poly_int: vect_no_alias_p
[official-gcc.git] / gcc / ChangeLog
blob1f70fcd9cc1c421ec64972bfffd434027b92d4ac
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-data-refs.c (vect_no_alias_p): Replace with...
6         (vect_compile_time_alias): ...this new function.  Do the calculation
7         on poly_ints rather than trees.
8         (vect_prune_runtime_alias_test_list): Update call accordingly.
10 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
11             Alan Hayward  <alan.hayward@arm.com>
12             David Sherwood  <david.sherwood@arm.com>
14         * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
15         numbers of units.
16         (vect_schedule_slp_instance): Likewise.
18 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
19             Alan Hayward  <alan.hayward@arm.com>
20             David Sherwood  <david.sherwood@arm.com>
22         * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
23         constant and extern definitions for variable-length vectors.
24         (vect_get_constant_vectors): Note that the number of units
25         is known to be constant.
27 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28             Alan Hayward  <alan.hayward@arm.com>
29             David Sherwood  <david.sherwood@arm.com>
31         * tree-vect-stmts.c (vectorizable_conversion): Treat the number
32         of units as polynomial.  Choose between WIDE and NARROW based
33         on multiple_p.
35 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
36             Alan Hayward  <alan.hayward@arm.com>
37             David Sherwood  <david.sherwood@arm.com>
39         * tree-vect-stmts.c (simd_clone_subparts): New function.
40         (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
42 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
43             Alan Hayward  <alan.hayward@arm.com>
44             David Sherwood  <david.sherwood@arm.com>
46         * tree-vect-stmts.c (vectorizable_call): Treat the number of
47         vectors as polynomial.  Use build_index_vector for
48         IFN_GOMP_SIMD_LANE.
50 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
51             Alan Hayward  <alan.hayward@arm.com>
52             David Sherwood  <david.sherwood@arm.com>
54         * tree-vect-stmts.c (get_load_store_type): Treat the number of
55         units as polynomial.  Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
56         for variable-length vectors.
57         (vectorizable_mask_load_store): Treat the number of units as
58         polynomial, asserting that it is constant if the condition has
59         already been enforced.
60         (vectorizable_store, vectorizable_load): Likewise.
62 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
63             Alan Hayward  <alan.hayward@arm.com>
64             David Sherwood  <david.sherwood@arm.com>
66         * tree-vect-loop.c (vectorizable_live_operation): Treat the number
67         of units as polynomial.  Punt if we can't tell at compile time
68         which vector contains the final result.
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-loop.c (vectorizable_induction): Treat the number
75         of units as polynomial.  Punt on SLP inductions.  Use an integer
76         VEC_SERIES_EXPR for variable-length integer reductions.  Use a
77         cast of such a series for variable-length floating-point
78         reductions.
80 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
81             Alan Hayward  <alan.hayward@arm.com>
82             David Sherwood  <david.sherwood@arm.com>
84         * tree.h (build_index_vector): Declare.
85         * tree.c (build_index_vector): New function.
86         * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
87         of units as polynomial, forcibly converting it to a constant if
88         vectorizable_reduction has already enforced the condition.
89         (vect_create_epilog_for_reduction): Likewise.  Use build_index_vector
90         to create a {1,2,3,...} vector.
91         (vectorizable_reduction): Treat the number of units as polynomial.
92         Choose vectype_in based on the largest scalar element size rather
93         than the smallest number of units.  Enforce the restrictions
94         relied on above.
96 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
97             Alan Hayward  <alan.hayward@arm.com>
98             David Sherwood  <david.sherwood@arm.com>
100         * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
101         number of units as polynomial.
103 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
104             Alan Hayward  <alan.hayward@arm.com>
105             David Sherwood  <david.sherwood@arm.com>
107         * target.h (vector_sizes, auto_vector_sizes): New typedefs.
108         * target.def (autovectorize_vector_sizes): Return the vector sizes
109         by pointer, using vector_sizes rather than a bitmask.
110         * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
111         * targhooks.c (default_autovectorize_vector_sizes): Likewise.
112         * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
113         Likewise.
114         * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
115         * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
116         * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
117         * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
118         * omp-general.c (omp_max_vf): Likewise.
119         * omp-low.c (omp_clause_aligned_alignment): Likewise.
120         * optabs-query.c (can_vec_mask_load_store_p): Likewise.
121         * tree-vect-loop.c (vect_analyze_loop): Likewise.
122         * tree-vect-slp.c (vect_slp_bb): Likewise.
123         * doc/tm.texi: Regenerate.
124         * tree-vectorizer.h (current_vector_size): Change from an unsigned int
125         to a poly_uint64.
126         * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
127         the vector size as a poly_uint64 rather than an unsigned int.
128         (current_vector_size): Change from an unsigned int to a poly_uint64.
129         (get_vectype_for_scalar_type): Update accordingly.
130         * tree.h (build_truth_vector_type): Take the size and number of
131         units as a poly_uint64 rather than an unsigned int.
132         (build_vector_type): Add a temporary overload that takes
133         the number of units as a poly_uint64 rather than an unsigned int.
134         * tree.c (make_vector_type): Likewise.
135         (build_truth_vector_type): Take the number of units as a poly_uint64
136         rather than an unsigned int.
138 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
139             Alan Hayward  <alan.hayward@arm.com>
140             David Sherwood  <david.sherwood@arm.com>
142         * target.def (get_mask_mode): Take the number of units and length
143         as poly_uint64s rather than unsigned ints.
144         * targhooks.h (default_get_mask_mode): Update accordingly.
145         * targhooks.c (default_get_mask_mode): Likewise.
146         * config/i386/i386.c (ix86_get_mask_mode): Likewise.
147         * doc/tm.texi: Regenerate.
149 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
150             Alan Hayward  <alan.hayward@arm.com>
151             David Sherwood  <david.sherwood@arm.com>
153         * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
154         * omp-general.c (omp_max_vf): Likewise.
155         * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
156         (expand_omp_simd): Handle polynomial safelen.
157         * omp-low.c (omplow_simd_context): Add a default constructor.
158         (omplow_simd_context::max_vf): Change from int to poly_uint64.
159         (lower_rec_simd_input_clauses): Update accordingly.
160         (lower_rec_input_clauses): Likewise.
162 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
163             Alan Hayward  <alan.hayward@arm.com>
164             David Sherwood  <david.sherwood@arm.com>
166         * tree-vectorizer.h (vect_nunits_for_cost): New function.
167         * tree-vect-loop.c (vect_model_reduction_cost): Use it.
168         * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
169         (vect_analyze_slp_cost): Likewise.
170         * tree-vect-stmts.c (vect_model_store_cost): Likewise.
171         (vect_model_load_cost): Likewise.
173 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
174             Alan Hayward  <alan.hayward@arm.com>
175             David Sherwood  <david.sherwood@arm.com>
177         * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
178         (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
179         from an unsigned int * to a poly_uint64_pod *.
180         (calculate_unrolling_factor): New function.
181         (vect_analyze_slp_instance): Use it.  Track polynomial max_nunits.
183 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
184             Alan Hayward  <alan.hayward@arm.com>
185             David Sherwood  <david.sherwood@arm.com>
187         * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
188         from an unsigned int to a poly_uint64.
189         (_loop_vec_info::slp_unrolling_factor): Likewise.
190         (_loop_vec_info::vectorization_factor): Change from an int
191         to a poly_uint64.
192         (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
193         (vect_get_num_vectors): New function.
194         (vect_update_max_nunits, vect_vf_for_cost): Likewise.
195         (vect_get_num_copies): Use vect_get_num_vectors.
196         (vect_analyze_data_ref_dependences): Change max_vf from an int *
197         to an unsigned int *.
198         (vect_analyze_data_refs): Change min_vf from an int * to a
199         poly_uint64 *.
200         (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
201         than an unsigned HOST_WIDE_INT.
202         * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
203         (vect_analyze_data_ref_dependence): Change max_vf from an int *
204         to an unsigned int *.
205         (vect_analyze_data_ref_dependences): Likewise.
206         (vect_compute_data_ref_alignment): Handle polynomial vf.
207         (vect_enhance_data_refs_alignment): Likewise.
208         (vect_prune_runtime_alias_test_list): Likewise.
209         (vect_shift_permute_load_chain): Likewise.
210         (vect_supportable_dr_alignment): Likewise.
211         (dependence_distance_ge_vf): Take the vectorization factor as a
212         poly_uint64 rather than an unsigned HOST_WIDE_INT.
213         (vect_analyze_data_refs): Change min_vf from an int * to a
214         poly_uint64 *.
215         * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
216         vfm1 as a poly_uint64 rather than an int.  Make the same change
217         for the returned bound_scalar.
218         (vect_gen_vector_loop_niters): Handle polynomial vf.
219         (vect_do_peeling): Likewise.  Update call to
220         vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
221         (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
222         be constant.
223         * tree-vect-loop.c (vect_determine_vectorization_factor)
224         (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
225         (vect_get_known_peeling_cost): Likewise.
226         (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
227         (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
228         (vect_transform_loop): Likewise.  Use the lowest possible VF when
229         updating the upper bounds of the loop.
230         (vect_min_worthwhile_factor): Make static.  Return an unsigned int
231         rather than an int.
232         * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
233         polynomial unroll factors.
234         (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
235         (vect_make_slp_decision): Likewise.
236         (vect_supported_load_permutation_p): Likewise, and polynomial
237         vf too.
238         (vect_analyze_slp_cost): Handle polynomial vf.
239         (vect_slp_analyze_node_operations): Likewise.
240         (vect_slp_analyze_bb_1): Likewise.
241         (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
242         than an unsigned HOST_WIDE_INT.
243         * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
244         (vectorizable_load): Handle polynomial vf.
245         * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
246         a poly_uint64.
247         (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
249 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
250             Alan Hayward  <alan.hayward@arm.com>
251             David Sherwood  <david.sherwood@arm.com>
253         * match.pd: Handle bit operations involving three constants
254         and try to fold one pair.
256 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
258         * tree-vect-loop-manip.c: Include gimple-fold.h.
259         (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
260         niters_maybe_zero parameters.  Handle other cases besides a step of 1.
261         (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
262         Add a path that uses a step of VF instead of 1, but disable it
263         for now.
264         (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
265         and niters_no_overflow parameters.  Update calls to
266         slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
267         Create a new SSA name if the latter choses to use a ste other
268         than zero, and return it via niters_vector_mult_vf_var.
269         * tree-vect-loop.c (vect_transform_loop): Update calls to
270         vect_do_peeling, vect_gen_vector_loop_niters and
271         slpeel_make_loop_iterate_ntimes.
272         * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
273         (vect_gen_vector_loop_niters): Update declarations after above changes.
275 2018-01-02  Michael Meissner  <meissner@linux.vnet.ibm.com>
277         * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
278         128-bit round to integer instructions.
279         (ceil<mode>2): Likewise.
280         (btrunc<mode>2): Likewise.
281         (round<mode>2): Likewise.
283 2018-01-02  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
285         * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
286         unaligned VSX load/store on P8/P9.
287         (expand_block_clear): Allow the use of unaligned VSX
288         load/store on P8/P9.
290 2018-01-02  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
292         * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
293         New function.
294         (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
295         swap associated with both a load and a store.
297 2018-01-02  Andrew Waterman  <andrew@sifive.com>
299         * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
300         * config/riscv/riscv.md (clear_cache): Use it.
302 2018-01-02  Artyom Skrobov  <tyomitch@gmail.com>
304         * web.c: Remove out-of-date comment.
306 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
308         * expr.c (fixup_args_size_notes): Check that any existing
309         REG_ARGS_SIZE notes are correct, and don't try to re-add them.
310         (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
311         (emit_single_push_insn): ...here.
313 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
315         * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
316         (const_vector_encoded_nelts): New function.
317         (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
318         (const_vector_int_elt, const_vector_elt): Declare.
319         * emit-rtl.c (const_vector_int_elt_1): New function.
320         (const_vector_elt): Likewise.
321         * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
322         of CONST_VECTOR_ELT.
324 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
326         * expr.c: Include rtx-vector-builder.h.
327         (const_vector_mask_from_tree): Use rtx_vector_builder and operate
328         directly on the tree encoding.
329         (const_vector_from_tree): Likewise.
330         * optabs.c: Include rtx-vector-builder.h.
331         (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
332         sequence of "u" values.
333         * vec-perm-indices.c: Include rtx-vector-builder.h.
334         (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
335         directly on the vec_perm_indices encoding.
337 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
339         * doc/rtl.texi (const_vector): Describe new encoding scheme.
340         * Makefile.in (OBJS): Add rtx-vector-builder.o.
341         * rtx-vector-builder.h: New file.
342         * rtx-vector-builder.c: Likewise.
343         * rtl.h (rtx_def::u2): Add a const_vector field.
344         (CONST_VECTOR_NPATTERNS): New macro.
345         (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
346         (CONST_VECTOR_DUPLICATE_P): Likewise.
347         (CONST_VECTOR_STEPPED_P): Likewise.
348         (CONST_VECTOR_ENCODED_ELT): Likewise.
349         (const_vec_duplicate_p): Check for a duplicated vector encoding.
350         (unwrap_const_vec_duplicate): Likewise.
351         (const_vec_series_p): Check for a non-duplicated vector encoding.
352         Say that the function only returns true for integer vectors.
353         * emit-rtl.c: Include rtx-vector-builder.h.
354         (gen_const_vec_duplicate_1): Delete.
355         (gen_const_vector): Call gen_const_vec_duplicate instead of
356         gen_const_vec_duplicate_1.
357         (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
358         (gen_const_vec_duplicate): Use rtx_vector_builder.
359         (gen_const_vec_series): Likewise.
360         (gen_rtx_CONST_VECTOR): Likewise.
361         * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
362         (swap_const_vector_halves): Take an rtx pointer rather than rtx.
363         Build a new vector rather than modifying a CONST_VECTOR in-place.
364         (handle_special_swappables): Update call accordingly.
365         * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
366         (swap_const_vector_halves): Take an rtx pointer rather than rtx.
367         Build a new vector rather than modifying a CONST_VECTOR in-place.
368         (handle_special_swappables): Update call accordingly.
370 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
372         * simplify-rtx.c (simplify_const_binary_operation): Use
373         CONST_VECTOR_ELT instead of XVECEXP.
375 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
377         * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
378         the selector elements to be different from the data elements
379         if the selector is a VECTOR_CST.
380         * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
381         ssizetype for the selector.
383 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
385         * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
386         before testing each element individually.
387         * tree-vect-generic.c (lower_vec_perm): Likewise.
389 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
391         * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
392         * selftest-run-tests.c (selftest::run_tests): Call it.
393         * vector-builder.h (vector_builder::operator ==): New function.
394         (vector_builder::operator !=): Likewise.
395         * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
396         (vec_perm_indices::all_from_input_p): New function.
397         * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
398         (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
399         * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
400         instead of reading the VECTOR_CST directly.  Detect whether both
401         vector inputs are the same before constructing the vec_perm_indices,
402         and update the number of inputs argument accordingly.  Use the
403         utility functions added above.  Only construct sel2 if we need to.
405 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
407         * optabs.c (expand_vec_perm_var): Use an explicit encoding for
408         the broadcast of the low byte.
409         (expand_mult_highpart): Use an explicit encoding for the permutes.
410         * optabs-query.c (can_mult_highpart_p): Likewise.
411         * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
412         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
413         (vectorizable_bswap): Likewise.
414         * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
415         explicit encoding for the power-of-2 permutes.
416         (vect_permute_store_chain): Likewise.
417         (vect_grouped_load_supported): Likewise.
418         (vect_permute_load_chain): Likewise.
420 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
422         * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
423         * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
424         * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
425         * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
426         * tree-vect-stmts.c (vectorizable_bswap): Likewise.
427         (vect_gen_perm_mask_any): Likewise.
429 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
431         * int-vector-builder.h: New file.
432         * vec-perm-indices.h: Include int-vector-builder.h.
433         (vec_perm_indices): Redefine as an int_vector_builder.
434         (auto_vec_perm_indices): Delete.
435         (vec_perm_builder): Redefine as a stand-alone class.
436         (vec_perm_indices::vec_perm_indices): New function.
437         (vec_perm_indices::clamp): Likewise.
438         * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
439         (vec_perm_indices::new_vector): New function.
440         (vec_perm_indices::new_expanded_vector): Update for new
441         vec_perm_indices class.
442         (vec_perm_indices::rotate_inputs): New function.
443         (vec_perm_indices::all_in_range_p): Operate directly on the
444         encoded form, without computing elided elements.
445         (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
446         encoding.  Update for new vec_perm_indices class.
447         * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
448         the given vec_perm_builder.
449         (expand_vec_perm_var): Update vec_perm_builder constructor.
450         (expand_mult_highpart): Use vec_perm_builder instead of
451         auto_vec_perm_indices.
452         * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
453         vec_perm_indices instead of auto_vec_perm_indices.  Use a single
454         or double series encoding as appropriate.
455         * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
456         vec_perm_indices instead of auto_vec_perm_indices.
457         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
458         * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
459         (vect_permute_store_chain): Likewise.
460         (vect_grouped_load_supported): Likewise.
461         (vect_permute_load_chain): Likewise.
462         (vect_shift_permute_load_chain): Likewise.
463         * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
464         (vect_transform_slp_perm_load): Likewise.
465         (vect_schedule_slp_instance): Likewise.
466         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
467         (vectorizable_mask_load_store): Likewise.
468         (vectorizable_bswap): Likewise.
469         (vectorizable_store): Likewise.
470         (vectorizable_load): Likewise.
471         * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
472         vec_perm_indices instead of auto_vec_perm_indices.  Use
473         tree_to_vec_perm_builder to read the vector from a tree.
474         * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
475         vec_perm_builder instead of a vec_perm_indices.
476         (have_whole_vector_shift): Use vec_perm_builder and
477         vec_perm_indices instead of auto_vec_perm_indices.  Leave the
478         truncation to calc_vec_perm_mask_for_shift.
479         (vect_create_epilog_for_reduction): Likewise.
480         * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
481         from auto_vec_perm_indices to vec_perm_indices.
482         (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
483         instead of changing individual elements.
484         (aarch64_vectorize_vec_perm_const): Use new_vector to install
485         the vector in d.perm.
486         * config/arm/arm.c (expand_vec_perm_d::perm): Change
487         from auto_vec_perm_indices to vec_perm_indices.
488         (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
489         instead of changing individual elements.
490         (arm_vectorize_vec_perm_const): Use new_vector to install
491         the vector in d.perm.
492         * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
493         Update vec_perm_builder constructor.
494         (rs6000_expand_interleave): Likewise.
495         * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
496         (rs6000_expand_interleave): Likewise.
498 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
500         * optabs-query.c (can_vec_perm_var_p): Check whether lowering
501         to qimode could truncate the indices.
502         * optabs.c (expand_vec_perm_var): Likewise.
504 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
506         * Makefile.in (OBJS): Add vec-perm-indices.o.
507         * vec-perm-indices.h: New file.
508         * vec-perm-indices.c: Likewise.
509         * target.h (vec_perm_indices): Replace with a forward class
510         declaration.
511         (auto_vec_perm_indices): Move to vec-perm-indices.h.
512         * optabs.h: Include vec-perm-indices.h.
513         (expand_vec_perm): Delete.
514         (selector_fits_mode_p, expand_vec_perm_var): Declare.
515         (expand_vec_perm_const): Declare.
516         * target.def (vec_perm_const_ok): Replace with...
517         (vec_perm_const): ...this new hook.
518         * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
519         (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
520         * doc/tm.texi: Regenerate.
521         * optabs.def (vec_perm_const): Delete.
522         * doc/md.texi (vec_perm_const): Likewise.
523         (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
524         * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
525         expand_vec_perm for constant permutation vectors.  Assert that
526         the mode of variable permutation vectors is the integer equivalent
527         of the mode that is being permuted.
528         * optabs-query.h (selector_fits_mode_p): Declare.
529         * optabs-query.c: Include vec-perm-indices.h.
530         (selector_fits_mode_p): New function.
531         (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
532         is defined, instead of checking whether the vec_perm_const_optab
533         exists.  Use targetm.vectorize.vec_perm_const instead of
534         targetm.vectorize.vec_perm_const_ok.  Check whether the indices
535         fit in the vector mode before using a variable permute.
536         * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
537         vec_perm_indices instead of an rtx.
538         (expand_vec_perm): Replace with...
539         (expand_vec_perm_const): ...this new function.  Take the selector
540         as a vec_perm_indices rather than an rtx.  Also take the mode of
541         the selector.  Update call to shift_amt_for_vec_perm_mask.
542         Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
543         Use vec_perm_indices::new_expanded_vector to expand the original
544         selector into bytes.  Check whether the indices fit in the vector
545         mode before using a variable permute.
546         (expand_vec_perm_var): Make global.
547         (expand_mult_highpart): Use expand_vec_perm_const.
548         * fold-const.c: Includes vec-perm-indices.h.
549         * tree-ssa-forwprop.c: Likewise.
550         * tree-vect-data-refs.c: Likewise.
551         * tree-vect-generic.c: Likewise.
552         * tree-vect-loop.c: Likewise.
553         * tree-vect-slp.c: Likewise.
554         * tree-vect-stmts.c: Likewise.
555         * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
556         Delete.
557         * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
558         * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
559         (aarch64_vectorize_vec_perm_const_ok): Fuse into...
560         (aarch64_vectorize_vec_perm_const): ...this new function.
561         (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
562         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
563         * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
564         * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
565         * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
566         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
567         (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
568         into...
569         (arm_vectorize_vec_perm_const): ...this new function.  Explicitly
570         check for NEON modes.
571         * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
572         * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
573         * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
574         (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
575         into...
576         (ix86_vectorize_vec_perm_const): ...this new function.  Incorporate
577         the old VEC_PERM_CONST conditions.
578         * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
579         * config/ia64/vect.md (vec_perm_const<mode>): Delete.
580         * config/ia64/ia64.c (ia64_expand_vec_perm_const)
581         (ia64_vectorize_vec_perm_const_ok): Merge into...
582         (ia64_vectorize_vec_perm_const): ...this new function.
583         * config/mips/loongson.md (vec_perm_const<mode>): Delete.
584         * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
585         * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
586         * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
587         * config/mips/mips.c (mips_expand_vec_perm_const)
588         (mips_vectorize_vec_perm_const_ok): Merge into...
589         (mips_vectorize_vec_perm_const): ...this new function.
590         * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
591         * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
592         * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
593         * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
594         * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
595         (rs6000_expand_vec_perm_const): Delete.
596         * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
597         Delete.
598         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
599         (altivec_expand_vec_perm_const_le): Take each operand individually.
600         Operate on constant selectors rather than rtxes.
601         (altivec_expand_vec_perm_const): Likewise.  Update call to
602         altivec_expand_vec_perm_const_le.
603         (rs6000_expand_vec_perm_const): Delete.
604         (rs6000_vectorize_vec_perm_const_ok): Delete.
605         (rs6000_vectorize_vec_perm_const): New function.
606         (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
607         an element count and rtx array.
608         (rs6000_expand_extract_even): Update call accordingly.
609         (rs6000_expand_interleave): Likewise.
610         * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
611         * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
612         * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
613         * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
614         (rs6000_expand_vec_perm_const): Delete.
615         * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
616         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
617         (altivec_expand_vec_perm_const_le): Take each operand individually.
618         Operate on constant selectors rather than rtxes.
619         (altivec_expand_vec_perm_const): Likewise.  Update call to
620         altivec_expand_vec_perm_const_le.
621         (rs6000_expand_vec_perm_const): Delete.
622         (rs6000_vectorize_vec_perm_const_ok): Delete.
623         (rs6000_vectorize_vec_perm_const): New function.  Remove stray
624         reference to the SPE evmerge intructions.
625         (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
626         an element count and rtx array.
627         (rs6000_expand_extract_even): Update call accordingly.
628         (rs6000_expand_interleave): Likewise.
629         * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
630         * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
631         new function.
632         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
634 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
636         * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
637         vector mode and that that mode matches the mode of the data
638         being permuted.
639         (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
640         out into expand_vec_perm_var.  Do all CONST_VECTOR handling here,
641         directly using expand_vec_perm_1 when forcing selectors into
642         registers.
643         (expand_vec_perm_var): New function, split out from expand_vec_perm.
645 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
647         * optabs-query.h (can_vec_perm_p): Delete.
648         (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
649         * optabs-query.c (can_vec_perm_p): Split into...
650         (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
651         (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
652         particular selector is valid.
653         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
654         * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
655         (vect_grouped_load_supported): Likewise.
656         (vect_shift_permute_load_chain): Likewise.
657         * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
658         (vect_transform_slp_perm_load): Likewise.
659         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
660         (vectorizable_bswap): Likewise.
661         (vect_gen_perm_mask_checked): Likewise.
662         * fold-const.c (fold_ternary_loc): Likewise.  Don't take
663         implementations of variable permutation vectors into account
664         when deciding which selector to use.
665         * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
666         vec_perm_const_optab is supported; instead use can_vec_perm_const_p
667         with a false third argument.
668         * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
669         to test whether the constant selector is valid and can_vec_perm_var_p
670         to test whether a variable selector is valid.
672 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
674         * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
675         * optabs-query.c (can_vec_perm_p): Likewise.
676         * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
677         instead of vec_perm_indices.
678         * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
679         (vect_gen_perm_mask_checked): Likewise,
680         * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
681         (vect_gen_perm_mask_checked): Likewise,
683 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
685         * optabs-query.h (qimode_for_vec_perm): Declare.
686         * optabs-query.c (can_vec_perm_p): Split out qimode search to...
687         (qimode_for_vec_perm): ...this new function.
688         * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
690 2018-01-02  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
692         * rtlanal.c (canonicalize_condition): Return 0 if final rtx
693         does not have a conditional at the top.
695 2018-01-02  Richard Biener  <rguenther@suse.de>
697         * ipa-inline.c (big_speedup_p): Fix expression.
699 2018-01-02  Jan Hubicka  <hubicka@ucw.cz>
701         PR target/81616
702         * config/i386/x86-tune-costs.h: Increase cost of integer load costs
703         for generic 4->6.
705 2018-01-02  Jan Hubicka  <hubicka@ucw.cz>
707         PR target/81616
708         Generic tuning.
709         * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
710         cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
711         and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
712         cond_taken_branch_cost 3->4.
714 2018-01-01  Jakub Jelinek  <jakub@redhat.com>
716         PR tree-optimization/83581
717         * tree-loop-distribution.c (pass_loop_distribution::execute): Return
718         TODO_cleanup_cfg if any changes have been made.
720         PR middle-end/83608
721         * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
722         convert_modes if target mode has the right side, but different mode
723         class.
725         PR middle-end/83609
726         * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
727         last argument when extracting from CONCAT.  If either from_real or
728         from_imag is NULL, use expansion through memory.  If result is not
729         a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
730         the parts directly to inner mode, if even that fails, use expansion
731         through memory.
733         PR middle-end/83623
734         * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
735         check for bswap in mode rather than HImode and use that in expand_unop
736         too.
738 2017-12-31  Jakub Jelinek  <jakub@redhat.com>
740         PR target/83536
741         * config/i386/i386.c (ix86_attribute_table): Remove excess
742         initializer for "shared" attribute.
744 2017-12-30  Tom de Vries  <tom@codesourcery.com>
746         PR libgomp/83046
747         * omp-expand.c (expand_omp_target): If in_lto_p, mark offload_funcs with
748         DECL_PRESERVE_P.
749         * lto-streamer-out.c (prune_offload_funcs): New function.  Remove
750         offload_funcs entries that no longer have a corresponding cgraph_node.
751         Mark the remaining ones as DECL_PRESERVE_P.
752         (output_lto): Call prune_offload_funcs.
754 2017-12-30  Jakub Jelinek  <jakub@redhat.com>
756         * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
757         vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>,
758         vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
759         vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
760         vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
761         vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1, vpdpbusd_<mode>,
762         vpdpbusd_<mode>_mask, vpdpbusd_<mode>_maskz, vpdpbusd_<mode>_maskz_1,
763         vpdpbusds_<mode>, vpdpbusds_<mode>_mask, vpdpbusds_<mode>_maskz,
764         vpdpbusds_<mode>_maskz_1, vpdpwssd_<mode>, vpdpwssd_<mode>_mask,
765         vpdpwssd_<mode>_maskz, vpdpwssd_<mode>_maskz_1, vpdpwssds_<mode>,
766         vpdpwssds_<mode>_mask, vpdpwssds_<mode>_maskz,
767         vpdpwssds_<mode>_maskz_1, vaesdec_<mode>, vaesdeclast_<mode>,
768         vaesenc_<mode>, vpclmulqdq_<mode>,
769         avx512vl_vpshufbitqmb<mode><mask_scalar_merge_name>): Formatting fixes.
771 2017-12-28  Michael Meissner  <meissner@linux.vnet.ibm.com>
773         * builtins.def: (_Float<N> and _Float<N>X BUILT_IN_CEIL): Add
774         _Float<N> and _Float<N>X variants for rounding built-in
775         functions.
776         (_Float<N> and _Float<N>X BUILT_IN_FLOOR): Likewise.
777         (_Float<N> and _Float<N>X BUILT_IN_NEARBYINT): Likewise.
778         (_Float<N> and _Float<N>X BUILT_IN_RINT): Likewise.
779         (_Float<N> and _Float<N>X BUILT_IN_ROUND): Likewise.
780         (_Float<N> and _Float<N>X BUILT_IN_TRUNC): Likewise.
781         * builtins.c (mathfn_built_in_2): Likewise.
782         * internal-fn.def (CEIL): Likewise.
783         (FLOOR): Likewise.
784         (NEARBYINT): Likewise.
785         (RINT): Likewise.
786         (ROUND): Likewise.
787         (TRUNC): Likewise.
788         * convert.c (convert_to_integer_1): Likewise.
789         * fold-const.c (tree_call_nonnegative_warnv_p): Likewise.
790         (integer_valued_real_call_p): Likewise.
791         * fold-const-call.c (fold_const_call_ss): Likewise.
792         * gencfn-macros.c (print_case_cfn): Change CFN and operator
793         printers to take a const char * suffix instead of a bool.
794         (print_define_operator_list): Likewise.
795         (fltall_suffixes): New list of suffixes, that include the
796         traditional suffixes as well as all of the _Float<N> and
797         _Float<N>X suffixes.
798         (main): For _Float<N> and _Float<N>X functions, emit both
799         <name>_FN and <name>_ALL variants.  The <macro>_FN variant only
800         has the _Float<N> and _Float<N>X case names or operators.  The
801         <name>_ALL variant has both the traditional and the
802         _Float<N>/_Float<N>X case names or operators.
803         * match.pd (COPYSIGN optimizations): Provide optimizations for
804         _Float<N> and _Float<N>X types where possible.
805         (MIN/MAX optimizations): Likewise.
806         (sqrt optimizations): Likewise.
807         (rounding optimizations): Likewise.
809 2017-12-28  Richard Sandiford  <richard.sandiford@linaro.org>
811         * config/powerpcspe/powerpcspe.md (ieee_128bit_negative_zero): Use
812         gen_int_mode rather than GEN_INT.
813         * config/rs6000/rs6000.md (ieee_128bit_negative_zero): Likewise.
815 2017-12-28  Richard Sandiford  <richard.sandiford@linaro.org>
817         * emit-rtl.c (gen_const_vec_series): Use valid_for_const_vector_p
818         instead of CONSTANT_P.
819         (gen_vec_series): Likewise.
820         * simplify-rtx.c (simplify_binary_operation_1): Likewise.
822 2017-12-28  Andreas Schwab  <schwab@linux-m68k.org>
824         * config/m68k/m68k.md (ashrdi3_const1, lshrdi3_const1): Add
825         CC_STATUS_INIT.
827 2017-12-27  Kugan Vivekanandarajah  <kuganv@linaro.org>
829         * config/aarch64/aarch64-simd.md (aarch64_ld1x2<VQ:mode>): New.
830         (aarch64_ld1x2<VDC:mode>): Likewise.
831         (aarch64_simd_ld1<mode>_x2): Likewise.
832         (aarch64_simd_ld1<mode>_x2): Likewise.
833         * config/aarch64/arm_neon.h (vld1_u8_x2): New.
834         (vld1_s8_x2): Likewise.
835         (vld1_u16_x2): Likewise.
836         (vld1_s16_x2): Likewise.
837         (vld1_u32_x2): Likewise.
838         (vld1_s32_x2): Likewise.
839         (vld1_u64_x2): Likewise.
840         (vld1_s64_x2): Likewise.
841         (vld1_f16_x2): Likewise.
842         (vld1_f32_x2): Likewise.
843         (vld1_f64_x2): Likewise.
844         (vld1_p8_x2): Likewise.
845         (vld1_p16_x2): Likewise.
846         (vld1_p64_x2): Likewise.
847         (vld1q_u8_x2): Likewise.
848         (vld1q_s8_x2): Likewise.
849         (vld1q_u16_x2): Likewise.
850         (vld1q_s16_x2): Likewise.
851         (vld1q_u32_x2): Likewise.
852         (vld1q_s32_x2): Likewise.
853         (vld1q_u64_x2): Likewise.
854         (vld1q_s64_x2): Likewise.
855         (vld1q_f16_x2): Likewise.
856         (vld1q_f32_x2): Likewise.
857         (vld1q_f64_x2): Likewise.
858         (vld1q_p8_x2): Likewise.
859         (vld1q_p16_x2): Likewise.
860         (vld1q_p64_x2): Likewise.
862 2017-12-27  Martin Liska  <mliska@suse.cz>
864         PR tree-optimization/83552
865         * tree-ssa-strlen.c (fold_strstr_to_strncmp): Assign result
866         of get_string_lenth to a SSA_NAME if not a GIMPLE value.
868 2017-12-27  Tom de Vries  <tom@codesourcery.com>
870         * config/nvptx/nvptx.c (nvptx_option_override): Disable
871         -gstatement-frontiers.
873 2017-12-26  Alexander Monakov  <amonakov@ispras.ru>
875         PR rtl-optimization/83513
876         * sel-sched.c (sel_rank_for_schedule): Order by non-zero usefulness
877         before priority comparison.
879 2017-12-25  Jakub Jelinek  <jakub@redhat.com>
881         PR target/83488
882         * config/i386/i386.opt (-mavx512vpopcntdq, -mmavx512bitalg): Move from
883         ix86_isa_flags2 to ix86_isa_flags.
884         * config/i386/i386-c.c (ix86_target_macros_internal): Test
885         OPTION_MASK_ISA_AVX512BITALG and OPTION_MASK_ISA_AVX512VPOPCNTDQ in
886         isa_flags rather than isa_flags2.
887         * config/i386/i386.c (ix86_target_string): Move -mavx512vpopcntdq
888         and -mavx512bitalg from isa2_opts to isa_opts.
889         (ix86_option_override_internal): Test OPTION_MASK_ISA_AVX512VPOPCNTDQ
890         in x_ix86_isa_flags_explicit rather than x_ix86_isa_flags2_explicit
891         and set it in x_ix86_isa_flags rather than x_ix86_isa_flags2.
892         Formatting fixes.
893         (def_builtin): Treat OPTION_MASK_ISA_AVX512BW or
894         OPTION_MASK_ISA_AVX512F ored with another option similarly to
895         OPTION_MASK_ISA_AVX512VL.  Even for OPTION_MASK_ISA_AVX512VL don't
896         clear it if mask is just OPTION_MASK_ISA_AVX512VL itself.
897         (ix86_expand_builtin): Don't handle OPTION_MASK_ISA_GFNI and
898         OPTION_MASK_ISA_VPCLMULQDQ specially, instead handle
899         OPTION_MASK_ISA_AVX512BW and OPTION_MASK_ISA_AVX512F that way.
900         * config/i386/i386-builtin.def: Move AVX512VPOPCNTDQ and AVX512BITALG
901         builtins from bdesc_args2 to bdesc_args section.
902         (__builtin_ia32_compressstoreuqi512_mask,
903         __builtin_ia32_compressstoreuhi512_mask,
904         __builtin_ia32_compressstoreuqi256_mask,
905         __builtin_ia32_expandloadqi512_mask,
906         __builtin_ia32_expandloadqi512_maskz,
907         __builtin_ia32_expandloadhi512_mask,
908         __builtin_ia32_expandloadhi512_maskz,
909         __builtin_ia32_compressqi512_mask, __builtin_ia32_compresshi512_mask,
910         __builtin_ia32_compressqi256_mask, __builtin_ia32_expandqi512_mask,
911         __builtin_ia32_expandqi512_maskz, __builtin_ia32_expandhi512_mask,
912         __builtin_ia32_expandhi512_maskz, __builtin_ia32_expandqi256_mask,
913         __builtin_ia32_expandqi256_maskz, __builtin_ia32_vpshrd_v32hi_mask,
914         __builtin_ia32_vpshld_v32hi_mask, __builtin_ia32_vpshrdv_v32hi_mask,
915         __builtin_ia32_vpshrdv_v32hi_maskz, __builtin_ia32_vpshldv_v32hi_mask,
916         __builtin_ia32_vpshldv_v32hi_maskz,
917         __builtin_ia32_vpopcountb_v64qi_mask,
918         __builtin_ia32_vpopcountw_v32hi_mask,
919         __builtin_ia32_vpshufbitqmb512_mask,
920         __builtin_ia32_vpshufbitqmb256_mask): Add
921         " | OPTION_MASK_ISA_AVX512BW".
922         (__builtin_ia32_expandloadqi256_mask,
923         __builtin_ia32_expandloadqi256_maskz,
924         __builtin_ia32_vpopcountb_v32qi_mask): Add
925         " | OPTION_MASK_ISA_AVX512VL | OPTION_MASK_ISA_AVX512BW".
926         (__builtin_ia32_expandloadhi256_mask,
927         __builtin_ia32_expandloadhi256_maskz,
928         __builtin_ia32_expandloadqi128_mask,
929         __builtin_ia32_expandloadqi128_maskz,
930         __builtin_ia32_expandloadhi128_mask,
931         __builtin_ia32_expandloadhi128_maskz,
932         __builtin_ia32_vpshrd_v16hi, __builtin_ia32_vpshrd_v16hi_mask,
933         __builtin_ia32_vpshrd_v8hi, __builtin_ia32_vpshrd_v8hi_mask,
934         __builtin_ia32_vpshrd_v8si, __builtin_ia32_vpshrd_v8si_mask,
935         __builtin_ia32_vpshrd_v4si, __builtin_ia32_vpshrd_v4si_mask,
936         __builtin_ia32_vpshrd_v4di, __builtin_ia32_vpshrd_v4di_mask,
937         __builtin_ia32_vpshrd_v2di, __builtin_ia32_vpshrd_v2di_mask,
938         __builtin_ia32_vpshld_v16hi, __builtin_ia32_vpshld_v16hi_mask,
939         __builtin_ia32_vpshld_v8hi, __builtin_ia32_vpshld_v8hi_mask,
940         __builtin_ia32_vpshld_v8si, __builtin_ia32_vpshld_v8si_mask,
941         __builtin_ia32_vpshld_v4si, __builtin_ia32_vpshld_v4si_mask,
942         __builtin_ia32_vpshld_v4di, __builtin_ia32_vpshld_v4di_mask,
943         __builtin_ia32_vpshld_v2di, __builtin_ia32_vpshld_v2di_mask,
944         __builtin_ia32_vpshrdv_v16hi, __builtin_ia32_vpshrdv_v16hi_mask,
945         __builtin_ia32_vpshrdv_v16hi_maskz, __builtin_ia32_vpshrdv_v8hi,
946         __builtin_ia32_vpshrdv_v8hi_mask, __builtin_ia32_vpshrdv_v8hi_maskz,
947         __builtin_ia32_vpshrdv_v8si, __builtin_ia32_vpshrdv_v8si_mask,
948         __builtin_ia32_vpshrdv_v8si_maskz, __builtin_ia32_vpshrdv_v4si,
949         __builtin_ia32_vpshrdv_v4si_mask, __builtin_ia32_vpshrdv_v4si_maskz,
950         __builtin_ia32_vpshrdv_v4di, __builtin_ia32_vpshrdv_v4di_mask,
951         __builtin_ia32_vpshrdv_v4di_maskz, __builtin_ia32_vpshrdv_v2di,
952         __builtin_ia32_vpshrdv_v2di_mask, __builtin_ia32_vpshrdv_v2di_maskz,
953         __builtin_ia32_vpshldv_v16hi, __builtin_ia32_vpshldv_v16hi_mask,
954         __builtin_ia32_vpshldv_v16hi_maskz, __builtin_ia32_vpshldv_v8hi,
955         __builtin_ia32_vpshldv_v8hi_mask, __builtin_ia32_vpshldv_v8hi_maskz,
956         __builtin_ia32_vpshldv_v8si, __builtin_ia32_vpshldv_v8si_mask,
957         __builtin_ia32_vpshldv_v8si_maskz, __builtin_ia32_vpshldv_v4si,
958         __builtin_ia32_vpshldv_v4si_mask, __builtin_ia32_vpshldv_v4si_maskz,
959         __builtin_ia32_vpshldv_v4di, __builtin_ia32_vpshldv_v4di_mask,
960         __builtin_ia32_vpshldv_v4di_maskz, __builtin_ia32_vpshldv_v2di,
961         __builtin_ia32_vpshldv_v2di_mask, __builtin_ia32_vpshldv_v2di_maskz,
962         __builtin_ia32_vpopcountb_v32qi, __builtin_ia32_vpopcountb_v16qi,
963         __builtin_ia32_vpopcountb_v16qi_mask, __builtin_ia32_vpopcountw_v16hi,
964         __builtin_ia32_vpopcountw_v16hi_mask, __builtin_ia32_vpopcountw_v8hi,
965         __builtin_ia32_vpopcountw_v8hi_mask): Add
966         " | OPTION_MASK_ISA_AVX512VL".
967         * config/i386/avx512vbmi2intrin.h (_mm512_shrdi_epi16,
968         _mm512_shrdi_epi32, _mm512_mask_shrdi_epi32, _mm512_maskz_shrdi_epi32,
969         _mm512_shrdi_epi64, _mm512_mask_shrdi_epi64, _mm512_maskz_shrdi_epi64,
970         _mm512_shldi_epi16, _mm512_shldi_epi32, _mm512_mask_shldi_epi32,
971         _mm512_maskz_shldi_epi32, _mm512_shldi_epi64, _mm512_mask_shldi_epi64,
972         _mm512_maskz_shldi_epi64, _mm512_shrdv_epi16, _mm512_shrdv_epi32,
973         _mm512_mask_shrdv_epi32, _mm512_maskz_shrdv_epi32, _mm512_shrdv_epi64,
974         _mm512_mask_shrdv_epi64, _mm512_maskz_shrdv_epi64, _mm512_shldv_epi16,
975         _mm512_shldv_epi32, _mm512_mask_shldv_epi32, _mm512_maskz_shldv_epi32,
976         _mm512_shldv_epi64, _mm512_mask_shldv_epi64,
977         _mm512_maskz_shldv_epi64): Don't require avx512bw for these intrinsics.
978         * config/i386/avx512bitalgintrin.h (_mm_bitshuffle_epi64_mask,
979         _mm_mask_bitshuffle_epi64_mask): Likewise.
980         * common/config/i386/i386-common.c
981         (OPTION_MASK_ISA_AVX512VPOPCNTDQ_SET,
982         OPTION_MASK_ISA_AVX512BITALG_SET): Or in OPTION_MASK_ISA_AVX512F_SET.
983         (OPTION_MASK_ISA_AVX512F_UNSET): Or in
984         OPTION_MASK_ISA_AVX512VPOPCNTDQ_UNSET and
985         OPTION_MASK_ISA_AVX512BITALG_UNSET.
986         (OPTION_MASK_ISA2_AVX512F_UNSET,
987         OPTION_MASK_ISA2_GENERAL_REGS_ONLY_UNSET): Define.
988         (ix86_handle_option): For -mno-general-regs-only, clear from
989         ix86_isa_flags2 OPTION_MASK_ISA2_GENERAL_REGS_ONLY_UNSET rather than
990         just OPTION_MASK_ISA_MPX.  For -mno-sse{,2,3,4,4.1,4.2,avx,avx2} and
991         -mno-ssse3 clear OPTION_MASK_ISA2_AVX512F_UNSET bits from
992         ix86_isa_flags2.  For -mno-avx512f likewise, instead of masking
993         individually listed ISAs.  For -m{,no-}avx512{vpopcntdq,bitalg} adjust
994         for moving from ix86_isa_flags2 to ix86_isa_flags.
996 2017-12-23  Jakub Jelinek  <jakub@redhat.com>
998         PR c++/83553
999         * fold-const.c (struct contains_label_data): New type.
1000         (contains_label_1): Return non-NULL even for CASE_LABEL_EXPR, unless
1001         inside of a SWITCH_BODY seen during the walk.
1002         (contains_label_p): Use walk_tree instead of
1003         walk_tree_without_duplicates, prepare data for contains_label_1 and
1004         provide own pset.
1006 2017-12-22  Jakub Jelinek  <jakub@redhat.com>
1008         PR debug/83547
1009         * tree-iterator.c (alloc_stmt_list): Start with cleared
1010         TREE_SIDE_EFFECTS regardless whether a new STATEMENT_LIST is allocated
1011         or old one reused.
1013         PR target/83488
1014         * config/i386/avx512vnniintrin.h: Don't check for __AVX512F__ nor
1015         enable avx512f explicitly in #pragma GCC target.
1016         * config/i386/i386-builtin.def (__builtin_ia32_vpdpbusd_v8si,
1017         __builtin_ia32_vpdpbusd_v8si_mask, __builtin_ia32_vpdpbusd_v8si_maskz,
1018         __builtin_ia32_vpdpbusd_v4si, __builtin_ia32_vpdpbusd_v4si_mask,
1019         __builtin_ia32_vpdpbusd_v4si_maskz, __builtin_ia32_vpdpbusds_v8si,
1020         __builtin_ia32_vpdpbusds_v8si_mask,
1021         __builtin_ia32_vpdpbusds_v8si_maskz, __builtin_ia32_vpdpbusds_v4si,
1022         __builtin_ia32_vpdpbusds_v4si_mask,
1023         __builtin_ia32_vpdpbusds_v4si_maskz, __builtin_ia32_vpdpwssd_v8si,
1024         __builtin_ia32_vpdpwssd_v8si_mask, __builtin_ia32_vpdpwssd_v8si_maskz,
1025         __builtin_ia32_vpdpwssd_v4si, __builtin_ia32_vpdpwssd_v4si_mask,
1026         __builtin_ia32_vpdpwssd_v4si_maskz, __builtin_ia32_vpdpwssds_v8si,
1027         __builtin_ia32_vpdpwssds_v8si_mask,
1028         __builtin_ia32_vpdpwssds_v8si_maskz, __builtin_ia32_vpdpwssds_v4si,
1029         __builtin_ia32_vpdpwssds_v4si_mask,
1030         __builtin_ia32_vpdpwssds_v4si_maskz): Use
1031         OPTION_MASK_ISA_AVX512VNNI | OPTION_MASK_ISA_AVX512VL instead of
1032         just OPTION_MASK_ISA_AVX512VNNI.
1034 2017-12-22  Martin Jambor  <mjambor@suse.cz>
1036         PR lto/82027
1037         * lto-cgraph.c (output_cgraph_opt_summary_p): Also check former
1038         clones.
1040 2017-12-22  Julia Koval  <julia.koval@intel.com>
1041             Sebastian Peryt  <sebastian.peryt@intel.com>
1043         * common/config/i386/i386-common.c (OPTION_MASK_ISA_AVX512BITALG_SET,
1044         OPTION_MASK_ISA_AVX512BITALG_UNSET): New.
1045         (ix86_handle_option): Handle -mavx512bitalg, fix 4VNNIW formatting.
1046         * config.gcc: Add avx512vpopcntdqvlintrin.h and avx512bitalgintrin.h.
1047         * config/i386/avx512bitalgintrin.h (_mm512_popcnt_epi8, _mm512_popcnt_epi16,
1048         _mm512_mask_popcnt_epi8, _mm512_maskz_popcnt_epi8, _mm512_mask_popcnt_epi16,
1049         _mm512_maskz_popcnt_epi16, _mm512_bitshuffle_epi64_mask, _mm256_popcnt_epi8,
1050         _mm512_mask_bitshuffle_epi64_mask, _mm256_mask_popcnt_epi8, _mm_popcnt_epi8,
1051         _mm256_maskz_popcnt_epi8, _mm_bitshuffle_epi64_mask, _mm256_popcnt_epi16,
1052         _mm_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
1053         _mm256_mask_bitshuffle_epi64_mask, _mm_popcnt_epi16, _mm_maskz_popcnt_epi8,
1054         _mm256_mask_popcnt_epi16, _mm256_maskz_popcnt_epi16, _mm_mask_popcnt_epi8,
1055         _mm_mask_popcnt_epi16, _mm_maskz_popcnt_epi16): New intrinsics.
1056         * config/i386/avx512vpopcntdqvlintrin.h (_mm_popcnt_epi32, _mm_popcnt_epi64,
1057         _mm_mask_popcnt_epi32, _mm_maskz_popcnt_epi32, _mm256_popcnt_epi32,
1058         _mm256_mask_popcnt_epi32, _mm256_maskz_popcnt_epi32, _mm_mask_popcnt_epi64,
1059         _mm_maskz_popcnt_epi64, _mm256_popcnt_epi64, _mm256_mask_popcnt_epi64,
1060         _mm256_maskz_popcnt_epi64): New intrinsics.
1061         * config/i386/cpuid.h (bit_AVX512BITALG): New bit.
1062         * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mavx512bitalg.
1063         * config/i386/i386-builtin-types.def (V64QI_FTYPE_V64QI, V64QI_FTYPE_V64QI,
1064         V4DI_FTYPE_V4DI, UHI_FTYPE_V2DI_V2DI_UHI, USI_FTYPE_V4DI_V4DI_USI,
1065         V4SI_FTYPE_V4SI_V4SI_UHI, V8SI_FTYPE_V8SI_V8SI_UHI): New types.
1066         * config/i386/i386-builtin.def (__builtin_ia32_vpopcountq_v4di,
1067         __builtin_ia32_vpopcountq_v4di_mask, __builtin_ia32_vpopcountq_v2di,
1068         __builtin_ia32_vpopcountq_v2di_mask, __builtin_ia32_vpopcountd_v4si,
1069         __builtin_ia32_vpopcountd_v4si_mask, __builtin_ia32_vpopcountd_v8si,
1070         __builtin_ia32_vpopcountd_v8si_mask, __builtin_ia32_vpopcountb_v64qi,
1071         __builtin_ia32_vpopcountb_v64qi_mask, __builtin_ia32_vpopcountb_v32qi,
1072         __builtin_ia32_vpopcountb_v32qi_mask, __builtin_ia32_vpopcountb_v16qi,
1073         __builtin_ia32_vpopcountb_v16qi_mask, __builtin_ia32_vpopcountw_v32hi,
1074         __builtin_ia32_vpopcountw_v32hi_mask, __builtin_ia32_vpopcountw_v16hi,
1075         __builtin_ia32_vpopcountw_v16hi_mask, __builtin_ia32_vpopcountw_v8hi,
1076         __builtin_ia32_vpopcountw_v8hi_mask, __builtin_ia32_vpshufbitqmb128_mask,
1077         __builtin_ia32_vpshufbitqmb256_mask,
1078         __builtin_ia32_vpshufbitqmb512_mask): New builtins.
1079         * config/i386/i386-c.c (__AVX512BITALG__): New.
1080         * config/i386/i386.c (isa2_opts): Add -mavx512bitalg.
1081         (ix86_valid_target_attribute_inner_p): Ditto.
1082         (ix86_expand_args_builtin): Handle new types.
1083         * config/i386/i386.h (TARGET_AVX512BITALG, TARGET_AVX512BITALG_P): New.
1084         * config/i386/i386.opt: Add -mavx512bitalg.
1085         * config/i386/immintrin.h: Add avx512vpopcntdqvlintrin.h and
1086         avx512bitalgintrin.h.
1087         * config/i386/sse.md (VI48_AVX512VLBW): New iterator.
1088         (vpopcount<mode><mask_name>): Add more types.
1089         (avx512vl_vpshufbitqmb<mode><mask_scalar_merge_name>): New.
1090         * doc/invoke.texi: Add -mavx512bitalg and -mavx512vpopcntdq.
1092 2017-12-22  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
1094         * common/config/i386/i386-common.c (OPTION_MASK_ISA_AVX512VNNI_SET):
1095         Or in OPTION_MASK_ISA_AVX512F_SET.
1096         (OPTION_MASK_ISA_AVX512F_UNSET): Or in
1097         OPTION_MASK_ISA_AVX512VNNI_UNSET.
1098         (ix86_handle_option): Adjust for
1099         OPTION_MASK_ISA_AVX512VNNI_*SET being in ix86_isa_flags.
1100         * config/i386/i386-builtin.def: Move VNNI builtins from ARGS2
1101         section to ARGS.
1102         * config/i386/i386-c.c: Check for OPTION_MASK_ISA_AVX512VNNI in
1103         isa_flag instead of isa_flag2.
1104         * config/i386/i386.c (ix86_target_string): Move -mavx512vnni from
1105         isa_opts2 to isa_opts.
1106         * config/i386/i386.opt (mavx512vnni): Move from ix86_isa_flags2
1107         to ix86_isa_flags.
1109 2017-12-22  Mike Stump  <mikestump@comcast.net>
1110             Eric Botcazou  <ebotcazou@adacore.com>
1112         * doc/extend.texi (Loop-Specific Pragmas): Document pragma GCC unroll.
1114 2017-12-21  Jakub Jelinek  <jakub@redhat.com>
1116         PR middle-end/83487
1117         * config/i386/i386.c (ix86_function_arg_boundary): Return
1118         PARM_BOUNDARY for TYPE_EMPTY_P types.
1120         PR c/83448
1121         * gimple-ssa-sprintf.c (maybe_warn): Don't call set_caret_index
1122         if navail is >= dir.len.
1124 2017-12-21  Steve Ellcey  <sellcey@cavium.com>
1126         * config/aarch64/t-aarch64-linux (MULTILIB_OSDIRNAMES): Fix
1127         triplet for ilp32.
1129 2017-12-21  Jakub Jelinek  <jakub@redhat.com>
1131         PR rtl-optimization/80747
1132         PR rtl-optimization/83512
1133         * cfgrtl.c (force_nonfallthru_and_redirect): When splitting
1134         succ edge from ENTRY, copy partition from e->dest to the newly
1135         created bb.
1136         * bb-reorder.c (reorder_basic_blocks_simple): If last_tail is
1137         ENTRY, use BB_PARTITION of its successor block as current_partition.
1138         Don't copy partition when splitting succ edge from ENTRY.
1140         PR tree-optimization/83523
1141         * tree-ssa-math-opts.c (is_widening_mult_p): Return false if
1142         for INTEGER_TYPE TYPE_OVERFLOW_TRAPS.
1143         (convert_mult_to_fma): Likewise.
1145         PR tree-optimization/83521
1146         * tree-ssa-phiopt.c (factor_out_conditional_conversion): Use
1147         gimple_build_assign without code on result of
1148         fold_build1 (VIEW_CONVERT_EXPR, ...), as it might not create
1149         a VIEW_CONVERT_EXPR.
1151 2017-12-21  Andrew Pinski  <apinski@cavium.com>
1152             Steve Ellcey  <sellcey@cavium.com>
1154         * config/aarch64/t-aarch64-linux (MULTILIB_OSDIRNAMES): Handle
1155         multi-arch for ilp32.
1157 2017-12-21  Uros Bizjak  <ubizjak@gmail.com>
1159         PR target/83467
1160         * config/i386/i386.md (*ashl<mode>3_mask): Add operand
1161         constraints to operand 2.
1162         (*ashl<mode>3_mask_1): Ditto.
1163         (*<shift_insn><mode>3_mask): Ditto.
1164         (*<shift_insn><mode>3_mask_1): Ditto.
1165         (*<rotate_insn><mode>3_mask): Ditto.
1166         (*<rotate_insn><mode>3_mask_1): Ditto.
1168 2017-12-21  Alexandre Oliva  <aoliva@redhat.com>
1170         * reorg.c (make_return_insns): Reemit each insn with its own location.
1172 2017-12-21  James Greenhalgh  <james.greenhalgh@arm.com>
1174         * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
1175         generation for cases where splatting a value is not useful.
1176         * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
1177         across a vec_duplicate and a paradoxical subreg forming a vector
1178         mode to a vec_concat.
1180 2017-12-21  James Greenhalgh  <james.greenhalgh@arm.com>
1182         * combine.c (simplify_set): Do not transform subregs to zero_extends
1183         if the destination is not a scalar int mode.
1185 2017-12-21  Eric Botcazou  <ebotcazou@adacore.com>
1187         PR c++/82872
1188         * convert.c (convert_to_integer_1) <POINTER_TYPE>: Do not return the
1189         shared zero if the input has overflowed.
1191 2017-12-21  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
1193         * config/arm/driver-arm.c (arm_cpu_table): Specify dotprod
1194         support for Cortex-A55 and Cortex-A75.
1196 2017-12-21  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
1198         * common/config/arm/arm-common.c (compare_opt_names): Add function
1199         comment.  Use strcmp instead of manual loop.
1201 2017-12-21  Martin Liska  <mliska@suse.cz>
1203         PR gcov-profile/83509
1204         * gcov-dump.c (dump_gcov_file): Do not read info about
1205         support_unexecuted_blocks for gcda files.
1207 2017-12-21  Jakub Jelinek  <jakub@redhat.com>
1209         PR rtl-optimization/82973
1210         * emit-rtl.h (valid_for_const_vec_duplicate_p): Rename to ...
1211         (valid_for_const_vector_p): ... this.
1212         * emit-rtl.c (valid_for_const_vec_duplicate_p): Rename to ...
1213         (valid_for_const_vector_p): ... this.  Adjust function comment.
1214         (gen_vec_duplicate): Adjust caller.
1215         * optabs.c (expand_vector_broadcast): Likewise.
1216         * simplify-rtx.c (simplify_const_unary_operation): Don't optimize into
1217         CONST_VECTOR if some element isn't simplified valid_for_const_vector_p
1218         constant.
1219         (simplify_const_binary_operation): Likewise.  Use CONST_FIXED_P macro
1220         instead of GET_CODE == CONST_FIXED.
1221         (simplify_subreg): Use CONST_FIXED_P macro instead of
1222         GET_CODE == CONST_FIXED.
1224         PR target/83488
1225         * config/i386/i386.c (ix86_target_string): Move -mavx512vbmi2 and
1226         -mshstk entries from isa_opts2 to isa_opts and -mhle, -mmovbe,
1227         -mclzero and -mmwaitx entries from isa_opts to isa_opts2.
1228         (ix86_option_override_internal): Adjust for
1229         OPTION_MASK_ISA_{HLE,MOVBE,CLZERO,MWAITX} moving to ix86_isa_flags2
1230         and OPTION_MASK_ISA_SHSTK moving to ix86_isa_flags.
1231         (BDESC_VERIFYS): Remove SPECIAL_ARGS2 related checks.
1232         (ix86_init_mmx_sse_builtins): Remove bdesc_special_args2 handling.
1233         Use def_builtin2 instead of def_builtin for OPTION_MASK_ISA_MWAITX
1234         and OPTION_MASK_ISA_CLZERO builtins.  Use def_builtin instead of
1235         def_builtin2 for CET builtins.
1236         (ix86_expand_builtin): Remove bdesc_special_args2 handling.  Fix
1237         up formatting in IX86_BUILTIN_RDPID code.
1238         * config/i386/i386-builtin.def: Move VBMI2 builtins from SPECIAL_ARGS2
1239         section to SPECIAL_ARGS and from ARGS2 section to ARGS.
1240         * config/i386/i386.opt (mavx512vbmi2, mshstk): Move from
1241         ix86_isa_flags2 to ix86_isa_flags.
1242         (mhle, mmovbe, mclzero, mmwaitx): Move from ix86_isa_flags to
1243         ix86_isa_flags2.
1244         * config/i386/i386-c.c (ix86_target_macros_internal): Check for
1245         OPTION_MASK_ISA_{CLZERO,MWAITX} in isa_flag2 instead of isa_flag.
1246         Check for OPTION_MASK_ISA_{SHSTK,AVX512VBMI2} in isa_flag instead
1247         of isa_flag2.
1248         * common/config/i386/i386-common.c (OPTION_MASK_ISA_AVX512VBMI2_SET):
1249         Or in OPTION_MASK_ISA_AVX512F_SET.
1250         (OPTION_MASK_ISA_AVX512F_UNSET): Or in
1251         OPTION_MASK_ISA_AVX512VBMI2_UNSET.
1252         (ix86_handle_option): Adjust for
1253         OPTION_MASK_ISA_{SHSTK,AVX512VBMI2}_*SET being in ix86_isa_flags
1254         and OPTION_MASK_ISA_{MOVBE,MWAITX,CLZERO}_*SET in ix86_isa_flags2.
1256 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1257             Alan Hayward  <alan.hayward@arm.com>
1258             David Sherwood  <david.sherwood@arm.com>
1260         * tree-data-ref.h (prune_runtime_alias_test_list): Take the
1261         factor as a poly_uint64 rather than an unsigned HOST_WIDE_INT.
1262         * tree-data-ref.c (prune_runtime_alias_test_list): Likewise.
1263         Track polynomial offsets.
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-vect-data-refs.c (vect_compute_data_ref_alignment):
1270         Treat drb->init as a poly_int.  Fail if its misalignment wrt
1271         vector_alignment isn't known.
1273 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1274             Alan Hayward  <alan.hayward@arm.com>
1275             David Sherwood  <david.sherwood@arm.com>
1277         * tree-vectorizer.h (_loop_vec_info): Add a versioning_threshold
1278         field.
1279         (LOOP_VINFO_VERSIONING_THRESHOLD): New macro
1280         (vect_loop_versioning): Take the loop versioning threshold as a
1281         separate parameter.
1282         * tree-vect-loop-manip.c (vect_loop_versioning): Likewise.
1283         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
1284         versioning_threshold.
1285         (vect_analyze_loop_2): Compute the loop versioning threshold
1286         whenever loop versioning is needed, and store it in the new
1287         field rather than combining it with the cost model threshold.
1288         (vect_transform_loop): Update call to vect_loop_versioning.
1289         Try to combine the loop versioning and cost thresholds here.
1291 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1292             Alan Hayward  <alan.hayward@arm.com>
1293             David Sherwood  <david.sherwood@arm.com>
1295         * tree-ssa-loop-ivopts.h (strip_offset): Return the offset as
1296         poly_uint64_pod rather than an unsigned HOST_WIDE_INT.
1297         * tree-loop-distribution.c (classify_builtin_st): Update accordingly.
1298         * tree-ssa-loop-ivopts.c (iv_use::addr_offset): Change from
1299         an unsigned HOST_WIDE_INT to a poly_uint64_pod.
1300         (group_compare_offset): Update accordingly.
1301         (split_small_address_groups_p): Likewise.
1302         (record_use): Take addr_offset as a poly_uint64 rather than
1303         an unsigned HOST_WIDE_INT.
1304         (strip_offset): Return the offset as a poly_uint64 rather than
1305         an unsigned HOST_WIDE_INT.
1306         (record_group_use, split_address_groups): Track polynomial offsets.
1307         (add_iv_candidate_for_use): Likewise.
1308         (addr_offset_valid_p): Take the offset as a poly_int64 rather
1309         than a HOST_WIDE_INT.
1310         (strip_offset_1): Return the offset as a poly_int64 rather than
1311         a HOST_WIDE_INT.
1313 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1314             Alan Hayward  <alan.hayward@arm.com>
1315             David Sherwood  <david.sherwood@arm.com>
1317         * tree.h (get_binfo_at_offset): Take the offset as a poly_int64
1318         rather than a HOST_WIDE_INT.
1319         * tree.c (get_binfo_at_offset): Likewise.
1321 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1322             Alan Hayward  <alan.hayward@arm.com>
1323             David Sherwood  <david.sherwood@arm.com>
1325         * ipa-prop.h (build_ref_for_offset): Take the offset as a poly_int64
1326         rather than a HOST_WIDE_INT.
1327         * tree-sra.c (build_ref_for_offset): Likewise.
1329 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1330             Alan Hayward  <alan.hayward@arm.com>
1331             David Sherwood  <david.sherwood@arm.com>
1333         * fold-const.h (mem_ref_offset): Return a poly_offset_int rather
1334         than an offset_int.
1335         * tree.c (mem_ref_offset): Likewise.
1336         (build_simple_mem_ref_loc): Treat MEM_REF offsets as poly_ints.
1337         * builtins.c (get_object_alignment_2): Likewise.
1338         * expr.c (get_inner_reference, expand_expr_real_1): Likewise.
1339         * gimple-fold.c (get_base_constructor): Likewise.
1340         * gimple-ssa-strength-reduction.c (restructure_reference): Likewise.
1341         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref):
1342         Likewise.
1343         * ipa-polymorphic-call.c
1344         (ipa_polymorphic_call_context::ipa_polymorphic_call_context): Likewise.
1345         * ipa-prop.c (compute_complex_assign_jump_func): Likewise.
1346         (get_ancestor_addr_info): Likewise.
1347         * ipa-param-manipulation.c (ipa_get_adjustment_candidate): Likewise.
1348         * match.pd: Likewise.
1349         * tree-data-ref.c (dr_analyze_innermost): Likewise.
1350         * tree-dfa.c (get_addr_base_and_unit_offset_1): Likewise.
1351         * tree-eh.c (tree_could_trap_p): Likewise.
1352         * tree-object-size.c (addr_object_size): Likewise.
1353         * tree-ssa-address.c (copy_ref_info): Likewise.
1354         * tree-ssa-alias.c (indirect_ref_may_alias_decl_p): Likewise.
1355         (indirect_refs_may_alias_p): Likewise.
1356         * tree-ssa-sccvn.c (copy_reference_ops_from_ref): Likewise.
1357         * tree-ssa.c (maybe_rewrite_mem_ref_base): Likewise.
1358         (non_rewritable_mem_ref_base): Likewise.
1359         * tree-vect-data-refs.c (vect_check_gather_scatter): Likewise.
1360         * tree-vrp.c (vrp_prop::check_array_ref): Likewise.
1361         * varasm.c (decode_addr_const): Likewise.
1363 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1364             Alan Hayward  <alan.hayward@arm.com>
1365             David Sherwood  <david.sherwood@arm.com>
1367         * gimple-ssa-stor-merging.c (find_bswap_or_nop_load): Track polynomial
1368         offsets for MEM_REFs.
1370 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1371             Alan Hayward  <alan.hayward@arm.com>
1372             David Sherwood  <david.sherwood@arm.com>
1374         * tree-ssanames.h (adjust_ptr_info_misalignment): Take the increment
1375         as a poly_uint64 rather than an unsigned int.
1376         * tree-ssanames.c (adjust_ptr_info_misalignment): Likewise.
1378 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1379             Alan Hayward  <alan.hayward@arm.com>
1380             David Sherwood  <david.sherwood@arm.com>
1382         * varasm.c (addr_const::offset): Change from HOST_WIDE_INT
1383         to poly_int64.
1384         (decode_addr_const): Update accordingly.
1386 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1387             Alan Hayward  <alan.hayward@arm.com>
1388             David Sherwood  <david.sherwood@arm.com>
1390         * tree.h (bit_field_size, bit_field_offset): New functions.
1391         * hsa-gen.c (gen_hsa_addr): Use them.
1392         * tree-ssa-forwprop.c (simplify_bitfield_ref): Likewise.
1393         (simplify_vector_constructor): Likewise.
1394         * tree-ssa-sccvn.c (copy_reference_ops_from_ref): Likewise.
1395         * tree-cfg.c (verify_expr): Require the sizes and offsets of a
1396         BIT_FIELD_REF to be poly_uint64s rather than uhwis.
1397         * fold-const.c (fold_ternary_loc): Protect tree_to_uhwi with
1398         tree_fits_uhwi_p.
1400 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1401             Alan Hayward  <alan.hayward@arm.com>
1402             David Sherwood  <david.sherwood@arm.com>
1404         * expr.h (emit_group_load, emit_group_load_into_temps)
1405         (emit_group_store): Take the size as a poly_int64 rather than an int.
1406         * expr.c (emit_group_load_1, emit_group_load): Likewise.
1407         (emit_group_load_into_temp, emit_group_store): Likewise.
1409 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1410             Alan Hayward  <alan.hayward@arm.com>
1411             David Sherwood  <david.sherwood@arm.com>
1413         * ira-int.h (ira_spilled_reg_stack_slot::width): Change from
1414         an unsigned int to a poly_uint64.
1415         * ira.h (ira_reuse_stack_slot, ira_mark_new_stack_slot): Take the
1416         sizes as poly_uint64s rather than unsigned ints.
1417         * ira-color.c (ira_reuse_stack_slot, ira_mark_new_stack_slot):
1418         Likewise.
1420 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1421             Alan Hayward  <alan.hayward@arm.com>
1422             David Sherwood  <david.sherwood@arm.com>
1424         * lra-constraints.c (emit_inc): Change inc_amount from an int
1425         to a poly_int64.
1427 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1428             Alan Hayward  <alan.hayward@arm.com>
1429             David Sherwood  <david.sherwood@arm.com>
1431         * cfgexpand.c (stack_var::size): Change from a HOST_WIDE_INT
1432         to a poly_uint64.
1433         (add_stack_var, stack_var_cmp, partition_stack_vars)
1434         (dump_stack_var_partition): Update accordingly.
1435         (alloc_stack_frame_space): Take the size as a poly_int64 rather
1436         than a HOST_WIDE_INT.
1437         (expand_stack_vars, expand_one_stack_var_1): Handle polynomial sizes.
1438         (defer_stack_allocation, estimated_stack_frame_size): Likewise.
1439         (account_stack_vars, expand_one_var): Likewise.  Return a poly_uint64
1440         rather than a HOST_WIDE_INT.
1442 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1443             Alan Hayward  <alan.hayward@arm.com>
1444             David Sherwood  <david.sherwood@arm.com>
1446         * target.def (return_pops_args): Treat both the input and output
1447         sizes as poly_int64s rather than HOST_WIDE_INTS.
1448         * targhooks.h (default_return_pops_args): Update accordingly.
1449         * targhooks.c (default_return_pops_args): Likewise.
1450         * doc/tm.texi: Regenerate.
1451         * emit-rtl.h (incoming_args): Change pops_args, size and
1452         outgoing_args_size from int to poly_int64_pod.
1453         * function.h (expr_status): Change x_pending_stack_adjust and
1454         x_stack_pointer_delta from int to poly_int64.
1455         (args_size::constant): Change from HOST_WIDE_INT to poly_int64.
1456         (ARGS_SIZE_RTX): Update accordingly.
1457         * calls.c (highest_outgoing_arg_in_use): Change from int to
1458         unsigned int.
1459         (stack_usage_watermark, stored_args_watermark): New variables.
1460         (stack_region_maybe_used_p, mark_stack_region_used): New functions.
1461         (emit_call_1): Change the stack_size and rounded_stack_size
1462         parameters from HOST_WIDE_INT to poly_int64.  Track n_popped
1463         as a poly_int64.
1464         (save_fixed_argument_area): Check stack_usage_watermark.
1465         (initialize_argument_information): Change old_pending_adj from
1466         a HOST_WIDE_INT * to a poly_int64_pod *.
1467         (compute_argument_block_size): Return the size as a poly_int64
1468         rather than an int.
1469         (finalize_must_preallocate): Track polynomial argument sizes.
1470         (compute_argument_addresses): Likewise.
1471         (internal_arg_pointer_based_exp): Track polynomial offsets.
1472         (mem_overlaps_already_clobbered_arg_p): Rename to...
1473         (mem_might_overlap_already_clobbered_arg_p): ...this and take the
1474         size as a poly_uint64 rather than an unsigned HOST_WIDE_INT.
1475         Check stored_args_used_watermark.
1476         (load_register_parameters): Update accordingly.
1477         (check_sibcall_argument_overlap_1): Likewise.
1478         (combine_pending_stack_adjustment_and_call): Take the unadjusted
1479         args size as a poly_int64 rather than an int.  Return a bool
1480         indicating whether the optimization was possible and return
1481         the new adjustment by reference.
1482         (check_sibcall_argument_overlap): Track polynomail argument sizes.
1483         Update stored_args_watermark.
1484         (can_implement_as_sibling_call_p): Handle polynomial argument sizes.
1485         (expand_call): Likewise.  Maintain stack_usage_watermark and
1486         stored_args_watermark.  Update calls to
1487         combine_pending_stack_adjustment_and_call.
1488         (emit_library_call_value_1): Handle polynomial argument sizes.
1489         Call stack_region_maybe_used_p and mark_stack_region_used.
1490         Maintain stack_usage_watermark.
1491         (store_one_arg): Likewise.  Update call to
1492         mem_overlaps_already_clobbered_arg_p.
1493         * config/arm/arm.c (arm_output_function_prologue): Add a cast to
1494         HOST_WIDE_INT.
1495         * config/avr/avr.c (avr_outgoing_args_size): Likewise.
1496         * config/microblaze/microblaze.c (microblaze_function_prologue):
1497         Likewise.
1498         * config/cr16/cr16.c (cr16_return_pops_args): Update for new
1499         TARGET_RETURN_POPS_ARGS interface.
1500         (cr16_compute_frame, cr16_initial_elimination_offset): Add casts
1501         to HOST_WIDE_INT.
1502         * config/ft32/ft32.c (ft32_compute_frame): Likewise.
1503         * config/i386/i386.c (ix86_return_pops_args): Update for new
1504         TARGET_RETURN_POPS_ARGS interface.
1505         (ix86_expand_split_stack_prologue): Add a cast to HOST_WIDE_INT.
1506         * config/moxie/moxie.c (moxie_compute_frame): Likewise.
1507         * config/m68k/m68k.c (m68k_return_pops_args): Update for new
1508         TARGET_RETURN_POPS_ARGS interface.
1509         * config/vax/vax.c (vax_return_pops_args): Likewise.
1510         * config/pa/pa.h (STACK_POINTER_OFFSET): Add a cast to poly_int64.
1511         (EXIT_IGNORE_STACK): Update reference to crtl->outgoing_args_size.
1512         * config/arm/arm.h (CALLER_INTERWORKING_SLOT_SIZE): Likewise.
1513         * config/powerpcspe/aix.h (STACK_DYNAMIC_OFFSET): Likewise.
1514         * config/powerpcspe/darwin.h (STACK_DYNAMIC_OFFSET): Likewise.
1515         * config/powerpcspe/powerpcspe.h (STACK_DYNAMIC_OFFSET): Likewise.
1516         * config/rs6000/aix.h (STACK_DYNAMIC_OFFSET): Likewise.
1517         * config/rs6000/darwin.h (STACK_DYNAMIC_OFFSET): Likewise.
1518         * config/rs6000/rs6000.h (STACK_DYNAMIC_OFFSET): Likewise.
1519         * dojump.h (saved_pending_stack_adjust): Change x_pending_stack_adjust
1520         and x_stack_pointer_delta from int to poly_int64.
1521         * dojump.c (do_pending_stack_adjust): Update accordingly.
1522         * explow.c (allocate_dynamic_stack_space): Handle polynomial
1523         stack_pointer_deltas.
1524         * function.c (STACK_DYNAMIC_OFFSET): Add a cast to poly_int64.
1525         (pad_to_arg_alignment): Track polynomial offsets.
1526         (assign_parm_find_stack_rtl): Likewise.
1527         (assign_parms, locate_and_pad_parm): Handle polynomial argument sizes.
1528         * toplev.c (output_stack_usage): Update reference to
1529         current_function_pushed_stack_size.
1531 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1532             Alan Hayward  <alan.hayward@arm.com>
1533             David Sherwood  <david.sherwood@arm.com>
1535         * function.c (in_arg_offset, var_offset, dynamic_offset)
1536         (out_arg_offset, cfa_offset): Change from int to poly_int64.
1537         (instantiate_new_reg): Return the new offset as a poly_int64_pod
1538         rather than a HOST_WIDE_INT.
1539         (instantiate_virtual_regs_in_rtx): Track polynomial offsets.
1540         (instantiate_virtual_regs_in_insn): Likewise.
1542 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1543             Alan Hayward  <alan.hayward@arm.com>
1544             David Sherwood  <david.sherwood@arm.com>
1546         * rtl.h (get_args_size, add_args_size_note): New functions.
1547         (find_args_size_adjust): Return a poly_int64 rather than a
1548         HOST_WIDE_INT.
1549         (fixup_args_size_notes): Likewise.  Make the same change to the
1550         end_args_size parameter.
1551         * rtlanal.c (get_args_size, add_args_size_note): New functions.
1552         * builtins.c (expand_builtin_trap): Use add_args_size_note.
1553         * calls.c (emit_call_1): Likewise.
1554         * explow.c (adjust_stack_1): Likewise.
1555         * cfgcleanup.c (old_insns_match_p): Update use of
1556         find_args_size_adjust.
1557         * combine.c (distribute_notes): Track polynomial arg sizes.
1558         * dwarf2cfi.c (dw_trace_info): Change beg_true_args_size,
1559         end_true_args_size, beg_delay_args_size and end_delay_args_size
1560         from HOST_WIDE_INT to poly_int64.
1561         (add_cfi_args_size): Take the args_size as a poly_int64 rather
1562         than a HOST_WIDE_INT.
1563         (notice_args_size, notice_eh_throw, maybe_record_trace_start)
1564         (maybe_record_trace_start_abnormal, scan_trace, connect_traces): Track
1565         polynomial arg sizes.
1566         * emit-rtl.c (try_split): Use get_args_size.
1567         * recog.c (peep2_attempt): Likewise.
1568         * reload1.c (reload_as_needed): Likewise.
1569         * expr.c (find_args_size_adjust): Return the adjustment as a
1570         poly_int64 rather than a HOST_WIDE_INT.
1571         (fixup_args_size_notes): Change end_args_size from a HOST_WIDE_INT
1572         to a poly_int64 and change the return type in the same way.
1573         (emit_single_push_insn): Track polynomial arg sizes.
1575 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1576             Alan Hayward  <alan.hayward@arm.com>
1577             David Sherwood  <david.sherwood@arm.com>
1579         * expr.h (push_block, emit_push_insn): Change the "extra" parameter
1580         from HOST_WIDE_INT to poly_int64.
1581         * expr.c (push_block, emit_push_insn): Likewise.
1583 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1584             Alan Hayward  <alan.hayward@arm.com>
1585             David Sherwood  <david.sherwood@arm.com>
1587         * function.h (frame_space): Change start and length from HOST_WIDE_INT
1588         to poly_int64.
1589         (get_frame_size): Return the size as a poly_int64 rather than a
1590         HOST_WIDE_INT.
1591         (frame_offset_overflow): Take the offset as a poly_int64 rather
1592         than a HOST_WIDE_INT.
1593         (assign_stack_local_1, assign_stack_local, assign_stack_temp_for_type)
1594         (assign_stack_temp): Likewise for the size.
1595         * function.c (get_frame_size): Return a poly_int64 rather than
1596         a HOST_WIDE_INT.
1597         (frame_offset_overflow): Take the offset as a poly_int64 rather
1598         than a HOST_WIDE_INT.
1599         (try_fit_stack_local): Take the start, length and size as poly_int64s
1600         rather than HOST_WIDE_INTs.  Return the offset as a poly_int64_pod
1601         rather than a HOST_WIDE_INT.
1602         (add_frame_space): Take the start and end as poly_int64s rather than
1603         HOST_WIDE_INTs.
1604         (assign_stack_local_1, assign_stack_local, assign_stack_temp_for_type)
1605         (assign_stack_temp): Likewise for the size.
1606         (temp_slot): Change size, base_offset and full_size from HOST_WIDE_INT
1607         to poly_int64.
1608         (find_temp_slot_from_address): Handle polynomial offsets.
1609         (combine_temp_slots): Likewise.
1610         * emit-rtl.h (rtl_data::x_frame_offset): Change from HOST_WIDE_INT
1611         to poly_int64.
1612         * cfgexpand.c (alloc_stack_frame_space): Return the offset as a
1613         poly_int64 rather than a HOST_WIDE_INT.
1614         (expand_one_stack_var_at): Take the offset as a poly_int64 rather
1615         than a HOST_WIDE_INT.
1616         (expand_stack_vars, expand_one_stack_var_1, expand_used_vars): Handle
1617         polynomial frame offsets.
1618         * config/m32r/m32r-protos.h (m32r_compute_frame_size): Take the size
1619         as a poly_int64 rather than an int.
1620         * config/m32r/m32r.c (m32r_compute_frame_size): Likewise.
1621         * config/v850/v850-protos.h (compute_frame_size): Likewise.
1622         * config/v850/v850.c (compute_frame_size): Likewise.
1623         * config/xtensa/xtensa-protos.h (compute_frame_size): Likewise.
1624         * config/xtensa/xtensa.c (compute_frame_size): Likewise.
1625         * config/pa/pa-protos.h (pa_compute_frame_size): Likewise.
1626         * config/pa/pa.c (pa_compute_frame_size): Likewise.
1627         * explow.h (get_dynamic_stack_base): Take the offset as a poly_int64
1628         rather than a HOST_WIDE_INT.
1629         * explow.c (get_dynamic_stack_base): Likewise.
1630         * final.c (final_start_function): Use the constant lower bound
1631         of the frame size for -Wframe-larger-than.
1632         * ira.c (do_reload): Adjust for new get_frame_size return type.
1633         * lra.c (lra): Likewise.
1634         * reload1.c (reload): Likewise.
1635         * config/avr/avr.c (avr_asm_function_end_prologue): Likewise.
1636         * config/pa/pa.h (EXIT_IGNORE_STACK): Likewise.
1637         * rtlanal.c (get_initial_register_offset): Return the offset as
1638         a poly_int64 rather than a HOST_WIDE_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         * reload1.c (elim_table): Change initial_offset, offset and
1645         previous_offset from HOST_WIDE_INT to poly_int64_pod.
1646         (offsets_at): Change the target array's element type from
1647         HOST_WIDE_INT to poly_int64_pod.
1648         (set_label_offsets, eliminate_regs_1, eliminate_regs_in_insn)
1649         (elimination_costs_in_insn, update_eliminable_offsets)
1650         (verify_initial_elim_offsets, set_offsets_for_label)
1651         (init_eliminable_invariants): Update after above changes.
1653 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1654             Alan Hayward  <alan.hayward@arm.com>
1655             David Sherwood  <david.sherwood@arm.com>
1657         * reload.h (reload::inc): Change from an int to a poly_int64_pod.
1658         * reload.c (combine_reloads, debug_reload_to_stream): Likewise.
1659         (decomposition): Change start and end from HOST_WIDE_INT
1660         to poly_int64_pod.
1661         (decompose, immune_p): Update accordingly.
1662         (find_inc_amount): Return a poly_int64 rather than an int.
1663         * reload1.c (inc_for_reload): Take the inc_amount as a poly_int64
1664         rather than an int.
1666 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1667             Alan Hayward  <alan.hayward@arm.com>
1668             David Sherwood  <david.sherwood@arm.com>
1670         * tree.h (get_inner_reference): Return the bitsize and bitpos
1671         as poly_int64_pods rather than HOST_WIDE_INT.
1672         * fold-const.h (ptr_difference_const): Return the pointer difference
1673         as a poly_int64_pod rather than a HOST_WIDE_INT.
1674         * expr.c (get_inner_reference): Return the bitsize and bitpos
1675         as poly_int64_pods rather than HOST_WIDE_INT.
1676         (expand_expr_addr_expr_1, expand_expr_real_1): Track polynomial
1677         offsets and sizes.
1678         * fold-const.c (make_bit_field_ref): Take the bitpos as a poly_int64
1679         rather than a HOST_WIDE_INT.  Update call to get_inner_reference.
1680         (optimize_bit_field_compare): Update call to get_inner_reference.
1681         (decode_field_reference): Likewise.
1682         (fold_unary_loc): Track polynomial offsets and sizes.
1683         (split_address_to_core_and_offset): Return the bitpos as a
1684         poly_int64_pod rather than a HOST_WIDE_INT.
1685         (ptr_difference_const): Likewise for the pointer difference.
1686         * asan.c (instrument_derefs): Track polynomial offsets and sizes.
1687         * config/mips/mips.c (r10k_safe_mem_expr_p): Likewise.
1688         * dbxout.c (dbxout_expand_expr): Likewise.
1689         * dwarf2out.c (loc_list_for_address_of_addr_expr_of_indirect_ref)
1690         (loc_list_from_tree_1, fortran_common): Likewise.
1691         * gimple-laddress.c (pass_laddress::execute): Likewise.
1692         * gimple-ssa-store-merging.c (find_bswap_or_nop_load): Likewise.
1693         * gimplify.c (gimplify_scan_omp_clauses): Likewise.
1694         * simplify-rtx.c (delegitimize_mem_from_attrs): Likewise.
1695         * tree-affine.c (tree_to_aff_combination): Likewise.
1696         (get_inner_reference_aff): Likewise.
1697         * tree-data-ref.c (split_constant_offset_1): Likewise.
1698         (dr_analyze_innermost): Likewise.
1699         * tree-scalar-evolution.c (interpret_rhs_expr): Likewise.
1700         * tree-sra.c (ipa_sra_check_caller): Likewise.
1701         * tree-vect-data-refs.c (vect_check_gather_scatter): Likewise.
1702         * ubsan.c (maybe_instrument_pointer_overflow): Likewise.
1703         (instrument_bool_enum_load, instrument_object_size): Likewise.
1704         * gimple-ssa-strength-reduction.c (slsr_process_ref): Update call
1705         to get_inner_reference.
1706         * hsa-gen.c (gen_hsa_addr): Likewise.
1707         * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Likewise.
1708         * tsan.c (instrument_expr): Likewise.
1709         * match.pd: Update call to ptr_difference_const.
1711 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1712             Alan Hayward  <alan.hayward@arm.com>
1713             David Sherwood  <david.sherwood@arm.com>
1715         * fold-const.c (fold_comparison): Track sizes and offsets as
1716         poly_int64s rather than HOST_WIDE_INTs when folding address
1717         comparisons.
1719 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1720             Alan Hayward  <alan.hayward@arm.com>
1721             David Sherwood  <david.sherwood@arm.com>
1723         * expr.h (get_bit_range): Return the bitstart and bitend as
1724         poly_uint64s rather than unsigned HOST_WIDE_INTs.  Return the bitpos
1725         as a poly_int64 rather than a HOST_WIDE_INT.
1726         * expr.c (get_bit_range): Likewise.
1727         (expand_assignment): Update call accordingly.
1728         * fold-const.c (optimize_bit_field_compare): Likewise.
1730 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1731             Alan Hayward  <alan.hayward@arm.com>
1732             David Sherwood  <david.sherwood@arm.com>
1734         * config/aarch64/aarch64-protos.h (aarch64_addr_query_type): New enum.
1735         (aarch64_legitimate_address_p): Use it instead of an rtx code,
1736         as an optional final parameter.
1737         * config/aarch64/aarch64.c (aarch64_classify_address): Likewise.
1738         (aarch64_legitimate_address_p): Likewise.
1739         (aarch64_print_address_internal): Take an aarch64_addr_query_type
1740         instead of an rtx code.
1741         (aarch64_address_valid_for_prefetch_p): Update calls accordingly.
1742         (aarch64_legitimate_address_hook_p): Likewise.
1743         (aarch64_print_ldpstp_address): Likewise.
1744         (aarch64_print_operand_address): Likewise.
1745         (aarch64_address_cost): Likewise.
1746         * config/aarch64/constraints.md (Uml, Umq, Ump, Utq): Likewise.
1747         * config/aarch64/predicates.md (aarch64_mem_pair_operand): Likewise.
1748         (aarch64_mem_pair_lanes_operand): Likewise.
1750 2017-12-20  Richard Biener  <rguenther@suse.de>
1752         * tree-ssa-dom.c (dom_opt_dom_walker::optimize_stmt): Call
1753         update_stmt_if_modified.
1755 2017-12-20  Wilco Dijkstra  <wdijkstr@arm.com>
1757         PR tree-optimization/83491
1758         * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Check for SSA_NAME
1759         before walking uses.  Improve coding style and comments.
1761 2017-12-20  Tom de Vries  <tom@codesourcery.com>
1763         * gimple-fold.c (fold_internal_goacc_dim): Simplify.
1765 2017-12-20  Jakub Jelinek  <jakub@redhat.com>
1767         PR ipa/83506
1768         * ipa-fnsummary.c (pass_data_ipa_free_fn_summary): Use 0 for
1769         todo_flags_finish.
1770         (pass_ipa_free_fn_summary): Add small_p private data member,
1771         initialize to false in the ctor.
1772         (pass_ipa_free_fn_summary::clone,
1773         pass_ipa_free_fn_summary::set_pass_param,
1774         pass_ipa_free_fn_summary::gate): New methods.
1775         (pass_ipa_free_fn_summary::execute): Return TODO_remove_functions
1776         | TODO_dump_symtab if small_p.
1777         * passes.def: Add true parm for the existing pass_ipa_free_fn_summary
1778         entry and add another instance of the pass with false parm after
1779         ipa-pure-const.
1780         * ipa-pure-const.c (pass_ipa_pure_const): Don't call
1781         ipa_free_fn_summary here.
1783 2017-12-20  Paolo Carlini  <paolo.carlini@oracle.com>
1785         * gimplify.c (gimplify_return_expr): Remove dead error_mark_node check.
1787 2017-12-20  Martin Sebor  <msebor@redhat.com>
1789         PR testsuite/83131
1790         * builtins.c (expand_builtin_strlen): Use get_callee_fndecl.
1791         (expand_builtin_strcmp): Call maybe_warn_nonstring_arg.
1792         (expand_builtin_strncmp): Same.
1794 2017-12-20  Alexandre Oliva  <aoliva@redhat.com>
1796         PR bootstrap/83396
1797         * cfgexpand.c (label_rtx_for_bb): Revert SFN changes that
1798         allowed debug stmts before labels.
1799         (expand_gimple_basic_block): Likewise.
1800         * gimple-iterator.c (gimple_find_edge_insert_loc): Likewise.
1801         * gimple-iterator.h (gsi_after_labels): Likewise.
1802         * tree-cfgcleanup (remove_forwarder_block): Likewise, but
1803         rename reused variable, and simplify using gsi_move_before.
1804         * tree-ssa-tail-merge.c (find_duplicate): Likewise.
1805         * tree-cfg.c (make_edges, cleanup_dead_labels): Likewise.
1806         (gimple_can_merge_blocks_p, verify_gimple_in_cfg): Likewise.
1807         (gimple_verify_flow_info, gimple_block_label): Likewise.
1808         (make_blocks): Move debug markers after adjacent labels.
1809         * cfgrtl.c (skip_insns_after_block): Revert SFN changes that
1810         allowed debug insns outside blocks.
1811         * df-scan.c (df_insn_delete): Likewise.
1812         * lra-constraints.c (update_ebb_live_info): Likewise.
1813         * var-tracking.c (get_first_insn, vt_emit_notes): Likewise.
1814         (vt_initialize, delete_vta_debug_insns): Likewise.
1815         (reemit_marker_as_note): Drop BB parm.  Adjust callers.
1817 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
1818             Alan Hayward  <alan.hayward@arm.com>
1819             David Sherwood  <david.sherwood@arm.com>
1821         * poly-int-types.h (round_down_to_byte_boundary): New macro.
1822         (round_up_to_byte_boundary): Likewise.
1823         * expr.h (get_bit_range): Add temporary shim.
1824         * gimple-ssa-store-merging.c (store_operand_info): Change the
1825         bitsize, bitpos, bitregion_start and bitregion_end fields from
1826         unsigned HOST_WIDE_INT to poly_uint64.
1827         (merged_store_group): Likewise load_align_base.
1828         (compatible_load_p, compatible_load_p): Update accordingly.
1829         (imm_store_chain_info::coalesce_immediate_stores): Likewise.
1830         (split_group, imm_store_chain_info::output_merged_store): Likewise.
1831         (mem_valid_for_store_merging): Return the bitsize, bitpos,
1832         bitregion_start and bitregion_end as poly_uint64s rather than
1833         unsigned HOST_WIDE_INTs.  Track polynomial offsets internally.
1834         (handled_load): Take the bitsize, bitpos,
1835         bitregion_start and bitregion_end as poly_uint64s rather than
1836         unsigned HOST_WIDE_INTs.
1837         (pass_store_merging::process_store): Update call to
1838         mem_valid_for_store_merging.
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         * builtins.c (get_object_alignment_2): Track polynomial offsets
1845         and sizes.  Update the alignment handling.
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         * tree.h (get_inner_reference): Add a version that returns the
1852         offset and size as poly_int64_pods rather than HOST_WIDE_INTs.
1853         * cfgexpand.c (expand_debug_expr): Track polynomial offsets.  Simply
1854         the case in which bitpos is not associated with the first byte.
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 (get_inner_reference_aff): Return the size as a
1861         poly_widest_int.
1862         * tree-affine.c (get_inner_reference_aff): Likewise.
1863         * tree-data-ref.c (dr_may_alias_p): Update accordingly.
1864         * tree-ssa-loop-im.c (mem_refs_may_alias_p): Likewise.
1866 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
1867             Alan Hayward  <alan.hayward@arm.com>
1868             David Sherwood  <david.sherwood@arm.com>
1870         * fold-const.c (pointer_may_wrap_p): Take the offset as a
1871         HOST_WIDE_INT rather than a poly_int64.
1873 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
1874             Alan Hayward  <alan.hayward@arm.com>
1875             David Sherwood  <david.sherwood@arm.com>
1877         * gimple-ssa-store-merging.c (symbolic_number::bytepos): Change from
1878         HOST_WIDE_INT to poly_int64_pod.
1879         (perform_symbolic_merge): Update accordingly.
1880         (bswap_replace): Likewise.
1882 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
1883             Alan Hayward  <alan.hayward@arm.com>
1884             David Sherwood  <david.sherwood@arm.com>
1886         * tree-affine.h (aff_tree::offset): Change from widest_int
1887         to poly_widest_int.
1888         (wide_int_ext_for_comb): Delete.
1889         (aff_combination_const, aff_comb_cannot_overlap_p): Take the
1890         constants as poly_widest_int rather than widest_int.
1891         (aff_combination_constant_multiple_p): Return the multiplier
1892         as a poly_widest_int.
1893         (aff_combination_zero_p, aff_combination_singleton_var_p): Handle
1894         polynomial offsets.
1895         * tree-affine.c (wide_int_ext_for_comb): Make original widest_int
1896         version static and add an overload for poly_widest_int.
1897         (aff_combination_const, aff_combination_add_cst)
1898         (wide_int_constant_multiple_p, aff_comb_cannot_overlap_p): Take
1899         the constants as poly_widest_int rather than widest_int.
1900         (tree_to_aff_combination): Generalize INTEGER_CST case to
1901         poly_int_tree_p.
1902         (aff_combination_to_tree): Track offsets as poly_widest_ints.
1903         (aff_combination_add_product, aff_combination_mult): Handle
1904         polynomial offsets.
1905         (aff_combination_constant_multiple_p): Return the multiplier
1906         as a poly_widest_int.
1907         * tree-predcom.c (determine_offset): Return the offset as a
1908         poly_widest_int.
1909         (split_data_refs_to_components, suitable_component_p): Update
1910         accordingly.
1911         (valid_initializer_p): Update call to
1912         aff_combination_constant_multiple_p.
1913         * tree-ssa-address.c (addr_to_parts): Handle polynomial offsets.
1914         * tree-ssa-loop-ivopts.c (get_address_cost_ainc): Take the step
1915         as a poly_int64 rather than a HOST_WIDE_INT.
1916         (get_address_cost): Handle polynomial offsets.
1917         (iv_elimination_compare_lt): Likewise.
1918         (rewrite_use_nonlinear_expr): Likewise.
1920 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
1921             Alan Hayward  <alan.hayward@arm.com>
1922             David Sherwood  <david.sherwood@arm.com>
1924         * tree-dfa.h (get_addr_base_and_unit_offset_1): Return the offset
1925         as a poly_int64_pod rather than a HOST_WIDE_INT.
1926         (get_addr_base_and_unit_offset): Likewise.
1927         * tree-dfa.c (get_addr_base_and_unit_offset_1): Likewise.
1928         (get_addr_base_and_unit_offset): Likewise.
1929         * doc/match-and-simplify.texi: Change off from HOST_WIDE_INT
1930         to poly_int64 in example.
1931         * fold-const.c (fold_binary_loc): Update call to
1932         get_addr_base_and_unit_offset.
1933         * gimple-fold.c (gimple_fold_builtin_memory_op): Likewise.
1934         (maybe_canonicalize_mem_ref_addr): Likewise.
1935         (gimple_fold_stmt_to_constant_1): Likewise.
1936         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref):
1937         Likewise.
1938         * ipa-param-manipulation.c (ipa_modify_call_arguments): Likewise.
1939         * match.pd: Likewise.
1940         * omp-low.c (lower_omp_target): Likewise.
1941         * tree-sra.c (build_ref_for_offset): Likewise.
1942         (build_debug_ref_for_model): Likewise.
1943         * tree-ssa-address.c (maybe_fold_tmr): Likewise.
1944         * tree-ssa-alias.c (ao_ref_init_from_ptr_and_size): Likewise.
1945         * tree-ssa-ccp.c (optimize_memcpy): Likewise.
1946         * tree-ssa-forwprop.c (forward_propagate_addr_expr_1): Likewise.
1947         (constant_pointer_difference): Likewise.
1948         * tree-ssa-loop-niter.c (expand_simple_operations): Likewise.
1949         * tree-ssa-phiopt.c (jump_function_from_stmt): Likewise.
1950         * tree-ssa-pre.c (create_component_ref_by_pieces_1): Likewise.
1951         * tree-ssa-sccvn.c (vn_reference_fold_indirect): Likewise.
1952         (vn_reference_maybe_forwprop_address, vn_reference_lookup_3): Likewise.
1953         (set_ssa_val_to): Likewise.
1954         * tree-ssa-strlen.c (get_addr_stridx, addr_stridxptr)
1955         (maybe_diag_stxncpy_trunc): Likewise.
1956         * tree-vrp.c (vrp_prop::check_array_ref): Likewise.
1957         * tree.c (build_simple_mem_ref_loc): Likewise.
1958         (array_at_struct_end_p): Likewise.
1960 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
1961             Alan Hayward  <alan.hayward@arm.com>
1962             David Sherwood  <david.sherwood@arm.com>
1964         * tree-dfa.h (get_ref_base_and_extent): Return the base, size and
1965         max_size as poly_int64_pods rather than HOST_WIDE_INTs.
1966         (get_ref_base_and_extent_hwi): Declare.
1967         * tree-dfa.c (get_ref_base_and_extent): Return the base, size and
1968         max_size as poly_int64_pods rather than HOST_WIDE_INTs.
1969         (get_ref_base_and_extent_hwi): New function.
1970         * cfgexpand.c (expand_debug_expr): Update call to
1971         get_ref_base_and_extent.
1972         * dwarf2out.c (add_var_loc_to_decl): Likewise.
1973         * gimple-fold.c (get_base_constructor): Return the offset as a
1974         poly_int64_pod rather than a HOST_WIDE_INT.
1975         (fold_const_aggregate_ref_1): Track polynomial sizes and offsets.
1976         * ipa-polymorphic-call.c
1977         (ipa_polymorphic_call_context::set_by_invariant)
1978         (extr_type_from_vtbl_ptr_store): Track polynomial offsets.
1979         (ipa_polymorphic_call_context::ipa_polymorphic_call_context)
1980         (check_stmt_for_type_change): Use get_ref_base_and_extent_hwi
1981         rather than get_ref_base_and_extent.
1982         (ipa_polymorphic_call_context::get_dynamic_type): Likewise.
1983         * ipa-prop.c (ipa_load_from_parm_agg, compute_complex_assign_jump_func)
1984         (get_ancestor_addr_info, determine_locally_known_aggregate_parts):
1985         Likewise.
1986         * ipa-param-manipulation.c (ipa_get_adjustment_candidate): Update
1987         call to get_ref_base_and_extent.
1988         * tree-sra.c (create_access, get_access_for_expr): Likewise.
1989         * tree-ssa-alias.c (ao_ref_base, aliasing_component_refs_p)
1990         (stmt_kills_ref_p): Likewise.
1991         * tree-ssa-dce.c (mark_aliased_reaching_defs_necessary_1): Likewise.
1992         * tree-ssa-scopedtables.c (avail_expr_hash, equal_mem_array_ref_p):
1993         Likewise.
1994         * tree-ssa-sccvn.c (vn_reference_lookup_3): Likewise.
1995         Use get_ref_base_and_extent_hwi rather than get_ref_base_and_extent
1996         when calling native_encode_expr.
1997         * tree-ssa-structalias.c (get_constraint_for_component_ref): Update
1998         call to get_ref_base_and_extent.
1999         (do_structure_copy): Use get_ref_base_and_extent_hwi rather than
2000         get_ref_base_and_extent.
2001         * var-tracking.c (track_expr_p): Likewise.
2003 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2004             Alan Hayward  <alan.hayward@arm.com>
2005             David Sherwood  <david.sherwood@arm.com>
2007         * ipa-param-manipulation.h (ipa_parm_adjustment::offset): Change from
2008         HOST_WIDE_INT to poly_int64_pod.
2009         * ipa-param-manipulation.c (ipa_modify_call_arguments): Track
2010         polynomail parameter offsets.
2012 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2013             Alan Hayward  <alan.hayward@arm.com>
2014             David Sherwood  <david.sherwood@arm.com>
2016         * gengtype.c (main): Handle poly_int64_pod.
2017         * dwarf2out.h (dw_cfi_oprnd_cfa_loc): New dw_cfi_oprnd_type.
2018         (dw_cfi_oprnd::dw_cfi_cfa_loc): New field.
2019         (dw_cfa_location::offset, dw_cfa_location::base_offset): Change
2020         from HOST_WIDE_INT to poly_int64_pod.
2021         * dwarf2cfi.c (queued_reg_save::cfa_offset): Likewise.
2022         (copy_cfa): New function.
2023         (lookup_cfa_1): Use the cached dw_cfi_cfa_loc, if it exists.
2024         (cfi_oprnd_equal_p): Handle dw_cfi_oprnd_cfa_loc.
2025         (cfa_equal_p, dwarf2out_frame_debug_adjust_cfa)
2026         (dwarf2out_frame_debug_cfa_offset, dwarf2out_frame_debug_expr)
2027         (initial_return_save): Treat offsets as poly_ints.
2028         (def_cfa_0): Likewise.  Cache the CFA in dw_cfi_cfa_loc if either
2029         offset is nonconstant.
2030         (reg_save): Take the offset as a poly_int64.  Fall back to
2031         DW_CFA_expression for nonconstant offsets.
2032         (queue_reg_save): Take the offset as a poly_int64.
2033         * dwarf2out.c (dw_cfi_oprnd2_desc): Handle DW_CFA_def_cfa_expression.
2035 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2036             Alan Hayward  <alan.hayward@arm.com>
2037             David Sherwood  <david.sherwood@arm.com>
2039         * rtl.h (operand_subword, operand_subword_force): Take the offset
2040         as a poly_uint64 an unsigned int.
2041         * emit-rtl.c (operand_subword, operand_subword_force): Likewise.
2043 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2044             Alan Hayward  <alan.hayward@arm.com>
2045             David Sherwood  <david.sherwood@arm.com>
2047         * doc/rtl.texi: Update documentation of SUBREG_BYTE.  Document the
2048         'p' format code.  Use INT_LIST rather than SUBREG as the example of
2049         a code with an XINT and an XEXP.  Remove the implication that
2050         accessing an rtx field using XINT is expected to work.
2051         * rtl.def (SUBREG): Change format from "ei" to "ep".
2052         * rtl.h (rtunion::rt_subreg): New field.
2053         (XCSUBREG): New macro.
2054         (SUBREG_BYTE): Use it.
2055         (subreg_shape): Change offset from an unsigned int to a poly_uint16.
2056         Update constructor accordingly.
2057         (subreg_shape::operator ==): Update accordingly.
2058         (subreg_shape::unique_id): Return an unsigned HOST_WIDE_INT rather
2059         than an unsigned int.
2060         (subreg_lsb, subreg_lowpart_offset, subreg_highpart_offset): Return
2061         a poly_uint64 rather than an unsigned int.
2062         (subreg_lsb_1): Likewise.  Take the offset as a poly_uint64 rather
2063         than an unsigned int.
2064         (subreg_size_offset_from_lsb, subreg_size_lowpart_offset)
2065         (subreg_size_highpart_offset): Return a poly_uint64 rather than
2066         an unsigned int.  Take the sizes as poly_uint64s.
2067         (subreg_offset_from_lsb): Return a poly_uint64 rather than
2068         an unsigned int.  Take the shift as a poly_uint64 rather than
2069         an unsigned int.
2070         (subreg_regno_offset, subreg_offset_representable_p): Take the offset
2071         as a poly_uint64 rather than an unsigned int.
2072         (simplify_subreg_regno): Likewise.
2073         (byte_lowpart_offset): Return the memory offset as a poly_int64
2074         rather than an int.
2075         (subreg_memory_offset): Likewise.  Take the subreg offset as a
2076         poly_uint64 rather than an unsigned int.
2077         (simplify_subreg, simplify_gen_subreg, subreg_get_info)
2078         (gen_rtx_SUBREG, validate_subreg): Take the subreg offset as a
2079         poly_uint64 rather than an unsigned int.
2080         * rtl.c (rtx_format): Describe 'p' in comment.
2081         (copy_rtx, rtx_equal_p_cb, rtx_equal_p): Handle 'p'.
2082         * emit-rtl.c (validate_subreg, gen_rtx_SUBREG): Take the subreg
2083         offset as a poly_uint64 rather than an unsigned int.
2084         (byte_lowpart_offset): Return the memory offset as a poly_int64
2085         rather than an int.
2086         (subreg_memory_offset): Likewise.  Take the subreg offset as a
2087         poly_uint64 rather than an unsigned int.
2088         (subreg_size_lowpart_offset, subreg_size_highpart_offset): Take the
2089         mode sizes as poly_uint64s rather than unsigned ints.  Return a
2090         poly_uint64 rather than an unsigned int.
2091         (subreg_lowpart_p): Treat subreg offsets as poly_ints.
2092         (copy_insn_1): Handle 'p'.
2093         * rtlanal.c (set_noop_p): Treat subregs offsets as poly_uint64s.
2094         (subreg_lsb_1): Take the subreg offset as a poly_uint64 rather than
2095         an unsigned int.  Return the shift in the same way.
2096         (subreg_lsb): Return the shift as a poly_uint64 rather than an
2097         unsigned int.
2098         (subreg_size_offset_from_lsb): Take the sizes and shift as
2099         poly_uint64s rather than unsigned ints.  Return the offset as
2100         a poly_uint64.
2101         (subreg_get_info, subreg_regno_offset, subreg_offset_representable_p)
2102         (simplify_subreg_regno): Take the offset as a poly_uint64 rather than
2103         an unsigned int.
2104         * rtlhash.c (add_rtx): Handle 'p'.
2105         * genemit.c (gen_exp): Likewise.
2106         * gengenrtl.c (type_from_format, gendef): Likewise.
2107         * gensupport.c (subst_pattern_match, get_alternatives_number)
2108         (collect_insn_data, alter_predicate_for_insn, alter_constraints)
2109         (subst_dup): Likewise.
2110         * gengtype.c (adjust_field_rtx_def): Likewise.
2111         * genrecog.c (find_operand, find_matching_operand, validate_pattern)
2112         (match_pattern_2): Likewise.
2113         (rtx_test::SUBREG_FIELD): New rtx_test::kind_enum.
2114         (rtx_test::subreg_field): New function.
2115         (operator ==, safe_to_hoist_p, transition_parameter_type)
2116         (print_nonbool_test, print_test): Handle SUBREG_FIELD.
2117         * genattrtab.c (attr_rtx_1): Say that 'p' is deliberately not handled.
2118         * genpeep.c (match_rtx): Likewise.
2119         * print-rtl.c (print_poly_int): Include if GENERATOR_FILE too.
2120         (rtx_writer::print_rtx_operand): Handle 'p'.
2121         (print_value): Handle SUBREG.
2122         * read-rtl.c (apply_int_iterator): Likewise.
2123         (rtx_reader::read_rtx_operand): Handle 'p'.
2124         * alias.c (rtx_equal_for_memref_p): Likewise.
2125         * cselib.c (rtx_equal_for_cselib_1, cselib_hash_rtx): Likewise.
2126         * caller-save.c (replace_reg_with_saved_mem): Treat subreg offsets
2127         as poly_ints.
2128         * calls.c (expand_call): Likewise.
2129         * combine.c (combine_simplify_rtx, expand_field_assignment): Likewise.
2130         (make_extraction, gen_lowpart_for_combine): Likewise.
2131         * loop-invariant.c (hash_invariant_expr_1, invariant_expr_equal_p):
2132         Likewise.
2133         * cse.c (remove_invalid_subreg_refs): Take the offset as a poly_uint64
2134         rather than an unsigned int.  Treat subreg offsets as poly_ints.
2135         (exp_equiv_p): Handle 'p'.
2136         (hash_rtx_cb): Likewise.  Treat subreg offsets as poly_ints.
2137         (equiv_constant, cse_insn): Treat subreg offsets as poly_ints.
2138         * dse.c (find_shift_sequence): Likewise.
2139         * dwarf2out.c (rtl_for_decl_location): Likewise.
2140         * expmed.c (extract_low_bits): Likewise.
2141         * expr.c (emit_group_store, undefined_operand_subword_p): Likewise.
2142         (expand_expr_real_2): Likewise.
2143         * final.c (alter_subreg): Likewise.
2144         (leaf_renumber_regs_insn): Handle 'p'.
2145         * function.c (assign_parm_find_stack_rtl, assign_parm_setup_stack):
2146         Treat subreg offsets as poly_ints.
2147         * fwprop.c (forward_propagate_and_simplify): Likewise.
2148         * ifcvt.c (noce_emit_move_insn, noce_emit_cmove): Likewise.
2149         * ira.c (get_subreg_tracking_sizes): Likewise.
2150         * ira-conflicts.c (go_through_subreg): Likewise.
2151         * ira-lives.c (process_single_reg_class_operands): Likewise.
2152         * jump.c (rtx_renumbered_equal_p): Likewise.  Handle 'p'.
2153         * lower-subreg.c (simplify_subreg_concatn): Take the subreg offset
2154         as a poly_uint64 rather than an unsigned int.
2155         (simplify_gen_subreg_concatn, resolve_simple_move): Treat
2156         subreg offsets as poly_ints.
2157         * lra-constraints.c (operands_match_p): Handle 'p'.
2158         (match_reload, curr_insn_transform): Treat subreg offsets as poly_ints.
2159         * lra-spills.c (assign_mem_slot): Likewise.
2160         * postreload.c (move2add_valid_value_p): Likewise.
2161         * recog.c (general_operand, indirect_operand): Likewise.
2162         * regcprop.c (copy_value, maybe_mode_change): Likewise.
2163         (copyprop_hardreg_forward_1): Likewise.
2164         * reginfo.c (simplifiable_subregs_hasher::hash, simplifiable_subregs)
2165         (record_subregs_of_mode): Likewise.
2166         * rtlhooks.c (gen_lowpart_general, gen_lowpart_if_possible): Likewise.
2167         * reload.c (operands_match_p): Handle 'p'.
2168         (find_reloads_subreg_address): Treat subreg offsets as poly_ints.
2169         * reload1.c (alter_reg, choose_reload_regs): Likewise.
2170         (compute_reload_subreg_offset): Likewise, and return an poly_int64.
2171         * simplify-rtx.c (simplify_truncation, simplify_binary_operation_1):
2172         (test_vector_ops_duplicate): Treat subreg offsets as poly_ints.
2173         (simplify_const_poly_int_tests<N>::run): Likewise.
2174         (simplify_subreg, simplify_gen_subreg): Take the subreg offset as
2175         a poly_uint64 rather than an unsigned int.
2176         * valtrack.c (debug_lowpart_subreg): Likewise.
2177         * var-tracking.c (var_lowpart): Likewise.
2178         (loc_cmp): Handle 'p'.
2180 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2181             Alan Hayward  <alan.hayward@arm.com>
2182             David Sherwood  <david.sherwood@arm.com>
2184         * ira.c (get_subreg_tracking_sizes): New function.
2185         (init_live_subregs): Take an integer size rather than a register.
2186         (build_insn_chain): Use get_subreg_tracking_sizes.  Update calls
2187         to init_live_subregs.
2189 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2190             Alan Hayward  <alan.hayward@arm.com>
2191             David Sherwood  <david.sherwood@arm.com>
2193         * expr.c (store_constructor_field): Change bitsize from a
2194         unsigned HOST_WIDE_INT to a poly_uint64 and bitpos from a
2195         HOST_WIDE_INT to a poly_int64.
2196         (store_constructor): Change size from a HOST_WIDE_INT to
2197         a poly_int64.
2198         (store_field): Likewise bitsize and bitpos.
2200 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2201             Alan Hayward  <alan.hayward@arm.com>
2202             David Sherwood  <david.sherwood@arm.com>
2204         * expmed.h (store_bit_field): Change bitregion_start and
2205         bitregion_end from unsigned HOST_WIDE_INT to poly_uint64.
2206         * expmed.c (adjust_bit_field_mem_for_reg, strict_volatile_bitfield_p)
2207         (store_bit_field_1, store_integral_bit_field, store_bit_field)
2208         (store_fixed_bit_field, store_split_bit_field): Likewise.
2209         * expr.c (store_constructor_field, store_field): Likewise.
2210         (optimize_bitfield_assignment_op): Likewise.  Make the same change
2211         to bitsize and bitpos.
2212         * machmode.h (bit_field_mode_iterator): Change m_bitregion_start
2213         and m_bitregion_end from HOST_WIDE_INT to poly_int64.  Make the
2214         same change in the constructor arguments.
2215         (get_best_mode): Change bitregion_start and bitregion_end from
2216         unsigned HOST_WIDE_INT to poly_uint64.
2217         * stor-layout.c (bit_field_mode_iterator::bit_field_mode_iterator):
2218         Change bitregion_start and bitregion_end from HOST_WIDE_INT to
2219         poly_int64.
2220         (bit_field_mode_iterator::next_mode): Update for new types
2221         of m_bitregion_start and m_bitregion_end.
2222         (get_best_mode): Change bitregion_start and bitregion_end from
2223         unsigned HOST_WIDE_INT to poly_uint64.
2225 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2226             Alan Hayward  <alan.hayward@arm.com>
2227             David Sherwood  <david.sherwood@arm.com>
2229         * rtl.h (simplify_gen_subreg): Add a temporary overload that
2230         accepts poly_uint64 offsets.
2231         * expmed.h (extract_bit_field): Take bitsize and bitnum as
2232         poly_uint64s rather than unsigned HOST_WIDE_INTs.
2233         * expmed.c (lowpart_bit_field_p): Likewise.
2234         (extract_bit_field_as_subreg): New function, split out from...
2235         (extract_bit_field_1): ...here.  Take bitsize and bitnum as
2236         poly_uint64s rather than unsigned HOST_WIDE_INTs.  For vector
2237         extractions, check that BITSIZE matches the size of the extracted
2238         value and that BITNUM is an exact multiple of that size.
2239         If all else fails, try forcing the value into memory if
2240         BITNUM is variable, and adjusting the address so that the
2241         offset is constant.  Split the part that can only handle constant
2242         bitsize and bitnum out into...
2243         (extract_integral_bit_field): ...this new function.
2244         (extract_bit_field): Take bitsize and bitnum as poly_uint64s
2245         rather than unsigned HOST_WIDE_INTs.
2247 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2248             Alan Hayward  <alan.hayward@arm.com>
2249             David Sherwood  <david.sherwood@arm.com>
2251         * expmed.h (store_bit_field): Take bitsize and bitnum as
2252         poly_uint64s rather than unsigned HOST_WIDE_INTs.
2253         * expmed.c (simple_mem_bitfield_p): Likewise.  Add a parameter
2254         that returns the byte size.
2255         (store_bit_field_1): Take bitsize and bitnum as
2256         poly_uint64s rather than unsigned HOST_WIDE_INTs.  Update call
2257         to simple_mem_bitfield_p.  Split the part that can only handle
2258         constant bitsize and bitnum out into...
2259         (store_integral_bit_field): ...this new function.
2260         (store_bit_field): Take bitsize and bitnum as poly_uint64s rather
2261         than unsigned HOST_WIDE_INTs.
2262         (extract_bit_field_1): Update call to simple_mem_bitfield_p.
2264 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2265             Alan Hayward  <alan.hayward@arm.com>
2266             David Sherwood  <david.sherwood@arm.com>
2268         * lra-int.h (lra_reg): Change offset from int to poly_int64.
2269         (lra_insn_recog_data): Change sp_offset from HOST_WIDE_INT
2270         to poly_int64.
2271         (lra_eliminate_regs_1, eliminate_regs_in_insn): Change
2272         update_sp_offset from a HOST_WIDE_INT to a poly_int64.
2273         (lra_update_reg_val_offset, lra_reg_val_equal_p): Take the
2274         offset as a poly_int64 rather than an int.
2275         * lra-assigns.c (find_hard_regno_for_1): Handle poly_int64 offsets.
2276         (setup_live_pseudos_and_spill_after_risky_transforms): Likewise.
2277         * lra-constraints.c (equiv_address_substitution): Track offsets
2278         as poly_int64s.
2279         (emit_inc): Check poly_int_rtx_p instead of CONST_INT_P.
2280         (curr_insn_transform): Handle the new form of sp_offset.
2281         * lra-eliminations.c (lra_elim_table): Change previous_offset
2282         and offset from HOST_WIDE_INT to poly_int64.
2283         (print_elim_table, update_reg_eliminate): Update accordingly.
2284         (self_elim_offsets): Change from HOST_WIDE_INT to poly_int64_pod.
2285         (get_elimination): Update accordingly.
2286         (form_sum): Check poly_int_rtx_p instead of CONST_INT_P.
2287         (lra_eliminate_regs_1, eliminate_regs_in_insn): Change
2288         update_sp_offset from a HOST_WIDE_INT to a poly_int64.  Handle
2289         poly_int64 offsets generally.
2290         (curr_sp_change): Change from HOST_WIDE_INT to poly_int64.
2291         (mark_not_eliminable, init_elimination): Update accordingly.
2292         (remove_reg_equal_offset_note): Return a bool and pass the new
2293         offset back by pointer as a poly_int64.
2294         * lra-remat.c (change_sp_offset): Take sp_offset as a poly_int64
2295         rather than a HOST_WIDE_INT.
2296         (do_remat): Track offsets poly_int64s.
2297         * lra.c (lra_update_insn_recog_data, setup_sp_offset): Likewise.
2299 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2300             Alan Hayward  <alan.hayward@arm.com>
2301             David Sherwood  <david.sherwood@arm.com>
2303         * rtl.h (mem_attrs): Add a default constructor.  Change size and
2304         offset from HOST_WIDE_INT to poly_int64.
2305         * emit-rtl.h (set_mem_offset, set_mem_size, adjust_address_1)
2306         (adjust_automodify_address_1, set_mem_attributes_minus_bitpos)
2307         (widen_memory_access): Take the sizes and offsets as poly_int64s
2308         rather than HOST_WIDE_INTs.
2309         * alias.c (ao_ref_from_mem): Handle the new form of MEM_OFFSET.
2310         (offset_overlap_p): Take poly_int64s rather than HOST_WIDE_INTs
2311         and ints.
2312         (adjust_offset_for_component_ref): Change the offset from a
2313         HOST_WIDE_INT to a poly_int64.
2314         (nonoverlapping_memrefs_p): Track polynomial offsets and sizes.
2315         * cfgcleanup.c (merge_memattrs): Update after mem_attrs changes.
2316         * dce.c (find_call_stack_args): Likewise.
2317         * dse.c (record_store): Likewise.
2318         * dwarf2out.c (tls_mem_loc_descriptor, dw_sra_loc_expr): Likewise.
2319         * print-rtl.c (rtx_writer::print_rtx): Likewise.
2320         * read-rtl-function.c (test_loading_mem): Likewise.
2321         * rtlanal.c (may_trap_p_1): Likewise.
2322         * simplify-rtx.c (delegitimize_mem_from_attrs): Likewise.
2323         * var-tracking.c (int_mem_offset, track_expr_p): Likewise.
2324         * emit-rtl.c (mem_attrs_eq_p, get_mem_align_offset): Likewise.
2325         (mem_attrs::mem_attrs): New function.
2326         (set_mem_attributes_minus_bitpos): Change bitpos from a
2327         HOST_WIDE_INT to poly_int64.
2328         (set_mem_alias_set, set_mem_addr_space, set_mem_align, set_mem_expr)
2329         (clear_mem_offset, clear_mem_size, change_address)
2330         (get_spill_slot_decl, set_mem_attrs_for_spill): Directly
2331         initialize mem_attrs.
2332         (set_mem_offset, set_mem_size, adjust_address_1)
2333         (adjust_automodify_address_1, offset_address, widen_memory_access):
2334         Likewise.  Take poly_int64s rather than HOST_WIDE_INT.
2336 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2337             Alan Hayward  <alan.hayward@arm.com>
2338             David Sherwood  <david.sherwood@arm.com>
2340         * rtlanal.c (rtx_addr_can_trap_p_1): Take the offset and size
2341         as poly_int64s rather than HOST_WIDE_INTs.  Use a size of -1
2342         rather than 0 to represent an unknown size.  Assert that the size
2343         is known when the mode isn't BLKmode.
2344         (may_trap_p_1): Use -1 for unknown sizes.
2345         (rtx_addr_can_trap_p): Likewise.  Pass BLKmode rather than VOIDmode.
2347 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2348             Alan Hayward  <alan.hayward@arm.com>
2349             David Sherwood  <david.sherwood@arm.com>
2351         * dse.c (store_info): Change offset and width from HOST_WIDE_INT
2352         to poly_int64.  Update commentary for positions_needed.large.
2353         (read_info_type): Change offset and width from HOST_WIDE_INT
2354         to poly_int64.
2355         (set_usage_bits): Likewise.
2356         (canon_address): Return the offset as a poly_int64 rather than
2357         a HOST_WIDE_INT.  Use strip_offset_and_add.
2358         (set_all_positions_unneeded, any_positions_needed_p): Use
2359         positions_needed.large to track stores with non-constant widths.
2360         (all_positions_needed_p): Likewise.  Take the offset and width
2361         as poly_int64s rather than ints.  Assert that rhs is nonnull.
2362         (record_store): Cope with non-constant offsets and widths.
2363         Nullify the rhs of an earlier store if we can't tell which bytes
2364         of it are needed.
2365         (find_shift_sequence): Take the access_size and shift as poly_int64s
2366         rather than ints.
2367         (get_stored_val): Take the read_offset and read_width as poly_int64s
2368         rather than HOST_WIDE_INTs.
2369         (check_mem_read_rtx, scan_stores, scan_reads, dse_step5): Handle
2370         non-constant offsets and widths.
2372 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2373             Alan Hayward  <alan.hayward@arm.com>
2374             David Sherwood  <david.sherwood@arm.com>
2376         * inchash.h (inchash::hash::add_poly_int): New function.
2377         * tree-ssa-alias.h (ao_ref::offset, ao_ref::size, ao_ref::max_size):
2378         Use poly_int64 rather than HOST_WIDE_INT.
2379         (ao_ref::max_size_known_p): New function.
2380         * tree-ssa-sccvn.h (vn_reference_op_struct::off): Use poly_int64_pod
2381         rather than HOST_WIDE_INT.
2382         * tree-ssa-alias.c (ao_ref_base): Apply get_ref_base_and_extent
2383         to temporaries until its interface is adjusted to match.
2384         (ao_ref_init_from_ptr_and_size): Handle polynomial offsets and sizes.
2385         (aliasing_component_refs_p, decl_refs_may_alias_p)
2386         (indirect_ref_may_alias_decl_p, indirect_refs_may_alias_p): Take
2387         the offsets and max_sizes as poly_int64s instead of HOST_WIDE_INTs.
2388         (refs_may_alias_p_1, stmt_kills_ref_p): Adjust for changes to
2389         ao_ref fields.
2390         * alias.c (ao_ref_from_mem): Likewise.
2391         * tree-ssa-dce.c (mark_aliased_reaching_defs_necessary_1): Likewise.
2392         * tree-ssa-dse.c (valid_ao_ref_for_dse, normalize_ref)
2393         (clear_bytes_written_by, setup_live_bytes_from_ref, compute_trims)
2394         (maybe_trim_complex_store, maybe_trim_constructor_store)
2395         (live_bytes_read, dse_classify_store): Likewise.
2396         * tree-ssa-sccvn.c (vn_reference_compute_hash, vn_reference_eq):
2397         (copy_reference_ops_from_ref, ao_ref_init_from_vn_reference)
2398         (fully_constant_vn_reference_p, valueize_refs_1): Likewise.
2399         (vn_reference_lookup_3): Likewise.
2400         * tree-ssa-uninit.c (warn_uninitialized_vars): Likewise.
2402 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2403             Alan Hayward  <alan.hayward@arm.com>
2404             David Sherwood  <david.sherwood@arm.com>
2406         * tree-ssa-alias.c (indirect_ref_may_alias_decl_p)
2407         (indirect_refs_may_alias_p): Use ranges_may_overlap_p
2408         instead of ranges_overlap_p.
2410 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2411             Alan Hayward  <alan.hayward@arm.com>
2412             David Sherwood  <david.sherwood@arm.com>
2414         * tree-ssa-alias.c (same_addr_size_stores_p): Take the offsets and
2415         sizes as poly_int64s rather than HOST_WIDE_INTs.
2417 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2418             Alan Hayward  <alan.hayward@arm.com>
2419             David Sherwood  <david.sherwood@arm.com>
2421         * gimple-fold.h (fold_ctor_reference): Take the offset and size
2422         as poly_uint64 rather than unsigned HOST_WIDE_INT.
2423         * gimple-fold.c (fold_ctor_reference): Likewise.
2425 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2426             Alan Hayward  <alan.hayward@arm.com>
2427             David Sherwood  <david.sherwood@arm.com>
2429         * target.def (dwarf_poly_indeterminate_value): New hook.
2430         * targhooks.h (default_dwarf_poly_indeterminate_value): Declare.
2431         * targhooks.c (default_dwarf_poly_indeterminate_value): New function.
2432         * doc/tm.texi.in (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Document.
2433         * doc/tm.texi: Regenerate.
2434         * dwarf2out.h (build_cfa_loc, build_cfa_aligned_loc): Take the
2435         offset as a poly_int64.
2436         * dwarf2out.c (new_reg_loc_descr): Move later in file.  Take the
2437         offset as a poly_int64.
2438         (loc_descr_plus_const, loc_list_plus_const, build_cfa_aligned_loc):
2439         Take the offset as a poly_int64.
2440         (build_cfa_loc): Likewise.  Use loc_descr_plus_const.
2441         (frame_pointer_fb_offset): Change to a poly_int64.
2442         (int_loc_descriptor): Take the offset as a poly_int64.  Use
2443         targetm.dwarf_poly_indeterminate_value for polynomial offsets.
2444         (based_loc_descr): Take the offset as a poly_int64.
2445         Use strip_offset_and_add to handle (plus X (const)).
2446         Use new_reg_loc_descr instead of an open-coded version of the
2447         previous implementation.
2448         (mem_loc_descriptor): Handle CONST_POLY_INT.
2449         (compute_frame_pointer_to_fb_displacement): Take the offset as a
2450         poly_int64.  Use strip_offset_and_add to handle (plus X (const)).
2452 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2453             Alan Hayward  <alan.hayward@arm.com>
2454             David Sherwood  <david.sherwood@arm.com>
2456         * rtl.h (reg_attrs::offset): Change from HOST_WIDE_INT to poly_int64.
2457         (gen_rtx_REG_offset): Take the offset as a poly_int64.
2458         * inchash.h (inchash::hash::add_poly_hwi): New function.
2459         * gengtype.c (main): Register poly_int64.
2460         * emit-rtl.c (reg_attr_hasher::hash): Use inchash.  Treat the
2461         offset as a poly_int.
2462         (reg_attr_hasher::equal): Use must_eq to compare offsets.
2463         (get_reg_attrs, update_reg_offset, gen_rtx_REG_offset): Take the
2464         offset as a poly_int64.
2465         (set_reg_attrs_from_value): Treat the offset as a poly_int64.
2466         * print-rtl.c (print_poly_int): New function.
2467         (rtx_writer::print_rtx_operand_code_r): Treat REG_OFFSET as
2468         a poly_int.
2469         * var-tracking.c (track_offset_p, get_tracked_reg_offset): New
2470         functions.
2471         (var_reg_set, var_reg_delete_and_set, var_reg_delete): Use them.
2472         (same_variable_part_p, track_loc_p): Take the offset as a poly_int64.
2473         (vt_get_decl_and_offset): Return the offset as a poly_int64.
2474         Enforce track_offset_p for parts of a PARALLEL.
2475         (vt_add_function_parameter): Use const_offset for the final
2476         offset to track.  Use get_tracked_reg_offset for the parts
2477         of a PARALLEL.
2479 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2480             Alan Hayward  <alan.hayward@arm.com>
2481             David Sherwood  <david.sherwood@arm.com>
2483         * target.def (truly_noop_truncation): Take poly_uint64s instead of
2484         unsigned ints.  Change default to hook_bool_puint64_puint64_true.
2485         * doc/tm.texi: Regenerate.
2486         * hooks.h (hook_bool_uint_uint_true): Delete.
2487         (hook_bool_puint64_puint64_true): Declare.
2488         * hooks.c (hook_bool_uint_uint_true): Delete.
2489         (hook_bool_puint64_puint64_true): New function.
2490         * config/mips/mips.c (mips_truly_noop_truncation): Take poly_uint64s
2491         instead of unsigned ints.
2492         * config/spu/spu.c (spu_truly_noop_truncation): Likewise.
2493         * config/tilegx/tilegx.c (tilegx_truly_noop_truncation): Likewise.
2495 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2496             Alan Hayward  <alan.hayward@arm.com>
2497             David Sherwood  <david.sherwood@arm.com>
2499         * optabs.h (expand_operand): Add an int_value field.
2500         (create_expand_operand): Add an int_value parameter and use it
2501         to initialize the new expand_operand field.
2502         (create_integer_operand): Replace with a declaration of a function
2503         that accepts poly_int64s.  Move the implementation to...
2504         * optabs.c (create_integer_operand): ...here.
2505         (maybe_legitimize_operand): For EXPAND_INTEGER, check whether
2506         the mode preserves the value of int_value, instead of calling
2507         const_int_operand on the rtx.  Use gen_int_mode to generate
2508         the new rtx.
2510 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2511             Alan Hayward  <alan.hayward@arm.com>
2512             David Sherwood  <david.sherwood@arm.com>
2514         * dumpfile.h (dump_dec): Declare.
2515         * dumpfile.c (dump_dec): New function.
2516         * pretty-print.h (pp_wide_integer): Turn into a function and
2517         declare a poly_int version.
2518         * pretty-print.c (pp_wide_integer): New function for poly_ints.
2520 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2521             Alan Hayward  <alan.hayward@arm.com>
2522             David Sherwood  <david.sherwood@arm.com>
2524         * doc/generic.texi (POLY_INT_CST): Document.
2525         * tree.def (POLY_INT_CST): New tree code.
2526         * treestruct.def (TS_POLY_INT_CST): New tree layout.
2527         * tree-core.h (tree_poly_int_cst): New struct.
2528         (tree_node): Add a poly_int_cst field.
2529         * tree.h (POLY_INT_CST_P, POLY_INT_CST_COEFF): New macros.
2530         (wide_int_to_tree, force_fit_type): Take a poly_wide_int_ref
2531         instead of a wide_int_ref.
2532         (build_int_cst, build_int_cst_type): Take a poly_int64 instead
2533         of a HOST_WIDE_INT.
2534         (build_int_cstu, build_array_type_nelts): Take a poly_uint64
2535         instead of an unsigned HOST_WIDE_INT.
2536         (build_poly_int_cst, tree_fits_poly_int64_p, tree_fits_poly_uint64_p)
2537         (ptrdiff_tree_p): Declare.
2538         (tree_to_poly_int64, tree_to_poly_uint64): Likewise.  Provide
2539         extern inline implementations if the target doesn't use POLY_INT_CST.
2540         (poly_int_tree_p): New function.
2541         (wi::unextended_tree): New class.
2542         (wi::int_traits <unextended_tree>): New override.
2543         (wi::extended_tree): Add a default constructor.
2544         (wi::extended_tree::get_tree): New function.
2545         (wi::widest_extended_tree, wi::offset_extended_tree): New typedefs.
2546         (wi::tree_to_widest_ref, wi::tree_to_offset_ref): Use them.
2547         (wi::tree_to_poly_widest_ref, wi::tree_to_poly_offset_ref)
2548         (wi::tree_to_poly_wide_ref): New typedefs.
2549         (wi::ints_for): Provide overloads for extended_tree and
2550         unextended_tree.
2551         (poly_int_cst_value, wi::to_poly_widest, wi::to_poly_offset)
2552         (wi::to_wide): New functions.
2553         (wi::fits_to_boolean_p, wi::fits_to_tree_p): Handle poly_ints.
2554         * tree.c (poly_int_cst_hasher): New struct.
2555         (poly_int_cst_hash_table): New variable.
2556         (tree_node_structure_for_code, tree_code_size, simple_cst_equal)
2557         (valid_constant_size_p, add_expr, drop_tree_overflow): Handle
2558         POLY_INT_CST.
2559         (initialize_tree_contains_struct): Handle TS_POLY_INT_CST.
2560         (init_ttree): Initialize poly_int_cst_hash_table.
2561         (build_int_cst, build_int_cst_type, build_invariant_address): Take
2562         a poly_int64 instead of a HOST_WIDE_INT.
2563         (build_int_cstu, build_array_type_nelts): Take a poly_uint64
2564         instead of an unsigned HOST_WIDE_INT.
2565         (wide_int_to_tree): Rename to...
2566         (wide_int_to_tree_1): ...this.
2567         (build_new_poly_int_cst, build_poly_int_cst): New functions.
2568         (force_fit_type): Take a poly_wide_int_ref instead of a wide_int_ref.
2569         (wide_int_to_tree): New function that takes a poly_wide_int_ref.
2570         (ptrdiff_tree_p, tree_to_poly_int64, tree_to_poly_uint64)
2571         (tree_fits_poly_int64_p, tree_fits_poly_uint64_p): New functions.
2572         * lto-streamer-out.c (DFS::DFS_write_tree_body, hash_tree): Handle
2573         TS_POLY_INT_CST.
2574         * tree-streamer-in.c (lto_input_ts_poly_tree_pointers): Likewise.
2575         (streamer_read_tree_body): Likewise.
2576         * tree-streamer-out.c (write_ts_poly_tree_pointers): Likewise.
2577         (streamer_write_tree_body): Likewise.
2578         * tree-streamer.c (streamer_check_handled_ts_structures): Likewise.
2579         * asan.c (asan_protect_global): Require the size to be an INTEGER_CST.
2580         * cfgexpand.c (expand_debug_expr): Handle POLY_INT_CST.
2581         * expr.c (expand_expr_real_1, const_vector_from_tree): Likewise.
2582         * gimple-expr.h (is_gimple_constant): Likewise.
2583         * gimplify.c (maybe_with_size_expr): Likewise.
2584         * print-tree.c (print_node): Likewise.
2585         * tree-data-ref.c (data_ref_compare_tree): Likewise.
2586         * tree-pretty-print.c (dump_generic_node): Likewise.
2587         * tree-ssa-address.c (addr_for_mem_ref): Likewise.
2588         * tree-vect-data-refs.c (dr_group_sort_cmp): Likewise.
2589         * tree-vrp.c (compare_values_warnv): Likewise.
2590         * tree-ssa-loop-ivopts.c (determine_base_object, constant_multiple_of)
2591         (get_loop_invariant_expr, add_candidate_1, get_computation_aff_1)
2592         (force_expr_to_var_cost): Likewise.
2593         * tree-ssa-loop.c (for_each_index): Likewise.
2594         * fold-const.h (build_invariant_address, size_int_kind): Take a
2595         poly_int64 instead of a HOST_WIDE_INT.
2596         * fold-const.c (fold_negate_expr_1, const_binop, const_unop)
2597         (fold_convert_const, multiple_of_p, fold_negate_const): Handle
2598         POLY_INT_CST.
2599         (size_binop_loc): Likewise.  Allow int_const_binop_1 to fail.
2600         (int_const_binop_2): New function, split out from...
2601         (int_const_binop_1): ...here.  Handle POLY_INT_CST.
2602         (size_int_kind): Take a poly_int64 instead of a HOST_WIDE_INT.
2603         * expmed.c (make_tree): Handle CONST_POLY_INT_P.
2604         * gimple-ssa-strength-reduction.c (slsr_process_add)
2605         (slsr_process_mul): Check for INTEGER_CSTs before using them
2606         as candidates.
2607         * stor-layout.c (bits_from_bytes): New function.
2608         (bit_from_pos): Use it.
2609         (layout_type): Likewise.  For vectors, multiply the TYPE_SIZE_UNIT
2610         by BITS_PER_UNIT to get the TYPE_SIZE.
2611         * tree-cfg.c (verify_expr, verify_types_in_gimple_reference): Allow
2612         MEM_REF and TARGET_MEM_REF offsets to be a POLY_INT_CST.
2614 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2615             Alan Hayward  <alan.hayward@arm.com>
2616             David Sherwood  <david.sherwood@arm.com>
2618         * doc/rtl.texi (const_poly_int): Document.  Also document the
2619         rtl sharing behavior.
2620         * gengenrtl.c (excluded_rtx): Return true for CONST_POLY_INT.
2621         * rtl.h (const_poly_int_def): New struct.
2622         (rtx_def::u): Add a cpi field.
2623         (CASE_CONST_UNIQUE, CASE_CONST_ANY): Add CONST_POLY_INT.
2624         (CONST_POLY_INT_P, CONST_POLY_INT_COEFFS): New macros.
2625         (wi::rtx_to_poly_wide_ref): New typedef
2626         (const_poly_int_value, wi::to_poly_wide, rtx_to_poly_int64)
2627         (poly_int_rtx_p): New functions.
2628         (trunc_int_for_mode): Declare a poly_int64 version.
2629         (plus_constant): Take a poly_int64 instead of a HOST_WIDE_INT.
2630         (immed_wide_int_const): Take a poly_wide_int_ref rather than
2631         a wide_int_ref.
2632         (strip_offset): Declare.
2633         (strip_offset_and_add): New function.
2634         * rtl.def (CONST_POLY_INT): New rtx code.
2635         * rtl.c (rtx_size): Handle CONST_POLY_INT.
2636         (shared_const_p): Use poly_int_rtx_p.
2637         * emit-rtl.h (gen_int_mode): Take a poly_int64 instead of a
2638         HOST_WIDE_INT.
2639         (gen_int_shift_amount): Likewise.
2640         * emit-rtl.c (const_poly_int_hasher): New class.
2641         (const_poly_int_htab): New variable.
2642         (init_emit_once): Initialize it when NUM_POLY_INT_COEFFS > 1.
2643         (const_poly_int_hasher::hash): New function.
2644         (const_poly_int_hasher::equal): Likewise.
2645         (gen_int_mode): Take a poly_int64 instead of a HOST_WIDE_INT.
2646         (immed_wide_int_const): Rename to...
2647         (immed_wide_int_const_1): ...this and make static.
2648         (immed_wide_int_const): New function, taking a poly_wide_int_ref
2649         instead of a wide_int_ref.
2650         (gen_int_shift_amount): Take a poly_int64 instead of a HOST_WIDE_INT.
2651         (gen_lowpart_common): Handle CONST_POLY_INT.
2652         * cse.c (hash_rtx_cb, equiv_constant): Likewise.
2653         * cselib.c (cselib_hash_rtx): Likewise.
2654         * dwarf2out.c (const_ok_for_output_1): Likewise.
2655         * expr.c (convert_modes): Likewise.
2656         * print-rtl.c (rtx_writer::print_rtx, print_value): Likewise.
2657         * rtlhash.c (add_rtx): Likewise.
2658         * explow.c (trunc_int_for_mode): Add a poly_int64 version.
2659         (plus_constant): Take a poly_int64 instead of a HOST_WIDE_INT.
2660         Handle existing CONST_POLY_INT rtxes.
2661         * expmed.h (expand_shift): Take a poly_int64 instead of a
2662         HOST_WIDE_INT.
2663         * expmed.c (expand_shift): Likewise.
2664         * rtlanal.c (strip_offset): New function.
2665         (commutative_operand_precedence): Give CONST_POLY_INT the same
2666         precedence as CONST_DOUBLE and put CONST_WIDE_INT between that
2667         and CONST_INT.
2668         * rtl-tests.c (const_poly_int_tests): New struct.
2669         (rtl_tests_c_tests): Use it.
2670         * simplify-rtx.c (simplify_const_unary_operation): Handle
2671         CONST_POLY_INT.
2672         (simplify_const_binary_operation): Likewise.
2673         (simplify_binary_operation_1): Fold additions of symbolic constants
2674         and CONST_POLY_INTs.
2675         (simplify_subreg): Handle extensions and truncations of
2676         CONST_POLY_INTs.
2677         (simplify_const_poly_int_tests): New struct.
2678         (simplify_rtx_c_tests): Use it.
2679         * wide-int.h (storage_ref): Add default constructor.
2680         (wide_int_ref_storage): Likewise.
2681         (trailing_wide_ints): Use GTY((user)).
2682         (trailing_wide_ints::operator[]): Add a const version.
2683         (trailing_wide_ints::get_precision): New function.
2684         (trailing_wide_ints::extra_size): Likewise.
2686 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2687             Alan Hayward  <alan.hayward@arm.com>
2688             David Sherwood  <david.sherwood@arm.com>
2690         * emit-rtl.h (gen_int_shift_amount): Declare.
2691         * emit-rtl.c (gen_int_shift_amount): New function.
2692         * asan.c (asan_emit_stack_protection): Use gen_int_shift_amount
2693         instead of GEN_INT.
2694         * calls.c (shift_return_value): Likewise.
2695         * cse.c (fold_rtx): Likewise.
2696         * dse.c (find_shift_sequence): Likewise.
2697         * expmed.c (init_expmed_one_mode, store_bit_field_1, expand_shift_1)
2698         (expand_shift, expand_smod_pow2): Likewise.
2699         * lower-subreg.c (shift_cost): Likewise.
2700         * optabs.c (expand_superword_shift, expand_doubleword_mult)
2701         (expand_unop, expand_binop, shift_amt_for_vec_perm_mask)
2702         (expand_vec_perm_var): Likewise.
2703         * simplify-rtx.c (simplify_unary_operation_1): Likewise.
2704         (simplify_binary_operation_1): Likewise.
2705         * combine.c (try_combine, find_split_point, force_int_to_mode)
2706         (simplify_shift_const_1, simplify_shift_const): Likewise.
2707         (change_zero_ext): Likewise.  Use simplify_gen_binary.
2709 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2711         * poly-int.h (multiple_p): Fix handling of two non-poly_ints.
2713 2017-12-20  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
2715         * doc/invoke.texi (ARM Options): Document accepted extension options
2716         for -march=armv8.3-a.
2718 2017-12-20  Richard Earnshaw  <rearnsha@arm.com>
2720         PR target/83105
2721         * config.gcc (arm*-*-linux*): When configured with --with-float=hard
2722         or --with-float=softfp, set the default CPU to arm10e.
2724 2017-12-20  Eric Botcazou  <ebotcazou@adacore.com>
2726         * config/visium/constraints.md (J, K, L): Use IN_RANGE macro.
2727         * config/visium/predicates.md (const_shift_operand): Likewise.
2728         * config/visium/visium.c (visium_legitimize_address): Fix oversight.
2729         (visium_legitimize_reload_address): Likewise.
2731 2017-12-20  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
2733         PR target/82975
2734         * config/arm/arm.h (TEST_REGNO): Adjust comment as expected in r255830.
2736 2017-12-20  Jakub Jelinek  <jakub@redhat.com>
2738         PR c++/83490
2739         * calls.c (compute_argument_addresses): Ignore TYPE_EMPTY_P arguments.
2741 2017-12-20  Julia Koval  <julia.koval@intel.com>
2743         * common/config/i386/i386-common.c (OPTION_MASK_ISA_VPCLMULQDQ_SET,
2744         OPTION_MASK_ISA_VPCLMULQDQ_UNSET): New.
2745         (ix86_handle_option): Handle -mvpclmulqdq, move cx6 to flags2.
2746         * config.gcc: Include vpclmulqdqintrin.h.
2747         * config/i386/cpuid.h: Handle bit_VPCLMULQDQ.
2748         * config/i386/driver-i386.c (host_detect_local_cpu): Handle
2749         -mvpclmulqdq.
2750         * config/i386/i386-builtin.def (__builtin_ia32_vpclmulqdq_v2di,
2751         __builtin_ia32_vpclmulqdq_v4di, __builtin_ia32_vpclmulqdq_v8di): New.
2752         * config/i386/i386-c.c (__VPCLMULQDQ__): New.
2753         * config/i386/i386.c (isa2_opts): Add -mcx16.
2754         (isa_opts): Add -mpclmulqdq, remove -mcx16.
2755         (ix86_option_override_internal): Move mcx16 to flags2.
2756         (ix86_valid_target_attribute_inner_p): Add vpclmulqdq.
2757         (ix86_expand_builtin): Handle OPTION_MASK_ISA_VPCLMULQDQ.
2758         * config/i386/i386.h (TARGET_VPCLMULQDQ, TARGET_VPCLMULQDQ_P): New.
2759         * config/i386/i386.opt: Add mvpclmulqdq, move mcx16 to flags2.
2760         * config/i386/immintrin.h: Include vpclmulqdqintrin.h.
2761         * config/i386/sse.md (vpclmulqdq_<mode>): New pattern.
2762         * config/i386/vpclmulqdqintrin.h (_mm512_clmulepi64_epi128,
2763         _mm_clmulepi64_epi128, _mm256_clmulepi64_epi128): New intrinsics.
2764         * doc/invoke.texi: Add -mvpclmulqdq.
2766 2017-12-20  Tom de Vries  <tom@codesourcery.com>
2768         PR middle-end/83423
2769         * config/i386/i386.c (ix86_static_chain): Move
2770         DECL_STATIC_CHAIN test ...
2771         * calls.c (rtx_for_static_chain): ... here.  New function.
2772         * calls.h (rtx_for_static_chain): Declare.
2773         * builtins.c (expand_builtin_setjmp_receiver): Use rtx_for_static_chain
2774         instead of targetm.calls.static_chain.
2775         * df-scan.c (df_get_entry_block_def_set): Same.
2777 2017-12-19  Tom de Vries  <tom@codesourcery.com>
2779         PR tree-optimization/83493
2780         * graphite-isl-ast-to-gimple.c (translate_isl_ast_node_for): Unshare ub
2781         and lb.
2783 2017-12-19  Jakub Jelinek  <jakub@redhat.com>
2785         * gimple-ssa-sprintf.c (format_directive): Use inform_n instead of
2786         inform with hardcoded english plural handling.
2788 2017-12-18  Jeff Law  <law@redhat.com>
2790         PR tree-optimization/83477
2791         * tree-ssa-threadedge.c (record_temporary_equivalences_from_phis): For
2792         a non-virtual PHI, always push a new range.
2794 2017-12-19  Martin Sebor  <msebor@redhat.com>
2796         PR middle-end/77608
2797         * builtins.c (compute_objsize): Handle non-constant offsets.
2799 2017-12-19  Jakub Jelinek  <jakub@redhat.com>
2801         PR tree-optimization/83444
2802         * tree-ssa-strlen.c (strlen_check_and_optimize_stmt): For the
2803         character load case, if get_stridx on MEM_REF's operand doesn't
2804         look usable, retry with get_addr_stridx.
2806 2017-12-19  Alexandre Oliva  <aoliva@redhat.com>
2808         PR debug/83422
2809         * var-tracking.c (vt_debug_insns_local): Do not drop markers.
2810         (variable_tracking_main_1): Keep markers even when VTA fails.
2812         PR bootstrap/83396
2813         * cfgexpand.c (expand_gimple_basic_block): Expand label first,
2814         even if there are markers before it.
2815         * cfgrtl.c (rtl_verify_bb_layout): Reject DEBUG_INSNs outside BBs.
2817 2017-12-19  Jakub Jelinek  <jakub@redhat.com>
2819         * read-rtl.c (parse_reg_note_name): Replace Yoda conditions with
2820         typical order conditions.
2821         * sel-sched.c (extract_new_fences_from): Likewise.
2822         * config/visium/constraints.md (J, K, L): Likewise.
2823         * config/visium/predicates.md (const_shift_operand): Likewise.
2824         * config/visium/visium.c (visium_legitimize_address,
2825         visium_legitimize_reload_address): Likewise.
2826         * config/m68k/m68k.c (output_reg_adjust, emit_reg_adjust): Likewise.
2827         * config/arm/arm.c (arm_block_move_unaligned_straight): Likewise.
2828         * config/avr/constraints.md (Y01, Ym1, Y02, Ym2): Likewise.
2829         * config/avr/avr-log.c (avr_vdump, avr_log_set_avr_log,
2830         SET_DUMP_DETAIL): Likewise.
2831         * config/avr/predicates.md (const_8_16_24_operand): Likewise.
2832         * config/avr/avr.c (STR_PREFIX_P, avr_popcount_each_byte,
2833         avr_is_casesi_sequence, avr_casei_sequence_check_operands,
2834         avr_set_core_architecture, avr_set_current_function,
2835         avr_legitimize_reload_address, avr_asm_len, avr_print_operand,
2836         output_movqi, output_movsisf, avr_out_plus, avr_out_bitop,
2837         avr_out_fract, avr_adjust_insn_length, avr_encode_section_info,
2838         avr_2word_insn_p, output_reload_in_const, avr_has_nibble_0xf,
2839         avr_map_decompose, avr_fold_builtin): Likewise.
2840         * config/avr/driver-avr.c (avr_devicespecs_file): Likewise.
2841         * config/avr/gen-avr-mmcu-specs.c (str_prefix_p, print_mcu): Likewise.
2842         * config/i386/i386.c (ix86_parse_stringop_strategy_string): Likewise.
2843         * config/m32c/m32c-pragma.c (m32c_pragma_memregs): Likewise.
2844         * config/m32c/m32c.c (m32c_conditional_register_usage,
2845         m32c_address_cost): Likewise.
2846         * config/m32c/predicates.md (shiftcount_operand,
2847         longshiftcount_operand): Likewise.
2848         * config/iq2000/iq2000.c (iq2000_expand_prologue): Likewise.
2849         * config/nios2/nios2.c (nios2_handle_custom_fpu_insn_option,
2850         can_use_cdx_ldstw): Likewise.
2851         * config/nios2/nios2.h (CDX_REG_P): Likewise.
2852         * config/cr16/cr16.h (RETURN_ADDR_RTX, REGNO_MODE_OK_FOR_BASE_P):
2853         Likewise.
2854         * config/cr16/cr16.md (*mov<mode>_double): Likewise.
2855         * config/cr16/cr16.c (cr16_create_dwarf_for_multi_push): Likewise.
2856         * config/h8300/h8300.c (h8300_rtx_costs, get_shift_alg): Likewise.
2857         * config/vax/constraints.md (U06, U08, U16, CN6, S08, S16): Likewise.
2858         * config/vax/vax.c (adjacent_operands_p): Likewise.
2859         * config/ft32/constraints.md (L, b, KA): Likewise.
2860         * config/ft32/ft32.c (ft32_load_immediate, ft32_expand_prologue):
2861         Likewise.
2862         * cfgexpand.c (expand_stack_alignment): Likewise.
2863         * gcse.c (insert_expr_in_table): Likewise.
2864         * print-rtl.c (rtx_writer::print_rtx_operand_codes_E_and_V): Likewise.
2865         * cgraphunit.c (cgraph_node::expand): Likewise.
2866         * ira-build.c (setup_min_max_allocno_live_range_point): Likewise.
2867         * emit-rtl.c (add_insn): Likewise.
2868         * input.c (dump_location_info): Likewise.
2869         * passes.c (NEXT_PASS): Likewise.
2870         * read-rtl-function.c (parse_note_insn_name,
2871         function_reader::read_rtx_operand_r, function_reader::parse_mem_expr):
2872         Likewise.
2873         * sched-rgn.c (sched_rgn_init): Likewise.
2874         * diagnostic-show-locus.c (layout::show_ruler): Likewise.
2875         * combine.c (find_split_point, simplify_if_then_else, force_to_mode,
2876         if_then_else_cond, simplify_shift_const_1, simplify_comparison):
2877         Likewise.
2878         * explow.c (eliminate_constant_term): Likewise.
2879         * final.c (leaf_renumber_regs_insn): Likewise.
2880         * cfgrtl.c (print_rtl_with_bb): Likewise.
2881         * genhooks.c (emit_init_macros): Likewise.
2882         * poly-int.h (maybe_ne, maybe_le, maybe_lt): Likewise.
2883         * tree-data-ref.c (conflict_fn): Likewise.
2884         * selftest.c (assert_streq): Likewise.
2885         * expr.c (store_constructor_field, expand_expr_real_1): Likewise.
2886         * fold-const.c (fold_range_test, extract_muldiv_1, fold_truth_andor,
2887         fold_binary_loc, multiple_of_p): Likewise.
2888         * reload.c (push_reload, find_equiv_reg): Likewise.
2889         * et-forest.c (et_nca, et_below): Likewise.
2890         * dbxout.c (dbxout_symbol_location): Likewise.
2891         * reorg.c (relax_delay_slots): Likewise.
2892         * dojump.c (do_compare_rtx_and_jump): Likewise.
2893         * gengtype-parse.c (type): Likewise.
2894         * simplify-rtx.c (simplify_gen_ternary, simplify_gen_relational,
2895         simplify_const_relational_operation): Likewise.
2896         * reload1.c (do_output_reload): Likewise.
2897         * dumpfile.c (get_dump_file_info_by_switch): Likewise.
2898         * gengtype.c (type_for_name): Likewise.
2899         * gimple-ssa-sprintf.c (format_directive): Likewise.
2901 2017-12-19  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
2903         PR target/82975
2904         * config/arm/arm.h (TEST_REGNO): Check reg_renumber is set before
2905         accessing it.  Adjust comment.
2907 2017-12-19  Jakub Jelinek  <jakub@redhat.com>
2909         PR middle-end/81914
2910         * predict.c (zero_one_minusone): New function.
2911         (apply_return_prediction): Avoid return prediction for functions
2912         returning only -1, 0 and 1 values, unless they only return -1 and 0
2913         or 0 and 1.
2915 2017-12-19  Claudiu Zissulescu  <claziss@synopsys.com>
2917         * config/arc/arc.c (legitimate_scaled_address_p): Clean
2918         fall-through warning.
2919         (arc_compute_frame_size): Remove unused variables.
2920         (arc_print_operand): Fix fprintif format.
2921         (arc_can_follow_jump): Clean fall-through warning.
2923 2017-12-19  Marek Polacek  <polacek@redhat.com>
2925         PR c++/83489
2926         * config/i386/i386.c (init_cumulative_args): Don't check TYPE_EMPTY_P
2927         on an error node.
2929 2017-12-19  Claudiu Zissulescu  <claziss@synopsys.com>
2931         * config/arc/arc.c (overriderregs): New variable.
2932         (arc_override_options): Track fixed/call saved/call options.
2933         (arc_conditional_register_usage): Check against overrideregs
2934         variable whenever we change register properties.
2936 2017-12-19  Nathan Sidwell  <nathan@acm.org>
2938         * opts.c (finish_options): Don't prefix dump_base_name if it
2939         already contains directories.
2941 2017-12-19  Martin Liska  <mliska@suse.cz>
2943         PR rtl-optimization/82675
2944         * loop-unroll.c (unroll_loop_constant_iterations): Allocate one
2945         more element in sbitmap.
2947 2017-12-19  Paolo Carlini  <paolo.carlini@oracle.com>
2949         * gimplify.c (gimplify_expr): Use error_operand_p.
2951 2017-12-19  Sebastian Huber  <sebastian.huber@embedded-brains.de>
2953         PR target/83387
2954         * config/rs6000/t-rtems (MULTILIB_REQUIRED): Remove 64-bit soft-float
2955         multilib.
2957 2017-12-19  Daniel Cederman  <cederman@gaisler.com>
2959         * config/sparc/sparc.c (sparc_do_work_around_errata): Make sure
2960         the jump is to a label.
2962 2017-12-19  Jakub Jelinek  <jakub@redhat.com>
2964         PR tree-optimization/83444
2965         * tree-ssa-strlen.c (strlen_check_and_optimize_stmt): Optimize
2966         character loads.
2968         PR ipa/82801
2969         PR ipa/83346
2970         * ipa-inline.c (flatten_remove_node_hook): New function.
2971         (ipa_inline): Keep only nodes with flatten attribute at the end of
2972         the array in the order from ipa_reverse_postorder, only walk that
2973         portion of array for flattening, if there is more than one such
2974         node, temporarily register a removal hook and ignore removed nodes.
2976         PR tree-optimization/80631
2977         * tree-vect-loop.c (vect_create_epilog_for_reduction): Compare
2978         induc_code against MAX_EXPR or MIN_EXPR instead of reduc_fn against
2979         IFN_REDUC_MAX or IFN_REDUC_MIN.
2981 2017-12-18  Jeff Law  <law@redhat.com>
2983         * tree-ssa-dom.c (record_equivalences_from_phis): Do not
2984         record symbolic equivalences from backedges in the CFG.
2986         Revert
2987         2017-11-19  Jeff Law  <law@redhat.com>
2989         * tree-ssa-dom.c (record_equivalences_from_phis): Fix handling
2990         of degenerates resulting from ignoring an edge.
2992 2017-12-18  Martin Sebor  <msebor@redhat.com>
2994         PR middle-end/83373
2995         PR tree-optimization/78450
2996         * tree-ssa-strlen.c (maybe_set_strlen_range): New function.
2997         (handle_builtin_strlen): Call it.
2999 2017-12-18  Segher Boessenkool  <segher@kernel.crashing.org>
3001         PR rtl-optimization/83424
3002         * rtlanal.c (dead_or_set_regno_p): Handle CLOBBER just like SET.
3004 2017-12-18  Marek Polacek  <polacek@redhat.com>
3006         PR middle-end/83463
3007         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref):
3008         Check if TYPE is INTEGRAL_TYPE_P before accessing its min/max
3009         values.
3011 2017-12-18  Claudiu Zissulescu  <claziss@synopsys.com>
3013         * config/arc/arc.md (maddsidi4, maddsidi4_split): Update pattern.
3014         (umaddsidi4, umaddsidi_split): Likewise.
3016 2017-12-18  Claudiu Zissulescu  <claziss@synopsys.com>
3018         * config/arc/arc.c (arc_legitimate_constant_p): Always check all
3019         constants.
3021 2017-12-18  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
3023         PR target/83420
3024         * config/s390/s390.c (s390_option_override): Avoid strncpy.
3026 2017-12-18  Richard Biener  <rguenther@suse.de>
3028         PR tree-optimization/81877
3029         * tree-ssa-loop-im.c (ref_indep_loop_p): Remove safelen parameters.
3030         (outermost_indep_loop): Adjust.
3031         (ref_indep_loop_p_1): Likewise.  Remove safelen handling again.
3032         (can_sm_ref_p): Adjust.
3034 2017-12-18  Richard Biener  <rguenther@suse.de>
3036         PR middle-end/77291
3037         * tree.c (array_at_struct_end_p): Return true if the underlying
3038         object has space for at least one element in excess of what
3039         the array domain specifies.
3041 2017-12-17  Sandra Loosemore  <sandra@codesourcery.com>
3043         * doc/extend.texi (x86 Function Attributes): Reformat nocf_check
3044         example to avoid overfull hbox.
3045         * doc/invoke.texi (Option Summary): Add missing @gol.
3046         (C++ Dialect Options): Reformat -Wnoexcept-type example to avoid
3047         overfull hbox.
3049 2017-12-17  Sandra Loosemore  <sandra@codesourcery.com>
3050             Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
3052         * doc/invoke.texi (Option Summary): Add -mverbose-cost-dump
3053         to AArch64 and ARM lists, plus missing -mflip-thumb for ARM.
3054         (AArch64 Options): Document -mverbose-cost-dump.
3055         (ARM Options): Likewise, plus -mflip-thumb.
3057 2017-12-17  Martin Sebor  <msebor@redhat.com>
3059         PR bootstrap/83446
3060         * gimple-ssa-warn-restrict.c
3061         (builtin_memref::offset_out_of_bounds): Correct the handling of
3062         anti-ranges.
3064 2017-12-17  Sandra Loosemore  <sandra@codesourcery.com>
3066         * doc/invoke.texi (ARC Options): Add missing -mlra entry.
3068 2017-12-17  John David Anglin  <danglin@gcc.gnu.org>
3070         * config/pa/pa.c (pa_som_asm_init_sections): Fix comment.
3072 2017-12-17  Markus Trippelsdorf  <markus@trippelsdorf.de>
3074         * x86-tune-costs.h (skylake_cost, core_cost): Decrease r64 multiply
3075         latencies.
3077 2017-12-16  Sandra Loosemore  <sandra@codesourcery.com>
3079         * doc/invoke.texi: Fix some typos.
3081 2017-12-16  Martin Sebor  <msebor@redhat.com>
3083         PR tree-optimization/78918
3084         * Makefile.in (OBJS): Add gimple-ssa-warn-restrict.o.
3085         * builtins.c (check_sizes): Rename...
3086         (check_access): ...to this.  Rename function arguments for clarity.
3087         (check_memop_sizes): Adjust names.
3088         (expand_builtin_memchr, expand_builtin_memcpy): Same.
3089         (expand_builtin_memmove, expand_builtin_mempcpy): Same.
3090         (expand_builtin_strcat, expand_builtin_stpncpy): Same.
3091         (check_strncat_sizes, expand_builtin_strncat): Same.
3092         (expand_builtin_strncpy, expand_builtin_memset): Same.
3093         (expand_builtin_bzero, expand_builtin_memcmp): Same.
3094         (expand_builtin_memory_chk, maybe_emit_chk_warning): Same.
3095         (maybe_emit_sprintf_chk_warning): Same.
3096         (expand_builtin_strcpy): Adjust.
3097         (expand_builtin_stpcpy): Same.
3098         (expand_builtin_with_bounds): Detect out-of-bounds accesses
3099         in pointer-checking forms of memcpy, memmove, and mempcpy.
3100         (gcall_to_tree_minimal, max_object_size): Define new functions.
3101         * builtins.h (max_object_size): Declare.
3102         * calls.c (alloc_max_size): Call max_object_size instead of
3103         hardcoding ssizetype limit.
3104         (get_size_range): Handle new argument.
3105         * calls.h (get_size_range): Add a new argument.
3106         * cfgexpand.c (expand_call_stmt): Propagate no-warning bit.
3107         * doc/invoke.texi (-Wrestrict): Adjust, add example.
3108         * gimple-fold.c (gimple_fold_builtin_memory_op): Detect overlapping
3109         operations.
3110         (gimple_fold_builtin_memory_chk): Same.
3111         (gimple_fold_builtin_stxcpy_chk): New function.
3112         * gimple-ssa-warn-restrict.c: New source.
3113         * gimple-ssa-warn-restrict.h: New header.
3114         * gimple.c (gimple_build_call_from_tree): Propagate location.
3115         * passes.def (pass_warn_restrict): Add new pass.
3116         * tree-pass.h (make_pass_warn_restrict): Declare.
3117         * tree-ssa-strlen.c (handle_builtin_strcpy): Detect overlapping
3118         operations.
3119         (handle_builtin_strcat): Same.
3120         (strlen_optimize_stmt): Rename...
3121         (strlen_check_and_optimize_stmt): ...to this.  Handle strncat,
3122         stpncpy, strncpy, and their checking forms.
3124 2017-12-16  Jan Hubicka  <hubicka@ucw.cz>
3126         PR rtl-optimization/82849
3127         * modulo-sched.c (sms_schedule): Use get_estimated_loop_iterations_int
3128         and get_max_loop_iterations_int.
3130 2017-12-16  Richard Sandiford  <richard.sandiford@linaro.org>
3131             Alan Hayward  <alan.hayward@arm.com>
3132             David Sherwood  <david.sherwood@arm.com>
3134         * machmode.h (mode_for_size, int_mode_for_size, float_mode_for_size)
3135         (smallest_mode_for_size, smallest_int_mode_for_size): Take the mode
3136         size as a poly_uint64.
3137         (mode_for_vector, mode_for_int_vector): Take the number of vector
3138         elements as a poly_uint64.
3139         * stor-layout.c (mode_for_size, smallest_mode_for_size): Take the mode
3140         size as a poly_uint64.
3141         (mode_for_vector, mode_for_int_vector): Take the number of vector
3142         elements as a poly_uint64.
3144 2017-12-16  Richard Sandiford  <richard.sandiford@linaro.org>
3145             Alan Hayward  <alan.hayward@arm.com>
3146             David Sherwood  <david.sherwood@arm.com>
3148         * machmode.h (MACRO_MODE): New macro.
3149         * addresses.h (base_reg_class, ok_for_base_p_1): Use it.
3150         * targhooks.c (default_libcall_value, default_secondary_reload)
3151         (default_memory_move_cost, default_register_move_cost)
3152         (default_class_max_nregs): Likewise.
3154 2017-12-16  Richard Sandiford  <richard.sandiford@linaro.org>
3155             Alan Hayward  <alan.hayward@arm.com>
3156             David Sherwood  <david.sherwood@arm.com>
3158         * doc/sourcebuild.texi: Document IN_TARGET_CODE.
3159         * genattrtab.c (write_header): Define IN_TARGET_CODE to 1 in the
3160         target C file.
3161         * genautomata.c (main): Likewise.
3162         * genconditions.c (write_header): Likewise.
3163         * genemit.c (main): Likewise.
3164         * genextract.c (print_header): Likewise.
3165         * genopinit.c (main): Likewise.
3166         * genoutput.c (output_prologue): Likewise.
3167         * genpeep.c (main): Likewise.
3168         * genpreds.c (write_insn_preds_c): Likewise.
3169         * genrecog.c (writer_header): Likewise.
3170         * config/aarch64/aarch64-builtins.c (IN_TARGET_CODE): Define.
3171         * config/aarch64/aarch64-c.c (IN_TARGET_CODE): Likewise.
3172         * config/aarch64/aarch64.c (IN_TARGET_CODE): Likewise.
3173         * config/aarch64/cortex-a57-fma-steering.c (IN_TARGET_CODE): Likewise.
3174         * config/aarch64/driver-aarch64.c (IN_TARGET_CODE): Likewise.
3175         * config/alpha/alpha.c (IN_TARGET_CODE): Likewise.
3176         * config/alpha/driver-alpha.c (IN_TARGET_CODE): Likewise.
3177         * config/arc/arc-c.c (IN_TARGET_CODE): Likewise.
3178         * config/arc/arc.c (IN_TARGET_CODE): Likewise.
3179         * config/arc/driver-arc.c (IN_TARGET_CODE): Likewise.
3180         * config/arm/aarch-common.c (IN_TARGET_CODE): Likewise.
3181         * config/arm/arm-builtins.c (IN_TARGET_CODE): Likewise.
3182         * config/arm/arm-c.c (IN_TARGET_CODE): Likewise.
3183         * config/arm/arm.c (IN_TARGET_CODE): Likewise.
3184         * config/arm/driver-arm.c (IN_TARGET_CODE): Likewise.
3185         * config/avr/avr-c.c (IN_TARGET_CODE): Likewise.
3186         * config/avr/avr-devices.c (IN_TARGET_CODE): Likewise.
3187         * config/avr/avr-log.c (IN_TARGET_CODE): Likewise.
3188         * config/avr/avr.c (IN_TARGET_CODE): Likewise.
3189         * config/avr/driver-avr.c (IN_TARGET_CODE): Likewise.
3190         * config/avr/gen-avr-mmcu-specs.c (IN_TARGET_CODE): Likewise.
3191         * config/bfin/bfin.c (IN_TARGET_CODE): Likewise.
3192         * config/c6x/c6x.c (IN_TARGET_CODE): Likewise.
3193         * config/cr16/cr16.c (IN_TARGET_CODE): Likewise.
3194         * config/cris/cris.c (IN_TARGET_CODE): Likewise.
3195         * config/darwin.c (IN_TARGET_CODE): Likewise.
3196         * config/epiphany/epiphany.c (IN_TARGET_CODE): Likewise.
3197         * config/epiphany/mode-switch-use.c (IN_TARGET_CODE): Likewise.
3198         * config/epiphany/resolve-sw-modes.c (IN_TARGET_CODE): Likewise.
3199         * config/fr30/fr30.c (IN_TARGET_CODE): Likewise.
3200         * config/frv/frv.c (IN_TARGET_CODE): Likewise.
3201         * config/ft32/ft32.c (IN_TARGET_CODE): Likewise.
3202         * config/h8300/h8300.c (IN_TARGET_CODE): Likewise.
3203         * config/i386/djgpp.c (IN_TARGET_CODE): Likewise.
3204         * config/i386/driver-i386.c (IN_TARGET_CODE): Likewise.
3205         * config/i386/driver-mingw32.c (IN_TARGET_CODE): Likewise.
3206         * config/i386/host-cygwin.c (IN_TARGET_CODE): Likewise.
3207         * config/i386/host-i386-darwin.c (IN_TARGET_CODE): Likewise.
3208         * config/i386/host-mingw32.c (IN_TARGET_CODE): Likewise.
3209         * config/i386/i386-c.c (IN_TARGET_CODE): Likewise.
3210         * config/i386/i386.c (IN_TARGET_CODE): Likewise.
3211         * config/i386/intelmic-mkoffload.c (IN_TARGET_CODE): Likewise.
3212         * config/i386/msformat-c.c (IN_TARGET_CODE): Likewise.
3213         * config/i386/winnt-cxx.c (IN_TARGET_CODE): Likewise.
3214         * config/i386/winnt-stubs.c (IN_TARGET_CODE): Likewise.
3215         * config/i386/winnt.c (IN_TARGET_CODE): Likewise.
3216         * config/i386/x86-tune-sched-atom.c (IN_TARGET_CODE): Likewise.
3217         * config/i386/x86-tune-sched-bd.c (IN_TARGET_CODE): Likewise.
3218         * config/i386/x86-tune-sched-core.c (IN_TARGET_CODE): Likewise.
3219         * config/i386/x86-tune-sched.c (IN_TARGET_CODE): Likewise.
3220         * config/ia64/ia64-c.c (IN_TARGET_CODE): Likewise.
3221         * config/ia64/ia64.c (IN_TARGET_CODE): Likewise.
3222         * config/iq2000/iq2000.c (IN_TARGET_CODE): Likewise.
3223         * config/lm32/lm32.c (IN_TARGET_CODE): Likewise.
3224         * config/m32c/m32c-pragma.c (IN_TARGET_CODE): Likewise.
3225         * config/m32c/m32c.c (IN_TARGET_CODE): Likewise.
3226         * config/m32r/m32r.c (IN_TARGET_CODE): Likewise.
3227         * config/m68k/m68k.c (IN_TARGET_CODE): Likewise.
3228         * config/mcore/mcore.c (IN_TARGET_CODE): Likewise.
3229         * config/microblaze/microblaze-c.c (IN_TARGET_CODE): Likewise.
3230         * config/microblaze/microblaze.c (IN_TARGET_CODE): Likewise.
3231         * config/mips/driver-native.c (IN_TARGET_CODE): Likewise.
3232         * config/mips/frame-header-opt.c (IN_TARGET_CODE): Likewise.
3233         * config/mips/mips.c (IN_TARGET_CODE): Likewise.
3234         * config/mmix/mmix.c (IN_TARGET_CODE): Likewise.
3235         * config/mn10300/mn10300.c (IN_TARGET_CODE): Likewise.
3236         * config/moxie/moxie.c (IN_TARGET_CODE): Likewise.
3237         * config/msp430/driver-msp430.c (IN_TARGET_CODE): Likewise.
3238         * config/msp430/msp430-c.c (IN_TARGET_CODE): Likewise.
3239         * config/msp430/msp430.c (IN_TARGET_CODE): Likewise.
3240         * config/nds32/nds32-cost.c (IN_TARGET_CODE): Likewise.
3241         * config/nds32/nds32-fp-as-gp.c (IN_TARGET_CODE): Likewise.
3242         * config/nds32/nds32-intrinsic.c (IN_TARGET_CODE): Likewise.
3243         * config/nds32/nds32-isr.c (IN_TARGET_CODE): Likewise.
3244         * config/nds32/nds32-md-auxiliary.c (IN_TARGET_CODE): Likewise.
3245         * config/nds32/nds32-memory-manipulation.c (IN_TARGET_CODE): Likewise.
3246         * config/nds32/nds32-pipelines-auxiliary.c (IN_TARGET_CODE): Likewise.
3247         * config/nds32/nds32-predicates.c (IN_TARGET_CODE): Likewise.
3248         * config/nds32/nds32.c (IN_TARGET_CODE): Likewise.
3249         * config/nios2/nios2.c (IN_TARGET_CODE): Likewise.
3250         * config/nvptx/mkoffload.c (IN_TARGET_CODE): Likewise.
3251         * config/nvptx/nvptx.c (IN_TARGET_CODE): Likewise.
3252         * config/pa/pa.c (IN_TARGET_CODE): Likewise.
3253         * config/pdp11/pdp11.c (IN_TARGET_CODE): Likewise.
3254         * config/powerpcspe/driver-powerpcspe.c (IN_TARGET_CODE): Likewise.
3255         * config/powerpcspe/host-darwin.c (IN_TARGET_CODE): Likewise.
3256         * config/powerpcspe/host-ppc64-darwin.c (IN_TARGET_CODE): Likewise.
3257         * config/powerpcspe/powerpcspe-c.c (IN_TARGET_CODE): Likewise.
3258         * config/powerpcspe/powerpcspe-linux.c (IN_TARGET_CODE): Likewise.
3259         * config/powerpcspe/powerpcspe.c (IN_TARGET_CODE): Likewise.
3260         * config/riscv/riscv-builtins.c (IN_TARGET_CODE): Likewise.
3261         * config/riscv/riscv-c.c (IN_TARGET_CODE): Likewise.
3262         * config/riscv/riscv.c (IN_TARGET_CODE): Likewise.
3263         * config/rl78/rl78-c.c (IN_TARGET_CODE): Likewise.
3264         * config/rl78/rl78.c (IN_TARGET_CODE): Likewise.
3265         * config/rs6000/driver-rs6000.c (IN_TARGET_CODE): Likewise.
3266         * config/rs6000/host-darwin.c (IN_TARGET_CODE): Likewise.
3267         * config/rs6000/host-ppc64-darwin.c (IN_TARGET_CODE): Likewise.
3268         * config/rs6000/rs6000-c.c (IN_TARGET_CODE): Likewise.
3269         * config/rs6000/rs6000-linux.c (IN_TARGET_CODE): Likewise.
3270         * config/rs6000/rs6000-p8swap.c (IN_TARGET_CODE): Likewise.
3271         * config/rs6000/rs6000-string.c (IN_TARGET_CODE): Likewise.
3272         * config/rs6000/rs6000.c (IN_TARGET_CODE): Likewise.
3273         * config/rx/rx.c (IN_TARGET_CODE): Likewise.
3274         * config/s390/driver-native.c (IN_TARGET_CODE): Likewise.
3275         * config/s390/s390-c.c (IN_TARGET_CODE): Likewise.
3276         * config/s390/s390.c (IN_TARGET_CODE): Likewise.
3277         * config/sh/sh-c.c (IN_TARGET_CODE): Likewise.
3278         * config/sh/sh-mem.cc (IN_TARGET_CODE): Likewise.
3279         * config/sh/sh.c (IN_TARGET_CODE): Likewise.
3280         * config/sh/sh_optimize_sett_clrt.cc (IN_TARGET_CODE): Likewise.
3281         * config/sh/sh_treg_combine.cc (IN_TARGET_CODE): Likewise.
3282         * config/sparc/driver-sparc.c (IN_TARGET_CODE): Likewise.
3283         * config/sparc/sparc-c.c (IN_TARGET_CODE): Likewise.
3284         * config/sparc/sparc.c (IN_TARGET_CODE): Likewise.
3285         * config/spu/spu-c.c (IN_TARGET_CODE): Likewise.
3286         * config/spu/spu.c (IN_TARGET_CODE): Likewise.
3287         * config/stormy16/stormy16.c (IN_TARGET_CODE): Likewise.
3288         * config/tilegx/mul-tables.c (IN_TARGET_CODE): Likewise.
3289         * config/tilegx/tilegx-c.c (IN_TARGET_CODE): Likewise.
3290         * config/tilegx/tilegx.c (IN_TARGET_CODE): Likewise.
3291         * config/tilepro/mul-tables.c (IN_TARGET_CODE): Likewise.
3292         * config/tilepro/tilepro-c.c (IN_TARGET_CODE): Likewise.
3293         * config/tilepro/tilepro.c (IN_TARGET_CODE): Likewise.
3294         * config/v850/v850-c.c (IN_TARGET_CODE): Likewise.
3295         * config/v850/v850.c (IN_TARGET_CODE): Likewise.
3296         * config/vax/vax.c (IN_TARGET_CODE): Likewise.
3297         * config/visium/visium.c (IN_TARGET_CODE): Likewise.
3298         * config/vms/vms-c.c (IN_TARGET_CODE): Likewise.
3299         * config/vms/vms-f.c (IN_TARGET_CODE): Likewise.
3300         * config/vms/vms.c (IN_TARGET_CODE): Likewise.
3301         * config/xtensa/xtensa.c (IN_TARGET_CODE): Likewise.
3303 2017-12-16  Richard Sandiford  <richard.sandiford@linaro.org>
3304             Alan Hayward  <alan.hayward@arm.com>
3305             David Sherwood  <david.sherwood@arm.com>
3307         * tree-ssa-sccvn.c (vn_reference_lookup_3): Avoid repeated
3308         checks for MEM_REF.
3310 2017-12-16  Richard Sandiford  <richard.sandiford@linaro.org>
3311             Alan Hayward  <alan.hayward@arm.com>
3312             David Sherwood  <david.sherwood@arm.com>
3314         * doc/generic.texi (VEC_SERIES_EXPR): Document.
3315         * doc/md.texi (vec_series@var{m}): Document.
3316         * tree.def (VEC_SERIES_EXPR): New tree code.
3317         * tree.h (build_vec_series): Declare.
3318         * tree.c (build_vec_series): New function.
3319         * cfgexpand.c (expand_debug_expr): Handle VEC_SERIES_EXPR.
3320         * tree-pretty-print.c (dump_generic_node): Likewise.
3321         * gimple-pretty-print.c (dump_binary_rhs): Likewise.
3322         * tree-inline.c (estimate_operator_cost): Likewise.
3323         * expr.c (expand_expr_real_2): Likewise.
3324         * optabs-tree.c (optab_for_tree_code): Likewise.
3325         * tree-cfg.c (verify_gimple_assign_binary): Likewise.
3326         * fold-const.c (const_binop): Fold VEC_SERIES_EXPRs of constants.
3327         * expmed.c (make_tree): Handle VEC_SERIES.
3328         * optabs.def (vec_series_optab): New optab.
3329         * optabs.h (expand_vec_series_expr): Declare.
3330         * optabs.c (expand_vec_series_expr): New function.
3331         * tree-vect-generic.c (expand_vector_operations_1): Check that
3332         the operands also have vector type.
3334 2017-12-16  Richard Sandiford  <richard.sandiford@linaro.org>
3335             Alan Hayward  <alan.hawyard@arm.com>
3336             David Sherwood  <david.sherwood@arm.com>
3338         * doc/generic.texi (VEC_DUPLICATE_EXPR): Document.
3339         (VEC_COND_EXPR): Add missing @tindex.
3340         * doc/md.texi (vec_duplicate@var{m}): Document.
3341         * tree.def (VEC_DUPLICATE_EXPR): New tree codes.
3342         * tree.c (build_vector_from_val): Add stubbed-out handling of
3343         variable-length vectors, using VEC_DUPLICATE_EXPR.
3344         (uniform_vector_p): Handle VEC_DUPLICATE_EXPR.
3345         * cfgexpand.c (expand_debug_expr): Likewise.
3346         * tree-cfg.c (verify_gimple_assign_unary): Likewise.
3347         * tree-inline.c (estimate_operator_cost): Likewise.
3348         * tree-pretty-print.c (dump_generic_node): Likewise.
3349         * tree-vect-generic.c (ssa_uniform_vector_p): Likewise.
3350         * fold-const.c (const_unop): Fold VEC_DUPLICATE_EXPRs of a constant.
3351         (test_vec_duplicate_folding): New function.
3352         (fold_const_c_tests): Call it.
3353         * optabs.def (vec_duplicate_optab): New optab.
3354         * optabs-tree.c (optab_for_tree_code): Handle VEC_DUPLICATE_EXPR.
3355         * optabs.h (expand_vector_broadcast): Declare.
3356         * optabs.c (expand_vector_broadcast): Make non-static.  Try using
3357         vec_duplicate_optab.
3358         * expr.c (store_constructor): Try using vec_duplicate_optab for
3359         uniform vectors.
3360         (expand_expr_real_2): Handle VEC_DUPLICATE_EXPR.
3362 2017-12-15  Markus Trippelsdorf  <markus@trippelsdorf.de>
3364         PR target/83358
3365         * config/i386/x86-tune-costs.h (skylake_cost, core_cost): Increase
3366         div/mod latencies a bit.
3368 2017-12-15  Jeff Law  <law@redhat.com>
3370         PR tree-optimization/36550
3371         * tree-ssa-threadupdate.c (count_stmts_and_phis_in_block): New.
3372         (mark_threaded_blocks): Rewrite code to avoid block copying when
3373         optimizing for size.  Don't pessimize blocks which will be
3374         copied, but all the statements will be dead.
3376 2017-12-15  Alexandre Oliva  <aoliva@redhat.com>
3378         PR tree-optimization/81165
3379         * tree-ssa-threadupdate.c (uses_in_bb): New.
3380         (estimate_threading_killed_stmts): New.
3381         * tree-ssa-threadupdate.h (estimate_threading_killed_stmts): Prototype.
3382         * tree-ssa-threadedge.c
3383         (record_temporary_equivalences_from_stmts_at_dest): Expand limit
3384         when its hit.
3386 2017-12-15  Jeff Law  <law@redhat.com>
3388         PR tree-optimization/83410
3389         * tree-ssa-threadupdate.c (thread_block_1): Avoid certain jump
3390         threads when parallelizing loops.
3392 2017-12-15  Jakub Jelinek  <jakub@redhat.com>
3394         * tree-core.h (struct attribute_spec): Swap affects_type_identity and
3395         handler fields.
3396         * config/alpha/alpha.c (vms_attribute_table): Swap
3397         affects_type_identity and handler fields, adjust comments.
3398         * config/mips/mips.c (mips_attribute_table): Likewise.
3399         * config/visium/visium.c (visium_attribute_table): Likewise.
3400         * config/epiphany/epiphany.c (epiphany_attribute_table): Likewise.
3401         * config/microblaze/microblaze.c (microblaze_attribute_table):
3402         Likewise.
3403         * config/spu/spu.c (spu_attribute_table): Likewise.
3404         * config/mcore/mcore.c (mcore_attribute_table): Likewise.
3405         * config/arc/arc.c (arc_attribute_table): Likewise.
3406         * config/m68k/m68k.c (m68k_attribute_table): Likewise.
3407         * config/v850/v850.c (v850_handle_interrupt_attribute,
3408         v850_handle_data_area_attribute): Formatting fixes.
3409         (v850_attribute_table): Swap affects_type_identity and handler
3410         fields, adjust comments.
3411         * config/m32r/m32r.c (m32r_attribute_table): Likewise.
3412         * config/arm/arm.c (arm_attribute_table): Likewise.
3413         * config/avr/avr.c (avr_attribute_table): Likewise.
3414         * config/s390/s390.c (s390_attribute_table): Likewise.
3415         * config/sh/sh.c (sh_attribute_table): Likewise.
3416         * config/i386/i386.c (ix86_handle_cconv_attribute,
3417         ix86_handle_callee_pop_aggregate_return): Formatting fixes.
3418         (ix86_attribute_table): Swap affects_type_identity and handler
3419         fields, adjust comments.
3420         * config/i386/cygming.h (SUBTARGET_ATTRIBUTE_TABLE): Likewise.
3421         * config/sparc/sparc.c (sparc_attribute_table): Likewise.
3422         * config/m32c/m32c.c (m32c_attribute_table): Likewise.
3423         * config/sol2.h (SOLARIS_ATTRIBUTE_TABLE): Likewise.
3424         * config/ia64/ia64.c (ia64_attribute_table): Likewise.
3425         * config/msp430/msp430.c (msp430_attribute_table): Likewise.
3426         * config/rx/rx.c (rx_attribute_table): Likewise.
3427         * config/cr16/cr16.c (cr16_attribute_table): Likewise.
3428         * config/h8300/h8300.c (h8300_attribute_table): Likewise.
3429         * config/nvptx/nvptx.c (nvptx_attribute_table): Likewise.
3430         * config/powerpcspe/powerpcspe.c (rs6000_attribute_table): Likewise.
3431         * config/darwin.h (SUBTARGET_ATTRIBUTE_TABLE): Likewise.
3432         * config/stormy16/stormy16.c (xstormy16_attribute_table): Likewise.
3433         * config/bfin/bfin.c (bfin_attribute_table): Likewise.
3434         * config/rs6000/rs6000.c (rs6000_attribute_table): Likewise.
3435         * config/rl78/rl78.c (rl78_attribute_table): Likewise.
3436         * config/nds32/nds32.c (nds32_attribute_table): Likewise.
3437         * doc/plugins.texi (user_attr): Likewise.  Add NULL for
3438         exclude.
3439         * attribs.c (empty_attribute_table): Swap affects_type_identity and
3440         handler fields.
3441         (register_scoped_attributes, decl_attributes): Formatting fixes.
3443         PR tree-optimization/83269
3444         * fold-const.c (fold_binary_loc): Perform (-A) - B -> (-B) - A
3445         subtraction in arg0's type if type is signed and arg0 is unsigned.
3446         Formatting fix.
3448         PR sanitizer/81281
3449         * match.pd ((T)(P + A) - (T)P -> (T) A): Use @@0 instead of @0 and
3450         convert? on @0 instead of convert.  Check type of @1, not @0.
3451         ((T)P - (T)(P + A) -> -(T) A): Use @@0 instead of @0 and
3452         convert? on @0 instead of convert.  Check type of @1, not @0.
3453         ((T)(P + A) - (T)(P + B) -> (T)A - (T)B): Use @@0 instead of @0,
3454         only optimize if either both @1 and @2 types are narrower
3455         precision, or both are wider or equal precision, and in the former
3456         case only if both have undefined overflow.
3458 2017-12-15  Richard Biener  <rguenther@suse.de>
3460         PR lto/83388
3461         * internal-fn.def (IFN_NOP): Add.
3462         * internal-fn.c (expand_NOP): Do nothing.
3463         * lto-streamer-in.c (input_function): Instead of removing
3464         sanitizer calls replace them with IFN_NOP calls.
3466 2017-12-15  Richard Sandiford  <richard.sandiford@linaro.org>
3467             Alan Hayward  <alan.hayward@arm.com>
3468             David Sherwood  <david.sherwood@arm.com>
3470         * dse.c (store_info, read_info_type): Replace begin and end with
3471         offset and width.
3472         (print_range): New function.
3473         (set_all_positions_unneeded, any_positions_needed_p)
3474         (check_mem_read_rtx, scan_stores, scan_reads, dse_step5): Update
3475         accordingly.
3476         (record_store): Likewise.  Optimize the case in which all positions
3477         are unneeded.
3478         (get_stored_val): Replace read_begin and read_end with read_offset
3479         and read_width.
3480         (replace_read): Update call accordingly.
3482 2017-12-15  Bin Cheng  <bin.cheng@arm.com>
3484         * gimple-loop-interchange.cc (STMT_COST_RATIO): New macro.
3485         (loop_cand::m_num_stmts, loop_cand::m_const_init_reduc): New members.
3486         (loop_cand::loop_cand): Initialize above members.
3487         (loop_cand::supported_operations): Delete.
3488         (loop_cand::can_interchange_p): Inline above function.
3489         (loop_cand::classify_simple_reduction): Record number of constant
3490         initialized simple reductions.
3491         (should_interchange_loops): New parameters.  Check stmt cost of loops
3492         to be interchange.
3493         (tree_loop_interchange::interchange): Prepare stmt cost of outer loop.
3494         Update call to should_interchange_loops.
3495         (should_interchange_loop_nest): Update call to
3496         should_interchange_loops.
3498 2017-12-15  Eric Botcazou  <ebotcazou@adacore.com>
3500         PR target/66488
3501         * ggc-page.c (HOST_BITS_PER_PTR): Do not define here...
3502         * hwint.h (HOST_BITS_PER_PTR): ...but here instead.
3503         * config/i386/xm-mingw32.h (HOST_BITS_PER_PTR): Delete.
3505 2017-12-15  Julia Koval  <julia.koval@intel.com>
3507         * config/i386/i386-builtin.def (__builtin_ia32_vaesenclast_v16qi,
3508         __builtin_ia32_vaesenclast_v32qi, __builtin_ia32_vaesenclast_v64qi):
3509         New.
3510         * config/i386/sse.md (vaesenclast_<mode>): New pattern.
3511         * config/i386/vaesintrin.h (_mm256_aesenclast_epi128,
3512         _mm512_aesenclast_epi128, _mm_aesenclast_epi128): New intrinsics.
3514 2017-12-15  Julia Koval  <julia.koval@intel.com>
3516         * config/i386/i386-builtin.def (__builtin_ia32_vaesenc_v16qi,
3517         __builtin_ia32_vaesenc_v32qi, __builtin_ia32_vaesenc_v64qi): New.
3518         * config/i386/sse.md (vaesenc_<mode>): New pattern.
3519         * config/i386/vaesintrin.h (_mm256_aesenc_epi128, _mm512_aesenc_epi128,
3520         _mm_aesenc_epi128): New intrinsics.
3522 2017-12-15  Julia Koval  <julia.koval@intel.com>
3524         * config/i386/i386-builtin.def (__builtin_ia32_vaesdeclast_v16qi,
3525         __builtin_ia32_vaesdeclast_v32qi, __builtin_ia32_vaesdeclast_v64qi):
3526         New.
3527         * config/i386/sse.md (vaesdeclast_<mode>): New pattern.
3528         * config/i386/vaesintrin.h (_mm256_aesdeclast_epi128,
3529         _mm512_aesdeclast_epi128, _mm_aesdeclast_epi128): New intrinsics.
3531 2017-12-14  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
3533         * gimple-ssa-strength-reduction.c (analyze_increments):
3534         Distinguish replacement costs for constant strides from those for
3535         unknown strides.
3537 2017-12-14  Jakub Jelinek  <jakub@redhat.com>
3539         * var-tracking.c (variable_tracking_main_1): Formatting fix.
3541 2017-12-14  Bernd Edlinger  <bernd.edlinger@hotmail.de>
3543         * doc/invoke.texi: Document -Wcast-function-type.
3544         * recog.h (stored_funcptr): Change signature.
3545         * tree-dump.c (dump_node): Avoid warning.
3546         * typed-splay-tree.h (typed_splay_tree): Avoid warning.
3548 2017-12-14  Qing Zhao  <qing.zhao@oracle.com>
3550         PR middle_end/79538
3551         * gimple-fold.c (get_range_strlen): Add the handling of non-member 
3552         array.
3554 2017-12-14  David Malcolm  <dmalcolm@redhat.com>
3556         PR tree-optimization/83312
3557         * domwalk.h (dom_walker::dom_walker): Fix typo in comment.
3558         * tree-cfg.c (find_taken_edge): Update to handle NULL_TREE for
3559         "val" param, and to cope with arbitrary basic blocks.
3560         (find_taken_edge_cond_expr): Add "cond_stmt" param and use it to
3561         handle NULL_TREE for "val", dropping "bb" param.
3562         (find_taken_edge_switch_expr): Make "switch_stmt" param const and
3563         drop "bb" param.  Handle NULL_TREE for "val".
3564         (find_case_label_for_value): Make "switch_stmt" param const.
3565         * tree-vrp.c (class check_array_bounds_dom_walker): New subclass
3566         of dom_walker.
3567         (vrp_prop::check_all_array_refs): Reimplement as...
3568         (check_array_bounds_dom_walker::before_dom_children): ...this new
3569         vfunc.  Replace linear search through BB block list, excluding
3570         those with non-executable in-edges via dominator walk.
3572 2017-12-14  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
3574         * config/arm/arm.opt (mverbose-cost-dump): New option.
3575         * config/arm/arm.c (arm_rtx_costs): Use it.
3577 2017-12-14  Andreas Schwab  <schwab@linux-m68k.org>
3579         PR bootstrap/83396
3580         * reload1.c (emit_input_reload_insns): Skip debug markers.
3582 2017-12-14  Alexandre Oliva  <aoliva@redhat.com>
3584         * config/i386/i386.c (rest_of_insert_endbranch): Use call loc
3585         for its nop_endbr.
3587         PR bootstrap/83396
3588         * config/arc/arc.c (hwloop_optimize): Skip debug insns.
3589         * config/sh/sh-protos.h (sh_find_set_of_reg): Adjust.
3590         * config/sh/sh.c: Skip debug insns besides notes.
3591         * config/sh/sh.md: Likewise.
3592         * config/sh/sh_treg_combine.cc: Likewise.
3593         * config/sh/sync.md: Likewise.
3595 2017-12-14  Tom de Vries  <tom@codesourcery.com>
3597         * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
3598         Add item for weak_undefined.
3600 2017-12-14  Richard Biener  <rguenther@suse.de>
3602         PR tree-optimization/67842
3603         * sese.h (bb_in_region): Remove #if 0'ed code.
3605 2017-12-14  Richard Biener  <rguenther@suse.de>
3607         PR tree-optimization/83326
3608         * tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Add
3609         may_be_zero parameter and handle it by not marking the first
3610         peeled copy as not exiting the loop.
3611         (try_peel_loop): Likewise.
3612         (canonicalize_loop_induction_variables): Use number_of_iterations_exit
3613         to handle the case of constant or zero iterations and perform
3614         loop header copying on-the-fly.
3616 2017-12-14  Richard Biener  <rguenther@suse.de>
3618         PR tree-optimization/83418
3619         * vr-values.c (vr_values::extract_range_for_var_from_comparison_expr):
3620         Instead of asserting we don't get unfolded comparisons deal with them.
3622 2017-12-14  Jakub Jelinek  <jakub@redhat.com>
3624         PR bootstrap/83396
3625         * var-tracking.c (vt_initialize): Ignore non-DEBUG_INSNs outside of
3626         basic blocks.  Assert debug bind insns don't appear outside of bbs,
3627         don't reset them.  Assert insns without BLOCK_FOR_INSN are outside of
3628         bb.  Simplify.
3630         PR tree-optimization/83198
3631         * gimple-ssa-sprintf.c (format_floating): Set type solely based on
3632         dir.modifier, regardless of TREE_TYPE (arg).  Assume non-REAL_CST
3633         value if arg is a REAL_CST with incompatible type.
3635 2017-12-14  Sudakshina Das  <sudi.das@arm.com>
3636             Bin Cheng  <bin.cheng@arm.com>
3638         PR target/81228
3639         * config/aarch64/aarch64.c (aarch64_select_cc_mode): Move LTGT
3640         to CCFPEmode.
3641         * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Add
3642         LTGT.
3644 2017-12-14  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
3646         * config/aarch64/aarch64-cores.def (cortex-a55, cortex-a75,
3647         cortex-a75.cortex-a55): Specify AARCH64_FL_F16 in the arch features.
3649 2017-12-14  Richard Sandiford  <richard.sandiford@linaro.org>
3650             Alan Hayward  <alan.hayward@arm.com>
3651             David Sherwood  <david.sherwood@arm.com>
3653         * poly-int.h: New file.
3654         * poly-int-types.h: Likewise.
3655         * coretypes.h: Include them.
3656         (POLY_INT_CONVERSION): Define.
3657         * target.def (estimated_poly_value): New hook.
3658         * doc/tm.texi.in (TARGET_ESTIMATED_POLY_VALUE): New hook.
3659         * doc/tm.texi: Regenerate.
3660         * doc/poly-int.texi: New file.
3661         * doc/gccint.texi: Include it.
3662         * doc/rtl.texi: Describe restrictions on subreg modes.
3663         * Makefile.in (TEXI_GCCINT_FILES): Add poly-int.texi.
3664         * genmodes.c (NUM_POLY_INT_COEFFS): Provide a default definition.
3665         (emit_insn_modes_h): Emit a definition of NUM_POLY_INT_COEFFS.
3666         * targhooks.h (default_estimated_poly_value): Declare.
3667         * targhooks.c (default_estimated_poly_value): New function.
3668         * target.h (estimated_poly_value): Likewise.
3669         * wide-int.h (WI_UNARY_RESULT): Use wi::binary_traits.
3670         (wi::unary_traits): Delete.
3671         (wi::binary_traits::signed_shift_result_type): Define for
3672         offset_int << HOST_WIDE_INT, etc.
3673         (generic_wide_int::operator <<=): Define for all types and use
3674         wi::lshift instead of <<.
3675         (wi::hwi_with_prec): Add a default constructor.
3676         (wi::ints_for): New class.
3677         (operator <<): Define for all wide-int types.
3678         (operator /): New function.
3679         (operator %): Likewise.
3680         * selftest.h (ASSERT_KNOWN_EQ, ASSERT_KNOWN_EQ_AT, ASSERT_MAYBE_NE)
3681         (ASSERT_MAYBE_NE_AT): New macros.
3683 2017-12-13  Eric Botcazou  <ebotcazou@adacore.com>
3684             Dominik Vogt  <vogt@linux.vnet.ibm.com>
3686         PR middle-end/78468
3687         * emit-rtl.c (init_emit): Remove ??? comment.
3688         * explow.c (get_dynamic_stack_size): Take known alignment of stack
3689         pointer + STACK_DYNAMIC_OFFSET into account in lieu of STACK_BOUNDARY.
3690         * config/sparc/sparc.h (INIT_EXPANDERS): In 32-bit mode, lower the
3691         alignment of 3 virtual registers to BITS_PER_WORD.
3693         * config/sparc/sparc.c (sparc_compute_frame_size): Simplify.
3695 2017-12-13  Peter Bergner  <bergner@vnet.ibm.com>
3697         * config/rs6000/ppc-auxv.h (PPC_FEATURE2_HTM_NO_SUSPEND): New define.
3698         * config/rs6000/rs6000.c (cpu_supports_info): Use it.
3700 2017-12-13  Alexandre Oliva  <aoliva@redhat.com>
3702         PR bootstrap/83396
3703         * reload1.c (eliminate_regs_in_insn): Skip debug markers.
3705 2017-12-13  Jakub Jelinek  <jakub@redhat.com>
3707         * tree-cfg.c (verify_gimple_in_cfg): Verify no non-label stmts
3708         with the exception of debug begin stmt markers appear before
3709         labels.
3711         PR bootstrap/83396
3712         * final.c (rest_of_handle_final): Call variable_tracking_main only
3713         if !flag_var_tracking.
3715 2017-12-13  Alexandre Oliva  <aoliva@redhat.com>
3716             Jakub Jelinek  <jakub@redhat.com>
3718         PR bootstrap/83396
3719         PR debug/83391
3720         * tree-cfgcleanup.c (remove_forwarder_block): Keep after
3721         labels debug stmts that can only appear after labels.
3723 2017-12-13  Alexander Monakov  <amonakov@ispras.ru>
3725         PR rtl-optimization/82398
3726         * sel-sched.c (sel_rank_for_schedule): Fix check for zero
3727         EXPR_USEFULNESS in priority comparison.
3729 2017-12-13  Segher Boessenkool  <segher@kernel.crashing.org>
3731         PR rtl-optimization/83393
3732         * combine.c (move_deaths): If reg_stat points to a too new insn in
3733         last_death, do not use it: find the proper insn instead.
3735 2017-12-12  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
3737         PR target/82190
3738         * config/rs6000/rs6000-string.c (expand_block_compare,
3739         expand_strn_compare): Fix set_mem_size() calls.
3741 2017-12-12  Jeff Law  <law@redhat.com>
3743         PR tree-optimization/83298
3744         PR tree-optimization/83362
3745         PR tree-optimization/83383
3746         * gimple-ssa-evrp-analyze.h (class evrp_range_analyzer): Make
3747         push_value_range a public interface.  Add new argument to
3748         record_ranges_from_stmt.
3749         * gimple-ssa-evrp-analyze.c
3750         (evrp_range_analyzer::record_ranges_from_stmt): Add new argument.
3751         Update comments.  Handle recording temporary equivalences.
3752         * tree-ssa-dom.c (dom_opt_opt_walker::before_dom_children): Add
3753         new argument to call to evrp_range_analyzer::record_ranges_from_stmt.
3754         * gimple-ssa-evrp.c (evrp_dom_walker::before_dom_children): Likewise.
3755         * tree-ssa-threadedge.c: Include alloc-pool.h, vr-values.h and
3756         gimple-ssa-evrp-analyze.h.
3757         (record_temporary_equivalences_from_phis): Add new argument.  When
3758         the PHI arg is an SSA_NAME, set the result's range to the range
3759         of the PHI arg.
3760         (record_temporary_equivalences_from_stmts_at_dest): Record ranges
3761         from statements too.
3762         (thread_through_normal_block): Accept new argument, evrp_range_analyzer.
3763         Pass it down to children as needed.
3764         (thread_outgoing_edges): Likewise.
3765         (thread_across_edge): Likewise.   Push/pop range state as needed.
3766         * tree-ssa-threadedge.h (thread_outgoing_edges): Update prototype.
3768 2017-12-12  Julia Koval  <julia.koval@intel.com>
3770         * config/i386/i386.c (PTA_SKYLAKE_AVX512): Add PTA_CLWB.
3771         (PTA_CANNONLAKE): Remove PTA_CLWB.
3773 2017-12-12  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
3775         PR target/83332
3776         * config/rs6000/vector.md (vcondv2dfv2di): New define_expand.
3777         (vcondv2div2df): Likewise.
3778         (vconduv2dfv2di): Likewise.
3780 2017-12-12  Jakub Jelinek  <jakub@redhat.com>
3782         * builtins.def (BUILT_IN_NEXTAFTER, BUILT_IN_NEXTAFTERF,
3783         BUILT_IN_NEXTAFTERL, BUILT_IN_NEXTTOWARD, BUILT_IN_NEXTTOWARDF,
3784         BUILT_IN_NEXTTOWARDL): Use ATTR_MATHFN_ERRNO instead of
3785         ATTR_MATHFN_FPROUNDING_ERRNO.
3787 2017-12-12  Richard Biener  <rguenther@suse.de>
3789         PR tree-optimization/83385
3790         * graphite-scop-detection.c (get_order, order): Remove.
3791         (bb_to_rpo): New global.
3792         (cmp_pbbs): Adjust.
3793         (build_scops): Sort pbbs in RPO order.
3795 2017-12-12  James Greenhalgh  <james.greenhalgh@arm.com>
3797         * combine.c (simplify_set): Do not transform subregs to zero_extends
3798         if the destination mode is a vector mode.
3800 2017-12-12  Jakub Jelinek  <jakub@redhat.com>
3802         PR tree-optimization/83359
3803         * tree-cfg.h (fold_loop_internal_call): Declare.
3804         * tree-vectorizer.c (fold_loop_internal_call): Moved to ...
3805         * tree-cfg.c (fold_loop_internal_call): ... here.  No longer static.
3806         (find_loop_dist_alias): New function.
3807         (move_sese_region_to_fn): If any dloop->orig_loop_num value is
3808         updated, also adjust any corresponding LOOP_DIST_ALIAS internal
3809         calls.
3811         PR tree-optimization/80631
3812         * tree-vect-loop.c (get_initial_def_for_reduction): Fix comment typo.
3813         (vect_create_epilog_for_reduction): Add INDUC_VAL and INDUC_CODE
3814         arguments, for INTEGER_INDUC_COND_REDUCTION use INDUC_VAL instead of
3815         hardcoding zero as the value if COND_EXPR is never true.  For
3816         INTEGER_INDUC_COND_REDUCTION don't emit the final COND_EXPR if
3817         INDUC_VAL is equal to INITIAL_DEF, and use INDUC_CODE instead of
3818         hardcoding MAX_EXPR as the reduction operation.
3819         (is_nonwrapping_integer_induction): Allow negative step.
3820         (vectorizable_reduction): Compute INDUC_VAL and INDUC_CODE for
3821         vect_create_epilog_for_reduction, if no value is suitable, don't
3822         use INTEGER_INDUC_COND_REDUCTION for now.  Formatting fixes.
3824 2017-12-12  Richard Biener  <rguenther@suse.de>
3826         PR tree-optimization/81889
3827         * tree-ssa-loop-niter.c (infer_loop_bounds_from_signedness): Use
3828         range info from the non-wrapping IV instead of just the range
3829         of the type.
3831 2017-12-12  Julia Koval  <julia.koval@intel.com>
3833         * config.gcc: Add vaesintrin.h.
3834         * config/i386/i386-builtin-types.def (V64QI_FTYPE_V64QI_V64QI):
3835         New type.
3836         * config/i386/i386-builtin.def (__builtin_ia32_vaesdec_v16qi,
3837         __builtin_ia32_vaesdec_v32qi, __builtin_ia32_vaesdec_v64qi):
3838         New builtins.
3839         * config/i386/i386.c (ix86_expand_args_builtin): Handle new type.
3840         * config/i386/immintrin.h: Include vaesintrin.h.
3841         * config/i386/sse.md (vaesdec_<mode>): New pattern.
3842         * config/i386/vaesintrin.h (_mm256_aesdec_epi128, _mm512_aesdec_epi128,
3843         _mm_aesdec_epi128): New intrinsics.
3845 2017-12-12  Julia Koval  <julia.koval@intel.com>
3847         * common/config/i386/i386-common.c (OPTION_MASK_ISA_VAES_SET,
3848         OPTION_MASK_ISA_VAES_UNSET): New.
3849         (ix86_handle_option): Handle -mvaes.
3850         * config/i386/cpuid.h: Define bit_VAES.
3851         * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mvaes.
3852         * config/i386/i386-c.c (__VAES__): New.
3853         * config/i386/i386.c (ix86_target_string): Add -mvaes.
3854         (ix86_valid_target_attribute_inner_p): Ditto.
3855         * config/i386/i386.h (TARGET_VAES, TARGET_VAES_P): New.
3856         * config/i386/i386.opt: Add -mvaes.
3857         * doc/invoke.texi: Ditto.
3859 2017-12-12  Alexandre Oliva  <aoliva@redhat.com>
3861         * debug.h (gcc_debug_hooks): Add inline_entry.
3862         * dbxout.c (dbx_debug_hooks, xcoff_debug_hooks): Likewise.
3863         * debug.c (do_nothing_debug_hooks): Likewise.
3864         * vmsdbgout.c (vmsdbg_debug_hooks): Likewise.
3865         * dwarf2out.c (dwarf2_debug_hooks): Likewise.
3866         (dwarf2_lineno_debug_hooks): Likewise.
3868         * common.opt (gstatement-frontiers): New, setting
3869         debug_nonbind_markers_p.
3870         * rtl.h (MAY_HAVE_DEBUG_MARKER_INSNS): Activate.
3871         * toplev.c (process_options): Autodetect value for debug statement
3872         frontiers option.
3873         * tree.h (MAY_HAVE_DEBUG_MARKER_STMTS): Activate.
3874         * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers): New.
3876         * cfgexpand.c (expand_gimple_basic_block): Handle begin stmt
3877         markers.  Integrate source bind into debug stmt expand loop.
3878         (pass_expand::execute): Check debug marker limit.  Avoid deep
3879         TER and expand debug locations for debug bind insns only.
3880         * cse.c (insn_live_p): Keep nonbind markers and debug bindings
3881         followed by them.
3882         * df-scan.c (df_insn_delete): Accept out-of-block debug insn.
3883         * final.c (reemit_insn_block_notes): Take current block from
3884         nonbind markers.  Declare note where it's first set.
3885         (final_scan_insn): Handle begin stmt notes.  Emit is_stmt according to
3886         begin stmt markers if enabled.
3887         (notice_source_line): Handle nonbind markers.  Fail if their
3888         location is unknown or that of builtins.
3889         (rest_of_handle_final): Convert begin stmt markers to notes if
3890         var-tracking didn't run.
3891         (rest_of_clean_state): Skip begin stmt markers.
3892         * gimple-pretty-print.c (dump_gimple_debug): Handle begin stmt
3893         markers.
3894         * function.c (allocate_struct_function): Set begin_stmt_markers.
3895         * function.h (struct function): Add debug_marker_count counter
3896         and debug_nonbind_markers flag.
3897         * gimple-iterator.c (gsi_remove): Adjust debug_marker_count.
3898         * gimple-low.c (lower_function_body): Adjust
3899         debug_nonbind_markers.
3900         (lower_stmt): Drop or skip gimple debug stmts.
3901         (lower_try_catch): Skip debug stmts.
3902         * gimple.c (gimple_build_debug_begin_stmt): New.
3903         (gimple_copy): Increment debug_marker_count if copying one.
3904         * gimple.h (gimple_build_debug_begin_stmt): Declare.
3905         * gimplify.c (rexpr_location): New.
3906         (rexpr_has_location): New.
3907         (warn_switch_unreachable_r): Handle gimple debug stmts.
3908         (shortcut_cond_r): Call expr_location.
3909         (find_goto): New.
3910         (find_goto_label): New.
3911         (shortcut_cond_expr): Call expr_has_location, expr_location, and
3912         find_goto_label.
3913         (gimplify_cond_expr): Call find_goto_label, expr_has_location, and
3914         expr_location.
3915         (gimplify_expr): Handle begin stmt markers.  Reject debug expr decls.
3916         * langhooks-def.h (LANG_HOOKS_EMITS_BEGIN_STMT): New.  Add to...
3917         (LANG_HOOKS_INITIALIZER): ... this.
3918         * langhooks.h (struct lang_hooks): Add emits_begin_stmt.
3919         * lra-contraints.c (inherit_reload_reg): Tolerate between-blocks
3920         debug insns.
3921         (update_ebb_live_info): Skip debug insn markers.
3922         * lra.c (debug_insn_static_data): Rename to...
3923         (debug_bind_static_data): ... this.
3924         (debug_marker_static_data): New.
3925         (lra_set_insn_recog_data): Select one of the above depending
3926         on debug insn kind.
3927         (lra_update_isn_regno_info): Don't assume debug insns have
3928         freqs.
3929         (push_insns): Skip debug insns.
3930         * lto-streamer-in.c (input_function): Drop debug stmts
3931         depending on active options.  Adjust debug_nonbind_markers.
3932         * params.def (PARAM_MAX_DEBUG_MARKER_COUNT): New.
3933         * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
3934         begin stmt marker notes.
3935         (print_insn): Likewise.
3936         * recog.c (extract_insn): Recognize rtl for debug markers.
3937         * rtl.def (DEBUG_MARKER): New.
3938         * tree-inline.c: Include params.h.
3939         (remap_gimple_stmt): Handle nonbind markers.
3940         (maybe_move_debug_stmts_to_successors): Likewise.
3941         (copy_debug_stmt): Likewise.
3942         * tree-iterator.c (append_to_statement_list_1): Append begin stmt
3943         markers regardless of no side effects.
3944         (tsi_link_before): Don't update container's side effects when adding
3945         a begin stmt marker.
3946         (tsi_link_after): Likewise.
3947         (expr_first): Skip begin stmt markers.
3948         (expr_last): Likewise.
3949         * tree-pretty-print (dump_generic_node): Handle begin stmt markers.
3950         * tree-ssa-threadedge.c (propagate_threaded_block_debug_info):
3951         Disregard nonbind markers.
3952         * tree.c (make_node_stat): Don't set side effects for begin stmt
3953         markers.
3954         (build1_stat): Likewise.
3955         * tree.def (DEBUG_BEGIN_STMT): New.
3956         * tree.h (GOTO_DESTINATION): Require a GOTO_EXPR.
3957         * var-tracking.c (delete_debug_insns): Renamed to...
3958         (delete_vta_debug_insns): ... this.
3959         (reemit_marker_as_note): New.
3960         (vt_initialize): Reemit markers.
3961         (delete_vta_debug_insns): Likewise.
3962         (vt_debug_insns_local): Reemit or delete markers.
3963         (variable_tracking_main_1): Likewise.
3964         * doc/generic.texi (DEBUG_BEGIN_STMT): Document.
3965         * doc/gimple.texi (gimple_debug_begin_stmt_p): New.
3966         (gimple_debug_nonbind_marker_p): New.
3967         (gimple_build_debug_bind): Adjust.
3968         (gimple_build_debug_begin_stmt): New.
3969         * doc/invoke.texi (max-debug-marker-count): New param.
3970         * doc/rtl.texi (debug_implicit_ptr, entry_value): New.
3971         (debug_parameter_ref, debug_marker): New.
3972         (NOTE_INSN_BEGIN_STMT): New.
3973         (DEBUG_INSN): Describe begin stmt markers.
3975         * cfgbuild.c (find_bb_boundaries): Don't purge dead edges if,
3976         without debug insns, we wouldn't, but clean up debug insns
3977         after a control flow insn nevertheless.
3979         * cfgcleanup.c (delete_unreachable_blocks): Use alternate
3980         block removal order if MAY_HAVE_DEBUG_BIND_INSNS.
3981         * cfgexpand.c (label_rtx_for_bb): Skip debug insns.
3982         * cfgrtl.c (try_redirect_by_replacing_jump): Skip debug insns.
3983         (rtl_tidy_fallthru_edge): Likewise.
3984         (rtl_verify_fallthru): Likewise.
3985         (rtl_verify_bb_layout): Likewise.
3986         (skip_insns_after_block): Likewise.
3987         (duplicate_insn_chain): Use DEBUG_BIND_INSN_P.
3988         * dwarf2out.c: Include print-rtl.h.
3989         (dwarf2out_next_real_insn): New.
3990         (dwarf2out_var_location): Call it.  Disregard begin stmt markers.
3991         Dump debug binds in asm comments.
3992         * gimple-iterator.c (gimple_find_edge_insert_loc): Skip debug stmts.
3993         * gimple-iterator.h (gsi_start_bb_nondebug): Remove; adjust
3994         callers to use gsi_start_nondebug_bb instead.
3995         (gsi_after_labels): Skip gimple debug stmts.
3996         (gsi_start_nondebug): New.
3997         * gimple-loop-interchange.c (find_deps_in_bb_for_stmt): Adjust.
3998         (proper_loop_form_for_interchange): Adjust.
3999         * gimple-low.c (gimple_seq_may_fallthru): Take last nondebug stmt.
4000         * gimple.h (gimple_seq_last_nondebug_stmt): New.
4001         * gimplify.c (last_stmt_in_scope): Skip debug stmts.
4002         (collect_fallthrough_labels): Likewise.
4003         (should_warn_for_implicit_fallthrough): Likewise.
4004         (warn_implicit_fallthrough_r): Likewise.
4005         (expand_FALLTHROUGH_r): Likewise.
4006         * graphite-isl-ast-to-gimple.c (gsi_insert_earliest): Adjust.
4007         (graphite_copy_stmts_from_block): Skip nonbind markers.
4008         * haifa-sched.c (sched_extend_bb): Skip debug insns.
4009         * ipa-icf-gimple.c (func_checker::compare_bb): Adjust.
4010         * jump.c (clean_barriers): Skip debug insns.
4011         * omp-expand.c (expand_parallel_call): Skip debug insns.
4012         (expand_task_call): Likewise.
4013         (remove_exit_barrier): Likewise.
4014         (expand_omp_taskreg): Likewise.
4015         (expand_omp_for_init_counts): Likewise.
4016         (expand_omp_for_generic): Likewise.
4017         (expand_omp_for_static_nochunk): Likewise.
4018         (expand_omp_for_static_chunk): Likewise.
4019         (expand_omp_simd): Likewise.
4020         (expand_omp_taskloop_for_outer): Likewise.
4021         (expand_omp_taskloop_for_inner): Likewise.
4022         (expand_oacc_for): Likewise.
4023         (expand_omp_sections): Likewise.
4024         (expand_omp_single): Likewise.
4025         (expand_omp_synch): Likewise.
4026         (expand_omp_atomic_load): Likewise.
4027         (expand_omp_atomic_store): Likewise.
4028         (expand_omp_atomic_fetch_op): Likewise.
4029         (expand_omp_atomic_pipeline): Likewise.
4030         (expand_omp_atomic_mutex): Likewise.
4031         (expand_omp_target): Likewise.
4032         (grid_expand_omp_for_loop): Likewise.
4033         (grid_expand_target_grid_body): Likewise.
4034         (build_omp_regions_1): Likewise.
4035         * omp-low.c (check_combined_parallel): Skip debug stmts.
4036         * postreload.c (fixup_debug_insns): Skip nonbind debug insns.
4037         * regcprop.c (find_oldest_value_reg): Ensure REGNO is not a pseudo.
4038         * sese.c (sese_trivially_empty_bb_p): Call is_gimple_debug in
4039         test.
4040         * tree-cfg.c (make_blobs_1): Skip debug stmts.
4041         (make_edges): Likewise.
4042         (cleanup_dead_labels): Likewise.
4043         (gimple_can_merge_blocks_p): Likewise.
4044         (stmt_starts_bb_p): Likewise.
4045         (gimple_block_label): Likewise.
4046         (gimple_redirect_edge_and_branch): Likewise.
4047         * tree-cfgcleanup.c (remove_forwarder_block): Rearrange skipping
4048         of debug stmts.
4049         (execute_cleanup_cfg_post_optimizing): Dump enumerated decls with
4050         TDF_SLIM.
4051         * tree-pretty-print (print_declaration): Omit initializer in slim
4052         dumps.
4053         * tree-ssa-dce.c (mark_stmt_if_obviously_necessary): Mark begin stmt
4054         markers.
4055         (eliminate_unnecessary_stmts): Stabilize block removal order.
4056         * tree-ssa-tail-merge.c (find_duplicate): Skip debug stmts.
4057         * var-tracking.c (get_first_insn): New.
4058         (vt_emit_notes): Call it.
4059         (vt_initialize): Walk any insns before the first BB.
4060         (delete_debug_insns): Likewise.
4062         * gimple.h (enum gimple_debug_subcode): Add
4063         GIMPLE_DEBUG_BEGIN_STMT.
4064         (gimple_debug_begin_stmt_p): New.
4065         (gimple_debug_nonbind_marker_p): New.
4066         * tree.h (MAY_HAVE_DEBUG_MARKER_STMTS): New.
4067         (MAY_HAVE_DEBUG_BIND_STMTS): Renamed from....
4068         (MAY_HAVE_DEBUG_STMTS): ... this.  Check both.
4069         * insn-notes.def (BEGIN_STMT): New.
4070         * rtl.h (MAY_HAVE_DEBUG_MARKER_INSNS): New.
4071         (MAY_HAVE_DEBUG_BIND_INSNS): Renamed from....
4072         (MAY_HAVE_DEBUG_INSNS): ... this.  Check both.
4073         (NOTE_MARKER_LOCATION, NOTE_MARKER_P): New.
4074         (DEBUG_BIND_INSN_P, DEBUG_MARKER_INSN_P): New.
4075         (INSN_DEBUG_MARKER_KIND): New.
4076         (GEN_RTX_DEBUG_MARKER_BEGIN_STMT_PAT): New.
4077         (INSN_VAR_LOCATION): Check for VAR_LOCATION.
4078         (INSN_VAR_LOCATION_PTR): New.
4079         * cfgexpand.c (expand_debug_locations): Handle debug bind insns
4080         only.
4081         (expand_gimple_basic_block): Likewise.  Emit debug temps for TER
4082         deps only if debug bind insns are enabled.
4083         (pass_expand::execute): Avoid deep TER and expand
4084         debug locations for debug bind insns only.
4085         * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Narrow
4086         debug stmts special handling down to debug bind stmts.
4087         * combine.c (try_combine): Narrow debug insns special handling
4088         down to debug bind insns.
4089         * cse.c (delete_trivially_dead_insns): Handle debug bindings.
4090         Narrow debug insns preexisting special handling down to debug
4091         bind insns.
4092         * dce.c (rest_of_handle_ud_dce): Narrow debug insns special
4093         handling down to debug bind insns.
4094         * function.c (instantiate_virtual_regs): Skip debug markers,
4095         adjust handling of debug binds.
4096         * gimple-ssa-backprop.c (backprop::prepare_change): Try debug
4097         temp insertion iff MAY_HAVE_DEBUG_BIND_STMTS.
4098         * haifa-sched.c (schedule_insn): Narrow special handling of debug
4099         insns to debug bind insns.
4100         * ipa-param-manipulation.c (ipa_modify_call_arguments): Narrow
4101         special handling of debug stmts to debug bind stmts.
4102         * ipa-split.c (split_function): Likewise.
4103         * ira.c (combine_and_move_insns): Adjust debug bind insns only.
4104         * loop-unroll.c (apply_opt_in_copies): Adjust tests on bind
4105         debug insns.
4106         * reg-stack.c (convert_regs_1): Use DEBUG_BIND_INSN_P.
4107         * regrename.c (build_def_use): Likewise.
4108         * regcprop.c (copyprop_hardreg_forward_1): Likewise.
4109         (pass_cprop_hardreg): Narrow special casing of debug insns to
4110         debug bind insns.
4111         * regstat.c (regstat_init_n_sets_and_refs): Likewise.
4112         * reload1.c (reload): Likewise.
4113         * sese.c (sese_insert_phis_for_liveouts): Narrow special
4114         casing of debug stmts to debug bind stmts.
4115         * shrink-wrap.c (move_insn_for_shrink_wrap): Likewise.
4116         * ssa-iterators.h (num_imm_uses): Likewise.
4117         * tree-cfg.c (gimple_merge_blocks): Narrow special casing of
4118         debug stmts to debug bind stmts.
4119         * tree-inline.c (tree_function_versioning): Narrow special casing
4120         of debug stmts to debug bind stmts.
4121         * tree-loop-distribution.c (generate_loops_for_partition):
4122         Narrow special casing of debug stmts to debug bind stmts.
4123         * tree-sra.c (analyze_access_subtree): Narrow special casing
4124         of debug stmts to debug bind stmts.
4125         * tree-ssa-dce.c (remove_dead_stmt): Narrow special casing of debug
4126         stmts to debug bind stmts.
4127         * tree-ssa-loop-ivopt.c (remove_unused_ivs): Narrow special
4128         casing of debug stmts to debug bind stmts.
4129         * tree-ssa-reassoc.c (reassoc_remove_stmt): Likewise.
4130         * tree-ssa-tail-merge.c (tail_merge_optimize): Narrow special
4131         casing of debug stmts to debug bind stmts.
4132         * tree-ssa-threadedge.c (propagate_threaded_block_debug_info):
4133         Likewise.
4134         * tree-ssa.c (flush_pending_stmts): Narrow special casing of
4135         debug stmts to debug bind stmts.
4136         (gimple_replace_ssa_lhs): Likewise.
4137         (insert_debug_temp_for_var_def): Likewise.
4138         (insert_debug_temps_for_defs): Likewise.
4139         (reset_debug_uses): Likewise.
4140         * tree-ssanames.c (release_ssa_name_fn): Likewise.
4141         * tree-vect-loop-manip.c (adjust_debug_stmts_now): Likewise.
4142         (adjust_debug_stmts): Likewise.
4143         (adjust_phi_and_debug_stmts): Likewise.
4144         (vect_do_peeling): Likewise.
4145         * tree-vect-loop.c (vect_transform_loop): Likewise.
4146         * valtrack.c (propagate_for_debug): Use BIND_DEBUG_INSN_P.
4147         * var-tracking.c (adjust_mems): Narrow special casing of debug
4148         insns to debug bind insns.
4149         (dv_onepart_p, dataflow_set_clar_at_call, use_type): Likewise.
4150         (compute_bb_dataflow, vt_find_locations): Likewise.
4151         (vt_expand_loc, emit_notes_for_changes): Likewise.
4152         (vt_init_cfa_base): Likewise.
4153         (vt_emit_notes): Likewise.
4154         (vt_initialize): Likewise.
4155         (vt_finalize): Likewise.
4157         * emit-rtl.c (next_nondebug_insn, prev_nondebug_insn): Reorder.
4158         (next_nonnote_nondebug_insn, prev_nonnote_nondebug_insn): Reorder.
4159         (next_nonnote_nondebug_insn_bb): New.
4160         (prev_nonnote_nondebug_insn_bb): New.
4161         (prev_nonnote_insn_bb, next_nonnote_insn_bb): Remove.
4162         * rtl.h (prev_nonnote_insn_bb, next_nonnote_insn_bb): Remove decls.
4163         (prev_nonnote_nondebug_insn_bb): Declare.
4164         (next_nonnote_nondebug_insn_bb): Declare.
4165         * cfgbuild.c (find_bb_boundaries): Adjust to skip debug insns.
4166         * cfgrtl.c (get_last_bb_insn): Likewise.
4167         * lra.c (push_insns): Likewise.
4169 2017-12-11  David Malcolm  <dmalcolm@redhat.com>
4171         PR c/82050
4172         * selftest-run-tests.c (selftest::run_tests): Move start/finish code
4173         to...
4174         * selftest.c (selftest::test_runner::test_runner): New ctor.
4175         (selftest::test_runner::~test_runner): New dtor.
4176         * selftest.h (class selftest::test_runner): New class.
4178 2017-12-11  Carl Love  <cel@us.ibm.com>
4180         * config/rs6000/altivec.h (vec_extract_fp32_from_shorth,
4181         vec_extract_fp32_from_shortl]): Add #defines.
4182         * config/rs6000/rs6000-builtin.def (VSLDOI_2DI): Add macro expansion.
4183         * config/rs6000/rs6000-c.c (ALTIVEC_BUILTIN_VEC_UNPACKH,
4184         ALTIVEC_BUILTIN_VEC_UNPACKL, ALTIVEC_BUILTIN_VEC_AND,
4185         ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SRL,
4186         ALTIVEC_BUILTIN_VEC_SRO, ALTIVEC_BUILTIN_VEC_SLD,
4187         ALTIVEC_BUILTIN_VEC_SLL): Add expansions.
4188         * doc/extend.texi: Add documentation for the added builtins.
4190 2017-12-11  Alexandre Oliva  <aoliva@redhat.com>
4192         PR rtl-optimization/80693
4193         PR rtl-optimization/81019
4194         PR rtl-optimization/81020
4195         * combine.c (distribute_notes): Reset any REG_UNUSED REGs that
4196         are not mentioned in i3.  Place the REG_UNUSED note on i2,
4197         possibly modified to REG_DEAD, if it did not originate in i3.
4199 2017-12-11  Jakub Jelinek  <jakub@redhat.com>
4201         * recog.c (store_data_bypass_p_1): New function.
4202         (store_data_bypass_p): Handle USE in a PARALLEL like CLOBBER.  Use
4203         store_data_bypass_p_1 to avoid code duplication.  Formatting fixes.
4205 2017-12-11  Segher Boessenkool  <segher@kernel.crashing.org>
4207         PR rtl-optimization/83361
4208         * ifcvt.c (if_convert): Call fixup_partitions.
4210 2017-12-11  Will Schmidt  <will_schmidt@vnet.ibm.com>
4212         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add support for
4213         early folding of splat_u{8,16,32}.
4215 2017-12-11  Jakub Jelinek  <jakub@redhat.com>
4217         * config/aarch64/aarch64.c (aarch64_print_operand): Don't start
4218         output_operand_lossage first argument with capital letter.
4219         (aarch64_override_options): Don't start error and sorry first argument
4220         with capital letter.
4222 2017-12-11  Andi Kleen  <ak@linux.intel.com>
4224         PR gcov-profile/83355
4225         * auto-profile.c (string_table::get_index_by_decl): Don't
4226         recurse when abstract origin points to itself.
4228 2017-12-11  Bin Cheng  <bin.cheng@arm.com>
4230         PR tree-optimization/83320
4231         * gimple-loop-interchange.cc (free_data_refs_with_aux): Use delete.
4232         (prune_datarefs_not_in_loop): Ditto.
4234 2017-12-10  Gerald Pfeifer  <gerald@pfeifer.com>
4236         * doc/install.texi (Specific): Tweak link to mkssoftware.com.
4238 2017-12-10  Jakub Jelinek  <jakub@redhat.com>
4240         PR tree-optimization/83337
4241         * gimple-loop-interchange.cc (compute_access_stride): Handle
4242         bitfield DRs properly.
4244 2017-12-09  Jakub Jelinek  <jakub@redhat.com>
4246         PR tree-optimization/83338
4247         * tree-vect-stmts.c (vectorizable_operation): Handle POINTER_DIFF_EXPR
4248         vectorization as MINUS_EXPR with a subsequent VIEW_CONVERT_EXPR from
4249         vector of unsigned integers to vector of signed integers.
4251 2017-12-08  Vladimir Makarov  <vmakarov@redhat.com>
4253         PR rtl-optimization/83317
4254         * lra-constraints.c (process_address_1): Add insn code check.
4256 2017-12-08  Michael Matz  <matz@suse.de>
4258         Fix PR tree-optimization/83323
4259         * gimple-loop-jam (unroll_jam_possible_p): Correct test for
4260         head-controlled loops and loop BBs.
4261         * common.opt (funroll-and-jam): Remove, instead ...
4262         (floop-unroll-and-jam): ... reuse this option.
4263         * opts.c (default_options_table): Use OPT_floop_unroll_and_jam.
4264         * doc/invoke.texi (-funroll-and-jam): Move docu to ...
4265         (-floop-unroll-and-jam): ... this option.
4267 2017-12-08  Jakub Jelinek  <jakub@redhat.com>
4269         * ipa-polymorphic-call.c (noncall_stmt_may_be_vtbl_ptr_store): Fix
4270         a comment typo, get_base_ref_and_offset -> get_ref_base_and_extent.
4271         * ipa-prop.c (stmt_may_be_vtbl_ptr_store): Likewise.
4273 2017-12-08  Richard Biener  <rguenther@suse.de>
4275         PR middle-end/81782
4276         * tree-ssa-uninit.c (warn_uninitialized_vars): Properly
4277         handle accesses outside of zero-sized vars.
4279 2017-12-08  Martin Jambor  <mjambor@suse.cz>
4281         PR tree-optimization/83141
4282         * tree-sra.c (contains_vce_or_bfcref_p): Move up in the file, also
4283         test for MEM_REFs implicitely changing types with padding.  Remove
4284         inline keyword.
4285         (build_accesses_from_assign): Added contains_vce_or_bfcref_p checks.
4287 2017-12-08  Jakub Jelinek  <jakub@redhat.com>
4289         * config/arc/arc.c (arc_attribute_table): Add exclusions to
4290         the comment.
4291         * config/avr/avr.c (avr_attribute_table): Likewise.
4292         * config/msp430/msp430.c (msp430_attribute_table): Likewise.
4293         * config/rl78/rl78.c (rl78_attribute_table): Likewise.
4294         * config/nds32/nds32.c (nds32_attribute_table): Likewise.
4295         * config/darwin.h (SUBTARGET_ATTRIBUTE_TABLE): Initialize new member
4296         of struct attribute_spec.
4297         * config/i386/cygming.h (SUBTARGET_ATTRIBUTE_TABLE): Likewise.
4299 2017-12-08  Ulrich Weigand  <uweigand@de.ibm.com>
4301         PR target/82960
4302         * config/spu/spu.c (pad_bb): Only check INSN_CODE when INSN_P is true.
4304 2017-12-08  Jan Hubicka  <hubicka@ucw.cz>
4306         PR middle-end/83609
4307         * profile-count.c (profile_count::from_gcov_type): Move from
4308         profile-count.h; handle overflow.
4309         * profile-count.h (profile_count::from_gcov_type): Move offline.
4311 2017-12-08  Segher Boessenkool  <segher@kernel.crashing.org>
4313         PR rtl-optimization/83304
4314         * combine.c (move_deaths): If we do not know where a register died,
4315         search for it.
4317 2017-12-08  Richard Biener  <rguenther@suse.de>
4319         * gimple-loop-interchange.cc (tree_loop_interchange::interchange):
4320         Provide -fopt-info-loop feedback when we interchange in a nest.
4322 2017-06-08  Richard Earnshaw  <rearnsha@arm.com>
4324         * config/arm/driver-arm.c (arm_cpu_table): Use fp-capable product names
4325         for armv6 ARM CPU IDs.
4327 2017-06-08  Richard Earnshaw  <rearnsha@arm.com>
4329         * common/config/arm/arm-common.c: Include <algorithm>.
4330         (INCLUDE_VECTOR): Define.
4331         (compare_opt_names): New function.
4332         (arm_rewrite_selected_arch): Only strip out extensions that can be
4333         expressed through -mfpu.  Sort the remaining extensions
4334         alphabetically.
4336 2017-06-08  Richard Earnshaw  <rearnsha@arm.com>
4338         * config/arm/arm.h (arm_asm_auto_mfpu): Declare.
4339         (ASM_CPU_SPEC_FUNCTIONS): Add new rule asm_auto_mfpu.
4340         (ASM_CPU_SPEC): Use it if -mfpu is set to auto.
4341         * common/config/arm/arm-common.c (arm_asm_auto_mfpu): New function.
4343 2017-06-08  Tristan Gingold  <gindold@adacore.com>
4345         PR ada/81470
4346         * dwarf2out.c (dwarf2out_do_cfi_startproc): Only emit
4347         .cfi_personality or .cfi_lsda if the eh data format is dwarf2.
4349 2017-12-08  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
4351         * config/sol2.h (SOLARIS_ATTRIBUTE_TABLE): Initialize new member
4352         of struct attribute_spec.
4354 2017-12-08  Julia Koval  <julia.koval@intel.com>
4356         * config/i386/avx512vnniintrin.h (_mm512_dpwssds_epi32,
4357         _mm512_mask_dpwssds_epi32, _mm512_maskz_dpwssds_epi32): New intrinsics.
4358         * config/i386/avx512vnnivlintrin.h (_mm256_dpwssds_epi32,
4359         _mm256_mask_dpwssds_epi32, _mm256_maskz_dpwssds_epi32,
4360         _mm_dpwssds_epi32, _mm_mask_dpwssds_epi32,
4361         _mm_maskz_dpwssds_epi32): Ditto.
4363 2017-12-08  Richard Biener  <rguenther@suse.de>
4365         PR tree-optimization/81303
4366         * tree-vect-stmts.c (vect_is_simple_cond): For invariant
4367         conditions try to create a comparison vector type matching
4368         the data vector type.
4369         (vectorizable_condition): Adjust.
4370         * tree-vect-patterns.c (vect_recog_mask_conversion_pattern):
4371         Leave invariant conditions alone in case we can vectorize those.
4373 2017-12-08  Julia Koval  <julia.koval@intel.com>
4375         * config/i386/avx512vnniintrin.h (_mm512_dpwssd_epi32,
4376         _mm512_mask_dpwssd_epi32, _mm512_maskz_dpwssd_epi32): New intrinsics.
4377         * config/i386/avx512vnnivlintrin.h (_mm256_dpwssd_epi32,
4378         _mm256_mask_dpwssd_epi32, _mm256_maskz_dpwssd_epi32, _mm_dpwssd_epi32,
4379         _mm_mask_dpwssd_epi32, _mm_maskz_dpwssd_epi32): Ditto.
4381 2017-12-08  Julia Koval  <julia.koval@intel.com>
4383         * config/i386/avx512vnniintrin.h (_mm512_dpbusds_epi32,
4384         _mm512_mask_dpbusds_epi32, _mm512_maskz_dpbusds_epi32): New.
4385         * config/i386/avx512vnnivlintrin.h (_mm256_dpbusds_epi32,
4386         _mm256_mask_dpbusds_epi32, _mm256_maskz_dpbusds_epi32,
4387         _mm_dpbusds_epi32, _mm_mask_dpbusds_epi32,
4388         _mm_maskz_dpbusds_epi32): New intrinsics.
4390 2017-12-07  Sandra Loosemore  <sandra@codesourcery.com>
4392         * config/nios2/nios2.md (ld<bhw_uns>io): Add splitter for memory
4393         operand.
4394         (ld<bh>io_signed): Likewise.
4395         (st<bhw>io): Likewise.
4396         * config/nios2/predicates.md (ldstio_memory_operand): Allow
4397         SMALL_INT12 constant integer operand.
4399 2017-12-07  Sandra Loosemore  <sandra@codesourcery.com>
4401         * config/nios2/nios2.c (nios2_symbolic_constant_allowed):
4402         Rename to...
4403         (nios2_large_constant_allowed): ...this.  Adjust uses.
4404         (nios2_plus_symbolic_constant_p): Rename to...
4405         (nios2_plus_large_constant_p): ...this.  Adjust uses.
4406         (nios2_legitimate_address_p): Correct CONST_INT handling.
4407         (nios2_symbolic_memory_operand_p): Rename to...
4408         (nios2_large_constant_memory_operand_p): ...this.  Adjust uses.
4409         (nios2_large_constant_p): Check for large constant integers too.
4410         (nios2_split_large_constant): Handle constant integers.
4411         (nios2_split_symbolic_memory_operand): Rename to...
4412         (nios2_split_large_constant_memory_operand): ...this.  Adjust uses.
4413         (nios2_legitimize_constant_address): Handle constant integers.
4414         (r0rel_constant_p): Handle small constant integers.
4415         (nios2_print_operand_address): Handle r0-relative integer addresses.
4416         * config/nios2/nios2-protos.h: Adjust for renamed functions.
4417         * config/nios2/nios2.md: Adjust for renamed functions.
4419 2017-12-07  Andrew Waterman  <andrew@sifive.com>
4421         * config/riscv/riscv.c (TARGET_ASM_SELECT_SECTION): New define.
4422         (TARGET_HAVE_SRODATA_SECTION): New define.
4423         (riscv_select_section): New function.
4425 2017-12-08  Joseph Myers  <joseph@codesourcery.com>
4426             Alexander Monakov  <amonakov@ispras.ru>
4427             Jakub Jelinek  <jakub@redhat.com>
4429         PR target/81906
4430         * config/i386/i386.c (ix86_expand_rint): Handle flag_rounding_math.
4432 2017-12-07  Richard Sandiford  <richard.sandiford@linaro.org>
4434         * config/aarch64/aarch64.c (aarch64_print_address_internal): Return
4435         a bool success value.  Don't call output_operand_lossage here.
4436         (aarch64_print_ldpstp_address): Return a bool success value.
4437         (aarch64_print_operand_address): Call output_addr_const if
4438         aarch64_print_address_internal fails.
4439         (aarch64_print_operand): Don't assert that the mode is 16 bytes for
4440         'y'; call output_operand_lossage instead.  Call output_operand_lossage
4441         if aarch64_print_ldpstp_address fails.
4443 2017-12-07  Richard Sandiford  <richard.sandiford@linaro.org>
4445         * tree-vector-builder.h
4446         (tree_vector_builder::binary_encoded_nelts): Declare.
4447         * tree-vector-builder.c
4448         (tree_vector_builder::binary_encoded_nelts): New function.
4449         * fold-const.c (negate_expr_p): Likewise.
4450         (operand_equal_p, fold_checksum_tree): Likewise.
4451         * tree-loop-distribution.c (const_with_all_bytes_same): Likewise.
4452         * tree.c (integer_zerop, integer_onep, integer_all_onesp, real_zerop)
4453         (real_onep, real_minus_onep, add_expr, initializer_zerop): Likewise.
4454         (uniform_vector_p): Likewise.
4455         * varasm.c (const_hash_1, compare_constant): Likewise.
4456         * tree-ssa-ccp.c: Include tree-vector-builder.h.
4457         (valid_lattice_transition): Operate directly on the VECTOR_CST
4458         encoding.
4459         * ipa-icf.c: Include tree-vector-builder.h.
4460         (sem_variable::equals): Operate directly on the VECTOR_CST encoding.
4461         * print-tree.c (print_node): Print encoding of VECTOR_CSTs.
4463 2017-12-07  Richard Sandiford  <richard.sandiford@linaro.org>
4465         * tree.c (build_vector): Delete.
4466         * tree.h (build_vector): Make static and move into the self-testing
4467         block.
4469 2017-12-07  Richard Sandiford  <richard.sandiford@linaro.org>
4471         * vector-builder.h (vector_builder::derived): New const overload.
4472         (vector_builder::elt): New function.
4473         * tree-vector-builder.h (tree_vector_builder::type): New function.
4474         (tree_vector_builder::apply_step): Declare.
4475         * tree-vector-builder.c (tree_vector_builder::apply_step): New
4476         function.
4477         * gimple-fold.h (tree_vector_builder): Declare.
4478         (gimple_build_vector): Take a tree_vector_builder instead of a
4479         type and vector of elements.
4480         * gimple-fold.c (gimple_build_vector): Likewise.
4481         * tree-vect-loop.c (get_initial_def_for_reduction): Update call
4482         accordingly.
4483         (get_initial_defs_for_reduction): Likewise.
4484         (vectorizable_induction): Likewise.
4486 2017-12-07  Richard Sandiford  <richard.sandiford@linaro.org>
4488         * tree-vector-builder.h
4489         (tree_vector_builder::new_binary_operation): Declare.
4490         * tree-vector-builder.c
4491         (tree_vector_builder::new_binary_operation): New function.
4492         * fold-const.c (fold_relational_const): Use it.
4493         (const_binop): Likewise.  Check that both input vectors have
4494         the same number of elements, thus excluding things like WIDEN_SUM.
4495         Check whether it is possible to operate directly on the encodings
4496         of stepped inputs.
4498 2017-12-07  Richard Sandiford  <richard.sandiford@linaro.org>
4500         * fold-const.c (fold_negate_expr_1): Use tree_vector_builder and
4501         new_unary_operation, operating only on the encoded elements.
4502         (const_unop): Likewise.
4503         (exact_inverse): Likewise.
4504         (distributes_over_addition_p): New function.
4505         (const_binop): Use tree_vector_builder and new_unary_operation
4506         for combinations of VECTOR_CST and INTEGER_CST.  Operate only
4507         on the encoded elements unless the encoding is strided and the
4508         operation does not distribute over addition.
4509         (fold_convert_const):  Use tree_vector_builder and
4510         new_unary_operation.  Operate only on the encoded elements
4511         for truncating integer conversions, or for non-stepped encodings.
4513 2017-12-07  Richard Sandiford  <richard.sandiford@linaro.org>
4515         * config/sparc/sparc.c: Include tree-vector-builder.h.
4516         (sparc_fold_builtin): Use tree_vector_builder instead of build_vector.
4517         * expmed.c: Include tree-vector-builder.h.
4518         (make_tree): Use tree_vector_builder instead of build_vector.
4519         * fold-const.c: Include tree-vector-builder.h.
4520         (const_binop): Use tree_vector_builder instead of build_vector.
4521         (const_unop): Likewise.
4522         (native_interpret_vector): Likewise.
4523         (fold_vec_perm): Likewise.
4524         (fold_ternary_loc): Likewise.
4525         * gimple-fold.c: Include tree-vector-builder.h.
4526         (gimple_fold_stmt_to_constant_1): Use tree_vector_builder instead
4527         of build_vector.
4528         * tree-ssa-forwprop.c: Include tree-vector-builder.h.
4529         (simplify_vector_constructor): Use tree_vector_builder instead
4530         of build_vector.
4531         * tree-vect-generic.c: Include tree-vector-builder.h.
4532         (add_rshift): Use tree_vector_builder instead of build_vector.
4533         (expand_vector_divmod): Likewise.
4534         (optimize_vector_constructor): Likewise.
4535         * tree-vect-loop.c: Include tree-vector-builder.h.
4536         (vect_create_epilog_for_reduction): Use tree_vector_builder instead
4537         of build_vector.  Explicitly use a stepped encoding for
4538         { 1, 2, 3, ... }.
4539         * tree-vect-slp.c: Include tree-vector-builder.h.
4540         (vect_get_constant_vectors): Use tree_vector_builder instead
4541         of build_vector.
4542         (vect_transform_slp_perm_load): Likewise.
4543         (vect_schedule_slp_instance): Likewise.
4544         * tree-vect-stmts.c: Include tree-vector-builder.h.
4545         (vectorizable_bswap): Use tree_vector_builder instead of build_vector.
4546         (vect_gen_perm_mask_any): Likewise.
4547         (vectorizable_call): Likewise.  Explicitly use a stepped encoding.
4548         * tree.c: (build_vector_from_ctor): Use tree_vector_builder instead
4549         of build_vector.
4550         (build_vector_from_val): Likewise.  Explicitly use a duplicate
4551         encoding.
4553 2017-12-07  Richard Sandiford  <richard.sandiford@arm.com>
4555         * doc/generic.texi (VECTOR_CST): Describe new representation of
4556         vector constants.
4557         * vector-builder.h: New file.
4558         * tree-vector-builder.h: Likewise.
4559         * tree-vector-builder.c: Likewise.
4560         * Makefile.in (OBJS): Add tree-vector-builder.o.
4561         * tree.def (VECTOR_CST): Update comment to refer to generic.texi.
4562         * tree-core.h (tree_base): Add a vector_cst field to the u union.
4563         (tree_vector): Change the number of elements to
4564         vector_cst_encoded_nelts.
4565         * tree.h (VECTOR_CST_NELTS): Redefine using TYPE_VECTOR_SUBPARTS.
4566         (VECTOR_CST_ELTS): Delete.
4567         (VECTOR_CST_ELT): Redefine using vector_cst_elt.
4568         (VECTOR_CST_LOG2_NPATTERNS, VECTOR_CST_NPATTERNS): New macros.
4569         (VECTOR_CST_NELTS_PER_PATTERN, VECTOR_CST_DUPLICATE_P): Likewise.
4570         (VECTOR_CST_STEPPED_P, VECTOR_CST_ENCODED_ELTS): Likewise.
4571         (VECTOR_CST_ENCODED_ELT): Likewise.
4572         (vector_cst_encoded_nelts): New function.
4573         (make_vector): Take the values of VECTOR_CST_LOG2_NPATTERNS and
4574         VECTOR_CST_NELTS_PER_PATTERN as arguments.
4575         (vector_cst_int_elt, vector_cst_elt): Declare.
4576         * tree.c: Include tree-vector-builder.h.
4577         (tree_code_size): Abort if passed VECTOR_CST.
4578         (tree_size): Update for new VECTOR_CST layout.
4579         (make_vector): Take the values of VECTOR_CST_LOG2_NPATTERNS and
4580         VECTOR_CST_NELTS_PER_PATTERN as arguments.
4581         (build_vector): Use tree_vector_builder.
4582         (vector_cst_int_elt, vector_cst_elt): New functions.
4583         (drop_tree_overflow): For VECTOR_CST, drop the TREE_OVERFLOW from the
4584         encoded elements and then create the vector in the canonical form.
4585         (check_vector_cst, check_vector_cst_duplicate, check_vector_cst_fill)
4586         (check_vector_cst_stepped, test_vector_cst_patterns): New functions.
4587         (tree_c_tests): Call test_vector_cst_patterns.
4588         * lto-streamer-out.c (DFS::DFS_write_tree_body): Handle the new
4589         VECTOR_CST fields.
4590         (hash_tree): Likewise.
4591         * tree-streamer-out.c (write_ts_vector_tree_pointers): Likewise.
4592         (streamer_write_tree_header): Likewise.
4593         * tree-streamer-in.c (lto_input_ts_vector_tree_pointers): Likewise.
4594         (streamer_alloc_tree): Likewise.  Update call to make_vector.
4595         * fold-const.c (fold_ternary_loc): Avoid using VECTOR_CST_ELTS.
4597 2017-12-07  Richard Sandiford  <richard.sandiford@linaro.org>
4599         * selftest.h (ASSERT_TRUE_AT, ASSERT_FALSE_AT, ASSERT_EQ_AT)
4600         (ASSERT_NE, ASSERT_PRED1): Add underscores to local variable names
4601         * selftest-rtl.h (ASSERT_RTX_EQ, ASSERT_RTX_PTR_EQ): Likewise.
4603 2017-12-07  Bin Cheng  <bin.cheng@arm.com>
4604             Richard Biener  <rguenther@suse.de>
4606         PR tree-optimization/81303
4607         * Makefile.in (gimple-loop-interchange.o): New object file.
4608         * common.opt (floop-interchange): Reuse the option from graphite.
4609         * doc/invoke.texi (-floop-interchange): Ditto.  New document for
4610         -floop-interchange and mention it for -O3.
4611         * opts.c (default_options_table): Enable -floop-interchange at -O3.
4612         * gimple-loop-interchange.cc: New file.
4613         * params.def (PARAM_LOOP_INTERCHANGE_MAX_NUM_STMTS): New parameter.
4614         (PARAM_LOOP_INTERCHANGE_STRIDE_RATIO): New parameter.
4615         * passes.def (pass_linterchange): New pass.
4616         * timevar.def (TV_LINTERCHANGE): New time var.
4617         * tree-pass.h (make_pass_linterchange): New declaration.
4618         * tree-ssa-loop-ivcanon.c (create_canonical_iv): Change to external
4619         interchange.  Record IV before/after increment in new parameters.
4620         * tree-ssa-loop-ivopts.h (create_canonical_iv): New declaration.
4621         * tree-vect-loop.c (vect_is_simple_reduction): Factor out reduction
4622         path check into...
4623         (check_reduction_path): ...New function here.
4624         * tree-vectorizer.h (check_reduction_path): New declaration.
4626 2017-12-07  Vladimir Makarov  <vmakarov@redhat.com>
4628         PR target/83252
4629         PR rtl-optimization/80818
4630         * lra.c (add_regs_to_insn_regno_info): Make a hard reg in CLOBBER
4631         always early clobbered.
4632         * lra-lives.c (process_bb_lives): Check input hard regs for early
4633         clobbered non-operand hard reg.
4635 2017-12-07  Jakub Jelinek  <jakub@redhat.com>
4637         PR middle-end/83164
4638         * tree-cfg.c (verify_gimple_assign_binary): Don't require
4639         types_compatible_p, just that TYPE_MODE is the same.
4641 2017-12-07  Martin Sebor  <msebor@redhat.com>
4643         PR c/81544
4644         * attribs.c (empty_attribute_table): Initialize new member of
4645         struct attribute_spec.
4646         (decl_attributes): Add argument.  Handle mutually exclusive
4647         combinations of attributes.
4648         (selftests::test_attribute_exclusions): New function.
4649         (selftests::attribute_c_tests): Ditto.
4650         * attribs.h (decl_attributes): Add default argument.
4651         * selftest.h (attribute_c_tests): Declare.
4652         * selftest-run-tests.c (selftest::run_tests): Call attribute_c_tests.
4653         * tree-core.h (attribute_spec::exclusions, exclude): New type and
4654         member.
4655         * doc/extend.texi (Common Function Attributes): Update const and pure.
4656         * config/alpha/alpha.c (vms_attribute_table): Initialize new member
4657         of struct attribute_spec.
4658         * config/arc/arc.c (arc_attribute_table): Same.
4659         * config/arm/arm.c (arm_attribute_table): Same.
4660         * config/avr/avr.c ( avr_attribute_table): Same.
4661         * config/bfin/bfin.c (bfin_attribute_table): Same.
4662         * config/cr16/cr16.c (cr16_attribute_table): Same.
4663         * config/epiphany/epiphany.c (epiphany_attribute_table): Same.
4664         * config/h8300/h8300.c (h8300_attribute_table): Same.
4665         * config/i386/i386.c (ix86_attribute_table): Same.
4666         * config/ia64/ia64.c (ia64_attribute_table): Same.
4667         * config/m32c/m32c.c (m32c_attribute_table): Same.
4668         * config/m32r/m32r.c (m32r_attribute_table): Same.
4669         * config/m68k/m68k.c (m68k_attribute_table): Same.
4670         * config/mcore/mcore.c (mcore_attribute_table): Same.
4671         * config/microblaze/microblaze.c (microblaze_attribute_table): Same.
4672         * config/mips/mips.c (mips_attribute_table): Same.
4673         * config/msp430/msp430.c (msp430_attribute_table): Same.
4674         * config/nds32/nds32.c (nds32_attribute_table): Same.
4675         * config/nvptx/nvptx.c (nvptx_attribute_table): Same.
4676         * config/powerpcspe/powerpcspe.c (rs6000_attribute_table): Same.
4677         * config/rl78/rl78.c (rl78__attribute_table): Same.
4678         * config/rs6000/rs6000.c (rs6000_attribute_table): Same.
4679         * onfig/rx/rx.c (rx_attribute_table): Same.
4680         * config/s390/s390.c (s390_handle_vectorbool_attribute): Same.
4681         * config/sh/sh.c (sh_attribute_table): Same.
4682         * config/sparc/sparc.c (sparc_attribute_table): Same.
4683         * config/spu/spu.c (spu_attribute_table): Same.
4684         * config/stormy16/stormy16.c (xstormy16_attribute_table): Same.
4685         * config/v850/v850.c (v850_attribute_table): Same.
4686         * config/visium/visium.c (visium_attribute_table): Same.
4688 2017-12-07  Tamar Christina  <tamar.christina@arm.com>
4690         PR target/82641
4691         * config/arm/arm.c (INCLUDE_STRING): Define.
4692         (arm_last_printed_arch_string, arm_last_printed_fpu_string): New.
4693         (arm_declare_function_name): Conservatively emit .arch, .arch_extensions
4694         and .fpu.
4696 2017-12-07  Michael Matz  <matz@suse.de>
4698         Add unroll and jam pass
4700         * gimple-loop-jam.c: New file.
4701         * Makefile.in (OBJS): Add gimple-loop-jam.o.
4702         * common.opt (funroll-and-jam): New option.
4703         * opts.c (default_options_table): Add unroll-and-jam at -O3.
4704         * params.def (PARAM_UNROLL_JAM_MIN_PERCENT): New param.
4705         (PARAM_UNROLL_JAM_MAX_UNROLL): Ditto.
4706         * passes.def: Add pass_loop_jam.
4707         * timevar.def (TV_LOOP_JAM): Add.
4708         * tree-pass.h (make_pass_loop_jam): Declare.
4709         * cfgloop.c (flow_loop_tree_node_add): Add AT argument.
4710         * cfgloop.h (flow_loop_tree_node_add): Adjust declaration.
4711         * cfgloopmanip.c (duplicate_loop): Add AT argument, adjust call
4712         to flow_loop_tree_node_add.
4713         (duplicate_subloops, copy_loops_to): Append to sibling list.
4714         * cfgloopmanip.h: (duplicate_loop): Adjust declaration.
4715         * doc/invoke.texi (-funroll-and-jam): Document new option.
4716         (unroll-jam-min-percent, unroll-jam-max-unroll): Document new params.
4718 2017-12-07  Richard Biener  <rguenther@suse.de>
4720         PR tree-optimization/83296
4721         PR tree-optimization/67769
4722         * tree-ssa-phiopt.c (conditional_replacement): Do not reset
4723         flow sensitive info in an unrelated BB.
4724         (value_replacement): Use reset_flow_sensitive_info.
4725         (minmax_replacement): Reset flow sensitive info on the def
4726         we move.  Do not reset flow sensitive info in the whole BB
4727         we move the stmt to.
4728         (abs_replacement): Likewise.
4730 2017-12-07  Segher Boessenkool  <segher@kernel.crashing.org>
4732         PR target/43871
4733         * config/rs6000/rs6000.c (rs6000_option_override_internal): Set
4734         rs6000_cpu to the given -mcpu=, or to the default processor.
4736 2017-12-07  Segher Boessenkool  <segher@kernel.crashing.org>
4738         * config/rs6000/rs6000.h (rs6000_cpu_attr): Delete.
4739         * config/rs6000/rs6000.c (rs6000_variable_issue_1): Use rs6000_tune
4740         instead of rs6000_cpu_attr.
4741         (rs6000_adjust_cost): Ditto.
4742         (is_microcoded_insn): Ditto.
4743         (rs6000_adjust_priority): Ditto.
4744         (rs6000_issue_rate): Ditto.
4745         (rs6000_use_sched_lookahead): Ditto.
4746         (rs6000_use_sched_lookahead_guard): Ditto.
4747         (rs6000_sched_reorder): Ditto.
4748         (force_new_group): Ditto.
4749         * config/rs6000/rs6000.md (cpu attribute): Ditto.
4750         (group_ending_nop): Ditto.
4752 2017-12-07  Segher Boessenkool  <segher@kernel.crashing.org>
4754         * config/rs6000/rs6000.opt (rs6000_tune): New variable.
4755         * config/rs6000/rs6000.c (rs6000_option_override_internal): Also set
4756         rs6000_tune.  Use rs6000_tune instead of rs6000_cpu where appropriate.
4757         (rs6000_loop_align): Use rs6000_tune instead of rs6000_cpu where
4758         appropriate.
4759         (rs6000_reassociation_width): Ditto.
4760         (rs6000_emit_epilogue): Ditto.
4761         (rs6000_adjust_cost): Ditto.
4762         (is_microcoded_insn): Ditto.
4763         (is_cracked_insn): Ditto.
4764         (rs6000_adjust_priority): Ditto.
4765         (rs6000_sched_reorder): Ditto.
4766         (rs6000_sched_reorder2): Ditto.
4767         (insn_must_be_first_in_group): Ditto.
4768         (insn_must_be_last_in_group): Ditto.
4769         (rs6000_register_move_cost): Ditto.
4770         * config/rs6000/rs6000.h (rs6000_cpu_attr): Use rs6000_tune instead of
4771         rs6000_cpu.
4773 2017-12-07  Julia Koval  <julia.koval@intel.com>
4775         * config.gcc: Add vaesintrin.h.
4776         * config/i386/i386-builtin-types.def (V64QI_FTYPE_V64QI_V64QI):
4777         New type.
4778         * config/i386/i386-builtin.def (__builtin_ia32_vaesdec_v16qi,
4779         __builtin_ia32_vaesdec_v32qi, __builtin_ia32_vaesdec_v64qi):
4780         New builtins.
4781         * config/i386/i386.c (ix86_expand_args_builtin): Handle new type.
4782         * config/i386/immintrin.h: Include vaesintrin.h.
4783         * config/i386/sse.md (vaesdec_<mode>): New pattern.
4784         * config/i386/vaesintrin.h (_mm256_aesdec_epi128, _mm512_aesdec_epi128,
4785         _mm_aesdec_epi128): New intrinsics.
4787 2017-12-06  David Malcolm  <dmalcolm@redhat.com>
4789         * Makefile.in (C_COMMON_OBJS): Add c-family/c-spellcheck.o.
4790         * spellcheck-tree.c (find_closest_macro_cpp_cb): Move to
4791         c-family/c-spellcheck.cc.
4792         (best_macro_match::best_macro_match): Likewise.
4793         * spellcheck-tree.h
4794         (struct edit_distance_traits<cpp_hashnode *>): Move to
4795         c-family/c-spellcheck.h.
4796         (class best_macro_match): Likewise.
4798 2017-12-06  Jakub Jelinek  <jakub@redhat.com>
4800         PR tree-optimization/83293
4801         * gimple-ssa-strength-reduction.c (insert_initializers): Use
4802         GSI_NEW_STMT instead of GSI_SAME_STMT in gsi_insert_after that
4803         might insert into empty bb.
4805         PR sanitizer/81281
4806         * match.pd ((T)(P + A) - (T)P -> (T) A): Split into separate
4807         simplify for plus with :c added, and pointer_plus without that.
4808         ((T)P - (T)(P + A) -> -(T) A): Likewise.  If type is integral
4809         with undefined overflow and the conversion is not widening,
4810         perform negation in utype and only convert to type afterwards.
4811         ((T)(P + A) - (T)(P + B) -> (T)A - (T)B): Split into separate
4812         simplify for plus with :c added, and pointer_plus without that.
4813         If type is integral with undefined overflow and the conversion is
4814         not widening, perform minus in utype and only convert to type
4815         afterwards.  Move the last pointer_diff_expr simplify into the
4816         two outermost ifs.
4818 2017-12-06  Martin Sebor  <msebor@redhat.com>
4820         PR tree-optimization/82646
4821         * builtins.c (maybe_emit_chk_warning): Use size as the bound for
4822         strncpy, not maxlen.
4824 2017-12-06  Martin Sebor  <msebor@redhat.com>
4826         * doc/invoke.texi (-Wstringop-truncation): Mention attribute
4827         nonstring.
4829         PR tree-optimization/83075
4830         * tree-ssa-strlen.c (handle_builtin_stxncpy): Avoid assuming
4831         strncat/strncpy don't change length of source string.
4833 2017-12-06  Eric Botcazou  <ebotcazou@adacore.com>
4835         Revert
4836         2017-11-29  Martin Aberg  <maberg@gaisler.com>
4838         * config/sparc/sparc.md (divdf3_fix): Add NOP and adjust length
4839         to prevent b2bst errata sequence.
4840         (sqrtdf2_fix): Likewise.
4842 2017-12-06  Jakub Jelinek  <jakub@redhat.com>
4844         PR tree-optimization/81945
4845         * cfgloop.h (FOR_EACH_LOOP_FN): Use FN instead of hardcoding fn.
4846         * tree-cfg.c (move_sese_region_to_fn): If any of the loops moved
4847         to dest_cfun has orig_loop_num set, either remap it to the new
4848         loop number if the loop got moved too, or clear it.
4850 2017-12-05  Steve Ellcey  <sellcey@cavium.com>
4852         * config/aarch64/thunderx2-t99.md (thunderx2t99_branch): Add trap
4853         to reservation.
4854         (thunderx2t99_nothing): New insn reservation.
4855         (thunderx2t99_mrs): New insn reservation.
4856         (thunderx2t99_multiple): New insn reservation.
4857         (thunderx2t99_alu_basi): Add bfx to reservation.
4858         (thunderx2t99_fp_cmp): Add fccmps and fccmpd to reservation.
4860 2017-12-05  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
4862         PR target/82248
4863         * config/arm/arm.md (probe_stack) : Use the 'o' constraint.
4865 2017-12-05  Bin Cheng  <bin.cheng@arm.com>
4867         * tree-ssa-dce.c (simple_dce_from_worklist): Move and rename from
4868         tree-ssa-pre.c::remove_dead_inserted_code.
4869         * tree-ssa-dce.h: New file.
4870         * tree-ssa-pre.c (tree-ssa-dce.h): Include new header file.
4871         (remove_dead_inserted_code): Move and rename to function
4872         tree-ssa-dce.c::simple_dce_from_worklist.
4873         (pass_pre::execute): Update use.
4875 2017-12-05  Richard Biener  <rguenther@suse.de>
4877         PR tree-optimization/83277
4878         * graphite-isl-ast-to-gimple.c (should_copy_to_new_region): Make sure
4879         to code-gen liveout vars.
4881 2017-12-05  Richard Sandiford  <richard.sandiford@linaro.org>
4883         * config/aarch64/aarch64-simd.md (aarch64_simd_bsldi_internal)
4884         (aarch64_simd_bsldi_alt): Check REG_P before GP_REGNUM_P.
4885         (aarch64_cm<optab>di, aarch64_cmtstdi): Add leading "&&" to
4886         split condition.
4888 2017-12-05  Max Filippov  <jcmvbkbc@gmail.com>
4890         * config/xtensa/xtensa.c (xtensa_asan_shadow_offset): New
4891         function.
4892         (TARGET_ASAN_SHADOW_OFFSET): New macro definition.
4893         * config/xtensa/xtensa.h (FRAME_GROWS_DOWNWARD): Set to 1 if
4894         ASAN is enabled.
4896 2017-12-05  Richard Biener  <rguenther@suse.de>
4898         * timevar.def (TV_TREE_RECIP, TV_TREE_SINCOS, TV_TREE_WIDEN_MUL): Add.
4899         * tree-ssa-math-opts.c (pass_data_cse_reciprocal): Use TV_TREE_RECIP.
4900         (pass_data_cse_sincos): Use TV_TREE_SINCOS.
4901         (pass_data_optimize_widening_mul): Use TV_TREE_WIDEN_MUL.
4903 2017-12-05  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
4905         * dbxout.c (dbxout_block): Grow buf to 30 bytes.
4907 2017-12-05  Martin Liska  <mliska@suse.cz>
4908             Jakub Jelinek  <jakub@redhat.com>
4910         * doc/invoke.texi: Document the options.
4911         * flag-types.h (enum sanitize_code): Add
4912         SANITIZE_POINTER_COMPARE and SANITIZE_POINTER_SUBTRACT.
4913         * ipa-inline.c (sanitize_attrs_match_for_inline_p): Add handling
4914         of SANITIZE_POINTER_COMPARE and SANITIZE_POINTER_SUBTRACT.
4915         * opts.c: Define new sanitizer options.
4916         * sanitizer.def (BUILT_IN_ASAN_POINTER_COMPARE): Likewise.
4917         (BUILT_IN_ASAN_POINTER_SUBTRACT): Likewise.
4919 2017-12-05  Julia Koval  <julia.koval@intel.com>
4921         * common/config/i386/i386-common.c (OPTION_MASK_ISA_AVX512VNNI_SET,
4922         OPTION_MASK_ISA_AVX512VNNI_UNSET): New.
4923         (ix86_handle_option): Handle -mavx512vnni.
4924         * config/i386/cpuid.h (bit_AVX512VNNI): New bit.
4925         * config/i386/driver-i386.c (host_detect_local_cpu): Handle new bit.
4926         * config/i386/i386-c (__AVX512VNNI__): New.
4927         * config/i386/i386.c (ix86_target_string): Handle new option.
4928         (ix86_valid_target_attribute_inner_p): Handle new option.
4929         * config/i386/i386.h (TARGET_AVX512VNNI, TARGET_AVX512VNNI_P): New.
4930         * config/i386/i386.opt (mavx512vnni): New option.
4932 2017-12-01  Jan Hubicka  <hubicka@ucw.cz>
4934         PR target/81616
4935         * athlon.md: Disable for generic.
4936         * haswell.md: Enable for generic.
4937         * i386.c (ix86_sched_init_global): Add core hooks for generic.
4938         * x86-tune-sched.c (ix86_issue_rate): Increase issue rate for generic
4939         to 4.
4940         (ix86_adjust_cost): Move generic to haswell path.
4942 2017-12-04  Eric Botcazou  <ebotcazou@adacore.com>
4944         * config/sparc/sparc.c (sparc_do_work_around_errata): Use mem_ref
4945         instead of MEM_P in a couple more places.  Fix formatting issues.
4947 2017-12-04  Jim Wilson  <jimw@sifive.com>
4949         * config/riscv/riscv.c (riscv_for_each_saved_reg): Use GP_REG_LAST
4950         instead of GP_REG_LAST-1.
4951         (riscv_adjust_libcall_cfi_prologue): Likewise.
4952         (riscv_adjust_libcall_cri_epilogue): Likewise.
4953         * config/riscv/riscv.h (CALL_USED_REGISTERS): Change a6 to t6 in
4954         comment.
4956 2017-12-04  Luis Machado  <luis.machado@linaro.org>
4958         * ipa-pure-const.c (check_decl): Add missing newline.
4959         (state_from_flags): Likewise.
4961 2017-12-04  Jeff Law  <law@redhat.com>
4963         PR tree-optimizatin/78496
4964         * gimple-ssa-evrp-analyze.h
4965         (evrp_range_analyzer::get_vr_values): Simplify.
4966         * gimple-ssa-evrp-analyze.c: Corresponding changes.
4967         * tree-ssa-dom.c: Include alloc-pool.h, tree-vrp.h, vr-values.h
4968         and gimple-ssa-evrp-analyze.h.
4969         (dom_opt_dom_walker class): Add evrp_range_analyzer member.
4970         (simplify_stmt_for_jump_threading): Copy a blob of code from
4971         tree-vrp.c to use ranges to simplify statements.
4972         (dom_opt_dom_walker::before_dom_children): Call
4973         evrp_range_analyzer::{enter,record_ranges_from_stmt} methods.
4974         (dom_opt_dom_walker::after_dom_children): Similarly for
4975         evrp_range_analyzer::leave.
4976         (dom_opt_dom_walker::optimize_stmt): Use EVRP ranges to optimize
4977         conditionals.
4979         * gimple-ssa-evrp-analyze.c
4980         (evrp_range_analyzer::extract_range_from_stmt):  Always use
4981         vr_values::update_value_range so preexisting range info is
4982         medged with new range info, even if the new range is VR_VARYING.
4984 2017-12-04  Segher Boessenkool  <segher@kernel.crashing.org>
4986         * combine.c: Adjust comment.
4987         (use_crosses_set_p): Delete.
4988         (can_combine_p): Use modified_between_p instead of use_crosses_set_p.
4989         (try_combine): Ditto.
4991 2017-12-04  Richard Biener  <rguenther@suse.de>
4993         PR tree-optimization/83255
4994         * graphite-isl-ast-to-gimple.c (translate_isl_ast_node_for):
4995         Re-add zero-iteration check.
4997 2017-12-04  Segher Boessenkool  <segher@kernel.crashing.org>
4999         PR rtl-optimization/83245
5000         * lra.c (collect_non_operand_hard_regs): Treat clobbers of non-operand
5001         hard registers as earlyclobber, also if not in an asm.
5003 2017-12-04  Segher Boessenkool  <segher@kernel.crashing.org>
5005         PR bootstrap/83265
5006         Revert
5007         2017-12-01  Segher Boessenkool  <segher@kernel.crashing.org>
5009         PR target/43871
5010         * config/rs6000/rs6000.c (rs6000_option_override_internal): Set
5011         rs6000_cpu based on cpu_index, not tune_index.
5013 2017-12-04  Richard Biener  <rguenther@suse.de>
5015         PR tree-optimization/83238
5016         * graphite-scop-detection.c (scop_detection::merge_sese): Make
5017         code match comment, rejecting invalid SESE regions.
5019 2017-12-03  John David Anglin  <danglin@gcc.gnu.org>
5021         * config/pa/pa.c (pa_legitimate_address_p): For scaled indexing,
5022         require base operand is a REG_POINTER prior to reload on targets
5023         with non-equivalent space registers.
5025 2017-12-01  Jan Hubicka  <hubicka@ucw.cz>
5027         * ipa-cp.c (ipcp_lattice<valtype>::print): Update dumping.
5028         (update_specialized_profile): Fix updating of counts.
5029         (perhaps_add_new_callers): Likewise.
5031 2017-12-01  Jan Hubicka  <hubicka@ucw.cz>
5033         PR target/81616
5034         * x86-tune.def: Remove obsolette FIXMEs.
5035         (X86_TUNE_PARTIAL_FLAG_REG_STALL): Disable for generic
5036         (X86_TUNE_FUSE_CMP_AND_BRANCH_32, X86_TUNE_FUSE_CMP_AND_BRANCH_64,
5037         X86_TUNE_FUSE_CMP_AND_BRANCH_SOFLAGS, X86_TUNE_FUSE_ALU_AND_BRANCH):
5038         Enable for generic.
5039         (X86_TUNE_PAD_RETURNS): Disable for generic.
5041 2017-12-02  Jakub Jelinek  <jakub@redhat.com>
5043         PR tree-optimization/83170
5044         PR tree-optimization/83241
5045         * gimple-ssa-store-merging.c
5046         (imm_store_chain_info::try_coalesce_bswap): Update vuse field from
5047         gimple_vuse (ins_stmt) in case it has changed.
5048         (imm_store_chain_info::output_merged_store): Likewise.
5050         * tree-chkp.c (chkp_compute_bounds_for_assignment): Handle
5051         POINTER_DIFF_EXPR.
5053         PR c++/81212
5054         * tree-cfg.c (pass_warn_function_return::execute): Handle
5055         __builtin_ubsan_handle_missing_return like __builtin_unreachable
5056         with BUILTINS_LOCATION.
5058         PR target/78643
5059         PR target/80583
5060         * expr.c (get_inner_reference): If DECL_MODE of a non-bitfield
5061         is BLKmode for vector field with vector raw mode, use TYPE_MODE
5062         instead of DECL_MODE.
5064         * config/i386/i386-protos.h (standard_sse_constant_opcode): Change
5065         last argument to rtx pointer.
5066         * config/i386/i386.c (standard_sse_constant_opcode): Replace X argument
5067         with OPERANDS.  For AVX+ 128-bit VEX encoded instructions over 256-bit
5068         or 512-bit.  If setting EXT_REX_SSE_REG_P, use EVEX encoded insn
5069         depending on the chosen ISAs.
5070         * config/i386/i386.md (*movxi_internal_avx512f, *movoi_internal_avx,
5071         *movti_internal, *movdi_internal, *movsi_internal, *movtf_internal,
5072         *movdf_internal, *movsf_internal): Adjust standard_sse_constant_opcode
5073         callers.
5074         * config/i386/sse.md (mov<mode>_internal): Likewise.
5075         * config/i386/mmx.md (*mov<mode>_internal): Likewise.
5077 2017-12-01  Segher Boessenkool  <segher@kernel.crashing.org>
5079         * doc/invoke.texi (-dp): Say that instruction cost is printed as well.
5081 2017-12-01  Segher Boessenkool  <segher@kernel.crashing.org>
5083         * config/rs6000/rs6000-protos.h (emit_fusion_addis): Remove last two
5084         parameters from prototype.
5085         * config/rs6000/rs6000.c (emit_fusion_addis): Remove last two
5086         parameters.  Don't print a comment.
5087         (emit_fusion_gpr_load): Adjust.
5088         (emit_fusion_load_store): Adjust.
5089         * config/rs6000/rs6000.md (*fusion_p9_<mode>_constant): Adjust.
5090         * config/rs6000/vsx.md (two peepholes): Print the "vector load fusion"
5091         comment on the second line.
5093 2017-12-01  Segher Boessenkool  <segher@kernel.crashing.org>
5095         PR target/43871
5096         * config/rs6000/rs6000.c (rs6000_option_override_internal): Set
5097         rs6000_cpu based on cpu_index, not tune_index.
5099 2017-12-01  Segher Boessenkool  <segher@kernel.crashing.org>
5101         * final.c (output_asm_name): Print insn_cost.  Shorten output.  Print
5102         which_alternative instead of which_alternative + 1.
5103         (output_asm_insn): Print an extra tab if the template is short.
5105 2017-12-01  Jim Wilson  <jimw@sifive.com>
5107         * common.opt (use_gnu_debug_info_extensions): Delete DWARF_DEBUG from
5108         comment.
5109         * config/vx-common.h (DWARF_DEBUGGING_INFO): Delete undef.
5110         * doc/tm.texi.in (PREFERRED_DEBUGGING_TYPE): Delete DWARF_DEBUG
5111         reference.
5112         * doc/tm.texi: Regenerate.
5114 2017-12-01  Michael Meissner  <meissner@linux.vnet.ibm.com>
5116         PR target/81959
5117         * config/rs6000/rs6000.c (rs6000_address_for_fpconvert): Check for
5118         whether we can allocate pseudos before trying to fix an address.
5119         * config/rs6000/rs6000.md (float_<mode>si2_hw): Make sure the
5120         memory address is indexed or indirect.
5121         (floatuns_<mode>si2_hw2): Likewise.
5123 2017-12-01  Jason Merrill  <jason@redhat.com>
5125         * Makefile.in (TAGS): Add c-family/*.cc.
5127 2017-12-01  Wilco Dijkstra  <wdijkstr@arm.com>
5129         * config/aarch64/aarch64.md (call_insn): Use %c rather than %a.
5130         (call_value_insn): Likewise.
5131         (sibcall_insn): Likewise.
5132         (sibcall_value_insn): Likewise.
5133         (movsi_aarch64): Likewise.
5134         (movdi_aarch64): Likewise.
5135         (add_losym_): Likewise.
5136         (ldr_got_small_): Likewise.
5137         (ldr_got_small_sidi): Likewise.
5138         (ldr_got_small_28k_): Likewise.
5139         (ldr_got_small_28k_sidi): Likewise.
5140         * config/aarch64/aarch64.c (aarch64_print_address_internal):
5141         Move output_addr_const to symbolic case. Add error check.
5143 2017-12-01  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
5145         * config/s390/predicates.md (plus16_Q_operand): New predicate.
5146         * config/s390/s390.md: Disable MVC merging peephole if it would
5147         disable operand forwarding.
5148         (new peephole2): Split MVCs if it would turn them into up to 2
5149         forwardable MVCs.
5151 2017-12-01  Richard Biener  <rguenther@suse.de>
5153         PR tree-optimization/83232
5154         * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Fix
5155         detection of same access. Instead of breaking the group here
5156         do not consider the duplicate.  Add comment explaining real fix.
5158 2017-12-01  Jonathan Wakely  <jwakely@redhat.com>
5160         * doc/md.texi (Insn Splitting): Fix "central flowgraph" typo.
5162 2017-12-01  Sudakshina Das  <sudi.das@arm.com>
5164         * config/arm/vfp.md (*movhf_vfp_fp16): Add conds attribute.
5166 2017-12-01  Jakub Jelinek  <jakub@redhat.com>
5168         * function.h (struct function): Remove cilk_frame_decl,
5169         is_cilk_function and calls_cilk_spawn fields.
5170         * tree-inline.h (struct copy_body_data): Remove remap_var_for_cilk
5171         field.
5172         * omp-simd-clone.c (simd_clone_clauses_extract): Don't clear
5173         cilk_elemental field.
5174         * cgraph.h (struct cgraph_simd_clone): Remove cilk_elemental field.
5175         * target.def: Adjust comment.
5176         * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
5177         Don't test cilk_elemental.
5179         PR tree-optimization/83233
5180         * gimple-ssa-store-merging.c (nop_stats, bswap_stats): Use
5181         bswap_stat name for the struct.
5183         PR c/79153
5184         * tree.h (SWITCH_BREAK_LABEL_P): Define.
5185         * gimplify.c (collect_fallthrough_labels): Handle GIMPLE_BIND
5186         starting with a GIMPLE_SWITCH and ending with GIMPLE_LABEL with
5187         SWITCH_BREAK_LABEL_P set on the label.
5188         (gimplify_switch_expr): Set SWITCH_BREAK_LABEL_P on the label
5189         added for default case if it was missing and not all cases covered.
5190         Wrap GIMPLE_SWITCH and the switch_body_seq into a GIMPLE_BIND if
5191         switch_body_seq ends with a GIMPLE_LABEL with SWITCH_BREAK_LABEL_P
5192         set on the label.
5193         * tree-chrec.c (evolution_function_is_univariate_p): Add return true;
5194         to avoid -Wimplicit-fallthrough warning.
5195         * config/i386/i386.c (ix86_expand_special_args_builtin): Add
5196         FALLTHRU comment to avoid -Wimplicit-fallthrough warning.
5198         PR tree-optimization/83221
5199         * tree-ssa-reassoc.c (sort_by_operand_rank): Shift bb_rank
5200         down by 16.
5201         (init_reassoc): Formatting fix.
5203         PR sanitizer/81275
5204         * tree-cfg.c (group_case_labels_stmt): Don't optimize away
5205         C++ FE implicitly added __builtin_unreachable () until -Wreturn-type
5206         is diagnosed.
5208         PR sanitizer/83219
5209         * tree-cfg.c: Include asan.h.
5210         (gimple_seq_unreachable_p): Return false for -fsanitize=unreachable.
5212 2017-12-01  Sergey Shalnov  <Sergey.Shalnov@intel.com>
5214         * config/i386/i386.md: Fix AVX512 register width in AVX512 instruction.
5216 2017-12-01  Sergey Shalnov  <Sergey.Shalnov@intel.com>
5218         * config/i386/i386.c (standard_sse_constant_opcode): Fix wrong form for
5219         vpcmpeqd instruction.
5221 2017-12-01  Sergey Shalnov  <Sergey.Shalnov@intel.com>
5223         * config/i386/i386.c (standard_sse_constant_opcode): Fix
5224         registers type for 128bit mode.
5226 2017-11-30  Jakub Jelinek  <jakub@redhat.com>
5228         * spellcheck-tree.c (test_find_closest_identifier): Use ; instead
5229         of ;;.
5230         * gengtype-state.c (read_state_pair): Likewise.
5231         * gimple-fold.c (gimple_fold_builtin_string_compare): Likewise.
5232         * sel-sched-dump.c (dump_insn_rtx_1): Likewise.
5233         * ipa-cp.c (intersect_aggregates_with_edge): Likewise.
5234         * ifcvt.c (noce_try_store_flag_constants): Likewise.
5235         * tree-ssa-ccp.c (ccp_finalize): Likewise.
5236         * omp-grid.c (grid_process_kernel_body_copy): Likewise.
5237         * builtins.c (fold_builtin_3): Likewise.
5238         * graphite-scop-detection.c
5239         (scop_detection::stmt_has_simple_data_refs_p): Likewise.
5240         * hsa-gen.c (hsa_function_representation::hsa_function_representation):
5241         Likewise.
5243 2017-12-01  Maxim Ostapenko  <m.ostapenko@samsung.com>
5245         PR sanitizer/81697
5246         * asan.c (asan_protect_global): Add new ignore_decl_rtl_set_p
5247         parameter. Return true if ignore_decl_rtl_set_p is true and other
5248         conditions are satisfied.
5249         * asan.h (asan_protect_global): Add new parameter.
5250         * varasm.c (categorize_decl_for_section): Pass true as second parameter
5251         to asan_protect_global calls.
5253 2017-11-30  Jim Wilson  <jimw@sifive.com>
5255         * doc/invoke.texi (RISC-V Options): Delete nonexistent -mmemcpy and
5256         -mno-memcpy options.  For -mplt, -mfdiv, -mdiv, -msave-restore, and
5257         -mstrict-align, add info on default value.  Delete redundant lines for
5258         -mabi.  Add missing -mexplicit-relocs docs.
5260 2017-11-30  Claudiu Zissulescu  <claziss@synopsys.com>
5262         * config/arc/arc.md (trap): New pattern.
5264 2017-11-30  Claudiu Zissulescu  <claziss@synopsys.com>
5266         * config/arc/arc.c (hwloop_optimize): Prevent the last ZOL
5267         instruction to end into a delay slot.
5268         * config/arc/arc.md (cond_delay_insn): Check if the instruction
5269         can be placed into a delay slot against reg_note.
5271 2017-11-30  Claudiu Zissulescu  <claziss@synopsys.com>
5273         * config/arc/arc.c (hwloop_optimize): Update hw-loop's end/start
5274         labels number of usages.
5276 2017-11-30  Claudiu Zissulescu  <claziss@synopsys.com>
5278         * config/arc/arc.c (arc_cannot_substitue_mem_equiv_p): New
5279         function.
5280         (TARGET_CANNOT_SUBSTITUTE_MEM_EQUIV_P): Define.
5282 2017-11-30  Jakub Jelinek  <jakub@redhat.com>
5284         PR target/83210
5285         * internal-fn.c (expand_mul_overflow): Optimize unsigned
5286         multiplication by power of 2 constant into two shifts + comparison.
5288 2017-11-30  Jan Hubicka  <hubicka@ucw.cz>
5290         PR target/81616
5291         * config/i386/x86-tune-costs.h (generic_cost): Revise for modern CPUs.
5293 2017-11-30  Richard Biener  <rguenther@suse.de>
5295         PR tree-optimization/83202
5296         * tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Add
5297         allow_peel argument and guard peeling.
5298         (canonicalize_loop_induction_variables): Likewise.
5299         (canonicalize_induction_variables): Pass false.
5300         (tree_unroll_loops_completely_1): Pass unroll_outer to disallow
5301         peeling from cunrolli.
5303 2017-11-29  Segher Boessenkool  <segher@kernel.crashing.org>
5305         * combine.c (try_combine): Print a message to dump file whenever
5306         I0, I1, or I2 cannot be combined into I3.
5308 2017-11-29  Segher Boessenkool  <segher@kernel.crashing.org>
5310         PR rtl-optimization/83156
5311         PR rtl-optimization/82621
5312         * combine.c (try_combine): Don't split an I2 if one of the dests is
5313         set again before I3.  Allow unused dests.
5315 2017-11-29  Segher Boessenkool  <segher@kernel.crashing.org>
5317         * config/rs6000/rs6000.md (*add<mode>3_carry_in_internal2): New.
5319 2017-11-29  Vladimir Makarov  <vmakarov@redhat.com>
5321         PR rtl-optimization/80818
5322         * lra.c (collect_non_operand_hard_regs): New arg insn.  Pass it
5323         recursively.  Use insn code for clobber.
5324         (lra_set_insn_recog_data): Pass the new arg to
5325         collect_non_operand_hard_regs.
5326         (add_regs_to_insn_regno_info): Pass insn instead of uid.  Use insn
5327         code for clobber.
5328         (lra_update_insn_regno_info): Pass insn to
5329         add_regs_to_insn_regno_info.
5331 2017-11-29  Jim Wilson  <jimw@sifive.com>
5332             Andrew Waterman  <andrew@sifive.com>
5334         * config/riscv/riscv.c (SINGLE_SHIFT_COST): New.
5335         (riscv_rtx_costs): Case ZERO_EXTRACT, match new pattern, and return
5336         SINGLE_SHIFT_COST.  Case LT and ZERO_EXTEND, likewise.  Case ASHIFT,
5337         use SINGLE_SHIFT_COST.
5338         * config/riscv/riscv.md (lshrsi3_zero_extend_1): New.
5339         (lshrsi3_zero_extend_2, lshrsi3_zero_extend_3): New.
5341 2017-11-29  Julia Koval  <julia.koval@intel.com>
5343         * config/i386/avx512vbmi2intrin.h (_mm512_shldv_epi16,
5344         _mm512_mask_shldv_epi16, _mm512_maskz_shldv_epi16, _mm512_shldv_epi32,
5345         _mm512_mask_shldv_epi32, _mm512_maskz_shldv_epi32, _mm512_shldv_epi64,
5346         _mm512_mask_shldv_epi64, _mm512_maskz_shldv_epi64): New intrinsics.
5347         * config/i386/avx512vbmi2vlintrin.h (_mm256_shldv_epi16,
5348         _mm256_mask_shldv_epi16, _mm256_maskz_shldv_epi16, _mm256_shldv_epi32,
5349         _mm256_mask_shldv_epi32, _mm256_maskz_shldv_epi32, _mm256_shldv_epi64,
5350         _mm256_mask_shldv_epi64, _mm256_maskz_shldv_epi64, _mm_shldv_epi16,
5351         _mm_mask_shldv_epi16, _mm_maskz_shldv_epi16, _mm_shldv_epi32,
5352         _mm_mask_shldv_epi32, _mm_maskz_shldv_epi32, _mm_shldv_epi64,
5353         _mm_mask_shldv_epi64, _mm_maskz_shldv_epi64): Ditto.
5354         * config/i386/i386-builtin.def (__builtin_ia32_vpshldv_v32hi,
5355         __builtin_ia32_vpshldv_v32hi_mask, __builtin_ia32_vpshldv_v32hi_maskz,
5356         __builtin_ia32_vpshldv_v16hi, __builtin_ia32_vpshldv_v16hi_mask,
5357         __builtin_ia32_vpshldv_v16hi_maskz, __builtin_ia32_vpshldv_v8hi,
5358         __builtin_ia32_vpshldv_v8hi_mask, __builtin_ia32_vpshldv_v8hi_maskz,
5359         __builtin_ia32_vpshldv_v16si, __builtin_ia32_vpshldv_v16si_mask,
5360         __builtin_ia32_vpshldv_v16si_maskz, __builtin_ia32_vpshldv_v8si,
5361         __builtin_ia32_vpshldv_v8si_mask, __builtin_ia32_vpshldv_v8si_maskz,
5362         __builtin_ia32_vpshldv_v4si, __builtin_ia32_vpshldv_v4si_mask,
5363         __builtin_ia32_vpshldv_v4si_maskz, __builtin_ia32_vpshldv_v8di,
5364         __builtin_ia32_vpshldv_v8di_mask, __builtin_ia32_vpshldv_v8di_maskz,
5365         __builtin_ia32_vpshldv_v4di, __builtin_ia32_vpshldv_v4di_mask,
5366         __builtin_ia32_vpshldv_v4di_maskz, __builtin_ia32_vpshldv_v2di,
5367         __builtin_ia32_vpshldv_v2di_mask,
5368         __builtin_ia32_vpshldv_v2di_maskz): New builtins.
5369         * config/i386/sse.md (vpshldv_<mode>, vpshldv_<mode>_mask,
5370         vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): New patterns.
5372 2017-11-29  Julia Koval  <julia.koval@intel.com>
5374         * config/i386/avx512vbmi2intrin.h (_mm512_shrdv_epi16,
5375         _mm512_mask_shrdv_epi16, _mm512_maskz_shrdv_epi16, _mm512_shrdv_epi32,
5376         _mm512_mask_shrdv_epi32, _mm512_maskz_shrdv_epi32, _mm512_shrdv_epi64,
5377         _mm512_mask_shrdv_epi64, _mm512_maskz_shrdv_epi64): New intrinsics.
5378         * config/i386/avx512vbmi2vlintrin.h (_mm256_shrdv_epi16,
5379         _mm256_mask_shrdv_epi16, _mm256_maskz_shrdv_epi16, _mm256_shrdv_epi32,
5380         _mm256_mask_shrdv_epi32, _mm256_maskz_shrdv_epi32, _mm256_shrdv_epi64,
5381         _mm256_mask_shrdv_epi64, _mm256_maskz_shrdv_epi64, _mm_shrdv_epi16,
5382         _mm_mask_shrdv_epi16, _mm_maskz_shrdv_epi16, _mm_shrdv_epi32,
5383         _mm_mask_shrdv_epi32, _mm_maskz_shrdv_epi32, _mm_shrdv_epi64,
5384         _mm_mask_shrdv_epi64, _mm_maskz_shrdv_epi64): Ditto.
5385         * config/i386/i386-builtin-types.def (V32HI_FTYPE_V32HI_V32HI_V32HI,
5386         V32HI_FTYPE_V32HI_V32HI_V32HI_INT, V16HI_FTYPE_V16HI_V16HI_V16HI_INT,
5387         V8HI_FTYPE_V8HI_V8HI_V8HI_INT, V8SI_FTYPE_V8SI_V8SI_V8SI_INT,
5388         V4SI_FTYPE_V4SI_V4SI_V4SI_INT, V8DI_FTYPE_V8DI_V8DI_V8DI,
5389         V8DI_FTYPE_V8DI_V8DI_V8DI_INT, V4DI_FTYPE_V4DI_V4DI_V4DI_INT,
5390         V16SI_FTYPE_V16SI_V16SI_V16SI, V16SI_FTYPE_V16SI_V16SI_V16SI_INT,
5391         V2DI_FTYPE_V2DI_V2DI_V2DI_INT): New types.
5392         * config/i386/i386.c (ix86_expand_args_builtin): Handle new types.
5393         * config/i386/sse.md (vpshrdv_<mode>, vpshrdv_<mode>_mask,
5394         vpshrdv_<mode>_maskz, vpshrdv_<mode>_maskz_1): New pattern.
5396 2017-11-29  Daniel Cederman  <cederman@gaisler.com>
5398         * config/sparc/sparc.c (sparc_do_work_around_errata): Treat the
5399         movsi_pic_gotdata_op instruction as a load for the UT699 errata
5400         workaround.
5402 2017-11-29  Martin Aberg  <maberg@gaisler.com>
5404         * config/sparc/sparc.md (divdf3_fix): Add NOP and adjust length
5405         to prevent b2bst errata sequence.
5406         (sqrtdf2_fix): Likewise.
5408 2017-11-29  Daniel Cederman  <cederman@gaisler.com>
5410         * config/sparc/sparc.c (fpop_reg_depend_p): New function.
5411         (div_sqrt_insn_p): New function.
5412         (sparc_do_work_around_errata): Insert NOP instructions to
5413         prevent sequences that could trigger the TN-0013 errata for
5414         certain LEON3 processors.
5415         (pass_work_around_errata::gate): Also test sparc_fix_lost_divsqrt.
5416         (sparc_option_override): Set sparc_fix_lost_divsqrt appropriately.
5417         * config/sparc/sparc.md (fix_lost_divsqrt): New attribute.
5418         (in_branch_delay): Prevent div and sqrt in delay slot if
5419         fix_lost_divsqrt.
5420         * config/sparc/sparc.opt (sparc_fix_lost_divsqrt): New variable.
5422 2017-11-29  Daniel Cederman  <cederman@gaisler.com>
5424         * config/sparc/sparc.c (atomic_insn_p): New function.
5425         (sparc_do_work_around_errata): Insert NOP instructions to
5426         prevent sequences that could trigger the TN-0010 errata for
5427         UT700.
5428         * config/sparc/sync.md (atomic_compare_and_swap_leon3_1): Make
5429         instruction referable in atomic_insns_p.
5431 2017-11-29  Daniel Cederman  <cederman@gaisler.com>
5433         * config/sparc/sync.md (swapsi): 16-byte align if sparc_fix_gr712rc.
5434         (atomic_compare_and_swap_leon3_1): Likewise.
5435         (ldstub): Likewise.
5437 2017-11-29  Daniel Cederman  <cederman@gaisler.com>
5439         * config/sparc/sparc.c (fpop_insn_p): New function.
5440         (sparc_do_work_around_errata): Insert NOP instructions to
5441         prevent sequences that could trigger the TN-0012 errata for
5442         GR712RC.
5443         (pass_work_around_errata::gate): Also test sparc_fix_gr712rc.
5444         * config/sparc/sparc.md (fix_gr712rc): New attribute.
5445         (in_branch_annul_delay): Prevent floating-point instructions
5446         in delay slot of annulled integer branch.
5448 2017-11-29  Richard Biener  <rguenther@suse.de>
5450         PR tree-optimization/83202
5451         * tree-vect-slp.c (scalar_stmts_set_t): New typedef.
5452         (bst_fail): Use it.
5453         (vect_analyze_slp_cost_1): Add visited set, do not account SLP
5454         nodes vectorized to the same stmts multiple times.
5455         (vect_analyze_slp_cost): Allocate a visited set and pass it down.
5456         (vect_analyze_slp_instance): Adjust.
5457         (scalar_stmts_to_slp_tree_map_t): New typedef.
5458         (vect_schedule_slp_instance): Add a map recording the SLP node
5459         representing the vectorized stmts for a set of scalar stmts.
5460         Avoid code-generating redundancies.
5461         (vect_schedule_slp): Allocate map and pass it down.
5463 2017-11-29  Nathan Sidwell  <nathan@acm.org>
5465         PR c++/83187
5466         * tree.c (build_complex_type): Fix canonicalization.  Only fill in
5467         type if it is new.
5469 2017-11-29  Wilco Dijkstra  <wdijkstr@arm.com>
5471         * config/aarch64/aarch64.c (aarch64_print_operand): Add new
5472         cases for printing LDP/STP memory addresses.
5473         (aarch64_print_address_internal): Renamed from
5474         aarch64_print_operand_address, added parameter, add Pmode check.
5475         (aarch64_print_ldpstp_address): New function for LDP/STP addresses.
5476         (aarch64_print_operand_address): Indirect to
5477         aarch64_print_address_internal.
5478         * config/aarch64/aarch64-simd.md (store_pair_lanes): Use new
5479         'y' operand output specifier.
5481 2017-11-29  Jakub Jelinek  <jakub@redhat.com>
5483         PR middle-end/83185
5484         * tree.c (build_simple_mem_ref_loc): Handle
5485         get_addr_base_and_unit_offset returning a MEM_REF.
5487         PR middle-end/80929
5488         * rtlanal.c (seq_cost): For non-single_set insns try to use insn_cost.
5490         PR target/80819
5491         * config/i386/sse.md (vec_concatv2di): Remove * from (=Yr,0,*rm)
5492         alternative.
5494 2017-11-28  David Malcolm  <dmalcolm@redhat.com>
5496         * diagnostic-show-locus.c (layout::print_trailing_fixits): Handle
5497         m_x_offset.
5498         (layout::move_to_column): Likewise.
5500 2017-11-28  Jakub Jelinek  <jakub@redhat.com>
5502         PR sanitizer/81275
5503         * tree.c (block_may_fallthru): Return false if SWITCH_ALL_CASES_P
5504         is set on SWITCH_EXPR and !block_may_fallthru (SWITCH_BODY ()).
5506 2017-11-28  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
5507             Martin Jambor  <mjambor@suse.cz>
5509         PR ipa/82808
5510         * tree.h (expr_type_first_operand_type_p): Declare
5511         * tree.c (expr_type_first_operand_type_p): New function.
5512         * ipa-prop.h (ipa_get_type): Allow i to be out of bounds.
5513         (ipa_value_from_jfunc): Adjust declaration.
5514         * ipa-cp.c (ipa_get_jf_pass_through_result): New parameter RES_TYPE.
5515         Use it as result type for arithmetics, unless it is NULL in which case
5516         be more conservative.
5517         (ipa_value_from_jfunc): New parameter PARM_TYPE, pass it to
5518         ipa_get_jf_pass_through_result.
5519         (propagate_vals_across_pass_through): Likewise.
5520         (propagate_scalar_across_jump_function): New parameter PARM_TYPE, pass
5521         is to propagate_vals_across_pass_through.
5522         (propagate_constants_across_call): Pass PARM_TYPE to
5523         propagate_scalar_across_jump_function.
5524         (find_more_scalar_values_for_callers_subset): Pass parameter type to
5525         ipa_value_from_jfunc.
5526         (cgraph_edge_brings_all_scalars_for_node): Likewise.
5527         * ipa-fnsummary.c (evaluate_properties_for_edge): Renamed parms_info
5528         to caller_parms_info, pass parameter type to ipa_value_from_jfunc.
5529         * ipa-prop.c (try_make_edge_direct_simple_call): New parameter
5530         target_type, pass it to ipa_value_from_jfunc.
5531         (update_indirect_edges_after_inlining): Pass parameter type to
5532         try_make_edge_direct_simple_call.
5534 2017-11-28  Jeff Law  <law@redhat.com>
5536         * gimple-ssa-evrp-analyze.c
5537         (evrp_range_analyzer::record_ranges_from_phis): Only use SCEV to
5538         refine ranges if scev_initialized_p returns true.
5539         * vr-values.c (vr_values::extract_range_from_phi_node): Likewise.
5541 2017-11-28  Julia Koval  <julia.koval@intel.com>
5543         * config/i386/avx512vbmi2intrin.h (_mm512_shrdi_epi16,
5544         _mm512_mask_shrdi_epi16, _mm512_maskz_shrdi_epi16, _mm512_shrdi_epi32,
5545         _mm512_mask_shrdi_epi32, _mm512_maskz_shrdi_epi32, _mm512_shrdi_epi64,
5546         _mm512_mask_shrdi_epi64, _mm512_maskz_shrdi_epi64): New intrinsics.
5547         * config/i386/avx512vbmi2vlintrin.h (_mm256_shrdi_epi16,
5548         _mm256_mask_shrdi_epi16, _mm256_maskz_shrdi_epi16,
5549         _mm256_mask_shrdi_epi32, _mm256_maskz_shrdi_epi32, _mm256_shrdi_epi32,
5550         _mm256_mask_shrdi_epi64, _mm256_maskz_shrdi_epi64, _mm256_shrdi_epi64,
5551         _mm_mask_shrdi_epi16, _mm_maskz_shrdi_epi16, _mm_shrdi_epi16,
5552         _mm_mask_shrdi_epi32, _mm_maskz_shrdi_epi32, _mm_shrdi_epi32,
5553         _mm_mask_shrdi_epi64, _mm_maskz_shrdi_epi64, _mm_shrdi_epi64): Ditto.
5554         * config/i386/i386-builtin.def (__builtin_ia32_vpshrd_v32hi,
5555         __builtin_ia32_vpshrd_v32hi_mask, __builtin_ia32_vpshrd_v16hi,
5556         __builtin_ia32_vpshrd_v16hi_mask, __builtin_ia32_vpshrd_v8hi,
5557         __builtin_ia32_vpshrd_v8hi_mask, __builtin_ia32_vpshrd_v16si,
5558         __builtin_ia32_vpshrd_v16si_mask, __builtin_ia32_vpshrd_v8si,
5559         __builtin_ia32_vpshrd_v8si_mask, __builtin_ia32_vpshrd_v4si,
5560         __builtin_ia32_vpshrd_v4si_mask, __builtin_ia32_vpshrd_v8di,
5561         __builtin_ia32_vpshrd_v8di_mask, __builtin_ia32_vpshrd_v4di,
5562         __builtin_ia32_vpshrd_v4di_mask, __builtin_ia32_vpshrd_v2di,
5563         __builtin_ia32_vpshrd_v2di_mask): New builtins.
5564         * config/i386/sse.md (vpshrd_<mode><mask_name>): New pattern.
5566 2017-11-28  Julia Koval  <julia.koval@intel.com>
5568         * config/i386/avx512vbmi2intrin.h (_mm512_shldi_epi16,
5569         _mm512_mask_shldi_epi16, _mm512_maskz_shldi_epi16, _mm512_shldi_epi32,
5570         _mm512_mask_shldi_epi32, _mm512_maskz_shldi_epi32, _mm512_shldi_epi64,
5571         _mm512_mask_shldi_epi64, _mm512_maskz_shldi_epi64): New intrinsics.
5572         * config/i386/avx512vbmi2vlintrin.h (_mm256_shldi_epi16,
5573         _mm256_mask_shldi_epi16, _mm256_maskz_shldi_epi16,
5574         _mm256_mask_shldi_epi32, _mm256_maskz_shldi_epi32, _mm256_shldi_epi32,
5575         _mm256_mask_shldi_epi64, _mm256_maskz_shldi_epi64, _mm256_shldi_epi64,
5576         _mm_mask_shldi_epi16, _mm_maskz_shldi_epi16, _mm_shldi_epi16,
5577         _mm_mask_shldi_epi32, _mm_maskz_shldi_epi32, _mm_shldi_epi32,
5578         _mm_mask_shldi_epi64, _mm_maskz_shldi_epi64, _mm_shldi_epi64): Ditto.
5579         * config/i386/i386-builtin-types.def (V32HI_FTYPE_V32HI_V32HI_INT,
5580         V32HI_FTYPE_V32HI_V32HI_INT_V32HI_INT, V16SI_FTYPE_V16SI_V16SI_INT,
5581         V16SI_FTYPE_V16SI_V16SI_INT_V16SI_INT,
5582         V8DI_FTYPE_V8DI_V8DI_INT_V8DI_INT, V8SI_FTYPE_V8SI_V8SI_INT_V8SI_INT,
5583         V16HI_FTYPE_V16HI_V16HI_INT_V16HI_INT,
5584         V4DI_FTYPE_V4DI_V4DI_INT_V4DI_INT,
5585         V8HI_FTYPE_V8HI_V8HI_INT_V8HI_INT,
5586         V4SI_FTYPE_V4SI_V4SI_INT_V4SI_INT,
5587         V2DI_FTYPE_V2DI_V2DI_INT_V2DI_INT): New types.
5588         * config/i386/i386-builtin.def (__builtin_ia32_vpshld_v32hi,
5589         __builtin_ia32_vpshld_v32hi_mask, __builtin_ia32_vpshld_v16hi,
5590         __builtin_ia32_vpshld_v16hi_mask, __builtin_ia32_vpshld_v8hi,
5591         __builtin_ia32_vpshld_v8hi_mask, __builtin_ia32_vpshld_v16si,
5592         __builtin_ia32_vpshld_v16si_mask, __builtin_ia32_vpshld_v8si,
5593         __builtin_ia32_vpshld_v8si_mask, __builtin_ia32_vpshld_v4si,
5594         __builtin_ia32_vpshld_v4si_mask, __builtin_ia32_vpshld_v8di,
5595         __builtin_ia32_vpshld_v8di_mask, __builtin_ia32_vpshld_v4di,
5596         __builtin_ia32_vpshld_v4di_mask, __builtin_ia32_vpshld_v2di,
5597         __builtin_ia32_vpshld_v2di_mask): New builtins.
5598         * config/i386/i386.c (ix86_expand_args_builtin): Handle new types.
5599         * config/i386/sse.md (vpshld_<mode><mask_name>): New pattern.
5601 2017-11-28  Richard Biener  <rguenther@suse.de>
5603         PR tree-optimization/80776
5604         * gimple-ssa-evrp-analyze.h (evrp_range_analyzer::set_ssa_range_info):
5605         Declare.
5606         * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::set_ssa_range_info):
5607         New function.
5608         (evrp_range_analyzer::record_ranges_from_incoming_edges):
5609         If the incoming edge is an effective fallthru because the other
5610         edge only reaches a __builtin_unreachable () then record ranges
5611         derived from the controlling condition in SSA info.
5612         (evrp_range_analyzer::record_ranges_from_phis): Use set_ssa_range_info.
5613         (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
5615 2017-11-28  Olivier Hainque  <hainque@adacore.com>
5617         * Makefile.in (SELFTEST_FLAGS): Use nul instead of /dev/null
5618         on mingw build hosts.
5620 2017-11-28  Jakub Jelinek  <jakub@redhat.com>
5622         PR debug/81307
5623         * dbxout.c (dbx_block_with_cold_children): Fix function comment.
5625 2017-11-28  Richard Biener  <rguenther@suse.de>
5627         PR middle-end/83141
5628         * gimple-fold.c (gimple_fold_builtin_memory_op): For aggregate
5629         copies generated from memcpy use a character array as reference
5630         type.
5632 2017-11-28  Julia Koval  <julia.koval@intel.com>
5633             Sebastian Peryt  <sebastian.peryt@intel.com>
5635         * Makefile.in (cilkplus.def, cilk-builtins.def, c-family/cilk.o,
5636         c-family/c-cilkplus.o, c-family/array-notation-common.o,
5637         cilk-common.o, cilk.h, cilk-common.c): Remove.
5638         * builtin-types.def
5639         (BT_FN_INT_PTR_PTR_PTR_FTYPE_BT_INT_BT_PTR_BT_PTR_BT_PTR): Remove.
5640         * builtins.c (is_builtin_name): Remove cilkplus condition.
5641         (BUILT_IN_CILK_DETACH, BUILT_IN_CILK_POP_FRAME): Remove.
5642         * builtins.def (DEF_CILK_BUILTIN_STUB, DEF_CILKPLUS_BUILTIN,
5643         cilk-builtins.def, cilkplus.def): Remove.
5644         * cif-code.def (CILK_SPAWN): Remove.
5645         * cilk-builtins.def: Delete.
5646         * cilk-common.c: Ditto.
5647         * cilk.h: Ditto.
5648         * cilkplus.def: Ditto.
5649         * config/darwin.h (fcilkplus): Delete.
5650         * cppbuiltin.c: Ditto.
5651         * doc/extend.texi: Remove cilkplus doc.
5652         * doc/generic.texi: Ditto.
5653         * doc/invoke.texi: Ditto.
5654         * doc/passes.texi: Ditto.
5655         * gcc.c (fcilkplus): Remove.
5656         * gengtype.c (cilk.h): Remove.
5657         * gimple-pretty-print.c (dump_gimple_omp_for): Remove cilkplus
5658         support.
5659         * gimple.h (GF_OMP_FOR_KIND_CILKFOR, GF_OMP_FOR_KIND_CILKSIMD):
5660         Remove.
5661         * gimplify.c (gimplify_return_expr, maybe_fold_stmt,
5662         gimplify_call_expr,
5663         is_gimple_stmt, gimplify_modify_expr, gimplify_scan_omp_clauses,
5664         gimplify_adjust_omp_clauses, gimplify_omp_for, gimplify_expr): Remove
5665         cilkplus conditions.
5666         * ipa-fnsummary.c (ipa_dump_fn_summary, compute_fn_summary,
5667         inline_read_section): Ditto.
5668         * ipa-inline-analysis.c (cilk.h): Remove.
5669         * ira.c (ira_setup_eliminable_regset): Remove cilkplus support.
5670         * lto-wrapper.c (merge_and_complain, append_compiler_options,
5671         append_linker_options): Remove condition for fcilkplus.
5672         * lto/lto-lang.c (cilk.h): Remove.
5673         (lto_init): Remove condition for fcilkplus.
5674         * omp-expand.c (expand_cilk_for_call): Delete.
5675         (expand_omp_taskreg, expand_omp_for_static_chunk,
5676         expand_omp_for): Remove cilkplus
5677         conditions.
5678         (expand_cilk_for): Delete.
5679         * omp-general.c (omp_extract_for_data): Remove cilkplus support.
5680         * omp-low.c (scan_sharing_clauses, create_omp_child_function,
5681         execute_lower_omp, diagnose_sb_0): Ditto.
5682         * omp-simd-clone.c (simd_clone_clauses_extract): Ditto.
5683         * tree-core.h (OMP_CLAUSE__CILK_FOR_COUNT_): Delete.
5684         * tree-nested.c: Ditto.
5685         * tree-pretty-print.c (dump_omp_clause): Remove cilkplus support.
5686         (dump_generic_node): Ditto.
5687         * tree.c (OMP_CLAUSE__CILK_FOR_COUNT_): Delete.
5688         * tree.def (cilk_simd, cilk_for, cilk_spawn_stmt, cilk_sync_stmt):
5689         Delete.
5690         * tree.h (CILK_SPAWN_FN, EXPR_CILK_SPAWN): Delete.
5692 2017-11-28  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
5694         * config/aarch64/aarch64.md (div<mode>3): Change check to TARGET_FLOAT.
5695         * config/aarch64/aarch64.c (aarch64_emit_approx_div): Add early exit
5696         for vector mode and !TARGET_SIMD.
5698 2017-11-28  Jakub Jelinek  <jakub@redhat.com>
5700         * tree.def (SWITCH_EXPR): Change from 3 operand to 2 operand tree.
5701         Adjust comment.
5702         * tree.h (SWITCH_LABELS): Remove.
5703         * gimplify.c (gimplify_switch_expr): Don't test SWITCH_LABELS,
5704         assert SWITCH_BODY is non-NULL.
5705         * tree-pretty-print.c (dump_generic_node): Remove SWITCH_LABELS
5706         handling.
5707         * tree.c (block_may_fallthru): Always return true; for SWITCH_EXPR.
5709         PR tree-optimization/80788
5710         * match.pd (X +- C1 CMP C2 -> X CMP C2 -+ C1): If res
5711         has TREE_OVERFLOW set, call drop_tree_overflow.
5713 2017-11-28  Richard Biener  <rguenther@suse.de>
5715         PR tree-optimization/83158
5716         * tree-vrp.c (intersect_ranges): Prefer ~[0, 0] in a few more cases.
5718 2017-11-28  Segher Boessenkool  <segher@kernel.crashing.org>
5720         PR 81288/target
5721         * config/rs6000/rs6000.c (rs6000_rtx_costs): Do not handle
5722         TARGET_ISEL && !TARGET_MFCRF differently.  Simplify code.
5724 2017-11-27  Segher Boessenkool  <segher@kernel.crashing.org>
5726         * config/rs6000/rs6000.md (<code><GPR:mode><GPR2:mode>2_isel): Change
5727         LT/GT/LTU/GTU to LE/GE/LEU/GEU where possible.
5729 2017-11-27  Michael Meissner  <meissner@linux.vnet.ibm.com>
5731         PR middle_end/82333
5732         * varasm.c (compare_constant): Take the mode of the constants into
5733         account when comparing floating point constants.
5735 2017-11-27  Gerald Pfeifer  <gerald@pfeifer.com>
5737         * hash-set.h (DEFINE_DEBUG_HASH_SET): Remove static qualifier
5738         from explicit instantiation of debug_helper.
5739         * vec.h (DEFINE_DEBUG_VEC): Ditto.
5741 2017-11-27  Richard Biener  <rguenther@suse.de>
5743         * gimple-fold.c (gimple_fold_builtin_memory_op): Remove dead code,
5744         refactor a bit.
5746 2017-11-27  Richard Biener  <rguenther@suse.de>
5748         * tree.c (wide_int_to_tree): Free discarded INTEGER_CST.
5749         (type_hash_canon): Also clear int_cst_hash_table entry for
5750         TYPE_MIN/MAX_VALUE.
5751         (build_nonstandard_integer_type): Hash all TYPE_MAX_VALUEs.
5753 2017-11-27  Tamar Christina  <tamar.christina@arm.com>
5755         * doc/extend.texi: Add -A suffix (ARMv8*-A, ARMv7-A).
5756         * doc/invoke.texi: Add -A suffix (ARMv8*-A, ARMv7-A).
5757         * doc/sourcebuild.texi: Add -A suffix (ARMv8*-A, ARMv7-A).
5759 2017-11-27  Markus Trippelsdorf  <markus@trippelsdorf.de>
5761         * hash-map.h (gt_cleare_cache): Avoid UB.
5763 2017-11-27  Eric Botcazou  <ebotcazou@adacore.com>
5765         * cfgloop.h (struct loop): Document usage of USHRT_MAX for unroll.
5766         * loop-unroll.c (decide_unroll_constant_iterations): Implement it.
5767         (decide_unroll_runtime_iterations): Likewise.
5768         (decide_unroll_stupid): Likewise.
5770 2017-11-27  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
5772         PR target/83109
5773         * config/i386/i386.md: Add a loop with incssp.
5775 2017-11-27  Martin Jambor  <mjambor@suse.cz>
5777         PR tree-optimization/81248
5778         * tree-sra.c (splice_param_accesses): Remove size check.
5779         (decide_one_param_reduction): Fix size check.
5780         * gimple-pretty-print.c (dump_profile): Silence warning.
5781         * params.def (PARAM_IPA_SRA_PTR_GROWTH_FACTOR): Adjust description.
5783 2017-11-27  Jakub Jelinek  <jakub@redhat.com>
5785         PR debug/81307
5786         * dbxout.c (lastlineno): New variable.
5787         (dbx_debug_hooks): Use dbxout_switch_text_section as
5788         switch_text_section debug hook.
5789         (dbxout_function_end): Switch to current_function_section
5790         rather than function_section.  If crtl->has_bb_partition,
5791         output just one N_FUN, depending on in_cold_section_p.
5792         (dbxout_source_line): Remember last lineno in lastlineno.
5793         (dbxout_switch_text_section): New function.
5794         (dbxout_function_decl): Adjust dbxout_block caller.
5795         (dbx_block_with_cold_children): New function.
5796         (dbxout_block): Return true if any LBRAC/RBRAC have been
5797         emitted.  Use dbx_block_with_cold_children at depth == 0
5798         in second partition.  Add PARENT_BLOCKNUM argument, pass
5799         it optionally adjusted to children.  Output LBRAC/RBRAC
5800         around recursive call only if the block is in the current
5801         partition, if not and anything was output, emit empty
5802         range LBRAC/RBRAC.
5803         * final.c (final_scan_insn): Compute cold_function_name
5804         before calling switch_text_section debug hook.  Call
5805         that hook even if dwarf2out_do_frame if not emitting
5806         dwarf debug info.
5808         PR target/83100
5809         * varasm.c (bss_initializer_p): Return true for DECL_COMMON
5810         TREE_READONLY decls.
5812 2017-11-27  Markus Trippelsdorf  <markus@trippelsdorf.de>
5814         PR rtl-optimization/82488
5815         * expr.c (fixup_args_size_notes): Avoid signed integer overflow.
5817 2017-11-26  Julia Koval  <julia.koval@intel.com>
5819         * config/i386/i386.c (processor_target_table): Add skylake_cost for
5820         skylake-avx512.
5821         * config/i386/x86-tune-costs.h (skylake_memcpy, skylake_memset,
5822         skylake_cost): New.
5824 2017-11-26  Julia Koval  <julia.koval@intel.com>
5826         * config/i386/driver-i386.c (host_detect_local_cpu):
5827         Detect skylake-avx512.
5829 2017-11-26  Julia Koval  <julia.koval@intel.com>
5831         * config.gcc: Add -march=cannonlake.
5832         * config/i386/driver-i386.c (host_detect_local_cpu): Detect cannonlake.
5833         * config/i386/i386-c.c (ix86_target_macros_internal): Handle cannonlake.
5834         * config/i386/i386.c (processor_costs): Add m_CANNONLAKE.
5835         (PTA_CANNONLAKE): New.
5836         (processor_target_table): Add cannonlake.
5837         (ix86_option_override_internal): Ditto.
5838         (fold_builtin_cpu): Ditto.
5839         (get_builtin_code_for_version): Handle cannonlake.
5840         (M_INTEL_COREI7_CANNONLAKE): New.
5841         * config/i386/i386.h (TARGET_CANNONLAKE, PROCESSOR_CANNONLAKE): New.
5842         * doc/invoke.texi: Add -march=cannonlake.
5844 2017-11-14  Boris Kolpackov  <boris@codesynthesis.com>
5846         * plugin.c (add_new_plugin): Use platform-specific library extensions.
5847         (try_init_one_plugin): Alternative implementation for MinGW.
5848         * Makefile.in (plugin_implib): New.
5849         (gengtype-lex.c): Fix broken AIX workaround.
5850         * configure: Regenerate.
5851         * doc/plugins.texi: Document support for MinGW.
5853 2017-11-25  Jakub Jelinek  <jakub@redhat.com>
5855         PR rtl-optimization/81553
5856         * combine.c (simplify_if_then_else): In (if_then_else COND (OP Z C1) Z)
5857         to (OP Z (mult COND (C1 * STORE_FLAG_VALUE))) optimization, if OP
5858         is a shift where C1 has different mode than the whole shift, use C1's
5859         mode for MULT rather than the shift's mode.
5861         PR target/82848
5862         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Don't fold
5863         builtins not enabled in the currently selected ISA.
5865 2017-11-24  Jackson Woodruff  <jackson.woodruff@arm.com>
5867         PR tree-optimization/71026
5868         * tree-ssa-math-opts (is_division_by_square, is_square_of): New.
5869         (insert_reciprocals): Change to insert reciprocals before a division
5870         by a square and to insert the square of a reciprocal.
5871         (execute_cse_reciprocals_1): Change to consider division by a square.
5872         (register_division_in): Add importance parameter.
5874 2017-11-24  Richard Biener  <rguenther@suse.de>
5876         PR tree-optimization/82402
5877         * tree-vect-loop-manip.c (create_lcssa_for_virtual_phi): Properly
5878         set SSA_NAME_OCCURS_IN_ABNORMAL_PHI.
5880 2017-11-24  Marc Glisse  <marc.glisse@inria.fr>
5882         * match.pd (0-ptr): New transformation.
5884 2017-11-24  Jan Hubicka  <hubicka@ucw.cz>
5886         PR bootstrap/83015
5887         * ipa-inline.c (inline_small_functions): Set current badnes correctly
5888         when skipping checking.
5890 2017-11-24  Richard Biener  <rguenther@suse.de>
5892         PR tree-optimization/83128
5893         * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Handle STRING_CSTs.
5894         (vn_reference_lookup_3): Likewise.
5896 2017-11-24  Jakub Jelinek  <jakub@redhat.com>
5898         PR sanitizer/83014
5899         * ubsan.c (ubsan_type_descriptor): Use pp_unsigned_wide_integer
5900         instead of pp_printf with HOST_WIDE_INT_PRINT_DEC.  Avoid calling
5901         tree_to_uhwi twice.
5903         * tree-object-size.c (pass_through_call): Use gimple_call_return_flags
5904         ERF_RETURN*ARG* for builtins other than BUILT_IN_ASSUME_ALIGNED,
5905         check for the latter with gimple_call_builtin_p.  Do not handle
5906         BUILT_IN_STPNCPY_CHK which is not a pass through call.
5908 2017-11-24  Christophe Lyon  <christophe.lyon@linaro.org>
5910         * config/arm/arm_neon.h: Fix pragma GCC push_options before
5911         vdot_u32.
5913 2017-11-23  Julia Koval  <julia.koval@intel.com>
5915         * config/i386/avx512vbmi2intrin.h (_mm512_mask_expand_epi8,
5916         _mm512_maskz_expand_epi8, _mm512_mask_expandloadu_epi8,
5917         _mm512_maskz_expandloadu_epi8, _mm512_mask_expand_epi16,
5918         _mm512_maskz_expand_epi16, _mm512_mask_expandloadu_epi16,
5919         _mm512_maskz_expandloadu_epi16): New intrinsics.
5920         * config/i386/avx512vbmi2vlintrin.h (_mm_mask_expand_epi8,
5921         _mm_maskz_expand_epi8, _mm_mask_expandloadu_epi8,
5922         _mm_maskz_expandloadu_epi8, _mm_mask_expand_epi16,
5923         _mm_maskz_expand_epi16, _mm_mask_expandloadu_epi16,
5924         _mm_maskz_expandloadu_epi16, _mm256_mask_expand_epi16,
5925         _mm256_maskz_expand_epi16, _mm256_mask_expandloadu_epi16,
5926         _mm256_maskz_expandloadu_epi16, _mm256_mask_expand_epi8,
5927         _mm256_maskz_expand_epi8, _mm256_mask_expandloadu_epi8,
5928         _mm256_maskz_expandloadu_epi8): New intrinsics.
5929         * config/i386/i386-builtin-types.def (V64QI_FTYPE_PCV64QI_V64QI_UDI,
5930         V32HI_FTYPE_PCV32HI_V32HI_USI, V32QI_FTYPE_PCV32QI_V32QI_USI,
5931         V16HI_FTYPE_PCV16HI_V16HI_UHI, V16QI_FTYPE_PCV16QI_V16QI_UHI,
5932         V8HI_FTYPE_PCV8HI_V8HI_UQI): New types.
5933         * config/i386/i386.c (ix86_expand_special_args_builtin): Use new types.
5934         * config/i386/sse.md (VI248_VLBW): New iterator.
5935         (expand<mode>_mask, expand<mode>_maskz): New patterns.
5937 2017-11-23  Julia Koval  <julia.koval@intel.com>
5939         * config.gcc (avx512vbmi2intrin.h, avx512vbmi2vlintrin): New headers.
5940         * config/i386/avx512vbmi2intrin.h (_mm512_mask_compress_epi8,
5941         _mm512_maskz_compress_epi8, _mm512_mask_compressstoreu_epi8,
5942         _mm512_mask_compress_epi16, _mm512_maskz_compress_epi16,
5943         _mm512_mask_compressstoreu_epi16): New.
5944         * config/i386/avx512vbmi2vlintrin.h (_mm_mask_compress_epi8,
5945         _mm_maskz_compress_epi8, _mm256_mask_compressstoreu_epi16,
5946         _mm_mask_compress_epi16, _mm_maskz_compress_epi16,
5947         _mm256_mask_compress_epi16, _mm256_maskz_compress_epi16,
5948         _mm_mask_compressstoreu_epi8, _mm_mask_compressstoreu_epi16,
5949         _mm256_mask_compress_epi8, _mm256_maskz_compress_epi8,
5950         _mm256_mask_compressstoreu_epi8): New.
5951         * config/i386/i386-builtin-types.def (VOID_FTYPE_PV64QI_V64QI_UDI,
5952         VOID_FTYPE_PV32HI_V32HI_USI, VOID_FTYPE_PV32QI_V32QI_USI,
5953         VOID_FTYPE_PV16QI_V16QI_UHI, VOID_FTYPE_PV16HI_V16HI_UHI,
5954         VOID_FTYPE_PV8HI_V8HI_UQI): New types.
5955         * config/i386/i386-builtin.def (__builtin_ia32_compressqi512_mask,
5956         __builtin_ia32_compresshi512_mask, __builtin_ia32_compressqi256_mask,
5957         __builtin_ia32_compressqi128_mask, __builtin_ia32_compresshi256_mask,
5958         __builtin_ia32_compresshi128_mask,
5959         __builtin_ia32_compressstoreuqi512_mask,
5960         __builtin_ia32_compressstoreuhi512_mask,
5961         __builtin_ia32_compressstoreuqi256_mask,
5962         __builtin_ia32_compressstoreuqi128_mask,
5963         __builtin_ia32_compressstoreuhi256_mask,
5964         __builtin_ia32_compressstoreuhi128_mask): New builtins.
5965         * config/i386/i386.c (ix86_init_mmx_sse_builtins): Create special args
5966         array for flags2.
5967         (ix86_expand_special_args_builtin): Handle new types.
5968         (s4fma_expand): Handle new builtin array.
5969         * config/i386/immintrin.h: Include new headers.
5970         * config/i386/sse.md (VI12_AVX512VLBW): New iterator.
5971         (compress<mode>_mask, compressstore<mode>_mask): New patterns.
5973 2017-11-23  Jakub Jelinek  <jakub@redhat.com>
5975         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Formatting
5976         fixes.  Declare temp and g variables at the top in order to avoid
5977         {} in most of the cases.
5979 2017-11-23  Marc Glisse  <marc.glisse@inria.fr>
5981         * match.pd (ptr-0): New transformation.
5983 2017-11-23  Charles Baylis  <charles.baylis@linaro.org>
5985         * config/arm/arm-protos.h (enum arm_addr_mode_op): New.
5986         (struct addr_mode_cost_table): New.
5987         (struct tune_params): Add field addr_mode_costs.
5988         * config/arm/arm.c (generic_addr_mode_costs): New.
5989         (arm_slowmul_tune): Initialise addr_mode_costs field.
5990         (arm_fastmul_tune): Likewise.
5991         (arm_strongarm_tune): Likewise.
5992         (arm_xscale_tune): Likewise.
5993         (arm_9e_tune): Likewise.
5994         (arm_marvell_pj4_tune): Likewise.
5995         (arm_v6t2_tune): Likewise.
5996         (arm_cortex_tune): Likewise.
5997         (arm_cortex_a8_tune): Likewise.
5998         (arm_cortex_a7_tune): Likewise.
5999         (arm_cortex_a15_tune): Likewise.
6000         (arm_cortex_a35_tune): Likewise.
6001         (arm_cortex_a53_tune): Likewise.
6002         (arm_cortex_a57_tune): Likewise.
6003         (arm_exynosm1_tune): Likewise.
6004         (arm_xgene1_tune): Likewise.
6005         (arm_cortex_a5_tune): Likewise.
6006         (arm_cortex_a9_tune): Likewise.
6007         (arm_cortex_a12_tune): Likewise.
6008         (arm_cortex_a73_tune): Likewise.
6009         (arm_v7m_tune): Likewise.
6010         (arm_cortex_m7_tune): Likewise.
6011         (arm_v6m_tune): Likewise.
6012         (arm_fa726te_tune): Likewise.
6013         (arm_mem_costs): Use table lookup to calculate cost of addressing mode.
6015 2017-11-23  Charles Baylis  <charles.baylis@linaro.org>
6017         * config/arm/arm.c (arm_mem_costs): New function.
6018         (arm_rtx_costs_internal): Use arm_mem_costs.
6020 2017-11-23  Mark Wielaard  <mark@klomp.org>
6022         * dwarf2out.c (init_sections_and_labels): Use generation to create
6023         unique ranges_section_label and ranges_base_label. Return generation.
6024         (output_rnglists): Add generation argument. Use generation to create
6025         unique ranges labels.
6026         (dwarf2out_finish): Get generation from init_sections_and_labels
6027         and pass generation to output_rnglists.
6029 2017-11-23  Mike Stump  <mikestump@comcast.net>
6030             Eric Botcazou  <ebotcazou@adacore.com>
6032         * doc/generic.texi (ANNOTATE_EXPR): Document 3rd operand.
6033         * cfgloop.h (struct loop): Add unroll field.
6034         * function.h (struct function): Add has_unroll bitfield.
6035         * gimplify.c (gimple_boolify) <ANNOTATE_EXPR>: Deal with unroll kind.
6036         (gimplify_expr) <ANNOTATE_EXPR>: Propagate 3rd operand.
6037         * loop-init.c (pass_loop2::gate): Return true if cfun->has_unroll.
6038         (pass_rtl_unroll_loops::gate): Likewise.
6039         * loop-unroll.c (decide_unrolling): Tweak note message.  Skip loops
6040         for which loop->unroll==1.
6041         (decide_unroll_constant_iterations): Use note for consistency and
6042         take loop->unroll into account.  Return early if loop->unroll is set.
6043         Fix thinko in existing test.
6044         (decide_unroll_runtime_iterations): Use note for consistency and
6045         take loop->unroll into account.
6046         (decide_unroll_stupid): Likewise.
6047         * lto-streamer-in.c (input_cfg): Read loop->unroll.
6048         * lto-streamer-out.c (output_cfg): Write loop->unroll.
6049         * tree-cfg.c (replace_loop_annotate_in_block) <annot_expr_unroll_kind>:
6050         New case.
6051         (replace_loop_annotate) <annot_expr_unroll_kind>: Likewise.
6052         (print_loop): Print loop->unroll if set.
6053         * tree-core.h (enum annot_expr_kind): Add annot_expr_unroll_kind.
6054         * tree-inline.c (copy_loops): Copy unroll and set cfun->has_unroll.
6055         * tree-pretty-print.c (dump_generic_node) <annot_expr_unroll_kind>:
6056         New case.
6057         * tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Bail out if
6058         loop->unroll is set and smaller than the trip count.  Otherwise bypass
6059         entirely the heuristics if loop->unroll is set.  Remove dead note.
6060         Fix off-by-one bug in other note.
6061         (try_peel_loop): Bail out if loop->unroll is set.  Fix formatting.
6062         (tree_unroll_loops_completely_1): Force unrolling if loop->unroll
6063         is greater than 1.
6064         (tree_unroll_loops_completely): Make static.
6065         (pass_complete_unroll::execute): Use correct type for variable.
6066         (pass_complete_unrolli::execute): Fix formatting.
6067         * tree.def (ANNOTATE_EXPR): Add 3rd operand.
6069 2017-11-23  Sergey Shalnov  <Sergey.Shalnov@intel.com>
6071         * config/i386/i386.h (TARGET_PREFER_AVX256): Also
6072         enable when TARGET_PREFER_AVX128 is set.
6074 2017-11-23  Jan Hubicka  <hubicka@ucw.cz>
6076         * ipa-profile.c (ipa_propagate_frequency_1): Use count instead of
6077         frequency.
6078         * cgraph.c (cgraph_edge::dump_edge_flags): Dump sreal frequencies.
6079         (cgraph_edge::maybe_hot_p): Use sreal frequencies.
6081 2017-11-23  Jan Hubicka  <hubicka@ucw.cz>
6083         * ipa-fnsummary.c (record_modified_bb_info): Use sreal
6084         frequencies. Fix estimation of aggregate parameters.
6086 2017-11-23  Jan Hubicka  <hubicka@ucw.cz>
6088         * cgraphclones.c (cgraph_node::create_clone): Fix updating of profile
6089         when inlining.
6091 2017-11-23  Tom de Vries  <tom@codesourcery.com>
6093         * config/sh/sh.h (ASM_OUTPUT_ADDR_VEC_ELT): Wrap in "do {} while (0)".
6095 2017-11-23  Tom de Vries  <tom@codesourcery.com>
6097         * config/ft32/ft32.h (ASM_OUTPUT_ADDR_VEC_ELT): Remove semicolon after
6098         macro.
6100 2017-11-23  Oleg Endo  <olegendo@gcc.gnu.org>
6102         PR target/83111
6103         * config/sh/sh.md (udivsi3, divsi3, sibcall_value_pcrel,
6104         sibcall_value_pcrel_fdpic): Use local variable instead of
6105         operands[3].
6106         (calli_tbr_rel): Add missing operand 2.
6107         (call_valuei_tbr_rel): Add missing operand 3.
6109 2017-11-23  Jakub Jelinek  <jakub@redhat.com>
6111         PR middle-end/82253
6112         * expr.c (expand_assignment): For CONCAT to_rtx, complex type from and
6113         bitpos/bitsize covering the whole destination, use store_expr only if
6114         the complex mode is the same.  Otherwise, use expand_normal and if
6115         it returns CONCAT, subreg each part separately instead of trying to
6116         subreg the whole result.
6118 2017-11-23  Richard Biener  <rguenther@suse.de>
6120         PR tree-optimization/23094
6121         * tree-ssa-sccvn.c (vuse_ssa_val): Handle VN_TOP when we
6122         come here from walking over backedges in the first iteration.
6123         (vn_reference_lookup_3): Skip clobbers that store the same value.
6125 2017-11-23  Richard Biener  <rguenther@suse.de>
6127         PR tree-optimization/81403
6128         * tree-ssa-pre.c (get_representative_for): Add parameter specifying
6129         a block we need a leader relative to.
6130         (phi_translate_1): For nary processing require a leader from
6131         get_representative_for given we run expression simplification
6132         using match-and-simplify.  Remove previous fix.
6134 2017-11-22  Jeff Law  <law@redhat.com>
6136         * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::try_find_new_range):
6137         Use new method allocate_value_range rather than accessing the
6138         vrp_value_range_pool data member directly.
6139         * tree-vrp.c (simplify_stmt_for_jump_threading): Tweak slightly
6140         to use extract_range_from_stmt method to avoid need for
6141         extract_range_from_assignment method.
6142         (vrp_prop::vrp_finalize): Use set_lattice_propagation_complete
6143         method rather than setting values_propgated data member directly.
6144         * vr-values.h (class vr_values): Privatize vrp_value_range_pool,
6145         and values propagated data members and extract_range_from_assignment
6146         method.  Reorder private data members to conform to standards.
6147         Add new methods set_lattice_propagation_complete and
6148         allocate_value_range.
6150 2017-11-22  Eric Botcazou  <ebotcazou@adacore.com>
6152         PR rtl-optimization/83030
6153         * doc/rtl.texi (Flags in an RTL Expression): Alphabetize, add entry
6154         for CROSSING_JUMP_P and mention usage of 'jump' for JUMP_INSNs.
6155         (Insns): Delete entry for REG_CROSSING_JUMP in register notes.
6156         * bb-reorder.c (update_crossing_jump_flags): Do not test whether the
6157         CROSSING_JUMP_P flag is already set before setting it.
6158         * cfgrtl.c (fixup_partition_crossing): Likewise.
6159         * reorg.c (relax_delay_slots): Do not consider a CROSSING_JUMP_P
6160         insn as useless.
6162 2017-11-22  Jakub Jelinek  <jakub@redhat.com>
6164         * simplify-rtx.c (simplify_binary_operation_1) <case VEC_SERIES>:
6165         Handle the case where both arguments are using gen_const_vec_series.
6167 2017-11-22  David Malcolm  <dmalcolm@redhat.com>
6169         PR c++/62170
6170         * pretty-print.c (pp_format): Move quoting implementation to
6171         pp_begin_quote and pp_end_quote.  Update pp_format_decoder call
6172         to pass address of "quote" local.
6173         (pp_begin_quote): New function.
6174         (pp_end_quote): New function.
6175         * pretty-print.h (printer_fn): Convert penultimate param from bool
6176         to bool *.
6177         (pp_begin_quote): New decl.
6178         (pp_end_quote): New decl.
6179         * tree-diagnostic.c (default_tree_printer): Convert penultimate
6180         param from bool to bool *.
6181         * tree-diagnostic.h (default_tree_printer): Likewise.
6183 2017-11-22  Jeff Law  <law@redhat.com>
6185         * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::evrp_range_analyzer)
6186         Initialize vr_values.
6187         (evrp_range_analyzer::try_find_new_range): Call methods attached to
6188         vr_values via vr_values class instance rather than delegators.
6189         (evrp_range_analyzer::record_ranges_from_phis): Likewise.
6190         (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
6191         (evrp_range_analyzer::push_value_range): Likewise.
6192         (evrp_range_analyzer::pop_value_range): Likewise.
6193         * gimple-ssa-evrp-analyze.h (class evrp_range_analyzer): Remove
6194         most delegators.  Those remaining are exposed as public interfaces.
6195         Make vr_values a pointer and private.
6196         (evrp_range_analyzer::~evrp_range_analyzer): Delete the attached
6197         vr_values.
6198         (evrp_range_analyzer::get_vr_value): New method.
6199         * gimple-ssa-evrp.c (class evrp_folder): Use DISABLE_COPY_AND_ASSIGN.
6200         (evrp_folder::evrp_folder): New ctor to initialize vr_values.
6201         (class evrp_dom_walker): Attach evrp_folder class, initialize
6202         it in the ctor.  Remove temporary delegators.
6203         (evrp_dom_walker::before_dom_children): Call methods in attached
6204         evrp_range_analyzer class via class instance pointer.  Use
6205         free value_range_constant_singleton to remove need for
6206         op_with_constant_singleton_value delegator method.  Do not
6207         create a vrp_prop class instance for every call!  Narrow
6208         scope of a couple variables.
6209         (evrp_dom_walker::cleanup): Call methods in attached
6210         evrp_range_analyzer class via class instance pointer.
6211         * vr-values.h (class vr_values): Privatize many methods and
6212         data members.
6214 2017-11-22  Richard Sandiford  <richard.sandiford@linaro.org>
6216         * tree.def (REDUC_MAX_EXPR, REDUC_MIN_EXPR, REDUC_PLUS_EXPR): Delete.
6217         * cfgexpand.c (expand_debug_expr): Remove handling for them.
6218         * expr.c (expand_expr_real_2): Likewise.
6219         * fold-const.c (const_unop): Likewise.
6220         * optabs-tree.c (optab_for_tree_code): Likewise.
6221         * tree-cfg.c (verify_gimple_assign_unary): Likewise.
6222         * tree-inline.c (estimate_operator_cost): Likewise.
6223         * tree-pretty-print.c (dump_generic_node): Likewise.
6224         (op_code_prio): Likewise.
6225         (op_symbol_code): Likewise.
6226         * internal-fn.def (DEF_INTERNAL_SIGNED_OPTAB_FN): Define.
6227         (IFN_REDUC_PLUS, IFN_REDUC_MAX, IFN_REDUC_MIN): New internal functions.
6228         * internal-fn.c (direct_internal_fn_optab): New function.
6229         (direct_internal_fn_array, direct_internal_fn_supported_p
6230         (internal_fn_expanders): Handle DEF_INTERNAL_SIGNED_OPTAB_FN.
6231         * fold-const-call.c (fold_const_reduction): New function.
6232         (fold_const_call): Handle CFN_REDUC_PLUS, CFN_REDUC_MAX and
6233         CFN_REDUC_MIN.
6234         * tree-vect-loop.c: Include internal-fn.h.
6235         (reduction_code_for_scalar_code): Rename to...
6236         (reduction_fn_for_scalar_code): ...this and return an internal
6237         function.
6238         (vect_model_reduction_cost): Take an internal_fn rather than
6239         a tree_code.
6240         (vect_create_epilog_for_reduction): Likewise.  Build calls rather
6241         than assignments.
6242         (vectorizable_reduction): Use internal functions rather than tree
6243         codes for the reduction operation.  Update calls to the functions
6244         above.
6245         * config/aarch64/aarch64-builtins.c (aarch64_gimple_fold_builtin):
6246         Use calls to internal functions rather than REDUC tree codes.
6247         * config/aarch64/aarch64-simd.md: Update comment accordingly.
6249 2017-11-22  Olivier Hainque  <hainque@adacore.com>
6251         * config/vxworks.c (vxworks_override_options): Pick default
6252         dwarf version from macro value, VXWORKS_DWARF_VERSION_DEFAULT.
6253         * config/vxworks.h: Define VXWORKS_DWARF_VERSION_DEFAULT and
6254         DWARF_GNAT_ENCODINGS_DEFAULT.
6255         * config/vxworksae.h: Likewise.
6257 2017-11-22  Marc Glisse  <marc.glisse@inria.fr>
6259         PR tree-optimization/83104
6260         * vr-values.c (simplify_stmt_using_ranges): Check integral argument,
6261         not result.
6263 2017-11-22  Marek Polacek  <polacek@redhat.com>
6264             H.J. Lu  <hongjiu.lu@intel.com>
6265             Jason Merrill  <jason@redhat.com>
6267         PR c++/60336
6268         PR middle-end/67239
6269         PR target/68355
6270         * calls.c (initialize_argument_information): Call
6271         warn_parameter_passing_abi target hook.
6272         (store_one_arg): Use 0 for empty record size.  Don't push 0 size
6273         argument onto stack.
6274         (must_pass_in_stack_var_size_or_pad): Return false for empty types.
6275         * common.opt: Update -fabi-version description.
6276         * config/i386/i386.c (init_cumulative_args): Set cum->warn_empty.
6277         (ix86_gimplify_va_arg): Call arg_int_size_in_bytes instead of
6278         int_size_in_bytes.
6279         (ix86_is_empty_record): New function.
6280         (ix86_warn_parameter_passing_abi): New function.
6281         (TARGET_EMPTY_RECORD_P): Redefine.
6282         (TARGET_WARN_PARAMETER_PASSING_ABI): Redefine.
6283         * config/i386/i386.h (CUMULATIVE_ARGS): Add warn_empty.
6284         * doc/tm.texi: Regenerated.
6285         * doc/tm.texi.in (TARGET_EMPTY_RECORD_P,
6286         TARGET_WARN_PARAMETER_PASSING_ABI): Add.
6287         * dwarf2out.c (get_ultimate_context): Move to tree.c.
6288         * explow.c (hard_function_value): Call arg_int_size_in_bytes
6289         instead of int_size_in_bytes.
6290         * expr.c (copy_blkmode_to_reg): Likewise.
6291         * function.c (aggregate_value_p): Return 0 for empty types.
6292         (assign_parm_find_entry_rtl): Call warn_parameter_passing_abi
6293         target hook.
6294         (locate_and_pad_parm): Call arg size_in_bytes instead
6295         size_in_bytes.
6296         * lto-streamer-out.c (hash_tree): Hash TYPE_EMPTY_P and DECL_PADDING_P.
6297         * stor-layout.c (finalize_type_size): Set TYPE_EMPTY_P.
6298         * target.def (empty_record_p, warn_parameter_passing_abi): New target
6299         hooks.
6300         * targhooks.c (hook_void_CUMULATIVE_ARGS_tree): New hook.
6301         (std_gimplify_va_arg_expr): Skip empty records.  Call
6302         arg_size_in_bytes instead size_in_bytes.
6303         * targhooks.h (hook_void_CUMULATIVE_ARGS_tree): Declare.
6304         * tree-core.h (tree_type_common): Add empty_flag.
6305         (tree_decl_common): Update comments.
6306         * tree-streamer-in.c (unpack_ts_decl_common_value_fields): Stream
6307         DECL_PADDING_P.
6308         (unpack_ts_type_common_value_fields): Stream TYPE_EMPTY_P.
6309         * tree-streamer-out.c (pack_ts_decl_common_value_fields): Stream
6310         DECL_PADDING_P.
6311         (pack_ts_type_common_value_fields): Stream TYPE_EMPTY_P.
6312         * tree.c (default_is_empty_type): New function.
6313         (default_is_empty_record): New function.
6314         (arg_int_size_in_bytes): New function.
6315         (arg_size_in_bytes): New function.
6316         (get_ultimate_context): New function.
6317         * tree.h: Define TYPE_EMPTY_P, DECL_PADDING_P and
6318         TRANSLATION_UNIT_WARN_EMPTY_P.
6319         (default_is_empty_record, arg_int_size_in_bytes,
6320         arg_size_in_bytes, get_ultimate_context): Declare.
6322 2017-11-22  Thomas Preud'homme  <thomas.preudhomme@arm.com>
6324         * config/arm/arm.c (cmse_clear_registers): New function.
6325         (cmse_nonsecure_call_clear_caller_saved): Replace register clearing
6326         code by call to cmse_clear_registers.
6327         (cmse_nonsecure_entry_clear_before_return): Likewise.
6329 2017-11-22  Tamar Christina  <tamar.christina@arm.com>
6331         * config/arm/arm_neon.h (vdot_u32, vdotq_u32)
6332         (vdot_s32, vdotq_s32): New.
6333         (vdot_lane_u32, vdotq_lane_u32): New.
6334         (vdot_lane_s32, vdotq_lane_s32): New.
6337 2017-11-22  Richard Sandiford  <richard.sandiford@linaro.org>
6339         PR middle-end/82547
6340         * wide-int.cc (wi::add_large, wi::sub_large): Fix overflow detection
6341         for unsigned values with fewer HWIs than the precision.
6342         (test_overflow): New function.
6343         (wide_int_cc_tests): Call it.
6345 2017-11-22  Richard Sandiford  <richard.sandiford@linaro.org>
6347         * emit-rtl.c (init_derived_machine_modes): Make sure ptr_mode
6348         has the same mode class as Pmode.
6350 2017-11-22  Thomas Preud'homme  <thomas.preudhomme@arm.com>
6352         * config/arm/arm.c (cmse_nonsecure_call_clear_caller_saved): Get rid of
6353         padding_bits_to_clear_ptr.
6354         (cmse_nonsecure_entry_clear_before_return): Likewise.
6356 2017-11-22  Thomas Preud'homme  <thomas.preudhomme@arm.com>
6358         * config/arm/arm.c (cmse_nonsecure_call_clear_caller_saved): Use
6359         auto_sbitap instead of integer bitfield to control register needing
6360         clearing.
6362 2017-11-22  Jakub Jelinek  <jakub@redhat.com>
6364         PR tree-optimization/83044
6365         * tree-vrp.c (vrp_prop::check_array_ref): If eltsize is not
6366         INTEGER_CST or is 0, clear up_bound{,_p1} and later ignore tests
6367         that need the upper bound.  Subtract offset from
6368         get_addr_base_and_unit_offset only if positive and subtract it
6369         before division by eltsize rather than after it.
6371         PR debug/83084
6372         * valtrack.c (propagate_for_debug_subst, propagate_for_debug): Reset
6373         debug insns if they would contain UNSPEC_VOLATILE or volatile asm.
6374         (dead_debug_insert_temp): Likewise, but also ignore even non-volatile
6375         asm.
6377         PR middle-end/82875
6378         * optabs.c (expand_doubleword_mult, expand_binop): Before calling
6379         expand_binop with *mul_widen_optab, make sure at least one of the
6380         operands doesn't have VOIDmode.
6382         PR debug/83034
6383         * dwarf2out.c (mem_loc_descriptor): Handle VEC_SERIES.
6385         PR rtl-optimization/82044
6386         PR tree-optimization/82042
6387         * dse.c (record_store): Check for overflow.
6388         (check_mem_read_rtx): Properly check for overflow if width == -1, call
6389         add_wild_read instead of clear_rhs_from_active_local_stores on
6390         overflow and log it into dump_file.
6392 2017-11-22  Richard Biener  <rguenther@suse.de>
6394         * gimple-iterator.c (gimple_find_edge_insert_loc): Ignore
6395         fake edges to exit when looking for a place to insert.
6396         * tree-ssa-pre.c (clear_expression_ids): Inline into callers
6397         and remove.
6398         (insert_into_preds_of_block): Commit edge insertion immediately,
6399         assert that doesn't require new BBs.
6400         (fini_pre): Release expressions.
6401         (pass_pre::execute): Shuffle things around a bit, if the fn
6402         is too large do not compute AVAIL either as this is really the
6403         quadratic bit.
6405 2017-11-22  Richard Biener  <rguenther@suse.de>
6407         PR tree-optimization/83089
6408         * tree-if-conv.c (pass_if_conversion::execute): If anything
6409         changed reset SCEV and free the number of iteration estimates.
6411 2017-11-21  Martin Sebor  <msebor@redhat.com>
6413         PR tree-optimization/82945
6414         * calls.h (warn_nonstring_bound): Remove unused function.
6416 2017-11-21  Martin Sebor  <msebor@redhat.com>
6418         PR tree-optimization/82945
6419         * builtins.c (expand_builtin_strlen): Call maybe_warn_nonstring_arg.
6420         * calls.h (maybe_warn_nonstring_arg): Declare new function.
6421         * calls.c (get_attr_nonstring_decl, maybe_warn_nonstring_arg): New
6422         functions.
6423         (initialize_argument_information): Call maybe_warn_nonstring_arg.
6424         * calls.h (get_attr_nonstring_decl): Declare new function.
6425         * doc/extend.texi (attribute nonstring): Update.
6426         * gimple-fold.c (gimple_fold_builtin_strncpy): Call
6427         get_attr_nonstring_decl and handle it.
6428         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Same.  Improve
6429         detection of nul-termination.
6430         (strlen_to_stridx): Change to a pointer.
6431         (handle_builtin_strlen, handle_builtin_stxncpy): Adjust.
6432         (pass_strlen::execute): Same.
6434 2017-11-21  Sergey Shalnov  <Sergey.Shalnov@intel.com>
6436         * config/i386/i386-opts.h (enum prefer_vector_width): Added new enum
6437         for the new option -mprefer-vector-width=[none|128|256|512].
6438         * config/i386/i386.c (ix86_target_string): remove old style options
6439         -mprefer-avx256 and make -mprefer-avx128 as alias.
6440         (ix86_option_override_internal):  Apply defaults for the
6441         -mprefer-vector-width=[128|256] option.
6442         * config/i386/i386.h (TARGET_PREFER_AVX128, TARGET_PREFER_AVX256):
6443         Implement macros to work with -mprefer-vector-width=.
6444         * config/i386/i386.opt: Implemented option
6445         -mprefer-vector-width=[none|128|256|512].
6446         * doc/invoke.texi: Documentation for
6447         -mprefer-vector-width=[none|128|256|512].
6449 2017-11-21  Pat Haugen  <pthaugen@us.ibm.com>
6451         * config/rs6000/ppc-asm.h (f50, vs50): Fix values.
6453 2017-11-21  Jonathan Wakely  <jwakely@redhat.com>
6455         * doc/invoke.texi (-Wmaybe-uninitialized): Rephrase for clarity.
6457 2017-11-21  Marc Glisse  <marc.glisse@inria.fr>
6459         * doc/generic.texi: Document POINTER_DIFF_EXPR, update
6460         POINTER_PLUS_EXPR.
6461         * cfgexpand.c (expand_debug_expr): Handle POINTER_DIFF_EXPR.
6462         * expr.c (expand_expr_real_2): Likewise.
6463         * fold-const.c (const_binop, fold_addr_of_array_ref_difference,
6464         fold_binary_loc): Likewise.
6465         * match.pd (X-X, P+(Q-P), &D-P, (P+N)-P, P-(P+N), (P+M)-(P+N),
6466         P-Q==0, -(A-B), X-Z<Y-Z, (X-Z)-(Y-Z), Z-X<Z-Y, (Z-X)-(Z-Y),
6467         (A-B)+(C-A)): New transformations for POINTER_DIFF_EXPR, based on
6468         MINUS_EXPR transformations.
6469         * optabs-tree.c (optab_for_tree_code): Handle POINTER_DIFF_EXPR.
6470         * tree-cfg.c (verify_expr, verify_gimple_assign_binary): Likewise.
6471         * tree-inline.c (estimate_operator_cost): Likewise.
6472         * tree-pretty-print.c (dump_generic_node, op_code_prio,
6473         op_symbol_code): Likewise.
6474         * tree-vect-stmts.c (vectorizable_operation): Likewise.
6475         * vr-values.c (extract_range_from_binary_expr): Likewise.
6476         * varasm.c (initializer_constant_valid_p_1): Likewise.
6477         * tree.def: New tree code POINTER_DIFF_EXPR.
6479 2017-11-21  Uros Bizjak  <ubizjak@gmail.com>
6481         * config/i386/i386.md (*bswap<mode>2_movbe): Add
6482         integer suffix to movbe mnemonic.
6483         (*bswaphi2_movbe): Ditto.
6484         (bswaphi_lowpart): Merge with *bswaphi_lowpart_1.
6486 2017-11-21  Jakub Jelinek  <jakub@redhat.com>
6488         PR c++/83045
6489         * tree-cfg.c (pass_warn_function_return::execute): Formatting fix.
6490         Also warn if seen __builtin_unreachable () call with BUILTINS_LOCATION.
6491         Use LOCATION_LOCUS when comparing against UNKNOWN_LOCATION.
6493 2017-11-21  Martin Liska  <mliska@suse.cz>
6495         * tree-inline.c (expand_call_inline): Remove not needed
6496         xstrdup_for_dump.
6498 2017-11-21  James Cowgill  <James.Cowgill@imgtec.com>
6499             Jakub Jelinek  <jakub@redhat.com>
6501         PR target/82880
6502         * config/mips/frame-header-opt.c (mips_register_frame_header_opt):
6503         Remove static keyword from f variable.
6505 2017-11-21  Jakub Jelinek  <jakub@redhat.com>
6507         PR tree-optimization/83086
6508         * gimple-ssa-store-merging.c
6509         (imm_store_chain_info::try_coalesce_bswap): Test this_n.base_addr
6510         rather than n.base_addr.
6512 2017-11-21  Martin Liska  <mliska@suse.cz>
6514         PR rtl-optimization/82044
6515         PR tree-optimization/82042
6516         * dse.c (check_mem_read_rtx): Check for overflow.
6518 2017-11-21  Aldy Hernandez  <aldyh@redhat.com>
6520         * tree-ssa-threadbackward.c (find_jump_threads_backwards): Fix
6521         typo in comment.
6523 2017-11-21  Claudiu Zissulescu  <claziss@synopsys.com>
6525         * cfgrtl.c (force_nonfallthru_and_redirect): Don't split a call
6526         and its corresponding call arg location note.
6528 2017-11-21  Claudiu Zissulescu  <claziss@synopsys.com>
6529             Andrew Burgess  <andrew.burgess@embecosm.com>
6531         * config/arc/arc-protos.h (arc_compute_frame_size): Delete
6532         declaration.
6533         (arc_return_slot_offset): Likewise.
6534         (arc_eh_return_address_location): New declaration.
6535         * config/arc/arc.c (TARGET_BUILTIN_SETJMP_FRAME_VALUE): Define.
6536         (MUST_SAVE_REGISTER): Add exception handler case.
6537         (MUST_SAVE_RETURN_ADDR): Likewise.
6538         (arc_frame_pointer_required): Likewise.
6539         (arc_frame_pointer_needed): New function.
6540         (arc_compute_frame_size): Changed.
6541         (arc_expand_prologue): Likewise.
6542         (arc_expand_epilogue): Likewise.
6543         (arc_initial_elimination_offset): Likewise.
6544         (arc_return_slot_offset): Delete.
6545         (arc_eh_return_address_location): New function.
6546         (arc_builtin_setjmp_frame_value): Likewise.
6547         * config/arc/arc.h (EH_RETURN_DATA_REGNO): Use 2 registers.
6548         (EH_RETURN_STACKADJ_RTX): Define.
6549         (EH_RETURN_HANDLER_RTX): Likewise.
6550         * config/arc/arc.md (eh_return): Delete.
6552 2017-11-21  Aldy Hernandez  <aldyh@redhat.com>
6554         * print-tree.h (debug_vec_tree): Remove prototype.
6555         * gdbinit.in (pvt): Remove macro.
6557 2017-11-21  Jakub Jelinek  <jakub@redhat.com>
6559         PR tree-optimization/83047
6560         * gimple-ssa-store-merging.c
6561         (imm_store_chain_info::output_merged_store): If the loads with the
6562         same vuse are in different basic blocks, for load_gsi pick a load
6563         location that is dominated by the other loads.
6565         PR c++/83059
6566         * config/i386/i386.c (ix86_memmodel_check): Start
6567         -Winvalid-memory-model diagnostics with lowercase letter.
6569         PR debug/82718
6570         * dwarf2out.c (dw_loc_list): If crtl->has_bb_partition, temporarily
6571         set in_cold_section_p to the partition containing loc_list->first.
6572         When seeing loc_list->last_before_switch node, update secname and
6573         perform range_across_switch second partition handling only after that.
6575         PR debug/82933
6576         * run-rtl-passes.c: Include debug.h.
6577         (run_rtl_passes): Call debug_hooks->assembly_start.
6578         * dwarf2out.c (dwarf2out_assembly_start): Return early if invoked
6579         multiple times.
6581         PR target/82981
6582         * internal-fn.c (expand_mul_overflow): Use OPTAB_WIDEN instead of
6583         OPTAB_DIRECT in calls to expand_simple_binop.
6585 2017-11-20  David Malcolm  <dmalcolm@redhat.com>
6587         PR c/81404
6588         * Makefile.in (C_COMMON_OBJS): Add c-family/known-headers.o.
6590 2017-11-20  David Malcolm  <dmalcolm@redhat.com>
6592         PR c++/72786
6593         * spellcheck.h (best_match::blithely_get_best_candidate): New
6594         accessor.
6596 2017-11-20  Jakub Jelinek  <jakub@redhat.com>
6598         * config/i386/i386.c (parse_mtune_ctrl_str): Start diagnostics
6599         with lower case letter.
6601 2017-11-20  Uros Bizjak  <ubizjak@gmail.com>
6603         * config/i386/i386.md (bswaphi2): New expander.
6604         (*bswaphi2_movbe): New insn pattern.
6605         (bswaphi -> rorhi pepehole2): New peephole pattern.
6607 2017-11-20  Jeff Law  <law@redhat.com>
6609         * Makefile.in (OBJS): Add gimple-ssa-evrp-analyze.o.
6610         * gimple-ssa-evrp-analyze.c: New file pulled from gimple-ssa-evrp.c.
6611         * gimple-ssa-evrp-analyze.h: New file pulled from gimple-ssa-evrp.c.
6612         * gimple-ssa-evrp.c: Remove bits moved into new files.  Include
6613         gimple-ssa-evrp-analyze.h.
6615         * gimple-ssa-evrp.c (evrp_dom_walker::before_dom_children): Do not
6616         set BB_VISITED here.
6617         (evrp_range_analyzer::enter): Set BB_VISITED here instead.
6619 2017-11-20  Bin Cheng  <bin.cheng@arm.com>
6621         * tree-predcom.c: Add general comment on Store-Store chains.
6622         (split_data_refs_to_components): Postpone clearing eliminate_store_p
6623         flag in component.
6624         (get_chain_last_ref_at): Rename into...
6625         (get_chain_last_write_at): ...this.
6626         (get_chain_last_write_before_load): New function.
6627         (add_ref_to_chain): Promote type of chain from CT_STORE_LOAD to
6628         CT_STORE_STORE when write reference is added.
6629         (determine_roots_comp): Support load ref in CT_STORE_STORE chains.
6630         (is_inv_store_elimination_chain): Update get_chain_last_write_at call.
6631         (initialize_root_vars_store_elim_1): Ditto.
6632         (initialize_root_vars_store_elim_2): Ditto.  Replace rhs once default
6633         definition is created.
6634         (execute_pred_commoning_chain): Support load ref in CT_STORE_STORE
6635         chain by replacing it with dominant stored value.
6637 2017-11-20  Bin Cheng  <bin.cheng@arm.com>
6639         * tree-predcom.c (add_ref_to_chain): Remove check on distance.
6641 2017-11-20  Marc Glisse  <marc.glisse@inria.fr>
6643         * vr-values.c (extract_range_from_binary_expr): Use a full range
6644         for VR_VARYING.
6646 2017-11-20  Thomas Preud'homme  <thomas.preudhomme@arm.com>
6648         * config/arm/arm.md (R4_REGNUM): Define constant.
6649         (nonsecure_call_internal): Remove r4 clobber.
6650         (nonsecure_call_value_internal): Likewise.
6651         * config/arm/thumb1.md (nonsecure_call_reg_thumb1_v5): Remove second
6652         clobber and resequence match_operands.
6653         (nonsecure_call_value_reg_thumb1_v5): Likewise.
6654         * config/arm/thumb2.md (nonsecure_call_reg_thumb2): Likewise.
6655         (nonsecure_call_value_reg_thumb2): Likewise.
6657 2017-11-20  Jakub Jelinek  <jakub@redhat.com>
6659         PR tree-optimization/78821
6660         * gimple-ssa-store-merging.c (find_bswap_or_nop_load): Give up
6661         if base is TARGET_MEM_REF.  If base is not MEM_REF, set base_addr
6662         to the address of the base rather than the base itself.
6663         (find_bswap_or_nop_1): Just use pointer comparison for vuse check.
6664         (find_bswap_or_nop_finalize): New function.
6665         (find_bswap_or_nop): Use it.
6666         (bswap_replace): Return a tree rather than bool, change first
6667         argument from gimple * to gimple_stmt_iterator, allow inserting
6668         into an empty sequence, allow ins_stmt to be NULL - then emit
6669         all stmts into gsi.  Fix up MEM_REF address gimplification.
6670         (pass_optimize_bswap::execute): Adjust bswap_replace caller.
6671         (struct store_immediate_info): Add N and INS_STMT non-static
6672         data members.
6673         (store_immediate_info::store_immediate_info): Initialize them
6674         from newly added ctor args.
6675         (merged_store_group::apply_stores): Formatting fixes.  Sort by
6676         bitpos at the end.
6677         (stmts_may_clobber_ref_p): For stores call also
6678         refs_anti_dependent_p.
6679         (gather_bswap_load_refs): New function.
6680         (imm_store_chain_info::try_coalesce_bswap): New method.
6681         (imm_store_chain_info::coalesce_immediate_stores): Use it.
6682         (split_group): Handle LROTATE_EXPR and NOP_EXPR rhs_code specially.
6683         (imm_store_chain_info::output_merged_store): Fail if number of
6684         new estimated stmts is bigger or equal than old.  Handle LROTATE_EXPR
6685         and NOP_EXPR rhs_code.
6686         (pass_store_merging::process_store): Compute n and ins_stmt, if
6687         ins_stmt is non-NULL and the store rhs is otherwise invalid, use
6688         LROTATE_EXPR rhs_code.  Pass n and ins_stmt to store_immediate_info
6689         ctor.
6690         (pass_store_merging::execute): Calculate dominators.
6692         * tree-ssa-math-opts.c (nop_stats, bswap_stats, struct symbolic_number,
6693         BITS_PER_MARKER, MARKER_MASK, MARKER_BYTE_UNKNOWN, HEAD_MARKER, CMPNOP,
6694         CMPXCHG, do_shift_rotate, verify_symbolic_number_p,
6695         init_symbolic_number, find_bswap_or_nop_load, perform_symbolic_merge,
6696         find_bswap_or_nop_1, find_bswap_or_nop, pass_data_optimize_bswap,
6697         class pass_optimize_bswap, bswap_replace,
6698         pass_optimize_bswap::execute): Moved to ...
6699         * gimple-ssa-store-merging.c: ... this file.
6700         Include optabs-tree.h.
6701         (nop_stats, bswap_stats, do_shift_rotate, verify_symbolic_number_p,
6702         init_symbolic_number, find_bswap_or_nop_load, perform_symbolic_merge,
6703         find_bswap_or_nop_1, find_bswap_or_nop, bswap_replace): Put into
6704         anonymous namespace, remove static keywords.
6705         (pass_optimize_bswap::gate): Test BITS_PER_UNIT == 8 here...
6706         (pass_optimize_bswap::execute): ... rather than here.  Formatting fix.
6708 2017-11-20  Jan Hubicka  <hubicka@ucw.cz>
6710         PR bootstrap/83062
6711         * ipa-inline.c (can_inline_edge_p): Fix typo in previous patch.
6713 2017-11-20  Aldy Hernandez  <aldyh@redhat.com>
6715         * vec.h (debug_helper): New function.
6716         (DEFINE_DEBUG_VEC): New macro.
6717         * hash-set.h (debug_helper): New function.
6718         (DEFINE_DEBUG_HASH_SET): New macro.
6719         * cfg.c (debug_slim (edge)): New function.
6720         Call DEFINE_DEBUG_VEC for edges.
6721         Call DEFINE_DEBUG_HASH_SET for edges.
6722         * cfghooks.c (debug_slim (basic_block)): New function.
6723         Call DEFINE_DEBUG_VEC for basic blocks.
6724         Call DEFINE_DEBUG_HASH_SET for basic blocks.
6725         * print-tree.c (debug_slim): New function to handle trees.
6726         Call DEFINE_DEBUG_VEC for trees.
6727         Call DEFINE_DEBUG_HASH_SET for trees.
6728         (debug (vec<tree, va_gc>) &): Remove.
6729         (debug (<vec<tree, va_gc>) *): Remove.
6730         * print-rtl.c (debug_slim): New function to handle const_rtx.
6731         Call DEFINE_DEBUG_VEC for rtx_def.
6732         Call DEFINE_DEBUG_VEC for rtx_insn.
6733         Call DEFINE_DEBUG_HASH_SET for rtx_def.
6734         Call DEFINE_DEBUG_HASH_SET for rtx_insn.
6735         * sel-sched-dump.c (debug (vec<rtx_insn *> &): Remove.
6736         (debug (vec<rtx_insn *> *ptr): Remove.
6737         (debug_insn_vector): Remove.
6738         * stor-layout.c (debug_rli): Call debug() instead of debug_vec_tree.
6740 2017-11-20  Tom de Vries  <tom@codesourcery.com>
6742         PR rtl-optimization/82020
6743         * simplify-rtx.c (simplify_ternary_operation): Fix comparison mode of
6744         IF_THEN_ELSE condition.
6746 2017-11-19  Jeff Law  <law@redhat.com>
6748         * tree-ssa-dom.c (record_equivalences_from_phis): Fix handling
6749         of degenerates resulting from ignoring an edge.
6751 2017-11-19  Jan Hubicka  <hubicka@ucw.cz>
6753         PR ipa/81360
6754         * ipa-inline.c (can_inline_edge_p): Also check that caller is optimized
6756 2017-11-19  Jan Hubicka  <hubicka@ucw.cz>
6758         PR ipa/83001
6759         * profile-count.c (profile_count::to_sreal_scale): Fix return value
6760         for uninitialied counts.
6762 2017-11-19  Jan Hubicka  <hubicka@ucw.cz>
6764         PR ipa/60243
6765         * tree-inline.c (estimate_num_insns): Set to 1 at least.
6767 2017-11-19  Jan Hubicka  <hubicka@ucw.cz>
6769         PR target/82713
6770         * i386.c (ix86_builtin_vectorization_cost): Be ready for insane types.
6772 2017-11-19  Tom de Vries  <tom@codesourcery.com>
6774         * config/arc/arc.h (FUNCTION_PROFILER): Remove semicolon after
6775         "do while (0)".
6777 2017-11-19  Tom de Vries  <tom@codesourcery.com>
6779         * config/phoenix.h (TARGET_OS_CPP_BUILTINS): Remove semicolon after
6780         "do {} while (0)".
6782 2017-11-19  Tom de Vries  <tom@codesourcery.com>
6784         * config/visium/visium.h (ASM_OUTPUT_CASE_END): Remove semicolon after
6785         macro body.
6787 2017-11-19  Tom de Vries  <tom@codesourcery.com>
6789         * config/ft32/ft32.h (REGISTER_TARGET_PRAGMAS): Remove semicolon after
6790         "do {} while (0)".
6791         * config/spu/spu.h (REGISTER_TARGET_PRAGMAS): Same.
6793 2017-11-19  Tom de Vries  <tom@codesourcery.com>
6795         * config/mcore/mcore-elf.h (MCORE_EXPORT_NAME): Remove semicolon after
6796         "do {} while (0)".
6797         * config/mcore/mcore.h (ASM_OUTPUT_ALIGNED_COMMON): After missing
6798         semicolon after MCORE_EXPORT_NAME call.
6800 2017-11-19  Tom de Vries  <tom@codesourcery.com>
6802         PR target/82961
6803         * vmsdbgout.c (vmsdbgout_early_finish): New function.
6804         (vmsdbg_debug_hooks): Set early_finish field to vmsdbgout_early_finish.
6806 2017-11-18  Jan Hubicka  <hubicka@ucw.cz>
6808         * cgraphclones.c (cgraph_edge::clone): Rename gcov_count to prof_count.
6809         (cgraph_edge::clone): Cleanup updating of profile.
6810         * ipa-cp.c (update_profiling_info): Likewise.
6811         * ipa-inline-transform.c (inline_transform): Likewise.
6812         * ipa-inline.c (inline_small_functions): Add missing space to dump.
6813         * ipa-split.c (execute_split_functions): Do not split when function
6814         is cold.
6815         * predict.c (estimate_bb_frequencies): Cleanup updating of profile.
6816         * profile-count.c (profile_count::dump): Add global0.
6817         (profile_count::to_cgraph_frequency): Do not ICE when entry is
6818         undefined.
6819         (profile_count::to_sreal_scale): Likewise.
6820         (profile_count::adjust_for_ipa_scaling): Fix typo in comment.
6821         (profile_count::combine_with_ipa_count): New function.
6822         * profile-count.h (profile_guessed_global0adjusted): New.
6823         (profile_count::adjusted_zero): New.
6824         (profile_count::global0adjusted): New.
6825         (profile_count::combine_with_ipa_count): New.
6826         * tree-inline.c (copy_edges_for_bb): Add NUM/DEN arugment;
6827         correct profile of return block of split functions.
6828         (copy_cfg_body): Remove unused profile_count.
6829         (copy_body): Likewise.
6830         (expand_call_inline): Update.
6831         (tree_function_versioning): Update.
6833 2017-11-18  Aldy Hernandez  <aldyh@redhat.com>
6835         * hash-set.h (hash_set::empty): New.
6836         * tree-ssa-threadbackward.h: Delete.
6837         * tree-ssa-threadbackward.c (class thread_jumps): New.
6838         Move max_threaded_paths into class.
6839         (fsm_find_thread_path): Remove arguments that are now in class.
6840         (profitable_jump_thread_path): Rename to...
6841         (thread_jumps::profitable_jump_thread_path): ...this.
6842         (convert_and_register_jump_thread_path): Rename to...
6843         (thread_jumps::convert_and_register_current_path): ...this.
6844         (check_subpath_and_update_thread_path): Rename to...
6845         (thread_jumps::check_subpath_and_update_thread_path): ...this.
6846         (register_jump_thread_path_if_profitable): Rename to...
6847         (thread_jumps::register_jump_thread_path_if_profitable): ...this.
6848         (handle_phi): Rename to...
6849         (thread_jumps::handle_phi): ...this.
6850         (handle_assignment): Rename to...
6851         (thread_jumps::handle_assignment): ...this.
6852         (fsm_find_control_statement_thread_paths): Rename to...
6853         (thread_jumps::fsm_find_control_statement_thread_paths): ...this.
6854         (find_jump_threads_backwards): Rename to...
6855         (thread_jumps::find_jump_threads_backwards): ...this.
6856         Initialize path local data.
6857         (pass_thread_jumps::execute): Call find_jump_threads_backwards
6858         from within thread_jumps class.
6859         (pass_early_thread_jumps::execute): Same.
6861 2017-11-17  Jan Hubicka  <hubicka@ucw.cz>
6863         * cgraph.c (cgraph_node::dump): Do IPA sanity checking on IPA counts.
6865 2017-11-17  Steve Ellcey  <sellcey@cavium.com>
6867         * config/aarch64/aarch64-simd.md (fnma<mode>4): Move neg operator
6868         to canonical location.
6870 2017-11-17  Steve Ellcey  <sellcey@cavium.com>
6872         PR target/81356
6873         * config/aarch64/aarch64.c (aarch64_use_by_pieces_infrastructure_p):
6874         Remove.
6875         (TARGET_USE_BY_PIECES_INFRASTRUCTURE_P): Remove define.
6877 2017-11-17  H.J. Lu  <hongjiu.lu@intel.com>
6879         * config/gnu-user.h (GNU_USER_TARGET_STARTFILE_SPEC): Use
6880         rcrt1.o%s/grcrt1.o%s for -static-pie.
6882 2017-11-17  Jan Hubicka  <hubicka@ucw.cz>
6884         * i386.c (ix86_multiplication_cost, ix86_division_cost,
6885         ix86_shift_rotate_cost): Break out from ...
6886         (ix86_rtx_costs): ... here.
6887         (ix86_add_stmt_cost): Use rtx cost machinery to compute cost of
6888         vector operations.
6890 2017-11-17  Jan Hubicka  <hubicka@ucw.cz>
6892         * predict.c (determine_unlikely_bbs): Set cgraph node count to 0
6893         when entry block was promoted unlikely.
6894         (estimate_bb_frequencies): Increase frequency scale.
6895         * profile-count.h (profile_count): Export precision info.
6897 2017-11-17  Jan Hubicka  <hubicka@ucw.cz>
6899         * tree-tailcall.c (eliminate_tail_call): Be more careful about not
6900         disturbin profile of entry block.
6902 2017-11-17  Jan Hubicka  <hubicka@ucw.cz>
6904         * ipa-fnsummary.c (estimate_node_size_and_time): Be more tolerant for
6905         roundoff errors.
6907 2017-11-17  Jan Hubicka  <hubicka@ucw.cz>
6909         * ipa-cp.c (update_profiling_info): Handle conversion to local profile.
6910         * tree-cfg.c (execute_fixup_cfg): Do fixup same way as inliner does.
6912 2017-11-17  Jeff Law  <law@redhat.com>
6914         * gimple-ssa-evrp.c (class evrp_range_analyzer): New class extracted
6915         from evrp_dom_walker class.  Various methods moved into new class.
6916         (evrp_range_analyzer::evrp_range_analyzer): Constructor for new class.
6917         (evrp_range_analyzer::enter): New method.
6918         (evrp_range_analyzer::leave): New method.
6919         (evrp_dom_walker): Remove delegators no longer needed by this class.
6920         Replace vr_values data member with evrp_range_analyzer
6922         * gimple-ssa-evrp.c (evrp_dom_walker::record_ranges_from_phis): New
6923         method extracted from evrp_dom_walker::before_dom_children.
6924         (evrp_dom_walker::record_ranges_from_stmt): Likewise.
6925         (evrp_dom_walker::record_ranges_from_incoming_edge): Likewise.
6927         * gimple-ssa-evrp.c (evrp_dom_walker): Add cleanup method.
6928         Add private copy constructor and move assignment operators.
6929         Privatize methods and class data where trivially possible.
6930         (evrp_dom_walker::cleanup): New function, extracted from
6931         execute_early_vrp.  Simplify access to class data.
6933         * vr-values.h (get_output_for_vrp): Prototype.
6934         * vr-values.c (get_output_for_vrp): New function extracted from
6935         vrp_visit_assignment_or_call and extract_range_from_stmt.
6936         (vrp_visit_assignment_or_call): Use get_output_for_vrp.  Simplify.
6938 2017-11-17  Luis Machado  <luis.machado@linaro.org>
6940         * config/aarch64/aarch64.c
6941         (qdf24xx_prefetch_tune) <default_opt_level>: Set to -1.
6942         (qdf24xx_tunings) <autoprefetcher_model>: Set to
6943         tune_params::AUTOPREFETCHER_WEAK.
6945 2017-11-17  Tamar Christina  <tamar.christina@arm.com>
6947         PR target/82641
6948         * config/arm/arm.c (arm_valid_target_attribute_rec):
6949         Parse "arch=" and "+<ext>".
6950         (arm_valid_target_attribute_tree): Re-init global options.
6951         (arm_option_override): Make non-static.
6952         (arm_options_perform_arch_sanity_checks): Make errors fatal.
6953         * gcc/config/arm/arm-c.c (__ARM_FEATURE_CMSE): Support undef.
6954         (__ARM_FEATURE_CRC32): Support undef.
6955         * config/arm/arm_acle.h (__ARM_FEATURE_CRC32): Replace with pragma.
6956         * doc/extend.texi (ARM Function Attributes): Add pragma and target.
6958 2017-11-17  David Malcolm  <dmalcolm@redhat.com>
6960         * gdbinit.in (break-on-diagnostic): New command.
6962 2017-11-17  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
6964         * config/i386/i386.c (ix86_expand_epilogue): Change simple
6965         return to indirect jump for EH return if control-flow
6966         protection is enabled. Change explicit 'false' argument in
6967         pro_epilogue_adjust_stack with a value of
6968         flag_cf_protection.
6969         * config/i386/i386.md (simple_return_indirect_internal):
6970         Remove SImode restriction to support 64-bit.
6972 2017-11-17  Segher Boessenkool  <segher@kernel.crashing.org>
6974         * combine.c (added_notes_insn): New.
6975         (try_combine): Handle added_notes_insn like added_links_insn.
6976         Rewrite return value code.
6977         (distribute_notes): Set added_notes_insn to the earliest insn we added
6978         a note to.
6980 2017-11-17  Segher Boessenkool  <segher@kernel.crashing.org>
6982         PR rtl-optimization/82621
6983         * combine.c (try_combine): Do not split PARALLELs of two SETs if the
6984         dest of one of those SETs is unused.
6986 2017-11-17  Richard Biener  <rguenther@suse.de>
6988         PR fortran/83017
6989         * tree-core.h (enum annot_expr_kind): Add annot_expr_parallel_kind.
6990         * tree-pretty-print.c (dump_generic_node): Handle
6991         annot_expr_parallel_kind.
6992         * tree-cfg.c (replace_loop_annotate_in_block): Likewise.
6993         * gimplify.c (gimple_boolify): Likewise.
6995 2017-11-17  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
6997         * config.gcc (extra_headers): Add cet.h for x86 targets.
6998         * config/i386/cet.h: New file.
6999         * doc/install.texi: Add --enable-cet/--disable-cet.
7001 2017-11-17  Richard Biener  <rguenther@suse.de>
7003         PR tree-optimization/83017
7004         * tree-parloops.c (MIN_PER_THREAD): Use --param parloops-min-per-thread.
7005         (gen_parallel_loop): Properly count iterations.
7006         (parallelize_loops): Handle loop->can_be_parallel independent
7007         of flag_loop_parallelize_all.  Make static profitability test match
7008         the runtime one.
7009         * params.def (PARAM_PARLOOPS_MIN_PER_THREAD): New.
7010         * invoke.texi (parloops-min-per-thread): Document.
7012 2017-11-17  Vineet Gupta  <vgupta@synopsys.com>
7014         * config/arc/linux.h: GLIBC_DYNAMIC_LINKER update per glibc
7015         upstreaming review comments.
7017 2017-11-17  Tamar Christina  <tamar.christina@arm.com>
7019         * expr.c (copy_blkmode_to_reg): Fix bitsize for targets
7020         with fast unaligned access.
7021         * doc/sourcebuild.texi (word_mode_no_slow_unalign): New.
7023 2017-11-17  Thomas Preud'homme  <thomas.preudhomme@arm.com>
7025         * config/arm/arm.c (cmse_nonsecure_entry_clear_before_return): Allocate
7026         to_clear_arg_regs_bitmap to the same size as to_clear_bitmap.
7028 2017-11-17  Richard Biener  <rguenther@suse.de>
7030         * tree-ssa-pre.c (phi_translate_1): Remove redundant constant
7031         folding of references.
7033 2017-11-17  Qing Zhao  <qing.zhao@oracle.com>
7035         PR middle-end/78809
7036         * gimple-fold.c (gimple_fold_builtin_string_compare): Add handling
7037         of replacing call to strncmp with corresponding call to strcmp when
7038         meeting conditions.
7040 2017-11-17  Sergey Shalnov  <Sergey.Shalnov@intel.com>
7042         * config/i386/x86-tune.def (X86_TUNE_AVX256_OPTIMAL): Add tuning
7043         option prefer-avx256 for skylake-avx512 configuration.
7044         * config/i386/i386.c (ix86_option_override_internal): Ditto.
7045         (get_builtin_code_for_version): Ditto.
7047 2017-11-17  Chung-Ju Wu  <jasonwucj@gmail.com>
7048             Monk Chiang  <sh.chiang04@gmail.com>
7050         * config/nds32/nds32.h (FIRST_PSEUDO_REGISTER): Modify.
7051         (FIXED_REGISTERS): Reserve more register numbers.
7052         (CALL_USED_REGISTERS): Likewise.
7053         (REG_ALLOC_ORDER): Likewise.
7054         (REG_CLASS_CONTENTS): Likewise.
7055         (REGISTER_NAMES): Likewise.
7057 2017-11-17  Chung-Ju Wu  <jasonwucj@gmail.com>
7058             Kito Cheng  <kito.cheng@gmail.com>
7060         * config/nds32/nds32-modes.def: Add vector mode V4QI V2HI V8QI V4HI
7061         V2SI.
7062         * config/nds32/iterators.md: Add vector mode iterators and attributes.
7064 2017-11-16  Steven Munroe  <munroesj@gcc.gnu.org>
7066         * config/rs6000/mmintrin.h (_mm_add_pi32[_ARCH_PWR]): Correct
7067         parameter list for vec_splats.
7069 2017-11-16  Joseph Myers  <joseph@codesourcery.com>
7071         * doc/invoke.texi (-std=c17): Refer to 2018 expected publication
7072         date of C17.
7073         (-std=c18, -std=iso9899:2018, -std=gnu18): Document option aliases.
7075 2017-11-16  Thomas Koenig  <tkoenig@gcc.gnu.org>
7077         PR bootstrap/82856
7078         * doc/install.texi: Document incompatibility of Perl >=5.6.26
7079         with the required version of automake 1.11.6.
7081 2017-11-16  Pat Haugen  <pthaugen@us.ibm.com>
7083         * rs6000/power9.md (power9fpdiv): New automaton and cpu_unit defined
7084         for it.
7085         (DU_C2_3_power9): Correct reservation combinations.
7086         (FP_DIV_power9, VEC_DIV_power9): New.
7087         (power9-alu): Split out rotate/shift...
7088         (power9-rot): ...to here, correct dispatch resource.
7089         (power9-cracked-alu, power9-mul, power9-mul-compare): Correct dispatch
7090         resource.
7091         (power9-fp): Correct latency.
7092         (power9-sdiv): Add div/sqrt resource.
7093         (power9-ddiv): Correct latency, add div/sqrt resource.
7094         (power9-sqrt, power9-dsqrt): Add div/sqrt resource.
7095         (power9-vecfdiv, power9-vecdiv): Correct latency, add div/sqrt
7096         resource.
7097         (power9-qpdiv, power9-qpmul): Adjust resource usage.
7099 2017-11-15  Michael Meissner  <meissner@linux.vnet.ibm.com>
7101         * config/rs6000/rs6000.c (rs6000_expand_builtin): Do not do the
7102         switch statement mapping KF built-ins to TF built-ins if we don't
7103         have the proper ISA 3.0 assembler support.
7105 2017-11-16  Jan Hubicka  <hubicka@ucw.cz>
7107         * tree-emutls.c (lower_emutls_data): Remove unused bb_freq.
7108         (lower_emutls_function_body): Do not compute it.
7110 2017-11-16  Jan Hubicka  <hubicka@ucw.cz>
7112         * ipa-split.c (split_bb_info): Turn time to sreal.
7113         (split_point): Likewise.
7114         (dump_split_point): Likewise.
7115         (fine_split_points): Likewise.
7116         (execute_split_functions): Only zero split_bbs; turn time to sreals.
7118 2017-11-16  Jan Hubicka  <hubicka@ucw.cz>
7120         * ipa-fnsummary.c (analyze_function_body): Accumulate time consistently
7121         in sreal.
7123 2017-11-16  Jan Hubicka  <hubicka@ucw.cz>
7125         * predict.c (combine_predictions_for_bb): Preserve zero predicted
7126         edges.
7127         (expensive_function_p): Remove useless assert.
7128         (determine_unlikely_bbs): Propagate also forward; determine cold blocks
7130 2017-11-16  Martin Sebor  <msebor@redhat.com>
7132         PR tree-optimization/82588
7133         PR tree-optimization/82583
7134         * tree-vrp.c (check_array_ref): Handle flexible array members,
7135         string literals, and inner indices.
7136         (search_for_addr_array): Add detail to diagnostics.
7138 2017-11-16  Nathan Sidwell  <nathan@acm.org>
7140         PR c++/82836
7141         PR c++/82737
7142         * tree.h (COPY_DECL_RTL): Rename parms for clarity.
7143         (SET_DECL_ASSEMBLER_NAME): Forward to
7144         overwrite_decl_assembler_name.
7145         (COPY_DECL_ASSEMBLER_NAME): Rename parms for clarity.
7146         (overwrite_decl_assembler_name): Declare.
7147         * tree.c (overwrite_decl_assembler_name): New.
7148         * langhooks-def.h (lhd_overwrite_decl_assembler_name): Declare.
7149         (LANG_HOOKS_OVERWRITE_DECL_ASSEMBLER_NAME): Provide default.
7150         (LANG_HOOKS_INITIALIZER): Add it.
7151         * langhooks.h (struct lang_hooks): Add overwrite_decl_assembler_name.
7152         * langhooks.c (lhd_set_decl_assembler_name): Use
7153         SET_DECL_ASSEMBLER_NAME.
7154         (lhd_overwrite_decl_assembler_name): Default implementation.
7156 2017-11-16  Wilco Dijkstra  <wdijkstr@arm.com>
7157             Jackson Woodruff  <jackson.woodruff@arm.com>
7159         PR tree-optimization/71026
7160         * match.pd: Canonicalize constant multiplies in division.
7162 2017-11-16  Wilco Dijkstra  <wdijkstr@arm.com>
7164         * opts.c (default_options_table): Add OPT_fomit_frame_pointer entry.
7165         * common/config/alpha/alpha-common.c (TARGET_OPTION_OPTIMIZATION_TABLE):
7166         Remove OPT_fomit_frame_pointer entry.
7167         * common/config/arc/arc-common.c: Likewise.
7168         * common/config/arm/arm-common.c: Likewise.
7169         * common/config/avr/avr-common.c: Likewise.
7170         * common/config/c6x/c6x-common.c: Likewise.
7171         * common/config/cr16/cr16-common.c: Likewise.
7172         * common/config/cris/cris-common.c: Likewise.
7173         * common/config/epiphany/epiphany-common.c: Likewise.
7174         * common/config/fr30/fr30-common.c: Likewise.
7175         * common/config/frv/frv-common.c: Likewise.
7176         * common/config/ia64/ia64-common.c: Likewise.
7177         * common/config/iq2000/iq2000-common.c: Likewise.
7178         * common/config/lm32/lm32-common.c: Likewise.
7179         * common/config/m32r/m32r-common.c: Likewise.
7180         * common/config/mcore/mcore-common.c: Likewise.
7181         * common/config/microblaze/microblaze-common.c: Likewise.
7182         * common/config/mips/mips-common.c: Likewise.
7183         * common/config/mmix/mmix-common.c: Likewise.
7184         * common/config/mn10300/mn10300-common.c: Likewise.
7185         * common/config/nios2/nios2-common.c: Likewise.
7186         * common/config/pa/pa-common.c: Likewise.
7187         * common/config/pdp11/pdp11-common.c: Likewise.
7188         * common/config/powerpcspe/powerpcspe-common.c: Likewise.
7189         * common/config/riscv/riscv-common.c: Likewise.
7190         * common/config/rs6000/rs6000-common.c: Likewise.
7191         * common/config/rx/rx-common.c: Likewise.
7192         * common/config/s390/s390-common.c: Likewise.
7193         * common/config/sh/sh-common.c: Likewise.
7194         * common/config/sparc/sparc-common.c: Likewise.
7195         * common/config/tilegx/tilegx-common.c: Likewise.
7196         * common/config/tilepro/tilepro-common.c: Likewise.
7197         * common/config/v850/v850-common.c: Likewise.
7198         * common/config/visium/visium-common.c: Likewise.
7199         * common/config/xstormy16/xstormy16-common.c: Likewise.
7200         * common/config/xtensa/xtensa-common.c: Likewise.
7201         * invoke.texi (-fomit-frame-pointer): Update documentation.
7203 2017-11-16  Jan Hubicka  <hubicka@ucw.cz>
7205         * tree-cfg.c (gimple_find_sub_bbs): Do not compute freq.
7207 2017-11-14  Jan Hubicka  <hubicka@ucw.cz>
7209         * cfg.c (scale_bbs_frequencies_int,
7210         cale_bbs_frequencies_gcov_type): Remove.
7211         * cfg.h (scale_bbs_frequencies_int,
7212         cale_bbs_frequencies_gcov_type): Remove.
7214 2017-11-14  Jan Hubicka  <hubicka@ucw.cz>
7216         * tree-ssa-loop-manip.c
7217         (scale_dominated_blocks_in_loop): Update to profile counts.
7218         (tree_transform_and_unroll_loop): Likewise.
7220 2017-11-14  Jan Hubicka  <hubicka@ucw.cz>
7222         * tree-vect-loop-manip.c (vect_do_peeling): Do not use
7223         scale_bbs_frequencies_int.
7225 2017-11-14  Jan Hubicka  <hubicka@ucw.cz>
7227         * final.c (compute_alignments): Use counts rather than frequencies.
7229 2017-11-14  Jan Hubicka  <hubicka@ucw.cz>
7231         * cfgloopanal.c: Include sreal.h
7232         (average_num_loop_insns): Use counts and sreal for accounting.
7234 2017-11-14  Jan Hubicka  <hubicka@ucw.cz>
7236         * cfgloopmanip.c (duplicate_loop_to_header_edge): Cleanup profile
7237         manipulation.
7239 2017-11-16  Chung-Ju Wu  <jasonwucj@gmail.com>
7240             Kito Cheng  <kito.cheng@gmail.com>
7242         * config/nds32/constraints.md: Provide more constraints.
7243         * config/nds32/nds32.h (enum reg_class, REG_CLASS_NAMES,
7244         REG_CLASS_CONTENTS): Define R5_REG, R8_REG, and FRAME_POINTER_REG to
7245         support constraints usage.
7247 2017-11-16  Chung-Ju Wu  <jasonwucj@gmail.com>
7249         * config/nds32/constants.md (UNSPEC_VOLATILE_FUNC_RETURN): Remove.
7251 2017-11-16  Chung-Ju Wu  <jasonwucj@gmail.com>
7252             Kito Cheng  <kito.cheng@gmail.com>
7254         * config/nds32/nds32.opt: Add mext-perf, mext-perf2, mext-string.
7255         * config/nds32/nds32.opt: Refine the layout.
7256         * config/nds32/nds32.c (TARGET_EXT_PERF, TARGET_EXT_PERF2,
7257         TARGET_EXT_STRING): Support new options.
7258         * config/nds32/nds32.h: Likewise.
7259         * config/nds32/nds32.md: Likewise.
7260         * config/nds32/nds32-predicates.c: Likewise.
7261         * config/nds32/constraints.md: Likewise.
7262         * common/config/nds32/nds32-common.c: Likewise.
7264 2017-11-16  Julia Koval  <julia.koval@intel.com>
7266         PR target/82983
7267         * config/i386/gfniintrin.h: Add sse check.
7268         * config/i386/i386.c (ix86_expand_builtin): Fix gfni check.
7270 2017-11-16  Julia Koval  <julia.koval@intel.com>
7272         * common/config/i386/i386-common.c (OPTION_MASK_ISA_AVX512VBMI2_SET,
7273         OPTION_MASK_ISA_AVX512VBMI2_UNSET): New.
7274         (ix86_handle_option): Handle -mavx512vbmi2.
7275         * config/i386/cpuid.h: Add bit_AVX512VBMI2.
7276         * config/i386/driver-i386.c (host_detect_local_cpu): Handle new bit.
7277         * config/i386/i386-c.c (__AVX512VBMI2__): New.
7278         * config/i386/i386.c (ix86_target_string): Handle -mavx512vbmi2.
7279         (ix86_valid_target_attribute_inner_p): Ditto.
7280         * config/i386/i386.h (TARGET_AVX512VBMI2, TARGET_AVX512VBMI2_P): New.
7281         * config/i386/i386.opt (mavx512vbmi2): New option.
7282         * doc/invoke.texi: Add new option.
7284 2017-11-16  Julia Koval  <julia.koval@intel.com>
7286         * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8, _mm256_gf2p8mul_epi8,
7287         _mm_mask_gf2p8mul_epi8, _mm_maskz_gf2p8mul_epi8,
7288         _mm256_mask_gf2p8mul_epi8, _mm256_maskz_gf2p8mul_epi8,
7289         _mm512_mask_gf2p8mul_epi8, _mm512_maskz_gf2p8mul_epi8,
7290         _mm512_gf2p8mul_epi8): New intrinsics.
7291         * config/i386/i386-builtin-types.def
7292         (V64QI_FTYPE_V64QI_V64QI): New type.
7293         * config/i386/i386-builtin.def (__builtin_ia32_vgf2p8mulb_v64qi,
7294         __builtin_ia32_vgf2p8mulb_v64qi_mask, __builtin_ia32_vgf2p8mulb_v32qi,
7295         __builtin_ia32_vgf2p8mulb_v32qi_mask, __builtin_ia32_vgf2p8mulb_v16qi,
7296         __builtin_ia32_vgf2p8mulb_v16qi_mask): New builtins.
7297         * config/i386/sse.md (vgf2p8mulb_*): New pattern.
7298         * config/i386/i386.c (ix86_expand_args_builtin): Handle new type.
7300 2017-11-15  Uros Bizjak  <ubizjak@gmail.com>
7302         * config/i386/i386.c (x86_print_call_or_nop): Emit 5 byte nop
7303         explicitly as a stream of bytes.
7305 2017-11-15  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
7307         * config/rs6000/altivec.h (vec_xst_be): New #define.
7308         * config/rs6000/altivec.md (altivec_vperm_<mode>_direct): Rename
7309         and externalize from *altivec_vperm_<mode>_internal.
7310         * config/rs6000/rs6000-builtin.def (XL_BE_V16QI): Remove macro
7311         instantiation.
7312         (XL_BE_V8HI): Likewise.
7313         (XL_BE_V4SI): Likewise.
7314         (XL_BE_V4SI): Likewise.
7315         (XL_BE_V2DI): Likewise.
7316         (XL_BE_V4SF): Likewise.
7317         (XL_BE_V2DF): Likewise.
7318         (XST_BE): Add BU_VSX_OVERLOAD_X macro instantiation.
7319         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Correct
7320         all array entries with these keys: VSX_BUILTIN_VEC_XL,
7321         VSX_BUILTIN_VEC_XL_BE, VSX_BUILTIN_VEC_XST.  Add entries for key
7322         VSX_BUILTIN_VEC_XST_BE.
7323         * config/rs6000/rs6000.c (altivec_expand_xl_be_builtin): Remove.
7324         (altivec_expand_builtin): Remove handling for VSX_BUILTIN_XL_BE_*
7325         built-ins.
7326         (altivec_init_builtins): Replace conditional calls to def_builtin
7327         for __builtin_vsx_ld_elemrev_{v8hi,v16qi} and
7328         __builtin_vsx_st_elemrev_{v8hi,v16qi} based on TARGET_P9_VECTOR
7329         with unconditional calls.  Remove calls to def_builtin for
7330         __builtin_vsx_le_be_<mode>.  Add a call to def_builtin for
7331         __builtin_vec_xst_be.
7332         * config/rs6000/vsx.md (vsx_ld_elemrev_v8hi): Convert define_insn
7333         to define_expand, and add alternate RTL generation for P8.
7334         (*vsx_ld_elemrev_v8hi_internal): New define_insn based on
7335         vsx_ld_elemrev_v8hi.
7336         (vsx_ld_elemrev_v16qi): Convert define_insn to define_expand, and
7337         add alternate RTL generation for P8.
7338         (*vsx_ld_elemrev_v16qi_internal): New define_insn based on
7339         vsx_ld_elemrev_v16qi.
7340         (vsx_st_elemrev_v8hi): Convert define_insn
7341         to define_expand, and add alternate RTL generation for P8.
7342         (*vsx_st_elemrev_v8hi_internal): New define_insn based on
7343         vsx_st_elemrev_v8hi.
7344         (vsx_st_elemrev_v16qi): Convert define_insn to define_expand, and
7345         add alternate RTL generation for P8.
7346         (*vsx_st_elemrev_v16qi_internal): New define_insn based on
7347         vsx_st_elemrev_v16qi.
7349 2017-11-15  H.J. Lu  <hongjiu.lu@intel.com>
7351         PR target/82990
7352         * config/i386/i386.c (pass_insert_vzeroupper::gate): Remove
7353         TARGET_AVX512ER check.
7354         (ix86_option_override_internal): Set MASK_VZEROUPPER if
7355         neither -mzeroupper nor -mno-zeroupper is used and
7356         TARGET_EMIT_VZEROUPPER is set.
7357         * config/i386/i386.h (TARGET_EMIT_VZEROUPPER): New.
7358         * config/i386/x86-tune.def: Add X86_TUNE_EMIT_VZEROUPPER.
7360 2017-11-15  Will Schmidt  <will_schmidt@vnet.ibm.com>
7362         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add support for
7363         folding of vector compares.
7364         (fold_build_vec_cmp): New helper function.
7365         (fold_compare_helper): New helper function.
7366         (builtin_function_type): Add compare builtins to the list of functions
7367         having unsigned arguments.  Cosmetic updates to comment indentation.
7368         * config/rs6000/vsx.md (vcmpneb, vcmpneh, vcmpnew): Update to specify
7369         the not+eq combination.
7371 2017-11-15  Bin Cheng  <bin.cheng@arm.com>
7373         PR tree-optimization/82726
7374         PR tree-optimization/70754
7375         * tree-predcom.c (order_drefs_by_pos): New function.
7376         (combine_chains): Move code setting has_max_use_after to...
7377         (try_combine_chains): ...here.  New parameter.  Sort combined chains
7378         according to position information.
7379         (tree_predictive_commoning_loop): Update call to above function.
7380         (update_pos_for_combined_chains, pcom_stmt_dominates_stmt_p): New.
7382 2017-11-15  Bin Cheng  <bin.cheng@arm.com>
7384         PR tree-optimization/82726
7385         Revert
7386         2017-01-23  Bin Cheng  <bin.cheng@arm.com>
7388         PR tree-optimization/70754
7389         * tree-predcom.c (stmt_combining_refs): New parameter INSERT_BEFORE.
7390         (reassociate_to_the_same_stmt): New parameter INSERT_BEFORE.  Insert
7391         combined stmt before it if not NULL.
7392         (combine_chains): Process refs reversely and compute dominance point
7393         for root ref.
7395         Revert
7396         2017-02-23  Bin Cheng  <bin.cheng@arm.com>
7398         PR tree-optimization/79663
7399         * tree-predcom.c (combine_chains): Process refs in reverse order
7400         only for ZERO length chains, and add explaining comment.
7402 2017-11-15  Tamar Christina  <tamar.christina@arm.com>
7404         * config/arm/arm-cpus.in (armv8_3, ARMv8_3a, armv8.3-a): New
7405         * config/arm/arm-tables.opt (armv8.3-a): Regenerated.
7406         * doc/invoke.texi (ARM Options): Add armv8.3-a.
7408 2017-11-15  Tamar Christina  <tamar.christina@arm.com>
7410         * config/arm/arm.h (TARGET_DOTPROD): Add arm_arch8_2.
7412 2017-11-15  Martin Liska  <mliska@suse.cz>
7414         * tree-cfg.c (pass_warn_function_return::execute):
7415         Compare warn_return_type for greater than zero.
7417 2017-11-15  Sebastian Peryt  <sebastian.peryt@intel.com>
7419         PR target/82941
7420         PR target/82942
7421         * config/i386/i386.c (pass_insert_vzeroupper): Modify gate condition
7422         to return true on Xeon and not on Xeon Phi.
7423         (ix86_check_avx256_register): Changed to ...
7424         (ix86_check_avx_upper_register): ... this. Add extra check for
7425         VALID_AVX512F_REG_OR_XI_MODE.
7426         (ix86_avx_u128_mode_needed): Changed
7427         ix86_check_avx256_register to ix86_check_avx_upper_register.
7428         (ix86_check_avx256_stores): Changed to ...
7429         (ix86_check_avx_upper_stores): ... this. Changed
7430         ix86_check_avx256_register to ix86_check_avx_upper_register.
7431         (ix86_avx_u128_mode_after): Changed
7432         avx_reg256_found to avx_upper_reg_found. Changed
7433         ix86_check_avx256_stores to ix86_check_avx_upper_stores.
7434         (ix86_avx_u128_mode_entry): Changed
7435         ix86_check_avx256_register to ix86_check_avx_upper_register.
7436         (ix86_avx_u128_mode_exit): Ditto.
7437         * config/i386/i386.h: (host_detect_local_cpu): New define.
7439 2017-11-15  Dominik Infuehr  <dominik.infuehr@theobroma-systems.com>
7441         * config/arm/xgene1.md (xgene1): Split into automatons
7442         xgene1_main, xgene1_decoder, xgene1_div, xgene1_simd.
7443         (xgene1_f_load): Adjust reservations and/or types.
7444         (xgene1_f_store): Likewise.
7445         (xgene1_load_pair): Likewise.
7446         (xgene1_store_pair): Likewise.
7447         (xgene1_fp_load1): Likewise.
7448         (xgene1_load1): Likewise.
7449         (xgene1_store1): Likewise.
7450         (xgene1_move): Likewise.
7451         (xgene1_alu): Likewise.
7452         (xgene1_simd): Likewise.
7453         (xgene1_bfm): Likewise.
7454         (xgene1_neon_load1): Likewise.
7455         (xgene1_neon_store1): Likewise.
7456         (xgene1_neon_logic): Likewise.
7457         (xgene1_neon_st1): Likewise.
7458         (xgene1_neon_ld1r): Likewise.
7459         (xgene1_alu_cond): Added.
7460         (xgene1_shift_reg): Likwise.
7461         (xgene1_bfx): Likewise.
7462         (xgene1_mul): Split into xgene1_mul32, xgene1_mul64.
7464 2017-11-15  Jakub Jelinek  <jakub@redhat.com>
7466         PR target/82981
7467         * internal-fn.c: Include gimple-ssa.h, tree-phinodes.h and
7468         ssa-iterators.h.
7469         (can_widen_mult_without_libcall): New function.
7470         (expand_mul_overflow): If only checking unsigned mul overflow,
7471         not result, and can do efficiently MULT_HIGHPART_EXPR, emit that.
7472         Don't use WIDEN_MULT_EXPR if it would involve a libcall, unless
7473         no other way works.  Add MULT_HIGHPART_EXPR + MULT_EXPR support.
7474         (expand_DIVMOD): Formatting fix.
7475         * expmed.h (expand_mult): Add NO_LIBCALL argument.
7476         * expmed.c (expand_mult): Likewise.  Use OPTAB_WIDEN rather
7477         than OPTAB_LIB_WIDEN if NO_LIBCALL is true, and allow it to fail.
7479         PR tree-optimization/82977
7480         * tree-ssa-strlen.c (strlen_optimize_stmt): Pass a reference to a copy
7481         constructed temporary to strlen_to_stridx.put.
7483 2017-11-15  Martin Liska  <mliska@suse.cz>
7485         * configure.ac: Remove -fkeep-inline-functions from coverage_flags.
7486         * configure: Regenerate.
7488 2017-11-15  Martin Liska  <mliska@suse.cz>
7490         PR target/82927
7491         * config/sh/sh-mem.cc: Use proper probability for
7492         REG_BR_PROB_NOTE.
7494 2017-11-14  Jeff Law  <law@redhat.com>
7496         * explow.c (anti_adjust_stack_and_probe_stack_clash): Avoid probing
7497         the red zone for stack_clash_protection_final_dynamic_probe targets
7498         when the total dynamic stack size is zero bytes.
7500         * tree-ssa-threadupdate.c (thread_through_all_blocks): Thread
7501         blocks is post order.
7503 2017-11-15  Alexandre Oliva  <aoliva@redhat.com>
7505         * dumpfile.h (TDF_COMPARE_DEBUG): New.
7506         * final.c (rest_of_clean_state): Set it for the
7507         -fcompare-debug dump.
7508         * tree-pretty-print.c (dump_generic_node): Omit OBJ_TYPE_REF
7509         class when TDF_COMPARE_DEBUG is set.
7511         * dwarf2out.c (gen_producer_string): Discard
7512         OPT_fcompare_debug.
7514 2017-11-15  Joseph Myers  <joseph@codesourcery.com>
7516         PR c/81156
7517         * doc/extend.texi (Other Builtins): Document __builtin_tgmath.
7518         * ginclude/tgmath.h (__tg_cplx, __tg_ldbl, __tg_dbl, __tg_choose)
7519         (__tg_choose_2, __tg_choose_3, __TGMATH_REAL_1_2)
7520         (__TGMATH_REAL_2_3): Remove macros.
7521         (__TGMATH_CPLX, __TGMATH_CPLX_2, __TGMATH_REAL, __TGMATH_REAL_2)
7522         (__TGMATH_REAL_3, __TGMATH_CPLX_ONLY): Define using
7523         __builtin_tgmath.
7524         (frexp, ldexp, nexttoward, scalbn, scalbln): Define using
7525         __TGMATH_REAL_2.
7526         (remquo): Define using __TGMATH_REAL_3.
7528 2017-11-14  Jeff Law  <law@redhat.com>
7530         * vr-values.c: New file with contents extracted from tree-vrp.c.
7531         * Makefile.in (OBJS): Add vr-values.o
7532         * tree-vrp.h (set_value_range_to_nonnull): Prototype.
7533         (set_value_range, set_and_canonicalize_value_range): Likewise.
7534         (vrp_bitmap_equal_p, range_is_nonnull): Likewise.
7535         (value_range_constant_singleton, symbolic_range_p): Likewise.
7536         (compare_values, compare_values_warnv, vrp_val_is_min): Likewise.
7537         (vrp_val_is_max, copy_value_range, set_value_range_to_value): Likewise.
7538         (extract_range_from_binary_expr_1, vrp_val_min, vrp_val_max): Likewise.
7539         (set_value_range_to_null, range_int_cst_p, opreand_less_p): Likewise.
7540         (find_case_label_range, find_case_label_index): Likewise.
7541         (zero_nonzero_bits_from_vr, overflow_comparison_p): Likewise.
7542         (range_int_cst_singleton_p, value_inside_range): Likewise.
7543         (get_single_symbol): Likewise.
7544         (switch_update): Move structure definition here.
7545         (to_remove_edges, to_update_switch_stmts): Provide externs.
7546         * tree-vrp.c: Move all methods for vr-values class to vr-values.c
7547         (vrp_val_max, vrp_val_min, vrp_val_is_max): Make externally visible.
7548         (vrp_val_is_min, set_value_range): Likewise.
7549         (set_and_canonicalize_value_range, copy_value_range): Likewise.
7550         (set_value_range_to_value, set_value_range_to_nonnull): Likewise.
7551         (set_value_range_to_null, vrp_bitmap_equal_p): Likewise.
7552         (range_is_nonnull, range_int_cst_p): Likewwise.
7553         (range_int_cst_singleton_p, symbolic_range_p): Likewise.
7554         (get_single_symbol, operand_less_p): Likewise
7555         (compare_values_warnv, compare_values): Likewise.
7556         (value_inside_range, value_range_constant_singleton): Likewise.
7557         (zero_nonzero_bitgs_from_vr): Likewise.
7558         (extract_range_from_binary_expr_1): Likewise.
7559         (overflow_comparison_p): Likewise.
7560         (to_remove_edges, to_update_switch_stmts): Likewise.
7561         (find_case_label-index, find_case_label_range): Likewise.
7562         (switch_update, set_value_range_to_nonnegative): Remove.
7563         (set_value_range_to_truthvalue): Likewise.
7564         (symbolic_range_based_on_p, gimple_assign_nonzero_p): Likewise.
7565         (gimple_stmt_nonzero_p, compare_ranges): Likewise.
7566         (compare_range_with_value, vrp_valueize, vrp_valueize_1): Likewise.
7567         (find_case_label_ranges, test_for_singularity): Likewise.
7568         (range_fits_type_p, simplify_conversion_using_ranges): LIkewise.
7569         (x_vr_values): Move to its remaining use site.
7571 2017-11-10  Jeff Law  <law@redhat.com>
7573         * vr-values.h (VR_INITIALIZER): Move #define here.
7574         * gimple-ssa-evrp.c: New file with contents extracted from tree-vrp.c
7575         * Makefile.in (OBJS): Add tree-evrp.o
7576         * tree-vrp.h (assert_info): Move structure definition here.
7577         (set_value_range_to_varying): Prototype.
7578         (vrp_operand_equal_p, range_includes_zero_p): Likewise.
7579         (infer_value_range, register_edge_assert_for): Likewise.
7580         (stmt_interesting_for_vrp): Likewise.
7581         * tree-vrp.c: Move all methods for evrp class into gimple-ssa-evrp.c.
7582         (set_value_range_to_varying): No longer static.
7583         (vrp_operand_equal_p, range_includes_zero_p): Likewise.
7584         (infer_value_range, register_edge_assert_for): Likewise.
7586 2017-11-14  Michael Meissner  <meissner@linux.vnet.ibm.com>
7588         * config/rs6000/rs6000.md (bswapdi2): On 32-bit ISA 3.0, don't
7589         generate the XXBRD instruction.
7591         * config/rs6000/rs6000-c.c (is_float128_p): New helper function.
7592         (rs6000_builtin_type_compatible): Treat _Float128 and long double
7593         as being compatible if -mabi=ieeelongdouble.
7594         * config/rs6000/rs6000-builtin.def (BU_FLOAT128_HW_1): New macros
7595         to setup float128 built-ins with hardware support.
7596         (BU_FLOAT128_HW_2): Likewise.
7597         (BU_FLOAT128_HW_3): Likewise.
7598         (BU_FLOAT128_HW_VSX_1): Likewise.
7599         (BU_FLOAT128_HW_VSX_2): Likewise.
7600         (scalar_extract_expq): Change float128 built-in functions to
7601         accommodate having both KFmode and TFmode functions.  Use the
7602         KFmode variant as the default.
7603         (scalar_extract_sigq): Likewise.
7604         (scalar_test_neg_qp): Likewise.
7605         (scalar_insert_exp_q): Likewise.
7606         (scalar_insert_exp_qp): Likewise.
7607         (scalar_test_data_class_qp): Likewise.
7608         (sqrtf128_round_to_odd): Delete processing the round to odd
7609         built-in functions as special built-in functions, and define them
7610         as float128 built-ins.  Use the KFmode variant as the default.
7611         (truncf128_round_to_odd): Likewise.
7612         (addf128_round_to_odd): Likewise.
7613         (subf128_round_to_odd): Likewise.
7614         (mulf128_round_to_odd): Likewise.
7615         (divf128_round_to_odd): Likewise.
7616         (fmaf128_round_to_odd): Likewise.
7617         * config/rs6000/rs6000.c (rs6000_expand_binop_builtin): Add
7618         support for KFmode and TFmode xststdcqp calls.
7619         (rs6000_expand_builtin): If long double is IEEE 128-bit floating
7620         point, switch the built-in handlers for the get/set float128
7621         exponent, get float128 mantissa, float128 test built-ins, and the
7622         float128 round to odd built-in functions.  Eliminate creating the
7623         float128 round to odd built-in functions as special built-ins.
7624         (rs6000_init_builtins): Eliminate special creation of the float128
7625         round to odd built-in functions.
7626         * config/rs6000/vsx.md (xsxexpqp_<mode>): Change float128 built-in
7627         function insns to support both TFmode and KFmode variants.
7628         (xsxsigqp_<mode>): Likewise.
7629         (xsiexpqpf_<mode>): Likewise.
7630         (xsiexpqp_<mode>): Likewise.
7631         (xststdcqp_<mode>): Likewise.
7632         (xststdcnegqp_<mode>): Likewise.
7633         (xststdcqp_<mode>): Likewise.
7635 2017-11-14  Jan Hubicka  <hubicka@ucw.cz>
7637         * tree-ssa-threadupdate.c (compute_path_counts): Remove
7638         unused path_in_freq_ptr parameter.
7639         (ssa_fix_duplicate_block_edges): Do not pass around path_in_freq
7641 2017-11-14  Jan Hubicka  <hubicka@ucw.cz>
7643         * ipa-inline.c (edge_badness): Dump sreal frequency.
7644         (compute_inlined_call_time): Match natural implementaiton ...
7645         * ipa-fnsummary.c (estimate_edge_size_and_time): ... here; remove
7646         forgotten division by CGRAPH_FREQ_BASE.
7648 2017-11-14  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
7650         * config.gcc (*-*-solaris2*): Enable default_use_cxa_atexit since
7651         Solaris 11.  Update comment.
7652         * configure.ac (gcc_cv_ld_pid): Adapt comment for Solaris 12
7653         renaming.
7654         * config/sol2.h (STARTFILE_SPEC): Likewise.
7655         * configure: Regenerate.
7657 2017-11-14  Carl Love  <cel@us.ibm.com>
7659         * config/rs6000/rs6000.c (swap_endian_selector_for_mode): Remove
7660         le_ and be_ prefixes to swap* variables.  Remove
7661         if (VECTOR_ELT_ORDER_BIG) statement. Remove E_V16QImode case
7662         statements.
7664 2017-11-14  Jason Merrill  <jason@redhat.com>
7666         Support GTY((cache)) on hash_map.
7667         * hash-traits.h (ggc_remove): Add ggc_maybe_mx member function.
7668         (ggc_cache_remove): Override it instead of ggc_mx.
7669         * hash-table.h (gt_ggc_mx): Call it instead of ggc_mx.
7670         (gt_cleare_cache): Call ggc_mx instead of gt_ggc_mx.
7671         * hash-map-traits.h (simple_hashmap_traits): Add maybe_mx member.
7672         (simple_cache_map_traits): Override maybe_mx.
7673         * hash-map.h (hash_entry): Add ggc_maybe_mx and keep_cache_entry.
7674         (hash_map): Friend gt_cleare_cache.
7675         (gt_cleare_cache): New.
7676         * tree.h (tree_cache_traits): New hash_map traits class.
7677         (tree_cache_map): New typedef.
7679 2017-11-14  Richard Biener  <rguenther@suse.de>
7681         * tree-cfgcleanup.c (cleanup_control_expr_graph): Remove first_p
7682         paramter and handling.
7683         (cleanup_control_flow_bb): Likewise.
7684         (cleanup_control_flow_pre): New helper performing a DFS walk
7685         to call cleanup_control_flow_bb in PRE order.
7686         (cleanup_tree_cfg_1): Do the first phase of cleanup_control_flow_bb
7687         via cleanup_control_flow_pre.
7689 2017-11-14  James Greenhalgh  <james.greenhalgh@arm.com>
7691         * config/aarch64/aarch64-simd.md
7692         (aarch64_simd_bsl<mode>_internal): Remove DImode.
7693         (*aarch64_simd_bsl<mode>_alt): Likewise.
7694         (aarch64_simd_bsldi_internal): New.
7695         (aarch64_simd_bsldi_alt): Likewise.
7697 2017-11-13  Jan Hubicka  <hubicka@ucw.cz>
7699         * tracer.c (better_p): Do not compare frequencies.
7700         * reg-stack.c (better_edge): Likewise.
7701         * shrink-wrap.c (try_shrink_wrapping): Do not convert to gcov counts
7702         and back.
7704 2017-11-13  Jan Hubicka  <hubicka@ucw.cz>
7706         * auto-profile.c (afdo_annotate_cfg): Use update_max_bb_count.
7707         * cgraphunit.c (cgraph_node::expand_thunk): Use update_max_bb_count.
7708         * ipa-utils.c (ipa_merge_profiles): Use update_max_bb_count.
7709         * lto-streamer-in.c (input_function): Use update_max_bb_count.
7710         * omp-expand.c (expand_omp_taskreg): Use update_max_bb_count.
7711         * predict.c (maybe_hot_frequency_p): Inline to ...
7712         (maybe_hot_count_p): ... here; rewrite to counts.
7713         (counts_to_freqs): Rename to ...
7714         (update_max_bb_count): ... this one.
7715         (expensive_function_p): Use counts.
7716         (estimate_bb_frequencies): Update.
7717         (rebuild_frequencies): Update.
7718         * predict.h (counts_to_freqs): Rename to ...
7719         (update_max_bb_count): ... this one.
7720         * profile.c (compute_branch_probabilities): Add debug info
7721         * tree-inline.c (expand_call_inline): Update debug info.
7722         (optimize_inline_calls): Use update_max_bb_count..
7723         (tree_function_versioning): Use update_max_bb_count..
7724         * value-prof.c (gimple_value_profile_transformations):
7725         Do not use update_max_bb_count.
7727 2017-11-13  Jan Hubicka  <hubicka@ucw.cz>
7729         * ipa-inline.c (compute_uninlined_call_time, compute_inlined_call_time):
7730         always use frequencies.
7732 2017-11-13  Jan Hubicka  <hubicka@ucw.cz>
7734         * bb-reorder.c: Remove frequencies from comments.
7735         (better_edge_p): Use profile counts.
7736         (find_traces): Dump profile counts.
7737         (rotate_loop): Use profile counts.
7738         (find_traces_1_round): Likewise.
7739         (connect_better_edge_p): Use counts instead of probabilities for
7740         reverse walk.
7741         (copy_bb_p): Drop early check for non-0 frequency.
7742         (sanitize_hot_paths): Update comments.
7744 2017-11-13  Jan Hubicka  <hubicka@ucw.cz>
7746         * ipa-split.c (struct split_point): Add count.
7747         (consider_split): Do not compute incoming frequency; compute incoming
7748         count and store it to split_point.
7749         (split_function): Set count of the call to split part correctly.
7751 2017-11-13  Carl Love  <cel@us.ibm.com>
7753         * config/rs6000/altivec.md (altivec_vsumsws_be): Add define_expand.
7755 2017-11-13  Tom Tromey  <tom@tromey.com>
7757         * doc/cpp.texi (Variadic Macros): Document __VA_OPT__.
7759 2017-11-13  Carl Love  <cel@us.ibm.com>
7761         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins):
7762         Add support for builtins:
7763         unsigned int vec_first_{,miss}_match_{,or_eos}index,
7764         vector {un,}signed {char,int,short},
7765         vector {un,}signed {char,int,short}) arguments.
7766         * config/rs6000/rs6000-builtin.def (VFIRSTMATCHINDEX,
7767         VFIRSTMATCHOREOSINDEX, VFIRSTMISMATCHINDEX, VFIRSTMISMATCHOREOSINDEX):
7768         Add BU_P9V_AV_2 expansions for the builtins.
7769         * config/rs6000/altivec.h (vec_first_match_index,
7770         vec_first_mismatch_index, vec_first_match_or_eos_index,
7771         vec_first_mismatch_or_eos_index): Add #defines for the builtins.
7772         * config/rs6000/rs6000-protos.h (bytes_in_mode): Add
7773         new extern declaration.
7774         * config/rs6000/rs6000.c (bytes_in_mode): Add new function.
7775         * config/rs6000/vsx.md (first_match_index_<mode>,
7776         first_match_or_eos_index_<mode>, first_mismatch_index_<mode>,
7777         first_mismatch_or_eos_index_<mode>): Add define expand.
7778         (vctzlsbb_<mode>): Add mode field to define_insn for vctzlsbb.
7779         * doc/extend.texi: Update the built-in documenation file for the new
7780         built-in functions.
7782 2017-11-13  Michael Meissner  <meissner@linux.vnet.ibm.com>
7784         * match.pd: Convert fminf<N>, fminf<N>x, fmax<N>, and fmax<N>x
7785         into the min/max operations for _Float<N> and _Float<N>X types.
7787 2017-11-13  Eric Botcazou  <ebotcazou@adacore.com>
7789         PR lto/81351
7790         * dwarf2out.c (do_eh_frame): New static variable.
7791         (dwarf2out_begin_prologue): Set it.
7792         (dwarf2out_frame_finish): Test it instead of dwarf2out_do_eh_frame.
7794 2017-11-13  Jan Hubicka  <hubicka@ucw.cz>
7796         * tree-ssa-coalesce.c (coalesce_cost): Fix formating.
7798         * tree-ssa-sink.c (select_best_block): Do not use frequencies.
7800 2017-11-13  Eric Botcazou  <ebotcazou@adacore.com>
7802         PR lto/81351
7803         * debug.h (dwarf2out_do_eh_frame): Declare.
7804         * dwarf2cfi.c (dwarf2out_do_eh_frame): New predicate.
7805         (dwarf2out_do_frame): Use it.
7806         (dwarf2out_do_cfi_asm): Likewise.
7807         * dwarf2out.c (dwarf2out_frame_finish): Likewise.
7808         (dwarf2out_assembly_start): Likewise.
7809         (dwarf2out_begin_prologue): Fix comment.
7810         * toplev.c (compile_file): Always call dwarf2out_frame_finish
7811         if the target needs either debug or unwind DWARF2 info.
7812         * lto-opts.c (lto_write_options): Do not save -fexceptions,
7813         -fnon-call-exceptions, -ffp-contract, -fmath-errno, -fsigned-zeros,
7814         -ftrapping-math, -ftrapv and -fwrapv.
7816 2017-11-13  Jan Hubicka  <hubicka@ucw.cz>
7818         * cgraph.c (cgraph_edge::sreal_frequency): New function.
7819         * cgraph.h (cgraph_edge::sreal_frequency): Declare.
7820         * ipa-fnsummary.c (dump_ipa_call_summary): Use sreal_frequency.
7821         (estimate_edge_size_and_time): Likewise.
7822         (ipa_merge_fn_summary_after_inlining): Likewise.
7823         * ipa-inline.c (cgraph_freq_base_rec): Remove.
7824         (compute_uninlined_call_time): Use sreal_frequency.
7825         (compute_inlined_call_time): Likewise.
7826         (ipa_inline): Do not initialize cgraph_freq_base_rec.
7827         * profile-count.c: Include sreal.h.
7828         (profile_count::to_sreal_scale): New.
7829         * profile-count.h: Forward declare sreal.
7830         (profile_count::to_sreal_scale): Declare.
7832 2017-11-13  Nathan Sidwell  <nathan@acm.org>
7834         * diagnostic.c (maybe_line_and_column): New.
7835         (diagnostic_get_location_text): Use it.
7836         (diagnostic_report_current_module): Likewise.
7837         (test_diagnostic_get_location_text): Add tests.
7839 2017-11-13  Luis Machado  <luis.machado@linaro.org>
7841         * doc/md.texi (Specifying processor pipeline description): Fix
7842         incorrect latency for the div instruction example.
7844 2017-11-13  Jakub Jelinek  <jakub@redhat.com>
7846         PR tree-optimization/78821
7847         * gimple-ssa-store-merging.c (compatible_load_p): Don't require
7848         that bit_not_p is the same.
7849         (imm_store_chain_info::coalesce_immediate_stores): Likewise.
7850         (split_group): Count precisely bit_not_p bits in each statement.
7851         (invert_op): New function.
7852         (imm_store_chain_info::output_merged_store): Use invert_op to
7853         emit BIT_XOR_EXPR with a xor_mask instead of BIT_NOT_EXPR if some
7854         but not all orig_stores have BIT_NOT_EXPR in the corresponding spots.
7856 2017-11-13  Martin Liska  <mliska@suse.cz>
7858         * gcov.c (struct coverage_info): Remove typedef of coverage_t.
7859         (struct source_info): Likewise.
7860         (add_branch_counts): Likewise.
7861         (add_line_counts): Likewise.
7862         (function_summary): Likewise.
7863         (output_intermediate_line): Likewise.
7864         (generate_results): Likewise.
7866 2017-11-13  Martin Liska  <mliska@suse.cz>
7868         * gcov.c (struct block_info): Remove typedef for block_t.
7869         (struct line_info): Likewise.
7870         (line_info::has_block): Likewise.
7871         (EXIT_BLOCK): Likewise.
7872         (unblock): Likewise.
7873         (circuit): Likewise.
7874         (get_cycles_count): Likewise.
7875         (process_file): Likewise.
7876         (read_graph_file): Likewise.
7877         (solve_flow_graph): Likewise.
7878         (find_exception_blocks): Likewise.
7879         (add_line_counts): Likewise.
7880         (accumulate_line_info): Likewise.
7881         (output_line_details): Likewise.
7883 2017-11-13  Martin Liska  <mliska@suse.cz>
7885         * gcov.c (struct arc_info): Remove typedef for arc_t.
7886         (struct line_info): Likewise.
7887         (add_branch_counts): Likewise.
7888         (output_branch_count): Likewise.
7889         (function_info::~function_info): Likewise.
7890         (circuit): Likewise.
7891         (output_intermediate_line): Likewise.
7892         (read_graph_file): Likewise.
7893         (solve_flow_graph): Likewise.
7894         (find_exception_blocks): Likewise.
7895         (add_line_counts): Likewise.
7896         (accumulate_line_info): Likewise.
7897         (output_line_details): Likewise.
7898         (output_function_details): Likewise.
7900 2017-11-13  Martin Liska  <mliska@suse.cz>
7902         * gcov.c (struct function_info): Remove typedef for function_t.
7903         (struct source_info): Likewise.
7904         (source_info::get_functions_at_location): Likewise.
7905         (solve_flow_graph): Likewise.
7906         (find_exception_blocks): Likewise.
7907         (add_line_counts): Likewise.
7908         (output_intermediate_file): Likewise.
7909         (process_file): Likewise.
7910         (generate_results): Likewise.
7911         (release_structures): Likewise.
7912         (read_graph_file): Likewise.
7913         (read_count_file): Likewise.
7914         (accumulate_line_counts): Likewise.
7915         (output_lines): Likewise.
7917 2017-11-13  Martin Liska  <mliska@suse.cz>
7919         * gcov.c (function_info::function_info): Remove num_counts
7920         and add vector<gcov_type>.
7921         (function_info::~function_info): Use the vector.
7922         (process_file): Likewise.
7923         (read_graph_file): Likewise.
7924         (read_count_file): Likewise.
7925         (solve_flow_graph): Likewise.
7927 2017-11-13  Martin Liska  <mliska@suse.cz>
7929         * gcov.c (function_info::is_artificial): New function.
7930         (process_file): Erase all artificial early.
7931         (generate_results): Skip as all artificial are already
7932         removed.
7934 2017-11-13  Martin Liska  <mliska@suse.cz>
7936         * gcov.c (read_graph_file): Store to global vector of functions.
7937         (read_count_file): Iterate the vector.
7938         (process_file): Likewise.
7939         (generate_results): Likewise.
7940         (release_structures): Likewise.
7942 2017-11-13  Jakub Jelinek  <jakub@redhat.com>
7944         PR tree-optimization/82954
7945         * gimple-ssa-store-merging.c
7946         (imm_store_chain_info::coalesce_immediate_stores): If
7947         !infof->ops[N].base_addr, split group if info->ops[N].base_addr.
7949 2017-11-13  Richard Sandiford  <richard.sandiford@linaro.org>
7951         * config/aarch64/aarch64-simd.md (aarch64_store_lane0<mode>):
7952         Upddate call to ENDIAN_LANE_N.
7953         (aarch64_<sur>dot_lane<vsi2qi>): Use aarch64_endian_lane_rtx.
7954         (aarch64_<sur>dot_laneq<vsi2qi>): Likewise.
7955         (*aarch64_simd_vec_copy_lane<mode>): Update calls to ENDIAN_LANE_N
7956         and use aarch64_endian_lane_rtx.
7957         (*aarch64_simd_vec_copy_lane_<vswap_width_name><mode>): Likewise.
7959 2017-11-12  Tom de Vries  <tom@codesourcery.com>
7961         * config/riscv/riscv.h (ASM_OUTPUT_LABELREF): Wrap in do {} while (0).
7963 2017-11-12  Tom de Vries  <tom@codesourcery.com>
7965         * config/elfos.h (ASM_OUTPUT_ASCII): Remove semicolon after macro body.
7967 2017-11-12  Tom de Vries  <tom@codesourcery.com>
7969         * config/cr16/cr16.h (ASM_OUTPUT_LABELREF): Remove semicolon after macro
7970         body.
7971         * config/powerpcspe/xcoff.h (ASM_OUTPUT_LABELREF): Same.
7972         * config/rs6000/xcoff.h (ASM_OUTPUT_LABELREF): Same.
7973         * defaults.h (ASM_OUTPUT_LABELREF): Same.
7975 2017-11-11  Martin Sebor  <msebor@redhat.com>
7977         PR c/81117
7978         * doc/extend.texi (attribute nonstring): Remove spurious argument.
7980         PR bootstrap/82948
7981         * prefic.c (translate_name): Replace strncpy with memcpy to
7982         avoid -Wstringop-truncation.
7984 2017-11-10  Jan Hubicka  <hubicka@ucw.cz>
7986         * tree-ssa-loop-im.c (execute_sm_if_changed): Do not compute freq_sum.
7988 2017-11-10  Jan Hubicka  <hubicka@ucw.cz>
7990         * predict.c (maybe_hot_frequency_p): Do not use cfun.
7992 2017-11-10  Jan Hubicka  <hubicka@ucw.cz>
7994         * tree-ssa-tail-merge.c (replace_block_by): Fix and re-enable profile
7995         merging.
7997 2017-11-10  Michael Meissner  <meissner@linux.vnet.ibm.com>
7999         * config/rs6000/rs6000.md (bswaphi2_reg): On ISA 3.0 systems,
8000         enable generating XXBRH if the value is in a vector register.
8001         (bswapsi2_reg): On ISA 3.0 systems, enable generating XXBRW if the
8002         value is in a vector register.
8003         (bswapdi2_reg): On ISA 3.0 systems, always use XXBRD to do
8004         register to register bswap64's instead of doing the GPR sequence
8005         used on previous machines.
8006         (bswapdi2_xxbrd): New insn.
8007         (bswapdi2_reg): Disallow on ISA 3.0.
8008         (register to register bswap64 splitter): Do not split the insn on
8009         ISA 3.0 systems that use XXBRD.
8011 2017-11-10  Martin Sebor  <msebor@redhat.com>
8013         PR c/81117
8014         * config/darwin-c.c (framework_construct_pathname): Replace strncpy
8015         with memcpy.
8016         (find_subframework_file): Same.
8018 2017-11-10  Jan Hubicka  <hubicka@ucw.cz>
8020         * auto-profile.c (afdo_indirect_call): Drop frequency.
8021         * cgraph.c (symbol_table::create_edge): Drop frequency argument.
8022         (cgraph_node::create_edge): Drop frequency argument.
8023         (cgraph_node::create_indirect_edge): Drop frequency argument.
8024         (cgraph_edge::make_speculative): Drop frequency arguments.
8025         (cgraph_edge::resolve_speculation): Do not update frequencies
8026         (cgraph_edge::dump_edge_flags): Do not dump frequency.
8027         (cgraph_node::dump): Check consistency in IPA mode.
8028         (cgraph_edge::maybe_hot_p): Use IPA counter.
8029         (cgraph_edge::verify_count_and_frequency): Rename to ...
8030         (cgraph_edge::verify_count): ... this one; drop frequency checking.
8031         (cgraph_node::verify_node): Update.
8032         * cgraph.h (struct cgraph_edge): Drop frequency.
8033         (cgraph_edge::frequency): New function.
8034         * cgraphbuild.c (pass_build_cgraph_edges::execute): Donot pass
8035         frequencies.
8036         (cgraph_edge::rebuild_edges): Likewise.
8037         * cgraphclones.c (cgraph_edge::clone): Scale only counts.
8038         (duplicate_thunk_for_node): Do not pass frequency.
8039         (cgraph_node::create_clone): Scale only counts.
8040         (cgraph_node::create_virtual_clone): Do not pass frequency.
8041         (cgraph_node::create_edge_including_clones): Do not pass frequency.
8042         (cgraph_node::create_version_clone): Do not pass frequency.
8043         * cgraphunit.c (cgraph_node::analyze): Do not pass frequency.
8044         (cgraph_node::expand_thunk): Do not pass frequency.
8045         (cgraph_node::create_wrapper): Do not pass frequency.
8046         * gimple-iterator.c (update_call_edge_frequencies): Do not pass
8047         frequency.
8048         * gimple-streamer-in.c (input_bb): Scale only IPA counts.
8049         * ipa-chkp.c (chkp_produce_thunks): Do not pass frequency.
8050         * ipa-cp.c (ipcp_lattice::print): Use frequency function.
8051         (gather_caller_stats): Use frequency function.
8052         (ipcp_cloning_candidate_p): Use frequency function.
8053         (ipcp_propagate_stage): Use frequency function.
8054         (get_info_about_necessary_edges): Use frequency function.
8055         (update_profiling_info): Update only IPA profile.
8056         (update_specialized_profile): Use frequency functoin.
8057         (perhaps_add_new_callers): Update only IPA profile.
8058         * ipa-devirt.c (ipa_devirt): Use IPA profile.
8059         * ipa-fnsummary.c (redirect_to_unreachable): Do not set frequrency.
8060         (dump_ipa_call_summary): Use frequency function.
8061         (estimate_edge_size_and_time): Use frequency function.
8062         (ipa_merge_fn_summary_after_inlining): Use frequency function.
8063         * ipa-inline-analysis.c (do_estimate_edge_time): Use IPA profile.
8064         * ipa-inline-transform.c (update_noncloned_frequencies): Rename to ..
8065         (update_noncloned_counts): ... ths one; scale counts only.
8066         (clone_inlined_nodes): Do not scale frequency.
8067         (inline_call): Do not pass frequency.
8068         * ipa-inline.c (compute_uninlined_call_time): Use IPA profile.
8069         (compute_inlined_call_time): Use IPA profile.
8070         (want_inline_small_function_p): Use IPA profile.
8071         (want_inline_self_recursive_call_p): Use IPA profile.
8072         (edge_badness): Use IPA profile.
8073         (lookup_recursive_calls): Use IPA profile.
8074         (recursive_inlining): Do not pass frequency.
8075         (resolve_noninline_speculation): Do not update frequency.
8076         (inline_small_functions): Collect max of IPA profile.
8077         (dump_overall_stats): Dump IPA porfile.
8078         (dump_inline_stats): Dump IPA porfile.
8079         (ipa_inline): Collect IPA stats.
8080         * ipa-inline.h (clone_inlined_nodes): Update prototype.
8081         * ipa-profile.c (ipa_propagate_frequency_1): Use frequency function.
8082         (ipa_propagate_frequency): Use frequency function.
8083         (ipa_profile): Cleanup.
8084         * ipa-prop.c (ipa_make_edge_direct_to_target): Do not pass frequency
8085         * ipa-utils.c (ipa_merge_profiles): Merge all profiles.
8086         * lto-cgraph.c (lto_output_edge): Do not stream frequency.
8087         (input_node): Do not stream frequency.
8088         (input_edge): Do not stream frequency.
8089         (merge_profile_summaries): Scale only IPA profiles.
8090         * omp-simd-clone.c (simd_clone_adjust): Do not pass frequency.
8091         * predict.c (drop_profile): Do not recompute frequency.
8092         * trans-mem.c (ipa_tm_insert_irr_call): Do not pass frequency.
8093         (ipa_tm_insert_gettmclone_call): Do not pass frequency.
8094         * tree-cfg.c (execute_fixup_cfg): Drop profile to global0 if needed.
8095         * tree-chkp.c (chkp_copy_bounds_for_assign): Do not pass frequency.
8096         * tree-emutls.c (gen_emutls_addr): Do not pass frequency.
8097         * tree-inline.c (copy_bb): Do not scale frequency.
8098         (expand_call_inline): Do not scale frequency.
8099         (tree_function_versioning): Do not scale frequency.
8100         * ubsan.c (ubsan_create_edge): Do not pass frequency.
8102 2017-11-10  Julia Koval  <julia.koval@intel.com>
8104         * config/i386/gfniintrin.h (_mm_gf2p8affine_epi64_epi8)
8105         (_mm256_gf2p8affine_epi64_epi8, _mm_mask_gf2p8affine_epi64_epi8)
8106         (_mm_maskz_gf2p8affine_epi64_epi8, _mm256_mask_gf2p8affine_epi64_epi8)
8107         (_mm256_maskz_gf2p8affine_epi64_epi8)
8108         (_mm512_mask_gf2p8affine_epi64_epi8, _mm512_gf2p8affine_epi64_epi8)
8109         (_mm512_maskz_gf2p8affine_epi64_epi8): New intrinsics.
8110         * config/i386/i386-builtin.def (__builtin_ia32_vgf2p8affineqb_v64qi)
8111         (__builtin_ia32_vgf2p8affineqb_v32qi)
8112         (__builtin_ia32_vgf2p8affineqb_v16qi): New builtins.
8113         * config/i386/sse.md (vgf2p8affineqb_<mode><mask_name>): New pattern.
8115 2017-11-10  Tamar Christina  <tamar.christina@arm.com>
8117         PR target/82641
8118         * config/arm/arm.c
8119         (arm_option_override): Refactor.
8120         (arm_option_reconfigure_globals): New.
8121         (arm_options_perform_arch_sanity_checks): New.
8122         * config/arm/arm-protos.h (arm_option_reconfigure_globals):
8123         New prototype.
8124         (arm_options_perform_arch_sanity_checks): Likewise
8126 2017-11-10  Pat Haugen  <pthaugen@us.ibm.com>
8128         * rs6000/power9.md (power9-qpdiv): Correct DFU pipe usage.
8129         (power9-qpmul): New.
8130         * rs6000/rs6000.md ("type" attr): Add qmul.
8131         (mul<mode>3, fma<mode>4_hw, *fms<mode>4_hw, *nfma<mode>4_hw,
8132         *nfms<mode>4_hw, mul<mode>3_odd, fma<mode>4_odd, *fms<mode>4_odd,
8133         *nfma<mode>4_odd, *nfms<mode>4_odd): Change type to qmul.
8135 2017-11-10  Martin Sebor  <msebor@redhat.com>
8137         PR c/81117
8138         * builtins.c (compute_objsize): Handle arrays that
8139         compute_builtin_object_size likes to fail for.  Make extern.
8140         * builtins.h (compute_objsize): Declare.
8141         (check_strncpy_sizes): New function.
8142         (expand_builtin_strncpy): Call check_strncpy_sizes.
8143         * gimple-fold.c (gimple_fold_builtin_strncpy): Implement
8144         -Wstringop-truncation.
8145         (gimple_fold_builtin_strncat): Same.
8146         * gimple.c (gimple_build_call_from_tree): Set call location.
8147         * tree-ssa-strlen.c (strlen_to_stridx): New global variable.
8148         (maybe_diag_bound_equal_length, is_strlen_related_p): New functions.
8149         (handle_builtin_stxncpy, handle_builtin_strncat): Same.
8150         (handle_builtin_strlen): Use strlen_to_stridx.
8151         (strlen_optimize_stmt): Handle flavors of strncat, strncpy, and
8152         stpncpy.
8153         Use strlen_to_stridx.
8154         (pass_strlen::execute): Release strlen_to_stridx.
8155         * doc/invoke.texi (-Wsizeof-pointer-memaccess): Document enhancement.
8156         (-Wstringop-truncation): Document new option.
8158 2017-11-10  Martin Liska  <mliska@suse.cz>
8160         PR gcov-profile/82702
8161         * gcov.c (main): Handle intermediate files in a different
8162         way.
8163         (get_gcov_intermediate_filename): New function.
8164         (output_gcov_file): Remove support of intermediate files.
8165         (generate_results): Allocate intermediate file.
8166         (release_structures): Clean-up properly fn_end.
8167         (output_intermediate_file): Start iterating with line 1.
8169 2017-11-10  Jakub Jelinek  <jakub@redhat.com>
8171         PR tree-optimization/82929
8172         * gimple-ssa-store-merging.c (struct store_immediate_info): Add
8173         ops_swapped_p non-static data member.
8174         (store_immediate_info::store_immediate_info): Clear it.
8175         (imm_store_chain_info::coalesce_immediate_stores): If swapping
8176         ops set ops_swapped_p.
8177         (count_multiple_uses): Handle ops_swapped_p.
8179 2017-11-10  Martin Liska  <mliska@suse.cz>
8181         * coverage.c (coverage_init): Stream information about
8182         support of has_unexecuted_blocks.
8183         * doc/gcov.texi: Document that.
8184         * gcov-dump.c (dump_gcov_file): Support it in gcov_dump tool.
8185         * gcov.c (read_graph_file): Likewise.
8186         (output_line_beginning): Fix a small issue with
8187         color output.
8189 2017-11-10  Bin Cheng  <bin.cheng@arm.com>
8191         * tree-predcom.c (determine_roots_comp): Avoid memory leak by freeing
8192         reference of trivial component.
8194 2017-11-10  Jakub Jelinek  <jakub@redhat.com>
8196         PR bootstrap/82916
8197         * gimple-ssa-store-merging.c
8198         (pass_store_merging::terminate_all_aliasing_chains): For
8199         gimple_store_p stmts also call refs_output_dependent_p.
8201         PR rtl-optimization/82913
8202         * compare-elim.c (try_merge_compare): Punt if def_insn is not
8203         single set.
8205 2017-11-09  Jeff Law  <law@redhat.com>
8207         * vr-values.h: New file with vr_values class.
8208         * tree-vrp.c: Include vr-values.h
8209         (vrp_value_range_pool, vrp_equiv_obstack, num_vr_values): Move static
8210         data objects into the vr_values class.
8211         (vr_value, values_propagated, vr_phi_edge_counts): Likewise.
8212         (get_value_range): Make it a member function within vr_values class.
8213         (set_defs_to_varying, update_value_range, add_equivalence): Likewise.
8214         (vrp_stmt_computes_nonzero_p, op_with_boolean_value_range_p): Likewise.
8215         (op_with_constant_singleton_value_range): Likewise.
8216         (extract_range_for_var_from_comparison_expr): Likewise.
8217         (extract_range_from_assert, extract_range_from_ssa_name): Likewise.
8218         (extract_range_from_binary_expr): Likewise.
8219         (extract_range_from_unary_expr): Likewise.
8220         (extract_range_from_cond_expr, extrat_range_from_comparison): Likewise.
8221         (check_for_binary_op_overflow, extract_range_basic): Likewise.
8222         (extract_range_from_assignment, adjust_range_with_scev): Likewise.
8223         (dump_all_value_ranges, get_vr_for_comparison): Likewise.
8224         (compare_name_with_value, compare_names): Likewise.
8225         (vrp_evaluate_conditional_warnv_with_ops_using_ranges): Likewise.
8226         (vrp_evaluate_conditional_warnv_with_ops): Likewise.  Remove prototype.
8227         (vrp_evaluate_conditional, vrp_visit_cond_stmt): Likewise.
8228         (vrp_visit_switch_stmt, extract_range_from_stmt): Likewise.
8229         (extract_range_from_phi_node): Likewise.
8230         (simplify_truth_ops_using_ranges): Likewise.
8231         (simplify_div_or_mod_using_ranges): Likewise.
8232         (simplify_min_or_max_using_ranges, simplify_abs_using_ranges): Likewise.
8233         (simplify_bit_ops_using_ranges, simplify_cond_using_ranges_1): Likewise.
8234         (simplify_cond_using_ranges_2, simplify_switch_using_ranges): Likewise.
8235         (simplify_float_conversion_using_ranges): Likewise.
8236         (simplify_internal_call_using_ranges): Likewise.
8237         (two_valued_val_range_p, simplify_stmt_using_ranges): Likewise.
8238         (vrp_visit_assignment_or_call): Likewise.  Smuggle class instance
8239         poitner via x_vr_values for calls into gimple folder.
8240         (vrp_initialize_lattice): Make this the vr_values ctor.
8241         (vrp_free_lattice): Make this the vr_values dtor.
8242         (set_vr_value): New function.
8243         (class vrp_prop): Add vr_values data member.  Add various member
8244         functions as well as member functions that delegate to vr_values.
8245         (check_array_ref): Make a member function within vrp_prop class.
8246         (search_for_addr_array, vrp_initialize): Likewise.
8247         (vrp_finalize): Likewise.  Revamp to avoid direct access to
8248         vr_value, values_propagated, etc.
8249         (check_array_bounds): Extract vrp_prop class instance pointer from
8250         walk info structure.  Use it to call member functions.
8251         (check_all_array_refs): Make a member function within vrp_prop class.
8252         Smuggle class instance pointer via walk info structure.
8253         (x_vr_values): New local static.
8254         (vrp_valueize): Use x_vr_values to get class instance.
8255         (vr_valueize_1): Likewise.
8256         (class vrp_folder): Add vr_values data member.  Add various member
8257         functions as well as member functions that delegate to vr_values.
8258         (fold_predicate_in): Make a mber fucntion within vrp_folder class.
8259         (simplify_stmt_for_jump_threading): Extract smuggled vr_values
8260         class instance from vr_values.  Use it to call member functions.
8261         (vrp_dom_walker): Add vr_values data member.
8262         (vrp_dom_walker::after_dom_children): Smuggle vr_values class
8263         instance via x_vr_values.
8264         (identify_jump_threads): Accept vr_values as argument.  Store
8265         it into the walker structure.
8266         (evrp_dom_walker): Add vr_values class data member.  Add various
8267         delegators.
8268         (evrp_dom_walker::try_find_new_range): Use vr_values data
8269         member to access the memory allocator.
8270         (evrp_dom_walker::before_dom_children): Store vr_values class
8271         instance into the vrp_folder class.
8272         (evrp_dom_walker::push_value_range): Rework to avoid direct
8273         access to num_vr_values and vr_value.
8274         (evrp_dom_walker::pop_value_range): Likewise.
8275         (execute_early_vrp): Remove call to vrp_initialize_lattice.
8276         Use vr_values to get to dump_all_value_ranges member function.
8277         Remove call to vrp_free_lattice.  Call vrp_initialize, vrp_finalize,
8278         and simplify_cond_using_ranges_2 via vrp_prop class instance.
8279         Pass vr_values class instance down to identify_jump_threads.
8280         Remove call to vrp_free_lattice.
8281         (debug_all_value_ranges): Remove.
8283         * tree-vrp.c (vrp_prop): Move class to earlier point in the file.
8284         (vrp_folder): Likewise.
8286         * tree-vrp.c (set_value_range): Do not reference vrp_equiv_obstack.
8287         Get it from the existing bitmap instead.
8288         (vrp_intersect_ranges_1): Likewise.
8290 2017-11-09  Jakub Jelinek  <jakub@redhat.com>
8292         * gimple-ssa-store-merging.c (struct store_immediate_info): Add
8293         bit_not_p field.
8294         (store_immediate_info::store_immediate_info): Add bitnotp argument,
8295         set bit_not_p to it.
8296         (imm_store_chain_info::coalesce_immediate_stores): Break group
8297         if bit_not_p is different.
8298         (count_multiple_uses, split_group,
8299         imm_store_chain_info::output_merged_store): Handle info->bit_not_p.
8300         (handled_load): Avoid multiple chained BIT_NOT_EXPRs.
8301         (pass_store_merging::process_store): Handle BIT_{AND,IOR,XOR}_EXPR
8302         result inverted using BIT_NOT_EXPR, compute bit_not_p, pass it
8303         to store_immediate_info ctor.
8305 2017-11-09  Jim Wilson  <jimw@sifive.com>
8307         * collect2.c (OBJECT_FORMAT_COFF): Remove EXTENDED_COFF support.
8308         (scan_prog_file): Likewise.
8310 2017-11-09  Jan Hubicka  <hubicka@ucw.cz>
8312         * bb-reorder.c (max_entry_frequency): Remove.
8313         (find_traces, rotate_loop, mark_bb_visited, connect_better_edge_p,
8314         connect_traces, push_to_next_round_p): Remove prototypes.
8315         (find_traces_1_round): Use counts only.
8316         (push_to_next_round_p): Likewise.
8317         (find_traces): Likewise.
8318         (rotate_loop): Likewise.
8319         (find_traces_1_round): Likewise.
8320         (connect_traces): Likewise.
8321         (edge_order): Likewise.
8323 2017-11-09  Thomas Preud'homme  <thomas.preudhomme@arm.com>
8325         * config/arm/arm.c (output_return_instruction): Add comments to
8326         indicate requirement for cmse_nonsecure_entry return to account
8327         for the size of clearing instruction output here.
8328         (thumb_exit): Likewise.
8329         * config/arm/thumb2.md (thumb2_cmse_entry_return): Fix length for
8330         return in hardfloat mode.
8332 2017-11-09  Segher Boessenkool  <segher@kernel.crashing.org>
8334         * config/rs6000/rs6000.c (machine_function): Add a bool,
8335         "toc_is_wrapped_separately".
8336         (rs6000_option_override_internal): Enable OPTION_MASK_SAVE_TOC_INDIRECT
8337         if it wasn't explicitly set or unset, we are optimizing for speed, and
8338         doing separate shrink-wrapping.
8339         (rs6000_get_separate_components): Enable the TOC component if
8340         saving the TOC register in the prologue.
8341         (rs6000_components_for_bb): Handle the TOC component.
8342         (rs6000_emit_prologue_components): Store the TOC register where needed.
8343         (rs6000_set_handled_components): Mark TOC as handled, if handled.
8344         (rs6000_emit_prologue): Don't save the TOC if that is already done.
8346 2017-11-09  Martin Jambor  <mjambor@suse.cz>
8348         * ipa-param-manipulation.c: New file.
8349         * ipa-param-manipulation.h: Likewise.
8350         * Makefile.in (OBJS): Add ipa-param-manipulation.o.
8351         (PLUGIN_HEADERS): Addded ipa-param-manipulation.h
8352         * ipa-param.h (ipa_parm_op): Moved to ipa-param-manipulation.h.
8353         (ipa_parm_adjustment): Likewise.
8354         (ipa_parm_adjustment_vec): Likewise.
8355         (ipa_get_vector_of_formal_parms): Moved declaration to
8356         ipa-param-manipulation.h.
8357         (ipa_get_vector_of_formal_parm_types): Likewise.
8358         (ipa_modify_formal_parameters): Likewise.
8359         (ipa_modify_call_arguments): Likewise.
8360         (ipa_combine_adjustments): Likewise.
8361         (ipa_dump_param_adjustments): Likewise.
8362         (ipa_modify_expr): Likewise.
8363         (ipa_get_adjustment_candidate): Likewise.
8364         * ipa-prop.c (ipa_get_vector_of_formal_parms): Moved to
8365         ipa-param-manipulation.c.
8366         (ipa_get_vector_of_formal_parm_types): Likewise.
8367         (ipa_modify_formal_parameters): Likewise.
8368         (ipa_modify_call_arguments): Likewise.
8369         (ipa_modify_expr): Likewise.
8370         (get_ssa_base_param): Likewise.
8371         (ipa_get_adjustment_candidate): Likewise.
8372         (index_in_adjustments_multiple_times_p): Likewise.
8373         (ipa_combine_adjustments): Likewise.
8374         (ipa_dump_param_adjustments): Likewise.
8375         * tree-sra.c: Also include ipa-param-manipulation.h
8376         * omp-simd-clone.c: Include ipa-param-manipulation.h instead of
8377         ipa-param.h.
8379 2017-11-09  Richard Sandiford  <richard.sandiford@linaro.org>
8380             Alan Hayward  <alan.hayward@arm.com>
8381             David Sherwood  <david.sherwood@arm.com>
8383         * doc/sourcebuild.texi (vect_masked_store): Document.
8385 2017-11-09  Richard Sandiford  <richard.sandiford@linaro.org>
8386             Alan Hayward  <alan.hayward@arm.com>
8387             David Sherwood  <david.sherwood@arm.com>
8389         * doc/sourcebuild.texi (vect_align_stack_vars): Document.
8391 2017-11-09  Richard Sandiford  <richard.sandiford@linaro.org>
8392             Alan Hayward  <alan.hayward@arm.com>
8393             David Sherwood  <david.sherwood@arm.com>
8395         * doc/sourcebuild.texi (vect_variable_length): Document.
8397 2017-11-09  Richard Sandiford  <richard.sandiford@linaro.org>
8398             Alan Hayward  <alan.hayward@arm.com>
8399             David Sherwood  <david.sherwood@arm.com>
8401         * doc/sourcebuild.texi (vect_unaligned_possible): Document.
8403 2017-11-09  Richard Sandiford  <richard.sandiford@linaro.org>
8404             Alan Hayward  <alan.hayward@arm.com>
8405             David Sherwood  <david.sherwood@arm.com>
8407         * doc/sourcebuild.texi (vect_element_align_preferred): Document.
8409 2017-11-09  Richard Sandiford  <richard.sandiford@linaro.org>
8410             Alan Hayward  <alan.hayward@arm.com>
8411             David Sherwood  <david.sherwood@arm.com>
8413         * doc/sourcebuild.texi (vect_perm_short, vect_perm_byte): Document
8414         previously undocumented selectors.
8415         (vect_perm3_byte, vect_perm3_short, vect_perm3_int): Document.
8417 2017-11-09  Richard Sandiford  <richard.sandiford@linaro.org>
8419         * doc/rtl.texi (const_vector): Say that elements can be
8420         const_wide_ints too.
8421         * emit-rtl.h (valid_for_const_vec_duplicate_p): Declare.
8422         * emit-rtl.c (valid_for_const_vec_duplicate_p): New function.
8423         (gen_vec_duplicate): Use it instead of CONSTANT_P.
8424         * optabs.c (expand_vector_broadcast): Likewise.
8426 2017-11-09  Richard Sandiford  <richard.sandiford@linaro.org>
8427             Alan Hayward  <alan.hayward@arm.com>
8428             David Sherwood  <david.sherwood@arm.com>
8430         * tree-ssa-loop-ivopts.c (get_address_cost): Try using a
8431         scaled index even if the unscaled address was invalid.
8432         Don't increase the complexity of using a scale in that case.
8434 2017-11-09  Richard Sandiford  <richard.sandiford@linaro.org>
8435             Alan Hayward  <alan.hayward@arm.com>
8436             David Sherwood  <david.sherwood@arm.com>
8438         * doc/rtl.texi: Rewrite the subreg rules so that they partition
8439         the inner register into REGMODE_NATURAL_SIZE bytes rather than
8440         UNITS_PER_WORD bytes.
8441         * emit-rtl.c (validate_subreg): Divide subregs into blocks
8442         based on REGMODE_NATURAL_SIZE of the inner mode.
8443         (gen_lowpart_common): Split the SCALAR_FLOAT_MODE_P and
8444         !SCALAR_FLOAT_MODE_P cases.  Use REGMODE_NATURAL_SIZE for the latter.
8445         * expmed.c (lowpart_bit_field_p): Divide the value up into
8446         chunks of REGMODE_NATURAL_SIZE rather than UNITS_PER_WORD.
8447         * expr.c (store_constructor): Use REGMODE_NATURAL_SIZE to test
8448         whether something is likely to occupy more than one register.
8450 2017-11-09  Jan Hubicka  <hubicka@ucw.cz>
8452         PR ipa/82879
8453         * ipa-inline-transform.c (update_noncloned_frequencies): Use
8454         profile_count::adjust_for_ipa_scaling.
8455         * tree-inline.c (copy_bb, copy_cfg_body): Likewise.
8456         * profile-count.c (profile_count::adjust_for_ipa_scaling): New member
8457         function.
8458         * profile-count.h (profile_count::adjust_for_ipa_scaling): Declare.
8460 2017-11-09  Jakub Jelinek  <jakub@redhat.com>
8462         * gimple-ssa-store-merging.c (count_multiple_uses): New function.
8463         (split_group): Add total_orig and total_new arguments, estimate the
8464         number of statements related to the store group without store merging
8465         and with store merging.
8466         (imm_store_chain_info::output_merged_store): Adjust split_group
8467         callers, punt if estimated number of statements with store merging
8468         is not smaller than estimated number of statements without it.
8469         Formatting fix.
8470         (handled_load): Remove has_single_use checks.
8471         (pass_store_merging::process_store): Likewise.
8473 2017-11-09  Richard Biener  <rguenther@suse.de>
8475         PR tree-optimization/82902
8476         * tree-ssa-phiprop.c (propagate_with_phi): Test proper type.
8478 2017-11-09  Martin Liska  <mliska@suse.cz>
8480         PR target/82863
8481         * emit-rtl.c (init_emit_regs): Initialize split_branch_probability to
8482         uninitialized.
8484 2017-11-09  Martin Liska  <mliska@suse.cz>
8486         PR tree-optimization/82669
8487         * sbitmap.h (bmp_iter_set_init): Remove non needed check.
8489 2017-11-09  Martin Liska  <mliska@suse.cz>
8491         PR gcov-profile/48463
8492         * coverage.c (coverage_begin_function): Output also end locus
8493         of a function and information whether the function is
8494         artificial.
8495         * gcov-dump.c (tag_function): Parse and print the information.
8496         * gcov.c (INCLUDE_MAP): Add include.
8497         (INCLUDE_SET): Likewise.
8498         (struct line_info): Move earlier in the source file because
8499         of vector<line_info> in function_info structure.
8500         (line_info::line_info): Likewise.
8501         (line_info::has_block): Likewise.
8502         (struct source_info): Add new member index.
8503         (source_info::get_functions_at_location): New function.
8504         (function_info::group_line_p): New function.
8505         (output_intermediate_line): New function.
8506         (output_intermediate_file): Use the mentioned function.
8507         (struct function_start): New.
8508         (struct function_start_pair_hash): Likewise.
8509         (process_file): Add code that identifies group functions.
8510         Assign lines either to global or function scope.
8511         (generate_results): Skip artificial functions.
8512         (find_source): Assign index for each source file.
8513         (read_graph_file): Read new flag artificial and end_line.
8514         (add_line_counts): Assign it either to global of function scope.
8515         (accumulate_line_counts): Isolate core of the function to
8516         accumulate_line_info and call it for both function and global
8517         scope lines.
8518         (accumulate_line_info): New function.
8519         (output_line_beginning): Fix GNU coding style.
8520         (print_source_line): New function.
8521         (output_line_details): Likewise.
8522         (output_function_details): Likewise.
8523         (output_lines): Iterate both source (global) scope and function
8524         scope.
8525         (struct function_line_start_cmp): New class.
8526         * doc/gcov.texi: Reflect changes in documentation.
8528 2017-11-09  Jakub Jelinek  <jakub@redhat.com>
8530         PR debug/82837
8531         * dwarf2out.c (const_ok_for_output_1): Reject NEG in addition to NOT.
8532         (mem_loc_descriptor): Handle (const (neg (...))) as (neg (const (...)))
8533         and similarly for not instead of neg.
8535 2017-11-08  Andi Kleen  <ak@linux.intel.com>
8537         * config/i386/i386.opt: Add -mforce-indirect-call.
8538         * config/i386/predicates.md: Check for flag_force_indirect_call.
8539         * doc/invoke.texi: Document -mforce-indirect-call
8541 2017-11-08  Kito Cheng  <kito.cheng@gmail.com>
8543         * config/riscv/riscv-protos.h (riscv_slow_unaligned_access_p):
8544         New extern.
8545         (MOVE_RATIO): Use riscv_slow_unaligned_access_p.
8546         config/riscv/riscv.c (predict.h): New include.
8547         (riscv_slow_unaligned_access_p): No longer static.
8548         (riscv_block_move_straight): Add require.
8549         * config/riscv/riscv-protos.h (riscv_hard_regno_nregs): Delete.
8551 2017-11-08  Jakub Jelinek  <jakub@redhat.com>
8553         PR target/82855
8554         * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>,
8555         <avx512>_eq<mode>3<mask_scalar_merge_name>_1): Use
8556         nonimmediate_operand predicate for operand 1 instead of
8557         register_operand.
8559 2017-11-08  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
8561         * config/aarch64/aarch64-simd.md (store_pair_lanes<mode>):
8562         New pattern.
8563         * config/aarch64/constraints.md (Uml): New constraint.
8564         * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): New
8565         predicate.
8567 2017-11-08  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
8569         * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
8570         of two vec_duplicates into a vec_concat.
8572 2017-11-08  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
8574         * simplify-rtx.c (simplify_ternary_operation, VEC_MERGE):
8575         Simplify vec_merge of vec_duplicate and vec_concat.
8576         * config/aarch64/constraints.md (Utq): New constraint.
8577         * config/aarch64/aarch64-simd.md (load_pair_lanes<mode>): New
8578         define_insn.
8580 2017-11-08  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
8582         * simplify-rtx.c (simplify_ternary_operation, VEC_MERGE):
8583         Simplify vec_merge of vec_duplicate and const_vector.
8584         * config/aarch64/predicates.md (aarch64_simd_or_scalar_imm_zero):
8585         New predicate.
8586         * config/aarch64/aarch64-simd.md (*aarch64_combinez<mode>): Use VDC
8587         mode iterator.  Update predicate on operand 1 to
8588         handle non-const_vec constants.  Delete constraints.
8589         (*aarch64_combinez_be<mode>): Likewise for operand 2.
8591 2017-11-08  Jakub Jelinek  <jakub@redhat.com>
8593         PR tree-optimization/78821
8594         * gimple-ssa-store-merging.c (struct store_operand_info): Add bit_not_p
8595         data member.
8596         (store_operand_info::store_operand_info): Initialize it to false.
8597         (pass_store_merging::terminate_all_aliasing_chains): Rewritten to use
8598         ref_maybe_used_by_stmt_p and stmt_may_clobber_ref_p on lhs of each
8599         store in the group, and if chain_info is non-NULL, to ignore altogether
8600         that chain.
8601         (compatible_load_p): Fail if bit_not_p does not match.
8602         (imm_store_chain_info::output_merged_store): Handle bit_not_p loads.
8603         (handled_load): Fill in bit_not_p.  Handle BIT_NOT_EXPR.
8604         (pass_store_merging::process_store): Adjust
8605         terminate_all_aliasing_chains calls to pass NULL in all current spots,
8606         call terminate_all_aliasing_chains newly when adding a store into
8607         a chain with non-NULL chain_info.
8609 2017-11-08  Wilco Dijkstra  <wdijkstr@arm.com>
8611         * config/aarch64/aarch64.c (aarch64_can_eliminate): Simplify logic.
8613 2017-11-08  Wilco Dijkstra  <wdijkstr@arm.com>
8615         * config/aarch64/aarch64.c (aarch64_frame_pointer_required)
8616         Remove.
8617         (aarch64_layout_frame): Initialise emit_frame_chain.
8618         (aarch64_can_eliminate): Remove omit leaf frame pointer code.
8619         (TARGET_FRAME_POINTER_REQUIRED): Remove define.
8621 2017-11-08  Martin Liska  <mliska@suse.cz>
8623         * gimplify.c (expand_FALLTHROUGH_r): Simplify usage
8624         of gimple_call_internal_p.
8626 2017-11-07  Tom de Vries  <tom@codesourcery.com>
8628         * config/mips/mips.h (ASM_OUTPUT_LABELREF): Wrap in "do {} while (0)".
8630 2017-11-07  Tom de Vries  <tom@codesourcery.com>
8632         * config/mips/mips.h (ASM_OUTPUT_CASE_END): Remove semicolon after
8633         "do {} while (0)".
8635 2017-11-08  Martin Liska  <mliska@suse.cz>
8637         PR sanitizer/82792
8638         * gimplify.c (expand_FALLTHROUGH_r): Skip IFN_ASAN_MARK.
8640 2017-11-07  Eric Botcazou  <ebotcazou@adacore.com>
8642         * gimple-pretty-print.c (dump_profile): Return "" instead of NULL.
8644 2017-11-07  Jakub Jelinek  <jakub@redhat.com>
8646         PR target/82855
8647         * config/i386/i386.md (SWI1248_AVX512BWDQ2_64): New mode iterator.
8648         (*cmp<mode>_ccz_1): New insn with $k alternative.
8650         PR target/82855
8651         * config/i386/i386.c (ix86_swap_binary_operands_p): Treat
8652         RTX_COMM_COMPARE as commutative as well.
8653         (ix86_binary_operator_ok): Formatting fix.
8654         * config/i386/sse.md (*mul<mode>3<mask_name><round_name>,
8655         *<code><mode>3<mask_name><round_saeonly_name>,
8656         *<code><mode>3<mask_name>, *<code>tf3, *mul<mode>3<mask_name>,
8657         *<s>mul<mode>3_highpart<mask_name>,
8658         *vec_widen_umult_even_v16si<mask_name>,
8659         *vec_widen_umult_even_v8si<mask_name>,
8660         *vec_widen_umult_even_v4si<mask_name>,
8661         *vec_widen_smult_even_v16si<mask_name>,
8662         *vec_widen_smult_even_v8si<mask_name>, *sse4_1_mulv2siv2di3<mask_name>,
8663         *avx2_pmaddwd, *sse2_pmaddwd, *<sse4_1_avx2>_mul<mode>3<mask_name>,
8664         *avx2_<code><mode>3, *avx512f_<code><mode>3<mask_name>,
8665         *sse4_1_<code><mode>3<mask_name>, *<code>v8hi3,
8666         *sse4_1_<code><mode>3<mask_name>, *<code>v16qi3, *avx2_eq<mode>3,
8667         <avx512>_eq<mode>3<mask_scalar_merge_name>_1, *sse4_1_eqv2di3,
8668         *sse2_eq<mode>3, <mask_codefor><code><mode>3<mask_name>,
8669         *<code><mode>3, *<sse2_avx2>_uavg<mode>3<mask_name>,
8670         *<ssse3_avx2>_pmulhrsw<mode>3<mask_name>, *ssse3_pmulhrswv4hi3): Use
8671         !(MEM_P (operands[1]) && MEM_P (operands[2])) condition instead of
8672         ix86_binary_operator_ok.  Formatting fixes.
8673         (*<plusminus_insn><mode>3<mask_name><round_name>,
8674         *<plusminus_insn><mode>3, *<plusminus_insn><mode>3_m): Formatting
8675         fixes.
8677 2017-11-07  Segher Boessenkool  <segher@kernel.crashing.org>
8679         * config/rs6000/rs6000.md (GPR2): New mode_iterator.
8680         ("cstore<mode>4"): Don't always expand with rs6000_emit_int_cmove for
8681         eq and ne if TARGET_ISEL.
8682         (cmp): New code_iterator.
8683         (UNS, UNSU_, UNSIK): New code_attrs.
8684         (<code><GPR:mode><GPR2:mode>2_isel): New define_insn_and_split.
8685         ("eq<mode>3"): New define_expand, rename the define_insn_and_split
8686         to...
8687         ("eq<mode>3"): ... this.
8688         ("ne<mode>3"): New define_expand, rename the define_insn_and_split
8689         to...
8690         ("ne<mode>3"): ... this.
8692 2017-11-07  Julia Koval  <julia.koval@intel.com>
8694         PR target/82812
8695         * common/config/i386/i386-common.c
8696         (OPTION_MASK_ISA_GENERAL_REGS_ONLY_UNSET): Remove MPX from flag.
8697         (ix86_handle_option): Move MPX to isa_flags2 and GFNI to isa_flags.
8698         * config/i386/i386-c.c (ix86_target_macros_internal): Ditto.
8699         * config/i386/i386.opt: Ditto.
8700         * config/i386/i386.c (ix86_target_string): Ditto.
8701         (ix86_option_override_internal): Ditto.
8702         (ix86_init_mpx_builtins): Move MPX to args2.
8703         (ix86_expand_builtin): Special handling for OPTION_MASK_ISA_GFNI.
8704         * config/i386/i386-builtin.def (__builtin_ia32_vgf2p8affineinvqb_v64qi,
8705         __builtin_ia32_vgf2p8affineinvqb_v64qi_mask,
8706         __builtin_ia32_vgf2p8affineinvqb_v32qi,
8707         __builtin_ia32_vgf2p8affineinvqb_v32qi_mask,
8708         __builtin_ia32_vgf2p8affineinvqb_v16qi,
8709         __builtin_ia32_vgf2p8affineinvqb_v16qi_mask): Move to ARGS array.
8711 2017-11-07  Uros Bizjak  <ubizjak@gmail.com>
8713         PR target/80425
8714         * config/i386.i386.md (*zero_extendsidi2): Change (?r,*Yj), (?*Yi,r)
8715         and (*x,m) to ($r,Yj), ($Yi,r) and ($x,m).
8716         (zero-extendsidi peephole2): Remove peephole.
8718 2017-11-07  Eric Botcazou  <ebotcazou@adacore.com>
8720         PR c/53037
8721         * stor-layout.c: Include attribs.h.
8722         (handle_warn_if_not_align): Replace test on TYPE_USER_ALIGN with
8723         explicit lookup of "aligned" attribute.
8725 2017-11-07  Andrew Waterman  <andrew@sifive.com>
8727         * config/riscv/riscv-protos.h (riscv_hard_regno_nregs): New prototype.
8728         (riscv_expand_block_move): Likewise.
8729         * config/riscv/riscv.h (MOVE_RATIO): Tune cost to movmemsi
8730         implementation.
8731         (RISCV_MAX_MOVE_BYTES_PER_LOOP_ITER): New define.
8732         (RISCV_MAX_MOVE_BYTES_STRAIGHT): New define.
8733         * config/riscv/riscv.c (riscv_block_move_straight): New function.
8734         (riscv_adjust_block_mem): Likewise.
8735         (riscv_block_move_loop): Likewise.
8736         (riscv_expand_block_move): Likewise.
8737         * config/riscv/riscv.md (movmemsi): New pattern.
8739 2017-11-07  Michael Clark  <michaeljclark@mac.com>
8741         * config/riscv/linux.h (MUSL_ABI_SUFFIX): New define.
8742         (MUSL_DYNAMIC_LINKER): Likewise.
8744 2017-11-07  Richard Sandiford  <richard.sandiford@linaro.org>
8746         * config/aarch64/aarch64.md (ashl<mode>3, ashr<mode>3, lshr<mode>3)
8747         (rotr<mode>3, rotl<mode>3): Use aarch64_reg_or_imm instead of
8748         nonmmory_operand.
8750 2017-11-07  Richard Biener  <rguenther@suse.de>
8752         * match.pd: Fix build.
8754 2017-11-07  Wilco Dijkstra  <wdijkstr@arm.com>
8755             Jackson Woodruff  <jackson.woodruff@arm.com>
8757         PR tree-optimization/71026
8758         * match.pd: Canonicalize negate in division.
8760 2017-11-07  Sudakshina Das  <sudi.das@arm.com>
8762         PR middle-end/80131
8763         * match.pd: Simplify 1 << (C - x) where C = precision (x) - 1.
8765 2017-11-07  Marc Glisse  <marc.glisse@inria.fr>
8767         * match.pd ((a&~b)|(a^b),(a&~b)^~a,(a|b)&~(a^b),a|~(a^b),
8768         (a|b)|(a&^b),(a&b)|~(a^b),~(~a&b),~X^Y): New transformations.
8770 2017-11-07  Marc Glisse  <marc.glisse@inria.fr>
8772         * fold-const.c (negate_expr_p) [PLUS_EXPR, MINUS_EXPR]: Handle
8773         non-scalar integral types.
8774         * match.pd (negate_expr_p): Handle MINUS_EXPR.
8775         (-(A-B), -(~A)): New transformations.
8777 2017-11-07  Tom de Vries  <tom@codesourcery.com>
8779         * config/powerpcspe/aix43.h (SUBTARGET_OVERRIDE_OPTIONS): Remove
8780         semicolon after "do {} while (0)".
8781         * config/powerpcspe/aix51.h (SUBTARGET_OVERRIDE_OPTIONS): Same.
8782         * config/powerpcspe/aix52.h (SUBTARGET_OVERRIDE_OPTIONS): Same.
8783         * config/powerpcspe/aix53.h (SUBTARGET_OVERRIDE_OPTIONS): Same.
8784         * config/powerpcspe/aix61.h (SUBTARGET_OVERRIDE_OPTIONS): Same.
8785         * config/powerpcspe/aix71.h (SUBTARGET_OVERRIDE_OPTIONS): Same.
8787 2017-11-07  Tom de Vries  <tom@codesourcery.com>
8789         * config/rs6000/aix43.h (SUBTARGET_OVERRIDE_OPTIONS): Remove semicolon
8790         after "do {} while (0)".
8791         * config/rs6000/aix51.h (SUBTARGET_OVERRIDE_OPTIONS): Same.
8792         * config/rs6000/aix52.h (SUBTARGET_OVERRIDE_OPTIONS): Same.
8793         * config/rs6000/aix53.h (SUBTARGET_OVERRIDE_OPTIONS): Same.
8794         * config/rs6000/aix61.h (SUBTARGET_OVERRIDE_OPTIONS): Same.
8795         * config/rs6000/aix71.h (SUBTARGET_OVERRIDE_OPTIONS): Same.
8797 2017-11-07  Tom de Vries  <tom@codesourcery.com>
8799         PR other/82784
8800         * config/arm/arm.c (HANDLE_NARROW_SHIFT_ARITH): Remove semicolon after
8801         "while {} do (0)".
8802         (arm_rtx_costs_internal): Add missing semicolon after
8803         HANDLE_NARROW_SHIFT_ARITH call.
8805 2017-11-06  Segher Boessenkool  <segher@kernel.crashing.org>
8807         * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
8808         disable isel if it was not set explicitly.
8810 2017-11-06  James Bowman  <james.bowman@ftdichip.com>
8812         * gcc/dwarf2out.c (modified_type_die): Retain ADDR_SPACE qualifiers.
8813         (add_type_attribute) likewise.
8815 2017-11-06  H.J. Lu  <hongjiu.lu@intel.com>
8817         * config/i386/i386.c (ix86_can_use_return_insn_p): Use reference
8818         of struct ix86_frame.
8819         (ix86_initial_elimination_offset): Likewise.
8820         (ix86_expand_split_stack_prologue): Likewise.
8822 2017-11-06  Marc Glisse  <marc.glisse@inria.fr>
8824         * tree-vrp.h (enum value_range_type): Update stale comment.
8826 2017-11-06  Richard Sandiford  <richard.sandiford@linaro.org>
8827             Alan Hayward  <alan.hayward@arm.com>
8828             David Sherwood  <david.sherwood@arm.com>
8830         * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm)
8831         (aarch64_expand_vec_perm_const): Take the number of units too.
8832         * config/aarch64/aarch64.c (aarch64_expand_vec_perm)
8833         (aarch64_expand_vec_perm_const): Likewise.
8834         * config/aarch64/aarch64-simd.md (vec_perm_const<mode>)
8835         (vec_perm<mode>): Update accordingly.
8837 2017-11-06  Richard Sandiford  <richard.sandiford@linaro.org>
8838             Alan Hayward  <alan.hayward@arm.com>
8839             David Sherwood  <david.sherwood@arm.com>
8841         * config/aarch64/aarch64-protos.h (aarch64_simd_vect_par_cnst_half):
8842         Take the number of units too.
8843         * config/aarch64/aarch64.c (aarch64_simd_vect_par_cnst_half): Likewise.
8844         (aarch64_simd_check_vect_par_cnst_half): Update call accordingly,
8845         but check for a vector mode before rather than after the call.
8846         * config/aarch64/aarch64-simd.md (aarch64_split_simd_mov<mode>)
8847         (move_hi_quad_<mode>, vec_unpack<su>_hi_<mode>)
8848         (vec_unpack<su>_lo_<mode, vec_widen_<su>mult_lo_<mode>)
8849         (vec_widen_<su>mult_hi_<mode>, vec_unpacks_lo_<mode>)
8850         (vec_unpacks_hi_<mode>, aarch64_saddl2<mode>, aarch64_uaddl2<mode>)
8851         (aarch64_ssubl2<mode>, aarch64_usubl2<mode>, widen_ssum<mode>3)
8852         (widen_usum<mode>3, aarch64_saddw2<mode>, aarch64_uaddw2<mode>)
8853         (aarch64_ssubw2<mode>, aarch64_usubw2<mode>, aarch64_sqdmlal2<mode>)
8854         (aarch64_sqdmlsl2<mode>, aarch64_sqdmlal2_lane<mode>)
8855         (aarch64_sqdmlal2_laneq<mode>, aarch64_sqdmlsl2_lane<mode>)
8856         (aarch64_sqdmlsl2_laneq<mode>, aarch64_sqdmlal2_n<mode>)
8857         (aarch64_sqdmlsl2_n<mode>, aarch64_sqdmull2<mode>)
8858         (aarch64_sqdmull2_lane<mode>, aarch64_sqdmull2_laneq<mode>)
8859         (aarch64_sqdmull2_n<mode>): Update accordingly.
8861 2017-11-06  Richard Sandiford  <richard.sandiford@linaro.org>
8862             Alan Hayward  <alan.hayward@arm.com>
8863             David Sherwood  <david.sherwood@arm.com>
8865         * config/aarch64/aarch64-protos.h (aarch64_reverse_mask): Take
8866         the number of units too.
8867         * config/aarch64/aarch64.c (aarch64_reverse_mask): Likewise.
8868         * config/aarch64/aarch64-simd.md (vec_load_lanesoi<mode>)
8869         (vec_store_lanesoi<mode>, vec_load_lanesci<mode>)
8870         (vec_store_lanesci<mode>, vec_load_lanesxi<mode>)
8871         (vec_store_lanesxi<mode>): Update accordingly.
8873 2017-11-06  Richard Sandiford  <richard.sandiford@linaro.org>
8874             Alan Hayward  <alan.hayward@arm.com>
8875             David Sherwood  <david.sherwood@arm.com>
8877         * config/aarch64/aarch64-protos.h (aarch64_endian_lane_rtx): Declare.
8878         * config/aarch64/aarch64.c (aarch64_endian_lane_rtx): New function.
8879         * config/aarch64/aarch64.h (ENDIAN_LANE_N): Take the number
8880         of units rather than the mode.
8881         * config/aarch64/iterators.md (nunits): New mode attribute.
8882         * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args):
8883         Use aarch64_endian_lane_rtx instead of GEN_INT (ENDIAN_LANE_N ...).
8884         * config/aarch64/aarch64-simd.md (aarch64_dup_lane<mode>)
8885         (aarch64_dup_lane_<vswap_width_name><mode>, *aarch64_mul3_elt<mode>)
8886         (*aarch64_mul3_elt_<vswap_width_name><mode>): Likewise.
8887         (*aarch64_mul3_elt_to_64v2df, *aarch64_mla_elt<mode>): Likewise.
8888         (*aarch64_mla_elt_<vswap_width_name><mode>, *aarch64_mls_elt<mode>)
8889         (*aarch64_mls_elt_<vswap_width_name><mode>, *aarch64_fma4_elt<mode>)
8890         (*aarch64_fma4_elt_<vswap_width_name><mode>):: Likewise.
8891         (*aarch64_fma4_elt_to_64v2df, *aarch64_fnma4_elt<mode>): Likewise.
8892         (*aarch64_fnma4_elt_<vswap_width_name><mode>): Likewise.
8893         (*aarch64_fnma4_elt_to_64v2df, reduc_plus_scal_<mode>): Likewise.
8894         (reduc_plus_scal_v4sf, reduc_<maxmin_uns>_scal_<mode>): Likewise.
8895         (reduc_<maxmin_uns>_scal_<mode>): Likewise.
8896         (*aarch64_get_lane_extend<GPI:mode><VDQQH:mode>): Likewise.
8897         (*aarch64_get_lane_zero_extendsi<mode>): Likewise.
8898         (aarch64_get_lane<mode>, *aarch64_mulx_elt_<vswap_width_name><mode>)
8899         (*aarch64_mulx_elt<mode>, *aarch64_vgetfmulx<mode>): Likewise.
8900         (aarch64_sq<r>dmulh_lane<mode>, aarch64_sq<r>dmulh_laneq<mode>)
8901         (aarch64_sqrdml<SQRDMLH_AS:rdma_as>h_lane<mode>): Likewise.
8902         (aarch64_sqrdml<SQRDMLH_AS:rdma_as>h_laneq<mode>): Likewise.
8903         (aarch64_sqdml<SBINQOPS:as>l_lane<mode>): Likewise.
8904         (aarch64_sqdml<SBINQOPS:as>l_laneq<mode>): Likewise.
8905         (aarch64_sqdml<SBINQOPS:as>l2_lane<mode>_internal): Likewise.
8906         (aarch64_sqdml<SBINQOPS:as>l2_laneq<mode>_internal): Likewise.
8907         (aarch64_sqdmull_lane<mode>, aarch64_sqdmull_laneq<mode>): Likewise.
8908         (aarch64_sqdmull2_lane<mode>_internal): Likewise.
8909         (aarch64_sqdmull2_laneq<mode>_internal): Likewise.
8910         (aarch64_vec_load_lanesoi_lane<mode>): Likewise.
8911         (aarch64_vec_store_lanesoi_lane<mode>): Likewise.
8912         (aarch64_vec_load_lanesci_lane<mode>): Likewise.
8913         (aarch64_vec_store_lanesci_lane<mode>): Likewise.
8914         (aarch64_vec_load_lanesxi_lane<mode>): Likewise.
8915         (aarch64_vec_store_lanesxi_lane<mode>): Likewise.
8916         (aarch64_simd_vec_set<mode>): Update use of ENDIAN_LANE_N.
8917         (aarch64_simd_vec_setv2di): Likewise.
8919 2017-11-06  Carl Love  <cel@us.ibm.com>
8921         * config/rs6000/rs6000-c.c (P8V_BUILTIN_VEC_REVB): Add power 8
8922         definitions.
8923         (P9V_BUILTIN_VEC_REVB): Remove the power 9 instance definitions.
8924         * config/rs6000/altivec.h (vec_revb): Change the #define from power 9
8925         to power 8.
8926         * config/rs6000/r6000-protos.h (swap_endian_selector_for_mode): Add new
8927         extern declaration.
8928         * config/rs6000/rs6000.c (swap_endian_selector_for_mode): Add function.
8929         * config/rs6000/rs6000-builtin.def (BU_P8V_VSX_1, BU_P8V_OVERLOAD_1):
8930         Add power 8 macro expansions.
8931         (BU_P9V_OVERLOAD_1): Remove power 9 overload expansion.
8932         * config/rs6000/vsx.md (revb_<mode>): Add define_expand to generate
8933         power 8 instructions.  (VSX_XXBR): Add iterator.
8935 2017-11-06  Wilco Dijkstra  <wdijkstr@arm.com>
8937         * config/arm/arm.md (predicable_short_it): Change default to "no",
8938         improve documentation, remove uses that are identical to the default.
8939         (enabled_for_depr_it): Rename to enabled_for_short_it.
8940         * gcc/config/arm/arm-fixed.md (predicable_short_it):
8941         Remove default uses.
8942         * gcc/config/arm/ldmstm.md (predicable_short_it): Likewise.
8943         * gcc/config/arm/sync.md (predicable_short_it): Likewise.
8944         * gcc/config/arm/thumb2.md (predicable_short_it): Likewise.
8945         * gcc/config/arm/vfp.md (predicable_short_it): Likewise.
8947 2017-11-06  Michael Meissner  <meissner@linux.vnet.ibm.com>
8949         PR target/82748
8950         * config/rs6000/rs6000-builtin.def (BU_FLOAT128_1): Delete
8951         float128 helper macros, which are no longer used after deleting
8952         the old 'q' built-in functions, and moving the round to odd
8953         built-in functions to being special built-in functions.
8954         (BU_FLOAT128_2): Likewise.
8955         (BU_FLOAT128_1_HW): Likewise.
8956         (BU_FLOAT128_2_HW): Likewise.
8957         (BU_FLOAT128_3_HW): Likewise.
8958         (FABSQ): Delete old 'q' built-in functions.
8959         (COPYSIGNQ): Likewise.
8960         (SQRTF128_ODD): Move round to odd built-in functions to be
8961         special built-in functions, so that we can handle
8962         -mabi=ieeelongdouble.
8963         (TRUNCF128_ODD): Likewise.
8964         (ADDF128_ODD): Likewise.
8965         (SUBF128_ODD): Likewise.
8966         (MULF128_ODD): Likewise.
8967         (DIVF128_ODD): Likewise.
8968         (FMAF128_ODD): Likewise.
8969         * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Map old 'q'
8970         built-in names to 'f128'.
8971         * config/rs6000/rs6000.c (rs6000_fold_builtin): Remove folding the
8972         old 'q' built-in functions, as the machine independent code for
8973         'f128' built-in functions handles this.
8974         (rs6000_expand_builtin): Add expansion for float128 round to odd
8975         functions, keying off on -mabi=ieeelongdouble of whether to use
8976         the KFmode or TFmode variant.
8977         (rs6000_init_builtins): Initialize the _Float128 round to odd
8978         built-in functions.
8979         * doc/extend.texi (PowerPC Built-in Functions): Document the old
8980         _Float128 'q' built-in functions are now mapped into the new
8981         'f128' built-in functions.
8983 2017-11-06  David Edelsohn  <dje.gcc@gmail.com>
8985         * collect2.c (add_lto_object): Compile for OBJECT_COFF.
8986         (scan_prog_file): Don't skip PASS_LTOINFO. Scan for LTO objects.
8988 2017-11-06  David Malcolm  <dmalcolm@redhat.com>
8990         PR jit/82826
8991         * ipa-fnsummary.c (ipa_fnsummary_c_finalize): New function.
8992         * ipa-fnsummary.h (ipa_fnsummary_c_finalize): New decl.
8993         * toplev.c: Include "ipa-fnsummary.h".
8994         (toplev::finalize): Call ipa_fnsummary_c_finalize.
8996 2017-11-06  Jakub Jelinek  <jakub@redhat.com>
8998         PR tree-optimization/82838
8999         * gimple-ssa-store-merging.c
9000         (imm_store_chain_info::output_merged_store): Call force_gimple_operand_1
9001         on a separate gimple_seq which is then appended to seq.
9003 2017-11-06  Jeff Law  <law@redhat.com>
9005         PR target/82788
9006         * config/i386/i386.c (PROBE_INTERVAL): Remove.
9007         (get_probe_interval): New functions.
9008         (ix86_adjust_stack_and_probe_stack_clash): Use get_probe_interval.
9009         (ix86_adjust_stack_and_probe): Likewise.
9010         (output_adjust_stack_and_probe): Likewise.
9011         (ix86_emit_probe_stack_range): Likewise.
9012         (ix86_expand_prologue): Likewise.
9014 2017-11-06  Richard Sandiford  <richard.sandiford@linaro.org>
9016         PR tree-optimization/82816
9017         * tree-ssa-math-opts.c (convert_mult_to_widen): Return false
9018         if the modes of the two types are the same.
9019         (convert_plusminus_to_widen): Likewise.
9021 2017-11-06  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
9023         * config/rs6000/altivec.md (*p9_vadu<mode>3) Rename to
9024         p9_vadu<mode>3.
9025         (usadv16qi): New define_expand.
9026         (usadv8hi): New define_expand.
9028 2017-11-06  Jan Hubicka  <hubicka@ucw.cz>
9030         PR bootstrap/82832
9031         * ipa-inline-transform.c (update_noncloned_frequencies): Always
9032         scale.
9033         (inline_transform): Likewise.
9034         * predict.c (counts_to_freqs): Remove useless conditional.
9035         * profile-count.h (profile_count::apply_scale): Move sanity check.
9036         * tree-inline.c (copy_bb): Always scale.
9037         (copy_cfg_body): Likewise.
9039 2017-11-06  Christophe Lyon  <christophe.lyon@linaro.org>
9041         PR target/67591
9042         * config/arm/arm.md (*sub_shiftsi): Add predicable_short_it
9043         attribute.
9044         (*cmp_ite0): Add enabled_for_depr_it attribute.
9045         (*cmp_ite1): Likewise.
9047 2017-11-06  Segher Boessenkool  <segher@kernel.crashing.org>
9049         * config/rs6000/rs6000.c (rs6000_insn_cost): Handle TYPE_MFCR and
9050         TYPE_MFCRF.
9052 2017-11-06  Richard Sandiford  <richard.sandiford@linaro.org>
9054         * tree-vrp.c (vrp_int_const_binop): Return true on success and
9055         return the value by pointer.
9056         (extract_range_from_multiplicative_op_1): Update accordingly.
9057         Return as soon as an operation fails.
9059 2017-11-05  Tom de Vries  <tom@codesourcery.com>
9061         PR other/82784
9062         * asan.c (DEF_SANITIZER_BUILTIN_1): Factor out of ...
9063         (DEF_SANITIZER_BUILTIN): ... here.
9064         (initialize_sanitizer_builtins): Use DEF_SANITIZER_BUILTIN_1 instead of
9065         DEF_SANITIZER_BUILTIN in if stmt.  Add missing semicolon.
9067 2017-11-05  Tom de Vries  <tom@codesourcery.com>
9069         PR other/82784
9070         * config/elfos.h (ASM_OUTPUT_BEFORE_CASE_LABEL): Remove semicolon after
9071         macro body.
9072         (ASM_OUTPUT_CASE_LABEL): Add semicolon after
9073         ASM_OUTPUT_BEFORE_CASE_LABEL call.
9074         * config/arc/arc.h (ASM_OUTPUT_BEFORE_CASE_LABEL): Remove semicolon
9075         after macro body.
9076         * config/m68k/m68kelf.h (ASM_OUTPUT_BEFORE_CASE_LABEL): Same.
9077         * config/mips/mips.h (ASM_OUTPUT_BEFORE_CASE_LABEL): Same.
9078         * config/v850/v850.h (ASM_OUTPUT_BEFORE_CASE_LABEL): Same.
9080 2017-11-05  Tom de Vries  <tom@codesourcery.com>
9082         PR other/82784
9083         * graphite-scop-detection.c (DEBUG_PRINT): Remove semicolon after
9084         "do {} while (0)".
9086 2017-11-04  Michael Clark  <michaeljclark@mac.com>
9088         * config/riscv/riscv.c (riscv_print_operand): Add a 'i' format.
9089         config/riscv/riscv.md (addsi3): Use 'i' for immediates.
9090         (adddi3): Likewise.
9091         (*addsi3_extended): Likewise.
9092         (*addsi3_extended2): Likewise.
9093         (<optab>si3): Likewise.
9094         (<optab>di3): Likewise.
9095         (<optab><mode>3): Likewise.
9096         (<*optabe>si3_internal): Likewise.
9097         (zero_extendqi<SUPERQI:mode>2): Likewise.
9098         (*add<mode>hi3): Likewise.
9099         (*xor<mode>hi3): Likewise.
9100         (<optab>di3): Likewise.
9101         (*<optab>si3_extend): Likewise.
9102         (*sge<u>_<X:mode><GPR:mode>): Likewise.
9103         (*slt<u>_<X:mode><GPR:mode>): Likewise.
9104         (*sle<u>_<X:mode><GPR:mode>): Likewise.
9106 2017-11-04  Andrew Waterman  <andrew@sifive.com>
9108         * config/riscv/riscv.c (riscv_option_override): Conditionally set
9109         TARGET_STRICT_ALIGN based upon -mtune argument.
9111 2017-11-04  Andrew Waterman  <andrew@sifive.com>
9113         * config/riscv/riscv.h (SLOW_BYTE_ACCESS): Change to 1.
9115 2017-11-04  Daniel Santos  <daniel.santos@pobox.com>
9117         * config/i386/i386.c (choose_basereg): Use optional scratch
9118         register and add assertion.
9119         (x86_emit_outlined_ms2sysv_save): Use scratch register when
9120         needed, and don't allocate stack.
9121         (ix86_expand_prologue): Rearrange where SSE saves/stub call is
9122         emitted, correct wrong allocation with -mcall-ms2sysv-xlogues.
9123         (ix86_emit_outlined_ms2sysv_restore): Fix non-immediate offsets.
9125 2017-11-03  Jeff Law  <law@redhat.com>
9127         * config/i386/i386.c (ix86_emit_restore_reg_using_pop): Prototype.
9128         (ix86_adjust_stack_and_probe_stack_clash): Use a push/pop sequence
9129         to probe at the start of a noreturn function.
9131 2017-11-03  Jakub Jelinek  <jakub@redhat.com>
9133         PR tree-optimization/78821
9134         * gimple-ssa-store-merging.c: Update the file comment.
9135         (MAX_STORE_ALIAS_CHECKS): Define.
9136         (struct store_operand_info): New type.
9137         (store_operand_info::store_operand_info): New constructor.
9138         (struct store_immediate_info): Add rhs_code and ops data members.
9139         (store_immediate_info::store_immediate_info): Add rhscode, op0r
9140         and op1r arguments to the ctor, initialize corresponding data members.
9141         (struct merged_store_group): Add load_align_base and load_align
9142         data members.
9143         (merged_store_group::merged_store_group): Initialize them.
9144         (merged_store_group::do_merge): Update them.
9145         (merged_store_group::apply_stores): Pick the constant for
9146         encode_tree_to_bitpos from one of the two operands, or skip
9147         encode_tree_to_bitpos if neither operand is a constant.
9148         (class pass_store_merging): Add process_store method decl.  Remove
9149         bool argument from terminate_all_aliasing_chains method decl.
9150         (pass_store_merging::terminate_all_aliasing_chains): Remove
9151         var_offset_p argument and corresponding handling.
9152         (stmts_may_clobber_ref_p): New function.
9153         (compatible_load_p): New function.
9154         (imm_store_chain_info::coalesce_immediate_stores): Terminate group
9155         if there is overlap and rhs_code is not INTEGER_CST.  For
9156         non-overlapping stores terminate group if rhs is not mergeable.
9157         (get_alias_type_for_stmts): Change first argument from
9158         auto_vec<gimple *> & to vec<gimple *> &.  Add IS_LOAD, CLIQUEP and
9159         BASEP arguments.  If IS_LOAD is true, look at rhs1 of the stmts
9160         instead of lhs.  Compute *CLIQUEP and *BASEP in addition to the
9161         alias type.
9162         (get_location_for_stmts): Change first argument from
9163         auto_vec<gimple *> & to vec<gimple *> &.
9164         (struct split_store): Remove orig_stmts data member, add orig_stores.
9165         (split_store::split_store): Create orig_stores rather than orig_stmts.
9166         (find_constituent_stmts): Renamed to ...
9167         (find_constituent_stores): ... this.  Change second argument from
9168         vec<gimple *> * to vec<store_immediate_info *> *, push pointers
9169         to info structures rather than the statements.
9170         (split_group): Rename ALLOW_UNALIGNED argument to
9171         ALLOW_UNALIGNED_STORE, add ALLOW_UNALIGNED_LOAD argument and handle
9172         it.  Adjust find_constituent_stores caller.
9173         (imm_store_chain_info::output_merged_store): Handle rhs_code other
9174         than INTEGER_CST, adjust split_group, get_alias_type_for_stmts and
9175         get_location_for_stmts callers.  Set MR_DEPENDENCE_CLIQUE and
9176         MR_DEPENDENCE_BASE on the MEM_REFs if they are the same in all stores.
9177         (mem_valid_for_store_merging): New function.
9178         (handled_load): New function.
9179         (pass_store_merging::process_store): New method.
9180         (pass_store_merging::execute): Use process_store method.  Adjust
9181         terminate_all_aliasing_chains caller.
9183 2017-11-03  Wilco Dijkstra  <wdijkstr@arm.com>
9185         * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
9186         Return true for more constants, symbols and label references.
9187         (aarch64_valid_floating_const): Remove unused function.
9189 2017-11-03  Wilco Dijkstra  <wdijkstr@arm.com>
9191         PR target/82786
9192         * config/aarch64/aarch64.c (aarch64_layout_frame):
9193         Undo forcing of LR at bottom of frame.
9195 2017-11-03  Jeff Law  <law@redhat.com>
9197         PR target/82823
9198         * config/i386/i386.c (ix86_expand_prologue): Tighten assert
9199         for int_registers_saved.
9201         * cfganal.c (single_pred_edge_ignoring_loop_edges): New function
9202         extracted from tree-ssa-dom.c.
9203         * cfganal.h (single_pred_edge_ignoring_loop_edges): Prototype.
9204         * tree-ssa-dom.c (single_incoming_edge_ignoring_loop_edges): Remove.
9205         (record_equivalences_from_incoming_edge): Add additional argument
9206         to single_pred_edge_ignoring_loop_edges call.
9207         * tree-ssa-uncprop.c (single_incoming_edge_ignoring_loop_edges): Remove.
9208         (uncprop_dom_walker::before_dom_children): Add additional argument
9209         to single_pred_edge_ignoring_loop_edges call.
9210         * tree-ssa-sccvn.c (sccvn_dom_walker::before_dom_children): Use
9211         single_pred_edge_ignoring_loop_edges rather than open coding.
9212         * tree-vrp.c (evrp_dom_walker::before_dom_children): Similarly.
9214 2017-11-03  Marc Glisse  <marc.glisse@inria.fr>
9216         * match.pd (-(-A)): Rewrite.
9218 2017-11-03  Segher Boessenkool  <segher@kernel.crashing.org>
9220         * config/rs6000/rs60000-protos.h (rs6000_emit_sISEL): Delete.
9221         (rs6000_emit_int_cmove): New declaration.
9222         * config/rs6000/rs6000.c (rs6000_emit_int_cmove): Delete declaration.
9223         (rs6000_emit_sISEL): Delete.
9224         (rs6000_emit_int_cmove): Make non-static.
9225         * config/rs6000/rs6000.md (cstore<mode>4): Use rs6000_emit_int_cmove
9226         instead of rs6000_emit_sISEL.
9228 2017-11-03  Jan Hubicka  <hubicka@ucw.cz>
9230         * asan.c (create_cond_insert_point): Maintain profile.
9231         * ipa-utils.c (ipa_merge_profiles): Be sure only IPA profiles are
9232         merged.
9233         * basic-block.h (struct basic_block_def): Remove frequency.
9234         (EDGE_FREQUENCY): Use to_frequency
9235         * bb-reorder.c (push_to_next_round_p): Use only IPA counts for global
9236         heuristics.
9237         (find_traces): Update to use to_frequency.
9238         (find_traces_1_round): Likewise; use only IPA counts.
9239         (bb_to_key): Likewise.
9240         (connect_traces): Use IPA counts only.
9241         (copy_bb_p): Update to use to_frequency.
9242         (fix_up_crossing_landing_pad): Likewise.
9243         (sanitize_hot_paths): Likewise.
9244         * bt-load.c (basic_block_freq): Likewise.
9245         * cfg.c (init_flow): Set count_max to uninitialized.
9246         (check_bb_profile): Remove frequencies; check counts.
9247         (dump_bb_info): Do not dump frequencies.
9248         (update_bb_profile_for_threading): Update counts only.
9249         (scale_bbs_frequencies_int): Likewise.
9250         (MAX_SAFE_MULTIPLIER): Remove.
9251         (scale_bbs_frequencies_gcov_type): Update counts only.
9252         (scale_bbs_frequencies_profile_count): Update counts only.
9253         (scale_bbs_frequencies): Update counts only.
9254         * cfg.h (struct control_flow_graph): Add count-max.
9255         (update_bb_profile_for_threading): Update prototype.
9256         * cfgbuild.c (find_bb_boundaries): Do not update frequencies.
9257         (find_many_sub_basic_blocks): Likewise.
9258         * cfgcleanup.c (try_forward_edges): Likewise.
9259         (try_crossjump_to_edge): Likewise.
9260         * cfgexpand.c (expand_gimple_cond): Likewise.
9261         (expand_gimple_tailcall): Likewise.
9262         (construct_init_block): Likewise.
9263         (construct_exit_block): Likewise.
9264         * cfghooks.c (verify_flow_info): Check consistency of counts.
9265         (dump_bb_for_graph): Do not dump frequencies.
9266         (split_block_1): Do not update frequencies.
9267         (split_edge): Do not update frequencies.
9268         (make_forwarder_block): Do not update frequencies.
9269         (duplicate_block): Do not update frequencies.
9270         (account_profile_record): Do not update frequencies.
9271         * cfgloop.c (find_subloop_latch_edge_by_profile): Use IPA counts
9272         for global heuristics.
9273         * cfgloopanal.c (average_num_loop_insns): Update to use to_frequency.
9274         (expected_loop_iterations_unbounded): Use counts only.
9275         * cfgloopmanip.c (scale_loop_profile): Simplify.
9276         (create_empty_loop_on_edge): Simplify
9277         (loopify): Simplify
9278         (duplicate_loop_to_header_edge): Simplify
9279         * cfgrtl.c (force_nonfallthru_and_redirect): Update profile.
9280         (update_br_prob_note): Take care of removing note when profile
9281         becomes undefined.
9282         (relink_block_chain): Do not dump frequency.
9283         (rtl_account_profile_record): Use to_frequency.
9284         * cgraph.c (symbol_table::create_edge): Convert count to ipa count.
9285         (cgraph_edge::redirect_call_stmt_to_calle): Conver tcount to ipa count.
9286         (cgraph_update_edges_for_call_stmt_node): Likewise.
9287         (cgraph_edge::verify_count_and_frequency): Update.
9288         (cgraph_node::verify_node): Temporarily disable frequency verification.
9289         * cgraphbuild.c (compute_call_stmt_bb_frequency): Use
9290         to_cgraph_frequency.
9291         (cgraph_edge::rebuild_edges): Convert to ipa counts.
9292         * cgraphunit.c (init_lowered_empty_function): Do not initialize
9293         frequencies.
9294         (cgraph_node::expand_thunk): Update profile.
9295         * except.c (dw2_build_landing_pads): Do not update frequency.
9296         * final.c (compute_alignments): Use to_frequency.
9297         (dump_basic_block_info): Do not dump frequency.
9298         * gimple-pretty-print.c (dump_profile): Do not dump frequency.
9299         (dump_gimple_bb_header): Do not dump frequency.
9300         * gimple-ssa-isolate-paths.c (isolate_path): Do not update frequency;
9301         do update count.
9302         * gimple-streamer-in.c (input_bb): Do not stream frequency.
9303         * gimple-streamer-out.c (output_bb): Do not stream frequency.
9304         * haifa-sched.c (sched_pressure_start_bb): Use to_freuqency.
9305         (init_before_recovery): Do not update frequency.
9306         (sched_create_recovery_edges): Do not update frequency.
9307         * hsa-gen.c (convert_switch_statements): Do not update frequency.
9308         * ipa-cp.c (ipcp_propagate_stage): Update search for max_count.
9309         (ipa_cp_c_finalize): Set max_count to uninitialized.
9310         * ipa-fnsummary.c (get_minimal_bb): Use counts.
9311         (param_change_prob): Use counts.
9312         * ipa-profile.c (ipa_profile_generate_summary): Do not summarize
9313         local profiles.
9314         * ipa-split.c (consider_split): Use to_frequency.
9315         (split_function): Use to_frequency.
9316         * ira-build.c (loop_compare_func): Likewise.
9317         (mark_loops_for_removal): Likewise.
9318         (mark_all_loops_for_removal): Likewise.
9319         * loop-doloop.c (doloop_modify): Do not update frequency.
9320         * loop-unroll.c (unroll_loop_runtime_iterations): Do not update
9321         frequency.
9322         * lto-streamer-in.c (input_function): Update count_max.
9323         * omp-expand.c (expand_omp_taskreg): Update count_max.
9324         * omp-simd-clone.c (simd_clone_adjust): Update profile.
9325         * predict.c (maybe_hot_frequency_p): Use to_frequency.
9326         (maybe_hot_count_p): Use ipa counts only.
9327         (maybe_hot_bb_p): Simplify.
9328         (maybe_hot_edge_p): Simplify.
9329         (probably_never_executed): Do not take frequency argument.
9330         (probably_never_executed_bb_p): Do not pass frequency.
9331         (probably_never_executed_edge_p): Likewise.
9332         (combine_predictions_for_bb): Check that profile is nonzero.
9333         (propagate_freq): Do not set frequency.
9334         (drop_profile): Simplify.
9335         (counts_to_freqs): Simplify.
9336         (expensive_function_p): Use to_frequency.
9337         (propagate_unlikely_bbs_forward): Simplify.
9338         (determine_unlikely_bbs): Simplify.
9339         (estimate_bb_frequencies): Add hack to silence graphite issues.
9340         (compute_function_frequency): Use ipa counts.
9341         (pass_profile::execute): Update.
9342         (rebuild_frequencies): Use counts only.
9343         (force_edge_cold): Use counts only.
9344         * profile-count.c (profile_count::dump): Dump new count types.
9345         (profile_count::differs_from_p): Check compatiblity.
9346         (profile_count::to_frequency): New function.
9347         (profile_count::to_cgraph_frequency): New function.
9348         * profile-count.h (struct function): Declare.
9349         (enum profile_quality): Add profile_guessed_local and
9350         profile_guessed_global0.
9351         (class profile_proability): Decrease number of bits to 29;
9352         update from_reg_br_prob_note and to_reg_br_prob_note.
9353         (class profile_count: Update comment; decrease number of bits
9354         to 61. Check compatibility.
9355         (profile_count::compatible_p): New private member function.
9356         (profile_count::ipa_p): New member function.
9357         (profile_count::operator<): Handle global zero correctly.
9358         (profile_count::operator>): Handle global zero correctly.
9359         (profile_count::operator<=): Handle global zero correctly.
9360         (profile_count::operator>=): Handle global zero correctly.
9361         (profile_count::nonzero_p): New member function.
9362         (profile_count::force_nonzero): New member function.
9363         (profile_count::max): New member function.
9364         (profile_count::apply_scale): Handle IPA scalling.
9365         (profile_count::guessed_local): New member function.
9366         (profile_count::global0): New member function.
9367         (profile_count::ipa): New member function.
9368         (profile_count::to_frequency): Declare.
9369         (profile_count::to_cgraph_frequency): Declare.
9370         * profile.c (OVERLAP_BASE): Delete.
9371         (compute_frequency_overlap): Delete.
9372         (compute_branch_probabilities): Do not use compute_frequency_overlap.
9373         * regs.h (REG_FREQ_FROM_BB): Use to_frequency.
9374         * sched-ebb.c (rank): Use counts only.
9375         * shrink-wrap.c (handle_simple_exit): Use counts only.
9376         (try_shrink_wrapping): Use counts only.
9377         (place_prologue_for_one_component): Use counts only.
9378         * tracer.c (find_best_predecessor): Use to_frequency.
9379         (find_trace): Use to_frequency.
9380         (tail_duplicate): Use to_frequency.
9381         * trans-mem.c (expand_transaction): Do not update frequency.
9382         * tree-call-cdce.c: Do not update frequency.
9383         * tree-cfg.c (gimple_find_sub_bbs): Likewise.
9384         (gimple_merge_blocks): Likewise.
9385         (gimple_split_edge): Likewise.
9386         (gimple_duplicate_sese_region): Likewise.
9387         (gimple_duplicate_sese_tail): Likewise.
9388         (move_sese_region_to_fn): Likewise.
9389         (gimple_account_profile_record): Likewise.
9390         (insert_cond_bb): Likewise.
9391         * tree-complex.c (expand_complex_div_wide): Likewise.
9392         * tree-eh.c (lower_resx): Update profile.
9393         * tree-inline.c (copy_bb): Simplify count scaling; do not scale
9394         frequencies.
9395         (initialize_cfun): Do not initialize frequencies
9396         (freqs_to_counts): Delete.
9397         (copy_cfg_body): Ignore count parameter.
9398         (copy_body): Update.
9399         (expand_call_inline): Update count_max.
9400         (optimize_inline_calls): Update count_max.
9401         (tree_function_versioning): Update count_max.
9402         * tree-ssa-coalesce.c (coalesce_cost_bb): Use to_frequency.
9403         * tree-ssa-ifcombine.c (update_profile_after_ifcombine): Do not update
9404         frequency.
9405         * tree-ssa-loop-im.c (execute_sm_if_changed): Use counts only.
9406         * tree-ssa-loop-ivcanon.c (unloop_loops): Do not update freuqency.
9407         (try_peel_loop): Likewise.
9408         * tree-ssa-loop-ivopts.c (get_scaled_computation_cost_at): Use
9409         to_frequency.
9410         * tree-ssa-loop-manip.c (niter_for_unrolled_loop): Pass -1.
9411         (tree_transform_and_unroll_loop): Do not use frequencies
9412         * tree-ssa-loop-niter.c (estimate_numbers_of_iterations):
9413         Use reliable prediction only.
9414         * tree-ssa-loop-unswitch.c (hoist_guard): Do not use frequencies.
9415         * tree-ssa-sink.c (select_best_block): Use to_frequency.
9416         * tree-ssa-tail-merge.c (replace_block_by): Temporarily disable
9417         probability scaling.
9418         * tree-ssa-threadupdate.c (create_block_for_threading): Do
9419         not update frequency
9420         (any_remaining_duplicated_blocks): Likewise.
9421         (update_profile): Likewise.
9422         (estimated_freqs_path): Delete.
9423         (freqs_to_counts_path): Delete.
9424         (clear_counts_path): Delete.
9425         (ssa_fix_duplicate_block_edges): Likewise.
9426         (duplicate_thread_path): Likewise.
9427         * tree-switch-conversion.c (gen_inbound_check): Use counts.
9428         * tree-tailcall.c (decrease_profile): Do not update frequency.
9429         (eliminate_tail_call): Likewise.
9430         * tree-vect-loop-manip.c (vect_do_peeling): Likewise.
9431         * tree-vect-loop.c (scale_profile_for_vect_loop): Likewise.
9432         (optimize_mask_stores): Likewise.
9433         * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
9434         * ubsan.c (ubsan_expand_null_ifn): Update profile.
9435         (ubsan_expand_ptr_ifn): Update profile.
9436         * value-prof.c (gimple_ic): Simplify.
9437         * value-prof.h (gimple_ic): Update prototype.
9438         * ipa-inline-transform.c (inline_transform): Fix scaling conditoins.
9439         * ipa-inline.c (compute_uninlined_call_time): Be sure that
9440         counts are nonzero.
9441         (want_inline_self_recursive_call_p): Likewise.
9442         (resolve_noninline_speculation): Only cummulate defined counts.
9443         (inline_small_functions): Use nonzero_p.
9444         (ipa_inline): Do not access freed node.
9446 2017-11-03  Wilco Dijkstra  <wdijkstr@arm.com>
9448         * config/aarch64/aarch64.c (aarch64_override_options_internal):
9449         Set PARAM_SCHED_PRESSURE_ALGORITHM to SCHED_PRESSURE_MODEL.
9451 2017-11-03  Kito Cheng  <kito.cheng@gmail.com>
9453         * config/riscv/riscv.c (riscv_legitimize_move): Handle
9454         non-legitimate address.
9456 2017-11-03  Segher Boessenkool  <segher@kernel.crashing.org>
9458         * config/rs6000/rs6000.md (*lt0_disi): Delete.
9459         (*lt0_<mode>di, *lt0_<mode>si): New.
9461 2017-11-03  Segher Boessenkool  <segher@kernel.crashing.org>
9463         * config/rs6000/rs6000.md (move_from_CR_ov_bit): Change condition to
9464         TARGET_PAIRED_FLOAT.
9466 2017-11-03  Siddhesh Poyarekar  <siddhesh.poyarekar@linaro.org>
9467             Jim Wilson  <jim.wilson@linaro.org>
9469         * config/aarch64/aarch64-cores.def (saphira): New CPU.
9470         * config/aarch64/aarch64-tune.md: Regenerated.
9471         * doc/invoke.texi (AArch64 Options/-mtune): Add "saphira".
9472         * gcc/config/aarch64/aarch64.c (saphira_tunings): New tuning table.
9474 2017-11-03  Cupertino Miranda  <cmiranda@synopsys.com>
9476         * config/arc/arc.c (arc_save_restore): Corrected CFA note.
9477         (arc_expand_prologue): Restore blink for millicode.
9478         * config/arc/linux.h (LINK_EH_SPEC): Defined.
9480 2017-11-03  Richard Sandiford  <richard.sandiford@linaro.org>
9482         PR target/82809
9483         * config/i386/i386.c (ix86_vector_duplicate_value): Use
9484         gen_vec_duplicate after forcing the scalar into a register.
9486 2017-11-02  Segher Boessenkool  <segher@kernel.crashing.org>
9488         * combine (try_combine): Print the insns input to try_combine to the
9489         dump file.
9491 2017-11-02  Steve Ellcey  <sellcey@cavium.com>
9493         PR target/79868
9494         * config/aarch64/aarch64-c.c (aarch64_pragma_target_parse):
9495         Remove second argument from aarch64_process_target_attr call.
9496         * config/aarch64/aarch64-protos.h (aarch64_process_target_attr):
9497         Ditto.
9498         * config/aarch64/aarch64.c (aarch64_attribute_info): Change
9499         field type.
9500         (aarch64_handle_attr_arch): Remove second argument.
9501         (aarch64_handle_attr_cpu): Ditto.
9502         (aarch64_handle_attr_tune): Ditto.
9503         (aarch64_handle_attr_isa_flags): Ditto.
9504         (aarch64_process_one_target_attr): Ditto.
9505         (aarch64_process_target_attr): Ditto.
9506         (aarch64_option_valid_attribute_p): Remove second argument.
9507         on aarch64_process_target_attr call.
9509 2017-11-02  David Malcolm  <dmalcolm@redhat.com>
9511         * diagnostic.c: Include "selftest-diagnostic.h".
9512         (selftest::assert_location_text): New function.
9513         (selftest::test_diagnostic_get_location_text): New function.
9514         (selftest::diagnostic_c_tests): Call it.
9516 2017-11-02  David Malcolm  <dmalcolm@redhat.com>
9518         * Makefile.in (OBJS-libcommon): Add selftest-diagnostic.o.
9519         * diagnostic-show-locus.c: Include "selftest-diagnostic.h".
9520         (class selftest::test_diagnostic_context): Move to...
9521         * selftest-diagnostic.c: New file.
9522         * selftest-diagnostic.h: New file.
9524 2017-11-02  James Bowman  <james.bowman@ftdichip.com>
9526         * config/ft32/ft32.c (ft32_addr_space_legitimate_address_p): increase
9527         offset range for FT32B.
9528         * config/ft32/ft32.h: option "mcompress" enables relaxation.
9529         * config/ft32/ft32.md: Add TARGET_NOPM.
9530         * config/ft32/ft32.opt: Add mft32b, mcompress, mnopm.
9531         * gcc/doc/invoke.texi: Add mft32b, mcompress, mnopm.
9533 2017-11-02  Wilco Dijkstra  <wdijkstr@arm.com>
9535         * config/aarch64/aarch64.h (MALLOC_ABI_ALIGNMENT): New define.
9537 2017-11-02  Jeff Law  <law@redhat.com>
9539         * gimple-ssa-sprintf.c (sprintf_dom_walker): Remove
9540         virtual keyword on FINAL OVERRIDE members.
9542         * tree-ssa-propagate.h (ssa_propagation_engine): Group
9543         virtuals together.  Add virtual destructor.
9544         (substitute_and_fold_engine): Similarly.
9546 2017-11-02  Jan Hubicka  <hubicka@ucw.cz>
9548         * x86-tune.def (X86_TUNE_USE_INCDEC): Enable for Haswell+.
9550 2017-11-02  Richard Biener  <rguenther@suse.de>
9552         PR tree-optimization/82795
9553         * tree-if-conv.c (predicate_mem_writes): Remove bogus assert.
9555 2017-11-02  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
9557         * acinclude.m4 (gcc_AC_INITFINI_ARRAY): Don't require
9558         gcc_SUN_LD_VERSION.
9559         (gcc_GAS_CHECK_FEATURE): Remove.
9560         * configure.ac (ld_vers) <*-*-solaris2*>: Move comments from
9561         gcc_AC_INITFINI_ARRAY here.  Update for Solaris 11.4 changes.
9562         * configure: Regenerate.
9564 2017-11-02  Claudiu Zissulescu <claziss@synopsys.com>
9566         * config/arc/arc.c (hwloop_optimize): Account for empty
9567         body loops.
9569 2017-11-02  Richard Biener  <rguenther@suse.de>
9571         PR middle-end/82765
9572         * varasm.c (decode_addr_const): Make offset HOST_WIDE_INT.
9573         Truncate ARRAY_REF index and element size.
9575 2017-11-01  Palmer Dabbelt  <palmer@dabbelt.com>
9577         * doc/invoke.texi (RISC-V Options): Use "@minus{}2 GB", not "-2 GB".
9579 2017-11-01  Jeff Law  <law@redhat.com>
9581         * tree-ssa-ccp.c (ccp_folder): New class derived from
9582         substitute_and_fold_engine.
9583         (ccp_folder::get_value): New member function.
9584         (ccp_folder::fold_stmt): Renamed from ccp_fold_stmt.
9585         (ccp_fold_stmt): Remove prototype.
9586         (ccp_finalize): Call substitute_and_fold from the ccp_class.
9587         * tree-ssa-copy.c (copy_folder): New class derived from
9588         substitute_and_fold_engine.
9589         (copy_folder::get_value): Renamed from get_value.
9590         (fini_copy_prop): Call substitute_and_fold from copy_folder class.
9591         * tree-vrp.c (vrp_folder): New class derived from
9592         substitute_and_fold_engine.
9593         (vrp_folder::fold_stmt): Renamed from vrp_fold_stmt.
9594         (vrp_folder::get_value): New member function.
9595         (vrp_finalize): Call substitute_and_fold from vrp_folder class.
9596         (evrp_dom_walker::before_dom_children): Similarly for replace_uses_in.
9597         * tree-ssa-propagate.h (substitute_and_fold_engine): New class to
9598         provide a class interface to folder/substitute routines.
9599         (ssa_prop_fold_stmt_fn): Remove typedef.
9600         (ssa_prop_get_value_fn): Likewise.
9601         (subsitute_and_fold): Remove prototype.
9602         (replace_uses_in): Likewise.
9603         * tree-ssa-propagate.c (substitute_and_fold_engine::replace_uses_in):
9604         Renamed from replace_uses_in.  Call the virtual member function
9605         (substitute_and_fold_engine::replace_phi_args_in): Similarly.
9606         (substitute_and_fold_dom_walker): Remove initialization of
9607         data member entries for calbacks.  Add substitute_and_fold_engine
9608         member and initialize it.
9609         (substitute_and_fold_dom_walker::before_dom_children0: Use the
9610         member functions for get_value, replace_phi_args_in c
9611         replace_uses_in, and fold_stmt calls.
9612         (substitute_and_fold_engine::substitute_and_fold): Renamed from
9613         substitute_and_fold.  Remove assert.   Update ctor call.
9615         * tree-ssa-propagate.h (ssa_prop_visit_stmt_fn): Remove typedef.
9616         (ssa_prop_visit_phi_fn): Likewise.
9617         (class ssa_propagation_engine): New class to provide an interface
9618         into ssa_propagate.
9619         * tree-ssa-propagate.c (ssa_prop_visit_stmt): Remove file scoped
9620         variable.
9621         (ssa_prop_visit_phi): Likewise.
9622         (ssa_propagation_engine::simulate_stmt): Moved into class.
9623         Call visit_phi/visit_stmt from the class rather than via
9624         file scoped static variables.
9625         (ssa_propagation_engine::simulate_block): Moved into class.
9626         (ssa_propagation_engine::process_ssa_edge_worklist): Similarly.
9627         (ssa_propagation_engine::ssa_propagate): Similarly.  No longer
9628         set file scoped statics for the visit_stmt/visit_phi callbacks.
9629         * tree-complex.c (complex_propagate): New class derived from
9630         ssa_propagation_engine.
9631         (complex_propagate::visit_stmt): Renamed from complex_visit_stmt.
9632         (complex_propagate::visit_phi): Renamed from complex_visit_phi.
9633         (tree_lower_complex): Call ssa_propagate via the complex_propagate
9634         class.
9635         * tree-ssa-ccp.c: (ccp_propagate): New class derived from
9636         ssa_propagation_engine.
9637         (ccp_propagate::visit_phi): Renamed from ccp_visit_phi_node.
9638         (ccp_propagate::visit_stmt): Renamed from ccp_visit_stmt.
9639         (do_ssa_ccp): Call ssa_propagate from the ccp_propagate class.
9640         * tree-ssa-copy.c (copy_prop): New class derived from
9641         ssa_propagation_engine.
9642         (copy_prop::visit_stmt): Renamed from copy_prop_visit_stmt.
9643         (copy_prop::visit_phi): Renamed from copy_prop_visit_phi_node.
9644         (execute_copy_prop): Call ssa_propagate from the copy_prop class.
9645         * tree-vrp.c (vrp_prop): New class derived from ssa_propagation_engine.
9646         (vrp_prop::visit_stmt): Renamed from vrp_visit_stmt.
9647         (vrp_prop::visit_phi): Renamed from vrp_visit_phi_node.
9648         (execute_vrp): Call ssa_propagate from the vrp_prop class.
9650 2017-11-01  Jakub Jelinek  <jakub@redhat.com>
9652         PR rtl-optimization/82778
9653         PR rtl-optimization/82597
9654         * compare-elim.c (struct comparison): Add in_a_setter field.
9655         (find_comparison_dom_walker::before_dom_children): Remove killed
9656         bitmap and df_simulate_find_defs call, instead walk the defs.
9657         Compute last_setter and initialize in_a_setter.  Merge definitions
9658         with first initialization for a few variables.
9659         (try_validate_parallel): Use insn_invalid_p instead of
9660         recog_memoized.  Return insn rather than just the pattern.
9661         (try_merge_compare): Fix up comment.  Don't uselessly test if
9662         in_a is a REG_P.  Use cmp->in_a_setter instead of walking UD
9663         chains.
9664         (execute_compare_elim_after_reload): Remove df_chain_add_problem
9665         call.
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_rtx_costs): Use
9672         aarch64_hard_regno_nregs to get the number of registers
9673         in a mode.
9675 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9676             Alan Hayward  <alan.hayward@arm.com>
9677             David Sherwood  <david.sherwood@arm.com>
9679         * config/aarch64/constraints.md (Upl): Rename to...
9680         (Uaa): ...this.
9681         * config/aarch64/aarch64.md
9682         (*zero_extend<SHORT:mode><GPI:mode>2_aarch64, *addsi3_aarch64_uxtw):
9683         Update accordingly.
9685 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9686             Alan Hayward  <alan.hayward@arm.com>
9687             David Sherwood  <david.sherwood@arm.com>
9689         * config/aarch64/aarch64.c (aarch64_add_constant_internal)
9690         (aarch64_add_constant, aarch64_add_sp, aarch64_sub_sp): Move
9691         earlier in file.
9693 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9694             Alan Hayward  <alan.hayward@arm.com>
9695             David Sherwood  <david.sherwood@arm.com>
9697         * config/aarch64/aarch64.c (aarch64_evpc_trn, aarch64_evpc_uzp)
9698         (aarch64_evpc_zip, aarch64_evpc_ext, aarch64_evpc_rev)
9699         (aarch64_evpc_dup): Generate rtl direcly, rather than using
9700         named expanders.
9701         (aarch64_expand_vec_perm_const_1): Explicitly check for permutes
9702         of a single element.
9703         * config/aarch64/iterators.md: Add a comment above the permute
9704         unspecs to say that they are generated directly by
9705         aarch64_expand_vec_perm_const.
9706         * config/aarch64/aarch64-simd.md: Likewise the permute instructions.
9708 2017-11-01  Nathan Sidwell  <nathan@acm.org>
9710         * tree-dump.c (dequeue_and_dump): Use HAS_DECL_ASSEMBLER_NAME_P.
9712 2017-11-01  Palmer Dabbelt  <palmer@dabbelt.com>
9714         * doc/invoke.texi (RISC-V Options): Explicitly name the medlow
9715         and medany code models, and describe what they do.
9717 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9719         Revert accidental duplicate:
9721         * combine.c (can_change_dest_mode): Reject changes in
9722         REGMODE_NATURAL_SIZE.
9724 2017-11-01  Segher Boessenkool  <segher@kernel.crashing.org>
9726         PR rtl-optimization/64682
9727         PR rtl-optimization/69567
9728         PR rtl-optimization/69737
9729         PR rtl-optimization/82683
9730         * combine.c (distribute_notes) <REG_DEAD>: If the new I2 sets the same
9731         register mentioned in the note, drop the note, unless it came from I3,
9732         in which case it should go to I3 again.
9734 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9736         * tree-ssa-dse.c (normalize_ref): Check whether the ranges overlap
9737         and return false if not.
9738         (clear_bytes_written_by, live_bytes_read): Update accordingly.
9740 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9742         * tree-ssa-alias.h (ranges_overlap_p): Return false if either
9743         range is known to be empty.
9745 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9746             Alan Hayward  <alan.hayward@arm.com>
9747             David Sherwood  <david.sherwood@arm.com>
9749         * simplify-rtx.c (simplify_const_unary_operation): Use GET_MODE_NUNITS
9750         and CONST_VECTOR_NUNITS instead of computing the number of units from
9751         the byte sizes of the vector and element.
9752         (simplify_binary_operation_1): Likewise.
9753         (simplify_const_binary_operation): Likewise.
9754         (simplify_ternary_operation): Likewise.
9756 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9757             Alan Hayward  <alan.hayward@arm.com>
9758             David Sherwood  <david.sherwood@arm.com>
9760         * var-tracking.c (INT_MEM_OFFSET): Replace with...
9761         (int_mem_offset): ...this new function.
9762         (var_mem_set, var_mem_delete_and_set, var_mem_delete)
9763         (find_mem_expr_in_1pdv, dataflow_set_preserve_mem_locs)
9764         (same_variable_part_p, use_type, add_stores, vt_get_decl_and_offset):
9765         Update accordingly.
9767 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9768             Alan Hayward  <alan.hayward@arm.com>
9769             David Sherwood  <david.sherwood@arm.com>
9771         * lower-subreg.c (interesting_mode_p): New function.
9772         (compute_costs, find_decomposable_subregs, decompose_register)
9773         (simplify_subreg_concatn, can_decompose_p, resolve_simple_move)
9774         (resolve_clobber, dump_choices): Use it.
9776 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9777             Alan Hayward  <alan.hayward@arm.com>
9778             David Sherwood  <david.sherwood@arm.com>
9780         * rtlhash.c (add_rtx): Use add_hwi for 'w' and add_int for 'i'.
9782 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9783             Alan Hayward  <alan.hayward@arm.com>
9784             David Sherwood  <david.sherwood@arm.com>
9786         * alias.c (find_base_value, find_base_term): Only process integer
9787         truncations.  Check the precision rather than the size.
9789 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9790             Alan Hayward  <alan.hayward@arm.com>
9791             David Sherwood  <david.sherwood@arm.com>
9793         * machmode.h (is_narrower_int_mode): New function
9794         * optabs.c (expand_float, expand_fix): Use it.
9795         * dwarf2out.c (rotate_loc_descriptor): Likewise.
9797 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9798             Alan Hayward  <alan.hayward@arm.com>
9799             David Sherwood  <david.sherwood@arm.com>
9801         * rtl.h (narrower_subreg_mode): New function.
9802         * ira-color.c (update_costs_from_allocno): Use it.
9804 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9805             Alan Hayward  <alan.hayward@arm.com>
9806             David Sherwood  <david.sherwood@arm.com>
9808         * optabs-query.h (convert_optab_p): New function, split out from...
9809         (convert_optab_handler): ...here.
9810         (widening_optab_handler): Delete.
9811         (find_widening_optab_handler): Remove permit_non_widening parameter.
9812         (find_widening_optab_handler_and_mode): Likewise.  Provide an
9813         override that operates on mode class wrappers.
9814         * optabs-query.c (widening_optab_handler): Delete.
9815         (find_widening_optab_handler_and_mode): Remove permit_non_widening
9816         parameter.  Assert that the two modes are the same class and that
9817         the "from" mode is narrower than the "to" mode.  Use
9818         convert_optab_handler instead of widening_optab_handler.
9819         * expmed.c (expmed_mult_highpart_optab): Use convert_optab_handler
9820         instead of widening_optab_handler.
9821         * expr.c (expand_expr_real_2): Update calls to
9822         find_widening_optab_handler.
9823         * optabs.c (expand_widen_pattern_expr): Likewise.
9824         (expand_binop_directly): Take the insn_code as a parameter.
9825         (expand_binop): Only call find_widening_optab_handler for
9826         conversion optabs; use optab_handler otherwise.  Update calls
9827         to find_widening_optab_handler and expand_binop_directly.
9828         Use convert_optab_handler instead of widening_optab_handler.
9829         * tree-ssa-math-opts.c (convert_mult_to_widen): Update calls to
9830         find_widening_optab_handler and use scalar_mode rather than
9831         machine_mode.
9832         (convert_plusminus_to_widen): Likewise.
9834 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9835             Alan Hayward  <alan.hayward@arm.com>
9836             David Sherwood  <david.sherwood@arm.com>
9838         * machmode.h (fixed_size_mode): New class.
9839         * rtl.h (get_pool_mode): Return fixed_size_mode.
9840         * gengtype.c (main): Add fixed_size_mode.
9841         * target.def (get_raw_result_mode): Return a fixed_size_mode.
9842         (get_raw_arg_mode): Likewise.
9843         * doc/tm.texi: Regenerate.
9844         * targhooks.h (default_get_reg_raw_mode): Return a fixed_size_mode.
9845         * targhooks.c (default_get_reg_raw_mode): Likewise.
9846         * config/ia64/ia64.c (ia64_get_reg_raw_mode): Likewise.
9847         * config/mips/mips.c (mips_get_reg_raw_mode): Likewise.
9848         * config/msp430/msp430.c (msp430_get_raw_arg_mode): Likewise.
9849         (msp430_get_raw_result_mode): Likewise.
9850         * config/avr/avr-protos.h (regmask): Use as_a <fixed_side_mode>
9851         * dbxout.c (dbxout_parms): Require fixed-size modes.
9852         * expr.c (copy_blkmode_from_reg, copy_blkmode_to_reg): Likewise.
9853         * gimple-ssa-store-merging.c (encode_tree_to_bitpos): Likewise.
9854         * omp-low.c (lower_oacc_reductions): Likewise.
9855         * simplify-rtx.c (simplify_immed_subreg): Take fixed_size_modes.
9856         (simplify_subreg): Update accordingly.
9857         * varasm.c (constant_descriptor_rtx::mode): Change to fixed_size_mode.
9858         (force_const_mem): Update accordingly.  Return NULL_RTX for modes
9859         that aren't fixed-size.
9860         (get_pool_mode): Return a fixed_size_mode.
9861         (output_constant_pool_2): Take a fixed_size_mode.
9863 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9864             Alan Hayward  <alan.hayward@arm.com>
9865             David Sherwood  <david.sherwood@arm.com>
9867         * doc/rtl.texi (vec_series): Document.
9868         (const): Say that the operand can be a vec_series.
9869         * rtl.def (VEC_SERIES): New rtx code.
9870         * rtl.h (const_vec_series_p_1): Declare.
9871         (const_vec_series_p): New function.
9872         * emit-rtl.h (gen_const_vec_series): Declare.
9873         (gen_vec_series): Likewise.
9874         * emit-rtl.c (const_vec_series_p_1, gen_const_vec_series)
9875         (gen_vec_series): Likewise.
9876         * optabs.c (expand_mult_highpart): Use gen_const_vec_series.
9877         * simplify-rtx.c (simplify_unary_operation): Handle negations
9878         of vector series.
9879         (simplify_binary_operation_series): New function.
9880         (simplify_binary_operation_1): Use it.  Handle VEC_SERIES.
9881         (test_vector_ops_series): New function.
9882         (test_vector_ops): Call it.
9883         * config/powerpcspe/altivec.md (altivec_lvsl): Use
9884         gen_const_vec_series.
9885         (altivec_lvsr): Likewise.
9886         * config/rs6000/altivec.md (altivec_lvsl, altivec_lvsr): Likewise.
9888 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9889             Alan Hayward  <alan.hayward@arm.com>
9890             David Sherwood  <david.sherwood@arm.com>
9892         * doc/rtl.texi (const): Update description of address constants.
9893         Say that vector constants are allowed too.
9894         * common.md (E, F): Use CONSTANT_P instead of checking for
9895         CONST_VECTOR.
9896         * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
9897         checking for CONST_VECTOR.
9898         * expmed.c (make_tree): Use build_vector_from_val for a CONST
9899         VEC_DUPLICATE.
9900         * expr.c (expand_expr_real_2): Check for vector modes instead
9901         of checking for CONST_VECTOR.
9902         * rtl.h (const_vec_p): New function.
9903         (const_vec_duplicate_p): Check for a CONST VEC_DUPLICATE.
9904         (unwrap_const_vec_duplicate): Handle them here too.
9906 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9907             David Malcolm  <dmalcolm@redhat.com>
9908             Alan Hayward  <alan.hayward@arm.com>
9909             David Sherwood  <david.sherwood@arm.com>
9911         * rtl.h (vec_duplicate_p): New function.
9912         * selftest-rtl.c (assert_rtx_eq_at): New function.
9913         * selftest-rtl.h (ASSERT_RTX_EQ): New macro.
9914         (assert_rtx_eq_at): Declare.
9915         * selftest.h (selftest::simplify_rtx_c_tests): Declare.
9916         * selftest-run-tests.c (selftest::run_tests): Call it.
9917         * simplify-rtx.c: Include selftest.h and selftest-rtl.h.
9918         (simplify_unary_operation_1): Recursively handle vector duplicates.
9919         (simplify_binary_operation_1): Likewise.  Handle VEC_SELECTs of
9920         vector duplicates.
9921         (simplify_subreg): Handle subregs of vector duplicates.
9922         (make_test_reg, test_vector_ops_duplicate, test_vector_ops)
9923         (selftest::simplify_rtx_c_tests): New functions.
9925 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9926             Alan Hayward  <alan.hayward@arm.com>
9927             David Sherwood  <david.sherwood@arm.com>
9929         * emit-rtl.h (gen_const_vec_duplicate): Declare.
9930         (gen_vec_duplicate): Likewise.
9931         * emit-rtl.c (gen_const_vec_duplicate_1): New function, split
9932         out from...
9933         (gen_const_vector): ...here.
9934         (gen_const_vec_duplicate, gen_vec_duplicate): New functions.
9935         (gen_rtx_CONST_VECTOR): Use gen_const_vec_duplicate for constants
9936         whose elements are all equal.
9937         * optabs.c (expand_vector_broadcast): Use gen_const_vec_duplicate.
9938         * simplify-rtx.c (simplify_const_unary_operation): Likewise.
9939         (simplify_relational_operation): Likewise.
9940         * config/aarch64/aarch64.c (aarch64_simd_gen_const_vector_dup):
9941         Likewise.
9942         (aarch64_simd_dup_constant): Use gen_vec_duplicate.
9943         (aarch64_expand_vector_init): Likewise.
9944         * config/arm/arm.c (neon_vdup_constant): Likewise.
9945         (neon_expand_vector_init): Likewise.
9946         (arm_expand_vec_perm): Use gen_const_vec_duplicate.
9947         (arm_block_set_unaligned_vect): Likewise.
9948         (arm_block_set_aligned_vect): Likewise.
9949         * config/arm/neon.md (neon_copysignf<mode>): Likewise.
9950         * config/i386/i386.c (ix86_expand_vec_perm): Likewise.
9951         (expand_vec_perm_even_odd_pack): Likewise.
9952         (ix86_vector_duplicate_value): Use gen_vec_duplicate.
9953         * config/i386/sse.md (one_cmpl<mode>2): Use CONSTM1_RTX.
9954         * config/ia64/ia64.c (ia64_expand_vecint_compare): Use
9955         gen_const_vec_duplicate.
9956         * config/ia64/vect.md (addv2sf3, subv2sf3): Use CONST1_RTX.
9957         * config/mips/mips.c (mips_gen_const_int_vector): Use
9958         gen_const_vec_duplicate.
9959         (mips_expand_vector_init): Use CONST0_RTX.
9960         * config/powerpcspe/altivec.md (abs<mode>2, nabs<mode>2): Likewise.
9961         (define_split): Use gen_const_vec_duplicate.
9962         * config/rs6000/altivec.md (abs<mode>2, nabs<mode>2): Use CONST0_RTX.
9963         (define_split): Use gen_const_vec_duplicate.
9964         * config/s390/vx-builtins.md (vec_genmask<mode>): Likewise.
9965         (vec_ctd_s64, vec_ctd_u64, vec_ctsl, vec_ctul): Likewise.
9966         * config/spu/spu.c (spu_const): Likewise.
9968 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9969             Alan Hayward  <alan.hayward@arm.com>
9970             David Sherwood  <david.sherwood@arm.com>
9972         * combine.c (can_change_dest_mode): Reject changes in
9973         REGMODE_NATURAL_SIZE.
9975 2017-10-31  Sandra Loosemore  <sandra@codesourcery.com>
9977         * configure.ac (--enable-libssp): New.
9978         (gcc_cv_libc_provides_ssp): Check for explicit setting before
9979         trying to determine target-specific default.  Adjust indentation.
9980         * configure: Regenerated.
9981         * doc/install.texi (Configuration): Expand --disable-libssp
9982         documentation.
9984 2017-10-31  Daniel Santos  <daniel.santos@pobox.com>
9986         config/i386/i386.c (ix86_expand_epilogue): Correct stack
9987         calculation.
9989 2017-10-31  Martin Jambor  <mjambor@suse.cz>
9991         PR c++/81702
9992         * gimple-fold.c (gimple_get_virt_method_for_vtable): Remove assert.
9994 2017-10-31  David Malcolm  <dmalcolm@redhat.com>
9996         * auto-profile.c (autofdo_source_profile::read): Use
9997         UNKNOWN_LOCATION rather than 0.
9998         * diagnostic-core.h (warning_at_rich_loc): Rename to...
9999         (warning_at): ...this overload.
10000         (warning_at_rich_loc_n): Rename to...
10001         (warning_n): ...this overload.
10002         (error_at_rich_loc): Rename to...
10003         (error_at): ...this overload.
10004         (pedwarn_at_rich_loc): Rename to...
10005         (pedwarn): ...this overload.
10006         (permerror_at_rich_loc): Rename to...
10007         (permerror): ...this overload.
10008         (inform_at_rich_loc): Rename to...
10009         (inform): ...this overload.
10010         * diagnostic.c: (diagnostic_n_impl): Delete location_t-based decl.
10011         (diagnostic_n_impl_richloc): Rename to...
10012         (diagnostic_n_impl): ...this rich_location *-based decl.
10013         (inform_at_rich_loc): Rename to...
10014         (inform): ...this, and add an assertion.
10015         (inform_n): Update for removal of location_t-based diagnostic_n_impl.
10016         (warning_at_rich_loc): Rename to...
10017         (warning_at): ...this, and add an assertion.
10018         (warning_at_rich_loc_n): Rename to...
10019         (warning_n): ...this, and add an assertion.
10020         (warning_n): Update location_t-based implementation for removal of
10021         location_t-based diagnostic_n_impl.
10022         (pedwarn_at_rich_loc): Rename to...
10023         (pedwarn): ...this, and add an assertion.
10024         (permerror_at_rich_loc): Rename to...
10025         (permerror): ...this, and add an assertion.
10026         (error_n): Update for removal of location_t-based diagnostic_n_impl.
10027         (error_at_rich_loc): Rename to...
10028         (error_at): ...this, and add an assertion.
10029         * gcc.c (do_spec_1): Use UNKNOWN_LOCATION rather than 0.
10030         (driver::do_spec_on_infiles): Likewise.
10031         * substring-locations.c (format_warning_va): Update for renaming
10032         of inform_at_rich_loc.
10034 2017-10-31  Michael Meissner  <meissner@linux.vnet.ibm.com>
10036         * builtins.def (DEF_FLOATN_BUILTIN): Change most _Float<N> and
10037         _Float<N>X built-in functions so that the variant without the
10038         "__builtin_" prefix is only enabled for the GNU C and Objective C
10039         languages when they are in non-strict ANSI/ISO mode.
10040         (DEF_EXT_LIB_FLOATN_NX_BUILTINS): Likewise.
10041         * target.def (floatn_builtin_p): Add a target hook to control
10042         whether _Float<N> and _Float<N>X built-in functions without the
10043         "__builtin_" prefix are enabled, and return true for C and
10044         Objective C in the default hook.  Include langhooks.h in
10045         targhooks.c.
10046         * targhooks.h (default_floatn_builtin_p): Likewise.
10047         * targhooks.c (default_floatn_builtin_p): Likewise.
10048         * doc/tm.texi.in (TARGET_FLOATN_BUILTIN_P): Document the
10049         floatn_builtin_p target hook.
10050         * doc/tm.texi (TARGET_FLOATN_BUILTIN_P): Likewise.
10052 2017-10-31  Matthew Fortune  <matthew.fortune@imgtec.com>
10053             Eric Botcazou  <ebotcazou@adacore.com>
10055         PR rtl-optimization/81803
10056         * lra-constraints.c (curr_insn_transform): Also reload the whole
10057         register for a strict subreg no wider than a word if this is for
10058         a WORD_REGISTER_OPERATIONS target.
10060 2017-10-31  Jason Merrill  <jason@redhat.com>
10062         * gdbinit.in: Skip over inlines from timevar.h.
10064 2017-10-31  Martin Liska  <mliska@suse.cz>
10066         * doc/gcov.texi: Document new option.
10067         * gcov.c (print_usage): Likewise print it.
10068         (process_args): Support the argument.
10069         (format_count): New function.
10070         (format_gcov): Use the function.
10072 2017-10-31  Martin Liska  <mliska@suse.cz>
10074         * gcov.c (struct name_map): do not use typedef.
10075         Define operator== and operator<.
10076         (name_search): Remove.
10077         (name_sort): Remove.
10078         (main): Do not allocate names.
10079         (process_file): Add vertical space.
10080         (generate_results): Use std::find.
10081         (release_structures): Do not release memory.
10082         (find_source): Use std::find.
10084 2017-10-31  Martin Liska  <mliska@suse.cz>
10086         * gcov.c (struct line_info): Remove it's typedef.
10087         (line_info::line_info): Add proper ctor.
10088         (line_info::has_block): Do not use a typedef.
10089         (struct source_info): Do not use typedef.
10090         (circuit): Likewise.
10091         (get_cycles_count): Likewise.
10092         (output_intermediate_file): Iterate via vector iterator.
10093         (add_line_counts): Use std::vector methods.
10094         (accumulate_line_counts): Likewise.
10095         (output_lines): Likewise.
10097 2017-10-31  Martin Liska  <mliska@suse.cz>
10099         * gcov.c (struct source_info): Remove typedef.
10100         (source_info::source_info): Add proper ctor.
10101         (accumulate_line_counts): Use struct, not it's typedef.
10102         (output_gcov_file): Likewise.
10103         (output_lines): Likewise.
10104         (main): Do not allocate an array.
10105         (output_intermediate_file): Use size of vector container.
10106         (process_file): Resize the vector.
10107         (generate_results): Do not preallocate, use newly added vector
10108         lines.
10109         (release_structures): Do not release sources.
10110         (find_source): Use vector methods.
10111         (add_line_counts): Do not use typedef.
10113 2017-10-31  Martin Liska  <mliska@suse.cz>
10115         * doc/gcov.texi: Document that.
10116         * gcov.c (add_line_counts): Mark lines with a non-executed
10117         statement.
10118         (output_line_beginning): Handle such lines.
10119         (output_lines): Pass new argument.
10120         (output_intermediate_file): Print it in intermediate format.
10122 2017-10-31  Martin Liska  <mliska@suse.cz>
10124         * color-macros.h: New file.
10125         * diagnostic-color.c: Factor out color related to macros to
10126         color-macros.h.
10127         * doc/gcov.texi: Document -k option.
10128         * gcov.c (INCLUDE_STRING): Include string.h.
10129         (print_usage): Add -k option.
10130         (process_args): Parse it.
10131         (pad_count_string): New function.
10132         (output_line_beginning): Likewise.
10133         (DEFAULT_LINE_START): New macro.
10134         (output_lines): Support color output.
10136 2017-10-31  Martin Liska  <mliska@suse.cz>
10138         PR gcov-profile/82633
10139         * doc/gcov.texi: Document -fkeep-{static,inline}-functions and
10140         their interaction with GCOV infrastructure.
10141         * configure.ac: Add -fkeep-{inline,static}-functions to
10142         coverage_flags.
10143         * configure: Regenerate.
10145 2017-10-31  Uros Bizjak  <ubizjak@gmail.com>
10147         PR target/82772
10148         * config/alpha/sync.md (fetchop_constr) <and>: Change to "rINM".
10150 2017-10-31  Segher Boessenkool  <segher@kernel.crashing.org>
10152         PR target/82674
10153         * config/rs6000/rs6000.md (allocate_stack): Force update interval
10154         into a register if it does not fit into an immediate offset field.
10156 2017-10-31  Olivier Hainque  <hainque@adacore.com>
10158         * gcc/Makefile.in (FLAGS_TO_PASS): Pass libsubdir as well.
10160 2017-10-31  Julia Koval  <julia.koval@intel.com>
10162         * config.gcc: Add gfniintrin.h.
10163         * config/i386/gfniintrin.h: New.
10164         * config/i386/i386-builtin-types.def
10165         (__builtin_ia32_vgf2p8affineinvqb_v64qi,
10166         __builtin_ia32_vgf2p8affineinvqb_v64qi_mask,
10167         __builtin_ia32_vgf2p8affineinvqb_v32qi,
10168         __builtin_ia32_vgf2p8affineinvqb_v32qi_mask,
10169         __builtin_ia32_vgf2p8affineinvqb_v16qi,
10170         __builtin_ia32_vgf2p8affineinvqb_v16qi_mask): New builtins.
10171         * config/i386/i386-builtin.def (V64QI_FTYPE_V64QI_V64QI_INT_V64QI_UDI,
10172         V32QI_FTYPE_V32QI_V32QI_INT_V32QI_USI,
10173         V16QI_FTYPE_V16QI_V16QI_INT_V16QI_UHI,
10174         V64QI_FTYPE_V64QI_V64QI_INT): New types.
10175         * config/i386/i386.c (ix86_expand_args_builtin): Handle new types.
10176         * config/i386/immintrin.h: Include gfniintrin.h.
10177         * config/i386/sse.md (vgf2p8affineinvqb_*) New pattern.
10179 2017-10-30  Eric Botcazou  <ebotcazou@adacore.com>
10181         * gcc.c (HAVE_TARGET_EXECUTABLE_SUFFIX): Remove old kludge.
10183 2017-10-30  Wilco Dijkstra  <wdijkstr@arm.com>
10185         * config/arm/arm.md (ashldi3): Remove shift by 1 expansion.
10186         (arm_ashldi3_1bit): Remove pattern.
10187         (ashrdi3): Remove shift by 1 expansion.
10188         (arm_ashrdi3_1bit): Remove pattern.
10189         (lshrdi3): Remove shift by 1 expansion.
10190         (arm_lshrdi3_1bit): Remove pattern.
10191         * config/arm/arm.c (arm_rtx_costs_internal): Slightly increase
10192         cost of ashldi3 by 1.
10193         * config/arm/neon.md (ashldi3_neon): Remove shift by 1 expansion.
10194         (<shift>di3_neon): Likewise.
10196 2017-10-30  Dominik Infuehr  <dominik.infuehr@theobroma-systems.com>
10198         * config/aarch64/aarch64-simd.md (*aarch64_simd_mov): Rename
10199         both identically named patterns to (*aarch64_simd_mov<VD:mode>)
10200         and (*aarch64_simd_mov<VQ:mode>).
10201         (*aarch64_simd_mov<VD:mode>): Change type attribute to match
10202         pattern alternative.
10203         (*aarch64_simd_mov<VQ:mode>): Re-order and change type
10204         attributes to match pattern alternative.
10206 2017-10-30  Steven Munroe  <munroesj@gcc.gnu.org>
10208         * config.gcc (powerpc*-*-*): Add emmintrin.h.
10209         * config/rs6000/emmintrin.h: New file.
10210         * config/rs6000/x86intrin.h [__ALTIVEC__]: Include emmintrin.h.
10212 2017-10-30  Wilco Dijkstra  <wdijkstr@arm.com>
10214         * config/arm/vfp.md (movdi_vfp): Merge changes from movdi_vfp_cortexa8.
10215         * (movdi_vfp_cortexa8): Remove pattern.
10217 2017-10-30  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
10219         * doc/install.texi (Specific, alpha*-*-*): Remove DEC OSF/1
10220         etc. reference.
10221         (Specific, alpha*-dec-osf5.1): Remove.
10222         (Specific, mips-sgi-irix5): Remove.
10223         (Specific, mips-sgi-irix6): Remove.
10225 2017-10-30  Jakub Jelinek  <jakub@redhat.com>
10227         PR middle-end/22141
10228         * gimple-ssa-store-merging.c (merged_store_group::apply_stores): Fix
10229         arguments to clear_bit_region_be.
10231 2017-10-30  Jim Wilson  <wilson@tuliptree.org>
10233         * gimplify.c: Include memmodel.h.
10235 2017-10-30  Martin Jambor  <mjambor@suse.cz>
10237         * omp-grid.c (grid_attempt_target_gridification): Also insert a
10238         condition whether loop should be executed at all.
10240 2017-10-30  Will Schmidt  <will_schmidt@vnet.ibm.com>
10242         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add support for
10243         gimple folding of vec_madd() intrinsics.
10244         * config/rs6000/altivec.md (mulv8hi3): Rename altivec_vmladduhm to
10245         fmav8hi4.  (altivec_vmladduhm): Rename to fmav8hi4.
10246         * config/rs6000/rs6000-builtin.def: Rename vmladduhm to fmav8hi4
10248 2017-10-30  Richard Biener  <rguenther@suse.de>
10250         PR tree-optimization/82762
10251         Revert
10252         2017-10-23  Richard Biener  <rguenther@suse.de>
10254         PR tree-optimization/82129
10255         Revert
10256         2017-08-01  Richard Biener  <rguenther@suse.de>
10258         PR tree-optimization/81181
10259         * tree-ssa-pre.c (compute_antic_aux): Defer clean() to ...
10260         (compute_antic): ... end of iteration here.
10262 2017-10-30  Joseph Myers  <joseph@codesourcery.com>
10264         * doc/invoke.texi (C Dialect Options): Document -std=c17,
10265         -std=iso9899:2017 and -std=gnu17.
10266         * doc/standards.texi (C Language): Document C17 support.
10267         * doc/cpp.texi (Overview): Mention -std=c17.
10268         (Standard Predefined Macros): Document C11 and C17 values of
10269         __STDC_VERSION__.  Do not refer to C99 support as incomplete.
10270         * doc/extend.texi (Inline): Do not list individual options for
10271         standards newer than C99.
10272         * dwarf2out.c (highest_c_language, gen_compile_unit_die): Handle
10273         "GNU C17".
10274         * config/rl78/rl78.c (rl78_option_override): Handle "GNU C17"
10275         language name.
10277 2017-10-30  Maxim Ostapenko  <m.ostapenko@samsung.com>
10279         * asan.c (asan_finish_file): Align asan globals array by shadow
10280         granularity.
10282 2017-10-30  Jakub Jelinek  <jakub@redhat.com>
10284         PR middle-end/22141
10285         * gimple-ssa-store-merging.c: Include rtl.h and expr.h.
10286         (struct store_immediate_info): Add bitregion_start and bitregion_end
10287         fields.
10288         (store_immediate_info::store_immediate_info): Add brs and bre
10289         arguments and initialize bitregion_{start,end} from those.
10290         (struct merged_store_group): Add bitregion_start, bitregion_end,
10291         align_base and mask fields.  Drop unnecessary struct keyword from
10292         struct store_immediate_info.  Add do_merge method.
10293         (clear_bit_region_be): Use memset instead of loop storing zeros.
10294         (merged_store_group::do_merge): New method.
10295         (merged_store_group::merge_into): Use do_merge.  Allow gaps in between
10296         stores as long as the surrounding bitregions have no gaps.
10297         (merged_store_group::merge_overlapping): Use do_merge.
10298         (merged_store_group::apply_stores): Test that bitregion_{start,end}
10299         is byte aligned, rather than requiring that start and width are
10300         byte aligned.  Drop unnecessary struct keyword from
10301         struct store_immediate_info.  Allocate and populate also mask array.
10302         Make start of the arrays relative to bitregion_start rather than
10303         start and size them according to bitregion_{end,start} difference.
10304         (struct imm_store_chain_info): Drop unnecessary struct keyword from
10305         struct store_immediate_info.
10306         (pass_store_merging::gate): Punt if BITS_PER_UNIT or CHAR_BIT is not 8.
10307         (pass_store_merging::terminate_all_aliasing_chains): Drop unnecessary
10308         struct keyword from struct store_immediate_info.
10309         (imm_store_chain_info::coalesce_immediate_stores): Allow gaps in
10310         between stores as long as the surrounding bitregions have no gaps.
10311         Formatting fixes.
10312         (struct split_store): Add orig non-static data member.
10313         (split_store::split_store): Initialize orig to false.
10314         (find_constituent_stmts): Return store_immediate_info *, non-NULL
10315         if there is exactly a single original stmt.  Change stmts argument
10316         to pointer from reference, if NULL, don't push anything to it.  Add
10317         first argument, use it to optimize skipping over orig stmts that
10318         are known to be before bitpos already.  Simplify.
10319         (split_group): Return unsigned int count how many stores are or
10320         would be needed rather than a bool.  Add allow_unaligned argument.
10321         Change split_stores argument from reference to pointer, if NULL,
10322         only do a dry run computing how many stores would be produced.
10323         Rewritten algorithm to use both alignment and misalign if
10324         !allow_unaligned and handle bitfield stores with gaps.
10325         (imm_store_chain_info::output_merged_store): Set start_byte_pos
10326         from bitregion_start instead of start.  Compute allow_unaligned
10327         here, if true, do 2 split_group dry runs to compute which one
10328         produces fewer stores and prefer aligned if equal.  Punt if
10329         new count is bigger or equal than original before emitting any
10330         statements, rather than during that.  Remove no longer needed
10331         new_ssa_names tracking.  Replace num_stmts with
10332         split_stores.length ().  Use 32-bit stack allocated entries
10333         in split_stores auto_vec.  Try to reuse original store lhs/rhs1
10334         if possible.  Handle bitfields with gaps.
10335         (pass_store_merging::execute): Ignore bitsize == 0 stores.
10336         Compute bitregion_{start,end} for the stores and construct
10337         store_immediate_info with that.  Formatting fixes.
10339 2017-10-30  Uros Bizjak  <ubizjak@gmail.com>
10341         PR target/82725
10342         * config/i386/i386.c (legitimate_pic_address_disp_p): Allow
10343         UNSPEC_DTPOFF and UNSPEC_NTPOFF with SImode immediate offset.
10345 2017-10-29  Jim Wilson  <wilson@tuliptree.org>
10347         * gimplify.c: Include tm_p.h.
10349         * common.opt (gcoff): Re-add as ignored option.
10350         (gcoff1, gcoff2, gcoff3): Likewise.
10352         * Makefile.in (OBJS): Delete sdbout.o.
10353         (GTFILES): Delete $(srcdir)/sdbout.c.
10354         * debug.h: Delete sdb_debug_hooks.
10355         * final.c: Delete sdbout.h include.
10356         (final_scan_insn): Delete SDB_DEBUG check.
10357         (rest_of_clean_state): Likewise.
10358         * output.h: Delete sdb_begin_function_line.
10359         * sdbout.c: Delete.
10360         * sdbout.h: Delete.
10361         * toplev.c: Delete sdbout.h include.
10362         (process_options): Delete SDB_DEBUG check.
10363         * tree-core.h (tree_type_common): Delete pointer field of
10364         tree_type_symtab.
10365         * tree.c (copy_node): Clear TYPE_SYMTAB_DIE instead of
10366         TYPE_SYMTAB_POINTER.
10367         * tree.h (TYPE_SYMTAB_POINTER): Delete.
10368         (TYPE_SYMTAB_IS_POINTER): Delete.
10369         (TYPE_SYMTAB_IS_DIE): Renumber.
10370         * xcoffout.c: Refer to former sdbout.c file.
10371         (xcoffout_begin_prologue): Use past tense for sdbout.c reference.
10373         * doc/install.texi (--with-stabs): Delete COFF and ECOFF info.
10374         * doc/invoke.texi (SEEALSO): Delete adb and sdb references.
10375         (Debugging Options): Delete -gcoff.
10376         (-gstabs): Delete SDB reference.
10377         (-gcoff): Delete.
10378         (-gcoff@var{level}): Delete.
10379         * doc/passes.texi (Debugging information output): Delete SDB and
10380         sdbout.c references.
10381         * doc/tm.texi: Regenerate.
10382         * doc/tm.texi.in (DWARF_CIE_DATA_ALIGNMENT): Delete SDB from xref.
10383         (SDB and DWARF): Change node name to DWARF and delete SDB and COFF
10384         references.
10385         (DEBUGGER_AUTO_OFFSET): Delete COFF and SDB references.
10386         (PREFERRED_DEBUGGING_TYPE): Delete SDB_DEBUG and -gcoff references.
10387         (SDB_DEBUGGING_INFO): Delete.
10388         (PUT_SDB_@dots{}, SDB_DELIM, SDB_ALLOW_UNKNOWN_REFERENCES)
10389         SDB_ALLOW_FORWARD_REFERENCES, SDB_OUTPUT_SOURCE_LINE): Delete.
10390         * target.def (output_source_filename): Delete COFF reference.
10392         * common.opt (gcoff): Delete.
10393         (gxcoff+): Update Negative chain.
10394         * defaults.h: Delete all references to SDB_DEBUGGING_INFO and
10395         SDB_DEBUG.
10396         * dwarf2out.c (gen_array_type_die): Change SDB to debuggers.
10397         * flag-types.h (enum debug_info_type): Delete SDB_DEBUG.
10398         * function.c (number_blocks): Delete SDB_DEBUGGING_INFO, SDB_DEBUG,
10399         and SDB references.
10400         (expand_function_start): Change sdb reference to past tense.
10401         (expand_function_end): Change sdb reference to past tense.
10402         * gcc.c (cpp_unique_options): Delete gcoff3 reference.
10403         * opts.c (debug_type_names): Delete coff entry.
10404         (common_handle_option): Delete OPT_gcoff case.
10405         * system.h (SDB_DEBUG, SDB_DEBUGGING_INFO): Poison.
10407         * config/dbxcoff.h (PREFERRED_DEBUGGING_TYPE): Set to DBX_DEBUG.
10408         * config/cris/cris.h: Delete SDB reference in comment.
10409         * config/i386/cygming.h: Don't define SDB_DEBUGGING_INFO.
10410         (ASM_DECLARE_FUNCTION_NAME): Delete SDB reference from comment.
10411         * config/i386/gas.h: Don't define SDB_DEBUGGING_INFO.
10412         * config/i386/i386.c (svr4_dbx_register_map): Change SDB references
10413         to past tense.
10414         (ix86_expand_prologue): Likewise.
10415         * config/i386/winnt.c (i386_pe_start_function): Don't check SDB_DEBUG.
10416         * config/ia64/ia64.h: Likewise.
10417         * config/m68k/m68kelf.h (DBX_REGISTER_NUMBER): Delete SDB reference.
10418         * config/mips/mips.h (SUBTARGET_ASM_DEBUGGING_SPEC): Delete gcoff*
10419         support.
10420         * config/mmix/mmix.h: Likewise.
10421         * config/nds32/nds32.c: Likewise.
10422         * config/stormy/storym16.h: Likewise.
10423         * config/visium/visium.h: Likewise.
10424         * config/vx-common.h (SDB_DEBUGGING_INFO): Delete undef.
10426 2017-10-28  Sandra Loosemore  <sandra@codesourcery.com>
10428         * config/nios2/nios2.h (FRAME_GROWS_DOWNWARD): Define to 1.
10429         * config/nios2/nios2.c (nios2_initial_elimination_offset):  Make
10430         FRAME_POINTER_REGNUM point at high end of local var area.
10432 2017-10-27  Eric Botcazou  <ebotcazou@adacore.com>
10434         * bb-reorder.c (find_traces_1_round): Fix off-by-one index.
10435         Move comment around.  Do not reset best_edge for a copiable
10436         destination if the copy would cause a partition change.
10437         (better_edge_p): Remove redundant check.
10439 2017-10-27  Uros Bizjak  <ubizjak@gmail.com>
10441         * config/i386/i386-protos.h (ix86_fp_compare_mode): Remove prototype.
10443 2017-10-27  Michael Meissner  <meissner@linux.vnet.ibm.com>
10445         * builtins.c (CASE_MATHFN_FLOATN): New helper macro to add cases
10446         for math functions that have _Float<N> and _Float<N>X variants.
10447         (mathfn_built_in_2): Add support for math functions that have
10448         _Float<N> and _Float<N>X variants.
10449         (DEF_INTERNAL_FLT_FLOATN_FN): New helper macro.
10450         (expand_builtin_mathfn_ternary): Add support for fma with
10451         _Float<N> and _Float<N>X variants.
10452         (expand_builtin): Likewise.
10453         (fold_builtin_3): Likewise.
10454         * builtins.def (DEF_EXT_LIB_FLOATN_NX_BUILTINS): New macro to
10455         create math function _Float<N> and _Float<N>X variants as external
10456         library builtins.
10457         (BUILT_IN_COPYSIGN _Float<N> and _Float<N>X variants) Use
10458         DEF_EXT_LIB_FLOATN_NX_BUILTINS to make built-in functions using
10459         the __builtin_ prefix and if not strict ansi, without the prefix.
10460         (BUILT_IN_FABS _Float<N> and _Float<N>X variants): Likewise.
10461         (BUILT_IN_FMA _Float<N> and _Float<N>X variants): Likewise.
10462         (BUILT_IN_FMAX _Float<N> and _Float<N>X variants): Likewise.
10463         (BUILT_IN_FMIN _Float<N> and _Float<N>X variants): Likewise.
10464         (BUILT_IN_NAN _Float<N> and _Float<N>X variants): Likewise.
10465         (BUILT_IN_SQRT _Float<N> and _Float<N>X variants): Likewise.
10466         * builtin-types.def (BT_FN_FLOAT16_FLOAT16_FLOAT16_FLOAT16): New
10467         function signatures for fma _Float<N> and _Float<N>X variants.
10468         (BT_FN_FLOAT32_FLOAT32_FLOAT32_FLOAT32): Likewise.
10469         (BT_FN_FLOAT64_FLOAT64_FLOAT64_FLOAT64): Likewise.
10470         (BT_FN_FLOAT128_FLOAT128_FLOAT128_FLOAT128): Likewise.
10471         (BT_FN_FLOAT32X_FLOAT32X_FLOAT32X_FLOAT32X): Likewise.
10472         (BT_FN_FLOAT64X_FLOAT64X_FLOAT64X_FLOAT64X): Likewise.
10473         (BT_FN_FLOAT128X_FLOAT128X_FLOAT128X_FLOAT128X): Likewise.
10474         * gencfn-macros.c (print_case_cfn): Add support for math functions
10475         that have _Float<N> and _Float<N>X variants.
10476         (print_define_operator_list): Likewise.
10477         (fltfn_suffixes): Likewise.
10478         (main): Likewise.
10479         * internal-fn.def (DEF_INTERNAL_FLT_FLOATN_FN): New helper macro
10480         for math functions that have _Float<N> and _Float<N>X variants.
10481         (SQRT): Add support for sqrt, copysign, fmin and fmax _Float<N>
10482         and _Float<N>X variants.
10483         (COPYSIGN): Likewise.
10484         (FMIN): Likewise.
10485         (FMAX): Likewise.
10486         * fold-const.c (tree_call_nonnegative_warnv_p): Add support for
10487         copysign, fma, fmax, fmin, and sqrt _Float<N> and _Float<N>X
10488         variants.
10489         (integer_valued_read_call_p): Likewise.
10490         * fold-const-call.c (fold_const_call_ss): Likewise.
10491         (fold_const_call_sss): Add support for copysign, fmin, and fmax
10492         _Float<N> and _Float<N>X variants.
10493         (fold_const_call_ssss): Add support for fma _Float<N> and
10494         _Float<N>X variants.
10495         * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Add
10496         support for copysign and fma _Float<N> and _Float<N>X variants.
10497         (backprop::process_builtin_call_use): Likewise.
10498         * tree-call-cdce.c (can_test_argument_range); Add support for
10499         sqrt _Float<N> and _Float<N>X variants.
10500         (edom_only_function): Likewise.
10501         (get_no_error_domain): Likewise.
10502         * tree-ssa-math-opts.c (internal_fn_reciprocal): Likewise.
10503         * tree-ssa-reassoc.c (attempt_builtin_copysign): Add support for
10504         copysign _Float<N> and _Float<N>X variants.
10505         * config/rs6000/rs6000-builtin.def (SQRTF128): Delete, this is now
10506         handled by machine independent code.
10507         (FMAF128): Likewise.
10508         * doc/cpp.texi (Common Predefined Macros): Document defining
10509         __FP_FAST_FMAF<N> and __FP_FAST_FMAF<N>X if the backend supports
10510         fma _Float<N> and _Float<N>X variants.
10512 2017-10-27  Uros Bizjak  <ubizjak@gmail.com>
10514         PR target/82692
10515         * config/i386/i386-modes.def (CCFPU): Remove definition.
10516         * config/i386/i386.c (put_condition_mode): Remove CCFPU mode handling.
10517         (ix86_cc_modes_compatible): Ditto.
10518         (ix86_expand_carry_flag_compare): Ditto.
10519         (ix86_expand_int_movcc): Ditto.
10520         (ix86_expand_int_addcc): Ditto.
10521         (ix86_reverse_condition): Ditto.
10522         (ix86_unordered_fp_compare): Rename from ix86_fp_compare_mode.
10523         Return true/false for unordered/ordered fp comparisons.
10524         (ix86_cc_mode): Always return CCFPmode for float mode comparisons.
10525         (ix86_prepare_fp_compare_args): Update for rename.
10526         (ix86_expand_fp_compare): Update for rename.  Generate unordered
10527         compare RTXes wrapped with UNSPEC_NOTRAP unspec.
10528         (ix86_expand_sse_compare_and_jump): Ditto.
10529         * config/i386/predicates.md (fcmov_comparison_operator):
10530         Remove CCFPU mode handling.
10531         (ix86_comparison_operator): Ditto.
10532         (ix86_carry_flag_operator): Ditto.
10533         * config/i386/i386.md (UNSPEC_NOTRAP): New unspec.
10534         (*cmpu<mode>_i387): Wrap compare RTX with UNSPEC_NOTRAP unspec.
10535         (*cmpu<mode>_cc_i387): Ditto.
10536         (FPCMP): Remove mode iterator.
10537         (unord): Remove mode attribute.
10538         (unord_subst): New define_subst transformation
10539         (unord): New define_subst attribute.
10540         (unordered): Ditto.
10541         (*cmpi<unord><MODEF:mode>): Rewrite using unord_subst transformation.
10542         (*cmpi<unord>xf_i387): Ditto.
10543         * config/i386/sse.md (<sse>_<unord>comi<round_saeonly_name>): Merge
10544         from <sse>_comi<round_saeonly_name> and <sse>_ucomi<round_saeonly_name>
10545         using unord_subst transformation.
10546         * config/i386/subst.md (SUBST_A): Remove CCFP and CCFPU modes.
10547         (round_saeonly): Also handle CCFP mode.
10548         * reg-stack.c (subst_stack_regs_pat): Handle UNSPEC_NOTRAP unspec.
10549         Remove UNSPEC_SAHF unspec handling.
10551 2017-10-27  Jan Hubicka  <hubicka@ucw.cz>
10553         * x86-tune.def (X86_TUNE_INTER_UNIT_MOVES_TO_VEC): Disable for Zen.
10555 2017-10-27  Jeff Law  <law@redhat.com>
10557         * gimple-ssa-sprintf.c: Include domwalk.h.
10558         (class sprintf_dom_walker): New class, derived from dom_walker.
10559         (sprintf_dom_walker::before_dom_children): New function.
10560         (struct call_info): Moved into sprintf_dom_walker class
10561         (compute_formath_length, handle_gimple_call): Likewise.
10562         (sprintf_length::execute): Call the dominator walker rather
10563         than walking the statements.
10565         * tree-vrp.c (check_all_array_refs): Do not use wi->info to smuggle
10566         gimple statement locations.
10567         (check_array_bounds): Corresponding changes.  Get the statement's
10568         location directly from wi->stmt.
10570 2017-10-27  Palmer Dabbelt  <palmer@dabbelt.com>
10572         PR target/82717
10573         * doc/invoke.texi (RISC-V) <-mabi>: Correct and improve.
10575 2017-10-27  Jan Hubicka  <hubicka@ucw.cz>
10577         * config/i386/x86-tune.def (X86_TUNE_PARTIAL_REG_DEPENDENCY,
10578         X86_TUNE_MOVX): Disable for Haswell and newer CPUs.
10580 2017-10-27  Jakub Jelinek  <jakub@redhat.com>
10582         PR target/82703
10583         * config/i386/i386-protos.h (maybe_get_pool_constant): Removed.
10584         * config/i386/i386.c (maybe_get_pool_constant): Removed.
10585         (ix86_split_to_parts): Use avoid_constant_pool_reference instead of
10586         maybe_get_pool_constant.
10587         * config/i386/predicates.md (zero_extended_scalar_load_operand):
10588         Likewise.
10590 2017-10-27  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
10592         * doc/install.texi (Specific, i?86-*-solaris2.10): Simplify gas
10593         2.26 caveat.  Update gas and gld versions.
10594         (Specific, *-*-solaris2*): Update binutils version.  Remove caveat
10595         reference.
10597 2017-10-27  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
10599         * cgraph.h (set_malloc_flag): Declare.
10600         * cgraph.c (set_malloc_flag_1): New function.
10601         (set_malloc_flag): Likewise.
10602         * ipa-fnsummary.h (ipa_call_summary): Add new field is_return_callee.
10603         * ipa-fnsummary.c (ipa_call_summary::reset): Set is_return_callee to
10604         false.
10605         (read_ipa_call_summary): Add support for reading is_return_callee.
10606         (write_ipa_call_summary): Stream is_return_callee.
10607         * ipa-inline.c (ipa_inline): Remove call to ipa_free_fn_summary.
10608         * ipa-pure-const.c: Add headers ssa.h, alloc-pool.h, symbol-summary.h,
10609         ipa-prop.h, ipa-fnsummary.h.
10610         (pure_const_names): Change to static.
10611         (malloc_state_e): Define.
10612         (malloc_state_names): Define.
10613         (funct_state_d): Add field malloc_state.
10614         (varying_state): Set malloc_state to STATE_MALLOC_BOTTOM.
10615         (check_retval_uses): New function.
10616         (malloc_candidate_p): Likewise.
10617         (analyze_function): Add support for malloc attribute.
10618         (pure_const_write_summary): Stream malloc_state.
10619         (pure_const_read_summary): Add support for reading malloc_state.
10620         (dump_malloc_lattice): New function.
10621         (propagate_malloc): New function.
10622         (warn_function_malloc): New function.
10623         (ipa_pure_const::execute): Call propagate_malloc and
10624         ipa_free_fn_summary.
10625         (pass_local_pure_const::execute): Add support for malloc attribute.
10626         * ssa-iterators.h (RETURN_FROM_IMM_USE_STMT): New macro.
10627         * doc/invoke.texi: Document Wsuggest-attribute=malloc.
10629 2017-10-27  Martin Liska  <mliska@suse.cz>
10631         PR gcov-profile/82457
10632         * doc/invoke.texi: Document that one needs a non-strict ISO mode
10633         for fork-like functions to be properly instrumented.
10635 2017-10-27  Richard Biener  <rguenther@suse.de>
10637         PR middle-end/81659
10638         * tree-eh.c (pass_lower_eh_dispatch::execute): Free dominator
10639         info when we redirected EH.
10641 2017-10-26  Michael Collison  <michael.collison@arm.com>
10643         * config/aarch64/aarch64.md(<optab>_trunc><vf><GPI:mode>2):
10644         New pattern.
10645         (<optab>_trunchf<GPI:mode>2: New pattern.
10646         (<optab>_trunc<vgp><GPI:mode>2: New pattern.
10647         * config/aarch64/iterators.md (wv): New mode attribute.
10648         (vf, VF): New mode attributes.
10649         (vgp, VGP): New mode attributes.
10650         (s): Update attribute with SImode and DImode prefixes.
10652 2017-10-26  Sandra Loosemore  <sandra@codesourcery.com>
10654         * config/nios2/constraints.md ("S"): Match r0rel_constant_p too.
10655         * config/nios2/nios2-protos.h (r0rel_constant_p): Declare.
10656         * config/nios2/nios2.c: (nios2_r0rel_sec_regex): New.
10657         (nios2_option_overide): Initialize it.  Don't allow R0-relative
10658         addressing with PIC.
10659         (nios2_rtx_costs): Handle r0rel_constant_p like gprel_constant_p.
10660         (nios2_symbolic_constant_p): Likewise.
10661         (nios2_legitimate_address_p): Likewise.
10662         (nios2_r0rel_section_name_p): New.
10663         (nios2_symbol_ref_in_r0rel_data_p): New.
10664         (nios2_emit_move_sequence): Handle r0rel_constant_p.
10665         (r0rel_constant_p): New.
10666         (nios2_print_operand_address): Handle r0rel_constant_p.
10667         (nios2_cdx_narrow_form_p): Likewise.
10668         * config/nios2/nios2.opt (mr0rel-sec=): New option.
10669         * doc/invoke.texi (Option Summary): Add -mr0rel-sec.
10670         (Nios II Options): Document -mr0rel-sec.
10672 2017-10-26  Sandra Loosemore  <sandra@codesourcery.com>
10674         * config/nios2/nios2.c: Include xregex.h.
10675         (nios2_gprel_sec_regex): New.
10676         (nios2_option_overide): Initialize it.  Don't allow GP-relative
10677         addressing with PIC.
10678         (nios2_small_section_name_p): Check for regex match.
10679         * config/nios2/nios2.opt (mgprel-sec=): New option.
10680         * doc/invoke.texi (Option Summary): Add -mgprel-sec.
10681         (Nios II Options): Document -mgprel-sec.
10683 2017-10-26  Jim Wilson  <wilson@tuliptree.org>
10685         * doc/invoke.texi (-fdebug-prefix-map): Expand documentation.
10687 2017-10-26  Tom de Vries  <tom@codesourcery.com>
10689         PR tree-optimization/82707
10690         * gimple.c (gimple_copy): Fix unsharing of
10691         GIMPLE_OMP_{SINGLE,TARGET,TEAMS}.
10693 2017-10-26  Olga Makhotina  <olga.makhotina@intel.com>
10695         * config/i386/avx512fintrin.h (_mm512_cmpeq_pd_mask,
10696         _mm512_cmple_pd_mask, _mm512_cmplt_pd_mask,
10697         _mm512_cmpneq_pd_mask, _mm512_cmpnle_pd_mask,
10698         _mm512_cmpnlt_pd_mask, _mm512_cmpord_pd_mask,
10699         _mm512_cmpunord_pd_mask, _mm512_mask_cmpeq_pd_mask,
10700         _mm512_mask_cmple_pd_mask, _mm512_mask_cmplt_pd_mask,
10701         _mm512_mask_cmpneq_pd_mask, _mm512_mask_cmpnle_pd_mask,
10702         _mm512_mask_cmpnlt_pd_mask, _mm512_mask_cmpord_pd_mask,
10703         _mm512_mask_cmpunord_pd_mask, _mm512_cmpeq_ps_mask,
10704         _mm512_cmple_ps_mask, _mm512_cmplt_ps_mask,
10705         _mm512_cmpneq_ps_mask, _mm512_cmpnle_ps_mask,
10706         _mm512_cmpnlt_ps_mask, _mm512_cmpord_ps_mask,
10707         _mm512_cmpunord_ps_mask, _mm512_mask_cmpeq_ps_mask,
10708         _mm512_mask_cmple_ps_mask, _mm512_mask_cmplt_ps_mask,
10709         _mm512_mask_cmpneq_ps_mask, _mm512_mask_cmpnle_ps_mask,
10710         _mm512_mask_cmpnlt_ps_mask, _mm512_mask_cmpord_ps_mask,
10711         _mm512_mask_cmpunord_ps_mask): New intrinsics.
10713 2017-10-26  Michael Meissner  <meissner@linux.vnet.ibm.com>
10715         * config/rs6000/aix.h (TARGET_IEEEQUAD_DEFAULT): Set long double
10716         default to IBM.
10717         * config/rs6000/darwin.h (TARGET_IEEEQUAD_DEFAULT): Likewise.
10718         * config/rs6000/rs6000.opt (-mabi=ieeelongdouble): Move the
10719         warning to rs6000.c.  Remove the Undocumented flag, since it has
10720         been documented.
10721         (-mabi=ibmlongdouble): Likewise.
10722         * config/rs6000/rs6000.c (TARGET_IEEEQUAD_DEFAULT): If it is not
10723         already set, set the default format for long double.
10724         (rs6000_debug_reg_global): Print whether long double is IBM or
10725         IEEE.
10726         (rs6000_option_override_internal): Rework setting long double
10727         format.  Only warn if the user is changing the long double default
10728         and they did not use -Wno-psabi.
10729         * doc/invoke.texi (PowerPC options): Update the documentation for
10730         -mabi=ieeelongdouble and -mabi=ibmlongdouble.
10732 2017-10-26  Richard Sandiford  <richard.sandiford@linaro.org>
10733             Alan Hayward  <alan.hayward@arm.com>
10734             David Sherwood  <david.sherwood@arm.com>
10736         * rtl.h (wider_subreg_mode): New function.
10737         * ira.h (ira_sort_regnos_for_alter_reg): Take a machine_mode *
10738         rather than an unsigned int *.
10739         * ira-color.c (regno_max_ref_width): Replace with...
10740         (regno_max_ref_mode): ...this new variable.
10741         (coalesced_pseudo_reg_slot_compare): Update accordingly.
10742         Use wider_subreg_mode.
10743         (ira_sort_regnos_for_alter_reg): Likewise.  Take a machine_mode *
10744         rather than an unsigned int *.
10745         * lra-constraints.c (uses_hard_regs_p): Use wider_subreg_mode.
10746         (process_alt_operands): Likewise.
10747         (invariant_p): Likewise.
10748         * lra-spills.c (assign_mem_slot): Likewise.
10749         (add_pseudo_to_slot): Likewise.
10750         * lra.c (collect_non_operand_hard_regs): Likewise.
10751         (add_regs_to_insn_regno_info): Likewise.
10752         * reload1.c (regno_max_ref_width): Replace with...
10753         (regno_max_ref_mode): ...this new variable.
10754         (reload): Update accordingly.  Update call to
10755         ira_sort_regnos_for_alter_reg.
10756         (alter_reg): Update to use regno_max_ref_mode.  Call wider_subreg_mode.
10757         (init_eliminable_invariants): Update to use regno_max_ref_mode.
10758         (scan_paradoxical_subregs): Likewise.
10760 2017-10-26  Wilco Dijkstra  <wdijkstr@arm.com>
10762         * config/aarch64/aarch64.h (EXIT_IGNORE_STACK): Set if alloca is used.
10763         (aarch64_frame): Add emit_frame_chain boolean.
10764         * config/aarch64/aarch64.c (aarch64_frame_pointer_required)
10765         Move eh_return case to aarch64_layout_frame.
10766         (aarch64_layout_frame): Initialize emit_frame_chain.
10767         (aarch64_expand_prologue): Use emit_frame_chain.
10769 2017-10-26  Wilco Dijkstra  <wdijkstr@arm.com>
10771         * config/aarch64/aarch64.c (aarch64_layout_frame):
10772         Ensure LR is always stored at the bottom of the callee-saves.
10773         Remove rarely used frame layout which saves callee-saves at top of
10774         frame, so the store of LR can be used as a valid probe in all cases.
10776 2017-10-26  Wilco Dijkstra  <wdijkstr@arm.com>
10778         * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
10779         Improve unaligned TImode/TFmode base/offset split.
10781 2017-10-26  Richard Sandiford  <richard.sandiford@linaro.org>
10782             Alan Hayward  <alan.hayward@arm.com>
10783             David Sherwood  <david.sherwood@arm.com>
10785         * caller-save.c (mark_referenced_regs):  Use read_modify_subreg_p.
10786         * combine.c (find_single_use_1): Likewise.
10787         (expand_field_assignment): Likewise.
10788         (move_deaths): Likewise.
10789         * lra-constraints.c (simplify_operand_subreg): Likewise.
10790         (curr_insn_transform): Likewise.
10791         * lra.c (collect_non_operand_hard_regs): Likewise.
10792         (add_regs_to_insn_regno_info): Likewise.
10793         * rtlanal.c (reg_referenced_p): Likewise.
10794         (covers_regno_no_parallel_p): Likewise.
10796 2017-10-26  Richard Sandiford  <richard.sandiford@linaro.org>
10798         * wide-int-print.cc (print_hex): Loop based on extract_uhwi.
10799         Don't print any bits outside the precision of the value.
10800         * wide-int.cc (test_printing): Add some new tests.
10802 2017-10-26  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
10804         * configure.ac (gcc_cv_as_ix86_xbrace_comment): Check if assembler
10805         supports -xbrace_comment option.
10806         * configure: Regenerate.
10807         * config.in: Regenerate.
10808         * config/i386/sol2.h (ASM_XBRACE_COMMENT_SPEC): Define.
10809         (ASM_CPU_SPEC): Use it.
10811 2017-10-26  Richard Sandiford  <richard.sandiford@linaro.org>
10813         * target.def (static_rtx_alignment): New hook.
10814         * targhooks.h (default_static_rtx_alignment): Declare.
10815         * targhooks.c (default_static_rtx_alignment): New function.
10816         * doc/tm.texi.in (TARGET_STATIC_RTX_ALIGNMENT): New hook.
10817         * doc/tm.texi: Regenerate.
10818         * varasm.c (force_const_mem): Use targetm.static_rtx_alignment
10819         instead of targetm.constant_alignment.  Remove call to
10820         set_mem_attributes.
10821         * config/cris/cris.c (TARGET_STATIC_RTX_ALIGNMENT): Redefine.
10822         (cris_preferred_mininum_alignment): New function, split out from...
10823         (cris_constant_alignment): ...here.
10824         (cris_static_rtx_alignment): New function.
10825         * config/i386/i386.c (ix86_static_rtx_alignment): New function,
10826         split out from...
10827         (ix86_constant_alignment): ...here.
10828         (TARGET_STATIC_RTX_ALIGNMENT): Redefine.
10829         * config/mmix/mmix.c (TARGET_STATIC_RTX_ALIGNMENT): Redefine.
10830         (mmix_static_rtx_alignment): New function.
10831         * config/spu/spu.c (spu_static_rtx_alignment): New function.
10832         (TARGET_STATIC_RTX_ALIGNMENT): Redefine.
10834 2017-10-26  Tamar Christina  <tamar.christina@arm.com>
10836         PR target/81800
10837         * config/aarch64/aarch64.md (lrint<GPF:mode><GPI:mode>2):
10838         Add flag_trapping_math and flag_fp_int_builtin_inexact.
10840 2017-10-25  Palmer Dabbelt  <palmer@dabbelt.com>
10842         * config/riscv/riscv.md (ZERO_EXTEND_LOAD): Define.
10843         * config/riscv/pic.md (local_pic_load): Rename to local_pic_load_s,
10844         mark as a sign-extending load.
10845         (local_pic_load_u): Define.
10847 2017-10-25  Eric Botcazou  <ebotcazou@adacore.com>
10849         PR middle-end/82062
10850         * fold-const.c (operand_equal_for_comparison_p): Also return true
10851         if ARG0 is a simple variant of ARG1 with narrower precision.
10852         (fold_ternary_loc): Always pass unstripped operands to the predicate.
10854 2017-10-25  Jan Hubicka  <hubicka@ucw.cz>
10856         * i386.c (ix86_builtin_vectorization_cost): Compute scatter/gather
10857         cost correctly.
10858         * i386.h (processor_costs): Add gather_static, gather_per_elt,
10859         scatter_static, scatter_per_elt.
10860         * x86-tune-costs.h: Add new cost entries.
10862 2017-10-25  Richard Biener  <rguenther@suse.de>
10864         * tree-ssa-sccvn.h (vn_eliminate): Declare.
10865         * tree-ssa-pre.c (class eliminate_dom_walker, eliminate,
10866         class pass_fre): Move to ...
10867         * tree-ssa-sccvn.c (class eliminate_dom_walker, vn_eliminate,
10868         class pass_fre): ... here and adjust for statistics.
10870 2017-10-25  Jakub Jelinek  <jakub@redhat.com>
10872         PR libstdc++/81706
10873         * attribs.c (attribute_value_equal): Use omp_declare_simd_clauses_equal
10874         for comparison of OMP_CLAUSEs regardless of flag_openmp{,_simd}.
10875         (duplicate_one_attribute, copy_attributes_to_builtin): New functions.
10876         * attribs.h (duplicate_one_attribute, copy_attributes_to_builtin): New
10877         declarations.
10879 2017-10-25  Richard Biener  <rguenther@suse.de>
10881         * tree-ssa-pre.c (need_eh_cleanup, need_ab_cleanup, el_to_remove,
10882         el_to_fixup, el_todo, el_avail, el_avail_stack, eliminate_avail,
10883         eliminate_push_avail, eliminate_insert): Move inside...
10884         (class eliminate_dom_walker): ... this class in preparation
10885         of move.
10886         (fini_eliminate): Remove by merging with ...
10887         (eliminate): ... this function.  Adjust for class changes.
10888         (pass_pre::execute): Remove fini_eliminate call.
10889         (pass_fre::execute): Likewise.
10891 2017-10-24  Jakub Jelinek  <jakub@redhat.com>
10893         PR target/82460
10894         * config/i386/sse.md (UNSPEC_VPERMI2, UNSPEC_VPERMI2_MASK): Remove.
10895         (VPERMI2, VPERMI2I): New mode iterators.
10896         (<avx512>_vpermi2var<mode>3_maskz): Remove 3 define_expand patterns.
10897         (<avx512>_vpermi2var<mode>3<sd_maskz_name>): Remove 3 define_insn
10898         patterns.
10899         (<avx512>_vpermi2var<mode>3_mask): New define_expand using VPERMI2
10900         mode iterator.  Remove 3 old define_insn patterns.
10901         (*<avx512>_vpermi2var<mode>3_mask): 2 new define_insn patterns.
10902         (<avx512>_vpermt2var<mode>3_maskz): Adjust 1 define_expand to use
10903         VPERMI2 mode iterator, remove the other two expanders.
10904         (<avx512>_vpermt2var<mode>3<sd_maskz_name>): Adjust 1 define_insn
10905         to use VPERMI2 mode iterator, add another alternative for vpermi2*
10906         instructions, remove the other two patterns.
10907         (<avx512>_vpermt2var<mode>3_mask): Adjust 1 define_insn to use VPERMI2
10908         mode iterator, remove the other two patterns.
10909         * config/i386/i386.c (ix86_expand_vec_perm_vpermi2): Renamed to ...
10910         (ix86_expand_vec_perm_vpermt2): ... this.  Swap mask and op0
10911         arguments, use gen_*vpermt2* expanders instead of gen_*vpermi2*
10912         and adjust argument order accordingly.
10913         (ix86_expand_vec_perm): Adjust caller.
10914         (expand_vec_perm_1): Likewise.
10915         (expand_vec_perm_vpermi2_vpshub2): Rename to ...
10916         (expand_vec_perm_vpermt2_vpshub2): ... this.
10917         (ix86_expand_vec_perm_const_1): Adjust caller.
10918         (ix86_vectorize_vec_perm_const_ok): Adjust comments.
10920         PR target/82370
10921         * config/i386/sse.md (VIMAX_AVX2): Remove V4TImode.
10922         (VIMAX_AVX2_AVX512BW, VIMAX_AVX512VL): New mode iterators.
10923         (vec_shl_<mode>): Remove unused expander.
10924         (avx512bw_<shift_insn><mode>3): New define_insn.
10925         (<sse2_avx2>_ashl<mode>3, <sse2_avx2>_lshr<mode>3): Replaced by ...
10926         (<sse2_avx2>_<shift_insn><mode>3): ... this.  New define_insn.
10928 2017-10-24  Paolo Carlini  <paolo.carlini@oracle.com>
10930         PR c++/82466
10931         * doc/invoke.texi ([Wbuiltin-declaration-mismatch]): Extend
10932         description.
10934 2017-10-24  Wilco Dijkstra  <wdijkstr@arm.com>
10936         PR rtl-optimization/82396
10937         * gcc/haifa-sched.c (ready_sort_real): Remove qsort workaround.
10938         (autopref_multipass_init): Simplify initialization.
10939         (autopref_rank_data): Simplify sort order.
10940         * gcc/sched-int.h (autopref_multipass_data_): Remove
10941         multi_mem_insn_p, min_offset and max_offset.
10943 2017-10-24  Wilco Dijkstra  <wdijkstr@arm.com>
10945         PR middle-end/60580
10946         * config/aarch64/aarch64.c (aarch64_frame_pointer_required)
10947         Check special value of flag_omit_frame_pointer.
10948         (aarch64_can_eliminate): Likewise.
10949         (aarch64_override_options_after_change_1): Simplify handling of
10950         -fomit-frame-pointer and -fomit-leaf-frame-pointer.
10952 2017-10-24  Richard Biener  <rguenther@suse.de>
10954         PR tree-optimization/82697
10955         * tree-ssa-phiopt.c (cond_store_replacement): Use alias-set
10956         zero for conditional load and unconditional store.
10958 2017-10-24  H.J. Lu  <hongjiu.lu@intel.com>
10960         * doc/install.texi: Document bootstrap-cet.
10962 2017-10-24  H.J. Lu  <hongjiu.lu@intel.com>
10964         PR target/82659
10965         * config/i386/i386.c (rest_of_insert_endbranch): Don't insert
10966         ENDBR instruction at function entrance if function is only
10967         called directly.
10969 2017-10-24  Jakub Jelinek  <jakub@redhat.com>
10971         PR target/82628
10972         * config/i386/i386.md (addcarry<mode>, subborrow<mode>): Change
10973         patterns to better describe from which operation the CF is computed.
10974         (addcarry<mode>_0, subborrow<mode>_0): New patterns.
10975         * config/i386/i386.c (ix86_expand_builtin) <case handlecarry>: Pass
10976         one LTU with [DT]Imode and another one with [SD]Imode.  If arg0
10977         is 0, use _0 suffixed expanders instead of emitting a comparison
10978         before it.
10980 2017-10-06  Sergey Shalnov  <Sergey.Shalnov@intel.com>
10982         * config/i386/i386.md(*movsf_internal, *movdf_internal):
10983         Avoid 512-bit AVX modes for TARGET_PREFER_AVX256.
10985 2017-10-24  Eric Botcazou  <ebotcazou@adacore.com>
10987         PR middle-end/82569
10988         * tree-outof-ssa.h (always_initialized_rtx_for_ssa_name_p): Delete.
10989         * expr.c (expand_expr_real_1) <expand_decl_rtl>: Revert latest change.
10990         * loop-iv.c (iv_get_reaching_def): Likewise.
10991         * cfgexpand.c (expand_one_ssa_partition): Initialize the RTX if the
10992         variable is promoted and the partition contains undefined values.
10994 2017-10-23  Sandra Loosemore  <sandra@codesourcery.com>
10996         * config/nios2/nios2.c (nios2_rtx_costs): Make costs better
10997         reflect reality.
10998         (nios2_address_cost): Define.
10999         (nios2_legitimize_address): Recognize (exp + constant) directly.
11000         (TARGET_ADDRESS_COST): Define.
11002 2017-10-23  Sandra Loosemore  <sandra@codesourcery.com>
11004         * config/nios2/nios2-protos.h (nios2_large_constant_p): Declare.
11005         (nios2_symbolic_memory_operand_p): Declare.
11006         (nios2_split_large_constant): Declare.
11007         (nios2_split_symbolic_memory_operand): Declare.
11008         * config/nios2/nios2.c: Adjust includes.
11009         (nios2_symbolic_constant_allowed): New.
11010         (nios2_symbolic_constant_p): New.
11011         (nios2_plus_symbolic_constant_p): New.
11012         (nios2_valid_addr_expr_p): Recognize addresses involving
11013         symbolic constants.
11014         (nios2_legitimate_address_p): Likewise, also LO_SUM.
11015         (nios2_symbolic_memory_operand_p): New.
11016         (nios2_large_constant_p): New.
11017         (nios2_split_large_constant): New.
11018         (nios2_split_plus_large_constant): New.
11019         (nios2_split_symbolic_memory_operand): New.
11020         (nios2_legitimize_address): Code refactoring.  Handle addresses
11021         involving symbolic constants.
11022         (nios2_emit_move_sequence): Likewise.
11023         (nios2_print_operand): Improve error output.
11024         (nios2_print_operand_address): Handle LO_SUM.
11025         (nios2_cdx_narrow_form_p): Likewise.
11026         * config/nios2/nios2.md (movqi_internal): Add splitter for memory
11027         operands involving symbolic constants.
11028         (movhi_internal, movsi_internal): Likewise.
11029         (zero_extendhisi2, zero_extendqi<mode>2): Likewise.
11030         (extendhisi2, extendqi<mode>2): Likewise.
11032 2017-10-23  Sandra Loosemore  <sandra@codesourcery.com>
11034         * tree-pass.h (PROP_rtl_split_insns): Define.
11035         * recog.c (pass_data_split_all_insns): Provide PROP_rtl_split_insns.
11037 2017-10-23  Sandra Loosemore  <sandra@codesourcery.com>
11039         * config/nios2/nios2.c (TARGET_LRA_P): Don't override.
11041 2017-10-23  Jakub Jelinek  <jakub@redhat.com>
11043         PR debug/82630
11044         * target.def (const_not_ok_for_debug_p): Default to
11045         default_const_not_ok_for_debug_p instead of hook_bool_rtx_false.
11046         * targhooks.h (default_const_not_ok_for_debug_p): New declaration.
11047         * targhooks.c (default_const_not_ok_for_debug_p): New function.
11048         * dwarf2out.c (const_ok_for_output_1): Only reject UNSPECs for
11049         which targetm.const_not_ok_for_debug_p returned true.
11050         * config/arm/arm.c (arm_const_not_ok_for_debug_p): Return true
11051         for UNSPECs.
11052         * config/powerpcspe/powerpcspe.c (rs6000_const_not_ok_for_debug_p):
11053         Likewise.
11054         * config/rs6000/rs6000.c (rs6000_const_not_ok_for_debug_p): Likewise.
11055         * config/i386/i386.c (ix86_delegitimize_address_1): Don't delegitimize
11056         UNSPEC_GOTOFF with addend into addend - _GLOBAL_OFFSET_TABLE_ + symbol
11057         if !base_term_p.
11058         (ix86_const_not_ok_for_debug_p): New function.
11059         (i386_asm_output_addr_const_extra): Handle UNSPEC_GOTOFF.
11060         (TARGET_CONST_NOT_OK_FOR_DEBUG_P): Redefine.
11062 2017-10-23  David Malcolm  <dmalcolm@redhat.com>
11064         PR bootstrap/82610
11065         * system.h: Conditionally include "unique-ptr.h" if
11066         INCLUDE_UNIQUE_PTR is defined.
11067         * unique-ptr-tests.cc: Remove include of "unique-ptr.h" in favor
11068         of defining INCLUDE_UNIQUE_PTR before including "system.h".
11070 2017-10-23  Sebastian Perta  <sebastian.perta@renesas.com>
11072         * config/rl78/rl78.md: New define_expand "subdi3".
11074 2017-10-23  H.J. Lu  <hongjiu.lu@intel.com>
11076         PR target/82673
11077         * config/i386/i386.c (ix86_finalize_stack_frame_flags): Skip
11078         DF_REF_INSN if DF_REF_INSN_INFO is false.
11080 2017-10-23  Jan Hubicka  <hubicka@ucw.cz>
11082         * i386.c (dimode_scalar_chain::compute_convert_gain): Use
11083         xmm_move instead of sse_move.
11084         (sse_store_index): New function.
11085         (ix86_register_move_cost): Be more sensible about mismatch stall;
11086         model AVX moves correctly; make difference between sse->integer and
11087         integer->sse.
11088         (ix86_builtin_vectorization_cost): Model correctly aligned and unaligned
11089         moves; make difference between SSE and AVX.
11090         * i386.h (processor_costs): Remove sse_move; add xmm_move, ymm_move
11091         and zmm_move. Increase size of sse load and store tables;
11092         add unaligned load and store tables; add ssemmx_to_integer.
11093         * x86-tune-costs.h: Update all entries according to real
11094         move latencies from Agner Fog's manual and chip documentation.
11096 2017-10-23  Jakub Jelinek  <jakub@redhat.com>
11098         PR target/82628
11099         * config/i386/predicates.md (x86_64_dwzext_immediate_operand): New.
11100         * config/i386/constraints.md (Wf): New constraint.
11101         * config/i386/i386.md (UNSPEC_SBB): New unspec.
11102         (cmp<dwi>_doubleword): Removed.
11103         (sub<mode>3_carry_ccc, *sub<mode>3_carry_ccc_1): New patterns.
11104         (sub<mode>3_carry_ccgz): Use unspec instead of compare.
11105         * config/i386/i386.c (ix86_expand_branch) <case E_TImode>: Don't
11106         expand with cmp<dwi>_doubleword.  For LTU and GEU use
11107         sub<mode>3_carry_ccc instead of sub<mode>3_carry_ccgz and use CCCmode.
11109         * common.opt (gcolumn-info): Enable by default.
11110         * doc/invoke.texi (gcolumn-info): Document new default.
11112 2017-10-23  Richard Biener  <rguenther@suse.de>
11114         PR tree-optimization/82672
11115         * graphite-isl-ast-to-gimple.c (graphite_copy_stmts_from_block):
11116         Fold the stmt if we propagated into it.
11118 2017-10-23  Richard Biener  <rguenther@suse.de>
11120         * tree-ssa-pre.c (bitmap_remove_from_set): Rename to...
11121         (bitmap_remove_expr_from_set): ... this.  All callers call this
11122         for non-constant values.
11123         (bitmap_set_subtract): Rename to...
11124         (bitmap_set_subtract_expressions): ... this.  Adjust and
11125         optimize.
11126         (bitmap_set_contains_value): Remove superfluous check.
11127         (bitmap_set_replace_value): Inline into single caller ...
11128         (bitmap_value_replace_in_set): ... here and simplify.
11129         (dependent_clean): Merge into ...
11130         (clean): ... this using an overload.  Adjust.
11131         (prune_clobbered_mems): Adjust.
11132         (compute_antic_aux): Likewise.
11133         (compute_partial_antic_aux): Likewise.
11135 2017-10-23  Richard Biener  <rguenther@suse.de>
11137         PR tree-optimization/82129
11138         Revert
11139         2017-08-01  Richard Biener  <rguenther@suse.de>
11141         PR tree-optimization/81181
11142         * tree-ssa-pre.c (compute_antic_aux): Defer clean() to ...
11143         (compute_antic): ... end of iteration here.
11145 2017-10-23  Richard Sandiford  <richard.sandiford@linaro.org>
11147         * target.def (starting_frame_offset): New hook.
11148         * doc/tm.texi (STARTING_FRAME_OFFSET): Remove in favor of...
11149         (TARGET_STARTING_FRAME_OFFSET): ...this new hook.
11150         * doc/tm.texi.in: Regenerate.
11151         * hooks.h (hook_hwi_void_0): Declare.
11152         * hooks.c (hook_hwi_void_0): New function.
11153         * doc/rtl.texi: Refer to TARGET_STARTING_FRAME_OFFSET instead of
11154         STARTING_FRAME_OFFSET.
11155         * builtins.c (expand_builtin_setjmp_receiver): Likewise.
11156         * reload1.c (reload): Likewise.
11157         * cfgexpand.c (expand_used_vars): Use targetm.starting_frame_offset
11158         instead of STARTING_FRAME_OFFSET.
11159         * function.c (try_fit_stack_local): Likewise.
11160         (assign_stack_local_1): Likewise
11161         (instantiate_virtual_regs): Likewise.
11162         * rtlanal.c (rtx_addr_can_trap_p_1): Likewise.
11163         * config/avr/avr.md (nonlocal_goto_receiver): Likewise.
11164         * config/aarch64/aarch64.h (STARTING_FRAME_OFFSET): Delete.
11165         * config/alpha/alpha.h (STARTING_FRAME_OFFSET): Likewise.
11166         * config/arc/arc.h (STARTING_FRAME_OFFSET): Likewise.
11167         * config/arm/arm.h (STARTING_FRAME_OFFSET): Likewise.
11168         * config/bfin/bfin.h (STARTING_FRAME_OFFSET): Likewise.
11169         * config/c6x/c6x.h (STARTING_FRAME_OFFSET): Likewise.
11170         * config/cr16/cr16.h (STARTING_FRAME_OFFSET): Likewise.
11171         * config/cris/cris.h (STARTING_FRAME_OFFSET): Likewise.
11172         * config/fr30/fr30.h (STARTING_FRAME_OFFSET): Likewise.
11173         * config/frv/frv.h (STARTING_FRAME_OFFSET): Likewise.
11174         * config/ft32/ft32.h (STARTING_FRAME_OFFSET): Likewise.
11175         * config/h8300/h8300.h (STARTING_FRAME_OFFSET): Likewise.
11176         * config/i386/i386.h (STARTING_FRAME_OFFSET): Likewise.
11177         * config/ia64/ia64.h (STARTING_FRAME_OFFSET): Likewise.
11178         * config/m32c/m32c.h (STARTING_FRAME_OFFSET): Likewise.
11179         * config/m68k/m68k.h (STARTING_FRAME_OFFSET): Likewise.
11180         * config/mcore/mcore.h (STARTING_FRAME_OFFSET): Likewise.
11181         * config/mn10300/mn10300.h (STARTING_FRAME_OFFSET): Likewise.
11182         * config/moxie/moxie.h (STARTING_FRAME_OFFSET): Likewise.
11183         * config/msp430/msp430.h (STARTING_FRAME_OFFSET): Likewise.
11184         * config/nds32/nds32.h (STARTING_FRAME_OFFSET): Likewise.
11185         * config/nios2/nios2.h (STARTING_FRAME_OFFSET): Likewise.
11186         * config/nvptx/nvptx.h (STARTING_FRAME_OFFSET): Likewise.
11187         * config/pdp11/pdp11.h (STARTING_FRAME_OFFSET): Likewise.
11188         * config/riscv/riscv.h (STARTING_FRAME_OFFSET): Likewise.
11189         * config/rl78/rl78.h (STARTING_FRAME_OFFSET): Likewise.
11190         * config/rx/rx.h (STARTING_FRAME_OFFSET): Likewise.
11191         * config/s390/s390.h (STARTING_FRAME_OFFSET): Likewise.
11192         * config/sh/sh.h (STARTING_FRAME_OFFSET): Likewise.
11193         * config/sparc/sparc.c (sparc_compute_frame_size): Likewise.
11194         * config/sparc/sparc.h (STARTING_FRAME_OFFSET): Likewise.
11195         * config/spu/spu.h (STARTING_FRAME_OFFSET): Likewise.
11196         * config/stormy16/stormy16.h (STARTING_FRAME_OFFSET): Likewise.
11197         * config/tilegx/tilegx.h (STARTING_FRAME_OFFSET): Likewise.
11198         * config/tilepro/tilepro.h (STARTING_FRAME_OFFSET): Likewise.
11199         * config/v850/v850.h (STARTING_FRAME_OFFSET): Likewise.
11200         * config/visium/visium.h (STARTING_FRAME_OFFSET): Likewise.
11201         * config/avr/avr.h (STARTING_FRAME_OFFSET): Likewise.
11202         * config/avr/avr-protos.h (avr_starting_frame_offset): Likewise.
11203         * config/avr/avr.c (avr_starting_frame_offset): Make static and
11204         return a HOST_WIDE_INT.
11205         (avr_builtin_setjmp_frame_value): Use it instead of
11206         STARTING_FRAME_OFFSET.
11207         (TARGET_STARTING_FRAME_OFFSET): Redefine.
11208         * config/epiphany/epiphany.h (STARTING_FRAME_OFFSET): Delete.
11209         * config/epiphany/epiphany.c (epiphany_starting_frame_offset):
11210         New function.
11211         (TARGET_STARTING_FRAME_OFFSET): Redefine.
11212         * config/iq2000/iq2000.h (STARTING_FRAME_OFFSET): Delete.
11213         * config/iq2000/iq2000.c (iq2000_starting_frame_offset): New function.
11214         (TARGET_CONSTANT_ALIGNMENT): Redefine.
11215         * config/lm32/lm32.h (STARTING_FRAME_OFFSET): Delete.
11216         * config/lm32/lm32.c (lm32_starting_frame_offset): New function.
11217         (TARGET_STARTING_FRAME_OFFSET): Redefine.
11218         * config/m32r/m32r.h (STARTING_FRAME_OFFSET): Delete.
11219         * config/m32r/m32r.c (m32r_starting_frame_offset): New function.
11220         (TARGET_STARTING_FRAME_OFFSET): Redefine.
11221         * config/microblaze/microblaze.h (STARTING_FRAME_OFFSET): Delete.
11222         * config/microblaze/microblaze.c (microblaze_starting_frame_offset):
11223         New function.
11224         (TARGET_STARTING_FRAME_OFFSET): Redefine.
11225         * config/mips/mips.h (STARTING_FRAME_OFFSET): Delete.
11226         * config/mips/mips.c (mips_compute_frame_info): Refer to
11227         TARGET_STARTING_FRAME_OFFSET instead of STARTING_FRAME_OFFSET.
11228         (mips_starting_frame_offset): New function.
11229         (TARGET_STARTING_FRAME_OFFSET): Redefine.
11230         * config/mmix/mmix.h (STARTING_FRAME_OFFSET): Delete.
11231         * config/mmix/mmix-protos.h (mmix_starting_frame_offset): Delete.
11232         * config/mmix/mmix.c (mmix_starting_frame_offset): Make static
11233         and return a HOST_WIDE_INT.
11234         (TARGET_STARTING_FRAME_OFFSET): Redefine.
11235         (mmix_initial_elimination_offset): Refer to
11236         TARGET_STARTING_FRAME_OFFSET instead of STARTING_FRAME_OFFSET.
11237         * config/pa/pa.h (STARTING_FRAME_OFFSET): Delete.
11238         * config/pa/pa.c (pa_starting_frame_offset): New function.
11239         (pa_compute_frame_size): Use it instead of STARTING_FRAME_OFFSET.
11240         (pa_expand_prologue): Likewise.
11241         (TARGET_STARTING_FRAME_OFFSET): Redefine.
11242         * config/powerpcspe/aix.h (STARTING_FRAME_OFFSET): Split out
11243         !FRAME_GROWS_DOWNWARD handling to...
11244         (RS6000_STARTING_FRAME_OFFSET): ...this new macro.
11245         * config/powerpcspe/darwin.h (STARTING_FRAME_OFFSET): Split out
11246         !FRAME_GROWS_DOWNWARD handling to...
11247         (RS6000_STARTING_FRAME_OFFSET): ...this new macro.
11248         * config/powerpcspe/powerpcspe.h (STARTING_FRAME_OFFSET): Split out
11249         !FRAME_GROWS_DOWNWARD handling to...
11250         (RS6000_STARTING_FRAME_OFFSET): ...this new macro.
11251         * config/powerpcspe/powerpcspe.c (TARGET_STARTING_FRAME_OFFSET):
11252         Redefine.
11253         (rs6000_starting_frame_offset): New function.
11254         * config/rs6000/aix.h (STARTING_FRAME_OFFSET): Split out
11255         !FRAME_GROWS_DOWNWARD handling to...
11256         (RS6000_STARTING_FRAME_OFFSET): ...this new macro.
11257         * config/rs6000/darwin.h (STARTING_FRAME_OFFSET): Split out
11258         !FRAME_GROWS_DOWNWARD handling to...
11259         (RS6000_STARTING_FRAME_OFFSET): ...this new macro.
11260         * config/rs6000/rs6000.h (STARTING_FRAME_OFFSET): Split out
11261         !FRAME_GROWS_DOWNWARD handling to...
11262         (RS6000_STARTING_FRAME_OFFSET): ...this new macro.
11263         * config/rs6000/rs6000.c (TARGET_STARTING_FRAME_OFFSET): Refine.
11264         (rs6000_starting_frame_offset): New function.
11265         * config/vax/elf.h (STARTING_FRAME_OFFSET): Delete.
11266         * config/vax/vax.h (STARTING_FRAME_OFFSET): Delete.
11267         * config/vax/vax.c (vax_starting_frame_offset): New function.
11268         (vax_expand_prologue): Use it instead of STARTING_FRAME_OFFSET.
11269         (TARGET_STARTING_FRAME_OFFSET): Redefine.
11270         * config/xtensa/xtensa.h (STARTING_FRAME_OFFSET): Delete.
11271         * config/xtensa/xtensa.c (xtensa_starting_frame_offset): New function.
11272         (TARGET_STARTING_FRAME_OFFSET): Redefine.
11273         * system.h (STARTING_FRAME_OFFSET): Poison.
11275 2017-10-23  Richard Sandiford  <richard.sandiford@linaro.org>
11277         * tree-vect-loop.c (vect_create_epilog_for_reduction): Use
11278         SCALAR_TYPE_MODE instead of TYPE_MODE.
11280 2017-10-23  Richard Sandiford  <richard.sandiford@linaro.org>
11281             Alan Hayward  <alan.hayward@arm.com>
11282             David Sherwood  <david.sherwood@arm.com>
11284         * dwarf2out.c (loc_list_from_tree_1): Use SCALAR_INT_TYPE_MODE
11286 2017-10-23  Richard Sandiford  <richard.sandiford@linaro.org>
11287             Alan Hayward  <alan.hayward@arm.com>
11288             David Sherwood  <david.sherwood@arm.com>
11290         * expmed.c (expand_shift_1): Use scalar_mode for scalar_mode.
11292 2017-10-23  Richard Biener  <rguenther@suse.de>
11294         PR tree-optimization/82129
11295         * tree-ssa-pre.c (bitmap_set_and): Remove.
11296         (compute_antic_aux): Compute ANTIC_OUT intersection in a way
11297         canonicalizing expressions in the set to those with lowest
11298         ID rather than taking that from the first edge.
11300 2017-10-23  Richard Sandiford  <richard.sandiford@linaro.org>
11302         * combine.c (rtx_equal_for_field_assignment_p): Use
11303         byte_lowpart_offset.
11305 2017-10-22  Richard Sandiford  <richard.sandiford@linaro.org>
11306             Alan Hayward  <alan.hayward@arm.com>
11307             David Sherwood  <david.sherwood@arm.com>
11309         * internal-fn.c (expand_direct_optab_fn): Don't assign directly
11310         to a SUBREG_PROMOTED_VAR.
11312 2017-10-22  Richard Sandiford  <richard.sandiford@linaro.org>
11313             Alan Hayward  <alan.hayward@arm.com>
11314             David Sherwood  <david.sherwood@arm.com>
11316         * cfgexpand.c (expand_debug_expr): Use GET_MODE_UNIT_PRECISION.
11317         (expand_debug_source_expr): Likewise.
11318         * combine.c (combine_simplify_rtx): Likewise.
11319         * cse.c (fold_rtx): Likewise.
11320         * optabs.c (expand_float): Likewise.
11321         * simplify-rtx.c (simplify_unary_operation_1): Likewise.
11322         (simplify_binary_operation_1): Likewise.
11324 2017-10-22  Richard Sandiford  <richard.sandiford@linaro.org>
11325             Alan Hayward  <alan.hayward@arm.com>
11326             David Sherwood  <david.sherwood@arm.com>
11328         * combine.c (simplify_comparison): Use HWI_COMPUTABLE_MODE_P.
11329         (record_promoted_value): Likewise.
11330         * expr.c (expand_expr_real_2): Likewise.
11331         * ree.c (update_reg_equal_equiv_notes): Likewise.
11332         (combine_set_extension): Likewise.
11333         * rtlanal.c (low_bitmask_len): Likewise.
11334         * simplify-rtx.c (neg_const_int): Likewise.
11335         (simplify_binary_operation_1): Likewise.
11337 2017-10-22  Richard Sandiford  <richard.sandiford@linaro.org>
11338             Alan Hayward  <alan.hayward@arm.com>
11339             David Sherwood  <david.sherwood@arm.com>
11341         * lra-spills.c (assign_mem_slot): Use subreg_size_lowpart_offset.
11342         * regcprop.c (maybe_mode_change): Likewise.
11343         * reload1.c (alter_reg): Likewise.
11345 2017-10-22  Richard Sandiford  <richard.sandiford@linaro.org>
11347         * inchash.h (inchash::hash::add_wide_int): New function.
11348         * lto-streamer-out.c (hash_tree): Use it.
11350 2017-10-22  Richard Sandiford  <richard.sandiford@linaro.org>
11352         * inchash.h (inchash::hash::add_wide_int): Rename to...
11353         (inchash::hash::add_hwi): ...this.
11354         * ipa-devirt.c (hash_odr_vtable): Update accordingly.
11355         (polymorphic_call_target_hasher::hash): Likewise.
11356         * ipa-icf.c (sem_function::get_hash, sem_function::init): Likewise.
11357         (sem_item::add_expr, sem_item::add_type, sem_variable::get_hash)
11358         (sem_item_optimizer::update_hash_by_addr_refs): Likewise.
11359         * lto-streamer-out.c (hash_tree): Likewise.
11360         * optc-save-gen.awk: Likewise.
11361         * tree.c (add_expr): Likewise.
11363 2017-10-22  Uros Bizjak  <ubizjak@gmail.com>
11365         PR target/52451
11366         * config/i386/i386.c (ix86_fp_compare_mode): Return CCFPmode
11367         for ordered inequality comparisons even with TARGET_IEEE_FP.
11369 2017-10-22  Uros Bizjak  <ubizjak@gmail.com>
11371         PR target/82628
11372         * config/i386/i386.md (cmp<dwi>_doubleword): New pattern.
11373         * config/i386/i386.c (ix86_expand_branch) <case E_TImode>:
11374         Expand with cmp<dwi>_doubleword.
11376 2017-10-21  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
11378         * extend.texi: Add x86 specific to 'nocf_check' attribute.
11379         List CET intrinsics.
11380         * invoke.texi: Add -mcet, -mibt, -mshstk options.  Add x86
11381         specific to -fcf-protection option.
11383 2017-10-21  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
11385         * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET): New.
11386         (OPTION_MASK_ISA_SHSTK_SET): Likewise.
11387         (OPTION_MASK_ISA_IBT_UNSET): Likewise.
11388         (OPTION_MASK_ISA_SHSTK_UNSET): Likewise.
11389         (ix86_handle_option): Add -mibt, -mshstk, -mcet handling.
11390         * config.gcc (extra_headers): Add cetintrin.h for x86 targets.
11391         (extra_objs): Add cet.o for Linux/x86 targets.
11392         (tmake_file): Add i386/t-cet for Linux/x86 targets.
11393         * config/i386/cet.c: New file.
11394         * config/i386/cetintrin.h: Likewise.
11395         * config/i386/t-cet: Likewise.
11396         * config/i386/cpuid.h (bit_SHSTK): New.
11397         (bit_IBT): Likewise.
11398         * config/i386/driver-i386.c (host_detect_local_cpu): Detect and
11399         pass IBT and SHSTK bits.
11400         * config/i386/i386-builtin-types.def
11401         (VOID_FTYPE_UNSIGNED_PVOID): New.
11402         (VOID_FTYPE_UINT64_PVOID): Likewise.
11403         * config/i386/i386-builtin.def: Add CET intrinsics.
11404         * config/i386/i386-c.c (ix86_target_macros_internal): Add
11405         OPTION_MASK_ISA_IBT, OPTION_MASK_ISA_SHSTK handling.
11406         * config/i386/i386-passes.def: Add pass_insert_endbranch pass.
11407         * config/i386/i386-protos.h (make_pass_insert_endbranch): New
11408         prototype.
11409         * config/i386/i386.c (rest_of_insert_endbranch): New.
11410         (pass_data_insert_endbranch): Likewise.
11411         (pass_insert_endbranch): Likewise.
11412         (make_pass_insert_endbranch): Likewise.
11413         (ix86_notrack_prefixed_insn_p): Likewise.
11414         (ix86_target_string): Add -mibt, -mshstk flags.
11415         (ix86_option_override_internal): Add flag_cf_protection
11416         processing.
11417         (ix86_valid_target_attribute_inner_p): Set OPT_mibt, OPT_mshstk.
11418         (ix86_print_operand): Add 'notrack' prefix output.
11419         (ix86_init_mmx_sse_builtins): Add CET intrinsics.
11420         (ix86_expand_builtin): Expand CET intrinsics.
11421         (x86_output_mi_thunk): Add 'endbranch' instruction.
11422         * config/i386/i386.h (TARGET_IBT): New.
11423         (TARGET_IBT_P): Likewise.
11424         (TARGET_SHSTK): Likewise.
11425         (TARGET_SHSTK_P): Likewise.
11426         * config/i386/i386.md (unspecv): Add UNSPECV_NOP_RDSSP,
11427         UNSPECV_INCSSP, UNSPECV_SAVEPREVSSP, UNSPECV_RSTORSSP,
11428         UNSPECV_WRSS, UNSPECV_WRUSS, UNSPECV_SETSSBSY, UNSPECV_CLRSSBSY.
11429         (builtin_setjmp_setup): New pattern.
11430         (builtin_longjmp): Likewise.
11431         (rdssp<mode>): Likewise.
11432         (incssp<mode>): Likewise.
11433         (saveprevssp): Likewise.
11434         (rstorssp): Likewise.
11435         (wrss<mode>): Likewise.
11436         (wruss<mode>): Likewise.
11437         (setssbsy): Likewise.
11438         (clrssbsy): Likewise.
11439         (nop_endbr): Likewise.
11440         * config/i386/i386.opt: Add -mcet, -mibt, -mshstk and -mcet-switch
11441         options.
11442         * config/i386/immintrin.h: Include <cetintrin.h>.
11443         * config/i386/linux-common.h
11444         (file_end_indicate_exec_stack_and_cet): New prototype.
11445         (TARGET_ASM_FILE_END): New.
11447 2017-10-20  Jan Hubicka  <hubicka@ucw.cz>
11449         * i386.c (ix86_builtin_vectorization_cost): Use existing rtx_cost
11450         latencies instead of having separate table; make difference between
11451         integer and float costs.
11452         * i386.h (processor_costs): Remove scalar_stmt_cost,
11453         scalar_load_cost, scalar_store_cost, vec_stmt_cost, vec_to_scalar_cost,
11454         scalar_to_vec_cost, vec_align_load_cost, vec_unalign_load_cost,
11455         vec_store_cost.
11456         * x86-tune-costs.h: Remove entries which has been removed in
11457         procesor_costs from all tables; make cond_taken_branch_cost
11458         and cond_not_taken_branch_cost COST_N_INSNS based.
11460 2017-10-20  Jan Hubicka  <hubicka@ucw.cz>
11462         * x86-tune-costs.h (intel_cost, generic_cost): Fix move costs.
11464 2017-10-20  Jakub Jelinek  <jakub@redhat.com>
11466         * config/i386/i386.md (isa): Remove fma_avx512f.
11467         * config/i386/sse.md (<avx512>_fmadd_<mode>_mask<round_name>,
11468         <avx512>_fmadd_<mode>_mask3<round_name>,
11469         <avx512>_fmsub_<mode>_mask<round_name>,
11470         <avx512>_fmsub_<mode>_mask3<round_name>,
11471         <avx512>_fnmadd_<mode>_mask<round_name>,
11472         <avx512>_fnmadd_<mode>_mask3<round_name>,
11473         <avx512>_fnmsub_<mode>_mask<round_name>,
11474         <avx512>_fnmsub_<mode>_mask3<round_name>,
11475         <avx512>_fmaddsub_<mode>_mask<round_name>,
11476         <avx512>_fmaddsub_<mode>_mask3<round_name>,
11477         <avx512>_fmsubadd_<mode>_mask<round_name>,
11478         <avx512>_fmsubadd_<mode>_mask3<round_name>): Remove isa attribute.
11479         (*vec_widen_umult_even_v16si<mask_name>,
11480         *vec_widen_smult_even_v16si<mask_name>): Likewise.
11481         (<mask_codefor>avx512bw_dbpsadbw<mode><mask_name>): Likewise.
11483 2017-10-20  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
11485         * extend.texi: Add 'nocf_check' documentation.
11486         * gimple.texi: Add second parameter to
11487         gimple_build_call_from_tree.
11488         * invoke.texi: Add -fcf-protection documentation.
11489         * rtl.texi: Add REG_CALL_NOTRACK documenation.
11491 2017-10-20  Richard Biener  <rguenther@suse.de>
11493         PR tree-optimization/82473
11494         * tree-vect-loop.c (vectorizable_reduction): Properly get at
11495         the largest input type.
11497 2017-10-20  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
11499         * c-attribs.c (handle_nocf_check_attribute): New function.
11500         (c_common_attribute_table): Add 'nocf_check' handling.
11501         * gimple-parser.c: Add second argument NULL to
11502         gimple_build_call_from_tree.
11503         * attrib.c (comp_type_attributes): Check nocf_check attribute.
11504         * cfgexpand.c (expand_call_stmt): Set REG_CALL_NOCF_CHECK for
11505         call insn.
11506         * combine.c (distribute_notes): Add REG_CALL_NOCF_CHECK handling.
11507         * common.opt: Add fcf-protection flag.
11508         * emit-rtl.c (try_split): Add REG_CALL_NOCF_CHECK handling.
11509         * flag-types.h: Add enum cf_protection_level.
11510         * gimple.c (gimple_build_call_from_tree): Add second parameter.
11511         Add 'nocf_check' attribute propagation to gimple call.
11512         * gimple.h (gf_mask): Add GF_CALL_NOCF_CHECK.
11513         (gimple_build_call_from_tree): Update prototype.
11514         (gimple_call_nocf_check_p): New function.
11515         (gimple_call_set_nocf_check): Likewise.
11516         * gimplify.c: Add second argument to gimple_build_call_from_tree.
11517         * ipa-icf.c: Add nocf_check attribute in statement hash.
11518         * recog.c (peep2_attempt): Add REG_CALL_NOCF_CHECK handling.
11519         * reg-notes.def: Add REG_NOTE (CALL_NOCF_CHECK).
11520         * toplev.c (process_options): Add flag_cf_protection handling.
11522 2017-10-19  Jan Hubicka  <hubicka@ucw.cz>
11524         * x86-tune-costs.h (core_cost): Fix div, move and sqrt latencies.
11526 2017-10-20  Richard Biener  <rguenther@suse.de>
11528         PR tree-optimization/82603
11529         * tree-if-conv.c (predicate_mem_writes): Make sure to only
11530         remove false predicated stores.
11532 2017-10-20  Richard Biener  <rguenther@suse.de>
11534         * graphite-isl-ast-to-gimple.c
11535         (translate_isl_ast_to_gimple::graphite_copy_stmts_from_block):
11536         Remove return value and simplify, dump copied stmt after lhs
11537         adjustment.
11538         (translate_isl_ast_to_gimple::translate_isl_ast_node_user):
11539         Reduce dump verbosity.
11540         (gsi_insert_earliest): Likewise.
11541         (translate_isl_ast_to_gimple::copy_bb_and_scalar_dependences): Adjust.
11542         * graphite.c (print_global_statistics): Adjust dumping.
11543         (print_graphite_scop_statistics): Likewise.
11544         (print_graphite_statistics): Do not dump loops here.
11545         (graphite_transform_loops): But here.
11547 2017-10-20  Nicolas Roche  <roche@adacore.com>
11549         * configure.ac (ACX_PROG_GNAT): Append "libgnat" to include search dir.
11550         * configure: Regenerate.
11552 2017-10-20  Jakub Jelinek  <jakub@redhat.com>
11554         PR target/82158
11555         * tree-cfg.c (pass_warn_function_return::execute): In noreturn
11556         functions when optimizing replace GIMPLE_RETURN stmts with
11557         calls to __builtin_unreachable ().
11559         PR sanitizer/82595
11560         * config/gnu-user.h (LIBTSAN_EARLY_SPEC): Add libtsan_preinit.o
11561         for -fsanitize=thread link of executables.
11562         (LIBLSAN_EARLY_SPEC): Add liblsan_preinit.o for -fsanitize=leak
11563         link of executables.
11565         PR target/82370
11566         * config/i386/sse.md (VI248_AVX2, VI248_AVX512BW, VI248_AVX512BW_2):
11567         New mode iterators.
11568         (<shift_insn><mode>3<mask_name>): Change the last of the 3
11569         define_insns for logical vector shifts to use VI248_AVX512BW
11570         iterator instead of VI48_AVX512, remove <mask_mode512bit_condition>
11571         condition, useless isa and prefix attributes.  Change the first
11572         2 of these define_insns to ...
11573         (<mask_codefor><shift_insn><mode>3<mask_name>): ... this, new
11574         define_insn for avx512vl.
11575         (<shift_insn><mode>3): ... and this, new define_insn without
11576         masking for non-avx512vl.
11578         PR target/82370
11579         * config/i386/sse.md (*andnot<mode>3,
11580         <mask_codefor><code><mode>3<mask_name>, *<code><mode>3): Split
11581         (=v,v,vm) alternative into (=x,x,xm) and (=v,v,vm), for 128-bit
11582         and 256-bit vectors, the (=x,x,xm) alternative and when mask is
11583         not applied use empty suffix even for TARGET_AVX512VL.
11584         * config/i386/subst.md (mask_prefix3, mask_prefix4): When mask
11585         is applied, supply evex,evex or evex,evex,evex instead of just
11586         evex.
11588 2017-10-20  Julia Koval  <julia.koval@intel.com>
11590         * common/config/i386/i386-common.c (OPTION_MASK_ISA_GFNI_SET,
11591         (OPTION_MASK_ISA_GFNI_UNSET): New.
11592         (ix86_handle_option): Handle OPT_mgfni.
11593         * config/i386/cpuid.h (bit_GFNI): New.
11594         * config/i386/driver-i386.c (host_detect_local_cpu): Detect gfni.
11595         * config/i386/i386-c.c (ix86_target_macros_internal): Define __GFNI__.
11596         * config/i386/i386.c (ix86_target_string): Add -mgfni.
11597         (ix86_valid_target_attribute_inner_p): Add OPT_mgfni.
11598         * config/i386/i386.h (TARGET_GFNI, TARGET_GFNI_P): New.
11599         * config/i386/i386.opt: Add mgfni.
11601 2017-10-20  Orlando Arias  <oarias@knights.ucf.edu>
11603         * config/msp430/msp430.c (msp430_option_override): Disable
11604         -fdelete-null-pointer-checks.
11605         * doc/invoke.text (-fdelete-null-pointer-checks): Document that.
11607 2017-10-19  Jan Hubicka  <hubicka@ucw.cz>
11609         * x86-tune-costs.h (generic_cost, core_cost): Correct costs
11610         of x87 and SSE instructions.
11612 2017-10-19  Jan Hubicka  <hubicka@ucw.cz>
11614         * asan.c (create_cond_insert_point): Do not update edge count.
11615         * auto-profile.c (afdo_propagate_edge): Update for edge count removal.
11616         (afdo_propagate_circuit): Likewise.
11617         (afdo_calculate_branch_prob): Likewise.
11618         (afdo_annotate_cfg): Likewise.
11619         * basic-block.h (struct edge_def): Remove count.
11620         (edge_def::count): New accessor.
11621         * bb-reorder.c (rotate_loop): Update.
11622         (find_traces_1_round): Update.
11623         (connect_traces): Update.
11624         (sanitize_hot_paths): Update.
11625         * cfg.c (unchecked_make_edge): Update.
11626         (make_single_succ_edge): Update.
11627         (check_bb_profile): Update.
11628         (dump_edge_info): Update.
11629         (update_bb_profile_for_threading): Update.
11630         (scale_bbs_frequencies_int): Update.
11631         (scale_bbs_frequencies_gcov_type): Update.
11632         (scale_bbs_frequencies_profile_count): Update.
11633         (scale_bbs_frequencies): Update.
11634         * cfganal.c (connect_infinite_loops_to_exit): Update.
11635         * cfgbuild.c (compute_outgoing_frequencies): Update.
11636         (find_many_sub_basic_blocks): Update.
11637         * cfgcleanup.c (try_forward_edges): Update.
11638         (try_crossjump_to_edge): Update
11639         * cfgexpand.c (expand_gimple_cond): Update
11640         (expand_gimple_tailcall): Update
11641         (construct_exit_block): Update
11642         * cfghooks.c (verify_flow_info): Update
11643         (redirect_edge_succ_nodup): Update
11644         (split_edge): Update
11645         (make_forwarder_block): Update
11646         (duplicate_block): Update
11647         (account_profile_record): Update
11648         * cfgloop.c (find_subloop_latch_edge_by_profile): Update.
11649         * cfgloopanal.c (expected_loop_iterations_unbounded): Update.
11650         * cfgloopmanip.c (scale_loop_profile): Update.
11651         (loopify): Update.
11652         (lv_adjust_loop_entry_edge): Update.
11653         * cfgrtl.c (try_redirect_by_replacing_jump): Update.
11654         (force_nonfallthru_and_redirect): Update.
11655         (purge_dead_edges): Update.
11656         (rtl_flow_call_edges_add): Update.
11657         * cgraphunit.c (init_lowered_empty_function): Update.
11658         (cgraph_node::expand_thunk): Update.
11659         * gimple-pretty-print.c (dump_probability): Update.
11660         (dump_edge_probability): Update.
11661         * gimple-ssa-isolate-paths.c (isolate_path): Update.
11662         * haifa-sched.c (sched_create_recovery_edges): Update.
11663         * hsa-gen.c (convert_switch_statements): Update.
11664         * ifcvt.c (dead_or_predicable): Update.
11665         * ipa-inline-transform.c (inline_transform): Update.
11666         * ipa-split.c (split_function): Update.
11667         * ipa-utils.c (ipa_merge_profiles): Update.
11668         * loop-doloop.c (add_test): Update.
11669         * loop-unroll.c (unroll_loop_runtime_iterations): Update.
11670         * lto-streamer-in.c (input_cfg): Update.
11671         (input_function): Update.
11672         * lto-streamer-out.c (output_cfg): Update.
11673         * modulo-sched.c (sms_schedule): Update.
11674         * postreload-gcse.c (eliminate_partially_redundant_load): Update.
11675         * predict.c (maybe_hot_edge_p): Update.
11676         (unlikely_executed_edge_p): Update.
11677         (probably_never_executed_edge_p): Update.
11678         (dump_prediction): Update.
11679         (drop_profile): Update.
11680         (propagate_unlikely_bbs_forward): Update.
11681         (determine_unlikely_bbs): Update.
11682         (force_edge_cold): Update.
11683         * profile.c (compute_branch_probabilities): Update.
11684         * reg-stack.c (better_edge): Update.
11685         * shrink-wrap.c (handle_simple_exit): Update.
11686         * tracer.c (better_p): Update.
11687         * trans-mem.c (expand_transaction): Update.
11688         (split_bb_make_tm_edge): Update.
11689         * tree-call-cdce.c: Update.
11690         * tree-cfg.c (gimple_find_sub_bbs): Update.
11691         (gimple_split_edge): Update.
11692         (gimple_duplicate_sese_region): Update.
11693         (gimple_duplicate_sese_tail): Update.
11694         (gimple_flow_call_edges_add): Update.
11695         (insert_cond_bb): Update.
11696         (execute_fixup_cfg): Update.
11697         * tree-cfgcleanup.c (cleanup_control_expr_graph): Update.
11698         * tree-complex.c (expand_complex_div_wide): Update.
11699         * tree-eh.c (lower_resx): Update.
11700         (unsplit_eh): Update.
11701         (cleanup_empty_eh_move_lp): Update.
11702         * tree-inline.c (copy_edges_for_bb): Update.
11703         (freqs_to_counts): Update.
11704         (copy_cfg_body): Update.
11705         * tree-ssa-dce.c (remove_dead_stmt): Update.
11706         * tree-ssa-ifcombine.c (update_profile_after_ifcombine): Update.
11707         * tree-ssa-loop-im.c (execute_sm_if_changed): Update.
11708         * tree-ssa-loop-ivcanon.c (remove_exits_and_undefined_stmts): Update.
11709         (unloop_loops): Update.
11710         * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Update.
11711         * tree-ssa-loop-split.c (connect_loops): Update.
11712         (split_loop): Update.
11713         * tree-ssa-loop-unswitch.c (hoist_guard): Update.
11714         * tree-ssa-phionlycprop.c (propagate_rhs_into_lhs): Update.
11715         * tree-ssa-phiopt.c (replace_phi_edge_with_variable): Update.
11716         * tree-ssa-reassoc.c (branch_fixup): Update.
11717         * tree-ssa-tail-merge.c (replace_block_by): Update.
11718         * tree-ssa-threadupdate.c (remove_ctrl_stmt_and_useless_edges): Update.
11719         (compute_path_counts): Update.
11720         (update_profile): Update.
11721         (recompute_probabilities): Update.
11722         (update_joiner_offpath_counts): Update.
11723         (estimated_freqs_path): Update.
11724         (freqs_to_counts_path): Update.
11725         (clear_counts_path): Update.
11726         (ssa_fix_duplicate_block_edges): Update.
11727         (duplicate_thread_path): Update.
11728         * tree-switch-conversion.c (hoist_edge_and_branch_if_true): Update.
11729         (case_bit_test_cmp): Update.
11730         (collect_switch_conv_info): Update.
11731         (gen_inbound_check): Update.
11732         (do_jump_if_equal): Update.
11733         (emit_cmp_and_jump_insns): Update.
11734         * tree-tailcall.c (decrease_profile): Update.
11735         (eliminate_tail_call): Update.
11736         * tree-vect-loop-manip.c (slpeel_add_loop_guard): Update.
11737         (vect_do_peeling): Update.
11738         * tree-vect-loop.c (scale_profile_for_vect_loop): Update.
11739         * ubsan.c (ubsan_expand_null_ifn): Update.
11740         (ubsan_expand_ptr_ifn): Update.
11741         * value-prof.c (gimple_divmod_fixed_value): Update.
11742         (gimple_mod_pow2): Update.
11743         (gimple_mod_subtract): Update.
11744         (gimple_ic): Update.
11745         (gimple_stringop_fixed_value): Update.
11747 2017-10-19  Uros Bizjak  <ubizjak@gmail.com>
11749         PR target/82618
11750         * config/i386/i386.md (sub to cmp): New peephole2 pattern.
11752 2017-10-19  Alexander Monakov  <amonakov@ispras.ru>
11754         PR rtl-optimization/82395
11755         * ira-color.c (allocno_priority_compare_func): Fix comparison step
11756         based on non_spilled_static_chain_regno_p.
11758 2017-10-19  Uros Bizjak  <ubizjak@gmail.com>
11760         * config/i386/i386.c (output_387_binary_op): Rewrite SSE part.
11761         (ix86_emit_mode_set): Rewrite insn mnemonic construction.
11762         (ix86_prepare_fp_compare_args): Redefine is_sse as bool.
11764 2017-10-19  Martin Sebor  <msebor@redhat.com>
11766         PR tree-optimization/82596
11767         * tree.c (array_at_struct_end_p): Handle STRING_CST.
11769 2017-10-19  Eric Botcazou  <ebotcazou@adacore.com>
11771         * asan.c (handle_builtin_alloca): Deal with all alloca variants.
11772         (get_mem_refs_of_builtin_call): Likewise.
11773         * builtins.c (expand_builtin_apply): Adjust call to
11774         allocate_dynamic_stack_space.
11775         (expand_builtin_alloca): For __builtin_alloca_with_align_and_max, pass
11776         the third argument to allocate_dynamic_stack_space, otherwise -1.
11777         (expand_builtin): Deal with all alloca variants.
11778         (is_inexpensive_builtin): Likewise.
11779         * builtins.def (BUILT_IN_ALLOCA_WITH_ALIGN_AND_MAX): New.
11780         * calls.c (special_function_p): Deal with all alloca variants.
11781         (initialize_argument_information): Adjust call to
11782         allocate_dynamic_stack_space.
11783         (expand_call): Likewise.
11784         * cfgexpand.c (expand_call_stmt): Deal with all alloca variants.
11785         * doc/extend.texi (Built-ins): Add __builtin_alloca_with_align_and_max
11786         * explow.c (allocate_dynamic_stack_space): Add MAX_SIZE parameter and
11787         use it for the stack usage computation.
11788         * explow.h (allocate_dynamic_stack_space): Adjust prototype.
11789         * function.c (gimplify_parameters): Call build_alloca_call_expr.
11790         * gimple-ssa-warn-alloca.c (alloca_call_type): Simplify control flow.
11791         Take into account 3rd argument of __builtin_alloca_with_align_and_max.
11792         (in_loop_p): Remove first argument and useless check.
11793         (pass_walloca::execute): Remove useless test and adjust call to above.
11794         * gimple.c (gimple_build_call_from_tree): Deal with all alloc variants
11795         * gimplify.c (gimplify_vla_decl): Call build_alloca_call_expr.
11796         (gimplify_call_expr): Deal with all alloca variants.
11797         * hsa-gen.c (gen_hsa_alloca): Likewise.
11798         (gen_hsa_insns_for_call): Likewise.
11799         * ipa-pure-const.c (special_builtin_state): Likewise.
11800         * tree-chkp.c (chkp_build_returned_bound): Likewise.
11801         * tree-object-size.c (alloc_object_size): Likewise.
11802         * tree-ssa-alias.c (ref_maybe_used_by_call_p_1): Likewise.
11803         (call_may_clobber_ref_p_1): Likewise.
11804         * tree-ssa-ccp.c (evaluate_stmt): Likewise.
11805         (ccp_fold_stmt): Likewise.
11806         (optimize_stack_restore): Likewise.
11807         * tree-ssa-dce.c (mark_stmt_if_obviously_necessary): Likewise.
11808         (mark_all_reaching_defs_necessary_1): Likewise.
11809         (propagate_necessity): Likewise.
11810         (eliminate_unnecessary_stmts): Likewise.
11811         * tree.c (build_common_builtin_nodes): Build
11812         BUILT_IN_ALLOCA_WITH_ALIGN_AND_MAX.
11813         (build_alloca_call_expr): New function.
11814         * tree.h (ALLOCA_FUNCTION_CODE_P): New macro.
11815         (CASE_BUILT_IN_ALLOCA): Likewise.
11816         (build_alloca_call_expr): Declare.
11817         * varasm.c (incorporeal_function_p): Deal with all alloca variants.
11819 2017-10-19  Eric Botcazou  <ebotcazou@adacore.com>
11821         PR debug/82509
11822         * dwarf2out.c (new_die_raw): New static inline function.
11823         (new_die): Use it to create the DIE.
11824         (add_AT_external_die_ref): Likewise.
11825         (clone_die): Likewise.
11826         (clone_as_declaration): Likewise.
11827         (dwarf2out_vms_debug_main_pointer): Likewise.
11828         (base_type_die): Likewise.  Remove early return for corner cases.
11829         Do not call add_pubtype on the DIE here.
11830         (is_base_type): Remove ERROR_MARK and return 0 for VOID_TYPE.
11831         (modified_type_die): Adjust the lookup for reverse order DIEs.  Skip
11832         typedefs for base types with DW_AT_endianity.  Make sure a DIE with
11833         native order exists for base types, attach the DIE manually and call
11834         add_pubtype on it.  Do not equate a reverse order DIE to the type.
11836 2017-10-19  Richard Earnshaw  <rearnsha@arm.com>
11838         * config/arm/arm.c (align_ok_ldrd_strd): New function.
11839         (mem_ok_for_ldrd_strd): New parameter align.  Extract the alignment of
11840         the mem into it.
11841         (gen_operands_ldrd_strd): Validate the alignment of the accesses.
11843 2017-10-19  Jakub Jelinek  <jakub@redhat.com>
11845         * flag-types.h (enum sanitize_code): Add SANITIZE_BUILTIN.  Or
11846         SANITIZE_BUILTIN into SANITIZE_UNDEFINED.
11847         * sanitizer.def (BUILT_IN_UBSAN_HANDLE_INVALID_BUILTIN,
11848         BUILT_IN_UBSAN_HANDLE_INVALID_BUILTIN_ABORT): New builtins.
11849         * opts.c (sanitizer_opts): Add builtin.
11850         * ubsan.c (instrument_builtin): New function.
11851         (pass_ubsan::execute): Call it.
11852         (pass_ubsan::gate): Enable even for SANITIZE_BUILTIN.
11853         * doc/invoke.texi: Document -fsanitize=builtin.
11855         * ubsan.c (ubsan_expand_null_ifn): Use _v1 suffixed type mismatch
11856         builtins, store max (log2 (align), 0) into uchar field instead of
11857         align into uptr field.
11858         (ubsan_expand_objsize_ifn): Use _v1 suffixed type mismatch builtins,
11859         store uchar 0 field instead of uptr 0 field.
11860         (instrument_nonnull_return): Use _v1 suffixed nonnull return builtin,
11861         instead of passing one address of struct with 2 locations pass
11862         two addresses of structs with 1 location each.
11863         * sanitizer.def (BUILT_IN_UBSAN_HANDLE_TYPE_MISMATCH,
11864         BUILT_IN_UBSAN_HANDLE_TYPE_MISMATCH_ABORT,
11865         BUILT_IN_UBSAN_HANDLE_NONNULL_RETURN,
11866         BUILT_IN_UBSAN_HANDLE_NONNULL_RETURN_ABORT): Removed.
11867         (BUILT_IN_UBSAN_HANDLE_TYPE_MISMATCH_V1,
11868         BUILT_IN_UBSAN_HANDLE_TYPE_MISMATCH_V1_ABORT,
11869         BUILT_IN_UBSAN_HANDLE_NONNULL_RETURN_V1,
11870         BUILT_IN_UBSAN_HANDLE_NONNULL_RETURN_V1_ABORT): New builtins.
11872 2017-10-19  Martin Liska  <mliska@suse.cz>
11874         PR driver/81829
11875         * file-find.c (remove_prefix): Remove.
11876         * file-find.h (remove_prefix): Likewise.
11877         * gcc-ar.c: Remove smartness of lookup.
11879 2017-10-19  Segher Boessenkool  <segher@kernel.crashing.org>
11881         * config/rs6000/rs6000.md (*call_indirect_aix<mode>,
11882         *call_value_indirect_aix<mode>, *call_indirect_elfv2<mode>,
11883         *call_value_indirect_elfv2<mode>): Add correct mode to the unspec.
11885 2017-10-19  Jakub Jelinek  <jakub@redhat.com>
11887         PR target/82580
11888         * config/i386/i386.md (setcc + movzbl to xor + setcc): New peephole2.
11889         (setcc + and to xor + setcc): New peephole2.
11891 2017-10-19  Tom de Vries  <tom@codesourcery.com>
11893         * doc/sourcebuild.texi (Test Directives, Variants of
11894         dg-require-support): Add dg-require-stack-size.
11896 2017-10-19  Martin Liska  <mliska@suse.cz>
11898         PR sanitizer/82517
11899         * gimplify.c (gimplify_decl_expr): Do not instrument variables
11900         that have a large alignment.
11901         (gimplify_target_expr): Likewise.
11903 2017-10-18  Segher Boessenkool  <segher@kernel.crashing.org>
11905         PR rtl-optimization/82602
11906         * ira.c (rtx_moveable_p): Return false for volatile asm.
11908 2017-10-18  Uros Bizjak  <ubizjak@gmail.com>
11910         PR target/82580
11911         * config/i386/i386-modes.def (CCGZ): New CC mode.
11912         * config/i386/i386.md (sub<mode>3_carry_ccgz): New insn pattern.
11913         * config/i386/predicates.md (ix86_comparison_operator):
11914         Handle CCGZmode.
11915         * config/i386/i386.c (ix86_expand_branch) <case E_TImode>:
11916         Emulate LE, LEU, GT, GTU, LT, LTU, GE and GEU double-word comparisons
11917         with double-word subtraction.
11918         (put_condition_code): Handle CCGZmode.
11920 2017-10-18  Aldy Hernandez  <aldyh@redhat.com>
11922         * wide-int.cc (debug (const wide_int &)): New.
11923         (debug (const wide_int *)): New.
11924         (debug (const widest_int &)): New.
11925         (debug (const widest_int *)): New.
11927 2017-10-18  Vladimir Makarov  <vmakarov@redhat.com>
11929         PR middle-end/82556
11930         * lra-constraints.c (curr_insn_transform): Use non-input operand
11931         instead of output one for matched reload.
11933 2017-10-18  Bin Cheng  <bin.cheng@arm.com>
11935         * tree-loop-distribution.c (INCLUDE_ALGORITHM): New header file.
11936         (tree-ssa-loop-ivopts.h): New header file.
11937         (struct builtin_info): New fields.
11938         (classify_builtin_1): Compute and record base and offset parts for
11939         memset builtin partition by calling strip_offset.
11940         (offset_cmp, fuse_memset_builtins): New functions.
11941         (finalize_partitions): Fuse adjacent memset partitions by calling
11942         above function.
11943         * tree-ssa-loop-ivopts.c (strip_offset): Delete static declaration.
11944         Expose the interface.
11945         * tree-ssa-loop-ivopts.h (strip_offset): New declaration.
11947 2017-10-18  Bin Cheng  <bin.cheng@arm.com>
11949         PR tree-optimization/82574
11950         * tree-loop-distribution.c (find_single_drs): New parameter.  Check
11951         that data reference must be executed exactly once per iteration
11952         against the outermost loop in nest.
11953         (classify_partition): Update call to above function.
11955 2017-10-18  Richard Biener  <rguenther@suse.de>
11957         PR tree-optimization/82591
11958         * graphite.c (graphite_transform_loops): Move code gen message
11959         printing ...
11960         * graphite-isl-ast-to-gimple.c (graphite_regenerate_ast_isl):
11961         Here.  Handle scop_to_isl_ast failing.
11962         (scop_to_isl_ast): Limit the number of ISL operations.
11964 2017-10-18  Richard Biener  <rguenther@suse.de>
11966         * graphite-isl-ast-to-gimple.c
11967         (translate_isl_ast_to_gimple::set_rename): Simplify.
11968         (translate_isl_ast_to_gimple::set_rename_for_each_def): Inline...
11969         (graphite_copy_stmts_from_block): ... here.
11970         (copy_bb_and_scalar_dependences): Simplify.
11971         (add_parameters_to_ivs_params): Canonicalize.
11972         (generate_entry_out_of_ssa_copies): Simplify.
11973         * graphite-sese-to-poly.c (extract_affine_name): Simplify
11974         by passing in ISL dimension.
11975         (parameter_index_in_region_1): Rename to ...
11976         (parameter_index_in_region): ... this.
11977         (extract_affine): Adjust assert, pass down parameter index.
11978         (add_param_constraints): Use range-info when available.
11979         (build_scop_context): Adjust.
11980         * sese.c (new_sese_info): Adjust.
11981         (free_sese_info): Likewise.
11982         * sese.h (bb_map_t, rename_map_t, phi_rename, init_back_edge_pair_t):
11983         Remove unused typedefs.
11984         (struct sese_info_t): Simplify rename_map, remove incomplete_phis.
11986 2017-10-18  Martin Liska  <mliska@suse.cz>
11988         * combine.c (simplify_compare_const): Add gcc_fallthrough.
11990 2017-10-18  Robin Dapp  <rdapp@linux.vnet.ibm.com>
11992         * config/s390/s390.c (s390_bb_fallthru_entry_likely): New function.
11993         (s390_sched_init): Do not reset s390_sched_state if we entered the
11994         current basic block via a fallthru edge and all others are unlikely.
11996 2017-10-18  Robin Dapp  <rdapp@linux.vnet.ibm.com>
11998         * config/s390/s390.c (NUM_SIDES): New variable.
11999         (LONGRUNNING_THRESHOLD): New variable.
12000         (LATENCY_FACTOR): New variable.
12001         (s390_sched_score): Decrease score for long-running instructions on
12002         wrong side.
12003         (s390_sched_variable_issue): Perform bookkeeping for long-running
12004         instructions.
12006 2017-10-18  Richard Biener  <rguenther@suse.de>
12008         * graphite-isl-ast-to-gimple.c (gcc_expression_from_isl_ast_expr_id):
12009         Simplify with removal of the parameter rename map.
12010         (set_rename): Likewise.
12011         (should_copy_to_new_region): Likewise.
12012         (graphite_copy_stmts_from_block): Likewise.
12013         (copy_bb_and_scalar_dependences): Remove initialization of
12014         unused copied_bb_map.
12015         (copy_def): Remove.
12016         (copy_internal_parameters): Likewise.
12017         (graphite_regenerate_ast_isl): Do not call copy_internal_parameters.
12018         * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
12019         Use INTEGRAL_TYPE_P.
12020         (parameter_index_in_region_1): Rename to ...
12021         (assign_parameter_index_in_region): ... this.  Assert we have
12022         a parameter we handle.
12023         (scan_tree_for_params): Adjust.
12024         * sese.h (parameter_rename_map_t): Remove.
12025         (struct sese_info_t): Remove unused parameter_rename_map and
12026         copied_bb_map members.
12027         * sese.c (new_sese_info): Adjust.
12028         (free_sese_info): Likewise.
12030 2017-10-18  Martin Liska  <mliska@suse.cz>
12032         PR sanitizer/82545
12033         * asan.c (asan_expand_poison_ifn): Do not put gimple stmt
12034         on an abnormal edge.
12036 2017-10-18  Sebastian Huber  <sebastian.huber@embedded-brains.de>
12038         * doc/invoke.texi (ffunction-sections and fdata-sections):
12039         Update.
12041 2017-10-17  Eric Botcazou  <ebotcazou@adacore.com>
12043         * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Bail out only if
12044         the use statement can throw internally.
12046 2017-10-17  Eric Botcazou  <ebotcazou@adacore.com>
12048         * config/visium/visium.c (visium_select_cc_mode): Return CCmode for
12049         any RTX present on the RHS of a SET.
12050         * compare-elim.c (try_eliminate_compare): Restore comment.
12052 2017-10-17  Jakub Jelinek  <jakub@redhat.com>
12054         * langhooks.h (struct lang_hooks): Document that tree_size langhook
12055         may be also called on tcc_type nodes.
12056         * langhooks.c (lhd_tree_size): Likewise.
12058 2017-10-17  David Malcolm  <dmalcolm@redhat.com>
12060         * gimple-ssa-sprintf.c (fmtwarn): Update for changed signature of
12061         format_warning_at_substring.
12062         (maybe_warn): Convert source_range * param to a location_t.  Pass
12063         UNKNOWN_LOCATION rather than NULL to fmtwarn.
12064         (format_directive): Remove code to extract source_ranges and
12065         source_range * in favor of just a location_t.
12066         (parse_directive): Pass UNKNOWN_LOCATION rather than NULL to
12067         fmtwarn.
12068         * substring-locations.c (format_warning_va): Convert
12069         source_range * param to a location_t.
12070         (format_warning_at_substring): Likewise.
12071         * substring-locations.h (format_warning_va): Likewise.
12072         (format_warning_at_substring): Likewise.
12074 2017-10-17  Jan Hubicka  <hubicka@ucw.cz>
12076         * target.h (enum vect_cost_for_stmt): Add vec_gather_load and
12077         vec_scatter_store
12078         * tree-vect-stmts.c (record_stmt_cost): Make difference between normal
12079         and scatter/gather ops.
12081         * aarch64/aarch64.c (aarch64_builtin_vectorization_cost): Add
12082         vec_gather_load and vec_scatter_store.
12083         * arm/arm.c (arm_builtin_vectorization_cost): Likewise.
12084         * powerpcspe/powerpcspe.c (rs6000_builtin_vectorization_cost): Likewise.
12085         * rs6000/rs6000.c (rs6000_builtin_vectorization_cost): Likewise.
12086         * s390/s390.c (s390_builtin_vectorization_cost): Likewise.
12087         * spu/spu.c (spu_builtin_vectorization_cost): Likewise.
12088         * i386/i386.c (x86_builtin_vectorization_cost): Likewise.
12090 2017-10-17  Uros Bizjak  <ubizjak@gmail.com>
12092         * reg-stack.c (compare_for_stack_reg): Add bool argument.
12093         Detect FTST instruction and handle its register pops.  Only pop
12094         second operand if can_pop_second_op is true.
12095         (subst_stack_regs_pat) <case COMPARE>: Detect FCOMI instruction to
12096         set can_pop_second_op to false in the compare_for_stack_reg call.
12098         * config/i386/i386.md (*cmpi<FPCMP:unord><MODEF:mode>): Only call
12099         output_fp_compare for stack register operands.
12100         * config/i386/i386.c (output_fp_compare): Do not output SSE compare
12101         instructions here.  Do not emit stack register pops here.  Assert
12102         that FCOMPP pops next to top stack register.  Rewrite function.
12104 2017-10-17  Nathan Sidwell  <nathan@acm.org>
12106         PR middle-end/82577
12107         * alias.c (compare_base_decls): Check HAS_DECL_ASSEMBLER_NAME_P,
12108         use DECL_ASSEMBLER_NAME_RAW.
12110         PR middle-end/82546
12111         * tree.c (tree_code_size): Reformat.  Punt to lang hook for unknown
12112         TYPE nodes.
12114 2017-10-17  Qing Zhao <qing.zhao@oracle.com>
12115             Wilco Dijkstra <wilco.dijkstra@arm.com>
12117         * builtins.c (expand_builtin_update_setjmp_buf): Add a
12118         converstion to Pmode from the buf_addr.
12120 2017-10-17  Richard Biener  <rguenther@suse.de>
12122         * graphite-dependences.c (scop_get_reads_and_writes): Change
12123         output parameters to references.
12125 2017-10-17  Jackson Woodruff  <jackson.woodruff@arm.com>
12127         PR 71026/tree-optimization
12128         * fold-const.c (distribute_real_division): Removed.
12129         (fold_binary_loc): Remove calls to distribute_real_divison.
12131 2017-10-17  Richard Biener  <rguenther@suse.de>
12133         * graphite-scop-detection.c
12134         (scop_detection::stmt_has_simple_data_refs_p): Always use
12135         the full nest as region.
12136         (try_generate_gimple_bb): Likewise.
12137         * sese.c (scalar_evolution_in_region): Simplify now that
12138         SCEV can handle instantiation in regions.
12139         * tree-scalar-evolution.c (instantiate_scev_name): Also instantiate
12140         in the non-loop part of a function if requested.
12142 2017-10-17  Richard Biener  <rguenther@suse.de>
12144         PR tree-optimization/82563
12145         * graphite-isl-ast-to-gimple.c (generate_entry_out_of_ssa_copies):
12146         New function.
12147         (graphite_regenerate_ast_isl): Call it.
12148         * graphite-scop-detection.c (build_scops): Remove entry edge split.
12150 2017-10-17  Jakub Jelinek  <jakub@redhat.com>
12152         PR tree-optimization/82549
12153         * fold-const.c (optimize_bit_field_compare, fold_truth_andor_1):
12154         Formatting fixes.  Instead of calling make_bit_field_ref with negative
12155         bitpos return 0.
12157 2017-10-17  Olga Makhotina  <olga.makhotina@intel.com>
12159         * config/i386/avx512dqintrin.h (_mm_mask_reduce_sd,
12160         _mm_maskz_reduce_sd, _mm_mask_reduce_ss,=20
12161         _mm_maskz_reduce_ss): New.
12162         * config/i386/i386-builtin.def (__builtin_ia32_reducesd_mask,
12163         __builtin_ia32_reducess_mask): Ditto..
12164         (__builtin_ia32_reducesd, __builtin_ia32_reducess): Remove.
12165         * config/i386/sse.md (reduces<mode>): Renamed to ...
12166         (reduces<mode><mask_scalar_name>): ... this.
12167         (vreduce<ssescalarmodesuffix>\t{%3, %2, %1, %0|%0, %1, %2, %3}):
12168         Changed to ...
12169         (vreduce<ssescalarmodesuffix>\t{%3, %2, %1, %0<mask_scalar_operand4>|
12170         %0<mask_scalar_operand4>, %1, %2, %3}): ... this.
12172 2017-10-16  David Malcolm  <dmalcolm@redhat.com>
12174         * Makefile.in (OBJS): Add unique-ptr-tests.o.
12175         * selftest-run-tests.c (selftest::run_tests): Call
12176         selftest::unique_ptr_tests_cc_tests.
12177         * selftest.h (selftest::unique_ptr_tests_cc_tests): New decl.
12178         * unique-ptr-tests.cc: New file.
12180 2017-10-16  Vladimir Makarov  <vmakarov@redhat.com>
12182         PR sanitizer/82353
12183         * lra.c (collect_non_operand_hard_regs): Don't ignore operator
12184         locations.
12185         * lra-lives.c (bb_killed_pseudos, bb_gen_pseudos): Move up.
12186         (make_hard_regno_born, make_hard_regno_dead): Update
12187         bb_killed_pseudos and bb_gen_pseudos for fixed regs.
12189 2017-10-16  Jeff Law  <law@redhat.com>
12191         * tree-ssa-dse.c (live_bytes_read): Fix thinko.
12193 2017-10-16  Jan Hubicka  <hubicka@ucw.cz>
12195         * x86-tune-costs.h (znver1_cost): Fix move cost tables.
12197 2017-10-16  Olivier Hainque  <hainque@adacore.com>
12199         * gcc/config.gcc (powerpc*-*-*spe*): Pick 8548 as the default
12200         with_cpu if we were configured for an e500v2 target cpu name.
12202 2017-10-16  Thomas Preud'homme  <thomas.preudhomme@arm.com>
12204         * config/arm/arm-cpus.in (cortex-m33): Add nodsp option.
12205         * doc/invoke.texi: Document +nodsp as a valid extension for
12206         -mcpu=cortex-m33.
12208 2017-10-16  Martin Liska  <mliska@suse.cz>
12210         * sbitmap.c (bitmap_bit_in_range_p_checking): New function.
12211         (test_set_range): Likewise.
12212         (test_range_functions): Rename to ...
12213         (test_bit_in_range): ... this.
12214         (sbitmap_c_tests): Add new test.
12216 2017-10-16  Tamar Christina  <tamar.christina@arm.com>
12218         * config/aarch64/arm_neon.h (vdot_u32, vdotq_u32, vdot_s32, vdotq_s32):
12219         New.
12220         (vdot_lane_u32, vdot_laneq_u32, vdotq_lane_u32, vdotq_laneq_u32): New.
12221         (vdot_lane_s32, vdot_laneq_s32, vdotq_lane_s32, vdotq_laneq_s32): New.
12223 2017-10-16  Tamar Christina  <tamar.christina@arm.com>
12225         * config/aarch64/aarch64-builtins.c
12226         (aarch64_types_quadopu_lane_qualifiers): New.
12227         (TYPES_QUADOPU_LANE): New.
12228         * config/aarch64/aarch64-simd.md (aarch64_<sur>dot<vsi2qi>): New.
12229         (<sur>dot_prod<vsi2qi>, aarch64_<sur>dot_lane<vsi2qi>): New.
12230         (aarch64_<sur>dot_laneq<vsi2qi>): New.
12231         * config/aarch64/aarch64-simd-builtins.def (sdot, udot): New.
12232         (sdot_lane, udot_lane, sdot_laneq, udot_laneq): New.
12233         * config/aarch64/iterators.md (sur): Add UNSPEC_SDOT, UNSPEC_UDOT.
12234         (Vdottype, DOTPROD): New.
12235         (sur): Add SDOT and UDOT.
12237 2017-10-16  Tamar Christina  <tamar.christina@arm.com>
12239         * config/aarch64/aarch64.h (AARCH64_FL_DOTPROD): New.
12240         (AARCH64_ISA_DOTPROD, TARGET_DOTPROD): New.
12241         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
12242         Add TARGET_DOTPROD.
12243         * config/aarch64/aarch64-option-extensions.def (dotprod): New.
12244         * config/aarch64/aarch64-cores.def (cortex-a55, cortex-a75):
12245         Enable TARGET_DOTPROD.
12246         (cortex-a75.cortex-a55): Likewise.
12247         * doc/invoke.texi (aarch64-feature-modifiers): Document dotprod.
12249 2017-10-16  Tamar Christina  <tamar.christina@arm.com>
12251         * config/arm/arm-builtins.c (arm_unsigned_uternop_qualifiers): New.
12252         (UTERNOP_QUALIFIERS, arm_umac_lane_qualifiers, UMAC_LANE_QUALIFIERS):
12253         New.
12254         * config/arm/arm_neon_builtins.def (sdot, udot, sdot_lane, udot_lane):
12255         New.
12256         * config/arm/iterators.md (DOTPROD, VSI2QI, vsi2qi): New.
12257         (UNSPEC_DOT_S, UNSPEC_DOT_U, opsuffix): New.
12258         * config/arm/neon.md (neon_<sup>dot<vsi2qi>): New.
12259         (neon_<sup>dot_lane<vsi2qi>, <sup>dot_prod<vsi2qi>): New.
12260         * config/arm/types.md (neon_dot, neon_dot_q): New.
12261         * config/arm/unspecs.md (sup): Add UNSPEC_DOT_S, UNSPEC_DOT_U.
12263 2017-10-16  Tamar Christina  <tamar.christina@arm.com>
12265         * config/arm/arm.h (TARGET_DOTPROD): New.
12266         * config/arm/arm.c (arm_arch_dotprod): New.
12267         (arm_option_reconfigure_globals): Add arm_arch_dotprod.
12268         * config/arm/arm-c.c (__ARM_FEATURE_DOTPROD): New.
12269         * config/arm/arm-cpus.in (armv8.2-a): Enabled +dotprod.
12270         (feature dotprod, group dotprod, ALL_SIMD_INTERNAL): New.
12271         (ALL_FPU_INTERNAL): Use ALL_SIMD_INTERNAL.
12272         * config/arm/t-multilib (v8_2_a_simd_variants): Add dotprod.
12273         * doc/invoke.texi (armv8.2-a): Document dotprod
12275 2017-10-14  Jan Hubicka  <hubicka@ucw.cz>
12277         * i386.c (ix86_vec_cost): New function.
12278         (ix86_rtx_costs): Handle vector operations better.
12279         * i386.h (struct processor_costs): Add sse_op, fmasd, fmass.
12280         * x86-tune-costs.h: Add new costs to all tables.
12282 2017-10-14  Jan Hubicka  <hubicka@ucw.cz>
12284         * i386.c (ix86_rtx_costs): Make difference between x87 and SSE
12285         operations.
12286         * i386.h (struct processor_costs): Add addss, mulss, mulsd, divss,
12287         divsd, sqrtss and sqrtsd
12288         * x86-tune-costs.h: Add new entries to all costs.
12289         (znver1_cost): Fix to match real instruction latencies.
12291 2017-10-14  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
12292             Michael Collison <michael.collison@arm.com>
12294         * compare-elim.c: Include emit-rtl.h.
12295         (can_merge_compare_into_arith): New function.
12296         (try_validate_parallel): Likewise.
12297         (try_merge_compare): Likewise.
12298         (try_eliminate_compare): Call the above when no previous clobber
12299         is available.
12300         (execute_compare_elim_after_reload): Add DF_UD_CHAIN and DF_DU_CHAIN
12301         dataflow problems.
12303 2017-10-14  Jakub Jelinek  <jakub@redhat.com>
12305         PR middle-end/62263
12306         PR middle-end/82498
12307         * tree-ssa-phiopt.c (value_replacement): Comment fix.  Handle
12308         up to 2 preparation statements for ASSIGN in MIDDLE_BB.
12310         PR middle-end/62263
12311         PR middle-end/82498
12312         * tree-ssa-forwprop.c (simplify_rotate): Allow def_arg1[N]
12313         to be any operand_equal_p operands.  For & (B - 1) require
12314         B to be power of 2.  Recognize
12315         (X << (Y & (B - 1))) | (X >> ((-Y) & (B - 1))) and similar patterns.
12317 2017-10-14  Uros Bizjak  <ubizjak@gmail.com>
12319         PR bootstrap/82553
12320         * optabs.c (expand_memory_blockage): Fix call of
12321         targetm.have_memory_blockage.
12323 2017-10-14  Jakub Jelinek  <jakub@redhat.com>
12325         PR bootstrap/82548
12326         * config.gcc (*-*-solaris2*, i[34567]86-*-cygwin*,
12327         x86_64-*-cygwin*, i[34567]86-*-mingw* | x86_64-*-mingw*): Append
12328         objects to extra_objs instead of overwriting it.
12330 2017-10-14  Uros Bizjak  <ubizjak@gmail.com>
12332         * config/i386/sync.md (FILD_ATOMIC/FIST_ATOMIC FP load peephole2):
12333         Use any_fp_register_operand as operand[3] predicate.  Simplify
12334         equality test for operands[2] and operands[4] memory location.
12335         (LDX_ATOMIC/STX_ATOMIC FP load peephole2): Ditto.
12336         (FILD_ATOMIC/FIST_ATOMIC FP load peephole2 with mem blockage): New.
12337         (LDX_ATOMIC/LDX_ATOMIC FP load peephole2 with mem blockage): Ditto.
12338         (FILD_ATOMIC/FIST_ATOMIC FP store peephole2): Use
12339         any_fp_register_operand as operand[1] predicate.  Simplify
12340         equality test for operands[0] and operands[3] memory location.
12341         (LDX_ATOMIC/STX_ATOMIC FP store peephole2): Ditto.
12342         (FILD_ATOMIC/FIST_ATOMIC FP store peephole2 with mem blockage): New.
12343         (LDX_ATOMIC/LDX_ATOMIC FP storepeephole2 with mem blockage): Ditto.
12345 2017-10-14  Uros Bizjak  <ubizjak@gmail.com>
12347         * target-insns.def: Add memory_blockage.
12348         * optabs.c (expand_memory_blockage): New function.
12349         (expand_asm_memory_barrier): Rename ...
12350         (expand_asm_memory_blockage): ... to this.
12351         (expand_mem_thread_fence): Call expand_memory_blockage
12352         instead of expand_asm_memory_barrier.
12353         (expand_mem_singnal_fence): Ditto.
12354         (expand_atomic_load): Ditto.
12355         (expand_atomic_store): Ditto.
12356         * doc/md.texi (Standard Pattern Names For Generation):
12357         Document memory_blockage instruction pattern.
12359 2017-10-13  Sebastian Perta  <sebastian.perta@renesas.com>
12361         * config/rl78/rl78.c (rl78_emit_libcall): New function.
12362         * config/rl78/rl78-protos.h (rl78_emit_libcall): New function.
12363         * config/rl78/rl78.md: New define_expand "adddi3".
12365 2017-10-13  Jan Hubicka  <hubicka@ucw.cz>
12367         * cfghooks.c (verify_flow_info): Disable check that all probabilities
12368         are set correctly.
12370 2017-10-13  Jeff Law  <law@redhat.com>
12372         * tree-ssa-reassoc.c (reassociate_bb): Clarify code slighly.
12374 2017-10-13  Jakub Jelinek  <jakub@redhat.com>
12376         PR target/82274
12377         * internal-fn.c (expand_mul_overflow): If both operands have
12378         the same highpart of -1 or 0 and the topmost bit of lowpart
12379         is different, overflow is if res <= 0 rather than res < 0.
12381 2017-10-13  Pat Haugen  <pthaugen@us.ibm.com>
12383         * config/rs6000/rs6000.c (rs6000_builtin_vectorization_cost): Remove
12384         TARGET_P9_VECTOR code for unaligned_load case.
12386 2017-10-13  Jan Hubicka  <hubicka@ucw.cz>
12388         * cfghooks.c (verify_flow_info): Check that edge probabilities are set.
12390 2017-10-13  Nathan Sidwell  <nathan@acm.org>
12392         * tree-core.h (tree_contains_struct): Make bool.
12393         * tree.c (tree_contains_struct): Likewise.
12394         * tree.h (MARK_TS_BASE): Remove do ... while (0) idiom.
12395         (MARK_TS_TYPED, MARK_TS_COMMON, MARK_TS_TYPE_COMMON,
12396         MARK_TS_TYPE_WITH_LANG_SPECIFIC, MARK_TS_DECL_MINIMAL,
12397         MARK_TS_DECL_COMMON, MARK_TS_DECL_WRTL, MARK_TS_DECL_WITH_VIS,
12398         MARK_TS_DECL_NON_COMMON): Likewise, use comma operator.
12400 2017-10-13  Richard Biener  <rguenther@suse.de>
12402         * graphite-isl-ast-to-gimple.c
12403         (translate_isl_ast_to_gimple::get_rename_from_scev): Remove unused
12404         parameters and dominance check.
12405         (translate_isl_ast_to_gimple::graphite_copy_stmts_from_block): Adjust.
12406         (translate_isl_ast_to_gimple::copy_bb_and_scalar_dependences): Likewise.
12407         (translate_isl_ast_to_gimple::graphite_regenerate_ast_isl):
12408         Do not update SSA form here or do intermediate IL verification.
12409         * graphite.c: Include tree-ssa.h and tree-into-ssa.h.
12410         (graphite_initialize): Remove check on the number of loops in
12411         the function and inline into graphite_transform_loops.
12412         (graphite_finalize): Inline into graphite_transform_loops.
12413         (graphite_transform_loops): Perform SSA update and IL verification
12414         here.
12415         * params.def (PARAM_GRAPHITE_MIN_LOOPS_PER_FUNCTION): Remove.
12417 2017-10-13  Richard Biener  <rguenther@suse.de>
12419         * graphite-isl-ast-to-gimple.c (max_mode_int_precision,
12420         graphite_expression_type_precision): Avoid global constructor
12421         by moving ...
12422         (translate_isl_ast_to_gimple::translate_isl_ast_to_gimple): Here.
12423         (translate_isl_ast_to_gimple::graphite_expr_type): Add type member.
12424         (translate_isl_ast_to_gimple::translate_isl_ast_node_for): Use it.
12425         (translate_isl_ast_to_gimple::build_iv_mapping): Likewise.
12426         (translate_isl_ast_to_gimple::graphite_create_new_guard): Likewise.
12427         * graphite-sese-to-poly.c (build_original_schedule): Return nothing.
12429 2017-10-13  H.J. Lu  <hongjiu.lu@intel.com>
12431         PR target/82499
12432         * config/i386/i386.h (ix86_red_zone_size): New.
12433         * config/i386/i386.md (push peephole2s): Replace
12434         "!ix86_using_red_zone ()" with "ix86_red_zone_size == 0".
12436 2017-10-13  Richard Sandiford  <richard.sandiford@linaro.org>
12437             Alan Hayward  <alan.hayward@arm.com>
12438             David Sherwood  <david.sherwood@arm.com>
12440         * combine.c (can_change_dest_mode): Reject changes in
12441         REGMODE_NATURAL_SIZE.
12443 2017-10-13  Richard Sandiford  <richard.sandiford@linaro.org>
12444             Alan Hayward  <alan.hayward@arm.com>
12445             David Sherwood  <david.sherwood@arm.com>
12447         * cfgexpand.c (expand_debug_expr): Use GET_MODE_UNIT_BITSIZE.
12448         (expand_debug_source_expr): Likewise.
12449         * combine.c (combine_simplify_rtx): Likewise.
12450         * cse.c (fold_rtx): Likewise.
12451         * fwprop.c (canonicalize_address): Likewise.
12452         * targhooks.c (default_shift_truncation_mask): Likewise.
12454 2017-10-13  Richard Sandiford  <richard.sandiford@linaro.org>
12455             Alan Hayward  <alan.hayward@arm.com>
12456             David Sherwood  <david.sherwood@arm.com>
12458         * optabs.c (add_equal_note): Use GET_MODE_UNIT_SIZE.
12459         (widened_mode): Likewise.
12460         (expand_unop): Likewise.
12461         * ree.c (transform_ifelse): Likewise.
12462         (merge_def_and_ext): Likewise.
12463         (combine_reaching_defs): Likewise.
12464         * simplify-rtx.c (simplify_unary_operation_1): Likewise.
12466 2017-10-13  Richard Sandiford  <richard.sandiford@linaro.org>
12467             Alan Hayward  <alan.hayward@arm.com>
12468             David Sherwood  <david.sherwood@arm.com>
12470         * caller-save.c (replace_reg_with_saved_mem): Use byte_lowpart_offset.
12471         * combine.c (gen_lowpart_for_combine): Likewise.
12472         * dwarf2out.c (rtl_for_decl_location): Likewise.
12473         * final.c (alter_subreg): Likewise.
12474         * rtlhooks.c (gen_lowpart_general): Likewise.
12475         (gen_lowpart_if_possible): Likewise.
12477 2017-10-13  Richard Sandiford  <richard.sandiford@linaro.org>
12478             Alan Hayward  <alan.hayward@arm.com>
12479             David Sherwood  <david.sherwood@arm.com>
12481         * calls.c (expand_call): Use subreg_lowpart_offset.
12482         * cse.c (cse_insn): Likewise.
12483         * regcprop.c (copy_value): Likewise.
12484         (copyprop_hardreg_forward_1): Likewise.
12486 2017-10-13  Jakub Jelinek  <jakub@redhat.com>
12488         PR target/82524
12489         * config/i386/i386.md (addqi_ext_1, andqi_ext_1,
12490         *andqi_ext_1_cc, *<code>qi_ext_1, *xorqi_ext_1_cc): Change
12491         =Q constraints to +Q and into insn condition add check
12492         that operands[0] and operands[1] are equal.
12493         (*addqi_ext_2, *andqi_ext_2, *<code>qi_ext_2): Change
12494         =Q constraints to +Q and into insn condition add check
12495         that operands[0] is equal to either operands[1] or operands[2].
12497         PR target/82498
12498         * fold-const.c (fold_binary_loc) <bit_rotate>: Code cleanups,
12499         instead of handling MINUS_EXPR twice (once for each argument),
12500         canonicalize operand order and handle just once, use rtype where
12501         possible.  Handle (A << B) | (A >> (-B & (Z - 1))).
12503         PR target/82498
12504         * config/i386/ia32intrin.h (__rold, __rord, __rolq, __rorq): Allow
12505         any values of __C while still being pattern recognizable as a simple
12506         rotate instruction.
12508 2017-10-13  Richard Biener  <rguenther@suse.de>
12510         PR tree-optimization/82451
12511         Revert
12512         2017-10-02  Richard Biener  <rguenther@suse.de>
12514         PR tree-optimization/82355
12515         * graphite-isl-ast-to-gimple.c (build_iv_mapping): Also build
12516         a mapping for the enclosing loop but avoid generating one for
12517         the loop tree root.
12518         (copy_bb_and_scalar_dependences): Remove premature codegen
12519         error on PHIs in blocks duplicated into multiple places.
12520         * graphite-scop-detection.c
12521         (scop_detection::stmt_has_simple_data_refs_p): For a loop not
12522         in the region use it as loop and nest to analyze the DR in.
12523         (try_generate_gimple_bb): Likewise.
12524         * graphite-sese-to-poly.c (extract_affine_chrec): Adjust.
12525         (add_loop_constraints): For blocks in a loop not in the region
12526         create a dimension with a single iteration.
12527         * sese.h (gbb_loop_at_index): Remove assert.
12529         * cfgloop.c (loop_preheader_edge): For the loop tree root
12530         return the single successor of the entry block.
12531         * graphite-isl-ast-to-gimple.c (graphite_regenerate_ast_isl):
12532         Reset the SCEV hashtable and niters.
12533         * graphite-scop-detection.c
12534         (scop_detection::graphite_can_represent_scev): Add SCOP parameter,
12535         assert that we only have POLYNOMIAL_CHREC that vary in loops
12536         contained in the region.
12537         (scop_detection::graphite_can_represent_expr): Adjust.
12538         (scop_detection::stmt_has_simple_data_refs_p): For loops
12539         not in the region set loop to NULL.  The nest is now the
12540         entry edge to the region.
12541         (try_generate_gimple_bb): Likewise.
12542         * sese.c (scalar_evolution_in_region): Adjust for
12543         instantiate_scev change.
12544         * tree-data-ref.h (graphite_find_data_references_in_stmt):
12545         Make nest parameter the edge into the region.
12546         (create_data_ref): Likewise.
12547         * tree-data-ref.c (dr_analyze_indices): Make nest parameter an
12548         entry edge into a region and adjust instantiate_scev calls.
12549         (create_data_ref): Likewise.
12550         (graphite_find_data_references_in_stmt): Likewise.
12551         (find_data_references_in_stmt): Pass the loop preheader edge
12552         from the nest argument.
12553         * tree-scalar-evolution.h (instantiate_scev): Make instantiate_below
12554         parameter the edge into the region.
12555         (instantiate_parameters): Use the loop preheader edge as entry.
12556         * tree-scalar-evolution.c (analyze_scalar_evolution): Handle
12557         NULL loop.
12558         (get_instantiated_value_entry): Make instantiate_below parameter
12559         the edge into the region.
12560         (instantiate_scev_name): Likewise.  Adjust dominance checks,
12561         when we cannot use loop-based instantiation instantiate by
12562         walking use-def chains.
12563         (instantiate_scev_poly): Adjust.
12564         (instantiate_scev_binary): Likewise.
12565         (instantiate_scev_convert): Likewise.
12566         (instantiate_scev_not): Likewise.
12567         (instantiate_array_ref): Remove.
12568         (instantiate_scev_3): Likewise.
12569         (instantiate_scev_2): Likewise.
12570         (instantiate_scev_1): Likewise.
12571         (instantiate_scev_r): Do not blindly handle N-operand trees.
12572         Do not instantiate array-refs.  Handle all constants and invariants.
12573         (instantiate_scev): Make instantiate_below parameter
12574         the edge into the region.
12575         (resolve_mixers): Use the loop preheader edge for the region
12576         parameter to instantiate_scev_r.
12577         * tree-ssa-loop-prefetch.c (determine_loop_nest_reuse): Adjust.
12579 2017-10-13  Richard Biener  <rguenther@suse.de>
12581         PR tree-optimization/82525
12582         * graphite-isl-ast-to-gimple.c
12583         (translate_isl_ast_to_gimple::widest_int_from_isl_expr_int): Split
12584         out from ...
12585         (translate_isl_ast_to_gimple::gcc_expression_from_isl_expr_int): Here.
12586         Fail code generation when we cannot represent the isl integer.
12587         (binary_op_to_tree): Elide modulo operations that are no-ops
12588         in the type we code generate.  Remove now superfluous code
12589         generation errors.
12591 2017-10-13  Richard Biener  <rguenther@suse.de>
12593         * graphite-scop-detection.c (loop_ivs_can_be_represented): Remove.
12594         (scop_detection::harmful_loop_in_region): Remove premature
12595         IV type restriction.
12596         (scop_detection::graphite_can_represent_scev): We can handle
12597         pointer IVs just fine.
12599 2017-10-13  Alan Modra  <amodra@gmail.com>
12601         * doc/extend.texi (Extended Asm <Clobbers>): Rename to
12602         "Clobbers and Scratch Registers".  Add paragraph on
12603         alternative to clobbers for scratch registers and OpenBLAS
12604         example.
12606 2017-10-13  Alan Modra  <amodra@gmail.com>
12608         * doc/extend.texi (Clobbers): Correct vax example.  Delete old
12609         example of a memory input for a string of known length.  Move
12610         commentary out of table.  Add a number of new examples
12611         covering array memory inputs.
12613 2017-10-12  Martin Liska  <mliska@suse.cz>
12615         PR tree-optimization/82493
12616         * sbitmap.c (bitmap_bit_in_range_p): Fix the implementation.
12617         (test_range_functions): New function.
12618         (sbitmap_c_tests): Likewise.
12619         * selftest-run-tests.c (selftest::run_tests): Run new tests.
12620         * selftest.h (sbitmap_c_tests): New function.
12622         * tree-ssa-dse.c (live_bytes_read): Fix thinko.
12624 2017-10-12  Michael Meissner  <meissner@linux.vnet.ibm.com>
12626         * config/rs6000/amo.h: Fix spacing issue.
12628 2017-10-12  Jakub Jelinek  <jakub@redhat.com>
12630         PR target/82498
12631         * config/i386/i386.md (*ashl<mode>3_mask_1,
12632         *<shift_insn><mode>3_mask_1, *<rotate_insn><mode>3_mask_1,
12633         *<btsc><mode>_mask_1, *btr<mode>_mask_1): New define_insn_and_split
12634         patterns.
12636 2017-10-12  Jan Hubicka  <hubicka@ucw.cz>
12638         * profile-count.h (safe_scale_64bit): Fix GCC4.x path.
12639         (profile_probability): Set max_probability
12640         to (uint32_t) 1 << (n_bits - 2) and update accessors to avoid overlfows
12641         in temporaries.
12642         * profile-count.c (profile_probability::differs_from_p): Do not
12643         rely on max_probaiblity == 10000
12645 2017-10-12  Jeff Law  <law@redhat.com>
12647         * tree-ssa-dse.c (valid_ao_ref_for_dse): Reject ao_refs with
12648         negative offsets.
12650 2017-10-12  Martin Sebor  <msebor@redhat.com>
12652         PR other/82301
12653         PR c/82435
12654         * cgraphunit.c (maybe_diag_incompatible_alias): New function.
12655         (handle_alias_pairs): Call it.
12656         * common.opt (-Wattribute-alias): New option.
12657         * doc/extend.texi (ifunc attribute): Discuss C++ specifics.
12658         * doc/invoke.texi (-Wattribute-alias): Document.
12660 2017-10-12  Vladimir Makarov  <vmakarov@redhat.com>
12662         Revert
12663         2017-10-11  Vladimir Makarov  <vmakarov@redhat.com>
12664         PR sanitizer/82353
12665         * lra.c (collect_non_operand_hard_regs): Don't ignore operator
12666         locations.
12667         * lra-lives.c (bb_killed_pseudos, bb_gen_pseudos): Move up.
12668         (make_hard_regno_born, make_hard_regno_dead): Update
12669         bb_killed_pseudos and bb_gen_pseudos.
12671 2017-10-12  Jan Hubicka  <hubicka@ucw.cz>
12673         * config/i386/x86-tune-sched.c (ix86_adjust_cost): Fix Zen support.
12675 2017-10-12  Uros Bizjak  <ubizjak@gmail.com>
12677         * config/alpha/alpha.c (alpha_split_conditional_move):
12678         Use std::swap instead of manually swapping.
12679         (alpha_stdarg_optimize_hook): Ditto.
12680         (alpha_canonicalize_comparison): Ditto.
12682 2017-10-12  Bin Cheng  <bin.cheng@arm.com>
12684         * tree-loop-distribution.c (struct builtin_info): New struct.
12685         (struct partition): Refactor fields into struct builtin_info.
12686         (partition_free): Free struct builtin_info.
12687         (build_size_arg_loc, build_addr_arg_loc): Delete.
12688         (generate_memset_builtin, generate_memcpy_builtin): Get memory range
12689         information from struct builtin_info.
12690         (find_single_drs): New function refactored from classify_partition.
12691         Also moved builtin validity checks to this function.
12692         (compute_access_range, alloc_builtin): New functions.
12693         (classify_builtin_st, classify_builtin_ldst): New functions.
12694         (classify_partition): Refactor code into functions find_single_drs,
12695         classify_builtin_st and classify_builtin_ldst.
12696         (distribute_loop): Don't do runtime alias check when distributing
12697         loop nest.
12698         (find_seed_stmts_for_distribution): New function.
12699         (pass_loop_distribution::execute): Refactor code finding seed
12700         stmts into above function.  Support distribution for the innermost
12701         two-level loop nest.  Adjust dump information.
12703 2017-10-12  Bin Cheng  <bin.cheng@arm.com>
12705         * tree-loop-distribution.c: Adjust the general comment.
12706         (NUM_PARTITION_THRESHOLD): New macro.
12707         (ssa_name_has_uses_outside_loop_p): Support loop nest distribution.
12708         (classify_partition): Skip builtin pattern of loop nest's inner loop.
12709         (merge_dep_scc_partitions): New parameter ignore_alias_p and use it
12710         in call to build_partition_graph.
12711         (finalize_partitions): New parameter.  Make loop distribution more
12712         conservative by fusing more partitions.
12713         (distribute_loop): Don't do runtime alias check in case of loop nest
12714         distribution.
12715         (find_seed_stmts_for_distribution): New function.
12716         (prepare_perfect_loop_nest): New function.
12717         (pass_loop_distribution::execute): Refactor code finding seed stmts
12718         and loop nest into above functions.  Support loop nest distribution.
12719         Adjust dump information accordingly.
12721 2017-10-12  Bin Cheng  <bin.cheng@arm.com>
12723         * tree-loop-distribution.c (break_alias_scc_partitions): Add comment
12724         and set PTYPE_SEQUENTIAL for merged partition.
12726 2017-10-12  Richard Biener  <rguenther@suse.de>
12728         PR tree-optimization/69728
12729         Revert
12730         2017-09-19  Richard Biener  <rguenther@suse.de>
12732         PR tree-optimization/69728
12733         * graphite-sese-to-poly.c (schedule_error): New global.
12734         (add_loop_schedule): Handle empty domain by failing the
12735         schedule.
12736         (build_original_schedule): Handle schedule_error.
12738         * graphite-sese-to-poly.c (add_loop_schedule): Handle empty
12739         domain by returning an unchanged schedule.
12741 2017-10-12  Jakub Jelinek  <jakub@redhat.com>
12743         * genrecog.c (validate_pattern): For VEC_SELECT verify that
12744         CONST_INT selectors are 0 to GET_MODE_NUNITS (imode) - 1.
12746 2017-10-12  Aldy Hernandez  <aldyh@redhat.com>
12748         * Makefile.in (TAGS): Merge all the *.def files into one pattern.
12749         Handle params.def.
12751 2017-10-12  Jakub Jelinek  <jakub@redhat.com>
12753         PR c++/82159
12754         * expr.c (store_field): Don't optimize away bitsize == 0 store
12755         from CALL_EXPR with addressable return type.
12757 2017-10-11  Segher Boessenkool  <segher@kernel.crashing.org>
12759         * config/rs6000/rs6000.h (TARGET_ISEL64): Delete.
12760         * config/rs6000/rs6000.md (sel): Delete mode attribute.
12761         (mov<mode>cc, isel_signed_<mode>, isel_unsigned_<mode>,
12762         *isel_reversed_signed_<mode>, *isel_reversed_unsigned_<mode>): Use
12763         TARGET_ISEL instead of TARGET_ISEL<sel>.
12765 2017-10-11  David Edelsohn  <dje.gcc@gmail.com>
12767         * config/rs6000/rs6000.c
12768         (rs6000_xcoff_asm_output_aligned_decl_common): Test for NULL decl.
12770 2017-10-11  Segher Boessenkool  <segher@kernel.crashing.org>
12772         * config/rs6000/predicates.md (zero_constant, all_ones_constant):
12773         Move up in file.
12774         (reg_or_cint_operand): Fix comment.
12775         (reg_or_zero_operand): New predicate.
12776         * config/rs6000/rs6000-protos.h (output_isel): Delete.
12777         * config/rs6000/rs6000.c (output_isel): Delete.
12778         * config/rs6000/rs6000.md (isel_signed_<mode>): Use reg_or_zero_operand
12779         instead of reg_or_cint_operand.  Output instruction directly (not via
12780         output_isel).
12781         (isel_unsigned_<mode>): Ditto.
12782         (*isel_reversed_signed_<mode>): Use reg_or_zero_operand instead of
12783         gpc_reg_operand.  Add an instruction alternative for this.  Output
12784         instruction directly.
12785         (*isel_reversed_unsigned_<mode>): Ditto.
12787 2017-10-11  Uros Bizjak  <ubizjak@gmail.com>
12789         * config/i386/i386.c (ix86_canonicalize_comparison): New function.
12790         (TARGET_CANONICALIZE_COMPARISON): Define.
12792 2017-10-11  Qing Zhao  <qing.zhao@oracle.com>
12794         PR target/81422
12795         * config/aarch64/aarch64.c (aarch64_load_symref_appropriately):
12796         Check whether the dest is REG before adding REG_EQUIV note.
12798 2017-10-11  Vladimir Makarov  <vmakarov@redhat.com>
12800         PR sanitizer/82353
12801         * lra.c (collect_non_operand_hard_regs): Don't ignore operator
12802         locations.
12803         * lra-lives.c (bb_killed_pseudos, bb_gen_pseudos): Move up.
12804         (make_hard_regno_born, make_hard_regno_dead): Update
12805         bb_killed_pseudos and bb_gen_pseudos.
12807 2017-10-11  Nathan Sidwell  <nathan@acm.org>
12809         * incpath.h (enum incpath_kind): Name enum, prefix values.
12810         (add_path, add_cpp_dir_path, get_added_cpp_dirs): Use incpath_kind.
12811         * incpath.c (heads, tails): Use INC_MAX.
12812         (add_env_var_paths, add_standard_paths): Use incpath_kind.
12813         (merge_include_chains, split_quote_chain,
12814         register_include_chains): Update incpath_kind names.
12815         (add_cpp_dir_path, add_path, get_added_cpp_dirs): Use incpath_kind.
12816         * config/darwin-c.c (add_system_framework_path): Update incpath_kind
12817         names.
12818         (add_framework_path, darwin_register_objc_includes): Likewise.
12819         * config/vms/vms-c.c (vms_c_register_includes): Likewise.
12821 2017-10-11  Uros Bizjak  <ubizjak@gmail.com>
12823         * config/i386/i386.md (*cmp<X87MODEF:mode>_<SWI24:mode>_i387):
12824         Do not use float_operator operator predicate.
12825         (*cmp<X87MODEF:mode>_<SWI24:mode>_cc_i387): Ditto.
12826         * config/i386/predicates.md (float_operator): Remove predicate.
12828 2017-10-11  Uros Bizjak  <ubizjak@gmail.com>
12830         * config/i386/i386.md (*jcc<mode>_0_i387): Remove insn pattern.
12831         (*jccxf_i387): Ditto.
12832         (*jcc<mode>_i387): Ditto.
12833         (*jccu<mode>_i387): Ditto.
12834         (*jcc<X87MODEF:mode>_<SWI24:mode>_i387): Ditto.
12835         (*jcc_*_i387 splitters): Remove.
12836         * config/i386/i386-protos.h (ix86_split_fp_branch): Remove prototype.
12837         * config/i386/i386.c (ix86_split_fp_branch): Remove.
12838         * config/i386/predicates.md (ix86_swapped_fp_comparison_operator):
12839         Remove predicate.
12841 2017-10-11  Jan Hubicka  <hubicka@ucw.cz>
12843         * profile-count.h (slow_safe_scale_64bit): New function.
12844         (safe_scale_64bit): New inline.
12845         (profile_count::max_safe_multiplier): Remove; use safe_scale_64bit.
12846         * profile-count.c: Include wide-int.h
12847         (slow_safe_scale_64bit): New.
12849 2017-10-11  Nathan Sidwell  <nathan@acm.org>
12851         * tree.h (DECL_ASSEMBLER_NAME_SET_P): Don't check
12852         HAS_DECL_ASSEMBLER_NAME_P.
12853         * gimple-expr.c (gimple_decl_printable_name: Check
12854         HAS_DECL_ASSEMBLER_NAME_P too.
12855         * ipa-utils.h (type_in_anonymous_namespace_p): Check
12856         DECL_ASSEMBLER_NAME_SET_P of TYPE_NAME.
12857         (odr_type_p): No need to assert TYPE_NAME is a TYPE_DECL.
12858         * passes.c (rest_of_decl_compilation): Check
12859         HAS_DECL_ASSEMBLER_NAME_P too.
12860         * recog.c (verify_changes): Likewise.
12861         * tree-pretty-print.c (dump_decl_name): Likewise.
12862         * tree-ssa-structalias.c (alias_get_name): Likewise.  Reimplement.
12864         * tree.h (DECL_ASSEMBLER_NAME_RAW): New.
12865         (SET_DECL_ASSEMBLER_NAME): Use it.
12866         (DECL_ASSEMBLER_NAME_SET_P): Likewise.
12867         (COPY_DECL_ASSEMBLER_NAME): Likewise.
12868         * tree.c (decl_assembler_name): Use DECL_ASSEMBLER_NAME_RAW.
12870 2017-10-11  Jan Hubicka  <hubicka@ucw.cz>
12872         * config.gcc (i386, x86_64): Add extra objects.
12873         * config/i386/i386-protos.h (ix86_rip_relative_addr_p): Declare.
12874         (ix86_min_insn_size): Declare.
12875         (ix86_issue_rate): Declare.
12876         (ix86_adjust_cost): Declare.
12877         (ia32_multipass_dfa_lookahead): Declare.
12878         (ix86_macro_fusion_p): Declare.
12879         (ix86_macro_fusion_pair_p): Declare.
12880         (ix86_bd_has_dispatch): Declare.
12881         (ix86_bd_do_dispatch): Declare.
12882         (ix86_core2i7_init_hooks): Declare.
12883         (ix86_atom_sched_reorder): Declare.
12884         * config/i386/i386.c Move all CPU cost tables to x86-tune-costs.h.
12885         (COSTS_N_BYTES): Move to x86-tune-costs.h.
12886         (DUMMY_STRINGOP_ALGS):Move to x86-tune-costs.h.
12887         (rip_relative_addr_p): Rename to ...
12888         (ix86_rip_relative_addr_p): ... this one; export.
12889         (memory_address_length): Update.
12890         (ix86_issue_rate): Move to x86-tune-sched.c.
12891         (ix86_flags_dependent): Move to x86-tune-sched.c.
12892         (ix86_agi_dependent): Move to x86-tune-sched.c.
12893         (exact_dependency_1): Move to x86-tune-sched.c.
12894         (exact_store_load_dependency): Move to x86-tune-sched.c.
12895         (ix86_adjust_cost): Move to x86-tune-sched.c.
12896         (ia32_multipass_dfa_lookahead): Move to x86-tune-sched.c.
12897         (ix86_macro_fusion_p): Move to x86-tune-sched.c.
12898         (ix86_macro_fusion_pair_p): Move to x86-tune-sched.c.
12899         (do_reorder_for_imul): Move to x86-tune-sched-atom.c.
12900         (swap_top_of_ready_list): Move to x86-tune-sched-atom.c.
12901         (ix86_sched_reorder): Move to x86-tune-sched-atom.c.
12902         (core2i7_first_cycle_multipass_init): Move to x86-tune-sched-core.c.
12903         (core2i7_dfa_post_advance_cycle): Move to x86-tune-sched-core.c.
12904         (min_insn_size): Rename to ...
12905         (ix86_min_insn_size): ... this one; export.
12906         (core2i7_first_cycle_multipass_begin): Move to x86-tune-sched-core.c.
12907         (core2i7_first_cycle_multipass_issue): Move to x86-tune-sched-core.c.
12908         (core2i7_first_cycle_multipass_backtrack): Move to
12909         x86-tune-sched-core.c.
12910         (core2i7_first_cycle_multipass_end): Move to x86-tune-sched-core.c.
12911         (core2i7_first_cycle_multipass_fini): Move to x86-tune-sched-core.c.
12912         (ix86_sched_init_global): Break up logic to ix86_core2i7_init_hooks.
12913         (ix86_avoid_jump_mispredicts): Update.
12914         (TARGET_SCHED_DISPATCH): Move to ix86-tune-sched-bd.c.
12915         (TARGET_SCHED_DISPATCH_DO): Move to ix86-tune-sched-bd.c.
12916         (TARGET_SCHED_REORDER): Move to ix86-tune-sched-bd.c.
12917         (DISPATCH_WINDOW_SIZE): Move to ix86-tune-sched-bd.c.
12918         (MAX_DISPATCH_WINDOWS): Move to ix86-tune-sched-bd.c.
12919         (MAX_INSN): Move to ix86-tune-sched-bd.c.
12920         (MAX_IMM): Move to ix86-tune-sched-bd.c.
12921         (MAX_IMM_SIZE): Move to ix86-tune-sched-bd.c.
12922         (MAX_IMM_32): Move to ix86-tune-sched-bd.c.
12923         (MAX_IMM_64): Move to ix86-tune-sched-bd.c.
12924         (MAX_LOAD): Move to ix86-tune-sched-bd.c.
12925         (MAX_STORE): Move to ix86-tune-sched-bd.c.
12926         (BIG): Move to ix86-tune-sched-bd.c.
12927         (enum dispatch_group): Move to ix86-tune-sched-bd.c.
12928         (enum insn_path): Move to ix86-tune-sched-bd.c.
12929         (get_mem_group): Move to ix86-tune-sched-bd.c.
12930         (is_cmp): Move to ix86-tune-sched-bd.c.
12931         (dispatch_violation): Move to ix86-tune-sched-bd.c.
12932         (is_branch): Move to ix86-tune-sched-bd.c.
12933         (is_prefetch): Move to ix86-tune-sched-bd.c.
12934         (init_window): Move to ix86-tune-sched-bd.c.
12935         (allocate_window): Move to ix86-tune-sched-bd.c.
12936         (init_dispatch_sched): Move to ix86-tune-sched-bd.c.
12937         (is_end_basic_block): Move to ix86-tune-sched-bd.c.
12938         (process_end_window): Move to ix86-tune-sched-bd.c.
12939         (allocate_next_window): Move to ix86-tune-sched-bd.c.
12940         (find_constant): Move to ix86-tune-sched-bd.c.
12941         (get_num_immediates): Move to ix86-tune-sched-bd.c.
12942         (has_immediate): Move to ix86-tune-sched-bd.c.
12943         (get_insn_path): Move to ix86-tune-sched-bd.c.
12944         (get_insn_group): Move to ix86-tune-sched-bd.c.
12945         (count_num_restricted): Move to ix86-tune-sched-bd.c.
12946         (fits_dispatch_window): Move to ix86-tune-sched-bd.c.
12947         (add_insn_window): Move to ix86-tune-sched-bd.c.
12948         (add_to_dispatch_window): Move to ix86-tune-sched-bd.c.
12949         (debug_dispatch_window_file): Move to ix86-tune-sched-bd.c.
12950         (debug_dispatch_window): Move to ix86-tune-sched-bd.c.
12951         (debug_insn_dispatch_info_file): Move to ix86-tune-sched-bd.c.
12952         (debug_ready_dispatch): Move to ix86-tune-sched-bd.c.
12953         (do_dispatch): Move to ix86-tune-sched-bd.c.
12954         (has_dispatch): Move to ix86-tune-sched-bd.c.
12955         * config/i386/t-i386: Add new object files.
12956         * config/i386/x86-tune-costs.h: New file.
12957         * config/i386/x86-tune-sched-atom.c: New file.
12958         * config/i386/x86-tune-sched-bd.c: New file.
12959         * config/i386/x86-tune-sched-core.c: New file.
12960         * config/i386/x86-tune-sched.c: New file.
12962 2017-10-11  Liu Hao  <lh_mouse@126.com>
12964         * pretty-print.c [_WIN32] (colorize_init): Remove.  Use
12965         the generic version below instead.
12966         (should_colorize): Recognize Windows consoles as terminals
12967         for MinGW targets.
12968         * pretty-print.c [__MINGW32__] (write_all): New function.
12969         [__MINGW32__] (find_esc_head): Likewise.
12970         [__MINGW32__] (find_esc_terminator): Likewise.
12971         [__MINGW32__] (eat_esc_sequence): Likewise.
12972         [__MINGW32__] (mingw_ansi_fputs): New function that handles
12973         ANSI escape codes.
12974         (pp_write_text_to_stream): Use mingw_ansi_fputs instead of fputs
12975         for MinGW targets.
12977 2017-10-11  Richard Biener  <rguenther@suse.de>
12979         * tree-ssa-loop-niter.c (infer_loop_bounds_from_pointer_arith):
12980         Properly call analyze_scalar_evolution with the loop of the stmt.
12982 2017-10-11  Richard Biener  <rguenther@suse.de>
12984         * tree.def (POLYNOMIAL_CHREC): Remove CHREC_VARIABLE tree operand.
12985         * tree-core.h (tree_base): Add chrec_var union member.
12986         * tree.h (CHREC_VAR): Remove.
12987         (CHREC_LEFT, CHREC_RIGHT, CHREC_VARIABLE): Adjust.
12988         * tree-chrec.h (build_polynomial_chrec): Adjust.
12989         * tree-chrec.c (reset_evolution_in_loop): Use build_polynomial_chrec.
12990         * tree-pretty-print.c (dump_generic_node): Use CHREC_VARIABLE.
12992 2017-10-11  Marc Glisse  <marc.glisse@inria.fr>
12994         * fold-const.c (fold_binary_loc) [X +- Y CMP X]: Move ...
12995         * match.pd: ... here.
12996         ((T) X == (T) Y): Relax condition.
12998 2017-10-11  Bin Cheng  <bin.cheng@arm.com>
13000         PR tree-optimization/82472
13001         * tree-loop-distribution.c (sort_partitions_by_post_order): Refine
13002         comment.
13003         (break_alias_scc_partitions): Update postorder number.
13005 2017-10-11  Martin Liska  <mliska@suse.cz>
13007         PR sanitizer/82490
13008         * opts.c (parse_no_sanitize_attribute): Do not use error_value
13009         variable.
13010         * opts.h (parse_no_sanitize_attribute): Remove last argument.
13012 2017-10-11  Martin Liska  <mliska@suse.cz>
13014         * print-rtl.c (print_insn): Move declaration of idbuf
13015         to same scope as name.
13017 2017-10-11  Martin Liska  <mliska@suse.cz>
13019         Revert r253637:
13021         PR sanitizer/82484
13022         * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
13023         volatile arguments.
13025 2017-10-11  Martin Liska  <mliska@suse.cz>
13027         PR sanitizer/82484
13028         * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
13029         volatile arguments.
13031 2017-10-11  Adhemerval Zanella  <adhemerval.zanella@linaro.org>
13033         * config.gcc (default_gnu_indirect_function): Default to yes for
13034         arm*-*-linux* with glibc.
13036 2017-10-11  Richard Biener  <rguenther@suse.de>
13038         * tree-scalar-evolution.c (get_scalar_evolution): Handle
13039         default-defs and types we do not want to analyze.
13040         (interpret_loop_phi): Replace unreachable code with an assert.
13041         (compute_scalar_evolution_in_loop): Remove and inline ...
13042         (analyze_scalar_evolution_1): ... here, replacing condition with
13043         what makes the intent clearer.  Remove handling of cases
13044         get_scalar_evolution now handles.
13046 2017-10-10  Jim Wilson  <wilson@tuliptree.org>
13048         PR rtl-optimization/81434
13049         * haifa-sched.c (prune_ready_list): Init min_cost_group to 0.  Update
13050         comment for main loop.  In sched_group_found if, also add checks for
13051         pass and min_cost_group.
13053 2017-10-10  Segher Boessenkool  <segher@kernel.crashing.org>
13055         * config/rs6000/rs6000.c (TARGET_INSN_COST): New.
13056         (rs6000_insn_cost): New function.
13057         * config/rs6000/rs6000.md (cost): New attribute.
13059 2017-10-10  Jakub Jelinek  <jakub@redhat.com>
13060             H.J. Lu  <hongjiu.lu@intel.com>
13062         PR target/79565
13063         PR target/82483
13064         * config/i386/i386.c (ix86_init_mmx_sse_builtins): Add
13065         OPTION_MASK_ISA_MMX for __builtin_ia32_maskmovq,
13066         __builtin_ia32_vec_ext_v4hi and __builtin_ia32_vec_set_v4hi.
13067         (ix86_expand_builtin): Treat OPTION_MASK_ISA_MMX similarly
13068         to OPTION_MASK_ISA_AVX512VL - builtins that have both
13069         OPTION_MASK_ISA_MMX and some other bit set require both
13070         mmx and the ISAs without the mmx bit.
13071         * config/i386/i386-builtin.def (__builtin_ia32_cvtps2pi,
13072         __builtin_ia32_cvttps2pi, __builtin_ia32_cvtpi2ps,
13073         __builtin_ia32_pavgb, __builtin_ia32_pavgw, __builtin_ia32_pmulhuw,
13074         __builtin_ia32_pmaxub, __builtin_ia32_pmaxsw, __builtin_ia32_pminub,
13075         __builtin_ia32_pminsw, __builtin_ia32_psadbw, __builtin_ia32_pmovmskb,
13076         __builtin_ia32_pshufw, __builtin_ia32_cvtpd2pi,
13077         __builtin_ia32_cvttpd2pi, __builtin_ia32_cvtpi2pd,
13078         __builtin_ia32_pmuludq, __builtin_ia32_pabsb, __builtin_ia32_pabsw,
13079         __builtin_ia32_pabsd, __builtin_ia32_phaddw, __builtin_ia32_phaddd,
13080         __builtin_ia32_phaddsw, __builtin_ia32_phsubw, __builtin_ia32_phsubd,
13081         __builtin_ia32_phsubsw, __builtin_ia32_pmaddubsw,
13082         __builtin_ia32_pmulhrsw, __builtin_ia32_pshufb, __builtin_ia32_psignb,
13083         __builtin_ia32_psignw, __builtin_ia32_psignd, __builtin_ia32_movntq,
13084         __builtin_ia32_paddq, __builtin_ia32_psubq, __builtin_ia32_palignr):
13085         Add OPTION_MASK_ISA_MMX.
13087 2017-10-10  Andreas Tobler  <andreast@gcc.gnu.org>
13089         * config.gcc (armv7*-*-freebsd*): New target.
13090         (armv6*-*-freebsd*): Remove obsolete TARGET_FREEBSD_ARMv6 define.
13092 2017-10-10  Jan Hubicka  <hubicka@ucw.cz>
13094         * x86-tune.def (X86_TUNE_AVOID_FALSE_DEP_FOR_BMI,
13095         X86_TUNE_ADJUST_UNROLL, X86_TUNE_ONE_IF_CONV_INSN): Move to right
13096         spot in the file.
13098 2017-10-10  Richard Sandiford  <richard.sandiford@linaro.org>
13100         * wide-int.h (wide_int_ref_storage): Make host_dependent_precision
13101         a template parameter.
13102         (WIDE_INT_REF_FOR): Update accordingly.
13103         * tree.h (wi::int_traits <const_tree>): Delete.
13104         (wi::tree_to_widest_ref, wi::tree_to_offset_ref): New typedefs.
13105         (wi::to_widest, wi::to_offset): Use them.  Expand commentary.
13106         (wi::tree_to_wide_ref): New typedef.
13107         (wi::to_wide): New function.
13108         * calls.c (get_size_range): Use wi::to_wide when operating on
13109         trees as wide_ints.
13110         * cgraph.c (cgraph_node::create_thunk): Likewise.
13111         * config/i386/i386.c (ix86_data_alignment): Likewise.
13112         (ix86_local_alignment): Likewise.
13113         * dbxout.c (stabstr_O): Likewise.
13114         * dwarf2out.c (add_scalar_info, gen_enumeration_type_die): Likewise.
13115         * expr.c (const_vector_from_tree): Likewise.
13116         * fold-const-call.c (host_size_t_cst_p, fold_const_call_1): Likewise.
13117         * fold-const.c (may_negate_without_overflow_p, negate_expr_p)
13118         (fold_negate_expr_1, int_const_binop_1, const_binop)
13119         (fold_convert_const_int_from_real, optimize_bit_field_compare)
13120         (all_ones_mask_p, sign_bit_p, unextend, extract_muldiv_1)
13121         (fold_div_compare, fold_single_bit_test, fold_plusminus_mult_expr)
13122         (pointer_may_wrap_p, expr_not_equal_to, fold_binary_loc)
13123         (fold_ternary_loc, multiple_of_p, fold_negate_const, fold_abs_const)
13124         (fold_not_const, round_up_loc): Likewise.
13125         * gimple-fold.c (gimple_fold_indirect_ref): Likewise.
13126         * gimple-ssa-warn-alloca.c (alloca_call_type_by_arg): Likewise.
13127         (alloca_call_type): Likewise.
13128         * gimple.c (preprocess_case_label_vec_for_gimple): Likewise.
13129         * godump.c (go_output_typedef): Likewise.
13130         * graphite-sese-to-poly.c (tree_int_to_gmp): Likewise.
13131         * internal-fn.c (get_min_precision): Likewise.
13132         * ipa-cp.c (ipcp_store_vr_results): Likewise.
13133         * ipa-polymorphic-call.c
13134         (ipa_polymorphic_call_context::ipa_polymorphic_call_context): Likewise.
13135         * ipa-prop.c (ipa_print_node_jump_functions_for_edge): Likewise.
13136         (ipa_modify_call_arguments): Likewise.
13137         * match.pd: Likewise.
13138         * omp-low.c (scan_omp_1_op, lower_omp_ordered_clauses): Likewise.
13139         * print-tree.c (print_node_brief, print_node): Likewise.
13140         * stmt.c (expand_case): Likewise.
13141         * stor-layout.c (layout_type): Likewise.
13142         * tree-affine.c (tree_to_aff_combination): Likewise.
13143         * tree-cfg.c (group_case_labels_stmt): Likewise.
13144         * tree-data-ref.c (dr_analyze_indices): Likewise.
13145         (prune_runtime_alias_test_list): Likewise.
13146         * tree-dump.c (dequeue_and_dump): Likewise.
13147         * tree-inline.c (remap_gimple_op_r, copy_tree_body_r): Likewise.
13148         * tree-predcom.c (is_inv_store_elimination_chain): Likewise.
13149         * tree-pretty-print.c (dump_generic_node): Likewise.
13150         * tree-scalar-evolution.c (iv_can_overflow_p): Likewise.
13151         (simple_iv_with_niters): Likewise.
13152         * tree-ssa-address.c (addr_for_mem_ref): Likewise.
13153         * tree-ssa-ccp.c (ccp_finalize, evaluate_stmt): Likewise.
13154         * tree-ssa-loop-ivopts.c (constant_multiple_of): Likewise.
13155         * tree-ssa-loop-niter.c (split_to_var_and_offset)
13156         (refine_value_range_using_guard, number_of_iterations_ne_max)
13157         (number_of_iterations_lt_to_ne, number_of_iterations_lt)
13158         (get_cst_init_from_scev, record_nonwrapping_iv)
13159         (scev_var_range_cant_overflow): Likewise.
13160         * tree-ssa-phiopt.c (minmax_replacement): Likewise.
13161         * tree-ssa-pre.c (compute_avail): Likewise.
13162         * tree-ssa-sccvn.c (vn_reference_fold_indirect): Likewise.
13163         (vn_reference_maybe_forwprop_address, valueized_wider_op): Likewise.
13164         * tree-ssa-structalias.c (get_constraint_for_ptr_offset): Likewise.
13165         * tree-ssa-uninit.c (is_pred_expr_subset_of): Likewise.
13166         * tree-ssanames.c (set_nonzero_bits, get_nonzero_bits): Likewise.
13167         * tree-switch-conversion.c (collect_switch_conv_info, array_value_type)
13168         (dump_case_nodes, try_switch_expansion): Likewise.
13169         * tree-vect-loop-manip.c (vect_gen_vector_loop_niters): Likewise.
13170         (vect_do_peeling): Likewise.
13171         * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
13172         * tree-vect-stmts.c (vectorizable_load): Likewise.
13173         * tree-vrp.c (compare_values_warnv, vrp_int_const_binop): Likewise.
13174         (zero_nonzero_bits_from_vr, ranges_from_anti_range): Likewise.
13175         (extract_range_from_binary_expr_1, adjust_range_with_scev): Likewise.
13176         (overflow_comparison_p_1, register_edge_assert_for_2): Likewise.
13177         (is_masked_range_test, find_switch_asserts, maybe_set_nonzero_bits)
13178         (vrp_evaluate_conditional_warnv_with_ops, intersect_ranges): Likewise.
13179         (range_fits_type_p, two_valued_val_range_p, vrp_finalize): Likewise.
13180         (evrp_dom_walker::before_dom_children): Likewise.
13181         * tree.c (cache_integer_cst, real_value_from_int_cst, integer_zerop)
13182         (integer_all_onesp, integer_pow2p, integer_nonzerop, tree_log2)
13183         (tree_floor_log2, tree_ctz, mem_ref_offset, tree_int_cst_sign_bit)
13184         (tree_int_cst_sgn, get_unwidened, int_fits_type_p): Likewise.
13185         (get_type_static_bounds, num_ending_zeros, drop_tree_overflow)
13186         (get_range_pos_neg): Likewise.
13187         * ubsan.c (ubsan_expand_ptr_ifn): Likewise.
13188         * config/darwin.c (darwin_mergeable_constant_section): Likewise.
13189         * config/aarch64/aarch64.c (aapcs_vfp_sub_candidate): Likewise.
13190         * config/arm/arm.c (aapcs_vfp_sub_candidate): Likewise.
13191         * config/avr/avr.c (avr_fold_builtin): Likewise.
13192         * config/bfin/bfin.c (bfin_local_alignment): Likewise.
13193         * config/msp430/msp430.c (msp430_attr): Likewise.
13194         * config/nds32/nds32.c (nds32_insert_attributes): Likewise.
13195         * config/powerpcspe/powerpcspe-c.c
13196         (altivec_resolve_overloaded_builtin): Likewise.
13197         * config/powerpcspe/powerpcspe.c (rs6000_aggregate_candidate)
13198         (rs6000_expand_ternop_builtin): Likewise.
13199         * config/rs6000/rs6000-c.c
13200         (altivec_resolve_overloaded_builtin): Likewise.
13201         * config/rs6000/rs6000.c (rs6000_aggregate_candidate): Likewise.
13202         (rs6000_expand_ternop_builtin): Likewise.
13203         * config/s390/s390.c (s390_handle_hotpatch_attribute): Likewise.
13205 2017-10-10  Bin Cheng  <bin.cheng@arm.com>
13207         * tree-vect-loop-manip.c (rename_variables_in_bb): Rename PHI nodes
13208         when copying loop nest with only one inner loop.
13210 2017-10-10  Richard Biener  <rguenther@suse.de>
13212         * tree-cfgcleanup.c (cleanup_tree_cfg_noloop): Avoid compacting
13213         blocks if SCEV is active.
13214         * tree-scalar-evolution.c (analyze_scalar_evolution_1): Remove
13215         dead code.
13216         (analyze_scalar_evolution): Handle cached evolutions the obvious way.
13217         (scev_initialize): Assert we are not yet initialized.
13219 2017-10-10  Bin Cheng  <bin.cheng@arm.com>
13221         * tree-loop-distribution.c (generate_loops_for_partition): Remove
13222         inner loop's exit stmt by making it always exit the loop, otherwise
13223         we would generate an infinite empty loop.
13225 2017-10-10  Bin Cheng  <bin.cheng@arm.com>
13227         * tree-vect-loop-manip.c (slpeel_tree_duplicate_loop_to_edge_cfg): Skip
13228         renaming variables in new preheader if it's deleted.
13230 2017-10-10  Bin Cheng  <bin.cheng@arm.com>
13232         * tree-loop-distribution.c (struct partition): Remove unused field
13233         loops of the structure.
13234         (partition_alloc, partition_free): Ditto.
13235         (build_rdg_partition_for_vertex): Ditto.
13237 2017-10-09  Jeff Law  <law@redhat.com>
13239         * targhooks.c (default_stack_clash_protection_final_dynamic_probe): Fix
13240         return type to match prototype and documentation.
13242 2010-10-09  Segher Boessenkool  <segher@kernel.crashing.org>
13244         * config/rs6000/rs6000.c (processor_costs): Move to ...
13245         * config/rs6000/rs6000.h: ... here.
13246         (rs6000_cost): Declare.
13248 2017-10-09  Eric Botcazou  <ebotcazou@adacore.com>
13250         * except.c (setjmp_fn): New global variable.
13251         (init_eh): Initialize it if DONT_USE_BUILTIN_SETJMP is defined.
13252         (sjlj_emit_function_enter): Call it instead of BUILTIN_SETJMP
13253         if DONT_USE_BUILTIN_SETJMP is defined.
13255 2017-10-09  Segher Boessenkool  <segher@kernel.crashing.org>
13257         * target.def (insn_cost): New hook.
13258         * doc/tm.texi.in (TARGET_INSN_COST): New hook.
13259         * doc/tm.texi: Regenerate.
13260         * rtlanal.c (insn_cost): Use the new hook.
13262 2017-10-09  Segher Boessenkool  <segher@kernel.crashing.org>
13264         * combine.c (combine_validate_cost): Compute the new insn_cost,
13265         not just pattern_cost.
13266         (try_combine): Adjust comment.
13268 2017-10-09  Segher Boessenkool  <segher@kernel.crashing.org>
13270         * cfgrtl.c (rtl_account_profile_record): Replace insn_rtx_cost with
13271         insn_cost.
13272         * combine.c (uid_insn_cost): Adjust comment.
13273         (combine_validate_cost): Adjust comment.  Use pattern_cost instead
13274         of insn_rtx_cost
13275         (combine_instructions): Use insn_cost instead of insn_rtx_cost.
13276         * dse.c (find_shift_sequence): Ditto.
13277         * ifcvt.c (cheap_bb_rtx_cost_p): Ditto.
13278         (bb_valid_for_noce_process_p): Use pattern_cost.
13279         * rtl.h (insn_rtx_cost): Delete.
13280         (pattern_cost): New prototype.
13281         (insn_cost): New prototype.
13282         * rtlanal.c (insn_rtx_cost): Rename to...
13283         (pattern_cost): ... this.
13284         (insn_cost): New.
13286 2017-10-09  Uros Bizjak  <ubizjak@gmail.com>
13288         * config/i386/i386.md (*jcc_2): Remove insn pattern.
13289         (*jcc<mode>_0_r_i387): Ditto.
13290         (*jccxf_r_i387): Ditto.
13291         (*jcc<mode>_r_i387): Ditto.
13292         (*jccu<mode>_r_i387): Ditto.
13293         (*jcc<X87MODEF:mode>_<SWI24:mode>_r_i387): Ditto.
13294         (*jcc): Rename from *jcc_1.
13296 2017-10-09  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
13298         * config/rs6000/rs6000-p8swap.c (rs6000_analyze_swaps): Process
13299         deferred rescans after the lvx/stvx recombination pre-pass.
13301 2017-10-09  Michael Meissner  <meissner@linux.vnet.ibm.com>
13303         * config/rs6000/amo.h: New include file to provide ISA 3.0 atomic
13304         memory operation instruction support.
13305         * config.gcc (powerpc*-*-*): Include amo.h as an extra header.
13306         (rs6000-ibm-aix[789]*): Likewise.
13307         * doc/extend.texi (PowerPC Atomic Memory Operation Functions):
13308         Document new functions.
13310 2017-10-09  Richard Biener  <rguenther@suse.de>
13312         PR tree-optimization/82397
13313         * tree-data-ref.c (data_ref_compare_tree): Make sure to return
13314         equality only for semantically equal trees.
13316 2017-10-09  Richard Biener  <rguenther@suse.de>
13318         PR tree-optimization/82449
13319         * sese.c (scev_analyzable_p): Check whether the SCEV is linear.
13320         * tree-chrec.h (evolution_function_is_constant_p): Adjust to
13321         allow constant addresses.
13322         * tree-chrec.c (scev_is_linear_expression): Constant evolutions
13323         are linear.
13325 2017-10-09  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
13327         * config/s390/s390-builtins.def (vec_nabs, vec_vfi): Fix builtin
13328         flags.
13330 2017-10-09  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
13332         PR target/82463
13333         * config/s390/vecintrin.h (vec_madd, vec_msub): Fix macro
13334         definitions.
13336 2017-10-09  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
13338         PR target/82465
13339         * config/s390/s390-builtins.def (vec_sqrt): Fix builtin flags.
13341 2017-10-09  Jakub Jelinek  <jakub@redhat.com>
13343         PR target/82464
13344         * config/s390/s390-builtins.def (s390_vec_xor_flt_a,
13345         s390_vec_xor_flt_b, s390_vec_xor_flt_c): New.
13347 2017-10-09  Richard Sandiford  <richard.sandiford@linaro.org>
13349         * wide-int.h (WI_BINARY_OPERATOR_RESULT): New macro.
13350         (WI_BINARY_PREDICATE_RESULT): Likewise.
13351         (wi::binary_traits::operator_result): New type.
13352         (wi::binary_traits::predicate_result): Likewise.
13353         (generic_wide_int::operator~, unary generic_wide_int::operator-)
13354         (generic_wide_int::operator==, generic_wide_int::operator!=)
13355         (generic_wide_int::operator&, generic_wide_int::and_not)
13356         (generic_wide_int::operator|, generic_wide_int::or_not)
13357         (generic_wide_int::operator^, generic_wide_int::operator+
13358         (binary generic_wide_int::operator-, generic_wide_int::operator*):
13359         Delete.
13360         (operator~, unary operator-, operator==, operator!=, operator&)
13361         (operator|, operator^, operator+, binary operator-, operator*): New
13362         functions.
13363         * expr.c (get_inner_reference): Use wi::bit_and_not.
13364         * fold-const.c (fold_binary_loc): Likewise.
13365         * ipa-prop.c (ipa_compute_jump_functions_for_edge): Likewise.
13366         * tree-ssa-ccp.c (get_value_from_alignment): Likewise.
13367         (bit_value_binop): Likewise.
13368         * tree-ssa-math-opts.c (find_bswap_or_nop_load): Likewise.
13369         * tree-vrp.c (zero_nonzero_bits_from_vr): Likewise.
13370         (extract_range_from_binary_expr_1): Likewise.
13371         (masked_increment): Likewise.
13372         (simplify_bit_ops_using_ranges): Likewise.
13374 2017-10-09  Martin Jambor  <mjambor@suse.cz>
13376         PR hsa/82416
13377         * hsa-common.h (hsa_op_with_type): New method extend_int_to_32bit.
13378         * hsa-gen.c (hsa_extend_inttype_to_32bit): New function.
13379         (hsa_type_for_scalar_tree_type): Use it.  Always force min32int for
13380         COMPLEX types.
13381         (hsa_fixup_mov_insn_type): New function.
13382         (hsa_op_with_type::get_in_type): Use it.
13383         (hsa_build_append_simple_mov): Likewise.  Allow sub-32bit
13384         immediates in an assert.
13385         (hsa_op_with_type::extend_int_to_32bit): New method.
13386         (gen_hsa_insns_for_bitfield): Fixup instruction and intermediary
13387         types.  Convert to dest type if necessary.
13388         (gen_hsa_insns_for_bitfield_load): Fixup load type if necessary.
13389         (reg_for_gimple_ssa): Pass false as min32int to
13390         hsa_type_for_scalar_tree_type.
13391         (gen_hsa_addr): Fixup type when creating addresable temporary.
13392         (gen_hsa_cmp_insn_from_gimple): Extend operands if necessary.
13393         (gen_hsa_unary_operation): Extend operands and convert to dest type if
13394         necessary.  Call hsa_fixup_mov_insn_type.
13395         (gen_hsa_binary_operation): Changed operand types to hsa_op_with_type,
13396         extend operands and convert to dest type if necessary.
13397         (gen_hsa_insns_for_operation_assignment): Extend operands and convert
13398         to dest type if necessary.
13399         (set_output_in_type): Call hsa_fixup_mov_insn_type.  Just ude dest
13400         if conversion nt necessary and size matches.
13401         (gen_hsa_insns_for_load): Call hsa_fixup_mov_insn_type, convert
13402         to dest type if necessary.
13403         (gen_hsa_insns_for_store): Call hsa_fixup_mov_insn_type.
13404         (gen_hsa_insns_for_switch_stmt): Likewise. Also extend operands if
13405         necessary.
13406         (gen_hsa_clrsb): Likewise.
13407         (gen_hsa_ffs): Likewise.
13408         (gen_hsa_divmod): Extend operands and convert to dest type if
13409         necessary.
13410         (gen_hsa_atomic_for_builtin): Change type of op to hsa_op_with_type.
13412 2017-10-08  Segher Boessenkool  <segher@kernel.crashing.org>
13414         * config/rs6000/rs6000.md (conditional branch): Clean up formatting.
13415         Remove empty default arguments.  Use a brace block as output
13416         statement.
13417         (conditional return): Ditto.
13418         (jump): Ditto.
13419         (indirect_jump): Ditto.  Use b%T0 instead of bctr/blr.
13420         (tablejump, tablejumpsi, tablejumpdi, *tablejump<mode>_internal1):
13421         Ditto.
13422         (group_ending_nop): Ditto.
13423         (doloop_end): Ditto.
13424         (ctr<mode>, ctr<mode>_internal1, ctr<mode>_internal2): Ditto.
13425         (splitters for those): Ditto.
13427 2017-10-08  Segher Boessenkool  <segher@kernel.crashing.org>
13429         * config/rs6000/rs6000-string.c (expand_strncmp_align_check): Invert
13430         a conditional jump (and the compare for it) so that pc_rtx is the
13431         last operand.
13432         * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Adjust
13433         for the deleted and renamed ctr<mode>_internal[234] patterns.
13434         * config/rs6000/rs6000.md: Delete second conditional branch pattern.
13435         Delete second conditional return pattern.
13436         (ctr<mode>_internal2): Delete this second bdnz pattern.
13437         (ctr<mode>_internal3): Rename to ctr<mode>_internal2.
13438         (ctr<mode>_internal4): Delete this second bdz pattern.
13440 2017-10-08  Eric Botcazou  <ebotcazou@adacore.com>
13442         * tree-outof-ssa.h (ssaexpand): Add partitions_for_undefined_values.
13443         (always_initialized_rtx_for_ssa_name_p): New predicate.
13444         * tree-outof-ssa.c (remove_ssa_form): Initialize new field of SA.
13445         (finish_out_of_ssa): Free new field of SA.
13446         * tree-ssa-coalesce.h (get_undefined_value_partitions): Declare.
13447         * tree-ssa-coalesce.c: Include tree-ssa.h.
13448         (get_parm_default_def_partitions): Remove extern keyword.
13449         (get_undefined_value_partitions): New function.
13450         * expr.c (expand_expr_real_1) <expand_decl_rtl>: For a SSA_NAME, do
13451         not set SUBREG_PROMOTED_VAR_P on the sub-register if it may contain
13452         uninitialized bits.
13453         * loop-iv.c (iv_get_reaching_def): Disqualify all subregs.
13455 2017-10-08  Eric Botcazou  <ebotcazou@adacore.com>
13457         * builtins.def (BUILT_IN_SETJMP): Revert latest change.
13459 2017-10-08  Jan Hubicka  <hubicka@ucw.cz>
13461         * config/i386/i386.c (ix86_expand_set_or_movmem): Disable 512bit loops
13462         for targets that preffer 128bit.
13464 2017-10-08  Jan Hubicka  <hubicka@ucw.cz>
13466         * config/i386/i386.c (has_dispatch): Disable for Ryzen.
13468 2017-10-08  Olivier Hainque  <hainque@adacore.com>
13470         * config/arm/arm.c (arm_set_return_address): Use MEM_VOLATILE_P
13471         on the target mem instead of RTX_FRAME_RELATED_P on the insn to
13472         prevent DSE.
13473         (thumb_set_return_address): Likewise.
13475 2017-10-08  Olivier Hainque  <hainque@adacore.com>
13477         * common/config/arm/arm-common.c (arm_except_unwind_info):
13478         Handle DWARF2_UNWIND_INFO.
13480 2017-10-07  Michael Collison <michael.collison@arm.com>
13482         * config/aarch64/aarch64.md (*aarch64_reg_<optab>_minus<mode>3):
13483         New pattern.
13485 2017-10-07  Eric Botcazou  <ebotcazou@adacore.com>
13487         * builtins.def (BUILT_IN_SETJMP): Declare as library builtin instead
13488         of GCC builtin if DONT_USE_BUILTIN_SETJMP is defined.
13489         * except.c (sjlj_emit_function_enter): If DONT_USE_BUILTIN_SETJMP is
13490         defined, force the creation of a new block for a dispatch label.
13492 2017-10-07  Jan Hubicka  <hubicka@ucw.cz>
13494         * invoke.texi (Wsuggest-attribute=cold): Document.
13495         * common.opt (Wsuggest-attribute=cold): New
13496         * ipa-pure-const.c (warn_function_cold): New function.
13497         * predict.c (compute_function_frequency): Use it.
13498         * predict.h (warn_function_cold): Declare.
13500 2017-10-06  Jan Hubicka  <hubicka@ucw.cz>
13502         * tree-switch-conversion.c (do_jump_if_equal, emit_cmp_and_jump_insns):
13503         Update profile.
13505 2017-10-06  Martin Liska  <mliska@suse.cz>
13507         * sanopt.c (struct sanopt_tree_triplet_hash): Remove inline
13508         keyword for member functions.
13509         (struct sanopt_tree_couple): New struct.
13510         (struct sanopt_tree_couple_hash): New function.
13511         (struct sanopt_ctx): Add new hash_map.
13512         (has_dominating_ubsan_ptr_check): New function.
13513         (record_ubsan_ptr_check_stmt): Likewise.
13514         (maybe_optimize_ubsan_ptr_ifn): Likewise.
13515         (sanopt_optimize_walker): Handle IFN_UBSAN_PTR.
13516         (pass_sanopt::execute): Handle also SANITIZE_POINTER_OVERFLOW.
13518 2017-10-06  Sudakshina Das  <sudi.das@arm.com>
13520         PR target/82440
13521         * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Only call
13522         aarch64_simd_valid_immediate on CONST_VECTORs.
13523         (aarch64_reg_or_bic_imm): Likewise.
13525 2017-10-06  Wilco Dijkstra  <wdijkstr@arm.com>
13527         PR rtl-optimization/82396
13528         * haifa-sched.c (ready_sort_real): Disable qsort checking.
13530 2017-10-06  Sebastian Pop  <sebpop@gmail.com>
13532         * graphite-dependences.c (scop_get_reads): Move code to...
13533         (scop_get_must_writes): Move code to...
13534         (scop_get_may_writes): Move code to...
13535         (scop_get_reads_and_writes): ... here.
13536         (scop_get_dependences): Call scop_get_reads_and_writes.
13538 2017-10-06  Jakub Jelinek  <jakub@redhat.com>
13540         PR tree-optimization/82434
13541         * fold-const.h (can_native_encode_type_p,
13542         can_native_encode_string_p): Remove.
13543         * fold-const.c (native_encode_int): Formatting fixes.  If ptr is NULL,
13544         don't encode anything, just return what would be otherwise returned.
13545         (native_encode_fixed, native_encode_complex, native_encode_vector):
13546         Likewise.
13547         (native_encode_string): Likewise.  Inline by hand
13548         can_native_encode_string_p.
13549         (can_native_encode_type_p): Remove.
13550         (can_native_encode_string_p): Remove.
13551         * tree-vect-stmts.c (vectorizable_store): Instead of testing just
13552         STRING_CSTs using can_native_encode_string_p, test all
13553         CONSTANT_CLASS_P values using native_encode_expr with NULL ptr.
13554         * gimple-ssa-store-merging.c (encode_tree_to_bitpos): Remove last
13555         argument from native_encode_expr.
13556         (rhs_valid_for_store_merging_p): Use native_encode_expr with NULL ptr.
13557         (pass_store_merging::execute): Don't unnecessarily look for 3 stmts,
13558         but just 2.
13560 2017-10-06  Richard Biener  <rguenther@suse.de>
13562         PR tree-optimization/82397
13563         * tree-vect-data-refs.c (dr_group_sort_cmp): Do not use
13564         operand_equal_p but rely on data_ref_compare_tree for detecting
13565         equalities.
13566         (vect_analyze_data_ref_accesses): Use data_ref_compare_tree
13567         to match up with dr_group_sort_cmp.
13569 2017-10-06  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
13571         PR target/82322
13572         * config/s390/s390-builtins.def (s390_vfi): Define new overloaded
13573         builtin.
13574         * config/s390/s390-builtin-types.def: Regenerate.
13576 2017-10-06  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
13578         PR target/82317
13579         * config/s390/s390-builtin-types.def: Regenerate.
13580         * config/s390/s390-builtins.def (s390_vfmaxdb_4, s390_vfmindb_4):
13581         Change flag from B_VXE to B_VX.
13582         (s390_vec_min_dbl): Remove B_VXE flag.
13584 2017-10-06  Richard Biener  <rguenther@suse.de>
13586         * graphite-isl-ast-to-gimple.c: Include ssa.h and tree-ssa.h.
13587         (translate_isl_ast_to_gimple::translate_pending_phi_nodes,
13588         translate_isl_ast_to_gimple::is_valid_rename,
13589         translate_isl_ast_to_gimple::get_rename,
13590         translate_isl_ast_to_gimple::get_def_bb_for_const,
13591         translate_isl_ast_to_gimple::get_new_name,
13592         translate_isl_ast_to_gimple::collect_all_ssa_names,
13593         translate_isl_ast_to_gimple::copy_loop_phi_args,
13594         translate_isl_ast_to_gimple::collect_all_ssa_names,
13595         translate_isl_ast_to_gimple::copy_loop_phi_args,
13596         translate_isl_ast_to_gimple::copy_loop_phi_nodes,
13597         translate_isl_ast_to_gimple::add_close_phis_to_merge_points,
13598         translate_isl_ast_to_gimple::add_close_phis_to_outer_loops,
13599         translate_isl_ast_to_gimple::copy_loop_close_phi_args,
13600         translate_isl_ast_to_gimple::copy_loop_close_phi_nodes,
13601         translate_isl_ast_to_gimple::copy_cond_phi_args,
13602         translate_isl_ast_to_gimple::copy_cond_phi_nodes,
13603         translate_isl_ast_to_gimple::edge_for_new_close_phis,
13604         translate_isl_ast_to_gimple::add_phi_arg_for_new_expr,
13605         translate_isl_ast_to_gimple::rename_uses,
13606         translate_isl_ast_to_gimple::rename_all_uses): Remove.
13607         (translate_isl_ast_to_gimple::get_rename_from_scev): Simplify.
13608         (set_rename_for_each_def): Likewise.
13609         (graphite_copy_stmts_from_block): Handle debug stmt resetting
13610         here.  Handle rewriting SCEV analyzable uses here.
13611         (copy_bb_and_scalar_dependences): Generate code for PHI
13612         copy-in/outs.
13613         (graphite_regenerate_ast_isl): Adjust.
13614         * graphite-scop-detection.c (trivially_empty_bb_p): Move to sese.[ch].
13615         (add_write, add_read): New functions.
13616         (build_cross_bb_scalars_def): Use it and simplify.
13617         (build_cross_bb_scalars_use): Likewise.
13618         (graphite_find_cross_bb_scalar_vars): Inline into...
13619         (try_generate_gimple_bb): ...here.  Add dependences for PHIs,
13620         simulating out-of-SSA.  Compute liveout and add dependencies.
13621         (build_scops): Force an empty entry block.
13622         * sese.h (sese_info_t::liveout, sese_info_t::debug_liveout): New
13623         members.
13624         (sese_build_liveouts): Declare.
13625         (sese_trivially_empty_bb_p): Likewise.
13626         * sese.c (sese_build_liveouts_bb): Properly handle PHIs,
13627         compute liveout and debug_liveout.
13628         (sese_bad_liveouts_use): Remove.
13629         (sese_reset_debug_liveouts_bb): Likewise.
13630         (sese_reset_debug_liveouts): Rewrite in terms of debug_liveout.
13631         (sese_build_liveouts): Build liveout and debug_liveout and store
13632         it in region.
13633         (new_sese_info): Adjust.
13634         (free_sese_info): Likewise.
13635         (sese_insert_phis_for_liveouts): Reset debug stmts from here,
13636         do not build liveout here.
13637         (move_sese_in_condition): Adjust region entry.
13638         (scev_analyzable_p): Match up with chrec_apply requirements.
13639         (sese_trivially_empty_bb_p): New.
13640         * tree-into-ssa.c (get_reaching_def): Properly support generating
13641         default-defs for incremental rewrite of anonymous names.
13643 2017-10-06  Richard Biener  <rguenther@suse.de>
13645         * graphite-sese-to-poly.c (extract_affine): For casts increasing
13646         precision do not perform modulo reduction.
13648 2017-10-06  Richard Biener  <rguenther@suse.de>
13650         PR tree-optimization/82436
13651         * tree-vect-slp.c (vect_supported_load_permutation_p): More
13652         conservatively choose the vectorization factor when checking
13653         whether we can perform the required load permutation.
13654         (vect_transform_slp_perm_load): Assert when we may not fail.
13656 2017-10-05  Segher Boessenkool  <segher@kernel.crashing.org>
13658         * config/rs6000/sysv4.h (SUBTARGET_OVERRIDE_OPTIONS): Correct error
13659         message for incompatible -msdata=* and -mcall-* options.
13661 2017-10-05  Jan Hubicka <hubicka@ucw.cz>
13663         * config/i386/i386.c (ia32_multipass_dfa_lookahead): Default to issue
13664         rate for post-reload scheduling.
13666 2017-10-05  Tamar Christina  <tamar.christina@arm.com>
13668         * doc/sourcebuild.texi (vect_sizes_16B_8B, vect_sizes_32B_16B): New.
13670 2017-10-05  Jan Hubicka <hubicka@ucw.cz>
13672         * config/i386/i386.c (znver1_cost): Set branch_cost to 3 (instead of 2)
13673         to improve monte carlo in scimark.
13675 2017-10-05  Jan Hubicka <hubicka@ucw.cz>
13677         * config/i386/i386.c (ix86_size_cost, i386_cost, i486_cost,
13678         pentium_cost, lakemont_cost, pentiumpro_cost, geode_cost, k6_cost,
13679         athlon_cost, k8_cost, amdfam10_cost, btver1_cost, btver2_cost,
13680         pentium4_cost, nocona_cost): Set reassociation width to 1.
13681         (bdver1_cost, bdver2_cost, bdver3_cost, bdver4_cost): Set reassociation
13682         width to 2 for fp operations and 1 otherwise.
13683         (znver1_cost): Set scalar reassoc width to 4 and vector to 3 and 6
13684         for int and fp.
13685         (atom_cost): Set reassociation width to 2.
13686         (slm_cost, generic_cost): Set fp reassociation width
13687         to 2 and 1 otherwise.
13688         (intel_cost): Set fp reassociation width to 4 and 1 otherwise.
13689         (core_cost): Set fp reassociation width to 4 and vector to 2.
13690         (ix86_reassociation_width): Rewrite using cost table; special case
13691         plus/minus on Zen; honor X86_TUNE_SSE_SPLIT_REGS
13692         and TARGET_AVX128_OPTIMAL.
13693         * config/i386/i386.h (processor_costs): Add
13694         reassoc_int, reassoc_fp, reassoc_vec_int, reassoc_vec_fp.
13695         (TARGET_VECTOR_PARALLEL_EXECUTION, TARGET_REASSOC_INT_TO_PARALLEL,
13696         TARGET_REASSOC_FP_TO_PARALLEL): Remove.
13697         * x86-tune.def (X86_TUNE_REASSOC_INT_TO_PARALLEL): Remove.
13698         (X86_TUNE_REASSOC_FP_TO_PARALLEL): Remove.
13699         (X86_TUNE_VECTOR_PARALLEL_EXECUTION):  Remove.
13701 2017-10-05  Nathan Sidwell  <nathan@acm.org>
13703         * doc/invoke.texi (Wparentheses): Document C++ MVP behaviour.
13705 2017-10-05  Tamar Christina  <tamar.christina@arm.com>
13707         * config/arm/arm.c (arm_test_fpu_data): New.
13708         (arm_run_selftests): Call arm_test_fpu_data.
13710 2017-10-04  Nathan Sidwell  <nathan@acm.org>
13712         * toplev.c (toplev::main): Remove excess parens on pretty_printer
13713         decl.
13714         * caller-save.c (insert_save): Remove excess parens on TO_SAVE parm.
13716 2017-10-04  Sudakshina Das  <sudi.das@arm.com>
13718         * config/aarch64/aarch64-protos.h (enum simd_immediate_check): New
13719         check type for aarch64_simd_valid_immediate.
13720         (aarch64_output_simd_mov_immediate): Update prototype.
13721         (aarch64_simd_valid_immediate): Update prototype.
13722         * config/aarch64/aarch64-simd.md (orr<mode>3): modified pattern to add
13723         support for ORR-immediate.
13724         (and<mode>3): modified pattern to add support for BIC-immediate.
13725         * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Function
13726         now checks for valid immediate for BIC and ORR based on new enum
13727         argument.
13728         (aarch64_output_simd_mov_immediate): Function now used to output
13729         BIC/ORR imm as well based on new enum argument.
13730         * config/aarch64/constraints.md (Do): New vector immediate constraint.
13731         (Db) : Likewise.
13732         * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): New predicate.
13733         (aarch64_reg_or_bic_imm): Likewise.
13735 2017-10-04  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
13737         * config/s390/vx-builtins.md ("vec_mergeh<mode>")
13738         ("vec_mergel<mode>"): Change mode iterator to V_128_NOSINGLE.
13740 2017-10-04  Wilco Dijkstra  <wdijkstr@arm.com>
13742         Revert r253399:
13744         PR rtl-optimization/82396
13745         * haifa-sched.c (autopref_multipass_init): Simplify
13746         initialization.
13747         (autopref_rank_data): Simplify sort order.
13748         * sched-int.h (autopref_multipass_data_): Remove
13749         multi_mem_insn_p, min_offset and max_offset.
13751 2017-10-04  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
13753         * doc/sourcebuild.texi: Document vect_peeling_profitable.
13755 2017-10-04  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
13757         * doc/sourcebuild.texi: Document vect_intdouble_cvt and
13758         vect_doubleint_cvt.
13760 2017-10-04  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
13762         * doc/sourcebuild.texi: Document vect_long_mult.
13764 2017-10-04  Richard Sandiford  <richard.sandiford@linaro.org>
13766         PR tree-optimization/82413
13767         * fold-const.c (build_range_check): Use widest_int when comparing
13768         the maximum ETYPE value with HIGH.
13770 2017-10-04  Wilco Dijkstra  <wdijkstr@arm.com>
13772         PR rtl-optimization/82396
13773         * haifa-sched.c (autopref_multipass_init): Simplify
13774         initialization.
13775         (autopref_rank_data): Simplify sort order.
13776         * sched-int.h (autopref_multipass_data_): Remove
13777         multi_mem_insn_p, min_offset and max_offset.
13779 2017-10-04  Jakub Jelinek  <jakub@redhat.com>
13781         PR tree-optimization/82381
13782         * tree-ssa-reassoc.c (sort_by_operand_rank): Check for different
13783         oeN->rank first.  Return 1 or -1 if one op is SSA_NAME and the other
13784         is not.
13786         PR tree-optimization/82374
13787         * omp-low.c (create_omp_child_function): Copy DECL_ATTRIBUTES,
13788         DECL_FUNCTION_SPECIFIC_OPTIMIZATION,
13789         DECL_FUNCTION_SPECIFIC_TARGET and DECL_FUNCTION_VERSIONED from
13790         current_function_decl to the new decl.
13792 2017-10-03  Michael Meissner  <meissner@linux.vnet.ibm.com>
13794         * config/rs6000/rs6000-builtin.def (BU_FLOAT128_2_HW): Define new
13795         helper macro for IEEE float128 hardware built-in functions.
13796         (SQRTF128_ODD): Add built-in functions with the round-to-odd
13797         semantics.
13798         (TRUNCF128_ODD): Likewise.
13799         (ADDF128_ODD): Likewise.
13800         (SUBF128_ODD): Likewise.
13801         (MULF128_ODD): Likewise.
13802         (DIVF128_ODD): Likewise.
13803         (FMAF128_ODD): Likewise.
13804         * config/rs6000/rs6000.md (UNSPEC_ROUND_TO_ODD): Rename to
13805         UNSPEC_TRUNC_ROUND_TO_ODD.
13806         (UNSPEC_TRUNC_ROUND_TO_ODD): Likewise.
13807         (UNSPEC_ADD_ROUND_TO_ODD): New unspec codes for the IEEE 128-bit
13808         floating point round to odd instructions.
13809         (UNSPEC_SUB_ROUND_TO_ODD): Likewise.
13810         (UNSPEC_MUL_ROUND_TO_ODD): Likewise.
13811         (UNSPEC_DIV_ROUND_TO_ODD): Likewise.
13812         (UNSPEC_FMA_ROUND_TO_ODD): Likewise.
13813         (UNSPEC_SQRT_ROUND_TO_ODD): Likewise.
13814         (trunc<mode>sf2_hw): Change the truncate with round to odd
13815         expansion to use UNSPEC_TRUNC_ROUND_TO_ODD.
13816         (add<mode>3_odd): Add insns for IEEE 128-bit floating point round
13817         to odd hardware instructions.
13818         (sub<mode>3_odd): Likewise.
13819         (mul<mode>3_odd): Likewise.
13820         (div<mode>3_odd): Likewise.
13821         (sqrt<mode>2_odd): Likewise.
13822         (fma<mode>4_odd): Likewise.
13823         (fms<mode>4_odd): Likewise.
13824         (nfma<mode>4_odd): Likewise.
13825         (nfms<mode>4_odd): Likewise.
13826         (trunc<mode>df2_odd): Change the truncate with round to odd
13827         expansion to use UNSPEC_TRUNC_ROUND_TO_ODD.  Add a generator
13828         function.
13829         * doc/extend.texi (PowerPC built-in functions): Update documentation
13830         for existing IEEE float128-bit built-in functions.  Add built-in
13831         functions that generate the IEEE 128-bit floating point round to
13832         odd instructions.
13834 2017-10-03  Segher Boessenkool  <segher@kernel.crashing.org>
13836         PR rtl-optimization/77729
13837         * simplify-rtx.c (simplify_binary_operation_1): Delete the (X&C1)|C2
13838         to (X&(C1&~C2))|C2 transformations.
13840 2017-10-03  Martin Jambor  <mjambor@suse.cz>
13842         PR tree-optimization/82363
13843         * tree-sra.c (propagate_subaccesses_across_link): In unrecoverable
13844         mismatch, mark lacc written regardless of racc.
13846 2017-10-03  Jakub Jelinek  <jakub@redhat.com>
13848         PR tree-optimization/82381
13849         * tree-ssa-reassoc.c (sort_by_operand_rank): Don't check
13850         stmt_to_insert nor wheather SSA_NAMEs are default defs.
13851         Return 1 or -1 if one of bba and bbb is NULL. If bb_rank is equal,
13852         fallthrough into reassoc_stmt_dominates_stmt_p.
13854         PR target/82386
13855         * combine.c (combine_instructions): Don't combine in unreachable
13856         basic blocks.
13858 2017-10-02  Peter Bergner  <bergner@vnet.ibm.com>
13860         PR target/80210
13861         * config/rs6000/rs6000.c (rs6000_option_override_internal): Rewrite
13862         function to not use the have_cpu variable.  Do not set cpu_index,
13863         rs6000_cpu_index or rs6000_tune_index if we end up using TARGET_DEFAULT
13864         or the default cpu.
13865         (rs6000_valid_attribute_p): Remove duplicate initializations of
13866         old_optimize and func_optimize.
13867         (rs6000_pragma_target_parse): Call rs6000_activate_target_options ().
13868         (rs6000_activate_target_options): Make global.
13869         * config/rs6000/rs6000-protos.h (rs6000_activate_target_options): Add
13870         prototype.
13872 2017-10-02  Jakub Jelinek  <jakub@redhat.com>
13874         * tree-dfa.c (get_ref_base_and_extent): Set *pmax_size to -1
13875         if *poffset + *pmax_size overflows in HOST_WIDE_INT.
13876         Set *poffset to 0 and *psize and *pmax_size to -1 if
13877         *poffset + *psize overflows in HOST_WIDE_INT.
13879         PR tree-optimization/82387
13880         PR tree-optimization/82388
13881         PR tree-optimization/82389
13882         * tree-ssa-dse.c (dse_classify_store): Test byte_tracking_enabled
13883         instead of live_bytes non-NULL.
13885 2017-10-02  Georg-Johann Lay  <avr@gjlay.de>
13887         PR target/41076
13888         * confg/avr/avr.md (*iorhi3.ashift8-ext.zerox): Add "r,r,0"
13889         alternative.
13891 2017-10-02  Richard Biener  <rguenther@suse.de>
13893         * graphite-isl-ast-to-gimple.c (set_codegen_error): With
13894         -fchecking and --param graphite-allow-codegen-errors=0 ICE.
13895         * params.def (PARAM_GRAPHITE_ALLOW_CODEGEN_ERRORS): New param.
13897 2017-10-02  Richard Sandiford  <richard.sandiford@linaro.org>
13899         * tree.h (wi::int_traits <const_tree>::decompose): Assert that the
13900         requested precision matches the type's.
13901         * calls.c (alloc_max_size): Calculate the new candidate size as
13902         a widest_int and use wi::to_widest when comparing it with the
13903         current candidate size.
13904         * gimple-ssa-warn-alloca.c (pass_walloca::execute): Compare with
13905         zero rather than integer_zero_node.
13906         * match.pd: Check for a no-op conversion before using wi::add
13907         rather than after.  Use tree_to_uhwi when summing small shift
13908         counts into an unsigned int.
13910 2017-10-02  Richard Sandiford  <richard.sandiford@linaro.org>
13911             Alan Hayward  <alan.hayward@arm.com>
13912             David Sherwood  <david.sherwood@arm.com>
13914         PR target/71307
13915         * config/aarch64/aarch64.h (POINTER_AND_FP_REGS): New reg class.
13916         (REG_CLASS_NAMES, REG_CLASS_CONTENTS): Update accordingly.
13917         * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
13918         POINTER_AND_FP_REGS.
13920 2017-10-02  Richard Biener  <rguenther@suse.de>
13922         PR tree-optimization/82355
13923         * graphite-isl-ast-to-gimple.c (build_iv_mapping): Also build
13924         a mapping for the enclosing loop but avoid generating one for
13925         the loop tree root.
13926         (copy_bb_and_scalar_dependences): Remove premature codegen
13927         error on PHIs in blocks duplicated into multiple places.
13928         * graphite-scop-detection.c
13929         (scop_detection::stmt_has_simple_data_refs_p): For a loop not
13930         in the region use it as loop and nest to analyze the DR in.
13931         (try_generate_gimple_bb): Likewise.
13932         * graphite-sese-to-poly.c (extract_affine_chrec): Adjust.
13933         (add_loop_constraints): For blocks in a loop not in the region
13934         create a dimension with a single iteration.
13935         * sese.h (gbb_loop_at_index): Remove assert.
13937 2017-10-01  Kevin Buettner  <kevinb@redhat.com>
13939         * omp-expand.c (adjust_context_scope): New function.
13940         (expand_parallel_call): Call adjust_context_scope.
13942 2017-10-01  Jeff Law  <law@redhat.com>
13944         * tree-ssa-dom.c (optimize_stmt): Make this a method within the
13945         dom_opt_dom_walker class with direct access to private members.
13946         Add comments.  Call test_for_singularity.
13947         (dom_opt_dom_walker::before_dom_children): Corresponding changes.
13948         (dom_opt_dom_walker::after_dom_children): Do not lazily initialize
13949         m_dummy_cond anymore.
13950         (class dom_opt_dom_walker): Initialize m_dummy_cond member in the
13951         class ctor.
13952         (pass_dominator:execute): Build the dummy_cond here and pass it
13953         to the dom_opt_dom_walker ctor.
13954         (test_for_singularity): New function.
13956 2017-09-30  Krister Walfridsson  <krister.walfridsson@gmail.com>
13957             Maya Rashish  <coypu@sdf.org>
13959         * config.gcc (*-*-netbsd*): New variable nbsd_tm_file containing
13960         netbsd.h, netbsd-stdint.h, and netbsd-elf.h.
13961         (alpha*-*-netbsd*) Use nbsd_tm_file.
13962         (arm*-*-netbsdelf*) Likewise.
13963         (i[34567]86-*-netbsdelf*) Likewise.
13964         (x86_64-*-netbsd*) Likewise.
13965         (mips*-*-netbsd*) Likewise.
13966         (powerpc-*-netbsd*) Likewise.
13967         (sh*-*-netbsd*) Likewise.
13968         (sparc-*-netbsdelf*) Likewise.
13969         (sparc64-*-netbsd*) Likewise.
13970         (m68k*-*-netbsdelf*) Use nbsd_tm_file and add CHAR_FAST8/SHORT_FAST16
13971         to tm_defines.
13972         (vax-*-netbsdelf*) Likewise.
13973         * config/netbsd-stdint.h (INT_FAST8_TYPE): Check CHAR_FAST8.
13974         (UINT_FAST8_TYPE) Likewise.
13975         (INT_FAST16_TYPE) Check CHAR_FAST16.
13976         (UINT_FAST16_TYPE) Likewise.
13978 2017-09-30  Jakub Jelinek  <jakub@redhat.com>
13980         PR target/82361
13981         * config/i386/i386.md
13982         (TARGET_USE_8BIT_IDIV zext divmodsi4 splitter): New define_split.
13983         (divmodsi4_zext_1, divmodsi4_zext_2, *divmodsi4_zext_1,
13984         *divmodsi4_zext_2): New define_insn_and_split.
13985         (*divmodsi4_noext_zext_1, *divmodsi4_noext_zext_2): New define_insn.
13986         (TARGET_USE_8BIT_IDIV zext udivmodsi4 splitter): New define_split.
13987         (udivmodsi4_zext_1, udivmodsi4_zext_2, *udivmodsi4_zext_1,
13988         *udivmodsi4_zext_2, *udivmodsi4_pow2_zext_1, *udivmodsi4_pow2_zext_2):
13989         New define_insn_and_split.
13990         (*udivmodsi4_noext_zext_1, *udivmodsi4_noext_zext_2): New define_insn.
13991         * config/i386/i386.c (ix86_split_idivmod): Handle operands[0] or
13992         operands[1] having DImode when mode is SImode.
13994         * config/i386/i386.c (ix86_split_idivmod): Use mode instead of
13995         always SImode for DIV and MOD in REG_EQUAL notes.
13997 2017-09-29  Yury Gribov  <tetra2005@gmail.com>
13999         PR middle-end/82319
14000         * match.pd: Fix handling of NaNs in pattern.
14002 2017-09-29  Jeff Law  <law@redhat.com>
14004         * sbitmap.c (bitmap_bit_in_range_p): New function.
14005         * sbitmap.h (bitmap_bit_in_range_p): Prototype.
14006         * tree-ssa-dse.c (live_bytes_read): New function.
14007         (dse_classify_store): Ignore reads of dead bytes.
14009         * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Fix
14010         typos and whitespace errors.
14011         * config/i386/predicates.md (address_no_seg_operand): Likewise.
14012         * config/s390/s390.c (s390_emit_prologue): Likewise.
14014 2017-09-29  Vladimir Makarov  <vmakarov@redhat.com>
14016         PR target/81481
14017         * ira-costs.c (scan_one_insn): Don't take into account PIC equiv
14018         with a symbol for LRA.
14020 2017-09-29  Vladimir Makarov  <vmakarov@redhat.com>
14022         PR rtl-optimization/82338
14023         * lra-constraints.c (inherit_in_ebb): Check usage_insns check.
14025 2017-09-29  Alexander Monakov  <amonakov@ispras.ru>
14027         * genmodes.c (calc_wider_mode): Suppress qsort macro.
14028         * system.h [CHECKING_P] (qsort): Redirect to qsort_chk.
14029         (qsort_chk): Declare.
14030         * vec.c [CHECKING_P] (qsort_chk_error): New static function.
14031         (qsort_chk): New function.
14033 2017-09-29  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
14035         PR tree-optimization/82337
14036         * gimple-ssa-strength-reduction.c (find_phi_def): Don't record a
14037         phi definition if the PHI result appears in an abnormal PHI.
14038         (find_basis_for_base_expr): Don't record a basis if the LHS of the
14039         basis appears in an abnormal PHI.
14041 2017-09-29  Richard Biener  <rguenther@suse.de>
14043         * graphite-isl-ast-to-gimple.c
14044         (translate_isl_ast_to_gimple::set_codegen_error): New function.
14045         (binary_op_to_tree): Use it.
14046         (get_rename_from_scev): Likewise.
14047         (copy_loop_phi_nodes): Likewise.
14048         (copy_bb_and_scalar_dependences): Likewise.
14049         (translate_pending_phi_nodes): Likewise.
14051 2017-09-29  Jakub Jelinek  <jakub@redhat.com>
14053         PR target/82339
14054         * config/i386/i386.md (*movdi_internal peephole2): New -Os peephole
14055         for movabsq $(i32 << shift), r64.
14057 2017-09-28  Uros Bizjak  <ubizjak@gmail.com>
14059         * config/i386/i386.c (ix86_print_operand_address_as): Do not check
14060         index when encoding %esp as %rsp to avoid 0x67 prefix.
14062 2017-09-28  Sergey Shalnov  <Sergey.Shalnov@intel.com>
14064         * config/i386/i386.md (*movsf_internal, *movdf_internal):
14065         Return 256-bit AVX modes for TARGET_PREFER_AVX256.
14067 2017-09-28  Thomas Preud'homme  <thomas.preudhomme@arm.com>
14069         * config/arm/arm.c (arm_option_override): Forbid ARMv8-M Security
14070         Extensions with more than 16 double VFP registers.
14071         (cmse_nonsecure_entry_clear_before_return): Remove second entry of
14072         to_clear_mask and all code related to it.  Replace the remaining
14073         entry by a sbitmap and adapt code accordingly.
14075 2017-09-28  Henry Linjamäki  <henry.linjamaki@parmance.com>
14077         * brig-builtins.def: Change pure attributes to const.
14079 2017-09-28  Joseph Myers  <joseph@codesourcery.com>
14081         * config.gcc (default_gnu_indirect_function): Default to yes for
14082         sparc*-*-linux* with glibc.
14084 2017-09-28  Joseph Myers  <joseph@codesourcery.com>
14086         * config/aarch64/aarch64.c (aarch64_elf_asm_constructor)
14087         (aarch64_elf_asm_destructor): Pass SECTION_NOTYPE to get_section
14088         when creating .init_array and .fini_array sections with priority
14089         specified.
14091 2017-09-27  Christophe Lyon  <christophe.lyon@linaro.org>
14093         PR target/71727
14094         * config/aarch64/aarch64.c
14095         (aarch64_builtin_support_vector_misalignment): Always return false
14096         when misalignment is unknown.
14098 2017-09-27  Kelvin Nilsen  <kelvin@gcc.gnu.org>
14100         * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Revise
14101         this function to return false if the definition used by the swap
14102         instruction is artificial, or if the memory address from which the
14103         constant value is loaded is not represented by a base address held
14104         in a register or if the base address register is a frame or stack
14105         pointer.  Additionally, return false if the base address of the
14106         loaded constant is a SYMBOL_REF but is not considered to be a
14107         constant.
14108         (replace_swapped_load_constant): New function.
14109         (rs6000_analyze_swaps): Add a new pass to replace a swap of a
14110         loaded constant vector with a load of a swapped constant vector.
14112 2017-09-27  Carl Love  <cel@us.ibm.com>
14114         * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
14115         (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
14116         * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
14117         fctiw instruction.
14119 2017-09-27  Alexander Monakov  <amonakov@ispras.ru>
14121         * haifa-sched.c (autopref_rank_for_schedule): Order 'irrelevant' insns
14122         first, always call autopref_rank_data otherwise.
14124 2017-09-27  Richard Biener  <rguenther@suse.de>
14126         * graphite-scop-detection.c (find_scop_parameters): Move
14127         loop bound handling ...
14128         (gather_bbs::before_dom_children): ... here, avoiding the need
14129         to build scop_info->loop_nest.
14130         (record_loop_in_sese): Remove.
14131         * sese.h (sese_info_t::loop_nest): Remove.
14132         * sese.c (new_sese_info): Do not allocate loop_nest.
14133         (free_sese_info): Do not free loop_nest.
14135 2017-09-27  Jakub Jelinek  <jakub@redhat.com>
14137         PR c++/82159
14138         * gimplify.c (gimplify_modify_expr): Don't optimize away zero sized
14139         lhs from calls if the lhs has addressable type.
14141 2017-09-27  Richard Biener  <rguenther@suse.de>
14143         * graphite.h (scop::max_alias_set): New member.
14144         * graphite-scop-detection.c: Remove references to non-existing
14145         --param in comments.
14146         (build_alias_sets): Record the maximum alias set used for drs.
14147         (build_scops): Support zero as unlimited for
14148         --param graphite-max-arrays-per-scop.
14149         * graphite-sese-to-poly.c (add_scalar_version_numbers): Remove
14150         and inline into ...
14151         (build_poly_sr_1): ... here.  Compute alias set based on the
14152         maximum alias set used for drs rather than
14153         PARAM_GRAPHITE_MAX_ARRAYS_PER_SCOP
14155 2017-09-27  Richard Biener  <rguenther@suse.de>
14157         * graphite-optimize-isl.c (get_schedule_for_node_st): Allow
14158         --param loop-block-tile-size=0 to disable tiling.
14160 2017-09-27  Richard Biener  <rguenther@suse.de>
14162         * doc/invoke.texi (graphite-max-bbs-per-function): Remove.
14163         (graphite-max-nb-scop-params): Document special value zero.
14164         * domwalk.h (dom_walker::STOP): New symbolical constant.
14165         (dom_walker::dom_walker): Add optional parameter for bb to
14166         RPO mapping.
14167         (dom_walker::~dom_walker): Declare.
14168         (dom_walker::before_dom_children): Document STOP return value.
14169         (dom_walker::m_user_bb_to_rpo): New member.
14170         (dom_walker::m_bb_to_rpo): Likewise.
14171         * domwalk.c (dom_walker::dom_walker): Compute bb to RPO
14172         mapping here if not provided by the user.
14173         (dom_walker::~dom_walker): Free bb to RPO mapping if not
14174         provided by the user.
14175         (dom_walker::STOP): Define.
14176         (dom_walker::walk): Do not compute bb to RPO mapping here.
14177         Support STOP return value from before_dom_children to stop
14178         walking.
14179         * graphite-optimize-isl.c (optimize_isl): If the schedule
14180         is the same still generate code if -fgraphite-identity
14181         or -floop-parallelize-all are given.
14182         * graphite-scop-detection.c: Include cfganal.h.
14183         (gather_bbs::gather_bbs): Get and pass through bb to RPO
14184         mapping.
14185         (gather_bbs::before_dom_children): Return STOP for BBs
14186         not in the region.
14187         (build_scops): Compute bb to RPO mapping and pass it to
14188         the domwalk.  Treat --param graphite-max-nb-scop-params=0
14189         as not limiting the number of params.
14190         * graphite.c (graphite_initialize): Remove limit on the
14191         number of basic-blocks in a function.
14192         * params.def (PARAM_GRAPHITE_MAX_BBS_PER_FUNCTION): Remove.
14193         (PARAM_GRAPHITE_MAX_NB_SCOP_PARAMS): Adjust to documented
14194         default value of 10.
14196 2017-09-26  Michael Meissner  <meissner@linux.vnet.ibm.com>
14198         * config/rs6000/vsx.md (peephole for optimizing move SF to GPR):
14199         Adjust code to eliminate needing to do the shift right 32-bits
14200         operation after XSCVDPSPN.
14202 2017-09-26  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
14204         * match.pd ((X / Y) == 0 -> X < Y): New pattern.
14205         ((X / Y) != 0 -> X >= Y): Likewise.
14207 2017-09-26  Carl Love  <cel@us.ibm.com>
14209         * config/rs6000/rs6000-c.c (P9V_BUILTIN_VEC_XL_LEN_R,
14210         P9V_BUILTIN_VEC_XST_LEN_R): Add support for builtins
14211         vector unsigned char vec_xl_len_r (unsigned char *, size_t);
14212         void vec_xst_len_r (vector unsigned char, unsigned char *, size_t);
14213         * config/rs6000/altivec.h (vec_xl_len_r, vec_xst_len_r): Add defines.
14214         * config/rs6000/rs6000-builtin.def (XL_LEN_R, XST_LEN_R): Add
14215         definitions and overloading.
14216         * config/rs6000/rs6000.c (altivec_expand_builtin): Add case
14217         statement for P9V_BUILTIN_XST_LEN_R.
14218         (altivec_init_builtins): Add def_builtin for P9V_BUILTIN_STXVLL.
14219         * config/rs6000/vsx.md (lxvll, stxvll, xl_len_r, xst_len_r): Add
14220         define_expand and define_insn for the instructions and builtins.
14221         * doc/extend.texi: Update the built-in documentation file for the new
14222         built-in functions.
14223         * config/rs6000/altivec.md (altivec_lvsl_reg, altivec_lvsr_reg): Add
14224         define_insn for the instructions
14226 2017-09-26  Krister Walfridsson  <krister.walfridsson@gmail.com>
14228         PR target/39570
14229         * gcc/config/netbsd-protos.h: New file.
14230         * gcc/config/netbsd.c: New file.
14231         * gcc/config/netbsd.h (SUBTARGET_INIT_BUILTINS): Define.
14232         * gcc/config/t-netbsd: New file.
14233         * gcc/config.gcc (tm_p_file): Add netbsd-protos.h.
14234         (tmake_file) Add t-netbsd.
14235         (extra_objs) Add netbsd.o.
14237 2017-09-26  Janus Weil  <janus@gcc.gnu.org>
14239         PR fortran/82143
14240         PR fortran/82324
14241         * doc/sourcebuild.texi: Document fortran_real_10 and fortran_real_16.
14243 2017-09-26  Michael Meissner  <meissner@linux.vnet.ibm.com>
14245         * config/rs6000/rs6000.md (extendsi<mode>2): Add a splitter to do
14246         sign extension from a vector register to a GPR by doing a 32-bit
14247         direct move and then an EXTSW.
14248         (extendsi<mode>2 splitter): Likewise.
14249         (movsi_from_sf): Adjust code to eliminate doing a 32-bit shift
14250         right or vector extract after doing XSCVDPSPN.  Use
14251         zero_extendsidi2 instead of p8_mfvsrd_4_disf to move the value to
14252         the GPRs.
14253         (movdi_from_sf_zero_ext): Likewise.
14254         (reload_gpr_from_vsxsf): Likewise.
14255         (p8_mfvsrd_4_disf): Delete, no longer used.
14256         (movsi_from_df): Optimize converting a DFmode to a SFmode, and
14257         then needing to move the SFmode to a GPR to use the XSCVDPSP
14258         instruction instead of FRSP and XSCVDPSPN.
14259         * config/rs6000/vsx.md (vsx_xscvspdp_scalar2): Move insn so that
14260         it is adjacent to the other XSCVSPDP insns.
14261         (vsx_xscvdpsp_scalar): Use "ww" constraint instead of "f" to allow
14262         SFmode to be in traditional Altivec registers.
14263         (vsx_xscvdpspn): Eliminate useless alternative constraint.
14264         (vsx_xscvspdpn): Likewise.
14265         (vsx_xscvspdpn_scalar): Likewise.
14267 2017-09-26  Martin Jambor  <mjambor@suse.cz>
14269         * tree-sra.c (compare_access_positions): Put integral types first,
14270         stabilize sorting of integral types, remove conditions putting
14271         non-full-precision integers last.
14272         (sort_and_splice_var_accesses): Disable scalarization if a
14273         non-integert would be represented by a non-full-precision integer.
14275 2017-09-26  Joseph Myers  <joseph@codesourcery.com>
14277         * config/microblaze/linux.h (TARGET_ASM_FILE_END): Likewise.
14278         * config/pa/pa.h (NEED_INDICATE_EXEC_STACK): Likewise.
14279         * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Likewise.
14280         * config/pa/pa.c (pa_hpux_file_end): Rename to pa_file_end.
14281         Define unconditionally, with [ASM_OUTPUT_EXTERNAL_REAL]
14282         conditionals inside the function instead of around it.  Call
14283         file_end_indicate_exec_stack if NEED_INDICATE_EXEC_STACK.
14284         (TARGET_ASM_FILE_END): Define unconditionally to pa_file_end.
14286 2017-09-26  Richard Biener  <rguenther@suse.de>
14288         * graphite-scop-detection.c (scop_detection::build_scop_depth): Rewrite,
14289         fold in ...
14290         (scop_detection::build_scop_breadth): ... this.  Removed.
14291         (scop_detection::loop_is_valid_in_scop): Fold into single caller.
14292         (scop_detection::harmful_stmt_in_bb): Likewise.
14293         (scop_detection::graphite_can_represent_stmt): Likewise.
14294         (scop_detection::loop_body_is_valid_scop): Likewise.  Remove recursion.
14295         (scop_detection::can_represent_loop): Remove recursion, fold in ...
14296         (scop_detection::can_represent_loop_1): ... this.  Removed.
14297         (scop_detection::harmful_loop_in_region): Simplify after inlining
14298         the above and remove more quadraticness.
14299         (build_scops): Adjust.
14300         * tree-data-ref.c (loop_nest_has_data_refs): Remove pointless
14301         quadraticness.
14303 2017-09-26  Jakub Jelinek  <jakub@redhat.com>
14305         PR target/82267
14306         * config/i386/i386.c (ix86_print_operand_address_as): Only test
14307         REGNO (base) == SP_REG if base is a REG.
14309         PR middle-end/35691
14310         * tree-ssa-reassoc.c (update_range_test): Dump r->exp each time
14311         if it is different SSA_NAME.
14312         (optimize_range_tests_cmp_bitwise): New function.
14313         (optimize_range_tests): Call it.
14315 2017-09-26  Richard Biener  <rguenther@suse.de>
14317         PR tree-optimization/82321
14318         * graphite.c (canonicalize_loop_closed_ssa): Properly check
14319         for the def being inside the loop.
14321 2017-09-26  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
14323         * config/s390/vx-builtins.md ("vmslg"): Add missing operand in
14324         assembler output.
14325         * config/s390/s390-builtins.def: Fix constraint on op4.
14327 2017-09-26  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
14329         * config/s390/s390.c (s390_expand_vec_compare): Use the new mode
14330         independent expanders.
14331         * config/s390/vector.md ("vec_cmpuneq", "vec_cmpltgt")
14332         ("vec_ordered", "vec_unordered"): New expanders.
14334 2017-09-26  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
14336         * config/s390/s390.c (s390_preferred_simd_mode): Return V4SFmode
14337         for SFmode.
14339 2017-09-26  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
14341         * config/s390/vector.md ("vec_unpacks_low_v16qi"): Rename to
14342         vec_unpacks_lo_v16qi.
14343         ("vec_unpacku_low_v16qi"): Rename to vec_unpacku_lo_v16qi.
14345 2017-09-26  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
14347         * config/s390/vector.md ("vec_unpacks_lo_v4sf")
14348         ("vec_unpacks_hi_v4sf", "vec_unpacks_lo_v2df")
14349         ("vec_unpacks_hi_v2df", "vec_pack_trunc_v2df"): New expanders.
14351 2017-09-26  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
14353         * config/s390/predicates.md ("const_shift_by_byte_operand"): New
14354         predicate.
14355         * config/s390/vector.md ("*vec_srb<mode>"): Change modes to V_128
14356         and V16QI.
14357         ("*vec_slb<mode>"): New insn pattern.
14358         ("vec_shr_<mode>"): New expander.
14359         * config/s390/vx-builtins.md ("vec_slb<mode>"): Turn into expander
14360         and force the shift count operand to V16QImode.
14361         ("vec_srb<mode>"): Set shift count mode to V16QI.
14363 2017-09-26  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
14365         * config/s390/vector.md ("vec_widen_umult_lo_<mode>")
14366         ("vec_widen_umult_hi_<mode>", "vec_widen_smult_lo_<mode>")
14367         ("vec_widen_smult_hi_<mode>"): New expander definitions.
14369 2017-09-26  Richard Earnshaw  <rearnsha@arm.com>
14371         PR target/82175
14372         * config/arm/arm.h (DRIVER_SELF_SPECS): Separate sub-rules with commas.
14374 2017-09-26  Richard Biener  <rguenther@suse.de>
14376         PR tree-optimization/82320
14377         * tree-ssa-sccvn.c (set_ssa_val_to): Changing undef to undef
14378         isn't a change.
14380 2017-09-25  Jeff Law  <law@redhat.com>
14382         * config/rs6000/rs6000-protos.h (output_probe_stack_range): Update
14383         prototype for new argument.
14384         * config/rs6000/rs6000.c (rs6000_emit_allocate_stack_1): New function,
14385         mostly extracted from rs6000_emit_allocate_stack.
14386         (rs6000_emit_probe_stack_range_stack_clash): New function.
14387         (rs6000_emit_allocate_stack): Call
14388         rs6000_emit_probe_stack_range_stack_clash as needed.
14389         (rs6000_emit_probe_stack_range): Add additional argument
14390         to call to gen_probe_stack_range{si,di}.
14391         (output_probe_stack_range): New.
14392         (output_probe_stack_range_1): Renamed from output_probe_stack_range.
14393         (output_probe_stack_range_stack_clash): New.
14394         (rs6000_emit_prologue): Emit notes into dump file as requested.
14395         * rs6000.md (allocate_stack): Handle -fstack-clash-protection.
14396         (probe_stack_range<P:mode>): Operand 0 is now early-clobbered.
14397         Add additional operand and pass it to output_probe_stack_range.
14399 2017-09-25  Bin Cheng  <bin.cheng@arm.com>
14401         PR tree-optimization/82163
14402         * tree-ssa-loop-manip.h (verify_loop_closed_ssa): New parameter.
14403         (checking_verify_loop_closed_ssa): New parameter.
14404         * tree-ssa-loop-manip.c (check_loop_closed_ssa_use): Delete.
14405         (check_loop_closed_ssa_stmt): Delete.
14406         (check_loop_closed_ssa_def, check_loop_closed_ssa_bb): New functions.
14407         (verify_loop_closed_ssa): Check loop closed ssa form for LOOP.
14408         (tree_transform_and_unroll_loop): Check loop closed ssa form only for
14409         changed loops.
14411 2017-09-25  Pekka Jaaskelainen <pekka@parmance.com>
14413         * brig-builtins.def: Treat HSAIL barrier builtins as
14414         setjmp/longjump style functions.
14416 2017-09-25  Richard Sandiford  <richard.sandiford@linaro.org>
14418         * target.def (constant_alignment): New hook.
14419         * defaults.h (CONSTANT_ALIGNMENT): Delete.
14420         * doc/tm.texi.in (CONSTANT_ALIGNMENT): Replace with...
14421         (TARGET_CONSTANT_ALIGNMENT): ...this new hook.
14422         * doc/tm.texi: Regenerate.
14423         * targhooks.h (default_constant_alignment): Declare.
14424         (constant_alignment_word_strings): Likewise.
14425         * targhooks.c (default_constant_alignment): New function.
14426         (constant_alignment_word_strings): Likewise.
14427         * builtins.c (get_object_alignment_2): Use targetm.constant_alignment
14428         instead of CONSTANT_ALIGNMENT.
14429         * varasm.c (align_variable, get_variable_align, build_constant_desc)
14430         (force_const_mem): Likewise.
14431         * config/aarch64/aarch64.h (CONSTANT_ALIGNMENT): Delete.
14432         * config/aarch64/aarch64.c (aarch64_constant_alignment): New function.
14433         (aarch64_classify_address): Call it instead of CONSTANT_ALIGNMENT.
14434         (TARGET_CONSTANT_ALIGNMENT): Redefine.
14435         * config/alpha/alpha.h (CONSTANT_ALIGNMENT): Delete commented-out
14436         definition.
14437         * config/arc/arc.h (CONSTANT_ALIGNMENT): Delete.
14438         * config/arc/arc.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14439         constant_alignment_word_strings.
14440         * config/arm/arm.h (CONSTANT_ALIGNMENT_FACTOR): Delete.
14441         (CONSTANT_ALIGNMENT): Likewise.
14442         * config/arm/arm.c (TARGET_CONSTANT_ALIGNMENT): Redefine.
14443         (arm_constant_alignment): New function.
14444         * config/bfin/bfin.h (CONSTANT_ALIGNMENT): Delete.
14445         * config/bfin/bfin.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14446         constant_alignment_word_strings.
14447         * config/cr16/cr16.h (CONSTANT_ALIGNMENT): Delete.
14448         * config/cr16/cr16.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14449         constant_alignment_word_strings.
14450         * config/cris/cris.h (CONSTANT_ALIGNMENT): Delete.
14451         * config/cris/cris.c (TARGET_CONSTANT_ALIGNMENT): Redefine.
14452         (cris_constant_alignment): New function.
14453         * config/epiphany/epiphany.h (CONSTANT_ALIGNMENT): Delete.
14454         * config/epiphany/epiphany.c (TARGET_CONSTANT_ALIGNMENT): Redefine.
14455         (epiphany_constant_alignment): New function.
14456         * config/fr30/fr30.h (CONSTANT_ALIGNMENT): Delete.
14457         * config/fr30/fr30.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14458         constant_alignment_word_strings.
14459         * config/frv/frv.h (CONSTANT_ALIGNMENT): Delete.
14460         * config/frv/frv.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14461         constant_alignment_word_strings.
14462         * config/ft32/ft32.h (CONSTANT_ALIGNMENT): Delete.
14463         * config/ft32/ft32.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14464         constant_alignment_word_strings.
14465         * config/i386/i386.h (CONSTANT_ALIGNMENT): Delete.
14466         * config/i386/i386-protos.h (ix86_constant_alignment): Delete.
14467         * config/i386/i386.c (ix86_constant_alignment): Make static.
14468         Use the same interface as the target hook.
14469         (TARGET_CONSTANT_ALIGNMENT): Redefine.
14470         * config/ia64/ia64.h (CONSTANT_ALIGNMENT): Delete.
14471         * config/ia64/ia64.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14472         constant_alignment_word_strings.
14473         * config/iq2000/iq2000.h (CONSTANT_ALIGNMENT): Delete.
14474         * config/iq2000/iq2000.c (iq2000_constant_alignment): New function.
14475         (TARGET_CONSTANT_ALIGNMENT): Redefine.
14476         * config/lm32/lm32.h (CONSTANT_ALIGNMENT): Delete.
14477         * config/lm32/lm32.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14478         constant_alignment_word_strings.
14479         * config/m32r/m32r.h (CONSTANT_ALIGNMENT): Delete.
14480         * config/m32r/m32r.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14481         constant_alignment_word_strings.
14482         * config/mcore/mcore.h (CONSTANT_ALIGNMENT): Delete.
14483         * config/mcore/mcore.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14484         constant_alignment_word_strings.
14485         * config/microblaze/microblaze.h (CONSTANT_ALIGNMENT): Delete.
14486         * config/microblaze/microblaze.c (microblaze_constant_alignment):
14487         New function.
14488         (TARGET_CONSTANT_ALIGNMENT): Redefine.
14489         * config/mips/mips.h (CONSTANT_ALIGNMENT): Delete.
14490         * config/mips/mips.c (mips_constant_alignment): New function.
14491         (TARGET_CONSTANT_ALIGNMENT): Redefine.
14492         * config/mmix/mmix.h (CONSTANT_ALIGNMENT): Delete.
14493         * config/mmix/mmix-protos.h (mmix_constant_alignment): Delete.
14494         * config/mmix/mmix.c (TARGET_CONSTANT_ALIGNMENT): Redefine.
14495         (mmix_constant_alignment): Make static.  Use the same interface
14496         as the target hook.
14497         * config/moxie/moxie.h (CONSTANT_ALIGNMENT): Delete.
14498         * config/moxie/moxie.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14499         constant_alignment_word_strings.
14500         * config/nios2/nios2.h (CONSTANT_ALIGNMENT): Delete.
14501         * config/nios2/nios2.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14502         constant_alignment_word_strings.
14503         * config/pa/pa.h (CONSTANT_ALIGNMENT): Delete.
14504         * config/pa/pa.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14505         constant_alignment_word_strings.
14506         * config/powerpcspe/powerpcspe.h (CONSTANT_ALIGNMENT): Delete.
14507         * config/powerpcspe/powerpcspe.c (TARGET_CONSTANT_ALIGNMENT): Redefine.
14508         (rs6000_constant_alignment): New function.
14509         * config/riscv/riscv.h (CONSTANT_ALIGNMENT): Delete.
14510         * config/riscv/riscv.c (riscv_constant_alignment): New function.
14511         (TARGET_CONSTANT_ALIGNMENT): Redefine.
14512         * config/rs6000/rs6000.h (CONSTANT_ALIGNMENT): Delete.
14513         * config/rs6000/rs6000.c (TARGET_CONSTANT_ALIGNMENT): Redefine.
14514         (rs6000_constant_alignment): New function.
14515         * config/s390/s390.h (CONSTANT_ALIGNMENT): Delete.
14516         * config/s390/s390.c (s390_constant_alignment): New function.
14517         (TARGET_CONSTANT_ALIGNMENT): Redefine.
14518         * config/sh/sh.h (CONSTANT_ALIGNMENT): Delete.
14519         * config/sh/sh.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14520         constant_alignment_word_strings.
14521         * config/sparc/sparc.h (CONSTANT_ALIGNMENT): Delete.
14522         * config/sparc/sparc.c (TARGET_CONSTANT_ALIGNMENT): Redefine.
14523         (sparc_constant_alignment): New function.
14524         * config/spu/spu.h (CONSTANT_ALIGNMENT): Delete.
14525         * config/spu/spu.c (spu_constant_alignment): New function.
14526         (TARGET_CONSTANT_ALIGNMENT): Redefine.
14527         * config/stormy16/stormy16.h (CONSTANT_ALIGNMENT): Delete.
14528         * config/stormy16/stormy16.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14529         constant_alignment_word_strings.
14530         * config/tilegx/tilegx.h (CONSTANT_ALIGNMENT): Delete.
14531         * config/tilegx/tilegx.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14532         constant_alignment_word_strings.
14533         * config/tilepro/tilepro.h (CONSTANT_ALIGNMENT): Delete.
14534         * config/tilepro/tilepro.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14535         constant_alignment_word_strings.
14536         * config/visium/visium.h (CONSTANT_ALIGNMENT): Delete.
14537         * config/visium/visium.c (TARGET_CONSTANT_ALIGNMENT): Redefine.
14538         (visium_constant_alignment): New function.
14539         * config/xtensa/xtensa.h (CONSTANT_ALIGNMENT): Delete.
14540         * config/xtensa/xtensa.c (TARGET_CONSTANT_ALIGNMENT): Redefine.
14541         (xtensa_constant_alignment): New function.
14542         * system.h (CONSTANT_ALIGNMENT): Poison.
14544 2017-09-25  Will Schmidt  <will_schmidt@vnet.ibm.com>
14546         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add handling
14547         for early folding of vector stores (ALTIVEC_BUILTIN_ST_*).
14548         (rs6000_builtin_valid_without_lhs): New helper function.
14549         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
14550         Remove obsoleted code for handling ALTIVEC_BUILTIN_VEC_ST.
14552 2017-09-25  Richard Sandiford  <richard.sandiford@linaro.org>
14554         * target.h (vec_perm_indices): Use unsigned short rather than
14555         unsigned char.
14556         (auto_vec_perm_indices): Likewise.
14557         * config/aarch64/aarch64.c (aarch64_vectorize_vec_perm_const_ok):
14558         Use unsigned int rather than unsigned char.
14559         * config/arm/arm.c (arm_vectorize_vec_perm_const_ok): Likewise.
14561 2017-09-25  Richard Biener  <rguenther@suse.de>
14563         * cfgloop.h (sort_sibling_loops): Declare.
14564         * cfgloop.c (sort_sibling_loops_cmp): New helper.
14565         (sort_sibling_loops): New function sorting the sibling loop list
14566         in RPO order.
14567         * graphite.c (graphite_transform_loops): Sort sibling loops.
14569 2017-09-25  Richard Sandiford  <richard.sandifird@linaro.org>
14571         * target.def (vec_perm_const_ok): Change sel parameter to
14572         vec_perm_indices.
14573         * optabs-query.c (can_vec_perm_p): Update accordingly.
14574         * doc/tm.texi: Regenerate.
14575         * config/aarch64/aarch64.c (expand_vec_perm_d): Change perm to
14576         auto_vec_perm_indices and remove separate nelt field.
14577         (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
14578         (aarch64_evpc_ext, aarch64_evpc_rev, aarch64_evpc_dup)
14579         (aarch64_evpc_tbl, aarch64_expand_vec_perm_const_1)
14580         (aarch64_expand_vec_perm_const): Update accordingly.
14581         (aarch64_vectorize_vec_perm_const_ok): Likewise.  Change sel
14582         to vec_perm_indices.
14583         * config/arm/arm.c (expand_vec_perm_d): Change perm to
14584         auto_vec_perm_indices and remove separate nelt field.
14585         (arm_evpc_neon_vuzp, arm_evpc_neon_vzip, arm_evpc_neon_vrev)
14586         (arm_evpc_neon_vtrn, arm_evpc_neon_vext, arm_evpc_neon_vtbl)
14587         (arm_expand_vec_perm_const_1, arm_expand_vec_perm_const): Update
14588         accordingly.
14589         (arm_vectorize_vec_perm_const_ok): Likewise.  Change sel
14590         to vec_perm_indices.
14591         * config/i386/i386.c (ix86_vectorize_vec_perm_const_ok): Change
14592         sel to vec_perm_indices.
14593         * config/ia64/ia64.c (ia64_vectorize_vec_perm_const_ok): Likewise.
14594         * config/mips/mips.c (mips_vectorize_vec_perm_const_ok): Likewise.
14595         * config/powerpcspe/powerpcspe.c (rs6000_vectorize_vec_perm_const_ok):
14596         Likewise.
14597         * config/rs6000/rs6000.c (rs6000_vectorize_vec_perm_const_ok):
14598         Likewise.
14600 2017-09-25  Pierre-Marie de Rodat  <derodat@adacore.com>
14602         PR debug/82155
14603         * dwarf2out.c (dwarf2out_early_global_decl): Call dwarf2out_decl
14604         on the FUNCTION_DECL function context if it has a DIE that is a
14605         declaration.
14607 2017-09-25  Richard Biener  <rguenther@suse.de>
14609         PR tree-optimization/82285
14610         * tree-vect-patterns.c (vect_recog_bool_pattern): Also handle
14611         enumeral types.
14613 2017-09-25  Tom de Vries  <tom@codesourcery.com>
14615         PR target/80035
14616         PR target/81069
14617         * config/nvptx/nvptx.c (nvptx_output_call_insn): Add exit after call to
14618         noreturn function.
14620 2017-09-25  Richard Biener  <rguenther@suse.de>
14622         * graphite-optimize-isl.c (optimize_isl): Fail and dump if
14623         ISL errors other than isl_error_quota happen.  Dump if the
14624         schedule is the same.
14625         * graphite-sese-to-poly.c (build_poly_scop): Fail on ISL
14626         errors instead of aborting inside ISL.
14628 2017-09-25  Iain Sandoe  <iain@codesourcery.com>
14630         PR target/80556
14631         * config/i386/darwin.h (REAL_LIB_SPEC): New; put libSystem ahead
14632         of libgcc_eh for m64.
14633         * config/i386/darwin64.h: Likewise.
14635 2017-09-25  Richard Biener  <rguenther@suse.de>
14637         PR middle-end/82144
14638         * dwarf2out.c (gen_enumeration_type_die): Do not add alignment
14639         attribute for incomplete types nor twice for complete ones.
14641 2017-09-24  Uros Bizjak  <ubizjak@gmail.com>
14643         PR target/82267
14644         * config/i386/i386.c (ix86_print_operand_address_as): Encode
14645         %esp as %rsp to avoid 0x67 prefix if there is no index or base
14646         register.
14648 2017-09-23  Uros Bizjak  <ubizjak@gmail.com>
14650         PR bootstrap/82306
14651         * config/i386/i386.opt (mprefer-avx256): Use
14652         ix86_target_flags variable.
14653         * config/i386/i386.c (ix86_target_string): Move
14654         -mprefer-avx256 to flag2_opts.
14656 2017-09-22  Jakub Jelinek  <jakub@redhat.com>
14658         PR middle-end/35691
14659         * match.pd: Simplify x == -1 & y == -1 into (x & y) == -1
14660         and x != -1 | y != -1 into (x & y) != -1.
14662 2017-09-22  Steve Ellcey  <sellcey@cavium.com>
14664         * config.gcc: Add new case statement to set
14665         default_gnu_indirect_function.  Remove it from x86_64-*-linux*,
14666         i[34567]86-*, powerpc*-*-linux*spe*, powerpc*-*-linux*, s390-*-linux*,
14667         s390x-*-linux* case statements.   Added aarch64 to the list of
14668         supported architectures.
14670 2017-09-22  Richard Sandiford  <richard.sandiford@linaro.org>
14672         PR tree-optimization/82289
14673         * tree-vect-data-refs.c (vect_get_peeling_costs_all_drs): Check
14674         STMT_VINFO_RELEVANT_P.
14676 2017-09-22  Richard Sandiford  <richard.sandiford@linaro.org>
14677             Alan Hayward  <alan.hayward@arm.com>
14678             David Sherwood  <david.sherwood@arm.com>
14680         * tree-vrp.c (extract_range_from_multiplicative_op_1): Assert
14681         for VR_RANGE only; don't allow VR_ANTI_RANGE.
14682         (extract_range_from_binary_expr_1): Don't call
14683         extract_range_from_multiplicative_op_1 if !range_int_cst_p.
14685 2017-09-22  Richard Sandiford  <richard.sandiford@linaro.org>
14686             Alan Hayward  <alan.hayward@arm.com>
14687             David Sherwood  <david.sherwood@arm.com>
14689         * target.def (preferred_vector_alignment): New hook.
14690         * doc/tm.texi.in (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): New
14691         hook.
14692         * doc/tm.texi: Regenerate.
14693         * targhooks.h (default_preferred_vector_alignment): Declare.
14694         * targhooks.c (default_preferred_vector_alignment): New function.
14695         * tree-vectorizer.h (dataref_aux): Add a target_alignment field.
14696         Expand commentary.
14697         (DR_TARGET_ALIGNMENT): New macro.
14698         (aligned_access_p): Update commentary.
14699         (vect_known_alignment_in_bytes): New function.
14700         * tree-vect-data-refs.c (vect_calculate_required_alignment): New
14701         function.
14702         (vect_compute_data_ref_alignment): Set DR_TARGET_ALIGNMENT.
14703         Calculate the misalignment based on the target alignment rather than
14704         the vector size.
14705         (vect_update_misalignment_for_peel): Use DR_TARGET_ALIGMENT
14706         rather than TYPE_ALIGN / BITS_PER_UNIT to update the misalignment.
14707         (vect_enhance_data_refs_alignment): Mask the byte misalignment with
14708         the target alignment, rather than masking the element misalignment
14709         with the number of elements in a vector.  Also use the target
14710         alignment when calculating the maximum number of peels.
14711         (vect_find_same_alignment_drs): Use vect_calculate_required_alignment
14712         instead of TYPE_ALIGN_UNIT.
14713         (vect_duplicate_ssa_name_ptr_info): Remove stmt_info parameter.
14714         Measure DR_MISALIGNMENT relative to DR_TARGET_ALIGNMENT.
14715         (vect_create_addr_base_for_vector_ref): Update call accordingly.
14716         (vect_create_data_ref_ptr): Likewise.
14717         (vect_setup_realignment): Realign by ANDing with
14718         -DR_TARGET_MISALIGNMENT.
14719         * tree-vect-loop-manip.c (vect_gen_prolog_loop_niters): Calculate
14720         the number of peels based on DR_TARGET_ALIGNMENT.
14721         * tree-vect-stmts.c (get_group_load_store_type): Compare the gap
14722         with the guaranteed alignment boundary when deciding whether
14723         overrun is OK.
14724         (vectorizable_mask_load_store): Interpret DR_MISALIGNMENT
14725         relative to DR_TARGET_ALIGNMENT instead of TYPE_ALIGN_UNIT.
14726         (ensure_base_align): Remove stmt_info parameter.  Get the
14727         target base alignment from DR_TARGET_ALIGNMENT.
14728         (vectorizable_store): Update call accordingly.   Interpret
14729         DR_MISALIGNMENT relative to DR_TARGET_ALIGNMENT instead of
14730         TYPE_ALIGN_UNIT.
14731         (vectorizable_load): Likewise.
14733 2017-09-22  Richard Sandiford  <richard.sandiford@linaro.org>
14734             Alan Hayward  <alan.hayward@arm.com>
14735             David Sherwood  <david.sherwood@arm.com>
14737         * tree-vectorizer.h (vect_get_scalar_dr_size): New function.
14738         * tree-vect-data-refs.c (vect_update_misalignment_for_peel): Use it.
14739         (vect_enhance_data_refs_alignment): Likewise.
14741 2017-09-22  Richard Earnshaw  <richard.earnshaw@arm.com>
14743         * config/arm/parsecpu.awk (fatal): Note that we've encountered an
14744         error.  Only quit immediately if parsing is complete.
14745         (BEGIN): Initialize fatal_err and parse_done.
14746         (begin fpu, end fpu): Check number of arguments.
14747         (begin arch, end arch): Likewise.
14748         (begin cpu, end cpu): Likewise.
14749         (cname, tune for, tune flags, architecture, fpu, option): Likewise.
14750         (optalias): Likewise.
14752 2017-09-22  Richard Earnshaw  <richard.earnshaw@arm.com>
14754         * config.gcc (arm*-*-*): Don't add arm-isa.h to tm_p_file.
14755         * config/arm/arm-isa.h: Delete.  Move definitions to ...
14756         * arm-cpus.in: ... here.  Use new feature and fgroup values.
14757         * config/arm/arm.c (arm_option_override): Use lower case for feature
14758         bit names.
14759         * config/arm/arm.h (TARGET_HARD_FLOAT): Likewise.
14760         (TARGET_VFP3, TARGET_VFP5, TARGET_FMA): Likewise.
14761         * config/arm/parsecpu.awk (END): Add new command 'isa'.
14762         (isa_pfx): Delete.
14763         (print_isa_bits_for): New function.
14764         (gen_isa): New function.
14765         (gen_comm_data): Use print_isa_bits_for.
14766         (define feature): New keyword.
14767         (define fgroup): New keyword.
14768         * config/arm/t-arm (TM_H): Remove.
14769         (GTM_H): Add arm-isa.h.
14770         (arm-isa.h): Add rule to generate file.
14771         * common/config/arm/arm-common.c: (arm_canon_arch_option): Use lower
14772         case for feature bit names.
14774 2017-09-22  Richard Biener  <rguenther@suse.de>
14776         * graphite-isl-ast-to-gimple.c (graphite_verify): Inline into
14777         single caller.
14778         (graphite_regenerate_ast_isl): Do not reset SCEV.  Move debug
14779         print of no dependency loops ...
14780         * graphite.c (graphite_transform_loops): ... here.
14781         (canonicalize_loop_closed_ssa_form): Work from inner to outer
14782         loops.
14783         (same_close_phi_node, remove_duplicate_close_phi,
14784         make_close_phi_nodes_unique, defined_in_loop_p): Fold into ...
14785         (canonicalize_loop_closed_ssa): ... here and simplify.
14786         * graphite-optimize-isl.c: Include tree-vectorizer.h.
14787         (optimize_isl): Use dump_printf_loc to tell when we stopped
14788         optimizing because of an ISL timeout.
14790 2017-09-22  Richard Biener  <rguenther@suse.de>
14792         PR tree-optimization/82291
14793         * tree-if-conv.c (predicate_mem_writes): Make sure to
14794         remove writes in blocks predicated with false.
14796 2017-09-22  Richard Biener  <rguenther@suse.de>
14798         * sese.c: Include cfganal.h.
14799         (if_region_set_false_region): Remove.
14800         (create_if_region_on_edge): Likewise.
14801         (move_sese_in_condition): Re-implement without destroying
14802         dominators.
14804 2017-09-22  Richard Biener  <rguenther@suse.de>
14806         * graphite-isl-ast-to-gimple.c (translate_pending_phi_nodes):
14807         Verify both BBs contain loop PHI nodes before dispatching to
14808         copy_loop_phi_args.
14809         (graphite_regenerate_ast_isl): Do not recompute dominators,
14810         do not verify three times.  Restructure for clarity.
14811         * graphite-scop-detection.c (same_close_phi_node,
14812         remove_duplicate_close_phi, make_close_phi_nodes_unique,
14813         defined_in_loop_p, canonicalize_loop_closed_ssa,
14814         canonicalize_loop_closed_ssa_form): Simplify, remove excess
14815         checking and SSA rewrite, move to ...
14816         * graphite.c: ... here.  Include ssa.h and tree-ssa-loop-manip.h.
14817         (graphite_initialize): Do not pass in ctx, do not reset the
14818         SCEV cache, compute only dominators.
14819         (graphite_transform_loops): Allocate ISL ctx after
14820         graphite_initialize.  Call canonicalize_loop_closed_ssa_form.
14821         Maintain post-dominators only around build_scops.
14822         * sese.c (if_region_set_false_region): Make static.  Free
14823         and recompute dominators.
14824         (move_sese_in_condition): Assert we don't get called with
14825         post-dominators computed.
14826         * sese.h (if_region_set_false_region): Remove.
14828 2017-09-22  Sergey Shalnov  <sergey.shalnov@intel.com>
14830         * config/i386/sse.md ("mov<mode>_internal"): Use <sseinsnmode>
14831         mode attribute for TARGET_AVX512VL.
14833 2017-09-21  Sergey Shalnov  <sergey.shalnov@intel.com>
14835         * config/i386/i386.opt (mprefer-avx256): New option.
14836         * config/i386/i386.c (ix86_target_string): Add -mprefer-avx256
14837         to flag_opts.
14838         (ix86_preferred_simd_mode): Return 256-bit AVX modes
14839         for TARGET_PREFER_AVX256.
14840         * doc/invoke.texi (x86 Options): Document -mprefer-avx256.
14842 2017-09-21  Jeff Law  <law@redhat.com>
14844         * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash):
14845         Fix dump output if the only stack space is for pushed registers.
14847 2017-09-21  Richard Sandiford  <richard.sandiford@linaro.org>
14849         * config/spu/spu.c (spu_sched_adjust_cost): Update after renaming
14850         of insn_cost.
14852 2017-09-21  Martin Sebor  <msebor@redhat.com>
14854         PR c/81882
14855         * doc/extend.texi (attribute ifunc): Avoid relying on ill-formed
14856         code (in C++) or code that triggers warnings.
14858 2017-09-21  Eric Botcazou  <ebotcazou@adacore.com>
14860         * stor-layout.c (bit_from_pos): Do not distribute the conversion.
14862 2017-09-21  Segher Boessenkool  <segher@kernel.crashing.org>
14864         * haifa-sched.c: Rename insn_cost to insn_sched_cost.
14865         * sched-rgn.c: Ditto.
14866         * sel-sched-ir.c: Ditto.
14868 2017-09-21  Alexander Monakov  <amonakov@ispras.ru>
14870         * toplev.h (set_random_seed): Adjust return type.
14871         * toplev.c (init_local_tick): Move eager initialization of random_seed
14872         to get_random_seed.  Adjust comment.
14873         (init_random_seed): Inline to get_random_seed, delete.
14874         (get_random_seed): Initialize random_seed lazily.
14875         (set_random_seed): Do not return previous value.
14876         (print_switch_value): Do not call get_random_seed.
14878 2017-09-21  Evgeny Kudryashov  <kudryashov@ispras.ru>
14880         * cgraph.c (delete_function_version): New, broken out from...
14881         (cgraph_node::delete_function_version): ...here.  Rename to
14882         cgraph_node::delete_function_version_by_decl.  Update all uses.
14883         (cgraph_node::remove): Call delete_function_version.
14885 2017-09-21  Jakub Jelinek  <jakub@redhat.com>
14887         PR sanitizer/81715
14888         * tree-inline.c (expand_call_inline): Emit clobber stmts for
14889         VAR_DECLs to which addressable non-volatile parameters are mapped
14890         and for id->retvar after the return value assignment.  Clear
14891         id->retval and id->retbnd after inlining.
14893 2017-09-21  Richard Biener  <rguenther@suse.de>
14895         PR tree-optimization/82276
14896         PR tree-optimization/82244
14897         * tree-vrp.c (build_assert_expr_for): Set
14898         SSA_NAME_OCCURS_IN_ABNORMAL_PHI if the variable we assert on
14899         has it set.
14900         (remove_range_assertions): Revert earlier change.
14902 2017-09-21  Wilco Dijkstra  <wdijkstr@arm.com>
14904         PR target/71951
14905         * config/aarch64/aarch64.h (LIBGCC2_UNWIND_ATTRIBUTE): Define.
14907 2017-09-21  Richard Biener  <rguenther@suse.de>
14909         * graphite-isl-ast-to-gimple.c (graphite_regenerate_ast_isl):
14910         Restore valid IL after code generation errors.
14911         * graphite.c (graphite_transform_loops): Diagnose code
14912         generation issues as MSG_MISSED_OPTIMIZATION and continue
14913         with processing SCOPs.
14915 2017-09-21  Richard Sandiford  <richard.sandiford@linaro.org>
14916             Alan Hayward  <alan.hayward@arm.com>
14917             David Sherwood  <david.sherwood@arm.com>
14919         * calls.c (compute_argument_addresses): Use simplify_gen_binary
14920         rather than choosing between plus_constant and gen_rtx_<CODE>.
14921         * expr.c (emit_push_insn): Likewise.
14922         (expand_expr_real_2): Likewise.
14924 2017-09-21  Richard Sandiford  <richard.sandiford@linaro.org>
14925             Alan Hayward  <alan.hayward@arm.com>
14926             David Sherwood  <david.sherwood@arm.com>
14928         * loop-unroll.c (split_iv): Call copy_rtx on the step.
14930 2017-09-21  Richard Sandiford  <richard.sandiford@linaro.org>
14931             Alan Hayward  <alan.hayward@arm.com>
14932             David Sherwood  <david.sherwood@arm.com>
14934         * tree.c (find_atomic_core_type): Check tree_fits_uhwi_p before
14935         calling tree_to_uhwi.
14937 2017-09-21  Richard Sandiford  <richard.sandiford@linaro.org>
14938             Alan Hayward  <alan.hayward@arm.com>
14939             David Sherwood  <david.sherwood@arm.com>
14941         * tree-ssa-ccp.c (get_value_for_expr): Use a positive test for
14942         INTEGER_CST rather than a negative test for ADDR_EXPR.
14944 2017-09-21  Richard Sandiford  <richard.sandiford@linaro.org>
14945             Alan Hayward  <alan.hayward@arm.com>
14946             David Sherwood  <david.sherwood@arm.com>
14948         * tree-vrp.c (extract_range_from_binary_expr_1): Check
14949         int_cst_rangeN before calling value_range_constant_singleton (&vrN).
14951 2017-09-21  Richard Biener  <rguenther@suse.de>
14953         PR tree-optimization/71351
14954         * graphite-isl-ast-to-gimple.c (translate_isl_ast_to_gimple::
14955         graphite_create_new_loop_guard): Remove, fold remaining parts
14956         into caller ...
14957         (translate_isl_ast_node_for): ... here and simplify.
14959 2017-09-21  Jakub Jelinek  <jakub@redhat.com>
14961         PR target/82260
14962         * config/i386/i386.md (*movqi_internal): Replace (=q,q) alternative
14963         with (=Q,Q), (=R,R) and (=r,r) alternatives, only enable the
14964         latter two for 64-bit, renumber alternatives, for -Os imov (=q,n)
14965         alternative always use QI mode, for -Os imov (=R,R) alternative
14966         always use SI mode, for imov (=Q,Q) or (=r,r) alternatives
14967         ignore -Os.
14969 2017-09-20  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
14970             Jeff Law  <law@redhat.com>
14972         * config/s390/s390.c (MIN_UNROLL_PROBES): Define.
14973         (allocate_stack_space): New function, partially extracted from
14974         s390_emit_prologue.
14975         (s390_emit_prologue): Track offset to most recent stack probe.
14976         Code to allocate space moved into allocate_stack_space.
14977         Dump actions when no stack is allocated.
14978         (s390_prologue_plus_offset): New function.
14979         (s390_emit_stack_probe): Likewise.
14981 2017-09-20  Alexandre Oliva  <aoliva@redhat.com>
14983         * common.opt (Wa, Wl, Wp, g, gz=): Add
14984         RejectNegative.
14985         (gno-column-info): Remove.
14986         (gcolumn-info): Drop RejectNegative.
14987         (gno-): New prefix.
14988         (gno-record-gcc-switches): Remove.
14989         (grecord-gcc-switches): Drop RejectNegative.
14990         (gno-split-dwarf): Remove.
14991         (gsplit-dwarf): Drop RejectNegative.
14992         (gno-strict-dwarf): Remove.
14993         (gstrict-dwarf): Drop RejectNegative.
14994         * config/darwin.opt (gfull, gused): Add RejectNegative.
14995         * dwarf2out.c (gen_producer_string): Drop
14996         gno-record-gcc-switches handler.
14997         * optc-gen.awk: Add g to prefixes with negative forms.
14998         * opts-common.c (remapping_prefix_p): New.
14999         (find_opt): Check it.
15000         (generate_canonical_option): Test g prefix.
15001         (option_map): Add -gno- mapping.
15002         (add_misspelling_candidates): Check remapping_prefix_p.
15004 2017-09-20  Jeff Law  <law@redhat.com>
15006         * config/powerpcspe/powerpcspe.c (rs6000_expand_prologue): Fix
15007         thinko in stack clash protection support.
15009         * explow.c (compute_stack_clash_protection_loop_data): Use
15010         CONST_INT_P instead of explicit test.  Verify object is a
15011         CONST_INT_P before looking at INTVAL.
15012         (anti_adjust_stack_and_probe_stack_clash): Use CONST_INT_P
15013         instead of explicit test.
15015 2017-09-20  Segher Boessenkool  <segher@kernel.crashing.org>
15017         PR target/77687
15018         * config/rs6000/rs6000.md (stack_restore_tie): Store to a scratch
15019         address instead of to r1 and r11.
15021 2017-09-20  Sebastian Peryt  <sebastian.peryt@intel.com>
15023         * config.gcc: Support "knm".
15024         * config/i386/driver-i386.c (host_detect_local_cpu): Detect "knm".
15025         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
15026         PROCESSOR_KNM.
15027         * config/i386/i386.c (m_KNM): Define.
15028         (processor_target_table): Add "knm".
15029         (PTA_KNM): Define.
15030         (ix86_option_override_internal): Add "knm".
15031         (ix86_issue_rate): Add PROCESSOR_KNM.
15032         (ix86_adjust_cost): Ditto.
15033         (ia32_multipass_dfa_lookahead): Ditto.
15034         (get_builtin_code_for_version): Handle PROCESSOR_KNM.
15035         (fold_builtin_cpu): Add M_INTEL_KNM.
15036         * config/i386/i386.h (processor_costs): Define TARGET_KNM.
15037         (processor_type): Add PROCESSOR_KNM.
15038         * config/i386/x86-tune.def: Add m_KNM.
15039         * doc/invoke.texi: Add knm as x86 -march=/-mtune= CPU type.
15041 2017-09-20  Richard Biener  <rguenther@suse.de>
15043         PR tree-optimization/80213
15044         * graphite-scop-detection.c (trivially_empty_bb_p): Labels
15045         are allowed in empty BBs as well.
15046         (canonicalize_loop_closed_ssa): Also look for other complex
15047         edges.
15048         (scop_detection::get_sese): Include the loop-closed PHI block
15049         in loop SESEs.
15050         (scop_detection::merge_sese): Remove code adding extra blocks.
15051         (scop_detection::region_has_one_loop): Adjust for get_sese changes.
15052         (build_scops): Assert the final returned scop is invalid.
15054 2017-09-20  Richard Biener  <rguenther@suse.de>
15056         PR tree-optimization/82264
15057         * tree-ssa-sccvn.c (vn_phi_eq): Use safe_dyn_cast to check
15058         for GIMPLE_CONDs.
15059         (vn_phi_lookup): Likewise.
15060         (vn_phi_insert): Likewise.
15062 2017-09-20  Jakub Jelinek  <jakub@redhat.com>
15064         * dwarf2out.c (tree_add_const_value_attribute): For INTEGER_CST
15065         that fits into uhwi or shwi, add DW_AT_const_value regardless
15066         of early_dwarf without going through RTL, using add_AT_unsigned
15067         or add_AT_int.
15069         * dwarf2out.c (DEBUG_LTO_DWO_INFO_SECTION): Reorder defines.
15070         (DEBUG_LTO_ABBREV_SECTION): Likewise.
15071         (DEBUG_LTO_MACINFO_SECTION): Likewise.
15072         (DEBUG_MACRO_SECTION): Likewise.
15073         (DEBUG_LTO_MACRO_SECTION): Likewise.
15074         (DEBUG_STR_DWO_SECTION): Likewise.
15075         (DEBUG_LTO_STR_DWO_SECTION): Likewise.
15076         (DEBUG_LTO_LINE_SECTION): Drop .dwo suffix from the name.
15077         (DEBUG_LTO_DWO_LINE_SECTION): Define.
15078         (DEBUG_LTO_LINE_STR_SECTION): Define.
15079         (init_sections_and_labels): Initialize debug_line_str_section
15080         variable.  Initialize debug_loc_section for -gdwarf-5 to
15081         DEBUG_LOCLISTS_SECTION.  Formatting fixes.
15083 2017-09-20  Richard Biener  <rguenther@suse.de>
15085         * graphite-sese-to-poly.c (extract_affine): Properly handle
15086         POINTER_PLUS_EXPR, BIT_NOT_EXPR and conversion to signed.
15088 2017-09-20  Richard Biener  <rguenther@suse.de>
15090         PR tree-optimization/81373
15091         * graphite-scop-detection.c (build_cross_bb_scalars_def):
15092         Force SESE live-out defs to be handled even if they are
15093         scev_analyzable_p.
15095 2017-09-19  Jeff Law  <law@redhat.com>
15097         * combine-stack-adj.c (combine_stack_adjustments_for_block): Do
15098         nothing for stack adjustments with REG_STACK_CHECK.
15099         * sched-deps.c (parse_add_or_inc): Reject insns with
15100         REG_STACK_CHECK from dependency breaking.
15101         * config/i386/i386.c (pro_epilogue_adjust_stack): Return insn.
15102         (ix86_adjust_satck_and_probe_stack_clash): Add REG_STACK_NOTEs.
15103         * reg-notes.def (STACK_CHECK): New note.
15105         * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): New.
15106         (ix86_expand_prologue): Dump stack clash info as needed.
15107         Call ix86_adjust_stack_and_probe_stack_clash as needed.
15109         * function.c (dump_stack_clash_frame_info): New function.
15110         * function.h (dump_stack_clash_frame_info): Prototype.
15111         (enum stack_clash_probes): New enum.
15113         * config/alpha/alpha.c (alpha_expand_prologue): Also check
15114         flag_stack_clash_protection.
15115         * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Likewise.
15116         (arm_expand_prologue, thumb1_expand_prologue): Likewise.
15117         (arm_frame_pointer_required): Likewise.
15118         * config/ia64/ia64.c (ia64_compute_frame_size): Likewise.
15119         (ia64_expand_prologue): Likewise.
15120         * config/mips/mips.c (mips_expand_prologue): Likewise.
15121         * config/powerpcspe/powerpcspe.c (rs6000_expand_prologue): Likewise.
15122         * config/sparc/sparc.c (sparc_expand_prologue): Likewise.
15123         (sparc_flat_expand_prologue): Likewise.
15124         * config/spu/spu.c (spu_expand_prologue): Likewise.
15126         * explow.c: Include "params.h".
15127         (anti_adjust_stack_and_probe_stack_clash): New function.
15128         (get_stack_check_protect): Likewise.
15129         (compute_stack_clash_protection_loop_data): Likewise.
15130         (emit_stack_clash_protection_loop_start): Likewise.
15131         (emit_stack_clash_protection_loop_end): Likewise.
15132         (allocate_dynamic_stack_space): Use get_stack_check_protect.
15133         Use anti_adjust_stack_and_probe_stack_clash.
15134         * explow.h (compute_stack_clash_protection_loop_data): Prototype.
15135         (emit_stack_clash_protection_loop_start): Likewise.
15136         (emit_stack_clash_protection_loop_end): Likewise.
15137         * rtl.h (get_stack_check_protect): Prototype.
15138         * target.def (stack_clash_protection_final_dynamic_probe): New hook.
15139         * targhooks.c (default_stack_clash_protection_final_dynamic_probe): New.
15140         * targhooks.h (default_stack_clash_protection_final_dynamic_probe):
15141         Prototype.
15142         * doc/tm.texi.in (TARGET_STACK_CLASH_PROTECTION_FINAL_DYNAMIC_PROBE):
15143         Add @hook.
15144         * doc/tm.texi: Rebuilt.
15145         * config/aarch64/aarch64.c (aarch64_expand_prologue): Use
15146         get_stack_check_protect.
15147         * config/alpha/alpha.c (alpha_expand_prologue): Likewise.
15148         * config/arm/arm.c (arm_expand_prologue): Likewise.
15149         (arm_frame_pointer_required): Likewise.
15150         * config/i386/i386.c (ix86_expand_prologue): Likewise.
15151         * config/ia64/ia64.c (ia64_expand_prologue): Likewise.
15152         * config/mips/mips.c (mips_expand_prologue): Likewise.
15153         * config/powerpcspe/powerpcspe.c (rs6000_emit_prologue): Likewise.
15154         * config/rs6000/rs6000.c (rs6000_emit_prologue): Likewise.
15155         * config/sparc/sparc.c (sparc_expand_prologue): Likewise.
15156         (sparc_flat_expand_prologue): Likewise.
15158         * common.opt (-fstack-clash-protection): New option.
15159         * flag-types.h (enum stack_check_type): Note difference between
15160         -fstack-check= and -fstack-clash-protection.
15161         * params.def (PARAM_STACK_CLASH_PROTECTION_GUARD_SIZE): New PARAM.
15162         (PARAM_STACK_CLASH_PROTECTION_PROBE_INTERVAL): Likewise.
15163         * toplev.c (process_options): Issue warnings/errors for cases
15164         not handled with -fstack-clash-protection.
15165         * doc/invoke.texi (-fstack-clash-protection): Document new option.
15166         (-fstack-check): Note additional problem with -fstack-check=generic.
15167         Note that -fstack-check is primarily for Ada and refer users
15168         to -fstack-clash-protection for stack-clash-protection.
15169         Document new params for stack clash protection.
15171 2017-09-19  Uros Bizjak  <ubizjak@gmail.com>
15173         * config/i386/i386.c (ix86_split_long_move): Do not handle
15174         address used for LEA in a special way.
15176 2017-09-19  Segher Boessenkool  <segher@kernel.crashing.org>
15178         * simplify-rtx.c (simplify_binary_operation_1): Fix typo in comment.
15180 2017-09-19  Martin Sebor  <msebor@redhat.com>
15182         PR c/81854
15183         * cgraphunit.c (handle_alias_pairs): Reject aliases between functions
15184         of incompatible types.
15186 2017-09-19  Will Schmidt  <will_schmidt@vnet.ibm.com>
15188         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add handling
15189         for early folding of vector loads (ALTIVEC_BUILTIN_LVX_*).
15190         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
15191         Remove obsoleted code for handling ALTIVEC_BUILTIN_VEC_LD.
15193 2017-09-19  Richard Biener  <rguenther@suse.de>
15195         PR tree-optimization/82244
15196         * tree-vrp.c (remove_range_assertions): Do not propagate
15197         a constant to abnormals but replace the assert with a copy.
15199 2017-09-19  Alexander Monakov  <amonakov@ispras.ru>
15201         PR rtl-optimization/57878
15202         PR rtl-optimization/68988
15203         * lra-assigns.c (reload_pseudo_compare_func): Remove fragmentation
15204         avoidance test involving non_reload_pseudos.  Move frequency test
15205         below the general fragmentation avoidance test.
15207 2017-09-19  Richard Biener  <rguenther@suse.de>
15209         PR tree-optimization/69728
15210         * graphite-sese-to-poly.c (schedule_error): New global.
15211         (add_loop_schedule): Handle empty domain by failing the
15212         schedule.
15213         (build_original_schedule): Handle schedule_error.
15215 2017-09-19  Richard Biener  <rguenther@suse.de>
15217         * graphite-scop-detection.c (scop_detection::can_represent_loop):
15218         Do not iterate to sibling loops but only to siblings of inner
15219         loops.
15221 2017-09-18  Andreas Schwab  <schwab@linux-m68k.org>
15223         PR target/81613
15224         * config/m68k/m68k.md (moveq feeding equality comparison): Check
15225         that the registers are different.
15227 2017-09-18  Uros Bizjak  <ubizjak@gmail.com>
15229         * config/i386/i386.c (fold_builtin_cpu): Add M_AMDFAM17H
15230         to processor_model and "amdfam17h" to arch_names_table.
15231         * doc/extend.texi (__builtin_cpu_is): Document amdfam17h CPU name.
15233 2017-09-18  Jakub Jelinek  <jakub@redhat.com>
15235         PR c/82234
15236         * doc/extend.texi: Add @findex entry for __builtin_shuffle.
15238 2017-09-18  Richard Sandiford  <richard.sandiford@linaro.org>
15239             Alan Hayward  <alan.hayward@arm.com>
15240             David Sherwood  <david.sherwood@arm.com>
15242         * tree-vectorizer.h (vect_slp_analyze_operations): Replace parameters
15243         with a vec_info *.
15244         * tree-vect-loop.c (vect_analyze_loop_operations): Update call
15245         accordingly.
15246         * tree-vect-slp.c (vect_slp_analyze_node_operations): Add vec_info *
15247         parameter.  Set SLP_TREE_NUMBER_OF_VEC_STMTS here rather than in
15248         vect_schedule_slp_instance.
15249         (vect_slp_analyze_operations): Replace parameters with a vec_info *.
15250         Update call to vect_slp_analyze_node_operations.  Simplify return
15251         value.
15252         (vect_slp_analyze_bb_1): Update call accordingly.
15253         (vect_schedule_slp_instance): Remove vectorization_factor parameter.
15254         Don't calculate SLP_TREE_NUMBER_OF_VEC_STMTS here.
15255         (vect_schedule_slp): Update call accordingly.
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-data-refs.c (vect_get_smallest_scalar_type): Cope
15262         with types that aren't in fact scalar.
15264 2017-09-18  Richard Sandiford  <richard.sandiford@linaro.org>
15266         * tree-vect-slp.c (vect_record_max_nunits): New function,
15267         split out from...
15268         (vect_build_slp_tree_1): ...here.
15269         (vect_build_slp_tree_2): Call it for phis too.
15271 2017-09-18  Richard Sandiford  <richard.sandiford@linaro.org>
15273         * tree-vect-stmts.c (vectorizable_mask_load_store): Pass mask_vectype
15274         to vect_get_vec_def_for_operand when getting the mask operand.
15276 2017-09-18  Richard Sandiford  <richard.sandiford@linaro.org>
15277             Alan Hayward  <alan.hayward@arm.com>
15278             David Sherwood  <david.sherwood@arm.com>
15280         * tree-vect-loop.c (vectorizable_live_operation): Fix type of
15281         bitstart.
15283 2017-09-18  Richard Sandiford  <richard.sandiford@linaro.org>
15284             Alan Hayward  <alan.hayward@arm.com>
15285             David Sherwood  <david.sherwood@arm.com>
15287         * tree-vect-loop.c (vectorizable_live_operation): Fix element size
15288         calculation for vector booleans.
15290 2017-09-18  Richard Sandiford  <richard.sandiford@linaro.org>
15291             Alan Hayward  <alan.hayward@arm.com>
15292             David Sherwood  <david.sherwood@arm.com>
15294         * tree-vect-stmts.c (can_vectorize_live_stmts): New function,
15295         split out from...
15296         (vect_transform_stmt): ...here.
15297         (vect_analyze_stmt): Use it instead of calling
15298         vectorizable_live_operation directly.
15300 2017-09-18  Cesar Philippidis  <cesar@codesourcery.com>
15302         * omp-offload.c (oacc_xform_loop): Enable SIMD vectorization on
15303         non-SIMT targets in acc vector loops.
15305 2017-09-18  Claudiu Zissulescu  <claziss@synopsys.com>
15307         * configure.ac: Add arc and check if assembler supports gdwarf2.
15308         * configure: Regenerate.
15310 2017-09-18  Richard Biener  <rguenther@suse.de>
15312         PR tree-optimization/82220
15313         * tree-vect-loop.c (vect_estimate_min_profitable_iters): Exclude
15314         epilogue niters from the min_profitable_iters compute.
15316 2017-09-18  Jakub Jelinek  <jakub@redhat.com>
15318         PR target/82145
15319         * config/i386/i386.c (ix86_init_large_pic_reg): Revert 2017-09-01
15320         changes.  Turn CODE_LABEL into NOTE_INSN_DELETED_LABEL immediately.
15321         (ix86_init_pic_reg): Revert 2017-09-01 changes.
15323 2017-09-18  Eric Botcazou  <ebotcazou@adacore.com>
15325         PR target/81361
15326         * dwarf2cfi.c (add_cfis_to_fde): Do not generate DW_CFA_set_loc after
15327         switching to a new text section.
15329 2017-09-18  Richard Biener  <rguenther@suse.de>
15331         * graphite-scop-detection.c (scop_detection::stmt_has_simple_data_ref):
15332         Simplify.
15333         (build_alias_set): Reject aliases with no access function.
15335 2017-09-18  Richard Biener  <rguenther@suse.de>
15337         PR tree-optimization/79622
15338         * graphite-scop-detection.c (build_cross_bb_scalars_def): Properly
15339         handle PHIs.
15340         (build_cross_bb_scalars_use): Likewise.
15342 2017-09-18  Pierre-Marie de Rodat  <derodat@adacore.com>
15344         * cgraph.h (cgraph_thunk_info): Fix a typo in a comment.
15346 2017-09-18  Alan Modra  <amodra@gmail.com>
15348         PR target/81996
15349         * gcc/config/rs6000/rs6000.c (rs6000_return_addr): Use
15350         stack_pointer_rtx for count 0.  Update comments.  Break up
15351         large rtl expression.
15353 2017-09-17  Daniel Santos  <daniel.santos@pobox.com>
15355         * config/i386/i386.c (xlogue_layout::STUB_NAME_MAX_LEN):
15356         Increase to 20 bytes.
15357         (xlogue_layout::s_stub_names): Add an additional size-2 diminsion.
15358         (xlogue_layout::get_stub_name): Modify to select the appropairate sse
15359         or avx version of the stub.
15361 2017-09-17  H.J. Lu  <hongjiu.lu@intel.com>
15363         PR target/82166
15364         * config/i386/i386.c (ix86_finalize_stack_frame_flags): Properly
15365         compute the minimum stack alignment.  Also update preferred stack
15366         boundary for leaf functions.
15368 2017-09-16  Richard Sandiford  <richard.sandiford@linaro.org>
15370         PR tree-optimization/82228
15371         * tree-vect-loop.c (vectorizable_live_operation): Move initialization
15372         of ncopies.
15374 2017-09-16  Chung-Ju Wu  <jasonwucj@gmail.com>
15376         * common/config/nds32/nds32-common.c
15377         (nds32_option_optimization_table): Refine formatting.
15378         (nds32_option_optimization_table): Use -fsched-pressure and
15379         -fomit-frame-pointer for specific optimization level.
15381 2017-09-16  Chung-Ju Wu  <jasonwucj@gmail.com>
15383         * config/nds32/nds32.c: Refine formatting and comments.
15384         * config/nds32/nds32.h: Likewise.
15385         * config/nds32/nds32.md: Likewise.
15386         * config/nds32/nds32-cost.c: Likewise.
15387         * config/nds32/nds32-isr.c: Likewise.
15388         * config/nds32/nds32-md-auxiliary.c: Likewise.
15389         * config/nds32/nds32-multiple.md: Likewise.
15390         * config/nds32/nds32-predicates.c: Likewise.
15392 2017-09-15  Andrew Sutton  <andrew.n.sutton@gmail.com>
15393             Jakub Jelinek  <jakub@redhat.com>
15395         Add support for -std=c++2a.
15396         * doc/cpp.texi (__cplusplus): Document value for -std=c++2a
15397         or -std=gnu+2a.
15398         * doc/invoke.texi: Document -std=c++2a and -std=gnu++2a.
15400 2017-09-15  Steve Ellcey  <sellcey@cavium.com>
15402         PR target/82066
15403         * doc/extend.texi (Common Function Attributes): Add 
15404         references to ARM, AArch64, and S/390 specific attributes.
15405         (Function Specific Option Pragmas): Add AArch64 and S/390
15406         to list of back ends that support the target pragma.
15408 2017-09-15  Nathan Sidwell  <nathan@acm.org>
15410         * doc/standards.texi: Fix C++17 description.  Update URLs for
15411         C++11 & 14.
15413 2017-09-15  Bernd Edlinger  <bernd.edlinger@hotmail.de>
15415         * common.opt (Wcast-align=strict): New warning option.
15416         * doc/invoke.texi: Document -Wcast-align=strict. 
15418 2017-09-15  Pierre-Marie de Rodat  <derodat@adacore.com>
15420         * cgraph.h (cgraph_thunk_info): Add comments.
15421         * cgraph.c (cgraph_node::create_thunk): Adjust comment, make
15422         assert for VIRTUAL_* arguments stricter.
15424 2017-09-15  Jackson Woodruff  <jackson.woodruff@arm.com>
15426         PR tree-optimization/71026
15427         * match.pd: Move RDIV patterns from fold-const.c
15428         * fold-const.c (distribute_real_division): Removed.
15429         (fold_binary_loc): Remove calls to distribute_real_divison.
15431 2017-09-15  Jakub Jelinek  <jakub@redhat.com>
15433         * doc/invoke.texi: Document -std=c++17 and -std=gnu++17 and document
15434         c++1z and gnu++1z as deprecated.  Change other references to
15435         -std=c++1z to -std=c++17 and -std=gnu++1z to -std=gnu++17.
15436         Change -Wc++1z-compat to -Wc++17-compat.
15437         * doc/cpp.texi: Document -std=c++17 defines __cplusplus 201703L.
15438         * dwarf2out.c (highest_c_language): Handle C++17.
15439         (gen_compile_unit_die): Likewise.
15441 2017-09-15  Jakub Jelinek  <jakub@redhat.com>
15443         PR rtl-optimization/82192
15444         * combine.c (make_extraction): Don't look through non-paradoxical
15445         SUBREGs or TRUNCATE if pos + len is or might be bigger than
15446         inner's mode.
15448 2017-09-15  Richard Sandiford  <richard.sandiford@linaro.org>
15449             Alan Hayward  <alan.hayward@arm.com>
15450             David Sherwood  <david.sherwood@arm.com>
15452         * target.def (function_arg_offset): New hook.
15453         * targhooks.h (default_function_arg_offset): Declare.
15454         * targhooks.c (default_function_arg_offset): New function.
15455         * function.c (locate_and_pad_parm): Use
15456         targetm.calls.function_arg_offset instead of FUNCTION_ARG_OFFSET.
15457         * doc/tm.texi.in (FUNCTION_ARG_OFFSET): Replace with...
15458         (TARGET_FUNCTION_ARG_OFFSET): ...this.
15459         * doc/tm.texi: Regenerate.
15460         * config/spu/spu.h (FUNCTION_ARG_OFFSET): Delete.
15461         * config/spu/spu.c (spu_function_arg_offset): New function.
15462         (TARGET_FUNCTION_ARG_OFFSET): Redefine.
15463         * system.h (FUNCTION_ARG_OFFSET): Poison.
15465 2017-09-15  Richard Sandiford  <richard.sandiford@linaro.org>
15466             Alan Hayard  <alan.hayward@arm.com>
15467             David Sherwood  <david.sherwood@arm.com>
15469         * target.def (truly_noop_truncation): New hook.
15470         (mode_rep_extended): Refer to TARGET_TRULY_NOOP_TRUNCATION rather
15471         than TRULY_NOOP_TRUNCATION.
15472         * hooks.h (hook_bool_uint_uint_true): Declare.
15473         * hooks.c (hook_bool_uint_uint_true): New function.
15474         * doc/tm.texi.in (TRULY_NOOP_TRUNCATION): Replace with...
15475         (TARGET_TRULY_NOOP_TRUNCATION): ...this.
15476         * doc/tm.texi: Regenerate.
15477         * combine.c (make_extraction): Refer to TARGET_TRULY_NOOP_TRUNCATION
15478         rather than TRULY_NOOP_TRUNCATION in comments.
15479         (simplify_comparison): Likewise.
15480         (record_truncated_value): Likewise.
15481         * expmed.c (extract_bit_field_1): Likewise.
15482         (extract_split_bit_field): Likewise.
15483         * convert.c (convert_to_integer_1): Use targetm.truly_noop_truncation
15484         instead of TRULY_NOOP_TRUNCATION.
15485         * function.c (assign_parm_setup_block): Likewise.
15486         * machmode.h (TRULY_NOOP_TRUNCATION_MODES_P): Likewise.
15487         * rtlhooks.c: Include target.h.
15488         * config/aarch64/aarch64.h (TRULY_NOOP_TRUNCATION): Delete.
15489         * config/alpha/alpha.h (TRULY_NOOP_TRUNCATION): Delete.
15490         * config/arc/arc.h (TRULY_NOOP_TRUNCATION): Delete.
15491         * config/arm/arm.h (TRULY_NOOP_TRUNCATION): Delete.
15492         * config/avr/avr.h (TRULY_NOOP_TRUNCATION): Delete.
15493         * config/bfin/bfin.h (TRULY_NOOP_TRUNCATION): Delete.
15494         * config/c6x/c6x.h (TRULY_NOOP_TRUNCATION): Delete.
15495         * config/cr16/cr16.h (TRULY_NOOP_TRUNCATION): Delete.
15496         * config/cris/cris.h (TRULY_NOOP_TRUNCATION): Delete.
15497         * config/epiphany/epiphany.h (TRULY_NOOP_TRUNCATION): Delete.
15498         * config/fr30/fr30.h (TRULY_NOOP_TRUNCATION): Delete.
15499         * config/frv/frv.h (TRULY_NOOP_TRUNCATION): Delete.
15500         * config/ft32/ft32.h (TRULY_NOOP_TRUNCATION): Delete.
15501         * config/h8300/h8300.h (TRULY_NOOP_TRUNCATION): Delete.
15502         * config/i386/i386.h (TRULY_NOOP_TRUNCATION): Delete.
15503         * config/ia64/ia64.h (TRULY_NOOP_TRUNCATION): Delete.
15504         * config/iq2000/iq2000.h (TRULY_NOOP_TRUNCATION): Delete.
15505         * config/lm32/lm32.h (TRULY_NOOP_TRUNCATION): Delete.
15506         * config/m32c/m32c.h (TRULY_NOOP_TRUNCATION): Delete.
15507         * config/m32r/m32r.h (TRULY_NOOP_TRUNCATION): Delete.
15508         * config/m68k/m68k.h (TRULY_NOOP_TRUNCATION): Delete.
15509         * config/mcore/mcore.h (TRULY_NOOP_TRUNCATION): Delete.
15510         * config/microblaze/microblaze.h (TRULY_NOOP_TRUNCATION): Delete.
15511         * config/mips/mips.h (TRULY_NOOP_TRUNCATION): Delete.
15512         * config/mips/mips.c (mips_truly_noop_truncation): New function.
15513         (TARGET_TRULY_NOOP_TRUNCATION): Redefine.
15514         * config/mips/mips.md: Refer to TARGET_TRULY_NOOP_TRUNCATION
15515         rather than TRULY_NOOP_TRUNCATION in comments.
15516         * config/mmix/mmix.h (TRULY_NOOP_TRUNCATION): Delete.
15517         * config/mn10300/mn10300.h (TRULY_NOOP_TRUNCATION): Delete.
15518         * config/moxie/moxie.h (TRULY_NOOP_TRUNCATION): Delete.
15519         * config/msp430/msp430.h (TRULY_NOOP_TRUNCATION): Delete.
15520         * config/nds32/nds32.h (TRULY_NOOP_TRUNCATION): Delete.
15521         * config/nios2/nios2.h (TRULY_NOOP_TRUNCATION): Delete.
15522         * config/nvptx/nvptx.h (TRULY_NOOP_TRUNCATION): Delete.
15523         * config/pa/pa.h (TRULY_NOOP_TRUNCATION): Delete.
15524         * config/pdp11/pdp11.h (TRULY_NOOP_TRUNCATION): Delete.
15525         * config/powerpcspe/powerpcspe.h (TRULY_NOOP_TRUNCATION): Delete.
15526         * config/riscv/riscv.h (TRULY_NOOP_TRUNCATION): Delete.
15527         * config/riscv/riscv.md: Refer to TARGET_TRULY_NOOP_TRUNCATION
15528         rather than TRULY_NOOP_TRUNCATION in comments.
15529         * config/rl78/rl78.h (TRULY_NOOP_TRUNCATION): Delete.
15530         * config/rs6000/rs6000.h (TRULY_NOOP_TRUNCATION): Delete.
15531         * config/rx/rx.h (TRULY_NOOP_TRUNCATION): Delete.
15532         * config/s390/s390.h (TRULY_NOOP_TRUNCATION): Delete.
15533         * config/sh/sh.h (MAYBE_BASE_REGISTER_RTX_P): Remove
15534         TRULY_NOOP_TRUNCATION condition.
15535         (MAYBE_INDEX_REGISTER_RTX_P): Likewise.
15536         (TRULY_NOOP_TRUNCATION): Delete.
15537         * config/sparc/sparc.h (TRULY_NOOP_TRUNCATION): Delete.
15538         * config/spu/spu.h (TRULY_NOOP_TRUNCATION): Delete.
15539         * config/spu/spu.c (spu_truly_noop_truncation): New function.
15540         (TARGET_TRULY_NOOP_TRUNCATION): Redefine.
15541         * config/stormy16/stormy16.h (TRULY_NOOP_TRUNCATION): Delete.
15542         * config/tilegx/tilegx.h (TRULY_NOOP_TRUNCATION): Delete.
15543         * config/tilegx/tilegx.c (tilegx_truly_noop_truncation): New fuction.
15544         (TARGET_TRULY_NOOP_TRUNCATION): Redefine.
15545         * config/tilegx/tilegx.md: Refer to TARGET_TRULY_NOOP_TRUNCATION
15546         rather than TRULY_NOOP_TRUNCATION in comments.
15547         * config/tilepro/tilepro.h (TRULY_NOOP_TRUNCATION): Delete.
15548         * config/v850/v850.h (TRULY_NOOP_TRUNCATION): Delete.
15549         * config/vax/vax.h (TRULY_NOOP_TRUNCATION): Delete.
15550         * config/visium/visium.h (TRULY_NOOP_TRUNCATION): Delete.
15551         * config/xtensa/xtensa.h (TRULY_NOOP_TRUNCATION): Delete.
15552         * system.h (TRULY_NOOP_TRUNCATION): Poison.
15554 2017-09-15  Christophe Lyon  <christophe.lyon@linaro.org>
15556         PR target/67591
15557         * config/arm/arm.md (*cmp_and): Add enabled_for_depr_it attribute.
15558         (*cmp_ior): Likewise.
15559         (*ior_scc_scc): Add alternative for enabled_for_depr_it attribute.
15560         (*ior_scc_scc_cmp): Likewise.
15561         (*and_scc_scc): Likewise.
15562         (*and_scc_scc_cmp): Likewise.
15564 2017-09-15  Richard Sandiford  <richard.sandiford@linaro.org>
15565             Alan Hayard  <alan.hayward@arm.com>
15566             David Sherwood  <david.sherwood@arm.com>
15568         * target.def (can_change_mode_class): New hook.
15569         (mode_rep_extended): Refer to it instead of CANNOT_CHANGE_MODE_CLASS.
15570         (hard_regno_nregs): Likewise.
15571         * hooks.h (hook_bool_mode_mode_reg_class_t_true): Declare.
15572         * hooks.c (hook_bool_mode_mode_reg_class_t_true): New function.
15573         * doc/tm.texi.in (CANNOT_CHANGE_MODE_CLASS): Replace with...
15574         (TARGET_CAN_CHANGE_MODE_CLASS): ...this.
15575         (LOAD_EXTEND_OP): Update accordingly.
15576         * doc/tm.texi: Regenerate.
15577         * doc/rtl.texi: Refer to TARGET_CAN_CHANGE_MODE_CLASS instead of
15578         CANNOT_CHANGE_MODE_CLASS.
15579         * hard-reg-set.h (REG_CANNOT_CHANGE_MODE_P): Replace with...
15580         (REG_CAN_CHANGE_MODE_P): ...this new macro.
15581         * combine.c (simplify_set): Update accordingly.
15582         * emit-rtl.c (validate_subreg): Likewise.
15583         * recog.c (general_operand): Likewise.
15584         * regcprop.c (mode_change_ok): Likewise.
15585         * reload1.c (choose_reload_regs): Likewise.
15586         (inherit_piecemeal_p): Likewise.
15587         * rtlanal.c (simplify_subreg_regno): Likewise.
15588         * postreload.c (reload_cse_simplify_set): Use REG_CAN_CHANGE_MODE_P
15589         instead of CANNOT_CHANGE_MODE_CLASS.
15590         (reload_cse_simplify_operands): Likewise.
15591         * reload.c (push_reload): Use targetm.can_change_mode_class
15592         instead of CANNOT_CHANGE_MODE_CLASS.
15593         (push_reload): Likewise.  Also use REG_CAN_CHANGE_MODE_P instead of
15594         REG_CANNOT_CHANGE_MODE_P.
15595         * config/alpha/alpha.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15596         * config/alpha/alpha.c (alpha_can_change_mode_class): New function.
15597         (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15598         * config/arm/arm.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15599         * config/arm/arm.c (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15600         (arm_can_change_mode_class): New function.
15601         * config/arm/neon.md: Refer to TARGET_CAN_CHANGE_MODE_CLASS rather
15602         than CANNOT_CHANGE_MODE_CLASS in comments.
15603         * config/i386/i386.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15604         * config/i386/i386-protos.h (ix86_cannot_change_mode_class): Delete.
15605         * config/i386/i386.c (ix86_cannot_change_mode_class): Replace with...
15606         (ix86_can_change_mode_class): ...this new function, inverting the
15607         sense of the return value.
15608         (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15609         * config/ia64/ia64.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15610         * config/ia64/ia64.c (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15611         (ia64_can_change_mode_class): New function.
15612         * config/m32c/m32c.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15613         * config/m32c/m32c-protos.h (m32c_cannot_change_mode_class): Delete.
15614         * config/m32c/m32c.c (m32c_cannot_change_mode_class): Replace with...
15615         (m32c_can_change_mode_class): ...this new function, inverting the
15616         sense of the return value.
15617         (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15618         * config/mips/mips.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15619         * config/mips/mips-protos.h (mips_cannot_change_mode_class): Delete.
15620         * config/mips/mips.c (mips_cannot_change_mode_class): Replace with...
15621         (mips_can_change_mode_class): ...this new function, inverting the
15622         sense of the return value.
15623         (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15624         * config/msp430/msp430.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15625         * config/msp430/msp430.c (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15626         (msp430_can_change_mode_class): New function.
15627         * config/nvptx/nvptx.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15628         * config/nvptx/nvptx.c (nvptx_can_change_mode_class): New function.
15629         (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15630         * config/pa/pa32-regs.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15631         * config/pa/pa64-regs.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15632         * config/pa/pa-protos.h (pa_cannot_change_mode_class): Delete.
15633         * config/pa/pa.c (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15634         (pa_cannot_change_mode_class): Replace with...
15635         (pa_can_change_mode_class): ...this new function, inverting the
15636         sense of the return value.
15637         (pa_modes_tieable_p): Refer to TARGET_CAN_CHANGE_MODE_CLASS rather
15638         than CANNOT_CHANGE_MODE_CLASS in comments.
15639         * config/pdp11/pdp11.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15640         * config/pdp11/pdp11-protos.h (pdp11_cannot_change_mode_class): Delete.
15641         * config/pdp11/pdp11.c (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15642         (pdp11_cannot_change_mode_class): Replace with...
15643         (pdp11_can_change_mode_class): ...this new function, inverting the
15644         sense of the return value.
15645         * config/powerpcspe/powerpcspe.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15646         * config/powerpcspe/powerpcspe-protos.h
15647         (rs6000_cannot_change_mode_class_ptr): Delete.
15648         * config/powerpcspe/powerpcspe.c
15649         (rs6000_cannot_change_mode_class_ptr): Delete.
15650         (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15651         (rs6000_option_override_internal): Assign to
15652         targetm.can_change_mode_class instead of
15653         rs6000_cannot_change_mode_class_ptr.
15654         (rs6000_cannot_change_mode_class): Replace with...
15655         (rs6000_can_change_mode_class): ...this new function, inverting the
15656         sense of the return value.
15657         (rs6000_debug_cannot_change_mode_class): Replace with...
15658         (rs6000_debug_can_change_mode_class): ...this new function.
15659         * config/riscv/riscv.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15660         * config/riscv/riscv.c (riscv_can_change_mode_class): New function.
15661         (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15662         * config/rs6000/rs6000.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15663         * config/rs6000/rs6000-protos.h (rs6000_cannot_change_mode_class_ptr):
15664         Delete.
15665         * config/rs6000/rs6000.c (rs6000_cannot_change_mode_class_ptr): Delete.
15666         (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15667         (rs6000_option_override_internal): Assign to
15668         targetm.can_change_mode_class instead of
15669         rs6000_cannot_change_mode_class_ptr.
15670         (rs6000_cannot_change_mode_class): Replace with...
15671         (rs6000_can_change_mode_class): ...this new function, inverting the
15672         sense of the return value.
15673         (rs6000_debug_cannot_change_mode_class): Replace with...
15674         (rs6000_debug_can_change_mode_class): ...this new function.
15675         * config/s390/s390.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15676         * config/s390/s390-protos.h (s390_cannot_change_mode_class): Delete.
15677         * config/s390/s390.c (s390_cannot_change_mode_class): Replace with...
15678         (s390_can_change_mode_class): ...this new function, inverting the
15679         sense of the return value.
15680         (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15681         * config/sh/sh.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15682         * config/sh/sh-protos.h (sh_cannot_change_mode_class): Delete.
15683         * config/sh/sh.c (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15684         (sh_cannot_change_mode_class): Replace with...
15685         (sh_can_change_mode_class): ...this new function, inverting the
15686         sense of the return value.
15687         * config/sparc/sparc.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15688         * config/sparc/sparc.c (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15689         (sparc_can_change_mode_class): New function.
15690         * config/spu/spu.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15691         * config/spu/spu.c (spu_can_change_mode_class): New function.
15692         (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15693         * config/visium/visium.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15694         * config/visium/visium.c (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15695         (visium_can_change_mode_class): New function.
15696         * system.h (CANNOT_CHANGE_MODE_CLASS): Poison.
15698 2017-09-15  Richard Biener  <rguenther@suse.de>
15700         PR tree-optimization/82217
15701         * tree-ssa-sccvn.c (visit_phi): Properly handle all VN_TOP
15702         but not undefined case.
15704 2017-09-15  Jakub Jelinek  <jakub@redhat.com>
15706         PR target/82145
15707         * postreload.c (reload_cse_simplify_operands): Skip
15708         NOTE_INSN_DELETED_LABEL similarly to skipping CODE_LABEL.
15710 2017-09-15  Richard Biener  <rguenther@suse.de>
15712         PR tree-optimization/68823
15713         * graphite-scop-detection.c (build_alias_set): If we have a
15714         possible dependence check whether we can handle them by just
15715         looking at the DRs DR_ACCESS_FNs.
15716         (build_scops): If build_alias_set fails, fail the SCOP.
15718 2017-09-14  Michael Meissner  <meissner@linux.vnet.ibm.com>
15720         * config/rs6000/rs6000-builtin.def (BU_FLOAT128_1_HW): New macros
15721         to support float128 built-in functions that require the ISA 3.0
15722         hardware.
15723         (BU_FLOAT128_3_HW): Likewise.
15724         (SQRTF128): Add support for the IEEE 128-bit square root and fma
15725         built-in functions.
15726         (FMAF128): Likewise.
15727         (FMAQ): Likewise.
15728         * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add
15729         support for built-in functions that need the ISA 3.0 IEEE 128-bit
15730         floating point instructions.
15731         (rs6000_invalid_builtin): Likewise.
15732         (rs6000_builtin_mask_names): Likewise.
15733         * config/rs6000/rs6000.h (MASK_FLOAT128_HW): Likewise.
15734         (RS6000_BTM_FLOAT128_HW): Likewise.
15735         (RS6000_BTM_COMMON): Likewise.
15736         * config/rs6000/rs6000.md (fma<mode>4_hw): Add a generator
15737         function.
15738         * doc/extend.texi (RS/6000 built-in functions): Document the
15739         IEEE 128-bit floating point square root and fused multiply-add
15740         built-in functions.
15742 2017-09-14  Pat Haugen  <pthaugen@us.ibm.com>
15744         * config/rs6000/rs6000.c (rs6000_set_up_by_prologue): Make sure the TOC
15745         reg (r2) isn't in the set of registers defined in the prologue.
15747 2017-09-14  Richard Sandiford  <richard.sandiford@linaro.org>
15748             Alan Hayward  <alan.hayward@arm.com>
15749             David Sherwood  <david.sherwood@arm.com>
15751         * tree-vectorizer.h (_loop_vec_info): Add max_vectorization_factor.
15752         (LOOP_VINFO_MAX_VECT_FACTOR): New macro.
15753         (LOOP_VINFO_ORIG_VECT_FACTOR): Replace with...
15754         (LOOP_VINFO_ORIG_MAX_VECT_FACTOR): ...this new macro.
15755         * tree-vect-data-refs.c (vect_analyze_data_ref_dependences): Update
15756         accordingly.
15757         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
15758         max_vectorization_factor.
15759         (vect_analyze_loop_2): Set LOOP_VINFO_MAX_VECT_FACTOR.
15761 2017-09-14  Richard Sandiford  <richard.sandiford@linaro.org>
15762             Alan Hayward  <alan.hayward@arm.com>
15763             David Sherwood  <david.sherwood@arm.com>
15765         * tree-vectorizer.h (vect_min_worthwhile_factor): Delete.
15766         (vect_worthwhile_without_simd_p): Declare.
15767         * tree-vect-loop.c (vect_worthwhile_without_simd_p): New function.
15768         (vectorizable_reduction): Use it.
15769         * tree-vect-stmts.c (vectorizable_shift): Likewise.
15770         (vectorizable_operation): Likewise.
15772 2017-09-14  Richard Sandiford  <richard.sandiford@linaro.org>
15773             Alan Hayward  <alan.hayward@arm.com>
15774             David Sherwood  <david.sherwood@arm.com>
15776         * tree-vectorizer.h (vect_get_num_copies): New function.
15777         * tree-vect-data-refs.c (vect_get_data_access_cost): Use it.
15778         * tree-vect-loop.c (vectorizable_reduction): Likewise.
15779         (vectorizable_induction): Likewise.
15780         (vectorizable_live_operation): Likewise.
15781         * tree-vect-stmts.c (vectorizable_mask_load_store): Likewise.
15782         (vectorizable_bswap): Likewise.
15783         (vectorizable_call): Likewise.
15784         (vectorizable_conversion): Likewise.
15785         (vectorizable_assignment): Likewise.
15786         (vectorizable_shift): Likewise.
15787         (vectorizable_operation): Likewise.
15788         (vectorizable_store): Likewise.
15789         (vectorizable_load): Likewise.
15790         (vectorizable_condition): Likewise.
15791         (vectorizable_comparison): Likewise.
15792         (vect_analyze_stmt): Pass the slp node to vectorizable_live_operation.
15794 2017-09-14  Richard Sandiford  <richard.sandiford@linaro.org>
15795             Alan Hayward  <alan.hayward@arm.com>
15796             David Sherwood  <david.sherwood@arm.com>
15798         * tree-vect-loop.c (vectorizable_induction): Use gimple_build instead
15799         of vect_init_vector.
15801 2017-09-14  Richard Sandiford  <richard.sandiford@linaro.org>
15802             Alan Hayward  <alan.hayward@arm.com>
15803             David Sherwood  <david.sherwood@arm.com>
15805         * gimple-fold.h (gimple_build_vector_from_val): Declare, and provide
15806         an inline wrapper that provides a location.
15807         (gimple_build_vector): Likewise.
15808         * gimple-fold.c (gimple_build_vector_from_val): New function.
15809         (gimple_build_vector): Likewise.
15810         * tree-vect-loop.c (get_initial_def_for_reduction): Use the new
15811         functions to build the initial value.  Always return a gimple value.
15812         (get_initial_defs_for_reduction): Likewise.  Only compute
15813         neutral_vec once.
15814         (vect_create_epilog_for_reduction): Don't call force_gimple_operand or
15815         vect_init_vector on the results from get_initial_def(s)_for_reduction.
15816         (vectorizable_induction): Use gimple_build_vector rather than
15817         vect_init_vector.
15819 2017-09-14  Richard Sandiford  <richard.sandiford@linaro.org>
15820             Alan Hayward  <alan.hayward@arm.com>
15821             David Sherwood  <david.sherwood@arm.com>
15823         * target.h (vec_perm_indices): New typedef.
15824         (auto_vec_perm_indices): Likewise.
15825         * optabs-query.h: Include target.h
15826         (can_vec_perm_p): Take a vec_perm_indices *.
15827         * optabs-query.c (can_vec_perm_p): Likewise.
15828         (can_mult_highpart_p): Update accordingly.  Use auto_vec_perm_indices.
15829         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
15830         * tree-vect-generic.c (lower_vec_perm): Likewise.
15831         * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
15832         (vect_grouped_load_supported): Likewise.
15833         (vect_shift_permute_load_chain): Likewise.
15834         (vect_permute_store_chain): Use auto_vec_perm_indices.
15835         (vect_permute_load_chain): Likewise.
15836         * fold-const.c (fold_vec_perm): Take vec_perm_indices.
15837         (fold_ternary_loc): Update accordingly.  Use auto_vec_perm_indices.
15838         Update uses of can_vec_perm_p.
15839         * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Replace the
15840         mode with a number of elements.  Take a vec_perm_indices *.
15841         (vect_create_epilog_for_reduction): Update accordingly.
15842         Use auto_vec_perm_indices.
15843         (have_whole_vector_shift): Likewise.  Update call to can_vec_perm_p.
15844         * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
15845         (vect_transform_slp_perm_load): Likewise.
15846         (vect_schedule_slp_instance): Use auto_vec_perm_indices.
15847         * tree-vectorizer.h (vect_gen_perm_mask_any): Take a vec_perm_indices.
15848         (vect_gen_perm_mask_checked): Likewise.
15849         * tree-vect-stmts.c (vect_gen_perm_mask_any): Take a vec_perm_indices.
15850         (vect_gen_perm_mask_checked): Likewise.
15851         (vectorizable_mask_load_store): Use auto_vec_perm_indices.
15852         (vectorizable_store): Likewise.
15853         (vectorizable_load): Likewise.
15854         (perm_mask_for_reverse): Likewise.  Update call to can_vec_perm_p.
15855         (vectorizable_bswap): Likewise.
15857 2017-09-14  Richard Sandiford  <richard.sandiford@linaro.org>
15858             Alan Hayward  <alan.hayward@arm.com>
15859             David Sherwood  <david.sherwood@arm.com>
15861         * tree.h (build_vector): Take a vec<tree> instead of a tree *.
15862         * tree.c (build_vector): Likewise.
15863         (build_vector_from_ctor): Update accordingly.
15864         (build_vector_from_val): Likewise.
15865         * gimple-fold.c (gimple_fold_stmt_to_constant_1): Likewise.
15866         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
15867         * tree-vect-generic.c (add_rshift): Likewise.
15868         (expand_vector_divmod): Likewise.
15869         (optimize_vector_constructor): Likewise.
15870         * tree-vect-slp.c (vect_get_constant_vectors): Likewise.
15871         (vect_transform_slp_perm_load): Likewise.
15872         (vect_schedule_slp_instance): Likewise.
15873         * tree-vect-stmts.c (vectorizable_bswap): Likewise.
15874         (vectorizable_call): Likewise.
15875         (vect_gen_perm_mask_any): Likewise.  Add elements in order.
15876         * expmed.c (make_tree): Likewise.
15877         * fold-const.c (fold_negate_expr_1): Use auto_vec<tree> when building
15878         a vector passed to build_vector.
15879         (fold_convert_const): Likewise.
15880         (exact_inverse): Likewise.
15881         (fold_ternary_loc): Likewise.
15882         (fold_relational_const): Likewise.
15883         (const_binop): Likewise.  Use VECTOR_CST_ELT directly when operating
15884         on VECTOR_CSTs, rather than going through vec_cst_ctor_to_array.
15885         (const_unop): Likewise.  Store the reduction accumulator in a
15886         variable rather than an array.
15887         (vec_cst_ctor_to_array): Take the number of elements as a parameter.
15888         (fold_vec_perm): Update calls accordingly.  Use auto_vec<tree> for
15889         the new vector, rather than constructing it after the input arrays.
15890         (native_interpret_vector): Use auto_vec<tree> when building
15891         a vector passed to build_vector.  Add elements in order.
15892         * tree-vect-loop.c (get_initial_defs_for_reduction): Use
15893         auto_vec<tree> when building a vector passed to build_vector.
15894         (vect_create_epilog_for_reduction): Likewise.
15895         (vectorizable_induction): Likewise.
15896         (get_initial_def_for_reduction): Likewise.  Fix indentation of
15897         case statements.
15898         * config/sparc/sparc.c (sparc_handle_vis_mul8x16): Change n_elts
15899         to a vec<tree> *.
15900         (sparc_fold_builtin): Use auto_vec<tree> when building a vector
15901         passed to build_vector.
15903 2017-09-14  Richard Sandiford  <richard.sandiford@linaro.org>
15904             Alan Hayward  <alan.hayward@arm.com>
15905             David Sherwood  <david.sherwood@arm.com>
15907         * tree-core.h (tree_base::u): Add an "nelts" field.
15908         (tree_vector): Use VECTOR_CST_NELTS as the length.
15909         * tree.c (tree_size): Likewise.
15910         (make_vector): Initialize VECTOR_CST_NELTS.
15911         * tree.h (VECTOR_CST_NELTS): Use the u.nelts field.
15912         * cfgexpand.c (expand_debug_expr): Use VECTOR_CST_NELTS instead of
15913         TYPE_VECTOR_SUBPARTS.
15914         * expr.c (const_vector_mask_from_tree): Consistently use "units"
15915         as the number of units, setting it from VECTOR_CST_NELTS.
15916         (const_vector_from_tree): Likewise.
15917         * fold-const.c (negate_expr_p): Use VECTOR_CST_NELTS instead of
15918         TYPE_VECTOR_SUBPARTS for the number of elements in a VECTOR_CST.
15919         (fold_negate_expr_1): Likewise.
15920         (fold_convert_const): Likewise.
15921         (const_binop): Likewise.  Differentiate the number of output and
15922         input elements.
15923         (const_unop): Likewise.
15924         (fold_ternary_loc): Use VECTOR_CST_NELTS for the number of elements
15925         in a VECTOR_CST, asserting that it is the same as TYPE_VECTOR_SUBPARTS
15926         in cases that did the opposite.
15928 2017-09-14  Richard Biener  <rguenther@suse.de>
15930         * tree-ssa-sccvn.c (visit_phi): Merge undefined values similar
15931         to VN_TOP.
15933 2017-09-14  Eric Botcazou  <ebotcazou@adacore.com>
15935         * dwarf2out.c (dwarf2out_source_line): Remove superfluous test.
15937 2017-09-14  Jakub Jelinek  <jakub@redhat.com>
15939         PR target/81325
15940         * cfgbuild.c (find_bb_boundaries): Ignore debug insns in decisions
15941         if and where to split a bb, except for splitting before debug insn
15942         sequences followed by non-label real insn.  Delete debug insns
15943         in between basic blocks.
15945         * combine.c (make_compound_operation_int): Formatting fixes.
15947         * config/alpha/elf.h (LINK_EH_SPEC): Add -static-pie support.
15948         * config/alpha/linux.h (LINK_GCC_C_SEQUENCE_SPEC): Likewise.
15949         * config/netbsd.h (LINK_EH_SPEC): Likewise.
15950         * config/sol2.h (LINK_EH_SPEC): Likewise.
15951         * config/arm/uclinux-elf.h (LINK_GCC_C_SEQUENCE_SPEC): Likewise.
15952         * config/s390/linux.h (LINK_SPEC): Likewise.
15953         * config/freebsd.h (LINK_EH_SPEC): Likewise.
15954         * config/openbsd.h (LINK_EH_SPEC): Likewise.
15955         * config/lm32/uclinux-elf.h (LINK_GCC_C_SEQUENCE_SPEC): Likewise.
15956         * config/aarch64/aarch64-linux.h (LINUX_TARGET_LINK_SPEC): Likewise.
15957         * config/powerpcspe/sysv4.h (LINK_EH_SPEC): Likewise.
15958         * config/bfin/linux.h (LINK_GCC_C_SEQUENCE_SPEC): Likewise.
15959         * config/i386/gnu-user64.h (GNU_USER_TARGET_LINK_SPEC): Fix a typo.
15960         * config/i386/gnu-user.h (GNU_USER_TARGET_LINK_SPEC): Formatting fix.
15962 2017-09-13  Jakub Jelinek  <jakub@redhat.com>
15964         * config/rs6000/sysv4.h (STARTFILE_LINUX_SPEC): Add -static-pie
15965         support.
15966         (ENDFILE_LINUX_SPEC): Likewise.
15967         (LINK_EH_SPEC): Likewise.
15968         * config/rs6000/linux64.h (LINK_SHLIB_SPEC): Likewise.
15969         (LINK_OS_LINUX_SPEC32): Likewise.
15970         (LINK_OS_LINUX_SPEC64): Likewise.
15971         * config/rs6000/linux.h (LINK_SHLIB_SPEC): Likewise.
15972         (LINK_OS_LINUX_SPEC): Likewise.
15974 2017-09-13  Martin Liska  <mliska@suse.cz>
15976         PR middle-end/82154
15977         * stmt.c (expand_sjlj_dispatch_table): Use CASE_LOW when
15978         CASE_HIGH is NULL_TREE.
15980 2017-09-13  Richard Sandiford  <richard.sandiford@linaro.org>
15981             Alan Hayward  <alan.hayward@arm.com>
15982             David Sherwood  <david.sherwood@arm.com>
15984         * target.def (secondary_memory_needed): New hook.
15985         (secondary_reload): Refer to TARGET_SECONDARY_MEMORY_NEEDED
15986         instead of SECONDARY_MEMORY_NEEDED.
15987         (secondary_memory_needed_mode): Likewise.
15988         * hooks.h (hook_bool_mode_reg_class_t_reg_class_t_false): Declare.
15989         * hooks.c (hook_bool_mode_reg_class_t_reg_class_t_false): New function.
15990         * doc/tm.texi.in (SECONDARY_MEMORY_NEEDED): Replace with...
15991         (TARGET_SECONDARY_MEMORY_NEEDED): ...this.
15992         (SECONDARY_MEMORY_NEEDED_RTX): Update reference accordingly.
15993         * doc/tm.texi: Regenerate.
15994         * config/alpha/alpha.h (SECONDARY_MEMORY_NEEDED): Delete.
15995         * config/alpha/alpha.c (alpha_secondary_memory_needed): New function.
15996         (TARGET_SECONDARY_MEMORY_NEEDED): Redefine.
15997         * config/i386/i386.h (SECONDARY_MEMORY_NEEDED): Delete.
15998         * config/i386/i386-protos.h (ix86_secondary_memory_needed): Delete.
15999         * config/i386/i386.c (inline_secondary_memory_needed): Put the
16000         mode argument first and change the reg_class arguments to reg_class_t.
16001         (ix86_secondary_memory_needed): Likewise.  Remove the strict parameter.
16002         Make static.  Update the call to inline_secondary_memory_needed.
16003         (ix86_register_move_cost): Update the call to
16004         inline_secondary_memory_needed.
16005         (TARGET_SECONDARY_MEMORY_NEEDED): Redefine.
16006         * config/ia64/ia64.h (SECONDARY_MEMORY_NEEDED): Delete commented-out
16007         definition.
16008         * config/ia64/ia64.c (spill_xfmode_rfmode_operand): Refer to
16009         TARGET_SECONDARY_MEMORY_NEEDED rather than SECONDARY_MEMORY_NEEDED
16010         in comment.
16011         * config/mips/mips.h (SECONDARY_MEMORY_NEEDED): Delete.
16012         * config/mips/mips-protos.h (mips_secondary_memory_needed): Delete.
16013         * config/mips/mips.c (mips_secondary_memory_needed): Make static
16014         and match hook interface.  Add comment from mips.h.
16015         (TARGET_SECONDARY_MEMORY_NEEDED): Redefine.
16016         * config/mmix/mmix.md (truncdfsf2): Refer to
16017         TARGET_SECONDARY_MEMORY_NEEDED rather than SECONDARY_MEMORY_NEEDED
16018         in comment.
16019         * config/pa/pa-64.h (SECONDARY_MEMORY_NEEDED): Rename to...
16020         (PA_SECONDARY_MEMORY_NEEDED): ...this, and put the mode argument first.
16021         * config/pa/pa.c (TARGET_SECONDARY_MEMORY_NEEDED): Redefine.
16022         (pa_secondary_memory_needed): New function.
16023         * config/pdp11/pdp11.h (SECONDARY_MEMORY_NEEDED): Delete.
16024         * config/pdp11/pdp11-protos.h (pdp11_secondary_memory_needed): Delete.
16025         * config/pdp11/pdp11.c (TARGET_SECONDARY_MEMORY_NEEDED): Redefine.
16026         (pdp11_secondary_memory_needed): Make static and match hook interface.
16027         * config/powerpcspe/powerpcspe.h (SECONDARY_MEMORY_NEEDED): Delete.
16028         * config/powerpcspe/powerpcspe-protos.h
16029         (rs6000_secondary_memory_needed_ptr): Delete.
16030         * config/powerpcspe/powerpcspe.c (rs6000_secondary_memory_needed_ptr):
16031         Delete.
16032         (TARGET_SECONDARY_MEMORY_NEEDED): Redefine.
16033         (rs6000_option_override_internal): Assign to
16034         targetm.secondary_memory_needed rather than
16035         rs6000_secondary_memory_needed_ptr.
16036         (rs6000_secondary_memory_needed): Match hook interface.
16037         (rs6000_debug_secondary_memory_needed): Likewise.
16038         * config/riscv/riscv.h (SECONDARY_MEMORY_NEEDED): Delete.
16039         * config/riscv/riscv.c (riscv_secondary_memory_needed): New function.
16040         (riscv_register_move_cost): Use it instead of SECONDARY_MEMORY_NEEDED.
16041         (TARGET_SECONDARY_MEMORY_NEEDED): Redefine.
16042         * config/rs6000/rs6000.h (SECONDARY_MEMORY_NEEDED): Delete.
16043         * config/rs6000/rs6000-protos.h (rs6000_secondary_memory_needed_ptr):
16044         Delete.
16045         * config/rs6000/rs6000.c (rs6000_secondary_memory_needed_ptr): Delete.
16046         (TARGET_SECONDARY_MEMORY_NEEDED): Redefine.
16047         (rs6000_option_override_internal): Assign to
16048         targetm.secondary_memory_needed rather than
16049         rs6000_secondary_memory_needed_ptr.
16050         (rs6000_secondary_memory_needed): Match hook interface.
16051         (rs6000_debug_secondary_memory_needed): Likewise.
16052         * config/s390/s390.h (SECONDARY_MEMORY_NEEDED): Delete.
16053         * config/s390/s390.c (s390_secondary_memory_needed): New function.
16054         (TARGET_SECONDARY_MEMORY_NEEDED): Redefine.
16055         * config/sparc/sparc.h (SECONDARY_MEMORY_NEEDED): Delete.
16056         * config/sparc/sparc.c (TARGET_SECONDARY_MEMORY_NEEDED): Redefine.
16057         (sparc_secondary_memory_needed): New function.
16058         * lra-constraints.c (check_and_process_move): Refer to
16059         TARGET_SECONDARY_MEMORY_NEEDED rather than SECONDARY_MEMORY_NEEDED
16060         in comment.
16061         (curr_insn_transform): Likewise.
16062         (process_alt_operands): Use targetm.secondary_memory_needed
16063         instead of TARGET_SECONDARY_MEMORY_NEEDED.
16064         (check_secondary_memory_needed_p): Likewise.
16065         (choose_split_class): Likewise.
16066         * reload.c: Unconditionally include code that was previously
16067         conditional on SECONDARY_MEMORY_NEEDED.
16068         (push_secondary_reload): Use targetm.secondary_memory_needed
16069         instead of TARGET_SECONDARY_MEMORY_NEEDED.
16070         (push_reload): Likewise.
16071         * reload1.c: Unconditionally include code that was previously
16072         conditional on SECONDARY_MEMORY_NEEDED.
16073         (choose_reload_regs): Use targetm.secondary_memory_needed
16074         instead of TARGET_SECONDARY_MEMORY_NEEDED.
16075         (gen_reload): Likewise.
16076         * system.h (SECONDARY_MEMORY_NEEDED): Poison.
16078 2017-09-13  Richard Sandiford  <richard.sandiford@linaro.org>
16079             Alan Hayward  <alan.hayward@arm.com>
16080             David Sherwood  <david.sherwood@arm.com>
16082         * target.def (secondary_memory_needed_mode): New hook:
16083         * targhooks.c (default_secondary_memory_needed_mode): Declare.
16084         * targhooks.h (default_secondary_memory_needed_mode): New function.
16085         * doc/tm.texi.in (SECONDARY_MEMORY_NEEDED_MODE): Replace with...
16086         (TARGET_SECONDARY_MEMORY_NEEDED_MODE): ...this.
16087         * doc/tm.texi: Regenerate.
16088         * lra-constraints.c (check_and_process_move): Use
16089         targetm.secondary_memory_needed_mode instead of
16090         TARGET_SECONDARY_MEMORY_NEEDED_MODE.
16091         (curr_insn_transform): Likewise.
16092         * reload.c (get_secondary_mem): Likewise.
16093         * config/alpha/alpha.h (SECONDARY_MEMORY_NEEDED_MODE): Delete.
16094         * config/alpha/alpha.c (alpha_secondary_memory_needed_mode): New
16095         function.
16096         (TARGET_SECONDARY_MEMORY_NEEDED_MODE): Redefine.
16097         * config/i386/i386.h (SECONDARY_MEMORY_NEEDED_MODE): Delete.
16098         * config/i386/i386.c (ix86_secondary_memory_needed_mode): New function.
16099         (TARGET_SECONDARY_MEMORY_NEEDED_MODE): Redefine.
16100         * config/powerpcspe/powerpcspe.h (SECONDARY_MEMORY_NEEDED_MODE):
16101         Delete.
16102         * config/powerpcspe/powerpcspe-protos.h
16103         (rs6000_secondary_memory_needed_mode): Delete.
16104         * config/powerpcspe/powerpcspe.c
16105         (TARGET_SECONDARY_MEMORY_NEEDED_MODE): Redefine.
16106         (rs6000_secondary_memory_needed_mode): Make static.
16107         * config/rs6000/rs6000.h (SECONDARY_MEMORY_NEEDED_MODE): Delete.
16108         * config/rs6000/rs6000-protos.h (rs6000_secondary_memory_needed_mode):
16109         Delete.
16110         * config/rs6000/rs6000.c (TARGET_SECONDARY_MEMORY_NEEDED_MODE):
16111         Redefine.
16112         (rs6000_secondary_memory_needed_mode): Make static.
16113         * config/s390/s390.h (SECONDARY_MEMORY_NEEDED_MODE): Delete.
16114         * config/s390/s390.c (s390_secondary_memory_needed_mode): New function.
16115         (TARGET_SECONDARY_MEMORY_NEEDED_MODE): Redefine.
16116         * config/sparc/sparc.h (SECONDARY_MEMORY_NEEDED_MODE): Delete.
16117         * config/sparc/sparc.c (TARGET_SECONDARY_MEMORY_NEEDED_MODE):
16118         Redefine.
16119         (sparc_secondary_memory_needed_mode): New function.
16120         * system.h (TARGET_SECONDARY_MEMORY_NEEDED_MODE): Poison.
16122 2017-09-13  Jackson Woodruff  <jackson.woodruff@arm.com>
16124         * config/aarch64/constraints.md (Umq): New constraint.
16125         * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>):
16126         Change to use Umq.
16127         (mov<mode>): Update condition.
16129 2017-09-13  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
16131         * gimple-ssa-store-merging.c (sort_by_bitpos): Compare store order
16132         when bitposition is the same.
16134 2017-09-13  Richard Biener  <rguenther@suse.de>
16136         * dwarf2out.c (output_die_symbol): Remove.
16137         (output_die): Do not output a DIEs symbol.
16139 2017-09-13  Richard Biener  <rguenther@suse.de>
16141         PR middle-end/82128
16142         * gimple-fold.c (gimple_fold_call): Update SSA name in-place to
16143         default-def to avoid breaking iterator update with the weird
16144         interaction with cgraph_update_edges_for_call_stmt_node.
16146 2017-09-13  Richard Biener  <rguenther@suse.de>
16148         * tree-cfg.c (verify_gimple_assign_binary): Add verification
16149         for WIDEN_SUM_EXPR, VEC_WIDEN_MULT_{HI,LO,EVEN,ODD}_EXPR,
16150         VEC_PACK_{TRUNC,SAT,FIX_TRUNC}_EXPR.
16151         (verify_gimple_assign_ternary): Add verification for DOT_PROD_EXPR.
16153 2017-09-13  Kugan Vivekanandarajah  <kuganv@linaro.org>
16155         * config/aarch64/aarch64.c (aarch64_override_options_after_change_1):
16156         Disable pc relative literal load irrespective of
16157         TARGET_FIX_ERR_A53_84341 for default.
16159 2017-09-12  Eric Botcazou  <ebotcazou@adacore.com>
16161         * config/sparc/sparc.c (output_return): Output the source location of
16162         the insn in the delay slot, if any.
16163         (output_sibcall): Likewise.
16165 2017-09-12  H.J. Lu  <hongjiu.lu@intel.com>
16167         PR driver/81498
16168         * common.opt (-static-pie): New alias.
16169         (shared): Negate static-pie.
16170         (-no-pie): Update help text.
16171         (-pie): Likewise.
16172         (static-pie): New option.
16173         * config/gnu-user.h (GNU_USER_TARGET_STARTFILE_SPEC): Add
16174         -static-pie support.
16175         (GNU_USER_TARGET_ENDFILE_SPEC): Likewise.
16176         (LINK_EH_SPEC): Likewise.
16177         (LINK_GCC_C_SEQUENCE_SPEC): Likewise.
16178         * config/i386/gnu-user.h (GNU_USER_TARGET_LINK_SPEC): Likewise.
16179         * config/i386/gnu-user64.h (GNU_USER_TARGET_LINK_SPEC): Likewise.
16180         * gcc.c (LINK_COMMAND_SPEC): Likewise.
16181         (init_gcc_specs): Likewise.
16182         (init_spec): Likewise.
16183         (display_help): Update help message for -pie.
16184         * doc/invoke.texi: Update -pie, -no-pie and -static.  Document
16185         -static-pie.
16187 2017-09-12  Wilco Dijkstra  <wdijkstr@arm.com>
16189         * config/aarch64/aarch64.md (movsi_aarch64): Remove all '*'.
16190         (movdi_aarch64): Likewise.
16191         (movti_aarch64): Likewise.
16193 2017-09-12 Simon Wright <simon@pushface.org>
16195         PR target/80204
16196         * config/darwin-driver.c (darwin_find_version_from_kernel): Eliminate
16197         calculation of the minor version, always output as 0.
16199 2017-09-12  Jakub Jelinek  <jakub@redhat.com>
16201         PR target/82112
16202         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): For
16203         ALTIVEC_BUILTIN_VEC_LD if arg1 has array type call default_conversion
16204         on it early, rather than manual conversion late.  For
16205         ALTIVEC_BUILTIN_VEC_ST if arg2 has array type call default_conversion
16206         instead of performing manual conversion.
16208 2017-09-12  Carl Love  <cel@us.ibm.com>
16210         * config/rs6000/altivec.md (vec_widen_umult_even_v4si,
16211         vec_widen_smult_even_v4si): Add define expands for vmuleuw, vmulesw,
16212         vmulouw, vmulosw.
16213         * config/rs6000/rs6000-builtin.def (VMLEUW, VMULESW, VMULOUW,
16214         VMULOSW): Add definitions.
16215         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
16216         ALTIVEC_BUILTIN_VMULESW, ALTIVEC_BUILTIN_VMULEUW,
16217         ALTIVEC_BUILTIN_VMULOSW, ALTIVEC_BUILTIN_VMULOUW entries.
16218         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin,
16219         builtin_function_type): Add ALTIVEC_BUILTIN_* case statements.
16221 2017-09-12  James Greenhalgh  <james.greenhalgh@arm.com>
16223         * config/aarch64/aarch64.md (movdi_aarch64): Set load/store
16224         types correctly.
16225         (movti_aarch64): Likewise.
16226         (movdf_aarch64): Likewise.
16227         (movtf_aarch64): Likewise.
16228         (load_pairdi): Likewise.
16229         (store_pairdi): Likewise.
16230         (load_pairdf): Likewise.
16231         (store_pairdf): Likewise.
16232         (loadwb_pair<GPI:mode>_<P:mode>): Likewise.
16233         (storewb_pair<GPI:mode>_<P:mode>): Likewise.
16234         (ldr_got_small_<mode>): Likewise.
16235         (ldr_got_small_28k_<mode>): Likewise.
16236         (ldr_got_tiny): Likewise.
16237         * config/aarch64/iterators.md (ldst_sz): New.
16238         (ldpstp_sz): Likewise.
16239         * config/aarch64/thunderx.md (thunderx_storepair): Split store_8
16240         to store_16.
16241         (thunderx_load): Split load_8 to load_16.
16242         * config/aarch64/thunderx2t99.md (thunderx2t99_loadpair): Split
16243         load_8 to load_16.
16244         (thunderx2t99_storepair_basic): Split store_8 to store_16.
16245         * config/arm/xgene1.md (xgene1_load_pair): Split load_8 to load_16.
16246         (xgene1_store_pair): Split store_8 to store_16.
16247         * config/aarch64/falkor.md (falkor_ld_3_ld): Split load_8 to load_16.
16248         (falkor_st_0_st_sd): Split store_8 to store_16.
16250 2017-09-12  James Greenhalgh  <james.greenhalgh@arm.com>
16252         * config/arm/types.md (type): Rename load1/2/3/4 to load_4/8/12/16
16253         and store1/2/3/4 to store_4/8/12/16.
16254         * config/aarch64/aarch64.md: Update for rename.
16255         * config/arm/arm.md: Likewise.: Likewise.
16256         * config/arm/arm.c: Likewise.
16257         * config/arm/thumb1.md: Likewise.
16258         * config/arm/thumb2.md: Likewise.
16259         * config/arm/vfp.md: Likewise.
16260         * config/arm/arm-generic.md: Likewise.
16261         * config/arm/arm1020e.md: Likewise.
16262         * config/arm/arm1026ejs.md: Likewise.
16263         * config/arm/arm1136jfs.md: Likewise.
16264         * config/arm/arm926ejs.md: Likewise.
16265         * config/arm/cortex-a15.md: Likewise.
16266         * config/arm/cortex-a17.md: Likewise.
16267         * config/arm/cortex-a5.md: Likewise.
16268         * config/arm/cortex-a53.md: Likewise.
16269         * config/arm/cortex-a57.md: Likewise.
16270         * config/arm/cortex-a7.md: Likewise.
16271         * config/arm/cortex-a8.md: Likewise.
16272         * config/arm/cortex-a9.md: Likewise.
16273         * config/arm/cortex-m4.md: Likewise.
16274         * config/arm/cortex-m7.md: Likewise.
16275         * config/arm/cortex-r4.md: Likewise.
16276         * config/arm/exynos-m1.md: Likewise.
16277         * config/arm/fa526.md: Likewise.
16278         * config/arm/fa606te.md: Likewise.
16279         * config/arm/fa626te.md: Likewise.
16280         * config/arm/fa726te.md: Likewise.
16281         * config/arm/fmp626.md: Likewise.
16282         * config/arm/iwmmxt.md: Likewise.
16283         * config/arm/ldmstm.md: Likewise.
16284         * config/arm/marvell-pj4.md: Likewise.
16285         * config/arm/xgene1.md: Likewise.
16286         * config/aarch64/thunderx.md: Likewise.
16287         * config/aarch64/thunderx2t99.md: Likewise.
16288         * config/aarch64/falkor.md: Likewise.
16290 2017-09-12  Martin Liska  <mliska@suse.cz>
16292         * attribs.c (private_lookup_attribute): New function.
16293         * attribs.h (private_lookup_attribute): Declared here.
16294         (lookup_attribute): Called from this place.
16296 2017-09-12  Richard Biener  <rguenther@suse.de>
16298         PR tree-optimization/82157
16299         * tree-ssa-pre.c (remove_dead_inserted_code): Do not remove
16300         stmts with side-effects.
16302 2017-09-12  Richard Sandiford  <richard.sandiford@linaro.org>
16303             Alan Hayward  <alan.hayward@arm.com>
16304             David Sherwood <david.sherwood@arm.com>
16306         * target.def (hard_regno_nregs): New hook.
16307         (class_max_nregs): Refer to it instead of HARD_REGNO_NREGS.
16308         * targhooks.h (default_hard_regno_nregs): Declare.
16309         * targhooks.c (default_hard_regno_nregs): New function.
16310         * doc/tm.texi.in (HARD_REGNO_NREGS): Replace with...
16311         (TARGET_HARD_REGNO_NREGS): ...this hook.
16312         (HARD_REGNO_NREGS_HAS_PADDING): Update accordingly.
16313         (CLASS_MAX_NREGS): Likewise.
16314         * doc/tm.texi: Regenerate.
16315         * reginfo.c (init_reg_modes_target): Use targetm.hard_regno_nregs
16316         instead of HARD_REGNO_NREGS.
16317         * rtl.h (REG_NREGS): Refer to TARGET_HARD_REGNO_NREGS rather than
16318         HARD_REGNO_NREGS in the comment.
16319         * config/aarch64/aarch64.h (HARD_REGNO_NREGS): Delete.
16320         * config/aarch64/aarch64-protos.h (aarch64_hard_regno_nregs): Delete.
16321         * config/aarch64/aarch64.c (aarch64_hard_regno_nregs): Make static.
16322         Return an unsigned int.
16323         (TARGET_HARD_REGNO_NREGS): Redefine.
16324         * config/alpha/alpha.h (HARD_REGNO_NREGS): Delete.
16325         * config/arc/arc.h (HARD_REGNO_NREGS): Delete.
16326         * config/arc/arc.c (TARGET_HARD_REGNO_NREGS): Redefine.
16327         (arc_hard_regno_nregs): New function.
16328         * config/arm/arm.h (HARD_REGNO_NREGS): Delete.
16329         * config/arm/arm.c (TARGET_HARD_REGNO_NREGS): Redefine.
16330         (arm_hard_regno_nregs): New function.
16331         * config/avr/avr.h (HARD_REGNO_NREGS): Delete.
16332         * config/bfin/bfin.h (HARD_REGNO_NREGS): Delete.
16333         * config/bfin/bfin.c (bfin_hard_regno_nregs): New function.
16334         (TARGET_HARD_REGNO_NREGS): Redefine.
16335         * config/c6x/c6x.h (HARD_REGNO_NREGS): Delete.
16336         * config/cr16/cr16.h (LONG_REG_P): Use targetm.hard_regno_nregs.
16337         (HARD_REGNO_NREGS): Delete.
16338         * config/cr16/cr16.c (TARGET_HARD_REGNO_NREGS): Redefine.
16339         (cr16_hard_regno_nregs): New function.
16340         (cr16_memory_move_cost): Use it instead of HARD_REGNO_NREGS.
16341         * config/cris/cris.h (HARD_REGNO_NREGS): Delete.
16342         * config/cris/cris.c (TARGET_HARD_REGNO_NREGS): Redefine.
16343         (cris_hard_regno_nregs): New function.
16344         * config/epiphany/epiphany.h (HARD_REGNO_NREGS): Delete.
16345         * config/fr30/fr30.h (HARD_REGNO_NREGS): Delete.
16346         (CLASS_MAX_NREGS): Use targetm.hard_regno_nregs.
16347         * config/frv/frv.h (HARD_REGNO_NREGS): Delete.
16348         (CLASS_MAX_NREGS): Remove outdated copy of documentation.
16349         * config/frv/frv-protos.h (frv_hard_regno_nregs): Delete.
16350         * config/frv/frv.c (TARGET_HARD_REGNO_NREGS): Redefine.
16351         (frv_hard_regno_nregs): Make static.  Take and return an
16352         unsigned int.
16353         (frv_class_max_nregs): Remove outdated copy of documentation.
16354         * config/ft32/ft32.h (HARD_REGNO_NREGS): Delete.
16355         * config/h8300/h8300.h (HARD_REGNO_NREGS): Delete.
16356         * config/h8300/h8300-protos.h (h8300_hard_regno_nregs): Delete.
16357         * config/h8300/h8300.c (h8300_hard_regno_nregs): Delete.
16358         * config/i386/i386.h (HARD_REGNO_NREGS): Delete.
16359         * config/i386/i386.c (ix86_hard_regno_nregs): New function.
16360         (TARGET_HARD_REGNO_NREGS): Redefine.
16361         * config/ia64/ia64.h (HARD_REGNO_NREGS): Delete.
16362         (CLASS_MAX_NREGS): Update comment.
16363         * config/ia64/ia64.c (TARGET_HARD_REGNO_NREGS): Redefine.
16364         (ia64_hard_regno_nregs): New function.
16365         * config/iq2000/iq2000.h (HARD_REGNO_NREGS): Delete.
16366         * config/lm32/lm32.h (HARD_REGNO_NREGS): Delete.
16367         * config/m32c/m32c.h (HARD_REGNO_NREGS): Delete.
16368         * config/m32c/m32c-protos.h (m32c_hard_regno_nregs): Delete.
16369         * config/m32c/m32c.c (m32c_hard_regno_nregs_1): Take and return
16370         an unsigned int.
16371         (m32c_hard_regno_nregs): Likewise.  Make static.
16372         (TARGET_HARD_REGNO_NREGS): Redefine.
16373         * config/m32r/m32r.h (HARD_REGNO_NREGS): Delete.
16374         * config/m68k/m68k.h (HARD_REGNO_NREGS): Delete.
16375         * config/m68k/m68k.c (TARGET_HARD_REGNO_NREGS): Redefine.
16376         (m68k_hard_regno_nregs): New function.
16377         * config/mcore/mcore.h (HARD_REGNO_NREGS): Delete.
16378         * config/microblaze/microblaze.h (HARD_REGNO_NREGS): Delete.
16379         * config/mips/mips.h (HARD_REGNO_NREGS): Delete.
16380         * config/mips/mips-protos.h (mips_hard_regno_nregs): Delete.
16381         * config/mips/mips.c (mips_hard_regno_nregs): Make static.
16382         Take and return an unsigned int.
16383         (TARGET_HARD_REGNO_NREGS): Redefine.
16384         * config/mmix/mmix.h (HARD_REGNO_NREGS): Delete.
16385         (CLASS_MAX_NREGS): Use targetm.hard_regno_nregs.
16386         * config/mn10300/mn10300.h (HARD_REGNO_NREGS): Delete.
16387         * config/moxie/moxie.h (HARD_REGNO_NREGS): Delete.
16388         * config/msp430/msp430.h (HARD_REGNO_NREGS): Delete.
16389         * config/msp430/msp430-protos.h (msp430_hard_regno_nregs): Delete.
16390         * config/msp430/msp430.c (TARGET_HARD_REGNO_NREGS): Redefine.
16391         (msp430_hard_regno_nregs): Make static.  Take and return an
16392         unsigned int.
16393         * config/nds32/nds32.h (HARD_REGNO_NREGS): Delete.
16394         * config/nds32/nds32-protos.h (nds32_hard_regno_nregs): Delete.
16395         * config/nds32/nds32.c (nds32_hard_regno_nregs): Delete.
16396         (nds32_hard_regno_mode_ok): Use targetm.hard_regno_nregs.
16397         * config/nios2/nios2.h (HARD_REGNO_NREGS): Delete.
16398         * config/nvptx/nvptx.h (HARD_REGNO_NREGS): Delete.
16399         * config/nvptx/nvptx.c (nvptx_hard_regno_nregs): New function.
16400         (TARGET_HARD_REGNO_NREGS): Redefine.
16401         * config/pa/pa32-regs.h (HARD_REGNO_NREGS): Rename to...
16402         (PA_HARD_REGNO_NREGS): ...this.
16403         * config/pa/pa64-regs.h (HARD_REGNO_NREGS): Rename to...
16404         (PA_HARD_REGNO_NREGS): ...this.
16405         * config/pa/pa.c (TARGET_HARD_REGNO_NREGS): Redefine.
16406         (pa_hard_regno_nregs): New function.
16407         * config/pdp11/pdp11.h (HARD_REGNO_NREGS): Delete.
16408         * config/pdp11/pdp11.c (TARGET_HARD_REGNO_NREGS): Redefine.
16409         (pdp11_hard_regno_nregs): New function.
16410         * config/powerpcspe/powerpcspe.h (HARD_REGNO_NREGS): Delete.
16411         * config/powerpcspe/powerpcspe.c (TARGET_HARD_REGNO_NREGS): Redefine.
16412         (rs6000_hard_regno_nregs_hook): New function.
16413         * config/riscv/riscv.h (HARD_REGNO_NREGS): Delete.
16414         * config/riscv/riscv-protos.h (riscv_hard_regno_nregs): Delete.
16415         * config/riscv/riscv.c (riscv_hard_regno_nregs): Make static.
16416         Take and return an unsigned int.  Move earlier in file.
16417         (TARGET_HARD_REGNO_NREGS): Redefine.
16418         * config/rl78/rl78.h (HARD_REGNO_NREGS): Delete.
16419         * config/rl78/rl78-protos.h (rl78_hard_regno_nregs): Delete.
16420         * config/rl78/rl78.c (TARGET_HARD_REGNO_NREGS): Reefine.
16421         (rl78_hard_regno_nregs): Make static.  Take and return an
16422         unsigned int.
16423         * config/rs6000/rs6000.h (HARD_REGNO_NREGS): Delete.
16424         * config/rs6000/rs6000.c (TARGET_HARD_REGNO_NREGS): Redefine.
16425         (rs6000_hard_regno_nregs_hook): New function.
16426         * config/rx/rx.h (HARD_REGNO_NREGS): Delete.
16427         * config/rx/rx.c (rx_hard_regno_nregs): New function.
16428         (TARGET_HARD_REGNO_NREGS): Redefine.
16429         * config/s390/s390.h (HARD_REGNO_NREGS): Delete.
16430         * config/s390/s390.c (REGNO_PAIR_OK): Use s390_hard_regno_nregs
16431         instead of HARD_REGNO_NREGS.
16432         (s390_hard_regno_nregs): New function.
16433         (s390_hard_regno_mode_ok): Add comment from s390.h.
16434         (TARGET_HARD_REGNO_NREGS): Redefine.
16435         * config/sh/sh.h (HARD_REGNO_NREGS): Delete.
16436         * config/sh/sh.c (TARGET_HARD_REGNO_NREGS): Redefine.
16437         (sh_hard_regno_nregs): New function.
16438         (sh_pass_in_reg_p): Use it.
16439         * config/sparc/sparc.h (HARD_REGNO_NREGS): Delete.
16440         * config/sparc/sparc.c (TARGET_HARD_REGNO_NREGS): Redefine.
16441         (sparc_hard_regno_nregs): New function.
16442         * config/spu/spu.h (HARD_REGNO_NREGS): Delete.
16443         * config/spu/spu.c (spu_hard_regno_nregs): New function.
16444         (spu_function_arg_advance): Use it, supplying a valid register number.
16445         (TARGET_HARD_REGNO_NREGS): Redefine.
16446         * config/stormy16/stormy16.h (HARD_REGNO_NREGS): Delete.
16447         * config/tilegx/tilegx.h (HARD_REGNO_NREGS): Delete.
16448         * config/tilepro/tilepro.h (HARD_REGNO_NREGS): Delete.
16449         * config/v850/v850.h (HARD_REGNO_NREGS): Delete.
16450         * config/vax/vax.h (HARD_REGNO_NREGS): Delete.
16451         * config/visium/visium.h (HARD_REGNO_NREGS): Delete.
16452         (CLASS_MAX_NREGS): Remove copy of old documentation.
16453         * config/visium/visium.c (TARGET_HARD_REGNO_NREGS): Redefine.
16454         (visium_hard_regno_nregs): New function.
16455         (visium_hard_regno_mode_ok): Use it instead of HARD_REGNO_NREGS.
16456         * config/xtensa/xtensa.h (HARD_REGNO_NREGS): Delete.
16457         * config/xtensa/xtensa.c (TARGET_HARD_REGNO_NREGS): Redefine.
16458         xtensa_hard_regno_nregs): New function.
16459         * system.h (HARD_REGNO_NREGS): Poison.
16461 2017-09-12  Richard Sandiford  <richard.sandiford@linaro.org>
16463         * config/arm/arm.h (THUMB_SECONDARY_INPUT_RELOAD_CLASS): Use
16464         hard_regno_nregs instead of HARD_REGNO_NREGS.
16465         (THUMB_SECONDARY_OUTPUT_RELOAD_CLASS): Likewise.
16466         * config/c6x/c6x.c (c6x_expand_prologue): Likewise.
16467         (c6x_expand_epilogue): Likewise.
16468         * config/frv/frv.c (frv_alloc_temp_reg): Likewise.
16469         (frv_read_iacc_argument): Likewise.
16470         * config/sh/sh.c: Include regs.h.
16471         (sh_print_operand): Use hard_regno_nregs instead of HARD_REGNO_NREGS.
16472         (regs_used): Likewise.
16473         (output_stack_adjust): Likewise.
16474         * config/xtensa/xtensa.c (xtensa_copy_incoming_a7): Likewise.
16475         * expmed.c: Include regs.h.
16476         (store_bit_field_1): Use hard_regno_nregs instead of HARD_REGNO_NREGS.
16477         * ree.c: Include regs.h.
16478         (combine_reaching_defs): Use hard_regno_nregs instead of
16479         HARD_REGNO_NREGS.
16480         (add_removable_extension): Likewise.
16482 2017-09-12  Richard Sandiford  <richard.sandiford@linaro.org>
16484         * regs.h (hard_regno_nregs): Turn into a function.
16485         (end_hard_regno): Update accordingly.
16486         * caller-save.c (setup_save_areas): Likewise.
16487         (save_call_clobbered_regs): Likewise.
16488         (replace_reg_with_saved_mem): Likewise.
16489         (insert_restore): Likewise.
16490         (insert_save): Likewise.
16491         * combine.c (can_change_dest_mode): Likewise.
16492         (move_deaths): Likewise.
16493         (distribute_notes): Likewise.
16494         * config/mips/mips.c (mips_hard_regno_call_part_clobbered): Likewise.
16495         * config/powerpcspe/powerpcspe.c (rs6000_cannot_change_mode_class)
16496         (rs6000_split_multireg_move): Likewise.
16497         (rs6000_register_move_cost): Likewise.
16498         (rs6000_memory_move_cost): Likewise.
16499         * config/rs6000/rs6000.c (rs6000_cannot_change_mode_class): Likewise.
16500         (rs6000_split_multireg_move): Likewise.
16501         (rs6000_register_move_cost): Likewise.
16502         (rs6000_memory_move_cost): Likewise.
16503         * cselib.c (cselib_reset_table): Likewise.
16504         (cselib_lookup_1): Likewise.
16505         * emit-rtl.c (set_mode_and_regno): Likewise.
16506         * function.c (aggregate_value_p): Likewise.
16507         * ira-color.c (setup_profitable_hard_regs): Likewise.
16508         (check_hard_reg_p): Likewise.
16509         (calculate_saved_nregs): Likewise.
16510         (assign_hard_reg): Likewise.
16511         (improve_allocation): Likewise.
16512         (calculate_spill_cost): Likewise.
16513         * ira-emit.c (modify_move_list): Likewise.
16514         * ira-int.h (ira_hard_reg_set_intersection_p): Likewise.
16515         (ira_hard_reg_in_set_p): Likewise.
16516         * ira.c (setup_reg_mode_hard_regset): Likewise.
16517         (clarify_prohibited_class_mode_regs): Likewise.
16518         (check_allocation): Likewise.
16519         * lra-assigns.c (find_hard_regno_for_1): Likewise.
16520         (lra_setup_reg_renumber): Likewise.
16521         (setup_try_hard_regno_pseudos): Likewise.
16522         (spill_for): Likewise.
16523         (assign_hard_regno): Likewise.
16524         (setup_live_pseudos_and_spill_after_risky_transforms): Likewise.
16525         * lra-constraints.c (in_class_p): Likewise.
16526         (lra_constraint_offset): Likewise.
16527         (simplify_operand_subreg): Likewise.
16528         (lra_constraints): Likewise.
16529         (split_reg): Likewise.
16530         (split_if_necessary): Likewise.
16531         (invariant_p): Likewise.
16532         (inherit_in_ebb): Likewise.
16533         * lra-lives.c (process_bb_lives): Likewise.
16534         * lra-remat.c (reg_overlap_for_remat_p): Likewise.
16535         (get_hard_regs): Likewise.
16536         (do_remat): Likewise.
16537         * lra-spills.c (assign_spill_hard_regs): Likewise.
16538         * mode-switching.c (create_pre_exit): Likewise.
16539         * postreload.c (reload_combine_recognize_pattern): Likewise.
16540         * recog.c (peep2_find_free_register): Likewise.
16541         * regcprop.c (kill_value_regno): Likewise.
16542         (set_value_regno): Likewise.
16543         (copy_value): Likewise.
16544         (maybe_mode_change): Likewise.
16545         (find_oldest_value_reg): Likewise.
16546         (copyprop_hardreg_forward_1): Likewise.
16547         * regrename.c (check_new_reg_p): Likewise.
16548         (regrename_do_replace): Likewise.
16549         * reload.c (push_reload): Likewise.
16550         (combine_reloads): Likewise.
16551         (find_dummy_reload): Likewise.
16552         (operands_match_p): Likewise.
16553         (find_reloads): Likewise.
16554         (find_equiv_reg): Likewise.
16555         (reload_adjust_reg_for_mode): Likewise.
16556         * reload1.c (count_pseudo): Likewise.
16557         (count_spilled_pseudo): Likewise.
16558         (find_reg): Likewise.
16559         (clear_reload_reg_in_use): Likewise.
16560         (free_for_value_p): Likewise.
16561         (allocate_reload_reg): Likewise.
16562         (choose_reload_regs): Likewise.
16563         (reload_adjust_reg_for_temp): Likewise.
16564         (emit_reload_insns): Likewise.
16565         (delete_output_reload): Likewise.
16566         * rtlanal.c (subreg_get_info): Likewise.
16567         * sched-deps.c (sched_analyze_reg): Likewise.
16568         * sel-sched.c (init_regs_for_mode): Likewise.
16569         (mark_unavailable_hard_regs): Likewise.
16570         (choose_best_reg_1): Likewise.
16571         (verify_target_availability): Likewise.
16572         * valtrack.c (dead_debug_insert_temp): Likewise.
16573         * var-tracking.c (track_loc_p): Likewise.
16574         (emit_note_insn_var_location): Likewise.
16575         * varasm.c (make_decl_rtl): Likewise.
16576         * reginfo.c (choose_hard_reg_mode): Likewise.
16577         (init_reg_modes_target): Refer directly to
16578         this_target_regs->x_hard_regno_nregs.
16580 2017-09-12  Richard Sandiford  <richard.sandiford@linaro.org>
16582         * ira-costs.c (record_operand_costs): Use in_hard_reg_set_p
16583         instead of hard_regno_nregs.
16585 2017-09-12  Richard Sandiford  <richard.sandiford@linaro.org>
16587         * config/aarch64/aarch64.c (aarch64_hard_regno_mode_ok): Use
16588         end_hard_regno instead of hard_regno_nregs.
16589         * config/s390/s390.c (s390_reg_clobbered_rtx): Likewise.
16590         * config/sparc/sparc.h (ASM_DECLARE_REGISTER_GLOBAL): Likewise.
16591         * config/visium/visium.c (visium_hard_regno_mode_ok): Likewise.
16592         * ira-color.c (improve_allocation): Likewise.
16593         * lra-assigns.c (find_hard_regno_for_1): Likewise.
16594         * lra-lives.c (mark_regno_live): Likewise.
16595         (mark_regno_dead): Likewise.
16596         * lra-remat.c (operand_to_remat): Likewise.
16597         * lra.c (collect_non_operand_hard_regs): Likewise.
16598         * postreload.c (reload_combine_note_store): Likewise.
16599         (move2add_valid_value_p): Likewise.
16600         * reload.c (regno_clobbered_p): Likewise.
16602 2017-09-12  Richard Sandiford  <richard.sandiford@linaro.org>
16604         * config/frv/frv.c (FOR_EACH_REGNO): Use END_REGNO instead of
16605         hard_regno_nregs.
16606         * config/v850/v850.c (v850_reorg): Likewise.
16607         * reload.c (refers_to_regno_for_reload_p): Likewise.
16608         (find_equiv_reg): Likewise.
16609         * reload1.c (reload_reg_reaches_end_p): Likewise.
16611 2017-09-12  Richard Sandiford  <richard.sandiford@linaro.org>
16613         * caller-save.c (add_used_regs): Use REG_NREGS instead of
16614         hard_regno_nregs.
16615         * config/aarch64/aarch64.c (aarch64_split_combinev16qi): Likewise.
16616         * config/arm/arm.c (output_move_neon): Likewise.
16617         (arm_attr_length_move_neon): Likewise.
16618         (neon_split_vcombine): Likewise.
16619         * config/c6x/c6x.c (c6x_mark_reg_read): Likewise.
16620         (c6x_mark_reg_written): Likewise.
16621         (c6x_dwarf_register_span): Likewise.
16622         * config/i386/i386.c (ix86_save_reg): Likewise.
16623         * config/ia64/ia64.c (mark_reg_gr_used_mask): Likewise.
16624         (rws_access_reg): Likewise.
16625         * config/s390/s390.c (s390_call_saved_register_used): Likewise.
16626         * mode-switching.c (create_pre_exit): Likewise.
16627         * ree.c (combine_reaching_defs): Likewise.
16628         (add_removable_extension): Likewise.
16629         * regcprop.c (find_oldest_value_reg): Likewise.
16630         (copyprop_hardreg_forward_1): Likewise.
16631         * reload.c (reload_inner_reg_of_subreg): Likewise.
16632         (push_reload): Likewise.
16633         (combine_reloads): Likewise.
16634         (find_dummy_reload): Likewise.
16635         (reload_adjust_reg_for_mode): Likewise.
16636         * reload1.c (find_reload_regs): Likewise.
16637         (forget_old_reloads_1): Likewise.
16638         (reload_reg_free_for_value_p): Likewise.
16639         (reload_adjust_reg_for_temp): Likewise.
16640         (emit_reload_insns): Likewise.
16641         (delete_output_reload): Likewise.
16642         * sel-sched.c (choose_best_reg_1): Likewise.
16643         (choose_best_pseudo_reg): Likewise.
16645 2017-09-12  Richard Sandiford  <richard.sandiford@linaro.org>
16646             Alan Hayward  <alan.hayward@arm.com>
16647             David Sherwood <david.sherwood@arm.com>
16649         * defaults.h (SLOW_UNALIGNED_ACCESS): Delete.
16650         * target.def (slow_unaligned_access): New hook.
16651         * targhooks.h (default_slow_unaligned_access): Declare.
16652         * targhooks.c (default_slow_unaligned_access): New function.
16653         * doc/tm.texi.in (SLOW_UNALIGNED_ACCESS): Replace with...
16654         (TARGET_SLOW_UNALIGNED_ACCESS): ...this.
16655         * doc/tm.texi: Regenerate.
16656         * config/alpha/alpha.h (SLOW_UNALIGNED_ACCESS): Delete.
16657         * config/arm/arm.h (SLOW_UNALIGNED_ACCESS): Delete.
16658         * config/i386/i386.h (SLOW_UNALIGNED_ACCESS): Delete commented-out
16659         definition.
16660         * config/powerpcspe/powerpcspe.h (SLOW_UNALIGNED_ACCESS): Delete.
16661         * config/powerpcspe/powerpcspe.c (TARGET_SLOW_UNALIGNED_ACCESS):
16662         Redefine.
16663         (rs6000_slow_unaligned_access): New function.
16664         (rs6000_emit_move): Use it instead of SLOW_UNALIGNED_ACCESS.
16665         (expand_block_compare): Likewise.
16666         (expand_strn_compare): Likewise.
16667         (rs6000_rtx_costs): Likewise.
16668         * config/riscv/riscv.h (SLOW_UNALIGNED_ACCESS): Delete.
16669         (riscv_slow_unaligned_access): Likewise.
16670         * config/riscv/riscv.c (riscv_slow_unaligned_access): Rename to...
16671         (riscv_slow_unaligned_access_p): ...this and make static.
16672         (riscv_option_override): Update accordingly.
16673         (riscv_slow_unaligned_access): New function.
16674         (TARGET_SLOW_UNALIGNED_ACCESS): Redefine.
16675         * config/rs6000/rs6000.h (SLOW_UNALIGNED_ACCESS): Delete.
16676         * config/rs6000/rs6000.c (TARGET_SLOW_UNALIGNED_ACCESS): Redefine.
16677         (rs6000_slow_unaligned_access): New function.
16678         (rs6000_emit_move): Use it instead of SLOW_UNALIGNED_ACCESS.
16679         (rs6000_rtx_costs): Likewise.
16680         * config/rs6000/rs6000-string.c (expand_block_compare)
16681         (expand_strn_compare): Use targetm.slow_unaligned_access instead
16682         of SLOW_UNALIGNED_ACCESS.
16683         * config/tilegx/tilegx.h (SLOW_UNALIGNED_ACCESS): Delete.
16684         * config/tilepro/tilepro.h (SLOW_UNALIGNED_ACCESS): Delete.
16685         * calls.c (expand_call): Use targetm.slow_unaligned_access instead
16686         of SLOW_UNALIGNED_ACCESS.
16687         * expmed.c (simple_mem_bitfield_p): Likewise.
16688         * expr.c (alignment_for_piecewise_move): Likewise.
16689         (emit_group_load_1): Likewise.
16690         (emit_group_store): Likewise.
16691         (copy_blkmode_from_reg): Likewise.
16692         (emit_push_insn): Likewise.
16693         (expand_assignment): Likewise.
16694         (store_field): Likewise.
16695         (expand_expr_real_1): Likewise.
16696         * gimple-fold.c (gimple_fold_builtin_memory_op): Likewise.
16697         * lra-constraints.c (simplify_operand_subreg): Likewise.
16698         * stor-layout.c (bit_field_mode_iterator::next_mode): Likewise.
16699         * gimple-ssa-store-merging.c: Likewise in block comment at start
16700         of file.
16701         * tree-ssa-strlen.c: Include target.h.
16702         (handle_builtin_memcmp): Use targetm.slow_unaligned_access instead
16703         of SLOW_UNALIGNED_ACCESS.
16704         * system.h (SLOW_UNALIGNED_ACCESS): Poison.
16706 2017-09-12  Richard Sandiford  <richard.sandiford@linaro.org>
16708         PR rtl-optimization/82185
16709         * expmed.c (emit_store_flag_int): Only test tem if it has been
16710         initialized.
16712 2017-09-12  Richard Biener  <rguenther@suse.de>
16714         PR middle-end/82149
16715         * match.pd ((FTYPE) N CMP CST): Fix typo.
16717 2017-09-12  Simon Atanasyan  <simon.atanasyan@imgtec.com>
16719         * config/mips/mips.c (mips_attribute_table): Add 'short_call'
16720         attribute.
16721         (mips_near_type_p): Add 'short_call' attribute as a synonym
16722         for 'near'.
16723         * doc/extend.texi (short_call): Document new function attribute.
16725 2017-09-12  Jakub Jelinek  <jakub@redhat.com>
16727         PR target/82112
16728         * c-common.c (sync_resolve_size): Instead of c_dialect_cxx ()
16729         assertion check that in the condition.
16730         (get_atomic_generic_size): Likewise.  Before testing if parameter
16731         has pointer type, if it has array type, call for C++
16732         default_conversion to perform array-to-pointer conversion.
16734 2017-09-12  Richard Biener  <rguenther@suse.de>
16736         * tree-vect-generic.c (expand_vector_operations_1): Do nothing
16737         for operations we cannot scalarize.
16739 2017-09-12  Aldy Hernandez  <aldyh@redhat.com>
16741         * tree-ssa-threadbackward.c (fsm_find_thread_path): Make GC
16742         vectors heap vectors.  Clean up comments.
16743         Make visited_bbs a reference.
16744         (profitable_jump_thread_path): Make GC
16745         vectors heap vectors.  Clean up comments.
16746         Misc cleanups.
16747         (convert_and_register_jump_thread_path): Make GC vectors heap
16748         vectors.
16749         (check_subpath_and_update_thread_path): Same.  Clean up comments.
16750         Make visited_bbs a reference.
16751         (handle_phi): Abstract common code to to
16752         register_jump_thread_path_if_profitable.
16753         Rename VAR_BB to DEF_BB.
16754         Update comments.
16755         Make GC vectors heap vectors.
16756         Make visited_bbs a reference.
16757         (handle_assignment): Same.
16758         (register_jump_thread_path_if_profitable): New.
16759         (fsm_find_control_statement_thread_paths): Rename VAR_BB to
16760         DEF_BB.
16761         Make GC vectors heap vectors.  Clean up comments.
16762         Make visited_bbs a reference.
16763         (find_jump_threads_backwards): Make visited_bbs live in the stack.
16764         * tree-ssa-threadupdate.c (delete_jump_thread_path): Fix typo in
16765         comment.
16767 2017-09-11  Max Filippov  <jcmvbkbc@gmail.com>
16769         PR target/82181
16770         * config/xtensa/xtensa.c (xtensa_mem_offset): Check that both
16771         words of E_DImode object are reachable by xtensa_uimm8x4 access.
16773 2017-09-11  Vidya Praveen  <vidyapraveen@arm.com>
16775         Revert r251800 and r251799.
16777 2017-09-11  Martin Jambor  <mjambor@suse.cz>
16779         PR hsa/82119
16780         * hsa-gen.c (gen_hsa_phi_from_gimple_phi): Process ADDR_EXPRs in
16781         arguments in advance.
16782         * hsa-regalloc.c (naive_process_phi): New parameter predecessors,
16783         use it to find predecessor edges.
16784         (naive_outof_ssa): Collect vector of predecessors.
16786 2017-09-08  Jason Merrill  <jason@redhat.com>
16788         PR c++/70029 - ICE with ref-qualifier and -flto
16789         * langhooks.h (struct lang_hooks_for_types): Add
16790         copy_lang_qualifiers.
16791         * attribs.c (build_type_attribute_qual_variant): Use it.
16792         * langhooks-def.h (LANG_HOOKS_COPY_LANG_QUALIFIERS): Default to
16793         NULL.
16794         (LANG_HOOKS_FOR_TYPES_INITIALIZER): Use it.
16795         * tree.c (verify_type): Re-enable TYPE_CANONICAL main variant check.
16797 2017-09-08  Eric Botcazou  <ebotcazou@adacore.com>
16799         PR target/81988
16800         * config/sparc/sparc.md (mulsi3): Rename into *mulsi3_sp32.
16801         (*mulsi3_sp64): New instruction.
16802         (mulsi3): New expander.
16804 2017-09-08  Uros Bizjak  <ubizjak@gmail.com>
16806         * config/alpha/alpha.c (alpha_print_operand) <case 'S'>: Remove.
16808 2017-09-08  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
16810         * sancov.c: Include memmodel.h.
16812 2017-09-07  Eric Botcazou  <ebotcazou@adacore.com>
16814         PR target/80897
16815         * config/sparc/sparc.c (sparc_emit_set_symbolic_const64): Deal with too
16816         large offsets.
16818 2017-09-07  Carl Love  <cel@us.ibm.com>
16820         * config/rs6000/vsx.md (define_insn "*stxvl"): Add missing argument to
16821         the sldi instruction.
16823 2017-09-07  David Edelsohn  <dje.gcc@gmail.com>
16825         * sancov.c: Include tm_p.h.
16827 2017-09-07  Jakub Jelinek  <jakub@redhat.com>
16829         PR target/81979
16830         * output.h (switch_to_other_text_partition): New declaration.
16831         * varasm.c (switch_to_other_text_partition): New function.
16832         * config/rs6000/rs6000.c (uses_TOC): Return 2 if
16833         NOTE_INSN_SWITCH_TEXT_SECTIONS is seen before finding load_toc_* insn.
16834         (rs6000_elf_declare_function_name): If uses_TOC returned 2, switch
16835         to the other text partition before emitting LCL label and switch back
16836         after emitting the word after it.
16838 2017-09-07  Richard Biener  <rguenther@suse.de>
16840         * passes.def (pass_split_crit_edges): Remove instance before PRE.
16841         * tree-ssa-pre.c (pass_pre::execute): Instead manually split
16842         critical edges here, after loop init.
16843         (pass_data_pre): Remove PROP_no_crit_edges flags.
16844         * tree-ssa-sccvn.c (vn_reference_lookup_3): Use vn_valueize
16845         for valueization of call args to avoid leaking VN_TOP.
16846         (visit_use): Assert we do not visit default defs.
16847         (init_scc_vn): Use build_decl for VN_TOP to make name nicer.
16848         Use error_mark_node to more easily detect leaking VN_TOP.
16849         All default-defs are varying, not VN_TOP.  Mark them visited.
16850         (run_scc_vn): Make code match comment.
16852 2017-09-07  Michael Meissner  <meissner@linux.vnet.ibm.com>
16854         * config/rs6000/rs6000-cpus.def (OTHER_VSX_VECTOR_MASKS): Delete
16855         OPTION_MASK_FLOAT128_KEYWORD.
16856         (POWERPC_MASKS): Likewise.
16857         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Delete
16858         support for the -mfloat128-type option, and make -mfloat128
16859         default on PowerPC Linux systems.  Define or undefine
16860         __FLOAT128__ and  __FLOAT128_HARDWARE__ for the current options.
16861         Define __float128 to be __ieee128 if IEEE 128-bit support is
16862         enabled, or undefine it.
16863         (rs6000_cpu_cpp_builtins): Delete defining __FLOAT128__ here.
16864         Delete defining __FLOAT128_TYPE__.
16865         * config/rs6000/rs6000.opt (x_TARGET_FLOAT128_TYPE): Delete the
16866         -mfloat128-type option and make -mfloat128 default on PowerPC
16867         Linux systems.
16868         (TARGET_FLOAT128_TYPE): Likewise.
16869         (-mfloat128-type): Likewise.
16870         * config/rs6000/rs6000.c (rs6000_option_override_internal):
16871         Delete the -mfloat128-type option and make -mfloat128 default on
16872         PowerPC Linux systems.  Always use __ieee128 to be the keyword for
16873         the IEEE 128-bit type, and map __float128 to __ieee128 if IEEE
16874         128-bit floating point is enabled.  Change tests from using
16875         -mfloat128-type to -mfloat128.
16876         (rs6000_mangle_type): Use the correct mangling for the __float128
16877         type even if normal long double is restricted to 64-bits.
16878         (floatn_mode): Enable the _Float128 type by default on VSX Linux
16879         systems.
16880         * config/rs6000/rs6000.h (MASK_FLOAT128_TYPE): Delete.
16881         (MASK_FLOAT128_KEYWORD): Define new shortcut macro.
16882         (RS6000BTM_FLOAT128): Define in terms of -mfloat128, not
16883         -mfloat128-type.
16884         * doc/invoke.texi (RS/6000 and PowerPC Options): Update
16885         documentation for -mfloat128.
16887 2017-09-06  Olivier Hainque  <hainque@adacore.com>
16889         * config.gcc (powerpc-wrs-vxworksspe): Now match as vxworks*spe.
16891 2017-09-06  Wish Wu  <wishwu007@gmail.com>
16892             Jakub Jelinek  <jakub@redhat.com>
16894         * asan.c (initialize_sanitizer_builtins): Add
16895         BT_FN_VOID_UINT8_UINT8, BT_FN_VOID_UINT16_UINT16,
16896         BT_FN_VOID_UINT32_UINT32, BT_FN_VOID_UINT64_UINT64,
16897         BT_FN_VOID_FLOAT_FLOAT, BT_FN_VOID_DOUBLE_DOUBLE and
16898         BT_FN_VOID_UINT64_PTR variables.
16899         * builtin-types.def (BT_FN_VOID_UINT8_UINT8): New fn type.
16900         (BT_FN_VOID_UINT16_UINT16): Likewise.
16901         (BT_FN_VOID_UINT32_UINT32): Likewise.
16902         (BT_FN_VOID_FLOAT_FLOAT): Likewise.
16903         (BT_FN_VOID_DOUBLE_DOUBLE): Likewise.
16904         (BT_FN_VOID_UINT64_PTR): Likewise.
16905         * common.opt (flag_sanitize_coverage): New variable.
16906         (fsanitize-coverage=trace-pc): Remove.
16907         (fsanitize-coverage=): Add.
16908         * flag-types.h (enum sanitize_coverage_code): New enum.
16909         * fold-const.c (fold_range_test): Disable non-short-circuit
16910         optimization if flag_sanitize_coverage.
16911         (fold_truth_andor): Likewise.
16912         * tree-ssa-ifcombine.c (ifcombine_ifandif): Likewise.
16913         * opts.c (COVERAGE_SANITIZER_OPT): Define.
16914         (coverage_sanitizer_opts): New array.
16915         (get_closest_sanitizer_option): Add OPTS argument, handle also
16916         OPT_fsanitize_coverage_.
16917         (parse_sanitizer_options): Adjusted to also handle
16918         OPT_fsanitize_coverage_.
16919         (common_handle_option): Add OPT_fsanitize_coverage_.
16920         * sancov.c (instrument_comparison, instrument_switch): New function.
16921         (sancov_pass): Add trace-cmp support.
16922         * sanitizer.def (BUILT_IN_SANITIZER_COV_TRACE_CMP1,
16923         BUILT_IN_SANITIZER_COV_TRACE_CMP2, BUILT_IN_SANITIZER_COV_TRACE_CMP4,
16924         BUILT_IN_SANITIZER_COV_TRACE_CMP8,
16925         BUILT_IN_SANITIZER_COV_TRACE_CONST_CMP1,
16926         BUILT_IN_SANITIZER_COV_TRACE_CONST_CMP2,
16927         BUILT_IN_SANITIZER_COV_TRACE_CONST_CMP4,
16928         BUILT_IN_SANITIZER_COV_TRACE_CONST_CMP8,
16929         BUILT_IN_SANITIZER_COV_TRACE_CMPF, BUILT_IN_SANITIZER_COV_TRACE_CMPD,
16930         BUILT_IN_SANITIZER_COV_TRACE_SWITCH): New builtins.
16931         * doc/invoke.texi: Document -fsanitize-coverage=trace-cmp.
16933 2017-09-06  Richard Earnshaw  <rearnsha@arm.com>
16935         * config/arm/parsecpu.awk (fatal): Note that we've encountered an
16936         error.  Only quit immediately if parsing is complete.
16937         (BEGIN): Initialize fatal_err and parse_done.
16938         (begin fpu, end fpu): Check number of arguments.
16939         (begin arch, end arch): Likewise.
16940         (begin cpu, end cpu): Likewise.
16941         (cname, tune for, tune flags, architecture, fpu, option): Likewise.
16942         (optalias): Likewise.
16944 2017-09-06  Richard Earnshaw  <rearnsha@arm.com>
16946         * config.gcc (arm*-*-*): Don't add arm-isa.h to tm_p_file.
16947         * config/arm/arm-isa.h: Delete.  Move definitions to ...
16948         * arm-cpus.in: ... here.  Use new feature and fgroup values.
16949         * config/arm/arm.c (arm_option_override): Use lower case for feature
16950         bit names.
16951         * config/arm/arm.h (TARGET_HARD_FLOAT): Likewise.
16952         (TARGET_VFP3, TARGET_VFP5, TARGET_FMA): Likewise.
16953         * config/arm/parsecpu.awk (END): Add new command 'isa'.
16954         (isa_pfx): Delete.
16955         (print_isa_bits_for): New function.
16956         (gen_isa): New function.
16957         (gen_comm_data): Use print_isa_bits_for.
16958         (define feature): New keyword.
16959         (define fgroup): New keyword.
16960         * config/arm/t-arm (OPTIONS_H_EXTRA): Add arm-isa.h
16961         (arm-isa.h): Add rule to generate file.
16962         * common/config/arm/arm-common.c: (arm_canon_arch_option): Use lower
16963         case for feature bit names.
16965 2017-09-06  Richard Biener  <rguenther@suse.de>
16967         * tree-ssa-pre.c (NECESSARY): Remove.
16968         (create_expression_by_pieces): Do not touch pass-local flags.
16969         (insert_into_preds_of_block): Likewise.
16970         (do_pre_regular_insertion): Likewise.
16971         (eliminate_insert): Likewise.
16972         (eliminate_dom_walker::before_dom_children): Likewise.
16973         (fini_eliminate): Do not look at inserted_exprs.
16974         (mark_operand_necessary): Remove.
16975         (remove_dead_inserted_code): Replace with simple work-list
16976         algorithm based on inserted_exprs and SSA uses.
16977         (pass_pre::execute): Re-order fini_eliminate and
16978         remove_dead_inserted_code.
16980 2017-09-06  Olivier Hainque  <hainque@adacore.com>
16982         * config/powerpcspe/vxworks.h (VXCPU_FOR_8548): Correct definition
16983         for VxWorks 7.  Adjust surrounding comments.
16985 2017-09-06  Richard Biener  <rguenther@suse.de>
16987         * gimple-ssa-strength-reduction.c
16988         (find_candidates_dom_walker::before_dom_children): Also allow
16989         pointer types.
16991 2017-09-06  Richard Biener  <rguenther@suse.de>
16993         PR tree-optimization/82108
16994         * tree-vect-stmts.c (vectorizable_load): Fix pointer adjustment
16995         for gap in the non-permutation SLP case.
16997 2017-09-06  Martin Jambor  <mjambor@suse.cz>
16999         PR tree-optimization/82078
17000         * tree-sra.c (sort_and_splice_var_accesses): Move call to
17001         add_access_to_work_queue...
17002         (build_accesses_from_assign): ...here.
17003         (propagate_all_subaccesses): Make sure racc is the group
17004         representative, if there is one.
17006 2017-09-06  Jakub Jelinek  <jakub@redhat.com>
17008         PR middle-end/82095
17009         * varasm.c (categorize_decl_for_section): Use SECCAT_TBSS for TLS vars with
17010         NULL DECL_INITIAL.
17012 2017-09-06  Richard Biener  <rguenther@suse.de>
17014         * gimple-ssa-strength-reduction.c
17015         (find_candidates_dom_walker::before_doom_children): Use a
17016         type and not a mode check.
17018 2017-09-06  Bernd Edlinger  <bernd.edlinger@hotmail.de>
17020         PR target/77308
17021         * config/arm/predicates.md (arm_general_adddi_operand): Create new
17022         non-vfp predicate.
17023         * config/arm/arm.md (*arm_adddi3, *arm_subdi3): Use new predicates.
17025 2017-09-05  Jeff Law  <law@redhat.com>
17027         PR tree-optimization/64910
17028         * tree-ssa-reassoc.c (reassociate_bb): Restrict last change to
17029         cases where we have 3 or more operands.
17031 2017-09-05  Jakub Jelinek  <jakub@redhat.com>
17033         PR middle-end/81768
17034         * omp-low.c (lower_omp_for): Recompute tree invariant if
17035         gimple_omp_for_initial/final is ADDR_EXPR.
17037         PR middle-end/81768
17038         * omp-expand.c (expand_omp_simd): Force second operands of COND_EXPR
17039         into gimple val before gimplification fo the COND_EXPR.
17041 2017-09-05  Aldy Hernandez  <aldyh@redhat.com>
17043         * tree-ssa-threadupdate.c (duplicate_thread_path): Remove unused
17044         REGION_COPY argument.
17045         (thread_through_all_blocks): Remove unused argument to
17046         duplicate_thread_path.
17048 2017-09-05  Richard Sandiford  <richard.sandiford@linaro.org>
17049             Alan Hayward  <alan.hayward@arm.com>
17050             David Sherwood  <david.sherwood@arm.com>
17052         * config/aarch64/aarch64-protos.h (aarch64_gen_adjusted_ldpstp):
17053         Take a scalar_mode rather than a machine_mode.
17054         (aarch64_operands_adjust_ok_for_ldpstp): Likewise.
17055         * config/aarch64/aarch64.c (aarch64_simd_container_mode): Likewise.
17056         (aarch64_operands_adjust_ok_for_ldpstp): Likewise.
17057         (aarch64_gen_adjusted_ldpstp): Likewise.
17058         (aarch64_expand_vector_init): Use scalar_mode instead of machine_mode.
17060 2017-09-05  Richard Sandiford  <richard.sandiford@linaro.org>
17061             Alan Hayward  <alan.hayward@arm.com>
17062             David Sherwood  <david.sherwood@arm.com>
17064         * config/aarch64/aarch64-protos.h (aarch64_is_extend_from_extract):
17065         Take a scalar_int_mode instead of a machine_mode.
17066         (aarch64_mask_and_shift_for_ubfiz_p): Likewise.
17067         (aarch64_output_scalar_simd_mov_immediate): Likewise.
17068         (aarch64_simd_scalar_immediate_valid_for_move): Likewise.
17069         (aarch64_simd_attr_length_rglist): Delete.
17070         * config/aarch64/aarch64.c (aarch64_is_extend_from_extract): Take
17071         a scalar_int_mode instead of a machine_mode.
17072         (aarch64_add_offset): Likewise.
17073         (aarch64_internal_mov_immediate): Likewise
17074         (aarch64_add_constant_internal): Likewise.
17075         (aarch64_add_constant): Likewise.
17076         (aarch64_movw_imm): Likewise.
17077         (aarch64_rtx_arith_op_extract_p): Likewise.
17078         (aarch64_mask_and_shift_for_ubfiz_p): Likewise.
17079         (aarch64_simd_scalar_immediate_valid_for_move): Likewise.
17080         Remove assert that the mode isn't a vector.
17081         (aarch64_output_scalar_simd_mov_immediate): Likewise.
17082         (aarch64_expand_mov_immediate): Update calls after above changes.
17083         (aarch64_output_casesi): Use as_a <scalar_int_mode>.
17084         (aarch64_and_bitmask_imm): Check for scalar integer modes.
17085         (aarch64_move_imm): Likewise.
17086         (aarch64_can_const_movi_rtx_p): Likewise.
17087         (aarch64_strip_extend): Likewise.
17088         (aarch64_extr_rtx_p): Likewise.
17089         (aarch64_rtx_costs): Likewise, using wode_mode as the mode of
17090         a CONST_INT when the mode parameter is VOIDmode.
17091         (aarch64_float_const_rtx_p): Use scalar_int_mode for a temporary.
17093 2017-09-05  Richard Sandiford  <richard.sandiford@linaro.org>
17095         * machmode.h (bitwise_mode_for_mode): Return opt_mode.
17096         * stor-layout.c (bitwise_mode_for_mode): Likewise.
17097         (bitwise_type_for_mode): Update accordingly.
17099 2017-09-05  Richard Sandiford  <richard.sandiford@linaro.org>
17101         * stor-layout.h (mode_for_size_tree): Return an opt_mode.
17102         * stor-layout.c (mode_for_size_tree): Likewise.
17103         (mode_for_array): Update accordingly.
17104         (layout_decl): Likewise.
17105         (compute_record_mode): Likewise.  Only set the mode once.
17107 2017-09-05  Richard Sandiford  <richard.sandiford@linaro.org>
17109         * target.def (get_mask_mode): Change return type to opt_mode.
17110         Expand commentary.
17111         * doc/tm.texi: Regenerate.
17112         * targhooks.h (default_get_mask_mode): Return an opt_mode.
17113         * targhooks.c (default_get_mask_mode): Likewise.
17114         * config/i386/i386.c (ix86_get_mask_mode): Likewise.
17115         * optabs-query.c (can_vec_mask_load_store_p): Update use of
17116         targetm.get_mask_mode.
17117         * tree.c (build_truth_vector_type): Likewise.
17119 2017-09-05  Richard Sandiford  <richard.sandiford@linaro.org>
17121         * machmode.h (mode_for_vector): Return an opt_mode.
17122         * stor-layout.c (mode_for_vector): Likewise.
17123         (mode_for_int_vector): Update accordingly.
17124         (layout_type): Likewise.
17125         * config/i386/i386.c (emit_memmov): Likewise.
17126         (ix86_expand_set_or_movmem): Likewise.
17127         (ix86_expand_vector_init): Likewise.
17128         (ix86_get_mask_mode): Likewise.
17129         * config/powerpcspe/powerpcspe.c (rs6000_expand_vec_perm_const_1):
17130         Likewise.
17131         * config/rs6000/rs6000.c (rs6000_expand_vec_perm_const_1): Likewise.
17132         * expmed.c (extract_bit_field_1): Likewise.
17133         * expr.c (expand_expr_real_2): Likewise.
17134         * optabs-query.c (can_vec_perm_p): Likewise.
17135         (can_vec_mask_load_store_p): Likewise.
17136         * optabs.c (expand_vec_perm): Likewise.
17137         * targhooks.c (default_get_mask_mode): Likewise.
17138         * tree-vect-stmts.c (vectorizable_store): Likewise.
17139         (vectorizable_load): Likewise.
17140         (get_vectype_for_scalar_type_and_size): Likewise.
17142 2017-09-05  Richard Sandiford  <richard.sandiford@linaro.org>
17144         * machmode.h (mode_for_int_vector): New function.
17145         * stor-layout.c (mode_for_int_vector): Likewise.
17146         * config/aarch64/aarch64.c (aarch64_emit_approx_sqrt): Use it.
17147         * config/powerpcspe/powerpcspe.c (rs6000_do_expand_vec_perm): Likewise.
17148         * config/rs6000/rs6000.c (rs6000_do_expand_vec_perm): Likewise.
17149         * config/s390/s390.c (s390_expand_vec_compare_cc): Likewise.
17150         (s390_expand_vcond): Likewise.
17152 2017-09-05  Richard Sandiford  <richard.sandiford@linaro.org>
17154         * machmode.h (opt_machine_mode): New type.
17155         (opt_mode<T>): Allow construction from anything that can be
17156         converted to a T.
17157         (is_a, as_a, dyn_cast): Add overloads for opt_mode.
17158         (mode_for_size): Return an opt_machine_mode.
17159         * stor-layout.c (mode_for_size): Likewise.
17160         (mode_for_size_tree): Update call accordingly.
17161         (bitwise_mode_for_mode): Likewise.
17162         (make_fract_type): Likewise.
17163         (make_accum_type): Likewise.
17164         * caller-save.c (replace_reg_with_saved_mem): Update call
17165         accordingly.
17166         * config/alpha/alpha.h (SECONDARY_MEMORY_NEEDED_MODE): Likewise.
17167         * config/i386/i386.h (SECONDARY_MEMORY_NEEDED_MODE): Likewise.
17168         * config/s390/s390.h (SECONDARY_MEMORY_NEEDED_MODE): Likewise.
17169         * config/sparc/sparc.h (SECONDARY_MEMORY_NEEDED_MODE): Likewise.
17170         * expmed.c (extract_bit_field_1): Likewise.
17171         * reload.c (get_secondary_mem): Likewise.
17172         * varasm.c (assemble_integer): Likewise.
17173         * lower-subreg.c (simplify_subreg_concatn): Likewise.  Move
17174         early-out.
17176 2017-09-05  Richard Sandiford  <richard.sandiford@linaro.org>
17178         * machmode.h (decimal_float_mode_for_size): New function.
17179         * real.h (REAL_VALUE_TO_TARGET_LONG_DOUBLE): Use float_mode_for_size.
17180         (REAL_VALUE_TO_TARGET_DOUBLE): Likewise.
17181         (REAL_VALUE_TO_TARGET_SINGLE): Likewise.
17182         (REAL_VALUE_TO_TARGET_DECIMAL128): Use decimal_float_mode_for_size.
17183         (REAL_VALUE_TO_TARGET_DECIMAL64): Likewise.
17184         (REAL_VALUE_TO_TARGET_DECIMAL32): Likewise.
17186 2017-09-05  Richard Sandiford  <richard.sandiford@linaro.org>
17188         * builtins.c (expand_builtin_powi): Use int_mode_for_size.
17189         (get_builtin_sync_mode): Likewise.
17190         (expand_ifn_atomic_compare_exchange): Likewise.
17191         (expand_builtin_atomic_clear): Likewise.
17192         (expand_builtin_atomic_test_and_set): Likewise.
17193         (fold_builtin_atomic_always_lock_free): Likewise.
17194         * calls.c (compute_argument_addresses): Likewise.
17195         (emit_library_call_value_1): Likewise.
17196         (store_one_arg): Likewise.
17197         * combine.c (combine_instructions): Likewise.
17198         * config/aarch64/aarch64.c (aarch64_function_value): Likewise.
17199         * config/arm/arm.c (arm_function_value): Likewise.
17200         (aapcs_allocate_return_reg): Likewise.
17201         * config/c6x/c6x.c (c6x_expand_movmem): Likewise.
17202         * config/i386/i386.c (construct_container): Likewise.
17203         (ix86_gimplify_va_arg): Likewise.
17204         (ix86_expand_sse_cmp): Likewise.
17205         (emit_memmov): Likewise.
17206         (emit_memset): Likewise.
17207         (expand_small_movmem_or_setmem): Likewise.
17208         (ix86_expand_pextr): Likewise.
17209         (ix86_expand_pinsr): Likewise.
17210         * config/lm32/lm32.c (lm32_block_move_inline): Likewise.
17211         * config/microblaze/microblaze.c (microblaze_block_move_straight):
17212         Likewise.
17213         * config/mips/mips.c (mips_function_value_1) Likewise.
17214         (mips_block_move_straight): Likewise.
17215         (mips_expand_ins_as_unaligned_store): Likewise.
17216         * config/powerpcspe/powerpcspe.c
17217         (rs6000_darwin64_record_arg_advance_flush): Likewise.
17218         (rs6000_darwin64_record_arg_flush): Likewise.
17219         * config/rs6000/rs6000.c
17220         (rs6000_darwin64_record_arg_advance_flush): Likewise.
17221         (rs6000_darwin64_record_arg_flush): Likewise.
17222         * config/sparc/sparc.c (sparc_function_arg_1): Likewise.
17223         (sparc_function_value_1): Likewise.
17224         * config/spu/spu.c (adjust_operand): Likewise.
17225         (spu_emit_branch_or_set): Likewise.
17226         (arith_immediate_p): Likewise.
17227         * emit-rtl.c (gen_lowpart_common): Likewise.
17228         * expr.c (expand_expr_real_1): Likewise.
17229         * function.c (assign_parm_setup_block): Likewise.
17230         * gimple-ssa-store-merging.c (encode_tree_to_bitpos): Likewise.
17231         * reload1.c (alter_reg): Likewise.
17232         * stor-layout.c (mode_for_vector): Likewise.
17233         (layout_type): Likewise.
17235 2017-09-05  Richard Sandiford  <richard.sandiford@linaro.org>
17237         * config/spu/spu.c (exp2_immediate_p): Use int_mode_for_mode.
17238         (spu_convert_move): Likewise.
17239         * lower-subreg.c (resolve_simple_move): Likewise.
17241 2017-09-05  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
17243         PR target/81833
17244         * config/rs6000/altivec.md (altivec_vsum2sws): Convert from a
17245         define_insn to a define_expand.
17246         (altivec_vsum2sws_direct): New define_insn.
17247         (altivec_vsumsws): Convert from a define_insn to a define_expand.
17249 2017-09-05  Wilco Dijkstra  <wdijkstr@arm.com>
17251         * config/arm/arm.c (arm_option_params_internal): Improve setting of
17252         max_insns_skipped.
17254 2017-09-05  H.J. Lu  <hongjiu.lu@intel.com>
17256         PR target/59501
17257         PR target/81624
17258         PR target/81769
17259         * config/i386/i386.c (ix86_finalize_stack_frame_flags): Don't
17260         realign stack if stack alignment needed is less than incoming
17261         stack boundary.
17263 2017-09-05  Marek Polacek  <polacek@redhat.com>
17265         PR sanitizer/82072
17266         * convert.c (convert_to_integer_1) <case NEGATE_EXPR>: Move the ubsan
17267         check earlier.
17269 2017-09-05  Wilco Dijkstra  <wdijkstr@arm.com>
17271         * explow.c (get_dynamic_stack_size): Improve dynamic alignment.
17273 2017-09-05  Richard Biener  <rguenther@suse.de>
17275         PR tree-optimization/82084
17276         * fold-const.c (can_native_encode_string_p): Handle wide characters.
17278 2017-09-05  Richard Biener  <rguenther@suse.de>
17280         PR tree-optimization/82102
17281         * tree-ssa-pre.c (fini_eliminate): Check if lhs is NULL.
17283 2017-09-05  Martin Liska  <mliska@suse.cz>
17285         PR tree-optimization/82032
17286         * tree-cfg.c (generate_range_test): New function.
17287         * tree-cfg.h (generate_range_test): Declared here.
17288         * tree-cfgcleanup.c (convert_single_case_switch): New function.
17289         (cleanup_control_expr_graph): Use it.
17290         * tree-switch-conversion.c (try_switch_expansion): Remove
17291         assert.
17292         (emit_case_nodes): Use generate_range_test.
17294 2017-09-04  Uros Bizjak  <ubizjak@gmail.com>
17296         PR target/82098
17297         * config/i386/i386.md (*<btsc><mode>_mask): Add
17298         TARGET_USE_BT to insn constraint.
17299         (*btr<mode>_mask): Ditto.
17301 2017-09-04  Wilco Dijkstra  <wdijkstr@arm.com>
17303         * config/arm/arm.c (arm_legitimate_index_p): Add comment.
17304         (thumb2_legitimate_index_p): Use correct range for DI/DF mode.
17306 2017-09-04  Bernd Edlinger  <bernd.edlinger@hotmail.de>
17308         PR target/77308
17309         * config/arm/arm.md (*arm_adddi3, *arm_subdi3): Split early except for
17310         TARGET_NEON and TARGET_IWMMXT.
17311         (anddi3, iordi3, xordi3, one_cmpldi2): Split while expanding except for
17312         TARGET_NEON and TARGET_IWMMXT.
17313         (*one_cmpldi2_insn): Moved the body of one_cmpldi2 here.
17315 2017-09-04  Uros Bizjak  <ubizjak@gmail.com>
17317         * config/i386/i386-protos.h (ix86_tls_address_pattern_p) New prototype.
17318         (ix86_rewrite_tls_address): Ditto.
17319         * config/i386/i386.c (ix86_tls_address_pattern_p) New function.
17320         (ix86_rewrite_tls_address_1): Ditto.
17321         (ix86_rewrite_tls_address): Ditto.
17322         * config/i386/predicates.md (tls_address_pattern): New predicate.
17323         * config/i386/i386.md (TLS address splitter): New splitter.
17325 2017-09-04  Richard Biener  <rguenther@suse.de>
17327         PR tree-optimization/82084
17328         * fold-const.h (can_native_encode_string_p): Declare.
17329         * fold-const.c (can_native_encode_string_p): Factor out from ...
17330         (native_encode_string): ... here.
17331         * tree-vect-stmts.c (vectorizable_store): Call it to avoid
17332         vectorizing stores from constants we later cannot handle.
17334 2017-09-04  Marek Polacek  <polacek@redhat.com>
17336         PR c/81783
17337         * doc/invoke.texi: Update -Wtautological-compare documentation.
17339 2017-09-04  Jeff Law  <law@redhat.com>
17341         PR tree-optimization/64910
17342         * tree-ssa-reassoc.c (reassociate_bb): For bitwise binary ops,
17343         swap the first and last operand if the last is a constant.
17345 2017-09-04  Marek Polacek  <polacek@redhat.com>
17347         PR sanitizer/82072
17348         * convert.c (do_narrow): When sanitizing signed integer overflows,
17349         bail out for signed types.
17350         (convert_to_integer_1) <case NEGATE_EXPR>: Likewise.
17352 2017-09-04  Richard Biener  <rguenther@suse.de>
17354         PR tree-optimization/82060
17355         * tree-ssa-pre.c (eliminate_dom_walker::before_dom_children):
17356         Move devirtualization after stmt folding and before EH/AB/noreturn
17357         cleanup to get the stmt refs canonicalized.  Use a bool instead
17358         of gimple_modified_p since that doesn't work for NOPs.  Schedule
17359         NOPs generated by folding for removal.
17361 2017-09-04  Richard Sandiford  <richard.sandiford@linaro.org>
17362             Alan Hayward  <alan.hayward@arm.com>
17363             David Sherwood  <david.sherwood@arm.com>
17365         * coretypes.h (pad_direction): New enum.
17366         * defaults.h (DEFAULT_FUNCTION_ARG_PADDING): Delete.
17367         (FUNCTION_ARG_PADDING): Likewise.
17368         * target.def (function_arg_padding): New hook.
17369         * targhooks.h (default_function_arg_padding): Declare.
17370         * targhooks.c (default_function_arg_padding): New function.
17371         * doc/tm.texi.in (FUNCTION_ARG_PADDING): Replace with...
17372         (TARGET_FUNCTION_ARG_PADDING): ...this.
17373         * doc/tm.texi: Regenerate.
17374         * calls.c (store_unaligned_arguments_into_pseudos): Use pad_direction
17375         instead of direction.
17376         (compute_argument_addresses): Likewise.
17377         (load_register_parameters): Likewise.
17378         (emit_library_call_value_1): Likewise.
17379         (store_one_arg): Use targetm.calls.function_arg_padding instead
17380         of FUNCTION_ARG_PADDING.
17381         (must_pass_in_stack_var_size_or_pad): Likewise.
17382         * expr.c (emit_group_load_1): Use pad_direction instead of direction.
17383         (emit_group_store): Likewise.
17384         (emit_single_push_insn_1): Use targetm.calls.function_arg_padding
17385         instead of FUNCTION_ARG_PADDING.
17386         (emit_push_insn): Likewise, and propagate enum change throughout
17387         function.
17388         * function.h (direction): Delete.
17389         (locate_and_pad_arg_data::where_pad): Use pad_direction instead
17390         of direction.
17391         * function.c (assign_parm_find_stack_rtl): Likewise.
17392         (assign_parm_setup_block_p): Likewise.
17393         (assign_parm_setup_block): Likewise.
17394         (gimplify_parameters): Likewise.
17395         (locate_and_pad_parm): Use targetm.calls.function_arg_padding
17396         instead of FUNCTION_ARG_PADDING, and propagate enum change throughout
17397         function.
17398         * config/aarch64/aarch64.h (FUNCTION_ARG_PADDING): Delete.
17399         (BLOCK_REG_PADDING): Use pad_direction instead of direction.
17400         * config/aarch64/aarch64-protos.h (aarch64_pad_arg_upward): Delete.
17401         * config/aarch64/aarch64.c (aarch64_pad_arg_upward): Replace with...
17402         (aarch64_function_arg_padding): ...this new function.
17403         (aarch64_gimplify_va_arg_expr): Use pad_direction instead of direction.
17404         (TARGET_FUNCTION_ARG_PADDING): Redefine.
17405         * config/arm/arm.h (FUNCTION_ARG_PADDING): Delete.
17406         (BLOCK_REG_PADDING): Use pad_direction instead of direction.
17407         * config/arm/arm-protos.h (arm_pad_arg_upward): Delete.
17408         * config/arm/arm.c (TARGET_FUNCTION_ARG_PADDING): Redefine.
17409         (arm_pad_arg_upward): Replace with...
17410         (arm_function_arg_padding): ...this new function.
17411         * config/c6x/c6x.h (BLOCK_REG_PADDING): Use pad_direction instead
17412         of direction.
17413         * config/ia64/hpux.h (FUNCTION_ARG_PADDING): Delete.
17414         * config/ia64/ia64-protos.h (ia64_hpux_function_arg_padding): Delete.
17415         * config/ia64/ia64.c (TARGET_FUNCTION_ARG_PADDING): Redefine.
17416         (ia64_hpux_function_arg_padding): Replace with...
17417         (ia64_function_arg_padding): ...this new function.  Use pad_direction
17418         instead of direction.  Check for TARGET_HPUX.
17419         * config/iq2000/iq2000.h (FUNCTION_ARG_PADDING): Delete.
17420         * config/iq2000/iq2000.c (TARGET_FUNCTION_ARG_PADDING): Redefine.
17421         (iq2000_function_arg_padding): New function.
17422         * config/mips/mips-protos.h (mips_pad_arg_upward): Delete.
17423         * config/mips/mips.c (mips_pad_arg_upward): Replace with...
17424         (mips_function_arg_padding): ...this new function.
17425         (mips_pad_reg_upward): Update accordingly.
17426         (TARGET_FUNCTION_ARG_PADDING): Redefine.
17427         * config/mips/mips.h (PAD_VARARGS_DOWN): Use
17428         targetm.calls.function_arg_padding.
17429         (FUNCTION_ARG_PADDING): Delete.
17430         (BLOCK_REG_PADDING): Use pad_direction instead of direction.
17431         * config/nios2/nios2.h (FUNCTION_ARG_PADDING): Delete.
17432         (PAD_VARARGS_DOWN): Use targetm.calls.function_arg_padding.
17433         * config/nios2/nios2-protos.h (nios2_function_arg_padding): Delete.
17434         (nios2_block_reg_padding): Return pad_direction instead of direction.
17435         * config/nios2/nios2.c (nios2_block_reg_padding): Return pad_direction
17436         instead of direction.
17437         (nios2_function_arg_padding): Likewise.  Make static.
17438         (TARGET_FUNCTION_ARG_PADDING): Redefine.
17439         * config/pa/pa.h (FUNCTION_ARG_PADDING): Delete.
17440         (BLOCK_REG_PADDING): Use targetm.calls.function_arg_padding.
17441         * config/pa/pa-protos.h (pa_function_arg_padding): Delete.
17442         * config/pa/pa.c (TARGET_FUNCTION_ARG_PADDING): Redefine.
17443         (pa_function_arg_padding): Make static.  Return pad_direction instead
17444         of direction.
17445         * config/powerpcspe/powerpcspe.h (FUNCTION_ARG_PADDING): Delete.
17446         (PAD_VARARGS_DOWN): Use targetm.calls.function_arg_padding.
17447         * config/powerpcspe/aix.h (BLOCK_REG_PADDING): Use pad_direction
17448         instead of direction.  Use targetm.calls.function_arg_padding.
17449         * config/powerpcspe/darwin.h (BLOCK_REG_PADDING): Likewise.
17450         * config/powerpcspe/freebsd64.h (BLOCK_REG_PADDING): Likewise.
17451         * config/powerpcspe/linux64.h (BLOCK_REG_PADDING): Likewise.
17452         * config/powerpcspe/powerpcspe-protos.h (function_arg_padding): Delete.
17453         * config/powerpcspe/powerpcspe.c (TARGET_FUNCTION_ARG_PADDING):
17454         Redefine.
17455         (function_arg_padding): Rename to...
17456         (rs6000_function_arg_padding): ...this.  Make static.  Return
17457         pad_direction instead of direction.
17458         (rs6000_return_in_msb): Use rs6000_function_arg_padding.
17459         * config/rs6000/rs6000.h (FUNCTION_ARG_PADDING): Delete.
17460         (PAD_VARARGS_DOWN): Use targetm.calls.function_arg_padding.
17461         * config/rs6000/aix.h (BLOCK_REG_PADDING): Use pad_direction
17462         instead of direction.  Use targetm.calls.function_arg_padding.
17463         * config/rs6000/darwin.h (BLOCK_REG_PADDING): Likewise.
17464         * config/rs6000/freebsd64.h (BLOCK_REG_PADDING): Likewise.
17465         * config/rs6000/linux64.h (BLOCK_REG_PADDING): Likewise.
17466         * config/rs6000/rs6000-protos.h (function_arg_padding): Delete.
17467         * config/rs6000/rs6000.c (TARGET_FUNCTION_ARG_PADDING): Redefine.
17468         (function_arg_padding): Rename to...
17469         (rs6000_function_arg_padding): ...this.  Make static.  Return
17470         pad_direction instead of direction.
17471         (rs6000_return_in_msb): Use rs6000_function_arg_padding.
17472         * config/s390/s390.h (FUNCTION_ARG_PADDING): Delete.
17473         * config/s390/s390.c (s390_function_arg_padding): New function.
17474         (TARGET_FUNCTION_ARG_PADDING): Redefine.
17475         * config/sparc/sparc.h (FUNCTION_ARG_PADDING): Delete.
17476         * config/sparc/sparc-protos.h (function_arg_padding): Delete.
17477         * config/sparc/sparc.c (TARGET_FUNCTION_ARG_PADDING): Redefine.
17478         (function_arg_padding): Rename to...
17479         (sparc_function_arg_padding): ...this.  Make static.  Return
17480         pad_direction instead of direction.
17481         * config/spu/spu.h (FUNCTION_ARG_PADDING): Delete.
17482         * config/spu/spu.c (spu_function_arg_padding): New function.
17483         (TARGET_FUNCTION_ARG_PADDING): Redefine.
17484         * system.h (FUNCTION_ARG_PADDING): Poison.
17486 2017-09-04  Richard Sandiford  <richard.sandiford@linaro.org>
17487             Alan Hayward  <alan.hayward@arm.com>
17488             David Sherwood  <david.sherwood@arm.com>
17490         * target.def (modes_tieable_p): New hook.
17491         * doc/tm.texi (MODES_TIEABLE_P): Replace with...
17492         (TARGET_MODES_TIEABLE_P): ...this.
17493         * doc/tm.texi.in: Regenerate.
17494         * hooks.h (hook_bool_mode_mode_true): Declare.
17495         * hooks.c (hook_bool_mode_mode_true): New function.
17496         * combine.c (subst): Use targetm.modes_tieable_p instead of
17497         MODES_TIEABLE_P.
17498         * dse.c (find_shift_sequence): Likewise.
17499         * expmed.c (extract_low_bits): Likewise.
17500         * lower-subreg.c: Include target.h.
17501         (find_decomposable_subregs): Use targetm.modes_tieable_p instead of
17502         MODES_TIEABLE_P.
17503         * rtlanal.c (rtx_cost): Likewise.
17504         * config/aarch64/aarch64.h (MODES_TIEABLE_P): Delete.
17505         * config/aarch64/aarch64-protos.h (aarch64_modes_tieable_p): Delete.
17506         * config/aarch64/aarch64.c (aarch64_modes_tieable_p): Make static.
17507         (TARGET_MODES_TIEABLE_P): Redefine.
17508         * config/alpha/alpha.h (MODES_TIEABLE_P): Delete.
17509         * config/alpha/alpha.c (alpha_modes_tieable_p): New function.
17510         (TARGET_MODES_TIEABLE_P): Redefine.
17511         * config/arc/arc.h (MODES_TIEABLE_P): Delete.
17512         * config/arc/arc.c (TARGET_MODES_TIEABLE_P): Redefine.
17513         (arc_modes_tieable_p): New function.
17514         * config/arm/arm.h (MODES_TIEABLE_P): Delete.
17515         * config/arm/arm-protos.h (arm_modes_tieable_p): Delete.
17516         * config/arm/arm.c (TARGET_MODES_TIEABLE_P): Redefine.
17517         (arm_modes_tieable_p): Make static.
17518         * config/avr/avr.h (MODES_TIEABLE_P): Delete.
17519         * config/bfin/bfin.h (MODES_TIEABLE_P): Delete.
17520         * config/bfin/bfin.c (bfin_modes_tieable_p): New function.
17521         (TARGET_MODES_TIEABLE_P): Redefine.
17522         * config/c6x/c6x.h (MODES_TIEABLE_P): Delete.
17523         * config/c6x/c6x.c (c6x_modes_tieable_p): New function.
17524         (TARGET_MODES_TIEABLE_P): Redefine.
17525         * config/cr16/cr16.h (MODES_TIEABLE_P): Delete.
17526         * config/cr16/cr16.c (TARGET_MODES_TIEABLE_P): Redefine.
17527         (cr16_modes_tieable_p): New function.
17528         * config/cris/cris.h (MODES_TIEABLE_P): Delete.
17529         * config/epiphany/epiphany.h (MODES_TIEABLE_P): Delete.
17530         * config/fr30/fr30.h (MODES_TIEABLE_P): Delete.
17531         (TRULY_NOOP_TRUNCATION): Update comment.
17532         * config/frv/frv.h (MODES_TIEABLE_P): Delete.
17533         (TRULY_NOOP_TRUNCATION): Update comment.
17534         * config/frv/frv.c (TARGET_MODES_TIEABLE_P): Redefine.
17535         (frv_modes_tieable_p): New function.
17536         * config/ft32/ft32.h (MODES_TIEABLE_P): Delete.
17537         * config/h8300/h8300.h (MODES_TIEABLE_P): Delete.
17538         * config/h8300/h8300.c (h8300_modes_tieable_p): New function.
17539         (TARGET_MODES_TIEABLE_P): Redefine.
17540         * config/i386/i386.h (MODES_TIEABLE_P): Delete.
17541         * config/i386/i386-protos.h (ix86_modes_tieable_p): Delete.
17542         * config/i386/i386.c (ix86_modes_tieable_p): Make static.
17543         (TARGET_MODES_TIEABLE_P): Redefine.
17544         * config/ia64/ia64.h (MODES_TIEABLE_P): Delete.
17545         * config/ia64/ia64.c (TARGET_MODES_TIEABLE_P): Redefine.
17546         (ia64_modes_tieable_p): New function.
17547         * config/iq2000/iq2000.h (MODES_TIEABLE_P): Delete.
17548         * config/iq2000/iq2000.c (TARGET_MODES_TIEABLE_P): Redefine.
17549         (iq2000_modes_tieable_p): New function.
17550         * config/lm32/lm32.h (MODES_TIEABLE_P): Delete.
17551         * config/lm32/lm32.c (TARGET_MODES_TIEABLE_P): Redefine.
17552         (lm32_modes_tieable_p): New function.
17553         * config/m32c/m32c.h (MODES_TIEABLE_P): Delete.
17554         * config/m32c/m32c-protos.h (m32c_modes_tieable_p): Delete.
17555         * config/m32c/m32c.c (m32c_modes_tieable_p): Make static.
17556         (TARGET_MODES_TIEABLE_P): Redefine.
17557         * config/m32r/m32r.h (MODES_TIEABLE_P): Delete.
17558         * config/m32r/m32r.c (TARGET_MODES_TIEABLE_P): Redefine.
17559         (m32r_modes_tieable_p): New function.
17560         * config/m68k/m68k.h (MODES_TIEABLE_P): Delete.
17561         * config/m68k/m68k.c (TARGET_MODES_TIEABLE_P): Redefine.
17562         (m68k_modes_tieable_p): New function.
17563         * config/mcore/mcore.h (MODES_TIEABLE_P): Delete.
17564         * config/mcore/mcore.c (TARGET_MODES_TIEABLE_P): Redefine.
17565         (mcore_modes_tieable_p): New function.
17566         * config/microblaze/microblaze.h (MODES_TIEABLE_P): Delete.
17567         * config/microblaze/microblaze.c (microblaze_modes_tieable_p): New
17568         function.
17569         (TARGET_MODES_TIEABLE_P): Redefine.
17570         * config/mips/mips.h (MODES_TIEABLE_P): Delete.
17571         * config/mips/mips-protos.h (mips_modes_tieable_p): Delete.
17572         * config/mips/mips.c (mips_modes_tieable_p): Make static.
17573         (TARGET_MODES_TIEABLE_P): Redefine.
17574         * config/mmix/mmix.h (MODES_TIEABLE_P): Delete.
17575         * config/mn10300/mn10300.h (MODES_TIEABLE_P): Delete.
17576         * config/mn10300/mn10300-protos.h (mn10300_modes_tieable): Delete.
17577         * config/mn10300/mn10300.c (mn10300_modes_tieable): Rename to...
17578         (mn10300_modes_tieable_p): ...this and make static.
17579         (TARGET_MODES_TIEABLE_P): Redefine.
17580         * config/moxie/moxie.h (MODES_TIEABLE_P): Delete.
17581         * config/msp430/msp430.h (MODES_TIEABLE_P): Delete.
17582         * config/msp430/msp430-protos.h (msp430_modes_tieable_p): Delete.
17583         * config/msp430/msp430.c (TARGET_MODES_TIEABLE_P): Redefine.
17584         (msp430_modes_tieable_p): Make static.
17585         * config/nds32/nds32.h (MODES_TIEABLE_P): Delete.
17586         * config/nds32/nds32.c (nds32_modes_tieable_p): New function.
17587         (TARGET_MODES_TIEABLE_P): Redefine.
17588         * config/nios2/nios2.h (MODES_TIEABLE_P): Delete.
17589         * config/nvptx/nvptx.h (MODES_TIEABLE_P): Delete.
17590         * config/nvptx/nvptx.c (nvptx_modes_tieable_p): New function.
17591         (TARGET_MODES_TIEABLE_P): Redefine.
17592         * config/pa/pa.h (MODES_TIEABLE_P): Delete.
17593         * config/pa/pa-protos.h (pa_modes_tieable_p): Delete.
17594         * config/pa/pa.c (pa_modes_tieable_p): Make static.
17595         (TARGET_MODES_TIEABLE_P): Redefine.
17596         * config/pdp11/pdp11.h (MODES_TIEABLE_P): Delete.
17597         * config/pdp11/pdp11.c (TARGET_MODES_TIEABLE_P): Redefine.
17598         (pdp11_modes_tieable_p): New function.
17599         * config/powerpcspe/powerpcspe.h (MODES_TIEABLE_P): Delete.
17600         * config/powerpcspe/powerpcspe.c (TARGET_MODES_TIEABLE_P): Redefine.
17601         (rs6000_modes_tieable_p): New function.
17602         (rs6000_debug_reg_global): Use it instead of MODES_TIEABLE_P.
17603         * config/powerpcspe/powerpcspe.md: Update comment.
17604         * config/riscv/riscv.h (MODES_TIEABLE_P): Delete.
17605         * config/riscv/riscv.c (riscv_modes_tieable_p): New function.
17606         (TARGET_MODES_TIEABLE_P): Redefine.
17607         * config/rl78/rl78.h (MODES_TIEABLE_P): Delete.
17608         * config/rl78/rl78.c (TARGET_MODES_TIEABLE_P): Redefine.
17609         (rl78_modes_tieable_p): New function.
17610         * config/rs6000/rs6000.h (MODES_TIEABLE_P): Delete.
17611         * config/rs6000/rs6000.c (TARGET_MODES_TIEABLE_P): Redefine.
17612         (rs6000_modes_tieable_p): New function.
17613         (rs6000_debug_reg_global): Use it instead of MODES_TIEABLE_P.
17614         * config/rs6000/rs6000.md: Update comment.
17615         * config/rx/rx.h (MODES_TIEABLE_P): Delete.
17616         * config/rx/rx.c (rx_modes_tieable_p): New function.
17617         (TARGET_MODES_TIEABLE_P): Redefine.
17618         * config/s390/s390.h (MODES_TIEABLE_P): Delete.
17619         * config/s390/s390.c (s390_modes_tieable_p): New function.
17620         (TARGET_MODES_TIEABLE_P): Redefine.
17621         * config/sh/sh.h (MODES_TIEABLE_P): Delete.
17622         * config/sh/sh.c (TARGET_MODES_TIEABLE_P): Redefine.
17623         (sh_modes_tieable_p): New function.
17624         * config/sparc/sparc.h (MODES_TIEABLE_P): Delete.
17625         * config/sparc/sparc-protos.h (sparc_modes_tieable_p): Delete.
17626         * config/sparc/sparc.c (TARGET_MODES_TIEABLE_P): Redefine.
17627         (sparc_modes_tieable_p): Make static.
17628         * config/spu/spu.h (MODES_TIEABLE_P): Delete.
17629         * config/spu/spu.c (spu_modes_tieable_p): New function.
17630         (TARGET_MODES_TIEABLE_P): Redefine.
17631         * config/stormy16/stormy16.h (MODES_TIEABLE_P): Delete.
17632         * config/stormy16/stormy16.c (xstormy16_modes_tieable_p): New function.
17633         (TARGET_MODES_TIEABLE_P): Redefine.
17634         * config/tilegx/tilegx.h (MODES_TIEABLE_P): Delete.
17635         * config/tilepro/tilepro.h (MODES_TIEABLE_P): Delete.
17636         * config/v850/v850.h (MODES_TIEABLE_P): Delete.
17637         * config/v850/v850.c (v850_modes_tieable_p): New function.
17638         (TARGET_MODES_TIEABLE_P): Redefine.
17639         * config/vax/vax.h (MODES_TIEABLE_P): Delete.
17640         * config/visium/visium.h (MODES_TIEABLE_P): Delete.
17641         * config/visium/visium.c (TARGET_MODES_TIEABLE_P): Redefine.
17642         (visium_modes_tieable_p): New function.
17643         * config/xtensa/xtensa.h (MODES_TIEABLE_P): Delete.
17644         * config/xtensa/xtensa.c (TARGET_MODES_TIEABLE_P): Redefine.
17645         (xtensa_modes_tieable_p): New function.
17646         * system.h (MODES_TIEABLE_P): Poison.
17648 2017-09-04  Richard Sandiford  <richard.sandiford@linaro.org>
17649             Alan Hayward  <alan.hayward@arm.com>
17650             David Sherwood  <david.sherwood@arm.com>
17652         * target.def (hard_regno_mode_ok): New hook.
17653         * doc/tm.texi (HARD_REGNO_MODE_OK): Replace with...
17654         (TARGET_HARD_REGNO_MODE_OK): ...this.
17655         * doc/tm.texi.in: Regenerate.
17656         * hooks.h (hook_bool_uint_mode_true): Declare.
17657         * hooks.c (hook_bool_uint_mode_true): New function.
17658         * doc/md.texi: Refer to targetm.hard_regno_mode_ok instead of
17659         HARD_REGNO_MODE_OK.
17660         * genpreds.c (write_insn_preds_c): Add an include of target.h.
17661         * alias.c (init_alias_target): Use targetm.hard_regno_mode_ok
17662         instead of HARD_REGNO_MODE_OK.
17663         * caller-save.c: Include target.h.
17664         (reg_save_code): Use targetm.hard_regno_mode_ok instead of
17665         HARD_REGNO_MODE_OK.
17666         * combine.c (can_combine_p): Likewise.
17667         (combinable_i3pat): Likewise.
17668         (can_change_dest_mode): Likewise.
17669         * expr.c (init_expr_target): Likewise.
17670         (convert_move): Likewise.
17671         (convert_modes): Likewise.
17672         * ira.c (setup_prohibited_class_mode_regs): Likewise.
17673         (setup_prohibited_mode_move_regs): Likewise.
17674         * ira.h (target_ira): Likewise.
17675         * lra-assigns.c (find_hard_regno_for_1): Likewise.
17676         * lra-constraints.c (process_alt_operands): Likewise.
17677         (split_reg): Likewise.
17678         * recog.c (peep2_find_free_register): Likewise.
17679         * ree.c (combine_reaching_defs): Likewise.
17680         * regcprop.c (maybe_mode_change): Likewise.
17681         * reginfo.c (init_reg_sets_1): Likewise.
17682         (choose_hard_reg_mode): Likewise.
17683         (simplifiable_subregs): Likewise.
17684         * regrename.c (check_new_reg_p): Likewise.
17685         * reload.c (find_valid_class): Likewise.
17686         (find_valid_class_1): Likewise.
17687         (reload_inner_reg_of_subreg): Likewise.
17688         (push_reload): Likewise.
17689         (combine_reloads): Likewise.
17690         (find_dummy_reload): Likewise.
17691         (find_reloads): Likewise.
17692         * reload1.c (find_reg): Likewise.
17693         (set_reload_reg): Likewise.
17694         (allocate_reload_reg): Likewise.
17695         (choose_reload_regs): Likewise.
17696         (reload_adjust_reg_for_temp): Likewise.
17697         * rtlanal.c (subreg_size_offset_from_lsb): Likewise.
17698         (simplify_subreg_regno): Likewise.
17699         * sel-sched.c (init_regs_for_mode): Likewise.
17700         * varasm.c (make_decl_rtl): Likewise.
17701         * config/aarch64/aarch64.h (HARD_REGNO_MODE_OK): Delete.
17702         (MODES_TIEABLE_P): Use targetm.hard_regno_mode_ok instead of
17703         HARD_REGNO_MODE_OK.
17704         * config/aarch64/aarch64-protos.h (aarch64_hard_regno_mode_ok): Delete.
17705         * config/aarch64/aarch64.c (aarch64_hard_regno_mode_ok): Make static.
17706         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17707         * config/alpha/alpha.h (HARD_REGNO_MODE_OK): Delete.
17708         * config/alpha/alpha.c (alpha_hard_regno_mode_ok): New function.
17709         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17710         * config/arc/arc.h (arc_hard_regno_mode_ok): Delete.
17711         (arc_mode_class): Delete.
17712         (HARD_REGNO_MODE_OK): Delete.
17713         * config/arc/arc.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17714         (arc_hard_regno_mode_ok): Rename old array to...
17715         (arc_hard_regno_mode_ok_modes): ...this.
17716         (arc_conditional_register_usage): Update accordingly.
17717         (arc_mode_class): Make static.
17718         (arc_hard_regno_mode_ok): New function.
17719         * config/arm/arm.h (HARD_REGNO_MODE_OK): Delete.
17720         * config/arm/arm-protos.h (arm_hard_regno_mode_ok): Delete.
17721         * config/arm/arm.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17722         (arm_hard_regno_mode_ok): Make static.
17723         * config/arm/arm.md (movdi): Use targetm.hard_regno_mode_ok instead of
17724         HARD_REGNO_MODE_OK.
17725         * config/avr/avr-protos.h (avr_hard_regno_mode_ok): Delete.
17726         * config/avr/avr.h (HARD_REGNO_MODE_OK): Delete.
17727         * config/avr/avr.c (avr_hard_regno_mode_ok): Make static and
17728         return a bool.
17729         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17730         * config/bfin/bfin-protos.h (hard_regno_mode_ok): Delete.
17731         * config/bfin/bfin.h (HARD_REGNO_MODE_OK): Delete.
17732         * config/bfin/bfin.c (hard_regno_mode_ok): Rename to...
17733         (bfin_hard_regno_mode_ok): ...this.  Make static and return a bool.
17734         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17735         * config/bfin/predicates.md (valid_reg_operand): Use
17736         targetm.hard_regno_mode_ok instead of HARD_REGNO_MODE_OK.
17737         * config/c6x/c6x.h (HARD_REGNO_MODE_OK): Delete.
17738         * config/c6x/c6x.c (c6x_hard_regno_mode_ok): New function.
17739         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17740         * config/cr16/cr16.h (HARD_REGNO_MODE_OK): Delete.
17741         * config/cr16/cr16-protos.h (cr16_hard_regno_mode_ok): Delete.
17742         * config/cr16/cr16.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17743         (cr16_hard_regno_mode_ok): Make static and return a bool.
17744         * config/cris/cris.h (HARD_REGNO_MODE_OK): Delete.
17745         * config/cris/cris.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17746         (cris_hard_regno_mode_ok): New function.
17747         * config/epiphany/epiphany.h (epiphany_hard_regno_mode_ok): Delete.
17748         (epiphany_mode_class): Delete.
17749         (HARD_REGNO_MODE_OK): Delete.
17750         * config/epiphany/epiphany-protos.h (hard_regno_mode_ok): Delete.
17751         * config/epiphany/epiphany.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17752         (hard_regno_mode_ok): Rename to...
17753         (epiphany_hard_regno_mode_ok): ...this.  Make static and return a bool.
17754         * config/fr30/fr30.h (HARD_REGNO_MODE_OK): Delete.
17755         * config/fr30/fr30.md: Refer to targetm.hard_regno_mode_ok instead of
17756         HARD_REGNO_MODE_OK.
17757         * config/frv/frv.h (HARD_REGNO_MODE_OK): Delete.
17758         * config/frv/frv-protos.h (frv_hard_regno_mode_ok): Delete.
17759         * config/frv/frv.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17760         (frv_hard_regno_mode_ok): Make static and return a bool.
17761         * config/frv/frv.md: Refer to targetm.hard_regno_mode_ok instead of
17762         HARD_REGNO_MODE_OK.
17763         * config/ft32/ft32.h (HARD_REGNO_MODE_OK): Delete.
17764         * config/h8300/h8300.h (HARD_REGNO_MODE_OK): Delete.
17765         * config/h8300/h8300-protos.h (h8300_hard_regno_mode_ok): Delete.
17766         * config/h8300/h8300.c (h8300_hard_regno_mode_ok): Make static
17767         and return a bool.
17768         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17769         * config/i386/i386.h (HARD_REGNO_MODE_OK): Delete.
17770         * config/i386/i386-protos.h (ix86_hard_regno_mode_ok): Delete.
17771         * config/i386/i386.c (ix86_hard_regno_mode_ok): Make static and
17772         return a bool.
17773         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17774         * config/ia64/ia64.h (HARD_REGNO_MODE_OK): Delete.
17775         * config/ia64/ia64.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17776         (ia64_hard_regno_mode_ok): New function.
17777         * config/iq2000/iq2000.h (HARD_REGNO_MODE_OK): Delete.
17778         * config/iq2000/iq2000.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17779         (iq2000_hard_regno_mode_ok): New function.
17780         * config/lm32/lm32.h (HARD_REGNO_MODE_OK): Delete.
17781         * config/lm32/lm32.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17782         (lm32_hard_regno_mode_ok): New function.
17783         * config/m32c/m32c.h (HARD_REGNO_MODE_OK): Delete.
17784         * config/m32c/m32c-protos.h (m32c_hard_regno_ok): Delete.
17785         * config/m32c/m32c.c (class_can_hold_mode): Use m32c_hard_regno_mode_ok
17786         instead of HARD_REGNO_MODE_OK.
17787         (m32c_hard_regno_ok): Rename to...
17788         (m32c_hard_regno_mode_ok): ...this.  Make static and return a bool.
17789         (m32c_cannot_change_mode_class): Update accordingly.
17790         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17791         * config/m32r/m32r.h (m32r_hard_regno_mode_ok): Delete.
17792         (m32r_mode_class): Delete.
17793         (HARD_REGNO_MODE_OK): Delete.
17794         * config/m32r/m32r.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17795         (m32r_hard_regno_mode_ok): Rename to...
17796         (m32r_hard_regno_modes): ...this.
17797         (m32r_mode_class): Make static.
17798         (m32r_hard_regno_mode_ok): New function.
17799         * config/m68k/m68k.h (HARD_REGNO_MODE_OK): Delete.
17800         * config/m68k/m68k-protos.h (m68k_regno_mode_ok): Delete.
17801         * config/m68k/m68k.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17802         (m68k_hard_regno_mode_ok): Make static.
17803         * config/mcore/mcore.h (HARD_REGNO_MODE_OK): Delete.
17804         * config/mcore/mcore.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17805         (mcore_hard_regno_mode_ok): New function.
17806         * config/microblaze/microblaze.h (microblaze_hard_regno_mode_ok)
17807         (HARD_REGNO_MODE_OK): Delete.
17808         * config/microblaze/microblaze.c (microblaze_hard_regno_mode_ok):
17809         Rename to...
17810         (microblaze_hard_regno_mode_ok_p): ...this and make static.
17811         (microblaze_hard_regno_mode_ok): New function.
17812         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17813         * config/mips/mips.h (HARD_REGNO_MODE_OK): Delete.
17814         (mips_hard_regno_mode_ok): Delete.
17815         * config/mips/mips.c (mips_hard_regno_mode_ok): Rename to...
17816         (mips_hard_regno_mode_ok_p): ...this and make static.
17817         (mips_hard_regno_mode_ok_p): Rename to...
17818         (mips_hard_regno_mode_ok_uncached): ...this.
17819         (mips_hard_regno_mode_ok): New function.
17820         (mips_class_max_nregs): Use mips_hard_regno_mode_ok instead
17821         of HARD_REGNO_MODE_OK.
17822         (mips_option_override): Update after above name changes.
17823         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17824         * config/mmix/mmix.h (HARD_REGNO_MODE_OK): Delete.
17825         * config/mn10300/mn10300.h (HARD_REGNO_MODE_OK): Delete.
17826         * config/mn10300/mn10300-protos.h (mn10300_hard_regno_mode_ok): Delete.
17827         * config/mn10300/mn10300.c (mn10300_hard_regno_mode_ok): Make static.
17828         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17829         * config/moxie/moxie.h (HARD_REGNO_MODE_OK): Delete.
17830         * config/msp430/msp430.h (HARD_REGNO_MODE_OK): Delete.
17831         * config/msp430/msp430-protos.h (msp430_hard_regno_mode_ok): Delete.
17832         * config/msp430/msp430.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17833         (msp430_hard_regno_mode_ok): Make static and return a bool.
17834         * config/nds32/nds32.h (HARD_REGNO_MODE_OK): Delete.
17835         * config/nds32/nds32-protos.h (nds32_hard_regno_mode_ok): Delete.
17836         * config/nds32/nds32.c (nds32_hard_regno_mode_ok): Make static
17837         and return a bool.
17838         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17839         * config/nios2/nios2.h (HARD_REGNO_MODE_OK): Delete.
17840         * config/nvptx/nvptx.h (HARD_REGNO_MODE_OK): Delete.
17841         * config/pa/pa.h (MODES_TIEABLE_P): Update commentary.
17842         * config/pa/pa32-regs.h (HARD_REGNO_MODE_OK): Rename to...
17843         (PA_HARD_REGNO_MODE_OK): ...this
17844         * config/pa/pa64-regs.h (HARD_REGNO_MODE_OK): Rename to...
17845         (PA_HARD_REGNO_MODE_OK): ...this.
17846         * config/pa/pa.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17847         (pa_hard_regno_mode_ok): New function.
17848         * config/pdp11/pdp11.h (HARD_REGNO_MODE_OK): Delete.
17849         * config/pdp11/pdp11.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17850         (pdp11_hard_regno_mode_ok): New function.
17851         * config/powerpcspe/powerpcspe.h (HARD_REGNO_MODE_OK): Delete.
17852         * config/powerpcspe/powerpcspe-protos.h (rs6000_hard_regno_mode_ok_p):
17853         Delete.
17854         * config/powerpcspe/powerpcspe.c (rs6000_hard_regno_mode_ok_p):
17855         Make static.
17856         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17857         (rs6000_hard_regno_mode_ok): Rename to...
17858         (rs6000_hard_regno_mode_ok_uncached): ...this.
17859         (rs6000_init_hard_regno_mode_ok): Update accordingly.
17860         (rs6000_hard_regno_mode_ok): New function.
17861         * config/riscv/riscv.h (HARD_REGNO_MODE_OK): Delete.
17862         * config/riscv/riscv-protos.h (riscv_hard_regno_mode_ok_p): Delete.
17863         * config/riscv/riscv.c (riscv_hard_regno_mode_ok_p): Rename to...
17864         (riscv_hard_regno_mode_ok): ...this and make static.
17865         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17866         * config/rl78/rl78.h (HARD_REGNO_MODE_OK): Delete.
17867         * config/rl78/rl78-protos.h (rl78_hard_regno_mode_ok): Delete.
17868         * config/rl78/rl78.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17869         (rl78_hard_regno_mode_ok): Make static and return bool.
17870         * config/rs6000/rs6000.h (HARD_REGNO_MODE_OK): Delete.
17871         * config/rs6000/rs6000-protos.h (rs6000_hard_regno_mode_ok_p):
17872         Delete.
17873         * config/rs6000/rs6000.c (rs6000_hard_regno_mode_ok_p): Make static.
17874         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17875         (rs6000_hard_regno_mode_ok): Rename to...
17876         (rs6000_hard_regno_mode_ok_uncached): ...this.
17877         (rs6000_init_hard_regno_mode_ok): Update accordingly.
17878         (rs6000_hard_regno_mode_ok): New function.
17879         * config/rx/rx.h (HARD_REGNO_MODE_OK): Delete.
17880         * config/rx/rx.c (rx_hard_regno_mode_ok): New function.
17881         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17882         * config/s390/s390.h (HARD_REGNO_MODE_OK): Delete.
17883         * config/s390/s390-protos.h (s390_hard_regno_mode_ok): Delete.
17884         * config/s390/s390.c (s390_hard_regno_mode_ok): Make static.
17885         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17886         * config/sh/sh.h (HARD_REGNO_MODE_OK): Delete.
17887         * config/sh/sh-protos.h (sh_hard_regno_mode_ok): Delete.
17888         * config/sh/sh.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17889         (sh_hard_regno_mode_ok): Make static.
17890         * config/sparc/constraints.md: Refer to targetm.hard_regno_mode_ok
17891         instead of HARD_REGNO_MODE_OK.
17892         * config/sparc/sparc.h (hard_regno_mode_classes): Delete.
17893         (sparc_mode_class): Delete.
17894         (HARD_REGNO_MODE_OK): Delete.
17895         * config/sparc/sparc.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17896         (hard_regno_mode_classes): Make static.
17897         (sparc_mode_class): Likewise.
17898         (sparc_hard_regno_mode_ok): New function.
17899         * config/spu/spu.h (HARD_REGNO_MODE_OK): Delete.
17900         * config/stormy16/stormy16.h (HARD_REGNO_MODE_OK): Delete.
17901         * config/stormy16/stormy16.c (xstormy16_hard_regno_mode_ok): New
17902         function.
17903         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17904         * config/tilegx/tilegx.h (HARD_REGNO_MODE_OK): Delete.
17905         * config/tilepro/tilepro.h (HARD_REGNO_MODE_OK): Delete.
17906         * config/v850/v850.h (HARD_REGNO_MODE_OK): Delete.
17907         * config/v850/v850.c (v850_hard_regno_mode_ok): New function.
17908         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17909         * config/vax/vax.h (HARD_REGNO_MODE_OK): Delete.
17910         * config/visium/visium.h (HARD_REGNO_MODE_OK): Delete.
17911         * config/visium/visium.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17912         (visium_hard_regno_mode_ok): New function.
17913         * config/visium/visium.md: Refer to targetm.hard_regno_mode_ok
17914         instead of HARD_REGNO_MODE_OK.
17915         * config/xtensa/xtensa.h (xtensa_hard_regno_mode_ok): Delete.
17916         (HARD_REGNO_MODE_OK): Delete.
17917         * config/xtensa/xtensa.c (xtensa_hard_regno_mode_ok): Rename to...
17918         (xtensa_hard_regno_mode_ok_p): ...this and make static.
17919         (xtensa_option_override): Update accordingly.
17920         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17921         (xtensa_hard_regno_mode_ok): New function.
17922         * system.h (HARD_REGNO_MODE_OK): Poison.
17924 2017-09-04  Richard Sandiford  <richard.sandiford@linaro.org>
17925             Alan Hayward  <alan.hayward@arm.com>
17926             David Sherwood  <david.sherwood@arm.com>
17928         * target.def (hard_regno_call_part_clobbered): New hook.
17929         * doc/tm.texi.in (HARD_REGNO_CALL_PART_CLOBBERED): Replace with...
17930         (TARGET_HARD_REGNO_CALL_PART_CLOBBERED): ...this hook.
17931         * doc/tm.texi: Regenerate.
17932         * hooks.h (hook_bool_uint_mode_false): Declare.
17933         * hooks.c (hook_bool_uint_mode_false): New function.
17934         * regs.h (HARD_REGNO_CALL_PART_CLOBBERED): Delete.
17935         * cselib.c (cselib_process_insn): Use
17936         targetm.hard_regno_call_part_clobbered instead of
17937         HARD_REGNO_CALL_PART_CLOBBERED.
17938         * ira-conflicts.c (ira_build_conflicts): Likewise.
17939         * ira-costs.c (ira_tune_allocno_costs): Likewise.
17940         * lra-constraints.c (need_for_call_save_p): Likewise.
17941         * lra-lives.c: Include target.h.
17942         (check_pseudos_live_through_calls): Use
17943         targetm.hard_regno_call_part_clobbered instead of
17944         HARD_REGNO_CALL_PART_CLOBBERED.
17945         * regcprop.c: Include target.h.
17946         (copyprop_hardreg_forward_1): Use
17947         targetm.hard_regno_call_part_clobbered instead of
17948         HARD_REGNO_CALL_PART_CLOBBERED.
17949         * reginfo.c (choose_hard_reg_mode): Likewise.
17950         * regrename.c (check_new_reg_p): Likewise.
17951         * reload.c (find_equiv_reg): Likewise.
17952         * reload1.c (emit_reload_insns): Likewise.
17953         * sched-deps.c (deps_analyze_insn): Likewise.
17954         * sel-sched.c (init_regs_for_mode): Likewise.
17955         (mark_unavailable_hard_regs): Likewise.
17956         * targhooks.c (default_dwarf_frame_reg_mode): Likewise.
17957         * config/aarch64/aarch64.h (HARD_REGNO_CALL_PART_CLOBBERED): Delete.
17958         * config/aarch64/aarch64.c (aarch64_hard_regno_call_part_clobbered):
17959         New function.
17960         (TARGET_HARD_REGNO_CALL_PART_CLOBBERED): Redefine.
17961         * config/avr/avr.h (HARD_REGNO_CALL_PART_CLOBBERED): Delete.
17962         * config/avr/avr-protos.h (avr_hard_regno_call_part_clobbered):
17963         Delete.
17964         * config/avr/avr.c (avr_hard_regno_call_part_clobbered): Make static
17965         and return a bool.
17966         (TARGET_HARD_REGNO_CALL_PART_CLOBBERED): Redefine.
17967         * config/i386/i386.h (HARD_REGNO_CALL_PART_CLOBBERED): Delete.
17968         * config/i386/i386.c (ix86_hard_regno_call_part_clobbered): New
17969         function.
17970         (TARGET_HARD_REGNO_CALL_PART_CLOBBERED): Redefine.
17971         * config/mips/mips.h (HARD_REGNO_CALL_PART_CLOBBERED): Delete.
17972         * config/mips/mips.c (mips_hard_regno_call_part_clobbered): New
17973         function.
17974         (TARGET_HARD_REGNO_CALL_PART_CLOBBERED): Redefine.
17975         * config/powerpcspe/powerpcspe.h (HARD_REGNO_CALL_PART_CLOBBERED):
17976         Delete.
17977         * config/powerpcspe/powerpcspe.c
17978         (rs6000_hard_regno_call_part_clobbered): New function.
17979         (TARGET_HARD_REGNO_CALL_PART_CLOBBERED): Redefine.
17980         * config/rs6000/rs6000.h (HARD_REGNO_CALL_PART_CLOBBERED): Delete.
17981         * config/rs6000/rs6000.c (rs6000_hard_regno_call_part_clobbered):
17982         New function.
17983         (TARGET_HARD_REGNO_CALL_PART_CLOBBERED): Redefine.
17984         * config/s390/s390.h (HARD_REGNO_CALL_PART_CLOBBERED): Delete.
17985         * config/s390/s390.c (s390_hard_regno_call_part_clobbered): New
17986         function.
17987         (TARGET_HARD_REGNO_CALL_PART_CLOBBERED): Redefine.
17988         * config/sh/sh.h (HARD_REGNO_CALL_PART_CLOBBERED): Delete.
17989         * system.h (HARD_REGNO_CALL_PART_CLOBBERED): Poison.
17991 2017-09-04  Richard Sandiford  <richard.sandiford@linaro.org>
17992             Alan Hayward  <alan.hayward@arm.com>
17993             David Sherwood  <david.sherwood@arm.com>
17995         * rtl.h (subreg_memory_offset): Declare.
17996         * emit-rtl.c (subreg_memory_offset): New function.
17997         * expmed.c (store_bit_field_1): Use it.
17998         * expr.c (undefined_operand_subword_p): Likewise.
17999         * simplify-rtx.c (simplify_subreg): Likewise.
18001 2017-09-04  Alexander Monakov  <amonakov@ispras.ru>
18003         PR rtl-optimization/57448
18004         PR target/67458
18005         PR target/81316
18006         * optabs.c (expand_atomic_load): Place compiler memory barriers if
18007         using atomic_load pattern.
18008         (expand_atomic_store): Likewise.
18010 2017-09-04  Jakub Jelinek  <jakub@redhat.com>
18012         PR sanitizer/81981
18013         * gimple-fold.c (gimple_fold_call): Optimize away useless UBSAN_PTR
18014         and UBSAN_BOUNDS internal calls.  Clean up IFN_UBSAN_OBJECT_SIZE
18015         handling.  Use replace_call_with_value with NULL instead of
18016         gsi_replace, unlink_stmt_vdef and release_defs.
18018         * gdbhooks.py (OptMachineModePrinter.to_string): Use 8 spaces
18019         instead of tab.
18021         * lra-remat.c (reg_overlap_for_remat_p): Fix a pasto.
18023 2017-09-04  Richard Sandiford  <richard.sandiford@linaro.org>
18025         PR bootstrap/82045
18026         * rtl.h (emit_library_call_value_1): Declare.
18027         (emit_library_call): Replace declaration with a series of overloads.
18028         Remove the parameter count argument.
18029         (emit_library_call_value): Likewise.
18030         * calls.c (emit_library_call_value_1): Make global.  Replace varargs
18031         with an "rtx_mode_t *".
18032         (emit_library_call_value): Delete.
18033         (emit_library_call): Likewise.
18034         * asan.c (asan_emit_stack_protection): Update calls accordingly.
18035         (asan_emit_allocas_unpoison): Likewise.
18036         * builtins.c (expand_builtin_powi): Likewise.
18037         (expand_asan_emit_allocas_unpoison): Likewise.
18038         * cfgexpand.c (expand_main_function): Likewise.
18039         * config/aarch64/aarch64.c (aarch64_trampoline_init): Likewise.
18040         * config/aarch64/aarch64.h (PROFILE_HOOK): Likewise.
18041         * config/alpha/alpha.c (alpha_trampoline_init): Likewise.
18042         * config/arm/arm.c (arm_trampoline_init): Likewise.
18043         (arm_call_tls_get_addr): Likewise.
18044         (arm_expand_divmod_libfunc): Likewise.
18045         * config/bfin/bfin.md (umulsi3_highpart): Likewise.
18046         (smulsi3_highpart): Likewise.
18047         * config/c6x/c6x.c (c6x_initialize_trampoline): Likewise.
18048         (c6x_expand_compare): Likewise.
18049         (c6x_expand_movmem): Likewise.
18050         * config/frv/frv.c (frv_trampoline_init): Likewise.
18051         * config/i386/i386.c (ix86_trampoline_init): Likewise.
18052         (ix86_expand_divmod_libfunc): Likewise.
18053         * config/ia64/ia64.c (ia64_expand_tls_address): Likewise.
18054         (ia64_expand_compare): Likewise.
18055         (ia64_profile_hook): Likewise.
18056         * config/ia64/ia64.md (save_stack_nonlocal): Likewise.
18057         (nonlocal_goto): Likewise.
18058         (restore_stack_nonlocal): Likewise.
18059         * config/m32r/m32r.c (block_move_call): Likewise.
18060         (m32r_trampoline_init): Likewise.
18061         * config/m68k/linux.h (FINALIZE_TRAMPOLINE): Likewise.
18062         * config/m68k/m68k.c (m68k_call_tls_get_addr): Likewise.
18063         (m68k_call_m68k_read_tp): Likewise.
18064         * config/microblaze/microblaze.c (microblaze_call_tls_get_addr)
18065         (microblaze_expand_divide): Likewise.
18066         * config/mips/mips.h (mips_args): Likewise.
18067         * config/mips/sdemtk.h (mips_sync_icache): Likewise.
18068         (MIPS_ICACHE_SYNC): Likewise.
18069         * config/nios2/nios2.c (nios2_emit_expensive_div): Likewise.
18070         (nios2_trampoline_init): Likewise.
18071         * config/pa/pa.c (hppa_tls_call): Likewise.
18072         (pa_trampoline_init): Likewise.
18073         * config/pa/pa.md (canonicalize_funcptr_for_compare): Likewise.
18074         * config/powerpcspe/powerpcspe.c (rs6000_legitimize_tls_address)
18075         (expand_strn_compare): Likewise.
18076         (rs6000_generate_compare): Likewise.
18077         (rs6000_expand_float128_convert): Likewise.
18078         (output_profile_hook): Likewise.
18079         (rs6000_trampoline_init): Likewise.
18080         * config/powerpcspe/powerpcspe.md (neg<mode>2): Likewise.
18081         * config/riscv/riscv.h (PROFILE_HOOK): Likewise.
18082         * config/rs6000/rs6000-string.c (expand_strn_compare): Likewise.
18083         * config/rs6000/rs6000.c (rs6000_legitimize_tls_address): Likewise.
18084         (rs6000_generate_compare): Likewise.
18085         (rs6000_expand_float128_convert): Likewise.
18086         (output_profile_hook): Likewise.
18087         (rs6000_trampoline_init): Likewise.
18088         * config/rs6000/rs6000.md (neg<mode>2): Likewise.
18089         * config/sh/sh.c (sh_trampoline_init): Likewise.
18090         * config/sparc/sparc.c (emit_soft_tfmode_libcall): Likewise.
18091         (sparc_emit_float_lib_cmp): Likewise.
18092         (sparc32_initialize_trampoline): Likewise.
18093         (sparc64_initialize_trampoline): Likewise.
18094         (sparc_profile_hook): Likewise.
18095         * config/spu/spu.c (ea_load_store): Likewise.
18096         * config/spu/spu.md (floatunssidf2): Likewise.
18097         * config/tilegx/tilegx.c (tilegx_trampoline_init): Likewise.
18098         * config/tilepro/tilepro.c (tilepro_trampoline_init): Likewise.
18099         * config/visium/visium.c (expand_block_move_4): Likewise.
18100         (expand_block_move_2): Likewise.
18101         (expand_block_move_1): Likewise.
18102         (expand_block_set_4): Likewise.
18103         (expand_block_set_2): Likewise.
18104         (expand_block_set_1): Likewise.
18105         (visium_trampoline_init): Likewise.
18106         (visium_profile_hook): Likewise.
18107         * config/xtensa/xtensa.c (xtensa_expand_nonlocal_goto): Likewise.
18108         (xtensa_setup_frame_addresses): Likewise.
18109         (xtensa_trampoline_init): Likewise.
18110         * except.c (sjlj_emit_function_enter): Likewise.
18111         (sjlj_emit_function_exit): Likewise.
18112         * explow.c (allocate_dynamic_stack_space): Likewise.
18113         (probe_stack_range): Likewise.
18114         * expr.c (convert_mode_scalar): Likewise.
18115         * optabs.c (expand_binop): Likewise.
18116         (expand_twoval_binop_libfunc): Likewise.
18117         (expand_unop): Likewise.
18118         (prepare_cmp_insn): Likewise.
18119         (prepare_float_lib_cmp): Likewise.
18120         (expand_float): Likewise.
18121         (expand_fix): Likewise.
18122         (expand_fixed_convert): Likewise.
18123         (maybe_emit_sync_lock_test_and_set): Likewise.
18124         (expand_atomic_compare_and_swap): Likewise.
18125         (expand_mem_thread_fence): Likewise.
18126         (expand_atomic_fetch_op): Likewise.
18128 2017-09-03  Gerald Pfeifer  <gerald@pfeifer.com>
18130         * doc/generic.texi (OpenACC): Adjust URL.
18131         * doc/invoke.texi (C Dialect Options): Ditto.
18133 2017-09-03  Uros Bizjak  <ubizjak@gmail.com>
18135         * config/i386/i386.md (*bt<mode>): Use nonimmediate_operand
18136         predicate for operand 1.  Add (m,<S>) constraint.
18137         (*jcc_bt<mode>): Use nonimmediate_operand predicate for operand 1.
18138         Prevent memory operand 1 with register operand 2.
18140 2017-09-01  Segher Boessenkool  <segher@kernel.crashing.org>
18142         PR rtl-optimization/82024
18143         * combine.c (try_combine): If the combination result is a PARALLEL,
18144         and we only need to retain the SET in there that would be placed
18145         at I2, check that we can place that at I3 instead, before doing so.
18147 2017-09-01  Jakub Jelinek  <jakub@redhat.com>
18149         PR target/81766
18150         * config/i386/i386.c (ix86_init_large_pic_reg): Return label
18151         instead of void.
18152         (ix86_init_pic_reg): Remember label from ix86_init_large_pic_reg,
18153         if non-NULL and preceded by NOTE_INSN_BASIC_BLOCK, swap the note
18154         and label.
18156 2017-09-01  Joerg Sonnenberger  <joerg@bec.de>
18157             Jeff Law  <law@redhat.com>
18159         * varasm.c (bss_initializer_p): Do not put constants into .bss
18160         (categorize_decl_for_section): Handle bss_initializer_p returning
18161         false when DECL_INITIAL is NULL.
18163 2017-09-01  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
18165         PR target/82012
18166         * config/s390/s390.c (s390_can_inline_p): New function.
18168 2017-09-01  Jeff Law  <law@redhat.com>
18170         PR tree-optimization/82052
18171         * tree-ssa-scopedtables.c (avail_exprs_stack::lookup_avail_expr):
18172         Always initialize the returned slot after a hash table miss
18173         when INSERT is true.
18175 2017-09-01  Alexander Monakov  <amonakov@ispras.ru>
18177         * config/s390/s390.md (mem_signal_fence): Remove.
18178         * doc/md.texi (mem_signal_fence): Remove.
18179         * optabs.c (expand_mem_signal_fence): Remove uses of mem_signal_fence.
18180         Update comments.
18181         * target-insns.def (mem_signal_fence): Remove.
18183 2017-09-01  Jakub Jelinek  <jakub@redhat.com>
18185         PR sanitizer/81902
18186         * doc/invoke.texi: Document -fsanitize=pointer-overflow.
18188         PR sanitizer/81923
18189         * asan.c (create_odr_indicator): Strip name encoding from assembler
18190         name before appending it after __odr_asan_.
18192 2017-09-01  Martin Liska  <mliska@suse.cz>
18194         PR tree-optimization/82059
18195         * gimple-ssa-isolate-paths.c (isolate_path): Add profile and
18196         frequency only when an edge is redirected.
18198 2017-09-01  Claudiu Zissulescu  <claziss@synopsys.com>
18200         * config/arc/arc-c.c (__ARC_LPC_WIDTH__): Add builtin define.
18201         * config/arc/arc.c (ARC_MAX_LOOP_LENGTH): Define.
18202         (arc_conditional_register_usage): Remove ARC600 lp_count
18203         exception.
18204         (arc_file_start): Emit Tag_ARC_CPU_variation.
18205         (arc_can_use_doloop_p): New conditions to use ZOLs.
18206         (hwloop_fail): New function.
18207         (hwloop_optimize): Likewise.
18208         (hwloop_pattern_reg): Likewise.
18209         (arc_doloop_hooks): New struct, to be used with reorg_loops.
18210         (arc_reorg_loops): New function, calls reorg_loops.
18211         (arc_reorg): Call arc_reorg_loops.  Remove old ZOL handling.
18212         (arc600_corereg_hazard): Remove ZOL checking, case handled by
18213         hwloop_optimize.
18214         (arc_loop_hazard): Remove function, functionality moved into
18215         hwloop_optimize.
18216         (arc_hazard): Remove arc_loop_hazard call.
18217         (arc_adjust_insn_length): Remove ZOL handling, functionality moved
18218         into hwloop_optimize.
18219         (arc_label_align): Remove ZOL handling.
18220         * config/arc/arc.h (LOOP_ALIGN): Changed to 0.
18221         * config/arc/arc.md (doloop_begin): Remove pattern.
18222         (doloop_begin_i): Likewise.
18223         (doloop_end_i): Likewise.
18224         (doloop_fallback): Likewise.
18225         (doloop_fallback_m): Likewise.
18226         (doloop_end): Reimplement expand.
18227         (arc_lp): New pattern for LP instruction.
18228         (loop_end): New pattern.
18229         (loop_fail): Likewise.
18230         (decrement_and_branch_until_zero): Likewise.
18231         * config/arc/arc.opt (mlpc-width): New option.
18232         * doc/invoke.texi (mlpc-width): Document option.
18234 2017-09-01  Claudiu Zissulescu  <claziss@synopsys.com>
18236         * config/arc/arc.c (arc_ifcvt): Remove use of merge_blocks call.
18237         (arc_ccfsm_advance): Fix checking for delay slots.
18238         (arc_reorg): Add rtl dump after each call to arc_ifcvt.
18240 2017-09-01  Claudiu Zissulescu  <claziss@synopsys.com>
18242         * config/arc/arc.md (movqi_insn): Add stores to save constant long
18243         immediates.
18244         (movhi_insn): Update store instruction constraint which are saving
18245         6-bit short immediates.
18246         (movsi_insn): Consider also short scaled load operations.
18247         (zero_extendhisi2_i): Use Usd constraint instead of T.
18248         (extendhisi2_i): Add q constraint.
18249         (arc_clzsi2): Add type and length attributes.
18250         (arc_ctzsi2): Likewise.
18251         * config/arc/constraints.md (Usc): Update constraint, the
18252         assembler can parse two relocations for a single instruction.
18254 2017-09-01  Claudiu Zissulescu  <claziss@synopsys.com>
18256         * config/arc/arc.c (arc_use_anchors_for_symbol_p): New function.
18257         (TARGET_USE_ANCHORS_FOR_SYMBOL_P): Define.
18259 2017-08-31  Olivier Hainque  <hainque@adacore.com>
18261         * config.gcc (powerpc-wrs-vxworks|vxworksae|vxworksmils): Now
18262         match as powerpc-wrs-vxworks*.
18264 2017-08-31  James Greenhalgh  <james.greenhalgh@arm.com>
18266         * config/aarch64/aarch64-simd.md (aarch64_mla_elt_merge<mode>): Fix
18267         register constraint for by-element operand.
18268         (aarch64_mls_elt_merge<mode>): Likewise.
18270 2017-08-31  Claudiu Zissulescu  <claziss@synopsys.com>
18272         * config/arc/arc.c (arc_can_follow_jump): Check for short
18273         branches.
18275 2017-08-31  Claudiu Zissulescu  <claziss@synopsys.com>
18277         * config.gcc: Use g.opt for arc.
18278         * config/arc/arc.c (LEGITIMATE_SCALED_ADDRESS_P): Deleted,
18279         functionality moved to ...
18280         (legitimate_scaled_address_p): New function, ...here.
18281         (LEGITIMATE_SMALL_DATA_OFFSET_P): New define.
18282         (LEGITIMATE_SMALL_DATA_ADDRESS_P): Use the above define.
18283         (legitimate_offset_address_p): Delete TARGET_NO_SDATA_SET
18284         condition.
18285         (arc_override_options): Handle G option.
18286         (arc_output_pic_addr_const): Correct function definition.
18287         (arc_legitimate_address_p): Use legitimate_scaled_address_p.
18288         (arc_decl_anon_ns_mem_p): Delete.
18289         (arc_in_small_data_p): Overhaul this function to take into
18290         consideration the value given via G option.
18291         (arc_rewrite_small_data_1): Renamed and corrected old
18292         arc_rewrite_small_data function.
18293         (arc_rewrite_small_data): New function.
18294         (small_data_pattern): Don't use pic_offset_table_rtx.
18295         * config/arc/arc.h (CC1_SPEC): Recognize G option.
18296         * config/arc/simdext.md (movmisalignv2hi): Use
18297         prepare_move_operands function.
18298         (mov*): Likewise.
18299         (movmisalign*): Likewise.
18300         * doc/invoke.texi (ARC options): Document -G option.
18302 2017-08-31  Claudiu Zissulescu  <claziss@synopsys.com>
18304         * config/arc/arc-protos.h (compact_sda_memory_operand): Update
18305         prototype.
18306         * config/arc/arc.c (arc_print_operand): Output scalled address for
18307         sdata whenever is possible.
18308         (arc_in_small_data_p): Allow sdata for 64bit datum when double
18309         load/stores are available.
18310         (compact_sda_memory_operand): Check for the alignment required by
18311         code density instructions.
18312         * config/arc/arc.md (movsi_insn): Use newly introduced Us0
18313         constraint.
18314         * config/arc/constraints.md (Usd): Update constraint.
18315         (Us0): New constraint.
18316         (Usc): Update constraint.
18318 2017-08-31  Richard Biener  <rguenther@suse.de>
18320         PR middle-end/82054
18321         * dwarf2out.c (dwarf2out_early_global_decl): Process each
18322         function only once.
18324 2017-08-31  Tamar Christina  <tamar.christina@arm.com>
18326         * config/aarch64/aarch64-builtins.c (aarch64_init_simd_builtins):
18327         Resize type_signature.
18329 2017-08-31  Richard Sandiford  <richard.sandiford@linaro.org>
18330             Alan Hayward  <alan.hayward@arm.com>
18331             David Sherwood  <david.sherwood@arm.com>
18333         * config/aarch64/aarch64.c (aarch64_base_register_rtx_p): Only allow
18334         subregs whose inner modes can be stored in GPRs.
18335         (aarch64_classify_index): Likewise.
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/iterators.md (V_cmp_result): Rename to...
18342         (V_INT_EQUIV): ...this.
18343         (v_cmp_result): Rename to...
18344         (v_int_equiv): ...this.
18345         * config/aarch64/aarch64.md (xorsign<mode>3): Update accordingly.
18346         * config/aarch64/aarch64-simd.md (xorsign<mode>3): Likewise.
18347         (copysign<mode>3): Likewise.
18348         (aarch64_simd_bsl<mode>_internal): Likewise.
18349         (aarch64_simd_bsl<mode>): Likewise.
18350         (vec_cmp<mode><mode>): Likewise.
18351         (vcond<mode><mode>): Likewise.
18352         (vcond<v_cmp_mixed><mode>): Likewise.
18353         (vcondu<mode><v_cmp_mixed>): Likewise.
18354         (aarch64_cm<optab><mode>): Likewise.
18355         (aarch64_cmtst<mode>): Likewise.
18356         (aarch64_fac<optab><mode>): Likewise.
18357         (vec_perm_const<mode>): Likewise.
18358         (vcond_mask_<mode><v_cmp_result>): Rename to...
18359         (vcond_mask_<mode><v_int_equiv>): ...this.
18360         (vec_cmp<mode><v_cmp_result>): Rename to...
18361         (vec_cmp<mode><v_int_equiv>): ...this.
18363 2017-08-31  Richard Sandiford  <richard.sandiford@linaro.org>
18364             Alan Hayward  <alan.hayward@arm.com>
18365             David Sherwood  <david.sherwood@arm.com>
18367         * config/aarch64/aarch64-modes.def: Remove 32-, 48- and 64-byte
18368         vector modes.
18369         * config/aarch64/iterators.md (VRL2, VRL3, VRL4): Delete.
18370         * config/aarch64/aarch64.md (UNSPEC_LD2_DREG, UNSPEC_LD3_DREG)
18371         (UNSPEC_LD4_DREG): New unspecs.
18372         * config/aarch64/aarch64-simd.md (aarch64_ld2<mode>_dreg_le)
18373         (aarch64_ld2<mode>_dreg_be): Replace with...
18374         (aarch64_ld2<mode>_dreg): ...this pattern and use the new DREG
18375         unspec.
18376         (aarch64_ld3<mode>_dreg_le)
18377         (aarch64_ld3<mode>_dreg_be): Replace with...
18378         (aarch64_ld3<mode>_dreg): ...this pattern and use the new DREG
18379         unspec.
18380         (aarch64_ld4<mode>_dreg_le)
18381         (aarch64_ld4<mode>_dreg_be): Replace with...
18382         (aarch64_ld4<mode>_dreg): ...this pattern and use the new DREG
18383         unspec.
18385 2017-08-30  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
18387         PR tree-optimization/81987
18388         * gimple-ssa-strength-reduction.c (insert_initializers): Don't
18389         insert an initializer in a location not dominated by the stride
18390         definition.
18392 2017-08-30  Eric Botcazou  <ebotcazou@adacore.com>
18394         * tree-eh.c (lower_try_finally_switch): Set the location of the finally
18395         on the entire header of the finally block in the fallthru case.
18397 2017-08-30  Eric Botcazou  <ebotcazou@adacore.com>
18399         * varasm.c (decode_addr_const): Deal with INDIRECT_REF <INTEGER_CST>.
18401 2017-08-30  Pat Haugen  <pthaugen@us.ibm.com>
18403         * config/rs6000/rs6000.c (rs6000_emit_prologue_move_from_cr): Rename from
18404         rs6000_emit_move_from_cr and call renamed function.
18405         (rs6000_emit_prologue): Call renamed functions.
18406         * config/rs6000/rs6000.md (prologue_movesi_from_cr): Rename from
18407         movesi_from_cr, remove volatile CRs.
18409 2017-08-30  Jon Beniston  <jon@beniston.com>
18410             Richard Biener  <rguenther@suse.de>
18412         * tree-vect-patterns.c (vect_pattern_recog_1): Use VECTOR_TYPE_P instead
18413         of VECTOR_MODE_P check.
18414         * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Allow single
18415         element vector types.
18417 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18419         * df.h (df_read_modify_subreg_p): Remove in favor of...
18420         * rtl.h (read_modify_subreg_p): ...this new function.  Take a
18421         const_rtx instead of an rtx.
18422         * cprop.c (local_cprop_find_used_regs): Update accordingly.
18423         * df-problems.c (df_word_lr_mark_ref): Likewise.
18424         * ira-lives.c (mark_pseudo_reg_live): Likewise.
18425         (mark_pseudo_reg_dead): Likewise.
18426         (mark_ref_dead): Likewise.
18427         * reginfo.c (init_subregs_of_mode): Likewise.
18428         * sched-deps.c (sched_analyze_1): Likewise.
18429         * df-scan.c (df_def_record_1): Likewise.
18430         (df_uses_record): Likewise.
18431         (df_read_modify_subreg_p): Remove in favor of...
18432         * rtlanal.c (read_modify_subreg_p): ...this new function.  Take a
18433         const_rtx instead of an rtx.
18435 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18436             Alan Hayward  <alan.hayward@arm.com>
18437             David Sherwood  <david.sherwood@arm.com>
18439         * rtl.h (partial_subreg_p): New function.
18440         * caller-save.c (save_call_clobbered_regs): Use it.
18441         * calls.c (expand_call): Likewise.
18442         * combine.c (combinable_i3pat): Likewise.
18443         (simplify_set): Likewise.
18444         (make_extraction): Likewise.
18445         (make_compound_operation_int): Likewise.
18446         (gen_lowpart_or_truncate): Likewise.
18447         (force_to_mode): Likewise.
18448         (make_field_assignment): Likewise.
18449         (reg_truncated_to_mode): Likewise.
18450         (record_truncated_value): Likewise.
18451         (move_deaths): Likewise.
18452         * cse.c (record_jump_cond): Likewise.
18453         (cse_insn): Likewise.
18454         * cselib.c (cselib_lookup_1): Likewise.
18455         * expmed.c (extract_bit_field_using_extv): Likewise.
18456         * function.c (assign_parm_setup_reg): Likewise.
18457         * ifcvt.c (noce_convert_multiple_sets): Likewise.
18458         * ira-build.c (create_insn_allocnos): Likewise.
18459         * lra-coalesce.c (merge_pseudos): Likewise.
18460         * lra-constraints.c (match_reload): Likewise.
18461         (simplify_operand_subreg): Likewise.
18462         (curr_insn_transform): Likewise.
18463         * lra-lives.c (process_bb_lives): Likewise.
18464         * lra.c (new_insn_reg): Likewise.
18465         (lra_substitute_pseudo): Likewise.
18466         * regcprop.c (mode_change_ok): Likewise.
18467         (maybe_mode_change): Likewise.
18468         (copyprop_hardreg_forward_1): Likewise.
18469         * reload.c (push_reload): Likewise.
18470         (find_reloads): Likewise.
18471         (find_reloads_subreg_address): Likewise.
18472         * reload1.c (alter_reg): Likewise.
18473         (eliminate_regs_1): Likewise.
18474         * simplify-rtx.c (simplify_unary_operation_1): Likewise.
18476 2017-08-30  David Edelsohn  <dje.gcc@gmail.com>
18478         * config/rs6000/rs6000.c (rs6000_expand_binop_builtin): Revert
18479         back to if statements, including unpack.
18481 2017-08-30  Martin Liska  <mliska@suse.cz>
18483         PR inline-asm/82001
18484         * ipa-icf-gimple.c (func_checker::compare_tree_list_operand):
18485         Rename to ...
18486         (func_checker::compare_asm_inputs_outputs): ... this function.
18487         (func_checker::compare_gimple_asm): Use the function to compare
18488         also ASM constrains.
18489         * ipa-icf-gimple.h: Rename the function.
18491 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18492             Alan Hayward  <alan.hayward@arm.com>
18493             David Sherwood  <david.sherwood@arm.com>
18495         * coretypes.h (complex_mode): New type.
18496         * gdbhooks.py (build_pretty_printer): Handle it.
18497         * machmode.h (complex_mode): New class.
18498         (complex_mode::includes_p): New function.
18499         (is_complex_int_mode): Likewise.
18500         (is_complex_float_mode): Likewise.
18501         * genmodes.c (get_mode_class): Handle complex mode classes.
18502         * function.c (expand_function_end): Use is_complex_int_mode.
18504 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18505             Alan Hayward  <alan.hayward@arm.com>
18506             David Sherwood  <david.sherwood@arm.com>
18508         * coretypes.h (scalar_mode_pod): New typedef.
18509         * gdbhooks.py (build_pretty_printer): Handle it.
18510         * machmode.h (gt_ggc_mx, gt_pch_nx): New functions.
18511         * fixed-value.h (fixed_value::mode): Change type to scalar_mode_pod.
18512         * fold-const.c (fold_convert_const_int_from_fixed): Use scalar_mode.
18513         * tree-streamer-in.c (unpack_ts_fixed_cst_value_fields): Use
18514         as_a <scalar_mode>.
18516 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18517             Alan Hayward  <alan.hayward@arm.com>
18518             David Sherwood  <david.sherwood@arm.com>
18520         * machmode.h (mode_for_vector): Take a scalar_mode instead
18521         of a machine_mode.
18522         * stor-layout.c (mode_for_vector): Likewise.
18523         * explow.c (promote_mode): Use as_a <scalar_mode>.
18524         * sdbout.c (sdbout_parms): Use is_a <scalar_mode>.
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 (preferred_simd_mode): Take a scalar_mode
18531         instead of a machine_mode.
18532         * targhooks.h (default_preferred_simd_mode): Likewise.
18533         * targhooks.c (default_preferred_simd_mode): Likewise.
18534         * config/arc/arc.c (arc_preferred_simd_mode): Likewise.
18535         * config/arm/arm.c (arm_preferred_simd_mode): Likewise.
18536         * config/c6x/c6x.c (c6x_preferred_simd_mode): Likewise.
18537         * config/epiphany/epiphany.c (epiphany_preferred_simd_mode): Likewise.
18538         * config/i386/i386.c (ix86_preferred_simd_mode): Likewise.
18539         * config/mips/mips.c (mips_preferred_simd_mode): Likewise.
18540         * config/nvptx/nvptx.c (nvptx_preferred_simd_mode): Likewise.
18541         * config/powerpcspe/powerpcspe.c (rs6000_preferred_simd_mode):
18542         Likewise.
18543         * config/rs6000/rs6000.c (rs6000_preferred_simd_mode): Likewise.
18544         * config/s390/s390.c (s390_preferred_simd_mode): Likewise.
18545         * config/sparc/sparc.c (sparc_preferred_simd_mode): Likewise.
18546         * config/aarch64/aarch64.c (aarch64_preferred_simd_mode): Likewise.
18547         (aarch64_simd_scalar_immediate_valid_for_move): Update accordingly.
18548         * doc/tm.texi: Regenerate.
18549         * optabs-query.c (can_vec_mask_load_store_p): Return false for
18550         non-scalar modes.
18552 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18553             Alan Hayward  <alan.hayward@arm.com>
18554             David Sherwood  <david.sherwood@arm.com>
18556         * target.def (scalar_mode_supported_p): Take a scalar_mode
18557         instead of a machine_mode.
18558         * targhooks.h (default_scalar_mode_supported_p): Likewise.
18559         * targhooks.c (default_scalar_mode_supported_p): Likewise.
18560         * config/aarch64/aarch64.c (aarch64_scalar_mode_supported_p): Likewise.
18561         * config/alpha/alpha.c (alpha_scalar_mode_supported_p): Likewise.
18562         * config/arm/arm.c (arm_scalar_mode_supported_p): Likewise.
18563         * config/avr/avr.c (avr_scalar_mode_supported_p): Likewise.
18564         * config/c6x/c6x.c (c6x_scalar_mode_supported_p): Likewise.
18565         * config/i386/i386.c (ix86_scalar_mode_supported_p): Likewise.
18566         * config/ia64/ia64.c (ia64_scalar_mode_supported_p): Likewise.
18567         * config/mips/mips.c (mips_scalar_mode_supported_p): Likewise.
18568         * config/msp430/msp430.c (msp430_scalar_mode_supported_p): Likewise.
18569         * config/pa/pa.c (pa_scalar_mode_supported_p): Likewise.
18570         * config/pdp11/pdp11.c (pdp11_scalar_mode_supported_p): Likewise.
18571         * config/powerpcspe/powerpcspe.c (rs6000_scalar_mode_supported_p):
18572         Likewise.
18573         * config/rs6000/rs6000.c (rs6000_scalar_mode_supported_p): Likewise.
18574         * config/s390/s390.c (s390_scalar_mode_supported_p): Likewise.
18575         * config/spu/spu.c (spu_scalar_mode_supported_p): Likewise.
18576         * config/tilegx/tilegx.c (tilegx_scalar_mode_supported_p): Likewise.
18577         * config/tilepro/tilepro.c (tilepro_scalar_mode_supported_p):
18578         Likewise.
18579         * doc/tm.texi: Regenerate.
18581 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18582             Alan Hayward  <alan.hayward@arm.com>
18583             David Sherwood  <david.sherwood@arm.com>
18585         * coretypes.h (opt_scalar_mode): New typedef.
18586         * gdbhooks.py (build_pretty_printers): Handle it.
18587         * machmode.h (mode_iterator::get_2xwider): Add overload for
18588         opt_mode<T>.
18589         * emit-rtl.c (init_emit_once): Use opt_scalar_mode when iterating
18590         over scalar modes.
18591         * expr.c (convert_mode_scalar): Likewise.
18592         * omp-low.c (omp_clause_aligned_alignment): Likewise.
18593         * optabs.c (expand_float): Likewise.
18594         (expand_fix): Likewise.
18595         * tree-vect-stmts.c (vectorizable_conversion): Likewise.
18597 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18598             Alan Hayward  <alan.hayward@arm.com>
18599             David Sherwood  <david.sherwood@arm.com>
18601         * optabs.c (expand_float): Explicitly check for scalars before
18602         using a branching expansion.
18603         (expand_fix): Likewise.
18605 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18606             Alan Hayward  <alan.hayward@arm.com>
18607             David Sherwood  <david.sherwood@arm.com>
18609         * expr.c (convert_mode): Split scalar handling out into...
18610         (convert_mode_scalar): ...this new function.  Treat the modes
18611         as scalar_modes.
18613 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18614             Alan Hayward  <alan.hayward@arm.com>
18615             David Sherwood  <david.sherwood@arm.com>
18617         * omp-expand.c (expand_omp_atomic): Use is_int_mode, is_float_mode
18618         and scalar_mode.
18619         * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Likewise.
18621 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18622             Alan Hayward  <alan.hayward@arm.com>
18623             David Sherwood  <david.sherwood@arm.com>
18625         * fixed-value.h (fixed_from_double_int): Take a scalar_mode
18626         rather than a machine_mode.
18627         (fixed_from_string): Likewise.
18628         (fixed_convert): Likewise.
18629         (fixed_convert_from_int): Likewise.
18630         (fixed_convert_from_real): Likewise.
18631         (real_convert_from_fixed): Likewise.
18632         * fixed-value.c (fixed_from_double_int): Likewise.
18633         (fixed_from_string): Likewise.
18634         (fixed_convert): Likewise.
18635         (fixed_convert_from_int): Likewise.
18636         (fixed_convert_from_real): Likewise.
18637         (real_convert_from_fixed): Likewise.
18638         * config/avr/avr.c (avr_out_round): Use as_a <scalar_mode>.
18640 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18641             Alan Hayward  <alan.hayward@arm.com>
18642             David Sherwood  <david.sherwood@arm.com>
18644         * emit-rtl.c (immed_double_const): Use is_a <scalar_mode> instead
18645         of separate mode class checks.  Do not allow vector modes here.
18646         (immed_wide_int_const): Use as_a <scalar_mode>.
18647         * explow.c (trunc_int_for_mode): Likewise.
18648         * rtl.h (wi::int_traits<rtx_mode_t>::get_precision): Likewise.
18649         (wi::shwi): Likewise.
18650         (wi::min_value): Likewise.
18651         (wi::max_value): Likewise.
18652         * dwarf2out.c (loc_descriptor): Likewise.
18653         * simplify-rtx.c (simplify_immed_subreg): Fix rtx_mode_t argument
18654         for CONST_WIDE_INT.
18656 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18657             Alan Hayward  <alan.hayward@arm.com>
18658             David Sherwood  <david.sherwood@arm.com>
18660         * tree.h (SCALAR_TYPE_MODE): New macro.
18661         * expr.c (expand_expr_addr_expr_1): Use it.
18662         (expand_expr_real_2): Likewise.
18663         * fold-const.c (fold_convert_const_fixed_from_fixed): Likeise.
18664         (fold_convert_const_fixed_from_int): Likewise.
18665         (fold_convert_const_fixed_from_real): Likewise.
18666         (native_encode_fixed): Likewise
18667         (native_encode_complex): Likewise
18668         (native_encode_vector): Likewise.
18669         (native_interpret_fixed): Likewise.
18670         (native_interpret_real): Likewise.
18671         (native_interpret_complex): Likewise.
18672         (native_interpret_vector): Likewise.
18673         * omp-simd-clone.c (simd_clone_adjust_return_type): Likewise.
18674         (simd_clone_adjust_argument_types): Likewise.
18675         (simd_clone_init_simd_arrays): Likewise.
18676         (simd_clone_adjust): Likewise.
18677         * stor-layout.c (layout_type): Likewise.
18678         * tree.c (build_minus_one_cst): Likewise.
18679         * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
18680         * tree-inline.c (estimate_move_cost): Likewise.
18681         * tree-ssa-math-opts.c (convert_plusminus_to_widen): Likewise.
18682         * tree-vect-loop.c (vect_create_epilog_for_reduction): Likewise.
18683         (vectorizable_reduction): Likewise.
18684         * tree-vect-patterns.c (vect_recog_widen_mult_pattern): Likewise.
18685         (vect_recog_mixed_size_cond_pattern): Likewise.
18686         (check_bool_pattern): Likewise.
18687         (adjust_bool_pattern): Likewise.
18688         (search_type_for_mask_1): Likewise.
18689         * tree-vect-slp.c (vect_schedule_slp_instance): Likewise.
18690         * tree-vect-stmts.c (vectorizable_conversion): Likewise.
18691         (vectorizable_load): Likewise.
18692         (vectorizable_store): Likewise.
18693         * ubsan.c (ubsan_encode_value): Likewise.
18694         * varasm.c (output_constant): Likewise.
18696 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18697             Alan Hayward  <alan.hayward@arm.com>
18698             David Sherwood  <david.sherwood@arm.com>
18700         * coretypes.h (scalar_mode): New class.
18701         * machmode.h (scalar_mode): Likewise.
18702         (scalar_mode::includes_p): New function.
18703         (mode_to_inner): Return a scalar_mode rather than a machine_mode.
18704         * gdbhooks.py (build_pretty_printers): Handle scalar_mode.
18705         * genmodes.c (get_mode_class): Handle remaining scalar modes.
18706         * cfgexpand.c (expand_debug_expr): Use scalar_mode.
18707         * expmed.c (store_bit_field_1): Likewise.
18708         (extract_bit_field_1): Likewise.
18709         * expr.c (write_complex_part): Likewise.
18710         (read_complex_part): Likewise.
18711         (emit_move_complex_push): Likewise.
18712         (expand_expr_real_2): Likewise.
18713         * function.c (assign_parm_setup_reg): Likewise.
18714         (assign_parms_unsplit_complex): Likewise.
18715         * optabs.c (expand_binop): Likewise.
18716         * rtlanal.c (subreg_get_info): Likewise.
18717         * simplify-rtx.c (simplify_immed_subreg): Likewise.
18718         * varasm.c (output_constant_pool_2): Likewise.
18720 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18721             Alan Hayward  <alan.hayward@arm.com>
18722             David Sherwood  <david.sherwood@arm.com>
18724         * expmed.c (extract_high_half): Use scalar_int_mode and remove
18725         assertion.
18726         (expmed_mult_highpart_optab): Likewise.
18727         (expmed_mult_highpart): Likewise.
18729 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18730             Alan Hayward  <alan.hayward@arm.com>
18731             David Sherwood  <david.sherwood@arm.com>
18733         * builtins.h (builtin_strncpy_read_str): Take a scalar_int_mode
18734         instead of a machine_mode.
18735         (builtin_memset_read_str): Likewise.
18736         * builtins.c (c_readstr): Likewise.
18737         (builtin_memcpy_read_str): Likewise.
18738         (builtin_strncpy_read_str): Likewise.
18739         (builtin_memset_read_str): Likewise.
18740         (builtin_memset_gen_str): Likewise.
18741         (expand_builtin_signbit): Use scalar_int_mode for local variables.
18742         * cfgexpand.c (convert_debug_memory_address): Take a scalar_int_mode
18743         instead of a machine_mode.
18744         * combine.c (simplify_if_then_else): Use scalar_int_mode for local
18745         variables.
18746         (make_extraction): Likewise.
18747         (try_widen_shift_mode): Take and return scalar_int_modes instead
18748         of machine_modes.
18749         * config/aarch64/aarch64.c (aarch64_libgcc_cmp_return_mode): Return
18750         a scalar_int_mode instead of a machine_mode.
18751         * config/avr/avr.c (avr_addr_space_address_mode): Likewise.
18752         (avr_addr_space_pointer_mode): Likewise.
18753         * config/cr16/cr16.c (cr16_unwind_word_mode): Likewise.
18754         * config/msp430/msp430.c (msp430_addr_space_pointer_mode): Likewise.
18755         (msp430_unwind_word_mode): Likewise.
18756         * config/spu/spu.c (spu_unwind_word_mode): Likewise.
18757         (spu_addr_space_pointer_mode): Likewise.
18758         (spu_addr_space_address_mode): Likewise.
18759         (spu_libgcc_cmp_return_mode): Likewise.
18760         (spu_libgcc_shift_count_mode): Likewise.
18761         * config/rl78/rl78.c (rl78_addr_space_address_mode): Likewise.
18762         (rl78_addr_space_pointer_mode): Likewise.
18763         (fl78_unwind_word_mode): Likewise.
18764         (rl78_valid_pointer_mode): Take a scalar_int_mode instead of a
18765         machine_mode.
18766         * config/alpha/alpha.c (vms_valid_pointer_mode): Likewise.
18767         * config/ia64/ia64.c (ia64_vms_valid_pointer_mode): Likewise.
18768         * config/mips/mips.c (mips_mode_rep_extended): Likewise.
18769         (mips_valid_pointer_mode): Likewise.
18770         * config/tilegx/tilegx.c (tilegx_mode_rep_extended): Likewise.
18771         * config/ft32/ft32.c (ft32_valid_pointer_mode): Likewise.
18772         (ft32_addr_space_pointer_mode): Return a scalar_int_mode instead
18773         of a machine_mode.
18774         (ft32_addr_space_address_mode): Likewise.
18775         * config/m32c/m32c.c (m32c_valid_pointer_mode): Take a
18776         scalar_int_mode instead of a machine_mode.
18777         (m32c_addr_space_pointer_mode): Return a scalar_int_mode instead
18778         of a machine_mode.
18779         (m32c_addr_space_address_mode): Likewise.
18780         * config/powerpcspe/powerpcspe.c (rs6000_abi_word_mode): Likewise.
18781         (rs6000_eh_return_filter_mode): Likewise.
18782         * config/rs6000/rs6000.c (rs6000_abi_word_mode): Likewise.
18783         (rs6000_eh_return_filter_mode): Likewise.
18784         * config/s390/s390.c (s390_libgcc_cmp_return_mode): Likewise.
18785         (s390_libgcc_shift_count_mode): Likewise.
18786         (s390_unwind_word_mode): Likewise.
18787         (s390_valid_pointer_mode): Take a scalar_int_mode rather than a
18788         machine_mode.
18789         * target.def (mode_rep_extended): Likewise.
18790         (valid_pointer_mode): Likewise.
18791         (addr_space.valid_pointer_mode): Likewise.
18792         (eh_return_filter_mode): Return a scalar_int_mode rather than
18793         a machine_mode.
18794         (libgcc_cmp_return_mode): Likewise.
18795         (libgcc_shift_count_mode): Likewise.
18796         (unwind_word_mode): Likewise.
18797         (addr_space.pointer_mode): Likewise.
18798         (addr_space.address_mode): Likewise.
18799         * doc/tm.texi: Regenerate.
18800         * dojump.c (prefer_and_bit_test): Take a scalar_int_mode rather than
18801         a machine_mode.
18802         (do_jump): Use scalar_int_mode for local variables.
18803         * dwarf2cfi.c (init_return_column_size): Take a scalar_int_mode
18804         rather than a machine_mode.
18805         * dwarf2out.c (convert_descriptor_to_mode): Likewise.
18806         (scompare_loc_descriptor_wide): Likewise.
18807         (scompare_loc_descriptor_narrow): Likewise.
18808         * emit-rtl.c (adjust_address_1): Use scalar_int_mode for local
18809         variables.
18810         * except.c (sjlj_emit_dispatch_table): Likewise.
18811         (expand_builtin_eh_copy_values): Likewise.
18812         * explow.c (convert_memory_address_addr_space_1): Likewise.
18813         Take a scalar_int_mode rather than a machine_mode.
18814         (convert_memory_address_addr_space): Take a scalar_int_mode rather
18815         than a machine_mode.
18816         (memory_address_addr_space): Use scalar_int_mode for local variables.
18817         * expmed.h (expand_mult_highpart_adjust): Take a scalar_int_mode
18818         rather than a machine_mode.
18819         * expmed.c (mask_rtx): Likewise.
18820         (init_expmed_one_conv): Likewise.
18821         (expand_mult_highpart_adjust): Likewise.
18822         (extract_high_half): Likewise.
18823         (expmed_mult_highpart_optab): Likewise.
18824         (expmed_mult_highpart): Likewise.
18825         (expand_smod_pow2): Likewise.
18826         (expand_sdiv_pow2): Likewise.
18827         (emit_store_flag_int): Likewise.
18828         (adjust_bit_field_mem_for_reg): Use scalar_int_mode for local
18829         variables.
18830         (extract_low_bits): Likewise.
18831         * expr.h (by_pieces_constfn): Take a scalar_int_mode rather than
18832         a machine_mode.
18833         * expr.c (pieces_addr::adjust):  Likewise.
18834         (can_store_by_pieces): Likewise.
18835         (store_by_pieces): Likewise.
18836         (clear_by_pieces_1): Likewise.
18837         (expand_expr_addr_expr_1): Likewise.
18838         (expand_expr_addr_expr): Use scalar_int_mode for local variables.
18839         (expand_expr_real_1): Likewise.
18840         (try_casesi): Likewise.
18841         * final.c (shorten_branches): Likewise.
18842         * fold-const.c (fold_convert_const_int_from_fixed): Change the
18843         type of "mode" to machine_mode.
18844         * internal-fn.c (expand_arith_overflow_result_store): Take a
18845         scalar_int_mode rather than a machine_mode.
18846         (expand_mul_overflow): Use scalar_int_mode for local variables.
18847         * loop-doloop.c (doloop_modify): Likewise.
18848         (doloop_optimize): Likewise.
18849         * optabs.c (expand_subword_shift): Take a scalar_int_mode rather
18850         than a machine_mode.
18851         (expand_doubleword_shift_condmove): Likewise.
18852         (expand_doubleword_shift): Likewise.
18853         (expand_doubleword_clz): Likewise.
18854         (expand_doubleword_popcount): Likewise.
18855         (expand_doubleword_parity): Likewise.
18856         (expand_absneg_bit): Use scalar_int_mode for local variables.
18857         (prepare_float_lib_cmp): Likewise.
18858         * rtl.h (convert_memory_address_addr_space_1): Take a scalar_int_mode
18859         rather than a machine_mode.
18860         (convert_memory_address_addr_space): Likewise.
18861         (get_mode_bounds): Likewise.
18862         (get_address_mode): Return a scalar_int_mode rather than a
18863         machine_mode.
18864         * rtlanal.c (get_address_mode): Likewise.
18865         * stor-layout.c (get_mode_bounds): Take a scalar_int_mode rather
18866         than a machine_mode.
18867         * targhooks.c (default_mode_rep_extended): Likewise.
18868         (default_valid_pointer_mode): Likewise.
18869         (default_addr_space_valid_pointer_mode): Likewise.
18870         (default_eh_return_filter_mode): Return a scalar_int_mode rather
18871         than a machine_mode.
18872         (default_libgcc_cmp_return_mode): Likewise.
18873         (default_libgcc_shift_count_mode): Likewise.
18874         (default_unwind_word_mode): Likewise.
18875         (default_addr_space_pointer_mode): Likewise.
18876         (default_addr_space_address_mode): Likewise.
18877         * targhooks.h (default_eh_return_filter_mode): Likewise.
18878         (default_libgcc_cmp_return_mode): Likewise.
18879         (default_libgcc_shift_count_mode): Likewise.
18880         (default_unwind_word_mode): Likewise.
18881         (default_addr_space_pointer_mode): Likewise.
18882         (default_addr_space_address_mode): Likewise.
18883         (default_mode_rep_extended): Take a scalar_int_mode rather than
18884         a machine_mode.
18885         (default_valid_pointer_mode): Likewise.
18886         (default_addr_space_valid_pointer_mode): Likewise.
18887         * tree-ssa-address.c (addr_for_mem_ref): Use scalar_int_mode for
18888         local variables.
18889         * tree-ssa-loop-ivopts.c (get_shiftadd_cost): Take a scalar_int_mode
18890         rather than a machine_mode.
18891         * tree-switch-conversion.c (array_value_type): Use scalar_int_mode
18892         for local variables.
18893         * tree-vrp.c (simplify_float_conversion_using_ranges): Likewise.
18894         * var-tracking.c (use_narrower_mode): Take a scalar_int_mode rather
18895         than a machine_mode.
18897 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18898             Alan Hayward  <alan.hayward@arm.com>
18899             David Sherwood  <david.sherwood@arm.com>
18901         * dojump.c (do_jump_by_parts_greater_rtx): Change the type of
18902         the mode argument to scalar_int_mode.
18903         (do_jump_by_parts_zero_rtx): Likewise.
18904         (do_jump_by_parts_equality_rtx): Likewise.
18905         (do_jump_by_parts_greater): Take a mode argument.
18906         (do_jump_by_parts_equality): Likewise.
18907         (do_jump_1): Update calls accordingly.
18909 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18910             Alan Hayward  <alan.hayward@arm.com>
18911             David Sherwood  <david.sherwood@arm.com>
18913         * is-a.h (safe_dyn_cast): New function.
18914         * rtl.h (rtx_jump_table_data::get_data_mode): New function.
18915         (jump_table_for_label): Likewise.
18916         * final.c (final_addr_vec_align): Take an rtx_jump_table_data *
18917         instead of an rtx_insn *.
18918         (shorten_branches): Use dyn_cast instead of LABEL_P and
18919         JUMP_TABLE_DATA_P.  Use jump_table_for_label and
18920         rtx_jump_table_data::get_data_mode.
18921         (final_scan_insn): Likewise.
18923 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18924             Alan Hayward  <alan.hayward@arm.com>
18925             David Sherwood  <david.sherwood@arm.com>
18927         * combine.c (try_combine): Use is_a <scalar_int_mode> when
18928         trying to combine a full-register integer set with a subreg
18929         integer set.
18931 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18932             Alan Hayward  <alan.hayward@arm.com>
18933             David Sherwood  <david.sherwood@arm.com>
18935         * expr.c (expand_expr_addr_expr): Add a new_tmode local variable
18936         that is always either address_mode or pointer_mode.
18938 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18939             Alan Hayward  <alan.hayward@arm.com>
18940             David Sherwood  <david.sherwood@arm.com>
18942         * expr.c (expand_expr_real_2): Use word_mode instead of innermode
18943         when the two are known to be equal.
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         * simplify-rtx.c (simplify_const_unary_operation): Use
18950         is_a <scalar_int_mode> instead of checking for a nonzero
18951         precision.  Forcibly convert op_mode to a scalar_int_mode
18952         in that case.  More clearly differentiate the operand and
18953         result modes and use the former when deciding what the value
18954         of a count-bits operation should be.  Use is_int_mode instead
18955         of checking for a MODE_INT.  Remove redundant check for whether
18956         this mode has a zero precision.
18958 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18959             Alan Hayward  <alan.hayward@arm.com>
18960             David Sherwood  <david.sherwood@arm.com>
18962         * optabs.c (widen_leading): Change the type of the mode argument
18963         to scalar_int_mode.  Use opt_scalar_int_mode for the mode iterator.
18964         (widen_bswap): Likewise.
18965         (expand_parity): Likewise.
18966         (expand_ctz): Change the type of the mode argument to scalar_int_mode.
18967         (expand_ffs): Likewise.
18968         (epand_unop): Check for scalar integer modes before calling the
18969         above routines.
18971 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18972             Alan Hayward  <alan.hayward@arm.com>
18973             David Sherwood  <david.sherwood@arm.com>
18975         * expr.c (const_scalar_mask_from_tree): Add a mode argument.
18976         Expand commentary.
18977         (expand_expr_real_1): Update call accordingly.
18979 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18980             Alan Hayward  <alan.hayward@arm.com>
18981             David Sherwood  <david.sherwood@arm.com>
18983         * expmed.c (store_bit_field_using_insv): Add op0_mode and
18984         value_mode arguments.  Use scalar_int_mode internally.
18985         (store_bit_field_1): Rename the new integer mode from imode
18986         to op0_mode and use it instead of GET_MODE (op0).  Update calls
18987         to store_split_bit_field, store_bit_field_using_insv and
18988         store_fixed_bit_field.
18989         (store_fixed_bit_field): Add op0_mode and value_mode arguments.
18990         Use scalar_int_mode internally.  Use a bit count rather than a mode
18991         when calculating the largest bit size for get_best_mode.
18992         Update calls to store_split_bit_field and store_fixed_bit_field_1.
18993         (store_fixed_bit_field_1): Add mode and value_mode arguments.
18994         Remove assertion that OP0 has a scalar integer mode.
18995         (store_split_bit_field): Add op0_mode and value_mode arguments.
18996         Update calls to extract_fixed_bit_field.
18997         (extract_bit_field_using_extv): Add an op0_mode argument.
18998         Use scalar_int_mode internally.
18999         (extract_bit_field_1): Rename the new integer mode from imode to
19000         op0_mode and use it instead of GET_MODE (op0).  Update calls to
19001         extract_split_bit_field, extract_bit_field_using_extv and
19002         extract_fixed_bit_field.
19003         (extract_fixed_bit_field): Add an op0_mode argument.  Update calls
19004         to extract_split_bit_field and extract_fixed_bit_field_1.
19005         (extract_fixed_bit_field_1): Add a mode argument.  Remove assertion
19006         that OP0 has a scalar integer mode.  Use as_a <scalar_int_mode>
19007         on the target mode.
19008         (extract_split_bit_field): Add an op0_mode argument.  Update call
19009         to extract_fixed_bit_field.
19011 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19012             Alan Hayward  <alan.hayward@arm.com>
19013             David Sherwood  <david.sherwood@arm.com>
19015         * cse.c (cse_insn): Use opt_scalar_int_mode for the mode iterator.
19016         * explow.c (hard_function_value): Likewise.
19017         * expmed.c (extract_fixed_bit_field_1): Likewise.  Move the
19018         convert_to_mode call outside the loop.
19019         * expr.c (alignment_for_piecewise_move): Use opt_scalar_int_mode
19020         for the mode iterator.  Require the mode specified by max_pieces
19021         to exist.
19022         (emit_block_move_via_movmem): Use opt_scalar_int_mode for the
19023         mode iterator.
19024         (copy_blkmode_to_reg): Likewise.
19025         (set_storage_via_setmem): Likewise.
19026         * optabs.c (prepare_cmp_insn): Likewise.
19027         * rtlanal.c (init_num_sign_bit_copies_in_rep): Likewise.
19028         * stor-layout.c (finish_bitfield_representative): Likewise.
19030 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19031             Alan Hayward  <alan.hayward@arm.com>
19032             David Sherwood  <david.sherwood@arm.com>
19034         * rtl.h (subreg_unpromoted_mode, subreg_promoted_mode): New functions.
19035         * expr.c (convert_move): Use them.
19036         (convert_modes): Likewise.
19037         (store_expr_with_bounds): Likewise.
19039 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19040             Alan Hayward  <alan.hayward@arm.com>
19041             David Sherwood  <david.sherwood@arm.com>
19043         * rtl.h (rtl_hooks::reg_nonzero_bits): Add a scalar_int_mode
19044         parameter for the mode of "x".  Remove the "known_x", "known_mode"
19045         and "known_ret" arguments.  Change the type of the mode argument
19046         to scalar_int_mode.
19047         (rtl_hooks:reg_num_sign_bit_copies): Likewise.
19048         * combine.c (reg_nonzero_bits_for_combine): Update accordingly.
19049         (reg_num_sign_bit_copies_for_combine): Likewise.
19050         * rtlanal.c (nonzero_bits1): Likewise.
19051         (num_sign_bit_copies1): Likewise.
19052         * rtlhooks-def.h (reg_nonzero_bits_general): Likewise.
19053         (reg_num_sign_bit_copies_general): Likewise.
19054         * rtlhooks.c (reg_num_sign_bit_copies_general): Likewise.
19055         (reg_nonzero_bits_general): Likewise.
19057 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19058             Alan Hayward  <alan.hayward@arm.com>
19059             David Sherwood  <david.sherwood@arm.com>
19061         * rtlanal.c (num_sign_bit_copies): Handle VOIDmode here rather
19062         than in subroutines.  Return 1 for non-integer modes.
19063         (cached_num_sign_bit_copies): Change the type of the mode parameter
19064         to scalar_int_mode.
19065         (num_sign_bit_copies1): Likewise.  Remove early exit for other mode
19066         classes.  Handle CONST_INT_P first and then check whether X also
19067         has a scalar integer mode.  Check the same thing for inner registers
19068         of a SUBREG and for values that are being extended or truncated.
19070 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19071             Alan Hayward  <alan.hayward@arm.com>
19072             David Sherwood  <david.sherwood@arm.com>
19074         * rtlanal.c (nonzero_bits): Handle VOIDmode here rather than
19075         in subroutines.  Return the mode mask for non-integer modes.
19076         (cached_nonzero_bits): Change the type of the mode parameter
19077         to scalar_int_mode.
19078         (nonzero_bits1): Likewise.  Remove early exit for other mode
19079         classes.  Handle CONST_INT_P first and then check whether X
19080         also has a scalar integer mode.
19082 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19083             Alan Hayward  <alan.hayward@arm.com>
19084             David Sherwood  <david.sherwood@arm.com>
19086         * expr.c (widest_int_mode_for_size): Make the comment match the code.
19087         Return a scalar_int_mode and assert that the size is greater than
19088         one byte.
19089         (by_pieces_ninsns): Update accordingly and remove VOIDmode handling.
19090         (op_by_pieces_d::op_by_pieces_d): Likewise.
19091         (op_by_pieces_d::run): Likewise.
19092         (can_store_by_pieces): Likewise.
19094 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19095             Alan Hayward  <alan.hayward@arm.com>
19096             David Sherwood  <david.sherwood@arm.com>
19098         * combine.c (extract_left_shift): Add a mode argument and update
19099         recursive calls.
19100         (make_compound_operation_int): Change the type of the mode parameter
19101         to scalar_int_mode and update the call to extract_left_shift.
19103 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19104             Alan Hayward  <alan.hayward@arm.com>
19105             David Sherwood  <david.sherwood@arm.com>
19107         * combine.c (simplify_and_const_int): Change the type of the mode
19108         parameter to scalar_int_mode.
19109         (simplify_and_const_int_1): Likewise.  Update recursive call.
19111 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19112             Alan Hayward  <alan.hayward@arm.com>
19113             David Sherwood  <david.sherwood@arm.com>
19115         * combine.c (simplify_compare_const): Check that the mode is a
19116         scalar_int_mode (rather than VOIDmode) before testing its
19117         precision.
19118         (simplify_comparison): Move COMPARISON_P handling out of the
19119         loop and restrict the latter part of the loop to scalar_int_modes.
19120         Check is_a <scalar_int_mode> before calling HWI_COMPUTABLE_MODE_P
19121         and when considering SUBREG_REGs.  Use is_int_mode instead of
19122         checking GET_MODE_CLASS against MODE_INT.
19124 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19125             Alan Hayward  <alan.hayward@arm.com>
19126             David Sherwood  <david.sherwood@arm.com>
19128         * combine.c (try_widen_shift_mode): Move check for equal modes to...
19129         (simplify_shift_const_1): ...here.  Use scalar_int_mode for
19130         shift_unit_mode and for modes involved in scalar shifts.
19132 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19133             Alan Hayward  <alan.hayward@arm.com>
19134             David Sherwood  <david.sherwood@arm.com>
19136         * combine.c (force_int_to_mode): New function, split out from...
19137         (force_to_mode): ...here.  Keep xmode up-to-date and use it
19138         instead of GET_MODE (x).
19140 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19141             Alan Hayward  <alan.hayward@arm.com>
19142             David Sherwood  <david.sherwood@arm.com>
19144         * optabs-query.h (extraction_insn::struct_mode): Change type to
19145         opt_scalar_int_mode and update comment.
19146         (extraction_insn::field_mode): Change type to scalar_int_mode.
19147         (extraction_insn::pos_mode): Likewise.
19148         * combine.c (make_extraction): Update accordingly.
19149         * optabs-query.c (get_traditional_extraction_insn): Likewise.
19150         (get_optab_extraction_insn): Likewise.
19151         * recog.c (simplify_while_replacing): Likewise.
19152         * expmed.c (narrow_bit_field_mem): Change the type of the mode
19153         parameter to opt_scalar_int_mode.
19155 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19156             Alan Hayward  <alan.hayward@arm.com>
19157             David Sherwood  <david.sherwood@arm.com>
19159         * machmode.h (bit_field_mode_iterator::next_mode): Take a pointer
19160         to a scalar_int_mode instead of a machine_mode.
19161         (bit_field_mode_iterator::m_mode): Change type to opt_scalar_int_mode.
19162         (get_best_mode): Return a boolean and use a pointer argument to store
19163         the selected mode.  Replace the limit mode parameter with a bit limit.
19164         * expmed.c (adjust_bit_field_mem_for_reg): Use scalar_int_mode
19165         for the values returned by bit_field_mode_iterator::next_mode.
19166         (store_bit_field): Update call to get_best_mode.
19167         (store_fixed_bit_field): Likewise.
19168         (extract_fixed_bit_field): Likewise.
19169         * expr.c (optimize_bitfield_assignment_op): Likewise.
19170         * fold-const.c (optimize_bit_field_compare): Likewise.
19171         (fold_truth_andor_1): Likewise.
19172         * stor-layout.c (bit_field_mode_iterator::next_mode): As above.
19173         Update for new type of m_mode.
19174         (get_best_mode): As above.
19176 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19177             Alan Hayward  <alan.hayward@arm.com>
19178             David Sherwood  <david.sherwood@arm.com>
19180         * expmed.c (strict_volatile_bitfield_p): Change the type of fieldmode
19181         to scalar_int_mode.  Remove check for SCALAR_INT_MODE_P.
19182         (store_bit_field): Check is_a <scalar_int_mode> before calling
19183         strict_volatile_bitfield_p.
19184         (extract_bit_field): Likewise.
19186 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19187             Alan Hayward  <alan.hayward@arm.com>
19188             David Sherwood  <david.sherwood@arm.com>
19190         * target.def (cstore_mode): Return a scalar_int_mode.
19191         * doc/tm.texi: Regenerate.
19192         * config/sparc/sparc.c (sparc_cstore_mode): Return a scalar_int_mode.
19193         * targhooks.h (default_cstore_mode): Likewise.
19194         * targhooks.c (default_cstore_mode): Likewise, using a forced
19195         conversion.
19196         * expmed.c (emit_cstore): Expect the target of the cstore to be
19197         a scalar_int_mode.
19199 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19200             Alan Hayward  <alan.hayward@arm.com>
19201             David Sherwood  <david.sherwood@arm.com>
19203         * cfgloop.h (rtx_iv): Change type of extend_mode and mode to
19204         scalar_int_mode.
19205         (niter_desc): Likewise mode.
19206         (iv_analyze): Add a mode parameter.
19207         (biv_p): Likewise.
19208         (iv_analyze_expr): Pass the mode paraeter before the rtx it describes
19209         and change its type to scalar_int_mode.
19210         * loop-iv.c: Update commentary at head of file.
19211         (iv_constant): Pass the mode paraeter before the rtx it describes
19212         and change its type to scalar_int_mode.  Remove VOIDmode handling.
19213         (iv_subreg): Change the type of the mode parameter to scalar_int_mode.
19214         (iv_extend): Likewise.
19215         (shorten_into_mode): Likewise.
19216         (iv_add): Use scalar_int_mode.
19217         (iv_mult): Likewise.
19218         (iv_shift): Likewise.
19219         (canonicalize_iv_subregs): Likewise.
19220         (get_biv_step_1): Pass the outer_mode parameter before the rtx
19221         it describes and change its mode to scalar_int_mode.   Also change
19222         the type of the returned inner_mode to scalar_int_mode.
19223         (get_biv_step): Likewise, turning outer_mode from a pointer
19224         into a direct parameter.  Update call to get_biv_step_1.
19225         (iv_analyze_biv): Add an outer_mode parameter.  Update calls to
19226         iv_constant and get_biv_step.
19227         (iv_analyze_expr): Pass the mode parameter before the rtx it describes
19228         and change its type to scalar_int_mode.  Don't initialise iv->mode
19229         to VOIDmode and remove later checks for its still being VOIDmode.
19230         Update calls to iv_analyze_op and iv_analyze_expr.  Check
19231         is_a <scalar_int_mode> when changing the mode under consideration.
19232         (iv_analyze_def): Ignore registers that don't have a scalar_int_mode.
19233         Update call to iv_analyze_expr.
19234         (iv_analyze_op): Add a mode parameter.  Reject subregs whose
19235         inner register is not also a scalar_int_mode.  Update call to
19236         iv_analyze_biv.
19237         (iv_analyze): Add a mode parameter.  Update call to iv_analyze_op.
19238         (biv_p): Add a mode parameter.  Update call to iv_analyze_biv.
19239         (iv_number_of_iterations): Use is_a <scalar_int_mode> instead of
19240         separate mode class checks.  Update calls to iv_analyze.  Remove
19241         fix-up of VOIDmodes after iv_analyze_biv.
19242         * loop-unroll.c (analyze_iv_to_split_insn): Reject registers that
19243         don't have a scalar_int_mode.  Update call to biv_p.
19245 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19246             Alan Hayward  <alan.hayward@arm.com>
19247             David Sherwood  <david.sherwood@arm.com>
19249         * cfgexpand.c (convert_debug_memory_address): Use
19250         as_a <scalar_int_mode>.
19251         * combine.c (expand_compound_operation): Likewise.
19252         (make_extraction): Likewise.
19253         (change_zero_ext): Likewise.
19254         (simplify_comparison): Likewise.
19255         * cse.c (cse_insn): Likewise.
19256         * dwarf2out.c (minmax_loc_descriptor): Likewise.
19257         (mem_loc_descriptor): Likewise.
19258         (loc_descriptor): Likewise.
19259         * expmed.c (init_expmed_one_mode): Likewise.
19260         (synth_mult): Likewise.
19261         (emit_store_flag_1): Likewise.
19262         (expand_divmod): Likewise.  Use HWI_COMPUTABLE_MODE_P instead
19263         of a comparison with size.
19264         * expr.c (expand_assignment): Use as_a <scalar_int_mode>.
19265         (reduce_to_bit_field_precision): Likewise.
19266         * function.c (expand_function_end): Likewise.
19267         * internal-fn.c (expand_arith_overflow_result_store): Likewise.
19268         * loop-doloop.c (doloop_modify): Likewise.
19269         * optabs.c (expand_binop): Likewise.
19270         (expand_unop): Likewise.
19271         (expand_copysign_absneg): Likewise.
19272         (prepare_cmp_insn): Likewise.
19273         (maybe_legitimize_operand): Likewise.
19274         * recog.c (const_scalar_int_operand): Likewise.
19275         * rtlanal.c (get_address_mode): Likewise.
19276         * simplify-rtx.c (simplify_unary_operation_1): Likewise.
19277         (simplify_cond_clz_ctz): Likewise.
19278         * tree-nested.c (get_nl_goto_field): Likewise.
19279         * tree.c (build_vector_type_for_mode): Likewise.
19280         * var-tracking.c (use_narrower_mode): Likewise.
19282 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19283             Alan Hayward  <alan.hayward@arm.com>
19284             David Sherwood  <david.sherwood@arm.com>
19286         * tree.h (SCALAR_INT_TYPE_MODE): New macro.
19287         * builtins.c (expand_builtin_signbit): Use it.
19288         * cfgexpand.c (expand_debug_expr): Likewise.
19289         * dojump.c (do_jump): Likewise.
19290         (do_compare_and_jump): Likewise.
19291         * dwarf2cfi.c (expand_builtin_init_dwarf_reg_sizes): Likewise.
19292         * expmed.c (make_tree): Likewise.
19293         * expr.c (expand_expr_real_2): Likewise.
19294         (expand_expr_real_1): Likewise.
19295         (try_casesi): Likewise.
19296         * fold-const-call.c (fold_const_call_ss): Likewise.
19297         * fold-const.c (unextend): Likewise.
19298         (extract_muldiv_1): Likewise.
19299         (fold_single_bit_test): Likewise.
19300         (native_encode_int): Likewise.
19301         (native_encode_string): Likewise.
19302         (native_interpret_int): Likewise.
19303         * gimple-fold.c (gimple_fold_builtin_memset): Likewise.
19304         * internal-fn.c (expand_addsub_overflow): Likewise.
19305         (expand_neg_overflow): Likewise.
19306         (expand_mul_overflow): Likewise.
19307         (expand_arith_overflow): Likewise.
19308         * match.pd: Likewise.
19309         * stor-layout.c (layout_type): Likewise.
19310         * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
19311         * tree-ssa-math-opts.c (convert_mult_to_widen): Likewise.
19312         * tree-ssanames.c (get_range_info): Likewise.
19313         * tree-switch-conversion.c (array_value_type) Likewise.
19314         * tree-vect-patterns.c (vect_recog_rotate_pattern): Likewise.
19315         (vect_recog_divmod_pattern): Likewise.
19316         (vect_recog_mixed_size_cond_pattern): Likewise.
19317         * tree-vrp.c (extract_range_basic): Likewise.
19318         (simplify_float_conversion_using_ranges): Likewise.
19319         * tree.c (int_fits_type_p): Likewise.
19320         * ubsan.c (instrument_bool_enum_load): Likewise.
19321         * varasm.c (mergeable_string_section): Likewise.
19322         (narrowing_initializer_constant_valid_p): Likewise.
19323         (output_constant): Likewise.
19325 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19326             Alan Hayward  <alan.hayward@arm.com>
19327             David Sherwood  <david.sherwood@arm.com>
19329         * machmode.h (NARROWEST_INT_MODE): New macro.
19330         * expr.c (alignment_for_piecewise_move): Use it instead of
19331         GET_CLASS_NARROWEST_MODE (MODE_INT).
19332         (push_block): Likewise.
19333         * stor-layout.c (bit_field_mode_iterator::bit_field_mode_iterator):
19334         Likewise.
19335         * tree-vrp.c (simplify_float_conversion_using_ranges): Likewise.
19337 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19338             Alan Hayward  <alan.hayward@arm.com>
19339             David Sherwood  <david.sherwood@arm.com>
19341         * postreload.c (move2add_valid_value_p): Change the type of the
19342         mode parameter to scalar_int_mode.
19343         (move2add_use_add2_insn): Add a mode parameter and use it instead
19344         of GET_MODE (reg).
19345         (move2add_use_add3_insn): Likewise.
19346         (reload_cse_move2add): Update accordingly.
19348 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19349             Alan Hayward  <alan.hayward@arm.com>
19350             David Sherwood  <david.sherwood@arm.com>
19352         * expr.c (expand_expr_real_2): Use scalar_int_mode for the
19353         double-word mode.
19354         * lower-subreg.c (resolve_shift_zext): Use is_a <scalar_int_mode>.
19355         * optabs.c (expand_unop): Likewise.
19357 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19358             Alan Hayward  <alan.hayward@arm.com>
19359             David Sherwood  <david.sherwood@arm.com>
19361         * dwarf2out.c (typed_binop): Change mode parameter to scalar_int_mode.
19362         (clz_loc_descriptor): Likewise.  Remove SCALAR_INT_MODE_P check.
19363         (popcount_loc_descriptor): Likewise.
19364         (bswap_loc_descriptor): Likewise.
19365         (rotate_loc_descriptor): Likewise.
19366         (mem_loc_descriptor): Add is_a <scalar_int_mode> checks before
19367         calling the functions above.
19369 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19370             Alan Hayward  <alan.hayward@arm.com>
19371             David Sherwood  <david.sherwood@arm.com>
19373         * combine.c (sign_extend_short_imm): Add is_a <scalar_int_mode>
19374         checks.
19375         (try_combine): Likewise.
19376         (simplify_if_then_else): Likewise.
19377         * cse.c (cse_insn): Likewise.
19378         * dwarf2out.c (mem_loc_descriptor): Likewise.
19379         * emit-rtl.c (gen_lowpart_common): Likewise.
19380         * simplify-rtx.c (simplify_truncation): Likewise.
19381         (simplify_binary_operation_1): Likewise.
19382         (simplify_const_relational_operation): Likewise.
19383         (simplify_ternary_operation): Likewise.
19384         * tree-ssa-loop-ivopts.c (force_expr_to_var_cost): Likewise.
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         * cse.c (cse_insn): Add is_a <scalar_int_mode> checks.
19391         * reload.c (push_reload): Likewise.
19392         (find_reloads): Likewise.
19394 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19395             Alan Hayward  <alan.hayward@arm.com>
19396             David Sherwood  <david.sherwood@arm.com>
19398         * combine.c (find_split_point): Add is_a <scalar_int_mode> checks.
19399         (make_compound_operation_int): Likewise.
19400         (change_zero_ext): Likewise.
19401         * expr.c (convert_move): Likewise.
19402         (convert_modes): Likewise.
19403         * fwprop.c (forward_propagate_subreg): Likewise.
19404         * loop-iv.c (get_biv_step_1): Likewise.
19405         * optabs.c (widen_operand): Likewise.
19406         * postreload.c (move2add_valid_value_p): Likewise.
19407         * recog.c (simplify_while_replacing): Likewise.
19408         * simplify-rtx.c (simplify_unary_operation_1): Likewise.
19409         (simplify_binary_operation_1): Likewise.  Remove redundant
19410         mode equality check.
19412 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19413             Alan Hayward  <alan.hayward@arm.com>
19414             David Sherwood  <david.sherwood@arm.com>
19416         * combine.c (combine_simplify_rtx): Add checks for
19417         is_a <scalar_int_mode>.
19418         (simplify_if_then_else): Likewise.
19419         (make_field_assignment): Likewise.
19420         (simplify_comparison): Likewise.
19421         * ifcvt.c (noce_try_bitop): Likewise.
19422         * loop-invariant.c (canonicalize_address_mult): Likewise.
19423         * simplify-rtx.c (simplify_unary_operation_1): Likewise.
19425 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19426             Alan Hayward  <alan.hayward@arm.com>
19427             David Sherwood  <david.sherwood@arm.com>
19429         * gimple-fold.c (gimple_fold_builtin_memory_op): Use
19430         is_a <scalar_int_mode> instead of != BLKmode.
19432 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19433             Alan Hayward  <alan.hayward@arm.com>
19434             David Sherwood  <david.sherwood@arm.com>
19436         * cfgexpand.c (expand_debug_expr): Use is_a <scalar_int_mode>
19437         instead of != VOIDmode.
19438         * combine.c (if_then_else_cond): Likewise.
19439         (change_zero_ext): Likewise.
19440         * dwarf2out.c (mem_loc_descriptor): Likewise.
19441         (loc_descriptor): Likewise.
19442         * rtlanal.c (canonicalize_condition): Likewise.
19443         * simplify-rtx.c (simplify_relational_operation_1): Likewise.
19445 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19446             Alan Hayward  <alan.hayward@arm.com>
19447             David Sherwood  <david.sherwood@arm.com>
19449         * simplify-rtx.c (simplify_binary_operation_1): Use
19450         is_a <scalar_int_mode> instead of !VECTOR_MODE_P.
19452 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19453             Alan Hayward  <alan.hayward@arm.com>
19454             David Sherwood  <david.sherwood@arm.com>
19456         * wide-int.h (int_traits<unsigned char>) New class.
19457         (int_traits<unsigned short>) Likewise.
19458         * cfgexpand.c (expand_debug_expr): Use is_a <scalar_int_mode>.
19459         Use GET_MODE_UNIT_PRECISION and remove redundant test for
19460         SCALAR_INT_MODE_P.
19461         * combine.c (set_nonzero_bits_and_sign_copies): Use
19462         is_a <scalar_int_mode>.
19463         (find_split_point): Likewise.
19464         (combine_simplify_rtx): Likewise.
19465         (simplify_logical): Likewise.
19466         (expand_compound_operation): Likewise.
19467         (expand_field_assignment): Likewise.
19468         (make_compound_operation): Likewise.
19469         (extended_count): Likewise.
19470         (change_zero_ext): Likewise.
19471         (simplify_comparison): Likewise.
19472         * dwarf2out.c (scompare_loc_descriptor): Likewise.
19473         (ucompare_loc_descriptor): Likewise.
19474         (minmax_loc_descriptor): Likewise.
19475         (mem_loc_descriptor): Likewise.
19476         (loc_descriptor): Likewise.
19477         * expmed.c (init_expmed_one_mode): Likewise.
19478         * lra-constraints.c (lra_constraint_offset): Likewise.
19479         * optabs.c (prepare_libcall_arg): Likewise.
19480         * postreload.c (move2add_note_store): Likewise.
19481         * reload.c (operands_match_p): Likewise.
19482         * rtl.h (load_extend_op): Likewise.
19483         * rtlhooks.c (gen_lowpart_general): Likewise.
19484         * simplify-rtx.c (simplify_truncation): Likewise.
19485         (simplify_unary_operation_1): Likewise.
19486         (simplify_binary_operation_1): Likewise.
19487         (simplify_const_binary_operation): Likewise.
19488         (simplify_const_relational_operation): Likewise.
19489         (simplify_subreg): Likewise.
19490         * stor-layout.c (bitwise_mode_for_mode): Likewise.
19491         * var-tracking.c (adjust_mems): Likewise.
19492         (prepare_call_arguments): Likewise.
19494 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19495             Alan Hayward  <alan.hayward@arm.com>
19496             David Sherwood  <david.sherwood@arm.com>
19498         * machmode.h (is_int_mode): New fuction.
19499         * combine.c (find_split_point): Use it.
19500         (combine_simplify_rtx): Likewise.
19501         (simplify_if_then_else): Likewise.
19502         (simplify_set): Likewise.
19503         (simplify_shift_const_1): Likewise.
19504         (simplify_comparison): Likewise.
19505         * config/aarch64/aarch64.c (aarch64_rtx_costs): Likewise.
19506         * cse.c (notreg_cost): Likewise.
19507         (cse_insn): Likewise.
19508         * cselib.c (cselib_lookup_1): Likewise.
19509         * dojump.c (do_jump_1): Likewise.
19510         (do_compare_rtx_and_jump): Likewise.
19511         * dse.c (get_call_args): Likewise.
19512         * dwarf2out.c (rtl_for_decl_init): Likewise.
19513         (native_encode_initializer): Likewise.
19514         * expmed.c (emit_store_flag_1): Likewise.
19515         (emit_store_flag): Likewise.
19516         * expr.c (convert_modes): Likewise.
19517         (store_field): Likewise.
19518         (expand_expr_real_1): Likewise.
19519         * fold-const.c (fold_read_from_constant_string): Likewise.
19520         * gimple-ssa-sprintf.c (get_format_string): Likewise.
19521         * optabs-libfuncs.c (gen_int_libfunc): Likewise.
19522         * optabs.c (expand_binop): Likewise.
19523         (expand_unop): Likewise.
19524         (expand_abs_nojump): Likewise.
19525         (expand_one_cmpl_abs_nojump): Likewise.
19526         * simplify-rtx.c (mode_signbit_p): Likewise.
19527         (val_signbit_p): Likewise.
19528         (val_signbit_known_set_p): Likewise.
19529         (val_signbit_known_clear_p): Likewise.
19530         (simplify_relational_operation_1): Likewise.
19531         * tree.c (vector_type_mode): Likewise.
19533 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19534             Alan Hayward  <alan.hayward@arm.com>
19535             David Sherwood  <david.sherwood@arm.com>
19537         * machmode.h (smallest_mode_for_size): Fix formatting.
19538         (smallest_int_mode_for_size): New function.
19539         * cfgexpand.c (expand_debug_expr): Use smallest_int_mode_for_size
19540         instead of smallest_mode_for_size.
19541         * combine.c (make_extraction): Likewise.
19542         * config/arc/arc.c (arc_expand_movmem): Likewise.
19543         * config/arm/arm.c (arm_expand_divmod_libfunc): Likewise.
19544         * config/i386/i386.c (ix86_get_mask_mode): Likewise.
19545         * config/s390/s390.c (s390_expand_insv): Likewise.
19546         * config/sparc/sparc.c (assign_int_registers): Likewise.
19547         * config/spu/spu.c (spu_function_value): Likewise.
19548         (spu_function_arg): Likewise.
19549         * coverage.c (get_gcov_type): Likewise.
19550         (get_gcov_unsigned_t): Likewise.
19551         * dse.c (find_shift_sequence): Likewise.
19552         * expmed.c (store_bit_field_1): Likewise.
19553         * expr.c (convert_move): Likewise.
19554         (store_field): Likewise.
19555         * internal-fn.c (expand_arith_overflow): Likewise.
19556         * optabs-query.c (get_best_extraction_insn): Likewise.
19557         * optabs.c (expand_twoval_binop_libfunc): Likewise.
19558         * stor-layout.c (layout_type): Likewise.
19559         (initialize_sizetypes): Likewise.
19560         * targhooks.c (default_get_mask_mode): Likewise.
19561         * tree-ssa-loop-manip.c (canonicalize_loop_ivs): Likewise.
19563 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19564             Alan Hayward  <alan.hayward@arm.com>
19565             David Sherwood  <david.sherwood@arm.com>
19567         * machmode.h (opt_mode::else_blk): New function.
19568         (int_mode_for_mode): Declare.
19569         * stor-layout.c (int_mode_for_mode): Return an opt_scalar_int_mode.
19570         * builtins.c (expand_builtin_signbit): Adjust for new int_mode_for_mode
19571         return type.
19572         * cfgexpand.c (expand_debug_expr): Likewise.
19573         * combine.c (gen_lowpart_or_truncate): Likewise.
19574         (gen_lowpart_for_combine): Likewise.
19575         * config/aarch64/aarch64.c (aarch64_emit_approx_sqrt): Likewise.
19576         * config/avr/avr.c (avr_to_int_mode): Likewise.
19577         (avr_out_plus_1): Likewise.
19578         (avr_out_plus): Likewise.
19579         (avr_out_round): Likewise.
19580         * config/i386/i386.c (ix86_split_to_parts): Likewise.
19581         * config/s390/s390.c (s390_expand_vec_compare_cc): Likewise.
19582         (s390_expand_vcond): Likewise.
19583         * config/spu/spu.c (spu_split_immediate): Likewise.
19584         (spu_expand_mov): Likewise.
19585         * dse.c (get_stored_val): Likewise.
19586         * expmed.c (store_bit_field_1): Likewise.
19587         (convert_extracted_bit_field): Use int_mode_for_mode instead of
19588         int_mode_for_size.
19589         (extract_bit_field_1): Adjust for new int_mode_for_mode return type.
19590         (extract_low_bits): Likewise.
19591         * expr.c (emit_group_load_1): Likewise.  Separate out the BLKmode
19592         handling rather than repeating the check.
19593         (emit_group_store): Likewise.
19594         (emit_move_via_integer): Adjust for new int_mode_for_mode return type.
19595         * optabs.c (expand_absneg_bit): Likewise.
19596         (expand_copysign_absneg): Likewise.
19597         (expand_copysign_bit): Likewise.
19598         * tree-if-conv.c (ifcvt_can_use_mask_load_store): Likewise.
19599         * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
19600         * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
19601         * var-tracking.c (prepare_call_arguments):  Likewise.
19602         * config/powerpcspe/powerpcspe.c (rs6000_do_expand_vec_perm): Use
19603         int_mode_for_mode instead of mode_for_size.
19604         * config/rs6000/rs6000.c (rs6000_do_expand_vec_perm): Likewise.
19606 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19607             Alan Hayward  <alan.hayward@arm.com>
19608             David Sherwood  <david.sherwood@arm.com>
19610         * machmode.h (int_mode_for_size): New function.
19611         * builtins.c (set_builtin_user_assembler_name): Use int_mode_for_size
19612         instead of mode_for_size.
19613         * calls.c (save_fixed_argument_area): Likewise.  Make use of BLKmode
19614         explicit.
19615         * combine.c (expand_field_assignment): Use int_mode_for_size
19616         instead of mode_for_size.
19617         (make_extraction): Likewise.
19618         (simplify_shift_const_1): Likewise.
19619         (simplify_comparison): Likewise.
19620         * dojump.c (do_jump): Likewise.
19621         * dwarf2out.c (mem_loc_descriptor): Likewise.
19622         * emit-rtl.c (init_derived_machine_modes): Likewise.
19623         * expmed.c (flip_storage_order): Likewise.
19624         (convert_extracted_bit_field): Likewise.
19625         * expr.c (copy_blkmode_from_reg): Likewise.
19626         * graphite-isl-ast-to-gimple.c (max_mode_int_precision): Likewise.
19627         * internal-fn.c (expand_mul_overflow): Likewise.
19628         * lower-subreg.c (simple_move): Likewise.
19629         * optabs-libfuncs.c (init_optabs): Likewise.
19630         * simplify-rtx.c (simplify_unary_operation_1): Likewise.
19631         * tree.c (vector_type_mode): Likewise.
19632         * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
19633         * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Likewise.
19634         * tree-vect-generic.c (expand_vector_parallel): Likewise.
19635         * tree-vect-stmts.c (vectorizable_load): Likewise.
19636         (vectorizable_store): Likewise.
19638 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19639             Alan Hayward  <alan.hayward@arm.com>
19640             David Sherwood  <david.sherwood@arm.com>
19642         * coretypes.h (pod_mode): New type.
19643         (scalar_int_mode_pod): New typedef.
19644         * machmode.h (pod_mode): New class.
19645         (int_n_data_t::m): Change type to scalar_int_mode_pod.
19646         * genmodes.c (emit_mode_int_n): Update accordingly.
19647         * lower-subreg.h (target_lower_subreg): Change type to
19648         scalar_int_mode_pod.
19649         * gdbhooks.py (build_pretty_printer): Handle pod_mode and
19650         scalar_int_mode_pod.
19652 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19653             Alan Hayward  <alan.hayward@arm.com>
19654             David Sherwood  <david.sherwood@arm.com>
19656         * config/powerpcspe/powerpcspe.h (rs6000_pmode): Change type from
19657         machine_mode to scalar_int_mode.
19658         * config/powerpcspe/powerpcspe.c (rs6000_pmode): Likewise.
19659         (rs6000_option_override_internal): Remove cast to int.
19660         * config/rs6000/rs6000.h (rs6000_pmode): Change type from
19661         machine_mode to scalar_int_mode.
19662         * config/rs6000/rs6000.c (rs6000_pmode): Likewise.
19663         (rs6000_option_override_internal): Remove cast to int.
19664         * config/s390/s390.h (Pmode): Remove cast to machine_mode.
19665         * config/epiphany/epiphany.h (RTX_OK_FOR_OFFSET_P): Add cast
19666         to machine_mode.
19667         * config/s390/s390.c (s390_expand_builtin): Likewise.
19668         * coretypes.h (scalar_int_mode): New type.
19669         (opt_scalar_int_mode): New typedef.
19670         * machmode.h (scalar_int_mode): New class.
19671         (scalar_int_mode::includes_p): New function.
19672         (byte_mode): Change type to scalar_int_mode.
19673         (word_mode): Likewise.
19674         (ptr_mode): Likewise.
19675         * emit-rtl.c (byte_mode): Likewise.
19676         (word_mode): Likewise.
19677         (ptr_mode): Likewise.
19678         (init_derived_machine_modes): Update accordingly.
19679         * genmodes.c (get_mode_class): Return scalar_int_mode for MODE_INT
19680         and MODE_PARTIAL_INT.
19681         * gdbhooks.py (build_pretty_printer): Handle scalar_int_mode and
19682         opt_scalar_int_mode.
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         * target.def (libgcc_floating_mode_supported_p): Take a
19689         scalar_float_mode.
19690         * doc/tm.texi: Regenerate.
19691         * targhooks.h (default_libgcc_floating_mode_supported_p): Take a
19692         scalar_float_mode.
19693         * targhooks.c (default_libgcc_floating_mode_supported_p): Likewise.
19694         * config/aarch64/aarch64.c (aarch64_libgcc_floating_mode_supported_p):
19695         Likewise.
19697 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19698             Alan Hayward  <alan.hayward@arm.com>
19699             David Sherwood  <david.sherwood@arm.com>
19701         * target.def (default_floatn_mode): Return an opt_scalar_float_mode.
19702         * doc/tm.texi: Regenerate.
19703         * config/arm/arm.c (arm_floatn_mode): Return an opt_scalar_float_mode.
19704         * config/powerpcspe/powerpcspe.c (rs6000_floatn_mode): Likewise.
19705         * config/rs6000/rs6000.c (rs6000_floatn_mode): Likewise.
19706         * targhooks.h (default_floatn_mode): Likewise.
19707         * targhooks.c (default_floatn_mode): Likewise.
19708         * tree.c (build_common_tree_nodes): Update accordingly.
19710 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19711             Alan Hayward  <alan.hayward@arm.com>
19712             David Sherwood  <david.sherwood@arm.com>
19714         * machmode.h (mode_iterator::start): Provide overload for opt_modes.
19715         (mode_iterator::iterate_p): Likewise.
19716         (mode_iterator::get_wider): Likewise.
19717         * expr.c (init_expr_target): Use opt_scalar_float_mode.
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         * coretypes.h (opt_scalar_float_mode): New typedef.
19724         * machmode.h (float_mode_for_size): New function.
19725         * emit-rtl.c (double_mode): Delete.
19726         (init_emit_once): Use float_mode_for_size.
19727         * stor-layout.c (layout_type): Likewise.
19728         * gdbhooks.py (build_pretty_printer): Handle opt_scalar_float_mode.
19730 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19731             Alan Hayward  <alan.hayward@arm.com>
19732             David Sherwood  <david.sherwood@arm.com>
19734         * output.h (assemble_real): Take a scalar_float_mode.
19735         * config/arm/arm.c (arm_assemble_integer): Update accordingly.
19736         * config/arm/arm.md (consttable_4): Likewise.
19737         (consttable_8): Likewise.
19738         (consttable_16): Likewise.
19739         * config/mips/mips.md (consttable_float): Likewise.
19740         * config/s390/s390.c (s390_output_pool_entry): Likewise.
19741         * varasm.c (assemble_real): Take a scalar_float_mode.
19742         (output_constant_pool_2): Update accordingly.
19743         (output_constant): Likewise.
19745 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19746             Alan Hayward  <alan.hayward@arm.com>
19747             David Sherwood  <david.sherwood@arm.com>
19749         * tree.h (SCALAR_FLOAT_TYPE_MODE): New macro.
19750         * builtins.c (expand_builtin_signbit): Use it instead of TYPE_MODE.
19751         * fold-const.c (fold_convert_const_real_from_fixed): Likewise.
19752         (native_encode_real): Likewise.
19753         (native_interpret_real): Likewise.
19754         * hsa-brig.c (emit_immediate_scalar_to_buffer): Likewise.
19755         * tree-vrp.c (simplify_float_conversion_using_ranges): Likewise.
19757 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19758             Alan Hayward  <alan.hayward@arm.com>
19759             David Sherwood  <david.sherwood@arm.com>
19761         * optabs-libfuncs.c (gen_trunc_conv_libfunc): Use is_a
19762         <scalar_float_mode>.  Simplify.
19763         (gen_extend_conv_libfunc): Likewise.
19765 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19766             Alan Hayward  <alan.hayward@arm.com>
19767             David Sherwood  <david.sherwood@arm.com>
19769         * coretypes.h (scalar_float_mode): New type.
19770         * machmode.h (mode_traits::from_int): Use machine_mode if
19771         USE_ENUM_MODES is defined.
19772         (is_a): New function.
19773         (as_a): Likewise.
19774         (dyn_cast): Likewise.
19775         (scalar_float_mode): New class.
19776         (scalar_float_mode::includes_p): New function.
19777         (is_float_mode): Likewise.
19778         * gdbhooks.py (MachineModePrinter): New class.
19779         (build_pretty_printer): Use it for scalar_float_mode.
19780         * real.h (FLOAT_MODE_FORMAT): Use as_a <scalar_float_mode>.
19781         (format_helper::format_helper): Turn into a template.
19782         * genmodes.c (get_mode_class): New function.
19783         (emit_insn_modes_h): Give modes the class returned by get_mode_class,
19784         or machine_mode if none.
19785         * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Use
19786         as_a <scalar_float_mode>.
19787         * dwarf2out.c (mem_loc_descriptor): Likewise.
19788         (insert_float): Likewise.
19789         (add_const_value_attribute): Likewise.
19790         * simplify-rtx.c (simplify_immed_subreg): Likewise.
19791         * optabs.c (expand_absneg_bit): Take a scalar_float_mode.
19792         (expand_unop): Update accordingly.
19793         (expand_abs_nojump): Likewise.
19794         (expand_copysign_absneg): Take a scalar_float_mode.
19795         (expand_copysign_bit): Likewise.
19796         (expand_copysign): Update accordingly.
19798 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19799             Alan Hayward  <alan.hayward@arm.com>
19800             David Sherwood  <david.sherwood@arm.com>
19802         * coretypes.h (opt_mode): New class.
19803         * machmode.h (opt_mode): Likewise.
19804         (opt_mode::else_void): New function.
19805         (opt_mode::require): Likewise.
19806         (opt_mode::exists): Likewise.
19807         (GET_MODE_WIDER_MODE): Turn into a function and return an opt_mode.
19808         (GET_MODE_2XWIDER_MODE): Likewise.
19809         (mode_iterator::get_wider): Update accordingly.
19810         (mode_iterator::get_2xwider): Likewise.
19811         (mode_iterator::get_known_wider): Likewise, turning into a template.
19812         * combine.c (make_extraction): Update use of GET_MODE_WIDER_MODE,
19813         forcing a wider mode to exist.
19814         * config/cr16/cr16.h (LONG_REG_P): Likewise.
19815         * rtlanal.c (init_num_sign_bit_copies_in_rep): Likewise.
19816         * config/c6x/c6x.c (c6x_rtx_costs): Update use of
19817         GET_MODE_2XWIDER_MODE, forcing a wider mode to exist.
19818         * lower-subreg.c (init_lower_subreg): Likewise.
19819         * optabs-libfuncs.c (init_sync_libfuncs_1): Likewise, but not
19820         on the final iteration.
19821         * config/i386/i386.c (ix86_expand_set_or_movmem): Check whether
19822         a wider mode exists before asking for a move pattern.
19823         (get_mode_wider_vector): Update use of GET_MODE_WIDER_MODE,
19824         forcing a wider mode to exist.
19825         (expand_vselect_vconcat): Update use of GET_MODE_2XWIDER_MODE,
19826         returning false if no such mode exists.
19827         * config/ia64/ia64.c (expand_vselect_vconcat): Likewise.
19828         * config/mips/mips.c (mips_expand_vselect_vconcat): Likewise.
19829         * expmed.c (init_expmed_one_mode): Update use of GET_MODE_WIDER_MODE.
19830         Avoid checking for a MODE_INT if we already know the mode is not a
19831         SCALAR_INT_MODE_P.
19832         (extract_high_half): Update use of GET_MODE_WIDER_MODE,
19833         forcing a wider mode to exist.
19834         (expmed_mult_highpart_optab): Likewise.
19835         (expmed_mult_highpart): Likewise.
19836         * expr.c (expand_expr_real_2): Update use of GET_MODE_WIDER_MODE,
19837         using else_void.
19838         * lto-streamer-in.c (lto_input_mode_table): Likewise.
19839         * optabs-query.c (find_widening_optab_handler_and_mode): Likewise.
19840         * stor-layout.c (bit_field_mode_iterator::next_mode): Likewise.
19841         * internal-fn.c (expand_mul_overflow): Update use of
19842         GET_MODE_2XWIDER_MODE.
19843         * omp-low.c (omp_clause_aligned_alignment): Likewise.
19844         * tree-ssa-math-opts.c (convert_mult_to_widen): Update use of
19845         GET_MODE_WIDER_MODE.
19846         (convert_plusminus_to_widen): Likewise.
19847         * tree-switch-conversion.c (array_value_type): Likewise.
19848         * var-tracking.c (emit_note_insn_var_location): Likewise.
19849         * tree-vrp.c (simplify_float_conversion_using_ranges): Likewise.
19850         Return false inside rather than outside the loop if no wider mode
19851         exists
19852         * optabs.c (expand_binop): Update use of GET_MODE_WIDER_MODE
19853         and GET_MODE_2XWIDER_MODE
19854         (can_compare_p): Use else_void.
19855         * gdbhooks.py (OptMachineModePrinter): New class.
19856         (build_pretty_printer): Use it for opt_mode.
19858 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19859             Alan Hayward  <alan.hayward@arm.com>
19860             David Sherwood  <david.sherwood@arm.com>
19862         * tree-switch-conversion.c (array_value_type): Only read TYPE_MODE
19863         once.  Use get_narrowest_mode instead of GET_CLASS_NARROWEST_MODE.
19865 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19866             Alan Hayward  <alan.hayward@arm.com>
19867             David Sherwood  <david.sherwood@arm.com>
19869         * machmode.h (mode_traits): New structure.
19870         (get_narrowest_mode): New function.
19871         (mode_iterator::start): Likewise.
19872         (mode_iterator::iterate_p): Likewise.
19873         (mode_iterator::get_wider): Likewise.
19874         (mode_iterator::get_known_wider): Likewise.
19875         (mode_iterator::get_2xwider): Likewise.
19876         (FOR_EACH_MODE_IN_CLASS): New mode iterator.
19877         (FOR_EACH_MODE): Likewise.
19878         (FOR_EACH_MODE_FROM): Likewise.
19879         (FOR_EACH_MODE_UNTIL): Likewise.
19880         (FOR_EACH_WIDER_MODE): Likewise.
19881         (FOR_EACH_2XWIDER_MODE): Likewise.
19882         * builtins.c (expand_builtin_strlen): Use new mode iterators.
19883         * combine.c (simplify_comparison): Likewise
19884         * config/i386/i386.c (type_natural_mode): Likewise.
19885         * cse.c (cse_insn): Likewise.
19886         * dse.c (find_shift_sequence): Likewise.
19887         * emit-rtl.c (init_derived_machine_modes): Likewise.
19888         (init_emit_once): Likewise.
19889         * explow.c (hard_function_value): Likewise.
19890         * expmed.c (extract_fixed_bit_field_1): Likewise.
19891         (extract_bit_field_1): Likewise.
19892         (expand_divmod): Likewise.
19893         (emit_store_flag_1): Likewise.
19894         * expr.c (init_expr_target): Likewise.
19895         (convert_move): Likewise.
19896         (alignment_for_piecewise_move): Likewise.
19897         (widest_int_mode_for_size): Likewise.
19898         (emit_block_move_via_movmem): Likewise.
19899         (copy_blkmode_to_reg): Likewise.
19900         (set_storage_via_setmem): Likewise.
19901         (compress_float_constant): Likewise.
19902         * omp-low.c (omp_clause_aligned_alignment): Likewise.
19903         * optabs-query.c (get_best_extraction_insn): Likewise.
19904         * optabs.c (expand_binop): Likewise.
19905         (expand_twoval_unop): Likewise.
19906         (expand_twoval_binop): Likewise.
19907         (widen_leading): Likewise.
19908         (widen_bswap): Likewise.
19909         (expand_parity): Likewise.
19910         (expand_unop): Likewise.
19911         (prepare_cmp_insn): Likewise.
19912         (prepare_float_lib_cmp): Likewise.
19913         (expand_float): Likewise.
19914         (expand_fix): Likewise.
19915         (expand_sfix_optab): Likewise.
19916         * postreload.c (move2add_use_add2_insn): Likewise.
19917         * reg-stack.c (reg_to_stack): Likewise.
19918         * reginfo.c (choose_hard_reg_mode): Likewise.
19919         * rtlanal.c (init_num_sign_bit_copies_in_rep): Likewise.
19920         * stor-layout.c (mode_for_size): Likewise.
19921         (smallest_mode_for_size): Likewise.
19922         (mode_for_vector): Likewise.
19923         (finish_bitfield_representative): Likewise.
19924         * tree-ssa-math-opts.c (target_supports_divmod_p): Likewise.
19925         * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
19926         * tree-vect-stmts.c (vectorizable_conversion): Likewise.
19927         * var-tracking.c (prepare_call_arguments): Likewise.
19929 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19930             Alan Hayward  <alan.hayward@arm.com>
19931             David Sherwood  <david.sherwood@arm.com>
19933         * genconditions.c (write_header): Add a "#define USE_ENUM_MODES".
19934         * genmodes.c (emit_insn_modes_h): Define FOOmode to E_FOOmode if
19935         USE_ENUM_MODES is defined and to ((void) 0, E_FOOmode) otherwise.
19936         * machmode.h (mode_size): Move earlier in file.
19937         (mode_precision): Likewise.
19938         (mode_inner): Likewise.
19939         (mode_nunits): Likewise.
19940         (mode_unit_size): Likewise.
19941         (unit_unit_precision): Likewise.
19942         (mode_wider): Likewise.
19943         (mode_2xwider): Likewise.
19944         (machine_mode): New class.
19945         (mode_to_bytes): New function.
19946         (mode_to_bits): Likewise.
19947         (mode_to_precision): Likewise.
19948         (mode_to_inner): Likewise.
19949         (mode_to_unit_size): Likewise.
19950         (mode_to_unit_precision): Likewise.
19951         (mode_to_nunits): Likewise.
19952         (GET_MODE_SIZE): Use mode_to_bytes.
19953         (GET_MODE_BITSIZE): Use mode_to_bits.
19954         (GET_MODE_PRECISION): Use mode_to_precision.
19955         (GET_MODE_INNER): Use mode_to_inner.
19956         (GET_MODE_UNIT_SIZE): Use mode_to_unit_size.
19957         (GET_MODE_UNIT_PRECISION): Use mode_to_unit_precision.
19958         (GET_MODE_NUNITS): Use mode_to_nunits.
19959         * system.h (ALWAYS_INLINE): New macro.
19960         * config/powerpcspe/powerpcspe-c.c
19961         (altivec_resolve_overloaded_builtin): Use machine_mode instead of
19962         int for arg1_mode and arg2_mode.
19964 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19965             Alan Hayward  <alan.hayward@arm.com>
19966             David Sherwood  <david.sherwood@arm.com>
19968         * config/aarch64/aarch64-builtins.c (aarch64_simd_builtin_std_type):
19969         Prefix mode names with E_ in case statements.
19970         * config/aarch64/aarch64-elf.h (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
19971         * config/aarch64/aarch64.c (aarch64_split_simd_combine): Likewise.
19972         (aarch64_split_simd_move): Likewise.
19973         (aarch64_gen_storewb_pair): Likewise.
19974         (aarch64_gen_loadwb_pair): Likewise.
19975         (aarch64_gen_store_pair): Likewise.
19976         (aarch64_gen_load_pair): Likewise.
19977         (aarch64_get_condition_code_1): Likewise.
19978         (aarch64_constant_pool_reload_icode): Likewise.
19979         (get_rsqrte_type): Likewise.
19980         (get_rsqrts_type): Likewise.
19981         (get_recpe_type): Likewise.
19982         (get_recps_type): Likewise.
19983         (aarch64_gimplify_va_arg_expr): Likewise.
19984         (aarch64_simd_container_mode): Likewise.
19985         (aarch64_emit_load_exclusive): Likewise.
19986         (aarch64_emit_store_exclusive): Likewise.
19987         (aarch64_expand_compare_and_swap): Likewise.
19988         (aarch64_gen_atomic_cas): Likewise.
19989         (aarch64_emit_bic): Likewise.
19990         (aarch64_emit_atomic_swap): Likewise.
19991         (aarch64_emit_atomic_load_op): Likewise.
19992         (aarch64_evpc_trn): Likewise.
19993         (aarch64_evpc_uzp): Likewise.
19994         (aarch64_evpc_zip): Likewise.
19995         (aarch64_evpc_ext): Likewise.
19996         (aarch64_evpc_rev): Likewise.
19997         (aarch64_evpc_dup): Likewise.
19998         (aarch64_gen_ccmp_first): Likewise.
19999         (aarch64_gen_ccmp_next): Likewise.
20000         * config/alpha/alpha.c (alpha_scalar_mode_supported_p): Likewise.
20001         (alpha_emit_xfloating_libcall): Likewise.
20002         (emit_insxl): Likewise.
20003         (alpha_arg_type): Likewise.
20004         * config/arc/arc.c (arc_vector_mode_supported_p): Likewise.
20005         (arc_preferred_simd_mode): Likewise.
20006         (arc_secondary_reload): Likewise.
20007         (get_arc_condition_code): Likewise.
20008         (arc_print_operand): Likewise.
20009         (arc_legitimate_constant_p): Likewise.
20010         * config/arc/arc.h (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
20011         * config/arc/arc.md (casesi_load): Likewise.
20012         (casesi_compact_jump): Likewise.
20013         * config/arc/predicates.md (proper_comparison_operator): Likewise.
20014         (cc_use_register): Likewise.
20015         * config/arm/aout.h (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
20016         * config/arm/arm-builtins.c (arm_simd_builtin_std_type): Likewise.
20017         (arm_init_iwmmxt_builtins): Likewise.
20018         * config/arm/arm.c (thumb1_size_rtx_costs): Likewise.
20019         (neon_expand_vector_init): Likewise.
20020         (arm_attr_length_move_neon): Likewise.
20021         (maybe_get_arm_condition_code): Likewise.
20022         (arm_emit_vector_const): Likewise.
20023         (arm_preferred_simd_mode): Likewise.
20024         (arm_output_iwmmxt_tinsr): Likewise.
20025         (thumb1_output_casesi): Likewise.
20026         (thumb2_output_casesi): Likewise.
20027         (arm_emit_load_exclusive): Likewise.
20028         (arm_emit_store_exclusive): Likewise.
20029         (arm_expand_compare_and_swap): Likewise.
20030         (arm_evpc_neon_vuzp): Likewise.
20031         (arm_evpc_neon_vzip): Likewise.
20032         (arm_evpc_neon_vrev): Likewise.
20033         (arm_evpc_neon_vtrn): Likewise.
20034         (arm_evpc_neon_vext): Likewise.
20035         (arm_validize_comparison): Likewise.
20036         * config/arm/neon.md (neon_vc<cmp_op><mode>): Likewise.
20037         * config/avr/avr-c.c (avr_resolve_overloaded_builtin): Likewise.
20038         * config/avr/avr.c (avr_rtx_costs_1): Likewise.
20039         * config/c6x/c6x.c (c6x_vector_mode_supported_p): Likewise.
20040         (c6x_preferred_simd_mode): Likewise.
20041         * config/epiphany/epiphany.c (get_epiphany_condition_code): Likewise.
20042         (epiphany_rtx_costs): Likewise.
20043         * config/epiphany/predicates.md (proper_comparison_operator):
20044         Likewise.
20045         * config/frv/frv.c (condexec_memory_operand): Likewise.
20046         (frv_emit_move): Likewise.
20047         (output_move_single): Likewise.
20048         (output_condmove_single): Likewise.
20049         (frv_hard_regno_mode_ok): Likewise.
20050         (frv_matching_accg_mode): Likewise.
20051         * config/h8300/h8300.c (split_adds_subs): Likewise.
20052         (h8300_rtx_costs): Likewise.
20053         (h8300_print_operand): Likewise.
20054         (compute_mov_length): Likewise.
20055         (output_logical_op): Likewise.
20056         (compute_logical_op_length): Likewise.
20057         (compute_logical_op_cc): Likewise.
20058         (h8300_shift_needs_scratch_p): Likewise.
20059         (output_a_shift): Likewise.
20060         (compute_a_shift_length): Likewise.
20061         (compute_a_shift_cc): Likewise.
20062         (expand_a_rotate): Likewise.
20063         (output_a_rotate): Likewise.
20064         * config/i386/i386.c (classify_argument): Likewise.
20065         (function_arg_advance_32): Likewise.
20066         (function_arg_32): Likewise.
20067         (function_arg_64): Likewise.
20068         (function_value_64): Likewise.
20069         (ix86_gimplify_va_arg): Likewise.
20070         (ix86_legitimate_constant_p): Likewise.
20071         (put_condition_code): Likewise.
20072         (split_double_mode): Likewise.
20073         (ix86_avx256_split_vector_move_misalign): Likewise.
20074         (ix86_expand_vector_logical_operator): Likewise.
20075         (ix86_split_idivmod): Likewise.
20076         (ix86_expand_adjust_ufix_to_sfix_si): Likewise.
20077         (ix86_build_const_vector): Likewise.
20078         (ix86_build_signbit_mask): Likewise.
20079         (ix86_match_ccmode): Likewise.
20080         (ix86_cc_modes_compatible): Likewise.
20081         (ix86_expand_branch): Likewise.
20082         (ix86_expand_sse_cmp): Likewise.
20083         (ix86_expand_sse_movcc): Likewise.
20084         (ix86_expand_int_sse_cmp): Likewise.
20085         (ix86_expand_vec_perm_vpermi2): Likewise.
20086         (ix86_expand_vec_perm): Likewise.
20087         (ix86_expand_sse_unpack): Likewise.
20088         (ix86_expand_int_addcc): Likewise.
20089         (ix86_split_to_parts): Likewise.
20090         (ix86_vectorize_builtin_gather): Likewise.
20091         (ix86_vectorize_builtin_scatter): Likewise.
20092         (avx_vpermilp_parallel): Likewise.
20093         (inline_memory_move_cost): Likewise.
20094         (ix86_tieable_integer_mode_p): Likewise.
20095         (x86_maybe_negate_const_int): Likewise.
20096         (ix86_expand_vector_init_duplicate): Likewise.
20097         (ix86_expand_vector_init_one_nonzero): Likewise.
20098         (ix86_expand_vector_init_one_var): Likewise.
20099         (ix86_expand_vector_init_concat): Likewise.
20100         (ix86_expand_vector_init_interleave): Likewise.
20101         (ix86_expand_vector_init_general): Likewise.
20102         (ix86_expand_vector_set): Likewise.
20103         (ix86_expand_vector_extract): Likewise.
20104         (emit_reduc_half): Likewise.
20105         (ix86_emit_i387_round): Likewise.
20106         (ix86_mangle_type): Likewise.
20107         (ix86_expand_round_sse4): Likewise.
20108         (expand_vec_perm_blend): Likewise.
20109         (canonicalize_vector_int_perm): Likewise.
20110         (ix86_expand_vec_one_operand_perm_avx512): Likewise.
20111         (expand_vec_perm_1): Likewise.
20112         (expand_vec_perm_interleave3): Likewise.
20113         (expand_vec_perm_even_odd_pack): Likewise.
20114         (expand_vec_perm_even_odd_1): Likewise.
20115         (expand_vec_perm_broadcast_1): Likewise.
20116         (ix86_vectorize_vec_perm_const_ok): Likewise.
20117         (ix86_expand_vecop_qihi): Likewise.
20118         (ix86_expand_mul_widen_hilo): Likewise.
20119         (ix86_expand_sse2_abs): Likewise.
20120         (ix86_expand_pextr): Likewise.
20121         (ix86_expand_pinsr): Likewise.
20122         (ix86_preferred_simd_mode): Likewise.
20123         (ix86_simd_clone_compute_vecsize_and_simdlen): Likewise.
20124         * config/i386/sse.md (*andnot<mode>3): Likewise.
20125         (<mask_codefor><code><mode>3<mask_name>): Likewise.
20126         (*<code><mode>3): Likewise.
20127         * config/ia64/ia64.c (ia64_expand_vecint_compare): Likewise.
20128         (ia64_expand_atomic_op): Likewise.
20129         (ia64_arg_type): Likewise.
20130         (ia64_mode_to_int): Likewise.
20131         (ia64_scalar_mode_supported_p): Likewise.
20132         (ia64_vector_mode_supported_p): Likewise.
20133         (expand_vec_perm_broadcast): Likewise.
20134         * config/iq2000/iq2000.c (iq2000_move_1word): Likewise.
20135         (iq2000_function_arg_advance): Likewise.
20136         (iq2000_function_arg): Likewise.
20137         * config/m32c/m32c.c (m32c_preferred_reload_class): Likewise.
20138         * config/m68k/m68k.c (output_dbcc_and_branch): Likewise.
20139         (m68k_libcall_value): Likewise.
20140         (m68k_function_value): Likewise.
20141         (sched_attr_op_type): Likewise.
20142         * config/mcore/mcore.c (mcore_output_move): Likewise.
20143         * config/microblaze/microblaze.c (microblaze_function_arg_advance):
20144         Likewise.
20145         (microblaze_function_arg): Likewise.
20146         * config/mips/mips.c (mips16_build_call_stub): Likewise.
20147         (mips_print_operand): Likewise.
20148         (mips_mode_ok_for_mov_fmt_p): Likewise.
20149         (mips_vector_mode_supported_p): Likewise.
20150         (mips_preferred_simd_mode): Likewise.
20151         (mips_expand_vpc_loongson_even_odd): Likewise.
20152         (mips_expand_vec_unpack): Likewise.
20153         (mips_expand_vi_broadcast): Likewise.
20154         (mips_expand_vector_init): Likewise.
20155         (mips_expand_vec_reduc): Likewise.
20156         (mips_expand_msa_cmp): Likewise.
20157         * config/mips/mips.md (casesi_internal_mips16_<mode>): Likewise.
20158         * config/mn10300/mn10300.c (mn10300_print_operand): Likewise.
20159         (cc_flags_for_mode): Likewise.
20160         * config/msp430/msp430.c (msp430_print_operand): Likewise.
20161         * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Likewise.
20162         (nds32_output_casesi_pc_relative): Likewise.
20163         * config/nds32/nds32.h (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
20164         * config/nvptx/nvptx.c (nvptx_ptx_type_from_mode): Likewise.
20165         (nvptx_gen_unpack): Likewise.
20166         (nvptx_gen_pack): Likewise.
20167         (nvptx_gen_shuffle): Likewise.
20168         (nvptx_gen_wcast): Likewise.
20169         (nvptx_preferred_simd_mode): Likewise.
20170         * config/pa/pa.c (pa_secondary_reload): Likewise.
20171         * config/pa/predicates.md (base14_operand): Likewise.
20172         * config/powerpcspe/powerpcspe-c.c
20173         (altivec_resolve_overloaded_builtin): Likewise.
20174         * config/powerpcspe/powerpcspe.c (rs6000_setup_reg_addr_masks):
20175         Likewise.
20176         (rs6000_preferred_simd_mode): Likewise.
20177         (output_vec_const_move): Likewise.
20178         (rs6000_expand_vector_extract): Likewise.
20179         (rs6000_split_vec_extract_var): Likewise.
20180         (reg_offset_addressing_ok_p): Likewise.
20181         (rs6000_legitimate_offset_address_p): Likewise.
20182         (rs6000_legitimize_address): Likewise.
20183         (rs6000_emit_set_const): Likewise.
20184         (rs6000_const_vec): Likewise.
20185         (rs6000_emit_move): Likewise.
20186         (spe_build_register_parallel): Likewise.
20187         (rs6000_darwin64_record_arg_recurse): Likewise.
20188         (swap_selector_for_mode): Likewise.
20189         (spe_init_builtins): Likewise.
20190         (paired_init_builtins): Likewise.
20191         (altivec_init_builtins): Likewise.
20192         (do_load_for_compare): Likewise.
20193         (rs6000_generate_compare): Likewise.
20194         (rs6000_expand_float128_convert): Likewise.
20195         (emit_load_locked): Likewise.
20196         (emit_store_conditional): Likewise.
20197         (rs6000_output_function_epilogue): Likewise.
20198         (rs6000_handle_altivec_attribute): Likewise.
20199         (rs6000_function_value): Likewise.
20200         (emit_fusion_gpr_load): Likewise.
20201         (emit_fusion_p9_load): Likewise.
20202         (emit_fusion_p9_store): Likewise.
20203         * config/powerpcspe/predicates.md (easy_fp_constant): Likewise.
20204         (fusion_gpr_mem_load): Likewise.
20205         (fusion_addis_mem_combo_load): Likewise.
20206         (fusion_addis_mem_combo_store): Likewise.
20207         * config/rs6000/predicates.md (easy_fp_constant): Likewise.
20208         (fusion_gpr_mem_load): Likewise.
20209         (fusion_addis_mem_combo_load): Likewise.
20210         (fusion_addis_mem_combo_store): Likewise.
20211         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
20212         Likewise.
20213         * config/rs6000/rs6000-string.c (do_load_for_compare): Likewise.
20214         * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): Likewise.
20215         (rs6000_preferred_simd_mode): Likewise.
20216         (output_vec_const_move): Likewise.
20217         (rs6000_expand_vector_extract): Likewise.
20218         (rs6000_split_vec_extract_var): Likewise.
20219         (reg_offset_addressing_ok_p): Likewise.
20220         (rs6000_legitimate_offset_address_p): Likewise.
20221         (rs6000_legitimize_address): Likewise.
20222         (rs6000_emit_set_const): Likewise.
20223         (rs6000_const_vec): Likewise.
20224         (rs6000_emit_move): Likewise.
20225         (rs6000_darwin64_record_arg_recurse): Likewise.
20226         (swap_selector_for_mode): Likewise.
20227         (paired_init_builtins): Likewise.
20228         (altivec_init_builtins): Likewise.
20229         (rs6000_expand_float128_convert): Likewise.
20230         (emit_load_locked): Likewise.
20231         (emit_store_conditional): Likewise.
20232         (rs6000_output_function_epilogue): Likewise.
20233         (rs6000_handle_altivec_attribute): Likewise.
20234         (rs6000_function_value): Likewise.
20235         (emit_fusion_gpr_load): Likewise.
20236         (emit_fusion_p9_load): Likewise.
20237         (emit_fusion_p9_store): Likewise.
20238         * config/rx/rx.c (rx_gen_move_template): Likewise.
20239         (flags_from_mode): Likewise.
20240         * config/s390/predicates.md (s390_alc_comparison): Likewise.
20241         (s390_slb_comparison): Likewise.
20242         * config/s390/s390.c (s390_handle_vectorbool_attribute): Likewise.
20243         (s390_vector_mode_supported_p): Likewise.
20244         (s390_cc_modes_compatible): Likewise.
20245         (s390_match_ccmode_set): Likewise.
20246         (s390_canonicalize_comparison): Likewise.
20247         (s390_emit_compare_and_swap): Likewise.
20248         (s390_branch_condition_mask): Likewise.
20249         (s390_rtx_costs): Likewise.
20250         (s390_secondary_reload): Likewise.
20251         (__SECONDARY_RELOAD_CASE): Likewise.
20252         (s390_expand_cs): Likewise.
20253         (s390_preferred_simd_mode): Likewise.
20254         * config/s390/vx-builtins.md (vec_packsu_u<mode>): Likewise.
20255         * config/sh/sh.c (sh_print_operand): Likewise.
20256         (dump_table): Likewise.
20257         (sh_secondary_reload): Likewise.
20258         * config/sh/sh.h (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
20259         * config/sh/sh.md (casesi_worker_1): Likewise.
20260         (casesi_worker_2): Likewise.
20261         * config/sparc/predicates.md (icc_comparison_operator): Likewise.
20262         (fcc_comparison_operator): Likewise.
20263         * config/sparc/sparc.c (sparc_expand_move): Likewise.
20264         (emit_soft_tfmode_cvt): Likewise.
20265         (sparc_preferred_simd_mode): Likewise.
20266         (output_cbranch): Likewise.
20267         (sparc_print_operand): Likewise.
20268         (sparc_expand_vec_perm_bmask): Likewise.
20269         (vector_init_bshuffle): Likewise.
20270         * config/spu/spu.c (spu_scalar_mode_supported_p): Likewise.
20271         (spu_vector_mode_supported_p): Likewise.
20272         (spu_expand_insv): Likewise.
20273         (spu_emit_branch_or_set): Likewise.
20274         (spu_handle_vector_attribute): Likewise.
20275         (spu_builtin_splats): Likewise.
20276         (spu_builtin_extract): Likewise.
20277         (spu_builtin_promote): Likewise.
20278         (spu_expand_sign_extend): Likewise.
20279         * config/tilegx/tilegx.c (tilegx_scalar_mode_supported_p): Likewise.
20280         (tilegx_simd_int): Likewise.
20281         * config/tilepro/tilepro.c (tilepro_scalar_mode_supported_p): Likewise.
20282         (tilepro_simd_int): Likewise.
20283         * config/v850/v850.c (const_double_split): Likewise.
20284         (v850_print_operand): Likewise.
20285         (ep_memory_offset): Likewise.
20286         * config/vax/vax.c (vax_rtx_costs): Likewise.
20287         (vax_output_int_move): Likewise.
20288         (vax_output_int_add): Likewise.
20289         (vax_output_int_subtract): Likewise.
20290         * config/visium/predicates.md (visium_branch_operator): Likewise.
20291         * config/visium/visium.c (rtx_ok_for_offset_p): Likewise.
20292         (visium_print_operand_address): Likewise.
20293         * config/visium/visium.h (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
20294         * config/xtensa/xtensa.c (xtensa_mem_offset): Likewise.
20295         (xtensa_expand_conditional_branch): Likewise.
20296         (xtensa_copy_incoming_a7): Likewise.
20297         (xtensa_output_literal): Likewise.
20298         * dfp.c (decimal_real_maxval): Likewise.
20299         * targhooks.c (default_libgcc_floating_mode_supported_p): Likewise.
20301 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
20302             Alan Hayward  <alan.hayward@arm.com>
20303             David Sherwood  <david.sherwood@arm.com>
20305         * genmodes.c (mode_size_inline): Add an E_ prefix to mode names.
20306         (mode_nunits_inline): Likewise.
20307         (mode_inner_inline): Likewise.
20308         (mode_unit_size_inline): Likewise.
20309         (mode_unit_precision_inline): Likewise.
20310         (emit_insn_modes_h): Likewise.  Also emit a #define of the
20311         unprefixed name.
20312         (emit_mode_wider): Add an E_ prefix to mode names.
20313         (emit_mode_complex): Likewise.
20314         (emit_mode_inner): Likewise.
20315         (emit_mode_adjustments): Likewise.
20316         (emit_mode_int_n): Likewise.
20317         * config/aarch64/aarch64-builtins.c (v8qi_UP, v4hi_UP, v4hf_UP)
20318         (v2si_UP, v2sf_UP, v1df_UP, di_UP, df_UP, v16qi_UP, v8hi_UP, v8hf_UP)
20319         (v4si_UP, v4sf_UP, v2di_UP, v2df_UP, ti_UP, oi_UP, ci_UP, xi_UP)
20320         (si_UP, sf_UP, hi_UP, hf_UP, qi_UP): Likewise.
20321         (CRC32_BUILTIN, ENTRY): Likewise.
20322         * config/aarch64/aarch64.c (aarch64_push_regs): Likewise.
20323         (aarch64_pop_regs): Likewise.
20324         (aarch64_process_components): Likewise.
20325         * config/alpha/alpha.c (alpha_emit_conditional_move): Likewise.
20326         * config/arm/arm-builtins.c (v8qi_UP, v4hi_UP, v4hf_UP, v2si_UP)
20327         (v2sf_UP, di_UP, v16qi_UP, v8hi_UP, v8hf_UP, v4si_UP, v4sf_UP)
20328         (v2di_UP, ti_UP, ei_UP, oi_UP, hf_UP, si_UP, void_UP): Likewise.
20329         * config/arm/arm.c (arm_init_libfuncs): Likewise.
20330         * config/i386/i386-builtin-types.awk (ix86_builtin_type_vect_mode):
20331         Likewise.
20332         * config/i386/i386-builtin.def (pcmpestr): Likewise.
20333         (pcmpistr): Likewise.
20334         * config/microblaze/microblaze.c (double_memory_operand): Likewise.
20335         * config/mmix/mmix.c (mmix_output_condition): Likewise.
20336         * config/powerpcspe/powerpcspe.c (rs6000_init_hard_regno_mode_ok):
20337         Likewise.
20338         * config/rl78/rl78.c (mduc_regs): Likewise.
20339         * config/rs6000/rs6000.c (rs6000_init_hard_regno_mode_ok): Likewise.
20340         (htm_expand_builtin): Likewise.
20341         * config/sh/sh.h (REGISTER_NATURAL_MODE): Likewise.
20342         * config/sparc/sparc.c (emit_save_or_restore_regs): Likewise.
20343         * config/xtensa/xtensa.c (print_operand): Likewise.
20344         * expmed.h (NUM_MODE_PARTIAL_INT): Likewise.
20345         (NUM_MODE_VECTOR_INT): Likewise.
20346         * genoutput.c (null_operand): Likewise.
20347         (output_operand_data): Likewise.
20348         * genrecog.c (print_parameter_value): Likewise.
20349         * lra.c (debug_operand_data): Likewise.
20351 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
20353         * dwarf2out.c (scompare_loc_descriptor_wide)
20354         (scompare_loc_descriptor_narrow): New functions, split out from...
20355         (scompare_loc_descriptor): ...here.
20356         * expmed.c (emit_store_flag_int): New function, split out from...
20357         (emit_store_flag): ...here.
20359 2017-08-30  Richard Biener  <rguenther@suse.de>
20361         * dwarf2out.c (dwarf2out_finish): Remove setting AT_pubnames.
20362         (dwarf2out_early_finish): Move setting of AT_pubnames from
20363         early debug output to early finish.
20365 2017-08-30  Jozef Lawrynowicz  <jozef.l@somniumtech.com>
20367         * gcc/config/msp430/msp430.h: Pass -mcode/data-region to the linker
20368         and -mdata-region to the assembler.
20370 2017-08-30  Richard Biener  <rguenther@suse.de>
20372         * dwarf2out.c (add_dwarf_attr): Check we don't add duplicate
20373         attributes.
20374         (gen_subprogram_die): Add DW_AT_object_pointer only early.
20375         (dwarf2out_early_global_decl): Only generate a DIE for the
20376         abstract origin if it doesn't already exist or is a declaration DIE.
20377         (resolve_addr): Do not add the linkage name twice when
20378         generating a stub DIE for the DW_TAG_GNU_call_site target.
20380 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
20382         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
20383         Use machine_mode rather than int for arg1_mode.
20385 2017-08-29  Michael Meissner  <meissner@linux.vnet.ibm.com>
20387         PR target/82015
20388         * config/rs6000/rs6000.c (rs6000_expand_binop_builtin): Insure
20389         that the second argument of the built-in functions to unpack
20390         128-bit scalar types to 64-bit values is 0 or 1.  Change to use a
20391         switch statement instead a lot of if statements.
20392         * config/rs6000/rs6000.md (unpack<mode>, FMOVE128_VSX iterator):
20393         Allow 64-bit values to be in Altivec registers as well as
20394         traditional floating point registers.
20395         (pack<mode>, FMOVE128_VSX iterator): Likewise.
20397 2017-08-29  Alexander Monakov  <amonakov@ispras.ru>
20399         * ira-costs.c (record_address_regs): Handle both operands of PLUS for
20400         MAX_REGS_PER_ADDRESS == 1.
20402 2017-08-29  Uros Bizjak  <ubizjak@gmail.com>
20404         * config/i386/i386.opt (flag_fentry): Do not init to -1.
20405         * config/i386/i386.c (ix86_option_override_internal): Simplify
20406         setting of opts->x_flag_entry.
20408 2017-08-29  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
20409             Jakub Jelinek  <jakub@redhat.com>
20410             Richard Biener  <rguenther@suse.de>
20412         PR tree-optimization/81503
20413         * gimple-ssa-strength-reduction.c (replace_mult_candidate): Ensure
20414         folded constant fits in the target type; reorder tests for clarity.
20416 2017-08-29  Martin Liska  <mliska@suse.cz>
20418         * passes.def: Include pass_lower_switch.
20419         * stmt.c (dump_case_nodes): Remove and move to
20420         tree-switch-conversion.
20421         (case_values_threshold): Likewise.
20422         (expand_switch_as_decision_tree_p): Likewise.
20423         (emit_case_decision_tree): Likewise.
20424         (expand_case): Likewise.
20425         (balance_case_nodes): Likewise.
20426         (node_has_low_bound): Likewise.
20427         (node_has_high_bound): Likewise.
20428         (node_is_bounded): Likewise.
20429         (emit_case_nodes): Likewise.
20430         (struct simple_case_node): New struct.
20431         (add_case_node): Remove.
20432         (emit_case_dispatch_table): Use vector instead of case_list.
20433         (reset_out_edges_aux): Remove.
20434         (compute_cases_per_edge): Likewise.
20435         (expand_case): Build list of simple_case_node.
20436         (expand_sjlj_dispatch_table): Use it.
20437         * tree-switch-conversion.c (struct case_node): Moved from
20438         stmt.c and adjusted.
20439         (emit_case_nodes): Likewise.
20440         (node_has_low_bound): Likewise.
20441         (node_has_high_bound): Likewise.
20442         (node_is_bounded): Likewise.
20443         (case_values_threshold): Likewise.
20444         (reset_out_edges_aux): Likewise.
20445         (compute_cases_per_edge): Likewise.
20446         (add_case_node): Likewise.
20447         (dump_case_nodes): Likewise.
20448         (balance_case_nodes): Likewise.
20449         (expand_switch_as_decision_tree_p): Likewise.
20450         (emit_jump): Likewise.
20451         (emit_case_decision_tree): Likewise.
20452         (try_switch_expansion): Likewise.
20453         (do_jump_if_equal): Likewise.
20454         (emit_cmp_and_jump_insns): Likewise.
20455         (fix_phi_operands_for_edge): New function.
20456         (record_phi_operand_mapping): Likewise.
20457         (class pass_lower_switch): New pass.
20458         (pass_lower_switch::execute): New function.
20459         (make_pass_lower_switch): Likewise.
20460         (conditional_probability):
20461         * timevar.def: Add TV_TREE_SWITCH_LOWERING.
20462         * tree-pass.h: Add make_pass_lower_switch.
20464 2017-08-29  Jozef Lawrynowicz  <jozef.l@somniumtech.com>
20466         PR target/80993
20467         * gcc/config/msp430/msp430.c (msp430_attr): Mark interrupt
20468         handlers as used.
20470 2017-08-29  Richard Biener  <rguenther@suse.de>
20472         * dwarf2out.c (add_dwarf_attr): When checking is enabled verify
20473         we do not add a DW_AT_inline attribute twice.
20474         (gen_subprogram_die): Remove code setting DW_AT_inline on
20475         DECL_ABSTRACT_P nodes.
20477 2017-08-29  Richard Sandiford  <richard.sandiford@linaro.org>
20479         * gimplify.c (gimplify_call_expr): Copy the nothrow flag to
20480         calls to internal functions.
20481         (gimplify_modify_expr): Likewise.
20482         * tree-call-cdce.c (use_internal_fn): Likewise.
20483         * tree-ssa-math-opts.c (pass_cse_reciprocals::execute): Likewise.
20484         (convert_to_divmod): Set the nothrow flag.
20485         * tree-if-conv.c (predicate_mem_writes):  Likewise.
20486         * tree-vect-stmts.c (vectorizable_mask_load_store): Likewise.
20487         (vectorizable_call): Likewise.
20488         (vectorizable_store): Likewise.
20489         (vectorizable_load): Likewise.
20490         * tree-vect-patterns.c (vect_recog_pow_pattern): Likewise.
20491         (vect_recog_mask_conversion_pattern): Likewise.
20493 2017-08-29  Martin Liska  <mliska@suse.cz>
20495         PR other/39851
20496         * gcc.c (driver_handle_option): Add new argument.
20497         * opts-common.c (handle_option): Pass
20498         target_option_override_hook.
20499         * opts-global.c (lang_handle_option): Add new option.
20500         (set_default_handlers):  Add new argument.
20501         (decode_options): Likewise.
20502         * opts.c (target_handle_option): Likewise.
20503         (common_handle_option): Call target_option_override_hook.
20504         * opts.h (struct cl_option_handler_func): Add hook for
20505         target option override.
20506         (struct cl_option_handlers): Likewise.
20507         (set_default_handlers): Add new argument.
20508         (decode_options): Likewise.
20509         (common_handle_option): Likewise.
20510         (target_handle_option): Likewise.
20511         * toplev.c (toplev::main): Pass targetm.target_option.override
20512         hook.
20514 2017-08-29  Richard Biener  <rguenther@suse.de>
20515         Dominik Infuehr <dominik.infuehr@theobroma-systems.com>
20517         * tree-vect-slp.c (vect_bb_slp_scalar_cost): Properly confine
20518         life to the active subtree.
20520 2017-08-28  Jeff Law  <law@redhat.com>
20522         * tree-ssa-dom.c (edge_info::record_simple_equiv): Call
20523         derive_equivalences.
20524         (derive_equivalences_from_bit_ior, record_temporary_equivalences):
20525         Code moved into....
20526         (edge_info::derive_equivalences): New private member function
20528         * tree-ssa-dom.c (class edge_info): Changed from a struct
20529         to a class.  Add ctor/dtor, methods and data members.
20530         (edge_info::edge_info): Renamed from allocate_edge_info.
20531         Initialize additional members.
20532         (edge_info::~edge_info): New.
20533         (free_dom_edge_info): Delete the edge info.
20534         (record_edge_info): Use new class & associated member functions.
20535         Tighten forms for testing for edge equivalences.
20536         (record_temporary_equivalences): Iterate over the simple
20537         equivalences rather than assuming there's only one per edge.
20538         (cprop_into_successor_phis): Iterate over the simple
20539         equivalences rather than assuming there's only one per edge.
20540         (optimize_stmt): Use operand_equal_p rather than pointer
20541         equality for mini-DSE code.
20543 2017-08-28  Nathan Sidwell  <nathan@acm.org>
20545         * gcc.c (execute): Fold SIGPIPE handling into switch
20546         statement.  Adjust internal error message.
20548 2017-08-28  Richard Biener  <rguenther@suse.de>
20550         PR debug/81993
20551         * dwarf2out.c (gen_remaining_tmpl_value_param_die_attributes):
20552         Do nothing for removed DIEs.
20554 2017-08-28  Richard Biener  <rguenther@suse.de>
20556         PR tree-optimization/81977
20557         * tree-ssa-sccvn.c (vn_reference_lookup_3): Fix look through
20558         memcpy.
20560 2017-08-28  Alexander Monakov  <amonakov@ispras.ru>
20562         PR target/80640
20563         * doc/md.texi (mem_thread_fence): Remove mention of mode.  Rewrite.
20564         * optabs.c (expand_mem_thread_fence): Emit a compiler barrier when
20565         using targetm.gen_mem_thread_fence.
20567 2017-08-27  Uros Bizjak  <ubizjak@gmail.com>
20569         PR target/81995
20570         * config/i386/i386.md (*<btsc><mode>): Change operand 2
20571         predicate to register_operand.  Reorder operands.
20572         (*btr<mode>): Ditto.
20573         (*<btsc><mode>_mask): Change operand 3 predicate to register_operand.
20574         (*btr<mode>_mask): Ditto.
20576 2017-08-25  Steven Munroe  <munroesj@gcc.gnu.org>
20578         * config.gcc (powerpc*-*-*): Add xmmintrin.h and mm_malloc.h.
20579         * config/rs6000/xmmintrin.h: New file.
20580         * config/rs6000/x86intrin.h [__ALTIVEC__]: Include xmmintrin.h.
20582 2017-08-25  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
20584         PR target/81504
20585         * config/rs6000/rs6000-p8swap.c (find_alignment_op): Add reference
20586         parameter and_insn and return it.
20587         (recombine_lvx_pattern): Insert a copy to ensure availability of
20588         the base register of the copied masking operation at the point of
20589         the instruction replacement.
20590         (recombine_stvx_pattern): Likewise.
20592 2017-08-25  Michael Meissner  <meissner@linux.vnet.ibm.com>
20594         * config/rs6000/rs6000.opt (-mpower9-dform-scalar): Delete
20595         undocumented switches.
20596         (-mpower9-dform-vector): Likewise.
20597         (-mpower9-dform): Likewise.
20598         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
20599         comments to delete references to -mpower9-dform* switches.
20600         * config/rs6000/predicates.md (vsx_quad_dform_memory_operand):
20601         Delete reference to -mpower9-dform* switches, test for
20602         -mpower9-vector instead.
20603         * config/rs6000/rs6000-cpus.def (ISA_3_0_MASKS_SERVER): Likewise.
20604         (OTHER_P9_VECTOR_MASKS): Likewise.
20605         (POWERPC_MASKS): Likewise.
20606         * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): Change
20607         tests against -mpower9-dform* to -mpower9-vector.  Delete code
20608         that checked for -mpower9-dform* consistancy with other options.
20609         Add test for -mpower9-misc to enable other power9 switches.
20610         (rs6000_init_hard_regno_mode_ok): Likewise.
20611         (rs6000_option_override_internal): Likewise.
20612         (rs6000_emit_prologue): Likewise.
20613         (rs6000_emit_epilogue): Likewise.
20614         (rs6000_opt_masks): Delete -mpower9-dform-{scalar,vector}.
20615         (rs6000_disable_incompatiable_switches): Delete -mpower9-dform.
20616         (emit_fusion_p9_load): Change tests for -mpower9-dform-scalar
20617         -mpower9-vector.
20618         (emit_fusion_p9_store): Likewise.
20619         * config/rs6000/rs6000.h (TARGET_P9_DFORM_SCALAR): Delete
20620         resetting these macros if the assembler does not support ISA 3.0
20621         instructions.
20622         (TARGET_P9_DFORM_VECTOR): Likewise.
20623         * config/rs6000/rs6000.md (peepholes to optimize altivec memory):
20624         Change to use -mpower9-vector instead of -mpower9-dform-scalar.
20626 2017-08-25  Alan Modra  <amodra@gmail.com>
20628         PR rtl-optimization/81747
20629         * cse.c (cse_extended_basic_block): Don't attempt to record
20630         equivalences for degenerate conditional branches that branch
20631         to their fall-through.
20633 2017-08-24  Martin Sebor  <msebor@redhat.com>
20635         PR middle-end/81908
20636         * gimple-fold.c (size_must_be_zero_p): New function.
20637         (gimple_fold_builtin_memory_op): Call it.
20639 2017-08-24  Steven Munroe  <munroesj@gcc.gnu.org>
20641         * config/rs6000/mm_malloc.h: New file.
20643 2017-08-24  Bin Cheng  <bin.cheng@arm.com>
20645         PR tree-optimization/81913
20646         * tree-ssa-loop-niter.c (number_of_iterations_cond): Skip niter
20647         analysis when either IVs in condition can wrap.
20649 2017-08-24  Uros Bizjak  <ubizjak@gmail.com>
20651         * dwarf2out.c (MAX_ARTIFICIAL_LABEL_BYTES): Increase to 40.
20652         * dwarf2cfi.c (MAX_ARTIFICIAL_LABEL_BYTES): Remove.
20654 2017-08-24  Richard Biener  <rguenther@suse.de>
20656         PR target/81921
20657         * targhooks.c (default_target_can_inline_p): Properly
20658         use target_option_default_node when no DECL_FUNCTION_SPECIFIC_TARGET
20659         is present and always compare.
20660         * config/i386/i386.c (ix86_valid_target_attribute_tree): Do not
20661         infer -mfpmath=sse from TARGET_SSE_P.
20662         (ix86_can_inline_p): Properly use target_option_default_node when
20663         no DECL_FUNCTION_SPECIFIC_TARGET is present and always compare.
20665 2017-08-24  Richard Biener  <rguenther@suse.de>
20667         PR debug/81936
20668         * dwarf2out.c (output_die): Handle flag_generate_offload like
20669         flag_generate_lto.
20670         (output_comp_unit): Likewise.
20671         (gen_array_type_die): Likewise.
20672         (dwarf2out_early_finish): Likewise.
20673         (note_variable_value_in_expr): Likewise.
20674         (dwarf2out_finish): Likewise.  Adjust assert.
20675         * cgraphunit.c (symbol_table::compile): Move setting of
20676         flag_generate_offload earlier ...
20677         (symbol_table::finalize_compilation_unit): ... here, before
20678         early debug finalization.
20680 2017-08-24  Richard Biener  <rguenther@suse.de>
20682         * config/i386/i386.c: Include symbol-summary.h, ipa-prop.h
20683         and ipa-fnsummary.h.
20684         (ix86_can_inline_p): When ix86_fpmath flags do not match
20685         check whether the callee uses FP math at all.
20687 2017-08-24  Aldy Hernandez  <aldyh@redhat.com>
20689         PR middle-end/81931
20690         * tree-ssanames.c (get_nonzero_bits): Use element_precision
20691         instead of TYPE_PRECISION.
20693 2017-08-24  Richard Sandiford  <richard.sandiford@linaro.org>
20694             Alan Hayward  <alan.hayward@arm.com>
20695             David Sherwood  <david.sherwood@arm.com>
20697         * combine.c (make_extraction): Use subreg_offset_from_lsb.
20699 2017-08-23  Daniel Santos  <daniel.santos@pobox.com>
20701         * config/i386/i386.h (ix86_frame::stack_realign_allocate_offset):
20702         Remove field.
20703         (ix86_frame::stack_realign_allocate): New field.
20704         (struct machine_frame_state): Modify comments.
20705         (machine_frame_state::sp_realigned_fp_end): New field.
20706         * config/i386/i386.c (ix86_compute_frame_layout): Rework stack frame
20707         layout calculation.
20708         (sp_valid_at): Add assertion to assure no attempt to access invalid
20709         offset of a realigned stack.
20710         (fp_valid_at): Likewise.
20711         (choose_baseaddr): Modify comments.
20712         (ix86_emit_outlined_ms2sysv_save): Adjust to changes in
20713         ix86_expand_prologue.
20714         (ix86_expand_prologue): Modify stack realignment and allocation.
20715         (ix86_expand_epilogue): Modify comments.
20716         * doc/sourcebuild.texi: Add documentation for target selectors avx2,
20717         avx2_runtime, avx512f, and avx512f_runtime.
20719 2017-08-23  Uros Bizjak  <ubizjak@gmail.com>
20721         * config/i386/i386.opt: Remove unneeded Init(0) initializations.
20722         (mstackrealign): Do not init to -1.
20723         * config/i386/i386.c (ix86_option_override_internal):
20724         Check opts_set, not opts when setting default value of
20725         opts->x_ix86_force_align_arg_pointer.
20727 2017-08-23  Richard Biener  <rguenther@suse.de>
20729         * function.c (fndecl_name): Use verbosity 1 (no arguments) for
20730         lang_hooks.decl_printable_name.
20731         * print-rtl-function.c (print_rtx_function): Likewise.
20732         * tree-pretty-print.c (dump_function_header): Likewise.
20734 2017-08-23  Richard Biener  <rguenther@suse.de>
20736         PR lto/81940
20737         * dwarf2out.c (dwarf2out_abstract_function): Handle LTO with
20738         -g0 at compile-time.
20740 2017-08-23  Tamar Christina  <tamar.christina@arm.com>
20742         PR middle-end/19706
20743         * doc/sourcebuild.texi (Other hardware attributes):
20744         Document xorsign.
20746 2017-08-23  Tamar Christina  <tamar.christina@arm.com>
20748         PR middle-end/19706
20749         * tree-ssa-math-opts.c (convert_expand_mult_copysign):
20750         Fix single-use check.
20752 2017-08-23  Thomas Preud'homme  <thomas.preudhomme@arm.com>
20754         * gcc.c (execute): Only test for SIGKILL and SIGQUIT if available.
20756 2017-08-22  Daniel Santos  <daniel.santos@pobox.com>
20758         * doc/install.texi: Modify to add more details on running selected
20759         tests.
20761 2017-08-22  Daniel Santos  <daniel.santos@pobox.com>
20763         * config/i386/i386.c (ix86_option_override_internal): Error when -mx32
20764         is combined with -mabi=ms.
20765         (ix86_function_type_abi): Limit errors for mixing -mx32 with attribute
20766         ms_abi.
20768 2017-08-22  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
20770         PR tree-optimization/81488
20771         * gimple-ssa-strength-reduction (struct slsr_cand_d): Add visited
20772         and cached_basis fields.
20773         (MAX_SPREAD): New constant.
20774         (alloc_cand_and_find_basis): Initialize new fields.
20775         (clear_visited): New function.
20776         (create_phi_basis_1): Rename from create_phi_basis, set visited
20777         and cached_basis fields.
20778         (create_phi_basis): New wrapper function.
20779         (phi_add_costs_1): Rename from phi_add_costs, add spread
20780         parameter, set visited field, short-circuit when limits reached.
20781         (phi_add_costs): New wrapper function.
20782         (record_phi_increments_1): Rename from record_phi_increments, set
20783         visited field.
20784         (record_phi_increments): New wrapper function.
20785         (phi_incr_cost_1): Rename from phi_incr_cost, set visited field.
20786         (phi_incr_cost): New wrapper function.
20787         (all_phi_incrs_profitable_1): Rename from
20788         all_phi_incrs_profitable, set visited field.
20789         (all_phi_incrs_profitable): New wrapper function.
20791 2017-08-22  Richard Sandiford  <richard.sandiford@linaro.org>
20792             Alan Hayward  <alan.hayward@arm.com>
20793             David Sherwood  <david.sherwood@arm.com>
20795         * rtl.h (paradoxical_subreg_p): Define inline, and add a version
20796         that takes the outer and inner modes.
20797         * doc/rtl.texi: Use paradoxical_subreg_p instead of a GET_MODE_SIZE
20798         comparison as the canonical test for a paradoxical subreg.
20799         * combine.c (simplify_set): Use paradoxical_subreg_p.
20800         (make_extraction): Likewise.
20801         (force_to_mode): Likewise.
20802         (rtx_equal_for_field_assignment_p): Likewise.
20803         (gen_lowpart_for_combine): Likewise.
20804         (simplify_comparison): Likewise.
20805         * cse.c (equiv_constant): Likewise.
20806         * expmed.c (store_bit_field_1): Likewise.
20807         * final.c (alter_subreg): Likewise.
20808         * fwprop.c (propagate_rtx): Likewise.
20809         (forward_propagate_subreg): Likewise.
20810         * ira-conflicts.c (ira_build_conflicts): Likewise.
20811         * lower-subreg.c (simplify_gen_subreg_concatn): Likewise.
20812         * lra-constraints.c (curr_insn_transform): Likewise.
20813         (split_reg): Likewise.
20814         * lra-eliminations.c (move_plus_up): Likewise.
20815         (lra_eliminate_regs_1): Likewise.
20816         * recog.c (general_operand): Likewise.
20817         * ree.c (combine_reaching_defs): Likewise.
20818         * reload.c (push_reload): Likewise.
20819         (find_reloads): Likewise.
20820         * reload1.c (elimination_effects): Likewise.
20821         (compute_reload_subreg_offset): Likewise.
20822         (choose_reload_regs): Likewise.
20823         * rtlanal.c (subreg_lsb_1): Likewise.
20824         * simplify-rtx.c (simplify_unary_operation_1): Likewise.
20825         (simplify_subreg): Likewise.
20826         * var-tracking.c (track_loc_p): Likewise.
20827         * emit-rtl.c (byte_lowpart_offset): Likewise.
20828         (paradoxical_subreg_p): Delete out-of-line definition.
20830 2017-08-22  Jeff Law  <law@redhat.com>
20832         PR tree-optimization/81741
20833         PR tree-optimization/71947
20834         * tree-ssa-dom.c: Include tree-inline.h.
20835         (record_temporary_equivalences): Only record SSA_NAME = SSA_NAME
20836         equivalences if one is more expensive to compute than the other.
20837         * tree-ssa-scopedtables.h (class const_or_copies): Make
20838         record_const_or_copy_raw method private.
20839         (class avail_exprs_stack): New method simplify_binary_operation.
20840         * tree-ssa-scopedtables.c (avail_exprs_stack::lookup_avail_expr): Call
20841         avail_exprs_stack::simplify_binary_operation as needed.
20842         (avail_exprs_stack::simplify_binary_operation): New function.
20844 2017-08-22  Sebastian Huber  <sebastian.huber@embedded-brains.de>
20846         * config.gcc (powerpc-*-rtems*): Add rs6000/linux64.opt.
20847         * config/rs6000/rtems.h (ASM_PREFERRED_EH_DATA_FORMAT): New define.
20848         (DOT_SYMBOLS): Likewise.
20849         (MINIMAL_TOC_SECTION_ASM_OP): Likewise.
20850         (RELOCATABLE_NEEDS_FIXUP): Likewise.
20851         (RS6000_ABI_NAME): Likewise.
20852         (TARGET_CMODEL): Likewise.
20853         (TOC_SECTION_ASM_OP): Likewise.
20854         (SET_CMODEL): New macro.
20855         (SUBSUBTARGET_OVERRIDE_OPTIONS): Evaluate cmodel options.
20857 2017-08-22  Richard Biener  <rguenther@suse.de>
20859         * tree-inline.c (remap_type_1): Change asserts on TYPE_SIZE[_UNIT]
20860         to allow for free-lang-data replacements similar to verify_type_variant.
20862 2017-08-22  Richard Sandiford  <richard.sandiford@linaro.org>
20863             Alan Hayward  <alan.hayward@arm.com>
20864             David Sherwood  <david.sherwood@arm.com>
20866         * config/aarch64/aarch64.md (casesi): Use DImode rather than
20867         VOIDmode for the LABEL_REF.
20869 2017-08-22  Richard Biener  <rguenther@suse.de>
20871         * tree-cfg.c (gimple_split_edge): Avoid reallocating target PHI nodes.
20873 2017-08-22  Richard Biener  <rguenther@suse.de>
20875         * common.opt (feliminate-dwarf2-dups): Ignore.
20876         * doc/invoke.texi (feliminate-dwarf2-dups): Remove documentation.
20877         * dwarf2out.c (push_new_compile_unit, pop_compile_unit,
20878         same_die_p_wrap, compute_section_prefix,
20879         is_symbol_die, assign_symbol_names, break_out_includes): Remove.
20880         (comdat_symbol_id, comdat_symbol_number): Likewise.
20881         (cu_hash_table_entry, cu_hash_table_entry_hasher, cu_hash_type):
20882         Likewise.
20883         (check_duplicate_cu, record_comdat_symbol_number): Likewise.
20884         (output_die): Mark unreachable path unreachable.
20885         (dwarf2out_start_source_file): Do not create DW_TAG_GNU_BINCL.
20886         (dwarf2out_end_source_file): Do not create DW_TAG_GNU_EINCL.
20887         (dwarf2out_init): Remove code handling flag_eliminate_dwarf2_dups.
20888         (dwarf2out_early_finish): Likewise.
20890 2017-08-22  Aldy Hernandez  <aldyh@redhat.com>
20892         * wide-int.h (hwi_with_prec::hwi_with_prec): Sign extend.
20894 2017-08-22  Georg-Johann Lay  <avr@gjlay.de>
20896         PR target/81910
20897         * config/avr/avr.c (avr_handle_addr_attribute): Early return if
20898         not VAR_P. Filter attribute warnings with OPT_Wattributes.
20899         (avr_attribute_table) <io, io_low, address>: Initialize
20900         .decl_required with true.
20902 2017-08-21  Michael Meissner  <meissner@linux.vnet.ibm.com>
20904         * config/rs6000/rs6000-cpus.def (-mvsx-scalar-float): Delete
20905         undocumented debugging options.
20906         (-mvsx-scalar-double): Likewise.
20907         (-mallow-df-permute): Likewise.
20908         (-mvectorize-builtins): Likewise.
20909         * config/rs6000/rs6000.c (rs6000_init_hard_regno_mode_ok): Likewise.
20910         (rs6000_builtin_vectorized_function): Likewise.
20911         (rs6000_builtin_md_vectorized_function): Likewise.
20912         (rs6000_opt_vars): Likewise.
20914 2017-08-21  Uros Bizjak  <ubizjak@gmail.com>
20916         PR target/46091
20917         * config/i386/i386.md (*btsq_imm): Rename from *btsq.
20918         (*btrq_imm): Rename from *btrq.
20919         (*btcq_imm): Rename from *btcq.
20920         (btsc): New code attribute.
20921         (*<btsc><mode>): New insn pattern.
20922         (*btr<mode>): Ditto.
20923         (*<btsc><mode>_mask): New insn_and_split pattern.
20924         (*btr<mode>_mask): Ditto.
20926 2017-08-21  Richard Sandiford  <richard.sandiford@linaro.org>
20927             Alan Hayward  <alan.hayward@arm.com>
20928             David Sherwood  <david.sherwood@arm.com>
20930         * function.c (pad_below): Simplify padding calculation.
20932 2017-08-21  Richard Sandiford  <richard.sandiford@linaro.org>
20933             Alan Hayward  <alan.hayward@arm.com>
20934             David Sherwood  <david.sherwood@arm.com>
20936         * target.def (function_prologue): Remove frame size argument.
20937         (function_epilogue): Likewise.
20938         * doc/tm.texi: Regenerate.
20939         * final.c (final_start_function): Update call to function_prologue.
20940         (final_end_function): Update call to function_epilogue.
20941         (default_function_pro_epilogue): Remove frame size argument.
20942         * output.h (default_function_pro_epilogue): Likewise.
20943         * config/arm/arm.c (arm_output_function_epilogue): Likewise.
20944         (arm_output_function_prologue): Likewise.
20945         * config/frv/frv.c (frv_function_prologue): Likewise.
20946         (frv_function_epilogue): Likewise.
20947         * config/i386/i386.c (ix86_output_function_epilogue): Likewise.
20948         * config/ia64/ia64.c (ia64_output_function_prologue): Likewise.
20949         (ia64_output_function_epilogue): Likewise.
20950         * config/m32r/m32r.c (m32r_output_function_prologue): Likewise.
20951         (m32r_output_function_epilogue): Likewise.
20952         * config/microblaze/microblaze.c (microblaze_function_prologue)
20953         (microblaze_function_epilogue): Likewise.
20954         * config/mips/mips.c (mips_output_function_prologue): Likewise.
20955         (mips_output_function_epilogue): Likewise.
20956         * config/mmix/mmix.c (mmix_target_asm_function_prologue): Likewise.
20957         (mmix_target_asm_function_epilogue): Likewise.
20958         * config/msp430/msp430.c (msp430_start_function): Likewise.
20959         * config/nds32/nds32.c (nds32_asm_function_prologue): Likewise.
20960         (nds32_asm_function_epilogue): Likewise.
20961         * config/nios2/nios2.c (nios2_asm_function_prologue): Likewise.
20962         * config/pa/pa.c (pa_output_function_prologue): Likewise.
20963         (pa_output_function_epilogue): Likewise.
20964         * config/powerpcspe/powerpcspe.c (rs6000_output_function_prologue)
20965         (rs6000_output_function_epilogue): Likewise.
20966         * config/rl78/rl78.c (rl78_start_function): Likewise.
20967         * config/rs6000/rs6000.c (rs6000_output_function_prologue): Likewise.
20968         (rs6000_output_function_epilogue): Likewise.
20969         * config/rx/rx.c (rx_output_function_prologue): Likewise.
20970         * config/sh/sh.c (sh_output_function_epilogue): Likewise.
20971         * config/sparc/sparc.c (sparc_asm_function_prologue): Likewise.
20972         (sparc_asm_function_epilogue): Likewise.
20974 2017-08-21  Richard Sandiford  <richard.sandiford@linaro.org>
20976         * tree.h (type_has_mode_precision_p): New function.
20977         * convert.c (convert_to_integer_1): Use it.
20978         * expr.c (expand_expr_real_2): Likewise.
20979         (expand_expr_real_1): Likewise.
20980         * fold-const.c (fold_single_bit_test_into_sign_test): Likewise.
20981         * match.pd: Likewise.
20982         * tree-ssa-forwprop.c (simplify_rotate): Likewise.
20983         * tree-ssa-math-opts.c (convert_mult_to_fma): Likewise.
20984         * tree-tailcall.c (process_assignment): Likewise.
20985         * tree-vect-loop.c (vectorizable_reduction): Likewise.
20986         * tree-vect-patterns.c (vect_recog_vector_vector_shift_pattern)
20987         (vect_recog_mult_pattern, vect_recog_divmod_pattern): Likewise.
20988         * tree-vect-stmts.c (vectorizable_conversion): Likewise.
20989         (vectorizable_assignment): Likewise.
20990         (vectorizable_shift): Likewise.
20991         (vectorizable_operation): Likewise.
20992         * tree-vrp.c (register_edge_assert_for_2): Likewise.
20994 2017-08-21  Wilco Dijkstra  <wdijkstr@arm.com>
20996         * match.pd: Add pow (C, x) simplification.
20998 2017-08-21  Richard Biener  <rguenther@suse.de>
21000         PR tree-optimization/81900
21001         * tree-ssa-pre.c (compute_antic_aux): Properly compute changed
21002         for blocks with abnormal predecessors.
21003         (compute_antic): Do not set visited flag prematurely.
21005 2017-08-21  Georg-Johann Lay  <avr@gjlay.de>
21007         PR target/79883
21008         * config/avr/avr.c (avr_set_current_function): Typo in diagnostic.
21010 2017-08-21  Richard Sandiford  <richard.sandiford@linaro.org>
21012         * stor-layout.h (vector_type_mode): Move to...
21013         * tree.h (vector_type_mode): ...here.
21014         * stor-layout.c (vector_type_mode): Move to...
21015         * tree.c (vector_type_mode): ...here.  Include rtl.h and regs.h.
21017 2017-08-21  Richard Biener  <rguenther@suse.de>
21019         * debug.h (struct gcc_debug_hooks): Add die_ref_for_decl and
21020         register_external_die hooks.
21021         (debug_false_tree_charstarstar_uhwistar): Declare.
21022         (debug_nothing_tree_charstar_uhwi): Likewise.
21023         * debug.c (do_nothing_debug_hooks): Adjust.
21024         (debug_false_tree_charstarstar_uhwistar): New do nothing.
21025         (debug_nothing_tree_charstar_uhwi): Likewise.
21026         * dbxout.c (dbx_debug_hooks): Adjust.
21027         (xcoff_debug_hooks): Likewise.
21028         * sdbout.c (sdb_debug_hooks): Likewise.
21029         * vmsdbgout.c (vmsdbg_debug_hooks): Likewise.
21030         * dwarf2out.c (macinfo_label_base): New global.
21031         (dwarf2out_register_external_die): New function for the
21032         register_external_die hook.
21033         (dwarf2out_die_ref_for_decl): Likewise for die_ref_for_decl.
21034         (dwarf2_debug_hooks): Use them.
21035         (dwarf2_lineno_debug_hooks): Adjust.
21036         (struct die_struct): Add with_offset flag.
21037         (DEBUG_LTO_DWO_INFO_SECTION, DEBUG_LTO_INFO_SECTION,
21038         DEBUG_LTO_DWO_ABBREV_SECTION, DEBUG_LTO_ABBREV_SECTION,
21039         DEBUG_LTO_DWO_MACINFO_SECTION, DEBUG_LTO_MACINFO_SECTION,
21040         DEBUG_LTO_DWO_MACRO_SECTION, DEBUG_LTO_MACRO_SECTION,
21041         DEBUG_LTO_LINE_SECTION, DEBUG_LTO_DWO_STR_OFFSETS_SECTION,
21042         DEBUG_LTO_STR_DWO_SECTION, DEBUG_STR_LTO_SECTION): New macros
21043         defining section names for the early LTO debug variants.
21044         (reset_indirect_string): New helper.
21045         (add_AT_external_die_ref): Helper for dwarf2out_register_external_die.
21046         (print_dw_val): Add support for offsetted symbol references.
21047         (get_ultimate_context): Split out from is_cxx.
21048         (is_cxx): Use get_ultimate_context.
21049         (is_fortran): Add decl overload.
21050         (compute_comp_unit_symbol): Split out worker from
21051         compute_section_prefix.
21052         (compute_section_prefix): Call compute_comp_unit_symbol and
21053         set comdat_type_p here.
21054         (output_die): Skip DIE symbol output for the LTO added one.
21055         Handle DIE symbol references with offset.
21056         (output_comp_unit): Guard section name mangling properly.
21057         For LTO debug sections emit a symbol at the section beginning
21058         which we use to refer to its DIEs.
21059         (add_abstract_origin_attribute): For DIEs registered via
21060         dwarf2out_register_external_die directly refer to the early
21061         DIE rather than indirectly through the shadow one we created.
21062         Remove obsolete call to dwarf2out_abstract_function for
21063         non-function/block origins.
21064         (gen_array_type_die): When generating early LTO debug do
21065         not emit DW_AT_string_length.
21066         (gen_formal_parameter_die): Do not re-create DIEs for PARM_DECLs
21067         late when in LTO.  As suggested place a gcc_unreachable for
21068         the DECL_ABSTRACT_P case.
21069         (gen_subprogram_die): Avoid another specification DIE
21070         for early built declarations/definitions for the late LTO case.
21071         (gen_variable_die): Add type references for late duplicated VLA dies
21072         when in late LTO.
21073         (gen_inlined_subroutine_die): Do not call dwarf2out_abstract_function,
21074         we have the abstract instance already.
21075         (process_scope_var): Adjust decl DIE contexts in LTO which
21076         first puts them in limbo.
21077         (gen_decl_die): Do not generate type DIEs late apart from
21078         types for VLAs or for decls we do not yet have a DIE.  Do not
21079         call dwarf2out_abstract_function late.
21080         (dwarf2out_early_global_decl): Make sure to create DIEs
21081         for abstract instances of a decl first.
21082         (dwarf2out_late_global_decl): Adjust comment.
21083         (output_macinfo_op): With multiple macro sections use
21084         macinfo_label_base to distinguish labels.
21085         (output_macinfo): Likewise.  Update macinfo_label_base.
21086         Pass in the line info label.
21087         (note_variable_value_in_expr): When generating LTO resolve
21088         all variable values here by generating DIEs as needed.
21089         (init_sections_and_labels): Add early LTO debug flag parameter
21090         and generate different sections and names if set.  Add generation
21091         counter for the labels so we can have multiple of them.
21092         (reset_dies): Helper to allow DIEs to be output multiple times.
21093         (dwarf2out_finish): When outputting DIEs to the fat part of an
21094         LTO object first reset DIEs.
21095         (dwarf2out_early_finish): Output early DIEs when generating LTO.
21096         (modified_type_die): Check for decl_ultimate_origin being self
21097         before recursing.
21098         (gen_type_die_with_usage): Likewise.
21099         (gen_typedef_die): Allow decl_ultimate_origin being self.
21100         (set_decl_abstract_flags): Remove.
21101         (set_block_abstract_flags): Likewise.
21102         (dwarf2out_abstract_function): Treat the early generated DIEs
21103         as the abstract copy and only add DW_AT_inline and
21104         DW_AT_artificial here and call set_decl_origin_self.
21105         If the DIE has an abstract origin don't do anything.
21106         * tree.c (free_lang_data): Build a dummy TRANSLATION_UNIT_DECL
21107         if we have none yet (Go fails to build one, PR78628).
21108         (variably_modified_type_p): Prevent endless recursion for Ada
21109         cyclic pointer types.
21110         * lto-streamer-in.c: Include debug.h.
21111         (dref_queue): New global.
21112         (lto_read_tree_1): Stream in DIE references.
21113         (lto_input_tree): Register DIE references.
21114         (input_function): Stream DECL_DEBUG_ARGS.
21115         * lto-streamer-out.c: Include debug.h.
21116         (lto_write_tree_1): Output DIE references.
21117         (DFS::DFS_write_tree_body): Follow DECL_ABSTRACT_ORIGIN.
21118         Force a TRANSLATION_UNIT_DECL DECL_CONTEXT for file-scope decls.
21119         (output_function): Stream DECL_DEBUG_ARGS.
21120         * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
21121         Stream DECL_ABSTRACT_ORIGIN.
21122         * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
21123         (write_ts_decl_minimal_tree_pointers): Force a TRANSLATION_UNIT_DECL
21124         DECL_CONTEXT for file-scope decls.
21125         * lto-streamer.h (struct dref_entry): Declare.
21126         (dref_queue): Likewise.
21127         * cfgexpand.c (pass_expand::execute): Do not call the
21128         outlining_inline_function hook here.
21129         * lto-wrapper.c (debug_obj): New global.
21130         (tool_cleanup): Unlink it if required.
21131         (debug_objcopy): New function.
21132         (run_gcc): Handle early debug sections in the IL files by
21133         extracting them to separate files, partially linkin them and
21134         feeding the result back as result to the linker.
21135         * config/darwin.h (DEBUG_LTO_INFO_SECTION, DEBUG_LTO_ABBREV_SECTION,
21136         DEBUG_LTO_MACINFO_SECTION, DEBUG_LTO_LINE_SECTION,
21137         DEBUG_STR_LTO_SECTION, DEBUG_LTO_MACRO_SECTION): Put early debug
21138         sections into a separate segment.
21139         * config/darwin.c (darwin_asm_named_section): Handle __GNU_DWARF_LTO
21140         segments.
21141         (darwin_asm_dwarf_section): Likewise.
21142         (darwin_asm_output_dwarf_offset): Likewise.
21143         * config/i386/i386.c (make_resolver_func): Set DECL_IGNORED_P.
21145 2017-08-21  Richard Sandiford  <richard.sandiford@linaro.org>
21146             Alan Hayward  <alan.hayward@arm.com>
21147             David Sherwood  <david.sherwood@arm.com>
21149         * read-md.h (md_reader::record_potential_iterator_use): Replace
21150         pointer argument with an rtx and an index.
21151         * read-rtl.c (iterator_group::apply_iterator): Likewise.
21152         (apply_mode_iterator): Likewise.
21153         (apply_code_iterator): Likewise.
21154         (apply_int_iterator): Likewise.
21155         (apply_subst_iterator): Likewise.
21156         (record_iterator_use): Likewise.
21157         (record_attribute_use): Likewise.
21158         (md_reader::record_potential_iterator_use): Likewise.  Update calls
21159         to record_iterator_use and apply_iterator.
21160         (iterator_use): Replace ptr with x and index.
21161         (attribute_use): Likewise.
21162         (apply_attribute_uses): Update calls to apply_iterator.
21163         (apply_iterators): Likewise.  Update initialization of iterator_use.
21164         (rtx_reader::read_rtx_code): Update calls to record_iterator_use
21165         and record_potential_iterator_use.
21166         (rtx_reader::read_rtx_operand): Likewise.
21168 2017-08-21  Richard Sandiford  <richard.sandiford@linaro.org>
21169             Alan Hayward  <alan.hayward@arm.com>
21170             David Sherwood  <david.sherwood@arm.com>
21172         * varasm.c (const_rtx_hash_1): Don't hash in the mode of a
21173         CONST_WIDE_INT.
21175 2017-08-21  Richard Biener  <rguenther@suse.de>
21177         PR middle-end/81884
21178         * tree-ssa-alias.c (stmt_kills_ref_p): Handle array accesses
21179         at struct end conservatively when comparing common bases.
21181 2017-08-21  Richard Biener  <rguenther@suse.de>
21183         * tree-ssa-loop-im.c (struct lim_aux_data): Add ref index member.
21184         (mem_ref_in_stmt): Remove.
21185         (determine_max_movement): Use ref index to get at the reference.
21186         (invariantness_dom_walker::before_dom_children): Deal with
21187         lim data already initialized.
21188         (gather_mem_refs_stmt): Initialize lim data and record ref index.
21190 2017-08-19  Uros Bizjak  <ubizjak@gmail.com>
21192         * config/i386/i386.h (OPTION_MASK_ISA_ROUND): Remove.
21193         (TARGET_ISA_ROUND): Ditto.
21194         (TARGET_ROUND): Ditto.
21195         * config/i386/i386.c: Substitute TARGET_ROUND with TARGET_SSE4_1.
21196         * config/i386/i386.md: Ditto.
21197         * config/i386/sse.md: Ditto.
21198         * config/i386/i386-builtin.def: Substitute OPTION_MASK_ISA_ROUND
21199         with OPTION_MASK_ISA_SSE4_1.
21201 2017-08-19  Uros Bizjak  <ubizjak@gmail.com>
21203         PR target/81894
21204         * doc/extend.texi (x86 Built-in Functions): Correct the name of
21205         __builtin_ia32_lzcnt_u16.
21207 2017-08-18  Peter Bergner  <bergner@vnet.ibm.com>
21209         PR target/80210
21210         * config/rs6000/rs6000.c (rs6000_activate_target_options): New function.
21211         (rs6000_set_current_function): Rewrite function to use it.
21213 2017-08-18  H.J. Lu  <hongjiu.lu@intel.com>
21215         PR c/53037
21216         * print-tree.c (print_node): Support DECL_WARN_IF_NOT_ALIGN
21217         and TYPE_WARN_IF_NOT_ALIGN.
21218         * stor-layout.c (do_type_align): Merge DECL_WARN_IF_NOT_ALIGN.
21219         (handle_warn_if_not_align): New.
21220         (place_union_field): Call handle_warn_if_not_align.
21221         (place_field): Call handle_warn_if_not_align.
21222         Copy TYPE_WARN_IF_NOT_ALIGN.
21223         (finish_builtin_struct): Copy TYPE_WARN_IF_NOT_ALIGN.
21224         (layout_type): Likewise.
21225         * tree-core.h (tree_type_common): Add warn_if_not_align.  Set
21226         spare to 18.
21227         (tree_decl_common): Add warn_if_not_align.
21228         * tree.c (build_range_type_1): Copy TYPE_WARN_IF_NOT_ALIGN.
21229         * tree.h (TYPE_WARN_IF_NOT_ALIGN): New.
21230         (SET_TYPE_WARN_IF_NOT_ALIGN): Likewise.
21231         (DECL_WARN_IF_NOT_ALIGN): Likewise.
21232         (SET_DECL_WARN_IF_NOT_ALIGN): Likewise.
21233         * doc/extend.texi: Document warn_if_not_aligned attribute.
21234         * doc/invoke.texi: Document -Wif-not-aligned and -Wpacked-not-aligned.
21236 2017-08-17  Martin Liska  <mliska@suse.cz>
21238         PR bootstrap/81864
21239         * tree-loop-distribution.c (ddrs_table): Change type to pointer type.
21240         (get_data_dependence): Use it as pointer type.
21241         (distribute_loop): Likewise.
21243 2017-08-17  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
21245         * config/rs6000/altivec.md (UNSPEC_VMRGOW_DIRECT): New constant.
21246         (p8_vmrgew_v4sf_direct): Generalize to p8_vmrgew_<mode>_direct.
21247         (p8_vmrgow_<mode>_direct): New define_insn.
21248         * config/rs6000/rs6000.c (altivec_expand_vec_perm_const): Properly
21249         handle endianness for vmrgew and vmrgow permute patterns.
21251 2017-08-17  Peter Bergner  <bergner@vnet.ibm.com>
21253         * config/rs6000/altivec.md (VParity): Remove TARGET_VSX_TIMODE.
21254         * config/rs6000/rs6000-cpus.def: Remove comment.
21255         (ISA_2_7_MASKS_SERVER): Delete OPTION_MASK_VSX_TIMODE;
21256         (POWERPC_MASKS): Likewise.
21257         * config/rs6000/rs6000.c (rs6000_hard_regno_mode_ok): Remove unneeded
21258         use of TARGET_VSX_TIMODE.
21259         (rs6000_setup_reg_addr_masks): Change TARGET_VSX_TIMODE to TARGET_VSX.
21260         (rs6000_init_hard_regno_mode_ok): Remove unneeded uses of
21261         TARGET_VSX_TIMODE.  Change use of TARGET_VSX_TIMODE to TARGET_VSX.
21262         (rs6000_option_override_internal): Remove dead code.
21263         (rs6000_legitimize_address): Change TARGET_VSX_TIMODE to TARGET_VSX.
21264         (rs6000_legitimize_reload_address): Likewise.
21265         (rs6000_legitimate_address_p): Likewise.
21266         (rs6000_opt_masks): Delete "vsx-timode".
21267         (rs6000_disable_incompatible_switches): Remove mention of -mvsx-timode
21268         from function comment.
21269         * config/rs6000/rs6000.h (MASK_VSX_TIMODE): Delete.
21270         * config/rs6000/rs6000.md (FMOVE128_GPR): Remove TARGET_VSX_TIMODE.
21271         (V16QI, V8HI, V4SI, V4SF, V2DI, V2DF, V1TI): Remove useless empty
21272         condition.
21273         * config/rs6000/rs6000.opt (mvsx-timode): Replace with stub.
21274         * config/rs6000/vector.md (VEC_IP): Remove TARGET_VSX_TIMODE.
21275         * config/rs6000/vsx.md (VSX_LE_128): Likewise.
21276         (VSX_TI): Likewise.
21277         (VSX_M): Likewise.
21278         (define_peephole2): Likewise.
21280 2017-08-17  Martin Sebor  <msebor@redhat.com>
21282         PR c/81859
21283         * pretty-print.c (pp_format): Use strnlen in %.*s to avoid reading
21284         past the end of an array.
21285         (test_pp_format): Add test cases.
21287 2017-08-17  Richard Sandiford  <richard.sandiford@linaro.org>
21289         * internal-fn.def (CLRSB, CLZ, CTZ, FFS, PARITY, POPCOUNT): Add
21290         missing ECF_NOTHROW flags.
21292 2017-08-17  Peter Bergner  <bergner@vnet.ibm.com>
21294         PR target/72804
21295         * config/rs6000/vsx.md (*vsx_le_permute_<mode>): Add support for
21296         operands residing in integer registers.
21297         (*vsx_le_perm_load_<mode>): Likewise.
21298         (*vsx_le_perm_store_<mode>): Likewise.
21299         (define_peephole2): Add peepholes to optimize the above.
21301 2017-08-17  Marek Polacek  <polacek@redhat.com>
21303         PR middle-end/81814
21304         * fold-const.c (operand_equal_for_comparison_p): Remove code that used
21305         to mimic what shorten_compare did.  Change the return type to bool.
21306         (fold_cond_expr_with_comparison): Update call to
21307         operand_equal_for_comparison_p.
21308         (fold_ternary_loc): Likewise.
21310 2017-08-17  Jackson Woodruff  <jackson.woodruff@arm.com>
21312         * aarch64-simd.md (mov<mode>): No longer force zero immediate into
21313         register.
21314         (*aarch64_simd_mov<mode>): Add new case for stp using zero immediate.
21316 2017-08-17  Richard Biener  <rguenther@suse.de>
21318         * tree-ssa-structalias.c (solve_graph): When propagating
21319         to successors update the graphs succ edges and avoid duplicate work.
21321 2017-08-17  Maxim Ostapenko  <m.ostapenko@samsung.com>
21323         PR target/81861
21324         * config/i386/i386.c (ix86_option_override_internal): Save target
21325         specific options after ix86_stack_protector_guard_reg was changed.
21327 2017-08-17  Richard Biener  <rguenther@suse.de>
21329         PR tree-optimization/81827
21330         * tree-ssa-structalias.c (struct variable_info): Add is_reg_var flag.
21331         (new_var_info): Initialize it conservatively.
21332         (get_call_vi): Mark register vars.
21333         (new_scalar_tmp_constraint_exp): Likewise.
21334         (handle_rhs_call): Likewise.
21335         (handle_const_call): Likewise.
21336         (create_function_info_for): Likewise.
21337         (solve_constraints): Sort varinfos to separate register from
21338         non-register vars to pack points-to solution bitmaps during
21339         iteration.
21341 2017-08-17  Marek Polacek  <polacek@redhat.com>
21343         * gimplify.c (gimplify_adjust_omp_clauses): Compare with 0 instead of 1.
21345 2017-08-17  Richard Biener  <rguenther@suse.de>
21347         * tree-vrp.c (vrp_int_const_binop): Do not set *overflow_p
21348         to true when overflow is undefined and we saturated the result.
21350 2017-08-17  Alan Modra  <amodra@gmail.com>
21352         PR target/80938
21353         * config/rs6000/rs6000.c (rs6000_savres_strategy): Revert 2017-08-09.
21354         Don't use store multiple if only one reg needs saving.
21355         (interesting_frame_related_regno): New function.
21356         (rs6000_frame_related): Don't emit frame info for regs that
21357         don't need saving.
21358         (rs6000_emit_epilogue): Likewise.
21360 2017-08-16  Nathan Sidwell  <nathan@acm.org>
21362         * tree-core.h (tree_type_non_common): Rename binfo to lang_1.
21363         * tree.h (TYPE_BINFO): Use type_non_common.maxval.
21364         (TYPE_LANG_SLOT_1): Use type_non_common.lang_1, for any type.
21365         * tree.c (free_lang_data_in_type): Use else-if chain.  Always
21366         clear TYPE_LANG_1.  Remove obsolete member-function stripping.
21367         (find_decls_types_r): Comment about TYPE_MAX_VALUES_RAW.
21368         (verify_type): Adjust for TYPE_BINFO move.
21369         * lto-streamer-out.c (DFS::DFS_write_tree_body): No need to
21370         process TYPE_BINFO directly.
21371         (hash_tree): Likewise.
21372         * tree-streamer-in.c (lto_input_ts_type_non_common_tree_pointers):
21373         Likewise.
21374         * tree-streamer-out.c (write_ts_type_non_common_tree_pointers):
21375         Likewise.
21377 2017-08-16  David Malcolm  <dmalcolm@redhat.com>
21379         * diagnostic-show-locus.c (colorizer::m_caret): Remove unused field.
21381 2017-08-16  Uros Bizjak  <ubizjak@gmail.com>
21383         PR target/46091
21384         * config/i386/i386.md (*anddi_1_btr): Change predicates of
21385         operand 0 and operand 1 to nomimmediate_operand. Add "m" constraint.
21386         Add ix86_binary_operator_ok to insn constraint.
21387         (*iordi_1_bts): Ditto.
21388         (*xordi_1_btc): Ditto.
21389         (*btsq): Change predicate of operand 0 to nonimmediate_operand.
21390         Update corresponding peephole2 pattern.
21391         (*btrq): Ditto.
21392         (*btcq): Ditto.
21394 2017-08-16  Bin Cheng  <bin.cheng@arm.com>
21396         PR tree-optimization/81832
21397         * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): Don't
21398         copy loop header which has IFN_LOOP_DIST_ALIAS call.
21400 2017-08-16  Marek Polacek  <polacek@redhat.com>
21402         PR middle/81695
21403         * fold-const.c (fold_indirect_ref_1): Restore original behavior
21404         regarding size_zero_node.
21406 2017-08-16  Martin Liska  <mliska@suse.cz>
21408         PR target/81753
21409         * config.gcc: Respect previously set extra_objs in case
21410         of darwin target.
21412 2017-08-16  Richard Sandiford  <richard.sandiford@linaro.org>
21414         PR tree-optimization/81835
21415         * tree-vect-loop.c (vect_is_simple_reduction): Simply checks for
21416         the phi SSA_NAME.  Check that the condition in a COND_EXPR does
21417         not depend on the phi.
21419 2017-08-16  Alan Modra  <amodra@gmail.com>
21421         * config/rs6000/rs6000.c (rs6000_init_hard_regno_mode_ok): Delete
21422         dead code.
21424 2017-08-16  Alan Modra  <amodra@gmail.com>
21426         * config/rs6000/rs6000.c (rs6000_reg_live_or_pic_offset_p): Merge..
21427         (save_reg_p): ..into this.  Update all callers.
21428         (first_reg_to_save): Simplify.
21430 2017-08-16  Alan Modra  <amodra@gmail.com>
21432         * config/rs6000/rs6000.c (rs6000_savres_strategy): Don't restore
21433         fixed regs.
21435 2017-08-15  Joseph Myers  <joseph@codesourcery.com>
21437         PR target/78460
21438         PR target/67712
21439         * config/sh/sh-mem.cc (sh_expand_cmpnstr): Only unroll for
21440         constant count if that count is less than 32.
21442 2017-08-15  Nathan Sidwell  <nathan@acm.org>
21444         * gcc.c (execute): Emit friendlier message if inferior is killed
21445         by an external cause.
21447 2017-08-15  Richard Biener  <rguenther@suse.de>
21449         PR tree-optimization/81790
21450         * tree-ssa-sccvn.c (vn_lookup_simplify_result): Handle both
21451         CONSTRUCTORs from simplifying and VN.
21453 2017-08-14  Martin Sebor  <msebor@redhat.com>
21455         * builtin-attrs.def: Add comments.
21457 2017-08-14  Martin Sebor  <msebor@redhat.com>
21459         PR c/81117
21460         * doc/extend.texi (attribute nonstring): Document new attribute.
21462 2017-08-14  Martin Sebor  <msebor@redhat.com>
21464         PR c/81117
21465         * tree-diagnostic.c (default_tree_printer): Handle %G.
21466         * gimple-pretty-print.h (percent_G_format): Declare new function.
21467         * gimple-pretty-print.c (percent_G_format): Define.
21468         * tree-pretty-print.c (percent_K_format): Add argument.
21470 2017-08-14  Martin Sebor  <msebor@redhat.com>
21472         PR translation/79998
21473         * gimple-ssa-sprintf.c (pass_sprintf_length::handle_gimple_call):
21474         Remove a stray space.
21476 2017-08-14  Uros Bizjak  <ubizjak@gmail.com>
21478         PR target/46091
21479         * config/i386/i386.md (*anddi_1_btr): New insn_and_split pattern.
21480         (*iordi_1_bts): Ditto.
21481         (*xordi_1_btc): Ditto.
21483 2017-08-14  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
21485         PR target/79845
21486         * config/rs6000/linux64.h (INVALID_64BIT): Use quoted strings.
21487         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
21488         Likewise.
21489         * config/rs6000/rs6000.c (rs6000_init_hard_regno_mode_ok): Use
21490         quoted strings, and make more translator-friendly.
21491         (darwin_rs6000_override_options): Likewise.
21492         (rs6000_option_override_internal): Likewise.
21493         (rs6000_return_in_memory): Fix overlong line.
21494         (init_cmulative_args): Use quoted strings, and make more
21495         translator-friendly.
21496         (rs6000_pass_by_reference): Fix overlong line.
21497         (def_builtin): Use quoted strings.
21498         (altivec_expand_predicate_builtin): Use quoted strings, and make
21499         more translator-friendly.
21500         (htm_expand_builtin): Use quoted strings.
21501         (cpu_expand_builtin): Use quoted strings, and make more
21502         translator-friendly.
21503         (altivec_expand_builtin): Likewise.
21504         (paired_expand_predicate_builtin): Likewise.
21505         (rs6000_invalid_builtin): Likewise.
21506         (builtin_function_type): Use quoted strings.
21507         (rs6000_expand_split_stack_prologue): Use quoted strings, and make
21508         more translator-friendly.
21509         (rs6000_trampoline_init): Likewise.
21510         (rs6000_handle_altivec_attribute): Likewise.
21511         (rs6000_inner_target_options): Use quoted strings.
21512         (rs6000_disable_incompatible_switches): Likewise.
21513         * config/rs6000/sysv4.h (SUBTARGET_OVERRIDE_OPTIONS): Use quoted
21514         strings, and make more translator-friendly.
21515         (SUBSUBTARGET_OVERRIDE_OPTIONS): Use quoted strings.
21517 2017-08-14  Bin Cheng  <bin.cheng@arm.com>
21519         PR tree-optimization/81799
21520         * tree-loop-distribution.c (version_loop_by_alias_check): Force
21521         cond_expr to simple gimple operand.
21523 2017-08-14  Wilco Dijkstra  <wdijkstr@arm.com>
21525         PR middle-end/46932
21526         * doc/sourcebuild.texi (autoincdec): Add autoincdec description.
21528 2017-08-14  Georg-Johann Lay  <avr@gjlay.de>
21530         PR target/81754
21531         PR target/81268
21532         * config/avr/avr.opt (mgas-isr-prologues): New Var avr_gasisr_prologues.
21533         * config/avr/avr.md (gasisr, *gasisr): Use it instead of
21534         TARGET_GASISR_PROLOGUES.
21535         * config/avr/avr.c (avr_option_override): Same.
21536         (avr_pass_pre_proep::execute): Same.
21538 2017-08-13  H.J. Lu  <hongjiu.lu@intel.com>
21540         PR target/81820
21541         * config/i386/i386.c (ix86_finalize_stack_frame_flags): Replace
21542         frame pointer with stack pointer - UNITS_PER_WORD in debug insns.
21544 2017-08-13  Uros Bizjak  <ubizjak@gmail.com>
21546         * config/i386/i386.md (*load_tp_<mode>): Redefine as
21547         define_insn_and_split.  Split to a memory load from 0 in
21548         DEFAULT_TLS_SEG_REG address space.  Merge with *load_tp_x32
21549         using PTR mode iterator.
21550         (*load_tp_x32_zext"): Redefine as define_insn_and_split.
21551         Split to a memory load from 0 in DEFAULT_TLS_SEG_REG address space.
21552         (*add_tp_<mode>): Redefine as define_insn_and_split.
21553         Split to an add with a memory load from 0 in DEFAULT_TLS_SEG_REG
21554         address space.  Merge with *add_tp_x32 using PTR mode iterator.
21555         (*add_tp_x32_zext"): Redefine as define_insn_and_split.
21556         Split to an add with a  memory load from 0 in
21557         DEFAULT_TLS_SEG_REG address space.
21559 2017-08-12  Andrew Pinski  <apinski@cavium.com>
21561         * config/aarch64/aarch64-option-extensions.def (rdma):
21562         Fix feature string to what Linux prints out in /proc/cpuinfo.
21564 2017-08-12  Pierre-Marie de Rodat  <derodat@adacore.com>
21566         PR ada/79542
21567         * dwarf2out.c (modified_type_die): For C typedef types that have
21568         an ultimate origin, process the ultimate origin instead of the
21569         input type.
21570         (gen_typedef_die): Assert that input DECLs have no ultimate
21571         origin.
21572         (gen_type_die_with_usage): For typedef variants that have an
21573         ultimate origin, just call gen_decl_die on the original DECL.
21574         (process_scope_var): Avoid creating DIEs for local typedefs and
21575         concrete static variables.
21577 2017-08-12  Alan Modra  <amodra@gmail.com>
21579         PR target/81170
21580         PR target/81295
21581         * config/rs6000/sysv4.h (STARTFILE_LINUX_SPEC): Upgrade to
21582         match gnu-user.h startfile.
21583         (ENDFILE_LINUX_SPEC): Similarly.
21585 2017-08-11  Thomas Schwinge  <thomas@codesourcery.com>
21587         PR lto/81430
21588         * config/nvptx/nvptx.c (nvptx_override_options_after_change):
21589         Remove function.
21590         (TARGET_OVERRIDE_OPTIONS_AFTER_CHANGE): Remove definition.
21592 2017-08-11  Tamar Christina  <tamar.christina@arm.com>
21593         * config/aarch64/aarch64.md (mov<mode>): Change.
21594         (*movhf_aarch64, *movsf_aarch64, *movdf_aarch64):
21595         aarch64_reg_or_fp_float into aarch64_reg_or_fp_zero.
21596         * config/aarch64/predicates.md (aarch64_reg_or_fp_float): Removed.
21598 2017-08-11  Eric Botcazou  <ebotcazou@adacore.com>
21600         * tree-sra.c (build_access_from_expr_1): Use more precise diagnostics
21601         for storage order barriers.
21603 2017-08-11  Martin Liska  <mliska@suse.cz>
21605         PR tree-opt/79987
21606         * tree-chkp.c (chkp_get_bounds_for_decl_addr): Do not instrument
21607         variables of void type.
21609 2017-08-11  Martin Liska  <mliska@suse.cz>
21611         * asan.c (asan_protect_global): Replace ASM_OUTPUT_DEF with
21612         TARGET_SUPPORTS_ALIASES.
21613         * cgraph.c (cgraph_node::create_same_body_alias): Likewise.
21614         * ipa-visibility.c (can_replace_by_local_alias): Likewise.
21615         (optimize_weakref): Likewise.
21616         * symtab.c (symtab_node::noninterposable_alias): Likewise.
21617         * varpool.c (varpool_node::create_extra_name_alias): Likewise.
21618         * defaults.h: Introduce TARGET_SUPPORTS_ALIASES.
21620 2017-08-11  Martin Liska  <mliska@suse.cz>
21622         PR ipa/81213
21623         * config/i386/i386.c (make_resolver_func): Do complete
21624         refactoring of the function.
21626 2017-08-10  Uros Bizjak  <ubizjak@gmail.com>
21628         PR target/81708
21629         * config/i386/i386.opt (mstack-protector-guard-symbol=): New option
21630         * config/i386/i386.c (ix86_stack_protect_guard): Use
21631         ix86_stack_protect_guard_symbol_str to generate varible declaration.
21632         * doc/invoke.texi (x86 Options): Document
21633         -mstack-protector-guard-symbol= option.
21635 2017-08-10  Uros Bizjak  <ubizjak@gmail.com>
21637         * config/i386/i386-protos.h (ix86_split_stack_guard): New prototype.
21638         * config/i386/i386.c (ix86_split_stack_guard): New function.
21639         (ix86_xpand_split_stack_prologue): Call ix86_split_stack_guard.
21640         (ix86_legitimate_address_p) <case UNSPEC_STACK_CHECK>: Remove.
21641         (i386_asm_output_addr_const_extra) <case UNSPEC_STACK_CHECK>: Ditto.
21642         (optput_pic_addr_const): Remove UNSPEC_STACK_CHECK handling.
21643         * config/i386/i386.md (unspec): Remove UNSPEC_STACK_CHECK.
21644         (split_stack_space_check): Call ix86_split_stack_guard.
21646 2017-08-10  Martin Sebor  <msebor@redhat.com>
21648         * print-tree.c (print_node): Print location using the established
21649         format %s:%i%i.
21650         Replace spaces with colons.
21651         (debug_raw, debug): Ditto.
21653 2017-08-10  Martin Sebor  <msebor@redhat.com>
21655         PR c++/81586
21656         * pretty-print.c (pp_format): Correct the handling of %s precision.
21658 2017-08-10  H.J. Lu  <hongjiu.lu@intel.com>
21660         PR target/81736
21661         * config/i386/i386.c (ix86_finalize_stack_realign_flags): Renamed
21662         to ...
21663         (ix86_finalize_stack_frame_flags): This.  Also clear
21664         frame_pointer_needed if -fno-omit-frame-pointer is used without
21665         stack access.
21666         (ix86_expand_prologue): Replace ix86_finalize_stack_realign_flags
21667         with ix86_finalize_stack_frame_flags.
21668         (ix86_expand_epilogue): Likewise.
21669         (ix86_expand_split_stack_prologue): Likewise.
21670         * doc/invoke.texi: Add a note for -fno-omit-frame-pointer.
21672 2017-08-10  Martin Liska  <mliska@suse.cz>
21674         PR c++/81355
21675         * c-attribs.c (handle_target_attribute):
21676         Report warning for an empty string argument of target attribute.
21678 2017-08-09  Jakub Jelinek  <jakub@redhat.com>
21680         PR c/81687
21681         * omp-low.c (omp_copy_decl): Don't remap FORCED_LABEL or DECL_NONLOCAL
21682         LABEL_DECLs.
21683         * tree-cfg.c (move_stmt_op): Don't adjust DECL_CONTEXT of FORCED_LABEL
21684         or DECL_NONLOCAL labels.
21685         (move_stmt_r) <case GIMPLE_LABEL>: Adjust DECL_CONTEXT of FORCED_LABEL
21686         or DECL_NONLOCAL labels here.
21688 2017-08-09  Will Schmidt  <will_schmidt@vnet.ibm.com>
21690         * config/rs6000/rs6000.c (rs6000_option_override_internal): Add blurb
21691         to indicate when early gimple folding has been disabled.
21692         (rs6000_gimple_fold_builtin): Add debug content.
21693         (rs6000_invalid_builtin): Fix whitespace.
21694         (rs6000_expand_builtin): Fix whitespace.
21695         * config/rs6000/rs6000.opt: Add option for -mfold-gimple.
21697 2017-08-09  Segher Boessenkool  <segher@kernel.crashing.org>
21699         PR target/80938
21700         * config/rs6000/rs6000.c (rs6000_savres_strategy): Don't use
21701         SAVE_MULTIPLE if not all the registers that saves, should be saved.
21703 2017-08-09  Jim Wilson  <jim.wilson@linaro.org>
21705         * config/aarch64/aarch64-cores.def (falkor): Use falkor pipeline.
21706         (qdf24xx): Likewise.
21707         * config/aarch64/aarch64.md: Include falkor.md.
21708         * config/aarch64/falkor.md: New.
21710 2017-08-09  Marek Polacek  <polacek@redhat.com>
21712         PR c/81233
21713         * diagnostic-core.h (emit_diagnostic_valist): Add declaration.
21714         * diagnostic.c (emit_diagnostic): Add a comment.
21715         (emit_diagnostic_valist): New function.
21717 2017-08-09  Marek Polacek  <polacek@redhat.com>
21719         PR c/81417
21720         * input.c (make_location): New overload.
21721         * input.h (make_location): Declare.
21723 2017-08-08  Alan Modra  <amodra@gmail.com>
21724             H.J. Lu  <hongjiu.lu@intel.com>
21726         PR driver/81523
21727         * gcc.c (NO_PIE_SPEC): Delete.
21728         (PIE_SPEC): Define as !no-pie/pie.  Move static|shared|r
21729         exclusion..
21730         (LINK_PIE_SPEC): ..to here.
21731         (LINK_COMMAND_SPEC): Support -no-pie.
21732         * config/gnu-user.h (GNU_USER_TARGET_STARTFILE_SPEC): Correct
21733         chain of crtbegin*.o selection, update for PIE_SPEC changes and
21734         format.
21735         (GNU_USER_TARGET_ENDFILE_SPEC): Similarly.
21736         * config/sol2.h (STARTFILE_CRTBEGIN_SPEC): Similarly.
21737         (ENDFILE_CRTEND_SPEC): Similarly.
21739 2017-08-08  Uros Bizjak  <ubizjak@gmail.com>
21741         PR target/81708
21742         * config/i386/i386.opt (mstack-protector-guard-reg=): New option
21743         (mstack-protector-guard-offset=): Ditto.
21744         * config/i386/i386.c (ix86_option_override): Handle
21745         -mstack-protector-guard-reg= and -mstack-protector-guard-offset=
21746         options.
21747         (ix86_stack_protect_guard): Use ix86_stack_protect_guard_reg and
21748         ix86_stack_protect_guard_offset variables.
21749         (TARGET_STACK_PROTECT_GUARD): Always define.
21750         * doc/invoke.texi (x86 Options): Document -mstack-protector-guard-reg=
21751         and -mstack-protector-guard-offset= options.
21753 2017-08-08  Bin Cheng  <bin.cheng@arm.com>
21755         * tree-ssa-loop-ivopts.c (relate_compare_use_with_all_cands): Handle
21756         boundary case for the last candidate.
21758 2017-08-08  Bin Cheng  <bin.cheng@arm.com>
21760         * doc/invoke.texi: Document -ftree-loop-distribution for O3.
21761         * opts.c (default_options_table): Add OPT_ftree_loop_distribution.
21763 2017-08-08  Tamar Christina  <tamar.christina@arm.com>
21765         PR middle-end/19706
21766         * config/aarch64/aarch64.md (xorsign<mode>3): New optabs.
21767         * config/aarch64/aarch64-builtins.c
21768         (aarch64_builtin_vectorized_function): Added CASE_CFN_XORSIGN.
21769         * config/aarch64/aarch64-simd-builtins.def: Added xorsign BINOP.
21770         * config/aarch64/aarch64-simd.md: Added xorsign<mode>3
21772 2017-08-08  Tamar Christina  <tamar.christina@arm.com>
21773             Andrew Pinski <pinskia@gmail.com>
21775         PR middle-end/19706
21776         * internal-fn.def (XORSIGN): New.
21777         * optabs.def (xorsign_optab): New.
21778         * tree-ssa-math-opts.c (is_copysign_call_with_1): New.
21779         (convert_expand_mult_copysign): New.
21780         (pass_optimize_widening_mul::execute): Call
21781         convert_expand_mult_copysign.
21783 2017-08-08  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
21785         PR tree-optimization/81354
21786         * gimple-ssa-strength-reduction.c (create_add_on_incoming_edge):
21787         Insert on edges rather than explicitly creating landing pads.
21788         (analyze_candidates_and_replace): Commit edge inserts.
21790 2017-08-08  Richard Biener  <rguenther@suse.de>
21792         PR middle-end/81719
21793         * tree-ssa-loop-niter.c: Include tree-dfa.h.
21794         (expand_simple_operations): Also look through ADDR_EXPRs with
21795         MEM_REF bases treating them as POINTER_PLUS_EXPR.
21797 2017-08-08  Richard Biener  <rguenther@suse.de>
21799         PR tree-optimization/81723
21800         * tree-vect-slp.c (struct bst_traits): New hash traits.
21801         (bst_fail): New global.
21802         (vect_build_slp_tree_2): New worker, split out from ...
21803         (vect_build_slp_tree): ... this now wrapping it with using
21804         bst_fail set to cache SLP tree build fails.  Properly handle
21805         max_tree_size.
21806         (vect_analyze_slp_instance): Allocate and free bst_fail.
21808 2017-08-08  Martin Liska  <mliska@suse.cz>
21810         PR tree-opt/81696
21811         * ipa-icf-gimple.c (func_checker::compare_cst_or_decl): Consider
21812         LABEL_DECLs that can be from a different function.
21814 2017-08-08  Bin Cheng  <bin.cheng@arm.com>
21816         PR tree-optimization/81744
21817         * tree-predcom.c (prepare_finalizers_chain): Deep copy expr of
21818         loop's number of iterations.
21820 2017-08-08  Martin Liska  <mliska@suse.cz>
21822         * asan.c: Include header files.
21823         * attribs.c (build_decl_attribute_variant): New function moved
21824         from tree.[ch].
21825         (build_type_attribute_qual_variant): Likewise.
21826         (cmp_attrib_identifiers): Likewise.
21827         (simple_cst_list_equal): Likewise.
21828         (omp_declare_simd_clauses_equal): Likewise.
21829         (attribute_value_equal): Likewise.
21830         (comp_type_attributes): Likewise.
21831         (build_type_attribute_variant): Likewise.
21832         (lookup_ident_attribute): Likewise.
21833         (remove_attribute): Likewise.
21834         (merge_attributes): Likewise.
21835         (merge_type_attributes): Likewise.
21836         (merge_decl_attributes): Likewise.
21837         (merge_dllimport_decl_attributes): Likewise.
21838         (handle_dll_attribute): Likewise.
21839         (attribute_list_equal): Likewise.
21840         (attribute_list_contained): Likewise.
21841         * attribs.h (lookup_attribute): New function moved from tree.[ch].
21842         (lookup_attribute_by_prefix): Likewise.
21843         * bb-reorder.c: Include header files.
21844         * builtins.c: Likewise.
21845         * calls.c: Likewise.
21846         * cfgexpand.c: Likewise.
21847         * cgraph.c: Likewise.
21848         * cgraphunit.c: Likewise.
21849         * convert.c: Likewise.
21850         * dwarf2out.c: Likewise.
21851         * final.c: Likewise.
21852         * fold-const.c: Likewise.
21853         * function.c: Likewise.
21854         * gimple-expr.c: Likewise.
21855         * gimple-fold.c: Likewise.
21856         * gimple-pretty-print.c: Likewise.
21857         * gimple.c: Likewise.
21858         * gimplify.c: Likewise.
21859         * hsa-common.c: Likewise.
21860         * hsa-gen.c: Likewise.
21861         * internal-fn.c: Likewise.
21862         * ipa-chkp.c: Likewise.
21863         * ipa-cp.c: Likewise.
21864         * ipa-devirt.c: Likewise.
21865         * ipa-fnsummary.c: Likewise.
21866         * ipa-inline.c: Likewise.
21867         * ipa-visibility.c: Likewise.
21868         * ipa.c: Likewise.
21869         * lto-cgraph.c: Likewise.
21870         * omp-expand.c: Likewise.
21871         * omp-general.c: Likewise.
21872         * omp-low.c: Likewise.
21873         * omp-offload.c: Likewise.
21874         * omp-simd-clone.c: Likewise.
21875         * opts-global.c: Likewise.
21876         * passes.c: Likewise.
21877         * predict.c: Likewise.
21878         * sancov.c: Likewise.
21879         * sanopt.c: Likewise.
21880         * symtab.c: Likewise.
21881         * toplev.c: Likewise.
21882         * trans-mem.c: Likewise.
21883         * tree-chkp.c: Likewise.
21884         * tree-eh.c: Likewise.
21885         * tree-into-ssa.c: Likewise.
21886         * tree-object-size.c: Likewise.
21887         * tree-parloops.c: Likewise.
21888         * tree-profile.c: Likewise.
21889         * tree-ssa-ccp.c: Likewise.
21890         * tree-ssa-live.c: Likewise.
21891         * tree-ssa-loop.c: Likewise.
21892         * tree-ssa-sccvn.c: Likewise.
21893         * tree-ssa-structalias.c: Likewise.
21894         * tree-ssa.c: Likewise.
21895         * tree-streamer-in.c: Likewise.
21896         * tree-vectorizer.c: Likewise.
21897         * tree-vrp.c: Likewise.
21898         * tsan.c: Likewise.
21899         * ubsan.c: Likewise.
21900         * varasm.c: Likewise.
21901         * varpool.c: Likewise.
21902         * tree.c: Remove functions moved to attribs.[ch].
21903         * tree.h: Likewise.
21904         * config/aarch64/aarch64.c: Add attrs.h header file.
21905         * config/alpha/alpha.c: Likewise.
21906         * config/arc/arc.c: Likewise.
21907         * config/arm/arm.c: Likewise.
21908         * config/avr/avr.c: Likewise.
21909         * config/bfin/bfin.c: Likewise.
21910         * config/c6x/c6x.c: Likewise.
21911         * config/cr16/cr16.c: Likewise.
21912         * config/cris/cris.c: Likewise.
21913         * config/darwin.c: Likewise.
21914         * config/epiphany/epiphany.c: Likewise.
21915         * config/fr30/fr30.c: Likewise.
21916         * config/frv/frv.c: Likewise.
21917         * config/ft32/ft32.c: Likewise.
21918         * config/h8300/h8300.c: Likewise.
21919         * config/i386/winnt.c: Likewise.
21920         * config/ia64/ia64.c: Likewise.
21921         * config/iq2000/iq2000.c: Likewise.
21922         * config/lm32/lm32.c: Likewise.
21923         * config/m32c/m32c.c: Likewise.
21924         * config/m32r/m32r.c: Likewise.
21925         * config/m68k/m68k.c: Likewise.
21926         * config/mcore/mcore.c: Likewise.
21927         * config/microblaze/microblaze.c: Likewise.
21928         * config/mips/mips.c: Likewise.
21929         * config/mmix/mmix.c: Likewise.
21930         * config/mn10300/mn10300.c: Likewise.
21931         * config/moxie/moxie.c: Likewise.
21932         * config/msp430/msp430.c: Likewise.
21933         * config/nds32/nds32-isr.c: Likewise.
21934         * config/nds32/nds32.c: Likewise.
21935         * config/nios2/nios2.c: Likewise.
21936         * config/nvptx/nvptx.c: Likewise.
21937         * config/pa/pa.c: Likewise.
21938         * config/pdp11/pdp11.c: Likewise.
21939         * config/powerpcspe/powerpcspe.c: Likewise.
21940         * config/riscv/riscv.c: Likewise.
21941         * config/rl78/rl78.c: Likewise.
21942         * config/rx/rx.c: Likewise.
21943         * config/s390/s390.c: Likewise.
21944         * config/sh/sh.c: Likewise.
21945         * config/sol2.c: Likewise.
21946         * config/sparc/sparc.c: Likewise.
21947         * config/spu/spu.c: Likewise.
21948         * config/stormy16/stormy16.c: Likewise.
21949         * config/tilegx/tilegx.c: Likewise.
21950         * config/tilepro/tilepro.c: Likewise.
21951         * config/v850/v850.c: Likewise.
21952         * config/vax/vax.c: Likewise.
21953         * config/visium/visium.c: Likewise.
21954         * config/xtensa/xtensa.c: Likewise.
21956 2017-08-07  Michael Meissner  <meissner@linux.vnet.ibm.com>
21958         PR target/81593
21959         * config/rs6000/vsx.md (vsx_concat_<mode>, VSX_D): Cleanup
21960         constraints since the -mupper-regs-* switches have been
21961         eliminated.
21962         (vsx_concat_<mode>_1): New combiner insns to recognize inserting
21963         into a vector from a double word element that was extracted from
21964         another vector, and eliminate extra XXPERMDI instructions.
21965         (vsx_concat_<mode>_2): Likewise.
21966         (vsx_concat_<mode>_3): Likewise.
21967         (vsx_set_<mode>, VSX_D): Rewrite vector set in terms of vector
21968         concat to allow optimizing inserts from previous extracts.
21970 2017-08-07  Uros Bizjak  <ubizjak@gmail.com>
21972         * config/i386/i386.c (ix86_stack_protect_guard): Generate
21973         memory reference to a SSP offset in TLS address space.
21974         (ix86_print_operand) <case '@'>: Remove.
21975         (ix86_print_operand_punct_valid_p): Remove '@' code.
21976         * config/i386/i386.md (unspec): Remove UNSPEC_SP_TLS_SET and
21977         UNSPEC_SP_TLS_TEST.
21978         (stack_tls_protect_set_<mode>): Remove.
21979         (stack_protect_set): Do not call gen_stack_tls_protect_set_<mode>.
21980         (stack_tls_protect_test_<mode>): Remove.
21981         (stack_protect_test): Do not call gen_stack_tls_protect_test_<mode>.
21983 2017-08-07  Olivier Hainque  <hainque@adacore.com>
21985         PR target/81755
21986         * config/vxworksae.h (VXWORKS_HAVE_TLS): Define.
21988 2017-08-07  Douglas Rupp  <rupp@adacore.com>
21990         * Makefile.in (install-mkheaders): Fix typo, where the multi_dir
21991         variable was referenced as multidir in command.
21993 2017-08-07  Jakub Jelinek  <jakub@redhat.com>
21995         PR c/69389
21996         * gimplify.c (goa_stabilize_expr): Handle BIT_INSERT_EXPR and
21997         BIT_FIELD_REF.
21999 2017-08-07  Martin Liska  <mliska@suse.cz>
22001         * config/m32c/m32c.c: Add include of stringpool.h and attribs.h.
22002         * config/rl78/rl78.c: Add include of attribs.h.
22003         * config/sh/sh.c: Likewise.
22004         * config/v850/v850.c: Likewise.
22006 2017-08-07  Tom de Vries  <tom@codesourcery.com>
22008         PR middle-end/78266
22009         * omp-expand.c (expand_oacc_for): Ensure diff_type is large enough.
22011 2017-08-07  Martin Liska  <mliska@suse.cz>
22013         * config/mips/mips.c: Include attribs.h.
22015 2017-08-07  Thomas Koenig  <tkoenig@gcc.gnu.org>
22017         PR fortran/68829
22018         * doc/invoke.texi: Document change in behvaior for -Ofast for
22019         Fortran.
22021 2017-08-07  Wilco Dijkstra  <wdijkstr@arm.com>
22023         * config/aarch64/aarch64.c (aarch64_pushwb_single_reg):
22024         Use gen_frame_mem.
22025         (aarch64_pop_regs): Likewise.
22026         (aarch64_gen_load_pair): Likewise.
22027         (aarch64_save_callee_saves): Likewise.
22028         (aarch64_restore_callee_saves): Likewise.
22030 2017-08-07  H.J. Lu  <hongjiu.lu@intel.com>
22032         * config/i386/i386.c: Revert the last change.
22034 2017-08-07  H.J. Lu  <hongjiu.lu@intel.com>
22036         PR target/81736
22037         * config/i386/i386.c (ix86_finalize_stack_realign_flags): Renamed
22038         to ...
22039         (ix86_finalize_stack_frame_flags): This.  Also clear
22040         frame_pointer_needed if -fno-omit-frame-pointer is used without
22041         stack access.
22042         (ix86_expand_prologue): Replace ix86_finalize_stack_realign_flags
22043         with ix86_finalize_stack_frame_flags.
22044         (ix86_expand_epilogue): Likewise.
22045         (ix86_expand_split_stack_prologue): Likewise.
22047 2017-08-07  H.J. Lu  <hongjiu.lu@intel.com>
22049         PR target/81743
22050         * config/i386/i386.c (get_builtin_code_for_version): Set priority
22051         to P_AES for Westmere.
22053 2017-08-07  Jonathan Yong  <10walls@gmail.com>
22055         * config/i386/mingw.opt (fset-stack-executable): Removed.
22056         * config/i386/cygming.opt (fset-stack-executable): Moved
22057         from mingw.opt.
22058         * config/i386/cygwin.h: Define CHECK_EXECUTE_STACK_ENABLED.
22060 2017-08-07  Segher Boessenkool  <segher@kernel.crashing.org>
22062         * print-rtl.c (print_exp): Print NOT as "~" instead of as "!".
22064 2017-08-07  Marek Polacek  <polacek@redhat.com>
22066         PR middle-end/81737
22067         * fold-const.c (fold_indirect_ref_1): Check type_domain.
22069 2017-08-07  Martin Liska  <mliska@suse.cz>
22071         * attribs.h (canonicalize_attr_name): New function.
22072         (cmp_attribs): Move from c-format.c and adjusted.
22073         (is_attribute_p): Moved from tree.h.
22074         * tree-inline.c: Add new includes.
22075         * tree.c (cmp_attrib_identifiers): Use cmp_attribs.
22076         (private_is_attribute_p): Remove.
22077         (private_lookup_attribute): Likewise.
22078         (private_lookup_attribute_by_prefix): Simplify.
22079         (remove_attribute): Use is_attribute_p.
22080         * tree.h: Remove removed declarations.
22082 2017-08-07  Jakub Jelinek  <jakub@redhat.com>
22084         PR middle-end/81698
22085         * stmt.c (emit_case_dispatch_table): Add DEFAULT_EDGE argument,
22086         instead of computing it in the function.  Formatting fix.
22087         (expand_case): Don't rely on default_edge being the first edge,
22088         clear it if removing it, pass default_edge to
22089         emit_case_dispatch_table.
22090         (expand_sjlj_dispatch_table): Pass NULL as DEFAULT_EDGE, formatting
22091         fix.
22093 2017-08-06  Uros Bizjak  <ubizjak@gmail.com>
22095         * config/alpha/alpha.c (alpha_reorg): If trap is the last active
22096         insn in the function, emit NOP after the insn.
22098 2017-08-06  Tom de Vries  <tom@codesourcery.com>
22100         * omp-expand.c (expand_oacc_for): Add missing edge probability for tile
22101         and element loops.
22103 2017-08-06  Tom de Vries  <tom@codesourcery.com>
22105         * omp-expand.c (expand_oacc_for): Add missing edge probability for chunk
22106         loop.
22108 2017-08-04  Yury Gribov  <tetra2005@gmail.com>
22110         PR tree-optimization/57371
22111         * match.pd: New pattern.
22113 2017-08-04  Marek Polacek  <polacek@redhat.com>
22115         PR middle-end/81695
22116         * fold-const.c (fold_indirect_ref_1): For ((int *)&a + 4 -> a[1],
22117         perform the computation in offset_int.
22119 2017-08-04  Richard Sandiford  <richard.sandiford@linaro.org>
22121         PR tree-optimization/81136
22122         * tree-vectorizer.h: Include tree-hash-traits.h.
22123         (vec_base_alignments): New typedef.
22124         (vec_info): Add a base_alignments field.
22125         (vect_record_base_alignments): Declare.
22126         * tree-data-ref.h (data_reference): Add an is_conditional_in_stmt
22127         field.
22128         (DR_IS_CONDITIONAL_IN_STMT): New macro.
22129         (create_data_ref): Add an is_conditional_in_stmt argument.
22130         * tree-data-ref.c (create_data_ref): Likewise.  Use it to initialize
22131         the is_conditional_in_stmt field.
22132         (data_ref_loc): Add an is_conditional_in_stmt field.
22133         (get_references_in_stmt): Set the is_conditional_in_stmt field.
22134         (find_data_references_in_stmt): Update call to create_data_ref.
22135         (graphite_find_data_references_in_stmt): Likewise.
22136         * tree-ssa-loop-prefetch.c (determine_loop_nest_reuse): Likewise.
22137         * tree-vect-data-refs.c (vect_analyze_data_refs): Likewise.
22138         (vect_record_base_alignment): New function.
22139         (vect_record_base_alignments): Likewise.
22140         (vect_compute_data_ref_alignment): Adjust base_addr and aligned_to
22141         for nested statements even if we fail to compute a misalignment.
22142         Use pooled base alignments for unconditional references.
22143         (vect_find_same_alignment_drs): Compare base addresses instead
22144         of base objects.
22145         (vect_analyze_data_refs_alignment): Call vect_record_base_alignments.
22146         * tree-vect-slp.c (vect_slp_analyze_bb_1): Likewise.
22148 2017-08-04  Richard Sandiford  <richard.sandiford@linaro.org>
22150         * tree-vectorizer.h (vec_info): Add a constructor and destructor.
22151         Add an explicit name for the enum.  Use auto_vec for slp_instances
22152         and grouped_stores.
22153         (_loop_vec_info): Add a constructor and destructor.  Use auto_vec
22154         for all vectors.
22155         (_bb_vec_info): Add a constructor and destructor.
22156         (vinfo_for_stmt): Return NULL for uids of -1 as well.
22157         (destroy_loop_vec_info): Delete.
22158         (vect_destroy_datarefs): Likewise.
22159         * tree-vectorizer.c (vect_destroy_datarefs): Delete.
22160         (vec_info::vec_info): New function.
22161         (vec_info::~vec_info): Likewise.
22162         (vectorize_loops): Use delete instead of destroy_loop_vec_info.
22163         * tree-parloops.c (gather_scalar_reductions): Use delete instead of
22164         destroy_loop_vec_info.
22165         * tree-vect-loop.c (new_loop_vec_info): Replace with...
22166         (_loop_vec_info::_loop_vec_info): ...this.
22167         (destroy_loop_vec_info): Replace with...
22168         (_loop_vec_info::~_loop_vec_info): ...this.  Unconditionally delete
22169         the stmt_vec_infos.  Leave handling of vec_info information to its
22170         destructor.  Remove explicit vector releases.
22171         (vect_analyze_loop_form): Use new instead of new_loop_vec_info.
22172         (vect_analyze_loop): Use delete instead of destroy_loop_vec_info.
22173         * tree-vect-slp.c (new_bb_vec_info): Replace with...
22174         (_bb_vec_info::_bb_vec_info): ...this.  Don't reserve space in
22175         BB_VINFO_GROUPED_STORES or BB_VINFO_SLP_INSTANCES.
22176         (destroy_bb_vec_info): Replace with...
22177         (_bb_vec_info::~_bb_vec_info): ...this.  Leave handling of vec_info
22178         information to its destructor.
22179         (vect_slp_analyze_bb_1): Use new and delete instead of
22180         new_bb_vec_info and destroy_bb_vec_info.
22181         (vect_slp_bb): Replace 2 calls to destroy_bb_vec_info with a
22182         single delete.
22184 2017-08-04  Richard Sandiford  <richard.sandiford@linaro.org>
22186         * tree-data-ref.h (subscript): Add access_fn field.
22187         (data_dependence_relation): Add could_be_independent_p.
22188         (SUB_ACCESS_FN, DDR_COULD_BE_INDEPENDENT_P): New macros.
22189         (same_access_functions): Move to tree-data-ref.c.
22190         * tree-data-ref.c (ref_contains_union_access_p): New function.
22191         (access_fn_component_p): Likewise.
22192         (access_fn_components_comparable_p): Likewise.
22193         (dr_analyze_indices): Add a reference to access_fn_component_p.
22194         (dump_data_dependence_relation): Use SUB_ACCESS_FN instead of
22195         DR_ACCESS_FN.
22196         (constant_access_functions): Likewise.
22197         (add_other_self_distances): Likewise.
22198         (same_access_functions): Likewise.  (Moved from tree-data-ref.h.)
22199         (initialize_data_dependence_relation): Use XCNEW and remove
22200         explicit zeroing of DDR_REVERSED_P.  Look for a subsequence
22201         of access functions that have the same type.  Allow the
22202         subsequence to end with different bases in some circumstances.
22203         Record the chosen access functions in SUB_ACCESS_FN.
22204         (build_classic_dist_vector_1): Replace ddr_a and ddr_b with
22205         a_index and b_index.  Use SUB_ACCESS_FN instead of DR_ACCESS_FN.
22206         (subscript_dependence_tester_1): Likewise dra and drb.
22207         (build_classic_dist_vector): Update calls accordingly.
22208         (subscript_dependence_tester): Likewise.
22209         * tree-ssa-loop-prefetch.c (determine_loop_nest_reuse): Check
22210         DDR_COULD_BE_INDEPENDENT_P.
22211         * tree-vectorizer.h (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Test
22212         comp_alias_ddrs instead of may_alias_ddrs.
22213         * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr):
22214         New function.
22215         (vect_analyze_data_ref_dependence): Use it if
22216         DDR_COULD_BE_INDEPENDENT_P, but fall back to using the recorded
22217         distance vectors if that fails.
22218         (dependence_distance_ge_vf): New function.
22219         (vect_prune_runtime_alias_test_list): Use it.  Don't clear
22220         LOOP_VINFO_MAY_ALIAS_DDRS.
22222 2017-08-04  Richard Biener  <rguenther@suse.de>
22224         PR middle-end/81705
22225         * fold-const.c (fold_binary_loc): Properly restrict
22226         minus_var0 && minus_var1 case when associating undefined overflow
22227         entities.
22229 2017-08-04  Tom de Vries  <tom@codesourcery.com>
22231         * omp-simd-clone.c (simd_clone_adjust): Add missing edge probability.
22233 2017-08-03  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
22235         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
22236         Don't start diagnostic messages with a capital letter.
22237         * config/rs6000/rs6000.c (rs6000_option_override_internal):
22238         Likewise.
22239         (rs6000_invalid_builtin): Likewise.
22240         (rs6000_trampoline_init): Likewise.
22242 2017-08-03  Jakub Jelinek  <jakub@redhat.com>
22244         PR target/81621
22245         * bb-reorder.c (pass_partition_blocks::execute): Return TODO_df_finish
22246         after setting changeable df flags.
22248 2017-08-03  Richard Biener  <rguenther@suse.de>
22250         * tree-ssa-reassoc.c (should_break_up_subtract): Also break
22251         up if the use is in USE - X.
22253 2017-08-03  Alexander Monakov  <amonakov@ispras.ru>
22255         * toplev.c (dumpfile.h): New include.
22256         (internal_error_reentered): New static function.  Use it...
22257         (internal_error_function): ...here to handle reentered internal_error.
22259 2017-08-03  Richard Biener  <rguenther@suse.de>
22261         PR middle-end/81148
22262         * fold-const.c (split_tree): Add minus_var and minus_con
22263         arguments, remove unused loc arg.  Never generate NEGATE_EXPRs
22264         here but always use minus_*.
22265         (associate_trees): Assert we never associate with MINUS_EXPR
22266         and NULL first operand.  Do not recurse for PLUS_EXPR operands
22267         when associating as MINUS_EXPR either.
22268         (fold_binary_loc): Track minus_var and minus_con.
22270 2017-08-03  Tom de Vries  <tom@codesourcery.com>
22272         PR lto/81430
22273         * tree-streamer-in.c (lto_input_ts_function_decl_tree_pointers): If
22274         ACCEL_COMPILER, apply finish_options on
22275         DECL_FUNCTION_SPECIFIC_OPTIMIZATION.
22277 2017-08-03  Tom de Vries  <tom@codesourcery.com>
22279         PR target/81662
22280         * config/nvptx/nvptx.c (nvptx_option_override): Emit sorry if
22281         function_entry_patch_area_size > 0.
22283 2017-08-03  Jakub Jelinek  <jakub@redhat.com>
22285         PR driver/81650
22286         * calls.c (alloc_max_size): Use HOST_WIDE_INT_UC (10??)
22287         instead of 10??LU, perform unit multiplication in wide_int,
22288         don't change alloc_object_size_limit if the limit is larger
22289         than SSIZE_MAX.
22291         PR tree-optimization/81655
22292         PR tree-optimization/81588
22293         * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Handle also
22294         the case when ranges[i].low and high are 1 for unsigned type with
22295         precision 1.
22297         PR middle-end/81052
22298         * omp-low.c (diagnose_sb_0): Handle flag_openmp_simd like flag_openmp.
22299         (pass_diagnose_omp_blocks::gate): Enable also for flag_openmp_simd.
22301 2017-08-03  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
22303         * tree-vrp.h: Add include guard.
22305 2017-08-02  Uros Bizjak  <ubizjak@gmail.com>
22307         PR target/81644
22308         * config/i386/i386.md (unspecv): Add UNSPECV_UD2.
22309         (ud2): New insn pattern.
22310         * config/i386/i386.c (ix86_expand_epilogue):
22311         For naked functions, generate ud2 instead of trap insn.
22313 2017-08-02  Marek Polacek  <polacek@redhat.com>
22315         PR other/81667
22316         * alloc-pool.h (base_pool_allocator): Initialize m_elt_size.
22318 2017-08-02  Tom de Vries  <tom@codesourcery.com>
22319             Cesar Philippidis  <cesar@codesourcery.com>
22321         * config/nvptx/nvptx.c (nvptx_lockless_update, nvptx_lockfull_update):
22322         Add missing edge probabilities.
22324 2017-08-02  Tamar Christina  <tamar.christina@arm.com>
22326         * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
22327         Correct endianness.
22329 2017-08-02  Jakub Jelinek  <jakub@redhat.com>
22331         PR middle-end/79499
22332         * function.c (thread_prologue_and_epilogue_insns): Determine blocks
22333         for find_many_sub_basic_blocks bitmap by looking up BLOCK_FOR_INSN
22334         of first NONDEBUG_INSN_P in each of the split_prologue_seq and
22335         prologue_seq sequences - if any.
22337 2017-08-02  Richard Biener  <rguenther@suse.de>
22339         * tree-vect-stmts.c (vectorizable_store): Perform vector extracts
22340         via vectors if supported, integer extracts via punning if supported
22341         or otherwise vector extracts.
22343 2017-08-02  Richard Biener  <rguenther@suse.de>
22345         * tree-ssa-pre.c (bitmap_insert_into_set_1): Remove and inline
22346         into ...
22347         (bitmap_insert_into_set): ... this.
22349 2017-08-02  Richard Biener  <rguenther@suse.de>
22351         PR tree-optimization/81633
22352         Revert
22353         2015-08-17  Alan Hayward  <alan.hayward@arm.com>
22355         PR tree-optimization/71752
22356         * tree-vect-slp.c (vect_get_slp_defs): Handle null operands.
22358 2017-08-01  Daniel Santos  <daniel.santos@pobox.com>
22360         * config/i386/i386.h (ix86_frame::outlined_save_offset): Remove field.
22361         (machine_function::call_ms2sysv_pad_out): Remove field.
22362         * config/i386/i386.c (xlogue_layout::get_stack_space_used): Modify.
22363         (ix86_compute_frame_layout): Likewise.
22365 2017-08-01  H.J. Lu  <hongjiu.lu@intel.com>
22367         PR target/81654
22368         * config/i386/i386.c (ix86_set_func_type): Disallow naked
22369         attribute with interrupt attribute.
22371 2017-08-01  Andrew Pinski  <apinski@cavium.com>
22373         * tree-ssa-scopedtables.c (hashable_expr_equal_p): Check
22374         BIT_INSERT_EXPR's operand 1
22375         to see if the types precision matches.
22377 2017-08-01  Martin Liska  <mliska@suse.cz>
22379         PR middle-end/70140
22380         * builtins.c (expand_builtin_memcpy_args): Remove.
22381         (expand_builtin_memcpy): Call newly added function
22382         expand_builtin_memory_copy_args.
22383         (expand_builtin_memcpy_with_bounds): Likewise.
22384         (expand_builtin_mempcpy): Remove last argument.
22385         (expand_builtin_mempcpy_with_bounds): Likewise.
22386         (expand_builtin_memory_copy_args): New function created from
22387         expand_builtin_mempcpy_args with small modifications.
22388         (expand_builtin_mempcpy_args): Remove.
22389         (expand_builtin_stpcpy): Remove unused argument.
22390         (expand_builtin): Likewise.
22391         (expand_builtin_with_bounds): Likewise.
22393 2017-08-01  Martin Liska  <mliska@suse.cz>
22395         Revert r250771
22396         Make mempcpy more optimal (PR middle-end/70140).
22398 2017-08-01  Jakub Jelinek  <jakub@redhat.com>
22400         PR target/81622
22401         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): For
22402         __builtin_vec_cmpne verify both arguments are compatible vectors
22403         before looking at TYPE_MODE on the element type.  For __builtin_vec_ld
22404         verify arg1_type is a pointer or array type.  For __builtin_vec_st,
22405         move computation of aligned to after checking the argument types.
22406         Formatting fixes.
22408         PR target/80846
22409         * config/rs6000/vsx.md (vextract_fp_from_shorth,
22410         vextract_fp_from_shortl): Add element mode after mode in gen_vec_init*
22411         calls.
22413 2017-08-01  Jerome Lambourg  <lambourg@adacore.com>
22414             Doug Rupp  <rupp@adacore.com>
22415             Olivier Hainque  <hainque@adacore.com>
22417         * config.gcc (arm-wrs-vxworks*): Rework to handle arm-wrs-vxworks7 as
22418         well as arm-wrs-vxworks. Update target_cpu_name from arm6 (arch v3) to
22419         arm8 (arch v4).
22420         * config/arm/vxworks.h (MAYBE_TARGET_BPABI_CPP_BUILTINS): New, helper
22421         for TARGET_OS_CPP_BUILTIN.
22422         (TARGET_OS_CPP_BUILTIN): Invoke MAYBE_TARGET_BPABI_CPP_BUILTINS(),
22423         refine CPU definitions for arm_arch5 and add those for arm_arch6 and
22424         arm_arch7.
22425         (MAYBE_ASM_ABI_SPEC): New, helper for SUBTARGET_EXTRA_ASM_SPEC,
22426         passing required abi options to the assembler for EABI configurations.
22427         (EXTRA_CC1_SPEC): New macro, to help prevent the implicit production
22428         of .text.hot and .text.unlikely sections for kernel modules when
22429         using ARM style exceptions.
22430         (CC1_SPEC): Remove obsolete attempt at mimicking Diab toolchain
22431         options. Add EXTRA_CC1_SPEC.
22432         (VXWORKS_ENDIAN_SPEC): Adjust comment and remove handling of Diab
22433         toolchain options.
22434         (DWARF2_UNWIND_INFO): Redefine to handle the pre/post VxWorks 7
22435         transition.
22436         (ARM_TARGET2_DWARF_FORMAT): Define.
22437         * config/arm/t-vxworks: Adjust multilib control to removal of the
22438         Diab command line options.
22440 2017-08-01  Martin Liska  <mliska@suse.cz>
22442         PR gcov-profile/81561
22443         * gcov.c (unblock): Make unblocking safe as we need to preserve
22444         index correspondence of blocks and block_lists.
22446 2017-08-01  Richard Biener  <rguenther@suse.de>
22448         PR tree-optimization/81181
22449         * tree-ssa-pre.c (compute_antic_aux): Defer clean() to ...
22450         (compute_antic): ... end of iteration here.
22452 2017-08-01  James Greenhalgh  <james.greenhalgh@arm.com>
22454         * common.opt (ftree-vectorize): No longer set flag_tree_vectorize.
22455         (ftree-loop-vectorize): Set as EnabledBy ftree-vectorize.
22456         (ftree-slp-vectorize): Likewise.
22457         * omp-expand (expand_omp_simd): Remove flag_tree_vectorize, as it
22458         can no longer be set independent of flag_tree_loop_vectorize.
22459         * omp-general.c (emp_max_vf): Likewise.
22460         * opts.c (enable_fdo_optimizations): Remove references to
22461         flag_tree_vectorize, these are now implicit.
22462         (common_handle_option): Remove handling for OPT_ftree_vectorize,
22463         and leave it for the options machinery.
22465 2017-08-01  Martin Liska  <mliska@suse.cz>
22467         PR middle-end/70140
22468         * builtins.c (expand_builtin_memcpy_args): Remove.
22469         (expand_builtin_memcpy): Call newly added function
22470         expand_builtin_memory_copy_args.
22471         (expand_builtin_memcpy_with_bounds): Likewise.
22472         (expand_builtin_mempcpy): Remove last argument.
22473         (expand_builtin_mempcpy_with_bounds): Likewise.
22474         (expand_builtin_memory_copy_args): New function created from
22475         expand_builtin_mempcpy_args with small modifications.
22476         (expand_builtin_mempcpy_args): Remove.
22477         (expand_builtin_stpcpy): Remove unused argument.
22478         (expand_builtin): Likewise.
22479         (expand_builtin_with_bounds): Likewise.
22481 2017-08-01  Uros Bizjak  <ubizjak@gmail.com>
22483         PR target/81641
22484         * config/i386/i386.c (ix86_print_operand_address_as): For -masm=intel
22485         print "ds:" only for immediates in generic address space.
22487 2017-08-01  Uros Bizjak  <ubizjak@gmail.com>
22489         PR target/81639
22490         * config/i386/i386.c (ix86_funciton_naked): New prototype.
22491         (ix86_function_ok_for_sibcall): Return false for naked functions.
22493 2017-08-01  Richard Biener  <rguenther@suse.de>
22495         * tree-ssa-pre.c (print_pre_expr): Handle NULL expr.
22496         (compute_antic): Seed worklist with exit block predecessors.
22497         * cfganal.c (dfs_find_deadend): For a cycle return the source
22498         of the edge closing it.
22500 2017-08-01  Tamar Christina  <tamar.christina@arm.com>
22502         * config/aarch64/aarch64.c
22503         (aarch64_can_const_movi_rtx_p): Move 0 check.
22505 2017-08-01  Bin Cheng  <bin.cheng@arm.com>
22507         * tree.h (POINTER_TYPE_OVERFLOW_UNDEFINED): Delete.
22508         * fold-const.c (fold_comparison, fold_binary_loc): Delete use of
22509         above macro.
22510         * match.pd: Ditto in address comparison pattern.
22512 2017-08-01  Bin Cheng  <bin.cheng@arm.com>
22514         PR tree-optimization/81627
22515         * tree-predcom.c (prepare_finalizers): Always rewrite into loop
22516         closed ssa form for store-store chain.
22518 2017-08-01  Bin Cheng  <bin.cheng@arm.com>
22520         PR tree-optimization/81620
22521         * tree-predcom.c (add_ref_to_chain): Don't set has_max_use_after
22522         for store-store chain.
22524 2017-08-01  Jakub Jelinek  <jakub@redhat.com>
22526         PR tree-optimization/81588
22527         * tree-ssa-reassoc.c (optimize_range_tests_var_bound): If
22528         ranges[i].in_p, invert comparison code ccode.  For >/>=,
22529         swap rhs1 and rhs2 and comparison code unconditionally,
22530         for </<= don't do that.  Don't swap rhs1/rhs2 again if
22531         ranges[i].in_p, instead invert comparison code ccode if
22532         opcode or oe->rank is BIT_IOR_EXPR.
22534         PR target/80846
22535         * optabs.def (vec_extract_optab, vec_init_optab): Change from
22536         a direct optab to conversion optab.
22537         * optabs.c (expand_vector_broadcast): Use convert_optab_handler
22538         with GET_MODE_INNER as last argument instead of optab_handler.
22539         * expmed.c (extract_bit_field_1): Likewise.  Use vector from
22540         vector extraction if possible and optab is available.
22541         * expr.c (store_constructor): Use convert_optab_handler instead
22542         of optab_handler.  Use vector initialization from smaller
22543         vectors if possible and optab is available.
22544         * tree-vect-stmts.c (vectorizable_load): Likewise.
22545         * doc/md.texi (vec_extract, vec_init): Document that the optabs
22546         now have two modes.
22547         * config/i386/i386.c (ix86_expand_vector_init): Handle expansion
22548         of vec_init from half-sized vectors with the same element mode.
22549         * config/i386/sse.md (ssehalfvecmode): Add V4TI case.
22550         (ssehalfvecmodelower, ssescalarmodelower): New mode attributes.
22551         (reduc_plus_scal_v8df, reduc_plus_scal_v4df, reduc_plus_scal_v2df,
22552         reduc_plus_scal_v16sf, reduc_plus_scal_v8sf, reduc_plus_scal_v4sf,
22553         reduc_<code>_scal_<mode>, reduc_umin_scal_v8hi): Add element mode
22554         after mode in gen_vec_extract* calls.
22555         (vec_extract<mode>): Renamed to ...
22556         (vec_extract<mode><ssescalarmodelower>): ... this.
22557         (vec_extract<mode><ssehalfvecmodelower>): New expander.
22558         (rotl<mode>3, rotr<mode>3, <shift_insn><mode>3, ashrv2di3): Add
22559         element mode after mode in gen_vec_init* calls.
22560         (VEC_INIT_HALF_MODE): New mode iterator.
22561         (vec_init<mode>): Renamed to ...
22562         (vec_init<mode><ssescalarmodelower>): ... this.
22563         (vec_init<mode><ssehalfvecmodelower>): New expander.
22564         * config/i386/mmx.md (vec_extractv2sf): Renamed to ...
22565         (vec_extractv2sfsf): ... this.
22566         (vec_initv2sf): Renamed to ...
22567         (vec_initv2sfsf): ... this.
22568         (vec_extractv2si): Renamed to ...
22569         (vec_extractv2sisi): ... this.
22570         (vec_initv2si): Renamed to ...
22571         (vec_initv2sisi): ... this.
22572         (vec_extractv4hi): Renamed to ...
22573         (vec_extractv4hihi): ... this.
22574         (vec_initv4hi): Renamed to ...
22575         (vec_initv4hihi): ... this.
22576         (vec_extractv8qi): Renamed to ...
22577         (vec_extractv8qiqi): ... this.
22578         (vec_initv8qi): Renamed to ...
22579         (vec_initv8qiqi): ... this.
22580         * config/rs6000/vector.md (VEC_base_l): New mode attribute.
22581         (vec_init<mode>): Renamed to ...
22582         (vec_init<mode><VEC_base_l>): ... this.
22583         (vec_extract<mode>): Renamed to ...
22584         (vec_extract<mode><VEC_base_l>): ... this.
22585         * config/rs6000/paired.md (vec_initv2sf): Renamed to ...
22586         (vec_initv2sfsf): ... this.
22587         * config/rs6000/altivec.md (splitter, altivec_copysign_v4sf3,
22588         vec_unpacku_hi_v16qi, vec_unpacku_hi_v8hi, vec_unpacku_lo_v16qi,
22589         vec_unpacku_lo_v8hi, mulv16qi3, altivec_vreve<mode>2): Add
22590         element mode after mode in gen_vec_init* calls.
22591         * config/aarch64/aarch64-simd.md (vec_init<mode>): Renamed to ...
22592         (vec_init<mode><Vel>): ... this.
22593         (vec_extract<mode>): Renamed to ...
22594         (vec_extract<mode><Vel>): ... this.
22595         * config/aarch64/iterators.md (Vel): New mode attribute.
22596         * config/s390/s390.c (s390_expand_vec_strlen, s390_expand_vec_movstr):
22597         Add element mode after mode in gen_vec_extract* calls.
22598         * config/s390/vector.md (non_vec_l): New mode attribute.
22599         (vec_extract<mode>): Renamed to ...
22600         (vec_extract<mode><non_vec_l>): ... this.
22601         (vec_init<mode>): Renamed to ...
22602         (vec_init<mode><non_vec_l>): ... this.
22603         * config/s390/s390-builtins.def (s390_vlgvb, s390_vlgvh, s390_vlgvf,
22604         s390_vlgvf_flt, s390_vlgvg, s390_vlgvg_dbl): Add element mode after
22605         vec_extract mode.
22606         * config/arm/iterators.md (V_elem_l): New mode attribute.
22607         * config/arm/neon.md (vec_extract<mode>): Renamed to ...
22608         (vec_extract<mode><V_elem_l>): ... this.
22609         (vec_extractv2di): Renamed to ...
22610         (vec_extractv2didi): ... this.
22611         (vec_init<mode>): Renamed to ...
22612         (vec_init<mode><V_elem_l>): ... this.
22613         (reduc_plus_scal_<mode>, reduc_plus_scal_v2di, reduc_smin_scal_<mode>,
22614         reduc_smax_scal_<mode>, reduc_umin_scal_<mode>,
22615         reduc_umax_scal_<mode>, neon_vget_lane<mode>, neon_vget_laneu<mode>):
22616         Add element mode after gen_vec_extract* calls.
22617         * config/mips/mips-msa.md (vec_init<mode>): Renamed to ...
22618         (vec_init<mode><unitmode>): ... this.
22619         (vec_extract<mode>): Renamed to ...
22620         (vec_extract<mode><unitmode>): ... this.
22621         * config/mips/loongson.md (vec_init<mode>): Renamed to ...
22622         (vec_init<mode><unitmode>): ... this.
22623         * config/mips/mips-ps-3d.md (vec_initv2sf): Renamed to ...
22624         (vec_initv2sfsf): ... this.
22625         (vec_extractv2sf): Renamed to ...
22626         (vec_extractv2sfsf): ... this.
22627         (reduc_plus_scal_v2sf, reduc_smin_scal_v2sf, reduc_smax_scal_v2sf):
22628         Add element mode after gen_vec_extract* calls.
22629         * config/mips/mips.md (unitmode): New mode iterator.
22630         * config/spu/spu.c (spu_expand_prologue, spu_allocate_stack,
22631         spu_builtin_extract): Add element mode after gen_vec_extract* calls.
22632         * config/spu/spu.md (inner_l): New mode attribute.
22633         (vec_init<mode>): Renamed to ...
22634         (vec_init<mode><inner_l>): ... this.
22635         (vec_extract<mode>): Renamed to ...
22636         (vec_extract<mode><inner_l>): ... this.
22637         * config/sparc/sparc.md (veltmode): New mode iterator.
22638         (vec_init<VMALL:mode>): Renamed to ...
22639         (vec_init<VMALL:mode><VMALL:veltmode>): ... this.
22640         * config/ia64/vect.md (vec_initv2si): Renamed to ...
22641         (vec_initv2sisi): ... this.
22642         (vec_initv2sf): Renamed to ...
22643         (vec_initv2sfsf): ... this.
22644         (vec_extractv2sf): Renamed to ...
22645         (vec_extractv2sfsf): ... this.
22646         * config/powerpcspe/vector.md (VEC_base_l): New mode attribute.
22647         (vec_init<mode>): Renamed to ...
22648         (vec_init<mode><VEC_base_l>): ... this.
22649         (vec_extract<mode>): Renamed to ...
22650         (vec_extract<mode><VEC_base_l>): ... this.
22651         * config/powerpcspe/paired.md (vec_initv2sf): Renamed to ...
22652         (vec_initv2sfsf): ... this.
22653         * config/powerpcspe/altivec.md (splitter, altivec_copysign_v4sf3,
22654         vec_unpacku_hi_v16qi, vec_unpacku_hi_v8hi, vec_unpacku_lo_v16qi,
22655         vec_unpacku_lo_v8hi, mulv16qi3): Add element mode after mode in
22656         gen_vec_init* calls.
22658 2017-08-01  Richard Biener  <rguenther@suse.de>
22660         PR tree-optimization/81297
22661         * tree-vrp.c (get_single_symbol): Remove assert, instead drop
22662         TREE_OVERFLOW from INTEGER_CSTs.
22664 2017-07-31  Segher Boessenkool  <segher@kernel.crashing.org>
22666         * config/rs6000/rs6000.c (enum rs6000_reg_type): Delete trailing comma.
22668 2017-07-31  Carl Love  <cel@us.ibm.com>
22670         * config/rs6000/rs6000-c: Add support for built-in functions
22671         vector signed char vec_xl_be (signed long long, signed char *);
22672         vector unsigned char vec_xl_be (signed long long, unsigned char *);
22673         vector signed int vec_xl_be (signed long long, signed int *);
22674         vector unsigned int vec_xl_be (signed long long, unsigned int *);
22675         vector signed long long vec_xl_be (signed long long, signed long long *);
22676         vector unsigned long long vec_xl_be (signed long long, unsigned long long *);
22677         vector signed short vec_xl_be (signed long long, signed short *);
22678         vector unsigned short vec_xl_be (signed long long, unsigned short *);
22679         vector double vec_xl_be (signed long long, double *);
22680         vector float vec_xl_be (signed long long, float *);
22681         * config/rs6000/altivec.h (vec_xl_be): Add #define.
22682         * config/rs6000/rs6000-builtin.def (XL_BE_V16QI, XL_BE_V8HI, XL_BE_V4SI)
22683         XL_BE_V2DI, XL_BE_V4SF, XL_BE_V2DF, XL_BE): Add definitions
22684         for the builtins.
22685         * config/rs6000/rs6000.c (altivec_expand_xl_be_builtin): Add function.
22686         (altivec_expand_builtin): Add switch statement to call
22687         altivec_expand_xl_be for each builtin.
22688         (altivec_init_builtins): Add def_builtin for _builtin_vsx_le_be_v8hi,
22689         __builtin_vsx_le_be_v4si, __builtin_vsx_le_be_v2di,
22690         __builtin_vsx_le_be_v4sf, __builtin_vsx_le_be_v2df,
22691         __builtin_vsx_le_be_v16qi.
22692         * doc/extend.texi: Update the built-in documentation file for the
22693         new built-in functions.
22695 2017-07-31  Uros Bizjak  <ubizjak@gmail.com>
22697         PR target/25967
22698         * config/i386/i386.c (ix86_allocate_stack_slots_for_args):
22699         New function.
22700         (TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS): Define.
22702 2017-07-31  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
22704         * config.gcc: Add z14.
22705         * config/s390/driver-native.c (s390_host_detect_local_cpu): Add
22706         CPU model numbers for z13s and z14.
22707         * config/s390/s390-c.c (s390_resolve_overloaded_builtin): Replace
22708         arch12 with z14.
22709         * config/s390/s390-opts.h (enum processor_type): Rename
22710         PROCESSOR_ARCH12 to PROCESSOR_3906_Z14.
22711         * config/s390/s390.c (processor_table): Add field for CPU name to
22712         be passed to Binutils.
22713         (s390_asm_output_machine_for_arch): Use the new field in
22714         processor_table for Binutils.
22715         (s390_expand_builtin): Replace arch12 with z14.
22716         (s390_issue_rate): Rename PROCESSOR_ARCH12 to PROCESSOR_3906_Z14.
22717         (s390_get_sched_attrmask): Likewise.
22718         (s390_get_unit_mask): Likewise.
22719         * config/s390/s390.opt: Add z14 to processor_type enum.
22721 2017-07-31  Martin Jambor  <mjambor@suse.cz>
22723         PR hsa/81477
22724         * ipa-fnsummary.c (ipa_fn_summary_generate): Set versionable
22725         regardless of optimization level.
22727 2017-07-31  Jan Hubicka <hubicka@ucw.cz>
22728             Martin Liska  <mliska@suse.cz>
22730         * predict.def: Remove old comment and adjust probability.
22731         * gimplify.c (should_warn_for_implicit_fallthrough): Ignore
22732         PREDICT statements.
22734 2017-07-31  Uros Bizjak  <ubizjak@gmail.com>
22736         PR target/25967
22737         * config/i386/i386.c (ix86_function_naked): New function.
22738         (ix86_can_use_return_insn_p): Return false for naked functions.
22739         (ix86_expand_prologue): Skip prologue for naked functions.
22740         (ix86_expand_epilogue): Skip epilogue for naked functions
22741         and emit trap instruction.
22742         (ix86_warn_func_return): New function.
22743         (ix86_attribute_table): Add "naked" attribute specification.
22744         (TARGET_WARN_FUNC_RETURN): Define.
22745         * doc/extend.texi (x86 Function Attributes) <naked>: Document it.
22747 2017-07-31  Martin Liska  <mliska@suse.cz>
22749         * gimple-pretty-print.c (dump_gimple_label): Never dump BB info.
22750         (dump_gimple_bb_header): Always dump BB info.
22751         (pp_cfg_jump): Do not append info about BB when dumping a jump.
22753 2017-07-31  Martin Liska  <mliska@suse.cz>
22755         PR sanitize/81530
22756         * convert.c (convert_to_integer_1): Guard condition with flag_sanitize_p
22757         also with current_function_decl non-null equality.
22759 2017-07-31  Jakub Jelinek  <jakub@redhat.com>
22761         PR sanitizer/81604
22762         * ubsan.c (ubsan_type_descriptor): For UBSAN_PRINT_ARRAY don't
22763         change type to the element type, instead add eltype variable and
22764         use it where we are interested in the element type.
22766         PR tree-optimization/81603
22767         * ipa-polymorphic-call.c
22768         (ipa_polymorphic_call_context::ipa_polymorphic_call_context): Perform
22769         offset arithmetic in offset_int, bail out if the resulting bit offset
22770         doesn't fit into shwi.
22772 2017-07-31  Martin Liska  <mliska@suse.cz>
22774         * gimplify.c (mostly_copy_tree_r): Remove Java specific hunk.
22775         (gimplify_save_expr): Fix comment.
22777 2017-07-30  H.J. Lu  <hongjiu.lu@intel.com>
22779         PR target/79793
22780         * config/i386/i386.c (ix86_function_arg): Update arguments for
22781         exception handler.
22782         (ix86_compute_frame_layout): Set the initial stack offset to
22783         INCOMING_FRAME_SP_OFFSET.  Update red-zone offset with
22784         INCOMING_FRAME_SP_OFFSET.
22785         (ix86_expand_epilogue): Don't pop the 'ERROR_CODE' off the
22786         stack before exception handler returns.
22787         * config/i386/i386.h (INCOMING_FRAME_SP_OFFSET): Add the
22788         the 'ERROR_CODE' for exception handler.
22790 2017-07-30  Uros Bizjak  <ubizjak@gmail.com>
22792         * config/i386/i386.h (ASM_PRINTF_EXTENSIONS): New macro.
22793         (ASM_OUTPUT_REG_PUSH): Rewrite with new operand modifiers.
22794         (ASM_OUTPUT_REG_POP): Ditto.
22795         * config/i386/i386.c (ix86_asm_output_function_label): Use fputs
22796         instead of asm_fprintf to output pure string.
22798 2017-07-29  Jakub Jelinek  <jakub@redhat.com>
22800         * debug.h (struct gcc_debug_hooks): Add IMPLICIT argument
22801         to imported_module_or_decl hook.
22802         (debug_nothing_tree_tree_tree_bool): Remove.
22803         (debug_nothing_tree_tree_tree_bool_bool): New declaration.
22804         * debug.c (do_nothing_debug_hooks): Use
22805         debug_nothing_tree_tree_tree_bool_bool instead of
22806         debug_nothing_tree_tree_tree_bool.
22807         * vmsdbgout.c (vmsdbg_debug_hooks): Likewise.
22808         * dbxout.c (dbx_debug_hooks, xcoff_debug_hooks): Likewise.
22809         * sdbout.c (sdb_debug_hooks): Likewise.
22810         * dwarf2out.c (dwarf2_lineno_debug_hooks): Likewise.
22811         (gen_namespace_die): Add DW_AT_export_symbols attribute if
22812         langhook wants it.
22813         (dwarf2out_imported_module_or_decl): Add IMPLICIT argument,
22814         if true, -gdwarf-5 and decl will have DW_AT_export_symbols
22815         attribute, don't add anything.
22817 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22819         * fold-const.c (fold_build1_stat_loc): Remove _stat from name.
22820         (fold_build2_stat_loc): Likewise.
22821         (fold_build3_stat_loc): Likewise.
22822         * fold-const.h (fold_build1, fold_build2, fold_build3): Adjust.
22823         (fold_build1_loc): Remove macro.
22824         (fold_build2_loc): Likewise.
22825         (fold_build3_loc): Likewise.
22827 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22829         * gimple.c (gimple_build_debug_bind_stat): Remove _stat from name.
22830         (gimple_build_debug_bind_source_stat): Likewise.
22831         * gimple.h (gimple_build_debug_bind): Remove macro.
22832         (gimple_build_debug_bind_source): Likewise.
22834 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22836         * bitmap.c (bitmap_alloc): Adjust.
22837         (bitmap_gc_alloc): Likewise.
22838         * bitmap.h (bitmap_initialize_stat): Remove _stat from name.
22840 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22842         * bitmap.c (bitmap_obstack_alloc_stat): Rename to bitmap_alloc.
22843         (bitmap_gc_alloc_stat): Rename to bitmap_gc_alloc.
22844         * bitmap.h (bitmap_obstack_alloc_stat): Adjust prototype.
22845         (bitmap_gc_alloc_stat): Likewise.
22846         (BITMAP_ALLOC, BITMAP_GGC_ALLOC): Adjust.
22848 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22850         * rtl.c (shallow_copy_rtx_stat): Remove _stat from name.
22851         * rtl.h (shallow_copy_rtx): Remove macro.
22853 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22855         * emit-rtl.c (gen_raw_REG): Adjust.
22856         * gengenrtl.c (gendef): Likewise.
22857         * rtl.c (rtx_alloc_stat): Remove _stat from name.
22858         * rtl.h (rtx_alloc): Remove macro.
22860 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22862         * tree.c (build_tree_list_vec_stat): Remove _stat from name.
22863         (build_tree_list_stat): Likewise.
22864         * tree.h (build_tree_list): Remove macro.
22865         (build_tree_list_vec): Likewise.
22867 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22869         * tree.c (make_vector_stat): Remove _stat from name.
22870         (build_vector_stat): Likewise.
22871         * tree.h (make_vector_stat): Remove macro.
22872         (build_vector_stat): Likewise.
22874 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22876         * tree.h (build_var_debug_value): Remove prototype.
22878 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22880         * tree.c (tree_cons_stat): Remove _stat from name.
22881         * tree.h (tree_cons): Remove macro.
22883 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22885         * tree.c (build_vl_exp_stat): Remove _stat from name.
22886         * tree.h (build_vl_exp): Remove macro.
22888 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22890         * tree.c (build_decl_stat): Remove _stat from name.
22891         * tree.h (build_decl): Remove macro.
22893 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22895         * gimple.c (gimple_build_with_ops_stat): Adjust.
22896         (gimple_alloc_stat): Remove _stat from name.
22897         * gimple.h (gimple_alloc): Remove macro.
22899 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22901         * tree.c (make_tree_vec_stat): Remove _stat from name.
22902         (grow_tree_vec_stat): Likewise.
22903         * tree.h (make_tree_vec_stat): Adjust prototype.
22904         (grow_tree_vec_stat): Likewise.
22905         (make_tree_vec): Remove macro.
22906         (grow_tree_vec): Likewise.
22908 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22910         * fold-const.c (fold_build1_stat_loc): Adjust.
22911         (fold_build2_stat_loc): Likewise.
22912         (fold_build3_stat_loc): Likewise.
22913         * tree.c (build0_stat): Remove _stat from name.
22914         (build1_stat): Likewise.
22915         (build2_stat): Likewise.
22916         (build3_stat): Likewise.
22917         (build4_stat): Likewise.
22918         (build5_stat): Likewise.
22919         * tree.h (build1_loc): Remove macro, and rename _stat function
22920         to this.
22921         (build2_loc): Likewise.
22922         (build3_loc): Likewise.
22923         (build4_loc): Likewise.
22924         (build5_loc): Likewise.
22926 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22928         * tree.c (make_int_cst_stat): Remove _stat from name.
22929         * tree.h (make_int_cst_stat): Adjust prototype.
22930         (make_int_cst): Remove macro.
22932 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22934         * tree.c (make_tre_binfo_stat): Remove _stat from name.
22935         * tree.h (make_tree_binfo_stat): Adjust prototype.
22936         (make_tree_binfo): Remove.
22938 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22940         * tree.c (copy_node_stat): Rename to copy_node.
22941         (build_distinct_type_copy): Adjust.
22942         * tree.h (copy_node_stat): Adjust prototype.
22943         (copy_node): Remove macro.
22945 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22947         * tree.c (make_node_stat): rename to make_node.
22948         (build_tree_list_stat): Adjust.
22949         (build0_stat): Likewise.
22950         (build2_stat): Likewise.
22951         (build3_stat): Likewise.
22952         (build4_stat): Likewise.
22953         (build5_stat): Likewise.
22954         (build_decl_stat): Likewise.
22955         * tree.h (make_node_stat): Adjust prototype.
22956         (make_node): remove macro.
22958 2017-07-28  Peter Bergner  <bergner@vnet.ibm.com>
22960         * config/rs6000/ppc-auxv.h (PPC_FEATURE2_DARN): New define.
22961         (PPC_FEATURE2_SCV): Likewise.
22962         * config/rs6000/rs6000.c (cpu_supports_info): Use them.
22964 2017-07-28  Tamar Christina  <tamar.christina@arm.com>
22966         * config/aarch64/aarch64.c
22967         (aarch64_internal_mov_immediate): Add new special pattern.
22968         * config/aarch64/aarch64.md (*movdi_aarch64):
22969         Add reg/32bit const mov case.
22971 2017-07-28  Tamar Christina  <tamar.christina@arm.com>
22972             Richard Sandiford <richard.sandiford@linaro.org>
22974         * config/aarch64/aarch64.md (mov<mode>): Generalize.
22975         (*movhf_aarch64, *movsf_aarch64, *movdf_aarch64):
22976         Add integer and movi cases.
22977         (movi-split-hf-df-sf split, fp16): New.
22978         (enabled): Added TARGET_FP_F16INST.
22979         * config/aarch64/iterators.md (GPF_HF): New.
22980         * config/aarch64/predicates.md (aarch64_reg_or_fp_float): New.
22982 2017-07-28  Tamar Christina  <tamar.christina@arm.com>
22984         * config/aarch64/aarch64.c
22985         (aarch64_simd_container_mode): Add prototype.
22986         (aarch64_expand_mov_immediate): Add HI support.
22987         (aarch64_reinterpret_float_as_int, aarch64_float_const_rtx_p: New.
22988         (aarch64_can_const_movi_rtx_p): New.
22989         (aarch64_preferred_reload_class):
22990         Remove restrictions of using FP registers for certain SIMD operations.
22991         (aarch64_rtx_costs): Added new cost for CONST_DOUBLE moves.
22992         (aarch64_valid_floating_const): Add integer move validation.
22993         (aarch64_simd_imm_scalar_p): Remove.
22994         (aarch64_output_scalar_simd_mov_immediate): Generalize function.
22995         (aarch64_legitimate_constant_p): Expand list of supported cases.
22996         * config/aarch64/aarch64-protos.h
22997         (aarch64_float_const_rtx_p, aarch64_can_const_movi_rtx_p): New.
22998         (aarch64_reinterpret_float_as_int): New.
22999         (aarch64_simd_imm_scalar_p): Remove.
23000         * config/aarch64/constraints.md (Uvi): New.
23001         (Dd): Split into Ds and new Dd.
23002         * config/aarch64/aarch64.md (*movsi_aarch64):
23003         Add SIMD mov case.
23004         (*movdi_aarch64): Add SIMD mov case.
23006 2017-07-28  Bin Cheng  <bin.cheng@arm.com>
23008         * tree-predcom.c: (struct chain): Handle store-store chain in which
23009         stores for elimination only store loop invariant values.
23010         (execute_pred_commoning_chain): Ditto.
23011         (prepare_initializers_chain_store_elim): Ditto.
23012         (prepare_finalizers): Ditto.
23013         (is_inv_store_elimination_chain): New function.
23014         (initialize_root_vars_store_elim_1): New function.
23016 2017-07-28  Bin Cheng  <bin.cheng@arm.com>
23018         * tree-predcom.c: Revise general description of the pass.
23019         (enum chain_type): New enum type for store elimination.
23020         (struct chain): New field supporting store elimination.
23021         (struct component): Ditto.
23022         (dump_chain): Dump store-stores chain.
23023         (release_chain): Release resources.
23024         (split_data_refs_to_components): Compute and create component
23025         contains only stores for elimination.
23026         (get_chain_last_ref_at): New function.
23027         (make_invariant_chain): Initialization.
23028         (make_rooted_chain): Specify chain type in parameter and record it.
23029         (add_looparound_copies): Skip for store-stores chain.
23030         (determine_roots_comp): Compute type of chain and pass it to
23031         make_rooted_chain.
23032         (initialize_root_vars_store_elim_2): New function.
23033         (finalize_eliminated_stores): New function.
23034         (remove_stmt): Handle store for elimination.
23035         (execute_pred_commoning_chain): Execute predictive commoning on
23036         store-store chains.
23037         (determine_unroll_factor): Skip unroll for store-stores chain.
23038         (prepare_initializers_chain_store_elim): New function.
23039         (prepare_initializers_chain): Hanlde store-store chain.
23040         (prepare_finalizers_chain, prepare_finalizers): New function.
23041         (tree_predictive_commoning_loop): Return integer value indicating
23042         if loop is unrolled or lcssa form is corrupted.
23043         (tree_predictive_commoning): Rewrite for lcssa form if necessary.
23045 2017-07-28  Bin Cheng  <bin.cheng@arm.com>
23047         * tree-predcom.c (initialize_root): Delete.
23048         (execute_pred_commoning_chain): Initialize root vars and replace
23049         reference of non-combined chain directly, rather than call above
23050         function.
23052 2017-07-28  Bin Cheng  <bin.cheng@arm.com>
23054         * tree-predcom.c (ref_at_iteration): Add parameter NITERS.  Compute
23055         memory reference to DR at (NITERS + ITERS)-th iteration of loop.
23057 2017-07-28  Bin Cheng  <bin.cheng@arm.com>
23059         * tree-predcom.c (struct chain): New field init_seq.
23060         (release_chain): Release init_seq.
23061         (prepare_initializers_chain): Record intialization stmts in above
23062         field.
23063         (insert_init_seqs): New function.
23064         (tree_predictive_commoning_loop): Call insert_init_seqs.
23066 2017-07-28  Bin Cheng  <bin.cheng@arm.com>
23068         * tree-predcom.c (determine_roots_comp): Skip trivial components.
23070 2017-07-28  Richard Biener  <rguenther@suse.de>
23072         * match.pd: Remove superfluous :c.
23073         * genmatch.c (simplify::id): Add member.
23074         (lower_commutative, lower_opt_convert, lower_cond, lower_for):
23075         Copy id.
23076         (current_id): New global.
23077         (dt_node::parent): Move from ...
23078         (dt_operand::parent): ... here.  Add for_id member.
23079         (is_a_helper <dt_operand *>::test): DT_TRUE is also a dt_operand.
23080         (decision_tree::find_node): Relax order requirement when
23081         merging DT_TRUE nodes to ones inbetween the current simplify
23082         and the one we try to merge with.  Add diagnostic whenever
23083         we need to enforce pattern order by not merging.
23084         (decision_tree::insert): Set current_id.
23085         (decision_tree::print_node): Dump parent node and for_id.
23086         (parser::last_id): Add member.
23087         (parser::push_simplify): Assign unique id.
23088         (parser::parser): Initialize last_id.
23090 2017-07-28  Martin Liska  <mliska@suse.cz>
23092         PR sanitizer/81340
23093         * sanopt.c (sanitize_rewrite_addressable_params): Set VALUE_EXPR after
23094         gimple_build_debug_bind.
23096 2017-07-28  Richard Biener  <rguenther@suse.de>
23098         PR tree-optimization/81502
23099         * match.pd: Add pattern combining BIT_INSERT_EXPR with
23100         BIT_FIELD_REF.
23101         * tree-cfg.c (verify_expr): Verify types of BIT_FIELD_REF
23102         size/pos operands.
23103         (verify_gimple_assign_ternary): Likewise for BIT_INSERT_EXPR pos.
23104         * gimple-fold.c (maybe_canonicalize_mem_ref_addr): Use bitsizetype
23105         for BIT_FIELD_REF args.
23106         * fold-const.c (make_bit_field_ref): Likewise.
23107         * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
23109 2017-07-28  Jakub Jelinek  <jakub@redhat.com>
23111         PR sanitizer/80998
23112         * sanopt.c (pass_sanopt::execute): Handle IFN_UBSAN_PTR.
23113         * tree-ssa-alias.c (call_may_clobber_ref_p_1): Likewise.
23114         * flag-types.h (enum sanitize_code): Add SANITIZER_POINTER_OVERFLOW.
23115         Or it into SANITIZER_UNDEFINED.
23116         * ubsan.c: Include gimple-fold.h and varasm.h.
23117         (ubsan_expand_ptr_ifn): New function.
23118         (instrument_pointer_overflow): New function.
23119         (maybe_instrument_pointer_overflow): New function.
23120         (instrument_object_size): Formatting fix.
23121         (pass_ubsan::execute): Call instrument_pointer_overflow
23122         and maybe_instrument_pointer_overflow.
23123         * internal-fn.c (expand_UBSAN_PTR): New function.
23124         * ubsan.h (ubsan_expand_ptr_ifn): Declare.
23125         * sanitizer.def (__ubsan_handle_pointer_overflow,
23126         __ubsan_handle_pointer_overflow_abort): New builtins.
23127         * tree-ssa-tail-merge.c (merge_stmts_p): Handle IFN_UBSAN_PTR.
23128         * internal-fn.def (UBSAN_PTR): New internal function.
23129         * opts.c (sanitizer_opts): Add pointer-overflow.
23130         * lto-streamer-in.c (input_function): Handle IFN_UBSAN_PTR.
23131         * fold-const.c (build_range_check): Compute pointer range check in
23132         integral type if pointer arithmetics would be needed.  Formatting
23133         fixes.
23135 2017-07-28  Martin Liska  <mliska@suse.cz>
23137         PR sanitizer/81460
23138         * sanopt.c (sanitize_rewrite_addressable_params): Do not rewrite
23139         parameters that are of a variable-length.
23141 2017-07-28  Sebastian Huber  <sebastian.huber@embedded-brains.de>
23143         * config.gcc (powerpc-*-rtems*): Remove rs6000/eabi.h.  Add
23144         rs6000/biarch64.h.
23145         * config/rs6000/rtems.h (ASM_DECLARE_FUNCTION_SIZE): New macro.
23146         (ASM_OUTPUT_SPECIAL_POOL_ENTRY_P): Likewise.
23147         (CRT_CALL_STATIC_FUNCTION): Likewise.
23148         (ASM_DEFAULT_SPEC): New define.
23149         (ASM_SPEC32): Likewise.
23150         (ASM_SPEC64): Likewise.
23151         (ASM_SPEC_COMMON): Likewise.
23152         (ASM_SPEC): Likewise.
23153         (INVALID_64BIT): Likewise.
23154         (LINK_OS_DEFAULT_SPEC): Likewise.
23155         (LINK_OS_SPEC32): Likewise.
23156         (LINK_OS_SPEC64): Likewise.
23157         (POWERPC_LINUX): Likewise.
23158         (PTRDIFF_TYPE): Likewise.
23159         (RESTORE_FP_PREFIX): Likewise.
23160         (RESTORE_FP_SUFFIX): Likewise.
23161         (SAVE_FP_PREFIX): Likewise.
23162         (SAVE_FP_SUFFIX): Likewise.
23163         (SIZE_TYPE): Likewise.
23164         (SUBSUBTARGET_OVERRIDE_OPTIONS): Likewise.
23165         (TARGET_64BIT): Likewise.
23166         (TARGET_64BIT): Likewise.
23167         (TARGET_AIX): Likewise.
23168         (WCHAR_TYPE_SIZE): Likewise.
23169         (WCHAR_TYPE): Undefine.
23170         (TARGET_OS_CPP_BUILTINS): Add 64-bit PowerPC defines.
23171         (CPP_OS_DEFAULT_SPEC): Use previous CPP_OS_RTEMS_SPEC.
23172         (CPP_OS_RTEMS_SPEC): Delete.
23173         (SUBSUBTARGET_EXTRA_SPECS): Remove cpp_os_rtems.  Add
23174         asm_spec_common, asm_spec32, asm_spec64, link_os_spec32, and
23175         link_os_spec64.
23176         * config/rs6000/t-rtems: Add mcpu=e6500/m64 multilibs.
23178 2017-07-28  Jakub Jelinek  <jakub@redhat.com>
23180         PR tree-optimization/81578
23181         * tree-parloops.c (build_new_reduction): Bail out if
23182         reduction_code isn't one of the standard OpenMP reductions.
23183         Move the details printing after that decision.
23185 2017-07-27  Peter Bergner  <bergner@vnet.ibm.com>
23187         * config/rs6000/predicates.md (volatile_mem_operand): Remove code
23188         related to reload_in_progress.
23189         (splat_input_operand): Likewise.
23190         * config/rs6000/rs6000-protos.h (rs6000_secondary_memory_needed_rtx):
23191         Delete prototype.
23192         * config/rs6000/rs6000.c (machine_function): Remove sdmode_stack_slot
23193         field.
23194         (TARGET_EXPAND_TO_RTL_HOOK): Delete.
23195         (TARGET_INSTANTIATE_DECLS): Likewise.
23196         (legitimate_indexed_address_p): Delete reload_in_progress code.
23197         (rs6000_debug_legitimate_address_p): Likewise.
23198         (rs6000_eliminate_indexed_memrefs): Likewise.
23199         (rs6000_emit_le_vsx_store): Likewise.
23200         (rs6000_emit_move_si_sf_subreg): Likewise.
23201         (rs6000_emit_move): Likewise.
23202         (register_to_reg_type): Likewise.
23203         (rs6000_pre_atomic_barrier): Likewise.
23204         (rs6000_machopic_legitimize_pic_address): Likewise.
23205         (rs6000_allocate_stack_temp): Likewise.
23206         (rs6000_address_for_fpconvert): Likewise.
23207         (rs6000_address_for_altivec): Likewise.
23208         (rs6000_secondary_memory_needed_rtx): Delete function.
23209         (rs6000_check_sdmode): Likewise.
23210         (rs6000_alloc_sdmode_stack_slot): Likewise.
23211         (rs6000_instantiate_decls): Likewise.
23212         * config/rs6000/rs6000.h (SECONDARY_MEMORY_NEEDED_RTX): Delete.
23213         * config/rs6000/rs6000.md (splitter for *movsi_got_internal):
23214         Delete reload_in_progress.
23215         (*vec_reload_and_plus_<mptrsize>): Likewise.
23216         * config/rs6000/vsx.md (vsx_mul_v2di): Likewise.
23217         (vsx_div_v2di): Likewise.
23218         (vsx_udiv_v2di): Likewise.
23220 2017-07-27  Peter Bergner  <bergner@vnet.ibm.com>
23222         * config/rs6000/rs6000.opt (mlra): Replace with stub.
23223         * config/rs6000/rs6000-cpus.def (POWERPC_MASKS): Delete OPTION_MASK_LRA.
23224         * config/rs6000/rs6000.c (TARGET_LRA_P): Delete.
23225         (rs6000_debug_reg_global): Delete print of LRA status.
23226         (rs6000_option_override_internal): Delete dead LRA related code.
23227         (rs6000_lra_p): Delete function.
23228         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mlra.
23230 2017-07-27  Sebastian Huber  <sebastian.huber@embedded-brains.de>
23232         * config.gcc (riscv*-*-elf*): Add (riscv*-*-rtems*).
23233         * config/riscv/rtems.h: New file.
23235 2017-07-27  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
23236             Sudakshina Das  <sudi.das@arm.com>
23238         * config/aarch64/aarch64.md
23239         (define_split for and<mode>3nr_compare): Move
23240         non aarch64_logical_operand to a register.
23241         (define_split for and_<SHIFT:optab><mode>3nr_compare0): Move non
23242         register immediate operand to a register.
23243         * config/aarch64/predicates.md (aarch64_mov_imm_operand): New.
23245 2017-07-27  Peter Bergner  <bergner@vnet.ibm.com>
23247         PR middle-end/81564
23248         * tree-cfg.c (group_case_labels_stmt): Handle already deleted blocks.
23250 2017-07-27  Richard Biener  <rguenther@suse.de>
23252         PR tree-optimization/81573
23253         PR tree-optimization/81494
23254         * tree-vect-loop.c (vect_create_epilog_for_reduction): Handle
23255         multi defuse cycle case.
23257 2017-07-27  Richard Biener  <rguenther@suse.de>
23259         PR tree-optimization/81571
23260         * tree-vect-slp.c (vect_build_slp_tree): Properly verify reduction
23261         PHIs.
23263 2017-07-27  Eric Botcazou  <ebotcazou@adacore.com>
23265         * config/sparc/sparc.c (sparc_option_override): Set MASK_FSMULD flag
23266         earlier and only if MASK_FPU is set.  Adjust formatting.
23268 2017-07-27  Martin Liska  <mliska@suse.cz>
23270         * opt-functions.awk: Add validation of value of Init.
23271         * optc-gen.awk: Pass new argument.
23273 2017-07-27  Martin Liska  <mliska@suse.cz>
23275         * auto-profile.c (autofdo_source_profile::update_inlined_ind_target):
23276         Fix wrong condition.
23278 2017-07-27  Martin Liska  <mliska@suse.cz>
23280         * auto-profile.c (afdo_annotate_cfg): Assign zero counts to
23281         BBs and edges seen by autoFDO.
23283 2017-07-27  Richard Biener  <rguenther@suse.de>
23285         PR tree-optimization/81502
23286         * tree-ssa.c (non_rewritable_lvalue_p): Handle BIT_INSERT_EXPR
23287         with incompatible but same sized type.
23288         (execute_update_addresses_taken): Likewise.
23290 2017-07-27  James Greenhalgh  <james.greenhalgh@arm.com>
23292         * tree-ssa-loop-ch.c (pass_ch::process_loop_p): Guard on
23293         flag_tree_loop_vectorize rather than flag_tree_vectorize.
23295 2017-07-27  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
23297         PR target/81534
23298         * config/s390/s390.md ("*atomic_compare_and_swap<mode>_1")
23299         ("*atomic_compare_and_swapdi_2", "*atomic_compare_and_swapsi_3"):
23300         Change s_operand to memory_operand.
23302 2017-07-27  Richard Sandiford  <richard.sandiford@linaro.org>
23304         * config/rs6000/rs6000-protos.h (rs6000_emit_le_vsx_permute): Declare.
23305         * config/rs6000/rs6000.c (rs6000_gen_le_vsx_permute): Replace with...
23306         (rs6000_emit_le_vsx_permute): ...this.  Take the destination as input.
23307         Emit instructions rather than returning an expression.  Handle TFmode
23308         and KFmode by casting to TImode.
23309         (rs6000_emit_le_vsx_load): Update to use rs6000_emit_le_vsx_permute.
23310         (rs6000_emit_le_vsx_store): Likewise.
23311         * config/rs6000/vsx.md (VSX_TI): New iterator.
23312         (*vsx_le_permute_<mode>): Use it instead of VSX_LE_128.
23313         (*vsx_le_undo_permute_<mode>): Likewise.
23314         (*vsx_le_perm_load_<mode>): Use rs6000_emit_le_vsx_permute to
23315         emit the split sequence.
23316         (*vsx_le_perm_store_<mode>): Likewise.
23318 2017-07-27  Jakub Jelinek  <jakub@redhat.com>
23320         PR tree-optimization/81555
23321         PR tree-optimization/81556
23322         * tree-ssa-reassoc.c (rewrite_expr_tree): Add NEXT_CHANGED argument,
23323         if true, force CHANGED for the recursive invocation.
23324         (reassociate_bb): Remember original length of ops array, pass
23325         len != orig_len as NEXT_CHANGED in rewrite_expr_tree call.
23327         * attribs.c (decl_attributes): Imply noinline, noclone and no_icf
23328         attributes for noipa attribute.  For naked attribute use
23329         lookup_attribute first before lookup_attribute_spec.
23330         * final.c (rest_of_handle_final): Disable IPA RA for functions with
23331         noipa attribute.
23332         * ipa-visibility.c (non_local_p): Fix comment typos.  Return true
23333         for functions with noipa attribute.
23334         (cgraph_externally_visible_p): Return true for functions with noipa
23335         attribute.
23336         * cgraph.c (cgraph_node::get_availability): Return AVAIL_INTERPOSABLE
23337         for functions with noipa attribute.
23338         * doc/extend.texi: Document noipa function attribute.
23339         * tree-ssa-structalias.c (refered_from_nonlocal_fn): Set *nonlocal_p
23340         also for functions with noipa attribute.
23341         (ipa_pta_execute): Set nonlocal_p also for nodes with noipa attribute.
23343 2017-07-26  Andrew Pinski  <apinski@cavium.com>
23345         * config/aarch64/aarch64.c (thunderx_vector_cost): Decrease cost of
23346         vec_unalign_load_cost and vec_unalign_store_cost.
23348 2017-07-26  Michael Meissner  <meissner@linux.vnet.ibm.com>
23350         * config/rs6000/rs6000-cpus.def (ISA_2_7_MASKS_SERVER): Delete
23351         -mvsx-small-integer option.
23352         (ISA_3_0_MASKS_IEEE): Likewise.
23353         (OTHER_VSX_VECTOR_MASKS): Likewise.
23354         (POWERPC_MASKS): Likewise.
23355         * config/rs6000/rs6000.opt (-mvsx-small-integer): Likewise.
23356         * config/rs6000/rs6000.c (rs6000_hard_regno_mode_ok): Simplify
23357         code, only testing for DImode being allowed in non-VSX floating
23358         point registers.
23359         (rs6000_init_hard_regno_mode_ok): Change TARGET_VSX_SMALL_INTEGER
23360         to TARGET_P8_VECTOR test.  Remove redundant VSX test inside of
23361         another VSX test.
23362         (rs6000_option_override_internal): Delete -mvsx-small-integer.
23363         (rs6000_expand_vector_set): Change TARGET_VSX_SMALL_INTEGER to
23364         TARGET_P8_VECTOR test.
23365         (rs6000_secondary_reload_simple_move): Likewise.
23366         (rs6000_preferred_reload_class): Delete TARGET_VSX_SMALL_INTEGER,
23367         since TARGET_P9_VECTOR was already tested.
23368         (rs6000_opt_masks): Remove -mvsx-small-integer.
23369         * config/rs6000/vsx.md (vsx_extract_<mode>): Delete
23370         TARGET_VSX_SMALL_INTEGER, since a test for TARGET_P9_VECTOR was
23371         used.
23372         (vsx_extract_<mode>_p9): Delete TARGET_VSX_SMALL_INTEGER, since a
23373         test for TARGET_VEXTRACTUB was used, and that uses
23374         TARGET_P9_VECTOR.
23375         (p9 extract splitter): Likewise.
23376         (vsx_extract_<mode>_di_p9): Likewise.
23377         (vsx_extract_<mode>_store_p9): Likewise.
23378         (vsx_extract_si): Delete TARGET_VSX_SMALL_INTEGER, since a test
23379         for TARGET_P9_VECTOR was used.  Delete code that is now dead with
23380         the elimination of TARGET_VSX_SMALL_INTEGER.
23381         (vsx_extract_<mode>_p8): Likewise.
23382         (vsx_ext_<VSX_EXTRACT_I:VS_scalar>_fl_<FL_CONV:mode>): Likewise.
23383         (vsx_ext_<VSX_EXTRACT_I:VS_scalar>_ufl_<FL_CONV:mode>): Likewise.
23384         (vsx_set_<mode>_p9): Likewise.
23385         (vsx_set_v4sf_p9): Likewise.
23386         (vsx_set_v4sf_p9_zero): Likewise.
23387         (vsx_insert_extract_v4sf_p9): Likewise.
23388         (vsx_insert_extract_v4sf_p9_2): Likewise.
23389         * config/rs6000/rs6000.md (sign extend splitter): Change
23390         TARGET_VSX_SMALL_INTEGER to TARGET_P8_VECTOR test.
23391         (floatsi<mode>2_lfiwax_mem): Likewise.
23392         (floatunssi<mode>2_lfiwzx_mem): Likewise.
23393         (float<QHI:mode><FP_ISA3:mode>2): Delete TARGET_VSX_SMALL_INTEGER,
23394         since a test for TARGET_P9_VECTOR was used.
23395         (float<QHI:mode><FP_ISA3:mode>2_internal): Likewise.
23396         (floatuns<QHI:mode><FP_ISA3:mode>2): Likewise.
23397         (floatuns<QHI:mode><FP_ISA3:mode>2_internal): Likewise.
23398         (fix_trunc<mode>si2): Change TARGET_VSX_SMALL_INTEGER to
23399         TARGET_P8_VECTOR test.
23400         (fix_trunc<mode>si2_stfiwx): Likewise.
23401         (fix_trunc<mode>si2_internal): Likewise.
23402         (fix_trunc<SFDF:mode><QHI:mode>2): Delete
23403         TARGET_VSX_SMALL_INTEGER, since a test for TARGET_P9_VECTOR was
23404         used.
23405         (fix_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
23406         (fixuns_trunc<mode>si2): Change TARGET_VSX_SMALL_INTEGER to
23407         TARGET_P8_VECTOR test.
23408         (fixuns_trunc<mode>si2_stfiwx): Likewise.
23409         (fixuns_trunc<SFDF:mode><QHI:mode>2): Delete
23410         TARGET_VSX_SMALL_INTEGER, since a test for TARGET_P9_VECTOR was
23411         used.
23412         (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
23413         (fctiw<u>z_<mode>_smallint): Delete TARGET_VSX_SMALL_INTEGER,
23414         since a test for TARGET_P9_VECTOR was used.
23415         (splitter for loading small constants): Likewise.
23417 2017-07-26  Andrew Pinski  <apinski@cavium.com>
23419         * config/aarch64/aarch64.c (thunderx_vector_cost): Fix
23420         vec_fp_stmt_cost.
23422 2017-07-26  H.J. Lu  <hongjiu.lu@intel.com>
23424         PR target/81563
23425         * config/i386/i386.c (sp_valid_at): Properly check CFA offset.
23426         (fp_valid_at): Likewise.
23428 2017-07-26  James Greenhalgh  <james.greenhalgh@arm.com>
23430         * config/aarch64/aarch64.c (cortexa57_addrcost_table): Remove.
23431         (qdf24xx_addrcost_table): Likewise.
23432         (cortexa57_tunings): Update to use generic_branch_cost.
23433         (cortexa72_tunings): Likewise.
23434         (cortexa73_tunings): Likewise.
23435         (qdf24xx_tunings): Likewise.
23437 2017-07-26  James Greenhalgh  <james.greenhalgh@arm.com>
23439         * config/aarch64/aarch64.c (cortexa57_branch_cost): Remove.
23440         (thunderx2t99_branch_cost): Likewise.
23441         (cortexa35_tunings): Update to use generic_branch_cost.
23442         (cortexa53_tunings): Likewise.
23443         (cortexa57_tunings): Likewise.
23444         (cortexa72_tunings): Likewise.
23445         (cortexa73_tunings): Likewise.
23446         (thunderx2t99_tunings): Likewise.
23448 2017-07-26  Sebastian Huber  <sebastian.huber@embedded-brains.de>
23450         * config/sparc/sparc.c (dump_target_flag_bits): Dump MASK_FSMULD.
23451         (sparc_option_override): Honour MASK_FSMULD.
23452         * config/sparc/sparc.h (MASK_FEATURES): Add MASK_FSMULD.
23453         * config/sparc/sparc.md (muldf3_extend): Use TARGET_FSMULD.
23454         * config/sparc/sparc.opt (mfsmuld): New option.
23455         * doc/invoke.texi (mfsmuld): Document option.
23457 2017-07-26  Marek Polacek  <polacek@redhat.com>
23459         PR middle-end/70992
23460         * tree.c (build2_stat): Don't set TREE_CONSTANT on divisions by zero.
23462 2017-07-26  Richard Biener  <rguenther@suse.de>
23464         * gimple-match-head.c (do_valueize): Return OP if valueize
23465         returns NULL_TREE.
23466         (get_def): New helper to get at the def stmt of a SSA name
23467         if valueize allows.
23468         * genmatch.c (dt_node::gen_kids_1): Use get_def instead of
23469         do_valueize to get at the def stmt.
23470         (dt_operand::gen_gimple_expr): Simplify do_valueize calls.
23472 2017-07-26  Wilco Dijkstra  <wdijkstr@arm.com>
23474         PR middle-end/46932
23475         * auto-inc-dec.c (parse_add_or_inc): Block autoinc on sfp.
23477 2017-07-26  Martin Liska  <mliska@suse.cz>
23479         PR sanitize/81186
23480         * function.c (expand_function_start): Make expansion of
23481         nonlocal_goto_save_area after parm_birth_insn.
23483 2017-07-26  Sebastian Huber  <sebastian.huber@embedded-brains.de>
23485         * config/sparc/sparc.c (sparc_option_override): Remove MASK_FPU
23486         from all CPU target flags enable members.
23488 2017-07-26  Richard Biener  <rguenther@suse.de>
23490         * genmatch.c (dt_simplify::gen): Make iterator vars const.
23491         (decision_tree::gen): Make 'type' const.
23492         (write_predicate): Likewise.
23494 2017-07-24  Michael Meissner  <meissner@linux.vnet.ibm.com>
23496         * config/rs6000/rs6000.c (rs6000_init_hard_regno_mode_ok):
23497         Eliminate TARGET_UPPER_REGS_{DF,DI,SF} usage.
23498         (rs6000_option_override_internal): Likewise.
23499         (rs6000_expand_vector_set): Likewise.
23500         * config/rs6000/rs6000.h (TARGET_UPPER_REGS_DF): Delete.
23501         (TARGET_UPPER_REGS_SF): Likewise.
23502         (TARGET_UPPER_REGS_DI): Likewise.
23503         (TARGET_VEXTRACTUB): Eliminate TARGET_UPPER_REGS_{DF,DI,SF}.
23504         (TARGET_DIRECT_MOVE_64BIT): Likewise.
23505         * config/rs6000/rs6000.md (ALTIVEC_DFORM): Likewise.
23506         (float<QHI:mode><FP_ISA3:mode>2_internal): Likewise.
23507         (Splitters for DI constants in Altivec registers): Likewise.
23508         * config/rs6000/vsx.md (vsx_set_<mode>_p9): Likewise.
23509         (vsx_set_v4sf_p9): Likewise.
23510         (vsx_set_v4sf_p9_zero): Likewise.
23511         (vsx_insert_extract_v4sf_p9): Likewise.
23512         (vsx_insert_extract_v4sf_p9_2): Likewise.
23514 2017-07-25  Carl Love  <cel@us.ibm.com>
23516         * doc/extend.texi: Update the built-in documentation file for the
23517         existing built-in functions
23518         vector signed char vec_cnttz (vector signed char);
23519         vector unsigned char vec_cnttz (vector unsigned char);
23520         vector signed short vec_cnttz (vector signed short);
23521         vector unsigned short vec_cnttz (vector unsigned short);
23522         vector signed int vec_cnttz (vector signed int);
23523         vector unsigned int vec_cnttz (vector unsigned int);
23524         vector signed long long vec_cnttz (vector signed long long);
23525         vector unsigned long long vec_cnttz (vector unsigned long long);
23527 2017-07-25  Andrew Pinski  <apinski@cavium.com>
23529         * tree-ssa-uninit.c (warn_uninitialized_vars): Don't warn about memory
23530         accesses where the use is for the first operand of a BIT_INSERT.
23532 2017-07-25  Jim Wilson  <jim.wilson@linaro.org>
23534         PR bootstrap/81521
23535         * config/i386/winnt-cxx.c (i386_pe_adjust_class_at_definition): Look
23536         for FUNCTION_DECLs in TYPE_FIELDS rather than TYPE_METHODS.
23538 2017-07-25  Jim Wilson  <jim.wilson@linaro.org>
23540         * config/i386/gstabs.h: Delete.
23541         * config/i386/openbsd.h, config/i386/t-openbsd: Likewise.
23543 2017-07-25  Uros Bizjak  <ubizjak@gmail.com>
23545         * config/i386/i386.c (ix86_decompose_address): Do not check for
23546         register RTX when looking at index_reg or base_reg.
23547         * config/i386/i386.h (INCOMING_RETURN_ADDR_RTX): Use stack_pointer_rtx.
23549 2017-07-25  Eric Botcazou  <ebotcazou@adacore.com>
23551         * gimple.c (gimple_assign_set_rhs_with_ops): Do not ask gsi_replace
23552         to update EH info here.
23554 2017-07-25  Alexander Monakov  <amonakov@ispras.ru>
23556         * match.pd ((X * CST1) * CST2): Simplify to X * (CST1 * CST2).
23558 2017-07-25  Alexander Monakov  <amonakov@ispras.ru>
23560         * match.pd ((X * CST) * Y): Reassociate to (X * Y) * CST.
23562 2017-07-25  Torsten Duwe  <duwe@suse.de>
23564         * common.opt: Introduce -fpatchable-function-entry
23565         command line option, and its variables function_entry_patch_area_size
23566         and function_entry_patch_area_start.
23567         * opts.c (common_handle_option): Add -fpatchable_function_entry_ case,
23568         including a two-value parser.
23569         * target.def (print_patchable_function_entry): New target hook.
23570         * targhooks.h (default_print_patchable_function_entry): New function.
23571         * targhooks.c (default_print_patchable_function_entry): Likewise.
23572         * toplev.c (process_options): Switch off IPA-RA if
23573         patchable function entries are being generated.
23574         * varasm.c (assemble_start_function): Look at the
23575         patchable-function-entry command line switch and current
23576         function attributes and maybe generate NOP instructions by
23577         calling the print_patchable_function_entry hook.
23578         * doc/extend.texi: Document patchable_function_entry attribute.
23579         * doc/invoke.texi: Document -fpatchable_function_entry
23580         command line option.
23581         * doc/tm.texi.in (TARGET_ASM_PRINT_PATCHABLE_FUNCTION_ENTRY):
23582         New target hook.
23583         * doc/tm.texi: Re-generate.
23585 2017-07-25  Jakub Jelinek  <jakub@redhat.com>
23587         PR target/81532
23588         * config/i386/constraints.md (Yd, Ye): Use ALL_SSE_REGS for
23589         TARGET_AVX512DQ rather than TARGET_AVX512BW.
23591 2017-07-25  Tamar Christina  <tamar.christina@arm.com>
23593         * config/arm/parsecpu.awk (all_cores): Remove duplicates.
23595 2017-07-25  Richard Biener  <rguenther@suse.de>
23597         PR tree-optimization/81455
23598         * tree-ssa-loop-unswitch.c (find_loop_guard): Make sure to
23599         not walk in cycles when looking for guards.
23601 2017-07-25  Richard Biener  <rguenther@suse.de>
23603         PR tree-optimization/81529
23604         * tree-vect-stmts.c (process_use): Disregard live induction PHIs
23605         when optimizing backedge uses.
23607 2017-07-25  David Edelsohn  <dje.gcc@gmail.com>
23609         * dwarf2asm.c (dw2_asm_output_nstring): Encode double quote
23610         character for AIX.
23611         * dwarf2out.c (output_macinfo): Copy debug_line_section_label
23612         to dl_section_ref.  On AIX, append an expression to subtract
23613         the size of the section length to dl_section_ref.
23615 2017-07-25  Segher Boessenkool  <segher@kernel.crashing.org>
23617         * configure.ac: If any of the config.* scripts fail, exit 1.
23618         * configure: Regenerate.
23620 2017-07-25  Richard Biener  <rguenther@suse.de>
23622         PR middle-end/81546
23623         * tree-ssa-operands.c (verify_imm_links): Remove cap on number
23624         of immediate uses, be more verbose on errors.
23626 2017-07-25  Richard Biener  <rguenther@suse.de>
23628         PR tree-optimization/81510
23629         * tree-vect-loop.c (vect_is_simple_reduction): When the
23630         reduction stmt is not inside the loop bail out.
23632 2017-07-25  Richard Biener  <rguenther@suse.de>
23634         PR tree-optimization/81303
23635         * tree-vect-loop-manip.c (vect_loop_versioning): Build
23636         profitability check against LOOP_VINFO_NITERSM1.
23638 2017-07-25  Alexander Monakov  <amonakov@ispras.ru>
23640         * domwalk.c (cmp_bb_postorder): Simplify.
23641         (sort_bbs_postorder): New function.  Use it...
23642         (dom_walker::walk): ...here to optimize common cases.
23644 2017-07-25  Martin Liska  <mliska@suse.cz>
23646         PR ipa/81520
23647         * ipa-visibility.c (function_and_variable_visibility): Make the
23648         redirection just on target that supports aliasing.
23649         Fix GNU coding style.
23651 2017-07-25  Sebastian Huber  <sebastian.huber@embedded-brains.de>
23653         PR libgcc/61152
23654         * config/aarch64/rtems.h: Add GCC Runtime Library Exception.
23655         Format changes.
23656         * config/arm/rtems.h: Likewise.
23657         * config/bfin/rtems.h: Likewise.
23658         * config/i386/rtemself.h: Likewise.
23659         * config/lm32/rtems.h: Likewise.
23660         * config/m32c/rtems.h: Likewise.
23661         * config/m68k/rtemself.h: Likewise.
23662         * config/microblaze/rtems.h: Likewise.
23663         * config/mips/rtems.h: Likewise.
23664         * config/moxie/rtems.h: Likewise.
23665         * config/nios2/rtems.h: Likewise.
23666         * config/powerpcspe/rtems.h: Likewise.
23667         * config/rs6000/rtems.h: Likewise.
23668         * config/rtems.h: Likewise.
23669         * config/sh/rtems.h: Likewise.
23670         * config/sh/rtemself.h: Likewise.
23671         * config/sparc/rtemself.h: Likewise.
23673 2017-07-25  Georg-Johann Lay  <avr@gjlay.de>
23675         PR 81487
23676         * hsa-brig.c (brig_init): Use xasprintf instead of asprintf.
23677         * gimple-pretty-print.c (dump_profile, dump_probability): Same.
23678         * tree-ssa-structalias.c (alias_get_name): Same.
23680 2017-07-25  Bin Cheng  <bin.cheng@arm.com>
23682         PR target/81414
23683         * config/aarch64/cortex-a57-fma-steering.c (analyze): Skip fmul/fmac
23684         instructions if no du chain is found.
23686 2017-07-25  Georg-Johann Lay  <avr@gjlay.de>
23688         * config/avr/avr-log.c (avr_log_vadump) ['T']: Print NULL-TREE.
23690 2017-07-25  Richard Biener  <rguenther@suse.de>
23692         PR middle-end/81505
23693         * fold-const.c (fold_negate_const): TREE_OVERFLOW should be
23694         sticky.
23696 2017-07-24  Michael Meissner  <meissner@linux.vnet.ibm.com>
23698         * config/rs6000/rs6000-cpus.def (ISA_2_6_MASKS_SERVER): Delete
23699         upper-regs options.
23700         (ISA_2_7_MASKS_SERVER): Likewise.
23701         (ISA_3_0_MASKS_IEEE): Likewise.
23702         (OTHER_P8_VECTOR_MASKS): Likewise.
23703         (OTHER_VSX_VECTOR_MASKS): Likewise.
23704         (POWERPC_MASKS): Likewise.
23705         (power7 cpu): Use ISA_2_6_MASKS_SERVER instead of using a
23706         duplicate list of options.
23707         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove
23708         explicit -mupper-regs options.
23709         * config/rs6000/rs6000.opt (-mvsx-scalar-memory): Delete
23710         -mupper-regs* options.  Delete -mvsx-scalar-memory, which was an
23711         alias for -mupper-regs-df.
23712         * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): Likewise.
23713         (rs6000_init_hard_regno_mode_ok): Likewise.
23714         (rs6000_option_override_internal): Likewise.
23715         (rs6000_opt_masks): Likewise.
23716         * config/rs6000/rs6000.h (TARGET_UPPER_REGS_DF): Define upper regs
23717         options in terms of whether -mvsx or -mpower8-vector was used.
23718         (TARGET_UPPER_REGS_DI): Likewise.
23719         (TARGET_UPPER_REGS_SF): Likewise.
23720         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete the
23721         -mupper-regs-* options.
23723 2017-07-24  Segher Boessenkool  <segher@kernel.crashing.org>
23725         * passes.c (emergency_dump_function): Print some empty lines and a
23726         header before the RTL dump.
23728 2017-07-24  Segher Boessenkool  <segher@kernel.crashing.org>
23730         * cfgrtl.c (rtl_dump_bb): Don't call NEXT_INSN on NULL.
23732 2017-07-24  Wilco Dijkstra  <wdijkstr@arm.com>
23734         PR target/79041
23735         * config/aarch64/aarch64.c (aarch64_classify_symbol):
23736         Avoid SYMBOL_SMALL_ABSOLUTE for literals with pc-relative literals.
23738 2017-07-24  Carl Love  <cel@us.ibm.com>
23740         * config/rs6000/rs6000-c.c: Add support for built-in functions
23741         vector float vec_extract_fp32_from_shorth (vector unsigned short);
23742         vector float vec_extract_fp32_from_shortl (vector unsigned short);
23743         * config/rs6000/altivec.h (vec_extract_fp_from_shorth,
23744         vec_extract_fp_from_shortl): Add defines for the two builtins.
23745         * config/rs6000/rs6000-builtin.def (VEXTRACT_FP_FROM_SHORTH,
23746         VEXTRACT_FP_FROM_SHORTL): Add BU_P9V_OVERLOAD_1 and BU_P9V_VSX_1
23747         new builtins.
23748         * config/rs6000/vsx.md vsx_xvcvhpsp): Add define_insn.
23749         (vextract_fp_from_shorth, vextract_fp_from_shortl): Add define_expands.
23750         * doc/extend.texi: Update the built-in documentation file for the
23751         new built-in function.
23753 2017-07-24  Jakub Jelinek  <jakub@redhat.com>
23755         PR bootstrap/81521
23756         * tree.def: Remove TYPE_METHODS documentation, adjust TYPE_FIELDS
23757         documentation.
23758         * doc/generic.texi: Likewise.
23759         * config/i386/winnt-cxx.c (i386_pe_adjust_class_at_definition): Look
23760         for FUNCTION_DECLs in TYPE_FIELDS rather than TYPE_METHODS.
23762 2017-07-24  Jackson Woodruff  <jackson.woodruff@arm.com>
23764         * config/aarch64/aarch64-simd.md (aarch64_mla_elt_merge<mode>): New.
23765         (aarch64_mls_elt_merge<mode>): Likewise.
23767 2017-07-23  Krister Walfridsson  <krister.walfridsson@gmail.com>
23769         * config.gcc (*-*-netbsd*): Remove check for NetBSD versions not
23770         having __cxa_atexit.
23772 2017-07-23  Michael Collison  <michael.collison@arm.com>
23774         * config/arm/arm.c (arm_option_override): Deprecate
23775         use of -mstructure-size-boundary.
23776         * config/arm/arm.opt: Deprecate -mstructure-size-boundary.
23777         * doc/invoke.texi: Deprecate -mstructure-size-boundary.
23779 2017-07-23  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
23781         PR target/80695
23782         * config/rs6000/rs6000.c (rs6000_builtin_vectorization_cost):
23783         Reduce cost estimate for direct moves.
23785 2017-07-23  Uros Bizjak  <ubizjak@gmail.com>
23787         PR target/80569
23788         * config/i386/i386.c (ix86_option_override_internal): Disable
23789         BMI, BMI2 and TBM instructions for -m16.
23791 2017-07-21  Carl Love  <cel@us.ibm.com>
23793         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
23794         ALTIVEC_BUILTIN_VMULESW, ALTIVEC_BUILTIN_VMULEUW,
23795         ALTIVEC_BUILTIN_VMULOSW, ALTIVEC_BUILTIN_VMULOUW entries.
23796         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin,
23797         builtin_function_type): Add ALTIVEC_BUILTIN_* case statements.
23798         * config/rs6000/altivec.md (MVULEUW, VMULESW, VMULOUW,
23799         VMULOSW): New enum "unspec" values.
23800         (altivec_vmuleuw, altivec_vmulesw, altivec_vmulouw,
23801         altivec_vmulosw): New patterns.
23802         * config/rs6000/rs6000-builtin.def (VMLEUW, VMULESW, VMULOUW,
23803         VMULOSW): Add definitions.
23805 2017-07-21  Jim Wilson  <jim.wilson@linaro.org>
23807         * config/aarch64/aarch64-cores.def (falkor): Add AARCH64_FL_RDMA.
23808         (qdf24xx): Likewise.
23809         * config/aarch64/aarch64-options-extensions.def (rdma); New.
23810         * config/aarch64/aarch64.h (AARCH64_FL_RDMA): New.
23811         (AARCH64_FL_V8_1): Renumber.
23812         (AARCH64_FL_FOR_ARCH8_1): Add AARCH64_FL_RDMA.
23813         (AARCH64_ISA_RDMA): Use AARCH64_FL_RDMA.
23814         * config/aarch64/arm_neon.h: Use +rdma instead of arch=armv8.1-a.
23815         * doc/invoke.texi (AArch64 Options): Mention +rmda in -march docs.  Add
23816         rdma to feature modifiers list.
23818 2017-07-21  Yury Gribov  <tetra2005@gmail.com>
23820         PR middle-end/56727
23821         * ipa-visibility (function_and_variable_visibility): Convert
23822         recursive PLT call to direct call if appropriate.
23824 2017-07-21  Andrew Pinski  <apinski@cavium.com>
23826         * tree-ssa-sccvn.c (vn_nary_op_eq): Check BIT_INSERT_EXPR's
23827         operand 1 to see if the types precision matches.
23828         * fold-const.c (operand_equal_p): Likewise.
23830 2017-07-21  Richard Biener  <rguenther@suse.de>
23832         PR tree-optimization/81303
23833         * tree-vect-data-refs.c (vect_get_peeling_costs_all_drs): Pass
23834         in datarefs vector.  Allow NULL dr0 for no peeling cost estimate.
23835         (vect_peeling_hash_get_lowest_cost): Adjust.
23836         (vect_enhance_data_refs_alignment): Likewise.  Use
23837         vect_get_peeling_costs_all_drs to compute the penalty for no
23838         peeling to match up costs.
23840 2017-07-21  Richard Biener  <rguenther@suse.de>
23842         PR tree-optimization/81500
23843         * tree-vect-loop.c (vect_is_simple_reduction): Properly fail if
23844         we didn't identify a reduction path.
23846 2017-07-21  Tom de Vries  <tom@codesourcery.com>
23847             Cesar Philippidis  <cesar@codesourcery.com>
23849         PR gcov-profile/81442
23850         * config/nvptx/nvptx.c (nvptx_goacc_reduction_init): Add missing edge
23851         probabilities.
23853 2017-07-21  Tom de Vries  <tom@codesourcery.com>
23855         PR lto/81430
23856         * config/nvptx/nvptx.c (nvptx_override_options_after_change): New
23857         function.
23858         (TARGET_OVERRIDE_OPTIONS_AFTER_CHANGE): Define to
23859         nvptx_override_options_after_change.
23861 2017-07-21  Ulrich Drepper  <drepper@redhat.com>
23863         * dwarf2out.c (output_file_names): Avoid double testing for
23864         dwarf_version >= 5.
23866 2017-07-21  Georg-Johann Lay  <avr@gjlay.de>
23868         * doc/invoke.texi (AVR Built-in Functions): Re-layout section.
23870 2016-07-21  Jan Hubicka  <hubicka@ucw.cz>
23872         * cfgcleanup.c (flow_find_cross_jump): Do not crossjump across
23873         hot/cold regions.
23874         (try_crossjump_to_edge): Do not punt on partitioned functions.
23876 2016-07-21  Jan Hubicka  <hubicka@ucw.cz>
23878         * bb-reorder.c (find_rarely_executed_basic_blocks_and_crossing_edges):
23879         Put all BBs reachable only via paths crossing cold region to cold
23880         region.
23881         * cfgrtl.c (find_bbs_reachable_by_hot_paths): New function.
23883 2016-07-21  Richard Biener  <rguenther@suse.de>
23885         PR tree-optimization/81303
23886         * tree-vect-loop.c (vect_estimate_min_profitable_iters): Take
23887         into account prologue and epilogue iterations when raising
23888         min_profitable_iters to sth at least covering one vector iteration.
23890 2017-07-21  Tamar Christina  <tamar.christina@arm.com>
23892         * config/arm/arm.c (arm_test_cpu_arch_dat):
23893         Check for overlap.
23895 2017-07-20  Nathan Sidwell  <nathan@acm.org>
23897         Remove TYPE_METHODS.
23898         * tree.h (TYPE_METHODS): Delete.
23899         * dwarf2out.c (gen_member_die): Member fns are on TYPE_FIELDS.
23900         * dbxout.c (dbxout_type_fields): Ignore FUNCTION_DECLs.
23901         (dbxout_type_methods): Scan TYPE_FIELDS.
23902         (dbxout_type): Don't check TYPE_METHODS here.
23903         * function.c (use_register_for_decl): Always ignore register for
23904         class types when not optimizing.
23905         * ipa-devirt.c (odr_types_equivalent_p): Delete TYPE_METHODS scan.
23906         * tree.c (free_lang_data_in_type): Stitch out member functions and
23907         templates from TYPE_FIELDS.
23908         (build_distinct_type_copy, verify_type_variant,
23909         verify_type): Member fns are on TYPE_FIELDS.
23910         * tree-dump.c (dequeue_and_dump): No TYPE_METHODS.
23911         * tree-pretty-print.c (dump_generic_node): Likewise.
23913 2017-07-20  Jakub Jelinek  <jakub@redhat.com>
23915         PR target/80846
23916         * config/i386/i386.c (ix86_expand_vector_init_general): Handle
23917         V2TImode and V4TImode.
23918         (ix86_expand_vector_extract): Likewise.
23919         * config/i386/sse.md (VMOVE): Enable V4TImode even for just
23920         TARGET_AVX512F, instead of only for TARGET_AVX512BW.
23921         (ssescalarmode): Handle V4TImode and V2TImode.
23922         (VEC_EXTRACT_MODE): Add V4TImode and V2TImode.
23923         (*vec_extractv2ti, *vec_extractv4ti): New insns.
23924         (VEXTRACTI128_MODE): New mode iterator.
23925         (splitter for *vec_extractv?ti first element): New.
23926         (VEC_INIT_MODE): New mode iterator.
23927         (vec_init<mode>): Consolidate 3 expanders into one using
23928         VEC_INIT_MODE mode iterator.
23930 2017-07-20  Alexander Monakov  <amonakov@ispras.ru>
23932         * lra-assigns.c (pseudo_compare_func): Fix comparison step based on
23933         non_spilled_static_chain_regno_p.
23935 2017-07-20  Alexander Monakov  <amonakov@ispras.ru>
23937         * gimple-ssa-store-merging.c (sort_by_bitpos): Return 0 on equal bitpos.
23939 2017-07-20  Jan Hubicka  <hubicka@ucw.cz>
23941         * bb-reorder.c (connect_traces): Allow copying of blocks within
23942         single partition.
23944 2017-07-20  Richard Biener  <rguenther@suse.de>
23946         * gimple.h (gimple_phi_result): Add gphi * overload.
23947         (gimple_phi_result_ptr): Likewise.
23948         (gimple_phi_arg): Likewise.  Adjust index assert to only
23949         allow actual argument accesses rather than all slots available
23950         by capacity.
23951         (gimple_phi_arg_def): Add gphi * overload.
23952         * tree-phinodes.c (make_phi_node): Initialize only actual
23953         arguments.
23954         (resize_phi_node): Clear memory not covered by old node,
23955         do not initialize excess argument slots.
23956         (reserve_phi_args_for_new_edge): Initialize new argument slot
23957         completely.
23959 2017-07-20  Bin Cheng  <bin.cheng@arm.com>
23961         PR tree-optimization/81388
23962         Revert r238585:
23963         2016-07-21  Bin Cheng  <bin.cheng@arm.com>
23965         * tree-ssa-loop-niter.c (number_of_iterations_lt_to_ne): Clean up
23966         by removing computation of may_be_zero.
23968 2017-07-18  Jan Hubicka  <hubicka@ucw.cz>
23969             Tom de Vries  <tom@codesourcery.com>
23971         PR middle-end/81030
23972         * cfgbuild.c (find_many_sub_basic_blocks): Update REG_BR_PROB note
23973         when gimple level profile disagrees with what RTL expander did.
23975 2017-07-20  Richard Biener  <rguenther@suse.de>
23977         PR tree-optimization/61171
23978         * tree-vectorizer.h (slp_instance): Add reduc_phis member.
23979         (vect_analyze_stmt): Add slp instance parameter.
23980         (vectorizable_reduction): Likewise.
23981         * tree-vect-loop.c (vect_analyze_loop_operations): Adjust.
23982         (vect_is_simple_reduction): Deal with chains not detected
23983         as SLP reduction chain, specifically not properly associated
23984         chains containing a mix of plus/minus.
23985         (get_reduction_op): Remove.
23986         (get_initial_defs_for_reduction): Simplify, pass in whether
23987         this is a reduction chain, pass in the SLP node for the PHIs.
23988         (vect_create_epilog_for_reduction): Get the SLP instance as
23989         arg and adjust.
23990         (vectorizable_reduction): Get the SLP instance as arg.
23991         During analysis remember the SLP node with the PHIs in the
23992         instance.  Simplify getting at the vectorized reduction PHIs.
23993         * tree-vect-slp.c (vect_slp_analyze_node_operations): Pass
23994         through SLP instance.
23995         (vect_slp_analyze_operations): Likewise.
23996         * tree-vect-stms.c (vect_analyze_stmt): Likewise.
23997         (vect_transform_stmt): Likewise.
23999 2017-07-20  Tom de Vries  <tom@codesourcery.com>
24001         PR tree-optimization/81489
24002         * gimple-ssa-isolate-paths.c (find_implicit_erroneous_behavior): Move
24003         read of phi arg location to before loop that modifies phi.
24005 2017-07-20  Naveen H.S  <Naveen.Hurugalawadi@cavium.com>
24007         * match.pd (((m1 >/</>=/<= m2) * d -> (m1 >/</>=/<= m2) ? d : 0):
24008         New pattern.
24010 2017-07-19  Jan Hubicka  <hubicka@ucw.cz>
24012         PR middle-end/81331
24013         * except.c (execute): Fix ordering issue.
24015 2018-07-19  Segher Boessenkool  <segher@kernel.crashing.org>
24017         PR rtl-optimization/81423
24018         * combine.c (make_compound_operation_int): Don't try to optimize
24019         the AND of a SUBREG of an LSHIFTRT if that SUBREG is paradoxical.
24021 2017-07-19  Segher Boessenkool  <segher@kernel.crashing.org>
24023         PR rtl-optimization/81423
24024         * simplify-rtx.c (simplify_truncation): Handle truncating an IOR
24025         with a constant that is -1 in the truncated to mode.
24027 2017-07-19  Jan Hubicka  <hubicka@ucw.cz>
24029         * predict.c (propagate_unlikely_bbs_forward): Break out from ...
24030         (determine_unlikely_bbs): ... here.
24031         * predict.h (propagate_unlikely_bbs_forward): Declare.
24032         * cfgexpand.c (pass_expand::execute): Use it.
24033         * bb-reorder.c (sanitize_hot_paths): Do not consider known to be
24034         unlikely edges.
24035         (find_rarely_executed_basic_blocks_and_crossing_edges): Use
24036         propagate_unlikely_bbs_forward.
24038 2017-07-19  Jan Hubicka  <hubicka@ucw.cz>
24040         PR middle-end/81331
24041         * except.c (maybe_add_nop_after_section_switch): New function.
24042         (execute): Use it.
24044 2017-07-19  Tom de Vries  <tom@codesourcery.com>
24046         * gimple.h (gimple_phi_set_arg): Make assert more strict.
24048 2017-07-19  Tom de Vries  <tom@codesourcery.com>
24050         * gimple.h (gimple_phi_arg): Make assert more strict.
24052 2017-07-19  Steven Munroe  <munroesj@gcc.gnu.org>
24054         * config.gcc (powerpc*-*-*): Add mmintrin.h.
24055         * config/rs6000/mmintrin.h: New file.
24056         * config/rs6000/x86intrin.h [__ALTIVEC__]: Include mmintrin.h.
24058 2017-07-19  Jakub Jelinek  <jakub@redhat.com>
24060         PR tree-optimization/81346
24061         * match.pd: Optimize (X - 1U) <= INT_MAX-1U into (int) X > 0.
24063 2017-07-19  Tom de Vries  <tom@codesourcery.com>
24065         * config/nvptx/nvptx.md (VECIM): Add V2DI.
24067 2017-07-19  Tom de Vries  <tom@codesourcery.com>
24069         * config/nvptx/nvptx-modes.def: Add V2DImode.
24070         * config/nvptx/nvptx-protos.h (nvptx_data_alignment): Declare.
24071         * config/nvptx/nvptx.c (nvptx_ptx_type_from_mode): Handle V2DImode.
24072         (nvptx_output_mov_insn): Handle lack of mov.b128.
24073         (nvptx_print_operand): Handle 'H' and 'L' codes.
24074         (nvptx_vector_mode_supported): Allow V2DImode.
24075         (nvptx_preferred_simd_mode): New function.
24076         (nvptx_data_alignment): New function.
24077         (TARGET_VECTORIZE_PREFERRED_SIMD_MODE): Redefine to
24078         nvptx_preferred_simd_mode.
24079         * config/nvptx/nvptx.h (STACK_BOUNDARY, BIGGEST_ALIGNMENT): Change from
24080         64 to 128 bits.
24081         (DATA_ALIGNMENT): Define.  Set to nvptx_data_alignment.
24083 2017-07-19  Tom de Vries  <tom@codesourcery.com>
24085         * config/nvptx/nvptx-modes.def: New file.  Add V2SImode.
24086         * config/nvptx/nvptx.c (nvptx_ptx_type_from_mode): Handle V2SImode.
24087         (nvptx_vector_mode_supported): New function.  Allow V2SImode.
24088         (TARGET_VECTOR_MODE_SUPPORTED_P): Redefine to nvptx_vector_mode_supported.
24089         * config/nvptx/nvptx.md (VECIM): New mode iterator. Add V2SI.
24090         (mov<VECIM>_insn): New define_insn.
24091         (define_expand "mov<VECIM>): New define_expand.
24093 2017-07-19  Tom de Vries  <tom@codesourcery.com>
24095         * config/nvptx/nvptx.c (nvptx_print_operand): Handle v2 vector mode.
24097 2017-07-19  Jakub Jelinek  <jakub@redhat.com>
24099         PR tree-optimization/81346
24100         * fold-const.h (fold_div_compare, range_check_type): Declare.
24101         * fold-const.c (range_check_type): New function.
24102         (build_range_check): Use range_check_type.
24103         (fold_div_compare): No longer static, rewritten into
24104         a match.pd helper function.
24105         (fold_comparison): Don't call fold_div_compare here.
24106         * match.pd (X / C1 op C2): New optimization using fold_div_compare
24107         as helper function.
24109 2017-07-19  Nathan Sidwell  <nathan@acm.org>
24111         * tree.h (TYPE_MINVAL, TYPE_MAXVAL): Rename to ...
24112         (TYPE_MIN_VALUE_RAW, TYPE_MAX_VALUE_RAW): ... these.
24113         * tree.c (find_decls_types_r, verify_type): Use
24114         TYPE_{MIN,MAX}_VALUE_RAW.
24115         * lto-streamer-out.c (DFS::DFS_write_tree_body): Likewise.
24116         (hash_tree): Likewise.
24117         * tree-streamer-in.c (lto_input_ts_type_non_common_tree_pointers):
24118         Likewise.
24119         * tree-streamer-out.c (write_ts_type_non_common_tree_pointers):
24120         Likewise.
24122 2017-07-18  Tom de Vries  <tom@codesourcery.com>
24124         PR middle-end/81464
24125         * omp-expand.c (expand_omp_for_static_chunk): Handle
24126         equal-argument loop exit phi.
24128 2017-07-18  Uros Bizjak  <ubizjak@gmail.com>
24130         PR target/81471
24131         * config/i386/i386.md (rorx_immediate_operand): New mode attribute.
24132         (*bmi2_rorx<mode>3_1): Use rorx_immediate_operand as
24133         operand 2 predicate.
24134         (*bmi2_rorxsi3_1_zext): Use const_0_to_31_operand as
24135         operand 2 predicate.
24136         (ror,rol -> rorx splitters): Use const_int_operand as
24137         operand 2 predicate.
24139 2017-06-18  Richard Biener  <rguenther@suse.de>
24141         PR tree-optimization/81410
24142         * tree-vect-stmts.c (vectorizable_load): Properly adjust for
24143         the gap in the ! slp_perm SLP case after each group.
24145 2017-07-18  Jan Hubicka  <hubicka@ucw.cz>
24147         PR middle-end/81463
24148         * cfgloopmanip.c (scale_loop_profile): Watch out for zero frequency
24149         again.
24151 2017-07-18  Jan Hubicka  <hubicka@ucw.cz>
24153         PR middle-end/81462
24154         * predict.c (set_even_probabilities): Cleanup; do not affect
24155         probabilities that are already known.
24156         (combine_predictions_for_bb): Call even when count is set.
24158 2017-07-18  Nathan Sidwell  <nathan@acm.org>
24160         * tree-parloops.c (try_transform_to_exit_first_loop_alt): Use
24161         TYPE_MAX_VALUE.
24163 2017-07-18  Bin Cheng  <bin.cheng@arm.com>
24165         PR target/81408
24166         * tree-ssa-loop-niter.c (number_of_iterations_exit): Dump missed
24167         optimization for loop niter analysis.
24169 2017-07-18  Georg-Johann Lay  <avr@gjlay.de>
24171         PR target/81473
24172         * config/avr/avr.c (avr_optimize_casesi): Don't use
24173         INT8_MIN, INT8_MAX, UINT8_MAX, INT16_MIN, INT16_MAX, UINT16_MAX.
24175 2017-07-18  Robin Dapp  <rdapp@linux.vnet.ibm.com>
24177         * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Remove
24178         body_cost_vec from _vect_peel_extended_info.
24179         (vect_peeling_hash_get_lowest_cost): Do not set body_cost_vec.
24180         (vect_peeling_hash_choose_best_peeling): Remove body_cost_vec and
24181         npeel.
24183 2017-07-18  Bin Cheng  <bin.cheng@arm.com>
24185         * config/arm/arm.c (emit_unlikely_jump): Remove unused var.
24187 2017-07-18  Richard Biener  <rguenther@suse.de>
24189         PR tree-optimization/80620
24190         PR tree-optimization/81403
24191         * tree-ssa-pre.c (phi_translate_1): Clear range and points-to
24192         info when re-using a VN table entry.
24194 2017-07-18  Richard Biener  <rguenther@suse.de>
24196         PR tree-optimization/81418
24197         * tree-vect-loop.c (vectorizable_reduction): Properly compute
24198         vectype_in.  Verify that with lane-reducing reduction operations
24199         we have a single def-use cycle.
24201 2017-07-17  Carl Love  <cel@us.ibm.com>
24203         Revert commit r249424   2017-06-20  Carl Love  <cel@us.ibm.com>
24205         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
24206         ALTIVEC_BUILTIN_VMULESW, ALTIVEC_BUILTIN_VMULEUW,
24207         ALTIVEC_BUILTIN_VMULOSW, ALTIVEC_BUILTIN_VMULOUW entries.
24208         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin,
24209         builtin_function_type): Add ALTIVEC_BUILTIN_* case statements.
24210         * config/rs6000/altivec.md (MVULEUW, VMULESW, VMULOUW,
24211         VMULOSW): New enum "unspec" values.
24212         (vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
24213         vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si,
24214         altivec_vmuleuw, altivec_vmulesw, altivec_vmulouw,
24215         altivec_vmulosw): New patterns.
24216         * config/rs6000/rs6000-builtin.def (VMLEUW, VMULESW, VMULOUW,
24217         VMULOSW): Add definitions.
24219 2017-07-17  Uros Bizjak  <ubizjak@gmail.com>
24221         * config/alpha/alpha.c: Include predict.h.
24223 2017-07-17  Yury Gribov  <tetra2005@gmail.com>
24225         * tree-vrp.c (compare_assert_loc): Fix comparison function
24226         to return predictable results.
24228 2017-07-17  Claudiu Zissulescu  <claziss@synopsys.com>
24230         * config/arc/arc.md (adddi3): Remove support for mexpand-adddi
24231         option.
24232         (subdi3): Likewise.
24233         * config/arc/arc.opt (mexpand-adddi): Deprecate it.
24234         * doc/invoke.texi (mexpand-adddi): Update text.
24236 2017-07-17  Claudiu Zissulescu  <claziss@synopsys.com>
24238         * config/arc/arc.md (clzsi2): Expand to an arc_clzsi2 instruction
24239         that also clobbers the CC register. The old expand code is moved
24240         to ...
24241         (*arc_clzsi2): ... here.
24242         (ctzsi2): Expand to an arc_ctzsi2 instruction that also clobbers
24243         the CC register. The old expand code is moved to ...
24244         (arc_ctzsi2): ... here.
24246 2017-07-17  Claudiu Zissulescu  <claziss@synopsys.com>
24248         * config/arc/arc.opt (mindexed-loads): Use initial value
24249         TARGET_INDEXED_LOADS_DEFAULT.
24250         (mauto-modify-reg): Use initial value
24251         TARGET_AUTO_MODIFY_REG_DEFAULT.
24252         * config/arc/elf.h (TARGET_INDEXED_LOADS_DEFAULT): Define.
24253         (TARGET_AUTO_MODIFY_REG_DEFAULT): Likewise.
24254         * config/arc/linux.h (TARGET_INDEXED_LOADS_DEFAULT): Define.
24255         (TARGET_AUTO_MODIFY_REG_DEFAULT): Likewise.
24257 2017-07-17  Martin Liska  <mliska@suse.cz>
24259         PR sanitizer/81302
24260         * opts.c (finish_options): Do not allow -fgnu-tm
24261         w/ -fsanitize={kernel-,}address.  Say sorry.
24263 2017-07-17  Bin Cheng  <bin.cheng@arm.com>
24265         PR target/81369
24266         * tree-loop-distribution.c (classify_partition): Only assert on
24267         numer of iterations.
24268         (merge_dep_scc_partitions): Delete prameter.  Update function call.
24269         (distribute_loop): Remove code handling loop with unknown niters.
24270         (pass_loop_distribution::execute): Skip loop with unknown niters.
24272 2017-07-17  Bin Cheng  <bin.cheng@arm.com>
24274         PR target/81369
24275         * tree-loop-distribution.c (merge_dep_scc_partitions): Sink call to
24276         function sort_partitions_by_post_order.
24278 2017-07-17  Bin Cheng  <bin.cheng@arm.com>
24280         PR tree-optimization/81374
24281         * tree-loop-distribution.c (pass_loop_distribution::execute): Record
24282         the max index of basic blocks, rather than number of basic blocks.
24284 2017-07-17  Claudiu Zissulescu  <claziss@synopsys.com>
24286         * config/arc/arc-protos.h (arc_legitimate_pc_offset_p): Remove
24287         proto.
24288         (arc_legitimate_pic_operand_p): Likewise.
24289         * config/arc/arc.c (arc_legitimate_pic_operand_p): Remove
24290         function.
24291         (arc_needs_pcl_p): Likewise.
24292         (arc_legitimate_pc_offset_p): Likewise.
24293         (arc_legitimate_pic_addr_p): Remove LABEL_REF case, as this
24294         function is also used in constrains.md.
24295         (arc_legitimate_constant_p): Use arc_legitimate_pic_addr_p to
24296         validate pic constants. Handle CONST_INT, CONST_DOUBLE, MINUS and
24297         PLUS.  Only return true/false in known cases, otherwise assert.
24298         (arc_legitimate_address_p): Remove arc_legitimate_pic_addr_p as it
24299         is already called in arc_legitimate_constant_p.
24300         * config/arc/arc.h (CONSTANT_ADDRESS_P): Consider also LABEL for
24301         pic addresses.
24302         (LEGITIMATE_PIC_OPERAND_P): Use
24303         arc_raw_symbolic_reference_mentioned_p function.
24304         * config/arc/constraints.md (Cpc): Use arc_legitimate_pic_addr_p
24305         function.
24306         (Cal): Likewise.
24307         (C32): Likewise.
24309 2017-07-17  Claudiu Zissulescu  <claziss@synopsys.com>
24310         Andrew Burgess  <andrew.burgess@embecosm.com>
24312         * config/arc/arc-protos.h (arc_compute_function_type): Change prototype.
24313         (arc_return_address_register): New function.
24314         * config/arc/arc.c (arc_handle_fndecl_attribute): New function.
24315         (arc_handle_fndecl_attribute): Add naked attribute.
24316         (TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS): Define.
24317         (TARGET_WARN_FUNC_RETURN): Likewise.
24318         (arc_allocate_stack_slots_for_args): New function.
24319         (arc_warn_func_return): Likewise.
24320         (machine_function): Change type fn_type.
24321         (arc_compute_function_type): Consider new naked function type,
24322         change function return type.
24323         (arc_must_save_register): Adapt to handle new
24324         arc_compute_function_type's return type.
24325         (arc_expand_prologue): Likewise.
24326         (arc_expand_epilogue): Likewise.
24327         (arc_return_address_regs): Delete.
24328         (arc_return_address_register): New function.
24329         (arc_epilogue_uses): Use above function.
24330         * config/arc/arc.h (arc_return_address_regs): Delete prototype.
24331         (arc_function_type): Change encoding, add naked type.
24332         (ARC_INTERRUPT_P): Change to handle the new encoding.
24333         (ARC_FAST_INTERRUPT_P): Likewise.
24334         (ARC_NORMAL_P): Define.
24335         (ARC_NAKED_P): Likewise.
24336         (arc_compute_function_type): Delete prototype.
24337         * config/arc/arc.md (in_ret_delay_slot): Use
24338         arc_return_address_register function.
24339         (simple_return): Likewise.
24340         (p_return_i): Likewise.
24342 2017-07-17  Jakub Jelinek  <jakub@redhat.com>
24344         PR tree-optimization/81428
24345         * match.pd (X / X -> one): Don't optimize _Fract divisions, as 1
24346         can't be built for those types.
24348 2017-07-17  Georg-Johann Lay  <avr@gjlay.de>
24350         Remove stuff dead since r239246.
24352         * config/avr/avr-arch.h (avr_inform_devices): Remove dead proto.
24353         * config/avr/avr-devices.c (mcu_name, comparator, avr_mcus_str)
24354         (avr_inform_devices): Remove dead stuff.
24356 2017-07-17  Tamar Christina  <tamar.christina@arm.com>
24358         * config/arm/arm_neon.h: Fix softp typo.
24360 2017-07-17  Jakub Jelinek  <jakub@redhat.com>
24362         PR tree-optimization/81365
24363         * tree-ssa-phiprop.c (propagate_with_phi): When considering hoisting
24364         aggregate moves onto bb predecessor edges, make sure there are no
24365         loads that could alias the lhs in between the start of bb and the
24366         loads from *phi.
24368 2017-07-17  Georg-Johann Lay  <avr@gjlay.de>
24370         PR 80929
24371         * config/avr/avr.c (avr_mul_highpart_cost): New static function.
24372         (avr_rtx_costs_1) [TRUNCATE]: Use it to compute mul_highpart cost.
24373         [LSHIFTRT, outer_code = TRUNCATE]: Same.
24375 2017-07-17  Jakub Jelinek  <jakub@redhat.com>
24377         PR tree-optimization/81396
24378         * tree-ssa-math-opts.c (struct symbolic_number): Add n_ops field.
24379         (init_symbolic_number): Initialize it to 1.
24380         (perform_symbolic_merge): Add n_ops from both operands into the new
24381         n_ops.
24382         (find_bswap_or_nop): Don't consider n->n == cmpnop computations
24383         without base_addr as useless if they need more than one operation.
24384         (bswap_replace): Handle !bswap case for NULL base_addr.
24386 2017-07-17  Tom de Vries  <tom@codesourcery.com>
24388         PR target/81069
24389         * config/nvptx/nvptx.c (nvptx_single): Insert diverging branch as late
24390         as possible.
24392 2017-07-17  Sebastian Huber  <sebastian.huber@embedded-brains.de>
24394         * config/sparc/rtemself.h (TARGET_OS_CPP_BUILTINS): Add
24395         conditional builtin define __FIX_LEON3FT_B2BST.
24397 2017-07-17  Daniel Cederman  <cederman@gaisler.com>
24399         * config/sparc/t-rtems: Add mfix-gr712rc multilibs. Replace
24400         MULTILIB_EXCEPTIONS with MULTILIB_REQUIRED. Match -mfix-gr712rc
24401         with -mfix-ut700.
24403 2017-07-16  Eric Botcazou  <ebotcazou@adacore.com>
24405         PR rtl-optimization/81424
24406         * optabs.c (prepare_cmp_insn): Use copy_to_reg instead of force_reg
24407         to remove potential trapping from operands if -fnon-call-exceptions.
24409 2017-07-16  Jan Hubicka  <hubicka@ucw.cz>
24411         * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Use
24412         profile_proability for scalling.
24413         * scale_profile_for_vect_loop.c (scale_profile_for_vect_loop): Likewise.
24415 2017-07-16  Jan Hubicka  <hubicka@ucw.cz>
24417         * cgraph.c (cgraph_edge::redirect_call_stmt_to_caller): Cleanup.
24419 2017-07-16  Jan Hubicka  <hubicka@ucw.cz>
24421         * cfgloopmanip.c (scale_loop_profile): Avoid use of REG_BR_PROB_BASE
24422         fixpoint arithmetics.
24424 2017-07-16  Jan Hubicka  <hubicka@ucw.cz>
24426         * tree-ssa-loop-unswitch.c (hoist_guard): Avoid use of REG_BR_PROB_BASE
24427         fixpoint arithmetics.
24429 2017-07-16  Jan Hubicka  <hubicka@ucw.cz>
24431         * asan.c (create_cond_insert_point): Avoid use of REG_BR_PROB_BASE
24432         fixpoint arithmetics.
24434 2017-07-16  Jan Hubicka  <hubicka@ucw.cz>
24436         * profile-count.h (profile_probability::from_reg_br_prob_note,
24437         profile_probability::to_reg_br_prob_note): New functions.
24438         * doc/rtl.texi (REG_BR_PROB_NOTE): Update documentation.
24439         * reg-notes.h (REG_BR_PROB, REG_BR_PRED): Update docs.
24440         * predict.c (probability_reliable_p): Update.
24441         (edge_probability_reliable_p): Update.
24442         (br_prob_note_reliable_p): Update.
24443         (invert_br_probabilities): Update.
24444         (add_reg_br_prob_note): New function.
24445         (combine_predictions_for_insn): Update.
24446         * asan.c (asan_clear_shadow): Update.
24447         * cfgbuild.c (compute_outgoing_frequencies): Update.
24448         * cfgrtl.c (force_nonfallthru_and_redirect): Update.
24449         (update_br_prob_note): Update.
24450         (rtl_verify_edges): Update.
24451         (purge_dead_edges): Update.
24452         (fixup_reorder_chain): Update.
24453         * emit-rtl.c (try_split): Update.
24454         * ifcvt.c (cond_exec_process_insns): Update.
24455         (cond_exec_process_if_block): Update.
24456         (dead_or_predicable): Update.
24457         * internal-fn.c (expand_addsub_overflow): Update.
24458         (expand_neg_overflow): Update.
24459         (expand_mul_overflow): Update.
24460         * loop-doloop.c (doloop_modify): Update.
24461         * loop-unroll.c (compare_and_jump_seq): Update.
24462         * optabs.c (emit_cmp_and_jump_insn_1): Update.
24463         * predict.h: Update.
24464         * reorg.c (mostly_true_jump): Update.
24465         * rtl.h: Update.
24466         * config/aarch64/aarch64.c (aarch64_emit_unlikely_jump): Update.
24467         * config/alpha/alpha.c (emit_unlikely_jump): Update.
24468         * config/arc/arc.c: (emit_unlikely_jump): Update.
24469         * config/arm/arm.c: (emit_unlikely_jump): Update.
24470         * config/bfin/bfin.c (cbranch_predicted_taken_p): Update.
24471         * config/frv/frv.c (frv_print_operand_jump_hint): Update.
24472         * config/i386/i386.c (ix86_expand_split_stack_prologue): Update.
24473         (ix86_print_operand): Update.
24474         (ix86_split_fp_branch): Update.
24475         (predict_jump): Update.
24476         * config/ia64/ia64.c (ia64_print_operand): Update.
24477         * config/mmix/mmix.c (mmix_print_operand): Update.
24478         * config/powerpcspe/powerpcspe.c (output_cbranch): Update.
24479         (rs6000_expand_split_stack_prologue): Update.
24480         * config/rs6000/rs6000.c: Update.
24481         * config/s390/s390.c (s390_expand_vec_strlen): Update.
24482         (s390_expand_vec_movstr): Update.
24483         (s390_expand_cs_tdsi): Update.
24484         (s390_expand_split_stack_prologue): Update.
24485         * config/sh/sh.c (sh_print_operand): Update.
24486         (expand_cbranchsi4): Update.
24487         (expand_cbranchdi4): Update.
24488         * config/sparc/sparc.c (output_v9branch): Update.
24489         * config/spu/spu.c (get_branch_target): Update.
24490         (ea_load_store_inline): Update.
24491         * config/tilegx/tilegx.c (cbranch_predicted_p): Update.
24492         * config/tilepro/tilepro.c: Update.
24494 2017-07-16  Eric Botcazou  <ebotcazou@adacore.com>
24496         * gimplify.c (mostly_copy_tree_r): Revert latest change.
24497         (gimplify_save_expr): Likewise.
24499 2017-07-07  Jan Hubicka  <hubicka@ucw.cz>
24501         * ipa-visibility.c (function_and_variable_visibility): Fix pasto.
24503 2017-07-07  Jan Hubicka  <hubicka@ucw.cz>
24505         * ipa-fnsummary.c (pass_data_ipa_fn_summary): Use
24506         TV_IPA_FNSUMMARY.
24507         * timevar.def (TV_IPA_FNSUMMARY): Define.
24509 2017-07-16  Daniel Cederman  <cederman@gaisler.com>
24511         * config/sparc/sparc.md (divdf3_fix): Add NOP to prevent back
24512         to back store errata sensitive sequence from being generated.
24513         (sqrtdf2_fix): Likewise.
24515 2017-07-07  Jan Hubicka  <hubicka@ucw.cz>
24517         * tree-ssa-threadupdate.c (compute_path_counts,
24518         update_joiner_offpath_counts): Use profile_probability.
24520 2017-07-15  Thomas Preud'homme  <thomas.preudhomme@arm.com>
24522         Revert:
24523         2017-07-14  Thomas Preud'homme  <thomas.preudhomme@arm.com>
24525         * config/arm/arm-c.c (arm_cpu_builtins): Define
24526         __ARM_FEATURE_NUMERIC_MAXMIN solely based on TARGET_VFP5.
24528 2017-07-14  Kelvin Nilsen  <kelvin@gcc.gnu.org>
24530         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
24531         array entries to represent __ieee128 versions of the
24532         scalar_test_data_class, scalar_test_neg, scalar_extract_exp,
24533         scalar_extract_sig, and scalar_insert_exp built-in functions.
24534         (altivec_resolve_overloaded_builtin): Add special case handling
24535         for the __builtin_scalar_insert_exp function, as represented by
24536         the P9V_BUILTIN_VEC_VSIEDP constant.
24537         * config/rs6000/rs6000-builtin.def (VSEEQP): Add scalar extract
24538         exponent support for __ieee128 argument.
24539         (VSESQP): Add scalar extract signature support for __ieee128
24540         argument.
24541         (VSTDCNQP): Add scalar test negative support for __ieee128
24542         argument.
24543         (VSIEQP): Add scalar insert exponent support for __int128 argument
24544         with __ieee128 result.
24545         (VSIEQPF): Add scalar insert exponent support for __ieee128
24546         argument with __ieee128 result.
24547         (VSTDCQP): Add scalar test data class support for __ieee128
24548         argument.
24549         (VSTDCNQP): Add overload support for scalar test negative with
24550         __ieee128 argument.
24551         (VSTDCQP): Add overload support for scalar test data class
24552         __ieee128 argument.
24553         * config/rs6000/vsx.md (UNSPEC_VSX_SXSIG) Replace
24554         UNSPEC_VSX_SXSIGDP.
24555         (UNSPEC_VSX_SIEXPQP): New constant.
24556         (xsxexpqp): New insn for VSX scalar extract exponent quad
24557         precision.
24558         (xsxsigqp): New insn for VSX scalar extract significand quad
24559         precision.
24560         (xsiexpqpf): New insn for VSX scalar insert exponent quad
24561         precision with floating point argument.
24562         (xststdcqp): New expand for VSX scalar test data class quad
24563         precision.
24564         (xststdcnegqp): New expand for VSX scalar test negative quad
24565         precision.
24566         (xststdcqp): New insn to match expansions for VSX scalar test data
24567         class quad precision and VSX scalar test negative quad precision.
24568         * config/rs6000/rs6000.c (rs6000_expand_binop_builtin): Add
24569         special case operand checking to enforce that second operand of
24570         VSX scalar test data class with quad precision argument is a 7-bit
24571         unsigned literal.
24572         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Add
24573         prototypes and descriptions of __ieee128 versions of
24574         scalar_extract_exp, scalar_extract_sig, scalar_insert_exp,
24575         scalar_test_data_class, and scalar_test_neg built-in functions.
24577 2016-07-14  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
24579         PR tree-optimization/81162
24580         * gimple-ssa-strength-reduction.c (replace_mult_candidate): Don't
24581         replace a negate with an add.
24583 2017-07-14  James Greenhalgh  <james.greenhalgh@arm.com>
24585         * doc/invoke.texi (arm/-mcpu): Document +crypto.
24587 2017-07-14  Thomas Preud'homme  <thomas.preudhomme@arm.com>
24589         * config/arm/arm-c.c (arm_cpu_builtins): Define
24590         __ARM_FEATURE_NUMERIC_MAXMIN solely based on TARGET_VFP5.
24592 2017-07-14  Thomas Preud'homme  <thomas.preudhomme@arm.com>
24594         * config/arm/arm-cpus.in (cortex-r52): Add new entry.
24595         (armv8-r): Set ARM Cortex-R52 as default CPU.
24596         * config/arm/arm-tables.opt: Regenerate.
24597         * config/arm/arm-tune.md: Regenerate.
24598         * config/arm/driver-arm.c (arm_cpu_table): Add entry for ARM
24599         Cortex-R52.
24600         * doc/invoke.texi: Mention -mtune=cortex-r52 and availability of fp.dp
24601         extension for -mcpu=cortex-r52.
24603 2017-07-14  Thomas Preud'homme  <thomas.preudhomme@arm.com>
24605         * config/arm/arm-isa.h (isa_bit_FP_ARMv8): Delete enumerator.
24606         (ISA_FP_ARMv8): Define as ISA_FPv5 and ISA_FP_D32.
24607         * config/arm/arm-cpus.in (armv8-r): Define fp.sp as enabling FPv5.
24608         (fp-armv8): Define it as FP_ARMv8 only.
24609         config/arm/arm.h (TARGET_FPU_ARMV8): Delete.
24610         (TARGET_VFP_FP16INST): Define using TARGET_VFP5 rather than
24611         TARGET_FPU_ARMV8.
24612         config/arm/arm.c (arm_rtx_costs_internal): Replace checks against
24613         TARGET_FPU_ARMV8 by checks against TARGET_VFP5.
24614         * config/arm/arm-builtins.c (arm_builtin_vectorized_function): Define
24615         first ARM_CHECK_BUILTIN_MODE definition using TARGET_VFP5 rather
24616         than TARGET_FPU_ARMV8.
24617         * config/arm/arm-c.c (arm_cpu_builtins): Likewise for
24618         __ARM_FEATURE_NUMERIC_MAXMIN macro definition.
24619         * config/arm/arm.md (cmov<mode>): Condition on TARGET_VFP5 rather than
24620         TARGET_FPU_ARMV8.
24621         * config/arm/neon.md (neon_vrint): Likewise.
24622         (neon_vcvt): Likewise.
24623         (neon_<fmaxmin_op><mode>): Likewise.
24624         (<fmaxmin><mode>3): Likewise.
24625         * config/arm/vfp.md (l<vrint_pattern><su_optab><mode>si2): Likewise.
24626         * config/arm/predicates.md (arm_cond_move_operator): Check against
24627         TARGET_VFP5 rather than TARGET_FPU_ARMV8 and fix spacing.
24629 2017-07-14  Jackson Woodruff  <jackson.woodruff@arm.com>
24631         * config/aarch64/aarch64.c (aarch64_print_operand): Move comments
24632         to top of function.
24634 2017-07-14  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
24636         * gimple-ssa-store-merging.c (clear_bit_region): Replace reference to
24637         loop in comment with memset.
24639 2017-07-14  Martin Liska  <mliska@suse.cz>
24641         * cfgexpand.c (expand_gimple_basic_block): Remove dead comment.
24642         * dwarf2out.c (is_java): Remove the function.
24643         (output_pubname): Remove usage of the function.
24644         (lower_bound_default): Remove usage of DW_LANG_Java.
24645         (gen_compile_unit_die): Likewise.
24646         * gcc.c: Remove compiler defaults for .java and .zip files.
24647         * gimple-expr.c (remove_suffix): Change as there's no longer
24648         extension than 4-letter one.
24649         * gimplify.c (mostly_copy_tree_r): Remove Java-special part.
24650         (gimplify_save_expr): Likewise.
24651         * ipa-utils.h (polymorphic_type_binfo_p): Remove the comment
24652         as it's possible even for other languages than Java.
24653         * langhooks.h (struct lang_hooks): Remove Java from a comment.
24654         * lto-opts.c (lto_write_options): Remove reference to Java.
24655         * opts.c (strip_off_ending): Update file extension handling.
24656         * tree-cfg.c (verify_gimple_call): Remove comment with Java.
24657         * tree-eh.c (lower_resx): Likewise.
24658         * tree.c (free_lang_data_in_type): Remove dead code.
24659         (find_decls_types_r): Likewise.
24660         (build_common_builtin_nodes): Remove Java from a comment.
24661         (verify_type): Remove dead code.
24662         * varasm.c (assemble_external): Remove Java from a comment.
24664 2017-07-14  Martin Liska  <mliska@suse.cz>
24666         * opts.c (finish_options): Add quotes.
24667         (common_handle_option): Likewise.
24669 2017-07-14  Martin Liska  <mliska@suse.cz>
24671         * dbxout.c (get_lang_number): Do not handle GNU Pascal.
24672         * dbxout.h (extern void dbxout_stab_value_internal_label_diff):
24673         Remove N_SO_PASCAL.
24674         * dwarf2out.c (lower_bound_default): Do not handle
24675         DW_LANG_Pascal83.
24676         (gen_compile_unit_die): Likewise.
24677         * gcc.c: Remove default extension binding for GNU Pascal.
24678         * stmt.c: Remove Pascal language from a comment.
24679         * xcoffout.c: Likewise.
24681 2017-07-13  David Malcolm  <dmalcolm@redhat.com>
24683         PR c/81405
24684         * diagnostic-show-locus.c (fixit_cmp): New function.
24685         (layout::layout): Sort m_fixit_hints.
24686         (column_range::column_range): Assert that the values are valid.
24687         (struct char_span): New struct.
24688         (correction::overwrite): New method.
24689         (struct source_line): New struct.
24690         (line_corrections::add_hint): Add assertions.  Reimplement memcpy
24691         calls in terms of classes source_line and char_span, and
24692         correction::overwrite.
24693         (selftest::test_overlapped_fixit_printing_2): New function.
24694         (selftest::diagnostic_show_locus_c_tests): Call it.
24696 2017-07-13  Will Schmidt  <will_schmidt@vnet.ibm.com>
24698         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Return
24699         early if there is no lhs.
24701 2017-07-13  Martin Liska  <mliska@suse.cz>
24703         * dwarf2out.c (gen_pointer_type_die): Remove dead code.
24704         (gen_reference_type_die): Likewise.
24705         * stor-layout.c: Remove Pascal-related comment.
24707 2017-07-13  Martin Liska  <mliska@suse.cz>
24709         * opts.c (finish_options): Add quotes to error messages.
24710         (parse_sanitizer_options): Likewise.
24712 2017-07-13  Thomas Preud'homme  <thomas.preudhomme@arm.com>
24714         * doc/invoke.texi (armv8-r): Document +fp.sp ARMv8-R extension.
24716 2017-07-13  Richard Earnshaw  <rearnsha@arm.com>
24718         * config/arm/vxworks.h (TARGET_ENDIAN_DEFAULT): Define.
24720 2017-07-13  Maxim Ostapenko  <m.ostapenko@samsung.com>
24722         * asan.c (asan_emit_allocas_unpoison): Use ptr_mode for arguments
24723         during expansion.
24724         * builtins.c (expand_asan_emit_allocas_unpoison): Likewise.
24726 2017-07-12  Michael Meissner  <meissner@linux.vnet.ibm.com>
24728         PR target/81193
24729         * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If GLIBC
24730         provides the hardware capability bits, define the macro
24731         __BUILTIN_CPU_SUPPORTS__.
24732         * config/rs6000/rs6000.c (cpu_expand_builtin): Generate a warning
24733         if GLIBC does not provide the hardware capability bits.  Add a
24734         gcc_unreachable call if the built-in cpu function is neither
24735         __builtin_cpu_is nor __builtin_cpu_supports.
24736         (rs6000_get_function_versions_dispatcher): Change the warning
24737         that an old GLIBC is used which does not export the capability
24738         bits to be an error.
24739         * doc/extend.texi (target_clones attribute): Document the
24740         restriction that GLIBC 2.23 or newer is needed on the PowerPC.
24741         (PowerPC built-in functions): Document that GLIBC 2.23 or newer is
24742         needed by __builtin_cpu_is and __builtin_cpu_supports.  Document
24743         the macros defined by GCC if the newer GLIBC is available.
24745 2017-07-12  Jeff Law  <law@redhat.com>
24747         * config/riscv/riscv.c: Remove unnecessary includes.  Reorder
24748         remaining includes slightly.
24749         * config/riscv/riscv-builtins.c: Include profile-count.h.
24751 2017-07-12  Georg-Johann Lay  <avr@gjlay.de>
24753         PR target/79883
24754         * config/avr/avr.c (avr_set_current_function): In diagnostic
24755         messages: Quote keywords and (parts of) identifiers.
24756         [WITH_AVRLIBC]: Warn for functions named "ISR", "SIGNAL" or
24757         "INTERUPT".
24759 2017-07-12  Carl Love  <cel@us.ibm.com>
24761         * config/rs6000/rs6000-c.c: Add support for built-in functions
24762         vector bool char vec_revb (vector bool char);
24763         vector bool short vec_revb (vector short char);
24764         vector bool int vec_revb (vector bool int);
24765         vector bool long long vec_revb (vector bool long long);
24766         * doc/extend.texi: Update the built-in documentation file for the
24767         new built-in functions.
24769 2017-07-12  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
24771         * config/s390/s390.md: Remove movcc splitter.
24773 2017-07-12  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
24775         * config/s390/s390.c (s390_rtx_costs): Return proper costs for
24776         load/store on condition.
24778 2017-07-12  Georg-Johann Lay  <avr@gjlay.de>
24780         PR target/81407
24781         * config/avr/avr.c (avr_encode_section_info)
24782         [progmem && !TREE_READONLY]: Error if progmem object needs
24783         constructing.
24785 2017-07-11  Michael Collison  <michael.collison@arm.com>
24787         * config/aarch64/aarch64-simd.md (aarch64_sub<mode>_compare0):
24788         New pattern.
24790 2017-07-11  Carl Love  <cel@us.ibm.com>
24792         * config/rs6000/rs6000-c.c: Add support for builtins
24793         vector unsigned int vec_parity_lsbb (vector signed int);
24794         vector unsigned int vec_parity_lsbb (vector unsigned int);
24795         vector unsigned __int128 vec_parity_lsbb (vector signed __int128);
24796         vector unsigned __int128 vec_parity_lsbb (vector unsigned __int128);
24797         vector unsigned long long vec_parity_lsbb (vector signed long long);
24798         vector unsigned long long vec_parity_lsbb (vector unsigned long long);
24799         * config/rs6000/rs6000-builtin.def (VPARITY_LSBB): Add BU_P9V_OVERLOAD1.
24800         * config/rs6000/altivec.h (vec_parity_lsbb): Add define.
24801         * doc/extend.texi: Update the built-in documentation file for the
24802         new built-in functions.
24804 2017-07-11  David Malcolm  <dmalcolm@redhat.com>
24806         * diagnostic-show-locus.c: Include "gcc-rich-location.h".
24807         (layout::m_primary_loc): New field.
24808         (layout::layout): Initialize new field.  Move location filtering
24809         logic from here to...
24810         (layout::maybe_add_location_range): ...this new method.  Add
24811         support for filtering to just the lines already specified by other
24812         locations.
24813         (layout::will_show_line_p): New method.
24814         (gcc_rich_location::add_location_if_nearby): New method.
24815         (selftest::test_add_location_if_nearby): New test function.
24816         (selftest::diagnostic_show_locus_c_tests): Call it.
24817         * gcc-rich-location.h (gcc_rich_location::add_location_if_nearby):
24818         New method.
24820 2017-07-11  Tom de Vries  <tom@codesourcery.com>
24822         * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG): New macro.
24823         (bb_first_real_insn): New function.
24824         (nvptx_single): Add extra initialization of broadcasted condition
24825         variables.
24827 2017-07-11  Nathan Sidwell  <nathan@acm.org>
24829         * dwarf2out.c (gen_member_die): Remove useless check for anon ctors.
24831 2017-07-11  Georg-Johann Lay  <avr@gjlay.de>
24833         * doc/extend.texi (AVR Function Attributes): Remove weblink to
24834         Binutils doc as TEXI will mess them up.
24835         * doc/invoke.texi (AVR Options): Same here.
24837 2017-07-11  Daniel Cederman  <cederman@gaisler.com>
24839         * config/sparc/sparc.opt (mfix-ut700): New option.
24840         (mfix-gr712rc): Likewise.
24841         (sparc_fix_b2bst): New variable.
24842         * doc/invoke.texi (SPARC options): Document them.
24843         (ARM options): Fix warnings.
24844         * config/sparc/sparc.c (sparc_do_work_around_errata): Insert NOP
24845         instructions to prevent sequences that can trigger the store-store
24846         errata for certain LEON3FT processors.
24847         (pass_work_around_errata::gate): Also test sparc_fix_b2bst.
24848         (sparc_option_override): Set sparc_fix_b2bst appropriately.
24849         * config/sparc/sparc.md (fix_b2bst): New attribute.
24850         (in_branch_delay): Prevent stores in delay slot if fix_b2bst.
24852 2017-07-10  Uros Bizjak  <ubizjak@gmail.com>
24854         PR target/81375
24855         * config/i386/i386.md (divsf3): Add TARGET_SSE to TARGET_SSE_MATH.
24856         (rcpps): Ditto.
24857         (*rsqrtsf2_sse): Ditto.
24858         (rsqrtsf2): Ditto.
24859         (div<mode>3): Macroize insn from divdf3 and divsf3
24860         using MODEF mode iterator.
24862 2017-07-10  Martin Sebor  <msebor@redhat.com>
24864         PR tree-optimization/80397
24865         * gimple-ssa-sprintf.c (format_integer): Use INTEGRAL_TYPE_P()
24866         instead of testing for equality to INTEGER_TYPE.
24868 2017-07-10  Vineet Gupta <vgupta@synopsys.com>
24870         * config.gcc: Remove uclibc from arc target spec.
24872 2017-07-10  Claudiu Zissulescu  <claziss@synopsys.com>
24874         * config/arc/arc.h (ADDITIONAL_REGISTER_NAMES): Define.
24876 2017-07-07  Jan Hubicka  <hubicka@ucw.cz>
24878         PR lto/80838
24879         * lto-wrapper.c (remove_option): New function.
24880         (merge_and_complain): Merge PIC/PIE options more realistically.
24882 2017-07-10  Georg-Johann Lay  <avr@gjlay.de>
24884         Better ISR prologues by supporting GASes __gcc_isr pseudo insn.
24886         PR target/20296
24887         PR target/81268
24888         * configure.ac [target=avr]: Add GAS check for -mgcc-isr.
24889         (HAVE_AS_AVR_MGCCISR_OPTION):  If so, AC_DEFINE it.
24890         * config.in: Regenerate.
24891         * configure: Regenerate.
24892         * doc/extend.texi (AVR Function Attributes) <no_gccisr>: Document it.
24893         * doc/invoke.texi (AVR Options) <-mgas-isr-prologues>: Document it.
24894         * config/avr/avr.opt (-mgas-isr-prologues): New option and...
24895         (TARGET_GASISR_PROLOGUES): ...target mask.
24896         * common/config/avr/avr-common.c
24897         (avr_option_optimization_table) [OPT_LEVELS_1_PLUS_NOT_DEBUG]:
24898         Set -mgas-isr-prologues.
24899         * config/avr/avr-passes.def (avr_pass_pre_proep): Add
24900         INSERT_PASS_BEFORE for it.
24901         * config/avr/avr-protos.h (make_avr_pass_pre_proep): New proto.
24902         * config/avr/avr.c (avr_option_override)
24903         [!HAVE_AS_AVR_MGCCISR_OPTION]: Unset TARGET_GASISR_PROLOGUES.
24904         (avr_no_gccisr_function_p, avr_hregs_split_reg): New static functions.
24905         (avr_attribute_table) <no_gccisr>: Add new function attribute.
24906         (avr_set_current_function) <is_no_gccisr>: Init machine field.
24907         (avr_pass_data_pre_proep, avr_pass_pre_proep): New pass data
24908         and rtl_opt_pass.
24909         (make_avr_pass_pre_proep): New function.
24910         (emit_push_sfr) <treg>: Add argument to function and use it
24911         instead of TMP_REG.
24912         (avr_expand_prologue) [machine->gasisr.maybe]: Emit gasisr insn
24913         and set machine->gasisr.yes.
24914         (avr_expand_epilogue) [machine->gasisr.yes]: Similar.
24915         (avr_asm_function_end_prologue) [machine->gasisr.yes]: Add
24916         __gcc_isr.n_pushed to .L__stack_usage.
24917         (TARGET_ASM_FINAL_POSTSCAN_INSN): Define to...
24918         (avr_asm_final_postscan_insn): ...this new static function.
24919         * config/avr/avr.h (machine_function)
24920         <is_no_gccisr, use_L__stack_usage>: New fields.
24921         <gasisr, gasisr.yes, gasisr.maybe, gasisr.regno>: New fields.
24922         * config/avr/avr.md (UNSPECV_GASISR): Add unspecv enum.
24923         (GASISR_Prologue, GASISR_Epilogue, GASISR_Done): New define_constants.
24924         (gasisr, *gasisr): New expander and insn.
24925         * config/avr/gen-avr-mmcu-specs.c (print_mcu)
24926         [HAVE_AS_AVR_MGCCISR_OPTION]: Print asm_gccisr spec.
24927         * config/avr/specs.h (ASM_SPEC) <asm_gccisr>: Add sub spec.
24929 2017-07-10  Richard Earnshaw  <rearnsha@arm.com>
24931         * config/arm/parsecpu.awk (gen_comm_data): Do not escape single quotes
24932         in quoted strings.
24934 2017-07-10  Georg-Johann Lay  <avr@gjlay.de>
24936         Move jump-tables out of .text again.
24938         PR target/81075
24939         * config/avr/avr.c (ASM_OUTPUT_ADDR_VEC_ELT): Remove function.
24940         (ASM_OUTPUT_ADDR_VEC): New function.
24941         (avr_adjust_insn_length) [JUMP_TABLE_DATA_P]: Return 0.
24942         (avr_final_prescan_insn) [avr_log.insn_addresses]: Dump
24943         INSN_ADDRESSes as asm comment.
24944         * config/avr/avr.h (JUMP_TABLES_IN_TEXT_SECTION): Adjust comment.
24945         (ASM_OUTPUT_ADDR_VEC_ELT): Remove define.
24946         (ASM_OUTPUT_ADDR_VEC): Define to avr_output_addr_vec.
24947         * config/avr/avr.md (*tablejump): Adjust comment.
24948         * config/avr/elf.h (ASM_OUTPUT_BEFORE_CASE_LABEL): Remove.
24949         * config/avr/avr-log.c (avr_log_set_avr_log) <insn_addresses>:
24950         New detail.
24951         * config/avr/avr-protos.h (avr_output_addr_vec_elt): Remove proto.
24952         (avr_output_addr_vec): New proto.
24953         (avr_log_t) <insn_addresses>: New field.
24955 2017-07-09  H.J. Lu  <hongjiu.lu@intel.com>
24957         PR target/81313
24958         * config/i386/i386.c (ix86_function_arg_advance): Set
24959         outgoing_args_on_stack to true if there are outgoing arguments
24960         on stack.
24961         (ix86_function_arg): Likewise.
24962         (ix86_get_drap_rtx): Use DRAP only if there are outgoing
24963         arguments on stack and ACCUMULATE_OUTGOING_ARGS is false.
24964         * config/i386/i386.h (machine_function): Add
24965         outgoing_args_on_stack.
24967 2017-07-09  Krister Walfridsson  <krister.walfridsson@gmail.com>
24969         * config.gcc (*-*-netbsd*): Remove check for NetBSD versions not
24970         supporting pthreds.
24971         * config/netbsd.h (NETBSD_LIBGCC_SPEC): Always enable pthreads.
24973 2017-07-08  Richard Sandiford  <richard.sandiford@linaro.org>
24975         * Makefile.in (HOOKS_H, RTL_BASE_H, FUNCTION_H, EXPR_H, REGS_H)
24976         (REAL_H): Remove $(MACHMODE_H).
24977         (FIXED_VALUE_H, TREE_CORE_H, CFGLOOP_H): Remove $(MACHMODE_H) and
24978         double-int.h.
24979         (CORETYPES_H): Add signop.h, wide-int.h, wide-int-print.h,
24980         $(MACHMODE_H) and double-int.h.
24981         (build/min-insn-modes.o): Depend on $(CORETYPES_H) rather than
24982         $(MACHMODE_H).
24983         (gengtype-state.o, gengtype.o, build/gengtype.o): Don't depend on
24984         double-int.h.
24986 2017-07-07  Andrew Pinski  <apinski@cavium.com>
24988         * config/aarch64/aarch64.c (aarch_macro_fusion_pair_p): Check
24989         prev_set and curr_set for AARCH64_FUSE_ALU_BRANCH.
24991 2017-07-07  Michael Meissner  <meissner@linux.vnet.ibm.com>
24993         * config/rs6000/rs6000.c (rs6000_get_function_versions_dispatcher):
24994         Add warning if GCC was not configured to link against a GLIBC that
24995         exports the hardware capability bits.
24996         (make_resolver_func): Make resolver function private and not a
24997         COMDAT function.  Create the name with clone_function_name instead
24998         of make_unique_name.
25000         PR target/81348
25001         * config/rs6000/rs6000.md (HI sign_extend splitter): Use the
25002         correct operand in doing the split.
25004 2017-07-07 Carl Love  <cel@us.ibm.com>
25006         * config/rs6000/rs6000-c: Add support for built-in function
25007         vector unsigned short vec_pack_to_short_fp32 (vector float,
25008                                                       vector float).
25009         * config/rs6000/rs6000-builtin.def (CONVERT_4F32_8I16): Add
25010         BU_P9V_AV_2 and BU_P9V_OVERLOAD_2 definitions.
25011         * config/rs6000/altivec.h (vec_pack_to_short_fp32): Add define.
25012         * config/rs6000/altivec.md(UNSPEC_CONVERT_4F32_8I16): Add UNSPEC.
25013         (convert_4f32_8i16): Add define_expand.
25014         * doc/extend.texi: Update the built-in documentation file for the
25015         new built-in function.
25017 2017-07-07  Jose E. Marchesi  <jose.marchesi@oracle.com>
25019         * config/sparc/m8.md: New file.
25020         * config/sparc/sparc.md: Include m8.md.
25022 2017-07-07  Jose E. Marchesi  <jose.marchesi@oracle.com>
25024         * config/sparc/sparc.opt: New option -mvis4b.
25025         * config/sparc/sparc.c (dump_target_flag_bits): Handle MASK_VIS4B.
25026         (sparc_option_override): Handle VIS4B.
25027         (enum sparc_builtins): Define
25028         SPARC_BUILTIN_DICTUNPACK{8,16,32},
25029         SPARC_BUILTIN_FPCMP{LE,GT,EQ,NE}{8,16,32}SHL,
25030         SPARC_BUILTIN_FPCMPU{LE,GT}{8,16,32}SHL,
25031         SPARC_BUILTIN_FPCMPDE{8,16,32}SHL and
25032         SPARC_BUILTIN_FPCMPUR{8,16,32}SHL.
25033         (check_constant_argument): New function.
25034         (sparc_vis_init_builtins): Define builtins
25035         __builtin_vis_dictunpack{8,16,32},
25036         __builtin_vis_fpcmp{le,gt,eq,ne}{8,16,32}shl,
25037         __builtin_vis_fpcmpu{le,gt}{8,16,32}shl,
25038         __builtin_vis_fpcmpde{8,16,32}shl and
25039         __builtin_vis_fpcmpur{8,16,32}shl.
25040         (sparc_expand_builtin): Check that the constant operands to
25041         __builtin_vis_fpcmp*shl and _builtin_vis_dictunpack* are indeed
25042         constant and in range.
25043         * config/sparc/sparc-c.c (sparc_target_macros): Handle
25044         TARGET_VIS4B.
25045         * config/sparc/sparc.h (SPARC_IMM2_P): Define.
25046         (SPARC_IMM5_P): Likewise.
25047         * config/sparc/sparc.md (cpu_feature): Add new feagure "vis4b".
25048         (enabled): Handle vis4b.
25049         (UNSPEC_DICTUNPACK): New unspec.
25050         (UNSPEC_FPCMPSHL): Likewise.
25051         (UNSPEC_FPUCMPSHL): Likewise.
25052         (UNSPEC_FPCMPDESHL): Likewise.
25053         (UNSPEC_FPCMPURSHL): Likewise.
25054         (cpu_feature): New CPU feature `vis4b'.
25055         (dictunpack{8,16,32}): New insns.
25056         (FPCSMODE): New mode iterator.
25057         (fpcscond): New code iterator.
25058         (fpcsucond): Likewise.
25059         (fpcmp{le,gt,eq,ne}{8,16,32}{si,di}shl): New insns.
25060         (fpcmpu{le,gt}{8,16,32}{si,di}shl): Likewise.
25061         (fpcmpde{8,16,32}{si,di}shl): Likewise.
25062         (fpcmpur{8,16,32}{si,di}shl): Likewise.
25063         * config/sparc/constraints.md: Define constraints `q' for unsigned
25064         2-bit integer constants and `t' for unsigned 5-bit integer
25065         constants.
25066         * config/sparc/predicates.md (imm5_operand_dictunpack8): New
25067         predicate.
25068         (imm5_operand_dictunpack16): Likewise.
25069         (imm5_operand_dictunpack32): Likewise.
25070         (imm2_operand): Likewise.
25071         * doc/invoke.texi (SPARC Options): Document -mvis4b.
25072         * doc/extend.texi (SPARC VIS Built-in Functions): Document the
25073         ditunpack* and fpcmp*shl builtins.
25075 2017-07-07  Jose E. Marchesi  <jose.marchesi@oracle.com>
25077         * config.gcc: Handle m8 in --with-{cpu,tune} options.
25078         * config.in: Add HAVE_AS_SPARC6 define.
25079         * config/sparc/driver-sparc.c (cpu_names): Add entry for the SPARC
25080         M8.
25081         * config/sparc/sol2.h (CPP_CPU64_DEFAULT_SPEC): Define for
25082         TARGET_CPU_m8.
25083         (ASM_CPU32_DEFAUILT_SPEC): Likewise.
25084         (CPP_CPU_SPEC): Handle m8.
25085         (ASM_CPU_SPEC): Likewise.
25086         * config/sparc/sparc-opts.h (enum processor_type): Add
25087         PROCESSOR_M8.
25088         * config/sparc/sparc.c (m8_costs): New struct.
25089         (sparc_option_override): Handle TARGET_CPU_m8.
25090         (sparc32_initialize_trampoline): Likewise.
25091         (sparc64_initialize_trampoline): Likewise.
25092         (sparc_issue_rate): Likewise.
25093         (sparc_register_move_cost): Likewise.
25094         * config/sparc/sparc.h (TARGET_CPU_m8): Define.
25095         (CPP_CPU64_DEFAULT_SPEC): Define for M8.
25096         (ASM_CPU64_DEFAULT_SPEC): Likewise.
25097         (CPP_CPU_SPEC): Handle M8.
25098         (ASM_CPU_SPEC): Likewise.
25099         (AS_M8_FLAG): Define.
25100         * config/sparc/sparc.md: Add m8 to the cpu attribute.
25101         * config/sparc/sparc.opt: New option -mcpu=m8 for sparc targets.
25102         * configure.ac (HAVE_AS_SPARC6): Check for assembler support for
25103         M8 instructions.
25104         * configure: Regenerate.
25105         * doc/invoke.texi (SPARC Options): Document -mcpu=m8 and
25106         -mtune=m8.
25108 2017-07-07  Jose E. Marchesi  <jose.marchesi@oracle.com>
25110         * config/sparc/niagara7.md: Rework the DFA scheduler to use insn
25111         subtypes.
25112         * config/sparc/sparc.md: Remove the `v3pipe' insn attribute.
25113         ("*movdi_insn_sp32"): Do not set v3pipe.
25114         ("*movsi_insn"): Likewise.
25115         ("*movdi_insn_sp64"): Likewise.
25116         ("*movsf_insn"): Likewise.
25117         ("*movdf_insn_sp32"): Likewise.
25118         ("*movdf_insn_sp64"): Likewise.
25119         ("*zero_extendsidi2_insn_sp64"): Likewise.
25120         ("*sign_extendsidi2_insn"): Likewise.
25121         ("*mov<VM32:mode>_insn"): Likewise.
25122         ("*mov<VM64:mode>_insn_sp64"): Likewise.
25123         ("*mov<VM64:mode>_insn_sp32"): Likewise.
25124         ("<plusminus_insn><VADDSUB:mode>3"): Likewise.
25125         ("<vlop:code><VL:mode>3"): Likewise.
25126         ("*not_<vlop:code><VL:mode>3"): Likewise.
25127         ("*nand<VL:mode>_vis"): Likewise.
25128         ("*<vlnotop:code>_not1<VL:mode>_vis"): Likewise.
25129         ("*<vlnotop:code>_not2<VL:mode>_vis"): Likewise.
25130         ("one_cmpl<VL:mode>2"): Likewise.
25131         ("faligndata<VM64:mode>_vis"): Likewise.
25132         ("alignaddrsi_vis"): Likewise.
25133         ("alignaddrdi_vis"): Likweise.
25134         ("alignaddrlsi_vis"): Likewise.
25135         ("alignaddrldi_vis"): Likewise.
25136         ("fcmp<gcond:code><GCM:gcm_name><P:mode>_vis"): Likewise.
25137         ("bmaskdi_vis"): Likewise.
25138         ("bmasksi_vis"): Likewise.
25139         ("bshuffle<VM64:mode>_vis"): Likewise.
25140         ("cmask8<P:mode>_vis"): Likewise.
25141         ("cmask16<P:mode>_vis"): Likewise.
25142         ("cmask32<P:mode>_vis"): Likewise.
25143         ("pdistn<P:mode>_vis"): Likewise.
25144         ("<vis3_addsub_ss_patname><VASS:mode>3"): Likewise.
25146 2017-07-07  Jose E. Marchesi  <jose.marchesi@oracle.com>
25148         * config/sparc/sparc.md ("subtype"): New insn attribute.
25149         ("*wrgsr_sp64"): Set insn subtype.
25150         ("*rdgsr_sp64"): Likewise.
25151         ("alignaddrsi_vis"): Likewise.
25152         ("alignaddrdi_vis"): Likewise.
25153         ("alignaddrlsi_vis"): Likewise.
25154         ("alignaddrldi_vis"): Likewise.
25155         ("<plusminus_insn><VADDSUB:mode>3"): Likewise.
25156         ("fexpand_vis"): Likewise.
25157         ("fpmerge_vis"): Likewise.
25158         ("faligndata<VM64:mode>_vis"): Likewise.
25159         ("bshuffle<VM64:mode>_vis"): Likewise.
25160         ("cmask8<P:mode>_vis"): Likewise.
25161         ("cmask16<P:mode>_vis"): Likewise.
25162         ("cmask32<P:mode>_vis"): Likewise.
25163         ("fchksm16_vis"): Likewise.
25164         ("v<vis3_shift_patname><GCM:mode>3"): Likewise.
25165         ("fmean16_vis"): Likewise.
25166         ("fp<plusminus_insn>64_vis"): Likewise.
25167         ("<plusminus_insn>v8qi3"): Likewise.
25168         ("<vis3_addsub_ss_patname><VASS:mode>3"): Likewise.
25169         ("<vis4_minmax_patname><VMMAX:mode>3"): Likewise.
25170         ("<vis4_uminmax_patname><VMMAX:mode>3"): Likewise.
25171         ("<vis3_addsub_ss_patname>v8qi3"): Likewise.
25172         ("<vis4_addsub_us_patname><VAUS:mode>3"): Likewise.
25173         ("*movqi_insn"): Likewise.
25174         ("*movhi_insn"): Likewise.
25175         ("*movsi_insn"): Likewise.
25176         ("movsi_pic_gotdata_op"): Likewise.
25177         ("*movdi_insn_sp32"): Likewise.
25178         ("*movdi_insn_sp64"): Likewise.
25179         ("movdi_pic_gotdata_op"): Likewise.
25180         ("*movsf_insn"): Likewise.
25181         ("*movdf_insn_sp32"): Likewise.
25182         ("*movdf_insn_sp64"): Likewise.
25183         ("*zero_extendhisi2_insn"): Likewise.
25184         ("*zero_extendqihi2_insn"): Likewise.
25185         ("*zero_extendqisi2_insn"): Likewise.
25186         ("*zero_extendqidi2_insn"): Likewise.
25187         ("*zero_extendhidi2_insn"): Likewise.
25188         ("*zero_extendsidi2_insn_sp64"): Likewise.
25189         ("ldfsr"): Likewise.
25190         ("prefetch_64"): Likewise.
25191         ("prefetch_32"): Likewise.
25192         ("tie_ld32"): Likewise.
25193         ("tie_ld64"): Likewise.
25194         ("*tldo_ldub_sp32"): Likewise.
25195         ("*tldo_ldub1_sp32"): Likewise.
25196         ("*tldo_ldub2_sp32"): Likewise.
25197         ("*tldo_ldub_sp64"): Likewise.
25198         ("*tldo_ldub1_sp64"): Likewise.
25199         ("*tldo_ldub2_sp64"): Likewise.
25200         ("*tldo_ldub3_sp64"): Likewise.
25201         ("*tldo_lduh_sp32"): Likewise.
25202         ("*tldo_lduh1_sp32"): Likewise.
25203         ("*tldo_lduh_sp64"): Likewise.
25204         ("*tldo_lduh1_sp64"): Likewise.
25205         ("*tldo_lduh2_sp64"): Likewise.
25206         ("*tldo_lduw_sp32"): Likewise.
25207         ("*tldo_lduw_sp64"): Likewise.
25208         ("*tldo_lduw1_sp64"): Likewise.
25209         ("*tldo_ldx_sp64"): Likewise.
25210         ("*mov<VM32:mode>_insn"): Likewise.
25211         ("*mov<VM64:mode>_insn_sp64"): Likewise.
25212         ("*mov<VM64:mode>_insn_sp32"): Likewise.
25214 2017-07-07  Jose E. Marchesi  <jose.marchesi@oracle.com>
25216         * config/sparc/sparc.md ("type"): New insn type viscmp.
25217         ("fcmp<gcond:code><GCM:gcm_name><P:mode>_vis"): Set insn type to
25218         viscmp.
25219         ("fpcmp<gcond:code>8<P:mode>_vis"): Likewise.
25220         ("fucmp<gcond:code>8<P:mode>_vis"): Likewise.
25221         ("fpcmpu<gcond:code><GCM:gcm_name><P:mode>_vis"): Likewise.
25222         * config/sparc/niagara7.md ("n7_vis_logical_v3pipe"): Handle
25223         viscmp.
25224         ("n7_vis_logical_11cycle"): Likewise.
25225         * config/sparc/niagara4.md ("n4_vis_logical"): Likewise.
25226         * config/sparc/niagara2.md ("niag3_vis": Likewise.
25227         * config/sparc/niagara.md ("niag_vis"): Likewise.
25228         * config/sparc/ultra3.md ("us3_fga"): Likewise.
25229         * config/sparc/ultra1_2.md ("us1_fga_double"): Likewise.
25231 2017-07-07  Jose E. Marchesi  <jose.marchesi@oracle.com>
25233         * config/sparc/sparc.md: New instruction type `bmask'.
25234         (bmaskdi_vis): Use the `bmask' type.
25235         (bmasksi_vis): Likewise.
25236         * config/sparc/ultra3.md (us3_array): Likewise.
25237         * config/sparc/niagara7.md (n7_array): Likewise.
25238         * config/sparc/niagara4.md (n4_array): Likewise.
25239         * config/sparc/niagara2.md (niag2_vis): Likewise.
25240         (niag3_vis): Likewise.
25241         * config/sparc/niagara.md (niag_vis): Likewise.
25243 2017-07-06  Jan Hubicka  <hubicka@ucw.cz>
25245         * ipa-comdats.c: Remove optimize check from gate.
25246         * ipa-fnsummary.c (ipa_fn_summary_generate): do not generate summary
25247         for functions not optimized.
25248         (ipa_fn_summary_read): Skip optimize check.
25249         (ipa_fn_summary_write): Likewise.
25250         * ipa-inline-analysis.c (do_estimate_growth_1): Check that caller
25251         is optimized.
25252         * ipa-inline.c (can_inline_edge_p): Not optimized functions are
25253         uninlinable.
25254         (can_inline_edge_p): Check flag_pcc_struct_return for match.
25255         (check_callers): Give up on caller which is not optimized.
25256         (inline_small_functions): Likewise.
25257         (ipa_inline): Do not give up when not optimizing.
25258         * ipa-visbility.c (function_and_variable_visibility): Do not optimize
25259         away unoptimizes cdtors.
25260         (whole_program_function_and_variable_visibility): Do
25261         ipa_discover_readonly_nonaddressable_vars in LTO mode.
25262         * ipa.c (process_references): Do not check optimize.
25263         (symbol_table::remove_unreachable_nodes): Update optimize check.
25264         (set_writeonly_bit): Update optimize check.
25265         (pass_ipa_cdtor_merge::gate): Do not check optimize.
25266         (pass_ipa_single_use::gate): Remove.
25268 2017-07-06  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
25270         * config/rs6000/rs6000.c (union_defs, union_uses, insn_is_load_p,
25271         insn_is_store_p, insn_is_swap_p, const_load_sequence_p, v2df_reduction_p,
25272         rtx_is_swappable_p, insn_is_swappable_p, chain_contains_only_swaps,
25273         mark_swaps_for_removal, swap_const_vector_halves, adjust_subreg_index,
25274         permute_load, permute_store, adjust_extract, adjust_splat,
25275         adjust_xxpermdi, adjust_concat, adjust_vperm, handle_special_swappables,
25276         replace_swap_with_copy, dump_swap_insn_table,
25277         alignment_with_canonical_addr, alignment_mask, find_alignment_op,
25278         recombine_lvx_pattern, recombine_stvx_pattern,
25279         recombine_lvx_stvx_patterns, rs6000_analyze_swaps,
25280         make_pass_analyze_swaps): Move all code related to p8 swap optimizations
25281         to file rs6000-p8swap.c.
25282         * config/rs6000/rs6000-p8swap.c: New file.
25283         * config/rs6000/t-rs6000: Add rule to build rs6000-p8swap.o.
25284         * config.gcc: Add rs6000-p8swap.o to extra_objs for powerpc*-*-*
25285         and rs6000*-*-* targets.
25287 2017-07-06  David Malcolm  <dmalcolm@redhat.com>
25289         * Makefile.in (selftest): Remove dependency on s-selftest-c++.
25291 2017-07-06  Jan Hubicka  <hubicka@ucw.cz>
25293         * lto-wrapper.c (merge_and_complain): Do not merge
25294         fexceptions, fnon_call_exceptions, ftrapv, ffp_contract_, fmath_errno,
25295         fsigned_zeros, ftrapping_math, fwrapv.
25296         (append_compiler_options): Do not track these options.
25297         (append_linker_options): Likewie
25299 2017-07-06  Jan Hubicka  <hubicka@ucw.cz>
25301         * cgraphunit.c (cgraph_node::finalize_function): When
25302         !flag_toplevel_reorde set no_reorder flag.
25303         (varpool_node::finalize_decl): Likewise.
25304         (symbol_table::compile): Drop no toplevel reorder path.
25306 2017-07-06  Jan Hubicka  <hubicka@ucw.cz>
25308         * bb-reorder.c (better_edge_p): Do not build traces across abnormal/eh
25309         edges; zero probability is not better than uninitialized.
25311 2017-07-06  Maxim Ostapenko  <m.ostapenko@samsung.com>
25313         * asan.h (asan_sanitize_allocas_p): Declare.
25314         * asan.c (asan_sanitize_allocas_p): New function.
25315         (handle_builtin_stack_restore): Bail out if !asan_sanitize_allocas_p.
25316         (handle_builtin_alloca): Likewise.
25317         * cfgexpand.c (expand_used_vars): Do not add allocas unpoisoning stuff
25318         if !asan_sanitize_allocas_p.
25319         * params.def (asan-instrument-allocas): Add new option.
25320         * params.h (ASAN_PROTECT_ALLOCAS): Define.
25321         * opts.c (common_handle_option): Disable allocas sanitization for
25322         KASan by default.
25324 2017-07-06  Maxim Ostapenko  <m.ostapenko@samsung.com>
25326         * asan.c: Include gimple-fold.h.
25327         (get_last_alloca_addr): New function.
25328         (handle_builtin_stackrestore): Likewise.
25329         (handle_builtin_alloca): Likewise.
25330         (asan_emit_allocas_unpoison): Likewise.
25331         (get_mem_refs_of_builtin_call): Add new parameter, remove const
25332         quallifier from first paramerer. Handle BUILT_IN_ALLOCA,
25333         BUILT_IN_ALLOCA_WITH_ALIGN and BUILT_IN_STACK_RESTORE builtins.
25334         (instrument_builtin_call): Pass gimple iterator to
25335         get_mem_refs_of_builtin_call.
25336         (last_alloca_addr): New global.
25337         * asan.h (asan_emit_allocas_unpoison): Declare.
25338         * builtins.c (expand_asan_emit_allocas_unpoison): New function.
25339         (expand_builtin): Handle BUILT_IN_ASAN_ALLOCAS_UNPOISON.
25340         * cfgexpand.c (expand_used_vars): Call asan_emit_allocas_unpoison
25341         if function calls alloca.
25342         * gimple-fold.c (replace_call_with_value): Remove static keyword.
25343         * gimple-fold.h (replace_call_with_value): Declare.
25344         * internal-fn.c: Include asan.h.
25345         * sanitizer.def (BUILT_IN_ASAN_ALLOCA_POISON,
25346         BUILT_IN_ASAN_ALLOCAS_UNPOISON): New builtins.
25348 2017-07-06  David Malcolm  <dmalcolm@redhat.com>
25350         * Makefile.in (SELFTEST_FLAGS): Drop "-x c", moving it to...
25351         (C_SELFTEST_FLAGS): New.
25352         (CPP_SELFTEST_FLAGS): New.
25353         (SELFTEST_DEPS): New, from deps of s-selftest.
25354         (C_SELFTEST_DEPS): New, from deps of s-selftest.
25355         (CPP_SELFTEST_DEPS): New.
25356         (selftest): Add dependency on s-selftest-c++.
25357         (s-selftest): Rename to...
25358         (s-selftest-c): ...this, moving deps to SELFTEST_DEPS
25359         and C_SELFTEST_DEPS, and using C_SELFTEST_FLAGS rather
25360         than SELFTEST_FLAGS.
25361         (selftest-gdb): Rename to...
25362         (selftest-c-gdb): ...this, using C_SELFTEST_DEPS and
25363         C_SELFTEST_FLAGS.
25364         (selftest-gdb): Reintroduce as an alias for selftest-c-gdb.
25365         (selftest-valgrind): Rename to...
25366         (selftest-c-valgrind): ...this, using C_SELFTEST_DEPS and
25367         C_SELFTEST_FLAGS.
25368         (selftest-valgrind): Reintroduce as an alias for
25369         selftest-c-valgrind.
25370         (s-selftest-c++): New.
25371         (selftest-c++-gdb): New.
25372         (selftest-c++-valgrind): New.
25374 2017-07-06  Olivier Hainque  <hainque@adacore.com>
25376         * gcc.c (process_command): When deciding if undefined variables
25377         should be ignored when processing specs, accept "gcc -v" as well.
25379 2017-07-06  Jan Hubicka  <hubicka@ucw.cz>
25381         * auto-profile.c (afdo_set_bb_count, afdo_propagate_edge,
25382         afdo_annotate_cfg): Set counts/probabilities as determined by afdo.
25384 2017-07-06  Thomas Preud'homme  <thomas.preudhomme@arm.com>
25386         * config/arm/arm-cpus.in (armv8-r): Add new entry.
25387         * config/arm/arm-isa.h (ISA_ARMv8r): Define macro.
25388         * config/arm/arm-tables.opt: Regenerate.
25389         * config/arm/arm.h (enum base_architecture): Add BASE_ARCH_8R
25390         enumerator.
25391         * doc/invoke.texi: Mention -march=armv8-r and its extensions.
25393 2017-07-06  Carl Love  <cel@us.ibm.com>
25395         * ChangeLog: Clean up from mid air collision
25397 2017-07-06  Carl Love  <cel@us.ibm.com>
25399         * config/rs6000/rs6000-c.c: Add support for built-in functions
25400         vector signed int vec_subc (vector signed int, vector signed int);
25401         vector signed __int128 vec_subc (vector signed __int128,
25402                                          vector signed __int128);
25403         vector unsigned __int128 vec_subc (vector unsigned __int128,
25404                                            vector unsigned __int128);
25405         vector signed int vec_sube (vector signed int, vector signed int,
25406                                     vector signed int);
25407         vector unsigned int vec_sube (vector unsigned int,
25408                                       vector unsigned int,
25409                                       vector unsigned int);
25410         vector signed __int128 vec_sube (vector signed __int128,
25411                                          vector signed __int128,
25412                                          vector signed__int128);
25413         vector unsigned __int128 vec_sube (vector unsigned __int128,
25414                                            vector unsigned __int128,
25415                                            vector unsigned __int128);
25416         vector signed int vec_subec (vector signed int, vector signed int,
25417                                      vector signed int);
25418         vector unsigned int vec_subec (vector unsigned int,
25419                                        vector unsigned int,
25420                                        vector unsigned int);
25421         vector signed __int128 vec_subec (vector signed __int128,
25422                                           vector signed __int128,
25423                                           vector signed__int128);
25424         vector unsigned __int128 vec_subec (vector unsigned __int128,
25425                                             vector unsigned __int128,
25426                                             vector unsigned __int128);
25427         * config/rs6000/rs6000.c (ALTIVEC_BUILTIN_VEC_SUBE,
25428         ALTIVEC_BUILTIN_VEC_SUBEC): Add ef_builtins.
25429         * config/rs6000/rs6000-builtin.def (SUBE, SUBEC): Add
25430         BU_ALTIVEC_OVERLOAD_X definitions.
25431         * config/rs6000/altivec.h (vec_sube, vec_subec): Add builtin defines.
25432         * doc/extend.texi: Update the built-in documentation file for the new
25433         built-in functions.
25435 2017-07-06  David Malcolm  <dmalcolm@redhat.com>
25437         PR c++/79300
25438         * diagnostic-show-locus.c (layout::layout): Use start and finish
25439         spelling location for the start and finish of each range.
25440         * genmatch.c (linemap_client_expand_location_to_spelling_point):
25441         Add unused aspect param.
25442         * input.c (expand_location_1): Add "aspect" param, and use it
25443         to access the correct part of the location.
25444         (expand_location): Pass LOCATION_ASPECT_CARET to new param of
25445         expand_location_1.
25446         (expand_location_to_spelling_point): Likewise.
25447         (linemap_client_expand_location_to_spelling_point): Add "aspect"
25448         param, and pass it to expand_location_1.
25450 2017-07-06  Sebastian Peryt  <sebastian.peryt@intel.com>
25452         * config/i386/avx512fintrin.h (_mm_mask_getexp_round_ss,
25453         _mm_maskz_getexp_round_ss, _mm_mask_getexp_round_sd,
25454         _mm_maskz_getexp_round_sd, _mm_mask_getmant_round_sd,
25455         _mm_maskz_getmant_round_sd, _mm_mask_getmant_round_ss,
25456         _mm_maskz_getmant_round_ss, _mm_mask_getexp_ss, _mm_maskz_getexp_ss,
25457         _mm_mask_getexp_sd, _mm_maskz_getexp_sd, _mm_mask_getmant_sd,
25458         _mm_maskz_getmant_sd, _mm_mask_getmant_ss,
25459         _mm_maskz_getmant_ss): New intrinsics.
25460         (__builtin_ia32_getexpss128_mask): Changed to ...
25461         __builtin_ia32_getexpss128_round ... this.
25462         (__builtin_ia32_getexpsd128_mask): Changed to ...
25463         __builtin_ia32_getexpsd128_round ... this.
25464         * config/i386/i386-builtin-types.def
25465         ((V2DF, V2DF, V2DF, INT, V2DF, UQI, INT),
25466         (V4SF, V4SF, V4SF, INT, V4SF, UQI, INT)): New function type aliases.
25467         * config/i386/i386-builtin.def (__builtin_ia32_getexpsd_mask_round,
25468         __builtin_ia32_getexpss_mask_round,     __builtin_ia32_getmantsd_mask_round,
25469         __builtin_ia32_getmantss_mask_round): New builtins.
25470         * config/i386/i386.c (V2DF_FTYPE_V2DF_V2DF_INT_V2DF_UQI_INT,
25471         V4SF_FTYPE_V4SF_V4SF_INT_V4SF_UQI_INT): Handle new types.
25472         (CODE_FOR_avx512f_vgetmantv2df_mask_round,
25473         CODE_FOR_avx512f_vgetmantv4sf_mask_round): New cases.
25474         * config/i386/sse.md
25475         (avx512f_sgetexp<mode><round_saeonly_name>): Changed to ...
25476         avx512f_sgetexp<mode><mask_scalar_name>
25477         <round_saeonly_scalar_name> ... this.
25478         (vgetexp<ssescalarmodesuffix>\t{<round_saeonly_op3>%2, %1, %0|
25479         %0, %1, %2<round_saeonly_op3>}): Changed to ...
25480         vgetexp<ssescalarmodesuffix>
25481         \t{<round_saeonly_scalar_mask_op3>%2, %1, %0<mask_scalar_operand3>|
25482         %0<mask_scalar_operand3>, %1, %2<round_saeonly_scalar_mask_op3>} ... this.
25483         (avx512f_vgetmant<mode><round_saeonly_name>): Changed to ...
25484         avx512f_vgetmant<mode><mask_scalar_name>
25485         <round_saeonly_scalar_name> ... this.
25486         (vgetmant<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
25487         %0, %1, %2<round_saeonly_op4>, %3}): Changed to ...
25488         vgetmant<ssescalarmodesuffix>
25489         \t{%3, <round_saeonly_scalar_mask_op4>%2, %1, %0<mask_scalar_operand4>|
25490         %0<mask_scalar_operand4>, %1, %2
25491         <round_saeonly_scalar_mask_op4>, %3} ... this.
25492         * config/i386/subst.md (mask_scalar_operand4,
25493         round_saeonly_scalar_mask_operand4,     round_saeonly_scalar_mask_op4,
25494         round_saeonly_scalar_nimm_predicate): New subst attributes.
25496 2017-07-06  Julia Koval  <julia.koval@intel.com>
25498         * config/i386/i386.c (ix86_erase_embedded_rounding):
25499         Remove code for old rounding pattern.
25501 2017-07-06  Richard Earnshaw  <rearnsha@arm.com>
25503         * config/arm/t-arm (GTM_H): Add arm-cpu.h.
25505 2017-07-06  Christophe Lyon  <christophe.lyon@linaro.org>
25507         * doc/sourcebuild.texi (Test Directives, Variants of
25508         dg-require-support): Add documentation for dg-require-stack-check.
25510 2017-07-05  Sebastian Peryt  <sebastian.peryt@intel.com>
25512         * config/i386/subst.md (mask_scalar, round_scalar,
25513         round_saeonly_scalar): New meta-templates.
25514         (mask_scalar_name, mask_scalar_operand3, round_scalar_name,
25515         round_scalar_mask_operand3, round_scalar_mask_op3,
25516         round_scalar_constraint, round_scalar_prefix, round_saeonly_scalar_name,
25517         round_saeonly_scalar_mask_operand3, round_saeonly_scalar_mask_op3,
25518         round_saeonly_scalar_constraint,
25519         round_saeonly_scalar_prefix): New subst attribute.
25520         * config/i386/sse.md
25521         (<sse>_vm<plusminus_insn><mode>3<mask_name><round_name>): Renamed to ...
25522         <sse>_vm<plusminus_insn><mode>3<mask_scalar_name>
25523         <round_scalar_name> ... this.
25524         (<sse>_vm<multdiv_mnemonic><mode>3<mask_name><round_name>): Renamed to ...
25525         <sse>_vm<multdiv_mnemonic><mode>3<mask_scalar_name>
25526         <round_scalar_name> ... this.
25527         (<sse>_vm<code><mode>3<mask_name><round_saeonly_name>): Renamed to ...
25528         <sse>_vm<code><mode>3<mask_scalar_name>
25529         <round_saeonly_scalar_name> ... this.
25530         (v<plusminus_mnemonic><ssescalarmodesuffix>
25531         \t{<round_mask_op3>%2, %1, %0<mask_operand3>|
25532         %0<mask_operand3>, %1, %<iptr>2<round_mask_op3>}): Changed to ...
25533         v<plusminus_mnemonic><ssescalarmodesuffix>
25534         \t{<round_scalar_mask_op3>%2, %1, %0<mask_scalar_operand3>|
25535         %0<mask_scalar_operand3>, %1, %<iptr>2<round_scalar_mask_op3>} ... this.
25536         (v<multdiv_mnemonic><ssescalarmodesuffix>
25537         \t{<round_mask_op3>%2, %1, %0<mask_operand3>|
25538         %0<mask_operand3>, %1, %<iptr>2<round_mask_op3>}): Changed to ...
25539         v<multdiv_mnemonic><ssescalarmodesuffix>
25540         \t{<round_scalar_mask_op3>%2, %1, %0<mask_scalar_operand3>|
25541         %0<mask_scalar_operand3>, %1, %<iptr>2<round_scalar_mask_op3>} ... this.
25542         (v<maxmin_float><ssescalarmodesuffix>
25543         \t{<round_saeonly_mask_op3>%2, %1, %0<mask_operand3>|
25544         %0<mask_operand3>, %1, %<iptr>2<round_saeonly_mask_op3>}): Changed to ...
25545         v<maxmin_float><ssescalarmodesuffix>
25546         \t{<round_saeonly_scalar_mask_op3>%2, %1, %0<mask_scalar_operand3>|
25547         %0<mask_scalar_operand3>, %1, %<iptr>2
25548         <round_saeonly_scalar_mask_op3>} ... this.
25550 2017-07-05  Richard Earnshaw  <rearnsha@arm.com>
25552         * config/arm/arm.c (arm_fixed_condition_code_regs): New function.
25553         (TARGET_FIXED_CONDITION_CODE_REGS): Redefine.
25555 2017-07-05  Richard Sandiford  <richard.sandiford@linaro.org>
25556             Alan Hayward  <alan.hayward@arm.com>
25557             David Sherwood  <david.sherwood@arm.com>
25559         * combine.c (simplify_if_then_else): Remove "enum" before
25560         "machine_mode".
25561         * compare-elim.c (can_eliminate_compare): Likewise.
25562         * config/aarch64/aarch64-builtins.c (aarch64_simd_builtin_std_type):
25563         Likewise.
25564         (aarch64_lookup_simd_builtin_type): Likewise.
25565         (aarch64_simd_builtin_type): Likewise.
25566         (aarch64_init_simd_builtin_types): Likewise.
25567         (aarch64_simd_expand_args): Likewise.
25568         * config/aarch64/aarch64-protos.h (aarch64_simd_attr_length_rglist):
25569         Likewise.
25570         (aarch64_reverse_mask): Likewise.
25571         (aarch64_simd_emit_reg_reg_move): Likewise.
25572         (aarch64_gen_adjusted_ldpstp): Likewise.
25573         (aarch64_ccmp_mode_to_code): Likewise.
25574         (aarch64_operands_ok_for_ldpstp): Likewise.
25575         (aarch64_operands_adjust_ok_for_ldpstp): Likewise.
25576         * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
25577         Likewise.
25578         (aarch64_min_divisions_for_recip_mul): Likewise.
25579         (aarch64_reassociation_width): Likewise.
25580         (aarch64_get_condition_code_1): Likewise.
25581         (aarch64_simd_emit_reg_reg_move): Likewise.
25582         (aarch64_simd_attr_length_rglist): Likewise.
25583         (aarch64_reverse_mask): Likewise.
25584         (aarch64_operands_ok_for_ldpstp): Likewise.
25585         (aarch64_operands_adjust_ok_for_ldpstp): Likewise.
25586         (aarch64_gen_adjusted_ldpstp): Likewise.
25587         * config/aarch64/cortex-a57-fma-steering.c (fma_node::rename):
25588         Likewise.
25589         * config/arc/arc.c (legitimate_offset_address_p): Likewise.
25590         * config/arm/arm-builtins.c (arm_simd_builtin_std_type): Likewise.
25591         (arm_lookup_simd_builtin_type): Likewise.
25592         (arm_simd_builtin_type): Likewise.
25593         (arm_init_simd_builtin_types): Likewise.
25594         (arm_expand_builtin_args): Likewise.
25595         * config/arm/arm-protos.h (arm_expand_builtin): Likewise.
25596         * config/ft32/ft32.c (ft32_libcall_value): Likewise.
25597         (ft32_setup_incoming_varargs): Likewise.
25598         (ft32_function_arg): Likewise.
25599         (ft32_function_arg_advance): Likewise.
25600         (ft32_pass_by_reference): Likewise.
25601         (ft32_arg_partial_bytes): Likewise.
25602         (ft32_valid_pointer_mode): Likewise.
25603         (ft32_addr_space_pointer_mode): Likewise.
25604         (ft32_addr_space_legitimate_address_p): Likewise.
25605         * config/i386/i386-protos.h (ix86_operands_ok_for_move_multiple):
25606         Likewise.
25607         * config/i386/i386.c (ix86_setup_incoming_vararg_bounds): Likewise.
25608         (ix86_emit_outlined_ms2sysv_restore): Likewise.
25609         (iamcu_alignment): Likewise.
25610         (canonicalize_vector_int_perm): Likewise.
25611         (ix86_noce_conversion_profitable_p): Likewise.
25612         (ix86_mpx_bound_mode): Likewise.
25613         (ix86_operands_ok_for_move_multiple): Likewise.
25614         * config/microblaze/microblaze-protos.h
25615         (microblaze_expand_conditional_branch_reg): Likewise.
25616         * config/microblaze/microblaze.c
25617         (microblaze_expand_conditional_branch_reg): Likewise.
25618         * config/powerpcspe/powerpcspe.c (rs6000_init_hard_regno_mode_ok):
25619         Likewise.
25620         (rs6000_reassociation_width): Likewise.
25621         (rs6000_invalid_binary_op): Likewise.
25622         (fusion_p9_p): Likewise.
25623         (emit_fusion_p9_load): Likewise.
25624         (emit_fusion_p9_store): Likewise.
25625         * config/riscv/riscv-protos.h (riscv_regno_mode_ok_for_base_p):
25626         Likewise.
25627         (riscv_hard_regno_mode_ok_p): Likewise.
25628         (riscv_address_insns): Likewise.
25629         (riscv_split_symbol): Likewise.
25630         (riscv_legitimize_move): Likewise.
25631         (riscv_function_value): Likewise.
25632         (riscv_hard_regno_nregs): Likewise.
25633         (riscv_expand_builtin): Likewise.
25634         * config/riscv/riscv.c (riscv_build_integer_1): Likewise.
25635         (riscv_build_integer): Likewise.
25636         (riscv_split_integer): Likewise.
25637         (riscv_legitimate_constant_p): Likewise.
25638         (riscv_cannot_force_const_mem): Likewise.
25639         (riscv_regno_mode_ok_for_base_p): Likewise.
25640         (riscv_valid_base_register_p): Likewise.
25641         (riscv_valid_offset_p): Likewise.
25642         (riscv_valid_lo_sum_p): Likewise.
25643         (riscv_classify_address): Likewise.
25644         (riscv_legitimate_address_p): Likewise.
25645         (riscv_address_insns): Likewise.
25646         (riscv_load_store_insns): Likewise.
25647         (riscv_force_binary): Likewise.
25648         (riscv_split_symbol): Likewise.
25649         (riscv_force_address): Likewise.
25650         (riscv_legitimize_address): Likewise.
25651         (riscv_move_integer): Likewise.
25652         (riscv_legitimize_const_move): Likewise.
25653         (riscv_legitimize_move): Likewise.
25654         (riscv_address_cost): Likewise.
25655         (riscv_subword): Likewise.
25656         (riscv_output_move): Likewise.
25657         (riscv_canonicalize_int_order_test): Likewise.
25658         (riscv_emit_int_order_test): Likewise.
25659         (riscv_function_arg_boundary): Likewise.
25660         (riscv_pass_mode_in_fpr_p): Likewise.
25661         (riscv_pass_fpr_single): Likewise.
25662         (riscv_pass_fpr_pair): Likewise.
25663         (riscv_get_arg_info): Likewise.
25664         (riscv_function_arg): Likewise.
25665         (riscv_function_arg_advance): Likewise.
25666         (riscv_arg_partial_bytes): Likewise.
25667         (riscv_function_value): Likewise.
25668         (riscv_pass_by_reference): Likewise.
25669         (riscv_setup_incoming_varargs): Likewise.
25670         (riscv_print_operand): Likewise.
25671         (riscv_elf_select_rtx_section): Likewise.
25672         (riscv_save_restore_reg): Likewise.
25673         (riscv_for_each_saved_reg): Likewise.
25674         (riscv_register_move_cost): Likewise.
25675         (riscv_hard_regno_mode_ok_p): Likewise.
25676         (riscv_hard_regno_nregs): Likewise.
25677         (riscv_class_max_nregs): Likewise.
25678         (riscv_memory_move_cost): Likewise.
25679         * config/rl78/rl78-protos.h (rl78_split_movsi): Likewise.
25680         * config/rl78/rl78.c (rl78_split_movsi): Likewise.
25681         (rl78_addr_space_address_mode): Likewise.
25682         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
25683         Likewise.
25684         * config/rs6000/rs6000.c (rs6000_init_hard_regno_mode_ok): Likewise.
25685         (rs6000_reassociation_width): Likewise.
25686         (rs6000_invalid_binary_op): Likewise.
25687         (fusion_p9_p): Likewise.
25688         (emit_fusion_p9_load): Likewise.
25689         (emit_fusion_p9_store): Likewise.
25690         * config/visium/visium-protos.h (prepare_move_operands): Likewise.
25691         (ok_for_simple_move_operands): Likewise.
25692         (ok_for_simple_move_strict_operands): Likewise.
25693         (ok_for_simple_arith_logic_operands): Likewise.
25694         (visium_legitimize_reload_address): Likewise.
25695         (visium_select_cc_mode): Likewise.
25696         (output_cbranch): Likewise.
25697         (visium_split_double_move): Likewise.
25698         (visium_expand_copysign): Likewise.
25699         (visium_expand_int_cstore): Likewise.
25700         (visium_expand_fp_cstore): Likewise.
25701         * config/visium/visium.c (visium_pass_by_reference): Likewise.
25702         (visium_function_arg): Likewise.
25703         (visium_function_arg_advance): Likewise.
25704         (visium_libcall_value): Likewise.
25705         (visium_setup_incoming_varargs): Likewise.
25706         (visium_legitimate_constant_p): Likewise.
25707         (visium_legitimate_address_p): Likewise.
25708         (visium_legitimize_address): Likewise.
25709         (visium_secondary_reload): Likewise.
25710         (visium_register_move_cost): Likewise.
25711         (visium_memory_move_cost): Likewise.
25712         (prepare_move_operands): Likewise.
25713         (ok_for_simple_move_operands): Likewise.
25714         (ok_for_simple_move_strict_operands): Likewise.
25715         (ok_for_simple_arith_logic_operands): Likewise.
25716         (visium_function_value_1): Likewise.
25717         (rtx_ok_for_offset_p): Likewise.
25718         (visium_legitimize_reload_address): Likewise.
25719         (visium_split_double_move): Likewise.
25720         (visium_expand_copysign): Likewise.
25721         (visium_expand_int_cstore): Likewise.
25722         (visium_expand_fp_cstore): Likewise.
25723         (visium_split_cstore): Likewise.
25724         (visium_select_cc_mode): Likewise.
25725         (visium_split_cbranch): Likewise.
25726         (output_cbranch): Likewise.
25727         (visium_print_operand_address): Likewise.
25728         * expmed.c (flip_storage_order): Likewise.
25729         * expmed.h (emit_cstore): Likewise.
25730         (flip_storage_order): Likewise.
25731         * genrecog.c (validate_pattern): Likewise.
25732         * hsa-gen.c (gen_hsa_addr): Likewise.
25733         * internal-fn.c (expand_arith_overflow): Likewise.
25734         * ira-color.c (allocno_copy_cost_saving): Likewise.
25735         * lra-assigns.c (find_hard_regno_for_1): Likewise.
25736         * lra-constraints.c (prohibited_class_reg_set_mode_p): Likewise.
25737         (process_invariant_for_inheritance): Likewise.
25738         * lra-eliminations.c (move_plus_up): Likewise.
25739         * omp-low.c (lower_oacc_reductions): Likewise.
25740         * simplify-rtx.c (simplify_subreg): Likewise.
25741         * target.def (TARGET_SETUP_INCOMING_VARARG_BOUNDS): Likewise.
25742         (TARGET_CHKP_BOUND_MODE): Likewise..
25743         * targhooks.c (default_chkp_bound_mode): Likewise.
25744         (default_setup_incoming_vararg_bounds): Likewise.
25745         * targhooks.h (default_chkp_bound_mode): Likewise.
25746         (default_setup_incoming_vararg_bounds): Likewise.
25747         * tree-ssa-math-opts.c (divmod_candidate_p): Likewise.
25748         * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
25749         (have_whole_vector_shift): Likewise.
25750         * tree-vect-stmts.c (vectorizable_load): Likewise.
25751         * doc/tm.texi: Regenerate.
25753 2017-07-05  Georg-Johann Lay  <avr@gjlay.de>
25755         Graceful degrade if Binutils PR21472 is not available.
25757         PR target/81072
25758         * configure.ac [target=avr]: WARN instead of ERROR if avrxmega3
25759         .rodata in flash test fails.
25760         (HAVE_LD_AVR_AVRXMEGA3_RODATA_IN_FLASH): Define it if test passes.
25761         * confgure: Regenerate.
25762         * config.in: Regenerate.
25763         * config/avr/avr.c (avr_asm_named_section)
25764         [HAVE_LD_AVR_AVRXMEGA3_RODATA_IN_FLASH]: Only trigger
25765         __do_copy_data for stuff in .rodata if flash_pm_offset = 0.
25766         (avr_asm_init_sections): Same.
25768 2017-07-05  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
25770         * config/arm/neon.md (fma<VCVTF:mode>4): Remove %?.
25771         (fma<VH:mode>4_intrinsic): Likewise.
25772         (*fmsub<VCVTF:mode>4): Likewise.
25773         (*fmsub<VH:mode>4_intrinsic): Likewise.
25775 2017-07-05  Georg-Johann Lay  <avr@gjlay.de>
25777         PR target/81305
25778         * config/avr/avr.c (avr_out_movhi_mr_r_xmega) [CONSTANT_ADDRESS_P]:
25779         Don't depend on "optimize > 0".
25780         (out_movhi_r_mr, out_movqi_mr_r): Same.
25781         (out_movhi_mr_r, out_movqi_r_mr): Same.
25782         (avr_address_cost) [CONSTANT_ADDRESS_P]: Don't depend cost for
25783         io_address_operand on "optimize > 0".
25785 2017-07-05  Bin Cheng  <bin.cheng@arm.com>
25787         * tree-loop-distribution.c: Add general explanantion on the pass.
25788         (generate_loops_for_partition): Mark distributed loop.
25789         (pg_add_dependence_edges): New parameter.  Handle alias data
25790         dependence specially and record it in the parameter if asked.
25791         (struct pg_vdata, pg_edata, pg_edge_callback_data): New structs.
25792         (init_partition_graph_vertices, add_partition_graph_edge): New.
25793         (pg_skip_alias_edge, free_partition_graph_edata_cb): New.
25794         (free_partition_graph_vdata, build_partition_graph): New.
25795         (sort_partitions_by_post_order, merge_dep_scc_partitions): New.
25796         (pg_collect_alias_ddrs, break_alias_scc_partitions): New.
25797         (data_ref_segment_size, latch_dominated_by_data_ref): New.
25798         (compute_alias_check_pairs, version_loop_by_alias_check): New.
25799         (version_for_distribution_p, finalize_partitions): New.
25800         (distribute_loop): Handle alias data dependence specially.  Factor
25801         out loop fusion code as functions and call these functions.
25803 2017-07-05  Bin Cheng  <bin.cheng@arm.com>
25805         * tree-loop-distribution.c (classify_partition): New parameter and
25806         better handle reduction statement.
25807         (rdg_build_partitions): Revise comment.
25808         (distribute_loop): Compute statements in all partitions and pass it
25809         to classify_partition.
25811 2017-07-05  Bin Cheng  <bin.cheng@arm.com>
25813         * tree-loop-distribution.c (enum partition_type): New.
25814         (struct partition): New field type.
25815         (partition_merge_into): Add parameter.  Update partition type.
25816         (data_dep_in_cycle_p, update_type_for_merge): New functions.
25817         (build_rdg_partition_for_vertex): Compute partition type.
25818         (rdg_build_partitions): Dump partition type.
25819         (distribute_loop): Update calls to partition_merge_into.
25821 2017-07-05  Bin Cheng  <bin.cheng@arm.com>
25823         * tree-loop-distribution.c (struct ddr_hasher): New.
25824         (ddr_hasher::hash, ::equal, get_data_dependence): New function.
25825         (ddrs_table): New.
25826         (classify_partition): Call get_data_dependence.
25827         (pg_add_dependence_edges): Ditto.
25828         (distribute_loop): Release data dependence hash table.
25830 2017-07-05  Bin Cheng  <bin.cheng@arm.com>
25832         * tree-loop-distribution.c (ref_base_address): Delete.
25833         (similar_memory_accesses): Rename ...
25834         (share_memory_accesses): ... to this.  Check if partitions access
25835         the same memory reference.
25836         (distribute_loop): Call share_memory_accesses.
25838 2017-07-05  Bin Cheng  <bin.cheng@arm.com>
25840         * tree-loop-distribution.c (struct partition): New field recording
25841         its data reference.
25842         (partition_alloc, partition_free): Init and release data refs.
25843         (partition_merge_into): Merge data refs.
25844         (build_rdg_partition_for_vertex): Collect data refs for partition.
25845         (pg_add_dependence_edges): Change parameters from vector to bitmap.
25846         Update uses.
25847         (distribute_loop): Remve data refs from vertice data of partition
25848         graph.
25850 2017-07-05  Bin Cheng  <bin.cheng@arm.com>
25852         * tree-loop-distribution.c (params.h): Include header file.
25853         (MAX_DATAREFS_NUM, DR_INDEX): New macro.
25854         (datarefs_vec): New global var.
25855         (create_rdg_vertices): Use datarefs_vec directly.
25856         (free_rdg): Don't free data references.
25857         (build_rdg): Update use.  Don't free data references.
25858         (distribute_loop): Compute global variable for data references.
25859         Bail out if there are too many data references.
25861 2017-07-05  Bin Cheng  <bin.cheng@arm.com>
25863         * tree-loop-distribution.c (loop_nest): New global var.
25864         (build_rdg): Use loop directly, rather than loop nest.
25865         (pg_add_dependence_edges): Remove loop nest parameter.  Use global
25866         variable directly.
25867         (distribute_loop): Compute global variable loop nest.  Update use.
25869 2017-07-05  Bin Cheng  <bin.cheng@arm.com>
25871         * tree-loop-distribution.c (enum fuse_type, fuse_message): New.
25872         (partition_merge_into): New parameter.  Dump reason for fusion.
25873         (distribute_loop): Update use of partition_merge_into.
25875 2017-07-05  Bin Cheng  <bin.cheng@arm.com>
25877         * tree-loop-distribution.c (bb_top_order_index): New.
25878         (bb_top_order_index_size, bb_top_order_cmp): New.
25879         (stmts_from_loop): Use topological order.
25880         (pass_loop_distribution::execute): Compute and release topological
25881         order for basic blocks.
25883 2017-07-05  Bin Cheng  <bin.cheng@arm.com>
25885         * tree-loop-distribution.c (pass_loop_distribution::execute): Skip
25886         if no loops.
25888 2017-07-05  Bin Cheng  <bin.cheng@arm.com>
25890         * cfgloop.h (struct loop): Add comment.  New field orig_loop_num.
25891         * cfgloopmanip.c (lv_adjust_loop_entry_edge): Comment change.
25892         * internal-fn.c (expand_LOOP_DIST_ALIAS): New function.
25893         * internal-fn.def (LOOP_DIST_ALIAS): New.
25894         * tree-vectorizer.c (fold_loop_vectorized_call): Rename to ...
25895         (fold_loop_internal_call): ... this.
25896         (vect_loop_dist_alias_call): New function.
25897         (set_uid_loop_bbs): Call fold_loop_internal_call.
25898         (vectorize_loops): Fold IFN_LOOP_VECTORIZED and IFN_LOOP_DIST_ALIAS
25899         internal calls.
25901 2017-07-04  Uros Bizjak  <ubizjak@gmail.com>
25903         PR target/81300
25904         * config/i386/i386.md (setcc + movzbl/and to xor + setcc peepholes):
25905         Require dead FLAGS_REG at the beginning of a peephole.
25907 2017-07-04  Uros Bizjak  <ubizjak@gmail.com>
25909         PR target/81294
25910         * config/i386/adxintrin.h (_subborrow_u32): Swap _X and _Y
25911         arguments in the call to __builtin_ia32_sbb_u32.
25912         (_subborrow_u64): Swap _X and _Y arguments in the call to
25913         __builtin_ia32_sbb_u64.
25915 2017-07-04  Jakub Jelinek  <jakub@redhat.com>
25917         PR debug/81278
25918         * tree-vrp.c (compare_assert_loc): Turn into a function template
25919         with stable template parameter.  Only test if a->e is NULL,
25920         !a->e == !b->e has been verified already.  Use e == NULL or
25921         e != NULL instead of e or ! e tests.  If stable is true, don't use
25922         iterative_hash_expr, on the other side allow a or b or both NULL
25923         and sort the NULLs last.
25924         (process_assert_insertions): Sort using compare_assert_loc<false>
25925         instead of compare_assert_loc, later sort using
25926         compare_assert_loc<true> before calling process_assert_insertions_for
25927         in a loop.  Use break instead of continue once seen NULL pointer.
25929 2017-07-04  Thomas Preud'homme  <thomas.preudhomme@arm.com>
25931         * config/arm/driver-arm.c (arm_cpu_table): Add entry for ARM
25932         Cortex-R7 and Cortex-R8 processors.
25934 2017-07-04  Jan Hubicka  <hubicka@ucw.cz>
25936         * ipa-utils.c (ipa_merge_profiles): Fix merging when dst is
25937         uninitialized while src is not.
25939 2017-07-04  Richard Earnshaw  <rearnsha@arm.com>
25941         * common/config/arm/arm-common.c: Adjust include path for
25942         arm-cpu-cdata.h
25943         * config/arm/t-arm (TM_H): Adjust path for arm-cpu.h.
25944         (arm-cpu.h): Create in build directory.  Adjust dependency rules.
25945         (arm-cpu-data.h): Likewise.
25946         (arm-cpu-cdata.h): Likewise.
25947         * config/arm/arm-cpu.h: Delete.
25948         * config/arm/arm-cpu-cdata.h: Delete.
25949         * config/arm/arm-cpu-data.h: Delete.
25951 2017-07-04  James Greenhalgh  <james.greenhalgh@arm.com>
25953         * config/arm/arm-cpus.in (cortex-a55): New.
25954         (cortex-a75): Likewise.
25955         (cortex-a75.cortex-a55): Likewise.
25956         * config/arm/driver-arm.c (arm_cpu_table): Add cortex-a55 and
25957         cortex-a75.
25958         * doc/invoke.texi (-mcpu): Document cortex-a55 and cortex-a75.
25959         * config/arm/arm-cpu-cdata.h: Regenerate.
25960         * config/arm/arm-cpu-data.h: Regenerate.
25961         * config/arm/arm-cpu.h: Regenerate.
25962         * config/arm/arm-tables.opt: Regenerate.
25963         * config/arm/arm-tune.md: Regenerate.
25965 2017-07-04  Jan Hubicka  <hubicka@ucw.cz>
25967         * haifa-sched.c (sched_create_recovery_edges): Update profile.
25969 2017-07-04  Jan Hubicka  <hubicka@ucw.cz>
25971         * bb-reorder.c (better_edge_p): Fix handling of uninitialized
25972         probability.
25974 2017-07-04  Richard Sandiford  <richard.sandiford@linaro.org>
25976         PR tree-optimization/81292
25977         * tree-ssa-strlen.c (handle_builtin_strlen): When setting
25978         full_string_p, also call adjust_related_strinfos if the adjustment
25979         is simple, otherwise invalidate related strinfos.
25981 2017-07-04  Martin Liska  <mliska@suse.cz>
25983         PR sanitizer/81040
25984         * sanopt.c (sanitize_rewrite_addressable_params): Mark the
25985         newly created variable as DECL_IGNORED_P.
25987 2017-07-04  Martin Liska  <mliska@suse.cz>
25989         PR ipa/81293
25990         * ipa-inline.c (inline_small_functions):
25991         Use xstrdup_for_dump.
25993 2017-07-04  Tom de Vries  <tom@codesourcery.com>
25995         * graph.c (draw_cfg_edges): Save and restore EDGE_DFS_BACK.
25997 2017-07-03  Dominique d'Humieres  <dominiq@lps.ens.fr>
25999         PR target/81033
26000         * config/darwin.c (darwin_function_switched_text_sections):
26001         Fix spaces.
26003 2017-07-03  Jan Hubicka  <hubicka@ucw.cz>
26005         * tree-vect-loop-manip.c (vect_do_peeling): Fix scaling up.
26007 2017-07-03  Richard Earnshaw  <rearnsha@arm.com>
26009         * doc/invoke.texi (ARM Options): Add -mbe8 and -mbe32 to option summary.
26011 2017-07-03  Richard Sandiford  <richard.sandiford@linaro.org>
26013         * tree-vect-loop.c (vect_analyze_loop_2): Treat min_scalar_loop_bound,
26014         min_profitable_iters, and th as inclusive lower bounds.
26015         Fix LOOP_VINFO_PEELING_FOR_GAPS condition.
26016         (vect_estimate_min_profitable_iters): Return inclusive lower bounds
26017         for min_profitable_iters and min_profitable_estimate.
26018         (vect_transform_loop): Treat th as an inclusive lower bound.
26019         * tree-vect-loop-manip.c (vect_loop_versioning): Likewise.
26021 2017-07-03  Dominique d'Humieres  <dominiq@lps.ens.fr>
26023         PR target/81033
26024         * config/darwin.c (darwin_function_switched_text_sections):
26025         Replace DECL_NAME with DECL_ASSEMBLER_NAME, split assemble_name_raw
26026         in two pieces, and suppress the use of buf.
26028 2017-07-03  Nathan Sidwell  <nathan@acm.org>
26030         * hash-table.h (hash_table_mod1): Fix indentation.
26032 2017-07-02  Jan Hubicka  <hubicka@ucw.cz>
26034         PR middle-end/81290
26035         * predict.c (force_edge_cold): Be more careful about propagation
26036         backward.
26037         * profile-count.h (profile_probability::guessed,
26038         profile_probability::fdo, profile_count::guessed, profile_count::fdo):
26039         New.
26040         * tree-ssa-threadupdate.c (recompute_probabilities): Result is guessed.
26042 2017-07-03  James Greenhalgh  <james.greenhalgh@arm.com>
26044         * doc/invoke.texi (rcpc architecture extension): Document it.
26046 2017-07-03  Richard Biener  <rguenther@suse.de>
26048         PR tree-optimization/60510
26049         * tree-vect-loop.c (vect_create_epilog_for_reduction): Pass in
26050         the scalar reduction PHI and use it.
26051         (vectorizable_reduction): Properly guard the single_defuse_cycle
26052         path for non-SLP reduction chains where we cannot use it.
26053         Rework reduc_def/index and vector type deduction.  Rework
26054         vector operand gathering during reduction op code-gen.
26055         * tree-vect-slp.c (vect_analyze_slp): For failed SLP reduction
26056         chains dissolve the chain and leave it to non-SLP reduction
26057         handling.
26059 2017-07-03  Richard Sandiford  <richard.sandiford@linaro.org>
26061         * tree-data-ref.h (dr_alignment): Declare.
26062         * tree-data-ref.c (dr_alignment): New function.
26063         * tree-vectorizer.h (dataref_aux): Remove base_element_aligned.
26064         * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
26065         set it.
26066         * tree-vect-stmts.c (vectorizable_store): Use dr_alignment.
26068 2017-07-03  Richard Sandiford  <richard.sandiford@linaro.org>
26070         * tree-data-ref.h (innermost_loop_behavior): Add base_alignment
26071         and base_misalignment fields.
26072         (DR_BASE_ALIGNMENT, DR_BASE_MISALIGNMENT): New macros.
26073         * tree-data-ref.c: Include builtins.h.
26074         (dr_analyze_innermost): Set up the new innmost_loop_behavior fields.
26075         * tree-vectorizer.h (STMT_VINFO_DR_BASE_ALIGNMENT): New macro.
26076         (STMT_VINFO_DR_BASE_MISALIGNMENT): Likewise.
26077         * tree-vect-data-refs.c: Include tree-cfg.h.
26078         (vect_compute_data_ref_alignment): Use the new innermost_loop_behavior
26079         fields instead of calculating an alignment here.
26080         (vect_analyze_data_refs): Use dr_analyze_innermost.  Dump the new
26081         innermost_loop_behavior fields.
26083 2017-07-03  Richard Sandiford  <richard.sandiford@linaro.org>
26085         * tree-data-ref.h (innermost_loop_behavior): Add a step_alignment
26086         field.
26087         (DR_STEP_ALIGNMENT): New macro.
26088         * tree-vectorizer.h (STMT_VINFO_DR_STEP_ALIGNMENT): Likewise.
26089         * tree-data-ref.c (dr_analyze_innermost): Initalize step_alignment.
26090         (create_data_ref): Print it.
26091         * tree-vect-stmts.c (vectorizable_load): Use the step alignment
26092         to tell whether the step preserves vector (mis)alignment.
26093         * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Likewise.
26094         Move the check for an integer step and generalise to all INTEGER_CST.
26095         (vect_analyze_data_refs): Set DR_STEP_ALIGNMENT when setting DR_STEP.
26096         Print the outer step alignment.
26098 2017-07-03  Richard Sandiford  <richard.sandiford@linaro.org>
26100         * tree-data-ref.h (innermost_loop_behavior): Replace aligned_to
26101         with offset_alignment.
26102         (DR_ALIGNED_TO): Delete.
26103         (DR_OFFSET_ALIGNMENT): New macro.
26104         * tree-vectorizer.h (STMT_VINFO_DR_ALIGNED_TO): Delete.
26105         (STMT_VINFO_DR_OFFSET_ALIGNMENT): New macro.
26106         * tree-data-ref.c (dr_analyze_innermost): Update after above changes.
26107         (create_data_ref): Likewise.
26108         * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Likewise.
26109         (vect_analyze_data_refs): Likewise.
26110         * tree-if-conv.c (if_convertible_loop_p_1): Use memset before
26111         creating dummy innermost behavior.
26113 2017-07-03  Richard Sandiford  <richard.sandiford@linaro.org>
26115         * tree-data-ref.h (dr_analyze_innermost): Replace the dr argument
26116         with a "innermost_loop_behavior *" and refeence tree.
26117         * tree-data-ref.c (dr_analyze_innermost): Likewise.
26118         (create_data_ref): Update call accordingly.
26119         * tree-predcom.c (find_looparound_phi): Likewise.
26121 2017-07-03  Richard Sandiford  <richard.sandiford@linaro.org>
26123         * tree-vectorizer.h (_stmt_vec_info): Replace individual dr_*
26124         fields with dr_wrt_vec_loop.
26125         (STMT_VINFO_DR_BASE_ADDRESS, STMT_VINFO_DR_INIT, STMT_VINFO_DR_OFFSET)
26126         (STMT_VINFO_DR_STEP, STMT_VINFO_DR_ALIGNED_TO): Update accordingly.
26127         (STMT_VINFO_DR_WRT_VEC_LOOP): New macro.
26128         (vect_dr_behavior): New function.
26129         (vect_create_addr_base_for_vector_ref): Remove loop parameter.
26130         * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
26131         vect_dr_behavior.  Use a step_preserves_misalignment_p boolean to
26132         track whether the step preserves the misalignment.
26133         (vect_create_addr_base_for_vector_ref): Remove loop parameter.
26134         Use vect_dr_behavior.
26135         (vect_setup_realignment): Update call accordingly.
26136         (vect_create_data_ref_ptr): Likewise.  Use vect_dr_behavior.
26137         * tree-vect-loop-manip.c (vect_gen_prolog_loop_niters): Update
26138         call to vect_create_addr_base_for_vector_ref.
26139         (vect_create_cond_for_align_checks): Likewise.
26140         * tree-vect-patterns.c (vect_recog_bool_pattern): Copy
26141         STMT_VINFO_DR_WRT_VEC_LOOP as a block.
26142         (vect_recog_mask_conversion_pattern): Likewise.
26143         * tree-vect-stmts.c (compare_step_with_zero): Use vect_dr_behavior.
26144         (new_stmt_vec_info): Remove redundant zeroing.
26146 2017-07-03  Richard Earnshaw  <rearnsha@arm.com>
26148         * common/config/arm/arm-common.c (arm_be8_option): New function.
26149         * config/arm/arm-isa.h (isa_feature): Add new feature bit isa_bit_be8.
26150         (ISA_ARMv6): Add isa_bit_be8.
26151         * config/arm/arm.h (arm_be8_option): Add prototype.
26152         (BE8_SPEC_FUNCTION): New define.
26153         (EXTRA_SPEC_FUNCTIONS): Add BE8_SPEC_FUNCTION.
26154         * config/arm/arm.opt (mbig-endian): Mark as Negative of mlittle-endian.
26155         (mlittle-endian): Similarly.
26156         (mbe8, mbe32): New options.
26157         * config/arm/bpabi.h (BE8_LINK_SPEC): Call arm_be8_option.
26158         * doc/invoke.texi (ARM Options): Document -mbe8 and -mbe32.
26160 2017-07-02  Jan Hubicka  <hubicka@ucw.cz>
26162         * tree-cfg.c (gimple_find_sub_bbs): Fix profile updating.
26164 2017-07-02  Jan Hubicka  <hubicka@ucw.cz>
26166         * tree-cfgcleanup.c (want_merge_blocks_p): New function.
26167         (cleanup_tree_cfg_bb): Use it.
26168         * profile-count.h (profile_count::of_for_merging, profile_count::merge):
26169         New functions.
26170         * tree-cfg.c (gimple_merge_blocks): Use profile_count::merge.
26172 2017-07-02  Jan Hubicka  <hubicka@ucw.cz>
26174         PR bootstrap/81285
26175         * loop-doloop.c (add_test): Update profile.
26177 2017-07-03  Martin Liska  <mliska@suse.cz>
26179         PR sanitize/81040
26180         * sanopt.c (rewrite_usage_of_param): New function.
26181         (sanitize_rewrite_addressable_params): Likewise.
26182         (pass_sanopt::execute): Call rewrite_usage_of_param.
26184 2017-07-03  Richard Biener  <rguenther@suse.de>
26186         * tree-vect-loop.c (vect_create_epilog_for_reduction): Revert
26187         back to using VIEW_CONVERT_EXPR.
26189 2017-07-03  Martin Liska  <mliska@suse.cz>
26191         PR other/78366
26192         * doc/extend.texi: Document when a resolver function is
26193         generated for target_clones.
26195 2017-07-03  Martin Liska  <mliska@suse.cz>
26197         * asan.c (asan_emit_stack_protection): Unpoison just red zones
26198         and shadow memory of auto variables which are subject of
26199         use-after-scope sanitization.
26200         (asan_expand_mark_ifn): Add do set only when is_poison.
26202 2016-07-03  Richard Biener  <rguenther@suse.de>
26204         * tree-vect-loop.c (vect_analyze_loop_operations): Also analyze
26205         reduction PHIs.
26206         (vect_force_simple_reduction): Record reduction def -> phi mapping.
26207         (vectorizable_reduction): Perform reduction PHI creation when
26208         visiting a reduction PHI and adjust and simplify code generation
26209         phase of the reduction op.  Cache dts, use fold_binary, not fold_build2.
26210         (vect_transform_loop): Visit reduction PHIs.
26211         * tree-vect-slp.c (vect_get_and_check_slp_defs): Record reduction
26212         defs into the SLP tree.
26213         (vect_build_slp_tree): Reduction defs terminate the recursion.
26214         * tree-vect-stmts.c (vect_get_vec_def_for_operand_1): Allow lookup
26215         of reduction defs.
26216         (vect_get_vec_defs_for_stmt_copy): Export.
26217         (vect_get_vec_defs): Likewise.
26218         * tree-vectorizer.h (struct _stmt_vec_info): Amend reduc_def
26219         purpose.
26220         (vect_get_vec_defs_for_stmt_copy): Declare.
26221         (vect_get_vec_defs): Likewise.
26223 2017-07-03  Richard Sandiford  <richard.sandiford@linaro.org>
26225         * tree-data-ref.c (dr_analyze_innermost): Replace the "nest"
26226         parameter with a "loop" parameter and use it instead of the
26227         loop containing DR_STMT.  Don't check simple_iv when doing
26228         BB analysis.  Describe the two analysis modes in the comment.
26230 2017-07-03  Tom de Vries  <tom@codesourcery.com>
26232         PR tree-optimization/69468
26233         * tree-ssa-tail-merge.c (ignore_edge_flags): New constant.
26234         (find_same_succ_bb): Handle ignore_edge_flags.
26236 2017-07-03  Tom de Vries  <tom@codesourcery.com>
26238         PR tree-optimization/81192
26239         * tree-ssa-tail-merge.c (same_succ_hash): Use bb->loop_father->num in
26240         hash.
26241         (same_succ::equal): Don't find bbs to be equal if bb->loop_father
26242         differs.
26243         (find_same_succ_bb): Remove obsolete test on bb->loop_father->latch.
26245 2017-07-03  Tom de Vries  <tom@codesourcery.com>
26247         PR tree-optimization/81192
26248         * tree-ssa-tail-merge.c (same_succ_flush_bb): Handle
26249         BB_SAME_SUCC (bb) == NULL.
26251 2017-07-02  Jan Hubicka  <hubicka@ucw.cz>
26253         * cfgrtl.c (rtl_verify_edges): Enable checking of profile_probability
26254         consistency.
26256 2017-07-02  Jan Hubicka  <hubicka@ucw.cz>
26258         * dumpfile.c: Include profile-count.h
26259         * tree-cfg.c (gimple_duplicate_sese_tail): Drop UNUSED attributes;
26260         update profile.
26261         (insert_cond_bb): Update profile.
26262         * tree-cfg.h (insert_cond_bb): Update prototype.
26263         * tree-chkp-opt.c (chkp_optimize_string_function_calls): Update.
26264         * tree-dump.c: Do not include tree-cfg.
26266 2017-07-02  Jan Hubicka  <hubicka@ucw.cz>
26268         * bb-reorder.c (fix_up_crossing_landing_pad): Update profile.
26270 2017-07-02  Jan Hubicka  <hubicka@ucw.cz>
26272         * expect.c (dw2_build_landing_pads): Update profile of the landing pad
26273         bb.
26275 2017-07-02  Jan Hubicka  <hubicka@ucw.cz>
26277         * tree-complex.c (expand_complex_div_wide): update profile.
26279 2017-07-02  Richard Sandiford  <richard.sandiford@linaro.org>
26280             Alan Hayward  <alan.hayward@arm.com>
26281             David Sherwood  <david.sherwood@arm.com>
26283         * Makefile.in (MACHMODE_H): Remove insn-modes.h
26284         (CORETYPES_H): New define.
26285         (MOSTLYCLEANFILES): Add insn-modes-inline.h.
26286         (insn-modes-inline.h, s-modes-inline-h): New rules.
26287         (generated_files): Add insn-modes-inline.h.
26288         (RTL_BASE_H, TREE_CORE_H): Use CORETYPES_H instead of coretypes.h.
26289         (build/gensupport.o, build/ggc-none.o, build/print-rtl.o): Likewise.
26290         (build/read-md.o, build/read-rtl.o, build/rtl.o): Likewise.
26291         (build/vec.o, build/hash-table.o, build/inchash.o): Likewise.
26292         (build/gencondmd.o, build/genattr.o, build/genattr-common.o): Likewise.
26293         (build/genattrtab.o, build/genautomata.o, build/gencheck.o): Likewise.
26294         (build/gencodes.o, build/genconditions.o): Likewise.
26295         (build/genconfig.o, build/genconstants.o, build/genemit.o): Likewise.
26296         (build/genenums.o, build/genextract.o, build/genflags.o): Likewise.
26297         (build/gentarget-def.o, build/genmddeps.o, build/genopinit.o)
26298         (build/genoutput.o, build/genpeep.o, build/genpreds.o): Likewise.
26299         (build/genrecog.o, build/genmddump.o, build/genmatch.o): Likewise.
26300         (build/gencfn-macros.o, build/gcov-iov.o): Likewise.
26301         * coretypes.h: Include everything up to real.h for generators.
26302         Include insn-modes.h first.  Include wide-int-print.h after
26303         wide-int.h.  Include insn-modes-inline.h and then machmode.h.
26304         * machmode.h: Don't include insn-modes.h here.
26305         * function-tests.c: Remove includes of signop.h, machmode.h,
26306         double-int.h and wide-int.h.
26307         * rtl.h: Likewise.
26308         * gcc-rich-location.c: Remove includes of machmode.h, double-int.h
26309         and wide-int.h.
26310         * optc-save-gen.awk: Likewise.
26311         * gencheck.c (BITS_PER_UNIT): Delete dummy definition.
26312         * godump.c: Remove include of wide-int-print.h.
26313         * pretty-print.h: Likewise.
26314         * wide-int-print.cc: Likewise.
26315         * wide-int.cc: Likewise.
26316         * hash-map-tests.c: Remove include of signop.h.
26317         * hash-set-tests.c: Likewise.
26318         * rtl-tests.c: Likewise.
26319         * mkconfig.sh: Remove include of machmode.h.
26320         * genmodes.c (emit_insn_modes_h): Split emission of inline functions
26321         into...
26322         (emit_insn_modes_inline_h): ...this new function.  Emit the code
26323         into an insn-modes-inline.h header file, adding appropriate
26324         include guards and end comments.
26325         (emit_insn_modes_c_header): Remove include of machmode.h.
26326         (emit_min_insn_modes_c_header): Include coretypes.h rather than
26327         machmode.h.
26328         (main): Handle -i flag and call emit_insn_modes_inline_h when
26329         it is passed.
26331 2017-07-02  Richard Sandiford  <richard.sandiford@linaro.org>
26333         * tree-ssa-strlen.c (strinfo): Rename the length field to
26334         nonzero_chars.  Add a full_string_p field.
26335         (compare_nonzero_chars, zero_length_string_p): New functions.
26336         (get_addr_stridx): Add an offset_out parameter.
26337         Use compare_nonzero_chars.
26338         (get_stridx): Update accordingly.  Use compare_nonzero_chars.
26339         (new_strinfo): Update after above changes to strinfo.
26340         (set_endptr_and_length): Set full_string_p.
26341         (get_string_length): Update after above changes to strinfo.
26342         (unshare_strinfo): Update call to new_strinfo.
26343         (maybe_invalidate): Likewise.
26344         (get_stridx_plus_constant): Change off to unsigned HOST_WIDE_INT.
26345         Use compare_nonzero_chars and zero_string_p.  Treat nonzero_chars
26346         as a uhwi instead of an shwi.  Update after above changes to
26347         strinfo and new_strinfo.
26348         (zero_length_string): Assert that chainsi contains full strings.
26349         Use zero_length_string_p.  Update call to new_strinfo.
26350         (adjust_related_strinfos): Update after above changes to strinfo.
26351         Copy full_string_p from origsi.
26352         (adjust_last_stmt): Use zero_length_string_p.
26353         (handle_builtin_strlen): Update after above changes to strinfo and
26354         new_strinfo.  Install the lhs as the string length if the previous
26355         entry didn't describe a full string.
26356         (handle_builtin_strchr): Update after above changes to strinfo
26357         and new_strinfo.
26358         (handle_builtin_strcpy): Likewise.
26359         (handle_builtin_strcat): Likewise.
26360         (handle_builtin_malloc): Likewise.
26361         (handle_pointer_plus): Likewise.
26362         (handle_builtin_memcpy): Likewise.  Track nonzero characters
26363         that aren't necessarily followed by a nul terminator.
26364         (handle_char_store): Likewise.
26366 2017-07-02  Richard Sandiford  <richard.sandiford@linaro.org>
26368         PR tree-optimization/80769
26369         * tree-ssa-strlen.c (strinfo): Document that "stmt" is also used
26370         for malloc and calloc.  Document the new invariant that all related
26371         strinfos have delayed lengths or none do.
26372         (verify_related_strinfos): Move earlier in file.
26373         (set_endptr_and_length): New function, split out from...
26374         (get_string_length): ...here.  Also set the lengths of related
26375         strinfos.
26376         (zero_length_string): Assert that chainsi has known (rather than
26377         delayed) lengths.
26378         (adjust_related_strinfos): Likewise.
26380 2017-07-02  Richard Sandiford  <richard.sandiford@linaro.org>
26382         PR tree-optimization/81136
26383         * tree-vect-data-refs.c (vect_update_misalignment_for_peel): Only
26384         assert that two references with the same misalignment have the same
26385         compile-time misalignment if those compile-time misalignments
26386         are known.
26388 2017-07-01  Andi Kleen  <ak@linux.intel.com>
26390         * print-tree.c (print_node): Print all attributes.
26392 2017-07-01  Jan Hubicka  <hubicka@ucw.cz>
26394         * cfg.c (scale_bbs_frequencies): New function.
26395         * cfg.h (scale_bbs_frequencies): Declare it.
26396         * cfgloopanal.c (single_likely_exit): Cleanup.
26397         * cfgloopmanip.c (scale_loop_frequencies): Take profile_probability
26398         as parameter.
26399         (scale_loop_profile): Likewise.
26400         (loop_version): Likewise.
26401         (create_empty_loop_on_edge): Update.
26402         * cfgloopmanip.h (scale_loop_frequencies, scale_loop_profile,
26403         scale_loop_frequencies, scale_loop_profile, loopify,
26404         loop_version): Update prototypes.
26405         * modulo-sched.c (sms_schedule): Update.
26406         * predict.c (unlikely_executed_edge_p): Also check probability.
26407         (probably_never_executed_edge_p): Fix typo.
26408         * tree-if-conv.c (version_loop_for_if_conversion): Update.
26409         * tree-parloops.c (gen_parallel_loop): Update.
26410         * tree-ssa-loop-ivcanon.c (try_peel_loop): Update.
26411         * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Update.
26412         * tree-ssa-loop-split.c (split_loop): Update.
26413         * tree-ssa-loop-unswitch.c (tree_unswitch_loop): Update.
26414         * tree-vect-loop-manip.c (vect_do_peeling): Update.
26415         (vect_loop_versioning): Update.
26416         * tree-vect-loop.c (scale_profile_for_vect_loop): Update.
26418 2017-07-01  Jan Hubicka  <hubicka@ucw.cz>
26420         * trans-mem.c (split_bb_make_tm_edge): Update profile.
26422 2017-07-01  Jan Hubicka  <hubicka@ucw.cz>
26424         * tree-if-conv.c (combine_blocks): Use make_single_succ_edge
26425         to keep profile consistent.
26427 2017-07-01  Jan Hubicka  <hubicka@ucw.cz>
26429         * cfgrtl.c (rtl_flow_call_edges_add): Update profile.
26430         * tree-cfg.c (gimple_flow_call_edges_add): Likewise.
26431         * profile-count.h (max_safe_multiplier): Make unsigned.
26432         (profile_count::guessed_zero): New.
26434 2017-07-01  Jan Hubicka  <hubicka@ucw.cz>
26436         * bb-reorder.c (fix_up_crossing_landing_pad,
26437         fix_crossing_conditional_branches): Use make_single_succ_edge
26438         to keep profile consistent.
26440 2017-07-01  Jan Hubicka  <hubicka@ucw.cz>
26442         * tree-vect-loop.c (optimize_mask_stores): Use make_single_succ_edge
26443         to update profile.
26445 2017-07-01  Jakub Jelinek  <jakub@redhat.com>
26447         PR sanitizer/81262
26448         * bb-reorder.c (fix_up_fall_thru_edges): Move variable declarations to
26449         the right scopes, make sure cond_jump isn't preserved between multiple
26450         iterations.  Search for fallthru edge whenever there are 3+ edges and
26451         use find_fallthru_edge for it.
26453 2017-06-29  Jan Hubicka  <hubicka@ucw.cz>
26455         Patch by Alexander Monakov <amonakov@ispras.ru>
26456         * sel-sched-ir.c (compute_succs_info): Handle uninitialized
26457         probabilities consistently.
26459 2017-06-29  Jan Hubicka  <hubicka@ucw.cz>
26461         * pa.c (pa_expand_compare_and_swap_loop): Update call of
26462         emit_cmp_and_jump_insns.
26464 2017-06-29  Jan Hubicka  <hubicka@ucw.cz>
26466         PR ipa/81261
26467         * tree-inline.c (expand_call_inline): Combine profile statuses.
26469 2017-06-30  Andrew Pinski  <apinski@cavium.com>
26471         * tree-if-conv.c (predicate_scalar_phi): Update new_stmt if
26472         fold_stmt returned true.
26474 2017-06-30  Nathan Sidwell  <nathan@acm.org>
26476         * ggc.h (empty_string): Delete.
26477         * cfgexpand.c (expand_asm_stmt): Use plain "".
26478         * optabs.c (expand_asm_memory_barrier): Likewise.
26479         * stringpool.c (empty_string): Delete.
26480         (digit_vector, digit_string): Delete.
26481         (ggc_alloc_string): Use plain "", don't optimize single digit
26482         strings.  Use ggc_alloc_atomic.
26484 2017-06-30  Richard Earnshaw  <rearnsha@arm.com>
26486         * rtlanal.c (insn_rtx_cost): If a parallel contains exactly one
26487         comparison set and one other set, use the cost of the non-comparison
26488         set.
26490 2017-06-30  Nathan Sidwell  <nathan@acm.org>
26492         * ggc.h: Replace all 'static inline' with plain 'inline'.  Fix
26493         some formatting.
26495 2017-06-30  Peter Bergner  <bergner@vnet.ibm.com>
26497         * tree-cfg.c (group_case_labels_stmt): Merge scanning and compressing
26498         loops.  Remove now unneeded calls to gimple_switch_set_label() that
26499         just set removed labels to NULL_TREE.
26501 2017-06-30  Aldy Hernandez  <aldyh@redhat.com>
26503         * tree-ssanames.c (set_range_info_raw): Abstract from ...
26504         (set_range_info): ...here.  Only call set_range_info_raw if domain
26505         is useful.
26506         (set_nonzero_bits): Call set_range_info_raw.
26507         * tree-ssanames.h (set_range_info_raw): New.
26509 2017-06-30  Jakub Jelinek  <jakub@redhat.com>
26511         PR target/81225
26512         * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): For V8FI,
26513         V16FI and VI8F_256 iterators, use <store_mask_predicate> instead
26514         of nonimmediate_operand and <store_mask_constraint> instead of m
26515         for the input operand.  For V8FI iterator, always split if input
26516         is a MEM.  For V16FI and V8SF_256 iterators, don't test if both
26517         operands are MEM if <mask_applied>.  For VI4F_256 iterator, use
26518         <store_mask_predicate> instead of register_operand and
26519         <store_mask_constraint> instead of v for the input operand.  Make
26520         sure both operands aren't MEMs for if not <mask_applied>.
26522 2017-06-30  Sylvestre Ledru  <sylvestre@debian.org>
26524         * lto-wrapper.c (copy_file) Close both file descriptors before
26525         exiting normally.
26527 2017-06-30  Martin Liska  <mliska@suse.cz>
26529         PR ipa/81214
26530         * multiple_target.c (create_dispatcher_calls): Make ifunc
26531         also for function that don't have calls or are not referenced.
26533 2017-06-30  Richard Biener  <rguenther@suse.de>
26535         * tree-vect-slp.c (vect_slp_analyze_node_operations): Only
26536         analyze the first scalar stmt.  Move vector type computation
26537         for the BB case here from ...
26538         * tree-vect-stmts.c (vect_analyze_stmt): ... here.  Guard
26539         live operation processing in the SLP case properly.
26541 2017-06-30  Richard Biener  <rguenther@suse.de>
26543         * graph.c (draw_cfg_node_succ_edges): Fix broken dot syntax.
26545 2017-06-30  Martin Liska  <mliska@suse.cz>
26547         PR sanitizer/81021
26548         * tree-eh.c (lower_resx): Call BUILT_IN_ASAN_HANDLE_NO_RETURN
26549         before BUILT_IN_UNWIND_RESUME when ASAN is used.
26551 2017-06-30  Yvan Roux  <yvan.roux@linaro.org>
26553         * doc/invoke.texi (AArch64): Add missing options and remove redundant
26554         ones.
26556 2017-06-30  Richard Biener  <rguenther@suse.de>
26558         PR tree-optimization/81249
26559         * tree-vect-loop.c (vect_create_epilog_for_reduction): Convert
26560         condition reduction result to original scalar type.
26562 2017-06-30  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
26564         * profile-count.h (enum profile_quality): Fix typos and whitespace
26565         issues.
26567 2017-06-30  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
26569         * config/s390/s390.c (s390_expand_setmem): Adjust to the new data
26570         type for branch probabilities.
26572 2017-06-29  Julian Brown  <julian@codesourcery.com>
26573             Naveen H.S  <Naveen.Hurugalawadi@cavium.com>
26575         * config/aarch64/aarch64-fusion-pairs.def: Add ALU_BRANCH entry.
26576         * config/aarch64/aarch64.c (AARCH64_FUSE_ALU_BRANCH): New fusion type.
26577         (thunderx2t99_tunings): Set AARCH64_FUSE_ALU_BRANCH flag.
26578         (aarch_macro_fusion_pair_p): Add support for AARCH64_FUSE_ALU_BRANCH.
26580 2017-06-29  Naveen H.S  <Naveen.Hurugalawadi@cavium.com>
26582         * config/aarch64/aarch64.c (aarch_macro_fusion_pair_p): Push the
26583         check for CC usage into AARCH64_FUSE_CMP_BRANCH.
26584         * config/i386/i386.c (ix86_macro_fusion_pair_p): Push the check for
26585         CC usage from generic code to here.
26586         * sched-deps.c (sched_macro_fuse_insns): Move the condition for
26587         CC usage into the target macros.
26589 2017-06-29  Maya Rashish  <coypu@sdf.org>
26591         * config/netbsd.h (NETBSD_LIB_SPEC): Add -lc when creating shared
26592         objects.
26594 2017-06-29  Jan Hubicka  <hubicka@ucw.cz>
26596         * arm/arm-builtins.c: Include profile-count.h
26597         * except.c (sjlj_emit_function_enter): Use
26598         profile_probability::unlikely.
26600 2017-06-29  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
26602         * config/rs6000/rs6000.c (toc_relative_expr_p): Make tocrel_base
26603         and tocrel_offset be pointer args rather than implicitly using
26604         static versions.
26605         (legitimate_constant_pool_address_p, rs6000_emit_move,
26606         const_load_sequence_p, adjust_vperm): Add local tocrel_base and
26607         tocrel_offset and use in toc_relative_expr_p call.
26608         (print_operand, print_operand_address): Use static tocrel_base_oac
26609         and tocrel_offset_oac.
26610         (rs6000_output_addr_const_extra): Use static tocrel_base_oac and
26611         tocrel_offset_oac.
26613 2017-06-29  Maya Rashish  <coypu@sdf.org>
26615         * config/vax/builtins.md (ffssi2_internal): Correct constraint.
26617 2017-06-29  Eric Botcazou  <ebotcazou@adacore.com>
26619         * expr.c (expand_expr) <normal_inner_ref>: When testing for unaligned
26620         objects, take into account only the alignment of 'op0' and 'mode1' if
26621         'op0' is a MEM.
26623 2017-06-29  Steve Ellcey  <sellcey@cavium.com>
26625         * ccmp.c (ccmp_tree_comparison_p): New function.
26626         (ccmp_candidate_p): Update to use above function.
26627         (get_compare_parts): New function.
26628         (expand_ccmp_next): Update to use new functions.
26629         (expand_ccmp_expr_1): Take tree arg instead of gimple, update to use
26630         new functions.
26631         (expand_ccmp_expr): Pass tree instead of gimple to expand_ccmp_expr_1,
26632         take mode as argument.
26633         * ccmp.h (expand_ccmp_expr): Add mode as argument.
26634         * expr.c (expand_expr_real_1): Pass mode as argument.
26636 2017-06-29  Segher Boessenkool  <segher@kernel.crashing.org>
26638         * combine.c (combine_instructions): Print insns to dump_file, together
26639         with their costs.
26641 2017-06-29  Jan Hubicka  <hubicka@ucw.cz>
26643         * asan.c (asan_emit_stack_protection): Update.
26644         (create_cond_insert_point): Update.
26645         * auto-profile.c (afdo_propagate_circuit): Update.
26646         * basic-block.h (struct edge_def): Turn probability to
26647         profile_probability.
26648         (EDGE_FREQUENCY): Update.
26649         * bb-reorder.c (find_traces_1_round): Update.
26650         (better_edge_p): Update.
26651         (sanitize_hot_paths): Update.
26652         * cfg.c (unchecked_make_edge): Initialize probability to uninitialized.
26653         (make_single_succ_edge): Update.
26654         (check_bb_profile): Update.
26655         (dump_edge_info): Update.
26656         (update_bb_profile_for_threading): Update.
26657         * cfganal.c (connect_infinite_loops_to_exit): Initialize new edge
26658         probabilitycount to 0.
26659         * cfgbuild.c (compute_outgoing_frequencies): Update.
26660         * cfgcleanup.c (try_forward_edges): Update.
26661         (outgoing_edges_match): Update.
26662         (try_crossjump_to_edge): Update.
26663         * cfgexpand.c (expand_gimple_cond): Update make_single_succ_edge.
26664         (expand_gimple_tailcall): Update.
26665         (construct_init_block): Use make_single_succ_edge.
26666         (construct_exit_block): Use make_single_succ_edge.
26667         * cfghooks.c (verify_flow_info): Update.
26668         (redirect_edge_succ_nodup): Update.
26669         (split_edge): Update.
26670         (account_profile_record): Update.
26671         * cfgloopanal.c (single_likely_exit): Update.
26672         * cfgloopmanip.c (scale_loop_profile): Update.
26673         (set_zero_probability): Remove.
26674         (duplicate_loop_to_header_edge): Update.
26675         * cfgloopmanip.h (loop_version): Update prototype.
26676         * cfgrtl.c (try_redirect_by_replacing_jump): Update.
26677         (force_nonfallthru_and_redirect): Update.
26678         (update_br_prob_note): Update.
26679         (rtl_verify_edges): Update.
26680         (purge_dead_edges): Update.
26681         (rtl_lv_add_condition_to_bb): Update.
26682         * cgraph.c: (cgraph_edge::redirect_call_stmt_to_calle): Update.
26683         * cgraphunit.c (init_lowered_empty_function): Update.
26684         (cgraph_node::expand_thunk): Update.
26685         * cilk-common.c: Include profile-count.h
26686         * dojump.c (inv): Remove.
26687         (jumpifnot): Update.
26688         (jumpifnot_1): Update.
26689         (do_jump_1): Update.
26690         (do_jump): Update.
26691         (do_jump_by_parts_greater_rtx): Update.
26692         (do_compare_rtx_and_jump): Update.
26693         * dojump.h (jumpifnot, jumpifnot_1, jumpif_1, jumpif, do_jump,
26694         do_jump_1. do_compare_rtx_and_jump): Update prototype.
26695         * dwarf2cfi.c: Include profile-count.h
26696         * except.c (dw2_build_landing_pads): Use make_single_succ_edge.
26697         (sjlj_emit_dispatch_table): Likewise.
26698         * explow.c: Include profile-count.h
26699         * expmed.c (emit_store_flag_force): Update.
26700         (do_cmp_and_jump): Update.
26701         * expr.c (compare_by_pieces_d::generate): Update.
26702         (compare_by_pieces_d::finish_mode): Update.
26703         (emit_block_move_via_loop): Update.
26704         (store_expr_with_bounds): Update.
26705         (store_constructor): Update.
26706         (expand_expr_real_2): Update.
26707         (expand_expr_real_1): Update.
26708         * expr.h (try_casesi, try_tablejump): Update prototypes.
26709         * gimple-pretty-print.c (dump_probability): Update.
26710         (dump_profile): New.
26711         (dump_gimple_label): Update.
26712         (dump_gimple_bb_header): Update.
26713         * graph.c (draw_cfg_node_succ_edges): Update.
26714         * hsa-gen.c (convert_switch_statements): Update.
26715         * ifcvt.c (cheap_bb_rtx_cost_p): Update.
26716         (find_if_case_1): Update.
26717         (find_if_case_2): Update.
26718         * internal-fn.c (expand_arith_overflow_result_store): Update.
26719         (expand_addsub_overflow): Update.
26720         (expand_neg_overflow): Update.
26721         (expand_mul_overflow): Update.
26722         (expand_vector_ubsan_overflow): Update.
26723         * ipa-cp.c (good_cloning_opportunity_p): Update.
26724         * ipa-split.c (split_function): Use make_single_succ_edge.
26725         * ipa-utils.c (ipa_merge_profiles): Update.
26726         * loop-doloop.c (add_test): Update.
26727         (doloop_modify): Update.
26728         * loop-unroll.c (compare_and_jump_seq): Update.
26729         (unroll_loop_runtime_iterations): Update.
26730         * lra-constraints.c (lra_inheritance): Update.
26731         * lto-streamer-in.c (input_cfg): Update.
26732         * lto-streamer-out.c (output_cfg): Update.
26733         * mcf.c (adjust_cfg_counts): Update.
26734         * modulo-sched.c (sms_schedule): Update.
26735         * omp-expand.c (expand_omp_for_init_counts): Update.
26736         (extract_omp_for_update_vars): Update.
26737         (expand_omp_ordered_sink): Update.
26738         (expand_omp_for_ordered_loops): Update.
26739         (expand_omp_for_generic): Update.
26740         (expand_omp_for_static_nochunk): Update.
26741         (expand_omp_for_static_chunk): Update.
26742         (expand_cilk_for): Update.
26743         (expand_omp_simd): Update.
26744         (expand_omp_taskloop_for_outer): Update.
26745         (expand_omp_taskloop_for_inner): Update.
26746         * omp-simd-clone.c (simd_clone_adjust): Update.
26747         * optabs.c (expand_doubleword_shift): Update.
26748         (expand_abs): Update.
26749         (emit_cmp_and_jump_insn_1): Update.
26750         (expand_compare_and_swap_loop): Update.
26751         * optabs.h (emit_cmp_and_jump_insns): Update prototype.
26752         * predict.c (predictable_edge_p): Update.
26753         (edge_probability_reliable_p): Update.
26754         (set_even_probabilities): Update.
26755         (combine_predictions_for_insn): Update.
26756         (combine_predictions_for_bb): Update.
26757         (propagate_freq): Update.
26758         (estimate_bb_frequencies): Update.
26759         (force_edge_cold): Update.
26760         * profile-count.c (profile_count::dump): Add missing space into dump.
26761         (profile_count::debug): Add newline.
26762         (profile_count::differs_from_p): Explicitly convert to unsigned.
26763         (profile_count::stream_in): Update.
26764         (profile_probability::dump): New member function.
26765         (profile_probability::debug): New member function.
26766         (profile_probability::differs_from_p): New member function.
26767         (profile_probability::differs_lot_from_p): New member function.
26768         (profile_probability::stream_in): New member function.
26769         (profile_probability::stream_out): New member function.
26770         * profile-count.h (profile_count_quality): Rename to ...
26771         (profile_quality): ... this one.
26772         (profile_probability): New.
26773         (profile_count): Update.
26774         * profile.c (compute_branch_probabilities): Update.
26775         * recog.c (peep2_attempt): Update.
26776         * sched-ebb.c (schedule_ebbs): Update.
26777         * sched-rgn.c (find_single_block_region): Update.
26778         (compute_dom_prob_ps): Update.
26779         (schedule_region): Update.
26780         * sel-sched-ir.c (compute_succs_info): Update.
26781         * stmt.c (struct case_node): Update.
26782         (do_jump_if_equal): Update.
26783         (get_outgoing_edge_probs): Update.
26784         (conditional_probability): Update.
26785         (emit_case_dispatch_table): Update.
26786         (expand_case): Update.
26787         (expand_sjlj_dispatch_table): Update.
26788         (emit_case_nodes): Update.
26789         * targhooks.c: Update.
26790         * tracer.c (better_p): Update.
26791         (find_best_successor): Update.
26792         * trans-mem.c (expand_transaction): Update.
26793         * tree-call-cdce.c: Update.
26794         * tree-cfg.c (gimple_split_edge): Upate.
26795         (move_sese_region_to_fn): Upate.
26796         * tree-cfgcleanup.c (cleanup_control_expr_graph): Upate.
26797         * tree-eh.c (lower_resx): Upate.
26798         (cleanup_empty_eh_move_lp): Upate.
26799         * tree-if-conv.c (version_loop_for_if_conversion): Update.
26800         * tree-inline.c (copy_edges_for_bb): Update.
26801         (copy_cfg_body): Update.
26802         * tree-parloops.c (gen_parallel_loop): Update.
26803         * tree-profile.c (gimple_gen_ic_func_profiler): Update.
26804         (gimple_gen_time_profiler): Update.
26805         * tree-ssa-dce.c (remove_dead_stmt): Update.
26806         * tree-ssa-ifcombine.c (update_profile_after_ifcombine): Update.
26807         * tree-ssa-loop-im.c (execute_sm_if_changed): Update.
26808         * tree-ssa-loop-ivcanon.c (remove_exits_and_undefined_stmts): Update.
26809         (unloop_loops): Update.
26810         (try_peel_loop): Update.
26811         * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Update.
26812         * tree-ssa-loop-split.c (connect_loops): Update.
26813         (split_loop): Update.
26814         * tree-ssa-loop-unswitch.c (tree_unswitch_loop): Update.
26815         (hoist_guard): Update.
26816         * tree-ssa-phionlycprop.c (propagate_rhs_into_lhs): Update.
26817         * tree-ssa-phiopt.c (replace_phi_edge_with_variable): Update.
26818         (value_replacement): Update.
26819         * tree-ssa-reassoc.c (branch_fixup): Update.
26820         * tree-ssa-tail-merge.c (replace_block_by): Update.
26821         * tree-ssa-threadupdate.c (remove_ctrl_stmt_and_useless_edges): Update.
26822         (create_edge_and_update_destination_phis): Update.
26823         (compute_path_counts): Update.
26824         (recompute_probabilities): Update.
26825         (update_joiner_offpath_counts): Update.
26826         (freqs_to_counts_path): Update.
26827         (duplicate_thread_path): Update.
26828         * tree-switch-conversion.c (hoist_edge_and_branch_if_true): Update.
26829         (struct switch_conv_info): Update.
26830         (gen_inbound_check): Update.
26831         * tree-vect-loop-manip.c (slpeel_add_loop_guard): Update.
26832         (vect_do_peeling): Update.
26833         (vect_loop_versioning): Update.
26834         * tree-vect-loop.c (scale_profile_for_vect_loop): Update.
26835         (optimize_mask_stores): Update.
26836         * ubsan.c (ubsan_expand_null_ifn): Update.
26837         * value-prof.c (gimple_divmod_fixed_value): Update.
26838         (gimple_divmod_fixed_value_transform): Update.
26839         (gimple_mod_pow2): Update.
26840         (gimple_mod_pow2_value_transform): Update.
26841         (gimple_mod_subtract): Update.
26842         (gimple_mod_subtract_transform): Update.
26843         (gimple_ic): Update.
26844         (gimple_stringop_fixed_value): Update.
26845         (gimple_stringops_transform): Update.
26846         * value-prof.h: Update.
26848 2017-06-29  Carl Love  <cel@us.ibm.com>
26850         * config/rs6000/rs6000-c.c: Add support for built-in functions
26851         vector signed int vec_signed (vector float);
26852         vector signed long long vec_signed (vector double);
26853         vector signed int vec_signed2 (vector double, vector double);
26854         vector signed int vec_signede (vector double);
26855         vector signed int vec_signedo (vector double);
26856         * config/rs6000/rs6000.c (rs6000_generate_vsigned2_code): Add
26857         instruction generator.
26858         * config/rs6000/vsx.md (UNSPEC_VSX_XVCVSPSXWS, UNSPEC_VSX_XVCVSPSXDS,
26859         UNSPEC_VSX_VSIGNED2): Add UNSPECS.
26860         (vsx_xvcvspsxws, vsx_xvcvdpuxds_scale, vsx_xvcvspuxws, vsigned2_v2df):
26861         Add define_insn.
26862         (vsignedo_v2df, vsignede_v2df, vunsigned2_v2df, vunsignedo_v2df,
26863         vunsignede_v2df): Add define_expands.
26864         * config/rs6000/rs6000-builtin.def (VEC_SIGNED, VEC_UNSIGNED,
26865         VEC_SIGNED2, VEC_UNSIGNED2, VEC_SIGNEDE, VEC_UNSIGNEDE, VEC_SIGNEDO,
26866         VEC_UNSIGNEDO): Add definitions.
26867         * config/vsx.md (UNSPEC_VSX_XVCVSPSXWS, UNSPEC_VSX_XVCVSPSXDS,
26868         UNSPEC_VSX_VSIGNED2): Add UNSPECs.
26869         (vsx_xvcvspsxws, vsx_xvcvspuxws): Add define_insn.
26870         (vsigned2_v2df, vsigendo_v2df, vsignede_v2df,
26871         vunsigned2_v2df, vunsignedo_v2df, vunsignede_v2df): Add define_expands.
26872         * config/rs6000/altivec.h (vec_signed, vec_signed2,
26873         vec_signede and vec_signedo, vec_unsigned, vec_unsigned2,
26874         vec_unsignede, vec_unsignedo): Add builtin defines.
26875         * config/rs6000-protos.h (rs6000_generate_vsigned2_code): Add extern
26876         declaration.
26877         * doc/extend.texi: Update the built-in documentation file for the
26878         new built-in functions.
26880 2017-06-29  Richard Biener  <rguenther@suse.de>
26882         * tree-vect-loop.c (vect_analyze_scalar_cycles_1): Do not add
26883         reduction chains to LOOP_VINFO_REDUCTIONS.
26884         * tree-vect-slp.c (vect_analyze_slp): Continue looking for
26885         SLP reductions after processing reduction chains.
26887 2017-06-29  Nathan Sidwell  <nathan@acm.org>
26889         * builtins.c (fold_builtin_FUNCTION): Use
26890         lang_hooks.decl_printable_name.
26892 2017-06-29  Peter Bergner  <bergner@vnet.ibm.com>
26894         PR middle-end/81194
26895         * cfgexpand.c (expand_gimple_stmt_1): Handle switch statements
26896         with only one label.
26897         * stmt.c (expand_case): Assert NCASES is greater than one.
26899 2017-06-29  Richard Biener  <rguenther@suse.de>
26901         * tree-cfg.c (group_case_labels_stmt): Return whether we changed
26902         anything.
26903         (group_case_labels): Likewise.
26904         (find_taken_edge): Push sanity checking on val to workers...
26905         (find_taken_edge_cond_expr): ... here
26906         (find_taken_edge_switch_expr): ... and here, handle cases
26907         with just a default label.
26908         * tree-cfg.h (group_case_labels_stmt): Adjust prototype.
26909         (group_case_labels): Likewise.
26910         * tree-cfgcleanup.c (execute_cleanup_cfg_post_optimizing): When
26911         group_case_labels does anything cleanup the CFG again.
26913 2017-06-29  Bin Cheng  <bin.cheng@arm.com>
26915         PR tree-optimization/81196
26916         * tree-ssa-loop-niter.c (number_of_iterations_cond): Handle loop
26917         exit condition comparing two IVs.
26919 2017-06-29  Richard Earnshaw  <rearnsha@arm.com>
26921         * config/arm/parsecpu.awk (gen_comm_data): Add initializer for
26922         profile to the dummy entry at the end of the list of architectures.
26923         * config/arm/arm-cpu-cdata.h: Regenerated.
26925 2017-06-29  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
26926             Michael Collison <michael.collison@arm.com>
26928         PR target/70119
26929         * config/aarch64/aarch64.md (*aarch64_<optab>_reg_<mode>3_mask1):
26930         New pattern.
26931         (*aarch64_reg_<mode>3_neg_mask2): New pattern.
26932         (*aarch64_reg_<mode>3_minus_mask): New pattern.
26933         (*aarch64_<optab>_reg_di3_mask2): New pattern.
26934         * config/aarch64/aarch64.c (aarch64_rtx_costs): Account for cost
26935         of shift when the shift amount is masked with constant equal to
26936         the size of the mode.
26937         * config/aarch64/predicates.md (subreg_lowpart_operator): New
26938         predicate.
26940 2017-06-29  Martin Liska  <mliska@suse.cz>
26942         * config/i386/i386.opt: Change range from [1,5] to [0,5].
26944 2017-06-29  Yury Gribov  <tetra2005@gmail.com>
26946         PR bootstrap/80565
26947         * ipa-cp.c (allocate_and_init_ipcp_value): Add initialization
26948         code.
26949         * ipa-inline.h
26950         (edge_growth_cache_entry::edge_growth_cache_entry): New
26951         function.
26952         (reset_edge_growth_cache): Update to use constructor.
26954 2017-06-28  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
26956         * config/aarch64/aarch64.h (AARCH64_EXPAND_ALIGNMENT): New.
26957         (DATA_ALIGNMENT): Update to use AARCH64_EXPAND_ALIGNMENT.
26958         (LOCAL_ALIGNMENT): Update to use AARCH64_EXPAND_ALIGNMENT.
26960 2017-06-28  Sebastian Peryt  <sebastian.peryt@intel.com>
26962         * config/i386/avx512vlintrin.h (_mm256_permutexvar_epi64)
26963         (_mm256_permutexvar_epi32, _mm256_permutex_epi64): New intrinsics.
26965 2017-06-28  Szabolcs Nagy  <szabolcs.nagy@arm.com>
26967         * config.gcc (*-linux-musl*): Add t-musl tmake_file.
26968         (*-linux-uclibc*): Add t-uclibc tmake_file.
26969         * config/t-musl: New.
26970         * config/t-uclibc: New.
26972 2017-06-28  Richard Earnshaw  <rearnsha@arm.com>
26974         * config/arm/parsecpu.awk (profile): Parse new keyword in an arch
26975         context.
26976         (gen_comm_data): Emit architectural setting of arch_prof.
26977         * config/arm/arm-cpus.in (armv6-m, armv6s-m, armv7-a, armv7ve): Set the
26978         profile.
26979         (armv7-r, armv7-m, armv7e-m, armv8-a, armv8.1-a, armv8.2-a): Likewise.
26980         (armv8-m.base, armv8-m.main): Likewise.
26981         * arm-protos.h (arm_build_target): Add profile field.
26982         (arch_option): Likewise.
26983         * config/arm/arm.c (arm_configure_build_target): Copy the profile to
26984         the active target.
26985         * config/arm/arm.h (TARGET_ARM_ARCH_PROFILE): Use
26986         arm_active_target.profile.
26988 2017-06-28  Richard Biener  <rguenther@suse.de>
26990         PR middle-end/81227
26991         * fold-const.c (negate_expr_p): Use TYPE_UNSIGNED, not
26992         TYPE_OVERFLOW_WRAPS.
26993         * match.pd (negate_expr_p): Likewise.
26994         * tree-ssa-reassoc.c (optimize_range_tests_diff): Use
26995         fold_build2, not fold_binary.
26997 2017-06-28  Wilco Dijkstra  <wdijkstr@arm.com>
26999         * config/aarch64/aarch64 (aarch64_expand_mov_immediate):
27000         Convert memory address to Pmode.
27001         (aarch64_print_operand): Assert MEM operands are always Pmode.
27003 2017-06-28  Wilco Dijkstra  <wdijkstr@arm.com>
27005         PR target/79665
27006         * config/arm/aarch-common.c (arm_no_early_alu_shift_dep):
27007         Remove redundant if.
27008         (aarch_forward_to_shift_is_not_shifted_reg): Remove.
27009         * config/arm/aarch-common-protos.h
27010         (aarch_forward_to_shift_is_not_shifted_re): Remove.
27011         * config/arm/cortex-a53.md: Use arm_no_early_alu_shift_dep in bypass.
27013 2017-06-28  Michael Meissner  <meissner@linux.vnet.ibm.com>
27015         PR ipa/81238
27016         * multiple_target.c (create_dispatcher_calls): Set the default
27017         clone to be static, not public.
27019 2017-06-28  Richard Biener  <rguenther@suse.de>
27021         * tree-vect-loop.c (vectorizable_reduction): Move special
27022         cond reduction IV var creation ...
27023         (vect_create_epilog_for_reduction): ... here.  Remove induction_index
27024         parameter.  Use STMT_VINFO_VECTYPE.
27025         * tree-vect-slp.c (vect_get_constant_vectors): Properly reset
27026         constant_p.
27028 2017-06-28  Martin Liska  <mliska@suse.cz>
27030         PR ipa/81128
27031         * ipa-visibility.c (non_local_p): Handle visibility.
27033 2017-06-28  Martin Liska  <mliska@suse.cz>
27035         PR driver/79659
27036         * common.opt: Add IntegerRange to various options.
27037         * opt-functions.awk (integer_range_info): New function.
27038         * optc-gen.awk: Add integer_range_info to cl_options struct.
27039         * opts-common.c (decode_cmdline_option): Handle
27040         CL_ERR_INT_RANGE_ARG.
27041         (cmdline_handle_error): Likewise.
27042         * opts.c (print_filtered_help): Show valid interval in
27043         when --help is provided.
27044         * opts.h (struct cl_option): Add range_min and range_max fields.
27045         * config/i386/i386.opt: Add IntegerRange for -mbranch-cost.
27047 2017-06-28  Marc Glisse  <marc.glisse@inria.fr>
27049         * match.pd ((X & ~Y) | (~X & Y)): Generalize to + and ^.
27050         (x * C EQ/NE y * C): New transformation.
27052 2017-06-28  Christophe Lyon  <christophe.lyon@linaro.org>
27054         * genmultilib (combination_space): Accept '+' in option names.
27056 2017-06-28  Martin Liska  <mliska@suse.cz>
27058         PR sanitizer/81224
27059         * asan.c (instrument_derefs): Bail out inner references
27060         that are hard register variables.
27062 2017-06-28  Jakub Jelinek  <jakub@redhat.com>
27064         PR target/81175
27065         * config/i386/i386.c (ix86_init_mmx_sse_builtins): Use def_builtin
27066         rather than def_builtin_pure for __builtin_ia32_gatherpf*.
27068 2017-06-28  Richard Biener  <rguenther@suse.de>
27070         * tree-vectorizer.h (vect_get_vec_defs): Remove.
27071         (vect_get_slp_defs): Adjust.
27072         * tree-vect-loop.c (get_initial_defs_for_reduction): Split
27073         out from ...
27074         * tree-vect-slp.c (vect_get_constant_vectors): ... here and
27075         simplify.
27076         * tree-vect-loop.c (vect_create_epilog_for_reduction): Use
27077         get_initial_defs_for_reduction instead of vect_get_vec_defs.
27078         (vectorizable_reduction): Adjust.
27079         * tree-vect-slp.c (vect_get_constant_vectors): Remove reduction
27080         handling.
27081         (vect_get_slp_defs): Likewise.
27082         * tree-vect-stmts.c (vect_get_vec_defs): Make static and adjust.
27083         (vectorizable_bswap): Adjust.
27084         (vectorizable_call): Likewise.
27085         (vectorizable_conversion): Likewise.
27086         (vectorizable_assignment): Likewise.
27087         (vectorizable_shift): Likewise.
27088         (vectorizable_operation): Likewise.
27089         (vectorizable_store): Likewise.
27090         (vectorizable_condition): Likewise.
27091         (vectorizable_comparison): Likewise.
27093 2017-06-28  Michael Collison  <michael.collison@arm.com>
27095         PR target/68535
27096         * config/arm/arm.c (gen_ldm_seq): Remove last unnecessary
27097         set of base_reg
27098         (arm_gen_movmemqi): Removed unused variable 'i'.
27099         Convert 'for' loop into 'while' loop.
27100         (arm_expand_prologue): Remove last unnecessary set of insn.
27101         (thumb_pop): Remove unused variable 'pushed_words'.
27102         (thumb_exit): Remove last unnecessary set of regs_to_pop.
27104 2017-06-28  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
27106         * config/s390/predicates.md: Use s390_rel_address_ok_p.
27107         * config/s390/s390-protos.h: Add prototype of
27108         s390_rel_address_ok_p.
27109         * config/s390/s390.c (s390_got_symbol): New function.
27110         (s390_rel_address_ok_p): New function.
27111         (legitimize_pic_address): Use s390_rel_address_ok_p.
27112         (s390_load_got): Use s390_got_symbol.
27113         (s390_option_override): Issue error if
27114         -mno-pic-data-is-text-relative is used without -fpic/-fPIC.
27115         * config/s390/s390.h (TARGET_DEFAULT_PIC_DATA_IS_TEXT_RELATIVE):
27116         New macro.
27117         * config/s390/s390.opt: New option mpic-data-is-text-relative.
27119 2017-06-27  Andrew Pinski  <apinski@cavium.com>
27121         * match.pd (X >/>=/</<= 0 ? 1.0 : -1.0): New patterns.
27122         (X * copysign (1.0, X)): New pattern.
27123         (X * copysign (1.0, -X)): New pattern.
27124         (copysign (-1.0, CST)): New pattern.
27126 2017-06-27  Joseph Myers  <joseph@codesourcery.com>
27128         * genmultilib (combination_space): Remove variable.
27129         Validate reuse rules against regular expression for any sequence
27130         of multilib options in any order.
27132 2017-06-27  Michael Collison  <michael.collison@arm.com>
27134         * config/aarch64/aarch64-simd.md (aarch64_combine<mode>): Directly
27135         call aarch64_split_simd_combine.
27136         * (aarch64_combine_internal<mode>): Delete pattern.
27137         * config/aarch64/aarch64.c (aarch64_split_simd_combine):
27138         Allow register and subreg operands.
27140 2017-06-27  Jerome Lambourg  <lambourg@adacore.com>
27142         * config/i386/vxworks.h (ASM_SPEC): Remove definition. No target
27143         specific need, just fallback on defaults.
27144         (ASM_OUTPUT_ALIGNED_BSS): Add #undef before #define.
27146 2017-06-27  Jerome Lambourg  <lambourg@adacore.com>
27147             Olivier Hainque  <hainque@adacore.com>
27149         * config/i386/vxworks.h (DBX_REGISTER_NUMBER): Pick distinct
27150         map for 64bits.
27151         (TARGET_OS_CPP_BUILTINS): builtin_define CPU to X86_64 for 64bit
27152         targets. Pick a default if no particular attempt applied.
27153         (STACK_CHECK_PROTECT): Double for 64bit targets, which have
27154         larger contexts.
27156 2017-06-27  Jerome Lambourg  <lambourg@adacore.com>
27158         * config.gcc (i*86-wrs-vxworks7): Handle new acceptable triplet.
27159         (x86_64-wrs-vxworks7): Likewise.
27161 2017-06-27  Marek Polacek  <polacek@redhat.com>
27163         PR sanitizer/81223
27164         * ubsan.c (instrument_null): Check get_base_address's result for null.
27166 2017-06-27  Marc Glisse  <marc.glisse@inria.fr>
27168         * match.pd ((A+-B)+(C-A), (A+B)-(A-C)): New transformations.
27170 2017-06-27  Marc Glisse  <marc.glisse@inria.fr>
27172         * builtin-types.def (BT_FENV_T_PTR, BT_CONST_FENV_T_PTR,
27173         BT_FEXCEPT_T_PTR, BT_CONST_FEXCEPT_T_PTR): New primitive types.
27174         (BT_FN_INT_FENV_T_PTR, BT_FN_INT_CONST_FENV_T_PTR,
27175         BT_FN_INT_FEXCEPT_T_PTR_INT, BT_FN_INT_CONST_FEXCEPT_T_PTR_INT):
27176         New function types.
27177         * builtins.def (BUILT_IN_FECLEAREXCEPT, BUILT_IN_FEGETENV,
27178         BUILT_IN_FEGETEXCEPTFLAG, BUILT_IN_FEGETROUND,
27179         BUILT_IN_FEHOLDEXCEPT, BUILT_IN_FERAISEEXCEPT,
27180         BUILT_IN_FESETENV, BUILT_IN_FESETEXCEPTFLAG,
27181         BUILT_IN_FESETROUND, BUILT_IN_FETESTEXCEPT,
27182         BUILT_IN_FEUPDATEENV): New builtins.
27183         * tree-core.h (TI_FENV_T_PTR_TYPE, TI_CONST_FENV_T_PTR_TYPE,
27184         TI_FEXCEPT_T_PTR_TYPE, TI_CONST_FEXCEPT_T_PTR_TYPE): New entries.
27185         * tree.h (fenv_t_ptr_type_node, const_fenv_t_ptr_type_node,
27186         fexcept_t_ptr_type_node, const_fexcept_t_ptr_type_node): New
27187         macros.
27188         (builtin_structptr_types): Adjust size.
27189         * tree.c (builtin_structptr_types): Add four entries.
27191 2017-06-27  Jerome Lambourg  <lambourg@adacore.com>
27192             Olivier Hainque  <hainque@adacore.com>
27194         * config/vxworks.h (VXWORKS_LIB_SPEC): Incorporate ...
27195         (TLS_SYM): New local macro, forcing reference to __tls__ on
27196         link command lines for VxWorks 7 RTPs, triggering initialization
27197         of tlsLib.
27198         (VXWORKS_HAVE_TLS): New macro. State whether the target VxWorks
27199         OS features TLS support, true for RTPs on VxWorks 7.
27200         * config/vxworks.c (vxworks_override_options): Setup emutls
27201         accordingly.
27203 2017-06-27  Jakub Jelinek  <jakub@redhat.com>
27205         * predict.c (test_prediction_value_range): Use -1U instead of -1
27206         to avoid narrowing conversion warning.
27207         * dumpfile.c (dump_options): Wrap all value into dump_flags_t cast
27208         to avoid narrowing conversion warning.
27209         * opt-functions.awk (var_ref): Return (unsigned short) -1 instead of
27210         -1.
27211         * optc-gen.awk (END): Expect (unsigned short) -1 instead of -1.
27213 2017-06-27  Jerome Lambourg  <lambourg@adacore.com>
27215         * config/vxworks.h (VXWORKS_LIBS_RTP): Alternative definition for
27216         64bit configurations.
27217         (PTR_DIFF_TYPE): Alternative definition for TARGET_LP64.
27218         (SIZE_TYPE): Likewise.
27219         * config/vxworks.c (vxworks_emutls_var_fields): Use
27220         long_unsigned_type_node instead of unsigned_type_node as the offset
27221         field type, which is "pointer" mode in emutls.c.
27223 2017-06-27  Jakub Jelinek  <jakub@redhat.com>
27225         PR sanitizer/81209
27226         * ubsan.c (ubsan_encode_value): Initialize DECL_CONTEXT on var.
27228         PR middle-end/81207
27229         * gimple-fold.c (replace_call_with_call_and_fold): Handle
27230         gimple_vuse copying separately from gimple_vdef copying.
27232 2017-06-27  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
27234         * value-prof.c (free_hist): Remove call to memset and the enclosing if
27235         condition.
27237 2017-06-26  Jerome Lambourg  <lambourg@adacore.com>
27238             Olivier Hainque  <hainque@adacore.com>
27240         * config.gcc (*-*-vxworks*): Add TARGET_VXWORKS7=1 to tm_defines
27241         for all vxworks7 targets.
27242         * config/vxworks.h (TARGET_VXWORKS7): If not defined, define to 0.
27243         (VXWORKS_ADDITIONAL_CPP_SPEC): Alternative definition for VXWORKS7.
27244         (VXWORKS_LIBS_RTP, VXWORKS_LIBS_RTP_DIR): New macros, allowing
27245         variations for VX6/VX7 and 32/64bits later on in ...
27246         (VXWORKS_LIB_SPEC): Leverage new macros.
27247         (VXWORKS_OS_CPP_BUILTINS): Define _VSB_CONFIG_FILE for VXWORKS7,
27248         as well as _ALLOW_KEYWORD_MACROS when "inline" is not a keyword.
27250 2017-06-26  Jerome Lambourg  <lambourg@adacore.com>
27252         * config/vxworks.h (VXWORKS_OS_CPP_BUILTINS): builtin_define
27253         _VX_TOOL_FAMILY and _VX_TOOL to gnu.
27255 2017-06-26  Carl Love  <cel@us.ibm.com>
27257         * config/rs6000/rs6000-c.c: Add support for built-in functions
27258         vector bool char vec_reve (vector bool char);
27259         vector signed char vec_reve (vector signed char);
27260         vector unsigned char vec_reve (vector unsigned char);
27261         vector bool int vec_reve (vector bool int);
27262         vector signed int vec_reve (vector signed int);
27263         vector unsigned int vec_reve (vector unsigned int);
27264         vector bool long long vec_reve (vector bool long long);
27265         vector signed long long vec_reve (vector signed long long);
27266         vector unsigned long long vec_reve (vector unsigned long long);
27267         vector bool short vec_reve (vector bool short);
27268         vector signed short vec_reve (vector signed short);
27269         vector double vec_reve (vector double);
27270         vector float vec_reve (vector float);
27271         * config/rs6000/rs6000-builtin.def (VREVE_V2DI, VREVE_V4SI,
27272         VREVE_V8HI, VREVE_V16QI, VREVE_V2DF, VREVE_V4SF, VREVE): New builtin.
27273         * config/rs6000/altivec.md (UNSPEC_VREVEV): New UNSPEC.
27274         (altivec_vreve): New pattern.
27275         * config/rs6000/altivec.h (vec_reve): New define.
27276         * doc/extend.texi (vec_rev): Update the built-in documentation file
27277         for the new built-in functions.
27279 2016-06-26  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
27281         PR tree-optimization/71815
27282         * gimple-ssa-strength-reduction.c (uses_consumed_by_stmt): New
27283         function.
27284         (find_basis_for_candidate): Call uses_consumed_by_stmt rather than
27285         has_single_use.
27286         (slsr_process_phi): Likewise.
27287         (replace_uncond_cands_and_profitable_phis): Don't replace a
27288         multiply candidate with a stride of 1 (copy or cast).
27289         (phi_incr_cost): Call uses_consumed_by_stmt rather than
27290         has_single_use.
27291         (lowest_cost_path): Likewise.
27292         (total_savings): Likewise.
27294 2017-06-26  Richard Biener  <rguenther@suse.de>
27296         PR target/81175
27297         * config/i386/i386.c (ix86_init_mmx_sse_builtins):
27298         Use def_builtin_pure for all gather builtins.
27300 2017-06-26  Richard Biener  <rguenther@suse.de>
27302         PR tree-optimization/81203
27303         * tree-tailcall.c (find_tail_calls): Do not move stmts into
27304         non-dominating BBs.
27306 2017-06-26  Marek Polacek  <polacek@redhat.com>
27308         PR c/80116
27309         * doc/invoke.texi: Document -Wmultistatement-macros.
27311 2017-06-26  Christophe Lyon  <christophe.lyon@linaro.org>
27313         * doc/sourcebuild.texi (ARM-specific attributes): Document new
27314         arm_neon_ok_no_float_abi effective target.
27316 2017-06-26  Richard Biener  <rguenther@suse.de>
27318         PR tree-optimization/80928
27319         * cfghooks.c (duplicate_block): Do not copy BB_DUPLICATED flag.
27320         (copy_bbs): Set BB_DUPLICATED flag early.
27321         (execute_on_growing_pred): Do not execute for BB_DUPLICATED
27322         marked blocks.
27323         (execute_on_shrinking_pred): Likewise.
27324         * tree-ssa.c (ssa_redirect_edge): Do not look for PHI args in
27325         BB_DUPLICATED blocks.
27326         * tree-ssa-phionlycoprop.c (eliminate_degenerate_phis_1): Properly
27327         iterate over all PHIs considering removal of *gsi.
27329 2017-06-23  Jim Wilson  <jim.wilson@linaro.org>
27331         * doc/invoke.texi (AArch64 Options, -mtune): Re-add falkor and
27332         qdf24xx.
27334 2017-06-23  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
27336         * config/rs6000/rs6000-string.c: (expand_block_clear,
27337         do_load_for_compare, select_block_compare_mode,
27338         compute_current_alignment, expand_block_compare,
27339         expand_strncmp_align_check, expand_strn_compare,
27340         expand_block_move, rs6000_output_load_multiple)
27341         Move functions related to string/block move/compare
27342         to a separate file.
27343         * config/rs6000/rs6000.c: Move above functions to rs6000-string.c.
27344         * config/rs6000/rs6000-protos.h (rs6000_emit_dot_insn): Add prototype
27345         for this function which is now used in two files.
27346         * config/rs6000/t-rs6000: Add rule to compile rs6000-string.o.
27347         * config.gcc: Add rs6000-string.o to extra_objs for
27348         targets powerpc*-*-* and rs6000*-*-*.
27350 2017-06-23  Michael Meissner  <meissner@linux.vnet.ibm.com>
27352         PR target/80510
27353         * config/rs6000/rs6000.md (ALTIVEC_DFORM): Do not allow DImode in
27354         32-bit, since indexed is not valid for DImode.
27355         (mov<mode>_hardfloat32): Reorder ISA 2.07 load/stores before ISA
27356         3.0 d-form load/stores to be the same as mov<mode>_hardfloat64.
27357         (define_peephole2 for Altivec d-form load): Add 32-bit support.
27358         (define_peephole2 for Altivec d-form store): Likewise.
27360         PR ipa/81185
27361         * multiple_target.c (create_dispatcher_calls): Only create the
27362         dispatcher call if the function is the default clone of a
27363         versioned function.
27365 2017-06-23  Segher Boessenkool  <segher@kernel.crashing.org>
27367         PR middle-end/80902
27368         * builtins.c (expand_builtin_atomic_fetch_op): If emitting code after
27369         a call, force the call to not be a tail call.
27371 2017-06-23  Jeff Law  <law@redhat.com>
27373         * doc/contrib.texi: Add entry for Steven Pemberton's work on
27374         enquire.
27376 2017-06-23  Will Schmidt  <will_schmidt@vnet.ibm.com>
27378         * config/rs6000/rs6000.c: Add include of ssa-propagate.h for
27379         update_call_from_tree().  (rs6000_gimple_fold_builtin): Add
27380         handling for early expansion of vector shifts (sl,sr,sra,rl).
27381         (builtin_function_type): Add vector shift right instructions
27382         to the unsigned argument list.
27384 2017-06-23  Bernd Edlinger  <bernd.edlinger@hotmail.de>
27386         rtl-optimizatoin/79286
27387         * ira.c (update_equiv_regs): Revert to using may_trap_or_fault_p again.
27388         * rtlanal.c (rtx_addr_can_trap_p_1): SYMBOL_REF_FUNCTION_P can never
27389         trap.  PIC register plus a const unspec without offset can never trap.
27391 2017-06-23  Marc Glisse  <marc.glisse@inria.fr>
27393         * tree.h (builtin_structptr_type): New type.
27394         (builtin_structptr_types): Declare new array.
27395         * tree.c (builtin_structptr_types): New array.
27396         (free_lang_data, build_common_tree_nodes): Use it.
27398 2017-06-23  Jonathan Wakely  <jwakely@redhat.com>
27400         PR c++/81187
27401         * doc/invoke.texi (-Wnoexcept-type): Fix name of option, from
27402         -Wnoexcept.
27404 2017-06-22  Matt Turner  <mattst88@gmail.com>
27406         * config/i386/driver-i386.c (host_detect_local_cpu): Add Kaby
27407         Lake models to skylake case.  Assume skylake for unknown
27408         models with clflushopt.
27410 2017-06-22  Jeff Law  <law@redhat.com>
27412         * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Handle
27413         frame sizes that do not satisfy aarch64_uimm12_shift.
27415 2017-06-22  Jan Hubicka <hubicka@ucw.cz>
27417         * profile-count.h (apply_probability,
27418         apply_scale, probability_in): Fix checks for zero.
27420 2017-06-22  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
27422         * incpath.c (add_sysroot_to_chain): Allow for $SYSROOT prefix.
27423         * doc/cppdiropts.texi (-I @var{dir}): Document it.
27425 2016-06-22  Richard Biener  <rguenther@suse.de>
27427         * tree-vect-loop.c (vect_model_reduction_cost): Handle
27428         COND_REDUCTION and INTEGER_INDUC_COND_REDUCTION without
27429         REDUC_MAX_EXPR support.
27430         (vectorizable_reduction): Likewise.
27431         (vect_create_epilog_for_reduction): Likewise.
27433 2017-06-22  James Greenhalgh  <james.greenhalgh@arm.com>
27435         * match.pd (A / (1 << B) -> A >> B): New.
27436         * generic-match-head.c: Include optabs-tree.h.
27437         * gimple-match-head.c: Likewise.
27438         * optabs-tree.h (target_supports_op_p): New.
27439         * optabs-tree.c (target_supports_op_p): New.
27441 2017-06-22  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
27443         * configure.ac (gcc_cv_ld_static_dynamic): Also check stderr for
27444         $gcc_cv_ld --help output.
27445         (gcc_cv_ld_demangle): Likewise.
27446         (gcc_cv_ld_eh_frame_hdr): Likewise.
27447         (gcc_cv_ld_pie): Likewise.
27448         (gcc_cv_ld_as_needed): Likewise.  Prefer native forms unless $gnu_ld.
27449         (gcc_cv_ld_buildid): Likewise.
27450         (gcc_cv_ld_sysroot): Likewise.
27451         (ld_bndplt_support): Likewise.
27452         (ld_pushpopstate_support): Likewise.
27453         * configure: Regenerate.
27454         * config/sol2.h [!USE_GLD] (SYSROOT_SPEC): Define.
27456 2017-06-21  Jakub Jelinek  <jakub@redhat.com>
27458         PR target/81151
27459         * config/i386/sse.md (round<mode>2): Renumber match_dup and
27460         operands indexes to avoid gap between operands and match_dups.
27462 2017-06-21  Andrew Pinski  <apinski@cavium.com>
27464         * config/aarch64/aarch64-cost-tables.h (thunderx_extra_costs):
27465         Increment Arith_shift and Arith_shift_reg by 1.
27466         * config/aarch64/aarch64-tuning-flags.def (cheap_shift_extend):
27467         New tuning flag.
27468         * config/aarch64/aarch64.c (thunderx_tunings): Enable
27469         AARCH64_EXTRA_TUNE_CHEAP_SHIFT_EXTEND.
27470         (aarch64_strip_extend): Add new argument and test for it.
27471         (aarch64_cheap_mult_shift_p): New function.
27472         (aarch64_rtx_mult_cost): Call aarch64_cheap_mult_shift_p and don't
27473         add a cost if it is true.
27474         Update calls to aarch64_strip_extend.
27475         (aarch64_rtx_costs): Update calls to aarch64_strip_extend.
27477 2017-06-21  Andrew Pinski  <apinski@cavium.com>
27479         * config/aarch64/aarch64-cores.def (thunderxt88p1): Use thunderxt88
27480         tunings.
27481         (thunderxt88): Likewise.
27482         * config/aarch64/aarch64.c (thunderxt88_prefetch_tune): New variable.
27483         (thunderx_prefetch_tune): New variable.
27484         (thunderx2t99_prefetch_tune): Update for the correct values.
27485         (thunderxt88_tunings): New variable.
27486         (thunderx_tunings): Use thunderx_prefetch_tune instead of
27487         generic_prefetch_tune.
27488         (thunderx2t99_tunings): Use AUTOPREFETCHER_WEAK.
27490 2017-06-21  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
27492         * config/aarch64/atomics.md (aarch64_compare_and_swap<mode>_lse,
27493         SHORT): Relax operand 3 to aarch64_reg_or_zero and constraint to Z.
27494         (aarch64_compare_and_swap<mode>_lse, GPI): Likewise.
27495         (aarch64_atomic_cas<mode>, SHORT): Likewise for operand 2.
27496         (aarch64_atomic_cas<mode>, GPI): Likewise.
27498 2017-06-21  Martin Liska  <mliska@suse.cz>
27500         * gimplify.c (gimplify_label_expr): Insert GIMPLE_PREDICT
27501         statements on cold and hot labels.
27502         * predict.c (tree_estimate_probability_bb): Remove the
27503         prediction from this place.
27505 2017-06-21  Martin Liska  <mliska@suse.cz>
27507         PR tree-optimization/79489
27508         * gimplify.c (maybe_add_early_return_predict_stmt): New
27509         function.
27510         (gimplify_return_expr): Call the function.
27511         * predict.c (tree_estimate_probability_bb): Remove handling
27512         of early return.
27513         * predict.def: Update comment about early return predictor.
27514         * gimple-predict.h (is_gimple_predict): New function.
27515         * predict.def: Change default value of early return to 66.
27516         * tree-tailcall.c (find_tail_calls): Skip GIMPLE_PREDICT
27517         statements.
27518         * passes.def: Put pass_strip_predict_hints to the beginning of
27519         IPA passes.
27521 2017-06-21  Pierre-Marie de Rodat  <derodat@adacore.com>
27523         * dwarf2out.c (gen_decl_die): Remove the guard to skip file-scope
27524         FUNCTION_DECL declarations.
27525         (dwarf2out_early_global_decl): Remove the guard to skip FUNCTION_DECL
27526         declarations.
27527         (dwaf2out_decl): Likewise.
27528         * godump.c (go_early_global_decl): Skip call to the real debug hook
27529         for FUNCTION_DECL declarations.
27530         * passes.c (rest_of_decl_compilation): Skip call to the
27531         early_global_decl debug hook for FUNCTION_DECL declarations, unless
27532         -fdump-go-spec is passed.
27534 2017-06-21  Marc Glisse  <marc.glisse@inria.fr>
27536         * config/i386/i386.c (struct builtin_isa): New field pure_p.
27537         Reorder for compactness.
27538         (def_builtin, def_builtin2, ix86_add_new_builtins): Handle pure_p.
27539         (def_builtin_pure, def_builtin_pure2): New functions.
27540         (ix86_init_mmx_sse_builtins) [__builtin_ia32_stmxcsr]: Mark as pure.
27542 2017-06-21  Marc Glisse  <marc.glisse@inria.fr>
27544         * match.pd (nop_convert): New predicate.
27545         ((A +- CST1) +- CST2): Allow some NOP conversions.
27547 2017-06-21  Jakub Jelinek  <jakub@redhat.com>
27549         PR c++/81130
27550         * gimplify.c (omp_add_variable): Don't force GOVD_SEEN for types
27551         with ctors/dtors if GOVD_SHARED is set.
27553 2017-06-21  Wilco Dijkstra  <wdijkstr@arm.com>
27555         * config/aarch64/aarch64.md (movti_aarch64):
27556         Emit mov rather than orr.
27557         (movtf_aarch64): Likewise.
27558         * config/aarch64/aarch64-simd.md (aarch64_simd_mov):
27559         Emit mov rather than orr.
27561 2017-06-21  Wilco Dijkstra  <wdijkstr@arm.com>
27563         * config/aarch64/aarch64-simd.md (aarch64_simd_dup):
27564         Swap alternatives, make integer dup more expensive.
27566 2017-06-21  Wilco Dijkstra  <wdijkstr@arm.com>
27568         * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
27569         Return true for non-tls symbols.
27571 2017-06-21  James Greenhalgh  <james.greenhalgh@arm.com>
27573         * config/aarch64/aarch64-cores.def (cortex-a55): New.
27574         (cortex-a75): Likewise.
27575         (cortex-a75.cortex-a55): Likewise.
27576         * config/aarch64/aarch64-tune.md: Regenerate.
27577         * doc/invoke.texi (-mtune): Document new values for -mtune.
27579 2017-06-21  Tom de Vries  <tom@codesourcery.com>
27581         * doc/sourcebuild.texi (Add Options, Features for dg-add-options): Add
27582         stack_size feature.
27583         (Effective-Target Keywords, Other attributes): Suggest using
27584         dg-add-options stack_size feature to get stack limit in stack_size
27585         effective target documentation.
27587 2017-06-21  Julian Brown  <julian@codesourcery.com>
27588             Naveen H.S  <Naveen.Hurugalawadi@cavium.com>
27590         * config/aarch64/aarch64-simd.md (aarch64_crypto_pmulldi)
27591         (aarch64_crypto_pmullv2di): Change type attribute to crypto_pmull.
27592         * config/aarch64/thunderx2t99.md (thunderx2t99_pmull): New
27593         reservation.
27594         * config/arm/cortex-a53.md (cortex_a53_advsimd_type): Add crypto_pmull to
27595         attribute type list for neon_multiply.
27596         * config/arm/cortex-a57.md (cortex_a57_neon_type): Add crypto_pmull to
27597         attribute type list for neon_multiply.
27598         * config/arm/crypto.md (crypto_vmullp64): Change type to crypto_pmull.
27599         * config/arm/exynos-m1.md (exynos_m1_neon_type): Add crypto_pmull to
27600         attribute type list for neon_multiply.
27601         * config/arm/types.md (crypto_pmull): Add.
27602         * config/arm/xgene1.md (xgene1_neon_pmull): Add crypto_pmull to
27603         attribute type list.
27605 2017-06-20  Andreas Tobler  <andreast@gcc.gnu.org>
27607         * config.gcc (armv6*-*-freebsd*): Change the target_cpu_cname to
27608         arm1176jzf-s.
27610 2017-06-20  Jakub Jelinek  <jakub@redhat.com>
27612         * ira-costs.c (find_costs_and_classes): Initialize cost_classes later
27613         to make sure not to dereference a NULL cost_classes_ptr pointer.
27615 2017-06-20  Carl Love  <cel@us.ibm.com>
27617         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
27618         ALTIVEC_BUILTIN_VMULESW, ALTIVEC_BUILTIN_VMULEUW,
27619         ALTIVEC_BUILTIN_VMULOSW, ALTIVEC_BUILTIN_VMULOUW entries.
27620         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin,
27621         builtin_function_type): Add ALTIVEC_BUILTIN_* case statements.
27622         * config/rs6000/altivec.md (MVULEUW, VMULESW, VMULOUW,
27623         VMULOSW): New enum "unspec" values.
27624         (vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
27625         vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si,
27626         altivec_vmuleuw, altivec_vmulesw, altivec_vmulouw,
27627         altivec_vmulosw): New patterns.
27628         * config/rs6000/rs6000-builtin.def (VMLEUW, VMULESW, VMULOUW,
27629         VMULOSW): Add definitions.
27631 2017-06-20  Julia Koval  <julia.koval@intel.com>
27633         * config/i386/i386.c: Fix rounding expand for new pattern.
27634         * config/i386/subst.md: Fix pattern (parallel -> unspec).
27636 2017-06-20  James Greenhalgh  <james.greenhalgh@arm.com>
27638         * config/aarch64/aarch64-option-extensions.def (rcpc): New.
27639         * config/aarch64/aarch64.h (AARCH64_FL_RCPC): New.
27641 2017-06-20  James Greenhalgh  <james.greenhalgh@arm.com>
27643         * config/aarch64/aarch64-option-extensions.def (fp16): Fix expected
27644         feature string.
27646 2017-06-20  James Greenhalgh  <james.greenhalgh@arm.com>
27648         * config/aarch64/aarch64-cores.def: Rearrange to sort by
27649         architecture, then by implementer ID.
27650         * config/aarch64/aarch64-tune.md: Regenerate.
27652 2017-06-20  Richard Biener  <rguenther@suse.de>
27654         PR middle-end/81097
27655         * fold-const.c (split_tree): Fold to type before negating.
27657 2017-06-20  David Malcolm  <dmalcolm@redhat.com>
27659         * diagnostic-show-locus.c
27660         (selftest::test_fixit_deletion_affecting_newline): New function.
27661         (selftest::diagnostic_show_locus_c_tests): Call it.
27663 2017-06-20  Andreas Schwab  <schwab@suse.de>
27665         PR target/80970
27666         * config/m68k/m68k.md (bsetdreg, bchgdreg, bclrdreg): Use "=d"
27667         instead of "+d".
27669 2017-06-20  Prakhar Bahuguna  <prakhar.bahuguna@arm.com>
27671         * config/arm/arm-c.c (arm_cpu_builtins): New block to define
27672         __ARM_FEATURE_COPROC according to support.
27674 2017-06-20  Jakub Jelinek  <jakub@redhat.com>
27676         * tree-chkp.c (chkp_get_hard_register_var_fake_base_address):
27677         Rewritten to avoid overflow for > 32-bit pointers.
27679         PR sanitizer/81125
27680         * ubsan.h (ubsan_encode_value): Workaround buggy clang++ parser
27681         by removing enum keyword.
27682         (ubsan_type_descriptor): Likewise.  Formatting fix.
27684         PR target/81121
27685         * config/i386/i386.md (TARGET_USE_VECTOR_CONVERTS float si->{sf,df}
27686         splitter): Require TARGET_SSE2 in the condition.
27688 2017-06-20  Michael Meissner  <meissner@linux.vnet.ibm.com>
27690         PR target/79799
27691         * config/rs6000/rs6000.c (rs6000_expand_vector_init): Add support
27692         for doing vector set of SFmode on ISA 3.0.
27693         * config/rs6000/vsx.md (vsx_set_v4sf_p9): Likewise.
27694         (vsx_set_v4sf_p9_zero): Special case setting 0.0f to a V4SF
27695         element.
27696         (vsx_insert_extract_v4sf_p9): Add an optimization for inserting a
27697         SFmode value into a V4SF variable that was extracted from another
27698         V4SF variable without converting the element to double precision
27699         and back to single precision vector format.
27700         (vsx_insert_extract_v4sf_p9_2): Likewise.
27702 2017-06-19  Jakub Jelinek  <jakub@redhat.com>
27704         * tree-ssa-structalias.c (get_constraint_for_ptr_offset): Multiply
27705         in UWHI to avoid undefined overflow.
27707         PR sanitizer/81125
27708         * ubsan.h (enum ubsan_encode_value_phase): New.
27709         (ubsan_encode_value): Change second argument to
27710         enum ubsan_encode_value_phase with default value of
27711         UBSAN_ENCODE_VALUE_GENERIC.
27712         * ubsan.c (ubsan_encode_value): Change second argument to
27713         enum ubsan_encode_value_phase PHASE from bool IN_EXPAND_P,
27714         adjust uses, for UBSAN_ENCODE_VALUE_GENERIC use just
27715         create_tmp_var_raw instead of create_tmp_var and use a
27716         TARGET_EXPR.
27717         (ubsan_expand_bounds_ifn, ubsan_build_overflow_builtin,
27718         instrument_bool_enum_load, ubsan_instrument_float_cast): Adjust
27719         ubsan_encode_value callers.
27721         PR sanitizer/81111
27722         * ubsan.c (ubsan_encode_value): If current_function_decl is NULL,
27723         use create_tmp_var_raw instead of create_tmp_var, mark it addressable
27724         just by setting TREE_ADDRESSABLE on the result and use a TARGET_EXPR.
27726 2017-06-19  Richard Biener  <rguenther@suse.de>
27728         PR middle-end/81118
27729         * tree-cfgcleanup.c (cleanup_tree_cfg_noloop): Clear niter
27730         estimates if we changed anything.
27732 2017-06-19  Richard Biener  <rguenther@suse.de>
27734         PR tree-optimization/80887
27735         * tree-ssa-sccvn.c (mprts_hook_cnt): New global.
27736         (vn_lookup_simplify_result): Allow only mprts_hook_cnt succesful
27737         simplified lookups, then reset mprts_hook.
27738         (vn_nary_build_or_lookup_1): Set mprts_hook_cnt to 9 before
27739         simplifying.
27740         (try_to_simplify): Likewise.
27742 2017-06-19  Martin Liska  <mliska@suse.cz>
27744         PR sanitizer/80879
27745         * gimplify.c (gimplify_switch_expr):
27746         Initialize live_switch_vars for SWITCH_BODY == STATEMENT_LIST.
27748 2017-06-19  Martin Liska  <mliska@suse.cz>
27750         * doc/install.texi: Document that PGO runs in 4 stages.
27752 2017-06-19  Martin Liska  <mliska@suse.cz>
27754         PR ipa/80732
27755         * attribs.c (make_dispatcher_decl): Do not append '.ifunc'
27756         to dispatcher function name.
27757         * multiple_target.c (replace_function_decl): New function.
27758         (create_dispatcher_calls): Redirect both edges and references.
27760 2017-06-19  Jan Hubicka <hubicka@ucw.cz>
27762         * profile-count.c (profile_count::dump): Dump quality.
27763         (profile_count::differs_from_p): Update for unsigned val.
27764         * profile-count.h (profile_count_quality): New enum.
27765         (profile_count): Turn m_val to 62bit unsigned, add quality tracking.
27767 2017-06-19  Richard Biener  <rguenther@suse.de>
27769         * tree-ssa-loop-niter.h (estimate_numbers_of_iterations): Take
27770         struct function as arg.
27771         (estimate_numbers_of_iterations): Export overload with loop arg.
27772         (free_numbers_of_iterations_estimates_loop): Use an overload of
27773         free_numbers_of_iterations_estimates instead.
27774         * tree-cfg.c (remove_bb): Adjust.
27775         * tree-cfgcleanup.c (remove_forwarder_block_with_phi): Likewise.
27776         * tree-parloops.c (gen_parallel_loop): Likewise.
27777         * tree-ssa-loop-ivcanon.c (canonicalize_induction_variables):
27778         Likewise.
27779         (tree_unroll_loops_completely): Likewise.
27780         * tree-ssa-loop-niter.c (estimate_numbers_of_iterations_loop):
27781         Use an overload instead and export.
27782         (estimated_loop_iterations): Adjust.
27783         (max_loop_iterations): Likewise.
27784         (likely_max_loop_iterations): Likewise.
27785         (estimate_numbers_of_iterations): Take struct function as arg
27786         and adjust.
27787         (loop_exits_before_overflow): Adjust.
27788         (free_numbers_of_iterations_estimates_loop): Use an overload.
27789         * tree-vect-loop.c (vect_analyze_loop_form): Adjust.
27790         * tree-vectorizer.c (vect_free_loop_info_assumptions): Likewise.
27792 2017-06-19  Richard Biener  <rguenther@suse.de>
27794         PR ipa/81112
27795         * ipa-prop.c (find_constructor_constant_at_offset): Handle
27796         RANGE_EXPR conservatively.
27798 2017-06-16  Carl Love  <cel@us.ibm.com>
27800         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
27801         definitions for vec_float, vec_float2, vec_floato,
27802         vec_floate built-ins.
27803         * config/rs6000/vsx.md (define_c_enum "unspec"): Add RTL code
27804         for instructions vsx_xvcvsxws vsx_xvcvuxwsp, float2, floato and
27805         floate.
27806         * config/rs6000/rs6000-builtin.def (FLOAT2_V2DI, FLOATE_V2DF,
27807         FLOATE_2DI, FLOATO_V2DF, FLOATEE_V2DI, XVCVSXWSP_V4SF,
27808         UNS_FLOATO_V2DI, UNS_FLOATE_V2DI): Add definitions.
27809         * config/altivec.md (define_insn "p8_vmrgew_<mode>",
27810         define_mode_attr VF_sxddp): Add V4SF type to p8_vmrgew.
27811         * config/rs6000/altivec.h (vec_float, vec_float2, vec_floate,
27812         vec_floato): Add builtin defines.
27813         * doc/extend.texi (vec_float, vec_float2, vec_floate, vec_floato):
27814         Update the built-in documentation file for the new built-in
27815         functions.
27817 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27819         * config/arm/arm.opt (marm): Mark as the negative of of -mthumb.
27820         (mthumb): Mark as the negative of -marm.
27822 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27824         * doc/invoke.texi (ARM Options, -mcpu): Document supported
27825         extension options.
27826         (ARM Options, -mtune): Document that this accepts the same
27827         extension options as -mcpu.
27828         (ARM Options, -mfpu): Document addition of -mfpu=auto.
27830 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27832         * doc/invoke.texi (ARM Options, -march=): Document new syntax and
27833         permitted extensions.
27835 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27837         * config/arm/arm-cpus.in (armv7): Add extension +nofp.
27838         (armv7-r): Add aliases vfpv3xd and vfpv3-d16.
27839         (armv8-m.main): Add option +nodsp.
27840         * config/arm/arm-cpu-cdata.h: Regenerated.
27842 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27844         * config/arm/t-fuchsia: New file.
27845         * config.gcc (arm*-*-fuchsia*): Use it.
27847 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27849         * config/arm/t-symbian: Rewrite for new option infrastructure.
27851 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27853         * config/arm/t-phoenix (MULTILIB_REUSE): Clear variable.
27854         (MULTILIB_REQUIRED): Likewise.
27856 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27858         * config/arm/t-linux-eabi (MULTILIB_EXCEPTIONS): Set to empty.
27859         (MULTILIB_RESUE): Likewise.
27860         (MULTILIB_MATCHES): Likewise.
27861         (MULTLIB_REQUIRED): Likewise.
27863 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27865         * config/arm/t-rtems: Rewrite for new option framework.
27867 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27869         * config/arm/t-aprofile (v7_a_nosimd_variants, v7_a_simd_variants)
27870         (v7ve_nosimd_variatns, v7ve_vfpv3_simd_variants)
27871         (v7ve_vfpv4_simd_variants, v8_a_nosimd_variants, v8_a_simd_variants)
27872         (v8_1_a_simd_variants, v8_2_a_simd_variants): Move to ...
27873         * config/arm/t-multilib: ... here.
27874         (MULTILIB_OPTIONS): Add armv7 and armv7+fp architectures.
27875         (MULTILIB_MATCHES): Use armv7 libraries for armv7-r.  Also use for
27876         armv7-a and armv8*-a when A-profile libraries have not been built.
27877         * config/arm/t-rmprofile: Rewrite.
27879 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27881         * genmultilib (multilib_reuse): Allow an explicit period to be escaped
27882         with a backslash.  Remove the backslash after substituting unescaped
27883         periods.
27884         * doc/fragments.texi (MULTILIB_REUSE): Document it.
27886 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27888         * config.gcc: (arm*-*-*): When building a-profile libraries, force
27889         the driver to pass through the default setting of -mfloat-abi.
27890         * common/config/arm/arm-common.c (arm_target_thumb_only): Return -marm
27891         rather than NULL.
27892         * config/arm/t-multilib (MULTILIB_REUSE): Initialize to empty.
27893         (all_feat_combs): New rule.
27894         (MULTILIB_OPTIONS): Use explicit ARM and Thumb directories.  Rework
27895         default libraries.
27896         * config/arm/t-aprofile: Rewrite.
27898 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27900         * config/arm/arm.h (FPUTYPE_AUTO): Define.
27901         * config/arm/arm.c (arm_option_override): Use FPUTYPE_AUTO if the
27902         fpu is not specified by the user/command-line.
27903         * config/arm/bpabi.h (FPUTYPE_DEFAULT): Delete.
27904         * config/arm/netbsd-elf.h (FPUTYPE_DEFAULT): Delete.
27905         * config/arm/linux-elf.h (FPUTYPE_DEFAULT): Delete.
27906         * config/arm/vxworks.h (FPUTYPE_DEFAULT): Delete.
27907         * common/config/arm/arm-common.c (arm_canon_arch_option): Use
27908         FPUTYPE_AUTO insted of FPUTYPE_DEFAULT.
27910 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27912         * config/arm/elf.h (MULTILIB_DEFAULTS): Delete.
27913         * config/arm/t-arm-elf: Rewritten.
27915 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27917         * config/arm/arm.h (TARGET_HARD_FLOAT): Also check that we
27918         have some floating-point instructions.
27919         (TARGET_SOFT_FLOAT): Define as inverse of TARGET_HARD_FLOAT.
27920         (TARGET_MAYBE_HARD_FLOAT): New macro.
27921         * config/arm/arm-builtins.c (arm_init_builtins): Use
27922         TARGET_MAYBE_HARD_FLOAT.
27923         * config/arm/arm.c (arm_option_override): Use TARGET_HARD_FLOAT_ABI.
27925 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27927         * common/config/arm/arm-common.c: Define INCLUDE_LIST.
27928         (configargs.h): Include it.
27929         (arm_print_hint_for_fpu_option): New function.
27930         (arm_parse_fpu_option): New function.
27931         (candidate_extension): New class.
27932         (arm_canon_for_multilib): New function.
27933         * config/arm/arm.h (CANON_ARCH_SPEC_FUNCTION): New macro.
27934         (EXTRA_SPEC_FUNCTIONS): Add CANON_ARCH_SPEC_FUNCTION.
27935         (ARCH_CANONICAL_SPECS): New macro.
27936         (DRIVER_SELF_SPECS): Add ARCH_CANONICAL_SPECS.
27938 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27940         * config.gcc (arm*-*-*): Ensure both target_cpu_cname and with_cpu
27941         are set after handling multilib fragments.  Set target_cpu_default2
27942         from with_cpu.
27944 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27946         * config.gcc (arm*-*-fucshia*): Set target_cpu_cname to the real
27947         cpu name.
27948         (arm*-*-*): Set target_cpu_default2 to a quoted string.
27949         * config/arm/parsecpu.awk (check_cpu): Validate any extension
27950         options.
27951         (check_arch): Likewise.
27952         * config/arm/arm.c (arm_configure_build_target): Handle
27953         TARGET_CPU_DEFAULT being a string constant.  Scan any feature
27954         options in the default.
27956 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27958         * config/arm/arm-protos.h (cpu_arch_extension): Add field to record
27959         when an option is an alias of another.
27960         * config/arm/parsecpu.awk (optalias): New parser token.
27961         (gen_comm_data): Mark non-alias options as such.  Emit entries
27962         for extension aliases.
27963         * config/arm/arm-cpus.in (armv5e): Make vfpv2 an alias.
27964         (armv5te, armv5tej, armv6, armv6j, armv6k, armv6z): Likewise.
27965         (armv6kz, armv6zk, armv6t2): Likewise.
27966         (armv7): Make vfpv3-d16 an alias.
27967         (armv7-a): Make vfpv3-d16, neon and neon-vfpv3 aliases.  Sort in
27968         canonical order.
27969         (armv7ve): Make vfpv4-d16, neon-vfpv3 and neon-vfpv4 aliases.
27970         Sort in canonical order.
27971         (armv8-a): Sort in canonical order.
27972         (armv8.1-a, armv8.2-a):  Likewise.
27973         (generic-armv7-a): Make neon and neon-vfpv3 aliases.  Sort in
27974         canonical order.
27975         (cortex-a9): Sort in canonical order.
27976         * config/arm/arm.c (selftests.h): Include it.
27977         (arm_test_cpu_arch_data): New function.
27978         (arm_run_self_tests): New function.
27979         (TARGET_RUN_TARGET_SELFTESTS): Redefine.
27980         (targetm): Move declaration to the end of the file.
27981         * arm-cpu-cdata.h: Regenerated.
27983 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27985         * config/arm/arm.h (TARGET_MODE_SPECS): Add additional parameter to
27986         call to target_mode_check describing the type of option passed.
27987         * common/config/arm/arm-common.c (arm_arch_core_flag): Delete.
27988         (arm_target_thumb_only): Use arm_parse_arch_option_name or
27989         arm_parse_cpu_option_name to match parameters against list of
27990         available targets.
27991         * config/arm/parsecpu.awk (gen_comm_data): Don't generate
27992         arm_arch_core_flags data structure.
27993         * config/arm/arm-cpu_cdata.h: Regenerated.
27995 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27997         * common/config/arm/arm-common.c (arm_initialize_isa): Moved here from
27998         config/arm/arm.c.
27999         (arm_print_hint_for_cpu_option): Likewise.
28000         (arm_print_hint_for_arch_option): Likewise.
28001         (arm_parse_cpu_option_name): Likewise.
28002         (arm_parse_arch_option_name): Likewise.
28003         * config/arm/arm.c (arm_identify_fpu_from_isa): Use the computed number
28004         of entries in the all_fpus list.
28005         * config/arm/arm-protos.h (all_architectures, all_cores): Declare.
28006         (arm_parse_cpu_option_name): Declare.
28007         (arm_parse_arch_option_name): Declare.
28008         (arm_parse_option_features): Declare.
28009         (arm_intialize_isa): Declare.
28010         * config/arm/parsecpu.awk (gen_data): Move CPU and architecture
28011         data tables to ...
28012         (gen_comm_data): ... here.  Make definitions non-static.
28013         * config/arm/arm-cpu-data.h: Regenerated.
28014         * config/arm/arm-cpu-cdata.h: Regenerated.
28016 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
28018         * config/arm/arm-protos.h (arm_build_target): Remove arch_core.
28019         (cpu_arch_extension): New structure.
28020         (cpu_arch_option, arch_option, cpu_option): New structures.
28021         * config/arm/parsecpu.awk (gen_headers): Build an enumeration of
28022         architecture types.
28023         (gen_data): Generate new format data tables.
28024         * config/arm/arm.c (cpu_tune): New structure.
28025         (cpu_option, processors): Delete.
28026         (arm_print_hint_for_core_or_arch): Delete.  Replace with ...
28027         (arm_print_hint_for_cpu_option): ... this and ...
28028         (arm_print_hint_for_arch_option): ... this.
28029         (arm_parse_arch_cpu_name): Delete.  Replace with ...
28030         (arm_parse_cpu_option_name): ... this and ...
28031         (arm_parse_arch_option_name): ... this.
28032         (arm_unrecognized_feature): Change type of target parameter to
28033         cpu_arch_option.
28034         (arm_parse_arch_cpu_features): Delete.  Replace with ...
28035         (arm_parse_option_features): ... this.
28036         (arm_configure_build_target): Rework to use new configuration data
28037         tables.
28038         (arm_print_tune_info): Rework for new configuration data tables.
28039         * config/arm/arm-cpu-data.h: Regenerated.
28040         * config/arm/arm-cpu.h: Regenerated.
28042 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
28044         * Makefile.in (OBJS): Move sbitmap.o from here ...
28045         (OBJS-libcommon): ... to here.
28047 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
28049         * config/arm/arm-isa.h (ISA_ALL_FPU_INTERNAL): Renamed from ISA_ALL_FPU.
28050         (ISA_ALL_CRYPTO): New macro.
28051         (ISA_ALL_SIMD): New macro
28052         (ISA_ALL_FP): New macro.
28053         * config/arm/arm.c (fpu_bitlist): Update initializer.
28054         * config/arm/arm-cpus.in: Use new ISA_ALL macros to disable crypto,
28055         simd or fp.
28056         (arm9e): Add fpu.  Add option for nofp
28057         (arm946e-s, arm966e-s, arm968e-s, arm10e, arm1020e, arm1022e): Likewise.
28058         (arm926ej-s, arm1026ej-s): Likewise.
28059         (generic-armv7-a): Add fpu.  Add options for simd, vfpv3, vfpv3-d16,
28060         vfpv3-fp16, vfpv3-d16-fp16, vfpv4, vfpv4-d16, neon, neon-vfp3,
28061         neon-fp16, neon-vfpv4, nofp and nosimd.
28062         (cortex-a5, cortex-a7): Add fpu.  Add options for nosimd and nofp.
28063         (cortex-a8): Add fpu.  Add option for nofp.
28064         (cortex-a9): Add fpu.  Add options for nosimd and nofp.
28065         (cortex-a12, cortex-a15, cortex-a17): Add fpu.  Add option for nofp.
28066         (cortex-r4f): Add fpu.
28067         (cortex-r5): Add fpu.  Add options for nofp.dp and nofp.
28068         (cortex-r7): Use idiv option from architecture.  Add fpu.  Add option
28069         for nofp.
28070         (cortex-r8): Likewise.
28071         (cortex-m4): Add fpu.  Add option for nofp.
28072         (cortex-a15.cortex-a7): Add fpu.  Add option for nofp.
28073         (cortex-a17.cortex-a7): Likewise.
28074         (cortex-a32): Add fpu.  Add options for crypto and nofp.
28075         (cortex-a35, cortex-a53): Likewise.
28076         (cortex-a57): Add fpu.  Add option for crypto.
28077         (cortex-a72, cortex-a73): Likewise.
28078         (exynos-m1): Likewise.
28079         (cortex-a57.cortex-a53, cortex-a72.cortex-a53): Likewise.
28080         (cortex-a73.cortex-a35, cortex-a73.cortex-a53): Likewise.
28081         (cortex-m33): Add fpu.  Add option for nofp.
28082         * config/arm/arm-cpu-cdata.h: Regenerated
28083         * config/arm/arm-cpu-data.h: Regenerated.
28085 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
28087         * arm-cpus.in (armv5e): Add options fp, vfpv2 and nofp.
28088         (armv5te, armv5tej): Likewise.
28089         (armv6, armv6j, armv6k, armv6z, armv6kz, armv6zk, armv6t2): Likewise.
28090         (armv7): Add options fp and vfpv3-d16.
28091         (armv7-a): Add options fp, simd, vfpv3, vfpv3-d16, vfpv3-d16-fp16,
28092         vfpv3-fp16, vfpv4, vfpv4-d16, neon, neon-vfpv3, neon-fp16, neon-vfpv4,
28093         nofp and nosimd.
28094         (armv7ve): Likewise.
28095         (armv7-r): Add options fp, fp.sp, idiv, nofp and noidiv.
28096         (armv7e-m): Add options fp, fpv5, fp.dp and nofp.
28097         (armv8-a): Add nocrypto option.
28098         (armv8.1-a, armv8.2-a): Likewise.
28099         (armv8-m.main): add options fp, fp.dp and nofp.
28101 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
28103         * config/arm/arm-cpus.in (armv8-a): Add options crc, simd crypto and
28104         nofp.
28105         (armv8-a+crc): Delete.
28106         (armv8.1-a): Add options simd, crypto and nofp.
28107         (armv8.2-a): Add options fp16, simd, crypto and nofp.
28108         (armv8.2-a+fp16): Delete.
28109         (armv8-m.main): Add option dsp.
28110         (armv8-m.main+dsp): Delete.
28111         (cortex-a8): Add fpu.  Add nofp option.
28112         (cortex-a9): Add fpu.  Add nofp and nosimd options.
28113         * config/arm/parsecpu.awk (gen_data): Generate option tables and
28114         link to main cpu and architecture data structures.
28115         (gen_comm_data): Only put isa attributes from the main architecture
28116         in common tables.
28117         (option): New statement for architecture and CPU entries.
28118         * arm.c (struct cpu_option): New structure.
28119         (struct processors): Add entry for options.
28120         (arm_unrecognized_feature): New function.
28121         (arm_parse_arch_cpu_name): Ignore any characters after the first
28122         '+' character.
28123         (arm_parse_arch_cpu_feature): New function.
28124         (arm_configure_build_target): Separate out any CPU and architecture
28125         features and parse separately.  Don't error out if -mfpu=auto is
28126         used with only an architecture string.
28127         (arm_print_asm_arch_directives): New function.
28128         (arm_file_start): Call it.
28129         * config/arm/arm-cpu-cdata.h: Regenerated.
28130         * config/arm/arm-cpu-data.h: Likewise.
28131         * config/arm/arm-tables.opt: Likewise.
28133 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
28135         * config/arm/elf.h (ASM_SPEC): Only pass -mfpu through to the
28136         assembler when it is not -mfpu=auto.
28138 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
28140         * config/arm/arm.h (BIG_LITTLE_SPEC): Delete macro.
28141         (ASM_REWRITE_SPEC_FUNCTIONS): New macro.
28142         (BIG_LITTLE_CPU_SPEC_FUNCTIONS): Delete macro.
28143         (ASM_CPU_SPEC): Rewrite.
28144         (MCPU_MTUNE_NATIVE_FUNCTIONS): New macro.
28145         (EXTRA_SPEC_FUNCTIONS): Move outside of ifdef.  Use
28146         MCPU_MTUNE_NATIVE_FUNCTIONS and ASM_REWRITE_SPEC_FUNCTIONS.  Remove
28147         reference to BIG_LITTLE_CPU_SPEC_FUNCTIONS.
28148         * common/config/arm/arm-common.c (arm_rewrite_selected_cpu): Ensure
28149         copied string is NUL-terminated.  Also strip any characters prefixed
28150         by '+'.
28151         (arm_rewrite_selected_arch): New function.
28152         (arm_rewrite_march): New function.
28154 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
28156         * config/arm/arm.opt (x_arm_arch_string): New TargetSave option.
28157         (x_arm_cpu_string, x_arm_tune_string): Likewise.
28158         (march, mcpu, mtune): Convert to string-based options.
28159         * config/arm/arm.c (arm_print_hint_for_core_or_arch): New function.
28160         (arm_parse_arch_cpu_name): New function.
28161         (arm_configure_build_target): Use arm_parse_arch_cpu_name to
28162         identify selected architecture or CPU.
28163         (arm_option_save): New function.
28164         (TARGET_OPTION_SAVE): Redefine.
28165         (arm_option_restore): Restore string options.
28166         (arm_option_print): Print string options.
28168 2017-06-16  Martin Sebor  <msebor@redhat.com>
28170         PR tree-optimization/80933
28171         PR tree-optimization/80934
28172         * builtins.c (fold_builtin_3): Do not handle bcmp here.
28173         * gimple-fold.c (gimple_fold_builtin_bcmp): New function.
28174         (gimple_fold_builtin_bcopy, gimple_fold_builtin_bzero): Likewise.
28175         (gimple_fold_builtin): Call them.
28177 2017-06-16  Jan Hubicka  <hubicka@ucw.cz>
28179         * gimple-ssa-isolate-paths.c (isolate_path): Set edge leading to path
28180         as unlikely; update profile.
28182 2017-06-16  Jan Hubicka  <hubicka@ucw.cz>
28184         * predict.c (force_edge_cold): Handle declaring edges impossible
28185         more aggresively.
28187 2017-06-16  Jan Hubicka  <hubicka@ucw.cz>
28189         * tree-ssa-loop-ivcanon.c (remove_exits_and_undefined_stmts): Update
28190         profile.
28191         (try_unroll_loop_completely): Fix reporting.
28193 2017-06-16  Jan Hubicka  <hubicka@ucw.cz>
28195         * tree-ssa-tail-merge.c (replace_block_by): Fix profile updating.
28197 2017-06-16  James Greenhalgh  <james.greenhalgh@arm.com>
28199         PR target/71778
28200         * config/arm/arm-builtins.c (arm_expand_builtin_args): Return TARGET
28201         if given a non-constant argument for an intrinsic which requires a
28202         constant.
28204 2017-06-16  Jan Hubicka  <hubicka@ucw.cz>
28206         * profile.c (compare_freqs): New function.
28207         (branch_prob): Sort edge list.
28208         (find_spanning_tree): Assume that the list is priority sorted.
28210 2017-06-16  Richard Biener  <rguenther@suse.de>
28212         PR tree-optimization/81090
28213         * passes.def (pass_record_bounds): Remove.
28214         * tree-pass.h (make_pass_record_bounds): Likewise.
28215         * tree-ssa-loop.c (pass_data_record_bounds, pass_record_bounds,
28216         make_pass_record_bounds): Likewise.
28217         * tree-ssa-loop-ivcanon.c (canonicalize_induction_variables): Do
28218         not free niter estimates at the beginning but at the end.
28219         * tree-scalar-evolution.c (scev_finalize): Free niter estimates.
28221 2017-06-16  Richard Biener  <rguenther@suse.de>
28223         * tree-switch-conversion.c (emit_case_bit_tests): Adjust
28224         initializer to workaround ICE in host GCC 4.8.
28226 2017-06-16  Jan Hubicka  <hubicka@ucw.cz>
28228         * ipa-inline-transform.c (update_noncloned_frequencies): Update also
28229         counts.
28230         (clone_inlined_nodes): Update.
28232 2017-06-16  Maxim Kuvyrkov  <maxim.kuvyrkov@linaro.org>
28234         * config/aarch64/aarch64.c (qdf24xx_prefetch_tune): Update
28235         prefetch settings, and enable prefetching by default at -O3.
28237 2017-06-16  Maxim Kuvyrkov  <maxim.kuvyrkov@linaro.org>
28239         * config/aarch64/aarch64.c (aarch64_override_options_internal):
28240         Set flag_prefetch_loop_arrays according to tuning data.
28242 2017-06-16  Maxim Kuvyrkov  <maxim.kuvyrkov@linaro.org>
28244         * config/aarch64/aarch64-protos.h (struct cpu_prefetch_tune):
28245         New tune structure.
28246         (struct tune_params): Use cpu_prefetch_tune instead of cache_line_size.
28247         [Unrelated to main purpose of the patch] Place the pointer field last
28248         to enable type checking errors when tune structure are wrongly merged.
28249         * config/aarch64/aarch64.c (generic_prefetch_tune,)
28250         (exynosm1_prefetch_tune, qdf24xx_prefetch_tune,)
28251         (thunderx2t99_prefetch_tune): New tune constants.
28252         (tune_params *_tunings): Update all tunings (no functional change).
28253         (aarch64_override_options_internal): Set PARAM_SIMULTANEOUS_PREFETCHES,
28254         PARAM_L1_CACHE_SIZE, PARAM_L1_CACHE_LINE_SIZE, and PARAM_L2_CACHE_SIZE
28255         from tunings structures.
28257 2017-06-16  Jakub Jelinek  <jakub@redhat.com>
28259         PR sanitizer/81094
28260         * ubsan.c (instrument_null): Add T argument, use it instead
28261         of computing it based on IS_LHS.
28262         (instrument_object_size): Likewise.
28263         (pass_ubsan::execute): Adjust instrument_null and
28264         instrument_object_size callers to pass gimple_get_lhs or
28265         gimple_assign_rhs1 result to it.  Use instrument_null instead of
28266         calling get_base_address and instrument_mem_ref.  Handle
28267         aggregate call arguments for object-size sanitization.
28269 2017-06-16  Yury Gribov  <tetra2005@gmail.com>
28271         PR tree-optimization/81089
28272         * tree-vrp.c (is_masked_range_test): Validate operands of
28273         subexpression.
28275 2017-06-15  Martin Sebor  <msebor@redhat.com>
28277         PR c++/80560
28278         * dumpfile.c (dump_register): Avoid calling memset to initialize
28279         a class with a default ctor.
28280         * gcc.c (struct compiler): Remove const qualification.
28281         * genattrtab.c (gen_insn_reserv): Replace memset with initialization.
28282         * hash-table.h: Ditto.
28283         * ipa-cp.c (allocate_and_init_ipcp_value): Replace memset with
28284           assignment.
28285         * ipa-prop.c (ipa_free_edge_args_substructures): Ditto.
28286         * omp-low.c (lower_omp_ordered_clauses): Replace memset with
28287         default ctor.
28288         * params.h (struct param_info): Make struct members non-const.
28289         * tree-switch-conversion.c (emit_case_bit_tests): Replace memset
28290         with default initialization.
28291         * vec.h (vec_copy_construct, vec_default_construct): New helper
28292         functions.
28293         (vec<T>::copy, vec<T>::splice, vec<T>::reserve): Replace memcpy
28294         with vec_copy_construct.
28295         (vect<T>::quick_grow_cleared): Replace memset with default ctor.
28296         (vect<T>::vec_safe_grow_cleared, vec_safe_grow_cleared): Same.
28297         * doc/invoke.texi (-Wclass-memaccess): Document.
28299 2017-06-15  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
28301         * emit-rtl.h (is_leaf): Update comment about local
28302         register allocator.
28304 2017-06-15  Jozef Lawrynowicz   <jozef.l@somniumtech.com>
28306         PR target/78818
28307         * config/msp430/msp430.c (msp430_data_attr): Check that it's possible
28308         for a variable to have a section before checking if the section has a
28309         name.
28310         Set section to.persistent if persistent attribute is set.
28311         Warn if .persistent attribute is used on an automatic variable.
28313 2017-06-15  Eric Botcazou  <ebotcazou@adacore.com>
28315         PR rtl-optimization/80474
28316         * reorg.c (update_block): Do not ignore instructions in a delay slot.
28318 2017-06-15  Segher Boessenkool  <segher@kernel.crashing.org>
28320         * config/rs6000/rs6000.md (add<mode>3): Use reg_or_subregno instead
28321         of REGNO.
28323 2017-06-14  Maciej W. Rozycki  <macro@imgtec.com>
28325         * config/mips/mips.md (MIPS16_T_REGNUM): Remove constant.
28326         (casesi): Emit bounds checking as RTL.
28327         (casesi_internal_mips16_<mode>): Remove bounds checking.
28329 2017-06-14  Max Filippov  <jcmvbkbc@gmail.com>
28331         * config/xtensa/xtensa.c (xtensa_option_override): Append
28332         MASK_CONST16 to target_flags in the absence of TARGET_L32R.
28333         (hwloop_optimize, hwloop_fail, hwloop_pattern_reg,
28334          xtensa_doloop_hooks): Define unconditionally.
28335         (xtensa_reorg_loops): Only call reorg_loops in the presence of
28336         TARGET_LOOPS.
28337         * config/xtensa/xtensa.h (TARGET_L32R): New definition.
28338         (TARGET_DEFAULT): Remove XCHAL_HAVE_L32R condition and account
28339         for it in xtensa_option_override.
28340         (HARD_FRAME_POINTER_IS_FRAME_POINTER,
28341          HARD_FRAME_POINTER_IS_ARG_POINTER): New definitions.
28343 2017-06-14  Boris Kolpackov  <boris@codesynthesis.com>
28345         * doc/cppopts.texi: Document '-' special value to -MF.
28347 2017-06-14  Wilco Dijkstra  <wdijkstr@arm.com>
28349         * config/arm/cortex-a53.md (cortex_a53_fpalu) Adjust latency.
28350         (cortex_a53_fconst): Likewise.
28351         (cortex_a53_fpmul): Likewise.
28352         (cortex_a53_f_load_64): Likewise.
28353         (cortex_a53_f_load_many): Likewise.
28354         (cortex_a53_advsimd_alu): Likewise.
28355         (cortex_a53_advsimd_alu_q): Likewise.
28356         (cortex_a53_advsimd_mul): Likewise.
28357         (cortex_a53_advsimd_mul_q): Likewise.
28358         (fpmac bypass): Add new bypass for fpmac-fpmac case.
28359         Add missing fmul, r2f_cvt and fconst cases.
28361 2017-06-14  Richard Biener  <rguenther@suse.de>
28363         PR middle-end/81088
28364         * fold-const.c (split_tree): Drop TREE_OVERFLOW flag from
28365         literal constants.
28366         (fold_binary_loc): When associating do not treat pre-existing
28367         TREE_OVERFLOW on literal constants as a reason to allow
28368         TREE_OVERFLOW on associated literal constants.
28370 2017-06-14  Eric Botcazou  <ebotcazou@adacore.com>
28372         * config/sparc/sparc.h (MASK_ISA): Add MASK_LEON and MASK_LEON3.
28373         (MASK_FEATURES): New macro.
28374         * config/sparc/sparc.c (sparc_option_override): Remove the special
28375         handling of -mfpu and generalize it to all MASK_FEATURES switches.
28377 2017-06-14  Eric Botcazou  <ebotcazou@adacore.com>
28379         * simplify-rtx.c (simplify_binary_operation_1) <UDIV>: Do not simplify
28380         a division of 0 if non-call exceptions are enabled.
28382 2017-06-14  Andrew Pinski  <apinski@cavium.com>
28383             Naveen H.S  <Naveen.Hurugalawadi@cavium.com>
28385         PR target/71663
28386         * config/aarch64/aarch64.c (aarch64_expand_vector_init):
28387         Improve vector initialization code gen for only variable case.
28389 2017-06-14  Eric Botcazou  <ebotcazou@adacore.com>
28391         * config/sparc/driver-sparc.c (cpu_names): Add SPARC-T5 entry.
28393 2017-06-14  Richard Biener  <rguenther@suse.de>
28395         PR tree-optimization/81083
28396         * tree-ssa-sccvn.c (vn_reference_lookup_3): Do not use abnormals
28397         as values.
28399 2017-06-13  Segher Boessenkool  <segher@kernel.crashing.org>
28401         * config/rs6000/rs6000.c: Update all comments that mentioned SPE.
28402         (rs6000_expand_builtin): Remove RS6000_BTC_EVSEL.
28403         * config/rs6000/rs6000.h (RS6000_BTC_EVSEL): Delete.
28404         * config/rs6000/vxworks.h (VXCPU_FOR_8548): Delete.  Adjust former use.
28405         * config/rs6000/vxworksae.h (VXCPU_FOR_8548): Delete.
28406         * config/rs6000/vxworksmils.h (VXCPU_FOR_8548): Delete.
28408 2017-06-13  Segher Boessenkool  <segher@kernel.crashing.org>
28410         * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete VECTOR_SPE.
28411         * config/rs6000/rs6000.c (rs6000_debug_vector_unit): Delete VECTOR_SPE.
28413 2017-06-13  Segher Boessenkool  <segher@kernel.crashing.org>
28415         * config/rs6000/rs6000.h (FIXED_SCRATCH): Delete.
28417 2017-06-13  Segher Boessenkool  <segher@kernel.crashing.org>
28419         * config/rs6000/t-rtems: Don't handle SPE.
28421 2017-06-13  Segher Boessenkool  <segher@kernel.crashing.org>
28423         * config/rs6000/t-linux: Don't handle SPE.
28425 2017-06-13  Segher Boessenkool  <segher@kernel.crashing.org>
28427         * config/rs6000/eabispe.h: Delete file.
28429 2017-06-13  Segher Boessenkool  <segher@kernel.crashing.org>
28431         * config/rs6000/t-spe: Delete file.
28433 2017-06-13  Segher Boessenkool  <segher@kernel.crashing.org>
28435         * config/rs6000/rs6000.c (SPE_CONST_OFFSET_OK): Delete.
28436         (rs6000_legitimate_offset_address_p): Return false for anything in
28437         V2SImode or V2SFmode.
28439 2017-06-13  Segher Boessenkool  <segher@kernel.crashing.org>
28441         * config/rs6000/rs6000-modes.def: Remove all 8-byte vector modes
28442         except V2SF and V2SI.  Rearrange the vector modes, and add comments.
28443         * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove V8QImode
28444         and V4HImode.
28445         (reg_offset_addressing_ok_p): Remove V4HImode and V1DImode.
28446         (rs6000_legitimate_offset_address_p): Ditto.
28447         (rs6000_emit_move): Ditto.
28448         (rs6000_init_builtins): Remove V4HI_type_node.
28450 2017-06-13  Martin Liska  <mliska@suse.cz>
28452         PR sanitize/78204
28453         * asan.c (asan_sanitize_stack_p): Use sanitize_flags_p.
28454         (gate_asan): Likewise.
28455         * asan.h (asan_no_sanitize_address_p): Remove the function.
28456         (sanitize_flags_p): New function.
28457         * builtins.def: Fix coding style.
28458         * common.opt: Use renamed enum value.
28459         * convert.c (convert_to_integer_1): Use sanitize_flags_p.
28460         * doc/extend.texi: Document no_sanitize attribute.
28461         * flag-types.h (enum sanitize_code): Rename SANITIZE_NONDEFAULT
28462         to SANITIZE_UNDEFINED_NONDEFAULT.
28463         * gcc.c (sanitize_spec_function): Use the renamed enum value.
28464         * gimple-fold.c (optimize_atomic_compare_exchange_p):
28465         Use sanitize_flags_p.
28466         * gimplify.c (gimplify_function_tree): Likewise.
28467         * ipa-inline.c (sanitize_attrs_match_for_inline_p): Likewise.
28468         * opts.c (parse_no_sanitize_attribute): New function.
28469         (common_handle_option): Use renamed enum value.
28470         * opts.h (parse_no_sanitize_attribute): Declare.
28471         * tree.c (sanitize_flags_p): New function.
28472         * tree.h: Declared here.
28473         * tsan.c: Use sanitize_flags_p.
28474         * ubsan.c (ubsan_expand_null_ifn): Likewise.
28475         (instrument_mem_ref): Likewise.
28476         (instrument_bool_enum_load): Likewise.
28477         (do_ubsan_in_current_function): Remove the function.
28478         (pass_ubsan::execute): Use sanitize_flags_p.
28479         * ubsan.h: Remove do_ubsan_in_current_function
28480         * tree-cfg.c (print_no_sanitize_attr_value): New function.
28481         (dump_function_to_file): Use it here.
28483 2017-06-13  Martin Jambor  <mjambor@suse.cz>
28485         PR tree-optimization/80803
28486         PR tree-optimization/81063
28487         * tree-sra.c (subtree_mark_written_and_enqueue): Move up in the file.
28488         (propagate_subaccesses_across_link): Enqueue subtree whenever
28489         necessary instead of relying on the caller.
28491 2017-06-13  Martin Jambor  <mjambor@suse.cz>
28493         * tree-sra.c (add_access_to_work_queue): Only enqueue accesses
28494         that have a first_link.
28495         (sort_and_splice_var_accesses): Do not check first_link before
28496         enquing.
28497         (subtree_mark_written_and_enqueue): Likewise.
28498         (propagate_all_subaccesses): Likewise and do not stop at first
28499         parent with a first_link.
28501 2017-06-13  Martin Jambor  <mjambor@suse.cz>
28503         * tree-sra.c (dump_access_tree_1): Fix accidental dumping to stderr
28504         instead of f.
28506 2017-06-13  Yury Gribov  <tetra2005@gmail.com>
28508         * match.pd: New pattern.
28510 2017-06-13  Yury Gribov  <tetra2005@gmail.com>
28512         * tree-vrp.c (is_masked_range_test): New function.
28513         (register_edge_assert_for): Determine ranges for
28514         some bit tests.
28516 2017-06-13  Yury Gribov  <tetra2005@gmail.com>
28518         PR tree-optimization/67328
28519         * fold-const.c (maskable_range_p): New function.
28520         (build_range_check): Generate bittests if possible.
28522 2017-06-13  Martin Liska  <mliska@suse.cz>
28524         * gimple-pretty-print.c (dump_probability): Add new argument.
28525         (dump_edge_probability): Dump both probability and count.
28526         (dump_gimple_label): Likewise.
28527         (dump_gimple_bb_header): Likewise.
28529 2017-06-13  Georg-Johann Lay  <avr@gjlay.de>
28531         PR target/81072
28532         * config/avr/avr-devices.c: Fix indentation.
28533         * config/avr/gen-avr-mmcu-specs.c: Dito.
28535 2017-06-13  Richard Biener  <rguenther@suse.de>
28537         * tree-vect-loop.c (vect_model_reduction_cost): Do not fail,
28538         instead get vector type from stmt_info.
28539         (vectorizable_reduction): Adjust.  Remove dead code.
28541 2017-06-13  Richard Biener  <rguenther@suse.de>
28543         PR middle-end/81065
28544         * fold-const.c (extract_muldiv_1): Remove bogus distribution
28545         case of C * (x * C2 + C3).
28546         (fold_addr_of_array_ref_difference): Properly fold index difference.
28548 2017-06-12  David S. Miller  <davem@davemloft.net>
28550         PR target/80968
28551         * config/sparc/sparc.md (return expander): Emit frame blockage if
28552         function uses alloca.
28554 2017-06-12  Richard Sandiford  <richard.sandiford@linaro.org>
28556         * combine.c (make_field_assignment): Check len rather than the mode
28557         precision when calling force_to_mode.
28559 2017-06-12  Georg-Johann Lay  <avr@gjlay.de>
28561         Support multilibs and devices that see flash in RAM address range.
28563         PR target/81072
28564         * config/avr/avr-arch.h (avr_arch_id) <ARCH_AVRXMEGA3>: New enum.
28565         (avr_mcu_t) <flash_pm_offset>: New field.
28566         (avr_device_specific_features) <AVR_ISA_RCALL>: New enum.
28567         * config/avr/avr.h (AVR_SHORT_CALLS): New define.
28568         (AVR_HAVE_JMP_CALL): Don't set if AVR_SHORT_CALLS.
28569         (AVR_TINY_PM_OFFSET): Remove macro.
28570         * config/avr/avr.opt (-mshort-calls): New option.
28571         * config/avr/gen-avr-mmcu-specs.c (print_mcu)
28572         [*self_spec]: Add / remove -mshort-calls depending on AVR_ISA_RCALL.
28573         * config/avr/avr-c.c (avr_cpu_cpp_builtins)
28574         <__AVR_SHORT_CALLS__>: Built-in define if AVR_SHORT_CALLS.
28575         <__AVR_HAVE_JMP_CALL__>: Use AVR_HAVE_JMP_CALL as condition
28576         instead of avr_arch->have_jmp_call.
28577         <__AVR_PM_BASE_ADDRESS__>: Built-in define if avr_arch->flash_pm_offset.
28578         [AVR_TINY] <__AVR_TINY_PM_BASE_ADDRESS__>: Use
28579         avr_arch->flash_pm_offset to define.
28580         * config/avr/avr-devices.c (avr_arch_types): Add initializers for
28581         new field flash_pm_offset.  Add entry for avrxmega3.
28582         (avr_texinfo): Add entry for avrxmega3.
28583         * config/avr/avr-mcus.def: Add entries for: avrxmega3,
28584         attiny212, attiny214,
28585         attiny412, attiny414, attiny416, attiny417,
28586         attiny814, attiny816, attiny817,
28587         attiny1614, attiny1616, attiny1617,
28588         attiny3214, attiny3216, attiny3217.
28589         * config/avr/avr.c (avr_assemble_integer)[AVR_TINY]: Use
28590         avr_arch->flash_pm_offset instead of AVR_TINY_PM_OFFSET.
28591         (avr_print_operand_address) [AVR_TINY]: Same.
28592         (avr_asm_init_sections) <readonly_data_section>: Only patch
28593         callback if avr_arch->flash_pm_offset = 0.
28594         (avr_asm_named_section) <avr_need_copy_data_p>: Skip setting it
28595         for rodata if avr_arch->flash_pm_offset != 0.
28596         (avr_encode_section_info) [AVR_TINY]: Adjust comment.
28597         * config/avr/genmultilib.awk (dir_rcall, opt_rcall): New vars.
28598         (opts) [AVR_ISA_RCALL]: Append opt_rcall.
28599         (m_options): Append opt_rcall.
28600         (m_dirnames): Append dir_rcall.
28601         * config/avr/t-multilib: Regenerate.
28603         * configure.ac [target=avr]: Check whether avrxmega3 default
28604         linker description file works as needed.
28605         * configure: Regenerate.
28606         * doc/avr-mmcu.texi: Regenerate.
28607         * doc/invoke.texi (AVR Options) <-mshort-calls>: Document it.
28608         <__AVR_ARCH__>: Document avrxmega3 and 103.
28609         <__AVR_HAVE_JMP_CALL__>: Adjust documentation.
28610         <__AVR_SHORT_CALLS__>: Document it.
28611         <__AVR_PM_BASE_ADDRESS__>: Document it.
28612         * doc/extend.texi (AVR Options) <-mshort-calls>: Document it.
28613         (AVR Variable Attributes) <progmem>: Document this is
28614         not needed for avrxmega3.
28615         (AVR Named Address Spaces) <__flash>: Dito.
28617 2017-06-12  Jan Hubicka  <hubicka@ucw.cz>
28619         * cgraph.c (cgraph_node::dump): Complain about profile insanities.
28621 2017-06-12  Doug Rupp  <rupp@adacore.com>
28623         * config.gcc (*-*-vxworks*): Set use_gcc_stdint to "provide".
28624         Append vxworks-stdint.h to the tm_file list.
28625         * config/vxworks-stdint.h: New file.
28627 2017-06-12  Martin Liska  <mliska@suse.cz>
28629         PR tree-optimization/81041
28630         * tree-profile.c (gimple_gen_ic_func_profiler):
28631         Create an extra BB in profile-generate
28632         (gimple_gen_time_profiler): Likewise.
28634 2017-06-12  Jakub Jelinek  <jakub@redhat.com>
28636         PR tree-optimization/81003
28637         * tree-ssa-reassoc.c (force_into_ssa_name): New function.
28638         (update_range_test): Use it instead of force_gimple_operand_gsi.
28640 2017-06-12  Richard Biener  <rguenther@suse.de>
28642         PR tree-optimization/81053
28643         * tree-vect-loop.c (vect_is_simple_reduction): Handle PHI
28644         with backedge value not defined in loop.  Simplify def stmt
28645         compute.
28647 2017-06-11  Tom de Vries  <tom@codesourcery.com>
28649         PR target/79939
28650         * config/nvptx/nvptx.c (nvptx_cannot_force_const_mem): New function.
28651         Return true.
28652         (TARGET_CANNOT_FORCE_CONST_MEM): Redefine to
28653         nvptx_cannot_force_const_mem.
28655 2017-06-10  Jan Hubicka  <hubicka@ucw.cz>
28657         * opts.c (finish_options): Move test for flag_split_stack after
28658         it has been initialized.
28660 2017-06-11  Jason Merrill  <jason@redhat.com>
28662         * tree.h (id_equal): New.
28663         * dwarf2out.c, hsa-gen.c, ipa-devirt.c, omp-expand.c,
28664         omp-simd-clone.c, read-rtl-function.c, tree-chkp.c, tree.c: Use it
28665         instead of strcmp of IDENTIFIER_POINTER.
28667 2017-06-10  Jan Hubicka  <hubicka@ucw.cz>
28669         * ipa-inline-transform.c: Include function.h, cfg.h and basic-block.h
28670         (mark_all_inlined_calls_cdtor): Fix formating.
28671         (inline_transform): Rescale profile before inlining.
28673 2017-06-10  Jan Hubicka  <hubicka@ucw.cz>
28675         * cgraph.h (cgraph_edge::clone): Update prototype.
28676         * cgraphclones.c (cgraph_edge::clone): Update profile scaling.
28677         (cgraph_node::create_clone): Update.
28678         (cgraph_node::create_version_clone): Update.
28679         * tree-inline.c (copy_bb): Update.
28680         (expand_call_inline): Update.
28682 2017-06-10  Segher Boessenkool  <segher@kernel.crashing.org>
28684         * config/rs6000/rs6000.c (emit_vrsave_prologue): New function,
28685         factored out from ...
28686         (rs6000_emit_prologue): ... here.
28688 2017-06-10  Segher Boessenkool  <segher@kernel.crashing.org>
28690         * config/rs6000/rs6000.c (emit_split_stack_prologue): New function,
28691         factored out from ...
28692         (rs6000_emit_prologue): ... here.
28694 2017-06-10  Jan Hubicka  <hubicka@ucw.cz>
28696         * predict.c (drop_profile): Also drop individual bb/edge and cgraph
28697         edge counts.
28698         (handle_missing_profiles): Fix computation of tp_first_run.
28699         (counts_to_freqs): Do not touch freqs when count is 0.
28701 2017-06-10  Jan Hubicka  <hubicka@ucw.cz>
28703         * cgraphbuild.c (cgraph_edge::rebuild_references): Do not touch
28704         profile.
28706 2017-06-10  Tom de Vries  <tom@codesourcery.com>
28708         * doc/sourcebuild.texi (Effective-Target Keywords, Environment
28709         attributes): Document signal effective target.
28711 2017-06-10  Tom de Vries  <tom@codesourcery.com>
28713         * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
28714         Document effective target stack_size.
28716 2017-06-09  David Malcolm  <dmalcolm@redhat.com>
28718         * diagnostic.c (diagnostic_report_diagnostic): Only add fixits
28719         to the edit_context if they can be auto-applied.
28721 2017-06-9  Ian Lance Taylor  <iant@golang.org>
28723         * opts.c (finish_options): If -fsplit-stack, disable implicit
28724         -forder-blocks-and-partition.
28725         * doc/invoke.texi (Optimize Options): Document that when using
28726         -fsplit-stack -forder-blocks-and-partition is not implicitly
28727         enabled.
28729 2017-06-09  Jan Hubicka  <hubicka@ucw.cz>
28731         * builtin-attrs.def (ATTR_NORETURN_NOTHROW_LEAF_COLD_LIST,
28732         ATTR_CONST_NORETURN_NOTHROW_LEAF_COLD_LIST,
28733         ATTR_TMPURE_NORETURN_NOTHROW_LEAF_COLD_LIST): New.
28734         * builtins.def (abort, trap, unreachable): Declare cold.
28735         * calls.c (flags_from_decl_or_type): Lookup ECF_COLD.
28736         * tree-core.h (ECF_COLD): New.
28737         * tree.c (set_call_expr_flags): Handle ECF_COLD.
28738         (build_common_builtin_nodes): Mark unreachable and abort as cold.
28740 2017-06-09  Jan Hubicka  <hubicka@ucw.cz>
28742         * predict.c (unlikely_executed_stmt_p): Cleanup.
28744 2017-06-09  Richard Biener  <rguenther@suse.de>
28746         * tree-ssa-loop-im.c (execute_sm): Do not force multi-threaded
28747         model if the ref is always written to.
28749 2017-06-09  Tamar Christina  <tamar.christina@arm.com>
28751         * config/aarch64/aarch64.md (lrint<GPF:mode><GPI:mode>2): New.
28753 2017-06-09  Tamar Christina  <tamar.christina@arm.com>
28755         * config/arm/arm.c (arm_rtx_costs_internal): Make sdiv more expensive
28756         than udiv.
28758 2017-06-09  Tom de Vries  <tom@codesourcery.com>
28760         PR target/80855
28761         * config/nvptx/nvptx.md (define_expand "mov<QHSDISDFM>"): Error out with
28762         "target cannot support label values" when encountering LABEL_REF.
28764 2017-06-09  Martin Liska  <mliska@suse.cz>
28766         * tree-profile.c (gimple_gen_ic_profiler): Update comment.
28767         (gimple_gen_ic_func_profiler): Emit direct comparison
28768         of __gcov_indirect_call_callee with NULL.
28769         (gimple_gen_time_profiler): Change probability from
28770         PROB_VERY_UNLIKELY to PROB_UNLIKELY.
28772 2017-06-09  Jan Hubicka  <hubicka@ucw.cz>
28774         * profile.c (edge_gcov_counts): Turn to pointer.
28775         (compute_branch_probabilities, compute_branch_probabilities): Update.
28776         (branch_prob): Do not clear edge_gcov_count.
28777         * profile.h (edge_gcov_counts): Turn to pointer.
28778         (edge_gcov_count): Update.
28780 2017-06-09  Jan Hubicka  <hubicka@ucw.cz>
28782         * gimple.h (gimple_check_failed): Mark cold.
28784 2017-06-09  Richard Biener  <rguenther@suse.de>
28786         PR tree-optimization/66623
28787         * tree-vect-loop.c (vect_is_simple_reduction): Cleanup,
28788         refactor check_reduction into two parts, properly computing
28789         whether we have to check reduction validity for outer loop
28790         vectorization.
28792 2017-06-09  Richard Biener  <rguenther@suse.de>
28794         PR tree-optimization/79483
28795         * graphite-scop-detection.c (order): New global.
28796         (get_order): Compute bb to order mapping that satisfies code
28797         generation constraints.
28798         (cmp_pbbs): New helper.
28799         (build_scops): Start domwalk at entry block, sort generated
28800         pbbs.
28802 2017-06-09  Richard Biener  <rguenther@suse.de>
28804         PR middle-end/81007
28805         * ipa-polymorphic-call.c
28806         (ipa_polymorphic_call_context::restrict_to_inner_class):
28807         Skip FIELD_DECLs with error_mark_node type.
28808         * passes.def (all_lowering_passes): Run pass_build_cgraph_edges
28809         last again.
28811 2017-06-09  Martin Liska  <mliska@suse.cz>
28813         * predict.c (struct branch_predictor): New struct.
28814         (test_prediction_value_range): New test.
28815         (predict_c_tests): New function.
28816         * selftest-run-tests.c (selftest::run_tests): Run the function.
28817         * selftest.h: Declare new tests.
28819 2017-06-09  Segher Boessenkool  <segher@kernel.crashing.org>
28821         PR target/80966
28822         * config/rs6000/rs6000.c (rs6000_emit_allocate_stack): Assert that
28823         gen_add3_insn did not fail.
28824         * config/rs6000/rs6000.md (add<mode>3): If asked to add a constant to
28825         r0, construct that number in a temporary reg and add that reg to r0.
28826         If asked to put the result in r0 as well, fail.
28828 2017-06-08  Will Schmidt  <will_schmidt@vnet.ibm.com>
28830         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add handling
28831         for early expansion of vec_eqv.
28833 2017-06-08  Jakub Jelinek  <jakub@redhat.com>
28835         PR middle-end/81005
28836         * ubsan.c (instrument_null): Avoid pointless code temporary.
28837         (pass_ubsan::execute): Instrument aggregate arguments of calls.
28839 2017-06-08  Uros Bizjak  <ubizjak@gmail.com>
28841         PR target/81015
28842         Revert:
28843         2016-12-14  Uros Bizjak  <ubizjak@gmail.com>
28845         PR target/59874
28846         * config/i386/i386.md (*ctzhi2): New insn_and_split pattern.
28847         (*clzhi2): Ditto.
28849 2017-06-08  Jan Hubicka  <hubicka@ucw.cz>
28851         * predict.c (unlikely_executed_edge_p): Move ahead.
28852         (probably_never_executed_edge_p): Use it.
28854 2017-06-08  Alexander Ivchenko  <alexander.ivchenko@intel.com>
28856         PR middle-end/79988
28857         * tree-chkp.c (chkp_gimple_call_builtin_p): Remove
28858         gimple_call_builtin_p call.
28860 2017-06-08  Jan Hubicka  <hubicka@ucw.cz>
28862         * system.h (fancy_abort): Annotate by ATTRIBUTE_COLD.
28863         * rtl.h (rtl_check_failed_bounds, rtl_check_failed_type1,
28864         rtl_check_failed_type2, rtl_check_failed_code1,
28865         rtl_check_failed_code2, rtl_check_failed_code_mode,
28866         rtl_check_failed_block_symbol, cwi_check_failed_bounds,
28867         rtvec_check_failed_bounds, rtl_check_failed_flag,
28868         _fatal_insn_not_found, _fatal_insn): Likewise.
28869         * tree.h (tree_contains_struct_check_failed,
28870         tree_check_failed, tree_not_check_failed,
28871         tree_class_check_failed, tree_range_check_failed,
28872         tree_not_class_check_failed, tree_int_cst_elt_check_failed,
28873         tree_vec_elt_check_failed, phi_node_elt_check_failed,
28874         tree_operand_check_failed, omp_clause_check_failed,
28875         omp_clause_operand_check_failed, omp_clause_range_check_failed):
28876         Likewise.
28878 2017-06-08  Jan Hubicka  <hubicka@ucw.cz>
28880         * cgraph.c (cgraph_edge::maybe_hot_p): Do not check
28881         flag_branch_probabilities.
28882         * ipa-inline.c (edge_badness): Likewise.
28883         * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
28884         * postreload-gcse.c (eliminate_partially_redundant_load): Likewise.
28885         * predict.c (maybe_hot_frequency_p): Likewise.
28886         (probably_never_executed): Likewise.
28887         * sched-ebb.c (schedule_ebbs): Likewise.
28888         * sched-rgn.c (find_single_block_region): Likewise.
28889         * tracer.c (tail_duplicate): Likewise.
28891 2017-06-08  Jan Hubicka  <hubicka@ucw.cz>
28893         * opts.c (finish_options): x_flag_reorder_blocks_and_partition no
28894         longer requires x_flag_profile_use.
28896 2017-06-08  Jan Hubicka  <hubicka@ucw.cz>
28898         * cfgrtl.c (cfg_layout_initialize): Check crtl->has_bb_partition
28899         instead of flag_reorder_blocks_and_partition.
28900         * dbxout.c (dbxout_function_end): Likewise.
28901         * dwarf2out.c (gen_subprogram_die): Likewise.
28902         * haifa-sched.c (sched_create_recovery_edges): Likewise.
28903         * hw-doloop.c (reorg_loops): Likewise.
28904         * varasm.c (assemble_start_function,
28905         assemble_end_function): Likewise.
28906         (decide_function_section): Do not check for
28907         flag_reorder_blocks_and_partition.
28909 2017-06-08  Alexander Ivchenko  <alexander.ivchenko@intel.com>
28911         * tree-chkp.c (chkp_get_hard_register_var_fake_base_address):
28912         New function.
28913         (chkp_get_hard_register_fake_addr_expr): Ditto.
28914         (chkp_build_addr_expr): Add check for hard reg case.
28915         (chkp_parse_array_and_component_ref): Ditto.
28916         (chkp_find_bounds_1): Ditto.
28917         (chkp_process_stmt): Don't generate bounds store for
28918         hard reg case.
28920 2017-06-08  Jan Hubicka  <hubicka@ucw.cz>
28922         * predict.c (maybe_hot_bb_p): Do not check profile status.
28923         (maybe_hot_edge_p): Likewise.
28924         (probably_never_executed): Check for zero counts even if profile
28925         is not read.
28926         (unlikely_executed_edge_p): New function.
28927         (unlikely_executed_stmt_p): New function.
28928         (unlikely_executed_bb_p): New function.
28929         (set_even_probabilities): Use unlikely predicates.
28930         (combine_predictions_for_bb): Likewise.
28931         (predict_paths_for_bb): Likewise.
28932         (predict_paths_leading_to_edge): Likewise.
28933         (determine_unlikely_bbs): New function.
28934         (estimate_bb_frequencies): Use it.
28935         (compute_function_frequency): Use zero counts even if profile is
28936         not read.
28937         * profile-count.h: Fix typo.
28939 2017-08-08  Julia Koval  <julia.koval@intel.com>
28941         * config/i386/avx512bwintrin.h (_mm512_mask_cvtepi16_storeu_epi8,
28942         _mm512_mask_cvtsepi16_storeu_epi8,
28943         _mm512_mask_cvtusepi16_storeu_epi8): New intrinsics.
28944         * config/i386/avx512vlbwintrin.h (_mm256_mask_cvtepi16_storeu_epi8,
28945         _mm_mask_cvtsepi16_storeu_epi8, _mm256_mask_cvtsepi16_storeu_epi8,
28946         _mm_mask_cvtusepi16_storeu_epi8, _mm256_mask_cvtusepi16_storeu_epi8,
28947         _mm_mask_cvtepi16_storeu_epi8): New intrinsics.
28948         * config/i386/i386-builtin-types.def (PV8Q, V8QI): New pointer type.
28949         (VOID_FTYPE_PV32QI_V32HI_USI, VOID_FTYPE_PV8QI_V8HI_UQI,
28950         VOID_FTYPE_PV16QI_V16HI_UHI): New function types.
28951         * config/i386/i386-builtin.def (__builtin_ia32_pmovwb128mem_mask,
28952         __builtin_ia32_pmovwb256mem_mask, __builtin_ia32_pmovswb128mem_mask,
28953         __builtin_ia32_pmovswb256mem_mask, __builtin_ia32_pmovuswb128mem_mask,
28954         __builtin_ia32_pmovuswb256mem_mask,
28955         __builtin_ia32_pmovuswb512mem_mask, __builtin_ia32_pmovswb512mem_mask)
28956         __builtin_ia32_pmovwb512mem_mask): New builtins.
28958 2017-08-08  Julia Koval  <julia.koval@intel.com>
28960         PR target/73350,80862
28961         * config/i386/subst.md (round): Fix round pattern.
28962         * config/i386/i386.c (ix86_erase_embedded_rounding):
28963         Fix erasing rounding for the fixed pattern.
28965 2017-06-08  Jan Hubicka  <hubicka@ucw.cz>
28967         * cfgbuild.c (find_many_sub_basic_blocks): Fix thinko.
28969 2017-06-08  Martin Liska  <mliska@suse.cz>
28971         PR gcov-profile/80911
28972         * gcov.c (block_info::block_info): New constructor.
28974 2017-06-07  Carl Love  <cel@us.ibm.com>
28976         * config/rs6000/rs6000-c: The return type of the following
28977         built-in functions was implemented as int not long long.  Fix sign
28978         of return value for the unsigned version of vec_mulo and vec_mule.
28979         vector unsigned long long vec_bperm (vector unsigned long long,
28980                                              vector unsigned char)
28981         vector signed long long vec_mule (vector signed int,
28982                                           vector signed int)
28983         vector unsigned long long vec_mule (vector unsigned int,
28984                                             vector unsigned int)
28985         vector signed long long vec_mulo (vector signed int,
28986                                           vector signed int)
28987         vector unsigned long long vec_mulo (vector unsigned int,
28988                                             vector unsigned int)
28989         * doc/extend.texi: Fix the documentation for the built-in
28990         functions.
28992 2017-06-07  Carl Love  <cel@us.ibm.com>
28994         PR target/80982
28995         * config/rs6000/altivec.md (double<mode>2): Fix the implementation of
28996         for BE.
28998 2017-06-07  Carl Love  <cel@us.ibm.com>
29000         * config/rs6000/altivec.md: Fix argument swizzle in vec_doublel
29001         support, Generate       doublehv for signed int/float for BE case only.
29003 2017-06-07  Alexander Monakov  <amonakov@ispras.ru>
29005         * doc/invoke.texi (mcx16): Rewrite.
29007 2017-06-07  Segher Boessenkool  <segher@kernel.crashing.org>
29009         * config/rs6000/predicates.md (rs6000_nonimmediate_operand): Delete.
29010         * config/rs6000/rs6000.md (*movsi_internal1, movsi_from_sf,
29011         *mov<mode>_softfloat, and an anonymous splitter): Use
29012         nonimmediate_operand instead of rs6000_nonimmediate_operand.
29014 2017-06-07  Segher Boessenkool  <segher@kernel.crashing.org>
29016         * config/rs6000/darwin.h (REGISTER_NAMES): Delete the SPE_ACC and
29017         SPEFSCR registers.
29018         * config/rs6000/rs6000.c (rs6000_reg_names, alt_reg_names): Ditto.
29019         (enum rs6000_reg_type): Delete SPE_ACC_TYPE and SPEFSCR_REG_TYPE.
29020         (rs6000_debug_reg_global): Adjust.
29021         (rs6000_init_hard_regno_mode_ok): Adjust.
29022         (rs6000_dbx_register_number): Adjust.
29023         * config/rs6000/rs6000.h (FIRST_PSEUDO_REGISTER): Change to 115.
29024         (FIXED_REGISTERS, CALL_USED_REGISTERS, CALL_REALLY_USED_REGISTERS):
29025         Remove SPE_ACC and SPEFSCR.
29026         (REG_ALLOC_ORDER): Ditto.
29027         (FRAME_POINTER_REGNUM): Change to 111.
29028         (enum reg_class): Remove the SPE_ACC and SPEFSCR registers.
29029         (REG_CLASS_NAMES): Ditto.
29030         (REG_CLASS_CONTENTS): Delete the SPE_ACC and SPEFSCR registers.
29031         (REGISTER_NAMES): Ditto.
29032         (ADDITIONAL_REG_NAMES): Ditto.
29033         (rs6000_reg_names): Ditto.
29034         * config/rs6000/rs6000.md: Renumber some register number
29035         define_constants.
29037 2017-06-07  Segher Boessenkool  <segher@kernel.crashing.org>
29039         * config/rs6000/darwin.h (REGISTER_NAMES): Delete the SPE high
29040         registers.
29041         * config/rs6000/rs6000.c (rs6000_reg_names, alt_reg_names): Ditto.
29042         * config/rs6000/rs6000.h (FIRST_PSEUDO_REGISTER): Change from 149
29043         to 117.
29044         (DWARF_REG_TO_UNWIND_COLUMN): Do not define.
29045         (FIXED_REGISTERS, CALL_USED_REGISTERS, CALL_REALLY_USED_REGISTERS):
29046         Delete the SPE high registers.
29047         (REG_ALLOC_ORDER): Ditto.
29048         (enum reg_class): Remove SPE_HIGH_REGS.
29049         (REG_CLASS_NAMES): Ditto.
29050         (REG_CLASS_CONTENTS): Delete the SPE high registers.
29051         (REGISTER_NAMES): Ditto.
29052         (rs6000_reg_names): Ditto.
29053         * doc/tm.texi.in: Remove SPE as example.
29054         * doc/tm.texi: Regenerate.
29056 2017-06-07  Segher Boessenkool  <segher@kernel.crashing.org>
29058         * config/rs6000/8540.md (ppc8540_brinc): Delete.
29059         * config/rs6000/e500mc.md (e500mc_brinc): Delete.
29060         * config/rs6000/e500mc64.md (e500mc64_brinc): Delete.
29061         * config/rs6000/rs6000.md (type): Remove "brinc".
29063 2017-06-07  Segher Boessenkool  <segher@kernel.crashing.org>
29065         * config.gcc (powerpc*-*-*): Don't add spe.h to extra_headers.
29066         (powerpc*-linux*spe*): Use ${cpu_type} instead of rs6000.
29067         * config/rs6000/linuxspe.h: Delete file.
29068         * config/rs6000/rs6000.md: Don't include spe.md.
29069         * config/rs6000/spe.h: Delete file.
29070         * config/rs6000/spe.md: Delete file.
29071         * config/rs6000/t-rs6000: Remove spe.md.
29073 2017-06-07  Segher Boessenkool  <segher@kernel.crashing.org>
29075         * config/rs6000/predicates.md (reg_or_mem_operand): Reformat.
29076         (reg_or_none500mem_operand): Delete.
29077         * config/rs6000/rs6000.md (extendsfdf2): Use reg_or_mem_operand
29078         instead of reg_or_none500mem_operand.
29080 2017-06-07  Segher Boessenkool  <segher@kernel.crashing.org>
29082         * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
29083         handling of SPE flags.
29084         * config/rs6000/rs6000.opt (-mspe, -mspe=no, -mspe=yes): Delete.
29086 2017-06-07  Segher Boessenkool  <segher@kernel.crashing.org>
29088         * config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
29089         SPE ABI handling.
29090         * config/rs6000/paired.md (paired_negv2sf2): Rename to negv2sf2.
29091         (paired_absv2sf2, paired_addv2sf3, paired_subv2sf3, paired_mulv2sf3,
29092         paired_divv2sf3): Similar.
29093         * config/rs6000/predicates.md: Replace TARGET_SPE, TARGET_SPE_ABI,
29094         SPE_VECTOR_MODE and SPE_HIGH_REGNO_P by 0; simplify.
29095         * config/rs6000/rs6000-builtin.def: Delete RS6000_BUILTIN_E and
29096         RS6000_BUILTIN_S.
29097         Delete BU_SPE_1, BU_SPE_2, BU_SPE_3, BU_SPE_E, BU_SPE_P, and BU_SPE_X.
29098         Rename the paired_* instruction patterns.
29099         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Do not
29100         define __SPE__.
29101         * config/rs6000/rs6000-protos.h (invalid_e500_subreg): Delete.
29102         * config/rs6000/rs6000.c: Delete RS6000_BUILTIN_E and RS6000_BUILTIN_S.
29103         (struct rs6000_stack): Delete fields spe_gp_save_offset, spe_gp_size,
29104         spe_padding_size, and spe_64bit_regs_used.  Replace TARGET_SPE and
29105         TARGET_SPE_ABI with 0, simplify.  Replace SPE_VECTOR_MODE with
29106         PAIRED_VECTOR_MODE.
29107         (struct machine_function): Delete field spe_insn_chain_scanned_p.
29108         (spe_func_has_64bit_regs_p): Delete.
29109         (spe_expand_predicate_builtin): Delete.
29110         (spe_expand_evsel_builtin): Delete.
29111         (TARGET_DWARF_REGISTER_SPAN): Do not define.
29112         (TARGET_MEMBER_TYPE_FORCES_BLK): Do not define.
29113         (invalid_e500_subreg): Delete.
29114         (rs6000_legitimize_address): Always force_reg op2 as well, for
29115         paired single memory accesses.
29116         (rs6000_member_type_forces_blk): Delete.
29117         (rs6000_spe_function_arg): Delete.
29118         (rs6000_expand_unop_builtin): Delete SPE handling.
29119         (rs6000_expand_binop_builtin): Ditto.
29120         (spe_expand_stv_builtin): Delete.
29121         (bdesc_2arg_spe): Delete.
29122         (spe_expand_builtin): Delete.
29123         (spe_expand_predicate_builtin): Delete.
29124         (spe_expand_evsel_builtin): Delete.
29125         (rs6000_invalid_builtin): Remove RS6000_BTM_SPE handling.
29126         (spe_init_builtins): Delete.
29127         (spe_func_has_64bit_regs_p): Delete.
29128         (savres_routine_name): Delete "info" parameter.  Adjust callers.
29129         (rs6000_emit_stack_reset): Ditto.
29130         (rs6000_dwarf_register_span): Delete.
29131         * config/rs6000/rs6000.h (TARGET_SPE_ABI, TARGET_SPE,
29132         UNITS_PER_SPE_WORD, SPE_HIGH_REGNO_P, SPE_SIMD_REGNO_P,
29133         SPE_VECTOR_MODE, RS6000_BTM_SPE, RS6000_BUILTIN_E, RS6000_BUILTIN_S):
29134         Delete.
29135         * config/rs6000/rs6000.md (FIRST_SPE_HIGH_REGNO, LAST_SPE_HIGH_REGNO):
29136         Delete.
29137         * config/rs6000/rs6000.opt (-mabi=spe, -mabi=no-spe): Delete.
29138         * config/rs6000/spe.md: Delete every pattern that uses TARGET_SPE.
29139         * config/rs6000/vector.md (absv2sf2, negv2sf2, addv2sf3, subv2sf3,
29140         mulv2sf3, divv2sf3): Delete expanders.
29142 2017-06-07  Segher Boessenkool  <segher@kernel.crashing.org>
29144         config/rs6000/rs6000.md (UNSPEC_MV_CR_GT): Delete.
29146 2017-06-07  Segher Boessenkool  <segher@kernel.crashing.org>
29148         * config/rs6000/rs6000-protos.h (output_e500_flip_gt_bit): Delete.
29149         * config/rs6000/rs6000.c: Ditto.
29151 2017-06-07  Segher Boessenkool  <segher@kernel.crashing.org>
29153         * config/rs6000/predicated.md (rs6000_cbranch_operator): Delete.
29154         * config/rs6000/rs6000.md: Replace rs6000_cbranch_operator by
29155         comparison_operator.
29157 2017-06-07  Segher Boessenkool  <segher@kernel.crashing.org>
29159         * config/rs6000/rs6000.c: Remove everything related to -mfloat-gprs.
29160         * config/rs6000/rs6000.opt: Ditto.
29161         * config/rs6000/t-rtems: Ditto.
29163 2017-06-07  Segher Boessenkool  <segher@kernel.crashing.org>
29165         * config/rs6000/predicates.md: Replace TARGET_E500_DOUBLE and
29166         TARGET_E500_SINGLE by 0, simplify.
29167         * config/rs6000/rs6000.c: Ditto.
29168         (rs6000_option_override_internal): Delete CHECK_E500_OPTIONS.
29169         (spe_build_register_parallel): Delete.
29170         * config/rs6000/rs6000.h: Delete TARGET_E500_SINGLE,
29171         TARGET_E500_DOUBLE, and CHECK_E500_OPTIONS.
29172         * config/rs6000/rs6000.md: Replace TARGET_E500_DOUBLE,
29173         TARGET_E500_SINGLE, and <E500_CONVERT> by 0, simplify.
29174         (E500_CONVERT): Delete.
29175         * config/rs6000/spe.md: Remove many patterns and all define_constants.
29177 2017-06-07  Segher Boessenkool  <segher@kernel.crashing.org>
29179         * config/rs6000/darwin.md: Replace TARGET_FPRS by 1 and simplify.
29180         * config/rs6000/dfp.md: Ditto.
29181         (negdd2, *negdd2_fpr): Merge.
29182         (absdd2, *absdd2_fpr): Merge.
29183         (negtd2, *negtd2_fpr): Merge.
29184         (abstd2, *abstd2_fpr): Merge.
29185         * config/rs6000/e500.h: Delete file.
29186         * config/rs6000/predicates.md (rs6000_cbranch_operator): Replace
29187         TARGET_FPRS by 1 and simplify.
29188         * config/rs6000/rs6000-c.c: Ditto.
29189         * config/rs6000/rs6000.c: Ditto.  Also replace TARGET_SF_SPE and
29190         TARGET_DF_SPE by 0.
29191         * config/rs6000/rs6000.h: Ditto.  Delete TARGET_SF_SPE and
29192         TARGET_DF_SPE.
29193         * config/rs6000/rs6000.md: Ditto.
29194         (floatdidf2, *floatdidf2_fpr): Merge.
29195         (move_from_CR_gt_bit): Delete.
29196         * config/rs6000/spe.md: Replace TARGET_FPRS by 1 and simplify.
29197         (E500_CR_IOR_COMPARE): Delete.
29198         (All patterns that require !TARGET_FPRS): Delete.
29199         * config/rs6000/vsx.md: Replace TARGET_FPRS by 1 and simplify.
29201 2017-06-07  Bin Cheng  <bin.cheng@arm.com>
29203         * passes.def (pass_iv_canon): Move before pass_loop_distribution.
29205 2017-06-07  Bin Cheng  <bin.cheng@arm.com>
29207         * graphds.c (add_edge): Intitialize edge's attached data.
29208         (foll_in_subgraph, dfs_fst_edge, dfs_next_edge): New function
29209         pointer parameter.  Call pointed function on each edge during
29210         graph traversing.  Skip traversing the edge when the function
29211         returns true.
29212         (graphds_dfs, graphds_scc): Ditto.
29213         (for_each_edge): New parameter.  Pass the new parameter to callback
29214         function.
29215         * graphds.h (skip_edge_callback): New function pointer type.
29216         (graphds_dfs, graphds_scc): New function pointer parameter.
29217         (graphds_edge_callback, for_each_edge): New parameter.
29219 2017-06-07  Bin Cheng  <bin.cheng@arm.com>
29221         * tree-vect-data-refs.c (vect_mark_for_runtime_alias_test): Factor
29222         out code checking if runtime alias check is possible to below ...
29223         Call the new function.
29224         * tree-data-ref.c (runtime_alias_check_p): ... to new function.
29225         * tree-data-ref.h (runtime_alias_check_p): New decalaration.
29227 2017-06-07  Marek Polacek  <polacek@redhat.com>
29229         PR sanitizer/80932
29230         * fold-const.c (extract_muldiv_1) <case MINUS_EXPR>: Add
29231         TYPE_OVERFLOW_WRAPS check.
29233 2017-06-07  Bin Cheng  <bin.cheng@arm.com>
29235         * tree-vect-loop-manip.c (vect_do_peeling): Don't skip vector loop
29236         if versioning is required.
29237         * tree-vect-loop.c (vect_analyze_loop_2): Merge niter check for loop
29238         peeling with the check for versioning.
29240 2017-06-07  Bin Cheng  <bin.cheng@arm.com>
29242         * tree-vectorizer.h (vect_build_loop_niters): New parameter.
29243         * tree-vect-loop-manip.c (vect_build_loop_niters): New parameter.
29244         Set true to new parameter if new ssa variable is defined.
29245         (vect_gen_vector_loop_niters): Refactor.  Set range information
29246         for the new vector loop bound variable.
29247         (vect_do_peeling): Ditto.
29249 2017-06-07  Bin Cheng  <bin.cheng@arm.com>
29251         * tree-affine.c (ssa.h): Include header file.
29252         (tree_to_aff_combination): Handle (T1)(X - CST) when inner type
29253         has wrapping overflow behavior.
29255 2017-06-07  Bin Cheng  <bin.cheng@arm.com>
29257         * tree-affine.c (tree_to_aff_combination): Handle (T1)(X + X).
29259 2017-06-07  Bin Cheng  <bin.cheng@arm.com>
29261         (aff_combination_expand): Move (T1)(X *+- CST) simplification to ...
29262         (tree_to_aff_combination): ... here.
29264 2017-06-07  Bin Cheng  <bin.cheng@arm.com>
29266         * tree-ssa-loop-ivopts.c (ivopts_estimate_reg_pressure): New
29267         reg_pressure model function.
29268         (ivopts_global_cost_for_size): Delete.
29269         (determine_set_costs, iv_ca_recount_cost): Call new model function
29270         ivopts_estimate_reg_pressure.
29272 2017-06-07  Tamar Christina  <tamar.christina@arm.com>
29274         * config/aarch64/aarch64.c (aarch64_rtx_costs): Make sdiv more
29275         expensive than udiv.  Remove floating point cases from mod.
29277 2017-06-07  Tamar Christina  <tamar.christina@arm.com>
29279         * config/arm/aarch-cost-tables.h (cortexa53_extra_cost):
29280         Increase idiv cost.
29282 2017-06-07  Tamar Christina  <tamar.christina@arm.com>
29284         * config/aarch64/aarch64.md
29285         (copysignsf3): Fix mask generation.
29287 2017-06-07  Jakub Jelinek  <jakub@redhat.com>
29289         * dumpfile.h (enum tree_dump_index): Rename TDI_generic to
29290         TDI_gimple.
29291         (class dump_manager): Add register_dumps method.
29292         * dumpfile.c: Include langhooks.h.
29293         (dump_files): Use 0 instead of 3/4/5 for TDI_{original,gimple,nested}.
29294         (FIRST_AUTO_NUMBERED_DUMP): Decrease to 1.
29295         (FIRST_ME_AUTO_NUMBERED_DUMP): Define.
29296         (dump_manager::dump_register): Start with 512 entries instead of 32.
29297         (dump_manager::register_dumps): New method.
29298         * toplev.c (general_init): Instead of invoking register_dumps
29299         langhook, invoke register_dumps method on the dump manager.
29300         * gimplify.c (gimplify_function_tree): Use TDI_gimple instead of
29301         TDI_generic.
29303 2017-06-07  Richard Sandiford  <richard.sandiford@linaro.org>
29305         * doc/md.texi: Clarify the restrictions on a define_insn condition.
29306         Say that # requires an associated define_split to exist, and that
29307         the define_split must be suitable for use after register allocation.
29309 2017-06-06  Jan Hubicka  <hubicka@ucw.cz>
29311         * cfgbuild.c (find_bb_boundaries): Initialize profile of split blocks.
29312         (compute_outgoing_frequencies): Also initialize zero counts.
29313         (find_many_sub_basic_blocks): Do not produce uninitialized profile
29314         around loops; preserve more of profile when nothing changes.
29316 2017-06-06  Jim Wilson  <jim.wilson@linaro.org>
29318         * config/aarch64/aarch64-cost-tables.h (qdf24xx_extra_costs): Move to
29319         here.
29320         * config/arm/aarch-cost-tables.h (qdf24xx_extra_costs): From here.
29321         * config/arm/arm-cpu-cdata.h: Regenerate.
29322         * config/arm/arm-cpu-data.h, config/arm/arm-cpu.h: Likewise.
29323         * config/arm/arm-tables.opt, config/arm/arm-tune.md: Likewise.
29324         * config/arm/arm-cpus.in: Delete falkor and qdf24xx entries.
29325         * config/arm/arm.c (arm_qdf24xx_tune): Delete.
29326         * config/arm/bpabi.h (BE8_LINK_SPEC): Delete falkor and qdf24xx
29327         support.
29328         * config/arm/t-aprofile (MULTILIB_MATCHES): Delete falkor and qdf24xx
29329         support.
29330         * config/arm/t-rmprofile: Likewise.
29331         * doc/invoke.texi (ARM Options): Drop falkor and qdf24xx support.
29333 2017-06-06  David S. Miller  <davem@davemloft.net>
29335         PR target/80968
29336         * config/sparc/sparc.c (sparc_expand_prologue): Emit frame
29337         blockage if function uses alloca.
29339 2017-06-06  Maxim Kuvyrkov  <maxim.kuvyrkov@linaro.org>
29341         * tree-ssa-loop-prefetch.c (struct mem_ref_group, struct mem_ref):
29342         New "uid" fields to hold pretty-print IDs of group and ref.
29343         Memory references are now identified as <group_id>:<ref_id>
29344         instead of using [random] addresses.
29345         (dump_mem_details): Simplify, no functional change.
29346         (dump_mem_ref): Simplify and make output more concise.
29347         Replace couple of fprintf's throughout code with calls to dump_mem_ref.
29348         (find_or_create_group): Initialize group uid.
29349         (record_ref): Initialize ref uid.  Improve debug output.
29350         (prune_group_by_reuse, should_issue_prefetch_p,)
29351         (should_issue_prefetch_p, schedule_prefetches, issue_prefetch_ref,)
29352         (mark_nontemporal_store, determine_loop_nest_reuse):
29353         Improve debug output.
29355 2017-06-06  Maxim Kuvyrkov  <maxim.kuvyrkov@linaro.org>
29357         * dbgcnt.def (prefetch): New debug counter.
29358         * tree-ssa-loop-prefetch.c (dbgcnt.h): New include.
29359         (schedule_prefetches): Stop issueing prefetches if debug counter
29360         tripped.
29362 2017-06-06  Tom de Vries  <tom@codesourcery.com>
29364         * doc/sourcebuild.texi (Testsuites, C Language Testsuites,
29365         gcc.c-torture/compile): Remove mention of NO_LABEL_VALUES in fixme.
29367 2017-06-06  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
29369         * config/aarch64/atomics.md (atomic_compare_and_swap<mode> expander):
29370         Use aarch64_reg_or_zero predicate for operand 4.
29371         (aarch64_compare_and_swap<mode> define_insn_and_split):
29372         Use aarch64_reg_or_zero predicate for operand 3.  Add 'Z' constraint.
29373         (aarch64_store_exclusive<mode>): Likewise for operand 2.
29375 2017-06-06  Thomas Preud'homme  <thomas.preudhomme@arm.com>
29377         * config/arm/arm.c (arm_compute_save_reg_mask): Rename into ...
29378         (arm_compute_save_core_reg_mask): This.
29379         (thumb1_compute_save_reg_mask): Rename into ...
29380         (thumb1_compute_save_core_reg_mask): This.
29381         (arm_compute_save_reg0_reg12_mask): Adapt comment.
29382         (arm_compute_frame_layout): Likewise.
29384 2017-06-06  Richard Biener  <rguenther@suse.de>
29386         PR tree-optimization/80974
29387         * tree-ssa-sccvn.c (set_ssa_val_to): Do not change but only
29388         keep or clear leaders SSA info.
29390 2017-06-06  Tom de Vries  <tom@codesourcery.com>
29392         * config/nvptx/nvptx.c (split_mode_p): New function.
29393         (nvptx_declare_function_name, nvptx_print_operand): Use split_mode_p.
29395 2017-06-06  Tom de Vries  <tom@codesourcery.com>
29397         * config/nvptx/nvptx.c (nvptx_print_operand): Use maybe_split_mode.
29399 2017-06-06  Jan Hubicka  <hubicka@ucw.cz>
29401         PR bootstrap/80978
29402         * tree-cfg.c (execute_fixup_cfg): Fix condition on when to rescale
29403         profile.
29405 2017-06-06  Jan Hubicka  <hubicka@ucw.cz>
29407         * shrink-wrap.c (handle_simple_exit): Update profile.
29408         (try_shrink_wrapping): Upate profile.
29410 2017-06-06  Jan Hubicka  <hubicka@ucw.cz>
29412         * predict.c (tree_estimate_probability_bb): Add LOCAL_ONLY.
29413         (tree_guess_outgoing_edge_probabilities): New.
29414         * predict.h (tree_guess_outgoing_edge_probabilities): Declare.
29415         * tree-cfg.c (gimple_find_sub_bbs): Propagate profile.
29417 2017-06-06  Jan Hubicka  <hubicka@ucw.cz>
29419         * ipa-split.c (split_function): Initialize return bb profile.
29421 2017-06-06  Jan Hubicka  <hubicka@ucw.cz>
29423         * profile.c (compute_branch_probabilities): Also initialize
29424         EXIT_BLOCK profile.
29426 2017-06-06  Richard Biener  <rguenther@suse.de>
29428         PR tree-optimization/80928
29429         * tree-vect-loop.c (vect_update_vf_for_slp): Amend dumps.
29430         (vect_analyze_loop_operations): Properly guard analysis for
29431         pure SLP case.
29432         (vect_transform_loop): Likewise.
29433         (vect_analyze_loop_2): Also reset SLP type on PHIs.
29434         (vect_model_induction_cost): Do not cost for pure SLP.
29435         (vectorizable_induction): Pass in SLP node, implement SLP vectorization
29436         of induction in inner loop vectorization.
29437         * tree-vect-slp.c (vect_create_new_slp_node): Handle PHIs.
29438         (vect_get_and_check_slp_defs): Handle vect_induction_def.
29439         (vect_build_slp_tree): Likewise.  Handle PHIs as terminating the
29440         recursion.
29441         (vect_analyze_slp_cost_1): Cost induction.
29442         (vect_detect_hybrid_slp_stmts): Handle PHIs.
29443         (vect_get_slp_vect_defs): Likewise.
29444         * tree-vect-stmts.c (vect_analyze_stmt): Handle induction.
29445         (vect_transform_stmt): Handle SLP reductions.
29446         * tree-vectorizer.h (vectorizable_induction): Adjust.
29448 2017-06-05  Michael Meissner  <meissner@linux.vnet.ibm.com>
29450         * config/rs6000/rs6000.c (make_resolver_func): Update
29451         init_lowered_empty_function call.
29453 2017-06-05  Bernd Edlinger  <bernd.edlinger@hotmail.de>
29455         * doc/invoke.texi: Document the -fprofile-abs-path option.
29456         * common.opt (fprofile-abs-path): New option.
29457         * gcov-io.h (gcov_write_filename): Declare.
29458         * gcov-io.c (gcov_write_filename): New function.
29459         * coverage.c (coverage_begin_function): Use gcov_write_filename.
29460         * profile.c (output_location): Likewise.
29462 2017-06-05  Jan Hubicka  <hubicka@ucw.cz>
29464         * shring-wrap.c: Revert accidental commit.
29466 2017-06-05  Volker Reichelt  <v.reichelt@netcologne.de>
29468         * doc/invoke.texi (-Wduplicated-branches): Add to warning list.
29470 2017-06-05  Jan Hubicka  <hubicka@ucw.cz>
29472         * cfgexpand.c (expand_gimple_tailcall): Initialize profile of
29473         new edge.
29474         * ipa-inline.c (want_inline_self_recursive_call_p): Watch for missing
29475         profile in callgraph edge.
29476         * profile-count.h (apply_probability): If THIS is 0, then result is 0
29477         (apply_scale): Likewise.
29478         * tree-inline.c (copy_bb, copy_edges_for_bb, copy_cfg_body):
29479         Also scale profile when inlining function with zero profile.
29480         (initialize_cfun): Update exit block profile even when it is zero.
29481         * tree-ssa-threadupdate.c (clear_counts_path): Handle correctly case
29482         when profile is read.
29484 2017-06-05  Michael Meissner  <meissner@linux.vnet.ibm.com>
29486         * config/rs6000/rs6000.c (toplevel): Include attribs.h.
29487         (CLONE_*): New constants to define the processors we can generate
29488         code for with the target_clone attribute.
29489         (rs6000_clone_map): New array to identify which clone processors
29490         the current program is running on.
29491         (TARGET_COMPARE_VERSION_PRIORITY): Define to enable the
29492         target_clone attribute.
29493         (TARGET_GENERATE_VERSION_DISPATCHER_BODY): Likewise.
29494         (TARGET_GET_FUNCTION_VERSIONS_DISPATCHER): Likewise.
29495         (TARGET_OPTION_FUNCTION_VERSIONS): Likewise.
29496         (cpu_expand_builtin): Add support for target_clone attribute.
29497         (rs6000_valid_attribute_p): Allow "default" attribute.
29498         (get_decl_name): New debug function to simplify printing the
29499         current function name in debugging statements.
29500         (rs6000_clone_priority): New functions to support the target_clone
29501         attribute, and be able to generate code to switch between ISA 2.05
29502         through ISA 3.0 (power6 through power9).
29503         (rs6000_compare_version_priority): Likewise.
29504         (rs6000_get_function_versions_dispatcher): Likewise.
29505         (make_resolver_func): Likewise.
29506         (add_condition_to_bb): Likewise.
29507         (dispatch_function_versions): Likewise.
29508         (rs6000_generate_version_dispatcher_body): Likewise.
29509         (rs6000_can_inline_p): Call get_decl_name for debugging usage.
29510         (fusion_gpr_load_p): Fix a spacing issue.
29511         * doc/extend.texi (Common Function Attributes): Document that the
29512         PowerPC supports the target_clone attribute.
29514 2017-06-05  Thomas Preud'homme  <thomas.preudhomme@arm.com>
29516         * config/arm/arm.h: explain F symbol found in description of ARM
29517         register allocation in its legend.
29519 2017-06-05  Jan Hubicka  <hubicka@ucw.cz>
29521         * config/mips/frame-header-opt.c: Include profile-count.h.
29522         * config/riscv/riscv.c: Include profile-count.h
29524 2017-06-05  Jan Hubicka  <hubicka@ucw.cz>
29526         * tree-ssa-loop-im.c (execute_sm_if_changed): Add FLAG_BBS parameter;
29527         update profile.
29528         (sm_set_flag_if_changed): Add bbs field.
29529         (execute_sm_if_changed_flag_set): Pass BBS.
29530         (execute_sm): Update.
29532 2017-06-05  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
29534         * config/aarch64/aarch64-simd.md (aarch64_store_lane0<mode>):
29535         New pattern.
29537 2017-06-05  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
29539         * config/aarch64/aarch64.md (sub<mode>3_compare1_imm): New define_insn.
29540         (peephole2): New peephole2 to emit the above.
29541         * config/aarch64/predicates.md (aarch64_sub_immediate): New predicate.
29543 2017-06-05  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
29545         * config/aarch64/aarch64.c (define_peephole2 above
29546         *sub_<shift>_<mode>): New peephole.
29548 2017-05-23  Jan Hubicka  <hubicka@ucw.cz>
29550         * config/i386/i386.c (make_resolver_func): Update.
29551         * Makefile.in: Add profile-count.h and profile-count.o
29552         * auto-profile.c (afdo_indirect_call): Update to new API.
29553         (afdo_set_bb_count): Update.
29554         (afdo_propagate_edge): Update.
29555         (afdo_propagate_circuit): Update.
29556         (afdo_calculate_branch_prob): Update.
29557         (afdo_annotate_cfg): Update.
29558         * basic-block.h: Include profile-count.h
29559         (struct edge_def): Turn count to profile_count.
29560         (struct basic_block_def): Likewie.
29561         (REG_BR_PROB_BASE): Move to profile-count.h
29562         (RDIV): Move to profile-count.h
29563         * bb-reorder.c (max_entry_count): Turn to profile_count.
29564         (find_traces): Update.
29565         (rotate_loop):Update.
29566         (connect_traces):Update.
29567         (sanitize_hot_paths):Update.
29568         * bt-load.c (migrate_btr_defs): Update.
29569         * cfg.c (RDIV): Remove.
29570         (init_flow): Use alloc_block.
29571         (alloc_block): Uninitialize count.
29572         (unchecked_make_edge): Uninitialize count.
29573         (check_bb_profile): Update.
29574         (dump_edge_info): Update.
29575         (dump_bb_info): Update.
29576         (update_bb_profile_for_threading): Update.
29577         (scale_bbs_frequencies_int): Update.
29578         (scale_bbs_frequencies_gcov_type): Update.
29579         (scale_bbs_frequencies_profile_count): New.
29580         * cfg.h (update_bb_profile_for_threading): Update.
29581         (scale_bbs_frequencies_profile_count): Declare.
29582         * cfgbuild.c (compute_outgoing_frequencies): Update.
29583         (find_many_sub_basic_blocks): Update.
29584         * cfgcleanup.c (try_forward_edges): Update.
29585         (try_crossjump_to_edge): Update.
29586         * cfgexpand.c (expand_gimple_tailcall): Update.
29587         (construct_exit_block): Update.
29588         * cfghooks.c (verify_flow_info): Update.
29589         (dump_bb_for_graph): Update.
29590         (split_edge): Update.
29591         (make_forwarder_block): Update.
29592         (duplicate_block): Update.
29593         (account_profile_record): Update.
29594         * cfgloop.c (find_subloop_latch_edge_by_profile): Update.
29595         (get_estimated_loop_iterations): Update.
29596         * cfgloopanal.c (expected_loop_iterations_unbounded): Update.
29597         (single_likely_exit): Update.
29598         * cfgloopmanip.c (scale_loop_profile): Update.
29599         (loopify): Update.
29600         (set_zero_probability): Update.
29601         (lv_adjust_loop_entry_edge): Update.
29602         * cfgrtl.c (force_nonfallthru_and_redirect): Update.
29603         (purge_dead_edges): Update.
29604         (rtl_account_profile_record): Update.
29605         * cgraph.c (cgraph_node::create): Uninitialize count.
29606         (symbol_table::create_edge): Uninitialize count.
29607         (cgraph_update_edges_for_call_stmt_node): Update.
29608         (cgraph_edge::dump_edge_flags): Update.
29609         (cgraph_node::dump): Update.
29610         (cgraph_edge::maybe_hot_p): Update.
29611         * cgraph.h: Include profile-count.h
29612         (create_clone), create_edge, create_indirect_edge): Update.
29613         (cgraph_node): Turn count to profile_count.
29614         (cgraph_edge0: Likewise.
29615         (make_speculative, clone): Update.
29616         (create_edge): Update.
29617         (init_lowered_empty_function): Update.
29618         * cgraphclones.c (cgraph_edge::clone): Update.
29619         (duplicate_thunk_for_node): Update.
29620         (cgraph_node::create_clone): Update.
29621         * cgraphunit.c (cgraph_node::analyze): Update.
29622         (cgraph_node::expand_thunk): Update.
29623         * final.c (dump_basic_block_info): Update.
29624         * gimple-streamer-in.c (input_bb): Update.
29625         * gimple-streamer-out.c (output_bb): Update.
29626         * graphite.c (print_global_statistics): Update.
29627         (print_graphite_scop_statistics): Update.
29628         * hsa-brig.c: Include basic-block.h.
29629         * hsa-dump.c: Include basic-block.h.
29630         * hsa-gen.c (T sum_slice): Update.
29631         (convert_switch_statements):Update.
29632         * hsa-regalloc.c: Include basic-block.h.
29633         * ipa-chkp.c (chkp_produce_thunks): Update.
29634         * ipa-cp.c (struct caller_statistics): Update.
29635         (init_caller_stats): Update.
29636         (gather_caller_stats): Update.
29637         (ipcp_cloning_candidate_p): Update.
29638         (good_cloning_opportunity_p): Update.
29639         (get_info_about_necessary_edges): Update.
29640         (dump_profile_updates): Update.
29641         (update_profiling_info): Update.
29642         (update_specialized_profile): Update.
29643         (perhaps_add_new_callers): Update.
29644         (decide_about_value): Update.
29645         (ipa_cp_c_finalize): Update.
29646         * ipa-devirt.c (struct odr_type_warn_count): Update.
29647         (struct decl_warn_count): Update.
29648         (struct final_warning_record): Update.
29649         (possible_polymorphic_call_targets): Update.
29650         (ipa_devirt): Update.
29651         * ipa-fnsummary.c (redirect_to_unreachable): Update.
29652         * ipa-icf.c (sem_function::merge): Update.
29653         * ipa-inline-analysis.c (do_estimate_edge_time): Update.
29654         * ipa-inline.c (compute_uninlined_call_time): Update.
29655         (compute_inlined_call_time): Update.
29656         (want_inline_small_function_p): Update.
29657         (want_inline_self_recursive_call_p): Update.
29658         (edge_badness): Update.
29659         (lookup_recursive_calls): Update.
29660         (recursive_inlining): Update.
29661         (inline_small_functions): Update.
29662         (dump_overall_stats): Update.
29663         (dump_inline_stats): Update.
29664         * ipa-profile.c (ipa_profile_generate_summary): Update.
29665         (ipa_propagate_frequency): Update.
29666         (ipa_profile): Update.
29667         * ipa-prop.c (ipa_make_edge_direct_to_target): Update.
29668         * ipa-utils.c (ipa_merge_profiles): Update.
29669         * loop-doloop.c (doloop_modify): Update.
29670         * loop-unroll.c (report_unroll): Update.
29671         (unroll_loop_runtime_iterations): Update.
29672         * lto-cgraph.c (lto_output_edge): Update.
29673         (lto_output_node): Update.
29674         (input_node): Update.
29675         (input_edge): Update.
29676         (merge_profile_summaries): Update.
29677         * lto-streamer-in.c (input_cfg): Update.
29678         * lto-streamer-out.c (output_cfg): Update.
29679         * mcf.c (create_fixup_graph): Update.
29680         (adjust_cfg_counts): Update.
29681         (sum_edge_counts): Update.
29682         * modulo-sched.c (sms_schedule): Update.
29683         * postreload-gcse.c (eliminate_partially_redundant_load): Update.
29684         * predict.c (maybe_hot_count_p): Update.
29685         (probably_never_executed): Update.
29686         (dump_prediction): Update.
29687         (combine_predictions_for_bb): Update.
29688         (propagate_freq): Update.
29689         (handle_missing_profiles): Update.
29690         (counts_to_freqs): Update.
29691         (rebuild_frequencies): Update.
29692         (force_edge_cold): Update.
29693         * predict.h: Include profile-count.h
29694         (maybe_hot_count_p, counts_to_freqs): UPdate.
29695         * print-rtl-function.c: Do not include cfg.h
29696         * print-rtl.c: Include basic-block.h
29697         * profile-count.c: New file.
29698         * profile-count.h: New file.
29699         * profile.c (is_edge_inconsistent): Update.
29700         (correct_negative_edge_counts): Update.
29701         (is_inconsistent): Update.
29702         (set_bb_counts): Update.
29703         (read_profile_edge_counts): Update.
29704         (compute_frequency_overlap): Update.
29705         (compute_branch_probabilities): Update; Initialize and deinitialize
29706         gcov_count tables.
29707         (branch_prob): Update.
29708         * profile.h (bb_gcov_counts, edge_gcov_counts): New.
29709         (edge_gcov_count): New.
29710         (bb_gcov_count): New.
29711         * shrink-wrap.c (try_shrink_wrapping): Update.
29712         * tracer.c (better_p): Update.
29713         * trans-mem.c (expand_transaction): Update.
29714         (ipa_tm_insert_irr_call): Update.
29715         (ipa_tm_insert_gettmclone_call): Update.
29716         * tree-call-cdce.c: Update.
29717         * tree-cfg.c (gimple_duplicate_sese_region): Update.
29718         (gimple_duplicate_sese_tail): Update.
29719         (gimple_account_profile_record): Update.
29720         (execute_fixup_cfg): Update.
29721         * tree-inline.c (copy_bb): Update.
29722         (copy_edges_for_bb): Update.
29723         (initialize_cfun): Update.
29724         (freqs_to_counts): Update.
29725         (copy_cfg_body): Update.
29726         (expand_call_inline): Update.
29727         * tree-ssa-ifcombine.c (update_profile_after_ifcombine): Update.
29728         * tree-ssa-loop-ivcanon.c (unloop_loops): Update.
29729         (try_unroll_loop_completely): Update.
29730         (try_peel_loop): Update.
29731         * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Update.
29732         * tree-ssa-loop-niter.c (estimate_numbers_of_iterations_loop): Update.
29733         * tree-ssa-loop-split.c (connect_loops): Update.
29734         * tree-ssa-loop-unswitch.c (hoist_guard): Update.
29735         * tree-ssa-reassoc.c (branch_fixup): Update.
29736         * tree-ssa-tail-merge.c (replace_block_by): Update.
29737         * tree-ssa-threadupdate.c (create_block_for_threading): Update.
29738         (compute_path_counts): Update.
29739         (update_profile): Update.
29740         (recompute_probabilities): Update.
29741         (update_joiner_offpath_counts): Update.
29742         (estimated_freqs_path): Update.
29743         (freqs_to_counts_path): Update.
29744         (clear_counts_path): Update.
29745         (ssa_fix_duplicate_block_edges): Update.
29746         (duplicate_thread_path): Update.
29747         * tree-switch-conversion.c (case_bit_test_cmp): Update.
29748         (struct switch_conv_info): Update.
29749         * tree-tailcall.c (decrease_profile): Update.
29750         * tree-vect-loop-manip.c (slpeel_add_loop_guard): Update.
29751         * tree-vect-loop.c (scale_profile_for_vect_loop): Update.
29752         * value-prof.c (check_counter): Update.
29753         (gimple_divmod_fixed_value): Update.
29754         (gimple_mod_pow2): Update.
29755         (gimple_mod_subtract): Update.
29756         (gimple_ic_transform): Update.
29757         (gimple_stringop_fixed_value): Update.
29758         * value-prof.h (gimple_ic): Update.
29760 2017-06-02  Carl Love  <cel@us.ibm.com>
29762         * config/rs6000/rs6000-c: Add support for built-in functions
29763         vector double vec_doublee (vector signed int);
29764         vector double vec_doublee (vector unsigned int);
29765         vector double vec_doublee (vector float);
29766         vector double vec_doubleh (vector signed int);
29767         vector double vec_doubleh (vector unsigned int);
29768         vector double vec_doubleh (vector float);
29769         vector double vec_doublel (vector signed int);
29770         vector double vec_doublel (vector unsigned int);
29771         vector double vec_doublel (vector float);
29772         vector double vec_doubleo (vector signed int);
29773         vector double vec_doubleo (vector unsigned int);
29774         vector double vec_doubleo (vector float);.
29775         * config/rs6000/rs6000-builtin.def: Add definitions for DOUBLEE,
29776         DOUBLEO, DOUBLEH, DOUBLEL, UNS_DOUBLEO, UNS_DOUBLEE, UNS_DOUBLEH,
29777         UNS_DOUBLEL.
29778         * config/rs6000/altivec.md: Add code generator for doublee<mode>2,
29779         unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2, doubleh<mode>2,
29780         unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2, add mode attribute
29781         VS_sxwsp.
29782         * config/rs6000/altivec.h: Add define for vec_doublee, vec_doubleo,
29783         vec_doublel, vec_doubleh.
29784         * doc/extend.texi: Update the built-in documentation file for the
29785         new built-in functions.
29787 2017-06-02  David Malcolm  <dmalcolm@redhat.com>
29789         PR jit/80954
29790         * ipa-inline-analysis.c (free_growth_caches): Set
29791         edge_removal_hook_holder to NULL after removing it.
29793 2017-06-02  Sudakshina Das  <sudi.das@arm.com>
29795         * config/aarch64/aarch64.c (aarch64_select_cc_mode): Return CC_SWP for
29796         comparision with zero.
29798 2017-06-02  Will Schmidt  <will_schmidt@vnet.ibm.com>
29799         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add handling
29800         for early expansion of vec_min and vec_max builtins.
29801         (builtin_function_type): Add min/max unsigned variants to those
29802         identified as having unsigned arguments.
29804 2017-06-02  Olivier Hainque  <hainque@adacore.com>
29806         * config/vx-common.h (DWARF_UNWIND_INFO): Switch #define to 1.
29808 2017-06-02  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
29810         * config/aarch64/aarch64-simd.md (*aarch64_simd_vec_copy_lane<mode>):
29811         Use VALL_F16 iterator rather than VALL.
29813 2017-06-02  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
29815         * config/aarch64/aarch64.c (aarch64_split_compare_and_swap):
29816         Emit CBNZ inside loop when doing a strong exchange and comparing
29817         against zero.  Generate the CC flags after the loop.
29819 2017-06-02  David Edelsohn  <dje.gcc@gmail.com>
29821         * dwarf2out.c (DWARF_INITIAL_LENGTH_SIZE_STR): New.
29822         (dl_section_ref): New.
29823         (dwarf2out_finish): Copy debug_line_section_label to dl_section_ref.
29824         On AIX, append an expression to subtract the size of the
29825         section length to dl_section_ref.
29827 2017-06-02  Will Schmidt  <will_schmidt@vnet.ibm.com>
29829         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add handling
29830         for early expansion of vector absolute builtins.
29832 2017-06-02  Richard Biener  <rguenther@suse.de>
29834         * tree-vect-slp.c (vect_detect_hybrid_slp_2): Match up
29835         what we consider a relevant use stmt with vect_detect_hybrid_slp_stmts.
29837 2017-06-02  Richard Biener  <rguenther@suse.de>
29839         PR tree-optimization/80948
29840         * tree-tailcall.c (find_tail_calls): Track stmts to move in
29841         stmt order as well.
29843 2017-06-02  Richard Biener  <rguenther@suse.de>
29845         * tree-vect-loop.c (vect_analyze_loop_operations): Not relevant
29846         PHIs are ok.
29847         * tree-vect-stmts.c (process_use): Do not mark backedge defs
29848         for inductions as relevant.
29850 2017-06-02  Richard Biener  <rguenther@suse.de>
29852         * tree-vect-loop.c (get_initial_def_for_induction): Inline into ...
29853         (vectorizable_induction): ... this.  Remove dead code.
29855 2017-06-02  Eric Botcazou  <ebotcazou@adacore.com>
29857         * builtins. (expand_builtin_alloca): Remove second parameter and
29858         infer its value from the first parameter instead.
29859         (expand_builtin) <BUILT_IN_ALLOCA>: Adjust call to above.
29861 2017-06-02  Jakub Jelinek  <jakub@redhat.com>
29863         PR rtl-optimization/80903
29864         * loop-doloop.c (add_test): Unshare sequence.
29866 2017-06-02  Bernd Edlinger  <bernd.edlinger@hotmail.de>
29868         * doc/invoke.texi: Document the -Wsizeof-pointer-div warning.
29870 2017-06-01  Bernd Edlinger  <bernd.edlinger@hotmail.de>
29872         * config/i386/i386.c (x86_64_ms_sysv_extra_clobbered_registers): Make
29873         static.
29874         (xlogue_layout::get_stack_space_used, xlogue_layout::s_instances,
29875         xlogue_layout::get_instance, logue_layout::xlogue_layout,
29876         sp_valid_at, fp_valid_at, choose_basereg): Formatting.
29877         (xlogue_layout::get_stub_rtx): Make static.
29878         (xlogue_layout::get_stub_name): Avoid const-cast, make static.
29879         (xlogue_layout::compute_stub_managed_regs): Rename to...
29880         (xlogue_layout::count_stub_managed_regs): ...this.
29881         (xlogue_layout::is_stub_managed_reg): New function.
29882         (xlogue_layout::m_stub_names): Rename to...
29883         (xlogue_layout::s_stub_names): ...this, make static.
29884         (xlogue_layout::STUB_INDEX_OFFSET, xlogue_layout::MIN_REGS,
29885         xlogue_layout::MAX_REGS, xlogue_layout::MAX_EXTRA_REGS,
29886         xlogue_layout::VARIANT_COUNT, xlogue_layout::STUB_NAME_MAX_LEN,
29887         xlogue_layout::s_stub_names): Instantiate statics.
29888         (stub_managed_regs): Remove.
29889         (ix86_save_reg): Use xlogue_layout::compute_stub_managed_regs.
29890         (disable_call_ms2sysv_xlogues): Rename to...
29891         (warn_once_call_ms2sysv_xlogues): ...this, and warn only once.
29892         (ix86_initial_elimination_offset, ix86_expand_call): Fix call_ms2sysv
29893         warning logic.
29894         (ix86_static_chain): Make sure that ix86_static_chain_on_stack can't
29895         change after reload_completed.
29896         (ix86_can_use_return_insn_p): Use the ix86_frame data structure
29897         directly.
29898         (ix86_expand_prologue): Likewise.
29899         (ix86_expand_epilogue): Likewise.
29900         (ix86_expand_split_stack_prologue): Likewise.
29901         (ix86_compute_frame_layout): Remove frame parameter ...
29902         (TARGET_COMPUTE_FRAME_LAYOUT): ... and export it as a target hook.
29903         (ix86_finalize_stack_realign_flags): Call ix86_compute_frame_layout
29904         only if necessary.
29905         (ix86_init_machine_status): Don't set use_fast_prologue_epilogue_nregs.
29906         (ix86_frame): Move from here ...
29907         * config/i386/i386.h (ix86_frame): ... to here.
29908         (machine_function): Remove use_fast_prologue_epilogue_nregs, cache the
29909         complete ix86_frame data structure instead.  Remove some_ld_name.
29911 2017-06-01  Pierre-Marie de Rodat  <derodat@adacore.com>
29913         * dwarf2out.c (dwarf2out_late_global_decl): Add locations for
29914         symbols that hold a DECL_VALUE_EXPR.
29916 2017-06-01  Martin Jambor  <mjambor@suse.cz>
29918         PR tree-optimization/80898
29919         * tree-sra.c (process_subtree_disqualification): Removed.
29920         (disqualify_candidate): Do not acll
29921         process_subtree_disqualification.
29922         (subtree_mark_written_and_enqueue): New function.
29923         (propagate_all_subaccesses): Set grp_write of LHS subtree if the
29924         RHS has been disqualified and re-queue LHS if necessary.  Apart
29925         from that, ignore disqualified RHS.
29927 2017-06-01  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
29929         * config/s390/s390.c (s390_emit_epilogue): Disable early return
29930         address fetch for z10 or later.
29932 2017-06-01  Claudiu Zissulescu  <claziss@synopsys.com>
29934         * config/arc/arc.md (tst_movb): Add guard when splitting.
29936 2017-06-01  Claudiu Zissulescu  <claziss@synopsys.com>
29938         * config/arc/arc.c (arc_can_eliminate): Test against
29939         arc_frame_pointer_needed.
29941 2017-06-01  Claudiu Zissulescu  <claziss@synopsys.com>
29943         * config/arc/arc.c (arc_expand_prologue): Emit a special barrier
29944         to prevent store reordering.
29945         * config/arc/arc.md (UNSPEC_ARC_STKTIE): Define.
29946         (type): Add block type.
29947         (stack_tie): Define special instruction to be used in
29948         expand_prologue.
29950 2017-06-01  Claudiu Zissulescu  <claziss@synopsys.com>
29952         * config/arc/arc.md (commutative_binary_comparison): Remove 'I'
29953         constraint. It is not valid for the pattern.
29954         (noncommutative_binary_comparison): Likewise.
29956 2017-06-01  Claudiu Zissulescu  <claziss@synopsys.com>
29958         * config/arc/simdext.md (movv2hi_insn): Change predicate to avoid
29959         scaled addresses.
29961 2017-06-01  Claudiu Zissulescu  <claziss@synopsys.com>
29963         * config/arc/arc.c (arc_conditional_register_usage): Allow r30 to
29964         be used by the reg-alloc.
29966 2017-06-01  Claudiu Zissulescu  <claziss@synopsys.com>
29968         * config/arc/arc.md (mulsi3): Avoid use of hard registers before
29969         reg-alloc when having mul64 or mul32x16 instructions.
29970         (mulsidi3): Likewise.
29971         (umulsidi3): Likewise.
29972         (mulsi32x16): New pattern.
29973         (mulsi64): Likewise.
29974         (mulsidi64): Likewise.
29975         (umulsidi64): Likewise.
29976         (MUL32x16_REG): Define.
29977         (mul64_600): Use MUL32x16_REG.
29978         (mac64_600): Likewise.
29979         (umul64_600): Likewise.
29980         (umac64_600): Likewise.
29982 2017-06-01  Claudiu Zissulescu  <claziss@synopsys.com>
29984         * config/arc/arc.md (mulsi3_700): Make it commutative.
29986 2017-06-01  Jose E. Marchesi  <jose.marchesi@oracle.com>
29988         * config/sparc/sparc.md (*zero_extendsidi2_insn_sp64): Set insn
29989         type for movstouw.
29990         (*sign_extendsidi2_insn): Likewise for movstosw.
29992 2017-06-01  Pierre-Marie de Rodat  <derodat@adacore.com>
29994         * dwarf2out.c (get_discr_value): Call the get_debug_type hook on
29995         the type of the input discriminant value.  Convert the
29996         discriminant value of signedness vary.
29998 2017-06-01  Volker Reichelt  <v.reichelt@netcologne.de>
30000         * doc/invoke.texi (-Wcatch-value): Document new shortcut.
30001         Add to -Wall section.
30003 2017-06-01  Richard Biener  <rguenther@suse.de>
30005         PR middle-end/66313
30006         * fold-const.c (fold_plusminus_mult_expr): If the factored
30007         factor may be zero use a wrapping type for the inner operation.
30008         * tree-tailcall.c (independent_of_stmt_p): Pass in to_move bitmap
30009         and handle moved defs.
30010         (process_assignment): Properly guard the unary op case.  Return a
30011         tri-state indicating that moving the stmt before the call may allow
30012         to continue.  Pass through to_move.
30013         (find_tail_calls): Handle moving unrelated defs before
30014         the call.
30016 2017-05-31  Segher Boessenkool  <segher@kernel.crashing.org>
30018         PR target/80618
30019         * config/rs6000/vector.md (*vector_uneq<mode>): Write the nor in the
30020         splitter result in the canonical way.
30022 2017-05-31  Uros Bizjak  <ubizjak@gmail.com>
30024         * config/i386/i386.md (*zero_extendsidi2): Enable alternative (?r, *Yj)
30025         also for 32bit target.  Update insn attributes.
30026         (zero-extendsidi2 splitter): Allow all registers for operand 1.
30028 2017-05-31  Sebastian Peryt  <sebastian.peryt@intel.com>
30030         * config/i386/avx512fintrin.h (_mm_mask_max_sd)
30031         (_mm_maskz_max_sd, _mm_mask_max_ss, _mm_maskz_max_ss)
30032         (_mm_mask_min_sd, _mm_maskz_min_sd, _mm_mask_min_ss)
30033         (_mm_maskz_min_ss): New intrinsics.
30035 2017-05-31  Martin Liska  <mliska@suse.cz>
30037         * tree-vect-loop.c (vect_create_epilog_for_reduction):
30038         Change comment style to one we normally use.
30039         (vectorizable_reduction): Likewise.
30040         (vectorizable_induction): Likewise.
30041         * tree-vect-stmts.c (vectorizable_mask_load_store): Likewise.
30042         (vectorizable_call): Likewise.
30043         (vectorizable_simd_clone_call): Likewise.
30044         (vectorizable_conversion): Likewise.
30045         (vectorizable_assignment): Likewise.
30046         (vectorizable_shift): Likewise.
30047         (vectorizable_operation): Likewise.
30048         (vectorizable_store): Likewise.
30049         (vectorizable_load): Likewise.
30050         * tree-vectorizer.h: Likewise.
30052 2017-05-31  Alexander Monakov  <amonakov@ispras.ru>
30054         * passes.c (emergency_dump_function): New.
30055         * tree-pass.h (emergency_dump_function): Declare.
30056         * plugin.c (plugins_internal_error_function): Remove.
30057         * plugin.h (plugins_internal_error_function): Remove declaration.
30058         * toplev.c (internal_error_function): New static function.  Use it...
30059         (general_init): ...here.
30061 2017-05-31  Graham Markall  <graham.markall@embecosm.com>
30063         * config/arc/arc.c (arc_print_operand): Handle constant operands.
30064         (arc_rtx_costs): Add costs for new patterns.
30065         * config/arc/arc.md: Additional *add_n and *sub_n patterns.
30066         * config/arc/predicates.md: Add _1_2_3_operand predicate.
30068 2017-05-31  Richard Sandiford  <richard.sandiford@linaro.org>
30070         * tree-ssa-strlen.c (get_next_strinfo): New function.
30071         (get_stridx_plus_constant): Use it.
30072         (zero_length_string): Likewise.
30073         (adjust_related_strinfos): Likewise.
30074         (adjust_last_stmt): Likewise.
30076 2017-05-31  Richard Biener  <rguenther@suse.de>
30078         PR target/80880
30079         * config/i386/i386.c (ix86_expand_builtin): Remove assert
30080         for arg being an SSA name when expanding IX86_BUILTIN_BNDRET.
30082 2017-05-31  Richard Sandiford  <richard.sandiford@linaro.org>
30084         * tree-vect-data-refs.c (vect_find_same_alignment_drs): Remove
30085         loop_vinfo argument and use of dependence distance vectors.
30086         Check instead whether the two references differ only in their
30087         initial value and assume that they have the same alignment if the
30088         difference is a multiple of the vector alignment.
30089         (vect_analyze_data_refs_alignment): Update call accordingly.
30091 2017-05-31  Martin Liska  <mliska@suse.cz>
30093         PR target/79155
30094         * config/i386/cpuid.h: Fix typo in a comment in cpuid.h.
30096 2017-05-31  Bin Cheng  <bin.cheng@arm.com>
30098         * tree-vect-loop-manip.c (create_intersect_range_checks_index)
30099         (create_intersect_range_checks): Move from ...
30100         * tree-data-ref.c (create_intersect_range_checks_index)
30101         (create_intersect_range_checks): ... to here.
30102         (create_runtime_alias_checks): New function factored from ...
30103         * tree-vect-loop-manip.c (vect_create_cond_for_alias_checks): ...
30104         here.  Call above function.
30105         * tree-data-ref.h (create_runtime_alias_checks): New function.
30107 2017-05-31  Bin Cheng  <bin.cheng@arm.com>
30109         * tree-data-ref.c (prune_runtime_alias_test_list): Relax minimal
30110         segment length for dr_b and compute it in wide_int.
30112 2017-05-31  Richard Biener  <rguenther@suse.de>
30114         PR tree-optimization/80906
30115         * graphite-isl-ast-to-gimple.c (copy_loop_close_phi_nodes): Get
30116         and pass through iv_map.
30117         (copy_bb_and_scalar_dependences): Adjust.
30118         (translate_pending_phi_nodes): Likewise.
30119         (copy_loop_close_phi_args): Handle code-generating IVs instead
30120         of ICEing.
30122 2017-05-30  David Malcolm  <dmalcolm@redhat.com>
30124         * diagnostic-color.c (color_dict): Add "type-diff".
30125         (parse_gcc_colors): Update comment.
30126         * doc/invoke.texi (Diagnostic Message Formatting Options): Add
30127         -fdiagnostics-show-template-tree and -fno-elide-type.
30128         (GCC_COLORS): Add type-diff to example.
30129         (type-diff=): New.
30130         (-fdiagnostics-show-template-tree): New.
30131         (-fno-elide-type): New.
30132         * pretty-print.c (pp_format): Pass quote and formatters[argno] to
30133         the pp_format_decoder callback.  Call any m_format_postprocessor's
30134         "handle" method.
30135         (pretty_printer::pretty_printer): Initialize
30136         m_format_postprocessor.
30137         (pretty_printer::~pretty_printer): Delete any
30138         m_format_postprocessor.
30139         * pretty-print.h (printer_fn): Add bool and const char ** parameters.
30140         (class format_postprocessor): New class.
30141         (struct pretty_printer::format_decoder): Document the new parameters.
30142         (struct pretty_printer::m_format_postprocessor): New field.
30143         * tree-diagnostic.c (default_tree_printer): Update for new
30144         bool and const char ** params.
30145         * tree-diagnostic.h (default_tree_printer): Likewise.
30147 2017-05-30  Segher Boessenkool  <segher@kernel.crashing.org>
30149         * config/rs6000/predicates.md (cc_reg_not_micro_cr0_operand): Delete.
30150         (lwa_operand): Delete rs6000_gen_cell_microcode test.
30151         * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
30152         rs6000_gen_cell_microcode code.
30153         (rs6000_final_prescan_insn): Delete.
30154         (rs6000_opt_vars): Delete the "gen-cell-microcode" and
30155         "warn-cell-microcode" entries.
30156         * config/rs6000/rs6000.h (FINAL_PRESCAN_INSN): Delete.
30157         * config/rs6000/rs6000.md: Delete rs6000_gen_cell_microcode tests
30158         throughout.  Change cc_reg_not_micro_cr0_operand to
30159         cc_reg_not_cr0_operand throughout.
30160         (*extendhi<mode>2_noload): Delete.
30161         * config/rs6000/rs6000.opt (mgen-cell-microcode): Replace by stub.
30162         (mwarn-cell-microcode): Delete.
30163         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete
30164         -mgen-cell-microcode and -mwarn-cell-microcode.
30166 2017-05-30  Uros Bizjak  <ubizjak@gmail.com>
30168         PR target/80833
30169         * config/i386/constraints.md (Yd): New constraint.
30170         (Ye): Ditto.
30171         * config/i386/i386.md (*movti_internal): Add (?r, Ye)
30172         and (?Yd, r) alternatives.  Update insn attributes.
30173         * config/i386/i386.md (*movti_internal): Add (?r, *Ye)
30174         and (?*Yd, r) alternatives.  Update insn attributes.
30175         (double-mode inter-unit splitters): Add new GR<->XMM splitters.
30177 2017-05-30  Pierre-Marie de Rodat  <derodat@adacore.com>
30179         * gimplify.c (gimplify_modify_expr): Don't create a
30180         DECL_DEBUG_EXPR link if *FROM_P does not belong to the current
30181         function.
30183 2017-05-30  Wilco Dijkstra  <wdijkstr@arm.com>
30185         * config/arm/arm-builtins.c (arm_expand_builtin): Remove const.
30187 2017-05-30  Richard Biener  <rguenther@suse.de>
30189         * tree-vectorizer.h (struct _stmt_vec_info): Add reduc_type
30190         and reduc_def fields.
30191         (STMT_VINFO_REDUC_TYPE): New define.
30192         (STMT_VINFO_REDUC_DEF): Likewise.
30193         (vect_force_simple_reduction): Adjust prototype.
30194         * tree-vect-loop.c (vect_analyze_scalar_cycles_1): Adjust.
30195         (vect_is_simple_reduction): Remove check_reduction argument.
30196         (vect_force_simple_reduction): Adjust and set
30197         STMT_VINFO_REDUC_TYPE and STMT_VINFO_REDUC_DEF.
30198         (vectorizable_reduction): Do not re-do reduction analysis
30199         but use STMT_VINFO_REDUC_TYPE and STMT_VINFO_REDUC_DEF.
30200         * tree-parloops.c (gather_scalar_reductions): Adjust.
30202 2017-05-30  Richard Biener  <rguenther@suse.de>
30204         PR middle-end/80901
30205         * cfgexpand.c (expand_gimple_cond): Match up loop fixup with
30206         split_edge code.
30208 2017-05-24  Robin Dapp  <rdapp@linux.vnet.ibm.com>
30210         * tree-vect-data-refs.c (vect_get_peeling_costs_all_drs):
30211         Introduce unknown_misalignment parameter and remove vf.
30212         (vect_peeling_hash_get_lowest_cost):
30213         Pass unknown_misalignment parameter.
30214         (vect_enhance_data_refs_alignment):
30215         Fix unsupportable data ref treatment.
30217 2017-05-30  Robin Dapp  <rdapp@linux.vnet.ibm.com>
30219         * tree-vect-data-refs.c (vect_get_data_access_cost):
30220         Workaround for SLP handling.
30221         (vect_enhance_data_refs_alignment):
30222         Compute costs for doing no peeling at all, compare to the best
30223         peeling costs so far and avoid peeling if cheaper.
30225 2017-05-30  Robin Dapp  <rdapp@linux.vnet.ibm.com>
30227         * tree-vect-data-refs.c (vect_peeling_hash_choose_best_peeling):
30228         Return peeling info and set costs to zero for unlimited cost
30229         model.
30230         (vect_enhance_data_refs_alignment): Also inspect all datarefs
30231         with unknown misalignment. Compute and costs for unknown
30232         misalignment, compare them to the costs for known misalignment
30233         and choose the cheapest for peeling.
30235 2017-05-30  Robin Dapp  <rdapp@linux.vnet.ibm.com>
30237         * tree-vect-data-refs.c (vect_update_misalignment_for_peel): Rename.
30238         (vect_get_peeling_costs_all_drs): Create function.
30239         (vect_peeling_hash_get_lowest_cost):
30240         Use vect_get_peeling_costs_all_drs.
30241         (vect_peeling_supportable): Create function.
30242         (vect_enhance_data_refs_alignment): Use vect_peeling_supportable.
30244 2017-05-30  Robin Dapp  <rdapp@linux.vnet.ibm.com>
30246         * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Create
30247         DR_HAS_NEGATIVE_STEP.
30248         (vect_update_misalignment_for_peel): Define DR_MISALIGNMENT.
30249         (vect_enhance_data_refs_alignment): Use.
30250         (vect_duplicate_ssa_name_ptr_info): Use.
30251         * tree-vectorizer.h (dr_misalignment): Use.
30252         (known_alignment_for_access_p): Use.
30254 2017-05-30  Jozef Lawrynowicz  <jozef.l@somniumtech.com>
30256         PR target/78838
30257         * config/msp430/msp430.c (gen_prefix): Return NULL when section name is
30258         .lowtext.
30259         (has_section_name): New function.
30261 2017-05-30  Martin Liska  <mliska@suse.cz>
30263         PR other/80909
30264         * auto-profile.c (get_function_decl_from_block): Fix
30265         parenthesis.
30267 2017-05-30  Richard Biener  <rguenther@suse.de>
30269         PR middle-end/80876
30270         * cfgexpand.c (expand_gimple_cond): Fixup preserving loops again.
30272 2017-05-30  Martin Liska  <mliska@suse.cz>
30274         * dumpfile.c: Use newly added macro DUMP_FILE_INFO.
30275         * dumpfile.h (struct dump_file_info): Remove ctors.
30277 2017-05-30  Martin Liska  <mliska@suse.cz>
30279         * predict.def: Fix GNU coding style.
30281 2017-05-29  Max Filippov  <jcmvbkbc@gmail.com>
30283         * config/xtensa/xtensa.c (xtensa_initial_elimination_offset):
30284         Mark 'to' argument with ATTRIBUTE_UNUSED.
30286 2017-05-29  Max Filippov  <jcmvbkbc@gmail.com>
30288         * config/xtensa/xtensa.c (xtensa_emit_call): Use
30289         HOST_WIDE_INT_PRINT_HEX instead of 0x%lx format string.
30290         (print_operand): Use HOST_WIDE_INT_PRINT_DEC instead of %ld
30291         format string.
30293 2017-05-29  Eric Botcazou  <ebotcazou@adacore.com>
30295         * doc/install.texi (Options specification): Restore entry of
30296         --enable-sjlj-exceptions.
30298 2017-05-27  Michael Eager  <eager@eagercon.com>
30300         Revert:
30301         2016-01-21  Ajit Agarwal  <ajitkum@xilinx.com>
30303         See https://gcc.gnu.org/ml/gcc/2017-05/msg00221.html.
30305         * config/microblaze/microblaze.h
30306         (FIXED_REGISTERS): Update in macro.
30307         (CALL_USED_REGISTERS): Update in macro.
30309 2017-05-27  François-Xavier Coudett  <fxcoudert@gcc.gnu.org>
30311         * doc/install.texi: Add links to macOS binary distributions.
30313 2017-05-27  Jakub Jelinek  <jakub@redhat.com>
30315         PR bootstrap/80887
30316         Revert:
30317         2017-05-25  Marc Glisse  <marc.glisse@inria.fr>
30319         * match.pd ((A +- CST1) +- CST2): Allow some conversions.
30321 2017-05-26  Martin Liska  <mliska@suse.cz>
30323         * dumpfile.h (enum dump_kind): Renumber TDF_* flags to be contiguous.
30325 2017-05-26  Martin Liska  <mliska@suse.cz>
30327         * cfg.c (check_bb_profile): Do not use TDF_COMMENT and print
30328         always leading ';; '.
30329         (dump_bb_info): Likewise.
30330         (brief_dump_cfg): Likewise.
30331         * cfgrtl.c (print_rtl_with_bb): Do not use TDF_COMMENT.
30332         * dumpfile.c: Remove usage of TDF_VERBOSE.
30333         * dumpfile.h (enum dump_kind): Likewise.
30334         (dump_gimple_bb_header): Do not use TDF_COMMENT.
30335         * print-tree.c (debug_verbose): Remove.
30336         * tree-cfg.c (gimple_dump_cfg): Do not use TDF_COMMENT.
30337         (dump_function_to_file): Remove dumps guarded with TDF_VERBOSE.
30338         * tree-diagnostic.c (default_tree_printer): Replace
30339         TDF_DIAGNOSTIC with TDF_SLIM.
30341 2017-05-26  Bin Cheng  <bin.cheng@arm.com>
30343         * tree-vect-loop-manip.c (create_intersect_range_checks_index): Pass
30344         in parameter loop, rather than loop_vinfo.
30345         (create_intersect_range_checks): Ditto.
30346         (vect_create_cond_for_alias_checks): Update call to above functions.
30348 2017-05-26  Bin Cheng  <bin.cheng@arm.com>
30350         PR tree-optimization/80815
30351         * tree-data-ref.c (prune_runtime_alias_test_list): Simplify condition
30352         for merging runtime alias checks.  Handle negative DR_STEPs.
30354 2017-05-26  Bin Cheng  <bin.cheng@arm.com>
30356         * tree-vect-data-refs.c (Operator==, comp_dr_with_seg_len_pair):
30357         Move from ...
30358         * tree-data-ref.c (Operator==, comp_dr_with_seg_len_pair): To here.
30359         * tree-vect-data-refs.c (vect_prune_runtime_alias_test_list): Factor
30360         out code pruning runtime alias checks.
30361         * tree-data-ref.c (prune_runtime_alias_test_list): New function
30362         factored out from above.
30363         * tree-vectorizer.h (struct dr_with_seg_len, dr_with_seg_len_pair_t):
30364         Move from ...
30365         * tree-data-ref.h (struct dr_with_seg_len, dr_with_seg_len_pair_t):
30366         ... to here.
30367         (prune_runtime_alias_test_list): New decalaration.
30369 2017-05-26  Bin Cheng  <bin.cheng@arm.com>
30371         * tree-vect-data-refs.c (compare_tree): Rename and move ...
30372         * tree-data-ref.c (data_ref_compare_tree): ... to here.
30373         * tree-data-ref.h (data_ref_compare_tree): New decalaration.
30374         * tree-vect-data-refs.c (dr_group_sort_cmp): Update uses.
30375         (operator==, comp_dr_with_seg_len_pair): Ditto.
30376         (vect_prune_runtime_alias_test_list): Ditto.
30378 2017-05-26  Martin Liska  <mliska@suse.cz>
30380         PR ipa/80663
30381         * params.def: Bound partial-inlining-entry-probability param.
30383 2017-05-26  Marek Polacek  <polacek@redhat.com>
30385         PR sanitizer/80875
30386         * fold-const.c (fold_binary_loc) <case MULT_EXPR>: Check if OP1
30387         can be negated.
30389 2017-05-26  Richard Biener  <rguenther@suse.de>
30391         PR tree-optimization/80842
30392         * tree-ssa-ccp.c (set_lattice_value): Always meet with the old
30393         value.
30395 2017-05-26  Richard Biener  <rguenther@suse.de>
30397         PR tree-optimization/80844
30398         * tree-vectorizer.c (adjust_simduid_builtins): Propagate results.
30400 2017-05-25  Sebastian Peryt  <sebastian.peryt@intel.com>
30402         * doc/md.texi (Machine Constraints): Update x86 family
30403         machine constraints section to match 'config/i386/constraints.md'.
30405 2017-05-25  Volker Reichelt  <v.reichelt@netcologne.de>
30407         * doc/invoke.texi (-Wcatch-value=): Document new warning option.
30409 2017-05-25  Nathan Sidwell  <nathan@acm.org>
30411         * doc/invoke.texi (--enable-languages): Update documentation.
30413 2017-05-25  Martin Liska  <mliska@suse.cz>
30415         * dumpfile.c: Add TDF_FOLDING.
30416         * dumpfile.h (enum dump_kind): Likewise.
30417         * genmatch.c (dt_simplify::gen_1): Use it.
30419 2017-05-25  Marc Glisse  <marc.glisse@inria.fr>
30421         * match.pd (view_convert (convert@0 @1)): Handle zero-extension.
30423 2017-05-25  Marc Glisse  <marc.glisse@inria.fr>
30425         * match.pd ((A +- CST1) +- CST2): Allow some conversions.
30426         * tree.c (drop_tree_overflow): Handle COMPLEX_CST and VECTOR_CST.
30428 2017-05-25  Marc Glisse  <marc.glisse@inria.fr>
30430         * fold-const.c (fold_binary_loc) [(A & C) == D]: Remove transformation.
30431         * match.pd (X == C): Rewrite it here.
30432         (with_possible_nonzero_bits, with_possible_nonzero_bits2,
30433         with_certain_nonzero_bits2): New predicates.
30434         * tree-ssanames.c (get_nonzero_bits): Handle INTEGER_CST.
30436 2017-05-24  Nathan Sidwell  <nathan@acm.org>
30438         * lto-streamer-in.c (lto_input_data_block): Adjust T const cast to
30439         avoid warning.
30441         * auto-profile.c (afdo_propagate): Adjust T const cast to avoid
30442         warning.
30444 2017-05-24  Segher Boessenkool  <segher@kernel.crashing.org>
30446         * config/powerpcspe: New port.  Files are copied from the rs6000
30447         port, with "rs6000" in filenames replaced by "powerpcspe".
30449 2017-05-24  Wilco Dijkstra  <wdijkstr@arm.com>
30451         PR rtl-optimization/80754
30452         * lra-remat.c (do_remat): Add overlap checks for dst_regno.
30454 2017-05-24  Sheldon Lobo  <smlobo@sheldon.us.oracle.com>
30456         * config/sparc/sparc.md (length): Return the correct value for -mflat
30457         sibcalls to match output_sibcall.
30459 2017-05-24  Segher Boessenkool  <segher@kernel.crashing.org>
30461         PR bootstrap/80860
30462         PR bootstrap/80843
30463         * config/rs6000/rs6000.c (struct machine_function): Add new field
30464         n_components.
30465         (rs6000_get_separate_components): Init that field, use it.
30466         (rs6000_components_for_bb): Use the field.
30468 2017-05-24  Thomas Preud'homme  <thomas.preudhomme@arm.com>
30470         * config/arm/arm.c (arm_expand_prologue): Fix typo in comment.
30472 2017-05-24  Peter Bergner  <bergner@vnet.ibm.com>
30474         PR middle-end/80823
30475         * tree-cfg.c (group_case_labels_stmt): Delete increment of "i";
30477 2017-05-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
30479         PR target/80725
30480         * config/s390/s390.c (s390_check_qrst_address): Check incoming
30481         address against address_operand predicate.
30482         * config/s390/s390.md ("*indirect_jump"): Swap alternatives.
30484 2017-05-24  Eric Botcazou  <ebotcazou@adacore.com>
30486         * var-tracking.c (track_expr_p): Do not return 0 for tracked record
30487         parameters passed indirectly.
30489 2017-05-23  Uros Bizjak  <ubizjak@gmail.com>
30491         * config/i386/i386.md (*movdi_internal): Remove SSE4
30492         alternative 18 (?r, *v).  Update insn attributes.
30493         (*movsi_internal): Remove SSE4 alternative 13 (?r, *v).
30494         Update insn attributes.
30495         (*zero_extendsidi2): Remove SSE4 alternative (?r, *x).
30496         Update insn attributes.
30497         * config/i386/sse.md (vec_extract<ssevecmodelower>_0): Remove SSE4
30498         alternative 1 (r, v). Remove isa attribute.
30499         * config/i386/i386.c (dimode_scalar_chain::make_vector_copies):
30500         Always move value through stack for !TARGET_INTER_UNIT_MOVES_TO_VEC
30501         and !TARGET_INTER_UNIT_MOVES_TO_VEC targets.
30503 2017-05-23  Tom de Vries  <tom@codesourcery.com>
30505         * doc/sourcebuild.texi (Directives, Verify compiler message): Document
30506         dg-line directive.
30508 2017-05-23  Jan Hubicka  <hubicka@ucw.cz>
30510         * cgraphunit.c (symbol_table::process_new_functions): Update.
30511         * ipa-fnsummary.c (pass_data_inline_parameters): Remove.
30512         (inline_generate_summary): Rename to ...
30513         (ipa_fn_summary_generate): ... this one.
30514         (inline_read_summary): Rename to ...
30515         (ipa_fn_summary_read): ... this one.
30516         (inline_write_summary): Rename to ...
30517         (ipa_fn_summary_write): ... this one.
30518         (inline_free_summary): Rename to ...
30519         (ipa_free_fn_summary): ... this one.
30520         (pass_data_local_fn_summary, pass_local_fn_summary,
30521         make_pass_local_fn_summary, pass_data_ipa_free_fn_summary,
30522         pass_ipa_free_fn_summary, make_pass_ipa_free_fn_summary,
30523         pass_data_ipa_fn_summary, pass_ipa_fn_summary,
30524         make_pass_ipa_fn_summary): New.
30525         * ipa-fnsummary.h (inline_generate_summary, inline_read_summary,
30526         inline_write_summary, inline_free_summary): Remove.
30527         (ipa_free_fn_summary) : New.
30528         * ipa-inline.c (ipa_inline): Update.
30529         (pass_ipa_inline): Do not generate summaries.
30530         * ipa.c (pass_data_ipa_free_fn_summary, pass_ipa_free_fn_summary):
30531         Remove.
30532         * passes.def: Replace pass_inline_parameters by pass_local_fn_summary
30533         and add pass_ipa_fn_summary.
30534         * tree-pass.h (make_pass_ipa_fn_summary, make_pass_local_fn_summary):
30535         New.
30536         (make_pass_inline_parameters): Remove.
30538 2017-05-23  Thomas Schwinge  <thomas@codesourcery.com>
30540         * omp-low.c (struct omp_context): Remove "default_kind" member.
30541         Adjust all users.
30543         * omp-offload.c (execute_oacc_device_lower): Remove the
30544         parallelism dimensions function attributes for unparallelized
30545         OpenACC kernels constructs.
30547 2017-05-23  Martin Liska  <mliska@suse.cz>
30549         * cgraph.c (cgraph_node::get_create): Use symtab_node::dump_{asm_,}name
30550         functions.
30551         (cgraph_edge::make_speculative): Likewise.
30552         (cgraph_edge::resolve_speculation): Likewise.
30553         (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
30554         (cgraph_node::dump): Likewise.
30555         * cgraph.h: Likewise.
30556         * cgraphunit.c (analyze_functions): Likewise.
30557         (symbol_table::compile): Likewise.
30558         * ipa-cp.c (print_all_lattices): Likewise.
30559         (determine_versionability): Likewise.
30560         (initialize_node_lattices): Likewise.
30561         (ipcp_verify_propagated_values): Likewise.
30562         (estimate_local_effects): Likewise.
30563         (update_profiling_info): Likewise.
30564         (create_specialized_node): Likewise.
30565         (perhaps_add_new_callers): Likewise.
30566         (decide_about_value): Likewise.
30567         (decide_whether_version_node): Likewise.
30568         (identify_dead_nodes): Likewise.
30569         (ipcp_store_bits_results): Likewise.
30570         * ipa-devirt.c (dump_targets): Likewise.
30571         (ipa_devirt): Likewise.
30572         * ipa-icf.c (sem_item::dump): Likewise.
30573         (sem_function::equals): Likewise.
30574         (sem_variable::equals): Likewise.
30575         (sem_item_optimizer::read_section): Likewise.
30576         (sem_item_optimizer::execute): Likewise.
30577         (congruence_class::dump): Likewise.
30578         * ipa-inline-analysis.c (dump_ipa_call_summary): Likewise.
30579         (dump_inline_summary): Likewise.
30580         (estimate_node_size_and_time): Likewise.
30581         (inline_analyze_function): Likewise.
30582         * ipa-inline-transform.c (inline_call): Likewise.
30583         * ipa-inline.c (report_inline_failed_reason): Likewise.
30584         (want_early_inline_function_p): Likewise.
30585         (edge_badness): Likewise.
30586         (update_edge_key): Likewise.
30587         (inline_small_functions): Likewise.
30588         * ipa-profile.c (ipa_profile): Likewise.
30589         * ipa-prop.c (ipa_print_node_jump_functions): Likewise.
30590         (ipa_make_edge_direct_to_target): Likewise.
30591         (remove_described_reference): Likewise.
30592         (ipa_impossible_devirt_target): Likewise.
30593         (propagate_controlled_uses): Likewise.
30594         (ipa_print_node_params): Likewise.
30595         (ipcp_transform_function): Likewise.
30596         * ipa-pure-const.c (pure_const_read_summary): Likewise.
30597         (propagate_pure_const): Likewise.
30598         * ipa-reference.c (generate_summary): Likewise.
30599         (read_write_all_from_decl): Likewise.
30600         (propagate): Likewise.
30601         (ipa_reference_read_optimization_summary): Likewise.
30602         * ipa-utils.c (ipa_merge_profiles): Likewise.
30603         * ipa.c (walk_polymorphic_call_targets): Likewise.
30604         (symbol_table::remove_unreachable_nodes): Likewise.
30605         (ipa_single_use): Likewise.
30606         * passes.c (execute_todo): Likewise.
30607         * predict.c (drop_profile): Likewise.
30608         * symtab.c (symtab_node::get_dump_name): New function.
30609         (symtab_node::dump_name): Likewise.
30610         (symtab_node::dump_asm_name): Likewise.
30611         (symtab_node::dump_references): Likewise.
30612         (symtab_node::dump_referring): Likewise.
30613         (symtab_node::dump_base): Likewise.
30614         (symtab_node::debug_symtab): Likewise.
30615         * tree-sra.c (convert_callers_for_node): Likewise.
30616         * tree-ssa-structalias.c (ipa_pta_execute): Likewise.
30617         * value-prof.c (init_node_map): Likewise.
30619 2017-05-23  Martin Liska  <mliska@suse.cz>
30621         * cgraph.h: Move symtab_node::dump_table to symbol_table::dump
30622         and symtab_node::debug_symtab to symbol_table::debug.
30623         * cgraphunit.c (analyze_functions): Use the renamed function.
30624         (symbol_table::compile): Likewise.
30625         * ipa-cp.c (ipcp_verify_propagated_values): Likewise.
30626         * ipa-icf.c (sem_item_optimizer::execute): Likewise.
30627         * passes.c (execute_todo): Likewise.
30628         * symtab.c (symbol_table::dump): New function.
30629         * tree-ssa-structalias.c (ipa_pta_execute): Use the renamed function.
30631 2017-05-23  Jan Hubicka  <hubicka@ucw.cz>
30633         * ipa-fnsummary.c (estimate_node_size_and_time): Do not sanity check
30634         that nonconst implies exec.
30636 2017-05-23  Jan Hubicka  <hubicka@ucw.cz>
30638         * ipa-inline-analysis.c (cgraph_2edge_hook_list, cgraph_edge_hook_list,
30639         inline_edge_removal_hook, inline_edge_duplication_hook): Remove.
30640         (inline_edge_summary_vec): Turn into ...
30641         (ipa_call_summaries): ... this one.
30642         (redirect_to_unreachable, edge_set_predicate,
30643         evaluate_properties_for_edge, inline_summary_alloc,
30644         reset_ipa_call_summary, reset_inline_summary,
30645         inline_summary_t::duplicate): Update.
30646         (inline_edge_duplication_hook): Turn to ...
30647         (ipa_call_summary_t::duplicate): ... this one.
30648         (inline_edge_removal_hook): Turn to ...
30649         (ipa_call_summary_t::remove): ... this one.
30650         (dump_inline_edge_summary): Turn to ...
30651         (dump_ipa_call_summary): ... this one.
30652         (estimate_function_body_sizes): Update.
30653         (inline_update_callee_summaries): Update.
30654         (remap_edge_change_prob): Update.
30655         (remap_edge_summaries): Update.
30656         (inline_merge_summary): Update.
30657         (do_estimate_edge_time): Update.
30658         (inline_generate_summary): Update.
30659         (inline_read_section): Update.
30660         (inline_read_summary): Update.
30661         (inline_free_summary): Update.
30662         * ipa-inline.c (can_inline_edge_p): Update.
30663         (compute_inlined_call_time): Update.
30664         (want_inline_small_function_p): Update.
30665         (edge_badness): Update.
30666         (early_inliner): Update.
30667         * ipa-inline.h (inline_edge_summary): Turn to ...
30668         (ipa_call_summary): ... this one.
30669         (ipa_call_summary_t): New class.
30670         (inline_edge_summary_t, inline_edge_summary_vec): Remove.
30671         (ipa_call_summaries): New.
30672         (inline_edge_summary): Remove.
30673         (estimate_edge_growth): Update.
30674         * ipa-profile.c (ipa_propagate_frequency_1): Update.
30675         * ipa-prop.c (ipa_make_edge_direct_to_target): Update.
30676         * ipa-split.c (execute_split_functions): Update.
30677         * ipa.c (symbol_table::remove_unreachable_nodes): Update.
30679 2017-05-23  Tom de Vries  <tom@codesourcery.com>
30681         * doc/sourcebuild.texi (Effective-Target Keywords, Other hardware
30682         attributes): Document rdrand effective target.
30684 2017-05-23  Tom de Vries  <tom@codesourcery.com>
30686         * doc/sourcebuild.texi (Effective-Target Keywords, Other hardware
30687         attributes): Sort alphabetically.
30689 2017-05-23  Georg-Johann Lay  <avr@gjlay.de>
30691         * config/avr/genmultilib.awk: Use gsub instead of gensub.
30693 2017-05-22  Michael Meissner  <meissner@linux.vnet.ibm.com>
30695         PR target/80718
30696         * config/rs6000/vsx.md (vsx_splat_<mode>, VSX_D iterator): Split
30697         V2DF/V2DI splat into two separate patterns, one that handles
30698         registers, and the other that only handles memory.  Drop support
30699         for splatting from a GPR on ISA 2.07 and then splitting the
30700         splat into direct move and splat.
30701         (vsx_splat_<mode>_reg): Likewise.
30702         (vsx_splat_<mode>_mem): Likewise.
30704 2017-05-22  Segher Boessenkool  <segher@kernel.crashing.org>
30706         * cfgcleanup.c (bb_is_just_return): Allow CLOBBERs.
30708 2017-05-22  Jakub Jelinek  <jakub@redhat.com>
30710         PR middle-end/80809
30711         * omp-low.c (finish_taskreg_remap): New function.
30712         (finish_taskreg_scan): If unit size of ctx->record_type
30713         is non-constant, unshare the size expression and replace
30714         decls in it with possible outer var refs.
30716         PR middle-end/80809
30717         * gimplify.c (omp_add_variable): For GOVD_DEBUG_PRIVATE use
30718         GOVD_SHARED rather than GOVD_PRIVATE with it.
30719         (gimplify_adjust_omp_clauses_1, gimplify_adjust_omp_clauses): Expect
30720         GOVD_SHARED rather than GOVD_PRIVATE with GOVD_DEBUG_PRIVATE.
30722         PR middle-end/80853
30723         * omp-low.c (lower_reduction_clauses): Pass OMP_CLAUSE_PRIVATE
30724         as last argument to build_outer_var_ref for pointer bases of array
30725         section reductions.
30727 2017-05-19  Martin Sebor  <msebor@redhat.com>
30729         * print-tree.c (print_node): Print DECL_READ_P flag.
30731 2017-05-22  Jan Hubicka  <hubicka@ucw.cz>
30733         * Makefile.in: Add ipa-fnsummary.o and ipa-fnsummary.h
30734         * auto-profile.c: Replace ipa-inline.h by ipa-fnsummary.h
30735         * cgraph.c: Likewise.
30736         * cgraphunit.c: Likewise.
30737         * gengtype.c: Likewise.
30738         * ipa-cp.c: Likewise.
30739         * ipa-devirt.c: Likewise.
30740         * ipa-icf.c: Likewise.
30741         * ipa-predicate.c: Likewise.
30742         * ipa-profile.c: Likewise.
30743         * ipa-prop.c: Likewise.
30744         * ipa-split.c: Likewise.
30745         * ipa.c: Likewise.
30746         * ipa-inline-analysis.c (inline_summaries, ipa_call_summaries,
30747         edge_predicate_pool, dump_inline_hints,
30748         inline_summary::account_size_time, redirect_to_unreachable,
30749         edge_set_predicate, set_hint_predicate,
30750         evaluate_conditions_for_known_args, evaluate_properties_for_edge,
30751         inline_summary_alloc, ipa_call_summary::reset, inline_summary::reset,
30752         inline_summary_t::remove, remap_hint_predicate_after_duplication,
30753         inline_summary_t::duplicate, ipa_call_summary_t::duplicate,
30754         ipa_call_summary_t::remove, initialize_growth_caches,
30755         free_growth_caches, dump_ipa_call_summary, dump_inline_summary,
30756         debug_inline_summary, dump_inline_summaries, initialize_inline_failed,
30757         mark_modified, unmodified_parm_1, unmodified_parm,
30758         unmodified_parm_or_parm_agg_item, eliminated_by_inlining_prob,
30759         set_cond_stmt_execution_predicate, set_switch_stmt_execution_predicate,
30760         compute_bb_predicates, will_be_nonconstant_expr_predicate,
30761         will_be_nonconstant_predicate, record_modified_bb_info,
30762         get_minimal_bb, record_modified, param_change_prob,
30763         phi_result_unknown_predicate, predicate_for_phi_result,
30764         array_index_predicate, clobber_only_eh_bb_p, fp_expression_p,
30765         estimate_function_body_sizes, compute_inline_parameters,
30766         compute_inline_parameters_for_curren, pass_data_inline_parameters,
30767         estimate_node_size_and_time, estimate_ipcp_clone_size_and_time,
30768         inline_update_callee_summaries, remap_edge_change_prob,
30769         remap_edge_summaries, remap_hint_predicate, inline_merge_summary,
30770         inline_update_overall_summary, inline_indirect_intraprocedural_analysis,
30771         inline_analyze_function, inline_summary_t::insert,
30772         inline_generate_summary, read_ipa_call_summary, inline_read_section,
30773         inline_read_summary, write_ipa_call_summary, inline_write_summary,
30774         inline_free_summary): Move to ipa-fnsummary.h
30775         (predicate_t): Remove.
30776         * ipa-fnsummary.c: New file.
30777         * ipa-inline.h:  Do not include sreal.h and ipa-predicate.h
30778         (enum inline_hints_vals, inline_hints, agg_position_info,
30779         INLINE_SIZE_SCALE, size_time_entry, inline_summary, inline_summary_t,
30780         inline_summaries, ipa_call_summary, ipa_call_summary_t,
30781         ipa_call_summaries, debug_inline_summary, dump_inline_summaries,
30782         dump_inline_summary, dump_inline_hints, inline_generate_summary,
30783         inline_read_summary, inline_write_summary, inline_free_summary,
30784         inline_analyze_function, initialize_inline_failed,
30785         inline_merge_summary, inline_update_overall_summary,
30786         compute_inline_parameters): Move to ipa-fnsummary.h
30787         * ipa-fnsummary.h: New file.
30788         * ipa-inline-transform.h: Include ipa-inline.h.
30789         * ipa-inline.c: LIkewise.
30791 2017-05-22  Jan Hubicka  <hubicka@ucw.cz>
30793         * ipa-inline.c (edge_badness): Use inlined_time instead of
30794         inline_summaries->get.
30796 2017-05-22  Jan Hubicka  <hubicka@ucw.cz>
30798         * ipa-inline.c (edge_badness): Use estimate_size_after_inlining.
30800 2017-05-22  Nathan Sidwell  <nathan@acm.org>
30802         * doc/invoke.texi (fdump-translation-unit): Delete documentation.
30803         (fdump-lang): Document 'raw' option.
30804         * dumpfile.h (TDI_tu): Delete.
30805         * dumpfile.c (dump_files): Remove translation-unit.
30806         (FIRST_AUTO_NUMBERED_DUMP): Decrement.
30808 2017-05-22  Georg-Johann Lay  <avr@gjlay.de>
30810         * config/avr/t-avr (AWK) [t-multilib]: Remove "-v FORMAT=Makefile"
30811         command option from $(AWK) call.
30812         * config/avr/genmultilib.awk: Simplify and rewrite so that it
30813         generates MULTILIB_REQUIRED instead of MULTILIB_EXCEPTIONS.
30814         [FORMAT]: Remove handling of variable.
30815         * config/avr/t-multilib: Regenerate.
30817 2017-05-22  Jan Hubicka  <hubicka@ucw.cz>
30819         * ipa-inline-analysis.c (inline_summary::reset): Do not reset
30820         self_time.
30821         (dump_inline_summary): Do not print self_time.
30822         (estimate_function_body_sizes): Do not set self_time.
30823         (compute_inline_parameters): Likewise.
30824         (inline_read_section, inline_write_summary): Do not stream self_time.
30825         * ipa-inline.h (inline_summary): Drop self_time.
30827 2017-05-22  Jan Hubicka  <hubicka@ucw.cz>
30829         * ipa-inline-analysis.c (account_size_time): Rename to ...
30830         (inline_summary::account_size_time): ... this one.
30831         (reset_ipa_call_summary): Turn to ...
30832         (ipa_call_summary::reset): ... this one.
30833         (reset_inline_summary): Turn to ...
30834         (inline_summary::reset): ... this one.
30835         (inline_summary_t::remove): Update.
30836         (inline_summary_t::duplicate): Update.
30837         (ipa_call_summary_t::remove): Update.
30838         (dump_inline_summary): Update.
30839         (estimate_function_body_sizes): Update.
30840         (compute_inline_parameters): Update.
30841         (estimate_node_size_and_time): Update.
30842         (inline_merge_summary): Update.
30843         (inline_update_overall_summary): Update.
30844         (inline_read_section): Update.
30845         (inline_write_summary): Update.
30846         * ipa-inline.h (inline_summary): Rename entry to size_time_table;
30847         add account_size_time and reset member functions.
30848         (ipa_call_summary): Add reset function.
30849         * ipa-predicate.h (predicate::operator &): Constify.
30851 2017-05-22  Richard Biener  <rguenther@suse.de>
30853         * df-scan.c (df_insn_refs_verify): Speedup when not verifying.
30855 2017-05-19  Jason Merrill  <jason@redhat.com>
30857         * tree.c (make_tree_vec_stat, grow_tree_vec_stat): Use size_t.
30859 2017-05-19  Marek Polacek  <polacek@redhat.com>
30861         PR sanitizer/80800
30862         * fold-const.c (extract_muldiv_1) <case TRUNC_DIV_EXPR>: Add
30863         TYPE_OVERFLOW_WRAPS checks.
30865 2017-05-19  Thomas Schwinge  <thomas@codesourcery.com>
30867         * tree-core.h (enum omp_clause_default_kind): Add
30868         "OMP_CLAUSE_DEFAULT_PRESENT".
30869         * tree-pretty-print.c (dump_omp_clause): Handle it.
30870         * gimplify.c (enum gimplify_omp_var_data): Add
30871         "GOVD_MAP_FORCE_PRESENT".
30872         (gimplify_adjust_omp_clauses_1): Map it to
30873         "GOMP_MAP_FORCE_PRESENT".
30874         (oacc_default_clause): Handle "OMP_CLAUSE_DEFAULT_PRESENT".
30876         * gimplify.c (oacc_default_clause): Clarify.
30878 2017-05-19  Nathan Sidwell  <nathan@acm.org>
30880         LANG_HOOK_REGISTER_DUMPS
30881         * toplev.c (general_init): Call register dump lang hook.
30882         * doc/invoke.texi: Document -fdump-lang option family.
30883         * dumpfile.c (dump_files): Remove class dump here.
30884         (FIRST_AUTO_NUMBERED_DUMP): Adjust.
30885         * dumpfile.h (tree_dump_index): Remove TDI_class.
30886         * langhooks-def.h (lhd_register_dumps): Declare.
30887         (LANG_HOOKS_REGISTER_DUMPS): Define.
30888         (LANG_HOOKS_INITIALIZER): Add it.
30889         * langhooks.c (lhd_register_dumps): Define.
30890         * langhooks.h (struct lang_hooks): Add register_dumps.
30892 2017-05-19  Nathan Sidwell  <nathan@acm.org>
30894         * context.h (context::set_passes): New.
30895         * context.c (context::context): Do not create pass manager.
30896         * toplev.c (general_init): Create pass manager here.
30898 2017-05-19  Segher Boessenkool  <segher@kernel.crashing.org>
30900         * config/rs6000/rs6000.md (splitter to load of -1 and mask): Don't
30901         use this splitter if two add or or instructions would also work for
30902         the constant we want to generate.
30904 2017-05-19  Richard Biener  <rguenther@suse.de>
30906         PR build/80821
30907         * genmatch.c (dt_node::gen_kids_1): Add missing scope around
30908         predicate evaluation.
30910 2017-05-19  Jan Hubicka  <hubicka@ucw.cz>
30912         * ipa-inline.h (ipa_call_summary): Turn sizes into signed;
30913         add ctor.
30914         * ipa-inline.c (want_inline_small_function_p): Do not cast to
30915         unsigned.
30917 2017-05-19  Jan Hubicka  <hubicka@ucw.cz>
30919         * ipa-inline-analysis.c (cgraph_2edge_hook_list, cgraph_edge_hook_list,
30920         inline_edge_removal_hook, inline_edge_duplication_hook): Remove.
30921         (inline_edge_summary_vec): Turn into ...
30922         (ipa_call_summaries): ... this one.
30923         (redirect_to_unreachable, edge_set_predicate,
30924         evaluate_properties_for_edge, inline_summary_alloc,
30925         reset_ipa_call_summary, reset_inline_summary,
30926         inline_summary_t::duplicate): Update.
30927         (inline_edge_duplication_hook): Turn to ...
30928         (ipa_call_summary_t::duplicate): ... this one.
30929         (inline_edge_removal_hook): Turn to ...
30930         (ipa_call_summary_t::remove): ... this one.
30931         (dump_inline_edge_summary): Turn to ...
30932         (dump_ipa_call_summary): ... this one.
30933         (estimate_function_body_sizes): Update.
30934         (inline_update_callee_summaries): Update.
30935         (remap_edge_change_prob): Update.
30936         (remap_edge_summaries): Update.
30937         (inline_merge_summary): Update.
30938         (do_estimate_edge_time): Update.
30939         (inline_generate_summary): Update.
30940         (inline_read_section): Update.
30941         (inline_read_summary): Update.
30942         (inline_free_summary): Update.
30943         * ipa-inline.c (can_inline_edge_p): Update.
30944         (compute_inlined_call_time): Update.
30945         (want_inline_small_function_p): Update.
30946         (edge_badness): Update.
30947         (early_inliner): Update.
30948         * ipa-inline.h (inline_edge_summary): Turn to ...
30949         (ipa_call_summary): ... this one.
30950         (ipa_call_summary_t): New class.
30951         (inline_edge_summary_t, inline_edge_summary_vec): Remove.
30952         (ipa_call_summaries): New.
30953         (inline_edge_summary): Remove.
30954         (estimate_edge_growth): Update.
30955         * ipa-profile.c (ipa_propagate_frequency_1): Update.
30956         * ipa-prop.c (ipa_make_edge_direct_to_target): Update.
30957         * ipa-split.c (execute_split_functions): Update.
30958         * ipa.c (symbol_table::remove_unreachable_nodes): Update.
30960 2017-05-19  Richard Biener  <rguenther@suse.de>
30962         PR middle-end/80764
30963         * cfgexpand.c (expand_gimple_cond): Fix loop fixup.
30965 2017-05-18  Segher Boessenkool  <segher@kernel.crashing.org>
30967         * config/rs6000/rs6000.c (struct machine_function): Add field
30968         fpr_is_wrapped_separately.
30969         (rs6000_get_separate_components): Use 64 components.  Handle the
30970         new FPR components.
30971         (rs6000_components_for_bb): Handle the FPR components.
30972         (rs6000_emit_prologue_components): Handle the FPR components.
30973         (rs6000_emit_epilogue_components): Handle the FPR components.
30974         (rs6000_set_handled_components): Handle the FPR components.
30975         (rs6000_emit_prologue): Don't output prologue code for those FPRs
30976         that are already separately shrink-wrapped.
30977         (rs6000_emit_epilogue): Don't output epilogue code for those FPRs
30978         that are already separately shrink-wrapped.
30980 2017-05-18  Michael Meissner  <meissner@linux.vnet.ibm.com>
30982         PR target/80510
30983         * config/rs6000/predicates.md (simple_offsettable_mem_operand):
30984         New predicate.
30986         * config/rs6000/rs6000.md (ALTIVEC_DFORM): New iterator.
30987         (define_peephole2 for Altivec d-form load): Add peepholes to catch
30988         cases where the register allocator uses a move and an offsettable
30989         memory operation to/from a FPR register on ISA 2.06/2.07.
30990         (define_peephole2 for Altivec d-form store): Likewise.
30992 2017-05-18  Uros Bizjak  <ubizjak@gmail.com>
30994         PR target/80799
30995         * config/i386/mmx.md (*mov<mode>_internal): Enable
30996         alternatives 11, 12, 13 and 14 also for 32bit targets.
30997         Remove alternatives 15, 16, 17 and 18.
30998         * config/i386/sse.md (vec_concatv2di): Change
30999         alternative (!x, *y) to (x, ?!*Yn).
31001 2017-05-18  Paolo Carlini  <paolo.carlini@oracle.com>
31003         * dumpfile.h (enum dump_kind): Remove stray comma.
31005 2017-05-18  Jan Hubicka  <hubicka@ucw.cz>
31007         * Makefile.in: Add ipa-predicate.o and ipa-predicate.h
31008         * ipa-inline-analysis.c (NUM_CONDITIONS): turn into
31009         predicate::num_conditions
31010         (IS_NOT_CONSTANT): turn into predicate::is_not_constant.
31011         (CHANGED): turn into predicate::changed.
31012         (agg_position_info): Move to ipa-predicate.h
31013         (add_condition, predicate::add_clause, predicate::operator &=,
31014         predicate::or_with, predicate::evaluate, predicate::probability,
31015         dump_condition, dump_clause, predicate::dump,
31016         predicate::remap_after_duplication, predicate::remap_after_inlining,
31017         predicate::stream_in, predicate::stream_out): Move to ipa-predicate.c
31018         (evaluate_conditions_for_known_args): Update.
31019         (set_cond_stmt_execution_predicate): Update.
31020         * ipa-inline.h: Include ipa-predicate.h
31021         (condition, inline_param_summary, conditions, agg_position_info,
31022         predicate): Move to ipa-predicate.h
31023         * ipa-predicate.c: New file.
31024         * ipa-predicate.h: New file.
31026 2017-05-18  Wilco Dijkstra  <wdijkstr@arm.com>
31028         * final.c (leaf_function_p): Check we are not in a sequence.
31030 2017-05-18  Martin Liska  <mliska@suse.cz>
31032         * cfgrtl.c (rtl_verify_edges): Remove usage of TDF_RTL.
31033         * dumpfile.c (dump_register): Use new enum dump_kind.
31034         (get_dump_file_name): Likewise.
31035         (dump_enable_all): Likewise.
31036         (dump_switch_p_1): Likewise.
31037         (enable_rtl_dump_file): Remove usage of TDF_RTL.
31038         * dumpfile.h (enum dump_kind): New enum type.
31039         (struct dump_file_info): Create constructor and
31040         format fields and comments.
31041         * passes.c (pass_manager::register_one_dump_file):
31042         Use num dump_kind.
31043         * statistics.c (statistics_early_init): Likewise.
31044         * tree-ssa-loop-prefetch.c (dump_mem_details): Replace
31045         TDF_TREE with TDF_SLIM.
31046         (gather_memory_references_ref): Likewise.
31048 2017-05-18  Martin Liska  <mliska@suse.cz>
31050         * vec.h (struct vnull): Use it.
31052 2017-05-18  Jan Hubicka  <hubicka@ucw.cz>
31054         * ipa-inline-analysis.c (predicate_conditions): Move to ipa-inline.h
31055         (true_predicate, false_predicate, true_predicate_p,
31056         false_predicate_p): Remove.
31057         (single_cond_predicate, not_inlined_predicate): Turn to member function
31058         in ipa-inline.h
31059         (add_condition): Update.
31060         (add_clause): Turn to...
31061         (predicate::add_clause): ... this one; update; allow passing NULL
31062         as parameter.
31063         (and_predicates): Turn to ...
31064         (predicate::operator &=): ... this one.
31065         (predicates_equal_p): Move to predicate::operator == in ipa-inline.h
31066         (or_predicates): Turn to ...
31067         (predicate::or_with): ... this one.
31068         (evaluate_predicate): Turn to ...
31069         (predicate::evaluate): ... this one.
31070         (predicate_probability): Turn to ...
31071         (predicate::probability): ... this one.
31072         (dump_condition): Update.
31073         (dump_predicate): Turn to ...
31074         (predicate::dump): ... this one.
31075         (account_size_time): Update.
31076         (edge_set_predicate): Update.
31077         (set_hint_predicate): UPdate.
31078         (evaluate_conditions_for_known_args): Update.
31079         (evaluate_properties_for_edge): Update.
31080         (remap_predicate_after_duplication): Turn to...
31081         (predicate::remap_after_duplication): ... this one.
31082         (remap_hint_predicate_after_duplication): Update.
31083         (inline_summary_t::duplicate): UPdate.
31084         (dump_inline_edge_summary): Update.
31085         (dump_inline_summary): Update.
31086         (set_cond_stmt_execution_predicate): Update.
31087         (set_switch_stmt_execution_predicate): Update.
31088         (compute_bb_predicates): Update.
31089         (will_be_nonconstant_expr_predicate): Update.
31090         (will_be_nonconstant_predicate): Update.
31091         (phi_result_unknown_predicate): Update.
31092         (predicate_for_phi_result): Update.
31093         (array_index_predicate): Update.
31094         (estimate_function_body_sizes): Update.
31095         (estimate_node_size_and_time): Update.
31096         (estimate_ipcp_clone_size_and_time): Update.
31097         (remap_predicate): Rename to ...
31098         (predicate::remap_after_inlining): ... this one.
31099         (remap_hint_predicate): Update.
31100         (inline_merge_summary): Update.
31101         (inline_update_overall_summary): Update.
31102         (estimate_size_after_inlining): Update.
31103         (read_predicate): Rename to ...
31104         (predicate::stream_in): ... this one.
31105         (read_inline_edge_summary): Update.
31106         (write_predicate): Rename to ...
31107         (predicate::stream_out): ... this one.
31108         (write_inline_edge_summary): Update.
31109         * ipa-inline.h (MAX_CLAUSES): Turn to predicate::max_clauses.
31110         (clause_t): Turn to uint32_t
31111         (predicate): Turn to class; implement constructor and operators
31112         ==, !=, &
31113         (size_time_entry): Update.
31114         (inline_summary): Update.
31115         (inline_edge_summary): Update.
31117 2017-05-18  Marc Glisse  <marc.glisse@inria.fr>
31119         * fold-const.c (fold_binary_loc): Move transformation...
31120         * match.pd (C - X CMP X): ... here.
31122 2017-05-18  Sheldon Lobo  <sheldon.lobo@oracle.com>
31124         * config/sparc/sparc.c (sparc_option_override): Set function
31125         alignment for -mcpu=niagara7 to 64 to match the I$ line.
31126         * config/sparc/sparc.h (BRANCH_COST): Set the SPARC M7 branch
31127         latency to 1.
31128         * config/sparc/sparc.h (BRANCH_COST): Set the SPARC T4 branch
31129         latency to 2.
31130         * config/sparc/sol2.h: Fix a ASM_CPU32_DEFAULT_SPEC typo.
31132 2017-05-18  Marek Polacek  <polacek@redhat.com>
31134         PR sanitizer/80797
31135         * ubsan.c (instrument_null): Unwrap ADDR_EXPRs.
31136         (pass_ubsan::execute): Call gimple_assign_single_p instead of
31137         gimple_assign_load_p.
31139 2017-05-17  Segher Boessenkool  <segher@kernel.crashing.org>
31141         PR middle-end/80692
31142         * real.c (do_compare): Give decimal_do_compare preference over
31143         comparing just the signs.
31145 2017-05-17  Uros Bizjak  <ubizjak@gmail.com>
31147         * doc/md.texi (Canonicalization of Instructions): Describe the
31148         canonical form of instructions that inherently set a condition
31149         code register.
31151 2017-05-17  Peter Bergner  <bergner@vnet.ibm.com>
31153         PR middle-end/80775
31154         * tree-cfg.c: Move deletion of unreachable case statements to after
31155         the merging of consecutive case labels.
31157 2017-05-17  Thomas Preud'homme  <thomas.preudhomme@arm.com>
31159         * config/arm/arm.c (cmse_nonsecure_call_clear_caller_saved): Refer
31160         readers to __gnu_cmse_nonsecure_call libcall for saving, clearing and
31161         restoring of callee-saved registers.
31163 2017-05-17  Eric Botcazou  <ebotcazou@adacore.com>
31165         * compare-elim.c (equivalent_reg_at_start): Adjust outdated comment.
31166         * config/visium/visium.c (single_set_and_flags): Likewise.
31167         * config/visium/visium.md (Substitutions): Likewise.
31169 2017-05-17  Martin Liska  <mliska@suse.cz>
31171         * cfg.c: Introduce dump_flags_t type and
31172         use it instead of int type.
31173         * cfg.h: Likewise.
31174         * cfghooks.c: Likewise.
31175         * cfghooks.h (struct cfg_hooks): Likewise.
31176         * cfgrtl.c: Likewise.
31177         * cfgrtl.h: Likewise.
31178         * cgraph.c (cgraph_node::get_body): Likewise.
31179         * coretypes.h: Likewise.
31180         * domwalk.c: Likewise.
31181         * domwalk.h: Likewise.
31182         * dumpfile.c (struct dump_option_value_info): Likewise.
31183         (dump_enable_all): Likewise.
31184         (dump_switch_p_1): Likewise.
31185         (opt_info_switch_p): Likewise.
31186         * dumpfile.h (enum tree_dump_index): Likewise.
31187         (struct dump_file_info): Likewise.
31188         * genemit.c: Likewise.
31189         * generic-match-head.c: Likewise.
31190         * gengtype.c (open_base_files): Likewise.
31191         * gimple-pretty-print.c: Likewise.
31192         * gimple-pretty-print.h: Likewise.
31193         * graph.c (print_graph_cfg): Likewise.
31194         * graphite-scop-detection.c (dot_all_sese): Likewise.
31195         * ipa-devirt.c (build_type_inheritance_graph): Likewise.
31196         * loop-unroll.c (report_unroll): Likewise.
31197         * passes.c (pass_manager::register_one_dump_file): Likewise.
31198         * print-tree.c: Likewise.
31199         * statistics.c: Likewise.
31200         * tree-cfg.c: Likewise.
31201         * tree-cfg.h: Likewise.
31202         * tree-dfa.c: Likewise.
31203         * tree-dfa.h: Likewise.
31204         * tree-dump.c (dump_function): Likewise.
31205         * tree-dump.h (struct dump_info): Likewise.
31206         * tree-pretty-print.c: Likewise.
31207         * tree-pretty-print.h: Likewise.
31208         * tree-ssa-live.c: Likewise.
31209         * tree-ssa-live.h: Likewise.
31210         * tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Likewise.
31211         * tree-vect-loop.c: Likewise.
31212         * tree-vect-slp.c: Likewise.
31214 2017-05-16  James Greenhalgh  <james.greenhalgh@arm.com>
31215             Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
31217         PR tree-optimization/80457
31218         * tree-vect-stmts.c (vect_model_simple_cost): Model the cost
31219         of all arguments to a statement as scalar_to_vec operations.
31220         (vectorizable_call): Adjust call to vect_model_simple_cost for
31221         new parameter.
31222         (vectorizable_conversion): Likewise.
31223         (vectorizable_assignment): Likewise.
31224         (vectorizable_shift): Likewise.
31225         (vectorizable_operation): Likewise.
31226         (vectorizable_comparison): Likewise.
31227         (vect_is_simple_cond): Record the def types for operands.
31228         (vectorizable_condition): Likewise, call vect_model_simple_cost.
31229         * tree-vectorizer.h (vect_model_simple_cost): Add new parameter
31230         for statement argument count.
31232 2017-05-16  Carl Love  <cel@us.ibm.com>
31234         * config/rs6000/rs6000-c: Add support for built-in functions
31235         vector unsigned long long vec_bperm (vector unsigned long long,
31236                                              vector unsigned char)
31237         vector signed long long vec_mule (vector signed int,
31238                                           vector signed int)
31239         vector unsigned long long vec_mule (vector unsigned int,
31240                                             vector unsigned int)
31241         vector signed long long vec_mulo (vector signed int,
31242                                           vector signed int)
31243         vector unsigned long long vec_mulo (vector unsigned int,
31244                                             vector unsigned int)
31245         vector signed char vec_sldw (vector signed char,
31246                                      vector signed char,
31247                                      const int)
31248         vector unsigned char vec_sldw (vector unsigned char,
31249                                        vector unsigned char,
31250                                        const int)
31251         vector signed short vec_sldw (vector signed short,
31252                                       vector signed short,
31253                                       const int)
31254         vector unsigned short vec_sldw (vector unsigned short,
31255                                         vector unsigned short,
31256                                         const int)
31257         vector signed int vec_sldw (vector signed int,
31258                                     vector signed int,
31259                                     const int)
31260         vector unsigned int vec_sldw (vector unsigned int,
31261                                       vector unsigned int,
31262                                       const int)
31263         vector signed long long vec_sldw (vector signed long long,
31264                                           vector signed long long,
31265                                           const int)
31266         vector unsigned long long vec_sldw (vector unsigned long long,
31267                                             vector unsigned long long,
31268                                             const int)
31269         * config/rs6000/rs6000-c: Add support for built-in functions
31270         * config/rs6000/rs6000-builtin.def: Add definition for SLDW.
31271         * config/rs6000/altivec.h: Add defintion for vec_sldw.
31272         * doc/extend.texi: Update the built-in documentation for the
31273         new built-in functions.
31275 2017-05-16  Marek Polacek  <polacek@redhat.com>
31277         PR sanitizer/80536
31278         PR sanitizer/80386
31279         * tree.c (save_expr): Don't fold the expression.
31281 2017-05-16  Uros Bizjak  <ubizjak@gmail.com>
31283         * config/i386/i386.md (*movsi_internal): Split (?rm,*y) alternative
31284         to (?r,*Yn) and (?m,*y) alternatives, and (?*y,rm) to (?*Ym,r)
31285         and (?*y,m).  Update insn attributes.
31287 2017-05-16  Martin Liska  <mliska@suse.cz>
31289         * cgraph.c (cgraph_edge::resolve_speculation): Add default value for
31290         flags argument of print_gimple_stmt, print_gimple_expr,
31291         print_generic_stmt and print_generic_expr.
31292         * cgraphclones.c (symbol_table::materialize_all_clones): Likewise.
31293         * coretypes.h: Likewise.
31294         * except.c (dump_eh_tree): Likewise.
31295         * gimple-fold.c (gimple_fold_stmt_to_constant_1): Likewise.
31296         * gimple-pretty-print.h: Likewise.
31297         * gimple-ssa-backprop.c (dump_usage_prefix): Likewise.
31298         (backprop::push_to_worklist): Likewise.
31299         (backprop::pop_from_worklist): Likewise.
31300         (backprop::process_use): Likewise.
31301         (backprop::intersect_uses): Likewise.
31302         (note_replacement): Likewise.
31303         * gimple-ssa-store-merging.c
31304         (pass_store_merging::terminate_all_aliasing_chains): Likewise.
31305         (imm_store_chain_info::coalesce_immediate_stores): Likewise.
31306         (pass_store_merging::execute): Likewise.
31307         * gimple-ssa-strength-reduction.c (dump_candidate): Likewise.
31308         (ssa_base_cand_dump_callback): Likewise.
31309         (dump_incr_vec): Likewise.
31310         (replace_refs): Likewise.
31311         (replace_mult_candidate): Likewise.
31312         (create_add_on_incoming_edge): Likewise.
31313         (create_phi_basis): Likewise.
31314         (insert_initializers): Likewise.
31315         (all_phi_incrs_profitable): Likewise.
31316         (introduce_cast_before_cand): Likewise.
31317         (replace_one_candidate): Likewise.
31318         * gimplify.c (gimplify_expr): Likewise.
31319         * graphite-isl-ast-to-gimple.c (is_valid_rename): Likewise.
31320         (set_rename): Likewise.
31321         (rename_uses): Likewise.
31322         (copy_loop_phi_nodes): Likewise.
31323         (add_close_phis_to_merge_points): Likewise.
31324         (copy_loop_close_phi_args): Likewise.
31325         (copy_cond_phi_args): Likewise.
31326         (graphite_copy_stmts_from_block): Likewise.
31327         (translate_pending_phi_nodes): Likewise.
31328         * graphite-poly.c (print_pdr): Likewise.
31329         (dump_gbb_cases): Likewise.
31330         (dump_gbb_conditions): Likewise.
31331         (print_scop_params): Likewise.
31332         * graphite-scop-detection.c (build_cross_bb_scalars_def): Likewise.
31333         (build_cross_bb_scalars_use): Likewise.
31334         (gather_bbs::before_dom_children): Likewise.
31335         * hsa-dump.c (dump_hsa_immed): Likewise.
31336         * ipa-cp.c (print_ipcp_constant_value): Likewise.
31337         (get_replacement_map): Likewise.
31338         * ipa-inline-analysis.c (dump_condition): Likewise.
31339         (estimate_function_body_sizes): Likewise.
31340         * ipa-polymorphic-call.c (check_stmt_for_type_change): Likewise.
31341         (ipa_polymorphic_call_context::get_dynamic_type): Likewise.
31342         * ipa-prop.c (ipa_dump_param): Likewise.
31343         (ipa_print_node_jump_functions_for_edge): Likewise.
31344         (ipa_modify_call_arguments): Likewise.
31345         (ipa_modify_expr): Likewise.
31346         (ipa_dump_param_adjustments): Likewise.
31347         (ipa_dump_agg_replacement_values): Likewise.
31348         (ipcp_modif_dom_walker::before_dom_children): Likewise.
31349         * ipa-pure-const.c (check_stmt): Likewise.
31350         (pass_nothrow::execute): Likewise.
31351         * ipa-split.c (execute_split_functions): Likewise.
31352         * omp-offload.c (dump_oacc_loop_part): Likewise.
31353         (dump_oacc_loop): Likewise.
31354         * trans-mem.c (tm_log_emit): Likewise.
31355         (tm_memopt_accumulate_memops): Likewise.
31356         (dump_tm_memopt_set): Likewise.
31357         (dump_tm_memopt_transform): Likewise.
31358         * tree-cfg.c (gimple_verify_flow_info): Likewise.
31359         (print_loop): Likewise.
31360         * tree-chkp-opt.c (chkp_print_addr): Likewise.
31361         (chkp_gather_checks_info): Likewise.
31362         (chkp_get_check_result): Likewise.
31363         (chkp_remove_check_if_pass): Likewise.
31364         (chkp_use_outer_bounds_if_possible): Likewise.
31365         (chkp_reduce_bounds_lifetime): Likewise.
31366         * tree-chkp.c (chkp_register_addr_bounds): Likewise.
31367         (chkp_mark_completed_bounds): Likewise.
31368         (chkp_register_incomplete_bounds): Likewise.
31369         (chkp_mark_invalid_bounds): Likewise.
31370         (chkp_maybe_copy_and_register_bounds): Likewise.
31371         (chkp_build_returned_bound): Likewise.
31372         (chkp_get_bound_for_parm): Likewise.
31373         (chkp_build_bndldx): Likewise.
31374         (chkp_get_bounds_by_definition): Likewise.
31375         (chkp_generate_extern_var_bounds): Likewise.
31376         (chkp_get_bounds_for_decl_addr): Likewise.
31377         * tree-chrec.c (chrec_apply): Likewise.
31378         * tree-data-ref.c (dump_data_reference): Likewise.
31379         (dump_subscript): Likewise.
31380         (dump_data_dependence_relation): Likewise.
31381         (analyze_overlapping_iterations): Likewise.
31382         * tree-inline.c (expand_call_inline): Likewise.
31383         (tree_function_versioning): Likewise.
31384         * tree-into-ssa.c (dump_defs_stack): Likewise.
31385         (dump_currdefs): Likewise.
31386         (dump_names_replaced_by): Likewise.
31387         (dump_update_ssa): Likewise.
31388         (update_ssa): Likewise.
31389         * tree-object-size.c (pass_object_sizes::execute): Likewise.
31390         * tree-parloops.c (build_new_reduction): Likewise.
31391         (try_create_reduction_list): Likewise.
31392         (ref_conflicts_with_region): Likewise.
31393         (oacc_entry_exit_ok_1): Likewise.
31394         (oacc_entry_exit_single_gang): Likewise.
31395         * tree-pretty-print.h: Likewise.
31396         * tree-scalar-evolution.c (set_scalar_evolution): Likewise.
31397         (get_scalar_evolution): Likewise.
31398         (add_to_evolution): Likewise.
31399         (get_loop_exit_condition): Likewise.
31400         (analyze_evolution_in_loop): Likewise.
31401         (analyze_initial_condition): Likewise.
31402         (analyze_scalar_evolution): Likewise.
31403         (instantiate_scev): Likewise.
31404         (number_of_latch_executions): Likewise.
31405         (gather_chrec_stats): Likewise.
31406         (final_value_replacement_loop): Likewise.
31407         (scev_const_prop): Likewise.
31408         * tree-sra.c (dump_access): Likewise.
31409         (disqualify_candidate): Likewise.
31410         (create_access): Likewise.
31411         (reject): Likewise.
31412         (maybe_add_sra_candidate): Likewise.
31413         (create_access_replacement): Likewise.
31414         (analyze_access_subtree): Likewise.
31415         (analyze_all_variable_accesses): Likewise.
31416         (sra_modify_assign): Likewise.
31417         (initialize_constant_pool_replacements): Likewise.
31418         (find_param_candidates): Likewise.
31419         (decide_one_param_reduction): Likewise.
31420         (replace_removed_params_ssa_names): Likewise.
31421         * tree-ssa-ccp.c (ccp_fold_stmt): Likewise.
31422         * tree-ssa-copy.c (dump_copy_of): Likewise.
31423         (copy_prop_visit_cond_stmt): Likewise.
31424         * tree-ssa-dce.c (mark_operand_necessary): Likewise.
31425         * tree-ssa-dom.c (pass_dominator::execute): Likewise.
31426         (record_equivalences_from_stmt): Likewise.
31427         * tree-ssa-dse.c (compute_trims): Likewise.
31428         (delete_dead_call): Likewise.
31429         (delete_dead_assignment): Likewise.
31430         * tree-ssa-forwprop.c (forward_propagate_into_gimple_cond): Likewise.
31431         (forward_propagate_into_cond): Likewise.
31432         (pass_forwprop::execute): Likewise.
31433         * tree-ssa-ifcombine.c (ifcombine_ifandif): Likewise.
31434         * tree-ssa-loop-im.c (invariantness_dom_walker::before_dom_children):
31435         Likewise.
31436         (move_computations_worker): Likewise.
31437         (execute_sm): Likewise.
31438         * tree-ssa-loop-ivcanon.c (tree_estimate_loop_size): Likewise.
31439         (remove_exits_and_undefined_stmts): Likewise.
31440         (remove_redundant_iv_tests): Likewise.
31441         * tree-ssa-loop-ivopts.c (dump_use): Likewise.
31442         (adjust_iv_update_pos): Likewise.
31443         * tree-ssa-math-opts.c (bswap_replace): Likewise.
31444         * tree-ssa-phiopt.c (factor_out_conditional_conversion): Likewise.
31445         (value_replacement): Likewise.
31446         * tree-ssa-phiprop.c (phiprop_insert_phi): Likewise.
31447         * tree-ssa-pre.c (print_pre_expr): Likewise.
31448         (get_representative_for): Likewise.
31449         (create_expression_by_pieces): Likewise.
31450         (insert_into_preds_of_block): Likewise.
31451         (eliminate_insert): Likewise.
31452         (eliminate_dom_walker::before_dom_children): Likewise.
31453         (eliminate): Likewise.
31454         (remove_dead_inserted_code): Likewise.
31455         * tree-ssa-propagate.c (substitute_and_fold): Likewise.
31456         * tree-ssa-reassoc.c (get_rank): Likewise.
31457         (eliminate_duplicate_pair): Likewise.
31458         (eliminate_plus_minus_pair): Likewise.
31459         (eliminate_not_pairs): Likewise.
31460         (undistribute_ops_list): Likewise.
31461         (eliminate_redundant_comparison): Likewise.
31462         (update_range_test): Likewise.
31463         (optimize_range_tests_var_bound): Likewise.
31464         (optimize_vec_cond_expr): Likewise.
31465         (rewrite_expr_tree): Likewise.
31466         (rewrite_expr_tree_parallel): Likewise.
31467         (linearize_expr): Likewise.
31468         (break_up_subtract): Likewise.
31469         (linearize_expr_tree): Likewise.
31470         (attempt_builtin_powi): Likewise.
31471         (attempt_builtin_copysign): Likewise.
31472         (transform_stmt_to_copy): Likewise.
31473         (transform_stmt_to_multiply): Likewise.
31474         (dump_ops_vector): Likewise.
31475         * tree-ssa-sccvn.c (vn_nary_build_or_lookup_1): Likewise.
31476         (print_scc): Likewise.
31477         (set_ssa_val_to): Likewise.
31478         (visit_reference_op_store): Likewise.
31479         (visit_use): Likewise.
31480         (sccvn_dom_walker::before_dom_children): Likewise.
31481         (run_scc_vn): Likewise.
31482         * tree-ssa-scopedtables.c (avail_exprs_stack::lookup_avail_expr):
31483         Likewise.
31484         (expr_hash_elt::print): Likewise.
31485         (const_and_copies::pop_to_marker): Likewise.
31486         (const_and_copies::record_const_or_copy_raw): Likewise.
31487         * tree-ssa-structalias.c (compute_dependence_clique): Likewise.
31488         * tree-ssa-uninit.c (collect_phi_def_edges): Likewise.
31489         (dump_predicates): Likewise.
31490         (find_uninit_use): Likewise.
31491         (warn_uninitialized_phi): Likewise.
31492         (pass_late_warn_uninitialized::execute): Likewise.
31493         * tree-ssa.c (verify_vssa): Likewise.
31494         (verify_ssa): Likewise.
31495         (maybe_optimize_var): Likewise.
31496         * tree-vrp.c (dump_value_range): Likewise.
31497         (dump_all_value_ranges): Likewise.
31498         (dump_asserts_for): Likewise.
31499         (register_edge_assert_for_2): Likewise.
31500         (vrp_visit_cond_stmt): Likewise.
31501         (vrp_visit_switch_stmt): Likewise.
31502         (vrp_visit_stmt): Likewise.
31503         (vrp_visit_phi_node): Likewise.
31504         (simplify_cond_using_ranges_1): Likewise.
31505         (fold_predicate_in): Likewise.
31506         (evrp_dom_walker::before_dom_children): Likewise.
31507         (evrp_dom_walker::push_value_range): Likewise.
31508         (evrp_dom_walker::pop_value_range): Likewise.
31509         (execute_early_vrp): Likewise.
31511 2017-05-16  Richard Biener  <rguenther@suse.de>
31513         * dwarf2out.c (loc_list_from_tree_1): Do not create
31514         DW_OP_GNU_variable_value for DECL_IGNORED_P decls.
31516 2017-05-16  Richard Biener  <rguenther@suse.de>
31518         * dwarf2out.c (resolve_variable_value_in_expr): Lookup DIE
31519         just generated.
31520         (note_variable_value_in_expr): If we resolved the decl ref
31521         do not push to the stack.
31523 2017-05-16  Matthew Wahab  <matthew.wahab@arm.com>
31525         * config/arm/arm_neon.h (vadd_f16): Use standard arithmetic
31526         operations in fast-math mode.
31527         (vaddq_f16): Likewise.
31528         (vmul_f16): Likewise.
31529         (vmulq_f16): Likewise.
31530         (vsub_f16): Likewise.
31531         (vsubq_f16): Likewise.
31532         * config/arm/neon.md (add<mode>3): New.
31533         (sub<mode>3): New.
31534         (fma:<VH:mode>3): New.  Also remove outdated comment.
31535         (mul<mode>3): New.
31537 2017-05-16  Martin Liska  <mliska@suse.cz>
31539         PR ipa/79849.
31540         PR ipa/79850.
31541         * ipa-devirt.c (warn_types_mismatch): Fix typo.
31542         (odr_types_equivalent_p): Likewise.
31544 2017-05-15  Sylvestre Ledru  <sylvestre@debian.org>
31546         * plugin.c (try_init_one_plugin): Fix ressource leaks (CID 726637).
31548 2017-05-15  Uros Bizjak  <ubizjak@gmail.com>
31550         PR target/80425
31551         * config/i386.i386.md (*zero_extendsidi2): Do not penalize
31552         non-interunit SSE move alternatives with '?'.
31553         (zero-extendsidi peephole2): New peephole to skip intermediate
31554         general register in SSE zero-extend sequence.
31556 2017-05-15  Jeff Law  <law@redhat.com>
31558         * reorg.c (relax_delay_slots): Create a new variable to hold
31559         the temporary target rather than clobbering TARGET_LABEL.
31561         * config/tilegx/tilegx.c (tilegx_expand_unaligned_load): Add
31562         missing argument to extract_bit_field call.
31563         * config/tilepro/tilepro.c (tilepro_expand_unaligned_load): Likewise.
31565 2017-05-15  Martin Liska  <mliska@suse.cz>
31567         PR driver/31468
31568         * gcc.c (process_command): Do not allow empty argument of -o option.
31570 2017-05-15  Renlin Li  <renlin.li@arm.com>
31572         * config/aarch64/aarch64-protos.h (aarch64_expand_call): Declare.
31573         * config/aarch64/aarch64.c (aarch64_expand_call): Define.
31574         * config/aarch64/constraints.md (Usf): Add long call check.
31575         * config/aarch64/aarch64.md (call): Use aarch64_expand_call.
31576         (call_value): Likewise.
31577         (sibcall): Likewise.
31578         (sibcall_value): Likewise.
31579         (call_insn): New.
31580         (call_value_insn): New.
31581         (sibcall_insn): Update rtx pattern.
31582         (sibcall_value_insn): Likewise.
31583         (call_internal): Remove.
31584         (call_value_internal): Likewise.
31585         (sibcall_internal): Likewise.
31586         (sibcall_value_internal): Likewise.
31587         (call_reg): Likewise.
31588         (call_symbol): Likewise.
31589         (call_value_reg): Likewise.
31590         (call_value_symbol): Likewise.
31592 2017-05-14  Krister Walfridsson  <krister.walfridsson@gmail.com>
31594         PR target/80600
31595         * config/netbsd.h (NETBSD_LIBGCC_SPEC): Always add -lgcc.
31597 2017-05-14  Uros Bizjak  <ubizjak@gmail.com>
31599         * config/i386.i386.c (ix86_cc_modes_compatible): CCNOmode is
31600         compatible with CCGOCmode and with CCZmode.
31602 2017-05-14  Martin Sebor  <msebor@redhat.com>
31604         PR middle-end/77671
31605         * gimple-fold.c (gimple_fold_builtin_sprintf): Make extern.
31606         (gimple_fold_builtin_snprintf): Same.
31607         * gimple-fold.h (gimple_fold_builtin_sprintf): Declare.
31608         (gimple_fold_builtin_snprintf): Same.
31609         * gimple-ssa-sprintf.c (get_format_string): Correct the detection
31610         of character types.
31611         (is_call_safe): New function.
31612         (try_substitute_return_value): Call it.
31613         (try_simplify_call): New function.
31614         (pass_sprintf_length::handle_gimple_call): Call it.
31616 2017-05-14  Martin Sebor  <msebor@redhat.com>
31618         PR middle-end/80669
31619         * builtins.c (expand_builtin_stpncpy): Simplify.
31621 2017-05-14  Daniel Santos  <daniel.santos@pobox.com>
31623         * config/i386/i386.opt: Add option -mcall-ms2sysv-xlogues.
31624         * config/i386/i386.h
31625         (x86_64_ms_sysv_extra_clobbered_registers): Change type to unsigned.
31626         (NUM_X86_64_MS_CLOBBERED_REGS): New macro.
31627         (struct machine_function): Add new members call_ms2sysv,
31628         call_ms2sysv_pad_in, call_ms2sysv_pad_out and call_ms2sysv_extra_regs.
31629         (struct machine_frame_state): New fields sp_realigned and
31630         sp_realigned_offset.
31631         * config/i386/i386.c
31632         (enum xlogue_stub): New enum.
31633         (enum xlogue_stub_sets): New enum.
31634         (class xlogue_layout): New class.
31635         (struct ix86_frame): New fields stack_realign_allocate_offset,
31636         stack_realign_offset and outlined_save_offset.  Modify comments to
31637         detail stack layout when using out-of-line stubs.
31638         (ix86_target_string): Add -mcall-ms2sysv-xlogues option.
31639         (ix86_option_override_internal): Add sorry() for TARGET_SEH and
31640         -mcall-ms2sysv-xlogues.
31641         (stub_managed_regs): New static variable.
31642         (ix86_save_reg): Add new parameter ignore_outlined to optionally omit
31643         registers managed by out-of-line stub.
31644         (disable_call_ms2sysv_xlogues): New function.
31645         (ix86_compute_frame_layout): Modify re-alignment calculations, disable
31646         m->call_ms2sysv when appropriate and compute frame layout for
31647         out-of-line stubs.
31648         (sp_valid_at, fp_valid_at): New inline functions.
31649         (choose_basereg): New function.
31650         (choose_baseaddr): Add align parameter, use choose_basereg and modify
31651         all callers.
31652         (ix86_emit_save_reg_using_mov, ix86_emit_restore_sse_regs_using_mov):
31653         Use align parameter of choose_baseaddr to generated aligned SSE movs
31654         when possible.
31655         (pro_epilogue_adjust_stack): Modify to track
31656         machine_frame_state::sp_realigned.
31657         (ix86_nsaved_regs): Modify to accommodate changes to ix86_save_reg.
31658         (ix86_nsaved_sseregs): Likewise.
31659         (ix86_emit_save_regs): Likewise.
31660         (ix86_emit_save_regs_using_mov): Likewise.
31661         (ix86_emit_save_sse_regs_using_mov): Likewise.
31662         (get_scratch_register_on_entry): Likewise.
31663         (gen_frame_set): New function.
31664         (gen_frame_load): Likewise.
31665         (gen_frame_store): Likewise.
31666         (emit_outlined_ms2sysv_save): Likewise.
31667         (emit_outlined_ms2sysv_restore): Likewise.
31668         (ix86_expand_prologue): Modify stack re-alignment code and call
31669         emit_outlined_ms2sysv_save when appropriate.
31670         (ix86_emit_leave): Clear machine_frame_state::sp_realigned.  Add
31671         parameter rtx_insn *insn, which allows the function to be used to only
31672         generate the notes.
31673         (ix86_expand_epilogue): Modify validity checks of frame and stack
31674         pointers, and call emit_outlined_ms2sysv_restore when appropriate.
31675         (ix86_expand_call): Modify to enable m->call_ms2sysv when appropriate.
31676         * config/i386/predicates.md
31677         (save_multiple): New predicate.
31678         (restore_multiple): Likewise.
31679         * config/i386/sse.md
31680         (save_multiple<mode>): New pattern.
31681         (save_multiple_realign<mode>): Likewise.
31682         (restore_multiple<mode>): Likewise.
31683         (restore_multiple_and_return<mode>): Likewise.
31684         (restore_multiple_leave_return<mode>): Likewise.
31685         * Makefile.in: Export HOSTCXX and HOSTCXXFLAGS to site.exp
31687 2017-05-14  Julia Koval  <julia.koval@intel.com>
31689         * config/i386/i386-builtin-types.def (VOID_FTYPE_INT_INT64): New type.
31690         * config/i386/i386-builtin.def (__builtin_ia32_xgetbv)
31691         (__builtin_ia32_xsetbv): New builtins.
31692         * config/i386/i386.c (ix86_expand_special_args_builtin):
31693         Process new types.
31694         (ix86_expand_builtin): Special expand for new intrinsics.
31695         * config/i386/i386.md (UNSPECV_XGETBV, UNSPECV_XSETBV): New.
31696         (xsetbv, xsetbv_rex64, xgetbv, xgetbv_rex64): New insn patterns.
31697         * config/i386/xsaveintrin.h (_xsetbv, _getbv): New intrinsics.
31699 2017-05-13  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
31701         * cfganal.c (inverted_post_order_compute): Change argument type
31702         to vec *.
31703         * cfganal.h (inverted_post_order_compute): Adjust prototype.
31704         * df-core.c (rest_of_handle_df_initialize): Adjust.
31705         (rest_of_handle_df_finish): Likewise.
31706         (df_analyze_1): Likewise.
31707         (df_analyze): Likewise.
31708         (loop_inverted_post_order_compute): Change argument to be a vec *.
31709         (df_analyze_loop): Adjust.
31710         (df_get_n_blocks): Likewise.
31711         (df_get_postorder): Likewise.
31712         * df.h (struct df_d): Change field to be a vec.
31713         * lcm.c (compute_laterin): Adjust.
31714         (compute_available): Likewise.
31715         * lra-lives.c (lra_create_live_ranges_1): Likewise.
31716         * tree-ssa-dce.c (remove_dead_stmt): Likewise.
31717         * tree-ssa-pre.c (compute_antic): Likewise.
31719 2017-05-13  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
31721         * cfganal.c (connect_infinite_loops_to_exit): Adjust.
31722         (depth_first_search::depth_first_search): Change structure init
31723         function to this constructor.
31724         (depth_first_search::add_bb): Rename function to this member.
31725         (depth_first_search::execute): Likewise.
31726         (flow_dfs_compute_reverse_finish): Adjust.
31728 2017-05-13  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
31730         * ddg.c (find_nodes_on_paths): Use auto_sbitmap.
31731         (longest_simple_path): Likewise.
31732         * shrink-wrap.c (spread_components): Likewise.
31733         (disqualify_problematic_components): Likewise.
31734         (emit_common_heads_for_components): Likewise.
31735         (emit_common_tails_for_components): Likewise.
31736         (insert_prologue_epilogue_for_components): Likewise.
31738 2017-05-13  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
31740         * tree-ssa-dse.c (dse_dom_walker): Make m_live_byes a
31741         auto_sbitmap.
31743 2017-05-13  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
31745         * df-core.c (df_set_blocks): Start using auto_bitmap.
31746         (df_compact_blocks): Likewise.
31747         * df-problems.c (df_rd_confluence_n): Likewise.
31748         * df-scan.c (df_insn_rescan_all): Likewise.
31749         (df_process_deferred_rescans): Likewise.
31750         (df_update_entry_block_defs): Likewise.
31751         (df_update_exit_block_uses): Likewise.
31752         (df_entry_block_bitmap_verify): Likewise.
31753         (df_exit_block_bitmap_verify): Likewise.
31754         (df_scan_verify): Likewise.
31755         * lra-constraints.c (lra_constraints): Likewise.
31756         (undo_optional_reloads): Likewise.
31757         (lra_undo_inheritance): Likewise.
31758         * lra-remat.c (calculate_gen_cands): Likewise.
31759         (do_remat): Likewise.
31760         * lra-spills.c (assign_spill_hard_regs): Likewise.
31761         (spill_pseudos): Likewise.
31762         * tree-ssa-pre.c (bitmap_set_and): Likewise.
31763         (bitmap_set_subtract_values): Likewise.
31765 2017-05-13  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
31767         * haifa-sched.c (estimate_shadow_tick): Replace manual bitmap
31768         management with auto_bitmap.
31769         (fix_inter_tick): Likewise.
31770         (fix_recovery_deps): Likewise.
31771         * ira.c (add_store_equivs): Likewise.
31772         (find_moveable_pseudos): Likewise.
31773         (split_live_ranges_for_shrink_wrap): Likewise.
31774         * print-rtl.c (rtx_reuse_manager::rtx_reuse_manager): Likewise.
31775         (rtx_reuse_manager::seen_def_p): Likewise.
31776         (rtx_reuse_manager::set_seen_def): Likewise.
31777         * print-rtl.h (class rtx_reuse_manager): Likewise.
31779 2017-05-13  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
31781         * bt-load.c (combine_btr_defs): Use auto_bitmap to manage bitmap
31782         lifetime.
31783         (migrate_btr_def): Likewise.
31784         * cfgloop.c (get_loop_body_in_bfs_order): Likewise.
31785         * df-core.c (loop_post_order_compute): Likewise.
31786         (loop_inverted_post_order_compute): Likewise.
31787         * hsa-common.h: Likewise.
31788         * hsa-gen.c (hsa_bb::~hsa_bb): Likewise.
31789         * init-regs.c (initialize_uninitialized_regs): Likewise.
31790         * ipa-inline.c (resolve_noninline_speculation): Likewise.
31791         (inline_small_functions): Likewise.
31792         * ipa-reference.c (ipa_reference_write_optimization_summary): Likewise.
31793         * ira.c (combine_and_move_insns): Likewise.
31794         (build_insn_chain): Likewise.
31795         * loop-invariant.c (find_invariants): Likewise.
31796         * lower-subreg.c (propagate_pseudo_copies): Likewise.
31797         * predict.c (tree_predict_by_opcode): Likewise.
31798         (predict_paths_leading_to): Likewise.
31799         (predict_paths_leading_to_edge): Likewise.
31800         (estimate_loops_at_level): Likewise.
31801         (estimate_loops): Likewise.
31802         * shrink-wrap.c (try_shrink_wrapping): Likewise.
31803         (spread_components): Likewise.
31804         * tree-cfg.c (remove_edge_and_dominated_blocks): Likewise.
31805         * tree-loop-distribution.c (rdg_build_partitions): Likewise.
31806         * tree-predcom.c (tree_predictive_commoning_loop): Likewise.
31807         * tree-ssa-coalesce.c (coalesce_ssa_name): Likewise.
31808         * tree-ssa-phionlycprop.c (pass_phi_only_cprop::execute): Likewise.
31809         * tree-ssa-pre.c (remove_dead_inserted_code): Likewise.
31810         * tree-ssa-sink.c (nearest_common_dominator_of_uses): Likewise.
31811         * tree-ssa-threadupdate.c (compute_path_counts): Likewise.
31812         (mark_threaded_blocks): Likewise.
31813         (thread_through_all_blocks): Likewise.
31814         * tree-ssa.c (verify_ssa): Likewise.
31815         (execute_update_addresses_taken): Likewise.
31816         * tree-ssanames.c (verify_ssaname_freelists): Likewise.
31818 2017-05-13  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
31820         * cfganal.c (mark_dfs_back_edges): Replace manual stack with
31821         auto_vec.
31822         (post_order_compute): Likewise.
31823         (inverted_post_order_compute): Likewise.
31824         (pre_and_rev_post_order_compute_fn): Likewise.
31826 2017-05-13  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
31828         * genrecog.c (int_set::int_set): Explicitly construct our
31829         auto_vec base class.
31830         * vec.h (auto_vec::auto_vec): New constructor.
31832 2017-05-13  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
31834         * bitmap.h (class auto_bitmap): New constructor taking
31835         bitmap_obstack * argument.
31837 2017-05-13  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
31839         * bitmap.h (class auto_bitmap): Change type of m_bits to
31840         bitmap_head, and adjust ctor / dtor and member operators.
31842 2017-05-13  Uros Bizjak  <ubizjak@gmail.com>
31844         * compare-elim.c (equivalent_reg_at_start): Return NULL_RTX
31845         when returned register mode doesn't match original mode.
31847 2017-05-12  Jeff Law  <law@redhat.com>
31848             Jakub Jelinek  <jakub@redhat.com>
31850         * config/mn10300/mn10300.c (mn10300_match_ccmode): Fix where
31851         we look for cc setter after the compare-elim changes.
31852         * config/mn10300/mn10300.md (addsi3_flags): Fix order of patterns
31853         within the vector to match what compare-elim now expects.
31854         (subsi3_flags, andsi3_flags, iorsi3_flags): Likewise.
31855         (xorsi3_flags, one_cmplsi2_flags): Likewise.
31857         * config/rx/rx.c (rx_match_ccmode): Fix where we look cc setter
31858         after the compare-elim changes.
31859         * config/rx/rx.md (abssi2_flags): Fix order of patterns within
31860         the vector to match what compare-elim now expects.
31861         (addsi3_flags, adc_flags, addsi3_flags peepholes): Likewise.
31862         (andsi3_flags, negsi2_flags, one_cmplsi2_flags): Likewise.
31863         (iorsi3_flags, rotlsi3_flags, rotrsi3_flags): Likewise.
31864         (ashrsi3_flags, lshrsi3_flags, ashlsi3_flags): Likewise.
31865         (ssaddsi3, subsi3_flags, sbb_flags, xorsi3_flags): Likewise.
31867         * config/visium/visium.c (single_set_and_flags): Fix where
31868         we look for cc setter after the compare-elim changes.
31869         * config/visium/visium.md (flags_subst_logic): Fix order of patterns
31870         with the vector to match what compare-elim now expects.
31871         (flags_subst_arith, add<mode>3_insn_set_carry): Likewise.
31872         (add<mode>3_insn_set_overflow, addsi3_insn_set_carry): Likewise.
31873         (addsi3_insn_set_overflow, sub<mode>3_insn_set_carry): Likewise.
31874         (sub<mode>3_insn_set_overflow, subsi3_insn_set_carry): Likewise.
31875         (subsi3_insn_set_overflow, negsi2_insn_set_carry): Likewise.
31876         (neg<mode>2_insn_set_overflow): Likewise.
31878 2017-05-12  Jim Wilson  <jim.wilson@linaro.org>
31880         PR middle-end/79794
31881         * expmed.c (extract_bit_field_1): Add alt_rtl argument.  Before
31882         maybe_expand_insn call, set ops[0].target.  If still set after call,
31883         set alt_rtl.  Add extra arg to recursive calls.
31884         (extract_bit_field): Add alt_rtl argument.  Pass to
31885         extract_bit_field.
31886         * expmed.h (extract_bit_field): Fix prototype.
31887         * expr.c (emit_group_load_1, copy_blkmode_from_reg)
31888         (copy_blkmode_to_reg, read_complex_part, store_field): Pass extra NULL
31889         to extract_bit_field_calls.
31890         (expand_expr_real_1): Pass alt_rtl to expand_expr_real instead of 0.
31891         Pass alt_rtl to extract_bit_field calls.
31892         * calls.c (store_unaligned_arguments_into_psuedos)
31893         load_register_parameters): Pass extra NULL to extract_bit_field calls.
31894         * optabs.c (maybe_legitimize_operand): Clear op->target when call
31895         gen_reg_rtx.
31896         * optabs.h (struct expand_operand): Add target bitfield.
31898 2017-05-12  Uros Bizjak  <ubizjak@gmail.com>
31900         * compare-elim.c (try_eliminate_compare): Canonicalize
31901         operation with embedded compare to
31902         [(set (reg:CCM) (compare:CCM (operation) (immediate)))
31903          (set (reg) (operation)].
31905         * config/i386/i386.c (TARGET_FLAGS_REGNUM): New define.
31907 2017-05-12  Uros Bizjak  <ubizjak@gmail.com>
31909         PR target/80723
31910         * config/i386/i386.c (ix86_rtx_cost) [case PLUS]: Ignore the
31911         cost of adding a carry flag for ADC instruction.
31912         [case MINUS]: Ignore the cost of subtracting a carry flag
31913         for SBB instruction.
31915 2017-05-12  Steven Munroe  <munroesj@gcc.gnu.org>
31917         * config.gcc (powerpc*-*-*): Add bmi2intrin.h, bmiintrin.h,
31918         and x86intrin.h
31919         * config/rs6000/bmiintrin.h: New file.
31920         * config/rs6000/bmi2intrin.h: New file.
31921         * config/rs6000/x86intrin.h: New file.
31923 2017-05-12  Jeff Law  <law@redhat.com>
31925         * tree-vrp.c (vrp_dom_walker::before_dom_children): Push unwinding
31926         markers.
31928 2017-05-12  Peter Bergner  <bergner@vnet.ibm.com>
31930         PR middle-end/80707
31931         * tree-cfg.c: Remove cfg edges of unreachable case statements.
31933 2017-05-12  Will Schmidt  <will_schmidt@vnet.ibm.com>
31935         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add handling for
31936         early expansion of vector divide builtins.
31937         (builtin_function_type): Add VSX_BUILTIN_UDIV_V2DI to the list of
31938         builtins identified as having unsigned arguments.
31940 2017-05-12  Will Schmidt  <will_schmidt@vnet.ibm.com>
31942         * config/rs6000/rs6000.c (gimple-fold.h): New #include.
31943         (rs6000_gimple_fold_builtin): Add handling for early GIMPLE
31944         expansion of vector logical operations (and, andc, or, xor,
31945         nor, orc, nand).
31947 2017-05-12  Will Schmidt  <will_schmidt@vnet.ibm.com>
31949         * gimple-fold.c (create_tmp_reg_or_ssa_name): Remove static declaration.
31950         * gimple-fold.h (create_tmp_reg_or_ssa_name): New prototype.
31952 2017-05-12  Will Schmidt  <will_schmidt@vnet.ibm.com>
31954         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add handling for
31955         early GIMPLE expansion of vector multiplies.
31957 2017-05-12  Prakhar Bahuguna  <prakhar.bahuguna@arm.com>
31959         * config/arm/arm.md (movsi): Add TARGET_32BIT in addition to the
31960         TARGET_HAVE_MOVT conditional.
31961         (movt splitter): Likewise.
31963 2017-05-12  Richard Biener  <rguenther@suse.de>
31965         * tree-ssa-sccvn.h (has_VN_INFO): Declare.
31966         * tree-ssa-pre.c (eliminate_dom_walker::before_dom_children):
31967         Fold all stmts not inplace.
31969 2017-05-12  Richard Biener  <rguenther@suse.de>
31971         PR tree-optimization/80713
31972         * tree-ssa-pre.c (remove_dead_inserted_code): Clear
31973         inserted_exprs bit for not removed stmts.
31975 2017-05-12  Thomas Schwinge  <thomas@codesourcery.com>
31977         PR middle-end/69921
31978         * tree-parloops.c (create_parallel_loop): Set "oacc kernels
31979         parallelized" attribute for parallelized OpenACC kernels.
31980         * omp-offload.c (execute_oacc_device_lower): Use it.
31982         * omp-expand.c (expand_omp_target) <GF_OMP_TARGET_KIND_OACC_KERNELS>:
31983         Set "oacc kernels" attribute.
31984         * omp-general.c (oacc_set_fn_attrib): Remove is_kernel formal
31985         parameter.  Adjust all users.
31986         (oacc_fn_attrib_kernels_p): Remove function.
31987         * omp-offload.c (execute_oacc_device_lower): Look for "oacc
31988         kernels" attribute instead of calling oacc_fn_attrib_kernels_p.
31989         * tree-ssa-loop.c (gate_oacc_kernels): Likewise.
31990         * tree-parloops.c (create_parallel_loop): If oacc_kernels_p,
31991         assert "oacc kernels" attribute is set.
31993 2017-05-11  Carl Love  <cel@us.ibm.com>
31995         * config/rs6000/rs6000-c: Add support for built-in functions
31996         vector unsigned char vec_popcnt (vector signed char)
31997         vector unsigned char vec_popcnt (vector unsigned char)
31998         vector unsigned short vec_popcnt (vector signed short)
31999         vector unsigned short vec_popcnt (vector unsigned short)
32000         vector unsigned int vec_popcnt (vector signed int)
32001         vector unsigned int vec_popcnt (vector unsigned int)
32002         vector unsigned long long vec_popcnt (vector signed long long)
32003         vector unsigned long long vec_popcnt (vector unsigned long long)
32004         vector signed long long vec_slo (vector signed long long,
32005                                          vector signed char)
32006         vector signed long long vec_slo (vector signed long long,
32007                                          vector unsigned char)
32008         vector unsigned long long vec_slo (vector unsigned long long,
32009                                            vector signed char)
32010         vector unsigned long long vec_slo (vector unsigned long long,
32011                                            vector unsigned char)
32012         * config/rs6000/rs6000-builtin.def: Add definitions for VPOPCNTUB,
32013         VPOPCNTUH, VPOPCNTUW, and VPOPCNTUD overloads.
32014         * config/rs6000/altivec.h: Add define for vec_popcnt, vec_popcntb,
32015         vec_popcnth, vec_popcntw and vec_popcntd built-in functions.
32016         * doc/extend.texi: Update the built-in documentation file for the
32017         new built-in functions.
32019 2017-05-11  Michael Meissner  <meissner@linux.vnet.ibm.com>
32021         * attribs.h (sorted_attr_string): Move machine independent
32022         functions for target clone support from the i386 port to common
32023         code.  Rename ix86_function_versions to common_function_versions.
32024         Rename make_name to make_unique_name.
32025         (common_function_versions): Likewise.
32026         (make_unique_name): Likewise.
32027         (make_dispatcher_decl): Likewise.
32028         (is_function_default_version): Likewise.
32029         * attribs.c (attr_strcmp): Likewise.
32030         (sorted_attr_string): Likewise.
32031         (common_function_versions): Likewise.
32032         (make_unique_name): Likewise.
32033         (make_dispatcher_decl): Likewise.
32034         (is_function_default_version): Likewise.
32035         * config/i386/i386.c (attr_strcmp): Likewise.
32036         (sorted_attr_string): Likewise.
32037         (ix86_function_versions): Likewise.
32038         (make_name): Likewise.
32039         (make_dispatcher_decl): Likewise.
32040         (is_function_default_version): Likewise.
32041         (TARGET_OPTION_FUNCTION_VERSIONS): Update target function hook.
32043 2017-05-11  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
32045         PR target/80695
32046         * config/rs6000/rs6000.c (rs6000_builtin_vectorization_cost):
32047         Account for direct move costs for vec_construct of integer
32048         vectors.
32050 2017-05-11  Uros Bizjak  <ubizjak@gmail.com>
32052         PR target/80706
32053         * config/i386/sync.md (UNSPEC_LDX_ATOMIC): New unspec.
32054         (UNSPEC_STX_ATOMIC): Ditto.
32055         (loaddi_via_sse): New insn.
32056         (storedi_via_sse): Ditto.
32057         (atomic_loaddi_fpu): Emit loaddi_via_sse and storedi_via_sse.
32058         Update corresponding peephole2 patterns.
32059         (atomic_storedi_fpu): Ditto.
32061 2017-05-11  Julia Koval  <julia.koval@intel.com>
32063         * config/i386/avx512fintrin.h (_mm_mask_rsqrt14_sd)
32064         (_mm_maskz_rsqrt14_sd, _mm_mask_rsqrt14_ss, _mm_maskz_rsqrt14_ss):
32065         New intrinsics.
32066         * config/i386/i386-builtin.def (__builtin_ia32_rsqrt14sd_mask)
32067         (__builtin_ia32_rsqrt14ss_mask): New builtins.
32068         * config/i386/sse.md (rsqrt14_<mode>_mask): New pattern.
32070 2017-05-11  Nathan Sidwell  <nathan@acm.org>
32072         * graphite-poly.c: Include dumpfile.h.
32074         * dumpfle.h (dump_function): Declare here ...
32075         * tree-dump.h (dump_function): ... not here.
32076         * dumpfile.c: #include tree-cfg.h.
32077         (dump_function): Move here from ...
32078         * tree-dump.c (dump_function): ... here.
32079         * gimplify.c: #include splay-tree.h, not tree-dump.h.
32080         * graphite-poly.c: Don't include tree-dump.h.
32081         * cgraphclones.c: Include dumpfile.h not tree-dump.h.
32082         * print-tree.c: Likewise.
32083         * stor-layout.c: Likewise.
32084         * tree-nested.c: Likewise.
32086         * dumpfile.c (dump_start): Use TDF_FLAGS.
32087         (dump_enable_all): Fix TDF_KIND check thinko.
32089 2017-05-11  Kelvin Nilsen  <kelvin@gcc.gnu.org>
32091         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
32092         array entries to represent two legal parameterizations of the
32093         overloaded __builtin_cmpb function, as represented by the
32094         P6_OV_BUILTIN_CMPB constant.
32095         (altivec_resolve_overloaded_builtin): Add special case handling
32096         for the __builtin_cmpb function, as represented by the
32097         P6_OV_BUILTIN_CMPB constant.
32098         * config/rs6000/rs6000-builtin.def (BU_P6_2): New macro.
32099         (BU_P6_64BIT_2): New macro.
32100         (BU_P6_OVERLOAD_2): New macro
32101         (CMPB_32): Add 32-bit compare-bytes support for 32-bit only targets.
32102         (CMPB): Add 64-bit compare-bytes support for 32-bit and 64-bit targets.
32103         (CMPB): Add overload support to represent both 32-bit and 64-bit
32104         compare-bytes function.
32105         * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add
32106         support for TARGET_CMPB.
32107         * config/rs6000/rs6000.h: Add support for RS6000_BTM_CMPB.
32108         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Add
32109         documentation of the __builtin_cmpb overloaded built-in function.
32111 2017-05-11  Richard Biener  <rguenther@suse.de>
32113         PR tree-optimization/80705
32114         * tree-vect-data-refs.c (vect_analyze_data_refs): DECL_NONALIASED
32115         bases are not vectorizable.
32117 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32119         * tree-ssa-loop-ivopts.c (determine_set_costs): Skip non-interger
32120         when counting register pressure.
32122 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32124         * tree-ssa-loop-ivopts.c (compare_cost_pair): New.
32125         (iv_ca_more_deps): Renamed to ...
32126         (iv_ca_compare_deps): ... this.
32127         (iv_ca_extend): Extend iv_ca if NEW_CP is cheaper than OLD_CP.
32129 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32131         * tree-ssa-loop-ivopts.c (find_interesting_uses): Move inv vars dump
32132         to ...
32133         (determine_group_iv_costs): ... here.
32134         (find_inv_vars_cb): Record inv var if it's not recorded before.
32136 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32138         * tree-ssa-loop-ivopts.c (allow_ip_end_pos_p): Refine comments.
32139         (get_shiftadd_cost): Ditto.
32141 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32143         * tree-ssa-address.c: Include header file.
32144         (move_hint_to_base): Return TRUE if BASE_HINT is moved to memory
32145         address.
32146         (add_to_parts): Refactor.
32147         (addr_to_parts): New parameter.  Update use of move_hint_to_base.
32148         (create_mem_ref): Update use of addr_to_parts.  Re-associate addr
32149         in new order.
32151 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32153         PR tree-optimization/53090
32154         * tree-ssa-loop-ivopts.c (enum comp_iv_rewrite): New enum value
32155         COMP_IV_EXPR_2.
32156         (extract_cond_operands): Detect condition with IV on both sides
32157         and return COMP_IV_EXPR_2.
32158         (find_interesting_uses_cond): Add iv_use for both IVs in condition.
32159         (rewrite_use_compare): Simplify by removing call to function
32160         extract_cond_operands.
32162 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32164         * tree-ssa-loop-ivopts.c (enum comp_iv_rewrite): New.
32165         (extract_cond_operands): Detect condition comparing against non-
32166         invariant bound and return appropriate enum value.
32167         (find_interesting_uses_cond): Update use of extract_cond_operands.
32168         Handle its return value accordingly.
32169         (determine_group_iv_cost_cond, rewrite_use_compare): Ditto.
32171 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32173         * tree-ssa-loop-ivopts.c (rewrite_use_nonlinear_expr): Re-associate
32174         nonlinear iv_use computation in loop invariant sensitive way.
32176 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32178         * tree-ssa-loop-ivopts.c (relate_compare_use_with_all_cands): New.
32179         (find_iv_candidates): Call relate_compare_use_with_all_cands.
32181 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32183         * tree-ssa-loop-ivopts.c (struct iv_cand): New field inv_exprs.
32184         (dump_cand): Support iv_cand.inv_exprs.
32185         (add_candidate_1): Record invariant exprs in iv_cand.inv_exprs
32186         for candidates.
32187         (iv_ca_set_no_cp, iv_ca_set_cp, free_loop_data): Support
32188         iv_cand.inv_exprs.
32190 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32192         * tree-ssa-loop-ivopts.c (multiplier_allowed_in_address_p): Move
32193         from ...
32194         * tree-ssa-address.c (multiplier_allowed_in_address_p): ... to here
32195         as local function.  Include necessary header files.
32196         * tree-ssa-loop-ivopts.h (multiplier_allowed_in_address_p): Delete.
32198 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32200         * tree-ssa-loop-ivopts.c (autoinc_possible_for_pair): Simplify.
32202 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32204         * tree-ssa-loop-ivopts.c (force_expr_to_var_cost): Handle more
32205         operators: TRUNC_DIV_EXPR, BIT_AND_EXPR, BIT_IOR_EXPR, LSHIFT_EXPR,
32206         RSHIFT_EXPR and BIT_NOT_EXPR.
32208 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32210         * tree-ssa-loop-ivopts.c (get_loop_invariant_expr): Simplify.
32211         (adjust_setup_cost): New parameter supporting round up adjustment.
32212         (struct address_cost_data): Delete.
32213         (force_expr_to_var_cost): Don't bound cost with spill_cost.
32214         (split_address_cost, ptr_difference_cost): Delete.
32215         (difference_cost, compare_aff_trees, record_inv_expr): Delete.
32216         (struct ainc_cost_data): New struct.
32217         (get_address_cost_ainc): New function.
32218         (get_address_cost, get_computation_cost): Reimplement.
32219         (determine_group_iv_cost_address): Record inv_expr for all uses of
32220         a group.
32221         (determine_group_iv_cost_cond): Call get_loop_invariant_expr.
32222         (iv_ca_has_deps): Reimplemented to ...
32223         (iv_ca_more_deps): ... this.  Check if NEW_CP introduces more deps
32224         than OLD_CP.
32225         (iv_ca_extend): Call iv_ca_more_deps.
32227 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32229         * tree-ssa-address.c (struct mem_address): Move to header file.
32230         (valid_mem_ref_p, move_fixed_address_to_symbol): Make it global.
32231         * tree-ssa-address.h (struct mem_address): Move from C file.
32232         (valid_mem_ref_p, move_fixed_address_to_symbol): Declare.
32234 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32236         * tree-affine.h (aff_combination_type): New interface.
32237         (aff_combination_zero_p): Remove static.
32238         (aff_combination_const_p): New interface.
32239         (aff_combination_singleton_var_p): New interfaces.
32241 2017-05-11  Richard Biener  <rguenther@suse.de>
32243         * tree-ssa-pre.c (eliminate_dom_walker::before_dom_children):
32244         Skip unreachable blocks and destinations.
32245         (eliminate): Move stmt removal and fixup ...
32246         (fini_eliminate): ... here.  Skip inserted exprs.
32247         (pass_pre::execute): Move fini_pre after fini_eliminate.
32248         * tree-ssa-tailmerge.c: Include tree-cfgcleanup.h.
32249         (tail_merge_optimize): Run cleanup_tree_cfg if requested by
32250         PRE to get rid of dead code that has invalid SSA form and
32251         split critical edges again.
32253 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32255         * rtlanal.c (rtx_cost): Handle TRUNCATE between tieable modes.
32257 2017-05-11  Richard Biener  <rguenther@suse.de>
32259         * passes.c (execute_function_todo): Verify loops if they are
32260         said to be up-to-date.
32261         * cfgexpand.c (pass_expand::execute): Discard loops for -dx.
32262         * trans-mem.c (pass_tm_edges::execute): Mark loops for fixup.
32264 2017-05-10  John David Anglin  <danglin@gcc.gnu.org>
32266         PR target/80090
32267         * config/pa/pa.c (pa_assemble_integer): When outputting a SYMBOL_REF,
32268         handle calling assemble_external ourself.
32270         PR target/79027
32271         * config/pa/pa.c (pa_cannot_change_mode_class): Reject changes to/from
32272         modes with zero size.  Enhance comment.
32274 2017-05-10  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
32276         * config/rs6000/rs6000.c (altivec_init_builtins): Define POWER8
32277         built-ins for vec_xl and vec_xst with short and char pointer
32278         arguments.
32280 2017-05-10  Sebastian Peryt  <sebastian.peryt@intel.com>
32282         * config/i386/avx512fintrin.h (_mm_mask_max_round_sd)
32283         (_mm_maskz_max_round_sd, _mm_mask_max_round_ss)
32284         (_mm_maskz_max_round_ss, _mm_mask_min_round_sd)
32285         (_mm_maskz_min_round_sd, _mm_mask_min_round_ss)
32286         (_mm_maskz_min_round_ss): New intrinsics.
32287         * config/i386/i386-builtin-types.def (V2DF, V2DF, V2DF, V2DF, UQI, INT)
32288         (V4SF, V4SF, V4SF, V4SF, UQI, INT): New function type aliases.
32289         * config/i386/i386-builtin.def (__builtin_ia32_maxsd_mask_round)
32290         (__builtin_ia32_maxss_mask_round, __builtin_ia32_minsd_mask_round)
32291         (__builtin_ia32_minss_mask_round): New builtins.
32292         * config/i386/i386.c (V2DF_FTYPE_V2DF_V2DF_V2DF_UQI_INT)
32293         (V4SF_FTYPE_V4SF_V4SF_V4SF_UQI_INT): Handle new types.
32294         * config/i386/sse.md (<sse>_vm<code><mode>3<round_saeonly_name>):
32295         Rename to ...
32296         (<sse>_vm<code><mode>3<mask_name><round_saeonly_name>): ... this.
32297         (v<maxmin_float><ssescalarmodesuffix>\t{<round_saeonly_op3>%2, %1, %0|%0, %1, %<iptr>2<round_saeonly_op3>}):
32298         Change to ...
32299         (v<maxmin_float><ssescalarmodesuffix>\t{<round_saeonly_mask_op3>%2, %1, %0<mask_operand3>|%0<mask_operand3>, %1, %<iptr>2<round_saeonly_mask_op3>}):
32300         ... this.
32302 2017-05-10  Sebastian Peryt  <sebastian.peryt@intel.com>
32304         * config/i386/avx512fintrin.h (_mm_mask_mul_round_sd)
32305         (_mm_maskz_mul_round_sd, _mm_mask_mul_round_ss)
32306         (_mm_maskz_mul_round_ss, _mm_mask_div_round_sd)
32307         (_mm_maskz_div_round_sd, _mm_mask_div_round_ss)
32308         (_mm_maskz_div_round_ss, _mm_mask_mul_sd, _mm_maskz_mul_sd)
32309         (_mm_mask_mul_ss, _mm_maskz_mul_ss, _mm_mask_div_sd)
32310         (_mm_maskz_div_sd, _mm_mask_div_ss, _mm_maskz_div_ss): New intrinsics.
32311         * config/i386/i386-builtin-types.def (V2DF_FTYPE_V2DF_V2DF_V2DF_UQI_INT)
32312         (V4SF_FTYPE_V4SF_V4SF_V4SF_UQI_INT): New function type aliases.
32313         * config/i386/i386-builtin.def (__builtin_ia32_divsd_mask_round)
32314         (__builtin_ia32_divss_mask_round, __builtin_ia32_mulsd_mask_round)
32315         (__builtin_ia32_mulss_mask_round): New builtins.
32316         * config/i386/i386.c (V2DF_FTYPE_V2DF_V2DF_V2DF_UQI_INT)
32317         (V4SF_FTYPE_V4SF_V4SF_V4SF_UQI_INT): Handle new types.
32318         * config/i386/sse.md (<sse>_vm<multdiv_mnemonic><mode>3<round_name>):
32319         Rename to ...
32320         (<sse>_vm<multdiv_mnemonic><mode>3<mask_name><round_name>): ... this.
32321         (v<multdiv_mnemonic><ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|%0, %1, %<iptr>2<round_op3>}):
32322         Change to ...
32323         (v<multdiv_mnemonic><ssescalarmodesuffix>\t{<round_mask_op3>%2, %1, %0<mask_operand3>|%0<mask_operand3>, %1, %<iptr>2<round_mask_op3>}):
32324         ... this.
32326 2017-05-10  Julia Koval  <julia.koval@intel.com>
32328         * config/i386/avxintrin.h (_mm256_set_m128, _mm256_set_m128d)
32329         (_mm256_set_m128i, _mm256_setr_m128, _mm256_setr_m128d)
32330         (_mm256_setr_m128i): New intrinsics.
32332 2017-05-10  Julia Koval  <julia.koval@intel.com>
32334         * config/i386/avx512fintrin.h (_mm_mask_rcp14_sd)
32335         (_mm_maskz_rcp14_sd, _mm_mask_rcp14_ss)
32336         (_mm_maskz_rcp14_ss): New intrinsics.
32337         * config/i386/i386-builtin.def (__builtin_ia32_rcp14sd_mask)
32338         (__builtin_ia32_rcp14ss_mask): New builtins.
32339         * config/i386/sse.md (srcp14<mode>_mask): New pattern.
32341 2017-05-10  Peter Bergner  <bergner@vnet.ibm.com>
32343         PR tree-optimization/51513
32344         * tree-cfg.c (gimple_seq_unreachable_p): New function.
32345         (assert_unreachable_fallthru_edge_p): Use it.
32346         (group_case_labels_stmt): Likewise.
32347         * tree-cfg.h: Prototype it.
32348         * stmt.c: Include cfghooks.h and tree-cfg.h.
32349         (emit_case_dispatch_table) <gap_label>: New local variable.
32350         Use it to fill dispatch table gaps.
32351         Test for default_label before updating probabilities.
32352         (expand_case) <default_label>: Remove unneeded initialization.
32353         Test for unreachable default case statement and remove its edge.
32354         Set default_label accordingly.
32355         * tree-ssa-ccp.c (optimize_unreachable): Update comment.
32357 2017-05-10  Carl Love  <cel@us.ibm.com>
32359         * config/rs6000/rs6000-c: Add support for built-in functions
32360         vector signed char      vec_neg (vector signed char)
32361         vector signed short int vec_neg (vector short int)
32362         vector signed int       vec_neg (vector signed int)
32363         vector signed long long vec_neg (vector signed long long)
32364         vector float            vec_neg (vector float)
32365         vector double           vec_neg (vector double)
32366         * config/rs6000/rs6000-builtin.def: Add definitions for NEG function
32367         overload.
32368         * config/rs6000/altivec.h: Add define for vec_neg
32369         * doc/extend.texi: Update the built-in documentation for the
32370         new built-in functions.
32372 2017-05-10  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
32374         PR tree-optimization/77644
32375         * match.pd (sqrt(x) cmp sqrt(y) -> x cmp y): New pattern.
32377 2017-05-10  Nathan Sidwell  <nathan@acm.org>
32379         * dumpfile.h (TDI_lang_all): New.
32380         (TDF_KIND): New. Renumber others
32381         (TDF_LANG, TDF_TREE, TDF_RTL, TDF_IPA): Enumerate value, rather
32382         than bits.
32383         * dumpfile.c (dump_files): Mark language dumps as TDF_LANG.  add
32384         lang-all.
32385         (get_dump_file_name): Adjust suffix generation.
32386         (dump_enable_all): Use TDF_KIND.
32387         * doc/invoke.texi (-fdump-lang-all): Document.
32389         * dumpfile.h: Tabify.
32391 2017-05-10  Wilco Dijkstra  <wdijkstr@arm.com>
32393         PR target/80671
32394         * config/aarch64/cortex-a57-fma-steering.c (merge_forest):
32395         Move member access before delete.
32397 2017-05-10  Alexandre Oliva  <aoliva@redhat.com>
32399         * tree-inline.c (expand_call_inline): Split block at stmt
32400         before the call.
32402 2017-05-09  Michael Meissner  <meissner@linux.vnet.ibm.com>
32404         PR target/68163
32405         * config/rs6000/rs6000.md (f32_lr): Delete mode attributes that
32406         are now unused after splitting mov{sf,sd}_hardfloat.
32407         (f32_lr2): Likewise.
32408         (f32_lm): Likewise.
32409         (f32_lm2): Likewise.
32410         (f32_li): Likewise.
32411         (f32_li2): Likewise.
32412         (f32_lv): Likewise.
32413         (f32_sr): Likewise.
32414         (f32_sr2): Likewise.
32415         (f32_sm): Likewise.
32416         (f32_sm2): Likewise.
32417         (f32_si): Likewise.
32418         (f32_si2): Likewise.
32419         (f32_sv): Likewise.
32420         (f32_dm): Likewise.
32421         (f32_vsx): Likewise.
32422         (f32_av): Likewise.
32423         (mov<mode>_hardfloat): Split into separate movsf and movsd pieces.
32424         For movsf, order stores so the VSX stores occur before the GPR
32425         store which encourages the register allocator to use a traditional
32426         FPR instead of a GPR.  For movsd, order the stores so that the GPR
32427         store comes before the VSX stores to allow the power6 to work.
32428         This is due to the power6 not having a 32-bit integer store
32429         instruction from a FPR.
32430         (movsf_hardfloat): Likewise.
32431         (movsd_hardfloat): Likewise.
32433 2017-05-09  Martin Sebor  <msebor@redhat.com>
32435         PR translation/80280
32436         * config/sol2-c.c (cmn_err_flag_specs): Initialize new data member
32437         added in r247778.
32439         PR translation/80280
32440         * config/i386/msformat-c.c (ms_printf_flag_specs): Initialize new
32441         data member added in r247778.
32442         (ms_scanf_flag_specs, ms_strftime_flag_specs): Same.
32444 2017-05-09  Nathan Sidwell  <nathan@acm.org>
32446         * tree.h (tree_fits_shwi_p, tree_fits_uhwi_p): Unconditionally pure.
32448         * ipa-devirt.c (default_hash_traits<type_pair>): Skip struct-scope
32449         typedefs.
32451 2017-05-09  Marek Polacek  <polacek@redhat.com>
32453         * doc/invoke.texi: Fix typo.
32455 2017-05-09  Richard Biener  <rguenther@suse.de>
32457         * tree-vrp.c (vrp_val_is_max): Adjust comment.
32458         (vrp_val_is_min): Likewise.
32459         (set_value_range_to_value): Likewise.
32460         (set_value_range_to_nonnegative): Likewise.
32461         (gimple_assign_nonzero_p): Likewise.
32462         (gimple_stmt_nonzero_p): Likewise.
32463         (vrp_int_const_binop): Likewise.  Remove unreachable case.
32464         (adjust_range_with_scev): Adjust comments.
32465         (compare_range_with_value): Likewise.
32466         (extract_range_from_phi_node): Likewise.
32467         (test_for_singularity): Likewise.
32469 2017-05-09  Richard Biener  <rguenther@suse.de>
32471         * tree-vrp.c (get_single_symbol): Add assert that we don't
32472         get overflowed constants as invariant part.
32473         (compare_values_warnv): Add comment before the TREE_NO_WARNING
32474         checks.  Use wi::cmp instead of recursing for integer constants.
32475         (compare_values): Just ignore whether we assumed undefined
32476         overflow instead of failing the compare.
32477         (extract_range_for_var_from_comparison_expr): Add comment before the
32478         TREE_NO_WARNING sets.
32479         (test_for_singularity): Likewise.
32480         (extract_range_from_comparison): Do not disable optimization
32481         when we assumed undefined overflow.
32482         (extract_range_basic): Remove init of unused var.
32484 2017-05-09  Richard Biener  <rguenther@suse.de>
32486         * tree-vrp.c (vrp_int_const_binop): Use wide-ints and simplify.
32487         (extract_range_from_multiplicative_op_1): Adjust.
32488         (extract_range_from_binary_expr_1): Use int_const_binop.
32490 2017-05-08  Kelvin Nilsen  <kelvin@gcc.gnu.org>
32492         PR target/80101
32493         * config/rs6000/power6.md: Replace store_data_bypass_p calls with
32494         rs6000_store_data_bypass_p in seven define_bypass directives and
32495         in several comments.
32496         * config/rs6000/rs6000-protos.h: Add prototype for
32497         rs6000_store_data_bypass_p function.
32498         * config/rs6000/rs6000.c (rs6000_store_data_bypass_p): New
32499         function implements slightly different (rs6000-specific) semantics
32500         than store_data_bypass_p, returning false rather than aborting
32501         with assertion error when arguments do not satisfy the
32502         requirements of store data bypass.
32503         (rs6000_adjust_cost): Replace six calls of store_data_bypass_p with
32504         rs6000_store_data_bypass_p.
32506 2017-05-08  Max Filippov  <jcmvbkbc@gmail.com>
32508         * config/xtensa/xtensa-protos.h
32509         (xtensa_initial_elimination_offset): New declaration.
32510         * config/xtensa/xtensa.c (xtensa_initial_elimination_offset):
32511         New function. Move its body from the INITIAL_ELIMINATION_OFFSET
32512         macro definition, add case for FRAME_POINTER_REGNUM when
32513         FRAME_GROWS_DOWNWARD.
32514         * config/xtensa/xtensa.h (FRAME_GROWS_DOWNWARD): New macro definition.
32515         (INITIAL_ELIMINATION_OFFSET): Replace body with call to
32516         xtensa_initial_elimination_offset.
32518 2017-05-08  Nathan Sidwell  <nathan@acm.org>
32520         * doc/invoke.texi: Alphabetize -fdump options.
32522 2017-05-08  Martin Sebor  <msebor@redhat.com>
32524         PR translation/80280
32525         * config/sol2-c.c (solaris_pragma_align): Correct quoting.
32527 2017-05-08  Bernd Edlinger  <bernd.edlinger@hotmail.de>
32529         * target.def (compute_frame_layout): New optional target hook.
32530         * doc/tm.texi.in (TARGET_COMPUTE_FRAME_LAYOUT): Add hook.
32531         * doc/tm.texi (TARGET_COMPUTE_FRAME_LAYOUT): Add documentation.
32532         * lra-eliminations.c (update_reg_eliminate): Call compute_frame_layout
32533         target hook.
32534         * reload1.c (verify_initial_elim_offsets): Likewise.
32535         * config/arm/arm.c (TARGET_COMPUTE_FRAME_LAYOUT): Define.
32536         (use_simple_return_p): Call arm_compute_frame_layout if needed.
32537         (arm_get_frame_offsets): Split up into this ...
32538         (arm_compute_frame_layout): ... and this function.
32540 2017-05-08  Richard Sandiford  <richard.sandiford@arm.com>
32542         * config/aarch64/constraints.md (Usa): New constraint.
32543         * config/aarch64/aarch64.md (*movsi_aarch64, *movdi_aarch64): Use it.
32545 2017-05-08  Thomas Preud'homme  <thomas.preudhomme@arm.com>
32547         * config.gcc (arm*-*-*): Set TM_MULTILIB_CONFIG from
32548         with_multilib_list after it has been checked.
32550 2017-05-08  Richard Biener  <rguenther@suse.de>
32552         * tree-ssa-pre.c (bitmap_set_and): Avoid bitmap copy.
32553         (bitmap_set_subtract_values): Likewise.
32555 2017-05-08  Richard Biener  <rguenther@suse.de>
32557         * tree-vrp.c (gimple_assign_nonzero_warnv_p): Rename to ...
32558         (gimple_assign_nonzero): ... this and remove strict_overflow_p
32559         argument.
32560         (gimple_stmt_nonzero_warnv_p): Rename to ...
32561         (gimple_stmt_nonzero_p): ... this and remove strict_overflow_p
32562         argument.
32563         (vrp_stmt_computes_nonzero): Remove strict_overflow_p argument.
32564         (extract_range_basic): Adjust, do not disable propagation on
32565         strict overflow sensitive simplification.
32566         (vrp_visit_cond_stmt): Likewise.
32568 2017-05-05  Jan Hubicka  <hubicka@ucw.cz>
32570         * ipa-inline-analysis.c (estimate_function_body_sizes): Recompute
32571         body size unconditionally.
32573 2017-05-07  Jeff Law  <law@redhat.com>
32575         Revert:
32576         2017-05-06  Jeff Law  <law@redhat.com>
32577         PR tree-optimization/78496
32578         * tree-vrp.c (simplify_assert_expr_using_ranges): Remove debugging
32579         code.
32581         PR tree-optimization/78496
32582         * tree-vrp.c (simplify_assert_expr_using_ranges): New function.
32583         (simplify_stmt_using_ranges): Call it.
32584         (vrp_dom_walker::before_dom_children): Extract equivalences
32585         from an ASSERT_EXPR with an equality comparison against a
32586         constant.
32588 2017-05-06  Jeff Law  <law@redhat.com>
32590         PR tree-optimization/78496
32591         * tree-vrp.c (simplify_assert_expr_using_ranges): Remove debugging
32592         code.
32594         PR tree-optimization/78496
32595         * tree-vrp.c (simplify_assert_expr_using_ranges): New function.
32596         (simplify_stmt_using_ranges): Call it.
32597         (vrp_dom_walker::before_dom_children): Extract equivalences
32598         from an ASSERT_EXPR with an equality comparison against a
32599         constant.
32601 2017-05-06  Richard Sandiford  <richard.sandiford@linaro.org>
32603         * lra-constraints.c (lra_copy_reg_equiv): New function.
32604         (split_reg): Use it to copy equivalence information from the
32605         original register to the spill register.
32607 2017-05-06  Richard Sandiford  <richard.sandiford@linaro.org>
32609         PR rtl-optimization/75964
32610         * simplify-rtx.c (simplify_const_relational_operation): Remove
32611         invalid handling of comparisons of integer ABS.
32613 2017-05-06  Uros Bizjak  <ubizjak@gmail.com>
32615         * config/i386/i386.c (ext_80387_constant_init): Do not explicitly
32616         initialize to zero.
32617         (init_regs): Remove declaration.
32618         (function_arg_advance_32): Initialize error_p as boolean variable.
32620 2017-05-05  Nathan Sidwell  <nathan@acm.org>
32622         * store-motion.c (remove_reachable_equiv_notes): Reformat long
32623         lines.  Use for (;;).
32625 2017-05-05  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
32627         * config/rs6000/rs6000.c (rs6000_vect_nonmem): New static var.
32628         (rs6000_init_cost): Initialize rs6000_vect_nonmem.
32629         (rs6000_add_stmt_cost): Update rs6000_vect_nonmem.
32630         (rs6000_finish_cost): Avoid vectorizing simple copy loops with
32631         VF=2 that require versioning.
32633 2017-05-05  David Malcolm  <dmalcolm@redhat.com>
32635         * diagnostic.h (CARET_LINE_MARGIN): Convert from macro to const
32636         int.
32638 2017-05-05  David Malcolm  <dmalcolm@redhat.com>
32640         * diagnostic.h (diagnostic_override_option_index): Convert from
32641         macro to inline function.
32643 2017-05-05  David Malcolm  <dmalcolm@redhat.com>
32645         * diagnostic.c (last_module_changed_p): New function.
32646         (set_last_module): New function.
32647         (diagnostic_report_current_module): Convert macro usage to
32648         the above functions.
32649         * diagnostic.h (diagnostic_context::last_module): Strengthen
32650         from const line_map * to const line_map_ordinary *.
32651         (diagnostic_last_module_changed): Delete macro.
32652         (diagnostic_set_last_module): Delete macro.
32654 2017-05-05  David Malcolm  <dmalcolm@redhat.com>
32656         * diagnostic.c (diagnostic_impl): Replace report_diagnostic
32657         with diagnostic_report_diagnostic.
32658         (diagnostic_n_impl_richloc): Likewise.
32659         * diagnostic.h (report_diagnostic): Delete macro.
32660         * rtl-error.c (diagnostic_for_asm): Replace report_diagnostic
32661         with diagnostic_report_diagnostic.
32662         * substring-locations.c (format_warning_va): Likewise.
32664 2017-05-05  David Malcolm  <dmalcolm@redhat.com>
32666         * diagnostic.c (diagnostic_report_diagnostic): Eliminate
32667         save/restor of format_spec.  Move option-printing code to...
32668         (print_option_information): ...this new function, and
32669         reimplement by simply printing to the pretty_printer,
32670         rather than appending to the format string.
32672 2017-05-05  David Malcolm  <dmalcolm@redhat.com>
32674         * diagnostic.c (diagnostic_report_diagnostic): Split out pragma
32675         handling logic into...
32676         (update_effective_level_from_pragmas): ...this new function.
32678 2017-05-04  Andrew Waterman  <andrew@sifive.com>
32680         * config/riscv/riscv.opt (mstrict-align): New option.
32681         * config/riscv/riscv.h (STRICT_ALIGNMENT): Use it.  Update comment.
32682         (SLOW_UNALIGNED_ACCESS): Define.
32683         (riscv_slow_unaligned_access): Declare.
32684         * config/riscv/riscv.c (riscv_tune_info): Add slow_unaligned_access
32685         field.
32686         (riscv_slow_unaligned_access): New variable.
32687         (rocket_tune_info): Set slow_unaligned_access to true.
32688         (optimize_size_tune_info): Set slow_unaligned_access to false.
32689         (riscv_cpu_info_table): Add entry for optimize_size_tune_info.
32690         (riscv_valid_lo_sum_p): Use TARGET_STRICT_ALIGN.
32691         (riscv_option_override): Set riscv_slow_unaligned_access.
32692         * doc/invoke.texi: Add -mstrict-align to RISC-V.
32694 2017-05-04  Kito Cheng  <kito.cheng@gmail.com>
32696         * config/riscv/riscv.md: Unify indentation.
32698 2017-05-05  Michael Meissner  <meissner@linux.vnet.ibm.com>
32700         PR target/79038
32701         PR target/79202
32702         PR target/79203
32703         * config/rs6000/rs6000.md (u code attribute): Add FIX and
32704         UNSIGNED_FIX.
32705         (extendsi<mode>2): Add support for doing sign extension via
32706         VUPKHSW and XXPERMDI if the value is in Altivec registers and we
32707         don't have ISA 3.0 instructions.
32708         (extendsi<mode>2 splitter): Likewise.
32709         (fix_trunc<mode>si2): If we are at ISA 2.07 (VSX small integer),
32710         generate the normal insns since SImode can now go in vector
32711         registers.  Disallow the special UNSPECs needed for previous
32712         machines to hide SImode being used.  Add new insns
32713         fctiw{,w}_<mode>_smallint if SImode can go in vector registers.
32714         (fix_trunc<mode>si2_stfiwx): Likewise.
32715         (fix_trunc<mode>si2_internal): Likewise.
32716         (fixuns_trunc<mode>si2): Likewise.
32717         (fixuns_trunc<mode>si2_stfiwx): Likewise.
32718         (fctiw<u>z_<mode>_smallint): Likewise.
32719         (fctiw<u>z_<mode>_mem): New combiner pattern to prevent conversion
32720         of floating point to 32-bit integer from doing a direct move to
32721         the GPR registers to do a store.
32722         (fctiwz_<mode>): Break long line.
32724 2017-05-05  Bin Cheng  <bin.cheng@arm.com>
32726         * Makefile.in (GTFILES): Add tree-ssa-loop-ivopts.c.
32727         * tree-ssa-loop-ivopts.c (compute_max_addr_offset): Delete.
32728         (addr_list, addr_offset_valid_p): New.
32729         (split_address_groups): Check offset validity with above function.
32730         (gt-tree-ssa-loop-ivopts.h): Include header file.
32732 2017-05-05  Nathan Sidwell  <nathan@acm.org>
32734         * config.gcc (arm*-*-*): Add missing 'fi'.
32736 2017-05-05  Steve Ellcey  <sellcey@cavium.com>
32738         * doc/invoke.texi (-fopt-info): Explicitly say order of options
32739         included in -fopt-info does not matter.
32740         * doc/optinfo.texi (-fopt-info): Fix description of default
32741         behavour. Explicitly say order of options included in -fopt-info
32742         does not matter.
32744 2017-05-05  Thomas Preud'homme  <thomas.preudhomme@arm.com>
32746         * config.gcc: Allow combinations of aprofile and rmprofile values for
32747         --with-multilib-list.
32748         * config/arm/t-multilib: New file.
32749         * config/arm/t-aprofile: Remove initialization of MULTILIB_*
32750         variables.  Remove setting of ISA and floating-point ABI in
32751         MULTILIB_OPTIONS and MULTILIB_DIRNAMES.  Set architecture and FPU in
32752         MULTI_ARCH_OPTS_A and MULTI_ARCH_DIRS_A rather than MULTILIB_OPTIONS
32753         and MULTILIB_DIRNAMES respectively.  Add comment to introduce all
32754         matches.  Add architecture matches for marvel-pj4 and generic-armv7-a
32755         CPU options.
32756         * config/arm/t-rmprofile: Likewise except for the matches changes.
32757         * doc/install.texi (--with-multilib-list): Document the combination of
32758         aprofile and rmprofile values and warn about pitfalls in doing that.
32760 2017-05-05  Wilco Dijkstra  <wdijkstr@arm.com>
32762         * config/aarch64/aarch64.md (movsi_aarch64): Remove '*' from r=w.
32763         (movdi_aarch64): Likewise.
32765 2017-05-05  Jakub Jelinek  <jakub@redhat.com>
32767         PR tree-optimization/80632
32768         * tree-switch-conversion.c (struct switch_conv_info): Add target_vop
32769         field.
32770         (build_arrays): Initialize it for virtual phis.
32771         (fix_phi_nodes): Use it for virtual phis.
32773         PR tree-optimization/80558
32774         * tree-vrp.c (extract_range_from_binary_expr_1): Optimize
32775         [x, y] op z into [x op, y op z] for op & or | if conditions
32776         are met.
32778 2017-05-05  Andre Vieira  <andre.simoesdiasvieira@arm.com>
32779             Prakhar Bahuguna  <prakhar.bahuguna@arm.com>
32781         PR target/71607
32782         * config/arm/arm.md (use_literal_pool): Remove.
32783         (64-bit immediate split): No longer takes cost into consideration
32784         if arm_disable_literal_pool is enabled.
32785         * config/arm/arm.c (arm_tls_referenced_p): Add diagnostic if TLS is
32786         used when arm_disable_literal_pool is enabled.
32787         (arm_max_const_double_inline_cost): Remove use of
32788         arm_disable_literal_pool.
32789         (push_minipool_fix): Add assert.
32790         (arm_reorg): Add return if arm_disable_literal_pool is enabled.
32791         * config/arm/vfp.md (no_literal_pool_df_immediate): New.
32792         (no_literal_pool_sf_immediate): New.
32794 2017-05-05  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
32796         PR tree-optimization/80613
32797         * tree-ssa-dce.c (propagate_necessity): Remove cases for
32798         BUILT_IN_STRDUP and BUILT_IN_STRNDUP.
32800 2017-05-05  Richard Biener  <rguenther@suse.de>
32802         * tree-ssa-pre.c (get_or_alloc_expr_for): Simplify.
32804 2017-05-05  Georg-Johann Lay  <avr@gjlay.de>
32806         * config/avr/avr.md [flag_strict_overflow]: Remove any occurence
32807         of this flag from insn conditions due to removal from r247495.
32809 2017-05-05  Wilco Dijkstra  <wdijkstr@arm.com>
32811         * config/arm/aarch-common.c (arm_early_load_addr_dep_ptr):
32812         New function.
32813         (arm_early_store_addr_dep_ptr): Likewise.
32814         * config/arm/aarch-common-protos.h
32815         (arm_early_load_addr_dep_ptr): Add prototype.
32816         (arm_early_store_addr_dep_ptr): Likewise.
32817         * config/arm/cortex-a53.md: Add new bypasses.
32819 2017-05-05  Jakub Jelinek  <jakub@redhat.com>
32821         * tree.c (next_type_uid): Change type to unsigned.
32822         (type_hash_canon): Decrement back next_type_uid if
32823         freeing a type node with the highest TYPE_UID.  For INTEGER_TYPEs
32824         also ggc_free TYPE_MIN_VALUE, TYPE_MAX_VALUE and TYPE_CACHED_VALUES
32825         if possible.
32827 2017-05-04  Martin Sebor  <msebor@redhat.com>
32829         * builtins.c: Fix a trivial typo in a comment.
32831         PR middle-end/79234
32832         * builtins.c (check_sizes): Adjust to handle reading past the end.
32833         Avoid printing excessive upper bound of ranges.  Use %E to print
32834         tree nodes instead of converting them to %wu.
32835         (expand_builtin_memchr): New function.
32836         (compute_dest_size): Rename...
32837         (compute_objsize): ...to this.
32838         (expand_builtin_memcpy): Adjust.
32839         (expand_builtin_mempcpy): Adjust.
32840         (expand_builtin_strcat): Adjust.
32841         (expand_builtin_strcpy): Adjust.
32842         (check_strncat_sizes): Adjust.
32843         (expand_builtin_strncat): Adjust.
32844         (expand_builtin_strncpy): Adjust and simplify.
32845         (expand_builtin_memset): Adjust.
32846         (expand_builtin_bzero): Adjust.
32847         (expand_builtin_memcmp): Adjust.
32848         (expand_builtin): Handle memcmp.
32849         (maybe_emit_chk_warning): Check strncat just once.
32851 2017-05-04  Martin Sebor  <msebor@redhat.com>
32853         PR preprocessor/79214
32854         PR middle-end/79222
32855         PR middle-end/79223
32856         * builtins.c (check_sizes): Add inlining context and issue
32857         warnings even when -Wno-system-headers is set.
32858         (check_strncat_sizes): Same.
32859         (expand_builtin_strncat): Same.
32860         (expand_builtin_memmove): New function.
32861         (expand_builtin_stpncpy): Same.
32862         (expand_builtin): Handle memmove and stpncpy.
32864 2017-05-04  Bin Cheng  <bin.cheng@arm.com>
32866         * tree-ssa-loop-ivopts.c (struct cost_pair): Remove field inv_expr
32867         which is not used any more.
32869 2017-05-04  Wilco Dijkstra  <wdijkstr@arm.com>
32871         * config/aarch64/aarch64.c (generic_tunings): Update prefetch model.
32873 2017-05-04  Wilco Dijkstra  <wdijkstr@arm.com>
32875         * config/aarch64/aarch64.c (cortexa35_tunings): Set jump alignment to 4.
32876         (cortexa53_tunings): Likewise.
32877         (cortexa57_tunings): Likewise.
32878         (cortexa72_tunings): Likewise.
32879         (cortexa73_tunings): Likewise.
32881 2017-05-04  Wilco Dijkstra  <wdijkstr@arm.com>
32883         * config/aarch64/aarch64.c (generic_tunings): Set jump alignment to 4.
32884         Set loop alignment to 8.
32886 2017-05-04  Martin Sebor  <msebor@redhat.com>
32888         PR translation/80280
32889         * builtins.c (expand_builtin_object_size): Add missing quoting to
32890         %D and like directives.
32891         * hsa-gen.c (hsa_type_for_scalar_tree_type): Same.
32892         (hsa_type_for_tree_type): Same.
32893         (verify_function_arguments): Same.
32894         * symtab.c (symbol_table::change_decl_assembler_name): Same.
32895         * varasm.c (get_section): Same.
32896         (mark_weak): Same.
32898 2017-05-04  Martin Sebor  <msebor@redhat.com>
32900         PR translation/80280
32901         * config/i386/i386.c (ix86_function_versions): Quote a %D directive.
32903 2017-05-04  Wilco Dijkstra  <wdijkstr@arm.com>
32905         * config/aarch64/aarch64.c (generic_addrcost_table):
32906         Change HI/TI mode setting.
32908 2017-05-04  Martin Jambor  <mjambor@suse.cz>
32910         PR tree-optimization/80622
32911         * tree-sra.c (comes_initialized_p): New function.
32912         (build_accesses_from_assign): Only set write lazily when
32913         comes_initialized_p is false.
32914         (analyze_access_subtree): Use comes_initialized_p.
32915         (propagate_subaccesses_across_link): Assert !comes_initialized_p
32916         instead of testing for PARM_DECL.
32918 2017-05-04  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
32920         * config/aarch64/aarch64.md (prefetch); Adjust predicate and
32921         constraint on operand 0 to allow more general addressing modes.
32922         Adjust output template.
32923         * config/aarch64/aarch64.c (aarch64_address_valid_for_prefetch_p):
32924         New function.
32925         * config/aarch64/aarch64-protos.h
32926         (aarch64_address_valid_for_prefetch_p): Declare prototype.
32927         * config/aarch64/constraints.md (Dp): New address constraint.
32928         * config/aarch64/predicates.md (aarch64_prefetch_operand): New
32929         predicate.
32931 2017-05-04  Jan Hubicka  <hubicka@ucw.cz>
32933         * ipa-cp.c (perform_estimation_of_a_value): Drop base_time parameter;
32934         update use of estimate_ipcp_clone_size_and_time.
32935         (estimate_local_effects): Update use of
32936         estimate_ipcp_clone_size_and_time and perform_estimation_of_a_value.
32937         * ipa-inline.h (estimate_ipcp_clone_size_and_time): Update prototype.
32938         * ipa-inline-analysis.c (estimate_ipcp_clone_size_and_time):
32939         Return nonspecialized time.
32941 2017-05-04  Richard Biener  <rguenther@suse.de>
32943         * tree-ssa-alias.c (get_continuation_for_phi): Improve looking
32944         for the last VUSE which def dominates the PHI.  Directly call
32945         maybe_skip_until.
32946         (get_continuation_for_phi_1): Remove.
32948 2017-05-04  Richard Sandiford  <richard.sandiford@linaro.org>
32950         * tree-ssa-loop-manip.c (niter_for_unrolled_loop): Add commentary
32951         to explain the use of truncating division.  Cap the number of
32952         iterations to the maximum given by nb_iterations_upper_bound,
32953         if defined.
32955 2017-05-04  Thomas Preud'homme  <thomas.preudhomme@arm.com>
32957         * configure.ac (--enable-mingw-wildcard): Add new configurable feature.
32958         * configure: Regenerate.
32959         * config.in: Regenerate.
32960         * config/i386/driver-mingw32.c: new file.
32961         * config/i386/x-mingw32: Add rule to build driver-mingw32.o.
32962         * config.host: Link driver-mingw32.o on MinGW host.
32963         * doc/install.texi: Document new --enable-mingw-wildcard configure
32964         option.
32966 2017-05-04  Marek Polacek  <polacek@redhat.com>
32968         PR tree-optimization/80612
32969         * calls.c (get_size_range): Check for INTEGRAL_TYPE_P.
32971 2017-05-04  Prakhar Bahuguna  <prakhar.bahuguna@arm.com>
32972             Andre Simoes Dias Vieira  <andre.simoesdiasvieira@arm.com>
32974         * config/arm/arm.md (movsi): Change TARGET_32BIT to TARGET_HAVE_MOVT.
32975         (movt splitter): Likewise.
32976         * config/arm/arm.c (arm_option_check_internal): Change arm_arch_thumb2
32977         to TARGET_HAVE_MOVT, and merge with -mslow-flash-data check.
32978         (const_ok_for_arm): Change else to else if (TARGET_THUMB2) and add else
32979         block for Thumb-1 with MOVT.
32980         (thumb2_legitimate_address_p): Move code block ...
32981         (can_avoid_literal_pool_for_label_p): ... into this new function.
32982         (thumb1_legitimate_address_p): Add check for TARGET_HAVE_MOVT and
32983         literal pool.
32984         (thumb_legitimate_constant_p): Add conditional on TARGET_HAVE_MOVT
32985         * doc/invoke.texi (-mpure-code): Change "ARMv7-M targets" for
32986         "M-profile targets with the MOVT instruction".
32988 2017-05-04  Prakhar Bahuguna  <prakhar.bahuguna@arm.com>
32990         * config/arm/arm-builtins.c (arm_init_builtins): Rename
32991         __builtin_arm_ldfscr to __builtin_arm_get_fpscr, and rename
32992         __builtin_arm_stfscr to __builtin_arm_set_fpscr.
32994 2017-05-04  Martin Liska  <mliska@suse.cz>
32996         * tree-vrp.c (simplify_cond_using_ranges_2): Remove unused
32997         variable cond_code.
32999 2017-05-04  Richard Biener  <rguenther@suse.de>
33001         * tree.c (array_at_struct_end_p): Handle arrays at struct
33002         end with flexarrays more conservatively.  Refactor and treat
33003         arrays of arrays or aggregates more strict.  Fix
33004         VIEW_CONVERT_EXPR handling.  Remove allow_compref argument.
33005         * tree.c (array_at_struct_end_p): Adjust prototype.
33006         * emit-rtl.c (set_mem_attributes_minus_bitpos): Adjust.
33007         * gimple-fold.c (get_range_strlen): Likewise.
33008         * tree-chkp.c (chkp_may_narrow_to_field): Likewise.
33010 2017-05-04  Richard Biener  <rguenther@suse.de>
33012         PR tree-optimization/31130
33013         * tree-vrp.c (needs_overflow_infinity): Remove as always returning
33014         false.
33015         (supports_overflow_infinity): Likewise.
33016         (is_negative_overflow_infinity): Likewise.
33017         (is_positive_overflow_infinity): Likewise.
33018         (is_overflow_infinity): Likewise.
33019         (stmt_overflow_infinity): Likewise.
33020         (overflow_infinity_range_p): Likewise.
33021         (usable_range_p): Remove as always returning true.
33022         (make_overflow_infinity): Remove.
33023         (negative_overflow_infinity): Likewise.
33024         (positive_overflow_infinity): Likewise.
33025         (avoid_overflow_infinity): Likewise.
33026         (set_value_range): Adjust accordingly.
33027         (set_value_range_to_nonnegative): Likewise, remove now unused
33028         overflow_infinity arg.
33029         (vrp_operand_equal_p): Adjust.
33030         (update_value_range): Likewise.
33031         (range_int_cst_singleton_p): Likewise.
33032         (operand_less_p): Likewise.
33033         (compare_values_warnv): Likewise.
33034         (extract_range_for_var_from_comparison_expr): Likewise.
33035         (vrp_int_const_binop): Likewise.
33036         (zero_nonzero_bits_from_vr): Likewise.
33037         (extract_range_from_multiplicative_op_1): Likewise.
33038         (extract_range_from_binary_expr_1): Likewise.
33039         (extract_range_from_unary_expr): Likewise.
33040         (extract_range_from_comparison): Likewise.
33041         (extract_range_basic): Likewise.
33042         (adjust_range_with_scev): Likewise.
33043         (compare_ranges): Likewise.
33044         (compare_range_with_value): Likewise.
33045         (dump_value_range): Likewise.
33046         (test_for_singularity): Likewise, remove strict_overflow_p parameter
33047         never used.
33048         (simplify_cond_using_ranges): Adjust.
33050 2017-05-04  Pekka Jääskeläinen  <pekka.jaaskelainen@parmance.com>
33052         * brig-builtins.def: Added a builtin for class_f64.
33053         * builtin-types.def: Added a builtin type needed by class_f64.
33055 2017-05-03  Jason Merrill  <jason@redhat.com>
33057         * timevar.def: Add TV_CONSTEXPR.
33059 2017-05-03  David Malcolm  <dmalcolm@redhat.com>
33061         * common.opt (fdiagnostics-parseable-fixits): Fix typo.
33063 2017-05-03  Martin Jambor  <mjambor@suse.cz>
33065         * ipa-prop.c (ipa_update_after_lto_read): Removed.
33066         * ipa-prop.h (ipa_update_after_lto_read): Remove declaration.
33067         * ipa-cp.c (ipcp_propagate_stage): Do not call
33068         ipa_update_after_lto_read.
33069         * ipa-inline.c (ipa_inline): Likewise.
33071 2017-05-03  Martin Jambor  <mjambor@suse.cz>
33073         * ipa-prop.h (ipa_edge_args): Make a class.  Mark with for_user GTY
33074         tag.  Added a default constructor and a destructor.
33075         (ipa_edge_args_sum_t): New class;
33076         (ipa_edge_args_sum): Declare.
33077         (ipa_edge_args_vector): Remove declaration.
33078         (IPA_EDGE_REF): Use ipa_edge_args_sum.
33079         (ipa_free_edge_args_substructures): Remove declaration.
33080         (ipa_check_create_edge_args): Use ipa_edge_args_sum.
33081         (ipa_edge_args_info_available_for_edge_p): Likewise.
33082         * ipa-prop.c (ipa_edge_args_vector): Removed.
33083         (edge_removal_hook_holder): Likewise.
33084         (edge_duplication_hook_holder): Likewise.
33085         (ipa_edge_args_sum): New variable.
33086         (ipa_propagate_indirect_call_infos): Test ipa_edge_args_sum instead of
33087         ipa_edge_args_vector.
33088         (ipa_free_edge_args_substructures): Likewise.
33089         (ipa_free_all_edge_args): Free ipa_edge_args_sum instead of
33090         ipa_edge_args_vector.
33091         (ipa_edge_removal_hook): Turned into method
33092         ipa_edge_args_sum_t::remove.
33093         (ipa_edge_duplication_hook): Turned into method
33094         ipa_edge_args_sum_t::duplicate.
33095         (ipa_register_cgraph_hooks): Create ipa_edge_args_sum instead of
33096         registering edge hooks.
33097         (ipa_unregister_cgraph_hooks): Do not unregister edge hooks.
33098         * ipa-inline-analysis.c (estimate_function_body_sizes): Test
33099         ipa_edge_args_sum instead of ipa_edge_args_vector.
33100         * ipa-profile.c (ipa_profile): Likewise.
33102 2017-05-03  Martin Jambor  <mjambor@suse.cz>
33104         * symbol-summary.h (function_summary): New method exists.
33105         (function_summary::symtab_removal): Deallocate through release.
33106         (call_summary): New class.
33107         (gt_ggc_mx): New overload.
33108         (gt_pch_nx): Likewise.
33109         (gt_pch_nx): Likewise.
33111 2017-05-03  Jeff Law  <law@redhat.com>
33113         PR tree-optimization/78496
33114         * tree-vrp.c (simplify_cond_using_ranges_1): Renamed
33115         from simplify_cond_using_ranges.  Split off code to walk
33116         backwards through casts into ...
33117         (simplify_cond_using_ranges_2): New function.
33118         (simplify_stmt_using_ranges): Call simplify_cond_using_ranges_1.
33119         (execute_vrp): After identifying jump threads, call
33120         simplify_cond_using_ranges_2.
33122 2017-05-03  Jan Hubicka  <hubicka@ucw.cz>
33124         PR bootstrap/80609
33125         * ipa-inline.h (inline_summary): Add ctor.
33126         (create_ggc): Do not use ggc_cleared_alloc.
33128 2017-05-03  Jeff Downs  <heydowns@somuchpressure.net>
33129             Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
33131         * gcc.c (handle_braces): Support escaping in switch matching
33132         text.
33133         * doc/invoke.texi (Spec Files): Document it.
33134         Remove superfluous @code markup in items.
33136 2017-05-03  David Malcolm  <dmalcolm@redhat.com>
33138         * diagnostic-show-locus.c (struct column_range): New struct.
33139         (get_affected_columns): New function.
33140         (get_printed_columns): New function.
33141         (struct correction): New struct.
33142         (correction::ensure_capacity): New function.
33143         (correction::ensure_terminated): New function.
33144         (struct line_corrections): New struct.
33145         (line_corrections::~line_corrections): New dtor.
33146         (line_corrections::add_hint): New function.
33147         (layout::print_trailing_fixits): Reimplement in terms of the new
33148         classes.
33149         (selftest::test_overlapped_fixit_printing): New function.
33150         (selftest::diagnostic_show_locus_c_tests): Call it.
33152 2017-05-03  Nathan Sidwell  <nathan@acm.org>
33154         Canonicalize canonical type hashing
33155         * tree.h (type_hash_canon_hash): Declare.
33156         * tree.c (type_hash_list, attribute_hash_list): Move into
33157         type_hash_canon_hash.
33158         (build_type_attribute_qual_variant): Break out hash code calc into
33159         type_hash_canon_hash.
33160         (type_hash_canon_hash): New.  Generic type hash computation.
33161         (build_range_type_1, build_array_type_1, build_function_type,
33162         build_method_type_directly, build_offset_type, build_complex_type,
33163         make_vector_type): Call it.
33165 2017-05-03  Richard Biener  <rguenther@suse.de>
33167         * tree-vect-data-refs.c (vect_enhance_data_refs_alignment):
33168         When all DRs have unknown misaligned do not always peel
33169         when there is a store but apply the same costing model as if
33170         there were only loads.
33172 2017-05-03  Richard Biener  <rguenther@suse.de>
33174         Revert
33175         PR tree-optimization/80492
33176         * tree-ssa-alias.c (decl_refs_may_alias_p): Handle
33177         compare_base_decls returning dont-know properly.
33179 2017-05-03  Thomas Preud'homme  <thomas.preudhomme@arm.com>
33181         * config/arm/iterators.md (CCSI): New mode iterator.
33182         (arch): New mode attribute.
33183         * config/arm/sync.md (atomic_compare_and_swap<mode>_1): Rename into ...
33184         (atomic_compare_and_swap<CCSI:arch><NARROW:mode>_1): This and ...
33185         (atomic_compare_and_swap<CCSI:arch><SIDI:mode>_1): This.  Use CCSI
33186         code iterator for success result mode.
33187         * config/arm/arm.c (arm_expand_compare_and_swap): Adapt code to use
33188         the corresponding new insn generators.
33190 2017-05-03  Bin Cheng  <bin.cheng@arm.com>
33192         Revert r247509
33193         2017-05-02  Bin Cheng  <bin.cheng@arm.com>
33194         * rtlanal.c (rtx_cost): Handle TRUNCATE between tieable modes.
33196 2017-05-03  Richard Sandiford  <richard.sandiford@linaro.org>
33198         * tree-data-ref.h (SUB_CONFLICTS_IN_A): Wrap SUB argument in brackets.
33199         (SUB_CONFLICTS_IN_B, SUB_LAST_CONFLICT, SUB_DISTANCE): Likewise.
33200         (DDR_A): Wrap DDR argument in brackets.
33201         (DDR_B, DDR_AFFINE_P, DDR_ARE_DEPENDENT, DDR_SUBSCRIPTS): Likewise.
33202         (DDR_LOOP_NEST, DDR_INNER_LOOP, DDR_SELF_REFERENCE): Likewise.
33203         (DDR_REVERSED_P): Likewise.
33205 2017-05-03  Jakub Jelinek  <jakub@redhat.com>
33207         PR tree-optimization/79472
33208         * tree-switch-conversion.c (struct switch_conv_info): Add
33209         contiguous_range and default_case_nonstandard fields.
33210         (collect_switch_conv_info): Compute contiguous_range and
33211         default_case_nonstandard fields, don't clear final_bb if
33212         contiguous_range and only the default case doesn't have the required
33213         structure.
33214         (check_all_empty_except_final): Set default_case_nonstandard instead
33215         of failing if contiguous_range and the default case doesn't have empty
33216         block.
33217         (check_final_bb): Add SWTCH argument, don't fail if contiguous_range
33218         and only the default case doesn't have the required constants.  Skip
33219         virtual phis.
33220         (gather_default_values): Skip virtual phis.  Allow non-NULL CASE_LOW
33221         if default_case_nonstandard.
33222         (build_constructors): Build constant 1 just once.  Assert that default
33223         values aren't inserted in between cases if contiguous_range.  Skip
33224         virtual phis.
33225         (build_arrays): Skip virtual phis.
33226         (prune_bbs): Add DEFAULT_BB argument, don't remove that bb.
33227         (fix_phi_nodes): Don't add e2f phi arg if default_case_nonstandard.
33228         Handle virtual phis.
33229         (gen_inbound_check): Handle default_case_nonstandard case.
33230         (process_switch): Adjust check_final_bb caller.  Call
33231         gather_default_values with the first non-default case instead of
33232         default case if default_case_nonstandard.
33234 2017-05-02  Nathan Sidwell  <nathan@acm.org>
33236         * ggc-page.c (move_ptes_to_front): Replace unsigned >0 with i--
33237         check.  Fix formatting.
33239 2017-05-02  Jan Hubicka  <hubicka@ucw.cz>
33241         * ipa-inline-analysis.c (estimate_node_size_and_time): Allow roundoff
33242         errors when comparing specialized and unspecialized times.
33244 2017-05-02  David Malcolm  <dmalcolm@redhat.com>
33246         * diagnostic-show-locus.c
33247         (layout::should_print_annotation_line_p): Make private.
33248         (layout::print_annotation_line): Make private.
33249         (layout::annotation_line_showed_range_p): Make private.
33250         (layout::show_ruler): Make private.
33251         (layout::print_source_line): Make private.  Pass in line and
33252         line_width, rather than calling location_get_source_line.  Drop
33253         returned value.
33254         (layout::print_leading_fixits): New method.
33255         (layout::print_any_fixits): Rename to...
33256         (layout::print_trailing_fixits): ...this, and make private.
33257         Don't print newline fixits.
33258         (diagnostic_show_locus): Move logic for printing one row into...
33259         (layout::print_line): ...this new function.  Move the
33260         location_get_source_line call and error-handling from
33261         print_source_line to here.  Call print_leading_fixits, and rename
33262         print_any_fixits to print_trailing_fixits.
33263         (selftest::test_fixit_insert_containing_newline): Update now that
33264         newlines are partially supported.
33265         (selftest::test_fixit_insert_containing_newline_2): New test.
33266         (selftest::test_fixit_replace_containing_newline): Update comments.
33267         (selftest::diagnostic_show_locus_c_tests): Call the new test.
33268         * edit-context.c (class added_line): New class.
33269         (class edited_line): Describe newline handling in comment.
33270         (edited_line::actually_edited_p): New method.
33271         (edited_line::print_content): Delete redundant decl.
33272         (edited_line::m_predecessors): New field.
33273         (edited_file::print_content): Call edited_line::print_content.
33274         (edited_file::print_diff): Update to support newlines.
33275         (edited_file::print_diff_hunk): Likewise.
33276         (edited_file::print_run_of_changed_lines): New function.
33277         (edited_file::print_diff_line): Convert to...
33278         (print_diff_line): ...this.
33279         (edited_file::get_effective_line_count): New function.
33280         (edited_line::edited_line): Initialize new field m_predecessors.
33281         (edited_line::~edited_line): Clean up m_predecessors.
33282         (edited_line::apply_fixit): Handle newlines.
33283         (edited_line::get_effective_line_count): New function.
33284         (edited_line::print_content): New function.
33285         (edited_line::print_diff_lines): New function.
33286         (selftest::test_applying_fixits_insert_containing_newline): New
33287         test.
33288         (selftest::test_applying_fixits_replace_containing_newline): New
33289         test.
33290         (selftest::insert_line): New function.
33291         (selftest::test_applying_fixits_multiple_lines): Add example of
33292         inserting a line.
33293         (selftest::edit_context_c_tests): Call the new tests.
33295 2017-05-02  Bin Cheng  <bin.cheng@arm.com>
33297         * tree-ssa-loop-ivopts.c (get_scaled_computation_cost_at): Delete
33298         parameter cand.  Update dump information.
33299         (get_computation_cost): Update uses.
33301 2017-05-02  Bin Cheng  <bin.cheng@arm.com>
33303         * tree-ssa-loop-ivopts.c (get_computation_aff_1): New.
33304         (get_computation_aff): Reorder parameters.  Use get_computation_aff_1.
33305         (get_computation_at, rewrite_use_address): Update use of
33306         get_computation_aff.
33308 2017-05-02  Bin Cheng  <bin.cheng@arm.com>
33310         * tree-ssa-loop-ivopts.c (get_computation_at): Reorder parameters.
33311         (get_computation): Delete.
33312         (get_computation_cost): Implement like get_computation_cost_at.
33313         Use get_computation_at.
33314         (get_computation_cost_at): Delete.
33315         (rewrite_use_nonlinear_expr): Use get_computation_at.
33316         (rewrite_use_compare, remove_unused_ivs): Ditto.
33318 2017-05-02  Bin Cheng  <bin.cheng@arm.com>
33320         * tree-ssa-loop-ivopts.c (rewrite_use_address): Simple refactor.
33322 2017-05-02  Bin Cheng  <bin.cheng@arm.com>
33324         * tree-ssa-loop-ivopts.c (struct iv_ca): Rename n_regs to n_invs.
33325         (ivopts_global_cost_for_size): Rename parameter and update uses.
33326         (iv_ca_recount_cost): Update uses.
33327         (iv_ca_set_remove_invs, iv_ca_set_no_cp): Record invariants and
33328         candidates seperately in n_invs and n_cands.
33329         (iv_ca_set_add_invs, iv_ca_set_cp, iv_ca_new): Ditto.
33331 2017-05-02  Bin Cheng  <bin.cheng@arm.com>
33333         * tree-ssa-loop-ivopts.c (struct walk_tree_data): New.
33334         (find_inv_vars_cb): New.
33335         (find_depends): Renamed to ...
33336         (find_inv_vars): ... this.
33337         (add_candidate_1, force_var_cost): Call find_inv_vars.
33338         (split_address_cost, determine_group_iv_cost_cond): Ditto.
33340 2017-05-02  Bin Cheng  <bin.cheng@arm.com>
33342         * tree-ssa-loop-ivopts.c (struct cost_pair): Rename depends_on to
33343         inv_vars.  Add inv_exprs.
33344         (struct iv_cand): Rename depends_on to inv_vars.
33345         (struct ivopts_data): Rename max_inv_id/n_invariant_uses to
33346         max_inv_var_id/n_inv_var_uses.  Move max_inv_expr_id around.
33347         Refactor field used_inv_exprs from has_map to array n_inv_expr_uses.
33348         (dump_cand): Dump inv_vars.
33349         (tree_ssa_iv_optimize_init): Support inv_vars and inv_exprs.
33350         (record_invariant, find_depends, add_candidate_1): Ditto.
33351         (set_group_iv_cost, force_var_cost): Ditto.
33352         (split_address_cost, ptr_difference_cost, difference_cost): Ditto.
33353         (get_computation_cost_at, get_computation_cost): Ditto.
33354         (determine_group_iv_cost_generic): Ditto.
33355         (determine_group_iv_cost_address): Ditto.
33356         (determine_group_iv_cost_cond, autoinc_possible_for_pair): Ditto.
33357         (determine_group_iv_costs): Ditto.
33358         (iv_ca_recount_cost): Update call to ivopts_global_cost_for_size.
33359         (iv_ca_set_remove_invariants): Renamed to ...
33360         (iv_ca_set_remove_invs): ... this.  Support inv_vars and inv_exprs.
33361         (iv_ca_set_no_cp): Use iv_ca_set_remove_invs.
33362         (iv_ca_set_add_invariants):  Renamed to ...
33363         (iv_ca_set_add_invs): ... this.  Support inv_vars and inv_exprs.
33364         (iv_ca_set_cp): Use iv_ca_set_add_invs.
33365         (iv_ca_has_deps): Support inv_vars and inv_exprs.
33366         (iv_ca_new, iv_ca_free, iv_ca_dump, free_loop_data): Ditto.
33367         (create_new_ivs): Remove useless dump.
33369 2017-05-02  Bin Cheng  <bin.cheng@arm.com>
33371         * tree-ssa-loop-ivopts.c (get_computation_cost_at): Remove pseudo
33372         iv_cand code.
33373         (determine_group_iv_cost_cond, determine_iv_cost): Ditto.
33374         (iv_ca_set_no_cp, create_new_iv): Ditto.
33376 2017-05-02  Bin Cheng  <bin.cheng@arm.com>
33378         * rtlanal.c (rtx_cost): Handle TRUNCATE between tieable modes.
33380 2017-05-02  Xi Ruoyao  <ryxi@stu.xidian.edu.cn>
33382         * tree.h (EXPR_CILK_SPAWN): Use macro TREE_CHECK2 instead of
33383         function tree_check2.
33385 2017-05-02  Martin Liska  <mliska@suse.cz>
33387         * doc/gcov.texi: Add missing preposition.
33388         * gcov.c (function_info::function_info): Properly fill up
33389         all member variables.
33391 2017-05-02  Tamar Christina  <tamar.christina@arm.com>
33393         * expr.c (expand_expr_real_2): Re-cost if previous costs are the same.
33395 2017-05-02  Tamar Christina  <tamar.christina@arm.com>
33397         * simplify-rtx.c (simplify_binary_operation_1): Add LSHIFTRT case.
33399 2017-05-02  Martin Liska  <mliska@suse.cz>
33401         PR lto/77954.
33402         * lto-streamer-in.c (lto_read_tree_1): Remove
33403         LTO_STREAMER_DEBUG.
33404         * lto-streamer.c (struct tree_hash_entry): Likewise.
33405         (struct tree_entry_hasher): Likewise.
33406         (tree_entry_hasher::hash): Likewise.
33407         (tree_entry_hasher::equal): Likewise.
33408         (lto_streamer_init): Likewise.
33409         (lto_orig_address_map): Likewise.
33410         (lto_orig_address_get): Likewise.
33411         (lto_orig_address_remove): Likewise.
33412         * lto-streamer.h: Likewise.
33413         * tree-streamer-in.c (streamer_alloc_tree): Likewise.
33414         * tree-streamer-out.c (streamer_write_tree_header): Likewise.
33416 2017-05-02  Sebastian Peryt  <sebastian.peryt@intel.com>
33418         * config/i386/avx512fintrin.h (_mm_mask_add_round_sd)
33419         (_mm_maskz_add_round_sd, _mm_mask_add_round_ss)
33420         (mm_maskz_add_round_ss, _mm_mask_sub_round_sd)
33421         (mm_maskz_sub_round_sd, _mm_mask_sub_round_ss)
33422         (mm_maskz_sub_round_ss, _mm_mask_add_sd)
33423         (mm_maskz_add_sd, _mm_mask_add_ss, _mm_maskz_add_ss)
33424         (mm_mask_sub_sd, _mm_maskz_sub_sd, _mm_mask_sub_ss)
33425         (mm_maskz_sub_ss): New intrinsics.
33426         * config/i386/i386-builtin-types.def (V2DF_FTYPE_V2DF_V2DF_V2DF_UQI_INT)
33427         (V4SF_FTYPE_V4SF_V4SF_V4SF_UQI_INT): New function type aliases.
33428         * config/i386/i386-builtin.def (__builtin_ia32_addsd_mask_round)
33429         (__builtin_ia32_addss_mask_round, __builtin_ia32_subsd_mask_round)
33430         (__builtin_ia32_subss_mask_round): New builtins.
33431         * config/i386/i386.c (V2DF_FTYPE_V2DF_V2DF_V2DF_UQI_INT)
33432         (V4SF_FTYPE_V4SF_V4SF_V4SF_UQI_INT): Handle new types.
33433         * config/i386/sse.md (<sse>_vm<plusminus_insn><mode>3<round_name>):
33434         Renamed to ...
33435         (<sse>_vm<plusminus_insn><mode>3<mask_name><round_name>): ... this.
33436         (v<plusminus_mnemonic><ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|%0, %1, %<iptr>2<round_op3>}):
33437         Changed to ...
33438         (v<plusminus_mnemonic><ssescalarmodesuffix>\t{<round_mask_op3>%2, %1, %0<mask_operand3>|%0<mask_operand3>, %1, %<iptr>2<round_mask_op3>}):
33439         ... this.
33441 2017-05-02  Martin Jambor  <mjambor@suse.cz>
33443         PR tree-optimization/78687
33444         * tree-sra.c (access): New field parent.
33445         (process_subtree_disqualification): New function.
33446         (disqualify_candidate): Call it.
33447         (build_accesses_from_assign): Reset write flag if creating an
33448         assighnment link.
33449         (build_access_subtree): Fill in parent field and also prpagate
33450         down grp_write flag.
33451         (create_artificial_child_access): New parameter set_grp_write, set
33452         grp_write to its value.
33453         (propagate_subaccesses_across_link): Also propagate grp_write flag
33454         values.
33455         (propagate_all_subaccesses): Push the closest parent back to work
33456         queue if add_access_to_work_queue returned true.
33458 2017-05-02  Richard Biener  <rguenther@suse.de>
33460         * common.opt (fstrict-overflow): Alias negative to fwrapv.
33461         * doc/invoke.texi (fstrict-overflow): Remove all traces of
33462         -fstrict-overflow documentation.
33463         * tree.h (TYPE_OVERFLOW_UNDEFINED): Do not test flag_strict_overflow.
33464         (POINTER_TYPE_OVERFLOW_UNDEFINED): Test !flag_wrapv instead of
33465         flag_strict_overflow.
33466         * ipa-inline.c (can_inline_edge_p): Do not test flag_strict_overflow.
33467         * lto-opts.c (lto_write_options): Do not stream it.
33468         * lto-wrapper.c (merge_and_complain): Do not handle it.
33469         * opts.c (default_options_table): Do not set -fstrict-overflow.
33470         (finish_options): Likewise do not clear it when sanitizing.
33471         * simplify-rtx.c (simplify_const_relational_operation): Do not
33472         test flag_strict_overflow.
33474 2017-05-02  Uros Bizjak  <ubizjak@gmail.com>
33476         * config/alpha/alpha.md (*add<mode>3_ieee): Merge to add<mode>3
33477         using enabled attribute.
33478         (*sub<mode>3_ieee): Merge to sub<mode>3 using enabled attribute.
33479         (*mul<mode>3_ieee): Merge to mul<mode>3 using enabled attribute.
33480         (*div<mode>3_ieee): Merge to div<mode>3 using enabled attribute.
33481         (*sqrt<mode>2_ieee): Merge to sqrt<mode>2 using enabled attribute.
33482         (*fix_truncdfdi_ieee): Merge to *fix_truncdfdi2 using enabled attribute.
33483         (*fix_truncsfdi_ieee): Merge to *fix_truncsfdi2 using enabled attribute.
33484         (*floatdisf_ieee): Merge to floatdisf2 using enabled attribute.
33485         (*floatdidf_ieee): Merge to floatdidf2 using enabled attribute.
33486         (*truncdfsf2_ieee): Merge to truncdfsf2 using enabled attribute.
33487         (*cmpdf_ieee): Merge to *cmpdf_internal using enabled attribute.
33489 2017-05-02  Uros Bizjak  <ubizjak@gmail.com>
33491         * config/i386/i386.c (ix86_code_end): Use {FIRST,LAST}_INT_REG.
33493 2017-05-02  Richard Biener  <rguenther@suse.de>
33495         PR tree-optimization/80591
33496         Revert
33497         2017-04-10  Richard Biener  <rguenther@suse.de>
33499         * tree-ssa-structalias.c (find_func_aliases): Properly handle
33500         asm inputs.
33502 2017-05-02  Richard Biener  <rguenther@suse.de>
33504         PR tree-optimization/80549
33505         * tree-cfgcleanup.c (mfb_keep_latches): New helper.
33506         (cleanup_tree_cfg_noloop): Create forwarders to known loop
33507         headers if they do not have a preheader.
33509 2017-05-02  Martin Liska  <mliska@suse.cz>
33511         PR other/80589
33512         * common.opt: Fix typo.
33513         * doc/invoke.texi: Likewise.
33515 2017-05-01  Jan Beulich  <jbeulich@suse.com>
33517         * config/i386/sse.md (xop_vpermil2<mode>3): Do not allow operand
33518         swapping, add (x,x,m,x,n) alternative.
33520 2017-05-01  Nathan Sidwell  <nathan@acm.org>
33522         * calls.c (combine_pending_stack_adjustment_and_call): Remove
33523         unnecessary unadjusted_alignment check.
33525 2017-05-01  Xi Ruoyao  <ryxi@stu.xidian.edu.cn>
33527         PR c++/80038
33528         * cilk_common.c (expand_builtin_cilk_detach): Move pedigree
33529         operations here.
33530         * gimplify.c (gimplify_cilk_detach): New function.
33531         (gimplify_call_expr, gimplify_modify_expr): Call it as needed.
33532         * tree-core.h: Document EXPR_CILK_SPAWN.
33533         * tree.h (EXPR_CILK_SPAWN): Define.
33535 2017-05-01  David Malcolm  <dmalcolm@redhat.com>
33537         * diagnostic-show-locus.c (layout::get_expanded_location): Rewrite
33538         to use new fixit_hint representation, using the "replace" logic.
33539         (get_line_span_for_fixit_hint): Likewise.
33540         (layout::print_any_fixits): Likewise.
33541         (selftest::test_one_liner_many_fixits): Rename to...
33542         (selftest::test_one_liner_many_fixits_1): ...this, and update
33543         comment and expected output to reflect that the multiple fix-it
33544         hints are now consolidated into one insertion.
33545         (selftest::test_one_liner_many_fixits_2): New test.
33546         (selftest::test_diagnostic_show_locus_one_liner): Update for
33547         above.
33548         (selftest::test_fixit_consolidation): Update for fix-it API
33549         change.
33550         * diagnostic.c (print_parseable_fixits): Likewise.
33551         * edit-context.c (edited_line::m_line_events): Convert from
33552         auto_vec <line_event *> to auto_vec <line_event>.
33553         (class line_event): Convert from abstract base class to a concrete
33554         class, taking over the role of replace_event.
33555         (class insert_event): Delete.
33556         (class replace_event): Rename to class line_event.  Convert to
33557         half-open range.
33558         (edit_context::add_fixits): Reimplement.
33559         (edit_context::apply_insert): Delete.
33560         (edit_context::apply_replace): Rename to...
33561         (edit_context::apply_fixit): ...this.  Convert to half-open range.
33562         (edited_file::apply_insert): Delete.
33563         (edited_file::apply_replace): Rename to...
33564         (edited_file::apply_fixit): ...this.
33565         (edited_line::~edited_line): Drop deletion of events.
33566         (edited_line::apply_insert): Delete.
33567         (edited_line::apply_replace): Rename to...
33568         (edited_line::apply_fixit): ...this.  Convert to half-open range.
33569         Update for change to type of m_line_events.
33570         * edit-context.h (edit_context::apply_insert): Delete.
33571         (edit_context::apply_replace): Rename to...
33572         (edit_context::apply_fixit): ...this.
33574 2017-05-01  Martin Sebor  <msebor@redhat.com>
33576         * gimple-ssa-sprintf.c (format_integer): Set knownrange when it's
33577         known.
33579 2017-05-01  Uros Bizjak  <ubizjak@gmail.com>
33581         PR target/68491
33582         * config/i386/cpuid.h (__get_cpuid): Always return 0 when
33583         __get_cpuid_max returns 0.
33584         (__get_cpuid_count): Ditto.
33586 2017-05-01  Eric Botcazou  <ebotcazou@adacore.com>
33588         * tree.c (substitute_in_expr) <tcc_vl_exp>: Also inline a call if the
33589         replacement expression is another instance of one of its arguments.
33591 2017-05-01  Jakub Jelinek  <jakub@redhat.com>
33593         PR target/79430
33594         * rtlanal.c (reg_set_p): If reg is a stack_pointer_rtx, also
33595         check for stack push/pop autoinc.
33596         * config/i386/i386.c (ix86_agi_dependent): Return false
33597         if the only reason why modified_in_p returned true is that
33598         addr is SP based and set_insn is a push or pop.
33600 2017-04-29  Jan Hubicka  <hubicka@ucw.cz>
33602         * ipa-inline.c (compute_inlined_call_time): Remove now unnecesary
33603         overflow check.
33605 2017-04-29  Jan Hubicka  <hubicka@ucw.cz>
33607         PR ipa/79224
33608         * ipa-inline-analysis.c (dump_predicate): Add optional parameter NL.
33609         (account_size_time): Use two predicates - exec_pred and
33610         nonconst_pred_ptr.
33611         (evaluate_conditions_for_known_args): Compute both clause and
33612         nonspec_clause.
33613         (evaluate_properties_for_edge): Evaulate both clause and nonspec_clause.
33614         (inline_summary_t::duplicate): Update.
33615         (estimate_function_body_sizes): Caluculate exec and nonconst predicates
33616         separately.
33617         (compute_inline_parameters): Likewise.
33618         (estimate_edge_size_and_time): Update caluclation of time.
33619         (estimate_node_size_and_time): Compute both time and nonspecialized
33620         time.
33621         (estimate_ipcp_clone_size_and_time): Update.
33622         (inline_merge_summary): Update.
33623         (do_estimate_edge_time): Update.
33624         (do_estimate_edge_size): Update.
33625         (do_estimate_edge_hints): Update.
33626         (inline_read_section, inline_write_summary): Stream both new predicates.
33627         * ipa-inline.c (compute_uninlined_call_time): Take uninlined_call_time
33628         as argument.
33629         (compute_inlined_call_time): Cleanup.
33630         (big_speedup_p): Update.
33631         (edge_badness): Update.
33632         * ipa-inline.h (INLINE_TIME_SCALE): Remove.
33633         (size_time_entry): Replace predicate by exec_predicate and
33634         nonconst_predicate.
33635         (edge_growth_cache_entry): Cache both time nad nonspecialized time.
33636         (estimate_edge_time): Return also nonspec_time.
33637         (reset_edge_growth_cache): Update.
33639 2017-04-29  Jakub Jelinek  <jakub@redhat.com>
33641         PR rtl-optimization/80491
33642         * ifcvt.c (noce_process_if_block): When looking for x setter
33643         with missing else_bb, don't check only the insn right before
33644         cond_earliest, but look for the last insn that x is modified in
33645         within the same bb.
33647         PR rtl-optimization/80491
33648         * alias.c (memory_modified_in_insn_p): Return true for CALL_INSNs.
33650 2017-04-29  Marc Glisse  <marc.glisse@inria.fr>
33652         PR tree-optimization/80487
33653         * tree-ssa-alias.c (stmt_kills_ref_p): Handle stpncpy and strncpy.
33655 2017-04-29  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
33657         PR tree-optimization/79697
33658         * tree-ssa-dce.c (mark_stmt_if_obviously_necessary): Check if callee
33659         is BUILT_IN_STRDUP, BUILT_IN_STRNDUP, BUILT_IN_REALLOC.
33660         (propagate_necessity): Check if def_callee is BUILT_IN_STRDUP or
33661         BUILT_IN_STRNDUP.
33662         * gimple-fold.c (gimple_fold_builtin_realloc): New function.
33663         (gimple_fold_builtin): Call gimple_fold_builtin_realloc.
33665 2017-04-28  Martin Sebor  <msebor@redhat.com>
33667         PR tree-optimization/80523
33668         * gimple-ssa-sprintf.c (target_to_host_charmap): New global variable.
33669         (init_target_to_host_charmap, target_to_host, target_strtol10): New
33670         functions.
33671         (maybe_warn, format_directive, parse_directive): Use new functions.
33672         (pass_sprintf_length::execute): Call init_target_to_host_charmap.
33674 2017-04-28  Marc Glisse  <marc.glisse@inria.fr>
33676         * match.pd (X+Z OP Y+Z, X-Z OP Y-Z, Z-X OP Z-Y): New transformations.
33678 2017-04-28  Bernd Edlinger  <bernd.edlinger@hotmail.de>
33680         * configure.ac (SYSTEM_HEADER_DIR, BUILD_SYSTEM_HEADER_DIR,
33681         target_header_dir): Set correctly.
33682         * configure: Regenerated.
33683         * Makefile.in (BUILD_SYSTEM_HEADER_DIR): New make variabe.
33684         (LIMITS_H_TEST, if_multiarch, stmp-fixinc): Use BUILD_SYSTEM_HEADER_DIR
33685         instead of SYSTEM_HEADER_DIR.
33687 2017-04-28  Jan Hubicka  <hubicka@ucw.cz>
33689         * ipa-cp.c (perform_estimation_of_a_value): Turn time to sreal.
33690         (estimate_local_effects): Likewise.
33691         * ipa-inline.c (compute_inlined_call_time, want_inline_small_function_p,
33692         edge_badness, inline_small_functions, dump_overall_stats): LIkewise.
33693         * ipa-inline.h (edge_growth_cache_entry, estimate_time_after_inlining,
33694         estimate_ipcp_clone_size_and_time, do_estimate_edge_time,
33695         do_estimate_edge_time, estimate_edge_time): Likewise.
33696         * ipa-inline-analysis.c (estimate_node_size_and_time,
33697         estimate_ipcp_clone_size_and_time, do_estimate_edge_time): Likewise.
33698         (estimate_time_after_inlining): Remove.
33700 2017-04-28  Martin Liska  <mliska@suse.cz>
33702         * doc/gcov.texi: Enhance documentation of gcov.
33704 2017-04-28  Martin Liska  <mliska@suse.cz>
33706         * doc/gcov.texi: Sort options in alphabetic order.
33707         * doc/gcov-dump.texi: Likewise.
33708         * doc/gcov-tool.texi: Likewise.
33709         * gcov.c (print_usage): Likewise.
33710         * gcov-dump.c (print_usage): Likewise.
33711         * gcov-tool.c (print_merge_usage_message): Likewise.
33712         (print_rewrite_usage_message): Likewise.
33713         (print_overlap_usage_message): Likewise.
33715 2017-04-28  Martin Liska  <mliska@suse.cz>
33717         PR gcov-profile/53915
33718         * gcov.c (format_gcov): Print 'NAN %' when top > bottom.
33720 2017-04-28  Martin Liska  <mliska@suse.cz>
33722         PR gcov-profile/79891
33723         * gcov.c (add_line_counts): Assign BBs to lines just if the BB
33724         is marked by compiler as living on a line.
33725         (get_cycles_count): Remove usage of the union.
33726         (output_intermediate_file): Likewise.
33727         (find_source): Fix GNU coding style.
33728         (accumulate_line_counts): Remove old non-all block mode.
33729         (output_lines): Remove usage of the union.
33730         * profile.c (output_location): Include all BBs, even if
33731         belonging to a same line (and file) as a previous BB.
33733 2017-04-28  Martin Liska  <mliska@suse.cz>
33735         * gcov.c (process_args): Handle new argument 'w'.
33736         (read_graph_file): Assign ID to BBs.
33737         (output_branch_count): Display BB # if verbose flag is set.
33738         (output_lines): Likewise for arcs.
33739         (print_usage): Add '--verbose' option help.
33740         * doc/gcov.texi: Document --verbose (-w) option.
33742 2017-04-28  Martin Liska  <mliska@suse.cz>
33744         * gcov.c (struct block_location_info): New struct.
33745         (process_file): Fill up the new structure.
33746         (read_graph_file): Replace usage of encoding by the newly added
33747         struct.
33748         (add_line_counts): Likewise.
33749         (accumulate_line_counts): Remove usage of the union.
33750         (function_info::function_info): New function.
33751         (function_info::~function_info): Likewise.
33752         (process_file): Call delete instead of release_function.
33753         (release_function): Release the function.
33754         (release_structures): Call delete instead of release_function.
33755         (solve_flow_graph): Replace usage of num_blocks.
33756         (find_exception_blocks): Likewise.
33757         (output_lines): Fix GNU coding style.
33759 2017-04-28  Martin Liska  <mliska@suse.cz>
33761         PR driver/56469
33762         * coverage.c (coverage_remove_note_file): New function.
33763         * coverage.h: Declare the function.
33764         * toplev.c (finalize): Clean if an error has been seen.
33766 2017-04-28  Martin Liska  <mliska@suse.cz>
33768         PR gcov-profile/80031
33769         * gcov-dump.c (tag_blocks): Just print number of basic blocks.
33770         * gcov-io.h (GCOV_TAG_BLOCKS_NUM): Remove unused macro.
33771         * gcov.c (read_graph_file): Read just number of blocks.
33772         * profile.c (branch_prob): Do not stream 0 flags per a basic
33773         block.
33775 2017-04-28  Martin Liska  <mliska@suse.cz>
33777         * gcov-dump.c (tag_*): Add new argument to declarations.
33778         (dump_gcov_file): Likewise.
33779         (tag_blocks): Add and use new argument depth.
33780         (tag_arcs): Likewise.
33781         (tag_lines): Likewise.
33782         (tag_counters): Likewise.
33783         (tag_summary): Likewise.
33784         (dump_working_sets): Use depth to do a proper indentation.
33786 2017-04-28  Jakub Jelinek  <jakub@redhat.com>
33788         PR bootstrap/80531
33789         * cgraph.h (symtab_node::debug_symtab): No longer inline.
33790         * symtab.c (symtab_node::debug_symtab): Move definition here.
33792 2017-04-28  Richard Biener  <rguenther@suse.de>
33794         * lto-streamer.h (LTO_major_version): Bump to 7.
33796 2017-04-28  Richard Biener  <rguenther@suse.de>
33798         * tree-vrp.c (assert_info): New struct.
33799         (add_assert_info): New helper.
33800         (register_edge_assert_for_2): Refactor to add asserts to a vector
33801         of assert_info.
33802         (register_edge_assert_for_1): Likewise.
33803         (register_edge_assert_for): Likewise.
33804         (finish_register_edge_assert_for): New helper actually registering
33805         asserts where live on edge.
33806         (find_conditional_asserts): Adjust.
33807         (find_switch_asserts): Likewise.
33808         (evrp_dom_walker::try_find_new_range): Generalize.
33809         (evrp_dom_walker::before_dom_children): Use register_edge_assert_for.
33811 2017-04-27  Marek Polacek  <polacek@redhat.com>
33813         PR sanitizer/80349
33814         * fold-const.c (fold_binary_loc) <case EQ_EXPR, NE_EXPR>: Convert
33815         arg10 and arg11 to itype.
33817 2017-04-27  Jonathan Wakely  <jwakely@redhat.com>
33819         * doc/extend.texi (Object Size Checking): Improve grammar.
33821 2017-04-27  Richard Earnshaw  <rearnsha@arm.com>
33823         PR target/80530
33824         * config/aarch64/aarch64.c (aarch64_emit_approx_sqrt): Ensure
33825         that the logic for permitting reciprocal estimates matches that
33826         in use_rsqrt_p.
33828 2017-04-27  Jakub Jelinek  <jakub@redhat.com>
33830         PR c++/80534
33831         * tree.c (type_cache_hasher::equal): Only compare
33832         TYPE_TYPELESS_STORAGE flag on non-aggregate element types.
33833         (build_array_type_1): Only hash TYPE_TYPELESS_STORAGE flag on
33834         non-aggregate element types.
33835         * tree.h (TYPE_TYPELESS_STORAGE): Fix comment typo, add more details
33836         about the flag on ARRAY_TYPEs in the comment, formatting fix.
33838 2017-04-27  Richard Biener  <rguenther@suse.de>
33840         PR middle-end/80533
33841         * emit-rtl.c (set_mem_attributes_minus_bitpos): When
33842         stripping ARRAY_REFs from MEM_EXPR make sure we're not
33843         keeping a reference to a trailing array.
33845 2017-04-27  Richard Biener  <rguenther@suse.de>
33847         PR middle-end/80539
33848         * tree-chrec.c (chrec_fold_plus_poly_poly): Deal with not
33849         being in loop-closed SSA form conservatively.
33850         (chrec_fold_multiply_poly_poly): Likewise.
33852 2017-04-27  Tamar Christina  <tamar.christina@arm.com>
33854         PR middle-end/79665
33855         * expr.c (expand_expr_real_2): Move TRUNC_MOD_EXPR, FLOOR_MOD_EXPR,
33856         CEIL_MOD_EXPR, ROUND_MOD_EXPR cases.
33858 2017-04-27  Jakub Jelinek  <jakub@redhat.com>
33860         PR target/77728
33861         * config/aarch64/aarch64.c (struct aarch64_fn_arg_alignment): Remove.
33862         (aarch64_function_arg_alignment): Return unsigned int again, but still
33863         ignore TYPE_FIELDS chain decls other than FIELD_DECLs.
33864         (aarch64_layout_arg): Adjust aarch64_function_arg_alignment caller.
33865         Don't emit -Wpsabi note.
33866         (aarch64_function_arg_boundary): Likewise.
33867         (aarch64_gimplify_va_arg_expr): Adjust aarch64_function_arg_alignment
33868         caller.
33870 2017-04-26  Nathan Sidwell  <nathan@acm.org>
33872         * tree.h (crc32_unsigned_n): Declare.
33873         (crc32_unsigned, crc32_unsigned): Make inline.
33874         * tree.c (crc32_unsigned_bits): Replace with ...
33875         (crc32_unsigned_n): ... this.
33876         (crc32_unsigned, crc32_byte): Remove.
33877         (crc32_string): Remove unnecessary braces.
33879 2017-04-25  Jan Hubicka  <hubicka@ucw.cz>
33881         * ipa-cp.c (estimate_local_effects): Convert sreal to int.
33882         * ipa-inline-analysis.c (MAX_TIME): Remove.
33883         (account_size_time): Use sreal for time.
33884         (dump_inline_summary): Update.
33885         (estimate_function_body_sizes): Update.
33886         (estimate_edge_size_and_time): Update.
33887         (estimate_calls_size_and_time): Update.
33888         (estimate_node_size_and_time): Update.
33889         (inline_merge_summary): Update.
33890         (inline_update_overall_summary): Update.
33891         (estimate_time_after_inlining): Update.
33892         (inline_read_section): Update.
33893         (inline_write_summary): Update.
33894         * ipa-inline.c (compute_uninlined_call_time): Update.
33895         (compute_inlined_call_time): Update.
33896         (recursive_inlining): Update.
33897         (inline_small_functions): Update.
33898         (dump_overall_stats): Update.
33899         * ipa-inline.h: Include sreal.h.
33900         (size_time_entry): Turn time to sreal.
33901         (inline_summary): Turn self_time nad time to sreal.
33903 2017-04-25  Jan Hubicka  <hubicka@ucw.cz>
33905         * sreal.c: Include backend.h, tree.h, gimple.h, cgraph.h and
33906         data-streamer.h
33907         (sreal::stream_out, sreal::stream_in): New.
33908         * sreal.h (sreal::stream_out, sreal::stream_in): Declare.
33910 2017-04-25  Jakub Jelinek  <jakub@redhat.com>
33912         * Makefile.in (s-options): Invoke opt-gather.awk with LC_ALL=C in the
33913         environment.
33915 2017-04-25  Uros Bizjak  <ubizjak@gmail.com>
33917         PR target/70799
33918         * config/i386/i386.c (dimode_scalar_to_vector_candidate_p):
33919         Handle ASHIFTRT.
33920         (dimode_scalar_chain::compute_convert_gain): Ditto.
33921         (dimode_scalar_chain::make_vector_copies): Ditto.
33922         (dimode_scalar_chain::convert_reg): Ditto.
33923         (dimode_scalar_chain::convert_insn): Ditto.
33924         * config/i386/sse.md (VI24_AVX512BW_1): Remove mode iterator.
33925         (VI248_AVX512BW_1): New mode iterator.
33926         (<mask_codefor>ashr<mode>3<mask_name>): Merge insn pattern with
33927         <mask_codefor>ashrv2di3<mask_name> insn using VI248_AVX512BW_1
33928         mode iterator.
33930 2017-04-25  Martin Sebor  <msebor@redhat.com>
33932         PR tree-optimization/80497
33933         * gimple-ssa-sprintf.c (get_int_range): Avoid assuming all integer
33934         constants are representable in HOST_WIDE_INT.
33935         (parse_directive): Ditto.
33937 2017-04-25  Martin Sebor  <msebor@redhat.com>
33939         PR bootstrap/80486
33940         * dominance.c (dom_info::m_n_basic_blocks): Change type to unsigned.
33941         (new_zero_array): Adjust signature.
33942         (dom_info::dom_init): Used unsigned rather that size_t.
33943         (dom_info::dom_info): Same.
33945 2017-04-25  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
33946             Jakub Jelinek  <jakub@redhat.com>
33948         PR target/77728
33949         * config/arm/arm.c: Include gimple.h.
33950         (aapcs_layout_arg): Emit -Wpsabi note if arm_needs_doubleword_align
33951         returns negative, increment ncrn only if it returned positive.
33952         (arm_needs_doubleword_align): Return int instead of bool,
33953         ignore DECL_ALIGN of non-FIELD_DECL TYPE_FIELDS chain
33954         members, but if there is any such non-FIELD_DECL
33955         > PARM_BOUNDARY aligned decl, return -1 instead of false.
33956         (arm_function_arg): Emit -Wpsabi note if arm_needs_doubleword_align
33957         returns negative, increment nregs only if it returned positive.
33958         (arm_setup_incoming_varargs): Likewise.
33959         (arm_function_arg_boundary): Emit -Wpsabi note if
33960         arm_needs_doubleword_align returns negative, return
33961         DOUBLEWORD_ALIGNMENT only if it returned positive.
33963 2017-04-25  Marek Polacek  <polacek@redhat.com>
33965         PR sanitizer/80349
33966         * fold-const.c (fold_binary_loc) <case BIT_IOR_EXPR>: Convert arg0's
33967         first argument to type.
33969 2017-04-25  Bill Seurer  <seurer@linux.vnet.ibm.com>
33971         PR target/80482
33972         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Change
33973         type checks to test for compatibility instead of equality.
33975 2017-04-25  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
33976             Jakub Jelinek  <jakub@redhat.com>
33978         PR target/77728
33979         * config/aarch64/aarch64.c (struct aarch64_fn_arg_alignment): New
33980         type.
33981         (aarch64_function_arg_alignment): Return aarch64_fn_arg_alignment
33982         struct.  Ignore DECL_ALIGN of decls other than FIELD_DECL for
33983         the alignment computation, but return their maximum in warn_alignment.
33984         (aarch64_layout_arg): Adjust aarch64_function_arg_alignment caller.
33985         Emit a -Wpsabi note if warn_alignment is 16 bytes, but alignment
33986         is smaller.
33987         (aarch64_function_arg_boundary): Likewise.  Simplify using MIN/MAX.
33988         (aarch64_gimplify_va_arg_expr): Adjust aarch64_function_arg_alignment
33989         caller.
33991 2017-04-25  Claudiu Zissulescu  <claziss@synopsys.com>
33993         * config/arc/simdext.md (dmpyh): Fix typo.
33995 2017-04-25  Richard Biener  <rguenther@suse.de>
33997         PR tree-optimization/80492
33998         * alias.c (compare_base_decls): Handle registers with asm
33999         specification conservatively.
34000         * tree-ssa-alias.c (decl_refs_may_alias_p): Handle
34001         compare_base_decls returning dont-know properly.
34003 2017-04-25  Claudiu Zissulescu  <claziss@synopsys.com>
34005         * config/arc/arc.c (LEGITIMATE_OFFSET_ADDRESS_P): Delete macro.
34006         (legitimate_offset_address_p): New function.
34007         (arc_legitimate_address_p): Use above function.
34009 2017-04-25  Claudiu Zissulescu  <claziss@synopsys.com>
34011         * config/arc/arc.c (arc_output_mi_thunk): Emit PIC calls.
34013 2017-04-25  Claudiu Zissulescu  <claziss@synopsys.com>
34015         * config/arc/arc.c (arc_conditional_register_usage): Use ACCL,
34016         ACCH registers whenever they are available.
34018 2017-04-25  Claudiu Zissulescu  <claziss@synopsys.com>
34020         * config/arc/arc.c (arc_conditional_register_usage): Make D0, D1
34021         double regs fix when not used.
34023 2017-04-25  Claudiu Zissulescu  <claziss@synopsys.com>
34025         * config/arc/arc.h (REGNO_OK_FOR_BASE_P): Consider also extension
34026         core registers.
34027         (REG_OK_FOR_INDEX_P_NONSTRICT): Likewise.
34028         (REG_OK_FOR_BASE_P_NONSTRICT): Likewise.
34030 2017-04-25  Claudiu Zissulescu  <claziss@synopsys.com>
34032         * config/arc/arc.c (arc_output_addsi): Check for h-register class
34033         when emitting short ADD instructions.
34035 2017-04-25  Claudiu Zissulescu  <claziss@synopsys.com>
34037         * config/arc/arc.md (cmpsi_cc_insn_mixed): Use 'h' register
34038         constraint.
34039         (cmpsi_cc_c_insn): Likewise.
34040         (cbranchsi4_scratch): Compute proper instruction length using
34041         compact_hreg_operand.
34042         * config/arc/predicates.md (compact_hreg_operand): New predicate.
34044 2017-04-25  Richard Biener  <rguenther@suse.de>
34046         PR middle-end/80509
34047         * passes.c (pass_manager::pass_manager): Initialize
34048         m_name_to_pass_map.
34050 2017-04-25  Richard Biener  <rguenther@suse.de>
34052         PR tree-optimization/79201
34053         * tree-ssa-sink.c (statement_sink_location): Handle calls.
34055 2017-04-25  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
34057         PR target/80464
34058         * config/s390/vector.md: Split MEM->GPR vector moves for
34059         non-s_operand addresses.
34061 2017-04-25  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
34063         PR target/79895
34064         * config/s390/predicates.md (reload_const_wide_int_operand): New
34065         predicate.
34066         * config/s390/s390.md ("movti"): Remove d/P alternative.
34067         ("movti_bigconst"): New pattern definition.
34069 2017-04-25  Dominik Vogt  <vogt@linux.vnet.ibm.com>
34071         PR target/80080
34072         * s390-protos.h (s390_expand_cs_hqi): Removed.
34073         (s390_expand_cs, s390_expand_atomic_exchange_tdsi): New prototypes.
34074         * config/s390/s390.c (s390_emit_compare_and_swap): Handle all integer
34075         modes as well as CCZ1mode and CCZmode.
34076         (s390_expand_atomic_exchange_tdsi, s390_expand_atomic): Adapt to new
34077         signature of s390_emit_compare_and_swap.
34078         (s390_expand_cs_hqi): Likewise, make static.
34079         (s390_expand_cs_tdsi): Generate an explicit compare before trying
34080         compare-and-swap, in some cases.
34081         (s390_expand_cs): Wrapper function.
34082         (s390_expand_atomic_exchange_tdsi): New backend specific expander for
34083         atomic_exchange.
34084         (s390_match_ccmode_set): Allow CCZmode <-> CCZ1 mode.
34085         * config/s390/s390.md ("atomic_compare_and_swap<mode>"): Merge the
34086         patterns for small and large integers.  Forbid symref memory operands.
34087         Move expander to s390.c.  Require cc register.
34088         ("atomic_compare_and_swap<DGPR:mode><CCZZ1:mode>_internal")
34089         ("*atomic_compare_and_swap<TDI:mode><CCZZ1:mode>_1")
34090         ("*atomic_compare_and_swapdi<CCZZ1:mode>_2")
34091         ("*atomic_compare_and_swapsi<CCZZ1:mode>_3"): Use s_operand to forbid
34092         symref memory operands.  Remove CC mode and call s390_match_ccmode
34093         instead.
34094         ("atomic_exchange<mode>"): Allow and implement all integer modes.
34096 2017-04-25  Dominik Vogt  <vogt@linux.vnet.ibm.com>
34098         * config/s390/s390.md (define_peephole2): New peephole to help
34099         combining the load-and-test pattern with volatile memory.
34101 2017-04-25  Dominik Vogt  <vogt@linux.vnet.ibm.com>
34103         * config/s390/s390.md ("cstorecc4"): Use load-on-condition and deal
34104         with CCZmode for TARGET_Z196.
34106 2017-04-25  Jakub Jelinek  <jakub@redhat.com>
34108         PR rtl-optimization/80501
34109         * combine.c (make_compound_operation_int): Set subreg_code to SET
34110         even for AND with mask of the sign bit of mode.
34112         PR rtl-optimization/80500
34113         * loop-unroll.c (combine_var_copies_in_loop_exit): Call copy_rtx on
34114         sum's initial value.
34116 2017-04-25  Julian Brown  <julian@codesourcery.com>
34117             Naveen H.S  <Naveen.Hurugalawadi@cavium.com>
34119         * config/aarch64/thunderx2t99.md (thunderx2t99_crc): New Reservation.
34121 2017-04-25  Marc Glisse  <marc.glisse@inria.fr>
34123         * fold-const.c (tree_single_nonzero_warnv_p): Handle SSA_NAME.
34125 2017-04-25  Julian Brown  <julian@codesourcery.com>
34126             Naveen H.S  <Naveen.Hurugalawadi@cavium.com>
34128         * config/aarch64/thunderx2t99.md (thunderx2t99_aes): New Reservation.
34129         (thunderx2t99_sha): New Reservation.
34131 2017-04-25  Julian Brown  <julian@codesourcery.com>
34132             Naveen H.S  <Naveen.Hurugalawadi@cavium.com>
34134         * config/aarch64/aarch64-simd.md (aarch64_simd_vec_set<mode>): Fix
34135         type for 1-element load.
34137 2017-04-24  Marc Glisse  <marc.glisse@inria.fr>
34139         * match.pd (X/[ex]C CMP Y/[ex]C): New transformation.
34141 2017-04-24  Martin Jambor  <mjambor@suse.cz>
34143         PR tree-optimization/80293
34144         * tree-sra.c (scalarizable_type_p): New parameter const_decl, make
34145         char arrays not totally scalarizable if it is false.
34146         (analyze_all_variable_accesses): Pass correct value in the new
34147         parameter.  Add a statistics counter.
34149 2017-04-24  Jan Hubicka  <hubicka@ucw.cz>
34151         PR middle-end/79931
34152         * ipa-devirt.c (dump_possible_polymorphic_call_targets): Fix ICE.
34154 2017-04-24  Richard Biener  <rguenther@suse.de>
34156         PR tree-optimization/80494
34157         * tree-scalar-evolution.c (analyze_scalar_evolution_1): Bail
34158         out for complex types.
34160 2017-04-24  Richard Biener  <rguenther@suse.de>
34162         * tree-ssa-sccvn.h (run_scc_vn): Adjust prototype.
34163         * tree-ssa-sccvn.c (print_scc): Print SCC size.
34164         (extract_and_process_scc_for_name): Never fail but drop SCC to varying.
34165         (DFS): Adjust and never fail.
34166         (sccvn_dom_walker::fail): Remove.
34167         (sccvn_dom_walker::before_dom_children): Adjust.
34168         (run_scc_vn): Likewise and never fail.
34169         * tree-ssa-pre.c (pass_pre::execute): Adjust.
34170         (pass_fre::execute): Likewise.
34172 2017-04-24  Richard Biener  <rguenther@suse.de>
34174         PR tree-optimization/79725
34175         * tree-ssa-sink.c (statement_sink_location): Return whether
34176         failure reason was zero uses.  Move that check later.
34177         (sink_code_in_bb): Deal with zero uses by removing the stmt
34178         if possible.
34180 2017-04-24  Richard Biener  <rguenther@suse.de>
34182         PR c++/2972
34183         * tree-ssa-uninit.c (warn_uninitialized_vars): Handle some
34184         pointer-based references.
34186 2017-04-24  Richard Biener  <rguenther@suse.de>
34188         PR bootstrap/79814
34189         * pass_manager.h (pass_manager::operator new): Remove.
34190         (pass_manager::operator delete): Likewise.
34191         * passes.c (pass_manager::operator new): Remove.
34192         (pass_manager::operator delete): Likewise.
34193         (pass_manager::pass_manager): Zero individual pass members.
34195 2017-04-23  Uros Bizjak  <ubizjak@gmail.com>
34197         PR target/70799
34198         * config/i386/i386.c (dimode_scalar_to_vector_candidate_p)
34199         <case ASHIFT, case LSHIFTRT>: Also consider variable shifts.
34200         Check "XEXP (src, 1)" operand here.
34201         <case PLUS, case MINUS, case IOR, case XOR, case AND>:
34202         Check "XEXP (src, 1)" operand here.
34203         (dimode_scalar_chain::make_vector_copies): Detect count register
34204         of a shift instruction.  Zero extend count register from QImode
34205         to DImode to satisfy vector shift pattern count operand predicate.
34206         Substitute vector shift count operand with a DImode copy.
34207         (dimode_scalar_chain::convert_reg): Ditto, zero-extend from
34208         vector register.
34210 2017-04-21  Uros Bizjak  <ubizjak@gmail.com>
34212         * config/i386/i386.md (*extzvqi_mem_rex64): Move above *extzv<mode>.
34213         Remove UNSPEC_NOREX_MEM tag.  Update corresponding peephole2 pattern.
34214         (*insvqi_1_mem_rex64): Move above insv<mode>_1.  Remove
34215         UNSPEC_NOREX_MEM tag.  Update corresponding peephole2 pattern.
34216         (UNSPEC_NOREX_MEM): Remove definition.
34218 2017-04-21  Richard Biener  <rguenther@suse.de>
34220         PR tree-optimization/79547
34221         * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
34222         Handle strlen, strcmp, strncmp, strcasecmp, strncasecmp, memcmp,
34223         bcmp, strspn, strcspn, __builtin_object_size and __builtin_constant_p
34224         without any constraints.
34226 2017-04-21  Richard Biener  <rguenther@suse.de>
34228         PR tree-optimization/78847
34229         * fold-const.c (split_tree): Handle POINTER_PLUS_EXPR.
34231 2017-04-21  Richard Biener  <rguenther@suse.de>
34233         * tree.h (build_qualified_type): Annotate with CXX_MEM_STAT_INFO.
34234         (build_distinct_type_copy): Likewise.
34235         (build_variant_type_copy): Likewise.
34236         * tree.c (build_qualified_type): Pass down mem-stat info.
34237         (build_distinct_type_copy): Likewise.
34238         (build_variant_type_copy): Likewise.
34240 2017-04-21  Richard Biener  <rguenther@suse.de>
34242         PR tree-optimization/80237
34243         * tree-ssa-pre.c (find_leader_in_sets): Add third set argument,
34244         defaulted to NULL.
34245         (phi_translate_1): Also allow a leader in AVAIL_OUT of pred
34246         for a simplified result.
34248 2016-04-21  Richard Biener  <rguenther@suse.de>
34250         * tree-ssa-loop-ivcanon.c (constant_after_peeling): Do not require
34251         sth as strict as a simple_iv but a chrec without symbols and an
34252         operand defined in the loop we are peeling (and not some subloop).
34253         (propagate_constants_for_unrolling): Propagate all constants.
34255 2017-04-20  Uros Bizjak  <ubizjak@gmail.com>
34257         PR target/79804
34258         * config/i386/i386.c (print_reg): Remove assert for disalowed
34259         regno values, call output_operand_lossage instead.
34261 2017-04-20  Uros Bizjak  <ubizjak@gmail.com>
34263         PR target/78090
34264         * config/i386/constraints.md (Yc): New register constraint.
34265         * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
34266         Use Yc constraint for alternative 2 of operand 0.  Remove
34267         preferred_for_speed attribute.
34269 2017-04-20  Alexander Monakov  <amonakov@ispras.ru>
34271         * omp-low.c (lower_lastprivate_clauses): Correct handling of linear and
34272         lastprivate clauses in SIMT case.
34274 2017-04-20  Volker Reichelt  <v.reichelt@netcologne.de>
34276         * doc/invoke.texi (-Wextra-semi): Document new warning option.
34278 2017-04-20  Richard Biener  <rguenther@suse.de>
34280         PR tree-optimization/57796
34281         * tree-vect-stmts.c (vect_model_store_cost): Cost scatters
34282         as N scalar stores.
34283         (vect_model_load_cost): Cost gathers as N scalar loads.
34285 2017-04-20  Richard Biener  <rguenther@suse.de>
34287         * ggc-page.c (ggc_allocated_p): Rename to ...
34288         (safe_lookup_page_table_entry): ... this and return the lookup
34289         result.
34290         (gt_ggc_m_S): Use safe_lookup_page_table_entry.
34292 2017-04-20  Richard Biener  <rguenther@suse.de>
34294         PR tree-optimization/80453
34295         * tree-ssa-sccvn.h (struct vn_phi_s): Add cclhs and ccrhs members.
34296         * tree-ssa-sccvn.c (cond_stmts_equal_p): Use recorded lhs and rhs
34297         from the conditions.
34298         (vn_phi_eq): Pass them down.
34299         (vn_phi_lookup): Record them.
34300         (vn_phi_insert): Likewise.
34302 2017-04-20  Matthew Fortune  <matthew.fortune@imgtec.com>
34304         * config/mips/mips.c (mips_expand_vec_perm_const): Re-fix
34305         uninitialized variable warning to avoid buffer overrun.
34307 2017-04-20  Alexander Monakov  <amonakov@ispras.ru>
34309         PR other/71250
34310         * doc/invoke.texi (-Wmissing-field-initializers): Mention that warning
34311         is suppressed for '{ 0 }' in C.
34313 2017-04-20  Jakub Jelinek  <jakub@redhat.com>
34315         * BASE-VER: Set to 8.0.0.
34317 2017-04-20  Thomas Preud'homme  <thomas.preudhomme@arm.com>
34319         * config/arm/arm.c (arm_elf_asm_cdtor): Create non-default
34320         priority .init_array and .fini_array section with SECTION_NOTYPE
34321         flag.
34323 2017-04-20  Jakub Jelinek  <jakub@redhat.com>
34325         PR middle-end/80423
34326         * tree.h (build_array_type): Add typeless_storage default argument.
34327         * tree.c (type_cache_hasher::equal): Also compare
34328         TYPE_TYPELESS_STORAGE flag for ARRAY_TYPEs.
34329         (build_array_type): Add typeless_storage argument, set
34330         TYPE_TYPELESS_STORAGE to it, if shared also hash it, and pass to
34331         recursive call.
34332         (build_nonshared_array_type): Adjust build_array_type_1 caller.
34333         (build_array_type): Likewise.  Add typeless_storage argument.
34335 2017-04-19  Eric Botcazou  <ebotcazou@adacore.com>
34336             Jakub Jelinek  <jakub@redhat.com>
34338         PR tree-optimization/80426
34339         * tree-vrp.c (extract_range_from_binary_expr_1): For an additive
34340         operation on symbolic operands, also compute the overflow for the
34341         invariant part when the operation degenerates into a negation.
34343 2017-04-19  Jakub Jelinek  <jakub@redhat.com>
34345         PR debug/80461
34346         * dwarf2out.c (modified_type_die, gen_type_die_with_usage):
34347         Check for t with zero TYPE_QUALS_NO_ADDR_SPACE.
34349         PR debug/80436
34350         * tree-ssa-loop-manip.c (find_uses_to_rename_def): Ignore debug uses.
34352 2017-04-19  Georg-Johann Lay  <avr@gjlay.de>
34354         PR target/80462
34355         * config/avr/avr.c (tree.h): Include it.
34356         (cgraph.h): Include it.
34357         (avr_encode_section_info): Don't warn for uninitialized progmem
34358         variable if it's just an alias.
34360 2017-04-19  Richard Biener  <rguenther@suse.de>
34362         PR ipa/65972
34363         * auto-profile.c (afdo_vpt_for_early_inline): Update SSA
34364         when needed by AutoPGO.
34366 2017-04-19  Paulo J. Matos  <paulo@matos-sorge.com>
34368         PR lto/50345
34369         * doc/lto.texi: Remove an extra 'that'.
34371 2017-04-19  Segher Boessenkool  <segher@kernel.crashing.org>
34373         PR rtl-optimization/80429
34374         * ira.c (split_live_ranges_for_shrink_wrap): Don't split regs that
34375         are only used in debug insns.
34377 2017-04-19  Eric Botcazou  <ebotcazou@adacore.com>
34378             Vladimir Makarov  <vmakarov@redhat.com>
34380         * config/sparc/predicates.md (input_operand): Add comment.  Return
34381         true for any memory operand when LRA is in progress.
34382         * config/sparc/sparc.c (sparc_expand_move): Minor formatting fix.
34384 2017-04-18  Jeff Law  <law@redhat.com>
34386         PR target/74563
34387         * mips.md ({return,simple_return}_internal): Do not overwrite
34388         operands[0].
34390 2017-04-18  Jakub Jelinek  <jakub@redhat.com>
34392         PR tree-optimization/80443
34393         * tree-vrp.c (intersect_ranges): For signed 1-bit precision type,
34394         instead of adding 1, subtract -1 and similarly instead of subtracting
34395         1 add -1.
34397 2017-04-18  Richard Sandiford  <richard.sandiford@arm.com>
34399         PR rtl-optimization/80357
34400         * haifa-sched.c (tmp_bitmap): New variable.
34401         (model_recompute): Handle duplicate use records.
34402         (alloc_global_sched_pressure_data): Initialize tmp_bitmap.
34403         (free_global_sched_pressure_data): Free it.
34405 2017-04-18  Bernd Edlinger  <bernd.edlinger@hotmail.de>
34407         Revert:
34408         2017-02-20  Bernd Edlinger  <bernd.edlinger@hotmail.de>
34409         * Makefile.in (BUILD_SYSTEM_HEADER_DIR): New make variabe.
34410         (LIMITS_H_TEST, if_multiarch, stmp-fixinc): Use BUILD_SYSTEM_HEADER_DIR
34411         instead of SYSTEM_HEADER_DIR.
34413 2017-04-18  Jeff Law  <law@redhat.com>
34415         PR middle-end/80422
34416         * cfgcleanup.c (try_crossjump_to_edge): Verify SRC1 and SRC2 have
34417         predecessors after walking up the insn chain.
34419 2017-04-18  Jakub Jelinek  <jakub@redhat.com>
34421         PR debug/80263
34422         * dwarf2out.c (modified_type_die): Try harder not to emit internal
34423         sizetype type into debug info.
34425 2017-04-18  Michael Meissner  <meissner@linux.vnet.ibm.com>
34427         PR target/80099
34428         * config/rs6000/rs6000.c (rs6000_expand_vector_extract): Eliminate
34429         unneeded test for TARGET_UPPER_REGS_SF.
34430         * config/rs6000/vsx.md (vsx_extract_v4sf_var): Likewise.
34432 2017-04-18  Jakub Jelinek  <jakub@redhat.com>
34434         PR sanitizer/80444
34435         * sancov.c (sancov_pass): Use gsi_start_nondebug_after_labels_bb
34436         instead of gsi_after_labels.
34438 2017-04-18  Jeff Law  <law@redhat.com>
34440         * regcprop.c (maybe_mode_change): Avoid creating copies of the
34441         stack pointer.
34443         Revert:
34444         2017-04-13  Jeff Law  <law@redhat.com>
34445         * config/mips.mips.md (zero_extendsidi2): Do not allow SP to appear
34446         in operands[1] if it is a MEM and TARGET_MIPS16 is active.
34448 2017-04-18  Georg-Johann Lay  <avr@gjlay.de>
34450         PR target/79453
34451         * config/avr/avr.c (intl.h): Include it.
34452         (avr_pgm_check_var_decl) [reason]: Wrap diagnostic snippets into _().
34454 2017-04-18  Martin Liska  <mliska@suse.cz>
34456         PR gcov-profile/78783
34457         * gcov-tool.c (gcov_output_files): Validate that destination
34458         file is either removed by the tool or by a user.
34460 2017-04-14  Andrew Burgess  <andrew.burgess@embecosm.com>
34461             Guy Benyei  <guybe@mellanox.com>
34463         * config/arc/arc.c (arc_reorg): Move loop_end_id into a more local
34464         block, and do not negate it, the stored id is already negative.
34466 2017-04-14  Andrew Burgess  <andrew.burgess@embecosm.com>
34468         * config/arc/arc.md (doloop_begin_i): Use @pcl assembler syntax.
34470 2017-04-14  Michael Meissner  <meissner@linux.vnet.ibm.com>
34472         PR target/80098
34473         * config/rs6000/rs6000-cpus.def (OTHER_P9_VECTOR_MASKS): Define
34474         masks of options that should be turned off if the VSX vector
34475         options are turned off.
34476         (OTHER_P8_VECTOR_MASKS): Likewise.
34477         (OTHER_VSX_VECTOR_MASKS): Likewise.
34478         * config/rs6000/rs6000.c (rs6000_option_override_internal): Call
34479         rs6000_disable_incompatible_switches to validate no type switches
34480         like -mvsx.
34481         (rs6000_incompatible_switch): New function to disallow turning on
34482         other vector options if -mno-vsx, -mno-power8-vector, or
34483         -mno-power9-vector are specified.
34485 2017-04-14  Claudiu Zissulescu  <claziss@synopsys.com>
34487         * config/arc/arc.h (CRT_CALL_STATIC_FUNCTION): Use long calls.
34489 2017-04-14  Claudiu Zissulescu  <claziss@synopsys.com>
34491         * config/arc/arc-protos.h (arc_decl_pretend_args): Remove.
34492         * config/arc/arc.c (arc_decl_pretend_args): Likewise.
34493         * config/arc/arc.h (CFA_FRAME_BASE_OFFSET): Likewise.
34494         (ARG_POINTER_CFA_OFFSET): Likewise.
34496 2017-04-14  Claudiu Zissulescu  <claziss@synopsys.com>
34498         * config/arc/arc.c (arc_mode_dependent_address_p): Relax
34499         conditions to take advantage of various optimizations.
34501 2017-04-13  Jeff Law  <law@redhat.com>
34503         * config/mips.mips.md (zero_extendsidi2): Do not allow SP to appear
34504         in operands[1] if it is a MEM and TARGET_MIPS16 is active.
34505         (zero_extendsidi2_dext): Likewise.
34507 2017-04-13  Jakub Jelinek  <jakub@redhat.com>
34509         PR sanitizer/80403
34510         * fold-const.c (fold_ternary_loc): Revert
34511         use op0 instead of fold_convert_loc (loc, type, arg0) part of
34512         2017-04-12 change.
34514 2017-04-13  Vladimir Makarov  <vmakarov@redhat.com>
34516         PR rtl-optimization/80343
34517         * lra-remat.c (update_scratch_ops): Assign original hard reg to
34518         new scratch pseudo.
34520 2017-04-13  Denis Khalikov <d.khalikov@partner.samsung.com>
34522         PR sanitizer/80414
34523         * ubsan.c (ubsan_expand_bounds_ifn): Pass original index
34524         to ubsan_encode_value.
34526 2017-04-13  Jeff Law  <law@redhat.com>
34528         * reload1.c (eliminate_regs_1): Call gen_rtx_raw_SUBREG for SUBREGs
34529         appearing in DEBUG_INSNs.
34531 2017-04-13  Martin Liska  <mliska@suse.cz>
34533         PR gcov-profile/80413
34534         * gcov-io.c (gcov_write_string): Copy to buffer just when
34535         allocated size is greater than zero.
34537 2017-04-13  Jakub Jelinek  <jakub@redhat.com>
34539         PR debug/80321
34540         * dwarf2out.c (decls_for_scope): Ignore declarations of
34541         current_function_decl in BLOCK_NONLOCALIZED_VARS.
34543 2017-04-12  Jan Hubicka  <hubicka@ucw.cz>
34545         PR lto/69953
34546         * ipa-visibility.c (non_local_p): Fix typos.
34547         (localize_node): When localizing symbol in same comdat group,
34548         dissolve the group only when we know external symbols are going
34549         to be privatized.
34550         (function_and_variable_visibility): Do not localize DECL_EXTERNAL.
34552 2017-04-12  Jakub Jelinek  <jakub@redhat.com>
34554         PR tree-optimization/79390
34555         * optabs.c (emit_conditional_move): If the preferred op2/op3 operand
34556         order does not result in usable sequence, retry with reversed operand
34557         order.
34559         PR sanitizer/80403
34560         PR sanitizer/80404
34561         PR sanitizer/80405
34562         * fold-const.c (fold_ternary_loc): Use op1 instead of arg1 as argument
34563         to fold_build2_loc.  Convert TREE_OPERAND (tem, 0) to type.  Use
34564         op0 instead of fold_convert_loc (loc, type, arg0).
34566 2017-04-12  Jeff Law  <law@redhat.com>
34568         * genattrtab.c (write_eligible_delay): Verify DELAY_INSN still
34569         has a delay slot in the generated code.
34571         * config/cris/cris.md (cris_preferred_reload_class): Return
34572         GENNONACR_REGS rather than GENERAL_REGS.
34574 2017-04-12  Jakub Jelinek  <jakub@redhat.com>
34576         PR c/80163
34577         * expr.c <CASE_CONVERT>: For EXPAND_INITIALIZER determine SIGN_EXTEND
34578         vs. ZERO_EXTEND based on signedness of treeop0's type rather than
34579         signedness of the result type.
34581 2017-04-12  Richard Biener  <rguenther@suse.de>
34582             Jeff Law  <law@redhat.com>
34584         PR tree-optimization/80359
34585         * tree-ssa-dse.c (maybe_trim_partially_dead_store): Do not
34586         trim stores to TARGET_MEM_REFs.
34588 2017-04-12  Richard Biener  <rguenther@suse.de>
34590         PR tree-optimization/79390
34591         * gimple-ssa-split-paths.c (is_feasible_trace): Restrict
34592         threading case even more.
34594 2017-04-12  Segher Boessenkool  <segher@kernel.crashing.org>
34596         PR target/80382
34597         * config/rs6000/sync.md (atomic_load<mode>, atomic_store<mode): Test
34598         for quad_address_p for TImode, instead of just not indexed_address.
34600 2017-04-12  Richard Biener  <rguenther@suse.de>
34601             Bernd Edlinger  <bernd.edlinger@hotmail.de>
34603         PR middle-end/79671
34604         * alias.c (component_uses_parent_alias_set_from): Handle
34605         TYPE_TYPELESS_STORAGE.
34606         (get_alias_set): Likewise.
34607         * tree-core.h (tree_type_common): Add typeless_storage flag.
34608         * tree.h (TYPE_TYPELESS_STORAGE): New macro.
34609         * stor-layout.c (place_union_field): Set TYPE_TYPELESS_STORAGE
34610         for types containing members with TYPE_TYPELESS_STORAGE.
34611         (place_field): Likewise.
34612         (layout_type): Likewise for ARRAY_TYPE.
34613         * lto-streamer-out.c (hash_tree): Hash TYPE_TYPELESS_STORAGE.
34614         * tree-streamer-in.c (unpack_ts_type_common_value_fields): Stream
34615         TYPE_TYPELESS_STORAGE.
34616         * tree-streamer-out.c (pack_ts_type_common_value_fields): Likewise.
34618 2017-04-12  Jakub Jelinek  <jakub@redhat.com>
34620         PR sanitizer/80349
34621         * fold-const.c (fold_binary_loc) <case BIT_IOR_EXPR>: Convert arg0's
34622         first argument to type.
34624 2017-04-11  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
34626         PR target/80376
34627         PR target/80315
34628         * config/rs6000/rs6000.c (rs6000_expand_unop_builtin): Return
34629         CONST0_RTX (mode) rather than const0_rtx where appropriate.
34630         (rs6000_expand_binop_builtin): Likewise.
34631         (rs6000_expand_ternop_builtin): Likewise; also add missing
34632         vsx_xxpermdi_* variants; also fix typo (arg1 => arg2) for
34633         vshasigma built-ins.
34634         * doc/extend.texi: Document that vec_xxpermdi's third argument
34635         must be a constant.
34637 2017-04-11  Uros Bizjak  <ubizjak@gmail.com>
34639         * config/i386/i386.c (dimode_scalar_chain::compute_convert_gain):
34640         Use shift_const cost parameter when calculating gain of STV shifts.
34642 2017-04-11  Vladimir Makarov  <vmakarov@redhat.com>
34644         PR rtl-optimization/70478
34645         * lra-constraints.c (process_alt_operands): Check memory for
34646         disfavoring memory insn operand.
34648 2017-04-11  Jakub Jelinek  <jakub@redhat.com>
34650         PR middle-end/80100
34651         * simplify-rtx.c (simplify_binary_operation_1) <case IOR>: Perform
34652         left shift in unsigned HOST_WIDE_INT type.
34654         PR rtl-optimization/80385
34655         * simplify-rtx.c (simplify_unary_operation_1): Don't transform
34656         (not (neg X)) into (plus X -1) for complex or non-integral modes.
34658         PR libgomp/80394
34659         * omp-low.c (scan_omp_task): Don't optimize away empty tasks
34660         if they have any depend clauses.
34662 2017-04-11  Martin Liska  <mliska@suse.cz>
34664         PR ipa/80212
34665         * cgraph.c (cgraph_node::dump): Dump calls_comdat_local.
34666         * ipa-split.c (split_function): Create a local comdat symbol
34667         if caller is in a comdat group.
34669 2017-04-11  Martin Liska  <mliska@suse.cz>
34671         PR ipa/80212
34672         * ipa-cp.c (determine_versionability): Handle calls_comdat_local
34673         flags.
34675 2017-04-11  Martin Sebor  <msebor@redhat.com>
34677         PR middle-end/80364
34678         * gimple-ssa-sprintf.c (get_int_range): Remove second argument and
34679         always use the int type.  Use INTEGRAL_TYPE_P() rather than testing
34680         for INTEGER_TYPE.
34681         (directive::set_width, directive::set_precision, format_character):
34682         Adjust.
34683         (parse_directive): Use INTEGRAL_TYPE_P() rather than testing for
34684         INTEGER_TYPE.
34686 2017-04-11  Richard Earnshaw  <rearnsha@arm.com>
34688         PR target/80389
34689         * config/arm/arm.c (arm_configure_build_target): When -mcpu and -arch
34690         conflict, set target->arch_name instead of target->cpu_name.
34692 2017-04-11  Richard Biener  <rguenther@suse.de>
34694         PR tree-optimization/80374
34695         * tree-ssa-dom.c (derive_equivalences_from_bit_ior): Use
34696         build_zero_cst, remove fold_convertible_p check again.
34698 2017-04-11  Martin Liska  <mliska@suse.cz>
34700         PR sanitizer/70878
34701         * ubsan.c (instrument_object_size): Do not instrument register
34702         variables.
34704 2017-04-11  Jakub Jelinek  <jakub@redhat.com>
34706         PR target/80381
34707         * config/i386/i386-builtin-types.def
34708         (V16HI_FTYPE_V16HI_INT_V16HI_UHI_COUNT,
34709         V16HI_FTYPE_V16HI_V8HI_V16HI_UHI_COUNT,
34710         V16SI_FTYPE_V16SI_INT_V16SI_UHI_COUNT,
34711         V16SI_FTYPE_V16SI_V4SI_V16SI_UHI_COUNT,
34712         V2DI_FTYPE_V2DI_INT_V2DI_UQI_COUNT,
34713         V2DI_FTYPE_V2DI_V2DI_V2DI_UQI_COUNT,
34714         V32HI_FTYPE_V32HI_INT_V32HI_USI_COUNT,
34715         V32HI_FTYPE_V32HI_V8HI_V32HI_USI_COUNT,
34716         V4DI_FTYPE_V4DI_INT_V4DI_UQI_COUNT,
34717         V4DI_FTYPE_V4DI_V2DI_V4DI_UQI_COUNT,
34718         V4SI_FTYPE_V4SI_INT_V4SI_UQI_COUNT,
34719         V4SI_FTYPE_V4SI_V4SI_V4SI_UQI_COUNT,
34720         V8DI_FTYPE_V8DI_INT_V8DI_UQI_COUNT,
34721         V8DI_FTYPE_V8DI_V2DI_V8DI_UQI_COUNT,
34722         V8HI_FTYPE_V8HI_INT_V8HI_UQI_COUNT,
34723         V8HI_FTYPE_V8HI_V8HI_V8HI_UQI_COUNT,
34724         V8SI_FTYPE_V8SI_INT_V8SI_UQI_COUNT,
34725         V8SI_FTYPE_V8SI_V4SI_V8SI_UQI_COUNT): New function type aliases.
34726         * config/i386/i386-builtin.def (__builtin_ia32_pslld512_mask,
34727         __builtin_ia32_pslldi512_mask, __builtin_ia32_psllq512_mask,
34728         __builtin_ia32_psllqi512_mask, __builtin_ia32_psrad512_mask,
34729         __builtin_ia32_psradi512_mask, __builtin_ia32_psraq512_mask,
34730         __builtin_ia32_psraqi512_mask, __builtin_ia32_psrld512_mask,
34731         __builtin_ia32_psrldi512_mask, __builtin_ia32_psrlq512_mask,
34732         __builtin_ia32_psrlqi512_mask, __builtin_ia32_psllwi128_mask,
34733         __builtin_ia32_pslldi128_mask, __builtin_ia32_psllqi128_mask,
34734         __builtin_ia32_psllw128_mask, __builtin_ia32_pslld128_mask,
34735         __builtin_ia32_psllq128_mask, __builtin_ia32_psllwi256_mask,
34736         __builtin_ia32_psllw256_mask, __builtin_ia32_pslldi256_mask,
34737         __builtin_ia32_pslld256_mask, __builtin_ia32_psllqi256_mask,
34738         __builtin_ia32_psllq256_mask, __builtin_ia32_psradi128_mask,
34739         __builtin_ia32_psrad128_mask, __builtin_ia32_psradi256_mask,
34740         __builtin_ia32_psrad256_mask, __builtin_ia32_psraqi128_mask,
34741         __builtin_ia32_psraq128_mask, __builtin_ia32_psraqi256_mask,
34742         __builtin_ia32_psraq256_mask, __builtin_ia32_psrldi128_mask,
34743         __builtin_ia32_psrld128_mask, __builtin_ia32_psrldi256_mask,
34744         __builtin_ia32_psrld256_mask, __builtin_ia32_psrlqi128_mask,
34745         __builtin_ia32_psrlq128_mask, __builtin_ia32_psrlqi256_mask,
34746         __builtin_ia32_psrlq256_mask, __builtin_ia32_psrawi256_mask,
34747         __builtin_ia32_psraw256_mask, __builtin_ia32_psrawi128_mask,
34748         __builtin_ia32_psraw128_mask, __builtin_ia32_psrlwi256_mask,
34749         __builtin_ia32_psrlw256_mask, __builtin_ia32_psrlwi128_mask,
34750         __builtin_ia32_psrlw128_mask, __builtin_ia32_psllwi512_mask,
34751         __builtin_ia32_psllw512_mask, __builtin_ia32_psrawi512_mask,
34752         __builtin_ia32_psraw512_mask, __builtin_ia32_psrlwi512_mask,
34753         __builtin_ia32_psrlw512_mask): Use _COUNT suffixed function type
34754         aliases.
34755         * config/i386/i386.c (ix86_expand_args_builtin): Rename last_arg_count
34756         flag to second_arg_count, handle 4 argument function type _COUNT
34757         aliases, handle second_arg_count on second argument rather than last.
34759 2017-04-10  Jeff Law  <law@redhat.com>
34761         PR tree-optimization/80374
34762         * tree-ssa-dom.c (derive_equivalences_from_bit_ior): Do not try to
34763         record anything if we can not convert integer_zero_node to the
34764         desired type.
34766 2017-04-10  Kelvin Nilsen  <kelvin@gcc.gnu.org>
34768         PR target/80108
34769         * config/rs6000/rs6000.c (rs6000_option_override_internal):
34770         Enhance special handling given to the TARGET_P9_MINMAX option in
34771         relation to certain other options.
34773 2017-04-10  Bin Cheng  <bin.cheng@arm.com>
34775         PR tree-optimization/80153
34776         * tree-ssa-loop-ivopts.c (add_iv_candidate_for_use): Check and
34777         remove POINTER_PLUS_EXPR's base part directly, rather than through
34778         aff_tree.
34780 2017-04-10  Richard Biener  <rguenther@suse.de>
34781             Bin Cheng  <bin.cheng@arm.com>
34783         PR tree-optimization/80153
34784         * tree-affine.c (aff_combination_to_tree): Get base pointer from
34785         the first element of pointer type aff_tree.  Build result expr in
34786         aff_tree's type.
34787         (add_elt_to_tree): Convert to type unconditionally.  Remove other
34788         fold_convert calls.
34789         * tree-ssa-loop-ivopts.c (alloc_iv): Pass in consistent types.
34790         (rewrite_use_nonlinear_expr): Check invariant using iv information.
34792 2017-04-10  Richard Biener  <rguenther@suse.de>
34794         * tree-ssa-structalias.c (find_func_aliases): Properly handle
34795         asm inputs.
34797 2017-04-10  Vladimir Makarov  <vmakarov@redhat.com>
34799         PR rtl-optimization/70478
34800         * lra-constraints.c (curr_small_class_check): New.
34801         (update_and_check_small_class_inputs): New.
34802         (process_alt_operands): Update curr_small_class_check.  Disfavor
34803         alternative insn memory operands.  Check available regs for small
34804         class operands.
34806 2017-03-31  Matthew Fortune  <matthew.fortune@imgtec.com>
34808         PR target/80057
34809         * config/mips/mips.opt (-mvirt): Update description.
34810         * doc/invoke.texi (-mvirt): Likewise.
34812 2017-04-10  Richard Biener  <rguenther@suse.de>
34814         PR middle-end/80362
34815         * fold-const.c (fold_binary_loc): Look at unstripped ops when
34816         looking for NEGATE_EXPR in -A / -B to A / B folding.
34818 2017-04-10  Martin Liska  <mliska@suse.cz>
34820         PR gcov-profile/80224
34821         * gcov.c (print_usage): Fix usage string.
34822         (get_gcov_intermediate_filename): Remove.
34823         (output_gcov_file): Use both for normal and intermediate format.
34824         (generate_results): Do not initialize special file for
34825         intermediate format.
34827 2017-04-10  Richard Biener  <rguenther@suse.de>
34829         PR tree-optimization/80304
34830         * tree-ssa-loop-im.c (ref_indep_loop_p_1): Also recurse
34831         for safelen.
34833 2017-04-10  Nathan Sidwell  <nathan@acm.org>
34835         PR target/79905
34836         * config/rs6000/rs6000.c (rs6000_vector_type): New.
34837         (rs6000_init_builtins): Use it.
34839 2016-04-10  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
34841         * config/arm/arm.md (<mrc>): Add mode to SET source.
34842         (<mrrc>): Likewise.
34844 2017-04-10  Richard Biener  <rguenther@suse.de>
34846         PR middle-end/80344
34847         * gimplify.c (is_gimple_mem_rhs_or_call): Allow CLOBBERs.
34849 2017-04-10  Jakub Jelinek  <jakub@redhat.com>
34851         PR target/80324
34852         * config/i386/avx512fintrin.h (_mm512_reduce_add_epi32,
34853         _mm512_reduce_mul_epi32, _mm512_reduce_and_epi32,
34854         _mm512_reduce_or_epi32, _mm512_mask_reduce_add_epi32,
34855         _mm512_mask_reduce_mul_epi32, _mm512_mask_reduce_and_epi32,
34856         _mm512_mask_reduce_or_epi32, _mm512_reduce_min_epi32,
34857         _mm512_reduce_max_epi32, _mm512_reduce_min_epu32,
34858         _mm512_reduce_max_epu32, _mm512_mask_reduce_min_epi32,
34859         _mm512_mask_reduce_max_epi32, _mm512_mask_reduce_min_epu32,
34860         _mm512_mask_reduce_max_epu32, _mm512_reduce_add_ps,
34861         _mm512_reduce_mul_ps, _mm512_mask_reduce_add_ps,
34862         _mm512_mask_reduce_mul_ps, _mm512_reduce_min_ps, _mm512_reduce_max_ps,
34863         _mm512_mask_reduce_min_ps, _mm512_mask_reduce_max_ps,
34864         _mm512_reduce_add_epi64, _mm512_reduce_mul_epi64,
34865         _mm512_reduce_and_epi64, _mm512_reduce_or_epi64,
34866         _mm512_mask_reduce_add_epi64, _mm512_mask_reduce_mul_epi64,
34867         _mm512_mask_reduce_and_epi64, _mm512_mask_reduce_or_epi64,
34868         _mm512_reduce_min_epi64, _mm512_reduce_max_epi64,
34869         _mm512_mask_reduce_min_epi64, _mm512_mask_reduce_max_epi64,
34870         _mm512_reduce_min_epu64, _mm512_reduce_max_epu64,
34871         _mm512_mask_reduce_min_epu64, _mm512_mask_reduce_max_epu64,
34872         _mm512_reduce_add_pd, _mm512_reduce_mul_pd, _mm512_mask_reduce_add_pd,
34873         _mm512_mask_reduce_mul_pd, _mm512_reduce_min_pd, _mm512_reduce_max_pd,
34874         _mm512_mask_reduce_min_pd, _mm512_mask_reduce_max_pd): New intrinsics.
34876 2017-04-08  Vladimir Makarov  <vmakarov@redhat.com>
34878         PR rtl-optimization/70478
34879         * lra-constraints.c: Reverse the last patch.
34881 2017-04-08  Andreas Tobler  <andreast@gcc.gnu.org>
34883         * config/aarch64/aarch64-freebsd.h: Define MCOUNT_NAME.
34884         Add comment for WCHAR_T.
34886 2017-04-08  Martin Liska  <mliska@suse.cz>
34888         Revert:
34889         2017-04-07  Martin Liska  <mliska@suse.cz>
34891         PR ipa/80212
34892         * ipa-split.c (split_function): Add function part to a same comdat
34893         group.
34895 2017-04-08  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
34897         PR target/80358
34898         * config/rs6000/rs6000.c (expand_block_compare): Fix boundary check.
34900 2017-04-07  Pat Haugen  <pthaugen@us.ibm.com>
34902         * rs6000/rs6000.c (vec_load_pendulum): Rename...
34903         (vec_pairing): ...to this.
34904         (power9_sched_reorder2): Rewrite code for pairing vector/vecload insns.
34905         (rs6000_sched_init): Adjust for name change.
34906         (struct rs6000_sched_context): Likewise.
34907         (rs6000_init_sched_context): Likewise.
34908         (rs6000_set_sched_context): Likewise.
34910 2017-04-07  Jakub Jelinek  <jakub@redhat.com>
34912         PR target/80322
34913         PR target/80323
34914         PR target/80325
34915         PR target/80326
34916         * config/i386/avxintrin.h (_mm256_cvtsd_f64, _mm256_cvtss_f32): New
34917         intrinsics.
34918         * config/i386/avx512fintrin.h (_mm512_int2mask, _mm512_mask2int,
34919         _mm512_abs_ps, _mm512_mask_abs_ps, _mm512_abs_pd, _mm512_mask_abs_pd,
34920         _mm512_cvtsd_f64, _mm512_cvtss_f32): Likewise.
34922 2017-04-07  Andreas Tobler  <andreast@gcc.gnu.org>
34924         * config/aarch64/aarch64-freebsd.h: Define WCHAR_TYPE.
34926 2017-04-07  Vladimir Makarov  <vmakarov@redhat.com>
34928         PR rtl-optimization/70703
34929         * ira-color.c (update_conflict_hard_regno_costs): Use
34930         int64_t instead of HOST_WIDE_INT.
34932 2017-04-07  Vladimir Makarov  <vmakarov@redhat.com>
34934         PR rtl-optimization/70478
34935         * lra-constraints.c (process_alt_operands): Disfavor alternative
34936         insn memory operands.
34938 2017-04-07  Jeff Law  <law@redhat.com>
34940         * config/iq2000/iq2000.c (final_prescan_insn): Do not separate a
34941         CALL and NOTE_INSN_CALL_ARG_LOCATION.
34943 2017-04-07  Martin Liska  <mliska@suse.cz>
34945         PR target/79889
34946         * config/aarch64/aarch64.c (aarch64_process_target_attr):
34947         Show error message instead of an ICE.
34949 2017-04-07  Martin Liska  <mliska@suse.cz>
34951         PR ipa/80212
34952         * ipa-split.c (split_function): Add function part to a same comdat
34953         group.
34955 2017-04-07  Richard Biener  <rguenther@suse.de>
34957         PR middle-end/80341
34958         * tree.c (get_unwidened): Also handle ! for_type case for
34959         INTEGER_CSTs.
34960         * convert.c (do_narrow): Split out from ...
34961         (convert_to_integer_1): ... here.  Do not pass final truncation
34962         type to get_unwidened for TRUNC_DIV_EXPR.
34964 2017-04-07  Richard Biener  <rguenther@suse.de>
34966         * tree-affine.c (wide_int_ext_for_comb): Take type rather
34967         than aff_tree.
34968         (aff_combination_const): Adjust.
34969         (aff_combination_scale): Likewise.
34970         (aff_combination_add_elt): Likewise.
34971         (aff_combination_add_cst): Likewise.
34972         (aff_combination_convert): Likewise.
34973         (add_elt_to_tree): Likewise.  Remove unused argument.
34974         (aff_combination_to_tree): Adjust calls to add_elt_to_tree.
34976 2017-04-07  Sebastian Huber  <sebastian.huber@embedded-brains.de>
34978         * config/arm/arm.h (ARM_DEFAULT_SHORT_ENUMS): Provide default
34979         definition.
34980         * config/arm/arm.c (arm_default_short_enums): Use
34981         ARM_DEFAULT_SHORT_ENUMS.
34982         * config/arm/rtems.h (ARM_DEFAULT_SHORT_ENUMS): Define.
34984 2017-04-06  Jakub Jelinek  <jakub@redhat.com>
34986         PR debug/80234
34987         * dwarf2out.c (gen_member_die): Handle C++17 inline static data
34988         members with redundant out-of-class redeclaration.
34990 2017-04-06  Uros Bizjak  <ubizjak@gmail.com>
34992         PR target/80286
34993         * config/i386/sse.md (*vec_extractv4si_0_zext_sse4): New pattern.
34994         * config/i386/i386.md (*zero_extendsidi2):
34995         Add (?*x,*x) and (?*v,*v) alternatives.
34997 2017-04-06  Uros Bizjak  <ubizjak@gmail.com>
34999         PR target/79733
35000         * config/i386/i386.c (ix86_expand_builtin)
35001         <case IX86_BUILTIN_K{,OR}TEST{C,Z}{8,16,32,64}>: Determine insn operand
35002         mode from insn data. Convert operands to insn operand mode.
35003         Copy operands that don't satisfy insn predicate to a register.
35005 2017-04-06  Sam Thursfield  <sam.thursfield@codethink.co.uk>
35007         * config/rs6000/x-aix: Increase memory limit for genautomata on AIX.
35008         Update comments.
35010 2017-04-06  Richard Biener  <rguenther@suse.de>
35012         PR tree-optimization/80334
35013         * tree-ssa-loop-ivopts.c (rewrite_use_address): Properly
35014         preserve alignment of accesses.
35016 2017-04-06  Richard Biener  <rguenther@suse.de>
35018         PR tree-optimization/80262
35019         * tree-sra.c (build_ref_for_offset): Preserve address-space
35020         information.
35021         * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address):
35022         Drop useless address-space information on MEM_REF offsets.
35024 2017-04-05  Andreas Schwab  <schwab@linux-m68k.org>
35026         * builtins.def (BUILT_IN_UPDATE_SETJMP_BUF): Fix type.
35028 2017-04-05  Vladimir Makarov  <vmakarov@redhat.com>
35030         PR rtl-optimization/70703
35031         * ira-color.c (update_conflict_hard_regno_costs): Use
35032         HOST_WIDE_INT instead of long.
35034 2017-04-05  Uros Bizjak  <ubizjak@gmail.com>
35036         PR target/80298
35037         * config/i386/mmintrin.h: Add -msse target option when __SSE__ is
35038         not defined for x86_64 target.  Add -mmmx target option when __SSE2__
35039         is not defined.
35040         * config/i386/mm3dnow.h: Add -msse target when __SSE__ is not defined
35041         for x86_64 target.  Handle -m3dnowa option.
35043 2017-04-05  Vladimir Makarov  <vmakarov@redhat.com>
35045         PR rtl-optimization/70703
35046         * ira-color.c (update_costs_from_allocno): Use the smallest mode.
35047         (update_conflict_hard_regno_costs): Use long instead of unsigned
35048         arithmetic for cost calculation.
35050 2017-04-05  Jakub Jelinek  <jakub@redhat.com>
35051             Bernd Edlinger  <bernd.edlinger@hotmail.de>
35053         PR sanitizer/80308
35054         * asan.c (asan_store_shadow_bytes): Fix location of last_chunk_value
35055         for big endian.
35057 2017-04-05  Eric Botcazou  <ebotcazou@adacore.com>
35059         PR target/78002
35060         * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Replace
35061         ptr_mode with Pmode throughout.
35062         * config/aarch64/aarch64.md (probe_stack_range_<PTR:mode): Rename
35063         into probe_stack_range and use DImode.
35065 2017-04-05  Dominik Vogt  <vogt@linux.vnet.ibm.com>
35067         PR target/79890
35068         * config/s390/s390.c (s390_register_info_gprtofpr): Return if
35069         call_eh_return is true.
35071 2017-04-05  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35073         * config/s390/s390-c.c (s390_resolve_overloaded_builtin):
35074         Initialize last_match_fntype_index.
35076 2017-04-05  Jakub Jelinek  <jakub@redhat.com>
35078         PR target/80310
35079         * tree-nvr.c: Include internal-fn.h.
35080         (pass_return_slot::execute): Ignore internal calls without
35081         direct optab.
35083 2017-04-04  Jakub Jelinek  <jakub@redhat.com>
35084             Richard Biener  <rguenther@suse.de>
35086         PR c++/80297
35087         * genmatch.c (capture::gen_transform): For GENERIC unshare_expr
35088         captures used multiple times, except for the last use.
35089         * generic-match-head.c: Include gimplify.h.
35091 2017-04-04  Jakub Jelinek  <jakub@redhat.com>
35093         PR tree-optimization/79390
35094         * target.h (struct noce_if_info): Declare.
35095         * targhooks.h (default_noce_conversion_profitable_p): Declare.
35096         * target.def (noce_conversion_profitable_p): New target hook.
35097         * ifcvt.h (struct noce_if_info): New type, moved from ...
35098         * ifcvt.c (struct noce_if_info): ... here.
35099         (noce_conversion_profitable_p): Renamed to ...
35100         (default_noce_conversion_profitable_p): ... this.  No longer
35101         static nor inline.
35102         (noce_try_store_flag_constants, noce_try_addcc,
35103         noce_try_store_flag_mask, noce_try_cmove, noce_try_cmove_arith,
35104         noce_convert_multiple_sets): Use targetm.noce_conversion_profitable_p
35105         instead of noce_conversion_profitable_p.
35106         * config/i386/i386.c: Include ifcvt.h.
35107         (ix86_option_override_internal): Don't override
35108         PARAM_MAX_RTL_IF_CONVERSION_INSNS default.
35109         (ix86_noce_conversion_profitable_p): New function.
35110         (TARGET_NOCE_CONVERSION_PROFITABLE_P): Redefine.
35111         * config/i386/x86-tune.def (X86_TUNE_ONE_IF_CONV_INSN): Adjust comment.
35112         * doc/tm.texi.in (TARGET_NOCE_CONVERSION_PROFITABLE_P): Add.
35113         * doc/tm.texi: Regenerated.
35115 2017-04-04  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
35117         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Grammar
35118         correction.
35120 2017-04-04  Thomas Preud'homme  <thomas.preudhomme@arm.com>
35122         PR target/80307
35123         * config/arm/arm.c (thumb1_rtx_costs): Give a cost of 32
35124         instructions for small multiply cores.
35126 2017-04-04  Jeff Law  <law@redhat.com>
35128         * config/mips/mips.c (mips_multi_add): Zero initialize the newly
35129         added member.
35130         (mips_expand_vec_perm_const): Initialize elements in orig_perm
35131         that are not set by the loop over the elements.
35133 2017-04-04  Jakub Jelinek  <jakub@redhat.com>
35135         PR target/80286
35136         * config/i386/i386.c (ix86_expand_args_builtin): If op has scalar
35137         int mode, convert_modes it to mode as unsigned, otherwise use
35138         lowpart_subreg to mode rather than SImode.
35139         * config/i386/sse.md (<mask_codefor>ashr<mode>3<mask_name>,
35140         ashr<mode>3, ashr<mode>3<mask_name>, <shift_insn><mode>3<mask_name>):
35141         Use DImode instead of SImode for the shift count operand.
35142         * config/i386/mmx.md (mmx_ashr<mode>3, mmx_<shift_insn><mode>3):
35143         Likewise.
35145 2017-04-04  Richard Biener  <rguenther@suse.de>
35147         PR middle-end/80281
35148         * match.pd (A + (-B) -> A - B): Make sure to preserve unsigned
35149         arithmetic done for the negate or the plus.  Simplify.
35150         (A - (-B) -> A + B): Likewise.
35151         * fold-const.c (split_tree): Make sure to not negate pointers.
35153 2017-04-04  Segher Boessenkool  <segher@kernel.crashing.org>
35155         PR rtl-optimization/60818
35156         * simplify-rtx.c (simplify_binary_operation_1): Do not replace
35157         a compare of comparisons with the thing compared if this results
35158         in a different machine mode.
35160 2017-04-03  Jonathan Wakely  <jwakely@redhat.com>
35162         * alias.c (base_alias_check): Fix typo in comment.
35163         * cgraph.h (class ipa_polymorphic_call_context): Likewise.
35164         * cgraphunit.c (symbol_table::compile): Likewise.
35165         * collect2.c (maybe_run_lto_and_relink): Likewise.
35166         * config/arm/arm.c (arm_thumb1_mi_thunk): Likewise.
35167         * config/avr/avr-arch.h (avr_arch_info_t): Likewise.
35168         * config/avr/avr.c (avr_map_op_t): Likewise.
35169         * config/cr16/cr16.h (DATA_ALIGNMENT): Likewise.
35170         * config/epiphany/epiphany.c (TARGET_ARG_PARTIAL_BYTES): Likewise.
35171         * config/epiphany/epiphany.md (movcc): Likewise.
35172         * config/i386/i386.c (legitimize_pe_coff_extern_decl): Likewise.
35173         * config/m68k/m68k.c (struct _sched_ib, m68k_sched_variable_issue):
35174         Likewise.
35175         * config/mips/mips.c (mips_save_restore_reg): Likewise.
35176         * config/rx/rx.c (rx_is_restricted_memory_address): Likewise.
35177         * config/s390/s390.c (Z10_EARLYLOAD_DISTANCE): Likewise.
35178         * config/sh/sh.c (sh_rtx_costs): Likewise.
35179         * fold-const.c (fold_truth_andor): Likewise.
35180         * genautomata.c (collapse_flag): Likewise.
35181         * gengtype.h (struct type::u::s): Likewise.
35182         * gensupport.c (has_subst_attribute, add_mnemonic_string): Likewise.
35183         * input.c (FORMAT_AMOUNT): Likewise.
35184         * ipa-cp.c (class ipcp_lattice, agg_replacements_to_vector)
35185         (known_aggs_to_agg_replacement_list): Likewise.
35186         * ipa-inline-analysis.c: Likewise.
35187         * ipa-inline.h (estimate_edge_time, estimate_edge_hints): Likewise.
35188         * ipa-polymorphic-call.c
35189         (ipa_polymorphic_call_context::restrict_to_inner_class): Likewise.
35190         * loop-unroll.c (analyze_insn_to_expand_var): Likewise.
35191         * lra.c (lra_optional_reload_pseudos, lra_subreg_reload_pseudos):
35192         Likewise.
35193         * modulo-sched.c (apply_reg_moves): Likewise.
35194         * omp-expand.c (build_omp_regions_1): Likewise.
35195         * trans-mem.c (struct tm_wrapper_hasher): Likewise.
35196         * tree-ssa-loop-ivopts.c (may_eliminate_iv): Likewise.
35197         * tree-ssa-loop-niter.c (maybe_lower_iteration_bound): Likewise.
35198         * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
35199         * value-prof.c: Likewise.
35200         * var-tracking.c (val_reset): Likewise.
35202 2017-04-03  Richard Biener  <rguenther@suse.de>
35204         PR tree-optimization/80275
35205         * fold-const.c (split_address_to_core_and_offset): Handle
35206         POINTER_PLUS_EXPR.
35208 2017-04-03  Eric Botcazou  <ebotcazou@adacore.com>
35210         * tree-nested.c (get_descriptor_type): Make sure that the alignment of
35211         descriptors is at least equal to that of functions.
35213 2017-04-02  Uros Bizjak  <ubizjak@gmail.com>
35215         * config/i386/sse.md (movdi_to_sse): Add missing DONE.
35217 2017-04-02  Uros Bizjak  <ubizjak@gmail.com>
35219         PR target/80250
35220         * config/i386/sse.md (mov<IMOD4:mode>): Remove insn pattern.
35221         (mov<IMOD4:mode>): New expander.
35222         (*mov<IMOD4:mode>_internal): New insn and split pattern.
35224 2017-03-31  Segher Boessenkool  <segher@kernel.crashing.org>
35226         PR rtl-optimization/79405
35227         * fwprop.c (propagations_left): New variable.
35228         (forward_propagate_into): Decrement it.
35229         (fwprop_init): Initialize it.
35230         (fw_prop): If the variable has reached zero, stop propagating.
35231         (fwprop_addr): Ditto.
35233 2017-03-31  Jakub Jelinek  <jakub@redhat.com>
35235         PR debug/79255
35236         * dwarf2out.c (decls_for_scope): If BLOCK_NONLOCALIZED_VAR is
35237         a FUNCTION_DECL, pass it as decl instead of origin to
35238         process_scope_var.
35240 2017-03-31  Alexander Monakov  <amonakov@ispras.ru>
35242         * config/nvptx/nvptx.c (nvptx_output_softstack_switch): Correct format
35243         string.
35245 2017-03-31  Pat Haugen  <pthaugen@us.ibm.com>
35247         PR target/80107
35248         * config/rs6000/rs6000.md (extendhi<mode>2): Add test for
35249         TARGET_VSX_SMALL_INTEGER.
35251 2017-03-31  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
35253         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Add
35254         reference to the OpenPOWER 64-Bit ELF V2 ABI Specification.
35256 2017-03-31  Matthew Fortune  <matthew.fortune@imgtec.com>
35258         * config/mips/mips-msa.md (msa_vec_extract_<msafmt_f>): Update
35259         extraction from odd-numbered MSA register.
35261 2017-03-31  Jakub Jelinek  <jakub@redhat.com>
35263         PR middle-end/80173
35264         * expmed.c (store_bit_field_1): Don't attempt to create
35265         a word subreg out of hard registers wider than word if they
35266         have HARD_REGNO_NREGS of 1 for their mode.
35268         PR middle-end/80163
35269         * varasm.c (initializer_constant_valid_p_1): Disallow sign-extending
35270         conversions to integer types wider than word and pointer.
35272         PR debug/80025
35273         * cselib.h (rtx_equal_for_cselib_1): Add depth argument.
35274         (rtx_equal_for_cselib_p): Pass 0 to it.
35275         * cselib.c (cselib_hasher::equal): Likewise.
35276         (rtx_equal_for_cselib_1): Add depth argument.  If depth
35277         is 128, don't look up VALUE locs and punt.  Increment
35278         depth in recursive calls when walking VALUE locs.
35280 2017-03-31  Bernd Edlinger  <bernd.edlinger@hotmail.de>
35282         * gcov.c (md5sum_to_hex): Fix output of MD5 hex bytes.
35283         (make_gcov_file_name): Use the canonical path name for generating
35284         the MD5 value.
35285         (read_line): Fix handling of files with ascii null bytes.
35287 2017-03-30  Matthew Fortune  <matthew.fortune@imgtec.com>
35289         * config/mips/mips.c (mips_expand_vector_init): Create a const_vector
35290         to initialise a vector register instead
35291         of using a const_int.
35293 2017-03-30  Jakub Jelinek  <jakub@redhat.com>
35295         PR translation/80189
35296         * gimplify.c (omp_default_clause): Use %qs instead of %s in
35297         diagnostic messages.
35299 2017-03-30  Peter Bergner  <bergner@vnet.ibm.com>
35301         PR target/80246
35302         * config/rs6000/dfp.md (dfp_dxex_<mode>): Update mode of operand 0.
35303         (dfp_diex_<mode>): Update mode of operand 1.
35304         * doc/extend.texi (dxex, dxexq): Document change to return type.
35305         (diex, diexq): Document change to argument type.
35307 2017-03-30  Martin Jambor  <mjambor@suse.cz>
35309         PR ipa/77333
35310         * cgraph.h (cgraph_build_function_type_skip_args): Declare.
35311         * cgraph.c (redirect_call_stmt_to_callee): Set gimple fntype so that
35312         it reflects the signature changes performed at the callee side.
35313         * cgraphclones.c (build_function_type_skip_args): Make public, renamed
35314         to cgraph_build_function_type_skip_args.
35315         (build_function_decl_skip_args): Adjust call to the above function.
35317 2017-03-30  Jakub Jelinek  <jakub@redhat.com>
35319         PR target/80206
35320         * config/i386/sse.md
35321         (<extract_type>_vextract<shuffletype><extract_suf>_mask): Use
35322         register as dest whenever it is a MEM not rtx_equal_p to the
35323         corresponding dup operand, and when forcing into reg move the
35324         reg into the memory afterwards.
35325         (<extract_type_2>_vextract<shuffletype><extract_suf_2>_mask):
35326         Likewise.  Use <ssehalfvecmode> instead of <ssequartermode>
35327         for the force_reg mode.
35328         (avx512vl_vextractf128<mode>): Use register as dest either
35329         always when a MEM, or when it is a MEM not rtx_equal_p to the
35330         corresponding dup operand, or even not when it is a CONST_VECTOR
35331         depending on the mode and lo vs. hi.
35332         (avx512dq_vextract<shuffletype>64x2_1_maskm): Remove extraneous
35333         parens.
35334         (avx512f_vextract<shuffletype>32x4_1_maskm): Likewise.
35335         (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name>):
35336         Likewise.  Require that operands[2] is even.
35337         (<mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name>):
35338         Remove extraneous parens.  Require that operands[2] is a multiple
35339         of 4.
35340         (vec_extract_lo_<mode><mask_name>): Don't bother testing if
35341         operands[0] is a MEM if <mask_applied>, the predicates/constraints
35342         disallow memory then.
35344 2017-03-30  Richard Biener  <rguenther@suse.de>
35346         PR tree-optimization/77498
35347         * tree-ssa-pre.c (phi_translate_1): Do not allow simplifications
35348         to non-constants over backedges.
35350 2017-03-29  Segher Boessenkool  <segher@kernel.crashing.org>
35352         PR rtl-optimization/80233
35353         * combine.c (combine_instructions): Only take NONDEBUG_INSN_P insns
35354         as last_combined_insn.  Do not test for BARRIER_P separately.
35356 2017-03-29  Andreas Schwab  <schwab@suse.de>
35358         PR ada/80146
35359         * calls.c (prepare_call_address): Convert funexp to Pmode before
35360         copying to temp reg.
35362 2017-03-29  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
35364         PR tree-optimization/80158
35365         * gimple-ssa-strength-reduction.c (replace_mult_candidate):
35366         Handle possible future case of more than one alternate
35367         interpretation.
35368         (replace_rhs_if_not_dup): Likewise.
35369         (replace_one_candidate): Likewise.
35371 2017-03-28  Vladimir Makarov  <vmakarov@redhat.com>
35373         PR rtl-optimization/80193
35374         * ira.c (ira): Do not check allocation for LRA.
35376 2017-03-28  Alexander Monakov  <amonakov@ispras.ru>
35378         * config/nvptx/nvptx-protos.h (nvptx_output_simt_enter): Declare.
35379         (nvptx_output_simt_exit): Declare.
35380         * config/nvptx/nvptx.c (nvptx_init_unisimt_predicate): Use
35381         cfun->machine->unisimt_location.  Handle NULL unisimt_predicate.
35382         (init_softstack_frame): Move initialization of crtl->is_leaf to...
35383         (nvptx_declare_function_name): ...here.  Emit declaration of local
35384         memory space buffer for omp_simt_enter insn.
35385         (nvptx_output_unisimt_switch): New.
35386         (nvptx_output_softstack_switch): New.
35387         (nvptx_output_simt_enter): New.
35388         (nvptx_output_simt_exit): New.
35389         * config/nvptx/nvptx.h (struct machine_function): New fields
35390         has_simtreg, unisimt_location, simt_stack_size, simt_stack_align.
35391         * config/nvptx/nvptx.md (UNSPECV_SIMT_ENTER): New unspec.
35392         (UNSPECV_SIMT_EXIT): Ditto.
35393         (omp_simt_enter_insn): New insn.
35394         (omp_simt_enter): New expansion.
35395         (omp_simt_exit): New insn.
35396         * config/nvptx/nvptx.opt (msoft-stack-reserve-local): New option.
35398         * internal-fn.c (expand_GOMP_SIMT_ENTER): New.
35399         (expand_GOMP_SIMT_ENTER_ALLOC): New.
35400         (expand_GOMP_SIMT_EXIT): New.
35401         * internal-fn.def (GOMP_SIMT_ENTER): New internal function.
35402         (GOMP_SIMT_ENTER_ALLOC): Ditto.
35403         (GOMP_SIMT_EXIT): Ditto.
35404         * target-insns.def (omp_simt_enter): New insn.
35405         (omp_simt_exit): Ditto.
35406         * omp-low.c (struct omplow_simd_context): New fields simt_eargs,
35407         simt_dlist.
35408         (lower_rec_simd_input_clauses): Implement SIMT privatization.
35409         (lower_rec_input_clauses): Likewise.
35410         (lower_lastprivate_clauses): Handle SIMT privatization.
35412         * omp-offload.c: Include langhooks.h, tree-nested.h, stor-layout.h.
35413         (ompdevlow_adjust_simt_enter): New.
35414         (find_simtpriv_var_op): New.
35415         (execute_omp_device_lower): Handle IFN_GOMP_SIMT_ENTER,
35416         IFN_GOMP_SIMT_ENTER_ALLOC, IFN_GOMP_SIMT_EXIT.
35418         * tree-inline.h (struct copy_body_data): New field dst_simt_vars.
35419         * tree-inline.c (expand_call_inline): Handle SIMT privatization.
35420         (copy_decl_for_dup_finish): Ditto.
35422         * tree-ssa.c (execute_update_addresses_taken): Handle GOMP_SIMT_ENTER.
35424 2017-03-28  Uros Bizjak  <ubizjak@gmail.com>
35426         PR target/53383
35427         * config/i386/i386.c (ix86_option_override_internal): Always
35428         allow -mpreferred-stack-boundary=3 for 64-bit targets.
35430 2017-03-28  Bin Cheng  <bin.cheng@arm.com>
35432         * tree-vect-loop.c (optimize_mask_stores): Add bb to the right loop.
35434 2017-03-28  Bin Cheng  <bin.cheng@arm.com>
35436         * tree-vect-loop-manip.c (slpeel_add_loop_guard): New param and
35437         mark new edge's irreducible flag accordign to it.
35438         (vect_do_peeling): Check loop preheader edge's irreducible flag
35439         and pass it to function slpeel_add_loop_guard.
35441 2017-03-28  Richard Sandiford  <richard.sandiford@arm.com>
35443         PR tree-optimization/80218
35444         * tree-call-cdce.c (shrink_wrap_one_built_in_call_with_conds):
35445         Update block frequencies and counts.
35447 2017-03-28  Richard Biener  <rguenther@suse.de>
35449         PR tree-optimization/78644
35450         * tree-ssa-ccp.c (evaluate_stmt): When we may not use the value
35451         of a simplification result we may not use it at all.
35453 2017-03-28  Richard Biener  <rguenther@suse.de>
35455         PR ipa/80205
35456         * tree-inline.c (copy_phis_for_bb): Do not create PHI node
35457         without arguments, generate default definition of a SSA name.
35459 2017-03-28  Richard Biener  <rguenther@suse.de>
35461         PR middle-end/80222
35462         * gimple-fold.c (gimple_fold_indirect_ref): Do not touch
35463         TYPE_REF_CAN_ALIAS_ALL references.
35464         * fold-const.c (fold_indirect_ref_1): Likewise.
35466 2017-03-28  Martin Liska  <mliska@suse.cz>
35468         PR ipa/80104
35469         * cgraphunit.c (cgraph_node::expand_thunk): Mark argument of a
35470         thunk call as DECL_GIMPLE_REG_P when vector or complex type.
35472 2017-03-28  Claudiu Zissulescu  <claziss@synopsys.com>
35473             Thomas Petazzoni  <thomas.petazzoni@free-electrons.com>
35475         * config/arc/arc.h (CPP_SPEC): Add subtarget_cpp_spec.
35476         (EXTRA_SPECS): Define.
35477         (SUBTARGET_EXTRA_SPECS): Likewise.
35478         (SUBTARGET_CPP_SPEC): Likewise.
35479         * config/arc/elf.h (EXTRA_SPECS): Renamed to
35480         SUBTARGET_EXTRA_SPECS.
35481         * config/arc/linux.h (SUBTARGET_CPP_SPEC): Define.
35483 2017-03-28  Claudiu Zissulescu  <claziss@synopsys.com>
35485         * config/arc/simdext.md (vst64_insn): Update pattern.
35486         (vld32wh_insn): Likewise.
35487         (vld32wl_insn): Likewise.
35488         (vld64_insn): Likewise.
35489         (vld32_insn): Likewise.
35491 2017-03-28  Marek Polacek  <polacek@redhat.com>
35493         PR sanitizer/80067
35494         * fold-const.c (fold_comparison): Use protected_set_expr_location
35495         instead of SET_EXPR_LOCATION.
35497 2017-03-28  Markus Trippelsdorf  <markus@trippelsdorf.de>
35499         * tree.c (add_expr): Avoid name lookup warning.
35501 2017-03-27  Jeff Law  <law@redhat.com>
35503         PR tree-optimization/80216
35504         * tree-ssa-dom.c (derive_equivalences_from_bit_ior): Fix typo in
35505         function name.  Limit recursion depth.
35506         (record_temporary_equivalences): Corresponding changes.
35508 2017-03-27  Jonathan Wakely  <jwakely@redhat.com>
35510         * doc/invoke.texi (-Wno-narrowing): Reorder so default behavior is
35511         covered first.
35513 2017-03-27  Jakub Jelinek  <jakub@redhat.com>
35515         PR target/80102
35516         * reg-notes.def (REG_CFA_NOTE): Define.  Use it for CFA related
35517         notes.
35518         * cfgcleanup.c (reg_note_cfa_p): New array.
35519         (insns_have_identical_cfa_notes): New function.
35520         (old_insns_match_p): Don't cross-jump in between /f
35521         and non-/f instructions.  If both i1 and i2 are frame related,
35522         verify all CFA notes, their order and content.
35524 2017-03-27  Michael Meissner  <meissner@linux.vnet.ibm.com>
35526         PR target/78543
35527         * config/rs6000/rs6000.md (bswaphi2_extenddi): Combine bswap
35528         HImode and SImode with zero extend to DImode to one insn.
35529         (bswap<mode>2_extenddi): Likewise.
35530         (bswapsi2_extenddi): Likewise.
35531         (bswaphi2_extendsi): Likewise.
35532         (bswaphi2): Combine bswap HImode and SImode into one insn.
35533         Separate memory insns from swapping register.
35534         (bswapsi2): Likewise.
35535         (bswap<mode>2): Likewise.
35536         (bswaphi2_internal): Delete, no longer used.
35537         (bswapsi2_internal): Likewise.
35538         (bswap<mode>2_load): Split bswap HImode/SImode into separate load,
35539         store, and gpr<-gpr swap insns.
35540         (bswap<mode>2_store): Likewise.
35541         (bswaphi2_reg): Register only splitter, combine with the splitter.
35542         (bswaphi2 splitter): Likewise.
35543         (bswapsi2_reg): Likewise.
35544         (bswapsi2 splitter): Likewise.
35545         (bswapdi2): If we have the LDBRX and STDBRX instructions, split
35546         the insns into load, store, and register/register insns.
35547         (bswapdi2_ldbrx): Likewise.
35548         (bswapdi2_load): Likewise.
35549         (bswapdi2_store): Likewise.
35550         (bswapdi2_reg): Likewise.
35552 2017-03-27  Gunther Nikl  <gnikl@users.sourceforge.net>
35554         * system.h (HAVE_DESIGNATED_INITIALIZERS): Fix non C++ case.
35555         (HAVE_DESIGNATED_UNION_INITIALIZERS): Likewise.
35557 2017-03-27  Kelvin Nilsen  <kelvin@gcc.gnu.org>
35559         PR target/80103
35560         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Edit and
35561         add comments.
35562         * config/rs6000/rs6000.c (rs6000_option_override_internal): Add
35563         special handling for target option conflicts between dform
35564         options (-mpower9-dform, -mpower9-dform-vector,
35565         -mpower9-dform-scalar) and -mno-direct-move.
35567 2017-03-27  Richard Biener  <rguenther@suse.de>
35569         PR tree-optimization/80181
35570         * tree-ssa-ccp.c (likely_value): UNDEFINED ^ X is UNDEFINED.
35572 2017-03-27  Claudiu Zissulescu  <claziss@synopsys.com>
35574         * config/arc/predicates.md (move_double_src_operand): Replace the
35575         call to move_double_src_operand with a call to address_operand.
35577 2017-03-27  Claudiu Zissulescu  <claziss@synopsys.com>
35579         * config/arc/elf.h (ARGET_ARC_TP_REGNO_DEFAULT): Define.
35580         * config/arc/linux.h (ARGET_ARC_TP_REGNO_DEFAULT): Likewise.
35581         * config/arc/arc.opt (mtp-regno): Use ARGET_ARC_TP_REGNO_DEFAULT.
35583 2017-03-27  Claudiu Zissulescu  <claziss@synopsys.com>
35585         * config/arc/predicates.md (long_immediate_loadstore_operand):
35586         Consider scaled addresses cases.
35588 2017-03-27  Claudiu Zissulescu  <claziss@synopsys.com>
35590         * config/arc/arc.c (arc_epilogue_uses): BLINK should be also
35591         restored when in interrupt.
35592         * config/arc/arc.md (simple_return): ARCv2 rtie instruction
35593         doesn't have delay slot.
35595 2017-03-27  Richard Biener  <rguenther@suse.de>
35597         PR ipa/79776
35598         * tree-ssa-structalias.c (associate_varinfo_to_alias): Skip
35599         inlined thunk clones.
35601 2017-03-27  Jakub Jelinek  <jakub@redhat.com>
35603         PR sanitizer/80168
35604         * asan.c (instrument_derefs): Copy over last operand from
35605         original COMPONENT_REF to the new COMPONENT_REF with
35606         DECL_BIT_FIELD_REPRESENTATIVE.
35607         * ubsan.c (instrument_object_size): Likewise.
35609 2017-03-27  Richard Biener  <rguenther@suse.de>
35611         PR tree-optimization/80170
35612         * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Make
35613         sure DR/SCEV didnt fold in constants we do not see when looking
35614         at the reference base alignment.
35616 2017-03-27  Richard Biener  <rguenther@suse.de>
35618         PR middle-end/80171
35619         * gimple-fold.c (fold_ctor_reference): Properly guard against
35620         NULL return value from canonicalize_constructor_val.
35622 2017-03-25  Uros Bizjak  <ubizjak@gmail.com>
35624         PR target/80180
35625         * config/i386/i386.c (ix86_expand_builtin)
35626         <IX86_BUILTIN_RDSEED{16,32,64}_STEP>: Do not expand arg0 between
35627         flags reg setting and flags reg using instructions.
35628         <IX86_BUILTIN_RDRAND{16,32,64}_STEP>: Ditto.  Use non-flags reg
35629         clobbering instructions to zero extend op2.
35631 2017-03-25  Gerald Pfeifer  <gerald@pfeifer.com>
35633         * doc/install.texi (Configuration) <--with-aix-soname>:
35634         Update link to AIX ld.
35636 2017-03-25  Bernd Schmidt  <bschmidt@redhat.com>
35638         PR rtl-optimization/80160
35639         PR rtl-optimization/80159
35640         * lra-assigns.c (must_not_spill_p): Tighten new test to also take
35641         reg_alternate_class into account.
35643 2017-03-24  Vladimir Makarov  <vmakarov@redhat.com>
35645         PR target/80148
35646         * lra-assigns.c (assign_by_spills): Add spilled non-reload pseudos
35647         to consider in curr_insn_transform.
35649 2017-03-24  Jakub Jelinek  <jakub@redhat.com>
35651         * genrecog.c (validate_pattern): Add VEC_SELECT validation.
35652         * genmodes.c (emit_min_insn_modes_c): Call emit_mode_nunits
35653         and emit_mode_inner.
35655 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35657         * config/s390/s390-builtins.def: Add VXE builtins.  Add a flags
35658         argument to the overloaded builtin variants.  Use the new flag to
35659         deprecate certain builtin variants.
35660         * config/s390/s390-builtin-types.def: Add new builtin types.
35661         * config/s390/s390-builtins.h: Support new flags field for
35662         overloaded builtins.
35663         * config/s390/s390-c.c (OB_DEF_VAR): New flags field.
35664         (s390_macro_to_expand): Enable vector float data type.
35665         (s390_cpu_cpp_builtins_internal): Indicate support of the new
35666         builtins by incrementing the __VEC__ version number.
35667         (s390_expand_overloaded_builtin): Support expansion of vec_xl and
35668         vec_xst.
35669         (s390_resolve_overloaded_builtin): Emit error messages depending
35670         on the builtin flags.
35671         * config/s390/s390.c (s390_expand_builtin): Support additional
35672         flags argument.  Change error message to match the messages
35673         emitted in s390-c.c.
35674         * config/s390/s390.md: New UNSPEC_* constants.
35675         (op_type): Add new instruction types.
35676         * config/s390/vecintrin.h: Add new builtins and test data class
35677         constants.
35678         * config/s390/vx-builtins.md (V_HW_32_64): Add V4SF.
35679         (V_HW_4, VEC_HW, VECF_HW): New mode iterators.
35680         (VEC_INEXACT, VEC_NOINEXACT): New constants.
35681         ("vec_splats<mode>", "vec_insert<mode>", "vec_promote<mode>")
35682         ("vec_insert_and_zero<mode>", "vec_mergeh<mode>")
35683         ("vec_mergel<mode>"): V_HW -> VEC_HW.
35685         ("vlrlrv16qi", "vstrlrv16qi", "vbpermv16qi", "vec_msumv2di")
35686         ("vmslg", "*vftci<mode>_cconly", "vftci<mode>_intcconly")
35687         ("*vftci<mode>", "vftci<mode>_intcc", "vec_double_s64")
35688         ("vec_double_u64", "vfmin<mode>", "vfmax<mode>"): New definition.
35690         ("and_av2df3", "and_cv2df3", "vec_andc_av2df3")
35691         ("vec_andc_cv2df3", "xor_av2df3", "xor_cv2df3", "vec_nor_av2df3")
35692         ("vec_nor_cv2df3", "ior_av2df3", "ior_cv2df3", "vec_nabs")
35693         ("*vftcidb", "*vftcidb_cconly", "vftcidb"): Remove definition.
35695         ("vec_all_<fpcmpcc:code>v2df", "vec_any_<fpcmpcc:code>v2df")
35696         ("vec_scatter_elementv4si_DI", "vec_cmp<fpcmp:code>v2df")
35697         ("vec_di_to_df_s64", "vec_di_to_df_u64", "vec_df_to_di_u64")
35698         ("vfidb", "*vldeb", "*vledb", "*vec_cmp<insn_cmp>v2df_cconly")
35699         ("vec_cmpeqv2df_cc", "vec_cmpeqv2df_cc", "vec_cmphv2df_cc")
35700         ("vec_cmphev2df_cc", "*vec_cmpeqv2df_cc")
35701         ("*vec_cmphv2df_cc", "*vec_cmphev2df_cc"): Enable new modes as ...
35703         ("vec_all_<fpcmpcc:code><mode>", "vec_any_<fpcmpcc:code><mode>")
35704         ("vec_scatter_element<V_HW_4:mode>_DI")
35705         ("vec_cmp<fpcmp:code><mode>", "vcdgb", "vcdlgb", "vclgdb")
35706         ("vec_fpint<mode>", "vflls")
35707         ("vflrd", "*vec_cmp<insn_cmp><mode>_cconly", "vec_cmpeq<mode>_cc")
35708         ("vec_cmpeq<mode>_cc", "vec_cmph<mode>_cc", "vec_cmphe<mode>_cc")
35709         ("*vec_cmpeq<mode>_cc", "*vec_cmph<mode>_cc")
35710         ("*vec_cmphe<mode>_cc"): ... these.
35712         ("vec_ctd_s64", "vec_ctsl", "vec_ctul", "vec_st2f"): Use rounding
35713         mode constant instead of magic value.
35715 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35717         * config/s390/s390.c (s390_expand_vec_compare): Support other
35718         vector floating point modes than just V2DF.
35719         (s390_expand_vcond): Likewise.
35720         (s390_hard_regno_mode_ok): Allow SFmode values in VRs.
35721         (s390_cannot_change_mode_class): Prevent mode changes between TF
35722         and V1TF in vector registers.
35723         * config/s390/s390.md (DF, SF): New mode attributes.
35724         ("*cmp<mode>_ccs", "add<mode>3", "sub<mode>3", "mul<mode>3")
35725         ("fma<mode>4", "fms<mode>4", "div<mode>3", "*neg<mode>2"): Add
35726         SFmode support for VRs.
35727         * config/s390/vector.md (V_HW, V_HW2, VT_HW, ti*, nonvec): Add new
35728         vector fp modes.
35729         (VFT, VF_HW): New mode iterators.
35730         (vw, sdx): New mode attributes.
35731         ("addv2df3", "subv2df3", "mulv2df3", "divv2df3", "sqrtv2df2")
35732         ("fmav2df4","fmsv2df4", "negv2df2", "absv2df2", "*negabsv2df2")
35733         ("smaxv2df3", "sminv2df3", "*vec_cmp<VFCMP_HW_OP:code>v2df_nocc")
35734         ("vec_cmpuneqv2df", "vec_cmpltgtv2df", "vec_orderedv2df")
35735         ("vec_unorderedv2df"): Adjust the v2df only patterns to support
35736         also the new vector floating point modes.  Renaming to ...
35738         ("add<mode>3", "sub<mode>3", "mul<mode>3", "div<mode>3")
35739         ("sqrt<mode>2", "fma<mode>4", "fms<mode>4", "neg<mode>2")
35740         ("abs<mode>2", "negabs<mode>2", "smax<mode>3")
35741         ("smin<mode>3", "*vec_cmp<VFCMP_HW_OP:code><mode>_nocc")
35742         ("vec_cmpuneq<mode>", "vec_cmpltgt<mode>", "vec_ordered<mode>")
35743         ("vec_unordered<mode>"): ... these.
35745         ("neg_fma<mode>4", "neg_fms<mode>4", "*smax<mode>3_vxe")
35746         ("*smin<mode>3_vxe", "*sminv2df3_vx", "*vec_extendv4sf")
35747         ("*vec_extendv2df"): New insn definitions.
35749 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35751         * config/s390/s390.md ("*adddi3_sign", "*subdi3_sign", "mulditi3")
35752         ("mulditi3_2", "*muldi3_sign"): New patterns.
35753         ("muldi3", "*muldi3", "mulsi3", "*mulsi3"): Add an expander and
35754         rename the pattern definition.
35756 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35758         * config/s390/s390.md ("indirect_jump"): Turn insn definition into
35759         expander.
35760         ("*indirect_jump", "*indirect2_jump"): New pattern definitions.
35762 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35764         * config/s390/s390.c (s390_expand_vec_init): Use vllezl
35765         instruction if possible.
35766         * config/s390/vector.md (vec_halfnumelts): New mode
35767         attribute.
35768         ("*vec_vllezlf<mode>"): New pattern.
35770 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35772         * config/s390/vector.md ("popcountv16qi2", "popcountv8hi2")
35773         ("popcountv4si2", "popcountv2di2"): Rename to ...
35774         ("popcount<mode>2", "popcountv8hi2_vx", "popcountv4si2_vx")
35775         ("popcountv2di2_vx"): ... these and add !TARGET_VXE to the
35776         condition.
35777         ("popcount<mode>2_vxe"): New pattern.
35779 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35781         * common/config/s390/s390-common.c (processor_flags_table): Add
35782         arch12.
35783         * config.gcc: Add arch12.
35784         * config/s390/driver-native.c (s390_host_detect_local_cpu):
35785         Default to arch12 for unknown CPU model numbers.
35786         * config/s390/s390-builtins.def: Add B_VXE builtin flag.
35787         * config/s390/s390-c.c (s390_cpu_cpp_builtins_internal): Adjust
35788         PROCESSOR_max sanity check.
35789         * config/s390/s390-opts.h (enum processor_type): Add
35790         PROCESSOR_ARCH12.
35791         * config/s390/s390.c (processor_table): Add arch12.
35792         (s390_expand_builtin): Add check for B_VXE flag.
35793         (s390_issue_rate): Add PROCESSOR_ARCH12.
35794         (s390_get_sched_attrmask): Likewise.
35795         (s390_get_unit_mask): Likewise.
35796         (s390_sched_score): Enable z13 scheduling for arch12.
35797         (s390_sched_reorder): Likewise.
35798         (s390_sched_variable_issue): Likewise.
35799         * config/s390/s390.h (enum processor_flags): Add PF_ARCH12 and
35800         PF_VXE.
35801         (s390_tune_attr): Use z13 scheduling also for arch12.
35802         (TARGET_CPU_ARCH12, TARGET_CPU_ARCH12_P, TARGET_CPU_VXE)
35803         (TARGET_CPU_VXE_P, TARGET_ARCH12, TARGET_ARCH12_P, TARGET_VXE)
35804         (TARGET_VXE_P): New macros.
35805         * config/s390/s390.md: Add arch12 to cpu attribute.  Add arch12
35806         and vxe to cpu_facility.  Add arch12 and vxe to enabled attribute.
35807         * config/s390/s390.opt: Add arch12 as processor_type.
35809 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35811         * config/s390/s390.md
35812         ("fixuns_truncdddi2", "fixuns_trunctddi2")
35813         ("fixuns_trunc<BFP:mode><GPR:mode>2"): Merge into ...
35814         ("fixuns_trunc<FP:mode><GPR:mode>2"): New expander.
35816         ("fixuns_trunc<BFP:mode><GPR:mode>2", "fixuns_trunc<mode>si2"):
35817         Rename expanders to ...
35819         ("fixuns_trunc<BFP:mode><GPR:mode>2_emu")
35820         ("fixuns_truncdddi2_emu"): ... these.
35822         ("fixuns_trunc<mode>si2_emu"): New expander.
35824         ("*fixuns_truncdfdi2_z13"): Rename to ...
35825         ("*fixuns_truncdfdi2_vx"): ... this.
35827 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35829         * config/s390/2964.md: Remove the single element vector compare
35830         instructions which are no longer used.
35831         * config/s390/s390.c (s390_select_ccmode): Remove handling of
35832         vector CCmodes.
35833         (s390_canonicalize_comparison): Remove handling of DFmode
35834         compares.
35835         (s390_expand_vec_compare_scalar): Remove function.
35836         (s390_emit_compare): Don't call s390_expand_vec_compare_scalar.
35837         * config/s390/s390.md ("*vec_cmp<insn_cmp>df_cconly"): Remove
35838         pattern.
35839         ("*cmp<mode>_ccs"): Add wfcdb instruction.
35841 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35843         * config/s390/s390.md ("mov<mode>_64dfp" DD_DF): Use vleig for loading a
35844         FP zero.
35845         ("*mov<mode>_64" DD_DF): Remove the vector instructions. These
35846         will anyway by matched by mov<mode>_64dfp.
35848 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35850         * config/s390/s390.md ("mov<mode>" SD_SF): Change vleg/vsteg to
35851         vlef/vstef.  Add missing operand to vleif.
35853 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35855         * config/s390/s390.c (s390_expand_vec_init): Enable vector load
35856         pair for all vector types with 64 bit elements.
35857         * config/s390/vx-builtins.md (V_HW_64): Move mode iterator to ...
35858         * config/s390/vector.md (V_HW_64): ... here.
35859         (V_128_NOSINGLE): New mode iterator.
35860         ("vec_init<V_HW:mode>"): Use V_128 as mode iterator.
35861         ("*vec_splat<mode>"): Use V_128_NOSINGLE mode iterator.
35862         ("*vec_tf_to_v1tf", "*vec_ti_to_v1ti"): New pattern definitions.
35863         ("*vec_load_pairv2di"): Change to ...
35864         ("*vec_load_pair<mode>"): ... this one.
35866 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35868         * config/s390/constraints.md: Add comments.
35869         (jKK): Reject element sizes > 8 bytes.
35870         * config/s390/s390.c (s390_split_ok_p): Enable splitting also for
35871         s_operands.
35872         * config/s390/s390.md: Add the s_operand checks formerly in
35873         s390_split_ok_p to various splitters where they are still
35874         required.
35875         * config/s390/vector.md ("mov<mode>" V_128): Add GPR alternatives
35876         for 128 bit vectors.  Plus two splitters.
35878 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35880         * config/s390/s390.md: Rename the cpu facilty vec to vx throughout
35881         the file.
35883 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35885         PR target/79893
35886         * config/s390/s390-c.c (s390_adjust_builtin_arglist): Issue an
35887         error if the boundary argument is not constant.
35889 2017-03-24  Jakub Jelinek  <jakub@redhat.com>
35891         PR rtl-optimization/80112
35892         * loop-doloop.c (doloop_condition_get): Don't check condition
35893         if cmp isn't SET with IF_THEN_ELSE src.
35895 2017-03-24  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
35897         PR tree-optimization/80158
35898         * gimple-ssa-strength-reduction.c (replace_mult_candidate): When
35899         replacing a candidate statement, also replace it for the
35900         candidate's alternate interpretation.
35901         (replace_rhs_if_not_dup): Likewise.
35902         (replace_one_candidate): Likewise.
35904 2017-03-24  Richard Biener  <rguenther@suse.de>
35906         PR tree-optimization/80167
35907         * graphite-isl-ast-to-gimple.c
35908         (translate_isl_ast_to_gimple::is_valid_rename): Handle default-defs
35909         properly.
35910         (translate_isl_ast_to_gimple::get_rename): Likewise.
35912 2017-03-23  Kelvin Nilsen  <kelvin@gcc.gnu.org>
35914         * config/rs6000/rs6000.c (rs6000_option_override_internal): Change
35915         handling of certain combinations of target options, including the
35916         combinations -mpower8-vector vs. -mno-vsx, -mpower9-vector vs.
35917         -mno-power8-vector, and -mpower9_dform vs. -mno-power9-vector.
35919 2017-03-23  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
35921         PR target/71436
35922         * config/arm/arm.md (*load_multiple): Add reload_completed to
35923         matching condition.
35925 2017-03-23  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
35926             Richard Biener  <rguenth@suse.de>
35928         PR tree-optimization/79908
35929         PR tree-optimization/80136
35930         * tree-stdarg.c (expand_ifn_va_arg_1): For a VA_ARG whose LHS has
35931         been cast away, gimplify_and_add suffices.
35933 2017-03-23  Markus Trippelsdorf  <markus@trippelsdorf.de>
35935         * tree-vrp.c (identify_jump_threads): Delete avail_exprs.
35937 2017-03-23  Richard Biener  <rguenther@suse.de>
35939         PR tree-optimization/80032
35940         * gimplify.c (gimple_push_cleanup): Forced unconditional
35941         cleanups still have to go to the conditional_cleanups
35942         sequence.
35944 2017-03-22  Jakub Jelinek  <jakub@redhat.com>
35946         PR tree-optimization/80072
35947         * tree-ssa-reassoc.c (struct operand_entry): Change id field type
35948         to unsigned int.
35949         (next_operand_entry_id): Change type to unsigned int.
35950         (sort_by_operand_rank): Make sure to return the right return value
35951         even if unsigned fields are bigger than INT_MAX.
35952         (struct oecount): Change cnt and id type to unsigned int.
35953         (oecount_hasher::equal): Formatting fix.
35954         (oecount_cmp): Make sure to return the right return value
35955         even if unsigned fields are bigger than INT_MAX.
35956         (undistribute_ops_list): Change next_oecount_id type to unsigned int.
35958         PR c++/80129
35959         * gimplify.c (gimplify_modify_expr_rhs) <case COND_EXPR>: Clear
35960         TREE_READONLY on result if writing it more than once.
35962         PR sanitizer/80110
35963         * doc/invoke.texi (-fsanitize=thread): Document that with
35964         -fnon-call-exceptions atomics are not able to throw
35965         exceptions.
35967         PR sanitizer/80110
35968         * tsan.c: Include tree-eh.h.
35969         (instrument_builtin_call): Call maybe_clean_eh_stmt or
35970         maybe_clean_or_replace_eh_stmt where needed.
35971         (instrument_memory_accesses): Add cfg_changed argument.
35972         Call gimple_purge_dead_eh_edges on each block and set *cfg_changed
35973         if it returned true.
35974         (tsan_pass): Adjust caller.  Return TODO_cleanup_cfg if cfg_changed.
35976         PR rtl-optimization/63191
35977         * config/i386/i386.c (ix86_delegitimize_address): Turn into small
35978         wrapper function, moved the whole old content into ...
35979         (ix86_delegitimize_address_1): ... this.  New inline function.
35980         (ix86_find_base_term): Use ix86_delegitimize_address_1 with
35981         true as last argument instead of ix86_delegitimize_address.
35983 2017-03-22  Wilco Dijkstra  <wdijkstr@arm.com>
35985         * config/aarch64/aarch64.c (generic_branch_cost): Copy
35986         cortexa57_branch_cost.
35988 2017-03-22  Wilco Dijkstra  <wdijkstr@arm.com>
35990         * config/aarch64/aarch64.c (generic_tunings): Add AES fusion.
35992 2017-03-21  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
35994         PR target/80123
35995         * doc/md.texi (Constraints): Document wA constraint.
35996         * config/rs6000/constraints.md (wA): New.
35997         * config/rs6000/rs6000.c (rs6000_debug_reg_global): Add wA reg_class.
35998         (rs6000_init_hard_regno_mode_ok): Init wA constraint.
35999         * config/rs6000/rs6000.h (RS6000_CONSTRAINT_wA): New.
36000         * config/rs6000/vsx.md (vsx_splat_<mode>): Use wA constraint.
36002 2017-03-22  Cesar Philippidis  <cesar@codesourcery.com>
36004         PR c++/80029
36005         * gimplify.c (is_oacc_declared): New function.
36006         (oacc_default_clause): Use it to set default flags for acc declared
36007         variables inside parallel regions.
36008         (gimplify_scan_omp_clauses): Strip firstprivate pointers for acc
36009         declared variables.
36010         (gimplify_oacc_declare): Gimplify the declare clauses.  Add the
36011         declare attribute to any decl as necessary.
36013 2017-03-22  Thomas Preud'homme  <thomas.preudhomme@arm.com>
36015         PR target/80082
36016         * config/arm/arm-isa.h (isa_bit_lpae): New feature bit.
36017         (ISA_ARMv7ve): Add isa_bit_lpae to the definition.
36018         * config/arm/arm-protos.h (arm_arch7ve): Rename into ...
36019         (arm_arch_lpae): This.
36020         * config/arm/arm.c (arm_arch7ve): Rename into ...
36021         (arm_arch_lpae): This.  Define it in term of isa_bit_lpae.
36022         * config/arm/arm.h (TARGET_HAVE_LPAE): Redefine in term of
36023         arm_arch_lpae.
36025 2017-03-22  Martin Liska  <mliska@suse.cz>
36027         PR target/79906
36028         * config/rs6000/rs6000.c (rs6000_inner_target_options): Show
36029         error message instead of an ICE.
36031 2017-03-21  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
36033         * doc/extend.texi (6.11 Additional Floating Types): Revise.
36035 2017-03-21  Kelvin Nilsen  <kelvin@gcc.gnu.org>
36037         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Add
36038         comments.
36039         * config/rs6000/rs6000.c (rs6000_option_override_internal): Add
36040         comments.
36042 2017-03-21  Martin Sebor  <msebor@redhat.com>
36044         * doc/extend.texi: Use "cannot" instead of "can't."
36045         * doc/hostconfig.texi: Same.
36046         * doc/install.texi: Same.
36047         * doc/invoke.texi: Same.
36048         * doc/loop.texi: Same.
36049         * doc/md.texi: Same.
36050         * doc/objc.texi: Same.
36051         * doc/rtl.texi: Same.
36052         * doc/tm.texi: Same.
36053         * doc/tm.texi.in: Same.
36054         * doc/trouble.texi: Same.
36056 2017-03-21  Alexandre Oliva  <aoliva@redhat.com>
36058         PR debug/63238
36059         * dwarf2out.c (struct checksum_attributes): Add at_alignment.
36060         (collect_checksum_attributes): Set it.
36061         (die_checksum_ordered): Use it.
36063 2017-03-21  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
36065         PR tree-optimization/79908
36066         * tree-stdarg.c (expand_ifn_va_arg_1): Revert the following
36067         change: For a VA_ARG whose LHS has been cast away, use
36068         force_gimple_operand to construct the side effects.
36070 2017-03-21  David Malcolm  <dmalcolm@redhat.com>
36072         PR translation/80001
36073         * omp-offload.c (oacc_loop_fixed_partitions): Make diagnostics
36074         more amenable to translation.
36075         (oacc_loop_auto_partitions): Likewise.
36077 2017-03-21  Marek Polacek  <polacek@redhat.com>
36078             Martin Sebor  <msebor@redhat.com>
36080         PR tree-optimization/80109
36081         * gimple-ssa-warn-alloca.c (alloca_call_type): Only call get_range_info
36082         on INTEGRAL_TYPE_P.
36084 2017-03-21  Jakub Jelinek  <jakub@redhat.com>
36085             Segher Boessenkool  <segher@kernel.crashing.org>
36087         PR target/80125
36088         * combine.c (can_combine_p): Revert the 2017-03-20 change, only
36089         check reg_used_between_p between insn and one of succ or succ2
36090         depending on if succ is artificial insn not inserted into insn
36091         stream.
36093 2017-03-21  Martin Liska  <mliska@suse.cz>
36095         PR gcov-profile/80081
36096         * Makefile.in: Add gcov-dump and fix installation of gcov-tool.
36097         * doc/gcc.texi: Include gcov-dump stuff.
36098         * doc/gcov-dump.texi: New file.
36100 2017-03-21  Toma Tabacu  <toma.tabacu@imgtec.com>
36102         PR rtl-optimization/79150
36103         * config/mips/mips.c (mips_block_move_loop): Emit a NOP after the
36104         conditional jump, if the jump is the last insn of the loop.
36106 2017-03-21  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
36107             Richard Biener  <rguenth@suse.de>
36109         PR tree-optimization/79908
36110         * tree-stdarg.c (expand_ifn_va_arg_1): For a VA_ARG whose LHS has
36111         been cast away, use force_gimple_operand to construct the side
36112         effects.
36114 2017-03-21  Martin Liska  <mliska@suse.cz>
36116         PR libfortran/79956
36117         * simplify-rtx.c (simplify_immed_subreg): Initialize a variable
36118         to NULL.
36120 2017-03-21  Brad Spengler <spender@grsecurity.net>
36122         PR plugins/80094
36123         * plugin.c (htab_hash_plugin): New function.
36124         (add_new_plugin): Use it and adjust.
36125         (parse_plugin_arg_opt): Adjust.
36126         (init_one_plugin): Likewise.
36128 2017-03-21  Richard Biener  <rguenther@suse.de>
36130         PR tree-optimization/80032
36131         * gimplify.c (gimple_push_cleanup): Add force_uncond parameter,
36132         if set force the cleanup to happen unconditionally.
36133         (gimplify_target_expr): Push inserted clobbers with force_uncond
36134         to avoid them being removed by control-dependent DCE.
36136 2017-03-21  Richard Biener  <rguenther@suse.de>
36138         PR tree-optimization/80122
36139         * tree-inline.c (copy_bb): Do not expans va-arg packs or
36140         va_arg_pack_len when the inlined call stmt requires pack
36141         expansion itself.
36142         * tree-inline.h (struct copy_body_data): Make call_stmt a gcall *.
36144 2017-03-21  Jakub Jelinek  <jakub@redhat.com>
36146         PR sanitizer/78158
36147         * tsan.c (instrument_builtin_call): If the memory model argument
36148         is not a constant, assume it is valid.
36150         PR c/67338
36151         * fold-const.c (round_up_loc): Negate divisor in unsigned type to
36152         avoid UB.
36154 2017-03-20  Segher Boessenkool  <segher@kernel.crashing.org>
36156         PR rtl-optimization/79910
36157         * combine.c (can_combine_p): Do not allow combining an I0 or I1
36158         if its dest is used by an insn before I2 (other than the combined
36159         insns themselves, which are properly handled already).
36161 2017-03-20  Segher Boessenkool  <segher@kernel.crashing.org>
36163         Revert:
36164         2017-03-17  Bernd Schmidt  <bschmidt@redhat.com>
36166         * combine.c (record_used_regs): New static function.
36167         (try_combine): Handle situations where there is an additional
36168         instruction between I2 and I3 which needs to have a LOG_LINK
36169         updated.
36171         Revert:
36172         2017-03-17  Jim Wilson  <jim.wilson@linaro.org>
36174         * combine.c (try_combine): Delete redundant i1 test.  Call
36175         prev_nonnote_nondebug_insn instead of prev_nonnote_insn.
36177 2017-03-20  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
36179         PR target/80083
36180         * config/rs6000/rs6000.md (*movsi_internal1): Fix constraints for
36181         alternatives 13/14.
36183 2017-03-20  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
36185         PR tree-optimization/80054
36186         * gimple-ssa-strength-reduction.c (all_phi_incrs_profitable): Fail
36187         the optimization if a PHI or any of its arguments is not dominated
36188         by the candidate's basis.  Use gphi* rather than gimple* as
36189         appropriate.
36190         (replace_profitable_candidates): Clean up a gimple* variable that
36191         should be a gphi* variable.
36193 2017-03-20  Martin Sebor  <msebor@redhat.com>
36195         PR c++/52477
36196         * doc/extend.texi (attribute constructor): Document present limitation.
36198 2017-03-20  Kelvin Nilsen  <kelvin@gcc.gnu.org>
36200         PR target/79963
36201         * config/rs6000/altivec.h (vec_all_ne): Under __cplusplus__ and
36202         __POWER9_VECTOR__ #ifdef control, change template definition to
36203         use Power9-specific built-in function.
36204         (vec_any_eq): Likewise.
36205         * config/rs6000/vector.md (vector_ae_v2di_p): Change the flag used
36206         to control outcomes from this test.
36207         (vector_ae_<mode>p): For VEC_F modes, likewise.
36209 2017-03-20  Ian Lance Taylor  <iant@google.com>
36211         * config/i386/i386.c (ix86_function_regparm): Save an extra
36212         register for -fsplit-stack with DECL_STATIC_CHAIN.
36214 2017-03-17  Palmer Dabbelt  <palmer@dabbelt.com>
36216         PR target/79912
36217         * config/riscv/riscv.c (riscv_preferred_reload_class): Remove.
36218         (TARGET_PREFERRED_RELOAD_CLASS): Likewise.
36220 2017-03-17  Palmer Dabbelt  <palmer@dabbelt.com>
36222         * config/riscv/riscv.c (riscv_print_operand): Use "fence
36223         iorw,ow".
36224         * config/riscv/sync.mc (mem_thread_fence_1): Use "fence
36225         iorw,iorw".
36227 2017-03-20  Marek Polacek  <polacek@redhat.com>
36229         PR sanitizer/80063
36230         * asan.c (DEF_SANITIZER_BUILTIN): Use do { } while (0).
36232 2017-03-20  Richard Biener  <rguenther@suse.de>
36234         PR tree-optimization/80113
36235         * graphite-isl-ast-to-gimple.c (copy_loop_phi_nodes): Do not
36236         allocate extra SSA name for PHI def.
36237         (add_close_phis_to_outer_loops): Likewise.
36238         (add_close_phis_to_merge_points): Likewise.
36239         (copy_loop_close_phi_args): Likewise.
36240         (copy_cond_phi_nodes): Likewise.
36242 2017-03-20  Martin Liska  <mliska@suse.cz>
36244         PR middle-end/79753
36245         * tree-chkp.c (chkp_build_returned_bound): Do not build
36246         returned bounds for a LHS that's not a BOUNDED_P type.
36248 2017-03-20  Martin Liska  <mliska@suse.cz>
36250         PR target/79769
36251         PR target/79770
36252         * tree-chkp.c (chkp_find_bounds_1): Handle REAL_CST,
36253         COMPLEX_CST and VECTOR_CST.
36255 2017-03-20  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
36257         PR target/78857
36258         * config/s390/s390.md ("cmp<mode>_ccs_0"): Add a clobber of the
36259         target operand.  A new splitter adds the clobber statement in case
36260         the target operand is dead anyway.
36262 2017-03-19  Gerald Pfeifer  <gerald@pfeifer.com>
36264         * doc/install.texi (Specific) <sparc-*-linux*>: No longer refer
36265         to age-old versions of binutils and glibc.
36267 2017-03-18  Segher Boessenkool  <segher@kernel.crashing.org>
36269         * doc/contrib.texi (Contributors): Remove duplicate entry for myself.
36271 2017-03-18  Gerald Pfeifer  <gerald@pfeifer.com>
36273         * doc/contrib.texi (Contributors): Add Segher Boessenkool.
36275 2017-03-18  Gerald Pfeifer  <gerald@pfeifer.com>
36277         * doc/install.texi (Specific) <arm-*-eabi>: Remove old
36278         requirement for binutils 2.13.
36280 2017-03-17  Jim Wilson  <jim.wilson@linaro.org>
36282         * combine.c (try_combine): Delete redundant i1 test.  Call
36283         prev_nonnote_nondebug_insn instead of prev_nonnote_insn.
36285 2017-03-17  Palmer Dabbelt  <palmer@dabbelt.com
36287         * doc/install.texi (Specific) <riscv32-*-elf>: Add riscv32-*-elf,
36288         riscv32-*-linux, riscv64-*-elf, riscv64-*-linux to the table of
36289         contents.
36290         <riscv64-*-elf>: Re-arrange section
36291         <riscv32-*-elf>: Add a note about requiring binutils 2.28.
36292         <riscv32-*-linux>: Likewise.
36293         <riscv64-*-elf>: Likewise
36294         <riscv64-*-linux>: Likewise.
36296 2017-03-17  Richard Earnshaw  <rearnsha@arm.com>
36298         PR target/80052
36299         * aarch64.opt(verbose-cost-dump): Fix typo.
36301 2017-03-17  Pat Haugen  <pthaugen@us.ibm.com>
36303         PR target/79951
36304         * config/rs6000/rs6000.md (copysign<mode>3_fcpsgn): Test
36305         for VECTOR_UNIT_VSX_P (<MODE>mode) too.
36307 2017-03-17  Bernd Schmidt  <bschmidt@redhat.com>
36309         * reload.c (find_reloads): When reloading a nonoffsettable address,
36310         use RELOAD_OTHER for it and its address reloads.
36312         PR rtl-optimization/79910
36313         * combine.c (record_used_regs): New static function.
36314         (try_combine): Handle situations where there is an additional
36315         instruction between I2 and I3 which needs to have a LOG_LINK
36316         updated.
36318 2017-03-17  Jeff Law  <law@redhat.com>
36320         PR tree-optimization/71437
36321         * tree-vrp.c (simplify_stmt_for_jump_threading): Lookup the
36322         conditional in the hash table first.
36323         (vrp_dom_walker::before_dom_children): Extract condition from
36324         ASSERT_EXPR.  Record condition, its inverion and any implied
36325         conditions as well.
36327 2017-03-17  Marek Polacek  <polacek@redhat.com>
36328             Markus Trippelsdorf  <markus@trippelsdorf.de>
36330         PR tree-optimization/80079
36331         * gimple-ssa-store-merging.c (class pass_store_merging): Initialize
36332         m_stores_head.
36334 2017-03-17  Richard Biener  <rguenther@suse.de>
36336         PR middle-end/80075
36337         * tree-eh.c (stmt_could_throw_1_p): Only handle gimple assigns.
36338         Properly verify the LHS before the RHS possibly claims to be
36339         handled.
36340         (stmt_could_throw_p): Hande gimple conds fully here.  Clobbers
36341         do not throw.
36343 2017-03-17  Martin Jambor  <mjambor@suse.cz>
36345         * doc/invoke.texi (Option Options): Include -fipa-vrp in the list.
36346         (List of -O2 options): Likewise.
36347         (-fipa-bit-cp): Replace "ipa" with "interprocedural."
36348         (-fipa-vrp) New.
36350 2017-03-17  Tom de Vries  <tom@codesourcery.com>
36352         * gcov-dump.c (print_usage): Print bug_report_url.
36354 2017-03-17  Richard Biener  <rguenther@suse.de>
36356         PR middle-end/80050
36357         * genmatch.c (parser::next): Remove pointless check for CPP_EOF.
36358         (parser::peek): Likewise.
36360 2017-03-17  Richard Biener  <rguenther@suse.de>
36362         PR tree-optimization/80048
36363         * sese.c (free_sese_info): Properly release rename_map and
36364         copied_bb_map elements.
36366 2017-03-16  Alexandre Oliva  <aoliva@redhat.com>
36368         * gimple-ssa-store-merging.c (struct imm_store_chain_info):
36369         Add linked-list forward and backlinks.  Insert on
36370         construction, remove on destruction.
36371         (class pass_store_merging): Add m_stores_head field.
36372         (pass_store_merging::terminate_and_process_all_chains):
36373         Iterate over m_stores_head list.
36374         (pass_store_merging::terminate_all_aliasing_chains):
36375         Likewise.
36376         (pass_store_merging::execute): Check for debug stmts first.
36377         Push new chains onto the m_stores_head stack.
36379 2017-03-16  Michael Meissner  <meissner@linux.vnet.ibm.com>
36381         PR target/71294
36382         * config/rs6000/vsx.md (vsx_splat_<mode>, VSX_D iterator): Allow a
36383         SPLAT operation on ISA 2.07 64-bit systems that have direct move,
36384         but no MTVSRDD support, by doing MTVSRD and XXPERMDI.
36386 2017-03-16  Jeff Law  <law@redhat.com>
36388         PR tree-optimization/71437
36389         * tree-ssa-dom.c (dom_opt_dom_walker): Remove thread_across_edge
36390         member function.  Implementation moved into after_dom_children
36391         member function and into the threader's thread_outgoing_edges
36392         function.
36393         (dom_opt_dom_walker::after_dom_children): Simplify by moving
36394         some code into new thread_outgoing_edges.
36395         * tree-ssa-threadedge.c (thread_across_edge): Make static and simplify
36396         definition.  Simplify marker handling (do it here).   Assume we always
36397         have the available expression and the const/copies tables.
36398         (thread_outgoing_edges): New function extracted from tree-ssa-dom.c
36399         and tree-vrp.c
36400         * tree-ssa-threadedge.h (thread_outgoing_edges): Declare.
36401         * tree-vrp.c (equiv_stack): No longer file scoped.
36402         (vrp_dom_walker): New class.
36403         (vrp_dom_walker::before_dom_children): New member function.
36404         (vrp_dom_walker::after_dom_children): Likewise.
36405         (identify_jump_threads):  Setup domwalker.  Use it rather than
36406         walking edges in a random order by hand.  Simplify setup/finalization.
36407         (finalize_jump_threads): Remove.
36408         (vrp_finalize): Do not call identify_jump_threads here.
36409         (execute_vrp): Do it here instead and call thread_through_all_blocks
36410         here too.
36412         PR tree-optimization/71437
36413         * tree-ssa-dom.c (pfn_simplify): Add basic_block argument.  All
36414         callers changed.
36415         (simplify_stmt_for_jump_threading): Add basic_block argument.  All
36416         callers changed.
36417         (lhs_of_dominating_assert): Moved from here into tree-vrp.c.
36418         (dom_opt_dom_walker::thread_across_edge): Remove
36419         handle_dominating_asserts argument.  All callers changed.
36420         (record_temporary_equivalences_from_stmts_at_dest): Corresponding
36421         changes.  Remove calls to lhs_of_dominating_assert.  Other
36422         uses of handle_dominating_asserts turn into unconditional code
36423         (simplify_control_stmt_condition_1): Likewise.
36424         (simplify_control_stmt_condition): Likewise.
36425         (thread_through_normal_block, thread_across_edge): Likewise.
36426         * tree-ssa-threadedge.h (thread_across_edge): Corresponding changes.
36427         * tree-vrp.c (lhs_of_dominating_assert): Move here.  Return original
36428         object if it is not an SSA_NAME.
36429         (simplify_stmt_for_jump_threading): Call lhs_of_dominating_assert
36430         before calling into the VRP specific simplifiers.
36431         (identify_jump_threads): Remove handle_dominating_asserts
36432         argument.
36434 2017-03-16  Jakub Jelinek  <jakub@redhat.com>
36436         PR fortran/79886
36437         * tree-diagnostic.c (default_tree_printer): No longer static.
36438         * tree-diagnostic.h (default_tree_printer): New prototype.
36440 2017-03-16  Tamar Christina  <tamar.christina@arm.com>
36442         * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>)
36443         Change ins into fmov.
36445 2017-03-16  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
36447         * config/aarch64/iterators.md (h_con): Return "x" for V4HF and V8HF.
36448         * config/aarch64/aarch64-simd.md (*aarch64_fma4_elt_from_dup<mode>):
36449         Use h_con constraint for operand 1.
36450         (*aarch64_fnma4_elt_from_dup<mode>): Likewise.
36451         (*aarch64_mulx_elt_from_dup<mode>): Likewise for operand 2.
36453 2017-03-15  Jeff Law  <law@redhat.com>
36455         PR tree-optimization/71437
36456         * tree-ssa-dom.c (derive_equivalences_from_bit_ior): New function.
36457         (record_temporary_equivalences): Use it.
36459         PR tree-optimization/71437
36460         * tree-ssa-dom.c (struct cond_equivalence): Moved from here into
36461         tree-ssa-scopedtables.
36462         (lookup_avail_expr, build_and_record_new_cond): Likewise.
36463         (record_conditions, record_cond, vuse_eq): Likewise.
36464         (record_edge_info): Adjust to API tweak of record_conditions.
36465         (simplify_stmt_for_jump_threading): Similarly for lookup_avail_expr.
36466         (record_temporary_equivalences, optimize_stmt): Likewise.
36467         (eliminate_redundant_computations): Likewise.
36468         (record_equivalences_from_stmt): Likewise.
36469         * tree-ssa-scopedtables.c: Include options.h and params.h.
36470         (vuse_eq): New function, moved from tree-ssa-dom.c
36471         (build_and_record_new_cond): Likewise.
36472         (record_conditions): Likewise.  Accept vector of conditions rather
36473         than edge_equivalence structure for first argument.
36474         for the first argument.
36475         (avail_exprs_stack::lookup_avail_expr): New member function, moved
36476         from tree-ssa-dom.c.
36477         (avail_exprs_stack::record_cond): Likewise.
36478         * tree-ssa-scopedtables.h (struct cond_equivalence): Moved here
36479         from tree-ssa-dom.c.
36480         (avail_exprs_stack): Add new member functions lookup_avail_expr
36481         and record_cond.
36482         (record_conditions): Declare.
36484 2017-03-15  Vladimir Makarov  <vmakarov@redhat.com>
36486         PR target/80017
36487         * lra-constraints.c (process_alt_operands): Increase reject for
36488         reloading an input/output operand.
36490 2017-03-15  Michael Meissner  <meissner@linux.vnet.ibm.com>
36492         PR target/79038
36493         * config/rs6000/rs6000.md (float<QHI:mode><IEEE128:mode>2): Define
36494         insns to convert from signed/unsigned char/short to IEEE 128-bit
36495         floating point.
36496         (floatuns<QHI:mode><IEEE128:mode>2): Likewise.
36498 2017-03-15  Uros Bizjak  <ubizjak@gmail.com>
36500         PR target/80019
36501         * config/i386/i386.c (ix86_vector_duplicate_value): Create
36502         subreg of inner mode for values already in registers.
36504 2017-03-15  Bernd Schmidt  <bschmidt@redhat.com>
36506         * config/c6x/c6x.c (hwloop_optimize): Handle case where the old
36507         iteration reg is used after the loop.
36509 2017-03-14  Martin Sebor  <msebor@redhat.com>
36511         PR tree-optimization/79800
36512         * gimple-ssa-sprintf.c (format_floating: Add argument.  Handle
36513         precision in negative-positive range.
36514         (format_floating): Call non-const overload with adjusted precision.
36516 2017-03-14  Michael Meissner  <meissner@linux.vnet.ibm.com>
36518         PR target/79947
36519         * config/rs6000/rs6000.h (TARGET_FRSQRTES): Add check for
36520         -mpowerpc-gfxopt.
36522 2017-03-14  Martin Sebor  <msebor@redhat.com>
36524         PR middle-end/80020
36525         * builtin-attrs.def (ATTR_ALLOC_SIZE_2_NOTHROW_LIST): New macro.
36526         * builtins.def (aligned_alloc): Use it.
36528         PR c/79936
36529         * Makefile.in (GTFILES): Add calls.c.
36530         * calls.c: Include "gt-calls.h".
36532 2017-03-14  Bernd Schmidt  <bschmidt@redhat.com>
36534         PR rtl-optimization/79728
36535         * regs.h (struct target_regs): New field
36536         x_contains_allocatable_regs_of_mode.
36537         (contains_allocatable_regs_of_mode): New macro.
36538         * reginfo.c (init_reg_sets_1): Initialize it, and change
36539         contains_reg_of_mode so it includes global regs as well.
36540         * reload.c (push_reload): Use contains_allocatable_regs_of_mode
36541         rather than contains_regs_of_mode.
36543 2017-03-14  Martin Liska  <mliska@suse.cz>
36545         * doc/invoke.texi: Document options that can't be combined with
36546         -fcheck-pointer-bounds.
36548 2017-03-14  Martin Liska  <mliska@suse.cz>
36550         PR middle-end/79831
36551         * doc/invoke.texi (-Wchkp): Document the option.
36553 2017-03-14  Martin Liska  <mliska@suse.cz>
36555         * Makefile.in: Install gcov-dump.
36557 2017-03-14  Martin Liska  <mliska@suse.cz>
36559         * multiple_target.c (expand_target_clones): Bail out for
36560         an invalid attribute.
36562 2017-03-14  Richard Biener  <rguenther@suse.de>
36564         * alias.c (struct alias_set_entry): Pack properly.
36565         * cfgloop.h (struct loop): Likewise.
36566         * cse.c (struct set): Likewise.
36567         * ipa-utils.c (struct searchc_env): Likewise.
36568         * loop-invariant.c (struct invariant): Likewise.
36569         * lra-remat.c (struct cand): Likewise.
36570         * recog.c (struct change_t): Likewise.
36571         * rtl.h (struct address_info): Likewise.
36572         * symbol-summary.h (function_summary): Likewise.
36573         * tree-loop-distribution.c (struct partition): Likewise.
36574         * tree-object-size.c (struct object_size_info): Likewise.
36575         * tree-ssa-loop-ivopts.c (struct cost_pair): Likewise.
36576         * tree-ssa-threadupdate.c (struct ssa_local_info_t): Likewise.
36577         * tree-vect-data-refs.c (struct _vect_peel_info): Likewise.
36578         * tree-vect-slp.c (struct _slp_oprnd_info): Likewise.
36579         * tree-vect-stmts.c (struct simd_call_arg_info): Likewise.
36580         * tree-vectorizer.h (struct _loop_vec_info): Likewise.
36581         (struct _stmt_vec_info): Likewise.
36583 2017-03-14  Martin Liska  <mliska@suse.cz>
36585         PR target/79892
36586         * multiple_target.c (create_dispatcher_calls): Check that
36587         a target can create a function dispatcher.
36589 2017-03-14  Martin Liska  <mliska@suse.cz>
36591         PR lto/66295
36592         * multiple_target.c (expand_target_clones): Drop local.local
36593         flag for default implementation.
36595 2017-03-14  Richard Biener  <rguenther@suse.de>
36597         PR tree-optimization/80030
36598         * tree-vect-stmts.c (vectorizable_store): Plug memleak.
36600 2017-03-13  Kito Cheng  <kito.cheng@gmail.com>
36602         * config/riscv/riscv.c (riscv_emit_float_compare>: Use
36603         gcc_fallthrough() instead of __attribute__((fallthrough));
36605 2017-03-13  Gerald Pfeifer  <gerald@pfeifer.com>
36607         * doc/gcc.texi: Remove "up" link to (DIR).
36608         * doc/gccint.texi: Ditto.
36610 2017-03-13  Gerald Pfeifer  <gerald@pfeifer.com>
36612         * doc/install.texi (Specific) <avr>: Remove reference to
36613         binutils 2.13.
36615 2017-03-13  Jeff Law  <law@redhat.com>
36617         * config/riscv/riscv.c (riscv_emit_float_compare): Use fallthru
36618         attribute rather than comments.
36620         * config/pdp11/pdp11.md (movmemhi): Adjust operand numbers to
36621         match_scratch operand is highest.
36623 2017-03-13  Martin Liska  <mliska@suse.cz>
36625         PR middle-end/78339
36626         * ipa-pure-const.c (warn_function_noreturn): If the declarations
36627         is a CHKP clone, use original declaration.
36629 2017-03-13  Claudiu Zissulescu  <claziss@synopsys.com>
36631         * config/arc/arc.c (arc_init): Use multiplier whenever we have it.
36632         (arc_conditional_register_usage): Use a different allocation order
36633         when optimizing for size.
36634         * common/config/arc/arc-common.c (arc_option_optimization_table):
36635         Section anchors default on when optimizing for size.
36637 2017-03-13  Claudiu Zissulescu  <claziss@synopsys.com>
36639         * config/arc/arc.md (*tst_bitfield_tst): Fix pattern.
36641 2017-03-13  Claudiu Zissulescu  <claziss@synopsys.com>
36643         * config/arc/arc.c (arc_output_addsi): Emit code density adds.
36644         * config/arc/arc.md (cpu_facility): Add cd variant.
36645         (*movqi_insn): Add code density variant.
36646         (*movhi_insn): Likewise.
36647         (*movqi_insn): Likewise.
36648         (*addsi3_mixed): Likewise.
36649         (subsi3_insn): Likewise.
36651 2017-03-13  Claudiu Zissulescu  <claziss@synopsys.com>
36653         * config/arc/arc.md (movsi_cond_exec): Update constraint.
36655 2017-03-13  Claudiu Zissulescu  <claziss@synopsys.com>
36657         * config/arc/arc.c (arc_legitimize_pic_address): Handle PIC
36658         expressions with MINUS and UNARY ops.
36660 2017-03-13  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
36662         PR target/79911
36663         * config/arm/neon.md (vec_sel_widen_ssum_lo<VQI:mode><VW:mode>3):
36664         Rename to...
36665         (vec_sel_widen_ssum_lo<mode><V_half>3): ... This. Avoid mismatch
36666         between vec_select and vector argument.
36667         (vec_sel_widen_ssum_hi<VQI:mode><VW:mode>3): Rename to...
36668         (vec_sel_widen_ssum_hi<mode><V_half>3): ... This. Likewise.
36669         (vec_sel_widen_usum_lo<VQI:mode><VW:mode>3): Rename to...
36670         (vec_sel_widen_usum_lo<mode><V_half>3): ... This.
36671         (vec_sel_widen_usum_hi<VQI:mode><VW:mode>3): Rename to...
36672         (vec_sel_widen_usum_hi<mode><V_half>3): ... This.
36674 2017-03-13  Richard Biener  <rguenther@suse.de>
36676         PR other/79991
36677         * params.def (vect-max-peeling-for-alignment): Fix typo.
36679 2017-03-12  Gerald Pfeifer  <gerald@pfeifer.com>
36681         * doc/install.texi (Specific) <mips-*-*>: Remove description of
36682         issue that only occurred with binutils below 2.18.
36684 2017-03-12  Gerald Pfeifer  <gerald@pfeifer.com>
36686         * doc/install.texi (Specific) <cris-axis-elf>: No longer
36687         refer to binutils 2.11/2.12 minimum.
36689 2017-03-12  Gerald Pfeifer  <gerald@pfeifer.com>
36691         * doc/install.texi (Specific) <powerpc-*-*>: Remove link to
36692         ftp.kernel.org and simplify binutils requirement.
36694 2017-03-11  Gerald Pfeifer  <gerald@pfeifer.com>
36696         * doc/invoke.texi (Warning Options): Fix spelling of link-time
36697         optimization.
36698         (Optimize Options): Ditto.  Also remove redundancy.
36700 2017-03-10  David Malcolm  <dmalcolm@redhat.com>
36702         PR translation/79848
36703         * ipa-devirt.c (warn_types_mismatch): Simplify uses of "%<%s%>" to
36704         "%qs".
36705         * ipa-pure-const.c (suggest_attribute): Likewise.  Convert _
36706         to G_ to avoid double translation.
36708 2017-03-10  David Malcolm  <dmalcolm@redhat.com>
36710         PR translation/79923
36711         * auto-profile.c (get_combined_location): Convert leading
36712         character of diagnostics to lower case and remove trailing period.
36713         (read_profile): Likewise for various diagnostics.
36714         * config/arm/arm.c (arm_option_override): Remove trailing period
36715         from various diagnostics.
36716         * config/msp430/msp430.c (msp430_expand_delay_cycles): Likewise.
36717         (msp430_expand_delay_cycles): Likewise.
36719 2017-03-10  David Malcolm  <dmalcolm@redhat.com>
36721         PR target/79925
36722         * config/aarch64/aarch64.c (aarch64_validate_mcpu): Quote the
36723         full command-line argument, rather than just "str".
36724         (aarch64_validate_march): Likewise.
36725         (aarch64_validate_mtune): Likewise.
36727 2017-03-10  Bernd Schmidt  <bschmidt@redhat.com>
36729         PR rtl-optimization/78911
36730         * lra-assigns.c (must_not_spill_p): New function.
36731         (spill_for): Use it.
36733 2017-03-10  Jakub Jelinek  <jakub@redhat.com>
36735         PR tree-optimization/79981
36736         * tree-vrp.c (extract_range_basic): Handle IMAGPART_EXPR of
36737         ATOMIC_COMPARE_EXCHANGE ifn result.
36738         (stmt_interesting_for_vrp, vrp_visit_stmt): Handle
36739         IFN_ATOMIC_COMPARE_EXCHANGE.
36741 2017-03-10  David Malcolm  <dmalcolm@redhat.com>
36743         PR driver/79875
36744         * opts.c (parse_sanitizer_options): Add missing question mark to
36745         "did you mean" message.
36747 2017-03-10  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
36749         * config/rs6000/rs6000-builtin.def (VMULEUB_UNS): Remove orphaned
36750         built-in.
36751         (VMULEUH_UNS): Likewise.
36752         (VMULOUB_UNS): Likewise.
36753         (VMULOUH_UNS): Likewise.
36754         * config/rs6000/rs6000.c (builtin_function_type): Remove
36755         references to ALTIVEC_BUILTIN_VMUL[EO]U[BH]_UNS.
36757 2017-03-10  David Malcolm  <dmalcolm@redhat.com>
36759         PR bootstrap/79952
36760         * read-rtl-function.c (function_reader::read_rtx_operand): Update
36761         x with result of extra_parsing_for_operand_code_0.
36762         (function_reader::extra_parsing_for_operand_code_0): Convert
36763         return type from void to rtx, returning x.  When reading
36764         SYMBOL_REF with SYMBOL_FLAG_HAS_BLOCK_INFO, reallocate x to the
36765         larger size containing struct block_symbol.
36767 2017-03-10  Segher Boessenkool  <segher@kernel.crashing.org>
36769         * config/rs6000/rs6000.c (rs6000_option_override_internal): Disallow
36770         -mfloat128-hardware without -m64.
36772 2017-03-10  Will Schmidt <will_schmidt@vnet.ibm.com>
36774         PR target/79941
36775         * config/rs6000/rs6000.c (builtin_function_type): Add VMUL*U[HB]
36776         entries to the case statement that marks unsigned arguments to
36777         overloaded functions.
36779 2017-03-10  Kelvin Nilsen  <kelvin@gcc.gnu.org>
36781         * config/rs6000/rs6000.c (rs6000_option_override_internal): Fix
36782         two typographic errors in the handling of TARGET_UPPER_REGS_DI.
36784 2017-03-10  Pat Haugen  <pthaugen@us.ibm.com>
36786         PR target/79907
36787         * config/rs6000/rs6000.c (rs6000_init_hard_regno_mode_ok): Test
36788         TARGET_UPPER_REGS_DI when setting 'wi' constraint regclass.
36790 2017-03-10  Martin Liska  <mliska@suse.cz>
36792         PR target/65705
36793         PR target/69804
36794         * toplev.c (process_options): Enable MPX with LSAN and UBSAN.
36795         * tree-chkp.c (chkp_walk_pointer_assignments): Verify that
36796         FIELD != NULL.
36798 2017-03-10  Olivier Hainque  <hainque@adacore.com>
36800         * tree-switch-conversion (array_value_type): Start by resetting
36801         candidate type to it's main variant.
36803 2017-03-10  Jakub Jelinek  <jakub@redhat.com>
36805         PR rtl-optimization/79909
36806         * combine.c (try_combine): Use simplify_replace_rtx on individual
36807         CALL_INSN_FUNCTION_USAGE elements instead of replace_rtx on copy_rtx
36808         of the whole CALL_INSN_FUNCTION_USAGE.
36810         PR tree-optimization/79972
36811         * gimple-ssa-warn-alloca.c (alloca_call_type): Only call
36812         get_range_info on SSA_NAMEs.  Formatting fixes.
36814 2017-03-10  Richard Biener  <rguenther@suse.de>
36815             Jakub Jelinek  <jakub@redhat.com>
36817         PR tree-optimization/77975
36818         * tree-ssa-loop-niter.c (get_base_for): Allow phi argument from latch
36819         edge to be constant.
36820         (get_val_for): For constant x return it.  Formatting fix.
36821         (loop_niter_by_eval): Avoid pointless looping if the next iteration
36822         would use the same bases as the current one.
36824 2017-03-09  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
36826         * config/rs6000/rs6000.c (rs6000_gen_le_vsx_permute): Use rotate
36827         instead of vec_select for V1TImode.
36828         * conifg/rs6000/vsx.md (VSX_LE): Remove mode iterator that is no
36829         longer needed.
36830         (VSX_LE_128): Add V1TI to this mode iterator.
36831         (*vsx_le_perm_load_<mode>): Change to use VSX_D mode iterator.
36832         (*vsx_le_perm_store_<mode>): Likewise.
36833         (pre-reload splitter for VSX stores): Likewise.
36834         (post-reload splitter for VSX stores): Likewise.
36835         (*vsx_xxpermdi2_le_<mode>): Likewise.
36836         (*vsx_lxvd2x2_le_<mode>): Likewise.
36837         (*vsx_stxvd2x2_le_<mode>): Likewise.
36839 2017-03-09  Michael Eager  <eager@eagercon.com>
36841         Correct failures with --enable-checking=yes,rtl.
36843         * config/microblaze/microblaze.c (microblaze_expand_shift):
36844         Replace GET_CODE test with CONST_INT_P and INTVAL test with
36845         test for const0_rtx.
36846         * config/microblaze/microblaze.md (ashlsi3_byone, ashrsi3_byone,
36847         lshrsi3_byone): Replace INTVAL with test for const1_rtx.
36849 2017-03-09  Richard Biener  <rguenther@suse.de>
36851         PR tree-optimization/79977
36852         * graphite-scop-detection.c (scop_detection::merge_sese):
36853         Handle the case of extra exits to blocks dominating the entry.
36855 2017-03-09  Toma Tabacu  <toma.tabacu@imgtec.com>
36857         * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
36858         Document rdynamic.
36860 2017-03-09  Vladimir Makarov  <vmakarov@redhat.com>
36862         PR rtl-optimization/79949
36863         * lra-constraints.c (process_alt_operands): Check memory when
36864         trying to predict a cycle.  Print about the overall increase.
36866 2017-03-09  Richard Biener  <rguenther@suse.de>
36868         PR middle-end/79971
36869         * gimple-expr.c (useless_type_conversion_p): Preserve
36870         TYPE_SATURATING for fixed-point types.
36872 2017-03-09  Richard Biener  <rguenther@suse.de>
36874         PR ipa/79970
36875         * ipa-prop.c (ipa_modify_formal_parameters): Avoid changing
36876         alignment of BLKmode params.
36878 2017-03-09  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
36880         PR target/79913
36881         * config/aarch64/iterators.md (VALL_F16_NO_V2Q): New mode iterator.
36882         (VALL_NO_V2Q): Likewise.
36883         (VDQF_DF): Delete.
36884         * config/aarch64/aarch64-simd.md
36885         (aarch64_dup_lane_<vswap_width_name><mode>): Use VALL_F16_NO_V2Q
36886         iterator.
36887         (*aarch64_simd_vec_copy_lane_<vswap_width_name><mode>): Use
36888         VALL_NO_V2Q mode iterator.
36889         (*aarch64_vgetfmulx<mode>): Use VDQF iterator.
36891 2017-03-09  Martin Liska  <mliska@suse.cz>
36893         PR tree-optimization/79631
36894         * tree-chkp-opt.c (chkp_is_constant_addr): Call
36895         tree_int_cst_sign_bit just for INTEGER constants.
36897 2017-03-09  Martin Liska  <mliska@suse.cz>
36899         PR target/65705
36900         PR target/69804
36901         * toplev.c (process_options): Disable -fcheck-pointer-bounds with
36902         sanitizers.
36904 2017-03-09  Marek Polacek  <polacek@redhat.com>
36906         PR c++/79672
36907         * tree.c (inchash::add_expr): Handle TREE_VEC.
36909 2017-03-09  Martin Liska  <mliska@suse.cz>
36911         PR ipa/79764
36912         (chkp_narrow_size_and_offset): New function.
36913         (chkp_parse_array_and_component_ref): Support BIT_FIELD_REF.
36914         (void chkp_parse_bit_field_ref): New function.
36915         (chkp_make_addressed_object_bounds): Add case for BIT_FIELD_REF.
36916         (chkp_process_stmt): Use chkp_parse_bit_field_ref.
36918 2017-03-09  Martin Liska  <mliska@suse.cz>
36920         PR ipa/79761
36921         * tree-chkp.c (chkp_get_bound_for_parm): Get bounds for a param.
36922         (chkp_find_bounds_1): Remove gcc_unreachable.
36924 2017-03-09  Jakub Jelinek  <jakub@redhat.com>
36926         PR sanitizer/79944
36927         * asan.c (get_mem_refs_of_builtin_call): For BUILT_IN_ATOMIC* and
36928         BUILT_IN_SYNC*, determine the access type from the size suffix and
36929         always build a MEM_REF with that type.  Handle forgotten
36930         BUILT_IN_SYNC_FETCH_AND_NAND_16 and BUILT_IN_SYNC_NAND_AND_FETCH_16.
36932         PR target/79932
36933         * config/i386/avx512vlintrin.h (_mm256_cmpge_epi32_mask,
36934         _mm256_cmpge_epi64_mask, _mm256_cmpge_epu32_mask,
36935         _mm256_cmpge_epu64_mask, _mm256_cmple_epi32_mask,
36936         _mm256_cmple_epi64_mask, _mm256_cmple_epu32_mask,
36937         _mm256_cmple_epu64_mask, _mm256_cmplt_epi32_mask,
36938         _mm256_cmplt_epi64_mask, _mm256_cmplt_epu32_mask,
36939         _mm256_cmplt_epu64_mask, _mm256_cmpneq_epi32_mask,
36940         _mm256_cmpneq_epi64_mask, _mm256_cmpneq_epu32_mask,
36941         _mm256_cmpneq_epu64_mask, _mm256_mask_cmpge_epi32_mask,
36942         _mm256_mask_cmpge_epi64_mask, _mm256_mask_cmpge_epu32_mask,
36943         _mm256_mask_cmpge_epu64_mask, _mm256_mask_cmple_epi32_mask,
36944         _mm256_mask_cmple_epi64_mask, _mm256_mask_cmple_epu32_mask,
36945         _mm256_mask_cmple_epu64_mask, _mm256_mask_cmplt_epi32_mask,
36946         _mm256_mask_cmplt_epi64_mask, _mm256_mask_cmplt_epu32_mask,
36947         _mm256_mask_cmplt_epu64_mask, _mm256_mask_cmpneq_epi32_mask,
36948         _mm256_mask_cmpneq_epi64_mask, _mm256_mask_cmpneq_epu32_mask,
36949         _mm256_mask_cmpneq_epu64_mask, _mm_cmpge_epi32_mask,
36950         _mm_cmpge_epi64_mask, _mm_cmpge_epu32_mask, _mm_cmpge_epu64_mask,
36951         _mm_cmple_epi32_mask, _mm_cmple_epi64_mask, _mm_cmple_epu32_mask,
36952         _mm_cmple_epu64_mask, _mm_cmplt_epi32_mask, _mm_cmplt_epi64_mask,
36953         _mm_cmplt_epu32_mask, _mm_cmplt_epu64_mask, _mm_cmpneq_epi32_mask,
36954         _mm_cmpneq_epi64_mask, _mm_cmpneq_epu32_mask, _mm_cmpneq_epu64_mask,
36955         _mm_mask_cmpge_epi32_mask, _mm_mask_cmpge_epi64_mask,
36956         _mm_mask_cmpge_epu32_mask, _mm_mask_cmpge_epu64_mask,
36957         _mm_mask_cmple_epi32_mask, _mm_mask_cmple_epi64_mask,
36958         _mm_mask_cmple_epu32_mask, _mm_mask_cmple_epu64_mask,
36959         _mm_mask_cmplt_epi32_mask, _mm_mask_cmplt_epi64_mask,
36960         _mm_mask_cmplt_epu32_mask, _mm_mask_cmplt_epu64_mask,
36961         _mm_mask_cmpneq_epi32_mask, _mm_mask_cmpneq_epi64_mask,
36962         _mm_mask_cmpneq_epu32_mask, _mm_mask_cmpneq_epu64_mask): Move
36963         definitions outside of __OPTIMIZE__ guarded section.
36965         PR target/79932
36966         * config/i386/avx512bwintrin.h (_mm512_packs_epi32,
36967         _mm512_maskz_packs_epi32, _mm512_mask_packs_epi32,
36968         _mm512_packus_epi32, _mm512_maskz_packus_epi32,
36969         _mm512_mask_packus_epi32): Move definitions outside of __OPTIMIZE__
36970         guarded section.
36972 2017-03-09  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
36974         * config/s390/vx-builtins.md ("vfee<mode>", "vfeez<mode>")
36975         ("vfenez<mode>"): Add missing constraints.
36977 2017-03-08  Martin Sebor  <msebor@redhat.com>
36979         PR target/79928
36980         * config/nds32/nds32.c (nds32_option_override):
36981         Fix misspelled diagnostic.
36983 2017-03-08  Jakub Jelinek  <jakub@redhat.com>
36985         PR c/79940
36986         * gimplify.c (gimplify_omp_for): Replace index var in outer
36987         taskloop statement with an artificial variable and add
36988         OMP_CLAUSE_PRIVATE clause for it.
36990 2017-03-08  Richard Biener  <rguenther@suse.de>
36992         PR tree-optimization/79955
36993         * tree-ssa-uninit.c (warn_uninitialized_vars): Do not warn
36994         for accesses that are completely outside of the variable.
36996 2017-03-08  Andrew Haley  <aph@redhat.com>
36998         PR tree-optimization/79943
36999         * tree-ssa-loop-split.c (compute_new_first_bound): When
37000         calculating the new upper bound, (END-BEG) should be added, not
37001         subtracted.
37003 2017-03-08  Jakub Jelinek  <jakub@redhat.com>
37005         * config/avr/avr.md (setmemhi): Make sure match_dup
37006         operand number comes before match_scratch.
37008 2017-03-08  Richard Biener  <rguenther@suse.de>
37010         PR tree-optimization/79920
37011         * tree-vect-slp.c (vect_create_mask_and_perm): Remove and inline
37012         with ncopies == 1 to ...
37013         (vect_transform_slp_perm_load): ... here.  Properly compute
37014         all element loads by iterating VF times over the group.  Do
37015         not handle ncopies (computed in a broken way) in
37016         vect_create_mask_and_perm.
37018 2017-03-08  Jakub Jelinek  <jakub@redhat.com>
37020         PR sanitizer/79904
37021         * internal-fn.c (expand_vector_ubsan_overflow): If arg0 or arg1
37022         is a uniform vector, use uniform_vector_p return value instead of
37023         building ARRAY_REF on folded VIEW_CONVERT_EXPR to array type.
37025 2017-03-07  Marek Polacek  <polacek@redhat.com>
37027         PR middle-end/79809
37028         * gimple-ssa-warn-alloca.c (pass_walloca::gate): Use HOST_WIDE_INT.
37029         (alloca_call_type): Likewise.
37031 2017-03-07  Martin Liska  <mliska@suse.cz>
37033         * gcov.c (process_args): Put comment to correct location.
37035 2017-03-07  Martin Liska  <mliska@suse.cz>
37037         PR middle-end/68270
37038         * tree-chkp.c (chkp_may_narrow_to_field): Add new argument ref.
37039         Use array_at_struct_end_p instead of DECL_CHAIN (field).
37040         (chkp_narrow_bounds_for_field): Likewise.
37041         (chkp_parse_array_and_component_ref): Pass one more argument to
37042         call.
37044 2017-03-07  Richard Biener  <rguenther@suse.de>
37046         * tree-vect-loop-manip.c (slpeel_add_loop_guard): Preserve
37047         preheaders.
37049 2017-03-07  Segher Boessenkool  <segher@kernel.crashing.org>
37051         * config/i386/i386.c (ix86_local_alignment): Align most aggregates
37052         of 16 bytes and more to 16 bytes, not those of 16 bits and more.
37054 2017-03-07  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
37056         PR c/79855
37057         * params.def (PARAM_STORE_MERGING_ALLOW_UNALIGNED): Add full stop
37058         to end of description.
37059         (PARAM_MAX_STORES_TO_MERGE): Likewise.
37061 2017-03-07  Jakub Jelinek  <jakub@redhat.com>
37063         PR rtl-optimization/79901
37064         * config/i386/sse.md (*avx512bw_<code><mode>3<mask_name>): Renamed to
37065         ...
37066         (*avx512f_<code><mode>3<mask_name>): ... this.
37067         (<code><mode>3 with maxmin code iterator): Use VI8_AVX2_AVX512F
37068         iterator instead of VI8_AVX2_AVX512BW.
37070         PR rtl-optimization/79901
37071         * expr.c (expand_expr_real_2): For vector MIN/MAX, if there is no
37072         min/max expander, expand it using expand_vec_cond_expr.
37074         PR sanitizer/79897
37075         * ubsan.c (ubsan_encode_value): Call mark_addressable on the
37076         temporary.
37078 2017-03-06  Jakub Jelinek  <jakub@redhat.com>
37080         PR c++/79821
37081         * dwarf2out.h (dw_vec_const): Change array type from unsigned char *
37082         to void * for PCH reasons.
37083         * dwarf2out.c (output_loc_operands, output_die): Cast
37084         v.val_vec.array to unsigned char *.
37086 2017-03-06  John David Anglin  <danglin@gcc.gnu.org>
37088         PR target/77850
37089         * config/pa/pa-64.h (PAD_VARARGS_DOWN): Don't pad down complex and
37090         vector types.
37092 2017-03-06  Vladimir Makarov  <vmakarov@redhat.com>
37094         PR rtl-optimization/79571
37095         * lra-constraints.c (process_alt_operands): Calculate static
37096         reject and subtract it from overall when only addresses will be
37097         reloaded.
37099 2017-03-06  Julia Koval  <julia.koval@intel.com>
37101         PR target/79793
37102         * config/i386/i386.c (ix86_minimum_incoming_stack_boundary): Set
37103         incoming stack boundary to 128 for 64-bit targets.
37105 2017-03-06  Richard Biener  <rguenther@suse.de>
37107         PR tree-optimization/79894
37108         * tree-vectorizer.c (vectorize_loops): Set loop_vectorized_call
37109         to NULL after folding it.
37111 2017-03-06  Richard Biener  <rguenther@suse.de>
37113         PR tree-optimization/79824
37114         * tree-vect-stmts.c (get_group_load_store_type): Fix alignment
37115         check disabling peeling for gaps.
37117 2017-03-06  Toma Tabacu  <toma.tabacu@imgtec.com>
37119         * doc/sourcebuild.texi (Effective-Target Keywords, Environment
37120         attributes): Document gettimeofday.
37122 2017-03-06  Robin Dapp  <rdapp@linux.vnet.ibm.com>
37124         * config/s390/s390.c (s390_option_override_internal): Set
37125         PARAM_MIN_VECT_LOOP_BOUND
37127 2017-03-06  Robin Dapp  <rdapp@linux.vnet.ibm.com>
37129         * config/s390/s390.c (s390_asm_output_function_label): Use nopr %r0.
37130         * config/s390/s390.md: Likewise.
37132 2017-03-06  Jakub Jelinek  <jakub@redhat.com>
37134         PR target/79812
37135         * config/i386/sse.md (VI8F_256_512): Remove mode iterator.
37136         (<avx2_avx512>_perm<mode>): Rename to ...
37137         (avx2_perm<mode>): ... this.  Use VI8F_256 iterator instead
37138         of VI8F_256_512.
37139         (<avx512>_perm<mode>_mask): Rename to ...
37140         (avx512vl_perm<mode>_mask): ... this.  Use VI8F_256 iterator instead
37141         of VI8F_256_512.
37142         (<avx2_avx512>_perm<mode>_1<mask_name>): Rename to ...
37143         (avx2_perm<mode>_1<mask_name): ... this.  Use VI8F_256 iterator
37144         instead of VI8F_256_512.
37145         (avx512f_perm<mode>): New define_expand.
37146         (avx512f_perm<mode>_mask): Likewise.
37147         (avx512f_perm<mode>_1<mask_name>): New define_insn.
37148         (<avx512>_vec_dup<mode>_1): Fix up vec_select mode.
37150 2017-03-06  Prachi Godbole  <prachi.godbole@imgtec.com>
37152         * config/mips/mips-msa.md (msa_fmax_a_<msafmt>, msa_fmin_a_<msafmt>,
37153         msa_max_a_<msafmt>, msa_min_a_<msafmt>): Introduce mode interator for
37154         if_then_else.
37155         (smin<mode>3, smax<mode>3): Change operand print code from 'B' to 'E'.
37157 2017-03-06  Martin Liska  <mliska@suse.cz>
37159         PR sanitize/79783
37160         * asan.c (asan_expand_poison_ifn): Do not expand ASAN_POISON
37161         when having a SSA NAME w/o VAR_DECL assigned to it.
37163 2017-03-06  Prachi Godbole  <prachi.godbole@imgtec.com>
37165         * config/mips/mips-msa.md (msa_dotp_<su>_d, msa_dpadd_<su>_d,
37166         msa_dpsub_<su>_d): Fix MODE for vec_select.
37168 2017-03-06  Prachi Godbole  <prachi.godbole@imgtec.com>
37170         * config/mips/mips.c (mips_gen_const_int_vector): Change type of last
37171         argument.
37172         * config/mips/mips-protos.h (mips_gen_const_int_vector): Likewise.
37174 2017-03-06  Richard Biener  <rguenther@suse.de>
37176         * lto-streamer.c (lto_check_version): Use %qs in diagnostics.
37177         * plugin.c (register_plugin_info): Likewise.
37178         * tree-chkp.c (chkp_make_static_const_bounds): Likewise.
37180 2017-03-05  Jakub Jelinek  <jakub@redhat.com>
37182         * config/i386/sse.md (sse_storehps, sse_storelps,
37183         avx_<castmode><avxsizesuffix>_<castmode>,
37184         avx512f_<castmode><avxsizesuffix>_<castmode>,
37185         avx512f_<castmode><avxsizesuffix>_256<castmode>): Require
37186         in condition that at least one operand is not a MEM.
37188 2017-03-03  Jakub Jelinek  <jakub@redhat.com>
37190         PR middle-end/79805
37191         * internal-fn.def (ATOMIC_BIT_TEST_AND_SET, ATOMIC_BIT_TEST_AND_RESET,
37192         ATOMIC_BIT_TEST_AND_COMPLEMENT, ATOMIC_COMPARE_EXCHANGE): Remove
37193         ECF_NOTHROW.
37194         * gimple-fold.c (fold_builtin_atomic_compare_exchange): Set
37195         gimple_call_nothrow_p flag based on whether original builtin can throw.
37196         If it can, emit following stmts on the fallthrough edge.
37197         * tree-ssa-ccp.c (optimize_atomic_bit_test_and): Similarly, except
37198         don't create new bb if inserting just debug stmts on the edge, try to
37199         insert them on the fallthru bb or just reset debug stmts.
37201 2017-03-03  Segher Boesssenkool  <segher@kernel.crashing.org>
37203         PR target/43763
37204         * config/rs6000/rs6000.c (rs6000_final_prescan_insn): Save and
37205         restore recog_data (including the operand rtxes inside it) around
37206         the call to get_insn_template.
37208 2017-03-03  Martin Sebor  <msebor@redhat.com>
37210         PR tree-optimization/79699
37211         * context.c (context::~context): Free MPFR caches to avoid
37212         a memory leak on program exit.
37214 2017-03-03  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
37216         * config/aarch64/aarch64.c (aarch64_float_const_representable_p):
37217         Use wide_int::ulow () instead of .elt (0).
37219 2017-03-03  Uros Bizjak  <ubizjak@gmail.com>
37221         * config/i386/i386.md (*pushtf): Change *roF constraint to *roC.
37222         (*pushxf): Limit oF constraint to 32bit targets and add oC
37223         constraint for 64bit targets.
37224         (pushxf splitter): Use PUSH_ROUNDING to calculate stack adjustment.
37225         (*pushdf): Change rmF constraint to rmC.
37227 2017-03-03  Martin Liska  <mliska@suse.cz>
37229         * tree-ssa-loop-prefetch.c (pass_loop_prefetch::execute):
37230         Remove unused variable.
37232 2017-03-03  Jakub Jelinek  <jakub@redhat.com>
37234         PR target/79807
37235         * config/i386/i386.c (ix86_expand_multi_arg_builtin): If target
37236         is a memory operand, increase num_memory.
37237         (ix86_expand_args_builtin): Likewise.
37239 2017-03-03  Jan Hubicka  <jh@suse.cz>
37241         PR lto/79760
37242         * ipa-devirt.c (maybe_record_node): Properly handle
37243         __cxa_pure_virtual visibility.
37245 2017-03-03  Martin Liska  <mliska@suse.cz>
37247         PR tree-optimization/79803
37248         * tree-ssa-loop-prefetch.c (tree_ssa_prefetch_arrays): Remove
37249         assert.
37250         (pass_loop_prefetch::execute): Disabled optimization if an
37251         assumption about L1 cache size is not met.
37253 2017-03-03  Martin Liska  <mliska@suse.cz>
37255         PR rtl-optimization/79574
37256         * gcse.c (struct gcse_expr): Use HOST_WIDE_INT instead of int.
37257         (hash_scan_set): Likewise.
37258         (dump_hash_table): Likewise.
37259         (hoist_code): Likewise.
37261 2017-03-03  Richard Biener  <rguenther@suse.de>
37263         * fixed-value.c (fixed_from_string): Restore use of elt (1)
37264         in place of uhigh ().
37265         (fixed_convert_from_real): Likewise.
37267 2017-03-03  Uros Bizjak  <ubizjak@gmail.com>
37269         PR target/79514
37270         * config/i386/i386.md (*pushxf_rounded): Use Pmode instead of DImode.
37272 2017-03-03  Richard Biener  <rguenther@suse.de>
37274         PR middle-end/79818
37275         * match.pd ( X +- C1 CMP C2 -> X CMP C2 -+ C1): Add missing
37276         TYPE_OVERFLOW_UNDEFINED check.
37278 2017-03-02  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
37280         * config/rs6000/vector.md (vector_ne_<mode>_p): Correct operand
37281         numbers.
37282         (vector_ae_<mode>_p): Likewise.
37283         (vector_nez_<mode>_p): Likewise.
37284         (vector_ne_v2di_p): Likewise.
37285         (vector_ae_v2di_p): Likewise.
37286         (vector_ne_<mode>_p): Likewise.
37287         * config/rs6000/vsx.md (vsx_tsqrt<mode>2_fg): Correct operand
37288         numbers.
37289         (vsx_tsqrt<mode>2_fe): Likewise.
37291 2017-03-02  Uros Bizjak  <ubizjak@gmail.com>
37293         PR target/79514
37294         * config/i386/i386.md (*pushxf_rounded): New insn_and_split pattern.
37296 2017-03-02  Jakub Jelinek  <jakub@redhat.com>
37298         PR rtl-optimization/79780
37299         * cprop.c (one_cprop_pass): When second and further conditional trap
37300         in a single basic block is turned into an unconditional trap, turn it
37301         into a deleted note to avoid RTL verification failures.
37303 2017-03-02  Richard Biener  <rguenther@suse.de>
37305         * fold-const.c (const_binop): Use ulow () instead of elt (0).
37307 2017-03-02  Richard Biener  <rguenther@suse.de>
37309         PR tree-optimization/79345
37310         PR c++/42000
37311         * tree-ssa-alias.c (walk_aliased_vdefs_1): Take a limit
37312         param and abort the walk, returning -1 if it is hit.
37313         (walk_aliased_vdefs): Take a limit param and pass it on.
37314         * tree-ssa-alias.h (walk_aliased_vdefs): Add a limit param,
37315         defaulting to 0 and return a signed int.
37316         * tree-ssa-uninit.c (struct check_defs_data): New struct.
37317         (check_defs): New helper.
37318         (warn_uninitialized_vars): Use walk_aliased_vdefs to warn
37319         about uninitialized memory.
37320         * fixed-value.c (fixed_from_string): Use ulow/uhigh to avoid
37321         bogus uninitialized warning.
37322         (fixed_convert_from_real): Likewise.
37324 2017-03-02  Bin Cheng  <bin.cheng@arm.com>
37326         PR tree-optimization/66768
37327         * tree-ssa-loop-ivopts.c (find_interesting_uses_address): Skip addr
37328         iv_use if base object can't be determined.
37330 2017-03-02  Jakub Jelinek  <jakub@redhat.com>
37332         PR tree-optimization/79345
37333         * gensupport.h (struct pattern_stats): Add min_scratch_opno field.
37334         * gensupport.c (get_pattern_stats_1) <case MATCH_SCRATCH>: Update it.
37335         (get_pattern_stats): Initialize it.
37336         * genemit.c (gen_expand): Verify match_scratch numbers come after
37337         match_operand/match_dup numbers.
37338         * config/i386/i386.md (<s>mul<mode>3_highpart): Swap match_dup and
37339         match_scratch numbers.
37340         * config/i386/sse.md (avx2_gathersi<mode>, avx2_gatherdi<mode>):
37341         Likewise.
37342         * config/s390/s390.md (trunctdsd2): Likewise.
37344 2017-03-02  Richard Biener  <rguenther@suse.de>
37346         * wide-int.h (wide_int_storage::operator=): Implement in terms
37347         of wi::copy.
37349 2017-03-02  Richard Biener  <rguenther@suse.de>
37351         PR tree-optimization/79777
37352         * tree-ssa-pre.c (eliminate_insert): Give up if we simplify
37353         the to insert expression to sth existing.
37355 2017-03-01  Martin Sebor  <msebor@redhat.com>
37357         PR middle-end/79692
37358         * gimple-ssa-sprintf.c
37359         (directive::known_width_and_precision): New function.
37360         (format_integer): Use it.
37361         (get_mpfr_format_length): Consider the full range of precision
37362         when computing %g output with the # flag.  Set the likely byte
37363         count to 3 rather than 1 when precision is indeterminate.
37364         (format_floating): Correct the lower bound of precision.
37366 2017-03-01  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
37368         * doc/invoke.texi: Document default code model for 64-bit Linux.
37370 2017-03-01  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
37372         PR target/79752
37373         * config/rs6000/rs6000.md (peephole2 for udiv/umod): Should emit
37374         udiv rather than div since input pattern is unsigned.
37376 2017-03-01  Uros Bizjak  <ubizjak@gmail.com>
37378         * config/i386/i386.c (print_reg): Warn for values of
37379         unsupported size in integer register.
37381 2017-03-01  Michael Meissner  <meissner@linux.vnet.ibm.com>
37383         PR target/79439
37384         * config/rs6000/predicates.md (current_file_function_operand): Do
37385         not allow self calls to be local if the function is replaceable.
37387 2017-03-01  Kelvin Nilsen  <kelvin@gcc.gnu.org>
37389         PR target/79395
37390         * config/rs6000/altivec.h (vec_ctz and others): Change the
37391         preprocessor macro that controls conditional compilation from
37392         _ARCH_PWR9 to __POWER9_VECTOR__.
37393         (vec_all_ne): Change parameterization of __altivec_scalar_pred
37394         macro expansion under preprocessor #ifdef __POWER9_VECTOR__
37395         control (instead of _ARCH_PWR9 control) so that template
37396         definition uses power9-specific function.
37397         (vec_any_eq): Likewise.
37398         (vec_all_ne): Change macro definition to use a power9-specific
37399         expansion under #ifdef __POWER9_VECTOR__ control (instead of
37400         _ARCH_PWR9 control).
37401         (vec_any_eq) Likewise.
37402         * config/rs6000/rs6000-builtin.def (CMPNEF): Remove BU_P9V_AV_2
37403         expansion for CMPNEF to remove support for xvcmpnesp instruction.
37404         (CMPNED): Remove BU_P9V_AV2 expansion for CMPNED to remove
37405         support for xvcmpnedp instruction.
37406         (VCMPNEB_P): Replace BU_P9V_AV_P macro expansion with BU_P9V_AV_2
37407         macro expansion so that Power9 implementation of vec_all_ne does
37408         not use the AltiVec predicate framework.
37409         (VCMPNEH_P): Likewise.
37410         (VCMPNEW_P): Likewise.
37411         (VCMPNED_P): Likewise.
37412         (VCMPNEFP_P): Likewise.
37413         (VCMPNEDP_P): Likewise.
37414         (VCMPAEB_P): Add BU_P9V_AV_2 macro expansion to change
37415         implementation of vec_any_eq to not use AltiVec predicate
37416         framework.
37417         (VCMPAEH_P): Likewise.
37418         (VCMPAEW_P): Likewise.
37419         (VCMPAED_P): Likewise.
37420         (VCMPAEFP_P): Likewise.
37421         (VCMPAEDP_P): Likewise.
37422         (VCMPNE_P): Replace BU_P9V_OVERLOAD_P macro expansion with
37423         BU_P9V_OVERLOAD_2 so that Power9 implementation of vec_all_ne does
37424         not use the AltiVec predicate framework.
37425         (VCMPAE_P): Add BU_P9V_OVERLOAD_2 macro to change implementation
37426         of vec_any_eq to not use AltiVec predicate framework.
37427         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Add
37428         support for predefined __POWER9_VECTOR__ macro to indicate that
37429         Power9 instruction selection is enabled.
37430         (altivec_overloaded_builtins): Remove extraneous
37431         ALTIVEC_BUILTIN_VEC_CMPNE entry for overloaded
37432         function argument types RS6000_BTI_bool_V16QI and
37433         RS6000_BTI_bool_V16QI.  Remove erroneous ALTIVEC_BUILTIN_VEC_CMPNE
37434         entry for overloaded function argument types RS6000_BTI_bool_V4SI
37435         andRS6000_BTI_bool_V4SI, mapping to P9V_BUILTIN_CMPNEB.  Remove
37436         two entries mapping to P9V_BUITIN_CMPNED and one entry mapping to
37437         P9V_BUILTIN_CMPNEF to force use of instructions not specific to
37438         Power9 for implementations of vec_cmpne.  Change the signature for
37439         all definitions of the overloaded P9V_BUILTIN_VEC_CMPNE_P function
37440         (representing vec_all_ne) to remove the previously described first
37441         argument of type RS6000_BTI_INTSI, as this was an artifact of
37442         reliance on the AltiVec predicate framework, which is no longer
37443         used in the implementation of these functions.  Add
37444         P9V_BUILTIN_VEC_VCMPAE_P entries (representing the vec_anyeq
37445         function) to match all of the P9V_BUILTIN_VEC_VCMNE_P entries
37446         since, unlike the AltiVec predicate framework implementation, we
37447         do not share function descriptors between vec_alle and vec_anyeq.
37448         (altivec_resolve_overloaded_builtin): Add SFmode and DFmode to the
37449         set of modes that receive special treatment even when
37450         TARGET_P9_VECTOR is true.  The special treatment emits code that
37451         does not depend on Power9 instructions.
37452         * config/rs6000/vector.md (vector_ne_<mode>_p): Change this
37453         define_expand to not rely on AltiVec predicate framework.
37454         (vector_ae_<mode>p): New define_expand to represent vec_any_eq
37455         function.
37456         (vector_ne_v2di_p): Change this define_expand to not rely on
37457         AltiVec predicate framework.
37458         (vector_ae_v2di_p): New define_expand to represent vec_any_eq
37459         function.
37460         (vector_ne_<mode>_p): Change this define_expand to not rely on
37461         AltiVec predicate framework.
37462         (vector_ae_<mode>p): New define_expand to represent vec_any_eq
37463         function.
37464         * config/rs6000/vsx.md (*vsx_ne_<mode>_p): For modes VSX_EXTRACT_I
37465         (V16QI, V8HI, V4SI), correct a typo in the code emitted for this
37466         define_insn pattern.
37467         (*vsx_ne_<mode>_p): For modes VSX_F (V4SF and V2DF), remove this
37468         define_insn pattern because the xvcmpne<VSs>. instruction is not
37469         supported.
37470         (vcmpne<VSs>): Remove this define_insn because xvcmpne<VSs>
37471         instruction is not supported.
37473 2017-03-01  Jakub Jelinek  <jakub@redhat.com>
37475         * config/nvptx/nvptx.c: Include intl.h.
37477 2017-03-01  Martin Jambor  <mjambor@suse.cz>
37479         PR lto/78140
37480         * ipa-prop.h (ipa_bits): Removed field known.
37481         (ipa_jump_func): Removed field vr_known.  Changed fields bits and m_vr
37482         to pointers.  Adjusted their comments to warn about their sharing.
37483         (ipcp_transformation_summary): Change bits to a vector of pointers.
37484         (ipa_check_create_edge_args): Moved to ipa-prop.c, declare.
37485         (ipa_get_ipa_bits_for_value): Declare.
37486         * tree-vrp.h (value_range): Mark as GTY((for_user)).
37487         * ipa-prop.c (ipa_bit_ggc_hash_traits): New.
37488         (ipa_bits_hash_table): Likewise.
37489         (ipa_vr_ggc_hash_traits): Likewise.
37490         (ipa_vr_hash_table): Likewise.
37491         (ipa_print_node_jump_functions_for_edge): Adjust for bits and m_vr
37492         being pointers and vr_known being removed.
37493         (ipa_set_jf_unknown): Likewise.
37494         (ipa_get_ipa_bits_for_value): New function.
37495         (ipa_set_jfunc_bits): Likewise.
37496         (ipa_get_value_range): New overloaded functions.
37497         (ipa_set_jfunc_vr): Likewise.
37498         (ipa_compute_jump_functions_for_edge): Use the above functions to
37499         construct bits and vr parts of jump functions.
37500         (ipa_check_create_edge_args): Move here from ipa-prop.h, also allocate
37501         ipa_bits_hash_table and ipa_vr_hash_table if they do not already
37502         exist.
37503         (ipcp_grow_transformations_if_necessary): Also allocate
37504         ipa_bits_hash_table and ipa_vr_hash_table if they do not already
37505         exist.
37506         (ipa_node_params_t::duplicate): Do not copy bits, just pointers to
37507         them.  Fix too long lines.
37508         (ipa_write_jump_function): Adjust for bits and m_vr being pointers and
37509         vr_known being removed.
37510         (ipa_read_jump_function): Use new setter functions to construct bits
37511         and vr parts of jump functions or set them to NULL.
37512         (write_ipcp_transformation_info): Adjust for bits being pointers.
37513         (read_ipcp_transformation_info): Likewise.
37514         (ipcp_update_bits): Likewise.  Fix excessively long lines a trailing
37515         space.
37516         Include gt-ipa-prop.h.
37517         * ipa-cp.c (propagate_bits_across_jump_function): Adjust for bits
37518         being pointers.
37519         (ipcp_store_bits_results): Likewise.
37520         (propagate_vr_across_jump_function): Adjust for m_vr being a pointer.
37521         Do not write to existing jump functions but use a temporary instead.
37523 2017-03-01  Jakub Jelinek  <jakub@redhat.com>
37525         PR c++/79681
37526         * fold-const.c (make_bit_field_ref): If orig_inner is COMPONENT_REF,
37527         attempt to use its first operand as BIT_FIELD_REF base.
37529 2017-03-01  Richard Biener  <rguenther@suse.de>
37531         PR middle-end/79721
37532         * tree-chrec.c (chrec_evaluate): Perform computation of Newtons
37533         interpolating formula in wrapping arithmetic.
37534         (chrec_apply): Convert chrec_evaluate return value to wanted type.
37536 2017-03-01  Jakub Jelinek  <jakub@redhat.com>
37538         PR tree-optimization/79734
37539         * tree-vect-generic.c (expand_vector_condition): Optimize
37540         AVX512 vector boolean VEC_COND_EXPRs into bitwise operations.
37541         Handle VEC_COND_EXPR where comparison has different inner width from
37542         type's inner width.
37544 2017-02-28  Sandra Loosemore  <sandra@codesourcery.com>
37546         * doc/invoke.texi (ARC Options): Copy-edit to fix punctuation,
37547         markup, and similar issues.  Remove @opindex entries for things
37548         that aren't options.  Add missing -mmpy-option entries.
37550 2017-02-28  Jakub Jelinek  <jakub@redhat.com>
37552         PR tree-optimization/79737
37553         * gimple-ssa-store-merging.c (encode_tree_to_bitpos): If bitlen is
37554         a multiple of BITS_PER_UNIT and !BYTES_BIG_ENDIAN, clear
37555         tmpbuf[byte_size - 1].  Call natice_encode_expr with byte_size - 1
37556         instead of byte_size.  Formatting fix.
37557         (shift_bytes_in_array_right): Formatting fix.
37559 2017-02-28  Eric Botcazou  <ebotcazou@adacore.com>
37561         PR target/79749
37562         * config/sparc/sparc.c (sparc_frame_pointer_required): Add missing
37563         condition on optimize for the leaf function test.
37565 2017-02-28  Martin Liska  <mliska@suse.cz>
37567         PR lto/79625
37568         * read-rtl-function.c (function_reader::handle_unknown_directive):
37569         Bail out when one uses -flto.
37571 2017-02-28  Martin Liska  <mliska@suse.cz>
37573         * common.opt: Replace space with tabular for options of <number>
37574         type.
37575         * config/i386/i386.opt: Show <number> value for
37576         -mlarge-data-threshold.
37577         * opts.c (print_filtered_help): Do not display number in hexadecimal
37578         format.
37580 2017-02-28  Martin Liska  <mliska@suse.cz>
37582         * common.opt: Fix --help=option -Q for options which are of
37583         an enum type.
37585 2017-02-28  Uros Bizjak  <ubizjak@gmail.com>
37587         * config/i386/i386.c (print_reg): Error out for values
37588         of 8-bit size in invalid integer register.
37590 2017-02-28  Martin Sebor  <msebor@redhat.com>
37592         PR tree-optimization/79691
37593         * passes.def (pass_all_optimizations_g): Enable pass_sprintf_length.
37595 2017-02-28  Jakub Jelinek  <jakub@redhat.com>
37597         PR target/79729
37598         * config/i386/i386.c (ix86_print_operand) <case 'R'>: Replace
37599         gcc_unreachable with output_operand_lossage.
37601 2017-02-28  Richard Biener  <rguenther@suse.de>
37603         PR tree-optimization/79740
37604         * tree-ssa-sccvn.c (vn_nary_op_insert_into): Allow redundant
37605         inserts.
37606         (visit_nary_op): Insert the nary into the hashtable if we
37607         pattern-matched sth.
37608         * tree-ssa-pre.c (eliminate_insert): Robustify.
37610 2017-02-28  Richard Biener  <rguenther@suse.de>
37612         PR middle-end/79731
37613         * fold-const.c (decode_field_reference): Reject out-of-bound
37614         accesses.
37616 2017-02-28  Jakub Jelinek  <jakub@redhat.com>
37618         * config/i386/i386.c: Include intl.h.
37619         (ix86_option_override_internal): Use cond ? G_("...") : G_("...")
37620         instead of just cond ? "..." : "...".
37621         * config/nvptx/nvptx.c (nvptx_goacc_validate_dims): Likewise.
37622         * coverage.c (read_counts_file): Likewise.
37623         * omp-offload.c: Include intl.h.
37624         (oacc_loop_fixed_partitions): Use cond ? G_("...") : G_("...") instead
37625         of just cond ? "..." : "...".
37626         * gcov.c (read_count_file): Use cond ? N_("...") : N_("...") instead
37627         of just cond ? "..." : "...".
37629 2017-02-28  Richard Earnshaw  <rearnsha@arm.com>
37631         PR target/79742
37632         * config/arm/parsecpu.awk (gen_data): Set tuning target to 'tune for'
37633         entry, if present.
37634         * config/arm/arm-cpus.in (cortex-m0plus.small-multiply): Correct
37635         'tune for' CPU name.
37636         * config/arm/arm-cpu-data.h: Regenerated.
37638 2017-02-28  Richard Biener  <rguenther@suse.de>
37640         PR tree-optimization/79732
37641         * tree-inline.c (expand_call_inline): Do not shadow var.
37643 2017-02-28  Richard Biener  <rguenther@suse.de>
37645         PR tree-optimization/79723
37646         * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Preserve
37647         address-space properly.
37649 2017-02-28  Thomas Schwinge  <thomas@codesourcery.com>
37651         * doc/optinfo.texi (Optimization groups): Fix option used for
37652         OPTGROUP_ALL.
37653         * doc/invoke.texi (-fopt-info): Document "omp".
37654         * dumpfile.h: Sort OPTGROUP_OMP before OPTGROUP_VEC.
37655         (OPTGROUP_ALL): Add OPTGROUP_OMP.
37656         * hsa-gen.c (pass_data_gen_hsail): Use OPTGROUP_OMP.
37657         * ipa-hsa.c (pass_data_ipa_hsa): Likewise.
37658         * omp-simd-clone.c (pass_data_omp_simd_clone): Likewise.
37660         * dumpfile.h (OPTGROUP_OPENMP): Rename to OPTGROUP_OMP.  Adjust
37661         all users.
37662         * dumpfile.c (optgroup_options): Instead of "openmp", associate
37663         OPTGROUP_OMP with "omp".
37665 2017-02-27  Pat Haugen  <pthaugen@us.ibm.com>
37667         PR target/79544
37668         * config/rs6000/rs6000-c.c (struct altivec_builtin_types): Use VSRAD
37669         for arithmetic shift of unsigned V2DI.
37671 2017-02-27  Claudiu Zissulescu  <claziss@synopsys.com>
37673         * config.gcc (arc*-): Clean up, use arc/big.h, arc/elf.h, and
37674         arc/linux.h headers.
37675         * config/arc/arc.h (TARGET_OS_CPP_BUILTINS): Remove.
37676         (LINK_SPEC): Likewise.
37677         (ARC_TLS_EXTRA_START_SPEC): Likewise.
37678         (EXTRA_SPECS): Likewise.
37679         (STARTFILE_SPEC): Likewise.
37680         (ENDFILE_SPEC): Likewise.
37681         (LIB_SPEC): Likewise.
37682         (TARGET_SDATA_DEFAULT): Likewise.
37683         (TARGET_MMEDIUM_CALLS_DEFAULT): Likewise.
37684         (MULTILIB_DEFAULTS): Likewise.
37685         (DWARF2_UNWIND_INFO): Likewise.
37686         * config/arc/big.h: New file.
37687         * config/arc/elf.h: Likewise.
37688         * config/arc/linux.h: Likewise.
37689         * config/arc/t-uClibc: Remove.
37691 2017-02-27  Bin Cheng  <bin.cheng@arm.com>
37693         PR tree-optimization/77536
37694         * tree-ssa-loop-manip.c (niter_for_unrolled_loop): New function.
37695         (tree_transform_and_unroll_loop): Use above function to compute the
37696         estimated niter of unrolled loop and use it when scaling profile.
37697         Also use count info rather than frequency if it's non-zero.
37698         * tree-ssa-loop-manip.h niter_for_unrolled_loop(): New declaration.
37699         * tree-vect-loop.c (scale_profile_for_vect_loop): New function.
37700         (vect_transform_loop): Call above function.
37702 2017-02-27  Richard Biener  <rguenther@suse.de>
37704         PR tree-optimization/45397
37705         * tree-ssa-pre.c (eliminate_insert): Handle BIT_AND_EXPR.
37706         * tree-ssa-sccvn.c (valueized_wider_op): New helper.
37707         (visit_nary_op): Add pattern matching for CSEing sign-changed
37708         or truncated operations with wider ones.
37710 2017-02-27  Richard Biener  <rguenther@suse.de>
37712         PR tree-optimization/79690
37713         * tree-vect-stmts.c (vectorizable_store): Use vector type
37714         built from the DR with address-space.
37716 2017-02-26  Gerald Pfeifer  <gerald@pfeifer.com>
37718         * doc/invoke.texi (Optimize Options): Refine the description
37719         of asan-use-after-return.
37721 2017-02-25  Alan Modra  <amodra@gmail.com>
37723         PR rtl-optimization/79584
37724         * lra-constraints.c (base_to_reg): Reload ad->base, the entire
37725         base, not ad->base_term, the reg within base.  Remove assertion
37726         that ad->base == ad->base_term.  Replace gen_int_mode using
37727         bogus mode with const0_rtx.
37729 2017-02-25  Jakub Jelinek  <jakub@redhat.com>
37731         PR middle-end/79396
37732         * tree-eh.c (operation_could_trap_p, stmt_could_throw_1_p): Handle
37733         FMA_EXPR like tcc_binary or tcc_unary.
37735         * tree-ssa-loop-niter.c (number_of_iterations_exit): Simplify warning.
37737         PR debug/77589
37738         * dwarf2out.c (struct dw_loc_list_struct): Add noted_variable_value
37739         bitfield.
37740         (size_of_loc_descr): Handle DW_OP_GNU_variable_value.
37741         (output_loc_operands): Handle DW_OP_call_ref and
37742         DW_OP_GNU_variable_value.
37743         (struct variable_value_struct): New type.
37744         (struct variable_value_hasher): Likewise.
37745         (variable_value_hash): New variable.
37746         (string_types): Remove.
37747         (copy_loc_descr): New function.
37748         (add_loc_descr_to_each): Clarify comment.  Use copy_loc_descr.
37749         (prepend_loc_descr_to_each): New function.
37750         (add_loc_list): Fix comment typo.  Use prepend_loc_descr_to_each
37751         instead of add_loc_descr_to_each if the first argument is single
37752         location list and the second has multiple.
37753         (resolve_args_picking_1): Handle DW_OP_GNU_variable_value.
37754         (loc_list_from_tree_1): For early_dwarf, emit DW_OP_GNU_variable_value
37755         when looking for variable value which doesn't have other location info.
37756         (loc_list_from_tree): Formatting fix.
37757         (gen_array_type_die): Simplify DW_AT_string_length handling.
37758         (adjust_string_types): Remove.
37759         (gen_subprogram_die): Don't call adjust_string_types nor test/set
37760         string_types.  Call resolve_variable_values.
37761         (prune_unused_types_walk_loc_descr): Handle DW_OP_GNU_variable_value.
37762         (resolve_addr_in_expr): Likewise.  Add A argument.
37763         (copy_deref_exprloc): Remove deref argument.  Adjust for the
37764         original expression being DW_OP_GNU_variable_value with optionally
37765         DW_OP_stack_value after it instead of DW_OP_call4 with DW_OP_deref
37766         optionally after it.
37767         (optimize_string_length): Rework for DW_OP_GNU_variable_value.
37768         (resolve_addr): Adjust optimize_string_length and resolve_addr_in_expr
37769         callers.  Set remove_AT_byte_size if removing DW_AT_string_length.
37770         (variable_value_hasher::hash, variable_value_hasher::equal): New
37771         methods.
37772         (resolve_variable_value_in_expr, resolve_variable_value,
37773         resolve_variable_values, note_variable_value_in_expr,
37774         note_variable_value): New functions.
37775         (dwarf2out_early_finish): Call note_variable_value on all toplevel
37776         DIEs.
37778 2017-02-24  Jakub Jelinek  <jakub@redhat.com>
37780         PR c/79677
37781         * opts.h (handle_generated_option): Add GENERATED_P argument.
37782         * opts-common.c (handle_option): Adjust function comment.
37783         (handle_generated_option): Add GENERATED_P argument, pass it to
37784         handle_option.
37785         (control_warning_option): Pass false to handle_generated_option
37786         GENERATED_P.
37787         * opts.c (maybe_default_option): Pass true to handle_generated_option
37788         GENERATED_P.
37789         * optc-gen.awk: Likewise.
37791 2017-02-24  Segher Boessenkool  <segher@kernel.crashing.org>
37793         * config/sh/sh.md (tstsi_t): If operands[0] is a SUBREG instead of
37794         a REG, look at the REG it is a SUBREG of.
37795         (splitter for cmpeqsi_t): Ditto.
37797 2017-02-24  Segher Boessenkool  <segher@kernel.crashing.org>
37799         * config/pa/pa.c (pa_combine_instructions): Do not share RTL.  Make
37800         the special USEs with the pattern of the insn, not the insn itself.
37802 2017-02-24  Matthew Fortune  <matthew.fortune@imgtec.com>
37804         PR target/79473
37805         * doc/invoke.texi: Document -mload-store-pairs.
37807 2017-02-24  Segher Boessenkool  <segher@kernel.crashing.org>
37808             Sandra Loosemore  <sandra@codesourcery.com>
37810         * config/nios2/nios2.c (nios2_simple_const_p): Returns false if the
37811         argument isn't a CONST_INT.
37812         (nios2_alternate_compare_const): Assert op is a CONST_INT.
37813         (nios2_valid_compare_const_p): Assert op is a CONST_INT.
37814         (nios2_validate_compare): Bypass alternate compare logic if *op2
37815         is not a CONST_INT.
37816         (ldstwm_operation_p): Return false if first_base is not a REG or
37817         if first_offset is not a CONST_INT.
37819 2017-02-24  Segher Boessenkool  <segher@kernel.crashing.org>
37821         * config/cris/cris.md: Use correct operand in a define_peephole2.
37823 2017-02-24  Segher Boessenkool  <segher@kernel.crashing.org>
37825         * config/c6x/c6x.c (predicate_insn): Do not incorrectly share RTL.
37827 2017-02-24  Segher Boessenkool  <segher@kernel.crashing.org>
37829         * config/arc/arc.c (arc_ccfsm_advance): Only take the PATTERN of
37830         this_insn if it is an INSN or JUMP_INSN.
37831         (force_offsettable): Look at base, not at addr.
37832         * config/arc/predicates.md (brcc_nolimm_operator): Don't call INTVAL
37833         on things that aren't necessarily CONST_INTs.
37835 2017-02-24  Uros Bizjak  <ubizjak@gmail.com>
37837         * doc/invoke.texi (x86 Options, -mfpmath=sse): Mention that
37838         -mfpmath=sse is the default also for x86-32 targets with SSE2
37839         instruction set when @option{-ffast-math} is enabled
37841 2017-02-24  Jeff Law  <law@redhat.com>
37843         PR rtl-optimizatoin/79286
37844         * ira.c (update_equiv_regs): Drop may_trap_p exception to
37845         dominance test.
37847 2017-02-24  Richard Biener  <rguenther@suse.de>
37849         PR tree-optimization/79389
37850         * gimple-ssa-split-paths.c (is_feasible_trace): Properly skip
37851         debug insns.
37853 2017-02-24  Aldy Hernandez  <aldyh@redhat.com>
37855         * tree-ssa-loop-niter.c (number_of_iterations_exit): Update
37856         function comment to reflect reality.
37857         (loop_exits_before_overflow): Fix typo in function description.
37859 2017-02-24  Richard Biener  <rguenther@suse.de>
37861         PR tree-optimization/79389
37862         * gimple-ssa-split-paths.c (is_feasible_trace): Verify more
37863         properly that a threading opportunity exists.  Detect conditional
37864         copy/constant propagation opportunities.
37866 2017-02-23  Eric Botcazou  <ebotcazou@adacore.com>
37868         * config/visium/visium.md (type): Add trap.
37869         (b): New mode attribute.
37870         (*btst): Rename into...
37871         (*btst<mode>): ...this and adjust.
37872         (*cbranchsi4_btst_insn): Rename into...
37873         (*cbranch<mode>4_btst_insn): ...this and adjust.
37874         (trap): New define_insn.
37876 2017-02-23  Jakub Jelinek  <jakub@redhat.com>
37878         PR tree-optimization/79389
37879         * ifcvt.c (struct noce_if_info): Add rev_cond field.
37880         (noce_reversed_cond_code): New function.
37881         (noce_emit_store_flag): Use rev_cond if non-NULL instead of
37882         reversed_comparison_code.  Formatting fix.
37883         (noce_try_store_flag): Test rev_cond != NULL in addition to
37884         reversed_comparison_code.
37885         (noce_try_store_flag_constants): Likewise.
37886         (noce_try_store_flag_mask): Likewise.
37887         (noce_try_addcc): Use rev_cond if non-NULL instead of
37888         reversed_comparison_code.
37889         (noce_try_cmove_arith): Likewise.  Formatting fixes.
37890         (noce_try_minmax, noce_try_abs): Clear rev_cond.
37891         (noce_find_if_block): Initialize rev_cond.
37892         (find_cond_trap): Call noce_get_condition with then_bb == trap_bb
37893         instead of false as last argument never attempt to reverse it
37894         afterwards.
37896 2017-02-23  Bin Cheng  <bin.cheng@arm.com>
37898         PR tree-optimization/79663
37899         * tree-predcom.c (combine_chains): Process refs in reverse order
37900         only for ZERO length chains, and add explaining comment.
37902 2017-02-23  Jeff Law  <law@redhat.com>
37904         PR tree-optimization/79578
37905         * tree-ssa-dse.c (clear_bytes_written_by): Use OEP_ADDRESS_OF
37906         in call to operand_equal_p.
37908 2017-01-23  Dominique d'Humieres  <dominiq@lps.ens.fr>
37910         PR target/71017
37911         * config/i386/cpuid.h: Fix another undefined behavior.
37913 2017-02-23  Richard Biener  <rguenther@suse.de>
37915         PR tree-optimization/79683
37916         * tree-vect-stmts.c (vect_analyze_stmt): Do not overwrite
37917         vector types for data-refs.
37919 2017-02-23  Martin Liska  <mliska@suse.cz>
37921         * params.def (PARAM_MIN_NONDEBUG_INSN_UID): Change default to 0.
37923 2017-02-23  Jakub Jelinek  <jakub@redhat.com>
37925         PR middle-end/79665
37926         * internal-fn.c (get_range_pos_neg): Moved to ...
37927         * tree.c (get_range_pos_neg): ... here.  No longer static.
37928         * tree.h (get_range_pos_neg): New prototype.
37929         * expr.c (expand_expr_real_2) <case TRUNC_DIV_EXPR>: If both arguments
37930         are known to be in between 0 and signed maximum inclusive, try to
37931         expand both unsigned and signed divmod and use the cheaper one from
37932         those.
37934 2017-02-22  Jeff Law  <law@redhat.com>
37936         PR tree-optimization/79578
37937         * tree-ssa-dse.c (clear_bytes_written_by): Use operand_equal_p
37938         to compare base operands.
37940 2017-02-22  Segher Boessenkool  <segher@kernel.crashing.org>
37942         PR target/79211
37943         * config/rs6000/rs6000.md (*fsel<SFDF:mode><SFDF2:mode>4): Use
37944         gpc_reg_operand instead of fpr_reg_operand.
37946 2017-02-22  Sameera Deshpande  <sameera.deshpande@imgtec.com>
37948         * config/mips/mips.c (mips_return_in_memory): Force FP
37949         vector types to be returned in memory for o32 ABI.
37951 2017-02-22  Jakub Jelinek  <jakub@redhat.com>
37953         * dwarf2out.c (gen_variable_die): For -gdwarf-5, use DW_TAG_variable
37954         instead of DW_TAG_member for static data member declarations and don't
37955         set no_linkage_name for static inline data members.
37956         (gen_member_die): For -gdwarf-5 don't change DW_TAG_variable
37957         to DW_TAG_member.
37959 2017-02-22  Martin Liska  <mliska@suse.cz>
37961         * doc/invoke.texi: Replace inequality signs with square brackets
37962         for -Wnormalized.
37964 2017-02-22  Matthew Fortune  <matthew.fortune@imgtec.com>
37966         PR target/78660
37967         * lra-constraints.c (simplify_operand_subreg): Handle
37968         WORD_REGISTER_OPERATIONS targets.
37970 2017-02-22  Jakub Jelinek  <jakub@redhat.com>
37972         PR target/70465
37973         * reg-stack.c (emit_swap_insn): Treat (float_extend:?F (mem:?F))
37974         and (const_double:?F) like (mem:?F) for the purpose of fxch %st(1)
37975         elimination by swapping fld*.
37977 2017-02-22  Richard Biener  <rguenther@suse.de>
37979         PR tree-optimization/79673
37980         * tree-ssa-pre.c (compute_avail): Use wide_int_to_tree to
37981         convert the [TARGET_]MEM_REF offset INTEGER_CST, scrapping off
37982         irrelevant address-space qualifiers and avoiding a
37983         ADDR_SPACE_CONVERT_EXPR from fold_convert.
37985 2017-02-22  Richard Biener  <rguenther@suse.de>
37987         PR tree-optimization/79666
37988         * tree-vrp.c (extract_range_from_binary_expr_1): Make sure
37989         to not symbolically negate if that may introduce undefined
37990         overflow.
37992 2017-02-22  Martin Liska  <mliska@suse.cz>
37994         PR lto/79587
37995         * data-streamer-in.c (streamer_read_gcov_count): Remove assert.
37996         * data-streamer-out.c (streamer_write_gcov_count_stream):
37997         Likewise.
37998         * value-prof.c (stream_out_histogram_value): Make assert more
37999         precise based on type of counter.
38001 2017-02-21  Uros Bizjak  <ubizjak@gmail.com>
38003         PR target/79593
38004         * config/i386/i386.md (standard_x87sse_constant_load splitter):
38005         Use nonimmediate_operand instead of memory_operand for operand 1.
38006         (float-extend standard_x87sse_constant_load splitter): Ditto.
38008 2017-02-21  Jeff Law  <law@redhat.com>
38010         PR tree-optimization/79621
38011         * gimple-ssa-isolate-paths.c (find_implicit_erroneous_behavior): Ignore
38012         blocks with edges to themselves.
38014 2017-02-21  Jakub Jelinek  <jakub@redhat.com>
38016         PR target/79633
38017         * tree-chkp-opt.c (chkp_optimize_string_function_calls): Use
38018         is_gimple_call instead of comparing gimple_code with GIMPLE_CALL.
38019         Use gimple_call_builtin_p.
38021         PR target/79570
38022         * sel-sched.c (moveup_expr_cached): Don't call sel_bb_head
38023         on temporarily removed DEBUG_INSNs.
38025         PR tree-optimization/79649
38026         * tree-loop-distribution.c (classify_partition): Give up on
38027         non-generic address space loads/stores.
38029 2017-02-21  Aldy Hernandez  <aldyh@redhat.com>
38031         * doc/loop.texi (Loop manipulation): Remove nonexistent
38032         tree_ssa_loop_version from the documentation.
38033         * cfgloopmanip.c (loop_version): Document CONDITION_BB argument.
38035 2017-02-21  Jakub Jelinek  <jakub@redhat.com>
38037         PR target/79494
38038         * config/i386/i386.c (ix86_expand_split_stack_prologue): Call
38039         make_reg_eh_region_note_nothrow_nononlocal on call_insn.
38040         * config/rs6000/rs6000.c: Include except.h.
38041         (rs6000_expand_split_stack_prologue): Call
38042         make_reg_eh_region_note_nothrow_nononlocal on the call insn.
38044 2017-02-21  Martin Jambor  <mjambor@suse.cz>
38046         PR lto/79579
38047         * ipa-prop.c (ipa_prop_write_jump_functions): Bail out if no edges
38048         have been analyzed.
38050 2017-02-21  Martin Jambor  <mjambor@suse.cz>
38052         * common.opt (-fipa-cp-alignment): Mark as ignored and preserved
38053         for backward compatibility only.
38054         * doc/invoke.texi (Option Summary): Remove all references to
38055         -fipa-cp-alignment.
38057 2017-02-21  Matthew Fortune  <matthew.fortune@imgtec.com>
38059         PR target/78660
38060         Revert:
38061         2017-02-20  Matthew Fortune  <matthew.fortune@imgtec.com>
38063         * lra-constraints.c (curr_insn_transform): Handle
38064         WORD_REGISTER_OPERATIONS requirements when reloading SUBREGs.
38066 2017-02-21  Martin Liska  <mliska@suse.cz>
38068         * config/i386/i386.opt: Replace -masm-dialect with -masm.
38070 2017-02-21  Thomas Schwinge  <thomas@codesourcery.com>
38072         PR translation/79638
38073         * config/nvptx/nvptx.c (ENTRY_TEMPLATE): Single out "%ntid.y".
38075 2017-02-21  Eric Botcazou  <ebotcazou@adacore.com>
38077         PR ada/67205
38078         * config/arm/arm.c (TARGET_CUSTOM_FUNCTION_DESCRIPTORS): Define.
38079         (arm_function_ok_for_sibcall): Return false for an indirect call by
38080         descriptor if all the argument registers are used.
38081         (arm_relayout_function): Use FUNCTION_ALIGNMENT macro to adjust the
38082         alignment of the function.
38084 2017-02-21  Jakub Jelinek  <jakub@redhat.com>
38086         PR tree-optimization/61441
38087         * simplify-rtx.c (simplify_const_unary_operation): For
38088         -fsignaling-nans and sNaN operand, return NULL_RTX rather than
38089         the sNaN unmodified.
38091 2017-02-20  Bernd Edlinger  <bernd.edlinger@hotmail.de>
38093         * Makefile.in (BUILD_SYSTEM_HEADER_DIR): New make variabe.
38094         (LIMITS_H_TEST, if_multiarch, stmp-fixinc): Use BUILD_SYSTEM_HEADER_DIR
38095         instead of SYSTEM_HEADER_DIR.
38097 2017-02-20  Gerald Pfeifer  <gerald@pfeifer.com>
38098             Martin LiÅ¡ka  <mliska@suse.cz>
38100         * doc/invoke.texi (use-after-scope-direct-emission-threshold):
38101         Fix typos and grammar, use active voice, and clarify.
38103 2017-02-20  Marek Polacek  <polacek@redhat.com>
38105         PR middle-end/79537
38106         * gimplify.c (gimplify_expr): Handle unused *&&L;.
38108         PR sanitizer/79558
38109         * ubsan.c (ubsan_type_descriptor): Check if TYPE_MAX_VALUE is null.
38111 2017-02-20  Jakub Jelinek  <jakub@redhat.com>
38113         PR target/79568
38114         * config/i386/i386.c (ix86_expand_builtin): Handle
38115         OPTION_MASK_ISA_AVX512VL and OPTION_MASK_ISA_64BIT in
38116         ix86_builtins_isa[fcode].isa as a requirement of those
38117         flags and any other flag in the bitmask.
38118         (ix86_init_mmx_sse_builtins): Use 0 instead of
38119         ~OPTION_MASK_ISA_64BIT as mask.
38120         * config/i386/i386-builtin.def (__builtin_ia32_rdtsc,
38121         __builtin_ia32_rdtscp, __builtin_ia32_pause, __builtin_ia32_bsrsi,
38122         __builtin_ia32_rdpmc, __builtin_ia32_rolqi, __builtin_ia32_rolhi,
38123         __builtin_ia32_rorqi, __builtin_ia32_rorhi): Likewise.
38125 2017-02-20  Matthew Fortune  <matthew.fortune@imgtec.com>
38127         PR target/78012
38128         * lra-constraints.c (split_reg): Check requested split mode
38129         is supported by the register.
38131 2017-02-20  Matthew Fortune  <matthew.fortune@imgtec.com>
38133         * lra-constraints.c (simplify_operand_subreg): Remove early
38134         return false.
38136 2017-02-20  Matthew Fortune  <matthew.fortune@imgtec.com>
38138         PR target/78660
38139         * lra-constraints.c (curr_insn_transform): Tighten condition
38140         for converting SUBREG reloads from OP_OUT to OP_INOUT.
38142 2017-02-20  Matthew Fortune  <matthew.fortune@imgtec.com>
38144         PR target/78660
38145         * lra-constraints.c (curr_insn_transform): Handle
38146         WORD_REGISTER_OPERATIONS requirements when reloading SUBREGs.
38148 2017-02-19  Uros Bizjak  <ubizjak@gmail.com>
38150         Revert:
38151         2016-05-30  Uros Bizjak  <ubizjak@gmail.com>
38153         * config/i386/sync.md (mfence_nosse): Use "lock orl $0, -4(%esp)".
38155 2017-02-19  Jonathan Wakely  <jwakely@redhat.com>
38157         PR c++/69523
38158         * doc/invoke.texi (C++ Dialect Options) [-Wliteral-suffix]: Update
38159         description.
38161 2017-02-19  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
38163         * gimple-pretty-print.c (dump_ternary_rhs): Adjust gimple dump format
38164         for FMA_EXPR.
38166 2017-02-18  Jakub Jelinek  <jakub@redhat.com>
38168         * final.c (last_columnnum, override_columnnum): New variables.
38169         (final_start_function): Set last_columnnum, pass it to begin_prologue
38170         hook and pass 0 to dwarf2out_begin_prologue.
38171         (final_scan_insn): Update override_columnnum.  Pass last_columnnum
38172         to source_line debug hook.
38173         (notice_source_line): Compute last_columnnum and for debug_column_info
38174         return true on column changes.
38175         * debug.h (struct gcc_debug_hooks): Add column argument to
38176         source_line and begin_prologue hooks.
38177         (debug_nothing_int_charstar_int_bool): Remove prototype.
38178         (debug_nothing_int_int_charstar,
38179         debug_nothing_int_int_charstar_int_bool): New prototypes.
38180         (dwarf2out_begin_prologue): Add column argument.
38181         * debug.c (do_nothing_debug_hooks): Adjust source_line and
38182         begin_prologue hooks.
38183         (debug_nothing_int_charstar_int_bool): Remove.
38184         (debug_nothing_int_int_charstar,
38185         debug_nothing_int_int_charstar_int_bool): New functions.
38186         * dwarf2out.c (dwarf2out_begin_prologue): Add column argument, pass it
38187         through to dwarf2out_source_line.
38188         (dwarf2_lineno_debug_hooks): Adjust begin_prologue hook.
38189         (dwarf2out_source_line): Add column argument, emit it if requested.
38190         * sdbout.c (sdbout_source_line, sdbout_begin_prologue): Add column
38191         arguments.
38192         * xcoffout.h (xcoffout_begin_prologue, xcoffout_source_line): Likewise.
38193         * xcoffout.c (xcoffout_begin_prologue, xcoffout_source_line): Likewise.
38194         * vmsdbgout.c (vmsdbgout_begin_prologue): Add column argument, pass it
38195         through to dwarf2out_begin_prologue.
38196         (vmsdbgout_source_line): Add column argument, pass it through to
38197         dwarf2out_source_line.
38198         * dbxout.c (dbxout_begin_prologue): Add column argument, adjust
38199         dbxout_source_line caller.
38200         (dbxout_source_line): Add column argument.
38202         * common.opt (gno-column-info, gcolumn-info): New options.
38203         * dwarf2out.c (dwarf2_lineno_debug_hooks): Formatting fix.
38204         (check_die): Also test for multiple DW_AT_decl_column attributes.
38205         (add_src_coords_attributes, dwarf2out_imported_module_or_decl_1): Add
38206         DW_AT_decl_column if requested.
38207         (gen_subprogram_die): Compare and/or add also DW_AT_decl_column
38208         if requested.
38209         (gen_variable_die): Likewise.
38210         (add_call_src_coords_attributes): Add DW_AT_call_column if requested.
38211         * doc/invoke.texi (-gcolumn-info, -gno-column-info): Document.
38213         PR target/79569
38214         * config/i386/i386.opt (m3dnowa): Replace Undocumented with Report.
38215         * common/config/i386/i386-common.c (OPTION_MASK_ISA_3DNOW_A_SET): Define.
38216         (ix86_handle_option): Handle OPT_m3dnowa.
38217         * doc/invoke.texi (-m3dnowa): Document.
38218         * doc/extend.texi (__builtin_ia32_pmulhuw, __builtin_ia32_pf2iw): Use
38219         -m3dnowa instead of -m3dnow -march=athlon.
38221         PR target/79559
38222         * config/i386/i386.c (ix86_print_operand): Use output_operand_lossage
38223         instead of gcc_assert for K, r and R code checks.  Formatting fixes.
38225 2017-02-17  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
38227         PR target/79261
38228         * config/rs6000/rs6000.c (rs6000_expand_ternop_builtin): Add
38229         support for CODE_FOR_vsx_xxpermdi_v2d[fi]_be.
38230         * config/rs6000/rs6000.md (reload_gpr_from_vsx<mode>): Call
38231         generator for vsx_xxpermdi_<mode>_be.
38232         * config/rs6000/vsx.md (vsx_xxpermdi_<mode>): Remove logic to
38233         force big-endian semantics.
38234         (vsx_xxpermdi_<mode>_be): New define_expand with same
38235         implementation as previous version of vsx_xxpermdi_<mode>.
38237 2017-02-17  Jakub Jelinek  <jakub@redhat.com>
38239         PR tree-optimization/79327
38240         * gimple-ssa-sprintf.c (format_integer): Remove likely_adjust
38241         variable, its initialization and use.
38243 2017-02-17  Julia Koval  <julia.koval@intel.com>
38245         * common/config/i386/i386-common.c (OPTION_MASK_ISA_RDPID_SET): New.
38246         (OPTION_MASK_ISA_PKU_UNSET): New.
38247         (ix86_handle_option): Handle -mrdpid.
38248         * config/i386/cpuid.h (bit_RDPID): New.
38249         * config/i386/driver-i386.c (host_detect_local_cpu):
38250         Detect RDPID feature.
38251         * config/i386/i386-builtin.def (__builtin_ia32_rdpid): New.
38252         * config/i386/i386-c.c (ix86_target_macros_internal):
38253         Handle RDPID flag.
38254         * config/i386/i386.c (ix86_target_string): Add -mrdpid to isa2_opts.
38255         (ix86_valid_target_attribute_inner_p): Add "rdpid".
38256         (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
38257         * config/i386/i386.h (TARGET_RDPID, TARGET_RDPID_P): New.
38258         * config/i386/i386.md (define_insn "rdpid"): New.
38259         * config/i386/i386.opt Add -mrdpid.
38260         * config/i386/immintrin.h (_rdpid_u32): New.
38262 2017-02-17  Vladimir Makarov  <vmakarov@redhat.com>
38264         PR rtl-optimization/79541
38265         * lra-constraints.c (curr_insn_transform): Remove wrong asm insn
38266         instead of transforming it into USE.
38268 2017-02-17  Segher Boessenkool  <segher@kernel.crashing.org>
38270         * config/rs6000/rs6000.md (extendsfdf2): Remove default arguments.
38271         If HONOR_SNANS (SFmode) force the input to a register.
38272         (*extendsfdf2_fpr): Add !HONOR_SNANS (SFmode) condition.
38273         (*extendsfdf2_snan): New pattern, used when using SNaNs; it generates
38274         an frsp or similar insn.
38276 2017-02-17  Martin Liska  <mliska@suse.cz>
38278         PR rtl-optimization/79577
38279         * params.def (selsched-max-sched-times): Increase minimum to 1.
38281 2017-02-17  Martin Liska  <mliska@suse.cz>
38283         PR rtl-optimization/79574
38284         * gcse.c (want_to_gcse_p): Prevent integer overflow.
38286 2017-02-17  Martin Liska  <mliska@suse.cz>
38288         PR tree-optimization/79529
38289         * tree-ssa-loop-unswitch.c (is_maybe_undefined): Use
38290         ssa_defined_default_def_p to handle cases which are implicitly
38291         defined.
38292         * tree-ssa.c (ssa_defined_default_def_p): New function.
38293         (ssa_undefined_value_p): Use ssa_defined_default_def_p to handle cases
38294         which are implicitly defined.
38295         * tree-ssa.h (ssa_defined_default_def_p): Declare.
38297 2017-02-17  Richard Biener  <rguenther@suse.de>
38299         PR middle-end/79576
38300         * params.def (max-ssa-name-query-depth): Limit to 10.
38302 2017-02-17  Richard Biener  <rguenther@suse.de>
38304         PR tree-optimization/79552
38305         * tree-ssa-structalias.c (visit_loadstore): Properly verify
38306         default defs.
38308 2017-02-17  Richard Biener  <rguenther@suse.de>
38310         PR bootstrap/79567
38311         * genmatch.c (output_line_directive): Handle DIR_SEPARATOR_2.
38313 2017-02-17  Marek Polacek  <polacek@redhat.com>
38315         PR middle-end/79536
38316         * fold-const.c (fold_negate_expr_1): Renamed from fold_negate_expr.
38317         (fold_negate_expr): New wrapper.
38319 2017-02-16  Sandra Loosemore  <sandra@codesourcery.com>
38321         * doc/invoke.texi (C++ Dialect Options) [-Wno-non-template-friend]: 
38322         Correct terminology and de-emphasize pre-standard behavior.
38324 2017-02-16  Alan Modra  <amodra@gmail.com>
38326         PR rtl-optimization/79286
38327         * ira.c (def_dominates_uses): New function.
38328         (update_equiv_regs): Don't create an equivalence for insns that
38329         may trap where the register def does not dominate the use.
38331 2017-02-16  Vladimir Makarov  <vmakarov@redhat.com>
38333         PR rtl-optimization/78127
38334         * lra.c (lra): Call lra_eliminate before finish the loop after
38335         lra_constraint.
38337 2017-02-16  Richard Biener  <rguenther@suse.de>
38339         * graphite.h: Do not include isl/isl_val_gmp.h, instead include
38340         isl/isl_val.h.
38341         * graphite-isl-ast-to-gimple.c (gmp_cst_to_tree): Remove.
38342         (gcc_expression_from_isl_expr_int): Use generic isl_val interface.
38343         * graphite-sese-to-poly.c: Do not include isl/isl_val_gmp.h.
38344         (isl_val_int_from_wi): New function.
38345         (extract_affine_gmp): Rename to ...
38346         (extract_affine_wi): ... this, take a widest_int.
38347         (extract_affine_int): Just wrap extract_affine_wi.
38348         (add_param_constraints): Use isl_val_int_from_wi.
38349         (add_loop_constraints): Likewise, and extract_affine_wi.
38351 2017-02-15  Jeff Law  <law@redhat.com>
38353         PR middle-end/79521
38354         * ira-costs.c (scan_one_insn): Check have_regs_of_mode before calling
38355         ira_init_register_move_cost_if_necessary.
38357 2017-02-15  Martin Sebor  <msebor@redhat.com>
38359         PR middle-end/32003
38360         * doc/invoke.texi (-fdump-final-insns): Replace option accidentally
38361         removed in a prior commit.
38363 2017-02-15  Bin Cheng  <bin.cheng@arm.com>
38365         PR tree-optimization/79347
38366         * tree-vect-loop-manip.c (vect_do_peeling): Maintain profile
38367         counters during peeling.
38369 2017-02-15  Thomas Schwinge  <thomas@codesourcery.com>
38371         * Makefile.in (site.exp): Remove "set ISLVER".
38373 2017-02-15  Jakub Jelinek  <jakub@redhat.com>
38375         PR target/79487
38376         * real.c (real_from_integer): Call real_convert even for decimal.
38378 2017-02-15  Dominik Vogt  <vogt@linux.vnet.ibm.com>
38380         PR target/79421
38381         * config/s390/s390.c: define TARGET_CUSTOM_FUNCTION_DESCRIPTORS.
38383 2017-02-14  Andrew Pinski  <apinski@cavium.com>
38385         * config/aarch64/aarch64-cores.def (thunderx2t99): Move to under 'C"
38386         cores and change the partno/implementer to be correct.
38387         (thunderx2t99p1): New core which replaces thunderx2t99 and still has
38388         the 'B" as the implementer.
38389         * config/aarch64/aarch64-tune.md: Regenerate.
38391 2017-02-14  Carl Love  <cel@us.ibm.com>
38393         * config/rs6000/rs6000.c: Add case statement entry to make the
38394         xvcvuxdsp built-in argument unsigned.
38395         * config/rs6000/vsx.md: Fix the source and return operand types so they
38396         match the instruction definitions from the ISA document.  Fix typo
38397         in the instruction generation for the (define_insn "vsx_xvcvuxdsp"
38398         statement.
38400 2017-02-14  Vladimir Makarov  <vmakarov@redhat.com>
38402         PR target/79282
38403         * lra-int.h (struct lra_operand_data, struct lra_insn_reg): Add
38404         member early_clobber_alts.
38405         * lra-lives.c (reg_early_clobber_p): New.
38406         (process_bb_lives): Use it.
38407         * lra.c (new_insn_reg): New arg early_clobber_alts.  Use it.
38408         (debug_operand_data): Initialize early_clobber_alts.
38409         (setup_operand_alternative): Set up early_clobber_alts.
38410         (collect_non_operand_hard_regs): Ditto.  Pass early clobber
38411         alternatives to new_insn_reg.
38412         (add_regs_to_insn_regno_info): Add arg early_clobber_alts.  Use
38413         it.
38414         (lra_update_insn_regno_info): Pass the new arg.
38416 2017-02-14  Jakub Jelinek  <jakub@redhat.com>
38418         PR middle-end/79505
38419         * omp-offload.c (free_oacc_loop): Release loop->ifns vector.
38420         (new_oacc_loop_raw): Don't clear already cleared fields.
38422         PR target/79481
38423         * config/i386/avx512pfintrin.h (_mm512_prefetch_i32gather_pd,
38424         _mm512_prefetch_i32gather_ps, _mm512_prefetch_i64gather_pd,
38425         _mm512_prefetch_i64gather_ps): New inline functions and macros.
38427 2017-02-14  Uros Bizjak  <ubizjak@gmail.com>
38429         PR target/79495
38430         * config/i386/i386.md (*movxf_internal): Add (o,rC) alternative.
38432 2017-02-14  H.J. Lu  <hongjiu.lu@intel.com>
38434         PR target/79498
38435         * config/i386/i386.c (timode_scalar_chain::convert_insn): Insert
38436         the extra instruction to the right place to store 128-bit constant
38437         when needed.
38439 2017-02-14  Martin Sebor  <msebor@redhat.com>
38441         PR middle-end/79448
38442         * gimple-ssa-sprintf.c (format_directive): Avoid issuing INT_MAX
38443           warning for strings of unknown length.
38445 2017-02-13  Segher Boessenkool  <segher@kernel.crashing.org>
38447         * config.gcc (supported_defaults) [powerpc*-*-*]: Update.
38449 2017-02-14  Jeff Law  <law@redhat.com>
38451         PR target/79404
38452         * ira-costs.c (scan_one_insn): Initialize register move costs
38453         for pseudos seen in USE/CLOBBER insns.
38455         PR tree-optimization/79095
38456         * tree-vrp.c (extract_range_from_binary_expr_1): For EXACT_DIV_EXPR,
38457         if the numerator has the range ~[0,0] make the resultant range ~[0,0].
38458         (extract_range_from_binary_expr): For MINUS_EXPR with no derived range,
38459         if the operands are known to be not equal, then the resulting range
38460         is ~[0,0].
38461         (intersect_ranges): If the new range is ~[0,0] and the old range is
38462         wide, then prefer ~[0,0].
38463         * tree-vrp.c (overflow_comparison_p_1): New function.
38464         (overflow_comparison_p): New function.
38465         * tree-vrp.c (register_edge_assert_for_2): Register additional asserts
38466         if NAME is used in an overflow test.
38467         (vrp_evaluate_conditional_warnv_with_ops): If the ops represent an
38468         overflow check that can be expressed as an equality test, then adjust
38469         ops to be that equality test.
38471 2017-02-14  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
38473         * config/s390/s390-builtin-types.def: Remove flags argument.
38474         * config/s390/s390.c (s390_init_builtins): Likewise.
38476 2017-02-14  Martin Liska  <mliska@suse.cz>
38478         * tree-ssa-loop-unswitch.c (hoist_guard): Release get_loop_body
38479         vector.  Fix trailing white spaces.
38481 2017-02-14  James Greenhalgh  <james.greenhalgh@arm.com>
38483         * config/aarch64/aarch64.c (aarch64_simd_container_mode): Handle
38484         HFmode.
38486 2017-02-14  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
38488         PR rtl-optimization/68664
38489         * config/arm/arm.c (arm_sched_can_speculate_insn):
38490         New function.  Declare prototype.
38491         (TARGET_SCHED_CAN_SPECULATE_INSN): Define.
38493 2017-02-14  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
38495         PR rtl-optimization/68664
38496         * config/aarch64/aarch64.c (aarch64_sched_can_speculate_insn):
38497         New function.
38498         (TARGET_SCHED_CAN_SPECULATE_INSN): Define.
38500 2017-02-14  Amit Pawar  <amit.pawar@amd.com>
38502         * config/i386/i386.c (znver1_cost): Fix the alignment for function and
38503         max skip bytes for function, loop and jump.
38505 2017-02-14  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
38507         * gimple-pretty-print.c (dump_unary_rhs): Adjust dump format for
38508         ABS_EXPR for gimple dump.
38510 2017-02-14  Jakub Jelinek  <jakub@redhat.com>
38512         PR target/79462
38513         * config/sh/sh.c (expand_cbranchdi4): Don't clear operands[4].
38515         PR tree-optimization/79408
38516         * tree-vrp.c (simplify_div_or_mod_using_ranges): Handle also the
38517         case when on TRUNC_MOD_EXPR op0 is INTEGER_CST.
38518         (simplify_stmt_using_ranges): Call simplify_div_or_mod_using_ranges
38519         also if rhs1 is INTEGER_CST.
38521 2017-02-14  Richard Biener  <rguenther@suse.de>
38523         PR middle-end/79432
38524         * tree-into-ssa.c (insert_phi_nodes): When the function can
38525         have abnormal edges rewrite SSA names with broken use-def
38526         dominance out of SSA and register them for PHI insertion.
38528 2017-02-13  Martin Sebor  <msebor@redhat.com>
38530         PR middle-end/79496
38531         * gimple-ssa-sprintf.c (pass_sprintf_length::handle_gimple_call): Avoid
38532         clearing info.nowrite flag when snprintf size argument is a range.
38534 2017-02-13  Jakub Jelinek  <jakub@redhat.com>
38536         * cprop.c (cprop_jump): Add missing space in string literal.
38537         * tree-ssa-structalias.c (rewrite_constraints): Likewise.
38538         (get_constraint_for_component_ref): Likewise.
38539         * df-core.c (df_worklist_dataflow_doublequeue): Likewise.
38540         * tree-outof-ssa.c (insert_partition_copy_on_edge): Likewise.
38541         * lra-constraints.c (process_alt_operands): Likewise.
38542         * ipa-inline.c (inline_small_functions): Likewise.
38543         * tree-ssa-sccvn.c (visit_reference_op_store): Likewise.
38544         * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
38545         * trans-mem.c (diagnose_tm_1_op): Likewise.
38546         * omp-grid.c (grid_find_single_omp_among_assignments): Likewise.
38547         (grid_parallel_clauses_gridifiable): Likewise.
38549         * config/nvptx/mkoffload.c (process): Add space in between
38550         , and %d.
38552         * config/i386/i386.h (REG_CLASS_NAMES): Add , in between
38553         "MOD4_SSE_REGS" and "ALL_REGS".
38555         * spellcheck.c (test_data): Add , in between "foo" and "food".
38557 2017-02-13  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
38559         PR target/79449
38560         * config/rs6000/rs6000.c (expand_block_compare): Make sure runtime
38561         boundary crossing check and subsequent code generation agree.
38563 2017-02-13  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
38565         * config/aarch64/aarch64.c (has_memory_op): Delete.
38566         (aarch64_madd_needs_nop): Use contains_mem_rtx_p instead of
38567         has_memory_op.
38569 2017-02-13  Jakub Jelinek  <jakub@redhat.com>
38571         PR rtl-optimization/79388
38572         PR rtl-optimization/79450
38573         * combine.c (distribute_notes): When removing TEM_INSN for which
38574         corresponding dest has last value recorded, invalidate that last
38575         value.
38577 2017-02-13  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
38579         * config/arm/arm.c (arm_print_tune_info): Use ASM_COMMENT_START instead
38580         of explicit '@'.  Add missing assembly comment marker on branch costs
38581         printout.
38583 2017-02-13  Nathan Sidwell  <nathan@acm.org>
38585         * gengtype-lex.l (<in_struct>): Add '/'.
38587 2017-02-13  Martin Liska  <mliska@suse.cz>
38589         PR c/79471
38590         * calls.c (expand_call): Replace XALLOCAVEC with XCNEWVEC.
38592 2017-02-13  Richard Biener  <rguenther@suse.de>
38594         * configure.ac (HAVE_ISL_OPTIONS_SET_SCHEDULE_SERIALIZE_SCCS):
38595         Remove.
38596         * configure: Re-generate.
38597         * config.in: Likewise.
38598         * graphite-dependences.c: Simplify as if
38599         HAVE_ISL_OPTIONS_SET_SCHEDULE_SERIALIZE_SCCS was defined.
38600         * graphite-isl-ast-to-gimple.c: Likewise.
38601         * graphite-optimize-isl.c: Likewise.
38602         * graphite-poly.c: Likewise.
38603         * graphite-sese-to-poly.c: Likewise.
38604         * graphite.h: Likewise.
38605         * toplev.c: Include isl/version.h and use isl_version () for
38606         printing the ISL version.
38607         * doc/install.texi: Update ISL requirement.
38609 2017-02-12  Gerald Pfeifer  <gerald@pfeifer.com>
38611         * doc/standards.texi (Standards): Update reference to
38612         Objective-C 2.0.
38614 2017-02-12  Gerald Pfeifer  <gerald@pfeifer.com>
38616         * doc/extend.texi (Named Address Spaces): sourceware.org now
38617         defaults to https.
38618         * doc/install.texi (Binaries): Ditto.
38619         (Specific): Ditto.
38621 2017-02-11  Sandra Loosemore  <sandra@codesourcery.com>
38623         * doc/cpp.texi: Replace "stringify"/"stringification" with C
38624         standard terminology "stringize"/"stringizing" throughout.
38625         * doc/cppinternals.texi: Likewise.
38627 2017-02-11  Sandra Loosemore  <sandra@codesourcery.com>
38629         * doc/extend.texi: Fix some spelling mistakes and typos.
38630         * doc/invoke.texi: Likewise.
38632 2017-02-11  Jan Hubicka  <hubicka@ucw.cz>
38634         PR ipa/79224
38635         * params.def (inline-min-speedup) Change from 10 to 8.
38637 2017-02-11  Jakub Jelinek  <jakub@redhat.com>
38639         * doc/invoke.texi (fopenmp): Bump OpenMP version from 4.0 to
38640         4.5.
38642 2017-02-11  Jan Hubicka  <hubicka@ucw.cz>
38644         PR ipa/79224
38645         * ipa-inline-analysis.c (get_minimal_bb): New function.
38646         (record_modified): Use it.
38647         (remap_edge_change_prob): Handle also ancestor functions.
38649 2017-02-11  Gerald Pfeifer  <gerald@pfeifer.com>
38651         * doc/contrib.texi (Contributors): Remove broken link into
38652         the Mauve CVS repository.
38654 2017-02-11  Jakub Jelinek  <jakub@redhat.com>
38656         PR middle-end/79454
38657         * internal-fn.c (expand_vector_ubsan_overflow): Use piece-wise
38658         result computation whenever lhs doesn't have vector mode, not
38659         just when it has BLKmode.
38661 2017-02-10  Gerald Pfeifer  <gerald@pfeifer.com>
38663         * doc/makefile.texi (profiledbootstrap): Refer to the
38664         installation instructions only in textual form.
38666 2017-02-10  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
38668         PR target/79295
38669         * config/rs6000/altivec.md (bcd<bcd_add_sub>): Fix constraints.
38671 2017-02-10  Gerald Pfeifer  <gerald@pfeifer.com>
38673         * doc/install.texi (Specific): Use https for blackfin.uclinux.org.
38674         (Specific): Update mingw-w64 reference.
38675         (Binaries): Ditto.
38676         (Specific): Remove broken link to Renesas RX processor.
38678 2017-02-10  Richard Biener  <rguenther@suse.de>
38680         * toplev.c (process_options): Do not mention obsolete graphite
38681         options when printing sorry message about missing graphite support.
38682         Mention -floop-nest-optimize.
38684 2017-02-10  Christophe Lyon  <christophe.lyon@linaro.org>
38686         * config/aarch64/arm_neon.h (vtst_p8): Rewrite without asm.
38687         (vtst_p16): Likewise.
38688         (vtstq_p8): Likewise.
38689         (vtstq_p16): Likewise.
38690         (vtst_p64): New.
38691         (vtstq_p64): Likewise.
38692         * config/arm/arm_neon.h (vgetq_lane_p64): New.
38693         (vset_lane_p64): New.
38694         (vsetq_lane_p64): New.
38696 2017-02-10  Jakub Jelinek  <jakub@redhat.com>
38698         PR tree-optimization/79411
38699         * tree-ssa-reassoc.c (is_reassociable_op): Return false if
38700         stmt operands are SSA_NAMEs used in abnormal phis.
38701         (can_reassociate_p): Return false if op is SSA_NAME used in abnormal
38702         phis.
38704 2017-02-09  Jan Hubicka  <hubicka@ucw.cz>
38706         PR ipa/70795
38707         * cgraphunit.c (cgraph_node::add_new_function): Set externally_visible
38708         flag if needed.
38710 2017-02-09  Jan Hubicka  <hubicka@ucw.cz>
38712         * tree-ssa-loop-unswitch.c (hoist_guard): Update profile.
38714 2017-02-09  Jakub Jelinek  <jakub@redhat.com>
38716         * omp-offload.c (oacc_loop_auto_partitions): Use || instead of |
38717         to avoid warning.
38719         PR c/79413
38720         * gimplify.h (is_gimple_sizepos): Only test for INTEGER_CST constants,
38721         not arbitrary TREE_CONSTANT.
38723         PR c/79431
38724         * gimplify.c (gimplify_adjust_omp_clauses): Ignore
38725         "omp declare target link" attribute unless is_global_var.
38726         * omp-offload.c (find_link_var_op): Likewise.
38728 2017-02-09  Nathan Sidwell  <nathan@codesourcery.com>
38729             Chung-Lin Tang  <cltang@codesourcery.com>
38731         * gimplify.c (gimplify_scan_omp_clauses): No special handling for
38732         OMP_CLAUSE_TILE.
38733         (gimplify_adjust_omp_clauses): Don't delete TILE.
38734         (gimplify_omp_for): Deal with TILE.
38735         * internal-fn.c (expand_GOACC_TILE): New function.
38736         * internal-fn.def (GOACC_DIM_POS): Comment may be overly conservative.
38737         (GOACC_TILE): New.
38738         * omp-expand.c (struct oacc_collapse): Add tile and outer fields.
38739         (expand_oacc_collapse_init): Add LOC paramter.  Initialize tile
38740         element fields.
38741         (expand_oacc_collapse_vars): Add INNER parm, adjust for tiling,
38742         avoid DIV for outermost collapse var.
38743         (expand_oacc_for): Insert tile element loop as needed.  Adjust.
38744         Remove out of date comments, fix whitespace.
38745         * omp-general.c (omp_extract_for_data): Deal with tiling.
38746         * omp-general.h (enum oacc_loop_flags): Add OLF_TILE flag,
38747         adjust OLF_DIM_BASE value.
38748         (struct omp_for_data): Add tiling field.
38749         * omp-low.c (scan_sharing_clauses): Allow OMP_CLAUSE_TILE.
38750         (lower_oacc_head_mark): Add OLF_TILE as appropriate.  Ensure 2 levels
38751         for auto loops.  Remove default auto determining, moved to
38752         oacc_loop_fixed_partitions.
38753         * omp-offload.c (struct oacc_loop): Change 'ifns' to vector of call
38754         stmts, add e_mask field.
38755         (oacc_dim_call): New function, abstracted out from oacc_thread_numbers.
38756         (oacc_thread_numbers): Use oacc_dim_call.
38757         (oacc_xform_tile): New.
38758         (new_oacc_loop_raw): Initialize e_mask, adjust for ifns vector.
38759         (finish_oacc_loop): Adjust for ifns vector.
38760         (oacc_loop_discover_walk): Append loop abstraction sites to list,
38761         add case for GOACC_TILE fns.
38762         (oacc_loop_xform_loop): Delete.
38763         (oacc_loop_process): Iterate over call list directly, and add
38764         handling for GOACC_TILE fns.
38765         (oacc_loop_fixed_partitions): Determine default auto, deal with TILE,
38766         dump partitioning.
38767         (oacc_loop_auto_partitions): Add outer_assign parm. Assign all but
38768         vector partitioning to outer loops.  Assign 2 partitions to loops
38769         when available. Add TILE handling.
38770         (oacc_loop_partition): Adjust oacc_loop_auto_partitions call.
38771         (execite_oacc_device_lower): Process GOACC_TILE fns,
38772         ignore unknown specs.
38773         * tree-nested.c (convert_nonlocal_omp_clauses): Allow OMP_CLAUSE_TILE.
38774         * tree.c (omp_clause_num_ops): Adjust TILE ops.
38775         * tree.h (OMP_CLAUSE_TILE_ITERVAR, OMP_CLAUSE_TILE_COUNT): New.
38777 2017-02-09  Gerald Pfeifer  <gerald@pfeifer.com>
38779         * configure.ac (ACX_BUGURL): Update.
38780         * configure: Regenerate.
38782 2017-02-09  Richard Biener  <rguenther@suse.de>
38784         PR tree-optimization/69823
38785         * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
38786         Properly enumerate all BBs in the region.  Use auto_vec/auto_bitmap.
38788 2017-02-09  Andrew Burgess  <andrew.burgess@embecosm.com>
38790         * config/arc/arc-c.def: Add __NPS400__ definition.
38791         * config/arc/arc.h (CPP_SPEC): Don't define __NPS400__ here.
38792         (TARGET_NPS400): Define.
38794 2017-02-09  Andrew Burgess  <andrew.burgess@embecosm.com>
38796         * config/arc/arc-arch.h (arc_arch_t): Move unchanged to earlier in
38797         file.
38798         (arc_cpu_t): Change base_architecture field, arch, to a arc_arc_t
38799         pointer, arch_info.
38800         (arc_cpu_types): Fill the arch_info field with a pointer into the
38801         arc_arch_types table.
38802         (arc_selected_cpu): Declare.
38803         * config/arc/arc.c (arc_selected_cpu): Make global.
38804         (arc_selected_arch): Delete.
38805         (arc_base_cpu): Delete.
38806         (arc_override_options): Remove references to deleted variables,
38807         update access to arch information.
38808         (ARC_OPT): Update access to arch information.
38809         (ARC_OPTX): Likewise.
38810         * config/arc/arc.h (arc_base_cpu): Remove declaration.
38811         (TARGET_ARC600): Update access to arch information.
38812         (TARGET_ARC601): Likewise.
38813         (TARGET_ARC700): Likewise.
38814         (TARGET_EM): Likewise.
38815         (TARGET_HS): Likewise.
38816         * config/arc/driver-arc.c (arc_cpu_to_as): Update access to arch
38817         information.
38819 2017-02-08  Pat Haugen  <pthaugen@us.ibm.com>
38821         PR target/78604
38822         * config/rs6000/rs6000.c (rs6000_emit_vector_cond_expr): Invert
38823         condition/operands for integer GE/LE/GEU/LEU operations.
38825 2017-02-08  Segher Boessenkool  <segher@kernel.crashing.org>
38827         PR translation/79397
38828         * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Fix spelling
38829         of AltiVec.
38831 2017-02-08  Martin Jambor  <mjambor@suse.cz>
38833         PR ipa/79375
38834         * ipa-prop.c (ipa_alloc_node_params): Make static, return bool
38835         whether allocation happened.
38836         (ipa_initialize_node_params): Do not call ipa_alloc_node_params if
38837         nothing was allocated.
38839 2017-02-08  Jakub Jelinek  <jakub@redhat.com>
38841         PR tree-optimization/79408
38842         * tree-vrp.c (simplify_div_or_mod_using_ranges): If op1 is not
38843         constant, but SSA_NAME with a known integer range, use the minimum
38844         of that range instead of op1 to determine if modulo can be replaced
38845         with its first operand.
38847 2016-02-08  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
38849         * config/riscv/riscv.c (riscv_build_integer_1): Avoid use of INT16_MAX.
38851 2017-02-08  Richard Biener  <rguenther@suse.de>
38853         PR tree-optimization/71824
38854         * graphite-scop-detection.c (scop_detection::build_scop_breadth):
38855         Check all loops contained in the merged region.
38857 2017-02-07  Andrew Pinski  <apinski@cavium.com>
38859         * config/aarch64/aarch64.md (popcount<mode>2): New pattern.
38861 2017-02-07  Andrew Pinski  <apinski@cavium.com>
38863         * config/aarch64/aarch64-cores.def (thunderx): Disable LSE.
38864         (thunderxt88): Likewise.
38865         (thunderxt81): Disable LSE and change v8.1 to v8.
38866         (thunderxt83): Likewise.
38868 2017-02-07  Jakub Jelinek  <jakub@redhat.com>
38869             Richard Biener  <rguenther@suse.de>
38871         PR middle-end/79399
38872         * ira-int.h (struct target_ira_int): Change x_max_struct_costs_size
38873         type from int to size_t.
38874         * ira-costs.c (struct_costs_size): Change type from int to size_t.
38876 2017-02-07  Jakub Jelinek  <jakub@redhat.com>
38878         PR rtl-optimization/79386
38879         * cprop.c (bypass_conditional_jumps): Initialize
38880         bypass_last_basic_block already before splitting bbs after
38881         unconditional traps...
38882         (bypass_conditional_jumps): ... rather than here.
38884         PR target/79299
38885         * config/i386/sse.md (xtg_mode, gatherq_mode): New mode attrs.
38886         (*avx512f_gathersi<mode>, *avx512f_gathersi<mode>_2,
38887         *avx512f_gatherdi<mode>, *avx512f_gatherdi<mode>_2): Use them,
38888         fix -masm=intel patterns.
38890 2017-02-07  Richard Biener  <rguenther@suse.de>
38892         PR tree-optimization/79256
38893         PR middle-end/79278
38894         * builtins.c (get_object_alignment_2): Use min_align_of_type
38895         to extract alignment for MEM_REFs to honor BIGGEST_FIELD_ALIGNMENT
38896         and ADJUST_FIELD_ALIGN.
38898         * doc/tm.texi.in (ADJUST_FIELD_ALIGN): Adjust to take additional
38899         type parameter.
38900         * doc/tm.texi: Regenerate.
38901         * stor-layout.c (layout_decl): Adjust.
38902         (update_alignment_for_field): Likewise.
38903         (place_field): Likewise.
38904         (min_align_of_type): Likewise.
38905         * config/arc/arc.h (ADJUST_FIELD_ALIGN): Adjust.
38906         * config/epiphany/epiphany.h (ADJUST_FIELD_ALIGN): Likewise.
38907         * config/epiphany/epiphany.c (epiphany_adjust_field_align): Likewise.
38908         * config/frv/frv.h (ADJUST_FIELD_ALIGN): Likewise.
38909         * config/frv/frv.c (frv_adjust_field_align): Likewise.
38910         * config/i386/i386.h (ADJUST_FIELD_ALIGN): Likewise.
38911         * config/i386/i386.c (x86_field_alignment): Likewise.
38912         * config/rs6000/aix.h (ADJUST_FIELD_ALIGN): Likewise.
38913         * config/rs6000/darwin.h (ADJUST_FIELD_ALIGN): Likewise.
38914         * config/rs6000/freebsd64.h (ADJUST_FIELD_ALIGN): Likewise.
38915         * config/rs6000/linux64.h (ADJUST_FIELD_ALIGN): Likewise.
38916         * config/rs6000/sysv4.h (ADJUST_FIELD_ALIGN): Likewise.
38917         * config/rs6000/rs6000.c (rs6000_special_adjust_field_align_p):
38918          Likewise.
38920         Revert
38921         2017-01-30  Richard Biener  <rguenther@suse.de>
38923         PR tree-optimization/79256
38924         * targhooks.c (default_builtin_vector_alignment_reachable): Honor
38925         BIGGEST_FIELD_ALIGNMENT and ADJUST_FIELD_ALIGN to fix up bogus
38926         alignment on TYPE.
38928 2017-02-07  Toma Tabacu  <toma.tabacu@imgtec.com>
38930         * config/mips/mips.c (mips_expand_builtin_insn): Convert the QImode
38931         argument of the pshufh, psllh, psllw, psrah, psraw, psrlh, psrlw
38932         builtins to SImode and emit a zero-extend, if necessary.
38934 2017-02-06  Palmer Dabbelt  <palmer@dabbelt.com>
38936         * docs/invoke.texi (RISC-V Options): Alphabetize.
38938 2017-02-06  Palmer Dabbelt  <palmer@dabbelt.com>
38940         * doc/invoke.texi (RISC-V Options): Use two spaces to separate
38941         options.
38943 2017-02-06  Palmer Dabbelt  <palmer@dabbelt.com>
38945         * config/riscv/riscv.c: New file.
38946         * common/config/riscv/riscv-common.c: Likewise.
38947         * config.gcc: Likewise.
38948         * config/riscv/constraints.md: Likewise.
38949         * config/riscv/elf.h: Likewise.
38950         * config/riscv/generic.md: Likewise.
38951         * config/riscv/linux.h: Likewise.
38952         * config/riscv/multilib-generator: Likewise.
38953         * config/riscv/peephole.md: Likewise.
38954         * config/riscv/pic.md: Likewise.
38955         * config/riscv/predicates.md: Likewise.
38956         * config/riscv/riscv-builtins.c: Likewise.
38957         * config/riscv/riscv-c.c: Likewise.
38958         * config/riscv/riscv-ftypes.def: Likewise.
38959         * config/riscv/riscv-modes.def: Likewise.
38960         * config/riscv/riscv-opts.h: Likewise.
38961         * config/riscv/riscv-protos.h: Likewise.
38962         * config/riscv/riscv.h: Likewise.
38963         * config/riscv/riscv.md: Likewise.
38964         * config/riscv/riscv.opt: Likewise.
38965         * config/riscv/sync.md: Likewise.
38966         * config/riscv/t-elf-multilib: Likewise.
38967         * config/riscv/t-linux: Likewise.
38968         * config/riscv/t-linux-multilib: Likewise.
38969         * config/riscv/t-riscv: Likewise.
38970         * configure.ac: Likewise.
38971         * doc/contrib.texi: Add Kito Cheng, Palmer Dabbelt, and Andrew
38972         Waterman as RISC-V maintainers.
38973         * doc/install.texi: Add RISC-V entries.
38974         * doc/invoke.texi: Add RISC-V options section.
38975         * doc/md.texi: Add RISC-V constraints section.
38976         * configure: Regenerated.
38978 2017-02-06  Michael Meissner  <meissner@linux.vnet.ibm.com>
38980         PR target/66144
38981         * config/rs6000/vector.md (vcond<mode><mode>): Allow the true and
38982         false values to be constant vectors with all 0 or all 1 bits set.
38983         (vcondu<mode><mode>): Likewise.
38984         * config/rs6000/predicates.md (vector_int_reg_or_same_bit): New
38985         predicate.
38986         (fpmask_comparison_operator): Update comment.
38987         (vecint_comparison_operator): New predicate.
38988         * config/rs6000/rs6000.c (rs6000_emit_vector_cond_expr): Optimize
38989         vector conditionals when the true and false values are constant
38990         vectors with all 0 bits or all 1 bits set.
38992 2017-02-06  Martin Sebor  <msebor@redhat.com>
38994         PR  tree-optimization/79376
38995         * gimple-fold.c (get_range_strlen): Set the minimum length to zero.
38997 2017-02-06  Uros Bizjak  <ubizjak@gmail.com>
38999         * config/i386/sse.md (vector modes -> vec_extract* splitter): Use
39000         explicit subreg RTX with operand 1.  Use VECTOR_MODE_P predicate
39001         to simplify split condition.
39003 2017-02-06  Jakub Jelinek  <jakub@redhat.com>
39005         * omp-expand.c (oxpand_omp_atomic_fetch_op,
39006         expand_omp_atomic_pipeline): Return false if can_atomic_load_p is
39007         false.
39009 2017-02-06  Segher Boessenkool  <segher@kernel.crashing.org>
39011         PR rtl-optimization/68664
39012         * target.def (can_speculate_insn): New hook.
39013         * doc/tm.texi.in (TARGET_SCHED_CAN_SPECULATE_INSN): New hook.
39014         * doc/tm.texi: Regenerate.
39015         * sched-rgn.c (can_schedule_ready_p): Use the new hook.
39016         * config/rs6000/rs6000.c (TARGET_SCHED_CAN_SPECULATE_INSN): New macro.
39017         (rs6000_sched_can_speculate_insn): New function.
39019 2017-02-06  Jakub Jelinek  <jakub@redhat.com>
39021         PR tree-optimization/79284
39022         * tree-vectorizer.h (VECT_SCALAR_BOOLEAN_TYPE_P): Define.
39023         * tree-vect-stmts.c (vect_get_vec_def_for_operand,
39024         vectorizable_mask_load_store, vectorizable_operation,
39025         vect_is_simple_cond, get_same_sized_vectype): Use it instead
39026         of comparing TREE_CODE of a type against BOOLEAN_TYPE.
39027         * tree-vect-patterns.c (check_bool_pattern, search_type_for_mask_1,
39028         vect_recog_bool_pattern, vect_recog_mask_conversion_pattern): Likewise.
39029         * tree-vect-slp.c (vect_get_constant_vectors): Likewise.
39030         * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
39031         Remove redundant gimple_code (stmt) == GIMPLE_ASSIGN test after
39032         is_gimple_assign (stmt).  Replace another such test with
39033         is_gimple_assign (stmt).
39035 2017-02-06  Georg-Johann Lay  <avr@gjlay.de>
39037         PR target/78883
39038         * config/avr/avr.c (rtl-iter.h): Include it.
39039         (TARGET_LEGITIMATE_COMBINED_INSN): New hook define...
39040         (avr_legitimate_combined_insn): ...and implementation.
39042 2017-02-06  Dominik Vogt  <vogt@linux.vnet.ibm.com>
39044         * config/s390/predicates.md ("larl_operand"): Use macros from hwint.h.
39045         * config/s390/s390.c (s390_const_operand_ok)
39046         (s390_canonicalize_comparison, s390_extract_part)
39047         (s390_single_part, s390_contiguous_bitmask_nowrap_p)
39048         (s390_contiguous_bitmask_p, s390_rtx_costs)
39049         (legitimize_pic_address): Likewise.
39050         * config/s390/s390.md ("clzdi2", "clztidi2"): Likewise.
39051         * config/s390/vx-builtins.md ("vec_genbytemaskv16qi")
39052         ("vec_permi<mode>", "vfae<mode>", "*vfaes<mode>", "vstrc<mode>")
39053         ("*vstrcs<mode>"): Use UINTVAL() to set unsigned HOST_WIDE_INT.
39054         * config/s390/vector.md ("vec_vfenes<mode>"): Likewise.
39056 2017-02-06  Georg-Johann Lay  <avr@gjlay.de>
39058         * config/avr/avr.md (*addhi3_zero_extend): Add alternative where
39059         REGNO($0) == REGNO($1).
39061 2017-02-06  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
39063         * config/s390/linux.h(SIZE_TYPE): Add comment.
39065 2017-02-06  Julian Brown  <julian@codesourcery.com>
39066             Naveen H.S  <Naveen.Hurugalawadi@cavium.com>
39067             Virendra Pathak  <virendra.pathak@broadcom.com>
39069         * config/aarch64/aarch64-cores.def: Change the scheduler
39070         to Thunderx2t99.
39071         * config/aarch64/aarch64.md: Include thunderx2t99.md.
39072         * config/aarch64/thunderx2t99.md: New file.
39074 2017-02-05  Gerald Pfeifer  <gerald@pfeifer.com>
39076         * doc/standards.texi (Go Language): Update link to language
39077         standard.
39079 2017-02-05  Jan Hubicka  <hubicka@ucw.cz>
39081         * tree-eh.c (lower_resx): Sanitize profile.
39082         (cleanup_empty_eh_move_lp): Likewise.
39084 2017-02-05  Jan Hubicka  <hubicka@ucw.cz>
39086         PR tree-ssa/79347
39087         * cfgloopmanip.c (lv_adjust_loop_entry_edge, loop_version): Add
39088         ELSE_PROB.
39089         * cfgloopmanip.h (loop_version): Update prototype.
39090         * modulo-sched.c (sms_schedule): Update call of loop_version.
39091         * tree-if-conv.c(version_loop_for_if_conversion): Likewise.
39092         * tree-parloops.c (gen_parallel_loop): Likewise.
39093         * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Likewise.
39094         * tree-ssa-loop-split.c (split_loop): Likewise.
39095         * tree-ssa-loop-unswitch.c (tree_unswitch_loop): Likewise.
39096         * tree-vect-loop-manip.c (vect_loop_versioning): Likewise.
39098 2017-02-05  Martin Liska  <mliska@suse.cz>
39100         PR bootstrap/78985
39101         * config/s390/s390.c (s390_gimplify_va_arg): Initialize local
39102         variable to NULL.
39103         (print_operand_address): Initialize a struct to zero.
39105 2017-02-05  Gerald Pfeifer  <gerald@pfeifer.com>
39107         * doc/contrib.texi (Contributors): Refer to Hans Boehm's
39108         garbage collector only in textual form.
39110 2017-02-05  Gerald Pfeifer  <gerald@pfeifer.com>
39112         * doc/extend.texi (x86 specific memory model extensions for
39113         transactional memory): Simplify a phrase.
39115 2017-02-05  Eric Botcazou  <ebotcazou@adacore.com>
39117         PR target/79353
39118         * config/sparc/sync.md (atomic_loaddi_1): Replace 'U' constraint with
39119         'r', 'm' constraint with 'T' and !TARGET_ARCH64 with TARGET_ARCH32.
39120         (atomic_storedi_1): Likewise.
39122 2017-02-04  Jakub Jelinek  <jakub@redhat.com>
39124         PR tree-optimization/79338
39125         * tree-parloops.c (gather_scalar_reductions): Don't call
39126         vect_analyze_loop_form for loop->inner before destroying loop's
39127         loop_vinfo.
39129 2017-02-03  Martin Sebor  <msebor@redhat.com>
39131         PR tree-optimization/79327
39132         * gimple-ssa-sprintf.c (tree_digits): Avoid adding the base prefix
39133         when precision has resulted in leading zeros.
39134         (format_integer): Adjust the likely counter to assume an unknown
39135         argument that may be zero is non-zero.
39137 2017-02-03  Jason Merrill  <jason@redhat.com>
39139         PR c++/78689
39140         * tree-inline.c (copy_tree_body_r) [COND_EXPR]: Revert change to
39141         avoid copying non-taken branch.
39143 2017-02-03  Jakub Jelinek  <jakub@redhat.com>
39145         PR tree-optimization/79340
39146         * tree-vect-loop.c (vectorizable_reduction): Release
39147         vec_defs elements after safe_splicing them into other vectors.
39148         Formatting fixes.
39150         PR tree-optimization/79327
39151         * gimple-ssa-sprintf.c (adjust_range_for_overflow): If returning
39152         true, always set *argmin and *argmax to TYPE_{MIN,MAX}_VALUE of
39153         dirtype.
39154         (format_integer): Use wide_int_to_tree instead of build_int_cst
39155         + to_?hwi.  If argmin is NULL, just set argmin and argmax to
39156         TYPE_{MIN,MAX}_VALUE of argtype.  Simplify and fix computation
39157         of shortest and longest sequence.
39159 2017-02-03  Uros Bizjak  <ubizjak@gmail.com>
39161         * config/i386/i386.c (dimode_scalar_chain::convert_reg):
39162         Use pextrd for TARGET_SSE4_1 when creating scalar copy.
39164 2017-02-03  Walter Lee  <walt@tilera.com>
39166         PR target/78862
39167         * config/tilegx/tilegx.md (tilegx_expand_prologue): Add blockage
39168         after initial stackframe link reg save.
39169         * config/tilepro/tilepro.md (tilepro_expand_prologue): Likewise.
39171 2017-02-03  Jakub Jelinek  <jakub@redhat.com>
39173         PR target/79354
39174         * config/rs6000/rs6000.md (movsi_from_sf): Use wb constraint instead of
39175         wu for stxssp alternative.
39177 2017-02-03  Martin Sebor  <msebor@redhat.com>
39179         PR tree-optimization/79352
39180         * gimple-fold.c (get_range_strlen): Add argument.
39181         (get_range_strlen): Change return type to bool.
39182         (get_maxval_strlen): Pass in a dummy argument.
39183         * gimple-fold.h (get_range_strlen): Change return type to bool.
39184         * gimple-ssa-sprintf.c (get_string_length): Set unlikely counter.
39185         * tree.h (array_at_struct_end_p): Add argument.
39186         * tree.c (array_at_struct_end_p): Handle it.
39188 2017-02-03  Martin Liska  <mliska@suse.cz>
39190         PR lto/66295
39191         * multiple_target.c (create_dispatcher_calls): Redirect edge
39192         from a caller of a dispatcher.
39193         (expand_target_clones): Make the clones local.
39194         (ipa_target_clone): Do both target clones and resolvers.
39195         (ipa_dispatcher_calls): Remove the pass.
39196         (pass_dispatcher_calls::gate): Likewise.
39197         (make_pass_dispatcher_calls): Likewise.
39198         * passes.def (pass_target_clone): Put as very first IPA early
39199         pass.
39201 2017-02-03  Martin Liska  <mliska@suse.cz>
39203         * symtab.c (symtab_node::binds_to_current_def_p): Bail out
39204         in case of a function with ifunc attribute.
39206 2017-02-03  Martin Liska  <mliska@suse.cz>
39208         * cgraph.c (cgraph_node::dump): Dump function version info.
39209         * symtab.c (symtab_node::dump_base): Add missing new line.
39211 2017-02-02  Jan Hubicka  <hubicka@ucw.cz>
39213         * tree-ssa-ifcombine.c (update_profile_after_ifcombine): New function.
39214         (ifcombine_ifandif): Use it.
39216 2017-02-03  Martin Liska  <mliska@suse.cz>
39218         * doc/invoke.texi: Document default value for
39219         use-after-scope-direct-emission-threshold.
39221 2017-02-03  Martin Liska  <mliska@suse.cz>
39223         PR tree-optimization/79339
39224         * gimple-ssa-sprintf.c (format_floating_max): Call mpfr_clear.
39225         (format_floating): Likewise.
39227 2017-02-03  Martin Liska  <mliska@suse.cz>
39229         PR ipa/79337
39230         * ipa-prop.c (ipa_node_params_t::insert): Remove current
39231         implementation.
39232         (ipa_node_params_t::remove): Likewise.
39233         * ipa-prop.h (ipa_node_params::ipa_node_params): Make default
39234         initialization from removed ipa_node_params_t::insert.
39235         (ipa_node_params::~ipa_node_params): Move from removed
39236         ipa_node_params_t::release.
39237         * symbol-summary.h (symbol_summary::m_released): New member.
39238         Do not release a summary twice.  Do not allow to call finalizer
39239         for types of a summary that live in GGC memory.
39241 2017-02-02  Naveen H.S  <Naveen.Hurugalawadi@cavium.com>
39243         * config/aarch64/aarch64.c (thunderx2t99_tunings): Enable AES and
39244         cmp_branch fusion.
39246 2017-02-02  Martin Sebor  <msebor@redhat.com>
39248         PR middle-end/79275
39249         * gimple-ssa-sprintf.c (get_string_length): Set lower bound to zero.
39250         (format_string): Tighten up the range of output for non-constant
39251         strings and correct the expected range for wide non-constant strings.
39253 2017-02-02  Martin Sebor  <msebor@redhat.com>
39255         * doc/invoke.texi (-maccumulate-args): Fix bad grammar.
39257         PR middle-end/32003
39258         * doc/invoke.texi (-fdump-tree-): Remove pass-specific options from
39259         index.
39260         (-fdump-tree-@var): Add to index and document how to come up
39261         with pass-specific option and dump file names.
39262         (-fdump-passes): Clarify where to look for output.
39264 2017-02-02  Jan Hubicka  <hubicka@ucw.cz>
39266         PR middle-end/77445
39267         * tree-ssa-threadbackward.c (profitable_jump_thread_path): Dump
39268         statistics of the analyzed path; allow threading for speed when
39269         any of BBs along the path are optimized for speed.
39271 2017-02-02  Eric Botcazou  <ebotcazou@adacore.com>
39273         PR middle-end/78468
39274         * emit-rtl.c (init_emit): Add ??? comment for problematic alignment
39275         settings of the virtual registers.
39277         Revert again
39278         2016-08-23  Dominik Vogt  <vogt@linux.vnet.ibm.com>
39280         * explow.c (get_dynamic_stack_size): Take known alignment of stack
39281         pointer + STACK_DYNAMIC_OFFSET into account when calculating the size
39282         needed.
39284 2017-02-02  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
39286         * config/s390/vx-builtins.md ("vec_ceil", "vec_floor")
39287         ("vec_trunc", "vec_roundc", "vec_round"): Remove expanders.
39289 2017-02-02  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
39291         * config/s390/s390.md: Add missing comments with the expanded
39292         mnemonics.
39293         * config/s390/vector.md: Likewise.
39294         * config/s390/vx-builtins.md: Likewise.
39296 2017-02-02  Jakub Jelinek  <jakub@redhat.com>
39298         PR target/79197
39299         * config/rs6000/rs6000.md (*fixuns_trunc<mode>di2_fctiduz): Rename to ...
39300         (fixuns_trunc<mode>di2): ... this, remove previous expander.  Put all
39301         conditions on a single line.
39303 2017-02-02  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
39305         * config/s390/s390-c.c (s390_cpu_cpp_builtins_internal): Rename
39306         __S390_VX__ to __VX__.
39308 2017-02-01  Andrew Pinski  <apinski@cavium.com>
39310         * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost): Pass
39311         stmt_info to record_stmt_cost.
39312         (vect_get_known_peeling_cost): Pass stmt_info if known to
39313         record_stmt_cost.
39314         * config/aarch64/aarch64-protos.h (cpu_vector_cost): Split
39315         cpu_vector_cost field into
39316         scalar_int_stmt_cost and scalar_fp_stmt_cost.  Split vec_stmt_cost
39317         field into vec_int_stmt_cost and vec_fp_stmt_cost.
39318         * config/aarch64/aarch64.c (generic_vector_cost): Update for the
39319         splitting of scalar_stmt_cost and vec_stmt_cost.
39320         (thunderx_vector_cost): Likewise.
39321         (cortexa57_vector_cost): LIkewise.
39322         (exynosm1_vector_cost): Likewise.
39323         (xgene1_vector_cost): Likewise.
39324         (thunderx2t99_vector_cost): Improve after the splitting of the two
39325         fields.
39326         (aarch64_builtin_vectorization_cost): Update for the splitting of
39327         scalar_stmt_cost and vec_stmt_cost.
39329 2017-02-01  Torvald Riegel  <triegel@redhat.com>
39330             Richard Henderson  <rth@redhat.com>
39332         * builtins.c (fold_builtin_atomic_always_lock_free): Make "lock-free"
39333         conditional on existance of a fast atomic load.
39334         * optabs-query.c (can_atomic_load_p): New function.
39335         * optabs-query.h (can_atomic_load_p): Declare it.
39336         * optabs.c (expand_atomic_exchange): Always delegate to libatomic if
39337         no fast atomic load is available for the particular size of access.
39338         (expand_atomic_compare_and_swap): Likewise.
39339         (expand_atomic_load): Likewise.
39340         (expand_atomic_store): Likewise.
39341         (expand_atomic_fetch_op): Likewise.
39342         * testsuite/lib/target-supports.exp
39343         (check_effective_target_sync_int_128): Remove x86 because it provides
39344         no fast atomic load.
39345         (check_effective_target_sync_int_128_runtime): Likewise.
39347 2017-02-01  Richard Biener  <rguenther@suse.de>
39349         * graphite.c: Include tree-vectorizer.h for find_loop_location.
39350         (graphite_transform_loops): Provide opt-info for optimized nests.
39351         * tree-parloop.c (parallelize_loops): Provide opt-info for
39352         parallelized loops.
39354 2017-02-01  Richard Biener  <rguenther@suse.de>
39356         PR middle-end/79315
39357         * tree-cfg.c (move_stmt_op): Never set TREE_BLOCK when it
39358         was not set before.
39360 2017-02-01  Richard Biener  <rguenther@suse.de>
39362         PR tree-optimization/71824
39363         * graphite-scop-detection.c (scop_detection::build_scop_breadth):
39364         Verify the loops are valid in the merged SESE region.
39365         (scop_detection::can_represent_loop_1): Check analyzing the
39366         evolution of the number of iterations in the region succeeds.
39368 2017-01-31  Ian Lance Taylor  <iant@golang.org>
39370         * config/i386/i386.c (ix86_expand_split_stack_prologue): Add
39371         REG_ARGS_SIZE note to 32-bit push insns and call insn.
39373 2017-01-31  David Malcolm  <dmalcolm@redhat.com>
39375         PR preprocessor/79210
39376         * input.c (get_substring_ranges_for_loc): Replace line_width
39377         assertion with error-handling.
39379 2017-01-31  Richard Biener  <rguenther@suse.de>
39381         PR tree-optimization/77318
39382         * graphite-sese-to-poly.c (extract_affine): Fix assert.
39383         (create_pw_aff_from_tree): Take loop parameter.
39384         (add_condition_to_pbb): Pass loop of the condition to
39385         create_pw_aff_from_tree.
39387 2017-01-31  Jakub Jelinek  <jakub@redhat.com>
39389         * config/s390/s390.c (s390_asan_shadow_offset): New function.
39390         (TARGET_ASAN_SHADOW_OFFSET): Redefine.
39392 2017-01-31  Michael Meissner  <meissner@linux.vnet.ibm.com>
39394         PR target/78597
39395         PR target/79038
39396         * config/rs6000/rs6000-protos.h (convert_float128_to_int): Delete,
39397         no longer used.
39398         (convert_int_to_float128): Likewise.
39399         * config/rs6000/rs6000.c (convert_float128_to_int): Likewise.
39400         (convert_int_to_float128): Likewise.
39401         * config/rs6000/rs6000.md (UNSPEC_IEEE128_MOVE): Likewise.
39402         (UNSPEC_IEEE128_CONVERT): Likewise.
39403         (floatsi<mode>2, FLOAT128 iterator): Bypass calling
39404         rs6000_expand_float128_convert if we have IEEE 128-bit hardware.
39405         Use local variables for IBM extended format.
39406         (fix_trunc<mode>si2, FLOAT128 iterator): Likewise.
39407         (fix_trunc<mode>si2_fprs): Likewise.
39408         (fixuns_trunc<IEEE128:mode><SDI:mode>2): Likewise.
39409         (floatuns<IEEE128:mode>2, IEEE128 iterator): Likewise.
39410         (fix<uns>_<mode>si2_hw): Rework the IEEE 128-bt hardware support
39411         to know that we can now have integers of all sizes in vector
39412         registers.
39413         (fix<uns>_<mode>di2_hw): Likewise.
39414         (float<uns>_<mode>si2_hw): Likewise.
39415         (fix_<mode>si2_hw): Likewise.
39416         (fixuns_<mode>si2_hw): Likewise.
39417         (float<uns>_<mode>di2_hw): Likewise.
39418         (float_<mode>di2_hw): Likewise.
39419         (float_<mode>si2_hw): Likewise.
39420         (floatuns_<mode>di2_hw): Likewise.
39421         (floatuns_<mode>si2_hw): Likewise.
39422         (xscvqp<su>wz_<mode>): Delete, no longer used.
39423         (xscvqp<su>dz_<mode>): Likewise.
39424         (xscv<su>dqp_<mode>): Likewise.
39425         (ieee128_mfvsrd_64bit): Likewise.
39426         (ieee128_mfvsrd_32bit): Likewise.
39427         (ieee128_mfvsrwz): Likewise.
39428         (ieee128_mtvsrw): Likewise.
39429         (ieee128_mtvsrd_64bit): Likewise.
39430         (ieee128_mtvsrd_32bit): Likewise.
39432 2017-01-31  Martin Liska  <mliska@suse.cz>
39434         PR ipa/79285
39435         * ipa-prop.c (ipa_free_all_node_params): Call release method
39436         instead of ~sumbol_summary to not to trigger double times
39437         dtor of hash_map.
39439 2017-01-31  Aldy Hernandez  <aldyh@redhat.com>
39441         PR tree-optimization/71691
39442         * bitmap.h (class auto_bitmap): New.
39443         * tree-ssa-loop-unswitch.c (tree_may_unswitch_on): Call
39444         is_maybe_undefined instead of ssa_undefined_value_p.
39446 2017-01-31  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
39448         * config/s390/s390-c.c (s390_cpu_cpp_builtins_internal): Rename
39449         __S390_ARCH_LEVEL__ to __ARCH__.
39451 2017-01-31  Jakub Jelinek  <jakub@redhat.com>
39453         PR tree-optimization/79267
39454         * value-prof.c (gimple_ic): Only drop lhs for noreturn calls
39455         if should_remove_lhs_p is true.
39457 2017-01-30  Alexandre Oliva  <aoliva@redhat.com>
39459         PR debug/63238
39460         * dwarf2out.c (clone_as_declaration): Drop DW_AT_alignment.
39461         (add_alignment_attribute): New.
39462         (base_type_die): Add alignment attribute.
39463         (subrange_type_die): Likewise.
39464         (modified_type_die): Likewise.
39465         (gen_array_type_die): Likewise.
39466         (gen_descr_array_type_die: Likewise.
39467         (gen_enumeration_type_die): Likewise.
39468         (gen_subprogram_die): Likewise.
39469         (gen_variable_die): Likewise.
39470         (gen_field_die): Likewise.
39471         (gen_ptr_to_mbr_type_die): Likewise.
39472         (gen_struct_or_union_type_die): Likewise.
39473         (gen_subroutine_type_die): Likewise.
39474         (gen_typedef_die): Likewise.
39475         (base_type_cmp): Compare alignment attribute.
39477 2017-01-30  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
39479         PR target/79170
39480         * config/rs6000/altivec.md (*setb_internal): Rename to setb_signed.
39481         (setb_unsigned) New pattern for setb with CCUNS.
39482         * config/rs6000/rs6000.c (expand_block_compare): Use a different
39483         subfc./subfe sequence to avoid overflow problems.  Generate a
39484         shorter sequence with cmpld/setb for power9.
39485         * config/rs6000/rs6000.md (subf<mode>3_carry_dot2): Add a new pattern
39486         for generating subfc. instruction.
39487         (cmpstrsi): Add TARGET_POPCNTD predicate as the generate sequence
39488         now uses this instruction.
39490 2017-01-30  Ian Lance Taylor  <iant@google.com>
39492         PR debug/79289
39493         * dwarf2out.c (gen_type_die_with_usage): When picking a variant
39494         for FUNCTION_TYPE/METHOD_TYPE, use the first matching one.
39496 2017-01-30  Martin Sebor  <msebor@redhat.com>
39498         * gimple-ssa-sprintf.c (fmtresult::adjust_for_width_or_precision):
39499         Move constant to the right of a relational operator.
39500         (get_mpfr_format_length, format_character, format_string): Ditto.
39501         (should_warn_p, maybe_warn): Same.
39503         * doc/invoke.texi (-Wformat-truncation=1): Fix typo.
39505 2017-01-30  Maxim Ostapenko  <m.ostapenko@samsung.com>
39507         PR lto/79061
39508         * asan.c (get_translation_unit_decl): Remove function.
39509         (asan_add_global): Force has_dynamic_init to zero in LTO mode.
39511 2017-01-30  Martin Liska  <mliska@suse.cz>
39513         PR gcov-profile/79259
39514         * opts.c (common_handle_option): Enable flag_ipa_bit_cp w/
39515         -fprofile-generate.
39517 2017-01-30  Martin Liska  <mliska@suse.cz>
39519         PR bootstrap/78985
39520         * config/aarch64/cortex-a57-fma-steering.c (func_fma_steering::analyze):
39521         Initialize variables with NULL value.
39523 2017-01-30  Richard Earnshaw  <rearnsh@arm.com>
39525         PR target/79260
39526         * config.gcc (arm*-*-*): Add arm/arm-flags.h and arm/arm-isa.h to
39527         tm_p_file.
39528         * arm/arm-protos.h: Don't directly include arm-flags.h and arm-isa.h.
39530 2017-01-30  Richard Biener  <rguenther@suse.de>
39532         PR tree-optimization/79276
39533         * tree-vrp.c (process_assert_insertions): Properly adjust common
39534         when removing a duplicate.
39536 2017-01-30  Richard Biener  <rguenther@suse.de>
39538         PR tree-optimization/79256
39539         * targhooks.c (default_builtin_vector_alignment_reachable): Honor
39540         BIGGEST_FIELD_ALIGNMENT and ADJUST_FIELD_ALIGN to fix up bogus
39541         alignment on TYPE.
39542         * tree.c (build_aligned_type): Set TYPE_USER_ALIGN.
39544 2017-01-30  Dominik Vogt  <vogt@linux.vnet.ibm.com>
39546         PR target/79240
39547         * config/s390/s390.md ("*r<noxa>sbg_<mode>_srl_bitmask")
39548         ("*r<noxa>sbg_<mode>_sll_bitmask")
39549         ("*extzv_<mode>_srl<clobbercc_or_nocc>")
39550         ("*extzv_<mode>_sll<clobbercc_or_nocc>"):
39551         Use contiguous_bitmask_nowrap_operand.
39553 2017-01-29  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
39555         PR target/79268
39556         * config/rs6000/altivec.h (vec_xl): Revise #define.
39557         (vec_xst): Likewise.
39559 2017-01-27  Uros Bizjak  <ubizjak@gmail.com>
39561         * config/i386/i386.c (print_reg): Use REGNO instead of true_regnum.
39563 2017-01-27  Bernd Schmidt  <bschmidt@redhat.com>
39565         PR rtl-optimization/79194
39566         * cprop.c (one_cprop_pass): Move deletion of code after unconditional
39567         traps before call to bypass_conditional_jumps.
39569 2017-01-27  Vladimir Makarov  <vmakarov@redhat.com>
39571         PR tree-optimization/71374
39572         * lra-constraints.c (check_conflict_input_operands): New.
39573         (match_reload): Use it.
39575 2017-01-27  Vladimir Makarov  <vmakarov@redhat.com>
39577         PR target/79131
39578         * lra-assigns.c (find_hard_regno_for_1): Take endianess for into
39579         account to calculate conflict_set.
39581 2017-01-27  Bin Cheng  <bin.cheng@arm.com>
39583         PR rtl-optimization/78559
39584         * combine.c (try_combine): Discard REG_EQUAL and REG_EQUIV for
39585         other_insn in combine.
39587 2017-01-27  Pekka Jääskeläinen  <pekka.jaaskelainen@parmance.com>
39589         * builtin-types.def: Use unsigned_char_type_node for BT_UINT8.  Use
39590         uint16_type_node for BT_UINT16.
39592 2017-01-27  David Malcolm  <dmalcolm@redhat.com>
39594         * doc/sourcebuild.texi (Testsuites): Add "GIMPLE Tests" and
39595         "RTL Tests" to menu.
39596         (GIMPLE Tests): New node.
39597         (RTL Tests): New node.
39599 2017-01-27  Richard Biener  <rguenther@suse.de>
39601         PR tree-optimization/79245
39602         * tree-loop-distribution.c (distribute_loop): Apply cost
39603         modeling also to detected patterns.
39605 2017-01-27  Richard Biener  <rguenther@suse.de>
39607         PR tree-optimization/71433
39608         * tree-vrp.c (register_new_assert_for): Revert earlier changes.
39609         (compare_assert_loc): New function.
39610         (process_assert_insertions): Sort and optimize assert locations
39611         to remove duplicates and push down identical assertions on
39612         edges to their destination block.
39614 2017-01-27  Richard Biener  <rguenther@suse.de>
39616         PR tree-optimization/79244
39617         * tree-vrp.c (remove_range_assertions): Forcefully propagate
39618         out SSA names even if abnormal.
39620 2017-01-27  Jakub Jelinek  <jakub@redhat.com>
39622         * realmpfr.h: Poison MPFR_RND{N,Z,U,D}.
39623         * gimple-ssa-sprintf.c (format_floating_max): Use GMP_RNDN
39624         instead of MPFR_RNDN.
39626 2017-01-27  Richard Earnshaw  <rearnsha@arm.com>
39628         PR target/79239
39629         * arm.c (arm_option_override): Don't call build_target_option_node
39630         until after doing all option overrides.
39631         (arm_valid_target_attribute_tree): Likewise.
39633 2017-01-27  Martin Liska  <mliska@suse.cz>
39635         * doc/invoke.texi (-fprofile-arcs): Document profiling support
39636         for {cd}tors and C++ {cd}tors.
39638 2017-01-27  Dominik Vogt  <vogt@linux.vnet.ibm.com>
39640         * config/s390/s390.md ("*setmem_long_and")
39641         ("*setmem_long_and_31z"): Use zero_extend instead of and.
39643 2017-01-26  Martin Sebor  <msebor@redhat.com>
39645         * gimple-ssa-sprintf.c (format_floating): Simplify the computation
39646         of precision.
39648 2017-01-26  Martin Sebor  <msebor@redhat.com>
39650         * gimple-ssa-sprintf.c (format_floating): Test HAVE_XFmode and
39651         HAVE_DFmode before using XFmode or DFmode.
39652         (parse_directive): Avoid using the z length modifier to avoid
39653         the ISO C++98 does not support the â€˜z’ gnu_printf length modifier.
39655         PR middle-end/78703
39656         * gimple-ssa-sprintf.c (adjust_for_width_or_precision): Change
39657         to accept adjustment as an array.
39658         (get_int_range): New function.
39659         (struct directive): Make width and prec arrays.
39660         (directive::set_width, directive::set_precision): Call get_int_range.
39661         (format_integer, format_floating): Handle width and precision ranges.
39662         (format_string, parse_directive): Same.
39664 2017-01-26  Jakub Jelinek  <jakub@redhat.com>
39666         PR debug/79129
39667         * dwarf2out.c (generate_skeleton_bottom_up): For children with
39668         comdat_type_p set, just clone them, but keep the children in the
39669         original DIE.
39671         PR debug/78835
39672         * dwarf2out.c (prune_unused_types): Mark all functions with DIEs
39673         which have direct callers with -fvar-tracking-assignments enabled
39674         in the current TU.
39675         (resolve_addr): Avoid adding skeleton DIEs for DW_AT_call_origin
39676         inside of type units.
39678 2017-01-26  Martin Sebor  <msebor@redhat.com>
39680         PR middle-end/78703
39681         * gimple-ssa-sprintf.c (struct result_range): Add likely and
39682         unlikely counters.
39683         (struct format_result): Replace number_chars, number_chars_min,
39684         and number_chars_max with a single member of struct result_range.
39685         Remove bounded.
39686         (format_result::operator+=): Adjust.
39687         (struct fmtresult): Remove bounded.  Handle likely and unlikely
39688         counters.
39689         (fmtresult::adjust_for_width_or_precision): New function.
39690         (fmtresult:type_max_digits): New function.
39691         (bytes_remaining): Handle likely and unlikely counters.
39692         (min_bytes_remaining): Remove.
39693         (format_percent): Simplify.
39694         (format_integer, format_floating): Set likely and unlikely counters.
39695         (get_string_length, format_character, format_string): Same.
39696         (format_plain, should_warn_p): New function.
39697         (maybe_warn): Call should_warn_p.  Update diagnostic messages
39698         and handle those for all directives, including plain strings.
39699         (format_directive): Handle likely and unlikely counters.
39700         Remove unnecessary quoting from diagnostics.  Add an informational
39701         note.
39702         (add_bytes): Remove.
39703         (pass_sprintf_length::compute_format_length): Simplify.
39704         (try_substitute_return_value): Handle likely and unlikely counters.
39706 2017-01-26  Carl Love  <cel@us.ibm.com>
39708         * config/rs6000/rs6000-c (altivec_overloaded_builtins): Remove
39709         bogus entries for the P8V_BUILTIN_VEC_VGBBD built-ins
39711 2017-01-26  Vladimir Makarov  <vmakarov@redhat.com>
39713         PR target/79131
39714         * lra-assigns.c (setup_live_pseudos_and_spill_after_risky): Take
39715         endianess for subregs into account.
39716         * lra-constraints.c (lra_constraints): Do risky transformations
39717         always on the first iteration.
39718         * lra-lives.c (check_pseudos_live_through_calls): Add arg
39719         last_call_used_reg_set.
39720         (process_bb_lives): Define and use last_call_used_reg_set.
39721         * lra.c (lra): Always continue after lra_constraints on the first
39722         iteration.
39724 2017-01-26  Kirill Yukhin  <kirill.yukhin@gmail.com>
39726         * gcc.target/i386/avx512bw-kshiftlq-2.c: Use unsigned long long
39727         constant.
39728         * gcc.target/i386/avx512bw-kshiftrq-2.c: Ditto.
39730 2017-01-26  Jakub Jelinek  <jakub@redhat.com>
39732         * config/i386/avx512fintrin.h (_ktest_mask16_u8,
39733         _ktestz_mask16_u8, _ktestc_mask16_u8, _kadd_mask16): Move to ...
39734         * config/i386/avx512dqintrin.h (_ktest_mask16_u8,
39735         _ktestz_mask16_u8, _ktestc_mask16_u8, _kadd_mask16): ... here.
39736         * config/i386/i386-builtin.def (__builtin_ia32_ktestchi,
39737         __builtin_ia32_ktestzhi, __builtin_ia32_kaddhi): Use
39738         OPTION_MASK_ISA_AVX512DQ instead of OPTION_MASK_ISA_AVX512F.
39739         * config/i386/sse.md (SWI1248_AVX512BWDQ2): New mode iterator.
39740         (kadd<mode>, ktest<mode>): Use it instead of SWI1248_AVX512BWDQ.
39742 2017-01-26  Marek Polacek  <polacek@redhat.com>
39744         PR c/79199
39745         * fold-const.c (operand_equal_p) [COND_EXPR]: Use OP_SAME_WITH_NULL
39746         for the third operand.
39748 2017-01-26  Jakub Jelinek  <jakub@redhat.com>
39750         PR middle-end/79236
39751         * omp-low.c (struct omp_context): Add simt_stmt field.
39752         (scan_omp_for): Return omp_context *.
39753         (scan_omp_simd): Set simt_stmt on the non-_simt_ SIMD
39754         context to the _simt_ SIMD stmt.
39755         (lower_omp_for): For combined SIMD with sibling _simt_
39756         SIMD, make sure to use the same decls in _looptemp_
39757         clauses as in the sibling.
39759 2017-01-26  David Sherwood  <david.sherwood@arm.com>
39761         PR middle-end/79212
39762         * gimplify.c (omp_notice_variable): Add GOVD_SEEN flag to variables in
39763         all contexts.
39765 2017-01-26  Jakub Jelinek  <jakub@redhat.com>
39767         PR target/70465
39768         * reg-stack.c (emit_swap_insn): Instead of fld a; fld b; fxchg %st(1);
39769         emit fld b; fld a; if possible.
39771         * brig-builtins.def: Update copyright years.
39772         * config/arm/arm_acle_builtins.def: Update copyright years.
39774 2017-01-25  Michael Meissner  <meissner@linux.vnet.ibm.com>
39776         PR target/79179
39777         * config/rs6000/vsx.md (vsx_extract_<mode>_store): Use wY
39778         constraint instead of o for the stxsd instruction.
39780 2017-01-25  Carl Love  <cel@us.ibm.com>
39782         * config/rs6000/rs6000-c (altivec_overloaded_builtins): Fix order
39783         of entries for ALTIVEC_BUILTIN_VEC_PACKS and P8V_BUILTIN_VEC_VGBBD.
39785 2017-01-25  Jonathan Wakely  <jwakely@redhat.com>
39787         * doc/invoke.texi (C++ Dialect Options): Fix typo.
39789 2017-01-25  Richard Biener  <rguenther@suse.de>
39791         PR tree-optimization/69264
39792         * target.def (vector_alignment_reachable): Improve documentation.
39793         * doc/tm.texi: Regenerate.
39794         * targhooks.c (default_builtin_vector_alignment_reachable): Simplify
39795         and add a comment.
39796         * tree-vect-data-refs.c (vect_supportable_dr_alignment): Revert
39797         earlier changes with respect to TYPE_USER_ALIGN.
39798         (vector_alignment_reachable_p): Likewise.  Improve dumping.
39800 2016-01-25  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
39802         PR target/79145
39803         * config/arm/arm.md (xordi3): Force constant operand into a register
39804         for TARGET_IWMMXT.
39806 2016-01-25  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
39808         * doc/invoke.texi (-fstore-merging): Correct default optimization
39809         levels at which it is enabled.
39810         (-O): Move -fstore-merging from list to...
39811         (-O2): ... Here.
39813 2017-01-25  Richard Biener  <rguenther@suse.de>
39815         PR debug/78363
39816         * omp-expand.c: Include debug.h.
39817         (expand_omp_taskreg): Make sure to generate early debug before
39818         outlining anything from a function.
39819         (expand_omp_target): Likewise.
39820         (grid_expand_target_grid_body): Likewise.
39822 2017-01-25  Maxim Ostapenko  <m.ostapenko@samsung.com>
39824         PR lto/79061
39825         * asan.c (get_translation_unit_decl): New function.
39826         (asan_add_global): Extract modules file name from globals
39827         TRANSLATION_UNIT_DECL name.
39829 2017-01-24  Eric Botcazou  <ebotcazou@adacore.com>
39831         PR target/77439
39832         * config/arm/arm.c (arm_function_ok_for_sibcall): Add back restriction
39833         for long calls with APCS frame and VFP.
39835 2017-01-24  David Malcolm  <dmalcolm@redhat.com>
39837         * cfg.c (original_copy_tables_initialized_p): New function.
39838         * cfg.h (original_copy_tables_initialized_p): New decl.
39839         * cfgrtl.c (relink_block_chain): Guard the call to
39840         free_original_copy_tables with a call to
39841         original_copy_tables_initialized_p.
39842         * cgraph.h (symtab_node::native_rtl_p): New decl.
39843         * cgraphunit.c (symtab_node::native_rtl_p): New function.
39844         (symtab_node::needed_p): Don't assert for early assembly output
39845         for __RTL functions.
39846         (cgraph_node::finalize_function): Set "force_output" for __RTL
39847         functions.
39848         (cgraph_node::analyze): Bail out early for __RTL functions.
39849         (analyze_functions): Update assertion to support __RTL functions.
39850         (cgraph_node::expand): Bail out early for __RTL functions.
39851         * final.c (rest_of_clean_state): Don't call delete_tree_ssa for
39852         __RTL functions.
39853         * function.h (struct function): Update comment for field
39854         "pass_startwith".
39855         * gimple-expr.c: Include "tree-pass.h".
39856         (gimple_has_body_p): Return false for __RTL functions.
39857         * Makefile.in (OBJS): Add run-rtl-passes.o.
39858         * pass_manager.h (gcc::pass_manager::get_rest_of_compilation): New
39859         accessor.
39860         (gcc::pass_manager::get_clean_slate): New accessor.
39861         * passes.c: Include "insn-addr.h".
39862         (should_skip_pass_p): Add logging.  Update logic for running
39863         "expand" to be compatible with both __GIMPLE and __RTL.  Guard
39864         property-provider override so it is only done for gimple passes.
39865         Don't skip dfinit.
39866         (skip_pass): New function.
39867         (execute_one_pass): Call skip_pass when skipping passes.
39868         * read-md.c (md_reader::read_char): Support filtering
39869         the input to a subset of line numbers.
39870         (md_reader::md_reader): Initialize fields
39871         m_first_line and m_last_line.
39872         (md_reader::read_file_fragment): New function.
39873         * read-md.h (md_reader::read_file_fragment): New decl.
39874         (md_reader::m_first_line): New field.
39875         (md_reader::m_last_line): New field.
39876         * read-rtl-function.c (function_reader::create_function): Only
39877         create cfun if it doesn't already exist.  Set PROP_rtl on cfun's
39878         curr_properties.  Set DECL_INITIAL to a dummy block.
39879         (read_rtl_function_body_from_file_range): New function.
39880         * read-rtl-function.h (read_rtl_function_body_from_file_range):
39881         New decl.
39882         * run-rtl-passes.c: New file.
39883         * run-rtl-passes.h: New file.
39885 2017-01-24  Jeff Law  <law@redhat.com>
39887         * config/microblaze/microblaze.h (ASM_FORMAT_PRIVATE_NAME): Increase
39888         buffer size.
39890 2017-01-24  Bin Cheng  <bin.cheng@arm.com>
39892         PR tree-optimization/79159
39893         * tree-ssa-loop-niter.c (get_cst_init_from_scev): New function.
39894         (record_nonwrapping_iv): Improve boundary using above function if no
39895         value range information.
39897 2017-01-24  Pekka Jääskeläinen  <pekka@parmance.com>
39898             Martin Jambor  <mjambor@suse.cz>
39900         * brig-builtins.def: New file.
39901         * builtins.def (DEF_HSAIL_BUILTIN): New macro.
39902         (DEF_HSAIL_ATOMIC_BUILTIN): Likewise.
39903         (DEF_HSAIL_SAT_BUILTIN): Likewise.
39904         (DEF_HSAIL_INTR_BUILTIN): Likewise.
39905         (DEF_HSAIL_CVT_ZEROI_SAT_BUILTIN): Likewise.
39906         * builtin-types.def (BT_INT8): New.
39907         (BT_INT16): Likewise.
39908         (BT_UINT8): Likewise.
39909         (BT_UINT16): Likewise.
39910         (BT_FN_ULONG): Likewise.
39911         (BT_FN_UINT_INT): Likewise.
39912         (BT_FN_UINT_ULONG): Likewise.
39913         (BT_FN_UINT_LONG): Likewise.
39914         (BT_FN_UINT_PTR): Likewise.
39915         (BT_FN_ULONG_PTR): Likewise.
39916         (BT_FN_INT8_FLOAT): Likewise.
39917         (BT_FN_INT16_FLOAT): Likewise.
39918         (BT_FN_UINT32_FLOAT): Likewise.
39919         (BT_FN_UINT16_FLOAT): Likewise.
39920         (BT_FN_UINT8_FLOAT): Likewise.
39921         (BT_FN_UINT64_FLOAT): Likewise.
39922         (BT_FN_UINT16_UINT32): Likewise.
39923         (BT_FN_UINT32_UINT16): Likewise.
39924         (BT_FN_UINT16_UINT16_UINT16): Likewise.
39925         (BT_FN_INT_PTR_INT): Likewise.
39926         (BT_FN_UINT_PTR_UINT): Likewise.
39927         (BT_FN_LONG_PTR_LONG): Likewise.
39928         (BT_FN_ULONG_PTR_ULONG): Likewise.
39929         (BT_FN_VOID_UINT64_UINT64): Likewise.
39930         (BT_FN_UINT8_UINT8_UINT8): Likewise.
39931         (BT_FN_INT8_INT8_INT8): Likewise.
39932         (BT_FN_INT16_INT16_INT16): Likewise.
39933         (BT_FN_INT_INT_INT): Likewise.
39934         (BT_FN_UINT_FLOAT_UINT): Likewise.
39935         (BT_FN_FLOAT_UINT_UINT): Likewise.
39936         (BT_FN_ULONG_UINT_UINT): Likewise.
39937         (BT_FN_ULONG_UINT_PTR): Likewise.
39938         (BT_FN_ULONG_ULONG_ULONG): Likewise.
39939         (BT_FN_UINT_UINT_UINT): Likewise.
39940         (BT_FN_VOID_UINT_PTR): Likewise.
39941         (BT_FN_UINT_UINT_PTR: Likewise.
39942         (BT_FN_UINT32_UINT64_PTR): Likewise.
39943         (BT_FN_INT_INT_UINT_UINT): Likewise.
39944         (BT_FN_UINT_UINT_UINT_UINT): Likewise.
39945         (BT_FN_UINT_UINT_UINT_PTR): Likewise.
39946         (BT_FN_UINT_ULONG_ULONG_UINT): Likewise.
39947         (BT_FN_ULONG_ULONG_ULONG_ULONG): Likewise.
39948         (BT_FN_LONG_LONG_UINT_UINT): Likewise.
39949         (BT_FN_ULONG_ULONG_UINT_UINT): Likewise.
39950         (BT_FN_VOID_UINT32_UINT64_PTR): Likewise.
39951         (BT_FN_VOID_UINT32_UINT32_PTR): Likewise.
39952         (BT_FN_UINT_UINT_UINT_UINT_UINT): Likewise.
39953         (BT_FN_UINT_FLOAT_FLOAT_FLOAT_FLOAT): Likewise.
39954         (BT_FN_ULONG_ULONG_ULONG_UINT_UINT): Likewise.
39955         * doc/frontends.texi: List BRIG FE.
39956         * doc/install.texi (Testing): Add BRIG tesring requirements.
39957         * doc/invoke.texi (Overall Options): Mention BRIG.
39958         * doc/standards.texi (Standards): Doucment BRIG HSA version.
39960 2017-01-24  Richard Biener  <rguenther@suse.de>
39962         PR translation/79208
39963         * ipa-devirt.c (odr_types_equivalent_p): Fix typo in diagnostic.
39965 2017-01-24  Martin Jambor  <mjambor@suse.cz>
39967         PR bootstrap/79198
39968         * ipa-prop.c (ipa_free_all_node_params): Call summary destructor.
39969         * ipa-prop.c (ipa_node_params_t::insert): Initialize fields known_csts
39970         and known_contexts.
39972 2017-01-24  Aldy Hernandez  <aldyh@redhat.com>
39974         PR middle-end/79123
39975         * gimple-ssa-warn-alloca.c (alloca_call_type): Make sure
39976         casts from signed to unsigned really don't have a range.
39978 2017-01-24  Markus Trippelsdorf  <markus@trippelsdorf.de>
39980         * gimple-ssa-sprintf.c (format_floating): Change MPFR_RNDx to
39981         GMP_RNDx for compatiblity.
39983 2017-01-24  Martin Liska  <mliska@suse.cz>
39985         PR bootstrap/79132
39986         * tree-ssa-reassoc.c (rewrite_expr_tree_parallel): Insert assert
39987         that would prevent us to call alloca with -1 as argument.
39989 2017-01-24  Jakub Jelinek  <jakub@redhat.com>
39991         * dwarf2out.c (output_compilation_unit_header, output_file_names):
39992         Avoid -Wformat-security warning.
39994 2017-01-23  Andrew Pinski  <apinski@cavium.com>
39996         * config/aarch64/aarch64.c (thunderx2t99_addrcost_table): Improve
39997         cost table.
39999 2017-01-23  Martin Sebor  <msebor@redhat.com>
40001         PR middle-end/78703
40002         * gimple-ssa-sprintf.c (warn_level): New global.
40003         (format_integer): Use it here and throughout the rest of the file.
40004         Use the same switch to compute sign as base.
40005         (maybe_warn): New function.
40006         (format_directive): Factor out warnings into maybe_warn.
40007         Add debugging output.  Use warn_level.
40008         (add_bytes): Use warn_level.
40009         (pass_sprintf_length::compute_format_length): Add debugging output.
40010         (try_substitute_return_value): Same.
40011         (pass_sprintf_length::handle_gimple_call): Set and use warn_level.
40013         PR middle-end/78703
40014         * gimple-ssa-sprintf.c (struct format_result): Remove constant member.
40015         (struct fmtresult, format_integer, format_floating): Adjust.
40016         (fmtresult::fmtresult): Set max correctly in two argument ctor.
40017         (get_string_length, format_string,format_directive): Same.
40018         (pass_sprintf_length::compute_format_length): Same.
40019         (try_substitute_return_value): Simplify slightly.
40021         PR middle-end/78703
40022         * gimple-ssa-sprintf.c (pass_sprintf_length::gate): Adjust formatting.
40023         (fmtresult::operator+=): Outlined.
40024         (struct fmtresult): Add ctors.
40025         (struct conversion_spec): Rename...
40026         (struct directive): ...to this.  Add and remove data members.
40027         (directive::set_width, directive::set_precision): New functions.
40028         (format_percent): Use fmtresult ctor.
40029         (get_width_and_precision): Remove.
40030         (format_integer): Make naming changes.  Avoid computing width and
40031         precision.
40032         (format_floating): Same.  Adjust indentation.
40033         (format_character, format_none): New functions.
40034         (format_string): Moved character handling to format_character.
40035         (format_directive): Remove arguments, change return type.
40036         (parse_directive): New function.
40037         (pass_sprintf_length::compute_format_length): Move directive
40038         parsing to parse_directive.
40040 2017-01-23  Jakub Jelinek  <jakub@redhat.com>
40042         * tree.h (assign_assembler_name_if_neeeded): Rename to ...
40043         (assign_assembler_name_if_needed): ... this.
40044         * tree.c (assign_assembler_name_if_neeeded): Rename to ...
40045         (assign_assembler_name_if_needed): ... this.
40046         (free_lang_data_in_cgraph): Adjust callers.
40047         * cgraphunit.c (cgraph_node::analyze): Likewise.
40048         * omp-expand.c (expand_omp_taskreg, expand_omp_target):
40049         Likewise.
40051 2017-01-23  Richard Biener  <rguenther@suse.de>
40053         PR tree-optimization/79088
40054         PR tree-optimization/79188
40055         * tree-ssa-threadupdate.c (mark_threaded_blocks): Move code
40056         resetting loop bounds after last path deletion.  Reset loop
40057         bounds of the target loop, make code match the comments.
40058         * tree-ssa-threadbackwards.c (pass_early_thread_jumps::execute):
40059         Make sure loops need no fixups.
40061 2017-01-23  Kelvin Nilsen  <kelvin@gcc.gnu.org>
40063         * config/rs6000/rs6000-builtin.def (VSIEDPF): Add scalar insert
40064         exponent support with double type for first argument.
40065         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Changed
40066         type returned by __builtin_vec_extract_sig,
40067         __builtin_vec_extract_sig_sp, and __builtin_vec_extract_sig_dp
40068         functions from "vector int" to "vector unsigned int" or from
40069         "vector long long int" to "vector unsigned long long int".
40070         Changed type returned by __builtin_vec_extract_exp,
40071         __builtin_vec_extract_exp_sp, and __builtin_vec_extract_exp_dp
40072         functions from "vector int" to "vector unsigned int" or from
40073         "vector long long int" to "vector unsigned long long int".
40074         Changed return type of __builtin_vec_test_data_class,
40075         __builtin_vec_test_data_class_sp, and
40076         __builtin_vec_test_data_class_dp from "vector int" to
40077         "vector bool int" or from "vector long long int" to "vector bool
40078         long long int" and changed second argument type from "unsigned
40079         int" to "int".  Added new overloaded function forms "vector float
40080         __builtin_vec_insert_exp (vector float, vector unsigned int)" and
40081         "vector float __builtin_vec_insert_exp_sp (vector float, vector
40082         unsigned int)" and "vector double __builtin_vec_insert_exp (vector
40083         double, vector unsigned long long int)" and "vector double
40084         __builtin_vec_insert_exp_dp (vector double, vector unsigned long
40085         long int)".  Changed return type of
40086         __builtin_scalar_test_data_class and
40087         __builtin_scalar_test_data_class_sp and
40088         __builtin_scalar_test_data_class_dp from "unsigned int" to "bool
40089         int" and changed second argument from "unsigned int" to "int".
40090         Changed type returned by __builtin_scalar_test_neg,
40091         __builtin_scalar_test_neg_sp, and __builtin_scalar_test_neg_dp
40092         from "int" to "bool int".  Added new overloaded function form
40093         "double __builtin_scalar_insert_exp (double, unsigned long long int)".
40094         * config/rs6000/vsx.md (xsiexpdpf): New insn for scalar insert
40095         exponent double-precision with floating point first argument.
40096         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Adjust
40097         documentation of scalar_test_data_class, scalar_test_neg,
40098         scalar_extract_sig, scalar_extract_exp, scalar_insert_exp,
40099         vector_extract_exp, vec_extract_sig, vec_insert_exp, and
40100         vec_test_data_class built-in functions to reflect refinements in
40101         their type signatures.
40103 2017-01-23  Andreas Tobler  <andreast@gcc.gnu.org>
40105         * config/aarch64/aarch64.c (aarch64_elf_asm_constructor): Increase
40106         size of buf.
40107         (aarch64_elf_asm_destructor): Likewise.
40109 2017-01-23  Bernd Schmidt  <bschmidt@redhat.com>
40111         PR rtl-optimization/78634
40112         * config/i386/i386.c (ix86_max_noce_ifcvt_seq_cost): New function.
40113         (TARGET_MAX_NOCE_IFCVT_SEQ_COST): Define.
40114         * ifcvt.c (noce_try_cmove): Add missing cost check.
40116         PR rtl-optimization/71724
40117         * combine.c (if_then_else_cond): Look for situations where it is
40118         beneficial to undo the work of one of the recursive calls.
40120 2017-01-23  Bin Cheng  <bin.cheng@arm.com>
40122         PR tree-optimization/70754
40123         * tree-predcom.c (stmt_combining_refs): New parameter INSERT_BEFORE.
40124         (reassociate_to_the_same_stmt): New parameter INSERT_BEFORE.  Insert
40125         combined stmt before it if not NULL.
40126         (combine_chains): Process refs reversely and compute dominance point
40127         for root ref.
40129 2017-01-23  Martin Liska  <mliska@suse.cz>
40131         PR tree-optimization/79196
40132         * tree-ssa-strlen.c (fold_strstr_to_memcmp): Rename to ...
40133         (fold_strstr_to_strncmp): ... this.  Fold the pattern to strncmp
40134         instead of memcmp.
40135         (strlen_optimize_stmt): Call the renamed function.
40137 2017-01-23  Michael Matz  <matz@suse.de>
40139         PR tree-optimization/78384
40140         * tree-ssa-loop-split.c (patch_loop_exit): Use correct edge.
40142 2017-01-23  Richard Biener  <rguenther@suse.de>
40144         PR tree-optimization/79186
40145         * tree-vrp.c (register_new_assert_for): Make sure we've seen
40146         both incoming edges before moving an assert.
40148 2017-01-23  Martin Jambor  <mjambor@suse.cz>
40150         * ipa-prop.c (load_from_param_1): Removed.
40151         (load_from_unmodified_param): Bits from load_from_param_1 put back
40152         here.
40153         (load_from_param): Removed.
40154         (compute_complex_assign_jump_func): Removed stmt2 and just replaced it
40155         with stmt.  Reverted back to use of load_from_unmodified_param.
40157 2017-01-23  Martin Jambor  <mjambor@suse.cz>
40159         PR ipa/79108
40160         * ipa-prop.h (ipa_param_descriptor): Anotate with with GTY(()).
40161         (ipa_node_params): Annotate with GTY((for_user)).  Make descriptors
40162         field a pointer to garbage collected vector, mark lattices and
40163         ipcp_orig_node with GTY((skip)).
40164         (ipa_get_param_count): Adjust to descriptors being a pointer.
40165         (ipa_get_param): Likewise.
40166         (ipa_get_type): Likewise.
40167         (ipa_get_param_move_cost): Likewise.
40168         (ipa_set_param_used): Likewise.
40169         (ipa_get_controlled_uses): Likewise.
40170         (ipa_set_controlled_uses): Likewise.
40171         (ipa_is_param_used): Likewise.
40172         (ipa_node_params_t): Move into garbage collector.  New methods insert
40173         and remove.
40174         (ipa_node_params_sum): Annotate wth GTY(()).
40175         (ipa_check_create_node_params): Adjust to ipa_node_params_sum being
40176         garbage collected.
40177         (ipa_load_from_parm_agg): Adjust declaration.
40178         * ipa-icf.c (param_used_p): Adjust to descriptors being a pointer.
40179         * ipa-profile.c (ipa_profile): Likewise.
40180         * ipa-prop.c (ipa_get_param_decl_index_1): Likewise.
40181         (ipa_populate_param_decls): Make descriptors parameter garbage
40182         collected.
40183         (ipa_dump_param): Adjust to descriptors being a pointer.
40184         (ipa_alloc_node_params): Likewise.
40185         (ipa_initialize_node_params): Likewise.
40186         (load_from_param_1): Make descriptors parameter garbage collected.
40187         (load_from_unmodified_param): Likewise.
40188         (load_from_param): Likewise.
40189         (ipa_load_from_parm_agg): Likewise.
40190         (ipa_node_params::~ipa_node_params): Removed.
40191         (ipa_free_all_node_params): Remove call to delete operator.
40192         (ipa_node_params_t::insert): New.
40193         (ipa_node_params_t::remove): Likewise.
40194         (ipa_node_params_t::duplicate): Adjust to descriptors being a pointer,
40195         copy known_csts and known_contexts vectors.
40196         (ipa_read_node_info): Adjust to descriptors being a pointer.
40197         (ipcp_modif_dom_walker): Make m_descriptors field garbage
40198         collected.
40199         (ipcp_transform_function): Make descriptors variable garbage
40200         collected.
40202 2017-01-23  Andrew Senkevich  <andrew.senkevich@intel.com>
40204         * config/i386/avx512bwintrin.h: Add k-mask test, kortest intrinsics.
40205         * config/i386/avx512dqintrin.h: Ditto.
40206         * config/i386/avx512fintrin.h: Ditto.
40207         * config/i386/i386.c: Handle new builtins.
40208         * config/i386/i386-builtin.def: Add new builtins.
40209         * config/i386/sse.md (ktest<mode>, kortest<mode>): New.
40210         (UNSPEC_KORTEST, UNSPEC_KTEST): New.
40212 2017-01-23  Jakub Jelinek  <jakub@redhat.com>
40213             Martin Liska  <mliska@suse.cz>
40215         * asan.h: Define ASAN_USE_AFTER_SCOPE_ATTRIBUTE.
40216         * asan.c (asan_expand_poison_ifn): Support stores and use
40217         appropriate ASAN report function.
40218         * internal-fn.c (expand_ASAN_POISON_USE): New function.
40219         * internal-fn.def (ASAN_POISON_USE): Declare.
40220         * tree-into-ssa.c (maybe_add_asan_poison_write): New function.
40221         (maybe_register_def): Create ASAN_POISON_USE when sanitizing.
40222         * tree-ssa-dce.c (eliminate_unnecessary_stmts): Remove
40223         ASAN_POISON calls w/o LHS.
40224         * tree-ssa.c (execute_update_addresses_taken): Create clobber
40225         for ASAN_MARK (UNPOISON, &x, ...) in order to prevent usage of a LHS
40226         from ASAN_MARK (POISON, &x, ...) coming to a PHI node.
40227         * gimplify.c (asan_poison_variables): Add attribute
40228         use_after_scope_memory to variables that really needs to live
40229         in memory.
40230         * tree-ssa.c (is_asan_mark_p): Do not rewrite into SSA when
40231         having the attribute.
40233 2017-01-23  Martin Liska  <mliska@suse.cz>
40235         * asan.c (create_asan_shadow_var): New function.
40236         (asan_expand_poison_ifn): Likewise.
40237         * asan.h (asan_expand_poison_ifn): New declaration.
40238         * internal-fn.c (expand_ASAN_POISON): Likewise.
40239         * internal-fn.def (ASAN_POISON): New builtin.
40240         * sanopt.c (pass_sanopt::execute): Expand
40241         asan_expand_poison_ifn.
40242         * tree-inline.c (copy_decl_for_dup_finish): Make function
40243         external.
40244         * tree-inline.h (copy_decl_for_dup_finish): Likewise.
40245         * tree-ssa.c (is_asan_mark_p): New function.
40246         (execute_update_addresses_taken): Rewrite local variables
40247         (identified just by use-after-scope as addressable) into SSA.
40249 2017-01-22  Gerald Pfeifer  <gerald@pfeifer.com>
40251         * doc/install.texi (Specific): opensource.apple.com uses https
40252         now. Remove trailing slash.
40254 2017-01-22  Gerald Pfeifer  <gerald@pfeifer.com>
40256         * README.Portability: Remove note on an Irix compatibility issue.
40258 2017-01-22  Dimitry Andric <dim@FreeBSD.org>
40260         * gcov.c (INCLUDE_ALGORITHM): Define.
40261         (INCLUDE_VECTOR): Define.
40262         No longer include <vector> and <algorithm> directly.
40264 2017-01-21  Gerald Pfeifer  <gerald@pfeifer.com>
40266         * doc/extend.texi (Thread-Local): Change www.akkadia.org reference
40267         to https.
40268         * doc/invoke.texi (Code Gen Options): Ditto.
40270 2017-01-21  Jan Hubicka  <hubicka@ucw.cz>
40272         PR lto/78407
40273         * cfg.c (update_bb_profile_for_threading): Fix updating of probablity.
40275 2017-01-21  Bernd Schmidt  <bschmidt@redhat.com>
40277         rtl-optimization/79125
40278         * cprop.c (local_cprop_pass): Handle cases where we make an
40279         unconditional trap.
40281 2017-01-20  Segher Boessenkool  <segher@kernel.crashing.org>
40283         PR target/61729
40284         PR target/77850
40285         * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Adjust address to
40286         read from, for big endian.
40288 2017-01-20  Jiong Wang  <jiong.wang@arm.com>
40290         * config/aarch64/aarch64-builtins.c (aarch64_init_builtins): Register
40291         register pauth builtins for LP64 only.
40293 2017-01-20  Marek Polacek  <polacek@redhat.com>
40295         PR c/79152
40296         * gimplify.c (should_warn_for_implicit_fallthrough): Handle consecutive
40297         non-case labels.
40299 2017-01-20  Alexander Monakov  <amonakov@ispras.ru>
40301         * omp-expand.c (expand_omp_simd): Clear PROP_gimple_lomp_dev regardless
40302         of safelen status.
40303         * omp-offload.c (pass_omp_device_lower::gate): Use PROP_gimple_lomp_dev.
40304         * passes.c (dump_properties): Handle PROP_gimple_lomp_dev.
40305         * tree-inline.c (expand_call_inline): Propagate PROP_gimple_lomp_dev.
40307 2017-01-20  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
40309         PR target/71270
40310         * config/arm/arm.c (neon_valid_immediate): Reject vector constants
40311         in big-endian mode when they are not a single duplicated value.
40313 2017-01-20  Richard Biener  <rguenther@suse.de>
40315         * BASE-VER: Bump to 7.0.1.
40317 2017-01-20  Alexander Monakov  <amonakov@ispras.ru>
40319         * omp-low.c (omplow_simd_context): New struct.  Use it...
40320         (lower_rec_simd_input_clauses): ...here and...
40321         (lower_rec_input_clauses): ...here to hold common data.  Adjust all
40322         references to idx, lane, max_vf, is_simt.
40324 2017-01-20  Graham Markall  <graham.markall@embecosm.com>
40326         * config/arc/arc.h (LINK_SPEC): Use arclinux_nps emulation when
40327         mcpu=nps400.
40329 2017-01-20  Martin Jambor  <mjambor@suse.cz>
40331         * hsa.h: Renaed to hsa-common.h.  Adjusted a comment.
40332         * hsa.c: Renaed to hsa-common.c.  Change include of gt-hsa.h to
40333         gt-hsa-common.h.
40334         * Makefile.in (OBJS): Rename hsa.o to hsa-common.o.
40335         (GTFILES): Rename hsa.c to hsa-common.c.
40336         * hsa-brig.c: Change include of hsa.h to hsa-common.h.
40337         * hsa-dump.c: Likewise.
40338         * hsa-gen.c: Likewise.
40339         * hsa-regalloc.c: Likewise.
40340         * ipa-hsa.c: Likewise.
40341         * omp-expand.c: Likewise.
40342         * omp-low.c: Likewise.
40343         * toplev.c: Likewise.
40345 2017-01-20  Marek Polacek  <polacek@redhat.com>
40347         PR c/64279
40348         * doc/invoke.texi: Document -Wduplicated-branches.
40349         * fold-const.c (operand_equal_p): Handle MODIFY_EXPR, INIT_EXPR,
40350         COMPOUND_EXPR, PREDECREMENT_EXPR, PREINCREMENT_EXPR,
40351         POSTDECREMENT_EXPR, POSTINCREMENT_EXPR, CLEANUP_POINT_EXPR, EXPR_STMT,
40352         STATEMENT_LIST, and RETURN_EXPR.  For non-pure non-const functions
40353         return 0 only when not OEP_LEXICOGRAPHIC.
40354         (fold_build_cleanup_point_expr): Use the expression
40355         location when building CLEANUP_POINT_EXPR.
40356         * tree-core.h (enum operand_equal_flag): Add OEP_LEXICOGRAPHIC.
40357         * tree.c (add_expr): Handle error_mark_node.
40359 2017-01-20  Martin Liska  <mliska@suse.cz>
40361         PR lto/69188
40362         * tree-profile.c (init_ic_make_global_vars): Do not call
40363         finalize_decl.
40364         (gimple_init_gcov_profiler): Likewise.
40366 2017-01-20  Martin Liska  <mliska@suse.cz>
40368         PR ipa/71190
40369         * cgraph.h (maybe_create_reference): Remove argument and
40370         update comment.
40371         * cgraphclones.c (cgraph_node::create_virtual_clone): Remove one
40372         argument.
40373         * ipa-cp.c (create_specialized_node): Likewise.
40374         * symtab.c (symtab_node::maybe_create_reference): Handle
40375         VAR_DECLs and ADDR_EXPRs and select ipa_ref_use type.
40377 2017-01-20  Martin Liska  <mliska@suse.cz>
40379         * read-rtl-function.c (function_reader::create_function): Use
40380         build_decl instread of build_decl_stat.
40382 2017-01-20  Andrew Senkevich  <andrew.senkevich@intel.com>
40384         * config/i386/avx512bwintrin.h: Add k-mask registers shift intrinsics.
40385         * config/i386/avx512dqintrin.h: Ditto.
40386         * config/i386/avx512fintrin.h: Ditto.
40387         * config/i386/i386-builtin-types.def: Add new types.
40388         * config/i386/i386.c: Handle new types.
40389         * config/i386/i386-builtin.def (__builtin_ia32_kshiftliqi)
40390         (__builtin_ia32_kshiftlihi, __builtin_ia32_kshiftlisi)
40391         (__builtin_ia32_kshiftlidi, __builtin_ia32_kshiftriqi)
40392         (__builtin_ia32_kshiftrihi, __builtin_ia32_kshiftrisi)
40393         (__builtin_ia32_kshiftridi): New.
40394         * config/i386/sse.md (k<code><mode>): Rename *k<code><mode>.
40396 2017-01-19  Segher Boessenkool  <segher@kernel.crashing.org>
40398         PR target/78875
40399         PR target/79140
40400         * config/rs6000/rs6000.c (TARGET_STACK_PROTECT_GUARD): Unconditionally
40401         define to rs6000_init_stack_protect_guard.
40402         (rs6000_init_stack_protect_guard): New function.
40404 2017-01-19  Matthew Fortune  <matthew.fortune@imgtec.com>
40405             Yunqiang Su  <yunqiang.su@imgtec.com>
40407         * config.gcc (supported_defaults): Add madd4.
40408         (with_madd4): Add validation.
40409         (all_defaults): Add madd4.
40410         * config/mips/mips.opt (mmadd4): New option.
40411         * config/mips/mips.h (OPTION_DEFAULT_SPECS): Add a default for
40412         mmadd4.
40413         (TARGET_CPU_CPP_BUILTINS): Add builtin_define for
40414         __mips_no_madd4.
40415         (ISA_HAS_UNFUSED_MADD4): Gate with mips_madd4.
40416         (ISA_HAS_FUSED_MADD4): Likewise.
40417         * doc/invoke.texi (-mmadd4): Document the new option.
40418         * doc/install.texi (--with-madd4): Document the new option.
40420 2017-01-19  Jiong Wang  <jiong.wang@arm.com>
40422         * config/aarch64/aarch64-builtins.c (enum aarch64_builtins): New
40423         entries for AARCH64_PAUTH_BUILTIN_XPACLRI,
40424         AARCH64_PAUTH_BUILTIN_PACIA1716, AARCH64_PAUTH_BUILTIN_AUTIA1716.
40425         (aarch64_init_pauth_hint_builtins): New.
40426         (aarch64_init_builtins): Call aarch64_init_pauth_hint_builtins.
40427         (aarch64_expand_builtin): Expand new builtins.
40429 2017-01-19  Jiong Wang  <jiong.wang@arm.com>
40431         * reg-notes.def (CFA_TOGGLE_RA_MANGLE): New reg-note.
40432         * combine-stack-adj.c (no_unhandled_cfa): Handle
40433         REG_CFA_TOGGLE_RA_MANGLE.
40434         * dwarf2cfi.c (dwarf2out_frame_debug): Handle REG_CFA_TOGGLE_RA_MANGLE.
40435         * config/aarch64/aarch64.c (aarch64_expand_prologue): Generates DWARF
40436         info for return address signing.
40437         (aarch64_expand_epilogue): Likewise.
40439 2017-01-19  Jiong Wang  <jiong.wang@arm.com>
40441         * config/aarch64/aarch64-opts.h (aarch64_function_type): New enum.
40442         * config/aarch64/aarch64-protos.h
40443         (aarch64_return_address_signing_enabled): New declaration.
40444         * config/aarch64/aarch64.c (aarch64_return_address_signing_enabled):
40445         New function.
40446         (aarch64_expand_prologue): Sign return address before it's pushed onto
40447         stack.
40448         (aarch64_expand_epilogue): Authenticate return address fetched from
40449         stack.
40450         (aarch64_override_options): Sanity check for ILP32 and ISA level.
40451         (aarch64_attributes): New function attributes for "sign-return-address".
40452         * config/aarch64/aarch64.md (UNSPEC_AUTI1716, UNSPEC_AUTISP,
40453         UNSPEC_PACI1716, UNSPEC_PACISP, UNSPEC_XPACLRI): New unspecs.
40454         ("*do_return"): Generate combined instructions according to key index.
40455         ("<pauth_mnem_prefix>sp", "<pauth_mnem_prefix1716", "xpaclri"): New.
40456         * config/aarch64/iterators.md (PAUTH_LR_SP, PAUTH_17_16): New integer
40457         iterators.
40458         (pauth_mnem_prefix, pauth_hint_num_a): New integer attributes.
40459         * config/aarch64/aarch64.opt (msign-return-address=): New.
40460         * doc/extend.texi (AArch64 Function Attributes): Documents
40461         "sign-return-address=".
40462         * doc/invoke.texi (AArch64 Options): Documents "-msign-return-address=".
40464 2017-01-19  Matthew Fortune  <matthew.fortune@imgtec.com>
40466         * doc/invoke.texi: Add missing -mlxc1-sxc1 options to
40467         overall option summary.
40469 2017-01-19  Jiong Wang  <jiong.wang@arm.com>
40471         * config/aarch64/aarch64-arches.def: New entry for "armv8.3-a".
40472         * config/aarch64/aarch64.h (AARCH64_FL_V8_3, AARCH64_FL_FOR_ARCH8_3,
40473         AARCH64_ISA_V8_3, TARGET_ARMV8_3): New.
40474         * doc/invoke.texi (AArch64 Options): Document "armv8.3-a".
40476 2017-01-19  Michael Meissner  <meissner@linux.vnet.ibm.com>
40478         * config/rs6000/rs6000-cpus.def (ISA_3_0_MASKS_SERVER): Enable
40479         -mpower9-minmax by default for -mcpu=power9.
40480         (ISA_3_MASKS_IEEE): Require -mvsx-small-integer to enable IEEE
40481         128-bit floating point.
40483 2017-01-20  Alan Modra  <amodra@gmail.com>
40485         * config/rs6000/rs6000.md (cmpstrnsi, cmpstrsi): Fail if
40486         optimizing for size.
40488 2017-01-20  Alan Modra  <amodra@gmail.com>
40490         PR target/79144
40491         * config/rs6000/rs6000.c (expand_strn_compare): Get the asm name
40492         for strcmp and strncmp from corresponding builtin decl.
40494 2017-01-19  Uros Bizjak  <ubizjak@gmail.com>
40496         * config.gcc (x86_64-*-rtems*): Use i386/rtemself.h
40497         instead of i386/rtems-64.h.
40498         * config/i386/rtems-64.h: Remove.
40500 2017-01-19  Uros Bizjak  <ubizjak@gmail.com>
40502         PR target/78478
40503         Revert:
40504         2013-11-05  Uros Bizjak  <ubizjak@gmail.com>
40506         * config/i386/rtemself.h (LONG_DOUBLE_TYPE_SIZE): New define.
40508 2017-01-19  Tamar Christina  <tamar.christina@arm.com>
40510         * config/aarch64/aarch64.c (aarch64_simd_gen_const_vector_dup):
40511         Change int to HOST_WIDE_INT.
40512         * config/aarch64/aarch64-protos.h
40513         (aarch64_simd_gen_const_vector_dup): Likewise.
40514         * config/aarch64/aarch64-simd.md: Add copysign<mode>3.
40516 2017-01-19  David Malcolm  <dmalcolm@redhat.com>
40518         * langhooks-def.h (lhd_type_for_size): New decl.
40519         (LANG_HOOKS_TYPE_FOR_SIZE): Define as lhd_type_for_size.
40520         * langhooks.c (lhd_type_for_size): New function, taken from
40521         lto_type_for_size.
40523 2017-01-19  Pat Haugen  <pthaugen@us.ibm.com>
40525         * config/rs6000/power9.md (power9-alu): Remove 'cmp' type and add
40526         define_bypass for CR latency.
40527         (power9-cracked-alu): Update bypass latency and remove power9-branch.
40528         (power9-alu2): Add define_bypass for CR latency.
40529         (power9-cmp): New.
40530         (power9-mul): Update insn latency.
40531         (power9-mul-compare): Update insn latency, bypass latency and remove
40532         power9-branch.
40534 2016-01-19  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
40536         * config/aarch64/aarch64-protos.h (aarch64_nopcrelative_literal_loads):
40537         Delete.
40538         * config/aarch64/aarch64.md
40539         (aarch64_reload_movcp<GPF_TF:mode><P:mode>): Delete reference to
40540         aarch64_nopcrelative_literal_loads.
40541         (aarch64_reload_movcp<VALL:mode><P:mode>): Likewise.
40543 2017-01-19  Chenghua Xu  <paul.hua.gm@gmail.com>
40545         * config/mips/mips.h (ISA_HAS_FUSED_MADD4): Enable for
40546         TARGET_LOONGSON_3A.
40547         (ISA_HAS_UNFUSED_MADD4): Exclude TARGET_LOONGSON_3A.
40549 2017-01-19  Doug Gilmore  <doug.gilmore@imgtec.com>
40551         PR target/78176
40552         * config.gcc (supported_defaults): Add lxc1-sxc1.
40553         (with_lxc1_sxc1): Add validation.
40554         (all_defaults): Add lxc1-sxc1.
40555         * config/mips/mips.opt (mlxc1-sxc1): New option.
40556         * config/mips/mips.h (OPTION_DEFAULT_SPECS): Add a default for
40557         mlxc1-sxc1.
40558         (TARGET_CPU_CPP_BUILTINS): Add builtin_define for
40559         __mips_no_lxc1_sxc1.
40560         (ISA_HAS_LXC1_SXC1): Gate with mips_lxc1_sxc1.
40561         * doc/invoke.texi (-mlxc1-sxc1): Document the new option.
40562         * doc/install.texi (--with-lxc1-sxc1): Document the new option.
40564 2017-01-19  Richard Biener  <rguenther@suse.de>
40566         PR tree-optimization/72488
40567         * tree-ssa-sccvn.c (run_scc_vn): When we abort the VN make
40568         sure to restore SSA info.
40569         * tree-ssa.c (verify_ssa): Verify SSA info is not shared.
40571 2017-01-19  Richard Earnshaw  <rearnsha@arm.com>
40573         PR rtl-optimization/79121
40574         * expr.c (expand_expr_real_2, case LSHIFT_EXPR): Look at the signedness
40575         of the inner type when shifting an extended value.
40577 2017-01-17  Jan Hubicka  <hubicka@ucw.cz>
40579         PR lto/78407
40580         * symtab.c (symtab_node::equal_address_to): Fix comparing of
40581         interposable aliases.
40583 2017-01-18  Peter Bergner  <bergner@vnet.ibm.com>
40585         PR target/78516
40586         * config/rs6000/spe.md (mov_si<mode>_e500_subreg0): Fix constraints.
40587         Use the evmergelohi instruction.
40588         (mov_si<mode>_e500_subreg4_2_le): Likewise.
40589         (mov_sitf_e500_subreg8_2_be): Likewise.
40590         (mov_sitf_e500_subreg12_2_le): Likewise.
40591         (mov_si<mode>_e500_subreg0_2_le): Fix constraints.
40592         (mov_si<mode>_e500_subreg4_2_be): Likewise.
40593         (mov_sitf_e500_subreg8_2_le): Likewise.
40594         (mov_sitf_e500_subreg12_2_be): Likewise.
40596 2017-01-18  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
40598         * config/rs6000/altivec.md (altivec_vbpermq): Change "type"
40599         attribute from vecsimple to vecperm.
40600         (altivec_vbpermq2): Likewise.
40602 2017-01-18  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
40604         PR target/79040
40605         * config/rs6000/altivec.h: Fix typo of vec_cntlz to vec_cnttz.
40607 2017-01-18  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
40608         * config/rs6000/rs6000-protos.h (expand_strn_compare): Add arg.
40609         * config/rs6000/rs6000.c (expand_strn_compare): Add ability to expand
40610         strcmp. Fix bug where comparison didn't stop with zero byte. Fix
40611         case where N arg is SIZE_MAX.
40612         * config/rs6000/rs6000.md (cmpstrnsi): Args to expand_strn_compare.
40613         (cmpstrsi): Add pattern.
40615 2017-01-18  Michael Meissner  <meissner@linux.vnet.ibm.com>
40617         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
40618         __builtin_vec_revb builtins.
40619         * config/rs6000/rs6000-builtins.def (P9V_BUILTIN_XXBRQ_V16QI): Add
40620         built-in functions to support generation of the ISA 3.0 XXBR<x>
40621         vector byte reverse instructions.
40622         (P9V_BUILTIN_XXBRQ_V1TI): Likewise.
40623         (P9V_BUILTIN_XXBRD_V2DI): Likewise.
40624         (P9V_BUILTIN_XXBRD_V2DF): Likewise.
40625         (P9V_BUILTIN_XXBGW_V4SI): Likewise.
40626         (P9V_BUILTIN_XXBGW_V4SF): Likewise.
40627         (P9V_BUILTIN_XXBGH_V8HI): Likewise.
40628         (P9V_BUILTIN_VEC_REVB): Likewise.
40629         * config/rs6000/vsx.md (p9_xxbrq_v1ti): New insns/expanders to
40630         generate the ISA 3.0 XXBR<x> vector byte reverse instructions.
40631         (p9_xxbrq_v16qi): Likewise.
40632         (p9_xxbrd_<mode>, VSX_D iterator): Likewise.
40633         (p9_xxbrw_<mode>, VSX_W iterator): Likewise.
40634         (p9_xxbrh_v8hi): Likewise.
40635         * config/rs6000/altivec.h (vec_revb): Define if ISA 3.0.
40636         * doc/extend.texi (RS/6000 Altivec Built-ins): Document the
40637         vec_revb built-in functions.
40639 2017-01-18  Uros Bizjak  <ubizjak@gmail.com>
40641         PR rtl-optimization/78952
40642         * config/i386/i386.md (any_extract): New code iterator.
40643         (*insvqi_2): Use any_extract for source operand.
40644         (*insvqi_3): Use any_shiftrt for source operand.
40646 2017-01-18  Wilco Dijkstra  <wdijkstr@arm.com>
40648         * config/aarch64/aarch64.c (aarch64_sched_adjust_priority)
40649         New function.
40650         (TARGET_SCHED_ADJUST_PRIORITY): Define target hook.
40652 2017-01-18  Matthias Klose  <doko@ubuntu.com>
40654         * doc/install.texi: Allow default for --with-target-bdw-gc-include.
40656 2016-01-18  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
40658         * config/rs6000/altivec.h (vec_bperm): Change #define.
40659         * config/rs6000/altivec.md (UNSPEC_VBPERMD): New enum constant.
40660         (altivec_vbpermq2): New define_insn.
40661         (altivec_vbpermd): Likewise.
40662         * config/rs6000/rs6000-builtin.def (VBPERMQ2): New monomorphic
40663         function interface.
40664         (VBPERMD): Likewise.
40665         (VBPERM): New polymorphic function interface.
40666         * config/rs6000/r6000-c.c (altivec_overloaded_builtins_table):
40667         Add entries for P9V_BUILTIN_VEC_VBPERM.
40668         * doc/extend.texi: Add interfaces for vec_bperm.
40670 2017-01-18  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
40672         * config/s390/s390-c.c (s390_expand_overloaded_builtin): Downcase
40673         first letter of error messages.
40674         (s390_resolve_overloaded_builtin): Likewise.
40675         * config/s390/s390.c (s390_expand_builtin): Likewise.
40676         (s390_invalid_arg_for_unprototyped_fn): Likewise.
40677         (s390_valid_target_attribute_inner_p): Likewise.
40678         * config/s390/s390.md ("tabort"): Likewise.
40680 2017-01-18  Toma Tabacu  <toma.tabacu@imgtec.com>
40682         * config/mips/mips.h (ISA_HAS_DIV3): Remove unused macro.
40683         (ISA_AVOID_DIV_HILO): New macro.
40684         (ISA_HAS_DIV): Use new ISA_AVOID_DIV_HILO macro.
40685         (ISA_HAS_DDIV): Likewise.
40687 2017-01-18  Markus Trippelsdorf  <markus@trippelsdorf.de>
40689         * doc/invoke.texi (fabi-version): Correct number of occurrences.
40691 2017-01-18  Markus Trippelsdorf  <markus@trippelsdorf.de>
40693         * doc/invoke.texi (fabi-version): Spelling fix.
40695 2017-01-18  Markus Trippelsdorf  <markus@trippelsdorf.de>
40697         PR c++/70182
40698         * doc/invoke.texi (fabi-version): Mention mangling fix for
40699         operator names.
40701 2017-01-18  Markus Trippelsdorf  <markus@trippelsdorf.de>
40703         PR c++/77489
40704         * doc/invoke.texi (fabi-version): Document discriminator mangling.
40706 2017-01-17  Segher Boessenkool  <segher@kernel.crashing.org>
40708         PR target/78875
40709         * config/rs6000/rs6000-opts.h (stack_protector_guard): New enum.
40710         * config/rs6000/rs6000.c (rs6000_option_override_internal): Handle
40711         the new options.
40712         * config/rs6000/rs6000.md (stack_protect_set): Handle the new more
40713         flexible settings.
40714         (stack_protect_test): Ditto.
40715         * config/rs6000/rs6000.opt (mstack-protector-guard=,
40716         mstack-protector-guard-reg=, mstack-protector-guard-offset=): New
40717         options.
40718         * doc/invoke.texi (Option Summary) [RS/6000 and PowerPC Options]:
40719         Add -mstack-protector-guard=, -mstack-protector-guard-reg=, and
40720         -mstack-protector-guard-offset=.
40721         (RS/6000 and PowerPC Options): Ditto.
40723 2017-01-17  Uros Bizjak  <ubizjak@gmail.com>
40725         * config/i386/i386.h (MASK_CLASS_P): New define.
40726         * config/i386/i386.c (inline_secondary_memory_needed): Ensure that
40727         there are no registers from different register sets also when
40728         mask registers are used.  Update function comment.
40729         * config/i386/i386.md (*movsi_internal): Split (*k/*krm) alternative
40730         to (*k/*r) and (*k/*km) alternatives.
40732 2017-01-17  Wilco Dijkstra  <wdijkstr@arm.com>
40734         * config/aarch64/aarch64.md (eh_return): Remove pattern and splitter.
40735         * config/aarch64/aarch64.h (AARCH64_EH_STACKADJ_REGNUM): Remove.
40736         (EH_RETURN_HANDLER_RTX): New define.
40737         * config/aarch64/aarch64.c (aarch64_frame_pointer_required):
40738         Force frame pointer in EH return functions.
40739         (aarch64_expand_epilogue): Add barrier for eh_return.
40740         (aarch64_final_eh_return_addr): Remove.
40741         (aarch64_eh_return_handler_rtx): New function.
40742         * config/aarch64/aarch64-protos.h (aarch64_final_eh_return_addr):
40743         Remove.
40744         (aarch64_eh_return_handler_rtx): New prototype.
40746 2017-01-17  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
40748         * config/rs6000/altivec.h (vec_rlmi): New #define.
40749         (vec_vrlnm): Likewise.
40750         (vec_rlnm): Likewise.
40751         * config/rs6000/altivec.md (UNSPEC_VRLMI): New UNSPEC enum value.
40752         (UNSPEC_VRLNM): Likewise.
40753         (VIlong): New mode iterator.
40754         (altivec_vrl<VI_char>mi): New define_insn.
40755         (altivec_vrl<VI_char>nm): Likewise.
40756         * config/rs6000/rs6000-builtin.def (VRLWNM): New monomorphic
40757         function entry.
40758         (VRLDNM): Likewise.
40759         (RLNM): New polymorphic function entry.
40760         (VRLWMI): New monomorphic function entry.
40761         (VRLDMI): Likewise.
40762         (RLMI): New polymorphic function entry.
40763         * config/rs6000/r6000-c.c (altivec_overloaded_builtin_table): Add
40764         new entries for P9V_BUILTIN_VEC_RLMI and P9V_BUILTIN_VEC_RLNM.
40765         * doc/extend.texi: Add description of vec_rlmi, vec_rlnm, and
40766         vec_vrlnm.
40768 2017-01-17  Jakub Jelinek  <jakub@redhat.com>
40770         PR debug/78839
40771         * dwarf2out.c (field_byte_offset): Restore the
40772         PCC_BITFIELD_TYPE_MATTERS behavior for INTEGER_CST DECL_FIELD_OFFSET
40773         and DECL_FIELD_BIT_OFFSET.  Use fold_build2 instead of build2 + fold.
40774         (analyze_variants_discr, gen_variant_part): Use fold_build2 instead
40775         of build2 + fold.
40777 2017-01-17  Eric Botcazou  <ebotcazou@adacore.com>
40779         PR ada/67205
40780         * config/aarch64/aarch64.c (TARGET_CUSTOM_FUNCTION_DESCRIPTORS): Define
40782 2017-01-17  Jakub Jelinek  <jakub@redhat.com>
40784         PR debug/71669
40785         * dwarf2out.c (add_data_member_location_attribute): For constant
40786         offset bitfield emit for -gdwarf-5 DW_AT_data_bit_offset attribute
40787         instead of DW_AT_data_member_location, DW_AT_bit_offset and
40788         DW_AT_byte_size attributes.
40790 2017-01-17  Eric Botcazou  <ebotcazou@adacore.com>
40792         * config/rs6000/rs6000.c (rs6000_emit_move): Also use a TOC reference
40793         after forcing to constant memory when the code model is medium.
40795 2017-01-17  Julia Koval  <julia.koval@intel.com>
40797         PR target/76731
40798         * config/i386/avx512fintrin.h
40799         (_mm512_i32gather_ps): Change __addr type to void const*.
40800         (_mm512_mask_i32gather_ps): Ditto.
40801         (_mm512_i32gather_pd): Ditto.
40802         (_mm512_mask_i32gather_pd): Ditto.
40803         (_mm512_i64gather_ps): Ditto.
40804         (_mm512_mask_i64gather_ps): Ditto.
40805         (_mm512_i64gather_pd): Ditto.
40806         (_mm512_mask_i64gather_pd): Ditto.
40807         (_mm512_i32gather_epi32): Ditto.
40808         (_mm512_mask_i32gather_epi32): Ditto.
40809         (_mm512_i32gather_epi64): Ditto.
40810         (_mm512_mask_i32gather_epi64): Ditto.
40811         (_mm512_i64gather_epi32): Ditto.
40812         (_mm512_mask_i64gather_epi32): Ditto.
40813         (_mm512_i64gather_epi64): Ditto.
40814         (_mm512_mask_i64gather_epi64): Ditto.
40815         (_mm512_i32scatter_ps): Change __addr type to void*.
40816         (_mm512_mask_i32scatter_ps): Ditto.
40817         (_mm512_i32scatter_pd): Ditto.
40818         (_mm512_mask_i32scatter_pd): Ditto.
40819         (_mm512_i64scatter_ps): Ditto.
40820         (_mm512_mask_i64scatter_ps): Ditto.
40821         (_mm512_i64scatter_pd): Ditto.
40822         (_mm512_mask_i64scatter_pd): Ditto.
40823         (_mm512_i32scatter_epi32): Ditto.
40824         (_mm512_mask_i32scatter_epi32): Ditto.
40825         (_mm512_i32scatter_epi64): Ditto.
40826         (_mm512_mask_i32scatter_epi64): Ditto.
40827         (_mm512_i64scatter_epi32): Ditto.
40828         (_mm512_mask_i64scatter_epi32): Ditto.
40829         (_mm512_i64scatter_epi64): Ditto.
40830         (_mm512_mask_i64scatter_epi64): Ditto.
40831         * config/i386/avx512pfintrin.h
40832         (_mm512_mask_prefetch_i32gather_pd): Change __addr type to void const*.
40833         (_mm512_mask_prefetch_i32gather_ps): Ditto.
40834         (_mm512_mask_prefetch_i64gather_pd): Ditto.
40835         (_mm512_mask_prefetch_i64gather_ps): Ditto.
40836         (_mm512_prefetch_i32scatter_pd): Change __addr type to void*.
40837         (_mm512_prefetch_i32scatter_ps): Ditto.
40838         (_mm512_mask_prefetch_i32scatter_pd): Ditto.
40839         (_mm512_mask_prefetch_i32scatter_ps): Ditto.
40840         (_mm512_prefetch_i64scatter_pd): Ditto.
40841         (_mm512_prefetch_i64scatter_ps): Ditto.
40842         (_mm512_mask_prefetch_i64scatter_pd): Ditto.
40843         (_mm512_mask_prefetch_i64scatter_ps): Ditto.
40844         * config/i386/avx512vlintrin.h
40845         (_mm256_mmask_i32gather_ps): Change __addr type to void const*.
40846         (_mm_mmask_i32gather_ps): Ditto.
40847         (_mm256_mmask_i32gather_pd): Ditto.
40848         (_mm_mmask_i32gather_pd): Ditto.
40849         (_mm256_mmask_i64gather_ps): Ditto.
40850         (_mm_mmask_i64gather_ps): Ditto.
40851         (_mm256_mmask_i64gather_pd): Ditto.
40852         (_mm_mmask_i64gather_pd): Ditto.
40853         (_mm256_mmask_i32gather_epi32): Ditto.
40854         (_mm_mmask_i32gather_epi32): Ditto.
40855         (_mm256_mmask_i32gather_epi64): Ditto.
40856         (_mm_mmask_i32gather_epi64): Ditto.
40857         (_mm256_mmask_i64gather_epi32): Ditto.
40858         (_mm_mmask_i64gather_epi32): Ditto.
40859         (_mm256_mmask_i64gather_epi64): Ditto.
40860         (_mm_mmask_i64gather_epi64): Ditto.
40861         (_mm256_i32scatter_ps): Change __addr type to void*.
40862         (_mm256_mask_i32scatter_ps): Ditto.
40863         (_mm_i32scatter_ps): Ditto.
40864         (_mm_mask_i32scatter_ps): Ditto.
40865         (_mm256_i32scatter_pd): Ditto.
40866         (_mm256_mask_i32scatter_pd): Ditto.
40867         (_mm_i32scatter_pd): Ditto.
40868         (_mm_mask_i32scatter_pd): Ditto.
40869         (_mm256_i64scatter_ps): Ditto.
40870         (_mm256_mask_i64scatter_ps): Ditto.
40871         (_mm_i64scatter_ps): Ditto.
40872         (_mm_mask_i64scatter_ps): Ditto.
40873         (_mm256_i64scatter_pd): Ditto.
40874         (_mm256_mask_i64scatter_pd): Ditto.
40875         (_mm_i64scatter_pd): Ditto.
40876         (_mm_mask_i64scatter_pd): Ditto.
40877         (_mm256_i32scatter_epi32): Ditto.
40878         (_mm256_mask_i32scatter_epi32): Ditto.
40879         (_mm_i32scatter_epi32): Ditto.
40880         (_mm_mask_i32scatter_epi32): Ditto.
40881         (_mm256_i32scatter_epi64): Ditto.
40882         (_mm256_mask_i32scatter_epi64): Ditto.
40883         (_mm_i32scatter_epi64): Ditto.
40884         (_mm_mask_i32scatter_epi64): Ditto.
40885         (_mm256_i64scatter_epi32): Ditto.
40886         (_mm256_mask_i64scatter_epi32): Ditto.
40887         (_mm_i64scatter_epi32): Ditto.
40888         (_mm_mask_i64scatter_epi32): Ditto.
40889         (_mm256_i64scatter_epi64): Ditto.
40890         (_mm256_mask_i64scatter_epi64): Ditto.
40891         (_mm_i64scatter_epi64): Ditto.
40892         (_mm_mask_i64scatter_epi64): Ditto.
40893         * config/i386/i386-builtin-types.def (V16SF_V16SF_PCFLOAT_V16SI_HI_INT)
40894         (V8DF_V8DF_PCDOUBLE_V8SI_QI_INT, V8SF_V8SF_PCFLOAT_V8DI_QI_INT)
40895         (V8DF_V8DF_PCDOUBLE_V8DI_QI_INT, V16SI_V16SI_PCINT_V16SI_HI_INT)
40896         (V8DI_V8DI_PCINT64_V8SI_QI_INT, V8SI_V8SI_PCINT_V8DI_QI_INT)
40897         (V8DI_V8DI_PCINT64_V8DI_QI_INT, V2DF_V2DF_PCDOUBLE_V4SI_QI_INT)
40898         (V4DF_V4DF_PCDOUBLE_V4SI_QI_INT, V2DF_V2DF_PCDOUBLE_V2DI_QI_INT)
40899         (V4DF_V4DF_PCDOUBLE_V4DI_QI_INT, V4SF_V4SF_PCFLOAT_V4SI_QI_INT)
40900         (V8SF_V8SF_PCFLOAT_V8SI_QI_INT, V4SF_V4SF_PCFLOAT_V2DI_QI_INT)
40901         (V4SF_V4SF_PCFLOAT_V4DI_QI_INT, V2DI_V2DI_PCINT64_V4SI_QI_INT)
40902         (V4DI_V4DI_PCINT64_V4SI_QI_INT, V2DI_V2DI_PCINT64_V2DI_QI_INT)
40903         (V4DI_V4DI_PCINT64_V4DI_QI_INT, V4SI_V4SI_PCINT_V4SI_QI_INT)
40904         (V8SI_V8SI_PCINT_V8SI_QI_INT, V4SI_V4SI_PCINT_V2DI_QI_INT)
40905         (V4SI_V4SI_PCINT_V4DI_QI_INT, VOID_PFLOAT_HI_V16SI_V16SF_INT)
40906         (VOID_PFLOAT_QI_V8SI_V8SF_INT, VOID_PFLOAT_QI_V4SI_V4SF_INT)
40907         (VOID_PDOUBLE_QI_V8SI_V8DF_INT, VOID_PDOUBLE_QI_V4SI_V4DF_INT)
40908         (VOID_PDOUBLE_QI_V4SI_V2DF_INT, VOID_PFLOAT_QI_V8DI_V8SF_INT)
40909         (VOID_PFLOAT_QI_V4DI_V4SF_INT, VOID_PFLOAT_QI_V2DI_V4SF_INT)
40910         (VOID_PDOUBLE_QI_V8DI_V8DF_INT, VOID_PDOUBLE_QI_V4DI_V4DF_INT)
40911         (VOID_PDOUBLE_QI_V2DI_V2DF_INT, VOID_PINT_HI_V16SI_V16SI_INT)
40912         (VOID_PINT_QI_V8SI_V8SI_INT, VOID_PINT_QI_V4SI_V4SI_INT)
40913         (VOID_PLONGLONG_QI_V8SI_V8DI_INT, VOID_PLONGLONG_QI_V4SI_V4DI_INT)
40914         (VOID_PLONGLONG_QI_V4SI_V2DI_INT, VOID_PINT_QI_V8DI_V8SI_INT)
40915         (VOID_PINT_QI_V4DI_V4SI_INT, VOID_PINT_QI_V2DI_V4SI_INT)
40916         (VOID_PLONGLONG_QI_V8DI_V8DI_INT, VOID_QI_V8SI_PCINT64_INT_INT)
40917         (VOID_PLONGLONG_QI_V4DI_V4DI_INT, VOID_PLONGLONG_QI_V2DI_V2DI_INT)
40918         (VOID_HI_V16SI_PCINT_INT_INT, VOID_QI_V8DI_PCINT64_INT_INT)
40919         (VOID_QI_V8DI_PCINT_INT_INT): Remove.
40920         (V16SF_V16SF_PCVOID_V16SI_HI_INT,  V8DF_V8DF_PCVOID_V8SI_QI_INT)
40921         (V8SF_V8SF_PCVOID_V8DI_QI_INT, V8DF_V8DF_PCVOID_V8DI_QI_INT)
40922         (V16SI_V16SI_PCVOID_V16SI_HI_INT, V8DI_V8DI_PCVOID_V8SI_QI_INT)
40923         (V8SI_V8SI_PCVOID_V8DI_QI_INT, V8DI_V8DI_PCVOID_V8DI_QI_INT)
40924         (VOID_PVOID_HI_V16SI_V16SF_INT, VOID_PVOID_QI_V8SI_V8DF_INT)
40925         (VOID_PVOID_QI_V8DI_V8SF_INT, VOID_PVOID_QI_V8DI_V8DF_INT)
40926         (VOID_PVOID_HI_V16SI_V16SI_INT, VOID_PVOID_QI_V8SI_V8DI_INT)
40927         (VOID_PVOID_QI_V8DI_V8SI_INT, VOID_PVOID_QI_V8DI_V8DI_INT)
40928         (V2DF_V2DF_PCVOID_V4SI_QI_INT, V4DF_V4DF_PCVOID_V4SI_QI_INT)
40929         (V2DF_V2DF_PCVOID_V2DI_QI_INT, V4DF_V4DF_PCVOID_V4DI_QI_INT
40930         (V4SF_V4SF_PCVOID_V4SI_QI_INT, V8SF_V8SF_PCVOID_V8SI_QI_INT)
40931         (V4SF_V4SF_PCVOID_V2DI_QI_INT, V4SF_V4SF_PCVOID_V4DI_QI_INT)
40932         (V2DI_V2DI_PCVOID_V4SI_QI_INT, V4DI_V4DI_PCVOID_V4SI_QI_INT)
40933         (V2DI_V2DI_PCVOID_V2DI_QI_INT, V4DI_V4DI_PCVOID_V4DI_QI_INT)
40934         (V4SI_V4SI_PCVOID_V4SI_QI_INT, V8SI_V8SI_PCVOID_V8SI_QI_INT)
40935         (V4SI_V4SI_PCVOID_V2DI_QI_INT, V4SI_V4SI_PCVOID_V4DI_QI_INT)
40936         (VOID_PVOID_QI_V8SI_V8SF_INT, VOID_PVOID_QI_V4SI_V4SF_INT)
40937         (VOID_PVOID_QI_V4SI_V4DF_INT, VOID_PVOID_QI_V4SI_V2DF_INT)
40938         (VOID_PVOID_QI_V4DI_V4SF_INT, VOID_PVOID_QI_V2DI_V4SF_INT)
40939         (VOID_PVOID_QI_V4DI_V4DF_INT, VOID_PVOID_QI_V2DI_V2DF_INT)
40940         (VOID_PVOID_QI_V8SI_V8SI_INT, VOID_PVOID_QI_V4SI_V4SI_INT)
40941         (VOID_PVOID_QI_V4SI_V4DI_INT, VOID_PVOID_QI_V4SI_V2DI_INT)
40942         (VOID_PVOID_QI_V4DI_V4SI_INT, VOID_PVOID_QI_V2DI_V4SI_INT)
40943         (VOID_PVOID_QI_V4DI_V4DI_INT, VOID_PVOID_QI_V2DI_V2DI_INT)
40944         (VOID_QI_V8SI_PCVOID_INT_INT, VOID_HI_V16SI_PCVOID_INT_INT)
40945         (VOID_QI_V8DI_PCVOID_INT_INT): Add.
40946         * config/i386/i386.c (ix86_init_mmx_sse_builtins): Adjust builtin
40947         definitions accordingly.
40949 2017-01-17  Kito Cheng  <kito.cheng@gmail.com>
40950             Kuan-Lin Chen  <kuanlinchentw@gmail.com>
40952         PR target/79079
40953         * internal-fn.c (expand_mul_overflow): Use convert_modes instead of
40954         gen_lowpart.
40956 2017-01-17  Vladimir Makarov  <vmakarov@redhat.com>
40958         PR target/79058
40959         * ira-conflicts.c (ira_build_conflicts): Update total conflict
40960         hard regs for inner regno.
40962 2017-01-17  Martin Liska  <mliska@suse.cz>
40964         PR ipa/71207
40965         * ipa-polymorphic-call.c (contains_type_p): Fix wrong
40966         assumption and add comment.
40968 2017-01-17  Nathan Sidwell  <nathan@acm.org>
40970         * ipa-visibility.c (localize_node): New function, broken out of ...
40971         (function_and_variable_visibility): ... here. Call it.
40973 2017-01-17  Jan Hubicka  <hubicka@ucw.cz>
40975         PR middle-end/77445
40976         * tree-ssa-threadupdate.c (remove_ctrl_stmt_and_useless_edges):
40977         correctly set frequency of oudgoing edge.
40978         (duplicate_thread_path): Fix profile updating.
40980 2017-01-17  Jakub Jelinek  <jakub@redhat.com>
40982         PR other/79046
40983         * configure.ac: Add GCC_BASE_VER.
40984         * Makefile.in (version): Use @get_gcc_base_ver@ instead of cat to get
40985         version from BASE-VER file.
40986         (CFLAGS-gcc.o): Add -DBASEVER=$(BASEVER_s).
40987         (gcc.o): Depend on $(BASEVER).
40988         * common.opt (dumpfullversion): New option.
40989         * gcc.c (driver_handle_option): Handle OPT_dumpfullversion.
40990         * doc/invoke.texi: Document -dumpfullversion.
40991         * doc/install.texi: Document --with-gcc-major-version-only.
40992         * configure: Regenerated.
40994 2017-01-17  Richard Biener  <rguenther@suse.de>
40996         PR tree-optimization/71433
40997         * tree-vrp.c (register_new_assert_for): Merge same asserts
40998         on all incoming edges.
40999         (process_assert_insertions_for): Handle insertions at the
41000         beginning of BBs.
41002 2017-01-17  Gerald Pfeifer  <gerald@pfeifer.com>
41004         * config/i386/cygwin.h (LIBGCJ_SONAME): Remove.
41005         * config/i386/mingw32.h (LIBGCJ_SONAME): Remove.
41007 2017-01-17  Kaz Kojima  <kkojima@gcc.gnu.org>
41009         PR target/78633
41010         * config/sh/sh.md (cmpeqsi_t+1): Call copy_rtx to avoid invalid
41011         RTL sharing.
41013 2017-01-17  Alan Modra  <amodra@gmail.com>
41015         PR target/79066
41016         * config/rs6000/rs6000.md (elf_high, elf_low): Disable when pic.
41017         * config/rs6000/rs6000.c (rs6000_emit_allocate_stack): Don't allow
41018         symbolic stack limit when pic.
41020 2017-01-16  Martin Sebor  <msebor@redhat.com>
41022         PR tree-optimization/78608
41023         * gimple-ssa-sprintf.c (tree_digits): Avoid negating TYPE_MIN.
41025 2017-01-16  Jeff Law  <law@redhat.com>
41027         Revert:
41028         2016-12-02  Tadek Kijkowski  <tkijkowski@gmail.com>
41029         * Makefile.in (PREPROCESSOR_DEFINES): Add a level of indirection
41030         for several include directories that may be relative to sysroot.
41031         * config/i386/x-mingw32 (gplus_includedir): Define.
41032         (gplus_tool_includedir, gplus_backward_include_dir): Likewise.
41033         (native_system_includedir): Likewise.
41034         * config/i386/mingw32.h (STANDARD_STARTFILE_PREFIX_1): Do not
41035         override if TARGET_SYSTEM_ROOT is defined.
41036         (NATIVE_SYSTEM_HEADER_DIR): Likewise.
41038         PR tree-optimization/79090
41039         PR tree-optimization/33562
41040         PR tree-optimization/61912
41041         PR tree-optimization/77485
41042         * tree-ssa-dse.c (compute_trims): Accept STMT argument.  Dump STMT
41043         and computed trims into the dump file.
41045 2017-01-17  Uros Bizjak  <ubizjak@gmail.com>
41047         * config/i386/i386.h (LIMIT_RELOAD_CLASS): Remove.
41049 2017-01-16  Jakub Jelinek  <jakub@redhat.com>
41051         PR c/79089
41052         * gimplify.c (gimplify_init_constructor): If want_value and
41053         object == lhs, unshare lhs to avoid invalid tree sharing.  Formatting
41054         fix.
41056         PR target/79080
41057         * loop-doloop.c (doloop_modify): Call unshare_all_rtl_in_chain on
41058         sequence.  Formatting fixes.
41059         (doloop_optimize): Formatting fixes.
41061         PR driver/49726
41062         * gcc.c (debug_level_greater_than_spec_func): New function.
41063         (static_spec_functions): Add debug-level-gt spec function.
41064         (ASM_DEBUG_SPEC, cpp_options): Use %:debug-level-gt(0) instead of
41065         !g0.
41066         * config/darwin.h (DSYMUTIL_SPEC, ASM_DEBUG_SPEC): Likewise.
41067         * config/darwin9.h (DSYMUTIL_SPEC, ASM_DEBUG_SPEC): Likewise.
41068         * common.opt (g, gcoff, gdwarf, gdwarf-, ggdb, gno-pubnames,
41069         gpubnames, ggnu-pubnames, gno-record-gcc-switches,
41070         grecord-gcc-switches, gno-strict-dwarf, gstrict-dwarf, gstabs,
41071         gstabs+, gtoggle, gvms, gxcoff, gxcoff+): Add Driver flag.
41073 2017-01-16  Uros Bizjak  <ubizjak@gmail.com>
41075         * config/i386/i386.h (HARD_REGNO_CALLER_SAVE_MODE): Apply HImode and
41076         QImode fixups to general and mask registers only.
41078 2017-01-16  Carl Love  <cel@us.ibm.com>
41080         * config/rs6000/rs6000-c (altivec_overloaded_builtins): Add support
41081         for built-in functions
41082         vector signed char vec_nabs (vector signed char)
41083         vector signed short vec_nabs (vector signed short)
41084         vector signed int vec_nabs (vector signed int)
41085         vector signed long long vec_nabs (vector signed long long)
41086         vector float vec_nabs (vector float)
41087         vector double vec_nabs (vector double)
41088         * config/rs6000/rs6000-builtin.def: Add definitions for NABS functions
41089         and NABS overload.
41090         * config/rs6000/altivec.md: New define_expand nabs<mode>2 types
41091         * config/rs6000/altivec.h: New define for vec_nabs built-in function.
41092         * doc/extend.texi: Update the documentation file for the new built-in
41093         functions.
41095 2017-01-16  Martin Sebor  <msebor@redhat.com>
41097         * gimple-ssa-sprintf.c (format_directive): Correct a typo in a warning
41098         message.
41100 2017-01-16  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
41102         * config/rs6000/rs6000.c (rtx_is_swappable_p): Change
41103         UNSPEC_VSX__XXSPLTD to require special splat handling.
41105 2017-01-16  David Malcolm  <dmalcolm@redhat.com>
41107         PR bootstrap/78616
41108         * system.h: Poison strndup.
41110 2017-01-16  Alan Modra  <amodra@gmail.com>
41112         PR target/79098
41113         * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Don't
41114         use a switch.
41116 2017-01-16  Georg-Johann Lay  <avr@gjlay.de>
41118         * config/avr/avr.h (BRANCH_COST) [reload_completed]: Increase by 4.
41120 2017-01-15  Uros Bizjak  <ubizjak@gmail.com>
41122         * config/i386/i386.c (ix86_legitimate_combined_insn): Do not
41123         call recog here.  Assert that INSN_CODE (insn) is non-negative.
41125 2017-01-15  Segher Boessenkool  <segher@kernel.crashing.org>
41127         PR target/72749
41128         * cfgrtl.c (rtl_split_edge): Also patch jump insns that jump to the
41129         fallthrough.
41130         * haifa-sched.c (dump_insn_stream): Don't crash if there is a label
41131         in the currently scheduled RTL fragment.
41133 2017-01-15  Segher Boessenkool  <segher@kernel.crashing.org>
41135         PR rtl-optimization/78751
41136         * ifcvt.c (find_cond_trap): If we generated a non-existing insn,
41137         give up.
41139 2017-01-14  Jeff Law  <law@redhat.com>
41141         PR tree-optimization/79090
41142         * tree-ssa-dse.c (valid_ao_ref_for_dse): Reject zero length and
41143         variable length stores.
41144         (compute_trims): Delete dead assignment to *trim_tail.
41145         (dse_dom_walker::dse_optimize_stmt): Optimize mem* calls with
41146         zero length.
41148 2017-01-14  Bernd Schmidt  <bschmidt@redhat.com>
41150         PR rtl-optimization/78626
41151         PR rtl-optimization/78727
41152         * cprop.c (one_cprop_pass): Collect unconditional traps in the middle
41153         of a block, and split such blocks after everything else is finished.
41155 2017-01-14  Alan Modra  <amodra@gmail.com>
41157         PR target/72749
41158         * combine.c (recog_for_combine_1): Set INSN_CODE before calling
41159         target legitimate_combined_insn.
41160         * config/rs6000/rs6000.c (TARGET_LEGITIMATE_COMBINED_INSN): Define.
41161         (rs6000_legitimate_combined_insn): New function.
41162         * config/rs6000/rs6000.md (UNSPEC_DOLOOP): Delete, and remove
41163         all uses.
41164         (ctr<mode>_internal3): Rename from *ctr<mode>_internal5.
41165         (ctr<mode>_internal4): Rename from *ctr<mode>_internal6.
41166         (ctr<mode>_internal1, ctr<mode>_internal2): Remove '*' from name.
41168 2017-01-14  Gerald Pfeifer  <gerald@pfeifer.com>
41170         * doc/frontends.texi (G++ and GCC): Remove references to Java.
41172 2017-01-13  Jeff Law  <law@redhat.com>
41174         PR tree-optimization/33562
41175         PR tree-optimization/61912
41176         PR tree-optimization/77485
41177         * tree-ssa-dse.c (delete_dead_call): Accept gsi rather than
41178         a statement.
41179         (delete_dead_assignment): Likewise.
41180         (dse_dom_walker::dse_optimize_stmt): Pass in the gsi rather than
41181         statement to delete_dead_call and delete_dead_assignment.
41183 2017-01-13  David Malcolm  <dmalcolm@redhat.com>
41185         PR c/78304
41186         * substring-locations.c (format_warning_va): Strengthen case 1 so
41187         that both endpoints of the substring must be within the format
41188         range for just the substring to be printed.
41190 2017-01-13  Uros Bizjak  <ubizjak@gmail.com>
41192         * config/i386/i386.opt (msgx): Use ix86_isa_flags2 variable.
41193         * config/i386/i386.c (ix86_target_string): Add missing options
41194         to isa_opts and reorder options by implied ISAs.  Rename isa_opts2 to
41195         isa2_opts, ix86_flag_opts to flag2_opts, ix86_target_other to
41196         flags_other and ix86_target_other to flags2_other.  Display unknown
41197         isa2 options.
41198         (ix86_valid_target_attribute_inner_p): Add missing options and
41199         reorder options by implied ISAs, as in ix86_target_string.
41201 2017-01-13  Richard Sandiford  <richard.sandiford@arm.com>
41203         * hash-table.h (hash_table::too_empty_p): New function.
41204         (hash_table::expand): Use it.
41205         (hash_table::traverse): Likewise.
41206         (hash_table::empty_slot): Use sizeof (value_type) instead of
41207         sizeof (PTR) to convert bytes to elements.  Shrink the table
41208         if the current size is excessive for the current number of
41209         elements.
41211 2017-01-13  Richard Sandiford  <richard.sandiford@arm.com>
41213         * ira-costs.c (record_reg_classes): Break from the inner loop
41214         early once alt_fail is known to be true.  Update outer loop
41215         handling accordingly.
41217 2017-01-13  Jeff Law  <law@redhat.com>
41219         * tree-ssa-dse.c (decrement_count): New function.
41220         (increment_start_addr, maybe_trim_memstar_call): Likewise.
41221         (dse_dom_walker::optimize_stmt): Call maybe_trim_memstar_call directly
41222         when we know the partially dead statement is a mem* function.
41224         PR tree-optimization/61912
41225         PR tree-optimization/77485
41226         * tree-ssa-dse.c: Include expr.h.
41227         (maybe_trim_constructor_store): New function.
41228         (maybe_trim_partially_dead_store): Call maybe_trim_constructor_store.
41230         PR tree-optimization/33562
41231         PR tree-optimization/61912
41232         PR tree-optimization/77485
41233         * doc/invoke.texi: Document new dse-max-object-size param.
41234         * params.def (PARM_DSE_MAX_OBJECT_SIZE): New PARAM.
41235         * tree-ssa-dse.c: Include params.h.
41236         (dse_store_status): New enum.
41237         (initialize_ao_ref_for_dse): New, partially extracted from
41238         dse_optimize_stmt.
41239         (valid_ao_ref_for_dse, normalize_ref): New.
41240         (setup_live_bytes_from_ref, compute_trims): Likewise.
41241         (clear_bytes_written_by, maybe_trim_complex_store): Likewise.
41242         (maybe_trim_partially_dead_store): Likewise.
41243         (maybe_trim_complex_store): Likewise.
41244         (dse_classify_store): Renamed from dse_possibly_dead_store_p.
41245         Track what bytes live from the original store.  Return tri-state
41246         for dead, partially dead or live.
41247         (dse_dom_walker): Add constructor, destructor and new private members.
41248         (delete_dead_call, delete_dead_assignment): New extracted from
41249         dse_optimize_stmt.
41250         (dse_optimize_stmt): Make a member of dse_dom_walker.
41251         Use initialize_ao_ref_for_dse.
41253         PR tree-optimization/33562
41254         PR tree-optimization/61912
41255         PR tree-optimization/77485
41256         * sbitmap.h (bitmap_count_bits): Prototype.
41257         (bitmap_clear_range, bitmap_set_range): Likewise.
41258         * sbitmap.c (bitmap_clear_range): New function.
41259         (bitmap_set_range, sbitmap_popcount, bitmap_count_bits): Likewise.
41261 2017-01-13  Martin Liska  <mliska@suse.cz>
41263         PR ipa/79043
41264         * function.c (set_cfun): Add new argument force.
41265         * function.h (set_cfun): Likewise.
41266         * ipa-inline-transform.c (inline_call): Use the function when
41267         strict alising from is dropped for function we inline to.
41269 2017-01-13  Richard Biener  <rguenther@suse.de>
41271         * tree-pretty-print.c (dump_generic_node): Fix inverted condition
41272         for dumping GIMPLE INTEGER_CSTs.
41274 2017-01-13  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
41276         * config/sol2.h (TARGET_OS_CPP_BUILTINS): Define __STDC_VERSION__
41277         to 201112L since C++17.
41279 2017-01-13  Maxim Ostapenko  <m.ostapenko@samsung.com>
41281         PR sanitizer/78887
41282         * asan.c (asan_needs_odr_indicator_p): Don't emit ODR indicators
41283         if -fsanitize=kernel-address is present.
41285 2017-01-13  Richard Biener  <rguenther@suse.de>
41287         * tree-pretty-print.c (dump_generic_node): Dump INTEGER_CSTs
41288         as _Literal ( type ) number in case usual suffixes do not
41289         preserve all information.
41291 2017-01-13  Richard Biener  <rguenther@suse.de>
41293         PR tree-optimization/77283
41294         * gimple-ssa-split-paths.c: Include gimple-ssa.h, tree-phinodes.h
41295         and ssa-iterators.h.
41296         (is_feasible_trace): Implement a cost model based on joiner
41297         PHI node uses.
41299 2017-01-12  Michael Meissner  <meissner@linux.vnet.ibm.com>
41301         PR target/79004
41302         * config/rs6000/rs6000.md (FP_ISA3): Do not optimize converting
41303         char or short to __float128/_Float128 directly.
41305 2017-01-12  Martin Sebor  <msebor@redhat.com>
41307         to -Wformat-overflow.
41308         * gimple-ssa-sprintf.c (pass_sprintf_length::gate): Adjust.
41309         (min_bytes_remaining): Same.
41310         (get_string_length): Same.
41311         (format_string): Same.
41312         (format_directive): Same.
41313         (add_bytes): Same.
41314         (pass_sprintf_length::handle_gimple_call): Same.
41316 2017-01-12  Jakub Jelinek  <jakub@redhat.com>
41318         * gimple-ssa-sprintf.c (try_substitute_return_value): Remove
41319         info.nowrite calls with no lhs that can't throw.  Return bool
41320         whether gsi_remove has been called or not.
41321         (pass_sprintf_length::handle_gimple_call): Return bool whether
41322         try_substitute_return_value called gsi_remove.  Formatting fix.
41323         (pass_sprintf_length::execute): Don't use gsi_remove if
41324         handle_gimple_call returned true.
41326         PR bootstrap/79069
41327         * cfgrtl.c (rtl_tidy_fallthru_edge): For any_uncondjump_p that can't
41328         be removed due to side-effects, don't remove following barrier nor
41329         turn the successor edge into fallthru edge.
41331 2017-01-12  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
41333         PR target/79044
41334         * config/rs6000/rs6000.c (insn_is_swappable_p): Mark
41335         element-reversing loads and stores as not swappable.
41337 2017-01-12  Nathan Sidwell  <nathan@acm.org>
41338             Nicolai Stange  <nicstange@gmail.com>
41340         * combine.c (try_combine): Don't ignore result of overlap checking
41341         loop.  Combine overlap & asm check into single loop.
41343 2017-01-12  Richard Biener  <rguenther@suse.de>
41345         * tree-pretty-print.c (dump_generic_node): Provide -gimple
41346         variant for MEM_REF.  Sanitize INTEGER_CST for -gimple.
41348 2017-01-12  Richard Biener  <rguenther@suse.de>
41350         * tree.c (initialize_tree_contains_struct): Make TS_OPTIMIZATION
41351         and TS_TARGET_OPTION directly derive from TS_BASE.
41352         * tree-core.h (tree_optimization_option): Derive from tree_base.
41353         (tree_target_option): Likewise.
41355 2017-01-11  Uros Bizjak  <ubizjak@gmail.com>
41357         * config/i386/i386.c (memory_address_length): Increase len
41358         only when rip_relative_addr_p returns false.
41360 2017-01-11  Julia Koval  <julia.koval@intel.com>
41362         * common/config/i386/i386-common.c (OPTION_MASK_ISA_SGX_UNSET): New.
41363         (OPTION_MASK_ISA_SGX_SET): New.
41364         (ix86_handle_option): Handle OPT_msgx.
41365         * config.gcc: Added sgxintrin.h.
41366         * config/i386/driver-i386.c (host_detect_local_cpu): Detect sgx.
41367         * config/i386/i386-c.c (ix86_target_macros_internal): Define __SGX__.
41368         * config/i386/i386.c (ix86_target_string): Add -msgx.
41369         (PTA_SGX): New.
41370         (ix86_option_override_internal): Handle new options.
41371         (ix86_valid_target_attribute_inner_p): Add sgx.
41372         * config/i386/i386.h (TARGET_SGX, TARGET_SGX_P): New.
41373         * config/i386/i386.opt: Add msgx.
41374         * config/i386/sgxintrin.h: New file.
41375         * config/i386/x86intrin.h: Add sgxintrin.h.
41377 2017-01-11  Jakub Jelinek  <jakub@redhat.com>
41379         PR c++/71537
41380         * fold-const.c (maybe_nonzero_address): Return 1 for function
41381         local objects.
41382         (tree_single_nonzero_warnv_p): Don't handle function local objects
41383         here.
41385         PR c++/72813
41386         * gcc.c (default_compilers): Don't add -o %g.s for -S -save-temps
41387         of c-header.
41389 2017-01-11  David Malcolm  <dmalcolm@redhat.com>
41391         PR driver/78877
41392         * opts.c: Include "spellcheck.h"
41393         (struct string_fragment): New struct.
41394         (struct edit_distance_traits<const string_fragment &>): New
41395         struct.
41396         (get_closest_sanitizer_option): New function.
41397         (parse_sanitizer_options): Offer suggestions for unrecognized arguments.
41399 2017-01-11  Jakub Jelinek  <jakub@redhat.com>
41401         * dwarf2out.c (DWARF_COMPILE_UNIT_HEADER_SIZE): For DWARF5 decrease
41402         by 12.
41403         (DWARF_COMDAT_TYPE_UNIT_HEADER_SIZE): Always
41404         DWARF_COMPILE_UNIT_HEADER_SIZE plus 12.
41405         (DWARF_COMPILE_UNIT_SKELETON_HEADER_SIZE): Define.
41406         (calc_base_type_die_sizes): Use DWARF_COMPILE_UNIT_SKELETON_HEADER_SIZE
41407         for initial die_offset if dwarf_split_debug_info.
41408         (output_comp_unit): Use DWARF_COMPILE_UNIT_SKELETON_HEADER_SIZE for
41409         initial next_die_offset if dwo_id is non-NULL.  Don't emit padding
41410         fields.
41411         (output_skeleton_debug_sections): Formatting fix.  Use
41412         DWARF_COMPILE_UNIT_SKELETON_HEADER_SIZE instead of
41413         DWARF_COMPILE_UNIT_HEADER_SIZE.  Don't emit padding.
41415 2017-01-11  Wilco Dijkstra  <wdijkstr@arm.com>
41417         * config/arm/cortex-a53.md: Add bypasses for
41418         cortex_a53_r2f_cvt.
41419         (cortex_a53_r2f): Only use for transfers.
41420         (cortex_a53_f2r): Likewise.
41421         (cortex_a53_r2f_cvt): Add reservation for conversions.
41422         (cortex_a53_f2r_cvt): Likewise.
41424 2017-01-11  Tamar Christina  <tamar.christina@arm.com>
41426         * config/arm/arm_neon.h: Add __artificial__ and gnu_inline
41427         to all inlined functions, change static to extern.
41429 2017-01-11  Christophe Lyon  <christophe.lyon@linaro.org>
41431         PR target/78253
41432         * config/arm/arm.c (legitimize_pic_address): Handle reference to
41433         weak symbol.
41434         (arm_assemble_integer): Likewise.
41436 2017-01-11  Richard Earnshaw  <rearnsha@arm.com>
41438         * config.gcc: Use new awk script to check CPU, FPU and architecture
41439         parameters for --with-... options.
41440         * config/arm/parsecpu.awk: New file
41441         * config/arm/arm-cpus.in: New file.
41442         * config/arm/arm-opts.h: Include arm-cpu.h instead of processing .def
41443         files.
41444         * config/arm/arm.c: Include arm-cpu-data.h instead of processing .def
41445         files.
41446         * config/arm/t-arm: Update dependency rules.
41447         * common/config/arm/arm-common.c: Include arm-cpu-cdata.h instead
41448         of processing .def files.
41449         * config/arm/genopt.sh: Deleted.
41450         * config/arm/gentune.sh: Deleted.
41451         * config/arm/arm-cores.def: Deleted.
41452         * config/arm/arm-arches.def: Deleted.
41453         * config/arm/arm-fpus.def: Deleted.
41454         * config/arm/arm-tune.md: Regenerated.
41455         * config/arm/arm-tables.opt: Regenerated.
41456         * config/arm/arm-cpu.h: New generated file.
41457         * config/arm/arm-cpu-data.h: New generated file.
41458         * config/arm/arm-cpu-cdata.h: New generated file.
41460 2017-01-11  Maxim Ostapenko  <m.ostapenko@samsung.com>
41462         PR lto/79042
41463         * lto-cgraph.c (lto_output_varpool_node): Pack dynamically_initialized
41464         bit.
41465         (input_varpool_node): Unpack dynamically_initialized bit.
41467 2017-01-11  Eric Botcazou  <ebotcazou@adacore.com>
41469         PR rtl-optimization/79032
41470         * lra-constraints.c (simplify_operand_subreg): In the MEM case, test
41471         the alignment of the adjusted memory reference against that of MODE,
41472         instead of the alignment of the original memory reference.
41474 2017-01-11  Martin Jambor  <mjambor@suse.cz>
41476         * hsa.c (hsa_callable_function_p): Revert addition of DECL_ARTIFICIAL
41477         test.
41478         * ipa-hsa.c (process_hsa_functions): Only duplicate non-artificial
41479         decorated functions.
41481 2017-01-11  Richard Biener  <rguenther@suse.de>
41483         * tree-vrp.c (evrp_dom_walker::before_dom_children): Also
41484         set range/nonnull info for PHI results.  Do not set it on
41485         stmts marked for removal.
41487 2017-01-10  Eric Botcazou  <ebotcazou@adacore.com>
41489         * expr.c (store_field): In the bitfield case, fetch the return value
41490         from the registers before applying a single big-endian adjustment.
41491         Always do a final load for a BLKmode value not larger than a word.
41493 2017-01-10  David Malcolm  <dmalcolm@redhat.com>
41495         PR c++/77949
41496         * input.c (selftest::test_accessing_ordinary_linemaps): Verify
41497         that we correctly handle column numbers greater than
41498         LINE_MAP_MAX_COLUMN_NUMBER.
41500 2017-01-10  Martin Sebor  <msebor@redhat.com>
41502         PR middle-end/78245
41503         * gimple-ssa-sprintf.c (get_destination_size): Call
41504         {init,fini}object_sizes.
41505         * tree-object-size.c (addr_object_size): Adjust.
41506         (pass_through_call): Adjust.
41507         (pass_object_sizes::execute): Adjust.
41508         * tree-object-size.h (fini_object_sizes): Declare.
41510 2017-01-10  Martin Sebor  <msebor@redhat.com>
41512         PR tree-optimization/78775
41513         * builtins.c (get_size_range): Move...
41514         * calls.c: ...to here.
41515         (alloc_max_size): Accept zero argument.
41516         (operand_signed_p): Remove.
41517         (maybe_warn_alloc_args_overflow): Call get_size_range.
41518         * calls.h (get_size_range): Declare.
41520 2017-01-10  Joe Seymour  <joe.s@somniumtech.com>
41522         * config/msp430/driver-msp430.c (msp430_mcu_data): Sync with data
41523         from TI's devices.csv file as of September 2016.
41524         * config/msp430/msp430.c (msp430_mcu_data): Likewise.
41526 2017-01-10  Sandra Loosemore  <sandra@codesourcery.com>
41528         * doc/extend.texi: Tweak formatting to fix overfull hbox warnings.
41529         * doc/invoke.texi: Likewise.
41530         * doc/md.texi: Likewise.
41531         * doc/objc.texi: Likewise.
41533 2017-01-10  Joshua Conner  <joshconner@google.com>
41535         * config/arm/fuchsia-elf.h: New file.
41536         * config/fuchsia.h: New file.
41537         * config.gcc (*-*-fuchsia*): Set native_system_header_dir.
41538         (aarch64*-*-fuchsia*, arm*-*-fuchsia*, x86_64-*-fuchsia*): Add to
41539         targets.
41540         * config.host: (aarch64*-*-fuchsia*, arm*-*-fuchsia*): Add to hosts.
41542 2016-01-10  Richard Biener  <rguenther@suse.de>
41544         PR tree-optimization/79034
41545         * tree-call-cdce.c (shrink_wrap_one_built_in_call_with_conds):
41546         Propagate out degenerate PHIs in the joiner.
41548 2017-01-10  Martin Liska  <mliska@suse.cz>
41550         * ipa-icf.c (sort_sem_items_by_decl_uid): New function.
41551         (sort_congruence_classes_by_decl_uid): Likewise.
41552         (sort_congruence_class_groups_by_decl_uid): Likewise.
41553         (sem_item_optimizer::merge_classes): Sort class, groups in these
41554         classes and members in the groups by DECL_UID of declarations.
41555         This would make merge operations stable.
41557 2017-01-10  Martin Liska  <mliska@suse.cz>
41559         * ipa-icf.c (sem_item_optimizer::sem_item_optimizer): Remove
41560         usage of m_classes_vec.
41561         (sem_item_optimizer::~sem_item_optimizer):  Likewise.
41562         (sem_item_optimizer::get_group_by_hash): Likewise.
41563         (sem_item_optimizer::subdivide_classes_by_equality): Likewise.
41564         (sem_item_optimizer::subdivide_classes_by_sensitive_refs): Likewise.
41565         (sem_item_optimizer::verify_classes): Likewise.
41566         (sem_item_optimizer::process_cong_reduction): Likewise.
41567         (sem_item_optimizer::dump_cong_classes): Likewise.
41568         (sem_item_optimizer::merge_classes): Likewise.
41569         * ipa-icf.h (congruence_class_hash): Rename from
41570         congruence_class_group_hash.  Remove declaration of m_classes_vec.
41572 2017-01-10  Andrew Senkevich  <andrew.senkevich@intel.com>
41574         * common/config/i386/i386-common.c (OPTION_MASK_ISA_AVX512VPOPCNTDQ_SET,
41575         OPTION_MASK_ISA_AVX512VPOPCNTDQ_UNSET): New.
41576         * config.gcc: Add avx512vpopcntdqintrin.h.
41577         * config/i386/avx512vpopcntdqintrin.h: New.
41578         * config/i386/cpuid.h (bit_AVX512VPOPCNTDQ): New.
41579         * config/i386/i386-builtin-types.def: Add new types.
41580         * config/i386/i386-builtin.def (__builtin_ia32_vpopcountd_v16si,
41581         __builtin_ia32_vpopcountd_v16si_mask, __builtin_ia32_vpopcountq_v8di,
41582         __builtin_ia32_vpopcountq_v8di_mask): New.
41583         * config/i386/i386-c.c (ix86_target_macros_internal): Define
41584         __AVX512VPOPCNTDQ__.
41585         * config/i386/i386.c (ix86_target_string): Add -mavx512vpopcntdq.
41586         (PTA_AVX512VPOPCNTDQ): Define.
41587         * config/i386/i386.h (TARGET_AVX512VPOPCNTDQ,
41588         TARGET_AVX512VPOPCNTDQ_P): Define.
41589         * config/i386/i386.opt: Add mavx512vpopcntdq.
41590         * config/i386/immintrin.h: Include avx512vpopcntdqintrin.h.
41591         * config/i386/sse.md (define_insn "vpopcount<mode><mask_name>"): New.
41593 2017-01-01  Jan Hubicka  <hubicka@ucw.cz>
41595         PR middle-end/77484
41596         * predict.def (PRED_CALL): Set to 67.
41598 2017-01-09  Eric Botcazou  <ebotcazou@adacore.com>
41600         * expr.c (store_field): In the bitfield case, if the value comes from
41601         a function call and is of an aggregate type returned in registers, do
41602         not modify the field mode; extract the value in all cases if the mode
41603         is BLKmode and the size is not larger than a word.
41605 2017-01-09  Dominique d'Humieres  <dominiq@lps.ens.fr>
41607         PR target/71017
41608         * config/i386/cpuid.h: Fix undefined behavior.
41610 2017-01-04  Jeff Law  <law@redhat.com>
41612         PR tree-optimization/79007
41613         PR tree-optimization/67955
41614         * tree-ssa-alias.c (same_addr_size_stores_p): Only need to be
41615         conservative for pt.null when flag_non_call_exceptions is on.
41617 2017-01-09  Jakub Jelinek  <jakub@redhat.com>
41619         PR translation/79019
41620         PR translation/79020
41621         * params.def (PARAM_INLINE_MIN_SPEEDUP,
41622         PARAM_IPA_CP_SINGLE_CALL_PENALTY,
41623         PARAM_USE_AFTER_SCOPE_DIRECT_EMISSION_THRESHOLD): Fix typos
41624         in descriptions.
41625         * config/avr/avr.opt (maccumulate-args): Likewise.
41626         * config/msp430/msp430.opt (mwarn-mcu): Likewise.
41627         * common.opt (freport-bug): Likewise.
41628         * cif-code.def (CIF_FINAL_ERROR): Likewise.
41629         * doc/invoke.texi (ipa-cp-single-call-penalty): Likewise.
41630         * config/s390/s390.c (s390_invalid_binary_op): Fix spelling in
41631         translatable string.
41632         * config/i386/i386.c (function_value_32): Likewise.
41633         * config/nios2/nios2.c (nios2_valid_target_attribute_rec): Likewise.
41634         * config/msp430/msp430.c (msp430_option_override, msp430_attr):
41635         Likewise.
41636         * config/msp430/driver-msp430.c (msp430_select_hwmult_lib): Likewise.
41637         * common/config/msp430/msp430-common.c (msp430_handle_option):
41638         Likewise.
41639         * symtab.c (symtab_node::verify_base): Likewise.
41640         * opts.c (set_debug_level): Likewise.
41641         * tree.c (verify_type_variant): Likewise.  Fix typo in comment.
41642         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Add
41643         missing whitespace to translatable strings.
41644         * config/avr/avr.md (bswapsi2): Fix typo in comment.
41645         * config/sh/superh.h: Likewise.
41646         * config/i386/xopintrin.h: Likewise.
41647         * config/i386/znver1.md: Likewise.
41648         * config/rs6000/rs6000.c (struct rs6000_opt_mask): Likewise.
41649         * ipa-inline-analysis.c (compute_inline_parameters): Likewise.
41650         * double-int.h (struct double_int): Likewise.
41651         * double-int.c (div_and_round_double): Likewise.
41652         * wide-int.cc: Likewise.
41653         * tree-ssa.c (non_rewritable_mem_ref_base): Likewise.
41654         * tree-ssa-sccvn.c (vn_reference_lookup_3): Likewise.
41655         * cfgcleanup.c (crossjumps_occured): Renamed to ...
41656         (crossjumps_occurred): ... this.
41657         (try_crossjump_bb, try_head_merge_bb, try_optimize_cfg, cleanup_cfg):
41658         Adjust all uses.
41660         PR tree-optimization/78899
41661         * tree-if-conv.c (version_loop_for_if_conversion): Instead of
41662         returning bool return struct loop *, NULL for failure and the new
41663         loop on success.
41664         (versionable_outer_loop_p): Don't version outer loop if it has
41665         dont_vectorized bit set.
41666         (tree_if_conversion): When versioning outer loop, ensure
41667         tree_if_conversion is performed also on the inner loop of the
41668         non-vectorizable outer loop copy.
41669         * tree-vectorizer.c (set_uid_loop_bbs): Formatting fix.  Fold
41670         LOOP_VECTORIZED in inner loop of the scalar outer loop and
41671         prevent vectorization of it.
41672         (vectorize_loops): For outer + inner LOOP_VECTORIZED, ensure
41673         the outer loop vectorization of the non-scalar version is attempted
41674         before vectorization of the inner loop in scalar version.  If
41675         outer LOOP_VECTORIZED guarded loop is not vectorized, prevent
41676         vectorization of its inner loop.
41677         * tree-vect-loop-manip.c (rename_variables_in_bb): If outer_loop
41678         has 2 inner loops, rename also on edges from bb whose single pred
41679         is outer_loop->header.  Fix typo in function comment.
41681 2017-01-09  Martin Sebor  <msebor@redhat.com>
41683         PR bootstrap/79033
41684         * asan.c (asan_emit_stack_protection): Increase local buffer size
41685         to avoid snprintf truncation warning.
41687 2017-01-09  Andrew Pinski  <apinski@cavium.com>
41689         * config/aarch64/aarch64-cores.def: Add thunderx2t99.  Change vulcan
41690         to reference thunderx2t99 for the tuning structure
41691         * config/aarch64/aarch64-cost-tables.h (vulcan_extra_costs):
41692         Rename to ...
41693         (thunderx2t99_extra_costs): This.
41694         * config/aarch64/aarch64-tune.md: Regenerate.
41695         * config/aarch64/aarch64.c (vulcan_addrcost_table): Rename to ...
41696         (vulcan_addrcost_table): This.
41697         (vulcan_regmove_cost): Rename to ...
41698         (thunderx2t99_regmove_cost): This.
41699         (vulcan_vector_cost): Rename to ...
41700         (thunderx2t99_vector_cost): this.
41701         (vulcan_branch_cost): Rename to ...
41702         (thunderx2t99_branch_cost): This.
41703         (vulcan_tunings): Rename to ...
41704         (thunderx2t99_tunings): This and s/vulcan/thunderx2t99 .
41705         * doc/invoke.texi (AARCH64/mtune): Add thunderx2t99.
41707 2017-01-09  Martin Jambor  <mjambor@suse.cz>
41709         PR ipa/78365
41710         PR ipa/78599
41711         * ipa-prop.h (ipa_jump_func): Swap positions of vr_known and m_vr.
41712         * ipa-cp.c (ipa_vr_operation_and_type_effects): New function.
41713         (propagate_vr_accross_jump_function): Use the above function for all
41714         value range computations for pass-through jump functions and type
41715         converasion from explicit value range values.
41716         (ipcp_propagate_stage): Do not attempt to deduce types of formal
41717         parameters from TYPE_ARG_TYPES.
41718         * ipa-prop.c (ipa_write_jump_function): Remove trailing whitespace.
41719         (ipa_write_node_info): Stream type of the actual argument.
41720         (ipa_read_node_info): Likewise. Also remove trailing whitespace.
41722 2017-01-09  Martin Liska  <mliska@suse.cz>
41724         PR pch/78970
41725         * gcc.c (driver_handle_option): Handle OPT_E and set have_E.
41726         (lookup_compiler): Do not show error message with have_E.
41728 2017-01-09  Jakub Jelinek  <jakub@redhat.com>
41730         PR tree-optimization/78938
41731         * tree-vect-stmts.c (vectorizable_condition): For non-masked COND_EXPR
41732         where comp_vectype is VECTOR_BOOLEAN_TYPE_P, use
41733         BIT_{NOT,XOR,AND,IOR}_EXPR on the comparison operands instead of
41734         {EQ,NE,GE,GT,LE,LT}_EXPR directly inside of VEC_COND_EXPR.  Formatting
41735         fixes.
41737 2017-01-09  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
41739         * tree-ssa-address.c (gen_addr_rtx): Don't handle index if it
41740         is const0_rtx.
41742 2017-01-09  Richard Biener  <rguenther@suse.de>
41744         PR tree-optimization/78997
41745         * tree-vect-slp.c (vect_mask_constant_operand_p): Handle SSA
41746         name condition properly.
41748 2017-01-09  Richard Biener  <rguenther@suse.de>
41750         PR debug/79000
41751         * dwarf2out.c (is_cxx): New overload with context.
41752         (is_naming_typedef_decl): Use it.
41754 2017-01-08  Sandra Loosemore  <sandra@codesourcery.com>
41756         * invoke.texi (Option Summary): Correct spacing in option lists
41757         and add line breaks to fix over-long lines.
41759 2017-01-08  Sandra Loosemore  <sandra@codesourcery.com>
41761         PR middle-end/17660
41763         * extend.texi (Common Variable Attributes): Add xref to GCC
41764         Internals manual to explain mode attribute keywords.
41766 2017-01-08  Sandra Loosemore  <sandra@codesourcery.com>
41768         PR other/16519
41769         * doc/invoke.texi (Option Summary): Move -pthread to Linker Options
41770         and Preprocessor Options.
41771         (Options for Linking): Document -pthread here....
41772         (RS/6000 and PowerPC Options): ...not here.
41773         (Solaris 2 Options): ...or here.
41774         * doc/cppopts.texi: Document -pthread.
41776 2017-01-08  Martin Sebor  <msebor@redhat.com>
41778         PR middle-end/77708
41779         * doc/invoke.texi (Warning Options): Document -Wformat-truncation.
41780         * gimple-ssa-sprintf.c (call_info::reval_used, call_info::warnopt):
41781         New member functions.
41782         (format_directive): Used them.
41783         (add_bytes): Same.
41784         (pass_sprintf_length::handle_gimple_call): Same.
41785         * graphite-sese-to-poly.c (tree_int_to_gmp): Increase buffer size
41786         to avoid truncation for any argument.
41787         (extract_affine_mul): Same.
41788         * tree.c (get_file_function_name): Same.
41790 2017-01-01  Jan Hubicka  <hubicka@ucw.cz>
41792         PR middle-end/77484
41793         * predict.def (PRED_INDIR_CALL): Set to 86.
41795 2017-01-07  Sandra Loosemore  <sandra@codesourcery.com>
41797         PR preprocessor/54124
41798         * doc/cppopts.texi: Reformat -d subtable to list the full name
41799         of the options.  Add cross-reference to the docs for the general
41800         compiler -d options.
41801         * doc/invoke.texi (Developer Options): Add cross-reference to the
41802         preprocessor-specific -d option documentation.
41804 2017-01-07  Sandra Loosemore  <sandra@codesourcery.com>
41806         PR preprocessor/13498
41807         * doc/cpp.texi (Search Path): Rewrite to remove obsolete and
41808         redudant material, and reflect new command-line options.
41809         (System Headers): Likewise.
41811 2017-01-07  Sandra Loosemore  <sandra@codesourcery.com>
41813         * doc/cppdiropts.texi: Merge documentation of -I, -iquote,
41814         -isystem, and -idirafter.  Copy-edit.
41815         * doc/cppopts.texi: Copy-edit.  Remove contradiction about
41816         default for -ftrack-macro-expansion.  Delete obsolete and
41817         badly-formatted implementation details about -fdebug-cpp output.
41818         * doc/cppwarnopts.texi: Copy-edit.
41820 2017-01-07  David Malcolm  <dmalcolm@redhat.com>
41822         PR c++/72803
41823         * input.c (selftest::test_accessing_ordinary_linemaps): Verify
41824         that the transition from a max line width >= 1<<10 to narrower
41825         lines works correctly.
41827 2017-01-07  Alexandre Oliva  <aoliva@redhat.com>
41829         * doc/options.texi (PerFunction): New.
41830         * opt-functions.awk (switch_flags): Map both Optimization and
41831         PerFunction to CL_OPTIMIZATION.
41832         * opth-gen.awk: Test for PerFunction flag along with
41833         Optimization.
41834         * optc-save-gen.awk: Likewise.  Introduce var_opt_hash and set
41835         it only when the latter is present.  Skip those that don't in
41836         the hash function generator.
41837         * common.opt (fvar-tracking): Mark as PerFunction instead of
41838         Optimization.
41839         (fvar-tracking-assignments): Likewise.
41840         (fvar-tracking-assignments-toggle): Likewise.
41841         (fvar-tracking-uninit): Likewise.
41843 2017-01-07  Jakub Jelinek  <jakub@redhat.com>
41845         PR translation/79018
41846         * params.def (PARAM_MAX_STORES_TO_MERGE): Add missing space between
41847         the and store.
41849 2017-01-06  Mikael Pettersson  <mikpelinux@gmail.com>
41851         PR target/57583
41852         * config/m68k/m68k.opt (LONG_JUMP_TABLE_OFFSETS): New option.
41853         * config/m68k/linux.h (ASM_RETURN_CASE_JUMP): Handle
41854         TARGET_LONG_JUMP_TABLE_OFFSETS.
41855         * config/m68k/m68kelf.h (ASM_RETURN_CASE_JUMP): Likewise.
41856         * config/m68k/netbsd-elf.h (ASM_RETURN_CASE_JUMP): Likewise.
41857         * config/m68k/m68k.h (CASE_VECTOR_MODE): Likewise.
41858         (ASM_OUTPUT_ADDR_DIFF_ELF): Likewise.
41859         * config/m68k/m68k.md (tablejump expander): Likewise.
41860         (*tablejump_pcrel_hi): Renamed from unnamed insn, reject
41861         TARGET_LONG_JUMP_TABLE_OFFSETS.
41862         (*tablejump_pcrel_si): New insn, handle TARGET_LONG_JUMP_TABLE_OFFSETS.
41863         * doc/invoke.texi (M68K options): Add -mlong-jump-table-offsets.
41865 2017-01-06  Edgar E. Iglesias <edgar.iglesias@xilinx.com>
41866             David Holsgrove <david.holsgrove@xilinx.com>
41868         * common/config/microblaze/microblaze-common.c
41869         (TARGET_EXCEPT_UNWIND_INFO): Remove.
41870         * config/microblaze/microblaze-protos.h (microblaze_eh_return):
41871         New prototype.
41872         * config/microblaze/microblaze.c (microblaze_must_save_register)
41873         (microblaze_expand_epilogue, microblaze_return_addr): Handle
41874         calls_eh_return.
41875         (microblaze_eh_return): New function.
41876         * config/microblaze/microblaze.h (RETURN_ADDR_OFFSET)
41877         (EH_RETURN_DATA_REGNO, MB_EH_STACKADJ_REGNUM)
41878         (EH_RETURN_STACKADJ_RTX, ASM_PREFERRED_EH_DATA_FORMAT): New macros.
41879         * config/microblaze/microblaze.md (eh_return): New pattern.
41881 2017-01-06  Jakub Jelinek  <jakub@redhat.com>
41883         * system.h (GCC_DIAGNOSTIC_PUSH_IGNORED, GCC_DIAGNOSTIC_POP,
41884         GCC_DIAGNOSTIC_STRINGIFY): Define.
41886         * read-rtl.c (rtx_reader::read_rtx_code): Avoid -Wsign-compare warning.
41888 2017-01-06  Andre Vieira  <andre.simoesdiasvieira@arm.com>
41890         * config/arm/arm.md (<mcrr>): New.
41891         (<mrrc>): New.
41892         * config/arm/arm.c (arm_arch5te): New.
41893         (arm_option_override): Set arm_arch5te.
41894         (arm_coproc_builtin_available): Add support for mcrr, mcrr2, mrrc
41895         and mrrc2.
41896         * config/arm/arm-builtins.c (MCRR_QUALIFIERS): Define to...
41897         (arm_mcrr_qualifiers): ... this. New.
41898         (MRRC_QUALIFIERS): Define to...
41899         (arm_mrrc_qualifiers): ... this. New.
41900         * config/arm/arm_acle.h (__arm_mcrr, __arm_mcrr2, __arm_mrrc,
41901         __arm_mrrc2): New.
41902         * config/arm/arm_acle_builtins.def (mcrr, mcrr2, mrrc, mrrc2): New.
41903         * config/arm/iterators.md (MCRRI, mcrr, MCRR): New.
41904         (MRRCI, mrrc, MRRC): New.
41905         * config/arm/unspecs.md (VUNSPEC_MCRR, VUNSPEC_MCRR2, VUNSPEC_MRRC,
41906         VUNSPEC_MRRC2): New.
41908 2017-01-06  Andre Vieira  <andre.simoesdiasvieira@arm.com>
41910         * config/arm/arm.md (<mcr>): New.
41911         (<mrc>): New.
41912         * config/arm/arm.c (arm_coproc_builtin_available): Add
41913         support for mcr, mrc, mcr2 and mrc2.
41914         * config/arm/arm-builtins.c (MCR_QUALIFIERS): Define to...
41915         (arm_mcr_qualifiers): ... this. New.
41916         (MRC_QUALIFIERS): Define to ...
41917         (arm_mrc_qualifiers): ... this. New.
41918         (MCR_QUALIFIERS): Define to ...
41919         (arm_mcr_qualifiers): ... this. New.
41920         * config/arm/arm_acle.h (__arm_mcr, __arm_mrc, __arm_mcr2,
41921         __arm_mrc2): New.
41922         * config/arm/arm_acle_builtins.def (mcr, mcr2, mrc, mrc2): New.
41923         * config/arm/iterators.md (MCRI, mcr, MCR, MRCI, mrc, MRC): New.
41924         * config/arm/unspecs.md (VUNSPEC_MCR, VUNSPEC_MCR2, VUNSPEC_MRC,
41925         VUNSPEC_MRC2): New.
41927 2017-01-06  Andre Vieira  <andre.simoesdiasvieira@arm.com>
41929         * config/arm/arm.md (*ldc): New.
41930         (*stc): New.
41931         (<ldc>): New.
41932         (<stc>): New.
41933         * config/arm/arm.c (arm_coproc_builtin_available): Add
41934         support for ldc,ldcl,stc,stcl,ldc2,ldc2l,stc2 and stc2l.
41935         (arm_coproc_ldc_stc_legitimate_address): New.
41936         * config/arm/arm-builtins.c (arm_type_qualifiers): Add
41937         'qualifier_const_pointer'.
41938         (LDC_QUALIFIERS): Define to...
41939         (arm_ldc_qualifiers): ... this. New.
41940         (STC_QUALIFIERS): Define to...
41941         (arm_stc_qualifiers): ... this. New.
41942         * config/arm/arm-protos.h
41943         (arm_coproc_ldc_stc_legitimate_address): New.
41944         * config/arm/arm_acle.h (__arm_ldc, __arm_ldcl, __arm_stc,
41945         __arm_stcl, __arm_ldc2, __arm_ldc2l, __arm_stc2, __arm_stc2l): New.
41946         * config/arm/arm_acle_builtins.def (ldc, ldc2, ldcl, ldc2l, stc,
41947         stc2, stcl, stc2l): New.
41948         * config/arm/constraints.md (Uz): New.
41949         * config/arm/iterators.md (LDCI, STCI, ldc, stc, LDC STC): New.
41950         * config/arm/unspecs.md (VUNSPEC_LDC, VUNSPEC_LDC2, VUNSPEC_LDCL,
41951         VUNSPEC_LDC2L, VUNSPEC_STC, VUNSPEC_STC2, VUNSPEC_STCL,
41952         VUNSPEC_STC2L): New.
41954 2017-01-06  Andre Vieira  <andre.simoesdiasvieira@arm.com>
41956         * config/arm/arm.md (<cdp>): New.
41957         * config/arm/arm.c (neon_const_bounds): Rename this ...
41958         (arm_const_bounds): ... this.
41959         (arm_coproc_builtin_available): New.
41960         * config/arm/arm-builtins.c (SIMD_MAX_BUILTIN_ARGS): Increase.
41961         (arm_type_qualifiers): Add 'qualifier_unsigned_immediate'.
41962         (CDP_QUALIFIERS): Define to...
41963         (arm_cdp_qualifiers): ... this. New.
41964         (void_UP): Define.
41965         (arm_expand_builtin_args): Add case for 6 arguments.
41966         * config/arm/arm-protos.h (neon_const_bounds): Rename this ...
41967         (arm_const_bounds): ... this.
41968         (arm_coproc_builtin_available): New.
41969         * config/arm/arm_acle.h (__arm_cdp): New.
41970         (__arm_cdp2): New.
41971         * config/arm/arm_acle_builtins.def (cdp): New.
41972         (cdp2): New.
41973         * config/arm/iterators.md (CDPI,CDP,cdp): New.
41974         * config/arm/neon.md: Rename all 'neon_const_bounds' to
41975         'arm_const_bounds'.
41976         * config/arm/types.md (coproc): New.
41977         * config/arm/unspecs.md (VUNSPEC_CDP, VUNSPEC_CDP2): New.
41978         * doc/extend.texi (ACLE): Add a mention of Coprocessor intrinsics.
41979         * doc/sourcebuild.texi (arm_coproc1_ok, arm_coproc2_ok,
41980         arm_coproc3_ok, arm_coproc4_ok): Document new effective targets.
41982 2017-01-06  Andre Vieira  <andre.simoesdiasvieira@arm.com>
41984         * config/arm/arm-builtins.c (arm_unsigned_binop_qualifiers): New.
41985         (UBINOP_QUALIFIERS): New.
41986         (si_UP): Define.
41987         (acle_builtin_data): New. Change comment.
41988         (arm_builtins): Remove ARM_BUILTIN_CRC32B, ARM_BUILTIN_CRC32H,
41989         ARM_BUILTIN_CRC32W, ARM_BUILTIN_CRC32CB, ARM_BUILTIN_CRC32CH,
41990         ARM_BUILTIN_CRC32CW. Add ARM_BUILTIN_ACLE_BASE and include
41991         arm_acle_builtins.def.
41992         (ARM_BUILTIN_ACLE_PATTERN_START): Define.
41993         (arm_init_acle_builtins): New.
41994         (CRC32_BUILTIN): Remove.
41995         (bdesc_2arg): Remove entries for crc32b, crc32h, crc32w,
41996         crc32cb, crc32ch and crc32cw.
41997         (arm_init_crc32_builtins): Remove.
41998         (arm_init_builtins): Use arm_init_acle_builtins rather
41999         than arm_init_crc32_builtins.
42000         (arm_expand_acle_builtin): New.
42001         (arm_expand_builtin): Use 'arm_expand_acle_builtin'.
42002         * config/arm/arm_acle_builtins.def: New.
42004 2017-01-06  Andre Vieira  <andre.simoesdiasvieira@arm.com>
42006         * config/arm/arm-builtins.c (neon_builtin_datum): Rename to ..
42007         (arm_builtin_datum): ... this.
42008         (arm_init_neon_builtin): Rename to ...
42009         (arm_init_builtin): ... this. Add a new parameters PREFIX
42010         and USE_SIG_IN_NAME.
42011         (arm_init_neon_builtins): Replace 'arm_init_neon_builtin' with
42012         'arm_init_builtin'. Replace type 'neon_builtin_datum' with
42013         'arm_builtin_datum'.
42014         (arm_init_vfp_builtins): Likewise.
42015         (builtin_arg): Rename enum's replacing 'NEON_ARG' with
42016         'ARG_BUILTIN' and add a 'ARG_BUILTIN_NEON_MEMORY.
42017         (arm_expand_neon_args): Rename to ...
42018         (arm_expand_builtin_args): ... this. Rename builtin_arg
42019         enum values and differentiate between ARG_BUILTIN_MEMORY
42020         and ARG_BUILTIN_NEON_MEMORY.
42021         (arm_expand_neon_builtin_1): Rename to ...
42022         (arm_expand_builtin_1): ... this. Rename builtin_arg enum
42023         values, arm_expand_builtin_args and add bool parameter NEON.
42024         (arm_expand_neon_builtin): Use arm_expand_builtin_1.
42025         (arm_expand_vfp_builtin): Likewise.
42026         (NEON_MAX_BUILTIN_ARGS): Remove, it was unused.
42028 2017-01-01  Jan Hubicka  <hubicka@ucw.cz>
42030         PR middle-end/77484
42031         * predict.def (PRED_POLYMORPHIC_CALL): Set to 59.
42032         * predict.c (tree_estimate_probability_bb): Reverse direction of
42033         polymorphic call predictor.
42035 2017-01-06  David Malcolm  <dmalcolm@redhat.com>
42037         * passes.c (execute_one_pass): Split out pass-skipping logic into...
42038         (determine_pass_name_match): ...this new function and...
42039         (should_skip_pass_p): ...this new function.
42041 2017-01-06  Nathan Sidwell  <nathan@acm.org>
42043         * ipa-visibility.c (function_and_variable_visibility): Reformat
42044         comments and long lines.  Remove extrneous if.
42045         * symtab.c (symtab_node::make_decl_local): Fix code format.
42046         (symtab_node::set_section_for_node): Fix comment typo.
42048 2017-01-06  Martin Liska  <mliska@suse.cz>
42050         PR bootstrap/79003
42051         * lra-constraints.c: Rename invariant to lra_invariant.
42052         * predict.c (set_even_probabilities): Initialize e to NULL.
42054 2017-01-05  Martin Sebor  <msebor@redhat.com>
42056         PR tree-optimization/78910
42057         * gimple-ssa-sprintf.c (tree_digits): Add an argument.
42058         (format_integer): Correct off-by-one error in the handling
42059         of precision with negative numbers in signed conversions..
42061 2017-01-05  Eric Botcazou  <ebotcazou@adacore.com>
42063         * doc/invoke.texi (C Dialect Options): Adjust -fsso-struct entry.
42065 2017-01-05  Jakub Jelinek  <jakub@redhat.com>
42067         PR tree-optimization/71016
42068         * tree-ssa-phiopt.c (tree_ssa_phiopt_worker): Pass cond_stmt to
42069         factor_out_conditional_conversion.  Formatting fix.
42070         (factor_out_conditional_conversion): Add cond_stmt argument.
42071         If arg1 is INTEGER_CST, punt if new_arg0 is not any operand of
42072         cond_stmt and if arg0_def_stmt is not the only stmt in its bb.
42073         Formatting fix.
42075 2017-01-05  David Malcolm  <dmalcolm@redhat.com>
42077         * Makefile.in (OBJS): Add read-md.o, read-rtl.o,
42078         read-rtl-function.o, and selftest-rtl.o.
42079         * config/aarch64/aarch64.c: Include selftest.h and selftest-rtl.h.
42080         (selftest::aarch64_test_loading_full_dump): New function.
42081         (selftest::aarch64_run_selftests): New function.
42082         (TARGET_RUN_TARGET_SELFTESTS): Wire it up to
42083         selftest::aarch64_run_selftests.
42084         * config/i386/i386.c
42085         (selftest::ix86_test_loading_dump_fragment_1): New function.
42086         (selftest::ix86_test_loading_call_insn): New function.
42087         (selftest::ix86_test_loading_full_dump): New function.
42088         (selftest::ix86_test_loading_unspec): New function.
42089         (selftest::ix86_run_selftests): Call the new functions.
42090         * emit-rtl.c (maybe_set_max_label_num): New function.
42091         * emit-rtl.h (maybe_set_max_label_num): New decl.
42092         * function.c (instantiate_decls): Guard call to
42093         instantiate_decls_1 with if (DECL_INITIAL (fndecl)).
42094         * function-tests.c (selftest::verify_three_block_rtl_cfg): Remove
42095         "static".
42096         * gensupport.c (gen_reader::gen_reader): Pass "false"
42097         for new "compact" param of rtx_reader.
42098         * print-rtl.c (rtx_writer::print_rtx_operand): Print "(nil)"
42099         rather than an empty string for NULL strings.
42100         * read-md.c: Potentially include config.h rather than bconfig.h.
42101         Wrap include of errors.h with #ifdef GENERATOR_FILE.
42102         (have_error): New global, copied from errors.c.
42103         (md_reader::read_name): Rename to...
42104         (md_reader::read_name_1): ...this, adding "out_loc" param,
42105         and converting "missing name or number" to returning false, rather
42106         than failing.
42107         (md_reader::read_name): Reimplement in terms of read_name_1.
42108         (md_reader::read_name_or_nil): New function.
42109         (md_reader::read_string): Handle "(nil)" by returning NULL.
42110         (md_reader::md_reader): Add new param "compact".
42111         (md_reader::read_md_files): Wrap with #ifdef GENERATOR_FILE.
42112         (md_reader::read_file): New method.
42113         * read-md.h (md_reader::md_reader): Add new param "compact".
42114         (md_reader::read_file): New method.
42115         (md_reader::is_compact): New accessor.
42116         (md_reader::read_name): Convert return type from void to file_location.
42117         (md_reader::read_name_or_nil): New decl.
42118         (md_reader::read_name_1): New decl.
42119         (md_reader::m_compact): New field.
42120         (noop_reader::noop_reader): Pass "false" for new "compact" param
42121         of rtx_reader.
42122         (rtx_reader::rtx_reader): Add new "compact" param.
42123         (rtx_reader::read_rtx_operand): Make virtual and convert return
42124         type from void to rtx.
42125         (rtx_reader::read_until): New decl.
42126         (rtx_reader::handle_any_trailing_information): New virtual function.
42127         (rtx_reader::postprocess): New virtual function.
42128         (rtx_reader::finalize_string): New virtual function.
42129         (rtx_reader::m_in_call_function_usage): New field.
42130         (rtx_reader::m_reuse_rtx_by_id): New field.
42131         * read-rtl-function.c: New file.
42132         * selftest-rtl.c (selftest::assert_rtx_ptr_eq_at): New function.
42133         * selftest-rtl.h (ASSERT_RTX_PTR_EQ): New macro.
42134         (selftest::verify_three_block_rtl_cfg): New decl.
42135         * read-rtl-function.h: New file.
42136         * read-rtl.c: Potentially include config.h rather than bconfig.h.
42137         For host, include function.h, memmodel.h, and emit-rtl.h.
42138         (one_time_initialization): New function.
42139         (struct compact_insn_name): New struct.
42140         (compact_insn_names): New array.
42141         (find_code): Handle insn codes in compact dumps.
42142         (apply_subst_iterator): Wrap with #ifdef GENERATOR_FILE.
42143         (bind_subst_iter_and_attr): Likewise.
42144         (add_condition_to_string): Likewise.
42145         (add_condition_to_rtx): Likewise.
42146         (apply_attribute_uses): Likewise.
42147         (add_current_iterators): Likewise.
42148         (apply_iterators): Likewise.
42149         (initialize_iterators): Guard usage of apply_subst_iterator with
42150         #ifdef GENERATOR_FILE.
42151         (read_conditions): Wrap with #ifdef GENERATOR_FILE.
42152         (md_reader::read_mapping): Likewise.
42153         (add_define_attr_for_define_subst): Likewise.
42154         (add_define_subst_attr): Likewise.
42155         (read_subst_mapping): Likewise.
42156         (check_code_iterator): Likewise.
42157         (rtx_reader::read_rtx): Likewise.  Move one-time initialization
42158         logic to...
42159         (one_time_initialization): New function.
42160         (rtx_reader::read_until): New method.
42161         (read_flags): New function.
42162         (parse_reg_note_name): New function.
42163         (rtx_reader::read_rtx_code): Initialize "iterator" to NULL.
42164         Handle reuse_rtx ids.
42165         Wrap iterator lookup within #ifdef GENERATOR_FILE.
42166         Add parsing support for RTL dumps, mirroring the special-cases in
42167         print_rtx, by calling read_flags, reading REG_NOTE names, INSN_UID
42168         values, and calling handle_any_trailing_information.
42169         (rtx_reader::read_rtx_operand): Convert return type from void
42170         to rtx, returning return_rtx.  Handle case 'e'.  Call
42171         finalize_string on XSTR and XTMPL fields.
42172         (rtx_reader::read_nested_rtx):  Handle dumps in which trailing
42173          "(nil)" values were omitted.  Call the postprocess vfunc on the
42174         return_rtx.
42175         (rtx_reader::rtx_reader): Add new "compact" param and pass to base
42176         class ctor.  Initialize m_in_call_function_usage.  Call
42177         one_time_initialization.
42178         * rtl-tests.c (selftest::test_uncond_jump): Call
42179         set_new_first_and_last_insn.
42180         * rtl.h (read_rtx): Wrap decl with #ifdef GENERATOR_FILE.
42181         * selftest-rtl.c: New file.
42182         * selftest-rtl.h (class selftest::rtl_dump_test): New class.
42183         (selftest::get_insn_by_uid): New decl.
42184         * selftest-run-tests.c (selftest::run_tests): Call
42185         read_rtl_function_c_tests.
42186         * selftest.h  (selftest::read_rtl_function_c_tests): New decl.
42187         * tree-dfa.c (ssa_default_def): Return NULL_TREE for rtl function
42188         dumps.
42190 2017-01-05  Uros Bizjak  <ubizjak@gmail.com>
42192         * config/i386/i386.md (*testqi_ext_3): No need to handle memory
42193         operands in a special way.  Assert that pos+len <= mode precision.
42195 2017-01-05  Jakub Jelinek  <jakub@redhat.com>
42197         * common.opt (fvect-cost-model): Remove RejectNegative flag, use
42198         3 argument Alias with unlimited for the negative form.
42199         (fno-vect-cost-model): Removed.
42201 2017-01-05  Martin Liska  <mliska@suse.cz>
42203         * hsa-gen.c (gen_hsa_divmod): New function.
42204         (gen_hsa_insn_for_internal_fn_call): Use the function for IFN_DIVMOD.
42206 2017-01-05  Martin Liska  <mliska@suse.cz>
42208         PR pch/78970
42209         * gcc.c (lookup_compiler): Reject '-' filename for a precompiled
42210         header.
42212 2017-01-05  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
42214         * config/s390/s390.c (s390_expand_setmem): Unroll the loop for
42215         small constant length operands.
42217 2017-01-05  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
42219         * config/s390/s390.c (s390_expand_setmem): Avoid overlapping bytes
42220         between loop iterations.
42222 2017-01-05  Martin Liska  <mliska@suse.cz>
42224         PR sanitizer/78815
42225         * gimplify.c (gimplify_decl_expr): Compare to
42226         asan_poisoned_variables instread of checking flags.
42227         (gimplify_target_expr): Likewise.
42228         (gimplify_expr): Likewise.
42229         (gimplify_function_tree): Conditionally initialize
42230         asan_poisoned_variables.
42232 2017-01-04  Jeff Law  <law@redhat.com>
42234         PR tree-optimizatin/78812
42235         * rtl.h (contains_mem_rtx_p): Prototype.
42236         * ifcvt.c (containts_mem_rtx_p): Move from here to...
42237         * rtlanal.c (contains_mem_rtx_p): Here and remove static linkage.
42238         * gcse.c (prune_expressions): Use contains_mem_rtx_p to discover
42239         and prune MEMs that are not at the toplevel of a SET_SRC rtx.  Look
42240         through ZERO_EXTEND and SIGN_EXTEND when trying to avoid pruning MEMs.
42242 2017-01-04  Alexandre Oliva  <aoliva@redhat.com>
42244         * input.c (assert_char_at_range): Default-initialize actual_range.
42246 2017-01-04  Alexandre Oliva  <aoliva@redhat.com>
42248         * df-scan.c (df_ref_create_structure): Make regno unsigned,
42249         to match the caller.
42251 2017-01-04  Alexandre Oliva  <aoliva@redhat.com>
42253         * cfgexpand.c (expand_gimple_basic_block): Disregard debug
42254         insns after final jump in test to emit dummy move.
42256 2017-01-04  Alexandre Oliva  <aoliva@redhat.com>
42258         * gimple-iterator.h (gsi_one_nondebug_before_end_p): New.
42259         * tree-eh.c (cleanup_empty_eh): Skip more debug stmts.
42261 2017-01-04  Alexandre Oliva  <aoliva@redhat.com>
42263         * multiple_target.c (create_dispatcher_calls): Init e_next.
42264         * tree-ssa-loop-split.c (split_loop): Init border.
42265         * tree-vect-loop.c (vect_determine_vectorization_factor): Init
42266         scalar_type.
42268 2017-01-04  Michael Meissner  <meissner@linux.vnet.ibm.com>
42270         PR target/71977
42271         PR target/70568
42272         PR target/78823
42273         * config/rs6000/predicates.md (sf_subreg_operand): New predicate.
42274         (altivec_register_operand): Do not return true if the operand
42275         contains a SUBREG mixing SImode and SFmode.
42276         (vsx_register_operand): Likewise.
42277         (vsx_reg_sfsubreg_ok): New predicate.
42278         (vfloat_operand): Do not return true if the operand contains a
42279         SUBREG mixing SImode and SFmode.
42280         (vint_operand): Likewise.
42281         (vlogical_operand): Likewise.
42282         (gpc_reg_operand): Likewise.
42283         (int_reg_operand): Likewise.
42284         * config/rs6000/rs6000-protos.h (valid_sf_si_move): Add declaration.
42285         * config/rs6000/rs6000.c (valid_sf_si_move): New function to
42286         determine if a MOVSI or MOVSF operation contains SUBREGs that mix
42287         SImode and SFmode.
42288         (rs6000_emit_move_si_sf_subreg): New helper function.
42289         (rs6000_emit_move): Call rs6000_emit_move_si_sf_subreg to possbily
42290         fixup SUBREGs involving SImode and SFmode.
42291         * config/rs6000/vsx.md (SFBOOL_*): New constants that are operand
42292         numbers for the new peephole2 optimization.
42293         (peephole2 for SFmode unions): New peephole2 to optimize cases in
42294         the GLIBC math library that do AND/IOR/XOR operations on single
42295         precision floating point.
42296         * config/rs6000/rs6000.h (TARGET_NO_SF_SUBREG): New internal
42297         target macros to say whether we need to avoid SUBREGs mixing
42298         SImode and SFmode.
42299         (TARGET_ALLOW_SF_SUBREG): Likewise.
42300         * config/rs6000/rs6000.md (UNSPEC_SF_FROM_SI): New unspecs.
42301         (UNSPEC_SI_FROM_SF): Likewise.
42302         (iorxor): Change spacing.
42303         (and_ior_xor): New iterator for AND, IOR, and XOR.
42304         (movsi_from_sf): New insns for SImode/SFmode SUBREG support.
42305         (movdi_from_sf_zero_ext): Likewise.
42306         (mov<mode>_hardfloat, FMOVE32 iterator): Use register_operand
42307         instead of gpc_reg_operand.  Add SImode/SFmode SUBREG support.
42308         (movsf_from_si): New insn for SImode/SFmode SUBREG support.
42309         (fma<mode>4): Use gpc_reg_operand instead of register_operand.
42310         (fms<mode>4): Likewise.
42311         (fnma<mode>4): Likewise.
42312         (fnms<mode>4): Likewise.
42313         (nfma<mode>4): Likewise.
42314         (nfms<mode>4): Likewise.
42316 2017-01-04  Marek Polacek  <polacek@redhat.com>
42318         PR c++/64767
42319         * doc/invoke.texi: Document -Wpointer-compare.
42321 2017-01-04  Jakub Jelinek  <jakub@redhat.com>
42323         * optc-gen.awk: Emit #error for -W*/-f*/-m* Enum without
42324         RejectNegative.
42326         * dwarf2out.c (output_loc_list): Don't throw away 64K+ location
42327         descriptions for -gdwarf-5 and emit them as uleb128 instead of
42328         2-byte data.
42330 2017-01-04  Kelvin Nilsen  <kelvin@gcc.gnu.org>
42332         PR target/78056
42333         * doc/sourcebuild.texi (PowerPC-specific attributes): Add
42334         documentation of the powerpc_popcntb_ok attribute.
42335         * config/rs6000/rs6000.c (rs6000_option_override_internal): Add
42336         code to issue warning messages if a requested CPU configuration is
42337         not supported by the binary (assembler and loader) toolchain.
42338         (spe_init_builtins): Add two assertions to prevent ICE if attempt is
42339         made to define a built-in function that has been disabled.
42340         (paired_init_builtins): Add assertion to prevent ICE if attempt is
42341         made to define a built-in function that has been disabled.
42342         (altivec_init_builtins): Add comment explaining why definition
42343         of the DST built-in functions is not preceded by an assertion
42344         check.  Add assertions to prevent ICE if attempts are made to
42345         define an altivec predicate or an abs* built-in function that has
42346         been disabled.
42347         (htm_init_builtins): Add comment explaining why definition of the
42348         htm built-in functions is not preceded by an assertion check.
42350 2017-01-04  Jeff Law  <law@redhat.com>
42352         PR tree-optimizatin/67955
42353         * tree-ssa-alias.c (same_addr_size_stores_p): Check offsets first.
42354         Allow any SSA_VAR_P as the base objects.  Use integer_zerop.  Verify
42355         the points-to solution does not include pt_null.  Use DECL_PT_UID
42356         unconditionally.
42358 2017-01-04  Uros Bizjak  <ubizjak@gmail.com>
42360         * config/i386/i386.md (HI/SImode test with imm to QImode splitters):
42361         Use gen_int_mode instead of gen_lopwart for const_int operands.
42363 2017-01-04  Jakub Jelinek  <jakub@redhat.com>
42365         PR tree-optimization/71563
42366         * match.pd: Simplify X << Y into X if Y is known to be 0 or
42367         out of range value - has low bits known to be zero.
42369 2017-01-04  Alan Modra  <amodra@gmail.com>
42371         * Makefile.in (aclocal_deps): Update and order as per aclocal.m4.
42372         * configure: Regenerate.
42373         * config.in: Regenerate.
42375 2017-01-04  Jakub Jelinek  <jakub@redhat.com>
42377         PR bootstrap/77569
42378         * input.c (ebcdic_execution_charset::on_error): Don't use strstr for
42379         a substring of the message, but strcmp with the whole message.  Ifdef
42380         ENABLE_NLS, translate the message first using dgettext.
42382 2017-01-03  Jeff Law  <law@redhat.com>
42384         PR tree-optimizatin/78856
42385         * tree-ssa-threadupdate.c: Include tree-vectorizer.h.
42386         (mark_threaded_blocks): Remove code to truncate thread paths that
42387         cross multiple loop headers.  Instead invalidate the cached loop
42388         iteration information and handle case of a thread path walking
42389         into an irreducible region.
42391 2017-01-03  Michael Meissner  <meissner@linux.vnet.ibm.com>
42393         PR target/78900
42394         * config/rs6000/rs6000.c (rs6000_split_signbit): Change some
42395         assertions.  Add support for doing the signbit if the IEEE 128-bit
42396         floating point value is in a GPR.
42397         * config/rs6000/rs6000.md (Fsignbit): Delete.
42398         (signbit<mode>2_dm): Delete using <Fsignbit> and just use "wa".
42399         Update the length attribute if the value is in a GPR.
42400         (signbit<mode>2_dm_<su>ext): Add combiner pattern to eliminate
42401         the sign or zero extension instruction, since the value is always 0/1.
42402         (signbit<mode>2_dm2): Delete using <Fsignbit>.
42404         PR target/78953
42405         * config/rs6000/vsx.md (vsx_extract_<mode>_store_p9): If we are
42406         extracting SImode to a GPR register so that we can generate a
42407         store, limit the vector to be in a traditional Altivec register
42408         for the vextuwrx instruction.
42410 2017-01-03  Ian Lance Taylor  <iant@google.com>
42412         * godump.c (go_format_type): Treat ENUMERAL_TYPE like INTEGER_TYPE.
42414 2017-01-03  Martin Sebor  <msebor@redhat.com>
42416         PR tree-optimization/78696
42417         * gimple-ssa-sprintf.c (format_floating): Correct handling of
42418         precision.  Use MPFR for %f for greater fidelity.  Correct handling
42419         of %g.
42420         (pass_sprintf_length::compute_format_length): Set width and precision
42421         specified by asrerisk to void_node for vararg functions.
42422         (try_substitute_return_value): Adjust dump output.
42424 2017-01-03  David Edelsohn  <dje.gcc@gmail.com>
42426         * doc/invoke.texi (RS6000 options): LRA is enabled by default.
42428 2017-01-03  Eric Botcazou  <ebotcazou@adacore.com>
42430         * doc/invoke.texi (SPARC options): Document -mlra as the default.
42431         * config/sparc/sparc.c (sparc_option_override): Force LRA unless
42432         -mlra/-mno-lra was passed to the compiler.
42434 2017-01-03  James Cowgill  <James.Cowgill@imgtec.com>
42436         PR rtl-optimization/65618
42437         * emit-rtl.c (try_split): Move initialization of "before" and
42438         "after" to just before the call to emit_insn_after_setloc.
42440 2017-01-03  Gerald Pfeifer  <gerald@pfeifer.com>
42442         * doc/md.texi (Standard Names): Remove reference to Java frontend.
42444 2017-01-03  Pierre-Marie de Rodat  <derodat@adacore.com>
42446         * dwarf2out.c (gen_enumeration_type_die): When
42447         -gno-strict-dwarf, add a DW_AT_encoding attribute.
42449 2017-01-03  Jakub Jelinek  <jakub@redhat.com>
42451         PR tree-optimization/78965
42452         * gimple-ssa-sprintf.c (pass_sprintf_length::compute_format_length):
42453         Change first argument from const call_info & to call_info &.  For %n
42454         set info.nowrite to false.
42456         PR middle-end/78901
42457         * gimple-ssa-sprintf.c (try_substitute_return_value): Don't change
42458         possibly throwing calls.
42460         * genmatch.c (dt_node::gen_kids_1): If generic_exprs include SSA_NAME
42461         and exprs_len || fns_len, emit the code for SSA_NAME next to the exprs
42462         and fns handling, rather than in a separate case SSA_NAME.
42464 2017-01-02  Jeff Law  <law@redhat.com>
42466         * config/darwin-driver.c (darwin_driver_init): Const-correctness
42467         fixes for first_period and second_period variables.
42469 2017-01-02  Uros Bizjak  <ubizjak@gmail.com>
42471         PR target/78967
42472         * config/i386/i386.md (UNSPEC_NOREX_MEM): New unspec.
42473         (*insvqi_1): New insn pattern.
42474         (*insvqi_1_mem_rex64): Ditto.
42475         (*insvqi_2): Ditto.
42476         (*insvqi_3): Rename from *insvqi.
42478         (*extzvqi_mem_rex64): Add UNSPEC_NOREX_MEM tag.
42480 2017-01-02  Gerald Pfeifer  <gerald@pfeifer.com>
42482         * doc/cfg.texi (Edges): Remove reference to Java.
42483         (Maintaining the CFG): Ditto.
42485 2017-01-01  Jan Hubicka  <hubicka@ucw.cz>
42487         PR middle-end/77674
42488         * symtab.c (symtab_node::binds_to_current_def_p): Fix handling of
42489         transparent aliases.
42491 2017-01-01  Jan Hubicka  <hubicka@ucw.cz>
42493         PR middle-end/77484
42494         * predict.def (PRED_CALL): Update hitrate.
42495         (PRED_INDIR_CALL, PRED_POLYMORPHIC_CALL): New predictors.
42496         * predict.c (tree_estimate_probability_bb): Split CALL predictor
42497         into direct/indirect/polymorphic variants.
42499 2017-01-01  Jakub Jelinek  <jakub@redhat.com>
42501         Update copyright years.
42503         * gcc.c (process_command): Update copyright notice dates.
42504         * gcov-dump.c (print_version): Ditto.
42505         * gcov.c (print_version): Ditto.
42506         * gcov-tool.c (print_version): Ditto.
42507         * gengtype.c (create_file): Ditto.
42508         * doc/cpp.texi: Bump @copying's copyright year.
42509         * doc/cppinternals.texi: Ditto.
42510         * doc/gcc.texi: Ditto.
42511         * doc/gccint.texi: Ditto.
42512         * doc/gcov.texi: Ditto.
42513         * doc/install.texi: Ditto.
42514         * doc/invoke.texi: Ditto.
42516 Copyright (C) 2017 Free Software Foundation, Inc.
42518 Copying and distribution of this file, with or without modification,
42519 are permitted in any medium without royalty provided the copyright
42520 notice and this notice are preserved.