
39939 2017-01-24  Martin Jambor  <mjambor@suse.cz>
39941         PR bootstrap/79198
39942         * ipa-prop.c (ipa_free_all_node_params): Call summary destructor.
39943         * ipa-prop.c (ipa_node_params_t::insert): Initialize fields known_csts
39944         and known_contexts.
39946 2017-01-24  Aldy Hernandez  <aldyh@redhat.com>
39948         PR middle-end/79123
39949         * gimple-ssa-warn-alloca.c (alloca_call_type): Make sure
39950         casts from signed to unsigned really don't have a range.
39952 2017-01-24  Markus Trippelsdorf  <markus@trippelsdorf.de>
39954         * gimple-ssa-sprintf.c (format_floating): Change MPFR_RNDx to
39955         GMP_RNDx for compatiblity.
39957 2017-01-24  Martin Liska  <mliska@suse.cz>
39959         PR bootstrap/79132
39960         * tree-ssa-reassoc.c (rewrite_expr_tree_parallel): Insert assert
39961         that would prevent us to call alloca with -1 as argument.
39963 2017-01-24  Jakub Jelinek  <jakub@redhat.com>
39965         * dwarf2out.c (output_compilation_unit_header, output_file_names):
39966         Avoid -Wformat-security warning.
39968 2017-01-23  Andrew Pinski  <apinski@cavium.com>
39970         * config/aarch64/aarch64.c (thunderx2t99_addrcost_table): Improve
39971         cost table.
39973 2017-01-23  Martin Sebor  <msebor@redhat.com>
39975         PR middle-end/78703
39976         * gimple-ssa-sprintf.c (warn_level): New global.
39977         (format_integer): Use it here and throughout the rest of the file.
39978         Use the same switch to compute sign as base.
39979         (maybe_warn): New function.
39980         (format_directive): Factor out warnings into maybe_warn.
39981         Add debugging output.  Use warn_level.
39982         (add_bytes): Use warn_level.
39983         (pass_sprintf_length::compute_format_length): Add debugging output.
39984         (try_substitute_return_value): Same.
39985         (pass_sprintf_length::handle_gimple_call): Set and use warn_level.
39987         PR middle-end/78703
39988         * gimple-ssa-sprintf.c (struct format_result): Remove constant member.
39989         (struct fmtresult, format_integer, format_floating): Adjust.
39990         (fmtresult::fmtresult): Set max correctly in two argument ctor.
39991         (get_string_length, format_string,format_directive): Same.
39992         (pass_sprintf_length::compute_format_length): Same.
39993         (try_substitute_return_value): Simplify slightly.
39995         PR middle-end/78703
39996         * gimple-ssa-sprintf.c (pass_sprintf_length::gate): Adjust formatting.
39997         (fmtresult::operator+=): Outlined.
39998         (struct fmtresult): Add ctors.
39999         (struct conversion_spec): Rename...
40000         (struct directive): ...to this.  Add and remove data members.
40001         (directive::set_width, directive::set_precision): New functions.
40002         (format_percent): Use fmtresult ctor.
40003         (get_width_and_precision): Remove.
40004         (format_integer): Make naming changes.  Avoid computing width and
40005         precision.
40006         (format_floating): Same.  Adjust indentation.
40007         (format_character, format_none): New functions.
40008         (format_string): Moved character handling to format_character.
40009         (format_directive): Remove arguments, change return type.
40010         (parse_directive): New function.
40011         (pass_sprintf_length::compute_format_length): Move directive
40012         parsing to parse_directive.
40014 2017-01-23  Jakub Jelinek  <jakub@redhat.com>
40016         * tree.h (assign_assembler_name_if_neeeded): Rename to ...
40017         (assign_assembler_name_if_needed): ... this.
40018         * tree.c (assign_assembler_name_if_neeeded): Rename to ...
40019         (assign_assembler_name_if_needed): ... this.
40020         (free_lang_data_in_cgraph): Adjust callers.
40021         * cgraphunit.c (cgraph_node::analyze): Likewise.
40022         * omp-expand.c (expand_omp_taskreg, expand_omp_target):
40023         Likewise.
40025 2017-01-23  Richard Biener  <rguenther@suse.de>
40027         PR tree-optimization/79088
40028         PR tree-optimization/79188
40029         * tree-ssa-threadupdate.c (mark_threaded_blocks): Move code
40030         resetting loop bounds after last path deletion.  Reset loop
40031         bounds of the target loop, make code match the comments.
40032         * tree-ssa-threadbackwards.c (pass_early_thread_jumps::execute):
40033         Make sure loops need no fixups.
40035 2017-01-23  Kelvin Nilsen  <kelvin@gcc.gnu.org>
40037         * config/rs6000/rs6000-builtin.def (VSIEDPF): Add scalar insert
40038         exponent support with double type for first argument.
40039         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Changed
40040         type returned by __builtin_vec_extract_sig,
40041         __builtin_vec_extract_sig_sp, and __builtin_vec_extract_sig_dp
40042         functions from "vector int" to "vector unsigned int" or from
40043         "vector long long int" to "vector unsigned long long int".
40044         Changed type returned by __builtin_vec_extract_exp,
40045         __builtin_vec_extract_exp_sp, and __builtin_vec_extract_exp_dp
40046         functions from "vector int" to "vector unsigned int" or from
40047         "vector long long int" to "vector unsigned long long int".
40048         Changed return type of __builtin_vec_test_data_class,
40049         __builtin_vec_test_data_class_sp, and
40050         __builtin_vec_test_data_class_dp from "vector int" to
40051         "vector bool int" or from "vector long long int" to "vector bool
40052         long long int" and changed second argument type from "unsigned
40053         int" to "int".  Added new overloaded function forms "vector float
40054         __builtin_vec_insert_exp (vector float, vector unsigned int)" and
40055         "vector float __builtin_vec_insert_exp_sp (vector float, vector
40056         unsigned int)" and "vector double __builtin_vec_insert_exp (vector
40057         double, vector unsigned long long int)" and "vector double
40058         __builtin_vec_insert_exp_dp (vector double, vector unsigned long
40059         long int)".  Changed return type of
40060         __builtin_scalar_test_data_class and
40061         __builtin_scalar_test_data_class_sp and
40062         __builtin_scalar_test_data_class_dp from "unsigned int" to "bool
40063         int" and changed second argument from "unsigned int" to "int".
40064         Changed type returned by __builtin_scalar_test_neg,
40065         __builtin_scalar_test_neg_sp, and __builtin_scalar_test_neg_dp
40066         from "int" to "bool int".  Added new overloaded function form
40067         "double __builtin_scalar_insert_exp (double, unsigned long long int)".
40068         * config/rs6000/vsx.md (xsiexpdpf): New insn for scalar insert
40069         exponent double-precision with floating point first argument.
40070         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Adjust
40071         documentation of scalar_test_data_class, scalar_test_neg,
40072         scalar_extract_sig, scalar_extract_exp, scalar_insert_exp,
40073         vector_extract_exp, vec_extract_sig, vec_insert_exp, and
40074         vec_test_data_class built-in functions to reflect refinements in
40075         their type signatures.
40077 2017-01-23  Andreas Tobler  <andreast@gcc.gnu.org>
40079         * config/aarch64/aarch64.c (aarch64_elf_asm_constructor): Increase
40080         size of buf.
40081         (aarch64_elf_asm_destructor): Likewise.
40083 2017-01-23  Bernd Schmidt  <bschmidt@redhat.com>
40085         PR rtl-optimization/78634
40086         * config/i386/i386.c (ix86_max_noce_ifcvt_seq_cost): New function.
40087         (TARGET_MAX_NOCE_IFCVT_SEQ_COST): Define.
40088         * ifcvt.c (noce_try_cmove): Add missing cost check.
40090         PR rtl-optimization/71724
40091         * combine.c (if_then_else_cond): Look for situations where it is
40092         beneficial to undo the work of one of the recursive calls.
40094 2017-01-23  Bin Cheng  <bin.cheng@arm.com>
40096         PR tree-optimization/70754
40097         * tree-predcom.c (stmt_combining_refs): New parameter INSERT_BEFORE.
40098         (reassociate_to_the_same_stmt): New parameter INSERT_BEFORE.  Insert
40099         combined stmt before it if not NULL.
40100         (combine_chains): Process refs reversely and compute dominance point
40101         for root ref.
40103 2017-01-23  Martin Liska  <mliska@suse.cz>
40105         PR tree-optimization/79196
40106         * tree-ssa-strlen.c (fold_strstr_to_memcmp): Rename to ...
40107         (fold_strstr_to_strncmp): ... this.  Fold the pattern to strncmp
40108         instead of memcmp.
40109         (strlen_optimize_stmt): Call the renamed function.
40111 2017-01-23  Michael Matz  <matz@suse.de>
40113         PR tree-optimization/78384
40114         * tree-ssa-loop-split.c (patch_loop_exit): Use correct edge.
40116 2017-01-23  Richard Biener  <rguenther@suse.de>
40118         PR tree-optimization/79186
40119         * tree-vrp.c (register_new_assert_for): Make sure we've seen
40120         both incoming edges before moving an assert.
40122 2017-01-23  Martin Jambor  <mjambor@suse.cz>
40124         * ipa-prop.c (load_from_param_1): Removed.
40125         (load_from_unmodified_param): Bits from load_from_param_1 put back
40126         here.
40127         (load_from_param): Removed.
40128         (compute_complex_assign_jump_func): Removed stmt2 and just replaced it
40129         with stmt.  Reverted back to use of load_from_unmodified_param.
40131 2017-01-23  Martin Jambor  <mjambor@suse.cz>
40133         PR ipa/79108
40134         * ipa-prop.h (ipa_param_descriptor): Anotate with with GTY(()).
40135         (ipa_node_params): Annotate with GTY((for_user)).  Make descriptors
40136         field a pointer to garbage collected vector, mark lattices and
40137         ipcp_orig_node with GTY((skip)).
40138         (ipa_get_param_count): Adjust to descriptors being a pointer.
40139         (ipa_get_param): Likewise.
40140         (ipa_get_type): Likewise.
40141         (ipa_get_param_move_cost): Likewise.
40142         (ipa_set_param_used): Likewise.
40143         (ipa_get_controlled_uses): Likewise.
40144         (ipa_set_controlled_uses): Likewise.
40145         (ipa_is_param_used): Likewise.
40146         (ipa_node_params_t): Move into garbage collector.  New methods insert
40147         and remove.
40148         (ipa_node_params_sum): Annotate wth GTY(()).
40149         (ipa_check_create_node_params): Adjust to ipa_node_params_sum being
40150         garbage collected.
40151         (ipa_load_from_parm_agg): Adjust declaration.
40152         * ipa-icf.c (param_used_p): Adjust to descriptors being a pointer.
40153         * ipa-profile.c (ipa_profile): Likewise.
40154         * ipa-prop.c (ipa_get_param_decl_index_1): Likewise.
40155         (ipa_populate_param_decls): Make descriptors parameter garbage
40156         collected.
40157         (ipa_dump_param): Adjust to descriptors being a pointer.
40158         (ipa_alloc_node_params): Likewise.
40159         (ipa_initialize_node_params): Likewise.
40160         (load_from_param_1): Make descriptors parameter garbage collected.
40161         (load_from_unmodified_param): Likewise.
40162         (load_from_param): Likewise.
40163         (ipa_load_from_parm_agg): Likewise.
40164         (ipa_node_params::~ipa_node_params): Removed.
40165         (ipa_free_all_node_params): Remove call to delete operator.
40166         (ipa_node_params_t::insert): New.
40167         (ipa_node_params_t::remove): Likewise.
40168         (ipa_node_params_t::duplicate): Adjust to descriptors being a pointer,
40169         copy known_csts and known_contexts vectors.
40170         (ipa_read_node_info): Adjust to descriptors being a pointer.
40171         (ipcp_modif_dom_walker): Make m_descriptors field garbage
40172         collected.
40173         (ipcp_transform_function): Make descriptors variable garbage
40174         collected.
40176 2017-01-23  Andrew Senkevich  <andrew.senkevich@intel.com>
40178         * config/i386/avx512bwintrin.h: Add k-mask test, kortest intrinsics.
40179         * config/i386/avx512dqintrin.h: Ditto.
40180         * config/i386/avx512fintrin.h: Ditto.
40181         * config/i386/i386.c: Handle new builtins.
40182         * config/i386/i386-builtin.def: Add new builtins.
40183         * config/i386/sse.md (ktest<mode>, kortest<mode>): New.
40184         (UNSPEC_KORTEST, UNSPEC_KTEST): New.
40186 2017-01-23  Jakub Jelinek  <jakub@redhat.com>
40187             Martin Liska  <mliska@suse.cz>
40189         * asan.h: Define ASAN_USE_AFTER_SCOPE_ATTRIBUTE.
40190         * asan.c (asan_expand_poison_ifn): Support stores and use
40191         appropriate ASAN report function.
40192         * internal-fn.c (expand_ASAN_POISON_USE): New function.
40193         * internal-fn.def (ASAN_POISON_USE): Declare.
40194         * tree-into-ssa.c (maybe_add_asan_poison_write): New function.
40195         (maybe_register_def): Create ASAN_POISON_USE when sanitizing.
40196         * tree-ssa-dce.c (eliminate_unnecessary_stmts): Remove
40197         ASAN_POISON calls w/o LHS.
40198         * tree-ssa.c (execute_update_addresses_taken): Create clobber
40199         for ASAN_MARK (UNPOISON, &x, ...) in order to prevent usage of a LHS
40200         from ASAN_MARK (POISON, &x, ...) coming to a PHI node.
40201         * gimplify.c (asan_poison_variables): Add attribute
40202         use_after_scope_memory to variables that really needs to live
40203         in memory.
40204         * tree-ssa.c (is_asan_mark_p): Do not rewrite into SSA when
40205         having the attribute.
40207 2017-01-23  Martin Liska  <mliska@suse.cz>
40209         * asan.c (create_asan_shadow_var): New function.
40210         (asan_expand_poison_ifn): Likewise.
40211         * asan.h (asan_expand_poison_ifn): New declaration.
40212         * internal-fn.c (expand_ASAN_POISON): Likewise.
40213         * internal-fn.def (ASAN_POISON): New builtin.
40214         * sanopt.c (pass_sanopt::execute): Expand
40215         asan_expand_poison_ifn.
40216         * tree-inline.c (copy_decl_for_dup_finish): Make function
40217         external.
40218         * tree-inline.h (copy_decl_for_dup_finish): Likewise.
40219         * tree-ssa.c (is_asan_mark_p): New function.
40220         (execute_update_addresses_taken): Rewrite local variables
40221         (identified just by use-after-scope as addressable) into SSA.
40223 2017-01-22  Gerald Pfeifer  <gerald@pfeifer.com>
40225         * doc/install.texi (Specific): opensource.apple.com uses https
40226         now. Remove trailing slash.
40228 2017-01-22  Gerald Pfeifer  <gerald@pfeifer.com>
40230         * README.Portability: Remove note on an Irix compatibility issue.
40232 2017-01-22  Dimitry Andric <dim@FreeBSD.org>
40234         * gcov.c (INCLUDE_ALGORITHM): Define.
40235         (INCLUDE_VECTOR): Define.
40236         No longer include <vector> and <algorithm> directly.
40238 2017-01-21  Gerald Pfeifer  <gerald@pfeifer.com>
40240         * doc/extend.texi (Thread-Local): Change www.akkadia.org reference
40241         to https.
40242         * doc/invoke.texi (Code Gen Options): Ditto.
40244 2017-01-21  Jan Hubicka  <hubicka@ucw.cz>
40246         PR lto/78407
40247         * cfg.c (update_bb_profile_for_threading): Fix updating of probablity.
40249 2017-01-21  Bernd Schmidt  <bschmidt@redhat.com>
40251         rtl-optimization/79125
40252         * cprop.c (local_cprop_pass): Handle cases where we make an
40253         unconditional trap.
40255 2017-01-20  Segher Boessenkool  <segher@kernel.crashing.org>
40257         PR target/61729
40258         PR target/77850
40259         * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Adjust address to
40260         read from, for big endian.
40262 2017-01-20  Jiong Wang  <jiong.wang@arm.com>
40264         * config/aarch64/aarch64-builtins.c (aarch64_init_builtins): Register
40265         register pauth builtins for LP64 only.
40267 2017-01-20  Marek Polacek  <polacek@redhat.com>
40269         PR c/79152
40270         * gimplify.c (should_warn_for_implicit_fallthrough): Handle consecutive
40271         non-case labels.
40273 2017-01-20  Alexander Monakov  <amonakov@ispras.ru>
40275         * omp-expand.c (expand_omp_simd): Clear PROP_gimple_lomp_dev regardless
40276         of safelen status.
40277         * omp-offload.c (pass_omp_device_lower::gate): Use PROP_gimple_lomp_dev.
40278         * passes.c (dump_properties): Handle PROP_gimple_lomp_dev.
40279         * tree-inline.c (expand_call_inline): Propagate PROP_gimple_lomp_dev.
40281 2017-01-20  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
40283         PR target/71270
40284         * config/arm/arm.c (neon_valid_immediate): Reject vector constants
40285         in big-endian mode when they are not a single duplicated value.
40287 2017-01-20  Richard Biener  <rguenther@suse.de>
40289         * BASE-VER: Bump to 7.0.1.
40291 2017-01-20  Alexander Monakov  <amonakov@ispras.ru>
40293         * omp-low.c (omplow_simd_context): New struct.  Use it...
40294         (lower_rec_simd_input_clauses): ...here and...
40295         (lower_rec_input_clauses): ...here to hold common data.  Adjust all
40296         references to idx, lane, max_vf, is_simt.
40298 2017-01-20  Graham Markall  <graham.markall@embecosm.com>
40300         * config/arc/arc.h (LINK_SPEC): Use arclinux_nps emulation when
40301         mcpu=nps400.
40303 2017-01-20  Martin Jambor  <mjambor@suse.cz>
40305         * hsa.h: Renaed to hsa-common.h.  Adjusted a comment.
40306         * hsa.c: Renaed to hsa-common.c.  Change include of gt-hsa.h to
40307         gt-hsa-common.h.
40308         * Makefile.in (OBJS): Rename hsa.o to hsa-common.o.
40309         (GTFILES): Rename hsa.c to hsa-common.c.
40310         * hsa-brig.c: Change include of hsa.h to hsa-common.h.
40311         * hsa-dump.c: Likewise.
40312         * hsa-gen.c: Likewise.
40313         * hsa-regalloc.c: Likewise.
40314         * ipa-hsa.c: Likewise.
40315         * omp-expand.c: Likewise.
40316         * omp-low.c: Likewise.
40317         * toplev.c: Likewise.
40319 2017-01-20  Marek Polacek  <polacek@redhat.com>
40321         PR c/64279
40322         * doc/invoke.texi: Document -Wduplicated-branches.
40323         * fold-const.c (operand_equal_p): Handle MODIFY_EXPR, INIT_EXPR,
40324         COMPOUND_EXPR, PREDECREMENT_EXPR, PREINCREMENT_EXPR,
40325         POSTDECREMENT_EXPR, POSTINCREMENT_EXPR, CLEANUP_POINT_EXPR, EXPR_STMT,
40326         STATEMENT_LIST, and RETURN_EXPR.  For non-pure non-const functions
40327         return 0 only when not OEP_LEXICOGRAPHIC.
40328         (fold_build_cleanup_point_expr): Use the expression
40329         location when building CLEANUP_POINT_EXPR.
40330         * tree-core.h (enum operand_equal_flag): Add OEP_LEXICOGRAPHIC.
40331         * tree.c (add_expr): Handle error_mark_node.
40333 2017-01-20  Martin Liska  <mliska@suse.cz>
40335         PR lto/69188
40336         * tree-profile.c (init_ic_make_global_vars): Do not call
40337         finalize_decl.
40338         (gimple_init_gcov_profiler): Likewise.
40340 2017-01-20  Martin Liska  <mliska@suse.cz>
40342         PR ipa/71190
40343         * cgraph.h (maybe_create_reference): Remove argument and
40344         update comment.
40345         * cgraphclones.c (cgraph_node::create_virtual_clone): Remove one
40346         argument.
40347         * ipa-cp.c (create_specialized_node): Likewise.
40348         * symtab.c (symtab_node::maybe_create_reference): Handle
40349         VAR_DECLs and ADDR_EXPRs and select ipa_ref_use type.
40351 2017-01-20  Martin Liska  <mliska@suse.cz>
40353         * read-rtl-function.c (function_reader::create_function): Use
40354         build_decl instread of build_decl_stat.
40356 2017-01-20  Andrew Senkevich  <andrew.senkevich@intel.com>
40358         * config/i386/avx512bwintrin.h: Add k-mask registers shift intrinsics.
40359         * config/i386/avx512dqintrin.h: Ditto.
40360         * config/i386/avx512fintrin.h: Ditto.
40361         * config/i386/i386-builtin-types.def: Add new types.
40362         * config/i386/i386.c: Handle new types.
40363         * config/i386/i386-builtin.def (__builtin_ia32_kshiftliqi)
40364         (__builtin_ia32_kshiftlihi, __builtin_ia32_kshiftlisi)
40365         (__builtin_ia32_kshiftlidi, __builtin_ia32_kshiftriqi)
40366         (__builtin_ia32_kshiftrihi, __builtin_ia32_kshiftrisi)
40367         (__builtin_ia32_kshiftridi): New.
40368         * config/i386/sse.md (k<code><mode>): Rename *k<code><mode>.
40370 2017-01-19  Segher Boessenkool  <segher@kernel.crashing.org>
40372         PR target/78875
40373         PR target/79140
40374         * config/rs6000/rs6000.c (TARGET_STACK_PROTECT_GUARD): Unconditionally
40375         define to rs6000_init_stack_protect_guard.
40376         (rs6000_init_stack_protect_guard): New function.
40378 2017-01-19  Matthew Fortune  <matthew.fortune@imgtec.com>
40379             Yunqiang Su  <yunqiang.su@imgtec.com>
40381         * config.gcc (supported_defaults): Add madd4.
40382         (with_madd4): Add validation.
40383         (all_defaults): Add madd4.
40384         * config/mips/mips.opt (mmadd4): New option.
40385         * config/mips/mips.h (OPTION_DEFAULT_SPECS): Add a default for
40386         mmadd4.
40387         (TARGET_CPU_CPP_BUILTINS): Add builtin_define for
40388         __mips_no_madd4.
40389         (ISA_HAS_UNFUSED_MADD4): Gate with mips_madd4.
40390         (ISA_HAS_FUSED_MADD4): Likewise.
40391         * doc/invoke.texi (-mmadd4): Document the new option.
40392         * doc/install.texi (--with-madd4): Document the new option.
40394 2017-01-19  Jiong Wang  <jiong.wang@arm.com>
40396         * config/aarch64/aarch64-builtins.c (enum aarch64_builtins): New
40397         entries for AARCH64_PAUTH_BUILTIN_XPACLRI,
40398         AARCH64_PAUTH_BUILTIN_PACIA1716, AARCH64_PAUTH_BUILTIN_AUTIA1716.
40399         (aarch64_init_pauth_hint_builtins): New.
40400         (aarch64_init_builtins): Call aarch64_init_pauth_hint_builtins.
40401         (aarch64_expand_builtin): Expand new builtins.
40403 2017-01-19  Jiong Wang  <jiong.wang@arm.com>
40405         * reg-notes.def (CFA_TOGGLE_RA_MANGLE): New reg-note.
40406         * combine-stack-adj.c (no_unhandled_cfa): Handle
40407         REG_CFA_TOGGLE_RA_MANGLE.
40408         * dwarf2cfi.c (dwarf2out_frame_debug): Handle REG_CFA_TOGGLE_RA_MANGLE.
40409         * config/aarch64/aarch64.c (aarch64_expand_prologue): Generates DWARF
40410         info for return address signing.
40411         (aarch64_expand_epilogue): Likewise.
40413 2017-01-19  Jiong Wang  <jiong.wang@arm.com>
40415         * config/aarch64/aarch64-opts.h (aarch64_function_type): New enum.
40416         * config/aarch64/aarch64-protos.h
40417         (aarch64_return_address_signing_enabled): New declaration.
40418         * config/aarch64/aarch64.c (aarch64_return_address_signing_enabled):
40419         New function.
40420         (aarch64_expand_prologue): Sign return address before it's pushed onto
40421         stack.
40422         (aarch64_expand_epilogue): Authenticate return address fetched from
40423         stack.
40424         (aarch64_override_options): Sanity check for ILP32 and ISA level.
40425         (aarch64_attributes): New function attributes for "sign-return-address".
40426         * config/aarch64/aarch64.md (UNSPEC_AUTI1716, UNSPEC_AUTISP,
40427         UNSPEC_PACI1716, UNSPEC_PACISP, UNSPEC_XPACLRI): New unspecs.
40428         ("*do_return"): Generate combined instructions according to key index.
40429         ("<pauth_mnem_prefix>sp", "<pauth_mnem_prefix1716", "xpaclri"): New.
40430         * config/aarch64/iterators.md (PAUTH_LR_SP, PAUTH_17_16): New integer
40431         iterators.
40432         (pauth_mnem_prefix, pauth_hint_num_a): New integer attributes.
40433         * config/aarch64/aarch64.opt (msign-return-address=): New.
40434         * doc/extend.texi (AArch64 Function Attributes): Documents
40435         "sign-return-address=".
40436         * doc/invoke.texi (AArch64 Options): Documents "-msign-return-address=".
40438 2017-01-19  Matthew Fortune  <matthew.fortune@imgtec.com>
40440         * doc/invoke.texi: Add missing -mlxc1-sxc1 options to
40441         overall option summary.
40443 2017-01-19  Jiong Wang  <jiong.wang@arm.com>
40445         * config/aarch64/aarch64-arches.def: New entry for "armv8.3-a".
40446         * config/aarch64/aarch64.h (AARCH64_FL_V8_3, AARCH64_FL_FOR_ARCH8_3,
40447         AARCH64_ISA_V8_3, TARGET_ARMV8_3): New.
40448         * doc/invoke.texi (AArch64 Options): Document "armv8.3-a".
40450 2017-01-19  Michael Meissner  <meissner@linux.vnet.ibm.com>
40452         * config/rs6000/rs6000-cpus.def (ISA_3_0_MASKS_SERVER): Enable
40453         -mpower9-minmax by default for -mcpu=power9.
40454         (ISA_3_MASKS_IEEE): Require -mvsx-small-integer to enable IEEE
40455         128-bit floating point.
40457 2017-01-20  Alan Modra  <amodra@gmail.com>
40459         * config/rs6000/rs6000.md (cmpstrnsi, cmpstrsi): Fail if
40460         optimizing for size.
40462 2017-01-20  Alan Modra  <amodra@gmail.com>
40464         PR target/79144
40465         * config/rs6000/rs6000.c (expand_strn_compare): Get the asm name
40466         for strcmp and strncmp from corresponding builtin decl.
40468 2017-01-19  Uros Bizjak  <ubizjak@gmail.com>
40470         * config.gcc (x86_64-*-rtems*): Use i386/rtemself.h
40471         instead of i386/rtems-64.h.
40472         * config/i386/rtems-64.h: Remove.
40474 2017-01-19  Uros Bizjak  <ubizjak@gmail.com>
40476         PR target/78478
40477         Revert:
40478         2013-11-05  Uros Bizjak  <ubizjak@gmail.com>
40480         * config/i386/rtemself.h (LONG_DOUBLE_TYPE_SIZE): New define.
40482 2017-01-19  Tamar Christina  <tamar.christina@arm.com>
40484         * config/aarch64/aarch64.c (aarch64_simd_gen_const_vector_dup):
40485         Change int to HOST_WIDE_INT.
40486         * config/aarch64/aarch64-protos.h
40487         (aarch64_simd_gen_const_vector_dup): Likewise.
40488         * config/aarch64/aarch64-simd.md: Add copysign<mode>3.
40490 2017-01-19  David Malcolm  <dmalcolm@redhat.com>
40492         * langhooks-def.h (lhd_type_for_size): New decl.
40493         (LANG_HOOKS_TYPE_FOR_SIZE): Define as lhd_type_for_size.
40494         * langhooks.c (lhd_type_for_size): New function, taken from
40495         lto_type_for_size.
40497 2017-01-19  Pat Haugen  <pthaugen@us.ibm.com>
40499         * config/rs6000/power9.md (power9-alu): Remove 'cmp' type and add
40500         define_bypass for CR latency.
40501         (power9-cracked-alu): Update bypass latency and remove power9-branch.
40502         (power9-alu2): Add define_bypass for CR latency.
40503         (power9-cmp): New.
40504         (power9-mul): Update insn latency.
40505         (power9-mul-compare): Update insn latency, bypass latency and remove
40506         power9-branch.
40508 2016-01-19  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
40510         * config/aarch64/aarch64-protos.h (aarch64_nopcrelative_literal_loads):
40511         Delete.
40512         * config/aarch64/aarch64.md
40513         (aarch64_reload_movcp<GPF_TF:mode><P:mode>): Delete reference to
40514         aarch64_nopcrelative_literal_loads.
40515         (aarch64_reload_movcp<VALL:mode><P:mode>): Likewise.
40517 2017-01-19  Chenghua Xu  <paul.hua.gm@gmail.com>
40519         * config/mips/mips.h (ISA_HAS_FUSED_MADD4): Enable for
40520         TARGET_LOONGSON_3A.
40521         (ISA_HAS_UNFUSED_MADD4): Exclude TARGET_LOONGSON_3A.
40523 2017-01-19  Doug Gilmore  <doug.gilmore@imgtec.com>
40525         PR target/78176
40526         * config.gcc (supported_defaults): Add lxc1-sxc1.
40527         (with_lxc1_sxc1): Add validation.
40528         (all_defaults): Add lxc1-sxc1.
40529         * config/mips/mips.opt (mlxc1-sxc1): New option.
40530         * config/mips/mips.h (OPTION_DEFAULT_SPECS): Add a default for
40531         mlxc1-sxc1.
40532         (TARGET_CPU_CPP_BUILTINS): Add builtin_define for
40533         __mips_no_lxc1_sxc1.
40534         (ISA_HAS_LXC1_SXC1): Gate with mips_lxc1_sxc1.
40535         * doc/invoke.texi (-mlxc1-sxc1): Document the new option.
40536         * doc/install.texi (--with-lxc1-sxc1): Document the new option.
40538 2017-01-19  Richard Biener  <rguenther@suse.de>
40540         PR tree-optimization/72488
40541         * tree-ssa-sccvn.c (run_scc_vn): When we abort the VN make
40542         sure to restore SSA info.
40543         * tree-ssa.c (verify_ssa): Verify SSA info is not shared.
40545 2017-01-19  Richard Earnshaw  <rearnsha@arm.com>
40547         PR rtl-optimization/79121
40548         * expr.c (expand_expr_real_2, case LSHIFT_EXPR): Look at the signedness
40549         of the inner type when shifting an extended value.
40551 2017-01-17  Jan Hubicka  <hubicka@ucw.cz>
40553         PR lto/78407
40554         * symtab.c (symtab_node::equal_address_to): Fix comparing of
40555         interposable aliases.
40557 2017-01-18  Peter Bergner  <bergner@vnet.ibm.com>
40559         PR target/78516
40560         * config/rs6000/spe.md (mov_si<mode>_e500_subreg0): Fix constraints.
40561         Use the evmergelohi instruction.
40562         (mov_si<mode>_e500_subreg4_2_le): Likewise.
40563         (mov_sitf_e500_subreg8_2_be): Likewise.
40564         (mov_sitf_e500_subreg12_2_le): Likewise.
40565         (mov_si<mode>_e500_subreg0_2_le): Fix constraints.
40566         (mov_si<mode>_e500_subreg4_2_be): Likewise.
40567         (mov_sitf_e500_subreg8_2_le): Likewise.
40568         (mov_sitf_e500_subreg12_2_be): Likewise.
40570 2017-01-18  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
40572         * config/rs6000/altivec.md (altivec_vbpermq): Change "type"
40573         attribute from vecsimple to vecperm.
40574         (altivec_vbpermq2): Likewise.
40576 2017-01-18  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
40578         PR target/79040
40579         * config/rs6000/altivec.h: Fix typo of vec_cntlz to vec_cnttz.
40581 2017-01-18  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
40582         * config/rs6000/rs6000-protos.h (expand_strn_compare): Add arg.
40583         * config/rs6000/rs6000.c (expand_strn_compare): Add ability to expand
40584         strcmp. Fix bug where comparison didn't stop with zero byte. Fix
40585         case where N arg is SIZE_MAX.
40586         * config/rs6000/rs6000.md (cmpstrnsi): Args to expand_strn_compare.
40587         (cmpstrsi): Add pattern.
40589 2017-01-18  Michael Meissner  <meissner@linux.vnet.ibm.com>
40591         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
40592         __builtin_vec_revb builtins.
40593         * config/rs6000/rs6000-builtins.def (P9V_BUILTIN_XXBRQ_V16QI): Add
40594         built-in functions to support generation of the ISA 3.0 XXBR<x>
40595         vector byte reverse instructions.
40596         (P9V_BUILTIN_XXBRQ_V1TI): Likewise.
40597         (P9V_BUILTIN_XXBRD_V2DI): Likewise.
40598         (P9V_BUILTIN_XXBRD_V2DF): Likewise.
40599         (P9V_BUILTIN_XXBGW_V4SI): Likewise.
40600         (P9V_BUILTIN_XXBGW_V4SF): Likewise.
40601         (P9V_BUILTIN_XXBGH_V8HI): Likewise.
40602         (P9V_BUILTIN_VEC_REVB): Likewise.
40603         * config/rs6000/vsx.md (p9_xxbrq_v1ti): New insns/expanders to
40604         generate the ISA 3.0 XXBR<x> vector byte reverse instructions.
40605         (p9_xxbrq_v16qi): Likewise.
40606         (p9_xxbrd_<mode>, VSX_D iterator): Likewise.
40607         (p9_xxbrw_<mode>, VSX_W iterator): Likewise.
40608         (p9_xxbrh_v8hi): Likewise.
40609         * config/rs6000/altivec.h (vec_revb): Define if ISA 3.0.
40610         * doc/extend.texi (RS/6000 Altivec Built-ins): Document the
40611         vec_revb built-in functions.
40613 2017-01-18  Uros Bizjak  <ubizjak@gmail.com>
40615         PR rtl-optimization/78952
40616         * config/i386/i386.md (any_extract): New code iterator.
40617         (*insvqi_2): Use any_extract for source operand.
40618         (*insvqi_3): Use any_shiftrt for source operand.
40620 2017-01-18  Wilco Dijkstra  <wdijkstr@arm.com>
40622         * config/aarch64/aarch64.c (aarch64_sched_adjust_priority)
40623         New function.
40624         (TARGET_SCHED_ADJUST_PRIORITY): Define target hook.
40626 2017-01-18  Matthias Klose  <doko@ubuntu.com>
40628         * doc/install.texi: Allow default for --with-target-bdw-gc-include.
40630 2016-01-18  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
40632         * config/rs6000/altivec.h (vec_bperm): Change #define.
40633         * config/rs6000/altivec.md (UNSPEC_VBPERMD): New enum constant.
40634         (altivec_vbpermq2): New define_insn.
40635         (altivec_vbpermd): Likewise.
40636         * config/rs6000/rs6000-builtin.def (VBPERMQ2): New monomorphic
40637         function interface.
40638         (VBPERMD): Likewise.
40639         (VBPERM): New polymorphic function interface.
40640         * config/rs6000/r6000-c.c (altivec_overloaded_builtins_table):
40641         Add entries for P9V_BUILTIN_VEC_VBPERM.
40642         * doc/extend.texi: Add interfaces for vec_bperm.
40644 2017-01-18  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
40646         * config/s390/s390-c.c (s390_expand_overloaded_builtin): Downcase
40647         first letter of error messages.
40648         (s390_resolve_overloaded_builtin): Likewise.
40649         * config/s390/s390.c (s390_expand_builtin): Likewise.
40650         (s390_invalid_arg_for_unprototyped_fn): Likewise.
40651         (s390_valid_target_attribute_inner_p): Likewise.
40652         * config/s390/s390.md ("tabort"): Likewise.
40654 2017-01-18  Toma Tabacu  <toma.tabacu@imgtec.com>
40656         * config/mips/mips.h (ISA_HAS_DIV3): Remove unused macro.
40657         (ISA_AVOID_DIV_HILO): New macro.
40658         (ISA_HAS_DIV): Use new ISA_AVOID_DIV_HILO macro.
40659         (ISA_HAS_DDIV): Likewise.
40661 2017-01-18  Markus Trippelsdorf  <markus@trippelsdorf.de>
40663         * doc/invoke.texi (fabi-version): Correct number of occurrences.
40665 2017-01-18  Markus Trippelsdorf  <markus@trippelsdorf.de>
40667         * doc/invoke.texi (fabi-version): Spelling fix.
40669 2017-01-18  Markus Trippelsdorf  <markus@trippelsdorf.de>
40671         PR c++/70182
40672         * doc/invoke.texi (fabi-version): Mention mangling fix for
40673         operator names.
40675 2017-01-18  Markus Trippelsdorf  <markus@trippelsdorf.de>
40677         PR c++/77489
40678         * doc/invoke.texi (fabi-version): Document discriminator mangling.
40680 2017-01-17  Segher Boessenkool  <segher@kernel.crashing.org>
40682         PR target/78875
40683         * config/rs6000/rs6000-opts.h (stack_protector_guard): New enum.
40684         * config/rs6000/rs6000.c (rs6000_option_override_internal): Handle
40685         the new options.
40686         * config/rs6000/rs6000.md (stack_protect_set): Handle the new more
40687         flexible settings.
40688         (stack_protect_test): Ditto.
40689         * config/rs6000/rs6000.opt (mstack-protector-guard=,
40690         mstack-protector-guard-reg=, mstack-protector-guard-offset=): New
40691         options.
40692         * doc/invoke.texi (Option Summary) [RS/6000 and PowerPC Options]:
40693         Add -mstack-protector-guard=, -mstack-protector-guard-reg=, and
40694         -mstack-protector-guard-offset=.
40695         (RS/6000 and PowerPC Options): Ditto.
40697 2017-01-17  Uros Bizjak  <ubizjak@gmail.com>
40699         * config/i386/i386.h (MASK_CLASS_P): New define.
40700         * config/i386/i386.c (inline_secondary_memory_needed): Ensure that
40701         there are no registers from different register sets also when
40702         mask registers are used.  Update function comment.
40703         * config/i386/i386.md (*movsi_internal): Split (*k/*krm) alternative
40704         to (*k/*r) and (*k/*km) alternatives.
40706 2017-01-17  Wilco Dijkstra  <wdijkstr@arm.com>
40708         * config/aarch64/aarch64.md (eh_return): Remove pattern and splitter.
40709         * config/aarch64/aarch64.h (AARCH64_EH_STACKADJ_REGNUM): Remove.
40710         (EH_RETURN_HANDLER_RTX): New define.
40711         * config/aarch64/aarch64.c (aarch64_frame_pointer_required):
40712         Force frame pointer in EH return functions.
40713         (aarch64_expand_epilogue): Add barrier for eh_return.
40714         (aarch64_final_eh_return_addr): Remove.
40715         (aarch64_eh_return_handler_rtx): New function.
40716         * config/aarch64/aarch64-protos.h (aarch64_final_eh_return_addr):
40717         Remove.
40718         (aarch64_eh_return_handler_rtx): New prototype.
40720 2017-01-17  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
40722         * config/rs6000/altivec.h (vec_rlmi): New #define.
40723         (vec_vrlnm): Likewise.
40724         (vec_rlnm): Likewise.
40725         * config/rs6000/altivec.md (UNSPEC_VRLMI): New UNSPEC enum value.
40726         (UNSPEC_VRLNM): Likewise.
40727         (VIlong): New mode iterator.
40728         (altivec_vrl<VI_char>mi): New define_insn.
40729         (altivec_vrl<VI_char>nm): Likewise.
40730         * config/rs6000/rs6000-builtin.def (VRLWNM): New monomorphic
40731         function entry.
40732         (VRLDNM): Likewise.
40733         (RLNM): New polymorphic function entry.
40734         (VRLWMI): New monomorphic function entry.
40735         (VRLDMI): Likewise.
40736         (RLMI): New polymorphic function entry.
40737         * config/rs6000/r6000-c.c (altivec_overloaded_builtin_table): Add
40738         new entries for P9V_BUILTIN_VEC_RLMI and P9V_BUILTIN_VEC_RLNM.
40739         * doc/extend.texi: Add description of vec_rlmi, vec_rlnm, and
40740         vec_vrlnm.
40742 2017-01-17  Jakub Jelinek  <jakub@redhat.com>
40744         PR debug/78839
40745         * dwarf2out.c (field_byte_offset): Restore the
40746         PCC_BITFIELD_TYPE_MATTERS behavior for INTEGER_CST DECL_FIELD_OFFSET
40747         and DECL_FIELD_BIT_OFFSET.  Use fold_build2 instead of build2 + fold.
40748         (analyze_variants_discr, gen_variant_part): Use fold_build2 instead
40749         of build2 + fold.
40751 2017-01-17  Eric Botcazou  <ebotcazou@adacore.com>
40753         PR ada/67205
40754         * config/aarch64/aarch64.c (TARGET_CUSTOM_FUNCTION_DESCRIPTORS): Define
40756 2017-01-17  Jakub Jelinek  <jakub@redhat.com>
40758         PR debug/71669
40759         * dwarf2out.c (add_data_member_location_attribute): For constant
40760         offset bitfield emit for -gdwarf-5 DW_AT_data_bit_offset attribute
40761         instead of DW_AT_data_member_location, DW_AT_bit_offset and
40762         DW_AT_byte_size attributes.
40764 2017-01-17  Eric Botcazou  <ebotcazou@adacore.com>
40766         * config/rs6000/rs6000.c (rs6000_emit_move): Also use a TOC reference
40767         after forcing to constant memory when the code model is medium.
40769 2017-01-17  Julia Koval  <julia.koval@intel.com>
40771         PR target/76731
40772         * config/i386/avx512fintrin.h
40773         (_mm512_i32gather_ps): Change __addr type to void const*.
40774         (_mm512_mask_i32gather_ps): Ditto.
40775         (_mm512_i32gather_pd): Ditto.
40776         (_mm512_mask_i32gather_pd): Ditto.
40777         (_mm512_i64gather_ps): Ditto.
40778         (_mm512_mask_i64gather_ps): Ditto.
40779         (_mm512_i64gather_pd): Ditto.
40780         (_mm512_mask_i64gather_pd): Ditto.
40781         (_mm512_i32gather_epi32): Ditto.
40782         (_mm512_mask_i32gather_epi32): Ditto.
40783         (_mm512_i32gather_epi64): Ditto.
40784         (_mm512_mask_i32gather_epi64): Ditto.
40785         (_mm512_i64gather_epi32): Ditto.
40786         (_mm512_mask_i64gather_epi32): Ditto.
40787         (_mm512_i64gather_epi64): Ditto.
40788         (_mm512_mask_i64gather_epi64): Ditto.
40789         (_mm512_i32scatter_ps): Change __addr type to void*.
40790         (_mm512_mask_i32scatter_ps): Ditto.
40791         (_mm512_i32scatter_pd): Ditto.
40792         (_mm512_mask_i32scatter_pd): Ditto.
40793         (_mm512_i64scatter_ps): Ditto.
40794         (_mm512_mask_i64scatter_ps): Ditto.
40795         (_mm512_i64scatter_pd): Ditto.
40796         (_mm512_mask_i64scatter_pd): Ditto.
40797         (_mm512_i32scatter_epi32): Ditto.
40798         (_mm512_mask_i32scatter_epi32): Ditto.
40799         (_mm512_i32scatter_epi64): Ditto.
40800         (_mm512_mask_i32scatter_epi64): Ditto.
40801         (_mm512_i64scatter_epi32): Ditto.
40802         (_mm512_mask_i64scatter_epi32): Ditto.
40803         (_mm512_i64scatter_epi64): Ditto.
40804         (_mm512_mask_i64scatter_epi64): Ditto.
40805         * config/i386/avx512pfintrin.h
40806         (_mm512_mask_prefetch_i32gather_pd): Change __addr type to void const*.
40807         (_mm512_mask_prefetch_i32gather_ps): Ditto.
40808         (_mm512_mask_prefetch_i64gather_pd): Ditto.
40809         (_mm512_mask_prefetch_i64gather_ps): Ditto.
40810         (_mm512_prefetch_i32scatter_pd): Change __addr type to void*.
40811         (_mm512_prefetch_i32scatter_ps): Ditto.
40812         (_mm512_mask_prefetch_i32scatter_pd): Ditto.
40813         (_mm512_mask_prefetch_i32scatter_ps): Ditto.
40814         (_mm512_prefetch_i64scatter_pd): Ditto.
40815         (_mm512_prefetch_i64scatter_ps): Ditto.
40816         (_mm512_mask_prefetch_i64scatter_pd): Ditto.
40817         (_mm512_mask_prefetch_i64scatter_ps): Ditto.
40818         * config/i386/avx512vlintrin.h
40819         (_mm256_mmask_i32gather_ps): Change __addr type to void const*.
40820         (_mm_mmask_i32gather_ps): Ditto.
40821         (_mm256_mmask_i32gather_pd): Ditto.
40822         (_mm_mmask_i32gather_pd): Ditto.
40823         (_mm256_mmask_i64gather_ps): Ditto.
40824         (_mm_mmask_i64gather_ps): Ditto.
40825         (_mm256_mmask_i64gather_pd): Ditto.
40826         (_mm_mmask_i64gather_pd): Ditto.
40827         (_mm256_mmask_i32gather_epi32): Ditto.
40828         (_mm_mmask_i32gather_epi32): Ditto.
40829         (_mm256_mmask_i32gather_epi64): Ditto.
40830         (_mm_mmask_i32gather_epi64): Ditto.
40831         (_mm256_mmask_i64gather_epi32): Ditto.
40832         (_mm_mmask_i64gather_epi32): Ditto.
40833         (_mm256_mmask_i64gather_epi64): Ditto.
40834         (_mm_mmask_i64gather_epi64): Ditto.
40835         (_mm256_i32scatter_ps): Change __addr type to void*.
40836         (_mm256_mask_i32scatter_ps): Ditto.
40837         (_mm_i32scatter_ps): Ditto.
40838         (_mm_mask_i32scatter_ps): Ditto.
40839         (_mm256_i32scatter_pd): Ditto.
40840         (_mm256_mask_i32scatter_pd): Ditto.
40841         (_mm_i32scatter_pd): Ditto.
40842         (_mm_mask_i32scatter_pd): Ditto.
40843         (_mm256_i64scatter_ps): Ditto.
40844         (_mm256_mask_i64scatter_ps): Ditto.
40845         (_mm_i64scatter_ps): Ditto.
40846         (_mm_mask_i64scatter_ps): Ditto.
40847         (_mm256_i64scatter_pd): Ditto.
40848         (_mm256_mask_i64scatter_pd): Ditto.
40849         (_mm_i64scatter_pd): Ditto.
40850         (_mm_mask_i64scatter_pd): Ditto.
40851         (_mm256_i32scatter_epi32): Ditto.
40852         (_mm256_mask_i32scatter_epi32): Ditto.
40853         (_mm_i32scatter_epi32): Ditto.
40854         (_mm_mask_i32scatter_epi32): Ditto.
40855         (_mm256_i32scatter_epi64): Ditto.
40856         (_mm256_mask_i32scatter_epi64): Ditto.
40857         (_mm_i32scatter_epi64): Ditto.
40858         (_mm_mask_i32scatter_epi64): Ditto.
40859         (_mm256_i64scatter_epi32): Ditto.
40860         (_mm256_mask_i64scatter_epi32): Ditto.
40861         (_mm_i64scatter_epi32): Ditto.
40862         (_mm_mask_i64scatter_epi32): Ditto.
40863         (_mm256_i64scatter_epi64): Ditto.
40864         (_mm256_mask_i64scatter_epi64): Ditto.
40865         (_mm_i64scatter_epi64): Ditto.
40866         (_mm_mask_i64scatter_epi64): Ditto.
40867         * config/i386/i386-builtin-types.def (V16SF_V16SF_PCFLOAT_V16SI_HI_INT)
40868         (V8DF_V8DF_PCDOUBLE_V8SI_QI_INT, V8SF_V8SF_PCFLOAT_V8DI_QI_INT)
40869         (V8DF_V8DF_PCDOUBLE_V8DI_QI_INT, V16SI_V16SI_PCINT_V16SI_HI_INT)
40870         (V8DI_V8DI_PCINT64_V8SI_QI_INT, V8SI_V8SI_PCINT_V8DI_QI_INT)
40871         (V8DI_V8DI_PCINT64_V8DI_QI_INT, V2DF_V2DF_PCDOUBLE_V4SI_QI_INT)
40872         (V4DF_V4DF_PCDOUBLE_V4SI_QI_INT, V2DF_V2DF_PCDOUBLE_V2DI_QI_INT)
40873         (V4DF_V4DF_PCDOUBLE_V4DI_QI_INT, V4SF_V4SF_PCFLOAT_V4SI_QI_INT)
40874         (V8SF_V8SF_PCFLOAT_V8SI_QI_INT, V4SF_V4SF_PCFLOAT_V2DI_QI_INT)
40875         (V4SF_V4SF_PCFLOAT_V4DI_QI_INT, V2DI_V2DI_PCINT64_V4SI_QI_INT)
40876         (V4DI_V4DI_PCINT64_V4SI_QI_INT, V2DI_V2DI_PCINT64_V2DI_QI_INT)
40877         (V4DI_V4DI_PCINT64_V4DI_QI_INT, V4SI_V4SI_PCINT_V4SI_QI_INT)
40878         (V8SI_V8SI_PCINT_V8SI_QI_INT, V4SI_V4SI_PCINT_V2DI_QI_INT)
40879         (V4SI_V4SI_PCINT_V4DI_QI_INT, VOID_PFLOAT_HI_V16SI_V16SF_INT)
40880         (VOID_PFLOAT_QI_V8SI_V8SF_INT, VOID_PFLOAT_QI_V4SI_V4SF_INT)
40881         (VOID_PDOUBLE_QI_V8SI_V8DF_INT, VOID_PDOUBLE_QI_V4SI_V4DF_INT)
40882         (VOID_PDOUBLE_QI_V4SI_V2DF_INT, VOID_PFLOAT_QI_V8DI_V8SF_INT)
40883         (VOID_PFLOAT_QI_V4DI_V4SF_INT, VOID_PFLOAT_QI_V2DI_V4SF_INT)
40884         (VOID_PDOUBLE_QI_V8DI_V8DF_INT, VOID_PDOUBLE_QI_V4DI_V4DF_INT)
40885         (VOID_PDOUBLE_QI_V2DI_V2DF_INT, VOID_PINT_HI_V16SI_V16SI_INT)
40886         (VOID_PINT_QI_V8SI_V8SI_INT, VOID_PINT_QI_V4SI_V4SI_INT)
40887         (VOID_PLONGLONG_QI_V8SI_V8DI_INT, VOID_PLONGLONG_QI_V4SI_V4DI_INT)
40888         (VOID_PLONGLONG_QI_V4SI_V2DI_INT, VOID_PINT_QI_V8DI_V8SI_INT)
40889         (VOID_PINT_QI_V4DI_V4SI_INT, VOID_PINT_QI_V2DI_V4SI_INT)
40890         (VOID_PLONGLONG_QI_V8DI_V8DI_INT, VOID_QI_V8SI_PCINT64_INT_INT)
40891         (VOID_PLONGLONG_QI_V4DI_V4DI_INT, VOID_PLONGLONG_QI_V2DI_V2DI_INT)
40892         (VOID_HI_V16SI_PCINT_INT_INT, VOID_QI_V8DI_PCINT64_INT_INT)
40893         (VOID_QI_V8DI_PCINT_INT_INT): Remove.
40894         (V16SF_V16SF_PCVOID_V16SI_HI_INT,  V8DF_V8DF_PCVOID_V8SI_QI_INT)
40895         (V8SF_V8SF_PCVOID_V8DI_QI_INT, V8DF_V8DF_PCVOID_V8DI_QI_INT)
40896         (V16SI_V16SI_PCVOID_V16SI_HI_INT, V8DI_V8DI_PCVOID_V8SI_QI_INT)
40897         (V8SI_V8SI_PCVOID_V8DI_QI_INT, V8DI_V8DI_PCVOID_V8DI_QI_INT)
40898         (VOID_PVOID_HI_V16SI_V16SF_INT, VOID_PVOID_QI_V8SI_V8DF_INT)
40899         (VOID_PVOID_QI_V8DI_V8SF_INT, VOID_PVOID_QI_V8DI_V8DF_INT)
40900         (VOID_PVOID_HI_V16SI_V16SI_INT, VOID_PVOID_QI_V8SI_V8DI_INT)
40901         (VOID_PVOID_QI_V8DI_V8SI_INT, VOID_PVOID_QI_V8DI_V8DI_INT)
40902         (V2DF_V2DF_PCVOID_V4SI_QI_INT, V4DF_V4DF_PCVOID_V4SI_QI_INT)
40903         (V2DF_V2DF_PCVOID_V2DI_QI_INT, V4DF_V4DF_PCVOID_V4DI_QI_INT
40904         (V4SF_V4SF_PCVOID_V4SI_QI_INT, V8SF_V8SF_PCVOID_V8SI_QI_INT)
40905         (V4SF_V4SF_PCVOID_V2DI_QI_INT, V4SF_V4SF_PCVOID_V4DI_QI_INT)
40906         (V2DI_V2DI_PCVOID_V4SI_QI_INT, V4DI_V4DI_PCVOID_V4SI_QI_INT)
40907         (V2DI_V2DI_PCVOID_V2DI_QI_INT, V4DI_V4DI_PCVOID_V4DI_QI_INT)
40908         (V4SI_V4SI_PCVOID_V4SI_QI_INT, V8SI_V8SI_PCVOID_V8SI_QI_INT)
40909         (V4SI_V4SI_PCVOID_V2DI_QI_INT, V4SI_V4SI_PCVOID_V4DI_QI_INT)
40910         (VOID_PVOID_QI_V8SI_V8SF_INT, VOID_PVOID_QI_V4SI_V4SF_INT)
40911         (VOID_PVOID_QI_V4SI_V4DF_INT, VOID_PVOID_QI_V4SI_V2DF_INT)
40912         (VOID_PVOID_QI_V4DI_V4SF_INT, VOID_PVOID_QI_V2DI_V4SF_INT)
40913         (VOID_PVOID_QI_V4DI_V4DF_INT, VOID_PVOID_QI_V2DI_V2DF_INT)
40914         (VOID_PVOID_QI_V8SI_V8SI_INT, VOID_PVOID_QI_V4SI_V4SI_INT)
40915         (VOID_PVOID_QI_V4SI_V4DI_INT, VOID_PVOID_QI_V4SI_V2DI_INT)
40916         (VOID_PVOID_QI_V4DI_V4SI_INT, VOID_PVOID_QI_V2DI_V4SI_INT)
40917         (VOID_PVOID_QI_V4DI_V4DI_INT, VOID_PVOID_QI_V2DI_V2DI_INT)
40918         (VOID_QI_V8SI_PCVOID_INT_INT, VOID_HI_V16SI_PCVOID_INT_INT)
40919         (VOID_QI_V8DI_PCVOID_INT_INT): Add.
40920         * config/i386/i386.c (ix86_init_mmx_sse_builtins): Adjust builtin
40921         definitions accordingly.
40923 2017-01-17  Kito Cheng  <kito.cheng@gmail.com>
40924             Kuan-Lin Chen  <kuanlinchentw@gmail.com>
40926         PR target/79079
40927         * internal-fn.c (expand_mul_overflow): Use convert_modes instead of
40928         gen_lowpart.
40930 2017-01-17  Vladimir Makarov  <vmakarov@redhat.com>
40932         PR target/79058
40933         * ira-conflicts.c (ira_build_conflicts): Update total conflict
40934         hard regs for inner regno.
40936 2017-01-17  Martin Liska  <mliska@suse.cz>
40938         PR ipa/71207
40939         * ipa-polymorphic-call.c (contains_type_p): Fix wrong
40940         assumption and add comment.
40942 2017-01-17  Nathan Sidwell  <nathan@acm.org>
40944         * ipa-visibility.c (localize_node): New function, broken out of ...
40945         (function_and_variable_visibility): ... here. Call it.
40947 2017-01-17  Jan Hubicka  <hubicka@ucw.cz>
40949         PR middle-end/77445
40950         * tree-ssa-threadupdate.c (remove_ctrl_stmt_and_useless_edges):
40951         correctly set frequency of oudgoing edge.
40952         (duplicate_thread_path): Fix profile updating.
40954 2017-01-17  Jakub Jelinek  <jakub@redhat.com>
40956         PR other/79046
40957         * configure.ac: Add GCC_BASE_VER.
40958         * Makefile.in (version): Use @get_gcc_base_ver@ instead of cat to get
40959         version from BASE-VER file.
40960         (CFLAGS-gcc.o): Add -DBASEVER=$(BASEVER_s).
40961         (gcc.o): Depend on $(BASEVER).
40962         * common.opt (dumpfullversion): New option.
40963         * gcc.c (driver_handle_option): Handle OPT_dumpfullversion.
40964         * doc/invoke.texi: Document -dumpfullversion.
40965         * doc/install.texi: Document --with-gcc-major-version-only.
40966         * configure: Regenerated.
40968 2017-01-17  Richard Biener  <rguenther@suse.de>
40970         PR tree-optimization/71433
40971         * tree-vrp.c (register_new_assert_for): Merge same asserts
40972         on all incoming edges.
40973         (process_assert_insertions_for): Handle insertions at the
40974         beginning of BBs.
40976 2017-01-17  Gerald Pfeifer  <gerald@pfeifer.com>
40978         * config/i386/cygwin.h (LIBGCJ_SONAME): Remove.
40979         * config/i386/mingw32.h (LIBGCJ_SONAME): Remove.
40981 2017-01-17  Kaz Kojima  <kkojima@gcc.gnu.org>
40983         PR target/78633
40984         * config/sh/sh.md (cmpeqsi_t+1): Call copy_rtx to avoid invalid
40985         RTL sharing.
40987 2017-01-17  Alan Modra  <amodra@gmail.com>
40989         PR target/79066
40990         * config/rs6000/rs6000.md (elf_high, elf_low): Disable when pic.
40991         * config/rs6000/rs6000.c (rs6000_emit_allocate_stack): Don't allow
40992         symbolic stack limit when pic.
40994 2017-01-16  Martin Sebor  <msebor@redhat.com>
40996         PR tree-optimization/78608
40997         * gimple-ssa-sprintf.c (tree_digits): Avoid negating TYPE_MIN.
40999 2017-01-16  Jeff Law  <law@redhat.com>
41001         Revert:
41002         2016-12-02  Tadek Kijkowski  <tkijkowski@gmail.com>
41003         * Makefile.in (PREPROCESSOR_DEFINES): Add a level of indirection
41004         for several include directories that may be relative to sysroot.
41005         * config/i386/x-mingw32 (gplus_includedir): Define.
41006         (gplus_tool_includedir, gplus_backward_include_dir): Likewise.
41007         (native_system_includedir): Likewise.
41008         * config/i386/mingw32.h (STANDARD_STARTFILE_PREFIX_1): Do not
41009         override if TARGET_SYSTEM_ROOT is defined.
41010         (NATIVE_SYSTEM_HEADER_DIR): Likewise.
41012         PR tree-optimization/79090
41013         PR tree-optimization/33562
41014         PR tree-optimization/61912
41015         PR tree-optimization/77485
41016         * tree-ssa-dse.c (compute_trims): Accept STMT argument.  Dump STMT
41017         and computed trims into the dump file.
41019 2017-01-17  Uros Bizjak  <ubizjak@gmail.com>
41021         * config/i386/i386.h (LIMIT_RELOAD_CLASS): Remove.
41023 2017-01-16  Jakub Jelinek  <jakub@redhat.com>
41025         PR c/79089
41026         * gimplify.c (gimplify_init_constructor): If want_value and
41027         object == lhs, unshare lhs to avoid invalid tree sharing.  Formatting
41028         fix.
41030         PR target/79080
41031         * loop-doloop.c (doloop_modify): Call unshare_all_rtl_in_chain on
41032         sequence.  Formatting fixes.
41033         (doloop_optimize): Formatting fixes.
41035         PR driver/49726
41036         * gcc.c (debug_level_greater_than_spec_func): New function.
41037         (static_spec_functions): Add debug-level-gt spec function.
41038         (ASM_DEBUG_SPEC, cpp_options): Use %:debug-level-gt(0) instead of
41039         !g0.
41040         * config/darwin.h (DSYMUTIL_SPEC, ASM_DEBUG_SPEC): Likewise.
41041         * config/darwin9.h (DSYMUTIL_SPEC, ASM_DEBUG_SPEC): Likewise.
41042         * common.opt (g, gcoff, gdwarf, gdwarf-, ggdb, gno-pubnames,
41043         gpubnames, ggnu-pubnames, gno-record-gcc-switches,
41044         grecord-gcc-switches, gno-strict-dwarf, gstrict-dwarf, gstabs,
41045         gstabs+, gtoggle, gvms, gxcoff, gxcoff+): Add Driver flag.
41047 2017-01-16  Uros Bizjak  <ubizjak@gmail.com>
41049         * config/i386/i386.h (HARD_REGNO_CALLER_SAVE_MODE): Apply HImode and
41050         QImode fixups to general and mask registers only.
41052 2017-01-16  Carl Love  <cel@us.ibm.com>
41054         * config/rs6000/rs6000-c (altivec_overloaded_builtins): Add support
41055         for built-in functions
41056         vector signed char vec_nabs (vector signed char)
41057         vector signed short vec_nabs (vector signed short)
41058         vector signed int vec_nabs (vector signed int)
41059         vector signed long long vec_nabs (vector signed long long)
41060         vector float vec_nabs (vector float)
41061         vector double vec_nabs (vector double)
41062         * config/rs6000/rs6000-builtin.def: Add definitions for NABS functions
41063         and NABS overload.
41064         * config/rs6000/altivec.md: New define_expand nabs<mode>2 types
41065         * config/rs6000/altivec.h: New define for vec_nabs built-in function.
41066         * doc/extend.texi: Update the documentation file for the new built-in
41067         functions.
41069 2017-01-16  Martin Sebor  <msebor@redhat.com>
41071         * gimple-ssa-sprintf.c (format_directive): Correct a typo in a warning
41072         message.
41074 2017-01-16  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
41076         * config/rs6000/rs6000.c (rtx_is_swappable_p): Change
41077         UNSPEC_VSX__XXSPLTD to require special splat handling.
41079 2017-01-16  David Malcolm  <dmalcolm@redhat.com>
41081         PR bootstrap/78616
41082         * system.h: Poison strndup.
41084 2017-01-16  Alan Modra  <amodra@gmail.com>
41086         PR target/79098
41087         * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Don't
41088         use a switch.
41090 2017-01-16  Georg-Johann Lay  <avr@gjlay.de>
41092         * config/avr/avr.h (BRANCH_COST) [reload_completed]: Increase by 4.
41094 2017-01-15  Uros Bizjak  <ubizjak@gmail.com>
41096         * config/i386/i386.c (ix86_legitimate_combined_insn): Do not
41097         call recog here.  Assert that INSN_CODE (insn) is non-negative.
41099 2017-01-15  Segher Boessenkool  <segher@kernel.crashing.org>
41101         PR target/72749
41102         * cfgrtl.c (rtl_split_edge): Also patch jump insns that jump to the
41103         fallthrough.
41104         * haifa-sched.c (dump_insn_stream): Don't crash if there is a label
41105         in the currently scheduled RTL fragment.
41107 2017-01-15  Segher Boessenkool  <segher@kernel.crashing.org>
41109         PR rtl-optimization/78751
41110         * ifcvt.c (find_cond_trap): If we generated a non-existing insn,
41111         give up.
41113 2017-01-14  Jeff Law  <law@redhat.com>
41115         PR tree-optimization/79090
41116         * tree-ssa-dse.c (valid_ao_ref_for_dse): Reject zero length and
41117         variable length stores.
41118         (compute_trims): Delete dead assignment to *trim_tail.
41119         (dse_dom_walker::dse_optimize_stmt): Optimize mem* calls with
41120         zero length.
41122 2017-01-14  Bernd Schmidt  <bschmidt@redhat.com>
41124         PR rtl-optimization/78626
41125         PR rtl-optimization/78727
41126         * cprop.c (one_cprop_pass): Collect unconditional traps in the middle
41127         of a block, and split such blocks after everything else is finished.
41129 2017-01-14  Alan Modra  <amodra@gmail.com>
41131         PR target/72749
41132         * combine.c (recog_for_combine_1): Set INSN_CODE before calling
41133         target legitimate_combined_insn.
41134         * config/rs6000/rs6000.c (TARGET_LEGITIMATE_COMBINED_INSN): Define.
41135         (rs6000_legitimate_combined_insn): New function.
41136         * config/rs6000/rs6000.md (UNSPEC_DOLOOP): Delete, and remove
41137         all uses.
41138         (ctr<mode>_internal3): Rename from *ctr<mode>_internal5.
41139         (ctr<mode>_internal4): Rename from *ctr<mode>_internal6.
41140         (ctr<mode>_internal1, ctr<mode>_internal2): Remove '*' from name.
41142 2017-01-14  Gerald Pfeifer  <gerald@pfeifer.com>
41144         * doc/frontends.texi (G++ and GCC): Remove references to Java.
41146 2017-01-13  Jeff Law  <law@redhat.com>
41148         PR tree-optimization/33562
41149         PR tree-optimization/61912
41150         PR tree-optimization/77485
41151         * tree-ssa-dse.c (delete_dead_call): Accept gsi rather than
41152         a statement.
41153         (delete_dead_assignment): Likewise.
41154         (dse_dom_walker::dse_optimize_stmt): Pass in the gsi rather than
41155         statement to delete_dead_call and delete_dead_assignment.
41157 2017-01-13  David Malcolm  <dmalcolm@redhat.com>
41159         PR c/78304
41160         * substring-locations.c (format_warning_va): Strengthen case 1 so
41161         that both endpoints of the substring must be within the format
41162         range for just the substring to be printed.
41164 2017-01-13  Uros Bizjak  <ubizjak@gmail.com>
41166         * config/i386/i386.opt (msgx): Use ix86_isa_flags2 variable.
41167         * config/i386/i386.c (ix86_target_string): Add missing options
41168         to isa_opts and reorder options by implied ISAs.  Rename isa_opts2 to
41169         isa2_opts, ix86_flag_opts to flag2_opts, ix86_target_other to
41170         flags_other and ix86_target_other to flags2_other.  Display unknown
41171         isa2 options.
41172         (ix86_valid_target_attribute_inner_p): Add missing options and
41173         reorder options by implied ISAs, as in ix86_target_string.
41175 2017-01-13  Richard Sandiford  <richard.sandiford@arm.com>
41177         * hash-table.h (hash_table::too_empty_p): New function.
41178         (hash_table::expand): Use it.
41179         (hash_table::traverse): Likewise.
41180         (hash_table::empty_slot): Use sizeof (value_type) instead of
41181         sizeof (PTR) to convert bytes to elements.  Shrink the table
41182         if the current size is excessive for the current number of
41183         elements.
41185 2017-01-13  Richard Sandiford  <richard.sandiford@arm.com>
41187         * ira-costs.c (record_reg_classes): Break from the inner loop
41188         early once alt_fail is known to be true.  Update outer loop
41189         handling accordingly.
41191 2017-01-13  Jeff Law  <law@redhat.com>
41193         * tree-ssa-dse.c (decrement_count): New function.
41194         (increment_start_addr, maybe_trim_memstar_call): Likewise.
41195         (dse_dom_walker::optimize_stmt): Call maybe_trim_memstar_call directly
41196         when we know the partially dead statement is a mem* function.
41198         PR tree-optimization/61912
41199         PR tree-optimization/77485
41200         * tree-ssa-dse.c: Include expr.h.
41201         (maybe_trim_constructor_store): New function.
41202         (maybe_trim_partially_dead_store): Call maybe_trim_constructor_store.
41204         PR tree-optimization/33562
41205         PR tree-optimization/61912
41206         PR tree-optimization/77485
41207         * doc/invoke.texi: Document new dse-max-object-size param.
41208         * params.def (PARM_DSE_MAX_OBJECT_SIZE): New PARAM.
41209         * tree-ssa-dse.c: Include params.h.
41210         (dse_store_status): New enum.
41211         (initialize_ao_ref_for_dse): New, partially extracted from
41212         dse_optimize_stmt.
41213         (valid_ao_ref_for_dse, normalize_ref): New.
41214         (setup_live_bytes_from_ref, compute_trims): Likewise.
41215         (clear_bytes_written_by, maybe_trim_complex_store): Likewise.
41216         (maybe_trim_partially_dead_store): Likewise.
41217         (maybe_trim_complex_store): Likewise.
41218         (dse_classify_store): Renamed from dse_possibly_dead_store_p.
41219         Track what bytes live from the original store.  Return tri-state
41220         for dead, partially dead or live.
41221         (dse_dom_walker): Add constructor, destructor and new private members.
41222         (delete_dead_call, delete_dead_assignment): New extracted from
41223         dse_optimize_stmt.
41224         (dse_optimize_stmt): Make a member of dse_dom_walker.
41225         Use initialize_ao_ref_for_dse.
41227         PR tree-optimization/33562
41228         PR tree-optimization/61912
41229         PR tree-optimization/77485
41230         * sbitmap.h (bitmap_count_bits): Prototype.
41231         (bitmap_clear_range, bitmap_set_range): Likewise.
41232         * sbitmap.c (bitmap_clear_range): New function.
41233         (bitmap_set_range, sbitmap_popcount, bitmap_count_bits): Likewise.
41235 2017-01-13  Martin Liska  <mliska@suse.cz>
41237         PR ipa/79043
41238         * function.c (set_cfun): Add new argument force.
41239         * function.h (set_cfun): Likewise.
41240         * ipa-inline-transform.c (inline_call): Use the function when
41241         strict alising from is dropped for function we inline to.
41243 2017-01-13  Richard Biener  <rguenther@suse.de>
41245         * tree-pretty-print.c (dump_generic_node): Fix inverted condition
41246         for dumping GIMPLE INTEGER_CSTs.
41248 2017-01-13  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
41250         * config/sol2.h (TARGET_OS_CPP_BUILTINS): Define __STDC_VERSION__
41251         to 201112L since C++17.
41253 2017-01-13  Maxim Ostapenko  <m.ostapenko@samsung.com>
41255         PR sanitizer/78887
41256         * asan.c (asan_needs_odr_indicator_p): Don't emit ODR indicators
41257         if -fsanitize=kernel-address is present.
41259 2017-01-13  Richard Biener  <rguenther@suse.de>
41261         * tree-pretty-print.c (dump_generic_node): Dump INTEGER_CSTs
41262         as _Literal ( type ) number in case usual suffixes do not
41263         preserve all information.
41265 2017-01-13  Richard Biener  <rguenther@suse.de>
41267         PR tree-optimization/77283
41268         * gimple-ssa-split-paths.c: Include gimple-ssa.h, tree-phinodes.h
41269         and ssa-iterators.h.
41270         (is_feasible_trace): Implement a cost model based on joiner
41271         PHI node uses.
41273 2017-01-12  Michael Meissner  <meissner@linux.vnet.ibm.com>
41275         PR target/79004
41276         * config/rs6000/rs6000.md (FP_ISA3): Do not optimize converting
41277         char or short to __float128/_Float128 directly.
41279 2017-01-12  Martin Sebor  <msebor@redhat.com>
41281         to -Wformat-overflow.
41282         * gimple-ssa-sprintf.c (pass_sprintf_length::gate): Adjust.
41283         (min_bytes_remaining): Same.
41284         (get_string_length): Same.
41285         (format_string): Same.
41286         (format_directive): Same.
41287         (add_bytes): Same.
41288         (pass_sprintf_length::handle_gimple_call): Same.
41290 2017-01-12  Jakub Jelinek  <jakub@redhat.com>
41292         * gimple-ssa-sprintf.c (try_substitute_return_value): Remove
41293         info.nowrite calls with no lhs that can't throw.  Return bool
41294         whether gsi_remove has been called or not.
41295         (pass_sprintf_length::handle_gimple_call): Return bool whether
41296         try_substitute_return_value called gsi_remove.  Formatting fix.
41297         (pass_sprintf_length::execute): Don't use gsi_remove if
41298         handle_gimple_call returned true.
41300         PR bootstrap/79069
41301         * cfgrtl.c (rtl_tidy_fallthru_edge): For any_uncondjump_p that can't
41302         be removed due to side-effects, don't remove following barrier nor
41303         turn the successor edge into fallthru edge.
41305 2017-01-12  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
41307         PR target/79044
41308         * config/rs6000/rs6000.c (insn_is_swappable_p): Mark
41309         element-reversing loads and stores as not swappable.
41311 2017-01-12  Nathan Sidwell  <nathan@acm.org>
41312             Nicolai Stange  <nicstange@gmail.com>
41314         * combine.c (try_combine): Don't ignore result of overlap checking
41315         loop.  Combine overlap & asm check into single loop.
41317 2017-01-12  Richard Biener  <rguenther@suse.de>
41319         * tree-pretty-print.c (dump_generic_node): Provide -gimple
41320         variant for MEM_REF.  Sanitize INTEGER_CST for -gimple.
41322 2017-01-12  Richard Biener  <rguenther@suse.de>
41324         * tree.c (initialize_tree_contains_struct): Make TS_OPTIMIZATION
41325         and TS_TARGET_OPTION directly derive from TS_BASE.
41326         * tree-core.h (tree_optimization_option): Derive from tree_base.
41327         (tree_target_option): Likewise.
41329 2017-01-11  Uros Bizjak  <ubizjak@gmail.com>
41331         * config/i386/i386.c (memory_address_length): Increase len
41332         only when rip_relative_addr_p returns false.
41334 2017-01-11  Julia Koval  <julia.koval@intel.com>
41336         * common/config/i386/i386-common.c (OPTION_MASK_ISA_SGX_UNSET): New.
41337         (OPTION_MASK_ISA_SGX_SET): New.
41338         (ix86_handle_option): Handle OPT_msgx.
41339         * config.gcc: Added sgxintrin.h.
41340         * config/i386/driver-i386.c (host_detect_local_cpu): Detect sgx.
41341         * config/i386/i386-c.c (ix86_target_macros_internal): Define __SGX__.
41342         * config/i386/i386.c (ix86_target_string): Add -msgx.
41343         (PTA_SGX): New.
41344         (ix86_option_override_internal): Handle new options.
41345         (ix86_valid_target_attribute_inner_p): Add sgx.
41346         * config/i386/i386.h (TARGET_SGX, TARGET_SGX_P): New.
41347         * config/i386/i386.opt: Add msgx.
41348         * config/i386/sgxintrin.h: New file.
41349         * config/i386/x86intrin.h: Add sgxintrin.h.
41351 2017-01-11  Jakub Jelinek  <jakub@redhat.com>
41353         PR c++/71537
41354         * fold-const.c (maybe_nonzero_address): Return 1 for function
41355         local objects.
41356         (tree_single_nonzero_warnv_p): Don't handle function local objects
41357         here.
41359         PR c++/72813
41360         * gcc.c (default_compilers): Don't add -o %g.s for -S -save-temps
41361         of c-header.
41363 2017-01-11  David Malcolm  <dmalcolm@redhat.com>
41365         PR driver/78877
41366         * opts.c: Include "spellcheck.h"
41367         (struct string_fragment): New struct.
41368         (struct edit_distance_traits<const string_fragment &>): New
41369         struct.
41370         (get_closest_sanitizer_option): New function.
41371         (parse_sanitizer_options): Offer suggestions for unrecognized arguments.
41373 2017-01-11  Jakub Jelinek  <jakub@redhat.com>
41375         * dwarf2out.c (DWARF_COMPILE_UNIT_HEADER_SIZE): For DWARF5 decrease
41376         by 12.
41377         (DWARF_COMDAT_TYPE_UNIT_HEADER_SIZE): Always
41378         DWARF_COMPILE_UNIT_HEADER_SIZE plus 12.
41379         (DWARF_COMPILE_UNIT_SKELETON_HEADER_SIZE): Define.
41380         (calc_base_type_die_sizes): Use DWARF_COMPILE_UNIT_SKELETON_HEADER_SIZE
41381         for initial die_offset if dwarf_split_debug_info.
41382         (output_comp_unit): Use DWARF_COMPILE_UNIT_SKELETON_HEADER_SIZE for
41383         initial next_die_offset if dwo_id is non-NULL.  Don't emit padding
41384         fields.
41385         (output_skeleton_debug_sections): Formatting fix.  Use
41386         DWARF_COMPILE_UNIT_SKELETON_HEADER_SIZE instead of
41387         DWARF_COMPILE_UNIT_HEADER_SIZE.  Don't emit padding.
41389 2017-01-11  Wilco Dijkstra  <wdijkstr@arm.com>
41391         * config/arm/cortex-a53.md: Add bypasses for
41392         cortex_a53_r2f_cvt.
41393         (cortex_a53_r2f): Only use for transfers.
41394         (cortex_a53_f2r): Likewise.
41395         (cortex_a53_r2f_cvt): Add reservation for conversions.
41396         (cortex_a53_f2r_cvt): Likewise.
41398 2017-01-11  Tamar Christina  <tamar.christina@arm.com>
41400         * config/arm/arm_neon.h: Add __artificial__ and gnu_inline
41401         to all inlined functions, change static to extern.
41403 2017-01-11  Christophe Lyon  <christophe.lyon@linaro.org>
41405         PR target/78253
41406         * config/arm/arm.c (legitimize_pic_address): Handle reference to
41407         weak symbol.
41408         (arm_assemble_integer): Likewise.
41410 2017-01-11  Richard Earnshaw  <rearnsha@arm.com>
41412         * config.gcc: Use new awk script to check CPU, FPU and architecture
41413         parameters for --with-... options.
41414         * config/arm/parsecpu.awk: New file
41415         * config/arm/arm-cpus.in: New file.
41416         * config/arm/arm-opts.h: Include arm-cpu.h instead of processing .def
41417         files.
41418         * config/arm/arm.c: Include arm-cpu-data.h instead of processing .def
41419         files.
41420         * config/arm/t-arm: Update dependency rules.
41421         * common/config/arm/arm-common.c: Include arm-cpu-cdata.h instead
41422         of processing .def files.
41423         * config/arm/genopt.sh: Deleted.
41424         * config/arm/gentune.sh: Deleted.
41425         * config/arm/arm-cores.def: Deleted.
41426         * config/arm/arm-arches.def: Deleted.
41427         * config/arm/arm-fpus.def: Deleted.
41428         * config/arm/arm-tune.md: Regenerated.
41429         * config/arm/arm-tables.opt: Regenerated.
41430         * config/arm/arm-cpu.h: New generated file.
41431         * config/arm/arm-cpu-data.h: New generated file.
41432         * config/arm/arm-cpu-cdata.h: New generated file.
41434 2017-01-11  Maxim Ostapenko  <m.ostapenko@samsung.com>
41436         PR lto/79042
41437         * lto-cgraph.c (lto_output_varpool_node): Pack dynamically_initialized
41438         bit.
41439         (input_varpool_node): Unpack dynamically_initialized bit.
41441 2017-01-11  Eric Botcazou  <ebotcazou@adacore.com>
41443         PR rtl-optimization/79032
41444         * lra-constraints.c (simplify_operand_subreg): In the MEM case, test
41445         the alignment of the adjusted memory reference against that of MODE,
41446         instead of the alignment of the original memory reference.
41448 2017-01-11  Martin Jambor  <mjambor@suse.cz>
41450         * hsa.c (hsa_callable_function_p): Revert addition of DECL_ARTIFICIAL
41451         test.
41452         * ipa-hsa.c (process_hsa_functions): Only duplicate non-artificial
41453         decorated functions.
41455 2017-01-11  Richard Biener  <rguenther@suse.de>
41457         * tree-vrp.c (evrp_dom_walker::before_dom_children): Also
41458         set range/nonnull info for PHI results.  Do not set it on
41459         stmts marked for removal.
41461 2017-01-10  Eric Botcazou  <ebotcazou@adacore.com>
41463         * expr.c (store_field): In the bitfield case, fetch the return value
41464         from the registers before applying a single big-endian adjustment.
41465         Always do a final load for a BLKmode value not larger than a word.
41467 2017-01-10  David Malcolm  <dmalcolm@redhat.com>
41469         PR c++/77949
41470         * input.c (selftest::test_accessing_ordinary_linemaps): Verify
41471         that we correctly handle column numbers greater than
41472         LINE_MAP_MAX_COLUMN_NUMBER.
41474 2017-01-10  Martin Sebor  <msebor@redhat.com>
41476         PR middle-end/78245
41477         * gimple-ssa-sprintf.c (get_destination_size): Call
41478         {init,fini}object_sizes.
41479         * tree-object-size.c (addr_object_size): Adjust.
41480         (pass_through_call): Adjust.
41481         (pass_object_sizes::execute): Adjust.
41482         * tree-object-size.h (fini_object_sizes): Declare.
41484 2017-01-10  Martin Sebor  <msebor@redhat.com>
41486         PR tree-optimization/78775
41487         * builtins.c (get_size_range): Move...
41488         * calls.c: ...to here.
41489         (alloc_max_size): Accept zero argument.
41490         (operand_signed_p): Remove.
41491         (maybe_warn_alloc_args_overflow): Call get_size_range.
41492         * calls.h (get_size_range): Declare.
41494 2017-01-10  Joe Seymour  <joe.s@somniumtech.com>
41496         * config/msp430/driver-msp430.c (msp430_mcu_data): Sync with data
41497         from TI's devices.csv file as of September 2016.
41498         * config/msp430/msp430.c (msp430_mcu_data): Likewise.
41500 2017-01-10  Sandra Loosemore  <sandra@codesourcery.com>
41502         * doc/extend.texi: Tweak formatting to fix overfull hbox warnings.
41503         * doc/invoke.texi: Likewise.
41504         * doc/md.texi: Likewise.
41505         * doc/objc.texi: Likewise.
41507 2017-01-10  Joshua Conner  <joshconner@google.com>
41509         * config/arm/fuchsia-elf.h: New file.
41510         * config/fuchsia.h: New file.
41511         * config.gcc (*-*-fuchsia*): Set native_system_header_dir.
41512         (aarch64*-*-fuchsia*, arm*-*-fuchsia*, x86_64-*-fuchsia*): Add to
41513         targets.
41514         * config.host: (aarch64*-*-fuchsia*, arm*-*-fuchsia*): Add to hosts.
41516 2016-01-10  Richard Biener  <rguenther@suse.de>
41518         PR tree-optimization/79034
41519         * tree-call-cdce.c (shrink_wrap_one_built_in_call_with_conds):
41520         Propagate out degenerate PHIs in the joiner.
41522 2017-01-10  Martin Liska  <mliska@suse.cz>
41524         * ipa-icf.c (sort_sem_items_by_decl_uid): New function.
41525         (sort_congruence_classes_by_decl_uid): Likewise.
41526         (sort_congruence_class_groups_by_decl_uid): Likewise.
41527         (sem_item_optimizer::merge_classes): Sort class, groups in these
41528         classes and members in the groups by DECL_UID of declarations.
41529         This would make merge operations stable.
41531 2017-01-10  Martin Liska  <mliska@suse.cz>
41533         * ipa-icf.c (sem_item_optimizer::sem_item_optimizer): Remove
41534         usage of m_classes_vec.
41535         (sem_item_optimizer::~sem_item_optimizer):  Likewise.
41536         (sem_item_optimizer::get_group_by_hash): Likewise.
41537         (sem_item_optimizer::subdivide_classes_by_equality): Likewise.
41538         (sem_item_optimizer::subdivide_classes_by_sensitive_refs): Likewise.
41539         (sem_item_optimizer::verify_classes): Likewise.
41540         (sem_item_optimizer::process_cong_reduction): Likewise.
41541         (sem_item_optimizer::dump_cong_classes): Likewise.
41542         (sem_item_optimizer::merge_classes): Likewise.
41543         * ipa-icf.h (congruence_class_hash): Rename from
41544         congruence_class_group_hash.  Remove declaration of m_classes_vec.
41546 2017-01-10  Andrew Senkevich  <andrew.senkevich@intel.com>
41548         * common/config/i386/i386-common.c (OPTION_MASK_ISA_AVX512VPOPCNTDQ_SET,
41549         OPTION_MASK_ISA_AVX512VPOPCNTDQ_UNSET): New.
41550         * config.gcc: Add avx512vpopcntdqintrin.h.
41551         * config/i386/avx512vpopcntdqintrin.h: New.
41552         * config/i386/cpuid.h (bit_AVX512VPOPCNTDQ): New.
41553         * config/i386/i386-builtin-types.def: Add new types.
41554         * config/i386/i386-builtin.def (__builtin_ia32_vpopcountd_v16si,
41555         __builtin_ia32_vpopcountd_v16si_mask, __builtin_ia32_vpopcountq_v8di,
41556         __builtin_ia32_vpopcountq_v8di_mask): New.
41557         * config/i386/i386-c.c (ix86_target_macros_internal): Define
41558         __AVX512VPOPCNTDQ__.
41559         * config/i386/i386.c (ix86_target_string): Add -mavx512vpopcntdq.
41560         (PTA_AVX512VPOPCNTDQ): Define.
41561         * config/i386/i386.h (TARGET_AVX512VPOPCNTDQ,
41562         TARGET_AVX512VPOPCNTDQ_P): Define.
41563         * config/i386/i386.opt: Add mavx512vpopcntdq.
41564         * config/i386/immintrin.h: Include avx512vpopcntdqintrin.h.
41565         * config/i386/sse.md (define_insn "vpopcount<mode><mask_name>"): New.
41567 2017-01-01  Jan Hubicka  <hubicka@ucw.cz>
41569         PR middle-end/77484
41570         * predict.def (PRED_CALL): Set to 67.
41572 2017-01-09  Eric Botcazou  <ebotcazou@adacore.com>
41574         * expr.c (store_field): In the bitfield case, if the value comes from
41575         a function call and is of an aggregate type returned in registers, do
41576         not modify the field mode; extract the value in all cases if the mode
41577         is BLKmode and the size is not larger than a word.
41579 2017-01-09  Dominique d'Humieres  <dominiq@lps.ens.fr>
41581         PR target/71017
41582         * config/i386/cpuid.h: Fix undefined behavior.
41584 2017-01-04  Jeff Law  <law@redhat.com>
41586         PR tree-optimization/79007
41587         PR tree-optimization/67955
41588         * tree-ssa-alias.c (same_addr_size_stores_p): Only need to be
41589         conservative for pt.null when flag_non_call_exceptions is on.
41591 2017-01-09  Jakub Jelinek  <jakub@redhat.com>
41593         PR translation/79019
41594         PR translation/79020
41595         * params.def (PARAM_INLINE_MIN_SPEEDUP,
41596         PARAM_IPA_CP_SINGLE_CALL_PENALTY,
41597         PARAM_USE_AFTER_SCOPE_DIRECT_EMISSION_THRESHOLD): Fix typos
41598         in descriptions.
41599         * config/avr/avr.opt (maccumulate-args): Likewise.
41600         * config/msp430/msp430.opt (mwarn-mcu): Likewise.
41601         * common.opt (freport-bug): Likewise.
41602         * cif-code.def (CIF_FINAL_ERROR): Likewise.
41603         * doc/invoke.texi (ipa-cp-single-call-penalty): Likewise.
41604         * config/s390/s390.c (s390_invalid_binary_op): Fix spelling in
41605         translatable string.
41606         * config/i386/i386.c (function_value_32): Likewise.
41607         * config/nios2/nios2.c (nios2_valid_target_attribute_rec): Likewise.
41608         * config/msp430/msp430.c (msp430_option_override, msp430_attr):
41609         Likewise.
41610         * config/msp430/driver-msp430.c (msp430_select_hwmult_lib): Likewise.
41611         * common/config/msp430/msp430-common.c (msp430_handle_option):
41612         Likewise.
41613         * symtab.c (symtab_node::verify_base): Likewise.
41614         * opts.c (set_debug_level): Likewise.
41615         * tree.c (verify_type_variant): Likewise.  Fix typo in comment.
41616         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Add
41617         missing whitespace to translatable strings.
41618         * config/avr/avr.md (bswapsi2): Fix typo in comment.
41619         * config/sh/superh.h: Likewise.
41620         * config/i386/xopintrin.h: Likewise.
41621         * config/i386/znver1.md: Likewise.
41622         * config/rs6000/rs6000.c (struct rs6000_opt_mask): Likewise.
41623         * ipa-inline-analysis.c (compute_inline_parameters): Likewise.
41624         * double-int.h (struct double_int): Likewise.
41625         * double-int.c (div_and_round_double): Likewise.
41626         * wide-int.cc: Likewise.
41627         * tree-ssa.c (non_rewritable_mem_ref_base): Likewise.
41628         * tree-ssa-sccvn.c (vn_reference_lookup_3): Likewise.
41629         * cfgcleanup.c (crossjumps_occured): Renamed to ...
41630         (crossjumps_occurred): ... this.
41631         (try_crossjump_bb, try_head_merge_bb, try_optimize_cfg, cleanup_cfg):
41632         Adjust all uses.
41634         PR tree-optimization/78899
41635         * tree-if-conv.c (version_loop_for_if_conversion): Instead of
41636         returning bool return struct loop *, NULL for failure and the new
41637         loop on success.
41638         (versionable_outer_loop_p): Don't version outer loop if it has
41639         dont_vectorized bit set.
41640         (tree_if_conversion): When versioning outer loop, ensure
41641         tree_if_conversion is performed also on the inner loop of the
41642         non-vectorizable outer loop copy.
41643         * tree-vectorizer.c (set_uid_loop_bbs): Formatting fix.  Fold
41644         LOOP_VECTORIZED in inner loop of the scalar outer loop and
41645         prevent vectorization of it.
41646         (vectorize_loops): For outer + inner LOOP_VECTORIZED, ensure
41647         the outer loop vectorization of the non-scalar version is attempted
41648         before vectorization of the inner loop in scalar version.  If
41649         outer LOOP_VECTORIZED guarded loop is not vectorized, prevent
41650         vectorization of its inner loop.
41651         * tree-vect-loop-manip.c (rename_variables_in_bb): If outer_loop
41652         has 2 inner loops, rename also on edges from bb whose single pred
41653         is outer_loop->header.  Fix typo in function comment.
41655 2017-01-09  Martin Sebor  <msebor@redhat.com>
41657         PR bootstrap/79033
41658         * asan.c (asan_emit_stack_protection): Increase local buffer size
41659         to avoid snprintf truncation warning.
41661 2017-01-09  Andrew Pinski  <apinski@cavium.com>
41663         * config/aarch64/aarch64-cores.def: Add thunderx2t99.  Change vulcan
41664         to reference thunderx2t99 for the tuning structure
41665         * config/aarch64/aarch64-cost-tables.h (vulcan_extra_costs):
41666         Rename to ...
41667         (thunderx2t99_extra_costs): This.
41668         * config/aarch64/aarch64-tune.md: Regenerate.
41669         * config/aarch64/aarch64.c (vulcan_addrcost_table): Rename to ...
41670         (vulcan_addrcost_table): This.
41671         (vulcan_regmove_cost): Rename to ...
41672         (thunderx2t99_regmove_cost): This.
41673         (vulcan_vector_cost): Rename to ...
41674         (thunderx2t99_vector_cost): this.
41675         (vulcan_branch_cost): Rename to ...
41676         (thunderx2t99_branch_cost): This.
41677         (vulcan_tunings): Rename to ...
41678         (thunderx2t99_tunings): This and s/vulcan/thunderx2t99 .
41679         * doc/invoke.texi (AARCH64/mtune): Add thunderx2t99.
41681 2017-01-09  Martin Jambor  <mjambor@suse.cz>
41683         PR ipa/78365
41684         PR ipa/78599
41685         * ipa-prop.h (ipa_jump_func): Swap positions of vr_known and m_vr.
41686         * ipa-cp.c (ipa_vr_operation_and_type_effects): New function.
41687         (propagate_vr_accross_jump_function): Use the above function for all
41688         value range computations for pass-through jump functions and type
41689         converasion from explicit value range values.
41690         (ipcp_propagate_stage): Do not attempt to deduce types of formal
41691         parameters from TYPE_ARG_TYPES.
41692         * ipa-prop.c (ipa_write_jump_function): Remove trailing whitespace.
41693         (ipa_write_node_info): Stream type of the actual argument.
41694         (ipa_read_node_info): Likewise. Also remove trailing whitespace.
41696 2017-01-09  Martin Liska  <mliska@suse.cz>
41698         PR pch/78970
41699         * gcc.c (driver_handle_option): Handle OPT_E and set have_E.
41700         (lookup_compiler): Do not show error message with have_E.
41702 2017-01-09  Jakub Jelinek  <jakub@redhat.com>
41704         PR tree-optimization/78938
41705         * tree-vect-stmts.c (vectorizable_condition): For non-masked COND_EXPR
41706         where comp_vectype is VECTOR_BOOLEAN_TYPE_P, use
41707         BIT_{NOT,XOR,AND,IOR}_EXPR on the comparison operands instead of
41708         {EQ,NE,GE,GT,LE,LT}_EXPR directly inside of VEC_COND_EXPR.  Formatting
41709         fixes.
41711 2017-01-09  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
41713         * tree-ssa-address.c (gen_addr_rtx): Don't handle index if it
41714         is const0_rtx.
41716 2017-01-09  Richard Biener  <rguenther@suse.de>
41718         PR tree-optimization/78997
41719         * tree-vect-slp.c (vect_mask_constant_operand_p): Handle SSA
41720         name condition properly.
41722 2017-01-09  Richard Biener  <rguenther@suse.de>
41724         PR debug/79000
41725         * dwarf2out.c (is_cxx): New overload with context.
41726         (is_naming_typedef_decl): Use it.
41728 2017-01-08  Sandra Loosemore  <sandra@codesourcery.com>
41730         * invoke.texi (Option Summary): Correct spacing in option lists
41731         and add line breaks to fix over-long lines.
41733 2017-01-08  Sandra Loosemore  <sandra@codesourcery.com>
41735         PR middle-end/17660
41737         * extend.texi (Common Variable Attributes): Add xref to GCC
41738         Internals manual to explain mode attribute keywords.
41740 2017-01-08  Sandra Loosemore  <sandra@codesourcery.com>
41742         PR other/16519
41743         * doc/invoke.texi (Option Summary): Move -pthread to Linker Options
41744         and Preprocessor Options.
41745         (Options for Linking): Document -pthread here....
41746         (RS/6000 and PowerPC Options): ...not here.
41747         (Solaris 2 Options): ...or here.
41748         * doc/cppopts.texi: Document -pthread.
41750 2017-01-08  Martin Sebor  <msebor@redhat.com>
41752         PR middle-end/77708
41753         * doc/invoke.texi (Warning Options): Document -Wformat-truncation.
41754         * gimple-ssa-sprintf.c (call_info::reval_used, call_info::warnopt):
41755         New member functions.
41756         (format_directive): Used them.
41757         (add_bytes): Same.
41758         (pass_sprintf_length::handle_gimple_call): Same.
41759         * graphite-sese-to-poly.c (tree_int_to_gmp): Increase buffer size
41760         to avoid truncation for any argument.
41761         (extract_affine_mul): Same.
41762         * tree.c (get_file_function_name): Same.
41764 2017-01-01  Jan Hubicka  <hubicka@ucw.cz>
41766         PR middle-end/77484
41767         * predict.def (PRED_INDIR_CALL): Set to 86.
41769 2017-01-07  Sandra Loosemore  <sandra@codesourcery.com>
41771         PR preprocessor/54124
41772         * doc/cppopts.texi: Reformat -d subtable to list the full name
41773         of the options.  Add cross-reference to the docs for the general
41774         compiler -d options.
41775         * doc/invoke.texi (Developer Options): Add cross-reference to the
41776         preprocessor-specific -d option documentation.
41778 2017-01-07  Sandra Loosemore  <sandra@codesourcery.com>
41780         PR preprocessor/13498
41781         * doc/cpp.texi (Search Path): Rewrite to remove obsolete and
41782         redudant material, and reflect new command-line options.
41783         (System Headers): Likewise.
41785 2017-01-07  Sandra Loosemore  <sandra@codesourcery.com>
41787         * doc/cppdiropts.texi: Merge documentation of -I, -iquote,
41788         -isystem, and -idirafter.  Copy-edit.
41789         * doc/cppopts.texi: Copy-edit.  Remove contradiction about
41790         default for -ftrack-macro-expansion.  Delete obsolete and
41791         badly-formatted implementation details about -fdebug-cpp output.
41792         * doc/cppwarnopts.texi: Copy-edit.
41794 2017-01-07  David Malcolm  <dmalcolm@redhat.com>
41796         PR c++/72803
41797         * input.c (selftest::test_accessing_ordinary_linemaps): Verify
41798         that the transition from a max line width >= 1<<10 to narrower
41799         lines works correctly.
41801 2017-01-07  Alexandre Oliva  <aoliva@redhat.com>
41803         * doc/options.texi (PerFunction): New.
41804         * opt-functions.awk (switch_flags): Map both Optimization and
41805         PerFunction to CL_OPTIMIZATION.
41806         * opth-gen.awk: Test for PerFunction flag along with
41807         Optimization.
41808         * optc-save-gen.awk: Likewise.  Introduce var_opt_hash and set
41809         it only when the latter is present.  Skip those that don't in
41810         the hash function generator.
41811         * common.opt (fvar-tracking): Mark as PerFunction instead of
41812         Optimization.
41813         (fvar-tracking-assignments): Likewise.
41814         (fvar-tracking-assignments-toggle): Likewise.
41815         (fvar-tracking-uninit): Likewise.
41817 2017-01-07  Jakub Jelinek  <jakub@redhat.com>
41819         PR translation/79018
41820         * params.def (PARAM_MAX_STORES_TO_MERGE): Add missing space between
41821         the and store.
41823 2017-01-06  Mikael Pettersson  <mikpelinux@gmail.com>
41825         PR target/57583
41826         * config/m68k/m68k.opt (LONG_JUMP_TABLE_OFFSETS): New option.
41827         * config/m68k/linux.h (ASM_RETURN_CASE_JUMP): Handle
41828         TARGET_LONG_JUMP_TABLE_OFFSETS.
41829         * config/m68k/m68kelf.h (ASM_RETURN_CASE_JUMP): Likewise.
41830         * config/m68k/netbsd-elf.h (ASM_RETURN_CASE_JUMP): Likewise.
41831         * config/m68k/m68k.h (CASE_VECTOR_MODE): Likewise.
41832         (ASM_OUTPUT_ADDR_DIFF_ELF): Likewise.
41833         * config/m68k/m68k.md (tablejump expander): Likewise.
41834         (*tablejump_pcrel_hi): Renamed from unnamed insn, reject
41835         TARGET_LONG_JUMP_TABLE_OFFSETS.
41836         (*tablejump_pcrel_si): New insn, handle TARGET_LONG_JUMP_TABLE_OFFSETS.
41837         * doc/invoke.texi (M68K options): Add -mlong-jump-table-offsets.
41839 2017-01-06  Edgar E. Iglesias <edgar.iglesias@xilinx.com>
41840             David Holsgrove <david.holsgrove@xilinx.com>
41842         * common/config/microblaze/microblaze-common.c
41843         (TARGET_EXCEPT_UNWIND_INFO): Remove.
41844         * config/microblaze/microblaze-protos.h (microblaze_eh_return):
41845         New prototype.
41846         * config/microblaze/microblaze.c (microblaze_must_save_register)
41847         (microblaze_expand_epilogue, microblaze_return_addr): Handle
41848         calls_eh_return.
41849         (microblaze_eh_return): New function.
41850         * config/microblaze/microblaze.h (RETURN_ADDR_OFFSET)
41851         (EH_RETURN_DATA_REGNO, MB_EH_STACKADJ_REGNUM)
41852         (EH_RETURN_STACKADJ_RTX, ASM_PREFERRED_EH_DATA_FORMAT): New macros.
41853         * config/microblaze/microblaze.md (eh_return): New pattern.
41855 2017-01-06  Jakub Jelinek  <jakub@redhat.com>
41857         * system.h (GCC_DIAGNOSTIC_PUSH_IGNORED, GCC_DIAGNOSTIC_POP,
41858         GCC_DIAGNOSTIC_STRINGIFY): Define.
41860         * read-rtl.c (rtx_reader::read_rtx_code): Avoid -Wsign-compare warning.
41862 2017-01-06  Andre Vieira  <andre.simoesdiasvieira@arm.com>
41864         * config/arm/arm.md (<mcrr>): New.
41865         (<mrrc>): New.
41866         * config/arm/arm.c (arm_arch5te): New.
41867         (arm_option_override): Set arm_arch5te.
41868         (arm_coproc_builtin_available): Add support for mcrr, mcrr2, mrrc
41869         and mrrc2.
41870         * config/arm/arm-builtins.c (MCRR_QUALIFIERS): Define to...
41871         (arm_mcrr_qualifiers): ... this. New.
41872         (MRRC_QUALIFIERS): Define to...
41873         (arm_mrrc_qualifiers): ... this. New.
41874         * config/arm/arm_acle.h (__arm_mcrr, __arm_mcrr2, __arm_mrrc,
41875         __arm_mrrc2): New.
41876         * config/arm/arm_acle_builtins.def (mcrr, mcrr2, mrrc, mrrc2): New.
41877         * config/arm/iterators.md (MCRRI, mcrr, MCRR): New.
41878         (MRRCI, mrrc, MRRC): New.
41879         * config/arm/unspecs.md (VUNSPEC_MCRR, VUNSPEC_MCRR2, VUNSPEC_MRRC,
41880         VUNSPEC_MRRC2): New.
41882 2017-01-06  Andre Vieira  <andre.simoesdiasvieira@arm.com>
41884         * config/arm/arm.md (<mcr>): New.
41885         (<mrc>): New.
41886         * config/arm/arm.c (arm_coproc_builtin_available): Add
41887         support for mcr, mrc, mcr2 and mrc2.
41888         * config/arm/arm-builtins.c (MCR_QUALIFIERS): Define to...
41889         (arm_mcr_qualifiers): ... this. New.
41890         (MRC_QUALIFIERS): Define to ...
41891         (arm_mrc_qualifiers): ... this. New.
41892         (MCR_QUALIFIERS): Define to ...
41893         (arm_mcr_qualifiers): ... this. New.
41894         * config/arm/arm_acle.h (__arm_mcr, __arm_mrc, __arm_mcr2,
41895         __arm_mrc2): New.
41896         * config/arm/arm_acle_builtins.def (mcr, mcr2, mrc, mrc2): New.
41897         * config/arm/iterators.md (MCRI, mcr, MCR, MRCI, mrc, MRC): New.
41898         * config/arm/unspecs.md (VUNSPEC_MCR, VUNSPEC_MCR2, VUNSPEC_MRC,
41899         VUNSPEC_MRC2): New.
41901 2017-01-06  Andre Vieira  <andre.simoesdiasvieira@arm.com>
41903         * config/arm/arm.md (*ldc): New.
41904         (*stc): New.
41905         (<ldc>): New.
41906         (<stc>): New.
41907         * config/arm/arm.c (arm_coproc_builtin_available): Add
41908         support for ldc,ldcl,stc,stcl,ldc2,ldc2l,stc2 and stc2l.
41909         (arm_coproc_ldc_stc_legitimate_address): New.
41910         * config/arm/arm-builtins.c (arm_type_qualifiers): Add
41911         'qualifier_const_pointer'.
41912         (LDC_QUALIFIERS): Define to...
41913         (arm_ldc_qualifiers): ... this. New.
41914         (STC_QUALIFIERS): Define to...
41915         (arm_stc_qualifiers): ... this. New.
41916         * config/arm/arm-protos.h
41917         (arm_coproc_ldc_stc_legitimate_address): New.
41918         * config/arm/arm_acle.h (__arm_ldc, __arm_ldcl, __arm_stc,
41919         __arm_stcl, __arm_ldc2, __arm_ldc2l, __arm_stc2, __arm_stc2l): New.
41920         * config/arm/arm_acle_builtins.def (ldc, ldc2, ldcl, ldc2l, stc,
41921         stc2, stcl, stc2l): New.
41922         * config/arm/constraints.md (Uz): New.
41923         * config/arm/iterators.md (LDCI, STCI, ldc, stc, LDC STC): New.
41924         * config/arm/unspecs.md (VUNSPEC_LDC, VUNSPEC_LDC2, VUNSPEC_LDCL,
41925         VUNSPEC_LDC2L, VUNSPEC_STC, VUNSPEC_STC2, VUNSPEC_STCL,
41926         VUNSPEC_STC2L): New.
41928 2017-01-06  Andre Vieira  <andre.simoesdiasvieira@arm.com>
41930         * config/arm/arm.md (<cdp>): New.
41931         * config/arm/arm.c (neon_const_bounds): Rename this ...
41932         (arm_const_bounds): ... this.
41933         (arm_coproc_builtin_available): New.
41934         * config/arm/arm-builtins.c (SIMD_MAX_BUILTIN_ARGS): Increase.
41935         (arm_type_qualifiers): Add 'qualifier_unsigned_immediate'.
41936         (CDP_QUALIFIERS): Define to...
41937         (arm_cdp_qualifiers): ... this. New.
41938         (void_UP): Define.
41939         (arm_expand_builtin_args): Add case for 6 arguments.
41940         * config/arm/arm-protos.h (neon_const_bounds): Rename this ...
41941         (arm_const_bounds): ... this.
41942         (arm_coproc_builtin_available): New.
41943         * config/arm/arm_acle.h (__arm_cdp): New.
41944         (__arm_cdp2): New.
41945         * config/arm/arm_acle_builtins.def (cdp): New.
41946         (cdp2): New.
41947         * config/arm/iterators.md (CDPI,CDP,cdp): New.
41948         * config/arm/neon.md: Rename all 'neon_const_bounds' to
41949         'arm_const_bounds'.
41950         * config/arm/types.md (coproc): New.
41951         * config/arm/unspecs.md (VUNSPEC_CDP, VUNSPEC_CDP2): New.
41952         * doc/extend.texi (ACLE): Add a mention of Coprocessor intrinsics.
41953         * doc/sourcebuild.texi (arm_coproc1_ok, arm_coproc2_ok,
41954         arm_coproc3_ok, arm_coproc4_ok): Document new effective targets.
41956 2017-01-06  Andre Vieira  <andre.simoesdiasvieira@arm.com>
41958         * config/arm/arm-builtins.c (arm_unsigned_binop_qualifiers): New.
41959         (UBINOP_QUALIFIERS): New.
41960         (si_UP): Define.
41961         (acle_builtin_data): New. Change comment.
41962         (arm_builtins): Remove ARM_BUILTIN_CRC32B, ARM_BUILTIN_CRC32H,
41963         ARM_BUILTIN_CRC32W, ARM_BUILTIN_CRC32CB, ARM_BUILTIN_CRC32CH,
41964         ARM_BUILTIN_CRC32CW. Add ARM_BUILTIN_ACLE_BASE and include
41965         arm_acle_builtins.def.
41966         (ARM_BUILTIN_ACLE_PATTERN_START): Define.
41967         (arm_init_acle_builtins): New.
41968         (CRC32_BUILTIN): Remove.
41969         (bdesc_2arg): Remove entries for crc32b, crc32h, crc32w,
41970         crc32cb, crc32ch and crc32cw.
41971         (arm_init_crc32_builtins): Remove.
41972         (arm_init_builtins): Use arm_init_acle_builtins rather
41973         than arm_init_crc32_builtins.
41974         (arm_expand_acle_builtin): New.
41975         (arm_expand_builtin): Use 'arm_expand_acle_builtin'.
41976         * config/arm/arm_acle_builtins.def: New.
41978 2017-01-06  Andre Vieira  <andre.simoesdiasvieira@arm.com>
41980         * config/arm/arm-builtins.c (neon_builtin_datum): Rename to ..
41981         (arm_builtin_datum): ... this.
41982         (arm_init_neon_builtin): Rename to ...
41983         (arm_init_builtin): ... this. Add a new parameters PREFIX
41984         and USE_SIG_IN_NAME.
41985         (arm_init_neon_builtins): Replace 'arm_init_neon_builtin' with
41986         'arm_init_builtin'. Replace type 'neon_builtin_datum' with
41987         'arm_builtin_datum'.
41988         (arm_init_vfp_builtins): Likewise.
41989         (builtin_arg): Rename enum's replacing 'NEON_ARG' with
41990         'ARG_BUILTIN' and add a 'ARG_BUILTIN_NEON_MEMORY.
41991         (arm_expand_neon_args): Rename to ...
41992         (arm_expand_builtin_args): ... this. Rename builtin_arg
41993         enum values and differentiate between ARG_BUILTIN_MEMORY
41994         and ARG_BUILTIN_NEON_MEMORY.
41995         (arm_expand_neon_builtin_1): Rename to ...
41996         (arm_expand_builtin_1): ... this. Rename builtin_arg enum
41997         values, arm_expand_builtin_args and add bool parameter NEON.
41998         (arm_expand_neon_builtin): Use arm_expand_builtin_1.
41999         (arm_expand_vfp_builtin): Likewise.
42000         (NEON_MAX_BUILTIN_ARGS): Remove, it was unused.
42002 2017-01-01  Jan Hubicka  <hubicka@ucw.cz>
42004         PR middle-end/77484
42005         * predict.def (PRED_POLYMORPHIC_CALL): Set to 59.
42006         * predict.c (tree_estimate_probability_bb): Reverse direction of
42007         polymorphic call predictor.
42009 2017-01-06  David Malcolm  <dmalcolm@redhat.com>
42011         * passes.c (execute_one_pass): Split out pass-skipping logic into...
42012         (determine_pass_name_match): ...this new function and...
42013         (should_skip_pass_p): ...this new function.
42015 2017-01-06  Nathan Sidwell  <nathan@acm.org>
42017         * ipa-visibility.c (function_and_variable_visibility): Reformat
42018         comments and long lines.  Remove extrneous if.
42019         * symtab.c (symtab_node::make_decl_local): Fix code format.
42020         (symtab_node::set_section_for_node): Fix comment typo.
42022 2017-01-06  Martin Liska  <mliska@suse.cz>
42024         PR bootstrap/79003
42025         * lra-constraints.c: Rename invariant to lra_invariant.
42026         * predict.c (set_even_probabilities): Initialize e to NULL.
42028 2017-01-05  Martin Sebor  <msebor@redhat.com>
42030         PR tree-optimization/78910
42031         * gimple-ssa-sprintf.c (tree_digits): Add an argument.
42032         (format_integer): Correct off-by-one error in the handling
42033         of precision with negative numbers in signed conversions..
42035 2017-01-05  Eric Botcazou  <ebotcazou@adacore.com>
42037         * doc/invoke.texi (C Dialect Options): Adjust -fsso-struct entry.
42039 2017-01-05  Jakub Jelinek  <jakub@redhat.com>
42041         PR tree-optimization/71016
42042         * tree-ssa-phiopt.c (tree_ssa_phiopt_worker): Pass cond_stmt to
42043         factor_out_conditional_conversion.  Formatting fix.
42044         (factor_out_conditional_conversion): Add cond_stmt argument.
42045         If arg1 is INTEGER_CST, punt if new_arg0 is not any operand of
42046         cond_stmt and if arg0_def_stmt is not the only stmt in its bb.
42047         Formatting fix.
42049 2017-01-05  David Malcolm  <dmalcolm@redhat.com>
42051         * Makefile.in (OBJS): Add read-md.o, read-rtl.o,
42052         read-rtl-function.o, and selftest-rtl.o.
42053         * config/aarch64/aarch64.c: Include selftest.h and selftest-rtl.h.
42054         (selftest::aarch64_test_loading_full_dump): New function.
42055         (selftest::aarch64_run_selftests): New function.
42056         (TARGET_RUN_TARGET_SELFTESTS): Wire it up to
42057         selftest::aarch64_run_selftests.
42058         * config/i386/i386.c
42059         (selftest::ix86_test_loading_dump_fragment_1): New function.
42060         (selftest::ix86_test_loading_call_insn): New function.
42061         (selftest::ix86_test_loading_full_dump): New function.
42062         (selftest::ix86_test_loading_unspec): New function.
42063         (selftest::ix86_run_selftests): Call the new functions.
42064         * emit-rtl.c (maybe_set_max_label_num): New function.
42065         * emit-rtl.h (maybe_set_max_label_num): New decl.
42066         * function.c (instantiate_decls): Guard call to
42067         instantiate_decls_1 with if (DECL_INITIAL (fndecl)).
42068         * function-tests.c (selftest::verify_three_block_rtl_cfg): Remove
42069         "static".
42070         * gensupport.c (gen_reader::gen_reader): Pass "false"
42071         for new "compact" param of rtx_reader.
42072         * print-rtl.c (rtx_writer::print_rtx_operand): Print "(nil)"
42073         rather than an empty string for NULL strings.
42074         * read-md.c: Potentially include config.h rather than bconfig.h.
42075         Wrap include of errors.h with #ifdef GENERATOR_FILE.
42076         (have_error): New global, copied from errors.c.
42077         (md_reader::read_name): Rename to...
42078         (md_reader::read_name_1): ...this, adding "out_loc" param,
42079         and converting "missing name or number" to returning false, rather
42080         than failing.
42081         (md_reader::read_name): Reimplement in terms of read_name_1.
42082         (md_reader::read_name_or_nil): New function.
42083         (md_reader::read_string): Handle "(nil)" by returning NULL.
42084         (md_reader::md_reader): Add new param "compact".
42085         (md_reader::read_md_files): Wrap with #ifdef GENERATOR_FILE.
42086         (md_reader::read_file): New method.
42087         * read-md.h (md_reader::md_reader): Add new param "compact".
42088         (md_reader::read_file): New method.
42089         (md_reader::is_compact): New accessor.
42090         (md_reader::read_name): Convert return type from void to file_location.
42091         (md_reader::read_name_or_nil): New decl.
42092         (md_reader::read_name_1): New decl.
42093         (md_reader::m_compact): New field.
42094         (noop_reader::noop_reader): Pass "false" for new "compact" param
42095         of rtx_reader.
42096         (rtx_reader::rtx_reader): Add new "compact" param.
42097         (rtx_reader::read_rtx_operand): Make virtual and convert return
42098         type from void to rtx.
42099         (rtx_reader::read_until): New decl.
42100         (rtx_reader::handle_any_trailing_information): New virtual function.
42101         (rtx_reader::postprocess): New virtual function.
42102         (rtx_reader::finalize_string): New virtual function.
42103         (rtx_reader::m_in_call_function_usage): New field.
42104         (rtx_reader::m_reuse_rtx_by_id): New field.
42105         * read-rtl-function.c: New file.
42106         * selftest-rtl.c (selftest::assert_rtx_ptr_eq_at): New function.
42107         * selftest-rtl.h (ASSERT_RTX_PTR_EQ): New macro.
42108         (selftest::verify_three_block_rtl_cfg): New decl.
42109         * read-rtl-function.h: New file.
42110         * read-rtl.c: Potentially include config.h rather than bconfig.h.
42111         For host, include function.h, memmodel.h, and emit-rtl.h.
42112         (one_time_initialization): New function.
42113         (struct compact_insn_name): New struct.
42114         (compact_insn_names): New array.
42115         (find_code): Handle insn codes in compact dumps.
42116         (apply_subst_iterator): Wrap with #ifdef GENERATOR_FILE.
42117         (bind_subst_iter_and_attr): Likewise.
42118         (add_condition_to_string): Likewise.
42119         (add_condition_to_rtx): Likewise.
42120         (apply_attribute_uses): Likewise.
42121         (add_current_iterators): Likewise.
42122         (apply_iterators): Likewise.
42123         (initialize_iterators): Guard usage of apply_subst_iterator with
42124         #ifdef GENERATOR_FILE.
42125         (read_conditions): Wrap with #ifdef GENERATOR_FILE.
42126         (md_reader::read_mapping): Likewise.
42127         (add_define_attr_for_define_subst): Likewise.
42128         (add_define_subst_attr): Likewise.
42129         (read_subst_mapping): Likewise.
42130         (check_code_iterator): Likewise.
42131         (rtx_reader::read_rtx): Likewise.  Move one-time initialization
42132         logic to...
42133         (one_time_initialization): New function.
42134         (rtx_reader::read_until): New method.
42135         (read_flags): New function.
42136         (parse_reg_note_name): New function.
42137         (rtx_reader::read_rtx_code): Initialize "iterator" to NULL.
42138         Handle reuse_rtx ids.
42139         Wrap iterator lookup within #ifdef GENERATOR_FILE.
42140         Add parsing support for RTL dumps, mirroring the special-cases in
42141         print_rtx, by calling read_flags, reading REG_NOTE names, INSN_UID
42142         values, and calling handle_any_trailing_information.
42143         (rtx_reader::read_rtx_operand): Convert return type from void
42144         to rtx, returning return_rtx.  Handle case 'e'.  Call
42145         finalize_string on XSTR and XTMPL fields.
42146         (rtx_reader::read_nested_rtx):  Handle dumps in which trailing
42147          "(nil)" values were omitted.  Call the postprocess vfunc on the
42148         return_rtx.
42149         (rtx_reader::rtx_reader): Add new "compact" param and pass to base
42150         class ctor.  Initialize m_in_call_function_usage.  Call
42151         one_time_initialization.
42152         * rtl-tests.c (selftest::test_uncond_jump): Call
42153         set_new_first_and_last_insn.
42154         * rtl.h (read_rtx): Wrap decl with #ifdef GENERATOR_FILE.
42155         * selftest-rtl.c: New file.
42156         * selftest-rtl.h (class selftest::rtl_dump_test): New class.
42157         (selftest::get_insn_by_uid): New decl.
42158         * selftest-run-tests.c (selftest::run_tests): Call
42159         read_rtl_function_c_tests.
42160         * selftest.h  (selftest::read_rtl_function_c_tests): New decl.
42161         * tree-dfa.c (ssa_default_def): Return NULL_TREE for rtl function
42162         dumps.
42164 2017-01-05  Uros Bizjak  <ubizjak@gmail.com>
42166         * config/i386/i386.md (*testqi_ext_3): No need to handle memory
42167         operands in a special way.  Assert that pos+len <= mode precision.
42169 2017-01-05  Jakub Jelinek  <jakub@redhat.com>
42171         * common.opt (fvect-cost-model): Remove RejectNegative flag, use
42172         3 argument Alias with unlimited for the negative form.
42173         (fno-vect-cost-model): Removed.
42175 2017-01-05  Martin Liska  <mliska@suse.cz>
42177         * hsa-gen.c (gen_hsa_divmod): New function.
42178         (gen_hsa_insn_for_internal_fn_call): Use the function for IFN_DIVMOD.
42180 2017-01-05  Martin Liska  <mliska@suse.cz>
42182         PR pch/78970
42183         * gcc.c (lookup_compiler): Reject '-' filename for a precompiled
42184         header.
42186 2017-01-05  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
42188         * config/s390/s390.c (s390_expand_setmem): Unroll the loop for
42189         small constant length operands.
42191 2017-01-05  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
42193         * config/s390/s390.c (s390_expand_setmem): Avoid overlapping bytes
42194         between loop iterations.
42196 2017-01-05  Martin Liska  <mliska@suse.cz>
42198         PR sanitizer/78815
42199         * gimplify.c (gimplify_decl_expr): Compare to
42200         asan_poisoned_variables instread of checking flags.
42201         (gimplify_target_expr): Likewise.
42202         (gimplify_expr): Likewise.
42203         (gimplify_function_tree): Conditionally initialize
42204         asan_poisoned_variables.
42206 2017-01-04  Jeff Law  <law@redhat.com>
42208         PR tree-optimizatin/78812
42209         * rtl.h (contains_mem_rtx_p): Prototype.
42210         * ifcvt.c (containts_mem_rtx_p): Move from here to...
42211         * rtlanal.c (contains_mem_rtx_p): Here and remove static linkage.
42212         * gcse.c (prune_expressions): Use contains_mem_rtx_p to discover
42213         and prune MEMs that are not at the toplevel of a SET_SRC rtx.  Look
42214         through ZERO_EXTEND and SIGN_EXTEND when trying to avoid pruning MEMs.
42216 2017-01-04  Alexandre Oliva  <aoliva@redhat.com>
42218         * input.c (assert_char_at_range): Default-initialize actual_range.
42220 2017-01-04  Alexandre Oliva  <aoliva@redhat.com>
42222         * df-scan.c (df_ref_create_structure): Make regno unsigned,
42223         to match the caller.
42225 2017-01-04  Alexandre Oliva  <aoliva@redhat.com>
42227         * cfgexpand.c (expand_gimple_basic_block): Disregard debug
42228         insns after final jump in test to emit dummy move.
42230 2017-01-04  Alexandre Oliva  <aoliva@redhat.com>
42232         * gimple-iterator.h (gsi_one_nondebug_before_end_p): New.
42233         * tree-eh.c (cleanup_empty_eh): Skip more debug stmts.
42235 2017-01-04  Alexandre Oliva  <aoliva@redhat.com>
42237         * multiple_target.c (create_dispatcher_calls): Init e_next.
42238         * tree-ssa-loop-split.c (split_loop): Init border.
42239         * tree-vect-loop.c (vect_determine_vectorization_factor): Init
42240         scalar_type.
42242 2017-01-04  Michael Meissner  <meissner@linux.vnet.ibm.com>
42244         PR target/71977
42245         PR target/70568
42246         PR target/78823
42247         * config/rs6000/predicates.md (sf_subreg_operand): New predicate.
42248         (altivec_register_operand): Do not return true if the operand
42249         contains a SUBREG mixing SImode and SFmode.
42250         (vsx_register_operand): Likewise.
42251         (vsx_reg_sfsubreg_ok): New predicate.
42252         (vfloat_operand): Do not return true if the operand contains a
42253         SUBREG mixing SImode and SFmode.
42254         (vint_operand): Likewise.
42255         (vlogical_operand): Likewise.
42256         (gpc_reg_operand): Likewise.
42257         (int_reg_operand): Likewise.
42258         * config/rs6000/rs6000-protos.h (valid_sf_si_move): Add declaration.
42259         * config/rs6000/rs6000.c (valid_sf_si_move): New function to
42260         determine if a MOVSI or MOVSF operation contains SUBREGs that mix
42261         SImode and SFmode.
42262         (rs6000_emit_move_si_sf_subreg): New helper function.
42263         (rs6000_emit_move): Call rs6000_emit_move_si_sf_subreg to possbily
42264         fixup SUBREGs involving SImode and SFmode.
42265         * config/rs6000/vsx.md (SFBOOL_*): New constants that are operand
42266         numbers for the new peephole2 optimization.
42267         (peephole2 for SFmode unions): New peephole2 to optimize cases in
42268         the GLIBC math library that do AND/IOR/XOR operations on single
42269         precision floating point.
42270         * config/rs6000/rs6000.h (TARGET_NO_SF_SUBREG): New internal
42271         target macros to say whether we need to avoid SUBREGs mixing
42272         SImode and SFmode.
42273         (TARGET_ALLOW_SF_SUBREG): Likewise.
42274         * config/rs6000/rs6000.md (UNSPEC_SF_FROM_SI): New unspecs.
42275         (UNSPEC_SI_FROM_SF): Likewise.
42276         (iorxor): Change spacing.
42277         (and_ior_xor): New iterator for AND, IOR, and XOR.
42278         (movsi_from_sf): New insns for SImode/SFmode SUBREG support.
42279         (movdi_from_sf_zero_ext): Likewise.
42280         (mov<mode>_hardfloat, FMOVE32 iterator): Use register_operand
42281         instead of gpc_reg_operand.  Add SImode/SFmode SUBREG support.
42282         (movsf_from_si): New insn for SImode/SFmode SUBREG support.
42283         (fma<mode>4): Use gpc_reg_operand instead of register_operand.
42284         (fms<mode>4): Likewise.
42285         (fnma<mode>4): Likewise.
42286         (fnms<mode>4): Likewise.
42287         (nfma<mode>4): Likewise.
42288         (nfms<mode>4): Likewise.
42290 2017-01-04  Marek Polacek  <polacek@redhat.com>
42292         PR c++/64767
42293         * doc/invoke.texi: Document -Wpointer-compare.
42295 2017-01-04  Jakub Jelinek  <jakub@redhat.com>
42297         * optc-gen.awk: Emit #error for -W*/-f*/-m* Enum without
42298         RejectNegative.
42300         * dwarf2out.c (output_loc_list): Don't throw away 64K+ location
42301         descriptions for -gdwarf-5 and emit them as uleb128 instead of
42302         2-byte data.
42304 2017-01-04  Kelvin Nilsen  <kelvin@gcc.gnu.org>
42306         PR target/78056
42307         * doc/sourcebuild.texi (PowerPC-specific attributes): Add
42308         documentation of the powerpc_popcntb_ok attribute.
42309         * config/rs6000/rs6000.c (rs6000_option_override_internal): Add
42310         code to issue warning messages if a requested CPU configuration is
42311         not supported by the binary (assembler and loader) toolchain.
42312         (spe_init_builtins): Add two assertions to prevent ICE if attempt is
42313         made to define a built-in function that has been disabled.
42314         (paired_init_builtins): Add assertion to prevent ICE if attempt is
42315         made to define a built-in function that has been disabled.
42316         (altivec_init_builtins): Add comment explaining why definition
42317         of the DST built-in functions is not preceded by an assertion
42318         check.  Add assertions to prevent ICE if attempts are made to
42319         define an altivec predicate or an abs* built-in function that has
42320         been disabled.
42321         (htm_init_builtins): Add comment explaining why definition of the
42322         htm built-in functions is not preceded by an assertion check.
42324 2017-01-04  Jeff Law  <law@redhat.com>
42326         PR tree-optimizatin/67955
42327         * tree-ssa-alias.c (same_addr_size_stores_p): Check offsets first.
42328         Allow any SSA_VAR_P as the base objects.  Use integer_zerop.  Verify
42329         the points-to solution does not include pt_null.  Use DECL_PT_UID
42330         unconditionally.
42332 2017-01-04  Uros Bizjak  <ubizjak@gmail.com>
42334         * config/i386/i386.md (HI/SImode test with imm to QImode splitters):
42335         Use gen_int_mode instead of gen_lopwart for const_int operands.
42337 2017-01-04  Jakub Jelinek  <jakub@redhat.com>
42339         PR tree-optimization/71563
42340         * match.pd: Simplify X << Y into X if Y is known to be 0 or
42341         out of range value - has low bits known to be zero.
42343 2017-01-04  Alan Modra  <amodra@gmail.com>
42345         * Makefile.in (aclocal_deps): Update and order as per aclocal.m4.
42346         * configure: Regenerate.
42347         * config.in: Regenerate.
42349 2017-01-04  Jakub Jelinek  <jakub@redhat.com>
42351         PR bootstrap/77569
42352         * input.c (ebcdic_execution_charset::on_error): Don't use strstr for
42353         a substring of the message, but strcmp with the whole message.  Ifdef
42354         ENABLE_NLS, translate the message first using dgettext.
42356 2017-01-03  Jeff Law  <law@redhat.com>
42358         PR tree-optimizatin/78856
42359         * tree-ssa-threadupdate.c: Include tree-vectorizer.h.
42360         (mark_threaded_blocks): Remove code to truncate thread paths that
42361         cross multiple loop headers.  Instead invalidate the cached loop
42362         iteration information and handle case of a thread path walking
42363         into an irreducible region.
42365 2017-01-03  Michael Meissner  <meissner@linux.vnet.ibm.com>
42367         PR target/78900
42368         * config/rs6000/rs6000.c (rs6000_split_signbit): Change some
42369         assertions.  Add support for doing the signbit if the IEEE 128-bit
42370         floating point value is in a GPR.
42371         * config/rs6000/rs6000.md (Fsignbit): Delete.
42372         (signbit<mode>2_dm): Delete using <Fsignbit> and just use "wa".
42373         Update the length attribute if the value is in a GPR.
42374         (signbit<mode>2_dm_<su>ext): Add combiner pattern to eliminate
42375         the sign or zero extension instruction, since the value is always 0/1.
42376         (signbit<mode>2_dm2): Delete using <Fsignbit>.
42378         PR target/78953
42379         * config/rs6000/vsx.md (vsx_extract_<mode>_store_p9): If we are
42380         extracting SImode to a GPR register so that we can generate a
42381         store, limit the vector to be in a traditional Altivec register
42382         for the vextuwrx instruction.
42384 2017-01-03  Ian Lance Taylor  <iant@google.com>
42386         * godump.c (go_format_type): Treat ENUMERAL_TYPE like INTEGER_TYPE.
42388 2017-01-03  Martin Sebor  <msebor@redhat.com>
42390         PR tree-optimization/78696
42391         * gimple-ssa-sprintf.c (format_floating): Correct handling of
42392         precision.  Use MPFR for %f for greater fidelity.  Correct handling
42393         of %g.
42394         (pass_sprintf_length::compute_format_length): Set width and precision
42395         specified by asrerisk to void_node for vararg functions.
42396         (try_substitute_return_value): Adjust dump output.
42398 2017-01-03  David Edelsohn  <dje.gcc@gmail.com>
42400         * doc/invoke.texi (RS6000 options): LRA is enabled by default.
42402 2017-01-03  Eric Botcazou  <ebotcazou@adacore.com>
42404         * doc/invoke.texi (SPARC options): Document -mlra as the default.
42405         * config/sparc/sparc.c (sparc_option_override): Force LRA unless
42406         -mlra/-mno-lra was passed to the compiler.
42408 2017-01-03  James Cowgill  <James.Cowgill@imgtec.com>
42410         PR rtl-optimization/65618
42411         * emit-rtl.c (try_split): Move initialization of "before" and
42412         "after" to just before the call to emit_insn_after_setloc.
42414 2017-01-03  Gerald Pfeifer  <gerald@pfeifer.com>
42416         * doc/md.texi (Standard Names): Remove reference to Java frontend.
42418 2017-01-03  Pierre-Marie de Rodat  <derodat@adacore.com>
42420         * dwarf2out.c (gen_enumeration_type_die): When
42421         -gno-strict-dwarf, add a DW_AT_encoding attribute.
42423 2017-01-03  Jakub Jelinek  <jakub@redhat.com>
42425         PR tree-optimization/78965
42426         * gimple-ssa-sprintf.c (pass_sprintf_length::compute_format_length):
42427         Change first argument from const call_info & to call_info &.  For %n
42428         set info.nowrite to false.
42430         PR middle-end/78901
42431         * gimple-ssa-sprintf.c (try_substitute_return_value): Don't change
42432         possibly throwing calls.
42434         * genmatch.c (dt_node::gen_kids_1): If generic_exprs include SSA_NAME
42435         and exprs_len || fns_len, emit the code for SSA_NAME next to the exprs
42436         and fns handling, rather than in a separate case SSA_NAME.
42438 2017-01-02  Jeff Law  <law@redhat.com>
42440         * config/darwin-driver.c (darwin_driver_init): Const-correctness
42441         fixes for first_period and second_period variables.
42443 2017-01-02  Uros Bizjak  <ubizjak@gmail.com>
42445         PR target/78967
42446         * config/i386/i386.md (UNSPEC_NOREX_MEM): New unspec.
42447         (*insvqi_1): New insn pattern.
42448         (*insvqi_1_mem_rex64): Ditto.
42449         (*insvqi_2): Ditto.
42450         (*insvqi_3): Rename from *insvqi.
42452         (*extzvqi_mem_rex64): Add UNSPEC_NOREX_MEM tag.
42454 2017-01-02  Gerald Pfeifer  <gerald@pfeifer.com>
42456         * doc/cfg.texi (Edges): Remove reference to Java.
42457         (Maintaining the CFG): Ditto.
42459 2017-01-01  Jan Hubicka  <hubicka@ucw.cz>
42461         PR middle-end/77674
42462         * symtab.c (symtab_node::binds_to_current_def_p): Fix handling of
42463         transparent aliases.
42465 2017-01-01  Jan Hubicka  <hubicka@ucw.cz>
42467         PR middle-end/77484
42468         * predict.def (PRED_CALL): Update hitrate.
42469         (PRED_INDIR_CALL, PRED_POLYMORPHIC_CALL): New predictors.
42470         * predict.c (tree_estimate_probability_bb): Split CALL predictor
42471         into direct/indirect/polymorphic variants.
42473 2017-01-01  Jakub Jelinek  <jakub@redhat.com>
42475         Update copyright years.
42477         * gcc.c (process_command): Update copyright notice dates.
42478         * gcov-dump.c (print_version): Ditto.
42479         * gcov.c (print_version): Ditto.
42480         * gcov-tool.c (print_version): Ditto.
42481         * gengtype.c (create_file): Ditto.
42482         * doc/cpp.texi: Bump @copying's copyright year.
42483         * doc/cppinternals.texi: Ditto.
42484         * doc/gcc.texi: Ditto.
42485         * doc/gccint.texi: Ditto.
42486         * doc/gcov.texi: Ditto.
42487         * doc/install.texi: Ditto.
42488         * doc/invoke.texi: Ditto.
42490 Copyright (C) 2017 Free Software Foundation, Inc.
42492 Copying and distribution of this file, with or without modification,
42493 are permitted in any medium without royalty provided the copyright
42494 notice and this notice are preserved.