poly_int: vectorizable_reduction
[official-gcc.git] / gcc / ChangeLog
blobeda66257b4708fdece1ffb39e5c9345204be1410
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.h (build_index_vector): Declare.
6         * tree.c (build_index_vector): New function.
7         * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
8         of units as polynomial, forcibly converting it to a constant if
9         vectorizable_reduction has already enforced the condition.
10         (vect_create_epilog_for_reduction): Likewise.  Use build_index_vector
11         to create a {1,2,3,...} vector.
12         (vectorizable_reduction): Treat the number of units as polynomial.
13         Choose vectype_in based on the largest scalar element size rather
14         than the smallest number of units.  Enforce the restrictions
15         relied on above.
17 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
18             Alan Hayward  <alan.hayward@arm.com>
19             David Sherwood  <david.sherwood@arm.com>
21         * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
22         number of units as polynomial.
24 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
25             Alan Hayward  <alan.hayward@arm.com>
26             David Sherwood  <david.sherwood@arm.com>
28         * target.h (vector_sizes, auto_vector_sizes): New typedefs.
29         * target.def (autovectorize_vector_sizes): Return the vector sizes
30         by pointer, using vector_sizes rather than a bitmask.
31         * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
32         * targhooks.c (default_autovectorize_vector_sizes): Likewise.
33         * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
34         Likewise.
35         * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
36         * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
37         * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
38         * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
39         * omp-general.c (omp_max_vf): Likewise.
40         * omp-low.c (omp_clause_aligned_alignment): Likewise.
41         * optabs-query.c (can_vec_mask_load_store_p): Likewise.
42         * tree-vect-loop.c (vect_analyze_loop): Likewise.
43         * tree-vect-slp.c (vect_slp_bb): Likewise.
44         * doc/tm.texi: Regenerate.
45         * tree-vectorizer.h (current_vector_size): Change from an unsigned int
46         to a poly_uint64.
47         * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
48         the vector size as a poly_uint64 rather than an unsigned int.
49         (current_vector_size): Change from an unsigned int to a poly_uint64.
50         (get_vectype_for_scalar_type): Update accordingly.
51         * tree.h (build_truth_vector_type): Take the size and number of
52         units as a poly_uint64 rather than an unsigned int.
53         (build_vector_type): Add a temporary overload that takes
54         the number of units as a poly_uint64 rather than an unsigned int.
55         * tree.c (make_vector_type): Likewise.
56         (build_truth_vector_type): Take the number of units as a poly_uint64
57         rather than an unsigned int.
59 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
60             Alan Hayward  <alan.hayward@arm.com>
61             David Sherwood  <david.sherwood@arm.com>
63         * target.def (get_mask_mode): Take the number of units and length
64         as poly_uint64s rather than unsigned ints.
65         * targhooks.h (default_get_mask_mode): Update accordingly.
66         * targhooks.c (default_get_mask_mode): Likewise.
67         * config/i386/i386.c (ix86_get_mask_mode): Likewise.
68         * doc/tm.texi: Regenerate.
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         * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
75         * omp-general.c (omp_max_vf): Likewise.
76         * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
77         (expand_omp_simd): Handle polynomial safelen.
78         * omp-low.c (omplow_simd_context): Add a default constructor.
79         (omplow_simd_context::max_vf): Change from int to poly_uint64.
80         (lower_rec_simd_input_clauses): Update accordingly.
81         (lower_rec_input_clauses): Likewise.
83 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
84             Alan Hayward  <alan.hayward@arm.com>
85             David Sherwood  <david.sherwood@arm.com>
87         * tree-vectorizer.h (vect_nunits_for_cost): New function.
88         * tree-vect-loop.c (vect_model_reduction_cost): Use it.
89         * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
90         (vect_analyze_slp_cost): Likewise.
91         * tree-vect-stmts.c (vect_model_store_cost): Likewise.
92         (vect_model_load_cost): Likewise.
94 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
95             Alan Hayward  <alan.hayward@arm.com>
96             David Sherwood  <david.sherwood@arm.com>
98         * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
99         (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
100         from an unsigned int * to a poly_uint64_pod *.
101         (calculate_unrolling_factor): New function.
102         (vect_analyze_slp_instance): Use it.  Track polynomial max_nunits.
104 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
105             Alan Hayward  <alan.hayward@arm.com>
106             David Sherwood  <david.sherwood@arm.com>
108         * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
109         from an unsigned int to a poly_uint64.
110         (_loop_vec_info::slp_unrolling_factor): Likewise.
111         (_loop_vec_info::vectorization_factor): Change from an int
112         to a poly_uint64.
113         (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
114         (vect_get_num_vectors): New function.
115         (vect_update_max_nunits, vect_vf_for_cost): Likewise.
116         (vect_get_num_copies): Use vect_get_num_vectors.
117         (vect_analyze_data_ref_dependences): Change max_vf from an int *
118         to an unsigned int *.
119         (vect_analyze_data_refs): Change min_vf from an int * to a
120         poly_uint64 *.
121         (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
122         than an unsigned HOST_WIDE_INT.
123         * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
124         (vect_analyze_data_ref_dependence): Change max_vf from an int *
125         to an unsigned int *.
126         (vect_analyze_data_ref_dependences): Likewise.
127         (vect_compute_data_ref_alignment): Handle polynomial vf.
128         (vect_enhance_data_refs_alignment): Likewise.
129         (vect_prune_runtime_alias_test_list): Likewise.
130         (vect_shift_permute_load_chain): Likewise.
131         (vect_supportable_dr_alignment): Likewise.
132         (dependence_distance_ge_vf): Take the vectorization factor as a
133         poly_uint64 rather than an unsigned HOST_WIDE_INT.
134         (vect_analyze_data_refs): Change min_vf from an int * to a
135         poly_uint64 *.
136         * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
137         vfm1 as a poly_uint64 rather than an int.  Make the same change
138         for the returned bound_scalar.
139         (vect_gen_vector_loop_niters): Handle polynomial vf.
140         (vect_do_peeling): Likewise.  Update call to
141         vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
142         (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
143         be constant.
144         * tree-vect-loop.c (vect_determine_vectorization_factor)
145         (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
146         (vect_get_known_peeling_cost): Likewise.
147         (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
148         (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
149         (vect_transform_loop): Likewise.  Use the lowest possible VF when
150         updating the upper bounds of the loop.
151         (vect_min_worthwhile_factor): Make static.  Return an unsigned int
152         rather than an int.
153         * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
154         polynomial unroll factors.
155         (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
156         (vect_make_slp_decision): Likewise.
157         (vect_supported_load_permutation_p): Likewise, and polynomial
158         vf too.
159         (vect_analyze_slp_cost): Handle polynomial vf.
160         (vect_slp_analyze_node_operations): Likewise.
161         (vect_slp_analyze_bb_1): Likewise.
162         (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
163         than an unsigned HOST_WIDE_INT.
164         * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
165         (vectorizable_load): Handle polynomial vf.
166         * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
167         a poly_uint64.
168         (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
170 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
171             Alan Hayward  <alan.hayward@arm.com>
172             David Sherwood  <david.sherwood@arm.com>
174         * match.pd: Handle bit operations involving three constants
175         and try to fold one pair.
177 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
179         * tree-vect-loop-manip.c: Include gimple-fold.h.
180         (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
181         niters_maybe_zero parameters.  Handle other cases besides a step of 1.
182         (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
183         Add a path that uses a step of VF instead of 1, but disable it
184         for now.
185         (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
186         and niters_no_overflow parameters.  Update calls to
187         slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
188         Create a new SSA name if the latter choses to use a ste other
189         than zero, and return it via niters_vector_mult_vf_var.
190         * tree-vect-loop.c (vect_transform_loop): Update calls to
191         vect_do_peeling, vect_gen_vector_loop_niters and
192         slpeel_make_loop_iterate_ntimes.
193         * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
194         (vect_gen_vector_loop_niters): Update declarations after above changes.
196 2018-01-02  Michael Meissner  <meissner@linux.vnet.ibm.com>
198         * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
199         128-bit round to integer instructions.
200         (ceil<mode>2): Likewise.
201         (btrunc<mode>2): Likewise.
202         (round<mode>2): Likewise.
204 2018-01-02  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
206         * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
207         unaligned VSX load/store on P8/P9.
208         (expand_block_clear): Allow the use of unaligned VSX
209         load/store on P8/P9.
211 2018-01-02  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
213         * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
214         New function.
215         (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
216         swap associated with both a load and a store.
218 2018-01-02  Andrew Waterman  <andrew@sifive.com>
220         * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
221         * config/riscv/riscv.md (clear_cache): Use it.
223 2018-01-02  Artyom Skrobov  <tyomitch@gmail.com>
225         * web.c: Remove out-of-date comment.
227 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
229         * expr.c (fixup_args_size_notes): Check that any existing
230         REG_ARGS_SIZE notes are correct, and don't try to re-add them.
231         (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
232         (emit_single_push_insn): ...here.
234 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
236         * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
237         (const_vector_encoded_nelts): New function.
238         (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
239         (const_vector_int_elt, const_vector_elt): Declare.
240         * emit-rtl.c (const_vector_int_elt_1): New function.
241         (const_vector_elt): Likewise.
242         * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
243         of CONST_VECTOR_ELT.
245 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
247         * expr.c: Include rtx-vector-builder.h.
248         (const_vector_mask_from_tree): Use rtx_vector_builder and operate
249         directly on the tree encoding.
250         (const_vector_from_tree): Likewise.
251         * optabs.c: Include rtx-vector-builder.h.
252         (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
253         sequence of "u" values.
254         * vec-perm-indices.c: Include rtx-vector-builder.h.
255         (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
256         directly on the vec_perm_indices encoding.
258 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
260         * doc/rtl.texi (const_vector): Describe new encoding scheme.
261         * Makefile.in (OBJS): Add rtx-vector-builder.o.
262         * rtx-vector-builder.h: New file.
263         * rtx-vector-builder.c: Likewise.
264         * rtl.h (rtx_def::u2): Add a const_vector field.
265         (CONST_VECTOR_NPATTERNS): New macro.
266         (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
267         (CONST_VECTOR_DUPLICATE_P): Likewise.
268         (CONST_VECTOR_STEPPED_P): Likewise.
269         (CONST_VECTOR_ENCODED_ELT): Likewise.
270         (const_vec_duplicate_p): Check for a duplicated vector encoding.
271         (unwrap_const_vec_duplicate): Likewise.
272         (const_vec_series_p): Check for a non-duplicated vector encoding.
273         Say that the function only returns true for integer vectors.
274         * emit-rtl.c: Include rtx-vector-builder.h.
275         (gen_const_vec_duplicate_1): Delete.
276         (gen_const_vector): Call gen_const_vec_duplicate instead of
277         gen_const_vec_duplicate_1.
278         (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
279         (gen_const_vec_duplicate): Use rtx_vector_builder.
280         (gen_const_vec_series): Likewise.
281         (gen_rtx_CONST_VECTOR): Likewise.
282         * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
283         (swap_const_vector_halves): Take an rtx pointer rather than rtx.
284         Build a new vector rather than modifying a CONST_VECTOR in-place.
285         (handle_special_swappables): Update call accordingly.
286         * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
287         (swap_const_vector_halves): Take an rtx pointer rather than rtx.
288         Build a new vector rather than modifying a CONST_VECTOR in-place.
289         (handle_special_swappables): Update call accordingly.
291 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
293         * simplify-rtx.c (simplify_const_binary_operation): Use
294         CONST_VECTOR_ELT instead of XVECEXP.
296 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
298         * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
299         the selector elements to be different from the data elements
300         if the selector is a VECTOR_CST.
301         * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
302         ssizetype for the selector.
304 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
306         * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
307         before testing each element individually.
308         * tree-vect-generic.c (lower_vec_perm): Likewise.
310 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
312         * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
313         * selftest-run-tests.c (selftest::run_tests): Call it.
314         * vector-builder.h (vector_builder::operator ==): New function.
315         (vector_builder::operator !=): Likewise.
316         * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
317         (vec_perm_indices::all_from_input_p): New function.
318         * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
319         (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
320         * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
321         instead of reading the VECTOR_CST directly.  Detect whether both
322         vector inputs are the same before constructing the vec_perm_indices,
323         and update the number of inputs argument accordingly.  Use the
324         utility functions added above.  Only construct sel2 if we need to.
326 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
328         * optabs.c (expand_vec_perm_var): Use an explicit encoding for
329         the broadcast of the low byte.
330         (expand_mult_highpart): Use an explicit encoding for the permutes.
331         * optabs-query.c (can_mult_highpart_p): Likewise.
332         * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
333         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
334         (vectorizable_bswap): Likewise.
335         * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
336         explicit encoding for the power-of-2 permutes.
337         (vect_permute_store_chain): Likewise.
338         (vect_grouped_load_supported): Likewise.
339         (vect_permute_load_chain): Likewise.
341 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
343         * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
344         * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
345         * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
346         * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
347         * tree-vect-stmts.c (vectorizable_bswap): Likewise.
348         (vect_gen_perm_mask_any): Likewise.
350 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
352         * int-vector-builder.h: New file.
353         * vec-perm-indices.h: Include int-vector-builder.h.
354         (vec_perm_indices): Redefine as an int_vector_builder.
355         (auto_vec_perm_indices): Delete.
356         (vec_perm_builder): Redefine as a stand-alone class.
357         (vec_perm_indices::vec_perm_indices): New function.
358         (vec_perm_indices::clamp): Likewise.
359         * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
360         (vec_perm_indices::new_vector): New function.
361         (vec_perm_indices::new_expanded_vector): Update for new
362         vec_perm_indices class.
363         (vec_perm_indices::rotate_inputs): New function.
364         (vec_perm_indices::all_in_range_p): Operate directly on the
365         encoded form, without computing elided elements.
366         (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
367         encoding.  Update for new vec_perm_indices class.
368         * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
369         the given vec_perm_builder.
370         (expand_vec_perm_var): Update vec_perm_builder constructor.
371         (expand_mult_highpart): Use vec_perm_builder instead of
372         auto_vec_perm_indices.
373         * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
374         vec_perm_indices instead of auto_vec_perm_indices.  Use a single
375         or double series encoding as appropriate.
376         * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
377         vec_perm_indices instead of auto_vec_perm_indices.
378         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
379         * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
380         (vect_permute_store_chain): Likewise.
381         (vect_grouped_load_supported): Likewise.
382         (vect_permute_load_chain): Likewise.
383         (vect_shift_permute_load_chain): Likewise.
384         * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
385         (vect_transform_slp_perm_load): Likewise.
386         (vect_schedule_slp_instance): Likewise.
387         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
388         (vectorizable_mask_load_store): Likewise.
389         (vectorizable_bswap): Likewise.
390         (vectorizable_store): Likewise.
391         (vectorizable_load): Likewise.
392         * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
393         vec_perm_indices instead of auto_vec_perm_indices.  Use
394         tree_to_vec_perm_builder to read the vector from a tree.
395         * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
396         vec_perm_builder instead of a vec_perm_indices.
397         (have_whole_vector_shift): Use vec_perm_builder and
398         vec_perm_indices instead of auto_vec_perm_indices.  Leave the
399         truncation to calc_vec_perm_mask_for_shift.
400         (vect_create_epilog_for_reduction): Likewise.
401         * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
402         from auto_vec_perm_indices to vec_perm_indices.
403         (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
404         instead of changing individual elements.
405         (aarch64_vectorize_vec_perm_const): Use new_vector to install
406         the vector in d.perm.
407         * config/arm/arm.c (expand_vec_perm_d::perm): Change
408         from auto_vec_perm_indices to vec_perm_indices.
409         (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
410         instead of changing individual elements.
411         (arm_vectorize_vec_perm_const): Use new_vector to install
412         the vector in d.perm.
413         * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
414         Update vec_perm_builder constructor.
415         (rs6000_expand_interleave): Likewise.
416         * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
417         (rs6000_expand_interleave): Likewise.
419 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
421         * optabs-query.c (can_vec_perm_var_p): Check whether lowering
422         to qimode could truncate the indices.
423         * optabs.c (expand_vec_perm_var): Likewise.
425 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
427         * Makefile.in (OBJS): Add vec-perm-indices.o.
428         * vec-perm-indices.h: New file.
429         * vec-perm-indices.c: Likewise.
430         * target.h (vec_perm_indices): Replace with a forward class
431         declaration.
432         (auto_vec_perm_indices): Move to vec-perm-indices.h.
433         * optabs.h: Include vec-perm-indices.h.
434         (expand_vec_perm): Delete.
435         (selector_fits_mode_p, expand_vec_perm_var): Declare.
436         (expand_vec_perm_const): Declare.
437         * target.def (vec_perm_const_ok): Replace with...
438         (vec_perm_const): ...this new hook.
439         * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
440         (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
441         * doc/tm.texi: Regenerate.
442         * optabs.def (vec_perm_const): Delete.
443         * doc/md.texi (vec_perm_const): Likewise.
444         (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
445         * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
446         expand_vec_perm for constant permutation vectors.  Assert that
447         the mode of variable permutation vectors is the integer equivalent
448         of the mode that is being permuted.
449         * optabs-query.h (selector_fits_mode_p): Declare.
450         * optabs-query.c: Include vec-perm-indices.h.
451         (selector_fits_mode_p): New function.
452         (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
453         is defined, instead of checking whether the vec_perm_const_optab
454         exists.  Use targetm.vectorize.vec_perm_const instead of
455         targetm.vectorize.vec_perm_const_ok.  Check whether the indices
456         fit in the vector mode before using a variable permute.
457         * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
458         vec_perm_indices instead of an rtx.
459         (expand_vec_perm): Replace with...
460         (expand_vec_perm_const): ...this new function.  Take the selector
461         as a vec_perm_indices rather than an rtx.  Also take the mode of
462         the selector.  Update call to shift_amt_for_vec_perm_mask.
463         Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
464         Use vec_perm_indices::new_expanded_vector to expand the original
465         selector into bytes.  Check whether the indices fit in the vector
466         mode before using a variable permute.
467         (expand_vec_perm_var): Make global.
468         (expand_mult_highpart): Use expand_vec_perm_const.
469         * fold-const.c: Includes vec-perm-indices.h.
470         * tree-ssa-forwprop.c: Likewise.
471         * tree-vect-data-refs.c: Likewise.
472         * tree-vect-generic.c: Likewise.
473         * tree-vect-loop.c: Likewise.
474         * tree-vect-slp.c: Likewise.
475         * tree-vect-stmts.c: Likewise.
476         * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
477         Delete.
478         * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
479         * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
480         (aarch64_vectorize_vec_perm_const_ok): Fuse into...
481         (aarch64_vectorize_vec_perm_const): ...this new function.
482         (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
483         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
484         * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
485         * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
486         * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
487         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
488         (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
489         into...
490         (arm_vectorize_vec_perm_const): ...this new function.  Explicitly
491         check for NEON modes.
492         * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
493         * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
494         * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
495         (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
496         into...
497         (ix86_vectorize_vec_perm_const): ...this new function.  Incorporate
498         the old VEC_PERM_CONST conditions.
499         * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
500         * config/ia64/vect.md (vec_perm_const<mode>): Delete.
501         * config/ia64/ia64.c (ia64_expand_vec_perm_const)
502         (ia64_vectorize_vec_perm_const_ok): Merge into...
503         (ia64_vectorize_vec_perm_const): ...this new function.
504         * config/mips/loongson.md (vec_perm_const<mode>): Delete.
505         * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
506         * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
507         * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
508         * config/mips/mips.c (mips_expand_vec_perm_const)
509         (mips_vectorize_vec_perm_const_ok): Merge into...
510         (mips_vectorize_vec_perm_const): ...this new function.
511         * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
512         * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
513         * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
514         * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
515         * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
516         (rs6000_expand_vec_perm_const): Delete.
517         * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
518         Delete.
519         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
520         (altivec_expand_vec_perm_const_le): Take each operand individually.
521         Operate on constant selectors rather than rtxes.
522         (altivec_expand_vec_perm_const): Likewise.  Update call to
523         altivec_expand_vec_perm_const_le.
524         (rs6000_expand_vec_perm_const): Delete.
525         (rs6000_vectorize_vec_perm_const_ok): Delete.
526         (rs6000_vectorize_vec_perm_const): New function.
527         (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
528         an element count and rtx array.
529         (rs6000_expand_extract_even): Update call accordingly.
530         (rs6000_expand_interleave): Likewise.
531         * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
532         * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
533         * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
534         * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
535         (rs6000_expand_vec_perm_const): Delete.
536         * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
537         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
538         (altivec_expand_vec_perm_const_le): Take each operand individually.
539         Operate on constant selectors rather than rtxes.
540         (altivec_expand_vec_perm_const): Likewise.  Update call to
541         altivec_expand_vec_perm_const_le.
542         (rs6000_expand_vec_perm_const): Delete.
543         (rs6000_vectorize_vec_perm_const_ok): Delete.
544         (rs6000_vectorize_vec_perm_const): New function.  Remove stray
545         reference to the SPE evmerge intructions.
546         (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
547         an element count and rtx array.
548         (rs6000_expand_extract_even): Update call accordingly.
549         (rs6000_expand_interleave): Likewise.
550         * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
551         * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
552         new function.
553         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
555 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
557         * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
558         vector mode and that that mode matches the mode of the data
559         being permuted.
560         (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
561         out into expand_vec_perm_var.  Do all CONST_VECTOR handling here,
562         directly using expand_vec_perm_1 when forcing selectors into
563         registers.
564         (expand_vec_perm_var): New function, split out from expand_vec_perm.
566 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
568         * optabs-query.h (can_vec_perm_p): Delete.
569         (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
570         * optabs-query.c (can_vec_perm_p): Split into...
571         (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
572         (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
573         particular selector is valid.
574         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
575         * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
576         (vect_grouped_load_supported): Likewise.
577         (vect_shift_permute_load_chain): Likewise.
578         * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
579         (vect_transform_slp_perm_load): Likewise.
580         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
581         (vectorizable_bswap): Likewise.
582         (vect_gen_perm_mask_checked): Likewise.
583         * fold-const.c (fold_ternary_loc): Likewise.  Don't take
584         implementations of variable permutation vectors into account
585         when deciding which selector to use.
586         * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
587         vec_perm_const_optab is supported; instead use can_vec_perm_const_p
588         with a false third argument.
589         * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
590         to test whether the constant selector is valid and can_vec_perm_var_p
591         to test whether a variable selector is valid.
593 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
595         * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
596         * optabs-query.c (can_vec_perm_p): Likewise.
597         * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
598         instead of vec_perm_indices.
599         * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
600         (vect_gen_perm_mask_checked): Likewise,
601         * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
602         (vect_gen_perm_mask_checked): Likewise,
604 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
606         * optabs-query.h (qimode_for_vec_perm): Declare.
607         * optabs-query.c (can_vec_perm_p): Split out qimode search to...
608         (qimode_for_vec_perm): ...this new function.
609         * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
611 2018-01-02  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
613         * rtlanal.c (canonicalize_condition): Return 0 if final rtx
614         does not have a conditional at the top.
616 2018-01-02  Richard Biener  <rguenther@suse.de>
618         * ipa-inline.c (big_speedup_p): Fix expression.
620 2018-01-02  Jan Hubicka  <hubicka@ucw.cz>
622         PR target/81616
623         * config/i386/x86-tune-costs.h: Increase cost of integer load costs
624         for generic 4->6.
626 2018-01-02  Jan Hubicka  <hubicka@ucw.cz>
628         PR target/81616
629         Generic tuning.
630         * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
631         cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
632         and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
633         cond_taken_branch_cost 3->4.
635 2018-01-01  Jakub Jelinek  <jakub@redhat.com>
637         PR tree-optimization/83581
638         * tree-loop-distribution.c (pass_loop_distribution::execute): Return
639         TODO_cleanup_cfg if any changes have been made.
641         PR middle-end/83608
642         * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
643         convert_modes if target mode has the right side, but different mode
644         class.
646         PR middle-end/83609
647         * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
648         last argument when extracting from CONCAT.  If either from_real or
649         from_imag is NULL, use expansion through memory.  If result is not
650         a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
651         the parts directly to inner mode, if even that fails, use expansion
652         through memory.
654         PR middle-end/83623
655         * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
656         check for bswap in mode rather than HImode and use that in expand_unop
657         too.
659 2017-12-31  Jakub Jelinek  <jakub@redhat.com>
661         PR target/83536
662         * config/i386/i386.c (ix86_attribute_table): Remove excess
663         initializer for "shared" attribute.
665 2017-12-30  Tom de Vries  <tom@codesourcery.com>
667         PR libgomp/83046
668         * omp-expand.c (expand_omp_target): If in_lto_p, mark offload_funcs with
669         DECL_PRESERVE_P.
670         * lto-streamer-out.c (prune_offload_funcs): New function.  Remove
671         offload_funcs entries that no longer have a corresponding cgraph_node.
672         Mark the remaining ones as DECL_PRESERVE_P.
673         (output_lto): Call prune_offload_funcs.
675 2017-12-30  Jakub Jelinek  <jakub@redhat.com>
677         * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
678         vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>,
679         vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
680         vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
681         vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
682         vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1, vpdpbusd_<mode>,
683         vpdpbusd_<mode>_mask, vpdpbusd_<mode>_maskz, vpdpbusd_<mode>_maskz_1,
684         vpdpbusds_<mode>, vpdpbusds_<mode>_mask, vpdpbusds_<mode>_maskz,
685         vpdpbusds_<mode>_maskz_1, vpdpwssd_<mode>, vpdpwssd_<mode>_mask,
686         vpdpwssd_<mode>_maskz, vpdpwssd_<mode>_maskz_1, vpdpwssds_<mode>,
687         vpdpwssds_<mode>_mask, vpdpwssds_<mode>_maskz,
688         vpdpwssds_<mode>_maskz_1, vaesdec_<mode>, vaesdeclast_<mode>,
689         vaesenc_<mode>, vpclmulqdq_<mode>,
690         avx512vl_vpshufbitqmb<mode><mask_scalar_merge_name>): Formatting fixes.
692 2017-12-28  Michael Meissner  <meissner@linux.vnet.ibm.com>
694         * builtins.def: (_Float<N> and _Float<N>X BUILT_IN_CEIL): Add
695         _Float<N> and _Float<N>X variants for rounding built-in
696         functions.
697         (_Float<N> and _Float<N>X BUILT_IN_FLOOR): Likewise.
698         (_Float<N> and _Float<N>X BUILT_IN_NEARBYINT): Likewise.
699         (_Float<N> and _Float<N>X BUILT_IN_RINT): Likewise.
700         (_Float<N> and _Float<N>X BUILT_IN_ROUND): Likewise.
701         (_Float<N> and _Float<N>X BUILT_IN_TRUNC): Likewise.
702         * builtins.c (mathfn_built_in_2): Likewise.
703         * internal-fn.def (CEIL): Likewise.
704         (FLOOR): Likewise.
705         (NEARBYINT): Likewise.
706         (RINT): Likewise.
707         (ROUND): Likewise.
708         (TRUNC): Likewise.
709         * convert.c (convert_to_integer_1): Likewise.
710         * fold-const.c (tree_call_nonnegative_warnv_p): Likewise.
711         (integer_valued_real_call_p): Likewise.
712         * fold-const-call.c (fold_const_call_ss): Likewise.
713         * gencfn-macros.c (print_case_cfn): Change CFN and operator
714         printers to take a const char * suffix instead of a bool.
715         (print_define_operator_list): Likewise.
716         (fltall_suffixes): New list of suffixes, that include the
717         traditional suffixes as well as all of the _Float<N> and
718         _Float<N>X suffixes.
719         (main): For _Float<N> and _Float<N>X functions, emit both
720         <name>_FN and <name>_ALL variants.  The <macro>_FN variant only
721         has the _Float<N> and _Float<N>X case names or operators.  The
722         <name>_ALL variant has both the traditional and the
723         _Float<N>/_Float<N>X case names or operators.
724         * match.pd (COPYSIGN optimizations): Provide optimizations for
725         _Float<N> and _Float<N>X types where possible.
726         (MIN/MAX optimizations): Likewise.
727         (sqrt optimizations): Likewise.
728         (rounding optimizations): Likewise.
730 2017-12-28  Richard Sandiford  <richard.sandiford@linaro.org>
732         * config/powerpcspe/powerpcspe.md (ieee_128bit_negative_zero): Use
733         gen_int_mode rather than GEN_INT.
734         * config/rs6000/rs6000.md (ieee_128bit_negative_zero): Likewise.
736 2017-12-28  Richard Sandiford  <richard.sandiford@linaro.org>
738         * emit-rtl.c (gen_const_vec_series): Use valid_for_const_vector_p
739         instead of CONSTANT_P.
740         (gen_vec_series): Likewise.
741         * simplify-rtx.c (simplify_binary_operation_1): Likewise.
743 2017-12-28  Andreas Schwab  <schwab@linux-m68k.org>
745         * config/m68k/m68k.md (ashrdi3_const1, lshrdi3_const1): Add
746         CC_STATUS_INIT.
748 2017-12-27  Kugan Vivekanandarajah  <kuganv@linaro.org>
750         * config/aarch64/aarch64-simd.md (aarch64_ld1x2<VQ:mode>): New.
751         (aarch64_ld1x2<VDC:mode>): Likewise.
752         (aarch64_simd_ld1<mode>_x2): Likewise.
753         (aarch64_simd_ld1<mode>_x2): Likewise.
754         * config/aarch64/arm_neon.h (vld1_u8_x2): New.
755         (vld1_s8_x2): Likewise.
756         (vld1_u16_x2): Likewise.
757         (vld1_s16_x2): Likewise.
758         (vld1_u32_x2): Likewise.
759         (vld1_s32_x2): Likewise.
760         (vld1_u64_x2): Likewise.
761         (vld1_s64_x2): Likewise.
762         (vld1_f16_x2): Likewise.
763         (vld1_f32_x2): Likewise.
764         (vld1_f64_x2): Likewise.
765         (vld1_p8_x2): Likewise.
766         (vld1_p16_x2): Likewise.
767         (vld1_p64_x2): Likewise.
768         (vld1q_u8_x2): Likewise.
769         (vld1q_s8_x2): Likewise.
770         (vld1q_u16_x2): Likewise.
771         (vld1q_s16_x2): Likewise.
772         (vld1q_u32_x2): Likewise.
773         (vld1q_s32_x2): Likewise.
774         (vld1q_u64_x2): Likewise.
775         (vld1q_s64_x2): Likewise.
776         (vld1q_f16_x2): Likewise.
777         (vld1q_f32_x2): Likewise.
778         (vld1q_f64_x2): Likewise.
779         (vld1q_p8_x2): Likewise.
780         (vld1q_p16_x2): Likewise.
781         (vld1q_p64_x2): Likewise.
783 2017-12-27  Martin Liska  <mliska@suse.cz>
785         PR tree-optimization/83552
786         * tree-ssa-strlen.c (fold_strstr_to_strncmp): Assign result
787         of get_string_lenth to a SSA_NAME if not a GIMPLE value.
789 2017-12-27  Tom de Vries  <tom@codesourcery.com>
791         * config/nvptx/nvptx.c (nvptx_option_override): Disable
792         -gstatement-frontiers.
794 2017-12-26  Alexander Monakov  <amonakov@ispras.ru>
796         PR rtl-optimization/83513
797         * sel-sched.c (sel_rank_for_schedule): Order by non-zero usefulness
798         before priority comparison.
800 2017-12-25  Jakub Jelinek  <jakub@redhat.com>
802         PR target/83488
803         * config/i386/i386.opt (-mavx512vpopcntdq, -mmavx512bitalg): Move from
804         ix86_isa_flags2 to ix86_isa_flags.
805         * config/i386/i386-c.c (ix86_target_macros_internal): Test
806         OPTION_MASK_ISA_AVX512BITALG and OPTION_MASK_ISA_AVX512VPOPCNTDQ in
807         isa_flags rather than isa_flags2.
808         * config/i386/i386.c (ix86_target_string): Move -mavx512vpopcntdq
809         and -mavx512bitalg from isa2_opts to isa_opts.
810         (ix86_option_override_internal): Test OPTION_MASK_ISA_AVX512VPOPCNTDQ
811         in x_ix86_isa_flags_explicit rather than x_ix86_isa_flags2_explicit
812         and set it in x_ix86_isa_flags rather than x_ix86_isa_flags2.
813         Formatting fixes.
814         (def_builtin): Treat OPTION_MASK_ISA_AVX512BW or
815         OPTION_MASK_ISA_AVX512F ored with another option similarly to
816         OPTION_MASK_ISA_AVX512VL.  Even for OPTION_MASK_ISA_AVX512VL don't
817         clear it if mask is just OPTION_MASK_ISA_AVX512VL itself.
818         (ix86_expand_builtin): Don't handle OPTION_MASK_ISA_GFNI and
819         OPTION_MASK_ISA_VPCLMULQDQ specially, instead handle
820         OPTION_MASK_ISA_AVX512BW and OPTION_MASK_ISA_AVX512F that way.
821         * config/i386/i386-builtin.def: Move AVX512VPOPCNTDQ and AVX512BITALG
822         builtins from bdesc_args2 to bdesc_args section.
823         (__builtin_ia32_compressstoreuqi512_mask,
824         __builtin_ia32_compressstoreuhi512_mask,
825         __builtin_ia32_compressstoreuqi256_mask,
826         __builtin_ia32_expandloadqi512_mask,
827         __builtin_ia32_expandloadqi512_maskz,
828         __builtin_ia32_expandloadhi512_mask,
829         __builtin_ia32_expandloadhi512_maskz,
830         __builtin_ia32_compressqi512_mask, __builtin_ia32_compresshi512_mask,
831         __builtin_ia32_compressqi256_mask, __builtin_ia32_expandqi512_mask,
832         __builtin_ia32_expandqi512_maskz, __builtin_ia32_expandhi512_mask,
833         __builtin_ia32_expandhi512_maskz, __builtin_ia32_expandqi256_mask,
834         __builtin_ia32_expandqi256_maskz, __builtin_ia32_vpshrd_v32hi_mask,
835         __builtin_ia32_vpshld_v32hi_mask, __builtin_ia32_vpshrdv_v32hi_mask,
836         __builtin_ia32_vpshrdv_v32hi_maskz, __builtin_ia32_vpshldv_v32hi_mask,
837         __builtin_ia32_vpshldv_v32hi_maskz,
838         __builtin_ia32_vpopcountb_v64qi_mask,
839         __builtin_ia32_vpopcountw_v32hi_mask,
840         __builtin_ia32_vpshufbitqmb512_mask,
841         __builtin_ia32_vpshufbitqmb256_mask): Add
842         " | OPTION_MASK_ISA_AVX512BW".
843         (__builtin_ia32_expandloadqi256_mask,
844         __builtin_ia32_expandloadqi256_maskz,
845         __builtin_ia32_vpopcountb_v32qi_mask): Add
846         " | OPTION_MASK_ISA_AVX512VL | OPTION_MASK_ISA_AVX512BW".
847         (__builtin_ia32_expandloadhi256_mask,
848         __builtin_ia32_expandloadhi256_maskz,
849         __builtin_ia32_expandloadqi128_mask,
850         __builtin_ia32_expandloadqi128_maskz,
851         __builtin_ia32_expandloadhi128_mask,
852         __builtin_ia32_expandloadhi128_maskz,
853         __builtin_ia32_vpshrd_v16hi, __builtin_ia32_vpshrd_v16hi_mask,
854         __builtin_ia32_vpshrd_v8hi, __builtin_ia32_vpshrd_v8hi_mask,
855         __builtin_ia32_vpshrd_v8si, __builtin_ia32_vpshrd_v8si_mask,
856         __builtin_ia32_vpshrd_v4si, __builtin_ia32_vpshrd_v4si_mask,
857         __builtin_ia32_vpshrd_v4di, __builtin_ia32_vpshrd_v4di_mask,
858         __builtin_ia32_vpshrd_v2di, __builtin_ia32_vpshrd_v2di_mask,
859         __builtin_ia32_vpshld_v16hi, __builtin_ia32_vpshld_v16hi_mask,
860         __builtin_ia32_vpshld_v8hi, __builtin_ia32_vpshld_v8hi_mask,
861         __builtin_ia32_vpshld_v8si, __builtin_ia32_vpshld_v8si_mask,
862         __builtin_ia32_vpshld_v4si, __builtin_ia32_vpshld_v4si_mask,
863         __builtin_ia32_vpshld_v4di, __builtin_ia32_vpshld_v4di_mask,
864         __builtin_ia32_vpshld_v2di, __builtin_ia32_vpshld_v2di_mask,
865         __builtin_ia32_vpshrdv_v16hi, __builtin_ia32_vpshrdv_v16hi_mask,
866         __builtin_ia32_vpshrdv_v16hi_maskz, __builtin_ia32_vpshrdv_v8hi,
867         __builtin_ia32_vpshrdv_v8hi_mask, __builtin_ia32_vpshrdv_v8hi_maskz,
868         __builtin_ia32_vpshrdv_v8si, __builtin_ia32_vpshrdv_v8si_mask,
869         __builtin_ia32_vpshrdv_v8si_maskz, __builtin_ia32_vpshrdv_v4si,
870         __builtin_ia32_vpshrdv_v4si_mask, __builtin_ia32_vpshrdv_v4si_maskz,
871         __builtin_ia32_vpshrdv_v4di, __builtin_ia32_vpshrdv_v4di_mask,
872         __builtin_ia32_vpshrdv_v4di_maskz, __builtin_ia32_vpshrdv_v2di,
873         __builtin_ia32_vpshrdv_v2di_mask, __builtin_ia32_vpshrdv_v2di_maskz,
874         __builtin_ia32_vpshldv_v16hi, __builtin_ia32_vpshldv_v16hi_mask,
875         __builtin_ia32_vpshldv_v16hi_maskz, __builtin_ia32_vpshldv_v8hi,
876         __builtin_ia32_vpshldv_v8hi_mask, __builtin_ia32_vpshldv_v8hi_maskz,
877         __builtin_ia32_vpshldv_v8si, __builtin_ia32_vpshldv_v8si_mask,
878         __builtin_ia32_vpshldv_v8si_maskz, __builtin_ia32_vpshldv_v4si,
879         __builtin_ia32_vpshldv_v4si_mask, __builtin_ia32_vpshldv_v4si_maskz,
880         __builtin_ia32_vpshldv_v4di, __builtin_ia32_vpshldv_v4di_mask,
881         __builtin_ia32_vpshldv_v4di_maskz, __builtin_ia32_vpshldv_v2di,
882         __builtin_ia32_vpshldv_v2di_mask, __builtin_ia32_vpshldv_v2di_maskz,
883         __builtin_ia32_vpopcountb_v32qi, __builtin_ia32_vpopcountb_v16qi,
884         __builtin_ia32_vpopcountb_v16qi_mask, __builtin_ia32_vpopcountw_v16hi,
885         __builtin_ia32_vpopcountw_v16hi_mask, __builtin_ia32_vpopcountw_v8hi,
886         __builtin_ia32_vpopcountw_v8hi_mask): Add
887         " | OPTION_MASK_ISA_AVX512VL".
888         * config/i386/avx512vbmi2intrin.h (_mm512_shrdi_epi16,
889         _mm512_shrdi_epi32, _mm512_mask_shrdi_epi32, _mm512_maskz_shrdi_epi32,
890         _mm512_shrdi_epi64, _mm512_mask_shrdi_epi64, _mm512_maskz_shrdi_epi64,
891         _mm512_shldi_epi16, _mm512_shldi_epi32, _mm512_mask_shldi_epi32,
892         _mm512_maskz_shldi_epi32, _mm512_shldi_epi64, _mm512_mask_shldi_epi64,
893         _mm512_maskz_shldi_epi64, _mm512_shrdv_epi16, _mm512_shrdv_epi32,
894         _mm512_mask_shrdv_epi32, _mm512_maskz_shrdv_epi32, _mm512_shrdv_epi64,
895         _mm512_mask_shrdv_epi64, _mm512_maskz_shrdv_epi64, _mm512_shldv_epi16,
896         _mm512_shldv_epi32, _mm512_mask_shldv_epi32, _mm512_maskz_shldv_epi32,
897         _mm512_shldv_epi64, _mm512_mask_shldv_epi64,
898         _mm512_maskz_shldv_epi64): Don't require avx512bw for these intrinsics.
899         * config/i386/avx512bitalgintrin.h (_mm_bitshuffle_epi64_mask,
900         _mm_mask_bitshuffle_epi64_mask): Likewise.
901         * common/config/i386/i386-common.c
902         (OPTION_MASK_ISA_AVX512VPOPCNTDQ_SET,
903         OPTION_MASK_ISA_AVX512BITALG_SET): Or in OPTION_MASK_ISA_AVX512F_SET.
904         (OPTION_MASK_ISA_AVX512F_UNSET): Or in
905         OPTION_MASK_ISA_AVX512VPOPCNTDQ_UNSET and
906         OPTION_MASK_ISA_AVX512BITALG_UNSET.
907         (OPTION_MASK_ISA2_AVX512F_UNSET,
908         OPTION_MASK_ISA2_GENERAL_REGS_ONLY_UNSET): Define.
909         (ix86_handle_option): For -mno-general-regs-only, clear from
910         ix86_isa_flags2 OPTION_MASK_ISA2_GENERAL_REGS_ONLY_UNSET rather than
911         just OPTION_MASK_ISA_MPX.  For -mno-sse{,2,3,4,4.1,4.2,avx,avx2} and
912         -mno-ssse3 clear OPTION_MASK_ISA2_AVX512F_UNSET bits from
913         ix86_isa_flags2.  For -mno-avx512f likewise, instead of masking
914         individually listed ISAs.  For -m{,no-}avx512{vpopcntdq,bitalg} adjust
915         for moving from ix86_isa_flags2 to ix86_isa_flags.
917 2017-12-23  Jakub Jelinek  <jakub@redhat.com>
919         PR c++/83553
920         * fold-const.c (struct contains_label_data): New type.
921         (contains_label_1): Return non-NULL even for CASE_LABEL_EXPR, unless
922         inside of a SWITCH_BODY seen during the walk.
923         (contains_label_p): Use walk_tree instead of
924         walk_tree_without_duplicates, prepare data for contains_label_1 and
925         provide own pset.
927 2017-12-22  Jakub Jelinek  <jakub@redhat.com>
929         PR debug/83547
930         * tree-iterator.c (alloc_stmt_list): Start with cleared
931         TREE_SIDE_EFFECTS regardless whether a new STATEMENT_LIST is allocated
932         or old one reused.
934         PR target/83488
935         * config/i386/avx512vnniintrin.h: Don't check for __AVX512F__ nor
936         enable avx512f explicitly in #pragma GCC target.
937         * config/i386/i386-builtin.def (__builtin_ia32_vpdpbusd_v8si,
938         __builtin_ia32_vpdpbusd_v8si_mask, __builtin_ia32_vpdpbusd_v8si_maskz,
939         __builtin_ia32_vpdpbusd_v4si, __builtin_ia32_vpdpbusd_v4si_mask,
940         __builtin_ia32_vpdpbusd_v4si_maskz, __builtin_ia32_vpdpbusds_v8si,
941         __builtin_ia32_vpdpbusds_v8si_mask,
942         __builtin_ia32_vpdpbusds_v8si_maskz, __builtin_ia32_vpdpbusds_v4si,
943         __builtin_ia32_vpdpbusds_v4si_mask,
944         __builtin_ia32_vpdpbusds_v4si_maskz, __builtin_ia32_vpdpwssd_v8si,
945         __builtin_ia32_vpdpwssd_v8si_mask, __builtin_ia32_vpdpwssd_v8si_maskz,
946         __builtin_ia32_vpdpwssd_v4si, __builtin_ia32_vpdpwssd_v4si_mask,
947         __builtin_ia32_vpdpwssd_v4si_maskz, __builtin_ia32_vpdpwssds_v8si,
948         __builtin_ia32_vpdpwssds_v8si_mask,
949         __builtin_ia32_vpdpwssds_v8si_maskz, __builtin_ia32_vpdpwssds_v4si,
950         __builtin_ia32_vpdpwssds_v4si_mask,
951         __builtin_ia32_vpdpwssds_v4si_maskz): Use
952         OPTION_MASK_ISA_AVX512VNNI | OPTION_MASK_ISA_AVX512VL instead of
953         just OPTION_MASK_ISA_AVX512VNNI.
955 2017-12-22  Martin Jambor  <mjambor@suse.cz>
957         PR lto/82027
958         * lto-cgraph.c (output_cgraph_opt_summary_p): Also check former
959         clones.
961 2017-12-22  Julia Koval  <julia.koval@intel.com>
962             Sebastian Peryt  <sebastian.peryt@intel.com>
964         * common/config/i386/i386-common.c (OPTION_MASK_ISA_AVX512BITALG_SET,
965         OPTION_MASK_ISA_AVX512BITALG_UNSET): New.
966         (ix86_handle_option): Handle -mavx512bitalg, fix 4VNNIW formatting.
967         * config.gcc: Add avx512vpopcntdqvlintrin.h and avx512bitalgintrin.h.
968         * config/i386/avx512bitalgintrin.h (_mm512_popcnt_epi8, _mm512_popcnt_epi16,
969         _mm512_mask_popcnt_epi8, _mm512_maskz_popcnt_epi8, _mm512_mask_popcnt_epi16,
970         _mm512_maskz_popcnt_epi16, _mm512_bitshuffle_epi64_mask, _mm256_popcnt_epi8,
971         _mm512_mask_bitshuffle_epi64_mask, _mm256_mask_popcnt_epi8, _mm_popcnt_epi8,
972         _mm256_maskz_popcnt_epi8, _mm_bitshuffle_epi64_mask, _mm256_popcnt_epi16,
973         _mm_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
974         _mm256_mask_bitshuffle_epi64_mask, _mm_popcnt_epi16, _mm_maskz_popcnt_epi8,
975         _mm256_mask_popcnt_epi16, _mm256_maskz_popcnt_epi16, _mm_mask_popcnt_epi8,
976         _mm_mask_popcnt_epi16, _mm_maskz_popcnt_epi16): New intrinsics.
977         * config/i386/avx512vpopcntdqvlintrin.h (_mm_popcnt_epi32, _mm_popcnt_epi64,
978         _mm_mask_popcnt_epi32, _mm_maskz_popcnt_epi32, _mm256_popcnt_epi32,
979         _mm256_mask_popcnt_epi32, _mm256_maskz_popcnt_epi32, _mm_mask_popcnt_epi64,
980         _mm_maskz_popcnt_epi64, _mm256_popcnt_epi64, _mm256_mask_popcnt_epi64,
981         _mm256_maskz_popcnt_epi64): New intrinsics.
982         * config/i386/cpuid.h (bit_AVX512BITALG): New bit.
983         * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mavx512bitalg.
984         * config/i386/i386-builtin-types.def (V64QI_FTYPE_V64QI, V64QI_FTYPE_V64QI,
985         V4DI_FTYPE_V4DI, UHI_FTYPE_V2DI_V2DI_UHI, USI_FTYPE_V4DI_V4DI_USI,
986         V4SI_FTYPE_V4SI_V4SI_UHI, V8SI_FTYPE_V8SI_V8SI_UHI): New types.
987         * config/i386/i386-builtin.def (__builtin_ia32_vpopcountq_v4di,
988         __builtin_ia32_vpopcountq_v4di_mask, __builtin_ia32_vpopcountq_v2di,
989         __builtin_ia32_vpopcountq_v2di_mask, __builtin_ia32_vpopcountd_v4si,
990         __builtin_ia32_vpopcountd_v4si_mask, __builtin_ia32_vpopcountd_v8si,
991         __builtin_ia32_vpopcountd_v8si_mask, __builtin_ia32_vpopcountb_v64qi,
992         __builtin_ia32_vpopcountb_v64qi_mask, __builtin_ia32_vpopcountb_v32qi,
993         __builtin_ia32_vpopcountb_v32qi_mask, __builtin_ia32_vpopcountb_v16qi,
994         __builtin_ia32_vpopcountb_v16qi_mask, __builtin_ia32_vpopcountw_v32hi,
995         __builtin_ia32_vpopcountw_v32hi_mask, __builtin_ia32_vpopcountw_v16hi,
996         __builtin_ia32_vpopcountw_v16hi_mask, __builtin_ia32_vpopcountw_v8hi,
997         __builtin_ia32_vpopcountw_v8hi_mask, __builtin_ia32_vpshufbitqmb128_mask,
998         __builtin_ia32_vpshufbitqmb256_mask,
999         __builtin_ia32_vpshufbitqmb512_mask): New builtins.
1000         * config/i386/i386-c.c (__AVX512BITALG__): New.
1001         * config/i386/i386.c (isa2_opts): Add -mavx512bitalg.
1002         (ix86_valid_target_attribute_inner_p): Ditto.
1003         (ix86_expand_args_builtin): Handle new types.
1004         * config/i386/i386.h (TARGET_AVX512BITALG, TARGET_AVX512BITALG_P): New.
1005         * config/i386/i386.opt: Add -mavx512bitalg.
1006         * config/i386/immintrin.h: Add avx512vpopcntdqvlintrin.h and
1007         avx512bitalgintrin.h.
1008         * config/i386/sse.md (VI48_AVX512VLBW): New iterator.
1009         (vpopcount<mode><mask_name>): Add more types.
1010         (avx512vl_vpshufbitqmb<mode><mask_scalar_merge_name>): New.
1011         * doc/invoke.texi: Add -mavx512bitalg and -mavx512vpopcntdq.
1013 2017-12-22  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
1015         * common/config/i386/i386-common.c (OPTION_MASK_ISA_AVX512VNNI_SET):
1016         Or in OPTION_MASK_ISA_AVX512F_SET.
1017         (OPTION_MASK_ISA_AVX512F_UNSET): Or in
1018         OPTION_MASK_ISA_AVX512VNNI_UNSET.
1019         (ix86_handle_option): Adjust for
1020         OPTION_MASK_ISA_AVX512VNNI_*SET being in ix86_isa_flags.
1021         * config/i386/i386-builtin.def: Move VNNI builtins from ARGS2
1022         section to ARGS.
1023         * config/i386/i386-c.c: Check for OPTION_MASK_ISA_AVX512VNNI in
1024         isa_flag instead of isa_flag2.
1025         * config/i386/i386.c (ix86_target_string): Move -mavx512vnni from
1026         isa_opts2 to isa_opts.
1027         * config/i386/i386.opt (mavx512vnni): Move from ix86_isa_flags2
1028         to ix86_isa_flags.
1030 2017-12-22  Mike Stump  <mikestump@comcast.net>
1031             Eric Botcazou  <ebotcazou@adacore.com>
1033         * doc/extend.texi (Loop-Specific Pragmas): Document pragma GCC unroll.
1035 2017-12-21  Jakub Jelinek  <jakub@redhat.com>
1037         PR middle-end/83487
1038         * config/i386/i386.c (ix86_function_arg_boundary): Return
1039         PARM_BOUNDARY for TYPE_EMPTY_P types.
1041         PR c/83448
1042         * gimple-ssa-sprintf.c (maybe_warn): Don't call set_caret_index
1043         if navail is >= dir.len.
1045 2017-12-21  Steve Ellcey  <sellcey@cavium.com>
1047         * config/aarch64/t-aarch64-linux (MULTILIB_OSDIRNAMES): Fix
1048         triplet for ilp32.
1050 2017-12-21  Jakub Jelinek  <jakub@redhat.com>
1052         PR rtl-optimization/80747
1053         PR rtl-optimization/83512
1054         * cfgrtl.c (force_nonfallthru_and_redirect): When splitting
1055         succ edge from ENTRY, copy partition from e->dest to the newly
1056         created bb.
1057         * bb-reorder.c (reorder_basic_blocks_simple): If last_tail is
1058         ENTRY, use BB_PARTITION of its successor block as current_partition.
1059         Don't copy partition when splitting succ edge from ENTRY.
1061         PR tree-optimization/83523
1062         * tree-ssa-math-opts.c (is_widening_mult_p): Return false if
1063         for INTEGER_TYPE TYPE_OVERFLOW_TRAPS.
1064         (convert_mult_to_fma): Likewise.
1066         PR tree-optimization/83521
1067         * tree-ssa-phiopt.c (factor_out_conditional_conversion): Use
1068         gimple_build_assign without code on result of
1069         fold_build1 (VIEW_CONVERT_EXPR, ...), as it might not create
1070         a VIEW_CONVERT_EXPR.
1072 2017-12-21  Andrew Pinski  <apinski@cavium.com>
1073             Steve Ellcey  <sellcey@cavium.com>
1075         * config/aarch64/t-aarch64-linux (MULTILIB_OSDIRNAMES): Handle
1076         multi-arch for ilp32.
1078 2017-12-21  Uros Bizjak  <ubizjak@gmail.com>
1080         PR target/83467
1081         * config/i386/i386.md (*ashl<mode>3_mask): Add operand
1082         constraints to operand 2.
1083         (*ashl<mode>3_mask_1): Ditto.
1084         (*<shift_insn><mode>3_mask): Ditto.
1085         (*<shift_insn><mode>3_mask_1): Ditto.
1086         (*<rotate_insn><mode>3_mask): Ditto.
1087         (*<rotate_insn><mode>3_mask_1): Ditto.
1089 2017-12-21  Alexandre Oliva  <aoliva@redhat.com>
1091         * reorg.c (make_return_insns): Reemit each insn with its own location.
1093 2017-12-21  James Greenhalgh  <james.greenhalgh@arm.com>
1095         * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
1096         generation for cases where splatting a value is not useful.
1097         * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
1098         across a vec_duplicate and a paradoxical subreg forming a vector
1099         mode to a vec_concat.
1101 2017-12-21  James Greenhalgh  <james.greenhalgh@arm.com>
1103         * combine.c (simplify_set): Do not transform subregs to zero_extends
1104         if the destination is not a scalar int mode.
1106 2017-12-21  Eric Botcazou  <ebotcazou@adacore.com>
1108         PR c++/82872
1109         * convert.c (convert_to_integer_1) <POINTER_TYPE>: Do not return the
1110         shared zero if the input has overflowed.
1112 2017-12-21  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
1114         * config/arm/driver-arm.c (arm_cpu_table): Specify dotprod
1115         support for Cortex-A55 and Cortex-A75.
1117 2017-12-21  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
1119         * common/config/arm/arm-common.c (compare_opt_names): Add function
1120         comment.  Use strcmp instead of manual loop.
1122 2017-12-21  Martin Liska  <mliska@suse.cz>
1124         PR gcov-profile/83509
1125         * gcov-dump.c (dump_gcov_file): Do not read info about
1126         support_unexecuted_blocks for gcda files.
1128 2017-12-21  Jakub Jelinek  <jakub@redhat.com>
1130         PR rtl-optimization/82973
1131         * emit-rtl.h (valid_for_const_vec_duplicate_p): Rename to ...
1132         (valid_for_const_vector_p): ... this.
1133         * emit-rtl.c (valid_for_const_vec_duplicate_p): Rename to ...
1134         (valid_for_const_vector_p): ... this.  Adjust function comment.
1135         (gen_vec_duplicate): Adjust caller.
1136         * optabs.c (expand_vector_broadcast): Likewise.
1137         * simplify-rtx.c (simplify_const_unary_operation): Don't optimize into
1138         CONST_VECTOR if some element isn't simplified valid_for_const_vector_p
1139         constant.
1140         (simplify_const_binary_operation): Likewise.  Use CONST_FIXED_P macro
1141         instead of GET_CODE == CONST_FIXED.
1142         (simplify_subreg): Use CONST_FIXED_P macro instead of
1143         GET_CODE == CONST_FIXED.
1145         PR target/83488
1146         * config/i386/i386.c (ix86_target_string): Move -mavx512vbmi2 and
1147         -mshstk entries from isa_opts2 to isa_opts and -mhle, -mmovbe,
1148         -mclzero and -mmwaitx entries from isa_opts to isa_opts2.
1149         (ix86_option_override_internal): Adjust for
1150         OPTION_MASK_ISA_{HLE,MOVBE,CLZERO,MWAITX} moving to ix86_isa_flags2
1151         and OPTION_MASK_ISA_SHSTK moving to ix86_isa_flags.
1152         (BDESC_VERIFYS): Remove SPECIAL_ARGS2 related checks.
1153         (ix86_init_mmx_sse_builtins): Remove bdesc_special_args2 handling.
1154         Use def_builtin2 instead of def_builtin for OPTION_MASK_ISA_MWAITX
1155         and OPTION_MASK_ISA_CLZERO builtins.  Use def_builtin instead of
1156         def_builtin2 for CET builtins.
1157         (ix86_expand_builtin): Remove bdesc_special_args2 handling.  Fix
1158         up formatting in IX86_BUILTIN_RDPID code.
1159         * config/i386/i386-builtin.def: Move VBMI2 builtins from SPECIAL_ARGS2
1160         section to SPECIAL_ARGS and from ARGS2 section to ARGS.
1161         * config/i386/i386.opt (mavx512vbmi2, mshstk): Move from
1162         ix86_isa_flags2 to ix86_isa_flags.
1163         (mhle, mmovbe, mclzero, mmwaitx): Move from ix86_isa_flags to
1164         ix86_isa_flags2.
1165         * config/i386/i386-c.c (ix86_target_macros_internal): Check for
1166         OPTION_MASK_ISA_{CLZERO,MWAITX} in isa_flag2 instead of isa_flag.
1167         Check for OPTION_MASK_ISA_{SHSTK,AVX512VBMI2} in isa_flag instead
1168         of isa_flag2.
1169         * common/config/i386/i386-common.c (OPTION_MASK_ISA_AVX512VBMI2_SET):
1170         Or in OPTION_MASK_ISA_AVX512F_SET.
1171         (OPTION_MASK_ISA_AVX512F_UNSET): Or in
1172         OPTION_MASK_ISA_AVX512VBMI2_UNSET.
1173         (ix86_handle_option): Adjust for
1174         OPTION_MASK_ISA_{SHSTK,AVX512VBMI2}_*SET being in ix86_isa_flags
1175         and OPTION_MASK_ISA_{MOVBE,MWAITX,CLZERO}_*SET in ix86_isa_flags2.
1177 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1178             Alan Hayward  <alan.hayward@arm.com>
1179             David Sherwood  <david.sherwood@arm.com>
1181         * tree-data-ref.h (prune_runtime_alias_test_list): Take the
1182         factor as a poly_uint64 rather than an unsigned HOST_WIDE_INT.
1183         * tree-data-ref.c (prune_runtime_alias_test_list): Likewise.
1184         Track polynomial offsets.
1186 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1187             Alan Hayward  <alan.hayward@arm.com>
1188             David Sherwood  <david.sherwood@arm.com>
1190         * tree-vect-data-refs.c (vect_compute_data_ref_alignment):
1191         Treat drb->init as a poly_int.  Fail if its misalignment wrt
1192         vector_alignment isn't known.
1194 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1195             Alan Hayward  <alan.hayward@arm.com>
1196             David Sherwood  <david.sherwood@arm.com>
1198         * tree-vectorizer.h (_loop_vec_info): Add a versioning_threshold
1199         field.
1200         (LOOP_VINFO_VERSIONING_THRESHOLD): New macro
1201         (vect_loop_versioning): Take the loop versioning threshold as a
1202         separate parameter.
1203         * tree-vect-loop-manip.c (vect_loop_versioning): Likewise.
1204         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
1205         versioning_threshold.
1206         (vect_analyze_loop_2): Compute the loop versioning threshold
1207         whenever loop versioning is needed, and store it in the new
1208         field rather than combining it with the cost model threshold.
1209         (vect_transform_loop): Update call to vect_loop_versioning.
1210         Try to combine the loop versioning and cost thresholds here.
1212 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1213             Alan Hayward  <alan.hayward@arm.com>
1214             David Sherwood  <david.sherwood@arm.com>
1216         * tree-ssa-loop-ivopts.h (strip_offset): Return the offset as
1217         poly_uint64_pod rather than an unsigned HOST_WIDE_INT.
1218         * tree-loop-distribution.c (classify_builtin_st): Update accordingly.
1219         * tree-ssa-loop-ivopts.c (iv_use::addr_offset): Change from
1220         an unsigned HOST_WIDE_INT to a poly_uint64_pod.
1221         (group_compare_offset): Update accordingly.
1222         (split_small_address_groups_p): Likewise.
1223         (record_use): Take addr_offset as a poly_uint64 rather than
1224         an unsigned HOST_WIDE_INT.
1225         (strip_offset): Return the offset as a poly_uint64 rather than
1226         an unsigned HOST_WIDE_INT.
1227         (record_group_use, split_address_groups): Track polynomial offsets.
1228         (add_iv_candidate_for_use): Likewise.
1229         (addr_offset_valid_p): Take the offset as a poly_int64 rather
1230         than a HOST_WIDE_INT.
1231         (strip_offset_1): Return the offset as a poly_int64 rather than
1232         a HOST_WIDE_INT.
1234 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1235             Alan Hayward  <alan.hayward@arm.com>
1236             David Sherwood  <david.sherwood@arm.com>
1238         * tree.h (get_binfo_at_offset): Take the offset as a poly_int64
1239         rather than a HOST_WIDE_INT.
1240         * tree.c (get_binfo_at_offset): Likewise.
1242 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1243             Alan Hayward  <alan.hayward@arm.com>
1244             David Sherwood  <david.sherwood@arm.com>
1246         * ipa-prop.h (build_ref_for_offset): Take the offset as a poly_int64
1247         rather than a HOST_WIDE_INT.
1248         * tree-sra.c (build_ref_for_offset): Likewise.
1250 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1251             Alan Hayward  <alan.hayward@arm.com>
1252             David Sherwood  <david.sherwood@arm.com>
1254         * fold-const.h (mem_ref_offset): Return a poly_offset_int rather
1255         than an offset_int.
1256         * tree.c (mem_ref_offset): Likewise.
1257         (build_simple_mem_ref_loc): Treat MEM_REF offsets as poly_ints.
1258         * builtins.c (get_object_alignment_2): Likewise.
1259         * expr.c (get_inner_reference, expand_expr_real_1): Likewise.
1260         * gimple-fold.c (get_base_constructor): Likewise.
1261         * gimple-ssa-strength-reduction.c (restructure_reference): Likewise.
1262         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref):
1263         Likewise.
1264         * ipa-polymorphic-call.c
1265         (ipa_polymorphic_call_context::ipa_polymorphic_call_context): Likewise.
1266         * ipa-prop.c (compute_complex_assign_jump_func): Likewise.
1267         (get_ancestor_addr_info): Likewise.
1268         * ipa-param-manipulation.c (ipa_get_adjustment_candidate): Likewise.
1269         * match.pd: Likewise.
1270         * tree-data-ref.c (dr_analyze_innermost): Likewise.
1271         * tree-dfa.c (get_addr_base_and_unit_offset_1): Likewise.
1272         * tree-eh.c (tree_could_trap_p): Likewise.
1273         * tree-object-size.c (addr_object_size): Likewise.
1274         * tree-ssa-address.c (copy_ref_info): Likewise.
1275         * tree-ssa-alias.c (indirect_ref_may_alias_decl_p): Likewise.
1276         (indirect_refs_may_alias_p): Likewise.
1277         * tree-ssa-sccvn.c (copy_reference_ops_from_ref): Likewise.
1278         * tree-ssa.c (maybe_rewrite_mem_ref_base): Likewise.
1279         (non_rewritable_mem_ref_base): Likewise.
1280         * tree-vect-data-refs.c (vect_check_gather_scatter): Likewise.
1281         * tree-vrp.c (vrp_prop::check_array_ref): Likewise.
1282         * varasm.c (decode_addr_const): Likewise.
1284 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1285             Alan Hayward  <alan.hayward@arm.com>
1286             David Sherwood  <david.sherwood@arm.com>
1288         * gimple-ssa-stor-merging.c (find_bswap_or_nop_load): Track polynomial
1289         offsets for MEM_REFs.
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-ssanames.h (adjust_ptr_info_misalignment): Take the increment
1296         as a poly_uint64 rather than an unsigned int.
1297         * tree-ssanames.c (adjust_ptr_info_misalignment): Likewise.
1299 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1300             Alan Hayward  <alan.hayward@arm.com>
1301             David Sherwood  <david.sherwood@arm.com>
1303         * varasm.c (addr_const::offset): Change from HOST_WIDE_INT
1304         to poly_int64.
1305         (decode_addr_const): Update accordingly.
1307 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1308             Alan Hayward  <alan.hayward@arm.com>
1309             David Sherwood  <david.sherwood@arm.com>
1311         * tree.h (bit_field_size, bit_field_offset): New functions.
1312         * hsa-gen.c (gen_hsa_addr): Use them.
1313         * tree-ssa-forwprop.c (simplify_bitfield_ref): Likewise.
1314         (simplify_vector_constructor): Likewise.
1315         * tree-ssa-sccvn.c (copy_reference_ops_from_ref): Likewise.
1316         * tree-cfg.c (verify_expr): Require the sizes and offsets of a
1317         BIT_FIELD_REF to be poly_uint64s rather than uhwis.
1318         * fold-const.c (fold_ternary_loc): Protect tree_to_uhwi with
1319         tree_fits_uhwi_p.
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         * expr.h (emit_group_load, emit_group_load_into_temps)
1326         (emit_group_store): Take the size as a poly_int64 rather than an int.
1327         * expr.c (emit_group_load_1, emit_group_load): Likewise.
1328         (emit_group_load_into_temp, emit_group_store): Likewise.
1330 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1331             Alan Hayward  <alan.hayward@arm.com>
1332             David Sherwood  <david.sherwood@arm.com>
1334         * ira-int.h (ira_spilled_reg_stack_slot::width): Change from
1335         an unsigned int to a poly_uint64.
1336         * ira.h (ira_reuse_stack_slot, ira_mark_new_stack_slot): Take the
1337         sizes as poly_uint64s rather than unsigned ints.
1338         * ira-color.c (ira_reuse_stack_slot, ira_mark_new_stack_slot):
1339         Likewise.
1341 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1342             Alan Hayward  <alan.hayward@arm.com>
1343             David Sherwood  <david.sherwood@arm.com>
1345         * lra-constraints.c (emit_inc): Change inc_amount from an int
1346         to a poly_int64.
1348 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1349             Alan Hayward  <alan.hayward@arm.com>
1350             David Sherwood  <david.sherwood@arm.com>
1352         * cfgexpand.c (stack_var::size): Change from a HOST_WIDE_INT
1353         to a poly_uint64.
1354         (add_stack_var, stack_var_cmp, partition_stack_vars)
1355         (dump_stack_var_partition): Update accordingly.
1356         (alloc_stack_frame_space): Take the size as a poly_int64 rather
1357         than a HOST_WIDE_INT.
1358         (expand_stack_vars, expand_one_stack_var_1): Handle polynomial sizes.
1359         (defer_stack_allocation, estimated_stack_frame_size): Likewise.
1360         (account_stack_vars, expand_one_var): Likewise.  Return a poly_uint64
1361         rather than a HOST_WIDE_INT.
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         * target.def (return_pops_args): Treat both the input and output
1368         sizes as poly_int64s rather than HOST_WIDE_INTS.
1369         * targhooks.h (default_return_pops_args): Update accordingly.
1370         * targhooks.c (default_return_pops_args): Likewise.
1371         * doc/tm.texi: Regenerate.
1372         * emit-rtl.h (incoming_args): Change pops_args, size and
1373         outgoing_args_size from int to poly_int64_pod.
1374         * function.h (expr_status): Change x_pending_stack_adjust and
1375         x_stack_pointer_delta from int to poly_int64.
1376         (args_size::constant): Change from HOST_WIDE_INT to poly_int64.
1377         (ARGS_SIZE_RTX): Update accordingly.
1378         * calls.c (highest_outgoing_arg_in_use): Change from int to
1379         unsigned int.
1380         (stack_usage_watermark, stored_args_watermark): New variables.
1381         (stack_region_maybe_used_p, mark_stack_region_used): New functions.
1382         (emit_call_1): Change the stack_size and rounded_stack_size
1383         parameters from HOST_WIDE_INT to poly_int64.  Track n_popped
1384         as a poly_int64.
1385         (save_fixed_argument_area): Check stack_usage_watermark.
1386         (initialize_argument_information): Change old_pending_adj from
1387         a HOST_WIDE_INT * to a poly_int64_pod *.
1388         (compute_argument_block_size): Return the size as a poly_int64
1389         rather than an int.
1390         (finalize_must_preallocate): Track polynomial argument sizes.
1391         (compute_argument_addresses): Likewise.
1392         (internal_arg_pointer_based_exp): Track polynomial offsets.
1393         (mem_overlaps_already_clobbered_arg_p): Rename to...
1394         (mem_might_overlap_already_clobbered_arg_p): ...this and take the
1395         size as a poly_uint64 rather than an unsigned HOST_WIDE_INT.
1396         Check stored_args_used_watermark.
1397         (load_register_parameters): Update accordingly.
1398         (check_sibcall_argument_overlap_1): Likewise.
1399         (combine_pending_stack_adjustment_and_call): Take the unadjusted
1400         args size as a poly_int64 rather than an int.  Return a bool
1401         indicating whether the optimization was possible and return
1402         the new adjustment by reference.
1403         (check_sibcall_argument_overlap): Track polynomail argument sizes.
1404         Update stored_args_watermark.
1405         (can_implement_as_sibling_call_p): Handle polynomial argument sizes.
1406         (expand_call): Likewise.  Maintain stack_usage_watermark and
1407         stored_args_watermark.  Update calls to
1408         combine_pending_stack_adjustment_and_call.
1409         (emit_library_call_value_1): Handle polynomial argument sizes.
1410         Call stack_region_maybe_used_p and mark_stack_region_used.
1411         Maintain stack_usage_watermark.
1412         (store_one_arg): Likewise.  Update call to
1413         mem_overlaps_already_clobbered_arg_p.
1414         * config/arm/arm.c (arm_output_function_prologue): Add a cast to
1415         HOST_WIDE_INT.
1416         * config/avr/avr.c (avr_outgoing_args_size): Likewise.
1417         * config/microblaze/microblaze.c (microblaze_function_prologue):
1418         Likewise.
1419         * config/cr16/cr16.c (cr16_return_pops_args): Update for new
1420         TARGET_RETURN_POPS_ARGS interface.
1421         (cr16_compute_frame, cr16_initial_elimination_offset): Add casts
1422         to HOST_WIDE_INT.
1423         * config/ft32/ft32.c (ft32_compute_frame): Likewise.
1424         * config/i386/i386.c (ix86_return_pops_args): Update for new
1425         TARGET_RETURN_POPS_ARGS interface.
1426         (ix86_expand_split_stack_prologue): Add a cast to HOST_WIDE_INT.
1427         * config/moxie/moxie.c (moxie_compute_frame): Likewise.
1428         * config/m68k/m68k.c (m68k_return_pops_args): Update for new
1429         TARGET_RETURN_POPS_ARGS interface.
1430         * config/vax/vax.c (vax_return_pops_args): Likewise.
1431         * config/pa/pa.h (STACK_POINTER_OFFSET): Add a cast to poly_int64.
1432         (EXIT_IGNORE_STACK): Update reference to crtl->outgoing_args_size.
1433         * config/arm/arm.h (CALLER_INTERWORKING_SLOT_SIZE): Likewise.
1434         * config/powerpcspe/aix.h (STACK_DYNAMIC_OFFSET): Likewise.
1435         * config/powerpcspe/darwin.h (STACK_DYNAMIC_OFFSET): Likewise.
1436         * config/powerpcspe/powerpcspe.h (STACK_DYNAMIC_OFFSET): Likewise.
1437         * config/rs6000/aix.h (STACK_DYNAMIC_OFFSET): Likewise.
1438         * config/rs6000/darwin.h (STACK_DYNAMIC_OFFSET): Likewise.
1439         * config/rs6000/rs6000.h (STACK_DYNAMIC_OFFSET): Likewise.
1440         * dojump.h (saved_pending_stack_adjust): Change x_pending_stack_adjust
1441         and x_stack_pointer_delta from int to poly_int64.
1442         * dojump.c (do_pending_stack_adjust): Update accordingly.
1443         * explow.c (allocate_dynamic_stack_space): Handle polynomial
1444         stack_pointer_deltas.
1445         * function.c (STACK_DYNAMIC_OFFSET): Add a cast to poly_int64.
1446         (pad_to_arg_alignment): Track polynomial offsets.
1447         (assign_parm_find_stack_rtl): Likewise.
1448         (assign_parms, locate_and_pad_parm): Handle polynomial argument sizes.
1449         * toplev.c (output_stack_usage): Update reference to
1450         current_function_pushed_stack_size.
1452 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1453             Alan Hayward  <alan.hayward@arm.com>
1454             David Sherwood  <david.sherwood@arm.com>
1456         * function.c (in_arg_offset, var_offset, dynamic_offset)
1457         (out_arg_offset, cfa_offset): Change from int to poly_int64.
1458         (instantiate_new_reg): Return the new offset as a poly_int64_pod
1459         rather than a HOST_WIDE_INT.
1460         (instantiate_virtual_regs_in_rtx): Track polynomial offsets.
1461         (instantiate_virtual_regs_in_insn): Likewise.
1463 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1464             Alan Hayward  <alan.hayward@arm.com>
1465             David Sherwood  <david.sherwood@arm.com>
1467         * rtl.h (get_args_size, add_args_size_note): New functions.
1468         (find_args_size_adjust): Return a poly_int64 rather than a
1469         HOST_WIDE_INT.
1470         (fixup_args_size_notes): Likewise.  Make the same change to the
1471         end_args_size parameter.
1472         * rtlanal.c (get_args_size, add_args_size_note): New functions.
1473         * builtins.c (expand_builtin_trap): Use add_args_size_note.
1474         * calls.c (emit_call_1): Likewise.
1475         * explow.c (adjust_stack_1): Likewise.
1476         * cfgcleanup.c (old_insns_match_p): Update use of
1477         find_args_size_adjust.
1478         * combine.c (distribute_notes): Track polynomial arg sizes.
1479         * dwarf2cfi.c (dw_trace_info): Change beg_true_args_size,
1480         end_true_args_size, beg_delay_args_size and end_delay_args_size
1481         from HOST_WIDE_INT to poly_int64.
1482         (add_cfi_args_size): Take the args_size as a poly_int64 rather
1483         than a HOST_WIDE_INT.
1484         (notice_args_size, notice_eh_throw, maybe_record_trace_start)
1485         (maybe_record_trace_start_abnormal, scan_trace, connect_traces): Track
1486         polynomial arg sizes.
1487         * emit-rtl.c (try_split): Use get_args_size.
1488         * recog.c (peep2_attempt): Likewise.
1489         * reload1.c (reload_as_needed): Likewise.
1490         * expr.c (find_args_size_adjust): Return the adjustment as a
1491         poly_int64 rather than a HOST_WIDE_INT.
1492         (fixup_args_size_notes): Change end_args_size from a HOST_WIDE_INT
1493         to a poly_int64 and change the return type in the same way.
1494         (emit_single_push_insn): Track polynomial arg sizes.
1496 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1497             Alan Hayward  <alan.hayward@arm.com>
1498             David Sherwood  <david.sherwood@arm.com>
1500         * expr.h (push_block, emit_push_insn): Change the "extra" parameter
1501         from HOST_WIDE_INT to poly_int64.
1502         * expr.c (push_block, emit_push_insn): Likewise.
1504 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1505             Alan Hayward  <alan.hayward@arm.com>
1506             David Sherwood  <david.sherwood@arm.com>
1508         * function.h (frame_space): Change start and length from HOST_WIDE_INT
1509         to poly_int64.
1510         (get_frame_size): Return the size as a poly_int64 rather than a
1511         HOST_WIDE_INT.
1512         (frame_offset_overflow): Take the offset as a poly_int64 rather
1513         than a HOST_WIDE_INT.
1514         (assign_stack_local_1, assign_stack_local, assign_stack_temp_for_type)
1515         (assign_stack_temp): Likewise for the size.
1516         * function.c (get_frame_size): Return a poly_int64 rather than
1517         a HOST_WIDE_INT.
1518         (frame_offset_overflow): Take the offset as a poly_int64 rather
1519         than a HOST_WIDE_INT.
1520         (try_fit_stack_local): Take the start, length and size as poly_int64s
1521         rather than HOST_WIDE_INTs.  Return the offset as a poly_int64_pod
1522         rather than a HOST_WIDE_INT.
1523         (add_frame_space): Take the start and end as poly_int64s rather than
1524         HOST_WIDE_INTs.
1525         (assign_stack_local_1, assign_stack_local, assign_stack_temp_for_type)
1526         (assign_stack_temp): Likewise for the size.
1527         (temp_slot): Change size, base_offset and full_size from HOST_WIDE_INT
1528         to poly_int64.
1529         (find_temp_slot_from_address): Handle polynomial offsets.
1530         (combine_temp_slots): Likewise.
1531         * emit-rtl.h (rtl_data::x_frame_offset): Change from HOST_WIDE_INT
1532         to poly_int64.
1533         * cfgexpand.c (alloc_stack_frame_space): Return the offset as a
1534         poly_int64 rather than a HOST_WIDE_INT.
1535         (expand_one_stack_var_at): Take the offset as a poly_int64 rather
1536         than a HOST_WIDE_INT.
1537         (expand_stack_vars, expand_one_stack_var_1, expand_used_vars): Handle
1538         polynomial frame offsets.
1539         * config/m32r/m32r-protos.h (m32r_compute_frame_size): Take the size
1540         as a poly_int64 rather than an int.
1541         * config/m32r/m32r.c (m32r_compute_frame_size): Likewise.
1542         * config/v850/v850-protos.h (compute_frame_size): Likewise.
1543         * config/v850/v850.c (compute_frame_size): Likewise.
1544         * config/xtensa/xtensa-protos.h (compute_frame_size): Likewise.
1545         * config/xtensa/xtensa.c (compute_frame_size): Likewise.
1546         * config/pa/pa-protos.h (pa_compute_frame_size): Likewise.
1547         * config/pa/pa.c (pa_compute_frame_size): Likewise.
1548         * explow.h (get_dynamic_stack_base): Take the offset as a poly_int64
1549         rather than a HOST_WIDE_INT.
1550         * explow.c (get_dynamic_stack_base): Likewise.
1551         * final.c (final_start_function): Use the constant lower bound
1552         of the frame size for -Wframe-larger-than.
1553         * ira.c (do_reload): Adjust for new get_frame_size return type.
1554         * lra.c (lra): Likewise.
1555         * reload1.c (reload): Likewise.
1556         * config/avr/avr.c (avr_asm_function_end_prologue): Likewise.
1557         * config/pa/pa.h (EXIT_IGNORE_STACK): Likewise.
1558         * rtlanal.c (get_initial_register_offset): Return the offset as
1559         a poly_int64 rather than a HOST_WIDE_INT.
1561 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1562             Alan Hayward  <alan.hayward@arm.com>
1563             David Sherwood  <david.sherwood@arm.com>
1565         * reload1.c (elim_table): Change initial_offset, offset and
1566         previous_offset from HOST_WIDE_INT to poly_int64_pod.
1567         (offsets_at): Change the target array's element type from
1568         HOST_WIDE_INT to poly_int64_pod.
1569         (set_label_offsets, eliminate_regs_1, eliminate_regs_in_insn)
1570         (elimination_costs_in_insn, update_eliminable_offsets)
1571         (verify_initial_elim_offsets, set_offsets_for_label)
1572         (init_eliminable_invariants): Update after above changes.
1574 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1575             Alan Hayward  <alan.hayward@arm.com>
1576             David Sherwood  <david.sherwood@arm.com>
1578         * reload.h (reload::inc): Change from an int to a poly_int64_pod.
1579         * reload.c (combine_reloads, debug_reload_to_stream): Likewise.
1580         (decomposition): Change start and end from HOST_WIDE_INT
1581         to poly_int64_pod.
1582         (decompose, immune_p): Update accordingly.
1583         (find_inc_amount): Return a poly_int64 rather than an int.
1584         * reload1.c (inc_for_reload): Take the inc_amount as a poly_int64
1585         rather than an int.
1587 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1588             Alan Hayward  <alan.hayward@arm.com>
1589             David Sherwood  <david.sherwood@arm.com>
1591         * tree.h (get_inner_reference): Return the bitsize and bitpos
1592         as poly_int64_pods rather than HOST_WIDE_INT.
1593         * fold-const.h (ptr_difference_const): Return the pointer difference
1594         as a poly_int64_pod rather than a HOST_WIDE_INT.
1595         * expr.c (get_inner_reference): Return the bitsize and bitpos
1596         as poly_int64_pods rather than HOST_WIDE_INT.
1597         (expand_expr_addr_expr_1, expand_expr_real_1): Track polynomial
1598         offsets and sizes.
1599         * fold-const.c (make_bit_field_ref): Take the bitpos as a poly_int64
1600         rather than a HOST_WIDE_INT.  Update call to get_inner_reference.
1601         (optimize_bit_field_compare): Update call to get_inner_reference.
1602         (decode_field_reference): Likewise.
1603         (fold_unary_loc): Track polynomial offsets and sizes.
1604         (split_address_to_core_and_offset): Return the bitpos as a
1605         poly_int64_pod rather than a HOST_WIDE_INT.
1606         (ptr_difference_const): Likewise for the pointer difference.
1607         * asan.c (instrument_derefs): Track polynomial offsets and sizes.
1608         * config/mips/mips.c (r10k_safe_mem_expr_p): Likewise.
1609         * dbxout.c (dbxout_expand_expr): Likewise.
1610         * dwarf2out.c (loc_list_for_address_of_addr_expr_of_indirect_ref)
1611         (loc_list_from_tree_1, fortran_common): Likewise.
1612         * gimple-laddress.c (pass_laddress::execute): Likewise.
1613         * gimple-ssa-store-merging.c (find_bswap_or_nop_load): Likewise.
1614         * gimplify.c (gimplify_scan_omp_clauses): Likewise.
1615         * simplify-rtx.c (delegitimize_mem_from_attrs): Likewise.
1616         * tree-affine.c (tree_to_aff_combination): Likewise.
1617         (get_inner_reference_aff): Likewise.
1618         * tree-data-ref.c (split_constant_offset_1): Likewise.
1619         (dr_analyze_innermost): Likewise.
1620         * tree-scalar-evolution.c (interpret_rhs_expr): Likewise.
1621         * tree-sra.c (ipa_sra_check_caller): Likewise.
1622         * tree-vect-data-refs.c (vect_check_gather_scatter): Likewise.
1623         * ubsan.c (maybe_instrument_pointer_overflow): Likewise.
1624         (instrument_bool_enum_load, instrument_object_size): Likewise.
1625         * gimple-ssa-strength-reduction.c (slsr_process_ref): Update call
1626         to get_inner_reference.
1627         * hsa-gen.c (gen_hsa_addr): Likewise.
1628         * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Likewise.
1629         * tsan.c (instrument_expr): Likewise.
1630         * match.pd: Update call to ptr_difference_const.
1632 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1633             Alan Hayward  <alan.hayward@arm.com>
1634             David Sherwood  <david.sherwood@arm.com>
1636         * fold-const.c (fold_comparison): Track sizes and offsets as
1637         poly_int64s rather than HOST_WIDE_INTs when folding address
1638         comparisons.
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         * expr.h (get_bit_range): Return the bitstart and bitend as
1645         poly_uint64s rather than unsigned HOST_WIDE_INTs.  Return the bitpos
1646         as a poly_int64 rather than a HOST_WIDE_INT.
1647         * expr.c (get_bit_range): Likewise.
1648         (expand_assignment): Update call accordingly.
1649         * fold-const.c (optimize_bit_field_compare): Likewise.
1651 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1652             Alan Hayward  <alan.hayward@arm.com>
1653             David Sherwood  <david.sherwood@arm.com>
1655         * config/aarch64/aarch64-protos.h (aarch64_addr_query_type): New enum.
1656         (aarch64_legitimate_address_p): Use it instead of an rtx code,
1657         as an optional final parameter.
1658         * config/aarch64/aarch64.c (aarch64_classify_address): Likewise.
1659         (aarch64_legitimate_address_p): Likewise.
1660         (aarch64_print_address_internal): Take an aarch64_addr_query_type
1661         instead of an rtx code.
1662         (aarch64_address_valid_for_prefetch_p): Update calls accordingly.
1663         (aarch64_legitimate_address_hook_p): Likewise.
1664         (aarch64_print_ldpstp_address): Likewise.
1665         (aarch64_print_operand_address): Likewise.
1666         (aarch64_address_cost): Likewise.
1667         * config/aarch64/constraints.md (Uml, Umq, Ump, Utq): Likewise.
1668         * config/aarch64/predicates.md (aarch64_mem_pair_operand): Likewise.
1669         (aarch64_mem_pair_lanes_operand): Likewise.
1671 2017-12-20  Richard Biener  <rguenther@suse.de>
1673         * tree-ssa-dom.c (dom_opt_dom_walker::optimize_stmt): Call
1674         update_stmt_if_modified.
1676 2017-12-20  Wilco Dijkstra  <wdijkstr@arm.com>
1678         PR tree-optimization/83491
1679         * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Check for SSA_NAME
1680         before walking uses.  Improve coding style and comments.
1682 2017-12-20  Tom de Vries  <tom@codesourcery.com>
1684         * gimple-fold.c (fold_internal_goacc_dim): Simplify.
1686 2017-12-20  Jakub Jelinek  <jakub@redhat.com>
1688         PR ipa/83506
1689         * ipa-fnsummary.c (pass_data_ipa_free_fn_summary): Use 0 for
1690         todo_flags_finish.
1691         (pass_ipa_free_fn_summary): Add small_p private data member,
1692         initialize to false in the ctor.
1693         (pass_ipa_free_fn_summary::clone,
1694         pass_ipa_free_fn_summary::set_pass_param,
1695         pass_ipa_free_fn_summary::gate): New methods.
1696         (pass_ipa_free_fn_summary::execute): Return TODO_remove_functions
1697         | TODO_dump_symtab if small_p.
1698         * passes.def: Add true parm for the existing pass_ipa_free_fn_summary
1699         entry and add another instance of the pass with false parm after
1700         ipa-pure-const.
1701         * ipa-pure-const.c (pass_ipa_pure_const): Don't call
1702         ipa_free_fn_summary here.
1704 2017-12-20  Paolo Carlini  <paolo.carlini@oracle.com>
1706         * gimplify.c (gimplify_return_expr): Remove dead error_mark_node check.
1708 2017-12-20  Martin Sebor  <msebor@redhat.com>
1710         PR testsuite/83131
1711         * builtins.c (expand_builtin_strlen): Use get_callee_fndecl.
1712         (expand_builtin_strcmp): Call maybe_warn_nonstring_arg.
1713         (expand_builtin_strncmp): Same.
1715 2017-12-20  Alexandre Oliva  <aoliva@redhat.com>
1717         PR bootstrap/83396
1718         * cfgexpand.c (label_rtx_for_bb): Revert SFN changes that
1719         allowed debug stmts before labels.
1720         (expand_gimple_basic_block): Likewise.
1721         * gimple-iterator.c (gimple_find_edge_insert_loc): Likewise.
1722         * gimple-iterator.h (gsi_after_labels): Likewise.
1723         * tree-cfgcleanup (remove_forwarder_block): Likewise, but
1724         rename reused variable, and simplify using gsi_move_before.
1725         * tree-ssa-tail-merge.c (find_duplicate): Likewise.
1726         * tree-cfg.c (make_edges, cleanup_dead_labels): Likewise.
1727         (gimple_can_merge_blocks_p, verify_gimple_in_cfg): Likewise.
1728         (gimple_verify_flow_info, gimple_block_label): Likewise.
1729         (make_blocks): Move debug markers after adjacent labels.
1730         * cfgrtl.c (skip_insns_after_block): Revert SFN changes that
1731         allowed debug insns outside blocks.
1732         * df-scan.c (df_insn_delete): Likewise.
1733         * lra-constraints.c (update_ebb_live_info): Likewise.
1734         * var-tracking.c (get_first_insn, vt_emit_notes): Likewise.
1735         (vt_initialize, delete_vta_debug_insns): Likewise.
1736         (reemit_marker_as_note): Drop BB parm.  Adjust callers.
1738 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
1739             Alan Hayward  <alan.hayward@arm.com>
1740             David Sherwood  <david.sherwood@arm.com>
1742         * poly-int-types.h (round_down_to_byte_boundary): New macro.
1743         (round_up_to_byte_boundary): Likewise.
1744         * expr.h (get_bit_range): Add temporary shim.
1745         * gimple-ssa-store-merging.c (store_operand_info): Change the
1746         bitsize, bitpos, bitregion_start and bitregion_end fields from
1747         unsigned HOST_WIDE_INT to poly_uint64.
1748         (merged_store_group): Likewise load_align_base.
1749         (compatible_load_p, compatible_load_p): Update accordingly.
1750         (imm_store_chain_info::coalesce_immediate_stores): Likewise.
1751         (split_group, imm_store_chain_info::output_merged_store): Likewise.
1752         (mem_valid_for_store_merging): Return the bitsize, bitpos,
1753         bitregion_start and bitregion_end as poly_uint64s rather than
1754         unsigned HOST_WIDE_INTs.  Track polynomial offsets internally.
1755         (handled_load): Take the bitsize, bitpos,
1756         bitregion_start and bitregion_end as poly_uint64s rather than
1757         unsigned HOST_WIDE_INTs.
1758         (pass_store_merging::process_store): Update call to
1759         mem_valid_for_store_merging.
1761 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
1762             Alan Hayward  <alan.hayward@arm.com>
1763             David Sherwood  <david.sherwood@arm.com>
1765         * builtins.c (get_object_alignment_2): Track polynomial offsets
1766         and sizes.  Update the alignment handling.
1768 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
1769             Alan Hayward  <alan.hayward@arm.com>
1770             David Sherwood  <david.sherwood@arm.com>
1772         * tree.h (get_inner_reference): Add a version that returns the
1773         offset and size as poly_int64_pods rather than HOST_WIDE_INTs.
1774         * cfgexpand.c (expand_debug_expr): Track polynomial offsets.  Simply
1775         the case in which bitpos is not associated with the first byte.
1777 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
1778             Alan Hayward  <alan.hayward@arm.com>
1779             David Sherwood  <david.sherwood@arm.com>
1781         * tree-affine.h (get_inner_reference_aff): Return the size as a
1782         poly_widest_int.
1783         * tree-affine.c (get_inner_reference_aff): Likewise.
1784         * tree-data-ref.c (dr_may_alias_p): Update accordingly.
1785         * tree-ssa-loop-im.c (mem_refs_may_alias_p): Likewise.
1787 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
1788             Alan Hayward  <alan.hayward@arm.com>
1789             David Sherwood  <david.sherwood@arm.com>
1791         * fold-const.c (pointer_may_wrap_p): Take the offset as a
1792         HOST_WIDE_INT rather than a poly_int64.
1794 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
1795             Alan Hayward  <alan.hayward@arm.com>
1796             David Sherwood  <david.sherwood@arm.com>
1798         * gimple-ssa-store-merging.c (symbolic_number::bytepos): Change from
1799         HOST_WIDE_INT to poly_int64_pod.
1800         (perform_symbolic_merge): Update accordingly.
1801         (bswap_replace): Likewise.
1803 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
1804             Alan Hayward  <alan.hayward@arm.com>
1805             David Sherwood  <david.sherwood@arm.com>
1807         * tree-affine.h (aff_tree::offset): Change from widest_int
1808         to poly_widest_int.
1809         (wide_int_ext_for_comb): Delete.
1810         (aff_combination_const, aff_comb_cannot_overlap_p): Take the
1811         constants as poly_widest_int rather than widest_int.
1812         (aff_combination_constant_multiple_p): Return the multiplier
1813         as a poly_widest_int.
1814         (aff_combination_zero_p, aff_combination_singleton_var_p): Handle
1815         polynomial offsets.
1816         * tree-affine.c (wide_int_ext_for_comb): Make original widest_int
1817         version static and add an overload for poly_widest_int.
1818         (aff_combination_const, aff_combination_add_cst)
1819         (wide_int_constant_multiple_p, aff_comb_cannot_overlap_p): Take
1820         the constants as poly_widest_int rather than widest_int.
1821         (tree_to_aff_combination): Generalize INTEGER_CST case to
1822         poly_int_tree_p.
1823         (aff_combination_to_tree): Track offsets as poly_widest_ints.
1824         (aff_combination_add_product, aff_combination_mult): Handle
1825         polynomial offsets.
1826         (aff_combination_constant_multiple_p): Return the multiplier
1827         as a poly_widest_int.
1828         * tree-predcom.c (determine_offset): Return the offset as a
1829         poly_widest_int.
1830         (split_data_refs_to_components, suitable_component_p): Update
1831         accordingly.
1832         (valid_initializer_p): Update call to
1833         aff_combination_constant_multiple_p.
1834         * tree-ssa-address.c (addr_to_parts): Handle polynomial offsets.
1835         * tree-ssa-loop-ivopts.c (get_address_cost_ainc): Take the step
1836         as a poly_int64 rather than a HOST_WIDE_INT.
1837         (get_address_cost): Handle polynomial offsets.
1838         (iv_elimination_compare_lt): Likewise.
1839         (rewrite_use_nonlinear_expr): Likewise.
1841 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
1842             Alan Hayward  <alan.hayward@arm.com>
1843             David Sherwood  <david.sherwood@arm.com>
1845         * tree-dfa.h (get_addr_base_and_unit_offset_1): Return the offset
1846         as a poly_int64_pod rather than a HOST_WIDE_INT.
1847         (get_addr_base_and_unit_offset): Likewise.
1848         * tree-dfa.c (get_addr_base_and_unit_offset_1): Likewise.
1849         (get_addr_base_and_unit_offset): Likewise.
1850         * doc/match-and-simplify.texi: Change off from HOST_WIDE_INT
1851         to poly_int64 in example.
1852         * fold-const.c (fold_binary_loc): Update call to
1853         get_addr_base_and_unit_offset.
1854         * gimple-fold.c (gimple_fold_builtin_memory_op): Likewise.
1855         (maybe_canonicalize_mem_ref_addr): Likewise.
1856         (gimple_fold_stmt_to_constant_1): Likewise.
1857         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref):
1858         Likewise.
1859         * ipa-param-manipulation.c (ipa_modify_call_arguments): Likewise.
1860         * match.pd: Likewise.
1861         * omp-low.c (lower_omp_target): Likewise.
1862         * tree-sra.c (build_ref_for_offset): Likewise.
1863         (build_debug_ref_for_model): Likewise.
1864         * tree-ssa-address.c (maybe_fold_tmr): Likewise.
1865         * tree-ssa-alias.c (ao_ref_init_from_ptr_and_size): Likewise.
1866         * tree-ssa-ccp.c (optimize_memcpy): Likewise.
1867         * tree-ssa-forwprop.c (forward_propagate_addr_expr_1): Likewise.
1868         (constant_pointer_difference): Likewise.
1869         * tree-ssa-loop-niter.c (expand_simple_operations): Likewise.
1870         * tree-ssa-phiopt.c (jump_function_from_stmt): Likewise.
1871         * tree-ssa-pre.c (create_component_ref_by_pieces_1): Likewise.
1872         * tree-ssa-sccvn.c (vn_reference_fold_indirect): Likewise.
1873         (vn_reference_maybe_forwprop_address, vn_reference_lookup_3): Likewise.
1874         (set_ssa_val_to): Likewise.
1875         * tree-ssa-strlen.c (get_addr_stridx, addr_stridxptr)
1876         (maybe_diag_stxncpy_trunc): Likewise.
1877         * tree-vrp.c (vrp_prop::check_array_ref): Likewise.
1878         * tree.c (build_simple_mem_ref_loc): Likewise.
1879         (array_at_struct_end_p): Likewise.
1881 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
1882             Alan Hayward  <alan.hayward@arm.com>
1883             David Sherwood  <david.sherwood@arm.com>
1885         * tree-dfa.h (get_ref_base_and_extent): Return the base, size and
1886         max_size as poly_int64_pods rather than HOST_WIDE_INTs.
1887         (get_ref_base_and_extent_hwi): Declare.
1888         * tree-dfa.c (get_ref_base_and_extent): Return the base, size and
1889         max_size as poly_int64_pods rather than HOST_WIDE_INTs.
1890         (get_ref_base_and_extent_hwi): New function.
1891         * cfgexpand.c (expand_debug_expr): Update call to
1892         get_ref_base_and_extent.
1893         * dwarf2out.c (add_var_loc_to_decl): Likewise.
1894         * gimple-fold.c (get_base_constructor): Return the offset as a
1895         poly_int64_pod rather than a HOST_WIDE_INT.
1896         (fold_const_aggregate_ref_1): Track polynomial sizes and offsets.
1897         * ipa-polymorphic-call.c
1898         (ipa_polymorphic_call_context::set_by_invariant)
1899         (extr_type_from_vtbl_ptr_store): Track polynomial offsets.
1900         (ipa_polymorphic_call_context::ipa_polymorphic_call_context)
1901         (check_stmt_for_type_change): Use get_ref_base_and_extent_hwi
1902         rather than get_ref_base_and_extent.
1903         (ipa_polymorphic_call_context::get_dynamic_type): Likewise.
1904         * ipa-prop.c (ipa_load_from_parm_agg, compute_complex_assign_jump_func)
1905         (get_ancestor_addr_info, determine_locally_known_aggregate_parts):
1906         Likewise.
1907         * ipa-param-manipulation.c (ipa_get_adjustment_candidate): Update
1908         call to get_ref_base_and_extent.
1909         * tree-sra.c (create_access, get_access_for_expr): Likewise.
1910         * tree-ssa-alias.c (ao_ref_base, aliasing_component_refs_p)
1911         (stmt_kills_ref_p): Likewise.
1912         * tree-ssa-dce.c (mark_aliased_reaching_defs_necessary_1): Likewise.
1913         * tree-ssa-scopedtables.c (avail_expr_hash, equal_mem_array_ref_p):
1914         Likewise.
1915         * tree-ssa-sccvn.c (vn_reference_lookup_3): Likewise.
1916         Use get_ref_base_and_extent_hwi rather than get_ref_base_and_extent
1917         when calling native_encode_expr.
1918         * tree-ssa-structalias.c (get_constraint_for_component_ref): Update
1919         call to get_ref_base_and_extent.
1920         (do_structure_copy): Use get_ref_base_and_extent_hwi rather than
1921         get_ref_base_and_extent.
1922         * var-tracking.c (track_expr_p): Likewise.
1924 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
1925             Alan Hayward  <alan.hayward@arm.com>
1926             David Sherwood  <david.sherwood@arm.com>
1928         * ipa-param-manipulation.h (ipa_parm_adjustment::offset): Change from
1929         HOST_WIDE_INT to poly_int64_pod.
1930         * ipa-param-manipulation.c (ipa_modify_call_arguments): Track
1931         polynomail parameter offsets.
1933 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
1934             Alan Hayward  <alan.hayward@arm.com>
1935             David Sherwood  <david.sherwood@arm.com>
1937         * gengtype.c (main): Handle poly_int64_pod.
1938         * dwarf2out.h (dw_cfi_oprnd_cfa_loc): New dw_cfi_oprnd_type.
1939         (dw_cfi_oprnd::dw_cfi_cfa_loc): New field.
1940         (dw_cfa_location::offset, dw_cfa_location::base_offset): Change
1941         from HOST_WIDE_INT to poly_int64_pod.
1942         * dwarf2cfi.c (queued_reg_save::cfa_offset): Likewise.
1943         (copy_cfa): New function.
1944         (lookup_cfa_1): Use the cached dw_cfi_cfa_loc, if it exists.
1945         (cfi_oprnd_equal_p): Handle dw_cfi_oprnd_cfa_loc.
1946         (cfa_equal_p, dwarf2out_frame_debug_adjust_cfa)
1947         (dwarf2out_frame_debug_cfa_offset, dwarf2out_frame_debug_expr)
1948         (initial_return_save): Treat offsets as poly_ints.
1949         (def_cfa_0): Likewise.  Cache the CFA in dw_cfi_cfa_loc if either
1950         offset is nonconstant.
1951         (reg_save): Take the offset as a poly_int64.  Fall back to
1952         DW_CFA_expression for nonconstant offsets.
1953         (queue_reg_save): Take the offset as a poly_int64.
1954         * dwarf2out.c (dw_cfi_oprnd2_desc): Handle DW_CFA_def_cfa_expression.
1956 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
1957             Alan Hayward  <alan.hayward@arm.com>
1958             David Sherwood  <david.sherwood@arm.com>
1960         * rtl.h (operand_subword, operand_subword_force): Take the offset
1961         as a poly_uint64 an unsigned int.
1962         * emit-rtl.c (operand_subword, operand_subword_force): Likewise.
1964 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
1965             Alan Hayward  <alan.hayward@arm.com>
1966             David Sherwood  <david.sherwood@arm.com>
1968         * doc/rtl.texi: Update documentation of SUBREG_BYTE.  Document the
1969         'p' format code.  Use INT_LIST rather than SUBREG as the example of
1970         a code with an XINT and an XEXP.  Remove the implication that
1971         accessing an rtx field using XINT is expected to work.
1972         * rtl.def (SUBREG): Change format from "ei" to "ep".
1973         * rtl.h (rtunion::rt_subreg): New field.
1974         (XCSUBREG): New macro.
1975         (SUBREG_BYTE): Use it.
1976         (subreg_shape): Change offset from an unsigned int to a poly_uint16.
1977         Update constructor accordingly.
1978         (subreg_shape::operator ==): Update accordingly.
1979         (subreg_shape::unique_id): Return an unsigned HOST_WIDE_INT rather
1980         than an unsigned int.
1981         (subreg_lsb, subreg_lowpart_offset, subreg_highpart_offset): Return
1982         a poly_uint64 rather than an unsigned int.
1983         (subreg_lsb_1): Likewise.  Take the offset as a poly_uint64 rather
1984         than an unsigned int.
1985         (subreg_size_offset_from_lsb, subreg_size_lowpart_offset)
1986         (subreg_size_highpart_offset): Return a poly_uint64 rather than
1987         an unsigned int.  Take the sizes as poly_uint64s.
1988         (subreg_offset_from_lsb): Return a poly_uint64 rather than
1989         an unsigned int.  Take the shift as a poly_uint64 rather than
1990         an unsigned int.
1991         (subreg_regno_offset, subreg_offset_representable_p): Take the offset
1992         as a poly_uint64 rather than an unsigned int.
1993         (simplify_subreg_regno): Likewise.
1994         (byte_lowpart_offset): Return the memory offset as a poly_int64
1995         rather than an int.
1996         (subreg_memory_offset): Likewise.  Take the subreg offset as a
1997         poly_uint64 rather than an unsigned int.
1998         (simplify_subreg, simplify_gen_subreg, subreg_get_info)
1999         (gen_rtx_SUBREG, validate_subreg): Take the subreg offset as a
2000         poly_uint64 rather than an unsigned int.
2001         * rtl.c (rtx_format): Describe 'p' in comment.
2002         (copy_rtx, rtx_equal_p_cb, rtx_equal_p): Handle 'p'.
2003         * emit-rtl.c (validate_subreg, gen_rtx_SUBREG): Take the subreg
2004         offset as a poly_uint64 rather than an unsigned int.
2005         (byte_lowpart_offset): Return the memory offset as a poly_int64
2006         rather than an int.
2007         (subreg_memory_offset): Likewise.  Take the subreg offset as a
2008         poly_uint64 rather than an unsigned int.
2009         (subreg_size_lowpart_offset, subreg_size_highpart_offset): Take the
2010         mode sizes as poly_uint64s rather than unsigned ints.  Return a
2011         poly_uint64 rather than an unsigned int.
2012         (subreg_lowpart_p): Treat subreg offsets as poly_ints.
2013         (copy_insn_1): Handle 'p'.
2014         * rtlanal.c (set_noop_p): Treat subregs offsets as poly_uint64s.
2015         (subreg_lsb_1): Take the subreg offset as a poly_uint64 rather than
2016         an unsigned int.  Return the shift in the same way.
2017         (subreg_lsb): Return the shift as a poly_uint64 rather than an
2018         unsigned int.
2019         (subreg_size_offset_from_lsb): Take the sizes and shift as
2020         poly_uint64s rather than unsigned ints.  Return the offset as
2021         a poly_uint64.
2022         (subreg_get_info, subreg_regno_offset, subreg_offset_representable_p)
2023         (simplify_subreg_regno): Take the offset as a poly_uint64 rather than
2024         an unsigned int.
2025         * rtlhash.c (add_rtx): Handle 'p'.
2026         * genemit.c (gen_exp): Likewise.
2027         * gengenrtl.c (type_from_format, gendef): Likewise.
2028         * gensupport.c (subst_pattern_match, get_alternatives_number)
2029         (collect_insn_data, alter_predicate_for_insn, alter_constraints)
2030         (subst_dup): Likewise.
2031         * gengtype.c (adjust_field_rtx_def): Likewise.
2032         * genrecog.c (find_operand, find_matching_operand, validate_pattern)
2033         (match_pattern_2): Likewise.
2034         (rtx_test::SUBREG_FIELD): New rtx_test::kind_enum.
2035         (rtx_test::subreg_field): New function.
2036         (operator ==, safe_to_hoist_p, transition_parameter_type)
2037         (print_nonbool_test, print_test): Handle SUBREG_FIELD.
2038         * genattrtab.c (attr_rtx_1): Say that 'p' is deliberately not handled.
2039         * genpeep.c (match_rtx): Likewise.
2040         * print-rtl.c (print_poly_int): Include if GENERATOR_FILE too.
2041         (rtx_writer::print_rtx_operand): Handle 'p'.
2042         (print_value): Handle SUBREG.
2043         * read-rtl.c (apply_int_iterator): Likewise.
2044         (rtx_reader::read_rtx_operand): Handle 'p'.
2045         * alias.c (rtx_equal_for_memref_p): Likewise.
2046         * cselib.c (rtx_equal_for_cselib_1, cselib_hash_rtx): Likewise.
2047         * caller-save.c (replace_reg_with_saved_mem): Treat subreg offsets
2048         as poly_ints.
2049         * calls.c (expand_call): Likewise.
2050         * combine.c (combine_simplify_rtx, expand_field_assignment): Likewise.
2051         (make_extraction, gen_lowpart_for_combine): Likewise.
2052         * loop-invariant.c (hash_invariant_expr_1, invariant_expr_equal_p):
2053         Likewise.
2054         * cse.c (remove_invalid_subreg_refs): Take the offset as a poly_uint64
2055         rather than an unsigned int.  Treat subreg offsets as poly_ints.
2056         (exp_equiv_p): Handle 'p'.
2057         (hash_rtx_cb): Likewise.  Treat subreg offsets as poly_ints.
2058         (equiv_constant, cse_insn): Treat subreg offsets as poly_ints.
2059         * dse.c (find_shift_sequence): Likewise.
2060         * dwarf2out.c (rtl_for_decl_location): Likewise.
2061         * expmed.c (extract_low_bits): Likewise.
2062         * expr.c (emit_group_store, undefined_operand_subword_p): Likewise.
2063         (expand_expr_real_2): Likewise.
2064         * final.c (alter_subreg): Likewise.
2065         (leaf_renumber_regs_insn): Handle 'p'.
2066         * function.c (assign_parm_find_stack_rtl, assign_parm_setup_stack):
2067         Treat subreg offsets as poly_ints.
2068         * fwprop.c (forward_propagate_and_simplify): Likewise.
2069         * ifcvt.c (noce_emit_move_insn, noce_emit_cmove): Likewise.
2070         * ira.c (get_subreg_tracking_sizes): Likewise.
2071         * ira-conflicts.c (go_through_subreg): Likewise.
2072         * ira-lives.c (process_single_reg_class_operands): Likewise.
2073         * jump.c (rtx_renumbered_equal_p): Likewise.  Handle 'p'.
2074         * lower-subreg.c (simplify_subreg_concatn): Take the subreg offset
2075         as a poly_uint64 rather than an unsigned int.
2076         (simplify_gen_subreg_concatn, resolve_simple_move): Treat
2077         subreg offsets as poly_ints.
2078         * lra-constraints.c (operands_match_p): Handle 'p'.
2079         (match_reload, curr_insn_transform): Treat subreg offsets as poly_ints.
2080         * lra-spills.c (assign_mem_slot): Likewise.
2081         * postreload.c (move2add_valid_value_p): Likewise.
2082         * recog.c (general_operand, indirect_operand): Likewise.
2083         * regcprop.c (copy_value, maybe_mode_change): Likewise.
2084         (copyprop_hardreg_forward_1): Likewise.
2085         * reginfo.c (simplifiable_subregs_hasher::hash, simplifiable_subregs)
2086         (record_subregs_of_mode): Likewise.
2087         * rtlhooks.c (gen_lowpart_general, gen_lowpart_if_possible): Likewise.
2088         * reload.c (operands_match_p): Handle 'p'.
2089         (find_reloads_subreg_address): Treat subreg offsets as poly_ints.
2090         * reload1.c (alter_reg, choose_reload_regs): Likewise.
2091         (compute_reload_subreg_offset): Likewise, and return an poly_int64.
2092         * simplify-rtx.c (simplify_truncation, simplify_binary_operation_1):
2093         (test_vector_ops_duplicate): Treat subreg offsets as poly_ints.
2094         (simplify_const_poly_int_tests<N>::run): Likewise.
2095         (simplify_subreg, simplify_gen_subreg): Take the subreg offset as
2096         a poly_uint64 rather than an unsigned int.
2097         * valtrack.c (debug_lowpart_subreg): Likewise.
2098         * var-tracking.c (var_lowpart): Likewise.
2099         (loc_cmp): Handle 'p'.
2101 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2102             Alan Hayward  <alan.hayward@arm.com>
2103             David Sherwood  <david.sherwood@arm.com>
2105         * ira.c (get_subreg_tracking_sizes): New function.
2106         (init_live_subregs): Take an integer size rather than a register.
2107         (build_insn_chain): Use get_subreg_tracking_sizes.  Update calls
2108         to init_live_subregs.
2110 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2111             Alan Hayward  <alan.hayward@arm.com>
2112             David Sherwood  <david.sherwood@arm.com>
2114         * expr.c (store_constructor_field): Change bitsize from a
2115         unsigned HOST_WIDE_INT to a poly_uint64 and bitpos from a
2116         HOST_WIDE_INT to a poly_int64.
2117         (store_constructor): Change size from a HOST_WIDE_INT to
2118         a poly_int64.
2119         (store_field): Likewise bitsize and bitpos.
2121 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2122             Alan Hayward  <alan.hayward@arm.com>
2123             David Sherwood  <david.sherwood@arm.com>
2125         * expmed.h (store_bit_field): Change bitregion_start and
2126         bitregion_end from unsigned HOST_WIDE_INT to poly_uint64.
2127         * expmed.c (adjust_bit_field_mem_for_reg, strict_volatile_bitfield_p)
2128         (store_bit_field_1, store_integral_bit_field, store_bit_field)
2129         (store_fixed_bit_field, store_split_bit_field): Likewise.
2130         * expr.c (store_constructor_field, store_field): Likewise.
2131         (optimize_bitfield_assignment_op): Likewise.  Make the same change
2132         to bitsize and bitpos.
2133         * machmode.h (bit_field_mode_iterator): Change m_bitregion_start
2134         and m_bitregion_end from HOST_WIDE_INT to poly_int64.  Make the
2135         same change in the constructor arguments.
2136         (get_best_mode): Change bitregion_start and bitregion_end from
2137         unsigned HOST_WIDE_INT to poly_uint64.
2138         * stor-layout.c (bit_field_mode_iterator::bit_field_mode_iterator):
2139         Change bitregion_start and bitregion_end from HOST_WIDE_INT to
2140         poly_int64.
2141         (bit_field_mode_iterator::next_mode): Update for new types
2142         of m_bitregion_start and m_bitregion_end.
2143         (get_best_mode): Change bitregion_start and bitregion_end from
2144         unsigned HOST_WIDE_INT to poly_uint64.
2146 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2147             Alan Hayward  <alan.hayward@arm.com>
2148             David Sherwood  <david.sherwood@arm.com>
2150         * rtl.h (simplify_gen_subreg): Add a temporary overload that
2151         accepts poly_uint64 offsets.
2152         * expmed.h (extract_bit_field): Take bitsize and bitnum as
2153         poly_uint64s rather than unsigned HOST_WIDE_INTs.
2154         * expmed.c (lowpart_bit_field_p): Likewise.
2155         (extract_bit_field_as_subreg): New function, split out from...
2156         (extract_bit_field_1): ...here.  Take bitsize and bitnum as
2157         poly_uint64s rather than unsigned HOST_WIDE_INTs.  For vector
2158         extractions, check that BITSIZE matches the size of the extracted
2159         value and that BITNUM is an exact multiple of that size.
2160         If all else fails, try forcing the value into memory if
2161         BITNUM is variable, and adjusting the address so that the
2162         offset is constant.  Split the part that can only handle constant
2163         bitsize and bitnum out into...
2164         (extract_integral_bit_field): ...this new function.
2165         (extract_bit_field): Take bitsize and bitnum as poly_uint64s
2166         rather than unsigned HOST_WIDE_INTs.
2168 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2169             Alan Hayward  <alan.hayward@arm.com>
2170             David Sherwood  <david.sherwood@arm.com>
2172         * expmed.h (store_bit_field): Take bitsize and bitnum as
2173         poly_uint64s rather than unsigned HOST_WIDE_INTs.
2174         * expmed.c (simple_mem_bitfield_p): Likewise.  Add a parameter
2175         that returns the byte size.
2176         (store_bit_field_1): Take bitsize and bitnum as
2177         poly_uint64s rather than unsigned HOST_WIDE_INTs.  Update call
2178         to simple_mem_bitfield_p.  Split the part that can only handle
2179         constant bitsize and bitnum out into...
2180         (store_integral_bit_field): ...this new function.
2181         (store_bit_field): Take bitsize and bitnum as poly_uint64s rather
2182         than unsigned HOST_WIDE_INTs.
2183         (extract_bit_field_1): Update call to simple_mem_bitfield_p.
2185 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2186             Alan Hayward  <alan.hayward@arm.com>
2187             David Sherwood  <david.sherwood@arm.com>
2189         * lra-int.h (lra_reg): Change offset from int to poly_int64.
2190         (lra_insn_recog_data): Change sp_offset from HOST_WIDE_INT
2191         to poly_int64.
2192         (lra_eliminate_regs_1, eliminate_regs_in_insn): Change
2193         update_sp_offset from a HOST_WIDE_INT to a poly_int64.
2194         (lra_update_reg_val_offset, lra_reg_val_equal_p): Take the
2195         offset as a poly_int64 rather than an int.
2196         * lra-assigns.c (find_hard_regno_for_1): Handle poly_int64 offsets.
2197         (setup_live_pseudos_and_spill_after_risky_transforms): Likewise.
2198         * lra-constraints.c (equiv_address_substitution): Track offsets
2199         as poly_int64s.
2200         (emit_inc): Check poly_int_rtx_p instead of CONST_INT_P.
2201         (curr_insn_transform): Handle the new form of sp_offset.
2202         * lra-eliminations.c (lra_elim_table): Change previous_offset
2203         and offset from HOST_WIDE_INT to poly_int64.
2204         (print_elim_table, update_reg_eliminate): Update accordingly.
2205         (self_elim_offsets): Change from HOST_WIDE_INT to poly_int64_pod.
2206         (get_elimination): Update accordingly.
2207         (form_sum): Check poly_int_rtx_p instead of CONST_INT_P.
2208         (lra_eliminate_regs_1, eliminate_regs_in_insn): Change
2209         update_sp_offset from a HOST_WIDE_INT to a poly_int64.  Handle
2210         poly_int64 offsets generally.
2211         (curr_sp_change): Change from HOST_WIDE_INT to poly_int64.
2212         (mark_not_eliminable, init_elimination): Update accordingly.
2213         (remove_reg_equal_offset_note): Return a bool and pass the new
2214         offset back by pointer as a poly_int64.
2215         * lra-remat.c (change_sp_offset): Take sp_offset as a poly_int64
2216         rather than a HOST_WIDE_INT.
2217         (do_remat): Track offsets poly_int64s.
2218         * lra.c (lra_update_insn_recog_data, setup_sp_offset): Likewise.
2220 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2221             Alan Hayward  <alan.hayward@arm.com>
2222             David Sherwood  <david.sherwood@arm.com>
2224         * rtl.h (mem_attrs): Add a default constructor.  Change size and
2225         offset from HOST_WIDE_INT to poly_int64.
2226         * emit-rtl.h (set_mem_offset, set_mem_size, adjust_address_1)
2227         (adjust_automodify_address_1, set_mem_attributes_minus_bitpos)
2228         (widen_memory_access): Take the sizes and offsets as poly_int64s
2229         rather than HOST_WIDE_INTs.
2230         * alias.c (ao_ref_from_mem): Handle the new form of MEM_OFFSET.
2231         (offset_overlap_p): Take poly_int64s rather than HOST_WIDE_INTs
2232         and ints.
2233         (adjust_offset_for_component_ref): Change the offset from a
2234         HOST_WIDE_INT to a poly_int64.
2235         (nonoverlapping_memrefs_p): Track polynomial offsets and sizes.
2236         * cfgcleanup.c (merge_memattrs): Update after mem_attrs changes.
2237         * dce.c (find_call_stack_args): Likewise.
2238         * dse.c (record_store): Likewise.
2239         * dwarf2out.c (tls_mem_loc_descriptor, dw_sra_loc_expr): Likewise.
2240         * print-rtl.c (rtx_writer::print_rtx): Likewise.
2241         * read-rtl-function.c (test_loading_mem): Likewise.
2242         * rtlanal.c (may_trap_p_1): Likewise.
2243         * simplify-rtx.c (delegitimize_mem_from_attrs): Likewise.
2244         * var-tracking.c (int_mem_offset, track_expr_p): Likewise.
2245         * emit-rtl.c (mem_attrs_eq_p, get_mem_align_offset): Likewise.
2246         (mem_attrs::mem_attrs): New function.
2247         (set_mem_attributes_minus_bitpos): Change bitpos from a
2248         HOST_WIDE_INT to poly_int64.
2249         (set_mem_alias_set, set_mem_addr_space, set_mem_align, set_mem_expr)
2250         (clear_mem_offset, clear_mem_size, change_address)
2251         (get_spill_slot_decl, set_mem_attrs_for_spill): Directly
2252         initialize mem_attrs.
2253         (set_mem_offset, set_mem_size, adjust_address_1)
2254         (adjust_automodify_address_1, offset_address, widen_memory_access):
2255         Likewise.  Take poly_int64s rather than HOST_WIDE_INT.
2257 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2258             Alan Hayward  <alan.hayward@arm.com>
2259             David Sherwood  <david.sherwood@arm.com>
2261         * rtlanal.c (rtx_addr_can_trap_p_1): Take the offset and size
2262         as poly_int64s rather than HOST_WIDE_INTs.  Use a size of -1
2263         rather than 0 to represent an unknown size.  Assert that the size
2264         is known when the mode isn't BLKmode.
2265         (may_trap_p_1): Use -1 for unknown sizes.
2266         (rtx_addr_can_trap_p): Likewise.  Pass BLKmode rather than VOIDmode.
2268 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2269             Alan Hayward  <alan.hayward@arm.com>
2270             David Sherwood  <david.sherwood@arm.com>
2272         * dse.c (store_info): Change offset and width from HOST_WIDE_INT
2273         to poly_int64.  Update commentary for positions_needed.large.
2274         (read_info_type): Change offset and width from HOST_WIDE_INT
2275         to poly_int64.
2276         (set_usage_bits): Likewise.
2277         (canon_address): Return the offset as a poly_int64 rather than
2278         a HOST_WIDE_INT.  Use strip_offset_and_add.
2279         (set_all_positions_unneeded, any_positions_needed_p): Use
2280         positions_needed.large to track stores with non-constant widths.
2281         (all_positions_needed_p): Likewise.  Take the offset and width
2282         as poly_int64s rather than ints.  Assert that rhs is nonnull.
2283         (record_store): Cope with non-constant offsets and widths.
2284         Nullify the rhs of an earlier store if we can't tell which bytes
2285         of it are needed.
2286         (find_shift_sequence): Take the access_size and shift as poly_int64s
2287         rather than ints.
2288         (get_stored_val): Take the read_offset and read_width as poly_int64s
2289         rather than HOST_WIDE_INTs.
2290         (check_mem_read_rtx, scan_stores, scan_reads, dse_step5): Handle
2291         non-constant offsets and widths.
2293 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2294             Alan Hayward  <alan.hayward@arm.com>
2295             David Sherwood  <david.sherwood@arm.com>
2297         * inchash.h (inchash::hash::add_poly_int): New function.
2298         * tree-ssa-alias.h (ao_ref::offset, ao_ref::size, ao_ref::max_size):
2299         Use poly_int64 rather than HOST_WIDE_INT.
2300         (ao_ref::max_size_known_p): New function.
2301         * tree-ssa-sccvn.h (vn_reference_op_struct::off): Use poly_int64_pod
2302         rather than HOST_WIDE_INT.
2303         * tree-ssa-alias.c (ao_ref_base): Apply get_ref_base_and_extent
2304         to temporaries until its interface is adjusted to match.
2305         (ao_ref_init_from_ptr_and_size): Handle polynomial offsets and sizes.
2306         (aliasing_component_refs_p, decl_refs_may_alias_p)
2307         (indirect_ref_may_alias_decl_p, indirect_refs_may_alias_p): Take
2308         the offsets and max_sizes as poly_int64s instead of HOST_WIDE_INTs.
2309         (refs_may_alias_p_1, stmt_kills_ref_p): Adjust for changes to
2310         ao_ref fields.
2311         * alias.c (ao_ref_from_mem): Likewise.
2312         * tree-ssa-dce.c (mark_aliased_reaching_defs_necessary_1): Likewise.
2313         * tree-ssa-dse.c (valid_ao_ref_for_dse, normalize_ref)
2314         (clear_bytes_written_by, setup_live_bytes_from_ref, compute_trims)
2315         (maybe_trim_complex_store, maybe_trim_constructor_store)
2316         (live_bytes_read, dse_classify_store): Likewise.
2317         * tree-ssa-sccvn.c (vn_reference_compute_hash, vn_reference_eq):
2318         (copy_reference_ops_from_ref, ao_ref_init_from_vn_reference)
2319         (fully_constant_vn_reference_p, valueize_refs_1): Likewise.
2320         (vn_reference_lookup_3): Likewise.
2321         * tree-ssa-uninit.c (warn_uninitialized_vars): Likewise.
2323 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2324             Alan Hayward  <alan.hayward@arm.com>
2325             David Sherwood  <david.sherwood@arm.com>
2327         * tree-ssa-alias.c (indirect_ref_may_alias_decl_p)
2328         (indirect_refs_may_alias_p): Use ranges_may_overlap_p
2329         instead of ranges_overlap_p.
2331 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2332             Alan Hayward  <alan.hayward@arm.com>
2333             David Sherwood  <david.sherwood@arm.com>
2335         * tree-ssa-alias.c (same_addr_size_stores_p): Take the offsets and
2336         sizes as poly_int64s rather than HOST_WIDE_INTs.
2338 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2339             Alan Hayward  <alan.hayward@arm.com>
2340             David Sherwood  <david.sherwood@arm.com>
2342         * gimple-fold.h (fold_ctor_reference): Take the offset and size
2343         as poly_uint64 rather than unsigned HOST_WIDE_INT.
2344         * gimple-fold.c (fold_ctor_reference): Likewise.
2346 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2347             Alan Hayward  <alan.hayward@arm.com>
2348             David Sherwood  <david.sherwood@arm.com>
2350         * target.def (dwarf_poly_indeterminate_value): New hook.
2351         * targhooks.h (default_dwarf_poly_indeterminate_value): Declare.
2352         * targhooks.c (default_dwarf_poly_indeterminate_value): New function.
2353         * doc/tm.texi.in (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Document.
2354         * doc/tm.texi: Regenerate.
2355         * dwarf2out.h (build_cfa_loc, build_cfa_aligned_loc): Take the
2356         offset as a poly_int64.
2357         * dwarf2out.c (new_reg_loc_descr): Move later in file.  Take the
2358         offset as a poly_int64.
2359         (loc_descr_plus_const, loc_list_plus_const, build_cfa_aligned_loc):
2360         Take the offset as a poly_int64.
2361         (build_cfa_loc): Likewise.  Use loc_descr_plus_const.
2362         (frame_pointer_fb_offset): Change to a poly_int64.
2363         (int_loc_descriptor): Take the offset as a poly_int64.  Use
2364         targetm.dwarf_poly_indeterminate_value for polynomial offsets.
2365         (based_loc_descr): Take the offset as a poly_int64.
2366         Use strip_offset_and_add to handle (plus X (const)).
2367         Use new_reg_loc_descr instead of an open-coded version of the
2368         previous implementation.
2369         (mem_loc_descriptor): Handle CONST_POLY_INT.
2370         (compute_frame_pointer_to_fb_displacement): Take the offset as a
2371         poly_int64.  Use strip_offset_and_add to handle (plus X (const)).
2373 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2374             Alan Hayward  <alan.hayward@arm.com>
2375             David Sherwood  <david.sherwood@arm.com>
2377         * rtl.h (reg_attrs::offset): Change from HOST_WIDE_INT to poly_int64.
2378         (gen_rtx_REG_offset): Take the offset as a poly_int64.
2379         * inchash.h (inchash::hash::add_poly_hwi): New function.
2380         * gengtype.c (main): Register poly_int64.
2381         * emit-rtl.c (reg_attr_hasher::hash): Use inchash.  Treat the
2382         offset as a poly_int.
2383         (reg_attr_hasher::equal): Use must_eq to compare offsets.
2384         (get_reg_attrs, update_reg_offset, gen_rtx_REG_offset): Take the
2385         offset as a poly_int64.
2386         (set_reg_attrs_from_value): Treat the offset as a poly_int64.
2387         * print-rtl.c (print_poly_int): New function.
2388         (rtx_writer::print_rtx_operand_code_r): Treat REG_OFFSET as
2389         a poly_int.
2390         * var-tracking.c (track_offset_p, get_tracked_reg_offset): New
2391         functions.
2392         (var_reg_set, var_reg_delete_and_set, var_reg_delete): Use them.
2393         (same_variable_part_p, track_loc_p): Take the offset as a poly_int64.
2394         (vt_get_decl_and_offset): Return the offset as a poly_int64.
2395         Enforce track_offset_p for parts of a PARALLEL.
2396         (vt_add_function_parameter): Use const_offset for the final
2397         offset to track.  Use get_tracked_reg_offset for the parts
2398         of a PARALLEL.
2400 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2401             Alan Hayward  <alan.hayward@arm.com>
2402             David Sherwood  <david.sherwood@arm.com>
2404         * target.def (truly_noop_truncation): Take poly_uint64s instead of
2405         unsigned ints.  Change default to hook_bool_puint64_puint64_true.
2406         * doc/tm.texi: Regenerate.
2407         * hooks.h (hook_bool_uint_uint_true): Delete.
2408         (hook_bool_puint64_puint64_true): Declare.
2409         * hooks.c (hook_bool_uint_uint_true): Delete.
2410         (hook_bool_puint64_puint64_true): New function.
2411         * config/mips/mips.c (mips_truly_noop_truncation): Take poly_uint64s
2412         instead of unsigned ints.
2413         * config/spu/spu.c (spu_truly_noop_truncation): Likewise.
2414         * config/tilegx/tilegx.c (tilegx_truly_noop_truncation): Likewise.
2416 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2417             Alan Hayward  <alan.hayward@arm.com>
2418             David Sherwood  <david.sherwood@arm.com>
2420         * optabs.h (expand_operand): Add an int_value field.
2421         (create_expand_operand): Add an int_value parameter and use it
2422         to initialize the new expand_operand field.
2423         (create_integer_operand): Replace with a declaration of a function
2424         that accepts poly_int64s.  Move the implementation to...
2425         * optabs.c (create_integer_operand): ...here.
2426         (maybe_legitimize_operand): For EXPAND_INTEGER, check whether
2427         the mode preserves the value of int_value, instead of calling
2428         const_int_operand on the rtx.  Use gen_int_mode to generate
2429         the new rtx.
2431 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2432             Alan Hayward  <alan.hayward@arm.com>
2433             David Sherwood  <david.sherwood@arm.com>
2435         * dumpfile.h (dump_dec): Declare.
2436         * dumpfile.c (dump_dec): New function.
2437         * pretty-print.h (pp_wide_integer): Turn into a function and
2438         declare a poly_int version.
2439         * pretty-print.c (pp_wide_integer): New function for poly_ints.
2441 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2442             Alan Hayward  <alan.hayward@arm.com>
2443             David Sherwood  <david.sherwood@arm.com>
2445         * doc/generic.texi (POLY_INT_CST): Document.
2446         * tree.def (POLY_INT_CST): New tree code.
2447         * treestruct.def (TS_POLY_INT_CST): New tree layout.
2448         * tree-core.h (tree_poly_int_cst): New struct.
2449         (tree_node): Add a poly_int_cst field.
2450         * tree.h (POLY_INT_CST_P, POLY_INT_CST_COEFF): New macros.
2451         (wide_int_to_tree, force_fit_type): Take a poly_wide_int_ref
2452         instead of a wide_int_ref.
2453         (build_int_cst, build_int_cst_type): Take a poly_int64 instead
2454         of a HOST_WIDE_INT.
2455         (build_int_cstu, build_array_type_nelts): Take a poly_uint64
2456         instead of an unsigned HOST_WIDE_INT.
2457         (build_poly_int_cst, tree_fits_poly_int64_p, tree_fits_poly_uint64_p)
2458         (ptrdiff_tree_p): Declare.
2459         (tree_to_poly_int64, tree_to_poly_uint64): Likewise.  Provide
2460         extern inline implementations if the target doesn't use POLY_INT_CST.
2461         (poly_int_tree_p): New function.
2462         (wi::unextended_tree): New class.
2463         (wi::int_traits <unextended_tree>): New override.
2464         (wi::extended_tree): Add a default constructor.
2465         (wi::extended_tree::get_tree): New function.
2466         (wi::widest_extended_tree, wi::offset_extended_tree): New typedefs.
2467         (wi::tree_to_widest_ref, wi::tree_to_offset_ref): Use them.
2468         (wi::tree_to_poly_widest_ref, wi::tree_to_poly_offset_ref)
2469         (wi::tree_to_poly_wide_ref): New typedefs.
2470         (wi::ints_for): Provide overloads for extended_tree and
2471         unextended_tree.
2472         (poly_int_cst_value, wi::to_poly_widest, wi::to_poly_offset)
2473         (wi::to_wide): New functions.
2474         (wi::fits_to_boolean_p, wi::fits_to_tree_p): Handle poly_ints.
2475         * tree.c (poly_int_cst_hasher): New struct.
2476         (poly_int_cst_hash_table): New variable.
2477         (tree_node_structure_for_code, tree_code_size, simple_cst_equal)
2478         (valid_constant_size_p, add_expr, drop_tree_overflow): Handle
2479         POLY_INT_CST.
2480         (initialize_tree_contains_struct): Handle TS_POLY_INT_CST.
2481         (init_ttree): Initialize poly_int_cst_hash_table.
2482         (build_int_cst, build_int_cst_type, build_invariant_address): Take
2483         a poly_int64 instead of a HOST_WIDE_INT.
2484         (build_int_cstu, build_array_type_nelts): Take a poly_uint64
2485         instead of an unsigned HOST_WIDE_INT.
2486         (wide_int_to_tree): Rename to...
2487         (wide_int_to_tree_1): ...this.
2488         (build_new_poly_int_cst, build_poly_int_cst): New functions.
2489         (force_fit_type): Take a poly_wide_int_ref instead of a wide_int_ref.
2490         (wide_int_to_tree): New function that takes a poly_wide_int_ref.
2491         (ptrdiff_tree_p, tree_to_poly_int64, tree_to_poly_uint64)
2492         (tree_fits_poly_int64_p, tree_fits_poly_uint64_p): New functions.
2493         * lto-streamer-out.c (DFS::DFS_write_tree_body, hash_tree): Handle
2494         TS_POLY_INT_CST.
2495         * tree-streamer-in.c (lto_input_ts_poly_tree_pointers): Likewise.
2496         (streamer_read_tree_body): Likewise.
2497         * tree-streamer-out.c (write_ts_poly_tree_pointers): Likewise.
2498         (streamer_write_tree_body): Likewise.
2499         * tree-streamer.c (streamer_check_handled_ts_structures): Likewise.
2500         * asan.c (asan_protect_global): Require the size to be an INTEGER_CST.
2501         * cfgexpand.c (expand_debug_expr): Handle POLY_INT_CST.
2502         * expr.c (expand_expr_real_1, const_vector_from_tree): Likewise.
2503         * gimple-expr.h (is_gimple_constant): Likewise.
2504         * gimplify.c (maybe_with_size_expr): Likewise.
2505         * print-tree.c (print_node): Likewise.
2506         * tree-data-ref.c (data_ref_compare_tree): Likewise.
2507         * tree-pretty-print.c (dump_generic_node): Likewise.
2508         * tree-ssa-address.c (addr_for_mem_ref): Likewise.
2509         * tree-vect-data-refs.c (dr_group_sort_cmp): Likewise.
2510         * tree-vrp.c (compare_values_warnv): Likewise.
2511         * tree-ssa-loop-ivopts.c (determine_base_object, constant_multiple_of)
2512         (get_loop_invariant_expr, add_candidate_1, get_computation_aff_1)
2513         (force_expr_to_var_cost): Likewise.
2514         * tree-ssa-loop.c (for_each_index): Likewise.
2515         * fold-const.h (build_invariant_address, size_int_kind): Take a
2516         poly_int64 instead of a HOST_WIDE_INT.
2517         * fold-const.c (fold_negate_expr_1, const_binop, const_unop)
2518         (fold_convert_const, multiple_of_p, fold_negate_const): Handle
2519         POLY_INT_CST.
2520         (size_binop_loc): Likewise.  Allow int_const_binop_1 to fail.
2521         (int_const_binop_2): New function, split out from...
2522         (int_const_binop_1): ...here.  Handle POLY_INT_CST.
2523         (size_int_kind): Take a poly_int64 instead of a HOST_WIDE_INT.
2524         * expmed.c (make_tree): Handle CONST_POLY_INT_P.
2525         * gimple-ssa-strength-reduction.c (slsr_process_add)
2526         (slsr_process_mul): Check for INTEGER_CSTs before using them
2527         as candidates.
2528         * stor-layout.c (bits_from_bytes): New function.
2529         (bit_from_pos): Use it.
2530         (layout_type): Likewise.  For vectors, multiply the TYPE_SIZE_UNIT
2531         by BITS_PER_UNIT to get the TYPE_SIZE.
2532         * tree-cfg.c (verify_expr, verify_types_in_gimple_reference): Allow
2533         MEM_REF and TARGET_MEM_REF offsets to be a POLY_INT_CST.
2535 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2536             Alan Hayward  <alan.hayward@arm.com>
2537             David Sherwood  <david.sherwood@arm.com>
2539         * doc/rtl.texi (const_poly_int): Document.  Also document the
2540         rtl sharing behavior.
2541         * gengenrtl.c (excluded_rtx): Return true for CONST_POLY_INT.
2542         * rtl.h (const_poly_int_def): New struct.
2543         (rtx_def::u): Add a cpi field.
2544         (CASE_CONST_UNIQUE, CASE_CONST_ANY): Add CONST_POLY_INT.
2545         (CONST_POLY_INT_P, CONST_POLY_INT_COEFFS): New macros.
2546         (wi::rtx_to_poly_wide_ref): New typedef
2547         (const_poly_int_value, wi::to_poly_wide, rtx_to_poly_int64)
2548         (poly_int_rtx_p): New functions.
2549         (trunc_int_for_mode): Declare a poly_int64 version.
2550         (plus_constant): Take a poly_int64 instead of a HOST_WIDE_INT.
2551         (immed_wide_int_const): Take a poly_wide_int_ref rather than
2552         a wide_int_ref.
2553         (strip_offset): Declare.
2554         (strip_offset_and_add): New function.
2555         * rtl.def (CONST_POLY_INT): New rtx code.
2556         * rtl.c (rtx_size): Handle CONST_POLY_INT.
2557         (shared_const_p): Use poly_int_rtx_p.
2558         * emit-rtl.h (gen_int_mode): Take a poly_int64 instead of a
2559         HOST_WIDE_INT.
2560         (gen_int_shift_amount): Likewise.
2561         * emit-rtl.c (const_poly_int_hasher): New class.
2562         (const_poly_int_htab): New variable.
2563         (init_emit_once): Initialize it when NUM_POLY_INT_COEFFS > 1.
2564         (const_poly_int_hasher::hash): New function.
2565         (const_poly_int_hasher::equal): Likewise.
2566         (gen_int_mode): Take a poly_int64 instead of a HOST_WIDE_INT.
2567         (immed_wide_int_const): Rename to...
2568         (immed_wide_int_const_1): ...this and make static.
2569         (immed_wide_int_const): New function, taking a poly_wide_int_ref
2570         instead of a wide_int_ref.
2571         (gen_int_shift_amount): Take a poly_int64 instead of a HOST_WIDE_INT.
2572         (gen_lowpart_common): Handle CONST_POLY_INT.
2573         * cse.c (hash_rtx_cb, equiv_constant): Likewise.
2574         * cselib.c (cselib_hash_rtx): Likewise.
2575         * dwarf2out.c (const_ok_for_output_1): Likewise.
2576         * expr.c (convert_modes): Likewise.
2577         * print-rtl.c (rtx_writer::print_rtx, print_value): Likewise.
2578         * rtlhash.c (add_rtx): Likewise.
2579         * explow.c (trunc_int_for_mode): Add a poly_int64 version.
2580         (plus_constant): Take a poly_int64 instead of a HOST_WIDE_INT.
2581         Handle existing CONST_POLY_INT rtxes.
2582         * expmed.h (expand_shift): Take a poly_int64 instead of a
2583         HOST_WIDE_INT.
2584         * expmed.c (expand_shift): Likewise.
2585         * rtlanal.c (strip_offset): New function.
2586         (commutative_operand_precedence): Give CONST_POLY_INT the same
2587         precedence as CONST_DOUBLE and put CONST_WIDE_INT between that
2588         and CONST_INT.
2589         * rtl-tests.c (const_poly_int_tests): New struct.
2590         (rtl_tests_c_tests): Use it.
2591         * simplify-rtx.c (simplify_const_unary_operation): Handle
2592         CONST_POLY_INT.
2593         (simplify_const_binary_operation): Likewise.
2594         (simplify_binary_operation_1): Fold additions of symbolic constants
2595         and CONST_POLY_INTs.
2596         (simplify_subreg): Handle extensions and truncations of
2597         CONST_POLY_INTs.
2598         (simplify_const_poly_int_tests): New struct.
2599         (simplify_rtx_c_tests): Use it.
2600         * wide-int.h (storage_ref): Add default constructor.
2601         (wide_int_ref_storage): Likewise.
2602         (trailing_wide_ints): Use GTY((user)).
2603         (trailing_wide_ints::operator[]): Add a const version.
2604         (trailing_wide_ints::get_precision): New function.
2605         (trailing_wide_ints::extra_size): Likewise.
2607 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2608             Alan Hayward  <alan.hayward@arm.com>
2609             David Sherwood  <david.sherwood@arm.com>
2611         * emit-rtl.h (gen_int_shift_amount): Declare.
2612         * emit-rtl.c (gen_int_shift_amount): New function.
2613         * asan.c (asan_emit_stack_protection): Use gen_int_shift_amount
2614         instead of GEN_INT.
2615         * calls.c (shift_return_value): Likewise.
2616         * cse.c (fold_rtx): Likewise.
2617         * dse.c (find_shift_sequence): Likewise.
2618         * expmed.c (init_expmed_one_mode, store_bit_field_1, expand_shift_1)
2619         (expand_shift, expand_smod_pow2): Likewise.
2620         * lower-subreg.c (shift_cost): Likewise.
2621         * optabs.c (expand_superword_shift, expand_doubleword_mult)
2622         (expand_unop, expand_binop, shift_amt_for_vec_perm_mask)
2623         (expand_vec_perm_var): Likewise.
2624         * simplify-rtx.c (simplify_unary_operation_1): Likewise.
2625         (simplify_binary_operation_1): Likewise.
2626         * combine.c (try_combine, find_split_point, force_int_to_mode)
2627         (simplify_shift_const_1, simplify_shift_const): Likewise.
2628         (change_zero_ext): Likewise.  Use simplify_gen_binary.
2630 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2632         * poly-int.h (multiple_p): Fix handling of two non-poly_ints.
2634 2017-12-20  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
2636         * doc/invoke.texi (ARM Options): Document accepted extension options
2637         for -march=armv8.3-a.
2639 2017-12-20  Richard Earnshaw  <rearnsha@arm.com>
2641         PR target/83105
2642         * config.gcc (arm*-*-linux*): When configured with --with-float=hard
2643         or --with-float=softfp, set the default CPU to arm10e.
2645 2017-12-20  Eric Botcazou  <ebotcazou@adacore.com>
2647         * config/visium/constraints.md (J, K, L): Use IN_RANGE macro.
2648         * config/visium/predicates.md (const_shift_operand): Likewise.
2649         * config/visium/visium.c (visium_legitimize_address): Fix oversight.
2650         (visium_legitimize_reload_address): Likewise.
2652 2017-12-20  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
2654         PR target/82975
2655         * config/arm/arm.h (TEST_REGNO): Adjust comment as expected in r255830.
2657 2017-12-20  Jakub Jelinek  <jakub@redhat.com>
2659         PR c++/83490
2660         * calls.c (compute_argument_addresses): Ignore TYPE_EMPTY_P arguments.
2662 2017-12-20  Julia Koval  <julia.koval@intel.com>
2664         * common/config/i386/i386-common.c (OPTION_MASK_ISA_VPCLMULQDQ_SET,
2665         OPTION_MASK_ISA_VPCLMULQDQ_UNSET): New.
2666         (ix86_handle_option): Handle -mvpclmulqdq, move cx6 to flags2.
2667         * config.gcc: Include vpclmulqdqintrin.h.
2668         * config/i386/cpuid.h: Handle bit_VPCLMULQDQ.
2669         * config/i386/driver-i386.c (host_detect_local_cpu): Handle
2670         -mvpclmulqdq.
2671         * config/i386/i386-builtin.def (__builtin_ia32_vpclmulqdq_v2di,
2672         __builtin_ia32_vpclmulqdq_v4di, __builtin_ia32_vpclmulqdq_v8di): New.
2673         * config/i386/i386-c.c (__VPCLMULQDQ__): New.
2674         * config/i386/i386.c (isa2_opts): Add -mcx16.
2675         (isa_opts): Add -mpclmulqdq, remove -mcx16.
2676         (ix86_option_override_internal): Move mcx16 to flags2.
2677         (ix86_valid_target_attribute_inner_p): Add vpclmulqdq.
2678         (ix86_expand_builtin): Handle OPTION_MASK_ISA_VPCLMULQDQ.
2679         * config/i386/i386.h (TARGET_VPCLMULQDQ, TARGET_VPCLMULQDQ_P): New.
2680         * config/i386/i386.opt: Add mvpclmulqdq, move mcx16 to flags2.
2681         * config/i386/immintrin.h: Include vpclmulqdqintrin.h.
2682         * config/i386/sse.md (vpclmulqdq_<mode>): New pattern.
2683         * config/i386/vpclmulqdqintrin.h (_mm512_clmulepi64_epi128,
2684         _mm_clmulepi64_epi128, _mm256_clmulepi64_epi128): New intrinsics.
2685         * doc/invoke.texi: Add -mvpclmulqdq.
2687 2017-12-20  Tom de Vries  <tom@codesourcery.com>
2689         PR middle-end/83423
2690         * config/i386/i386.c (ix86_static_chain): Move
2691         DECL_STATIC_CHAIN test ...
2692         * calls.c (rtx_for_static_chain): ... here.  New function.
2693         * calls.h (rtx_for_static_chain): Declare.
2694         * builtins.c (expand_builtin_setjmp_receiver): Use rtx_for_static_chain
2695         instead of targetm.calls.static_chain.
2696         * df-scan.c (df_get_entry_block_def_set): Same.
2698 2017-12-19  Tom de Vries  <tom@codesourcery.com>
2700         PR tree-optimization/83493
2701         * graphite-isl-ast-to-gimple.c (translate_isl_ast_node_for): Unshare ub
2702         and lb.
2704 2017-12-19  Jakub Jelinek  <jakub@redhat.com>
2706         * gimple-ssa-sprintf.c (format_directive): Use inform_n instead of
2707         inform with hardcoded english plural handling.
2709 2017-12-18  Jeff Law  <law@redhat.com>
2711         PR tree-optimization/83477
2712         * tree-ssa-threadedge.c (record_temporary_equivalences_from_phis): For
2713         a non-virtual PHI, always push a new range.
2715 2017-12-19  Martin Sebor  <msebor@redhat.com>
2717         PR middle-end/77608
2718         * builtins.c (compute_objsize): Handle non-constant offsets.
2720 2017-12-19  Jakub Jelinek  <jakub@redhat.com>
2722         PR tree-optimization/83444
2723         * tree-ssa-strlen.c (strlen_check_and_optimize_stmt): For the
2724         character load case, if get_stridx on MEM_REF's operand doesn't
2725         look usable, retry with get_addr_stridx.
2727 2017-12-19  Alexandre Oliva  <aoliva@redhat.com>
2729         PR debug/83422
2730         * var-tracking.c (vt_debug_insns_local): Do not drop markers.
2731         (variable_tracking_main_1): Keep markers even when VTA fails.
2733         PR bootstrap/83396
2734         * cfgexpand.c (expand_gimple_basic_block): Expand label first,
2735         even if there are markers before it.
2736         * cfgrtl.c (rtl_verify_bb_layout): Reject DEBUG_INSNs outside BBs.
2738 2017-12-19  Jakub Jelinek  <jakub@redhat.com>
2740         * read-rtl.c (parse_reg_note_name): Replace Yoda conditions with
2741         typical order conditions.
2742         * sel-sched.c (extract_new_fences_from): Likewise.
2743         * config/visium/constraints.md (J, K, L): Likewise.
2744         * config/visium/predicates.md (const_shift_operand): Likewise.
2745         * config/visium/visium.c (visium_legitimize_address,
2746         visium_legitimize_reload_address): Likewise.
2747         * config/m68k/m68k.c (output_reg_adjust, emit_reg_adjust): Likewise.
2748         * config/arm/arm.c (arm_block_move_unaligned_straight): Likewise.
2749         * config/avr/constraints.md (Y01, Ym1, Y02, Ym2): Likewise.
2750         * config/avr/avr-log.c (avr_vdump, avr_log_set_avr_log,
2751         SET_DUMP_DETAIL): Likewise.
2752         * config/avr/predicates.md (const_8_16_24_operand): Likewise.
2753         * config/avr/avr.c (STR_PREFIX_P, avr_popcount_each_byte,
2754         avr_is_casesi_sequence, avr_casei_sequence_check_operands,
2755         avr_set_core_architecture, avr_set_current_function,
2756         avr_legitimize_reload_address, avr_asm_len, avr_print_operand,
2757         output_movqi, output_movsisf, avr_out_plus, avr_out_bitop,
2758         avr_out_fract, avr_adjust_insn_length, avr_encode_section_info,
2759         avr_2word_insn_p, output_reload_in_const, avr_has_nibble_0xf,
2760         avr_map_decompose, avr_fold_builtin): Likewise.
2761         * config/avr/driver-avr.c (avr_devicespecs_file): Likewise.
2762         * config/avr/gen-avr-mmcu-specs.c (str_prefix_p, print_mcu): Likewise.
2763         * config/i386/i386.c (ix86_parse_stringop_strategy_string): Likewise.
2764         * config/m32c/m32c-pragma.c (m32c_pragma_memregs): Likewise.
2765         * config/m32c/m32c.c (m32c_conditional_register_usage,
2766         m32c_address_cost): Likewise.
2767         * config/m32c/predicates.md (shiftcount_operand,
2768         longshiftcount_operand): Likewise.
2769         * config/iq2000/iq2000.c (iq2000_expand_prologue): Likewise.
2770         * config/nios2/nios2.c (nios2_handle_custom_fpu_insn_option,
2771         can_use_cdx_ldstw): Likewise.
2772         * config/nios2/nios2.h (CDX_REG_P): Likewise.
2773         * config/cr16/cr16.h (RETURN_ADDR_RTX, REGNO_MODE_OK_FOR_BASE_P):
2774         Likewise.
2775         * config/cr16/cr16.md (*mov<mode>_double): Likewise.
2776         * config/cr16/cr16.c (cr16_create_dwarf_for_multi_push): Likewise.
2777         * config/h8300/h8300.c (h8300_rtx_costs, get_shift_alg): Likewise.
2778         * config/vax/constraints.md (U06, U08, U16, CN6, S08, S16): Likewise.
2779         * config/vax/vax.c (adjacent_operands_p): Likewise.
2780         * config/ft32/constraints.md (L, b, KA): Likewise.
2781         * config/ft32/ft32.c (ft32_load_immediate, ft32_expand_prologue):
2782         Likewise.
2783         * cfgexpand.c (expand_stack_alignment): Likewise.
2784         * gcse.c (insert_expr_in_table): Likewise.
2785         * print-rtl.c (rtx_writer::print_rtx_operand_codes_E_and_V): Likewise.
2786         * cgraphunit.c (cgraph_node::expand): Likewise.
2787         * ira-build.c (setup_min_max_allocno_live_range_point): Likewise.
2788         * emit-rtl.c (add_insn): Likewise.
2789         * input.c (dump_location_info): Likewise.
2790         * passes.c (NEXT_PASS): Likewise.
2791         * read-rtl-function.c (parse_note_insn_name,
2792         function_reader::read_rtx_operand_r, function_reader::parse_mem_expr):
2793         Likewise.
2794         * sched-rgn.c (sched_rgn_init): Likewise.
2795         * diagnostic-show-locus.c (layout::show_ruler): Likewise.
2796         * combine.c (find_split_point, simplify_if_then_else, force_to_mode,
2797         if_then_else_cond, simplify_shift_const_1, simplify_comparison):
2798         Likewise.
2799         * explow.c (eliminate_constant_term): Likewise.
2800         * final.c (leaf_renumber_regs_insn): Likewise.
2801         * cfgrtl.c (print_rtl_with_bb): Likewise.
2802         * genhooks.c (emit_init_macros): Likewise.
2803         * poly-int.h (maybe_ne, maybe_le, maybe_lt): Likewise.
2804         * tree-data-ref.c (conflict_fn): Likewise.
2805         * selftest.c (assert_streq): Likewise.
2806         * expr.c (store_constructor_field, expand_expr_real_1): Likewise.
2807         * fold-const.c (fold_range_test, extract_muldiv_1, fold_truth_andor,
2808         fold_binary_loc, multiple_of_p): Likewise.
2809         * reload.c (push_reload, find_equiv_reg): Likewise.
2810         * et-forest.c (et_nca, et_below): Likewise.
2811         * dbxout.c (dbxout_symbol_location): Likewise.
2812         * reorg.c (relax_delay_slots): Likewise.
2813         * dojump.c (do_compare_rtx_and_jump): Likewise.
2814         * gengtype-parse.c (type): Likewise.
2815         * simplify-rtx.c (simplify_gen_ternary, simplify_gen_relational,
2816         simplify_const_relational_operation): Likewise.
2817         * reload1.c (do_output_reload): Likewise.
2818         * dumpfile.c (get_dump_file_info_by_switch): Likewise.
2819         * gengtype.c (type_for_name): Likewise.
2820         * gimple-ssa-sprintf.c (format_directive): Likewise.
2822 2017-12-19  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
2824         PR target/82975
2825         * config/arm/arm.h (TEST_REGNO): Check reg_renumber is set before
2826         accessing it.  Adjust comment.
2828 2017-12-19  Jakub Jelinek  <jakub@redhat.com>
2830         PR middle-end/81914
2831         * predict.c (zero_one_minusone): New function.
2832         (apply_return_prediction): Avoid return prediction for functions
2833         returning only -1, 0 and 1 values, unless they only return -1 and 0
2834         or 0 and 1.
2836 2017-12-19  Claudiu Zissulescu  <claziss@synopsys.com>
2838         * config/arc/arc.c (legitimate_scaled_address_p): Clean
2839         fall-through warning.
2840         (arc_compute_frame_size): Remove unused variables.
2841         (arc_print_operand): Fix fprintif format.
2842         (arc_can_follow_jump): Clean fall-through warning.
2844 2017-12-19  Marek Polacek  <polacek@redhat.com>
2846         PR c++/83489
2847         * config/i386/i386.c (init_cumulative_args): Don't check TYPE_EMPTY_P
2848         on an error node.
2850 2017-12-19  Claudiu Zissulescu  <claziss@synopsys.com>
2852         * config/arc/arc.c (overriderregs): New variable.
2853         (arc_override_options): Track fixed/call saved/call options.
2854         (arc_conditional_register_usage): Check against overrideregs
2855         variable whenever we change register properties.
2857 2017-12-19  Nathan Sidwell  <nathan@acm.org>
2859         * opts.c (finish_options): Don't prefix dump_base_name if it
2860         already contains directories.
2862 2017-12-19  Martin Liska  <mliska@suse.cz>
2864         PR rtl-optimization/82675
2865         * loop-unroll.c (unroll_loop_constant_iterations): Allocate one
2866         more element in sbitmap.
2868 2017-12-19  Paolo Carlini  <paolo.carlini@oracle.com>
2870         * gimplify.c (gimplify_expr): Use error_operand_p.
2872 2017-12-19  Sebastian Huber  <sebastian.huber@embedded-brains.de>
2874         PR target/83387
2875         * config/rs6000/t-rtems (MULTILIB_REQUIRED): Remove 64-bit soft-float
2876         multilib.
2878 2017-12-19  Daniel Cederman  <cederman@gaisler.com>
2880         * config/sparc/sparc.c (sparc_do_work_around_errata): Make sure
2881         the jump is to a label.
2883 2017-12-19  Jakub Jelinek  <jakub@redhat.com>
2885         PR tree-optimization/83444
2886         * tree-ssa-strlen.c (strlen_check_and_optimize_stmt): Optimize
2887         character loads.
2889         PR ipa/82801
2890         PR ipa/83346
2891         * ipa-inline.c (flatten_remove_node_hook): New function.
2892         (ipa_inline): Keep only nodes with flatten attribute at the end of
2893         the array in the order from ipa_reverse_postorder, only walk that
2894         portion of array for flattening, if there is more than one such
2895         node, temporarily register a removal hook and ignore removed nodes.
2897         PR tree-optimization/80631
2898         * tree-vect-loop.c (vect_create_epilog_for_reduction): Compare
2899         induc_code against MAX_EXPR or MIN_EXPR instead of reduc_fn against
2900         IFN_REDUC_MAX or IFN_REDUC_MIN.
2902 2017-12-18  Jeff Law  <law@redhat.com>
2904         * tree-ssa-dom.c (record_equivalences_from_phis): Do not
2905         record symbolic equivalences from backedges in the CFG.
2907         Revert
2908         2017-11-19  Jeff Law  <law@redhat.com>
2910         * tree-ssa-dom.c (record_equivalences_from_phis): Fix handling
2911         of degenerates resulting from ignoring an edge.
2913 2017-12-18  Martin Sebor  <msebor@redhat.com>
2915         PR middle-end/83373
2916         PR tree-optimization/78450
2917         * tree-ssa-strlen.c (maybe_set_strlen_range): New function.
2918         (handle_builtin_strlen): Call it.
2920 2017-12-18  Segher Boessenkool  <segher@kernel.crashing.org>
2922         PR rtl-optimization/83424
2923         * rtlanal.c (dead_or_set_regno_p): Handle CLOBBER just like SET.
2925 2017-12-18  Marek Polacek  <polacek@redhat.com>
2927         PR middle-end/83463
2928         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref):
2929         Check if TYPE is INTEGRAL_TYPE_P before accessing its min/max
2930         values.
2932 2017-12-18  Claudiu Zissulescu  <claziss@synopsys.com>
2934         * config/arc/arc.md (maddsidi4, maddsidi4_split): Update pattern.
2935         (umaddsidi4, umaddsidi_split): Likewise.
2937 2017-12-18  Claudiu Zissulescu  <claziss@synopsys.com>
2939         * config/arc/arc.c (arc_legitimate_constant_p): Always check all
2940         constants.
2942 2017-12-18  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
2944         PR target/83420
2945         * config/s390/s390.c (s390_option_override): Avoid strncpy.
2947 2017-12-18  Richard Biener  <rguenther@suse.de>
2949         PR tree-optimization/81877
2950         * tree-ssa-loop-im.c (ref_indep_loop_p): Remove safelen parameters.
2951         (outermost_indep_loop): Adjust.
2952         (ref_indep_loop_p_1): Likewise.  Remove safelen handling again.
2953         (can_sm_ref_p): Adjust.
2955 2017-12-18  Richard Biener  <rguenther@suse.de>
2957         PR middle-end/77291
2958         * tree.c (array_at_struct_end_p): Return true if the underlying
2959         object has space for at least one element in excess of what
2960         the array domain specifies.
2962 2017-12-17  Sandra Loosemore  <sandra@codesourcery.com>
2964         * doc/extend.texi (x86 Function Attributes): Reformat nocf_check
2965         example to avoid overfull hbox.
2966         * doc/invoke.texi (Option Summary): Add missing @gol.
2967         (C++ Dialect Options): Reformat -Wnoexcept-type example to avoid
2968         overfull hbox.
2970 2017-12-17  Sandra Loosemore  <sandra@codesourcery.com>
2971             Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
2973         * doc/invoke.texi (Option Summary): Add -mverbose-cost-dump
2974         to AArch64 and ARM lists, plus missing -mflip-thumb for ARM.
2975         (AArch64 Options): Document -mverbose-cost-dump.
2976         (ARM Options): Likewise, plus -mflip-thumb.
2978 2017-12-17  Martin Sebor  <msebor@redhat.com>
2980         PR bootstrap/83446
2981         * gimple-ssa-warn-restrict.c
2982         (builtin_memref::offset_out_of_bounds): Correct the handling of
2983         anti-ranges.
2985 2017-12-17  Sandra Loosemore  <sandra@codesourcery.com>
2987         * doc/invoke.texi (ARC Options): Add missing -mlra entry.
2989 2017-12-17  John David Anglin  <danglin@gcc.gnu.org>
2991         * config/pa/pa.c (pa_som_asm_init_sections): Fix comment.
2993 2017-12-17  Markus Trippelsdorf  <markus@trippelsdorf.de>
2995         * x86-tune-costs.h (skylake_cost, core_cost): Decrease r64 multiply
2996         latencies.
2998 2017-12-16  Sandra Loosemore  <sandra@codesourcery.com>
3000         * doc/invoke.texi: Fix some typos.
3002 2017-12-16  Martin Sebor  <msebor@redhat.com>
3004         PR tree-optimization/78918
3005         * Makefile.in (OBJS): Add gimple-ssa-warn-restrict.o.
3006         * builtins.c (check_sizes): Rename...
3007         (check_access): ...to this.  Rename function arguments for clarity.
3008         (check_memop_sizes): Adjust names.
3009         (expand_builtin_memchr, expand_builtin_memcpy): Same.
3010         (expand_builtin_memmove, expand_builtin_mempcpy): Same.
3011         (expand_builtin_strcat, expand_builtin_stpncpy): Same.
3012         (check_strncat_sizes, expand_builtin_strncat): Same.
3013         (expand_builtin_strncpy, expand_builtin_memset): Same.
3014         (expand_builtin_bzero, expand_builtin_memcmp): Same.
3015         (expand_builtin_memory_chk, maybe_emit_chk_warning): Same.
3016         (maybe_emit_sprintf_chk_warning): Same.
3017         (expand_builtin_strcpy): Adjust.
3018         (expand_builtin_stpcpy): Same.
3019         (expand_builtin_with_bounds): Detect out-of-bounds accesses
3020         in pointer-checking forms of memcpy, memmove, and mempcpy.
3021         (gcall_to_tree_minimal, max_object_size): Define new functions.
3022         * builtins.h (max_object_size): Declare.
3023         * calls.c (alloc_max_size): Call max_object_size instead of
3024         hardcoding ssizetype limit.
3025         (get_size_range): Handle new argument.
3026         * calls.h (get_size_range): Add a new argument.
3027         * cfgexpand.c (expand_call_stmt): Propagate no-warning bit.
3028         * doc/invoke.texi (-Wrestrict): Adjust, add example.
3029         * gimple-fold.c (gimple_fold_builtin_memory_op): Detect overlapping
3030         operations.
3031         (gimple_fold_builtin_memory_chk): Same.
3032         (gimple_fold_builtin_stxcpy_chk): New function.
3033         * gimple-ssa-warn-restrict.c: New source.
3034         * gimple-ssa-warn-restrict.h: New header.
3035         * gimple.c (gimple_build_call_from_tree): Propagate location.
3036         * passes.def (pass_warn_restrict): Add new pass.
3037         * tree-pass.h (make_pass_warn_restrict): Declare.
3038         * tree-ssa-strlen.c (handle_builtin_strcpy): Detect overlapping
3039         operations.
3040         (handle_builtin_strcat): Same.
3041         (strlen_optimize_stmt): Rename...
3042         (strlen_check_and_optimize_stmt): ...to this.  Handle strncat,
3043         stpncpy, strncpy, and their checking forms.
3045 2017-12-16  Jan Hubicka  <hubicka@ucw.cz>
3047         PR rtl-optimization/82849
3048         * modulo-sched.c (sms_schedule): Use get_estimated_loop_iterations_int
3049         and get_max_loop_iterations_int.
3051 2017-12-16  Richard Sandiford  <richard.sandiford@linaro.org>
3052             Alan Hayward  <alan.hayward@arm.com>
3053             David Sherwood  <david.sherwood@arm.com>
3055         * machmode.h (mode_for_size, int_mode_for_size, float_mode_for_size)
3056         (smallest_mode_for_size, smallest_int_mode_for_size): Take the mode
3057         size as a poly_uint64.
3058         (mode_for_vector, mode_for_int_vector): Take the number of vector
3059         elements as a poly_uint64.
3060         * stor-layout.c (mode_for_size, smallest_mode_for_size): Take the mode
3061         size as a poly_uint64.
3062         (mode_for_vector, mode_for_int_vector): Take the number of vector
3063         elements as a poly_uint64.
3065 2017-12-16  Richard Sandiford  <richard.sandiford@linaro.org>
3066             Alan Hayward  <alan.hayward@arm.com>
3067             David Sherwood  <david.sherwood@arm.com>
3069         * machmode.h (MACRO_MODE): New macro.
3070         * addresses.h (base_reg_class, ok_for_base_p_1): Use it.
3071         * targhooks.c (default_libcall_value, default_secondary_reload)
3072         (default_memory_move_cost, default_register_move_cost)
3073         (default_class_max_nregs): Likewise.
3075 2017-12-16  Richard Sandiford  <richard.sandiford@linaro.org>
3076             Alan Hayward  <alan.hayward@arm.com>
3077             David Sherwood  <david.sherwood@arm.com>
3079         * doc/sourcebuild.texi: Document IN_TARGET_CODE.
3080         * genattrtab.c (write_header): Define IN_TARGET_CODE to 1 in the
3081         target C file.
3082         * genautomata.c (main): Likewise.
3083         * genconditions.c (write_header): Likewise.
3084         * genemit.c (main): Likewise.
3085         * genextract.c (print_header): Likewise.
3086         * genopinit.c (main): Likewise.
3087         * genoutput.c (output_prologue): Likewise.
3088         * genpeep.c (main): Likewise.
3089         * genpreds.c (write_insn_preds_c): Likewise.
3090         * genrecog.c (writer_header): Likewise.
3091         * config/aarch64/aarch64-builtins.c (IN_TARGET_CODE): Define.
3092         * config/aarch64/aarch64-c.c (IN_TARGET_CODE): Likewise.
3093         * config/aarch64/aarch64.c (IN_TARGET_CODE): Likewise.
3094         * config/aarch64/cortex-a57-fma-steering.c (IN_TARGET_CODE): Likewise.
3095         * config/aarch64/driver-aarch64.c (IN_TARGET_CODE): Likewise.
3096         * config/alpha/alpha.c (IN_TARGET_CODE): Likewise.
3097         * config/alpha/driver-alpha.c (IN_TARGET_CODE): Likewise.
3098         * config/arc/arc-c.c (IN_TARGET_CODE): Likewise.
3099         * config/arc/arc.c (IN_TARGET_CODE): Likewise.
3100         * config/arc/driver-arc.c (IN_TARGET_CODE): Likewise.
3101         * config/arm/aarch-common.c (IN_TARGET_CODE): Likewise.
3102         * config/arm/arm-builtins.c (IN_TARGET_CODE): Likewise.
3103         * config/arm/arm-c.c (IN_TARGET_CODE): Likewise.
3104         * config/arm/arm.c (IN_TARGET_CODE): Likewise.
3105         * config/arm/driver-arm.c (IN_TARGET_CODE): Likewise.
3106         * config/avr/avr-c.c (IN_TARGET_CODE): Likewise.
3107         * config/avr/avr-devices.c (IN_TARGET_CODE): Likewise.
3108         * config/avr/avr-log.c (IN_TARGET_CODE): Likewise.
3109         * config/avr/avr.c (IN_TARGET_CODE): Likewise.
3110         * config/avr/driver-avr.c (IN_TARGET_CODE): Likewise.
3111         * config/avr/gen-avr-mmcu-specs.c (IN_TARGET_CODE): Likewise.
3112         * config/bfin/bfin.c (IN_TARGET_CODE): Likewise.
3113         * config/c6x/c6x.c (IN_TARGET_CODE): Likewise.
3114         * config/cr16/cr16.c (IN_TARGET_CODE): Likewise.
3115         * config/cris/cris.c (IN_TARGET_CODE): Likewise.
3116         * config/darwin.c (IN_TARGET_CODE): Likewise.
3117         * config/epiphany/epiphany.c (IN_TARGET_CODE): Likewise.
3118         * config/epiphany/mode-switch-use.c (IN_TARGET_CODE): Likewise.
3119         * config/epiphany/resolve-sw-modes.c (IN_TARGET_CODE): Likewise.
3120         * config/fr30/fr30.c (IN_TARGET_CODE): Likewise.
3121         * config/frv/frv.c (IN_TARGET_CODE): Likewise.
3122         * config/ft32/ft32.c (IN_TARGET_CODE): Likewise.
3123         * config/h8300/h8300.c (IN_TARGET_CODE): Likewise.
3124         * config/i386/djgpp.c (IN_TARGET_CODE): Likewise.
3125         * config/i386/driver-i386.c (IN_TARGET_CODE): Likewise.
3126         * config/i386/driver-mingw32.c (IN_TARGET_CODE): Likewise.
3127         * config/i386/host-cygwin.c (IN_TARGET_CODE): Likewise.
3128         * config/i386/host-i386-darwin.c (IN_TARGET_CODE): Likewise.
3129         * config/i386/host-mingw32.c (IN_TARGET_CODE): Likewise.
3130         * config/i386/i386-c.c (IN_TARGET_CODE): Likewise.
3131         * config/i386/i386.c (IN_TARGET_CODE): Likewise.
3132         * config/i386/intelmic-mkoffload.c (IN_TARGET_CODE): Likewise.
3133         * config/i386/msformat-c.c (IN_TARGET_CODE): Likewise.
3134         * config/i386/winnt-cxx.c (IN_TARGET_CODE): Likewise.
3135         * config/i386/winnt-stubs.c (IN_TARGET_CODE): Likewise.
3136         * config/i386/winnt.c (IN_TARGET_CODE): Likewise.
3137         * config/i386/x86-tune-sched-atom.c (IN_TARGET_CODE): Likewise.
3138         * config/i386/x86-tune-sched-bd.c (IN_TARGET_CODE): Likewise.
3139         * config/i386/x86-tune-sched-core.c (IN_TARGET_CODE): Likewise.
3140         * config/i386/x86-tune-sched.c (IN_TARGET_CODE): Likewise.
3141         * config/ia64/ia64-c.c (IN_TARGET_CODE): Likewise.
3142         * config/ia64/ia64.c (IN_TARGET_CODE): Likewise.
3143         * config/iq2000/iq2000.c (IN_TARGET_CODE): Likewise.
3144         * config/lm32/lm32.c (IN_TARGET_CODE): Likewise.
3145         * config/m32c/m32c-pragma.c (IN_TARGET_CODE): Likewise.
3146         * config/m32c/m32c.c (IN_TARGET_CODE): Likewise.
3147         * config/m32r/m32r.c (IN_TARGET_CODE): Likewise.
3148         * config/m68k/m68k.c (IN_TARGET_CODE): Likewise.
3149         * config/mcore/mcore.c (IN_TARGET_CODE): Likewise.
3150         * config/microblaze/microblaze-c.c (IN_TARGET_CODE): Likewise.
3151         * config/microblaze/microblaze.c (IN_TARGET_CODE): Likewise.
3152         * config/mips/driver-native.c (IN_TARGET_CODE): Likewise.
3153         * config/mips/frame-header-opt.c (IN_TARGET_CODE): Likewise.
3154         * config/mips/mips.c (IN_TARGET_CODE): Likewise.
3155         * config/mmix/mmix.c (IN_TARGET_CODE): Likewise.
3156         * config/mn10300/mn10300.c (IN_TARGET_CODE): Likewise.
3157         * config/moxie/moxie.c (IN_TARGET_CODE): Likewise.
3158         * config/msp430/driver-msp430.c (IN_TARGET_CODE): Likewise.
3159         * config/msp430/msp430-c.c (IN_TARGET_CODE): Likewise.
3160         * config/msp430/msp430.c (IN_TARGET_CODE): Likewise.
3161         * config/nds32/nds32-cost.c (IN_TARGET_CODE): Likewise.
3162         * config/nds32/nds32-fp-as-gp.c (IN_TARGET_CODE): Likewise.
3163         * config/nds32/nds32-intrinsic.c (IN_TARGET_CODE): Likewise.
3164         * config/nds32/nds32-isr.c (IN_TARGET_CODE): Likewise.
3165         * config/nds32/nds32-md-auxiliary.c (IN_TARGET_CODE): Likewise.
3166         * config/nds32/nds32-memory-manipulation.c (IN_TARGET_CODE): Likewise.
3167         * config/nds32/nds32-pipelines-auxiliary.c (IN_TARGET_CODE): Likewise.
3168         * config/nds32/nds32-predicates.c (IN_TARGET_CODE): Likewise.
3169         * config/nds32/nds32.c (IN_TARGET_CODE): Likewise.
3170         * config/nios2/nios2.c (IN_TARGET_CODE): Likewise.
3171         * config/nvptx/mkoffload.c (IN_TARGET_CODE): Likewise.
3172         * config/nvptx/nvptx.c (IN_TARGET_CODE): Likewise.
3173         * config/pa/pa.c (IN_TARGET_CODE): Likewise.
3174         * config/pdp11/pdp11.c (IN_TARGET_CODE): Likewise.
3175         * config/powerpcspe/driver-powerpcspe.c (IN_TARGET_CODE): Likewise.
3176         * config/powerpcspe/host-darwin.c (IN_TARGET_CODE): Likewise.
3177         * config/powerpcspe/host-ppc64-darwin.c (IN_TARGET_CODE): Likewise.
3178         * config/powerpcspe/powerpcspe-c.c (IN_TARGET_CODE): Likewise.
3179         * config/powerpcspe/powerpcspe-linux.c (IN_TARGET_CODE): Likewise.
3180         * config/powerpcspe/powerpcspe.c (IN_TARGET_CODE): Likewise.
3181         * config/riscv/riscv-builtins.c (IN_TARGET_CODE): Likewise.
3182         * config/riscv/riscv-c.c (IN_TARGET_CODE): Likewise.
3183         * config/riscv/riscv.c (IN_TARGET_CODE): Likewise.
3184         * config/rl78/rl78-c.c (IN_TARGET_CODE): Likewise.
3185         * config/rl78/rl78.c (IN_TARGET_CODE): Likewise.
3186         * config/rs6000/driver-rs6000.c (IN_TARGET_CODE): Likewise.
3187         * config/rs6000/host-darwin.c (IN_TARGET_CODE): Likewise.
3188         * config/rs6000/host-ppc64-darwin.c (IN_TARGET_CODE): Likewise.
3189         * config/rs6000/rs6000-c.c (IN_TARGET_CODE): Likewise.
3190         * config/rs6000/rs6000-linux.c (IN_TARGET_CODE): Likewise.
3191         * config/rs6000/rs6000-p8swap.c (IN_TARGET_CODE): Likewise.
3192         * config/rs6000/rs6000-string.c (IN_TARGET_CODE): Likewise.
3193         * config/rs6000/rs6000.c (IN_TARGET_CODE): Likewise.
3194         * config/rx/rx.c (IN_TARGET_CODE): Likewise.
3195         * config/s390/driver-native.c (IN_TARGET_CODE): Likewise.
3196         * config/s390/s390-c.c (IN_TARGET_CODE): Likewise.
3197         * config/s390/s390.c (IN_TARGET_CODE): Likewise.
3198         * config/sh/sh-c.c (IN_TARGET_CODE): Likewise.
3199         * config/sh/sh-mem.cc (IN_TARGET_CODE): Likewise.
3200         * config/sh/sh.c (IN_TARGET_CODE): Likewise.
3201         * config/sh/sh_optimize_sett_clrt.cc (IN_TARGET_CODE): Likewise.
3202         * config/sh/sh_treg_combine.cc (IN_TARGET_CODE): Likewise.
3203         * config/sparc/driver-sparc.c (IN_TARGET_CODE): Likewise.
3204         * config/sparc/sparc-c.c (IN_TARGET_CODE): Likewise.
3205         * config/sparc/sparc.c (IN_TARGET_CODE): Likewise.
3206         * config/spu/spu-c.c (IN_TARGET_CODE): Likewise.
3207         * config/spu/spu.c (IN_TARGET_CODE): Likewise.
3208         * config/stormy16/stormy16.c (IN_TARGET_CODE): Likewise.
3209         * config/tilegx/mul-tables.c (IN_TARGET_CODE): Likewise.
3210         * config/tilegx/tilegx-c.c (IN_TARGET_CODE): Likewise.
3211         * config/tilegx/tilegx.c (IN_TARGET_CODE): Likewise.
3212         * config/tilepro/mul-tables.c (IN_TARGET_CODE): Likewise.
3213         * config/tilepro/tilepro-c.c (IN_TARGET_CODE): Likewise.
3214         * config/tilepro/tilepro.c (IN_TARGET_CODE): Likewise.
3215         * config/v850/v850-c.c (IN_TARGET_CODE): Likewise.
3216         * config/v850/v850.c (IN_TARGET_CODE): Likewise.
3217         * config/vax/vax.c (IN_TARGET_CODE): Likewise.
3218         * config/visium/visium.c (IN_TARGET_CODE): Likewise.
3219         * config/vms/vms-c.c (IN_TARGET_CODE): Likewise.
3220         * config/vms/vms-f.c (IN_TARGET_CODE): Likewise.
3221         * config/vms/vms.c (IN_TARGET_CODE): Likewise.
3222         * config/xtensa/xtensa.c (IN_TARGET_CODE): Likewise.
3224 2017-12-16  Richard Sandiford  <richard.sandiford@linaro.org>
3225             Alan Hayward  <alan.hayward@arm.com>
3226             David Sherwood  <david.sherwood@arm.com>
3228         * tree-ssa-sccvn.c (vn_reference_lookup_3): Avoid repeated
3229         checks for MEM_REF.
3231 2017-12-16  Richard Sandiford  <richard.sandiford@linaro.org>
3232             Alan Hayward  <alan.hayward@arm.com>
3233             David Sherwood  <david.sherwood@arm.com>
3235         * doc/generic.texi (VEC_SERIES_EXPR): Document.
3236         * doc/md.texi (vec_series@var{m}): Document.
3237         * tree.def (VEC_SERIES_EXPR): New tree code.
3238         * tree.h (build_vec_series): Declare.
3239         * tree.c (build_vec_series): New function.
3240         * cfgexpand.c (expand_debug_expr): Handle VEC_SERIES_EXPR.
3241         * tree-pretty-print.c (dump_generic_node): Likewise.
3242         * gimple-pretty-print.c (dump_binary_rhs): Likewise.
3243         * tree-inline.c (estimate_operator_cost): Likewise.
3244         * expr.c (expand_expr_real_2): Likewise.
3245         * optabs-tree.c (optab_for_tree_code): Likewise.
3246         * tree-cfg.c (verify_gimple_assign_binary): Likewise.
3247         * fold-const.c (const_binop): Fold VEC_SERIES_EXPRs of constants.
3248         * expmed.c (make_tree): Handle VEC_SERIES.
3249         * optabs.def (vec_series_optab): New optab.
3250         * optabs.h (expand_vec_series_expr): Declare.
3251         * optabs.c (expand_vec_series_expr): New function.
3252         * tree-vect-generic.c (expand_vector_operations_1): Check that
3253         the operands also have vector type.
3255 2017-12-16  Richard Sandiford  <richard.sandiford@linaro.org>
3256             Alan Hayward  <alan.hawyard@arm.com>
3257             David Sherwood  <david.sherwood@arm.com>
3259         * doc/generic.texi (VEC_DUPLICATE_EXPR): Document.
3260         (VEC_COND_EXPR): Add missing @tindex.
3261         * doc/md.texi (vec_duplicate@var{m}): Document.
3262         * tree.def (VEC_DUPLICATE_EXPR): New tree codes.
3263         * tree.c (build_vector_from_val): Add stubbed-out handling of
3264         variable-length vectors, using VEC_DUPLICATE_EXPR.
3265         (uniform_vector_p): Handle VEC_DUPLICATE_EXPR.
3266         * cfgexpand.c (expand_debug_expr): Likewise.
3267         * tree-cfg.c (verify_gimple_assign_unary): Likewise.
3268         * tree-inline.c (estimate_operator_cost): Likewise.
3269         * tree-pretty-print.c (dump_generic_node): Likewise.
3270         * tree-vect-generic.c (ssa_uniform_vector_p): Likewise.
3271         * fold-const.c (const_unop): Fold VEC_DUPLICATE_EXPRs of a constant.
3272         (test_vec_duplicate_folding): New function.
3273         (fold_const_c_tests): Call it.
3274         * optabs.def (vec_duplicate_optab): New optab.
3275         * optabs-tree.c (optab_for_tree_code): Handle VEC_DUPLICATE_EXPR.
3276         * optabs.h (expand_vector_broadcast): Declare.
3277         * optabs.c (expand_vector_broadcast): Make non-static.  Try using
3278         vec_duplicate_optab.
3279         * expr.c (store_constructor): Try using vec_duplicate_optab for
3280         uniform vectors.
3281         (expand_expr_real_2): Handle VEC_DUPLICATE_EXPR.
3283 2017-12-15  Markus Trippelsdorf  <markus@trippelsdorf.de>
3285         PR target/83358
3286         * config/i386/x86-tune-costs.h (skylake_cost, core_cost): Increase
3287         div/mod latencies a bit.
3289 2017-12-15  Jeff Law  <law@redhat.com>
3291         PR tree-optimization/36550
3292         * tree-ssa-threadupdate.c (count_stmts_and_phis_in_block): New.
3293         (mark_threaded_blocks): Rewrite code to avoid block copying when
3294         optimizing for size.  Don't pessimize blocks which will be
3295         copied, but all the statements will be dead.
3297 2017-12-15  Alexandre Oliva  <aoliva@redhat.com>
3299         PR tree-optimization/81165
3300         * tree-ssa-threadupdate.c (uses_in_bb): New.
3301         (estimate_threading_killed_stmts): New.
3302         * tree-ssa-threadupdate.h (estimate_threading_killed_stmts): Prototype.
3303         * tree-ssa-threadedge.c
3304         (record_temporary_equivalences_from_stmts_at_dest): Expand limit
3305         when its hit.
3307 2017-12-15  Jeff Law  <law@redhat.com>
3309         PR tree-optimization/83410
3310         * tree-ssa-threadupdate.c (thread_block_1): Avoid certain jump
3311         threads when parallelizing loops.
3313 2017-12-15  Jakub Jelinek  <jakub@redhat.com>
3315         * tree-core.h (struct attribute_spec): Swap affects_type_identity and
3316         handler fields.
3317         * config/alpha/alpha.c (vms_attribute_table): Swap
3318         affects_type_identity and handler fields, adjust comments.
3319         * config/mips/mips.c (mips_attribute_table): Likewise.
3320         * config/visium/visium.c (visium_attribute_table): Likewise.
3321         * config/epiphany/epiphany.c (epiphany_attribute_table): Likewise.
3322         * config/microblaze/microblaze.c (microblaze_attribute_table):
3323         Likewise.
3324         * config/spu/spu.c (spu_attribute_table): Likewise.
3325         * config/mcore/mcore.c (mcore_attribute_table): Likewise.
3326         * config/arc/arc.c (arc_attribute_table): Likewise.
3327         * config/m68k/m68k.c (m68k_attribute_table): Likewise.
3328         * config/v850/v850.c (v850_handle_interrupt_attribute,
3329         v850_handle_data_area_attribute): Formatting fixes.
3330         (v850_attribute_table): Swap affects_type_identity and handler
3331         fields, adjust comments.
3332         * config/m32r/m32r.c (m32r_attribute_table): Likewise.
3333         * config/arm/arm.c (arm_attribute_table): Likewise.
3334         * config/avr/avr.c (avr_attribute_table): Likewise.
3335         * config/s390/s390.c (s390_attribute_table): Likewise.
3336         * config/sh/sh.c (sh_attribute_table): Likewise.
3337         * config/i386/i386.c (ix86_handle_cconv_attribute,
3338         ix86_handle_callee_pop_aggregate_return): Formatting fixes.
3339         (ix86_attribute_table): Swap affects_type_identity and handler
3340         fields, adjust comments.
3341         * config/i386/cygming.h (SUBTARGET_ATTRIBUTE_TABLE): Likewise.
3342         * config/sparc/sparc.c (sparc_attribute_table): Likewise.
3343         * config/m32c/m32c.c (m32c_attribute_table): Likewise.
3344         * config/sol2.h (SOLARIS_ATTRIBUTE_TABLE): Likewise.
3345         * config/ia64/ia64.c (ia64_attribute_table): Likewise.
3346         * config/msp430/msp430.c (msp430_attribute_table): Likewise.
3347         * config/rx/rx.c (rx_attribute_table): Likewise.
3348         * config/cr16/cr16.c (cr16_attribute_table): Likewise.
3349         * config/h8300/h8300.c (h8300_attribute_table): Likewise.
3350         * config/nvptx/nvptx.c (nvptx_attribute_table): Likewise.
3351         * config/powerpcspe/powerpcspe.c (rs6000_attribute_table): Likewise.
3352         * config/darwin.h (SUBTARGET_ATTRIBUTE_TABLE): Likewise.
3353         * config/stormy16/stormy16.c (xstormy16_attribute_table): Likewise.
3354         * config/bfin/bfin.c (bfin_attribute_table): Likewise.
3355         * config/rs6000/rs6000.c (rs6000_attribute_table): Likewise.
3356         * config/rl78/rl78.c (rl78_attribute_table): Likewise.
3357         * config/nds32/nds32.c (nds32_attribute_table): Likewise.
3358         * doc/plugins.texi (user_attr): Likewise.  Add NULL for
3359         exclude.
3360         * attribs.c (empty_attribute_table): Swap affects_type_identity and
3361         handler fields.
3362         (register_scoped_attributes, decl_attributes): Formatting fixes.
3364         PR tree-optimization/83269
3365         * fold-const.c (fold_binary_loc): Perform (-A) - B -> (-B) - A
3366         subtraction in arg0's type if type is signed and arg0 is unsigned.
3367         Formatting fix.
3369         PR sanitizer/81281
3370         * match.pd ((T)(P + A) - (T)P -> (T) A): Use @@0 instead of @0 and
3371         convert? on @0 instead of convert.  Check type of @1, not @0.
3372         ((T)P - (T)(P + A) -> -(T) A): Use @@0 instead of @0 and
3373         convert? on @0 instead of convert.  Check type of @1, not @0.
3374         ((T)(P + A) - (T)(P + B) -> (T)A - (T)B): Use @@0 instead of @0,
3375         only optimize if either both @1 and @2 types are narrower
3376         precision, or both are wider or equal precision, and in the former
3377         case only if both have undefined overflow.
3379 2017-12-15  Richard Biener  <rguenther@suse.de>
3381         PR lto/83388
3382         * internal-fn.def (IFN_NOP): Add.
3383         * internal-fn.c (expand_NOP): Do nothing.
3384         * lto-streamer-in.c (input_function): Instead of removing
3385         sanitizer calls replace them with IFN_NOP calls.
3387 2017-12-15  Richard Sandiford  <richard.sandiford@linaro.org>
3388             Alan Hayward  <alan.hayward@arm.com>
3389             David Sherwood  <david.sherwood@arm.com>
3391         * dse.c (store_info, read_info_type): Replace begin and end with
3392         offset and width.
3393         (print_range): New function.
3394         (set_all_positions_unneeded, any_positions_needed_p)
3395         (check_mem_read_rtx, scan_stores, scan_reads, dse_step5): Update
3396         accordingly.
3397         (record_store): Likewise.  Optimize the case in which all positions
3398         are unneeded.
3399         (get_stored_val): Replace read_begin and read_end with read_offset
3400         and read_width.
3401         (replace_read): Update call accordingly.
3403 2017-12-15  Bin Cheng  <bin.cheng@arm.com>
3405         * gimple-loop-interchange.cc (STMT_COST_RATIO): New macro.
3406         (loop_cand::m_num_stmts, loop_cand::m_const_init_reduc): New members.
3407         (loop_cand::loop_cand): Initialize above members.
3408         (loop_cand::supported_operations): Delete.
3409         (loop_cand::can_interchange_p): Inline above function.
3410         (loop_cand::classify_simple_reduction): Record number of constant
3411         initialized simple reductions.
3412         (should_interchange_loops): New parameters.  Check stmt cost of loops
3413         to be interchange.
3414         (tree_loop_interchange::interchange): Prepare stmt cost of outer loop.
3415         Update call to should_interchange_loops.
3416         (should_interchange_loop_nest): Update call to
3417         should_interchange_loops.
3419 2017-12-15  Eric Botcazou  <ebotcazou@adacore.com>
3421         PR target/66488
3422         * ggc-page.c (HOST_BITS_PER_PTR): Do not define here...
3423         * hwint.h (HOST_BITS_PER_PTR): ...but here instead.
3424         * config/i386/xm-mingw32.h (HOST_BITS_PER_PTR): Delete.
3426 2017-12-15  Julia Koval  <julia.koval@intel.com>
3428         * config/i386/i386-builtin.def (__builtin_ia32_vaesenclast_v16qi,
3429         __builtin_ia32_vaesenclast_v32qi, __builtin_ia32_vaesenclast_v64qi):
3430         New.
3431         * config/i386/sse.md (vaesenclast_<mode>): New pattern.
3432         * config/i386/vaesintrin.h (_mm256_aesenclast_epi128,
3433         _mm512_aesenclast_epi128, _mm_aesenclast_epi128): New intrinsics.
3435 2017-12-15  Julia Koval  <julia.koval@intel.com>
3437         * config/i386/i386-builtin.def (__builtin_ia32_vaesenc_v16qi,
3438         __builtin_ia32_vaesenc_v32qi, __builtin_ia32_vaesenc_v64qi): New.
3439         * config/i386/sse.md (vaesenc_<mode>): New pattern.
3440         * config/i386/vaesintrin.h (_mm256_aesenc_epi128, _mm512_aesenc_epi128,
3441         _mm_aesenc_epi128): New intrinsics.
3443 2017-12-15  Julia Koval  <julia.koval@intel.com>
3445         * config/i386/i386-builtin.def (__builtin_ia32_vaesdeclast_v16qi,
3446         __builtin_ia32_vaesdeclast_v32qi, __builtin_ia32_vaesdeclast_v64qi):
3447         New.
3448         * config/i386/sse.md (vaesdeclast_<mode>): New pattern.
3449         * config/i386/vaesintrin.h (_mm256_aesdeclast_epi128,
3450         _mm512_aesdeclast_epi128, _mm_aesdeclast_epi128): New intrinsics.
3452 2017-12-14  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
3454         * gimple-ssa-strength-reduction.c (analyze_increments):
3455         Distinguish replacement costs for constant strides from those for
3456         unknown strides.
3458 2017-12-14  Jakub Jelinek  <jakub@redhat.com>
3460         * var-tracking.c (variable_tracking_main_1): Formatting fix.
3462 2017-12-14  Bernd Edlinger  <bernd.edlinger@hotmail.de>
3464         * doc/invoke.texi: Document -Wcast-function-type.
3465         * recog.h (stored_funcptr): Change signature.
3466         * tree-dump.c (dump_node): Avoid warning.
3467         * typed-splay-tree.h (typed_splay_tree): Avoid warning.
3469 2017-12-14  Qing Zhao  <qing.zhao@oracle.com>
3471         PR middle_end/79538
3472         * gimple-fold.c (get_range_strlen): Add the handling of non-member 
3473         array.
3475 2017-12-14  David Malcolm  <dmalcolm@redhat.com>
3477         PR tree-optimization/83312
3478         * domwalk.h (dom_walker::dom_walker): Fix typo in comment.
3479         * tree-cfg.c (find_taken_edge): Update to handle NULL_TREE for
3480         "val" param, and to cope with arbitrary basic blocks.
3481         (find_taken_edge_cond_expr): Add "cond_stmt" param and use it to
3482         handle NULL_TREE for "val", dropping "bb" param.
3483         (find_taken_edge_switch_expr): Make "switch_stmt" param const and
3484         drop "bb" param.  Handle NULL_TREE for "val".
3485         (find_case_label_for_value): Make "switch_stmt" param const.
3486         * tree-vrp.c (class check_array_bounds_dom_walker): New subclass
3487         of dom_walker.
3488         (vrp_prop::check_all_array_refs): Reimplement as...
3489         (check_array_bounds_dom_walker::before_dom_children): ...this new
3490         vfunc.  Replace linear search through BB block list, excluding
3491         those with non-executable in-edges via dominator walk.
3493 2017-12-14  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
3495         * config/arm/arm.opt (mverbose-cost-dump): New option.
3496         * config/arm/arm.c (arm_rtx_costs): Use it.
3498 2017-12-14  Andreas Schwab  <schwab@linux-m68k.org>
3500         PR bootstrap/83396
3501         * reload1.c (emit_input_reload_insns): Skip debug markers.
3503 2017-12-14  Alexandre Oliva  <aoliva@redhat.com>
3505         * config/i386/i386.c (rest_of_insert_endbranch): Use call loc
3506         for its nop_endbr.
3508         PR bootstrap/83396
3509         * config/arc/arc.c (hwloop_optimize): Skip debug insns.
3510         * config/sh/sh-protos.h (sh_find_set_of_reg): Adjust.
3511         * config/sh/sh.c: Skip debug insns besides notes.
3512         * config/sh/sh.md: Likewise.
3513         * config/sh/sh_treg_combine.cc: Likewise.
3514         * config/sh/sync.md: Likewise.
3516 2017-12-14  Tom de Vries  <tom@codesourcery.com>
3518         * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
3519         Add item for weak_undefined.
3521 2017-12-14  Richard Biener  <rguenther@suse.de>
3523         PR tree-optimization/67842
3524         * sese.h (bb_in_region): Remove #if 0'ed code.
3526 2017-12-14  Richard Biener  <rguenther@suse.de>
3528         PR tree-optimization/83326
3529         * tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Add
3530         may_be_zero parameter and handle it by not marking the first
3531         peeled copy as not exiting the loop.
3532         (try_peel_loop): Likewise.
3533         (canonicalize_loop_induction_variables): Use number_of_iterations_exit
3534         to handle the case of constant or zero iterations and perform
3535         loop header copying on-the-fly.
3537 2017-12-14  Richard Biener  <rguenther@suse.de>
3539         PR tree-optimization/83418
3540         * vr-values.c (vr_values::extract_range_for_var_from_comparison_expr):
3541         Instead of asserting we don't get unfolded comparisons deal with them.
3543 2017-12-14  Jakub Jelinek  <jakub@redhat.com>
3545         PR bootstrap/83396
3546         * var-tracking.c (vt_initialize): Ignore non-DEBUG_INSNs outside of
3547         basic blocks.  Assert debug bind insns don't appear outside of bbs,
3548         don't reset them.  Assert insns without BLOCK_FOR_INSN are outside of
3549         bb.  Simplify.
3551         PR tree-optimization/83198
3552         * gimple-ssa-sprintf.c (format_floating): Set type solely based on
3553         dir.modifier, regardless of TREE_TYPE (arg).  Assume non-REAL_CST
3554         value if arg is a REAL_CST with incompatible type.
3556 2017-12-14  Sudakshina Das  <sudi.das@arm.com>
3557             Bin Cheng  <bin.cheng@arm.com>
3559         PR target/81228
3560         * config/aarch64/aarch64.c (aarch64_select_cc_mode): Move LTGT
3561         to CCFPEmode.
3562         * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Add
3563         LTGT.
3565 2017-12-14  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
3567         * config/aarch64/aarch64-cores.def (cortex-a55, cortex-a75,
3568         cortex-a75.cortex-a55): Specify AARCH64_FL_F16 in the arch features.
3570 2017-12-14  Richard Sandiford  <richard.sandiford@linaro.org>
3571             Alan Hayward  <alan.hayward@arm.com>
3572             David Sherwood  <david.sherwood@arm.com>
3574         * poly-int.h: New file.
3575         * poly-int-types.h: Likewise.
3576         * coretypes.h: Include them.
3577         (POLY_INT_CONVERSION): Define.
3578         * target.def (estimated_poly_value): New hook.
3579         * doc/tm.texi.in (TARGET_ESTIMATED_POLY_VALUE): New hook.
3580         * doc/tm.texi: Regenerate.
3581         * doc/poly-int.texi: New file.
3582         * doc/gccint.texi: Include it.
3583         * doc/rtl.texi: Describe restrictions on subreg modes.
3584         * Makefile.in (TEXI_GCCINT_FILES): Add poly-int.texi.
3585         * genmodes.c (NUM_POLY_INT_COEFFS): Provide a default definition.
3586         (emit_insn_modes_h): Emit a definition of NUM_POLY_INT_COEFFS.
3587         * targhooks.h (default_estimated_poly_value): Declare.
3588         * targhooks.c (default_estimated_poly_value): New function.
3589         * target.h (estimated_poly_value): Likewise.
3590         * wide-int.h (WI_UNARY_RESULT): Use wi::binary_traits.
3591         (wi::unary_traits): Delete.
3592         (wi::binary_traits::signed_shift_result_type): Define for
3593         offset_int << HOST_WIDE_INT, etc.
3594         (generic_wide_int::operator <<=): Define for all types and use
3595         wi::lshift instead of <<.
3596         (wi::hwi_with_prec): Add a default constructor.
3597         (wi::ints_for): New class.
3598         (operator <<): Define for all wide-int types.
3599         (operator /): New function.
3600         (operator %): Likewise.
3601         * selftest.h (ASSERT_KNOWN_EQ, ASSERT_KNOWN_EQ_AT, ASSERT_MAYBE_NE)
3602         (ASSERT_MAYBE_NE_AT): New macros.
3604 2017-12-13  Eric Botcazou  <ebotcazou@adacore.com>
3605             Dominik Vogt  <vogt@linux.vnet.ibm.com>
3607         PR middle-end/78468
3608         * emit-rtl.c (init_emit): Remove ??? comment.
3609         * explow.c (get_dynamic_stack_size): Take known alignment of stack
3610         pointer + STACK_DYNAMIC_OFFSET into account in lieu of STACK_BOUNDARY.
3611         * config/sparc/sparc.h (INIT_EXPANDERS): In 32-bit mode, lower the
3612         alignment of 3 virtual registers to BITS_PER_WORD.
3614         * config/sparc/sparc.c (sparc_compute_frame_size): Simplify.
3616 2017-12-13  Peter Bergner  <bergner@vnet.ibm.com>
3618         * config/rs6000/ppc-auxv.h (PPC_FEATURE2_HTM_NO_SUSPEND): New define.
3619         * config/rs6000/rs6000.c (cpu_supports_info): Use it.
3621 2017-12-13  Alexandre Oliva  <aoliva@redhat.com>
3623         PR bootstrap/83396
3624         * reload1.c (eliminate_regs_in_insn): Skip debug markers.
3626 2017-12-13  Jakub Jelinek  <jakub@redhat.com>
3628         * tree-cfg.c (verify_gimple_in_cfg): Verify no non-label stmts
3629         with the exception of debug begin stmt markers appear before
3630         labels.
3632         PR bootstrap/83396
3633         * final.c (rest_of_handle_final): Call variable_tracking_main only
3634         if !flag_var_tracking.
3636 2017-12-13  Alexandre Oliva  <aoliva@redhat.com>
3637             Jakub Jelinek  <jakub@redhat.com>
3639         PR bootstrap/83396
3640         PR debug/83391
3641         * tree-cfgcleanup.c (remove_forwarder_block): Keep after
3642         labels debug stmts that can only appear after labels.
3644 2017-12-13  Alexander Monakov  <amonakov@ispras.ru>
3646         PR rtl-optimization/82398
3647         * sel-sched.c (sel_rank_for_schedule): Fix check for zero
3648         EXPR_USEFULNESS in priority comparison.
3650 2017-12-13  Segher Boessenkool  <segher@kernel.crashing.org>
3652         PR rtl-optimization/83393
3653         * combine.c (move_deaths): If reg_stat points to a too new insn in
3654         last_death, do not use it: find the proper insn instead.
3656 2017-12-12  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
3658         PR target/82190
3659         * config/rs6000/rs6000-string.c (expand_block_compare,
3660         expand_strn_compare): Fix set_mem_size() calls.
3662 2017-12-12  Jeff Law  <law@redhat.com>
3664         PR tree-optimization/83298
3665         PR tree-optimization/83362
3666         PR tree-optimization/83383
3667         * gimple-ssa-evrp-analyze.h (class evrp_range_analyzer): Make
3668         push_value_range a public interface.  Add new argument to
3669         record_ranges_from_stmt.
3670         * gimple-ssa-evrp-analyze.c
3671         (evrp_range_analyzer::record_ranges_from_stmt): Add new argument.
3672         Update comments.  Handle recording temporary equivalences.
3673         * tree-ssa-dom.c (dom_opt_opt_walker::before_dom_children): Add
3674         new argument to call to evrp_range_analyzer::record_ranges_from_stmt.
3675         * gimple-ssa-evrp.c (evrp_dom_walker::before_dom_children): Likewise.
3676         * tree-ssa-threadedge.c: Include alloc-pool.h, vr-values.h and
3677         gimple-ssa-evrp-analyze.h.
3678         (record_temporary_equivalences_from_phis): Add new argument.  When
3679         the PHI arg is an SSA_NAME, set the result's range to the range
3680         of the PHI arg.
3681         (record_temporary_equivalences_from_stmts_at_dest): Record ranges
3682         from statements too.
3683         (thread_through_normal_block): Accept new argument, evrp_range_analyzer.
3684         Pass it down to children as needed.
3685         (thread_outgoing_edges): Likewise.
3686         (thread_across_edge): Likewise.   Push/pop range state as needed.
3687         * tree-ssa-threadedge.h (thread_outgoing_edges): Update prototype.
3689 2017-12-12  Julia Koval  <julia.koval@intel.com>
3691         * config/i386/i386.c (PTA_SKYLAKE_AVX512): Add PTA_CLWB.
3692         (PTA_CANNONLAKE): Remove PTA_CLWB.
3694 2017-12-12  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
3696         PR target/83332
3697         * config/rs6000/vector.md (vcondv2dfv2di): New define_expand.
3698         (vcondv2div2df): Likewise.
3699         (vconduv2dfv2di): Likewise.
3701 2017-12-12  Jakub Jelinek  <jakub@redhat.com>
3703         * builtins.def (BUILT_IN_NEXTAFTER, BUILT_IN_NEXTAFTERF,
3704         BUILT_IN_NEXTAFTERL, BUILT_IN_NEXTTOWARD, BUILT_IN_NEXTTOWARDF,
3705         BUILT_IN_NEXTTOWARDL): Use ATTR_MATHFN_ERRNO instead of
3706         ATTR_MATHFN_FPROUNDING_ERRNO.
3708 2017-12-12  Richard Biener  <rguenther@suse.de>
3710         PR tree-optimization/83385
3711         * graphite-scop-detection.c (get_order, order): Remove.
3712         (bb_to_rpo): New global.
3713         (cmp_pbbs): Adjust.
3714         (build_scops): Sort pbbs in RPO order.
3716 2017-12-12  James Greenhalgh  <james.greenhalgh@arm.com>
3718         * combine.c (simplify_set): Do not transform subregs to zero_extends
3719         if the destination mode is a vector mode.
3721 2017-12-12  Jakub Jelinek  <jakub@redhat.com>
3723         PR tree-optimization/83359
3724         * tree-cfg.h (fold_loop_internal_call): Declare.
3725         * tree-vectorizer.c (fold_loop_internal_call): Moved to ...
3726         * tree-cfg.c (fold_loop_internal_call): ... here.  No longer static.
3727         (find_loop_dist_alias): New function.
3728         (move_sese_region_to_fn): If any dloop->orig_loop_num value is
3729         updated, also adjust any corresponding LOOP_DIST_ALIAS internal
3730         calls.
3732         PR tree-optimization/80631
3733         * tree-vect-loop.c (get_initial_def_for_reduction): Fix comment typo.
3734         (vect_create_epilog_for_reduction): Add INDUC_VAL and INDUC_CODE
3735         arguments, for INTEGER_INDUC_COND_REDUCTION use INDUC_VAL instead of
3736         hardcoding zero as the value if COND_EXPR is never true.  For
3737         INTEGER_INDUC_COND_REDUCTION don't emit the final COND_EXPR if
3738         INDUC_VAL is equal to INITIAL_DEF, and use INDUC_CODE instead of
3739         hardcoding MAX_EXPR as the reduction operation.
3740         (is_nonwrapping_integer_induction): Allow negative step.
3741         (vectorizable_reduction): Compute INDUC_VAL and INDUC_CODE for
3742         vect_create_epilog_for_reduction, if no value is suitable, don't
3743         use INTEGER_INDUC_COND_REDUCTION for now.  Formatting fixes.
3745 2017-12-12  Richard Biener  <rguenther@suse.de>
3747         PR tree-optimization/81889
3748         * tree-ssa-loop-niter.c (infer_loop_bounds_from_signedness): Use
3749         range info from the non-wrapping IV instead of just the range
3750         of the type.
3752 2017-12-12  Julia Koval  <julia.koval@intel.com>
3754         * config.gcc: Add vaesintrin.h.
3755         * config/i386/i386-builtin-types.def (V64QI_FTYPE_V64QI_V64QI):
3756         New type.
3757         * config/i386/i386-builtin.def (__builtin_ia32_vaesdec_v16qi,
3758         __builtin_ia32_vaesdec_v32qi, __builtin_ia32_vaesdec_v64qi):
3759         New builtins.
3760         * config/i386/i386.c (ix86_expand_args_builtin): Handle new type.
3761         * config/i386/immintrin.h: Include vaesintrin.h.
3762         * config/i386/sse.md (vaesdec_<mode>): New pattern.
3763         * config/i386/vaesintrin.h (_mm256_aesdec_epi128, _mm512_aesdec_epi128,
3764         _mm_aesdec_epi128): New intrinsics.
3766 2017-12-12  Julia Koval  <julia.koval@intel.com>
3768         * common/config/i386/i386-common.c (OPTION_MASK_ISA_VAES_SET,
3769         OPTION_MASK_ISA_VAES_UNSET): New.
3770         (ix86_handle_option): Handle -mvaes.
3771         * config/i386/cpuid.h: Define bit_VAES.
3772         * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mvaes.
3773         * config/i386/i386-c.c (__VAES__): New.
3774         * config/i386/i386.c (ix86_target_string): Add -mvaes.
3775         (ix86_valid_target_attribute_inner_p): Ditto.
3776         * config/i386/i386.h (TARGET_VAES, TARGET_VAES_P): New.
3777         * config/i386/i386.opt: Add -mvaes.
3778         * doc/invoke.texi: Ditto.
3780 2017-12-12  Alexandre Oliva  <aoliva@redhat.com>
3782         * debug.h (gcc_debug_hooks): Add inline_entry.
3783         * dbxout.c (dbx_debug_hooks, xcoff_debug_hooks): Likewise.
3784         * debug.c (do_nothing_debug_hooks): Likewise.
3785         * vmsdbgout.c (vmsdbg_debug_hooks): Likewise.
3786         * dwarf2out.c (dwarf2_debug_hooks): Likewise.
3787         (dwarf2_lineno_debug_hooks): Likewise.
3789         * common.opt (gstatement-frontiers): New, setting
3790         debug_nonbind_markers_p.
3791         * rtl.h (MAY_HAVE_DEBUG_MARKER_INSNS): Activate.
3792         * toplev.c (process_options): Autodetect value for debug statement
3793         frontiers option.
3794         * tree.h (MAY_HAVE_DEBUG_MARKER_STMTS): Activate.
3795         * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers): New.
3797         * cfgexpand.c (expand_gimple_basic_block): Handle begin stmt
3798         markers.  Integrate source bind into debug stmt expand loop.
3799         (pass_expand::execute): Check debug marker limit.  Avoid deep
3800         TER and expand debug locations for debug bind insns only.
3801         * cse.c (insn_live_p): Keep nonbind markers and debug bindings
3802         followed by them.
3803         * df-scan.c (df_insn_delete): Accept out-of-block debug insn.
3804         * final.c (reemit_insn_block_notes): Take current block from
3805         nonbind markers.  Declare note where it's first set.
3806         (final_scan_insn): Handle begin stmt notes.  Emit is_stmt according to
3807         begin stmt markers if enabled.
3808         (notice_source_line): Handle nonbind markers.  Fail if their
3809         location is unknown or that of builtins.
3810         (rest_of_handle_final): Convert begin stmt markers to notes if
3811         var-tracking didn't run.
3812         (rest_of_clean_state): Skip begin stmt markers.
3813         * gimple-pretty-print.c (dump_gimple_debug): Handle begin stmt
3814         markers.
3815         * function.c (allocate_struct_function): Set begin_stmt_markers.
3816         * function.h (struct function): Add debug_marker_count counter
3817         and debug_nonbind_markers flag.
3818         * gimple-iterator.c (gsi_remove): Adjust debug_marker_count.
3819         * gimple-low.c (lower_function_body): Adjust
3820         debug_nonbind_markers.
3821         (lower_stmt): Drop or skip gimple debug stmts.
3822         (lower_try_catch): Skip debug stmts.
3823         * gimple.c (gimple_build_debug_begin_stmt): New.
3824         (gimple_copy): Increment debug_marker_count if copying one.
3825         * gimple.h (gimple_build_debug_begin_stmt): Declare.
3826         * gimplify.c (rexpr_location): New.
3827         (rexpr_has_location): New.
3828         (warn_switch_unreachable_r): Handle gimple debug stmts.
3829         (shortcut_cond_r): Call expr_location.
3830         (find_goto): New.
3831         (find_goto_label): New.
3832         (shortcut_cond_expr): Call expr_has_location, expr_location, and
3833         find_goto_label.
3834         (gimplify_cond_expr): Call find_goto_label, expr_has_location, and
3835         expr_location.
3836         (gimplify_expr): Handle begin stmt markers.  Reject debug expr decls.
3837         * langhooks-def.h (LANG_HOOKS_EMITS_BEGIN_STMT): New.  Add to...
3838         (LANG_HOOKS_INITIALIZER): ... this.
3839         * langhooks.h (struct lang_hooks): Add emits_begin_stmt.
3840         * lra-contraints.c (inherit_reload_reg): Tolerate between-blocks
3841         debug insns.
3842         (update_ebb_live_info): Skip debug insn markers.
3843         * lra.c (debug_insn_static_data): Rename to...
3844         (debug_bind_static_data): ... this.
3845         (debug_marker_static_data): New.
3846         (lra_set_insn_recog_data): Select one of the above depending
3847         on debug insn kind.
3848         (lra_update_isn_regno_info): Don't assume debug insns have
3849         freqs.
3850         (push_insns): Skip debug insns.
3851         * lto-streamer-in.c (input_function): Drop debug stmts
3852         depending on active options.  Adjust debug_nonbind_markers.
3853         * params.def (PARAM_MAX_DEBUG_MARKER_COUNT): New.
3854         * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
3855         begin stmt marker notes.
3856         (print_insn): Likewise.
3857         * recog.c (extract_insn): Recognize rtl for debug markers.
3858         * rtl.def (DEBUG_MARKER): New.
3859         * tree-inline.c: Include params.h.
3860         (remap_gimple_stmt): Handle nonbind markers.
3861         (maybe_move_debug_stmts_to_successors): Likewise.
3862         (copy_debug_stmt): Likewise.
3863         * tree-iterator.c (append_to_statement_list_1): Append begin stmt
3864         markers regardless of no side effects.
3865         (tsi_link_before): Don't update container's side effects when adding
3866         a begin stmt marker.
3867         (tsi_link_after): Likewise.
3868         (expr_first): Skip begin stmt markers.
3869         (expr_last): Likewise.
3870         * tree-pretty-print (dump_generic_node): Handle begin stmt markers.
3871         * tree-ssa-threadedge.c (propagate_threaded_block_debug_info):
3872         Disregard nonbind markers.
3873         * tree.c (make_node_stat): Don't set side effects for begin stmt
3874         markers.
3875         (build1_stat): Likewise.
3876         * tree.def (DEBUG_BEGIN_STMT): New.
3877         * tree.h (GOTO_DESTINATION): Require a GOTO_EXPR.
3878         * var-tracking.c (delete_debug_insns): Renamed to...
3879         (delete_vta_debug_insns): ... this.
3880         (reemit_marker_as_note): New.
3881         (vt_initialize): Reemit markers.
3882         (delete_vta_debug_insns): Likewise.
3883         (vt_debug_insns_local): Reemit or delete markers.
3884         (variable_tracking_main_1): Likewise.
3885         * doc/generic.texi (DEBUG_BEGIN_STMT): Document.
3886         * doc/gimple.texi (gimple_debug_begin_stmt_p): New.
3887         (gimple_debug_nonbind_marker_p): New.
3888         (gimple_build_debug_bind): Adjust.
3889         (gimple_build_debug_begin_stmt): New.
3890         * doc/invoke.texi (max-debug-marker-count): New param.
3891         * doc/rtl.texi (debug_implicit_ptr, entry_value): New.
3892         (debug_parameter_ref, debug_marker): New.
3893         (NOTE_INSN_BEGIN_STMT): New.
3894         (DEBUG_INSN): Describe begin stmt markers.
3896         * cfgbuild.c (find_bb_boundaries): Don't purge dead edges if,
3897         without debug insns, we wouldn't, but clean up debug insns
3898         after a control flow insn nevertheless.
3900         * cfgcleanup.c (delete_unreachable_blocks): Use alternate
3901         block removal order if MAY_HAVE_DEBUG_BIND_INSNS.
3902         * cfgexpand.c (label_rtx_for_bb): Skip debug insns.
3903         * cfgrtl.c (try_redirect_by_replacing_jump): Skip debug insns.
3904         (rtl_tidy_fallthru_edge): Likewise.
3905         (rtl_verify_fallthru): Likewise.
3906         (rtl_verify_bb_layout): Likewise.
3907         (skip_insns_after_block): Likewise.
3908         (duplicate_insn_chain): Use DEBUG_BIND_INSN_P.
3909         * dwarf2out.c: Include print-rtl.h.
3910         (dwarf2out_next_real_insn): New.
3911         (dwarf2out_var_location): Call it.  Disregard begin stmt markers.
3912         Dump debug binds in asm comments.
3913         * gimple-iterator.c (gimple_find_edge_insert_loc): Skip debug stmts.
3914         * gimple-iterator.h (gsi_start_bb_nondebug): Remove; adjust
3915         callers to use gsi_start_nondebug_bb instead.
3916         (gsi_after_labels): Skip gimple debug stmts.
3917         (gsi_start_nondebug): New.
3918         * gimple-loop-interchange.c (find_deps_in_bb_for_stmt): Adjust.
3919         (proper_loop_form_for_interchange): Adjust.
3920         * gimple-low.c (gimple_seq_may_fallthru): Take last nondebug stmt.
3921         * gimple.h (gimple_seq_last_nondebug_stmt): New.
3922         * gimplify.c (last_stmt_in_scope): Skip debug stmts.
3923         (collect_fallthrough_labels): Likewise.
3924         (should_warn_for_implicit_fallthrough): Likewise.
3925         (warn_implicit_fallthrough_r): Likewise.
3926         (expand_FALLTHROUGH_r): Likewise.
3927         * graphite-isl-ast-to-gimple.c (gsi_insert_earliest): Adjust.
3928         (graphite_copy_stmts_from_block): Skip nonbind markers.
3929         * haifa-sched.c (sched_extend_bb): Skip debug insns.
3930         * ipa-icf-gimple.c (func_checker::compare_bb): Adjust.
3931         * jump.c (clean_barriers): Skip debug insns.
3932         * omp-expand.c (expand_parallel_call): Skip debug insns.
3933         (expand_task_call): Likewise.
3934         (remove_exit_barrier): Likewise.
3935         (expand_omp_taskreg): Likewise.
3936         (expand_omp_for_init_counts): Likewise.
3937         (expand_omp_for_generic): Likewise.
3938         (expand_omp_for_static_nochunk): Likewise.
3939         (expand_omp_for_static_chunk): Likewise.
3940         (expand_omp_simd): Likewise.
3941         (expand_omp_taskloop_for_outer): Likewise.
3942         (expand_omp_taskloop_for_inner): Likewise.
3943         (expand_oacc_for): Likewise.
3944         (expand_omp_sections): Likewise.
3945         (expand_omp_single): Likewise.
3946         (expand_omp_synch): Likewise.
3947         (expand_omp_atomic_load): Likewise.
3948         (expand_omp_atomic_store): Likewise.
3949         (expand_omp_atomic_fetch_op): Likewise.
3950         (expand_omp_atomic_pipeline): Likewise.
3951         (expand_omp_atomic_mutex): Likewise.
3952         (expand_omp_target): Likewise.
3953         (grid_expand_omp_for_loop): Likewise.
3954         (grid_expand_target_grid_body): Likewise.
3955         (build_omp_regions_1): Likewise.
3956         * omp-low.c (check_combined_parallel): Skip debug stmts.
3957         * postreload.c (fixup_debug_insns): Skip nonbind debug insns.
3958         * regcprop.c (find_oldest_value_reg): Ensure REGNO is not a pseudo.
3959         * sese.c (sese_trivially_empty_bb_p): Call is_gimple_debug in
3960         test.
3961         * tree-cfg.c (make_blobs_1): Skip debug stmts.
3962         (make_edges): Likewise.
3963         (cleanup_dead_labels): Likewise.
3964         (gimple_can_merge_blocks_p): Likewise.
3965         (stmt_starts_bb_p): Likewise.
3966         (gimple_block_label): Likewise.
3967         (gimple_redirect_edge_and_branch): Likewise.
3968         * tree-cfgcleanup.c (remove_forwarder_block): Rearrange skipping
3969         of debug stmts.
3970         (execute_cleanup_cfg_post_optimizing): Dump enumerated decls with
3971         TDF_SLIM.
3972         * tree-pretty-print (print_declaration): Omit initializer in slim
3973         dumps.
3974         * tree-ssa-dce.c (mark_stmt_if_obviously_necessary): Mark begin stmt
3975         markers.
3976         (eliminate_unnecessary_stmts): Stabilize block removal order.
3977         * tree-ssa-tail-merge.c (find_duplicate): Skip debug stmts.
3978         * var-tracking.c (get_first_insn): New.
3979         (vt_emit_notes): Call it.
3980         (vt_initialize): Walk any insns before the first BB.
3981         (delete_debug_insns): Likewise.
3983         * gimple.h (enum gimple_debug_subcode): Add
3984         GIMPLE_DEBUG_BEGIN_STMT.
3985         (gimple_debug_begin_stmt_p): New.
3986         (gimple_debug_nonbind_marker_p): New.
3987         * tree.h (MAY_HAVE_DEBUG_MARKER_STMTS): New.
3988         (MAY_HAVE_DEBUG_BIND_STMTS): Renamed from....
3989         (MAY_HAVE_DEBUG_STMTS): ... this.  Check both.
3990         * insn-notes.def (BEGIN_STMT): New.
3991         * rtl.h (MAY_HAVE_DEBUG_MARKER_INSNS): New.
3992         (MAY_HAVE_DEBUG_BIND_INSNS): Renamed from....
3993         (MAY_HAVE_DEBUG_INSNS): ... this.  Check both.
3994         (NOTE_MARKER_LOCATION, NOTE_MARKER_P): New.
3995         (DEBUG_BIND_INSN_P, DEBUG_MARKER_INSN_P): New.
3996         (INSN_DEBUG_MARKER_KIND): New.
3997         (GEN_RTX_DEBUG_MARKER_BEGIN_STMT_PAT): New.
3998         (INSN_VAR_LOCATION): Check for VAR_LOCATION.
3999         (INSN_VAR_LOCATION_PTR): New.
4000         * cfgexpand.c (expand_debug_locations): Handle debug bind insns
4001         only.
4002         (expand_gimple_basic_block): Likewise.  Emit debug temps for TER
4003         deps only if debug bind insns are enabled.
4004         (pass_expand::execute): Avoid deep TER and expand
4005         debug locations for debug bind insns only.
4006         * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Narrow
4007         debug stmts special handling down to debug bind stmts.
4008         * combine.c (try_combine): Narrow debug insns special handling
4009         down to debug bind insns.
4010         * cse.c (delete_trivially_dead_insns): Handle debug bindings.
4011         Narrow debug insns preexisting special handling down to debug
4012         bind insns.
4013         * dce.c (rest_of_handle_ud_dce): Narrow debug insns special
4014         handling down to debug bind insns.
4015         * function.c (instantiate_virtual_regs): Skip debug markers,
4016         adjust handling of debug binds.
4017         * gimple-ssa-backprop.c (backprop::prepare_change): Try debug
4018         temp insertion iff MAY_HAVE_DEBUG_BIND_STMTS.
4019         * haifa-sched.c (schedule_insn): Narrow special handling of debug
4020         insns to debug bind insns.
4021         * ipa-param-manipulation.c (ipa_modify_call_arguments): Narrow
4022         special handling of debug stmts to debug bind stmts.
4023         * ipa-split.c (split_function): Likewise.
4024         * ira.c (combine_and_move_insns): Adjust debug bind insns only.
4025         * loop-unroll.c (apply_opt_in_copies): Adjust tests on bind
4026         debug insns.
4027         * reg-stack.c (convert_regs_1): Use DEBUG_BIND_INSN_P.
4028         * regrename.c (build_def_use): Likewise.
4029         * regcprop.c (copyprop_hardreg_forward_1): Likewise.
4030         (pass_cprop_hardreg): Narrow special casing of debug insns to
4031         debug bind insns.
4032         * regstat.c (regstat_init_n_sets_and_refs): Likewise.
4033         * reload1.c (reload): Likewise.
4034         * sese.c (sese_insert_phis_for_liveouts): Narrow special
4035         casing of debug stmts to debug bind stmts.
4036         * shrink-wrap.c (move_insn_for_shrink_wrap): Likewise.
4037         * ssa-iterators.h (num_imm_uses): Likewise.
4038         * tree-cfg.c (gimple_merge_blocks): Narrow special casing of
4039         debug stmts to debug bind stmts.
4040         * tree-inline.c (tree_function_versioning): Narrow special casing
4041         of debug stmts to debug bind stmts.
4042         * tree-loop-distribution.c (generate_loops_for_partition):
4043         Narrow special casing of debug stmts to debug bind stmts.
4044         * tree-sra.c (analyze_access_subtree): Narrow special casing
4045         of debug stmts to debug bind stmts.
4046         * tree-ssa-dce.c (remove_dead_stmt): Narrow special casing of debug
4047         stmts to debug bind stmts.
4048         * tree-ssa-loop-ivopt.c (remove_unused_ivs): Narrow special
4049         casing of debug stmts to debug bind stmts.
4050         * tree-ssa-reassoc.c (reassoc_remove_stmt): Likewise.
4051         * tree-ssa-tail-merge.c (tail_merge_optimize): Narrow special
4052         casing of debug stmts to debug bind stmts.
4053         * tree-ssa-threadedge.c (propagate_threaded_block_debug_info):
4054         Likewise.
4055         * tree-ssa.c (flush_pending_stmts): Narrow special casing of
4056         debug stmts to debug bind stmts.
4057         (gimple_replace_ssa_lhs): Likewise.
4058         (insert_debug_temp_for_var_def): Likewise.
4059         (insert_debug_temps_for_defs): Likewise.
4060         (reset_debug_uses): Likewise.
4061         * tree-ssanames.c (release_ssa_name_fn): Likewise.
4062         * tree-vect-loop-manip.c (adjust_debug_stmts_now): Likewise.
4063         (adjust_debug_stmts): Likewise.
4064         (adjust_phi_and_debug_stmts): Likewise.
4065         (vect_do_peeling): Likewise.
4066         * tree-vect-loop.c (vect_transform_loop): Likewise.
4067         * valtrack.c (propagate_for_debug): Use BIND_DEBUG_INSN_P.
4068         * var-tracking.c (adjust_mems): Narrow special casing of debug
4069         insns to debug bind insns.
4070         (dv_onepart_p, dataflow_set_clar_at_call, use_type): Likewise.
4071         (compute_bb_dataflow, vt_find_locations): Likewise.
4072         (vt_expand_loc, emit_notes_for_changes): Likewise.
4073         (vt_init_cfa_base): Likewise.
4074         (vt_emit_notes): Likewise.
4075         (vt_initialize): Likewise.
4076         (vt_finalize): Likewise.
4078         * emit-rtl.c (next_nondebug_insn, prev_nondebug_insn): Reorder.
4079         (next_nonnote_nondebug_insn, prev_nonnote_nondebug_insn): Reorder.
4080         (next_nonnote_nondebug_insn_bb): New.
4081         (prev_nonnote_nondebug_insn_bb): New.
4082         (prev_nonnote_insn_bb, next_nonnote_insn_bb): Remove.
4083         * rtl.h (prev_nonnote_insn_bb, next_nonnote_insn_bb): Remove decls.
4084         (prev_nonnote_nondebug_insn_bb): Declare.
4085         (next_nonnote_nondebug_insn_bb): Declare.
4086         * cfgbuild.c (find_bb_boundaries): Adjust to skip debug insns.
4087         * cfgrtl.c (get_last_bb_insn): Likewise.
4088         * lra.c (push_insns): Likewise.
4090 2017-12-11  David Malcolm  <dmalcolm@redhat.com>
4092         PR c/82050
4093         * selftest-run-tests.c (selftest::run_tests): Move start/finish code
4094         to...
4095         * selftest.c (selftest::test_runner::test_runner): New ctor.
4096         (selftest::test_runner::~test_runner): New dtor.
4097         * selftest.h (class selftest::test_runner): New class.
4099 2017-12-11  Carl Love  <cel@us.ibm.com>
4101         * config/rs6000/altivec.h (vec_extract_fp32_from_shorth,
4102         vec_extract_fp32_from_shortl]): Add #defines.
4103         * config/rs6000/rs6000-builtin.def (VSLDOI_2DI): Add macro expansion.
4104         * config/rs6000/rs6000-c.c (ALTIVEC_BUILTIN_VEC_UNPACKH,
4105         ALTIVEC_BUILTIN_VEC_UNPACKL, ALTIVEC_BUILTIN_VEC_AND,
4106         ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SRL,
4107         ALTIVEC_BUILTIN_VEC_SRO, ALTIVEC_BUILTIN_VEC_SLD,
4108         ALTIVEC_BUILTIN_VEC_SLL): Add expansions.
4109         * doc/extend.texi: Add documentation for the added builtins.
4111 2017-12-11  Alexandre Oliva  <aoliva@redhat.com>
4113         PR rtl-optimization/80693
4114         PR rtl-optimization/81019
4115         PR rtl-optimization/81020
4116         * combine.c (distribute_notes): Reset any REG_UNUSED REGs that
4117         are not mentioned in i3.  Place the REG_UNUSED note on i2,
4118         possibly modified to REG_DEAD, if it did not originate in i3.
4120 2017-12-11  Jakub Jelinek  <jakub@redhat.com>
4122         * recog.c (store_data_bypass_p_1): New function.
4123         (store_data_bypass_p): Handle USE in a PARALLEL like CLOBBER.  Use
4124         store_data_bypass_p_1 to avoid code duplication.  Formatting fixes.
4126 2017-12-11  Segher Boessenkool  <segher@kernel.crashing.org>
4128         PR rtl-optimization/83361
4129         * ifcvt.c (if_convert): Call fixup_partitions.
4131 2017-12-11  Will Schmidt  <will_schmidt@vnet.ibm.com>
4133         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add support for
4134         early folding of splat_u{8,16,32}.
4136 2017-12-11  Jakub Jelinek  <jakub@redhat.com>
4138         * config/aarch64/aarch64.c (aarch64_print_operand): Don't start
4139         output_operand_lossage first argument with capital letter.
4140         (aarch64_override_options): Don't start error and sorry first argument
4141         with capital letter.
4143 2017-12-11  Andi Kleen  <ak@linux.intel.com>
4145         PR gcov-profile/83355
4146         * auto-profile.c (string_table::get_index_by_decl): Don't
4147         recurse when abstract origin points to itself.
4149 2017-12-11  Bin Cheng  <bin.cheng@arm.com>
4151         PR tree-optimization/83320
4152         * gimple-loop-interchange.cc (free_data_refs_with_aux): Use delete.
4153         (prune_datarefs_not_in_loop): Ditto.
4155 2017-12-10  Gerald Pfeifer  <gerald@pfeifer.com>
4157         * doc/install.texi (Specific): Tweak link to mkssoftware.com.
4159 2017-12-10  Jakub Jelinek  <jakub@redhat.com>
4161         PR tree-optimization/83337
4162         * gimple-loop-interchange.cc (compute_access_stride): Handle
4163         bitfield DRs properly.
4165 2017-12-09  Jakub Jelinek  <jakub@redhat.com>
4167         PR tree-optimization/83338
4168         * tree-vect-stmts.c (vectorizable_operation): Handle POINTER_DIFF_EXPR
4169         vectorization as MINUS_EXPR with a subsequent VIEW_CONVERT_EXPR from
4170         vector of unsigned integers to vector of signed integers.
4172 2017-12-08  Vladimir Makarov  <vmakarov@redhat.com>
4174         PR rtl-optimization/83317
4175         * lra-constraints.c (process_address_1): Add insn code check.
4177 2017-12-08  Michael Matz  <matz@suse.de>
4179         Fix PR tree-optimization/83323
4180         * gimple-loop-jam (unroll_jam_possible_p): Correct test for
4181         head-controlled loops and loop BBs.
4182         * common.opt (funroll-and-jam): Remove, instead ...
4183         (floop-unroll-and-jam): ... reuse this option.
4184         * opts.c (default_options_table): Use OPT_floop_unroll_and_jam.
4185         * doc/invoke.texi (-funroll-and-jam): Move docu to ...
4186         (-floop-unroll-and-jam): ... this option.
4188 2017-12-08  Jakub Jelinek  <jakub@redhat.com>
4190         * ipa-polymorphic-call.c (noncall_stmt_may_be_vtbl_ptr_store): Fix
4191         a comment typo, get_base_ref_and_offset -> get_ref_base_and_extent.
4192         * ipa-prop.c (stmt_may_be_vtbl_ptr_store): Likewise.
4194 2017-12-08  Richard Biener  <rguenther@suse.de>
4196         PR middle-end/81782
4197         * tree-ssa-uninit.c (warn_uninitialized_vars): Properly
4198         handle accesses outside of zero-sized vars.
4200 2017-12-08  Martin Jambor  <mjambor@suse.cz>
4202         PR tree-optimization/83141
4203         * tree-sra.c (contains_vce_or_bfcref_p): Move up in the file, also
4204         test for MEM_REFs implicitely changing types with padding.  Remove
4205         inline keyword.
4206         (build_accesses_from_assign): Added contains_vce_or_bfcref_p checks.
4208 2017-12-08  Jakub Jelinek  <jakub@redhat.com>
4210         * config/arc/arc.c (arc_attribute_table): Add exclusions to
4211         the comment.
4212         * config/avr/avr.c (avr_attribute_table): Likewise.
4213         * config/msp430/msp430.c (msp430_attribute_table): Likewise.
4214         * config/rl78/rl78.c (rl78_attribute_table): Likewise.
4215         * config/nds32/nds32.c (nds32_attribute_table): Likewise.
4216         * config/darwin.h (SUBTARGET_ATTRIBUTE_TABLE): Initialize new member
4217         of struct attribute_spec.
4218         * config/i386/cygming.h (SUBTARGET_ATTRIBUTE_TABLE): Likewise.
4220 2017-12-08  Ulrich Weigand  <uweigand@de.ibm.com>
4222         PR target/82960
4223         * config/spu/spu.c (pad_bb): Only check INSN_CODE when INSN_P is true.
4225 2017-12-08  Jan Hubicka  <hubicka@ucw.cz>
4227         PR middle-end/83609
4228         * profile-count.c (profile_count::from_gcov_type): Move from
4229         profile-count.h; handle overflow.
4230         * profile-count.h (profile_count::from_gcov_type): Move offline.
4232 2017-12-08  Segher Boessenkool  <segher@kernel.crashing.org>
4234         PR rtl-optimization/83304
4235         * combine.c (move_deaths): If we do not know where a register died,
4236         search for it.
4238 2017-12-08  Richard Biener  <rguenther@suse.de>
4240         * gimple-loop-interchange.cc (tree_loop_interchange::interchange):
4241         Provide -fopt-info-loop feedback when we interchange in a nest.
4243 2017-06-08  Richard Earnshaw  <rearnsha@arm.com>
4245         * config/arm/driver-arm.c (arm_cpu_table): Use fp-capable product names
4246         for armv6 ARM CPU IDs.
4248 2017-06-08  Richard Earnshaw  <rearnsha@arm.com>
4250         * common/config/arm/arm-common.c: Include <algorithm>.
4251         (INCLUDE_VECTOR): Define.
4252         (compare_opt_names): New function.
4253         (arm_rewrite_selected_arch): Only strip out extensions that can be
4254         expressed through -mfpu.  Sort the remaining extensions
4255         alphabetically.
4257 2017-06-08  Richard Earnshaw  <rearnsha@arm.com>
4259         * config/arm/arm.h (arm_asm_auto_mfpu): Declare.
4260         (ASM_CPU_SPEC_FUNCTIONS): Add new rule asm_auto_mfpu.
4261         (ASM_CPU_SPEC): Use it if -mfpu is set to auto.
4262         * common/config/arm/arm-common.c (arm_asm_auto_mfpu): New function.
4264 2017-06-08  Tristan Gingold  <gindold@adacore.com>
4266         PR ada/81470
4267         * dwarf2out.c (dwarf2out_do_cfi_startproc): Only emit
4268         .cfi_personality or .cfi_lsda if the eh data format is dwarf2.
4270 2017-12-08  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
4272         * config/sol2.h (SOLARIS_ATTRIBUTE_TABLE): Initialize new member
4273         of struct attribute_spec.
4275 2017-12-08  Julia Koval  <julia.koval@intel.com>
4277         * config/i386/avx512vnniintrin.h (_mm512_dpwssds_epi32,
4278         _mm512_mask_dpwssds_epi32, _mm512_maskz_dpwssds_epi32): New intrinsics.
4279         * config/i386/avx512vnnivlintrin.h (_mm256_dpwssds_epi32,
4280         _mm256_mask_dpwssds_epi32, _mm256_maskz_dpwssds_epi32,
4281         _mm_dpwssds_epi32, _mm_mask_dpwssds_epi32,
4282         _mm_maskz_dpwssds_epi32): Ditto.
4284 2017-12-08  Richard Biener  <rguenther@suse.de>
4286         PR tree-optimization/81303
4287         * tree-vect-stmts.c (vect_is_simple_cond): For invariant
4288         conditions try to create a comparison vector type matching
4289         the data vector type.
4290         (vectorizable_condition): Adjust.
4291         * tree-vect-patterns.c (vect_recog_mask_conversion_pattern):
4292         Leave invariant conditions alone in case we can vectorize those.
4294 2017-12-08  Julia Koval  <julia.koval@intel.com>
4296         * config/i386/avx512vnniintrin.h (_mm512_dpwssd_epi32,
4297         _mm512_mask_dpwssd_epi32, _mm512_maskz_dpwssd_epi32): New intrinsics.
4298         * config/i386/avx512vnnivlintrin.h (_mm256_dpwssd_epi32,
4299         _mm256_mask_dpwssd_epi32, _mm256_maskz_dpwssd_epi32, _mm_dpwssd_epi32,
4300         _mm_mask_dpwssd_epi32, _mm_maskz_dpwssd_epi32): Ditto.
4302 2017-12-08  Julia Koval  <julia.koval@intel.com>
4304         * config/i386/avx512vnniintrin.h (_mm512_dpbusds_epi32,
4305         _mm512_mask_dpbusds_epi32, _mm512_maskz_dpbusds_epi32): New.
4306         * config/i386/avx512vnnivlintrin.h (_mm256_dpbusds_epi32,
4307         _mm256_mask_dpbusds_epi32, _mm256_maskz_dpbusds_epi32,
4308         _mm_dpbusds_epi32, _mm_mask_dpbusds_epi32,
4309         _mm_maskz_dpbusds_epi32): New intrinsics.
4311 2017-12-07  Sandra Loosemore  <sandra@codesourcery.com>
4313         * config/nios2/nios2.md (ld<bhw_uns>io): Add splitter for memory
4314         operand.
4315         (ld<bh>io_signed): Likewise.
4316         (st<bhw>io): Likewise.
4317         * config/nios2/predicates.md (ldstio_memory_operand): Allow
4318         SMALL_INT12 constant integer operand.
4320 2017-12-07  Sandra Loosemore  <sandra@codesourcery.com>
4322         * config/nios2/nios2.c (nios2_symbolic_constant_allowed):
4323         Rename to...
4324         (nios2_large_constant_allowed): ...this.  Adjust uses.
4325         (nios2_plus_symbolic_constant_p): Rename to...
4326         (nios2_plus_large_constant_p): ...this.  Adjust uses.
4327         (nios2_legitimate_address_p): Correct CONST_INT handling.
4328         (nios2_symbolic_memory_operand_p): Rename to...
4329         (nios2_large_constant_memory_operand_p): ...this.  Adjust uses.
4330         (nios2_large_constant_p): Check for large constant integers too.
4331         (nios2_split_large_constant): Handle constant integers.
4332         (nios2_split_symbolic_memory_operand): Rename to...
4333         (nios2_split_large_constant_memory_operand): ...this.  Adjust uses.
4334         (nios2_legitimize_constant_address): Handle constant integers.
4335         (r0rel_constant_p): Handle small constant integers.
4336         (nios2_print_operand_address): Handle r0-relative integer addresses.
4337         * config/nios2/nios2-protos.h: Adjust for renamed functions.
4338         * config/nios2/nios2.md: Adjust for renamed functions.
4340 2017-12-07  Andrew Waterman  <andrew@sifive.com>
4342         * config/riscv/riscv.c (TARGET_ASM_SELECT_SECTION): New define.
4343         (TARGET_HAVE_SRODATA_SECTION): New define.
4344         (riscv_select_section): New function.
4346 2017-12-08  Joseph Myers  <joseph@codesourcery.com>
4347             Alexander Monakov  <amonakov@ispras.ru>
4348             Jakub Jelinek  <jakub@redhat.com>
4350         PR target/81906
4351         * config/i386/i386.c (ix86_expand_rint): Handle flag_rounding_math.
4353 2017-12-07  Richard Sandiford  <richard.sandiford@linaro.org>
4355         * config/aarch64/aarch64.c (aarch64_print_address_internal): Return
4356         a bool success value.  Don't call output_operand_lossage here.
4357         (aarch64_print_ldpstp_address): Return a bool success value.
4358         (aarch64_print_operand_address): Call output_addr_const if
4359         aarch64_print_address_internal fails.
4360         (aarch64_print_operand): Don't assert that the mode is 16 bytes for
4361         'y'; call output_operand_lossage instead.  Call output_operand_lossage
4362         if aarch64_print_ldpstp_address fails.
4364 2017-12-07  Richard Sandiford  <richard.sandiford@linaro.org>
4366         * tree-vector-builder.h
4367         (tree_vector_builder::binary_encoded_nelts): Declare.
4368         * tree-vector-builder.c
4369         (tree_vector_builder::binary_encoded_nelts): New function.
4370         * fold-const.c (negate_expr_p): Likewise.
4371         (operand_equal_p, fold_checksum_tree): Likewise.
4372         * tree-loop-distribution.c (const_with_all_bytes_same): Likewise.
4373         * tree.c (integer_zerop, integer_onep, integer_all_onesp, real_zerop)
4374         (real_onep, real_minus_onep, add_expr, initializer_zerop): Likewise.
4375         (uniform_vector_p): Likewise.
4376         * varasm.c (const_hash_1, compare_constant): Likewise.
4377         * tree-ssa-ccp.c: Include tree-vector-builder.h.
4378         (valid_lattice_transition): Operate directly on the VECTOR_CST
4379         encoding.
4380         * ipa-icf.c: Include tree-vector-builder.h.
4381         (sem_variable::equals): Operate directly on the VECTOR_CST encoding.
4382         * print-tree.c (print_node): Print encoding of VECTOR_CSTs.
4384 2017-12-07  Richard Sandiford  <richard.sandiford@linaro.org>
4386         * tree.c (build_vector): Delete.
4387         * tree.h (build_vector): Make static and move into the self-testing
4388         block.
4390 2017-12-07  Richard Sandiford  <richard.sandiford@linaro.org>
4392         * vector-builder.h (vector_builder::derived): New const overload.
4393         (vector_builder::elt): New function.
4394         * tree-vector-builder.h (tree_vector_builder::type): New function.
4395         (tree_vector_builder::apply_step): Declare.
4396         * tree-vector-builder.c (tree_vector_builder::apply_step): New
4397         function.
4398         * gimple-fold.h (tree_vector_builder): Declare.
4399         (gimple_build_vector): Take a tree_vector_builder instead of a
4400         type and vector of elements.
4401         * gimple-fold.c (gimple_build_vector): Likewise.
4402         * tree-vect-loop.c (get_initial_def_for_reduction): Update call
4403         accordingly.
4404         (get_initial_defs_for_reduction): Likewise.
4405         (vectorizable_induction): Likewise.
4407 2017-12-07  Richard Sandiford  <richard.sandiford@linaro.org>
4409         * tree-vector-builder.h
4410         (tree_vector_builder::new_binary_operation): Declare.
4411         * tree-vector-builder.c
4412         (tree_vector_builder::new_binary_operation): New function.
4413         * fold-const.c (fold_relational_const): Use it.
4414         (const_binop): Likewise.  Check that both input vectors have
4415         the same number of elements, thus excluding things like WIDEN_SUM.
4416         Check whether it is possible to operate directly on the encodings
4417         of stepped inputs.
4419 2017-12-07  Richard Sandiford  <richard.sandiford@linaro.org>
4421         * fold-const.c (fold_negate_expr_1): Use tree_vector_builder and
4422         new_unary_operation, operating only on the encoded elements.
4423         (const_unop): Likewise.
4424         (exact_inverse): Likewise.
4425         (distributes_over_addition_p): New function.
4426         (const_binop): Use tree_vector_builder and new_unary_operation
4427         for combinations of VECTOR_CST and INTEGER_CST.  Operate only
4428         on the encoded elements unless the encoding is strided and the
4429         operation does not distribute over addition.
4430         (fold_convert_const):  Use tree_vector_builder and
4431         new_unary_operation.  Operate only on the encoded elements
4432         for truncating integer conversions, or for non-stepped encodings.
4434 2017-12-07  Richard Sandiford  <richard.sandiford@linaro.org>
4436         * config/sparc/sparc.c: Include tree-vector-builder.h.
4437         (sparc_fold_builtin): Use tree_vector_builder instead of build_vector.
4438         * expmed.c: Include tree-vector-builder.h.
4439         (make_tree): Use tree_vector_builder instead of build_vector.
4440         * fold-const.c: Include tree-vector-builder.h.
4441         (const_binop): Use tree_vector_builder instead of build_vector.
4442         (const_unop): Likewise.
4443         (native_interpret_vector): Likewise.
4444         (fold_vec_perm): Likewise.
4445         (fold_ternary_loc): Likewise.
4446         * gimple-fold.c: Include tree-vector-builder.h.
4447         (gimple_fold_stmt_to_constant_1): Use tree_vector_builder instead
4448         of build_vector.
4449         * tree-ssa-forwprop.c: Include tree-vector-builder.h.
4450         (simplify_vector_constructor): Use tree_vector_builder instead
4451         of build_vector.
4452         * tree-vect-generic.c: Include tree-vector-builder.h.
4453         (add_rshift): Use tree_vector_builder instead of build_vector.
4454         (expand_vector_divmod): Likewise.
4455         (optimize_vector_constructor): Likewise.
4456         * tree-vect-loop.c: Include tree-vector-builder.h.
4457         (vect_create_epilog_for_reduction): Use tree_vector_builder instead
4458         of build_vector.  Explicitly use a stepped encoding for
4459         { 1, 2, 3, ... }.
4460         * tree-vect-slp.c: Include tree-vector-builder.h.
4461         (vect_get_constant_vectors): Use tree_vector_builder instead
4462         of build_vector.
4463         (vect_transform_slp_perm_load): Likewise.
4464         (vect_schedule_slp_instance): Likewise.
4465         * tree-vect-stmts.c: Include tree-vector-builder.h.
4466         (vectorizable_bswap): Use tree_vector_builder instead of build_vector.
4467         (vect_gen_perm_mask_any): Likewise.
4468         (vectorizable_call): Likewise.  Explicitly use a stepped encoding.
4469         * tree.c: (build_vector_from_ctor): Use tree_vector_builder instead
4470         of build_vector.
4471         (build_vector_from_val): Likewise.  Explicitly use a duplicate
4472         encoding.
4474 2017-12-07  Richard Sandiford  <richard.sandiford@arm.com>
4476         * doc/generic.texi (VECTOR_CST): Describe new representation of
4477         vector constants.
4478         * vector-builder.h: New file.
4479         * tree-vector-builder.h: Likewise.
4480         * tree-vector-builder.c: Likewise.
4481         * Makefile.in (OBJS): Add tree-vector-builder.o.
4482         * tree.def (VECTOR_CST): Update comment to refer to generic.texi.
4483         * tree-core.h (tree_base): Add a vector_cst field to the u union.
4484         (tree_vector): Change the number of elements to
4485         vector_cst_encoded_nelts.
4486         * tree.h (VECTOR_CST_NELTS): Redefine using TYPE_VECTOR_SUBPARTS.
4487         (VECTOR_CST_ELTS): Delete.
4488         (VECTOR_CST_ELT): Redefine using vector_cst_elt.
4489         (VECTOR_CST_LOG2_NPATTERNS, VECTOR_CST_NPATTERNS): New macros.
4490         (VECTOR_CST_NELTS_PER_PATTERN, VECTOR_CST_DUPLICATE_P): Likewise.
4491         (VECTOR_CST_STEPPED_P, VECTOR_CST_ENCODED_ELTS): Likewise.
4492         (VECTOR_CST_ENCODED_ELT): Likewise.
4493         (vector_cst_encoded_nelts): New function.
4494         (make_vector): Take the values of VECTOR_CST_LOG2_NPATTERNS and
4495         VECTOR_CST_NELTS_PER_PATTERN as arguments.
4496         (vector_cst_int_elt, vector_cst_elt): Declare.
4497         * tree.c: Include tree-vector-builder.h.
4498         (tree_code_size): Abort if passed VECTOR_CST.
4499         (tree_size): Update for new VECTOR_CST layout.
4500         (make_vector): Take the values of VECTOR_CST_LOG2_NPATTERNS and
4501         VECTOR_CST_NELTS_PER_PATTERN as arguments.
4502         (build_vector): Use tree_vector_builder.
4503         (vector_cst_int_elt, vector_cst_elt): New functions.
4504         (drop_tree_overflow): For VECTOR_CST, drop the TREE_OVERFLOW from the
4505         encoded elements and then create the vector in the canonical form.
4506         (check_vector_cst, check_vector_cst_duplicate, check_vector_cst_fill)
4507         (check_vector_cst_stepped, test_vector_cst_patterns): New functions.
4508         (tree_c_tests): Call test_vector_cst_patterns.
4509         * lto-streamer-out.c (DFS::DFS_write_tree_body): Handle the new
4510         VECTOR_CST fields.
4511         (hash_tree): Likewise.
4512         * tree-streamer-out.c (write_ts_vector_tree_pointers): Likewise.
4513         (streamer_write_tree_header): Likewise.
4514         * tree-streamer-in.c (lto_input_ts_vector_tree_pointers): Likewise.
4515         (streamer_alloc_tree): Likewise.  Update call to make_vector.
4516         * fold-const.c (fold_ternary_loc): Avoid using VECTOR_CST_ELTS.
4518 2017-12-07  Richard Sandiford  <richard.sandiford@linaro.org>
4520         * selftest.h (ASSERT_TRUE_AT, ASSERT_FALSE_AT, ASSERT_EQ_AT)
4521         (ASSERT_NE, ASSERT_PRED1): Add underscores to local variable names
4522         * selftest-rtl.h (ASSERT_RTX_EQ, ASSERT_RTX_PTR_EQ): Likewise.
4524 2017-12-07  Bin Cheng  <bin.cheng@arm.com>
4525             Richard Biener  <rguenther@suse.de>
4527         PR tree-optimization/81303
4528         * Makefile.in (gimple-loop-interchange.o): New object file.
4529         * common.opt (floop-interchange): Reuse the option from graphite.
4530         * doc/invoke.texi (-floop-interchange): Ditto.  New document for
4531         -floop-interchange and mention it for -O3.
4532         * opts.c (default_options_table): Enable -floop-interchange at -O3.
4533         * gimple-loop-interchange.cc: New file.
4534         * params.def (PARAM_LOOP_INTERCHANGE_MAX_NUM_STMTS): New parameter.
4535         (PARAM_LOOP_INTERCHANGE_STRIDE_RATIO): New parameter.
4536         * passes.def (pass_linterchange): New pass.
4537         * timevar.def (TV_LINTERCHANGE): New time var.
4538         * tree-pass.h (make_pass_linterchange): New declaration.
4539         * tree-ssa-loop-ivcanon.c (create_canonical_iv): Change to external
4540         interchange.  Record IV before/after increment in new parameters.
4541         * tree-ssa-loop-ivopts.h (create_canonical_iv): New declaration.
4542         * tree-vect-loop.c (vect_is_simple_reduction): Factor out reduction
4543         path check into...
4544         (check_reduction_path): ...New function here.
4545         * tree-vectorizer.h (check_reduction_path): New declaration.
4547 2017-12-07  Vladimir Makarov  <vmakarov@redhat.com>
4549         PR target/83252
4550         PR rtl-optimization/80818
4551         * lra.c (add_regs_to_insn_regno_info): Make a hard reg in CLOBBER
4552         always early clobbered.
4553         * lra-lives.c (process_bb_lives): Check input hard regs for early
4554         clobbered non-operand hard reg.
4556 2017-12-07  Jakub Jelinek  <jakub@redhat.com>
4558         PR middle-end/83164
4559         * tree-cfg.c (verify_gimple_assign_binary): Don't require
4560         types_compatible_p, just that TYPE_MODE is the same.
4562 2017-12-07  Martin Sebor  <msebor@redhat.com>
4564         PR c/81544
4565         * attribs.c (empty_attribute_table): Initialize new member of
4566         struct attribute_spec.
4567         (decl_attributes): Add argument.  Handle mutually exclusive
4568         combinations of attributes.
4569         (selftests::test_attribute_exclusions): New function.
4570         (selftests::attribute_c_tests): Ditto.
4571         * attribs.h (decl_attributes): Add default argument.
4572         * selftest.h (attribute_c_tests): Declare.
4573         * selftest-run-tests.c (selftest::run_tests): Call attribute_c_tests.
4574         * tree-core.h (attribute_spec::exclusions, exclude): New type and
4575         member.
4576         * doc/extend.texi (Common Function Attributes): Update const and pure.
4577         * config/alpha/alpha.c (vms_attribute_table): Initialize new member
4578         of struct attribute_spec.
4579         * config/arc/arc.c (arc_attribute_table): Same.
4580         * config/arm/arm.c (arm_attribute_table): Same.
4581         * config/avr/avr.c ( avr_attribute_table): Same.
4582         * config/bfin/bfin.c (bfin_attribute_table): Same.
4583         * config/cr16/cr16.c (cr16_attribute_table): Same.
4584         * config/epiphany/epiphany.c (epiphany_attribute_table): Same.
4585         * config/h8300/h8300.c (h8300_attribute_table): Same.
4586         * config/i386/i386.c (ix86_attribute_table): Same.
4587         * config/ia64/ia64.c (ia64_attribute_table): Same.
4588         * config/m32c/m32c.c (m32c_attribute_table): Same.
4589         * config/m32r/m32r.c (m32r_attribute_table): Same.
4590         * config/m68k/m68k.c (m68k_attribute_table): Same.
4591         * config/mcore/mcore.c (mcore_attribute_table): Same.
4592         * config/microblaze/microblaze.c (microblaze_attribute_table): Same.
4593         * config/mips/mips.c (mips_attribute_table): Same.
4594         * config/msp430/msp430.c (msp430_attribute_table): Same.
4595         * config/nds32/nds32.c (nds32_attribute_table): Same.
4596         * config/nvptx/nvptx.c (nvptx_attribute_table): Same.
4597         * config/powerpcspe/powerpcspe.c (rs6000_attribute_table): Same.
4598         * config/rl78/rl78.c (rl78__attribute_table): Same.
4599         * config/rs6000/rs6000.c (rs6000_attribute_table): Same.
4600         * onfig/rx/rx.c (rx_attribute_table): Same.
4601         * config/s390/s390.c (s390_handle_vectorbool_attribute): Same.
4602         * config/sh/sh.c (sh_attribute_table): Same.
4603         * config/sparc/sparc.c (sparc_attribute_table): Same.
4604         * config/spu/spu.c (spu_attribute_table): Same.
4605         * config/stormy16/stormy16.c (xstormy16_attribute_table): Same.
4606         * config/v850/v850.c (v850_attribute_table): Same.
4607         * config/visium/visium.c (visium_attribute_table): Same.
4609 2017-12-07  Tamar Christina  <tamar.christina@arm.com>
4611         PR target/82641
4612         * config/arm/arm.c (INCLUDE_STRING): Define.
4613         (arm_last_printed_arch_string, arm_last_printed_fpu_string): New.
4614         (arm_declare_function_name): Conservatively emit .arch, .arch_extensions
4615         and .fpu.
4617 2017-12-07  Michael Matz  <matz@suse.de>
4619         Add unroll and jam pass
4621         * gimple-loop-jam.c: New file.
4622         * Makefile.in (OBJS): Add gimple-loop-jam.o.
4623         * common.opt (funroll-and-jam): New option.
4624         * opts.c (default_options_table): Add unroll-and-jam at -O3.
4625         * params.def (PARAM_UNROLL_JAM_MIN_PERCENT): New param.
4626         (PARAM_UNROLL_JAM_MAX_UNROLL): Ditto.
4627         * passes.def: Add pass_loop_jam.
4628         * timevar.def (TV_LOOP_JAM): Add.
4629         * tree-pass.h (make_pass_loop_jam): Declare.
4630         * cfgloop.c (flow_loop_tree_node_add): Add AT argument.
4631         * cfgloop.h (flow_loop_tree_node_add): Adjust declaration.
4632         * cfgloopmanip.c (duplicate_loop): Add AT argument, adjust call
4633         to flow_loop_tree_node_add.
4634         (duplicate_subloops, copy_loops_to): Append to sibling list.
4635         * cfgloopmanip.h: (duplicate_loop): Adjust declaration.
4636         * doc/invoke.texi (-funroll-and-jam): Document new option.
4637         (unroll-jam-min-percent, unroll-jam-max-unroll): Document new params.
4639 2017-12-07  Richard Biener  <rguenther@suse.de>
4641         PR tree-optimization/83296
4642         PR tree-optimization/67769
4643         * tree-ssa-phiopt.c (conditional_replacement): Do not reset
4644         flow sensitive info in an unrelated BB.
4645         (value_replacement): Use reset_flow_sensitive_info.
4646         (minmax_replacement): Reset flow sensitive info on the def
4647         we move.  Do not reset flow sensitive info in the whole BB
4648         we move the stmt to.
4649         (abs_replacement): Likewise.
4651 2017-12-07  Segher Boessenkool  <segher@kernel.crashing.org>
4653         PR target/43871
4654         * config/rs6000/rs6000.c (rs6000_option_override_internal): Set
4655         rs6000_cpu to the given -mcpu=, or to the default processor.
4657 2017-12-07  Segher Boessenkool  <segher@kernel.crashing.org>
4659         * config/rs6000/rs6000.h (rs6000_cpu_attr): Delete.
4660         * config/rs6000/rs6000.c (rs6000_variable_issue_1): Use rs6000_tune
4661         instead of rs6000_cpu_attr.
4662         (rs6000_adjust_cost): Ditto.
4663         (is_microcoded_insn): Ditto.
4664         (rs6000_adjust_priority): Ditto.
4665         (rs6000_issue_rate): Ditto.
4666         (rs6000_use_sched_lookahead): Ditto.
4667         (rs6000_use_sched_lookahead_guard): Ditto.
4668         (rs6000_sched_reorder): Ditto.
4669         (force_new_group): Ditto.
4670         * config/rs6000/rs6000.md (cpu attribute): Ditto.
4671         (group_ending_nop): Ditto.
4673 2017-12-07  Segher Boessenkool  <segher@kernel.crashing.org>
4675         * config/rs6000/rs6000.opt (rs6000_tune): New variable.
4676         * config/rs6000/rs6000.c (rs6000_option_override_internal): Also set
4677         rs6000_tune.  Use rs6000_tune instead of rs6000_cpu where appropriate.
4678         (rs6000_loop_align): Use rs6000_tune instead of rs6000_cpu where
4679         appropriate.
4680         (rs6000_reassociation_width): Ditto.
4681         (rs6000_emit_epilogue): Ditto.
4682         (rs6000_adjust_cost): Ditto.
4683         (is_microcoded_insn): Ditto.
4684         (is_cracked_insn): Ditto.
4685         (rs6000_adjust_priority): Ditto.
4686         (rs6000_sched_reorder): Ditto.
4687         (rs6000_sched_reorder2): Ditto.
4688         (insn_must_be_first_in_group): Ditto.
4689         (insn_must_be_last_in_group): Ditto.
4690         (rs6000_register_move_cost): Ditto.
4691         * config/rs6000/rs6000.h (rs6000_cpu_attr): Use rs6000_tune instead of
4692         rs6000_cpu.
4694 2017-12-07  Julia Koval  <julia.koval@intel.com>
4696         * config.gcc: Add vaesintrin.h.
4697         * config/i386/i386-builtin-types.def (V64QI_FTYPE_V64QI_V64QI):
4698         New type.
4699         * config/i386/i386-builtin.def (__builtin_ia32_vaesdec_v16qi,
4700         __builtin_ia32_vaesdec_v32qi, __builtin_ia32_vaesdec_v64qi):
4701         New builtins.
4702         * config/i386/i386.c (ix86_expand_args_builtin): Handle new type.
4703         * config/i386/immintrin.h: Include vaesintrin.h.
4704         * config/i386/sse.md (vaesdec_<mode>): New pattern.
4705         * config/i386/vaesintrin.h (_mm256_aesdec_epi128, _mm512_aesdec_epi128,
4706         _mm_aesdec_epi128): New intrinsics.
4708 2017-12-06  David Malcolm  <dmalcolm@redhat.com>
4710         * Makefile.in (C_COMMON_OBJS): Add c-family/c-spellcheck.o.
4711         * spellcheck-tree.c (find_closest_macro_cpp_cb): Move to
4712         c-family/c-spellcheck.cc.
4713         (best_macro_match::best_macro_match): Likewise.
4714         * spellcheck-tree.h
4715         (struct edit_distance_traits<cpp_hashnode *>): Move to
4716         c-family/c-spellcheck.h.
4717         (class best_macro_match): Likewise.
4719 2017-12-06  Jakub Jelinek  <jakub@redhat.com>
4721         PR tree-optimization/83293
4722         * gimple-ssa-strength-reduction.c (insert_initializers): Use
4723         GSI_NEW_STMT instead of GSI_SAME_STMT in gsi_insert_after that
4724         might insert into empty bb.
4726         PR sanitizer/81281
4727         * match.pd ((T)(P + A) - (T)P -> (T) A): Split into separate
4728         simplify for plus with :c added, and pointer_plus without that.
4729         ((T)P - (T)(P + A) -> -(T) A): Likewise.  If type is integral
4730         with undefined overflow and the conversion is not widening,
4731         perform negation in utype and only convert to type afterwards.
4732         ((T)(P + A) - (T)(P + B) -> (T)A - (T)B): Split into separate
4733         simplify for plus with :c added, and pointer_plus without that.
4734         If type is integral with undefined overflow and the conversion is
4735         not widening, perform minus in utype and only convert to type
4736         afterwards.  Move the last pointer_diff_expr simplify into the
4737         two outermost ifs.
4739 2017-12-06  Martin Sebor  <msebor@redhat.com>
4741         PR tree-optimization/82646
4742         * builtins.c (maybe_emit_chk_warning): Use size as the bound for
4743         strncpy, not maxlen.
4745 2017-12-06  Martin Sebor  <msebor@redhat.com>
4747         * doc/invoke.texi (-Wstringop-truncation): Mention attribute
4748         nonstring.
4750         PR tree-optimization/83075
4751         * tree-ssa-strlen.c (handle_builtin_stxncpy): Avoid assuming
4752         strncat/strncpy don't change length of source string.
4754 2017-12-06  Eric Botcazou  <ebotcazou@adacore.com>
4756         Revert
4757         2017-11-29  Martin Aberg  <maberg@gaisler.com>
4759         * config/sparc/sparc.md (divdf3_fix): Add NOP and adjust length
4760         to prevent b2bst errata sequence.
4761         (sqrtdf2_fix): Likewise.
4763 2017-12-06  Jakub Jelinek  <jakub@redhat.com>
4765         PR tree-optimization/81945
4766         * cfgloop.h (FOR_EACH_LOOP_FN): Use FN instead of hardcoding fn.
4767         * tree-cfg.c (move_sese_region_to_fn): If any of the loops moved
4768         to dest_cfun has orig_loop_num set, either remap it to the new
4769         loop number if the loop got moved too, or clear it.
4771 2017-12-05  Steve Ellcey  <sellcey@cavium.com>
4773         * config/aarch64/thunderx2-t99.md (thunderx2t99_branch): Add trap
4774         to reservation.
4775         (thunderx2t99_nothing): New insn reservation.
4776         (thunderx2t99_mrs): New insn reservation.
4777         (thunderx2t99_multiple): New insn reservation.
4778         (thunderx2t99_alu_basi): Add bfx to reservation.
4779         (thunderx2t99_fp_cmp): Add fccmps and fccmpd to reservation.
4781 2017-12-05  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
4783         PR target/82248
4784         * config/arm/arm.md (probe_stack) : Use the 'o' constraint.
4786 2017-12-05  Bin Cheng  <bin.cheng@arm.com>
4788         * tree-ssa-dce.c (simple_dce_from_worklist): Move and rename from
4789         tree-ssa-pre.c::remove_dead_inserted_code.
4790         * tree-ssa-dce.h: New file.
4791         * tree-ssa-pre.c (tree-ssa-dce.h): Include new header file.
4792         (remove_dead_inserted_code): Move and rename to function
4793         tree-ssa-dce.c::simple_dce_from_worklist.
4794         (pass_pre::execute): Update use.
4796 2017-12-05  Richard Biener  <rguenther@suse.de>
4798         PR tree-optimization/83277
4799         * graphite-isl-ast-to-gimple.c (should_copy_to_new_region): Make sure
4800         to code-gen liveout vars.
4802 2017-12-05  Richard Sandiford  <richard.sandiford@linaro.org>
4804         * config/aarch64/aarch64-simd.md (aarch64_simd_bsldi_internal)
4805         (aarch64_simd_bsldi_alt): Check REG_P before GP_REGNUM_P.
4806         (aarch64_cm<optab>di, aarch64_cmtstdi): Add leading "&&" to
4807         split condition.
4809 2017-12-05  Max Filippov  <jcmvbkbc@gmail.com>
4811         * config/xtensa/xtensa.c (xtensa_asan_shadow_offset): New
4812         function.
4813         (TARGET_ASAN_SHADOW_OFFSET): New macro definition.
4814         * config/xtensa/xtensa.h (FRAME_GROWS_DOWNWARD): Set to 1 if
4815         ASAN is enabled.
4817 2017-12-05  Richard Biener  <rguenther@suse.de>
4819         * timevar.def (TV_TREE_RECIP, TV_TREE_SINCOS, TV_TREE_WIDEN_MUL): Add.
4820         * tree-ssa-math-opts.c (pass_data_cse_reciprocal): Use TV_TREE_RECIP.
4821         (pass_data_cse_sincos): Use TV_TREE_SINCOS.
4822         (pass_data_optimize_widening_mul): Use TV_TREE_WIDEN_MUL.
4824 2017-12-05  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
4826         * dbxout.c (dbxout_block): Grow buf to 30 bytes.
4828 2017-12-05  Martin Liska  <mliska@suse.cz>
4829             Jakub Jelinek  <jakub@redhat.com>
4831         * doc/invoke.texi: Document the options.
4832         * flag-types.h (enum sanitize_code): Add
4833         SANITIZE_POINTER_COMPARE and SANITIZE_POINTER_SUBTRACT.
4834         * ipa-inline.c (sanitize_attrs_match_for_inline_p): Add handling
4835         of SANITIZE_POINTER_COMPARE and SANITIZE_POINTER_SUBTRACT.
4836         * opts.c: Define new sanitizer options.
4837         * sanitizer.def (BUILT_IN_ASAN_POINTER_COMPARE): Likewise.
4838         (BUILT_IN_ASAN_POINTER_SUBTRACT): Likewise.
4840 2017-12-05  Julia Koval  <julia.koval@intel.com>
4842         * common/config/i386/i386-common.c (OPTION_MASK_ISA_AVX512VNNI_SET,
4843         OPTION_MASK_ISA_AVX512VNNI_UNSET): New.
4844         (ix86_handle_option): Handle -mavx512vnni.
4845         * config/i386/cpuid.h (bit_AVX512VNNI): New bit.
4846         * config/i386/driver-i386.c (host_detect_local_cpu): Handle new bit.
4847         * config/i386/i386-c (__AVX512VNNI__): New.
4848         * config/i386/i386.c (ix86_target_string): Handle new option.
4849         (ix86_valid_target_attribute_inner_p): Handle new option.
4850         * config/i386/i386.h (TARGET_AVX512VNNI, TARGET_AVX512VNNI_P): New.
4851         * config/i386/i386.opt (mavx512vnni): New option.
4853 2017-12-01  Jan Hubicka  <hubicka@ucw.cz>
4855         PR target/81616
4856         * athlon.md: Disable for generic.
4857         * haswell.md: Enable for generic.
4858         * i386.c (ix86_sched_init_global): Add core hooks for generic.
4859         * x86-tune-sched.c (ix86_issue_rate): Increase issue rate for generic
4860         to 4.
4861         (ix86_adjust_cost): Move generic to haswell path.
4863 2017-12-04  Eric Botcazou  <ebotcazou@adacore.com>
4865         * config/sparc/sparc.c (sparc_do_work_around_errata): Use mem_ref
4866         instead of MEM_P in a couple more places.  Fix formatting issues.
4868 2017-12-04  Jim Wilson  <jimw@sifive.com>
4870         * config/riscv/riscv.c (riscv_for_each_saved_reg): Use GP_REG_LAST
4871         instead of GP_REG_LAST-1.
4872         (riscv_adjust_libcall_cfi_prologue): Likewise.
4873         (riscv_adjust_libcall_cri_epilogue): Likewise.
4874         * config/riscv/riscv.h (CALL_USED_REGISTERS): Change a6 to t6 in
4875         comment.
4877 2017-12-04  Luis Machado  <luis.machado@linaro.org>
4879         * ipa-pure-const.c (check_decl): Add missing newline.
4880         (state_from_flags): Likewise.
4882 2017-12-04  Jeff Law  <law@redhat.com>
4884         PR tree-optimizatin/78496
4885         * gimple-ssa-evrp-analyze.h
4886         (evrp_range_analyzer::get_vr_values): Simplify.
4887         * gimple-ssa-evrp-analyze.c: Corresponding changes.
4888         * tree-ssa-dom.c: Include alloc-pool.h, tree-vrp.h, vr-values.h
4889         and gimple-ssa-evrp-analyze.h.
4890         (dom_opt_dom_walker class): Add evrp_range_analyzer member.
4891         (simplify_stmt_for_jump_threading): Copy a blob of code from
4892         tree-vrp.c to use ranges to simplify statements.
4893         (dom_opt_dom_walker::before_dom_children): Call
4894         evrp_range_analyzer::{enter,record_ranges_from_stmt} methods.
4895         (dom_opt_dom_walker::after_dom_children): Similarly for
4896         evrp_range_analyzer::leave.
4897         (dom_opt_dom_walker::optimize_stmt): Use EVRP ranges to optimize
4898         conditionals.
4900         * gimple-ssa-evrp-analyze.c
4901         (evrp_range_analyzer::extract_range_from_stmt):  Always use
4902         vr_values::update_value_range so preexisting range info is
4903         medged with new range info, even if the new range is VR_VARYING.
4905 2017-12-04  Segher Boessenkool  <segher@kernel.crashing.org>
4907         * combine.c: Adjust comment.
4908         (use_crosses_set_p): Delete.
4909         (can_combine_p): Use modified_between_p instead of use_crosses_set_p.
4910         (try_combine): Ditto.
4912 2017-12-04  Richard Biener  <rguenther@suse.de>
4914         PR tree-optimization/83255
4915         * graphite-isl-ast-to-gimple.c (translate_isl_ast_node_for):
4916         Re-add zero-iteration check.
4918 2017-12-04  Segher Boessenkool  <segher@kernel.crashing.org>
4920         PR rtl-optimization/83245
4921         * lra.c (collect_non_operand_hard_regs): Treat clobbers of non-operand
4922         hard registers as earlyclobber, also if not in an asm.
4924 2017-12-04  Segher Boessenkool  <segher@kernel.crashing.org>
4926         PR bootstrap/83265
4927         Revert
4928         2017-12-01  Segher Boessenkool  <segher@kernel.crashing.org>
4930         PR target/43871
4931         * config/rs6000/rs6000.c (rs6000_option_override_internal): Set
4932         rs6000_cpu based on cpu_index, not tune_index.
4934 2017-12-04  Richard Biener  <rguenther@suse.de>
4936         PR tree-optimization/83238
4937         * graphite-scop-detection.c (scop_detection::merge_sese): Make
4938         code match comment, rejecting invalid SESE regions.
4940 2017-12-03  John David Anglin  <danglin@gcc.gnu.org>
4942         * config/pa/pa.c (pa_legitimate_address_p): For scaled indexing,
4943         require base operand is a REG_POINTER prior to reload on targets
4944         with non-equivalent space registers.
4946 2017-12-01  Jan Hubicka  <hubicka@ucw.cz>
4948         * ipa-cp.c (ipcp_lattice<valtype>::print): Update dumping.
4949         (update_specialized_profile): Fix updating of counts.
4950         (perhaps_add_new_callers): Likewise.
4952 2017-12-01  Jan Hubicka  <hubicka@ucw.cz>
4954         PR target/81616
4955         * x86-tune.def: Remove obsolette FIXMEs.
4956         (X86_TUNE_PARTIAL_FLAG_REG_STALL): Disable for generic
4957         (X86_TUNE_FUSE_CMP_AND_BRANCH_32, X86_TUNE_FUSE_CMP_AND_BRANCH_64,
4958         X86_TUNE_FUSE_CMP_AND_BRANCH_SOFLAGS, X86_TUNE_FUSE_ALU_AND_BRANCH):
4959         Enable for generic.
4960         (X86_TUNE_PAD_RETURNS): Disable for generic.
4962 2017-12-02  Jakub Jelinek  <jakub@redhat.com>
4964         PR tree-optimization/83170
4965         PR tree-optimization/83241
4966         * gimple-ssa-store-merging.c
4967         (imm_store_chain_info::try_coalesce_bswap): Update vuse field from
4968         gimple_vuse (ins_stmt) in case it has changed.
4969         (imm_store_chain_info::output_merged_store): Likewise.
4971         * tree-chkp.c (chkp_compute_bounds_for_assignment): Handle
4972         POINTER_DIFF_EXPR.
4974         PR c++/81212
4975         * tree-cfg.c (pass_warn_function_return::execute): Handle
4976         __builtin_ubsan_handle_missing_return like __builtin_unreachable
4977         with BUILTINS_LOCATION.
4979         PR target/78643
4980         PR target/80583
4981         * expr.c (get_inner_reference): If DECL_MODE of a non-bitfield
4982         is BLKmode for vector field with vector raw mode, use TYPE_MODE
4983         instead of DECL_MODE.
4985         * config/i386/i386-protos.h (standard_sse_constant_opcode): Change
4986         last argument to rtx pointer.
4987         * config/i386/i386.c (standard_sse_constant_opcode): Replace X argument
4988         with OPERANDS.  For AVX+ 128-bit VEX encoded instructions over 256-bit
4989         or 512-bit.  If setting EXT_REX_SSE_REG_P, use EVEX encoded insn
4990         depending on the chosen ISAs.
4991         * config/i386/i386.md (*movxi_internal_avx512f, *movoi_internal_avx,
4992         *movti_internal, *movdi_internal, *movsi_internal, *movtf_internal,
4993         *movdf_internal, *movsf_internal): Adjust standard_sse_constant_opcode
4994         callers.
4995         * config/i386/sse.md (mov<mode>_internal): Likewise.
4996         * config/i386/mmx.md (*mov<mode>_internal): Likewise.
4998 2017-12-01  Segher Boessenkool  <segher@kernel.crashing.org>
5000         * doc/invoke.texi (-dp): Say that instruction cost is printed as well.
5002 2017-12-01  Segher Boessenkool  <segher@kernel.crashing.org>
5004         * config/rs6000/rs6000-protos.h (emit_fusion_addis): Remove last two
5005         parameters from prototype.
5006         * config/rs6000/rs6000.c (emit_fusion_addis): Remove last two
5007         parameters.  Don't print a comment.
5008         (emit_fusion_gpr_load): Adjust.
5009         (emit_fusion_load_store): Adjust.
5010         * config/rs6000/rs6000.md (*fusion_p9_<mode>_constant): Adjust.
5011         * config/rs6000/vsx.md (two peepholes): Print the "vector load fusion"
5012         comment on the second line.
5014 2017-12-01  Segher Boessenkool  <segher@kernel.crashing.org>
5016         PR target/43871
5017         * config/rs6000/rs6000.c (rs6000_option_override_internal): Set
5018         rs6000_cpu based on cpu_index, not tune_index.
5020 2017-12-01  Segher Boessenkool  <segher@kernel.crashing.org>
5022         * final.c (output_asm_name): Print insn_cost.  Shorten output.  Print
5023         which_alternative instead of which_alternative + 1.
5024         (output_asm_insn): Print an extra tab if the template is short.
5026 2017-12-01  Jim Wilson  <jimw@sifive.com>
5028         * common.opt (use_gnu_debug_info_extensions): Delete DWARF_DEBUG from
5029         comment.
5030         * config/vx-common.h (DWARF_DEBUGGING_INFO): Delete undef.
5031         * doc/tm.texi.in (PREFERRED_DEBUGGING_TYPE): Delete DWARF_DEBUG
5032         reference.
5033         * doc/tm.texi: Regenerate.
5035 2017-12-01  Michael Meissner  <meissner@linux.vnet.ibm.com>
5037         PR target/81959
5038         * config/rs6000/rs6000.c (rs6000_address_for_fpconvert): Check for
5039         whether we can allocate pseudos before trying to fix an address.
5040         * config/rs6000/rs6000.md (float_<mode>si2_hw): Make sure the
5041         memory address is indexed or indirect.
5042         (floatuns_<mode>si2_hw2): Likewise.
5044 2017-12-01  Jason Merrill  <jason@redhat.com>
5046         * Makefile.in (TAGS): Add c-family/*.cc.
5048 2017-12-01  Wilco Dijkstra  <wdijkstr@arm.com>
5050         * config/aarch64/aarch64.md (call_insn): Use %c rather than %a.
5051         (call_value_insn): Likewise.
5052         (sibcall_insn): Likewise.
5053         (sibcall_value_insn): Likewise.
5054         (movsi_aarch64): Likewise.
5055         (movdi_aarch64): Likewise.
5056         (add_losym_): Likewise.
5057         (ldr_got_small_): Likewise.
5058         (ldr_got_small_sidi): Likewise.
5059         (ldr_got_small_28k_): Likewise.
5060         (ldr_got_small_28k_sidi): Likewise.
5061         * config/aarch64/aarch64.c (aarch64_print_address_internal):
5062         Move output_addr_const to symbolic case. Add error check.
5064 2017-12-01  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
5066         * config/s390/predicates.md (plus16_Q_operand): New predicate.
5067         * config/s390/s390.md: Disable MVC merging peephole if it would
5068         disable operand forwarding.
5069         (new peephole2): Split MVCs if it would turn them into up to 2
5070         forwardable MVCs.
5072 2017-12-01  Richard Biener  <rguenther@suse.de>
5074         PR tree-optimization/83232
5075         * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Fix
5076         detection of same access. Instead of breaking the group here
5077         do not consider the duplicate.  Add comment explaining real fix.
5079 2017-12-01  Jonathan Wakely  <jwakely@redhat.com>
5081         * doc/md.texi (Insn Splitting): Fix "central flowgraph" typo.
5083 2017-12-01  Sudakshina Das  <sudi.das@arm.com>
5085         * config/arm/vfp.md (*movhf_vfp_fp16): Add conds attribute.
5087 2017-12-01  Jakub Jelinek  <jakub@redhat.com>
5089         * function.h (struct function): Remove cilk_frame_decl,
5090         is_cilk_function and calls_cilk_spawn fields.
5091         * tree-inline.h (struct copy_body_data): Remove remap_var_for_cilk
5092         field.
5093         * omp-simd-clone.c (simd_clone_clauses_extract): Don't clear
5094         cilk_elemental field.
5095         * cgraph.h (struct cgraph_simd_clone): Remove cilk_elemental field.
5096         * target.def: Adjust comment.
5097         * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
5098         Don't test cilk_elemental.
5100         PR tree-optimization/83233
5101         * gimple-ssa-store-merging.c (nop_stats, bswap_stats): Use
5102         bswap_stat name for the struct.
5104         PR c/79153
5105         * tree.h (SWITCH_BREAK_LABEL_P): Define.
5106         * gimplify.c (collect_fallthrough_labels): Handle GIMPLE_BIND
5107         starting with a GIMPLE_SWITCH and ending with GIMPLE_LABEL with
5108         SWITCH_BREAK_LABEL_P set on the label.
5109         (gimplify_switch_expr): Set SWITCH_BREAK_LABEL_P on the label
5110         added for default case if it was missing and not all cases covered.
5111         Wrap GIMPLE_SWITCH and the switch_body_seq into a GIMPLE_BIND if
5112         switch_body_seq ends with a GIMPLE_LABEL with SWITCH_BREAK_LABEL_P
5113         set on the label.
5114         * tree-chrec.c (evolution_function_is_univariate_p): Add return true;
5115         to avoid -Wimplicit-fallthrough warning.
5116         * config/i386/i386.c (ix86_expand_special_args_builtin): Add
5117         FALLTHRU comment to avoid -Wimplicit-fallthrough warning.
5119         PR tree-optimization/83221
5120         * tree-ssa-reassoc.c (sort_by_operand_rank): Shift bb_rank
5121         down by 16.
5122         (init_reassoc): Formatting fix.
5124         PR sanitizer/81275
5125         * tree-cfg.c (group_case_labels_stmt): Don't optimize away
5126         C++ FE implicitly added __builtin_unreachable () until -Wreturn-type
5127         is diagnosed.
5129         PR sanitizer/83219
5130         * tree-cfg.c: Include asan.h.
5131         (gimple_seq_unreachable_p): Return false for -fsanitize=unreachable.
5133 2017-12-01  Sergey Shalnov  <Sergey.Shalnov@intel.com>
5135         * config/i386/i386.md: Fix AVX512 register width in AVX512 instruction.
5137 2017-12-01  Sergey Shalnov  <Sergey.Shalnov@intel.com>
5139         * config/i386/i386.c (standard_sse_constant_opcode): Fix wrong form for
5140         vpcmpeqd instruction.
5142 2017-12-01  Sergey Shalnov  <Sergey.Shalnov@intel.com>
5144         * config/i386/i386.c (standard_sse_constant_opcode): Fix
5145         registers type for 128bit mode.
5147 2017-11-30  Jakub Jelinek  <jakub@redhat.com>
5149         * spellcheck-tree.c (test_find_closest_identifier): Use ; instead
5150         of ;;.
5151         * gengtype-state.c (read_state_pair): Likewise.
5152         * gimple-fold.c (gimple_fold_builtin_string_compare): Likewise.
5153         * sel-sched-dump.c (dump_insn_rtx_1): Likewise.
5154         * ipa-cp.c (intersect_aggregates_with_edge): Likewise.
5155         * ifcvt.c (noce_try_store_flag_constants): Likewise.
5156         * tree-ssa-ccp.c (ccp_finalize): Likewise.
5157         * omp-grid.c (grid_process_kernel_body_copy): Likewise.
5158         * builtins.c (fold_builtin_3): Likewise.
5159         * graphite-scop-detection.c
5160         (scop_detection::stmt_has_simple_data_refs_p): Likewise.
5161         * hsa-gen.c (hsa_function_representation::hsa_function_representation):
5162         Likewise.
5164 2017-12-01  Maxim Ostapenko  <m.ostapenko@samsung.com>
5166         PR sanitizer/81697
5167         * asan.c (asan_protect_global): Add new ignore_decl_rtl_set_p
5168         parameter. Return true if ignore_decl_rtl_set_p is true and other
5169         conditions are satisfied.
5170         * asan.h (asan_protect_global): Add new parameter.
5171         * varasm.c (categorize_decl_for_section): Pass true as second parameter
5172         to asan_protect_global calls.
5174 2017-11-30  Jim Wilson  <jimw@sifive.com>
5176         * doc/invoke.texi (RISC-V Options): Delete nonexistent -mmemcpy and
5177         -mno-memcpy options.  For -mplt, -mfdiv, -mdiv, -msave-restore, and
5178         -mstrict-align, add info on default value.  Delete redundant lines for
5179         -mabi.  Add missing -mexplicit-relocs docs.
5181 2017-11-30  Claudiu Zissulescu  <claziss@synopsys.com>
5183         * config/arc/arc.md (trap): New pattern.
5185 2017-11-30  Claudiu Zissulescu  <claziss@synopsys.com>
5187         * config/arc/arc.c (hwloop_optimize): Prevent the last ZOL
5188         instruction to end into a delay slot.
5189         * config/arc/arc.md (cond_delay_insn): Check if the instruction
5190         can be placed into a delay slot against reg_note.
5192 2017-11-30  Claudiu Zissulescu  <claziss@synopsys.com>
5194         * config/arc/arc.c (hwloop_optimize): Update hw-loop's end/start
5195         labels number of usages.
5197 2017-11-30  Claudiu Zissulescu  <claziss@synopsys.com>
5199         * config/arc/arc.c (arc_cannot_substitue_mem_equiv_p): New
5200         function.
5201         (TARGET_CANNOT_SUBSTITUTE_MEM_EQUIV_P): Define.
5203 2017-11-30  Jakub Jelinek  <jakub@redhat.com>
5205         PR target/83210
5206         * internal-fn.c (expand_mul_overflow): Optimize unsigned
5207         multiplication by power of 2 constant into two shifts + comparison.
5209 2017-11-30  Jan Hubicka  <hubicka@ucw.cz>
5211         PR target/81616
5212         * config/i386/x86-tune-costs.h (generic_cost): Revise for modern CPUs.
5214 2017-11-30  Richard Biener  <rguenther@suse.de>
5216         PR tree-optimization/83202
5217         * tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Add
5218         allow_peel argument and guard peeling.
5219         (canonicalize_loop_induction_variables): Likewise.
5220         (canonicalize_induction_variables): Pass false.
5221         (tree_unroll_loops_completely_1): Pass unroll_outer to disallow
5222         peeling from cunrolli.
5224 2017-11-29  Segher Boessenkool  <segher@kernel.crashing.org>
5226         * combine.c (try_combine): Print a message to dump file whenever
5227         I0, I1, or I2 cannot be combined into I3.
5229 2017-11-29  Segher Boessenkool  <segher@kernel.crashing.org>
5231         PR rtl-optimization/83156
5232         PR rtl-optimization/82621
5233         * combine.c (try_combine): Don't split an I2 if one of the dests is
5234         set again before I3.  Allow unused dests.
5236 2017-11-29  Segher Boessenkool  <segher@kernel.crashing.org>
5238         * config/rs6000/rs6000.md (*add<mode>3_carry_in_internal2): New.
5240 2017-11-29  Vladimir Makarov  <vmakarov@redhat.com>
5242         PR rtl-optimization/80818
5243         * lra.c (collect_non_operand_hard_regs): New arg insn.  Pass it
5244         recursively.  Use insn code for clobber.
5245         (lra_set_insn_recog_data): Pass the new arg to
5246         collect_non_operand_hard_regs.
5247         (add_regs_to_insn_regno_info): Pass insn instead of uid.  Use insn
5248         code for clobber.
5249         (lra_update_insn_regno_info): Pass insn to
5250         add_regs_to_insn_regno_info.
5252 2017-11-29  Jim Wilson  <jimw@sifive.com>
5253             Andrew Waterman  <andrew@sifive.com>
5255         * config/riscv/riscv.c (SINGLE_SHIFT_COST): New.
5256         (riscv_rtx_costs): Case ZERO_EXTRACT, match new pattern, and return
5257         SINGLE_SHIFT_COST.  Case LT and ZERO_EXTEND, likewise.  Case ASHIFT,
5258         use SINGLE_SHIFT_COST.
5259         * config/riscv/riscv.md (lshrsi3_zero_extend_1): New.
5260         (lshrsi3_zero_extend_2, lshrsi3_zero_extend_3): New.
5262 2017-11-29  Julia Koval  <julia.koval@intel.com>
5264         * config/i386/avx512vbmi2intrin.h (_mm512_shldv_epi16,
5265         _mm512_mask_shldv_epi16, _mm512_maskz_shldv_epi16, _mm512_shldv_epi32,
5266         _mm512_mask_shldv_epi32, _mm512_maskz_shldv_epi32, _mm512_shldv_epi64,
5267         _mm512_mask_shldv_epi64, _mm512_maskz_shldv_epi64): New intrinsics.
5268         * config/i386/avx512vbmi2vlintrin.h (_mm256_shldv_epi16,
5269         _mm256_mask_shldv_epi16, _mm256_maskz_shldv_epi16, _mm256_shldv_epi32,
5270         _mm256_mask_shldv_epi32, _mm256_maskz_shldv_epi32, _mm256_shldv_epi64,
5271         _mm256_mask_shldv_epi64, _mm256_maskz_shldv_epi64, _mm_shldv_epi16,
5272         _mm_mask_shldv_epi16, _mm_maskz_shldv_epi16, _mm_shldv_epi32,
5273         _mm_mask_shldv_epi32, _mm_maskz_shldv_epi32, _mm_shldv_epi64,
5274         _mm_mask_shldv_epi64, _mm_maskz_shldv_epi64): Ditto.
5275         * config/i386/i386-builtin.def (__builtin_ia32_vpshldv_v32hi,
5276         __builtin_ia32_vpshldv_v32hi_mask, __builtin_ia32_vpshldv_v32hi_maskz,
5277         __builtin_ia32_vpshldv_v16hi, __builtin_ia32_vpshldv_v16hi_mask,
5278         __builtin_ia32_vpshldv_v16hi_maskz, __builtin_ia32_vpshldv_v8hi,
5279         __builtin_ia32_vpshldv_v8hi_mask, __builtin_ia32_vpshldv_v8hi_maskz,
5280         __builtin_ia32_vpshldv_v16si, __builtin_ia32_vpshldv_v16si_mask,
5281         __builtin_ia32_vpshldv_v16si_maskz, __builtin_ia32_vpshldv_v8si,
5282         __builtin_ia32_vpshldv_v8si_mask, __builtin_ia32_vpshldv_v8si_maskz,
5283         __builtin_ia32_vpshldv_v4si, __builtin_ia32_vpshldv_v4si_mask,
5284         __builtin_ia32_vpshldv_v4si_maskz, __builtin_ia32_vpshldv_v8di,
5285         __builtin_ia32_vpshldv_v8di_mask, __builtin_ia32_vpshldv_v8di_maskz,
5286         __builtin_ia32_vpshldv_v4di, __builtin_ia32_vpshldv_v4di_mask,
5287         __builtin_ia32_vpshldv_v4di_maskz, __builtin_ia32_vpshldv_v2di,
5288         __builtin_ia32_vpshldv_v2di_mask,
5289         __builtin_ia32_vpshldv_v2di_maskz): New builtins.
5290         * config/i386/sse.md (vpshldv_<mode>, vpshldv_<mode>_mask,
5291         vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): New patterns.
5293 2017-11-29  Julia Koval  <julia.koval@intel.com>
5295         * config/i386/avx512vbmi2intrin.h (_mm512_shrdv_epi16,
5296         _mm512_mask_shrdv_epi16, _mm512_maskz_shrdv_epi16, _mm512_shrdv_epi32,
5297         _mm512_mask_shrdv_epi32, _mm512_maskz_shrdv_epi32, _mm512_shrdv_epi64,
5298         _mm512_mask_shrdv_epi64, _mm512_maskz_shrdv_epi64): New intrinsics.
5299         * config/i386/avx512vbmi2vlintrin.h (_mm256_shrdv_epi16,
5300         _mm256_mask_shrdv_epi16, _mm256_maskz_shrdv_epi16, _mm256_shrdv_epi32,
5301         _mm256_mask_shrdv_epi32, _mm256_maskz_shrdv_epi32, _mm256_shrdv_epi64,
5302         _mm256_mask_shrdv_epi64, _mm256_maskz_shrdv_epi64, _mm_shrdv_epi16,
5303         _mm_mask_shrdv_epi16, _mm_maskz_shrdv_epi16, _mm_shrdv_epi32,
5304         _mm_mask_shrdv_epi32, _mm_maskz_shrdv_epi32, _mm_shrdv_epi64,
5305         _mm_mask_shrdv_epi64, _mm_maskz_shrdv_epi64): Ditto.
5306         * config/i386/i386-builtin-types.def (V32HI_FTYPE_V32HI_V32HI_V32HI,
5307         V32HI_FTYPE_V32HI_V32HI_V32HI_INT, V16HI_FTYPE_V16HI_V16HI_V16HI_INT,
5308         V8HI_FTYPE_V8HI_V8HI_V8HI_INT, V8SI_FTYPE_V8SI_V8SI_V8SI_INT,
5309         V4SI_FTYPE_V4SI_V4SI_V4SI_INT, V8DI_FTYPE_V8DI_V8DI_V8DI,
5310         V8DI_FTYPE_V8DI_V8DI_V8DI_INT, V4DI_FTYPE_V4DI_V4DI_V4DI_INT,
5311         V16SI_FTYPE_V16SI_V16SI_V16SI, V16SI_FTYPE_V16SI_V16SI_V16SI_INT,
5312         V2DI_FTYPE_V2DI_V2DI_V2DI_INT): New types.
5313         * config/i386/i386.c (ix86_expand_args_builtin): Handle new types.
5314         * config/i386/sse.md (vpshrdv_<mode>, vpshrdv_<mode>_mask,
5315         vpshrdv_<mode>_maskz, vpshrdv_<mode>_maskz_1): New pattern.
5317 2017-11-29  Daniel Cederman  <cederman@gaisler.com>
5319         * config/sparc/sparc.c (sparc_do_work_around_errata): Treat the
5320         movsi_pic_gotdata_op instruction as a load for the UT699 errata
5321         workaround.
5323 2017-11-29  Martin Aberg  <maberg@gaisler.com>
5325         * config/sparc/sparc.md (divdf3_fix): Add NOP and adjust length
5326         to prevent b2bst errata sequence.
5327         (sqrtdf2_fix): Likewise.
5329 2017-11-29  Daniel Cederman  <cederman@gaisler.com>
5331         * config/sparc/sparc.c (fpop_reg_depend_p): New function.
5332         (div_sqrt_insn_p): New function.
5333         (sparc_do_work_around_errata): Insert NOP instructions to
5334         prevent sequences that could trigger the TN-0013 errata for
5335         certain LEON3 processors.
5336         (pass_work_around_errata::gate): Also test sparc_fix_lost_divsqrt.
5337         (sparc_option_override): Set sparc_fix_lost_divsqrt appropriately.
5338         * config/sparc/sparc.md (fix_lost_divsqrt): New attribute.
5339         (in_branch_delay): Prevent div and sqrt in delay slot if
5340         fix_lost_divsqrt.
5341         * config/sparc/sparc.opt (sparc_fix_lost_divsqrt): New variable.
5343 2017-11-29  Daniel Cederman  <cederman@gaisler.com>
5345         * config/sparc/sparc.c (atomic_insn_p): New function.
5346         (sparc_do_work_around_errata): Insert NOP instructions to
5347         prevent sequences that could trigger the TN-0010 errata for
5348         UT700.
5349         * config/sparc/sync.md (atomic_compare_and_swap_leon3_1): Make
5350         instruction referable in atomic_insns_p.
5352 2017-11-29  Daniel Cederman  <cederman@gaisler.com>
5354         * config/sparc/sync.md (swapsi): 16-byte align if sparc_fix_gr712rc.
5355         (atomic_compare_and_swap_leon3_1): Likewise.
5356         (ldstub): Likewise.
5358 2017-11-29  Daniel Cederman  <cederman@gaisler.com>
5360         * config/sparc/sparc.c (fpop_insn_p): New function.
5361         (sparc_do_work_around_errata): Insert NOP instructions to
5362         prevent sequences that could trigger the TN-0012 errata for
5363         GR712RC.
5364         (pass_work_around_errata::gate): Also test sparc_fix_gr712rc.
5365         * config/sparc/sparc.md (fix_gr712rc): New attribute.
5366         (in_branch_annul_delay): Prevent floating-point instructions
5367         in delay slot of annulled integer branch.
5369 2017-11-29  Richard Biener  <rguenther@suse.de>
5371         PR tree-optimization/83202
5372         * tree-vect-slp.c (scalar_stmts_set_t): New typedef.
5373         (bst_fail): Use it.
5374         (vect_analyze_slp_cost_1): Add visited set, do not account SLP
5375         nodes vectorized to the same stmts multiple times.
5376         (vect_analyze_slp_cost): Allocate a visited set and pass it down.
5377         (vect_analyze_slp_instance): Adjust.
5378         (scalar_stmts_to_slp_tree_map_t): New typedef.
5379         (vect_schedule_slp_instance): Add a map recording the SLP node
5380         representing the vectorized stmts for a set of scalar stmts.
5381         Avoid code-generating redundancies.
5382         (vect_schedule_slp): Allocate map and pass it down.
5384 2017-11-29  Nathan Sidwell  <nathan@acm.org>
5386         PR c++/83187
5387         * tree.c (build_complex_type): Fix canonicalization.  Only fill in
5388         type if it is new.
5390 2017-11-29  Wilco Dijkstra  <wdijkstr@arm.com>
5392         * config/aarch64/aarch64.c (aarch64_print_operand): Add new
5393         cases for printing LDP/STP memory addresses.
5394         (aarch64_print_address_internal): Renamed from
5395         aarch64_print_operand_address, added parameter, add Pmode check.
5396         (aarch64_print_ldpstp_address): New function for LDP/STP addresses.
5397         (aarch64_print_operand_address): Indirect to
5398         aarch64_print_address_internal.
5399         * config/aarch64/aarch64-simd.md (store_pair_lanes): Use new
5400         'y' operand output specifier.
5402 2017-11-29  Jakub Jelinek  <jakub@redhat.com>
5404         PR middle-end/83185
5405         * tree.c (build_simple_mem_ref_loc): Handle
5406         get_addr_base_and_unit_offset returning a MEM_REF.
5408         PR middle-end/80929
5409         * rtlanal.c (seq_cost): For non-single_set insns try to use insn_cost.
5411         PR target/80819
5412         * config/i386/sse.md (vec_concatv2di): Remove * from (=Yr,0,*rm)
5413         alternative.
5415 2017-11-28  David Malcolm  <dmalcolm@redhat.com>
5417         * diagnostic-show-locus.c (layout::print_trailing_fixits): Handle
5418         m_x_offset.
5419         (layout::move_to_column): Likewise.
5421 2017-11-28  Jakub Jelinek  <jakub@redhat.com>
5423         PR sanitizer/81275
5424         * tree.c (block_may_fallthru): Return false if SWITCH_ALL_CASES_P
5425         is set on SWITCH_EXPR and !block_may_fallthru (SWITCH_BODY ()).
5427 2017-11-28  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
5428             Martin Jambor  <mjambor@suse.cz>
5430         PR ipa/82808
5431         * tree.h (expr_type_first_operand_type_p): Declare
5432         * tree.c (expr_type_first_operand_type_p): New function.
5433         * ipa-prop.h (ipa_get_type): Allow i to be out of bounds.
5434         (ipa_value_from_jfunc): Adjust declaration.
5435         * ipa-cp.c (ipa_get_jf_pass_through_result): New parameter RES_TYPE.
5436         Use it as result type for arithmetics, unless it is NULL in which case
5437         be more conservative.
5438         (ipa_value_from_jfunc): New parameter PARM_TYPE, pass it to
5439         ipa_get_jf_pass_through_result.
5440         (propagate_vals_across_pass_through): Likewise.
5441         (propagate_scalar_across_jump_function): New parameter PARM_TYPE, pass
5442         is to propagate_vals_across_pass_through.
5443         (propagate_constants_across_call): Pass PARM_TYPE to
5444         propagate_scalar_across_jump_function.
5445         (find_more_scalar_values_for_callers_subset): Pass parameter type to
5446         ipa_value_from_jfunc.
5447         (cgraph_edge_brings_all_scalars_for_node): Likewise.
5448         * ipa-fnsummary.c (evaluate_properties_for_edge): Renamed parms_info
5449         to caller_parms_info, pass parameter type to ipa_value_from_jfunc.
5450         * ipa-prop.c (try_make_edge_direct_simple_call): New parameter
5451         target_type, pass it to ipa_value_from_jfunc.
5452         (update_indirect_edges_after_inlining): Pass parameter type to
5453         try_make_edge_direct_simple_call.
5455 2017-11-28  Jeff Law  <law@redhat.com>
5457         * gimple-ssa-evrp-analyze.c
5458         (evrp_range_analyzer::record_ranges_from_phis): Only use SCEV to
5459         refine ranges if scev_initialized_p returns true.
5460         * vr-values.c (vr_values::extract_range_from_phi_node): Likewise.
5462 2017-11-28  Julia Koval  <julia.koval@intel.com>
5464         * config/i386/avx512vbmi2intrin.h (_mm512_shrdi_epi16,
5465         _mm512_mask_shrdi_epi16, _mm512_maskz_shrdi_epi16, _mm512_shrdi_epi32,
5466         _mm512_mask_shrdi_epi32, _mm512_maskz_shrdi_epi32, _mm512_shrdi_epi64,
5467         _mm512_mask_shrdi_epi64, _mm512_maskz_shrdi_epi64): New intrinsics.
5468         * config/i386/avx512vbmi2vlintrin.h (_mm256_shrdi_epi16,
5469         _mm256_mask_shrdi_epi16, _mm256_maskz_shrdi_epi16,
5470         _mm256_mask_shrdi_epi32, _mm256_maskz_shrdi_epi32, _mm256_shrdi_epi32,
5471         _mm256_mask_shrdi_epi64, _mm256_maskz_shrdi_epi64, _mm256_shrdi_epi64,
5472         _mm_mask_shrdi_epi16, _mm_maskz_shrdi_epi16, _mm_shrdi_epi16,
5473         _mm_mask_shrdi_epi32, _mm_maskz_shrdi_epi32, _mm_shrdi_epi32,
5474         _mm_mask_shrdi_epi64, _mm_maskz_shrdi_epi64, _mm_shrdi_epi64): Ditto.
5475         * config/i386/i386-builtin.def (__builtin_ia32_vpshrd_v32hi,
5476         __builtin_ia32_vpshrd_v32hi_mask, __builtin_ia32_vpshrd_v16hi,
5477         __builtin_ia32_vpshrd_v16hi_mask, __builtin_ia32_vpshrd_v8hi,
5478         __builtin_ia32_vpshrd_v8hi_mask, __builtin_ia32_vpshrd_v16si,
5479         __builtin_ia32_vpshrd_v16si_mask, __builtin_ia32_vpshrd_v8si,
5480         __builtin_ia32_vpshrd_v8si_mask, __builtin_ia32_vpshrd_v4si,
5481         __builtin_ia32_vpshrd_v4si_mask, __builtin_ia32_vpshrd_v8di,
5482         __builtin_ia32_vpshrd_v8di_mask, __builtin_ia32_vpshrd_v4di,
5483         __builtin_ia32_vpshrd_v4di_mask, __builtin_ia32_vpshrd_v2di,
5484         __builtin_ia32_vpshrd_v2di_mask): New builtins.
5485         * config/i386/sse.md (vpshrd_<mode><mask_name>): New pattern.
5487 2017-11-28  Julia Koval  <julia.koval@intel.com>
5489         * config/i386/avx512vbmi2intrin.h (_mm512_shldi_epi16,
5490         _mm512_mask_shldi_epi16, _mm512_maskz_shldi_epi16, _mm512_shldi_epi32,
5491         _mm512_mask_shldi_epi32, _mm512_maskz_shldi_epi32, _mm512_shldi_epi64,
5492         _mm512_mask_shldi_epi64, _mm512_maskz_shldi_epi64): New intrinsics.
5493         * config/i386/avx512vbmi2vlintrin.h (_mm256_shldi_epi16,
5494         _mm256_mask_shldi_epi16, _mm256_maskz_shldi_epi16,
5495         _mm256_mask_shldi_epi32, _mm256_maskz_shldi_epi32, _mm256_shldi_epi32,
5496         _mm256_mask_shldi_epi64, _mm256_maskz_shldi_epi64, _mm256_shldi_epi64,
5497         _mm_mask_shldi_epi16, _mm_maskz_shldi_epi16, _mm_shldi_epi16,
5498         _mm_mask_shldi_epi32, _mm_maskz_shldi_epi32, _mm_shldi_epi32,
5499         _mm_mask_shldi_epi64, _mm_maskz_shldi_epi64, _mm_shldi_epi64): Ditto.
5500         * config/i386/i386-builtin-types.def (V32HI_FTYPE_V32HI_V32HI_INT,
5501         V32HI_FTYPE_V32HI_V32HI_INT_V32HI_INT, V16SI_FTYPE_V16SI_V16SI_INT,
5502         V16SI_FTYPE_V16SI_V16SI_INT_V16SI_INT,
5503         V8DI_FTYPE_V8DI_V8DI_INT_V8DI_INT, V8SI_FTYPE_V8SI_V8SI_INT_V8SI_INT,
5504         V16HI_FTYPE_V16HI_V16HI_INT_V16HI_INT,
5505         V4DI_FTYPE_V4DI_V4DI_INT_V4DI_INT,
5506         V8HI_FTYPE_V8HI_V8HI_INT_V8HI_INT,
5507         V4SI_FTYPE_V4SI_V4SI_INT_V4SI_INT,
5508         V2DI_FTYPE_V2DI_V2DI_INT_V2DI_INT): New types.
5509         * config/i386/i386-builtin.def (__builtin_ia32_vpshld_v32hi,
5510         __builtin_ia32_vpshld_v32hi_mask, __builtin_ia32_vpshld_v16hi,
5511         __builtin_ia32_vpshld_v16hi_mask, __builtin_ia32_vpshld_v8hi,
5512         __builtin_ia32_vpshld_v8hi_mask, __builtin_ia32_vpshld_v16si,
5513         __builtin_ia32_vpshld_v16si_mask, __builtin_ia32_vpshld_v8si,
5514         __builtin_ia32_vpshld_v8si_mask, __builtin_ia32_vpshld_v4si,
5515         __builtin_ia32_vpshld_v4si_mask, __builtin_ia32_vpshld_v8di,
5516         __builtin_ia32_vpshld_v8di_mask, __builtin_ia32_vpshld_v4di,
5517         __builtin_ia32_vpshld_v4di_mask, __builtin_ia32_vpshld_v2di,
5518         __builtin_ia32_vpshld_v2di_mask): New builtins.
5519         * config/i386/i386.c (ix86_expand_args_builtin): Handle new types.
5520         * config/i386/sse.md (vpshld_<mode><mask_name>): New pattern.
5522 2017-11-28  Richard Biener  <rguenther@suse.de>
5524         PR tree-optimization/80776
5525         * gimple-ssa-evrp-analyze.h (evrp_range_analyzer::set_ssa_range_info):
5526         Declare.
5527         * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::set_ssa_range_info):
5528         New function.
5529         (evrp_range_analyzer::record_ranges_from_incoming_edges):
5530         If the incoming edge is an effective fallthru because the other
5531         edge only reaches a __builtin_unreachable () then record ranges
5532         derived from the controlling condition in SSA info.
5533         (evrp_range_analyzer::record_ranges_from_phis): Use set_ssa_range_info.
5534         (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
5536 2017-11-28  Olivier Hainque  <hainque@adacore.com>
5538         * Makefile.in (SELFTEST_FLAGS): Use nul instead of /dev/null
5539         on mingw build hosts.
5541 2017-11-28  Jakub Jelinek  <jakub@redhat.com>
5543         PR debug/81307
5544         * dbxout.c (dbx_block_with_cold_children): Fix function comment.
5546 2017-11-28  Richard Biener  <rguenther@suse.de>
5548         PR middle-end/83141
5549         * gimple-fold.c (gimple_fold_builtin_memory_op): For aggregate
5550         copies generated from memcpy use a character array as reference
5551         type.
5553 2017-11-28  Julia Koval  <julia.koval@intel.com>
5554             Sebastian Peryt  <sebastian.peryt@intel.com>
5556         * Makefile.in (cilkplus.def, cilk-builtins.def, c-family/cilk.o,
5557         c-family/c-cilkplus.o, c-family/array-notation-common.o,
5558         cilk-common.o, cilk.h, cilk-common.c): Remove.
5559         * builtin-types.def
5560         (BT_FN_INT_PTR_PTR_PTR_FTYPE_BT_INT_BT_PTR_BT_PTR_BT_PTR): Remove.
5561         * builtins.c (is_builtin_name): Remove cilkplus condition.
5562         (BUILT_IN_CILK_DETACH, BUILT_IN_CILK_POP_FRAME): Remove.
5563         * builtins.def (DEF_CILK_BUILTIN_STUB, DEF_CILKPLUS_BUILTIN,
5564         cilk-builtins.def, cilkplus.def): Remove.
5565         * cif-code.def (CILK_SPAWN): Remove.
5566         * cilk-builtins.def: Delete.
5567         * cilk-common.c: Ditto.
5568         * cilk.h: Ditto.
5569         * cilkplus.def: Ditto.
5570         * config/darwin.h (fcilkplus): Delete.
5571         * cppbuiltin.c: Ditto.
5572         * doc/extend.texi: Remove cilkplus doc.
5573         * doc/generic.texi: Ditto.
5574         * doc/invoke.texi: Ditto.
5575         * doc/passes.texi: Ditto.
5576         * gcc.c (fcilkplus): Remove.
5577         * gengtype.c (cilk.h): Remove.
5578         * gimple-pretty-print.c (dump_gimple_omp_for): Remove cilkplus
5579         support.
5580         * gimple.h (GF_OMP_FOR_KIND_CILKFOR, GF_OMP_FOR_KIND_CILKSIMD):
5581         Remove.
5582         * gimplify.c (gimplify_return_expr, maybe_fold_stmt,
5583         gimplify_call_expr,
5584         is_gimple_stmt, gimplify_modify_expr, gimplify_scan_omp_clauses,
5585         gimplify_adjust_omp_clauses, gimplify_omp_for, gimplify_expr): Remove
5586         cilkplus conditions.
5587         * ipa-fnsummary.c (ipa_dump_fn_summary, compute_fn_summary,
5588         inline_read_section): Ditto.
5589         * ipa-inline-analysis.c (cilk.h): Remove.
5590         * ira.c (ira_setup_eliminable_regset): Remove cilkplus support.
5591         * lto-wrapper.c (merge_and_complain, append_compiler_options,
5592         append_linker_options): Remove condition for fcilkplus.
5593         * lto/lto-lang.c (cilk.h): Remove.
5594         (lto_init): Remove condition for fcilkplus.
5595         * omp-expand.c (expand_cilk_for_call): Delete.
5596         (expand_omp_taskreg, expand_omp_for_static_chunk,
5597         expand_omp_for): Remove cilkplus
5598         conditions.
5599         (expand_cilk_for): Delete.
5600         * omp-general.c (omp_extract_for_data): Remove cilkplus support.
5601         * omp-low.c (scan_sharing_clauses, create_omp_child_function,
5602         execute_lower_omp, diagnose_sb_0): Ditto.
5603         * omp-simd-clone.c (simd_clone_clauses_extract): Ditto.
5604         * tree-core.h (OMP_CLAUSE__CILK_FOR_COUNT_): Delete.
5605         * tree-nested.c: Ditto.
5606         * tree-pretty-print.c (dump_omp_clause): Remove cilkplus support.
5607         (dump_generic_node): Ditto.
5608         * tree.c (OMP_CLAUSE__CILK_FOR_COUNT_): Delete.
5609         * tree.def (cilk_simd, cilk_for, cilk_spawn_stmt, cilk_sync_stmt):
5610         Delete.
5611         * tree.h (CILK_SPAWN_FN, EXPR_CILK_SPAWN): Delete.
5613 2017-11-28  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
5615         * config/aarch64/aarch64.md (div<mode>3): Change check to TARGET_FLOAT.
5616         * config/aarch64/aarch64.c (aarch64_emit_approx_div): Add early exit
5617         for vector mode and !TARGET_SIMD.
5619 2017-11-28  Jakub Jelinek  <jakub@redhat.com>
5621         * tree.def (SWITCH_EXPR): Change from 3 operand to 2 operand tree.
5622         Adjust comment.
5623         * tree.h (SWITCH_LABELS): Remove.
5624         * gimplify.c (gimplify_switch_expr): Don't test SWITCH_LABELS,
5625         assert SWITCH_BODY is non-NULL.
5626         * tree-pretty-print.c (dump_generic_node): Remove SWITCH_LABELS
5627         handling.
5628         * tree.c (block_may_fallthru): Always return true; for SWITCH_EXPR.
5630         PR tree-optimization/80788
5631         * match.pd (X +- C1 CMP C2 -> X CMP C2 -+ C1): If res
5632         has TREE_OVERFLOW set, call drop_tree_overflow.
5634 2017-11-28  Richard Biener  <rguenther@suse.de>
5636         PR tree-optimization/83158
5637         * tree-vrp.c (intersect_ranges): Prefer ~[0, 0] in a few more cases.
5639 2017-11-28  Segher Boessenkool  <segher@kernel.crashing.org>
5641         PR 81288/target
5642         * config/rs6000/rs6000.c (rs6000_rtx_costs): Do not handle
5643         TARGET_ISEL && !TARGET_MFCRF differently.  Simplify code.
5645 2017-11-27  Segher Boessenkool  <segher@kernel.crashing.org>
5647         * config/rs6000/rs6000.md (<code><GPR:mode><GPR2:mode>2_isel): Change
5648         LT/GT/LTU/GTU to LE/GE/LEU/GEU where possible.
5650 2017-11-27  Michael Meissner  <meissner@linux.vnet.ibm.com>
5652         PR middle_end/82333
5653         * varasm.c (compare_constant): Take the mode of the constants into
5654         account when comparing floating point constants.
5656 2017-11-27  Gerald Pfeifer  <gerald@pfeifer.com>
5658         * hash-set.h (DEFINE_DEBUG_HASH_SET): Remove static qualifier
5659         from explicit instantiation of debug_helper.
5660         * vec.h (DEFINE_DEBUG_VEC): Ditto.
5662 2017-11-27  Richard Biener  <rguenther@suse.de>
5664         * gimple-fold.c (gimple_fold_builtin_memory_op): Remove dead code,
5665         refactor a bit.
5667 2017-11-27  Richard Biener  <rguenther@suse.de>
5669         * tree.c (wide_int_to_tree): Free discarded INTEGER_CST.
5670         (type_hash_canon): Also clear int_cst_hash_table entry for
5671         TYPE_MIN/MAX_VALUE.
5672         (build_nonstandard_integer_type): Hash all TYPE_MAX_VALUEs.
5674 2017-11-27  Tamar Christina  <tamar.christina@arm.com>
5676         * doc/extend.texi: Add -A suffix (ARMv8*-A, ARMv7-A).
5677         * doc/invoke.texi: Add -A suffix (ARMv8*-A, ARMv7-A).
5678         * doc/sourcebuild.texi: Add -A suffix (ARMv8*-A, ARMv7-A).
5680 2017-11-27  Markus Trippelsdorf  <markus@trippelsdorf.de>
5682         * hash-map.h (gt_cleare_cache): Avoid UB.
5684 2017-11-27  Eric Botcazou  <ebotcazou@adacore.com>
5686         * cfgloop.h (struct loop): Document usage of USHRT_MAX for unroll.
5687         * loop-unroll.c (decide_unroll_constant_iterations): Implement it.
5688         (decide_unroll_runtime_iterations): Likewise.
5689         (decide_unroll_stupid): Likewise.
5691 2017-11-27  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
5693         PR target/83109
5694         * config/i386/i386.md: Add a loop with incssp.
5696 2017-11-27  Martin Jambor  <mjambor@suse.cz>
5698         PR tree-optimization/81248
5699         * tree-sra.c (splice_param_accesses): Remove size check.
5700         (decide_one_param_reduction): Fix size check.
5701         * gimple-pretty-print.c (dump_profile): Silence warning.
5702         * params.def (PARAM_IPA_SRA_PTR_GROWTH_FACTOR): Adjust description.
5704 2017-11-27  Jakub Jelinek  <jakub@redhat.com>
5706         PR debug/81307
5707         * dbxout.c (lastlineno): New variable.
5708         (dbx_debug_hooks): Use dbxout_switch_text_section as
5709         switch_text_section debug hook.
5710         (dbxout_function_end): Switch to current_function_section
5711         rather than function_section.  If crtl->has_bb_partition,
5712         output just one N_FUN, depending on in_cold_section_p.
5713         (dbxout_source_line): Remember last lineno in lastlineno.
5714         (dbxout_switch_text_section): New function.
5715         (dbxout_function_decl): Adjust dbxout_block caller.
5716         (dbx_block_with_cold_children): New function.
5717         (dbxout_block): Return true if any LBRAC/RBRAC have been
5718         emitted.  Use dbx_block_with_cold_children at depth == 0
5719         in second partition.  Add PARENT_BLOCKNUM argument, pass
5720         it optionally adjusted to children.  Output LBRAC/RBRAC
5721         around recursive call only if the block is in the current
5722         partition, if not and anything was output, emit empty
5723         range LBRAC/RBRAC.
5724         * final.c (final_scan_insn): Compute cold_function_name
5725         before calling switch_text_section debug hook.  Call
5726         that hook even if dwarf2out_do_frame if not emitting
5727         dwarf debug info.
5729         PR target/83100
5730         * varasm.c (bss_initializer_p): Return true for DECL_COMMON
5731         TREE_READONLY decls.
5733 2017-11-27  Markus Trippelsdorf  <markus@trippelsdorf.de>
5735         PR rtl-optimization/82488
5736         * expr.c (fixup_args_size_notes): Avoid signed integer overflow.
5738 2017-11-26  Julia Koval  <julia.koval@intel.com>
5740         * config/i386/i386.c (processor_target_table): Add skylake_cost for
5741         skylake-avx512.
5742         * config/i386/x86-tune-costs.h (skylake_memcpy, skylake_memset,
5743         skylake_cost): New.
5745 2017-11-26  Julia Koval  <julia.koval@intel.com>
5747         * config/i386/driver-i386.c (host_detect_local_cpu):
5748         Detect skylake-avx512.
5750 2017-11-26  Julia Koval  <julia.koval@intel.com>
5752         * config.gcc: Add -march=cannonlake.
5753         * config/i386/driver-i386.c (host_detect_local_cpu): Detect cannonlake.
5754         * config/i386/i386-c.c (ix86_target_macros_internal): Handle cannonlake.
5755         * config/i386/i386.c (processor_costs): Add m_CANNONLAKE.
5756         (PTA_CANNONLAKE): New.
5757         (processor_target_table): Add cannonlake.
5758         (ix86_option_override_internal): Ditto.
5759         (fold_builtin_cpu): Ditto.
5760         (get_builtin_code_for_version): Handle cannonlake.
5761         (M_INTEL_COREI7_CANNONLAKE): New.
5762         * config/i386/i386.h (TARGET_CANNONLAKE, PROCESSOR_CANNONLAKE): New.
5763         * doc/invoke.texi: Add -march=cannonlake.
5765 2017-11-14  Boris Kolpackov  <boris@codesynthesis.com>
5767         * plugin.c (add_new_plugin): Use platform-specific library extensions.
5768         (try_init_one_plugin): Alternative implementation for MinGW.
5769         * Makefile.in (plugin_implib): New.
5770         (gengtype-lex.c): Fix broken AIX workaround.
5771         * configure: Regenerate.
5772         * doc/plugins.texi: Document support for MinGW.
5774 2017-11-25  Jakub Jelinek  <jakub@redhat.com>
5776         PR rtl-optimization/81553
5777         * combine.c (simplify_if_then_else): In (if_then_else COND (OP Z C1) Z)
5778         to (OP Z (mult COND (C1 * STORE_FLAG_VALUE))) optimization, if OP
5779         is a shift where C1 has different mode than the whole shift, use C1's
5780         mode for MULT rather than the shift's mode.
5782         PR target/82848
5783         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Don't fold
5784         builtins not enabled in the currently selected ISA.
5786 2017-11-24  Jackson Woodruff  <jackson.woodruff@arm.com>
5788         PR tree-optimization/71026
5789         * tree-ssa-math-opts (is_division_by_square, is_square_of): New.
5790         (insert_reciprocals): Change to insert reciprocals before a division
5791         by a square and to insert the square of a reciprocal.
5792         (execute_cse_reciprocals_1): Change to consider division by a square.
5793         (register_division_in): Add importance parameter.
5795 2017-11-24  Richard Biener  <rguenther@suse.de>
5797         PR tree-optimization/82402
5798         * tree-vect-loop-manip.c (create_lcssa_for_virtual_phi): Properly
5799         set SSA_NAME_OCCURS_IN_ABNORMAL_PHI.
5801 2017-11-24  Marc Glisse  <marc.glisse@inria.fr>
5803         * match.pd (0-ptr): New transformation.
5805 2017-11-24  Jan Hubicka  <hubicka@ucw.cz>
5807         PR bootstrap/83015
5808         * ipa-inline.c (inline_small_functions): Set current badnes correctly
5809         when skipping checking.
5811 2017-11-24  Richard Biener  <rguenther@suse.de>
5813         PR tree-optimization/83128
5814         * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Handle STRING_CSTs.
5815         (vn_reference_lookup_3): Likewise.
5817 2017-11-24  Jakub Jelinek  <jakub@redhat.com>
5819         PR sanitizer/83014
5820         * ubsan.c (ubsan_type_descriptor): Use pp_unsigned_wide_integer
5821         instead of pp_printf with HOST_WIDE_INT_PRINT_DEC.  Avoid calling
5822         tree_to_uhwi twice.
5824         * tree-object-size.c (pass_through_call): Use gimple_call_return_flags
5825         ERF_RETURN*ARG* for builtins other than BUILT_IN_ASSUME_ALIGNED,
5826         check for the latter with gimple_call_builtin_p.  Do not handle
5827         BUILT_IN_STPNCPY_CHK which is not a pass through call.
5829 2017-11-24  Christophe Lyon  <christophe.lyon@linaro.org>
5831         * config/arm/arm_neon.h: Fix pragma GCC push_options before
5832         vdot_u32.
5834 2017-11-23  Julia Koval  <julia.koval@intel.com>
5836         * config/i386/avx512vbmi2intrin.h (_mm512_mask_expand_epi8,
5837         _mm512_maskz_expand_epi8, _mm512_mask_expandloadu_epi8,
5838         _mm512_maskz_expandloadu_epi8, _mm512_mask_expand_epi16,
5839         _mm512_maskz_expand_epi16, _mm512_mask_expandloadu_epi16,
5840         _mm512_maskz_expandloadu_epi16): New intrinsics.
5841         * config/i386/avx512vbmi2vlintrin.h (_mm_mask_expand_epi8,
5842         _mm_maskz_expand_epi8, _mm_mask_expandloadu_epi8,
5843         _mm_maskz_expandloadu_epi8, _mm_mask_expand_epi16,
5844         _mm_maskz_expand_epi16, _mm_mask_expandloadu_epi16,
5845         _mm_maskz_expandloadu_epi16, _mm256_mask_expand_epi16,
5846         _mm256_maskz_expand_epi16, _mm256_mask_expandloadu_epi16,
5847         _mm256_maskz_expandloadu_epi16, _mm256_mask_expand_epi8,
5848         _mm256_maskz_expand_epi8, _mm256_mask_expandloadu_epi8,
5849         _mm256_maskz_expandloadu_epi8): New intrinsics.
5850         * config/i386/i386-builtin-types.def (V64QI_FTYPE_PCV64QI_V64QI_UDI,
5851         V32HI_FTYPE_PCV32HI_V32HI_USI, V32QI_FTYPE_PCV32QI_V32QI_USI,
5852         V16HI_FTYPE_PCV16HI_V16HI_UHI, V16QI_FTYPE_PCV16QI_V16QI_UHI,
5853         V8HI_FTYPE_PCV8HI_V8HI_UQI): New types.
5854         * config/i386/i386.c (ix86_expand_special_args_builtin): Use new types.
5855         * config/i386/sse.md (VI248_VLBW): New iterator.
5856         (expand<mode>_mask, expand<mode>_maskz): New patterns.
5858 2017-11-23  Julia Koval  <julia.koval@intel.com>
5860         * config.gcc (avx512vbmi2intrin.h, avx512vbmi2vlintrin): New headers.
5861         * config/i386/avx512vbmi2intrin.h (_mm512_mask_compress_epi8,
5862         _mm512_maskz_compress_epi8, _mm512_mask_compressstoreu_epi8,
5863         _mm512_mask_compress_epi16, _mm512_maskz_compress_epi16,
5864         _mm512_mask_compressstoreu_epi16): New.
5865         * config/i386/avx512vbmi2vlintrin.h (_mm_mask_compress_epi8,
5866         _mm_maskz_compress_epi8, _mm256_mask_compressstoreu_epi16,
5867         _mm_mask_compress_epi16, _mm_maskz_compress_epi16,
5868         _mm256_mask_compress_epi16, _mm256_maskz_compress_epi16,
5869         _mm_mask_compressstoreu_epi8, _mm_mask_compressstoreu_epi16,
5870         _mm256_mask_compress_epi8, _mm256_maskz_compress_epi8,
5871         _mm256_mask_compressstoreu_epi8): New.
5872         * config/i386/i386-builtin-types.def (VOID_FTYPE_PV64QI_V64QI_UDI,
5873         VOID_FTYPE_PV32HI_V32HI_USI, VOID_FTYPE_PV32QI_V32QI_USI,
5874         VOID_FTYPE_PV16QI_V16QI_UHI, VOID_FTYPE_PV16HI_V16HI_UHI,
5875         VOID_FTYPE_PV8HI_V8HI_UQI): New types.
5876         * config/i386/i386-builtin.def (__builtin_ia32_compressqi512_mask,
5877         __builtin_ia32_compresshi512_mask, __builtin_ia32_compressqi256_mask,
5878         __builtin_ia32_compressqi128_mask, __builtin_ia32_compresshi256_mask,
5879         __builtin_ia32_compresshi128_mask,
5880         __builtin_ia32_compressstoreuqi512_mask,
5881         __builtin_ia32_compressstoreuhi512_mask,
5882         __builtin_ia32_compressstoreuqi256_mask,
5883         __builtin_ia32_compressstoreuqi128_mask,
5884         __builtin_ia32_compressstoreuhi256_mask,
5885         __builtin_ia32_compressstoreuhi128_mask): New builtins.
5886         * config/i386/i386.c (ix86_init_mmx_sse_builtins): Create special args
5887         array for flags2.
5888         (ix86_expand_special_args_builtin): Handle new types.
5889         (s4fma_expand): Handle new builtin array.
5890         * config/i386/immintrin.h: Include new headers.
5891         * config/i386/sse.md (VI12_AVX512VLBW): New iterator.
5892         (compress<mode>_mask, compressstore<mode>_mask): New patterns.
5894 2017-11-23  Jakub Jelinek  <jakub@redhat.com>
5896         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Formatting
5897         fixes.  Declare temp and g variables at the top in order to avoid
5898         {} in most of the cases.
5900 2017-11-23  Marc Glisse  <marc.glisse@inria.fr>
5902         * match.pd (ptr-0): New transformation.
5904 2017-11-23  Charles Baylis  <charles.baylis@linaro.org>
5906         * config/arm/arm-protos.h (enum arm_addr_mode_op): New.
5907         (struct addr_mode_cost_table): New.
5908         (struct tune_params): Add field addr_mode_costs.
5909         * config/arm/arm.c (generic_addr_mode_costs): New.
5910         (arm_slowmul_tune): Initialise addr_mode_costs field.
5911         (arm_fastmul_tune): Likewise.
5912         (arm_strongarm_tune): Likewise.
5913         (arm_xscale_tune): Likewise.
5914         (arm_9e_tune): Likewise.
5915         (arm_marvell_pj4_tune): Likewise.
5916         (arm_v6t2_tune): Likewise.
5917         (arm_cortex_tune): Likewise.
5918         (arm_cortex_a8_tune): Likewise.
5919         (arm_cortex_a7_tune): Likewise.
5920         (arm_cortex_a15_tune): Likewise.
5921         (arm_cortex_a35_tune): Likewise.
5922         (arm_cortex_a53_tune): Likewise.
5923         (arm_cortex_a57_tune): Likewise.
5924         (arm_exynosm1_tune): Likewise.
5925         (arm_xgene1_tune): Likewise.
5926         (arm_cortex_a5_tune): Likewise.
5927         (arm_cortex_a9_tune): Likewise.
5928         (arm_cortex_a12_tune): Likewise.
5929         (arm_cortex_a73_tune): Likewise.
5930         (arm_v7m_tune): Likewise.
5931         (arm_cortex_m7_tune): Likewise.
5932         (arm_v6m_tune): Likewise.
5933         (arm_fa726te_tune): Likewise.
5934         (arm_mem_costs): Use table lookup to calculate cost of addressing mode.
5936 2017-11-23  Charles Baylis  <charles.baylis@linaro.org>
5938         * config/arm/arm.c (arm_mem_costs): New function.
5939         (arm_rtx_costs_internal): Use arm_mem_costs.
5941 2017-11-23  Mark Wielaard  <mark@klomp.org>
5943         * dwarf2out.c (init_sections_and_labels): Use generation to create
5944         unique ranges_section_label and ranges_base_label. Return generation.
5945         (output_rnglists): Add generation argument. Use generation to create
5946         unique ranges labels.
5947         (dwarf2out_finish): Get generation from init_sections_and_labels
5948         and pass generation to output_rnglists.
5950 2017-11-23  Mike Stump  <mikestump@comcast.net>
5951             Eric Botcazou  <ebotcazou@adacore.com>
5953         * doc/generic.texi (ANNOTATE_EXPR): Document 3rd operand.
5954         * cfgloop.h (struct loop): Add unroll field.
5955         * function.h (struct function): Add has_unroll bitfield.
5956         * gimplify.c (gimple_boolify) <ANNOTATE_EXPR>: Deal with unroll kind.
5957         (gimplify_expr) <ANNOTATE_EXPR>: Propagate 3rd operand.
5958         * loop-init.c (pass_loop2::gate): Return true if cfun->has_unroll.
5959         (pass_rtl_unroll_loops::gate): Likewise.
5960         * loop-unroll.c (decide_unrolling): Tweak note message.  Skip loops
5961         for which loop->unroll==1.
5962         (decide_unroll_constant_iterations): Use note for consistency and
5963         take loop->unroll into account.  Return early if loop->unroll is set.
5964         Fix thinko in existing test.
5965         (decide_unroll_runtime_iterations): Use note for consistency and
5966         take loop->unroll into account.
5967         (decide_unroll_stupid): Likewise.
5968         * lto-streamer-in.c (input_cfg): Read loop->unroll.
5969         * lto-streamer-out.c (output_cfg): Write loop->unroll.
5970         * tree-cfg.c (replace_loop_annotate_in_block) <annot_expr_unroll_kind>:
5971         New case.
5972         (replace_loop_annotate) <annot_expr_unroll_kind>: Likewise.
5973         (print_loop): Print loop->unroll if set.
5974         * tree-core.h (enum annot_expr_kind): Add annot_expr_unroll_kind.
5975         * tree-inline.c (copy_loops): Copy unroll and set cfun->has_unroll.
5976         * tree-pretty-print.c (dump_generic_node) <annot_expr_unroll_kind>:
5977         New case.
5978         * tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Bail out if
5979         loop->unroll is set and smaller than the trip count.  Otherwise bypass
5980         entirely the heuristics if loop->unroll is set.  Remove dead note.
5981         Fix off-by-one bug in other note.
5982         (try_peel_loop): Bail out if loop->unroll is set.  Fix formatting.
5983         (tree_unroll_loops_completely_1): Force unrolling if loop->unroll
5984         is greater than 1.
5985         (tree_unroll_loops_completely): Make static.
5986         (pass_complete_unroll::execute): Use correct type for variable.
5987         (pass_complete_unrolli::execute): Fix formatting.
5988         * tree.def (ANNOTATE_EXPR): Add 3rd operand.
5990 2017-11-23  Sergey Shalnov  <Sergey.Shalnov@intel.com>
5992         * config/i386/i386.h (TARGET_PREFER_AVX256): Also
5993         enable when TARGET_PREFER_AVX128 is set.
5995 2017-11-23  Jan Hubicka  <hubicka@ucw.cz>
5997         * ipa-profile.c (ipa_propagate_frequency_1): Use count instead of
5998         frequency.
5999         * cgraph.c (cgraph_edge::dump_edge_flags): Dump sreal frequencies.
6000         (cgraph_edge::maybe_hot_p): Use sreal frequencies.
6002 2017-11-23  Jan Hubicka  <hubicka@ucw.cz>
6004         * ipa-fnsummary.c (record_modified_bb_info): Use sreal
6005         frequencies. Fix estimation of aggregate parameters.
6007 2017-11-23  Jan Hubicka  <hubicka@ucw.cz>
6009         * cgraphclones.c (cgraph_node::create_clone): Fix updating of profile
6010         when inlining.
6012 2017-11-23  Tom de Vries  <tom@codesourcery.com>
6014         * config/sh/sh.h (ASM_OUTPUT_ADDR_VEC_ELT): Wrap in "do {} while (0)".
6016 2017-11-23  Tom de Vries  <tom@codesourcery.com>
6018         * config/ft32/ft32.h (ASM_OUTPUT_ADDR_VEC_ELT): Remove semicolon after
6019         macro.
6021 2017-11-23  Oleg Endo  <olegendo@gcc.gnu.org>
6023         PR target/83111
6024         * config/sh/sh.md (udivsi3, divsi3, sibcall_value_pcrel,
6025         sibcall_value_pcrel_fdpic): Use local variable instead of
6026         operands[3].
6027         (calli_tbr_rel): Add missing operand 2.
6028         (call_valuei_tbr_rel): Add missing operand 3.
6030 2017-11-23  Jakub Jelinek  <jakub@redhat.com>
6032         PR middle-end/82253
6033         * expr.c (expand_assignment): For CONCAT to_rtx, complex type from and
6034         bitpos/bitsize covering the whole destination, use store_expr only if
6035         the complex mode is the same.  Otherwise, use expand_normal and if
6036         it returns CONCAT, subreg each part separately instead of trying to
6037         subreg the whole result.
6039 2017-11-23  Richard Biener  <rguenther@suse.de>
6041         PR tree-optimization/23094
6042         * tree-ssa-sccvn.c (vuse_ssa_val): Handle VN_TOP when we
6043         come here from walking over backedges in the first iteration.
6044         (vn_reference_lookup_3): Skip clobbers that store the same value.
6046 2017-11-23  Richard Biener  <rguenther@suse.de>
6048         PR tree-optimization/81403
6049         * tree-ssa-pre.c (get_representative_for): Add parameter specifying
6050         a block we need a leader relative to.
6051         (phi_translate_1): For nary processing require a leader from
6052         get_representative_for given we run expression simplification
6053         using match-and-simplify.  Remove previous fix.
6055 2017-11-22  Jeff Law  <law@redhat.com>
6057         * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::try_find_new_range):
6058         Use new method allocate_value_range rather than accessing the
6059         vrp_value_range_pool data member directly.
6060         * tree-vrp.c (simplify_stmt_for_jump_threading): Tweak slightly
6061         to use extract_range_from_stmt method to avoid need for
6062         extract_range_from_assignment method.
6063         (vrp_prop::vrp_finalize): Use set_lattice_propagation_complete
6064         method rather than setting values_propgated data member directly.
6065         * vr-values.h (class vr_values): Privatize vrp_value_range_pool,
6066         and values propagated data members and extract_range_from_assignment
6067         method.  Reorder private data members to conform to standards.
6068         Add new methods set_lattice_propagation_complete and
6069         allocate_value_range.
6071 2017-11-22  Eric Botcazou  <ebotcazou@adacore.com>
6073         PR rtl-optimization/83030
6074         * doc/rtl.texi (Flags in an RTL Expression): Alphabetize, add entry
6075         for CROSSING_JUMP_P and mention usage of 'jump' for JUMP_INSNs.
6076         (Insns): Delete entry for REG_CROSSING_JUMP in register notes.
6077         * bb-reorder.c (update_crossing_jump_flags): Do not test whether the
6078         CROSSING_JUMP_P flag is already set before setting it.
6079         * cfgrtl.c (fixup_partition_crossing): Likewise.
6080         * reorg.c (relax_delay_slots): Do not consider a CROSSING_JUMP_P
6081         insn as useless.
6083 2017-11-22  Jakub Jelinek  <jakub@redhat.com>
6085         * simplify-rtx.c (simplify_binary_operation_1) <case VEC_SERIES>:
6086         Handle the case where both arguments are using gen_const_vec_series.
6088 2017-11-22  David Malcolm  <dmalcolm@redhat.com>
6090         PR c++/62170
6091         * pretty-print.c (pp_format): Move quoting implementation to
6092         pp_begin_quote and pp_end_quote.  Update pp_format_decoder call
6093         to pass address of "quote" local.
6094         (pp_begin_quote): New function.
6095         (pp_end_quote): New function.
6096         * pretty-print.h (printer_fn): Convert penultimate param from bool
6097         to bool *.
6098         (pp_begin_quote): New decl.
6099         (pp_end_quote): New decl.
6100         * tree-diagnostic.c (default_tree_printer): Convert penultimate
6101         param from bool to bool *.
6102         * tree-diagnostic.h (default_tree_printer): Likewise.
6104 2017-11-22  Jeff Law  <law@redhat.com>
6106         * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::evrp_range_analyzer)
6107         Initialize vr_values.
6108         (evrp_range_analyzer::try_find_new_range): Call methods attached to
6109         vr_values via vr_values class instance rather than delegators.
6110         (evrp_range_analyzer::record_ranges_from_phis): Likewise.
6111         (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
6112         (evrp_range_analyzer::push_value_range): Likewise.
6113         (evrp_range_analyzer::pop_value_range): Likewise.
6114         * gimple-ssa-evrp-analyze.h (class evrp_range_analyzer): Remove
6115         most delegators.  Those remaining are exposed as public interfaces.
6116         Make vr_values a pointer and private.
6117         (evrp_range_analyzer::~evrp_range_analyzer): Delete the attached
6118         vr_values.
6119         (evrp_range_analyzer::get_vr_value): New method.
6120         * gimple-ssa-evrp.c (class evrp_folder): Use DISABLE_COPY_AND_ASSIGN.
6121         (evrp_folder::evrp_folder): New ctor to initialize vr_values.
6122         (class evrp_dom_walker): Attach evrp_folder class, initialize
6123         it in the ctor.  Remove temporary delegators.
6124         (evrp_dom_walker::before_dom_children): Call methods in attached
6125         evrp_range_analyzer class via class instance pointer.  Use
6126         free value_range_constant_singleton to remove need for
6127         op_with_constant_singleton_value delegator method.  Do not
6128         create a vrp_prop class instance for every call!  Narrow
6129         scope of a couple variables.
6130         (evrp_dom_walker::cleanup): Call methods in attached
6131         evrp_range_analyzer class via class instance pointer.
6132         * vr-values.h (class vr_values): Privatize many methods and
6133         data members.
6135 2017-11-22  Richard Sandiford  <richard.sandiford@linaro.org>
6137         * tree.def (REDUC_MAX_EXPR, REDUC_MIN_EXPR, REDUC_PLUS_EXPR): Delete.
6138         * cfgexpand.c (expand_debug_expr): Remove handling for them.
6139         * expr.c (expand_expr_real_2): Likewise.
6140         * fold-const.c (const_unop): Likewise.
6141         * optabs-tree.c (optab_for_tree_code): Likewise.
6142         * tree-cfg.c (verify_gimple_assign_unary): Likewise.
6143         * tree-inline.c (estimate_operator_cost): Likewise.
6144         * tree-pretty-print.c (dump_generic_node): Likewise.
6145         (op_code_prio): Likewise.
6146         (op_symbol_code): Likewise.
6147         * internal-fn.def (DEF_INTERNAL_SIGNED_OPTAB_FN): Define.
6148         (IFN_REDUC_PLUS, IFN_REDUC_MAX, IFN_REDUC_MIN): New internal functions.
6149         * internal-fn.c (direct_internal_fn_optab): New function.
6150         (direct_internal_fn_array, direct_internal_fn_supported_p
6151         (internal_fn_expanders): Handle DEF_INTERNAL_SIGNED_OPTAB_FN.
6152         * fold-const-call.c (fold_const_reduction): New function.
6153         (fold_const_call): Handle CFN_REDUC_PLUS, CFN_REDUC_MAX and
6154         CFN_REDUC_MIN.
6155         * tree-vect-loop.c: Include internal-fn.h.
6156         (reduction_code_for_scalar_code): Rename to...
6157         (reduction_fn_for_scalar_code): ...this and return an internal
6158         function.
6159         (vect_model_reduction_cost): Take an internal_fn rather than
6160         a tree_code.
6161         (vect_create_epilog_for_reduction): Likewise.  Build calls rather
6162         than assignments.
6163         (vectorizable_reduction): Use internal functions rather than tree
6164         codes for the reduction operation.  Update calls to the functions
6165         above.
6166         * config/aarch64/aarch64-builtins.c (aarch64_gimple_fold_builtin):
6167         Use calls to internal functions rather than REDUC tree codes.
6168         * config/aarch64/aarch64-simd.md: Update comment accordingly.
6170 2017-11-22  Olivier Hainque  <hainque@adacore.com>
6172         * config/vxworks.c (vxworks_override_options): Pick default
6173         dwarf version from macro value, VXWORKS_DWARF_VERSION_DEFAULT.
6174         * config/vxworks.h: Define VXWORKS_DWARF_VERSION_DEFAULT and
6175         DWARF_GNAT_ENCODINGS_DEFAULT.
6176         * config/vxworksae.h: Likewise.
6178 2017-11-22  Marc Glisse  <marc.glisse@inria.fr>
6180         PR tree-optimization/83104
6181         * vr-values.c (simplify_stmt_using_ranges): Check integral argument,
6182         not result.
6184 2017-11-22  Marek Polacek  <polacek@redhat.com>
6185             H.J. Lu  <hongjiu.lu@intel.com>
6186             Jason Merrill  <jason@redhat.com>
6188         PR c++/60336
6189         PR middle-end/67239
6190         PR target/68355
6191         * calls.c (initialize_argument_information): Call
6192         warn_parameter_passing_abi target hook.
6193         (store_one_arg): Use 0 for empty record size.  Don't push 0 size
6194         argument onto stack.
6195         (must_pass_in_stack_var_size_or_pad): Return false for empty types.
6196         * common.opt: Update -fabi-version description.
6197         * config/i386/i386.c (init_cumulative_args): Set cum->warn_empty.
6198         (ix86_gimplify_va_arg): Call arg_int_size_in_bytes instead of
6199         int_size_in_bytes.
6200         (ix86_is_empty_record): New function.
6201         (ix86_warn_parameter_passing_abi): New function.
6202         (TARGET_EMPTY_RECORD_P): Redefine.
6203         (TARGET_WARN_PARAMETER_PASSING_ABI): Redefine.
6204         * config/i386/i386.h (CUMULATIVE_ARGS): Add warn_empty.
6205         * doc/tm.texi: Regenerated.
6206         * doc/tm.texi.in (TARGET_EMPTY_RECORD_P,
6207         TARGET_WARN_PARAMETER_PASSING_ABI): Add.
6208         * dwarf2out.c (get_ultimate_context): Move to tree.c.
6209         * explow.c (hard_function_value): Call arg_int_size_in_bytes
6210         instead of int_size_in_bytes.
6211         * expr.c (copy_blkmode_to_reg): Likewise.
6212         * function.c (aggregate_value_p): Return 0 for empty types.
6213         (assign_parm_find_entry_rtl): Call warn_parameter_passing_abi
6214         target hook.
6215         (locate_and_pad_parm): Call arg size_in_bytes instead
6216         size_in_bytes.
6217         * lto-streamer-out.c (hash_tree): Hash TYPE_EMPTY_P and DECL_PADDING_P.
6218         * stor-layout.c (finalize_type_size): Set TYPE_EMPTY_P.
6219         * target.def (empty_record_p, warn_parameter_passing_abi): New target
6220         hooks.
6221         * targhooks.c (hook_void_CUMULATIVE_ARGS_tree): New hook.
6222         (std_gimplify_va_arg_expr): Skip empty records.  Call
6223         arg_size_in_bytes instead size_in_bytes.
6224         * targhooks.h (hook_void_CUMULATIVE_ARGS_tree): Declare.
6225         * tree-core.h (tree_type_common): Add empty_flag.
6226         (tree_decl_common): Update comments.
6227         * tree-streamer-in.c (unpack_ts_decl_common_value_fields): Stream
6228         DECL_PADDING_P.
6229         (unpack_ts_type_common_value_fields): Stream TYPE_EMPTY_P.
6230         * tree-streamer-out.c (pack_ts_decl_common_value_fields): Stream
6231         DECL_PADDING_P.
6232         (pack_ts_type_common_value_fields): Stream TYPE_EMPTY_P.
6233         * tree.c (default_is_empty_type): New function.
6234         (default_is_empty_record): New function.
6235         (arg_int_size_in_bytes): New function.
6236         (arg_size_in_bytes): New function.
6237         (get_ultimate_context): New function.
6238         * tree.h: Define TYPE_EMPTY_P, DECL_PADDING_P and
6239         TRANSLATION_UNIT_WARN_EMPTY_P.
6240         (default_is_empty_record, arg_int_size_in_bytes,
6241         arg_size_in_bytes, get_ultimate_context): Declare.
6243 2017-11-22  Thomas Preud'homme  <thomas.preudhomme@arm.com>
6245         * config/arm/arm.c (cmse_clear_registers): New function.
6246         (cmse_nonsecure_call_clear_caller_saved): Replace register clearing
6247         code by call to cmse_clear_registers.
6248         (cmse_nonsecure_entry_clear_before_return): Likewise.
6250 2017-11-22  Tamar Christina  <tamar.christina@arm.com>
6252         * config/arm/arm_neon.h (vdot_u32, vdotq_u32)
6253         (vdot_s32, vdotq_s32): New.
6254         (vdot_lane_u32, vdotq_lane_u32): New.
6255         (vdot_lane_s32, vdotq_lane_s32): New.
6258 2017-11-22  Richard Sandiford  <richard.sandiford@linaro.org>
6260         PR middle-end/82547
6261         * wide-int.cc (wi::add_large, wi::sub_large): Fix overflow detection
6262         for unsigned values with fewer HWIs than the precision.
6263         (test_overflow): New function.
6264         (wide_int_cc_tests): Call it.
6266 2017-11-22  Richard Sandiford  <richard.sandiford@linaro.org>
6268         * emit-rtl.c (init_derived_machine_modes): Make sure ptr_mode
6269         has the same mode class as Pmode.
6271 2017-11-22  Thomas Preud'homme  <thomas.preudhomme@arm.com>
6273         * config/arm/arm.c (cmse_nonsecure_call_clear_caller_saved): Get rid of
6274         padding_bits_to_clear_ptr.
6275         (cmse_nonsecure_entry_clear_before_return): Likewise.
6277 2017-11-22  Thomas Preud'homme  <thomas.preudhomme@arm.com>
6279         * config/arm/arm.c (cmse_nonsecure_call_clear_caller_saved): Use
6280         auto_sbitap instead of integer bitfield to control register needing
6281         clearing.
6283 2017-11-22  Jakub Jelinek  <jakub@redhat.com>
6285         PR tree-optimization/83044
6286         * tree-vrp.c (vrp_prop::check_array_ref): If eltsize is not
6287         INTEGER_CST or is 0, clear up_bound{,_p1} and later ignore tests
6288         that need the upper bound.  Subtract offset from
6289         get_addr_base_and_unit_offset only if positive and subtract it
6290         before division by eltsize rather than after it.
6292         PR debug/83084
6293         * valtrack.c (propagate_for_debug_subst, propagate_for_debug): Reset
6294         debug insns if they would contain UNSPEC_VOLATILE or volatile asm.
6295         (dead_debug_insert_temp): Likewise, but also ignore even non-volatile
6296         asm.
6298         PR middle-end/82875
6299         * optabs.c (expand_doubleword_mult, expand_binop): Before calling
6300         expand_binop with *mul_widen_optab, make sure at least one of the
6301         operands doesn't have VOIDmode.
6303         PR debug/83034
6304         * dwarf2out.c (mem_loc_descriptor): Handle VEC_SERIES.
6306         PR rtl-optimization/82044
6307         PR tree-optimization/82042
6308         * dse.c (record_store): Check for overflow.
6309         (check_mem_read_rtx): Properly check for overflow if width == -1, call
6310         add_wild_read instead of clear_rhs_from_active_local_stores on
6311         overflow and log it into dump_file.
6313 2017-11-22  Richard Biener  <rguenther@suse.de>
6315         * gimple-iterator.c (gimple_find_edge_insert_loc): Ignore
6316         fake edges to exit when looking for a place to insert.
6317         * tree-ssa-pre.c (clear_expression_ids): Inline into callers
6318         and remove.
6319         (insert_into_preds_of_block): Commit edge insertion immediately,
6320         assert that doesn't require new BBs.
6321         (fini_pre): Release expressions.
6322         (pass_pre::execute): Shuffle things around a bit, if the fn
6323         is too large do not compute AVAIL either as this is really the
6324         quadratic bit.
6326 2017-11-22  Richard Biener  <rguenther@suse.de>
6328         PR tree-optimization/83089
6329         * tree-if-conv.c (pass_if_conversion::execute): If anything
6330         changed reset SCEV and free the number of iteration estimates.
6332 2017-11-21  Martin Sebor  <msebor@redhat.com>
6334         PR tree-optimization/82945
6335         * calls.h (warn_nonstring_bound): Remove unused function.
6337 2017-11-21  Martin Sebor  <msebor@redhat.com>
6339         PR tree-optimization/82945
6340         * builtins.c (expand_builtin_strlen): Call maybe_warn_nonstring_arg.
6341         * calls.h (maybe_warn_nonstring_arg): Declare new function.
6342         * calls.c (get_attr_nonstring_decl, maybe_warn_nonstring_arg): New
6343         functions.
6344         (initialize_argument_information): Call maybe_warn_nonstring_arg.
6345         * calls.h (get_attr_nonstring_decl): Declare new function.
6346         * doc/extend.texi (attribute nonstring): Update.
6347         * gimple-fold.c (gimple_fold_builtin_strncpy): Call
6348         get_attr_nonstring_decl and handle it.
6349         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Same.  Improve
6350         detection of nul-termination.
6351         (strlen_to_stridx): Change to a pointer.
6352         (handle_builtin_strlen, handle_builtin_stxncpy): Adjust.
6353         (pass_strlen::execute): Same.
6355 2017-11-21  Sergey Shalnov  <Sergey.Shalnov@intel.com>
6357         * config/i386/i386-opts.h (enum prefer_vector_width): Added new enum
6358         for the new option -mprefer-vector-width=[none|128|256|512].
6359         * config/i386/i386.c (ix86_target_string): remove old style options
6360         -mprefer-avx256 and make -mprefer-avx128 as alias.
6361         (ix86_option_override_internal):  Apply defaults for the
6362         -mprefer-vector-width=[128|256] option.
6363         * config/i386/i386.h (TARGET_PREFER_AVX128, TARGET_PREFER_AVX256):
6364         Implement macros to work with -mprefer-vector-width=.
6365         * config/i386/i386.opt: Implemented option
6366         -mprefer-vector-width=[none|128|256|512].
6367         * doc/invoke.texi: Documentation for
6368         -mprefer-vector-width=[none|128|256|512].
6370 2017-11-21  Pat Haugen  <pthaugen@us.ibm.com>
6372         * config/rs6000/ppc-asm.h (f50, vs50): Fix values.
6374 2017-11-21  Jonathan Wakely  <jwakely@redhat.com>
6376         * doc/invoke.texi (-Wmaybe-uninitialized): Rephrase for clarity.
6378 2017-11-21  Marc Glisse  <marc.glisse@inria.fr>
6380         * doc/generic.texi: Document POINTER_DIFF_EXPR, update
6381         POINTER_PLUS_EXPR.
6382         * cfgexpand.c (expand_debug_expr): Handle POINTER_DIFF_EXPR.
6383         * expr.c (expand_expr_real_2): Likewise.
6384         * fold-const.c (const_binop, fold_addr_of_array_ref_difference,
6385         fold_binary_loc): Likewise.
6386         * match.pd (X-X, P+(Q-P), &D-P, (P+N)-P, P-(P+N), (P+M)-(P+N),
6387         P-Q==0, -(A-B), X-Z<Y-Z, (X-Z)-(Y-Z), Z-X<Z-Y, (Z-X)-(Z-Y),
6388         (A-B)+(C-A)): New transformations for POINTER_DIFF_EXPR, based on
6389         MINUS_EXPR transformations.
6390         * optabs-tree.c (optab_for_tree_code): Handle POINTER_DIFF_EXPR.
6391         * tree-cfg.c (verify_expr, verify_gimple_assign_binary): Likewise.
6392         * tree-inline.c (estimate_operator_cost): Likewise.
6393         * tree-pretty-print.c (dump_generic_node, op_code_prio,
6394         op_symbol_code): Likewise.
6395         * tree-vect-stmts.c (vectorizable_operation): Likewise.
6396         * vr-values.c (extract_range_from_binary_expr): Likewise.
6397         * varasm.c (initializer_constant_valid_p_1): Likewise.
6398         * tree.def: New tree code POINTER_DIFF_EXPR.
6400 2017-11-21  Uros Bizjak  <ubizjak@gmail.com>
6402         * config/i386/i386.md (*bswap<mode>2_movbe): Add
6403         integer suffix to movbe mnemonic.
6404         (*bswaphi2_movbe): Ditto.
6405         (bswaphi_lowpart): Merge with *bswaphi_lowpart_1.
6407 2017-11-21  Jakub Jelinek  <jakub@redhat.com>
6409         PR c++/83045
6410         * tree-cfg.c (pass_warn_function_return::execute): Formatting fix.
6411         Also warn if seen __builtin_unreachable () call with BUILTINS_LOCATION.
6412         Use LOCATION_LOCUS when comparing against UNKNOWN_LOCATION.
6414 2017-11-21  Martin Liska  <mliska@suse.cz>
6416         * tree-inline.c (expand_call_inline): Remove not needed
6417         xstrdup_for_dump.
6419 2017-11-21  James Cowgill  <James.Cowgill@imgtec.com>
6420             Jakub Jelinek  <jakub@redhat.com>
6422         PR target/82880
6423         * config/mips/frame-header-opt.c (mips_register_frame_header_opt):
6424         Remove static keyword from f variable.
6426 2017-11-21  Jakub Jelinek  <jakub@redhat.com>
6428         PR tree-optimization/83086
6429         * gimple-ssa-store-merging.c
6430         (imm_store_chain_info::try_coalesce_bswap): Test this_n.base_addr
6431         rather than n.base_addr.
6433 2017-11-21  Martin Liska  <mliska@suse.cz>
6435         PR rtl-optimization/82044
6436         PR tree-optimization/82042
6437         * dse.c (check_mem_read_rtx): Check for overflow.
6439 2017-11-21  Aldy Hernandez  <aldyh@redhat.com>
6441         * tree-ssa-threadbackward.c (find_jump_threads_backwards): Fix
6442         typo in comment.
6444 2017-11-21  Claudiu Zissulescu  <claziss@synopsys.com>
6446         * cfgrtl.c (force_nonfallthru_and_redirect): Don't split a call
6447         and its corresponding call arg location note.
6449 2017-11-21  Claudiu Zissulescu  <claziss@synopsys.com>
6450             Andrew Burgess  <andrew.burgess@embecosm.com>
6452         * config/arc/arc-protos.h (arc_compute_frame_size): Delete
6453         declaration.
6454         (arc_return_slot_offset): Likewise.
6455         (arc_eh_return_address_location): New declaration.
6456         * config/arc/arc.c (TARGET_BUILTIN_SETJMP_FRAME_VALUE): Define.
6457         (MUST_SAVE_REGISTER): Add exception handler case.
6458         (MUST_SAVE_RETURN_ADDR): Likewise.
6459         (arc_frame_pointer_required): Likewise.
6460         (arc_frame_pointer_needed): New function.
6461         (arc_compute_frame_size): Changed.
6462         (arc_expand_prologue): Likewise.
6463         (arc_expand_epilogue): Likewise.
6464         (arc_initial_elimination_offset): Likewise.
6465         (arc_return_slot_offset): Delete.
6466         (arc_eh_return_address_location): New function.
6467         (arc_builtin_setjmp_frame_value): Likewise.
6468         * config/arc/arc.h (EH_RETURN_DATA_REGNO): Use 2 registers.
6469         (EH_RETURN_STACKADJ_RTX): Define.
6470         (EH_RETURN_HANDLER_RTX): Likewise.
6471         * config/arc/arc.md (eh_return): Delete.
6473 2017-11-21  Aldy Hernandez  <aldyh@redhat.com>
6475         * print-tree.h (debug_vec_tree): Remove prototype.
6476         * gdbinit.in (pvt): Remove macro.
6478 2017-11-21  Jakub Jelinek  <jakub@redhat.com>
6480         PR tree-optimization/83047
6481         * gimple-ssa-store-merging.c
6482         (imm_store_chain_info::output_merged_store): If the loads with the
6483         same vuse are in different basic blocks, for load_gsi pick a load
6484         location that is dominated by the other loads.
6486         PR c++/83059
6487         * config/i386/i386.c (ix86_memmodel_check): Start
6488         -Winvalid-memory-model diagnostics with lowercase letter.
6490         PR debug/82718
6491         * dwarf2out.c (dw_loc_list): If crtl->has_bb_partition, temporarily
6492         set in_cold_section_p to the partition containing loc_list->first.
6493         When seeing loc_list->last_before_switch node, update secname and
6494         perform range_across_switch second partition handling only after that.
6496         PR debug/82933
6497         * run-rtl-passes.c: Include debug.h.
6498         (run_rtl_passes): Call debug_hooks->assembly_start.
6499         * dwarf2out.c (dwarf2out_assembly_start): Return early if invoked
6500         multiple times.
6502         PR target/82981
6503         * internal-fn.c (expand_mul_overflow): Use OPTAB_WIDEN instead of
6504         OPTAB_DIRECT in calls to expand_simple_binop.
6506 2017-11-20  David Malcolm  <dmalcolm@redhat.com>
6508         PR c/81404
6509         * Makefile.in (C_COMMON_OBJS): Add c-family/known-headers.o.
6511 2017-11-20  David Malcolm  <dmalcolm@redhat.com>
6513         PR c++/72786
6514         * spellcheck.h (best_match::blithely_get_best_candidate): New
6515         accessor.
6517 2017-11-20  Jakub Jelinek  <jakub@redhat.com>
6519         * config/i386/i386.c (parse_mtune_ctrl_str): Start diagnostics
6520         with lower case letter.
6522 2017-11-20  Uros Bizjak  <ubizjak@gmail.com>
6524         * config/i386/i386.md (bswaphi2): New expander.
6525         (*bswaphi2_movbe): New insn pattern.
6526         (bswaphi -> rorhi pepehole2): New peephole pattern.
6528 2017-11-20  Jeff Law  <law@redhat.com>
6530         * Makefile.in (OBJS): Add gimple-ssa-evrp-analyze.o.
6531         * gimple-ssa-evrp-analyze.c: New file pulled from gimple-ssa-evrp.c.
6532         * gimple-ssa-evrp-analyze.h: New file pulled from gimple-ssa-evrp.c.
6533         * gimple-ssa-evrp.c: Remove bits moved into new files.  Include
6534         gimple-ssa-evrp-analyze.h.
6536         * gimple-ssa-evrp.c (evrp_dom_walker::before_dom_children): Do not
6537         set BB_VISITED here.
6538         (evrp_range_analyzer::enter): Set BB_VISITED here instead.
6540 2017-11-20  Bin Cheng  <bin.cheng@arm.com>
6542         * tree-predcom.c: Add general comment on Store-Store chains.
6543         (split_data_refs_to_components): Postpone clearing eliminate_store_p
6544         flag in component.
6545         (get_chain_last_ref_at): Rename into...
6546         (get_chain_last_write_at): ...this.
6547         (get_chain_last_write_before_load): New function.
6548         (add_ref_to_chain): Promote type of chain from CT_STORE_LOAD to
6549         CT_STORE_STORE when write reference is added.
6550         (determine_roots_comp): Support load ref in CT_STORE_STORE chains.
6551         (is_inv_store_elimination_chain): Update get_chain_last_write_at call.
6552         (initialize_root_vars_store_elim_1): Ditto.
6553         (initialize_root_vars_store_elim_2): Ditto.  Replace rhs once default
6554         definition is created.
6555         (execute_pred_commoning_chain): Support load ref in CT_STORE_STORE
6556         chain by replacing it with dominant stored value.
6558 2017-11-20  Bin Cheng  <bin.cheng@arm.com>
6560         * tree-predcom.c (add_ref_to_chain): Remove check on distance.
6562 2017-11-20  Marc Glisse  <marc.glisse@inria.fr>
6564         * vr-values.c (extract_range_from_binary_expr): Use a full range
6565         for VR_VARYING.
6567 2017-11-20  Thomas Preud'homme  <thomas.preudhomme@arm.com>
6569         * config/arm/arm.md (R4_REGNUM): Define constant.
6570         (nonsecure_call_internal): Remove r4 clobber.
6571         (nonsecure_call_value_internal): Likewise.
6572         * config/arm/thumb1.md (nonsecure_call_reg_thumb1_v5): Remove second
6573         clobber and resequence match_operands.
6574         (nonsecure_call_value_reg_thumb1_v5): Likewise.
6575         * config/arm/thumb2.md (nonsecure_call_reg_thumb2): Likewise.
6576         (nonsecure_call_value_reg_thumb2): Likewise.
6578 2017-11-20  Jakub Jelinek  <jakub@redhat.com>
6580         PR tree-optimization/78821
6581         * gimple-ssa-store-merging.c (find_bswap_or_nop_load): Give up
6582         if base is TARGET_MEM_REF.  If base is not MEM_REF, set base_addr
6583         to the address of the base rather than the base itself.
6584         (find_bswap_or_nop_1): Just use pointer comparison for vuse check.
6585         (find_bswap_or_nop_finalize): New function.
6586         (find_bswap_or_nop): Use it.
6587         (bswap_replace): Return a tree rather than bool, change first
6588         argument from gimple * to gimple_stmt_iterator, allow inserting
6589         into an empty sequence, allow ins_stmt to be NULL - then emit
6590         all stmts into gsi.  Fix up MEM_REF address gimplification.
6591         (pass_optimize_bswap::execute): Adjust bswap_replace caller.
6592         (struct store_immediate_info): Add N and INS_STMT non-static
6593         data members.
6594         (store_immediate_info::store_immediate_info): Initialize them
6595         from newly added ctor args.
6596         (merged_store_group::apply_stores): Formatting fixes.  Sort by
6597         bitpos at the end.
6598         (stmts_may_clobber_ref_p): For stores call also
6599         refs_anti_dependent_p.
6600         (gather_bswap_load_refs): New function.
6601         (imm_store_chain_info::try_coalesce_bswap): New method.
6602         (imm_store_chain_info::coalesce_immediate_stores): Use it.
6603         (split_group): Handle LROTATE_EXPR and NOP_EXPR rhs_code specially.
6604         (imm_store_chain_info::output_merged_store): Fail if number of
6605         new estimated stmts is bigger or equal than old.  Handle LROTATE_EXPR
6606         and NOP_EXPR rhs_code.
6607         (pass_store_merging::process_store): Compute n and ins_stmt, if
6608         ins_stmt is non-NULL and the store rhs is otherwise invalid, use
6609         LROTATE_EXPR rhs_code.  Pass n and ins_stmt to store_immediate_info
6610         ctor.
6611         (pass_store_merging::execute): Calculate dominators.
6613         * tree-ssa-math-opts.c (nop_stats, bswap_stats, struct symbolic_number,
6614         BITS_PER_MARKER, MARKER_MASK, MARKER_BYTE_UNKNOWN, HEAD_MARKER, CMPNOP,
6615         CMPXCHG, do_shift_rotate, verify_symbolic_number_p,
6616         init_symbolic_number, find_bswap_or_nop_load, perform_symbolic_merge,
6617         find_bswap_or_nop_1, find_bswap_or_nop, pass_data_optimize_bswap,
6618         class pass_optimize_bswap, bswap_replace,
6619         pass_optimize_bswap::execute): Moved to ...
6620         * gimple-ssa-store-merging.c: ... this file.
6621         Include optabs-tree.h.
6622         (nop_stats, bswap_stats, do_shift_rotate, verify_symbolic_number_p,
6623         init_symbolic_number, find_bswap_or_nop_load, perform_symbolic_merge,
6624         find_bswap_or_nop_1, find_bswap_or_nop, bswap_replace): Put into
6625         anonymous namespace, remove static keywords.
6626         (pass_optimize_bswap::gate): Test BITS_PER_UNIT == 8 here...
6627         (pass_optimize_bswap::execute): ... rather than here.  Formatting fix.
6629 2017-11-20  Jan Hubicka  <hubicka@ucw.cz>
6631         PR bootstrap/83062
6632         * ipa-inline.c (can_inline_edge_p): Fix typo in previous patch.
6634 2017-11-20  Aldy Hernandez  <aldyh@redhat.com>
6636         * vec.h (debug_helper): New function.
6637         (DEFINE_DEBUG_VEC): New macro.
6638         * hash-set.h (debug_helper): New function.
6639         (DEFINE_DEBUG_HASH_SET): New macro.
6640         * cfg.c (debug_slim (edge)): New function.
6641         Call DEFINE_DEBUG_VEC for edges.
6642         Call DEFINE_DEBUG_HASH_SET for edges.
6643         * cfghooks.c (debug_slim (basic_block)): New function.
6644         Call DEFINE_DEBUG_VEC for basic blocks.
6645         Call DEFINE_DEBUG_HASH_SET for basic blocks.
6646         * print-tree.c (debug_slim): New function to handle trees.
6647         Call DEFINE_DEBUG_VEC for trees.
6648         Call DEFINE_DEBUG_HASH_SET for trees.
6649         (debug (vec<tree, va_gc>) &): Remove.
6650         (debug (<vec<tree, va_gc>) *): Remove.
6651         * print-rtl.c (debug_slim): New function to handle const_rtx.
6652         Call DEFINE_DEBUG_VEC for rtx_def.
6653         Call DEFINE_DEBUG_VEC for rtx_insn.
6654         Call DEFINE_DEBUG_HASH_SET for rtx_def.
6655         Call DEFINE_DEBUG_HASH_SET for rtx_insn.
6656         * sel-sched-dump.c (debug (vec<rtx_insn *> &): Remove.
6657         (debug (vec<rtx_insn *> *ptr): Remove.
6658         (debug_insn_vector): Remove.
6659         * stor-layout.c (debug_rli): Call debug() instead of debug_vec_tree.
6661 2017-11-20  Tom de Vries  <tom@codesourcery.com>
6663         PR rtl-optimization/82020
6664         * simplify-rtx.c (simplify_ternary_operation): Fix comparison mode of
6665         IF_THEN_ELSE condition.
6667 2017-11-19  Jeff Law  <law@redhat.com>
6669         * tree-ssa-dom.c (record_equivalences_from_phis): Fix handling
6670         of degenerates resulting from ignoring an edge.
6672 2017-11-19  Jan Hubicka  <hubicka@ucw.cz>
6674         PR ipa/81360
6675         * ipa-inline.c (can_inline_edge_p): Also check that caller is optimized
6677 2017-11-19  Jan Hubicka  <hubicka@ucw.cz>
6679         PR ipa/83001
6680         * profile-count.c (profile_count::to_sreal_scale): Fix return value
6681         for uninitialied counts.
6683 2017-11-19  Jan Hubicka  <hubicka@ucw.cz>
6685         PR ipa/60243
6686         * tree-inline.c (estimate_num_insns): Set to 1 at least.
6688 2017-11-19  Jan Hubicka  <hubicka@ucw.cz>
6690         PR target/82713
6691         * i386.c (ix86_builtin_vectorization_cost): Be ready for insane types.
6693 2017-11-19  Tom de Vries  <tom@codesourcery.com>
6695         * config/arc/arc.h (FUNCTION_PROFILER): Remove semicolon after
6696         "do while (0)".
6698 2017-11-19  Tom de Vries  <tom@codesourcery.com>
6700         * config/phoenix.h (TARGET_OS_CPP_BUILTINS): Remove semicolon after
6701         "do {} while (0)".
6703 2017-11-19  Tom de Vries  <tom@codesourcery.com>
6705         * config/visium/visium.h (ASM_OUTPUT_CASE_END): Remove semicolon after
6706         macro body.
6708 2017-11-19  Tom de Vries  <tom@codesourcery.com>
6710         * config/ft32/ft32.h (REGISTER_TARGET_PRAGMAS): Remove semicolon after
6711         "do {} while (0)".
6712         * config/spu/spu.h (REGISTER_TARGET_PRAGMAS): Same.
6714 2017-11-19  Tom de Vries  <tom@codesourcery.com>
6716         * config/mcore/mcore-elf.h (MCORE_EXPORT_NAME): Remove semicolon after
6717         "do {} while (0)".
6718         * config/mcore/mcore.h (ASM_OUTPUT_ALIGNED_COMMON): After missing
6719         semicolon after MCORE_EXPORT_NAME call.
6721 2017-11-19  Tom de Vries  <tom@codesourcery.com>
6723         PR target/82961
6724         * vmsdbgout.c (vmsdbgout_early_finish): New function.
6725         (vmsdbg_debug_hooks): Set early_finish field to vmsdbgout_early_finish.
6727 2017-11-18  Jan Hubicka  <hubicka@ucw.cz>
6729         * cgraphclones.c (cgraph_edge::clone): Rename gcov_count to prof_count.
6730         (cgraph_edge::clone): Cleanup updating of profile.
6731         * ipa-cp.c (update_profiling_info): Likewise.
6732         * ipa-inline-transform.c (inline_transform): Likewise.
6733         * ipa-inline.c (inline_small_functions): Add missing space to dump.
6734         * ipa-split.c (execute_split_functions): Do not split when function
6735         is cold.
6736         * predict.c (estimate_bb_frequencies): Cleanup updating of profile.
6737         * profile-count.c (profile_count::dump): Add global0.
6738         (profile_count::to_cgraph_frequency): Do not ICE when entry is
6739         undefined.
6740         (profile_count::to_sreal_scale): Likewise.
6741         (profile_count::adjust_for_ipa_scaling): Fix typo in comment.
6742         (profile_count::combine_with_ipa_count): New function.
6743         * profile-count.h (profile_guessed_global0adjusted): New.
6744         (profile_count::adjusted_zero): New.
6745         (profile_count::global0adjusted): New.
6746         (profile_count::combine_with_ipa_count): New.
6747         * tree-inline.c (copy_edges_for_bb): Add NUM/DEN arugment;
6748         correct profile of return block of split functions.
6749         (copy_cfg_body): Remove unused profile_count.
6750         (copy_body): Likewise.
6751         (expand_call_inline): Update.
6752         (tree_function_versioning): Update.
6754 2017-11-18  Aldy Hernandez  <aldyh@redhat.com>
6756         * hash-set.h (hash_set::empty): New.
6757         * tree-ssa-threadbackward.h: Delete.
6758         * tree-ssa-threadbackward.c (class thread_jumps): New.
6759         Move max_threaded_paths into class.
6760         (fsm_find_thread_path): Remove arguments that are now in class.
6761         (profitable_jump_thread_path): Rename to...
6762         (thread_jumps::profitable_jump_thread_path): ...this.
6763         (convert_and_register_jump_thread_path): Rename to...
6764         (thread_jumps::convert_and_register_current_path): ...this.
6765         (check_subpath_and_update_thread_path): Rename to...
6766         (thread_jumps::check_subpath_and_update_thread_path): ...this.
6767         (register_jump_thread_path_if_profitable): Rename to...
6768         (thread_jumps::register_jump_thread_path_if_profitable): ...this.
6769         (handle_phi): Rename to...
6770         (thread_jumps::handle_phi): ...this.
6771         (handle_assignment): Rename to...
6772         (thread_jumps::handle_assignment): ...this.
6773         (fsm_find_control_statement_thread_paths): Rename to...
6774         (thread_jumps::fsm_find_control_statement_thread_paths): ...this.
6775         (find_jump_threads_backwards): Rename to...
6776         (thread_jumps::find_jump_threads_backwards): ...this.
6777         Initialize path local data.
6778         (pass_thread_jumps::execute): Call find_jump_threads_backwards
6779         from within thread_jumps class.
6780         (pass_early_thread_jumps::execute): Same.
6782 2017-11-17  Jan Hubicka  <hubicka@ucw.cz>
6784         * cgraph.c (cgraph_node::dump): Do IPA sanity checking on IPA counts.
6786 2017-11-17  Steve Ellcey  <sellcey@cavium.com>
6788         * config/aarch64/aarch64-simd.md (fnma<mode>4): Move neg operator
6789         to canonical location.
6791 2017-11-17  Steve Ellcey  <sellcey@cavium.com>
6793         PR target/81356
6794         * config/aarch64/aarch64.c (aarch64_use_by_pieces_infrastructure_p):
6795         Remove.
6796         (TARGET_USE_BY_PIECES_INFRASTRUCTURE_P): Remove define.
6798 2017-11-17  H.J. Lu  <hongjiu.lu@intel.com>
6800         * config/gnu-user.h (GNU_USER_TARGET_STARTFILE_SPEC): Use
6801         rcrt1.o%s/grcrt1.o%s for -static-pie.
6803 2017-11-17  Jan Hubicka  <hubicka@ucw.cz>
6805         * i386.c (ix86_multiplication_cost, ix86_division_cost,
6806         ix86_shift_rotate_cost): Break out from ...
6807         (ix86_rtx_costs): ... here.
6808         (ix86_add_stmt_cost): Use rtx cost machinery to compute cost of
6809         vector operations.
6811 2017-11-17  Jan Hubicka  <hubicka@ucw.cz>
6813         * predict.c (determine_unlikely_bbs): Set cgraph node count to 0
6814         when entry block was promoted unlikely.
6815         (estimate_bb_frequencies): Increase frequency scale.
6816         * profile-count.h (profile_count): Export precision info.
6818 2017-11-17  Jan Hubicka  <hubicka@ucw.cz>
6820         * tree-tailcall.c (eliminate_tail_call): Be more careful about not
6821         disturbin profile of entry block.
6823 2017-11-17  Jan Hubicka  <hubicka@ucw.cz>
6825         * ipa-fnsummary.c (estimate_node_size_and_time): Be more tolerant for
6826         roundoff errors.
6828 2017-11-17  Jan Hubicka  <hubicka@ucw.cz>
6830         * ipa-cp.c (update_profiling_info): Handle conversion to local profile.
6831         * tree-cfg.c (execute_fixup_cfg): Do fixup same way as inliner does.
6833 2017-11-17  Jeff Law  <law@redhat.com>
6835         * gimple-ssa-evrp.c (class evrp_range_analyzer): New class extracted
6836         from evrp_dom_walker class.  Various methods moved into new class.
6837         (evrp_range_analyzer::evrp_range_analyzer): Constructor for new class.
6838         (evrp_range_analyzer::enter): New method.
6839         (evrp_range_analyzer::leave): New method.
6840         (evrp_dom_walker): Remove delegators no longer needed by this class.
6841         Replace vr_values data member with evrp_range_analyzer
6843         * gimple-ssa-evrp.c (evrp_dom_walker::record_ranges_from_phis): New
6844         method extracted from evrp_dom_walker::before_dom_children.
6845         (evrp_dom_walker::record_ranges_from_stmt): Likewise.
6846         (evrp_dom_walker::record_ranges_from_incoming_edge): Likewise.
6848         * gimple-ssa-evrp.c (evrp_dom_walker): Add cleanup method.
6849         Add private copy constructor and move assignment operators.
6850         Privatize methods and class data where trivially possible.
6851         (evrp_dom_walker::cleanup): New function, extracted from
6852         execute_early_vrp.  Simplify access to class data.
6854         * vr-values.h (get_output_for_vrp): Prototype.
6855         * vr-values.c (get_output_for_vrp): New function extracted from
6856         vrp_visit_assignment_or_call and extract_range_from_stmt.
6857         (vrp_visit_assignment_or_call): Use get_output_for_vrp.  Simplify.
6859 2017-11-17  Luis Machado  <luis.machado@linaro.org>
6861         * config/aarch64/aarch64.c
6862         (qdf24xx_prefetch_tune) <default_opt_level>: Set to -1.
6863         (qdf24xx_tunings) <autoprefetcher_model>: Set to
6864         tune_params::AUTOPREFETCHER_WEAK.
6866 2017-11-17  Tamar Christina  <tamar.christina@arm.com>
6868         PR target/82641
6869         * config/arm/arm.c (arm_valid_target_attribute_rec):
6870         Parse "arch=" and "+<ext>".
6871         (arm_valid_target_attribute_tree): Re-init global options.
6872         (arm_option_override): Make non-static.
6873         (arm_options_perform_arch_sanity_checks): Make errors fatal.
6874         * gcc/config/arm/arm-c.c (__ARM_FEATURE_CMSE): Support undef.
6875         (__ARM_FEATURE_CRC32): Support undef.
6876         * config/arm/arm_acle.h (__ARM_FEATURE_CRC32): Replace with pragma.
6877         * doc/extend.texi (ARM Function Attributes): Add pragma and target.
6879 2017-11-17  David Malcolm  <dmalcolm@redhat.com>
6881         * gdbinit.in (break-on-diagnostic): New command.
6883 2017-11-17  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
6885         * config/i386/i386.c (ix86_expand_epilogue): Change simple
6886         return to indirect jump for EH return if control-flow
6887         protection is enabled. Change explicit 'false' argument in
6888         pro_epilogue_adjust_stack with a value of
6889         flag_cf_protection.
6890         * config/i386/i386.md (simple_return_indirect_internal):
6891         Remove SImode restriction to support 64-bit.
6893 2017-11-17  Segher Boessenkool  <segher@kernel.crashing.org>
6895         * combine.c (added_notes_insn): New.
6896         (try_combine): Handle added_notes_insn like added_links_insn.
6897         Rewrite return value code.
6898         (distribute_notes): Set added_notes_insn to the earliest insn we added
6899         a note to.
6901 2017-11-17  Segher Boessenkool  <segher@kernel.crashing.org>
6903         PR rtl-optimization/82621
6904         * combine.c (try_combine): Do not split PARALLELs of two SETs if the
6905         dest of one of those SETs is unused.
6907 2017-11-17  Richard Biener  <rguenther@suse.de>
6909         PR fortran/83017
6910         * tree-core.h (enum annot_expr_kind): Add annot_expr_parallel_kind.
6911         * tree-pretty-print.c (dump_generic_node): Handle
6912         annot_expr_parallel_kind.
6913         * tree-cfg.c (replace_loop_annotate_in_block): Likewise.
6914         * gimplify.c (gimple_boolify): Likewise.
6916 2017-11-17  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
6918         * config.gcc (extra_headers): Add cet.h for x86 targets.
6919         * config/i386/cet.h: New file.
6920         * doc/install.texi: Add --enable-cet/--disable-cet.
6922 2017-11-17  Richard Biener  <rguenther@suse.de>
6924         PR tree-optimization/83017
6925         * tree-parloops.c (MIN_PER_THREAD): Use --param parloops-min-per-thread.
6926         (gen_parallel_loop): Properly count iterations.
6927         (parallelize_loops): Handle loop->can_be_parallel independent
6928         of flag_loop_parallelize_all.  Make static profitability test match
6929         the runtime one.
6930         * params.def (PARAM_PARLOOPS_MIN_PER_THREAD): New.
6931         * invoke.texi (parloops-min-per-thread): Document.
6933 2017-11-17  Vineet Gupta  <vgupta@synopsys.com>
6935         * config/arc/linux.h: GLIBC_DYNAMIC_LINKER update per glibc
6936         upstreaming review comments.
6938 2017-11-17  Tamar Christina  <tamar.christina@arm.com>
6940         * expr.c (copy_blkmode_to_reg): Fix bitsize for targets
6941         with fast unaligned access.
6942         * doc/sourcebuild.texi (word_mode_no_slow_unalign): New.
6944 2017-11-17  Thomas Preud'homme  <thomas.preudhomme@arm.com>
6946         * config/arm/arm.c (cmse_nonsecure_entry_clear_before_return): Allocate
6947         to_clear_arg_regs_bitmap to the same size as to_clear_bitmap.
6949 2017-11-17  Richard Biener  <rguenther@suse.de>
6951         * tree-ssa-pre.c (phi_translate_1): Remove redundant constant
6952         folding of references.
6954 2017-11-17  Qing Zhao  <qing.zhao@oracle.com>
6956         PR middle-end/78809
6957         * gimple-fold.c (gimple_fold_builtin_string_compare): Add handling
6958         of replacing call to strncmp with corresponding call to strcmp when
6959         meeting conditions.
6961 2017-11-17  Sergey Shalnov  <Sergey.Shalnov@intel.com>
6963         * config/i386/x86-tune.def (X86_TUNE_AVX256_OPTIMAL): Add tuning
6964         option prefer-avx256 for skylake-avx512 configuration.
6965         * config/i386/i386.c (ix86_option_override_internal): Ditto.
6966         (get_builtin_code_for_version): Ditto.
6968 2017-11-17  Chung-Ju Wu  <jasonwucj@gmail.com>
6969             Monk Chiang  <sh.chiang04@gmail.com>
6971         * config/nds32/nds32.h (FIRST_PSEUDO_REGISTER): Modify.
6972         (FIXED_REGISTERS): Reserve more register numbers.
6973         (CALL_USED_REGISTERS): Likewise.
6974         (REG_ALLOC_ORDER): Likewise.
6975         (REG_CLASS_CONTENTS): Likewise.
6976         (REGISTER_NAMES): Likewise.
6978 2017-11-17  Chung-Ju Wu  <jasonwucj@gmail.com>
6979             Kito Cheng  <kito.cheng@gmail.com>
6981         * config/nds32/nds32-modes.def: Add vector mode V4QI V2HI V8QI V4HI
6982         V2SI.
6983         * config/nds32/iterators.md: Add vector mode iterators and attributes.
6985 2017-11-16  Steven Munroe  <munroesj@gcc.gnu.org>
6987         * config/rs6000/mmintrin.h (_mm_add_pi32[_ARCH_PWR]): Correct
6988         parameter list for vec_splats.
6990 2017-11-16  Joseph Myers  <joseph@codesourcery.com>
6992         * doc/invoke.texi (-std=c17): Refer to 2018 expected publication
6993         date of C17.
6994         (-std=c18, -std=iso9899:2018, -std=gnu18): Document option aliases.
6996 2017-11-16  Thomas Koenig  <tkoenig@gcc.gnu.org>
6998         PR bootstrap/82856
6999         * doc/install.texi: Document incompatibility of Perl >=5.6.26
7000         with the required version of automake 1.11.6.
7002 2017-11-16  Pat Haugen  <pthaugen@us.ibm.com>
7004         * rs6000/power9.md (power9fpdiv): New automaton and cpu_unit defined
7005         for it.
7006         (DU_C2_3_power9): Correct reservation combinations.
7007         (FP_DIV_power9, VEC_DIV_power9): New.
7008         (power9-alu): Split out rotate/shift...
7009         (power9-rot): ...to here, correct dispatch resource.
7010         (power9-cracked-alu, power9-mul, power9-mul-compare): Correct dispatch
7011         resource.
7012         (power9-fp): Correct latency.
7013         (power9-sdiv): Add div/sqrt resource.
7014         (power9-ddiv): Correct latency, add div/sqrt resource.
7015         (power9-sqrt, power9-dsqrt): Add div/sqrt resource.
7016         (power9-vecfdiv, power9-vecdiv): Correct latency, add div/sqrt
7017         resource.
7018         (power9-qpdiv, power9-qpmul): Adjust resource usage.
7020 2017-11-15  Michael Meissner  <meissner@linux.vnet.ibm.com>
7022         * config/rs6000/rs6000.c (rs6000_expand_builtin): Do not do the
7023         switch statement mapping KF built-ins to TF built-ins if we don't
7024         have the proper ISA 3.0 assembler support.
7026 2017-11-16  Jan Hubicka  <hubicka@ucw.cz>
7028         * tree-emutls.c (lower_emutls_data): Remove unused bb_freq.
7029         (lower_emutls_function_body): Do not compute it.
7031 2017-11-16  Jan Hubicka  <hubicka@ucw.cz>
7033         * ipa-split.c (split_bb_info): Turn time to sreal.
7034         (split_point): Likewise.
7035         (dump_split_point): Likewise.
7036         (fine_split_points): Likewise.
7037         (execute_split_functions): Only zero split_bbs; turn time to sreals.
7039 2017-11-16  Jan Hubicka  <hubicka@ucw.cz>
7041         * ipa-fnsummary.c (analyze_function_body): Accumulate time consistently
7042         in sreal.
7044 2017-11-16  Jan Hubicka  <hubicka@ucw.cz>
7046         * predict.c (combine_predictions_for_bb): Preserve zero predicted
7047         edges.
7048         (expensive_function_p): Remove useless assert.
7049         (determine_unlikely_bbs): Propagate also forward; determine cold blocks
7051 2017-11-16  Martin Sebor  <msebor@redhat.com>
7053         PR tree-optimization/82588
7054         PR tree-optimization/82583
7055         * tree-vrp.c (check_array_ref): Handle flexible array members,
7056         string literals, and inner indices.
7057         (search_for_addr_array): Add detail to diagnostics.
7059 2017-11-16  Nathan Sidwell  <nathan@acm.org>
7061         PR c++/82836
7062         PR c++/82737
7063         * tree.h (COPY_DECL_RTL): Rename parms for clarity.
7064         (SET_DECL_ASSEMBLER_NAME): Forward to
7065         overwrite_decl_assembler_name.
7066         (COPY_DECL_ASSEMBLER_NAME): Rename parms for clarity.
7067         (overwrite_decl_assembler_name): Declare.
7068         * tree.c (overwrite_decl_assembler_name): New.
7069         * langhooks-def.h (lhd_overwrite_decl_assembler_name): Declare.
7070         (LANG_HOOKS_OVERWRITE_DECL_ASSEMBLER_NAME): Provide default.
7071         (LANG_HOOKS_INITIALIZER): Add it.
7072         * langhooks.h (struct lang_hooks): Add overwrite_decl_assembler_name.
7073         * langhooks.c (lhd_set_decl_assembler_name): Use
7074         SET_DECL_ASSEMBLER_NAME.
7075         (lhd_overwrite_decl_assembler_name): Default implementation.
7077 2017-11-16  Wilco Dijkstra  <wdijkstr@arm.com>
7078             Jackson Woodruff  <jackson.woodruff@arm.com>
7080         PR tree-optimization/71026
7081         * match.pd: Canonicalize constant multiplies in division.
7083 2017-11-16  Wilco Dijkstra  <wdijkstr@arm.com>
7085         * opts.c (default_options_table): Add OPT_fomit_frame_pointer entry.
7086         * common/config/alpha/alpha-common.c (TARGET_OPTION_OPTIMIZATION_TABLE):
7087         Remove OPT_fomit_frame_pointer entry.
7088         * common/config/arc/arc-common.c: Likewise.
7089         * common/config/arm/arm-common.c: Likewise.
7090         * common/config/avr/avr-common.c: Likewise.
7091         * common/config/c6x/c6x-common.c: Likewise.
7092         * common/config/cr16/cr16-common.c: Likewise.
7093         * common/config/cris/cris-common.c: Likewise.
7094         * common/config/epiphany/epiphany-common.c: Likewise.
7095         * common/config/fr30/fr30-common.c: Likewise.
7096         * common/config/frv/frv-common.c: Likewise.
7097         * common/config/ia64/ia64-common.c: Likewise.
7098         * common/config/iq2000/iq2000-common.c: Likewise.
7099         * common/config/lm32/lm32-common.c: Likewise.
7100         * common/config/m32r/m32r-common.c: Likewise.
7101         * common/config/mcore/mcore-common.c: Likewise.
7102         * common/config/microblaze/microblaze-common.c: Likewise.
7103         * common/config/mips/mips-common.c: Likewise.
7104         * common/config/mmix/mmix-common.c: Likewise.
7105         * common/config/mn10300/mn10300-common.c: Likewise.
7106         * common/config/nios2/nios2-common.c: Likewise.
7107         * common/config/pa/pa-common.c: Likewise.
7108         * common/config/pdp11/pdp11-common.c: Likewise.
7109         * common/config/powerpcspe/powerpcspe-common.c: Likewise.
7110         * common/config/riscv/riscv-common.c: Likewise.
7111         * common/config/rs6000/rs6000-common.c: Likewise.
7112         * common/config/rx/rx-common.c: Likewise.
7113         * common/config/s390/s390-common.c: Likewise.
7114         * common/config/sh/sh-common.c: Likewise.
7115         * common/config/sparc/sparc-common.c: Likewise.
7116         * common/config/tilegx/tilegx-common.c: Likewise.
7117         * common/config/tilepro/tilepro-common.c: Likewise.
7118         * common/config/v850/v850-common.c: Likewise.
7119         * common/config/visium/visium-common.c: Likewise.
7120         * common/config/xstormy16/xstormy16-common.c: Likewise.
7121         * common/config/xtensa/xtensa-common.c: Likewise.
7122         * invoke.texi (-fomit-frame-pointer): Update documentation.
7124 2017-11-16  Jan Hubicka  <hubicka@ucw.cz>
7126         * tree-cfg.c (gimple_find_sub_bbs): Do not compute freq.
7128 2017-11-14  Jan Hubicka  <hubicka@ucw.cz>
7130         * cfg.c (scale_bbs_frequencies_int,
7131         cale_bbs_frequencies_gcov_type): Remove.
7132         * cfg.h (scale_bbs_frequencies_int,
7133         cale_bbs_frequencies_gcov_type): Remove.
7135 2017-11-14  Jan Hubicka  <hubicka@ucw.cz>
7137         * tree-ssa-loop-manip.c
7138         (scale_dominated_blocks_in_loop): Update to profile counts.
7139         (tree_transform_and_unroll_loop): Likewise.
7141 2017-11-14  Jan Hubicka  <hubicka@ucw.cz>
7143         * tree-vect-loop-manip.c (vect_do_peeling): Do not use
7144         scale_bbs_frequencies_int.
7146 2017-11-14  Jan Hubicka  <hubicka@ucw.cz>
7148         * final.c (compute_alignments): Use counts rather than frequencies.
7150 2017-11-14  Jan Hubicka  <hubicka@ucw.cz>
7152         * cfgloopanal.c: Include sreal.h
7153         (average_num_loop_insns): Use counts and sreal for accounting.
7155 2017-11-14  Jan Hubicka  <hubicka@ucw.cz>
7157         * cfgloopmanip.c (duplicate_loop_to_header_edge): Cleanup profile
7158         manipulation.
7160 2017-11-16  Chung-Ju Wu  <jasonwucj@gmail.com>
7161             Kito Cheng  <kito.cheng@gmail.com>
7163         * config/nds32/constraints.md: Provide more constraints.
7164         * config/nds32/nds32.h (enum reg_class, REG_CLASS_NAMES,
7165         REG_CLASS_CONTENTS): Define R5_REG, R8_REG, and FRAME_POINTER_REG to
7166         support constraints usage.
7168 2017-11-16  Chung-Ju Wu  <jasonwucj@gmail.com>
7170         * config/nds32/constants.md (UNSPEC_VOLATILE_FUNC_RETURN): Remove.
7172 2017-11-16  Chung-Ju Wu  <jasonwucj@gmail.com>
7173             Kito Cheng  <kito.cheng@gmail.com>
7175         * config/nds32/nds32.opt: Add mext-perf, mext-perf2, mext-string.
7176         * config/nds32/nds32.opt: Refine the layout.
7177         * config/nds32/nds32.c (TARGET_EXT_PERF, TARGET_EXT_PERF2,
7178         TARGET_EXT_STRING): Support new options.
7179         * config/nds32/nds32.h: Likewise.
7180         * config/nds32/nds32.md: Likewise.
7181         * config/nds32/nds32-predicates.c: Likewise.
7182         * config/nds32/constraints.md: Likewise.
7183         * common/config/nds32/nds32-common.c: Likewise.
7185 2017-11-16  Julia Koval  <julia.koval@intel.com>
7187         PR target/82983
7188         * config/i386/gfniintrin.h: Add sse check.
7189         * config/i386/i386.c (ix86_expand_builtin): Fix gfni check.
7191 2017-11-16  Julia Koval  <julia.koval@intel.com>
7193         * common/config/i386/i386-common.c (OPTION_MASK_ISA_AVX512VBMI2_SET,
7194         OPTION_MASK_ISA_AVX512VBMI2_UNSET): New.
7195         (ix86_handle_option): Handle -mavx512vbmi2.
7196         * config/i386/cpuid.h: Add bit_AVX512VBMI2.
7197         * config/i386/driver-i386.c (host_detect_local_cpu): Handle new bit.
7198         * config/i386/i386-c.c (__AVX512VBMI2__): New.
7199         * config/i386/i386.c (ix86_target_string): Handle -mavx512vbmi2.
7200         (ix86_valid_target_attribute_inner_p): Ditto.
7201         * config/i386/i386.h (TARGET_AVX512VBMI2, TARGET_AVX512VBMI2_P): New.
7202         * config/i386/i386.opt (mavx512vbmi2): New option.
7203         * doc/invoke.texi: Add new option.
7205 2017-11-16  Julia Koval  <julia.koval@intel.com>
7207         * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8, _mm256_gf2p8mul_epi8,
7208         _mm_mask_gf2p8mul_epi8, _mm_maskz_gf2p8mul_epi8,
7209         _mm256_mask_gf2p8mul_epi8, _mm256_maskz_gf2p8mul_epi8,
7210         _mm512_mask_gf2p8mul_epi8, _mm512_maskz_gf2p8mul_epi8,
7211         _mm512_gf2p8mul_epi8): New intrinsics.
7212         * config/i386/i386-builtin-types.def
7213         (V64QI_FTYPE_V64QI_V64QI): New type.
7214         * config/i386/i386-builtin.def (__builtin_ia32_vgf2p8mulb_v64qi,
7215         __builtin_ia32_vgf2p8mulb_v64qi_mask, __builtin_ia32_vgf2p8mulb_v32qi,
7216         __builtin_ia32_vgf2p8mulb_v32qi_mask, __builtin_ia32_vgf2p8mulb_v16qi,
7217         __builtin_ia32_vgf2p8mulb_v16qi_mask): New builtins.
7218         * config/i386/sse.md (vgf2p8mulb_*): New pattern.
7219         * config/i386/i386.c (ix86_expand_args_builtin): Handle new type.
7221 2017-11-15  Uros Bizjak  <ubizjak@gmail.com>
7223         * config/i386/i386.c (x86_print_call_or_nop): Emit 5 byte nop
7224         explicitly as a stream of bytes.
7226 2017-11-15  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
7228         * config/rs6000/altivec.h (vec_xst_be): New #define.
7229         * config/rs6000/altivec.md (altivec_vperm_<mode>_direct): Rename
7230         and externalize from *altivec_vperm_<mode>_internal.
7231         * config/rs6000/rs6000-builtin.def (XL_BE_V16QI): Remove macro
7232         instantiation.
7233         (XL_BE_V8HI): Likewise.
7234         (XL_BE_V4SI): Likewise.
7235         (XL_BE_V4SI): Likewise.
7236         (XL_BE_V2DI): Likewise.
7237         (XL_BE_V4SF): Likewise.
7238         (XL_BE_V2DF): Likewise.
7239         (XST_BE): Add BU_VSX_OVERLOAD_X macro instantiation.
7240         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Correct
7241         all array entries with these keys: VSX_BUILTIN_VEC_XL,
7242         VSX_BUILTIN_VEC_XL_BE, VSX_BUILTIN_VEC_XST.  Add entries for key
7243         VSX_BUILTIN_VEC_XST_BE.
7244         * config/rs6000/rs6000.c (altivec_expand_xl_be_builtin): Remove.
7245         (altivec_expand_builtin): Remove handling for VSX_BUILTIN_XL_BE_*
7246         built-ins.
7247         (altivec_init_builtins): Replace conditional calls to def_builtin
7248         for __builtin_vsx_ld_elemrev_{v8hi,v16qi} and
7249         __builtin_vsx_st_elemrev_{v8hi,v16qi} based on TARGET_P9_VECTOR
7250         with unconditional calls.  Remove calls to def_builtin for
7251         __builtin_vsx_le_be_<mode>.  Add a call to def_builtin for
7252         __builtin_vec_xst_be.
7253         * config/rs6000/vsx.md (vsx_ld_elemrev_v8hi): Convert define_insn
7254         to define_expand, and add alternate RTL generation for P8.
7255         (*vsx_ld_elemrev_v8hi_internal): New define_insn based on
7256         vsx_ld_elemrev_v8hi.
7257         (vsx_ld_elemrev_v16qi): Convert define_insn to define_expand, and
7258         add alternate RTL generation for P8.
7259         (*vsx_ld_elemrev_v16qi_internal): New define_insn based on
7260         vsx_ld_elemrev_v16qi.
7261         (vsx_st_elemrev_v8hi): Convert define_insn
7262         to define_expand, and add alternate RTL generation for P8.
7263         (*vsx_st_elemrev_v8hi_internal): New define_insn based on
7264         vsx_st_elemrev_v8hi.
7265         (vsx_st_elemrev_v16qi): Convert define_insn to define_expand, and
7266         add alternate RTL generation for P8.
7267         (*vsx_st_elemrev_v16qi_internal): New define_insn based on
7268         vsx_st_elemrev_v16qi.
7270 2017-11-15  H.J. Lu  <hongjiu.lu@intel.com>
7272         PR target/82990
7273         * config/i386/i386.c (pass_insert_vzeroupper::gate): Remove
7274         TARGET_AVX512ER check.
7275         (ix86_option_override_internal): Set MASK_VZEROUPPER if
7276         neither -mzeroupper nor -mno-zeroupper is used and
7277         TARGET_EMIT_VZEROUPPER is set.
7278         * config/i386/i386.h (TARGET_EMIT_VZEROUPPER): New.
7279         * config/i386/x86-tune.def: Add X86_TUNE_EMIT_VZEROUPPER.
7281 2017-11-15  Will Schmidt  <will_schmidt@vnet.ibm.com>
7283         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add support for
7284         folding of vector compares.
7285         (fold_build_vec_cmp): New helper function.
7286         (fold_compare_helper): New helper function.
7287         (builtin_function_type): Add compare builtins to the list of functions
7288         having unsigned arguments.  Cosmetic updates to comment indentation.
7289         * config/rs6000/vsx.md (vcmpneb, vcmpneh, vcmpnew): Update to specify
7290         the not+eq combination.
7292 2017-11-15  Bin Cheng  <bin.cheng@arm.com>
7294         PR tree-optimization/82726
7295         PR tree-optimization/70754
7296         * tree-predcom.c (order_drefs_by_pos): New function.
7297         (combine_chains): Move code setting has_max_use_after to...
7298         (try_combine_chains): ...here.  New parameter.  Sort combined chains
7299         according to position information.
7300         (tree_predictive_commoning_loop): Update call to above function.
7301         (update_pos_for_combined_chains, pcom_stmt_dominates_stmt_p): New.
7303 2017-11-15  Bin Cheng  <bin.cheng@arm.com>
7305         PR tree-optimization/82726
7306         Revert
7307         2017-01-23  Bin Cheng  <bin.cheng@arm.com>
7309         PR tree-optimization/70754
7310         * tree-predcom.c (stmt_combining_refs): New parameter INSERT_BEFORE.
7311         (reassociate_to_the_same_stmt): New parameter INSERT_BEFORE.  Insert
7312         combined stmt before it if not NULL.
7313         (combine_chains): Process refs reversely and compute dominance point
7314         for root ref.
7316         Revert
7317         2017-02-23  Bin Cheng  <bin.cheng@arm.com>
7319         PR tree-optimization/79663
7320         * tree-predcom.c (combine_chains): Process refs in reverse order
7321         only for ZERO length chains, and add explaining comment.
7323 2017-11-15  Tamar Christina  <tamar.christina@arm.com>
7325         * config/arm/arm-cpus.in (armv8_3, ARMv8_3a, armv8.3-a): New
7326         * config/arm/arm-tables.opt (armv8.3-a): Regenerated.
7327         * doc/invoke.texi (ARM Options): Add armv8.3-a.
7329 2017-11-15  Tamar Christina  <tamar.christina@arm.com>
7331         * config/arm/arm.h (TARGET_DOTPROD): Add arm_arch8_2.
7333 2017-11-15  Martin Liska  <mliska@suse.cz>
7335         * tree-cfg.c (pass_warn_function_return::execute):
7336         Compare warn_return_type for greater than zero.
7338 2017-11-15  Sebastian Peryt  <sebastian.peryt@intel.com>
7340         PR target/82941
7341         PR target/82942
7342         * config/i386/i386.c (pass_insert_vzeroupper): Modify gate condition
7343         to return true on Xeon and not on Xeon Phi.
7344         (ix86_check_avx256_register): Changed to ...
7345         (ix86_check_avx_upper_register): ... this. Add extra check for
7346         VALID_AVX512F_REG_OR_XI_MODE.
7347         (ix86_avx_u128_mode_needed): Changed
7348         ix86_check_avx256_register to ix86_check_avx_upper_register.
7349         (ix86_check_avx256_stores): Changed to ...
7350         (ix86_check_avx_upper_stores): ... this. Changed
7351         ix86_check_avx256_register to ix86_check_avx_upper_register.
7352         (ix86_avx_u128_mode_after): Changed
7353         avx_reg256_found to avx_upper_reg_found. Changed
7354         ix86_check_avx256_stores to ix86_check_avx_upper_stores.
7355         (ix86_avx_u128_mode_entry): Changed
7356         ix86_check_avx256_register to ix86_check_avx_upper_register.
7357         (ix86_avx_u128_mode_exit): Ditto.
7358         * config/i386/i386.h: (host_detect_local_cpu): New define.
7360 2017-11-15  Dominik Infuehr  <dominik.infuehr@theobroma-systems.com>
7362         * config/arm/xgene1.md (xgene1): Split into automatons
7363         xgene1_main, xgene1_decoder, xgene1_div, xgene1_simd.
7364         (xgene1_f_load): Adjust reservations and/or types.
7365         (xgene1_f_store): Likewise.
7366         (xgene1_load_pair): Likewise.
7367         (xgene1_store_pair): Likewise.
7368         (xgene1_fp_load1): Likewise.
7369         (xgene1_load1): Likewise.
7370         (xgene1_store1): Likewise.
7371         (xgene1_move): Likewise.
7372         (xgene1_alu): Likewise.
7373         (xgene1_simd): Likewise.
7374         (xgene1_bfm): Likewise.
7375         (xgene1_neon_load1): Likewise.
7376         (xgene1_neon_store1): Likewise.
7377         (xgene1_neon_logic): Likewise.
7378         (xgene1_neon_st1): Likewise.
7379         (xgene1_neon_ld1r): Likewise.
7380         (xgene1_alu_cond): Added.
7381         (xgene1_shift_reg): Likwise.
7382         (xgene1_bfx): Likewise.
7383         (xgene1_mul): Split into xgene1_mul32, xgene1_mul64.
7385 2017-11-15  Jakub Jelinek  <jakub@redhat.com>
7387         PR target/82981
7388         * internal-fn.c: Include gimple-ssa.h, tree-phinodes.h and
7389         ssa-iterators.h.
7390         (can_widen_mult_without_libcall): New function.
7391         (expand_mul_overflow): If only checking unsigned mul overflow,
7392         not result, and can do efficiently MULT_HIGHPART_EXPR, emit that.
7393         Don't use WIDEN_MULT_EXPR if it would involve a libcall, unless
7394         no other way works.  Add MULT_HIGHPART_EXPR + MULT_EXPR support.
7395         (expand_DIVMOD): Formatting fix.
7396         * expmed.h (expand_mult): Add NO_LIBCALL argument.
7397         * expmed.c (expand_mult): Likewise.  Use OPTAB_WIDEN rather
7398         than OPTAB_LIB_WIDEN if NO_LIBCALL is true, and allow it to fail.
7400         PR tree-optimization/82977
7401         * tree-ssa-strlen.c (strlen_optimize_stmt): Pass a reference to a copy
7402         constructed temporary to strlen_to_stridx.put.
7404 2017-11-15  Martin Liska  <mliska@suse.cz>
7406         * configure.ac: Remove -fkeep-inline-functions from coverage_flags.
7407         * configure: Regenerate.
7409 2017-11-15  Martin Liska  <mliska@suse.cz>
7411         PR target/82927
7412         * config/sh/sh-mem.cc: Use proper probability for
7413         REG_BR_PROB_NOTE.
7415 2017-11-14  Jeff Law  <law@redhat.com>
7417         * explow.c (anti_adjust_stack_and_probe_stack_clash): Avoid probing
7418         the red zone for stack_clash_protection_final_dynamic_probe targets
7419         when the total dynamic stack size is zero bytes.
7421         * tree-ssa-threadupdate.c (thread_through_all_blocks): Thread
7422         blocks is post order.
7424 2017-11-15  Alexandre Oliva  <aoliva@redhat.com>
7426         * dumpfile.h (TDF_COMPARE_DEBUG): New.
7427         * final.c (rest_of_clean_state): Set it for the
7428         -fcompare-debug dump.
7429         * tree-pretty-print.c (dump_generic_node): Omit OBJ_TYPE_REF
7430         class when TDF_COMPARE_DEBUG is set.
7432         * dwarf2out.c (gen_producer_string): Discard
7433         OPT_fcompare_debug.
7435 2017-11-15  Joseph Myers  <joseph@codesourcery.com>
7437         PR c/81156
7438         * doc/extend.texi (Other Builtins): Document __builtin_tgmath.
7439         * ginclude/tgmath.h (__tg_cplx, __tg_ldbl, __tg_dbl, __tg_choose)
7440         (__tg_choose_2, __tg_choose_3, __TGMATH_REAL_1_2)
7441         (__TGMATH_REAL_2_3): Remove macros.
7442         (__TGMATH_CPLX, __TGMATH_CPLX_2, __TGMATH_REAL, __TGMATH_REAL_2)
7443         (__TGMATH_REAL_3, __TGMATH_CPLX_ONLY): Define using
7444         __builtin_tgmath.
7445         (frexp, ldexp, nexttoward, scalbn, scalbln): Define using
7446         __TGMATH_REAL_2.
7447         (remquo): Define using __TGMATH_REAL_3.
7449 2017-11-14  Jeff Law  <law@redhat.com>
7451         * vr-values.c: New file with contents extracted from tree-vrp.c.
7452         * Makefile.in (OBJS): Add vr-values.o
7453         * tree-vrp.h (set_value_range_to_nonnull): Prototype.
7454         (set_value_range, set_and_canonicalize_value_range): Likewise.
7455         (vrp_bitmap_equal_p, range_is_nonnull): Likewise.
7456         (value_range_constant_singleton, symbolic_range_p): Likewise.
7457         (compare_values, compare_values_warnv, vrp_val_is_min): Likewise.
7458         (vrp_val_is_max, copy_value_range, set_value_range_to_value): Likewise.
7459         (extract_range_from_binary_expr_1, vrp_val_min, vrp_val_max): Likewise.
7460         (set_value_range_to_null, range_int_cst_p, opreand_less_p): Likewise.
7461         (find_case_label_range, find_case_label_index): Likewise.
7462         (zero_nonzero_bits_from_vr, overflow_comparison_p): Likewise.
7463         (range_int_cst_singleton_p, value_inside_range): Likewise.
7464         (get_single_symbol): Likewise.
7465         (switch_update): Move structure definition here.
7466         (to_remove_edges, to_update_switch_stmts): Provide externs.
7467         * tree-vrp.c: Move all methods for vr-values class to vr-values.c
7468         (vrp_val_max, vrp_val_min, vrp_val_is_max): Make externally visible.
7469         (vrp_val_is_min, set_value_range): Likewise.
7470         (set_and_canonicalize_value_range, copy_value_range): Likewise.
7471         (set_value_range_to_value, set_value_range_to_nonnull): Likewise.
7472         (set_value_range_to_null, vrp_bitmap_equal_p): Likewise.
7473         (range_is_nonnull, range_int_cst_p): Likewwise.
7474         (range_int_cst_singleton_p, symbolic_range_p): Likewise.
7475         (get_single_symbol, operand_less_p): Likewise
7476         (compare_values_warnv, compare_values): Likewise.
7477         (value_inside_range, value_range_constant_singleton): Likewise.
7478         (zero_nonzero_bitgs_from_vr): Likewise.
7479         (extract_range_from_binary_expr_1): Likewise.
7480         (overflow_comparison_p): Likewise.
7481         (to_remove_edges, to_update_switch_stmts): Likewise.
7482         (find_case_label-index, find_case_label_range): Likewise.
7483         (switch_update, set_value_range_to_nonnegative): Remove.
7484         (set_value_range_to_truthvalue): Likewise.
7485         (symbolic_range_based_on_p, gimple_assign_nonzero_p): Likewise.
7486         (gimple_stmt_nonzero_p, compare_ranges): Likewise.
7487         (compare_range_with_value, vrp_valueize, vrp_valueize_1): Likewise.
7488         (find_case_label_ranges, test_for_singularity): Likewise.
7489         (range_fits_type_p, simplify_conversion_using_ranges): LIkewise.
7490         (x_vr_values): Move to its remaining use site.
7492 2017-11-10  Jeff Law  <law@redhat.com>
7494         * vr-values.h (VR_INITIALIZER): Move #define here.
7495         * gimple-ssa-evrp.c: New file with contents extracted from tree-vrp.c
7496         * Makefile.in (OBJS): Add tree-evrp.o
7497         * tree-vrp.h (assert_info): Move structure definition here.
7498         (set_value_range_to_varying): Prototype.
7499         (vrp_operand_equal_p, range_includes_zero_p): Likewise.
7500         (infer_value_range, register_edge_assert_for): Likewise.
7501         (stmt_interesting_for_vrp): Likewise.
7502         * tree-vrp.c: Move all methods for evrp class into gimple-ssa-evrp.c.
7503         (set_value_range_to_varying): No longer static.
7504         (vrp_operand_equal_p, range_includes_zero_p): Likewise.
7505         (infer_value_range, register_edge_assert_for): Likewise.
7507 2017-11-14  Michael Meissner  <meissner@linux.vnet.ibm.com>
7509         * config/rs6000/rs6000.md (bswapdi2): On 32-bit ISA 3.0, don't
7510         generate the XXBRD instruction.
7512         * config/rs6000/rs6000-c.c (is_float128_p): New helper function.
7513         (rs6000_builtin_type_compatible): Treat _Float128 and long double
7514         as being compatible if -mabi=ieeelongdouble.
7515         * config/rs6000/rs6000-builtin.def (BU_FLOAT128_HW_1): New macros
7516         to setup float128 built-ins with hardware support.
7517         (BU_FLOAT128_HW_2): Likewise.
7518         (BU_FLOAT128_HW_3): Likewise.
7519         (BU_FLOAT128_HW_VSX_1): Likewise.
7520         (BU_FLOAT128_HW_VSX_2): Likewise.
7521         (scalar_extract_expq): Change float128 built-in functions to
7522         accommodate having both KFmode and TFmode functions.  Use the
7523         KFmode variant as the default.
7524         (scalar_extract_sigq): Likewise.
7525         (scalar_test_neg_qp): Likewise.
7526         (scalar_insert_exp_q): Likewise.
7527         (scalar_insert_exp_qp): Likewise.
7528         (scalar_test_data_class_qp): Likewise.
7529         (sqrtf128_round_to_odd): Delete processing the round to odd
7530         built-in functions as special built-in functions, and define them
7531         as float128 built-ins.  Use the KFmode variant as the default.
7532         (truncf128_round_to_odd): Likewise.
7533         (addf128_round_to_odd): Likewise.
7534         (subf128_round_to_odd): Likewise.
7535         (mulf128_round_to_odd): Likewise.
7536         (divf128_round_to_odd): Likewise.
7537         (fmaf128_round_to_odd): Likewise.
7538         * config/rs6000/rs6000.c (rs6000_expand_binop_builtin): Add
7539         support for KFmode and TFmode xststdcqp calls.
7540         (rs6000_expand_builtin): If long double is IEEE 128-bit floating
7541         point, switch the built-in handlers for the get/set float128
7542         exponent, get float128 mantissa, float128 test built-ins, and the
7543         float128 round to odd built-in functions.  Eliminate creating the
7544         float128 round to odd built-in functions as special built-ins.
7545         (rs6000_init_builtins): Eliminate special creation of the float128
7546         round to odd built-in functions.
7547         * config/rs6000/vsx.md (xsxexpqp_<mode>): Change float128 built-in
7548         function insns to support both TFmode and KFmode variants.
7549         (xsxsigqp_<mode>): Likewise.
7550         (xsiexpqpf_<mode>): Likewise.
7551         (xsiexpqp_<mode>): Likewise.
7552         (xststdcqp_<mode>): Likewise.
7553         (xststdcnegqp_<mode>): Likewise.
7554         (xststdcqp_<mode>): Likewise.
7556 2017-11-14  Jan Hubicka  <hubicka@ucw.cz>
7558         * tree-ssa-threadupdate.c (compute_path_counts): Remove
7559         unused path_in_freq_ptr parameter.
7560         (ssa_fix_duplicate_block_edges): Do not pass around path_in_freq
7562 2017-11-14  Jan Hubicka  <hubicka@ucw.cz>
7564         * ipa-inline.c (edge_badness): Dump sreal frequency.
7565         (compute_inlined_call_time): Match natural implementaiton ...
7566         * ipa-fnsummary.c (estimate_edge_size_and_time): ... here; remove
7567         forgotten division by CGRAPH_FREQ_BASE.
7569 2017-11-14  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
7571         * config.gcc (*-*-solaris2*): Enable default_use_cxa_atexit since
7572         Solaris 11.  Update comment.
7573         * configure.ac (gcc_cv_ld_pid): Adapt comment for Solaris 12
7574         renaming.
7575         * config/sol2.h (STARTFILE_SPEC): Likewise.
7576         * configure: Regenerate.
7578 2017-11-14  Carl Love  <cel@us.ibm.com>
7580         * config/rs6000/rs6000.c (swap_endian_selector_for_mode): Remove
7581         le_ and be_ prefixes to swap* variables.  Remove
7582         if (VECTOR_ELT_ORDER_BIG) statement. Remove E_V16QImode case
7583         statements.
7585 2017-11-14  Jason Merrill  <jason@redhat.com>
7587         Support GTY((cache)) on hash_map.
7588         * hash-traits.h (ggc_remove): Add ggc_maybe_mx member function.
7589         (ggc_cache_remove): Override it instead of ggc_mx.
7590         * hash-table.h (gt_ggc_mx): Call it instead of ggc_mx.
7591         (gt_cleare_cache): Call ggc_mx instead of gt_ggc_mx.
7592         * hash-map-traits.h (simple_hashmap_traits): Add maybe_mx member.
7593         (simple_cache_map_traits): Override maybe_mx.
7594         * hash-map.h (hash_entry): Add ggc_maybe_mx and keep_cache_entry.
7595         (hash_map): Friend gt_cleare_cache.
7596         (gt_cleare_cache): New.
7597         * tree.h (tree_cache_traits): New hash_map traits class.
7598         (tree_cache_map): New typedef.
7600 2017-11-14  Richard Biener  <rguenther@suse.de>
7602         * tree-cfgcleanup.c (cleanup_control_expr_graph): Remove first_p
7603         paramter and handling.
7604         (cleanup_control_flow_bb): Likewise.
7605         (cleanup_control_flow_pre): New helper performing a DFS walk
7606         to call cleanup_control_flow_bb in PRE order.
7607         (cleanup_tree_cfg_1): Do the first phase of cleanup_control_flow_bb
7608         via cleanup_control_flow_pre.
7610 2017-11-14  James Greenhalgh  <james.greenhalgh@arm.com>
7612         * config/aarch64/aarch64-simd.md
7613         (aarch64_simd_bsl<mode>_internal): Remove DImode.
7614         (*aarch64_simd_bsl<mode>_alt): Likewise.
7615         (aarch64_simd_bsldi_internal): New.
7616         (aarch64_simd_bsldi_alt): Likewise.
7618 2017-11-13  Jan Hubicka  <hubicka@ucw.cz>
7620         * tracer.c (better_p): Do not compare frequencies.
7621         * reg-stack.c (better_edge): Likewise.
7622         * shrink-wrap.c (try_shrink_wrapping): Do not convert to gcov counts
7623         and back.
7625 2017-11-13  Jan Hubicka  <hubicka@ucw.cz>
7627         * auto-profile.c (afdo_annotate_cfg): Use update_max_bb_count.
7628         * cgraphunit.c (cgraph_node::expand_thunk): Use update_max_bb_count.
7629         * ipa-utils.c (ipa_merge_profiles): Use update_max_bb_count.
7630         * lto-streamer-in.c (input_function): Use update_max_bb_count.
7631         * omp-expand.c (expand_omp_taskreg): Use update_max_bb_count.
7632         * predict.c (maybe_hot_frequency_p): Inline to ...
7633         (maybe_hot_count_p): ... here; rewrite to counts.
7634         (counts_to_freqs): Rename to ...
7635         (update_max_bb_count): ... this one.
7636         (expensive_function_p): Use counts.
7637         (estimate_bb_frequencies): Update.
7638         (rebuild_frequencies): Update.
7639         * predict.h (counts_to_freqs): Rename to ...
7640         (update_max_bb_count): ... this one.
7641         * profile.c (compute_branch_probabilities): Add debug info
7642         * tree-inline.c (expand_call_inline): Update debug info.
7643         (optimize_inline_calls): Use update_max_bb_count..
7644         (tree_function_versioning): Use update_max_bb_count..
7645         * value-prof.c (gimple_value_profile_transformations):
7646         Do not use update_max_bb_count.
7648 2017-11-13  Jan Hubicka  <hubicka@ucw.cz>
7650         * ipa-inline.c (compute_uninlined_call_time, compute_inlined_call_time):
7651         always use frequencies.
7653 2017-11-13  Jan Hubicka  <hubicka@ucw.cz>
7655         * bb-reorder.c: Remove frequencies from comments.
7656         (better_edge_p): Use profile counts.
7657         (find_traces): Dump profile counts.
7658         (rotate_loop): Use profile counts.
7659         (find_traces_1_round): Likewise.
7660         (connect_better_edge_p): Use counts instead of probabilities for
7661         reverse walk.
7662         (copy_bb_p): Drop early check for non-0 frequency.
7663         (sanitize_hot_paths): Update comments.
7665 2017-11-13  Jan Hubicka  <hubicka@ucw.cz>
7667         * ipa-split.c (struct split_point): Add count.
7668         (consider_split): Do not compute incoming frequency; compute incoming
7669         count and store it to split_point.
7670         (split_function): Set count of the call to split part correctly.
7672 2017-11-13  Carl Love  <cel@us.ibm.com>
7674         * config/rs6000/altivec.md (altivec_vsumsws_be): Add define_expand.
7676 2017-11-13  Tom Tromey  <tom@tromey.com>
7678         * doc/cpp.texi (Variadic Macros): Document __VA_OPT__.
7680 2017-11-13  Carl Love  <cel@us.ibm.com>
7682         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins):
7683         Add support for builtins:
7684         unsigned int vec_first_{,miss}_match_{,or_eos}index,
7685         vector {un,}signed {char,int,short},
7686         vector {un,}signed {char,int,short}) arguments.
7687         * config/rs6000/rs6000-builtin.def (VFIRSTMATCHINDEX,
7688         VFIRSTMATCHOREOSINDEX, VFIRSTMISMATCHINDEX, VFIRSTMISMATCHOREOSINDEX):
7689         Add BU_P9V_AV_2 expansions for the builtins.
7690         * config/rs6000/altivec.h (vec_first_match_index,
7691         vec_first_mismatch_index, vec_first_match_or_eos_index,
7692         vec_first_mismatch_or_eos_index): Add #defines for the builtins.
7693         * config/rs6000/rs6000-protos.h (bytes_in_mode): Add
7694         new extern declaration.
7695         * config/rs6000/rs6000.c (bytes_in_mode): Add new function.
7696         * config/rs6000/vsx.md (first_match_index_<mode>,
7697         first_match_or_eos_index_<mode>, first_mismatch_index_<mode>,
7698         first_mismatch_or_eos_index_<mode>): Add define expand.
7699         (vctzlsbb_<mode>): Add mode field to define_insn for vctzlsbb.
7700         * doc/extend.texi: Update the built-in documenation file for the new
7701         built-in functions.
7703 2017-11-13  Michael Meissner  <meissner@linux.vnet.ibm.com>
7705         * match.pd: Convert fminf<N>, fminf<N>x, fmax<N>, and fmax<N>x
7706         into the min/max operations for _Float<N> and _Float<N>X types.
7708 2017-11-13  Eric Botcazou  <ebotcazou@adacore.com>
7710         PR lto/81351
7711         * dwarf2out.c (do_eh_frame): New static variable.
7712         (dwarf2out_begin_prologue): Set it.
7713         (dwarf2out_frame_finish): Test it instead of dwarf2out_do_eh_frame.
7715 2017-11-13  Jan Hubicka  <hubicka@ucw.cz>
7717         * tree-ssa-coalesce.c (coalesce_cost): Fix formating.
7719         * tree-ssa-sink.c (select_best_block): Do not use frequencies.
7721 2017-11-13  Eric Botcazou  <ebotcazou@adacore.com>
7723         PR lto/81351
7724         * debug.h (dwarf2out_do_eh_frame): Declare.
7725         * dwarf2cfi.c (dwarf2out_do_eh_frame): New predicate.
7726         (dwarf2out_do_frame): Use it.
7727         (dwarf2out_do_cfi_asm): Likewise.
7728         * dwarf2out.c (dwarf2out_frame_finish): Likewise.
7729         (dwarf2out_assembly_start): Likewise.
7730         (dwarf2out_begin_prologue): Fix comment.
7731         * toplev.c (compile_file): Always call dwarf2out_frame_finish
7732         if the target needs either debug or unwind DWARF2 info.
7733         * lto-opts.c (lto_write_options): Do not save -fexceptions,
7734         -fnon-call-exceptions, -ffp-contract, -fmath-errno, -fsigned-zeros,
7735         -ftrapping-math, -ftrapv and -fwrapv.
7737 2017-11-13  Jan Hubicka  <hubicka@ucw.cz>
7739         * cgraph.c (cgraph_edge::sreal_frequency): New function.
7740         * cgraph.h (cgraph_edge::sreal_frequency): Declare.
7741         * ipa-fnsummary.c (dump_ipa_call_summary): Use sreal_frequency.
7742         (estimate_edge_size_and_time): Likewise.
7743         (ipa_merge_fn_summary_after_inlining): Likewise.
7744         * ipa-inline.c (cgraph_freq_base_rec): Remove.
7745         (compute_uninlined_call_time): Use sreal_frequency.
7746         (compute_inlined_call_time): Likewise.
7747         (ipa_inline): Do not initialize cgraph_freq_base_rec.
7748         * profile-count.c: Include sreal.h.
7749         (profile_count::to_sreal_scale): New.
7750         * profile-count.h: Forward declare sreal.
7751         (profile_count::to_sreal_scale): Declare.
7753 2017-11-13  Nathan Sidwell  <nathan@acm.org>
7755         * diagnostic.c (maybe_line_and_column): New.
7756         (diagnostic_get_location_text): Use it.
7757         (diagnostic_report_current_module): Likewise.
7758         (test_diagnostic_get_location_text): Add tests.
7760 2017-11-13  Luis Machado  <luis.machado@linaro.org>
7762         * doc/md.texi (Specifying processor pipeline description): Fix
7763         incorrect latency for the div instruction example.
7765 2017-11-13  Jakub Jelinek  <jakub@redhat.com>
7767         PR tree-optimization/78821
7768         * gimple-ssa-store-merging.c (compatible_load_p): Don't require
7769         that bit_not_p is the same.
7770         (imm_store_chain_info::coalesce_immediate_stores): Likewise.
7771         (split_group): Count precisely bit_not_p bits in each statement.
7772         (invert_op): New function.
7773         (imm_store_chain_info::output_merged_store): Use invert_op to
7774         emit BIT_XOR_EXPR with a xor_mask instead of BIT_NOT_EXPR if some
7775         but not all orig_stores have BIT_NOT_EXPR in the corresponding spots.
7777 2017-11-13  Martin Liska  <mliska@suse.cz>
7779         * gcov.c (struct coverage_info): Remove typedef of coverage_t.
7780         (struct source_info): Likewise.
7781         (add_branch_counts): Likewise.
7782         (add_line_counts): Likewise.
7783         (function_summary): Likewise.
7784         (output_intermediate_line): Likewise.
7785         (generate_results): Likewise.
7787 2017-11-13  Martin Liska  <mliska@suse.cz>
7789         * gcov.c (struct block_info): Remove typedef for block_t.
7790         (struct line_info): Likewise.
7791         (line_info::has_block): Likewise.
7792         (EXIT_BLOCK): Likewise.
7793         (unblock): Likewise.
7794         (circuit): Likewise.
7795         (get_cycles_count): Likewise.
7796         (process_file): Likewise.
7797         (read_graph_file): Likewise.
7798         (solve_flow_graph): Likewise.
7799         (find_exception_blocks): Likewise.
7800         (add_line_counts): Likewise.
7801         (accumulate_line_info): Likewise.
7802         (output_line_details): Likewise.
7804 2017-11-13  Martin Liska  <mliska@suse.cz>
7806         * gcov.c (struct arc_info): Remove typedef for arc_t.
7807         (struct line_info): Likewise.
7808         (add_branch_counts): Likewise.
7809         (output_branch_count): Likewise.
7810         (function_info::~function_info): Likewise.
7811         (circuit): Likewise.
7812         (output_intermediate_line): Likewise.
7813         (read_graph_file): Likewise.
7814         (solve_flow_graph): Likewise.
7815         (find_exception_blocks): Likewise.
7816         (add_line_counts): Likewise.
7817         (accumulate_line_info): Likewise.
7818         (output_line_details): Likewise.
7819         (output_function_details): Likewise.
7821 2017-11-13  Martin Liska  <mliska@suse.cz>
7823         * gcov.c (struct function_info): Remove typedef for function_t.
7824         (struct source_info): Likewise.
7825         (source_info::get_functions_at_location): Likewise.
7826         (solve_flow_graph): Likewise.
7827         (find_exception_blocks): Likewise.
7828         (add_line_counts): Likewise.
7829         (output_intermediate_file): Likewise.
7830         (process_file): Likewise.
7831         (generate_results): Likewise.
7832         (release_structures): Likewise.
7833         (read_graph_file): Likewise.
7834         (read_count_file): Likewise.
7835         (accumulate_line_counts): Likewise.
7836         (output_lines): Likewise.
7838 2017-11-13  Martin Liska  <mliska@suse.cz>
7840         * gcov.c (function_info::function_info): Remove num_counts
7841         and add vector<gcov_type>.
7842         (function_info::~function_info): Use the vector.
7843         (process_file): Likewise.
7844         (read_graph_file): Likewise.
7845         (read_count_file): Likewise.
7846         (solve_flow_graph): Likewise.
7848 2017-11-13  Martin Liska  <mliska@suse.cz>
7850         * gcov.c (function_info::is_artificial): New function.
7851         (process_file): Erase all artificial early.
7852         (generate_results): Skip as all artificial are already
7853         removed.
7855 2017-11-13  Martin Liska  <mliska@suse.cz>
7857         * gcov.c (read_graph_file): Store to global vector of functions.
7858         (read_count_file): Iterate the vector.
7859         (process_file): Likewise.
7860         (generate_results): Likewise.
7861         (release_structures): Likewise.
7863 2017-11-13  Jakub Jelinek  <jakub@redhat.com>
7865         PR tree-optimization/82954
7866         * gimple-ssa-store-merging.c
7867         (imm_store_chain_info::coalesce_immediate_stores): If
7868         !infof->ops[N].base_addr, split group if info->ops[N].base_addr.
7870 2017-11-13  Richard Sandiford  <richard.sandiford@linaro.org>
7872         * config/aarch64/aarch64-simd.md (aarch64_store_lane0<mode>):
7873         Upddate call to ENDIAN_LANE_N.
7874         (aarch64_<sur>dot_lane<vsi2qi>): Use aarch64_endian_lane_rtx.
7875         (aarch64_<sur>dot_laneq<vsi2qi>): Likewise.
7876         (*aarch64_simd_vec_copy_lane<mode>): Update calls to ENDIAN_LANE_N
7877         and use aarch64_endian_lane_rtx.
7878         (*aarch64_simd_vec_copy_lane_<vswap_width_name><mode>): Likewise.
7880 2017-11-12  Tom de Vries  <tom@codesourcery.com>
7882         * config/riscv/riscv.h (ASM_OUTPUT_LABELREF): Wrap in do {} while (0).
7884 2017-11-12  Tom de Vries  <tom@codesourcery.com>
7886         * config/elfos.h (ASM_OUTPUT_ASCII): Remove semicolon after macro body.
7888 2017-11-12  Tom de Vries  <tom@codesourcery.com>
7890         * config/cr16/cr16.h (ASM_OUTPUT_LABELREF): Remove semicolon after macro
7891         body.
7892         * config/powerpcspe/xcoff.h (ASM_OUTPUT_LABELREF): Same.
7893         * config/rs6000/xcoff.h (ASM_OUTPUT_LABELREF): Same.
7894         * defaults.h (ASM_OUTPUT_LABELREF): Same.
7896 2017-11-11  Martin Sebor  <msebor@redhat.com>
7898         PR c/81117
7899         * doc/extend.texi (attribute nonstring): Remove spurious argument.
7901         PR bootstrap/82948
7902         * prefic.c (translate_name): Replace strncpy with memcpy to
7903         avoid -Wstringop-truncation.
7905 2017-11-10  Jan Hubicka  <hubicka@ucw.cz>
7907         * tree-ssa-loop-im.c (execute_sm_if_changed): Do not compute freq_sum.
7909 2017-11-10  Jan Hubicka  <hubicka@ucw.cz>
7911         * predict.c (maybe_hot_frequency_p): Do not use cfun.
7913 2017-11-10  Jan Hubicka  <hubicka@ucw.cz>
7915         * tree-ssa-tail-merge.c (replace_block_by): Fix and re-enable profile
7916         merging.
7918 2017-11-10  Michael Meissner  <meissner@linux.vnet.ibm.com>
7920         * config/rs6000/rs6000.md (bswaphi2_reg): On ISA 3.0 systems,
7921         enable generating XXBRH if the value is in a vector register.
7922         (bswapsi2_reg): On ISA 3.0 systems, enable generating XXBRW if the
7923         value is in a vector register.
7924         (bswapdi2_reg): On ISA 3.0 systems, always use XXBRD to do
7925         register to register bswap64's instead of doing the GPR sequence
7926         used on previous machines.
7927         (bswapdi2_xxbrd): New insn.
7928         (bswapdi2_reg): Disallow on ISA 3.0.
7929         (register to register bswap64 splitter): Do not split the insn on
7930         ISA 3.0 systems that use XXBRD.
7932 2017-11-10  Martin Sebor  <msebor@redhat.com>
7934         PR c/81117
7935         * config/darwin-c.c (framework_construct_pathname): Replace strncpy
7936         with memcpy.
7937         (find_subframework_file): Same.
7939 2017-11-10  Jan Hubicka  <hubicka@ucw.cz>
7941         * auto-profile.c (afdo_indirect_call): Drop frequency.
7942         * cgraph.c (symbol_table::create_edge): Drop frequency argument.
7943         (cgraph_node::create_edge): Drop frequency argument.
7944         (cgraph_node::create_indirect_edge): Drop frequency argument.
7945         (cgraph_edge::make_speculative): Drop frequency arguments.
7946         (cgraph_edge::resolve_speculation): Do not update frequencies
7947         (cgraph_edge::dump_edge_flags): Do not dump frequency.
7948         (cgraph_node::dump): Check consistency in IPA mode.
7949         (cgraph_edge::maybe_hot_p): Use IPA counter.
7950         (cgraph_edge::verify_count_and_frequency): Rename to ...
7951         (cgraph_edge::verify_count): ... this one; drop frequency checking.
7952         (cgraph_node::verify_node): Update.
7953         * cgraph.h (struct cgraph_edge): Drop frequency.
7954         (cgraph_edge::frequency): New function.
7955         * cgraphbuild.c (pass_build_cgraph_edges::execute): Donot pass
7956         frequencies.
7957         (cgraph_edge::rebuild_edges): Likewise.
7958         * cgraphclones.c (cgraph_edge::clone): Scale only counts.
7959         (duplicate_thunk_for_node): Do not pass frequency.
7960         (cgraph_node::create_clone): Scale only counts.
7961         (cgraph_node::create_virtual_clone): Do not pass frequency.
7962         (cgraph_node::create_edge_including_clones): Do not pass frequency.
7963         (cgraph_node::create_version_clone): Do not pass frequency.
7964         * cgraphunit.c (cgraph_node::analyze): Do not pass frequency.
7965         (cgraph_node::expand_thunk): Do not pass frequency.
7966         (cgraph_node::create_wrapper): Do not pass frequency.
7967         * gimple-iterator.c (update_call_edge_frequencies): Do not pass
7968         frequency.
7969         * gimple-streamer-in.c (input_bb): Scale only IPA counts.
7970         * ipa-chkp.c (chkp_produce_thunks): Do not pass frequency.
7971         * ipa-cp.c (ipcp_lattice::print): Use frequency function.
7972         (gather_caller_stats): Use frequency function.
7973         (ipcp_cloning_candidate_p): Use frequency function.
7974         (ipcp_propagate_stage): Use frequency function.
7975         (get_info_about_necessary_edges): Use frequency function.
7976         (update_profiling_info): Update only IPA profile.
7977         (update_specialized_profile): Use frequency functoin.
7978         (perhaps_add_new_callers): Update only IPA profile.
7979         * ipa-devirt.c (ipa_devirt): Use IPA profile.
7980         * ipa-fnsummary.c (redirect_to_unreachable): Do not set frequrency.
7981         (dump_ipa_call_summary): Use frequency function.
7982         (estimate_edge_size_and_time): Use frequency function.
7983         (ipa_merge_fn_summary_after_inlining): Use frequency function.
7984         * ipa-inline-analysis.c (do_estimate_edge_time): Use IPA profile.
7985         * ipa-inline-transform.c (update_noncloned_frequencies): Rename to ..
7986         (update_noncloned_counts): ... ths one; scale counts only.
7987         (clone_inlined_nodes): Do not scale frequency.
7988         (inline_call): Do not pass frequency.
7989         * ipa-inline.c (compute_uninlined_call_time): Use IPA profile.
7990         (compute_inlined_call_time): Use IPA profile.
7991         (want_inline_small_function_p): Use IPA profile.
7992         (want_inline_self_recursive_call_p): Use IPA profile.
7993         (edge_badness): Use IPA profile.
7994         (lookup_recursive_calls): Use IPA profile.
7995         (recursive_inlining): Do not pass frequency.
7996         (resolve_noninline_speculation): Do not update frequency.
7997         (inline_small_functions): Collect max of IPA profile.
7998         (dump_overall_stats): Dump IPA porfile.
7999         (dump_inline_stats): Dump IPA porfile.
8000         (ipa_inline): Collect IPA stats.
8001         * ipa-inline.h (clone_inlined_nodes): Update prototype.
8002         * ipa-profile.c (ipa_propagate_frequency_1): Use frequency function.
8003         (ipa_propagate_frequency): Use frequency function.
8004         (ipa_profile): Cleanup.
8005         * ipa-prop.c (ipa_make_edge_direct_to_target): Do not pass frequency
8006         * ipa-utils.c (ipa_merge_profiles): Merge all profiles.
8007         * lto-cgraph.c (lto_output_edge): Do not stream frequency.
8008         (input_node): Do not stream frequency.
8009         (input_edge): Do not stream frequency.
8010         (merge_profile_summaries): Scale only IPA profiles.
8011         * omp-simd-clone.c (simd_clone_adjust): Do not pass frequency.
8012         * predict.c (drop_profile): Do not recompute frequency.
8013         * trans-mem.c (ipa_tm_insert_irr_call): Do not pass frequency.
8014         (ipa_tm_insert_gettmclone_call): Do not pass frequency.
8015         * tree-cfg.c (execute_fixup_cfg): Drop profile to global0 if needed.
8016         * tree-chkp.c (chkp_copy_bounds_for_assign): Do not pass frequency.
8017         * tree-emutls.c (gen_emutls_addr): Do not pass frequency.
8018         * tree-inline.c (copy_bb): Do not scale frequency.
8019         (expand_call_inline): Do not scale frequency.
8020         (tree_function_versioning): Do not scale frequency.
8021         * ubsan.c (ubsan_create_edge): Do not pass frequency.
8023 2017-11-10  Julia Koval  <julia.koval@intel.com>
8025         * config/i386/gfniintrin.h (_mm_gf2p8affine_epi64_epi8)
8026         (_mm256_gf2p8affine_epi64_epi8, _mm_mask_gf2p8affine_epi64_epi8)
8027         (_mm_maskz_gf2p8affine_epi64_epi8, _mm256_mask_gf2p8affine_epi64_epi8)
8028         (_mm256_maskz_gf2p8affine_epi64_epi8)
8029         (_mm512_mask_gf2p8affine_epi64_epi8, _mm512_gf2p8affine_epi64_epi8)
8030         (_mm512_maskz_gf2p8affine_epi64_epi8): New intrinsics.
8031         * config/i386/i386-builtin.def (__builtin_ia32_vgf2p8affineqb_v64qi)
8032         (__builtin_ia32_vgf2p8affineqb_v32qi)
8033         (__builtin_ia32_vgf2p8affineqb_v16qi): New builtins.
8034         * config/i386/sse.md (vgf2p8affineqb_<mode><mask_name>): New pattern.
8036 2017-11-10  Tamar Christina  <tamar.christina@arm.com>
8038         PR target/82641
8039         * config/arm/arm.c
8040         (arm_option_override): Refactor.
8041         (arm_option_reconfigure_globals): New.
8042         (arm_options_perform_arch_sanity_checks): New.
8043         * config/arm/arm-protos.h (arm_option_reconfigure_globals):
8044         New prototype.
8045         (arm_options_perform_arch_sanity_checks): Likewise
8047 2017-11-10  Pat Haugen  <pthaugen@us.ibm.com>
8049         * rs6000/power9.md (power9-qpdiv): Correct DFU pipe usage.
8050         (power9-qpmul): New.
8051         * rs6000/rs6000.md ("type" attr): Add qmul.
8052         (mul<mode>3, fma<mode>4_hw, *fms<mode>4_hw, *nfma<mode>4_hw,
8053         *nfms<mode>4_hw, mul<mode>3_odd, fma<mode>4_odd, *fms<mode>4_odd,
8054         *nfma<mode>4_odd, *nfms<mode>4_odd): Change type to qmul.
8056 2017-11-10  Martin Sebor  <msebor@redhat.com>
8058         PR c/81117
8059         * builtins.c (compute_objsize): Handle arrays that
8060         compute_builtin_object_size likes to fail for.  Make extern.
8061         * builtins.h (compute_objsize): Declare.
8062         (check_strncpy_sizes): New function.
8063         (expand_builtin_strncpy): Call check_strncpy_sizes.
8064         * gimple-fold.c (gimple_fold_builtin_strncpy): Implement
8065         -Wstringop-truncation.
8066         (gimple_fold_builtin_strncat): Same.
8067         * gimple.c (gimple_build_call_from_tree): Set call location.
8068         * tree-ssa-strlen.c (strlen_to_stridx): New global variable.
8069         (maybe_diag_bound_equal_length, is_strlen_related_p): New functions.
8070         (handle_builtin_stxncpy, handle_builtin_strncat): Same.
8071         (handle_builtin_strlen): Use strlen_to_stridx.
8072         (strlen_optimize_stmt): Handle flavors of strncat, strncpy, and
8073         stpncpy.
8074         Use strlen_to_stridx.
8075         (pass_strlen::execute): Release strlen_to_stridx.
8076         * doc/invoke.texi (-Wsizeof-pointer-memaccess): Document enhancement.
8077         (-Wstringop-truncation): Document new option.
8079 2017-11-10  Martin Liska  <mliska@suse.cz>
8081         PR gcov-profile/82702
8082         * gcov.c (main): Handle intermediate files in a different
8083         way.
8084         (get_gcov_intermediate_filename): New function.
8085         (output_gcov_file): Remove support of intermediate files.
8086         (generate_results): Allocate intermediate file.
8087         (release_structures): Clean-up properly fn_end.
8088         (output_intermediate_file): Start iterating with line 1.
8090 2017-11-10  Jakub Jelinek  <jakub@redhat.com>
8092         PR tree-optimization/82929
8093         * gimple-ssa-store-merging.c (struct store_immediate_info): Add
8094         ops_swapped_p non-static data member.
8095         (store_immediate_info::store_immediate_info): Clear it.
8096         (imm_store_chain_info::coalesce_immediate_stores): If swapping
8097         ops set ops_swapped_p.
8098         (count_multiple_uses): Handle ops_swapped_p.
8100 2017-11-10  Martin Liska  <mliska@suse.cz>
8102         * coverage.c (coverage_init): Stream information about
8103         support of has_unexecuted_blocks.
8104         * doc/gcov.texi: Document that.
8105         * gcov-dump.c (dump_gcov_file): Support it in gcov_dump tool.
8106         * gcov.c (read_graph_file): Likewise.
8107         (output_line_beginning): Fix a small issue with
8108         color output.
8110 2017-11-10  Bin Cheng  <bin.cheng@arm.com>
8112         * tree-predcom.c (determine_roots_comp): Avoid memory leak by freeing
8113         reference of trivial component.
8115 2017-11-10  Jakub Jelinek  <jakub@redhat.com>
8117         PR bootstrap/82916
8118         * gimple-ssa-store-merging.c
8119         (pass_store_merging::terminate_all_aliasing_chains): For
8120         gimple_store_p stmts also call refs_output_dependent_p.
8122         PR rtl-optimization/82913
8123         * compare-elim.c (try_merge_compare): Punt if def_insn is not
8124         single set.
8126 2017-11-09  Jeff Law  <law@redhat.com>
8128         * vr-values.h: New file with vr_values class.
8129         * tree-vrp.c: Include vr-values.h
8130         (vrp_value_range_pool, vrp_equiv_obstack, num_vr_values): Move static
8131         data objects into the vr_values class.
8132         (vr_value, values_propagated, vr_phi_edge_counts): Likewise.
8133         (get_value_range): Make it a member function within vr_values class.
8134         (set_defs_to_varying, update_value_range, add_equivalence): Likewise.
8135         (vrp_stmt_computes_nonzero_p, op_with_boolean_value_range_p): Likewise.
8136         (op_with_constant_singleton_value_range): Likewise.
8137         (extract_range_for_var_from_comparison_expr): Likewise.
8138         (extract_range_from_assert, extract_range_from_ssa_name): Likewise.
8139         (extract_range_from_binary_expr): Likewise.
8140         (extract_range_from_unary_expr): Likewise.
8141         (extract_range_from_cond_expr, extrat_range_from_comparison): Likewise.
8142         (check_for_binary_op_overflow, extract_range_basic): Likewise.
8143         (extract_range_from_assignment, adjust_range_with_scev): Likewise.
8144         (dump_all_value_ranges, get_vr_for_comparison): Likewise.
8145         (compare_name_with_value, compare_names): Likewise.
8146         (vrp_evaluate_conditional_warnv_with_ops_using_ranges): Likewise.
8147         (vrp_evaluate_conditional_warnv_with_ops): Likewise.  Remove prototype.
8148         (vrp_evaluate_conditional, vrp_visit_cond_stmt): Likewise.
8149         (vrp_visit_switch_stmt, extract_range_from_stmt): Likewise.
8150         (extract_range_from_phi_node): Likewise.
8151         (simplify_truth_ops_using_ranges): Likewise.
8152         (simplify_div_or_mod_using_ranges): Likewise.
8153         (simplify_min_or_max_using_ranges, simplify_abs_using_ranges): Likewise.
8154         (simplify_bit_ops_using_ranges, simplify_cond_using_ranges_1): Likewise.
8155         (simplify_cond_using_ranges_2, simplify_switch_using_ranges): Likewise.
8156         (simplify_float_conversion_using_ranges): Likewise.
8157         (simplify_internal_call_using_ranges): Likewise.
8158         (two_valued_val_range_p, simplify_stmt_using_ranges): Likewise.
8159         (vrp_visit_assignment_or_call): Likewise.  Smuggle class instance
8160         poitner via x_vr_values for calls into gimple folder.
8161         (vrp_initialize_lattice): Make this the vr_values ctor.
8162         (vrp_free_lattice): Make this the vr_values dtor.
8163         (set_vr_value): New function.
8164         (class vrp_prop): Add vr_values data member.  Add various member
8165         functions as well as member functions that delegate to vr_values.
8166         (check_array_ref): Make a member function within vrp_prop class.
8167         (search_for_addr_array, vrp_initialize): Likewise.
8168         (vrp_finalize): Likewise.  Revamp to avoid direct access to
8169         vr_value, values_propagated, etc.
8170         (check_array_bounds): Extract vrp_prop class instance pointer from
8171         walk info structure.  Use it to call member functions.
8172         (check_all_array_refs): Make a member function within vrp_prop class.
8173         Smuggle class instance pointer via walk info structure.
8174         (x_vr_values): New local static.
8175         (vrp_valueize): Use x_vr_values to get class instance.
8176         (vr_valueize_1): Likewise.
8177         (class vrp_folder): Add vr_values data member.  Add various member
8178         functions as well as member functions that delegate to vr_values.
8179         (fold_predicate_in): Make a mber fucntion within vrp_folder class.
8180         (simplify_stmt_for_jump_threading): Extract smuggled vr_values
8181         class instance from vr_values.  Use it to call member functions.
8182         (vrp_dom_walker): Add vr_values data member.
8183         (vrp_dom_walker::after_dom_children): Smuggle vr_values class
8184         instance via x_vr_values.
8185         (identify_jump_threads): Accept vr_values as argument.  Store
8186         it into the walker structure.
8187         (evrp_dom_walker): Add vr_values class data member.  Add various
8188         delegators.
8189         (evrp_dom_walker::try_find_new_range): Use vr_values data
8190         member to access the memory allocator.
8191         (evrp_dom_walker::before_dom_children): Store vr_values class
8192         instance into the vrp_folder class.
8193         (evrp_dom_walker::push_value_range): Rework to avoid direct
8194         access to num_vr_values and vr_value.
8195         (evrp_dom_walker::pop_value_range): Likewise.
8196         (execute_early_vrp): Remove call to vrp_initialize_lattice.
8197         Use vr_values to get to dump_all_value_ranges member function.
8198         Remove call to vrp_free_lattice.  Call vrp_initialize, vrp_finalize,
8199         and simplify_cond_using_ranges_2 via vrp_prop class instance.
8200         Pass vr_values class instance down to identify_jump_threads.
8201         Remove call to vrp_free_lattice.
8202         (debug_all_value_ranges): Remove.
8204         * tree-vrp.c (vrp_prop): Move class to earlier point in the file.
8205         (vrp_folder): Likewise.
8207         * tree-vrp.c (set_value_range): Do not reference vrp_equiv_obstack.
8208         Get it from the existing bitmap instead.
8209         (vrp_intersect_ranges_1): Likewise.
8211 2017-11-09  Jakub Jelinek  <jakub@redhat.com>
8213         * gimple-ssa-store-merging.c (struct store_immediate_info): Add
8214         bit_not_p field.
8215         (store_immediate_info::store_immediate_info): Add bitnotp argument,
8216         set bit_not_p to it.
8217         (imm_store_chain_info::coalesce_immediate_stores): Break group
8218         if bit_not_p is different.
8219         (count_multiple_uses, split_group,
8220         imm_store_chain_info::output_merged_store): Handle info->bit_not_p.
8221         (handled_load): Avoid multiple chained BIT_NOT_EXPRs.
8222         (pass_store_merging::process_store): Handle BIT_{AND,IOR,XOR}_EXPR
8223         result inverted using BIT_NOT_EXPR, compute bit_not_p, pass it
8224         to store_immediate_info ctor.
8226 2017-11-09  Jim Wilson  <jimw@sifive.com>
8228         * collect2.c (OBJECT_FORMAT_COFF): Remove EXTENDED_COFF support.
8229         (scan_prog_file): Likewise.
8231 2017-11-09  Jan Hubicka  <hubicka@ucw.cz>
8233         * bb-reorder.c (max_entry_frequency): Remove.
8234         (find_traces, rotate_loop, mark_bb_visited, connect_better_edge_p,
8235         connect_traces, push_to_next_round_p): Remove prototypes.
8236         (find_traces_1_round): Use counts only.
8237         (push_to_next_round_p): Likewise.
8238         (find_traces): Likewise.
8239         (rotate_loop): Likewise.
8240         (find_traces_1_round): Likewise.
8241         (connect_traces): Likewise.
8242         (edge_order): Likewise.
8244 2017-11-09  Thomas Preud'homme  <thomas.preudhomme@arm.com>
8246         * config/arm/arm.c (output_return_instruction): Add comments to
8247         indicate requirement for cmse_nonsecure_entry return to account
8248         for the size of clearing instruction output here.
8249         (thumb_exit): Likewise.
8250         * config/arm/thumb2.md (thumb2_cmse_entry_return): Fix length for
8251         return in hardfloat mode.
8253 2017-11-09  Segher Boessenkool  <segher@kernel.crashing.org>
8255         * config/rs6000/rs6000.c (machine_function): Add a bool,
8256         "toc_is_wrapped_separately".
8257         (rs6000_option_override_internal): Enable OPTION_MASK_SAVE_TOC_INDIRECT
8258         if it wasn't explicitly set or unset, we are optimizing for speed, and
8259         doing separate shrink-wrapping.
8260         (rs6000_get_separate_components): Enable the TOC component if
8261         saving the TOC register in the prologue.
8262         (rs6000_components_for_bb): Handle the TOC component.
8263         (rs6000_emit_prologue_components): Store the TOC register where needed.
8264         (rs6000_set_handled_components): Mark TOC as handled, if handled.
8265         (rs6000_emit_prologue): Don't save the TOC if that is already done.
8267 2017-11-09  Martin Jambor  <mjambor@suse.cz>
8269         * ipa-param-manipulation.c: New file.
8270         * ipa-param-manipulation.h: Likewise.
8271         * Makefile.in (OBJS): Add ipa-param-manipulation.o.
8272         (PLUGIN_HEADERS): Addded ipa-param-manipulation.h
8273         * ipa-param.h (ipa_parm_op): Moved to ipa-param-manipulation.h.
8274         (ipa_parm_adjustment): Likewise.
8275         (ipa_parm_adjustment_vec): Likewise.
8276         (ipa_get_vector_of_formal_parms): Moved declaration to
8277         ipa-param-manipulation.h.
8278         (ipa_get_vector_of_formal_parm_types): Likewise.
8279         (ipa_modify_formal_parameters): Likewise.
8280         (ipa_modify_call_arguments): Likewise.
8281         (ipa_combine_adjustments): Likewise.
8282         (ipa_dump_param_adjustments): Likewise.
8283         (ipa_modify_expr): Likewise.
8284         (ipa_get_adjustment_candidate): Likewise.
8285         * ipa-prop.c (ipa_get_vector_of_formal_parms): Moved to
8286         ipa-param-manipulation.c.
8287         (ipa_get_vector_of_formal_parm_types): Likewise.
8288         (ipa_modify_formal_parameters): Likewise.
8289         (ipa_modify_call_arguments): Likewise.
8290         (ipa_modify_expr): Likewise.
8291         (get_ssa_base_param): Likewise.
8292         (ipa_get_adjustment_candidate): Likewise.
8293         (index_in_adjustments_multiple_times_p): Likewise.
8294         (ipa_combine_adjustments): Likewise.
8295         (ipa_dump_param_adjustments): Likewise.
8296         * tree-sra.c: Also include ipa-param-manipulation.h
8297         * omp-simd-clone.c: Include ipa-param-manipulation.h instead of
8298         ipa-param.h.
8300 2017-11-09  Richard Sandiford  <richard.sandiford@linaro.org>
8301             Alan Hayward  <alan.hayward@arm.com>
8302             David Sherwood  <david.sherwood@arm.com>
8304         * doc/sourcebuild.texi (vect_masked_store): Document.
8306 2017-11-09  Richard Sandiford  <richard.sandiford@linaro.org>
8307             Alan Hayward  <alan.hayward@arm.com>
8308             David Sherwood  <david.sherwood@arm.com>
8310         * doc/sourcebuild.texi (vect_align_stack_vars): Document.
8312 2017-11-09  Richard Sandiford  <richard.sandiford@linaro.org>
8313             Alan Hayward  <alan.hayward@arm.com>
8314             David Sherwood  <david.sherwood@arm.com>
8316         * doc/sourcebuild.texi (vect_variable_length): Document.
8318 2017-11-09  Richard Sandiford  <richard.sandiford@linaro.org>
8319             Alan Hayward  <alan.hayward@arm.com>
8320             David Sherwood  <david.sherwood@arm.com>
8322         * doc/sourcebuild.texi (vect_unaligned_possible): Document.
8324 2017-11-09  Richard Sandiford  <richard.sandiford@linaro.org>
8325             Alan Hayward  <alan.hayward@arm.com>
8326             David Sherwood  <david.sherwood@arm.com>
8328         * doc/sourcebuild.texi (vect_element_align_preferred): Document.
8330 2017-11-09  Richard Sandiford  <richard.sandiford@linaro.org>
8331             Alan Hayward  <alan.hayward@arm.com>
8332             David Sherwood  <david.sherwood@arm.com>
8334         * doc/sourcebuild.texi (vect_perm_short, vect_perm_byte): Document
8335         previously undocumented selectors.
8336         (vect_perm3_byte, vect_perm3_short, vect_perm3_int): Document.
8338 2017-11-09  Richard Sandiford  <richard.sandiford@linaro.org>
8340         * doc/rtl.texi (const_vector): Say that elements can be
8341         const_wide_ints too.
8342         * emit-rtl.h (valid_for_const_vec_duplicate_p): Declare.
8343         * emit-rtl.c (valid_for_const_vec_duplicate_p): New function.
8344         (gen_vec_duplicate): Use it instead of CONSTANT_P.
8345         * optabs.c (expand_vector_broadcast): Likewise.
8347 2017-11-09  Richard Sandiford  <richard.sandiford@linaro.org>
8348             Alan Hayward  <alan.hayward@arm.com>
8349             David Sherwood  <david.sherwood@arm.com>
8351         * tree-ssa-loop-ivopts.c (get_address_cost): Try using a
8352         scaled index even if the unscaled address was invalid.
8353         Don't increase the complexity of using a scale in that case.
8355 2017-11-09  Richard Sandiford  <richard.sandiford@linaro.org>
8356             Alan Hayward  <alan.hayward@arm.com>
8357             David Sherwood  <david.sherwood@arm.com>
8359         * doc/rtl.texi: Rewrite the subreg rules so that they partition
8360         the inner register into REGMODE_NATURAL_SIZE bytes rather than
8361         UNITS_PER_WORD bytes.
8362         * emit-rtl.c (validate_subreg): Divide subregs into blocks
8363         based on REGMODE_NATURAL_SIZE of the inner mode.
8364         (gen_lowpart_common): Split the SCALAR_FLOAT_MODE_P and
8365         !SCALAR_FLOAT_MODE_P cases.  Use REGMODE_NATURAL_SIZE for the latter.
8366         * expmed.c (lowpart_bit_field_p): Divide the value up into
8367         chunks of REGMODE_NATURAL_SIZE rather than UNITS_PER_WORD.
8368         * expr.c (store_constructor): Use REGMODE_NATURAL_SIZE to test
8369         whether something is likely to occupy more than one register.
8371 2017-11-09  Jan Hubicka  <hubicka@ucw.cz>
8373         PR ipa/82879
8374         * ipa-inline-transform.c (update_noncloned_frequencies): Use
8375         profile_count::adjust_for_ipa_scaling.
8376         * tree-inline.c (copy_bb, copy_cfg_body): Likewise.
8377         * profile-count.c (profile_count::adjust_for_ipa_scaling): New member
8378         function.
8379         * profile-count.h (profile_count::adjust_for_ipa_scaling): Declare.
8381 2017-11-09  Jakub Jelinek  <jakub@redhat.com>
8383         * gimple-ssa-store-merging.c (count_multiple_uses): New function.
8384         (split_group): Add total_orig and total_new arguments, estimate the
8385         number of statements related to the store group without store merging
8386         and with store merging.
8387         (imm_store_chain_info::output_merged_store): Adjust split_group
8388         callers, punt if estimated number of statements with store merging
8389         is not smaller than estimated number of statements without it.
8390         Formatting fix.
8391         (handled_load): Remove has_single_use checks.
8392         (pass_store_merging::process_store): Likewise.
8394 2017-11-09  Richard Biener  <rguenther@suse.de>
8396         PR tree-optimization/82902
8397         * tree-ssa-phiprop.c (propagate_with_phi): Test proper type.
8399 2017-11-09  Martin Liska  <mliska@suse.cz>
8401         PR target/82863
8402         * emit-rtl.c (init_emit_regs): Initialize split_branch_probability to
8403         uninitialized.
8405 2017-11-09  Martin Liska  <mliska@suse.cz>
8407         PR tree-optimization/82669
8408         * sbitmap.h (bmp_iter_set_init): Remove non needed check.
8410 2017-11-09  Martin Liska  <mliska@suse.cz>
8412         PR gcov-profile/48463
8413         * coverage.c (coverage_begin_function): Output also end locus
8414         of a function and information whether the function is
8415         artificial.
8416         * gcov-dump.c (tag_function): Parse and print the information.
8417         * gcov.c (INCLUDE_MAP): Add include.
8418         (INCLUDE_SET): Likewise.
8419         (struct line_info): Move earlier in the source file because
8420         of vector<line_info> in function_info structure.
8421         (line_info::line_info): Likewise.
8422         (line_info::has_block): Likewise.
8423         (struct source_info): Add new member index.
8424         (source_info::get_functions_at_location): New function.
8425         (function_info::group_line_p): New function.
8426         (output_intermediate_line): New function.
8427         (output_intermediate_file): Use the mentioned function.
8428         (struct function_start): New.
8429         (struct function_start_pair_hash): Likewise.
8430         (process_file): Add code that identifies group functions.
8431         Assign lines either to global or function scope.
8432         (generate_results): Skip artificial functions.
8433         (find_source): Assign index for each source file.
8434         (read_graph_file): Read new flag artificial and end_line.
8435         (add_line_counts): Assign it either to global of function scope.
8436         (accumulate_line_counts): Isolate core of the function to
8437         accumulate_line_info and call it for both function and global
8438         scope lines.
8439         (accumulate_line_info): New function.
8440         (output_line_beginning): Fix GNU coding style.
8441         (print_source_line): New function.
8442         (output_line_details): Likewise.
8443         (output_function_details): Likewise.
8444         (output_lines): Iterate both source (global) scope and function
8445         scope.
8446         (struct function_line_start_cmp): New class.
8447         * doc/gcov.texi: Reflect changes in documentation.
8449 2017-11-09  Jakub Jelinek  <jakub@redhat.com>
8451         PR debug/82837
8452         * dwarf2out.c (const_ok_for_output_1): Reject NEG in addition to NOT.
8453         (mem_loc_descriptor): Handle (const (neg (...))) as (neg (const (...)))
8454         and similarly for not instead of neg.
8456 2017-11-08  Andi Kleen  <ak@linux.intel.com>
8458         * config/i386/i386.opt: Add -mforce-indirect-call.
8459         * config/i386/predicates.md: Check for flag_force_indirect_call.
8460         * doc/invoke.texi: Document -mforce-indirect-call
8462 2017-11-08  Kito Cheng  <kito.cheng@gmail.com>
8464         * config/riscv/riscv-protos.h (riscv_slow_unaligned_access_p):
8465         New extern.
8466         (MOVE_RATIO): Use riscv_slow_unaligned_access_p.
8467         config/riscv/riscv.c (predict.h): New include.
8468         (riscv_slow_unaligned_access_p): No longer static.
8469         (riscv_block_move_straight): Add require.
8470         * config/riscv/riscv-protos.h (riscv_hard_regno_nregs): Delete.
8472 2017-11-08  Jakub Jelinek  <jakub@redhat.com>
8474         PR target/82855
8475         * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>,
8476         <avx512>_eq<mode>3<mask_scalar_merge_name>_1): Use
8477         nonimmediate_operand predicate for operand 1 instead of
8478         register_operand.
8480 2017-11-08  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
8482         * config/aarch64/aarch64-simd.md (store_pair_lanes<mode>):
8483         New pattern.
8484         * config/aarch64/constraints.md (Uml): New constraint.
8485         * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): New
8486         predicate.
8488 2017-11-08  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
8490         * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
8491         of two vec_duplicates into a vec_concat.
8493 2017-11-08  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
8495         * simplify-rtx.c (simplify_ternary_operation, VEC_MERGE):
8496         Simplify vec_merge of vec_duplicate and vec_concat.
8497         * config/aarch64/constraints.md (Utq): New constraint.
8498         * config/aarch64/aarch64-simd.md (load_pair_lanes<mode>): New
8499         define_insn.
8501 2017-11-08  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
8503         * simplify-rtx.c (simplify_ternary_operation, VEC_MERGE):
8504         Simplify vec_merge of vec_duplicate and const_vector.
8505         * config/aarch64/predicates.md (aarch64_simd_or_scalar_imm_zero):
8506         New predicate.
8507         * config/aarch64/aarch64-simd.md (*aarch64_combinez<mode>): Use VDC
8508         mode iterator.  Update predicate on operand 1 to
8509         handle non-const_vec constants.  Delete constraints.
8510         (*aarch64_combinez_be<mode>): Likewise for operand 2.
8512 2017-11-08  Jakub Jelinek  <jakub@redhat.com>
8514         PR tree-optimization/78821
8515         * gimple-ssa-store-merging.c (struct store_operand_info): Add bit_not_p
8516         data member.
8517         (store_operand_info::store_operand_info): Initialize it to false.
8518         (pass_store_merging::terminate_all_aliasing_chains): Rewritten to use
8519         ref_maybe_used_by_stmt_p and stmt_may_clobber_ref_p on lhs of each
8520         store in the group, and if chain_info is non-NULL, to ignore altogether
8521         that chain.
8522         (compatible_load_p): Fail if bit_not_p does not match.
8523         (imm_store_chain_info::output_merged_store): Handle bit_not_p loads.
8524         (handled_load): Fill in bit_not_p.  Handle BIT_NOT_EXPR.
8525         (pass_store_merging::process_store): Adjust
8526         terminate_all_aliasing_chains calls to pass NULL in all current spots,
8527         call terminate_all_aliasing_chains newly when adding a store into
8528         a chain with non-NULL chain_info.
8530 2017-11-08  Wilco Dijkstra  <wdijkstr@arm.com>
8532         * config/aarch64/aarch64.c (aarch64_can_eliminate): Simplify logic.
8534 2017-11-08  Wilco Dijkstra  <wdijkstr@arm.com>
8536         * config/aarch64/aarch64.c (aarch64_frame_pointer_required)
8537         Remove.
8538         (aarch64_layout_frame): Initialise emit_frame_chain.
8539         (aarch64_can_eliminate): Remove omit leaf frame pointer code.
8540         (TARGET_FRAME_POINTER_REQUIRED): Remove define.
8542 2017-11-08  Martin Liska  <mliska@suse.cz>
8544         * gimplify.c (expand_FALLTHROUGH_r): Simplify usage
8545         of gimple_call_internal_p.
8547 2017-11-07  Tom de Vries  <tom@codesourcery.com>
8549         * config/mips/mips.h (ASM_OUTPUT_LABELREF): Wrap in "do {} while (0)".
8551 2017-11-07  Tom de Vries  <tom@codesourcery.com>
8553         * config/mips/mips.h (ASM_OUTPUT_CASE_END): Remove semicolon after
8554         "do {} while (0)".
8556 2017-11-08  Martin Liska  <mliska@suse.cz>
8558         PR sanitizer/82792
8559         * gimplify.c (expand_FALLTHROUGH_r): Skip IFN_ASAN_MARK.
8561 2017-11-07  Eric Botcazou  <ebotcazou@adacore.com>
8563         * gimple-pretty-print.c (dump_profile): Return "" instead of NULL.
8565 2017-11-07  Jakub Jelinek  <jakub@redhat.com>
8567         PR target/82855
8568         * config/i386/i386.md (SWI1248_AVX512BWDQ2_64): New mode iterator.
8569         (*cmp<mode>_ccz_1): New insn with $k alternative.
8571         PR target/82855
8572         * config/i386/i386.c (ix86_swap_binary_operands_p): Treat
8573         RTX_COMM_COMPARE as commutative as well.
8574         (ix86_binary_operator_ok): Formatting fix.
8575         * config/i386/sse.md (*mul<mode>3<mask_name><round_name>,
8576         *<code><mode>3<mask_name><round_saeonly_name>,
8577         *<code><mode>3<mask_name>, *<code>tf3, *mul<mode>3<mask_name>,
8578         *<s>mul<mode>3_highpart<mask_name>,
8579         *vec_widen_umult_even_v16si<mask_name>,
8580         *vec_widen_umult_even_v8si<mask_name>,
8581         *vec_widen_umult_even_v4si<mask_name>,
8582         *vec_widen_smult_even_v16si<mask_name>,
8583         *vec_widen_smult_even_v8si<mask_name>, *sse4_1_mulv2siv2di3<mask_name>,
8584         *avx2_pmaddwd, *sse2_pmaddwd, *<sse4_1_avx2>_mul<mode>3<mask_name>,
8585         *avx2_<code><mode>3, *avx512f_<code><mode>3<mask_name>,
8586         *sse4_1_<code><mode>3<mask_name>, *<code>v8hi3,
8587         *sse4_1_<code><mode>3<mask_name>, *<code>v16qi3, *avx2_eq<mode>3,
8588         <avx512>_eq<mode>3<mask_scalar_merge_name>_1, *sse4_1_eqv2di3,
8589         *sse2_eq<mode>3, <mask_codefor><code><mode>3<mask_name>,
8590         *<code><mode>3, *<sse2_avx2>_uavg<mode>3<mask_name>,
8591         *<ssse3_avx2>_pmulhrsw<mode>3<mask_name>, *ssse3_pmulhrswv4hi3): Use
8592         !(MEM_P (operands[1]) && MEM_P (operands[2])) condition instead of
8593         ix86_binary_operator_ok.  Formatting fixes.
8594         (*<plusminus_insn><mode>3<mask_name><round_name>,
8595         *<plusminus_insn><mode>3, *<plusminus_insn><mode>3_m): Formatting
8596         fixes.
8598 2017-11-07  Segher Boessenkool  <segher@kernel.crashing.org>
8600         * config/rs6000/rs6000.md (GPR2): New mode_iterator.
8601         ("cstore<mode>4"): Don't always expand with rs6000_emit_int_cmove for
8602         eq and ne if TARGET_ISEL.
8603         (cmp): New code_iterator.
8604         (UNS, UNSU_, UNSIK): New code_attrs.
8605         (<code><GPR:mode><GPR2:mode>2_isel): New define_insn_and_split.
8606         ("eq<mode>3"): New define_expand, rename the define_insn_and_split
8607         to...
8608         ("eq<mode>3"): ... this.
8609         ("ne<mode>3"): New define_expand, rename the define_insn_and_split
8610         to...
8611         ("ne<mode>3"): ... this.
8613 2017-11-07  Julia Koval  <julia.koval@intel.com>
8615         PR target/82812
8616         * common/config/i386/i386-common.c
8617         (OPTION_MASK_ISA_GENERAL_REGS_ONLY_UNSET): Remove MPX from flag.
8618         (ix86_handle_option): Move MPX to isa_flags2 and GFNI to isa_flags.
8619         * config/i386/i386-c.c (ix86_target_macros_internal): Ditto.
8620         * config/i386/i386.opt: Ditto.
8621         * config/i386/i386.c (ix86_target_string): Ditto.
8622         (ix86_option_override_internal): Ditto.
8623         (ix86_init_mpx_builtins): Move MPX to args2.
8624         (ix86_expand_builtin): Special handling for OPTION_MASK_ISA_GFNI.
8625         * config/i386/i386-builtin.def (__builtin_ia32_vgf2p8affineinvqb_v64qi,
8626         __builtin_ia32_vgf2p8affineinvqb_v64qi_mask,
8627         __builtin_ia32_vgf2p8affineinvqb_v32qi,
8628         __builtin_ia32_vgf2p8affineinvqb_v32qi_mask,
8629         __builtin_ia32_vgf2p8affineinvqb_v16qi,
8630         __builtin_ia32_vgf2p8affineinvqb_v16qi_mask): Move to ARGS array.
8632 2017-11-07  Uros Bizjak  <ubizjak@gmail.com>
8634         PR target/80425
8635         * config/i386.i386.md (*zero_extendsidi2): Change (?r,*Yj), (?*Yi,r)
8636         and (*x,m) to ($r,Yj), ($Yi,r) and ($x,m).
8637         (zero-extendsidi peephole2): Remove peephole.
8639 2017-11-07  Eric Botcazou  <ebotcazou@adacore.com>
8641         PR c/53037
8642         * stor-layout.c: Include attribs.h.
8643         (handle_warn_if_not_align): Replace test on TYPE_USER_ALIGN with
8644         explicit lookup of "aligned" attribute.
8646 2017-11-07  Andrew Waterman  <andrew@sifive.com>
8648         * config/riscv/riscv-protos.h (riscv_hard_regno_nregs): New prototype.
8649         (riscv_expand_block_move): Likewise.
8650         * config/riscv/riscv.h (MOVE_RATIO): Tune cost to movmemsi
8651         implementation.
8652         (RISCV_MAX_MOVE_BYTES_PER_LOOP_ITER): New define.
8653         (RISCV_MAX_MOVE_BYTES_STRAIGHT): New define.
8654         * config/riscv/riscv.c (riscv_block_move_straight): New function.
8655         (riscv_adjust_block_mem): Likewise.
8656         (riscv_block_move_loop): Likewise.
8657         (riscv_expand_block_move): Likewise.
8658         * config/riscv/riscv.md (movmemsi): New pattern.
8660 2017-11-07  Michael Clark  <michaeljclark@mac.com>
8662         * config/riscv/linux.h (MUSL_ABI_SUFFIX): New define.
8663         (MUSL_DYNAMIC_LINKER): Likewise.
8665 2017-11-07  Richard Sandiford  <richard.sandiford@linaro.org>
8667         * config/aarch64/aarch64.md (ashl<mode>3, ashr<mode>3, lshr<mode>3)
8668         (rotr<mode>3, rotl<mode>3): Use aarch64_reg_or_imm instead of
8669         nonmmory_operand.
8671 2017-11-07  Richard Biener  <rguenther@suse.de>
8673         * match.pd: Fix build.
8675 2017-11-07  Wilco Dijkstra  <wdijkstr@arm.com>
8676             Jackson Woodruff  <jackson.woodruff@arm.com>
8678         PR tree-optimization/71026
8679         * match.pd: Canonicalize negate in division.
8681 2017-11-07  Sudakshina Das  <sudi.das@arm.com>
8683         PR middle-end/80131
8684         * match.pd: Simplify 1 << (C - x) where C = precision (x) - 1.
8686 2017-11-07  Marc Glisse  <marc.glisse@inria.fr>
8688         * match.pd ((a&~b)|(a^b),(a&~b)^~a,(a|b)&~(a^b),a|~(a^b),
8689         (a|b)|(a&^b),(a&b)|~(a^b),~(~a&b),~X^Y): New transformations.
8691 2017-11-07  Marc Glisse  <marc.glisse@inria.fr>
8693         * fold-const.c (negate_expr_p) [PLUS_EXPR, MINUS_EXPR]: Handle
8694         non-scalar integral types.
8695         * match.pd (negate_expr_p): Handle MINUS_EXPR.
8696         (-(A-B), -(~A)): New transformations.
8698 2017-11-07  Tom de Vries  <tom@codesourcery.com>
8700         * config/powerpcspe/aix43.h (SUBTARGET_OVERRIDE_OPTIONS): Remove
8701         semicolon after "do {} while (0)".
8702         * config/powerpcspe/aix51.h (SUBTARGET_OVERRIDE_OPTIONS): Same.
8703         * config/powerpcspe/aix52.h (SUBTARGET_OVERRIDE_OPTIONS): Same.
8704         * config/powerpcspe/aix53.h (SUBTARGET_OVERRIDE_OPTIONS): Same.
8705         * config/powerpcspe/aix61.h (SUBTARGET_OVERRIDE_OPTIONS): Same.
8706         * config/powerpcspe/aix71.h (SUBTARGET_OVERRIDE_OPTIONS): Same.
8708 2017-11-07  Tom de Vries  <tom@codesourcery.com>
8710         * config/rs6000/aix43.h (SUBTARGET_OVERRIDE_OPTIONS): Remove semicolon
8711         after "do {} while (0)".
8712         * config/rs6000/aix51.h (SUBTARGET_OVERRIDE_OPTIONS): Same.
8713         * config/rs6000/aix52.h (SUBTARGET_OVERRIDE_OPTIONS): Same.
8714         * config/rs6000/aix53.h (SUBTARGET_OVERRIDE_OPTIONS): Same.
8715         * config/rs6000/aix61.h (SUBTARGET_OVERRIDE_OPTIONS): Same.
8716         * config/rs6000/aix71.h (SUBTARGET_OVERRIDE_OPTIONS): Same.
8718 2017-11-07  Tom de Vries  <tom@codesourcery.com>
8720         PR other/82784
8721         * config/arm/arm.c (HANDLE_NARROW_SHIFT_ARITH): Remove semicolon after
8722         "while {} do (0)".
8723         (arm_rtx_costs_internal): Add missing semicolon after
8724         HANDLE_NARROW_SHIFT_ARITH call.
8726 2017-11-06  Segher Boessenkool  <segher@kernel.crashing.org>
8728         * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
8729         disable isel if it was not set explicitly.
8731 2017-11-06  James Bowman  <james.bowman@ftdichip.com>
8733         * gcc/dwarf2out.c (modified_type_die): Retain ADDR_SPACE qualifiers.
8734         (add_type_attribute) likewise.
8736 2017-11-06  H.J. Lu  <hongjiu.lu@intel.com>
8738         * config/i386/i386.c (ix86_can_use_return_insn_p): Use reference
8739         of struct ix86_frame.
8740         (ix86_initial_elimination_offset): Likewise.
8741         (ix86_expand_split_stack_prologue): Likewise.
8743 2017-11-06  Marc Glisse  <marc.glisse@inria.fr>
8745         * tree-vrp.h (enum value_range_type): Update stale comment.
8747 2017-11-06  Richard Sandiford  <richard.sandiford@linaro.org>
8748             Alan Hayward  <alan.hayward@arm.com>
8749             David Sherwood  <david.sherwood@arm.com>
8751         * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm)
8752         (aarch64_expand_vec_perm_const): Take the number of units too.
8753         * config/aarch64/aarch64.c (aarch64_expand_vec_perm)
8754         (aarch64_expand_vec_perm_const): Likewise.
8755         * config/aarch64/aarch64-simd.md (vec_perm_const<mode>)
8756         (vec_perm<mode>): Update accordingly.
8758 2017-11-06  Richard Sandiford  <richard.sandiford@linaro.org>
8759             Alan Hayward  <alan.hayward@arm.com>
8760             David Sherwood  <david.sherwood@arm.com>
8762         * config/aarch64/aarch64-protos.h (aarch64_simd_vect_par_cnst_half):
8763         Take the number of units too.
8764         * config/aarch64/aarch64.c (aarch64_simd_vect_par_cnst_half): Likewise.
8765         (aarch64_simd_check_vect_par_cnst_half): Update call accordingly,
8766         but check for a vector mode before rather than after the call.
8767         * config/aarch64/aarch64-simd.md (aarch64_split_simd_mov<mode>)
8768         (move_hi_quad_<mode>, vec_unpack<su>_hi_<mode>)
8769         (vec_unpack<su>_lo_<mode, vec_widen_<su>mult_lo_<mode>)
8770         (vec_widen_<su>mult_hi_<mode>, vec_unpacks_lo_<mode>)
8771         (vec_unpacks_hi_<mode>, aarch64_saddl2<mode>, aarch64_uaddl2<mode>)
8772         (aarch64_ssubl2<mode>, aarch64_usubl2<mode>, widen_ssum<mode>3)
8773         (widen_usum<mode>3, aarch64_saddw2<mode>, aarch64_uaddw2<mode>)
8774         (aarch64_ssubw2<mode>, aarch64_usubw2<mode>, aarch64_sqdmlal2<mode>)
8775         (aarch64_sqdmlsl2<mode>, aarch64_sqdmlal2_lane<mode>)
8776         (aarch64_sqdmlal2_laneq<mode>, aarch64_sqdmlsl2_lane<mode>)
8777         (aarch64_sqdmlsl2_laneq<mode>, aarch64_sqdmlal2_n<mode>)
8778         (aarch64_sqdmlsl2_n<mode>, aarch64_sqdmull2<mode>)
8779         (aarch64_sqdmull2_lane<mode>, aarch64_sqdmull2_laneq<mode>)
8780         (aarch64_sqdmull2_n<mode>): Update accordingly.
8782 2017-11-06  Richard Sandiford  <richard.sandiford@linaro.org>
8783             Alan Hayward  <alan.hayward@arm.com>
8784             David Sherwood  <david.sherwood@arm.com>
8786         * config/aarch64/aarch64-protos.h (aarch64_reverse_mask): Take
8787         the number of units too.
8788         * config/aarch64/aarch64.c (aarch64_reverse_mask): Likewise.
8789         * config/aarch64/aarch64-simd.md (vec_load_lanesoi<mode>)
8790         (vec_store_lanesoi<mode>, vec_load_lanesci<mode>)
8791         (vec_store_lanesci<mode>, vec_load_lanesxi<mode>)
8792         (vec_store_lanesxi<mode>): Update accordingly.
8794 2017-11-06  Richard Sandiford  <richard.sandiford@linaro.org>
8795             Alan Hayward  <alan.hayward@arm.com>
8796             David Sherwood  <david.sherwood@arm.com>
8798         * config/aarch64/aarch64-protos.h (aarch64_endian_lane_rtx): Declare.
8799         * config/aarch64/aarch64.c (aarch64_endian_lane_rtx): New function.
8800         * config/aarch64/aarch64.h (ENDIAN_LANE_N): Take the number
8801         of units rather than the mode.
8802         * config/aarch64/iterators.md (nunits): New mode attribute.
8803         * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args):
8804         Use aarch64_endian_lane_rtx instead of GEN_INT (ENDIAN_LANE_N ...).
8805         * config/aarch64/aarch64-simd.md (aarch64_dup_lane<mode>)
8806         (aarch64_dup_lane_<vswap_width_name><mode>, *aarch64_mul3_elt<mode>)
8807         (*aarch64_mul3_elt_<vswap_width_name><mode>): Likewise.
8808         (*aarch64_mul3_elt_to_64v2df, *aarch64_mla_elt<mode>): Likewise.
8809         (*aarch64_mla_elt_<vswap_width_name><mode>, *aarch64_mls_elt<mode>)
8810         (*aarch64_mls_elt_<vswap_width_name><mode>, *aarch64_fma4_elt<mode>)
8811         (*aarch64_fma4_elt_<vswap_width_name><mode>):: Likewise.
8812         (*aarch64_fma4_elt_to_64v2df, *aarch64_fnma4_elt<mode>): Likewise.
8813         (*aarch64_fnma4_elt_<vswap_width_name><mode>): Likewise.
8814         (*aarch64_fnma4_elt_to_64v2df, reduc_plus_scal_<mode>): Likewise.
8815         (reduc_plus_scal_v4sf, reduc_<maxmin_uns>_scal_<mode>): Likewise.
8816         (reduc_<maxmin_uns>_scal_<mode>): Likewise.
8817         (*aarch64_get_lane_extend<GPI:mode><VDQQH:mode>): Likewise.
8818         (*aarch64_get_lane_zero_extendsi<mode>): Likewise.
8819         (aarch64_get_lane<mode>, *aarch64_mulx_elt_<vswap_width_name><mode>)
8820         (*aarch64_mulx_elt<mode>, *aarch64_vgetfmulx<mode>): Likewise.
8821         (aarch64_sq<r>dmulh_lane<mode>, aarch64_sq<r>dmulh_laneq<mode>)
8822         (aarch64_sqrdml<SQRDMLH_AS:rdma_as>h_lane<mode>): Likewise.
8823         (aarch64_sqrdml<SQRDMLH_AS:rdma_as>h_laneq<mode>): Likewise.
8824         (aarch64_sqdml<SBINQOPS:as>l_lane<mode>): Likewise.
8825         (aarch64_sqdml<SBINQOPS:as>l_laneq<mode>): Likewise.
8826         (aarch64_sqdml<SBINQOPS:as>l2_lane<mode>_internal): Likewise.
8827         (aarch64_sqdml<SBINQOPS:as>l2_laneq<mode>_internal): Likewise.
8828         (aarch64_sqdmull_lane<mode>, aarch64_sqdmull_laneq<mode>): Likewise.
8829         (aarch64_sqdmull2_lane<mode>_internal): Likewise.
8830         (aarch64_sqdmull2_laneq<mode>_internal): Likewise.
8831         (aarch64_vec_load_lanesoi_lane<mode>): Likewise.
8832         (aarch64_vec_store_lanesoi_lane<mode>): Likewise.
8833         (aarch64_vec_load_lanesci_lane<mode>): Likewise.
8834         (aarch64_vec_store_lanesci_lane<mode>): Likewise.
8835         (aarch64_vec_load_lanesxi_lane<mode>): Likewise.
8836         (aarch64_vec_store_lanesxi_lane<mode>): Likewise.
8837         (aarch64_simd_vec_set<mode>): Update use of ENDIAN_LANE_N.
8838         (aarch64_simd_vec_setv2di): Likewise.
8840 2017-11-06  Carl Love  <cel@us.ibm.com>
8842         * config/rs6000/rs6000-c.c (P8V_BUILTIN_VEC_REVB): Add power 8
8843         definitions.
8844         (P9V_BUILTIN_VEC_REVB): Remove the power 9 instance definitions.
8845         * config/rs6000/altivec.h (vec_revb): Change the #define from power 9
8846         to power 8.
8847         * config/rs6000/r6000-protos.h (swap_endian_selector_for_mode): Add new
8848         extern declaration.
8849         * config/rs6000/rs6000.c (swap_endian_selector_for_mode): Add function.
8850         * config/rs6000/rs6000-builtin.def (BU_P8V_VSX_1, BU_P8V_OVERLOAD_1):
8851         Add power 8 macro expansions.
8852         (BU_P9V_OVERLOAD_1): Remove power 9 overload expansion.
8853         * config/rs6000/vsx.md (revb_<mode>): Add define_expand to generate
8854         power 8 instructions.  (VSX_XXBR): Add iterator.
8856 2017-11-06  Wilco Dijkstra  <wdijkstr@arm.com>
8858         * config/arm/arm.md (predicable_short_it): Change default to "no",
8859         improve documentation, remove uses that are identical to the default.
8860         (enabled_for_depr_it): Rename to enabled_for_short_it.
8861         * gcc/config/arm/arm-fixed.md (predicable_short_it):
8862         Remove default uses.
8863         * gcc/config/arm/ldmstm.md (predicable_short_it): Likewise.
8864         * gcc/config/arm/sync.md (predicable_short_it): Likewise.
8865         * gcc/config/arm/thumb2.md (predicable_short_it): Likewise.
8866         * gcc/config/arm/vfp.md (predicable_short_it): Likewise.
8868 2017-11-06  Michael Meissner  <meissner@linux.vnet.ibm.com>
8870         PR target/82748
8871         * config/rs6000/rs6000-builtin.def (BU_FLOAT128_1): Delete
8872         float128 helper macros, which are no longer used after deleting
8873         the old 'q' built-in functions, and moving the round to odd
8874         built-in functions to being special built-in functions.
8875         (BU_FLOAT128_2): Likewise.
8876         (BU_FLOAT128_1_HW): Likewise.
8877         (BU_FLOAT128_2_HW): Likewise.
8878         (BU_FLOAT128_3_HW): Likewise.
8879         (FABSQ): Delete old 'q' built-in functions.
8880         (COPYSIGNQ): Likewise.
8881         (SQRTF128_ODD): Move round to odd built-in functions to be
8882         special built-in functions, so that we can handle
8883         -mabi=ieeelongdouble.
8884         (TRUNCF128_ODD): Likewise.
8885         (ADDF128_ODD): Likewise.
8886         (SUBF128_ODD): Likewise.
8887         (MULF128_ODD): Likewise.
8888         (DIVF128_ODD): Likewise.
8889         (FMAF128_ODD): Likewise.
8890         * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Map old 'q'
8891         built-in names to 'f128'.
8892         * config/rs6000/rs6000.c (rs6000_fold_builtin): Remove folding the
8893         old 'q' built-in functions, as the machine independent code for
8894         'f128' built-in functions handles this.
8895         (rs6000_expand_builtin): Add expansion for float128 round to odd
8896         functions, keying off on -mabi=ieeelongdouble of whether to use
8897         the KFmode or TFmode variant.
8898         (rs6000_init_builtins): Initialize the _Float128 round to odd
8899         built-in functions.
8900         * doc/extend.texi (PowerPC Built-in Functions): Document the old
8901         _Float128 'q' built-in functions are now mapped into the new
8902         'f128' built-in functions.
8904 2017-11-06  David Edelsohn  <dje.gcc@gmail.com>
8906         * collect2.c (add_lto_object): Compile for OBJECT_COFF.
8907         (scan_prog_file): Don't skip PASS_LTOINFO. Scan for LTO objects.
8909 2017-11-06  David Malcolm  <dmalcolm@redhat.com>
8911         PR jit/82826
8912         * ipa-fnsummary.c (ipa_fnsummary_c_finalize): New function.
8913         * ipa-fnsummary.h (ipa_fnsummary_c_finalize): New decl.
8914         * toplev.c: Include "ipa-fnsummary.h".
8915         (toplev::finalize): Call ipa_fnsummary_c_finalize.
8917 2017-11-06  Jakub Jelinek  <jakub@redhat.com>
8919         PR tree-optimization/82838
8920         * gimple-ssa-store-merging.c
8921         (imm_store_chain_info::output_merged_store): Call force_gimple_operand_1
8922         on a separate gimple_seq which is then appended to seq.
8924 2017-11-06  Jeff Law  <law@redhat.com>
8926         PR target/82788
8927         * config/i386/i386.c (PROBE_INTERVAL): Remove.
8928         (get_probe_interval): New functions.
8929         (ix86_adjust_stack_and_probe_stack_clash): Use get_probe_interval.
8930         (ix86_adjust_stack_and_probe): Likewise.
8931         (output_adjust_stack_and_probe): Likewise.
8932         (ix86_emit_probe_stack_range): Likewise.
8933         (ix86_expand_prologue): Likewise.
8935 2017-11-06  Richard Sandiford  <richard.sandiford@linaro.org>
8937         PR tree-optimization/82816
8938         * tree-ssa-math-opts.c (convert_mult_to_widen): Return false
8939         if the modes of the two types are the same.
8940         (convert_plusminus_to_widen): Likewise.
8942 2017-11-06  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
8944         * config/rs6000/altivec.md (*p9_vadu<mode>3) Rename to
8945         p9_vadu<mode>3.
8946         (usadv16qi): New define_expand.
8947         (usadv8hi): New define_expand.
8949 2017-11-06  Jan Hubicka  <hubicka@ucw.cz>
8951         PR bootstrap/82832
8952         * ipa-inline-transform.c (update_noncloned_frequencies): Always
8953         scale.
8954         (inline_transform): Likewise.
8955         * predict.c (counts_to_freqs): Remove useless conditional.
8956         * profile-count.h (profile_count::apply_scale): Move sanity check.
8957         * tree-inline.c (copy_bb): Always scale.
8958         (copy_cfg_body): Likewise.
8960 2017-11-06  Christophe Lyon  <christophe.lyon@linaro.org>
8962         PR target/67591
8963         * config/arm/arm.md (*sub_shiftsi): Add predicable_short_it
8964         attribute.
8965         (*cmp_ite0): Add enabled_for_depr_it attribute.
8966         (*cmp_ite1): Likewise.
8968 2017-11-06  Segher Boessenkool  <segher@kernel.crashing.org>
8970         * config/rs6000/rs6000.c (rs6000_insn_cost): Handle TYPE_MFCR and
8971         TYPE_MFCRF.
8973 2017-11-06  Richard Sandiford  <richard.sandiford@linaro.org>
8975         * tree-vrp.c (vrp_int_const_binop): Return true on success and
8976         return the value by pointer.
8977         (extract_range_from_multiplicative_op_1): Update accordingly.
8978         Return as soon as an operation fails.
8980 2017-11-05  Tom de Vries  <tom@codesourcery.com>
8982         PR other/82784
8983         * asan.c (DEF_SANITIZER_BUILTIN_1): Factor out of ...
8984         (DEF_SANITIZER_BUILTIN): ... here.
8985         (initialize_sanitizer_builtins): Use DEF_SANITIZER_BUILTIN_1 instead of
8986         DEF_SANITIZER_BUILTIN in if stmt.  Add missing semicolon.
8988 2017-11-05  Tom de Vries  <tom@codesourcery.com>
8990         PR other/82784
8991         * config/elfos.h (ASM_OUTPUT_BEFORE_CASE_LABEL): Remove semicolon after
8992         macro body.
8993         (ASM_OUTPUT_CASE_LABEL): Add semicolon after
8994         ASM_OUTPUT_BEFORE_CASE_LABEL call.
8995         * config/arc/arc.h (ASM_OUTPUT_BEFORE_CASE_LABEL): Remove semicolon
8996         after macro body.
8997         * config/m68k/m68kelf.h (ASM_OUTPUT_BEFORE_CASE_LABEL): Same.
8998         * config/mips/mips.h (ASM_OUTPUT_BEFORE_CASE_LABEL): Same.
8999         * config/v850/v850.h (ASM_OUTPUT_BEFORE_CASE_LABEL): Same.
9001 2017-11-05  Tom de Vries  <tom@codesourcery.com>
9003         PR other/82784
9004         * graphite-scop-detection.c (DEBUG_PRINT): Remove semicolon after
9005         "do {} while (0)".
9007 2017-11-04  Michael Clark  <michaeljclark@mac.com>
9009         * config/riscv/riscv.c (riscv_print_operand): Add a 'i' format.
9010         config/riscv/riscv.md (addsi3): Use 'i' for immediates.
9011         (adddi3): Likewise.
9012         (*addsi3_extended): Likewise.
9013         (*addsi3_extended2): Likewise.
9014         (<optab>si3): Likewise.
9015         (<optab>di3): Likewise.
9016         (<optab><mode>3): Likewise.
9017         (<*optabe>si3_internal): Likewise.
9018         (zero_extendqi<SUPERQI:mode>2): Likewise.
9019         (*add<mode>hi3): Likewise.
9020         (*xor<mode>hi3): Likewise.
9021         (<optab>di3): Likewise.
9022         (*<optab>si3_extend): Likewise.
9023         (*sge<u>_<X:mode><GPR:mode>): Likewise.
9024         (*slt<u>_<X:mode><GPR:mode>): Likewise.
9025         (*sle<u>_<X:mode><GPR:mode>): Likewise.
9027 2017-11-04  Andrew Waterman  <andrew@sifive.com>
9029         * config/riscv/riscv.c (riscv_option_override): Conditionally set
9030         TARGET_STRICT_ALIGN based upon -mtune argument.
9032 2017-11-04  Andrew Waterman  <andrew@sifive.com>
9034         * config/riscv/riscv.h (SLOW_BYTE_ACCESS): Change to 1.
9036 2017-11-04  Daniel Santos  <daniel.santos@pobox.com>
9038         * config/i386/i386.c (choose_basereg): Use optional scratch
9039         register and add assertion.
9040         (x86_emit_outlined_ms2sysv_save): Use scratch register when
9041         needed, and don't allocate stack.
9042         (ix86_expand_prologue): Rearrange where SSE saves/stub call is
9043         emitted, correct wrong allocation with -mcall-ms2sysv-xlogues.
9044         (ix86_emit_outlined_ms2sysv_restore): Fix non-immediate offsets.
9046 2017-11-03  Jeff Law  <law@redhat.com>
9048         * config/i386/i386.c (ix86_emit_restore_reg_using_pop): Prototype.
9049         (ix86_adjust_stack_and_probe_stack_clash): Use a push/pop sequence
9050         to probe at the start of a noreturn function.
9052 2017-11-03  Jakub Jelinek  <jakub@redhat.com>
9054         PR tree-optimization/78821
9055         * gimple-ssa-store-merging.c: Update the file comment.
9056         (MAX_STORE_ALIAS_CHECKS): Define.
9057         (struct store_operand_info): New type.
9058         (store_operand_info::store_operand_info): New constructor.
9059         (struct store_immediate_info): Add rhs_code and ops data members.
9060         (store_immediate_info::store_immediate_info): Add rhscode, op0r
9061         and op1r arguments to the ctor, initialize corresponding data members.
9062         (struct merged_store_group): Add load_align_base and load_align
9063         data members.
9064         (merged_store_group::merged_store_group): Initialize them.
9065         (merged_store_group::do_merge): Update them.
9066         (merged_store_group::apply_stores): Pick the constant for
9067         encode_tree_to_bitpos from one of the two operands, or skip
9068         encode_tree_to_bitpos if neither operand is a constant.
9069         (class pass_store_merging): Add process_store method decl.  Remove
9070         bool argument from terminate_all_aliasing_chains method decl.
9071         (pass_store_merging::terminate_all_aliasing_chains): Remove
9072         var_offset_p argument and corresponding handling.
9073         (stmts_may_clobber_ref_p): New function.
9074         (compatible_load_p): New function.
9075         (imm_store_chain_info::coalesce_immediate_stores): Terminate group
9076         if there is overlap and rhs_code is not INTEGER_CST.  For
9077         non-overlapping stores terminate group if rhs is not mergeable.
9078         (get_alias_type_for_stmts): Change first argument from
9079         auto_vec<gimple *> & to vec<gimple *> &.  Add IS_LOAD, CLIQUEP and
9080         BASEP arguments.  If IS_LOAD is true, look at rhs1 of the stmts
9081         instead of lhs.  Compute *CLIQUEP and *BASEP in addition to the
9082         alias type.
9083         (get_location_for_stmts): Change first argument from
9084         auto_vec<gimple *> & to vec<gimple *> &.
9085         (struct split_store): Remove orig_stmts data member, add orig_stores.
9086         (split_store::split_store): Create orig_stores rather than orig_stmts.
9087         (find_constituent_stmts): Renamed to ...
9088         (find_constituent_stores): ... this.  Change second argument from
9089         vec<gimple *> * to vec<store_immediate_info *> *, push pointers
9090         to info structures rather than the statements.
9091         (split_group): Rename ALLOW_UNALIGNED argument to
9092         ALLOW_UNALIGNED_STORE, add ALLOW_UNALIGNED_LOAD argument and handle
9093         it.  Adjust find_constituent_stores caller.
9094         (imm_store_chain_info::output_merged_store): Handle rhs_code other
9095         than INTEGER_CST, adjust split_group, get_alias_type_for_stmts and
9096         get_location_for_stmts callers.  Set MR_DEPENDENCE_CLIQUE and
9097         MR_DEPENDENCE_BASE on the MEM_REFs if they are the same in all stores.
9098         (mem_valid_for_store_merging): New function.
9099         (handled_load): New function.
9100         (pass_store_merging::process_store): New method.
9101         (pass_store_merging::execute): Use process_store method.  Adjust
9102         terminate_all_aliasing_chains caller.
9104 2017-11-03  Wilco Dijkstra  <wdijkstr@arm.com>
9106         * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
9107         Return true for more constants, symbols and label references.
9108         (aarch64_valid_floating_const): Remove unused function.
9110 2017-11-03  Wilco Dijkstra  <wdijkstr@arm.com>
9112         PR target/82786
9113         * config/aarch64/aarch64.c (aarch64_layout_frame):
9114         Undo forcing of LR at bottom of frame.
9116 2017-11-03  Jeff Law  <law@redhat.com>
9118         PR target/82823
9119         * config/i386/i386.c (ix86_expand_prologue): Tighten assert
9120         for int_registers_saved.
9122         * cfganal.c (single_pred_edge_ignoring_loop_edges): New function
9123         extracted from tree-ssa-dom.c.
9124         * cfganal.h (single_pred_edge_ignoring_loop_edges): Prototype.
9125         * tree-ssa-dom.c (single_incoming_edge_ignoring_loop_edges): Remove.
9126         (record_equivalences_from_incoming_edge): Add additional argument
9127         to single_pred_edge_ignoring_loop_edges call.
9128         * tree-ssa-uncprop.c (single_incoming_edge_ignoring_loop_edges): Remove.
9129         (uncprop_dom_walker::before_dom_children): Add additional argument
9130         to single_pred_edge_ignoring_loop_edges call.
9131         * tree-ssa-sccvn.c (sccvn_dom_walker::before_dom_children): Use
9132         single_pred_edge_ignoring_loop_edges rather than open coding.
9133         * tree-vrp.c (evrp_dom_walker::before_dom_children): Similarly.
9135 2017-11-03  Marc Glisse  <marc.glisse@inria.fr>
9137         * match.pd (-(-A)): Rewrite.
9139 2017-11-03  Segher Boessenkool  <segher@kernel.crashing.org>
9141         * config/rs6000/rs60000-protos.h (rs6000_emit_sISEL): Delete.
9142         (rs6000_emit_int_cmove): New declaration.
9143         * config/rs6000/rs6000.c (rs6000_emit_int_cmove): Delete declaration.
9144         (rs6000_emit_sISEL): Delete.
9145         (rs6000_emit_int_cmove): Make non-static.
9146         * config/rs6000/rs6000.md (cstore<mode>4): Use rs6000_emit_int_cmove
9147         instead of rs6000_emit_sISEL.
9149 2017-11-03  Jan Hubicka  <hubicka@ucw.cz>
9151         * asan.c (create_cond_insert_point): Maintain profile.
9152         * ipa-utils.c (ipa_merge_profiles): Be sure only IPA profiles are
9153         merged.
9154         * basic-block.h (struct basic_block_def): Remove frequency.
9155         (EDGE_FREQUENCY): Use to_frequency
9156         * bb-reorder.c (push_to_next_round_p): Use only IPA counts for global
9157         heuristics.
9158         (find_traces): Update to use to_frequency.
9159         (find_traces_1_round): Likewise; use only IPA counts.
9160         (bb_to_key): Likewise.
9161         (connect_traces): Use IPA counts only.
9162         (copy_bb_p): Update to use to_frequency.
9163         (fix_up_crossing_landing_pad): Likewise.
9164         (sanitize_hot_paths): Likewise.
9165         * bt-load.c (basic_block_freq): Likewise.
9166         * cfg.c (init_flow): Set count_max to uninitialized.
9167         (check_bb_profile): Remove frequencies; check counts.
9168         (dump_bb_info): Do not dump frequencies.
9169         (update_bb_profile_for_threading): Update counts only.
9170         (scale_bbs_frequencies_int): Likewise.
9171         (MAX_SAFE_MULTIPLIER): Remove.
9172         (scale_bbs_frequencies_gcov_type): Update counts only.
9173         (scale_bbs_frequencies_profile_count): Update counts only.
9174         (scale_bbs_frequencies): Update counts only.
9175         * cfg.h (struct control_flow_graph): Add count-max.
9176         (update_bb_profile_for_threading): Update prototype.
9177         * cfgbuild.c (find_bb_boundaries): Do not update frequencies.
9178         (find_many_sub_basic_blocks): Likewise.
9179         * cfgcleanup.c (try_forward_edges): Likewise.
9180         (try_crossjump_to_edge): Likewise.
9181         * cfgexpand.c (expand_gimple_cond): Likewise.
9182         (expand_gimple_tailcall): Likewise.
9183         (construct_init_block): Likewise.
9184         (construct_exit_block): Likewise.
9185         * cfghooks.c (verify_flow_info): Check consistency of counts.
9186         (dump_bb_for_graph): Do not dump frequencies.
9187         (split_block_1): Do not update frequencies.
9188         (split_edge): Do not update frequencies.
9189         (make_forwarder_block): Do not update frequencies.
9190         (duplicate_block): Do not update frequencies.
9191         (account_profile_record): Do not update frequencies.
9192         * cfgloop.c (find_subloop_latch_edge_by_profile): Use IPA counts
9193         for global heuristics.
9194         * cfgloopanal.c (average_num_loop_insns): Update to use to_frequency.
9195         (expected_loop_iterations_unbounded): Use counts only.
9196         * cfgloopmanip.c (scale_loop_profile): Simplify.
9197         (create_empty_loop_on_edge): Simplify
9198         (loopify): Simplify
9199         (duplicate_loop_to_header_edge): Simplify
9200         * cfgrtl.c (force_nonfallthru_and_redirect): Update profile.
9201         (update_br_prob_note): Take care of removing note when profile
9202         becomes undefined.
9203         (relink_block_chain): Do not dump frequency.
9204         (rtl_account_profile_record): Use to_frequency.
9205         * cgraph.c (symbol_table::create_edge): Convert count to ipa count.
9206         (cgraph_edge::redirect_call_stmt_to_calle): Conver tcount to ipa count.
9207         (cgraph_update_edges_for_call_stmt_node): Likewise.
9208         (cgraph_edge::verify_count_and_frequency): Update.
9209         (cgraph_node::verify_node): Temporarily disable frequency verification.
9210         * cgraphbuild.c (compute_call_stmt_bb_frequency): Use
9211         to_cgraph_frequency.
9212         (cgraph_edge::rebuild_edges): Convert to ipa counts.
9213         * cgraphunit.c (init_lowered_empty_function): Do not initialize
9214         frequencies.
9215         (cgraph_node::expand_thunk): Update profile.
9216         * except.c (dw2_build_landing_pads): Do not update frequency.
9217         * final.c (compute_alignments): Use to_frequency.
9218         (dump_basic_block_info): Do not dump frequency.
9219         * gimple-pretty-print.c (dump_profile): Do not dump frequency.
9220         (dump_gimple_bb_header): Do not dump frequency.
9221         * gimple-ssa-isolate-paths.c (isolate_path): Do not update frequency;
9222         do update count.
9223         * gimple-streamer-in.c (input_bb): Do not stream frequency.
9224         * gimple-streamer-out.c (output_bb): Do not stream frequency.
9225         * haifa-sched.c (sched_pressure_start_bb): Use to_freuqency.
9226         (init_before_recovery): Do not update frequency.
9227         (sched_create_recovery_edges): Do not update frequency.
9228         * hsa-gen.c (convert_switch_statements): Do not update frequency.
9229         * ipa-cp.c (ipcp_propagate_stage): Update search for max_count.
9230         (ipa_cp_c_finalize): Set max_count to uninitialized.
9231         * ipa-fnsummary.c (get_minimal_bb): Use counts.
9232         (param_change_prob): Use counts.
9233         * ipa-profile.c (ipa_profile_generate_summary): Do not summarize
9234         local profiles.
9235         * ipa-split.c (consider_split): Use to_frequency.
9236         (split_function): Use to_frequency.
9237         * ira-build.c (loop_compare_func): Likewise.
9238         (mark_loops_for_removal): Likewise.
9239         (mark_all_loops_for_removal): Likewise.
9240         * loop-doloop.c (doloop_modify): Do not update frequency.
9241         * loop-unroll.c (unroll_loop_runtime_iterations): Do not update
9242         frequency.
9243         * lto-streamer-in.c (input_function): Update count_max.
9244         * omp-expand.c (expand_omp_taskreg): Update count_max.
9245         * omp-simd-clone.c (simd_clone_adjust): Update profile.
9246         * predict.c (maybe_hot_frequency_p): Use to_frequency.
9247         (maybe_hot_count_p): Use ipa counts only.
9248         (maybe_hot_bb_p): Simplify.
9249         (maybe_hot_edge_p): Simplify.
9250         (probably_never_executed): Do not take frequency argument.
9251         (probably_never_executed_bb_p): Do not pass frequency.
9252         (probably_never_executed_edge_p): Likewise.
9253         (combine_predictions_for_bb): Check that profile is nonzero.
9254         (propagate_freq): Do not set frequency.
9255         (drop_profile): Simplify.
9256         (counts_to_freqs): Simplify.
9257         (expensive_function_p): Use to_frequency.
9258         (propagate_unlikely_bbs_forward): Simplify.
9259         (determine_unlikely_bbs): Simplify.
9260         (estimate_bb_frequencies): Add hack to silence graphite issues.
9261         (compute_function_frequency): Use ipa counts.
9262         (pass_profile::execute): Update.
9263         (rebuild_frequencies): Use counts only.
9264         (force_edge_cold): Use counts only.
9265         * profile-count.c (profile_count::dump): Dump new count types.
9266         (profile_count::differs_from_p): Check compatiblity.
9267         (profile_count::to_frequency): New function.
9268         (profile_count::to_cgraph_frequency): New function.
9269         * profile-count.h (struct function): Declare.
9270         (enum profile_quality): Add profile_guessed_local and
9271         profile_guessed_global0.
9272         (class profile_proability): Decrease number of bits to 29;
9273         update from_reg_br_prob_note and to_reg_br_prob_note.
9274         (class profile_count: Update comment; decrease number of bits
9275         to 61. Check compatibility.
9276         (profile_count::compatible_p): New private member function.
9277         (profile_count::ipa_p): New member function.
9278         (profile_count::operator<): Handle global zero correctly.
9279         (profile_count::operator>): Handle global zero correctly.
9280         (profile_count::operator<=): Handle global zero correctly.
9281         (profile_count::operator>=): Handle global zero correctly.
9282         (profile_count::nonzero_p): New member function.
9283         (profile_count::force_nonzero): New member function.
9284         (profile_count::max): New member function.
9285         (profile_count::apply_scale): Handle IPA scalling.
9286         (profile_count::guessed_local): New member function.
9287         (profile_count::global0): New member function.
9288         (profile_count::ipa): New member function.
9289         (profile_count::to_frequency): Declare.
9290         (profile_count::to_cgraph_frequency): Declare.
9291         * profile.c (OVERLAP_BASE): Delete.
9292         (compute_frequency_overlap): Delete.
9293         (compute_branch_probabilities): Do not use compute_frequency_overlap.
9294         * regs.h (REG_FREQ_FROM_BB): Use to_frequency.
9295         * sched-ebb.c (rank): Use counts only.
9296         * shrink-wrap.c (handle_simple_exit): Use counts only.
9297         (try_shrink_wrapping): Use counts only.
9298         (place_prologue_for_one_component): Use counts only.
9299         * tracer.c (find_best_predecessor): Use to_frequency.
9300         (find_trace): Use to_frequency.
9301         (tail_duplicate): Use to_frequency.
9302         * trans-mem.c (expand_transaction): Do not update frequency.
9303         * tree-call-cdce.c: Do not update frequency.
9304         * tree-cfg.c (gimple_find_sub_bbs): Likewise.
9305         (gimple_merge_blocks): Likewise.
9306         (gimple_split_edge): Likewise.
9307         (gimple_duplicate_sese_region): Likewise.
9308         (gimple_duplicate_sese_tail): Likewise.
9309         (move_sese_region_to_fn): Likewise.
9310         (gimple_account_profile_record): Likewise.
9311         (insert_cond_bb): Likewise.
9312         * tree-complex.c (expand_complex_div_wide): Likewise.
9313         * tree-eh.c (lower_resx): Update profile.
9314         * tree-inline.c (copy_bb): Simplify count scaling; do not scale
9315         frequencies.
9316         (initialize_cfun): Do not initialize frequencies
9317         (freqs_to_counts): Delete.
9318         (copy_cfg_body): Ignore count parameter.
9319         (copy_body): Update.
9320         (expand_call_inline): Update count_max.
9321         (optimize_inline_calls): Update count_max.
9322         (tree_function_versioning): Update count_max.
9323         * tree-ssa-coalesce.c (coalesce_cost_bb): Use to_frequency.
9324         * tree-ssa-ifcombine.c (update_profile_after_ifcombine): Do not update
9325         frequency.
9326         * tree-ssa-loop-im.c (execute_sm_if_changed): Use counts only.
9327         * tree-ssa-loop-ivcanon.c (unloop_loops): Do not update freuqency.
9328         (try_peel_loop): Likewise.
9329         * tree-ssa-loop-ivopts.c (get_scaled_computation_cost_at): Use
9330         to_frequency.
9331         * tree-ssa-loop-manip.c (niter_for_unrolled_loop): Pass -1.
9332         (tree_transform_and_unroll_loop): Do not use frequencies
9333         * tree-ssa-loop-niter.c (estimate_numbers_of_iterations):
9334         Use reliable prediction only.
9335         * tree-ssa-loop-unswitch.c (hoist_guard): Do not use frequencies.
9336         * tree-ssa-sink.c (select_best_block): Use to_frequency.
9337         * tree-ssa-tail-merge.c (replace_block_by): Temporarily disable
9338         probability scaling.
9339         * tree-ssa-threadupdate.c (create_block_for_threading): Do
9340         not update frequency
9341         (any_remaining_duplicated_blocks): Likewise.
9342         (update_profile): Likewise.
9343         (estimated_freqs_path): Delete.
9344         (freqs_to_counts_path): Delete.
9345         (clear_counts_path): Delete.
9346         (ssa_fix_duplicate_block_edges): Likewise.
9347         (duplicate_thread_path): Likewise.
9348         * tree-switch-conversion.c (gen_inbound_check): Use counts.
9349         * tree-tailcall.c (decrease_profile): Do not update frequency.
9350         (eliminate_tail_call): Likewise.
9351         * tree-vect-loop-manip.c (vect_do_peeling): Likewise.
9352         * tree-vect-loop.c (scale_profile_for_vect_loop): Likewise.
9353         (optimize_mask_stores): Likewise.
9354         * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
9355         * ubsan.c (ubsan_expand_null_ifn): Update profile.
9356         (ubsan_expand_ptr_ifn): Update profile.
9357         * value-prof.c (gimple_ic): Simplify.
9358         * value-prof.h (gimple_ic): Update prototype.
9359         * ipa-inline-transform.c (inline_transform): Fix scaling conditoins.
9360         * ipa-inline.c (compute_uninlined_call_time): Be sure that
9361         counts are nonzero.
9362         (want_inline_self_recursive_call_p): Likewise.
9363         (resolve_noninline_speculation): Only cummulate defined counts.
9364         (inline_small_functions): Use nonzero_p.
9365         (ipa_inline): Do not access freed node.
9367 2017-11-03  Wilco Dijkstra  <wdijkstr@arm.com>
9369         * config/aarch64/aarch64.c (aarch64_override_options_internal):
9370         Set PARAM_SCHED_PRESSURE_ALGORITHM to SCHED_PRESSURE_MODEL.
9372 2017-11-03  Kito Cheng  <kito.cheng@gmail.com>
9374         * config/riscv/riscv.c (riscv_legitimize_move): Handle
9375         non-legitimate address.
9377 2017-11-03  Segher Boessenkool  <segher@kernel.crashing.org>
9379         * config/rs6000/rs6000.md (*lt0_disi): Delete.
9380         (*lt0_<mode>di, *lt0_<mode>si): New.
9382 2017-11-03  Segher Boessenkool  <segher@kernel.crashing.org>
9384         * config/rs6000/rs6000.md (move_from_CR_ov_bit): Change condition to
9385         TARGET_PAIRED_FLOAT.
9387 2017-11-03  Siddhesh Poyarekar  <siddhesh.poyarekar@linaro.org>
9388             Jim Wilson  <jim.wilson@linaro.org>
9390         * config/aarch64/aarch64-cores.def (saphira): New CPU.
9391         * config/aarch64/aarch64-tune.md: Regenerated.
9392         * doc/invoke.texi (AArch64 Options/-mtune): Add "saphira".
9393         * gcc/config/aarch64/aarch64.c (saphira_tunings): New tuning table.
9395 2017-11-03  Cupertino Miranda  <cmiranda@synopsys.com>
9397         * config/arc/arc.c (arc_save_restore): Corrected CFA note.
9398         (arc_expand_prologue): Restore blink for millicode.
9399         * config/arc/linux.h (LINK_EH_SPEC): Defined.
9401 2017-11-03  Richard Sandiford  <richard.sandiford@linaro.org>
9403         PR target/82809
9404         * config/i386/i386.c (ix86_vector_duplicate_value): Use
9405         gen_vec_duplicate after forcing the scalar into a register.
9407 2017-11-02  Segher Boessenkool  <segher@kernel.crashing.org>
9409         * combine (try_combine): Print the insns input to try_combine to the
9410         dump file.
9412 2017-11-02  Steve Ellcey  <sellcey@cavium.com>
9414         PR target/79868
9415         * config/aarch64/aarch64-c.c (aarch64_pragma_target_parse):
9416         Remove second argument from aarch64_process_target_attr call.
9417         * config/aarch64/aarch64-protos.h (aarch64_process_target_attr):
9418         Ditto.
9419         * config/aarch64/aarch64.c (aarch64_attribute_info): Change
9420         field type.
9421         (aarch64_handle_attr_arch): Remove second argument.
9422         (aarch64_handle_attr_cpu): Ditto.
9423         (aarch64_handle_attr_tune): Ditto.
9424         (aarch64_handle_attr_isa_flags): Ditto.
9425         (aarch64_process_one_target_attr): Ditto.
9426         (aarch64_process_target_attr): Ditto.
9427         (aarch64_option_valid_attribute_p): Remove second argument.
9428         on aarch64_process_target_attr call.
9430 2017-11-02  David Malcolm  <dmalcolm@redhat.com>
9432         * diagnostic.c: Include "selftest-diagnostic.h".
9433         (selftest::assert_location_text): New function.
9434         (selftest::test_diagnostic_get_location_text): New function.
9435         (selftest::diagnostic_c_tests): Call it.
9437 2017-11-02  David Malcolm  <dmalcolm@redhat.com>
9439         * Makefile.in (OBJS-libcommon): Add selftest-diagnostic.o.
9440         * diagnostic-show-locus.c: Include "selftest-diagnostic.h".
9441         (class selftest::test_diagnostic_context): Move to...
9442         * selftest-diagnostic.c: New file.
9443         * selftest-diagnostic.h: New file.
9445 2017-11-02  James Bowman  <james.bowman@ftdichip.com>
9447         * config/ft32/ft32.c (ft32_addr_space_legitimate_address_p): increase
9448         offset range for FT32B.
9449         * config/ft32/ft32.h: option "mcompress" enables relaxation.
9450         * config/ft32/ft32.md: Add TARGET_NOPM.
9451         * config/ft32/ft32.opt: Add mft32b, mcompress, mnopm.
9452         * gcc/doc/invoke.texi: Add mft32b, mcompress, mnopm.
9454 2017-11-02  Wilco Dijkstra  <wdijkstr@arm.com>
9456         * config/aarch64/aarch64.h (MALLOC_ABI_ALIGNMENT): New define.
9458 2017-11-02  Jeff Law  <law@redhat.com>
9460         * gimple-ssa-sprintf.c (sprintf_dom_walker): Remove
9461         virtual keyword on FINAL OVERRIDE members.
9463         * tree-ssa-propagate.h (ssa_propagation_engine): Group
9464         virtuals together.  Add virtual destructor.
9465         (substitute_and_fold_engine): Similarly.
9467 2017-11-02  Jan Hubicka  <hubicka@ucw.cz>
9469         * x86-tune.def (X86_TUNE_USE_INCDEC): Enable for Haswell+.
9471 2017-11-02  Richard Biener  <rguenther@suse.de>
9473         PR tree-optimization/82795
9474         * tree-if-conv.c (predicate_mem_writes): Remove bogus assert.
9476 2017-11-02  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
9478         * acinclude.m4 (gcc_AC_INITFINI_ARRAY): Don't require
9479         gcc_SUN_LD_VERSION.
9480         (gcc_GAS_CHECK_FEATURE): Remove.
9481         * configure.ac (ld_vers) <*-*-solaris2*>: Move comments from
9482         gcc_AC_INITFINI_ARRAY here.  Update for Solaris 11.4 changes.
9483         * configure: Regenerate.
9485 2017-11-02  Claudiu Zissulescu <claziss@synopsys.com>
9487         * config/arc/arc.c (hwloop_optimize): Account for empty
9488         body loops.
9490 2017-11-02  Richard Biener  <rguenther@suse.de>
9492         PR middle-end/82765
9493         * varasm.c (decode_addr_const): Make offset HOST_WIDE_INT.
9494         Truncate ARRAY_REF index and element size.
9496 2017-11-01  Palmer Dabbelt  <palmer@dabbelt.com>
9498         * doc/invoke.texi (RISC-V Options): Use "@minus{}2 GB", not "-2 GB".
9500 2017-11-01  Jeff Law  <law@redhat.com>
9502         * tree-ssa-ccp.c (ccp_folder): New class derived from
9503         substitute_and_fold_engine.
9504         (ccp_folder::get_value): New member function.
9505         (ccp_folder::fold_stmt): Renamed from ccp_fold_stmt.
9506         (ccp_fold_stmt): Remove prototype.
9507         (ccp_finalize): Call substitute_and_fold from the ccp_class.
9508         * tree-ssa-copy.c (copy_folder): New class derived from
9509         substitute_and_fold_engine.
9510         (copy_folder::get_value): Renamed from get_value.
9511         (fini_copy_prop): Call substitute_and_fold from copy_folder class.
9512         * tree-vrp.c (vrp_folder): New class derived from
9513         substitute_and_fold_engine.
9514         (vrp_folder::fold_stmt): Renamed from vrp_fold_stmt.
9515         (vrp_folder::get_value): New member function.
9516         (vrp_finalize): Call substitute_and_fold from vrp_folder class.
9517         (evrp_dom_walker::before_dom_children): Similarly for replace_uses_in.
9518         * tree-ssa-propagate.h (substitute_and_fold_engine): New class to
9519         provide a class interface to folder/substitute routines.
9520         (ssa_prop_fold_stmt_fn): Remove typedef.
9521         (ssa_prop_get_value_fn): Likewise.
9522         (subsitute_and_fold): Remove prototype.
9523         (replace_uses_in): Likewise.
9524         * tree-ssa-propagate.c (substitute_and_fold_engine::replace_uses_in):
9525         Renamed from replace_uses_in.  Call the virtual member function
9526         (substitute_and_fold_engine::replace_phi_args_in): Similarly.
9527         (substitute_and_fold_dom_walker): Remove initialization of
9528         data member entries for calbacks.  Add substitute_and_fold_engine
9529         member and initialize it.
9530         (substitute_and_fold_dom_walker::before_dom_children0: Use the
9531         member functions for get_value, replace_phi_args_in c
9532         replace_uses_in, and fold_stmt calls.
9533         (substitute_and_fold_engine::substitute_and_fold): Renamed from
9534         substitute_and_fold.  Remove assert.   Update ctor call.
9536         * tree-ssa-propagate.h (ssa_prop_visit_stmt_fn): Remove typedef.
9537         (ssa_prop_visit_phi_fn): Likewise.
9538         (class ssa_propagation_engine): New class to provide an interface
9539         into ssa_propagate.
9540         * tree-ssa-propagate.c (ssa_prop_visit_stmt): Remove file scoped
9541         variable.
9542         (ssa_prop_visit_phi): Likewise.
9543         (ssa_propagation_engine::simulate_stmt): Moved into class.
9544         Call visit_phi/visit_stmt from the class rather than via
9545         file scoped static variables.
9546         (ssa_propagation_engine::simulate_block): Moved into class.
9547         (ssa_propagation_engine::process_ssa_edge_worklist): Similarly.
9548         (ssa_propagation_engine::ssa_propagate): Similarly.  No longer
9549         set file scoped statics for the visit_stmt/visit_phi callbacks.
9550         * tree-complex.c (complex_propagate): New class derived from
9551         ssa_propagation_engine.
9552         (complex_propagate::visit_stmt): Renamed from complex_visit_stmt.
9553         (complex_propagate::visit_phi): Renamed from complex_visit_phi.
9554         (tree_lower_complex): Call ssa_propagate via the complex_propagate
9555         class.
9556         * tree-ssa-ccp.c: (ccp_propagate): New class derived from
9557         ssa_propagation_engine.
9558         (ccp_propagate::visit_phi): Renamed from ccp_visit_phi_node.
9559         (ccp_propagate::visit_stmt): Renamed from ccp_visit_stmt.
9560         (do_ssa_ccp): Call ssa_propagate from the ccp_propagate class.
9561         * tree-ssa-copy.c (copy_prop): New class derived from
9562         ssa_propagation_engine.
9563         (copy_prop::visit_stmt): Renamed from copy_prop_visit_stmt.
9564         (copy_prop::visit_phi): Renamed from copy_prop_visit_phi_node.
9565         (execute_copy_prop): Call ssa_propagate from the copy_prop class.
9566         * tree-vrp.c (vrp_prop): New class derived from ssa_propagation_engine.
9567         (vrp_prop::visit_stmt): Renamed from vrp_visit_stmt.
9568         (vrp_prop::visit_phi): Renamed from vrp_visit_phi_node.
9569         (execute_vrp): Call ssa_propagate from the vrp_prop class.
9571 2017-11-01  Jakub Jelinek  <jakub@redhat.com>
9573         PR rtl-optimization/82778
9574         PR rtl-optimization/82597
9575         * compare-elim.c (struct comparison): Add in_a_setter field.
9576         (find_comparison_dom_walker::before_dom_children): Remove killed
9577         bitmap and df_simulate_find_defs call, instead walk the defs.
9578         Compute last_setter and initialize in_a_setter.  Merge definitions
9579         with first initialization for a few variables.
9580         (try_validate_parallel): Use insn_invalid_p instead of
9581         recog_memoized.  Return insn rather than just the pattern.
9582         (try_merge_compare): Fix up comment.  Don't uselessly test if
9583         in_a is a REG_P.  Use cmp->in_a_setter instead of walking UD
9584         chains.
9585         (execute_compare_elim_after_reload): Remove df_chain_add_problem
9586         call.
9588 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9589             Alan Hayward  <alan.hayward@arm.com>
9590             David Sherwood  <david.sherwood@arm.com>
9592         * config/aarch64/aarch64.c (aarch64_rtx_costs): Use
9593         aarch64_hard_regno_nregs to get the number of registers
9594         in a mode.
9596 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9597             Alan Hayward  <alan.hayward@arm.com>
9598             David Sherwood  <david.sherwood@arm.com>
9600         * config/aarch64/constraints.md (Upl): Rename to...
9601         (Uaa): ...this.
9602         * config/aarch64/aarch64.md
9603         (*zero_extend<SHORT:mode><GPI:mode>2_aarch64, *addsi3_aarch64_uxtw):
9604         Update accordingly.
9606 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9607             Alan Hayward  <alan.hayward@arm.com>
9608             David Sherwood  <david.sherwood@arm.com>
9610         * config/aarch64/aarch64.c (aarch64_add_constant_internal)
9611         (aarch64_add_constant, aarch64_add_sp, aarch64_sub_sp): Move
9612         earlier in file.
9614 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9615             Alan Hayward  <alan.hayward@arm.com>
9616             David Sherwood  <david.sherwood@arm.com>
9618         * config/aarch64/aarch64.c (aarch64_evpc_trn, aarch64_evpc_uzp)
9619         (aarch64_evpc_zip, aarch64_evpc_ext, aarch64_evpc_rev)
9620         (aarch64_evpc_dup): Generate rtl direcly, rather than using
9621         named expanders.
9622         (aarch64_expand_vec_perm_const_1): Explicitly check for permutes
9623         of a single element.
9624         * config/aarch64/iterators.md: Add a comment above the permute
9625         unspecs to say that they are generated directly by
9626         aarch64_expand_vec_perm_const.
9627         * config/aarch64/aarch64-simd.md: Likewise the permute instructions.
9629 2017-11-01  Nathan Sidwell  <nathan@acm.org>
9631         * tree-dump.c (dequeue_and_dump): Use HAS_DECL_ASSEMBLER_NAME_P.
9633 2017-11-01  Palmer Dabbelt  <palmer@dabbelt.com>
9635         * doc/invoke.texi (RISC-V Options): Explicitly name the medlow
9636         and medany code models, and describe what they do.
9638 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9640         Revert accidental duplicate:
9642         * combine.c (can_change_dest_mode): Reject changes in
9643         REGMODE_NATURAL_SIZE.
9645 2017-11-01  Segher Boessenkool  <segher@kernel.crashing.org>
9647         PR rtl-optimization/64682
9648         PR rtl-optimization/69567
9649         PR rtl-optimization/69737
9650         PR rtl-optimization/82683
9651         * combine.c (distribute_notes) <REG_DEAD>: If the new I2 sets the same
9652         register mentioned in the note, drop the note, unless it came from I3,
9653         in which case it should go to I3 again.
9655 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9657         * tree-ssa-dse.c (normalize_ref): Check whether the ranges overlap
9658         and return false if not.
9659         (clear_bytes_written_by, live_bytes_read): Update accordingly.
9661 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9663         * tree-ssa-alias.h (ranges_overlap_p): Return false if either
9664         range is known to be empty.
9666 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9667             Alan Hayward  <alan.hayward@arm.com>
9668             David Sherwood  <david.sherwood@arm.com>
9670         * simplify-rtx.c (simplify_const_unary_operation): Use GET_MODE_NUNITS
9671         and CONST_VECTOR_NUNITS instead of computing the number of units from
9672         the byte sizes of the vector and element.
9673         (simplify_binary_operation_1): Likewise.
9674         (simplify_const_binary_operation): Likewise.
9675         (simplify_ternary_operation): Likewise.
9677 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9678             Alan Hayward  <alan.hayward@arm.com>
9679             David Sherwood  <david.sherwood@arm.com>
9681         * var-tracking.c (INT_MEM_OFFSET): Replace with...
9682         (int_mem_offset): ...this new function.
9683         (var_mem_set, var_mem_delete_and_set, var_mem_delete)
9684         (find_mem_expr_in_1pdv, dataflow_set_preserve_mem_locs)
9685         (same_variable_part_p, use_type, add_stores, vt_get_decl_and_offset):
9686         Update accordingly.
9688 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9689             Alan Hayward  <alan.hayward@arm.com>
9690             David Sherwood  <david.sherwood@arm.com>
9692         * lower-subreg.c (interesting_mode_p): New function.
9693         (compute_costs, find_decomposable_subregs, decompose_register)
9694         (simplify_subreg_concatn, can_decompose_p, resolve_simple_move)
9695         (resolve_clobber, dump_choices): Use it.
9697 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9698             Alan Hayward  <alan.hayward@arm.com>
9699             David Sherwood  <david.sherwood@arm.com>
9701         * rtlhash.c (add_rtx): Use add_hwi for 'w' and add_int for 'i'.
9703 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9704             Alan Hayward  <alan.hayward@arm.com>
9705             David Sherwood  <david.sherwood@arm.com>
9707         * alias.c (find_base_value, find_base_term): Only process integer
9708         truncations.  Check the precision rather than the size.
9710 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9711             Alan Hayward  <alan.hayward@arm.com>
9712             David Sherwood  <david.sherwood@arm.com>
9714         * machmode.h (is_narrower_int_mode): New function
9715         * optabs.c (expand_float, expand_fix): Use it.
9716         * dwarf2out.c (rotate_loc_descriptor): Likewise.
9718 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9719             Alan Hayward  <alan.hayward@arm.com>
9720             David Sherwood  <david.sherwood@arm.com>
9722         * rtl.h (narrower_subreg_mode): New function.
9723         * ira-color.c (update_costs_from_allocno): Use it.
9725 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9726             Alan Hayward  <alan.hayward@arm.com>
9727             David Sherwood  <david.sherwood@arm.com>
9729         * optabs-query.h (convert_optab_p): New function, split out from...
9730         (convert_optab_handler): ...here.
9731         (widening_optab_handler): Delete.
9732         (find_widening_optab_handler): Remove permit_non_widening parameter.
9733         (find_widening_optab_handler_and_mode): Likewise.  Provide an
9734         override that operates on mode class wrappers.
9735         * optabs-query.c (widening_optab_handler): Delete.
9736         (find_widening_optab_handler_and_mode): Remove permit_non_widening
9737         parameter.  Assert that the two modes are the same class and that
9738         the "from" mode is narrower than the "to" mode.  Use
9739         convert_optab_handler instead of widening_optab_handler.
9740         * expmed.c (expmed_mult_highpart_optab): Use convert_optab_handler
9741         instead of widening_optab_handler.
9742         * expr.c (expand_expr_real_2): Update calls to
9743         find_widening_optab_handler.
9744         * optabs.c (expand_widen_pattern_expr): Likewise.
9745         (expand_binop_directly): Take the insn_code as a parameter.
9746         (expand_binop): Only call find_widening_optab_handler for
9747         conversion optabs; use optab_handler otherwise.  Update calls
9748         to find_widening_optab_handler and expand_binop_directly.
9749         Use convert_optab_handler instead of widening_optab_handler.
9750         * tree-ssa-math-opts.c (convert_mult_to_widen): Update calls to
9751         find_widening_optab_handler and use scalar_mode rather than
9752         machine_mode.
9753         (convert_plusminus_to_widen): Likewise.
9755 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9756             Alan Hayward  <alan.hayward@arm.com>
9757             David Sherwood  <david.sherwood@arm.com>
9759         * machmode.h (fixed_size_mode): New class.
9760         * rtl.h (get_pool_mode): Return fixed_size_mode.
9761         * gengtype.c (main): Add fixed_size_mode.
9762         * target.def (get_raw_result_mode): Return a fixed_size_mode.
9763         (get_raw_arg_mode): Likewise.
9764         * doc/tm.texi: Regenerate.
9765         * targhooks.h (default_get_reg_raw_mode): Return a fixed_size_mode.
9766         * targhooks.c (default_get_reg_raw_mode): Likewise.
9767         * config/ia64/ia64.c (ia64_get_reg_raw_mode): Likewise.
9768         * config/mips/mips.c (mips_get_reg_raw_mode): Likewise.
9769         * config/msp430/msp430.c (msp430_get_raw_arg_mode): Likewise.
9770         (msp430_get_raw_result_mode): Likewise.
9771         * config/avr/avr-protos.h (regmask): Use as_a <fixed_side_mode>
9772         * dbxout.c (dbxout_parms): Require fixed-size modes.
9773         * expr.c (copy_blkmode_from_reg, copy_blkmode_to_reg): Likewise.
9774         * gimple-ssa-store-merging.c (encode_tree_to_bitpos): Likewise.
9775         * omp-low.c (lower_oacc_reductions): Likewise.
9776         * simplify-rtx.c (simplify_immed_subreg): Take fixed_size_modes.
9777         (simplify_subreg): Update accordingly.
9778         * varasm.c (constant_descriptor_rtx::mode): Change to fixed_size_mode.
9779         (force_const_mem): Update accordingly.  Return NULL_RTX for modes
9780         that aren't fixed-size.
9781         (get_pool_mode): Return a fixed_size_mode.
9782         (output_constant_pool_2): Take a fixed_size_mode.
9784 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9785             Alan Hayward  <alan.hayward@arm.com>
9786             David Sherwood  <david.sherwood@arm.com>
9788         * doc/rtl.texi (vec_series): Document.
9789         (const): Say that the operand can be a vec_series.
9790         * rtl.def (VEC_SERIES): New rtx code.
9791         * rtl.h (const_vec_series_p_1): Declare.
9792         (const_vec_series_p): New function.
9793         * emit-rtl.h (gen_const_vec_series): Declare.
9794         (gen_vec_series): Likewise.
9795         * emit-rtl.c (const_vec_series_p_1, gen_const_vec_series)
9796         (gen_vec_series): Likewise.
9797         * optabs.c (expand_mult_highpart): Use gen_const_vec_series.
9798         * simplify-rtx.c (simplify_unary_operation): Handle negations
9799         of vector series.
9800         (simplify_binary_operation_series): New function.
9801         (simplify_binary_operation_1): Use it.  Handle VEC_SERIES.
9802         (test_vector_ops_series): New function.
9803         (test_vector_ops): Call it.
9804         * config/powerpcspe/altivec.md (altivec_lvsl): Use
9805         gen_const_vec_series.
9806         (altivec_lvsr): Likewise.
9807         * config/rs6000/altivec.md (altivec_lvsl, altivec_lvsr): Likewise.
9809 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9810             Alan Hayward  <alan.hayward@arm.com>
9811             David Sherwood  <david.sherwood@arm.com>
9813         * doc/rtl.texi (const): Update description of address constants.
9814         Say that vector constants are allowed too.
9815         * common.md (E, F): Use CONSTANT_P instead of checking for
9816         CONST_VECTOR.
9817         * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
9818         checking for CONST_VECTOR.
9819         * expmed.c (make_tree): Use build_vector_from_val for a CONST
9820         VEC_DUPLICATE.
9821         * expr.c (expand_expr_real_2): Check for vector modes instead
9822         of checking for CONST_VECTOR.
9823         * rtl.h (const_vec_p): New function.
9824         (const_vec_duplicate_p): Check for a CONST VEC_DUPLICATE.
9825         (unwrap_const_vec_duplicate): Handle them here too.
9827 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9828             David Malcolm  <dmalcolm@redhat.com>
9829             Alan Hayward  <alan.hayward@arm.com>
9830             David Sherwood  <david.sherwood@arm.com>
9832         * rtl.h (vec_duplicate_p): New function.
9833         * selftest-rtl.c (assert_rtx_eq_at): New function.
9834         * selftest-rtl.h (ASSERT_RTX_EQ): New macro.
9835         (assert_rtx_eq_at): Declare.
9836         * selftest.h (selftest::simplify_rtx_c_tests): Declare.
9837         * selftest-run-tests.c (selftest::run_tests): Call it.
9838         * simplify-rtx.c: Include selftest.h and selftest-rtl.h.
9839         (simplify_unary_operation_1): Recursively handle vector duplicates.
9840         (simplify_binary_operation_1): Likewise.  Handle VEC_SELECTs of
9841         vector duplicates.
9842         (simplify_subreg): Handle subregs of vector duplicates.
9843         (make_test_reg, test_vector_ops_duplicate, test_vector_ops)
9844         (selftest::simplify_rtx_c_tests): New functions.
9846 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9847             Alan Hayward  <alan.hayward@arm.com>
9848             David Sherwood  <david.sherwood@arm.com>
9850         * emit-rtl.h (gen_const_vec_duplicate): Declare.
9851         (gen_vec_duplicate): Likewise.
9852         * emit-rtl.c (gen_const_vec_duplicate_1): New function, split
9853         out from...
9854         (gen_const_vector): ...here.
9855         (gen_const_vec_duplicate, gen_vec_duplicate): New functions.
9856         (gen_rtx_CONST_VECTOR): Use gen_const_vec_duplicate for constants
9857         whose elements are all equal.
9858         * optabs.c (expand_vector_broadcast): Use gen_const_vec_duplicate.
9859         * simplify-rtx.c (simplify_const_unary_operation): Likewise.
9860         (simplify_relational_operation): Likewise.
9861         * config/aarch64/aarch64.c (aarch64_simd_gen_const_vector_dup):
9862         Likewise.
9863         (aarch64_simd_dup_constant): Use gen_vec_duplicate.
9864         (aarch64_expand_vector_init): Likewise.
9865         * config/arm/arm.c (neon_vdup_constant): Likewise.
9866         (neon_expand_vector_init): Likewise.
9867         (arm_expand_vec_perm): Use gen_const_vec_duplicate.
9868         (arm_block_set_unaligned_vect): Likewise.
9869         (arm_block_set_aligned_vect): Likewise.
9870         * config/arm/neon.md (neon_copysignf<mode>): Likewise.
9871         * config/i386/i386.c (ix86_expand_vec_perm): Likewise.
9872         (expand_vec_perm_even_odd_pack): Likewise.
9873         (ix86_vector_duplicate_value): Use gen_vec_duplicate.
9874         * config/i386/sse.md (one_cmpl<mode>2): Use CONSTM1_RTX.
9875         * config/ia64/ia64.c (ia64_expand_vecint_compare): Use
9876         gen_const_vec_duplicate.
9877         * config/ia64/vect.md (addv2sf3, subv2sf3): Use CONST1_RTX.
9878         * config/mips/mips.c (mips_gen_const_int_vector): Use
9879         gen_const_vec_duplicate.
9880         (mips_expand_vector_init): Use CONST0_RTX.
9881         * config/powerpcspe/altivec.md (abs<mode>2, nabs<mode>2): Likewise.
9882         (define_split): Use gen_const_vec_duplicate.
9883         * config/rs6000/altivec.md (abs<mode>2, nabs<mode>2): Use CONST0_RTX.
9884         (define_split): Use gen_const_vec_duplicate.
9885         * config/s390/vx-builtins.md (vec_genmask<mode>): Likewise.
9886         (vec_ctd_s64, vec_ctd_u64, vec_ctsl, vec_ctul): Likewise.
9887         * config/spu/spu.c (spu_const): Likewise.
9889 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9890             Alan Hayward  <alan.hayward@arm.com>
9891             David Sherwood  <david.sherwood@arm.com>
9893         * combine.c (can_change_dest_mode): Reject changes in
9894         REGMODE_NATURAL_SIZE.
9896 2017-10-31  Sandra Loosemore  <sandra@codesourcery.com>
9898         * configure.ac (--enable-libssp): New.
9899         (gcc_cv_libc_provides_ssp): Check for explicit setting before
9900         trying to determine target-specific default.  Adjust indentation.
9901         * configure: Regenerated.
9902         * doc/install.texi (Configuration): Expand --disable-libssp
9903         documentation.
9905 2017-10-31  Daniel Santos  <daniel.santos@pobox.com>
9907         config/i386/i386.c (ix86_expand_epilogue): Correct stack
9908         calculation.
9910 2017-10-31  Martin Jambor  <mjambor@suse.cz>
9912         PR c++/81702
9913         * gimple-fold.c (gimple_get_virt_method_for_vtable): Remove assert.
9915 2017-10-31  David Malcolm  <dmalcolm@redhat.com>
9917         * auto-profile.c (autofdo_source_profile::read): Use
9918         UNKNOWN_LOCATION rather than 0.
9919         * diagnostic-core.h (warning_at_rich_loc): Rename to...
9920         (warning_at): ...this overload.
9921         (warning_at_rich_loc_n): Rename to...
9922         (warning_n): ...this overload.
9923         (error_at_rich_loc): Rename to...
9924         (error_at): ...this overload.
9925         (pedwarn_at_rich_loc): Rename to...
9926         (pedwarn): ...this overload.
9927         (permerror_at_rich_loc): Rename to...
9928         (permerror): ...this overload.
9929         (inform_at_rich_loc): Rename to...
9930         (inform): ...this overload.
9931         * diagnostic.c: (diagnostic_n_impl): Delete location_t-based decl.
9932         (diagnostic_n_impl_richloc): Rename to...
9933         (diagnostic_n_impl): ...this rich_location *-based decl.
9934         (inform_at_rich_loc): Rename to...
9935         (inform): ...this, and add an assertion.
9936         (inform_n): Update for removal of location_t-based diagnostic_n_impl.
9937         (warning_at_rich_loc): Rename to...
9938         (warning_at): ...this, and add an assertion.
9939         (warning_at_rich_loc_n): Rename to...
9940         (warning_n): ...this, and add an assertion.
9941         (warning_n): Update location_t-based implementation for removal of
9942         location_t-based diagnostic_n_impl.
9943         (pedwarn_at_rich_loc): Rename to...
9944         (pedwarn): ...this, and add an assertion.
9945         (permerror_at_rich_loc): Rename to...
9946         (permerror): ...this, and add an assertion.
9947         (error_n): Update for removal of location_t-based diagnostic_n_impl.
9948         (error_at_rich_loc): Rename to...
9949         (error_at): ...this, and add an assertion.
9950         * gcc.c (do_spec_1): Use UNKNOWN_LOCATION rather than 0.
9951         (driver::do_spec_on_infiles): Likewise.
9952         * substring-locations.c (format_warning_va): Update for renaming
9953         of inform_at_rich_loc.
9955 2017-10-31  Michael Meissner  <meissner@linux.vnet.ibm.com>
9957         * builtins.def (DEF_FLOATN_BUILTIN): Change most _Float<N> and
9958         _Float<N>X built-in functions so that the variant without the
9959         "__builtin_" prefix is only enabled for the GNU C and Objective C
9960         languages when they are in non-strict ANSI/ISO mode.
9961         (DEF_EXT_LIB_FLOATN_NX_BUILTINS): Likewise.
9962         * target.def (floatn_builtin_p): Add a target hook to control
9963         whether _Float<N> and _Float<N>X built-in functions without the
9964         "__builtin_" prefix are enabled, and return true for C and
9965         Objective C in the default hook.  Include langhooks.h in
9966         targhooks.c.
9967         * targhooks.h (default_floatn_builtin_p): Likewise.
9968         * targhooks.c (default_floatn_builtin_p): Likewise.
9969         * doc/tm.texi.in (TARGET_FLOATN_BUILTIN_P): Document the
9970         floatn_builtin_p target hook.
9971         * doc/tm.texi (TARGET_FLOATN_BUILTIN_P): Likewise.
9973 2017-10-31  Matthew Fortune  <matthew.fortune@imgtec.com>
9974             Eric Botcazou  <ebotcazou@adacore.com>
9976         PR rtl-optimization/81803
9977         * lra-constraints.c (curr_insn_transform): Also reload the whole
9978         register for a strict subreg no wider than a word if this is for
9979         a WORD_REGISTER_OPERATIONS target.
9981 2017-10-31  Jason Merrill  <jason@redhat.com>
9983         * gdbinit.in: Skip over inlines from timevar.h.
9985 2017-10-31  Martin Liska  <mliska@suse.cz>
9987         * doc/gcov.texi: Document new option.
9988         * gcov.c (print_usage): Likewise print it.
9989         (process_args): Support the argument.
9990         (format_count): New function.
9991         (format_gcov): Use the function.
9993 2017-10-31  Martin Liska  <mliska@suse.cz>
9995         * gcov.c (struct name_map): do not use typedef.
9996         Define operator== and operator<.
9997         (name_search): Remove.
9998         (name_sort): Remove.
9999         (main): Do not allocate names.
10000         (process_file): Add vertical space.
10001         (generate_results): Use std::find.
10002         (release_structures): Do not release memory.
10003         (find_source): Use std::find.
10005 2017-10-31  Martin Liska  <mliska@suse.cz>
10007         * gcov.c (struct line_info): Remove it's typedef.
10008         (line_info::line_info): Add proper ctor.
10009         (line_info::has_block): Do not use a typedef.
10010         (struct source_info): Do not use typedef.
10011         (circuit): Likewise.
10012         (get_cycles_count): Likewise.
10013         (output_intermediate_file): Iterate via vector iterator.
10014         (add_line_counts): Use std::vector methods.
10015         (accumulate_line_counts): Likewise.
10016         (output_lines): Likewise.
10018 2017-10-31  Martin Liska  <mliska@suse.cz>
10020         * gcov.c (struct source_info): Remove typedef.
10021         (source_info::source_info): Add proper ctor.
10022         (accumulate_line_counts): Use struct, not it's typedef.
10023         (output_gcov_file): Likewise.
10024         (output_lines): Likewise.
10025         (main): Do not allocate an array.
10026         (output_intermediate_file): Use size of vector container.
10027         (process_file): Resize the vector.
10028         (generate_results): Do not preallocate, use newly added vector
10029         lines.
10030         (release_structures): Do not release sources.
10031         (find_source): Use vector methods.
10032         (add_line_counts): Do not use typedef.
10034 2017-10-31  Martin Liska  <mliska@suse.cz>
10036         * doc/gcov.texi: Document that.
10037         * gcov.c (add_line_counts): Mark lines with a non-executed
10038         statement.
10039         (output_line_beginning): Handle such lines.
10040         (output_lines): Pass new argument.
10041         (output_intermediate_file): Print it in intermediate format.
10043 2017-10-31  Martin Liska  <mliska@suse.cz>
10045         * color-macros.h: New file.
10046         * diagnostic-color.c: Factor out color related to macros to
10047         color-macros.h.
10048         * doc/gcov.texi: Document -k option.
10049         * gcov.c (INCLUDE_STRING): Include string.h.
10050         (print_usage): Add -k option.
10051         (process_args): Parse it.
10052         (pad_count_string): New function.
10053         (output_line_beginning): Likewise.
10054         (DEFAULT_LINE_START): New macro.
10055         (output_lines): Support color output.
10057 2017-10-31  Martin Liska  <mliska@suse.cz>
10059         PR gcov-profile/82633
10060         * doc/gcov.texi: Document -fkeep-{static,inline}-functions and
10061         their interaction with GCOV infrastructure.
10062         * configure.ac: Add -fkeep-{inline,static}-functions to
10063         coverage_flags.
10064         * configure: Regenerate.
10066 2017-10-31  Uros Bizjak  <ubizjak@gmail.com>
10068         PR target/82772
10069         * config/alpha/sync.md (fetchop_constr) <and>: Change to "rINM".
10071 2017-10-31  Segher Boessenkool  <segher@kernel.crashing.org>
10073         PR target/82674
10074         * config/rs6000/rs6000.md (allocate_stack): Force update interval
10075         into a register if it does not fit into an immediate offset field.
10077 2017-10-31  Olivier Hainque  <hainque@adacore.com>
10079         * gcc/Makefile.in (FLAGS_TO_PASS): Pass libsubdir as well.
10081 2017-10-31  Julia Koval  <julia.koval@intel.com>
10083         * config.gcc: Add gfniintrin.h.
10084         * config/i386/gfniintrin.h: New.
10085         * config/i386/i386-builtin-types.def
10086         (__builtin_ia32_vgf2p8affineinvqb_v64qi,
10087         __builtin_ia32_vgf2p8affineinvqb_v64qi_mask,
10088         __builtin_ia32_vgf2p8affineinvqb_v32qi,
10089         __builtin_ia32_vgf2p8affineinvqb_v32qi_mask,
10090         __builtin_ia32_vgf2p8affineinvqb_v16qi,
10091         __builtin_ia32_vgf2p8affineinvqb_v16qi_mask): New builtins.
10092         * config/i386/i386-builtin.def (V64QI_FTYPE_V64QI_V64QI_INT_V64QI_UDI,
10093         V32QI_FTYPE_V32QI_V32QI_INT_V32QI_USI,
10094         V16QI_FTYPE_V16QI_V16QI_INT_V16QI_UHI,
10095         V64QI_FTYPE_V64QI_V64QI_INT): New types.
10096         * config/i386/i386.c (ix86_expand_args_builtin): Handle new types.
10097         * config/i386/immintrin.h: Include gfniintrin.h.
10098         * config/i386/sse.md (vgf2p8affineinvqb_*) New pattern.
10100 2017-10-30  Eric Botcazou  <ebotcazou@adacore.com>
10102         * gcc.c (HAVE_TARGET_EXECUTABLE_SUFFIX): Remove old kludge.
10104 2017-10-30  Wilco Dijkstra  <wdijkstr@arm.com>
10106         * config/arm/arm.md (ashldi3): Remove shift by 1 expansion.
10107         (arm_ashldi3_1bit): Remove pattern.
10108         (ashrdi3): Remove shift by 1 expansion.
10109         (arm_ashrdi3_1bit): Remove pattern.
10110         (lshrdi3): Remove shift by 1 expansion.
10111         (arm_lshrdi3_1bit): Remove pattern.
10112         * config/arm/arm.c (arm_rtx_costs_internal): Slightly increase
10113         cost of ashldi3 by 1.
10114         * config/arm/neon.md (ashldi3_neon): Remove shift by 1 expansion.
10115         (<shift>di3_neon): Likewise.
10117 2017-10-30  Dominik Infuehr  <dominik.infuehr@theobroma-systems.com>
10119         * config/aarch64/aarch64-simd.md (*aarch64_simd_mov): Rename
10120         both identically named patterns to (*aarch64_simd_mov<VD:mode>)
10121         and (*aarch64_simd_mov<VQ:mode>).
10122         (*aarch64_simd_mov<VD:mode>): Change type attribute to match
10123         pattern alternative.
10124         (*aarch64_simd_mov<VQ:mode>): Re-order and change type
10125         attributes to match pattern alternative.
10127 2017-10-30  Steven Munroe  <munroesj@gcc.gnu.org>
10129         * config.gcc (powerpc*-*-*): Add emmintrin.h.
10130         * config/rs6000/emmintrin.h: New file.
10131         * config/rs6000/x86intrin.h [__ALTIVEC__]: Include emmintrin.h.
10133 2017-10-30  Wilco Dijkstra  <wdijkstr@arm.com>
10135         * config/arm/vfp.md (movdi_vfp): Merge changes from movdi_vfp_cortexa8.
10136         * (movdi_vfp_cortexa8): Remove pattern.
10138 2017-10-30  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
10140         * doc/install.texi (Specific, alpha*-*-*): Remove DEC OSF/1
10141         etc. reference.
10142         (Specific, alpha*-dec-osf5.1): Remove.
10143         (Specific, mips-sgi-irix5): Remove.
10144         (Specific, mips-sgi-irix6): Remove.
10146 2017-10-30  Jakub Jelinek  <jakub@redhat.com>
10148         PR middle-end/22141
10149         * gimple-ssa-store-merging.c (merged_store_group::apply_stores): Fix
10150         arguments to clear_bit_region_be.
10152 2017-10-30  Jim Wilson  <wilson@tuliptree.org>
10154         * gimplify.c: Include memmodel.h.
10156 2017-10-30  Martin Jambor  <mjambor@suse.cz>
10158         * omp-grid.c (grid_attempt_target_gridification): Also insert a
10159         condition whether loop should be executed at all.
10161 2017-10-30  Will Schmidt  <will_schmidt@vnet.ibm.com>
10163         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add support for
10164         gimple folding of vec_madd() intrinsics.
10165         * config/rs6000/altivec.md (mulv8hi3): Rename altivec_vmladduhm to
10166         fmav8hi4.  (altivec_vmladduhm): Rename to fmav8hi4.
10167         * config/rs6000/rs6000-builtin.def: Rename vmladduhm to fmav8hi4
10169 2017-10-30  Richard Biener  <rguenther@suse.de>
10171         PR tree-optimization/82762
10172         Revert
10173         2017-10-23  Richard Biener  <rguenther@suse.de>
10175         PR tree-optimization/82129
10176         Revert
10177         2017-08-01  Richard Biener  <rguenther@suse.de>
10179         PR tree-optimization/81181
10180         * tree-ssa-pre.c (compute_antic_aux): Defer clean() to ...
10181         (compute_antic): ... end of iteration here.
10183 2017-10-30  Joseph Myers  <joseph@codesourcery.com>
10185         * doc/invoke.texi (C Dialect Options): Document -std=c17,
10186         -std=iso9899:2017 and -std=gnu17.
10187         * doc/standards.texi (C Language): Document C17 support.
10188         * doc/cpp.texi (Overview): Mention -std=c17.
10189         (Standard Predefined Macros): Document C11 and C17 values of
10190         __STDC_VERSION__.  Do not refer to C99 support as incomplete.
10191         * doc/extend.texi (Inline): Do not list individual options for
10192         standards newer than C99.
10193         * dwarf2out.c (highest_c_language, gen_compile_unit_die): Handle
10194         "GNU C17".
10195         * config/rl78/rl78.c (rl78_option_override): Handle "GNU C17"
10196         language name.
10198 2017-10-30  Maxim Ostapenko  <m.ostapenko@samsung.com>
10200         * asan.c (asan_finish_file): Align asan globals array by shadow
10201         granularity.
10203 2017-10-30  Jakub Jelinek  <jakub@redhat.com>
10205         PR middle-end/22141
10206         * gimple-ssa-store-merging.c: Include rtl.h and expr.h.
10207         (struct store_immediate_info): Add bitregion_start and bitregion_end
10208         fields.
10209         (store_immediate_info::store_immediate_info): Add brs and bre
10210         arguments and initialize bitregion_{start,end} from those.
10211         (struct merged_store_group): Add bitregion_start, bitregion_end,
10212         align_base and mask fields.  Drop unnecessary struct keyword from
10213         struct store_immediate_info.  Add do_merge method.
10214         (clear_bit_region_be): Use memset instead of loop storing zeros.
10215         (merged_store_group::do_merge): New method.
10216         (merged_store_group::merge_into): Use do_merge.  Allow gaps in between
10217         stores as long as the surrounding bitregions have no gaps.
10218         (merged_store_group::merge_overlapping): Use do_merge.
10219         (merged_store_group::apply_stores): Test that bitregion_{start,end}
10220         is byte aligned, rather than requiring that start and width are
10221         byte aligned.  Drop unnecessary struct keyword from
10222         struct store_immediate_info.  Allocate and populate also mask array.
10223         Make start of the arrays relative to bitregion_start rather than
10224         start and size them according to bitregion_{end,start} difference.
10225         (struct imm_store_chain_info): Drop unnecessary struct keyword from
10226         struct store_immediate_info.
10227         (pass_store_merging::gate): Punt if BITS_PER_UNIT or CHAR_BIT is not 8.
10228         (pass_store_merging::terminate_all_aliasing_chains): Drop unnecessary
10229         struct keyword from struct store_immediate_info.
10230         (imm_store_chain_info::coalesce_immediate_stores): Allow gaps in
10231         between stores as long as the surrounding bitregions have no gaps.
10232         Formatting fixes.
10233         (struct split_store): Add orig non-static data member.
10234         (split_store::split_store): Initialize orig to false.
10235         (find_constituent_stmts): Return store_immediate_info *, non-NULL
10236         if there is exactly a single original stmt.  Change stmts argument
10237         to pointer from reference, if NULL, don't push anything to it.  Add
10238         first argument, use it to optimize skipping over orig stmts that
10239         are known to be before bitpos already.  Simplify.
10240         (split_group): Return unsigned int count how many stores are or
10241         would be needed rather than a bool.  Add allow_unaligned argument.
10242         Change split_stores argument from reference to pointer, if NULL,
10243         only do a dry run computing how many stores would be produced.
10244         Rewritten algorithm to use both alignment and misalign if
10245         !allow_unaligned and handle bitfield stores with gaps.
10246         (imm_store_chain_info::output_merged_store): Set start_byte_pos
10247         from bitregion_start instead of start.  Compute allow_unaligned
10248         here, if true, do 2 split_group dry runs to compute which one
10249         produces fewer stores and prefer aligned if equal.  Punt if
10250         new count is bigger or equal than original before emitting any
10251         statements, rather than during that.  Remove no longer needed
10252         new_ssa_names tracking.  Replace num_stmts with
10253         split_stores.length ().  Use 32-bit stack allocated entries
10254         in split_stores auto_vec.  Try to reuse original store lhs/rhs1
10255         if possible.  Handle bitfields with gaps.
10256         (pass_store_merging::execute): Ignore bitsize == 0 stores.
10257         Compute bitregion_{start,end} for the stores and construct
10258         store_immediate_info with that.  Formatting fixes.
10260 2017-10-30  Uros Bizjak  <ubizjak@gmail.com>
10262         PR target/82725
10263         * config/i386/i386.c (legitimate_pic_address_disp_p): Allow
10264         UNSPEC_DTPOFF and UNSPEC_NTPOFF with SImode immediate offset.
10266 2017-10-29  Jim Wilson  <wilson@tuliptree.org>
10268         * gimplify.c: Include tm_p.h.
10270         * common.opt (gcoff): Re-add as ignored option.
10271         (gcoff1, gcoff2, gcoff3): Likewise.
10273         * Makefile.in (OBJS): Delete sdbout.o.
10274         (GTFILES): Delete $(srcdir)/sdbout.c.
10275         * debug.h: Delete sdb_debug_hooks.
10276         * final.c: Delete sdbout.h include.
10277         (final_scan_insn): Delete SDB_DEBUG check.
10278         (rest_of_clean_state): Likewise.
10279         * output.h: Delete sdb_begin_function_line.
10280         * sdbout.c: Delete.
10281         * sdbout.h: Delete.
10282         * toplev.c: Delete sdbout.h include.
10283         (process_options): Delete SDB_DEBUG check.
10284         * tree-core.h (tree_type_common): Delete pointer field of
10285         tree_type_symtab.
10286         * tree.c (copy_node): Clear TYPE_SYMTAB_DIE instead of
10287         TYPE_SYMTAB_POINTER.
10288         * tree.h (TYPE_SYMTAB_POINTER): Delete.
10289         (TYPE_SYMTAB_IS_POINTER): Delete.
10290         (TYPE_SYMTAB_IS_DIE): Renumber.
10291         * xcoffout.c: Refer to former sdbout.c file.
10292         (xcoffout_begin_prologue): Use past tense for sdbout.c reference.
10294         * doc/install.texi (--with-stabs): Delete COFF and ECOFF info.
10295         * doc/invoke.texi (SEEALSO): Delete adb and sdb references.
10296         (Debugging Options): Delete -gcoff.
10297         (-gstabs): Delete SDB reference.
10298         (-gcoff): Delete.
10299         (-gcoff@var{level}): Delete.
10300         * doc/passes.texi (Debugging information output): Delete SDB and
10301         sdbout.c references.
10302         * doc/tm.texi: Regenerate.
10303         * doc/tm.texi.in (DWARF_CIE_DATA_ALIGNMENT): Delete SDB from xref.
10304         (SDB and DWARF): Change node name to DWARF and delete SDB and COFF
10305         references.
10306         (DEBUGGER_AUTO_OFFSET): Delete COFF and SDB references.
10307         (PREFERRED_DEBUGGING_TYPE): Delete SDB_DEBUG and -gcoff references.
10308         (SDB_DEBUGGING_INFO): Delete.
10309         (PUT_SDB_@dots{}, SDB_DELIM, SDB_ALLOW_UNKNOWN_REFERENCES)
10310         SDB_ALLOW_FORWARD_REFERENCES, SDB_OUTPUT_SOURCE_LINE): Delete.
10311         * target.def (output_source_filename): Delete COFF reference.
10313         * common.opt (gcoff): Delete.
10314         (gxcoff+): Update Negative chain.
10315         * defaults.h: Delete all references to SDB_DEBUGGING_INFO and
10316         SDB_DEBUG.
10317         * dwarf2out.c (gen_array_type_die): Change SDB to debuggers.
10318         * flag-types.h (enum debug_info_type): Delete SDB_DEBUG.
10319         * function.c (number_blocks): Delete SDB_DEBUGGING_INFO, SDB_DEBUG,
10320         and SDB references.
10321         (expand_function_start): Change sdb reference to past tense.
10322         (expand_function_end): Change sdb reference to past tense.
10323         * gcc.c (cpp_unique_options): Delete gcoff3 reference.
10324         * opts.c (debug_type_names): Delete coff entry.
10325         (common_handle_option): Delete OPT_gcoff case.
10326         * system.h (SDB_DEBUG, SDB_DEBUGGING_INFO): Poison.
10328         * config/dbxcoff.h (PREFERRED_DEBUGGING_TYPE): Set to DBX_DEBUG.
10329         * config/cris/cris.h: Delete SDB reference in comment.
10330         * config/i386/cygming.h: Don't define SDB_DEBUGGING_INFO.
10331         (ASM_DECLARE_FUNCTION_NAME): Delete SDB reference from comment.
10332         * config/i386/gas.h: Don't define SDB_DEBUGGING_INFO.
10333         * config/i386/i386.c (svr4_dbx_register_map): Change SDB references
10334         to past tense.
10335         (ix86_expand_prologue): Likewise.
10336         * config/i386/winnt.c (i386_pe_start_function): Don't check SDB_DEBUG.
10337         * config/ia64/ia64.h: Likewise.
10338         * config/m68k/m68kelf.h (DBX_REGISTER_NUMBER): Delete SDB reference.
10339         * config/mips/mips.h (SUBTARGET_ASM_DEBUGGING_SPEC): Delete gcoff*
10340         support.
10341         * config/mmix/mmix.h: Likewise.
10342         * config/nds32/nds32.c: Likewise.
10343         * config/stormy/storym16.h: Likewise.
10344         * config/visium/visium.h: Likewise.
10345         * config/vx-common.h (SDB_DEBUGGING_INFO): Delete undef.
10347 2017-10-28  Sandra Loosemore  <sandra@codesourcery.com>
10349         * config/nios2/nios2.h (FRAME_GROWS_DOWNWARD): Define to 1.
10350         * config/nios2/nios2.c (nios2_initial_elimination_offset):  Make
10351         FRAME_POINTER_REGNUM point at high end of local var area.
10353 2017-10-27  Eric Botcazou  <ebotcazou@adacore.com>
10355         * bb-reorder.c (find_traces_1_round): Fix off-by-one index.
10356         Move comment around.  Do not reset best_edge for a copiable
10357         destination if the copy would cause a partition change.
10358         (better_edge_p): Remove redundant check.
10360 2017-10-27  Uros Bizjak  <ubizjak@gmail.com>
10362         * config/i386/i386-protos.h (ix86_fp_compare_mode): Remove prototype.
10364 2017-10-27  Michael Meissner  <meissner@linux.vnet.ibm.com>
10366         * builtins.c (CASE_MATHFN_FLOATN): New helper macro to add cases
10367         for math functions that have _Float<N> and _Float<N>X variants.
10368         (mathfn_built_in_2): Add support for math functions that have
10369         _Float<N> and _Float<N>X variants.
10370         (DEF_INTERNAL_FLT_FLOATN_FN): New helper macro.
10371         (expand_builtin_mathfn_ternary): Add support for fma with
10372         _Float<N> and _Float<N>X variants.
10373         (expand_builtin): Likewise.
10374         (fold_builtin_3): Likewise.
10375         * builtins.def (DEF_EXT_LIB_FLOATN_NX_BUILTINS): New macro to
10376         create math function _Float<N> and _Float<N>X variants as external
10377         library builtins.
10378         (BUILT_IN_COPYSIGN _Float<N> and _Float<N>X variants) Use
10379         DEF_EXT_LIB_FLOATN_NX_BUILTINS to make built-in functions using
10380         the __builtin_ prefix and if not strict ansi, without the prefix.
10381         (BUILT_IN_FABS _Float<N> and _Float<N>X variants): Likewise.
10382         (BUILT_IN_FMA _Float<N> and _Float<N>X variants): Likewise.
10383         (BUILT_IN_FMAX _Float<N> and _Float<N>X variants): Likewise.
10384         (BUILT_IN_FMIN _Float<N> and _Float<N>X variants): Likewise.
10385         (BUILT_IN_NAN _Float<N> and _Float<N>X variants): Likewise.
10386         (BUILT_IN_SQRT _Float<N> and _Float<N>X variants): Likewise.
10387         * builtin-types.def (BT_FN_FLOAT16_FLOAT16_FLOAT16_FLOAT16): New
10388         function signatures for fma _Float<N> and _Float<N>X variants.
10389         (BT_FN_FLOAT32_FLOAT32_FLOAT32_FLOAT32): Likewise.
10390         (BT_FN_FLOAT64_FLOAT64_FLOAT64_FLOAT64): Likewise.
10391         (BT_FN_FLOAT128_FLOAT128_FLOAT128_FLOAT128): Likewise.
10392         (BT_FN_FLOAT32X_FLOAT32X_FLOAT32X_FLOAT32X): Likewise.
10393         (BT_FN_FLOAT64X_FLOAT64X_FLOAT64X_FLOAT64X): Likewise.
10394         (BT_FN_FLOAT128X_FLOAT128X_FLOAT128X_FLOAT128X): Likewise.
10395         * gencfn-macros.c (print_case_cfn): Add support for math functions
10396         that have _Float<N> and _Float<N>X variants.
10397         (print_define_operator_list): Likewise.
10398         (fltfn_suffixes): Likewise.
10399         (main): Likewise.
10400         * internal-fn.def (DEF_INTERNAL_FLT_FLOATN_FN): New helper macro
10401         for math functions that have _Float<N> and _Float<N>X variants.
10402         (SQRT): Add support for sqrt, copysign, fmin and fmax _Float<N>
10403         and _Float<N>X variants.
10404         (COPYSIGN): Likewise.
10405         (FMIN): Likewise.
10406         (FMAX): Likewise.
10407         * fold-const.c (tree_call_nonnegative_warnv_p): Add support for
10408         copysign, fma, fmax, fmin, and sqrt _Float<N> and _Float<N>X
10409         variants.
10410         (integer_valued_read_call_p): Likewise.
10411         * fold-const-call.c (fold_const_call_ss): Likewise.
10412         (fold_const_call_sss): Add support for copysign, fmin, and fmax
10413         _Float<N> and _Float<N>X variants.
10414         (fold_const_call_ssss): Add support for fma _Float<N> and
10415         _Float<N>X variants.
10416         * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Add
10417         support for copysign and fma _Float<N> and _Float<N>X variants.
10418         (backprop::process_builtin_call_use): Likewise.
10419         * tree-call-cdce.c (can_test_argument_range); Add support for
10420         sqrt _Float<N> and _Float<N>X variants.
10421         (edom_only_function): Likewise.
10422         (get_no_error_domain): Likewise.
10423         * tree-ssa-math-opts.c (internal_fn_reciprocal): Likewise.
10424         * tree-ssa-reassoc.c (attempt_builtin_copysign): Add support for
10425         copysign _Float<N> and _Float<N>X variants.
10426         * config/rs6000/rs6000-builtin.def (SQRTF128): Delete, this is now
10427         handled by machine independent code.
10428         (FMAF128): Likewise.
10429         * doc/cpp.texi (Common Predefined Macros): Document defining
10430         __FP_FAST_FMAF<N> and __FP_FAST_FMAF<N>X if the backend supports
10431         fma _Float<N> and _Float<N>X variants.
10433 2017-10-27  Uros Bizjak  <ubizjak@gmail.com>
10435         PR target/82692
10436         * config/i386/i386-modes.def (CCFPU): Remove definition.
10437         * config/i386/i386.c (put_condition_mode): Remove CCFPU mode handling.
10438         (ix86_cc_modes_compatible): Ditto.
10439         (ix86_expand_carry_flag_compare): Ditto.
10440         (ix86_expand_int_movcc): Ditto.
10441         (ix86_expand_int_addcc): Ditto.
10442         (ix86_reverse_condition): Ditto.
10443         (ix86_unordered_fp_compare): Rename from ix86_fp_compare_mode.
10444         Return true/false for unordered/ordered fp comparisons.
10445         (ix86_cc_mode): Always return CCFPmode for float mode comparisons.
10446         (ix86_prepare_fp_compare_args): Update for rename.
10447         (ix86_expand_fp_compare): Update for rename.  Generate unordered
10448         compare RTXes wrapped with UNSPEC_NOTRAP unspec.
10449         (ix86_expand_sse_compare_and_jump): Ditto.
10450         * config/i386/predicates.md (fcmov_comparison_operator):
10451         Remove CCFPU mode handling.
10452         (ix86_comparison_operator): Ditto.
10453         (ix86_carry_flag_operator): Ditto.
10454         * config/i386/i386.md (UNSPEC_NOTRAP): New unspec.
10455         (*cmpu<mode>_i387): Wrap compare RTX with UNSPEC_NOTRAP unspec.
10456         (*cmpu<mode>_cc_i387): Ditto.
10457         (FPCMP): Remove mode iterator.
10458         (unord): Remove mode attribute.
10459         (unord_subst): New define_subst transformation
10460         (unord): New define_subst attribute.
10461         (unordered): Ditto.
10462         (*cmpi<unord><MODEF:mode>): Rewrite using unord_subst transformation.
10463         (*cmpi<unord>xf_i387): Ditto.
10464         * config/i386/sse.md (<sse>_<unord>comi<round_saeonly_name>): Merge
10465         from <sse>_comi<round_saeonly_name> and <sse>_ucomi<round_saeonly_name>
10466         using unord_subst transformation.
10467         * config/i386/subst.md (SUBST_A): Remove CCFP and CCFPU modes.
10468         (round_saeonly): Also handle CCFP mode.
10469         * reg-stack.c (subst_stack_regs_pat): Handle UNSPEC_NOTRAP unspec.
10470         Remove UNSPEC_SAHF unspec handling.
10472 2017-10-27  Jan Hubicka  <hubicka@ucw.cz>
10474         * x86-tune.def (X86_TUNE_INTER_UNIT_MOVES_TO_VEC): Disable for Zen.
10476 2017-10-27  Jeff Law  <law@redhat.com>
10478         * gimple-ssa-sprintf.c: Include domwalk.h.
10479         (class sprintf_dom_walker): New class, derived from dom_walker.
10480         (sprintf_dom_walker::before_dom_children): New function.
10481         (struct call_info): Moved into sprintf_dom_walker class
10482         (compute_formath_length, handle_gimple_call): Likewise.
10483         (sprintf_length::execute): Call the dominator walker rather
10484         than walking the statements.
10486         * tree-vrp.c (check_all_array_refs): Do not use wi->info to smuggle
10487         gimple statement locations.
10488         (check_array_bounds): Corresponding changes.  Get the statement's
10489         location directly from wi->stmt.
10491 2017-10-27  Palmer Dabbelt  <palmer@dabbelt.com>
10493         PR target/82717
10494         * doc/invoke.texi (RISC-V) <-mabi>: Correct and improve.
10496 2017-10-27  Jan Hubicka  <hubicka@ucw.cz>
10498         * config/i386/x86-tune.def (X86_TUNE_PARTIAL_REG_DEPENDENCY,
10499         X86_TUNE_MOVX): Disable for Haswell and newer CPUs.
10501 2017-10-27  Jakub Jelinek  <jakub@redhat.com>
10503         PR target/82703
10504         * config/i386/i386-protos.h (maybe_get_pool_constant): Removed.
10505         * config/i386/i386.c (maybe_get_pool_constant): Removed.
10506         (ix86_split_to_parts): Use avoid_constant_pool_reference instead of
10507         maybe_get_pool_constant.
10508         * config/i386/predicates.md (zero_extended_scalar_load_operand):
10509         Likewise.
10511 2017-10-27  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
10513         * doc/install.texi (Specific, i?86-*-solaris2.10): Simplify gas
10514         2.26 caveat.  Update gas and gld versions.
10515         (Specific, *-*-solaris2*): Update binutils version.  Remove caveat
10516         reference.
10518 2017-10-27  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
10520         * cgraph.h (set_malloc_flag): Declare.
10521         * cgraph.c (set_malloc_flag_1): New function.
10522         (set_malloc_flag): Likewise.
10523         * ipa-fnsummary.h (ipa_call_summary): Add new field is_return_callee.
10524         * ipa-fnsummary.c (ipa_call_summary::reset): Set is_return_callee to
10525         false.
10526         (read_ipa_call_summary): Add support for reading is_return_callee.
10527         (write_ipa_call_summary): Stream is_return_callee.
10528         * ipa-inline.c (ipa_inline): Remove call to ipa_free_fn_summary.
10529         * ipa-pure-const.c: Add headers ssa.h, alloc-pool.h, symbol-summary.h,
10530         ipa-prop.h, ipa-fnsummary.h.
10531         (pure_const_names): Change to static.
10532         (malloc_state_e): Define.
10533         (malloc_state_names): Define.
10534         (funct_state_d): Add field malloc_state.
10535         (varying_state): Set malloc_state to STATE_MALLOC_BOTTOM.
10536         (check_retval_uses): New function.
10537         (malloc_candidate_p): Likewise.
10538         (analyze_function): Add support for malloc attribute.
10539         (pure_const_write_summary): Stream malloc_state.
10540         (pure_const_read_summary): Add support for reading malloc_state.
10541         (dump_malloc_lattice): New function.
10542         (propagate_malloc): New function.
10543         (warn_function_malloc): New function.
10544         (ipa_pure_const::execute): Call propagate_malloc and
10545         ipa_free_fn_summary.
10546         (pass_local_pure_const::execute): Add support for malloc attribute.
10547         * ssa-iterators.h (RETURN_FROM_IMM_USE_STMT): New macro.
10548         * doc/invoke.texi: Document Wsuggest-attribute=malloc.
10550 2017-10-27  Martin Liska  <mliska@suse.cz>
10552         PR gcov-profile/82457
10553         * doc/invoke.texi: Document that one needs a non-strict ISO mode
10554         for fork-like functions to be properly instrumented.
10556 2017-10-27  Richard Biener  <rguenther@suse.de>
10558         PR middle-end/81659
10559         * tree-eh.c (pass_lower_eh_dispatch::execute): Free dominator
10560         info when we redirected EH.
10562 2017-10-26  Michael Collison  <michael.collison@arm.com>
10564         * config/aarch64/aarch64.md(<optab>_trunc><vf><GPI:mode>2):
10565         New pattern.
10566         (<optab>_trunchf<GPI:mode>2: New pattern.
10567         (<optab>_trunc<vgp><GPI:mode>2: New pattern.
10568         * config/aarch64/iterators.md (wv): New mode attribute.
10569         (vf, VF): New mode attributes.
10570         (vgp, VGP): New mode attributes.
10571         (s): Update attribute with SImode and DImode prefixes.
10573 2017-10-26  Sandra Loosemore  <sandra@codesourcery.com>
10575         * config/nios2/constraints.md ("S"): Match r0rel_constant_p too.
10576         * config/nios2/nios2-protos.h (r0rel_constant_p): Declare.
10577         * config/nios2/nios2.c: (nios2_r0rel_sec_regex): New.
10578         (nios2_option_overide): Initialize it.  Don't allow R0-relative
10579         addressing with PIC.
10580         (nios2_rtx_costs): Handle r0rel_constant_p like gprel_constant_p.
10581         (nios2_symbolic_constant_p): Likewise.
10582         (nios2_legitimate_address_p): Likewise.
10583         (nios2_r0rel_section_name_p): New.
10584         (nios2_symbol_ref_in_r0rel_data_p): New.
10585         (nios2_emit_move_sequence): Handle r0rel_constant_p.
10586         (r0rel_constant_p): New.
10587         (nios2_print_operand_address): Handle r0rel_constant_p.
10588         (nios2_cdx_narrow_form_p): Likewise.
10589         * config/nios2/nios2.opt (mr0rel-sec=): New option.
10590         * doc/invoke.texi (Option Summary): Add -mr0rel-sec.
10591         (Nios II Options): Document -mr0rel-sec.
10593 2017-10-26  Sandra Loosemore  <sandra@codesourcery.com>
10595         * config/nios2/nios2.c: Include xregex.h.
10596         (nios2_gprel_sec_regex): New.
10597         (nios2_option_overide): Initialize it.  Don't allow GP-relative
10598         addressing with PIC.
10599         (nios2_small_section_name_p): Check for regex match.
10600         * config/nios2/nios2.opt (mgprel-sec=): New option.
10601         * doc/invoke.texi (Option Summary): Add -mgprel-sec.
10602         (Nios II Options): Document -mgprel-sec.
10604 2017-10-26  Jim Wilson  <wilson@tuliptree.org>
10606         * doc/invoke.texi (-fdebug-prefix-map): Expand documentation.
10608 2017-10-26  Tom de Vries  <tom@codesourcery.com>
10610         PR tree-optimization/82707
10611         * gimple.c (gimple_copy): Fix unsharing of
10612         GIMPLE_OMP_{SINGLE,TARGET,TEAMS}.
10614 2017-10-26  Olga Makhotina  <olga.makhotina@intel.com>
10616         * config/i386/avx512fintrin.h (_mm512_cmpeq_pd_mask,
10617         _mm512_cmple_pd_mask, _mm512_cmplt_pd_mask,
10618         _mm512_cmpneq_pd_mask, _mm512_cmpnle_pd_mask,
10619         _mm512_cmpnlt_pd_mask, _mm512_cmpord_pd_mask,
10620         _mm512_cmpunord_pd_mask, _mm512_mask_cmpeq_pd_mask,
10621         _mm512_mask_cmple_pd_mask, _mm512_mask_cmplt_pd_mask,
10622         _mm512_mask_cmpneq_pd_mask, _mm512_mask_cmpnle_pd_mask,
10623         _mm512_mask_cmpnlt_pd_mask, _mm512_mask_cmpord_pd_mask,
10624         _mm512_mask_cmpunord_pd_mask, _mm512_cmpeq_ps_mask,
10625         _mm512_cmple_ps_mask, _mm512_cmplt_ps_mask,
10626         _mm512_cmpneq_ps_mask, _mm512_cmpnle_ps_mask,
10627         _mm512_cmpnlt_ps_mask, _mm512_cmpord_ps_mask,
10628         _mm512_cmpunord_ps_mask, _mm512_mask_cmpeq_ps_mask,
10629         _mm512_mask_cmple_ps_mask, _mm512_mask_cmplt_ps_mask,
10630         _mm512_mask_cmpneq_ps_mask, _mm512_mask_cmpnle_ps_mask,
10631         _mm512_mask_cmpnlt_ps_mask, _mm512_mask_cmpord_ps_mask,
10632         _mm512_mask_cmpunord_ps_mask): New intrinsics.
10634 2017-10-26  Michael Meissner  <meissner@linux.vnet.ibm.com>
10636         * config/rs6000/aix.h (TARGET_IEEEQUAD_DEFAULT): Set long double
10637         default to IBM.
10638         * config/rs6000/darwin.h (TARGET_IEEEQUAD_DEFAULT): Likewise.
10639         * config/rs6000/rs6000.opt (-mabi=ieeelongdouble): Move the
10640         warning to rs6000.c.  Remove the Undocumented flag, since it has
10641         been documented.
10642         (-mabi=ibmlongdouble): Likewise.
10643         * config/rs6000/rs6000.c (TARGET_IEEEQUAD_DEFAULT): If it is not
10644         already set, set the default format for long double.
10645         (rs6000_debug_reg_global): Print whether long double is IBM or
10646         IEEE.
10647         (rs6000_option_override_internal): Rework setting long double
10648         format.  Only warn if the user is changing the long double default
10649         and they did not use -Wno-psabi.
10650         * doc/invoke.texi (PowerPC options): Update the documentation for
10651         -mabi=ieeelongdouble and -mabi=ibmlongdouble.
10653 2017-10-26  Richard Sandiford  <richard.sandiford@linaro.org>
10654             Alan Hayward  <alan.hayward@arm.com>
10655             David Sherwood  <david.sherwood@arm.com>
10657         * rtl.h (wider_subreg_mode): New function.
10658         * ira.h (ira_sort_regnos_for_alter_reg): Take a machine_mode *
10659         rather than an unsigned int *.
10660         * ira-color.c (regno_max_ref_width): Replace with...
10661         (regno_max_ref_mode): ...this new variable.
10662         (coalesced_pseudo_reg_slot_compare): Update accordingly.
10663         Use wider_subreg_mode.
10664         (ira_sort_regnos_for_alter_reg): Likewise.  Take a machine_mode *
10665         rather than an unsigned int *.
10666         * lra-constraints.c (uses_hard_regs_p): Use wider_subreg_mode.
10667         (process_alt_operands): Likewise.
10668         (invariant_p): Likewise.
10669         * lra-spills.c (assign_mem_slot): Likewise.
10670         (add_pseudo_to_slot): Likewise.
10671         * lra.c (collect_non_operand_hard_regs): Likewise.
10672         (add_regs_to_insn_regno_info): Likewise.
10673         * reload1.c (regno_max_ref_width): Replace with...
10674         (regno_max_ref_mode): ...this new variable.
10675         (reload): Update accordingly.  Update call to
10676         ira_sort_regnos_for_alter_reg.
10677         (alter_reg): Update to use regno_max_ref_mode.  Call wider_subreg_mode.
10678         (init_eliminable_invariants): Update to use regno_max_ref_mode.
10679         (scan_paradoxical_subregs): Likewise.
10681 2017-10-26  Wilco Dijkstra  <wdijkstr@arm.com>
10683         * config/aarch64/aarch64.h (EXIT_IGNORE_STACK): Set if alloca is used.
10684         (aarch64_frame): Add emit_frame_chain boolean.
10685         * config/aarch64/aarch64.c (aarch64_frame_pointer_required)
10686         Move eh_return case to aarch64_layout_frame.
10687         (aarch64_layout_frame): Initialize emit_frame_chain.
10688         (aarch64_expand_prologue): Use emit_frame_chain.
10690 2017-10-26  Wilco Dijkstra  <wdijkstr@arm.com>
10692         * config/aarch64/aarch64.c (aarch64_layout_frame):
10693         Ensure LR is always stored at the bottom of the callee-saves.
10694         Remove rarely used frame layout which saves callee-saves at top of
10695         frame, so the store of LR can be used as a valid probe in all cases.
10697 2017-10-26  Wilco Dijkstra  <wdijkstr@arm.com>
10699         * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
10700         Improve unaligned TImode/TFmode base/offset split.
10702 2017-10-26  Richard Sandiford  <richard.sandiford@linaro.org>
10703             Alan Hayward  <alan.hayward@arm.com>
10704             David Sherwood  <david.sherwood@arm.com>
10706         * caller-save.c (mark_referenced_regs):  Use read_modify_subreg_p.
10707         * combine.c (find_single_use_1): Likewise.
10708         (expand_field_assignment): Likewise.
10709         (move_deaths): Likewise.
10710         * lra-constraints.c (simplify_operand_subreg): Likewise.
10711         (curr_insn_transform): Likewise.
10712         * lra.c (collect_non_operand_hard_regs): Likewise.
10713         (add_regs_to_insn_regno_info): Likewise.
10714         * rtlanal.c (reg_referenced_p): Likewise.
10715         (covers_regno_no_parallel_p): Likewise.
10717 2017-10-26  Richard Sandiford  <richard.sandiford@linaro.org>
10719         * wide-int-print.cc (print_hex): Loop based on extract_uhwi.
10720         Don't print any bits outside the precision of the value.
10721         * wide-int.cc (test_printing): Add some new tests.
10723 2017-10-26  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
10725         * configure.ac (gcc_cv_as_ix86_xbrace_comment): Check if assembler
10726         supports -xbrace_comment option.
10727         * configure: Regenerate.
10728         * config.in: Regenerate.
10729         * config/i386/sol2.h (ASM_XBRACE_COMMENT_SPEC): Define.
10730         (ASM_CPU_SPEC): Use it.
10732 2017-10-26  Richard Sandiford  <richard.sandiford@linaro.org>
10734         * target.def (static_rtx_alignment): New hook.
10735         * targhooks.h (default_static_rtx_alignment): Declare.
10736         * targhooks.c (default_static_rtx_alignment): New function.
10737         * doc/tm.texi.in (TARGET_STATIC_RTX_ALIGNMENT): New hook.
10738         * doc/tm.texi: Regenerate.
10739         * varasm.c (force_const_mem): Use targetm.static_rtx_alignment
10740         instead of targetm.constant_alignment.  Remove call to
10741         set_mem_attributes.
10742         * config/cris/cris.c (TARGET_STATIC_RTX_ALIGNMENT): Redefine.
10743         (cris_preferred_mininum_alignment): New function, split out from...
10744         (cris_constant_alignment): ...here.
10745         (cris_static_rtx_alignment): New function.
10746         * config/i386/i386.c (ix86_static_rtx_alignment): New function,
10747         split out from...
10748         (ix86_constant_alignment): ...here.
10749         (TARGET_STATIC_RTX_ALIGNMENT): Redefine.
10750         * config/mmix/mmix.c (TARGET_STATIC_RTX_ALIGNMENT): Redefine.
10751         (mmix_static_rtx_alignment): New function.
10752         * config/spu/spu.c (spu_static_rtx_alignment): New function.
10753         (TARGET_STATIC_RTX_ALIGNMENT): Redefine.
10755 2017-10-26  Tamar Christina  <tamar.christina@arm.com>
10757         PR target/81800
10758         * config/aarch64/aarch64.md (lrint<GPF:mode><GPI:mode>2):
10759         Add flag_trapping_math and flag_fp_int_builtin_inexact.
10761 2017-10-25  Palmer Dabbelt  <palmer@dabbelt.com>
10763         * config/riscv/riscv.md (ZERO_EXTEND_LOAD): Define.
10764         * config/riscv/pic.md (local_pic_load): Rename to local_pic_load_s,
10765         mark as a sign-extending load.
10766         (local_pic_load_u): Define.
10768 2017-10-25  Eric Botcazou  <ebotcazou@adacore.com>
10770         PR middle-end/82062
10771         * fold-const.c (operand_equal_for_comparison_p): Also return true
10772         if ARG0 is a simple variant of ARG1 with narrower precision.
10773         (fold_ternary_loc): Always pass unstripped operands to the predicate.
10775 2017-10-25  Jan Hubicka  <hubicka@ucw.cz>
10777         * i386.c (ix86_builtin_vectorization_cost): Compute scatter/gather
10778         cost correctly.
10779         * i386.h (processor_costs): Add gather_static, gather_per_elt,
10780         scatter_static, scatter_per_elt.
10781         * x86-tune-costs.h: Add new cost entries.
10783 2017-10-25  Richard Biener  <rguenther@suse.de>
10785         * tree-ssa-sccvn.h (vn_eliminate): Declare.
10786         * tree-ssa-pre.c (class eliminate_dom_walker, eliminate,
10787         class pass_fre): Move to ...
10788         * tree-ssa-sccvn.c (class eliminate_dom_walker, vn_eliminate,
10789         class pass_fre): ... here and adjust for statistics.
10791 2017-10-25  Jakub Jelinek  <jakub@redhat.com>
10793         PR libstdc++/81706
10794         * attribs.c (attribute_value_equal): Use omp_declare_simd_clauses_equal
10795         for comparison of OMP_CLAUSEs regardless of flag_openmp{,_simd}.
10796         (duplicate_one_attribute, copy_attributes_to_builtin): New functions.
10797         * attribs.h (duplicate_one_attribute, copy_attributes_to_builtin): New
10798         declarations.
10800 2017-10-25  Richard Biener  <rguenther@suse.de>
10802         * tree-ssa-pre.c (need_eh_cleanup, need_ab_cleanup, el_to_remove,
10803         el_to_fixup, el_todo, el_avail, el_avail_stack, eliminate_avail,
10804         eliminate_push_avail, eliminate_insert): Move inside...
10805         (class eliminate_dom_walker): ... this class in preparation
10806         of move.
10807         (fini_eliminate): Remove by merging with ...
10808         (eliminate): ... this function.  Adjust for class changes.
10809         (pass_pre::execute): Remove fini_eliminate call.
10810         (pass_fre::execute): Likewise.
10812 2017-10-24  Jakub Jelinek  <jakub@redhat.com>
10814         PR target/82460
10815         * config/i386/sse.md (UNSPEC_VPERMI2, UNSPEC_VPERMI2_MASK): Remove.
10816         (VPERMI2, VPERMI2I): New mode iterators.
10817         (<avx512>_vpermi2var<mode>3_maskz): Remove 3 define_expand patterns.
10818         (<avx512>_vpermi2var<mode>3<sd_maskz_name>): Remove 3 define_insn
10819         patterns.
10820         (<avx512>_vpermi2var<mode>3_mask): New define_expand using VPERMI2
10821         mode iterator.  Remove 3 old define_insn patterns.
10822         (*<avx512>_vpermi2var<mode>3_mask): 2 new define_insn patterns.
10823         (<avx512>_vpermt2var<mode>3_maskz): Adjust 1 define_expand to use
10824         VPERMI2 mode iterator, remove the other two expanders.
10825         (<avx512>_vpermt2var<mode>3<sd_maskz_name>): Adjust 1 define_insn
10826         to use VPERMI2 mode iterator, add another alternative for vpermi2*
10827         instructions, remove the other two patterns.
10828         (<avx512>_vpermt2var<mode>3_mask): Adjust 1 define_insn to use VPERMI2
10829         mode iterator, remove the other two patterns.
10830         * config/i386/i386.c (ix86_expand_vec_perm_vpermi2): Renamed to ...
10831         (ix86_expand_vec_perm_vpermt2): ... this.  Swap mask and op0
10832         arguments, use gen_*vpermt2* expanders instead of gen_*vpermi2*
10833         and adjust argument order accordingly.
10834         (ix86_expand_vec_perm): Adjust caller.
10835         (expand_vec_perm_1): Likewise.
10836         (expand_vec_perm_vpermi2_vpshub2): Rename to ...
10837         (expand_vec_perm_vpermt2_vpshub2): ... this.
10838         (ix86_expand_vec_perm_const_1): Adjust caller.
10839         (ix86_vectorize_vec_perm_const_ok): Adjust comments.
10841         PR target/82370
10842         * config/i386/sse.md (VIMAX_AVX2): Remove V4TImode.
10843         (VIMAX_AVX2_AVX512BW, VIMAX_AVX512VL): New mode iterators.
10844         (vec_shl_<mode>): Remove unused expander.
10845         (avx512bw_<shift_insn><mode>3): New define_insn.
10846         (<sse2_avx2>_ashl<mode>3, <sse2_avx2>_lshr<mode>3): Replaced by ...
10847         (<sse2_avx2>_<shift_insn><mode>3): ... this.  New define_insn.
10849 2017-10-24  Paolo Carlini  <paolo.carlini@oracle.com>
10851         PR c++/82466
10852         * doc/invoke.texi ([Wbuiltin-declaration-mismatch]): Extend
10853         description.
10855 2017-10-24  Wilco Dijkstra  <wdijkstr@arm.com>
10857         PR rtl-optimization/82396
10858         * gcc/haifa-sched.c (ready_sort_real): Remove qsort workaround.
10859         (autopref_multipass_init): Simplify initialization.
10860         (autopref_rank_data): Simplify sort order.
10861         * gcc/sched-int.h (autopref_multipass_data_): Remove
10862         multi_mem_insn_p, min_offset and max_offset.
10864 2017-10-24  Wilco Dijkstra  <wdijkstr@arm.com>
10866         PR middle-end/60580
10867         * config/aarch64/aarch64.c (aarch64_frame_pointer_required)
10868         Check special value of flag_omit_frame_pointer.
10869         (aarch64_can_eliminate): Likewise.
10870         (aarch64_override_options_after_change_1): Simplify handling of
10871         -fomit-frame-pointer and -fomit-leaf-frame-pointer.
10873 2017-10-24  Richard Biener  <rguenther@suse.de>
10875         PR tree-optimization/82697
10876         * tree-ssa-phiopt.c (cond_store_replacement): Use alias-set
10877         zero for conditional load and unconditional store.
10879 2017-10-24  H.J. Lu  <hongjiu.lu@intel.com>
10881         * doc/install.texi: Document bootstrap-cet.
10883 2017-10-24  H.J. Lu  <hongjiu.lu@intel.com>
10885         PR target/82659
10886         * config/i386/i386.c (rest_of_insert_endbranch): Don't insert
10887         ENDBR instruction at function entrance if function is only
10888         called directly.
10890 2017-10-24  Jakub Jelinek  <jakub@redhat.com>
10892         PR target/82628
10893         * config/i386/i386.md (addcarry<mode>, subborrow<mode>): Change
10894         patterns to better describe from which operation the CF is computed.
10895         (addcarry<mode>_0, subborrow<mode>_0): New patterns.
10896         * config/i386/i386.c (ix86_expand_builtin) <case handlecarry>: Pass
10897         one LTU with [DT]Imode and another one with [SD]Imode.  If arg0
10898         is 0, use _0 suffixed expanders instead of emitting a comparison
10899         before it.
10901 2017-10-06  Sergey Shalnov  <Sergey.Shalnov@intel.com>
10903         * config/i386/i386.md(*movsf_internal, *movdf_internal):
10904         Avoid 512-bit AVX modes for TARGET_PREFER_AVX256.
10906 2017-10-24  Eric Botcazou  <ebotcazou@adacore.com>
10908         PR middle-end/82569
10909         * tree-outof-ssa.h (always_initialized_rtx_for_ssa_name_p): Delete.
10910         * expr.c (expand_expr_real_1) <expand_decl_rtl>: Revert latest change.
10911         * loop-iv.c (iv_get_reaching_def): Likewise.
10912         * cfgexpand.c (expand_one_ssa_partition): Initialize the RTX if the
10913         variable is promoted and the partition contains undefined values.
10915 2017-10-23  Sandra Loosemore  <sandra@codesourcery.com>
10917         * config/nios2/nios2.c (nios2_rtx_costs): Make costs better
10918         reflect reality.
10919         (nios2_address_cost): Define.
10920         (nios2_legitimize_address): Recognize (exp + constant) directly.
10921         (TARGET_ADDRESS_COST): Define.
10923 2017-10-23  Sandra Loosemore  <sandra@codesourcery.com>
10925         * config/nios2/nios2-protos.h (nios2_large_constant_p): Declare.
10926         (nios2_symbolic_memory_operand_p): Declare.
10927         (nios2_split_large_constant): Declare.
10928         (nios2_split_symbolic_memory_operand): Declare.
10929         * config/nios2/nios2.c: Adjust includes.
10930         (nios2_symbolic_constant_allowed): New.
10931         (nios2_symbolic_constant_p): New.
10932         (nios2_plus_symbolic_constant_p): New.
10933         (nios2_valid_addr_expr_p): Recognize addresses involving
10934         symbolic constants.
10935         (nios2_legitimate_address_p): Likewise, also LO_SUM.
10936         (nios2_symbolic_memory_operand_p): New.
10937         (nios2_large_constant_p): New.
10938         (nios2_split_large_constant): New.
10939         (nios2_split_plus_large_constant): New.
10940         (nios2_split_symbolic_memory_operand): New.
10941         (nios2_legitimize_address): Code refactoring.  Handle addresses
10942         involving symbolic constants.
10943         (nios2_emit_move_sequence): Likewise.
10944         (nios2_print_operand): Improve error output.
10945         (nios2_print_operand_address): Handle LO_SUM.
10946         (nios2_cdx_narrow_form_p): Likewise.
10947         * config/nios2/nios2.md (movqi_internal): Add splitter for memory
10948         operands involving symbolic constants.
10949         (movhi_internal, movsi_internal): Likewise.
10950         (zero_extendhisi2, zero_extendqi<mode>2): Likewise.
10951         (extendhisi2, extendqi<mode>2): Likewise.
10953 2017-10-23  Sandra Loosemore  <sandra@codesourcery.com>
10955         * tree-pass.h (PROP_rtl_split_insns): Define.
10956         * recog.c (pass_data_split_all_insns): Provide PROP_rtl_split_insns.
10958 2017-10-23  Sandra Loosemore  <sandra@codesourcery.com>
10960         * config/nios2/nios2.c (TARGET_LRA_P): Don't override.
10962 2017-10-23  Jakub Jelinek  <jakub@redhat.com>
10964         PR debug/82630
10965         * target.def (const_not_ok_for_debug_p): Default to
10966         default_const_not_ok_for_debug_p instead of hook_bool_rtx_false.
10967         * targhooks.h (default_const_not_ok_for_debug_p): New declaration.
10968         * targhooks.c (default_const_not_ok_for_debug_p): New function.
10969         * dwarf2out.c (const_ok_for_output_1): Only reject UNSPECs for
10970         which targetm.const_not_ok_for_debug_p returned true.
10971         * config/arm/arm.c (arm_const_not_ok_for_debug_p): Return true
10972         for UNSPECs.
10973         * config/powerpcspe/powerpcspe.c (rs6000_const_not_ok_for_debug_p):
10974         Likewise.
10975         * config/rs6000/rs6000.c (rs6000_const_not_ok_for_debug_p): Likewise.
10976         * config/i386/i386.c (ix86_delegitimize_address_1): Don't delegitimize
10977         UNSPEC_GOTOFF with addend into addend - _GLOBAL_OFFSET_TABLE_ + symbol
10978         if !base_term_p.
10979         (ix86_const_not_ok_for_debug_p): New function.
10980         (i386_asm_output_addr_const_extra): Handle UNSPEC_GOTOFF.
10981         (TARGET_CONST_NOT_OK_FOR_DEBUG_P): Redefine.
10983 2017-10-23  David Malcolm  <dmalcolm@redhat.com>
10985         PR bootstrap/82610
10986         * system.h: Conditionally include "unique-ptr.h" if
10987         INCLUDE_UNIQUE_PTR is defined.
10988         * unique-ptr-tests.cc: Remove include of "unique-ptr.h" in favor
10989         of defining INCLUDE_UNIQUE_PTR before including "system.h".
10991 2017-10-23  Sebastian Perta  <sebastian.perta@renesas.com>
10993         * config/rl78/rl78.md: New define_expand "subdi3".
10995 2017-10-23  H.J. Lu  <hongjiu.lu@intel.com>
10997         PR target/82673
10998         * config/i386/i386.c (ix86_finalize_stack_frame_flags): Skip
10999         DF_REF_INSN if DF_REF_INSN_INFO is false.
11001 2017-10-23  Jan Hubicka  <hubicka@ucw.cz>
11003         * i386.c (dimode_scalar_chain::compute_convert_gain): Use
11004         xmm_move instead of sse_move.
11005         (sse_store_index): New function.
11006         (ix86_register_move_cost): Be more sensible about mismatch stall;
11007         model AVX moves correctly; make difference between sse->integer and
11008         integer->sse.
11009         (ix86_builtin_vectorization_cost): Model correctly aligned and unaligned
11010         moves; make difference between SSE and AVX.
11011         * i386.h (processor_costs): Remove sse_move; add xmm_move, ymm_move
11012         and zmm_move. Increase size of sse load and store tables;
11013         add unaligned load and store tables; add ssemmx_to_integer.
11014         * x86-tune-costs.h: Update all entries according to real
11015         move latencies from Agner Fog's manual and chip documentation.
11017 2017-10-23  Jakub Jelinek  <jakub@redhat.com>
11019         PR target/82628
11020         * config/i386/predicates.md (x86_64_dwzext_immediate_operand): New.
11021         * config/i386/constraints.md (Wf): New constraint.
11022         * config/i386/i386.md (UNSPEC_SBB): New unspec.
11023         (cmp<dwi>_doubleword): Removed.
11024         (sub<mode>3_carry_ccc, *sub<mode>3_carry_ccc_1): New patterns.
11025         (sub<mode>3_carry_ccgz): Use unspec instead of compare.
11026         * config/i386/i386.c (ix86_expand_branch) <case E_TImode>: Don't
11027         expand with cmp<dwi>_doubleword.  For LTU and GEU use
11028         sub<mode>3_carry_ccc instead of sub<mode>3_carry_ccgz and use CCCmode.
11030         * common.opt (gcolumn-info): Enable by default.
11031         * doc/invoke.texi (gcolumn-info): Document new default.
11033 2017-10-23  Richard Biener  <rguenther@suse.de>
11035         PR tree-optimization/82672
11036         * graphite-isl-ast-to-gimple.c (graphite_copy_stmts_from_block):
11037         Fold the stmt if we propagated into it.
11039 2017-10-23  Richard Biener  <rguenther@suse.de>
11041         * tree-ssa-pre.c (bitmap_remove_from_set): Rename to...
11042         (bitmap_remove_expr_from_set): ... this.  All callers call this
11043         for non-constant values.
11044         (bitmap_set_subtract): Rename to...
11045         (bitmap_set_subtract_expressions): ... this.  Adjust and
11046         optimize.
11047         (bitmap_set_contains_value): Remove superfluous check.
11048         (bitmap_set_replace_value): Inline into single caller ...
11049         (bitmap_value_replace_in_set): ... here and simplify.
11050         (dependent_clean): Merge into ...
11051         (clean): ... this using an overload.  Adjust.
11052         (prune_clobbered_mems): Adjust.
11053         (compute_antic_aux): Likewise.
11054         (compute_partial_antic_aux): Likewise.
11056 2017-10-23  Richard Biener  <rguenther@suse.de>
11058         PR tree-optimization/82129
11059         Revert
11060         2017-08-01  Richard Biener  <rguenther@suse.de>
11062         PR tree-optimization/81181
11063         * tree-ssa-pre.c (compute_antic_aux): Defer clean() to ...
11064         (compute_antic): ... end of iteration here.
11066 2017-10-23  Richard Sandiford  <richard.sandiford@linaro.org>
11068         * target.def (starting_frame_offset): New hook.
11069         * doc/tm.texi (STARTING_FRAME_OFFSET): Remove in favor of...
11070         (TARGET_STARTING_FRAME_OFFSET): ...this new hook.
11071         * doc/tm.texi.in: Regenerate.
11072         * hooks.h (hook_hwi_void_0): Declare.
11073         * hooks.c (hook_hwi_void_0): New function.
11074         * doc/rtl.texi: Refer to TARGET_STARTING_FRAME_OFFSET instead of
11075         STARTING_FRAME_OFFSET.
11076         * builtins.c (expand_builtin_setjmp_receiver): Likewise.
11077         * reload1.c (reload): Likewise.
11078         * cfgexpand.c (expand_used_vars): Use targetm.starting_frame_offset
11079         instead of STARTING_FRAME_OFFSET.
11080         * function.c (try_fit_stack_local): Likewise.
11081         (assign_stack_local_1): Likewise
11082         (instantiate_virtual_regs): Likewise.
11083         * rtlanal.c (rtx_addr_can_trap_p_1): Likewise.
11084         * config/avr/avr.md (nonlocal_goto_receiver): Likewise.
11085         * config/aarch64/aarch64.h (STARTING_FRAME_OFFSET): Delete.
11086         * config/alpha/alpha.h (STARTING_FRAME_OFFSET): Likewise.
11087         * config/arc/arc.h (STARTING_FRAME_OFFSET): Likewise.
11088         * config/arm/arm.h (STARTING_FRAME_OFFSET): Likewise.
11089         * config/bfin/bfin.h (STARTING_FRAME_OFFSET): Likewise.
11090         * config/c6x/c6x.h (STARTING_FRAME_OFFSET): Likewise.
11091         * config/cr16/cr16.h (STARTING_FRAME_OFFSET): Likewise.
11092         * config/cris/cris.h (STARTING_FRAME_OFFSET): Likewise.
11093         * config/fr30/fr30.h (STARTING_FRAME_OFFSET): Likewise.
11094         * config/frv/frv.h (STARTING_FRAME_OFFSET): Likewise.
11095         * config/ft32/ft32.h (STARTING_FRAME_OFFSET): Likewise.
11096         * config/h8300/h8300.h (STARTING_FRAME_OFFSET): Likewise.
11097         * config/i386/i386.h (STARTING_FRAME_OFFSET): Likewise.
11098         * config/ia64/ia64.h (STARTING_FRAME_OFFSET): Likewise.
11099         * config/m32c/m32c.h (STARTING_FRAME_OFFSET): Likewise.
11100         * config/m68k/m68k.h (STARTING_FRAME_OFFSET): Likewise.
11101         * config/mcore/mcore.h (STARTING_FRAME_OFFSET): Likewise.
11102         * config/mn10300/mn10300.h (STARTING_FRAME_OFFSET): Likewise.
11103         * config/moxie/moxie.h (STARTING_FRAME_OFFSET): Likewise.
11104         * config/msp430/msp430.h (STARTING_FRAME_OFFSET): Likewise.
11105         * config/nds32/nds32.h (STARTING_FRAME_OFFSET): Likewise.
11106         * config/nios2/nios2.h (STARTING_FRAME_OFFSET): Likewise.
11107         * config/nvptx/nvptx.h (STARTING_FRAME_OFFSET): Likewise.
11108         * config/pdp11/pdp11.h (STARTING_FRAME_OFFSET): Likewise.
11109         * config/riscv/riscv.h (STARTING_FRAME_OFFSET): Likewise.
11110         * config/rl78/rl78.h (STARTING_FRAME_OFFSET): Likewise.
11111         * config/rx/rx.h (STARTING_FRAME_OFFSET): Likewise.
11112         * config/s390/s390.h (STARTING_FRAME_OFFSET): Likewise.
11113         * config/sh/sh.h (STARTING_FRAME_OFFSET): Likewise.
11114         * config/sparc/sparc.c (sparc_compute_frame_size): Likewise.
11115         * config/sparc/sparc.h (STARTING_FRAME_OFFSET): Likewise.
11116         * config/spu/spu.h (STARTING_FRAME_OFFSET): Likewise.
11117         * config/stormy16/stormy16.h (STARTING_FRAME_OFFSET): Likewise.
11118         * config/tilegx/tilegx.h (STARTING_FRAME_OFFSET): Likewise.
11119         * config/tilepro/tilepro.h (STARTING_FRAME_OFFSET): Likewise.
11120         * config/v850/v850.h (STARTING_FRAME_OFFSET): Likewise.
11121         * config/visium/visium.h (STARTING_FRAME_OFFSET): Likewise.
11122         * config/avr/avr.h (STARTING_FRAME_OFFSET): Likewise.
11123         * config/avr/avr-protos.h (avr_starting_frame_offset): Likewise.
11124         * config/avr/avr.c (avr_starting_frame_offset): Make static and
11125         return a HOST_WIDE_INT.
11126         (avr_builtin_setjmp_frame_value): Use it instead of
11127         STARTING_FRAME_OFFSET.
11128         (TARGET_STARTING_FRAME_OFFSET): Redefine.
11129         * config/epiphany/epiphany.h (STARTING_FRAME_OFFSET): Delete.
11130         * config/epiphany/epiphany.c (epiphany_starting_frame_offset):
11131         New function.
11132         (TARGET_STARTING_FRAME_OFFSET): Redefine.
11133         * config/iq2000/iq2000.h (STARTING_FRAME_OFFSET): Delete.
11134         * config/iq2000/iq2000.c (iq2000_starting_frame_offset): New function.
11135         (TARGET_CONSTANT_ALIGNMENT): Redefine.
11136         * config/lm32/lm32.h (STARTING_FRAME_OFFSET): Delete.
11137         * config/lm32/lm32.c (lm32_starting_frame_offset): New function.
11138         (TARGET_STARTING_FRAME_OFFSET): Redefine.
11139         * config/m32r/m32r.h (STARTING_FRAME_OFFSET): Delete.
11140         * config/m32r/m32r.c (m32r_starting_frame_offset): New function.
11141         (TARGET_STARTING_FRAME_OFFSET): Redefine.
11142         * config/microblaze/microblaze.h (STARTING_FRAME_OFFSET): Delete.
11143         * config/microblaze/microblaze.c (microblaze_starting_frame_offset):
11144         New function.
11145         (TARGET_STARTING_FRAME_OFFSET): Redefine.
11146         * config/mips/mips.h (STARTING_FRAME_OFFSET): Delete.
11147         * config/mips/mips.c (mips_compute_frame_info): Refer to
11148         TARGET_STARTING_FRAME_OFFSET instead of STARTING_FRAME_OFFSET.
11149         (mips_starting_frame_offset): New function.
11150         (TARGET_STARTING_FRAME_OFFSET): Redefine.
11151         * config/mmix/mmix.h (STARTING_FRAME_OFFSET): Delete.
11152         * config/mmix/mmix-protos.h (mmix_starting_frame_offset): Delete.
11153         * config/mmix/mmix.c (mmix_starting_frame_offset): Make static
11154         and return a HOST_WIDE_INT.
11155         (TARGET_STARTING_FRAME_OFFSET): Redefine.
11156         (mmix_initial_elimination_offset): Refer to
11157         TARGET_STARTING_FRAME_OFFSET instead of STARTING_FRAME_OFFSET.
11158         * config/pa/pa.h (STARTING_FRAME_OFFSET): Delete.
11159         * config/pa/pa.c (pa_starting_frame_offset): New function.
11160         (pa_compute_frame_size): Use it instead of STARTING_FRAME_OFFSET.
11161         (pa_expand_prologue): Likewise.
11162         (TARGET_STARTING_FRAME_OFFSET): Redefine.
11163         * config/powerpcspe/aix.h (STARTING_FRAME_OFFSET): Split out
11164         !FRAME_GROWS_DOWNWARD handling to...
11165         (RS6000_STARTING_FRAME_OFFSET): ...this new macro.
11166         * config/powerpcspe/darwin.h (STARTING_FRAME_OFFSET): Split out
11167         !FRAME_GROWS_DOWNWARD handling to...
11168         (RS6000_STARTING_FRAME_OFFSET): ...this new macro.
11169         * config/powerpcspe/powerpcspe.h (STARTING_FRAME_OFFSET): Split out
11170         !FRAME_GROWS_DOWNWARD handling to...
11171         (RS6000_STARTING_FRAME_OFFSET): ...this new macro.
11172         * config/powerpcspe/powerpcspe.c (TARGET_STARTING_FRAME_OFFSET):
11173         Redefine.
11174         (rs6000_starting_frame_offset): New function.
11175         * config/rs6000/aix.h (STARTING_FRAME_OFFSET): Split out
11176         !FRAME_GROWS_DOWNWARD handling to...
11177         (RS6000_STARTING_FRAME_OFFSET): ...this new macro.
11178         * config/rs6000/darwin.h (STARTING_FRAME_OFFSET): Split out
11179         !FRAME_GROWS_DOWNWARD handling to...
11180         (RS6000_STARTING_FRAME_OFFSET): ...this new macro.
11181         * config/rs6000/rs6000.h (STARTING_FRAME_OFFSET): Split out
11182         !FRAME_GROWS_DOWNWARD handling to...
11183         (RS6000_STARTING_FRAME_OFFSET): ...this new macro.
11184         * config/rs6000/rs6000.c (TARGET_STARTING_FRAME_OFFSET): Refine.
11185         (rs6000_starting_frame_offset): New function.
11186         * config/vax/elf.h (STARTING_FRAME_OFFSET): Delete.
11187         * config/vax/vax.h (STARTING_FRAME_OFFSET): Delete.
11188         * config/vax/vax.c (vax_starting_frame_offset): New function.
11189         (vax_expand_prologue): Use it instead of STARTING_FRAME_OFFSET.
11190         (TARGET_STARTING_FRAME_OFFSET): Redefine.
11191         * config/xtensa/xtensa.h (STARTING_FRAME_OFFSET): Delete.
11192         * config/xtensa/xtensa.c (xtensa_starting_frame_offset): New function.
11193         (TARGET_STARTING_FRAME_OFFSET): Redefine.
11194         * system.h (STARTING_FRAME_OFFSET): Poison.
11196 2017-10-23  Richard Sandiford  <richard.sandiford@linaro.org>
11198         * tree-vect-loop.c (vect_create_epilog_for_reduction): Use
11199         SCALAR_TYPE_MODE instead of TYPE_MODE.
11201 2017-10-23  Richard Sandiford  <richard.sandiford@linaro.org>
11202             Alan Hayward  <alan.hayward@arm.com>
11203             David Sherwood  <david.sherwood@arm.com>
11205         * dwarf2out.c (loc_list_from_tree_1): Use SCALAR_INT_TYPE_MODE
11207 2017-10-23  Richard Sandiford  <richard.sandiford@linaro.org>
11208             Alan Hayward  <alan.hayward@arm.com>
11209             David Sherwood  <david.sherwood@arm.com>
11211         * expmed.c (expand_shift_1): Use scalar_mode for scalar_mode.
11213 2017-10-23  Richard Biener  <rguenther@suse.de>
11215         PR tree-optimization/82129
11216         * tree-ssa-pre.c (bitmap_set_and): Remove.
11217         (compute_antic_aux): Compute ANTIC_OUT intersection in a way
11218         canonicalizing expressions in the set to those with lowest
11219         ID rather than taking that from the first edge.
11221 2017-10-23  Richard Sandiford  <richard.sandiford@linaro.org>
11223         * combine.c (rtx_equal_for_field_assignment_p): Use
11224         byte_lowpart_offset.
11226 2017-10-22  Richard Sandiford  <richard.sandiford@linaro.org>
11227             Alan Hayward  <alan.hayward@arm.com>
11228             David Sherwood  <david.sherwood@arm.com>
11230         * internal-fn.c (expand_direct_optab_fn): Don't assign directly
11231         to a SUBREG_PROMOTED_VAR.
11233 2017-10-22  Richard Sandiford  <richard.sandiford@linaro.org>
11234             Alan Hayward  <alan.hayward@arm.com>
11235             David Sherwood  <david.sherwood@arm.com>
11237         * cfgexpand.c (expand_debug_expr): Use GET_MODE_UNIT_PRECISION.
11238         (expand_debug_source_expr): Likewise.
11239         * combine.c (combine_simplify_rtx): Likewise.
11240         * cse.c (fold_rtx): Likewise.
11241         * optabs.c (expand_float): Likewise.
11242         * simplify-rtx.c (simplify_unary_operation_1): Likewise.
11243         (simplify_binary_operation_1): Likewise.
11245 2017-10-22  Richard Sandiford  <richard.sandiford@linaro.org>
11246             Alan Hayward  <alan.hayward@arm.com>
11247             David Sherwood  <david.sherwood@arm.com>
11249         * combine.c (simplify_comparison): Use HWI_COMPUTABLE_MODE_P.
11250         (record_promoted_value): Likewise.
11251         * expr.c (expand_expr_real_2): Likewise.
11252         * ree.c (update_reg_equal_equiv_notes): Likewise.
11253         (combine_set_extension): Likewise.
11254         * rtlanal.c (low_bitmask_len): Likewise.
11255         * simplify-rtx.c (neg_const_int): Likewise.
11256         (simplify_binary_operation_1): Likewise.
11258 2017-10-22  Richard Sandiford  <richard.sandiford@linaro.org>
11259             Alan Hayward  <alan.hayward@arm.com>
11260             David Sherwood  <david.sherwood@arm.com>
11262         * lra-spills.c (assign_mem_slot): Use subreg_size_lowpart_offset.
11263         * regcprop.c (maybe_mode_change): Likewise.
11264         * reload1.c (alter_reg): Likewise.
11266 2017-10-22  Richard Sandiford  <richard.sandiford@linaro.org>
11268         * inchash.h (inchash::hash::add_wide_int): New function.
11269         * lto-streamer-out.c (hash_tree): Use it.
11271 2017-10-22  Richard Sandiford  <richard.sandiford@linaro.org>
11273         * inchash.h (inchash::hash::add_wide_int): Rename to...
11274         (inchash::hash::add_hwi): ...this.
11275         * ipa-devirt.c (hash_odr_vtable): Update accordingly.
11276         (polymorphic_call_target_hasher::hash): Likewise.
11277         * ipa-icf.c (sem_function::get_hash, sem_function::init): Likewise.
11278         (sem_item::add_expr, sem_item::add_type, sem_variable::get_hash)
11279         (sem_item_optimizer::update_hash_by_addr_refs): Likewise.
11280         * lto-streamer-out.c (hash_tree): Likewise.
11281         * optc-save-gen.awk: Likewise.
11282         * tree.c (add_expr): Likewise.
11284 2017-10-22  Uros Bizjak  <ubizjak@gmail.com>
11286         PR target/52451
11287         * config/i386/i386.c (ix86_fp_compare_mode): Return CCFPmode
11288         for ordered inequality comparisons even with TARGET_IEEE_FP.
11290 2017-10-22  Uros Bizjak  <ubizjak@gmail.com>
11292         PR target/82628
11293         * config/i386/i386.md (cmp<dwi>_doubleword): New pattern.
11294         * config/i386/i386.c (ix86_expand_branch) <case E_TImode>:
11295         Expand with cmp<dwi>_doubleword.
11297 2017-10-21  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
11299         * extend.texi: Add x86 specific to 'nocf_check' attribute.
11300         List CET intrinsics.
11301         * invoke.texi: Add -mcet, -mibt, -mshstk options.  Add x86
11302         specific to -fcf-protection option.
11304 2017-10-21  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
11306         * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET): New.
11307         (OPTION_MASK_ISA_SHSTK_SET): Likewise.
11308         (OPTION_MASK_ISA_IBT_UNSET): Likewise.
11309         (OPTION_MASK_ISA_SHSTK_UNSET): Likewise.
11310         (ix86_handle_option): Add -mibt, -mshstk, -mcet handling.
11311         * config.gcc (extra_headers): Add cetintrin.h for x86 targets.
11312         (extra_objs): Add cet.o for Linux/x86 targets.
11313         (tmake_file): Add i386/t-cet for Linux/x86 targets.
11314         * config/i386/cet.c: New file.
11315         * config/i386/cetintrin.h: Likewise.
11316         * config/i386/t-cet: Likewise.
11317         * config/i386/cpuid.h (bit_SHSTK): New.
11318         (bit_IBT): Likewise.
11319         * config/i386/driver-i386.c (host_detect_local_cpu): Detect and
11320         pass IBT and SHSTK bits.
11321         * config/i386/i386-builtin-types.def
11322         (VOID_FTYPE_UNSIGNED_PVOID): New.
11323         (VOID_FTYPE_UINT64_PVOID): Likewise.
11324         * config/i386/i386-builtin.def: Add CET intrinsics.
11325         * config/i386/i386-c.c (ix86_target_macros_internal): Add
11326         OPTION_MASK_ISA_IBT, OPTION_MASK_ISA_SHSTK handling.
11327         * config/i386/i386-passes.def: Add pass_insert_endbranch pass.
11328         * config/i386/i386-protos.h (make_pass_insert_endbranch): New
11329         prototype.
11330         * config/i386/i386.c (rest_of_insert_endbranch): New.
11331         (pass_data_insert_endbranch): Likewise.
11332         (pass_insert_endbranch): Likewise.
11333         (make_pass_insert_endbranch): Likewise.
11334         (ix86_notrack_prefixed_insn_p): Likewise.
11335         (ix86_target_string): Add -mibt, -mshstk flags.
11336         (ix86_option_override_internal): Add flag_cf_protection
11337         processing.
11338         (ix86_valid_target_attribute_inner_p): Set OPT_mibt, OPT_mshstk.
11339         (ix86_print_operand): Add 'notrack' prefix output.
11340         (ix86_init_mmx_sse_builtins): Add CET intrinsics.
11341         (ix86_expand_builtin): Expand CET intrinsics.
11342         (x86_output_mi_thunk): Add 'endbranch' instruction.
11343         * config/i386/i386.h (TARGET_IBT): New.
11344         (TARGET_IBT_P): Likewise.
11345         (TARGET_SHSTK): Likewise.
11346         (TARGET_SHSTK_P): Likewise.
11347         * config/i386/i386.md (unspecv): Add UNSPECV_NOP_RDSSP,
11348         UNSPECV_INCSSP, UNSPECV_SAVEPREVSSP, UNSPECV_RSTORSSP,
11349         UNSPECV_WRSS, UNSPECV_WRUSS, UNSPECV_SETSSBSY, UNSPECV_CLRSSBSY.
11350         (builtin_setjmp_setup): New pattern.
11351         (builtin_longjmp): Likewise.
11352         (rdssp<mode>): Likewise.
11353         (incssp<mode>): Likewise.
11354         (saveprevssp): Likewise.
11355         (rstorssp): Likewise.
11356         (wrss<mode>): Likewise.
11357         (wruss<mode>): Likewise.
11358         (setssbsy): Likewise.
11359         (clrssbsy): Likewise.
11360         (nop_endbr): Likewise.
11361         * config/i386/i386.opt: Add -mcet, -mibt, -mshstk and -mcet-switch
11362         options.
11363         * config/i386/immintrin.h: Include <cetintrin.h>.
11364         * config/i386/linux-common.h
11365         (file_end_indicate_exec_stack_and_cet): New prototype.
11366         (TARGET_ASM_FILE_END): New.
11368 2017-10-20  Jan Hubicka  <hubicka@ucw.cz>
11370         * i386.c (ix86_builtin_vectorization_cost): Use existing rtx_cost
11371         latencies instead of having separate table; make difference between
11372         integer and float costs.
11373         * i386.h (processor_costs): Remove scalar_stmt_cost,
11374         scalar_load_cost, scalar_store_cost, vec_stmt_cost, vec_to_scalar_cost,
11375         scalar_to_vec_cost, vec_align_load_cost, vec_unalign_load_cost,
11376         vec_store_cost.
11377         * x86-tune-costs.h: Remove entries which has been removed in
11378         procesor_costs from all tables; make cond_taken_branch_cost
11379         and cond_not_taken_branch_cost COST_N_INSNS based.
11381 2017-10-20  Jan Hubicka  <hubicka@ucw.cz>
11383         * x86-tune-costs.h (intel_cost, generic_cost): Fix move costs.
11385 2017-10-20  Jakub Jelinek  <jakub@redhat.com>
11387         * config/i386/i386.md (isa): Remove fma_avx512f.
11388         * config/i386/sse.md (<avx512>_fmadd_<mode>_mask<round_name>,
11389         <avx512>_fmadd_<mode>_mask3<round_name>,
11390         <avx512>_fmsub_<mode>_mask<round_name>,
11391         <avx512>_fmsub_<mode>_mask3<round_name>,
11392         <avx512>_fnmadd_<mode>_mask<round_name>,
11393         <avx512>_fnmadd_<mode>_mask3<round_name>,
11394         <avx512>_fnmsub_<mode>_mask<round_name>,
11395         <avx512>_fnmsub_<mode>_mask3<round_name>,
11396         <avx512>_fmaddsub_<mode>_mask<round_name>,
11397         <avx512>_fmaddsub_<mode>_mask3<round_name>,
11398         <avx512>_fmsubadd_<mode>_mask<round_name>,
11399         <avx512>_fmsubadd_<mode>_mask3<round_name>): Remove isa attribute.
11400         (*vec_widen_umult_even_v16si<mask_name>,
11401         *vec_widen_smult_even_v16si<mask_name>): Likewise.
11402         (<mask_codefor>avx512bw_dbpsadbw<mode><mask_name>): Likewise.
11404 2017-10-20  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
11406         * extend.texi: Add 'nocf_check' documentation.
11407         * gimple.texi: Add second parameter to
11408         gimple_build_call_from_tree.
11409         * invoke.texi: Add -fcf-protection documentation.
11410         * rtl.texi: Add REG_CALL_NOTRACK documenation.
11412 2017-10-20  Richard Biener  <rguenther@suse.de>
11414         PR tree-optimization/82473
11415         * tree-vect-loop.c (vectorizable_reduction): Properly get at
11416         the largest input type.
11418 2017-10-20  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
11420         * c-attribs.c (handle_nocf_check_attribute): New function.
11421         (c_common_attribute_table): Add 'nocf_check' handling.
11422         * gimple-parser.c: Add second argument NULL to
11423         gimple_build_call_from_tree.
11424         * attrib.c (comp_type_attributes): Check nocf_check attribute.
11425         * cfgexpand.c (expand_call_stmt): Set REG_CALL_NOCF_CHECK for
11426         call insn.
11427         * combine.c (distribute_notes): Add REG_CALL_NOCF_CHECK handling.
11428         * common.opt: Add fcf-protection flag.
11429         * emit-rtl.c (try_split): Add REG_CALL_NOCF_CHECK handling.
11430         * flag-types.h: Add enum cf_protection_level.
11431         * gimple.c (gimple_build_call_from_tree): Add second parameter.
11432         Add 'nocf_check' attribute propagation to gimple call.
11433         * gimple.h (gf_mask): Add GF_CALL_NOCF_CHECK.
11434         (gimple_build_call_from_tree): Update prototype.
11435         (gimple_call_nocf_check_p): New function.
11436         (gimple_call_set_nocf_check): Likewise.
11437         * gimplify.c: Add second argument to gimple_build_call_from_tree.
11438         * ipa-icf.c: Add nocf_check attribute in statement hash.
11439         * recog.c (peep2_attempt): Add REG_CALL_NOCF_CHECK handling.
11440         * reg-notes.def: Add REG_NOTE (CALL_NOCF_CHECK).
11441         * toplev.c (process_options): Add flag_cf_protection handling.
11443 2017-10-19  Jan Hubicka  <hubicka@ucw.cz>
11445         * x86-tune-costs.h (core_cost): Fix div, move and sqrt latencies.
11447 2017-10-20  Richard Biener  <rguenther@suse.de>
11449         PR tree-optimization/82603
11450         * tree-if-conv.c (predicate_mem_writes): Make sure to only
11451         remove false predicated stores.
11453 2017-10-20  Richard Biener  <rguenther@suse.de>
11455         * graphite-isl-ast-to-gimple.c
11456         (translate_isl_ast_to_gimple::graphite_copy_stmts_from_block):
11457         Remove return value and simplify, dump copied stmt after lhs
11458         adjustment.
11459         (translate_isl_ast_to_gimple::translate_isl_ast_node_user):
11460         Reduce dump verbosity.
11461         (gsi_insert_earliest): Likewise.
11462         (translate_isl_ast_to_gimple::copy_bb_and_scalar_dependences): Adjust.
11463         * graphite.c (print_global_statistics): Adjust dumping.
11464         (print_graphite_scop_statistics): Likewise.
11465         (print_graphite_statistics): Do not dump loops here.
11466         (graphite_transform_loops): But here.
11468 2017-10-20  Nicolas Roche  <roche@adacore.com>
11470         * configure.ac (ACX_PROG_GNAT): Append "libgnat" to include search dir.
11471         * configure: Regenerate.
11473 2017-10-20  Jakub Jelinek  <jakub@redhat.com>
11475         PR target/82158
11476         * tree-cfg.c (pass_warn_function_return::execute): In noreturn
11477         functions when optimizing replace GIMPLE_RETURN stmts with
11478         calls to __builtin_unreachable ().
11480         PR sanitizer/82595
11481         * config/gnu-user.h (LIBTSAN_EARLY_SPEC): Add libtsan_preinit.o
11482         for -fsanitize=thread link of executables.
11483         (LIBLSAN_EARLY_SPEC): Add liblsan_preinit.o for -fsanitize=leak
11484         link of executables.
11486         PR target/82370
11487         * config/i386/sse.md (VI248_AVX2, VI248_AVX512BW, VI248_AVX512BW_2):
11488         New mode iterators.
11489         (<shift_insn><mode>3<mask_name>): Change the last of the 3
11490         define_insns for logical vector shifts to use VI248_AVX512BW
11491         iterator instead of VI48_AVX512, remove <mask_mode512bit_condition>
11492         condition, useless isa and prefix attributes.  Change the first
11493         2 of these define_insns to ...
11494         (<mask_codefor><shift_insn><mode>3<mask_name>): ... this, new
11495         define_insn for avx512vl.
11496         (<shift_insn><mode>3): ... and this, new define_insn without
11497         masking for non-avx512vl.
11499         PR target/82370
11500         * config/i386/sse.md (*andnot<mode>3,
11501         <mask_codefor><code><mode>3<mask_name>, *<code><mode>3): Split
11502         (=v,v,vm) alternative into (=x,x,xm) and (=v,v,vm), for 128-bit
11503         and 256-bit vectors, the (=x,x,xm) alternative and when mask is
11504         not applied use empty suffix even for TARGET_AVX512VL.
11505         * config/i386/subst.md (mask_prefix3, mask_prefix4): When mask
11506         is applied, supply evex,evex or evex,evex,evex instead of just
11507         evex.
11509 2017-10-20  Julia Koval  <julia.koval@intel.com>
11511         * common/config/i386/i386-common.c (OPTION_MASK_ISA_GFNI_SET,
11512         (OPTION_MASK_ISA_GFNI_UNSET): New.
11513         (ix86_handle_option): Handle OPT_mgfni.
11514         * config/i386/cpuid.h (bit_GFNI): New.
11515         * config/i386/driver-i386.c (host_detect_local_cpu): Detect gfni.
11516         * config/i386/i386-c.c (ix86_target_macros_internal): Define __GFNI__.
11517         * config/i386/i386.c (ix86_target_string): Add -mgfni.
11518         (ix86_valid_target_attribute_inner_p): Add OPT_mgfni.
11519         * config/i386/i386.h (TARGET_GFNI, TARGET_GFNI_P): New.
11520         * config/i386/i386.opt: Add mgfni.
11522 2017-10-20  Orlando Arias  <oarias@knights.ucf.edu>
11524         * config/msp430/msp430.c (msp430_option_override): Disable
11525         -fdelete-null-pointer-checks.
11526         * doc/invoke.text (-fdelete-null-pointer-checks): Document that.
11528 2017-10-19  Jan Hubicka  <hubicka@ucw.cz>
11530         * x86-tune-costs.h (generic_cost, core_cost): Correct costs
11531         of x87 and SSE instructions.
11533 2017-10-19  Jan Hubicka  <hubicka@ucw.cz>
11535         * asan.c (create_cond_insert_point): Do not update edge count.
11536         * auto-profile.c (afdo_propagate_edge): Update for edge count removal.
11537         (afdo_propagate_circuit): Likewise.
11538         (afdo_calculate_branch_prob): Likewise.
11539         (afdo_annotate_cfg): Likewise.
11540         * basic-block.h (struct edge_def): Remove count.
11541         (edge_def::count): New accessor.
11542         * bb-reorder.c (rotate_loop): Update.
11543         (find_traces_1_round): Update.
11544         (connect_traces): Update.
11545         (sanitize_hot_paths): Update.
11546         * cfg.c (unchecked_make_edge): Update.
11547         (make_single_succ_edge): Update.
11548         (check_bb_profile): Update.
11549         (dump_edge_info): Update.
11550         (update_bb_profile_for_threading): Update.
11551         (scale_bbs_frequencies_int): Update.
11552         (scale_bbs_frequencies_gcov_type): Update.
11553         (scale_bbs_frequencies_profile_count): Update.
11554         (scale_bbs_frequencies): Update.
11555         * cfganal.c (connect_infinite_loops_to_exit): Update.
11556         * cfgbuild.c (compute_outgoing_frequencies): Update.
11557         (find_many_sub_basic_blocks): Update.
11558         * cfgcleanup.c (try_forward_edges): Update.
11559         (try_crossjump_to_edge): Update
11560         * cfgexpand.c (expand_gimple_cond): Update
11561         (expand_gimple_tailcall): Update
11562         (construct_exit_block): Update
11563         * cfghooks.c (verify_flow_info): Update
11564         (redirect_edge_succ_nodup): Update
11565         (split_edge): Update
11566         (make_forwarder_block): Update
11567         (duplicate_block): Update
11568         (account_profile_record): Update
11569         * cfgloop.c (find_subloop_latch_edge_by_profile): Update.
11570         * cfgloopanal.c (expected_loop_iterations_unbounded): Update.
11571         * cfgloopmanip.c (scale_loop_profile): Update.
11572         (loopify): Update.
11573         (lv_adjust_loop_entry_edge): Update.
11574         * cfgrtl.c (try_redirect_by_replacing_jump): Update.
11575         (force_nonfallthru_and_redirect): Update.
11576         (purge_dead_edges): Update.
11577         (rtl_flow_call_edges_add): Update.
11578         * cgraphunit.c (init_lowered_empty_function): Update.
11579         (cgraph_node::expand_thunk): Update.
11580         * gimple-pretty-print.c (dump_probability): Update.
11581         (dump_edge_probability): Update.
11582         * gimple-ssa-isolate-paths.c (isolate_path): Update.
11583         * haifa-sched.c (sched_create_recovery_edges): Update.
11584         * hsa-gen.c (convert_switch_statements): Update.
11585         * ifcvt.c (dead_or_predicable): Update.
11586         * ipa-inline-transform.c (inline_transform): Update.
11587         * ipa-split.c (split_function): Update.
11588         * ipa-utils.c (ipa_merge_profiles): Update.
11589         * loop-doloop.c (add_test): Update.
11590         * loop-unroll.c (unroll_loop_runtime_iterations): Update.
11591         * lto-streamer-in.c (input_cfg): Update.
11592         (input_function): Update.
11593         * lto-streamer-out.c (output_cfg): Update.
11594         * modulo-sched.c (sms_schedule): Update.
11595         * postreload-gcse.c (eliminate_partially_redundant_load): Update.
11596         * predict.c (maybe_hot_edge_p): Update.
11597         (unlikely_executed_edge_p): Update.
11598         (probably_never_executed_edge_p): Update.
11599         (dump_prediction): Update.
11600         (drop_profile): Update.
11601         (propagate_unlikely_bbs_forward): Update.
11602         (determine_unlikely_bbs): Update.
11603         (force_edge_cold): Update.
11604         * profile.c (compute_branch_probabilities): Update.
11605         * reg-stack.c (better_edge): Update.
11606         * shrink-wrap.c (handle_simple_exit): Update.
11607         * tracer.c (better_p): Update.
11608         * trans-mem.c (expand_transaction): Update.
11609         (split_bb_make_tm_edge): Update.
11610         * tree-call-cdce.c: Update.
11611         * tree-cfg.c (gimple_find_sub_bbs): Update.
11612         (gimple_split_edge): Update.
11613         (gimple_duplicate_sese_region): Update.
11614         (gimple_duplicate_sese_tail): Update.
11615         (gimple_flow_call_edges_add): Update.
11616         (insert_cond_bb): Update.
11617         (execute_fixup_cfg): Update.
11618         * tree-cfgcleanup.c (cleanup_control_expr_graph): Update.
11619         * tree-complex.c (expand_complex_div_wide): Update.
11620         * tree-eh.c (lower_resx): Update.
11621         (unsplit_eh): Update.
11622         (cleanup_empty_eh_move_lp): Update.
11623         * tree-inline.c (copy_edges_for_bb): Update.
11624         (freqs_to_counts): Update.
11625         (copy_cfg_body): Update.
11626         * tree-ssa-dce.c (remove_dead_stmt): Update.
11627         * tree-ssa-ifcombine.c (update_profile_after_ifcombine): Update.
11628         * tree-ssa-loop-im.c (execute_sm_if_changed): Update.
11629         * tree-ssa-loop-ivcanon.c (remove_exits_and_undefined_stmts): Update.
11630         (unloop_loops): Update.
11631         * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Update.
11632         * tree-ssa-loop-split.c (connect_loops): Update.
11633         (split_loop): Update.
11634         * tree-ssa-loop-unswitch.c (hoist_guard): Update.
11635         * tree-ssa-phionlycprop.c (propagate_rhs_into_lhs): Update.
11636         * tree-ssa-phiopt.c (replace_phi_edge_with_variable): Update.
11637         * tree-ssa-reassoc.c (branch_fixup): Update.
11638         * tree-ssa-tail-merge.c (replace_block_by): Update.
11639         * tree-ssa-threadupdate.c (remove_ctrl_stmt_and_useless_edges): Update.
11640         (compute_path_counts): Update.
11641         (update_profile): Update.
11642         (recompute_probabilities): Update.
11643         (update_joiner_offpath_counts): Update.
11644         (estimated_freqs_path): Update.
11645         (freqs_to_counts_path): Update.
11646         (clear_counts_path): Update.
11647         (ssa_fix_duplicate_block_edges): Update.
11648         (duplicate_thread_path): Update.
11649         * tree-switch-conversion.c (hoist_edge_and_branch_if_true): Update.
11650         (case_bit_test_cmp): Update.
11651         (collect_switch_conv_info): Update.
11652         (gen_inbound_check): Update.
11653         (do_jump_if_equal): Update.
11654         (emit_cmp_and_jump_insns): Update.
11655         * tree-tailcall.c (decrease_profile): Update.
11656         (eliminate_tail_call): Update.
11657         * tree-vect-loop-manip.c (slpeel_add_loop_guard): Update.
11658         (vect_do_peeling): Update.
11659         * tree-vect-loop.c (scale_profile_for_vect_loop): Update.
11660         * ubsan.c (ubsan_expand_null_ifn): Update.
11661         (ubsan_expand_ptr_ifn): Update.
11662         * value-prof.c (gimple_divmod_fixed_value): Update.
11663         (gimple_mod_pow2): Update.
11664         (gimple_mod_subtract): Update.
11665         (gimple_ic): Update.
11666         (gimple_stringop_fixed_value): Update.
11668 2017-10-19  Uros Bizjak  <ubizjak@gmail.com>
11670         PR target/82618
11671         * config/i386/i386.md (sub to cmp): New peephole2 pattern.
11673 2017-10-19  Alexander Monakov  <amonakov@ispras.ru>
11675         PR rtl-optimization/82395
11676         * ira-color.c (allocno_priority_compare_func): Fix comparison step
11677         based on non_spilled_static_chain_regno_p.
11679 2017-10-19  Uros Bizjak  <ubizjak@gmail.com>
11681         * config/i386/i386.c (output_387_binary_op): Rewrite SSE part.
11682         (ix86_emit_mode_set): Rewrite insn mnemonic construction.
11683         (ix86_prepare_fp_compare_args): Redefine is_sse as bool.
11685 2017-10-19  Martin Sebor  <msebor@redhat.com>
11687         PR tree-optimization/82596
11688         * tree.c (array_at_struct_end_p): Handle STRING_CST.
11690 2017-10-19  Eric Botcazou  <ebotcazou@adacore.com>
11692         * asan.c (handle_builtin_alloca): Deal with all alloca variants.
11693         (get_mem_refs_of_builtin_call): Likewise.
11694         * builtins.c (expand_builtin_apply): Adjust call to
11695         allocate_dynamic_stack_space.
11696         (expand_builtin_alloca): For __builtin_alloca_with_align_and_max, pass
11697         the third argument to allocate_dynamic_stack_space, otherwise -1.
11698         (expand_builtin): Deal with all alloca variants.
11699         (is_inexpensive_builtin): Likewise.
11700         * builtins.def (BUILT_IN_ALLOCA_WITH_ALIGN_AND_MAX): New.
11701         * calls.c (special_function_p): Deal with all alloca variants.
11702         (initialize_argument_information): Adjust call to
11703         allocate_dynamic_stack_space.
11704         (expand_call): Likewise.
11705         * cfgexpand.c (expand_call_stmt): Deal with all alloca variants.
11706         * doc/extend.texi (Built-ins): Add __builtin_alloca_with_align_and_max
11707         * explow.c (allocate_dynamic_stack_space): Add MAX_SIZE parameter and
11708         use it for the stack usage computation.
11709         * explow.h (allocate_dynamic_stack_space): Adjust prototype.
11710         * function.c (gimplify_parameters): Call build_alloca_call_expr.
11711         * gimple-ssa-warn-alloca.c (alloca_call_type): Simplify control flow.
11712         Take into account 3rd argument of __builtin_alloca_with_align_and_max.
11713         (in_loop_p): Remove first argument and useless check.
11714         (pass_walloca::execute): Remove useless test and adjust call to above.
11715         * gimple.c (gimple_build_call_from_tree): Deal with all alloc variants
11716         * gimplify.c (gimplify_vla_decl): Call build_alloca_call_expr.
11717         (gimplify_call_expr): Deal with all alloca variants.
11718         * hsa-gen.c (gen_hsa_alloca): Likewise.
11719         (gen_hsa_insns_for_call): Likewise.
11720         * ipa-pure-const.c (special_builtin_state): Likewise.
11721         * tree-chkp.c (chkp_build_returned_bound): Likewise.
11722         * tree-object-size.c (alloc_object_size): Likewise.
11723         * tree-ssa-alias.c (ref_maybe_used_by_call_p_1): Likewise.
11724         (call_may_clobber_ref_p_1): Likewise.
11725         * tree-ssa-ccp.c (evaluate_stmt): Likewise.
11726         (ccp_fold_stmt): Likewise.
11727         (optimize_stack_restore): Likewise.
11728         * tree-ssa-dce.c (mark_stmt_if_obviously_necessary): Likewise.
11729         (mark_all_reaching_defs_necessary_1): Likewise.
11730         (propagate_necessity): Likewise.
11731         (eliminate_unnecessary_stmts): Likewise.
11732         * tree.c (build_common_builtin_nodes): Build
11733         BUILT_IN_ALLOCA_WITH_ALIGN_AND_MAX.
11734         (build_alloca_call_expr): New function.
11735         * tree.h (ALLOCA_FUNCTION_CODE_P): New macro.
11736         (CASE_BUILT_IN_ALLOCA): Likewise.
11737         (build_alloca_call_expr): Declare.
11738         * varasm.c (incorporeal_function_p): Deal with all alloca variants.
11740 2017-10-19  Eric Botcazou  <ebotcazou@adacore.com>
11742         PR debug/82509
11743         * dwarf2out.c (new_die_raw): New static inline function.
11744         (new_die): Use it to create the DIE.
11745         (add_AT_external_die_ref): Likewise.
11746         (clone_die): Likewise.
11747         (clone_as_declaration): Likewise.
11748         (dwarf2out_vms_debug_main_pointer): Likewise.
11749         (base_type_die): Likewise.  Remove early return for corner cases.
11750         Do not call add_pubtype on the DIE here.
11751         (is_base_type): Remove ERROR_MARK and return 0 for VOID_TYPE.
11752         (modified_type_die): Adjust the lookup for reverse order DIEs.  Skip
11753         typedefs for base types with DW_AT_endianity.  Make sure a DIE with
11754         native order exists for base types, attach the DIE manually and call
11755         add_pubtype on it.  Do not equate a reverse order DIE to the type.
11757 2017-10-19  Richard Earnshaw  <rearnsha@arm.com>
11759         * config/arm/arm.c (align_ok_ldrd_strd): New function.
11760         (mem_ok_for_ldrd_strd): New parameter align.  Extract the alignment of
11761         the mem into it.
11762         (gen_operands_ldrd_strd): Validate the alignment of the accesses.
11764 2017-10-19  Jakub Jelinek  <jakub@redhat.com>
11766         * flag-types.h (enum sanitize_code): Add SANITIZE_BUILTIN.  Or
11767         SANITIZE_BUILTIN into SANITIZE_UNDEFINED.
11768         * sanitizer.def (BUILT_IN_UBSAN_HANDLE_INVALID_BUILTIN,
11769         BUILT_IN_UBSAN_HANDLE_INVALID_BUILTIN_ABORT): New builtins.
11770         * opts.c (sanitizer_opts): Add builtin.
11771         * ubsan.c (instrument_builtin): New function.
11772         (pass_ubsan::execute): Call it.
11773         (pass_ubsan::gate): Enable even for SANITIZE_BUILTIN.
11774         * doc/invoke.texi: Document -fsanitize=builtin.
11776         * ubsan.c (ubsan_expand_null_ifn): Use _v1 suffixed type mismatch
11777         builtins, store max (log2 (align), 0) into uchar field instead of
11778         align into uptr field.
11779         (ubsan_expand_objsize_ifn): Use _v1 suffixed type mismatch builtins,
11780         store uchar 0 field instead of uptr 0 field.
11781         (instrument_nonnull_return): Use _v1 suffixed nonnull return builtin,
11782         instead of passing one address of struct with 2 locations pass
11783         two addresses of structs with 1 location each.
11784         * sanitizer.def (BUILT_IN_UBSAN_HANDLE_TYPE_MISMATCH,
11785         BUILT_IN_UBSAN_HANDLE_TYPE_MISMATCH_ABORT,
11786         BUILT_IN_UBSAN_HANDLE_NONNULL_RETURN,
11787         BUILT_IN_UBSAN_HANDLE_NONNULL_RETURN_ABORT): Removed.
11788         (BUILT_IN_UBSAN_HANDLE_TYPE_MISMATCH_V1,
11789         BUILT_IN_UBSAN_HANDLE_TYPE_MISMATCH_V1_ABORT,
11790         BUILT_IN_UBSAN_HANDLE_NONNULL_RETURN_V1,
11791         BUILT_IN_UBSAN_HANDLE_NONNULL_RETURN_V1_ABORT): New builtins.
11793 2017-10-19  Martin Liska  <mliska@suse.cz>
11795         PR driver/81829
11796         * file-find.c (remove_prefix): Remove.
11797         * file-find.h (remove_prefix): Likewise.
11798         * gcc-ar.c: Remove smartness of lookup.
11800 2017-10-19  Segher Boessenkool  <segher@kernel.crashing.org>
11802         * config/rs6000/rs6000.md (*call_indirect_aix<mode>,
11803         *call_value_indirect_aix<mode>, *call_indirect_elfv2<mode>,
11804         *call_value_indirect_elfv2<mode>): Add correct mode to the unspec.
11806 2017-10-19  Jakub Jelinek  <jakub@redhat.com>
11808         PR target/82580
11809         * config/i386/i386.md (setcc + movzbl to xor + setcc): New peephole2.
11810         (setcc + and to xor + setcc): New peephole2.
11812 2017-10-19  Tom de Vries  <tom@codesourcery.com>
11814         * doc/sourcebuild.texi (Test Directives, Variants of
11815         dg-require-support): Add dg-require-stack-size.
11817 2017-10-19  Martin Liska  <mliska@suse.cz>
11819         PR sanitizer/82517
11820         * gimplify.c (gimplify_decl_expr): Do not instrument variables
11821         that have a large alignment.
11822         (gimplify_target_expr): Likewise.
11824 2017-10-18  Segher Boessenkool  <segher@kernel.crashing.org>
11826         PR rtl-optimization/82602
11827         * ira.c (rtx_moveable_p): Return false for volatile asm.
11829 2017-10-18  Uros Bizjak  <ubizjak@gmail.com>
11831         PR target/82580
11832         * config/i386/i386-modes.def (CCGZ): New CC mode.
11833         * config/i386/i386.md (sub<mode>3_carry_ccgz): New insn pattern.
11834         * config/i386/predicates.md (ix86_comparison_operator):
11835         Handle CCGZmode.
11836         * config/i386/i386.c (ix86_expand_branch) <case E_TImode>:
11837         Emulate LE, LEU, GT, GTU, LT, LTU, GE and GEU double-word comparisons
11838         with double-word subtraction.
11839         (put_condition_code): Handle CCGZmode.
11841 2017-10-18  Aldy Hernandez  <aldyh@redhat.com>
11843         * wide-int.cc (debug (const wide_int &)): New.
11844         (debug (const wide_int *)): New.
11845         (debug (const widest_int &)): New.
11846         (debug (const widest_int *)): New.
11848 2017-10-18  Vladimir Makarov  <vmakarov@redhat.com>
11850         PR middle-end/82556
11851         * lra-constraints.c (curr_insn_transform): Use non-input operand
11852         instead of output one for matched reload.
11854 2017-10-18  Bin Cheng  <bin.cheng@arm.com>
11856         * tree-loop-distribution.c (INCLUDE_ALGORITHM): New header file.
11857         (tree-ssa-loop-ivopts.h): New header file.
11858         (struct builtin_info): New fields.
11859         (classify_builtin_1): Compute and record base and offset parts for
11860         memset builtin partition by calling strip_offset.
11861         (offset_cmp, fuse_memset_builtins): New functions.
11862         (finalize_partitions): Fuse adjacent memset partitions by calling
11863         above function.
11864         * tree-ssa-loop-ivopts.c (strip_offset): Delete static declaration.
11865         Expose the interface.
11866         * tree-ssa-loop-ivopts.h (strip_offset): New declaration.
11868 2017-10-18  Bin Cheng  <bin.cheng@arm.com>
11870         PR tree-optimization/82574
11871         * tree-loop-distribution.c (find_single_drs): New parameter.  Check
11872         that data reference must be executed exactly once per iteration
11873         against the outermost loop in nest.
11874         (classify_partition): Update call to above function.
11876 2017-10-18  Richard Biener  <rguenther@suse.de>
11878         PR tree-optimization/82591
11879         * graphite.c (graphite_transform_loops): Move code gen message
11880         printing ...
11881         * graphite-isl-ast-to-gimple.c (graphite_regenerate_ast_isl):
11882         Here.  Handle scop_to_isl_ast failing.
11883         (scop_to_isl_ast): Limit the number of ISL operations.
11885 2017-10-18  Richard Biener  <rguenther@suse.de>
11887         * graphite-isl-ast-to-gimple.c
11888         (translate_isl_ast_to_gimple::set_rename): Simplify.
11889         (translate_isl_ast_to_gimple::set_rename_for_each_def): Inline...
11890         (graphite_copy_stmts_from_block): ... here.
11891         (copy_bb_and_scalar_dependences): Simplify.
11892         (add_parameters_to_ivs_params): Canonicalize.
11893         (generate_entry_out_of_ssa_copies): Simplify.
11894         * graphite-sese-to-poly.c (extract_affine_name): Simplify
11895         by passing in ISL dimension.
11896         (parameter_index_in_region_1): Rename to ...
11897         (parameter_index_in_region): ... this.
11898         (extract_affine): Adjust assert, pass down parameter index.
11899         (add_param_constraints): Use range-info when available.
11900         (build_scop_context): Adjust.
11901         * sese.c (new_sese_info): Adjust.
11902         (free_sese_info): Likewise.
11903         * sese.h (bb_map_t, rename_map_t, phi_rename, init_back_edge_pair_t):
11904         Remove unused typedefs.
11905         (struct sese_info_t): Simplify rename_map, remove incomplete_phis.
11907 2017-10-18  Martin Liska  <mliska@suse.cz>
11909         * combine.c (simplify_compare_const): Add gcc_fallthrough.
11911 2017-10-18  Robin Dapp  <rdapp@linux.vnet.ibm.com>
11913         * config/s390/s390.c (s390_bb_fallthru_entry_likely): New function.
11914         (s390_sched_init): Do not reset s390_sched_state if we entered the
11915         current basic block via a fallthru edge and all others are unlikely.
11917 2017-10-18  Robin Dapp  <rdapp@linux.vnet.ibm.com>
11919         * config/s390/s390.c (NUM_SIDES): New variable.
11920         (LONGRUNNING_THRESHOLD): New variable.
11921         (LATENCY_FACTOR): New variable.
11922         (s390_sched_score): Decrease score for long-running instructions on
11923         wrong side.
11924         (s390_sched_variable_issue): Perform bookkeeping for long-running
11925         instructions.
11927 2017-10-18  Richard Biener  <rguenther@suse.de>
11929         * graphite-isl-ast-to-gimple.c (gcc_expression_from_isl_ast_expr_id):
11930         Simplify with removal of the parameter rename map.
11931         (set_rename): Likewise.
11932         (should_copy_to_new_region): Likewise.
11933         (graphite_copy_stmts_from_block): Likewise.
11934         (copy_bb_and_scalar_dependences): Remove initialization of
11935         unused copied_bb_map.
11936         (copy_def): Remove.
11937         (copy_internal_parameters): Likewise.
11938         (graphite_regenerate_ast_isl): Do not call copy_internal_parameters.
11939         * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
11940         Use INTEGRAL_TYPE_P.
11941         (parameter_index_in_region_1): Rename to ...
11942         (assign_parameter_index_in_region): ... this.  Assert we have
11943         a parameter we handle.
11944         (scan_tree_for_params): Adjust.
11945         * sese.h (parameter_rename_map_t): Remove.
11946         (struct sese_info_t): Remove unused parameter_rename_map and
11947         copied_bb_map members.
11948         * sese.c (new_sese_info): Adjust.
11949         (free_sese_info): Likewise.
11951 2017-10-18  Martin Liska  <mliska@suse.cz>
11953         PR sanitizer/82545
11954         * asan.c (asan_expand_poison_ifn): Do not put gimple stmt
11955         on an abnormal edge.
11957 2017-10-18  Sebastian Huber  <sebastian.huber@embedded-brains.de>
11959         * doc/invoke.texi (ffunction-sections and fdata-sections):
11960         Update.
11962 2017-10-17  Eric Botcazou  <ebotcazou@adacore.com>
11964         * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Bail out only if
11965         the use statement can throw internally.
11967 2017-10-17  Eric Botcazou  <ebotcazou@adacore.com>
11969         * config/visium/visium.c (visium_select_cc_mode): Return CCmode for
11970         any RTX present on the RHS of a SET.
11971         * compare-elim.c (try_eliminate_compare): Restore comment.
11973 2017-10-17  Jakub Jelinek  <jakub@redhat.com>
11975         * langhooks.h (struct lang_hooks): Document that tree_size langhook
11976         may be also called on tcc_type nodes.
11977         * langhooks.c (lhd_tree_size): Likewise.
11979 2017-10-17  David Malcolm  <dmalcolm@redhat.com>
11981         * gimple-ssa-sprintf.c (fmtwarn): Update for changed signature of
11982         format_warning_at_substring.
11983         (maybe_warn): Convert source_range * param to a location_t.  Pass
11984         UNKNOWN_LOCATION rather than NULL to fmtwarn.
11985         (format_directive): Remove code to extract source_ranges and
11986         source_range * in favor of just a location_t.
11987         (parse_directive): Pass UNKNOWN_LOCATION rather than NULL to
11988         fmtwarn.
11989         * substring-locations.c (format_warning_va): Convert
11990         source_range * param to a location_t.
11991         (format_warning_at_substring): Likewise.
11992         * substring-locations.h (format_warning_va): Likewise.
11993         (format_warning_at_substring): Likewise.
11995 2017-10-17  Jan Hubicka  <hubicka@ucw.cz>
11997         * target.h (enum vect_cost_for_stmt): Add vec_gather_load and
11998         vec_scatter_store
11999         * tree-vect-stmts.c (record_stmt_cost): Make difference between normal
12000         and scatter/gather ops.
12002         * aarch64/aarch64.c (aarch64_builtin_vectorization_cost): Add
12003         vec_gather_load and vec_scatter_store.
12004         * arm/arm.c (arm_builtin_vectorization_cost): Likewise.
12005         * powerpcspe/powerpcspe.c (rs6000_builtin_vectorization_cost): Likewise.
12006         * rs6000/rs6000.c (rs6000_builtin_vectorization_cost): Likewise.
12007         * s390/s390.c (s390_builtin_vectorization_cost): Likewise.
12008         * spu/spu.c (spu_builtin_vectorization_cost): Likewise.
12009         * i386/i386.c (x86_builtin_vectorization_cost): Likewise.
12011 2017-10-17  Uros Bizjak  <ubizjak@gmail.com>
12013         * reg-stack.c (compare_for_stack_reg): Add bool argument.
12014         Detect FTST instruction and handle its register pops.  Only pop
12015         second operand if can_pop_second_op is true.
12016         (subst_stack_regs_pat) <case COMPARE>: Detect FCOMI instruction to
12017         set can_pop_second_op to false in the compare_for_stack_reg call.
12019         * config/i386/i386.md (*cmpi<FPCMP:unord><MODEF:mode>): Only call
12020         output_fp_compare for stack register operands.
12021         * config/i386/i386.c (output_fp_compare): Do not output SSE compare
12022         instructions here.  Do not emit stack register pops here.  Assert
12023         that FCOMPP pops next to top stack register.  Rewrite function.
12025 2017-10-17  Nathan Sidwell  <nathan@acm.org>
12027         PR middle-end/82577
12028         * alias.c (compare_base_decls): Check HAS_DECL_ASSEMBLER_NAME_P,
12029         use DECL_ASSEMBLER_NAME_RAW.
12031         PR middle-end/82546
12032         * tree.c (tree_code_size): Reformat.  Punt to lang hook for unknown
12033         TYPE nodes.
12035 2017-10-17  Qing Zhao <qing.zhao@oracle.com>
12036             Wilco Dijkstra <wilco.dijkstra@arm.com>
12038         * builtins.c (expand_builtin_update_setjmp_buf): Add a
12039         converstion to Pmode from the buf_addr.
12041 2017-10-17  Richard Biener  <rguenther@suse.de>
12043         * graphite-dependences.c (scop_get_reads_and_writes): Change
12044         output parameters to references.
12046 2017-10-17  Jackson Woodruff  <jackson.woodruff@arm.com>
12048         PR 71026/tree-optimization
12049         * fold-const.c (distribute_real_division): Removed.
12050         (fold_binary_loc): Remove calls to distribute_real_divison.
12052 2017-10-17  Richard Biener  <rguenther@suse.de>
12054         * graphite-scop-detection.c
12055         (scop_detection::stmt_has_simple_data_refs_p): Always use
12056         the full nest as region.
12057         (try_generate_gimple_bb): Likewise.
12058         * sese.c (scalar_evolution_in_region): Simplify now that
12059         SCEV can handle instantiation in regions.
12060         * tree-scalar-evolution.c (instantiate_scev_name): Also instantiate
12061         in the non-loop part of a function if requested.
12063 2017-10-17  Richard Biener  <rguenther@suse.de>
12065         PR tree-optimization/82563
12066         * graphite-isl-ast-to-gimple.c (generate_entry_out_of_ssa_copies):
12067         New function.
12068         (graphite_regenerate_ast_isl): Call it.
12069         * graphite-scop-detection.c (build_scops): Remove entry edge split.
12071 2017-10-17  Jakub Jelinek  <jakub@redhat.com>
12073         PR tree-optimization/82549
12074         * fold-const.c (optimize_bit_field_compare, fold_truth_andor_1):
12075         Formatting fixes.  Instead of calling make_bit_field_ref with negative
12076         bitpos return 0.
12078 2017-10-17  Olga Makhotina  <olga.makhotina@intel.com>
12080         * config/i386/avx512dqintrin.h (_mm_mask_reduce_sd,
12081         _mm_maskz_reduce_sd, _mm_mask_reduce_ss,=20
12082         _mm_maskz_reduce_ss): New.
12083         * config/i386/i386-builtin.def (__builtin_ia32_reducesd_mask,
12084         __builtin_ia32_reducess_mask): Ditto..
12085         (__builtin_ia32_reducesd, __builtin_ia32_reducess): Remove.
12086         * config/i386/sse.md (reduces<mode>): Renamed to ...
12087         (reduces<mode><mask_scalar_name>): ... this.
12088         (vreduce<ssescalarmodesuffix>\t{%3, %2, %1, %0|%0, %1, %2, %3}):
12089         Changed to ...
12090         (vreduce<ssescalarmodesuffix>\t{%3, %2, %1, %0<mask_scalar_operand4>|
12091         %0<mask_scalar_operand4>, %1, %2, %3}): ... this.
12093 2017-10-16  David Malcolm  <dmalcolm@redhat.com>
12095         * Makefile.in (OBJS): Add unique-ptr-tests.o.
12096         * selftest-run-tests.c (selftest::run_tests): Call
12097         selftest::unique_ptr_tests_cc_tests.
12098         * selftest.h (selftest::unique_ptr_tests_cc_tests): New decl.
12099         * unique-ptr-tests.cc: New file.
12101 2017-10-16  Vladimir Makarov  <vmakarov@redhat.com>
12103         PR sanitizer/82353
12104         * lra.c (collect_non_operand_hard_regs): Don't ignore operator
12105         locations.
12106         * lra-lives.c (bb_killed_pseudos, bb_gen_pseudos): Move up.
12107         (make_hard_regno_born, make_hard_regno_dead): Update
12108         bb_killed_pseudos and bb_gen_pseudos for fixed regs.
12110 2017-10-16  Jeff Law  <law@redhat.com>
12112         * tree-ssa-dse.c (live_bytes_read): Fix thinko.
12114 2017-10-16  Jan Hubicka  <hubicka@ucw.cz>
12116         * x86-tune-costs.h (znver1_cost): Fix move cost tables.
12118 2017-10-16  Olivier Hainque  <hainque@adacore.com>
12120         * gcc/config.gcc (powerpc*-*-*spe*): Pick 8548 as the default
12121         with_cpu if we were configured for an e500v2 target cpu name.
12123 2017-10-16  Thomas Preud'homme  <thomas.preudhomme@arm.com>
12125         * config/arm/arm-cpus.in (cortex-m33): Add nodsp option.
12126         * doc/invoke.texi: Document +nodsp as a valid extension for
12127         -mcpu=cortex-m33.
12129 2017-10-16  Martin Liska  <mliska@suse.cz>
12131         * sbitmap.c (bitmap_bit_in_range_p_checking): New function.
12132         (test_set_range): Likewise.
12133         (test_range_functions): Rename to ...
12134         (test_bit_in_range): ... this.
12135         (sbitmap_c_tests): Add new test.
12137 2017-10-16  Tamar Christina  <tamar.christina@arm.com>
12139         * config/aarch64/arm_neon.h (vdot_u32, vdotq_u32, vdot_s32, vdotq_s32):
12140         New.
12141         (vdot_lane_u32, vdot_laneq_u32, vdotq_lane_u32, vdotq_laneq_u32): New.
12142         (vdot_lane_s32, vdot_laneq_s32, vdotq_lane_s32, vdotq_laneq_s32): New.
12144 2017-10-16  Tamar Christina  <tamar.christina@arm.com>
12146         * config/aarch64/aarch64-builtins.c
12147         (aarch64_types_quadopu_lane_qualifiers): New.
12148         (TYPES_QUADOPU_LANE): New.
12149         * config/aarch64/aarch64-simd.md (aarch64_<sur>dot<vsi2qi>): New.
12150         (<sur>dot_prod<vsi2qi>, aarch64_<sur>dot_lane<vsi2qi>): New.
12151         (aarch64_<sur>dot_laneq<vsi2qi>): New.
12152         * config/aarch64/aarch64-simd-builtins.def (sdot, udot): New.
12153         (sdot_lane, udot_lane, sdot_laneq, udot_laneq): New.
12154         * config/aarch64/iterators.md (sur): Add UNSPEC_SDOT, UNSPEC_UDOT.
12155         (Vdottype, DOTPROD): New.
12156         (sur): Add SDOT and UDOT.
12158 2017-10-16  Tamar Christina  <tamar.christina@arm.com>
12160         * config/aarch64/aarch64.h (AARCH64_FL_DOTPROD): New.
12161         (AARCH64_ISA_DOTPROD, TARGET_DOTPROD): New.
12162         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
12163         Add TARGET_DOTPROD.
12164         * config/aarch64/aarch64-option-extensions.def (dotprod): New.
12165         * config/aarch64/aarch64-cores.def (cortex-a55, cortex-a75):
12166         Enable TARGET_DOTPROD.
12167         (cortex-a75.cortex-a55): Likewise.
12168         * doc/invoke.texi (aarch64-feature-modifiers): Document dotprod.
12170 2017-10-16  Tamar Christina  <tamar.christina@arm.com>
12172         * config/arm/arm-builtins.c (arm_unsigned_uternop_qualifiers): New.
12173         (UTERNOP_QUALIFIERS, arm_umac_lane_qualifiers, UMAC_LANE_QUALIFIERS):
12174         New.
12175         * config/arm/arm_neon_builtins.def (sdot, udot, sdot_lane, udot_lane):
12176         New.
12177         * config/arm/iterators.md (DOTPROD, VSI2QI, vsi2qi): New.
12178         (UNSPEC_DOT_S, UNSPEC_DOT_U, opsuffix): New.
12179         * config/arm/neon.md (neon_<sup>dot<vsi2qi>): New.
12180         (neon_<sup>dot_lane<vsi2qi>, <sup>dot_prod<vsi2qi>): New.
12181         * config/arm/types.md (neon_dot, neon_dot_q): New.
12182         * config/arm/unspecs.md (sup): Add UNSPEC_DOT_S, UNSPEC_DOT_U.
12184 2017-10-16  Tamar Christina  <tamar.christina@arm.com>
12186         * config/arm/arm.h (TARGET_DOTPROD): New.
12187         * config/arm/arm.c (arm_arch_dotprod): New.
12188         (arm_option_reconfigure_globals): Add arm_arch_dotprod.
12189         * config/arm/arm-c.c (__ARM_FEATURE_DOTPROD): New.
12190         * config/arm/arm-cpus.in (armv8.2-a): Enabled +dotprod.
12191         (feature dotprod, group dotprod, ALL_SIMD_INTERNAL): New.
12192         (ALL_FPU_INTERNAL): Use ALL_SIMD_INTERNAL.
12193         * config/arm/t-multilib (v8_2_a_simd_variants): Add dotprod.
12194         * doc/invoke.texi (armv8.2-a): Document dotprod
12196 2017-10-14  Jan Hubicka  <hubicka@ucw.cz>
12198         * i386.c (ix86_vec_cost): New function.
12199         (ix86_rtx_costs): Handle vector operations better.
12200         * i386.h (struct processor_costs): Add sse_op, fmasd, fmass.
12201         * x86-tune-costs.h: Add new costs to all tables.
12203 2017-10-14  Jan Hubicka  <hubicka@ucw.cz>
12205         * i386.c (ix86_rtx_costs): Make difference between x87 and SSE
12206         operations.
12207         * i386.h (struct processor_costs): Add addss, mulss, mulsd, divss,
12208         divsd, sqrtss and sqrtsd
12209         * x86-tune-costs.h: Add new entries to all costs.
12210         (znver1_cost): Fix to match real instruction latencies.
12212 2017-10-14  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
12213             Michael Collison <michael.collison@arm.com>
12215         * compare-elim.c: Include emit-rtl.h.
12216         (can_merge_compare_into_arith): New function.
12217         (try_validate_parallel): Likewise.
12218         (try_merge_compare): Likewise.
12219         (try_eliminate_compare): Call the above when no previous clobber
12220         is available.
12221         (execute_compare_elim_after_reload): Add DF_UD_CHAIN and DF_DU_CHAIN
12222         dataflow problems.
12224 2017-10-14  Jakub Jelinek  <jakub@redhat.com>
12226         PR middle-end/62263
12227         PR middle-end/82498
12228         * tree-ssa-phiopt.c (value_replacement): Comment fix.  Handle
12229         up to 2 preparation statements for ASSIGN in MIDDLE_BB.
12231         PR middle-end/62263
12232         PR middle-end/82498
12233         * tree-ssa-forwprop.c (simplify_rotate): Allow def_arg1[N]
12234         to be any operand_equal_p operands.  For & (B - 1) require
12235         B to be power of 2.  Recognize
12236         (X << (Y & (B - 1))) | (X >> ((-Y) & (B - 1))) and similar patterns.
12238 2017-10-14  Uros Bizjak  <ubizjak@gmail.com>
12240         PR bootstrap/82553
12241         * optabs.c (expand_memory_blockage): Fix call of
12242         targetm.have_memory_blockage.
12244 2017-10-14  Jakub Jelinek  <jakub@redhat.com>
12246         PR bootstrap/82548
12247         * config.gcc (*-*-solaris2*, i[34567]86-*-cygwin*,
12248         x86_64-*-cygwin*, i[34567]86-*-mingw* | x86_64-*-mingw*): Append
12249         objects to extra_objs instead of overwriting it.
12251 2017-10-14  Uros Bizjak  <ubizjak@gmail.com>
12253         * config/i386/sync.md (FILD_ATOMIC/FIST_ATOMIC FP load peephole2):
12254         Use any_fp_register_operand as operand[3] predicate.  Simplify
12255         equality test for operands[2] and operands[4] memory location.
12256         (LDX_ATOMIC/STX_ATOMIC FP load peephole2): Ditto.
12257         (FILD_ATOMIC/FIST_ATOMIC FP load peephole2 with mem blockage): New.
12258         (LDX_ATOMIC/LDX_ATOMIC FP load peephole2 with mem blockage): Ditto.
12259         (FILD_ATOMIC/FIST_ATOMIC FP store peephole2): Use
12260         any_fp_register_operand as operand[1] predicate.  Simplify
12261         equality test for operands[0] and operands[3] memory location.
12262         (LDX_ATOMIC/STX_ATOMIC FP store peephole2): Ditto.
12263         (FILD_ATOMIC/FIST_ATOMIC FP store peephole2 with mem blockage): New.
12264         (LDX_ATOMIC/LDX_ATOMIC FP storepeephole2 with mem blockage): Ditto.
12266 2017-10-14  Uros Bizjak  <ubizjak@gmail.com>
12268         * target-insns.def: Add memory_blockage.
12269         * optabs.c (expand_memory_blockage): New function.
12270         (expand_asm_memory_barrier): Rename ...
12271         (expand_asm_memory_blockage): ... to this.
12272         (expand_mem_thread_fence): Call expand_memory_blockage
12273         instead of expand_asm_memory_barrier.
12274         (expand_mem_singnal_fence): Ditto.
12275         (expand_atomic_load): Ditto.
12276         (expand_atomic_store): Ditto.
12277         * doc/md.texi (Standard Pattern Names For Generation):
12278         Document memory_blockage instruction pattern.
12280 2017-10-13  Sebastian Perta  <sebastian.perta@renesas.com>
12282         * config/rl78/rl78.c (rl78_emit_libcall): New function.
12283         * config/rl78/rl78-protos.h (rl78_emit_libcall): New function.
12284         * config/rl78/rl78.md: New define_expand "adddi3".
12286 2017-10-13  Jan Hubicka  <hubicka@ucw.cz>
12288         * cfghooks.c (verify_flow_info): Disable check that all probabilities
12289         are set correctly.
12291 2017-10-13  Jeff Law  <law@redhat.com>
12293         * tree-ssa-reassoc.c (reassociate_bb): Clarify code slighly.
12295 2017-10-13  Jakub Jelinek  <jakub@redhat.com>
12297         PR target/82274
12298         * internal-fn.c (expand_mul_overflow): If both operands have
12299         the same highpart of -1 or 0 and the topmost bit of lowpart
12300         is different, overflow is if res <= 0 rather than res < 0.
12302 2017-10-13  Pat Haugen  <pthaugen@us.ibm.com>
12304         * config/rs6000/rs6000.c (rs6000_builtin_vectorization_cost): Remove
12305         TARGET_P9_VECTOR code for unaligned_load case.
12307 2017-10-13  Jan Hubicka  <hubicka@ucw.cz>
12309         * cfghooks.c (verify_flow_info): Check that edge probabilities are set.
12311 2017-10-13  Nathan Sidwell  <nathan@acm.org>
12313         * tree-core.h (tree_contains_struct): Make bool.
12314         * tree.c (tree_contains_struct): Likewise.
12315         * tree.h (MARK_TS_BASE): Remove do ... while (0) idiom.
12316         (MARK_TS_TYPED, MARK_TS_COMMON, MARK_TS_TYPE_COMMON,
12317         MARK_TS_TYPE_WITH_LANG_SPECIFIC, MARK_TS_DECL_MINIMAL,
12318         MARK_TS_DECL_COMMON, MARK_TS_DECL_WRTL, MARK_TS_DECL_WITH_VIS,
12319         MARK_TS_DECL_NON_COMMON): Likewise, use comma operator.
12321 2017-10-13  Richard Biener  <rguenther@suse.de>
12323         * graphite-isl-ast-to-gimple.c
12324         (translate_isl_ast_to_gimple::get_rename_from_scev): Remove unused
12325         parameters and dominance check.
12326         (translate_isl_ast_to_gimple::graphite_copy_stmts_from_block): Adjust.
12327         (translate_isl_ast_to_gimple::copy_bb_and_scalar_dependences): Likewise.
12328         (translate_isl_ast_to_gimple::graphite_regenerate_ast_isl):
12329         Do not update SSA form here or do intermediate IL verification.
12330         * graphite.c: Include tree-ssa.h and tree-into-ssa.h.
12331         (graphite_initialize): Remove check on the number of loops in
12332         the function and inline into graphite_transform_loops.
12333         (graphite_finalize): Inline into graphite_transform_loops.
12334         (graphite_transform_loops): Perform SSA update and IL verification
12335         here.
12336         * params.def (PARAM_GRAPHITE_MIN_LOOPS_PER_FUNCTION): Remove.
12338 2017-10-13  Richard Biener  <rguenther@suse.de>
12340         * graphite-isl-ast-to-gimple.c (max_mode_int_precision,
12341         graphite_expression_type_precision): Avoid global constructor
12342         by moving ...
12343         (translate_isl_ast_to_gimple::translate_isl_ast_to_gimple): Here.
12344         (translate_isl_ast_to_gimple::graphite_expr_type): Add type member.
12345         (translate_isl_ast_to_gimple::translate_isl_ast_node_for): Use it.
12346         (translate_isl_ast_to_gimple::build_iv_mapping): Likewise.
12347         (translate_isl_ast_to_gimple::graphite_create_new_guard): Likewise.
12348         * graphite-sese-to-poly.c (build_original_schedule): Return nothing.
12350 2017-10-13  H.J. Lu  <hongjiu.lu@intel.com>
12352         PR target/82499
12353         * config/i386/i386.h (ix86_red_zone_size): New.
12354         * config/i386/i386.md (push peephole2s): Replace
12355         "!ix86_using_red_zone ()" with "ix86_red_zone_size == 0".
12357 2017-10-13  Richard Sandiford  <richard.sandiford@linaro.org>
12358             Alan Hayward  <alan.hayward@arm.com>
12359             David Sherwood  <david.sherwood@arm.com>
12361         * combine.c (can_change_dest_mode): Reject changes in
12362         REGMODE_NATURAL_SIZE.
12364 2017-10-13  Richard Sandiford  <richard.sandiford@linaro.org>
12365             Alan Hayward  <alan.hayward@arm.com>
12366             David Sherwood  <david.sherwood@arm.com>
12368         * cfgexpand.c (expand_debug_expr): Use GET_MODE_UNIT_BITSIZE.
12369         (expand_debug_source_expr): Likewise.
12370         * combine.c (combine_simplify_rtx): Likewise.
12371         * cse.c (fold_rtx): Likewise.
12372         * fwprop.c (canonicalize_address): Likewise.
12373         * targhooks.c (default_shift_truncation_mask): Likewise.
12375 2017-10-13  Richard Sandiford  <richard.sandiford@linaro.org>
12376             Alan Hayward  <alan.hayward@arm.com>
12377             David Sherwood  <david.sherwood@arm.com>
12379         * optabs.c (add_equal_note): Use GET_MODE_UNIT_SIZE.
12380         (widened_mode): Likewise.
12381         (expand_unop): Likewise.
12382         * ree.c (transform_ifelse): Likewise.
12383         (merge_def_and_ext): Likewise.
12384         (combine_reaching_defs): Likewise.
12385         * simplify-rtx.c (simplify_unary_operation_1): Likewise.
12387 2017-10-13  Richard Sandiford  <richard.sandiford@linaro.org>
12388             Alan Hayward  <alan.hayward@arm.com>
12389             David Sherwood  <david.sherwood@arm.com>
12391         * caller-save.c (replace_reg_with_saved_mem): Use byte_lowpart_offset.
12392         * combine.c (gen_lowpart_for_combine): Likewise.
12393         * dwarf2out.c (rtl_for_decl_location): Likewise.
12394         * final.c (alter_subreg): Likewise.
12395         * rtlhooks.c (gen_lowpart_general): Likewise.
12396         (gen_lowpart_if_possible): Likewise.
12398 2017-10-13  Richard Sandiford  <richard.sandiford@linaro.org>
12399             Alan Hayward  <alan.hayward@arm.com>
12400             David Sherwood  <david.sherwood@arm.com>
12402         * calls.c (expand_call): Use subreg_lowpart_offset.
12403         * cse.c (cse_insn): Likewise.
12404         * regcprop.c (copy_value): Likewise.
12405         (copyprop_hardreg_forward_1): Likewise.
12407 2017-10-13  Jakub Jelinek  <jakub@redhat.com>
12409         PR target/82524
12410         * config/i386/i386.md (addqi_ext_1, andqi_ext_1,
12411         *andqi_ext_1_cc, *<code>qi_ext_1, *xorqi_ext_1_cc): Change
12412         =Q constraints to +Q and into insn condition add check
12413         that operands[0] and operands[1] are equal.
12414         (*addqi_ext_2, *andqi_ext_2, *<code>qi_ext_2): Change
12415         =Q constraints to +Q and into insn condition add check
12416         that operands[0] is equal to either operands[1] or operands[2].
12418         PR target/82498
12419         * fold-const.c (fold_binary_loc) <bit_rotate>: Code cleanups,
12420         instead of handling MINUS_EXPR twice (once for each argument),
12421         canonicalize operand order and handle just once, use rtype where
12422         possible.  Handle (A << B) | (A >> (-B & (Z - 1))).
12424         PR target/82498
12425         * config/i386/ia32intrin.h (__rold, __rord, __rolq, __rorq): Allow
12426         any values of __C while still being pattern recognizable as a simple
12427         rotate instruction.
12429 2017-10-13  Richard Biener  <rguenther@suse.de>
12431         PR tree-optimization/82451
12432         Revert
12433         2017-10-02  Richard Biener  <rguenther@suse.de>
12435         PR tree-optimization/82355
12436         * graphite-isl-ast-to-gimple.c (build_iv_mapping): Also build
12437         a mapping for the enclosing loop but avoid generating one for
12438         the loop tree root.
12439         (copy_bb_and_scalar_dependences): Remove premature codegen
12440         error on PHIs in blocks duplicated into multiple places.
12441         * graphite-scop-detection.c
12442         (scop_detection::stmt_has_simple_data_refs_p): For a loop not
12443         in the region use it as loop and nest to analyze the DR in.
12444         (try_generate_gimple_bb): Likewise.
12445         * graphite-sese-to-poly.c (extract_affine_chrec): Adjust.
12446         (add_loop_constraints): For blocks in a loop not in the region
12447         create a dimension with a single iteration.
12448         * sese.h (gbb_loop_at_index): Remove assert.
12450         * cfgloop.c (loop_preheader_edge): For the loop tree root
12451         return the single successor of the entry block.
12452         * graphite-isl-ast-to-gimple.c (graphite_regenerate_ast_isl):
12453         Reset the SCEV hashtable and niters.
12454         * graphite-scop-detection.c
12455         (scop_detection::graphite_can_represent_scev): Add SCOP parameter,
12456         assert that we only have POLYNOMIAL_CHREC that vary in loops
12457         contained in the region.
12458         (scop_detection::graphite_can_represent_expr): Adjust.
12459         (scop_detection::stmt_has_simple_data_refs_p): For loops
12460         not in the region set loop to NULL.  The nest is now the
12461         entry edge to the region.
12462         (try_generate_gimple_bb): Likewise.
12463         * sese.c (scalar_evolution_in_region): Adjust for
12464         instantiate_scev change.
12465         * tree-data-ref.h (graphite_find_data_references_in_stmt):
12466         Make nest parameter the edge into the region.
12467         (create_data_ref): Likewise.
12468         * tree-data-ref.c (dr_analyze_indices): Make nest parameter an
12469         entry edge into a region and adjust instantiate_scev calls.
12470         (create_data_ref): Likewise.
12471         (graphite_find_data_references_in_stmt): Likewise.
12472         (find_data_references_in_stmt): Pass the loop preheader edge
12473         from the nest argument.
12474         * tree-scalar-evolution.h (instantiate_scev): Make instantiate_below
12475         parameter the edge into the region.
12476         (instantiate_parameters): Use the loop preheader edge as entry.
12477         * tree-scalar-evolution.c (analyze_scalar_evolution): Handle
12478         NULL loop.
12479         (get_instantiated_value_entry): Make instantiate_below parameter
12480         the edge into the region.
12481         (instantiate_scev_name): Likewise.  Adjust dominance checks,
12482         when we cannot use loop-based instantiation instantiate by
12483         walking use-def chains.
12484         (instantiate_scev_poly): Adjust.
12485         (instantiate_scev_binary): Likewise.
12486         (instantiate_scev_convert): Likewise.
12487         (instantiate_scev_not): Likewise.
12488         (instantiate_array_ref): Remove.
12489         (instantiate_scev_3): Likewise.
12490         (instantiate_scev_2): Likewise.
12491         (instantiate_scev_1): Likewise.
12492         (instantiate_scev_r): Do not blindly handle N-operand trees.
12493         Do not instantiate array-refs.  Handle all constants and invariants.
12494         (instantiate_scev): Make instantiate_below parameter
12495         the edge into the region.
12496         (resolve_mixers): Use the loop preheader edge for the region
12497         parameter to instantiate_scev_r.
12498         * tree-ssa-loop-prefetch.c (determine_loop_nest_reuse): Adjust.
12500 2017-10-13  Richard Biener  <rguenther@suse.de>
12502         PR tree-optimization/82525
12503         * graphite-isl-ast-to-gimple.c
12504         (translate_isl_ast_to_gimple::widest_int_from_isl_expr_int): Split
12505         out from ...
12506         (translate_isl_ast_to_gimple::gcc_expression_from_isl_expr_int): Here.
12507         Fail code generation when we cannot represent the isl integer.
12508         (binary_op_to_tree): Elide modulo operations that are no-ops
12509         in the type we code generate.  Remove now superfluous code
12510         generation errors.
12512 2017-10-13  Richard Biener  <rguenther@suse.de>
12514         * graphite-scop-detection.c (loop_ivs_can_be_represented): Remove.
12515         (scop_detection::harmful_loop_in_region): Remove premature
12516         IV type restriction.
12517         (scop_detection::graphite_can_represent_scev): We can handle
12518         pointer IVs just fine.
12520 2017-10-13  Alan Modra  <amodra@gmail.com>
12522         * doc/extend.texi (Extended Asm <Clobbers>): Rename to
12523         "Clobbers and Scratch Registers".  Add paragraph on
12524         alternative to clobbers for scratch registers and OpenBLAS
12525         example.
12527 2017-10-13  Alan Modra  <amodra@gmail.com>
12529         * doc/extend.texi (Clobbers): Correct vax example.  Delete old
12530         example of a memory input for a string of known length.  Move
12531         commentary out of table.  Add a number of new examples
12532         covering array memory inputs.
12534 2017-10-12  Martin Liska  <mliska@suse.cz>
12536         PR tree-optimization/82493
12537         * sbitmap.c (bitmap_bit_in_range_p): Fix the implementation.
12538         (test_range_functions): New function.
12539         (sbitmap_c_tests): Likewise.
12540         * selftest-run-tests.c (selftest::run_tests): Run new tests.
12541         * selftest.h (sbitmap_c_tests): New function.
12543         * tree-ssa-dse.c (live_bytes_read): Fix thinko.
12545 2017-10-12  Michael Meissner  <meissner@linux.vnet.ibm.com>
12547         * config/rs6000/amo.h: Fix spacing issue.
12549 2017-10-12  Jakub Jelinek  <jakub@redhat.com>
12551         PR target/82498
12552         * config/i386/i386.md (*ashl<mode>3_mask_1,
12553         *<shift_insn><mode>3_mask_1, *<rotate_insn><mode>3_mask_1,
12554         *<btsc><mode>_mask_1, *btr<mode>_mask_1): New define_insn_and_split
12555         patterns.
12557 2017-10-12  Jan Hubicka  <hubicka@ucw.cz>
12559         * profile-count.h (safe_scale_64bit): Fix GCC4.x path.
12560         (profile_probability): Set max_probability
12561         to (uint32_t) 1 << (n_bits - 2) and update accessors to avoid overlfows
12562         in temporaries.
12563         * profile-count.c (profile_probability::differs_from_p): Do not
12564         rely on max_probaiblity == 10000
12566 2017-10-12  Jeff Law  <law@redhat.com>
12568         * tree-ssa-dse.c (valid_ao_ref_for_dse): Reject ao_refs with
12569         negative offsets.
12571 2017-10-12  Martin Sebor  <msebor@redhat.com>
12573         PR other/82301
12574         PR c/82435
12575         * cgraphunit.c (maybe_diag_incompatible_alias): New function.
12576         (handle_alias_pairs): Call it.
12577         * common.opt (-Wattribute-alias): New option.
12578         * doc/extend.texi (ifunc attribute): Discuss C++ specifics.
12579         * doc/invoke.texi (-Wattribute-alias): Document.
12581 2017-10-12  Vladimir Makarov  <vmakarov@redhat.com>
12583         Revert
12584         2017-10-11  Vladimir Makarov  <vmakarov@redhat.com>
12585         PR sanitizer/82353
12586         * lra.c (collect_non_operand_hard_regs): Don't ignore operator
12587         locations.
12588         * lra-lives.c (bb_killed_pseudos, bb_gen_pseudos): Move up.
12589         (make_hard_regno_born, make_hard_regno_dead): Update
12590         bb_killed_pseudos and bb_gen_pseudos.
12592 2017-10-12  Jan Hubicka  <hubicka@ucw.cz>
12594         * config/i386/x86-tune-sched.c (ix86_adjust_cost): Fix Zen support.
12596 2017-10-12  Uros Bizjak  <ubizjak@gmail.com>
12598         * config/alpha/alpha.c (alpha_split_conditional_move):
12599         Use std::swap instead of manually swapping.
12600         (alpha_stdarg_optimize_hook): Ditto.
12601         (alpha_canonicalize_comparison): Ditto.
12603 2017-10-12  Bin Cheng  <bin.cheng@arm.com>
12605         * tree-loop-distribution.c (struct builtin_info): New struct.
12606         (struct partition): Refactor fields into struct builtin_info.
12607         (partition_free): Free struct builtin_info.
12608         (build_size_arg_loc, build_addr_arg_loc): Delete.
12609         (generate_memset_builtin, generate_memcpy_builtin): Get memory range
12610         information from struct builtin_info.
12611         (find_single_drs): New function refactored from classify_partition.
12612         Also moved builtin validity checks to this function.
12613         (compute_access_range, alloc_builtin): New functions.
12614         (classify_builtin_st, classify_builtin_ldst): New functions.
12615         (classify_partition): Refactor code into functions find_single_drs,
12616         classify_builtin_st and classify_builtin_ldst.
12617         (distribute_loop): Don't do runtime alias check when distributing
12618         loop nest.
12619         (find_seed_stmts_for_distribution): New function.
12620         (pass_loop_distribution::execute): Refactor code finding seed
12621         stmts into above function.  Support distribution for the innermost
12622         two-level loop nest.  Adjust dump information.
12624 2017-10-12  Bin Cheng  <bin.cheng@arm.com>
12626         * tree-loop-distribution.c: Adjust the general comment.
12627         (NUM_PARTITION_THRESHOLD): New macro.
12628         (ssa_name_has_uses_outside_loop_p): Support loop nest distribution.
12629         (classify_partition): Skip builtin pattern of loop nest's inner loop.
12630         (merge_dep_scc_partitions): New parameter ignore_alias_p and use it
12631         in call to build_partition_graph.
12632         (finalize_partitions): New parameter.  Make loop distribution more
12633         conservative by fusing more partitions.
12634         (distribute_loop): Don't do runtime alias check in case of loop nest
12635         distribution.
12636         (find_seed_stmts_for_distribution): New function.
12637         (prepare_perfect_loop_nest): New function.
12638         (pass_loop_distribution::execute): Refactor code finding seed stmts
12639         and loop nest into above functions.  Support loop nest distribution.
12640         Adjust dump information accordingly.
12642 2017-10-12  Bin Cheng  <bin.cheng@arm.com>
12644         * tree-loop-distribution.c (break_alias_scc_partitions): Add comment
12645         and set PTYPE_SEQUENTIAL for merged partition.
12647 2017-10-12  Richard Biener  <rguenther@suse.de>
12649         PR tree-optimization/69728
12650         Revert
12651         2017-09-19  Richard Biener  <rguenther@suse.de>
12653         PR tree-optimization/69728
12654         * graphite-sese-to-poly.c (schedule_error): New global.
12655         (add_loop_schedule): Handle empty domain by failing the
12656         schedule.
12657         (build_original_schedule): Handle schedule_error.
12659         * graphite-sese-to-poly.c (add_loop_schedule): Handle empty
12660         domain by returning an unchanged schedule.
12662 2017-10-12  Jakub Jelinek  <jakub@redhat.com>
12664         * genrecog.c (validate_pattern): For VEC_SELECT verify that
12665         CONST_INT selectors are 0 to GET_MODE_NUNITS (imode) - 1.
12667 2017-10-12  Aldy Hernandez  <aldyh@redhat.com>
12669         * Makefile.in (TAGS): Merge all the *.def files into one pattern.
12670         Handle params.def.
12672 2017-10-12  Jakub Jelinek  <jakub@redhat.com>
12674         PR c++/82159
12675         * expr.c (store_field): Don't optimize away bitsize == 0 store
12676         from CALL_EXPR with addressable return type.
12678 2017-10-11  Segher Boessenkool  <segher@kernel.crashing.org>
12680         * config/rs6000/rs6000.h (TARGET_ISEL64): Delete.
12681         * config/rs6000/rs6000.md (sel): Delete mode attribute.
12682         (mov<mode>cc, isel_signed_<mode>, isel_unsigned_<mode>,
12683         *isel_reversed_signed_<mode>, *isel_reversed_unsigned_<mode>): Use
12684         TARGET_ISEL instead of TARGET_ISEL<sel>.
12686 2017-10-11  David Edelsohn  <dje.gcc@gmail.com>
12688         * config/rs6000/rs6000.c
12689         (rs6000_xcoff_asm_output_aligned_decl_common): Test for NULL decl.
12691 2017-10-11  Segher Boessenkool  <segher@kernel.crashing.org>
12693         * config/rs6000/predicates.md (zero_constant, all_ones_constant):
12694         Move up in file.
12695         (reg_or_cint_operand): Fix comment.
12696         (reg_or_zero_operand): New predicate.
12697         * config/rs6000/rs6000-protos.h (output_isel): Delete.
12698         * config/rs6000/rs6000.c (output_isel): Delete.
12699         * config/rs6000/rs6000.md (isel_signed_<mode>): Use reg_or_zero_operand
12700         instead of reg_or_cint_operand.  Output instruction directly (not via
12701         output_isel).
12702         (isel_unsigned_<mode>): Ditto.
12703         (*isel_reversed_signed_<mode>): Use reg_or_zero_operand instead of
12704         gpc_reg_operand.  Add an instruction alternative for this.  Output
12705         instruction directly.
12706         (*isel_reversed_unsigned_<mode>): Ditto.
12708 2017-10-11  Uros Bizjak  <ubizjak@gmail.com>
12710         * config/i386/i386.c (ix86_canonicalize_comparison): New function.
12711         (TARGET_CANONICALIZE_COMPARISON): Define.
12713 2017-10-11  Qing Zhao  <qing.zhao@oracle.com>
12715         PR target/81422
12716         * config/aarch64/aarch64.c (aarch64_load_symref_appropriately):
12717         Check whether the dest is REG before adding REG_EQUIV note.
12719 2017-10-11  Vladimir Makarov  <vmakarov@redhat.com>
12721         PR sanitizer/82353
12722         * lra.c (collect_non_operand_hard_regs): Don't ignore operator
12723         locations.
12724         * lra-lives.c (bb_killed_pseudos, bb_gen_pseudos): Move up.
12725         (make_hard_regno_born, make_hard_regno_dead): Update
12726         bb_killed_pseudos and bb_gen_pseudos.
12728 2017-10-11  Nathan Sidwell  <nathan@acm.org>
12730         * incpath.h (enum incpath_kind): Name enum, prefix values.
12731         (add_path, add_cpp_dir_path, get_added_cpp_dirs): Use incpath_kind.
12732         * incpath.c (heads, tails): Use INC_MAX.
12733         (add_env_var_paths, add_standard_paths): Use incpath_kind.
12734         (merge_include_chains, split_quote_chain,
12735         register_include_chains): Update incpath_kind names.
12736         (add_cpp_dir_path, add_path, get_added_cpp_dirs): Use incpath_kind.
12737         * config/darwin-c.c (add_system_framework_path): Update incpath_kind
12738         names.
12739         (add_framework_path, darwin_register_objc_includes): Likewise.
12740         * config/vms/vms-c.c (vms_c_register_includes): Likewise.
12742 2017-10-11  Uros Bizjak  <ubizjak@gmail.com>
12744         * config/i386/i386.md (*cmp<X87MODEF:mode>_<SWI24:mode>_i387):
12745         Do not use float_operator operator predicate.
12746         (*cmp<X87MODEF:mode>_<SWI24:mode>_cc_i387): Ditto.
12747         * config/i386/predicates.md (float_operator): Remove predicate.
12749 2017-10-11  Uros Bizjak  <ubizjak@gmail.com>
12751         * config/i386/i386.md (*jcc<mode>_0_i387): Remove insn pattern.
12752         (*jccxf_i387): Ditto.
12753         (*jcc<mode>_i387): Ditto.
12754         (*jccu<mode>_i387): Ditto.
12755         (*jcc<X87MODEF:mode>_<SWI24:mode>_i387): Ditto.
12756         (*jcc_*_i387 splitters): Remove.
12757         * config/i386/i386-protos.h (ix86_split_fp_branch): Remove prototype.
12758         * config/i386/i386.c (ix86_split_fp_branch): Remove.
12759         * config/i386/predicates.md (ix86_swapped_fp_comparison_operator):
12760         Remove predicate.
12762 2017-10-11  Jan Hubicka  <hubicka@ucw.cz>
12764         * profile-count.h (slow_safe_scale_64bit): New function.
12765         (safe_scale_64bit): New inline.
12766         (profile_count::max_safe_multiplier): Remove; use safe_scale_64bit.
12767         * profile-count.c: Include wide-int.h
12768         (slow_safe_scale_64bit): New.
12770 2017-10-11  Nathan Sidwell  <nathan@acm.org>
12772         * tree.h (DECL_ASSEMBLER_NAME_SET_P): Don't check
12773         HAS_DECL_ASSEMBLER_NAME_P.
12774         * gimple-expr.c (gimple_decl_printable_name: Check
12775         HAS_DECL_ASSEMBLER_NAME_P too.
12776         * ipa-utils.h (type_in_anonymous_namespace_p): Check
12777         DECL_ASSEMBLER_NAME_SET_P of TYPE_NAME.
12778         (odr_type_p): No need to assert TYPE_NAME is a TYPE_DECL.
12779         * passes.c (rest_of_decl_compilation): Check
12780         HAS_DECL_ASSEMBLER_NAME_P too.
12781         * recog.c (verify_changes): Likewise.
12782         * tree-pretty-print.c (dump_decl_name): Likewise.
12783         * tree-ssa-structalias.c (alias_get_name): Likewise.  Reimplement.
12785         * tree.h (DECL_ASSEMBLER_NAME_RAW): New.
12786         (SET_DECL_ASSEMBLER_NAME): Use it.
12787         (DECL_ASSEMBLER_NAME_SET_P): Likewise.
12788         (COPY_DECL_ASSEMBLER_NAME): Likewise.
12789         * tree.c (decl_assembler_name): Use DECL_ASSEMBLER_NAME_RAW.
12791 2017-10-11  Jan Hubicka  <hubicka@ucw.cz>
12793         * config.gcc (i386, x86_64): Add extra objects.
12794         * config/i386/i386-protos.h (ix86_rip_relative_addr_p): Declare.
12795         (ix86_min_insn_size): Declare.
12796         (ix86_issue_rate): Declare.
12797         (ix86_adjust_cost): Declare.
12798         (ia32_multipass_dfa_lookahead): Declare.
12799         (ix86_macro_fusion_p): Declare.
12800         (ix86_macro_fusion_pair_p): Declare.
12801         (ix86_bd_has_dispatch): Declare.
12802         (ix86_bd_do_dispatch): Declare.
12803         (ix86_core2i7_init_hooks): Declare.
12804         (ix86_atom_sched_reorder): Declare.
12805         * config/i386/i386.c Move all CPU cost tables to x86-tune-costs.h.
12806         (COSTS_N_BYTES): Move to x86-tune-costs.h.
12807         (DUMMY_STRINGOP_ALGS):Move to x86-tune-costs.h.
12808         (rip_relative_addr_p): Rename to ...
12809         (ix86_rip_relative_addr_p): ... this one; export.
12810         (memory_address_length): Update.
12811         (ix86_issue_rate): Move to x86-tune-sched.c.
12812         (ix86_flags_dependent): Move to x86-tune-sched.c.
12813         (ix86_agi_dependent): Move to x86-tune-sched.c.
12814         (exact_dependency_1): Move to x86-tune-sched.c.
12815         (exact_store_load_dependency): Move to x86-tune-sched.c.
12816         (ix86_adjust_cost): Move to x86-tune-sched.c.
12817         (ia32_multipass_dfa_lookahead): Move to x86-tune-sched.c.
12818         (ix86_macro_fusion_p): Move to x86-tune-sched.c.
12819         (ix86_macro_fusion_pair_p): Move to x86-tune-sched.c.
12820         (do_reorder_for_imul): Move to x86-tune-sched-atom.c.
12821         (swap_top_of_ready_list): Move to x86-tune-sched-atom.c.
12822         (ix86_sched_reorder): Move to x86-tune-sched-atom.c.
12823         (core2i7_first_cycle_multipass_init): Move to x86-tune-sched-core.c.
12824         (core2i7_dfa_post_advance_cycle): Move to x86-tune-sched-core.c.
12825         (min_insn_size): Rename to ...
12826         (ix86_min_insn_size): ... this one; export.
12827         (core2i7_first_cycle_multipass_begin): Move to x86-tune-sched-core.c.
12828         (core2i7_first_cycle_multipass_issue): Move to x86-tune-sched-core.c.
12829         (core2i7_first_cycle_multipass_backtrack): Move to
12830         x86-tune-sched-core.c.
12831         (core2i7_first_cycle_multipass_end): Move to x86-tune-sched-core.c.
12832         (core2i7_first_cycle_multipass_fini): Move to x86-tune-sched-core.c.
12833         (ix86_sched_init_global): Break up logic to ix86_core2i7_init_hooks.
12834         (ix86_avoid_jump_mispredicts): Update.
12835         (TARGET_SCHED_DISPATCH): Move to ix86-tune-sched-bd.c.
12836         (TARGET_SCHED_DISPATCH_DO): Move to ix86-tune-sched-bd.c.
12837         (TARGET_SCHED_REORDER): Move to ix86-tune-sched-bd.c.
12838         (DISPATCH_WINDOW_SIZE): Move to ix86-tune-sched-bd.c.
12839         (MAX_DISPATCH_WINDOWS): Move to ix86-tune-sched-bd.c.
12840         (MAX_INSN): Move to ix86-tune-sched-bd.c.
12841         (MAX_IMM): Move to ix86-tune-sched-bd.c.
12842         (MAX_IMM_SIZE): Move to ix86-tune-sched-bd.c.
12843         (MAX_IMM_32): Move to ix86-tune-sched-bd.c.
12844         (MAX_IMM_64): Move to ix86-tune-sched-bd.c.
12845         (MAX_LOAD): Move to ix86-tune-sched-bd.c.
12846         (MAX_STORE): Move to ix86-tune-sched-bd.c.
12847         (BIG): Move to ix86-tune-sched-bd.c.
12848         (enum dispatch_group): Move to ix86-tune-sched-bd.c.
12849         (enum insn_path): Move to ix86-tune-sched-bd.c.
12850         (get_mem_group): Move to ix86-tune-sched-bd.c.
12851         (is_cmp): Move to ix86-tune-sched-bd.c.
12852         (dispatch_violation): Move to ix86-tune-sched-bd.c.
12853         (is_branch): Move to ix86-tune-sched-bd.c.
12854         (is_prefetch): Move to ix86-tune-sched-bd.c.
12855         (init_window): Move to ix86-tune-sched-bd.c.
12856         (allocate_window): Move to ix86-tune-sched-bd.c.
12857         (init_dispatch_sched): Move to ix86-tune-sched-bd.c.
12858         (is_end_basic_block): Move to ix86-tune-sched-bd.c.
12859         (process_end_window): Move to ix86-tune-sched-bd.c.
12860         (allocate_next_window): Move to ix86-tune-sched-bd.c.
12861         (find_constant): Move to ix86-tune-sched-bd.c.
12862         (get_num_immediates): Move to ix86-tune-sched-bd.c.
12863         (has_immediate): Move to ix86-tune-sched-bd.c.
12864         (get_insn_path): Move to ix86-tune-sched-bd.c.
12865         (get_insn_group): Move to ix86-tune-sched-bd.c.
12866         (count_num_restricted): Move to ix86-tune-sched-bd.c.
12867         (fits_dispatch_window): Move to ix86-tune-sched-bd.c.
12868         (add_insn_window): Move to ix86-tune-sched-bd.c.
12869         (add_to_dispatch_window): Move to ix86-tune-sched-bd.c.
12870         (debug_dispatch_window_file): Move to ix86-tune-sched-bd.c.
12871         (debug_dispatch_window): Move to ix86-tune-sched-bd.c.
12872         (debug_insn_dispatch_info_file): Move to ix86-tune-sched-bd.c.
12873         (debug_ready_dispatch): Move to ix86-tune-sched-bd.c.
12874         (do_dispatch): Move to ix86-tune-sched-bd.c.
12875         (has_dispatch): Move to ix86-tune-sched-bd.c.
12876         * config/i386/t-i386: Add new object files.
12877         * config/i386/x86-tune-costs.h: New file.
12878         * config/i386/x86-tune-sched-atom.c: New file.
12879         * config/i386/x86-tune-sched-bd.c: New file.
12880         * config/i386/x86-tune-sched-core.c: New file.
12881         * config/i386/x86-tune-sched.c: New file.
12883 2017-10-11  Liu Hao  <lh_mouse@126.com>
12885         * pretty-print.c [_WIN32] (colorize_init): Remove.  Use
12886         the generic version below instead.
12887         (should_colorize): Recognize Windows consoles as terminals
12888         for MinGW targets.
12889         * pretty-print.c [__MINGW32__] (write_all): New function.
12890         [__MINGW32__] (find_esc_head): Likewise.
12891         [__MINGW32__] (find_esc_terminator): Likewise.
12892         [__MINGW32__] (eat_esc_sequence): Likewise.
12893         [__MINGW32__] (mingw_ansi_fputs): New function that handles
12894         ANSI escape codes.
12895         (pp_write_text_to_stream): Use mingw_ansi_fputs instead of fputs
12896         for MinGW targets.
12898 2017-10-11  Richard Biener  <rguenther@suse.de>
12900         * tree-ssa-loop-niter.c (infer_loop_bounds_from_pointer_arith):
12901         Properly call analyze_scalar_evolution with the loop of the stmt.
12903 2017-10-11  Richard Biener  <rguenther@suse.de>
12905         * tree.def (POLYNOMIAL_CHREC): Remove CHREC_VARIABLE tree operand.
12906         * tree-core.h (tree_base): Add chrec_var union member.
12907         * tree.h (CHREC_VAR): Remove.
12908         (CHREC_LEFT, CHREC_RIGHT, CHREC_VARIABLE): Adjust.
12909         * tree-chrec.h (build_polynomial_chrec): Adjust.
12910         * tree-chrec.c (reset_evolution_in_loop): Use build_polynomial_chrec.
12911         * tree-pretty-print.c (dump_generic_node): Use CHREC_VARIABLE.
12913 2017-10-11  Marc Glisse  <marc.glisse@inria.fr>
12915         * fold-const.c (fold_binary_loc) [X +- Y CMP X]: Move ...
12916         * match.pd: ... here.
12917         ((T) X == (T) Y): Relax condition.
12919 2017-10-11  Bin Cheng  <bin.cheng@arm.com>
12921         PR tree-optimization/82472
12922         * tree-loop-distribution.c (sort_partitions_by_post_order): Refine
12923         comment.
12924         (break_alias_scc_partitions): Update postorder number.
12926 2017-10-11  Martin Liska  <mliska@suse.cz>
12928         PR sanitizer/82490
12929         * opts.c (parse_no_sanitize_attribute): Do not use error_value
12930         variable.
12931         * opts.h (parse_no_sanitize_attribute): Remove last argument.
12933 2017-10-11  Martin Liska  <mliska@suse.cz>
12935         * print-rtl.c (print_insn): Move declaration of idbuf
12936         to same scope as name.
12938 2017-10-11  Martin Liska  <mliska@suse.cz>
12940         Revert r253637:
12942         PR sanitizer/82484
12943         * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
12944         volatile arguments.
12946 2017-10-11  Martin Liska  <mliska@suse.cz>
12948         PR sanitizer/82484
12949         * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
12950         volatile arguments.
12952 2017-10-11  Adhemerval Zanella  <adhemerval.zanella@linaro.org>
12954         * config.gcc (default_gnu_indirect_function): Default to yes for
12955         arm*-*-linux* with glibc.
12957 2017-10-11  Richard Biener  <rguenther@suse.de>
12959         * tree-scalar-evolution.c (get_scalar_evolution): Handle
12960         default-defs and types we do not want to analyze.
12961         (interpret_loop_phi): Replace unreachable code with an assert.
12962         (compute_scalar_evolution_in_loop): Remove and inline ...
12963         (analyze_scalar_evolution_1): ... here, replacing condition with
12964         what makes the intent clearer.  Remove handling of cases
12965         get_scalar_evolution now handles.
12967 2017-10-10  Jim Wilson  <wilson@tuliptree.org>
12969         PR rtl-optimization/81434
12970         * haifa-sched.c (prune_ready_list): Init min_cost_group to 0.  Update
12971         comment for main loop.  In sched_group_found if, also add checks for
12972         pass and min_cost_group.
12974 2017-10-10  Segher Boessenkool  <segher@kernel.crashing.org>
12976         * config/rs6000/rs6000.c (TARGET_INSN_COST): New.
12977         (rs6000_insn_cost): New function.
12978         * config/rs6000/rs6000.md (cost): New attribute.
12980 2017-10-10  Jakub Jelinek  <jakub@redhat.com>
12981             H.J. Lu  <hongjiu.lu@intel.com>
12983         PR target/79565
12984         PR target/82483
12985         * config/i386/i386.c (ix86_init_mmx_sse_builtins): Add
12986         OPTION_MASK_ISA_MMX for __builtin_ia32_maskmovq,
12987         __builtin_ia32_vec_ext_v4hi and __builtin_ia32_vec_set_v4hi.
12988         (ix86_expand_builtin): Treat OPTION_MASK_ISA_MMX similarly
12989         to OPTION_MASK_ISA_AVX512VL - builtins that have both
12990         OPTION_MASK_ISA_MMX and some other bit set require both
12991         mmx and the ISAs without the mmx bit.
12992         * config/i386/i386-builtin.def (__builtin_ia32_cvtps2pi,
12993         __builtin_ia32_cvttps2pi, __builtin_ia32_cvtpi2ps,
12994         __builtin_ia32_pavgb, __builtin_ia32_pavgw, __builtin_ia32_pmulhuw,
12995         __builtin_ia32_pmaxub, __builtin_ia32_pmaxsw, __builtin_ia32_pminub,
12996         __builtin_ia32_pminsw, __builtin_ia32_psadbw, __builtin_ia32_pmovmskb,
12997         __builtin_ia32_pshufw, __builtin_ia32_cvtpd2pi,
12998         __builtin_ia32_cvttpd2pi, __builtin_ia32_cvtpi2pd,
12999         __builtin_ia32_pmuludq, __builtin_ia32_pabsb, __builtin_ia32_pabsw,
13000         __builtin_ia32_pabsd, __builtin_ia32_phaddw, __builtin_ia32_phaddd,
13001         __builtin_ia32_phaddsw, __builtin_ia32_phsubw, __builtin_ia32_phsubd,
13002         __builtin_ia32_phsubsw, __builtin_ia32_pmaddubsw,
13003         __builtin_ia32_pmulhrsw, __builtin_ia32_pshufb, __builtin_ia32_psignb,
13004         __builtin_ia32_psignw, __builtin_ia32_psignd, __builtin_ia32_movntq,
13005         __builtin_ia32_paddq, __builtin_ia32_psubq, __builtin_ia32_palignr):
13006         Add OPTION_MASK_ISA_MMX.
13008 2017-10-10  Andreas Tobler  <andreast@gcc.gnu.org>
13010         * config.gcc (armv7*-*-freebsd*): New target.
13011         (armv6*-*-freebsd*): Remove obsolete TARGET_FREEBSD_ARMv6 define.
13013 2017-10-10  Jan Hubicka  <hubicka@ucw.cz>
13015         * x86-tune.def (X86_TUNE_AVOID_FALSE_DEP_FOR_BMI,
13016         X86_TUNE_ADJUST_UNROLL, X86_TUNE_ONE_IF_CONV_INSN): Move to right
13017         spot in the file.
13019 2017-10-10  Richard Sandiford  <richard.sandiford@linaro.org>
13021         * wide-int.h (wide_int_ref_storage): Make host_dependent_precision
13022         a template parameter.
13023         (WIDE_INT_REF_FOR): Update accordingly.
13024         * tree.h (wi::int_traits <const_tree>): Delete.
13025         (wi::tree_to_widest_ref, wi::tree_to_offset_ref): New typedefs.
13026         (wi::to_widest, wi::to_offset): Use them.  Expand commentary.
13027         (wi::tree_to_wide_ref): New typedef.
13028         (wi::to_wide): New function.
13029         * calls.c (get_size_range): Use wi::to_wide when operating on
13030         trees as wide_ints.
13031         * cgraph.c (cgraph_node::create_thunk): Likewise.
13032         * config/i386/i386.c (ix86_data_alignment): Likewise.
13033         (ix86_local_alignment): Likewise.
13034         * dbxout.c (stabstr_O): Likewise.
13035         * dwarf2out.c (add_scalar_info, gen_enumeration_type_die): Likewise.
13036         * expr.c (const_vector_from_tree): Likewise.
13037         * fold-const-call.c (host_size_t_cst_p, fold_const_call_1): Likewise.
13038         * fold-const.c (may_negate_without_overflow_p, negate_expr_p)
13039         (fold_negate_expr_1, int_const_binop_1, const_binop)
13040         (fold_convert_const_int_from_real, optimize_bit_field_compare)
13041         (all_ones_mask_p, sign_bit_p, unextend, extract_muldiv_1)
13042         (fold_div_compare, fold_single_bit_test, fold_plusminus_mult_expr)
13043         (pointer_may_wrap_p, expr_not_equal_to, fold_binary_loc)
13044         (fold_ternary_loc, multiple_of_p, fold_negate_const, fold_abs_const)
13045         (fold_not_const, round_up_loc): Likewise.
13046         * gimple-fold.c (gimple_fold_indirect_ref): Likewise.
13047         * gimple-ssa-warn-alloca.c (alloca_call_type_by_arg): Likewise.
13048         (alloca_call_type): Likewise.
13049         * gimple.c (preprocess_case_label_vec_for_gimple): Likewise.
13050         * godump.c (go_output_typedef): Likewise.
13051         * graphite-sese-to-poly.c (tree_int_to_gmp): Likewise.
13052         * internal-fn.c (get_min_precision): Likewise.
13053         * ipa-cp.c (ipcp_store_vr_results): Likewise.
13054         * ipa-polymorphic-call.c
13055         (ipa_polymorphic_call_context::ipa_polymorphic_call_context): Likewise.
13056         * ipa-prop.c (ipa_print_node_jump_functions_for_edge): Likewise.
13057         (ipa_modify_call_arguments): Likewise.
13058         * match.pd: Likewise.
13059         * omp-low.c (scan_omp_1_op, lower_omp_ordered_clauses): Likewise.
13060         * print-tree.c (print_node_brief, print_node): Likewise.
13061         * stmt.c (expand_case): Likewise.
13062         * stor-layout.c (layout_type): Likewise.
13063         * tree-affine.c (tree_to_aff_combination): Likewise.
13064         * tree-cfg.c (group_case_labels_stmt): Likewise.
13065         * tree-data-ref.c (dr_analyze_indices): Likewise.
13066         (prune_runtime_alias_test_list): Likewise.
13067         * tree-dump.c (dequeue_and_dump): Likewise.
13068         * tree-inline.c (remap_gimple_op_r, copy_tree_body_r): Likewise.
13069         * tree-predcom.c (is_inv_store_elimination_chain): Likewise.
13070         * tree-pretty-print.c (dump_generic_node): Likewise.
13071         * tree-scalar-evolution.c (iv_can_overflow_p): Likewise.
13072         (simple_iv_with_niters): Likewise.
13073         * tree-ssa-address.c (addr_for_mem_ref): Likewise.
13074         * tree-ssa-ccp.c (ccp_finalize, evaluate_stmt): Likewise.
13075         * tree-ssa-loop-ivopts.c (constant_multiple_of): Likewise.
13076         * tree-ssa-loop-niter.c (split_to_var_and_offset)
13077         (refine_value_range_using_guard, number_of_iterations_ne_max)
13078         (number_of_iterations_lt_to_ne, number_of_iterations_lt)
13079         (get_cst_init_from_scev, record_nonwrapping_iv)
13080         (scev_var_range_cant_overflow): Likewise.
13081         * tree-ssa-phiopt.c (minmax_replacement): Likewise.
13082         * tree-ssa-pre.c (compute_avail): Likewise.
13083         * tree-ssa-sccvn.c (vn_reference_fold_indirect): Likewise.
13084         (vn_reference_maybe_forwprop_address, valueized_wider_op): Likewise.
13085         * tree-ssa-structalias.c (get_constraint_for_ptr_offset): Likewise.
13086         * tree-ssa-uninit.c (is_pred_expr_subset_of): Likewise.
13087         * tree-ssanames.c (set_nonzero_bits, get_nonzero_bits): Likewise.
13088         * tree-switch-conversion.c (collect_switch_conv_info, array_value_type)
13089         (dump_case_nodes, try_switch_expansion): Likewise.
13090         * tree-vect-loop-manip.c (vect_gen_vector_loop_niters): Likewise.
13091         (vect_do_peeling): Likewise.
13092         * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
13093         * tree-vect-stmts.c (vectorizable_load): Likewise.
13094         * tree-vrp.c (compare_values_warnv, vrp_int_const_binop): Likewise.
13095         (zero_nonzero_bits_from_vr, ranges_from_anti_range): Likewise.
13096         (extract_range_from_binary_expr_1, adjust_range_with_scev): Likewise.
13097         (overflow_comparison_p_1, register_edge_assert_for_2): Likewise.
13098         (is_masked_range_test, find_switch_asserts, maybe_set_nonzero_bits)
13099         (vrp_evaluate_conditional_warnv_with_ops, intersect_ranges): Likewise.
13100         (range_fits_type_p, two_valued_val_range_p, vrp_finalize): Likewise.
13101         (evrp_dom_walker::before_dom_children): Likewise.
13102         * tree.c (cache_integer_cst, real_value_from_int_cst, integer_zerop)
13103         (integer_all_onesp, integer_pow2p, integer_nonzerop, tree_log2)
13104         (tree_floor_log2, tree_ctz, mem_ref_offset, tree_int_cst_sign_bit)
13105         (tree_int_cst_sgn, get_unwidened, int_fits_type_p): Likewise.
13106         (get_type_static_bounds, num_ending_zeros, drop_tree_overflow)
13107         (get_range_pos_neg): Likewise.
13108         * ubsan.c (ubsan_expand_ptr_ifn): Likewise.
13109         * config/darwin.c (darwin_mergeable_constant_section): Likewise.
13110         * config/aarch64/aarch64.c (aapcs_vfp_sub_candidate): Likewise.
13111         * config/arm/arm.c (aapcs_vfp_sub_candidate): Likewise.
13112         * config/avr/avr.c (avr_fold_builtin): Likewise.
13113         * config/bfin/bfin.c (bfin_local_alignment): Likewise.
13114         * config/msp430/msp430.c (msp430_attr): Likewise.
13115         * config/nds32/nds32.c (nds32_insert_attributes): Likewise.
13116         * config/powerpcspe/powerpcspe-c.c
13117         (altivec_resolve_overloaded_builtin): Likewise.
13118         * config/powerpcspe/powerpcspe.c (rs6000_aggregate_candidate)
13119         (rs6000_expand_ternop_builtin): Likewise.
13120         * config/rs6000/rs6000-c.c
13121         (altivec_resolve_overloaded_builtin): Likewise.
13122         * config/rs6000/rs6000.c (rs6000_aggregate_candidate): Likewise.
13123         (rs6000_expand_ternop_builtin): Likewise.
13124         * config/s390/s390.c (s390_handle_hotpatch_attribute): Likewise.
13126 2017-10-10  Bin Cheng  <bin.cheng@arm.com>
13128         * tree-vect-loop-manip.c (rename_variables_in_bb): Rename PHI nodes
13129         when copying loop nest with only one inner loop.
13131 2017-10-10  Richard Biener  <rguenther@suse.de>
13133         * tree-cfgcleanup.c (cleanup_tree_cfg_noloop): Avoid compacting
13134         blocks if SCEV is active.
13135         * tree-scalar-evolution.c (analyze_scalar_evolution_1): Remove
13136         dead code.
13137         (analyze_scalar_evolution): Handle cached evolutions the obvious way.
13138         (scev_initialize): Assert we are not yet initialized.
13140 2017-10-10  Bin Cheng  <bin.cheng@arm.com>
13142         * tree-loop-distribution.c (generate_loops_for_partition): Remove
13143         inner loop's exit stmt by making it always exit the loop, otherwise
13144         we would generate an infinite empty loop.
13146 2017-10-10  Bin Cheng  <bin.cheng@arm.com>
13148         * tree-vect-loop-manip.c (slpeel_tree_duplicate_loop_to_edge_cfg): Skip
13149         renaming variables in new preheader if it's deleted.
13151 2017-10-10  Bin Cheng  <bin.cheng@arm.com>
13153         * tree-loop-distribution.c (struct partition): Remove unused field
13154         loops of the structure.
13155         (partition_alloc, partition_free): Ditto.
13156         (build_rdg_partition_for_vertex): Ditto.
13158 2017-10-09  Jeff Law  <law@redhat.com>
13160         * targhooks.c (default_stack_clash_protection_final_dynamic_probe): Fix
13161         return type to match prototype and documentation.
13163 2010-10-09  Segher Boessenkool  <segher@kernel.crashing.org>
13165         * config/rs6000/rs6000.c (processor_costs): Move to ...
13166         * config/rs6000/rs6000.h: ... here.
13167         (rs6000_cost): Declare.
13169 2017-10-09  Eric Botcazou  <ebotcazou@adacore.com>
13171         * except.c (setjmp_fn): New global variable.
13172         (init_eh): Initialize it if DONT_USE_BUILTIN_SETJMP is defined.
13173         (sjlj_emit_function_enter): Call it instead of BUILTIN_SETJMP
13174         if DONT_USE_BUILTIN_SETJMP is defined.
13176 2017-10-09  Segher Boessenkool  <segher@kernel.crashing.org>
13178         * target.def (insn_cost): New hook.
13179         * doc/tm.texi.in (TARGET_INSN_COST): New hook.
13180         * doc/tm.texi: Regenerate.
13181         * rtlanal.c (insn_cost): Use the new hook.
13183 2017-10-09  Segher Boessenkool  <segher@kernel.crashing.org>
13185         * combine.c (combine_validate_cost): Compute the new insn_cost,
13186         not just pattern_cost.
13187         (try_combine): Adjust comment.
13189 2017-10-09  Segher Boessenkool  <segher@kernel.crashing.org>
13191         * cfgrtl.c (rtl_account_profile_record): Replace insn_rtx_cost with
13192         insn_cost.
13193         * combine.c (uid_insn_cost): Adjust comment.
13194         (combine_validate_cost): Adjust comment.  Use pattern_cost instead
13195         of insn_rtx_cost
13196         (combine_instructions): Use insn_cost instead of insn_rtx_cost.
13197         * dse.c (find_shift_sequence): Ditto.
13198         * ifcvt.c (cheap_bb_rtx_cost_p): Ditto.
13199         (bb_valid_for_noce_process_p): Use pattern_cost.
13200         * rtl.h (insn_rtx_cost): Delete.
13201         (pattern_cost): New prototype.
13202         (insn_cost): New prototype.
13203         * rtlanal.c (insn_rtx_cost): Rename to...
13204         (pattern_cost): ... this.
13205         (insn_cost): New.
13207 2017-10-09  Uros Bizjak  <ubizjak@gmail.com>
13209         * config/i386/i386.md (*jcc_2): Remove insn pattern.
13210         (*jcc<mode>_0_r_i387): Ditto.
13211         (*jccxf_r_i387): Ditto.
13212         (*jcc<mode>_r_i387): Ditto.
13213         (*jccu<mode>_r_i387): Ditto.
13214         (*jcc<X87MODEF:mode>_<SWI24:mode>_r_i387): Ditto.
13215         (*jcc): Rename from *jcc_1.
13217 2017-10-09  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
13219         * config/rs6000/rs6000-p8swap.c (rs6000_analyze_swaps): Process
13220         deferred rescans after the lvx/stvx recombination pre-pass.
13222 2017-10-09  Michael Meissner  <meissner@linux.vnet.ibm.com>
13224         * config/rs6000/amo.h: New include file to provide ISA 3.0 atomic
13225         memory operation instruction support.
13226         * config.gcc (powerpc*-*-*): Include amo.h as an extra header.
13227         (rs6000-ibm-aix[789]*): Likewise.
13228         * doc/extend.texi (PowerPC Atomic Memory Operation Functions):
13229         Document new functions.
13231 2017-10-09  Richard Biener  <rguenther@suse.de>
13233         PR tree-optimization/82397
13234         * tree-data-ref.c (data_ref_compare_tree): Make sure to return
13235         equality only for semantically equal trees.
13237 2017-10-09  Richard Biener  <rguenther@suse.de>
13239         PR tree-optimization/82449
13240         * sese.c (scev_analyzable_p): Check whether the SCEV is linear.
13241         * tree-chrec.h (evolution_function_is_constant_p): Adjust to
13242         allow constant addresses.
13243         * tree-chrec.c (scev_is_linear_expression): Constant evolutions
13244         are linear.
13246 2017-10-09  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
13248         * config/s390/s390-builtins.def (vec_nabs, vec_vfi): Fix builtin
13249         flags.
13251 2017-10-09  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
13253         PR target/82463
13254         * config/s390/vecintrin.h (vec_madd, vec_msub): Fix macro
13255         definitions.
13257 2017-10-09  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
13259         PR target/82465
13260         * config/s390/s390-builtins.def (vec_sqrt): Fix builtin flags.
13262 2017-10-09  Jakub Jelinek  <jakub@redhat.com>
13264         PR target/82464
13265         * config/s390/s390-builtins.def (s390_vec_xor_flt_a,
13266         s390_vec_xor_flt_b, s390_vec_xor_flt_c): New.
13268 2017-10-09  Richard Sandiford  <richard.sandiford@linaro.org>
13270         * wide-int.h (WI_BINARY_OPERATOR_RESULT): New macro.
13271         (WI_BINARY_PREDICATE_RESULT): Likewise.
13272         (wi::binary_traits::operator_result): New type.
13273         (wi::binary_traits::predicate_result): Likewise.
13274         (generic_wide_int::operator~, unary generic_wide_int::operator-)
13275         (generic_wide_int::operator==, generic_wide_int::operator!=)
13276         (generic_wide_int::operator&, generic_wide_int::and_not)
13277         (generic_wide_int::operator|, generic_wide_int::or_not)
13278         (generic_wide_int::operator^, generic_wide_int::operator+
13279         (binary generic_wide_int::operator-, generic_wide_int::operator*):
13280         Delete.
13281         (operator~, unary operator-, operator==, operator!=, operator&)
13282         (operator|, operator^, operator+, binary operator-, operator*): New
13283         functions.
13284         * expr.c (get_inner_reference): Use wi::bit_and_not.
13285         * fold-const.c (fold_binary_loc): Likewise.
13286         * ipa-prop.c (ipa_compute_jump_functions_for_edge): Likewise.
13287         * tree-ssa-ccp.c (get_value_from_alignment): Likewise.
13288         (bit_value_binop): Likewise.
13289         * tree-ssa-math-opts.c (find_bswap_or_nop_load): Likewise.
13290         * tree-vrp.c (zero_nonzero_bits_from_vr): Likewise.
13291         (extract_range_from_binary_expr_1): Likewise.
13292         (masked_increment): Likewise.
13293         (simplify_bit_ops_using_ranges): Likewise.
13295 2017-10-09  Martin Jambor  <mjambor@suse.cz>
13297         PR hsa/82416
13298         * hsa-common.h (hsa_op_with_type): New method extend_int_to_32bit.
13299         * hsa-gen.c (hsa_extend_inttype_to_32bit): New function.
13300         (hsa_type_for_scalar_tree_type): Use it.  Always force min32int for
13301         COMPLEX types.
13302         (hsa_fixup_mov_insn_type): New function.
13303         (hsa_op_with_type::get_in_type): Use it.
13304         (hsa_build_append_simple_mov): Likewise.  Allow sub-32bit
13305         immediates in an assert.
13306         (hsa_op_with_type::extend_int_to_32bit): New method.
13307         (gen_hsa_insns_for_bitfield): Fixup instruction and intermediary
13308         types.  Convert to dest type if necessary.
13309         (gen_hsa_insns_for_bitfield_load): Fixup load type if necessary.
13310         (reg_for_gimple_ssa): Pass false as min32int to
13311         hsa_type_for_scalar_tree_type.
13312         (gen_hsa_addr): Fixup type when creating addresable temporary.
13313         (gen_hsa_cmp_insn_from_gimple): Extend operands if necessary.
13314         (gen_hsa_unary_operation): Extend operands and convert to dest type if
13315         necessary.  Call hsa_fixup_mov_insn_type.
13316         (gen_hsa_binary_operation): Changed operand types to hsa_op_with_type,
13317         extend operands and convert to dest type if necessary.
13318         (gen_hsa_insns_for_operation_assignment): Extend operands and convert
13319         to dest type if necessary.
13320         (set_output_in_type): Call hsa_fixup_mov_insn_type.  Just ude dest
13321         if conversion nt necessary and size matches.
13322         (gen_hsa_insns_for_load): Call hsa_fixup_mov_insn_type, convert
13323         to dest type if necessary.
13324         (gen_hsa_insns_for_store): Call hsa_fixup_mov_insn_type.
13325         (gen_hsa_insns_for_switch_stmt): Likewise. Also extend operands if
13326         necessary.
13327         (gen_hsa_clrsb): Likewise.
13328         (gen_hsa_ffs): Likewise.
13329         (gen_hsa_divmod): Extend operands and convert to dest type if
13330         necessary.
13331         (gen_hsa_atomic_for_builtin): Change type of op to hsa_op_with_type.
13333 2017-10-08  Segher Boessenkool  <segher@kernel.crashing.org>
13335         * config/rs6000/rs6000.md (conditional branch): Clean up formatting.
13336         Remove empty default arguments.  Use a brace block as output
13337         statement.
13338         (conditional return): Ditto.
13339         (jump): Ditto.
13340         (indirect_jump): Ditto.  Use b%T0 instead of bctr/blr.
13341         (tablejump, tablejumpsi, tablejumpdi, *tablejump<mode>_internal1):
13342         Ditto.
13343         (group_ending_nop): Ditto.
13344         (doloop_end): Ditto.
13345         (ctr<mode>, ctr<mode>_internal1, ctr<mode>_internal2): Ditto.
13346         (splitters for those): Ditto.
13348 2017-10-08  Segher Boessenkool  <segher@kernel.crashing.org>
13350         * config/rs6000/rs6000-string.c (expand_strncmp_align_check): Invert
13351         a conditional jump (and the compare for it) so that pc_rtx is the
13352         last operand.
13353         * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Adjust
13354         for the deleted and renamed ctr<mode>_internal[234] patterns.
13355         * config/rs6000/rs6000.md: Delete second conditional branch pattern.
13356         Delete second conditional return pattern.
13357         (ctr<mode>_internal2): Delete this second bdnz pattern.
13358         (ctr<mode>_internal3): Rename to ctr<mode>_internal2.
13359         (ctr<mode>_internal4): Delete this second bdz pattern.
13361 2017-10-08  Eric Botcazou  <ebotcazou@adacore.com>
13363         * tree-outof-ssa.h (ssaexpand): Add partitions_for_undefined_values.
13364         (always_initialized_rtx_for_ssa_name_p): New predicate.
13365         * tree-outof-ssa.c (remove_ssa_form): Initialize new field of SA.
13366         (finish_out_of_ssa): Free new field of SA.
13367         * tree-ssa-coalesce.h (get_undefined_value_partitions): Declare.
13368         * tree-ssa-coalesce.c: Include tree-ssa.h.
13369         (get_parm_default_def_partitions): Remove extern keyword.
13370         (get_undefined_value_partitions): New function.
13371         * expr.c (expand_expr_real_1) <expand_decl_rtl>: For a SSA_NAME, do
13372         not set SUBREG_PROMOTED_VAR_P on the sub-register if it may contain
13373         uninitialized bits.
13374         * loop-iv.c (iv_get_reaching_def): Disqualify all subregs.
13376 2017-10-08  Eric Botcazou  <ebotcazou@adacore.com>
13378         * builtins.def (BUILT_IN_SETJMP): Revert latest change.
13380 2017-10-08  Jan Hubicka  <hubicka@ucw.cz>
13382         * config/i386/i386.c (ix86_expand_set_or_movmem): Disable 512bit loops
13383         for targets that preffer 128bit.
13385 2017-10-08  Jan Hubicka  <hubicka@ucw.cz>
13387         * config/i386/i386.c (has_dispatch): Disable for Ryzen.
13389 2017-10-08  Olivier Hainque  <hainque@adacore.com>
13391         * config/arm/arm.c (arm_set_return_address): Use MEM_VOLATILE_P
13392         on the target mem instead of RTX_FRAME_RELATED_P on the insn to
13393         prevent DSE.
13394         (thumb_set_return_address): Likewise.
13396 2017-10-08  Olivier Hainque  <hainque@adacore.com>
13398         * common/config/arm/arm-common.c (arm_except_unwind_info):
13399         Handle DWARF2_UNWIND_INFO.
13401 2017-10-07  Michael Collison <michael.collison@arm.com>
13403         * config/aarch64/aarch64.md (*aarch64_reg_<optab>_minus<mode>3):
13404         New pattern.
13406 2017-10-07  Eric Botcazou  <ebotcazou@adacore.com>
13408         * builtins.def (BUILT_IN_SETJMP): Declare as library builtin instead
13409         of GCC builtin if DONT_USE_BUILTIN_SETJMP is defined.
13410         * except.c (sjlj_emit_function_enter): If DONT_USE_BUILTIN_SETJMP is
13411         defined, force the creation of a new block for a dispatch label.
13413 2017-10-07  Jan Hubicka  <hubicka@ucw.cz>
13415         * invoke.texi (Wsuggest-attribute=cold): Document.
13416         * common.opt (Wsuggest-attribute=cold): New
13417         * ipa-pure-const.c (warn_function_cold): New function.
13418         * predict.c (compute_function_frequency): Use it.
13419         * predict.h (warn_function_cold): Declare.
13421 2017-10-06  Jan Hubicka  <hubicka@ucw.cz>
13423         * tree-switch-conversion.c (do_jump_if_equal, emit_cmp_and_jump_insns):
13424         Update profile.
13426 2017-10-06  Martin Liska  <mliska@suse.cz>
13428         * sanopt.c (struct sanopt_tree_triplet_hash): Remove inline
13429         keyword for member functions.
13430         (struct sanopt_tree_couple): New struct.
13431         (struct sanopt_tree_couple_hash): New function.
13432         (struct sanopt_ctx): Add new hash_map.
13433         (has_dominating_ubsan_ptr_check): New function.
13434         (record_ubsan_ptr_check_stmt): Likewise.
13435         (maybe_optimize_ubsan_ptr_ifn): Likewise.
13436         (sanopt_optimize_walker): Handle IFN_UBSAN_PTR.
13437         (pass_sanopt::execute): Handle also SANITIZE_POINTER_OVERFLOW.
13439 2017-10-06  Sudakshina Das  <sudi.das@arm.com>
13441         PR target/82440
13442         * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Only call
13443         aarch64_simd_valid_immediate on CONST_VECTORs.
13444         (aarch64_reg_or_bic_imm): Likewise.
13446 2017-10-06  Wilco Dijkstra  <wdijkstr@arm.com>
13448         PR rtl-optimization/82396
13449         * haifa-sched.c (ready_sort_real): Disable qsort checking.
13451 2017-10-06  Sebastian Pop  <sebpop@gmail.com>
13453         * graphite-dependences.c (scop_get_reads): Move code to...
13454         (scop_get_must_writes): Move code to...
13455         (scop_get_may_writes): Move code to...
13456         (scop_get_reads_and_writes): ... here.
13457         (scop_get_dependences): Call scop_get_reads_and_writes.
13459 2017-10-06  Jakub Jelinek  <jakub@redhat.com>
13461         PR tree-optimization/82434
13462         * fold-const.h (can_native_encode_type_p,
13463         can_native_encode_string_p): Remove.
13464         * fold-const.c (native_encode_int): Formatting fixes.  If ptr is NULL,
13465         don't encode anything, just return what would be otherwise returned.
13466         (native_encode_fixed, native_encode_complex, native_encode_vector):
13467         Likewise.
13468         (native_encode_string): Likewise.  Inline by hand
13469         can_native_encode_string_p.
13470         (can_native_encode_type_p): Remove.
13471         (can_native_encode_string_p): Remove.
13472         * tree-vect-stmts.c (vectorizable_store): Instead of testing just
13473         STRING_CSTs using can_native_encode_string_p, test all
13474         CONSTANT_CLASS_P values using native_encode_expr with NULL ptr.
13475         * gimple-ssa-store-merging.c (encode_tree_to_bitpos): Remove last
13476         argument from native_encode_expr.
13477         (rhs_valid_for_store_merging_p): Use native_encode_expr with NULL ptr.
13478         (pass_store_merging::execute): Don't unnecessarily look for 3 stmts,
13479         but just 2.
13481 2017-10-06  Richard Biener  <rguenther@suse.de>
13483         PR tree-optimization/82397
13484         * tree-vect-data-refs.c (dr_group_sort_cmp): Do not use
13485         operand_equal_p but rely on data_ref_compare_tree for detecting
13486         equalities.
13487         (vect_analyze_data_ref_accesses): Use data_ref_compare_tree
13488         to match up with dr_group_sort_cmp.
13490 2017-10-06  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
13492         PR target/82322
13493         * config/s390/s390-builtins.def (s390_vfi): Define new overloaded
13494         builtin.
13495         * config/s390/s390-builtin-types.def: Regenerate.
13497 2017-10-06  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
13499         PR target/82317
13500         * config/s390/s390-builtin-types.def: Regenerate.
13501         * config/s390/s390-builtins.def (s390_vfmaxdb_4, s390_vfmindb_4):
13502         Change flag from B_VXE to B_VX.
13503         (s390_vec_min_dbl): Remove B_VXE flag.
13505 2017-10-06  Richard Biener  <rguenther@suse.de>
13507         * graphite-isl-ast-to-gimple.c: Include ssa.h and tree-ssa.h.
13508         (translate_isl_ast_to_gimple::translate_pending_phi_nodes,
13509         translate_isl_ast_to_gimple::is_valid_rename,
13510         translate_isl_ast_to_gimple::get_rename,
13511         translate_isl_ast_to_gimple::get_def_bb_for_const,
13512         translate_isl_ast_to_gimple::get_new_name,
13513         translate_isl_ast_to_gimple::collect_all_ssa_names,
13514         translate_isl_ast_to_gimple::copy_loop_phi_args,
13515         translate_isl_ast_to_gimple::collect_all_ssa_names,
13516         translate_isl_ast_to_gimple::copy_loop_phi_args,
13517         translate_isl_ast_to_gimple::copy_loop_phi_nodes,
13518         translate_isl_ast_to_gimple::add_close_phis_to_merge_points,
13519         translate_isl_ast_to_gimple::add_close_phis_to_outer_loops,
13520         translate_isl_ast_to_gimple::copy_loop_close_phi_args,
13521         translate_isl_ast_to_gimple::copy_loop_close_phi_nodes,
13522         translate_isl_ast_to_gimple::copy_cond_phi_args,
13523         translate_isl_ast_to_gimple::copy_cond_phi_nodes,
13524         translate_isl_ast_to_gimple::edge_for_new_close_phis,
13525         translate_isl_ast_to_gimple::add_phi_arg_for_new_expr,
13526         translate_isl_ast_to_gimple::rename_uses,
13527         translate_isl_ast_to_gimple::rename_all_uses): Remove.
13528         (translate_isl_ast_to_gimple::get_rename_from_scev): Simplify.
13529         (set_rename_for_each_def): Likewise.
13530         (graphite_copy_stmts_from_block): Handle debug stmt resetting
13531         here.  Handle rewriting SCEV analyzable uses here.
13532         (copy_bb_and_scalar_dependences): Generate code for PHI
13533         copy-in/outs.
13534         (graphite_regenerate_ast_isl): Adjust.
13535         * graphite-scop-detection.c (trivially_empty_bb_p): Move to sese.[ch].
13536         (add_write, add_read): New functions.
13537         (build_cross_bb_scalars_def): Use it and simplify.
13538         (build_cross_bb_scalars_use): Likewise.
13539         (graphite_find_cross_bb_scalar_vars): Inline into...
13540         (try_generate_gimple_bb): ...here.  Add dependences for PHIs,
13541         simulating out-of-SSA.  Compute liveout and add dependencies.
13542         (build_scops): Force an empty entry block.
13543         * sese.h (sese_info_t::liveout, sese_info_t::debug_liveout): New
13544         members.
13545         (sese_build_liveouts): Declare.
13546         (sese_trivially_empty_bb_p): Likewise.
13547         * sese.c (sese_build_liveouts_bb): Properly handle PHIs,
13548         compute liveout and debug_liveout.
13549         (sese_bad_liveouts_use): Remove.
13550         (sese_reset_debug_liveouts_bb): Likewise.
13551         (sese_reset_debug_liveouts): Rewrite in terms of debug_liveout.
13552         (sese_build_liveouts): Build liveout and debug_liveout and store
13553         it in region.
13554         (new_sese_info): Adjust.
13555         (free_sese_info): Likewise.
13556         (sese_insert_phis_for_liveouts): Reset debug stmts from here,
13557         do not build liveout here.
13558         (move_sese_in_condition): Adjust region entry.
13559         (scev_analyzable_p): Match up with chrec_apply requirements.
13560         (sese_trivially_empty_bb_p): New.
13561         * tree-into-ssa.c (get_reaching_def): Properly support generating
13562         default-defs for incremental rewrite of anonymous names.
13564 2017-10-06  Richard Biener  <rguenther@suse.de>
13566         * graphite-sese-to-poly.c (extract_affine): For casts increasing
13567         precision do not perform modulo reduction.
13569 2017-10-06  Richard Biener  <rguenther@suse.de>
13571         PR tree-optimization/82436
13572         * tree-vect-slp.c (vect_supported_load_permutation_p): More
13573         conservatively choose the vectorization factor when checking
13574         whether we can perform the required load permutation.
13575         (vect_transform_slp_perm_load): Assert when we may not fail.
13577 2017-10-05  Segher Boessenkool  <segher@kernel.crashing.org>
13579         * config/rs6000/sysv4.h (SUBTARGET_OVERRIDE_OPTIONS): Correct error
13580         message for incompatible -msdata=* and -mcall-* options.
13582 2017-10-05  Jan Hubicka <hubicka@ucw.cz>
13584         * config/i386/i386.c (ia32_multipass_dfa_lookahead): Default to issue
13585         rate for post-reload scheduling.
13587 2017-10-05  Tamar Christina  <tamar.christina@arm.com>
13589         * doc/sourcebuild.texi (vect_sizes_16B_8B, vect_sizes_32B_16B): New.
13591 2017-10-05  Jan Hubicka <hubicka@ucw.cz>
13593         * config/i386/i386.c (znver1_cost): Set branch_cost to 3 (instead of 2)
13594         to improve monte carlo in scimark.
13596 2017-10-05  Jan Hubicka <hubicka@ucw.cz>
13598         * config/i386/i386.c (ix86_size_cost, i386_cost, i486_cost,
13599         pentium_cost, lakemont_cost, pentiumpro_cost, geode_cost, k6_cost,
13600         athlon_cost, k8_cost, amdfam10_cost, btver1_cost, btver2_cost,
13601         pentium4_cost, nocona_cost): Set reassociation width to 1.
13602         (bdver1_cost, bdver2_cost, bdver3_cost, bdver4_cost): Set reassociation
13603         width to 2 for fp operations and 1 otherwise.
13604         (znver1_cost): Set scalar reassoc width to 4 and vector to 3 and 6
13605         for int and fp.
13606         (atom_cost): Set reassociation width to 2.
13607         (slm_cost, generic_cost): Set fp reassociation width
13608         to 2 and 1 otherwise.
13609         (intel_cost): Set fp reassociation width to 4 and 1 otherwise.
13610         (core_cost): Set fp reassociation width to 4 and vector to 2.
13611         (ix86_reassociation_width): Rewrite using cost table; special case
13612         plus/minus on Zen; honor X86_TUNE_SSE_SPLIT_REGS
13613         and TARGET_AVX128_OPTIMAL.
13614         * config/i386/i386.h (processor_costs): Add
13615         reassoc_int, reassoc_fp, reassoc_vec_int, reassoc_vec_fp.
13616         (TARGET_VECTOR_PARALLEL_EXECUTION, TARGET_REASSOC_INT_TO_PARALLEL,
13617         TARGET_REASSOC_FP_TO_PARALLEL): Remove.
13618         * x86-tune.def (X86_TUNE_REASSOC_INT_TO_PARALLEL): Remove.
13619         (X86_TUNE_REASSOC_FP_TO_PARALLEL): Remove.
13620         (X86_TUNE_VECTOR_PARALLEL_EXECUTION):  Remove.
13622 2017-10-05  Nathan Sidwell  <nathan@acm.org>
13624         * doc/invoke.texi (Wparentheses): Document C++ MVP behaviour.
13626 2017-10-05  Tamar Christina  <tamar.christina@arm.com>
13628         * config/arm/arm.c (arm_test_fpu_data): New.
13629         (arm_run_selftests): Call arm_test_fpu_data.
13631 2017-10-04  Nathan Sidwell  <nathan@acm.org>
13633         * toplev.c (toplev::main): Remove excess parens on pretty_printer
13634         decl.
13635         * caller-save.c (insert_save): Remove excess parens on TO_SAVE parm.
13637 2017-10-04  Sudakshina Das  <sudi.das@arm.com>
13639         * config/aarch64/aarch64-protos.h (enum simd_immediate_check): New
13640         check type for aarch64_simd_valid_immediate.
13641         (aarch64_output_simd_mov_immediate): Update prototype.
13642         (aarch64_simd_valid_immediate): Update prototype.
13643         * config/aarch64/aarch64-simd.md (orr<mode>3): modified pattern to add
13644         support for ORR-immediate.
13645         (and<mode>3): modified pattern to add support for BIC-immediate.
13646         * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Function
13647         now checks for valid immediate for BIC and ORR based on new enum
13648         argument.
13649         (aarch64_output_simd_mov_immediate): Function now used to output
13650         BIC/ORR imm as well based on new enum argument.
13651         * config/aarch64/constraints.md (Do): New vector immediate constraint.
13652         (Db) : Likewise.
13653         * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): New predicate.
13654         (aarch64_reg_or_bic_imm): Likewise.
13656 2017-10-04  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
13658         * config/s390/vx-builtins.md ("vec_mergeh<mode>")
13659         ("vec_mergel<mode>"): Change mode iterator to V_128_NOSINGLE.
13661 2017-10-04  Wilco Dijkstra  <wdijkstr@arm.com>
13663         Revert r253399:
13665         PR rtl-optimization/82396
13666         * haifa-sched.c (autopref_multipass_init): Simplify
13667         initialization.
13668         (autopref_rank_data): Simplify sort order.
13669         * sched-int.h (autopref_multipass_data_): Remove
13670         multi_mem_insn_p, min_offset and max_offset.
13672 2017-10-04  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
13674         * doc/sourcebuild.texi: Document vect_peeling_profitable.
13676 2017-10-04  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
13678         * doc/sourcebuild.texi: Document vect_intdouble_cvt and
13679         vect_doubleint_cvt.
13681 2017-10-04  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
13683         * doc/sourcebuild.texi: Document vect_long_mult.
13685 2017-10-04  Richard Sandiford  <richard.sandiford@linaro.org>
13687         PR tree-optimization/82413
13688         * fold-const.c (build_range_check): Use widest_int when comparing
13689         the maximum ETYPE value with HIGH.
13691 2017-10-04  Wilco Dijkstra  <wdijkstr@arm.com>
13693         PR rtl-optimization/82396
13694         * haifa-sched.c (autopref_multipass_init): Simplify
13695         initialization.
13696         (autopref_rank_data): Simplify sort order.
13697         * sched-int.h (autopref_multipass_data_): Remove
13698         multi_mem_insn_p, min_offset and max_offset.
13700 2017-10-04  Jakub Jelinek  <jakub@redhat.com>
13702         PR tree-optimization/82381
13703         * tree-ssa-reassoc.c (sort_by_operand_rank): Check for different
13704         oeN->rank first.  Return 1 or -1 if one op is SSA_NAME and the other
13705         is not.
13707         PR tree-optimization/82374
13708         * omp-low.c (create_omp_child_function): Copy DECL_ATTRIBUTES,
13709         DECL_FUNCTION_SPECIFIC_OPTIMIZATION,
13710         DECL_FUNCTION_SPECIFIC_TARGET and DECL_FUNCTION_VERSIONED from
13711         current_function_decl to the new decl.
13713 2017-10-03  Michael Meissner  <meissner@linux.vnet.ibm.com>
13715         * config/rs6000/rs6000-builtin.def (BU_FLOAT128_2_HW): Define new
13716         helper macro for IEEE float128 hardware built-in functions.
13717         (SQRTF128_ODD): Add built-in functions with the round-to-odd
13718         semantics.
13719         (TRUNCF128_ODD): Likewise.
13720         (ADDF128_ODD): Likewise.
13721         (SUBF128_ODD): Likewise.
13722         (MULF128_ODD): Likewise.
13723         (DIVF128_ODD): Likewise.
13724         (FMAF128_ODD): Likewise.
13725         * config/rs6000/rs6000.md (UNSPEC_ROUND_TO_ODD): Rename to
13726         UNSPEC_TRUNC_ROUND_TO_ODD.
13727         (UNSPEC_TRUNC_ROUND_TO_ODD): Likewise.
13728         (UNSPEC_ADD_ROUND_TO_ODD): New unspec codes for the IEEE 128-bit
13729         floating point round to odd instructions.
13730         (UNSPEC_SUB_ROUND_TO_ODD): Likewise.
13731         (UNSPEC_MUL_ROUND_TO_ODD): Likewise.
13732         (UNSPEC_DIV_ROUND_TO_ODD): Likewise.
13733         (UNSPEC_FMA_ROUND_TO_ODD): Likewise.
13734         (UNSPEC_SQRT_ROUND_TO_ODD): Likewise.
13735         (trunc<mode>sf2_hw): Change the truncate with round to odd
13736         expansion to use UNSPEC_TRUNC_ROUND_TO_ODD.
13737         (add<mode>3_odd): Add insns for IEEE 128-bit floating point round
13738         to odd hardware instructions.
13739         (sub<mode>3_odd): Likewise.
13740         (mul<mode>3_odd): Likewise.
13741         (div<mode>3_odd): Likewise.
13742         (sqrt<mode>2_odd): Likewise.
13743         (fma<mode>4_odd): Likewise.
13744         (fms<mode>4_odd): Likewise.
13745         (nfma<mode>4_odd): Likewise.
13746         (nfms<mode>4_odd): Likewise.
13747         (trunc<mode>df2_odd): Change the truncate with round to odd
13748         expansion to use UNSPEC_TRUNC_ROUND_TO_ODD.  Add a generator
13749         function.
13750         * doc/extend.texi (PowerPC built-in functions): Update documentation
13751         for existing IEEE float128-bit built-in functions.  Add built-in
13752         functions that generate the IEEE 128-bit floating point round to
13753         odd instructions.
13755 2017-10-03  Segher Boessenkool  <segher@kernel.crashing.org>
13757         PR rtl-optimization/77729
13758         * simplify-rtx.c (simplify_binary_operation_1): Delete the (X&C1)|C2
13759         to (X&(C1&~C2))|C2 transformations.
13761 2017-10-03  Martin Jambor  <mjambor@suse.cz>
13763         PR tree-optimization/82363
13764         * tree-sra.c (propagate_subaccesses_across_link): In unrecoverable
13765         mismatch, mark lacc written regardless of racc.
13767 2017-10-03  Jakub Jelinek  <jakub@redhat.com>
13769         PR tree-optimization/82381
13770         * tree-ssa-reassoc.c (sort_by_operand_rank): Don't check
13771         stmt_to_insert nor wheather SSA_NAMEs are default defs.
13772         Return 1 or -1 if one of bba and bbb is NULL. If bb_rank is equal,
13773         fallthrough into reassoc_stmt_dominates_stmt_p.
13775         PR target/82386
13776         * combine.c (combine_instructions): Don't combine in unreachable
13777         basic blocks.
13779 2017-10-02  Peter Bergner  <bergner@vnet.ibm.com>
13781         PR target/80210
13782         * config/rs6000/rs6000.c (rs6000_option_override_internal): Rewrite
13783         function to not use the have_cpu variable.  Do not set cpu_index,
13784         rs6000_cpu_index or rs6000_tune_index if we end up using TARGET_DEFAULT
13785         or the default cpu.
13786         (rs6000_valid_attribute_p): Remove duplicate initializations of
13787         old_optimize and func_optimize.
13788         (rs6000_pragma_target_parse): Call rs6000_activate_target_options ().
13789         (rs6000_activate_target_options): Make global.
13790         * config/rs6000/rs6000-protos.h (rs6000_activate_target_options): Add
13791         prototype.
13793 2017-10-02  Jakub Jelinek  <jakub@redhat.com>
13795         * tree-dfa.c (get_ref_base_and_extent): Set *pmax_size to -1
13796         if *poffset + *pmax_size overflows in HOST_WIDE_INT.
13797         Set *poffset to 0 and *psize and *pmax_size to -1 if
13798         *poffset + *psize overflows in HOST_WIDE_INT.
13800         PR tree-optimization/82387
13801         PR tree-optimization/82388
13802         PR tree-optimization/82389
13803         * tree-ssa-dse.c (dse_classify_store): Test byte_tracking_enabled
13804         instead of live_bytes non-NULL.
13806 2017-10-02  Georg-Johann Lay  <avr@gjlay.de>
13808         PR target/41076
13809         * confg/avr/avr.md (*iorhi3.ashift8-ext.zerox): Add "r,r,0"
13810         alternative.
13812 2017-10-02  Richard Biener  <rguenther@suse.de>
13814         * graphite-isl-ast-to-gimple.c (set_codegen_error): With
13815         -fchecking and --param graphite-allow-codegen-errors=0 ICE.
13816         * params.def (PARAM_GRAPHITE_ALLOW_CODEGEN_ERRORS): New param.
13818 2017-10-02  Richard Sandiford  <richard.sandiford@linaro.org>
13820         * tree.h (wi::int_traits <const_tree>::decompose): Assert that the
13821         requested precision matches the type's.
13822         * calls.c (alloc_max_size): Calculate the new candidate size as
13823         a widest_int and use wi::to_widest when comparing it with the
13824         current candidate size.
13825         * gimple-ssa-warn-alloca.c (pass_walloca::execute): Compare with
13826         zero rather than integer_zero_node.
13827         * match.pd: Check for a no-op conversion before using wi::add
13828         rather than after.  Use tree_to_uhwi when summing small shift
13829         counts into an unsigned int.
13831 2017-10-02  Richard Sandiford  <richard.sandiford@linaro.org>
13832             Alan Hayward  <alan.hayward@arm.com>
13833             David Sherwood  <david.sherwood@arm.com>
13835         PR target/71307
13836         * config/aarch64/aarch64.h (POINTER_AND_FP_REGS): New reg class.
13837         (REG_CLASS_NAMES, REG_CLASS_CONTENTS): Update accordingly.
13838         * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
13839         POINTER_AND_FP_REGS.
13841 2017-10-02  Richard Biener  <rguenther@suse.de>
13843         PR tree-optimization/82355
13844         * graphite-isl-ast-to-gimple.c (build_iv_mapping): Also build
13845         a mapping for the enclosing loop but avoid generating one for
13846         the loop tree root.
13847         (copy_bb_and_scalar_dependences): Remove premature codegen
13848         error on PHIs in blocks duplicated into multiple places.
13849         * graphite-scop-detection.c
13850         (scop_detection::stmt_has_simple_data_refs_p): For a loop not
13851         in the region use it as loop and nest to analyze the DR in.
13852         (try_generate_gimple_bb): Likewise.
13853         * graphite-sese-to-poly.c (extract_affine_chrec): Adjust.
13854         (add_loop_constraints): For blocks in a loop not in the region
13855         create a dimension with a single iteration.
13856         * sese.h (gbb_loop_at_index): Remove assert.
13858 2017-10-01  Kevin Buettner  <kevinb@redhat.com>
13860         * omp-expand.c (adjust_context_scope): New function.
13861         (expand_parallel_call): Call adjust_context_scope.
13863 2017-10-01  Jeff Law  <law@redhat.com>
13865         * tree-ssa-dom.c (optimize_stmt): Make this a method within the
13866         dom_opt_dom_walker class with direct access to private members.
13867         Add comments.  Call test_for_singularity.
13868         (dom_opt_dom_walker::before_dom_children): Corresponding changes.
13869         (dom_opt_dom_walker::after_dom_children): Do not lazily initialize
13870         m_dummy_cond anymore.
13871         (class dom_opt_dom_walker): Initialize m_dummy_cond member in the
13872         class ctor.
13873         (pass_dominator:execute): Build the dummy_cond here and pass it
13874         to the dom_opt_dom_walker ctor.
13875         (test_for_singularity): New function.
13877 2017-09-30  Krister Walfridsson  <krister.walfridsson@gmail.com>
13878             Maya Rashish  <coypu@sdf.org>
13880         * config.gcc (*-*-netbsd*): New variable nbsd_tm_file containing
13881         netbsd.h, netbsd-stdint.h, and netbsd-elf.h.
13882         (alpha*-*-netbsd*) Use nbsd_tm_file.
13883         (arm*-*-netbsdelf*) Likewise.
13884         (i[34567]86-*-netbsdelf*) Likewise.
13885         (x86_64-*-netbsd*) Likewise.
13886         (mips*-*-netbsd*) Likewise.
13887         (powerpc-*-netbsd*) Likewise.
13888         (sh*-*-netbsd*) Likewise.
13889         (sparc-*-netbsdelf*) Likewise.
13890         (sparc64-*-netbsd*) Likewise.
13891         (m68k*-*-netbsdelf*) Use nbsd_tm_file and add CHAR_FAST8/SHORT_FAST16
13892         to tm_defines.
13893         (vax-*-netbsdelf*) Likewise.
13894         * config/netbsd-stdint.h (INT_FAST8_TYPE): Check CHAR_FAST8.
13895         (UINT_FAST8_TYPE) Likewise.
13896         (INT_FAST16_TYPE) Check CHAR_FAST16.
13897         (UINT_FAST16_TYPE) Likewise.
13899 2017-09-30  Jakub Jelinek  <jakub@redhat.com>
13901         PR target/82361
13902         * config/i386/i386.md
13903         (TARGET_USE_8BIT_IDIV zext divmodsi4 splitter): New define_split.
13904         (divmodsi4_zext_1, divmodsi4_zext_2, *divmodsi4_zext_1,
13905         *divmodsi4_zext_2): New define_insn_and_split.
13906         (*divmodsi4_noext_zext_1, *divmodsi4_noext_zext_2): New define_insn.
13907         (TARGET_USE_8BIT_IDIV zext udivmodsi4 splitter): New define_split.
13908         (udivmodsi4_zext_1, udivmodsi4_zext_2, *udivmodsi4_zext_1,
13909         *udivmodsi4_zext_2, *udivmodsi4_pow2_zext_1, *udivmodsi4_pow2_zext_2):
13910         New define_insn_and_split.
13911         (*udivmodsi4_noext_zext_1, *udivmodsi4_noext_zext_2): New define_insn.
13912         * config/i386/i386.c (ix86_split_idivmod): Handle operands[0] or
13913         operands[1] having DImode when mode is SImode.
13915         * config/i386/i386.c (ix86_split_idivmod): Use mode instead of
13916         always SImode for DIV and MOD in REG_EQUAL notes.
13918 2017-09-29  Yury Gribov  <tetra2005@gmail.com>
13920         PR middle-end/82319
13921         * match.pd: Fix handling of NaNs in pattern.
13923 2017-09-29  Jeff Law  <law@redhat.com>
13925         * sbitmap.c (bitmap_bit_in_range_p): New function.
13926         * sbitmap.h (bitmap_bit_in_range_p): Prototype.
13927         * tree-ssa-dse.c (live_bytes_read): New function.
13928         (dse_classify_store): Ignore reads of dead bytes.
13930         * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Fix
13931         typos and whitespace errors.
13932         * config/i386/predicates.md (address_no_seg_operand): Likewise.
13933         * config/s390/s390.c (s390_emit_prologue): Likewise.
13935 2017-09-29  Vladimir Makarov  <vmakarov@redhat.com>
13937         PR target/81481
13938         * ira-costs.c (scan_one_insn): Don't take into account PIC equiv
13939         with a symbol for LRA.
13941 2017-09-29  Vladimir Makarov  <vmakarov@redhat.com>
13943         PR rtl-optimization/82338
13944         * lra-constraints.c (inherit_in_ebb): Check usage_insns check.
13946 2017-09-29  Alexander Monakov  <amonakov@ispras.ru>
13948         * genmodes.c (calc_wider_mode): Suppress qsort macro.
13949         * system.h [CHECKING_P] (qsort): Redirect to qsort_chk.
13950         (qsort_chk): Declare.
13951         * vec.c [CHECKING_P] (qsort_chk_error): New static function.
13952         (qsort_chk): New function.
13954 2017-09-29  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
13956         PR tree-optimization/82337
13957         * gimple-ssa-strength-reduction.c (find_phi_def): Don't record a
13958         phi definition if the PHI result appears in an abnormal PHI.
13959         (find_basis_for_base_expr): Don't record a basis if the LHS of the
13960         basis appears in an abnormal PHI.
13962 2017-09-29  Richard Biener  <rguenther@suse.de>
13964         * graphite-isl-ast-to-gimple.c
13965         (translate_isl_ast_to_gimple::set_codegen_error): New function.
13966         (binary_op_to_tree): Use it.
13967         (get_rename_from_scev): Likewise.
13968         (copy_loop_phi_nodes): Likewise.
13969         (copy_bb_and_scalar_dependences): Likewise.
13970         (translate_pending_phi_nodes): Likewise.
13972 2017-09-29  Jakub Jelinek  <jakub@redhat.com>
13974         PR target/82339
13975         * config/i386/i386.md (*movdi_internal peephole2): New -Os peephole
13976         for movabsq $(i32 << shift), r64.
13978 2017-09-28  Uros Bizjak  <ubizjak@gmail.com>
13980         * config/i386/i386.c (ix86_print_operand_address_as): Do not check
13981         index when encoding %esp as %rsp to avoid 0x67 prefix.
13983 2017-09-28  Sergey Shalnov  <Sergey.Shalnov@intel.com>
13985         * config/i386/i386.md (*movsf_internal, *movdf_internal):
13986         Return 256-bit AVX modes for TARGET_PREFER_AVX256.
13988 2017-09-28  Thomas Preud'homme  <thomas.preudhomme@arm.com>
13990         * config/arm/arm.c (arm_option_override): Forbid ARMv8-M Security
13991         Extensions with more than 16 double VFP registers.
13992         (cmse_nonsecure_entry_clear_before_return): Remove second entry of
13993         to_clear_mask and all code related to it.  Replace the remaining
13994         entry by a sbitmap and adapt code accordingly.
13996 2017-09-28  Henry Linjamäki  <henry.linjamaki@parmance.com>
13998         * brig-builtins.def: Change pure attributes to const.
14000 2017-09-28  Joseph Myers  <joseph@codesourcery.com>
14002         * config.gcc (default_gnu_indirect_function): Default to yes for
14003         sparc*-*-linux* with glibc.
14005 2017-09-28  Joseph Myers  <joseph@codesourcery.com>
14007         * config/aarch64/aarch64.c (aarch64_elf_asm_constructor)
14008         (aarch64_elf_asm_destructor): Pass SECTION_NOTYPE to get_section
14009         when creating .init_array and .fini_array sections with priority
14010         specified.
14012 2017-09-27  Christophe Lyon  <christophe.lyon@linaro.org>
14014         PR target/71727
14015         * config/aarch64/aarch64.c
14016         (aarch64_builtin_support_vector_misalignment): Always return false
14017         when misalignment is unknown.
14019 2017-09-27  Kelvin Nilsen  <kelvin@gcc.gnu.org>
14021         * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Revise
14022         this function to return false if the definition used by the swap
14023         instruction is artificial, or if the memory address from which the
14024         constant value is loaded is not represented by a base address held
14025         in a register or if the base address register is a frame or stack
14026         pointer.  Additionally, return false if the base address of the
14027         loaded constant is a SYMBOL_REF but is not considered to be a
14028         constant.
14029         (replace_swapped_load_constant): New function.
14030         (rs6000_analyze_swaps): Add a new pass to replace a swap of a
14031         loaded constant vector with a load of a swapped constant vector.
14033 2017-09-27  Carl Love  <cel@us.ibm.com>
14035         * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
14036         (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
14037         * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
14038         fctiw instruction.
14040 2017-09-27  Alexander Monakov  <amonakov@ispras.ru>
14042         * haifa-sched.c (autopref_rank_for_schedule): Order 'irrelevant' insns
14043         first, always call autopref_rank_data otherwise.
14045 2017-09-27  Richard Biener  <rguenther@suse.de>
14047         * graphite-scop-detection.c (find_scop_parameters): Move
14048         loop bound handling ...
14049         (gather_bbs::before_dom_children): ... here, avoiding the need
14050         to build scop_info->loop_nest.
14051         (record_loop_in_sese): Remove.
14052         * sese.h (sese_info_t::loop_nest): Remove.
14053         * sese.c (new_sese_info): Do not allocate loop_nest.
14054         (free_sese_info): Do not free loop_nest.
14056 2017-09-27  Jakub Jelinek  <jakub@redhat.com>
14058         PR c++/82159
14059         * gimplify.c (gimplify_modify_expr): Don't optimize away zero sized
14060         lhs from calls if the lhs has addressable type.
14062 2017-09-27  Richard Biener  <rguenther@suse.de>
14064         * graphite.h (scop::max_alias_set): New member.
14065         * graphite-scop-detection.c: Remove references to non-existing
14066         --param in comments.
14067         (build_alias_sets): Record the maximum alias set used for drs.
14068         (build_scops): Support zero as unlimited for
14069         --param graphite-max-arrays-per-scop.
14070         * graphite-sese-to-poly.c (add_scalar_version_numbers): Remove
14071         and inline into ...
14072         (build_poly_sr_1): ... here.  Compute alias set based on the
14073         maximum alias set used for drs rather than
14074         PARAM_GRAPHITE_MAX_ARRAYS_PER_SCOP
14076 2017-09-27  Richard Biener  <rguenther@suse.de>
14078         * graphite-optimize-isl.c (get_schedule_for_node_st): Allow
14079         --param loop-block-tile-size=0 to disable tiling.
14081 2017-09-27  Richard Biener  <rguenther@suse.de>
14083         * doc/invoke.texi (graphite-max-bbs-per-function): Remove.
14084         (graphite-max-nb-scop-params): Document special value zero.
14085         * domwalk.h (dom_walker::STOP): New symbolical constant.
14086         (dom_walker::dom_walker): Add optional parameter for bb to
14087         RPO mapping.
14088         (dom_walker::~dom_walker): Declare.
14089         (dom_walker::before_dom_children): Document STOP return value.
14090         (dom_walker::m_user_bb_to_rpo): New member.
14091         (dom_walker::m_bb_to_rpo): Likewise.
14092         * domwalk.c (dom_walker::dom_walker): Compute bb to RPO
14093         mapping here if not provided by the user.
14094         (dom_walker::~dom_walker): Free bb to RPO mapping if not
14095         provided by the user.
14096         (dom_walker::STOP): Define.
14097         (dom_walker::walk): Do not compute bb to RPO mapping here.
14098         Support STOP return value from before_dom_children to stop
14099         walking.
14100         * graphite-optimize-isl.c (optimize_isl): If the schedule
14101         is the same still generate code if -fgraphite-identity
14102         or -floop-parallelize-all are given.
14103         * graphite-scop-detection.c: Include cfganal.h.
14104         (gather_bbs::gather_bbs): Get and pass through bb to RPO
14105         mapping.
14106         (gather_bbs::before_dom_children): Return STOP for BBs
14107         not in the region.
14108         (build_scops): Compute bb to RPO mapping and pass it to
14109         the domwalk.  Treat --param graphite-max-nb-scop-params=0
14110         as not limiting the number of params.
14111         * graphite.c (graphite_initialize): Remove limit on the
14112         number of basic-blocks in a function.
14113         * params.def (PARAM_GRAPHITE_MAX_BBS_PER_FUNCTION): Remove.
14114         (PARAM_GRAPHITE_MAX_NB_SCOP_PARAMS): Adjust to documented
14115         default value of 10.
14117 2017-09-26  Michael Meissner  <meissner@linux.vnet.ibm.com>
14119         * config/rs6000/vsx.md (peephole for optimizing move SF to GPR):
14120         Adjust code to eliminate needing to do the shift right 32-bits
14121         operation after XSCVDPSPN.
14123 2017-09-26  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
14125         * match.pd ((X / Y) == 0 -> X < Y): New pattern.
14126         ((X / Y) != 0 -> X >= Y): Likewise.
14128 2017-09-26  Carl Love  <cel@us.ibm.com>
14130         * config/rs6000/rs6000-c.c (P9V_BUILTIN_VEC_XL_LEN_R,
14131         P9V_BUILTIN_VEC_XST_LEN_R): Add support for builtins
14132         vector unsigned char vec_xl_len_r (unsigned char *, size_t);
14133         void vec_xst_len_r (vector unsigned char, unsigned char *, size_t);
14134         * config/rs6000/altivec.h (vec_xl_len_r, vec_xst_len_r): Add defines.
14135         * config/rs6000/rs6000-builtin.def (XL_LEN_R, XST_LEN_R): Add
14136         definitions and overloading.
14137         * config/rs6000/rs6000.c (altivec_expand_builtin): Add case
14138         statement for P9V_BUILTIN_XST_LEN_R.
14139         (altivec_init_builtins): Add def_builtin for P9V_BUILTIN_STXVLL.
14140         * config/rs6000/vsx.md (lxvll, stxvll, xl_len_r, xst_len_r): Add
14141         define_expand and define_insn for the instructions and builtins.
14142         * doc/extend.texi: Update the built-in documentation file for the new
14143         built-in functions.
14144         * config/rs6000/altivec.md (altivec_lvsl_reg, altivec_lvsr_reg): Add
14145         define_insn for the instructions
14147 2017-09-26  Krister Walfridsson  <krister.walfridsson@gmail.com>
14149         PR target/39570
14150         * gcc/config/netbsd-protos.h: New file.
14151         * gcc/config/netbsd.c: New file.
14152         * gcc/config/netbsd.h (SUBTARGET_INIT_BUILTINS): Define.
14153         * gcc/config/t-netbsd: New file.
14154         * gcc/config.gcc (tm_p_file): Add netbsd-protos.h.
14155         (tmake_file) Add t-netbsd.
14156         (extra_objs) Add netbsd.o.
14158 2017-09-26  Janus Weil  <janus@gcc.gnu.org>
14160         PR fortran/82143
14161         PR fortran/82324
14162         * doc/sourcebuild.texi: Document fortran_real_10 and fortran_real_16.
14164 2017-09-26  Michael Meissner  <meissner@linux.vnet.ibm.com>
14166         * config/rs6000/rs6000.md (extendsi<mode>2): Add a splitter to do
14167         sign extension from a vector register to a GPR by doing a 32-bit
14168         direct move and then an EXTSW.
14169         (extendsi<mode>2 splitter): Likewise.
14170         (movsi_from_sf): Adjust code to eliminate doing a 32-bit shift
14171         right or vector extract after doing XSCVDPSPN.  Use
14172         zero_extendsidi2 instead of p8_mfvsrd_4_disf to move the value to
14173         the GPRs.
14174         (movdi_from_sf_zero_ext): Likewise.
14175         (reload_gpr_from_vsxsf): Likewise.
14176         (p8_mfvsrd_4_disf): Delete, no longer used.
14177         (movsi_from_df): Optimize converting a DFmode to a SFmode, and
14178         then needing to move the SFmode to a GPR to use the XSCVDPSP
14179         instruction instead of FRSP and XSCVDPSPN.
14180         * config/rs6000/vsx.md (vsx_xscvspdp_scalar2): Move insn so that
14181         it is adjacent to the other XSCVSPDP insns.
14182         (vsx_xscvdpsp_scalar): Use "ww" constraint instead of "f" to allow
14183         SFmode to be in traditional Altivec registers.
14184         (vsx_xscvdpspn): Eliminate useless alternative constraint.
14185         (vsx_xscvspdpn): Likewise.
14186         (vsx_xscvspdpn_scalar): Likewise.
14188 2017-09-26  Martin Jambor  <mjambor@suse.cz>
14190         * tree-sra.c (compare_access_positions): Put integral types first,
14191         stabilize sorting of integral types, remove conditions putting
14192         non-full-precision integers last.
14193         (sort_and_splice_var_accesses): Disable scalarization if a
14194         non-integert would be represented by a non-full-precision integer.
14196 2017-09-26  Joseph Myers  <joseph@codesourcery.com>
14198         * config/microblaze/linux.h (TARGET_ASM_FILE_END): Likewise.
14199         * config/pa/pa.h (NEED_INDICATE_EXEC_STACK): Likewise.
14200         * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Likewise.
14201         * config/pa/pa.c (pa_hpux_file_end): Rename to pa_file_end.
14202         Define unconditionally, with [ASM_OUTPUT_EXTERNAL_REAL]
14203         conditionals inside the function instead of around it.  Call
14204         file_end_indicate_exec_stack if NEED_INDICATE_EXEC_STACK.
14205         (TARGET_ASM_FILE_END): Define unconditionally to pa_file_end.
14207 2017-09-26  Richard Biener  <rguenther@suse.de>
14209         * graphite-scop-detection.c (scop_detection::build_scop_depth): Rewrite,
14210         fold in ...
14211         (scop_detection::build_scop_breadth): ... this.  Removed.
14212         (scop_detection::loop_is_valid_in_scop): Fold into single caller.
14213         (scop_detection::harmful_stmt_in_bb): Likewise.
14214         (scop_detection::graphite_can_represent_stmt): Likewise.
14215         (scop_detection::loop_body_is_valid_scop): Likewise.  Remove recursion.
14216         (scop_detection::can_represent_loop): Remove recursion, fold in ...
14217         (scop_detection::can_represent_loop_1): ... this.  Removed.
14218         (scop_detection::harmful_loop_in_region): Simplify after inlining
14219         the above and remove more quadraticness.
14220         (build_scops): Adjust.
14221         * tree-data-ref.c (loop_nest_has_data_refs): Remove pointless
14222         quadraticness.
14224 2017-09-26  Jakub Jelinek  <jakub@redhat.com>
14226         PR target/82267
14227         * config/i386/i386.c (ix86_print_operand_address_as): Only test
14228         REGNO (base) == SP_REG if base is a REG.
14230         PR middle-end/35691
14231         * tree-ssa-reassoc.c (update_range_test): Dump r->exp each time
14232         if it is different SSA_NAME.
14233         (optimize_range_tests_cmp_bitwise): New function.
14234         (optimize_range_tests): Call it.
14236 2017-09-26  Richard Biener  <rguenther@suse.de>
14238         PR tree-optimization/82321
14239         * graphite.c (canonicalize_loop_closed_ssa): Properly check
14240         for the def being inside the loop.
14242 2017-09-26  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
14244         * config/s390/vx-builtins.md ("vmslg"): Add missing operand in
14245         assembler output.
14246         * config/s390/s390-builtins.def: Fix constraint on op4.
14248 2017-09-26  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
14250         * config/s390/s390.c (s390_expand_vec_compare): Use the new mode
14251         independent expanders.
14252         * config/s390/vector.md ("vec_cmpuneq", "vec_cmpltgt")
14253         ("vec_ordered", "vec_unordered"): New expanders.
14255 2017-09-26  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
14257         * config/s390/s390.c (s390_preferred_simd_mode): Return V4SFmode
14258         for SFmode.
14260 2017-09-26  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
14262         * config/s390/vector.md ("vec_unpacks_low_v16qi"): Rename to
14263         vec_unpacks_lo_v16qi.
14264         ("vec_unpacku_low_v16qi"): Rename to vec_unpacku_lo_v16qi.
14266 2017-09-26  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
14268         * config/s390/vector.md ("vec_unpacks_lo_v4sf")
14269         ("vec_unpacks_hi_v4sf", "vec_unpacks_lo_v2df")
14270         ("vec_unpacks_hi_v2df", "vec_pack_trunc_v2df"): New expanders.
14272 2017-09-26  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
14274         * config/s390/predicates.md ("const_shift_by_byte_operand"): New
14275         predicate.
14276         * config/s390/vector.md ("*vec_srb<mode>"): Change modes to V_128
14277         and V16QI.
14278         ("*vec_slb<mode>"): New insn pattern.
14279         ("vec_shr_<mode>"): New expander.
14280         * config/s390/vx-builtins.md ("vec_slb<mode>"): Turn into expander
14281         and force the shift count operand to V16QImode.
14282         ("vec_srb<mode>"): Set shift count mode to V16QI.
14284 2017-09-26  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
14286         * config/s390/vector.md ("vec_widen_umult_lo_<mode>")
14287         ("vec_widen_umult_hi_<mode>", "vec_widen_smult_lo_<mode>")
14288         ("vec_widen_smult_hi_<mode>"): New expander definitions.
14290 2017-09-26  Richard Earnshaw  <rearnsha@arm.com>
14292         PR target/82175
14293         * config/arm/arm.h (DRIVER_SELF_SPECS): Separate sub-rules with commas.
14295 2017-09-26  Richard Biener  <rguenther@suse.de>
14297         PR tree-optimization/82320
14298         * tree-ssa-sccvn.c (set_ssa_val_to): Changing undef to undef
14299         isn't a change.
14301 2017-09-25  Jeff Law  <law@redhat.com>
14303         * config/rs6000/rs6000-protos.h (output_probe_stack_range): Update
14304         prototype for new argument.
14305         * config/rs6000/rs6000.c (rs6000_emit_allocate_stack_1): New function,
14306         mostly extracted from rs6000_emit_allocate_stack.
14307         (rs6000_emit_probe_stack_range_stack_clash): New function.
14308         (rs6000_emit_allocate_stack): Call
14309         rs6000_emit_probe_stack_range_stack_clash as needed.
14310         (rs6000_emit_probe_stack_range): Add additional argument
14311         to call to gen_probe_stack_range{si,di}.
14312         (output_probe_stack_range): New.
14313         (output_probe_stack_range_1): Renamed from output_probe_stack_range.
14314         (output_probe_stack_range_stack_clash): New.
14315         (rs6000_emit_prologue): Emit notes into dump file as requested.
14316         * rs6000.md (allocate_stack): Handle -fstack-clash-protection.
14317         (probe_stack_range<P:mode>): Operand 0 is now early-clobbered.
14318         Add additional operand and pass it to output_probe_stack_range.
14320 2017-09-25  Bin Cheng  <bin.cheng@arm.com>
14322         PR tree-optimization/82163
14323         * tree-ssa-loop-manip.h (verify_loop_closed_ssa): New parameter.
14324         (checking_verify_loop_closed_ssa): New parameter.
14325         * tree-ssa-loop-manip.c (check_loop_closed_ssa_use): Delete.
14326         (check_loop_closed_ssa_stmt): Delete.
14327         (check_loop_closed_ssa_def, check_loop_closed_ssa_bb): New functions.
14328         (verify_loop_closed_ssa): Check loop closed ssa form for LOOP.
14329         (tree_transform_and_unroll_loop): Check loop closed ssa form only for
14330         changed loops.
14332 2017-09-25  Pekka Jaaskelainen <pekka@parmance.com>
14334         * brig-builtins.def: Treat HSAIL barrier builtins as
14335         setjmp/longjump style functions.
14337 2017-09-25  Richard Sandiford  <richard.sandiford@linaro.org>
14339         * target.def (constant_alignment): New hook.
14340         * defaults.h (CONSTANT_ALIGNMENT): Delete.
14341         * doc/tm.texi.in (CONSTANT_ALIGNMENT): Replace with...
14342         (TARGET_CONSTANT_ALIGNMENT): ...this new hook.
14343         * doc/tm.texi: Regenerate.
14344         * targhooks.h (default_constant_alignment): Declare.
14345         (constant_alignment_word_strings): Likewise.
14346         * targhooks.c (default_constant_alignment): New function.
14347         (constant_alignment_word_strings): Likewise.
14348         * builtins.c (get_object_alignment_2): Use targetm.constant_alignment
14349         instead of CONSTANT_ALIGNMENT.
14350         * varasm.c (align_variable, get_variable_align, build_constant_desc)
14351         (force_const_mem): Likewise.
14352         * config/aarch64/aarch64.h (CONSTANT_ALIGNMENT): Delete.
14353         * config/aarch64/aarch64.c (aarch64_constant_alignment): New function.
14354         (aarch64_classify_address): Call it instead of CONSTANT_ALIGNMENT.
14355         (TARGET_CONSTANT_ALIGNMENT): Redefine.
14356         * config/alpha/alpha.h (CONSTANT_ALIGNMENT): Delete commented-out
14357         definition.
14358         * config/arc/arc.h (CONSTANT_ALIGNMENT): Delete.
14359         * config/arc/arc.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14360         constant_alignment_word_strings.
14361         * config/arm/arm.h (CONSTANT_ALIGNMENT_FACTOR): Delete.
14362         (CONSTANT_ALIGNMENT): Likewise.
14363         * config/arm/arm.c (TARGET_CONSTANT_ALIGNMENT): Redefine.
14364         (arm_constant_alignment): New function.
14365         * config/bfin/bfin.h (CONSTANT_ALIGNMENT): Delete.
14366         * config/bfin/bfin.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14367         constant_alignment_word_strings.
14368         * config/cr16/cr16.h (CONSTANT_ALIGNMENT): Delete.
14369         * config/cr16/cr16.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14370         constant_alignment_word_strings.
14371         * config/cris/cris.h (CONSTANT_ALIGNMENT): Delete.
14372         * config/cris/cris.c (TARGET_CONSTANT_ALIGNMENT): Redefine.
14373         (cris_constant_alignment): New function.
14374         * config/epiphany/epiphany.h (CONSTANT_ALIGNMENT): Delete.
14375         * config/epiphany/epiphany.c (TARGET_CONSTANT_ALIGNMENT): Redefine.
14376         (epiphany_constant_alignment): New function.
14377         * config/fr30/fr30.h (CONSTANT_ALIGNMENT): Delete.
14378         * config/fr30/fr30.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14379         constant_alignment_word_strings.
14380         * config/frv/frv.h (CONSTANT_ALIGNMENT): Delete.
14381         * config/frv/frv.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14382         constant_alignment_word_strings.
14383         * config/ft32/ft32.h (CONSTANT_ALIGNMENT): Delete.
14384         * config/ft32/ft32.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14385         constant_alignment_word_strings.
14386         * config/i386/i386.h (CONSTANT_ALIGNMENT): Delete.
14387         * config/i386/i386-protos.h (ix86_constant_alignment): Delete.
14388         * config/i386/i386.c (ix86_constant_alignment): Make static.
14389         Use the same interface as the target hook.
14390         (TARGET_CONSTANT_ALIGNMENT): Redefine.
14391         * config/ia64/ia64.h (CONSTANT_ALIGNMENT): Delete.
14392         * config/ia64/ia64.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14393         constant_alignment_word_strings.
14394         * config/iq2000/iq2000.h (CONSTANT_ALIGNMENT): Delete.
14395         * config/iq2000/iq2000.c (iq2000_constant_alignment): New function.
14396         (TARGET_CONSTANT_ALIGNMENT): Redefine.
14397         * config/lm32/lm32.h (CONSTANT_ALIGNMENT): Delete.
14398         * config/lm32/lm32.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14399         constant_alignment_word_strings.
14400         * config/m32r/m32r.h (CONSTANT_ALIGNMENT): Delete.
14401         * config/m32r/m32r.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14402         constant_alignment_word_strings.
14403         * config/mcore/mcore.h (CONSTANT_ALIGNMENT): Delete.
14404         * config/mcore/mcore.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14405         constant_alignment_word_strings.
14406         * config/microblaze/microblaze.h (CONSTANT_ALIGNMENT): Delete.
14407         * config/microblaze/microblaze.c (microblaze_constant_alignment):
14408         New function.
14409         (TARGET_CONSTANT_ALIGNMENT): Redefine.
14410         * config/mips/mips.h (CONSTANT_ALIGNMENT): Delete.
14411         * config/mips/mips.c (mips_constant_alignment): New function.
14412         (TARGET_CONSTANT_ALIGNMENT): Redefine.
14413         * config/mmix/mmix.h (CONSTANT_ALIGNMENT): Delete.
14414         * config/mmix/mmix-protos.h (mmix_constant_alignment): Delete.
14415         * config/mmix/mmix.c (TARGET_CONSTANT_ALIGNMENT): Redefine.
14416         (mmix_constant_alignment): Make static.  Use the same interface
14417         as the target hook.
14418         * config/moxie/moxie.h (CONSTANT_ALIGNMENT): Delete.
14419         * config/moxie/moxie.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14420         constant_alignment_word_strings.
14421         * config/nios2/nios2.h (CONSTANT_ALIGNMENT): Delete.
14422         * config/nios2/nios2.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14423         constant_alignment_word_strings.
14424         * config/pa/pa.h (CONSTANT_ALIGNMENT): Delete.
14425         * config/pa/pa.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14426         constant_alignment_word_strings.
14427         * config/powerpcspe/powerpcspe.h (CONSTANT_ALIGNMENT): Delete.
14428         * config/powerpcspe/powerpcspe.c (TARGET_CONSTANT_ALIGNMENT): Redefine.
14429         (rs6000_constant_alignment): New function.
14430         * config/riscv/riscv.h (CONSTANT_ALIGNMENT): Delete.
14431         * config/riscv/riscv.c (riscv_constant_alignment): New function.
14432         (TARGET_CONSTANT_ALIGNMENT): Redefine.
14433         * config/rs6000/rs6000.h (CONSTANT_ALIGNMENT): Delete.
14434         * config/rs6000/rs6000.c (TARGET_CONSTANT_ALIGNMENT): Redefine.
14435         (rs6000_constant_alignment): New function.
14436         * config/s390/s390.h (CONSTANT_ALIGNMENT): Delete.
14437         * config/s390/s390.c (s390_constant_alignment): New function.
14438         (TARGET_CONSTANT_ALIGNMENT): Redefine.
14439         * config/sh/sh.h (CONSTANT_ALIGNMENT): Delete.
14440         * config/sh/sh.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14441         constant_alignment_word_strings.
14442         * config/sparc/sparc.h (CONSTANT_ALIGNMENT): Delete.
14443         * config/sparc/sparc.c (TARGET_CONSTANT_ALIGNMENT): Redefine.
14444         (sparc_constant_alignment): New function.
14445         * config/spu/spu.h (CONSTANT_ALIGNMENT): Delete.
14446         * config/spu/spu.c (spu_constant_alignment): New function.
14447         (TARGET_CONSTANT_ALIGNMENT): Redefine.
14448         * config/stormy16/stormy16.h (CONSTANT_ALIGNMENT): Delete.
14449         * config/stormy16/stormy16.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14450         constant_alignment_word_strings.
14451         * config/tilegx/tilegx.h (CONSTANT_ALIGNMENT): Delete.
14452         * config/tilegx/tilegx.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14453         constant_alignment_word_strings.
14454         * config/tilepro/tilepro.h (CONSTANT_ALIGNMENT): Delete.
14455         * config/tilepro/tilepro.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14456         constant_alignment_word_strings.
14457         * config/visium/visium.h (CONSTANT_ALIGNMENT): Delete.
14458         * config/visium/visium.c (TARGET_CONSTANT_ALIGNMENT): Redefine.
14459         (visium_constant_alignment): New function.
14460         * config/xtensa/xtensa.h (CONSTANT_ALIGNMENT): Delete.
14461         * config/xtensa/xtensa.c (TARGET_CONSTANT_ALIGNMENT): Redefine.
14462         (xtensa_constant_alignment): New function.
14463         * system.h (CONSTANT_ALIGNMENT): Poison.
14465 2017-09-25  Will Schmidt  <will_schmidt@vnet.ibm.com>
14467         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add handling
14468         for early folding of vector stores (ALTIVEC_BUILTIN_ST_*).
14469         (rs6000_builtin_valid_without_lhs): New helper function.
14470         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
14471         Remove obsoleted code for handling ALTIVEC_BUILTIN_VEC_ST.
14473 2017-09-25  Richard Sandiford  <richard.sandiford@linaro.org>
14475         * target.h (vec_perm_indices): Use unsigned short rather than
14476         unsigned char.
14477         (auto_vec_perm_indices): Likewise.
14478         * config/aarch64/aarch64.c (aarch64_vectorize_vec_perm_const_ok):
14479         Use unsigned int rather than unsigned char.
14480         * config/arm/arm.c (arm_vectorize_vec_perm_const_ok): Likewise.
14482 2017-09-25  Richard Biener  <rguenther@suse.de>
14484         * cfgloop.h (sort_sibling_loops): Declare.
14485         * cfgloop.c (sort_sibling_loops_cmp): New helper.
14486         (sort_sibling_loops): New function sorting the sibling loop list
14487         in RPO order.
14488         * graphite.c (graphite_transform_loops): Sort sibling loops.
14490 2017-09-25  Richard Sandiford  <richard.sandifird@linaro.org>
14492         * target.def (vec_perm_const_ok): Change sel parameter to
14493         vec_perm_indices.
14494         * optabs-query.c (can_vec_perm_p): Update accordingly.
14495         * doc/tm.texi: Regenerate.
14496         * config/aarch64/aarch64.c (expand_vec_perm_d): Change perm to
14497         auto_vec_perm_indices and remove separate nelt field.
14498         (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
14499         (aarch64_evpc_ext, aarch64_evpc_rev, aarch64_evpc_dup)
14500         (aarch64_evpc_tbl, aarch64_expand_vec_perm_const_1)
14501         (aarch64_expand_vec_perm_const): Update accordingly.
14502         (aarch64_vectorize_vec_perm_const_ok): Likewise.  Change sel
14503         to vec_perm_indices.
14504         * config/arm/arm.c (expand_vec_perm_d): Change perm to
14505         auto_vec_perm_indices and remove separate nelt field.
14506         (arm_evpc_neon_vuzp, arm_evpc_neon_vzip, arm_evpc_neon_vrev)
14507         (arm_evpc_neon_vtrn, arm_evpc_neon_vext, arm_evpc_neon_vtbl)
14508         (arm_expand_vec_perm_const_1, arm_expand_vec_perm_const): Update
14509         accordingly.
14510         (arm_vectorize_vec_perm_const_ok): Likewise.  Change sel
14511         to vec_perm_indices.
14512         * config/i386/i386.c (ix86_vectorize_vec_perm_const_ok): Change
14513         sel to vec_perm_indices.
14514         * config/ia64/ia64.c (ia64_vectorize_vec_perm_const_ok): Likewise.
14515         * config/mips/mips.c (mips_vectorize_vec_perm_const_ok): Likewise.
14516         * config/powerpcspe/powerpcspe.c (rs6000_vectorize_vec_perm_const_ok):
14517         Likewise.
14518         * config/rs6000/rs6000.c (rs6000_vectorize_vec_perm_const_ok):
14519         Likewise.
14521 2017-09-25  Pierre-Marie de Rodat  <derodat@adacore.com>
14523         PR debug/82155
14524         * dwarf2out.c (dwarf2out_early_global_decl): Call dwarf2out_decl
14525         on the FUNCTION_DECL function context if it has a DIE that is a
14526         declaration.
14528 2017-09-25  Richard Biener  <rguenther@suse.de>
14530         PR tree-optimization/82285
14531         * tree-vect-patterns.c (vect_recog_bool_pattern): Also handle
14532         enumeral types.
14534 2017-09-25  Tom de Vries  <tom@codesourcery.com>
14536         PR target/80035
14537         PR target/81069
14538         * config/nvptx/nvptx.c (nvptx_output_call_insn): Add exit after call to
14539         noreturn function.
14541 2017-09-25  Richard Biener  <rguenther@suse.de>
14543         * graphite-optimize-isl.c (optimize_isl): Fail and dump if
14544         ISL errors other than isl_error_quota happen.  Dump if the
14545         schedule is the same.
14546         * graphite-sese-to-poly.c (build_poly_scop): Fail on ISL
14547         errors instead of aborting inside ISL.
14549 2017-09-25  Iain Sandoe  <iain@codesourcery.com>
14551         PR target/80556
14552         * config/i386/darwin.h (REAL_LIB_SPEC): New; put libSystem ahead
14553         of libgcc_eh for m64.
14554         * config/i386/darwin64.h: Likewise.
14556 2017-09-25  Richard Biener  <rguenther@suse.de>
14558         PR middle-end/82144
14559         * dwarf2out.c (gen_enumeration_type_die): Do not add alignment
14560         attribute for incomplete types nor twice for complete ones.
14562 2017-09-24  Uros Bizjak  <ubizjak@gmail.com>
14564         PR target/82267
14565         * config/i386/i386.c (ix86_print_operand_address_as): Encode
14566         %esp as %rsp to avoid 0x67 prefix if there is no index or base
14567         register.
14569 2017-09-23  Uros Bizjak  <ubizjak@gmail.com>
14571         PR bootstrap/82306
14572         * config/i386/i386.opt (mprefer-avx256): Use
14573         ix86_target_flags variable.
14574         * config/i386/i386.c (ix86_target_string): Move
14575         -mprefer-avx256 to flag2_opts.
14577 2017-09-22  Jakub Jelinek  <jakub@redhat.com>
14579         PR middle-end/35691
14580         * match.pd: Simplify x == -1 & y == -1 into (x & y) == -1
14581         and x != -1 | y != -1 into (x & y) != -1.
14583 2017-09-22  Steve Ellcey  <sellcey@cavium.com>
14585         * config.gcc: Add new case statement to set
14586         default_gnu_indirect_function.  Remove it from x86_64-*-linux*,
14587         i[34567]86-*, powerpc*-*-linux*spe*, powerpc*-*-linux*, s390-*-linux*,
14588         s390x-*-linux* case statements.   Added aarch64 to the list of
14589         supported architectures.
14591 2017-09-22  Richard Sandiford  <richard.sandiford@linaro.org>
14593         PR tree-optimization/82289
14594         * tree-vect-data-refs.c (vect_get_peeling_costs_all_drs): Check
14595         STMT_VINFO_RELEVANT_P.
14597 2017-09-22  Richard Sandiford  <richard.sandiford@linaro.org>
14598             Alan Hayward  <alan.hayward@arm.com>
14599             David Sherwood  <david.sherwood@arm.com>
14601         * tree-vrp.c (extract_range_from_multiplicative_op_1): Assert
14602         for VR_RANGE only; don't allow VR_ANTI_RANGE.
14603         (extract_range_from_binary_expr_1): Don't call
14604         extract_range_from_multiplicative_op_1 if !range_int_cst_p.
14606 2017-09-22  Richard Sandiford  <richard.sandiford@linaro.org>
14607             Alan Hayward  <alan.hayward@arm.com>
14608             David Sherwood  <david.sherwood@arm.com>
14610         * target.def (preferred_vector_alignment): New hook.
14611         * doc/tm.texi.in (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): New
14612         hook.
14613         * doc/tm.texi: Regenerate.
14614         * targhooks.h (default_preferred_vector_alignment): Declare.
14615         * targhooks.c (default_preferred_vector_alignment): New function.
14616         * tree-vectorizer.h (dataref_aux): Add a target_alignment field.
14617         Expand commentary.
14618         (DR_TARGET_ALIGNMENT): New macro.
14619         (aligned_access_p): Update commentary.
14620         (vect_known_alignment_in_bytes): New function.
14621         * tree-vect-data-refs.c (vect_calculate_required_alignment): New
14622         function.
14623         (vect_compute_data_ref_alignment): Set DR_TARGET_ALIGNMENT.
14624         Calculate the misalignment based on the target alignment rather than
14625         the vector size.
14626         (vect_update_misalignment_for_peel): Use DR_TARGET_ALIGMENT
14627         rather than TYPE_ALIGN / BITS_PER_UNIT to update the misalignment.
14628         (vect_enhance_data_refs_alignment): Mask the byte misalignment with
14629         the target alignment, rather than masking the element misalignment
14630         with the number of elements in a vector.  Also use the target
14631         alignment when calculating the maximum number of peels.
14632         (vect_find_same_alignment_drs): Use vect_calculate_required_alignment
14633         instead of TYPE_ALIGN_UNIT.
14634         (vect_duplicate_ssa_name_ptr_info): Remove stmt_info parameter.
14635         Measure DR_MISALIGNMENT relative to DR_TARGET_ALIGNMENT.
14636         (vect_create_addr_base_for_vector_ref): Update call accordingly.
14637         (vect_create_data_ref_ptr): Likewise.
14638         (vect_setup_realignment): Realign by ANDing with
14639         -DR_TARGET_MISALIGNMENT.
14640         * tree-vect-loop-manip.c (vect_gen_prolog_loop_niters): Calculate
14641         the number of peels based on DR_TARGET_ALIGNMENT.
14642         * tree-vect-stmts.c (get_group_load_store_type): Compare the gap
14643         with the guaranteed alignment boundary when deciding whether
14644         overrun is OK.
14645         (vectorizable_mask_load_store): Interpret DR_MISALIGNMENT
14646         relative to DR_TARGET_ALIGNMENT instead of TYPE_ALIGN_UNIT.
14647         (ensure_base_align): Remove stmt_info parameter.  Get the
14648         target base alignment from DR_TARGET_ALIGNMENT.
14649         (vectorizable_store): Update call accordingly.   Interpret
14650         DR_MISALIGNMENT relative to DR_TARGET_ALIGNMENT instead of
14651         TYPE_ALIGN_UNIT.
14652         (vectorizable_load): Likewise.
14654 2017-09-22  Richard Sandiford  <richard.sandiford@linaro.org>
14655             Alan Hayward  <alan.hayward@arm.com>
14656             David Sherwood  <david.sherwood@arm.com>
14658         * tree-vectorizer.h (vect_get_scalar_dr_size): New function.
14659         * tree-vect-data-refs.c (vect_update_misalignment_for_peel): Use it.
14660         (vect_enhance_data_refs_alignment): Likewise.
14662 2017-09-22  Richard Earnshaw  <richard.earnshaw@arm.com>
14664         * config/arm/parsecpu.awk (fatal): Note that we've encountered an
14665         error.  Only quit immediately if parsing is complete.
14666         (BEGIN): Initialize fatal_err and parse_done.
14667         (begin fpu, end fpu): Check number of arguments.
14668         (begin arch, end arch): Likewise.
14669         (begin cpu, end cpu): Likewise.
14670         (cname, tune for, tune flags, architecture, fpu, option): Likewise.
14671         (optalias): Likewise.
14673 2017-09-22  Richard Earnshaw  <richard.earnshaw@arm.com>
14675         * config.gcc (arm*-*-*): Don't add arm-isa.h to tm_p_file.
14676         * config/arm/arm-isa.h: Delete.  Move definitions to ...
14677         * arm-cpus.in: ... here.  Use new feature and fgroup values.
14678         * config/arm/arm.c (arm_option_override): Use lower case for feature
14679         bit names.
14680         * config/arm/arm.h (TARGET_HARD_FLOAT): Likewise.
14681         (TARGET_VFP3, TARGET_VFP5, TARGET_FMA): Likewise.
14682         * config/arm/parsecpu.awk (END): Add new command 'isa'.
14683         (isa_pfx): Delete.
14684         (print_isa_bits_for): New function.
14685         (gen_isa): New function.
14686         (gen_comm_data): Use print_isa_bits_for.
14687         (define feature): New keyword.
14688         (define fgroup): New keyword.
14689         * config/arm/t-arm (TM_H): Remove.
14690         (GTM_H): Add arm-isa.h.
14691         (arm-isa.h): Add rule to generate file.
14692         * common/config/arm/arm-common.c: (arm_canon_arch_option): Use lower
14693         case for feature bit names.
14695 2017-09-22  Richard Biener  <rguenther@suse.de>
14697         * graphite-isl-ast-to-gimple.c (graphite_verify): Inline into
14698         single caller.
14699         (graphite_regenerate_ast_isl): Do not reset SCEV.  Move debug
14700         print of no dependency loops ...
14701         * graphite.c (graphite_transform_loops): ... here.
14702         (canonicalize_loop_closed_ssa_form): Work from inner to outer
14703         loops.
14704         (same_close_phi_node, remove_duplicate_close_phi,
14705         make_close_phi_nodes_unique, defined_in_loop_p): Fold into ...
14706         (canonicalize_loop_closed_ssa): ... here and simplify.
14707         * graphite-optimize-isl.c: Include tree-vectorizer.h.
14708         (optimize_isl): Use dump_printf_loc to tell when we stopped
14709         optimizing because of an ISL timeout.
14711 2017-09-22  Richard Biener  <rguenther@suse.de>
14713         PR tree-optimization/82291
14714         * tree-if-conv.c (predicate_mem_writes): Make sure to
14715         remove writes in blocks predicated with false.
14717 2017-09-22  Richard Biener  <rguenther@suse.de>
14719         * sese.c: Include cfganal.h.
14720         (if_region_set_false_region): Remove.
14721         (create_if_region_on_edge): Likewise.
14722         (move_sese_in_condition): Re-implement without destroying
14723         dominators.
14725 2017-09-22  Richard Biener  <rguenther@suse.de>
14727         * graphite-isl-ast-to-gimple.c (translate_pending_phi_nodes):
14728         Verify both BBs contain loop PHI nodes before dispatching to
14729         copy_loop_phi_args.
14730         (graphite_regenerate_ast_isl): Do not recompute dominators,
14731         do not verify three times.  Restructure for clarity.
14732         * graphite-scop-detection.c (same_close_phi_node,
14733         remove_duplicate_close_phi, make_close_phi_nodes_unique,
14734         defined_in_loop_p, canonicalize_loop_closed_ssa,
14735         canonicalize_loop_closed_ssa_form): Simplify, remove excess
14736         checking and SSA rewrite, move to ...
14737         * graphite.c: ... here.  Include ssa.h and tree-ssa-loop-manip.h.
14738         (graphite_initialize): Do not pass in ctx, do not reset the
14739         SCEV cache, compute only dominators.
14740         (graphite_transform_loops): Allocate ISL ctx after
14741         graphite_initialize.  Call canonicalize_loop_closed_ssa_form.
14742         Maintain post-dominators only around build_scops.
14743         * sese.c (if_region_set_false_region): Make static.  Free
14744         and recompute dominators.
14745         (move_sese_in_condition): Assert we don't get called with
14746         post-dominators computed.
14747         * sese.h (if_region_set_false_region): Remove.
14749 2017-09-22  Sergey Shalnov  <sergey.shalnov@intel.com>
14751         * config/i386/sse.md ("mov<mode>_internal"): Use <sseinsnmode>
14752         mode attribute for TARGET_AVX512VL.
14754 2017-09-21  Sergey Shalnov  <sergey.shalnov@intel.com>
14756         * config/i386/i386.opt (mprefer-avx256): New option.
14757         * config/i386/i386.c (ix86_target_string): Add -mprefer-avx256
14758         to flag_opts.
14759         (ix86_preferred_simd_mode): Return 256-bit AVX modes
14760         for TARGET_PREFER_AVX256.
14761         * doc/invoke.texi (x86 Options): Document -mprefer-avx256.
14763 2017-09-21  Jeff Law  <law@redhat.com>
14765         * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash):
14766         Fix dump output if the only stack space is for pushed registers.
14768 2017-09-21  Richard Sandiford  <richard.sandiford@linaro.org>
14770         * config/spu/spu.c (spu_sched_adjust_cost): Update after renaming
14771         of insn_cost.
14773 2017-09-21  Martin Sebor  <msebor@redhat.com>
14775         PR c/81882
14776         * doc/extend.texi (attribute ifunc): Avoid relying on ill-formed
14777         code (in C++) or code that triggers warnings.
14779 2017-09-21  Eric Botcazou  <ebotcazou@adacore.com>
14781         * stor-layout.c (bit_from_pos): Do not distribute the conversion.
14783 2017-09-21  Segher Boessenkool  <segher@kernel.crashing.org>
14785         * haifa-sched.c: Rename insn_cost to insn_sched_cost.
14786         * sched-rgn.c: Ditto.
14787         * sel-sched-ir.c: Ditto.
14789 2017-09-21  Alexander Monakov  <amonakov@ispras.ru>
14791         * toplev.h (set_random_seed): Adjust return type.
14792         * toplev.c (init_local_tick): Move eager initialization of random_seed
14793         to get_random_seed.  Adjust comment.
14794         (init_random_seed): Inline to get_random_seed, delete.
14795         (get_random_seed): Initialize random_seed lazily.
14796         (set_random_seed): Do not return previous value.
14797         (print_switch_value): Do not call get_random_seed.
14799 2017-09-21  Evgeny Kudryashov  <kudryashov@ispras.ru>
14801         * cgraph.c (delete_function_version): New, broken out from...
14802         (cgraph_node::delete_function_version): ...here.  Rename to
14803         cgraph_node::delete_function_version_by_decl.  Update all uses.
14804         (cgraph_node::remove): Call delete_function_version.
14806 2017-09-21  Jakub Jelinek  <jakub@redhat.com>
14808         PR sanitizer/81715
14809         * tree-inline.c (expand_call_inline): Emit clobber stmts for
14810         VAR_DECLs to which addressable non-volatile parameters are mapped
14811         and for id->retvar after the return value assignment.  Clear
14812         id->retval and id->retbnd after inlining.
14814 2017-09-21  Richard Biener  <rguenther@suse.de>
14816         PR tree-optimization/82276
14817         PR tree-optimization/82244
14818         * tree-vrp.c (build_assert_expr_for): Set
14819         SSA_NAME_OCCURS_IN_ABNORMAL_PHI if the variable we assert on
14820         has it set.
14821         (remove_range_assertions): Revert earlier change.
14823 2017-09-21  Wilco Dijkstra  <wdijkstr@arm.com>
14825         PR target/71951
14826         * config/aarch64/aarch64.h (LIBGCC2_UNWIND_ATTRIBUTE): Define.
14828 2017-09-21  Richard Biener  <rguenther@suse.de>
14830         * graphite-isl-ast-to-gimple.c (graphite_regenerate_ast_isl):
14831         Restore valid IL after code generation errors.
14832         * graphite.c (graphite_transform_loops): Diagnose code
14833         generation issues as MSG_MISSED_OPTIMIZATION and continue
14834         with processing SCOPs.
14836 2017-09-21  Richard Sandiford  <richard.sandiford@linaro.org>
14837             Alan Hayward  <alan.hayward@arm.com>
14838             David Sherwood  <david.sherwood@arm.com>
14840         * calls.c (compute_argument_addresses): Use simplify_gen_binary
14841         rather than choosing between plus_constant and gen_rtx_<CODE>.
14842         * expr.c (emit_push_insn): Likewise.
14843         (expand_expr_real_2): Likewise.
14845 2017-09-21  Richard Sandiford  <richard.sandiford@linaro.org>
14846             Alan Hayward  <alan.hayward@arm.com>
14847             David Sherwood  <david.sherwood@arm.com>
14849         * loop-unroll.c (split_iv): Call copy_rtx on the step.
14851 2017-09-21  Richard Sandiford  <richard.sandiford@linaro.org>
14852             Alan Hayward  <alan.hayward@arm.com>
14853             David Sherwood  <david.sherwood@arm.com>
14855         * tree.c (find_atomic_core_type): Check tree_fits_uhwi_p before
14856         calling tree_to_uhwi.
14858 2017-09-21  Richard Sandiford  <richard.sandiford@linaro.org>
14859             Alan Hayward  <alan.hayward@arm.com>
14860             David Sherwood  <david.sherwood@arm.com>
14862         * tree-ssa-ccp.c (get_value_for_expr): Use a positive test for
14863         INTEGER_CST rather than a negative test for ADDR_EXPR.
14865 2017-09-21  Richard Sandiford  <richard.sandiford@linaro.org>
14866             Alan Hayward  <alan.hayward@arm.com>
14867             David Sherwood  <david.sherwood@arm.com>
14869         * tree-vrp.c (extract_range_from_binary_expr_1): Check
14870         int_cst_rangeN before calling value_range_constant_singleton (&vrN).
14872 2017-09-21  Richard Biener  <rguenther@suse.de>
14874         PR tree-optimization/71351
14875         * graphite-isl-ast-to-gimple.c (translate_isl_ast_to_gimple::
14876         graphite_create_new_loop_guard): Remove, fold remaining parts
14877         into caller ...
14878         (translate_isl_ast_node_for): ... here and simplify.
14880 2017-09-21  Jakub Jelinek  <jakub@redhat.com>
14882         PR target/82260
14883         * config/i386/i386.md (*movqi_internal): Replace (=q,q) alternative
14884         with (=Q,Q), (=R,R) and (=r,r) alternatives, only enable the
14885         latter two for 64-bit, renumber alternatives, for -Os imov (=q,n)
14886         alternative always use QI mode, for -Os imov (=R,R) alternative
14887         always use SI mode, for imov (=Q,Q) or (=r,r) alternatives
14888         ignore -Os.
14890 2017-09-20  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
14891             Jeff Law  <law@redhat.com>
14893         * config/s390/s390.c (MIN_UNROLL_PROBES): Define.
14894         (allocate_stack_space): New function, partially extracted from
14895         s390_emit_prologue.
14896         (s390_emit_prologue): Track offset to most recent stack probe.
14897         Code to allocate space moved into allocate_stack_space.
14898         Dump actions when no stack is allocated.
14899         (s390_prologue_plus_offset): New function.
14900         (s390_emit_stack_probe): Likewise.
14902 2017-09-20  Alexandre Oliva  <aoliva@redhat.com>
14904         * common.opt (Wa, Wl, Wp, g, gz=): Add
14905         RejectNegative.
14906         (gno-column-info): Remove.
14907         (gcolumn-info): Drop RejectNegative.
14908         (gno-): New prefix.
14909         (gno-record-gcc-switches): Remove.
14910         (grecord-gcc-switches): Drop RejectNegative.
14911         (gno-split-dwarf): Remove.
14912         (gsplit-dwarf): Drop RejectNegative.
14913         (gno-strict-dwarf): Remove.
14914         (gstrict-dwarf): Drop RejectNegative.
14915         * config/darwin.opt (gfull, gused): Add RejectNegative.
14916         * dwarf2out.c (gen_producer_string): Drop
14917         gno-record-gcc-switches handler.
14918         * optc-gen.awk: Add g to prefixes with negative forms.
14919         * opts-common.c (remapping_prefix_p): New.
14920         (find_opt): Check it.
14921         (generate_canonical_option): Test g prefix.
14922         (option_map): Add -gno- mapping.
14923         (add_misspelling_candidates): Check remapping_prefix_p.
14925 2017-09-20  Jeff Law  <law@redhat.com>
14927         * config/powerpcspe/powerpcspe.c (rs6000_expand_prologue): Fix
14928         thinko in stack clash protection support.
14930         * explow.c (compute_stack_clash_protection_loop_data): Use
14931         CONST_INT_P instead of explicit test.  Verify object is a
14932         CONST_INT_P before looking at INTVAL.
14933         (anti_adjust_stack_and_probe_stack_clash): Use CONST_INT_P
14934         instead of explicit test.
14936 2017-09-20  Segher Boessenkool  <segher@kernel.crashing.org>
14938         PR target/77687
14939         * config/rs6000/rs6000.md (stack_restore_tie): Store to a scratch
14940         address instead of to r1 and r11.
14942 2017-09-20  Sebastian Peryt  <sebastian.peryt@intel.com>
14944         * config.gcc: Support "knm".
14945         * config/i386/driver-i386.c (host_detect_local_cpu): Detect "knm".
14946         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
14947         PROCESSOR_KNM.
14948         * config/i386/i386.c (m_KNM): Define.
14949         (processor_target_table): Add "knm".
14950         (PTA_KNM): Define.
14951         (ix86_option_override_internal): Add "knm".
14952         (ix86_issue_rate): Add PROCESSOR_KNM.
14953         (ix86_adjust_cost): Ditto.
14954         (ia32_multipass_dfa_lookahead): Ditto.
14955         (get_builtin_code_for_version): Handle PROCESSOR_KNM.
14956         (fold_builtin_cpu): Add M_INTEL_KNM.
14957         * config/i386/i386.h (processor_costs): Define TARGET_KNM.
14958         (processor_type): Add PROCESSOR_KNM.
14959         * config/i386/x86-tune.def: Add m_KNM.
14960         * doc/invoke.texi: Add knm as x86 -march=/-mtune= CPU type.
14962 2017-09-20  Richard Biener  <rguenther@suse.de>
14964         PR tree-optimization/80213
14965         * graphite-scop-detection.c (trivially_empty_bb_p): Labels
14966         are allowed in empty BBs as well.
14967         (canonicalize_loop_closed_ssa): Also look for other complex
14968         edges.
14969         (scop_detection::get_sese): Include the loop-closed PHI block
14970         in loop SESEs.
14971         (scop_detection::merge_sese): Remove code adding extra blocks.
14972         (scop_detection::region_has_one_loop): Adjust for get_sese changes.
14973         (build_scops): Assert the final returned scop is invalid.
14975 2017-09-20  Richard Biener  <rguenther@suse.de>
14977         PR tree-optimization/82264
14978         * tree-ssa-sccvn.c (vn_phi_eq): Use safe_dyn_cast to check
14979         for GIMPLE_CONDs.
14980         (vn_phi_lookup): Likewise.
14981         (vn_phi_insert): Likewise.
14983 2017-09-20  Jakub Jelinek  <jakub@redhat.com>
14985         * dwarf2out.c (tree_add_const_value_attribute): For INTEGER_CST
14986         that fits into uhwi or shwi, add DW_AT_const_value regardless
14987         of early_dwarf without going through RTL, using add_AT_unsigned
14988         or add_AT_int.
14990         * dwarf2out.c (DEBUG_LTO_DWO_INFO_SECTION): Reorder defines.
14991         (DEBUG_LTO_ABBREV_SECTION): Likewise.
14992         (DEBUG_LTO_MACINFO_SECTION): Likewise.
14993         (DEBUG_MACRO_SECTION): Likewise.
14994         (DEBUG_LTO_MACRO_SECTION): Likewise.
14995         (DEBUG_STR_DWO_SECTION): Likewise.
14996         (DEBUG_LTO_STR_DWO_SECTION): Likewise.
14997         (DEBUG_LTO_LINE_SECTION): Drop .dwo suffix from the name.
14998         (DEBUG_LTO_DWO_LINE_SECTION): Define.
14999         (DEBUG_LTO_LINE_STR_SECTION): Define.
15000         (init_sections_and_labels): Initialize debug_line_str_section
15001         variable.  Initialize debug_loc_section for -gdwarf-5 to
15002         DEBUG_LOCLISTS_SECTION.  Formatting fixes.
15004 2017-09-20  Richard Biener  <rguenther@suse.de>
15006         * graphite-sese-to-poly.c (extract_affine): Properly handle
15007         POINTER_PLUS_EXPR, BIT_NOT_EXPR and conversion to signed.
15009 2017-09-20  Richard Biener  <rguenther@suse.de>
15011         PR tree-optimization/81373
15012         * graphite-scop-detection.c (build_cross_bb_scalars_def):
15013         Force SESE live-out defs to be handled even if they are
15014         scev_analyzable_p.
15016 2017-09-19  Jeff Law  <law@redhat.com>
15018         * combine-stack-adj.c (combine_stack_adjustments_for_block): Do
15019         nothing for stack adjustments with REG_STACK_CHECK.
15020         * sched-deps.c (parse_add_or_inc): Reject insns with
15021         REG_STACK_CHECK from dependency breaking.
15022         * config/i386/i386.c (pro_epilogue_adjust_stack): Return insn.
15023         (ix86_adjust_satck_and_probe_stack_clash): Add REG_STACK_NOTEs.
15024         * reg-notes.def (STACK_CHECK): New note.
15026         * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): New.
15027         (ix86_expand_prologue): Dump stack clash info as needed.
15028         Call ix86_adjust_stack_and_probe_stack_clash as needed.
15030         * function.c (dump_stack_clash_frame_info): New function.
15031         * function.h (dump_stack_clash_frame_info): Prototype.
15032         (enum stack_clash_probes): New enum.
15034         * config/alpha/alpha.c (alpha_expand_prologue): Also check
15035         flag_stack_clash_protection.
15036         * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Likewise.
15037         (arm_expand_prologue, thumb1_expand_prologue): Likewise.
15038         (arm_frame_pointer_required): Likewise.
15039         * config/ia64/ia64.c (ia64_compute_frame_size): Likewise.
15040         (ia64_expand_prologue): Likewise.
15041         * config/mips/mips.c (mips_expand_prologue): Likewise.
15042         * config/powerpcspe/powerpcspe.c (rs6000_expand_prologue): Likewise.
15043         * config/sparc/sparc.c (sparc_expand_prologue): Likewise.
15044         (sparc_flat_expand_prologue): Likewise.
15045         * config/spu/spu.c (spu_expand_prologue): Likewise.
15047         * explow.c: Include "params.h".
15048         (anti_adjust_stack_and_probe_stack_clash): New function.
15049         (get_stack_check_protect): Likewise.
15050         (compute_stack_clash_protection_loop_data): Likewise.
15051         (emit_stack_clash_protection_loop_start): Likewise.
15052         (emit_stack_clash_protection_loop_end): Likewise.
15053         (allocate_dynamic_stack_space): Use get_stack_check_protect.
15054         Use anti_adjust_stack_and_probe_stack_clash.
15055         * explow.h (compute_stack_clash_protection_loop_data): Prototype.
15056         (emit_stack_clash_protection_loop_start): Likewise.
15057         (emit_stack_clash_protection_loop_end): Likewise.
15058         * rtl.h (get_stack_check_protect): Prototype.
15059         * target.def (stack_clash_protection_final_dynamic_probe): New hook.
15060         * targhooks.c (default_stack_clash_protection_final_dynamic_probe): New.
15061         * targhooks.h (default_stack_clash_protection_final_dynamic_probe):
15062         Prototype.
15063         * doc/tm.texi.in (TARGET_STACK_CLASH_PROTECTION_FINAL_DYNAMIC_PROBE):
15064         Add @hook.
15065         * doc/tm.texi: Rebuilt.
15066         * config/aarch64/aarch64.c (aarch64_expand_prologue): Use
15067         get_stack_check_protect.
15068         * config/alpha/alpha.c (alpha_expand_prologue): Likewise.
15069         * config/arm/arm.c (arm_expand_prologue): Likewise.
15070         (arm_frame_pointer_required): Likewise.
15071         * config/i386/i386.c (ix86_expand_prologue): Likewise.
15072         * config/ia64/ia64.c (ia64_expand_prologue): Likewise.
15073         * config/mips/mips.c (mips_expand_prologue): Likewise.
15074         * config/powerpcspe/powerpcspe.c (rs6000_emit_prologue): Likewise.
15075         * config/rs6000/rs6000.c (rs6000_emit_prologue): Likewise.
15076         * config/sparc/sparc.c (sparc_expand_prologue): Likewise.
15077         (sparc_flat_expand_prologue): Likewise.
15079         * common.opt (-fstack-clash-protection): New option.
15080         * flag-types.h (enum stack_check_type): Note difference between
15081         -fstack-check= and -fstack-clash-protection.
15082         * params.def (PARAM_STACK_CLASH_PROTECTION_GUARD_SIZE): New PARAM.
15083         (PARAM_STACK_CLASH_PROTECTION_PROBE_INTERVAL): Likewise.
15084         * toplev.c (process_options): Issue warnings/errors for cases
15085         not handled with -fstack-clash-protection.
15086         * doc/invoke.texi (-fstack-clash-protection): Document new option.
15087         (-fstack-check): Note additional problem with -fstack-check=generic.
15088         Note that -fstack-check is primarily for Ada and refer users
15089         to -fstack-clash-protection for stack-clash-protection.
15090         Document new params for stack clash protection.
15092 2017-09-19  Uros Bizjak  <ubizjak@gmail.com>
15094         * config/i386/i386.c (ix86_split_long_move): Do not handle
15095         address used for LEA in a special way.
15097 2017-09-19  Segher Boessenkool  <segher@kernel.crashing.org>
15099         * simplify-rtx.c (simplify_binary_operation_1): Fix typo in comment.
15101 2017-09-19  Martin Sebor  <msebor@redhat.com>
15103         PR c/81854
15104         * cgraphunit.c (handle_alias_pairs): Reject aliases between functions
15105         of incompatible types.
15107 2017-09-19  Will Schmidt  <will_schmidt@vnet.ibm.com>
15109         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add handling
15110         for early folding of vector loads (ALTIVEC_BUILTIN_LVX_*).
15111         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
15112         Remove obsoleted code for handling ALTIVEC_BUILTIN_VEC_LD.
15114 2017-09-19  Richard Biener  <rguenther@suse.de>
15116         PR tree-optimization/82244
15117         * tree-vrp.c (remove_range_assertions): Do not propagate
15118         a constant to abnormals but replace the assert with a copy.
15120 2017-09-19  Alexander Monakov  <amonakov@ispras.ru>
15122         PR rtl-optimization/57878
15123         PR rtl-optimization/68988
15124         * lra-assigns.c (reload_pseudo_compare_func): Remove fragmentation
15125         avoidance test involving non_reload_pseudos.  Move frequency test
15126         below the general fragmentation avoidance test.
15128 2017-09-19  Richard Biener  <rguenther@suse.de>
15130         PR tree-optimization/69728
15131         * graphite-sese-to-poly.c (schedule_error): New global.
15132         (add_loop_schedule): Handle empty domain by failing the
15133         schedule.
15134         (build_original_schedule): Handle schedule_error.
15136 2017-09-19  Richard Biener  <rguenther@suse.de>
15138         * graphite-scop-detection.c (scop_detection::can_represent_loop):
15139         Do not iterate to sibling loops but only to siblings of inner
15140         loops.
15142 2017-09-18  Andreas Schwab  <schwab@linux-m68k.org>
15144         PR target/81613
15145         * config/m68k/m68k.md (moveq feeding equality comparison): Check
15146         that the registers are different.
15148 2017-09-18  Uros Bizjak  <ubizjak@gmail.com>
15150         * config/i386/i386.c (fold_builtin_cpu): Add M_AMDFAM17H
15151         to processor_model and "amdfam17h" to arch_names_table.
15152         * doc/extend.texi (__builtin_cpu_is): Document amdfam17h CPU name.
15154 2017-09-18  Jakub Jelinek  <jakub@redhat.com>
15156         PR c/82234
15157         * doc/extend.texi: Add @findex entry for __builtin_shuffle.
15159 2017-09-18  Richard Sandiford  <richard.sandiford@linaro.org>
15160             Alan Hayward  <alan.hayward@arm.com>
15161             David Sherwood  <david.sherwood@arm.com>
15163         * tree-vectorizer.h (vect_slp_analyze_operations): Replace parameters
15164         with a vec_info *.
15165         * tree-vect-loop.c (vect_analyze_loop_operations): Update call
15166         accordingly.
15167         * tree-vect-slp.c (vect_slp_analyze_node_operations): Add vec_info *
15168         parameter.  Set SLP_TREE_NUMBER_OF_VEC_STMTS here rather than in
15169         vect_schedule_slp_instance.
15170         (vect_slp_analyze_operations): Replace parameters with a vec_info *.
15171         Update call to vect_slp_analyze_node_operations.  Simplify return
15172         value.
15173         (vect_slp_analyze_bb_1): Update call accordingly.
15174         (vect_schedule_slp_instance): Remove vectorization_factor parameter.
15175         Don't calculate SLP_TREE_NUMBER_OF_VEC_STMTS here.
15176         (vect_schedule_slp): Update call accordingly.
15178 2017-09-18  Richard Sandiford  <richard.sandiford@linaro.org>
15179             Alan Hayward  <alan.hayward@arm.com>
15180             David Sherwood  <david.sherwood@arm.com>
15182         * tree-vect-data-refs.c (vect_get_smallest_scalar_type): Cope
15183         with types that aren't in fact scalar.
15185 2017-09-18  Richard Sandiford  <richard.sandiford@linaro.org>
15187         * tree-vect-slp.c (vect_record_max_nunits): New function,
15188         split out from...
15189         (vect_build_slp_tree_1): ...here.
15190         (vect_build_slp_tree_2): Call it for phis too.
15192 2017-09-18  Richard Sandiford  <richard.sandiford@linaro.org>
15194         * tree-vect-stmts.c (vectorizable_mask_load_store): Pass mask_vectype
15195         to vect_get_vec_def_for_operand when getting the mask operand.
15197 2017-09-18  Richard Sandiford  <richard.sandiford@linaro.org>
15198             Alan Hayward  <alan.hayward@arm.com>
15199             David Sherwood  <david.sherwood@arm.com>
15201         * tree-vect-loop.c (vectorizable_live_operation): Fix type of
15202         bitstart.
15204 2017-09-18  Richard Sandiford  <richard.sandiford@linaro.org>
15205             Alan Hayward  <alan.hayward@arm.com>
15206             David Sherwood  <david.sherwood@arm.com>
15208         * tree-vect-loop.c (vectorizable_live_operation): Fix element size
15209         calculation for vector booleans.
15211 2017-09-18  Richard Sandiford  <richard.sandiford@linaro.org>
15212             Alan Hayward  <alan.hayward@arm.com>
15213             David Sherwood  <david.sherwood@arm.com>
15215         * tree-vect-stmts.c (can_vectorize_live_stmts): New function,
15216         split out from...
15217         (vect_transform_stmt): ...here.
15218         (vect_analyze_stmt): Use it instead of calling
15219         vectorizable_live_operation directly.
15221 2017-09-18  Cesar Philippidis  <cesar@codesourcery.com>
15223         * omp-offload.c (oacc_xform_loop): Enable SIMD vectorization on
15224         non-SIMT targets in acc vector loops.
15226 2017-09-18  Claudiu Zissulescu  <claziss@synopsys.com>
15228         * configure.ac: Add arc and check if assembler supports gdwarf2.
15229         * configure: Regenerate.
15231 2017-09-18  Richard Biener  <rguenther@suse.de>
15233         PR tree-optimization/82220
15234         * tree-vect-loop.c (vect_estimate_min_profitable_iters): Exclude
15235         epilogue niters from the min_profitable_iters compute.
15237 2017-09-18  Jakub Jelinek  <jakub@redhat.com>
15239         PR target/82145
15240         * config/i386/i386.c (ix86_init_large_pic_reg): Revert 2017-09-01
15241         changes.  Turn CODE_LABEL into NOTE_INSN_DELETED_LABEL immediately.
15242         (ix86_init_pic_reg): Revert 2017-09-01 changes.
15244 2017-09-18  Eric Botcazou  <ebotcazou@adacore.com>
15246         PR target/81361
15247         * dwarf2cfi.c (add_cfis_to_fde): Do not generate DW_CFA_set_loc after
15248         switching to a new text section.
15250 2017-09-18  Richard Biener  <rguenther@suse.de>
15252         * graphite-scop-detection.c (scop_detection::stmt_has_simple_data_ref):
15253         Simplify.
15254         (build_alias_set): Reject aliases with no access function.
15256 2017-09-18  Richard Biener  <rguenther@suse.de>
15258         PR tree-optimization/79622
15259         * graphite-scop-detection.c (build_cross_bb_scalars_def): Properly
15260         handle PHIs.
15261         (build_cross_bb_scalars_use): Likewise.
15263 2017-09-18  Pierre-Marie de Rodat  <derodat@adacore.com>
15265         * cgraph.h (cgraph_thunk_info): Fix a typo in a comment.
15267 2017-09-18  Alan Modra  <amodra@gmail.com>
15269         PR target/81996
15270         * gcc/config/rs6000/rs6000.c (rs6000_return_addr): Use
15271         stack_pointer_rtx for count 0.  Update comments.  Break up
15272         large rtl expression.
15274 2017-09-17  Daniel Santos  <daniel.santos@pobox.com>
15276         * config/i386/i386.c (xlogue_layout::STUB_NAME_MAX_LEN):
15277         Increase to 20 bytes.
15278         (xlogue_layout::s_stub_names): Add an additional size-2 diminsion.
15279         (xlogue_layout::get_stub_name): Modify to select the appropairate sse
15280         or avx version of the stub.
15282 2017-09-17  H.J. Lu  <hongjiu.lu@intel.com>
15284         PR target/82166
15285         * config/i386/i386.c (ix86_finalize_stack_frame_flags): Properly
15286         compute the minimum stack alignment.  Also update preferred stack
15287         boundary for leaf functions.
15289 2017-09-16  Richard Sandiford  <richard.sandiford@linaro.org>
15291         PR tree-optimization/82228
15292         * tree-vect-loop.c (vectorizable_live_operation): Move initialization
15293         of ncopies.
15295 2017-09-16  Chung-Ju Wu  <jasonwucj@gmail.com>
15297         * common/config/nds32/nds32-common.c
15298         (nds32_option_optimization_table): Refine formatting.
15299         (nds32_option_optimization_table): Use -fsched-pressure and
15300         -fomit-frame-pointer for specific optimization level.
15302 2017-09-16  Chung-Ju Wu  <jasonwucj@gmail.com>
15304         * config/nds32/nds32.c: Refine formatting and comments.
15305         * config/nds32/nds32.h: Likewise.
15306         * config/nds32/nds32.md: Likewise.
15307         * config/nds32/nds32-cost.c: Likewise.
15308         * config/nds32/nds32-isr.c: Likewise.
15309         * config/nds32/nds32-md-auxiliary.c: Likewise.
15310         * config/nds32/nds32-multiple.md: Likewise.
15311         * config/nds32/nds32-predicates.c: Likewise.
15313 2017-09-15  Andrew Sutton  <andrew.n.sutton@gmail.com>
15314             Jakub Jelinek  <jakub@redhat.com>
15316         Add support for -std=c++2a.
15317         * doc/cpp.texi (__cplusplus): Document value for -std=c++2a
15318         or -std=gnu+2a.
15319         * doc/invoke.texi: Document -std=c++2a and -std=gnu++2a.
15321 2017-09-15  Steve Ellcey  <sellcey@cavium.com>
15323         PR target/82066
15324         * doc/extend.texi (Common Function Attributes): Add 
15325         references to ARM, AArch64, and S/390 specific attributes.
15326         (Function Specific Option Pragmas): Add AArch64 and S/390
15327         to list of back ends that support the target pragma.
15329 2017-09-15  Nathan Sidwell  <nathan@acm.org>
15331         * doc/standards.texi: Fix C++17 description.  Update URLs for
15332         C++11 & 14.
15334 2017-09-15  Bernd Edlinger  <bernd.edlinger@hotmail.de>
15336         * common.opt (Wcast-align=strict): New warning option.
15337         * doc/invoke.texi: Document -Wcast-align=strict. 
15339 2017-09-15  Pierre-Marie de Rodat  <derodat@adacore.com>
15341         * cgraph.h (cgraph_thunk_info): Add comments.
15342         * cgraph.c (cgraph_node::create_thunk): Adjust comment, make
15343         assert for VIRTUAL_* arguments stricter.
15345 2017-09-15  Jackson Woodruff  <jackson.woodruff@arm.com>
15347         PR tree-optimization/71026
15348         * match.pd: Move RDIV patterns from fold-const.c
15349         * fold-const.c (distribute_real_division): Removed.
15350         (fold_binary_loc): Remove calls to distribute_real_divison.
15352 2017-09-15  Jakub Jelinek  <jakub@redhat.com>
15354         * doc/invoke.texi: Document -std=c++17 and -std=gnu++17 and document
15355         c++1z and gnu++1z as deprecated.  Change other references to
15356         -std=c++1z to -std=c++17 and -std=gnu++1z to -std=gnu++17.
15357         Change -Wc++1z-compat to -Wc++17-compat.
15358         * doc/cpp.texi: Document -std=c++17 defines __cplusplus 201703L.
15359         * dwarf2out.c (highest_c_language): Handle C++17.
15360         (gen_compile_unit_die): Likewise.
15362 2017-09-15  Jakub Jelinek  <jakub@redhat.com>
15364         PR rtl-optimization/82192
15365         * combine.c (make_extraction): Don't look through non-paradoxical
15366         SUBREGs or TRUNCATE if pos + len is or might be bigger than
15367         inner's mode.
15369 2017-09-15  Richard Sandiford  <richard.sandiford@linaro.org>
15370             Alan Hayward  <alan.hayward@arm.com>
15371             David Sherwood  <david.sherwood@arm.com>
15373         * target.def (function_arg_offset): New hook.
15374         * targhooks.h (default_function_arg_offset): Declare.
15375         * targhooks.c (default_function_arg_offset): New function.
15376         * function.c (locate_and_pad_parm): Use
15377         targetm.calls.function_arg_offset instead of FUNCTION_ARG_OFFSET.
15378         * doc/tm.texi.in (FUNCTION_ARG_OFFSET): Replace with...
15379         (TARGET_FUNCTION_ARG_OFFSET): ...this.
15380         * doc/tm.texi: Regenerate.
15381         * config/spu/spu.h (FUNCTION_ARG_OFFSET): Delete.
15382         * config/spu/spu.c (spu_function_arg_offset): New function.
15383         (TARGET_FUNCTION_ARG_OFFSET): Redefine.
15384         * system.h (FUNCTION_ARG_OFFSET): Poison.
15386 2017-09-15  Richard Sandiford  <richard.sandiford@linaro.org>
15387             Alan Hayard  <alan.hayward@arm.com>
15388             David Sherwood  <david.sherwood@arm.com>
15390         * target.def (truly_noop_truncation): New hook.
15391         (mode_rep_extended): Refer to TARGET_TRULY_NOOP_TRUNCATION rather
15392         than TRULY_NOOP_TRUNCATION.
15393         * hooks.h (hook_bool_uint_uint_true): Declare.
15394         * hooks.c (hook_bool_uint_uint_true): New function.
15395         * doc/tm.texi.in (TRULY_NOOP_TRUNCATION): Replace with...
15396         (TARGET_TRULY_NOOP_TRUNCATION): ...this.
15397         * doc/tm.texi: Regenerate.
15398         * combine.c (make_extraction): Refer to TARGET_TRULY_NOOP_TRUNCATION
15399         rather than TRULY_NOOP_TRUNCATION in comments.
15400         (simplify_comparison): Likewise.
15401         (record_truncated_value): Likewise.
15402         * expmed.c (extract_bit_field_1): Likewise.
15403         (extract_split_bit_field): Likewise.
15404         * convert.c (convert_to_integer_1): Use targetm.truly_noop_truncation
15405         instead of TRULY_NOOP_TRUNCATION.
15406         * function.c (assign_parm_setup_block): Likewise.
15407         * machmode.h (TRULY_NOOP_TRUNCATION_MODES_P): Likewise.
15408         * rtlhooks.c: Include target.h.
15409         * config/aarch64/aarch64.h (TRULY_NOOP_TRUNCATION): Delete.
15410         * config/alpha/alpha.h (TRULY_NOOP_TRUNCATION): Delete.
15411         * config/arc/arc.h (TRULY_NOOP_TRUNCATION): Delete.
15412         * config/arm/arm.h (TRULY_NOOP_TRUNCATION): Delete.
15413         * config/avr/avr.h (TRULY_NOOP_TRUNCATION): Delete.
15414         * config/bfin/bfin.h (TRULY_NOOP_TRUNCATION): Delete.
15415         * config/c6x/c6x.h (TRULY_NOOP_TRUNCATION): Delete.
15416         * config/cr16/cr16.h (TRULY_NOOP_TRUNCATION): Delete.
15417         * config/cris/cris.h (TRULY_NOOP_TRUNCATION): Delete.
15418         * config/epiphany/epiphany.h (TRULY_NOOP_TRUNCATION): Delete.
15419         * config/fr30/fr30.h (TRULY_NOOP_TRUNCATION): Delete.
15420         * config/frv/frv.h (TRULY_NOOP_TRUNCATION): Delete.
15421         * config/ft32/ft32.h (TRULY_NOOP_TRUNCATION): Delete.
15422         * config/h8300/h8300.h (TRULY_NOOP_TRUNCATION): Delete.
15423         * config/i386/i386.h (TRULY_NOOP_TRUNCATION): Delete.
15424         * config/ia64/ia64.h (TRULY_NOOP_TRUNCATION): Delete.
15425         * config/iq2000/iq2000.h (TRULY_NOOP_TRUNCATION): Delete.
15426         * config/lm32/lm32.h (TRULY_NOOP_TRUNCATION): Delete.
15427         * config/m32c/m32c.h (TRULY_NOOP_TRUNCATION): Delete.
15428         * config/m32r/m32r.h (TRULY_NOOP_TRUNCATION): Delete.
15429         * config/m68k/m68k.h (TRULY_NOOP_TRUNCATION): Delete.
15430         * config/mcore/mcore.h (TRULY_NOOP_TRUNCATION): Delete.
15431         * config/microblaze/microblaze.h (TRULY_NOOP_TRUNCATION): Delete.
15432         * config/mips/mips.h (TRULY_NOOP_TRUNCATION): Delete.
15433         * config/mips/mips.c (mips_truly_noop_truncation): New function.
15434         (TARGET_TRULY_NOOP_TRUNCATION): Redefine.
15435         * config/mips/mips.md: Refer to TARGET_TRULY_NOOP_TRUNCATION
15436         rather than TRULY_NOOP_TRUNCATION in comments.
15437         * config/mmix/mmix.h (TRULY_NOOP_TRUNCATION): Delete.
15438         * config/mn10300/mn10300.h (TRULY_NOOP_TRUNCATION): Delete.
15439         * config/moxie/moxie.h (TRULY_NOOP_TRUNCATION): Delete.
15440         * config/msp430/msp430.h (TRULY_NOOP_TRUNCATION): Delete.
15441         * config/nds32/nds32.h (TRULY_NOOP_TRUNCATION): Delete.
15442         * config/nios2/nios2.h (TRULY_NOOP_TRUNCATION): Delete.
15443         * config/nvptx/nvptx.h (TRULY_NOOP_TRUNCATION): Delete.
15444         * config/pa/pa.h (TRULY_NOOP_TRUNCATION): Delete.
15445         * config/pdp11/pdp11.h (TRULY_NOOP_TRUNCATION): Delete.
15446         * config/powerpcspe/powerpcspe.h (TRULY_NOOP_TRUNCATION): Delete.
15447         * config/riscv/riscv.h (TRULY_NOOP_TRUNCATION): Delete.
15448         * config/riscv/riscv.md: Refer to TARGET_TRULY_NOOP_TRUNCATION
15449         rather than TRULY_NOOP_TRUNCATION in comments.
15450         * config/rl78/rl78.h (TRULY_NOOP_TRUNCATION): Delete.
15451         * config/rs6000/rs6000.h (TRULY_NOOP_TRUNCATION): Delete.
15452         * config/rx/rx.h (TRULY_NOOP_TRUNCATION): Delete.
15453         * config/s390/s390.h (TRULY_NOOP_TRUNCATION): Delete.
15454         * config/sh/sh.h (MAYBE_BASE_REGISTER_RTX_P): Remove
15455         TRULY_NOOP_TRUNCATION condition.
15456         (MAYBE_INDEX_REGISTER_RTX_P): Likewise.
15457         (TRULY_NOOP_TRUNCATION): Delete.
15458         * config/sparc/sparc.h (TRULY_NOOP_TRUNCATION): Delete.
15459         * config/spu/spu.h (TRULY_NOOP_TRUNCATION): Delete.
15460         * config/spu/spu.c (spu_truly_noop_truncation): New function.
15461         (TARGET_TRULY_NOOP_TRUNCATION): Redefine.
15462         * config/stormy16/stormy16.h (TRULY_NOOP_TRUNCATION): Delete.
15463         * config/tilegx/tilegx.h (TRULY_NOOP_TRUNCATION): Delete.
15464         * config/tilegx/tilegx.c (tilegx_truly_noop_truncation): New fuction.
15465         (TARGET_TRULY_NOOP_TRUNCATION): Redefine.
15466         * config/tilegx/tilegx.md: Refer to TARGET_TRULY_NOOP_TRUNCATION
15467         rather than TRULY_NOOP_TRUNCATION in comments.
15468         * config/tilepro/tilepro.h (TRULY_NOOP_TRUNCATION): Delete.
15469         * config/v850/v850.h (TRULY_NOOP_TRUNCATION): Delete.
15470         * config/vax/vax.h (TRULY_NOOP_TRUNCATION): Delete.
15471         * config/visium/visium.h (TRULY_NOOP_TRUNCATION): Delete.
15472         * config/xtensa/xtensa.h (TRULY_NOOP_TRUNCATION): Delete.
15473         * system.h (TRULY_NOOP_TRUNCATION): Poison.
15475 2017-09-15  Christophe Lyon  <christophe.lyon@linaro.org>
15477         PR target/67591
15478         * config/arm/arm.md (*cmp_and): Add enabled_for_depr_it attribute.
15479         (*cmp_ior): Likewise.
15480         (*ior_scc_scc): Add alternative for enabled_for_depr_it attribute.
15481         (*ior_scc_scc_cmp): Likewise.
15482         (*and_scc_scc): Likewise.
15483         (*and_scc_scc_cmp): Likewise.
15485 2017-09-15  Richard Sandiford  <richard.sandiford@linaro.org>
15486             Alan Hayard  <alan.hayward@arm.com>
15487             David Sherwood  <david.sherwood@arm.com>
15489         * target.def (can_change_mode_class): New hook.
15490         (mode_rep_extended): Refer to it instead of CANNOT_CHANGE_MODE_CLASS.
15491         (hard_regno_nregs): Likewise.
15492         * hooks.h (hook_bool_mode_mode_reg_class_t_true): Declare.
15493         * hooks.c (hook_bool_mode_mode_reg_class_t_true): New function.
15494         * doc/tm.texi.in (CANNOT_CHANGE_MODE_CLASS): Replace with...
15495         (TARGET_CAN_CHANGE_MODE_CLASS): ...this.
15496         (LOAD_EXTEND_OP): Update accordingly.
15497         * doc/tm.texi: Regenerate.
15498         * doc/rtl.texi: Refer to TARGET_CAN_CHANGE_MODE_CLASS instead of
15499         CANNOT_CHANGE_MODE_CLASS.
15500         * hard-reg-set.h (REG_CANNOT_CHANGE_MODE_P): Replace with...
15501         (REG_CAN_CHANGE_MODE_P): ...this new macro.
15502         * combine.c (simplify_set): Update accordingly.
15503         * emit-rtl.c (validate_subreg): Likewise.
15504         * recog.c (general_operand): Likewise.
15505         * regcprop.c (mode_change_ok): Likewise.
15506         * reload1.c (choose_reload_regs): Likewise.
15507         (inherit_piecemeal_p): Likewise.
15508         * rtlanal.c (simplify_subreg_regno): Likewise.
15509         * postreload.c (reload_cse_simplify_set): Use REG_CAN_CHANGE_MODE_P
15510         instead of CANNOT_CHANGE_MODE_CLASS.
15511         (reload_cse_simplify_operands): Likewise.
15512         * reload.c (push_reload): Use targetm.can_change_mode_class
15513         instead of CANNOT_CHANGE_MODE_CLASS.
15514         (push_reload): Likewise.  Also use REG_CAN_CHANGE_MODE_P instead of
15515         REG_CANNOT_CHANGE_MODE_P.
15516         * config/alpha/alpha.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15517         * config/alpha/alpha.c (alpha_can_change_mode_class): New function.
15518         (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15519         * config/arm/arm.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15520         * config/arm/arm.c (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15521         (arm_can_change_mode_class): New function.
15522         * config/arm/neon.md: Refer to TARGET_CAN_CHANGE_MODE_CLASS rather
15523         than CANNOT_CHANGE_MODE_CLASS in comments.
15524         * config/i386/i386.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15525         * config/i386/i386-protos.h (ix86_cannot_change_mode_class): Delete.
15526         * config/i386/i386.c (ix86_cannot_change_mode_class): Replace with...
15527         (ix86_can_change_mode_class): ...this new function, inverting the
15528         sense of the return value.
15529         (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15530         * config/ia64/ia64.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15531         * config/ia64/ia64.c (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15532         (ia64_can_change_mode_class): New function.
15533         * config/m32c/m32c.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15534         * config/m32c/m32c-protos.h (m32c_cannot_change_mode_class): Delete.
15535         * config/m32c/m32c.c (m32c_cannot_change_mode_class): Replace with...
15536         (m32c_can_change_mode_class): ...this new function, inverting the
15537         sense of the return value.
15538         (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15539         * config/mips/mips.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15540         * config/mips/mips-protos.h (mips_cannot_change_mode_class): Delete.
15541         * config/mips/mips.c (mips_cannot_change_mode_class): Replace with...
15542         (mips_can_change_mode_class): ...this new function, inverting the
15543         sense of the return value.
15544         (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15545         * config/msp430/msp430.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15546         * config/msp430/msp430.c (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15547         (msp430_can_change_mode_class): New function.
15548         * config/nvptx/nvptx.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15549         * config/nvptx/nvptx.c (nvptx_can_change_mode_class): New function.
15550         (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15551         * config/pa/pa32-regs.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15552         * config/pa/pa64-regs.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15553         * config/pa/pa-protos.h (pa_cannot_change_mode_class): Delete.
15554         * config/pa/pa.c (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15555         (pa_cannot_change_mode_class): Replace with...
15556         (pa_can_change_mode_class): ...this new function, inverting the
15557         sense of the return value.
15558         (pa_modes_tieable_p): Refer to TARGET_CAN_CHANGE_MODE_CLASS rather
15559         than CANNOT_CHANGE_MODE_CLASS in comments.
15560         * config/pdp11/pdp11.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15561         * config/pdp11/pdp11-protos.h (pdp11_cannot_change_mode_class): Delete.
15562         * config/pdp11/pdp11.c (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15563         (pdp11_cannot_change_mode_class): Replace with...
15564         (pdp11_can_change_mode_class): ...this new function, inverting the
15565         sense of the return value.
15566         * config/powerpcspe/powerpcspe.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15567         * config/powerpcspe/powerpcspe-protos.h
15568         (rs6000_cannot_change_mode_class_ptr): Delete.
15569         * config/powerpcspe/powerpcspe.c
15570         (rs6000_cannot_change_mode_class_ptr): Delete.
15571         (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15572         (rs6000_option_override_internal): Assign to
15573         targetm.can_change_mode_class instead of
15574         rs6000_cannot_change_mode_class_ptr.
15575         (rs6000_cannot_change_mode_class): Replace with...
15576         (rs6000_can_change_mode_class): ...this new function, inverting the
15577         sense of the return value.
15578         (rs6000_debug_cannot_change_mode_class): Replace with...
15579         (rs6000_debug_can_change_mode_class): ...this new function.
15580         * config/riscv/riscv.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15581         * config/riscv/riscv.c (riscv_can_change_mode_class): New function.
15582         (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15583         * config/rs6000/rs6000.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15584         * config/rs6000/rs6000-protos.h (rs6000_cannot_change_mode_class_ptr):
15585         Delete.
15586         * config/rs6000/rs6000.c (rs6000_cannot_change_mode_class_ptr): Delete.
15587         (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15588         (rs6000_option_override_internal): Assign to
15589         targetm.can_change_mode_class instead of
15590         rs6000_cannot_change_mode_class_ptr.
15591         (rs6000_cannot_change_mode_class): Replace with...
15592         (rs6000_can_change_mode_class): ...this new function, inverting the
15593         sense of the return value.
15594         (rs6000_debug_cannot_change_mode_class): Replace with...
15595         (rs6000_debug_can_change_mode_class): ...this new function.
15596         * config/s390/s390.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15597         * config/s390/s390-protos.h (s390_cannot_change_mode_class): Delete.
15598         * config/s390/s390.c (s390_cannot_change_mode_class): Replace with...
15599         (s390_can_change_mode_class): ...this new function, inverting the
15600         sense of the return value.
15601         (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15602         * config/sh/sh.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15603         * config/sh/sh-protos.h (sh_cannot_change_mode_class): Delete.
15604         * config/sh/sh.c (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15605         (sh_cannot_change_mode_class): Replace with...
15606         (sh_can_change_mode_class): ...this new function, inverting the
15607         sense of the return value.
15608         * config/sparc/sparc.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15609         * config/sparc/sparc.c (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15610         (sparc_can_change_mode_class): New function.
15611         * config/spu/spu.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15612         * config/spu/spu.c (spu_can_change_mode_class): New function.
15613         (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15614         * config/visium/visium.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15615         * config/visium/visium.c (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15616         (visium_can_change_mode_class): New function.
15617         * system.h (CANNOT_CHANGE_MODE_CLASS): Poison.
15619 2017-09-15  Richard Biener  <rguenther@suse.de>
15621         PR tree-optimization/82217
15622         * tree-ssa-sccvn.c (visit_phi): Properly handle all VN_TOP
15623         but not undefined case.
15625 2017-09-15  Jakub Jelinek  <jakub@redhat.com>
15627         PR target/82145
15628         * postreload.c (reload_cse_simplify_operands): Skip
15629         NOTE_INSN_DELETED_LABEL similarly to skipping CODE_LABEL.
15631 2017-09-15  Richard Biener  <rguenther@suse.de>
15633         PR tree-optimization/68823
15634         * graphite-scop-detection.c (build_alias_set): If we have a
15635         possible dependence check whether we can handle them by just
15636         looking at the DRs DR_ACCESS_FNs.
15637         (build_scops): If build_alias_set fails, fail the SCOP.
15639 2017-09-14  Michael Meissner  <meissner@linux.vnet.ibm.com>
15641         * config/rs6000/rs6000-builtin.def (BU_FLOAT128_1_HW): New macros
15642         to support float128 built-in functions that require the ISA 3.0
15643         hardware.
15644         (BU_FLOAT128_3_HW): Likewise.
15645         (SQRTF128): Add support for the IEEE 128-bit square root and fma
15646         built-in functions.
15647         (FMAF128): Likewise.
15648         (FMAQ): Likewise.
15649         * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add
15650         support for built-in functions that need the ISA 3.0 IEEE 128-bit
15651         floating point instructions.
15652         (rs6000_invalid_builtin): Likewise.
15653         (rs6000_builtin_mask_names): Likewise.
15654         * config/rs6000/rs6000.h (MASK_FLOAT128_HW): Likewise.
15655         (RS6000_BTM_FLOAT128_HW): Likewise.
15656         (RS6000_BTM_COMMON): Likewise.
15657         * config/rs6000/rs6000.md (fma<mode>4_hw): Add a generator
15658         function.
15659         * doc/extend.texi (RS/6000 built-in functions): Document the
15660         IEEE 128-bit floating point square root and fused multiply-add
15661         built-in functions.
15663 2017-09-14  Pat Haugen  <pthaugen@us.ibm.com>
15665         * config/rs6000/rs6000.c (rs6000_set_up_by_prologue): Make sure the TOC
15666         reg (r2) isn't in the set of registers defined in the prologue.
15668 2017-09-14  Richard Sandiford  <richard.sandiford@linaro.org>
15669             Alan Hayward  <alan.hayward@arm.com>
15670             David Sherwood  <david.sherwood@arm.com>
15672         * tree-vectorizer.h (_loop_vec_info): Add max_vectorization_factor.
15673         (LOOP_VINFO_MAX_VECT_FACTOR): New macro.
15674         (LOOP_VINFO_ORIG_VECT_FACTOR): Replace with...
15675         (LOOP_VINFO_ORIG_MAX_VECT_FACTOR): ...this new macro.
15676         * tree-vect-data-refs.c (vect_analyze_data_ref_dependences): Update
15677         accordingly.
15678         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
15679         max_vectorization_factor.
15680         (vect_analyze_loop_2): Set LOOP_VINFO_MAX_VECT_FACTOR.
15682 2017-09-14  Richard Sandiford  <richard.sandiford@linaro.org>
15683             Alan Hayward  <alan.hayward@arm.com>
15684             David Sherwood  <david.sherwood@arm.com>
15686         * tree-vectorizer.h (vect_min_worthwhile_factor): Delete.
15687         (vect_worthwhile_without_simd_p): Declare.
15688         * tree-vect-loop.c (vect_worthwhile_without_simd_p): New function.
15689         (vectorizable_reduction): Use it.
15690         * tree-vect-stmts.c (vectorizable_shift): Likewise.
15691         (vectorizable_operation): Likewise.
15693 2017-09-14  Richard Sandiford  <richard.sandiford@linaro.org>
15694             Alan Hayward  <alan.hayward@arm.com>
15695             David Sherwood  <david.sherwood@arm.com>
15697         * tree-vectorizer.h (vect_get_num_copies): New function.
15698         * tree-vect-data-refs.c (vect_get_data_access_cost): Use it.
15699         * tree-vect-loop.c (vectorizable_reduction): Likewise.
15700         (vectorizable_induction): Likewise.
15701         (vectorizable_live_operation): Likewise.
15702         * tree-vect-stmts.c (vectorizable_mask_load_store): Likewise.
15703         (vectorizable_bswap): Likewise.
15704         (vectorizable_call): Likewise.
15705         (vectorizable_conversion): Likewise.
15706         (vectorizable_assignment): Likewise.
15707         (vectorizable_shift): Likewise.
15708         (vectorizable_operation): Likewise.
15709         (vectorizable_store): Likewise.
15710         (vectorizable_load): Likewise.
15711         (vectorizable_condition): Likewise.
15712         (vectorizable_comparison): Likewise.
15713         (vect_analyze_stmt): Pass the slp node to vectorizable_live_operation.
15715 2017-09-14  Richard Sandiford  <richard.sandiford@linaro.org>
15716             Alan Hayward  <alan.hayward@arm.com>
15717             David Sherwood  <david.sherwood@arm.com>
15719         * tree-vect-loop.c (vectorizable_induction): Use gimple_build instead
15720         of vect_init_vector.
15722 2017-09-14  Richard Sandiford  <richard.sandiford@linaro.org>
15723             Alan Hayward  <alan.hayward@arm.com>
15724             David Sherwood  <david.sherwood@arm.com>
15726         * gimple-fold.h (gimple_build_vector_from_val): Declare, and provide
15727         an inline wrapper that provides a location.
15728         (gimple_build_vector): Likewise.
15729         * gimple-fold.c (gimple_build_vector_from_val): New function.
15730         (gimple_build_vector): Likewise.
15731         * tree-vect-loop.c (get_initial_def_for_reduction): Use the new
15732         functions to build the initial value.  Always return a gimple value.
15733         (get_initial_defs_for_reduction): Likewise.  Only compute
15734         neutral_vec once.
15735         (vect_create_epilog_for_reduction): Don't call force_gimple_operand or
15736         vect_init_vector on the results from get_initial_def(s)_for_reduction.
15737         (vectorizable_induction): Use gimple_build_vector rather than
15738         vect_init_vector.
15740 2017-09-14  Richard Sandiford  <richard.sandiford@linaro.org>
15741             Alan Hayward  <alan.hayward@arm.com>
15742             David Sherwood  <david.sherwood@arm.com>
15744         * target.h (vec_perm_indices): New typedef.
15745         (auto_vec_perm_indices): Likewise.
15746         * optabs-query.h: Include target.h
15747         (can_vec_perm_p): Take a vec_perm_indices *.
15748         * optabs-query.c (can_vec_perm_p): Likewise.
15749         (can_mult_highpart_p): Update accordingly.  Use auto_vec_perm_indices.
15750         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
15751         * tree-vect-generic.c (lower_vec_perm): Likewise.
15752         * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
15753         (vect_grouped_load_supported): Likewise.
15754         (vect_shift_permute_load_chain): Likewise.
15755         (vect_permute_store_chain): Use auto_vec_perm_indices.
15756         (vect_permute_load_chain): Likewise.
15757         * fold-const.c (fold_vec_perm): Take vec_perm_indices.
15758         (fold_ternary_loc): Update accordingly.  Use auto_vec_perm_indices.
15759         Update uses of can_vec_perm_p.
15760         * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Replace the
15761         mode with a number of elements.  Take a vec_perm_indices *.
15762         (vect_create_epilog_for_reduction): Update accordingly.
15763         Use auto_vec_perm_indices.
15764         (have_whole_vector_shift): Likewise.  Update call to can_vec_perm_p.
15765         * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
15766         (vect_transform_slp_perm_load): Likewise.
15767         (vect_schedule_slp_instance): Use auto_vec_perm_indices.
15768         * tree-vectorizer.h (vect_gen_perm_mask_any): Take a vec_perm_indices.
15769         (vect_gen_perm_mask_checked): Likewise.
15770         * tree-vect-stmts.c (vect_gen_perm_mask_any): Take a vec_perm_indices.
15771         (vect_gen_perm_mask_checked): Likewise.
15772         (vectorizable_mask_load_store): Use auto_vec_perm_indices.
15773         (vectorizable_store): Likewise.
15774         (vectorizable_load): Likewise.
15775         (perm_mask_for_reverse): Likewise.  Update call to can_vec_perm_p.
15776         (vectorizable_bswap): Likewise.
15778 2017-09-14  Richard Sandiford  <richard.sandiford@linaro.org>
15779             Alan Hayward  <alan.hayward@arm.com>
15780             David Sherwood  <david.sherwood@arm.com>
15782         * tree.h (build_vector): Take a vec<tree> instead of a tree *.
15783         * tree.c (build_vector): Likewise.
15784         (build_vector_from_ctor): Update accordingly.
15785         (build_vector_from_val): Likewise.
15786         * gimple-fold.c (gimple_fold_stmt_to_constant_1): Likewise.
15787         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
15788         * tree-vect-generic.c (add_rshift): Likewise.
15789         (expand_vector_divmod): Likewise.
15790         (optimize_vector_constructor): Likewise.
15791         * tree-vect-slp.c (vect_get_constant_vectors): Likewise.
15792         (vect_transform_slp_perm_load): Likewise.
15793         (vect_schedule_slp_instance): Likewise.
15794         * tree-vect-stmts.c (vectorizable_bswap): Likewise.
15795         (vectorizable_call): Likewise.
15796         (vect_gen_perm_mask_any): Likewise.  Add elements in order.
15797         * expmed.c (make_tree): Likewise.
15798         * fold-const.c (fold_negate_expr_1): Use auto_vec<tree> when building
15799         a vector passed to build_vector.
15800         (fold_convert_const): Likewise.
15801         (exact_inverse): Likewise.
15802         (fold_ternary_loc): Likewise.
15803         (fold_relational_const): Likewise.
15804         (const_binop): Likewise.  Use VECTOR_CST_ELT directly when operating
15805         on VECTOR_CSTs, rather than going through vec_cst_ctor_to_array.
15806         (const_unop): Likewise.  Store the reduction accumulator in a
15807         variable rather than an array.
15808         (vec_cst_ctor_to_array): Take the number of elements as a parameter.
15809         (fold_vec_perm): Update calls accordingly.  Use auto_vec<tree> for
15810         the new vector, rather than constructing it after the input arrays.
15811         (native_interpret_vector): Use auto_vec<tree> when building
15812         a vector passed to build_vector.  Add elements in order.
15813         * tree-vect-loop.c (get_initial_defs_for_reduction): Use
15814         auto_vec<tree> when building a vector passed to build_vector.
15815         (vect_create_epilog_for_reduction): Likewise.
15816         (vectorizable_induction): Likewise.
15817         (get_initial_def_for_reduction): Likewise.  Fix indentation of
15818         case statements.
15819         * config/sparc/sparc.c (sparc_handle_vis_mul8x16): Change n_elts
15820         to a vec<tree> *.
15821         (sparc_fold_builtin): Use auto_vec<tree> when building a vector
15822         passed to build_vector.
15824 2017-09-14  Richard Sandiford  <richard.sandiford@linaro.org>
15825             Alan Hayward  <alan.hayward@arm.com>
15826             David Sherwood  <david.sherwood@arm.com>
15828         * tree-core.h (tree_base::u): Add an "nelts" field.
15829         (tree_vector): Use VECTOR_CST_NELTS as the length.
15830         * tree.c (tree_size): Likewise.
15831         (make_vector): Initialize VECTOR_CST_NELTS.
15832         * tree.h (VECTOR_CST_NELTS): Use the u.nelts field.
15833         * cfgexpand.c (expand_debug_expr): Use VECTOR_CST_NELTS instead of
15834         TYPE_VECTOR_SUBPARTS.
15835         * expr.c (const_vector_mask_from_tree): Consistently use "units"
15836         as the number of units, setting it from VECTOR_CST_NELTS.
15837         (const_vector_from_tree): Likewise.
15838         * fold-const.c (negate_expr_p): Use VECTOR_CST_NELTS instead of
15839         TYPE_VECTOR_SUBPARTS for the number of elements in a VECTOR_CST.
15840         (fold_negate_expr_1): Likewise.
15841         (fold_convert_const): Likewise.
15842         (const_binop): Likewise.  Differentiate the number of output and
15843         input elements.
15844         (const_unop): Likewise.
15845         (fold_ternary_loc): Use VECTOR_CST_NELTS for the number of elements
15846         in a VECTOR_CST, asserting that it is the same as TYPE_VECTOR_SUBPARTS
15847         in cases that did the opposite.
15849 2017-09-14  Richard Biener  <rguenther@suse.de>
15851         * tree-ssa-sccvn.c (visit_phi): Merge undefined values similar
15852         to VN_TOP.
15854 2017-09-14  Eric Botcazou  <ebotcazou@adacore.com>
15856         * dwarf2out.c (dwarf2out_source_line): Remove superfluous test.
15858 2017-09-14  Jakub Jelinek  <jakub@redhat.com>
15860         PR target/81325
15861         * cfgbuild.c (find_bb_boundaries): Ignore debug insns in decisions
15862         if and where to split a bb, except for splitting before debug insn
15863         sequences followed by non-label real insn.  Delete debug insns
15864         in between basic blocks.
15866         * combine.c (make_compound_operation_int): Formatting fixes.
15868         * config/alpha/elf.h (LINK_EH_SPEC): Add -static-pie support.
15869         * config/alpha/linux.h (LINK_GCC_C_SEQUENCE_SPEC): Likewise.
15870         * config/netbsd.h (LINK_EH_SPEC): Likewise.
15871         * config/sol2.h (LINK_EH_SPEC): Likewise.
15872         * config/arm/uclinux-elf.h (LINK_GCC_C_SEQUENCE_SPEC): Likewise.
15873         * config/s390/linux.h (LINK_SPEC): Likewise.
15874         * config/freebsd.h (LINK_EH_SPEC): Likewise.
15875         * config/openbsd.h (LINK_EH_SPEC): Likewise.
15876         * config/lm32/uclinux-elf.h (LINK_GCC_C_SEQUENCE_SPEC): Likewise.
15877         * config/aarch64/aarch64-linux.h (LINUX_TARGET_LINK_SPEC): Likewise.
15878         * config/powerpcspe/sysv4.h (LINK_EH_SPEC): Likewise.
15879         * config/bfin/linux.h (LINK_GCC_C_SEQUENCE_SPEC): Likewise.
15880         * config/i386/gnu-user64.h (GNU_USER_TARGET_LINK_SPEC): Fix a typo.
15881         * config/i386/gnu-user.h (GNU_USER_TARGET_LINK_SPEC): Formatting fix.
15883 2017-09-13  Jakub Jelinek  <jakub@redhat.com>
15885         * config/rs6000/sysv4.h (STARTFILE_LINUX_SPEC): Add -static-pie
15886         support.
15887         (ENDFILE_LINUX_SPEC): Likewise.
15888         (LINK_EH_SPEC): Likewise.
15889         * config/rs6000/linux64.h (LINK_SHLIB_SPEC): Likewise.
15890         (LINK_OS_LINUX_SPEC32): Likewise.
15891         (LINK_OS_LINUX_SPEC64): Likewise.
15892         * config/rs6000/linux.h (LINK_SHLIB_SPEC): Likewise.
15893         (LINK_OS_LINUX_SPEC): Likewise.
15895 2017-09-13  Martin Liska  <mliska@suse.cz>
15897         PR middle-end/82154
15898         * stmt.c (expand_sjlj_dispatch_table): Use CASE_LOW when
15899         CASE_HIGH is NULL_TREE.
15901 2017-09-13  Richard Sandiford  <richard.sandiford@linaro.org>
15902             Alan Hayward  <alan.hayward@arm.com>
15903             David Sherwood  <david.sherwood@arm.com>
15905         * target.def (secondary_memory_needed): New hook.
15906         (secondary_reload): Refer to TARGET_SECONDARY_MEMORY_NEEDED
15907         instead of SECONDARY_MEMORY_NEEDED.
15908         (secondary_memory_needed_mode): Likewise.
15909         * hooks.h (hook_bool_mode_reg_class_t_reg_class_t_false): Declare.
15910         * hooks.c (hook_bool_mode_reg_class_t_reg_class_t_false): New function.
15911         * doc/tm.texi.in (SECONDARY_MEMORY_NEEDED): Replace with...
15912         (TARGET_SECONDARY_MEMORY_NEEDED): ...this.
15913         (SECONDARY_MEMORY_NEEDED_RTX): Update reference accordingly.
15914         * doc/tm.texi: Regenerate.
15915         * config/alpha/alpha.h (SECONDARY_MEMORY_NEEDED): Delete.
15916         * config/alpha/alpha.c (alpha_secondary_memory_needed): New function.
15917         (TARGET_SECONDARY_MEMORY_NEEDED): Redefine.
15918         * config/i386/i386.h (SECONDARY_MEMORY_NEEDED): Delete.
15919         * config/i386/i386-protos.h (ix86_secondary_memory_needed): Delete.
15920         * config/i386/i386.c (inline_secondary_memory_needed): Put the
15921         mode argument first and change the reg_class arguments to reg_class_t.
15922         (ix86_secondary_memory_needed): Likewise.  Remove the strict parameter.
15923         Make static.  Update the call to inline_secondary_memory_needed.
15924         (ix86_register_move_cost): Update the call to
15925         inline_secondary_memory_needed.
15926         (TARGET_SECONDARY_MEMORY_NEEDED): Redefine.
15927         * config/ia64/ia64.h (SECONDARY_MEMORY_NEEDED): Delete commented-out
15928         definition.
15929         * config/ia64/ia64.c (spill_xfmode_rfmode_operand): Refer to
15930         TARGET_SECONDARY_MEMORY_NEEDED rather than SECONDARY_MEMORY_NEEDED
15931         in comment.
15932         * config/mips/mips.h (SECONDARY_MEMORY_NEEDED): Delete.
15933         * config/mips/mips-protos.h (mips_secondary_memory_needed): Delete.
15934         * config/mips/mips.c (mips_secondary_memory_needed): Make static
15935         and match hook interface.  Add comment from mips.h.
15936         (TARGET_SECONDARY_MEMORY_NEEDED): Redefine.
15937         * config/mmix/mmix.md (truncdfsf2): Refer to
15938         TARGET_SECONDARY_MEMORY_NEEDED rather than SECONDARY_MEMORY_NEEDED
15939         in comment.
15940         * config/pa/pa-64.h (SECONDARY_MEMORY_NEEDED): Rename to...
15941         (PA_SECONDARY_MEMORY_NEEDED): ...this, and put the mode argument first.
15942         * config/pa/pa.c (TARGET_SECONDARY_MEMORY_NEEDED): Redefine.
15943         (pa_secondary_memory_needed): New function.
15944         * config/pdp11/pdp11.h (SECONDARY_MEMORY_NEEDED): Delete.
15945         * config/pdp11/pdp11-protos.h (pdp11_secondary_memory_needed): Delete.
15946         * config/pdp11/pdp11.c (TARGET_SECONDARY_MEMORY_NEEDED): Redefine.
15947         (pdp11_secondary_memory_needed): Make static and match hook interface.
15948         * config/powerpcspe/powerpcspe.h (SECONDARY_MEMORY_NEEDED): Delete.
15949         * config/powerpcspe/powerpcspe-protos.h
15950         (rs6000_secondary_memory_needed_ptr): Delete.
15951         * config/powerpcspe/powerpcspe.c (rs6000_secondary_memory_needed_ptr):
15952         Delete.
15953         (TARGET_SECONDARY_MEMORY_NEEDED): Redefine.
15954         (rs6000_option_override_internal): Assign to
15955         targetm.secondary_memory_needed rather than
15956         rs6000_secondary_memory_needed_ptr.
15957         (rs6000_secondary_memory_needed): Match hook interface.
15958         (rs6000_debug_secondary_memory_needed): Likewise.
15959         * config/riscv/riscv.h (SECONDARY_MEMORY_NEEDED): Delete.
15960         * config/riscv/riscv.c (riscv_secondary_memory_needed): New function.
15961         (riscv_register_move_cost): Use it instead of SECONDARY_MEMORY_NEEDED.
15962         (TARGET_SECONDARY_MEMORY_NEEDED): Redefine.
15963         * config/rs6000/rs6000.h (SECONDARY_MEMORY_NEEDED): Delete.
15964         * config/rs6000/rs6000-protos.h (rs6000_secondary_memory_needed_ptr):
15965         Delete.
15966         * config/rs6000/rs6000.c (rs6000_secondary_memory_needed_ptr): Delete.
15967         (TARGET_SECONDARY_MEMORY_NEEDED): Redefine.
15968         (rs6000_option_override_internal): Assign to
15969         targetm.secondary_memory_needed rather than
15970         rs6000_secondary_memory_needed_ptr.
15971         (rs6000_secondary_memory_needed): Match hook interface.
15972         (rs6000_debug_secondary_memory_needed): Likewise.
15973         * config/s390/s390.h (SECONDARY_MEMORY_NEEDED): Delete.
15974         * config/s390/s390.c (s390_secondary_memory_needed): New function.
15975         (TARGET_SECONDARY_MEMORY_NEEDED): Redefine.
15976         * config/sparc/sparc.h (SECONDARY_MEMORY_NEEDED): Delete.
15977         * config/sparc/sparc.c (TARGET_SECONDARY_MEMORY_NEEDED): Redefine.
15978         (sparc_secondary_memory_needed): New function.
15979         * lra-constraints.c (check_and_process_move): Refer to
15980         TARGET_SECONDARY_MEMORY_NEEDED rather than SECONDARY_MEMORY_NEEDED
15981         in comment.
15982         (curr_insn_transform): Likewise.
15983         (process_alt_operands): Use targetm.secondary_memory_needed
15984         instead of TARGET_SECONDARY_MEMORY_NEEDED.
15985         (check_secondary_memory_needed_p): Likewise.
15986         (choose_split_class): Likewise.
15987         * reload.c: Unconditionally include code that was previously
15988         conditional on SECONDARY_MEMORY_NEEDED.
15989         (push_secondary_reload): Use targetm.secondary_memory_needed
15990         instead of TARGET_SECONDARY_MEMORY_NEEDED.
15991         (push_reload): Likewise.
15992         * reload1.c: Unconditionally include code that was previously
15993         conditional on SECONDARY_MEMORY_NEEDED.
15994         (choose_reload_regs): Use targetm.secondary_memory_needed
15995         instead of TARGET_SECONDARY_MEMORY_NEEDED.
15996         (gen_reload): Likewise.
15997         * system.h (SECONDARY_MEMORY_NEEDED): Poison.
15999 2017-09-13  Richard Sandiford  <richard.sandiford@linaro.org>
16000             Alan Hayward  <alan.hayward@arm.com>
16001             David Sherwood  <david.sherwood@arm.com>
16003         * target.def (secondary_memory_needed_mode): New hook:
16004         * targhooks.c (default_secondary_memory_needed_mode): Declare.
16005         * targhooks.h (default_secondary_memory_needed_mode): New function.
16006         * doc/tm.texi.in (SECONDARY_MEMORY_NEEDED_MODE): Replace with...
16007         (TARGET_SECONDARY_MEMORY_NEEDED_MODE): ...this.
16008         * doc/tm.texi: Regenerate.
16009         * lra-constraints.c (check_and_process_move): Use
16010         targetm.secondary_memory_needed_mode instead of
16011         TARGET_SECONDARY_MEMORY_NEEDED_MODE.
16012         (curr_insn_transform): Likewise.
16013         * reload.c (get_secondary_mem): Likewise.
16014         * config/alpha/alpha.h (SECONDARY_MEMORY_NEEDED_MODE): Delete.
16015         * config/alpha/alpha.c (alpha_secondary_memory_needed_mode): New
16016         function.
16017         (TARGET_SECONDARY_MEMORY_NEEDED_MODE): Redefine.
16018         * config/i386/i386.h (SECONDARY_MEMORY_NEEDED_MODE): Delete.
16019         * config/i386/i386.c (ix86_secondary_memory_needed_mode): New function.
16020         (TARGET_SECONDARY_MEMORY_NEEDED_MODE): Redefine.
16021         * config/powerpcspe/powerpcspe.h (SECONDARY_MEMORY_NEEDED_MODE):
16022         Delete.
16023         * config/powerpcspe/powerpcspe-protos.h
16024         (rs6000_secondary_memory_needed_mode): Delete.
16025         * config/powerpcspe/powerpcspe.c
16026         (TARGET_SECONDARY_MEMORY_NEEDED_MODE): Redefine.
16027         (rs6000_secondary_memory_needed_mode): Make static.
16028         * config/rs6000/rs6000.h (SECONDARY_MEMORY_NEEDED_MODE): Delete.
16029         * config/rs6000/rs6000-protos.h (rs6000_secondary_memory_needed_mode):
16030         Delete.
16031         * config/rs6000/rs6000.c (TARGET_SECONDARY_MEMORY_NEEDED_MODE):
16032         Redefine.
16033         (rs6000_secondary_memory_needed_mode): Make static.
16034         * config/s390/s390.h (SECONDARY_MEMORY_NEEDED_MODE): Delete.
16035         * config/s390/s390.c (s390_secondary_memory_needed_mode): New function.
16036         (TARGET_SECONDARY_MEMORY_NEEDED_MODE): Redefine.
16037         * config/sparc/sparc.h (SECONDARY_MEMORY_NEEDED_MODE): Delete.
16038         * config/sparc/sparc.c (TARGET_SECONDARY_MEMORY_NEEDED_MODE):
16039         Redefine.
16040         (sparc_secondary_memory_needed_mode): New function.
16041         * system.h (TARGET_SECONDARY_MEMORY_NEEDED_MODE): Poison.
16043 2017-09-13  Jackson Woodruff  <jackson.woodruff@arm.com>
16045         * config/aarch64/constraints.md (Umq): New constraint.
16046         * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>):
16047         Change to use Umq.
16048         (mov<mode>): Update condition.
16050 2017-09-13  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
16052         * gimple-ssa-store-merging.c (sort_by_bitpos): Compare store order
16053         when bitposition is the same.
16055 2017-09-13  Richard Biener  <rguenther@suse.de>
16057         * dwarf2out.c (output_die_symbol): Remove.
16058         (output_die): Do not output a DIEs symbol.
16060 2017-09-13  Richard Biener  <rguenther@suse.de>
16062         PR middle-end/82128
16063         * gimple-fold.c (gimple_fold_call): Update SSA name in-place to
16064         default-def to avoid breaking iterator update with the weird
16065         interaction with cgraph_update_edges_for_call_stmt_node.
16067 2017-09-13  Richard Biener  <rguenther@suse.de>
16069         * tree-cfg.c (verify_gimple_assign_binary): Add verification
16070         for WIDEN_SUM_EXPR, VEC_WIDEN_MULT_{HI,LO,EVEN,ODD}_EXPR,
16071         VEC_PACK_{TRUNC,SAT,FIX_TRUNC}_EXPR.
16072         (verify_gimple_assign_ternary): Add verification for DOT_PROD_EXPR.
16074 2017-09-13  Kugan Vivekanandarajah  <kuganv@linaro.org>
16076         * config/aarch64/aarch64.c (aarch64_override_options_after_change_1):
16077         Disable pc relative literal load irrespective of
16078         TARGET_FIX_ERR_A53_84341 for default.
16080 2017-09-12  Eric Botcazou  <ebotcazou@adacore.com>
16082         * config/sparc/sparc.c (output_return): Output the source location of
16083         the insn in the delay slot, if any.
16084         (output_sibcall): Likewise.
16086 2017-09-12  H.J. Lu  <hongjiu.lu@intel.com>
16088         PR driver/81498
16089         * common.opt (-static-pie): New alias.
16090         (shared): Negate static-pie.
16091         (-no-pie): Update help text.
16092         (-pie): Likewise.
16093         (static-pie): New option.
16094         * config/gnu-user.h (GNU_USER_TARGET_STARTFILE_SPEC): Add
16095         -static-pie support.
16096         (GNU_USER_TARGET_ENDFILE_SPEC): Likewise.
16097         (LINK_EH_SPEC): Likewise.
16098         (LINK_GCC_C_SEQUENCE_SPEC): Likewise.
16099         * config/i386/gnu-user.h (GNU_USER_TARGET_LINK_SPEC): Likewise.
16100         * config/i386/gnu-user64.h (GNU_USER_TARGET_LINK_SPEC): Likewise.
16101         * gcc.c (LINK_COMMAND_SPEC): Likewise.
16102         (init_gcc_specs): Likewise.
16103         (init_spec): Likewise.
16104         (display_help): Update help message for -pie.
16105         * doc/invoke.texi: Update -pie, -no-pie and -static.  Document
16106         -static-pie.
16108 2017-09-12  Wilco Dijkstra  <wdijkstr@arm.com>
16110         * config/aarch64/aarch64.md (movsi_aarch64): Remove all '*'.
16111         (movdi_aarch64): Likewise.
16112         (movti_aarch64): Likewise.
16114 2017-09-12 Simon Wright <simon@pushface.org>
16116         PR target/80204
16117         * config/darwin-driver.c (darwin_find_version_from_kernel): Eliminate
16118         calculation of the minor version, always output as 0.
16120 2017-09-12  Jakub Jelinek  <jakub@redhat.com>
16122         PR target/82112
16123         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): For
16124         ALTIVEC_BUILTIN_VEC_LD if arg1 has array type call default_conversion
16125         on it early, rather than manual conversion late.  For
16126         ALTIVEC_BUILTIN_VEC_ST if arg2 has array type call default_conversion
16127         instead of performing manual conversion.
16129 2017-09-12  Carl Love  <cel@us.ibm.com>
16131         * config/rs6000/altivec.md (vec_widen_umult_even_v4si,
16132         vec_widen_smult_even_v4si): Add define expands for vmuleuw, vmulesw,
16133         vmulouw, vmulosw.
16134         * config/rs6000/rs6000-builtin.def (VMLEUW, VMULESW, VMULOUW,
16135         VMULOSW): Add definitions.
16136         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
16137         ALTIVEC_BUILTIN_VMULESW, ALTIVEC_BUILTIN_VMULEUW,
16138         ALTIVEC_BUILTIN_VMULOSW, ALTIVEC_BUILTIN_VMULOUW entries.
16139         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin,
16140         builtin_function_type): Add ALTIVEC_BUILTIN_* case statements.
16142 2017-09-12  James Greenhalgh  <james.greenhalgh@arm.com>
16144         * config/aarch64/aarch64.md (movdi_aarch64): Set load/store
16145         types correctly.
16146         (movti_aarch64): Likewise.
16147         (movdf_aarch64): Likewise.
16148         (movtf_aarch64): Likewise.
16149         (load_pairdi): Likewise.
16150         (store_pairdi): Likewise.
16151         (load_pairdf): Likewise.
16152         (store_pairdf): Likewise.
16153         (loadwb_pair<GPI:mode>_<P:mode>): Likewise.
16154         (storewb_pair<GPI:mode>_<P:mode>): Likewise.
16155         (ldr_got_small_<mode>): Likewise.
16156         (ldr_got_small_28k_<mode>): Likewise.
16157         (ldr_got_tiny): Likewise.
16158         * config/aarch64/iterators.md (ldst_sz): New.
16159         (ldpstp_sz): Likewise.
16160         * config/aarch64/thunderx.md (thunderx_storepair): Split store_8
16161         to store_16.
16162         (thunderx_load): Split load_8 to load_16.
16163         * config/aarch64/thunderx2t99.md (thunderx2t99_loadpair): Split
16164         load_8 to load_16.
16165         (thunderx2t99_storepair_basic): Split store_8 to store_16.
16166         * config/arm/xgene1.md (xgene1_load_pair): Split load_8 to load_16.
16167         (xgene1_store_pair): Split store_8 to store_16.
16168         * config/aarch64/falkor.md (falkor_ld_3_ld): Split load_8 to load_16.
16169         (falkor_st_0_st_sd): Split store_8 to store_16.
16171 2017-09-12  James Greenhalgh  <james.greenhalgh@arm.com>
16173         * config/arm/types.md (type): Rename load1/2/3/4 to load_4/8/12/16
16174         and store1/2/3/4 to store_4/8/12/16.
16175         * config/aarch64/aarch64.md: Update for rename.
16176         * config/arm/arm.md: Likewise.: Likewise.
16177         * config/arm/arm.c: Likewise.
16178         * config/arm/thumb1.md: Likewise.
16179         * config/arm/thumb2.md: Likewise.
16180         * config/arm/vfp.md: Likewise.
16181         * config/arm/arm-generic.md: Likewise.
16182         * config/arm/arm1020e.md: Likewise.
16183         * config/arm/arm1026ejs.md: Likewise.
16184         * config/arm/arm1136jfs.md: Likewise.
16185         * config/arm/arm926ejs.md: Likewise.
16186         * config/arm/cortex-a15.md: Likewise.
16187         * config/arm/cortex-a17.md: Likewise.
16188         * config/arm/cortex-a5.md: Likewise.
16189         * config/arm/cortex-a53.md: Likewise.
16190         * config/arm/cortex-a57.md: Likewise.
16191         * config/arm/cortex-a7.md: Likewise.
16192         * config/arm/cortex-a8.md: Likewise.
16193         * config/arm/cortex-a9.md: Likewise.
16194         * config/arm/cortex-m4.md: Likewise.
16195         * config/arm/cortex-m7.md: Likewise.
16196         * config/arm/cortex-r4.md: Likewise.
16197         * config/arm/exynos-m1.md: Likewise.
16198         * config/arm/fa526.md: Likewise.
16199         * config/arm/fa606te.md: Likewise.
16200         * config/arm/fa626te.md: Likewise.
16201         * config/arm/fa726te.md: Likewise.
16202         * config/arm/fmp626.md: Likewise.
16203         * config/arm/iwmmxt.md: Likewise.
16204         * config/arm/ldmstm.md: Likewise.
16205         * config/arm/marvell-pj4.md: Likewise.
16206         * config/arm/xgene1.md: Likewise.
16207         * config/aarch64/thunderx.md: Likewise.
16208         * config/aarch64/thunderx2t99.md: Likewise.
16209         * config/aarch64/falkor.md: Likewise.
16211 2017-09-12  Martin Liska  <mliska@suse.cz>
16213         * attribs.c (private_lookup_attribute): New function.
16214         * attribs.h (private_lookup_attribute): Declared here.
16215         (lookup_attribute): Called from this place.
16217 2017-09-12  Richard Biener  <rguenther@suse.de>
16219         PR tree-optimization/82157
16220         * tree-ssa-pre.c (remove_dead_inserted_code): Do not remove
16221         stmts with side-effects.
16223 2017-09-12  Richard Sandiford  <richard.sandiford@linaro.org>
16224             Alan Hayward  <alan.hayward@arm.com>
16225             David Sherwood <david.sherwood@arm.com>
16227         * target.def (hard_regno_nregs): New hook.
16228         (class_max_nregs): Refer to it instead of HARD_REGNO_NREGS.
16229         * targhooks.h (default_hard_regno_nregs): Declare.
16230         * targhooks.c (default_hard_regno_nregs): New function.
16231         * doc/tm.texi.in (HARD_REGNO_NREGS): Replace with...
16232         (TARGET_HARD_REGNO_NREGS): ...this hook.
16233         (HARD_REGNO_NREGS_HAS_PADDING): Update accordingly.
16234         (CLASS_MAX_NREGS): Likewise.
16235         * doc/tm.texi: Regenerate.
16236         * reginfo.c (init_reg_modes_target): Use targetm.hard_regno_nregs
16237         instead of HARD_REGNO_NREGS.
16238         * rtl.h (REG_NREGS): Refer to TARGET_HARD_REGNO_NREGS rather than
16239         HARD_REGNO_NREGS in the comment.
16240         * config/aarch64/aarch64.h (HARD_REGNO_NREGS): Delete.
16241         * config/aarch64/aarch64-protos.h (aarch64_hard_regno_nregs): Delete.
16242         * config/aarch64/aarch64.c (aarch64_hard_regno_nregs): Make static.
16243         Return an unsigned int.
16244         (TARGET_HARD_REGNO_NREGS): Redefine.
16245         * config/alpha/alpha.h (HARD_REGNO_NREGS): Delete.
16246         * config/arc/arc.h (HARD_REGNO_NREGS): Delete.
16247         * config/arc/arc.c (TARGET_HARD_REGNO_NREGS): Redefine.
16248         (arc_hard_regno_nregs): New function.
16249         * config/arm/arm.h (HARD_REGNO_NREGS): Delete.
16250         * config/arm/arm.c (TARGET_HARD_REGNO_NREGS): Redefine.
16251         (arm_hard_regno_nregs): New function.
16252         * config/avr/avr.h (HARD_REGNO_NREGS): Delete.
16253         * config/bfin/bfin.h (HARD_REGNO_NREGS): Delete.
16254         * config/bfin/bfin.c (bfin_hard_regno_nregs): New function.
16255         (TARGET_HARD_REGNO_NREGS): Redefine.
16256         * config/c6x/c6x.h (HARD_REGNO_NREGS): Delete.
16257         * config/cr16/cr16.h (LONG_REG_P): Use targetm.hard_regno_nregs.
16258         (HARD_REGNO_NREGS): Delete.
16259         * config/cr16/cr16.c (TARGET_HARD_REGNO_NREGS): Redefine.
16260         (cr16_hard_regno_nregs): New function.
16261         (cr16_memory_move_cost): Use it instead of HARD_REGNO_NREGS.
16262         * config/cris/cris.h (HARD_REGNO_NREGS): Delete.
16263         * config/cris/cris.c (TARGET_HARD_REGNO_NREGS): Redefine.
16264         (cris_hard_regno_nregs): New function.
16265         * config/epiphany/epiphany.h (HARD_REGNO_NREGS): Delete.
16266         * config/fr30/fr30.h (HARD_REGNO_NREGS): Delete.
16267         (CLASS_MAX_NREGS): Use targetm.hard_regno_nregs.
16268         * config/frv/frv.h (HARD_REGNO_NREGS): Delete.
16269         (CLASS_MAX_NREGS): Remove outdated copy of documentation.
16270         * config/frv/frv-protos.h (frv_hard_regno_nregs): Delete.
16271         * config/frv/frv.c (TARGET_HARD_REGNO_NREGS): Redefine.
16272         (frv_hard_regno_nregs): Make static.  Take and return an
16273         unsigned int.
16274         (frv_class_max_nregs): Remove outdated copy of documentation.
16275         * config/ft32/ft32.h (HARD_REGNO_NREGS): Delete.
16276         * config/h8300/h8300.h (HARD_REGNO_NREGS): Delete.
16277         * config/h8300/h8300-protos.h (h8300_hard_regno_nregs): Delete.
16278         * config/h8300/h8300.c (h8300_hard_regno_nregs): Delete.
16279         * config/i386/i386.h (HARD_REGNO_NREGS): Delete.
16280         * config/i386/i386.c (ix86_hard_regno_nregs): New function.
16281         (TARGET_HARD_REGNO_NREGS): Redefine.
16282         * config/ia64/ia64.h (HARD_REGNO_NREGS): Delete.
16283         (CLASS_MAX_NREGS): Update comment.
16284         * config/ia64/ia64.c (TARGET_HARD_REGNO_NREGS): Redefine.
16285         (ia64_hard_regno_nregs): New function.
16286         * config/iq2000/iq2000.h (HARD_REGNO_NREGS): Delete.
16287         * config/lm32/lm32.h (HARD_REGNO_NREGS): Delete.
16288         * config/m32c/m32c.h (HARD_REGNO_NREGS): Delete.
16289         * config/m32c/m32c-protos.h (m32c_hard_regno_nregs): Delete.
16290         * config/m32c/m32c.c (m32c_hard_regno_nregs_1): Take and return
16291         an unsigned int.
16292         (m32c_hard_regno_nregs): Likewise.  Make static.
16293         (TARGET_HARD_REGNO_NREGS): Redefine.
16294         * config/m32r/m32r.h (HARD_REGNO_NREGS): Delete.
16295         * config/m68k/m68k.h (HARD_REGNO_NREGS): Delete.
16296         * config/m68k/m68k.c (TARGET_HARD_REGNO_NREGS): Redefine.
16297         (m68k_hard_regno_nregs): New function.
16298         * config/mcore/mcore.h (HARD_REGNO_NREGS): Delete.
16299         * config/microblaze/microblaze.h (HARD_REGNO_NREGS): Delete.
16300         * config/mips/mips.h (HARD_REGNO_NREGS): Delete.
16301         * config/mips/mips-protos.h (mips_hard_regno_nregs): Delete.
16302         * config/mips/mips.c (mips_hard_regno_nregs): Make static.
16303         Take and return an unsigned int.
16304         (TARGET_HARD_REGNO_NREGS): Redefine.
16305         * config/mmix/mmix.h (HARD_REGNO_NREGS): Delete.
16306         (CLASS_MAX_NREGS): Use targetm.hard_regno_nregs.
16307         * config/mn10300/mn10300.h (HARD_REGNO_NREGS): Delete.
16308         * config/moxie/moxie.h (HARD_REGNO_NREGS): Delete.
16309         * config/msp430/msp430.h (HARD_REGNO_NREGS): Delete.
16310         * config/msp430/msp430-protos.h (msp430_hard_regno_nregs): Delete.
16311         * config/msp430/msp430.c (TARGET_HARD_REGNO_NREGS): Redefine.
16312         (msp430_hard_regno_nregs): Make static.  Take and return an
16313         unsigned int.
16314         * config/nds32/nds32.h (HARD_REGNO_NREGS): Delete.
16315         * config/nds32/nds32-protos.h (nds32_hard_regno_nregs): Delete.
16316         * config/nds32/nds32.c (nds32_hard_regno_nregs): Delete.
16317         (nds32_hard_regno_mode_ok): Use targetm.hard_regno_nregs.
16318         * config/nios2/nios2.h (HARD_REGNO_NREGS): Delete.
16319         * config/nvptx/nvptx.h (HARD_REGNO_NREGS): Delete.
16320         * config/nvptx/nvptx.c (nvptx_hard_regno_nregs): New function.
16321         (TARGET_HARD_REGNO_NREGS): Redefine.
16322         * config/pa/pa32-regs.h (HARD_REGNO_NREGS): Rename to...
16323         (PA_HARD_REGNO_NREGS): ...this.
16324         * config/pa/pa64-regs.h (HARD_REGNO_NREGS): Rename to...
16325         (PA_HARD_REGNO_NREGS): ...this.
16326         * config/pa/pa.c (TARGET_HARD_REGNO_NREGS): Redefine.
16327         (pa_hard_regno_nregs): New function.
16328         * config/pdp11/pdp11.h (HARD_REGNO_NREGS): Delete.
16329         * config/pdp11/pdp11.c (TARGET_HARD_REGNO_NREGS): Redefine.
16330         (pdp11_hard_regno_nregs): New function.
16331         * config/powerpcspe/powerpcspe.h (HARD_REGNO_NREGS): Delete.
16332         * config/powerpcspe/powerpcspe.c (TARGET_HARD_REGNO_NREGS): Redefine.
16333         (rs6000_hard_regno_nregs_hook): New function.
16334         * config/riscv/riscv.h (HARD_REGNO_NREGS): Delete.
16335         * config/riscv/riscv-protos.h (riscv_hard_regno_nregs): Delete.
16336         * config/riscv/riscv.c (riscv_hard_regno_nregs): Make static.
16337         Take and return an unsigned int.  Move earlier in file.
16338         (TARGET_HARD_REGNO_NREGS): Redefine.
16339         * config/rl78/rl78.h (HARD_REGNO_NREGS): Delete.
16340         * config/rl78/rl78-protos.h (rl78_hard_regno_nregs): Delete.
16341         * config/rl78/rl78.c (TARGET_HARD_REGNO_NREGS): Reefine.
16342         (rl78_hard_regno_nregs): Make static.  Take and return an
16343         unsigned int.
16344         * config/rs6000/rs6000.h (HARD_REGNO_NREGS): Delete.
16345         * config/rs6000/rs6000.c (TARGET_HARD_REGNO_NREGS): Redefine.
16346         (rs6000_hard_regno_nregs_hook): New function.
16347         * config/rx/rx.h (HARD_REGNO_NREGS): Delete.
16348         * config/rx/rx.c (rx_hard_regno_nregs): New function.
16349         (TARGET_HARD_REGNO_NREGS): Redefine.
16350         * config/s390/s390.h (HARD_REGNO_NREGS): Delete.
16351         * config/s390/s390.c (REGNO_PAIR_OK): Use s390_hard_regno_nregs
16352         instead of HARD_REGNO_NREGS.
16353         (s390_hard_regno_nregs): New function.
16354         (s390_hard_regno_mode_ok): Add comment from s390.h.
16355         (TARGET_HARD_REGNO_NREGS): Redefine.
16356         * config/sh/sh.h (HARD_REGNO_NREGS): Delete.
16357         * config/sh/sh.c (TARGET_HARD_REGNO_NREGS): Redefine.
16358         (sh_hard_regno_nregs): New function.
16359         (sh_pass_in_reg_p): Use it.
16360         * config/sparc/sparc.h (HARD_REGNO_NREGS): Delete.
16361         * config/sparc/sparc.c (TARGET_HARD_REGNO_NREGS): Redefine.
16362         (sparc_hard_regno_nregs): New function.
16363         * config/spu/spu.h (HARD_REGNO_NREGS): Delete.
16364         * config/spu/spu.c (spu_hard_regno_nregs): New function.
16365         (spu_function_arg_advance): Use it, supplying a valid register number.
16366         (TARGET_HARD_REGNO_NREGS): Redefine.
16367         * config/stormy16/stormy16.h (HARD_REGNO_NREGS): Delete.
16368         * config/tilegx/tilegx.h (HARD_REGNO_NREGS): Delete.
16369         * config/tilepro/tilepro.h (HARD_REGNO_NREGS): Delete.
16370         * config/v850/v850.h (HARD_REGNO_NREGS): Delete.
16371         * config/vax/vax.h (HARD_REGNO_NREGS): Delete.
16372         * config/visium/visium.h (HARD_REGNO_NREGS): Delete.
16373         (CLASS_MAX_NREGS): Remove copy of old documentation.
16374         * config/visium/visium.c (TARGET_HARD_REGNO_NREGS): Redefine.
16375         (visium_hard_regno_nregs): New function.
16376         (visium_hard_regno_mode_ok): Use it instead of HARD_REGNO_NREGS.
16377         * config/xtensa/xtensa.h (HARD_REGNO_NREGS): Delete.
16378         * config/xtensa/xtensa.c (TARGET_HARD_REGNO_NREGS): Redefine.
16379         xtensa_hard_regno_nregs): New function.
16380         * system.h (HARD_REGNO_NREGS): Poison.
16382 2017-09-12  Richard Sandiford  <richard.sandiford@linaro.org>
16384         * config/arm/arm.h (THUMB_SECONDARY_INPUT_RELOAD_CLASS): Use
16385         hard_regno_nregs instead of HARD_REGNO_NREGS.
16386         (THUMB_SECONDARY_OUTPUT_RELOAD_CLASS): Likewise.
16387         * config/c6x/c6x.c (c6x_expand_prologue): Likewise.
16388         (c6x_expand_epilogue): Likewise.
16389         * config/frv/frv.c (frv_alloc_temp_reg): Likewise.
16390         (frv_read_iacc_argument): Likewise.
16391         * config/sh/sh.c: Include regs.h.
16392         (sh_print_operand): Use hard_regno_nregs instead of HARD_REGNO_NREGS.
16393         (regs_used): Likewise.
16394         (output_stack_adjust): Likewise.
16395         * config/xtensa/xtensa.c (xtensa_copy_incoming_a7): Likewise.
16396         * expmed.c: Include regs.h.
16397         (store_bit_field_1): Use hard_regno_nregs instead of HARD_REGNO_NREGS.
16398         * ree.c: Include regs.h.
16399         (combine_reaching_defs): Use hard_regno_nregs instead of
16400         HARD_REGNO_NREGS.
16401         (add_removable_extension): Likewise.
16403 2017-09-12  Richard Sandiford  <richard.sandiford@linaro.org>
16405         * regs.h (hard_regno_nregs): Turn into a function.
16406         (end_hard_regno): Update accordingly.
16407         * caller-save.c (setup_save_areas): Likewise.
16408         (save_call_clobbered_regs): Likewise.
16409         (replace_reg_with_saved_mem): Likewise.
16410         (insert_restore): Likewise.
16411         (insert_save): Likewise.
16412         * combine.c (can_change_dest_mode): Likewise.
16413         (move_deaths): Likewise.
16414         (distribute_notes): Likewise.
16415         * config/mips/mips.c (mips_hard_regno_call_part_clobbered): Likewise.
16416         * config/powerpcspe/powerpcspe.c (rs6000_cannot_change_mode_class)
16417         (rs6000_split_multireg_move): Likewise.
16418         (rs6000_register_move_cost): Likewise.
16419         (rs6000_memory_move_cost): Likewise.
16420         * config/rs6000/rs6000.c (rs6000_cannot_change_mode_class): Likewise.
16421         (rs6000_split_multireg_move): Likewise.
16422         (rs6000_register_move_cost): Likewise.
16423         (rs6000_memory_move_cost): Likewise.
16424         * cselib.c (cselib_reset_table): Likewise.
16425         (cselib_lookup_1): Likewise.
16426         * emit-rtl.c (set_mode_and_regno): Likewise.
16427         * function.c (aggregate_value_p): Likewise.
16428         * ira-color.c (setup_profitable_hard_regs): Likewise.
16429         (check_hard_reg_p): Likewise.
16430         (calculate_saved_nregs): Likewise.
16431         (assign_hard_reg): Likewise.
16432         (improve_allocation): Likewise.
16433         (calculate_spill_cost): Likewise.
16434         * ira-emit.c (modify_move_list): Likewise.
16435         * ira-int.h (ira_hard_reg_set_intersection_p): Likewise.
16436         (ira_hard_reg_in_set_p): Likewise.
16437         * ira.c (setup_reg_mode_hard_regset): Likewise.
16438         (clarify_prohibited_class_mode_regs): Likewise.
16439         (check_allocation): Likewise.
16440         * lra-assigns.c (find_hard_regno_for_1): Likewise.
16441         (lra_setup_reg_renumber): Likewise.
16442         (setup_try_hard_regno_pseudos): Likewise.
16443         (spill_for): Likewise.
16444         (assign_hard_regno): Likewise.
16445         (setup_live_pseudos_and_spill_after_risky_transforms): Likewise.
16446         * lra-constraints.c (in_class_p): Likewise.
16447         (lra_constraint_offset): Likewise.
16448         (simplify_operand_subreg): Likewise.
16449         (lra_constraints): Likewise.
16450         (split_reg): Likewise.
16451         (split_if_necessary): Likewise.
16452         (invariant_p): Likewise.
16453         (inherit_in_ebb): Likewise.
16454         * lra-lives.c (process_bb_lives): Likewise.
16455         * lra-remat.c (reg_overlap_for_remat_p): Likewise.
16456         (get_hard_regs): Likewise.
16457         (do_remat): Likewise.
16458         * lra-spills.c (assign_spill_hard_regs): Likewise.
16459         * mode-switching.c (create_pre_exit): Likewise.
16460         * postreload.c (reload_combine_recognize_pattern): Likewise.
16461         * recog.c (peep2_find_free_register): Likewise.
16462         * regcprop.c (kill_value_regno): Likewise.
16463         (set_value_regno): Likewise.
16464         (copy_value): Likewise.
16465         (maybe_mode_change): Likewise.
16466         (find_oldest_value_reg): Likewise.
16467         (copyprop_hardreg_forward_1): Likewise.
16468         * regrename.c (check_new_reg_p): Likewise.
16469         (regrename_do_replace): Likewise.
16470         * reload.c (push_reload): Likewise.
16471         (combine_reloads): Likewise.
16472         (find_dummy_reload): Likewise.
16473         (operands_match_p): Likewise.
16474         (find_reloads): Likewise.
16475         (find_equiv_reg): Likewise.
16476         (reload_adjust_reg_for_mode): Likewise.
16477         * reload1.c (count_pseudo): Likewise.
16478         (count_spilled_pseudo): Likewise.
16479         (find_reg): Likewise.
16480         (clear_reload_reg_in_use): Likewise.
16481         (free_for_value_p): Likewise.
16482         (allocate_reload_reg): Likewise.
16483         (choose_reload_regs): Likewise.
16484         (reload_adjust_reg_for_temp): Likewise.
16485         (emit_reload_insns): Likewise.
16486         (delete_output_reload): Likewise.
16487         * rtlanal.c (subreg_get_info): Likewise.
16488         * sched-deps.c (sched_analyze_reg): Likewise.
16489         * sel-sched.c (init_regs_for_mode): Likewise.
16490         (mark_unavailable_hard_regs): Likewise.
16491         (choose_best_reg_1): Likewise.
16492         (verify_target_availability): Likewise.
16493         * valtrack.c (dead_debug_insert_temp): Likewise.
16494         * var-tracking.c (track_loc_p): Likewise.
16495         (emit_note_insn_var_location): Likewise.
16496         * varasm.c (make_decl_rtl): Likewise.
16497         * reginfo.c (choose_hard_reg_mode): Likewise.
16498         (init_reg_modes_target): Refer directly to
16499         this_target_regs->x_hard_regno_nregs.
16501 2017-09-12  Richard Sandiford  <richard.sandiford@linaro.org>
16503         * ira-costs.c (record_operand_costs): Use in_hard_reg_set_p
16504         instead of hard_regno_nregs.
16506 2017-09-12  Richard Sandiford  <richard.sandiford@linaro.org>
16508         * config/aarch64/aarch64.c (aarch64_hard_regno_mode_ok): Use
16509         end_hard_regno instead of hard_regno_nregs.
16510         * config/s390/s390.c (s390_reg_clobbered_rtx): Likewise.
16511         * config/sparc/sparc.h (ASM_DECLARE_REGISTER_GLOBAL): Likewise.
16512         * config/visium/visium.c (visium_hard_regno_mode_ok): Likewise.
16513         * ira-color.c (improve_allocation): Likewise.
16514         * lra-assigns.c (find_hard_regno_for_1): Likewise.
16515         * lra-lives.c (mark_regno_live): Likewise.
16516         (mark_regno_dead): Likewise.
16517         * lra-remat.c (operand_to_remat): Likewise.
16518         * lra.c (collect_non_operand_hard_regs): Likewise.
16519         * postreload.c (reload_combine_note_store): Likewise.
16520         (move2add_valid_value_p): Likewise.
16521         * reload.c (regno_clobbered_p): Likewise.
16523 2017-09-12  Richard Sandiford  <richard.sandiford@linaro.org>
16525         * config/frv/frv.c (FOR_EACH_REGNO): Use END_REGNO instead of
16526         hard_regno_nregs.
16527         * config/v850/v850.c (v850_reorg): Likewise.
16528         * reload.c (refers_to_regno_for_reload_p): Likewise.
16529         (find_equiv_reg): Likewise.
16530         * reload1.c (reload_reg_reaches_end_p): Likewise.
16532 2017-09-12  Richard Sandiford  <richard.sandiford@linaro.org>
16534         * caller-save.c (add_used_regs): Use REG_NREGS instead of
16535         hard_regno_nregs.
16536         * config/aarch64/aarch64.c (aarch64_split_combinev16qi): Likewise.
16537         * config/arm/arm.c (output_move_neon): Likewise.
16538         (arm_attr_length_move_neon): Likewise.
16539         (neon_split_vcombine): Likewise.
16540         * config/c6x/c6x.c (c6x_mark_reg_read): Likewise.
16541         (c6x_mark_reg_written): Likewise.
16542         (c6x_dwarf_register_span): Likewise.
16543         * config/i386/i386.c (ix86_save_reg): Likewise.
16544         * config/ia64/ia64.c (mark_reg_gr_used_mask): Likewise.
16545         (rws_access_reg): Likewise.
16546         * config/s390/s390.c (s390_call_saved_register_used): Likewise.
16547         * mode-switching.c (create_pre_exit): Likewise.
16548         * ree.c (combine_reaching_defs): Likewise.
16549         (add_removable_extension): Likewise.
16550         * regcprop.c (find_oldest_value_reg): Likewise.
16551         (copyprop_hardreg_forward_1): Likewise.
16552         * reload.c (reload_inner_reg_of_subreg): Likewise.
16553         (push_reload): Likewise.
16554         (combine_reloads): Likewise.
16555         (find_dummy_reload): Likewise.
16556         (reload_adjust_reg_for_mode): Likewise.
16557         * reload1.c (find_reload_regs): Likewise.
16558         (forget_old_reloads_1): Likewise.
16559         (reload_reg_free_for_value_p): Likewise.
16560         (reload_adjust_reg_for_temp): Likewise.
16561         (emit_reload_insns): Likewise.
16562         (delete_output_reload): Likewise.
16563         * sel-sched.c (choose_best_reg_1): Likewise.
16564         (choose_best_pseudo_reg): Likewise.
16566 2017-09-12  Richard Sandiford  <richard.sandiford@linaro.org>
16567             Alan Hayward  <alan.hayward@arm.com>
16568             David Sherwood <david.sherwood@arm.com>
16570         * defaults.h (SLOW_UNALIGNED_ACCESS): Delete.
16571         * target.def (slow_unaligned_access): New hook.
16572         * targhooks.h (default_slow_unaligned_access): Declare.
16573         * targhooks.c (default_slow_unaligned_access): New function.
16574         * doc/tm.texi.in (SLOW_UNALIGNED_ACCESS): Replace with...
16575         (TARGET_SLOW_UNALIGNED_ACCESS): ...this.
16576         * doc/tm.texi: Regenerate.
16577         * config/alpha/alpha.h (SLOW_UNALIGNED_ACCESS): Delete.
16578         * config/arm/arm.h (SLOW_UNALIGNED_ACCESS): Delete.
16579         * config/i386/i386.h (SLOW_UNALIGNED_ACCESS): Delete commented-out
16580         definition.
16581         * config/powerpcspe/powerpcspe.h (SLOW_UNALIGNED_ACCESS): Delete.
16582         * config/powerpcspe/powerpcspe.c (TARGET_SLOW_UNALIGNED_ACCESS):
16583         Redefine.
16584         (rs6000_slow_unaligned_access): New function.
16585         (rs6000_emit_move): Use it instead of SLOW_UNALIGNED_ACCESS.
16586         (expand_block_compare): Likewise.
16587         (expand_strn_compare): Likewise.
16588         (rs6000_rtx_costs): Likewise.
16589         * config/riscv/riscv.h (SLOW_UNALIGNED_ACCESS): Delete.
16590         (riscv_slow_unaligned_access): Likewise.
16591         * config/riscv/riscv.c (riscv_slow_unaligned_access): Rename to...
16592         (riscv_slow_unaligned_access_p): ...this and make static.
16593         (riscv_option_override): Update accordingly.
16594         (riscv_slow_unaligned_access): New function.
16595         (TARGET_SLOW_UNALIGNED_ACCESS): Redefine.
16596         * config/rs6000/rs6000.h (SLOW_UNALIGNED_ACCESS): Delete.
16597         * config/rs6000/rs6000.c (TARGET_SLOW_UNALIGNED_ACCESS): Redefine.
16598         (rs6000_slow_unaligned_access): New function.
16599         (rs6000_emit_move): Use it instead of SLOW_UNALIGNED_ACCESS.
16600         (rs6000_rtx_costs): Likewise.
16601         * config/rs6000/rs6000-string.c (expand_block_compare)
16602         (expand_strn_compare): Use targetm.slow_unaligned_access instead
16603         of SLOW_UNALIGNED_ACCESS.
16604         * config/tilegx/tilegx.h (SLOW_UNALIGNED_ACCESS): Delete.
16605         * config/tilepro/tilepro.h (SLOW_UNALIGNED_ACCESS): Delete.
16606         * calls.c (expand_call): Use targetm.slow_unaligned_access instead
16607         of SLOW_UNALIGNED_ACCESS.
16608         * expmed.c (simple_mem_bitfield_p): Likewise.
16609         * expr.c (alignment_for_piecewise_move): Likewise.
16610         (emit_group_load_1): Likewise.
16611         (emit_group_store): Likewise.
16612         (copy_blkmode_from_reg): Likewise.
16613         (emit_push_insn): Likewise.
16614         (expand_assignment): Likewise.
16615         (store_field): Likewise.
16616         (expand_expr_real_1): Likewise.
16617         * gimple-fold.c (gimple_fold_builtin_memory_op): Likewise.
16618         * lra-constraints.c (simplify_operand_subreg): Likewise.
16619         * stor-layout.c (bit_field_mode_iterator::next_mode): Likewise.
16620         * gimple-ssa-store-merging.c: Likewise in block comment at start
16621         of file.
16622         * tree-ssa-strlen.c: Include target.h.
16623         (handle_builtin_memcmp): Use targetm.slow_unaligned_access instead
16624         of SLOW_UNALIGNED_ACCESS.
16625         * system.h (SLOW_UNALIGNED_ACCESS): Poison.
16627 2017-09-12  Richard Sandiford  <richard.sandiford@linaro.org>
16629         PR rtl-optimization/82185
16630         * expmed.c (emit_store_flag_int): Only test tem if it has been
16631         initialized.
16633 2017-09-12  Richard Biener  <rguenther@suse.de>
16635         PR middle-end/82149
16636         * match.pd ((FTYPE) N CMP CST): Fix typo.
16638 2017-09-12  Simon Atanasyan  <simon.atanasyan@imgtec.com>
16640         * config/mips/mips.c (mips_attribute_table): Add 'short_call'
16641         attribute.
16642         (mips_near_type_p): Add 'short_call' attribute as a synonym
16643         for 'near'.
16644         * doc/extend.texi (short_call): Document new function attribute.
16646 2017-09-12  Jakub Jelinek  <jakub@redhat.com>
16648         PR target/82112
16649         * c-common.c (sync_resolve_size): Instead of c_dialect_cxx ()
16650         assertion check that in the condition.
16651         (get_atomic_generic_size): Likewise.  Before testing if parameter
16652         has pointer type, if it has array type, call for C++
16653         default_conversion to perform array-to-pointer conversion.
16655 2017-09-12  Richard Biener  <rguenther@suse.de>
16657         * tree-vect-generic.c (expand_vector_operations_1): Do nothing
16658         for operations we cannot scalarize.
16660 2017-09-12  Aldy Hernandez  <aldyh@redhat.com>
16662         * tree-ssa-threadbackward.c (fsm_find_thread_path): Make GC
16663         vectors heap vectors.  Clean up comments.
16664         Make visited_bbs a reference.
16665         (profitable_jump_thread_path): Make GC
16666         vectors heap vectors.  Clean up comments.
16667         Misc cleanups.
16668         (convert_and_register_jump_thread_path): Make GC vectors heap
16669         vectors.
16670         (check_subpath_and_update_thread_path): Same.  Clean up comments.
16671         Make visited_bbs a reference.
16672         (handle_phi): Abstract common code to to
16673         register_jump_thread_path_if_profitable.
16674         Rename VAR_BB to DEF_BB.
16675         Update comments.
16676         Make GC vectors heap vectors.
16677         Make visited_bbs a reference.
16678         (handle_assignment): Same.
16679         (register_jump_thread_path_if_profitable): New.
16680         (fsm_find_control_statement_thread_paths): Rename VAR_BB to
16681         DEF_BB.
16682         Make GC vectors heap vectors.  Clean up comments.
16683         Make visited_bbs a reference.
16684         (find_jump_threads_backwards): Make visited_bbs live in the stack.
16685         * tree-ssa-threadupdate.c (delete_jump_thread_path): Fix typo in
16686         comment.
16688 2017-09-11  Max Filippov  <jcmvbkbc@gmail.com>
16690         PR target/82181
16691         * config/xtensa/xtensa.c (xtensa_mem_offset): Check that both
16692         words of E_DImode object are reachable by xtensa_uimm8x4 access.
16694 2017-09-11  Vidya Praveen  <vidyapraveen@arm.com>
16696         Revert r251800 and r251799.
16698 2017-09-11  Martin Jambor  <mjambor@suse.cz>
16700         PR hsa/82119
16701         * hsa-gen.c (gen_hsa_phi_from_gimple_phi): Process ADDR_EXPRs in
16702         arguments in advance.
16703         * hsa-regalloc.c (naive_process_phi): New parameter predecessors,
16704         use it to find predecessor edges.
16705         (naive_outof_ssa): Collect vector of predecessors.
16707 2017-09-08  Jason Merrill  <jason@redhat.com>
16709         PR c++/70029 - ICE with ref-qualifier and -flto
16710         * langhooks.h (struct lang_hooks_for_types): Add
16711         copy_lang_qualifiers.
16712         * attribs.c (build_type_attribute_qual_variant): Use it.
16713         * langhooks-def.h (LANG_HOOKS_COPY_LANG_QUALIFIERS): Default to
16714         NULL.
16715         (LANG_HOOKS_FOR_TYPES_INITIALIZER): Use it.
16716         * tree.c (verify_type): Re-enable TYPE_CANONICAL main variant check.
16718 2017-09-08  Eric Botcazou  <ebotcazou@adacore.com>
16720         PR target/81988
16721         * config/sparc/sparc.md (mulsi3): Rename into *mulsi3_sp32.
16722         (*mulsi3_sp64): New instruction.
16723         (mulsi3): New expander.
16725 2017-09-08  Uros Bizjak  <ubizjak@gmail.com>
16727         * config/alpha/alpha.c (alpha_print_operand) <case 'S'>: Remove.
16729 2017-09-08  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
16731         * sancov.c: Include memmodel.h.
16733 2017-09-07  Eric Botcazou  <ebotcazou@adacore.com>
16735         PR target/80897
16736         * config/sparc/sparc.c (sparc_emit_set_symbolic_const64): Deal with too
16737         large offsets.
16739 2017-09-07  Carl Love  <cel@us.ibm.com>
16741         * config/rs6000/vsx.md (define_insn "*stxvl"): Add missing argument to
16742         the sldi instruction.
16744 2017-09-07  David Edelsohn  <dje.gcc@gmail.com>
16746         * sancov.c: Include tm_p.h.
16748 2017-09-07  Jakub Jelinek  <jakub@redhat.com>
16750         PR target/81979
16751         * output.h (switch_to_other_text_partition): New declaration.
16752         * varasm.c (switch_to_other_text_partition): New function.
16753         * config/rs6000/rs6000.c (uses_TOC): Return 2 if
16754         NOTE_INSN_SWITCH_TEXT_SECTIONS is seen before finding load_toc_* insn.
16755         (rs6000_elf_declare_function_name): If uses_TOC returned 2, switch
16756         to the other text partition before emitting LCL label and switch back
16757         after emitting the word after it.
16759 2017-09-07  Richard Biener  <rguenther@suse.de>
16761         * passes.def (pass_split_crit_edges): Remove instance before PRE.
16762         * tree-ssa-pre.c (pass_pre::execute): Instead manually split
16763         critical edges here, after loop init.
16764         (pass_data_pre): Remove PROP_no_crit_edges flags.
16765         * tree-ssa-sccvn.c (vn_reference_lookup_3): Use vn_valueize
16766         for valueization of call args to avoid leaking VN_TOP.
16767         (visit_use): Assert we do not visit default defs.
16768         (init_scc_vn): Use build_decl for VN_TOP to make name nicer.
16769         Use error_mark_node to more easily detect leaking VN_TOP.
16770         All default-defs are varying, not VN_TOP.  Mark them visited.
16771         (run_scc_vn): Make code match comment.
16773 2017-09-07  Michael Meissner  <meissner@linux.vnet.ibm.com>
16775         * config/rs6000/rs6000-cpus.def (OTHER_VSX_VECTOR_MASKS): Delete
16776         OPTION_MASK_FLOAT128_KEYWORD.
16777         (POWERPC_MASKS): Likewise.
16778         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Delete
16779         support for the -mfloat128-type option, and make -mfloat128
16780         default on PowerPC Linux systems.  Define or undefine
16781         __FLOAT128__ and  __FLOAT128_HARDWARE__ for the current options.
16782         Define __float128 to be __ieee128 if IEEE 128-bit support is
16783         enabled, or undefine it.
16784         (rs6000_cpu_cpp_builtins): Delete defining __FLOAT128__ here.
16785         Delete defining __FLOAT128_TYPE__.
16786         * config/rs6000/rs6000.opt (x_TARGET_FLOAT128_TYPE): Delete the
16787         -mfloat128-type option and make -mfloat128 default on PowerPC
16788         Linux systems.
16789         (TARGET_FLOAT128_TYPE): Likewise.
16790         (-mfloat128-type): Likewise.
16791         * config/rs6000/rs6000.c (rs6000_option_override_internal):
16792         Delete the -mfloat128-type option and make -mfloat128 default on
16793         PowerPC Linux systems.  Always use __ieee128 to be the keyword for
16794         the IEEE 128-bit type, and map __float128 to __ieee128 if IEEE
16795         128-bit floating point is enabled.  Change tests from using
16796         -mfloat128-type to -mfloat128.
16797         (rs6000_mangle_type): Use the correct mangling for the __float128
16798         type even if normal long double is restricted to 64-bits.
16799         (floatn_mode): Enable the _Float128 type by default on VSX Linux
16800         systems.
16801         * config/rs6000/rs6000.h (MASK_FLOAT128_TYPE): Delete.
16802         (MASK_FLOAT128_KEYWORD): Define new shortcut macro.
16803         (RS6000BTM_FLOAT128): Define in terms of -mfloat128, not
16804         -mfloat128-type.
16805         * doc/invoke.texi (RS/6000 and PowerPC Options): Update
16806         documentation for -mfloat128.
16808 2017-09-06  Olivier Hainque  <hainque@adacore.com>
16810         * config.gcc (powerpc-wrs-vxworksspe): Now match as vxworks*spe.
16812 2017-09-06  Wish Wu  <wishwu007@gmail.com>
16813             Jakub Jelinek  <jakub@redhat.com>
16815         * asan.c (initialize_sanitizer_builtins): Add
16816         BT_FN_VOID_UINT8_UINT8, BT_FN_VOID_UINT16_UINT16,
16817         BT_FN_VOID_UINT32_UINT32, BT_FN_VOID_UINT64_UINT64,
16818         BT_FN_VOID_FLOAT_FLOAT, BT_FN_VOID_DOUBLE_DOUBLE and
16819         BT_FN_VOID_UINT64_PTR variables.
16820         * builtin-types.def (BT_FN_VOID_UINT8_UINT8): New fn type.
16821         (BT_FN_VOID_UINT16_UINT16): Likewise.
16822         (BT_FN_VOID_UINT32_UINT32): Likewise.
16823         (BT_FN_VOID_FLOAT_FLOAT): Likewise.
16824         (BT_FN_VOID_DOUBLE_DOUBLE): Likewise.
16825         (BT_FN_VOID_UINT64_PTR): Likewise.
16826         * common.opt (flag_sanitize_coverage): New variable.
16827         (fsanitize-coverage=trace-pc): Remove.
16828         (fsanitize-coverage=): Add.
16829         * flag-types.h (enum sanitize_coverage_code): New enum.
16830         * fold-const.c (fold_range_test): Disable non-short-circuit
16831         optimization if flag_sanitize_coverage.
16832         (fold_truth_andor): Likewise.
16833         * tree-ssa-ifcombine.c (ifcombine_ifandif): Likewise.
16834         * opts.c (COVERAGE_SANITIZER_OPT): Define.
16835         (coverage_sanitizer_opts): New array.
16836         (get_closest_sanitizer_option): Add OPTS argument, handle also
16837         OPT_fsanitize_coverage_.
16838         (parse_sanitizer_options): Adjusted to also handle
16839         OPT_fsanitize_coverage_.
16840         (common_handle_option): Add OPT_fsanitize_coverage_.
16841         * sancov.c (instrument_comparison, instrument_switch): New function.
16842         (sancov_pass): Add trace-cmp support.
16843         * sanitizer.def (BUILT_IN_SANITIZER_COV_TRACE_CMP1,
16844         BUILT_IN_SANITIZER_COV_TRACE_CMP2, BUILT_IN_SANITIZER_COV_TRACE_CMP4,
16845         BUILT_IN_SANITIZER_COV_TRACE_CMP8,
16846         BUILT_IN_SANITIZER_COV_TRACE_CONST_CMP1,
16847         BUILT_IN_SANITIZER_COV_TRACE_CONST_CMP2,
16848         BUILT_IN_SANITIZER_COV_TRACE_CONST_CMP4,
16849         BUILT_IN_SANITIZER_COV_TRACE_CONST_CMP8,
16850         BUILT_IN_SANITIZER_COV_TRACE_CMPF, BUILT_IN_SANITIZER_COV_TRACE_CMPD,
16851         BUILT_IN_SANITIZER_COV_TRACE_SWITCH): New builtins.
16852         * doc/invoke.texi: Document -fsanitize-coverage=trace-cmp.
16854 2017-09-06  Richard Earnshaw  <rearnsha@arm.com>
16856         * config/arm/parsecpu.awk (fatal): Note that we've encountered an
16857         error.  Only quit immediately if parsing is complete.
16858         (BEGIN): Initialize fatal_err and parse_done.
16859         (begin fpu, end fpu): Check number of arguments.
16860         (begin arch, end arch): Likewise.
16861         (begin cpu, end cpu): Likewise.
16862         (cname, tune for, tune flags, architecture, fpu, option): Likewise.
16863         (optalias): Likewise.
16865 2017-09-06  Richard Earnshaw  <rearnsha@arm.com>
16867         * config.gcc (arm*-*-*): Don't add arm-isa.h to tm_p_file.
16868         * config/arm/arm-isa.h: Delete.  Move definitions to ...
16869         * arm-cpus.in: ... here.  Use new feature and fgroup values.
16870         * config/arm/arm.c (arm_option_override): Use lower case for feature
16871         bit names.
16872         * config/arm/arm.h (TARGET_HARD_FLOAT): Likewise.
16873         (TARGET_VFP3, TARGET_VFP5, TARGET_FMA): Likewise.
16874         * config/arm/parsecpu.awk (END): Add new command 'isa'.
16875         (isa_pfx): Delete.
16876         (print_isa_bits_for): New function.
16877         (gen_isa): New function.
16878         (gen_comm_data): Use print_isa_bits_for.
16879         (define feature): New keyword.
16880         (define fgroup): New keyword.
16881         * config/arm/t-arm (OPTIONS_H_EXTRA): Add arm-isa.h
16882         (arm-isa.h): Add rule to generate file.
16883         * common/config/arm/arm-common.c: (arm_canon_arch_option): Use lower
16884         case for feature bit names.
16886 2017-09-06  Richard Biener  <rguenther@suse.de>
16888         * tree-ssa-pre.c (NECESSARY): Remove.
16889         (create_expression_by_pieces): Do not touch pass-local flags.
16890         (insert_into_preds_of_block): Likewise.
16891         (do_pre_regular_insertion): Likewise.
16892         (eliminate_insert): Likewise.
16893         (eliminate_dom_walker::before_dom_children): Likewise.
16894         (fini_eliminate): Do not look at inserted_exprs.
16895         (mark_operand_necessary): Remove.
16896         (remove_dead_inserted_code): Replace with simple work-list
16897         algorithm based on inserted_exprs and SSA uses.
16898         (pass_pre::execute): Re-order fini_eliminate and
16899         remove_dead_inserted_code.
16901 2017-09-06  Olivier Hainque  <hainque@adacore.com>
16903         * config/powerpcspe/vxworks.h (VXCPU_FOR_8548): Correct definition
16904         for VxWorks 7.  Adjust surrounding comments.
16906 2017-09-06  Richard Biener  <rguenther@suse.de>
16908         * gimple-ssa-strength-reduction.c
16909         (find_candidates_dom_walker::before_dom_children): Also allow
16910         pointer types.
16912 2017-09-06  Richard Biener  <rguenther@suse.de>
16914         PR tree-optimization/82108
16915         * tree-vect-stmts.c (vectorizable_load): Fix pointer adjustment
16916         for gap in the non-permutation SLP case.
16918 2017-09-06  Martin Jambor  <mjambor@suse.cz>
16920         PR tree-optimization/82078
16921         * tree-sra.c (sort_and_splice_var_accesses): Move call to
16922         add_access_to_work_queue...
16923         (build_accesses_from_assign): ...here.
16924         (propagate_all_subaccesses): Make sure racc is the group
16925         representative, if there is one.
16927 2017-09-06  Jakub Jelinek  <jakub@redhat.com>
16929         PR middle-end/82095
16930         * varasm.c (categorize_decl_for_section): Use SECCAT_TBSS for TLS vars with
16931         NULL DECL_INITIAL.
16933 2017-09-06  Richard Biener  <rguenther@suse.de>
16935         * gimple-ssa-strength-reduction.c
16936         (find_candidates_dom_walker::before_doom_children): Use a
16937         type and not a mode check.
16939 2017-09-06  Bernd Edlinger  <bernd.edlinger@hotmail.de>
16941         PR target/77308
16942         * config/arm/predicates.md (arm_general_adddi_operand): Create new
16943         non-vfp predicate.
16944         * config/arm/arm.md (*arm_adddi3, *arm_subdi3): Use new predicates.
16946 2017-09-05  Jeff Law  <law@redhat.com>
16948         PR tree-optimization/64910
16949         * tree-ssa-reassoc.c (reassociate_bb): Restrict last change to
16950         cases where we have 3 or more operands.
16952 2017-09-05  Jakub Jelinek  <jakub@redhat.com>
16954         PR middle-end/81768
16955         * omp-low.c (lower_omp_for): Recompute tree invariant if
16956         gimple_omp_for_initial/final is ADDR_EXPR.
16958         PR middle-end/81768
16959         * omp-expand.c (expand_omp_simd): Force second operands of COND_EXPR
16960         into gimple val before gimplification fo the COND_EXPR.
16962 2017-09-05  Aldy Hernandez  <aldyh@redhat.com>
16964         * tree-ssa-threadupdate.c (duplicate_thread_path): Remove unused
16965         REGION_COPY argument.
16966         (thread_through_all_blocks): Remove unused argument to
16967         duplicate_thread_path.
16969 2017-09-05  Richard Sandiford  <richard.sandiford@linaro.org>
16970             Alan Hayward  <alan.hayward@arm.com>
16971             David Sherwood  <david.sherwood@arm.com>
16973         * config/aarch64/aarch64-protos.h (aarch64_gen_adjusted_ldpstp):
16974         Take a scalar_mode rather than a machine_mode.
16975         (aarch64_operands_adjust_ok_for_ldpstp): Likewise.
16976         * config/aarch64/aarch64.c (aarch64_simd_container_mode): Likewise.
16977         (aarch64_operands_adjust_ok_for_ldpstp): Likewise.
16978         (aarch64_gen_adjusted_ldpstp): Likewise.
16979         (aarch64_expand_vector_init): Use scalar_mode instead of machine_mode.
16981 2017-09-05  Richard Sandiford  <richard.sandiford@linaro.org>
16982             Alan Hayward  <alan.hayward@arm.com>
16983             David Sherwood  <david.sherwood@arm.com>
16985         * config/aarch64/aarch64-protos.h (aarch64_is_extend_from_extract):
16986         Take a scalar_int_mode instead of a machine_mode.
16987         (aarch64_mask_and_shift_for_ubfiz_p): Likewise.
16988         (aarch64_output_scalar_simd_mov_immediate): Likewise.
16989         (aarch64_simd_scalar_immediate_valid_for_move): Likewise.
16990         (aarch64_simd_attr_length_rglist): Delete.
16991         * config/aarch64/aarch64.c (aarch64_is_extend_from_extract): Take
16992         a scalar_int_mode instead of a machine_mode.
16993         (aarch64_add_offset): Likewise.
16994         (aarch64_internal_mov_immediate): Likewise
16995         (aarch64_add_constant_internal): Likewise.
16996         (aarch64_add_constant): Likewise.
16997         (aarch64_movw_imm): Likewise.
16998         (aarch64_rtx_arith_op_extract_p): Likewise.
16999         (aarch64_mask_and_shift_for_ubfiz_p): Likewise.
17000         (aarch64_simd_scalar_immediate_valid_for_move): Likewise.
17001         Remove assert that the mode isn't a vector.
17002         (aarch64_output_scalar_simd_mov_immediate): Likewise.
17003         (aarch64_expand_mov_immediate): Update calls after above changes.
17004         (aarch64_output_casesi): Use as_a <scalar_int_mode>.
17005         (aarch64_and_bitmask_imm): Check for scalar integer modes.
17006         (aarch64_move_imm): Likewise.
17007         (aarch64_can_const_movi_rtx_p): Likewise.
17008         (aarch64_strip_extend): Likewise.
17009         (aarch64_extr_rtx_p): Likewise.
17010         (aarch64_rtx_costs): Likewise, using wode_mode as the mode of
17011         a CONST_INT when the mode parameter is VOIDmode.
17012         (aarch64_float_const_rtx_p): Use scalar_int_mode for a temporary.
17014 2017-09-05  Richard Sandiford  <richard.sandiford@linaro.org>
17016         * machmode.h (bitwise_mode_for_mode): Return opt_mode.
17017         * stor-layout.c (bitwise_mode_for_mode): Likewise.
17018         (bitwise_type_for_mode): Update accordingly.
17020 2017-09-05  Richard Sandiford  <richard.sandiford@linaro.org>
17022         * stor-layout.h (mode_for_size_tree): Return an opt_mode.
17023         * stor-layout.c (mode_for_size_tree): Likewise.
17024         (mode_for_array): Update accordingly.
17025         (layout_decl): Likewise.
17026         (compute_record_mode): Likewise.  Only set the mode once.
17028 2017-09-05  Richard Sandiford  <richard.sandiford@linaro.org>
17030         * target.def (get_mask_mode): Change return type to opt_mode.
17031         Expand commentary.
17032         * doc/tm.texi: Regenerate.
17033         * targhooks.h (default_get_mask_mode): Return an opt_mode.
17034         * targhooks.c (default_get_mask_mode): Likewise.
17035         * config/i386/i386.c (ix86_get_mask_mode): Likewise.
17036         * optabs-query.c (can_vec_mask_load_store_p): Update use of
17037         targetm.get_mask_mode.
17038         * tree.c (build_truth_vector_type): Likewise.
17040 2017-09-05  Richard Sandiford  <richard.sandiford@linaro.org>
17042         * machmode.h (mode_for_vector): Return an opt_mode.
17043         * stor-layout.c (mode_for_vector): Likewise.
17044         (mode_for_int_vector): Update accordingly.
17045         (layout_type): Likewise.
17046         * config/i386/i386.c (emit_memmov): Likewise.
17047         (ix86_expand_set_or_movmem): Likewise.
17048         (ix86_expand_vector_init): Likewise.
17049         (ix86_get_mask_mode): Likewise.
17050         * config/powerpcspe/powerpcspe.c (rs6000_expand_vec_perm_const_1):
17051         Likewise.
17052         * config/rs6000/rs6000.c (rs6000_expand_vec_perm_const_1): Likewise.
17053         * expmed.c (extract_bit_field_1): Likewise.
17054         * expr.c (expand_expr_real_2): Likewise.
17055         * optabs-query.c (can_vec_perm_p): Likewise.
17056         (can_vec_mask_load_store_p): Likewise.
17057         * optabs.c (expand_vec_perm): Likewise.
17058         * targhooks.c (default_get_mask_mode): Likewise.
17059         * tree-vect-stmts.c (vectorizable_store): Likewise.
17060         (vectorizable_load): Likewise.
17061         (get_vectype_for_scalar_type_and_size): Likewise.
17063 2017-09-05  Richard Sandiford  <richard.sandiford@linaro.org>
17065         * machmode.h (mode_for_int_vector): New function.
17066         * stor-layout.c (mode_for_int_vector): Likewise.
17067         * config/aarch64/aarch64.c (aarch64_emit_approx_sqrt): Use it.
17068         * config/powerpcspe/powerpcspe.c (rs6000_do_expand_vec_perm): Likewise.
17069         * config/rs6000/rs6000.c (rs6000_do_expand_vec_perm): Likewise.
17070         * config/s390/s390.c (s390_expand_vec_compare_cc): Likewise.
17071         (s390_expand_vcond): Likewise.
17073 2017-09-05  Richard Sandiford  <richard.sandiford@linaro.org>
17075         * machmode.h (opt_machine_mode): New type.
17076         (opt_mode<T>): Allow construction from anything that can be
17077         converted to a T.
17078         (is_a, as_a, dyn_cast): Add overloads for opt_mode.
17079         (mode_for_size): Return an opt_machine_mode.
17080         * stor-layout.c (mode_for_size): Likewise.
17081         (mode_for_size_tree): Update call accordingly.
17082         (bitwise_mode_for_mode): Likewise.
17083         (make_fract_type): Likewise.
17084         (make_accum_type): Likewise.
17085         * caller-save.c (replace_reg_with_saved_mem): Update call
17086         accordingly.
17087         * config/alpha/alpha.h (SECONDARY_MEMORY_NEEDED_MODE): Likewise.
17088         * config/i386/i386.h (SECONDARY_MEMORY_NEEDED_MODE): Likewise.
17089         * config/s390/s390.h (SECONDARY_MEMORY_NEEDED_MODE): Likewise.
17090         * config/sparc/sparc.h (SECONDARY_MEMORY_NEEDED_MODE): Likewise.
17091         * expmed.c (extract_bit_field_1): Likewise.
17092         * reload.c (get_secondary_mem): Likewise.
17093         * varasm.c (assemble_integer): Likewise.
17094         * lower-subreg.c (simplify_subreg_concatn): Likewise.  Move
17095         early-out.
17097 2017-09-05  Richard Sandiford  <richard.sandiford@linaro.org>
17099         * machmode.h (decimal_float_mode_for_size): New function.
17100         * real.h (REAL_VALUE_TO_TARGET_LONG_DOUBLE): Use float_mode_for_size.
17101         (REAL_VALUE_TO_TARGET_DOUBLE): Likewise.
17102         (REAL_VALUE_TO_TARGET_SINGLE): Likewise.
17103         (REAL_VALUE_TO_TARGET_DECIMAL128): Use decimal_float_mode_for_size.
17104         (REAL_VALUE_TO_TARGET_DECIMAL64): Likewise.
17105         (REAL_VALUE_TO_TARGET_DECIMAL32): Likewise.
17107 2017-09-05  Richard Sandiford  <richard.sandiford@linaro.org>
17109         * builtins.c (expand_builtin_powi): Use int_mode_for_size.
17110         (get_builtin_sync_mode): Likewise.
17111         (expand_ifn_atomic_compare_exchange): Likewise.
17112         (expand_builtin_atomic_clear): Likewise.
17113         (expand_builtin_atomic_test_and_set): Likewise.
17114         (fold_builtin_atomic_always_lock_free): Likewise.
17115         * calls.c (compute_argument_addresses): Likewise.
17116         (emit_library_call_value_1): Likewise.
17117         (store_one_arg): Likewise.
17118         * combine.c (combine_instructions): Likewise.
17119         * config/aarch64/aarch64.c (aarch64_function_value): Likewise.
17120         * config/arm/arm.c (arm_function_value): Likewise.
17121         (aapcs_allocate_return_reg): Likewise.
17122         * config/c6x/c6x.c (c6x_expand_movmem): Likewise.
17123         * config/i386/i386.c (construct_container): Likewise.
17124         (ix86_gimplify_va_arg): Likewise.
17125         (ix86_expand_sse_cmp): Likewise.
17126         (emit_memmov): Likewise.
17127         (emit_memset): Likewise.
17128         (expand_small_movmem_or_setmem): Likewise.
17129         (ix86_expand_pextr): Likewise.
17130         (ix86_expand_pinsr): Likewise.
17131         * config/lm32/lm32.c (lm32_block_move_inline): Likewise.
17132         * config/microblaze/microblaze.c (microblaze_block_move_straight):
17133         Likewise.
17134         * config/mips/mips.c (mips_function_value_1) Likewise.
17135         (mips_block_move_straight): Likewise.
17136         (mips_expand_ins_as_unaligned_store): Likewise.
17137         * config/powerpcspe/powerpcspe.c
17138         (rs6000_darwin64_record_arg_advance_flush): Likewise.
17139         (rs6000_darwin64_record_arg_flush): Likewise.
17140         * config/rs6000/rs6000.c
17141         (rs6000_darwin64_record_arg_advance_flush): Likewise.
17142         (rs6000_darwin64_record_arg_flush): Likewise.
17143         * config/sparc/sparc.c (sparc_function_arg_1): Likewise.
17144         (sparc_function_value_1): Likewise.
17145         * config/spu/spu.c (adjust_operand): Likewise.
17146         (spu_emit_branch_or_set): Likewise.
17147         (arith_immediate_p): Likewise.
17148         * emit-rtl.c (gen_lowpart_common): Likewise.
17149         * expr.c (expand_expr_real_1): Likewise.
17150         * function.c (assign_parm_setup_block): Likewise.
17151         * gimple-ssa-store-merging.c (encode_tree_to_bitpos): Likewise.
17152         * reload1.c (alter_reg): Likewise.
17153         * stor-layout.c (mode_for_vector): Likewise.
17154         (layout_type): Likewise.
17156 2017-09-05  Richard Sandiford  <richard.sandiford@linaro.org>
17158         * config/spu/spu.c (exp2_immediate_p): Use int_mode_for_mode.
17159         (spu_convert_move): Likewise.
17160         * lower-subreg.c (resolve_simple_move): Likewise.
17162 2017-09-05  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
17164         PR target/81833
17165         * config/rs6000/altivec.md (altivec_vsum2sws): Convert from a
17166         define_insn to a define_expand.
17167         (altivec_vsum2sws_direct): New define_insn.
17168         (altivec_vsumsws): Convert from a define_insn to a define_expand.
17170 2017-09-05  Wilco Dijkstra  <wdijkstr@arm.com>
17172         * config/arm/arm.c (arm_option_params_internal): Improve setting of
17173         max_insns_skipped.
17175 2017-09-05  H.J. Lu  <hongjiu.lu@intel.com>
17177         PR target/59501
17178         PR target/81624
17179         PR target/81769
17180         * config/i386/i386.c (ix86_finalize_stack_frame_flags): Don't
17181         realign stack if stack alignment needed is less than incoming
17182         stack boundary.
17184 2017-09-05  Marek Polacek  <polacek@redhat.com>
17186         PR sanitizer/82072
17187         * convert.c (convert_to_integer_1) <case NEGATE_EXPR>: Move the ubsan
17188         check earlier.
17190 2017-09-05  Wilco Dijkstra  <wdijkstr@arm.com>
17192         * explow.c (get_dynamic_stack_size): Improve dynamic alignment.
17194 2017-09-05  Richard Biener  <rguenther@suse.de>
17196         PR tree-optimization/82084
17197         * fold-const.c (can_native_encode_string_p): Handle wide characters.
17199 2017-09-05  Richard Biener  <rguenther@suse.de>
17201         PR tree-optimization/82102
17202         * tree-ssa-pre.c (fini_eliminate): Check if lhs is NULL.
17204 2017-09-05  Martin Liska  <mliska@suse.cz>
17206         PR tree-optimization/82032
17207         * tree-cfg.c (generate_range_test): New function.
17208         * tree-cfg.h (generate_range_test): Declared here.
17209         * tree-cfgcleanup.c (convert_single_case_switch): New function.
17210         (cleanup_control_expr_graph): Use it.
17211         * tree-switch-conversion.c (try_switch_expansion): Remove
17212         assert.
17213         (emit_case_nodes): Use generate_range_test.
17215 2017-09-04  Uros Bizjak  <ubizjak@gmail.com>
17217         PR target/82098
17218         * config/i386/i386.md (*<btsc><mode>_mask): Add
17219         TARGET_USE_BT to insn constraint.
17220         (*btr<mode>_mask): Ditto.
17222 2017-09-04  Wilco Dijkstra  <wdijkstr@arm.com>
17224         * config/arm/arm.c (arm_legitimate_index_p): Add comment.
17225         (thumb2_legitimate_index_p): Use correct range for DI/DF mode.
17227 2017-09-04  Bernd Edlinger  <bernd.edlinger@hotmail.de>
17229         PR target/77308
17230         * config/arm/arm.md (*arm_adddi3, *arm_subdi3): Split early except for
17231         TARGET_NEON and TARGET_IWMMXT.
17232         (anddi3, iordi3, xordi3, one_cmpldi2): Split while expanding except for
17233         TARGET_NEON and TARGET_IWMMXT.
17234         (*one_cmpldi2_insn): Moved the body of one_cmpldi2 here.
17236 2017-09-04  Uros Bizjak  <ubizjak@gmail.com>
17238         * config/i386/i386-protos.h (ix86_tls_address_pattern_p) New prototype.
17239         (ix86_rewrite_tls_address): Ditto.
17240         * config/i386/i386.c (ix86_tls_address_pattern_p) New function.
17241         (ix86_rewrite_tls_address_1): Ditto.
17242         (ix86_rewrite_tls_address): Ditto.
17243         * config/i386/predicates.md (tls_address_pattern): New predicate.
17244         * config/i386/i386.md (TLS address splitter): New splitter.
17246 2017-09-04  Richard Biener  <rguenther@suse.de>
17248         PR tree-optimization/82084
17249         * fold-const.h (can_native_encode_string_p): Declare.
17250         * fold-const.c (can_native_encode_string_p): Factor out from ...
17251         (native_encode_string): ... here.
17252         * tree-vect-stmts.c (vectorizable_store): Call it to avoid
17253         vectorizing stores from constants we later cannot handle.
17255 2017-09-04  Marek Polacek  <polacek@redhat.com>
17257         PR c/81783
17258         * doc/invoke.texi: Update -Wtautological-compare documentation.
17260 2017-09-04  Jeff Law  <law@redhat.com>
17262         PR tree-optimization/64910
17263         * tree-ssa-reassoc.c (reassociate_bb): For bitwise binary ops,
17264         swap the first and last operand if the last is a constant.
17266 2017-09-04  Marek Polacek  <polacek@redhat.com>
17268         PR sanitizer/82072
17269         * convert.c (do_narrow): When sanitizing signed integer overflows,
17270         bail out for signed types.
17271         (convert_to_integer_1) <case NEGATE_EXPR>: Likewise.
17273 2017-09-04  Richard Biener  <rguenther@suse.de>
17275         PR tree-optimization/82060
17276         * tree-ssa-pre.c (eliminate_dom_walker::before_dom_children):
17277         Move devirtualization after stmt folding and before EH/AB/noreturn
17278         cleanup to get the stmt refs canonicalized.  Use a bool instead
17279         of gimple_modified_p since that doesn't work for NOPs.  Schedule
17280         NOPs generated by folding for removal.
17282 2017-09-04  Richard Sandiford  <richard.sandiford@linaro.org>
17283             Alan Hayward  <alan.hayward@arm.com>
17284             David Sherwood  <david.sherwood@arm.com>
17286         * coretypes.h (pad_direction): New enum.
17287         * defaults.h (DEFAULT_FUNCTION_ARG_PADDING): Delete.
17288         (FUNCTION_ARG_PADDING): Likewise.
17289         * target.def (function_arg_padding): New hook.
17290         * targhooks.h (default_function_arg_padding): Declare.
17291         * targhooks.c (default_function_arg_padding): New function.
17292         * doc/tm.texi.in (FUNCTION_ARG_PADDING): Replace with...
17293         (TARGET_FUNCTION_ARG_PADDING): ...this.
17294         * doc/tm.texi: Regenerate.
17295         * calls.c (store_unaligned_arguments_into_pseudos): Use pad_direction
17296         instead of direction.
17297         (compute_argument_addresses): Likewise.
17298         (load_register_parameters): Likewise.
17299         (emit_library_call_value_1): Likewise.
17300         (store_one_arg): Use targetm.calls.function_arg_padding instead
17301         of FUNCTION_ARG_PADDING.
17302         (must_pass_in_stack_var_size_or_pad): Likewise.
17303         * expr.c (emit_group_load_1): Use pad_direction instead of direction.
17304         (emit_group_store): Likewise.
17305         (emit_single_push_insn_1): Use targetm.calls.function_arg_padding
17306         instead of FUNCTION_ARG_PADDING.
17307         (emit_push_insn): Likewise, and propagate enum change throughout
17308         function.
17309         * function.h (direction): Delete.
17310         (locate_and_pad_arg_data::where_pad): Use pad_direction instead
17311         of direction.
17312         * function.c (assign_parm_find_stack_rtl): Likewise.
17313         (assign_parm_setup_block_p): Likewise.
17314         (assign_parm_setup_block): Likewise.
17315         (gimplify_parameters): Likewise.
17316         (locate_and_pad_parm): Use targetm.calls.function_arg_padding
17317         instead of FUNCTION_ARG_PADDING, and propagate enum change throughout
17318         function.
17319         * config/aarch64/aarch64.h (FUNCTION_ARG_PADDING): Delete.
17320         (BLOCK_REG_PADDING): Use pad_direction instead of direction.
17321         * config/aarch64/aarch64-protos.h (aarch64_pad_arg_upward): Delete.
17322         * config/aarch64/aarch64.c (aarch64_pad_arg_upward): Replace with...
17323         (aarch64_function_arg_padding): ...this new function.
17324         (aarch64_gimplify_va_arg_expr): Use pad_direction instead of direction.
17325         (TARGET_FUNCTION_ARG_PADDING): Redefine.
17326         * config/arm/arm.h (FUNCTION_ARG_PADDING): Delete.
17327         (BLOCK_REG_PADDING): Use pad_direction instead of direction.
17328         * config/arm/arm-protos.h (arm_pad_arg_upward): Delete.
17329         * config/arm/arm.c (TARGET_FUNCTION_ARG_PADDING): Redefine.
17330         (arm_pad_arg_upward): Replace with...
17331         (arm_function_arg_padding): ...this new function.
17332         * config/c6x/c6x.h (BLOCK_REG_PADDING): Use pad_direction instead
17333         of direction.
17334         * config/ia64/hpux.h (FUNCTION_ARG_PADDING): Delete.
17335         * config/ia64/ia64-protos.h (ia64_hpux_function_arg_padding): Delete.
17336         * config/ia64/ia64.c (TARGET_FUNCTION_ARG_PADDING): Redefine.
17337         (ia64_hpux_function_arg_padding): Replace with...
17338         (ia64_function_arg_padding): ...this new function.  Use pad_direction
17339         instead of direction.  Check for TARGET_HPUX.
17340         * config/iq2000/iq2000.h (FUNCTION_ARG_PADDING): Delete.
17341         * config/iq2000/iq2000.c (TARGET_FUNCTION_ARG_PADDING): Redefine.
17342         (iq2000_function_arg_padding): New function.
17343         * config/mips/mips-protos.h (mips_pad_arg_upward): Delete.
17344         * config/mips/mips.c (mips_pad_arg_upward): Replace with...
17345         (mips_function_arg_padding): ...this new function.
17346         (mips_pad_reg_upward): Update accordingly.
17347         (TARGET_FUNCTION_ARG_PADDING): Redefine.
17348         * config/mips/mips.h (PAD_VARARGS_DOWN): Use
17349         targetm.calls.function_arg_padding.
17350         (FUNCTION_ARG_PADDING): Delete.
17351         (BLOCK_REG_PADDING): Use pad_direction instead of direction.
17352         * config/nios2/nios2.h (FUNCTION_ARG_PADDING): Delete.
17353         (PAD_VARARGS_DOWN): Use targetm.calls.function_arg_padding.
17354         * config/nios2/nios2-protos.h (nios2_function_arg_padding): Delete.
17355         (nios2_block_reg_padding): Return pad_direction instead of direction.
17356         * config/nios2/nios2.c (nios2_block_reg_padding): Return pad_direction
17357         instead of direction.
17358         (nios2_function_arg_padding): Likewise.  Make static.
17359         (TARGET_FUNCTION_ARG_PADDING): Redefine.
17360         * config/pa/pa.h (FUNCTION_ARG_PADDING): Delete.
17361         (BLOCK_REG_PADDING): Use targetm.calls.function_arg_padding.
17362         * config/pa/pa-protos.h (pa_function_arg_padding): Delete.
17363         * config/pa/pa.c (TARGET_FUNCTION_ARG_PADDING): Redefine.
17364         (pa_function_arg_padding): Make static.  Return pad_direction instead
17365         of direction.
17366         * config/powerpcspe/powerpcspe.h (FUNCTION_ARG_PADDING): Delete.
17367         (PAD_VARARGS_DOWN): Use targetm.calls.function_arg_padding.
17368         * config/powerpcspe/aix.h (BLOCK_REG_PADDING): Use pad_direction
17369         instead of direction.  Use targetm.calls.function_arg_padding.
17370         * config/powerpcspe/darwin.h (BLOCK_REG_PADDING): Likewise.
17371         * config/powerpcspe/freebsd64.h (BLOCK_REG_PADDING): Likewise.
17372         * config/powerpcspe/linux64.h (BLOCK_REG_PADDING): Likewise.
17373         * config/powerpcspe/powerpcspe-protos.h (function_arg_padding): Delete.
17374         * config/powerpcspe/powerpcspe.c (TARGET_FUNCTION_ARG_PADDING):
17375         Redefine.
17376         (function_arg_padding): Rename to...
17377         (rs6000_function_arg_padding): ...this.  Make static.  Return
17378         pad_direction instead of direction.
17379         (rs6000_return_in_msb): Use rs6000_function_arg_padding.
17380         * config/rs6000/rs6000.h (FUNCTION_ARG_PADDING): Delete.
17381         (PAD_VARARGS_DOWN): Use targetm.calls.function_arg_padding.
17382         * config/rs6000/aix.h (BLOCK_REG_PADDING): Use pad_direction
17383         instead of direction.  Use targetm.calls.function_arg_padding.
17384         * config/rs6000/darwin.h (BLOCK_REG_PADDING): Likewise.
17385         * config/rs6000/freebsd64.h (BLOCK_REG_PADDING): Likewise.
17386         * config/rs6000/linux64.h (BLOCK_REG_PADDING): Likewise.
17387         * config/rs6000/rs6000-protos.h (function_arg_padding): Delete.
17388         * config/rs6000/rs6000.c (TARGET_FUNCTION_ARG_PADDING): Redefine.
17389         (function_arg_padding): Rename to...
17390         (rs6000_function_arg_padding): ...this.  Make static.  Return
17391         pad_direction instead of direction.
17392         (rs6000_return_in_msb): Use rs6000_function_arg_padding.
17393         * config/s390/s390.h (FUNCTION_ARG_PADDING): Delete.
17394         * config/s390/s390.c (s390_function_arg_padding): New function.
17395         (TARGET_FUNCTION_ARG_PADDING): Redefine.
17396         * config/sparc/sparc.h (FUNCTION_ARG_PADDING): Delete.
17397         * config/sparc/sparc-protos.h (function_arg_padding): Delete.
17398         * config/sparc/sparc.c (TARGET_FUNCTION_ARG_PADDING): Redefine.
17399         (function_arg_padding): Rename to...
17400         (sparc_function_arg_padding): ...this.  Make static.  Return
17401         pad_direction instead of direction.
17402         * config/spu/spu.h (FUNCTION_ARG_PADDING): Delete.
17403         * config/spu/spu.c (spu_function_arg_padding): New function.
17404         (TARGET_FUNCTION_ARG_PADDING): Redefine.
17405         * system.h (FUNCTION_ARG_PADDING): Poison.
17407 2017-09-04  Richard Sandiford  <richard.sandiford@linaro.org>
17408             Alan Hayward  <alan.hayward@arm.com>
17409             David Sherwood  <david.sherwood@arm.com>
17411         * target.def (modes_tieable_p): New hook.
17412         * doc/tm.texi (MODES_TIEABLE_P): Replace with...
17413         (TARGET_MODES_TIEABLE_P): ...this.
17414         * doc/tm.texi.in: Regenerate.
17415         * hooks.h (hook_bool_mode_mode_true): Declare.
17416         * hooks.c (hook_bool_mode_mode_true): New function.
17417         * combine.c (subst): Use targetm.modes_tieable_p instead of
17418         MODES_TIEABLE_P.
17419         * dse.c (find_shift_sequence): Likewise.
17420         * expmed.c (extract_low_bits): Likewise.
17421         * lower-subreg.c: Include target.h.
17422         (find_decomposable_subregs): Use targetm.modes_tieable_p instead of
17423         MODES_TIEABLE_P.
17424         * rtlanal.c (rtx_cost): Likewise.
17425         * config/aarch64/aarch64.h (MODES_TIEABLE_P): Delete.
17426         * config/aarch64/aarch64-protos.h (aarch64_modes_tieable_p): Delete.
17427         * config/aarch64/aarch64.c (aarch64_modes_tieable_p): Make static.
17428         (TARGET_MODES_TIEABLE_P): Redefine.
17429         * config/alpha/alpha.h (MODES_TIEABLE_P): Delete.
17430         * config/alpha/alpha.c (alpha_modes_tieable_p): New function.
17431         (TARGET_MODES_TIEABLE_P): Redefine.
17432         * config/arc/arc.h (MODES_TIEABLE_P): Delete.
17433         * config/arc/arc.c (TARGET_MODES_TIEABLE_P): Redefine.
17434         (arc_modes_tieable_p): New function.
17435         * config/arm/arm.h (MODES_TIEABLE_P): Delete.
17436         * config/arm/arm-protos.h (arm_modes_tieable_p): Delete.
17437         * config/arm/arm.c (TARGET_MODES_TIEABLE_P): Redefine.
17438         (arm_modes_tieable_p): Make static.
17439         * config/avr/avr.h (MODES_TIEABLE_P): Delete.
17440         * config/bfin/bfin.h (MODES_TIEABLE_P): Delete.
17441         * config/bfin/bfin.c (bfin_modes_tieable_p): New function.
17442         (TARGET_MODES_TIEABLE_P): Redefine.
17443         * config/c6x/c6x.h (MODES_TIEABLE_P): Delete.
17444         * config/c6x/c6x.c (c6x_modes_tieable_p): New function.
17445         (TARGET_MODES_TIEABLE_P): Redefine.
17446         * config/cr16/cr16.h (MODES_TIEABLE_P): Delete.
17447         * config/cr16/cr16.c (TARGET_MODES_TIEABLE_P): Redefine.
17448         (cr16_modes_tieable_p): New function.
17449         * config/cris/cris.h (MODES_TIEABLE_P): Delete.
17450         * config/epiphany/epiphany.h (MODES_TIEABLE_P): Delete.
17451         * config/fr30/fr30.h (MODES_TIEABLE_P): Delete.
17452         (TRULY_NOOP_TRUNCATION): Update comment.
17453         * config/frv/frv.h (MODES_TIEABLE_P): Delete.
17454         (TRULY_NOOP_TRUNCATION): Update comment.
17455         * config/frv/frv.c (TARGET_MODES_TIEABLE_P): Redefine.
17456         (frv_modes_tieable_p): New function.
17457         * config/ft32/ft32.h (MODES_TIEABLE_P): Delete.
17458         * config/h8300/h8300.h (MODES_TIEABLE_P): Delete.
17459         * config/h8300/h8300.c (h8300_modes_tieable_p): New function.
17460         (TARGET_MODES_TIEABLE_P): Redefine.
17461         * config/i386/i386.h (MODES_TIEABLE_P): Delete.
17462         * config/i386/i386-protos.h (ix86_modes_tieable_p): Delete.
17463         * config/i386/i386.c (ix86_modes_tieable_p): Make static.
17464         (TARGET_MODES_TIEABLE_P): Redefine.
17465         * config/ia64/ia64.h (MODES_TIEABLE_P): Delete.
17466         * config/ia64/ia64.c (TARGET_MODES_TIEABLE_P): Redefine.
17467         (ia64_modes_tieable_p): New function.
17468         * config/iq2000/iq2000.h (MODES_TIEABLE_P): Delete.
17469         * config/iq2000/iq2000.c (TARGET_MODES_TIEABLE_P): Redefine.
17470         (iq2000_modes_tieable_p): New function.
17471         * config/lm32/lm32.h (MODES_TIEABLE_P): Delete.
17472         * config/lm32/lm32.c (TARGET_MODES_TIEABLE_P): Redefine.
17473         (lm32_modes_tieable_p): New function.
17474         * config/m32c/m32c.h (MODES_TIEABLE_P): Delete.
17475         * config/m32c/m32c-protos.h (m32c_modes_tieable_p): Delete.
17476         * config/m32c/m32c.c (m32c_modes_tieable_p): Make static.
17477         (TARGET_MODES_TIEABLE_P): Redefine.
17478         * config/m32r/m32r.h (MODES_TIEABLE_P): Delete.
17479         * config/m32r/m32r.c (TARGET_MODES_TIEABLE_P): Redefine.
17480         (m32r_modes_tieable_p): New function.
17481         * config/m68k/m68k.h (MODES_TIEABLE_P): Delete.
17482         * config/m68k/m68k.c (TARGET_MODES_TIEABLE_P): Redefine.
17483         (m68k_modes_tieable_p): New function.
17484         * config/mcore/mcore.h (MODES_TIEABLE_P): Delete.
17485         * config/mcore/mcore.c (TARGET_MODES_TIEABLE_P): Redefine.
17486         (mcore_modes_tieable_p): New function.
17487         * config/microblaze/microblaze.h (MODES_TIEABLE_P): Delete.
17488         * config/microblaze/microblaze.c (microblaze_modes_tieable_p): New
17489         function.
17490         (TARGET_MODES_TIEABLE_P): Redefine.
17491         * config/mips/mips.h (MODES_TIEABLE_P): Delete.
17492         * config/mips/mips-protos.h (mips_modes_tieable_p): Delete.
17493         * config/mips/mips.c (mips_modes_tieable_p): Make static.
17494         (TARGET_MODES_TIEABLE_P): Redefine.
17495         * config/mmix/mmix.h (MODES_TIEABLE_P): Delete.
17496         * config/mn10300/mn10300.h (MODES_TIEABLE_P): Delete.
17497         * config/mn10300/mn10300-protos.h (mn10300_modes_tieable): Delete.
17498         * config/mn10300/mn10300.c (mn10300_modes_tieable): Rename to...
17499         (mn10300_modes_tieable_p): ...this and make static.
17500         (TARGET_MODES_TIEABLE_P): Redefine.
17501         * config/moxie/moxie.h (MODES_TIEABLE_P): Delete.
17502         * config/msp430/msp430.h (MODES_TIEABLE_P): Delete.
17503         * config/msp430/msp430-protos.h (msp430_modes_tieable_p): Delete.
17504         * config/msp430/msp430.c (TARGET_MODES_TIEABLE_P): Redefine.
17505         (msp430_modes_tieable_p): Make static.
17506         * config/nds32/nds32.h (MODES_TIEABLE_P): Delete.
17507         * config/nds32/nds32.c (nds32_modes_tieable_p): New function.
17508         (TARGET_MODES_TIEABLE_P): Redefine.
17509         * config/nios2/nios2.h (MODES_TIEABLE_P): Delete.
17510         * config/nvptx/nvptx.h (MODES_TIEABLE_P): Delete.
17511         * config/nvptx/nvptx.c (nvptx_modes_tieable_p): New function.
17512         (TARGET_MODES_TIEABLE_P): Redefine.
17513         * config/pa/pa.h (MODES_TIEABLE_P): Delete.
17514         * config/pa/pa-protos.h (pa_modes_tieable_p): Delete.
17515         * config/pa/pa.c (pa_modes_tieable_p): Make static.
17516         (TARGET_MODES_TIEABLE_P): Redefine.
17517         * config/pdp11/pdp11.h (MODES_TIEABLE_P): Delete.
17518         * config/pdp11/pdp11.c (TARGET_MODES_TIEABLE_P): Redefine.
17519         (pdp11_modes_tieable_p): New function.
17520         * config/powerpcspe/powerpcspe.h (MODES_TIEABLE_P): Delete.
17521         * config/powerpcspe/powerpcspe.c (TARGET_MODES_TIEABLE_P): Redefine.
17522         (rs6000_modes_tieable_p): New function.
17523         (rs6000_debug_reg_global): Use it instead of MODES_TIEABLE_P.
17524         * config/powerpcspe/powerpcspe.md: Update comment.
17525         * config/riscv/riscv.h (MODES_TIEABLE_P): Delete.
17526         * config/riscv/riscv.c (riscv_modes_tieable_p): New function.
17527         (TARGET_MODES_TIEABLE_P): Redefine.
17528         * config/rl78/rl78.h (MODES_TIEABLE_P): Delete.
17529         * config/rl78/rl78.c (TARGET_MODES_TIEABLE_P): Redefine.
17530         (rl78_modes_tieable_p): New function.
17531         * config/rs6000/rs6000.h (MODES_TIEABLE_P): Delete.
17532         * config/rs6000/rs6000.c (TARGET_MODES_TIEABLE_P): Redefine.
17533         (rs6000_modes_tieable_p): New function.
17534         (rs6000_debug_reg_global): Use it instead of MODES_TIEABLE_P.
17535         * config/rs6000/rs6000.md: Update comment.
17536         * config/rx/rx.h (MODES_TIEABLE_P): Delete.
17537         * config/rx/rx.c (rx_modes_tieable_p): New function.
17538         (TARGET_MODES_TIEABLE_P): Redefine.
17539         * config/s390/s390.h (MODES_TIEABLE_P): Delete.
17540         * config/s390/s390.c (s390_modes_tieable_p): New function.
17541         (TARGET_MODES_TIEABLE_P): Redefine.
17542         * config/sh/sh.h (MODES_TIEABLE_P): Delete.
17543         * config/sh/sh.c (TARGET_MODES_TIEABLE_P): Redefine.
17544         (sh_modes_tieable_p): New function.
17545         * config/sparc/sparc.h (MODES_TIEABLE_P): Delete.
17546         * config/sparc/sparc-protos.h (sparc_modes_tieable_p): Delete.
17547         * config/sparc/sparc.c (TARGET_MODES_TIEABLE_P): Redefine.
17548         (sparc_modes_tieable_p): Make static.
17549         * config/spu/spu.h (MODES_TIEABLE_P): Delete.
17550         * config/spu/spu.c (spu_modes_tieable_p): New function.
17551         (TARGET_MODES_TIEABLE_P): Redefine.
17552         * config/stormy16/stormy16.h (MODES_TIEABLE_P): Delete.
17553         * config/stormy16/stormy16.c (xstormy16_modes_tieable_p): New function.
17554         (TARGET_MODES_TIEABLE_P): Redefine.
17555         * config/tilegx/tilegx.h (MODES_TIEABLE_P): Delete.
17556         * config/tilepro/tilepro.h (MODES_TIEABLE_P): Delete.
17557         * config/v850/v850.h (MODES_TIEABLE_P): Delete.
17558         * config/v850/v850.c (v850_modes_tieable_p): New function.
17559         (TARGET_MODES_TIEABLE_P): Redefine.
17560         * config/vax/vax.h (MODES_TIEABLE_P): Delete.
17561         * config/visium/visium.h (MODES_TIEABLE_P): Delete.
17562         * config/visium/visium.c (TARGET_MODES_TIEABLE_P): Redefine.
17563         (visium_modes_tieable_p): New function.
17564         * config/xtensa/xtensa.h (MODES_TIEABLE_P): Delete.
17565         * config/xtensa/xtensa.c (TARGET_MODES_TIEABLE_P): Redefine.
17566         (xtensa_modes_tieable_p): New function.
17567         * system.h (MODES_TIEABLE_P): Poison.
17569 2017-09-04  Richard Sandiford  <richard.sandiford@linaro.org>
17570             Alan Hayward  <alan.hayward@arm.com>
17571             David Sherwood  <david.sherwood@arm.com>
17573         * target.def (hard_regno_mode_ok): New hook.
17574         * doc/tm.texi (HARD_REGNO_MODE_OK): Replace with...
17575         (TARGET_HARD_REGNO_MODE_OK): ...this.
17576         * doc/tm.texi.in: Regenerate.
17577         * hooks.h (hook_bool_uint_mode_true): Declare.
17578         * hooks.c (hook_bool_uint_mode_true): New function.
17579         * doc/md.texi: Refer to targetm.hard_regno_mode_ok instead of
17580         HARD_REGNO_MODE_OK.
17581         * genpreds.c (write_insn_preds_c): Add an include of target.h.
17582         * alias.c (init_alias_target): Use targetm.hard_regno_mode_ok
17583         instead of HARD_REGNO_MODE_OK.
17584         * caller-save.c: Include target.h.
17585         (reg_save_code): Use targetm.hard_regno_mode_ok instead of
17586         HARD_REGNO_MODE_OK.
17587         * combine.c (can_combine_p): Likewise.
17588         (combinable_i3pat): Likewise.
17589         (can_change_dest_mode): Likewise.
17590         * expr.c (init_expr_target): Likewise.
17591         (convert_move): Likewise.
17592         (convert_modes): Likewise.
17593         * ira.c (setup_prohibited_class_mode_regs): Likewise.
17594         (setup_prohibited_mode_move_regs): Likewise.
17595         * ira.h (target_ira): Likewise.
17596         * lra-assigns.c (find_hard_regno_for_1): Likewise.
17597         * lra-constraints.c (process_alt_operands): Likewise.
17598         (split_reg): Likewise.
17599         * recog.c (peep2_find_free_register): Likewise.
17600         * ree.c (combine_reaching_defs): Likewise.
17601         * regcprop.c (maybe_mode_change): Likewise.
17602         * reginfo.c (init_reg_sets_1): Likewise.
17603         (choose_hard_reg_mode): Likewise.
17604         (simplifiable_subregs): Likewise.
17605         * regrename.c (check_new_reg_p): Likewise.
17606         * reload.c (find_valid_class): Likewise.
17607         (find_valid_class_1): Likewise.
17608         (reload_inner_reg_of_subreg): Likewise.
17609         (push_reload): Likewise.
17610         (combine_reloads): Likewise.
17611         (find_dummy_reload): Likewise.
17612         (find_reloads): Likewise.
17613         * reload1.c (find_reg): Likewise.
17614         (set_reload_reg): Likewise.
17615         (allocate_reload_reg): Likewise.
17616         (choose_reload_regs): Likewise.
17617         (reload_adjust_reg_for_temp): Likewise.
17618         * rtlanal.c (subreg_size_offset_from_lsb): Likewise.
17619         (simplify_subreg_regno): Likewise.
17620         * sel-sched.c (init_regs_for_mode): Likewise.
17621         * varasm.c (make_decl_rtl): Likewise.
17622         * config/aarch64/aarch64.h (HARD_REGNO_MODE_OK): Delete.
17623         (MODES_TIEABLE_P): Use targetm.hard_regno_mode_ok instead of
17624         HARD_REGNO_MODE_OK.
17625         * config/aarch64/aarch64-protos.h (aarch64_hard_regno_mode_ok): Delete.
17626         * config/aarch64/aarch64.c (aarch64_hard_regno_mode_ok): Make static.
17627         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17628         * config/alpha/alpha.h (HARD_REGNO_MODE_OK): Delete.
17629         * config/alpha/alpha.c (alpha_hard_regno_mode_ok): New function.
17630         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17631         * config/arc/arc.h (arc_hard_regno_mode_ok): Delete.
17632         (arc_mode_class): Delete.
17633         (HARD_REGNO_MODE_OK): Delete.
17634         * config/arc/arc.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17635         (arc_hard_regno_mode_ok): Rename old array to...
17636         (arc_hard_regno_mode_ok_modes): ...this.
17637         (arc_conditional_register_usage): Update accordingly.
17638         (arc_mode_class): Make static.
17639         (arc_hard_regno_mode_ok): New function.
17640         * config/arm/arm.h (HARD_REGNO_MODE_OK): Delete.
17641         * config/arm/arm-protos.h (arm_hard_regno_mode_ok): Delete.
17642         * config/arm/arm.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17643         (arm_hard_regno_mode_ok): Make static.
17644         * config/arm/arm.md (movdi): Use targetm.hard_regno_mode_ok instead of
17645         HARD_REGNO_MODE_OK.
17646         * config/avr/avr-protos.h (avr_hard_regno_mode_ok): Delete.
17647         * config/avr/avr.h (HARD_REGNO_MODE_OK): Delete.
17648         * config/avr/avr.c (avr_hard_regno_mode_ok): Make static and
17649         return a bool.
17650         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17651         * config/bfin/bfin-protos.h (hard_regno_mode_ok): Delete.
17652         * config/bfin/bfin.h (HARD_REGNO_MODE_OK): Delete.
17653         * config/bfin/bfin.c (hard_regno_mode_ok): Rename to...
17654         (bfin_hard_regno_mode_ok): ...this.  Make static and return a bool.
17655         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17656         * config/bfin/predicates.md (valid_reg_operand): Use
17657         targetm.hard_regno_mode_ok instead of HARD_REGNO_MODE_OK.
17658         * config/c6x/c6x.h (HARD_REGNO_MODE_OK): Delete.
17659         * config/c6x/c6x.c (c6x_hard_regno_mode_ok): New function.
17660         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17661         * config/cr16/cr16.h (HARD_REGNO_MODE_OK): Delete.
17662         * config/cr16/cr16-protos.h (cr16_hard_regno_mode_ok): Delete.
17663         * config/cr16/cr16.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17664         (cr16_hard_regno_mode_ok): Make static and return a bool.
17665         * config/cris/cris.h (HARD_REGNO_MODE_OK): Delete.
17666         * config/cris/cris.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17667         (cris_hard_regno_mode_ok): New function.
17668         * config/epiphany/epiphany.h (epiphany_hard_regno_mode_ok): Delete.
17669         (epiphany_mode_class): Delete.
17670         (HARD_REGNO_MODE_OK): Delete.
17671         * config/epiphany/epiphany-protos.h (hard_regno_mode_ok): Delete.
17672         * config/epiphany/epiphany.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17673         (hard_regno_mode_ok): Rename to...
17674         (epiphany_hard_regno_mode_ok): ...this.  Make static and return a bool.
17675         * config/fr30/fr30.h (HARD_REGNO_MODE_OK): Delete.
17676         * config/fr30/fr30.md: Refer to targetm.hard_regno_mode_ok instead of
17677         HARD_REGNO_MODE_OK.
17678         * config/frv/frv.h (HARD_REGNO_MODE_OK): Delete.
17679         * config/frv/frv-protos.h (frv_hard_regno_mode_ok): Delete.
17680         * config/frv/frv.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17681         (frv_hard_regno_mode_ok): Make static and return a bool.
17682         * config/frv/frv.md: Refer to targetm.hard_regno_mode_ok instead of
17683         HARD_REGNO_MODE_OK.
17684         * config/ft32/ft32.h (HARD_REGNO_MODE_OK): Delete.
17685         * config/h8300/h8300.h (HARD_REGNO_MODE_OK): Delete.
17686         * config/h8300/h8300-protos.h (h8300_hard_regno_mode_ok): Delete.
17687         * config/h8300/h8300.c (h8300_hard_regno_mode_ok): Make static
17688         and return a bool.
17689         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17690         * config/i386/i386.h (HARD_REGNO_MODE_OK): Delete.
17691         * config/i386/i386-protos.h (ix86_hard_regno_mode_ok): Delete.
17692         * config/i386/i386.c (ix86_hard_regno_mode_ok): Make static and
17693         return a bool.
17694         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17695         * config/ia64/ia64.h (HARD_REGNO_MODE_OK): Delete.
17696         * config/ia64/ia64.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17697         (ia64_hard_regno_mode_ok): New function.
17698         * config/iq2000/iq2000.h (HARD_REGNO_MODE_OK): Delete.
17699         * config/iq2000/iq2000.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17700         (iq2000_hard_regno_mode_ok): New function.
17701         * config/lm32/lm32.h (HARD_REGNO_MODE_OK): Delete.
17702         * config/lm32/lm32.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17703         (lm32_hard_regno_mode_ok): New function.
17704         * config/m32c/m32c.h (HARD_REGNO_MODE_OK): Delete.
17705         * config/m32c/m32c-protos.h (m32c_hard_regno_ok): Delete.
17706         * config/m32c/m32c.c (class_can_hold_mode): Use m32c_hard_regno_mode_ok
17707         instead of HARD_REGNO_MODE_OK.
17708         (m32c_hard_regno_ok): Rename to...
17709         (m32c_hard_regno_mode_ok): ...this.  Make static and return a bool.
17710         (m32c_cannot_change_mode_class): Update accordingly.
17711         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17712         * config/m32r/m32r.h (m32r_hard_regno_mode_ok): Delete.
17713         (m32r_mode_class): Delete.
17714         (HARD_REGNO_MODE_OK): Delete.
17715         * config/m32r/m32r.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17716         (m32r_hard_regno_mode_ok): Rename to...
17717         (m32r_hard_regno_modes): ...this.
17718         (m32r_mode_class): Make static.
17719         (m32r_hard_regno_mode_ok): New function.
17720         * config/m68k/m68k.h (HARD_REGNO_MODE_OK): Delete.
17721         * config/m68k/m68k-protos.h (m68k_regno_mode_ok): Delete.
17722         * config/m68k/m68k.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17723         (m68k_hard_regno_mode_ok): Make static.
17724         * config/mcore/mcore.h (HARD_REGNO_MODE_OK): Delete.
17725         * config/mcore/mcore.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17726         (mcore_hard_regno_mode_ok): New function.
17727         * config/microblaze/microblaze.h (microblaze_hard_regno_mode_ok)
17728         (HARD_REGNO_MODE_OK): Delete.
17729         * config/microblaze/microblaze.c (microblaze_hard_regno_mode_ok):
17730         Rename to...
17731         (microblaze_hard_regno_mode_ok_p): ...this and make static.
17732         (microblaze_hard_regno_mode_ok): New function.
17733         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17734         * config/mips/mips.h (HARD_REGNO_MODE_OK): Delete.
17735         (mips_hard_regno_mode_ok): Delete.
17736         * config/mips/mips.c (mips_hard_regno_mode_ok): Rename to...
17737         (mips_hard_regno_mode_ok_p): ...this and make static.
17738         (mips_hard_regno_mode_ok_p): Rename to...
17739         (mips_hard_regno_mode_ok_uncached): ...this.
17740         (mips_hard_regno_mode_ok): New function.
17741         (mips_class_max_nregs): Use mips_hard_regno_mode_ok instead
17742         of HARD_REGNO_MODE_OK.
17743         (mips_option_override): Update after above name changes.
17744         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17745         * config/mmix/mmix.h (HARD_REGNO_MODE_OK): Delete.
17746         * config/mn10300/mn10300.h (HARD_REGNO_MODE_OK): Delete.
17747         * config/mn10300/mn10300-protos.h (mn10300_hard_regno_mode_ok): Delete.
17748         * config/mn10300/mn10300.c (mn10300_hard_regno_mode_ok): Make static.
17749         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17750         * config/moxie/moxie.h (HARD_REGNO_MODE_OK): Delete.
17751         * config/msp430/msp430.h (HARD_REGNO_MODE_OK): Delete.
17752         * config/msp430/msp430-protos.h (msp430_hard_regno_mode_ok): Delete.
17753         * config/msp430/msp430.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17754         (msp430_hard_regno_mode_ok): Make static and return a bool.
17755         * config/nds32/nds32.h (HARD_REGNO_MODE_OK): Delete.
17756         * config/nds32/nds32-protos.h (nds32_hard_regno_mode_ok): Delete.
17757         * config/nds32/nds32.c (nds32_hard_regno_mode_ok): Make static
17758         and return a bool.
17759         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17760         * config/nios2/nios2.h (HARD_REGNO_MODE_OK): Delete.
17761         * config/nvptx/nvptx.h (HARD_REGNO_MODE_OK): Delete.
17762         * config/pa/pa.h (MODES_TIEABLE_P): Update commentary.
17763         * config/pa/pa32-regs.h (HARD_REGNO_MODE_OK): Rename to...
17764         (PA_HARD_REGNO_MODE_OK): ...this
17765         * config/pa/pa64-regs.h (HARD_REGNO_MODE_OK): Rename to...
17766         (PA_HARD_REGNO_MODE_OK): ...this.
17767         * config/pa/pa.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17768         (pa_hard_regno_mode_ok): New function.
17769         * config/pdp11/pdp11.h (HARD_REGNO_MODE_OK): Delete.
17770         * config/pdp11/pdp11.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17771         (pdp11_hard_regno_mode_ok): New function.
17772         * config/powerpcspe/powerpcspe.h (HARD_REGNO_MODE_OK): Delete.
17773         * config/powerpcspe/powerpcspe-protos.h (rs6000_hard_regno_mode_ok_p):
17774         Delete.
17775         * config/powerpcspe/powerpcspe.c (rs6000_hard_regno_mode_ok_p):
17776         Make static.
17777         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17778         (rs6000_hard_regno_mode_ok): Rename to...
17779         (rs6000_hard_regno_mode_ok_uncached): ...this.
17780         (rs6000_init_hard_regno_mode_ok): Update accordingly.
17781         (rs6000_hard_regno_mode_ok): New function.
17782         * config/riscv/riscv.h (HARD_REGNO_MODE_OK): Delete.
17783         * config/riscv/riscv-protos.h (riscv_hard_regno_mode_ok_p): Delete.
17784         * config/riscv/riscv.c (riscv_hard_regno_mode_ok_p): Rename to...
17785         (riscv_hard_regno_mode_ok): ...this and make static.
17786         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17787         * config/rl78/rl78.h (HARD_REGNO_MODE_OK): Delete.
17788         * config/rl78/rl78-protos.h (rl78_hard_regno_mode_ok): Delete.
17789         * config/rl78/rl78.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17790         (rl78_hard_regno_mode_ok): Make static and return bool.
17791         * config/rs6000/rs6000.h (HARD_REGNO_MODE_OK): Delete.
17792         * config/rs6000/rs6000-protos.h (rs6000_hard_regno_mode_ok_p):
17793         Delete.
17794         * config/rs6000/rs6000.c (rs6000_hard_regno_mode_ok_p): Make static.
17795         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17796         (rs6000_hard_regno_mode_ok): Rename to...
17797         (rs6000_hard_regno_mode_ok_uncached): ...this.
17798         (rs6000_init_hard_regno_mode_ok): Update accordingly.
17799         (rs6000_hard_regno_mode_ok): New function.
17800         * config/rx/rx.h (HARD_REGNO_MODE_OK): Delete.
17801         * config/rx/rx.c (rx_hard_regno_mode_ok): New function.
17802         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17803         * config/s390/s390.h (HARD_REGNO_MODE_OK): Delete.
17804         * config/s390/s390-protos.h (s390_hard_regno_mode_ok): Delete.
17805         * config/s390/s390.c (s390_hard_regno_mode_ok): Make static.
17806         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17807         * config/sh/sh.h (HARD_REGNO_MODE_OK): Delete.
17808         * config/sh/sh-protos.h (sh_hard_regno_mode_ok): Delete.
17809         * config/sh/sh.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17810         (sh_hard_regno_mode_ok): Make static.
17811         * config/sparc/constraints.md: Refer to targetm.hard_regno_mode_ok
17812         instead of HARD_REGNO_MODE_OK.
17813         * config/sparc/sparc.h (hard_regno_mode_classes): Delete.
17814         (sparc_mode_class): Delete.
17815         (HARD_REGNO_MODE_OK): Delete.
17816         * config/sparc/sparc.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17817         (hard_regno_mode_classes): Make static.
17818         (sparc_mode_class): Likewise.
17819         (sparc_hard_regno_mode_ok): New function.
17820         * config/spu/spu.h (HARD_REGNO_MODE_OK): Delete.
17821         * config/stormy16/stormy16.h (HARD_REGNO_MODE_OK): Delete.
17822         * config/stormy16/stormy16.c (xstormy16_hard_regno_mode_ok): New
17823         function.
17824         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17825         * config/tilegx/tilegx.h (HARD_REGNO_MODE_OK): Delete.
17826         * config/tilepro/tilepro.h (HARD_REGNO_MODE_OK): Delete.
17827         * config/v850/v850.h (HARD_REGNO_MODE_OK): Delete.
17828         * config/v850/v850.c (v850_hard_regno_mode_ok): New function.
17829         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17830         * config/vax/vax.h (HARD_REGNO_MODE_OK): Delete.
17831         * config/visium/visium.h (HARD_REGNO_MODE_OK): Delete.
17832         * config/visium/visium.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17833         (visium_hard_regno_mode_ok): New function.
17834         * config/visium/visium.md: Refer to targetm.hard_regno_mode_ok
17835         instead of HARD_REGNO_MODE_OK.
17836         * config/xtensa/xtensa.h (xtensa_hard_regno_mode_ok): Delete.
17837         (HARD_REGNO_MODE_OK): Delete.
17838         * config/xtensa/xtensa.c (xtensa_hard_regno_mode_ok): Rename to...
17839         (xtensa_hard_regno_mode_ok_p): ...this and make static.
17840         (xtensa_option_override): Update accordingly.
17841         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17842         (xtensa_hard_regno_mode_ok): New function.
17843         * system.h (HARD_REGNO_MODE_OK): Poison.
17845 2017-09-04  Richard Sandiford  <richard.sandiford@linaro.org>
17846             Alan Hayward  <alan.hayward@arm.com>
17847             David Sherwood  <david.sherwood@arm.com>
17849         * target.def (hard_regno_call_part_clobbered): New hook.
17850         * doc/tm.texi.in (HARD_REGNO_CALL_PART_CLOBBERED): Replace with...
17851         (TARGET_HARD_REGNO_CALL_PART_CLOBBERED): ...this hook.
17852         * doc/tm.texi: Regenerate.
17853         * hooks.h (hook_bool_uint_mode_false): Declare.
17854         * hooks.c (hook_bool_uint_mode_false): New function.
17855         * regs.h (HARD_REGNO_CALL_PART_CLOBBERED): Delete.
17856         * cselib.c (cselib_process_insn): Use
17857         targetm.hard_regno_call_part_clobbered instead of
17858         HARD_REGNO_CALL_PART_CLOBBERED.
17859         * ira-conflicts.c (ira_build_conflicts): Likewise.
17860         * ira-costs.c (ira_tune_allocno_costs): Likewise.
17861         * lra-constraints.c (need_for_call_save_p): Likewise.
17862         * lra-lives.c: Include target.h.
17863         (check_pseudos_live_through_calls): Use
17864         targetm.hard_regno_call_part_clobbered instead of
17865         HARD_REGNO_CALL_PART_CLOBBERED.
17866         * regcprop.c: Include target.h.
17867         (copyprop_hardreg_forward_1): Use
17868         targetm.hard_regno_call_part_clobbered instead of
17869         HARD_REGNO_CALL_PART_CLOBBERED.
17870         * reginfo.c (choose_hard_reg_mode): Likewise.
17871         * regrename.c (check_new_reg_p): Likewise.
17872         * reload.c (find_equiv_reg): Likewise.
17873         * reload1.c (emit_reload_insns): Likewise.
17874         * sched-deps.c (deps_analyze_insn): Likewise.
17875         * sel-sched.c (init_regs_for_mode): Likewise.
17876         (mark_unavailable_hard_regs): Likewise.
17877         * targhooks.c (default_dwarf_frame_reg_mode): Likewise.
17878         * config/aarch64/aarch64.h (HARD_REGNO_CALL_PART_CLOBBERED): Delete.
17879         * config/aarch64/aarch64.c (aarch64_hard_regno_call_part_clobbered):
17880         New function.
17881         (TARGET_HARD_REGNO_CALL_PART_CLOBBERED): Redefine.
17882         * config/avr/avr.h (HARD_REGNO_CALL_PART_CLOBBERED): Delete.
17883         * config/avr/avr-protos.h (avr_hard_regno_call_part_clobbered):
17884         Delete.
17885         * config/avr/avr.c (avr_hard_regno_call_part_clobbered): Make static
17886         and return a bool.
17887         (TARGET_HARD_REGNO_CALL_PART_CLOBBERED): Redefine.
17888         * config/i386/i386.h (HARD_REGNO_CALL_PART_CLOBBERED): Delete.
17889         * config/i386/i386.c (ix86_hard_regno_call_part_clobbered): New
17890         function.
17891         (TARGET_HARD_REGNO_CALL_PART_CLOBBERED): Redefine.
17892         * config/mips/mips.h (HARD_REGNO_CALL_PART_CLOBBERED): Delete.
17893         * config/mips/mips.c (mips_hard_regno_call_part_clobbered): New
17894         function.
17895         (TARGET_HARD_REGNO_CALL_PART_CLOBBERED): Redefine.
17896         * config/powerpcspe/powerpcspe.h (HARD_REGNO_CALL_PART_CLOBBERED):
17897         Delete.
17898         * config/powerpcspe/powerpcspe.c
17899         (rs6000_hard_regno_call_part_clobbered): New function.
17900         (TARGET_HARD_REGNO_CALL_PART_CLOBBERED): Redefine.
17901         * config/rs6000/rs6000.h (HARD_REGNO_CALL_PART_CLOBBERED): Delete.
17902         * config/rs6000/rs6000.c (rs6000_hard_regno_call_part_clobbered):
17903         New function.
17904         (TARGET_HARD_REGNO_CALL_PART_CLOBBERED): Redefine.
17905         * config/s390/s390.h (HARD_REGNO_CALL_PART_CLOBBERED): Delete.
17906         * config/s390/s390.c (s390_hard_regno_call_part_clobbered): New
17907         function.
17908         (TARGET_HARD_REGNO_CALL_PART_CLOBBERED): Redefine.
17909         * config/sh/sh.h (HARD_REGNO_CALL_PART_CLOBBERED): Delete.
17910         * system.h (HARD_REGNO_CALL_PART_CLOBBERED): Poison.
17912 2017-09-04  Richard Sandiford  <richard.sandiford@linaro.org>
17913             Alan Hayward  <alan.hayward@arm.com>
17914             David Sherwood  <david.sherwood@arm.com>
17916         * rtl.h (subreg_memory_offset): Declare.
17917         * emit-rtl.c (subreg_memory_offset): New function.
17918         * expmed.c (store_bit_field_1): Use it.
17919         * expr.c (undefined_operand_subword_p): Likewise.
17920         * simplify-rtx.c (simplify_subreg): Likewise.
17922 2017-09-04  Alexander Monakov  <amonakov@ispras.ru>
17924         PR rtl-optimization/57448
17925         PR target/67458
17926         PR target/81316
17927         * optabs.c (expand_atomic_load): Place compiler memory barriers if
17928         using atomic_load pattern.
17929         (expand_atomic_store): Likewise.
17931 2017-09-04  Jakub Jelinek  <jakub@redhat.com>
17933         PR sanitizer/81981
17934         * gimple-fold.c (gimple_fold_call): Optimize away useless UBSAN_PTR
17935         and UBSAN_BOUNDS internal calls.  Clean up IFN_UBSAN_OBJECT_SIZE
17936         handling.  Use replace_call_with_value with NULL instead of
17937         gsi_replace, unlink_stmt_vdef and release_defs.
17939         * gdbhooks.py (OptMachineModePrinter.to_string): Use 8 spaces
17940         instead of tab.
17942         * lra-remat.c (reg_overlap_for_remat_p): Fix a pasto.
17944 2017-09-04  Richard Sandiford  <richard.sandiford@linaro.org>
17946         PR bootstrap/82045
17947         * rtl.h (emit_library_call_value_1): Declare.
17948         (emit_library_call): Replace declaration with a series of overloads.
17949         Remove the parameter count argument.
17950         (emit_library_call_value): Likewise.
17951         * calls.c (emit_library_call_value_1): Make global.  Replace varargs
17952         with an "rtx_mode_t *".
17953         (emit_library_call_value): Delete.
17954         (emit_library_call): Likewise.
17955         * asan.c (asan_emit_stack_protection): Update calls accordingly.
17956         (asan_emit_allocas_unpoison): Likewise.
17957         * builtins.c (expand_builtin_powi): Likewise.
17958         (expand_asan_emit_allocas_unpoison): Likewise.
17959         * cfgexpand.c (expand_main_function): Likewise.
17960         * config/aarch64/aarch64.c (aarch64_trampoline_init): Likewise.
17961         * config/aarch64/aarch64.h (PROFILE_HOOK): Likewise.
17962         * config/alpha/alpha.c (alpha_trampoline_init): Likewise.
17963         * config/arm/arm.c (arm_trampoline_init): Likewise.
17964         (arm_call_tls_get_addr): Likewise.
17965         (arm_expand_divmod_libfunc): Likewise.
17966         * config/bfin/bfin.md (umulsi3_highpart): Likewise.
17967         (smulsi3_highpart): Likewise.
17968         * config/c6x/c6x.c (c6x_initialize_trampoline): Likewise.
17969         (c6x_expand_compare): Likewise.
17970         (c6x_expand_movmem): Likewise.
17971         * config/frv/frv.c (frv_trampoline_init): Likewise.
17972         * config/i386/i386.c (ix86_trampoline_init): Likewise.
17973         (ix86_expand_divmod_libfunc): Likewise.
17974         * config/ia64/ia64.c (ia64_expand_tls_address): Likewise.
17975         (ia64_expand_compare): Likewise.
17976         (ia64_profile_hook): Likewise.
17977         * config/ia64/ia64.md (save_stack_nonlocal): Likewise.
17978         (nonlocal_goto): Likewise.
17979         (restore_stack_nonlocal): Likewise.
17980         * config/m32r/m32r.c (block_move_call): Likewise.
17981         (m32r_trampoline_init): Likewise.
17982         * config/m68k/linux.h (FINALIZE_TRAMPOLINE): Likewise.
17983         * config/m68k/m68k.c (m68k_call_tls_get_addr): Likewise.
17984         (m68k_call_m68k_read_tp): Likewise.
17985         * config/microblaze/microblaze.c (microblaze_call_tls_get_addr)
17986         (microblaze_expand_divide): Likewise.
17987         * config/mips/mips.h (mips_args): Likewise.
17988         * config/mips/sdemtk.h (mips_sync_icache): Likewise.
17989         (MIPS_ICACHE_SYNC): Likewise.
17990         * config/nios2/nios2.c (nios2_emit_expensive_div): Likewise.
17991         (nios2_trampoline_init): Likewise.
17992         * config/pa/pa.c (hppa_tls_call): Likewise.
17993         (pa_trampoline_init): Likewise.
17994         * config/pa/pa.md (canonicalize_funcptr_for_compare): Likewise.
17995         * config/powerpcspe/powerpcspe.c (rs6000_legitimize_tls_address)
17996         (expand_strn_compare): Likewise.
17997         (rs6000_generate_compare): Likewise.
17998         (rs6000_expand_float128_convert): Likewise.
17999         (output_profile_hook): Likewise.
18000         (rs6000_trampoline_init): Likewise.
18001         * config/powerpcspe/powerpcspe.md (neg<mode>2): Likewise.
18002         * config/riscv/riscv.h (PROFILE_HOOK): Likewise.
18003         * config/rs6000/rs6000-string.c (expand_strn_compare): Likewise.
18004         * config/rs6000/rs6000.c (rs6000_legitimize_tls_address): Likewise.
18005         (rs6000_generate_compare): Likewise.
18006         (rs6000_expand_float128_convert): Likewise.
18007         (output_profile_hook): Likewise.
18008         (rs6000_trampoline_init): Likewise.
18009         * config/rs6000/rs6000.md (neg<mode>2): Likewise.
18010         * config/sh/sh.c (sh_trampoline_init): Likewise.
18011         * config/sparc/sparc.c (emit_soft_tfmode_libcall): Likewise.
18012         (sparc_emit_float_lib_cmp): Likewise.
18013         (sparc32_initialize_trampoline): Likewise.
18014         (sparc64_initialize_trampoline): Likewise.
18015         (sparc_profile_hook): Likewise.
18016         * config/spu/spu.c (ea_load_store): Likewise.
18017         * config/spu/spu.md (floatunssidf2): Likewise.
18018         * config/tilegx/tilegx.c (tilegx_trampoline_init): Likewise.
18019         * config/tilepro/tilepro.c (tilepro_trampoline_init): Likewise.
18020         * config/visium/visium.c (expand_block_move_4): Likewise.
18021         (expand_block_move_2): Likewise.
18022         (expand_block_move_1): Likewise.
18023         (expand_block_set_4): Likewise.
18024         (expand_block_set_2): Likewise.
18025         (expand_block_set_1): Likewise.
18026         (visium_trampoline_init): Likewise.
18027         (visium_profile_hook): Likewise.
18028         * config/xtensa/xtensa.c (xtensa_expand_nonlocal_goto): Likewise.
18029         (xtensa_setup_frame_addresses): Likewise.
18030         (xtensa_trampoline_init): Likewise.
18031         * except.c (sjlj_emit_function_enter): Likewise.
18032         (sjlj_emit_function_exit): Likewise.
18033         * explow.c (allocate_dynamic_stack_space): Likewise.
18034         (probe_stack_range): Likewise.
18035         * expr.c (convert_mode_scalar): Likewise.
18036         * optabs.c (expand_binop): Likewise.
18037         (expand_twoval_binop_libfunc): Likewise.
18038         (expand_unop): Likewise.
18039         (prepare_cmp_insn): Likewise.
18040         (prepare_float_lib_cmp): Likewise.
18041         (expand_float): Likewise.
18042         (expand_fix): Likewise.
18043         (expand_fixed_convert): Likewise.
18044         (maybe_emit_sync_lock_test_and_set): Likewise.
18045         (expand_atomic_compare_and_swap): Likewise.
18046         (expand_mem_thread_fence): Likewise.
18047         (expand_atomic_fetch_op): Likewise.
18049 2017-09-03  Gerald Pfeifer  <gerald@pfeifer.com>
18051         * doc/generic.texi (OpenACC): Adjust URL.
18052         * doc/invoke.texi (C Dialect Options): Ditto.
18054 2017-09-03  Uros Bizjak  <ubizjak@gmail.com>
18056         * config/i386/i386.md (*bt<mode>): Use nonimmediate_operand
18057         predicate for operand 1.  Add (m,<S>) constraint.
18058         (*jcc_bt<mode>): Use nonimmediate_operand predicate for operand 1.
18059         Prevent memory operand 1 with register operand 2.
18061 2017-09-01  Segher Boessenkool  <segher@kernel.crashing.org>
18063         PR rtl-optimization/82024
18064         * combine.c (try_combine): If the combination result is a PARALLEL,
18065         and we only need to retain the SET in there that would be placed
18066         at I2, check that we can place that at I3 instead, before doing so.
18068 2017-09-01  Jakub Jelinek  <jakub@redhat.com>
18070         PR target/81766
18071         * config/i386/i386.c (ix86_init_large_pic_reg): Return label
18072         instead of void.
18073         (ix86_init_pic_reg): Remember label from ix86_init_large_pic_reg,
18074         if non-NULL and preceded by NOTE_INSN_BASIC_BLOCK, swap the note
18075         and label.
18077 2017-09-01  Joerg Sonnenberger  <joerg@bec.de>
18078             Jeff Law  <law@redhat.com>
18080         * varasm.c (bss_initializer_p): Do not put constants into .bss
18081         (categorize_decl_for_section): Handle bss_initializer_p returning
18082         false when DECL_INITIAL is NULL.
18084 2017-09-01  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
18086         PR target/82012
18087         * config/s390/s390.c (s390_can_inline_p): New function.
18089 2017-09-01  Jeff Law  <law@redhat.com>
18091         PR tree-optimization/82052
18092         * tree-ssa-scopedtables.c (avail_exprs_stack::lookup_avail_expr):
18093         Always initialize the returned slot after a hash table miss
18094         when INSERT is true.
18096 2017-09-01  Alexander Monakov  <amonakov@ispras.ru>
18098         * config/s390/s390.md (mem_signal_fence): Remove.
18099         * doc/md.texi (mem_signal_fence): Remove.
18100         * optabs.c (expand_mem_signal_fence): Remove uses of mem_signal_fence.
18101         Update comments.
18102         * target-insns.def (mem_signal_fence): Remove.
18104 2017-09-01  Jakub Jelinek  <jakub@redhat.com>
18106         PR sanitizer/81902
18107         * doc/invoke.texi: Document -fsanitize=pointer-overflow.
18109         PR sanitizer/81923
18110         * asan.c (create_odr_indicator): Strip name encoding from assembler
18111         name before appending it after __odr_asan_.
18113 2017-09-01  Martin Liska  <mliska@suse.cz>
18115         PR tree-optimization/82059
18116         * gimple-ssa-isolate-paths.c (isolate_path): Add profile and
18117         frequency only when an edge is redirected.
18119 2017-09-01  Claudiu Zissulescu  <claziss@synopsys.com>
18121         * config/arc/arc-c.c (__ARC_LPC_WIDTH__): Add builtin define.
18122         * config/arc/arc.c (ARC_MAX_LOOP_LENGTH): Define.
18123         (arc_conditional_register_usage): Remove ARC600 lp_count
18124         exception.
18125         (arc_file_start): Emit Tag_ARC_CPU_variation.
18126         (arc_can_use_doloop_p): New conditions to use ZOLs.
18127         (hwloop_fail): New function.
18128         (hwloop_optimize): Likewise.
18129         (hwloop_pattern_reg): Likewise.
18130         (arc_doloop_hooks): New struct, to be used with reorg_loops.
18131         (arc_reorg_loops): New function, calls reorg_loops.
18132         (arc_reorg): Call arc_reorg_loops.  Remove old ZOL handling.
18133         (arc600_corereg_hazard): Remove ZOL checking, case handled by
18134         hwloop_optimize.
18135         (arc_loop_hazard): Remove function, functionality moved into
18136         hwloop_optimize.
18137         (arc_hazard): Remove arc_loop_hazard call.
18138         (arc_adjust_insn_length): Remove ZOL handling, functionality moved
18139         into hwloop_optimize.
18140         (arc_label_align): Remove ZOL handling.
18141         * config/arc/arc.h (LOOP_ALIGN): Changed to 0.
18142         * config/arc/arc.md (doloop_begin): Remove pattern.
18143         (doloop_begin_i): Likewise.
18144         (doloop_end_i): Likewise.
18145         (doloop_fallback): Likewise.
18146         (doloop_fallback_m): Likewise.
18147         (doloop_end): Reimplement expand.
18148         (arc_lp): New pattern for LP instruction.
18149         (loop_end): New pattern.
18150         (loop_fail): Likewise.
18151         (decrement_and_branch_until_zero): Likewise.
18152         * config/arc/arc.opt (mlpc-width): New option.
18153         * doc/invoke.texi (mlpc-width): Document option.
18155 2017-09-01  Claudiu Zissulescu  <claziss@synopsys.com>
18157         * config/arc/arc.c (arc_ifcvt): Remove use of merge_blocks call.
18158         (arc_ccfsm_advance): Fix checking for delay slots.
18159         (arc_reorg): Add rtl dump after each call to arc_ifcvt.
18161 2017-09-01  Claudiu Zissulescu  <claziss@synopsys.com>
18163         * config/arc/arc.md (movqi_insn): Add stores to save constant long
18164         immediates.
18165         (movhi_insn): Update store instruction constraint which are saving
18166         6-bit short immediates.
18167         (movsi_insn): Consider also short scaled load operations.
18168         (zero_extendhisi2_i): Use Usd constraint instead of T.
18169         (extendhisi2_i): Add q constraint.
18170         (arc_clzsi2): Add type and length attributes.
18171         (arc_ctzsi2): Likewise.
18172         * config/arc/constraints.md (Usc): Update constraint, the
18173         assembler can parse two relocations for a single instruction.
18175 2017-09-01  Claudiu Zissulescu  <claziss@synopsys.com>
18177         * config/arc/arc.c (arc_use_anchors_for_symbol_p): New function.
18178         (TARGET_USE_ANCHORS_FOR_SYMBOL_P): Define.
18180 2017-08-31  Olivier Hainque  <hainque@adacore.com>
18182         * config.gcc (powerpc-wrs-vxworks|vxworksae|vxworksmils): Now
18183         match as powerpc-wrs-vxworks*.
18185 2017-08-31  James Greenhalgh  <james.greenhalgh@arm.com>
18187         * config/aarch64/aarch64-simd.md (aarch64_mla_elt_merge<mode>): Fix
18188         register constraint for by-element operand.
18189         (aarch64_mls_elt_merge<mode>): Likewise.
18191 2017-08-31  Claudiu Zissulescu  <claziss@synopsys.com>
18193         * config/arc/arc.c (arc_can_follow_jump): Check for short
18194         branches.
18196 2017-08-31  Claudiu Zissulescu  <claziss@synopsys.com>
18198         * config.gcc: Use g.opt for arc.
18199         * config/arc/arc.c (LEGITIMATE_SCALED_ADDRESS_P): Deleted,
18200         functionality moved to ...
18201         (legitimate_scaled_address_p): New function, ...here.
18202         (LEGITIMATE_SMALL_DATA_OFFSET_P): New define.
18203         (LEGITIMATE_SMALL_DATA_ADDRESS_P): Use the above define.
18204         (legitimate_offset_address_p): Delete TARGET_NO_SDATA_SET
18205         condition.
18206         (arc_override_options): Handle G option.
18207         (arc_output_pic_addr_const): Correct function definition.
18208         (arc_legitimate_address_p): Use legitimate_scaled_address_p.
18209         (arc_decl_anon_ns_mem_p): Delete.
18210         (arc_in_small_data_p): Overhaul this function to take into
18211         consideration the value given via G option.
18212         (arc_rewrite_small_data_1): Renamed and corrected old
18213         arc_rewrite_small_data function.
18214         (arc_rewrite_small_data): New function.
18215         (small_data_pattern): Don't use pic_offset_table_rtx.
18216         * config/arc/arc.h (CC1_SPEC): Recognize G option.
18217         * config/arc/simdext.md (movmisalignv2hi): Use
18218         prepare_move_operands function.
18219         (mov*): Likewise.
18220         (movmisalign*): Likewise.
18221         * doc/invoke.texi (ARC options): Document -G option.
18223 2017-08-31  Claudiu Zissulescu  <claziss@synopsys.com>
18225         * config/arc/arc-protos.h (compact_sda_memory_operand): Update
18226         prototype.
18227         * config/arc/arc.c (arc_print_operand): Output scalled address for
18228         sdata whenever is possible.
18229         (arc_in_small_data_p): Allow sdata for 64bit datum when double
18230         load/stores are available.
18231         (compact_sda_memory_operand): Check for the alignment required by
18232         code density instructions.
18233         * config/arc/arc.md (movsi_insn): Use newly introduced Us0
18234         constraint.
18235         * config/arc/constraints.md (Usd): Update constraint.
18236         (Us0): New constraint.
18237         (Usc): Update constraint.
18239 2017-08-31  Richard Biener  <rguenther@suse.de>
18241         PR middle-end/82054
18242         * dwarf2out.c (dwarf2out_early_global_decl): Process each
18243         function only once.
18245 2017-08-31  Tamar Christina  <tamar.christina@arm.com>
18247         * config/aarch64/aarch64-builtins.c (aarch64_init_simd_builtins):
18248         Resize type_signature.
18250 2017-08-31  Richard Sandiford  <richard.sandiford@linaro.org>
18251             Alan Hayward  <alan.hayward@arm.com>
18252             David Sherwood  <david.sherwood@arm.com>
18254         * config/aarch64/aarch64.c (aarch64_base_register_rtx_p): Only allow
18255         subregs whose inner modes can be stored in GPRs.
18256         (aarch64_classify_index): Likewise.
18258 2017-08-31  Richard Sandiford  <richard.sandiford@linaro.org>
18259             Alan Hayward  <alan.hayward@arm.com>
18260             David Sherwood  <david.sherwood@arm.com>
18262         * config/aarch64/iterators.md (V_cmp_result): Rename to...
18263         (V_INT_EQUIV): ...this.
18264         (v_cmp_result): Rename to...
18265         (v_int_equiv): ...this.
18266         * config/aarch64/aarch64.md (xorsign<mode>3): Update accordingly.
18267         * config/aarch64/aarch64-simd.md (xorsign<mode>3): Likewise.
18268         (copysign<mode>3): Likewise.
18269         (aarch64_simd_bsl<mode>_internal): Likewise.
18270         (aarch64_simd_bsl<mode>): Likewise.
18271         (vec_cmp<mode><mode>): Likewise.
18272         (vcond<mode><mode>): Likewise.
18273         (vcond<v_cmp_mixed><mode>): Likewise.
18274         (vcondu<mode><v_cmp_mixed>): Likewise.
18275         (aarch64_cm<optab><mode>): Likewise.
18276         (aarch64_cmtst<mode>): Likewise.
18277         (aarch64_fac<optab><mode>): Likewise.
18278         (vec_perm_const<mode>): Likewise.
18279         (vcond_mask_<mode><v_cmp_result>): Rename to...
18280         (vcond_mask_<mode><v_int_equiv>): ...this.
18281         (vec_cmp<mode><v_cmp_result>): Rename to...
18282         (vec_cmp<mode><v_int_equiv>): ...this.
18284 2017-08-31  Richard Sandiford  <richard.sandiford@linaro.org>
18285             Alan Hayward  <alan.hayward@arm.com>
18286             David Sherwood  <david.sherwood@arm.com>
18288         * config/aarch64/aarch64-modes.def: Remove 32-, 48- and 64-byte
18289         vector modes.
18290         * config/aarch64/iterators.md (VRL2, VRL3, VRL4): Delete.
18291         * config/aarch64/aarch64.md (UNSPEC_LD2_DREG, UNSPEC_LD3_DREG)
18292         (UNSPEC_LD4_DREG): New unspecs.
18293         * config/aarch64/aarch64-simd.md (aarch64_ld2<mode>_dreg_le)
18294         (aarch64_ld2<mode>_dreg_be): Replace with...
18295         (aarch64_ld2<mode>_dreg): ...this pattern and use the new DREG
18296         unspec.
18297         (aarch64_ld3<mode>_dreg_le)
18298         (aarch64_ld3<mode>_dreg_be): Replace with...
18299         (aarch64_ld3<mode>_dreg): ...this pattern and use the new DREG
18300         unspec.
18301         (aarch64_ld4<mode>_dreg_le)
18302         (aarch64_ld4<mode>_dreg_be): Replace with...
18303         (aarch64_ld4<mode>_dreg): ...this pattern and use the new DREG
18304         unspec.
18306 2017-08-30  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
18308         PR tree-optimization/81987
18309         * gimple-ssa-strength-reduction.c (insert_initializers): Don't
18310         insert an initializer in a location not dominated by the stride
18311         definition.
18313 2017-08-30  Eric Botcazou  <ebotcazou@adacore.com>
18315         * tree-eh.c (lower_try_finally_switch): Set the location of the finally
18316         on the entire header of the finally block in the fallthru case.
18318 2017-08-30  Eric Botcazou  <ebotcazou@adacore.com>
18320         * varasm.c (decode_addr_const): Deal with INDIRECT_REF <INTEGER_CST>.
18322 2017-08-30  Pat Haugen  <pthaugen@us.ibm.com>
18324         * config/rs6000/rs6000.c (rs6000_emit_prologue_move_from_cr): Rename from
18325         rs6000_emit_move_from_cr and call renamed function.
18326         (rs6000_emit_prologue): Call renamed functions.
18327         * config/rs6000/rs6000.md (prologue_movesi_from_cr): Rename from
18328         movesi_from_cr, remove volatile CRs.
18330 2017-08-30  Jon Beniston  <jon@beniston.com>
18331             Richard Biener  <rguenther@suse.de>
18333         * tree-vect-patterns.c (vect_pattern_recog_1): Use VECTOR_TYPE_P instead
18334         of VECTOR_MODE_P check.
18335         * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Allow single
18336         element vector types.
18338 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18340         * df.h (df_read_modify_subreg_p): Remove in favor of...
18341         * rtl.h (read_modify_subreg_p): ...this new function.  Take a
18342         const_rtx instead of an rtx.
18343         * cprop.c (local_cprop_find_used_regs): Update accordingly.
18344         * df-problems.c (df_word_lr_mark_ref): Likewise.
18345         * ira-lives.c (mark_pseudo_reg_live): Likewise.
18346         (mark_pseudo_reg_dead): Likewise.
18347         (mark_ref_dead): Likewise.
18348         * reginfo.c (init_subregs_of_mode): Likewise.
18349         * sched-deps.c (sched_analyze_1): Likewise.
18350         * df-scan.c (df_def_record_1): Likewise.
18351         (df_uses_record): Likewise.
18352         (df_read_modify_subreg_p): Remove in favor of...
18353         * rtlanal.c (read_modify_subreg_p): ...this new function.  Take a
18354         const_rtx instead of an rtx.
18356 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18357             Alan Hayward  <alan.hayward@arm.com>
18358             David Sherwood  <david.sherwood@arm.com>
18360         * rtl.h (partial_subreg_p): New function.
18361         * caller-save.c (save_call_clobbered_regs): Use it.
18362         * calls.c (expand_call): Likewise.
18363         * combine.c (combinable_i3pat): Likewise.
18364         (simplify_set): Likewise.
18365         (make_extraction): Likewise.
18366         (make_compound_operation_int): Likewise.
18367         (gen_lowpart_or_truncate): Likewise.
18368         (force_to_mode): Likewise.
18369         (make_field_assignment): Likewise.
18370         (reg_truncated_to_mode): Likewise.
18371         (record_truncated_value): Likewise.
18372         (move_deaths): Likewise.
18373         * cse.c (record_jump_cond): Likewise.
18374         (cse_insn): Likewise.
18375         * cselib.c (cselib_lookup_1): Likewise.
18376         * expmed.c (extract_bit_field_using_extv): Likewise.
18377         * function.c (assign_parm_setup_reg): Likewise.
18378         * ifcvt.c (noce_convert_multiple_sets): Likewise.
18379         * ira-build.c (create_insn_allocnos): Likewise.
18380         * lra-coalesce.c (merge_pseudos): Likewise.
18381         * lra-constraints.c (match_reload): Likewise.
18382         (simplify_operand_subreg): Likewise.
18383         (curr_insn_transform): Likewise.
18384         * lra-lives.c (process_bb_lives): Likewise.
18385         * lra.c (new_insn_reg): Likewise.
18386         (lra_substitute_pseudo): Likewise.
18387         * regcprop.c (mode_change_ok): Likewise.
18388         (maybe_mode_change): Likewise.
18389         (copyprop_hardreg_forward_1): Likewise.
18390         * reload.c (push_reload): Likewise.
18391         (find_reloads): Likewise.
18392         (find_reloads_subreg_address): Likewise.
18393         * reload1.c (alter_reg): Likewise.
18394         (eliminate_regs_1): Likewise.
18395         * simplify-rtx.c (simplify_unary_operation_1): Likewise.
18397 2017-08-30  David Edelsohn  <dje.gcc@gmail.com>
18399         * config/rs6000/rs6000.c (rs6000_expand_binop_builtin): Revert
18400         back to if statements, including unpack.
18402 2017-08-30  Martin Liska  <mliska@suse.cz>
18404         PR inline-asm/82001
18405         * ipa-icf-gimple.c (func_checker::compare_tree_list_operand):
18406         Rename to ...
18407         (func_checker::compare_asm_inputs_outputs): ... this function.
18408         (func_checker::compare_gimple_asm): Use the function to compare
18409         also ASM constrains.
18410         * ipa-icf-gimple.h: Rename the function.
18412 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18413             Alan Hayward  <alan.hayward@arm.com>
18414             David Sherwood  <david.sherwood@arm.com>
18416         * coretypes.h (complex_mode): New type.
18417         * gdbhooks.py (build_pretty_printer): Handle it.
18418         * machmode.h (complex_mode): New class.
18419         (complex_mode::includes_p): New function.
18420         (is_complex_int_mode): Likewise.
18421         (is_complex_float_mode): Likewise.
18422         * genmodes.c (get_mode_class): Handle complex mode classes.
18423         * function.c (expand_function_end): Use is_complex_int_mode.
18425 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18426             Alan Hayward  <alan.hayward@arm.com>
18427             David Sherwood  <david.sherwood@arm.com>
18429         * coretypes.h (scalar_mode_pod): New typedef.
18430         * gdbhooks.py (build_pretty_printer): Handle it.
18431         * machmode.h (gt_ggc_mx, gt_pch_nx): New functions.
18432         * fixed-value.h (fixed_value::mode): Change type to scalar_mode_pod.
18433         * fold-const.c (fold_convert_const_int_from_fixed): Use scalar_mode.
18434         * tree-streamer-in.c (unpack_ts_fixed_cst_value_fields): Use
18435         as_a <scalar_mode>.
18437 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18438             Alan Hayward  <alan.hayward@arm.com>
18439             David Sherwood  <david.sherwood@arm.com>
18441         * machmode.h (mode_for_vector): Take a scalar_mode instead
18442         of a machine_mode.
18443         * stor-layout.c (mode_for_vector): Likewise.
18444         * explow.c (promote_mode): Use as_a <scalar_mode>.
18445         * sdbout.c (sdbout_parms): Use is_a <scalar_mode>.
18447 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18448             Alan Hayward  <alan.hayward@arm.com>
18449             David Sherwood  <david.sherwood@arm.com>
18451         * target.def (preferred_simd_mode): Take a scalar_mode
18452         instead of a machine_mode.
18453         * targhooks.h (default_preferred_simd_mode): Likewise.
18454         * targhooks.c (default_preferred_simd_mode): Likewise.
18455         * config/arc/arc.c (arc_preferred_simd_mode): Likewise.
18456         * config/arm/arm.c (arm_preferred_simd_mode): Likewise.
18457         * config/c6x/c6x.c (c6x_preferred_simd_mode): Likewise.
18458         * config/epiphany/epiphany.c (epiphany_preferred_simd_mode): Likewise.
18459         * config/i386/i386.c (ix86_preferred_simd_mode): Likewise.
18460         * config/mips/mips.c (mips_preferred_simd_mode): Likewise.
18461         * config/nvptx/nvptx.c (nvptx_preferred_simd_mode): Likewise.
18462         * config/powerpcspe/powerpcspe.c (rs6000_preferred_simd_mode):
18463         Likewise.
18464         * config/rs6000/rs6000.c (rs6000_preferred_simd_mode): Likewise.
18465         * config/s390/s390.c (s390_preferred_simd_mode): Likewise.
18466         * config/sparc/sparc.c (sparc_preferred_simd_mode): Likewise.
18467         * config/aarch64/aarch64.c (aarch64_preferred_simd_mode): Likewise.
18468         (aarch64_simd_scalar_immediate_valid_for_move): Update accordingly.
18469         * doc/tm.texi: Regenerate.
18470         * optabs-query.c (can_vec_mask_load_store_p): Return false for
18471         non-scalar modes.
18473 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18474             Alan Hayward  <alan.hayward@arm.com>
18475             David Sherwood  <david.sherwood@arm.com>
18477         * target.def (scalar_mode_supported_p): Take a scalar_mode
18478         instead of a machine_mode.
18479         * targhooks.h (default_scalar_mode_supported_p): Likewise.
18480         * targhooks.c (default_scalar_mode_supported_p): Likewise.
18481         * config/aarch64/aarch64.c (aarch64_scalar_mode_supported_p): Likewise.
18482         * config/alpha/alpha.c (alpha_scalar_mode_supported_p): Likewise.
18483         * config/arm/arm.c (arm_scalar_mode_supported_p): Likewise.
18484         * config/avr/avr.c (avr_scalar_mode_supported_p): Likewise.
18485         * config/c6x/c6x.c (c6x_scalar_mode_supported_p): Likewise.
18486         * config/i386/i386.c (ix86_scalar_mode_supported_p): Likewise.
18487         * config/ia64/ia64.c (ia64_scalar_mode_supported_p): Likewise.
18488         * config/mips/mips.c (mips_scalar_mode_supported_p): Likewise.
18489         * config/msp430/msp430.c (msp430_scalar_mode_supported_p): Likewise.
18490         * config/pa/pa.c (pa_scalar_mode_supported_p): Likewise.
18491         * config/pdp11/pdp11.c (pdp11_scalar_mode_supported_p): Likewise.
18492         * config/powerpcspe/powerpcspe.c (rs6000_scalar_mode_supported_p):
18493         Likewise.
18494         * config/rs6000/rs6000.c (rs6000_scalar_mode_supported_p): Likewise.
18495         * config/s390/s390.c (s390_scalar_mode_supported_p): Likewise.
18496         * config/spu/spu.c (spu_scalar_mode_supported_p): Likewise.
18497         * config/tilegx/tilegx.c (tilegx_scalar_mode_supported_p): Likewise.
18498         * config/tilepro/tilepro.c (tilepro_scalar_mode_supported_p):
18499         Likewise.
18500         * doc/tm.texi: Regenerate.
18502 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18503             Alan Hayward  <alan.hayward@arm.com>
18504             David Sherwood  <david.sherwood@arm.com>
18506         * coretypes.h (opt_scalar_mode): New typedef.
18507         * gdbhooks.py (build_pretty_printers): Handle it.
18508         * machmode.h (mode_iterator::get_2xwider): Add overload for
18509         opt_mode<T>.
18510         * emit-rtl.c (init_emit_once): Use opt_scalar_mode when iterating
18511         over scalar modes.
18512         * expr.c (convert_mode_scalar): Likewise.
18513         * omp-low.c (omp_clause_aligned_alignment): Likewise.
18514         * optabs.c (expand_float): Likewise.
18515         (expand_fix): Likewise.
18516         * tree-vect-stmts.c (vectorizable_conversion): Likewise.
18518 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18519             Alan Hayward  <alan.hayward@arm.com>
18520             David Sherwood  <david.sherwood@arm.com>
18522         * optabs.c (expand_float): Explicitly check for scalars before
18523         using a branching expansion.
18524         (expand_fix): Likewise.
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         * expr.c (convert_mode): Split scalar handling out into...
18531         (convert_mode_scalar): ...this new function.  Treat the modes
18532         as scalar_modes.
18534 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18535             Alan Hayward  <alan.hayward@arm.com>
18536             David Sherwood  <david.sherwood@arm.com>
18538         * omp-expand.c (expand_omp_atomic): Use is_int_mode, is_float_mode
18539         and scalar_mode.
18540         * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Likewise.
18542 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18543             Alan Hayward  <alan.hayward@arm.com>
18544             David Sherwood  <david.sherwood@arm.com>
18546         * fixed-value.h (fixed_from_double_int): Take a scalar_mode
18547         rather than a machine_mode.
18548         (fixed_from_string): Likewise.
18549         (fixed_convert): Likewise.
18550         (fixed_convert_from_int): Likewise.
18551         (fixed_convert_from_real): Likewise.
18552         (real_convert_from_fixed): Likewise.
18553         * fixed-value.c (fixed_from_double_int): Likewise.
18554         (fixed_from_string): Likewise.
18555         (fixed_convert): Likewise.
18556         (fixed_convert_from_int): Likewise.
18557         (fixed_convert_from_real): Likewise.
18558         (real_convert_from_fixed): Likewise.
18559         * config/avr/avr.c (avr_out_round): Use as_a <scalar_mode>.
18561 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18562             Alan Hayward  <alan.hayward@arm.com>
18563             David Sherwood  <david.sherwood@arm.com>
18565         * emit-rtl.c (immed_double_const): Use is_a <scalar_mode> instead
18566         of separate mode class checks.  Do not allow vector modes here.
18567         (immed_wide_int_const): Use as_a <scalar_mode>.
18568         * explow.c (trunc_int_for_mode): Likewise.
18569         * rtl.h (wi::int_traits<rtx_mode_t>::get_precision): Likewise.
18570         (wi::shwi): Likewise.
18571         (wi::min_value): Likewise.
18572         (wi::max_value): Likewise.
18573         * dwarf2out.c (loc_descriptor): Likewise.
18574         * simplify-rtx.c (simplify_immed_subreg): Fix rtx_mode_t argument
18575         for CONST_WIDE_INT.
18577 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18578             Alan Hayward  <alan.hayward@arm.com>
18579             David Sherwood  <david.sherwood@arm.com>
18581         * tree.h (SCALAR_TYPE_MODE): New macro.
18582         * expr.c (expand_expr_addr_expr_1): Use it.
18583         (expand_expr_real_2): Likewise.
18584         * fold-const.c (fold_convert_const_fixed_from_fixed): Likeise.
18585         (fold_convert_const_fixed_from_int): Likewise.
18586         (fold_convert_const_fixed_from_real): Likewise.
18587         (native_encode_fixed): Likewise
18588         (native_encode_complex): Likewise
18589         (native_encode_vector): Likewise.
18590         (native_interpret_fixed): Likewise.
18591         (native_interpret_real): Likewise.
18592         (native_interpret_complex): Likewise.
18593         (native_interpret_vector): Likewise.
18594         * omp-simd-clone.c (simd_clone_adjust_return_type): Likewise.
18595         (simd_clone_adjust_argument_types): Likewise.
18596         (simd_clone_init_simd_arrays): Likewise.
18597         (simd_clone_adjust): Likewise.
18598         * stor-layout.c (layout_type): Likewise.
18599         * tree.c (build_minus_one_cst): Likewise.
18600         * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
18601         * tree-inline.c (estimate_move_cost): Likewise.
18602         * tree-ssa-math-opts.c (convert_plusminus_to_widen): Likewise.
18603         * tree-vect-loop.c (vect_create_epilog_for_reduction): Likewise.
18604         (vectorizable_reduction): Likewise.
18605         * tree-vect-patterns.c (vect_recog_widen_mult_pattern): Likewise.
18606         (vect_recog_mixed_size_cond_pattern): Likewise.
18607         (check_bool_pattern): Likewise.
18608         (adjust_bool_pattern): Likewise.
18609         (search_type_for_mask_1): Likewise.
18610         * tree-vect-slp.c (vect_schedule_slp_instance): Likewise.
18611         * tree-vect-stmts.c (vectorizable_conversion): Likewise.
18612         (vectorizable_load): Likewise.
18613         (vectorizable_store): Likewise.
18614         * ubsan.c (ubsan_encode_value): Likewise.
18615         * varasm.c (output_constant): Likewise.
18617 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18618             Alan Hayward  <alan.hayward@arm.com>
18619             David Sherwood  <david.sherwood@arm.com>
18621         * coretypes.h (scalar_mode): New class.
18622         * machmode.h (scalar_mode): Likewise.
18623         (scalar_mode::includes_p): New function.
18624         (mode_to_inner): Return a scalar_mode rather than a machine_mode.
18625         * gdbhooks.py (build_pretty_printers): Handle scalar_mode.
18626         * genmodes.c (get_mode_class): Handle remaining scalar modes.
18627         * cfgexpand.c (expand_debug_expr): Use scalar_mode.
18628         * expmed.c (store_bit_field_1): Likewise.
18629         (extract_bit_field_1): Likewise.
18630         * expr.c (write_complex_part): Likewise.
18631         (read_complex_part): Likewise.
18632         (emit_move_complex_push): Likewise.
18633         (expand_expr_real_2): Likewise.
18634         * function.c (assign_parm_setup_reg): Likewise.
18635         (assign_parms_unsplit_complex): Likewise.
18636         * optabs.c (expand_binop): Likewise.
18637         * rtlanal.c (subreg_get_info): Likewise.
18638         * simplify-rtx.c (simplify_immed_subreg): Likewise.
18639         * varasm.c (output_constant_pool_2): Likewise.
18641 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18642             Alan Hayward  <alan.hayward@arm.com>
18643             David Sherwood  <david.sherwood@arm.com>
18645         * expmed.c (extract_high_half): Use scalar_int_mode and remove
18646         assertion.
18647         (expmed_mult_highpart_optab): Likewise.
18648         (expmed_mult_highpart): Likewise.
18650 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18651             Alan Hayward  <alan.hayward@arm.com>
18652             David Sherwood  <david.sherwood@arm.com>
18654         * builtins.h (builtin_strncpy_read_str): Take a scalar_int_mode
18655         instead of a machine_mode.
18656         (builtin_memset_read_str): Likewise.
18657         * builtins.c (c_readstr): Likewise.
18658         (builtin_memcpy_read_str): Likewise.
18659         (builtin_strncpy_read_str): Likewise.
18660         (builtin_memset_read_str): Likewise.
18661         (builtin_memset_gen_str): Likewise.
18662         (expand_builtin_signbit): Use scalar_int_mode for local variables.
18663         * cfgexpand.c (convert_debug_memory_address): Take a scalar_int_mode
18664         instead of a machine_mode.
18665         * combine.c (simplify_if_then_else): Use scalar_int_mode for local
18666         variables.
18667         (make_extraction): Likewise.
18668         (try_widen_shift_mode): Take and return scalar_int_modes instead
18669         of machine_modes.
18670         * config/aarch64/aarch64.c (aarch64_libgcc_cmp_return_mode): Return
18671         a scalar_int_mode instead of a machine_mode.
18672         * config/avr/avr.c (avr_addr_space_address_mode): Likewise.
18673         (avr_addr_space_pointer_mode): Likewise.
18674         * config/cr16/cr16.c (cr16_unwind_word_mode): Likewise.
18675         * config/msp430/msp430.c (msp430_addr_space_pointer_mode): Likewise.
18676         (msp430_unwind_word_mode): Likewise.
18677         * config/spu/spu.c (spu_unwind_word_mode): Likewise.
18678         (spu_addr_space_pointer_mode): Likewise.
18679         (spu_addr_space_address_mode): Likewise.
18680         (spu_libgcc_cmp_return_mode): Likewise.
18681         (spu_libgcc_shift_count_mode): Likewise.
18682         * config/rl78/rl78.c (rl78_addr_space_address_mode): Likewise.
18683         (rl78_addr_space_pointer_mode): Likewise.
18684         (fl78_unwind_word_mode): Likewise.
18685         (rl78_valid_pointer_mode): Take a scalar_int_mode instead of a
18686         machine_mode.
18687         * config/alpha/alpha.c (vms_valid_pointer_mode): Likewise.
18688         * config/ia64/ia64.c (ia64_vms_valid_pointer_mode): Likewise.
18689         * config/mips/mips.c (mips_mode_rep_extended): Likewise.
18690         (mips_valid_pointer_mode): Likewise.
18691         * config/tilegx/tilegx.c (tilegx_mode_rep_extended): Likewise.
18692         * config/ft32/ft32.c (ft32_valid_pointer_mode): Likewise.
18693         (ft32_addr_space_pointer_mode): Return a scalar_int_mode instead
18694         of a machine_mode.
18695         (ft32_addr_space_address_mode): Likewise.
18696         * config/m32c/m32c.c (m32c_valid_pointer_mode): Take a
18697         scalar_int_mode instead of a machine_mode.
18698         (m32c_addr_space_pointer_mode): Return a scalar_int_mode instead
18699         of a machine_mode.
18700         (m32c_addr_space_address_mode): Likewise.
18701         * config/powerpcspe/powerpcspe.c (rs6000_abi_word_mode): Likewise.
18702         (rs6000_eh_return_filter_mode): Likewise.
18703         * config/rs6000/rs6000.c (rs6000_abi_word_mode): Likewise.
18704         (rs6000_eh_return_filter_mode): Likewise.
18705         * config/s390/s390.c (s390_libgcc_cmp_return_mode): Likewise.
18706         (s390_libgcc_shift_count_mode): Likewise.
18707         (s390_unwind_word_mode): Likewise.
18708         (s390_valid_pointer_mode): Take a scalar_int_mode rather than a
18709         machine_mode.
18710         * target.def (mode_rep_extended): Likewise.
18711         (valid_pointer_mode): Likewise.
18712         (addr_space.valid_pointer_mode): Likewise.
18713         (eh_return_filter_mode): Return a scalar_int_mode rather than
18714         a machine_mode.
18715         (libgcc_cmp_return_mode): Likewise.
18716         (libgcc_shift_count_mode): Likewise.
18717         (unwind_word_mode): Likewise.
18718         (addr_space.pointer_mode): Likewise.
18719         (addr_space.address_mode): Likewise.
18720         * doc/tm.texi: Regenerate.
18721         * dojump.c (prefer_and_bit_test): Take a scalar_int_mode rather than
18722         a machine_mode.
18723         (do_jump): Use scalar_int_mode for local variables.
18724         * dwarf2cfi.c (init_return_column_size): Take a scalar_int_mode
18725         rather than a machine_mode.
18726         * dwarf2out.c (convert_descriptor_to_mode): Likewise.
18727         (scompare_loc_descriptor_wide): Likewise.
18728         (scompare_loc_descriptor_narrow): Likewise.
18729         * emit-rtl.c (adjust_address_1): Use scalar_int_mode for local
18730         variables.
18731         * except.c (sjlj_emit_dispatch_table): Likewise.
18732         (expand_builtin_eh_copy_values): Likewise.
18733         * explow.c (convert_memory_address_addr_space_1): Likewise.
18734         Take a scalar_int_mode rather than a machine_mode.
18735         (convert_memory_address_addr_space): Take a scalar_int_mode rather
18736         than a machine_mode.
18737         (memory_address_addr_space): Use scalar_int_mode for local variables.
18738         * expmed.h (expand_mult_highpart_adjust): Take a scalar_int_mode
18739         rather than a machine_mode.
18740         * expmed.c (mask_rtx): Likewise.
18741         (init_expmed_one_conv): Likewise.
18742         (expand_mult_highpart_adjust): Likewise.
18743         (extract_high_half): Likewise.
18744         (expmed_mult_highpart_optab): Likewise.
18745         (expmed_mult_highpart): Likewise.
18746         (expand_smod_pow2): Likewise.
18747         (expand_sdiv_pow2): Likewise.
18748         (emit_store_flag_int): Likewise.
18749         (adjust_bit_field_mem_for_reg): Use scalar_int_mode for local
18750         variables.
18751         (extract_low_bits): Likewise.
18752         * expr.h (by_pieces_constfn): Take a scalar_int_mode rather than
18753         a machine_mode.
18754         * expr.c (pieces_addr::adjust):  Likewise.
18755         (can_store_by_pieces): Likewise.
18756         (store_by_pieces): Likewise.
18757         (clear_by_pieces_1): Likewise.
18758         (expand_expr_addr_expr_1): Likewise.
18759         (expand_expr_addr_expr): Use scalar_int_mode for local variables.
18760         (expand_expr_real_1): Likewise.
18761         (try_casesi): Likewise.
18762         * final.c (shorten_branches): Likewise.
18763         * fold-const.c (fold_convert_const_int_from_fixed): Change the
18764         type of "mode" to machine_mode.
18765         * internal-fn.c (expand_arith_overflow_result_store): Take a
18766         scalar_int_mode rather than a machine_mode.
18767         (expand_mul_overflow): Use scalar_int_mode for local variables.
18768         * loop-doloop.c (doloop_modify): Likewise.
18769         (doloop_optimize): Likewise.
18770         * optabs.c (expand_subword_shift): Take a scalar_int_mode rather
18771         than a machine_mode.
18772         (expand_doubleword_shift_condmove): Likewise.
18773         (expand_doubleword_shift): Likewise.
18774         (expand_doubleword_clz): Likewise.
18775         (expand_doubleword_popcount): Likewise.
18776         (expand_doubleword_parity): Likewise.
18777         (expand_absneg_bit): Use scalar_int_mode for local variables.
18778         (prepare_float_lib_cmp): Likewise.
18779         * rtl.h (convert_memory_address_addr_space_1): Take a scalar_int_mode
18780         rather than a machine_mode.
18781         (convert_memory_address_addr_space): Likewise.
18782         (get_mode_bounds): Likewise.
18783         (get_address_mode): Return a scalar_int_mode rather than a
18784         machine_mode.
18785         * rtlanal.c (get_address_mode): Likewise.
18786         * stor-layout.c (get_mode_bounds): Take a scalar_int_mode rather
18787         than a machine_mode.
18788         * targhooks.c (default_mode_rep_extended): Likewise.
18789         (default_valid_pointer_mode): Likewise.
18790         (default_addr_space_valid_pointer_mode): Likewise.
18791         (default_eh_return_filter_mode): Return a scalar_int_mode rather
18792         than a machine_mode.
18793         (default_libgcc_cmp_return_mode): Likewise.
18794         (default_libgcc_shift_count_mode): Likewise.
18795         (default_unwind_word_mode): Likewise.
18796         (default_addr_space_pointer_mode): Likewise.
18797         (default_addr_space_address_mode): Likewise.
18798         * targhooks.h (default_eh_return_filter_mode): Likewise.
18799         (default_libgcc_cmp_return_mode): Likewise.
18800         (default_libgcc_shift_count_mode): Likewise.
18801         (default_unwind_word_mode): Likewise.
18802         (default_addr_space_pointer_mode): Likewise.
18803         (default_addr_space_address_mode): Likewise.
18804         (default_mode_rep_extended): Take a scalar_int_mode rather than
18805         a machine_mode.
18806         (default_valid_pointer_mode): Likewise.
18807         (default_addr_space_valid_pointer_mode): Likewise.
18808         * tree-ssa-address.c (addr_for_mem_ref): Use scalar_int_mode for
18809         local variables.
18810         * tree-ssa-loop-ivopts.c (get_shiftadd_cost): Take a scalar_int_mode
18811         rather than a machine_mode.
18812         * tree-switch-conversion.c (array_value_type): Use scalar_int_mode
18813         for local variables.
18814         * tree-vrp.c (simplify_float_conversion_using_ranges): Likewise.
18815         * var-tracking.c (use_narrower_mode): Take a scalar_int_mode rather
18816         than a machine_mode.
18818 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18819             Alan Hayward  <alan.hayward@arm.com>
18820             David Sherwood  <david.sherwood@arm.com>
18822         * dojump.c (do_jump_by_parts_greater_rtx): Change the type of
18823         the mode argument to scalar_int_mode.
18824         (do_jump_by_parts_zero_rtx): Likewise.
18825         (do_jump_by_parts_equality_rtx): Likewise.
18826         (do_jump_by_parts_greater): Take a mode argument.
18827         (do_jump_by_parts_equality): Likewise.
18828         (do_jump_1): Update calls accordingly.
18830 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18831             Alan Hayward  <alan.hayward@arm.com>
18832             David Sherwood  <david.sherwood@arm.com>
18834         * is-a.h (safe_dyn_cast): New function.
18835         * rtl.h (rtx_jump_table_data::get_data_mode): New function.
18836         (jump_table_for_label): Likewise.
18837         * final.c (final_addr_vec_align): Take an rtx_jump_table_data *
18838         instead of an rtx_insn *.
18839         (shorten_branches): Use dyn_cast instead of LABEL_P and
18840         JUMP_TABLE_DATA_P.  Use jump_table_for_label and
18841         rtx_jump_table_data::get_data_mode.
18842         (final_scan_insn): Likewise.
18844 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18845             Alan Hayward  <alan.hayward@arm.com>
18846             David Sherwood  <david.sherwood@arm.com>
18848         * combine.c (try_combine): Use is_a <scalar_int_mode> when
18849         trying to combine a full-register integer set with a subreg
18850         integer set.
18852 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18853             Alan Hayward  <alan.hayward@arm.com>
18854             David Sherwood  <david.sherwood@arm.com>
18856         * expr.c (expand_expr_addr_expr): Add a new_tmode local variable
18857         that is always either address_mode or pointer_mode.
18859 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18860             Alan Hayward  <alan.hayward@arm.com>
18861             David Sherwood  <david.sherwood@arm.com>
18863         * expr.c (expand_expr_real_2): Use word_mode instead of innermode
18864         when the two are known to be equal.
18866 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18867             Alan Hayward  <alan.hayward@arm.com>
18868             David Sherwood  <david.sherwood@arm.com>
18870         * simplify-rtx.c (simplify_const_unary_operation): Use
18871         is_a <scalar_int_mode> instead of checking for a nonzero
18872         precision.  Forcibly convert op_mode to a scalar_int_mode
18873         in that case.  More clearly differentiate the operand and
18874         result modes and use the former when deciding what the value
18875         of a count-bits operation should be.  Use is_int_mode instead
18876         of checking for a MODE_INT.  Remove redundant check for whether
18877         this mode has a zero precision.
18879 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18880             Alan Hayward  <alan.hayward@arm.com>
18881             David Sherwood  <david.sherwood@arm.com>
18883         * optabs.c (widen_leading): Change the type of the mode argument
18884         to scalar_int_mode.  Use opt_scalar_int_mode for the mode iterator.
18885         (widen_bswap): Likewise.
18886         (expand_parity): Likewise.
18887         (expand_ctz): Change the type of the mode argument to scalar_int_mode.
18888         (expand_ffs): Likewise.
18889         (epand_unop): Check for scalar integer modes before calling the
18890         above routines.
18892 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18893             Alan Hayward  <alan.hayward@arm.com>
18894             David Sherwood  <david.sherwood@arm.com>
18896         * expr.c (const_scalar_mask_from_tree): Add a mode argument.
18897         Expand commentary.
18898         (expand_expr_real_1): Update call accordingly.
18900 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18901             Alan Hayward  <alan.hayward@arm.com>
18902             David Sherwood  <david.sherwood@arm.com>
18904         * expmed.c (store_bit_field_using_insv): Add op0_mode and
18905         value_mode arguments.  Use scalar_int_mode internally.
18906         (store_bit_field_1): Rename the new integer mode from imode
18907         to op0_mode and use it instead of GET_MODE (op0).  Update calls
18908         to store_split_bit_field, store_bit_field_using_insv and
18909         store_fixed_bit_field.
18910         (store_fixed_bit_field): Add op0_mode and value_mode arguments.
18911         Use scalar_int_mode internally.  Use a bit count rather than a mode
18912         when calculating the largest bit size for get_best_mode.
18913         Update calls to store_split_bit_field and store_fixed_bit_field_1.
18914         (store_fixed_bit_field_1): Add mode and value_mode arguments.
18915         Remove assertion that OP0 has a scalar integer mode.
18916         (store_split_bit_field): Add op0_mode and value_mode arguments.
18917         Update calls to extract_fixed_bit_field.
18918         (extract_bit_field_using_extv): Add an op0_mode argument.
18919         Use scalar_int_mode internally.
18920         (extract_bit_field_1): Rename the new integer mode from imode to
18921         op0_mode and use it instead of GET_MODE (op0).  Update calls to
18922         extract_split_bit_field, extract_bit_field_using_extv and
18923         extract_fixed_bit_field.
18924         (extract_fixed_bit_field): Add an op0_mode argument.  Update calls
18925         to extract_split_bit_field and extract_fixed_bit_field_1.
18926         (extract_fixed_bit_field_1): Add a mode argument.  Remove assertion
18927         that OP0 has a scalar integer mode.  Use as_a <scalar_int_mode>
18928         on the target mode.
18929         (extract_split_bit_field): Add an op0_mode argument.  Update call
18930         to extract_fixed_bit_field.
18932 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18933             Alan Hayward  <alan.hayward@arm.com>
18934             David Sherwood  <david.sherwood@arm.com>
18936         * cse.c (cse_insn): Use opt_scalar_int_mode for the mode iterator.
18937         * explow.c (hard_function_value): Likewise.
18938         * expmed.c (extract_fixed_bit_field_1): Likewise.  Move the
18939         convert_to_mode call outside the loop.
18940         * expr.c (alignment_for_piecewise_move): Use opt_scalar_int_mode
18941         for the mode iterator.  Require the mode specified by max_pieces
18942         to exist.
18943         (emit_block_move_via_movmem): Use opt_scalar_int_mode for the
18944         mode iterator.
18945         (copy_blkmode_to_reg): Likewise.
18946         (set_storage_via_setmem): Likewise.
18947         * optabs.c (prepare_cmp_insn): Likewise.
18948         * rtlanal.c (init_num_sign_bit_copies_in_rep): Likewise.
18949         * stor-layout.c (finish_bitfield_representative): Likewise.
18951 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18952             Alan Hayward  <alan.hayward@arm.com>
18953             David Sherwood  <david.sherwood@arm.com>
18955         * rtl.h (subreg_unpromoted_mode, subreg_promoted_mode): New functions.
18956         * expr.c (convert_move): Use them.
18957         (convert_modes): Likewise.
18958         (store_expr_with_bounds): Likewise.
18960 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18961             Alan Hayward  <alan.hayward@arm.com>
18962             David Sherwood  <david.sherwood@arm.com>
18964         * rtl.h (rtl_hooks::reg_nonzero_bits): Add a scalar_int_mode
18965         parameter for the mode of "x".  Remove the "known_x", "known_mode"
18966         and "known_ret" arguments.  Change the type of the mode argument
18967         to scalar_int_mode.
18968         (rtl_hooks:reg_num_sign_bit_copies): Likewise.
18969         * combine.c (reg_nonzero_bits_for_combine): Update accordingly.
18970         (reg_num_sign_bit_copies_for_combine): Likewise.
18971         * rtlanal.c (nonzero_bits1): Likewise.
18972         (num_sign_bit_copies1): Likewise.
18973         * rtlhooks-def.h (reg_nonzero_bits_general): Likewise.
18974         (reg_num_sign_bit_copies_general): Likewise.
18975         * rtlhooks.c (reg_num_sign_bit_copies_general): Likewise.
18976         (reg_nonzero_bits_general): Likewise.
18978 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18979             Alan Hayward  <alan.hayward@arm.com>
18980             David Sherwood  <david.sherwood@arm.com>
18982         * rtlanal.c (num_sign_bit_copies): Handle VOIDmode here rather
18983         than in subroutines.  Return 1 for non-integer modes.
18984         (cached_num_sign_bit_copies): Change the type of the mode parameter
18985         to scalar_int_mode.
18986         (num_sign_bit_copies1): Likewise.  Remove early exit for other mode
18987         classes.  Handle CONST_INT_P first and then check whether X also
18988         has a scalar integer mode.  Check the same thing for inner registers
18989         of a SUBREG and for values that are being extended or truncated.
18991 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18992             Alan Hayward  <alan.hayward@arm.com>
18993             David Sherwood  <david.sherwood@arm.com>
18995         * rtlanal.c (nonzero_bits): Handle VOIDmode here rather than
18996         in subroutines.  Return the mode mask for non-integer modes.
18997         (cached_nonzero_bits): Change the type of the mode parameter
18998         to scalar_int_mode.
18999         (nonzero_bits1): Likewise.  Remove early exit for other mode
19000         classes.  Handle CONST_INT_P first and then check whether X
19001         also has a scalar integer mode.
19003 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19004             Alan Hayward  <alan.hayward@arm.com>
19005             David Sherwood  <david.sherwood@arm.com>
19007         * expr.c (widest_int_mode_for_size): Make the comment match the code.
19008         Return a scalar_int_mode and assert that the size is greater than
19009         one byte.
19010         (by_pieces_ninsns): Update accordingly and remove VOIDmode handling.
19011         (op_by_pieces_d::op_by_pieces_d): Likewise.
19012         (op_by_pieces_d::run): Likewise.
19013         (can_store_by_pieces): Likewise.
19015 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19016             Alan Hayward  <alan.hayward@arm.com>
19017             David Sherwood  <david.sherwood@arm.com>
19019         * combine.c (extract_left_shift): Add a mode argument and update
19020         recursive calls.
19021         (make_compound_operation_int): Change the type of the mode parameter
19022         to scalar_int_mode and update the call to extract_left_shift.
19024 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19025             Alan Hayward  <alan.hayward@arm.com>
19026             David Sherwood  <david.sherwood@arm.com>
19028         * combine.c (simplify_and_const_int): Change the type of the mode
19029         parameter to scalar_int_mode.
19030         (simplify_and_const_int_1): Likewise.  Update recursive call.
19032 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19033             Alan Hayward  <alan.hayward@arm.com>
19034             David Sherwood  <david.sherwood@arm.com>
19036         * combine.c (simplify_compare_const): Check that the mode is a
19037         scalar_int_mode (rather than VOIDmode) before testing its
19038         precision.
19039         (simplify_comparison): Move COMPARISON_P handling out of the
19040         loop and restrict the latter part of the loop to scalar_int_modes.
19041         Check is_a <scalar_int_mode> before calling HWI_COMPUTABLE_MODE_P
19042         and when considering SUBREG_REGs.  Use is_int_mode instead of
19043         checking GET_MODE_CLASS against MODE_INT.
19045 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19046             Alan Hayward  <alan.hayward@arm.com>
19047             David Sherwood  <david.sherwood@arm.com>
19049         * combine.c (try_widen_shift_mode): Move check for equal modes to...
19050         (simplify_shift_const_1): ...here.  Use scalar_int_mode for
19051         shift_unit_mode and for modes involved in scalar shifts.
19053 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19054             Alan Hayward  <alan.hayward@arm.com>
19055             David Sherwood  <david.sherwood@arm.com>
19057         * combine.c (force_int_to_mode): New function, split out from...
19058         (force_to_mode): ...here.  Keep xmode up-to-date and use it
19059         instead of GET_MODE (x).
19061 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19062             Alan Hayward  <alan.hayward@arm.com>
19063             David Sherwood  <david.sherwood@arm.com>
19065         * optabs-query.h (extraction_insn::struct_mode): Change type to
19066         opt_scalar_int_mode and update comment.
19067         (extraction_insn::field_mode): Change type to scalar_int_mode.
19068         (extraction_insn::pos_mode): Likewise.
19069         * combine.c (make_extraction): Update accordingly.
19070         * optabs-query.c (get_traditional_extraction_insn): Likewise.
19071         (get_optab_extraction_insn): Likewise.
19072         * recog.c (simplify_while_replacing): Likewise.
19073         * expmed.c (narrow_bit_field_mem): Change the type of the mode
19074         parameter to opt_scalar_int_mode.
19076 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19077             Alan Hayward  <alan.hayward@arm.com>
19078             David Sherwood  <david.sherwood@arm.com>
19080         * machmode.h (bit_field_mode_iterator::next_mode): Take a pointer
19081         to a scalar_int_mode instead of a machine_mode.
19082         (bit_field_mode_iterator::m_mode): Change type to opt_scalar_int_mode.
19083         (get_best_mode): Return a boolean and use a pointer argument to store
19084         the selected mode.  Replace the limit mode parameter with a bit limit.
19085         * expmed.c (adjust_bit_field_mem_for_reg): Use scalar_int_mode
19086         for the values returned by bit_field_mode_iterator::next_mode.
19087         (store_bit_field): Update call to get_best_mode.
19088         (store_fixed_bit_field): Likewise.
19089         (extract_fixed_bit_field): Likewise.
19090         * expr.c (optimize_bitfield_assignment_op): Likewise.
19091         * fold-const.c (optimize_bit_field_compare): Likewise.
19092         (fold_truth_andor_1): Likewise.
19093         * stor-layout.c (bit_field_mode_iterator::next_mode): As above.
19094         Update for new type of m_mode.
19095         (get_best_mode): As above.
19097 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19098             Alan Hayward  <alan.hayward@arm.com>
19099             David Sherwood  <david.sherwood@arm.com>
19101         * expmed.c (strict_volatile_bitfield_p): Change the type of fieldmode
19102         to scalar_int_mode.  Remove check for SCALAR_INT_MODE_P.
19103         (store_bit_field): Check is_a <scalar_int_mode> before calling
19104         strict_volatile_bitfield_p.
19105         (extract_bit_field): Likewise.
19107 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19108             Alan Hayward  <alan.hayward@arm.com>
19109             David Sherwood  <david.sherwood@arm.com>
19111         * target.def (cstore_mode): Return a scalar_int_mode.
19112         * doc/tm.texi: Regenerate.
19113         * config/sparc/sparc.c (sparc_cstore_mode): Return a scalar_int_mode.
19114         * targhooks.h (default_cstore_mode): Likewise.
19115         * targhooks.c (default_cstore_mode): Likewise, using a forced
19116         conversion.
19117         * expmed.c (emit_cstore): Expect the target of the cstore to be
19118         a scalar_int_mode.
19120 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19121             Alan Hayward  <alan.hayward@arm.com>
19122             David Sherwood  <david.sherwood@arm.com>
19124         * cfgloop.h (rtx_iv): Change type of extend_mode and mode to
19125         scalar_int_mode.
19126         (niter_desc): Likewise mode.
19127         (iv_analyze): Add a mode parameter.
19128         (biv_p): Likewise.
19129         (iv_analyze_expr): Pass the mode paraeter before the rtx it describes
19130         and change its type to scalar_int_mode.
19131         * loop-iv.c: Update commentary at head of file.
19132         (iv_constant): Pass the mode paraeter before the rtx it describes
19133         and change its type to scalar_int_mode.  Remove VOIDmode handling.
19134         (iv_subreg): Change the type of the mode parameter to scalar_int_mode.
19135         (iv_extend): Likewise.
19136         (shorten_into_mode): Likewise.
19137         (iv_add): Use scalar_int_mode.
19138         (iv_mult): Likewise.
19139         (iv_shift): Likewise.
19140         (canonicalize_iv_subregs): Likewise.
19141         (get_biv_step_1): Pass the outer_mode parameter before the rtx
19142         it describes and change its mode to scalar_int_mode.   Also change
19143         the type of the returned inner_mode to scalar_int_mode.
19144         (get_biv_step): Likewise, turning outer_mode from a pointer
19145         into a direct parameter.  Update call to get_biv_step_1.
19146         (iv_analyze_biv): Add an outer_mode parameter.  Update calls to
19147         iv_constant and get_biv_step.
19148         (iv_analyze_expr): Pass the mode parameter before the rtx it describes
19149         and change its type to scalar_int_mode.  Don't initialise iv->mode
19150         to VOIDmode and remove later checks for its still being VOIDmode.
19151         Update calls to iv_analyze_op and iv_analyze_expr.  Check
19152         is_a <scalar_int_mode> when changing the mode under consideration.
19153         (iv_analyze_def): Ignore registers that don't have a scalar_int_mode.
19154         Update call to iv_analyze_expr.
19155         (iv_analyze_op): Add a mode parameter.  Reject subregs whose
19156         inner register is not also a scalar_int_mode.  Update call to
19157         iv_analyze_biv.
19158         (iv_analyze): Add a mode parameter.  Update call to iv_analyze_op.
19159         (biv_p): Add a mode parameter.  Update call to iv_analyze_biv.
19160         (iv_number_of_iterations): Use is_a <scalar_int_mode> instead of
19161         separate mode class checks.  Update calls to iv_analyze.  Remove
19162         fix-up of VOIDmodes after iv_analyze_biv.
19163         * loop-unroll.c (analyze_iv_to_split_insn): Reject registers that
19164         don't have a scalar_int_mode.  Update call to biv_p.
19166 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19167             Alan Hayward  <alan.hayward@arm.com>
19168             David Sherwood  <david.sherwood@arm.com>
19170         * cfgexpand.c (convert_debug_memory_address): Use
19171         as_a <scalar_int_mode>.
19172         * combine.c (expand_compound_operation): Likewise.
19173         (make_extraction): Likewise.
19174         (change_zero_ext): Likewise.
19175         (simplify_comparison): Likewise.
19176         * cse.c (cse_insn): Likewise.
19177         * dwarf2out.c (minmax_loc_descriptor): Likewise.
19178         (mem_loc_descriptor): Likewise.
19179         (loc_descriptor): Likewise.
19180         * expmed.c (init_expmed_one_mode): Likewise.
19181         (synth_mult): Likewise.
19182         (emit_store_flag_1): Likewise.
19183         (expand_divmod): Likewise.  Use HWI_COMPUTABLE_MODE_P instead
19184         of a comparison with size.
19185         * expr.c (expand_assignment): Use as_a <scalar_int_mode>.
19186         (reduce_to_bit_field_precision): Likewise.
19187         * function.c (expand_function_end): Likewise.
19188         * internal-fn.c (expand_arith_overflow_result_store): Likewise.
19189         * loop-doloop.c (doloop_modify): Likewise.
19190         * optabs.c (expand_binop): Likewise.
19191         (expand_unop): Likewise.
19192         (expand_copysign_absneg): Likewise.
19193         (prepare_cmp_insn): Likewise.
19194         (maybe_legitimize_operand): Likewise.
19195         * recog.c (const_scalar_int_operand): Likewise.
19196         * rtlanal.c (get_address_mode): Likewise.
19197         * simplify-rtx.c (simplify_unary_operation_1): Likewise.
19198         (simplify_cond_clz_ctz): Likewise.
19199         * tree-nested.c (get_nl_goto_field): Likewise.
19200         * tree.c (build_vector_type_for_mode): Likewise.
19201         * var-tracking.c (use_narrower_mode): Likewise.
19203 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19204             Alan Hayward  <alan.hayward@arm.com>
19205             David Sherwood  <david.sherwood@arm.com>
19207         * tree.h (SCALAR_INT_TYPE_MODE): New macro.
19208         * builtins.c (expand_builtin_signbit): Use it.
19209         * cfgexpand.c (expand_debug_expr): Likewise.
19210         * dojump.c (do_jump): Likewise.
19211         (do_compare_and_jump): Likewise.
19212         * dwarf2cfi.c (expand_builtin_init_dwarf_reg_sizes): Likewise.
19213         * expmed.c (make_tree): Likewise.
19214         * expr.c (expand_expr_real_2): Likewise.
19215         (expand_expr_real_1): Likewise.
19216         (try_casesi): Likewise.
19217         * fold-const-call.c (fold_const_call_ss): Likewise.
19218         * fold-const.c (unextend): Likewise.
19219         (extract_muldiv_1): Likewise.
19220         (fold_single_bit_test): Likewise.
19221         (native_encode_int): Likewise.
19222         (native_encode_string): Likewise.
19223         (native_interpret_int): Likewise.
19224         * gimple-fold.c (gimple_fold_builtin_memset): Likewise.
19225         * internal-fn.c (expand_addsub_overflow): Likewise.
19226         (expand_neg_overflow): Likewise.
19227         (expand_mul_overflow): Likewise.
19228         (expand_arith_overflow): Likewise.
19229         * match.pd: Likewise.
19230         * stor-layout.c (layout_type): Likewise.
19231         * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
19232         * tree-ssa-math-opts.c (convert_mult_to_widen): Likewise.
19233         * tree-ssanames.c (get_range_info): Likewise.
19234         * tree-switch-conversion.c (array_value_type) Likewise.
19235         * tree-vect-patterns.c (vect_recog_rotate_pattern): Likewise.
19236         (vect_recog_divmod_pattern): Likewise.
19237         (vect_recog_mixed_size_cond_pattern): Likewise.
19238         * tree-vrp.c (extract_range_basic): Likewise.
19239         (simplify_float_conversion_using_ranges): Likewise.
19240         * tree.c (int_fits_type_p): Likewise.
19241         * ubsan.c (instrument_bool_enum_load): Likewise.
19242         * varasm.c (mergeable_string_section): Likewise.
19243         (narrowing_initializer_constant_valid_p): Likewise.
19244         (output_constant): Likewise.
19246 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19247             Alan Hayward  <alan.hayward@arm.com>
19248             David Sherwood  <david.sherwood@arm.com>
19250         * machmode.h (NARROWEST_INT_MODE): New macro.
19251         * expr.c (alignment_for_piecewise_move): Use it instead of
19252         GET_CLASS_NARROWEST_MODE (MODE_INT).
19253         (push_block): Likewise.
19254         * stor-layout.c (bit_field_mode_iterator::bit_field_mode_iterator):
19255         Likewise.
19256         * tree-vrp.c (simplify_float_conversion_using_ranges): Likewise.
19258 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19259             Alan Hayward  <alan.hayward@arm.com>
19260             David Sherwood  <david.sherwood@arm.com>
19262         * postreload.c (move2add_valid_value_p): Change the type of the
19263         mode parameter to scalar_int_mode.
19264         (move2add_use_add2_insn): Add a mode parameter and use it instead
19265         of GET_MODE (reg).
19266         (move2add_use_add3_insn): Likewise.
19267         (reload_cse_move2add): Update accordingly.
19269 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19270             Alan Hayward  <alan.hayward@arm.com>
19271             David Sherwood  <david.sherwood@arm.com>
19273         * expr.c (expand_expr_real_2): Use scalar_int_mode for the
19274         double-word mode.
19275         * lower-subreg.c (resolve_shift_zext): Use is_a <scalar_int_mode>.
19276         * optabs.c (expand_unop): Likewise.
19278 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19279             Alan Hayward  <alan.hayward@arm.com>
19280             David Sherwood  <david.sherwood@arm.com>
19282         * dwarf2out.c (typed_binop): Change mode parameter to scalar_int_mode.
19283         (clz_loc_descriptor): Likewise.  Remove SCALAR_INT_MODE_P check.
19284         (popcount_loc_descriptor): Likewise.
19285         (bswap_loc_descriptor): Likewise.
19286         (rotate_loc_descriptor): Likewise.
19287         (mem_loc_descriptor): Add is_a <scalar_int_mode> checks before
19288         calling the functions above.
19290 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19291             Alan Hayward  <alan.hayward@arm.com>
19292             David Sherwood  <david.sherwood@arm.com>
19294         * combine.c (sign_extend_short_imm): Add is_a <scalar_int_mode>
19295         checks.
19296         (try_combine): Likewise.
19297         (simplify_if_then_else): Likewise.
19298         * cse.c (cse_insn): Likewise.
19299         * dwarf2out.c (mem_loc_descriptor): Likewise.
19300         * emit-rtl.c (gen_lowpart_common): Likewise.
19301         * simplify-rtx.c (simplify_truncation): Likewise.
19302         (simplify_binary_operation_1): Likewise.
19303         (simplify_const_relational_operation): Likewise.
19304         (simplify_ternary_operation): Likewise.
19305         * tree-ssa-loop-ivopts.c (force_expr_to_var_cost): Likewise.
19307 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19308             Alan Hayward  <alan.hayward@arm.com>
19309             David Sherwood  <david.sherwood@arm.com>
19311         * cse.c (cse_insn): Add is_a <scalar_int_mode> checks.
19312         * reload.c (push_reload): Likewise.
19313         (find_reloads): Likewise.
19315 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19316             Alan Hayward  <alan.hayward@arm.com>
19317             David Sherwood  <david.sherwood@arm.com>
19319         * combine.c (find_split_point): Add is_a <scalar_int_mode> checks.
19320         (make_compound_operation_int): Likewise.
19321         (change_zero_ext): Likewise.
19322         * expr.c (convert_move): Likewise.
19323         (convert_modes): Likewise.
19324         * fwprop.c (forward_propagate_subreg): Likewise.
19325         * loop-iv.c (get_biv_step_1): Likewise.
19326         * optabs.c (widen_operand): Likewise.
19327         * postreload.c (move2add_valid_value_p): Likewise.
19328         * recog.c (simplify_while_replacing): Likewise.
19329         * simplify-rtx.c (simplify_unary_operation_1): Likewise.
19330         (simplify_binary_operation_1): Likewise.  Remove redundant
19331         mode equality check.
19333 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19334             Alan Hayward  <alan.hayward@arm.com>
19335             David Sherwood  <david.sherwood@arm.com>
19337         * combine.c (combine_simplify_rtx): Add checks for
19338         is_a <scalar_int_mode>.
19339         (simplify_if_then_else): Likewise.
19340         (make_field_assignment): Likewise.
19341         (simplify_comparison): Likewise.
19342         * ifcvt.c (noce_try_bitop): Likewise.
19343         * loop-invariant.c (canonicalize_address_mult): Likewise.
19344         * simplify-rtx.c (simplify_unary_operation_1): Likewise.
19346 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19347             Alan Hayward  <alan.hayward@arm.com>
19348             David Sherwood  <david.sherwood@arm.com>
19350         * gimple-fold.c (gimple_fold_builtin_memory_op): Use
19351         is_a <scalar_int_mode> instead of != BLKmode.
19353 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19354             Alan Hayward  <alan.hayward@arm.com>
19355             David Sherwood  <david.sherwood@arm.com>
19357         * cfgexpand.c (expand_debug_expr): Use is_a <scalar_int_mode>
19358         instead of != VOIDmode.
19359         * combine.c (if_then_else_cond): Likewise.
19360         (change_zero_ext): Likewise.
19361         * dwarf2out.c (mem_loc_descriptor): Likewise.
19362         (loc_descriptor): Likewise.
19363         * rtlanal.c (canonicalize_condition): Likewise.
19364         * simplify-rtx.c (simplify_relational_operation_1): Likewise.
19366 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19367             Alan Hayward  <alan.hayward@arm.com>
19368             David Sherwood  <david.sherwood@arm.com>
19370         * simplify-rtx.c (simplify_binary_operation_1): Use
19371         is_a <scalar_int_mode> instead of !VECTOR_MODE_P.
19373 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19374             Alan Hayward  <alan.hayward@arm.com>
19375             David Sherwood  <david.sherwood@arm.com>
19377         * wide-int.h (int_traits<unsigned char>) New class.
19378         (int_traits<unsigned short>) Likewise.
19379         * cfgexpand.c (expand_debug_expr): Use is_a <scalar_int_mode>.
19380         Use GET_MODE_UNIT_PRECISION and remove redundant test for
19381         SCALAR_INT_MODE_P.
19382         * combine.c (set_nonzero_bits_and_sign_copies): Use
19383         is_a <scalar_int_mode>.
19384         (find_split_point): Likewise.
19385         (combine_simplify_rtx): Likewise.
19386         (simplify_logical): Likewise.
19387         (expand_compound_operation): Likewise.
19388         (expand_field_assignment): Likewise.
19389         (make_compound_operation): Likewise.
19390         (extended_count): Likewise.
19391         (change_zero_ext): Likewise.
19392         (simplify_comparison): Likewise.
19393         * dwarf2out.c (scompare_loc_descriptor): Likewise.
19394         (ucompare_loc_descriptor): Likewise.
19395         (minmax_loc_descriptor): Likewise.
19396         (mem_loc_descriptor): Likewise.
19397         (loc_descriptor): Likewise.
19398         * expmed.c (init_expmed_one_mode): Likewise.
19399         * lra-constraints.c (lra_constraint_offset): Likewise.
19400         * optabs.c (prepare_libcall_arg): Likewise.
19401         * postreload.c (move2add_note_store): Likewise.
19402         * reload.c (operands_match_p): Likewise.
19403         * rtl.h (load_extend_op): Likewise.
19404         * rtlhooks.c (gen_lowpart_general): Likewise.
19405         * simplify-rtx.c (simplify_truncation): Likewise.
19406         (simplify_unary_operation_1): Likewise.
19407         (simplify_binary_operation_1): Likewise.
19408         (simplify_const_binary_operation): Likewise.
19409         (simplify_const_relational_operation): Likewise.
19410         (simplify_subreg): Likewise.
19411         * stor-layout.c (bitwise_mode_for_mode): Likewise.
19412         * var-tracking.c (adjust_mems): Likewise.
19413         (prepare_call_arguments): Likewise.
19415 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19416             Alan Hayward  <alan.hayward@arm.com>
19417             David Sherwood  <david.sherwood@arm.com>
19419         * machmode.h (is_int_mode): New fuction.
19420         * combine.c (find_split_point): Use it.
19421         (combine_simplify_rtx): Likewise.
19422         (simplify_if_then_else): Likewise.
19423         (simplify_set): Likewise.
19424         (simplify_shift_const_1): Likewise.
19425         (simplify_comparison): Likewise.
19426         * config/aarch64/aarch64.c (aarch64_rtx_costs): Likewise.
19427         * cse.c (notreg_cost): Likewise.
19428         (cse_insn): Likewise.
19429         * cselib.c (cselib_lookup_1): Likewise.
19430         * dojump.c (do_jump_1): Likewise.
19431         (do_compare_rtx_and_jump): Likewise.
19432         * dse.c (get_call_args): Likewise.
19433         * dwarf2out.c (rtl_for_decl_init): Likewise.
19434         (native_encode_initializer): Likewise.
19435         * expmed.c (emit_store_flag_1): Likewise.
19436         (emit_store_flag): Likewise.
19437         * expr.c (convert_modes): Likewise.
19438         (store_field): Likewise.
19439         (expand_expr_real_1): Likewise.
19440         * fold-const.c (fold_read_from_constant_string): Likewise.
19441         * gimple-ssa-sprintf.c (get_format_string): Likewise.
19442         * optabs-libfuncs.c (gen_int_libfunc): Likewise.
19443         * optabs.c (expand_binop): Likewise.
19444         (expand_unop): Likewise.
19445         (expand_abs_nojump): Likewise.
19446         (expand_one_cmpl_abs_nojump): Likewise.
19447         * simplify-rtx.c (mode_signbit_p): Likewise.
19448         (val_signbit_p): Likewise.
19449         (val_signbit_known_set_p): Likewise.
19450         (val_signbit_known_clear_p): Likewise.
19451         (simplify_relational_operation_1): Likewise.
19452         * tree.c (vector_type_mode): Likewise.
19454 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19455             Alan Hayward  <alan.hayward@arm.com>
19456             David Sherwood  <david.sherwood@arm.com>
19458         * machmode.h (smallest_mode_for_size): Fix formatting.
19459         (smallest_int_mode_for_size): New function.
19460         * cfgexpand.c (expand_debug_expr): Use smallest_int_mode_for_size
19461         instead of smallest_mode_for_size.
19462         * combine.c (make_extraction): Likewise.
19463         * config/arc/arc.c (arc_expand_movmem): Likewise.
19464         * config/arm/arm.c (arm_expand_divmod_libfunc): Likewise.
19465         * config/i386/i386.c (ix86_get_mask_mode): Likewise.
19466         * config/s390/s390.c (s390_expand_insv): Likewise.
19467         * config/sparc/sparc.c (assign_int_registers): Likewise.
19468         * config/spu/spu.c (spu_function_value): Likewise.
19469         (spu_function_arg): Likewise.
19470         * coverage.c (get_gcov_type): Likewise.
19471         (get_gcov_unsigned_t): Likewise.
19472         * dse.c (find_shift_sequence): Likewise.
19473         * expmed.c (store_bit_field_1): Likewise.
19474         * expr.c (convert_move): Likewise.
19475         (store_field): Likewise.
19476         * internal-fn.c (expand_arith_overflow): Likewise.
19477         * optabs-query.c (get_best_extraction_insn): Likewise.
19478         * optabs.c (expand_twoval_binop_libfunc): Likewise.
19479         * stor-layout.c (layout_type): Likewise.
19480         (initialize_sizetypes): Likewise.
19481         * targhooks.c (default_get_mask_mode): Likewise.
19482         * tree-ssa-loop-manip.c (canonicalize_loop_ivs): Likewise.
19484 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19485             Alan Hayward  <alan.hayward@arm.com>
19486             David Sherwood  <david.sherwood@arm.com>
19488         * machmode.h (opt_mode::else_blk): New function.
19489         (int_mode_for_mode): Declare.
19490         * stor-layout.c (int_mode_for_mode): Return an opt_scalar_int_mode.
19491         * builtins.c (expand_builtin_signbit): Adjust for new int_mode_for_mode
19492         return type.
19493         * cfgexpand.c (expand_debug_expr): Likewise.
19494         * combine.c (gen_lowpart_or_truncate): Likewise.
19495         (gen_lowpart_for_combine): Likewise.
19496         * config/aarch64/aarch64.c (aarch64_emit_approx_sqrt): Likewise.
19497         * config/avr/avr.c (avr_to_int_mode): Likewise.
19498         (avr_out_plus_1): Likewise.
19499         (avr_out_plus): Likewise.
19500         (avr_out_round): Likewise.
19501         * config/i386/i386.c (ix86_split_to_parts): Likewise.
19502         * config/s390/s390.c (s390_expand_vec_compare_cc): Likewise.
19503         (s390_expand_vcond): Likewise.
19504         * config/spu/spu.c (spu_split_immediate): Likewise.
19505         (spu_expand_mov): Likewise.
19506         * dse.c (get_stored_val): Likewise.
19507         * expmed.c (store_bit_field_1): Likewise.
19508         (convert_extracted_bit_field): Use int_mode_for_mode instead of
19509         int_mode_for_size.
19510         (extract_bit_field_1): Adjust for new int_mode_for_mode return type.
19511         (extract_low_bits): Likewise.
19512         * expr.c (emit_group_load_1): Likewise.  Separate out the BLKmode
19513         handling rather than repeating the check.
19514         (emit_group_store): Likewise.
19515         (emit_move_via_integer): Adjust for new int_mode_for_mode return type.
19516         * optabs.c (expand_absneg_bit): Likewise.
19517         (expand_copysign_absneg): Likewise.
19518         (expand_copysign_bit): Likewise.
19519         * tree-if-conv.c (ifcvt_can_use_mask_load_store): Likewise.
19520         * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
19521         * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
19522         * var-tracking.c (prepare_call_arguments):  Likewise.
19523         * config/powerpcspe/powerpcspe.c (rs6000_do_expand_vec_perm): Use
19524         int_mode_for_mode instead of mode_for_size.
19525         * config/rs6000/rs6000.c (rs6000_do_expand_vec_perm): Likewise.
19527 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19528             Alan Hayward  <alan.hayward@arm.com>
19529             David Sherwood  <david.sherwood@arm.com>
19531         * machmode.h (int_mode_for_size): New function.
19532         * builtins.c (set_builtin_user_assembler_name): Use int_mode_for_size
19533         instead of mode_for_size.
19534         * calls.c (save_fixed_argument_area): Likewise.  Make use of BLKmode
19535         explicit.
19536         * combine.c (expand_field_assignment): Use int_mode_for_size
19537         instead of mode_for_size.
19538         (make_extraction): Likewise.
19539         (simplify_shift_const_1): Likewise.
19540         (simplify_comparison): Likewise.
19541         * dojump.c (do_jump): Likewise.
19542         * dwarf2out.c (mem_loc_descriptor): Likewise.
19543         * emit-rtl.c (init_derived_machine_modes): Likewise.
19544         * expmed.c (flip_storage_order): Likewise.
19545         (convert_extracted_bit_field): Likewise.
19546         * expr.c (copy_blkmode_from_reg): Likewise.
19547         * graphite-isl-ast-to-gimple.c (max_mode_int_precision): Likewise.
19548         * internal-fn.c (expand_mul_overflow): Likewise.
19549         * lower-subreg.c (simple_move): Likewise.
19550         * optabs-libfuncs.c (init_optabs): Likewise.
19551         * simplify-rtx.c (simplify_unary_operation_1): Likewise.
19552         * tree.c (vector_type_mode): Likewise.
19553         * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
19554         * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Likewise.
19555         * tree-vect-generic.c (expand_vector_parallel): Likewise.
19556         * tree-vect-stmts.c (vectorizable_load): Likewise.
19557         (vectorizable_store): Likewise.
19559 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19560             Alan Hayward  <alan.hayward@arm.com>
19561             David Sherwood  <david.sherwood@arm.com>
19563         * coretypes.h (pod_mode): New type.
19564         (scalar_int_mode_pod): New typedef.
19565         * machmode.h (pod_mode): New class.
19566         (int_n_data_t::m): Change type to scalar_int_mode_pod.
19567         * genmodes.c (emit_mode_int_n): Update accordingly.
19568         * lower-subreg.h (target_lower_subreg): Change type to
19569         scalar_int_mode_pod.
19570         * gdbhooks.py (build_pretty_printer): Handle pod_mode and
19571         scalar_int_mode_pod.
19573 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19574             Alan Hayward  <alan.hayward@arm.com>
19575             David Sherwood  <david.sherwood@arm.com>
19577         * config/powerpcspe/powerpcspe.h (rs6000_pmode): Change type from
19578         machine_mode to scalar_int_mode.
19579         * config/powerpcspe/powerpcspe.c (rs6000_pmode): Likewise.
19580         (rs6000_option_override_internal): Remove cast to int.
19581         * config/rs6000/rs6000.h (rs6000_pmode): Change type from
19582         machine_mode to scalar_int_mode.
19583         * config/rs6000/rs6000.c (rs6000_pmode): Likewise.
19584         (rs6000_option_override_internal): Remove cast to int.
19585         * config/s390/s390.h (Pmode): Remove cast to machine_mode.
19586         * config/epiphany/epiphany.h (RTX_OK_FOR_OFFSET_P): Add cast
19587         to machine_mode.
19588         * config/s390/s390.c (s390_expand_builtin): Likewise.
19589         * coretypes.h (scalar_int_mode): New type.
19590         (opt_scalar_int_mode): New typedef.
19591         * machmode.h (scalar_int_mode): New class.
19592         (scalar_int_mode::includes_p): New function.
19593         (byte_mode): Change type to scalar_int_mode.
19594         (word_mode): Likewise.
19595         (ptr_mode): Likewise.
19596         * emit-rtl.c (byte_mode): Likewise.
19597         (word_mode): Likewise.
19598         (ptr_mode): Likewise.
19599         (init_derived_machine_modes): Update accordingly.
19600         * genmodes.c (get_mode_class): Return scalar_int_mode for MODE_INT
19601         and MODE_PARTIAL_INT.
19602         * gdbhooks.py (build_pretty_printer): Handle scalar_int_mode and
19603         opt_scalar_int_mode.
19605 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19606             Alan Hayward  <alan.hayward@arm.com>
19607             David Sherwood  <david.sherwood@arm.com>
19609         * target.def (libgcc_floating_mode_supported_p): Take a
19610         scalar_float_mode.
19611         * doc/tm.texi: Regenerate.
19612         * targhooks.h (default_libgcc_floating_mode_supported_p): Take a
19613         scalar_float_mode.
19614         * targhooks.c (default_libgcc_floating_mode_supported_p): Likewise.
19615         * config/aarch64/aarch64.c (aarch64_libgcc_floating_mode_supported_p):
19616         Likewise.
19618 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19619             Alan Hayward  <alan.hayward@arm.com>
19620             David Sherwood  <david.sherwood@arm.com>
19622         * target.def (default_floatn_mode): Return an opt_scalar_float_mode.
19623         * doc/tm.texi: Regenerate.
19624         * config/arm/arm.c (arm_floatn_mode): Return an opt_scalar_float_mode.
19625         * config/powerpcspe/powerpcspe.c (rs6000_floatn_mode): Likewise.
19626         * config/rs6000/rs6000.c (rs6000_floatn_mode): Likewise.
19627         * targhooks.h (default_floatn_mode): Likewise.
19628         * targhooks.c (default_floatn_mode): Likewise.
19629         * tree.c (build_common_tree_nodes): Update accordingly.
19631 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19632             Alan Hayward  <alan.hayward@arm.com>
19633             David Sherwood  <david.sherwood@arm.com>
19635         * machmode.h (mode_iterator::start): Provide overload for opt_modes.
19636         (mode_iterator::iterate_p): Likewise.
19637         (mode_iterator::get_wider): Likewise.
19638         * expr.c (init_expr_target): Use opt_scalar_float_mode.
19640 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19641             Alan Hayward  <alan.hayward@arm.com>
19642             David Sherwood  <david.sherwood@arm.com>
19644         * coretypes.h (opt_scalar_float_mode): New typedef.
19645         * machmode.h (float_mode_for_size): New function.
19646         * emit-rtl.c (double_mode): Delete.
19647         (init_emit_once): Use float_mode_for_size.
19648         * stor-layout.c (layout_type): Likewise.
19649         * gdbhooks.py (build_pretty_printer): Handle opt_scalar_float_mode.
19651 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19652             Alan Hayward  <alan.hayward@arm.com>
19653             David Sherwood  <david.sherwood@arm.com>
19655         * output.h (assemble_real): Take a scalar_float_mode.
19656         * config/arm/arm.c (arm_assemble_integer): Update accordingly.
19657         * config/arm/arm.md (consttable_4): Likewise.
19658         (consttable_8): Likewise.
19659         (consttable_16): Likewise.
19660         * config/mips/mips.md (consttable_float): Likewise.
19661         * config/s390/s390.c (s390_output_pool_entry): Likewise.
19662         * varasm.c (assemble_real): Take a scalar_float_mode.
19663         (output_constant_pool_2): Update accordingly.
19664         (output_constant): Likewise.
19666 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19667             Alan Hayward  <alan.hayward@arm.com>
19668             David Sherwood  <david.sherwood@arm.com>
19670         * tree.h (SCALAR_FLOAT_TYPE_MODE): New macro.
19671         * builtins.c (expand_builtin_signbit): Use it instead of TYPE_MODE.
19672         * fold-const.c (fold_convert_const_real_from_fixed): Likewise.
19673         (native_encode_real): Likewise.
19674         (native_interpret_real): Likewise.
19675         * hsa-brig.c (emit_immediate_scalar_to_buffer): Likewise.
19676         * tree-vrp.c (simplify_float_conversion_using_ranges): Likewise.
19678 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19679             Alan Hayward  <alan.hayward@arm.com>
19680             David Sherwood  <david.sherwood@arm.com>
19682         * optabs-libfuncs.c (gen_trunc_conv_libfunc): Use is_a
19683         <scalar_float_mode>.  Simplify.
19684         (gen_extend_conv_libfunc): Likewise.
19686 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19687             Alan Hayward  <alan.hayward@arm.com>
19688             David Sherwood  <david.sherwood@arm.com>
19690         * coretypes.h (scalar_float_mode): New type.
19691         * machmode.h (mode_traits::from_int): Use machine_mode if
19692         USE_ENUM_MODES is defined.
19693         (is_a): New function.
19694         (as_a): Likewise.
19695         (dyn_cast): Likewise.
19696         (scalar_float_mode): New class.
19697         (scalar_float_mode::includes_p): New function.
19698         (is_float_mode): Likewise.
19699         * gdbhooks.py (MachineModePrinter): New class.
19700         (build_pretty_printer): Use it for scalar_float_mode.
19701         * real.h (FLOAT_MODE_FORMAT): Use as_a <scalar_float_mode>.
19702         (format_helper::format_helper): Turn into a template.
19703         * genmodes.c (get_mode_class): New function.
19704         (emit_insn_modes_h): Give modes the class returned by get_mode_class,
19705         or machine_mode if none.
19706         * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Use
19707         as_a <scalar_float_mode>.
19708         * dwarf2out.c (mem_loc_descriptor): Likewise.
19709         (insert_float): Likewise.
19710         (add_const_value_attribute): Likewise.
19711         * simplify-rtx.c (simplify_immed_subreg): Likewise.
19712         * optabs.c (expand_absneg_bit): Take a scalar_float_mode.
19713         (expand_unop): Update accordingly.
19714         (expand_abs_nojump): Likewise.
19715         (expand_copysign_absneg): Take a scalar_float_mode.
19716         (expand_copysign_bit): Likewise.
19717         (expand_copysign): Update accordingly.
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_mode): New class.
19724         * machmode.h (opt_mode): Likewise.
19725         (opt_mode::else_void): New function.
19726         (opt_mode::require): Likewise.
19727         (opt_mode::exists): Likewise.
19728         (GET_MODE_WIDER_MODE): Turn into a function and return an opt_mode.
19729         (GET_MODE_2XWIDER_MODE): Likewise.
19730         (mode_iterator::get_wider): Update accordingly.
19731         (mode_iterator::get_2xwider): Likewise.
19732         (mode_iterator::get_known_wider): Likewise, turning into a template.
19733         * combine.c (make_extraction): Update use of GET_MODE_WIDER_MODE,
19734         forcing a wider mode to exist.
19735         * config/cr16/cr16.h (LONG_REG_P): Likewise.
19736         * rtlanal.c (init_num_sign_bit_copies_in_rep): Likewise.
19737         * config/c6x/c6x.c (c6x_rtx_costs): Update use of
19738         GET_MODE_2XWIDER_MODE, forcing a wider mode to exist.
19739         * lower-subreg.c (init_lower_subreg): Likewise.
19740         * optabs-libfuncs.c (init_sync_libfuncs_1): Likewise, but not
19741         on the final iteration.
19742         * config/i386/i386.c (ix86_expand_set_or_movmem): Check whether
19743         a wider mode exists before asking for a move pattern.
19744         (get_mode_wider_vector): Update use of GET_MODE_WIDER_MODE,
19745         forcing a wider mode to exist.
19746         (expand_vselect_vconcat): Update use of GET_MODE_2XWIDER_MODE,
19747         returning false if no such mode exists.
19748         * config/ia64/ia64.c (expand_vselect_vconcat): Likewise.
19749         * config/mips/mips.c (mips_expand_vselect_vconcat): Likewise.
19750         * expmed.c (init_expmed_one_mode): Update use of GET_MODE_WIDER_MODE.
19751         Avoid checking for a MODE_INT if we already know the mode is not a
19752         SCALAR_INT_MODE_P.
19753         (extract_high_half): Update use of GET_MODE_WIDER_MODE,
19754         forcing a wider mode to exist.
19755         (expmed_mult_highpart_optab): Likewise.
19756         (expmed_mult_highpart): Likewise.
19757         * expr.c (expand_expr_real_2): Update use of GET_MODE_WIDER_MODE,
19758         using else_void.
19759         * lto-streamer-in.c (lto_input_mode_table): Likewise.
19760         * optabs-query.c (find_widening_optab_handler_and_mode): Likewise.
19761         * stor-layout.c (bit_field_mode_iterator::next_mode): Likewise.
19762         * internal-fn.c (expand_mul_overflow): Update use of
19763         GET_MODE_2XWIDER_MODE.
19764         * omp-low.c (omp_clause_aligned_alignment): Likewise.
19765         * tree-ssa-math-opts.c (convert_mult_to_widen): Update use of
19766         GET_MODE_WIDER_MODE.
19767         (convert_plusminus_to_widen): Likewise.
19768         * tree-switch-conversion.c (array_value_type): Likewise.
19769         * var-tracking.c (emit_note_insn_var_location): Likewise.
19770         * tree-vrp.c (simplify_float_conversion_using_ranges): Likewise.
19771         Return false inside rather than outside the loop if no wider mode
19772         exists
19773         * optabs.c (expand_binop): Update use of GET_MODE_WIDER_MODE
19774         and GET_MODE_2XWIDER_MODE
19775         (can_compare_p): Use else_void.
19776         * gdbhooks.py (OptMachineModePrinter): New class.
19777         (build_pretty_printer): Use it for opt_mode.
19779 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19780             Alan Hayward  <alan.hayward@arm.com>
19781             David Sherwood  <david.sherwood@arm.com>
19783         * tree-switch-conversion.c (array_value_type): Only read TYPE_MODE
19784         once.  Use get_narrowest_mode instead of GET_CLASS_NARROWEST_MODE.
19786 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19787             Alan Hayward  <alan.hayward@arm.com>
19788             David Sherwood  <david.sherwood@arm.com>
19790         * machmode.h (mode_traits): New structure.
19791         (get_narrowest_mode): New function.
19792         (mode_iterator::start): Likewise.
19793         (mode_iterator::iterate_p): Likewise.
19794         (mode_iterator::get_wider): Likewise.
19795         (mode_iterator::get_known_wider): Likewise.
19796         (mode_iterator::get_2xwider): Likewise.
19797         (FOR_EACH_MODE_IN_CLASS): New mode iterator.
19798         (FOR_EACH_MODE): Likewise.
19799         (FOR_EACH_MODE_FROM): Likewise.
19800         (FOR_EACH_MODE_UNTIL): Likewise.
19801         (FOR_EACH_WIDER_MODE): Likewise.
19802         (FOR_EACH_2XWIDER_MODE): Likewise.
19803         * builtins.c (expand_builtin_strlen): Use new mode iterators.
19804         * combine.c (simplify_comparison): Likewise
19805         * config/i386/i386.c (type_natural_mode): Likewise.
19806         * cse.c (cse_insn): Likewise.
19807         * dse.c (find_shift_sequence): Likewise.
19808         * emit-rtl.c (init_derived_machine_modes): Likewise.
19809         (init_emit_once): Likewise.
19810         * explow.c (hard_function_value): Likewise.
19811         * expmed.c (extract_fixed_bit_field_1): Likewise.
19812         (extract_bit_field_1): Likewise.
19813         (expand_divmod): Likewise.
19814         (emit_store_flag_1): Likewise.
19815         * expr.c (init_expr_target): Likewise.
19816         (convert_move): Likewise.
19817         (alignment_for_piecewise_move): Likewise.
19818         (widest_int_mode_for_size): Likewise.
19819         (emit_block_move_via_movmem): Likewise.
19820         (copy_blkmode_to_reg): Likewise.
19821         (set_storage_via_setmem): Likewise.
19822         (compress_float_constant): Likewise.
19823         * omp-low.c (omp_clause_aligned_alignment): Likewise.
19824         * optabs-query.c (get_best_extraction_insn): Likewise.
19825         * optabs.c (expand_binop): Likewise.
19826         (expand_twoval_unop): Likewise.
19827         (expand_twoval_binop): Likewise.
19828         (widen_leading): Likewise.
19829         (widen_bswap): Likewise.
19830         (expand_parity): Likewise.
19831         (expand_unop): Likewise.
19832         (prepare_cmp_insn): Likewise.
19833         (prepare_float_lib_cmp): Likewise.
19834         (expand_float): Likewise.
19835         (expand_fix): Likewise.
19836         (expand_sfix_optab): Likewise.
19837         * postreload.c (move2add_use_add2_insn): Likewise.
19838         * reg-stack.c (reg_to_stack): Likewise.
19839         * reginfo.c (choose_hard_reg_mode): Likewise.
19840         * rtlanal.c (init_num_sign_bit_copies_in_rep): Likewise.
19841         * stor-layout.c (mode_for_size): Likewise.
19842         (smallest_mode_for_size): Likewise.
19843         (mode_for_vector): Likewise.
19844         (finish_bitfield_representative): Likewise.
19845         * tree-ssa-math-opts.c (target_supports_divmod_p): Likewise.
19846         * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
19847         * tree-vect-stmts.c (vectorizable_conversion): Likewise.
19848         * var-tracking.c (prepare_call_arguments): Likewise.
19850 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19851             Alan Hayward  <alan.hayward@arm.com>
19852             David Sherwood  <david.sherwood@arm.com>
19854         * genconditions.c (write_header): Add a "#define USE_ENUM_MODES".
19855         * genmodes.c (emit_insn_modes_h): Define FOOmode to E_FOOmode if
19856         USE_ENUM_MODES is defined and to ((void) 0, E_FOOmode) otherwise.
19857         * machmode.h (mode_size): Move earlier in file.
19858         (mode_precision): Likewise.
19859         (mode_inner): Likewise.
19860         (mode_nunits): Likewise.
19861         (mode_unit_size): Likewise.
19862         (unit_unit_precision): Likewise.
19863         (mode_wider): Likewise.
19864         (mode_2xwider): Likewise.
19865         (machine_mode): New class.
19866         (mode_to_bytes): New function.
19867         (mode_to_bits): Likewise.
19868         (mode_to_precision): Likewise.
19869         (mode_to_inner): Likewise.
19870         (mode_to_unit_size): Likewise.
19871         (mode_to_unit_precision): Likewise.
19872         (mode_to_nunits): Likewise.
19873         (GET_MODE_SIZE): Use mode_to_bytes.
19874         (GET_MODE_BITSIZE): Use mode_to_bits.
19875         (GET_MODE_PRECISION): Use mode_to_precision.
19876         (GET_MODE_INNER): Use mode_to_inner.
19877         (GET_MODE_UNIT_SIZE): Use mode_to_unit_size.
19878         (GET_MODE_UNIT_PRECISION): Use mode_to_unit_precision.
19879         (GET_MODE_NUNITS): Use mode_to_nunits.
19880         * system.h (ALWAYS_INLINE): New macro.
19881         * config/powerpcspe/powerpcspe-c.c
19882         (altivec_resolve_overloaded_builtin): Use machine_mode instead of
19883         int for arg1_mode and arg2_mode.
19885 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19886             Alan Hayward  <alan.hayward@arm.com>
19887             David Sherwood  <david.sherwood@arm.com>
19889         * config/aarch64/aarch64-builtins.c (aarch64_simd_builtin_std_type):
19890         Prefix mode names with E_ in case statements.
19891         * config/aarch64/aarch64-elf.h (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
19892         * config/aarch64/aarch64.c (aarch64_split_simd_combine): Likewise.
19893         (aarch64_split_simd_move): Likewise.
19894         (aarch64_gen_storewb_pair): Likewise.
19895         (aarch64_gen_loadwb_pair): Likewise.
19896         (aarch64_gen_store_pair): Likewise.
19897         (aarch64_gen_load_pair): Likewise.
19898         (aarch64_get_condition_code_1): Likewise.
19899         (aarch64_constant_pool_reload_icode): Likewise.
19900         (get_rsqrte_type): Likewise.
19901         (get_rsqrts_type): Likewise.
19902         (get_recpe_type): Likewise.
19903         (get_recps_type): Likewise.
19904         (aarch64_gimplify_va_arg_expr): Likewise.
19905         (aarch64_simd_container_mode): Likewise.
19906         (aarch64_emit_load_exclusive): Likewise.
19907         (aarch64_emit_store_exclusive): Likewise.
19908         (aarch64_expand_compare_and_swap): Likewise.
19909         (aarch64_gen_atomic_cas): Likewise.
19910         (aarch64_emit_bic): Likewise.
19911         (aarch64_emit_atomic_swap): Likewise.
19912         (aarch64_emit_atomic_load_op): Likewise.
19913         (aarch64_evpc_trn): Likewise.
19914         (aarch64_evpc_uzp): Likewise.
19915         (aarch64_evpc_zip): Likewise.
19916         (aarch64_evpc_ext): Likewise.
19917         (aarch64_evpc_rev): Likewise.
19918         (aarch64_evpc_dup): Likewise.
19919         (aarch64_gen_ccmp_first): Likewise.
19920         (aarch64_gen_ccmp_next): Likewise.
19921         * config/alpha/alpha.c (alpha_scalar_mode_supported_p): Likewise.
19922         (alpha_emit_xfloating_libcall): Likewise.
19923         (emit_insxl): Likewise.
19924         (alpha_arg_type): Likewise.
19925         * config/arc/arc.c (arc_vector_mode_supported_p): Likewise.
19926         (arc_preferred_simd_mode): Likewise.
19927         (arc_secondary_reload): Likewise.
19928         (get_arc_condition_code): Likewise.
19929         (arc_print_operand): Likewise.
19930         (arc_legitimate_constant_p): Likewise.
19931         * config/arc/arc.h (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
19932         * config/arc/arc.md (casesi_load): Likewise.
19933         (casesi_compact_jump): Likewise.
19934         * config/arc/predicates.md (proper_comparison_operator): Likewise.
19935         (cc_use_register): Likewise.
19936         * config/arm/aout.h (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
19937         * config/arm/arm-builtins.c (arm_simd_builtin_std_type): Likewise.
19938         (arm_init_iwmmxt_builtins): Likewise.
19939         * config/arm/arm.c (thumb1_size_rtx_costs): Likewise.
19940         (neon_expand_vector_init): Likewise.
19941         (arm_attr_length_move_neon): Likewise.
19942         (maybe_get_arm_condition_code): Likewise.
19943         (arm_emit_vector_const): Likewise.
19944         (arm_preferred_simd_mode): Likewise.
19945         (arm_output_iwmmxt_tinsr): Likewise.
19946         (thumb1_output_casesi): Likewise.
19947         (thumb2_output_casesi): Likewise.
19948         (arm_emit_load_exclusive): Likewise.
19949         (arm_emit_store_exclusive): Likewise.
19950         (arm_expand_compare_and_swap): Likewise.
19951         (arm_evpc_neon_vuzp): Likewise.
19952         (arm_evpc_neon_vzip): Likewise.
19953         (arm_evpc_neon_vrev): Likewise.
19954         (arm_evpc_neon_vtrn): Likewise.
19955         (arm_evpc_neon_vext): Likewise.
19956         (arm_validize_comparison): Likewise.
19957         * config/arm/neon.md (neon_vc<cmp_op><mode>): Likewise.
19958         * config/avr/avr-c.c (avr_resolve_overloaded_builtin): Likewise.
19959         * config/avr/avr.c (avr_rtx_costs_1): Likewise.
19960         * config/c6x/c6x.c (c6x_vector_mode_supported_p): Likewise.
19961         (c6x_preferred_simd_mode): Likewise.
19962         * config/epiphany/epiphany.c (get_epiphany_condition_code): Likewise.
19963         (epiphany_rtx_costs): Likewise.
19964         * config/epiphany/predicates.md (proper_comparison_operator):
19965         Likewise.
19966         * config/frv/frv.c (condexec_memory_operand): Likewise.
19967         (frv_emit_move): Likewise.
19968         (output_move_single): Likewise.
19969         (output_condmove_single): Likewise.
19970         (frv_hard_regno_mode_ok): Likewise.
19971         (frv_matching_accg_mode): Likewise.
19972         * config/h8300/h8300.c (split_adds_subs): Likewise.
19973         (h8300_rtx_costs): Likewise.
19974         (h8300_print_operand): Likewise.
19975         (compute_mov_length): Likewise.
19976         (output_logical_op): Likewise.
19977         (compute_logical_op_length): Likewise.
19978         (compute_logical_op_cc): Likewise.
19979         (h8300_shift_needs_scratch_p): Likewise.
19980         (output_a_shift): Likewise.
19981         (compute_a_shift_length): Likewise.
19982         (compute_a_shift_cc): Likewise.
19983         (expand_a_rotate): Likewise.
19984         (output_a_rotate): Likewise.
19985         * config/i386/i386.c (classify_argument): Likewise.
19986         (function_arg_advance_32): Likewise.
19987         (function_arg_32): Likewise.
19988         (function_arg_64): Likewise.
19989         (function_value_64): Likewise.
19990         (ix86_gimplify_va_arg): Likewise.
19991         (ix86_legitimate_constant_p): Likewise.
19992         (put_condition_code): Likewise.
19993         (split_double_mode): Likewise.
19994         (ix86_avx256_split_vector_move_misalign): Likewise.
19995         (ix86_expand_vector_logical_operator): Likewise.
19996         (ix86_split_idivmod): Likewise.
19997         (ix86_expand_adjust_ufix_to_sfix_si): Likewise.
19998         (ix86_build_const_vector): Likewise.
19999         (ix86_build_signbit_mask): Likewise.
20000         (ix86_match_ccmode): Likewise.
20001         (ix86_cc_modes_compatible): Likewise.
20002         (ix86_expand_branch): Likewise.
20003         (ix86_expand_sse_cmp): Likewise.
20004         (ix86_expand_sse_movcc): Likewise.
20005         (ix86_expand_int_sse_cmp): Likewise.
20006         (ix86_expand_vec_perm_vpermi2): Likewise.
20007         (ix86_expand_vec_perm): Likewise.
20008         (ix86_expand_sse_unpack): Likewise.
20009         (ix86_expand_int_addcc): Likewise.
20010         (ix86_split_to_parts): Likewise.
20011         (ix86_vectorize_builtin_gather): Likewise.
20012         (ix86_vectorize_builtin_scatter): Likewise.
20013         (avx_vpermilp_parallel): Likewise.
20014         (inline_memory_move_cost): Likewise.
20015         (ix86_tieable_integer_mode_p): Likewise.
20016         (x86_maybe_negate_const_int): Likewise.
20017         (ix86_expand_vector_init_duplicate): Likewise.
20018         (ix86_expand_vector_init_one_nonzero): Likewise.
20019         (ix86_expand_vector_init_one_var): Likewise.
20020         (ix86_expand_vector_init_concat): Likewise.
20021         (ix86_expand_vector_init_interleave): Likewise.
20022         (ix86_expand_vector_init_general): Likewise.
20023         (ix86_expand_vector_set): Likewise.
20024         (ix86_expand_vector_extract): Likewise.
20025         (emit_reduc_half): Likewise.
20026         (ix86_emit_i387_round): Likewise.
20027         (ix86_mangle_type): Likewise.
20028         (ix86_expand_round_sse4): Likewise.
20029         (expand_vec_perm_blend): Likewise.
20030         (canonicalize_vector_int_perm): Likewise.
20031         (ix86_expand_vec_one_operand_perm_avx512): Likewise.
20032         (expand_vec_perm_1): Likewise.
20033         (expand_vec_perm_interleave3): Likewise.
20034         (expand_vec_perm_even_odd_pack): Likewise.
20035         (expand_vec_perm_even_odd_1): Likewise.
20036         (expand_vec_perm_broadcast_1): Likewise.
20037         (ix86_vectorize_vec_perm_const_ok): Likewise.
20038         (ix86_expand_vecop_qihi): Likewise.
20039         (ix86_expand_mul_widen_hilo): Likewise.
20040         (ix86_expand_sse2_abs): Likewise.
20041         (ix86_expand_pextr): Likewise.
20042         (ix86_expand_pinsr): Likewise.
20043         (ix86_preferred_simd_mode): Likewise.
20044         (ix86_simd_clone_compute_vecsize_and_simdlen): Likewise.
20045         * config/i386/sse.md (*andnot<mode>3): Likewise.
20046         (<mask_codefor><code><mode>3<mask_name>): Likewise.
20047         (*<code><mode>3): Likewise.
20048         * config/ia64/ia64.c (ia64_expand_vecint_compare): Likewise.
20049         (ia64_expand_atomic_op): Likewise.
20050         (ia64_arg_type): Likewise.
20051         (ia64_mode_to_int): Likewise.
20052         (ia64_scalar_mode_supported_p): Likewise.
20053         (ia64_vector_mode_supported_p): Likewise.
20054         (expand_vec_perm_broadcast): Likewise.
20055         * config/iq2000/iq2000.c (iq2000_move_1word): Likewise.
20056         (iq2000_function_arg_advance): Likewise.
20057         (iq2000_function_arg): Likewise.
20058         * config/m32c/m32c.c (m32c_preferred_reload_class): Likewise.
20059         * config/m68k/m68k.c (output_dbcc_and_branch): Likewise.
20060         (m68k_libcall_value): Likewise.
20061         (m68k_function_value): Likewise.
20062         (sched_attr_op_type): Likewise.
20063         * config/mcore/mcore.c (mcore_output_move): Likewise.
20064         * config/microblaze/microblaze.c (microblaze_function_arg_advance):
20065         Likewise.
20066         (microblaze_function_arg): Likewise.
20067         * config/mips/mips.c (mips16_build_call_stub): Likewise.
20068         (mips_print_operand): Likewise.
20069         (mips_mode_ok_for_mov_fmt_p): Likewise.
20070         (mips_vector_mode_supported_p): Likewise.
20071         (mips_preferred_simd_mode): Likewise.
20072         (mips_expand_vpc_loongson_even_odd): Likewise.
20073         (mips_expand_vec_unpack): Likewise.
20074         (mips_expand_vi_broadcast): Likewise.
20075         (mips_expand_vector_init): Likewise.
20076         (mips_expand_vec_reduc): Likewise.
20077         (mips_expand_msa_cmp): Likewise.
20078         * config/mips/mips.md (casesi_internal_mips16_<mode>): Likewise.
20079         * config/mn10300/mn10300.c (mn10300_print_operand): Likewise.
20080         (cc_flags_for_mode): Likewise.
20081         * config/msp430/msp430.c (msp430_print_operand): Likewise.
20082         * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Likewise.
20083         (nds32_output_casesi_pc_relative): Likewise.
20084         * config/nds32/nds32.h (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
20085         * config/nvptx/nvptx.c (nvptx_ptx_type_from_mode): Likewise.
20086         (nvptx_gen_unpack): Likewise.
20087         (nvptx_gen_pack): Likewise.
20088         (nvptx_gen_shuffle): Likewise.
20089         (nvptx_gen_wcast): Likewise.
20090         (nvptx_preferred_simd_mode): Likewise.
20091         * config/pa/pa.c (pa_secondary_reload): Likewise.
20092         * config/pa/predicates.md (base14_operand): Likewise.
20093         * config/powerpcspe/powerpcspe-c.c
20094         (altivec_resolve_overloaded_builtin): Likewise.
20095         * config/powerpcspe/powerpcspe.c (rs6000_setup_reg_addr_masks):
20096         Likewise.
20097         (rs6000_preferred_simd_mode): Likewise.
20098         (output_vec_const_move): Likewise.
20099         (rs6000_expand_vector_extract): Likewise.
20100         (rs6000_split_vec_extract_var): Likewise.
20101         (reg_offset_addressing_ok_p): Likewise.
20102         (rs6000_legitimate_offset_address_p): Likewise.
20103         (rs6000_legitimize_address): Likewise.
20104         (rs6000_emit_set_const): Likewise.
20105         (rs6000_const_vec): Likewise.
20106         (rs6000_emit_move): Likewise.
20107         (spe_build_register_parallel): Likewise.
20108         (rs6000_darwin64_record_arg_recurse): Likewise.
20109         (swap_selector_for_mode): Likewise.
20110         (spe_init_builtins): Likewise.
20111         (paired_init_builtins): Likewise.
20112         (altivec_init_builtins): Likewise.
20113         (do_load_for_compare): Likewise.
20114         (rs6000_generate_compare): Likewise.
20115         (rs6000_expand_float128_convert): Likewise.
20116         (emit_load_locked): Likewise.
20117         (emit_store_conditional): Likewise.
20118         (rs6000_output_function_epilogue): Likewise.
20119         (rs6000_handle_altivec_attribute): Likewise.
20120         (rs6000_function_value): Likewise.
20121         (emit_fusion_gpr_load): Likewise.
20122         (emit_fusion_p9_load): Likewise.
20123         (emit_fusion_p9_store): Likewise.
20124         * config/powerpcspe/predicates.md (easy_fp_constant): Likewise.
20125         (fusion_gpr_mem_load): Likewise.
20126         (fusion_addis_mem_combo_load): Likewise.
20127         (fusion_addis_mem_combo_store): Likewise.
20128         * config/rs6000/predicates.md (easy_fp_constant): Likewise.
20129         (fusion_gpr_mem_load): Likewise.
20130         (fusion_addis_mem_combo_load): Likewise.
20131         (fusion_addis_mem_combo_store): Likewise.
20132         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
20133         Likewise.
20134         * config/rs6000/rs6000-string.c (do_load_for_compare): Likewise.
20135         * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): Likewise.
20136         (rs6000_preferred_simd_mode): Likewise.
20137         (output_vec_const_move): Likewise.
20138         (rs6000_expand_vector_extract): Likewise.
20139         (rs6000_split_vec_extract_var): Likewise.
20140         (reg_offset_addressing_ok_p): Likewise.
20141         (rs6000_legitimate_offset_address_p): Likewise.
20142         (rs6000_legitimize_address): Likewise.
20143         (rs6000_emit_set_const): Likewise.
20144         (rs6000_const_vec): Likewise.
20145         (rs6000_emit_move): Likewise.
20146         (rs6000_darwin64_record_arg_recurse): Likewise.
20147         (swap_selector_for_mode): Likewise.
20148         (paired_init_builtins): Likewise.
20149         (altivec_init_builtins): Likewise.
20150         (rs6000_expand_float128_convert): Likewise.
20151         (emit_load_locked): Likewise.
20152         (emit_store_conditional): Likewise.
20153         (rs6000_output_function_epilogue): Likewise.
20154         (rs6000_handle_altivec_attribute): Likewise.
20155         (rs6000_function_value): Likewise.
20156         (emit_fusion_gpr_load): Likewise.
20157         (emit_fusion_p9_load): Likewise.
20158         (emit_fusion_p9_store): Likewise.
20159         * config/rx/rx.c (rx_gen_move_template): Likewise.
20160         (flags_from_mode): Likewise.
20161         * config/s390/predicates.md (s390_alc_comparison): Likewise.
20162         (s390_slb_comparison): Likewise.
20163         * config/s390/s390.c (s390_handle_vectorbool_attribute): Likewise.
20164         (s390_vector_mode_supported_p): Likewise.
20165         (s390_cc_modes_compatible): Likewise.
20166         (s390_match_ccmode_set): Likewise.
20167         (s390_canonicalize_comparison): Likewise.
20168         (s390_emit_compare_and_swap): Likewise.
20169         (s390_branch_condition_mask): Likewise.
20170         (s390_rtx_costs): Likewise.
20171         (s390_secondary_reload): Likewise.
20172         (__SECONDARY_RELOAD_CASE): Likewise.
20173         (s390_expand_cs): Likewise.
20174         (s390_preferred_simd_mode): Likewise.
20175         * config/s390/vx-builtins.md (vec_packsu_u<mode>): Likewise.
20176         * config/sh/sh.c (sh_print_operand): Likewise.
20177         (dump_table): Likewise.
20178         (sh_secondary_reload): Likewise.
20179         * config/sh/sh.h (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
20180         * config/sh/sh.md (casesi_worker_1): Likewise.
20181         (casesi_worker_2): Likewise.
20182         * config/sparc/predicates.md (icc_comparison_operator): Likewise.
20183         (fcc_comparison_operator): Likewise.
20184         * config/sparc/sparc.c (sparc_expand_move): Likewise.
20185         (emit_soft_tfmode_cvt): Likewise.
20186         (sparc_preferred_simd_mode): Likewise.
20187         (output_cbranch): Likewise.
20188         (sparc_print_operand): Likewise.
20189         (sparc_expand_vec_perm_bmask): Likewise.
20190         (vector_init_bshuffle): Likewise.
20191         * config/spu/spu.c (spu_scalar_mode_supported_p): Likewise.
20192         (spu_vector_mode_supported_p): Likewise.
20193         (spu_expand_insv): Likewise.
20194         (spu_emit_branch_or_set): Likewise.
20195         (spu_handle_vector_attribute): Likewise.
20196         (spu_builtin_splats): Likewise.
20197         (spu_builtin_extract): Likewise.
20198         (spu_builtin_promote): Likewise.
20199         (spu_expand_sign_extend): Likewise.
20200         * config/tilegx/tilegx.c (tilegx_scalar_mode_supported_p): Likewise.
20201         (tilegx_simd_int): Likewise.
20202         * config/tilepro/tilepro.c (tilepro_scalar_mode_supported_p): Likewise.
20203         (tilepro_simd_int): Likewise.
20204         * config/v850/v850.c (const_double_split): Likewise.
20205         (v850_print_operand): Likewise.
20206         (ep_memory_offset): Likewise.
20207         * config/vax/vax.c (vax_rtx_costs): Likewise.
20208         (vax_output_int_move): Likewise.
20209         (vax_output_int_add): Likewise.
20210         (vax_output_int_subtract): Likewise.
20211         * config/visium/predicates.md (visium_branch_operator): Likewise.
20212         * config/visium/visium.c (rtx_ok_for_offset_p): Likewise.
20213         (visium_print_operand_address): Likewise.
20214         * config/visium/visium.h (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
20215         * config/xtensa/xtensa.c (xtensa_mem_offset): Likewise.
20216         (xtensa_expand_conditional_branch): Likewise.
20217         (xtensa_copy_incoming_a7): Likewise.
20218         (xtensa_output_literal): Likewise.
20219         * dfp.c (decimal_real_maxval): Likewise.
20220         * targhooks.c (default_libgcc_floating_mode_supported_p): Likewise.
20222 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
20223             Alan Hayward  <alan.hayward@arm.com>
20224             David Sherwood  <david.sherwood@arm.com>
20226         * genmodes.c (mode_size_inline): Add an E_ prefix to mode names.
20227         (mode_nunits_inline): Likewise.
20228         (mode_inner_inline): Likewise.
20229         (mode_unit_size_inline): Likewise.
20230         (mode_unit_precision_inline): Likewise.
20231         (emit_insn_modes_h): Likewise.  Also emit a #define of the
20232         unprefixed name.
20233         (emit_mode_wider): Add an E_ prefix to mode names.
20234         (emit_mode_complex): Likewise.
20235         (emit_mode_inner): Likewise.
20236         (emit_mode_adjustments): Likewise.
20237         (emit_mode_int_n): Likewise.
20238         * config/aarch64/aarch64-builtins.c (v8qi_UP, v4hi_UP, v4hf_UP)
20239         (v2si_UP, v2sf_UP, v1df_UP, di_UP, df_UP, v16qi_UP, v8hi_UP, v8hf_UP)
20240         (v4si_UP, v4sf_UP, v2di_UP, v2df_UP, ti_UP, oi_UP, ci_UP, xi_UP)
20241         (si_UP, sf_UP, hi_UP, hf_UP, qi_UP): Likewise.
20242         (CRC32_BUILTIN, ENTRY): Likewise.
20243         * config/aarch64/aarch64.c (aarch64_push_regs): Likewise.
20244         (aarch64_pop_regs): Likewise.
20245         (aarch64_process_components): Likewise.
20246         * config/alpha/alpha.c (alpha_emit_conditional_move): Likewise.
20247         * config/arm/arm-builtins.c (v8qi_UP, v4hi_UP, v4hf_UP, v2si_UP)
20248         (v2sf_UP, di_UP, v16qi_UP, v8hi_UP, v8hf_UP, v4si_UP, v4sf_UP)
20249         (v2di_UP, ti_UP, ei_UP, oi_UP, hf_UP, si_UP, void_UP): Likewise.
20250         * config/arm/arm.c (arm_init_libfuncs): Likewise.
20251         * config/i386/i386-builtin-types.awk (ix86_builtin_type_vect_mode):
20252         Likewise.
20253         * config/i386/i386-builtin.def (pcmpestr): Likewise.
20254         (pcmpistr): Likewise.
20255         * config/microblaze/microblaze.c (double_memory_operand): Likewise.
20256         * config/mmix/mmix.c (mmix_output_condition): Likewise.
20257         * config/powerpcspe/powerpcspe.c (rs6000_init_hard_regno_mode_ok):
20258         Likewise.
20259         * config/rl78/rl78.c (mduc_regs): Likewise.
20260         * config/rs6000/rs6000.c (rs6000_init_hard_regno_mode_ok): Likewise.
20261         (htm_expand_builtin): Likewise.
20262         * config/sh/sh.h (REGISTER_NATURAL_MODE): Likewise.
20263         * config/sparc/sparc.c (emit_save_or_restore_regs): Likewise.
20264         * config/xtensa/xtensa.c (print_operand): Likewise.
20265         * expmed.h (NUM_MODE_PARTIAL_INT): Likewise.
20266         (NUM_MODE_VECTOR_INT): Likewise.
20267         * genoutput.c (null_operand): Likewise.
20268         (output_operand_data): Likewise.
20269         * genrecog.c (print_parameter_value): Likewise.
20270         * lra.c (debug_operand_data): Likewise.
20272 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
20274         * dwarf2out.c (scompare_loc_descriptor_wide)
20275         (scompare_loc_descriptor_narrow): New functions, split out from...
20276         (scompare_loc_descriptor): ...here.
20277         * expmed.c (emit_store_flag_int): New function, split out from...
20278         (emit_store_flag): ...here.
20280 2017-08-30  Richard Biener  <rguenther@suse.de>
20282         * dwarf2out.c (dwarf2out_finish): Remove setting AT_pubnames.
20283         (dwarf2out_early_finish): Move setting of AT_pubnames from
20284         early debug output to early finish.
20286 2017-08-30  Jozef Lawrynowicz  <jozef.l@somniumtech.com>
20288         * gcc/config/msp430/msp430.h: Pass -mcode/data-region to the linker
20289         and -mdata-region to the assembler.
20291 2017-08-30  Richard Biener  <rguenther@suse.de>
20293         * dwarf2out.c (add_dwarf_attr): Check we don't add duplicate
20294         attributes.
20295         (gen_subprogram_die): Add DW_AT_object_pointer only early.
20296         (dwarf2out_early_global_decl): Only generate a DIE for the
20297         abstract origin if it doesn't already exist or is a declaration DIE.
20298         (resolve_addr): Do not add the linkage name twice when
20299         generating a stub DIE for the DW_TAG_GNU_call_site target.
20301 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
20303         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
20304         Use machine_mode rather than int for arg1_mode.
20306 2017-08-29  Michael Meissner  <meissner@linux.vnet.ibm.com>
20308         PR target/82015
20309         * config/rs6000/rs6000.c (rs6000_expand_binop_builtin): Insure
20310         that the second argument of the built-in functions to unpack
20311         128-bit scalar types to 64-bit values is 0 or 1.  Change to use a
20312         switch statement instead a lot of if statements.
20313         * config/rs6000/rs6000.md (unpack<mode>, FMOVE128_VSX iterator):
20314         Allow 64-bit values to be in Altivec registers as well as
20315         traditional floating point registers.
20316         (pack<mode>, FMOVE128_VSX iterator): Likewise.
20318 2017-08-29  Alexander Monakov  <amonakov@ispras.ru>
20320         * ira-costs.c (record_address_regs): Handle both operands of PLUS for
20321         MAX_REGS_PER_ADDRESS == 1.
20323 2017-08-29  Uros Bizjak  <ubizjak@gmail.com>
20325         * config/i386/i386.opt (flag_fentry): Do not init to -1.
20326         * config/i386/i386.c (ix86_option_override_internal): Simplify
20327         setting of opts->x_flag_entry.
20329 2017-08-29  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
20330             Jakub Jelinek  <jakub@redhat.com>
20331             Richard Biener  <rguenther@suse.de>
20333         PR tree-optimization/81503
20334         * gimple-ssa-strength-reduction.c (replace_mult_candidate): Ensure
20335         folded constant fits in the target type; reorder tests for clarity.
20337 2017-08-29  Martin Liska  <mliska@suse.cz>
20339         * passes.def: Include pass_lower_switch.
20340         * stmt.c (dump_case_nodes): Remove and move to
20341         tree-switch-conversion.
20342         (case_values_threshold): Likewise.
20343         (expand_switch_as_decision_tree_p): Likewise.
20344         (emit_case_decision_tree): Likewise.
20345         (expand_case): Likewise.
20346         (balance_case_nodes): Likewise.
20347         (node_has_low_bound): Likewise.
20348         (node_has_high_bound): Likewise.
20349         (node_is_bounded): Likewise.
20350         (emit_case_nodes): Likewise.
20351         (struct simple_case_node): New struct.
20352         (add_case_node): Remove.
20353         (emit_case_dispatch_table): Use vector instead of case_list.
20354         (reset_out_edges_aux): Remove.
20355         (compute_cases_per_edge): Likewise.
20356         (expand_case): Build list of simple_case_node.
20357         (expand_sjlj_dispatch_table): Use it.
20358         * tree-switch-conversion.c (struct case_node): Moved from
20359         stmt.c and adjusted.
20360         (emit_case_nodes): Likewise.
20361         (node_has_low_bound): Likewise.
20362         (node_has_high_bound): Likewise.
20363         (node_is_bounded): Likewise.
20364         (case_values_threshold): Likewise.
20365         (reset_out_edges_aux): Likewise.
20366         (compute_cases_per_edge): Likewise.
20367         (add_case_node): Likewise.
20368         (dump_case_nodes): Likewise.
20369         (balance_case_nodes): Likewise.
20370         (expand_switch_as_decision_tree_p): Likewise.
20371         (emit_jump): Likewise.
20372         (emit_case_decision_tree): Likewise.
20373         (try_switch_expansion): Likewise.
20374         (do_jump_if_equal): Likewise.
20375         (emit_cmp_and_jump_insns): Likewise.
20376         (fix_phi_operands_for_edge): New function.
20377         (record_phi_operand_mapping): Likewise.
20378         (class pass_lower_switch): New pass.
20379         (pass_lower_switch::execute): New function.
20380         (make_pass_lower_switch): Likewise.
20381         (conditional_probability):
20382         * timevar.def: Add TV_TREE_SWITCH_LOWERING.
20383         * tree-pass.h: Add make_pass_lower_switch.
20385 2017-08-29  Jozef Lawrynowicz  <jozef.l@somniumtech.com>
20387         PR target/80993
20388         * gcc/config/msp430/msp430.c (msp430_attr): Mark interrupt
20389         handlers as used.
20391 2017-08-29  Richard Biener  <rguenther@suse.de>
20393         * dwarf2out.c (add_dwarf_attr): When checking is enabled verify
20394         we do not add a DW_AT_inline attribute twice.
20395         (gen_subprogram_die): Remove code setting DW_AT_inline on
20396         DECL_ABSTRACT_P nodes.
20398 2017-08-29  Richard Sandiford  <richard.sandiford@linaro.org>
20400         * gimplify.c (gimplify_call_expr): Copy the nothrow flag to
20401         calls to internal functions.
20402         (gimplify_modify_expr): Likewise.
20403         * tree-call-cdce.c (use_internal_fn): Likewise.
20404         * tree-ssa-math-opts.c (pass_cse_reciprocals::execute): Likewise.
20405         (convert_to_divmod): Set the nothrow flag.
20406         * tree-if-conv.c (predicate_mem_writes):  Likewise.
20407         * tree-vect-stmts.c (vectorizable_mask_load_store): Likewise.
20408         (vectorizable_call): Likewise.
20409         (vectorizable_store): Likewise.
20410         (vectorizable_load): Likewise.
20411         * tree-vect-patterns.c (vect_recog_pow_pattern): Likewise.
20412         (vect_recog_mask_conversion_pattern): Likewise.
20414 2017-08-29  Martin Liska  <mliska@suse.cz>
20416         PR other/39851
20417         * gcc.c (driver_handle_option): Add new argument.
20418         * opts-common.c (handle_option): Pass
20419         target_option_override_hook.
20420         * opts-global.c (lang_handle_option): Add new option.
20421         (set_default_handlers):  Add new argument.
20422         (decode_options): Likewise.
20423         * opts.c (target_handle_option): Likewise.
20424         (common_handle_option): Call target_option_override_hook.
20425         * opts.h (struct cl_option_handler_func): Add hook for
20426         target option override.
20427         (struct cl_option_handlers): Likewise.
20428         (set_default_handlers): Add new argument.
20429         (decode_options): Likewise.
20430         (common_handle_option): Likewise.
20431         (target_handle_option): Likewise.
20432         * toplev.c (toplev::main): Pass targetm.target_option.override
20433         hook.
20435 2017-08-29  Richard Biener  <rguenther@suse.de>
20436         Dominik Infuehr <dominik.infuehr@theobroma-systems.com>
20438         * tree-vect-slp.c (vect_bb_slp_scalar_cost): Properly confine
20439         life to the active subtree.
20441 2017-08-28  Jeff Law  <law@redhat.com>
20443         * tree-ssa-dom.c (edge_info::record_simple_equiv): Call
20444         derive_equivalences.
20445         (derive_equivalences_from_bit_ior, record_temporary_equivalences):
20446         Code moved into....
20447         (edge_info::derive_equivalences): New private member function
20449         * tree-ssa-dom.c (class edge_info): Changed from a struct
20450         to a class.  Add ctor/dtor, methods and data members.
20451         (edge_info::edge_info): Renamed from allocate_edge_info.
20452         Initialize additional members.
20453         (edge_info::~edge_info): New.
20454         (free_dom_edge_info): Delete the edge info.
20455         (record_edge_info): Use new class & associated member functions.
20456         Tighten forms for testing for edge equivalences.
20457         (record_temporary_equivalences): Iterate over the simple
20458         equivalences rather than assuming there's only one per edge.
20459         (cprop_into_successor_phis): Iterate over the simple
20460         equivalences rather than assuming there's only one per edge.
20461         (optimize_stmt): Use operand_equal_p rather than pointer
20462         equality for mini-DSE code.
20464 2017-08-28  Nathan Sidwell  <nathan@acm.org>
20466         * gcc.c (execute): Fold SIGPIPE handling into switch
20467         statement.  Adjust internal error message.
20469 2017-08-28  Richard Biener  <rguenther@suse.de>
20471         PR debug/81993
20472         * dwarf2out.c (gen_remaining_tmpl_value_param_die_attributes):
20473         Do nothing for removed DIEs.
20475 2017-08-28  Richard Biener  <rguenther@suse.de>
20477         PR tree-optimization/81977
20478         * tree-ssa-sccvn.c (vn_reference_lookup_3): Fix look through
20479         memcpy.
20481 2017-08-28  Alexander Monakov  <amonakov@ispras.ru>
20483         PR target/80640
20484         * doc/md.texi (mem_thread_fence): Remove mention of mode.  Rewrite.
20485         * optabs.c (expand_mem_thread_fence): Emit a compiler barrier when
20486         using targetm.gen_mem_thread_fence.
20488 2017-08-27  Uros Bizjak  <ubizjak@gmail.com>
20490         PR target/81995
20491         * config/i386/i386.md (*<btsc><mode>): Change operand 2
20492         predicate to register_operand.  Reorder operands.
20493         (*btr<mode>): Ditto.
20494         (*<btsc><mode>_mask): Change operand 3 predicate to register_operand.
20495         (*btr<mode>_mask): Ditto.
20497 2017-08-25  Steven Munroe  <munroesj@gcc.gnu.org>
20499         * config.gcc (powerpc*-*-*): Add xmmintrin.h and mm_malloc.h.
20500         * config/rs6000/xmmintrin.h: New file.
20501         * config/rs6000/x86intrin.h [__ALTIVEC__]: Include xmmintrin.h.
20503 2017-08-25  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
20505         PR target/81504
20506         * config/rs6000/rs6000-p8swap.c (find_alignment_op): Add reference
20507         parameter and_insn and return it.
20508         (recombine_lvx_pattern): Insert a copy to ensure availability of
20509         the base register of the copied masking operation at the point of
20510         the instruction replacement.
20511         (recombine_stvx_pattern): Likewise.
20513 2017-08-25  Michael Meissner  <meissner@linux.vnet.ibm.com>
20515         * config/rs6000/rs6000.opt (-mpower9-dform-scalar): Delete
20516         undocumented switches.
20517         (-mpower9-dform-vector): Likewise.
20518         (-mpower9-dform): Likewise.
20519         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
20520         comments to delete references to -mpower9-dform* switches.
20521         * config/rs6000/predicates.md (vsx_quad_dform_memory_operand):
20522         Delete reference to -mpower9-dform* switches, test for
20523         -mpower9-vector instead.
20524         * config/rs6000/rs6000-cpus.def (ISA_3_0_MASKS_SERVER): Likewise.
20525         (OTHER_P9_VECTOR_MASKS): Likewise.
20526         (POWERPC_MASKS): Likewise.
20527         * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): Change
20528         tests against -mpower9-dform* to -mpower9-vector.  Delete code
20529         that checked for -mpower9-dform* consistancy with other options.
20530         Add test for -mpower9-misc to enable other power9 switches.
20531         (rs6000_init_hard_regno_mode_ok): Likewise.
20532         (rs6000_option_override_internal): Likewise.
20533         (rs6000_emit_prologue): Likewise.
20534         (rs6000_emit_epilogue): Likewise.
20535         (rs6000_opt_masks): Delete -mpower9-dform-{scalar,vector}.
20536         (rs6000_disable_incompatiable_switches): Delete -mpower9-dform.
20537         (emit_fusion_p9_load): Change tests for -mpower9-dform-scalar
20538         -mpower9-vector.
20539         (emit_fusion_p9_store): Likewise.
20540         * config/rs6000/rs6000.h (TARGET_P9_DFORM_SCALAR): Delete
20541         resetting these macros if the assembler does not support ISA 3.0
20542         instructions.
20543         (TARGET_P9_DFORM_VECTOR): Likewise.
20544         * config/rs6000/rs6000.md (peepholes to optimize altivec memory):
20545         Change to use -mpower9-vector instead of -mpower9-dform-scalar.
20547 2017-08-25  Alan Modra  <amodra@gmail.com>
20549         PR rtl-optimization/81747
20550         * cse.c (cse_extended_basic_block): Don't attempt to record
20551         equivalences for degenerate conditional branches that branch
20552         to their fall-through.
20554 2017-08-24  Martin Sebor  <msebor@redhat.com>
20556         PR middle-end/81908
20557         * gimple-fold.c (size_must_be_zero_p): New function.
20558         (gimple_fold_builtin_memory_op): Call it.
20560 2017-08-24  Steven Munroe  <munroesj@gcc.gnu.org>
20562         * config/rs6000/mm_malloc.h: New file.
20564 2017-08-24  Bin Cheng  <bin.cheng@arm.com>
20566         PR tree-optimization/81913
20567         * tree-ssa-loop-niter.c (number_of_iterations_cond): Skip niter
20568         analysis when either IVs in condition can wrap.
20570 2017-08-24  Uros Bizjak  <ubizjak@gmail.com>
20572         * dwarf2out.c (MAX_ARTIFICIAL_LABEL_BYTES): Increase to 40.
20573         * dwarf2cfi.c (MAX_ARTIFICIAL_LABEL_BYTES): Remove.
20575 2017-08-24  Richard Biener  <rguenther@suse.de>
20577         PR target/81921
20578         * targhooks.c (default_target_can_inline_p): Properly
20579         use target_option_default_node when no DECL_FUNCTION_SPECIFIC_TARGET
20580         is present and always compare.
20581         * config/i386/i386.c (ix86_valid_target_attribute_tree): Do not
20582         infer -mfpmath=sse from TARGET_SSE_P.
20583         (ix86_can_inline_p): Properly use target_option_default_node when
20584         no DECL_FUNCTION_SPECIFIC_TARGET is present and always compare.
20586 2017-08-24  Richard Biener  <rguenther@suse.de>
20588         PR debug/81936
20589         * dwarf2out.c (output_die): Handle flag_generate_offload like
20590         flag_generate_lto.
20591         (output_comp_unit): Likewise.
20592         (gen_array_type_die): Likewise.
20593         (dwarf2out_early_finish): Likewise.
20594         (note_variable_value_in_expr): Likewise.
20595         (dwarf2out_finish): Likewise.  Adjust assert.
20596         * cgraphunit.c (symbol_table::compile): Move setting of
20597         flag_generate_offload earlier ...
20598         (symbol_table::finalize_compilation_unit): ... here, before
20599         early debug finalization.
20601 2017-08-24  Richard Biener  <rguenther@suse.de>
20603         * config/i386/i386.c: Include symbol-summary.h, ipa-prop.h
20604         and ipa-fnsummary.h.
20605         (ix86_can_inline_p): When ix86_fpmath flags do not match
20606         check whether the callee uses FP math at all.
20608 2017-08-24  Aldy Hernandez  <aldyh@redhat.com>
20610         PR middle-end/81931
20611         * tree-ssanames.c (get_nonzero_bits): Use element_precision
20612         instead of TYPE_PRECISION.
20614 2017-08-24  Richard Sandiford  <richard.sandiford@linaro.org>
20615             Alan Hayward  <alan.hayward@arm.com>
20616             David Sherwood  <david.sherwood@arm.com>
20618         * combine.c (make_extraction): Use subreg_offset_from_lsb.
20620 2017-08-23  Daniel Santos  <daniel.santos@pobox.com>
20622         * config/i386/i386.h (ix86_frame::stack_realign_allocate_offset):
20623         Remove field.
20624         (ix86_frame::stack_realign_allocate): New field.
20625         (struct machine_frame_state): Modify comments.
20626         (machine_frame_state::sp_realigned_fp_end): New field.
20627         * config/i386/i386.c (ix86_compute_frame_layout): Rework stack frame
20628         layout calculation.
20629         (sp_valid_at): Add assertion to assure no attempt to access invalid
20630         offset of a realigned stack.
20631         (fp_valid_at): Likewise.
20632         (choose_baseaddr): Modify comments.
20633         (ix86_emit_outlined_ms2sysv_save): Adjust to changes in
20634         ix86_expand_prologue.
20635         (ix86_expand_prologue): Modify stack realignment and allocation.
20636         (ix86_expand_epilogue): Modify comments.
20637         * doc/sourcebuild.texi: Add documentation for target selectors avx2,
20638         avx2_runtime, avx512f, and avx512f_runtime.
20640 2017-08-23  Uros Bizjak  <ubizjak@gmail.com>
20642         * config/i386/i386.opt: Remove unneeded Init(0) initializations.
20643         (mstackrealign): Do not init to -1.
20644         * config/i386/i386.c (ix86_option_override_internal):
20645         Check opts_set, not opts when setting default value of
20646         opts->x_ix86_force_align_arg_pointer.
20648 2017-08-23  Richard Biener  <rguenther@suse.de>
20650         * function.c (fndecl_name): Use verbosity 1 (no arguments) for
20651         lang_hooks.decl_printable_name.
20652         * print-rtl-function.c (print_rtx_function): Likewise.
20653         * tree-pretty-print.c (dump_function_header): Likewise.
20655 2017-08-23  Richard Biener  <rguenther@suse.de>
20657         PR lto/81940
20658         * dwarf2out.c (dwarf2out_abstract_function): Handle LTO with
20659         -g0 at compile-time.
20661 2017-08-23  Tamar Christina  <tamar.christina@arm.com>
20663         PR middle-end/19706
20664         * doc/sourcebuild.texi (Other hardware attributes):
20665         Document xorsign.
20667 2017-08-23  Tamar Christina  <tamar.christina@arm.com>
20669         PR middle-end/19706
20670         * tree-ssa-math-opts.c (convert_expand_mult_copysign):
20671         Fix single-use check.
20673 2017-08-23  Thomas Preud'homme  <thomas.preudhomme@arm.com>
20675         * gcc.c (execute): Only test for SIGKILL and SIGQUIT if available.
20677 2017-08-22  Daniel Santos  <daniel.santos@pobox.com>
20679         * doc/install.texi: Modify to add more details on running selected
20680         tests.
20682 2017-08-22  Daniel Santos  <daniel.santos@pobox.com>
20684         * config/i386/i386.c (ix86_option_override_internal): Error when -mx32
20685         is combined with -mabi=ms.
20686         (ix86_function_type_abi): Limit errors for mixing -mx32 with attribute
20687         ms_abi.
20689 2017-08-22  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
20691         PR tree-optimization/81488
20692         * gimple-ssa-strength-reduction (struct slsr_cand_d): Add visited
20693         and cached_basis fields.
20694         (MAX_SPREAD): New constant.
20695         (alloc_cand_and_find_basis): Initialize new fields.
20696         (clear_visited): New function.
20697         (create_phi_basis_1): Rename from create_phi_basis, set visited
20698         and cached_basis fields.
20699         (create_phi_basis): New wrapper function.
20700         (phi_add_costs_1): Rename from phi_add_costs, add spread
20701         parameter, set visited field, short-circuit when limits reached.
20702         (phi_add_costs): New wrapper function.
20703         (record_phi_increments_1): Rename from record_phi_increments, set
20704         visited field.
20705         (record_phi_increments): New wrapper function.
20706         (phi_incr_cost_1): Rename from phi_incr_cost, set visited field.
20707         (phi_incr_cost): New wrapper function.
20708         (all_phi_incrs_profitable_1): Rename from
20709         all_phi_incrs_profitable, set visited field.
20710         (all_phi_incrs_profitable): New wrapper function.
20712 2017-08-22  Richard Sandiford  <richard.sandiford@linaro.org>
20713             Alan Hayward  <alan.hayward@arm.com>
20714             David Sherwood  <david.sherwood@arm.com>
20716         * rtl.h (paradoxical_subreg_p): Define inline, and add a version
20717         that takes the outer and inner modes.
20718         * doc/rtl.texi: Use paradoxical_subreg_p instead of a GET_MODE_SIZE
20719         comparison as the canonical test for a paradoxical subreg.
20720         * combine.c (simplify_set): Use paradoxical_subreg_p.
20721         (make_extraction): Likewise.
20722         (force_to_mode): Likewise.
20723         (rtx_equal_for_field_assignment_p): Likewise.
20724         (gen_lowpart_for_combine): Likewise.
20725         (simplify_comparison): Likewise.
20726         * cse.c (equiv_constant): Likewise.
20727         * expmed.c (store_bit_field_1): Likewise.
20728         * final.c (alter_subreg): Likewise.
20729         * fwprop.c (propagate_rtx): Likewise.
20730         (forward_propagate_subreg): Likewise.
20731         * ira-conflicts.c (ira_build_conflicts): Likewise.
20732         * lower-subreg.c (simplify_gen_subreg_concatn): Likewise.
20733         * lra-constraints.c (curr_insn_transform): Likewise.
20734         (split_reg): Likewise.
20735         * lra-eliminations.c (move_plus_up): Likewise.
20736         (lra_eliminate_regs_1): Likewise.
20737         * recog.c (general_operand): Likewise.
20738         * ree.c (combine_reaching_defs): Likewise.
20739         * reload.c (push_reload): Likewise.
20740         (find_reloads): Likewise.
20741         * reload1.c (elimination_effects): Likewise.
20742         (compute_reload_subreg_offset): Likewise.
20743         (choose_reload_regs): Likewise.
20744         * rtlanal.c (subreg_lsb_1): Likewise.
20745         * simplify-rtx.c (simplify_unary_operation_1): Likewise.
20746         (simplify_subreg): Likewise.
20747         * var-tracking.c (track_loc_p): Likewise.
20748         * emit-rtl.c (byte_lowpart_offset): Likewise.
20749         (paradoxical_subreg_p): Delete out-of-line definition.
20751 2017-08-22  Jeff Law  <law@redhat.com>
20753         PR tree-optimization/81741
20754         PR tree-optimization/71947
20755         * tree-ssa-dom.c: Include tree-inline.h.
20756         (record_temporary_equivalences): Only record SSA_NAME = SSA_NAME
20757         equivalences if one is more expensive to compute than the other.
20758         * tree-ssa-scopedtables.h (class const_or_copies): Make
20759         record_const_or_copy_raw method private.
20760         (class avail_exprs_stack): New method simplify_binary_operation.
20761         * tree-ssa-scopedtables.c (avail_exprs_stack::lookup_avail_expr): Call
20762         avail_exprs_stack::simplify_binary_operation as needed.
20763         (avail_exprs_stack::simplify_binary_operation): New function.
20765 2017-08-22  Sebastian Huber  <sebastian.huber@embedded-brains.de>
20767         * config.gcc (powerpc-*-rtems*): Add rs6000/linux64.opt.
20768         * config/rs6000/rtems.h (ASM_PREFERRED_EH_DATA_FORMAT): New define.
20769         (DOT_SYMBOLS): Likewise.
20770         (MINIMAL_TOC_SECTION_ASM_OP): Likewise.
20771         (RELOCATABLE_NEEDS_FIXUP): Likewise.
20772         (RS6000_ABI_NAME): Likewise.
20773         (TARGET_CMODEL): Likewise.
20774         (TOC_SECTION_ASM_OP): Likewise.
20775         (SET_CMODEL): New macro.
20776         (SUBSUBTARGET_OVERRIDE_OPTIONS): Evaluate cmodel options.
20778 2017-08-22  Richard Biener  <rguenther@suse.de>
20780         * tree-inline.c (remap_type_1): Change asserts on TYPE_SIZE[_UNIT]
20781         to allow for free-lang-data replacements similar to verify_type_variant.
20783 2017-08-22  Richard Sandiford  <richard.sandiford@linaro.org>
20784             Alan Hayward  <alan.hayward@arm.com>
20785             David Sherwood  <david.sherwood@arm.com>
20787         * config/aarch64/aarch64.md (casesi): Use DImode rather than
20788         VOIDmode for the LABEL_REF.
20790 2017-08-22  Richard Biener  <rguenther@suse.de>
20792         * tree-cfg.c (gimple_split_edge): Avoid reallocating target PHI nodes.
20794 2017-08-22  Richard Biener  <rguenther@suse.de>
20796         * common.opt (feliminate-dwarf2-dups): Ignore.
20797         * doc/invoke.texi (feliminate-dwarf2-dups): Remove documentation.
20798         * dwarf2out.c (push_new_compile_unit, pop_compile_unit,
20799         same_die_p_wrap, compute_section_prefix,
20800         is_symbol_die, assign_symbol_names, break_out_includes): Remove.
20801         (comdat_symbol_id, comdat_symbol_number): Likewise.
20802         (cu_hash_table_entry, cu_hash_table_entry_hasher, cu_hash_type):
20803         Likewise.
20804         (check_duplicate_cu, record_comdat_symbol_number): Likewise.
20805         (output_die): Mark unreachable path unreachable.
20806         (dwarf2out_start_source_file): Do not create DW_TAG_GNU_BINCL.
20807         (dwarf2out_end_source_file): Do not create DW_TAG_GNU_EINCL.
20808         (dwarf2out_init): Remove code handling flag_eliminate_dwarf2_dups.
20809         (dwarf2out_early_finish): Likewise.
20811 2017-08-22  Aldy Hernandez  <aldyh@redhat.com>
20813         * wide-int.h (hwi_with_prec::hwi_with_prec): Sign extend.
20815 2017-08-22  Georg-Johann Lay  <avr@gjlay.de>
20817         PR target/81910
20818         * config/avr/avr.c (avr_handle_addr_attribute): Early return if
20819         not VAR_P. Filter attribute warnings with OPT_Wattributes.
20820         (avr_attribute_table) <io, io_low, address>: Initialize
20821         .decl_required with true.
20823 2017-08-21  Michael Meissner  <meissner@linux.vnet.ibm.com>
20825         * config/rs6000/rs6000-cpus.def (-mvsx-scalar-float): Delete
20826         undocumented debugging options.
20827         (-mvsx-scalar-double): Likewise.
20828         (-mallow-df-permute): Likewise.
20829         (-mvectorize-builtins): Likewise.
20830         * config/rs6000/rs6000.c (rs6000_init_hard_regno_mode_ok): Likewise.
20831         (rs6000_builtin_vectorized_function): Likewise.
20832         (rs6000_builtin_md_vectorized_function): Likewise.
20833         (rs6000_opt_vars): Likewise.
20835 2017-08-21  Uros Bizjak  <ubizjak@gmail.com>
20837         PR target/46091
20838         * config/i386/i386.md (*btsq_imm): Rename from *btsq.
20839         (*btrq_imm): Rename from *btrq.
20840         (*btcq_imm): Rename from *btcq.
20841         (btsc): New code attribute.
20842         (*<btsc><mode>): New insn pattern.
20843         (*btr<mode>): Ditto.
20844         (*<btsc><mode>_mask): New insn_and_split pattern.
20845         (*btr<mode>_mask): Ditto.
20847 2017-08-21  Richard Sandiford  <richard.sandiford@linaro.org>
20848             Alan Hayward  <alan.hayward@arm.com>
20849             David Sherwood  <david.sherwood@arm.com>
20851         * function.c (pad_below): Simplify padding calculation.
20853 2017-08-21  Richard Sandiford  <richard.sandiford@linaro.org>
20854             Alan Hayward  <alan.hayward@arm.com>
20855             David Sherwood  <david.sherwood@arm.com>
20857         * target.def (function_prologue): Remove frame size argument.
20858         (function_epilogue): Likewise.
20859         * doc/tm.texi: Regenerate.
20860         * final.c (final_start_function): Update call to function_prologue.
20861         (final_end_function): Update call to function_epilogue.
20862         (default_function_pro_epilogue): Remove frame size argument.
20863         * output.h (default_function_pro_epilogue): Likewise.
20864         * config/arm/arm.c (arm_output_function_epilogue): Likewise.
20865         (arm_output_function_prologue): Likewise.
20866         * config/frv/frv.c (frv_function_prologue): Likewise.
20867         (frv_function_epilogue): Likewise.
20868         * config/i386/i386.c (ix86_output_function_epilogue): Likewise.
20869         * config/ia64/ia64.c (ia64_output_function_prologue): Likewise.
20870         (ia64_output_function_epilogue): Likewise.
20871         * config/m32r/m32r.c (m32r_output_function_prologue): Likewise.
20872         (m32r_output_function_epilogue): Likewise.
20873         * config/microblaze/microblaze.c (microblaze_function_prologue)
20874         (microblaze_function_epilogue): Likewise.
20875         * config/mips/mips.c (mips_output_function_prologue): Likewise.
20876         (mips_output_function_epilogue): Likewise.
20877         * config/mmix/mmix.c (mmix_target_asm_function_prologue): Likewise.
20878         (mmix_target_asm_function_epilogue): Likewise.
20879         * config/msp430/msp430.c (msp430_start_function): Likewise.
20880         * config/nds32/nds32.c (nds32_asm_function_prologue): Likewise.
20881         (nds32_asm_function_epilogue): Likewise.
20882         * config/nios2/nios2.c (nios2_asm_function_prologue): Likewise.
20883         * config/pa/pa.c (pa_output_function_prologue): Likewise.
20884         (pa_output_function_epilogue): Likewise.
20885         * config/powerpcspe/powerpcspe.c (rs6000_output_function_prologue)
20886         (rs6000_output_function_epilogue): Likewise.
20887         * config/rl78/rl78.c (rl78_start_function): Likewise.
20888         * config/rs6000/rs6000.c (rs6000_output_function_prologue): Likewise.
20889         (rs6000_output_function_epilogue): Likewise.
20890         * config/rx/rx.c (rx_output_function_prologue): Likewise.
20891         * config/sh/sh.c (sh_output_function_epilogue): Likewise.
20892         * config/sparc/sparc.c (sparc_asm_function_prologue): Likewise.
20893         (sparc_asm_function_epilogue): Likewise.
20895 2017-08-21  Richard Sandiford  <richard.sandiford@linaro.org>
20897         * tree.h (type_has_mode_precision_p): New function.
20898         * convert.c (convert_to_integer_1): Use it.
20899         * expr.c (expand_expr_real_2): Likewise.
20900         (expand_expr_real_1): Likewise.
20901         * fold-const.c (fold_single_bit_test_into_sign_test): Likewise.
20902         * match.pd: Likewise.
20903         * tree-ssa-forwprop.c (simplify_rotate): Likewise.
20904         * tree-ssa-math-opts.c (convert_mult_to_fma): Likewise.
20905         * tree-tailcall.c (process_assignment): Likewise.
20906         * tree-vect-loop.c (vectorizable_reduction): Likewise.
20907         * tree-vect-patterns.c (vect_recog_vector_vector_shift_pattern)
20908         (vect_recog_mult_pattern, vect_recog_divmod_pattern): Likewise.
20909         * tree-vect-stmts.c (vectorizable_conversion): Likewise.
20910         (vectorizable_assignment): Likewise.
20911         (vectorizable_shift): Likewise.
20912         (vectorizable_operation): Likewise.
20913         * tree-vrp.c (register_edge_assert_for_2): Likewise.
20915 2017-08-21  Wilco Dijkstra  <wdijkstr@arm.com>
20917         * match.pd: Add pow (C, x) simplification.
20919 2017-08-21  Richard Biener  <rguenther@suse.de>
20921         PR tree-optimization/81900
20922         * tree-ssa-pre.c (compute_antic_aux): Properly compute changed
20923         for blocks with abnormal predecessors.
20924         (compute_antic): Do not set visited flag prematurely.
20926 2017-08-21  Georg-Johann Lay  <avr@gjlay.de>
20928         PR target/79883
20929         * config/avr/avr.c (avr_set_current_function): Typo in diagnostic.
20931 2017-08-21  Richard Sandiford  <richard.sandiford@linaro.org>
20933         * stor-layout.h (vector_type_mode): Move to...
20934         * tree.h (vector_type_mode): ...here.
20935         * stor-layout.c (vector_type_mode): Move to...
20936         * tree.c (vector_type_mode): ...here.  Include rtl.h and regs.h.
20938 2017-08-21  Richard Biener  <rguenther@suse.de>
20940         * debug.h (struct gcc_debug_hooks): Add die_ref_for_decl and
20941         register_external_die hooks.
20942         (debug_false_tree_charstarstar_uhwistar): Declare.
20943         (debug_nothing_tree_charstar_uhwi): Likewise.
20944         * debug.c (do_nothing_debug_hooks): Adjust.
20945         (debug_false_tree_charstarstar_uhwistar): New do nothing.
20946         (debug_nothing_tree_charstar_uhwi): Likewise.
20947         * dbxout.c (dbx_debug_hooks): Adjust.
20948         (xcoff_debug_hooks): Likewise.
20949         * sdbout.c (sdb_debug_hooks): Likewise.
20950         * vmsdbgout.c (vmsdbg_debug_hooks): Likewise.
20951         * dwarf2out.c (macinfo_label_base): New global.
20952         (dwarf2out_register_external_die): New function for the
20953         register_external_die hook.
20954         (dwarf2out_die_ref_for_decl): Likewise for die_ref_for_decl.
20955         (dwarf2_debug_hooks): Use them.
20956         (dwarf2_lineno_debug_hooks): Adjust.
20957         (struct die_struct): Add with_offset flag.
20958         (DEBUG_LTO_DWO_INFO_SECTION, DEBUG_LTO_INFO_SECTION,
20959         DEBUG_LTO_DWO_ABBREV_SECTION, DEBUG_LTO_ABBREV_SECTION,
20960         DEBUG_LTO_DWO_MACINFO_SECTION, DEBUG_LTO_MACINFO_SECTION,
20961         DEBUG_LTO_DWO_MACRO_SECTION, DEBUG_LTO_MACRO_SECTION,
20962         DEBUG_LTO_LINE_SECTION, DEBUG_LTO_DWO_STR_OFFSETS_SECTION,
20963         DEBUG_LTO_STR_DWO_SECTION, DEBUG_STR_LTO_SECTION): New macros
20964         defining section names for the early LTO debug variants.
20965         (reset_indirect_string): New helper.
20966         (add_AT_external_die_ref): Helper for dwarf2out_register_external_die.
20967         (print_dw_val): Add support for offsetted symbol references.
20968         (get_ultimate_context): Split out from is_cxx.
20969         (is_cxx): Use get_ultimate_context.
20970         (is_fortran): Add decl overload.
20971         (compute_comp_unit_symbol): Split out worker from
20972         compute_section_prefix.
20973         (compute_section_prefix): Call compute_comp_unit_symbol and
20974         set comdat_type_p here.
20975         (output_die): Skip DIE symbol output for the LTO added one.
20976         Handle DIE symbol references with offset.
20977         (output_comp_unit): Guard section name mangling properly.
20978         For LTO debug sections emit a symbol at the section beginning
20979         which we use to refer to its DIEs.
20980         (add_abstract_origin_attribute): For DIEs registered via
20981         dwarf2out_register_external_die directly refer to the early
20982         DIE rather than indirectly through the shadow one we created.
20983         Remove obsolete call to dwarf2out_abstract_function for
20984         non-function/block origins.
20985         (gen_array_type_die): When generating early LTO debug do
20986         not emit DW_AT_string_length.
20987         (gen_formal_parameter_die): Do not re-create DIEs for PARM_DECLs
20988         late when in LTO.  As suggested place a gcc_unreachable for
20989         the DECL_ABSTRACT_P case.
20990         (gen_subprogram_die): Avoid another specification DIE
20991         for early built declarations/definitions for the late LTO case.
20992         (gen_variable_die): Add type references for late duplicated VLA dies
20993         when in late LTO.
20994         (gen_inlined_subroutine_die): Do not call dwarf2out_abstract_function,
20995         we have the abstract instance already.
20996         (process_scope_var): Adjust decl DIE contexts in LTO which
20997         first puts them in limbo.
20998         (gen_decl_die): Do not generate type DIEs late apart from
20999         types for VLAs or for decls we do not yet have a DIE.  Do not
21000         call dwarf2out_abstract_function late.
21001         (dwarf2out_early_global_decl): Make sure to create DIEs
21002         for abstract instances of a decl first.
21003         (dwarf2out_late_global_decl): Adjust comment.
21004         (output_macinfo_op): With multiple macro sections use
21005         macinfo_label_base to distinguish labels.
21006         (output_macinfo): Likewise.  Update macinfo_label_base.
21007         Pass in the line info label.
21008         (note_variable_value_in_expr): When generating LTO resolve
21009         all variable values here by generating DIEs as needed.
21010         (init_sections_and_labels): Add early LTO debug flag parameter
21011         and generate different sections and names if set.  Add generation
21012         counter for the labels so we can have multiple of them.
21013         (reset_dies): Helper to allow DIEs to be output multiple times.
21014         (dwarf2out_finish): When outputting DIEs to the fat part of an
21015         LTO object first reset DIEs.
21016         (dwarf2out_early_finish): Output early DIEs when generating LTO.
21017         (modified_type_die): Check for decl_ultimate_origin being self
21018         before recursing.
21019         (gen_type_die_with_usage): Likewise.
21020         (gen_typedef_die): Allow decl_ultimate_origin being self.
21021         (set_decl_abstract_flags): Remove.
21022         (set_block_abstract_flags): Likewise.
21023         (dwarf2out_abstract_function): Treat the early generated DIEs
21024         as the abstract copy and only add DW_AT_inline and
21025         DW_AT_artificial here and call set_decl_origin_self.
21026         If the DIE has an abstract origin don't do anything.
21027         * tree.c (free_lang_data): Build a dummy TRANSLATION_UNIT_DECL
21028         if we have none yet (Go fails to build one, PR78628).
21029         (variably_modified_type_p): Prevent endless recursion for Ada
21030         cyclic pointer types.
21031         * lto-streamer-in.c: Include debug.h.
21032         (dref_queue): New global.
21033         (lto_read_tree_1): Stream in DIE references.
21034         (lto_input_tree): Register DIE references.
21035         (input_function): Stream DECL_DEBUG_ARGS.
21036         * lto-streamer-out.c: Include debug.h.
21037         (lto_write_tree_1): Output DIE references.
21038         (DFS::DFS_write_tree_body): Follow DECL_ABSTRACT_ORIGIN.
21039         Force a TRANSLATION_UNIT_DECL DECL_CONTEXT for file-scope decls.
21040         (output_function): Stream DECL_DEBUG_ARGS.
21041         * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
21042         Stream DECL_ABSTRACT_ORIGIN.
21043         * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
21044         (write_ts_decl_minimal_tree_pointers): Force a TRANSLATION_UNIT_DECL
21045         DECL_CONTEXT for file-scope decls.
21046         * lto-streamer.h (struct dref_entry): Declare.
21047         (dref_queue): Likewise.
21048         * cfgexpand.c (pass_expand::execute): Do not call the
21049         outlining_inline_function hook here.
21050         * lto-wrapper.c (debug_obj): New global.
21051         (tool_cleanup): Unlink it if required.
21052         (debug_objcopy): New function.
21053         (run_gcc): Handle early debug sections in the IL files by
21054         extracting them to separate files, partially linkin them and
21055         feeding the result back as result to the linker.
21056         * config/darwin.h (DEBUG_LTO_INFO_SECTION, DEBUG_LTO_ABBREV_SECTION,
21057         DEBUG_LTO_MACINFO_SECTION, DEBUG_LTO_LINE_SECTION,
21058         DEBUG_STR_LTO_SECTION, DEBUG_LTO_MACRO_SECTION): Put early debug
21059         sections into a separate segment.
21060         * config/darwin.c (darwin_asm_named_section): Handle __GNU_DWARF_LTO
21061         segments.
21062         (darwin_asm_dwarf_section): Likewise.
21063         (darwin_asm_output_dwarf_offset): Likewise.
21064         * config/i386/i386.c (make_resolver_func): Set DECL_IGNORED_P.
21066 2017-08-21  Richard Sandiford  <richard.sandiford@linaro.org>
21067             Alan Hayward  <alan.hayward@arm.com>
21068             David Sherwood  <david.sherwood@arm.com>
21070         * read-md.h (md_reader::record_potential_iterator_use): Replace
21071         pointer argument with an rtx and an index.
21072         * read-rtl.c (iterator_group::apply_iterator): Likewise.
21073         (apply_mode_iterator): Likewise.
21074         (apply_code_iterator): Likewise.
21075         (apply_int_iterator): Likewise.
21076         (apply_subst_iterator): Likewise.
21077         (record_iterator_use): Likewise.
21078         (record_attribute_use): Likewise.
21079         (md_reader::record_potential_iterator_use): Likewise.  Update calls
21080         to record_iterator_use and apply_iterator.
21081         (iterator_use): Replace ptr with x and index.
21082         (attribute_use): Likewise.
21083         (apply_attribute_uses): Update calls to apply_iterator.
21084         (apply_iterators): Likewise.  Update initialization of iterator_use.
21085         (rtx_reader::read_rtx_code): Update calls to record_iterator_use
21086         and record_potential_iterator_use.
21087         (rtx_reader::read_rtx_operand): Likewise.
21089 2017-08-21  Richard Sandiford  <richard.sandiford@linaro.org>
21090             Alan Hayward  <alan.hayward@arm.com>
21091             David Sherwood  <david.sherwood@arm.com>
21093         * varasm.c (const_rtx_hash_1): Don't hash in the mode of a
21094         CONST_WIDE_INT.
21096 2017-08-21  Richard Biener  <rguenther@suse.de>
21098         PR middle-end/81884
21099         * tree-ssa-alias.c (stmt_kills_ref_p): Handle array accesses
21100         at struct end conservatively when comparing common bases.
21102 2017-08-21  Richard Biener  <rguenther@suse.de>
21104         * tree-ssa-loop-im.c (struct lim_aux_data): Add ref index member.
21105         (mem_ref_in_stmt): Remove.
21106         (determine_max_movement): Use ref index to get at the reference.
21107         (invariantness_dom_walker::before_dom_children): Deal with
21108         lim data already initialized.
21109         (gather_mem_refs_stmt): Initialize lim data and record ref index.
21111 2017-08-19  Uros Bizjak  <ubizjak@gmail.com>
21113         * config/i386/i386.h (OPTION_MASK_ISA_ROUND): Remove.
21114         (TARGET_ISA_ROUND): Ditto.
21115         (TARGET_ROUND): Ditto.
21116         * config/i386/i386.c: Substitute TARGET_ROUND with TARGET_SSE4_1.
21117         * config/i386/i386.md: Ditto.
21118         * config/i386/sse.md: Ditto.
21119         * config/i386/i386-builtin.def: Substitute OPTION_MASK_ISA_ROUND
21120         with OPTION_MASK_ISA_SSE4_1.
21122 2017-08-19  Uros Bizjak  <ubizjak@gmail.com>
21124         PR target/81894
21125         * doc/extend.texi (x86 Built-in Functions): Correct the name of
21126         __builtin_ia32_lzcnt_u16.
21128 2017-08-18  Peter Bergner  <bergner@vnet.ibm.com>
21130         PR target/80210
21131         * config/rs6000/rs6000.c (rs6000_activate_target_options): New function.
21132         (rs6000_set_current_function): Rewrite function to use it.
21134 2017-08-18  H.J. Lu  <hongjiu.lu@intel.com>
21136         PR c/53037
21137         * print-tree.c (print_node): Support DECL_WARN_IF_NOT_ALIGN
21138         and TYPE_WARN_IF_NOT_ALIGN.
21139         * stor-layout.c (do_type_align): Merge DECL_WARN_IF_NOT_ALIGN.
21140         (handle_warn_if_not_align): New.
21141         (place_union_field): Call handle_warn_if_not_align.
21142         (place_field): Call handle_warn_if_not_align.
21143         Copy TYPE_WARN_IF_NOT_ALIGN.
21144         (finish_builtin_struct): Copy TYPE_WARN_IF_NOT_ALIGN.
21145         (layout_type): Likewise.
21146         * tree-core.h (tree_type_common): Add warn_if_not_align.  Set
21147         spare to 18.
21148         (tree_decl_common): Add warn_if_not_align.
21149         * tree.c (build_range_type_1): Copy TYPE_WARN_IF_NOT_ALIGN.
21150         * tree.h (TYPE_WARN_IF_NOT_ALIGN): New.
21151         (SET_TYPE_WARN_IF_NOT_ALIGN): Likewise.
21152         (DECL_WARN_IF_NOT_ALIGN): Likewise.
21153         (SET_DECL_WARN_IF_NOT_ALIGN): Likewise.
21154         * doc/extend.texi: Document warn_if_not_aligned attribute.
21155         * doc/invoke.texi: Document -Wif-not-aligned and -Wpacked-not-aligned.
21157 2017-08-17  Martin Liska  <mliska@suse.cz>
21159         PR bootstrap/81864
21160         * tree-loop-distribution.c (ddrs_table): Change type to pointer type.
21161         (get_data_dependence): Use it as pointer type.
21162         (distribute_loop): Likewise.
21164 2017-08-17  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
21166         * config/rs6000/altivec.md (UNSPEC_VMRGOW_DIRECT): New constant.
21167         (p8_vmrgew_v4sf_direct): Generalize to p8_vmrgew_<mode>_direct.
21168         (p8_vmrgow_<mode>_direct): New define_insn.
21169         * config/rs6000/rs6000.c (altivec_expand_vec_perm_const): Properly
21170         handle endianness for vmrgew and vmrgow permute patterns.
21172 2017-08-17  Peter Bergner  <bergner@vnet.ibm.com>
21174         * config/rs6000/altivec.md (VParity): Remove TARGET_VSX_TIMODE.
21175         * config/rs6000/rs6000-cpus.def: Remove comment.
21176         (ISA_2_7_MASKS_SERVER): Delete OPTION_MASK_VSX_TIMODE;
21177         (POWERPC_MASKS): Likewise.
21178         * config/rs6000/rs6000.c (rs6000_hard_regno_mode_ok): Remove unneeded
21179         use of TARGET_VSX_TIMODE.
21180         (rs6000_setup_reg_addr_masks): Change TARGET_VSX_TIMODE to TARGET_VSX.
21181         (rs6000_init_hard_regno_mode_ok): Remove unneeded uses of
21182         TARGET_VSX_TIMODE.  Change use of TARGET_VSX_TIMODE to TARGET_VSX.
21183         (rs6000_option_override_internal): Remove dead code.
21184         (rs6000_legitimize_address): Change TARGET_VSX_TIMODE to TARGET_VSX.
21185         (rs6000_legitimize_reload_address): Likewise.
21186         (rs6000_legitimate_address_p): Likewise.
21187         (rs6000_opt_masks): Delete "vsx-timode".
21188         (rs6000_disable_incompatible_switches): Remove mention of -mvsx-timode
21189         from function comment.
21190         * config/rs6000/rs6000.h (MASK_VSX_TIMODE): Delete.
21191         * config/rs6000/rs6000.md (FMOVE128_GPR): Remove TARGET_VSX_TIMODE.
21192         (V16QI, V8HI, V4SI, V4SF, V2DI, V2DF, V1TI): Remove useless empty
21193         condition.
21194         * config/rs6000/rs6000.opt (mvsx-timode): Replace with stub.
21195         * config/rs6000/vector.md (VEC_IP): Remove TARGET_VSX_TIMODE.
21196         * config/rs6000/vsx.md (VSX_LE_128): Likewise.
21197         (VSX_TI): Likewise.
21198         (VSX_M): Likewise.
21199         (define_peephole2): Likewise.
21201 2017-08-17  Martin Sebor  <msebor@redhat.com>
21203         PR c/81859
21204         * pretty-print.c (pp_format): Use strnlen in %.*s to avoid reading
21205         past the end of an array.
21206         (test_pp_format): Add test cases.
21208 2017-08-17  Richard Sandiford  <richard.sandiford@linaro.org>
21210         * internal-fn.def (CLRSB, CLZ, CTZ, FFS, PARITY, POPCOUNT): Add
21211         missing ECF_NOTHROW flags.
21213 2017-08-17  Peter Bergner  <bergner@vnet.ibm.com>
21215         PR target/72804
21216         * config/rs6000/vsx.md (*vsx_le_permute_<mode>): Add support for
21217         operands residing in integer registers.
21218         (*vsx_le_perm_load_<mode>): Likewise.
21219         (*vsx_le_perm_store_<mode>): Likewise.
21220         (define_peephole2): Add peepholes to optimize the above.
21222 2017-08-17  Marek Polacek  <polacek@redhat.com>
21224         PR middle-end/81814
21225         * fold-const.c (operand_equal_for_comparison_p): Remove code that used
21226         to mimic what shorten_compare did.  Change the return type to bool.
21227         (fold_cond_expr_with_comparison): Update call to
21228         operand_equal_for_comparison_p.
21229         (fold_ternary_loc): Likewise.
21231 2017-08-17  Jackson Woodruff  <jackson.woodruff@arm.com>
21233         * aarch64-simd.md (mov<mode>): No longer force zero immediate into
21234         register.
21235         (*aarch64_simd_mov<mode>): Add new case for stp using zero immediate.
21237 2017-08-17  Richard Biener  <rguenther@suse.de>
21239         * tree-ssa-structalias.c (solve_graph): When propagating
21240         to successors update the graphs succ edges and avoid duplicate work.
21242 2017-08-17  Maxim Ostapenko  <m.ostapenko@samsung.com>
21244         PR target/81861
21245         * config/i386/i386.c (ix86_option_override_internal): Save target
21246         specific options after ix86_stack_protector_guard_reg was changed.
21248 2017-08-17  Richard Biener  <rguenther@suse.de>
21250         PR tree-optimization/81827
21251         * tree-ssa-structalias.c (struct variable_info): Add is_reg_var flag.
21252         (new_var_info): Initialize it conservatively.
21253         (get_call_vi): Mark register vars.
21254         (new_scalar_tmp_constraint_exp): Likewise.
21255         (handle_rhs_call): Likewise.
21256         (handle_const_call): Likewise.
21257         (create_function_info_for): Likewise.
21258         (solve_constraints): Sort varinfos to separate register from
21259         non-register vars to pack points-to solution bitmaps during
21260         iteration.
21262 2017-08-17  Marek Polacek  <polacek@redhat.com>
21264         * gimplify.c (gimplify_adjust_omp_clauses): Compare with 0 instead of 1.
21266 2017-08-17  Richard Biener  <rguenther@suse.de>
21268         * tree-vrp.c (vrp_int_const_binop): Do not set *overflow_p
21269         to true when overflow is undefined and we saturated the result.
21271 2017-08-17  Alan Modra  <amodra@gmail.com>
21273         PR target/80938
21274         * config/rs6000/rs6000.c (rs6000_savres_strategy): Revert 2017-08-09.
21275         Don't use store multiple if only one reg needs saving.
21276         (interesting_frame_related_regno): New function.
21277         (rs6000_frame_related): Don't emit frame info for regs that
21278         don't need saving.
21279         (rs6000_emit_epilogue): Likewise.
21281 2017-08-16  Nathan Sidwell  <nathan@acm.org>
21283         * tree-core.h (tree_type_non_common): Rename binfo to lang_1.
21284         * tree.h (TYPE_BINFO): Use type_non_common.maxval.
21285         (TYPE_LANG_SLOT_1): Use type_non_common.lang_1, for any type.
21286         * tree.c (free_lang_data_in_type): Use else-if chain.  Always
21287         clear TYPE_LANG_1.  Remove obsolete member-function stripping.
21288         (find_decls_types_r): Comment about TYPE_MAX_VALUES_RAW.
21289         (verify_type): Adjust for TYPE_BINFO move.
21290         * lto-streamer-out.c (DFS::DFS_write_tree_body): No need to
21291         process TYPE_BINFO directly.
21292         (hash_tree): Likewise.
21293         * tree-streamer-in.c (lto_input_ts_type_non_common_tree_pointers):
21294         Likewise.
21295         * tree-streamer-out.c (write_ts_type_non_common_tree_pointers):
21296         Likewise.
21298 2017-08-16  David Malcolm  <dmalcolm@redhat.com>
21300         * diagnostic-show-locus.c (colorizer::m_caret): Remove unused field.
21302 2017-08-16  Uros Bizjak  <ubizjak@gmail.com>
21304         PR target/46091
21305         * config/i386/i386.md (*anddi_1_btr): Change predicates of
21306         operand 0 and operand 1 to nomimmediate_operand. Add "m" constraint.
21307         Add ix86_binary_operator_ok to insn constraint.
21308         (*iordi_1_bts): Ditto.
21309         (*xordi_1_btc): Ditto.
21310         (*btsq): Change predicate of operand 0 to nonimmediate_operand.
21311         Update corresponding peephole2 pattern.
21312         (*btrq): Ditto.
21313         (*btcq): Ditto.
21315 2017-08-16  Bin Cheng  <bin.cheng@arm.com>
21317         PR tree-optimization/81832
21318         * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): Don't
21319         copy loop header which has IFN_LOOP_DIST_ALIAS call.
21321 2017-08-16  Marek Polacek  <polacek@redhat.com>
21323         PR middle/81695
21324         * fold-const.c (fold_indirect_ref_1): Restore original behavior
21325         regarding size_zero_node.
21327 2017-08-16  Martin Liska  <mliska@suse.cz>
21329         PR target/81753
21330         * config.gcc: Respect previously set extra_objs in case
21331         of darwin target.
21333 2017-08-16  Richard Sandiford  <richard.sandiford@linaro.org>
21335         PR tree-optimization/81835
21336         * tree-vect-loop.c (vect_is_simple_reduction): Simply checks for
21337         the phi SSA_NAME.  Check that the condition in a COND_EXPR does
21338         not depend on the phi.
21340 2017-08-16  Alan Modra  <amodra@gmail.com>
21342         * config/rs6000/rs6000.c (rs6000_init_hard_regno_mode_ok): Delete
21343         dead code.
21345 2017-08-16  Alan Modra  <amodra@gmail.com>
21347         * config/rs6000/rs6000.c (rs6000_reg_live_or_pic_offset_p): Merge..
21348         (save_reg_p): ..into this.  Update all callers.
21349         (first_reg_to_save): Simplify.
21351 2017-08-16  Alan Modra  <amodra@gmail.com>
21353         * config/rs6000/rs6000.c (rs6000_savres_strategy): Don't restore
21354         fixed regs.
21356 2017-08-15  Joseph Myers  <joseph@codesourcery.com>
21358         PR target/78460
21359         PR target/67712
21360         * config/sh/sh-mem.cc (sh_expand_cmpnstr): Only unroll for
21361         constant count if that count is less than 32.
21363 2017-08-15  Nathan Sidwell  <nathan@acm.org>
21365         * gcc.c (execute): Emit friendlier message if inferior is killed
21366         by an external cause.
21368 2017-08-15  Richard Biener  <rguenther@suse.de>
21370         PR tree-optimization/81790
21371         * tree-ssa-sccvn.c (vn_lookup_simplify_result): Handle both
21372         CONSTRUCTORs from simplifying and VN.
21374 2017-08-14  Martin Sebor  <msebor@redhat.com>
21376         * builtin-attrs.def: Add comments.
21378 2017-08-14  Martin Sebor  <msebor@redhat.com>
21380         PR c/81117
21381         * doc/extend.texi (attribute nonstring): Document new attribute.
21383 2017-08-14  Martin Sebor  <msebor@redhat.com>
21385         PR c/81117
21386         * tree-diagnostic.c (default_tree_printer): Handle %G.
21387         * gimple-pretty-print.h (percent_G_format): Declare new function.
21388         * gimple-pretty-print.c (percent_G_format): Define.
21389         * tree-pretty-print.c (percent_K_format): Add argument.
21391 2017-08-14  Martin Sebor  <msebor@redhat.com>
21393         PR translation/79998
21394         * gimple-ssa-sprintf.c (pass_sprintf_length::handle_gimple_call):
21395         Remove a stray space.
21397 2017-08-14  Uros Bizjak  <ubizjak@gmail.com>
21399         PR target/46091
21400         * config/i386/i386.md (*anddi_1_btr): New insn_and_split pattern.
21401         (*iordi_1_bts): Ditto.
21402         (*xordi_1_btc): Ditto.
21404 2017-08-14  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
21406         PR target/79845
21407         * config/rs6000/linux64.h (INVALID_64BIT): Use quoted strings.
21408         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
21409         Likewise.
21410         * config/rs6000/rs6000.c (rs6000_init_hard_regno_mode_ok): Use
21411         quoted strings, and make more translator-friendly.
21412         (darwin_rs6000_override_options): Likewise.
21413         (rs6000_option_override_internal): Likewise.
21414         (rs6000_return_in_memory): Fix overlong line.
21415         (init_cmulative_args): Use quoted strings, and make more
21416         translator-friendly.
21417         (rs6000_pass_by_reference): Fix overlong line.
21418         (def_builtin): Use quoted strings.
21419         (altivec_expand_predicate_builtin): Use quoted strings, and make
21420         more translator-friendly.
21421         (htm_expand_builtin): Use quoted strings.
21422         (cpu_expand_builtin): Use quoted strings, and make more
21423         translator-friendly.
21424         (altivec_expand_builtin): Likewise.
21425         (paired_expand_predicate_builtin): Likewise.
21426         (rs6000_invalid_builtin): Likewise.
21427         (builtin_function_type): Use quoted strings.
21428         (rs6000_expand_split_stack_prologue): Use quoted strings, and make
21429         more translator-friendly.
21430         (rs6000_trampoline_init): Likewise.
21431         (rs6000_handle_altivec_attribute): Likewise.
21432         (rs6000_inner_target_options): Use quoted strings.
21433         (rs6000_disable_incompatible_switches): Likewise.
21434         * config/rs6000/sysv4.h (SUBTARGET_OVERRIDE_OPTIONS): Use quoted
21435         strings, and make more translator-friendly.
21436         (SUBSUBTARGET_OVERRIDE_OPTIONS): Use quoted strings.
21438 2017-08-14  Bin Cheng  <bin.cheng@arm.com>
21440         PR tree-optimization/81799
21441         * tree-loop-distribution.c (version_loop_by_alias_check): Force
21442         cond_expr to simple gimple operand.
21444 2017-08-14  Wilco Dijkstra  <wdijkstr@arm.com>
21446         PR middle-end/46932
21447         * doc/sourcebuild.texi (autoincdec): Add autoincdec description.
21449 2017-08-14  Georg-Johann Lay  <avr@gjlay.de>
21451         PR target/81754
21452         PR target/81268
21453         * config/avr/avr.opt (mgas-isr-prologues): New Var avr_gasisr_prologues.
21454         * config/avr/avr.md (gasisr, *gasisr): Use it instead of
21455         TARGET_GASISR_PROLOGUES.
21456         * config/avr/avr.c (avr_option_override): Same.
21457         (avr_pass_pre_proep::execute): Same.
21459 2017-08-13  H.J. Lu  <hongjiu.lu@intel.com>
21461         PR target/81820
21462         * config/i386/i386.c (ix86_finalize_stack_frame_flags): Replace
21463         frame pointer with stack pointer - UNITS_PER_WORD in debug insns.
21465 2017-08-13  Uros Bizjak  <ubizjak@gmail.com>
21467         * config/i386/i386.md (*load_tp_<mode>): Redefine as
21468         define_insn_and_split.  Split to a memory load from 0 in
21469         DEFAULT_TLS_SEG_REG address space.  Merge with *load_tp_x32
21470         using PTR mode iterator.
21471         (*load_tp_x32_zext"): Redefine as define_insn_and_split.
21472         Split to a memory load from 0 in DEFAULT_TLS_SEG_REG address space.
21473         (*add_tp_<mode>): Redefine as define_insn_and_split.
21474         Split to an add with a memory load from 0 in DEFAULT_TLS_SEG_REG
21475         address space.  Merge with *add_tp_x32 using PTR mode iterator.
21476         (*add_tp_x32_zext"): Redefine as define_insn_and_split.
21477         Split to an add with a  memory load from 0 in
21478         DEFAULT_TLS_SEG_REG address space.
21480 2017-08-12  Andrew Pinski  <apinski@cavium.com>
21482         * config/aarch64/aarch64-option-extensions.def (rdma):
21483         Fix feature string to what Linux prints out in /proc/cpuinfo.
21485 2017-08-12  Pierre-Marie de Rodat  <derodat@adacore.com>
21487         PR ada/79542
21488         * dwarf2out.c (modified_type_die): For C typedef types that have
21489         an ultimate origin, process the ultimate origin instead of the
21490         input type.
21491         (gen_typedef_die): Assert that input DECLs have no ultimate
21492         origin.
21493         (gen_type_die_with_usage): For typedef variants that have an
21494         ultimate origin, just call gen_decl_die on the original DECL.
21495         (process_scope_var): Avoid creating DIEs for local typedefs and
21496         concrete static variables.
21498 2017-08-12  Alan Modra  <amodra@gmail.com>
21500         PR target/81170
21501         PR target/81295
21502         * config/rs6000/sysv4.h (STARTFILE_LINUX_SPEC): Upgrade to
21503         match gnu-user.h startfile.
21504         (ENDFILE_LINUX_SPEC): Similarly.
21506 2017-08-11  Thomas Schwinge  <thomas@codesourcery.com>
21508         PR lto/81430
21509         * config/nvptx/nvptx.c (nvptx_override_options_after_change):
21510         Remove function.
21511         (TARGET_OVERRIDE_OPTIONS_AFTER_CHANGE): Remove definition.
21513 2017-08-11  Tamar Christina  <tamar.christina@arm.com>
21514         * config/aarch64/aarch64.md (mov<mode>): Change.
21515         (*movhf_aarch64, *movsf_aarch64, *movdf_aarch64):
21516         aarch64_reg_or_fp_float into aarch64_reg_or_fp_zero.
21517         * config/aarch64/predicates.md (aarch64_reg_or_fp_float): Removed.
21519 2017-08-11  Eric Botcazou  <ebotcazou@adacore.com>
21521         * tree-sra.c (build_access_from_expr_1): Use more precise diagnostics
21522         for storage order barriers.
21524 2017-08-11  Martin Liska  <mliska@suse.cz>
21526         PR tree-opt/79987
21527         * tree-chkp.c (chkp_get_bounds_for_decl_addr): Do not instrument
21528         variables of void type.
21530 2017-08-11  Martin Liska  <mliska@suse.cz>
21532         * asan.c (asan_protect_global): Replace ASM_OUTPUT_DEF with
21533         TARGET_SUPPORTS_ALIASES.
21534         * cgraph.c (cgraph_node::create_same_body_alias): Likewise.
21535         * ipa-visibility.c (can_replace_by_local_alias): Likewise.
21536         (optimize_weakref): Likewise.
21537         * symtab.c (symtab_node::noninterposable_alias): Likewise.
21538         * varpool.c (varpool_node::create_extra_name_alias): Likewise.
21539         * defaults.h: Introduce TARGET_SUPPORTS_ALIASES.
21541 2017-08-11  Martin Liska  <mliska@suse.cz>
21543         PR ipa/81213
21544         * config/i386/i386.c (make_resolver_func): Do complete
21545         refactoring of the function.
21547 2017-08-10  Uros Bizjak  <ubizjak@gmail.com>
21549         PR target/81708
21550         * config/i386/i386.opt (mstack-protector-guard-symbol=): New option
21551         * config/i386/i386.c (ix86_stack_protect_guard): Use
21552         ix86_stack_protect_guard_symbol_str to generate varible declaration.
21553         * doc/invoke.texi (x86 Options): Document
21554         -mstack-protector-guard-symbol= option.
21556 2017-08-10  Uros Bizjak  <ubizjak@gmail.com>
21558         * config/i386/i386-protos.h (ix86_split_stack_guard): New prototype.
21559         * config/i386/i386.c (ix86_split_stack_guard): New function.
21560         (ix86_xpand_split_stack_prologue): Call ix86_split_stack_guard.
21561         (ix86_legitimate_address_p) <case UNSPEC_STACK_CHECK>: Remove.
21562         (i386_asm_output_addr_const_extra) <case UNSPEC_STACK_CHECK>: Ditto.
21563         (optput_pic_addr_const): Remove UNSPEC_STACK_CHECK handling.
21564         * config/i386/i386.md (unspec): Remove UNSPEC_STACK_CHECK.
21565         (split_stack_space_check): Call ix86_split_stack_guard.
21567 2017-08-10  Martin Sebor  <msebor@redhat.com>
21569         * print-tree.c (print_node): Print location using the established
21570         format %s:%i%i.
21571         Replace spaces with colons.
21572         (debug_raw, debug): Ditto.
21574 2017-08-10  Martin Sebor  <msebor@redhat.com>
21576         PR c++/81586
21577         * pretty-print.c (pp_format): Correct the handling of %s precision.
21579 2017-08-10  H.J. Lu  <hongjiu.lu@intel.com>
21581         PR target/81736
21582         * config/i386/i386.c (ix86_finalize_stack_realign_flags): Renamed
21583         to ...
21584         (ix86_finalize_stack_frame_flags): This.  Also clear
21585         frame_pointer_needed if -fno-omit-frame-pointer is used without
21586         stack access.
21587         (ix86_expand_prologue): Replace ix86_finalize_stack_realign_flags
21588         with ix86_finalize_stack_frame_flags.
21589         (ix86_expand_epilogue): Likewise.
21590         (ix86_expand_split_stack_prologue): Likewise.
21591         * doc/invoke.texi: Add a note for -fno-omit-frame-pointer.
21593 2017-08-10  Martin Liska  <mliska@suse.cz>
21595         PR c++/81355
21596         * c-attribs.c (handle_target_attribute):
21597         Report warning for an empty string argument of target attribute.
21599 2017-08-09  Jakub Jelinek  <jakub@redhat.com>
21601         PR c/81687
21602         * omp-low.c (omp_copy_decl): Don't remap FORCED_LABEL or DECL_NONLOCAL
21603         LABEL_DECLs.
21604         * tree-cfg.c (move_stmt_op): Don't adjust DECL_CONTEXT of FORCED_LABEL
21605         or DECL_NONLOCAL labels.
21606         (move_stmt_r) <case GIMPLE_LABEL>: Adjust DECL_CONTEXT of FORCED_LABEL
21607         or DECL_NONLOCAL labels here.
21609 2017-08-09  Will Schmidt  <will_schmidt@vnet.ibm.com>
21611         * config/rs6000/rs6000.c (rs6000_option_override_internal): Add blurb
21612         to indicate when early gimple folding has been disabled.
21613         (rs6000_gimple_fold_builtin): Add debug content.
21614         (rs6000_invalid_builtin): Fix whitespace.
21615         (rs6000_expand_builtin): Fix whitespace.
21616         * config/rs6000/rs6000.opt: Add option for -mfold-gimple.
21618 2017-08-09  Segher Boessenkool  <segher@kernel.crashing.org>
21620         PR target/80938
21621         * config/rs6000/rs6000.c (rs6000_savres_strategy): Don't use
21622         SAVE_MULTIPLE if not all the registers that saves, should be saved.
21624 2017-08-09  Jim Wilson  <jim.wilson@linaro.org>
21626         * config/aarch64/aarch64-cores.def (falkor): Use falkor pipeline.
21627         (qdf24xx): Likewise.
21628         * config/aarch64/aarch64.md: Include falkor.md.
21629         * config/aarch64/falkor.md: New.
21631 2017-08-09  Marek Polacek  <polacek@redhat.com>
21633         PR c/81233
21634         * diagnostic-core.h (emit_diagnostic_valist): Add declaration.
21635         * diagnostic.c (emit_diagnostic): Add a comment.
21636         (emit_diagnostic_valist): New function.
21638 2017-08-09  Marek Polacek  <polacek@redhat.com>
21640         PR c/81417
21641         * input.c (make_location): New overload.
21642         * input.h (make_location): Declare.
21644 2017-08-08  Alan Modra  <amodra@gmail.com>
21645             H.J. Lu  <hongjiu.lu@intel.com>
21647         PR driver/81523
21648         * gcc.c (NO_PIE_SPEC): Delete.
21649         (PIE_SPEC): Define as !no-pie/pie.  Move static|shared|r
21650         exclusion..
21651         (LINK_PIE_SPEC): ..to here.
21652         (LINK_COMMAND_SPEC): Support -no-pie.
21653         * config/gnu-user.h (GNU_USER_TARGET_STARTFILE_SPEC): Correct
21654         chain of crtbegin*.o selection, update for PIE_SPEC changes and
21655         format.
21656         (GNU_USER_TARGET_ENDFILE_SPEC): Similarly.
21657         * config/sol2.h (STARTFILE_CRTBEGIN_SPEC): Similarly.
21658         (ENDFILE_CRTEND_SPEC): Similarly.
21660 2017-08-08  Uros Bizjak  <ubizjak@gmail.com>
21662         PR target/81708
21663         * config/i386/i386.opt (mstack-protector-guard-reg=): New option
21664         (mstack-protector-guard-offset=): Ditto.
21665         * config/i386/i386.c (ix86_option_override): Handle
21666         -mstack-protector-guard-reg= and -mstack-protector-guard-offset=
21667         options.
21668         (ix86_stack_protect_guard): Use ix86_stack_protect_guard_reg and
21669         ix86_stack_protect_guard_offset variables.
21670         (TARGET_STACK_PROTECT_GUARD): Always define.
21671         * doc/invoke.texi (x86 Options): Document -mstack-protector-guard-reg=
21672         and -mstack-protector-guard-offset= options.
21674 2017-08-08  Bin Cheng  <bin.cheng@arm.com>
21676         * tree-ssa-loop-ivopts.c (relate_compare_use_with_all_cands): Handle
21677         boundary case for the last candidate.
21679 2017-08-08  Bin Cheng  <bin.cheng@arm.com>
21681         * doc/invoke.texi: Document -ftree-loop-distribution for O3.
21682         * opts.c (default_options_table): Add OPT_ftree_loop_distribution.
21684 2017-08-08  Tamar Christina  <tamar.christina@arm.com>
21686         PR middle-end/19706
21687         * config/aarch64/aarch64.md (xorsign<mode>3): New optabs.
21688         * config/aarch64/aarch64-builtins.c
21689         (aarch64_builtin_vectorized_function): Added CASE_CFN_XORSIGN.
21690         * config/aarch64/aarch64-simd-builtins.def: Added xorsign BINOP.
21691         * config/aarch64/aarch64-simd.md: Added xorsign<mode>3
21693 2017-08-08  Tamar Christina  <tamar.christina@arm.com>
21694             Andrew Pinski <pinskia@gmail.com>
21696         PR middle-end/19706
21697         * internal-fn.def (XORSIGN): New.
21698         * optabs.def (xorsign_optab): New.
21699         * tree-ssa-math-opts.c (is_copysign_call_with_1): New.
21700         (convert_expand_mult_copysign): New.
21701         (pass_optimize_widening_mul::execute): Call
21702         convert_expand_mult_copysign.
21704 2017-08-08  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
21706         PR tree-optimization/81354
21707         * gimple-ssa-strength-reduction.c (create_add_on_incoming_edge):
21708         Insert on edges rather than explicitly creating landing pads.
21709         (analyze_candidates_and_replace): Commit edge inserts.
21711 2017-08-08  Richard Biener  <rguenther@suse.de>
21713         PR middle-end/81719
21714         * tree-ssa-loop-niter.c: Include tree-dfa.h.
21715         (expand_simple_operations): Also look through ADDR_EXPRs with
21716         MEM_REF bases treating them as POINTER_PLUS_EXPR.
21718 2017-08-08  Richard Biener  <rguenther@suse.de>
21720         PR tree-optimization/81723
21721         * tree-vect-slp.c (struct bst_traits): New hash traits.
21722         (bst_fail): New global.
21723         (vect_build_slp_tree_2): New worker, split out from ...
21724         (vect_build_slp_tree): ... this now wrapping it with using
21725         bst_fail set to cache SLP tree build fails.  Properly handle
21726         max_tree_size.
21727         (vect_analyze_slp_instance): Allocate and free bst_fail.
21729 2017-08-08  Martin Liska  <mliska@suse.cz>
21731         PR tree-opt/81696
21732         * ipa-icf-gimple.c (func_checker::compare_cst_or_decl): Consider
21733         LABEL_DECLs that can be from a different function.
21735 2017-08-08  Bin Cheng  <bin.cheng@arm.com>
21737         PR tree-optimization/81744
21738         * tree-predcom.c (prepare_finalizers_chain): Deep copy expr of
21739         loop's number of iterations.
21741 2017-08-08  Martin Liska  <mliska@suse.cz>
21743         * asan.c: Include header files.
21744         * attribs.c (build_decl_attribute_variant): New function moved
21745         from tree.[ch].
21746         (build_type_attribute_qual_variant): Likewise.
21747         (cmp_attrib_identifiers): Likewise.
21748         (simple_cst_list_equal): Likewise.
21749         (omp_declare_simd_clauses_equal): Likewise.
21750         (attribute_value_equal): Likewise.
21751         (comp_type_attributes): Likewise.
21752         (build_type_attribute_variant): Likewise.
21753         (lookup_ident_attribute): Likewise.
21754         (remove_attribute): Likewise.
21755         (merge_attributes): Likewise.
21756         (merge_type_attributes): Likewise.
21757         (merge_decl_attributes): Likewise.
21758         (merge_dllimport_decl_attributes): Likewise.
21759         (handle_dll_attribute): Likewise.
21760         (attribute_list_equal): Likewise.
21761         (attribute_list_contained): Likewise.
21762         * attribs.h (lookup_attribute): New function moved from tree.[ch].
21763         (lookup_attribute_by_prefix): Likewise.
21764         * bb-reorder.c: Include header files.
21765         * builtins.c: Likewise.
21766         * calls.c: Likewise.
21767         * cfgexpand.c: Likewise.
21768         * cgraph.c: Likewise.
21769         * cgraphunit.c: Likewise.
21770         * convert.c: Likewise.
21771         * dwarf2out.c: Likewise.
21772         * final.c: Likewise.
21773         * fold-const.c: Likewise.
21774         * function.c: Likewise.
21775         * gimple-expr.c: Likewise.
21776         * gimple-fold.c: Likewise.
21777         * gimple-pretty-print.c: Likewise.
21778         * gimple.c: Likewise.
21779         * gimplify.c: Likewise.
21780         * hsa-common.c: Likewise.
21781         * hsa-gen.c: Likewise.
21782         * internal-fn.c: Likewise.
21783         * ipa-chkp.c: Likewise.
21784         * ipa-cp.c: Likewise.
21785         * ipa-devirt.c: Likewise.
21786         * ipa-fnsummary.c: Likewise.
21787         * ipa-inline.c: Likewise.
21788         * ipa-visibility.c: Likewise.
21789         * ipa.c: Likewise.
21790         * lto-cgraph.c: Likewise.
21791         * omp-expand.c: Likewise.
21792         * omp-general.c: Likewise.
21793         * omp-low.c: Likewise.
21794         * omp-offload.c: Likewise.
21795         * omp-simd-clone.c: Likewise.
21796         * opts-global.c: Likewise.
21797         * passes.c: Likewise.
21798         * predict.c: Likewise.
21799         * sancov.c: Likewise.
21800         * sanopt.c: Likewise.
21801         * symtab.c: Likewise.
21802         * toplev.c: Likewise.
21803         * trans-mem.c: Likewise.
21804         * tree-chkp.c: Likewise.
21805         * tree-eh.c: Likewise.
21806         * tree-into-ssa.c: Likewise.
21807         * tree-object-size.c: Likewise.
21808         * tree-parloops.c: Likewise.
21809         * tree-profile.c: Likewise.
21810         * tree-ssa-ccp.c: Likewise.
21811         * tree-ssa-live.c: Likewise.
21812         * tree-ssa-loop.c: Likewise.
21813         * tree-ssa-sccvn.c: Likewise.
21814         * tree-ssa-structalias.c: Likewise.
21815         * tree-ssa.c: Likewise.
21816         * tree-streamer-in.c: Likewise.
21817         * tree-vectorizer.c: Likewise.
21818         * tree-vrp.c: Likewise.
21819         * tsan.c: Likewise.
21820         * ubsan.c: Likewise.
21821         * varasm.c: Likewise.
21822         * varpool.c: Likewise.
21823         * tree.c: Remove functions moved to attribs.[ch].
21824         * tree.h: Likewise.
21825         * config/aarch64/aarch64.c: Add attrs.h header file.
21826         * config/alpha/alpha.c: Likewise.
21827         * config/arc/arc.c: Likewise.
21828         * config/arm/arm.c: Likewise.
21829         * config/avr/avr.c: Likewise.
21830         * config/bfin/bfin.c: Likewise.
21831         * config/c6x/c6x.c: Likewise.
21832         * config/cr16/cr16.c: Likewise.
21833         * config/cris/cris.c: Likewise.
21834         * config/darwin.c: Likewise.
21835         * config/epiphany/epiphany.c: Likewise.
21836         * config/fr30/fr30.c: Likewise.
21837         * config/frv/frv.c: Likewise.
21838         * config/ft32/ft32.c: Likewise.
21839         * config/h8300/h8300.c: Likewise.
21840         * config/i386/winnt.c: Likewise.
21841         * config/ia64/ia64.c: Likewise.
21842         * config/iq2000/iq2000.c: Likewise.
21843         * config/lm32/lm32.c: Likewise.
21844         * config/m32c/m32c.c: Likewise.
21845         * config/m32r/m32r.c: Likewise.
21846         * config/m68k/m68k.c: Likewise.
21847         * config/mcore/mcore.c: Likewise.
21848         * config/microblaze/microblaze.c: Likewise.
21849         * config/mips/mips.c: Likewise.
21850         * config/mmix/mmix.c: Likewise.
21851         * config/mn10300/mn10300.c: Likewise.
21852         * config/moxie/moxie.c: Likewise.
21853         * config/msp430/msp430.c: Likewise.
21854         * config/nds32/nds32-isr.c: Likewise.
21855         * config/nds32/nds32.c: Likewise.
21856         * config/nios2/nios2.c: Likewise.
21857         * config/nvptx/nvptx.c: Likewise.
21858         * config/pa/pa.c: Likewise.
21859         * config/pdp11/pdp11.c: Likewise.
21860         * config/powerpcspe/powerpcspe.c: Likewise.
21861         * config/riscv/riscv.c: Likewise.
21862         * config/rl78/rl78.c: Likewise.
21863         * config/rx/rx.c: Likewise.
21864         * config/s390/s390.c: Likewise.
21865         * config/sh/sh.c: Likewise.
21866         * config/sol2.c: Likewise.
21867         * config/sparc/sparc.c: Likewise.
21868         * config/spu/spu.c: Likewise.
21869         * config/stormy16/stormy16.c: Likewise.
21870         * config/tilegx/tilegx.c: Likewise.
21871         * config/tilepro/tilepro.c: Likewise.
21872         * config/v850/v850.c: Likewise.
21873         * config/vax/vax.c: Likewise.
21874         * config/visium/visium.c: Likewise.
21875         * config/xtensa/xtensa.c: Likewise.
21877 2017-08-07  Michael Meissner  <meissner@linux.vnet.ibm.com>
21879         PR target/81593
21880         * config/rs6000/vsx.md (vsx_concat_<mode>, VSX_D): Cleanup
21881         constraints since the -mupper-regs-* switches have been
21882         eliminated.
21883         (vsx_concat_<mode>_1): New combiner insns to recognize inserting
21884         into a vector from a double word element that was extracted from
21885         another vector, and eliminate extra XXPERMDI instructions.
21886         (vsx_concat_<mode>_2): Likewise.
21887         (vsx_concat_<mode>_3): Likewise.
21888         (vsx_set_<mode>, VSX_D): Rewrite vector set in terms of vector
21889         concat to allow optimizing inserts from previous extracts.
21891 2017-08-07  Uros Bizjak  <ubizjak@gmail.com>
21893         * config/i386/i386.c (ix86_stack_protect_guard): Generate
21894         memory reference to a SSP offset in TLS address space.
21895         (ix86_print_operand) <case '@'>: Remove.
21896         (ix86_print_operand_punct_valid_p): Remove '@' code.
21897         * config/i386/i386.md (unspec): Remove UNSPEC_SP_TLS_SET and
21898         UNSPEC_SP_TLS_TEST.
21899         (stack_tls_protect_set_<mode>): Remove.
21900         (stack_protect_set): Do not call gen_stack_tls_protect_set_<mode>.
21901         (stack_tls_protect_test_<mode>): Remove.
21902         (stack_protect_test): Do not call gen_stack_tls_protect_test_<mode>.
21904 2017-08-07  Olivier Hainque  <hainque@adacore.com>
21906         PR target/81755
21907         * config/vxworksae.h (VXWORKS_HAVE_TLS): Define.
21909 2017-08-07  Douglas Rupp  <rupp@adacore.com>
21911         * Makefile.in (install-mkheaders): Fix typo, where the multi_dir
21912         variable was referenced as multidir in command.
21914 2017-08-07  Jakub Jelinek  <jakub@redhat.com>
21916         PR c/69389
21917         * gimplify.c (goa_stabilize_expr): Handle BIT_INSERT_EXPR and
21918         BIT_FIELD_REF.
21920 2017-08-07  Martin Liska  <mliska@suse.cz>
21922         * config/m32c/m32c.c: Add include of stringpool.h and attribs.h.
21923         * config/rl78/rl78.c: Add include of attribs.h.
21924         * config/sh/sh.c: Likewise.
21925         * config/v850/v850.c: Likewise.
21927 2017-08-07  Tom de Vries  <tom@codesourcery.com>
21929         PR middle-end/78266
21930         * omp-expand.c (expand_oacc_for): Ensure diff_type is large enough.
21932 2017-08-07  Martin Liska  <mliska@suse.cz>
21934         * config/mips/mips.c: Include attribs.h.
21936 2017-08-07  Thomas Koenig  <tkoenig@gcc.gnu.org>
21938         PR fortran/68829
21939         * doc/invoke.texi: Document change in behvaior for -Ofast for
21940         Fortran.
21942 2017-08-07  Wilco Dijkstra  <wdijkstr@arm.com>
21944         * config/aarch64/aarch64.c (aarch64_pushwb_single_reg):
21945         Use gen_frame_mem.
21946         (aarch64_pop_regs): Likewise.
21947         (aarch64_gen_load_pair): Likewise.
21948         (aarch64_save_callee_saves): Likewise.
21949         (aarch64_restore_callee_saves): Likewise.
21951 2017-08-07  H.J. Lu  <hongjiu.lu@intel.com>
21953         * config/i386/i386.c: Revert the last change.
21955 2017-08-07  H.J. Lu  <hongjiu.lu@intel.com>
21957         PR target/81736
21958         * config/i386/i386.c (ix86_finalize_stack_realign_flags): Renamed
21959         to ...
21960         (ix86_finalize_stack_frame_flags): This.  Also clear
21961         frame_pointer_needed if -fno-omit-frame-pointer is used without
21962         stack access.
21963         (ix86_expand_prologue): Replace ix86_finalize_stack_realign_flags
21964         with ix86_finalize_stack_frame_flags.
21965         (ix86_expand_epilogue): Likewise.
21966         (ix86_expand_split_stack_prologue): Likewise.
21968 2017-08-07  H.J. Lu  <hongjiu.lu@intel.com>
21970         PR target/81743
21971         * config/i386/i386.c (get_builtin_code_for_version): Set priority
21972         to P_AES for Westmere.
21974 2017-08-07  Jonathan Yong  <10walls@gmail.com>
21976         * config/i386/mingw.opt (fset-stack-executable): Removed.
21977         * config/i386/cygming.opt (fset-stack-executable): Moved
21978         from mingw.opt.
21979         * config/i386/cygwin.h: Define CHECK_EXECUTE_STACK_ENABLED.
21981 2017-08-07  Segher Boessenkool  <segher@kernel.crashing.org>
21983         * print-rtl.c (print_exp): Print NOT as "~" instead of as "!".
21985 2017-08-07  Marek Polacek  <polacek@redhat.com>
21987         PR middle-end/81737
21988         * fold-const.c (fold_indirect_ref_1): Check type_domain.
21990 2017-08-07  Martin Liska  <mliska@suse.cz>
21992         * attribs.h (canonicalize_attr_name): New function.
21993         (cmp_attribs): Move from c-format.c and adjusted.
21994         (is_attribute_p): Moved from tree.h.
21995         * tree-inline.c: Add new includes.
21996         * tree.c (cmp_attrib_identifiers): Use cmp_attribs.
21997         (private_is_attribute_p): Remove.
21998         (private_lookup_attribute): Likewise.
21999         (private_lookup_attribute_by_prefix): Simplify.
22000         (remove_attribute): Use is_attribute_p.
22001         * tree.h: Remove removed declarations.
22003 2017-08-07  Jakub Jelinek  <jakub@redhat.com>
22005         PR middle-end/81698
22006         * stmt.c (emit_case_dispatch_table): Add DEFAULT_EDGE argument,
22007         instead of computing it in the function.  Formatting fix.
22008         (expand_case): Don't rely on default_edge being the first edge,
22009         clear it if removing it, pass default_edge to
22010         emit_case_dispatch_table.
22011         (expand_sjlj_dispatch_table): Pass NULL as DEFAULT_EDGE, formatting
22012         fix.
22014 2017-08-06  Uros Bizjak  <ubizjak@gmail.com>
22016         * config/alpha/alpha.c (alpha_reorg): If trap is the last active
22017         insn in the function, emit NOP after the insn.
22019 2017-08-06  Tom de Vries  <tom@codesourcery.com>
22021         * omp-expand.c (expand_oacc_for): Add missing edge probability for tile
22022         and element loops.
22024 2017-08-06  Tom de Vries  <tom@codesourcery.com>
22026         * omp-expand.c (expand_oacc_for): Add missing edge probability for chunk
22027         loop.
22029 2017-08-04  Yury Gribov  <tetra2005@gmail.com>
22031         PR tree-optimization/57371
22032         * match.pd: New pattern.
22034 2017-08-04  Marek Polacek  <polacek@redhat.com>
22036         PR middle-end/81695
22037         * fold-const.c (fold_indirect_ref_1): For ((int *)&a + 4 -> a[1],
22038         perform the computation in offset_int.
22040 2017-08-04  Richard Sandiford  <richard.sandiford@linaro.org>
22042         PR tree-optimization/81136
22043         * tree-vectorizer.h: Include tree-hash-traits.h.
22044         (vec_base_alignments): New typedef.
22045         (vec_info): Add a base_alignments field.
22046         (vect_record_base_alignments): Declare.
22047         * tree-data-ref.h (data_reference): Add an is_conditional_in_stmt
22048         field.
22049         (DR_IS_CONDITIONAL_IN_STMT): New macro.
22050         (create_data_ref): Add an is_conditional_in_stmt argument.
22051         * tree-data-ref.c (create_data_ref): Likewise.  Use it to initialize
22052         the is_conditional_in_stmt field.
22053         (data_ref_loc): Add an is_conditional_in_stmt field.
22054         (get_references_in_stmt): Set the is_conditional_in_stmt field.
22055         (find_data_references_in_stmt): Update call to create_data_ref.
22056         (graphite_find_data_references_in_stmt): Likewise.
22057         * tree-ssa-loop-prefetch.c (determine_loop_nest_reuse): Likewise.
22058         * tree-vect-data-refs.c (vect_analyze_data_refs): Likewise.
22059         (vect_record_base_alignment): New function.
22060         (vect_record_base_alignments): Likewise.
22061         (vect_compute_data_ref_alignment): Adjust base_addr and aligned_to
22062         for nested statements even if we fail to compute a misalignment.
22063         Use pooled base alignments for unconditional references.
22064         (vect_find_same_alignment_drs): Compare base addresses instead
22065         of base objects.
22066         (vect_analyze_data_refs_alignment): Call vect_record_base_alignments.
22067         * tree-vect-slp.c (vect_slp_analyze_bb_1): Likewise.
22069 2017-08-04  Richard Sandiford  <richard.sandiford@linaro.org>
22071         * tree-vectorizer.h (vec_info): Add a constructor and destructor.
22072         Add an explicit name for the enum.  Use auto_vec for slp_instances
22073         and grouped_stores.
22074         (_loop_vec_info): Add a constructor and destructor.  Use auto_vec
22075         for all vectors.
22076         (_bb_vec_info): Add a constructor and destructor.
22077         (vinfo_for_stmt): Return NULL for uids of -1 as well.
22078         (destroy_loop_vec_info): Delete.
22079         (vect_destroy_datarefs): Likewise.
22080         * tree-vectorizer.c (vect_destroy_datarefs): Delete.
22081         (vec_info::vec_info): New function.
22082         (vec_info::~vec_info): Likewise.
22083         (vectorize_loops): Use delete instead of destroy_loop_vec_info.
22084         * tree-parloops.c (gather_scalar_reductions): Use delete instead of
22085         destroy_loop_vec_info.
22086         * tree-vect-loop.c (new_loop_vec_info): Replace with...
22087         (_loop_vec_info::_loop_vec_info): ...this.
22088         (destroy_loop_vec_info): Replace with...
22089         (_loop_vec_info::~_loop_vec_info): ...this.  Unconditionally delete
22090         the stmt_vec_infos.  Leave handling of vec_info information to its
22091         destructor.  Remove explicit vector releases.
22092         (vect_analyze_loop_form): Use new instead of new_loop_vec_info.
22093         (vect_analyze_loop): Use delete instead of destroy_loop_vec_info.
22094         * tree-vect-slp.c (new_bb_vec_info): Replace with...
22095         (_bb_vec_info::_bb_vec_info): ...this.  Don't reserve space in
22096         BB_VINFO_GROUPED_STORES or BB_VINFO_SLP_INSTANCES.
22097         (destroy_bb_vec_info): Replace with...
22098         (_bb_vec_info::~_bb_vec_info): ...this.  Leave handling of vec_info
22099         information to its destructor.
22100         (vect_slp_analyze_bb_1): Use new and delete instead of
22101         new_bb_vec_info and destroy_bb_vec_info.
22102         (vect_slp_bb): Replace 2 calls to destroy_bb_vec_info with a
22103         single delete.
22105 2017-08-04  Richard Sandiford  <richard.sandiford@linaro.org>
22107         * tree-data-ref.h (subscript): Add access_fn field.
22108         (data_dependence_relation): Add could_be_independent_p.
22109         (SUB_ACCESS_FN, DDR_COULD_BE_INDEPENDENT_P): New macros.
22110         (same_access_functions): Move to tree-data-ref.c.
22111         * tree-data-ref.c (ref_contains_union_access_p): New function.
22112         (access_fn_component_p): Likewise.
22113         (access_fn_components_comparable_p): Likewise.
22114         (dr_analyze_indices): Add a reference to access_fn_component_p.
22115         (dump_data_dependence_relation): Use SUB_ACCESS_FN instead of
22116         DR_ACCESS_FN.
22117         (constant_access_functions): Likewise.
22118         (add_other_self_distances): Likewise.
22119         (same_access_functions): Likewise.  (Moved from tree-data-ref.h.)
22120         (initialize_data_dependence_relation): Use XCNEW and remove
22121         explicit zeroing of DDR_REVERSED_P.  Look for a subsequence
22122         of access functions that have the same type.  Allow the
22123         subsequence to end with different bases in some circumstances.
22124         Record the chosen access functions in SUB_ACCESS_FN.
22125         (build_classic_dist_vector_1): Replace ddr_a and ddr_b with
22126         a_index and b_index.  Use SUB_ACCESS_FN instead of DR_ACCESS_FN.
22127         (subscript_dependence_tester_1): Likewise dra and drb.
22128         (build_classic_dist_vector): Update calls accordingly.
22129         (subscript_dependence_tester): Likewise.
22130         * tree-ssa-loop-prefetch.c (determine_loop_nest_reuse): Check
22131         DDR_COULD_BE_INDEPENDENT_P.
22132         * tree-vectorizer.h (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Test
22133         comp_alias_ddrs instead of may_alias_ddrs.
22134         * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr):
22135         New function.
22136         (vect_analyze_data_ref_dependence): Use it if
22137         DDR_COULD_BE_INDEPENDENT_P, but fall back to using the recorded
22138         distance vectors if that fails.
22139         (dependence_distance_ge_vf): New function.
22140         (vect_prune_runtime_alias_test_list): Use it.  Don't clear
22141         LOOP_VINFO_MAY_ALIAS_DDRS.
22143 2017-08-04  Richard Biener  <rguenther@suse.de>
22145         PR middle-end/81705
22146         * fold-const.c (fold_binary_loc): Properly restrict
22147         minus_var0 && minus_var1 case when associating undefined overflow
22148         entities.
22150 2017-08-04  Tom de Vries  <tom@codesourcery.com>
22152         * omp-simd-clone.c (simd_clone_adjust): Add missing edge probability.
22154 2017-08-03  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
22156         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
22157         Don't start diagnostic messages with a capital letter.
22158         * config/rs6000/rs6000.c (rs6000_option_override_internal):
22159         Likewise.
22160         (rs6000_invalid_builtin): Likewise.
22161         (rs6000_trampoline_init): Likewise.
22163 2017-08-03  Jakub Jelinek  <jakub@redhat.com>
22165         PR target/81621
22166         * bb-reorder.c (pass_partition_blocks::execute): Return TODO_df_finish
22167         after setting changeable df flags.
22169 2017-08-03  Richard Biener  <rguenther@suse.de>
22171         * tree-ssa-reassoc.c (should_break_up_subtract): Also break
22172         up if the use is in USE - X.
22174 2017-08-03  Alexander Monakov  <amonakov@ispras.ru>
22176         * toplev.c (dumpfile.h): New include.
22177         (internal_error_reentered): New static function.  Use it...
22178         (internal_error_function): ...here to handle reentered internal_error.
22180 2017-08-03  Richard Biener  <rguenther@suse.de>
22182         PR middle-end/81148
22183         * fold-const.c (split_tree): Add minus_var and minus_con
22184         arguments, remove unused loc arg.  Never generate NEGATE_EXPRs
22185         here but always use minus_*.
22186         (associate_trees): Assert we never associate with MINUS_EXPR
22187         and NULL first operand.  Do not recurse for PLUS_EXPR operands
22188         when associating as MINUS_EXPR either.
22189         (fold_binary_loc): Track minus_var and minus_con.
22191 2017-08-03  Tom de Vries  <tom@codesourcery.com>
22193         PR lto/81430
22194         * tree-streamer-in.c (lto_input_ts_function_decl_tree_pointers): If
22195         ACCEL_COMPILER, apply finish_options on
22196         DECL_FUNCTION_SPECIFIC_OPTIMIZATION.
22198 2017-08-03  Tom de Vries  <tom@codesourcery.com>
22200         PR target/81662
22201         * config/nvptx/nvptx.c (nvptx_option_override): Emit sorry if
22202         function_entry_patch_area_size > 0.
22204 2017-08-03  Jakub Jelinek  <jakub@redhat.com>
22206         PR driver/81650
22207         * calls.c (alloc_max_size): Use HOST_WIDE_INT_UC (10??)
22208         instead of 10??LU, perform unit multiplication in wide_int,
22209         don't change alloc_object_size_limit if the limit is larger
22210         than SSIZE_MAX.
22212         PR tree-optimization/81655
22213         PR tree-optimization/81588
22214         * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Handle also
22215         the case when ranges[i].low and high are 1 for unsigned type with
22216         precision 1.
22218         PR middle-end/81052
22219         * omp-low.c (diagnose_sb_0): Handle flag_openmp_simd like flag_openmp.
22220         (pass_diagnose_omp_blocks::gate): Enable also for flag_openmp_simd.
22222 2017-08-03  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
22224         * tree-vrp.h: Add include guard.
22226 2017-08-02  Uros Bizjak  <ubizjak@gmail.com>
22228         PR target/81644
22229         * config/i386/i386.md (unspecv): Add UNSPECV_UD2.
22230         (ud2): New insn pattern.
22231         * config/i386/i386.c (ix86_expand_epilogue):
22232         For naked functions, generate ud2 instead of trap insn.
22234 2017-08-02  Marek Polacek  <polacek@redhat.com>
22236         PR other/81667
22237         * alloc-pool.h (base_pool_allocator): Initialize m_elt_size.
22239 2017-08-02  Tom de Vries  <tom@codesourcery.com>
22240             Cesar Philippidis  <cesar@codesourcery.com>
22242         * config/nvptx/nvptx.c (nvptx_lockless_update, nvptx_lockfull_update):
22243         Add missing edge probabilities.
22245 2017-08-02  Tamar Christina  <tamar.christina@arm.com>
22247         * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
22248         Correct endianness.
22250 2017-08-02  Jakub Jelinek  <jakub@redhat.com>
22252         PR middle-end/79499
22253         * function.c (thread_prologue_and_epilogue_insns): Determine blocks
22254         for find_many_sub_basic_blocks bitmap by looking up BLOCK_FOR_INSN
22255         of first NONDEBUG_INSN_P in each of the split_prologue_seq and
22256         prologue_seq sequences - if any.
22258 2017-08-02  Richard Biener  <rguenther@suse.de>
22260         * tree-vect-stmts.c (vectorizable_store): Perform vector extracts
22261         via vectors if supported, integer extracts via punning if supported
22262         or otherwise vector extracts.
22264 2017-08-02  Richard Biener  <rguenther@suse.de>
22266         * tree-ssa-pre.c (bitmap_insert_into_set_1): Remove and inline
22267         into ...
22268         (bitmap_insert_into_set): ... this.
22270 2017-08-02  Richard Biener  <rguenther@suse.de>
22272         PR tree-optimization/81633
22273         Revert
22274         2015-08-17  Alan Hayward  <alan.hayward@arm.com>
22276         PR tree-optimization/71752
22277         * tree-vect-slp.c (vect_get_slp_defs): Handle null operands.
22279 2017-08-01  Daniel Santos  <daniel.santos@pobox.com>
22281         * config/i386/i386.h (ix86_frame::outlined_save_offset): Remove field.
22282         (machine_function::call_ms2sysv_pad_out): Remove field.
22283         * config/i386/i386.c (xlogue_layout::get_stack_space_used): Modify.
22284         (ix86_compute_frame_layout): Likewise.
22286 2017-08-01  H.J. Lu  <hongjiu.lu@intel.com>
22288         PR target/81654
22289         * config/i386/i386.c (ix86_set_func_type): Disallow naked
22290         attribute with interrupt attribute.
22292 2017-08-01  Andrew Pinski  <apinski@cavium.com>
22294         * tree-ssa-scopedtables.c (hashable_expr_equal_p): Check
22295         BIT_INSERT_EXPR's operand 1
22296         to see if the types precision matches.
22298 2017-08-01  Martin Liska  <mliska@suse.cz>
22300         PR middle-end/70140
22301         * builtins.c (expand_builtin_memcpy_args): Remove.
22302         (expand_builtin_memcpy): Call newly added function
22303         expand_builtin_memory_copy_args.
22304         (expand_builtin_memcpy_with_bounds): Likewise.
22305         (expand_builtin_mempcpy): Remove last argument.
22306         (expand_builtin_mempcpy_with_bounds): Likewise.
22307         (expand_builtin_memory_copy_args): New function created from
22308         expand_builtin_mempcpy_args with small modifications.
22309         (expand_builtin_mempcpy_args): Remove.
22310         (expand_builtin_stpcpy): Remove unused argument.
22311         (expand_builtin): Likewise.
22312         (expand_builtin_with_bounds): Likewise.
22314 2017-08-01  Martin Liska  <mliska@suse.cz>
22316         Revert r250771
22317         Make mempcpy more optimal (PR middle-end/70140).
22319 2017-08-01  Jakub Jelinek  <jakub@redhat.com>
22321         PR target/81622
22322         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): For
22323         __builtin_vec_cmpne verify both arguments are compatible vectors
22324         before looking at TYPE_MODE on the element type.  For __builtin_vec_ld
22325         verify arg1_type is a pointer or array type.  For __builtin_vec_st,
22326         move computation of aligned to after checking the argument types.
22327         Formatting fixes.
22329         PR target/80846
22330         * config/rs6000/vsx.md (vextract_fp_from_shorth,
22331         vextract_fp_from_shortl): Add element mode after mode in gen_vec_init*
22332         calls.
22334 2017-08-01  Jerome Lambourg  <lambourg@adacore.com>
22335             Doug Rupp  <rupp@adacore.com>
22336             Olivier Hainque  <hainque@adacore.com>
22338         * config.gcc (arm-wrs-vxworks*): Rework to handle arm-wrs-vxworks7 as
22339         well as arm-wrs-vxworks. Update target_cpu_name from arm6 (arch v3) to
22340         arm8 (arch v4).
22341         * config/arm/vxworks.h (MAYBE_TARGET_BPABI_CPP_BUILTINS): New, helper
22342         for TARGET_OS_CPP_BUILTIN.
22343         (TARGET_OS_CPP_BUILTIN): Invoke MAYBE_TARGET_BPABI_CPP_BUILTINS(),
22344         refine CPU definitions for arm_arch5 and add those for arm_arch6 and
22345         arm_arch7.
22346         (MAYBE_ASM_ABI_SPEC): New, helper for SUBTARGET_EXTRA_ASM_SPEC,
22347         passing required abi options to the assembler for EABI configurations.
22348         (EXTRA_CC1_SPEC): New macro, to help prevent the implicit production
22349         of .text.hot and .text.unlikely sections for kernel modules when
22350         using ARM style exceptions.
22351         (CC1_SPEC): Remove obsolete attempt at mimicking Diab toolchain
22352         options. Add EXTRA_CC1_SPEC.
22353         (VXWORKS_ENDIAN_SPEC): Adjust comment and remove handling of Diab
22354         toolchain options.
22355         (DWARF2_UNWIND_INFO): Redefine to handle the pre/post VxWorks 7
22356         transition.
22357         (ARM_TARGET2_DWARF_FORMAT): Define.
22358         * config/arm/t-vxworks: Adjust multilib control to removal of the
22359         Diab command line options.
22361 2017-08-01  Martin Liska  <mliska@suse.cz>
22363         PR gcov-profile/81561
22364         * gcov.c (unblock): Make unblocking safe as we need to preserve
22365         index correspondence of blocks and block_lists.
22367 2017-08-01  Richard Biener  <rguenther@suse.de>
22369         PR tree-optimization/81181
22370         * tree-ssa-pre.c (compute_antic_aux): Defer clean() to ...
22371         (compute_antic): ... end of iteration here.
22373 2017-08-01  James Greenhalgh  <james.greenhalgh@arm.com>
22375         * common.opt (ftree-vectorize): No longer set flag_tree_vectorize.
22376         (ftree-loop-vectorize): Set as EnabledBy ftree-vectorize.
22377         (ftree-slp-vectorize): Likewise.
22378         * omp-expand (expand_omp_simd): Remove flag_tree_vectorize, as it
22379         can no longer be set independent of flag_tree_loop_vectorize.
22380         * omp-general.c (emp_max_vf): Likewise.
22381         * opts.c (enable_fdo_optimizations): Remove references to
22382         flag_tree_vectorize, these are now implicit.
22383         (common_handle_option): Remove handling for OPT_ftree_vectorize,
22384         and leave it for the options machinery.
22386 2017-08-01  Martin Liska  <mliska@suse.cz>
22388         PR middle-end/70140
22389         * builtins.c (expand_builtin_memcpy_args): Remove.
22390         (expand_builtin_memcpy): Call newly added function
22391         expand_builtin_memory_copy_args.
22392         (expand_builtin_memcpy_with_bounds): Likewise.
22393         (expand_builtin_mempcpy): Remove last argument.
22394         (expand_builtin_mempcpy_with_bounds): Likewise.
22395         (expand_builtin_memory_copy_args): New function created from
22396         expand_builtin_mempcpy_args with small modifications.
22397         (expand_builtin_mempcpy_args): Remove.
22398         (expand_builtin_stpcpy): Remove unused argument.
22399         (expand_builtin): Likewise.
22400         (expand_builtin_with_bounds): Likewise.
22402 2017-08-01  Uros Bizjak  <ubizjak@gmail.com>
22404         PR target/81641
22405         * config/i386/i386.c (ix86_print_operand_address_as): For -masm=intel
22406         print "ds:" only for immediates in generic address space.
22408 2017-08-01  Uros Bizjak  <ubizjak@gmail.com>
22410         PR target/81639
22411         * config/i386/i386.c (ix86_funciton_naked): New prototype.
22412         (ix86_function_ok_for_sibcall): Return false for naked functions.
22414 2017-08-01  Richard Biener  <rguenther@suse.de>
22416         * tree-ssa-pre.c (print_pre_expr): Handle NULL expr.
22417         (compute_antic): Seed worklist with exit block predecessors.
22418         * cfganal.c (dfs_find_deadend): For a cycle return the source
22419         of the edge closing it.
22421 2017-08-01  Tamar Christina  <tamar.christina@arm.com>
22423         * config/aarch64/aarch64.c
22424         (aarch64_can_const_movi_rtx_p): Move 0 check.
22426 2017-08-01  Bin Cheng  <bin.cheng@arm.com>
22428         * tree.h (POINTER_TYPE_OVERFLOW_UNDEFINED): Delete.
22429         * fold-const.c (fold_comparison, fold_binary_loc): Delete use of
22430         above macro.
22431         * match.pd: Ditto in address comparison pattern.
22433 2017-08-01  Bin Cheng  <bin.cheng@arm.com>
22435         PR tree-optimization/81627
22436         * tree-predcom.c (prepare_finalizers): Always rewrite into loop
22437         closed ssa form for store-store chain.
22439 2017-08-01  Bin Cheng  <bin.cheng@arm.com>
22441         PR tree-optimization/81620
22442         * tree-predcom.c (add_ref_to_chain): Don't set has_max_use_after
22443         for store-store chain.
22445 2017-08-01  Jakub Jelinek  <jakub@redhat.com>
22447         PR tree-optimization/81588
22448         * tree-ssa-reassoc.c (optimize_range_tests_var_bound): If
22449         ranges[i].in_p, invert comparison code ccode.  For >/>=,
22450         swap rhs1 and rhs2 and comparison code unconditionally,
22451         for </<= don't do that.  Don't swap rhs1/rhs2 again if
22452         ranges[i].in_p, instead invert comparison code ccode if
22453         opcode or oe->rank is BIT_IOR_EXPR.
22455         PR target/80846
22456         * optabs.def (vec_extract_optab, vec_init_optab): Change from
22457         a direct optab to conversion optab.
22458         * optabs.c (expand_vector_broadcast): Use convert_optab_handler
22459         with GET_MODE_INNER as last argument instead of optab_handler.
22460         * expmed.c (extract_bit_field_1): Likewise.  Use vector from
22461         vector extraction if possible and optab is available.
22462         * expr.c (store_constructor): Use convert_optab_handler instead
22463         of optab_handler.  Use vector initialization from smaller
22464         vectors if possible and optab is available.
22465         * tree-vect-stmts.c (vectorizable_load): Likewise.
22466         * doc/md.texi (vec_extract, vec_init): Document that the optabs
22467         now have two modes.
22468         * config/i386/i386.c (ix86_expand_vector_init): Handle expansion
22469         of vec_init from half-sized vectors with the same element mode.
22470         * config/i386/sse.md (ssehalfvecmode): Add V4TI case.
22471         (ssehalfvecmodelower, ssescalarmodelower): New mode attributes.
22472         (reduc_plus_scal_v8df, reduc_plus_scal_v4df, reduc_plus_scal_v2df,
22473         reduc_plus_scal_v16sf, reduc_plus_scal_v8sf, reduc_plus_scal_v4sf,
22474         reduc_<code>_scal_<mode>, reduc_umin_scal_v8hi): Add element mode
22475         after mode in gen_vec_extract* calls.
22476         (vec_extract<mode>): Renamed to ...
22477         (vec_extract<mode><ssescalarmodelower>): ... this.
22478         (vec_extract<mode><ssehalfvecmodelower>): New expander.
22479         (rotl<mode>3, rotr<mode>3, <shift_insn><mode>3, ashrv2di3): Add
22480         element mode after mode in gen_vec_init* calls.
22481         (VEC_INIT_HALF_MODE): New mode iterator.
22482         (vec_init<mode>): Renamed to ...
22483         (vec_init<mode><ssescalarmodelower>): ... this.
22484         (vec_init<mode><ssehalfvecmodelower>): New expander.
22485         * config/i386/mmx.md (vec_extractv2sf): Renamed to ...
22486         (vec_extractv2sfsf): ... this.
22487         (vec_initv2sf): Renamed to ...
22488         (vec_initv2sfsf): ... this.
22489         (vec_extractv2si): Renamed to ...
22490         (vec_extractv2sisi): ... this.
22491         (vec_initv2si): Renamed to ...
22492         (vec_initv2sisi): ... this.
22493         (vec_extractv4hi): Renamed to ...
22494         (vec_extractv4hihi): ... this.
22495         (vec_initv4hi): Renamed to ...
22496         (vec_initv4hihi): ... this.
22497         (vec_extractv8qi): Renamed to ...
22498         (vec_extractv8qiqi): ... this.
22499         (vec_initv8qi): Renamed to ...
22500         (vec_initv8qiqi): ... this.
22501         * config/rs6000/vector.md (VEC_base_l): New mode attribute.
22502         (vec_init<mode>): Renamed to ...
22503         (vec_init<mode><VEC_base_l>): ... this.
22504         (vec_extract<mode>): Renamed to ...
22505         (vec_extract<mode><VEC_base_l>): ... this.
22506         * config/rs6000/paired.md (vec_initv2sf): Renamed to ...
22507         (vec_initv2sfsf): ... this.
22508         * config/rs6000/altivec.md (splitter, altivec_copysign_v4sf3,
22509         vec_unpacku_hi_v16qi, vec_unpacku_hi_v8hi, vec_unpacku_lo_v16qi,
22510         vec_unpacku_lo_v8hi, mulv16qi3, altivec_vreve<mode>2): Add
22511         element mode after mode in gen_vec_init* calls.
22512         * config/aarch64/aarch64-simd.md (vec_init<mode>): Renamed to ...
22513         (vec_init<mode><Vel>): ... this.
22514         (vec_extract<mode>): Renamed to ...
22515         (vec_extract<mode><Vel>): ... this.
22516         * config/aarch64/iterators.md (Vel): New mode attribute.
22517         * config/s390/s390.c (s390_expand_vec_strlen, s390_expand_vec_movstr):
22518         Add element mode after mode in gen_vec_extract* calls.
22519         * config/s390/vector.md (non_vec_l): New mode attribute.
22520         (vec_extract<mode>): Renamed to ...
22521         (vec_extract<mode><non_vec_l>): ... this.
22522         (vec_init<mode>): Renamed to ...
22523         (vec_init<mode><non_vec_l>): ... this.
22524         * config/s390/s390-builtins.def (s390_vlgvb, s390_vlgvh, s390_vlgvf,
22525         s390_vlgvf_flt, s390_vlgvg, s390_vlgvg_dbl): Add element mode after
22526         vec_extract mode.
22527         * config/arm/iterators.md (V_elem_l): New mode attribute.
22528         * config/arm/neon.md (vec_extract<mode>): Renamed to ...
22529         (vec_extract<mode><V_elem_l>): ... this.
22530         (vec_extractv2di): Renamed to ...
22531         (vec_extractv2didi): ... this.
22532         (vec_init<mode>): Renamed to ...
22533         (vec_init<mode><V_elem_l>): ... this.
22534         (reduc_plus_scal_<mode>, reduc_plus_scal_v2di, reduc_smin_scal_<mode>,
22535         reduc_smax_scal_<mode>, reduc_umin_scal_<mode>,
22536         reduc_umax_scal_<mode>, neon_vget_lane<mode>, neon_vget_laneu<mode>):
22537         Add element mode after gen_vec_extract* calls.
22538         * config/mips/mips-msa.md (vec_init<mode>): Renamed to ...
22539         (vec_init<mode><unitmode>): ... this.
22540         (vec_extract<mode>): Renamed to ...
22541         (vec_extract<mode><unitmode>): ... this.
22542         * config/mips/loongson.md (vec_init<mode>): Renamed to ...
22543         (vec_init<mode><unitmode>): ... this.
22544         * config/mips/mips-ps-3d.md (vec_initv2sf): Renamed to ...
22545         (vec_initv2sfsf): ... this.
22546         (vec_extractv2sf): Renamed to ...
22547         (vec_extractv2sfsf): ... this.
22548         (reduc_plus_scal_v2sf, reduc_smin_scal_v2sf, reduc_smax_scal_v2sf):
22549         Add element mode after gen_vec_extract* calls.
22550         * config/mips/mips.md (unitmode): New mode iterator.
22551         * config/spu/spu.c (spu_expand_prologue, spu_allocate_stack,
22552         spu_builtin_extract): Add element mode after gen_vec_extract* calls.
22553         * config/spu/spu.md (inner_l): New mode attribute.
22554         (vec_init<mode>): Renamed to ...
22555         (vec_init<mode><inner_l>): ... this.
22556         (vec_extract<mode>): Renamed to ...
22557         (vec_extract<mode><inner_l>): ... this.
22558         * config/sparc/sparc.md (veltmode): New mode iterator.
22559         (vec_init<VMALL:mode>): Renamed to ...
22560         (vec_init<VMALL:mode><VMALL:veltmode>): ... this.
22561         * config/ia64/vect.md (vec_initv2si): Renamed to ...
22562         (vec_initv2sisi): ... this.
22563         (vec_initv2sf): Renamed to ...
22564         (vec_initv2sfsf): ... this.
22565         (vec_extractv2sf): Renamed to ...
22566         (vec_extractv2sfsf): ... this.
22567         * config/powerpcspe/vector.md (VEC_base_l): New mode attribute.
22568         (vec_init<mode>): Renamed to ...
22569         (vec_init<mode><VEC_base_l>): ... this.
22570         (vec_extract<mode>): Renamed to ...
22571         (vec_extract<mode><VEC_base_l>): ... this.
22572         * config/powerpcspe/paired.md (vec_initv2sf): Renamed to ...
22573         (vec_initv2sfsf): ... this.
22574         * config/powerpcspe/altivec.md (splitter, altivec_copysign_v4sf3,
22575         vec_unpacku_hi_v16qi, vec_unpacku_hi_v8hi, vec_unpacku_lo_v16qi,
22576         vec_unpacku_lo_v8hi, mulv16qi3): Add element mode after mode in
22577         gen_vec_init* calls.
22579 2017-08-01  Richard Biener  <rguenther@suse.de>
22581         PR tree-optimization/81297
22582         * tree-vrp.c (get_single_symbol): Remove assert, instead drop
22583         TREE_OVERFLOW from INTEGER_CSTs.
22585 2017-07-31  Segher Boessenkool  <segher@kernel.crashing.org>
22587         * config/rs6000/rs6000.c (enum rs6000_reg_type): Delete trailing comma.
22589 2017-07-31  Carl Love  <cel@us.ibm.com>
22591         * config/rs6000/rs6000-c: Add support for built-in functions
22592         vector signed char vec_xl_be (signed long long, signed char *);
22593         vector unsigned char vec_xl_be (signed long long, unsigned char *);
22594         vector signed int vec_xl_be (signed long long, signed int *);
22595         vector unsigned int vec_xl_be (signed long long, unsigned int *);
22596         vector signed long long vec_xl_be (signed long long, signed long long *);
22597         vector unsigned long long vec_xl_be (signed long long, unsigned long long *);
22598         vector signed short vec_xl_be (signed long long, signed short *);
22599         vector unsigned short vec_xl_be (signed long long, unsigned short *);
22600         vector double vec_xl_be (signed long long, double *);
22601         vector float vec_xl_be (signed long long, float *);
22602         * config/rs6000/altivec.h (vec_xl_be): Add #define.
22603         * config/rs6000/rs6000-builtin.def (XL_BE_V16QI, XL_BE_V8HI, XL_BE_V4SI)
22604         XL_BE_V2DI, XL_BE_V4SF, XL_BE_V2DF, XL_BE): Add definitions
22605         for the builtins.
22606         * config/rs6000/rs6000.c (altivec_expand_xl_be_builtin): Add function.
22607         (altivec_expand_builtin): Add switch statement to call
22608         altivec_expand_xl_be for each builtin.
22609         (altivec_init_builtins): Add def_builtin for _builtin_vsx_le_be_v8hi,
22610         __builtin_vsx_le_be_v4si, __builtin_vsx_le_be_v2di,
22611         __builtin_vsx_le_be_v4sf, __builtin_vsx_le_be_v2df,
22612         __builtin_vsx_le_be_v16qi.
22613         * doc/extend.texi: Update the built-in documentation file for the
22614         new built-in functions.
22616 2017-07-31  Uros Bizjak  <ubizjak@gmail.com>
22618         PR target/25967
22619         * config/i386/i386.c (ix86_allocate_stack_slots_for_args):
22620         New function.
22621         (TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS): Define.
22623 2017-07-31  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
22625         * config.gcc: Add z14.
22626         * config/s390/driver-native.c (s390_host_detect_local_cpu): Add
22627         CPU model numbers for z13s and z14.
22628         * config/s390/s390-c.c (s390_resolve_overloaded_builtin): Replace
22629         arch12 with z14.
22630         * config/s390/s390-opts.h (enum processor_type): Rename
22631         PROCESSOR_ARCH12 to PROCESSOR_3906_Z14.
22632         * config/s390/s390.c (processor_table): Add field for CPU name to
22633         be passed to Binutils.
22634         (s390_asm_output_machine_for_arch): Use the new field in
22635         processor_table for Binutils.
22636         (s390_expand_builtin): Replace arch12 with z14.
22637         (s390_issue_rate): Rename PROCESSOR_ARCH12 to PROCESSOR_3906_Z14.
22638         (s390_get_sched_attrmask): Likewise.
22639         (s390_get_unit_mask): Likewise.
22640         * config/s390/s390.opt: Add z14 to processor_type enum.
22642 2017-07-31  Martin Jambor  <mjambor@suse.cz>
22644         PR hsa/81477
22645         * ipa-fnsummary.c (ipa_fn_summary_generate): Set versionable
22646         regardless of optimization level.
22648 2017-07-31  Jan Hubicka <hubicka@ucw.cz>
22649             Martin Liska  <mliska@suse.cz>
22651         * predict.def: Remove old comment and adjust probability.
22652         * gimplify.c (should_warn_for_implicit_fallthrough): Ignore
22653         PREDICT statements.
22655 2017-07-31  Uros Bizjak  <ubizjak@gmail.com>
22657         PR target/25967
22658         * config/i386/i386.c (ix86_function_naked): New function.
22659         (ix86_can_use_return_insn_p): Return false for naked functions.
22660         (ix86_expand_prologue): Skip prologue for naked functions.
22661         (ix86_expand_epilogue): Skip epilogue for naked functions
22662         and emit trap instruction.
22663         (ix86_warn_func_return): New function.
22664         (ix86_attribute_table): Add "naked" attribute specification.
22665         (TARGET_WARN_FUNC_RETURN): Define.
22666         * doc/extend.texi (x86 Function Attributes) <naked>: Document it.
22668 2017-07-31  Martin Liska  <mliska@suse.cz>
22670         * gimple-pretty-print.c (dump_gimple_label): Never dump BB info.
22671         (dump_gimple_bb_header): Always dump BB info.
22672         (pp_cfg_jump): Do not append info about BB when dumping a jump.
22674 2017-07-31  Martin Liska  <mliska@suse.cz>
22676         PR sanitize/81530
22677         * convert.c (convert_to_integer_1): Guard condition with flag_sanitize_p
22678         also with current_function_decl non-null equality.
22680 2017-07-31  Jakub Jelinek  <jakub@redhat.com>
22682         PR sanitizer/81604
22683         * ubsan.c (ubsan_type_descriptor): For UBSAN_PRINT_ARRAY don't
22684         change type to the element type, instead add eltype variable and
22685         use it where we are interested in the element type.
22687         PR tree-optimization/81603
22688         * ipa-polymorphic-call.c
22689         (ipa_polymorphic_call_context::ipa_polymorphic_call_context): Perform
22690         offset arithmetic in offset_int, bail out if the resulting bit offset
22691         doesn't fit into shwi.
22693 2017-07-31  Martin Liska  <mliska@suse.cz>
22695         * gimplify.c (mostly_copy_tree_r): Remove Java specific hunk.
22696         (gimplify_save_expr): Fix comment.
22698 2017-07-30  H.J. Lu  <hongjiu.lu@intel.com>
22700         PR target/79793
22701         * config/i386/i386.c (ix86_function_arg): Update arguments for
22702         exception handler.
22703         (ix86_compute_frame_layout): Set the initial stack offset to
22704         INCOMING_FRAME_SP_OFFSET.  Update red-zone offset with
22705         INCOMING_FRAME_SP_OFFSET.
22706         (ix86_expand_epilogue): Don't pop the 'ERROR_CODE' off the
22707         stack before exception handler returns.
22708         * config/i386/i386.h (INCOMING_FRAME_SP_OFFSET): Add the
22709         the 'ERROR_CODE' for exception handler.
22711 2017-07-30  Uros Bizjak  <ubizjak@gmail.com>
22713         * config/i386/i386.h (ASM_PRINTF_EXTENSIONS): New macro.
22714         (ASM_OUTPUT_REG_PUSH): Rewrite with new operand modifiers.
22715         (ASM_OUTPUT_REG_POP): Ditto.
22716         * config/i386/i386.c (ix86_asm_output_function_label): Use fputs
22717         instead of asm_fprintf to output pure string.
22719 2017-07-29  Jakub Jelinek  <jakub@redhat.com>
22721         * debug.h (struct gcc_debug_hooks): Add IMPLICIT argument
22722         to imported_module_or_decl hook.
22723         (debug_nothing_tree_tree_tree_bool): Remove.
22724         (debug_nothing_tree_tree_tree_bool_bool): New declaration.
22725         * debug.c (do_nothing_debug_hooks): Use
22726         debug_nothing_tree_tree_tree_bool_bool instead of
22727         debug_nothing_tree_tree_tree_bool.
22728         * vmsdbgout.c (vmsdbg_debug_hooks): Likewise.
22729         * dbxout.c (dbx_debug_hooks, xcoff_debug_hooks): Likewise.
22730         * sdbout.c (sdb_debug_hooks): Likewise.
22731         * dwarf2out.c (dwarf2_lineno_debug_hooks): Likewise.
22732         (gen_namespace_die): Add DW_AT_export_symbols attribute if
22733         langhook wants it.
22734         (dwarf2out_imported_module_or_decl): Add IMPLICIT argument,
22735         if true, -gdwarf-5 and decl will have DW_AT_export_symbols
22736         attribute, don't add anything.
22738 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22740         * fold-const.c (fold_build1_stat_loc): Remove _stat from name.
22741         (fold_build2_stat_loc): Likewise.
22742         (fold_build3_stat_loc): Likewise.
22743         * fold-const.h (fold_build1, fold_build2, fold_build3): Adjust.
22744         (fold_build1_loc): Remove macro.
22745         (fold_build2_loc): Likewise.
22746         (fold_build3_loc): Likewise.
22748 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22750         * gimple.c (gimple_build_debug_bind_stat): Remove _stat from name.
22751         (gimple_build_debug_bind_source_stat): Likewise.
22752         * gimple.h (gimple_build_debug_bind): Remove macro.
22753         (gimple_build_debug_bind_source): Likewise.
22755 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22757         * bitmap.c (bitmap_alloc): Adjust.
22758         (bitmap_gc_alloc): Likewise.
22759         * bitmap.h (bitmap_initialize_stat): Remove _stat from name.
22761 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22763         * bitmap.c (bitmap_obstack_alloc_stat): Rename to bitmap_alloc.
22764         (bitmap_gc_alloc_stat): Rename to bitmap_gc_alloc.
22765         * bitmap.h (bitmap_obstack_alloc_stat): Adjust prototype.
22766         (bitmap_gc_alloc_stat): Likewise.
22767         (BITMAP_ALLOC, BITMAP_GGC_ALLOC): Adjust.
22769 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22771         * rtl.c (shallow_copy_rtx_stat): Remove _stat from name.
22772         * rtl.h (shallow_copy_rtx): Remove macro.
22774 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22776         * emit-rtl.c (gen_raw_REG): Adjust.
22777         * gengenrtl.c (gendef): Likewise.
22778         * rtl.c (rtx_alloc_stat): Remove _stat from name.
22779         * rtl.h (rtx_alloc): Remove macro.
22781 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22783         * tree.c (build_tree_list_vec_stat): Remove _stat from name.
22784         (build_tree_list_stat): Likewise.
22785         * tree.h (build_tree_list): Remove macro.
22786         (build_tree_list_vec): Likewise.
22788 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22790         * tree.c (make_vector_stat): Remove _stat from name.
22791         (build_vector_stat): Likewise.
22792         * tree.h (make_vector_stat): Remove macro.
22793         (build_vector_stat): Likewise.
22795 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22797         * tree.h (build_var_debug_value): Remove prototype.
22799 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22801         * tree.c (tree_cons_stat): Remove _stat from name.
22802         * tree.h (tree_cons): Remove macro.
22804 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22806         * tree.c (build_vl_exp_stat): Remove _stat from name.
22807         * tree.h (build_vl_exp): Remove macro.
22809 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22811         * tree.c (build_decl_stat): Remove _stat from name.
22812         * tree.h (build_decl): Remove macro.
22814 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22816         * gimple.c (gimple_build_with_ops_stat): Adjust.
22817         (gimple_alloc_stat): Remove _stat from name.
22818         * gimple.h (gimple_alloc): Remove macro.
22820 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22822         * tree.c (make_tree_vec_stat): Remove _stat from name.
22823         (grow_tree_vec_stat): Likewise.
22824         * tree.h (make_tree_vec_stat): Adjust prototype.
22825         (grow_tree_vec_stat): Likewise.
22826         (make_tree_vec): Remove macro.
22827         (grow_tree_vec): Likewise.
22829 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22831         * fold-const.c (fold_build1_stat_loc): Adjust.
22832         (fold_build2_stat_loc): Likewise.
22833         (fold_build3_stat_loc): Likewise.
22834         * tree.c (build0_stat): Remove _stat from name.
22835         (build1_stat): Likewise.
22836         (build2_stat): Likewise.
22837         (build3_stat): Likewise.
22838         (build4_stat): Likewise.
22839         (build5_stat): Likewise.
22840         * tree.h (build1_loc): Remove macro, and rename _stat function
22841         to this.
22842         (build2_loc): Likewise.
22843         (build3_loc): Likewise.
22844         (build4_loc): Likewise.
22845         (build5_loc): Likewise.
22847 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22849         * tree.c (make_int_cst_stat): Remove _stat from name.
22850         * tree.h (make_int_cst_stat): Adjust prototype.
22851         (make_int_cst): Remove macro.
22853 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22855         * tree.c (make_tre_binfo_stat): Remove _stat from name.
22856         * tree.h (make_tree_binfo_stat): Adjust prototype.
22857         (make_tree_binfo): Remove.
22859 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22861         * tree.c (copy_node_stat): Rename to copy_node.
22862         (build_distinct_type_copy): Adjust.
22863         * tree.h (copy_node_stat): Adjust prototype.
22864         (copy_node): Remove macro.
22866 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22868         * tree.c (make_node_stat): rename to make_node.
22869         (build_tree_list_stat): Adjust.
22870         (build0_stat): Likewise.
22871         (build2_stat): Likewise.
22872         (build3_stat): Likewise.
22873         (build4_stat): Likewise.
22874         (build5_stat): Likewise.
22875         (build_decl_stat): Likewise.
22876         * tree.h (make_node_stat): Adjust prototype.
22877         (make_node): remove macro.
22879 2017-07-28  Peter Bergner  <bergner@vnet.ibm.com>
22881         * config/rs6000/ppc-auxv.h (PPC_FEATURE2_DARN): New define.
22882         (PPC_FEATURE2_SCV): Likewise.
22883         * config/rs6000/rs6000.c (cpu_supports_info): Use them.
22885 2017-07-28  Tamar Christina  <tamar.christina@arm.com>
22887         * config/aarch64/aarch64.c
22888         (aarch64_internal_mov_immediate): Add new special pattern.
22889         * config/aarch64/aarch64.md (*movdi_aarch64):
22890         Add reg/32bit const mov case.
22892 2017-07-28  Tamar Christina  <tamar.christina@arm.com>
22893             Richard Sandiford <richard.sandiford@linaro.org>
22895         * config/aarch64/aarch64.md (mov<mode>): Generalize.
22896         (*movhf_aarch64, *movsf_aarch64, *movdf_aarch64):
22897         Add integer and movi cases.
22898         (movi-split-hf-df-sf split, fp16): New.
22899         (enabled): Added TARGET_FP_F16INST.
22900         * config/aarch64/iterators.md (GPF_HF): New.
22901         * config/aarch64/predicates.md (aarch64_reg_or_fp_float): New.
22903 2017-07-28  Tamar Christina  <tamar.christina@arm.com>
22905         * config/aarch64/aarch64.c
22906         (aarch64_simd_container_mode): Add prototype.
22907         (aarch64_expand_mov_immediate): Add HI support.
22908         (aarch64_reinterpret_float_as_int, aarch64_float_const_rtx_p: New.
22909         (aarch64_can_const_movi_rtx_p): New.
22910         (aarch64_preferred_reload_class):
22911         Remove restrictions of using FP registers for certain SIMD operations.
22912         (aarch64_rtx_costs): Added new cost for CONST_DOUBLE moves.
22913         (aarch64_valid_floating_const): Add integer move validation.
22914         (aarch64_simd_imm_scalar_p): Remove.
22915         (aarch64_output_scalar_simd_mov_immediate): Generalize function.
22916         (aarch64_legitimate_constant_p): Expand list of supported cases.
22917         * config/aarch64/aarch64-protos.h
22918         (aarch64_float_const_rtx_p, aarch64_can_const_movi_rtx_p): New.
22919         (aarch64_reinterpret_float_as_int): New.
22920         (aarch64_simd_imm_scalar_p): Remove.
22921         * config/aarch64/constraints.md (Uvi): New.
22922         (Dd): Split into Ds and new Dd.
22923         * config/aarch64/aarch64.md (*movsi_aarch64):
22924         Add SIMD mov case.
22925         (*movdi_aarch64): Add SIMD mov case.
22927 2017-07-28  Bin Cheng  <bin.cheng@arm.com>
22929         * tree-predcom.c: (struct chain): Handle store-store chain in which
22930         stores for elimination only store loop invariant values.
22931         (execute_pred_commoning_chain): Ditto.
22932         (prepare_initializers_chain_store_elim): Ditto.
22933         (prepare_finalizers): Ditto.
22934         (is_inv_store_elimination_chain): New function.
22935         (initialize_root_vars_store_elim_1): New function.
22937 2017-07-28  Bin Cheng  <bin.cheng@arm.com>
22939         * tree-predcom.c: Revise general description of the pass.
22940         (enum chain_type): New enum type for store elimination.
22941         (struct chain): New field supporting store elimination.
22942         (struct component): Ditto.
22943         (dump_chain): Dump store-stores chain.
22944         (release_chain): Release resources.
22945         (split_data_refs_to_components): Compute and create component
22946         contains only stores for elimination.
22947         (get_chain_last_ref_at): New function.
22948         (make_invariant_chain): Initialization.
22949         (make_rooted_chain): Specify chain type in parameter and record it.
22950         (add_looparound_copies): Skip for store-stores chain.
22951         (determine_roots_comp): Compute type of chain and pass it to
22952         make_rooted_chain.
22953         (initialize_root_vars_store_elim_2): New function.
22954         (finalize_eliminated_stores): New function.
22955         (remove_stmt): Handle store for elimination.
22956         (execute_pred_commoning_chain): Execute predictive commoning on
22957         store-store chains.
22958         (determine_unroll_factor): Skip unroll for store-stores chain.
22959         (prepare_initializers_chain_store_elim): New function.
22960         (prepare_initializers_chain): Hanlde store-store chain.
22961         (prepare_finalizers_chain, prepare_finalizers): New function.
22962         (tree_predictive_commoning_loop): Return integer value indicating
22963         if loop is unrolled or lcssa form is corrupted.
22964         (tree_predictive_commoning): Rewrite for lcssa form if necessary.
22966 2017-07-28  Bin Cheng  <bin.cheng@arm.com>
22968         * tree-predcom.c (initialize_root): Delete.
22969         (execute_pred_commoning_chain): Initialize root vars and replace
22970         reference of non-combined chain directly, rather than call above
22971         function.
22973 2017-07-28  Bin Cheng  <bin.cheng@arm.com>
22975         * tree-predcom.c (ref_at_iteration): Add parameter NITERS.  Compute
22976         memory reference to DR at (NITERS + ITERS)-th iteration of loop.
22978 2017-07-28  Bin Cheng  <bin.cheng@arm.com>
22980         * tree-predcom.c (struct chain): New field init_seq.
22981         (release_chain): Release init_seq.
22982         (prepare_initializers_chain): Record intialization stmts in above
22983         field.
22984         (insert_init_seqs): New function.
22985         (tree_predictive_commoning_loop): Call insert_init_seqs.
22987 2017-07-28  Bin Cheng  <bin.cheng@arm.com>
22989         * tree-predcom.c (determine_roots_comp): Skip trivial components.
22991 2017-07-28  Richard Biener  <rguenther@suse.de>
22993         * match.pd: Remove superfluous :c.
22994         * genmatch.c (simplify::id): Add member.
22995         (lower_commutative, lower_opt_convert, lower_cond, lower_for):
22996         Copy id.
22997         (current_id): New global.
22998         (dt_node::parent): Move from ...
22999         (dt_operand::parent): ... here.  Add for_id member.
23000         (is_a_helper <dt_operand *>::test): DT_TRUE is also a dt_operand.
23001         (decision_tree::find_node): Relax order requirement when
23002         merging DT_TRUE nodes to ones inbetween the current simplify
23003         and the one we try to merge with.  Add diagnostic whenever
23004         we need to enforce pattern order by not merging.
23005         (decision_tree::insert): Set current_id.
23006         (decision_tree::print_node): Dump parent node and for_id.
23007         (parser::last_id): Add member.
23008         (parser::push_simplify): Assign unique id.
23009         (parser::parser): Initialize last_id.
23011 2017-07-28  Martin Liska  <mliska@suse.cz>
23013         PR sanitizer/81340
23014         * sanopt.c (sanitize_rewrite_addressable_params): Set VALUE_EXPR after
23015         gimple_build_debug_bind.
23017 2017-07-28  Richard Biener  <rguenther@suse.de>
23019         PR tree-optimization/81502
23020         * match.pd: Add pattern combining BIT_INSERT_EXPR with
23021         BIT_FIELD_REF.
23022         * tree-cfg.c (verify_expr): Verify types of BIT_FIELD_REF
23023         size/pos operands.
23024         (verify_gimple_assign_ternary): Likewise for BIT_INSERT_EXPR pos.
23025         * gimple-fold.c (maybe_canonicalize_mem_ref_addr): Use bitsizetype
23026         for BIT_FIELD_REF args.
23027         * fold-const.c (make_bit_field_ref): Likewise.
23028         * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
23030 2017-07-28  Jakub Jelinek  <jakub@redhat.com>
23032         PR sanitizer/80998
23033         * sanopt.c (pass_sanopt::execute): Handle IFN_UBSAN_PTR.
23034         * tree-ssa-alias.c (call_may_clobber_ref_p_1): Likewise.
23035         * flag-types.h (enum sanitize_code): Add SANITIZER_POINTER_OVERFLOW.
23036         Or it into SANITIZER_UNDEFINED.
23037         * ubsan.c: Include gimple-fold.h and varasm.h.
23038         (ubsan_expand_ptr_ifn): New function.
23039         (instrument_pointer_overflow): New function.
23040         (maybe_instrument_pointer_overflow): New function.
23041         (instrument_object_size): Formatting fix.
23042         (pass_ubsan::execute): Call instrument_pointer_overflow
23043         and maybe_instrument_pointer_overflow.
23044         * internal-fn.c (expand_UBSAN_PTR): New function.
23045         * ubsan.h (ubsan_expand_ptr_ifn): Declare.
23046         * sanitizer.def (__ubsan_handle_pointer_overflow,
23047         __ubsan_handle_pointer_overflow_abort): New builtins.
23048         * tree-ssa-tail-merge.c (merge_stmts_p): Handle IFN_UBSAN_PTR.
23049         * internal-fn.def (UBSAN_PTR): New internal function.
23050         * opts.c (sanitizer_opts): Add pointer-overflow.
23051         * lto-streamer-in.c (input_function): Handle IFN_UBSAN_PTR.
23052         * fold-const.c (build_range_check): Compute pointer range check in
23053         integral type if pointer arithmetics would be needed.  Formatting
23054         fixes.
23056 2017-07-28  Martin Liska  <mliska@suse.cz>
23058         PR sanitizer/81460
23059         * sanopt.c (sanitize_rewrite_addressable_params): Do not rewrite
23060         parameters that are of a variable-length.
23062 2017-07-28  Sebastian Huber  <sebastian.huber@embedded-brains.de>
23064         * config.gcc (powerpc-*-rtems*): Remove rs6000/eabi.h.  Add
23065         rs6000/biarch64.h.
23066         * config/rs6000/rtems.h (ASM_DECLARE_FUNCTION_SIZE): New macro.
23067         (ASM_OUTPUT_SPECIAL_POOL_ENTRY_P): Likewise.
23068         (CRT_CALL_STATIC_FUNCTION): Likewise.
23069         (ASM_DEFAULT_SPEC): New define.
23070         (ASM_SPEC32): Likewise.
23071         (ASM_SPEC64): Likewise.
23072         (ASM_SPEC_COMMON): Likewise.
23073         (ASM_SPEC): Likewise.
23074         (INVALID_64BIT): Likewise.
23075         (LINK_OS_DEFAULT_SPEC): Likewise.
23076         (LINK_OS_SPEC32): Likewise.
23077         (LINK_OS_SPEC64): Likewise.
23078         (POWERPC_LINUX): Likewise.
23079         (PTRDIFF_TYPE): Likewise.
23080         (RESTORE_FP_PREFIX): Likewise.
23081         (RESTORE_FP_SUFFIX): Likewise.
23082         (SAVE_FP_PREFIX): Likewise.
23083         (SAVE_FP_SUFFIX): Likewise.
23084         (SIZE_TYPE): Likewise.
23085         (SUBSUBTARGET_OVERRIDE_OPTIONS): Likewise.
23086         (TARGET_64BIT): Likewise.
23087         (TARGET_64BIT): Likewise.
23088         (TARGET_AIX): Likewise.
23089         (WCHAR_TYPE_SIZE): Likewise.
23090         (WCHAR_TYPE): Undefine.
23091         (TARGET_OS_CPP_BUILTINS): Add 64-bit PowerPC defines.
23092         (CPP_OS_DEFAULT_SPEC): Use previous CPP_OS_RTEMS_SPEC.
23093         (CPP_OS_RTEMS_SPEC): Delete.
23094         (SUBSUBTARGET_EXTRA_SPECS): Remove cpp_os_rtems.  Add
23095         asm_spec_common, asm_spec32, asm_spec64, link_os_spec32, and
23096         link_os_spec64.
23097         * config/rs6000/t-rtems: Add mcpu=e6500/m64 multilibs.
23099 2017-07-28  Jakub Jelinek  <jakub@redhat.com>
23101         PR tree-optimization/81578
23102         * tree-parloops.c (build_new_reduction): Bail out if
23103         reduction_code isn't one of the standard OpenMP reductions.
23104         Move the details printing after that decision.
23106 2017-07-27  Peter Bergner  <bergner@vnet.ibm.com>
23108         * config/rs6000/predicates.md (volatile_mem_operand): Remove code
23109         related to reload_in_progress.
23110         (splat_input_operand): Likewise.
23111         * config/rs6000/rs6000-protos.h (rs6000_secondary_memory_needed_rtx):
23112         Delete prototype.
23113         * config/rs6000/rs6000.c (machine_function): Remove sdmode_stack_slot
23114         field.
23115         (TARGET_EXPAND_TO_RTL_HOOK): Delete.
23116         (TARGET_INSTANTIATE_DECLS): Likewise.
23117         (legitimate_indexed_address_p): Delete reload_in_progress code.
23118         (rs6000_debug_legitimate_address_p): Likewise.
23119         (rs6000_eliminate_indexed_memrefs): Likewise.
23120         (rs6000_emit_le_vsx_store): Likewise.
23121         (rs6000_emit_move_si_sf_subreg): Likewise.
23122         (rs6000_emit_move): Likewise.
23123         (register_to_reg_type): Likewise.
23124         (rs6000_pre_atomic_barrier): Likewise.
23125         (rs6000_machopic_legitimize_pic_address): Likewise.
23126         (rs6000_allocate_stack_temp): Likewise.
23127         (rs6000_address_for_fpconvert): Likewise.
23128         (rs6000_address_for_altivec): Likewise.
23129         (rs6000_secondary_memory_needed_rtx): Delete function.
23130         (rs6000_check_sdmode): Likewise.
23131         (rs6000_alloc_sdmode_stack_slot): Likewise.
23132         (rs6000_instantiate_decls): Likewise.
23133         * config/rs6000/rs6000.h (SECONDARY_MEMORY_NEEDED_RTX): Delete.
23134         * config/rs6000/rs6000.md (splitter for *movsi_got_internal):
23135         Delete reload_in_progress.
23136         (*vec_reload_and_plus_<mptrsize>): Likewise.
23137         * config/rs6000/vsx.md (vsx_mul_v2di): Likewise.
23138         (vsx_div_v2di): Likewise.
23139         (vsx_udiv_v2di): Likewise.
23141 2017-07-27  Peter Bergner  <bergner@vnet.ibm.com>
23143         * config/rs6000/rs6000.opt (mlra): Replace with stub.
23144         * config/rs6000/rs6000-cpus.def (POWERPC_MASKS): Delete OPTION_MASK_LRA.
23145         * config/rs6000/rs6000.c (TARGET_LRA_P): Delete.
23146         (rs6000_debug_reg_global): Delete print of LRA status.
23147         (rs6000_option_override_internal): Delete dead LRA related code.
23148         (rs6000_lra_p): Delete function.
23149         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mlra.
23151 2017-07-27  Sebastian Huber  <sebastian.huber@embedded-brains.de>
23153         * config.gcc (riscv*-*-elf*): Add (riscv*-*-rtems*).
23154         * config/riscv/rtems.h: New file.
23156 2017-07-27  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
23157             Sudakshina Das  <sudi.das@arm.com>
23159         * config/aarch64/aarch64.md
23160         (define_split for and<mode>3nr_compare): Move
23161         non aarch64_logical_operand to a register.
23162         (define_split for and_<SHIFT:optab><mode>3nr_compare0): Move non
23163         register immediate operand to a register.
23164         * config/aarch64/predicates.md (aarch64_mov_imm_operand): New.
23166 2017-07-27  Peter Bergner  <bergner@vnet.ibm.com>
23168         PR middle-end/81564
23169         * tree-cfg.c (group_case_labels_stmt): Handle already deleted blocks.
23171 2017-07-27  Richard Biener  <rguenther@suse.de>
23173         PR tree-optimization/81573
23174         PR tree-optimization/81494
23175         * tree-vect-loop.c (vect_create_epilog_for_reduction): Handle
23176         multi defuse cycle case.
23178 2017-07-27  Richard Biener  <rguenther@suse.de>
23180         PR tree-optimization/81571
23181         * tree-vect-slp.c (vect_build_slp_tree): Properly verify reduction
23182         PHIs.
23184 2017-07-27  Eric Botcazou  <ebotcazou@adacore.com>
23186         * config/sparc/sparc.c (sparc_option_override): Set MASK_FSMULD flag
23187         earlier and only if MASK_FPU is set.  Adjust formatting.
23189 2017-07-27  Martin Liska  <mliska@suse.cz>
23191         * opt-functions.awk: Add validation of value of Init.
23192         * optc-gen.awk: Pass new argument.
23194 2017-07-27  Martin Liska  <mliska@suse.cz>
23196         * auto-profile.c (autofdo_source_profile::update_inlined_ind_target):
23197         Fix wrong condition.
23199 2017-07-27  Martin Liska  <mliska@suse.cz>
23201         * auto-profile.c (afdo_annotate_cfg): Assign zero counts to
23202         BBs and edges seen by autoFDO.
23204 2017-07-27  Richard Biener  <rguenther@suse.de>
23206         PR tree-optimization/81502
23207         * tree-ssa.c (non_rewritable_lvalue_p): Handle BIT_INSERT_EXPR
23208         with incompatible but same sized type.
23209         (execute_update_addresses_taken): Likewise.
23211 2017-07-27  James Greenhalgh  <james.greenhalgh@arm.com>
23213         * tree-ssa-loop-ch.c (pass_ch::process_loop_p): Guard on
23214         flag_tree_loop_vectorize rather than flag_tree_vectorize.
23216 2017-07-27  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
23218         PR target/81534
23219         * config/s390/s390.md ("*atomic_compare_and_swap<mode>_1")
23220         ("*atomic_compare_and_swapdi_2", "*atomic_compare_and_swapsi_3"):
23221         Change s_operand to memory_operand.
23223 2017-07-27  Richard Sandiford  <richard.sandiford@linaro.org>
23225         * config/rs6000/rs6000-protos.h (rs6000_emit_le_vsx_permute): Declare.
23226         * config/rs6000/rs6000.c (rs6000_gen_le_vsx_permute): Replace with...
23227         (rs6000_emit_le_vsx_permute): ...this.  Take the destination as input.
23228         Emit instructions rather than returning an expression.  Handle TFmode
23229         and KFmode by casting to TImode.
23230         (rs6000_emit_le_vsx_load): Update to use rs6000_emit_le_vsx_permute.
23231         (rs6000_emit_le_vsx_store): Likewise.
23232         * config/rs6000/vsx.md (VSX_TI): New iterator.
23233         (*vsx_le_permute_<mode>): Use it instead of VSX_LE_128.
23234         (*vsx_le_undo_permute_<mode>): Likewise.
23235         (*vsx_le_perm_load_<mode>): Use rs6000_emit_le_vsx_permute to
23236         emit the split sequence.
23237         (*vsx_le_perm_store_<mode>): Likewise.
23239 2017-07-27  Jakub Jelinek  <jakub@redhat.com>
23241         PR tree-optimization/81555
23242         PR tree-optimization/81556
23243         * tree-ssa-reassoc.c (rewrite_expr_tree): Add NEXT_CHANGED argument,
23244         if true, force CHANGED for the recursive invocation.
23245         (reassociate_bb): Remember original length of ops array, pass
23246         len != orig_len as NEXT_CHANGED in rewrite_expr_tree call.
23248         * attribs.c (decl_attributes): Imply noinline, noclone and no_icf
23249         attributes for noipa attribute.  For naked attribute use
23250         lookup_attribute first before lookup_attribute_spec.
23251         * final.c (rest_of_handle_final): Disable IPA RA for functions with
23252         noipa attribute.
23253         * ipa-visibility.c (non_local_p): Fix comment typos.  Return true
23254         for functions with noipa attribute.
23255         (cgraph_externally_visible_p): Return true for functions with noipa
23256         attribute.
23257         * cgraph.c (cgraph_node::get_availability): Return AVAIL_INTERPOSABLE
23258         for functions with noipa attribute.
23259         * doc/extend.texi: Document noipa function attribute.
23260         * tree-ssa-structalias.c (refered_from_nonlocal_fn): Set *nonlocal_p
23261         also for functions with noipa attribute.
23262         (ipa_pta_execute): Set nonlocal_p also for nodes with noipa attribute.
23264 2017-07-26  Andrew Pinski  <apinski@cavium.com>
23266         * config/aarch64/aarch64.c (thunderx_vector_cost): Decrease cost of
23267         vec_unalign_load_cost and vec_unalign_store_cost.
23269 2017-07-26  Michael Meissner  <meissner@linux.vnet.ibm.com>
23271         * config/rs6000/rs6000-cpus.def (ISA_2_7_MASKS_SERVER): Delete
23272         -mvsx-small-integer option.
23273         (ISA_3_0_MASKS_IEEE): Likewise.
23274         (OTHER_VSX_VECTOR_MASKS): Likewise.
23275         (POWERPC_MASKS): Likewise.
23276         * config/rs6000/rs6000.opt (-mvsx-small-integer): Likewise.
23277         * config/rs6000/rs6000.c (rs6000_hard_regno_mode_ok): Simplify
23278         code, only testing for DImode being allowed in non-VSX floating
23279         point registers.
23280         (rs6000_init_hard_regno_mode_ok): Change TARGET_VSX_SMALL_INTEGER
23281         to TARGET_P8_VECTOR test.  Remove redundant VSX test inside of
23282         another VSX test.
23283         (rs6000_option_override_internal): Delete -mvsx-small-integer.
23284         (rs6000_expand_vector_set): Change TARGET_VSX_SMALL_INTEGER to
23285         TARGET_P8_VECTOR test.
23286         (rs6000_secondary_reload_simple_move): Likewise.
23287         (rs6000_preferred_reload_class): Delete TARGET_VSX_SMALL_INTEGER,
23288         since TARGET_P9_VECTOR was already tested.
23289         (rs6000_opt_masks): Remove -mvsx-small-integer.
23290         * config/rs6000/vsx.md (vsx_extract_<mode>): Delete
23291         TARGET_VSX_SMALL_INTEGER, since a test for TARGET_P9_VECTOR was
23292         used.
23293         (vsx_extract_<mode>_p9): Delete TARGET_VSX_SMALL_INTEGER, since a
23294         test for TARGET_VEXTRACTUB was used, and that uses
23295         TARGET_P9_VECTOR.
23296         (p9 extract splitter): Likewise.
23297         (vsx_extract_<mode>_di_p9): Likewise.
23298         (vsx_extract_<mode>_store_p9): Likewise.
23299         (vsx_extract_si): Delete TARGET_VSX_SMALL_INTEGER, since a test
23300         for TARGET_P9_VECTOR was used.  Delete code that is now dead with
23301         the elimination of TARGET_VSX_SMALL_INTEGER.
23302         (vsx_extract_<mode>_p8): Likewise.
23303         (vsx_ext_<VSX_EXTRACT_I:VS_scalar>_fl_<FL_CONV:mode>): Likewise.
23304         (vsx_ext_<VSX_EXTRACT_I:VS_scalar>_ufl_<FL_CONV:mode>): Likewise.
23305         (vsx_set_<mode>_p9): Likewise.
23306         (vsx_set_v4sf_p9): Likewise.
23307         (vsx_set_v4sf_p9_zero): Likewise.
23308         (vsx_insert_extract_v4sf_p9): Likewise.
23309         (vsx_insert_extract_v4sf_p9_2): Likewise.
23310         * config/rs6000/rs6000.md (sign extend splitter): Change
23311         TARGET_VSX_SMALL_INTEGER to TARGET_P8_VECTOR test.
23312         (floatsi<mode>2_lfiwax_mem): Likewise.
23313         (floatunssi<mode>2_lfiwzx_mem): Likewise.
23314         (float<QHI:mode><FP_ISA3:mode>2): Delete TARGET_VSX_SMALL_INTEGER,
23315         since a test for TARGET_P9_VECTOR was used.
23316         (float<QHI:mode><FP_ISA3:mode>2_internal): Likewise.
23317         (floatuns<QHI:mode><FP_ISA3:mode>2): Likewise.
23318         (floatuns<QHI:mode><FP_ISA3:mode>2_internal): Likewise.
23319         (fix_trunc<mode>si2): Change TARGET_VSX_SMALL_INTEGER to
23320         TARGET_P8_VECTOR test.
23321         (fix_trunc<mode>si2_stfiwx): Likewise.
23322         (fix_trunc<mode>si2_internal): Likewise.
23323         (fix_trunc<SFDF:mode><QHI:mode>2): Delete
23324         TARGET_VSX_SMALL_INTEGER, since a test for TARGET_P9_VECTOR was
23325         used.
23326         (fix_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
23327         (fixuns_trunc<mode>si2): Change TARGET_VSX_SMALL_INTEGER to
23328         TARGET_P8_VECTOR test.
23329         (fixuns_trunc<mode>si2_stfiwx): Likewise.
23330         (fixuns_trunc<SFDF:mode><QHI:mode>2): Delete
23331         TARGET_VSX_SMALL_INTEGER, since a test for TARGET_P9_VECTOR was
23332         used.
23333         (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
23334         (fctiw<u>z_<mode>_smallint): Delete TARGET_VSX_SMALL_INTEGER,
23335         since a test for TARGET_P9_VECTOR was used.
23336         (splitter for loading small constants): Likewise.
23338 2017-07-26  Andrew Pinski  <apinski@cavium.com>
23340         * config/aarch64/aarch64.c (thunderx_vector_cost): Fix
23341         vec_fp_stmt_cost.
23343 2017-07-26  H.J. Lu  <hongjiu.lu@intel.com>
23345         PR target/81563
23346         * config/i386/i386.c (sp_valid_at): Properly check CFA offset.
23347         (fp_valid_at): Likewise.
23349 2017-07-26  James Greenhalgh  <james.greenhalgh@arm.com>
23351         * config/aarch64/aarch64.c (cortexa57_addrcost_table): Remove.
23352         (qdf24xx_addrcost_table): Likewise.
23353         (cortexa57_tunings): Update to use generic_branch_cost.
23354         (cortexa72_tunings): Likewise.
23355         (cortexa73_tunings): Likewise.
23356         (qdf24xx_tunings): Likewise.
23358 2017-07-26  James Greenhalgh  <james.greenhalgh@arm.com>
23360         * config/aarch64/aarch64.c (cortexa57_branch_cost): Remove.
23361         (thunderx2t99_branch_cost): Likewise.
23362         (cortexa35_tunings): Update to use generic_branch_cost.
23363         (cortexa53_tunings): Likewise.
23364         (cortexa57_tunings): Likewise.
23365         (cortexa72_tunings): Likewise.
23366         (cortexa73_tunings): Likewise.
23367         (thunderx2t99_tunings): Likewise.
23369 2017-07-26  Sebastian Huber  <sebastian.huber@embedded-brains.de>
23371         * config/sparc/sparc.c (dump_target_flag_bits): Dump MASK_FSMULD.
23372         (sparc_option_override): Honour MASK_FSMULD.
23373         * config/sparc/sparc.h (MASK_FEATURES): Add MASK_FSMULD.
23374         * config/sparc/sparc.md (muldf3_extend): Use TARGET_FSMULD.
23375         * config/sparc/sparc.opt (mfsmuld): New option.
23376         * doc/invoke.texi (mfsmuld): Document option.
23378 2017-07-26  Marek Polacek  <polacek@redhat.com>
23380         PR middle-end/70992
23381         * tree.c (build2_stat): Don't set TREE_CONSTANT on divisions by zero.
23383 2017-07-26  Richard Biener  <rguenther@suse.de>
23385         * gimple-match-head.c (do_valueize): Return OP if valueize
23386         returns NULL_TREE.
23387         (get_def): New helper to get at the def stmt of a SSA name
23388         if valueize allows.
23389         * genmatch.c (dt_node::gen_kids_1): Use get_def instead of
23390         do_valueize to get at the def stmt.
23391         (dt_operand::gen_gimple_expr): Simplify do_valueize calls.
23393 2017-07-26  Wilco Dijkstra  <wdijkstr@arm.com>
23395         PR middle-end/46932
23396         * auto-inc-dec.c (parse_add_or_inc): Block autoinc on sfp.
23398 2017-07-26  Martin Liska  <mliska@suse.cz>
23400         PR sanitize/81186
23401         * function.c (expand_function_start): Make expansion of
23402         nonlocal_goto_save_area after parm_birth_insn.
23404 2017-07-26  Sebastian Huber  <sebastian.huber@embedded-brains.de>
23406         * config/sparc/sparc.c (sparc_option_override): Remove MASK_FPU
23407         from all CPU target flags enable members.
23409 2017-07-26  Richard Biener  <rguenther@suse.de>
23411         * genmatch.c (dt_simplify::gen): Make iterator vars const.
23412         (decision_tree::gen): Make 'type' const.
23413         (write_predicate): Likewise.
23415 2017-07-24  Michael Meissner  <meissner@linux.vnet.ibm.com>
23417         * config/rs6000/rs6000.c (rs6000_init_hard_regno_mode_ok):
23418         Eliminate TARGET_UPPER_REGS_{DF,DI,SF} usage.
23419         (rs6000_option_override_internal): Likewise.
23420         (rs6000_expand_vector_set): Likewise.
23421         * config/rs6000/rs6000.h (TARGET_UPPER_REGS_DF): Delete.
23422         (TARGET_UPPER_REGS_SF): Likewise.
23423         (TARGET_UPPER_REGS_DI): Likewise.
23424         (TARGET_VEXTRACTUB): Eliminate TARGET_UPPER_REGS_{DF,DI,SF}.
23425         (TARGET_DIRECT_MOVE_64BIT): Likewise.
23426         * config/rs6000/rs6000.md (ALTIVEC_DFORM): Likewise.
23427         (float<QHI:mode><FP_ISA3:mode>2_internal): Likewise.
23428         (Splitters for DI constants in Altivec registers): Likewise.
23429         * config/rs6000/vsx.md (vsx_set_<mode>_p9): Likewise.
23430         (vsx_set_v4sf_p9): Likewise.
23431         (vsx_set_v4sf_p9_zero): Likewise.
23432         (vsx_insert_extract_v4sf_p9): Likewise.
23433         (vsx_insert_extract_v4sf_p9_2): Likewise.
23435 2017-07-25  Carl Love  <cel@us.ibm.com>
23437         * doc/extend.texi: Update the built-in documentation file for the
23438         existing built-in functions
23439         vector signed char vec_cnttz (vector signed char);
23440         vector unsigned char vec_cnttz (vector unsigned char);
23441         vector signed short vec_cnttz (vector signed short);
23442         vector unsigned short vec_cnttz (vector unsigned short);
23443         vector signed int vec_cnttz (vector signed int);
23444         vector unsigned int vec_cnttz (vector unsigned int);
23445         vector signed long long vec_cnttz (vector signed long long);
23446         vector unsigned long long vec_cnttz (vector unsigned long long);
23448 2017-07-25  Andrew Pinski  <apinski@cavium.com>
23450         * tree-ssa-uninit.c (warn_uninitialized_vars): Don't warn about memory
23451         accesses where the use is for the first operand of a BIT_INSERT.
23453 2017-07-25  Jim Wilson  <jim.wilson@linaro.org>
23455         PR bootstrap/81521
23456         * config/i386/winnt-cxx.c (i386_pe_adjust_class_at_definition): Look
23457         for FUNCTION_DECLs in TYPE_FIELDS rather than TYPE_METHODS.
23459 2017-07-25  Jim Wilson  <jim.wilson@linaro.org>
23461         * config/i386/gstabs.h: Delete.
23462         * config/i386/openbsd.h, config/i386/t-openbsd: Likewise.
23464 2017-07-25  Uros Bizjak  <ubizjak@gmail.com>
23466         * config/i386/i386.c (ix86_decompose_address): Do not check for
23467         register RTX when looking at index_reg or base_reg.
23468         * config/i386/i386.h (INCOMING_RETURN_ADDR_RTX): Use stack_pointer_rtx.
23470 2017-07-25  Eric Botcazou  <ebotcazou@adacore.com>
23472         * gimple.c (gimple_assign_set_rhs_with_ops): Do not ask gsi_replace
23473         to update EH info here.
23475 2017-07-25  Alexander Monakov  <amonakov@ispras.ru>
23477         * match.pd ((X * CST1) * CST2): Simplify to X * (CST1 * CST2).
23479 2017-07-25  Alexander Monakov  <amonakov@ispras.ru>
23481         * match.pd ((X * CST) * Y): Reassociate to (X * Y) * CST.
23483 2017-07-25  Torsten Duwe  <duwe@suse.de>
23485         * common.opt: Introduce -fpatchable-function-entry
23486         command line option, and its variables function_entry_patch_area_size
23487         and function_entry_patch_area_start.
23488         * opts.c (common_handle_option): Add -fpatchable_function_entry_ case,
23489         including a two-value parser.
23490         * target.def (print_patchable_function_entry): New target hook.
23491         * targhooks.h (default_print_patchable_function_entry): New function.
23492         * targhooks.c (default_print_patchable_function_entry): Likewise.
23493         * toplev.c (process_options): Switch off IPA-RA if
23494         patchable function entries are being generated.
23495         * varasm.c (assemble_start_function): Look at the
23496         patchable-function-entry command line switch and current
23497         function attributes and maybe generate NOP instructions by
23498         calling the print_patchable_function_entry hook.
23499         * doc/extend.texi: Document patchable_function_entry attribute.
23500         * doc/invoke.texi: Document -fpatchable_function_entry
23501         command line option.
23502         * doc/tm.texi.in (TARGET_ASM_PRINT_PATCHABLE_FUNCTION_ENTRY):
23503         New target hook.
23504         * doc/tm.texi: Re-generate.
23506 2017-07-25  Jakub Jelinek  <jakub@redhat.com>
23508         PR target/81532
23509         * config/i386/constraints.md (Yd, Ye): Use ALL_SSE_REGS for
23510         TARGET_AVX512DQ rather than TARGET_AVX512BW.
23512 2017-07-25  Tamar Christina  <tamar.christina@arm.com>
23514         * config/arm/parsecpu.awk (all_cores): Remove duplicates.
23516 2017-07-25  Richard Biener  <rguenther@suse.de>
23518         PR tree-optimization/81455
23519         * tree-ssa-loop-unswitch.c (find_loop_guard): Make sure to
23520         not walk in cycles when looking for guards.
23522 2017-07-25  Richard Biener  <rguenther@suse.de>
23524         PR tree-optimization/81529
23525         * tree-vect-stmts.c (process_use): Disregard live induction PHIs
23526         when optimizing backedge uses.
23528 2017-07-25  David Edelsohn  <dje.gcc@gmail.com>
23530         * dwarf2asm.c (dw2_asm_output_nstring): Encode double quote
23531         character for AIX.
23532         * dwarf2out.c (output_macinfo): Copy debug_line_section_label
23533         to dl_section_ref.  On AIX, append an expression to subtract
23534         the size of the section length to dl_section_ref.
23536 2017-07-25  Segher Boessenkool  <segher@kernel.crashing.org>
23538         * configure.ac: If any of the config.* scripts fail, exit 1.
23539         * configure: Regenerate.
23541 2017-07-25  Richard Biener  <rguenther@suse.de>
23543         PR middle-end/81546
23544         * tree-ssa-operands.c (verify_imm_links): Remove cap on number
23545         of immediate uses, be more verbose on errors.
23547 2017-07-25  Richard Biener  <rguenther@suse.de>
23549         PR tree-optimization/81510
23550         * tree-vect-loop.c (vect_is_simple_reduction): When the
23551         reduction stmt is not inside the loop bail out.
23553 2017-07-25  Richard Biener  <rguenther@suse.de>
23555         PR tree-optimization/81303
23556         * tree-vect-loop-manip.c (vect_loop_versioning): Build
23557         profitability check against LOOP_VINFO_NITERSM1.
23559 2017-07-25  Alexander Monakov  <amonakov@ispras.ru>
23561         * domwalk.c (cmp_bb_postorder): Simplify.
23562         (sort_bbs_postorder): New function.  Use it...
23563         (dom_walker::walk): ...here to optimize common cases.
23565 2017-07-25  Martin Liska  <mliska@suse.cz>
23567         PR ipa/81520
23568         * ipa-visibility.c (function_and_variable_visibility): Make the
23569         redirection just on target that supports aliasing.
23570         Fix GNU coding style.
23572 2017-07-25  Sebastian Huber  <sebastian.huber@embedded-brains.de>
23574         PR libgcc/61152
23575         * config/aarch64/rtems.h: Add GCC Runtime Library Exception.
23576         Format changes.
23577         * config/arm/rtems.h: Likewise.
23578         * config/bfin/rtems.h: Likewise.
23579         * config/i386/rtemself.h: Likewise.
23580         * config/lm32/rtems.h: Likewise.
23581         * config/m32c/rtems.h: Likewise.
23582         * config/m68k/rtemself.h: Likewise.
23583         * config/microblaze/rtems.h: Likewise.
23584         * config/mips/rtems.h: Likewise.
23585         * config/moxie/rtems.h: Likewise.
23586         * config/nios2/rtems.h: Likewise.
23587         * config/powerpcspe/rtems.h: Likewise.
23588         * config/rs6000/rtems.h: Likewise.
23589         * config/rtems.h: Likewise.
23590         * config/sh/rtems.h: Likewise.
23591         * config/sh/rtemself.h: Likewise.
23592         * config/sparc/rtemself.h: Likewise.
23594 2017-07-25  Georg-Johann Lay  <avr@gjlay.de>
23596         PR 81487
23597         * hsa-brig.c (brig_init): Use xasprintf instead of asprintf.
23598         * gimple-pretty-print.c (dump_profile, dump_probability): Same.
23599         * tree-ssa-structalias.c (alias_get_name): Same.
23601 2017-07-25  Bin Cheng  <bin.cheng@arm.com>
23603         PR target/81414
23604         * config/aarch64/cortex-a57-fma-steering.c (analyze): Skip fmul/fmac
23605         instructions if no du chain is found.
23607 2017-07-25  Georg-Johann Lay  <avr@gjlay.de>
23609         * config/avr/avr-log.c (avr_log_vadump) ['T']: Print NULL-TREE.
23611 2017-07-25  Richard Biener  <rguenther@suse.de>
23613         PR middle-end/81505
23614         * fold-const.c (fold_negate_const): TREE_OVERFLOW should be
23615         sticky.
23617 2017-07-24  Michael Meissner  <meissner@linux.vnet.ibm.com>
23619         * config/rs6000/rs6000-cpus.def (ISA_2_6_MASKS_SERVER): Delete
23620         upper-regs options.
23621         (ISA_2_7_MASKS_SERVER): Likewise.
23622         (ISA_3_0_MASKS_IEEE): Likewise.
23623         (OTHER_P8_VECTOR_MASKS): Likewise.
23624         (OTHER_VSX_VECTOR_MASKS): Likewise.
23625         (POWERPC_MASKS): Likewise.
23626         (power7 cpu): Use ISA_2_6_MASKS_SERVER instead of using a
23627         duplicate list of options.
23628         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove
23629         explicit -mupper-regs options.
23630         * config/rs6000/rs6000.opt (-mvsx-scalar-memory): Delete
23631         -mupper-regs* options.  Delete -mvsx-scalar-memory, which was an
23632         alias for -mupper-regs-df.
23633         * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): Likewise.
23634         (rs6000_init_hard_regno_mode_ok): Likewise.
23635         (rs6000_option_override_internal): Likewise.
23636         (rs6000_opt_masks): Likewise.
23637         * config/rs6000/rs6000.h (TARGET_UPPER_REGS_DF): Define upper regs
23638         options in terms of whether -mvsx or -mpower8-vector was used.
23639         (TARGET_UPPER_REGS_DI): Likewise.
23640         (TARGET_UPPER_REGS_SF): Likewise.
23641         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete the
23642         -mupper-regs-* options.
23644 2017-07-24  Segher Boessenkool  <segher@kernel.crashing.org>
23646         * passes.c (emergency_dump_function): Print some empty lines and a
23647         header before the RTL dump.
23649 2017-07-24  Segher Boessenkool  <segher@kernel.crashing.org>
23651         * cfgrtl.c (rtl_dump_bb): Don't call NEXT_INSN on NULL.
23653 2017-07-24  Wilco Dijkstra  <wdijkstr@arm.com>
23655         PR target/79041
23656         * config/aarch64/aarch64.c (aarch64_classify_symbol):
23657         Avoid SYMBOL_SMALL_ABSOLUTE for literals with pc-relative literals.
23659 2017-07-24  Carl Love  <cel@us.ibm.com>
23661         * config/rs6000/rs6000-c.c: Add support for built-in functions
23662         vector float vec_extract_fp32_from_shorth (vector unsigned short);
23663         vector float vec_extract_fp32_from_shortl (vector unsigned short);
23664         * config/rs6000/altivec.h (vec_extract_fp_from_shorth,
23665         vec_extract_fp_from_shortl): Add defines for the two builtins.
23666         * config/rs6000/rs6000-builtin.def (VEXTRACT_FP_FROM_SHORTH,
23667         VEXTRACT_FP_FROM_SHORTL): Add BU_P9V_OVERLOAD_1 and BU_P9V_VSX_1
23668         new builtins.
23669         * config/rs6000/vsx.md vsx_xvcvhpsp): Add define_insn.
23670         (vextract_fp_from_shorth, vextract_fp_from_shortl): Add define_expands.
23671         * doc/extend.texi: Update the built-in documentation file for the
23672         new built-in function.
23674 2017-07-24  Jakub Jelinek  <jakub@redhat.com>
23676         PR bootstrap/81521
23677         * tree.def: Remove TYPE_METHODS documentation, adjust TYPE_FIELDS
23678         documentation.
23679         * doc/generic.texi: Likewise.
23680         * config/i386/winnt-cxx.c (i386_pe_adjust_class_at_definition): Look
23681         for FUNCTION_DECLs in TYPE_FIELDS rather than TYPE_METHODS.
23683 2017-07-24  Jackson Woodruff  <jackson.woodruff@arm.com>
23685         * config/aarch64/aarch64-simd.md (aarch64_mla_elt_merge<mode>): New.
23686         (aarch64_mls_elt_merge<mode>): Likewise.
23688 2017-07-23  Krister Walfridsson  <krister.walfridsson@gmail.com>
23690         * config.gcc (*-*-netbsd*): Remove check for NetBSD versions not
23691         having __cxa_atexit.
23693 2017-07-23  Michael Collison  <michael.collison@arm.com>
23695         * config/arm/arm.c (arm_option_override): Deprecate
23696         use of -mstructure-size-boundary.
23697         * config/arm/arm.opt: Deprecate -mstructure-size-boundary.
23698         * doc/invoke.texi: Deprecate -mstructure-size-boundary.
23700 2017-07-23  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
23702         PR target/80695
23703         * config/rs6000/rs6000.c (rs6000_builtin_vectorization_cost):
23704         Reduce cost estimate for direct moves.
23706 2017-07-23  Uros Bizjak  <ubizjak@gmail.com>
23708         PR target/80569
23709         * config/i386/i386.c (ix86_option_override_internal): Disable
23710         BMI, BMI2 and TBM instructions for -m16.
23712 2017-07-21  Carl Love  <cel@us.ibm.com>
23714         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
23715         ALTIVEC_BUILTIN_VMULESW, ALTIVEC_BUILTIN_VMULEUW,
23716         ALTIVEC_BUILTIN_VMULOSW, ALTIVEC_BUILTIN_VMULOUW entries.
23717         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin,
23718         builtin_function_type): Add ALTIVEC_BUILTIN_* case statements.
23719         * config/rs6000/altivec.md (MVULEUW, VMULESW, VMULOUW,
23720         VMULOSW): New enum "unspec" values.
23721         (altivec_vmuleuw, altivec_vmulesw, altivec_vmulouw,
23722         altivec_vmulosw): New patterns.
23723         * config/rs6000/rs6000-builtin.def (VMLEUW, VMULESW, VMULOUW,
23724         VMULOSW): Add definitions.
23726 2017-07-21  Jim Wilson  <jim.wilson@linaro.org>
23728         * config/aarch64/aarch64-cores.def (falkor): Add AARCH64_FL_RDMA.
23729         (qdf24xx): Likewise.
23730         * config/aarch64/aarch64-options-extensions.def (rdma); New.
23731         * config/aarch64/aarch64.h (AARCH64_FL_RDMA): New.
23732         (AARCH64_FL_V8_1): Renumber.
23733         (AARCH64_FL_FOR_ARCH8_1): Add AARCH64_FL_RDMA.
23734         (AARCH64_ISA_RDMA): Use AARCH64_FL_RDMA.
23735         * config/aarch64/arm_neon.h: Use +rdma instead of arch=armv8.1-a.
23736         * doc/invoke.texi (AArch64 Options): Mention +rmda in -march docs.  Add
23737         rdma to feature modifiers list.
23739 2017-07-21  Yury Gribov  <tetra2005@gmail.com>
23741         PR middle-end/56727
23742         * ipa-visibility (function_and_variable_visibility): Convert
23743         recursive PLT call to direct call if appropriate.
23745 2017-07-21  Andrew Pinski  <apinski@cavium.com>
23747         * tree-ssa-sccvn.c (vn_nary_op_eq): Check BIT_INSERT_EXPR's
23748         operand 1 to see if the types precision matches.
23749         * fold-const.c (operand_equal_p): Likewise.
23751 2017-07-21  Richard Biener  <rguenther@suse.de>
23753         PR tree-optimization/81303
23754         * tree-vect-data-refs.c (vect_get_peeling_costs_all_drs): Pass
23755         in datarefs vector.  Allow NULL dr0 for no peeling cost estimate.
23756         (vect_peeling_hash_get_lowest_cost): Adjust.
23757         (vect_enhance_data_refs_alignment): Likewise.  Use
23758         vect_get_peeling_costs_all_drs to compute the penalty for no
23759         peeling to match up costs.
23761 2017-07-21  Richard Biener  <rguenther@suse.de>
23763         PR tree-optimization/81500
23764         * tree-vect-loop.c (vect_is_simple_reduction): Properly fail if
23765         we didn't identify a reduction path.
23767 2017-07-21  Tom de Vries  <tom@codesourcery.com>
23768             Cesar Philippidis  <cesar@codesourcery.com>
23770         PR gcov-profile/81442
23771         * config/nvptx/nvptx.c (nvptx_goacc_reduction_init): Add missing edge
23772         probabilities.
23774 2017-07-21  Tom de Vries  <tom@codesourcery.com>
23776         PR lto/81430
23777         * config/nvptx/nvptx.c (nvptx_override_options_after_change): New
23778         function.
23779         (TARGET_OVERRIDE_OPTIONS_AFTER_CHANGE): Define to
23780         nvptx_override_options_after_change.
23782 2017-07-21  Ulrich Drepper  <drepper@redhat.com>
23784         * dwarf2out.c (output_file_names): Avoid double testing for
23785         dwarf_version >= 5.
23787 2017-07-21  Georg-Johann Lay  <avr@gjlay.de>
23789         * doc/invoke.texi (AVR Built-in Functions): Re-layout section.
23791 2016-07-21  Jan Hubicka  <hubicka@ucw.cz>
23793         * cfgcleanup.c (flow_find_cross_jump): Do not crossjump across
23794         hot/cold regions.
23795         (try_crossjump_to_edge): Do not punt on partitioned functions.
23797 2016-07-21  Jan Hubicka  <hubicka@ucw.cz>
23799         * bb-reorder.c (find_rarely_executed_basic_blocks_and_crossing_edges):
23800         Put all BBs reachable only via paths crossing cold region to cold
23801         region.
23802         * cfgrtl.c (find_bbs_reachable_by_hot_paths): New function.
23804 2016-07-21  Richard Biener  <rguenther@suse.de>
23806         PR tree-optimization/81303
23807         * tree-vect-loop.c (vect_estimate_min_profitable_iters): Take
23808         into account prologue and epilogue iterations when raising
23809         min_profitable_iters to sth at least covering one vector iteration.
23811 2017-07-21  Tamar Christina  <tamar.christina@arm.com>
23813         * config/arm/arm.c (arm_test_cpu_arch_dat):
23814         Check for overlap.
23816 2017-07-20  Nathan Sidwell  <nathan@acm.org>
23818         Remove TYPE_METHODS.
23819         * tree.h (TYPE_METHODS): Delete.
23820         * dwarf2out.c (gen_member_die): Member fns are on TYPE_FIELDS.
23821         * dbxout.c (dbxout_type_fields): Ignore FUNCTION_DECLs.
23822         (dbxout_type_methods): Scan TYPE_FIELDS.
23823         (dbxout_type): Don't check TYPE_METHODS here.
23824         * function.c (use_register_for_decl): Always ignore register for
23825         class types when not optimizing.
23826         * ipa-devirt.c (odr_types_equivalent_p): Delete TYPE_METHODS scan.
23827         * tree.c (free_lang_data_in_type): Stitch out member functions and
23828         templates from TYPE_FIELDS.
23829         (build_distinct_type_copy, verify_type_variant,
23830         verify_type): Member fns are on TYPE_FIELDS.
23831         * tree-dump.c (dequeue_and_dump): No TYPE_METHODS.
23832         * tree-pretty-print.c (dump_generic_node): Likewise.
23834 2017-07-20  Jakub Jelinek  <jakub@redhat.com>
23836         PR target/80846
23837         * config/i386/i386.c (ix86_expand_vector_init_general): Handle
23838         V2TImode and V4TImode.
23839         (ix86_expand_vector_extract): Likewise.
23840         * config/i386/sse.md (VMOVE): Enable V4TImode even for just
23841         TARGET_AVX512F, instead of only for TARGET_AVX512BW.
23842         (ssescalarmode): Handle V4TImode and V2TImode.
23843         (VEC_EXTRACT_MODE): Add V4TImode and V2TImode.
23844         (*vec_extractv2ti, *vec_extractv4ti): New insns.
23845         (VEXTRACTI128_MODE): New mode iterator.
23846         (splitter for *vec_extractv?ti first element): New.
23847         (VEC_INIT_MODE): New mode iterator.
23848         (vec_init<mode>): Consolidate 3 expanders into one using
23849         VEC_INIT_MODE mode iterator.
23851 2017-07-20  Alexander Monakov  <amonakov@ispras.ru>
23853         * lra-assigns.c (pseudo_compare_func): Fix comparison step based on
23854         non_spilled_static_chain_regno_p.
23856 2017-07-20  Alexander Monakov  <amonakov@ispras.ru>
23858         * gimple-ssa-store-merging.c (sort_by_bitpos): Return 0 on equal bitpos.
23860 2017-07-20  Jan Hubicka  <hubicka@ucw.cz>
23862         * bb-reorder.c (connect_traces): Allow copying of blocks within
23863         single partition.
23865 2017-07-20  Richard Biener  <rguenther@suse.de>
23867         * gimple.h (gimple_phi_result): Add gphi * overload.
23868         (gimple_phi_result_ptr): Likewise.
23869         (gimple_phi_arg): Likewise.  Adjust index assert to only
23870         allow actual argument accesses rather than all slots available
23871         by capacity.
23872         (gimple_phi_arg_def): Add gphi * overload.
23873         * tree-phinodes.c (make_phi_node): Initialize only actual
23874         arguments.
23875         (resize_phi_node): Clear memory not covered by old node,
23876         do not initialize excess argument slots.
23877         (reserve_phi_args_for_new_edge): Initialize new argument slot
23878         completely.
23880 2017-07-20  Bin Cheng  <bin.cheng@arm.com>
23882         PR tree-optimization/81388
23883         Revert r238585:
23884         2016-07-21  Bin Cheng  <bin.cheng@arm.com>
23886         * tree-ssa-loop-niter.c (number_of_iterations_lt_to_ne): Clean up
23887         by removing computation of may_be_zero.
23889 2017-07-18  Jan Hubicka  <hubicka@ucw.cz>
23890             Tom de Vries  <tom@codesourcery.com>
23892         PR middle-end/81030
23893         * cfgbuild.c (find_many_sub_basic_blocks): Update REG_BR_PROB note
23894         when gimple level profile disagrees with what RTL expander did.
23896 2017-07-20  Richard Biener  <rguenther@suse.de>
23898         PR tree-optimization/61171
23899         * tree-vectorizer.h (slp_instance): Add reduc_phis member.
23900         (vect_analyze_stmt): Add slp instance parameter.
23901         (vectorizable_reduction): Likewise.
23902         * tree-vect-loop.c (vect_analyze_loop_operations): Adjust.
23903         (vect_is_simple_reduction): Deal with chains not detected
23904         as SLP reduction chain, specifically not properly associated
23905         chains containing a mix of plus/minus.
23906         (get_reduction_op): Remove.
23907         (get_initial_defs_for_reduction): Simplify, pass in whether
23908         this is a reduction chain, pass in the SLP node for the PHIs.
23909         (vect_create_epilog_for_reduction): Get the SLP instance as
23910         arg and adjust.
23911         (vectorizable_reduction): Get the SLP instance as arg.
23912         During analysis remember the SLP node with the PHIs in the
23913         instance.  Simplify getting at the vectorized reduction PHIs.
23914         * tree-vect-slp.c (vect_slp_analyze_node_operations): Pass
23915         through SLP instance.
23916         (vect_slp_analyze_operations): Likewise.
23917         * tree-vect-stms.c (vect_analyze_stmt): Likewise.
23918         (vect_transform_stmt): Likewise.
23920 2017-07-20  Tom de Vries  <tom@codesourcery.com>
23922         PR tree-optimization/81489
23923         * gimple-ssa-isolate-paths.c (find_implicit_erroneous_behavior): Move
23924         read of phi arg location to before loop that modifies phi.
23926 2017-07-20  Naveen H.S  <Naveen.Hurugalawadi@cavium.com>
23928         * match.pd (((m1 >/</>=/<= m2) * d -> (m1 >/</>=/<= m2) ? d : 0):
23929         New pattern.
23931 2017-07-19  Jan Hubicka  <hubicka@ucw.cz>
23933         PR middle-end/81331
23934         * except.c (execute): Fix ordering issue.
23936 2018-07-19  Segher Boessenkool  <segher@kernel.crashing.org>
23938         PR rtl-optimization/81423
23939         * combine.c (make_compound_operation_int): Don't try to optimize
23940         the AND of a SUBREG of an LSHIFTRT if that SUBREG is paradoxical.
23942 2017-07-19  Segher Boessenkool  <segher@kernel.crashing.org>
23944         PR rtl-optimization/81423
23945         * simplify-rtx.c (simplify_truncation): Handle truncating an IOR
23946         with a constant that is -1 in the truncated to mode.
23948 2017-07-19  Jan Hubicka  <hubicka@ucw.cz>
23950         * predict.c (propagate_unlikely_bbs_forward): Break out from ...
23951         (determine_unlikely_bbs): ... here.
23952         * predict.h (propagate_unlikely_bbs_forward): Declare.
23953         * cfgexpand.c (pass_expand::execute): Use it.
23954         * bb-reorder.c (sanitize_hot_paths): Do not consider known to be
23955         unlikely edges.
23956         (find_rarely_executed_basic_blocks_and_crossing_edges): Use
23957         propagate_unlikely_bbs_forward.
23959 2017-07-19  Jan Hubicka  <hubicka@ucw.cz>
23961         PR middle-end/81331
23962         * except.c (maybe_add_nop_after_section_switch): New function.
23963         (execute): Use it.
23965 2017-07-19  Tom de Vries  <tom@codesourcery.com>
23967         * gimple.h (gimple_phi_set_arg): Make assert more strict.
23969 2017-07-19  Tom de Vries  <tom@codesourcery.com>
23971         * gimple.h (gimple_phi_arg): Make assert more strict.
23973 2017-07-19  Steven Munroe  <munroesj@gcc.gnu.org>
23975         * config.gcc (powerpc*-*-*): Add mmintrin.h.
23976         * config/rs6000/mmintrin.h: New file.
23977         * config/rs6000/x86intrin.h [__ALTIVEC__]: Include mmintrin.h.
23979 2017-07-19  Jakub Jelinek  <jakub@redhat.com>
23981         PR tree-optimization/81346
23982         * match.pd: Optimize (X - 1U) <= INT_MAX-1U into (int) X > 0.
23984 2017-07-19  Tom de Vries  <tom@codesourcery.com>
23986         * config/nvptx/nvptx.md (VECIM): Add V2DI.
23988 2017-07-19  Tom de Vries  <tom@codesourcery.com>
23990         * config/nvptx/nvptx-modes.def: Add V2DImode.
23991         * config/nvptx/nvptx-protos.h (nvptx_data_alignment): Declare.
23992         * config/nvptx/nvptx.c (nvptx_ptx_type_from_mode): Handle V2DImode.
23993         (nvptx_output_mov_insn): Handle lack of mov.b128.
23994         (nvptx_print_operand): Handle 'H' and 'L' codes.
23995         (nvptx_vector_mode_supported): Allow V2DImode.
23996         (nvptx_preferred_simd_mode): New function.
23997         (nvptx_data_alignment): New function.
23998         (TARGET_VECTORIZE_PREFERRED_SIMD_MODE): Redefine to
23999         nvptx_preferred_simd_mode.
24000         * config/nvptx/nvptx.h (STACK_BOUNDARY, BIGGEST_ALIGNMENT): Change from
24001         64 to 128 bits.
24002         (DATA_ALIGNMENT): Define.  Set to nvptx_data_alignment.
24004 2017-07-19  Tom de Vries  <tom@codesourcery.com>
24006         * config/nvptx/nvptx-modes.def: New file.  Add V2SImode.
24007         * config/nvptx/nvptx.c (nvptx_ptx_type_from_mode): Handle V2SImode.
24008         (nvptx_vector_mode_supported): New function.  Allow V2SImode.
24009         (TARGET_VECTOR_MODE_SUPPORTED_P): Redefine to nvptx_vector_mode_supported.
24010         * config/nvptx/nvptx.md (VECIM): New mode iterator. Add V2SI.
24011         (mov<VECIM>_insn): New define_insn.
24012         (define_expand "mov<VECIM>): New define_expand.
24014 2017-07-19  Tom de Vries  <tom@codesourcery.com>
24016         * config/nvptx/nvptx.c (nvptx_print_operand): Handle v2 vector mode.
24018 2017-07-19  Jakub Jelinek  <jakub@redhat.com>
24020         PR tree-optimization/81346
24021         * fold-const.h (fold_div_compare, range_check_type): Declare.
24022         * fold-const.c (range_check_type): New function.
24023         (build_range_check): Use range_check_type.
24024         (fold_div_compare): No longer static, rewritten into
24025         a match.pd helper function.
24026         (fold_comparison): Don't call fold_div_compare here.
24027         * match.pd (X / C1 op C2): New optimization using fold_div_compare
24028         as helper function.
24030 2017-07-19  Nathan Sidwell  <nathan@acm.org>
24032         * tree.h (TYPE_MINVAL, TYPE_MAXVAL): Rename to ...
24033         (TYPE_MIN_VALUE_RAW, TYPE_MAX_VALUE_RAW): ... these.
24034         * tree.c (find_decls_types_r, verify_type): Use
24035         TYPE_{MIN,MAX}_VALUE_RAW.
24036         * lto-streamer-out.c (DFS::DFS_write_tree_body): Likewise.
24037         (hash_tree): Likewise.
24038         * tree-streamer-in.c (lto_input_ts_type_non_common_tree_pointers):
24039         Likewise.
24040         * tree-streamer-out.c (write_ts_type_non_common_tree_pointers):
24041         Likewise.
24043 2017-07-18  Tom de Vries  <tom@codesourcery.com>
24045         PR middle-end/81464
24046         * omp-expand.c (expand_omp_for_static_chunk): Handle
24047         equal-argument loop exit phi.
24049 2017-07-18  Uros Bizjak  <ubizjak@gmail.com>
24051         PR target/81471
24052         * config/i386/i386.md (rorx_immediate_operand): New mode attribute.
24053         (*bmi2_rorx<mode>3_1): Use rorx_immediate_operand as
24054         operand 2 predicate.
24055         (*bmi2_rorxsi3_1_zext): Use const_0_to_31_operand as
24056         operand 2 predicate.
24057         (ror,rol -> rorx splitters): Use const_int_operand as
24058         operand 2 predicate.
24060 2017-06-18  Richard Biener  <rguenther@suse.de>
24062         PR tree-optimization/81410
24063         * tree-vect-stmts.c (vectorizable_load): Properly adjust for
24064         the gap in the ! slp_perm SLP case after each group.
24066 2017-07-18  Jan Hubicka  <hubicka@ucw.cz>
24068         PR middle-end/81463
24069         * cfgloopmanip.c (scale_loop_profile): Watch out for zero frequency
24070         again.
24072 2017-07-18  Jan Hubicka  <hubicka@ucw.cz>
24074         PR middle-end/81462
24075         * predict.c (set_even_probabilities): Cleanup; do not affect
24076         probabilities that are already known.
24077         (combine_predictions_for_bb): Call even when count is set.
24079 2017-07-18  Nathan Sidwell  <nathan@acm.org>
24081         * tree-parloops.c (try_transform_to_exit_first_loop_alt): Use
24082         TYPE_MAX_VALUE.
24084 2017-07-18  Bin Cheng  <bin.cheng@arm.com>
24086         PR target/81408
24087         * tree-ssa-loop-niter.c (number_of_iterations_exit): Dump missed
24088         optimization for loop niter analysis.
24090 2017-07-18  Georg-Johann Lay  <avr@gjlay.de>
24092         PR target/81473
24093         * config/avr/avr.c (avr_optimize_casesi): Don't use
24094         INT8_MIN, INT8_MAX, UINT8_MAX, INT16_MIN, INT16_MAX, UINT16_MAX.
24096 2017-07-18  Robin Dapp  <rdapp@linux.vnet.ibm.com>
24098         * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Remove
24099         body_cost_vec from _vect_peel_extended_info.
24100         (vect_peeling_hash_get_lowest_cost): Do not set body_cost_vec.
24101         (vect_peeling_hash_choose_best_peeling): Remove body_cost_vec and
24102         npeel.
24104 2017-07-18  Bin Cheng  <bin.cheng@arm.com>
24106         * config/arm/arm.c (emit_unlikely_jump): Remove unused var.
24108 2017-07-18  Richard Biener  <rguenther@suse.de>
24110         PR tree-optimization/80620
24111         PR tree-optimization/81403
24112         * tree-ssa-pre.c (phi_translate_1): Clear range and points-to
24113         info when re-using a VN table entry.
24115 2017-07-18  Richard Biener  <rguenther@suse.de>
24117         PR tree-optimization/81418
24118         * tree-vect-loop.c (vectorizable_reduction): Properly compute
24119         vectype_in.  Verify that with lane-reducing reduction operations
24120         we have a single def-use cycle.
24122 2017-07-17  Carl Love  <cel@us.ibm.com>
24124         Revert commit r249424   2017-06-20  Carl Love  <cel@us.ibm.com>
24126         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
24127         ALTIVEC_BUILTIN_VMULESW, ALTIVEC_BUILTIN_VMULEUW,
24128         ALTIVEC_BUILTIN_VMULOSW, ALTIVEC_BUILTIN_VMULOUW entries.
24129         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin,
24130         builtin_function_type): Add ALTIVEC_BUILTIN_* case statements.
24131         * config/rs6000/altivec.md (MVULEUW, VMULESW, VMULOUW,
24132         VMULOSW): New enum "unspec" values.
24133         (vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
24134         vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si,
24135         altivec_vmuleuw, altivec_vmulesw, altivec_vmulouw,
24136         altivec_vmulosw): New patterns.
24137         * config/rs6000/rs6000-builtin.def (VMLEUW, VMULESW, VMULOUW,
24138         VMULOSW): Add definitions.
24140 2017-07-17  Uros Bizjak  <ubizjak@gmail.com>
24142         * config/alpha/alpha.c: Include predict.h.
24144 2017-07-17  Yury Gribov  <tetra2005@gmail.com>
24146         * tree-vrp.c (compare_assert_loc): Fix comparison function
24147         to return predictable results.
24149 2017-07-17  Claudiu Zissulescu  <claziss@synopsys.com>
24151         * config/arc/arc.md (adddi3): Remove support for mexpand-adddi
24152         option.
24153         (subdi3): Likewise.
24154         * config/arc/arc.opt (mexpand-adddi): Deprecate it.
24155         * doc/invoke.texi (mexpand-adddi): Update text.
24157 2017-07-17  Claudiu Zissulescu  <claziss@synopsys.com>
24159         * config/arc/arc.md (clzsi2): Expand to an arc_clzsi2 instruction
24160         that also clobbers the CC register. The old expand code is moved
24161         to ...
24162         (*arc_clzsi2): ... here.
24163         (ctzsi2): Expand to an arc_ctzsi2 instruction that also clobbers
24164         the CC register. The old expand code is moved to ...
24165         (arc_ctzsi2): ... here.
24167 2017-07-17  Claudiu Zissulescu  <claziss@synopsys.com>
24169         * config/arc/arc.opt (mindexed-loads): Use initial value
24170         TARGET_INDEXED_LOADS_DEFAULT.
24171         (mauto-modify-reg): Use initial value
24172         TARGET_AUTO_MODIFY_REG_DEFAULT.
24173         * config/arc/elf.h (TARGET_INDEXED_LOADS_DEFAULT): Define.
24174         (TARGET_AUTO_MODIFY_REG_DEFAULT): Likewise.
24175         * config/arc/linux.h (TARGET_INDEXED_LOADS_DEFAULT): Define.
24176         (TARGET_AUTO_MODIFY_REG_DEFAULT): Likewise.
24178 2017-07-17  Martin Liska  <mliska@suse.cz>
24180         PR sanitizer/81302
24181         * opts.c (finish_options): Do not allow -fgnu-tm
24182         w/ -fsanitize={kernel-,}address.  Say sorry.
24184 2017-07-17  Bin Cheng  <bin.cheng@arm.com>
24186         PR target/81369
24187         * tree-loop-distribution.c (classify_partition): Only assert on
24188         numer of iterations.
24189         (merge_dep_scc_partitions): Delete prameter.  Update function call.
24190         (distribute_loop): Remove code handling loop with unknown niters.
24191         (pass_loop_distribution::execute): Skip loop with unknown niters.
24193 2017-07-17  Bin Cheng  <bin.cheng@arm.com>
24195         PR target/81369
24196         * tree-loop-distribution.c (merge_dep_scc_partitions): Sink call to
24197         function sort_partitions_by_post_order.
24199 2017-07-17  Bin Cheng  <bin.cheng@arm.com>
24201         PR tree-optimization/81374
24202         * tree-loop-distribution.c (pass_loop_distribution::execute): Record
24203         the max index of basic blocks, rather than number of basic blocks.
24205 2017-07-17  Claudiu Zissulescu  <claziss@synopsys.com>
24207         * config/arc/arc-protos.h (arc_legitimate_pc_offset_p): Remove
24208         proto.
24209         (arc_legitimate_pic_operand_p): Likewise.
24210         * config/arc/arc.c (arc_legitimate_pic_operand_p): Remove
24211         function.
24212         (arc_needs_pcl_p): Likewise.
24213         (arc_legitimate_pc_offset_p): Likewise.
24214         (arc_legitimate_pic_addr_p): Remove LABEL_REF case, as this
24215         function is also used in constrains.md.
24216         (arc_legitimate_constant_p): Use arc_legitimate_pic_addr_p to
24217         validate pic constants. Handle CONST_INT, CONST_DOUBLE, MINUS and
24218         PLUS.  Only return true/false in known cases, otherwise assert.
24219         (arc_legitimate_address_p): Remove arc_legitimate_pic_addr_p as it
24220         is already called in arc_legitimate_constant_p.
24221         * config/arc/arc.h (CONSTANT_ADDRESS_P): Consider also LABEL for
24222         pic addresses.
24223         (LEGITIMATE_PIC_OPERAND_P): Use
24224         arc_raw_symbolic_reference_mentioned_p function.
24225         * config/arc/constraints.md (Cpc): Use arc_legitimate_pic_addr_p
24226         function.
24227         (Cal): Likewise.
24228         (C32): Likewise.
24230 2017-07-17  Claudiu Zissulescu  <claziss@synopsys.com>
24231         Andrew Burgess  <andrew.burgess@embecosm.com>
24233         * config/arc/arc-protos.h (arc_compute_function_type): Change prototype.
24234         (arc_return_address_register): New function.
24235         * config/arc/arc.c (arc_handle_fndecl_attribute): New function.
24236         (arc_handle_fndecl_attribute): Add naked attribute.
24237         (TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS): Define.
24238         (TARGET_WARN_FUNC_RETURN): Likewise.
24239         (arc_allocate_stack_slots_for_args): New function.
24240         (arc_warn_func_return): Likewise.
24241         (machine_function): Change type fn_type.
24242         (arc_compute_function_type): Consider new naked function type,
24243         change function return type.
24244         (arc_must_save_register): Adapt to handle new
24245         arc_compute_function_type's return type.
24246         (arc_expand_prologue): Likewise.
24247         (arc_expand_epilogue): Likewise.
24248         (arc_return_address_regs): Delete.
24249         (arc_return_address_register): New function.
24250         (arc_epilogue_uses): Use above function.
24251         * config/arc/arc.h (arc_return_address_regs): Delete prototype.
24252         (arc_function_type): Change encoding, add naked type.
24253         (ARC_INTERRUPT_P): Change to handle the new encoding.
24254         (ARC_FAST_INTERRUPT_P): Likewise.
24255         (ARC_NORMAL_P): Define.
24256         (ARC_NAKED_P): Likewise.
24257         (arc_compute_function_type): Delete prototype.
24258         * config/arc/arc.md (in_ret_delay_slot): Use
24259         arc_return_address_register function.
24260         (simple_return): Likewise.
24261         (p_return_i): Likewise.
24263 2017-07-17  Jakub Jelinek  <jakub@redhat.com>
24265         PR tree-optimization/81428
24266         * match.pd (X / X -> one): Don't optimize _Fract divisions, as 1
24267         can't be built for those types.
24269 2017-07-17  Georg-Johann Lay  <avr@gjlay.de>
24271         Remove stuff dead since r239246.
24273         * config/avr/avr-arch.h (avr_inform_devices): Remove dead proto.
24274         * config/avr/avr-devices.c (mcu_name, comparator, avr_mcus_str)
24275         (avr_inform_devices): Remove dead stuff.
24277 2017-07-17  Tamar Christina  <tamar.christina@arm.com>
24279         * config/arm/arm_neon.h: Fix softp typo.
24281 2017-07-17  Jakub Jelinek  <jakub@redhat.com>
24283         PR tree-optimization/81365
24284         * tree-ssa-phiprop.c (propagate_with_phi): When considering hoisting
24285         aggregate moves onto bb predecessor edges, make sure there are no
24286         loads that could alias the lhs in between the start of bb and the
24287         loads from *phi.
24289 2017-07-17  Georg-Johann Lay  <avr@gjlay.de>
24291         PR 80929
24292         * config/avr/avr.c (avr_mul_highpart_cost): New static function.
24293         (avr_rtx_costs_1) [TRUNCATE]: Use it to compute mul_highpart cost.
24294         [LSHIFTRT, outer_code = TRUNCATE]: Same.
24296 2017-07-17  Jakub Jelinek  <jakub@redhat.com>
24298         PR tree-optimization/81396
24299         * tree-ssa-math-opts.c (struct symbolic_number): Add n_ops field.
24300         (init_symbolic_number): Initialize it to 1.
24301         (perform_symbolic_merge): Add n_ops from both operands into the new
24302         n_ops.
24303         (find_bswap_or_nop): Don't consider n->n == cmpnop computations
24304         without base_addr as useless if they need more than one operation.
24305         (bswap_replace): Handle !bswap case for NULL base_addr.
24307 2017-07-17  Tom de Vries  <tom@codesourcery.com>
24309         PR target/81069
24310         * config/nvptx/nvptx.c (nvptx_single): Insert diverging branch as late
24311         as possible.
24313 2017-07-17  Sebastian Huber  <sebastian.huber@embedded-brains.de>
24315         * config/sparc/rtemself.h (TARGET_OS_CPP_BUILTINS): Add
24316         conditional builtin define __FIX_LEON3FT_B2BST.
24318 2017-07-17  Daniel Cederman  <cederman@gaisler.com>
24320         * config/sparc/t-rtems: Add mfix-gr712rc multilibs. Replace
24321         MULTILIB_EXCEPTIONS with MULTILIB_REQUIRED. Match -mfix-gr712rc
24322         with -mfix-ut700.
24324 2017-07-16  Eric Botcazou  <ebotcazou@adacore.com>
24326         PR rtl-optimization/81424
24327         * optabs.c (prepare_cmp_insn): Use copy_to_reg instead of force_reg
24328         to remove potential trapping from operands if -fnon-call-exceptions.
24330 2017-07-16  Jan Hubicka  <hubicka@ucw.cz>
24332         * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Use
24333         profile_proability for scalling.
24334         * scale_profile_for_vect_loop.c (scale_profile_for_vect_loop): Likewise.
24336 2017-07-16  Jan Hubicka  <hubicka@ucw.cz>
24338         * cgraph.c (cgraph_edge::redirect_call_stmt_to_caller): Cleanup.
24340 2017-07-16  Jan Hubicka  <hubicka@ucw.cz>
24342         * cfgloopmanip.c (scale_loop_profile): Avoid use of REG_BR_PROB_BASE
24343         fixpoint arithmetics.
24345 2017-07-16  Jan Hubicka  <hubicka@ucw.cz>
24347         * tree-ssa-loop-unswitch.c (hoist_guard): Avoid use of REG_BR_PROB_BASE
24348         fixpoint arithmetics.
24350 2017-07-16  Jan Hubicka  <hubicka@ucw.cz>
24352         * asan.c (create_cond_insert_point): Avoid use of REG_BR_PROB_BASE
24353         fixpoint arithmetics.
24355 2017-07-16  Jan Hubicka  <hubicka@ucw.cz>
24357         * profile-count.h (profile_probability::from_reg_br_prob_note,
24358         profile_probability::to_reg_br_prob_note): New functions.
24359         * doc/rtl.texi (REG_BR_PROB_NOTE): Update documentation.
24360         * reg-notes.h (REG_BR_PROB, REG_BR_PRED): Update docs.
24361         * predict.c (probability_reliable_p): Update.
24362         (edge_probability_reliable_p): Update.
24363         (br_prob_note_reliable_p): Update.
24364         (invert_br_probabilities): Update.
24365         (add_reg_br_prob_note): New function.
24366         (combine_predictions_for_insn): Update.
24367         * asan.c (asan_clear_shadow): Update.
24368         * cfgbuild.c (compute_outgoing_frequencies): Update.
24369         * cfgrtl.c (force_nonfallthru_and_redirect): Update.
24370         (update_br_prob_note): Update.
24371         (rtl_verify_edges): Update.
24372         (purge_dead_edges): Update.
24373         (fixup_reorder_chain): Update.
24374         * emit-rtl.c (try_split): Update.
24375         * ifcvt.c (cond_exec_process_insns): Update.
24376         (cond_exec_process_if_block): Update.
24377         (dead_or_predicable): Update.
24378         * internal-fn.c (expand_addsub_overflow): Update.
24379         (expand_neg_overflow): Update.
24380         (expand_mul_overflow): Update.
24381         * loop-doloop.c (doloop_modify): Update.
24382         * loop-unroll.c (compare_and_jump_seq): Update.
24383         * optabs.c (emit_cmp_and_jump_insn_1): Update.
24384         * predict.h: Update.
24385         * reorg.c (mostly_true_jump): Update.
24386         * rtl.h: Update.
24387         * config/aarch64/aarch64.c (aarch64_emit_unlikely_jump): Update.
24388         * config/alpha/alpha.c (emit_unlikely_jump): Update.
24389         * config/arc/arc.c: (emit_unlikely_jump): Update.
24390         * config/arm/arm.c: (emit_unlikely_jump): Update.
24391         * config/bfin/bfin.c (cbranch_predicted_taken_p): Update.
24392         * config/frv/frv.c (frv_print_operand_jump_hint): Update.
24393         * config/i386/i386.c (ix86_expand_split_stack_prologue): Update.
24394         (ix86_print_operand): Update.
24395         (ix86_split_fp_branch): Update.
24396         (predict_jump): Update.
24397         * config/ia64/ia64.c (ia64_print_operand): Update.
24398         * config/mmix/mmix.c (mmix_print_operand): Update.
24399         * config/powerpcspe/powerpcspe.c (output_cbranch): Update.
24400         (rs6000_expand_split_stack_prologue): Update.
24401         * config/rs6000/rs6000.c: Update.
24402         * config/s390/s390.c (s390_expand_vec_strlen): Update.
24403         (s390_expand_vec_movstr): Update.
24404         (s390_expand_cs_tdsi): Update.
24405         (s390_expand_split_stack_prologue): Update.
24406         * config/sh/sh.c (sh_print_operand): Update.
24407         (expand_cbranchsi4): Update.
24408         (expand_cbranchdi4): Update.
24409         * config/sparc/sparc.c (output_v9branch): Update.
24410         * config/spu/spu.c (get_branch_target): Update.
24411         (ea_load_store_inline): Update.
24412         * config/tilegx/tilegx.c (cbranch_predicted_p): Update.
24413         * config/tilepro/tilepro.c: Update.
24415 2017-07-16  Eric Botcazou  <ebotcazou@adacore.com>
24417         * gimplify.c (mostly_copy_tree_r): Revert latest change.
24418         (gimplify_save_expr): Likewise.
24420 2017-07-07  Jan Hubicka  <hubicka@ucw.cz>
24422         * ipa-visibility.c (function_and_variable_visibility): Fix pasto.
24424 2017-07-07  Jan Hubicka  <hubicka@ucw.cz>
24426         * ipa-fnsummary.c (pass_data_ipa_fn_summary): Use
24427         TV_IPA_FNSUMMARY.
24428         * timevar.def (TV_IPA_FNSUMMARY): Define.
24430 2017-07-16  Daniel Cederman  <cederman@gaisler.com>
24432         * config/sparc/sparc.md (divdf3_fix): Add NOP to prevent back
24433         to back store errata sensitive sequence from being generated.
24434         (sqrtdf2_fix): Likewise.
24436 2017-07-07  Jan Hubicka  <hubicka@ucw.cz>
24438         * tree-ssa-threadupdate.c (compute_path_counts,
24439         update_joiner_offpath_counts): Use profile_probability.
24441 2017-07-15  Thomas Preud'homme  <thomas.preudhomme@arm.com>
24443         Revert:
24444         2017-07-14  Thomas Preud'homme  <thomas.preudhomme@arm.com>
24446         * config/arm/arm-c.c (arm_cpu_builtins): Define
24447         __ARM_FEATURE_NUMERIC_MAXMIN solely based on TARGET_VFP5.
24449 2017-07-14  Kelvin Nilsen  <kelvin@gcc.gnu.org>
24451         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
24452         array entries to represent __ieee128 versions of the
24453         scalar_test_data_class, scalar_test_neg, scalar_extract_exp,
24454         scalar_extract_sig, and scalar_insert_exp built-in functions.
24455         (altivec_resolve_overloaded_builtin): Add special case handling
24456         for the __builtin_scalar_insert_exp function, as represented by
24457         the P9V_BUILTIN_VEC_VSIEDP constant.
24458         * config/rs6000/rs6000-builtin.def (VSEEQP): Add scalar extract
24459         exponent support for __ieee128 argument.
24460         (VSESQP): Add scalar extract signature support for __ieee128
24461         argument.
24462         (VSTDCNQP): Add scalar test negative support for __ieee128
24463         argument.
24464         (VSIEQP): Add scalar insert exponent support for __int128 argument
24465         with __ieee128 result.
24466         (VSIEQPF): Add scalar insert exponent support for __ieee128
24467         argument with __ieee128 result.
24468         (VSTDCQP): Add scalar test data class support for __ieee128
24469         argument.
24470         (VSTDCNQP): Add overload support for scalar test negative with
24471         __ieee128 argument.
24472         (VSTDCQP): Add overload support for scalar test data class
24473         __ieee128 argument.
24474         * config/rs6000/vsx.md (UNSPEC_VSX_SXSIG) Replace
24475         UNSPEC_VSX_SXSIGDP.
24476         (UNSPEC_VSX_SIEXPQP): New constant.
24477         (xsxexpqp): New insn for VSX scalar extract exponent quad
24478         precision.
24479         (xsxsigqp): New insn for VSX scalar extract significand quad
24480         precision.
24481         (xsiexpqpf): New insn for VSX scalar insert exponent quad
24482         precision with floating point argument.
24483         (xststdcqp): New expand for VSX scalar test data class quad
24484         precision.
24485         (xststdcnegqp): New expand for VSX scalar test negative quad
24486         precision.
24487         (xststdcqp): New insn to match expansions for VSX scalar test data
24488         class quad precision and VSX scalar test negative quad precision.
24489         * config/rs6000/rs6000.c (rs6000_expand_binop_builtin): Add
24490         special case operand checking to enforce that second operand of
24491         VSX scalar test data class with quad precision argument is a 7-bit
24492         unsigned literal.
24493         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Add
24494         prototypes and descriptions of __ieee128 versions of
24495         scalar_extract_exp, scalar_extract_sig, scalar_insert_exp,
24496         scalar_test_data_class, and scalar_test_neg built-in functions.
24498 2016-07-14  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
24500         PR tree-optimization/81162
24501         * gimple-ssa-strength-reduction.c (replace_mult_candidate): Don't
24502         replace a negate with an add.
24504 2017-07-14  James Greenhalgh  <james.greenhalgh@arm.com>
24506         * doc/invoke.texi (arm/-mcpu): Document +crypto.
24508 2017-07-14  Thomas Preud'homme  <thomas.preudhomme@arm.com>
24510         * config/arm/arm-c.c (arm_cpu_builtins): Define
24511         __ARM_FEATURE_NUMERIC_MAXMIN solely based on TARGET_VFP5.
24513 2017-07-14  Thomas Preud'homme  <thomas.preudhomme@arm.com>
24515         * config/arm/arm-cpus.in (cortex-r52): Add new entry.
24516         (armv8-r): Set ARM Cortex-R52 as default CPU.
24517         * config/arm/arm-tables.opt: Regenerate.
24518         * config/arm/arm-tune.md: Regenerate.
24519         * config/arm/driver-arm.c (arm_cpu_table): Add entry for ARM
24520         Cortex-R52.
24521         * doc/invoke.texi: Mention -mtune=cortex-r52 and availability of fp.dp
24522         extension for -mcpu=cortex-r52.
24524 2017-07-14  Thomas Preud'homme  <thomas.preudhomme@arm.com>
24526         * config/arm/arm-isa.h (isa_bit_FP_ARMv8): Delete enumerator.
24527         (ISA_FP_ARMv8): Define as ISA_FPv5 and ISA_FP_D32.
24528         * config/arm/arm-cpus.in (armv8-r): Define fp.sp as enabling FPv5.
24529         (fp-armv8): Define it as FP_ARMv8 only.
24530         config/arm/arm.h (TARGET_FPU_ARMV8): Delete.
24531         (TARGET_VFP_FP16INST): Define using TARGET_VFP5 rather than
24532         TARGET_FPU_ARMV8.
24533         config/arm/arm.c (arm_rtx_costs_internal): Replace checks against
24534         TARGET_FPU_ARMV8 by checks against TARGET_VFP5.
24535         * config/arm/arm-builtins.c (arm_builtin_vectorized_function): Define
24536         first ARM_CHECK_BUILTIN_MODE definition using TARGET_VFP5 rather
24537         than TARGET_FPU_ARMV8.
24538         * config/arm/arm-c.c (arm_cpu_builtins): Likewise for
24539         __ARM_FEATURE_NUMERIC_MAXMIN macro definition.
24540         * config/arm/arm.md (cmov<mode>): Condition on TARGET_VFP5 rather than
24541         TARGET_FPU_ARMV8.
24542         * config/arm/neon.md (neon_vrint): Likewise.
24543         (neon_vcvt): Likewise.
24544         (neon_<fmaxmin_op><mode>): Likewise.
24545         (<fmaxmin><mode>3): Likewise.
24546         * config/arm/vfp.md (l<vrint_pattern><su_optab><mode>si2): Likewise.
24547         * config/arm/predicates.md (arm_cond_move_operator): Check against
24548         TARGET_VFP5 rather than TARGET_FPU_ARMV8 and fix spacing.
24550 2017-07-14  Jackson Woodruff  <jackson.woodruff@arm.com>
24552         * config/aarch64/aarch64.c (aarch64_print_operand): Move comments
24553         to top of function.
24555 2017-07-14  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
24557         * gimple-ssa-store-merging.c (clear_bit_region): Replace reference to
24558         loop in comment with memset.
24560 2017-07-14  Martin Liska  <mliska@suse.cz>
24562         * cfgexpand.c (expand_gimple_basic_block): Remove dead comment.
24563         * dwarf2out.c (is_java): Remove the function.
24564         (output_pubname): Remove usage of the function.
24565         (lower_bound_default): Remove usage of DW_LANG_Java.
24566         (gen_compile_unit_die): Likewise.
24567         * gcc.c: Remove compiler defaults for .java and .zip files.
24568         * gimple-expr.c (remove_suffix): Change as there's no longer
24569         extension than 4-letter one.
24570         * gimplify.c (mostly_copy_tree_r): Remove Java-special part.
24571         (gimplify_save_expr): Likewise.
24572         * ipa-utils.h (polymorphic_type_binfo_p): Remove the comment
24573         as it's possible even for other languages than Java.
24574         * langhooks.h (struct lang_hooks): Remove Java from a comment.
24575         * lto-opts.c (lto_write_options): Remove reference to Java.
24576         * opts.c (strip_off_ending): Update file extension handling.
24577         * tree-cfg.c (verify_gimple_call): Remove comment with Java.
24578         * tree-eh.c (lower_resx): Likewise.
24579         * tree.c (free_lang_data_in_type): Remove dead code.
24580         (find_decls_types_r): Likewise.
24581         (build_common_builtin_nodes): Remove Java from a comment.
24582         (verify_type): Remove dead code.
24583         * varasm.c (assemble_external): Remove Java from a comment.
24585 2017-07-14  Martin Liska  <mliska@suse.cz>
24587         * opts.c (finish_options): Add quotes.
24588         (common_handle_option): Likewise.
24590 2017-07-14  Martin Liska  <mliska@suse.cz>
24592         * dbxout.c (get_lang_number): Do not handle GNU Pascal.
24593         * dbxout.h (extern void dbxout_stab_value_internal_label_diff):
24594         Remove N_SO_PASCAL.
24595         * dwarf2out.c (lower_bound_default): Do not handle
24596         DW_LANG_Pascal83.
24597         (gen_compile_unit_die): Likewise.
24598         * gcc.c: Remove default extension binding for GNU Pascal.
24599         * stmt.c: Remove Pascal language from a comment.
24600         * xcoffout.c: Likewise.
24602 2017-07-13  David Malcolm  <dmalcolm@redhat.com>
24604         PR c/81405
24605         * diagnostic-show-locus.c (fixit_cmp): New function.
24606         (layout::layout): Sort m_fixit_hints.
24607         (column_range::column_range): Assert that the values are valid.
24608         (struct char_span): New struct.
24609         (correction::overwrite): New method.
24610         (struct source_line): New struct.
24611         (line_corrections::add_hint): Add assertions.  Reimplement memcpy
24612         calls in terms of classes source_line and char_span, and
24613         correction::overwrite.
24614         (selftest::test_overlapped_fixit_printing_2): New function.
24615         (selftest::diagnostic_show_locus_c_tests): Call it.
24617 2017-07-13  Will Schmidt  <will_schmidt@vnet.ibm.com>
24619         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Return
24620         early if there is no lhs.
24622 2017-07-13  Martin Liska  <mliska@suse.cz>
24624         * dwarf2out.c (gen_pointer_type_die): Remove dead code.
24625         (gen_reference_type_die): Likewise.
24626         * stor-layout.c: Remove Pascal-related comment.
24628 2017-07-13  Martin Liska  <mliska@suse.cz>
24630         * opts.c (finish_options): Add quotes to error messages.
24631         (parse_sanitizer_options): Likewise.
24633 2017-07-13  Thomas Preud'homme  <thomas.preudhomme@arm.com>
24635         * doc/invoke.texi (armv8-r): Document +fp.sp ARMv8-R extension.
24637 2017-07-13  Richard Earnshaw  <rearnsha@arm.com>
24639         * config/arm/vxworks.h (TARGET_ENDIAN_DEFAULT): Define.
24641 2017-07-13  Maxim Ostapenko  <m.ostapenko@samsung.com>
24643         * asan.c (asan_emit_allocas_unpoison): Use ptr_mode for arguments
24644         during expansion.
24645         * builtins.c (expand_asan_emit_allocas_unpoison): Likewise.
24647 2017-07-12  Michael Meissner  <meissner@linux.vnet.ibm.com>
24649         PR target/81193
24650         * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If GLIBC
24651         provides the hardware capability bits, define the macro
24652         __BUILTIN_CPU_SUPPORTS__.
24653         * config/rs6000/rs6000.c (cpu_expand_builtin): Generate a warning
24654         if GLIBC does not provide the hardware capability bits.  Add a
24655         gcc_unreachable call if the built-in cpu function is neither
24656         __builtin_cpu_is nor __builtin_cpu_supports.
24657         (rs6000_get_function_versions_dispatcher): Change the warning
24658         that an old GLIBC is used which does not export the capability
24659         bits to be an error.
24660         * doc/extend.texi (target_clones attribute): Document the
24661         restriction that GLIBC 2.23 or newer is needed on the PowerPC.
24662         (PowerPC built-in functions): Document that GLIBC 2.23 or newer is
24663         needed by __builtin_cpu_is and __builtin_cpu_supports.  Document
24664         the macros defined by GCC if the newer GLIBC is available.
24666 2017-07-12  Jeff Law  <law@redhat.com>
24668         * config/riscv/riscv.c: Remove unnecessary includes.  Reorder
24669         remaining includes slightly.
24670         * config/riscv/riscv-builtins.c: Include profile-count.h.
24672 2017-07-12  Georg-Johann Lay  <avr@gjlay.de>
24674         PR target/79883
24675         * config/avr/avr.c (avr_set_current_function): In diagnostic
24676         messages: Quote keywords and (parts of) identifiers.
24677         [WITH_AVRLIBC]: Warn for functions named "ISR", "SIGNAL" or
24678         "INTERUPT".
24680 2017-07-12  Carl Love  <cel@us.ibm.com>
24682         * config/rs6000/rs6000-c.c: Add support for built-in functions
24683         vector bool char vec_revb (vector bool char);
24684         vector bool short vec_revb (vector short char);
24685         vector bool int vec_revb (vector bool int);
24686         vector bool long long vec_revb (vector bool long long);
24687         * doc/extend.texi: Update the built-in documentation file for the
24688         new built-in functions.
24690 2017-07-12  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
24692         * config/s390/s390.md: Remove movcc splitter.
24694 2017-07-12  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
24696         * config/s390/s390.c (s390_rtx_costs): Return proper costs for
24697         load/store on condition.
24699 2017-07-12  Georg-Johann Lay  <avr@gjlay.de>
24701         PR target/81407
24702         * config/avr/avr.c (avr_encode_section_info)
24703         [progmem && !TREE_READONLY]: Error if progmem object needs
24704         constructing.
24706 2017-07-11  Michael Collison  <michael.collison@arm.com>
24708         * config/aarch64/aarch64-simd.md (aarch64_sub<mode>_compare0):
24709         New pattern.
24711 2017-07-11  Carl Love  <cel@us.ibm.com>
24713         * config/rs6000/rs6000-c.c: Add support for builtins
24714         vector unsigned int vec_parity_lsbb (vector signed int);
24715         vector unsigned int vec_parity_lsbb (vector unsigned int);
24716         vector unsigned __int128 vec_parity_lsbb (vector signed __int128);
24717         vector unsigned __int128 vec_parity_lsbb (vector unsigned __int128);
24718         vector unsigned long long vec_parity_lsbb (vector signed long long);
24719         vector unsigned long long vec_parity_lsbb (vector unsigned long long);
24720         * config/rs6000/rs6000-builtin.def (VPARITY_LSBB): Add BU_P9V_OVERLOAD1.
24721         * config/rs6000/altivec.h (vec_parity_lsbb): Add define.
24722         * doc/extend.texi: Update the built-in documentation file for the
24723         new built-in functions.
24725 2017-07-11  David Malcolm  <dmalcolm@redhat.com>
24727         * diagnostic-show-locus.c: Include "gcc-rich-location.h".
24728         (layout::m_primary_loc): New field.
24729         (layout::layout): Initialize new field.  Move location filtering
24730         logic from here to...
24731         (layout::maybe_add_location_range): ...this new method.  Add
24732         support for filtering to just the lines already specified by other
24733         locations.
24734         (layout::will_show_line_p): New method.
24735         (gcc_rich_location::add_location_if_nearby): New method.
24736         (selftest::test_add_location_if_nearby): New test function.
24737         (selftest::diagnostic_show_locus_c_tests): Call it.
24738         * gcc-rich-location.h (gcc_rich_location::add_location_if_nearby):
24739         New method.
24741 2017-07-11  Tom de Vries  <tom@codesourcery.com>
24743         * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG): New macro.
24744         (bb_first_real_insn): New function.
24745         (nvptx_single): Add extra initialization of broadcasted condition
24746         variables.
24748 2017-07-11  Nathan Sidwell  <nathan@acm.org>
24750         * dwarf2out.c (gen_member_die): Remove useless check for anon ctors.
24752 2017-07-11  Georg-Johann Lay  <avr@gjlay.de>
24754         * doc/extend.texi (AVR Function Attributes): Remove weblink to
24755         Binutils doc as TEXI will mess them up.
24756         * doc/invoke.texi (AVR Options): Same here.
24758 2017-07-11  Daniel Cederman  <cederman@gaisler.com>
24760         * config/sparc/sparc.opt (mfix-ut700): New option.
24761         (mfix-gr712rc): Likewise.
24762         (sparc_fix_b2bst): New variable.
24763         * doc/invoke.texi (SPARC options): Document them.
24764         (ARM options): Fix warnings.
24765         * config/sparc/sparc.c (sparc_do_work_around_errata): Insert NOP
24766         instructions to prevent sequences that can trigger the store-store
24767         errata for certain LEON3FT processors.
24768         (pass_work_around_errata::gate): Also test sparc_fix_b2bst.
24769         (sparc_option_override): Set sparc_fix_b2bst appropriately.
24770         * config/sparc/sparc.md (fix_b2bst): New attribute.
24771         (in_branch_delay): Prevent stores in delay slot if fix_b2bst.
24773 2017-07-10  Uros Bizjak  <ubizjak@gmail.com>
24775         PR target/81375
24776         * config/i386/i386.md (divsf3): Add TARGET_SSE to TARGET_SSE_MATH.
24777         (rcpps): Ditto.
24778         (*rsqrtsf2_sse): Ditto.
24779         (rsqrtsf2): Ditto.
24780         (div<mode>3): Macroize insn from divdf3 and divsf3
24781         using MODEF mode iterator.
24783 2017-07-10  Martin Sebor  <msebor@redhat.com>
24785         PR tree-optimization/80397
24786         * gimple-ssa-sprintf.c (format_integer): Use INTEGRAL_TYPE_P()
24787         instead of testing for equality to INTEGER_TYPE.
24789 2017-07-10  Vineet Gupta <vgupta@synopsys.com>
24791         * config.gcc: Remove uclibc from arc target spec.
24793 2017-07-10  Claudiu Zissulescu  <claziss@synopsys.com>
24795         * config/arc/arc.h (ADDITIONAL_REGISTER_NAMES): Define.
24797 2017-07-07  Jan Hubicka  <hubicka@ucw.cz>
24799         PR lto/80838
24800         * lto-wrapper.c (remove_option): New function.
24801         (merge_and_complain): Merge PIC/PIE options more realistically.
24803 2017-07-10  Georg-Johann Lay  <avr@gjlay.de>
24805         Better ISR prologues by supporting GASes __gcc_isr pseudo insn.
24807         PR target/20296
24808         PR target/81268
24809         * configure.ac [target=avr]: Add GAS check for -mgcc-isr.
24810         (HAVE_AS_AVR_MGCCISR_OPTION):  If so, AC_DEFINE it.
24811         * config.in: Regenerate.
24812         * configure: Regenerate.
24813         * doc/extend.texi (AVR Function Attributes) <no_gccisr>: Document it.
24814         * doc/invoke.texi (AVR Options) <-mgas-isr-prologues>: Document it.
24815         * config/avr/avr.opt (-mgas-isr-prologues): New option and...
24816         (TARGET_GASISR_PROLOGUES): ...target mask.
24817         * common/config/avr/avr-common.c
24818         (avr_option_optimization_table) [OPT_LEVELS_1_PLUS_NOT_DEBUG]:
24819         Set -mgas-isr-prologues.
24820         * config/avr/avr-passes.def (avr_pass_pre_proep): Add
24821         INSERT_PASS_BEFORE for it.
24822         * config/avr/avr-protos.h (make_avr_pass_pre_proep): New proto.
24823         * config/avr/avr.c (avr_option_override)
24824         [!HAVE_AS_AVR_MGCCISR_OPTION]: Unset TARGET_GASISR_PROLOGUES.
24825         (avr_no_gccisr_function_p, avr_hregs_split_reg): New static functions.
24826         (avr_attribute_table) <no_gccisr>: Add new function attribute.
24827         (avr_set_current_function) <is_no_gccisr>: Init machine field.
24828         (avr_pass_data_pre_proep, avr_pass_pre_proep): New pass data
24829         and rtl_opt_pass.
24830         (make_avr_pass_pre_proep): New function.
24831         (emit_push_sfr) <treg>: Add argument to function and use it
24832         instead of TMP_REG.
24833         (avr_expand_prologue) [machine->gasisr.maybe]: Emit gasisr insn
24834         and set machine->gasisr.yes.
24835         (avr_expand_epilogue) [machine->gasisr.yes]: Similar.
24836         (avr_asm_function_end_prologue) [machine->gasisr.yes]: Add
24837         __gcc_isr.n_pushed to .L__stack_usage.
24838         (TARGET_ASM_FINAL_POSTSCAN_INSN): Define to...
24839         (avr_asm_final_postscan_insn): ...this new static function.
24840         * config/avr/avr.h (machine_function)
24841         <is_no_gccisr, use_L__stack_usage>: New fields.
24842         <gasisr, gasisr.yes, gasisr.maybe, gasisr.regno>: New fields.
24843         * config/avr/avr.md (UNSPECV_GASISR): Add unspecv enum.
24844         (GASISR_Prologue, GASISR_Epilogue, GASISR_Done): New define_constants.
24845         (gasisr, *gasisr): New expander and insn.
24846         * config/avr/gen-avr-mmcu-specs.c (print_mcu)
24847         [HAVE_AS_AVR_MGCCISR_OPTION]: Print asm_gccisr spec.
24848         * config/avr/specs.h (ASM_SPEC) <asm_gccisr>: Add sub spec.
24850 2017-07-10  Richard Earnshaw  <rearnsha@arm.com>
24852         * config/arm/parsecpu.awk (gen_comm_data): Do not escape single quotes
24853         in quoted strings.
24855 2017-07-10  Georg-Johann Lay  <avr@gjlay.de>
24857         Move jump-tables out of .text again.
24859         PR target/81075
24860         * config/avr/avr.c (ASM_OUTPUT_ADDR_VEC_ELT): Remove function.
24861         (ASM_OUTPUT_ADDR_VEC): New function.
24862         (avr_adjust_insn_length) [JUMP_TABLE_DATA_P]: Return 0.
24863         (avr_final_prescan_insn) [avr_log.insn_addresses]: Dump
24864         INSN_ADDRESSes as asm comment.
24865         * config/avr/avr.h (JUMP_TABLES_IN_TEXT_SECTION): Adjust comment.
24866         (ASM_OUTPUT_ADDR_VEC_ELT): Remove define.
24867         (ASM_OUTPUT_ADDR_VEC): Define to avr_output_addr_vec.
24868         * config/avr/avr.md (*tablejump): Adjust comment.
24869         * config/avr/elf.h (ASM_OUTPUT_BEFORE_CASE_LABEL): Remove.
24870         * config/avr/avr-log.c (avr_log_set_avr_log) <insn_addresses>:
24871         New detail.
24872         * config/avr/avr-protos.h (avr_output_addr_vec_elt): Remove proto.
24873         (avr_output_addr_vec): New proto.
24874         (avr_log_t) <insn_addresses>: New field.
24876 2017-07-09  H.J. Lu  <hongjiu.lu@intel.com>
24878         PR target/81313
24879         * config/i386/i386.c (ix86_function_arg_advance): Set
24880         outgoing_args_on_stack to true if there are outgoing arguments
24881         on stack.
24882         (ix86_function_arg): Likewise.
24883         (ix86_get_drap_rtx): Use DRAP only if there are outgoing
24884         arguments on stack and ACCUMULATE_OUTGOING_ARGS is false.
24885         * config/i386/i386.h (machine_function): Add
24886         outgoing_args_on_stack.
24888 2017-07-09  Krister Walfridsson  <krister.walfridsson@gmail.com>
24890         * config.gcc (*-*-netbsd*): Remove check for NetBSD versions not
24891         supporting pthreds.
24892         * config/netbsd.h (NETBSD_LIBGCC_SPEC): Always enable pthreads.
24894 2017-07-08  Richard Sandiford  <richard.sandiford@linaro.org>
24896         * Makefile.in (HOOKS_H, RTL_BASE_H, FUNCTION_H, EXPR_H, REGS_H)
24897         (REAL_H): Remove $(MACHMODE_H).
24898         (FIXED_VALUE_H, TREE_CORE_H, CFGLOOP_H): Remove $(MACHMODE_H) and
24899         double-int.h.
24900         (CORETYPES_H): Add signop.h, wide-int.h, wide-int-print.h,
24901         $(MACHMODE_H) and double-int.h.
24902         (build/min-insn-modes.o): Depend on $(CORETYPES_H) rather than
24903         $(MACHMODE_H).
24904         (gengtype-state.o, gengtype.o, build/gengtype.o): Don't depend on
24905         double-int.h.
24907 2017-07-07  Andrew Pinski  <apinski@cavium.com>
24909         * config/aarch64/aarch64.c (aarch_macro_fusion_pair_p): Check
24910         prev_set and curr_set for AARCH64_FUSE_ALU_BRANCH.
24912 2017-07-07  Michael Meissner  <meissner@linux.vnet.ibm.com>
24914         * config/rs6000/rs6000.c (rs6000_get_function_versions_dispatcher):
24915         Add warning if GCC was not configured to link against a GLIBC that
24916         exports the hardware capability bits.
24917         (make_resolver_func): Make resolver function private and not a
24918         COMDAT function.  Create the name with clone_function_name instead
24919         of make_unique_name.
24921         PR target/81348
24922         * config/rs6000/rs6000.md (HI sign_extend splitter): Use the
24923         correct operand in doing the split.
24925 2017-07-07 Carl Love  <cel@us.ibm.com>
24927         * config/rs6000/rs6000-c: Add support for built-in function
24928         vector unsigned short vec_pack_to_short_fp32 (vector float,
24929                                                       vector float).
24930         * config/rs6000/rs6000-builtin.def (CONVERT_4F32_8I16): Add
24931         BU_P9V_AV_2 and BU_P9V_OVERLOAD_2 definitions.
24932         * config/rs6000/altivec.h (vec_pack_to_short_fp32): Add define.
24933         * config/rs6000/altivec.md(UNSPEC_CONVERT_4F32_8I16): Add UNSPEC.
24934         (convert_4f32_8i16): Add define_expand.
24935         * doc/extend.texi: Update the built-in documentation file for the
24936         new built-in function.
24938 2017-07-07  Jose E. Marchesi  <jose.marchesi@oracle.com>
24940         * config/sparc/m8.md: New file.
24941         * config/sparc/sparc.md: Include m8.md.
24943 2017-07-07  Jose E. Marchesi  <jose.marchesi@oracle.com>
24945         * config/sparc/sparc.opt: New option -mvis4b.
24946         * config/sparc/sparc.c (dump_target_flag_bits): Handle MASK_VIS4B.
24947         (sparc_option_override): Handle VIS4B.
24948         (enum sparc_builtins): Define
24949         SPARC_BUILTIN_DICTUNPACK{8,16,32},
24950         SPARC_BUILTIN_FPCMP{LE,GT,EQ,NE}{8,16,32}SHL,
24951         SPARC_BUILTIN_FPCMPU{LE,GT}{8,16,32}SHL,
24952         SPARC_BUILTIN_FPCMPDE{8,16,32}SHL and
24953         SPARC_BUILTIN_FPCMPUR{8,16,32}SHL.
24954         (check_constant_argument): New function.
24955         (sparc_vis_init_builtins): Define builtins
24956         __builtin_vis_dictunpack{8,16,32},
24957         __builtin_vis_fpcmp{le,gt,eq,ne}{8,16,32}shl,
24958         __builtin_vis_fpcmpu{le,gt}{8,16,32}shl,
24959         __builtin_vis_fpcmpde{8,16,32}shl and
24960         __builtin_vis_fpcmpur{8,16,32}shl.
24961         (sparc_expand_builtin): Check that the constant operands to
24962         __builtin_vis_fpcmp*shl and _builtin_vis_dictunpack* are indeed
24963         constant and in range.
24964         * config/sparc/sparc-c.c (sparc_target_macros): Handle
24965         TARGET_VIS4B.
24966         * config/sparc/sparc.h (SPARC_IMM2_P): Define.
24967         (SPARC_IMM5_P): Likewise.
24968         * config/sparc/sparc.md (cpu_feature): Add new feagure "vis4b".
24969         (enabled): Handle vis4b.
24970         (UNSPEC_DICTUNPACK): New unspec.
24971         (UNSPEC_FPCMPSHL): Likewise.
24972         (UNSPEC_FPUCMPSHL): Likewise.
24973         (UNSPEC_FPCMPDESHL): Likewise.
24974         (UNSPEC_FPCMPURSHL): Likewise.
24975         (cpu_feature): New CPU feature `vis4b'.
24976         (dictunpack{8,16,32}): New insns.
24977         (FPCSMODE): New mode iterator.
24978         (fpcscond): New code iterator.
24979         (fpcsucond): Likewise.
24980         (fpcmp{le,gt,eq,ne}{8,16,32}{si,di}shl): New insns.
24981         (fpcmpu{le,gt}{8,16,32}{si,di}shl): Likewise.
24982         (fpcmpde{8,16,32}{si,di}shl): Likewise.
24983         (fpcmpur{8,16,32}{si,di}shl): Likewise.
24984         * config/sparc/constraints.md: Define constraints `q' for unsigned
24985         2-bit integer constants and `t' for unsigned 5-bit integer
24986         constants.
24987         * config/sparc/predicates.md (imm5_operand_dictunpack8): New
24988         predicate.
24989         (imm5_operand_dictunpack16): Likewise.
24990         (imm5_operand_dictunpack32): Likewise.
24991         (imm2_operand): Likewise.
24992         * doc/invoke.texi (SPARC Options): Document -mvis4b.
24993         * doc/extend.texi (SPARC VIS Built-in Functions): Document the
24994         ditunpack* and fpcmp*shl builtins.
24996 2017-07-07  Jose E. Marchesi  <jose.marchesi@oracle.com>
24998         * config.gcc: Handle m8 in --with-{cpu,tune} options.
24999         * config.in: Add HAVE_AS_SPARC6 define.
25000         * config/sparc/driver-sparc.c (cpu_names): Add entry for the SPARC
25001         M8.
25002         * config/sparc/sol2.h (CPP_CPU64_DEFAULT_SPEC): Define for
25003         TARGET_CPU_m8.
25004         (ASM_CPU32_DEFAUILT_SPEC): Likewise.
25005         (CPP_CPU_SPEC): Handle m8.
25006         (ASM_CPU_SPEC): Likewise.
25007         * config/sparc/sparc-opts.h (enum processor_type): Add
25008         PROCESSOR_M8.
25009         * config/sparc/sparc.c (m8_costs): New struct.
25010         (sparc_option_override): Handle TARGET_CPU_m8.
25011         (sparc32_initialize_trampoline): Likewise.
25012         (sparc64_initialize_trampoline): Likewise.
25013         (sparc_issue_rate): Likewise.
25014         (sparc_register_move_cost): Likewise.
25015         * config/sparc/sparc.h (TARGET_CPU_m8): Define.
25016         (CPP_CPU64_DEFAULT_SPEC): Define for M8.
25017         (ASM_CPU64_DEFAULT_SPEC): Likewise.
25018         (CPP_CPU_SPEC): Handle M8.
25019         (ASM_CPU_SPEC): Likewise.
25020         (AS_M8_FLAG): Define.
25021         * config/sparc/sparc.md: Add m8 to the cpu attribute.
25022         * config/sparc/sparc.opt: New option -mcpu=m8 for sparc targets.
25023         * configure.ac (HAVE_AS_SPARC6): Check for assembler support for
25024         M8 instructions.
25025         * configure: Regenerate.
25026         * doc/invoke.texi (SPARC Options): Document -mcpu=m8 and
25027         -mtune=m8.
25029 2017-07-07  Jose E. Marchesi  <jose.marchesi@oracle.com>
25031         * config/sparc/niagara7.md: Rework the DFA scheduler to use insn
25032         subtypes.
25033         * config/sparc/sparc.md: Remove the `v3pipe' insn attribute.
25034         ("*movdi_insn_sp32"): Do not set v3pipe.
25035         ("*movsi_insn"): Likewise.
25036         ("*movdi_insn_sp64"): Likewise.
25037         ("*movsf_insn"): Likewise.
25038         ("*movdf_insn_sp32"): Likewise.
25039         ("*movdf_insn_sp64"): Likewise.
25040         ("*zero_extendsidi2_insn_sp64"): Likewise.
25041         ("*sign_extendsidi2_insn"): Likewise.
25042         ("*mov<VM32:mode>_insn"): Likewise.
25043         ("*mov<VM64:mode>_insn_sp64"): Likewise.
25044         ("*mov<VM64:mode>_insn_sp32"): Likewise.
25045         ("<plusminus_insn><VADDSUB:mode>3"): Likewise.
25046         ("<vlop:code><VL:mode>3"): Likewise.
25047         ("*not_<vlop:code><VL:mode>3"): Likewise.
25048         ("*nand<VL:mode>_vis"): Likewise.
25049         ("*<vlnotop:code>_not1<VL:mode>_vis"): Likewise.
25050         ("*<vlnotop:code>_not2<VL:mode>_vis"): Likewise.
25051         ("one_cmpl<VL:mode>2"): Likewise.
25052         ("faligndata<VM64:mode>_vis"): Likewise.
25053         ("alignaddrsi_vis"): Likewise.
25054         ("alignaddrdi_vis"): Likweise.
25055         ("alignaddrlsi_vis"): Likewise.
25056         ("alignaddrldi_vis"): Likewise.
25057         ("fcmp<gcond:code><GCM:gcm_name><P:mode>_vis"): Likewise.
25058         ("bmaskdi_vis"): Likewise.
25059         ("bmasksi_vis"): Likewise.
25060         ("bshuffle<VM64:mode>_vis"): Likewise.
25061         ("cmask8<P:mode>_vis"): Likewise.
25062         ("cmask16<P:mode>_vis"): Likewise.
25063         ("cmask32<P:mode>_vis"): Likewise.
25064         ("pdistn<P:mode>_vis"): Likewise.
25065         ("<vis3_addsub_ss_patname><VASS:mode>3"): Likewise.
25067 2017-07-07  Jose E. Marchesi  <jose.marchesi@oracle.com>
25069         * config/sparc/sparc.md ("subtype"): New insn attribute.
25070         ("*wrgsr_sp64"): Set insn subtype.
25071         ("*rdgsr_sp64"): Likewise.
25072         ("alignaddrsi_vis"): Likewise.
25073         ("alignaddrdi_vis"): Likewise.
25074         ("alignaddrlsi_vis"): Likewise.
25075         ("alignaddrldi_vis"): Likewise.
25076         ("<plusminus_insn><VADDSUB:mode>3"): Likewise.
25077         ("fexpand_vis"): Likewise.
25078         ("fpmerge_vis"): Likewise.
25079         ("faligndata<VM64:mode>_vis"): Likewise.
25080         ("bshuffle<VM64:mode>_vis"): Likewise.
25081         ("cmask8<P:mode>_vis"): Likewise.
25082         ("cmask16<P:mode>_vis"): Likewise.
25083         ("cmask32<P:mode>_vis"): Likewise.
25084         ("fchksm16_vis"): Likewise.
25085         ("v<vis3_shift_patname><GCM:mode>3"): Likewise.
25086         ("fmean16_vis"): Likewise.
25087         ("fp<plusminus_insn>64_vis"): Likewise.
25088         ("<plusminus_insn>v8qi3"): Likewise.
25089         ("<vis3_addsub_ss_patname><VASS:mode>3"): Likewise.
25090         ("<vis4_minmax_patname><VMMAX:mode>3"): Likewise.
25091         ("<vis4_uminmax_patname><VMMAX:mode>3"): Likewise.
25092         ("<vis3_addsub_ss_patname>v8qi3"): Likewise.
25093         ("<vis4_addsub_us_patname><VAUS:mode>3"): Likewise.
25094         ("*movqi_insn"): Likewise.
25095         ("*movhi_insn"): Likewise.
25096         ("*movsi_insn"): Likewise.
25097         ("movsi_pic_gotdata_op"): Likewise.
25098         ("*movdi_insn_sp32"): Likewise.
25099         ("*movdi_insn_sp64"): Likewise.
25100         ("movdi_pic_gotdata_op"): Likewise.
25101         ("*movsf_insn"): Likewise.
25102         ("*movdf_insn_sp32"): Likewise.
25103         ("*movdf_insn_sp64"): Likewise.
25104         ("*zero_extendhisi2_insn"): Likewise.
25105         ("*zero_extendqihi2_insn"): Likewise.
25106         ("*zero_extendqisi2_insn"): Likewise.
25107         ("*zero_extendqidi2_insn"): Likewise.
25108         ("*zero_extendhidi2_insn"): Likewise.
25109         ("*zero_extendsidi2_insn_sp64"): Likewise.
25110         ("ldfsr"): Likewise.
25111         ("prefetch_64"): Likewise.
25112         ("prefetch_32"): Likewise.
25113         ("tie_ld32"): Likewise.
25114         ("tie_ld64"): Likewise.
25115         ("*tldo_ldub_sp32"): Likewise.
25116         ("*tldo_ldub1_sp32"): Likewise.
25117         ("*tldo_ldub2_sp32"): Likewise.
25118         ("*tldo_ldub_sp64"): Likewise.
25119         ("*tldo_ldub1_sp64"): Likewise.
25120         ("*tldo_ldub2_sp64"): Likewise.
25121         ("*tldo_ldub3_sp64"): Likewise.
25122         ("*tldo_lduh_sp32"): Likewise.
25123         ("*tldo_lduh1_sp32"): Likewise.
25124         ("*tldo_lduh_sp64"): Likewise.
25125         ("*tldo_lduh1_sp64"): Likewise.
25126         ("*tldo_lduh2_sp64"): Likewise.
25127         ("*tldo_lduw_sp32"): Likewise.
25128         ("*tldo_lduw_sp64"): Likewise.
25129         ("*tldo_lduw1_sp64"): Likewise.
25130         ("*tldo_ldx_sp64"): Likewise.
25131         ("*mov<VM32:mode>_insn"): Likewise.
25132         ("*mov<VM64:mode>_insn_sp64"): Likewise.
25133         ("*mov<VM64:mode>_insn_sp32"): Likewise.
25135 2017-07-07  Jose E. Marchesi  <jose.marchesi@oracle.com>
25137         * config/sparc/sparc.md ("type"): New insn type viscmp.
25138         ("fcmp<gcond:code><GCM:gcm_name><P:mode>_vis"): Set insn type to
25139         viscmp.
25140         ("fpcmp<gcond:code>8<P:mode>_vis"): Likewise.
25141         ("fucmp<gcond:code>8<P:mode>_vis"): Likewise.
25142         ("fpcmpu<gcond:code><GCM:gcm_name><P:mode>_vis"): Likewise.
25143         * config/sparc/niagara7.md ("n7_vis_logical_v3pipe"): Handle
25144         viscmp.
25145         ("n7_vis_logical_11cycle"): Likewise.
25146         * config/sparc/niagara4.md ("n4_vis_logical"): Likewise.
25147         * config/sparc/niagara2.md ("niag3_vis": Likewise.
25148         * config/sparc/niagara.md ("niag_vis"): Likewise.
25149         * config/sparc/ultra3.md ("us3_fga"): Likewise.
25150         * config/sparc/ultra1_2.md ("us1_fga_double"): Likewise.
25152 2017-07-07  Jose E. Marchesi  <jose.marchesi@oracle.com>
25154         * config/sparc/sparc.md: New instruction type `bmask'.
25155         (bmaskdi_vis): Use the `bmask' type.
25156         (bmasksi_vis): Likewise.
25157         * config/sparc/ultra3.md (us3_array): Likewise.
25158         * config/sparc/niagara7.md (n7_array): Likewise.
25159         * config/sparc/niagara4.md (n4_array): Likewise.
25160         * config/sparc/niagara2.md (niag2_vis): Likewise.
25161         (niag3_vis): Likewise.
25162         * config/sparc/niagara.md (niag_vis): Likewise.
25164 2017-07-06  Jan Hubicka  <hubicka@ucw.cz>
25166         * ipa-comdats.c: Remove optimize check from gate.
25167         * ipa-fnsummary.c (ipa_fn_summary_generate): do not generate summary
25168         for functions not optimized.
25169         (ipa_fn_summary_read): Skip optimize check.
25170         (ipa_fn_summary_write): Likewise.
25171         * ipa-inline-analysis.c (do_estimate_growth_1): Check that caller
25172         is optimized.
25173         * ipa-inline.c (can_inline_edge_p): Not optimized functions are
25174         uninlinable.
25175         (can_inline_edge_p): Check flag_pcc_struct_return for match.
25176         (check_callers): Give up on caller which is not optimized.
25177         (inline_small_functions): Likewise.
25178         (ipa_inline): Do not give up when not optimizing.
25179         * ipa-visbility.c (function_and_variable_visibility): Do not optimize
25180         away unoptimizes cdtors.
25181         (whole_program_function_and_variable_visibility): Do
25182         ipa_discover_readonly_nonaddressable_vars in LTO mode.
25183         * ipa.c (process_references): Do not check optimize.
25184         (symbol_table::remove_unreachable_nodes): Update optimize check.
25185         (set_writeonly_bit): Update optimize check.
25186         (pass_ipa_cdtor_merge::gate): Do not check optimize.
25187         (pass_ipa_single_use::gate): Remove.
25189 2017-07-06  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
25191         * config/rs6000/rs6000.c (union_defs, union_uses, insn_is_load_p,
25192         insn_is_store_p, insn_is_swap_p, const_load_sequence_p, v2df_reduction_p,
25193         rtx_is_swappable_p, insn_is_swappable_p, chain_contains_only_swaps,
25194         mark_swaps_for_removal, swap_const_vector_halves, adjust_subreg_index,
25195         permute_load, permute_store, adjust_extract, adjust_splat,
25196         adjust_xxpermdi, adjust_concat, adjust_vperm, handle_special_swappables,
25197         replace_swap_with_copy, dump_swap_insn_table,
25198         alignment_with_canonical_addr, alignment_mask, find_alignment_op,
25199         recombine_lvx_pattern, recombine_stvx_pattern,
25200         recombine_lvx_stvx_patterns, rs6000_analyze_swaps,
25201         make_pass_analyze_swaps): Move all code related to p8 swap optimizations
25202         to file rs6000-p8swap.c.
25203         * config/rs6000/rs6000-p8swap.c: New file.
25204         * config/rs6000/t-rs6000: Add rule to build rs6000-p8swap.o.
25205         * config.gcc: Add rs6000-p8swap.o to extra_objs for powerpc*-*-*
25206         and rs6000*-*-* targets.
25208 2017-07-06  David Malcolm  <dmalcolm@redhat.com>
25210         * Makefile.in (selftest): Remove dependency on s-selftest-c++.
25212 2017-07-06  Jan Hubicka  <hubicka@ucw.cz>
25214         * lto-wrapper.c (merge_and_complain): Do not merge
25215         fexceptions, fnon_call_exceptions, ftrapv, ffp_contract_, fmath_errno,
25216         fsigned_zeros, ftrapping_math, fwrapv.
25217         (append_compiler_options): Do not track these options.
25218         (append_linker_options): Likewie
25220 2017-07-06  Jan Hubicka  <hubicka@ucw.cz>
25222         * cgraphunit.c (cgraph_node::finalize_function): When
25223         !flag_toplevel_reorde set no_reorder flag.
25224         (varpool_node::finalize_decl): Likewise.
25225         (symbol_table::compile): Drop no toplevel reorder path.
25227 2017-07-06  Jan Hubicka  <hubicka@ucw.cz>
25229         * bb-reorder.c (better_edge_p): Do not build traces across abnormal/eh
25230         edges; zero probability is not better than uninitialized.
25232 2017-07-06  Maxim Ostapenko  <m.ostapenko@samsung.com>
25234         * asan.h (asan_sanitize_allocas_p): Declare.
25235         * asan.c (asan_sanitize_allocas_p): New function.
25236         (handle_builtin_stack_restore): Bail out if !asan_sanitize_allocas_p.
25237         (handle_builtin_alloca): Likewise.
25238         * cfgexpand.c (expand_used_vars): Do not add allocas unpoisoning stuff
25239         if !asan_sanitize_allocas_p.
25240         * params.def (asan-instrument-allocas): Add new option.
25241         * params.h (ASAN_PROTECT_ALLOCAS): Define.
25242         * opts.c (common_handle_option): Disable allocas sanitization for
25243         KASan by default.
25245 2017-07-06  Maxim Ostapenko  <m.ostapenko@samsung.com>
25247         * asan.c: Include gimple-fold.h.
25248         (get_last_alloca_addr): New function.
25249         (handle_builtin_stackrestore): Likewise.
25250         (handle_builtin_alloca): Likewise.
25251         (asan_emit_allocas_unpoison): Likewise.
25252         (get_mem_refs_of_builtin_call): Add new parameter, remove const
25253         quallifier from first paramerer. Handle BUILT_IN_ALLOCA,
25254         BUILT_IN_ALLOCA_WITH_ALIGN and BUILT_IN_STACK_RESTORE builtins.
25255         (instrument_builtin_call): Pass gimple iterator to
25256         get_mem_refs_of_builtin_call.
25257         (last_alloca_addr): New global.
25258         * asan.h (asan_emit_allocas_unpoison): Declare.
25259         * builtins.c (expand_asan_emit_allocas_unpoison): New function.
25260         (expand_builtin): Handle BUILT_IN_ASAN_ALLOCAS_UNPOISON.
25261         * cfgexpand.c (expand_used_vars): Call asan_emit_allocas_unpoison
25262         if function calls alloca.
25263         * gimple-fold.c (replace_call_with_value): Remove static keyword.
25264         * gimple-fold.h (replace_call_with_value): Declare.
25265         * internal-fn.c: Include asan.h.
25266         * sanitizer.def (BUILT_IN_ASAN_ALLOCA_POISON,
25267         BUILT_IN_ASAN_ALLOCAS_UNPOISON): New builtins.
25269 2017-07-06  David Malcolm  <dmalcolm@redhat.com>
25271         * Makefile.in (SELFTEST_FLAGS): Drop "-x c", moving it to...
25272         (C_SELFTEST_FLAGS): New.
25273         (CPP_SELFTEST_FLAGS): New.
25274         (SELFTEST_DEPS): New, from deps of s-selftest.
25275         (C_SELFTEST_DEPS): New, from deps of s-selftest.
25276         (CPP_SELFTEST_DEPS): New.
25277         (selftest): Add dependency on s-selftest-c++.
25278         (s-selftest): Rename to...
25279         (s-selftest-c): ...this, moving deps to SELFTEST_DEPS
25280         and C_SELFTEST_DEPS, and using C_SELFTEST_FLAGS rather
25281         than SELFTEST_FLAGS.
25282         (selftest-gdb): Rename to...
25283         (selftest-c-gdb): ...this, using C_SELFTEST_DEPS and
25284         C_SELFTEST_FLAGS.
25285         (selftest-gdb): Reintroduce as an alias for selftest-c-gdb.
25286         (selftest-valgrind): Rename to...
25287         (selftest-c-valgrind): ...this, using C_SELFTEST_DEPS and
25288         C_SELFTEST_FLAGS.
25289         (selftest-valgrind): Reintroduce as an alias for
25290         selftest-c-valgrind.
25291         (s-selftest-c++): New.
25292         (selftest-c++-gdb): New.
25293         (selftest-c++-valgrind): New.
25295 2017-07-06  Olivier Hainque  <hainque@adacore.com>
25297         * gcc.c (process_command): When deciding if undefined variables
25298         should be ignored when processing specs, accept "gcc -v" as well.
25300 2017-07-06  Jan Hubicka  <hubicka@ucw.cz>
25302         * auto-profile.c (afdo_set_bb_count, afdo_propagate_edge,
25303         afdo_annotate_cfg): Set counts/probabilities as determined by afdo.
25305 2017-07-06  Thomas Preud'homme  <thomas.preudhomme@arm.com>
25307         * config/arm/arm-cpus.in (armv8-r): Add new entry.
25308         * config/arm/arm-isa.h (ISA_ARMv8r): Define macro.
25309         * config/arm/arm-tables.opt: Regenerate.
25310         * config/arm/arm.h (enum base_architecture): Add BASE_ARCH_8R
25311         enumerator.
25312         * doc/invoke.texi: Mention -march=armv8-r and its extensions.
25314 2017-07-06  Carl Love  <cel@us.ibm.com>
25316         * ChangeLog: Clean up from mid air collision
25318 2017-07-06  Carl Love  <cel@us.ibm.com>
25320         * config/rs6000/rs6000-c.c: Add support for built-in functions
25321         vector signed int vec_subc (vector signed int, vector signed int);
25322         vector signed __int128 vec_subc (vector signed __int128,
25323                                          vector signed __int128);
25324         vector unsigned __int128 vec_subc (vector unsigned __int128,
25325                                            vector unsigned __int128);
25326         vector signed int vec_sube (vector signed int, vector signed int,
25327                                     vector signed int);
25328         vector unsigned int vec_sube (vector unsigned int,
25329                                       vector unsigned int,
25330                                       vector unsigned int);
25331         vector signed __int128 vec_sube (vector signed __int128,
25332                                          vector signed __int128,
25333                                          vector signed__int128);
25334         vector unsigned __int128 vec_sube (vector unsigned __int128,
25335                                            vector unsigned __int128,
25336                                            vector unsigned __int128);
25337         vector signed int vec_subec (vector signed int, vector signed int,
25338                                      vector signed int);
25339         vector unsigned int vec_subec (vector unsigned int,
25340                                        vector unsigned int,
25341                                        vector unsigned int);
25342         vector signed __int128 vec_subec (vector signed __int128,
25343                                           vector signed __int128,
25344                                           vector signed__int128);
25345         vector unsigned __int128 vec_subec (vector unsigned __int128,
25346                                             vector unsigned __int128,
25347                                             vector unsigned __int128);
25348         * config/rs6000/rs6000.c (ALTIVEC_BUILTIN_VEC_SUBE,
25349         ALTIVEC_BUILTIN_VEC_SUBEC): Add ef_builtins.
25350         * config/rs6000/rs6000-builtin.def (SUBE, SUBEC): Add
25351         BU_ALTIVEC_OVERLOAD_X definitions.
25352         * config/rs6000/altivec.h (vec_sube, vec_subec): Add builtin defines.
25353         * doc/extend.texi: Update the built-in documentation file for the new
25354         built-in functions.
25356 2017-07-06  David Malcolm  <dmalcolm@redhat.com>
25358         PR c++/79300
25359         * diagnostic-show-locus.c (layout::layout): Use start and finish
25360         spelling location for the start and finish of each range.
25361         * genmatch.c (linemap_client_expand_location_to_spelling_point):
25362         Add unused aspect param.
25363         * input.c (expand_location_1): Add "aspect" param, and use it
25364         to access the correct part of the location.
25365         (expand_location): Pass LOCATION_ASPECT_CARET to new param of
25366         expand_location_1.
25367         (expand_location_to_spelling_point): Likewise.
25368         (linemap_client_expand_location_to_spelling_point): Add "aspect"
25369         param, and pass it to expand_location_1.
25371 2017-07-06  Sebastian Peryt  <sebastian.peryt@intel.com>
25373         * config/i386/avx512fintrin.h (_mm_mask_getexp_round_ss,
25374         _mm_maskz_getexp_round_ss, _mm_mask_getexp_round_sd,
25375         _mm_maskz_getexp_round_sd, _mm_mask_getmant_round_sd,
25376         _mm_maskz_getmant_round_sd, _mm_mask_getmant_round_ss,
25377         _mm_maskz_getmant_round_ss, _mm_mask_getexp_ss, _mm_maskz_getexp_ss,
25378         _mm_mask_getexp_sd, _mm_maskz_getexp_sd, _mm_mask_getmant_sd,
25379         _mm_maskz_getmant_sd, _mm_mask_getmant_ss,
25380         _mm_maskz_getmant_ss): New intrinsics.
25381         (__builtin_ia32_getexpss128_mask): Changed to ...
25382         __builtin_ia32_getexpss128_round ... this.
25383         (__builtin_ia32_getexpsd128_mask): Changed to ...
25384         __builtin_ia32_getexpsd128_round ... this.
25385         * config/i386/i386-builtin-types.def
25386         ((V2DF, V2DF, V2DF, INT, V2DF, UQI, INT),
25387         (V4SF, V4SF, V4SF, INT, V4SF, UQI, INT)): New function type aliases.
25388         * config/i386/i386-builtin.def (__builtin_ia32_getexpsd_mask_round,
25389         __builtin_ia32_getexpss_mask_round,     __builtin_ia32_getmantsd_mask_round,
25390         __builtin_ia32_getmantss_mask_round): New builtins.
25391         * config/i386/i386.c (V2DF_FTYPE_V2DF_V2DF_INT_V2DF_UQI_INT,
25392         V4SF_FTYPE_V4SF_V4SF_INT_V4SF_UQI_INT): Handle new types.
25393         (CODE_FOR_avx512f_vgetmantv2df_mask_round,
25394         CODE_FOR_avx512f_vgetmantv4sf_mask_round): New cases.
25395         * config/i386/sse.md
25396         (avx512f_sgetexp<mode><round_saeonly_name>): Changed to ...
25397         avx512f_sgetexp<mode><mask_scalar_name>
25398         <round_saeonly_scalar_name> ... this.
25399         (vgetexp<ssescalarmodesuffix>\t{<round_saeonly_op3>%2, %1, %0|
25400         %0, %1, %2<round_saeonly_op3>}): Changed to ...
25401         vgetexp<ssescalarmodesuffix>
25402         \t{<round_saeonly_scalar_mask_op3>%2, %1, %0<mask_scalar_operand3>|
25403         %0<mask_scalar_operand3>, %1, %2<round_saeonly_scalar_mask_op3>} ... this.
25404         (avx512f_vgetmant<mode><round_saeonly_name>): Changed to ...
25405         avx512f_vgetmant<mode><mask_scalar_name>
25406         <round_saeonly_scalar_name> ... this.
25407         (vgetmant<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
25408         %0, %1, %2<round_saeonly_op4>, %3}): Changed to ...
25409         vgetmant<ssescalarmodesuffix>
25410         \t{%3, <round_saeonly_scalar_mask_op4>%2, %1, %0<mask_scalar_operand4>|
25411         %0<mask_scalar_operand4>, %1, %2
25412         <round_saeonly_scalar_mask_op4>, %3} ... this.
25413         * config/i386/subst.md (mask_scalar_operand4,
25414         round_saeonly_scalar_mask_operand4,     round_saeonly_scalar_mask_op4,
25415         round_saeonly_scalar_nimm_predicate): New subst attributes.
25417 2017-07-06  Julia Koval  <julia.koval@intel.com>
25419         * config/i386/i386.c (ix86_erase_embedded_rounding):
25420         Remove code for old rounding pattern.
25422 2017-07-06  Richard Earnshaw  <rearnsha@arm.com>
25424         * config/arm/t-arm (GTM_H): Add arm-cpu.h.
25426 2017-07-06  Christophe Lyon  <christophe.lyon@linaro.org>
25428         * doc/sourcebuild.texi (Test Directives, Variants of
25429         dg-require-support): Add documentation for dg-require-stack-check.
25431 2017-07-05  Sebastian Peryt  <sebastian.peryt@intel.com>
25433         * config/i386/subst.md (mask_scalar, round_scalar,
25434         round_saeonly_scalar): New meta-templates.
25435         (mask_scalar_name, mask_scalar_operand3, round_scalar_name,
25436         round_scalar_mask_operand3, round_scalar_mask_op3,
25437         round_scalar_constraint, round_scalar_prefix, round_saeonly_scalar_name,
25438         round_saeonly_scalar_mask_operand3, round_saeonly_scalar_mask_op3,
25439         round_saeonly_scalar_constraint,
25440         round_saeonly_scalar_prefix): New subst attribute.
25441         * config/i386/sse.md
25442         (<sse>_vm<plusminus_insn><mode>3<mask_name><round_name>): Renamed to ...
25443         <sse>_vm<plusminus_insn><mode>3<mask_scalar_name>
25444         <round_scalar_name> ... this.
25445         (<sse>_vm<multdiv_mnemonic><mode>3<mask_name><round_name>): Renamed to ...
25446         <sse>_vm<multdiv_mnemonic><mode>3<mask_scalar_name>
25447         <round_scalar_name> ... this.
25448         (<sse>_vm<code><mode>3<mask_name><round_saeonly_name>): Renamed to ...
25449         <sse>_vm<code><mode>3<mask_scalar_name>
25450         <round_saeonly_scalar_name> ... this.
25451         (v<plusminus_mnemonic><ssescalarmodesuffix>
25452         \t{<round_mask_op3>%2, %1, %0<mask_operand3>|
25453         %0<mask_operand3>, %1, %<iptr>2<round_mask_op3>}): Changed to ...
25454         v<plusminus_mnemonic><ssescalarmodesuffix>
25455         \t{<round_scalar_mask_op3>%2, %1, %0<mask_scalar_operand3>|
25456         %0<mask_scalar_operand3>, %1, %<iptr>2<round_scalar_mask_op3>} ... this.
25457         (v<multdiv_mnemonic><ssescalarmodesuffix>
25458         \t{<round_mask_op3>%2, %1, %0<mask_operand3>|
25459         %0<mask_operand3>, %1, %<iptr>2<round_mask_op3>}): Changed to ...
25460         v<multdiv_mnemonic><ssescalarmodesuffix>
25461         \t{<round_scalar_mask_op3>%2, %1, %0<mask_scalar_operand3>|
25462         %0<mask_scalar_operand3>, %1, %<iptr>2<round_scalar_mask_op3>} ... this.
25463         (v<maxmin_float><ssescalarmodesuffix>
25464         \t{<round_saeonly_mask_op3>%2, %1, %0<mask_operand3>|
25465         %0<mask_operand3>, %1, %<iptr>2<round_saeonly_mask_op3>}): Changed to ...
25466         v<maxmin_float><ssescalarmodesuffix>
25467         \t{<round_saeonly_scalar_mask_op3>%2, %1, %0<mask_scalar_operand3>|
25468         %0<mask_scalar_operand3>, %1, %<iptr>2
25469         <round_saeonly_scalar_mask_op3>} ... this.
25471 2017-07-05  Richard Earnshaw  <rearnsha@arm.com>
25473         * config/arm/arm.c (arm_fixed_condition_code_regs): New function.
25474         (TARGET_FIXED_CONDITION_CODE_REGS): Redefine.
25476 2017-07-05  Richard Sandiford  <richard.sandiford@linaro.org>
25477             Alan Hayward  <alan.hayward@arm.com>
25478             David Sherwood  <david.sherwood@arm.com>
25480         * combine.c (simplify_if_then_else): Remove "enum" before
25481         "machine_mode".
25482         * compare-elim.c (can_eliminate_compare): Likewise.
25483         * config/aarch64/aarch64-builtins.c (aarch64_simd_builtin_std_type):
25484         Likewise.
25485         (aarch64_lookup_simd_builtin_type): Likewise.
25486         (aarch64_simd_builtin_type): Likewise.
25487         (aarch64_init_simd_builtin_types): Likewise.
25488         (aarch64_simd_expand_args): Likewise.
25489         * config/aarch64/aarch64-protos.h (aarch64_simd_attr_length_rglist):
25490         Likewise.
25491         (aarch64_reverse_mask): Likewise.
25492         (aarch64_simd_emit_reg_reg_move): Likewise.
25493         (aarch64_gen_adjusted_ldpstp): Likewise.
25494         (aarch64_ccmp_mode_to_code): Likewise.
25495         (aarch64_operands_ok_for_ldpstp): Likewise.
25496         (aarch64_operands_adjust_ok_for_ldpstp): Likewise.
25497         * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
25498         Likewise.
25499         (aarch64_min_divisions_for_recip_mul): Likewise.
25500         (aarch64_reassociation_width): Likewise.
25501         (aarch64_get_condition_code_1): Likewise.
25502         (aarch64_simd_emit_reg_reg_move): Likewise.
25503         (aarch64_simd_attr_length_rglist): Likewise.
25504         (aarch64_reverse_mask): Likewise.
25505         (aarch64_operands_ok_for_ldpstp): Likewise.
25506         (aarch64_operands_adjust_ok_for_ldpstp): Likewise.
25507         (aarch64_gen_adjusted_ldpstp): Likewise.
25508         * config/aarch64/cortex-a57-fma-steering.c (fma_node::rename):
25509         Likewise.
25510         * config/arc/arc.c (legitimate_offset_address_p): Likewise.
25511         * config/arm/arm-builtins.c (arm_simd_builtin_std_type): Likewise.
25512         (arm_lookup_simd_builtin_type): Likewise.
25513         (arm_simd_builtin_type): Likewise.
25514         (arm_init_simd_builtin_types): Likewise.
25515         (arm_expand_builtin_args): Likewise.
25516         * config/arm/arm-protos.h (arm_expand_builtin): Likewise.
25517         * config/ft32/ft32.c (ft32_libcall_value): Likewise.
25518         (ft32_setup_incoming_varargs): Likewise.
25519         (ft32_function_arg): Likewise.
25520         (ft32_function_arg_advance): Likewise.
25521         (ft32_pass_by_reference): Likewise.
25522         (ft32_arg_partial_bytes): Likewise.
25523         (ft32_valid_pointer_mode): Likewise.
25524         (ft32_addr_space_pointer_mode): Likewise.
25525         (ft32_addr_space_legitimate_address_p): Likewise.
25526         * config/i386/i386-protos.h (ix86_operands_ok_for_move_multiple):
25527         Likewise.
25528         * config/i386/i386.c (ix86_setup_incoming_vararg_bounds): Likewise.
25529         (ix86_emit_outlined_ms2sysv_restore): Likewise.
25530         (iamcu_alignment): Likewise.
25531         (canonicalize_vector_int_perm): Likewise.
25532         (ix86_noce_conversion_profitable_p): Likewise.
25533         (ix86_mpx_bound_mode): Likewise.
25534         (ix86_operands_ok_for_move_multiple): Likewise.
25535         * config/microblaze/microblaze-protos.h
25536         (microblaze_expand_conditional_branch_reg): Likewise.
25537         * config/microblaze/microblaze.c
25538         (microblaze_expand_conditional_branch_reg): Likewise.
25539         * config/powerpcspe/powerpcspe.c (rs6000_init_hard_regno_mode_ok):
25540         Likewise.
25541         (rs6000_reassociation_width): Likewise.
25542         (rs6000_invalid_binary_op): Likewise.
25543         (fusion_p9_p): Likewise.
25544         (emit_fusion_p9_load): Likewise.
25545         (emit_fusion_p9_store): Likewise.
25546         * config/riscv/riscv-protos.h (riscv_regno_mode_ok_for_base_p):
25547         Likewise.
25548         (riscv_hard_regno_mode_ok_p): Likewise.
25549         (riscv_address_insns): Likewise.
25550         (riscv_split_symbol): Likewise.
25551         (riscv_legitimize_move): Likewise.
25552         (riscv_function_value): Likewise.
25553         (riscv_hard_regno_nregs): Likewise.
25554         (riscv_expand_builtin): Likewise.
25555         * config/riscv/riscv.c (riscv_build_integer_1): Likewise.
25556         (riscv_build_integer): Likewise.
25557         (riscv_split_integer): Likewise.
25558         (riscv_legitimate_constant_p): Likewise.
25559         (riscv_cannot_force_const_mem): Likewise.
25560         (riscv_regno_mode_ok_for_base_p): Likewise.
25561         (riscv_valid_base_register_p): Likewise.
25562         (riscv_valid_offset_p): Likewise.
25563         (riscv_valid_lo_sum_p): Likewise.
25564         (riscv_classify_address): Likewise.
25565         (riscv_legitimate_address_p): Likewise.
25566         (riscv_address_insns): Likewise.
25567         (riscv_load_store_insns): Likewise.
25568         (riscv_force_binary): Likewise.
25569         (riscv_split_symbol): Likewise.
25570         (riscv_force_address): Likewise.
25571         (riscv_legitimize_address): Likewise.
25572         (riscv_move_integer): Likewise.
25573         (riscv_legitimize_const_move): Likewise.
25574         (riscv_legitimize_move): Likewise.
25575         (riscv_address_cost): Likewise.
25576         (riscv_subword): Likewise.
25577         (riscv_output_move): Likewise.
25578         (riscv_canonicalize_int_order_test): Likewise.
25579         (riscv_emit_int_order_test): Likewise.
25580         (riscv_function_arg_boundary): Likewise.
25581         (riscv_pass_mode_in_fpr_p): Likewise.
25582         (riscv_pass_fpr_single): Likewise.
25583         (riscv_pass_fpr_pair): Likewise.
25584         (riscv_get_arg_info): Likewise.
25585         (riscv_function_arg): Likewise.
25586         (riscv_function_arg_advance): Likewise.
25587         (riscv_arg_partial_bytes): Likewise.
25588         (riscv_function_value): Likewise.
25589         (riscv_pass_by_reference): Likewise.
25590         (riscv_setup_incoming_varargs): Likewise.
25591         (riscv_print_operand): Likewise.
25592         (riscv_elf_select_rtx_section): Likewise.
25593         (riscv_save_restore_reg): Likewise.
25594         (riscv_for_each_saved_reg): Likewise.
25595         (riscv_register_move_cost): Likewise.
25596         (riscv_hard_regno_mode_ok_p): Likewise.
25597         (riscv_hard_regno_nregs): Likewise.
25598         (riscv_class_max_nregs): Likewise.
25599         (riscv_memory_move_cost): Likewise.
25600         * config/rl78/rl78-protos.h (rl78_split_movsi): Likewise.
25601         * config/rl78/rl78.c (rl78_split_movsi): Likewise.
25602         (rl78_addr_space_address_mode): Likewise.
25603         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
25604         Likewise.
25605         * config/rs6000/rs6000.c (rs6000_init_hard_regno_mode_ok): Likewise.
25606         (rs6000_reassociation_width): Likewise.
25607         (rs6000_invalid_binary_op): Likewise.
25608         (fusion_p9_p): Likewise.
25609         (emit_fusion_p9_load): Likewise.
25610         (emit_fusion_p9_store): Likewise.
25611         * config/visium/visium-protos.h (prepare_move_operands): Likewise.
25612         (ok_for_simple_move_operands): Likewise.
25613         (ok_for_simple_move_strict_operands): Likewise.
25614         (ok_for_simple_arith_logic_operands): Likewise.
25615         (visium_legitimize_reload_address): Likewise.
25616         (visium_select_cc_mode): Likewise.
25617         (output_cbranch): Likewise.
25618         (visium_split_double_move): Likewise.
25619         (visium_expand_copysign): Likewise.
25620         (visium_expand_int_cstore): Likewise.
25621         (visium_expand_fp_cstore): Likewise.
25622         * config/visium/visium.c (visium_pass_by_reference): Likewise.
25623         (visium_function_arg): Likewise.
25624         (visium_function_arg_advance): Likewise.
25625         (visium_libcall_value): Likewise.
25626         (visium_setup_incoming_varargs): Likewise.
25627         (visium_legitimate_constant_p): Likewise.
25628         (visium_legitimate_address_p): Likewise.
25629         (visium_legitimize_address): Likewise.
25630         (visium_secondary_reload): Likewise.
25631         (visium_register_move_cost): Likewise.
25632         (visium_memory_move_cost): Likewise.
25633         (prepare_move_operands): Likewise.
25634         (ok_for_simple_move_operands): Likewise.
25635         (ok_for_simple_move_strict_operands): Likewise.
25636         (ok_for_simple_arith_logic_operands): Likewise.
25637         (visium_function_value_1): Likewise.
25638         (rtx_ok_for_offset_p): Likewise.
25639         (visium_legitimize_reload_address): Likewise.
25640         (visium_split_double_move): Likewise.
25641         (visium_expand_copysign): Likewise.
25642         (visium_expand_int_cstore): Likewise.
25643         (visium_expand_fp_cstore): Likewise.
25644         (visium_split_cstore): Likewise.
25645         (visium_select_cc_mode): Likewise.
25646         (visium_split_cbranch): Likewise.
25647         (output_cbranch): Likewise.
25648         (visium_print_operand_address): Likewise.
25649         * expmed.c (flip_storage_order): Likewise.
25650         * expmed.h (emit_cstore): Likewise.
25651         (flip_storage_order): Likewise.
25652         * genrecog.c (validate_pattern): Likewise.
25653         * hsa-gen.c (gen_hsa_addr): Likewise.
25654         * internal-fn.c (expand_arith_overflow): Likewise.
25655         * ira-color.c (allocno_copy_cost_saving): Likewise.
25656         * lra-assigns.c (find_hard_regno_for_1): Likewise.
25657         * lra-constraints.c (prohibited_class_reg_set_mode_p): Likewise.
25658         (process_invariant_for_inheritance): Likewise.
25659         * lra-eliminations.c (move_plus_up): Likewise.
25660         * omp-low.c (lower_oacc_reductions): Likewise.
25661         * simplify-rtx.c (simplify_subreg): Likewise.
25662         * target.def (TARGET_SETUP_INCOMING_VARARG_BOUNDS): Likewise.
25663         (TARGET_CHKP_BOUND_MODE): Likewise..
25664         * targhooks.c (default_chkp_bound_mode): Likewise.
25665         (default_setup_incoming_vararg_bounds): Likewise.
25666         * targhooks.h (default_chkp_bound_mode): Likewise.
25667         (default_setup_incoming_vararg_bounds): Likewise.
25668         * tree-ssa-math-opts.c (divmod_candidate_p): Likewise.
25669         * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
25670         (have_whole_vector_shift): Likewise.
25671         * tree-vect-stmts.c (vectorizable_load): Likewise.
25672         * doc/tm.texi: Regenerate.
25674 2017-07-05  Georg-Johann Lay  <avr@gjlay.de>
25676         Graceful degrade if Binutils PR21472 is not available.
25678         PR target/81072
25679         * configure.ac [target=avr]: WARN instead of ERROR if avrxmega3
25680         .rodata in flash test fails.
25681         (HAVE_LD_AVR_AVRXMEGA3_RODATA_IN_FLASH): Define it if test passes.
25682         * confgure: Regenerate.
25683         * config.in: Regenerate.
25684         * config/avr/avr.c (avr_asm_named_section)
25685         [HAVE_LD_AVR_AVRXMEGA3_RODATA_IN_FLASH]: Only trigger
25686         __do_copy_data for stuff in .rodata if flash_pm_offset = 0.
25687         (avr_asm_init_sections): Same.
25689 2017-07-05  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
25691         * config/arm/neon.md (fma<VCVTF:mode>4): Remove %?.
25692         (fma<VH:mode>4_intrinsic): Likewise.
25693         (*fmsub<VCVTF:mode>4): Likewise.
25694         (*fmsub<VH:mode>4_intrinsic): Likewise.
25696 2017-07-05  Georg-Johann Lay  <avr@gjlay.de>
25698         PR target/81305
25699         * config/avr/avr.c (avr_out_movhi_mr_r_xmega) [CONSTANT_ADDRESS_P]:
25700         Don't depend on "optimize > 0".
25701         (out_movhi_r_mr, out_movqi_mr_r): Same.
25702         (out_movhi_mr_r, out_movqi_r_mr): Same.
25703         (avr_address_cost) [CONSTANT_ADDRESS_P]: Don't depend cost for
25704         io_address_operand on "optimize > 0".
25706 2017-07-05  Bin Cheng  <bin.cheng@arm.com>
25708         * tree-loop-distribution.c: Add general explanantion on the pass.
25709         (generate_loops_for_partition): Mark distributed loop.
25710         (pg_add_dependence_edges): New parameter.  Handle alias data
25711         dependence specially and record it in the parameter if asked.
25712         (struct pg_vdata, pg_edata, pg_edge_callback_data): New structs.
25713         (init_partition_graph_vertices, add_partition_graph_edge): New.
25714         (pg_skip_alias_edge, free_partition_graph_edata_cb): New.
25715         (free_partition_graph_vdata, build_partition_graph): New.
25716         (sort_partitions_by_post_order, merge_dep_scc_partitions): New.
25717         (pg_collect_alias_ddrs, break_alias_scc_partitions): New.
25718         (data_ref_segment_size, latch_dominated_by_data_ref): New.
25719         (compute_alias_check_pairs, version_loop_by_alias_check): New.
25720         (version_for_distribution_p, finalize_partitions): New.
25721         (distribute_loop): Handle alias data dependence specially.  Factor
25722         out loop fusion code as functions and call these functions.
25724 2017-07-05  Bin Cheng  <bin.cheng@arm.com>
25726         * tree-loop-distribution.c (classify_partition): New parameter and
25727         better handle reduction statement.
25728         (rdg_build_partitions): Revise comment.
25729         (distribute_loop): Compute statements in all partitions and pass it
25730         to classify_partition.
25732 2017-07-05  Bin Cheng  <bin.cheng@arm.com>
25734         * tree-loop-distribution.c (enum partition_type): New.
25735         (struct partition): New field type.
25736         (partition_merge_into): Add parameter.  Update partition type.
25737         (data_dep_in_cycle_p, update_type_for_merge): New functions.
25738         (build_rdg_partition_for_vertex): Compute partition type.
25739         (rdg_build_partitions): Dump partition type.
25740         (distribute_loop): Update calls to partition_merge_into.
25742 2017-07-05  Bin Cheng  <bin.cheng@arm.com>
25744         * tree-loop-distribution.c (struct ddr_hasher): New.
25745         (ddr_hasher::hash, ::equal, get_data_dependence): New function.
25746         (ddrs_table): New.
25747         (classify_partition): Call get_data_dependence.
25748         (pg_add_dependence_edges): Ditto.
25749         (distribute_loop): Release data dependence hash table.
25751 2017-07-05  Bin Cheng  <bin.cheng@arm.com>
25753         * tree-loop-distribution.c (ref_base_address): Delete.
25754         (similar_memory_accesses): Rename ...
25755         (share_memory_accesses): ... to this.  Check if partitions access
25756         the same memory reference.
25757         (distribute_loop): Call share_memory_accesses.
25759 2017-07-05  Bin Cheng  <bin.cheng@arm.com>
25761         * tree-loop-distribution.c (struct partition): New field recording
25762         its data reference.
25763         (partition_alloc, partition_free): Init and release data refs.
25764         (partition_merge_into): Merge data refs.
25765         (build_rdg_partition_for_vertex): Collect data refs for partition.
25766         (pg_add_dependence_edges): Change parameters from vector to bitmap.
25767         Update uses.
25768         (distribute_loop): Remve data refs from vertice data of partition
25769         graph.
25771 2017-07-05  Bin Cheng  <bin.cheng@arm.com>
25773         * tree-loop-distribution.c (params.h): Include header file.
25774         (MAX_DATAREFS_NUM, DR_INDEX): New macro.
25775         (datarefs_vec): New global var.
25776         (create_rdg_vertices): Use datarefs_vec directly.
25777         (free_rdg): Don't free data references.
25778         (build_rdg): Update use.  Don't free data references.
25779         (distribute_loop): Compute global variable for data references.
25780         Bail out if there are too many data references.
25782 2017-07-05  Bin Cheng  <bin.cheng@arm.com>
25784         * tree-loop-distribution.c (loop_nest): New global var.
25785         (build_rdg): Use loop directly, rather than loop nest.
25786         (pg_add_dependence_edges): Remove loop nest parameter.  Use global
25787         variable directly.
25788         (distribute_loop): Compute global variable loop nest.  Update use.
25790 2017-07-05  Bin Cheng  <bin.cheng@arm.com>
25792         * tree-loop-distribution.c (enum fuse_type, fuse_message): New.
25793         (partition_merge_into): New parameter.  Dump reason for fusion.
25794         (distribute_loop): Update use of partition_merge_into.
25796 2017-07-05  Bin Cheng  <bin.cheng@arm.com>
25798         * tree-loop-distribution.c (bb_top_order_index): New.
25799         (bb_top_order_index_size, bb_top_order_cmp): New.
25800         (stmts_from_loop): Use topological order.
25801         (pass_loop_distribution::execute): Compute and release topological
25802         order for basic blocks.
25804 2017-07-05  Bin Cheng  <bin.cheng@arm.com>
25806         * tree-loop-distribution.c (pass_loop_distribution::execute): Skip
25807         if no loops.
25809 2017-07-05  Bin Cheng  <bin.cheng@arm.com>
25811         * cfgloop.h (struct loop): Add comment.  New field orig_loop_num.
25812         * cfgloopmanip.c (lv_adjust_loop_entry_edge): Comment change.
25813         * internal-fn.c (expand_LOOP_DIST_ALIAS): New function.
25814         * internal-fn.def (LOOP_DIST_ALIAS): New.
25815         * tree-vectorizer.c (fold_loop_vectorized_call): Rename to ...
25816         (fold_loop_internal_call): ... this.
25817         (vect_loop_dist_alias_call): New function.
25818         (set_uid_loop_bbs): Call fold_loop_internal_call.
25819         (vectorize_loops): Fold IFN_LOOP_VECTORIZED and IFN_LOOP_DIST_ALIAS
25820         internal calls.
25822 2017-07-04  Uros Bizjak  <ubizjak@gmail.com>
25824         PR target/81300
25825         * config/i386/i386.md (setcc + movzbl/and to xor + setcc peepholes):
25826         Require dead FLAGS_REG at the beginning of a peephole.
25828 2017-07-04  Uros Bizjak  <ubizjak@gmail.com>
25830         PR target/81294
25831         * config/i386/adxintrin.h (_subborrow_u32): Swap _X and _Y
25832         arguments in the call to __builtin_ia32_sbb_u32.
25833         (_subborrow_u64): Swap _X and _Y arguments in the call to
25834         __builtin_ia32_sbb_u64.
25836 2017-07-04  Jakub Jelinek  <jakub@redhat.com>
25838         PR debug/81278
25839         * tree-vrp.c (compare_assert_loc): Turn into a function template
25840         with stable template parameter.  Only test if a->e is NULL,
25841         !a->e == !b->e has been verified already.  Use e == NULL or
25842         e != NULL instead of e or ! e tests.  If stable is true, don't use
25843         iterative_hash_expr, on the other side allow a or b or both NULL
25844         and sort the NULLs last.
25845         (process_assert_insertions): Sort using compare_assert_loc<false>
25846         instead of compare_assert_loc, later sort using
25847         compare_assert_loc<true> before calling process_assert_insertions_for
25848         in a loop.  Use break instead of continue once seen NULL pointer.
25850 2017-07-04  Thomas Preud'homme  <thomas.preudhomme@arm.com>
25852         * config/arm/driver-arm.c (arm_cpu_table): Add entry for ARM
25853         Cortex-R7 and Cortex-R8 processors.
25855 2017-07-04  Jan Hubicka  <hubicka@ucw.cz>
25857         * ipa-utils.c (ipa_merge_profiles): Fix merging when dst is
25858         uninitialized while src is not.
25860 2017-07-04  Richard Earnshaw  <rearnsha@arm.com>
25862         * common/config/arm/arm-common.c: Adjust include path for
25863         arm-cpu-cdata.h
25864         * config/arm/t-arm (TM_H): Adjust path for arm-cpu.h.
25865         (arm-cpu.h): Create in build directory.  Adjust dependency rules.
25866         (arm-cpu-data.h): Likewise.
25867         (arm-cpu-cdata.h): Likewise.
25868         * config/arm/arm-cpu.h: Delete.
25869         * config/arm/arm-cpu-cdata.h: Delete.
25870         * config/arm/arm-cpu-data.h: Delete.
25872 2017-07-04  James Greenhalgh  <james.greenhalgh@arm.com>
25874         * config/arm/arm-cpus.in (cortex-a55): New.
25875         (cortex-a75): Likewise.
25876         (cortex-a75.cortex-a55): Likewise.
25877         * config/arm/driver-arm.c (arm_cpu_table): Add cortex-a55 and
25878         cortex-a75.
25879         * doc/invoke.texi (-mcpu): Document cortex-a55 and cortex-a75.
25880         * config/arm/arm-cpu-cdata.h: Regenerate.
25881         * config/arm/arm-cpu-data.h: Regenerate.
25882         * config/arm/arm-cpu.h: Regenerate.
25883         * config/arm/arm-tables.opt: Regenerate.
25884         * config/arm/arm-tune.md: Regenerate.
25886 2017-07-04  Jan Hubicka  <hubicka@ucw.cz>
25888         * haifa-sched.c (sched_create_recovery_edges): Update profile.
25890 2017-07-04  Jan Hubicka  <hubicka@ucw.cz>
25892         * bb-reorder.c (better_edge_p): Fix handling of uninitialized
25893         probability.
25895 2017-07-04  Richard Sandiford  <richard.sandiford@linaro.org>
25897         PR tree-optimization/81292
25898         * tree-ssa-strlen.c (handle_builtin_strlen): When setting
25899         full_string_p, also call adjust_related_strinfos if the adjustment
25900         is simple, otherwise invalidate related strinfos.
25902 2017-07-04  Martin Liska  <mliska@suse.cz>
25904         PR sanitizer/81040
25905         * sanopt.c (sanitize_rewrite_addressable_params): Mark the
25906         newly created variable as DECL_IGNORED_P.
25908 2017-07-04  Martin Liska  <mliska@suse.cz>
25910         PR ipa/81293
25911         * ipa-inline.c (inline_small_functions):
25912         Use xstrdup_for_dump.
25914 2017-07-04  Tom de Vries  <tom@codesourcery.com>
25916         * graph.c (draw_cfg_edges): Save and restore EDGE_DFS_BACK.
25918 2017-07-03  Dominique d'Humieres  <dominiq@lps.ens.fr>
25920         PR target/81033
25921         * config/darwin.c (darwin_function_switched_text_sections):
25922         Fix spaces.
25924 2017-07-03  Jan Hubicka  <hubicka@ucw.cz>
25926         * tree-vect-loop-manip.c (vect_do_peeling): Fix scaling up.
25928 2017-07-03  Richard Earnshaw  <rearnsha@arm.com>
25930         * doc/invoke.texi (ARM Options): Add -mbe8 and -mbe32 to option summary.
25932 2017-07-03  Richard Sandiford  <richard.sandiford@linaro.org>
25934         * tree-vect-loop.c (vect_analyze_loop_2): Treat min_scalar_loop_bound,
25935         min_profitable_iters, and th as inclusive lower bounds.
25936         Fix LOOP_VINFO_PEELING_FOR_GAPS condition.
25937         (vect_estimate_min_profitable_iters): Return inclusive lower bounds
25938         for min_profitable_iters and min_profitable_estimate.
25939         (vect_transform_loop): Treat th as an inclusive lower bound.
25940         * tree-vect-loop-manip.c (vect_loop_versioning): Likewise.
25942 2017-07-03  Dominique d'Humieres  <dominiq@lps.ens.fr>
25944         PR target/81033
25945         * config/darwin.c (darwin_function_switched_text_sections):
25946         Replace DECL_NAME with DECL_ASSEMBLER_NAME, split assemble_name_raw
25947         in two pieces, and suppress the use of buf.
25949 2017-07-03  Nathan Sidwell  <nathan@acm.org>
25951         * hash-table.h (hash_table_mod1): Fix indentation.
25953 2017-07-02  Jan Hubicka  <hubicka@ucw.cz>
25955         PR middle-end/81290
25956         * predict.c (force_edge_cold): Be more careful about propagation
25957         backward.
25958         * profile-count.h (profile_probability::guessed,
25959         profile_probability::fdo, profile_count::guessed, profile_count::fdo):
25960         New.
25961         * tree-ssa-threadupdate.c (recompute_probabilities): Result is guessed.
25963 2017-07-03  James Greenhalgh  <james.greenhalgh@arm.com>
25965         * doc/invoke.texi (rcpc architecture extension): Document it.
25967 2017-07-03  Richard Biener  <rguenther@suse.de>
25969         PR tree-optimization/60510
25970         * tree-vect-loop.c (vect_create_epilog_for_reduction): Pass in
25971         the scalar reduction PHI and use it.
25972         (vectorizable_reduction): Properly guard the single_defuse_cycle
25973         path for non-SLP reduction chains where we cannot use it.
25974         Rework reduc_def/index and vector type deduction.  Rework
25975         vector operand gathering during reduction op code-gen.
25976         * tree-vect-slp.c (vect_analyze_slp): For failed SLP reduction
25977         chains dissolve the chain and leave it to non-SLP reduction
25978         handling.
25980 2017-07-03  Richard Sandiford  <richard.sandiford@linaro.org>
25982         * tree-data-ref.h (dr_alignment): Declare.
25983         * tree-data-ref.c (dr_alignment): New function.
25984         * tree-vectorizer.h (dataref_aux): Remove base_element_aligned.
25985         * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
25986         set it.
25987         * tree-vect-stmts.c (vectorizable_store): Use dr_alignment.
25989 2017-07-03  Richard Sandiford  <richard.sandiford@linaro.org>
25991         * tree-data-ref.h (innermost_loop_behavior): Add base_alignment
25992         and base_misalignment fields.
25993         (DR_BASE_ALIGNMENT, DR_BASE_MISALIGNMENT): New macros.
25994         * tree-data-ref.c: Include builtins.h.
25995         (dr_analyze_innermost): Set up the new innmost_loop_behavior fields.
25996         * tree-vectorizer.h (STMT_VINFO_DR_BASE_ALIGNMENT): New macro.
25997         (STMT_VINFO_DR_BASE_MISALIGNMENT): Likewise.
25998         * tree-vect-data-refs.c: Include tree-cfg.h.
25999         (vect_compute_data_ref_alignment): Use the new innermost_loop_behavior
26000         fields instead of calculating an alignment here.
26001         (vect_analyze_data_refs): Use dr_analyze_innermost.  Dump the new
26002         innermost_loop_behavior fields.
26004 2017-07-03  Richard Sandiford  <richard.sandiford@linaro.org>
26006         * tree-data-ref.h (innermost_loop_behavior): Add a step_alignment
26007         field.
26008         (DR_STEP_ALIGNMENT): New macro.
26009         * tree-vectorizer.h (STMT_VINFO_DR_STEP_ALIGNMENT): Likewise.
26010         * tree-data-ref.c (dr_analyze_innermost): Initalize step_alignment.
26011         (create_data_ref): Print it.
26012         * tree-vect-stmts.c (vectorizable_load): Use the step alignment
26013         to tell whether the step preserves vector (mis)alignment.
26014         * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Likewise.
26015         Move the check for an integer step and generalise to all INTEGER_CST.
26016         (vect_analyze_data_refs): Set DR_STEP_ALIGNMENT when setting DR_STEP.
26017         Print the outer step alignment.
26019 2017-07-03  Richard Sandiford  <richard.sandiford@linaro.org>
26021         * tree-data-ref.h (innermost_loop_behavior): Replace aligned_to
26022         with offset_alignment.
26023         (DR_ALIGNED_TO): Delete.
26024         (DR_OFFSET_ALIGNMENT): New macro.
26025         * tree-vectorizer.h (STMT_VINFO_DR_ALIGNED_TO): Delete.
26026         (STMT_VINFO_DR_OFFSET_ALIGNMENT): New macro.
26027         * tree-data-ref.c (dr_analyze_innermost): Update after above changes.
26028         (create_data_ref): Likewise.
26029         * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Likewise.
26030         (vect_analyze_data_refs): Likewise.
26031         * tree-if-conv.c (if_convertible_loop_p_1): Use memset before
26032         creating dummy innermost behavior.
26034 2017-07-03  Richard Sandiford  <richard.sandiford@linaro.org>
26036         * tree-data-ref.h (dr_analyze_innermost): Replace the dr argument
26037         with a "innermost_loop_behavior *" and refeence tree.
26038         * tree-data-ref.c (dr_analyze_innermost): Likewise.
26039         (create_data_ref): Update call accordingly.
26040         * tree-predcom.c (find_looparound_phi): Likewise.
26042 2017-07-03  Richard Sandiford  <richard.sandiford@linaro.org>
26044         * tree-vectorizer.h (_stmt_vec_info): Replace individual dr_*
26045         fields with dr_wrt_vec_loop.
26046         (STMT_VINFO_DR_BASE_ADDRESS, STMT_VINFO_DR_INIT, STMT_VINFO_DR_OFFSET)
26047         (STMT_VINFO_DR_STEP, STMT_VINFO_DR_ALIGNED_TO): Update accordingly.
26048         (STMT_VINFO_DR_WRT_VEC_LOOP): New macro.
26049         (vect_dr_behavior): New function.
26050         (vect_create_addr_base_for_vector_ref): Remove loop parameter.
26051         * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
26052         vect_dr_behavior.  Use a step_preserves_misalignment_p boolean to
26053         track whether the step preserves the misalignment.
26054         (vect_create_addr_base_for_vector_ref): Remove loop parameter.
26055         Use vect_dr_behavior.
26056         (vect_setup_realignment): Update call accordingly.
26057         (vect_create_data_ref_ptr): Likewise.  Use vect_dr_behavior.
26058         * tree-vect-loop-manip.c (vect_gen_prolog_loop_niters): Update
26059         call to vect_create_addr_base_for_vector_ref.
26060         (vect_create_cond_for_align_checks): Likewise.
26061         * tree-vect-patterns.c (vect_recog_bool_pattern): Copy
26062         STMT_VINFO_DR_WRT_VEC_LOOP as a block.
26063         (vect_recog_mask_conversion_pattern): Likewise.
26064         * tree-vect-stmts.c (compare_step_with_zero): Use vect_dr_behavior.
26065         (new_stmt_vec_info): Remove redundant zeroing.
26067 2017-07-03  Richard Earnshaw  <rearnsha@arm.com>
26069         * common/config/arm/arm-common.c (arm_be8_option): New function.
26070         * config/arm/arm-isa.h (isa_feature): Add new feature bit isa_bit_be8.
26071         (ISA_ARMv6): Add isa_bit_be8.
26072         * config/arm/arm.h (arm_be8_option): Add prototype.
26073         (BE8_SPEC_FUNCTION): New define.
26074         (EXTRA_SPEC_FUNCTIONS): Add BE8_SPEC_FUNCTION.
26075         * config/arm/arm.opt (mbig-endian): Mark as Negative of mlittle-endian.
26076         (mlittle-endian): Similarly.
26077         (mbe8, mbe32): New options.
26078         * config/arm/bpabi.h (BE8_LINK_SPEC): Call arm_be8_option.
26079         * doc/invoke.texi (ARM Options): Document -mbe8 and -mbe32.
26081 2017-07-02  Jan Hubicka  <hubicka@ucw.cz>
26083         * tree-cfg.c (gimple_find_sub_bbs): Fix profile updating.
26085 2017-07-02  Jan Hubicka  <hubicka@ucw.cz>
26087         * tree-cfgcleanup.c (want_merge_blocks_p): New function.
26088         (cleanup_tree_cfg_bb): Use it.
26089         * profile-count.h (profile_count::of_for_merging, profile_count::merge):
26090         New functions.
26091         * tree-cfg.c (gimple_merge_blocks): Use profile_count::merge.
26093 2017-07-02  Jan Hubicka  <hubicka@ucw.cz>
26095         PR bootstrap/81285
26096         * loop-doloop.c (add_test): Update profile.
26098 2017-07-03  Martin Liska  <mliska@suse.cz>
26100         PR sanitize/81040
26101         * sanopt.c (rewrite_usage_of_param): New function.
26102         (sanitize_rewrite_addressable_params): Likewise.
26103         (pass_sanopt::execute): Call rewrite_usage_of_param.
26105 2017-07-03  Richard Biener  <rguenther@suse.de>
26107         * tree-vect-loop.c (vect_create_epilog_for_reduction): Revert
26108         back to using VIEW_CONVERT_EXPR.
26110 2017-07-03  Martin Liska  <mliska@suse.cz>
26112         PR other/78366
26113         * doc/extend.texi: Document when a resolver function is
26114         generated for target_clones.
26116 2017-07-03  Martin Liska  <mliska@suse.cz>
26118         * asan.c (asan_emit_stack_protection): Unpoison just red zones
26119         and shadow memory of auto variables which are subject of
26120         use-after-scope sanitization.
26121         (asan_expand_mark_ifn): Add do set only when is_poison.
26123 2016-07-03  Richard Biener  <rguenther@suse.de>
26125         * tree-vect-loop.c (vect_analyze_loop_operations): Also analyze
26126         reduction PHIs.
26127         (vect_force_simple_reduction): Record reduction def -> phi mapping.
26128         (vectorizable_reduction): Perform reduction PHI creation when
26129         visiting a reduction PHI and adjust and simplify code generation
26130         phase of the reduction op.  Cache dts, use fold_binary, not fold_build2.
26131         (vect_transform_loop): Visit reduction PHIs.
26132         * tree-vect-slp.c (vect_get_and_check_slp_defs): Record reduction
26133         defs into the SLP tree.
26134         (vect_build_slp_tree): Reduction defs terminate the recursion.
26135         * tree-vect-stmts.c (vect_get_vec_def_for_operand_1): Allow lookup
26136         of reduction defs.
26137         (vect_get_vec_defs_for_stmt_copy): Export.
26138         (vect_get_vec_defs): Likewise.
26139         * tree-vectorizer.h (struct _stmt_vec_info): Amend reduc_def
26140         purpose.
26141         (vect_get_vec_defs_for_stmt_copy): Declare.
26142         (vect_get_vec_defs): Likewise.
26144 2017-07-03  Richard Sandiford  <richard.sandiford@linaro.org>
26146         * tree-data-ref.c (dr_analyze_innermost): Replace the "nest"
26147         parameter with a "loop" parameter and use it instead of the
26148         loop containing DR_STMT.  Don't check simple_iv when doing
26149         BB analysis.  Describe the two analysis modes in the comment.
26151 2017-07-03  Tom de Vries  <tom@codesourcery.com>
26153         PR tree-optimization/69468
26154         * tree-ssa-tail-merge.c (ignore_edge_flags): New constant.
26155         (find_same_succ_bb): Handle ignore_edge_flags.
26157 2017-07-03  Tom de Vries  <tom@codesourcery.com>
26159         PR tree-optimization/81192
26160         * tree-ssa-tail-merge.c (same_succ_hash): Use bb->loop_father->num in
26161         hash.
26162         (same_succ::equal): Don't find bbs to be equal if bb->loop_father
26163         differs.
26164         (find_same_succ_bb): Remove obsolete test on bb->loop_father->latch.
26166 2017-07-03  Tom de Vries  <tom@codesourcery.com>
26168         PR tree-optimization/81192
26169         * tree-ssa-tail-merge.c (same_succ_flush_bb): Handle
26170         BB_SAME_SUCC (bb) == NULL.
26172 2017-07-02  Jan Hubicka  <hubicka@ucw.cz>
26174         * cfgrtl.c (rtl_verify_edges): Enable checking of profile_probability
26175         consistency.
26177 2017-07-02  Jan Hubicka  <hubicka@ucw.cz>
26179         * dumpfile.c: Include profile-count.h
26180         * tree-cfg.c (gimple_duplicate_sese_tail): Drop UNUSED attributes;
26181         update profile.
26182         (insert_cond_bb): Update profile.
26183         * tree-cfg.h (insert_cond_bb): Update prototype.
26184         * tree-chkp-opt.c (chkp_optimize_string_function_calls): Update.
26185         * tree-dump.c: Do not include tree-cfg.
26187 2017-07-02  Jan Hubicka  <hubicka@ucw.cz>
26189         * bb-reorder.c (fix_up_crossing_landing_pad): Update profile.
26191 2017-07-02  Jan Hubicka  <hubicka@ucw.cz>
26193         * expect.c (dw2_build_landing_pads): Update profile of the landing pad
26194         bb.
26196 2017-07-02  Jan Hubicka  <hubicka@ucw.cz>
26198         * tree-complex.c (expand_complex_div_wide): update profile.
26200 2017-07-02  Richard Sandiford  <richard.sandiford@linaro.org>
26201             Alan Hayward  <alan.hayward@arm.com>
26202             David Sherwood  <david.sherwood@arm.com>
26204         * Makefile.in (MACHMODE_H): Remove insn-modes.h
26205         (CORETYPES_H): New define.
26206         (MOSTLYCLEANFILES): Add insn-modes-inline.h.
26207         (insn-modes-inline.h, s-modes-inline-h): New rules.
26208         (generated_files): Add insn-modes-inline.h.
26209         (RTL_BASE_H, TREE_CORE_H): Use CORETYPES_H instead of coretypes.h.
26210         (build/gensupport.o, build/ggc-none.o, build/print-rtl.o): Likewise.
26211         (build/read-md.o, build/read-rtl.o, build/rtl.o): Likewise.
26212         (build/vec.o, build/hash-table.o, build/inchash.o): Likewise.
26213         (build/gencondmd.o, build/genattr.o, build/genattr-common.o): Likewise.
26214         (build/genattrtab.o, build/genautomata.o, build/gencheck.o): Likewise.
26215         (build/gencodes.o, build/genconditions.o): Likewise.
26216         (build/genconfig.o, build/genconstants.o, build/genemit.o): Likewise.
26217         (build/genenums.o, build/genextract.o, build/genflags.o): Likewise.
26218         (build/gentarget-def.o, build/genmddeps.o, build/genopinit.o)
26219         (build/genoutput.o, build/genpeep.o, build/genpreds.o): Likewise.
26220         (build/genrecog.o, build/genmddump.o, build/genmatch.o): Likewise.
26221         (build/gencfn-macros.o, build/gcov-iov.o): Likewise.
26222         * coretypes.h: Include everything up to real.h for generators.
26223         Include insn-modes.h first.  Include wide-int-print.h after
26224         wide-int.h.  Include insn-modes-inline.h and then machmode.h.
26225         * machmode.h: Don't include insn-modes.h here.
26226         * function-tests.c: Remove includes of signop.h, machmode.h,
26227         double-int.h and wide-int.h.
26228         * rtl.h: Likewise.
26229         * gcc-rich-location.c: Remove includes of machmode.h, double-int.h
26230         and wide-int.h.
26231         * optc-save-gen.awk: Likewise.
26232         * gencheck.c (BITS_PER_UNIT): Delete dummy definition.
26233         * godump.c: Remove include of wide-int-print.h.
26234         * pretty-print.h: Likewise.
26235         * wide-int-print.cc: Likewise.
26236         * wide-int.cc: Likewise.
26237         * hash-map-tests.c: Remove include of signop.h.
26238         * hash-set-tests.c: Likewise.
26239         * rtl-tests.c: Likewise.
26240         * mkconfig.sh: Remove include of machmode.h.
26241         * genmodes.c (emit_insn_modes_h): Split emission of inline functions
26242         into...
26243         (emit_insn_modes_inline_h): ...this new function.  Emit the code
26244         into an insn-modes-inline.h header file, adding appropriate
26245         include guards and end comments.
26246         (emit_insn_modes_c_header): Remove include of machmode.h.
26247         (emit_min_insn_modes_c_header): Include coretypes.h rather than
26248         machmode.h.
26249         (main): Handle -i flag and call emit_insn_modes_inline_h when
26250         it is passed.
26252 2017-07-02  Richard Sandiford  <richard.sandiford@linaro.org>
26254         * tree-ssa-strlen.c (strinfo): Rename the length field to
26255         nonzero_chars.  Add a full_string_p field.
26256         (compare_nonzero_chars, zero_length_string_p): New functions.
26257         (get_addr_stridx): Add an offset_out parameter.
26258         Use compare_nonzero_chars.
26259         (get_stridx): Update accordingly.  Use compare_nonzero_chars.
26260         (new_strinfo): Update after above changes to strinfo.
26261         (set_endptr_and_length): Set full_string_p.
26262         (get_string_length): Update after above changes to strinfo.
26263         (unshare_strinfo): Update call to new_strinfo.
26264         (maybe_invalidate): Likewise.
26265         (get_stridx_plus_constant): Change off to unsigned HOST_WIDE_INT.
26266         Use compare_nonzero_chars and zero_string_p.  Treat nonzero_chars
26267         as a uhwi instead of an shwi.  Update after above changes to
26268         strinfo and new_strinfo.
26269         (zero_length_string): Assert that chainsi contains full strings.
26270         Use zero_length_string_p.  Update call to new_strinfo.
26271         (adjust_related_strinfos): Update after above changes to strinfo.
26272         Copy full_string_p from origsi.
26273         (adjust_last_stmt): Use zero_length_string_p.
26274         (handle_builtin_strlen): Update after above changes to strinfo and
26275         new_strinfo.  Install the lhs as the string length if the previous
26276         entry didn't describe a full string.
26277         (handle_builtin_strchr): Update after above changes to strinfo
26278         and new_strinfo.
26279         (handle_builtin_strcpy): Likewise.
26280         (handle_builtin_strcat): Likewise.
26281         (handle_builtin_malloc): Likewise.
26282         (handle_pointer_plus): Likewise.
26283         (handle_builtin_memcpy): Likewise.  Track nonzero characters
26284         that aren't necessarily followed by a nul terminator.
26285         (handle_char_store): Likewise.
26287 2017-07-02  Richard Sandiford  <richard.sandiford@linaro.org>
26289         PR tree-optimization/80769
26290         * tree-ssa-strlen.c (strinfo): Document that "stmt" is also used
26291         for malloc and calloc.  Document the new invariant that all related
26292         strinfos have delayed lengths or none do.
26293         (verify_related_strinfos): Move earlier in file.
26294         (set_endptr_and_length): New function, split out from...
26295         (get_string_length): ...here.  Also set the lengths of related
26296         strinfos.
26297         (zero_length_string): Assert that chainsi has known (rather than
26298         delayed) lengths.
26299         (adjust_related_strinfos): Likewise.
26301 2017-07-02  Richard Sandiford  <richard.sandiford@linaro.org>
26303         PR tree-optimization/81136
26304         * tree-vect-data-refs.c (vect_update_misalignment_for_peel): Only
26305         assert that two references with the same misalignment have the same
26306         compile-time misalignment if those compile-time misalignments
26307         are known.
26309 2017-07-01  Andi Kleen  <ak@linux.intel.com>
26311         * print-tree.c (print_node): Print all attributes.
26313 2017-07-01  Jan Hubicka  <hubicka@ucw.cz>
26315         * cfg.c (scale_bbs_frequencies): New function.
26316         * cfg.h (scale_bbs_frequencies): Declare it.
26317         * cfgloopanal.c (single_likely_exit): Cleanup.
26318         * cfgloopmanip.c (scale_loop_frequencies): Take profile_probability
26319         as parameter.
26320         (scale_loop_profile): Likewise.
26321         (loop_version): Likewise.
26322         (create_empty_loop_on_edge): Update.
26323         * cfgloopmanip.h (scale_loop_frequencies, scale_loop_profile,
26324         scale_loop_frequencies, scale_loop_profile, loopify,
26325         loop_version): Update prototypes.
26326         * modulo-sched.c (sms_schedule): Update.
26327         * predict.c (unlikely_executed_edge_p): Also check probability.
26328         (probably_never_executed_edge_p): Fix typo.
26329         * tree-if-conv.c (version_loop_for_if_conversion): Update.
26330         * tree-parloops.c (gen_parallel_loop): Update.
26331         * tree-ssa-loop-ivcanon.c (try_peel_loop): Update.
26332         * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Update.
26333         * tree-ssa-loop-split.c (split_loop): Update.
26334         * tree-ssa-loop-unswitch.c (tree_unswitch_loop): Update.
26335         * tree-vect-loop-manip.c (vect_do_peeling): Update.
26336         (vect_loop_versioning): Update.
26337         * tree-vect-loop.c (scale_profile_for_vect_loop): Update.
26339 2017-07-01  Jan Hubicka  <hubicka@ucw.cz>
26341         * trans-mem.c (split_bb_make_tm_edge): Update profile.
26343 2017-07-01  Jan Hubicka  <hubicka@ucw.cz>
26345         * tree-if-conv.c (combine_blocks): Use make_single_succ_edge
26346         to keep profile consistent.
26348 2017-07-01  Jan Hubicka  <hubicka@ucw.cz>
26350         * cfgrtl.c (rtl_flow_call_edges_add): Update profile.
26351         * tree-cfg.c (gimple_flow_call_edges_add): Likewise.
26352         * profile-count.h (max_safe_multiplier): Make unsigned.
26353         (profile_count::guessed_zero): New.
26355 2017-07-01  Jan Hubicka  <hubicka@ucw.cz>
26357         * bb-reorder.c (fix_up_crossing_landing_pad,
26358         fix_crossing_conditional_branches): Use make_single_succ_edge
26359         to keep profile consistent.
26361 2017-07-01  Jan Hubicka  <hubicka@ucw.cz>
26363         * tree-vect-loop.c (optimize_mask_stores): Use make_single_succ_edge
26364         to update profile.
26366 2017-07-01  Jakub Jelinek  <jakub@redhat.com>
26368         PR sanitizer/81262
26369         * bb-reorder.c (fix_up_fall_thru_edges): Move variable declarations to
26370         the right scopes, make sure cond_jump isn't preserved between multiple
26371         iterations.  Search for fallthru edge whenever there are 3+ edges and
26372         use find_fallthru_edge for it.
26374 2017-06-29  Jan Hubicka  <hubicka@ucw.cz>
26376         Patch by Alexander Monakov <amonakov@ispras.ru>
26377         * sel-sched-ir.c (compute_succs_info): Handle uninitialized
26378         probabilities consistently.
26380 2017-06-29  Jan Hubicka  <hubicka@ucw.cz>
26382         * pa.c (pa_expand_compare_and_swap_loop): Update call of
26383         emit_cmp_and_jump_insns.
26385 2017-06-29  Jan Hubicka  <hubicka@ucw.cz>
26387         PR ipa/81261
26388         * tree-inline.c (expand_call_inline): Combine profile statuses.
26390 2017-06-30  Andrew Pinski  <apinski@cavium.com>
26392         * tree-if-conv.c (predicate_scalar_phi): Update new_stmt if
26393         fold_stmt returned true.
26395 2017-06-30  Nathan Sidwell  <nathan@acm.org>
26397         * ggc.h (empty_string): Delete.
26398         * cfgexpand.c (expand_asm_stmt): Use plain "".
26399         * optabs.c (expand_asm_memory_barrier): Likewise.
26400         * stringpool.c (empty_string): Delete.
26401         (digit_vector, digit_string): Delete.
26402         (ggc_alloc_string): Use plain "", don't optimize single digit
26403         strings.  Use ggc_alloc_atomic.
26405 2017-06-30  Richard Earnshaw  <rearnsha@arm.com>
26407         * rtlanal.c (insn_rtx_cost): If a parallel contains exactly one
26408         comparison set and one other set, use the cost of the non-comparison
26409         set.
26411 2017-06-30  Nathan Sidwell  <nathan@acm.org>
26413         * ggc.h: Replace all 'static inline' with plain 'inline'.  Fix
26414         some formatting.
26416 2017-06-30  Peter Bergner  <bergner@vnet.ibm.com>
26418         * tree-cfg.c (group_case_labels_stmt): Merge scanning and compressing
26419         loops.  Remove now unneeded calls to gimple_switch_set_label() that
26420         just set removed labels to NULL_TREE.
26422 2017-06-30  Aldy Hernandez  <aldyh@redhat.com>
26424         * tree-ssanames.c (set_range_info_raw): Abstract from ...
26425         (set_range_info): ...here.  Only call set_range_info_raw if domain
26426         is useful.
26427         (set_nonzero_bits): Call set_range_info_raw.
26428         * tree-ssanames.h (set_range_info_raw): New.
26430 2017-06-30  Jakub Jelinek  <jakub@redhat.com>
26432         PR target/81225
26433         * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): For V8FI,
26434         V16FI and VI8F_256 iterators, use <store_mask_predicate> instead
26435         of nonimmediate_operand and <store_mask_constraint> instead of m
26436         for the input operand.  For V8FI iterator, always split if input
26437         is a MEM.  For V16FI and V8SF_256 iterators, don't test if both
26438         operands are MEM if <mask_applied>.  For VI4F_256 iterator, use
26439         <store_mask_predicate> instead of register_operand and
26440         <store_mask_constraint> instead of v for the input operand.  Make
26441         sure both operands aren't MEMs for if not <mask_applied>.
26443 2017-06-30  Sylvestre Ledru  <sylvestre@debian.org>
26445         * lto-wrapper.c (copy_file) Close both file descriptors before
26446         exiting normally.
26448 2017-06-30  Martin Liska  <mliska@suse.cz>
26450         PR ipa/81214
26451         * multiple_target.c (create_dispatcher_calls): Make ifunc
26452         also for function that don't have calls or are not referenced.
26454 2017-06-30  Richard Biener  <rguenther@suse.de>
26456         * tree-vect-slp.c (vect_slp_analyze_node_operations): Only
26457         analyze the first scalar stmt.  Move vector type computation
26458         for the BB case here from ...
26459         * tree-vect-stmts.c (vect_analyze_stmt): ... here.  Guard
26460         live operation processing in the SLP case properly.
26462 2017-06-30  Richard Biener  <rguenther@suse.de>
26464         * graph.c (draw_cfg_node_succ_edges): Fix broken dot syntax.
26466 2017-06-30  Martin Liska  <mliska@suse.cz>
26468         PR sanitizer/81021
26469         * tree-eh.c (lower_resx): Call BUILT_IN_ASAN_HANDLE_NO_RETURN
26470         before BUILT_IN_UNWIND_RESUME when ASAN is used.
26472 2017-06-30  Yvan Roux  <yvan.roux@linaro.org>
26474         * doc/invoke.texi (AArch64): Add missing options and remove redundant
26475         ones.
26477 2017-06-30  Richard Biener  <rguenther@suse.de>
26479         PR tree-optimization/81249
26480         * tree-vect-loop.c (vect_create_epilog_for_reduction): Convert
26481         condition reduction result to original scalar type.
26483 2017-06-30  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
26485         * profile-count.h (enum profile_quality): Fix typos and whitespace
26486         issues.
26488 2017-06-30  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
26490         * config/s390/s390.c (s390_expand_setmem): Adjust to the new data
26491         type for branch probabilities.
26493 2017-06-29  Julian Brown  <julian@codesourcery.com>
26494             Naveen H.S  <Naveen.Hurugalawadi@cavium.com>
26496         * config/aarch64/aarch64-fusion-pairs.def: Add ALU_BRANCH entry.
26497         * config/aarch64/aarch64.c (AARCH64_FUSE_ALU_BRANCH): New fusion type.
26498         (thunderx2t99_tunings): Set AARCH64_FUSE_ALU_BRANCH flag.
26499         (aarch_macro_fusion_pair_p): Add support for AARCH64_FUSE_ALU_BRANCH.
26501 2017-06-29  Naveen H.S  <Naveen.Hurugalawadi@cavium.com>
26503         * config/aarch64/aarch64.c (aarch_macro_fusion_pair_p): Push the
26504         check for CC usage into AARCH64_FUSE_CMP_BRANCH.
26505         * config/i386/i386.c (ix86_macro_fusion_pair_p): Push the check for
26506         CC usage from generic code to here.
26507         * sched-deps.c (sched_macro_fuse_insns): Move the condition for
26508         CC usage into the target macros.
26510 2017-06-29  Maya Rashish  <coypu@sdf.org>
26512         * config/netbsd.h (NETBSD_LIB_SPEC): Add -lc when creating shared
26513         objects.
26515 2017-06-29  Jan Hubicka  <hubicka@ucw.cz>
26517         * arm/arm-builtins.c: Include profile-count.h
26518         * except.c (sjlj_emit_function_enter): Use
26519         profile_probability::unlikely.
26521 2017-06-29  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
26523         * config/rs6000/rs6000.c (toc_relative_expr_p): Make tocrel_base
26524         and tocrel_offset be pointer args rather than implicitly using
26525         static versions.
26526         (legitimate_constant_pool_address_p, rs6000_emit_move,
26527         const_load_sequence_p, adjust_vperm): Add local tocrel_base and
26528         tocrel_offset and use in toc_relative_expr_p call.
26529         (print_operand, print_operand_address): Use static tocrel_base_oac
26530         and tocrel_offset_oac.
26531         (rs6000_output_addr_const_extra): Use static tocrel_base_oac and
26532         tocrel_offset_oac.
26534 2017-06-29  Maya Rashish  <coypu@sdf.org>
26536         * config/vax/builtins.md (ffssi2_internal): Correct constraint.
26538 2017-06-29  Eric Botcazou  <ebotcazou@adacore.com>
26540         * expr.c (expand_expr) <normal_inner_ref>: When testing for unaligned
26541         objects, take into account only the alignment of 'op0' and 'mode1' if
26542         'op0' is a MEM.
26544 2017-06-29  Steve Ellcey  <sellcey@cavium.com>
26546         * ccmp.c (ccmp_tree_comparison_p): New function.
26547         (ccmp_candidate_p): Update to use above function.
26548         (get_compare_parts): New function.
26549         (expand_ccmp_next): Update to use new functions.
26550         (expand_ccmp_expr_1): Take tree arg instead of gimple, update to use
26551         new functions.
26552         (expand_ccmp_expr): Pass tree instead of gimple to expand_ccmp_expr_1,
26553         take mode as argument.
26554         * ccmp.h (expand_ccmp_expr): Add mode as argument.
26555         * expr.c (expand_expr_real_1): Pass mode as argument.
26557 2017-06-29  Segher Boessenkool  <segher@kernel.crashing.org>
26559         * combine.c (combine_instructions): Print insns to dump_file, together
26560         with their costs.
26562 2017-06-29  Jan Hubicka  <hubicka@ucw.cz>
26564         * asan.c (asan_emit_stack_protection): Update.
26565         (create_cond_insert_point): Update.
26566         * auto-profile.c (afdo_propagate_circuit): Update.
26567         * basic-block.h (struct edge_def): Turn probability to
26568         profile_probability.
26569         (EDGE_FREQUENCY): Update.
26570         * bb-reorder.c (find_traces_1_round): Update.
26571         (better_edge_p): Update.
26572         (sanitize_hot_paths): Update.
26573         * cfg.c (unchecked_make_edge): Initialize probability to uninitialized.
26574         (make_single_succ_edge): Update.
26575         (check_bb_profile): Update.
26576         (dump_edge_info): Update.
26577         (update_bb_profile_for_threading): Update.
26578         * cfganal.c (connect_infinite_loops_to_exit): Initialize new edge
26579         probabilitycount to 0.
26580         * cfgbuild.c (compute_outgoing_frequencies): Update.
26581         * cfgcleanup.c (try_forward_edges): Update.
26582         (outgoing_edges_match): Update.
26583         (try_crossjump_to_edge): Update.
26584         * cfgexpand.c (expand_gimple_cond): Update make_single_succ_edge.
26585         (expand_gimple_tailcall): Update.
26586         (construct_init_block): Use make_single_succ_edge.
26587         (construct_exit_block): Use make_single_succ_edge.
26588         * cfghooks.c (verify_flow_info): Update.
26589         (redirect_edge_succ_nodup): Update.
26590         (split_edge): Update.
26591         (account_profile_record): Update.
26592         * cfgloopanal.c (single_likely_exit): Update.
26593         * cfgloopmanip.c (scale_loop_profile): Update.
26594         (set_zero_probability): Remove.
26595         (duplicate_loop_to_header_edge): Update.
26596         * cfgloopmanip.h (loop_version): Update prototype.
26597         * cfgrtl.c (try_redirect_by_replacing_jump): Update.
26598         (force_nonfallthru_and_redirect): Update.
26599         (update_br_prob_note): Update.
26600         (rtl_verify_edges): Update.
26601         (purge_dead_edges): Update.
26602         (rtl_lv_add_condition_to_bb): Update.
26603         * cgraph.c: (cgraph_edge::redirect_call_stmt_to_calle): Update.
26604         * cgraphunit.c (init_lowered_empty_function): Update.
26605         (cgraph_node::expand_thunk): Update.
26606         * cilk-common.c: Include profile-count.h
26607         * dojump.c (inv): Remove.
26608         (jumpifnot): Update.
26609         (jumpifnot_1): Update.
26610         (do_jump_1): Update.
26611         (do_jump): Update.
26612         (do_jump_by_parts_greater_rtx): Update.
26613         (do_compare_rtx_and_jump): Update.
26614         * dojump.h (jumpifnot, jumpifnot_1, jumpif_1, jumpif, do_jump,
26615         do_jump_1. do_compare_rtx_and_jump): Update prototype.
26616         * dwarf2cfi.c: Include profile-count.h
26617         * except.c (dw2_build_landing_pads): Use make_single_succ_edge.
26618         (sjlj_emit_dispatch_table): Likewise.
26619         * explow.c: Include profile-count.h
26620         * expmed.c (emit_store_flag_force): Update.
26621         (do_cmp_and_jump): Update.
26622         * expr.c (compare_by_pieces_d::generate): Update.
26623         (compare_by_pieces_d::finish_mode): Update.
26624         (emit_block_move_via_loop): Update.
26625         (store_expr_with_bounds): Update.
26626         (store_constructor): Update.
26627         (expand_expr_real_2): Update.
26628         (expand_expr_real_1): Update.
26629         * expr.h (try_casesi, try_tablejump): Update prototypes.
26630         * gimple-pretty-print.c (dump_probability): Update.
26631         (dump_profile): New.
26632         (dump_gimple_label): Update.
26633         (dump_gimple_bb_header): Update.
26634         * graph.c (draw_cfg_node_succ_edges): Update.
26635         * hsa-gen.c (convert_switch_statements): Update.
26636         * ifcvt.c (cheap_bb_rtx_cost_p): Update.
26637         (find_if_case_1): Update.
26638         (find_if_case_2): Update.
26639         * internal-fn.c (expand_arith_overflow_result_store): Update.
26640         (expand_addsub_overflow): Update.
26641         (expand_neg_overflow): Update.
26642         (expand_mul_overflow): Update.
26643         (expand_vector_ubsan_overflow): Update.
26644         * ipa-cp.c (good_cloning_opportunity_p): Update.
26645         * ipa-split.c (split_function): Use make_single_succ_edge.
26646         * ipa-utils.c (ipa_merge_profiles): Update.
26647         * loop-doloop.c (add_test): Update.
26648         (doloop_modify): Update.
26649         * loop-unroll.c (compare_and_jump_seq): Update.
26650         (unroll_loop_runtime_iterations): Update.
26651         * lra-constraints.c (lra_inheritance): Update.
26652         * lto-streamer-in.c (input_cfg): Update.
26653         * lto-streamer-out.c (output_cfg): Update.
26654         * mcf.c (adjust_cfg_counts): Update.
26655         * modulo-sched.c (sms_schedule): Update.
26656         * omp-expand.c (expand_omp_for_init_counts): Update.
26657         (extract_omp_for_update_vars): Update.
26658         (expand_omp_ordered_sink): Update.
26659         (expand_omp_for_ordered_loops): Update.
26660         (expand_omp_for_generic): Update.
26661         (expand_omp_for_static_nochunk): Update.
26662         (expand_omp_for_static_chunk): Update.
26663         (expand_cilk_for): Update.
26664         (expand_omp_simd): Update.
26665         (expand_omp_taskloop_for_outer): Update.
26666         (expand_omp_taskloop_for_inner): Update.
26667         * omp-simd-clone.c (simd_clone_adjust): Update.
26668         * optabs.c (expand_doubleword_shift): Update.
26669         (expand_abs): Update.
26670         (emit_cmp_and_jump_insn_1): Update.
26671         (expand_compare_and_swap_loop): Update.
26672         * optabs.h (emit_cmp_and_jump_insns): Update prototype.
26673         * predict.c (predictable_edge_p): Update.
26674         (edge_probability_reliable_p): Update.
26675         (set_even_probabilities): Update.
26676         (combine_predictions_for_insn): Update.
26677         (combine_predictions_for_bb): Update.
26678         (propagate_freq): Update.
26679         (estimate_bb_frequencies): Update.
26680         (force_edge_cold): Update.
26681         * profile-count.c (profile_count::dump): Add missing space into dump.
26682         (profile_count::debug): Add newline.
26683         (profile_count::differs_from_p): Explicitly convert to unsigned.
26684         (profile_count::stream_in): Update.
26685         (profile_probability::dump): New member function.
26686         (profile_probability::debug): New member function.
26687         (profile_probability::differs_from_p): New member function.
26688         (profile_probability::differs_lot_from_p): New member function.
26689         (profile_probability::stream_in): New member function.
26690         (profile_probability::stream_out): New member function.
26691         * profile-count.h (profile_count_quality): Rename to ...
26692         (profile_quality): ... this one.
26693         (profile_probability): New.
26694         (profile_count): Update.
26695         * profile.c (compute_branch_probabilities): Update.
26696         * recog.c (peep2_attempt): Update.
26697         * sched-ebb.c (schedule_ebbs): Update.
26698         * sched-rgn.c (find_single_block_region): Update.
26699         (compute_dom_prob_ps): Update.
26700         (schedule_region): Update.
26701         * sel-sched-ir.c (compute_succs_info): Update.
26702         * stmt.c (struct case_node): Update.
26703         (do_jump_if_equal): Update.
26704         (get_outgoing_edge_probs): Update.
26705         (conditional_probability): Update.
26706         (emit_case_dispatch_table): Update.
26707         (expand_case): Update.
26708         (expand_sjlj_dispatch_table): Update.
26709         (emit_case_nodes): Update.
26710         * targhooks.c: Update.
26711         * tracer.c (better_p): Update.
26712         (find_best_successor): Update.
26713         * trans-mem.c (expand_transaction): Update.
26714         * tree-call-cdce.c: Update.
26715         * tree-cfg.c (gimple_split_edge): Upate.
26716         (move_sese_region_to_fn): Upate.
26717         * tree-cfgcleanup.c (cleanup_control_expr_graph): Upate.
26718         * tree-eh.c (lower_resx): Upate.
26719         (cleanup_empty_eh_move_lp): Upate.
26720         * tree-if-conv.c (version_loop_for_if_conversion): Update.
26721         * tree-inline.c (copy_edges_for_bb): Update.
26722         (copy_cfg_body): Update.
26723         * tree-parloops.c (gen_parallel_loop): Update.
26724         * tree-profile.c (gimple_gen_ic_func_profiler): Update.
26725         (gimple_gen_time_profiler): Update.
26726         * tree-ssa-dce.c (remove_dead_stmt): Update.
26727         * tree-ssa-ifcombine.c (update_profile_after_ifcombine): Update.
26728         * tree-ssa-loop-im.c (execute_sm_if_changed): Update.
26729         * tree-ssa-loop-ivcanon.c (remove_exits_and_undefined_stmts): Update.
26730         (unloop_loops): Update.
26731         (try_peel_loop): Update.
26732         * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Update.
26733         * tree-ssa-loop-split.c (connect_loops): Update.
26734         (split_loop): Update.
26735         * tree-ssa-loop-unswitch.c (tree_unswitch_loop): Update.
26736         (hoist_guard): Update.
26737         * tree-ssa-phionlycprop.c (propagate_rhs_into_lhs): Update.
26738         * tree-ssa-phiopt.c (replace_phi_edge_with_variable): Update.
26739         (value_replacement): Update.
26740         * tree-ssa-reassoc.c (branch_fixup): Update.
26741         * tree-ssa-tail-merge.c (replace_block_by): Update.
26742         * tree-ssa-threadupdate.c (remove_ctrl_stmt_and_useless_edges): Update.
26743         (create_edge_and_update_destination_phis): Update.
26744         (compute_path_counts): Update.
26745         (recompute_probabilities): Update.
26746         (update_joiner_offpath_counts): Update.
26747         (freqs_to_counts_path): Update.
26748         (duplicate_thread_path): Update.
26749         * tree-switch-conversion.c (hoist_edge_and_branch_if_true): Update.
26750         (struct switch_conv_info): Update.
26751         (gen_inbound_check): Update.
26752         * tree-vect-loop-manip.c (slpeel_add_loop_guard): Update.
26753         (vect_do_peeling): Update.
26754         (vect_loop_versioning): Update.
26755         * tree-vect-loop.c (scale_profile_for_vect_loop): Update.
26756         (optimize_mask_stores): Update.
26757         * ubsan.c (ubsan_expand_null_ifn): Update.
26758         * value-prof.c (gimple_divmod_fixed_value): Update.
26759         (gimple_divmod_fixed_value_transform): Update.
26760         (gimple_mod_pow2): Update.
26761         (gimple_mod_pow2_value_transform): Update.
26762         (gimple_mod_subtract): Update.
26763         (gimple_mod_subtract_transform): Update.
26764         (gimple_ic): Update.
26765         (gimple_stringop_fixed_value): Update.
26766         (gimple_stringops_transform): Update.
26767         * value-prof.h: Update.
26769 2017-06-29  Carl Love  <cel@us.ibm.com>
26771         * config/rs6000/rs6000-c.c: Add support for built-in functions
26772         vector signed int vec_signed (vector float);
26773         vector signed long long vec_signed (vector double);
26774         vector signed int vec_signed2 (vector double, vector double);
26775         vector signed int vec_signede (vector double);
26776         vector signed int vec_signedo (vector double);
26777         * config/rs6000/rs6000.c (rs6000_generate_vsigned2_code): Add
26778         instruction generator.
26779         * config/rs6000/vsx.md (UNSPEC_VSX_XVCVSPSXWS, UNSPEC_VSX_XVCVSPSXDS,
26780         UNSPEC_VSX_VSIGNED2): Add UNSPECS.
26781         (vsx_xvcvspsxws, vsx_xvcvdpuxds_scale, vsx_xvcvspuxws, vsigned2_v2df):
26782         Add define_insn.
26783         (vsignedo_v2df, vsignede_v2df, vunsigned2_v2df, vunsignedo_v2df,
26784         vunsignede_v2df): Add define_expands.
26785         * config/rs6000/rs6000-builtin.def (VEC_SIGNED, VEC_UNSIGNED,
26786         VEC_SIGNED2, VEC_UNSIGNED2, VEC_SIGNEDE, VEC_UNSIGNEDE, VEC_SIGNEDO,
26787         VEC_UNSIGNEDO): Add definitions.
26788         * config/vsx.md (UNSPEC_VSX_XVCVSPSXWS, UNSPEC_VSX_XVCVSPSXDS,
26789         UNSPEC_VSX_VSIGNED2): Add UNSPECs.
26790         (vsx_xvcvspsxws, vsx_xvcvspuxws): Add define_insn.
26791         (vsigned2_v2df, vsigendo_v2df, vsignede_v2df,
26792         vunsigned2_v2df, vunsignedo_v2df, vunsignede_v2df): Add define_expands.
26793         * config/rs6000/altivec.h (vec_signed, vec_signed2,
26794         vec_signede and vec_signedo, vec_unsigned, vec_unsigned2,
26795         vec_unsignede, vec_unsignedo): Add builtin defines.
26796         * config/rs6000-protos.h (rs6000_generate_vsigned2_code): Add extern
26797         declaration.
26798         * doc/extend.texi: Update the built-in documentation file for the
26799         new built-in functions.
26801 2017-06-29  Richard Biener  <rguenther@suse.de>
26803         * tree-vect-loop.c (vect_analyze_scalar_cycles_1): Do not add
26804         reduction chains to LOOP_VINFO_REDUCTIONS.
26805         * tree-vect-slp.c (vect_analyze_slp): Continue looking for
26806         SLP reductions after processing reduction chains.
26808 2017-06-29  Nathan Sidwell  <nathan@acm.org>
26810         * builtins.c (fold_builtin_FUNCTION): Use
26811         lang_hooks.decl_printable_name.
26813 2017-06-29  Peter Bergner  <bergner@vnet.ibm.com>
26815         PR middle-end/81194
26816         * cfgexpand.c (expand_gimple_stmt_1): Handle switch statements
26817         with only one label.
26818         * stmt.c (expand_case): Assert NCASES is greater than one.
26820 2017-06-29  Richard Biener  <rguenther@suse.de>
26822         * tree-cfg.c (group_case_labels_stmt): Return whether we changed
26823         anything.
26824         (group_case_labels): Likewise.
26825         (find_taken_edge): Push sanity checking on val to workers...
26826         (find_taken_edge_cond_expr): ... here
26827         (find_taken_edge_switch_expr): ... and here, handle cases
26828         with just a default label.
26829         * tree-cfg.h (group_case_labels_stmt): Adjust prototype.
26830         (group_case_labels): Likewise.
26831         * tree-cfgcleanup.c (execute_cleanup_cfg_post_optimizing): When
26832         group_case_labels does anything cleanup the CFG again.
26834 2017-06-29  Bin Cheng  <bin.cheng@arm.com>
26836         PR tree-optimization/81196
26837         * tree-ssa-loop-niter.c (number_of_iterations_cond): Handle loop
26838         exit condition comparing two IVs.
26840 2017-06-29  Richard Earnshaw  <rearnsha@arm.com>
26842         * config/arm/parsecpu.awk (gen_comm_data): Add initializer for
26843         profile to the dummy entry at the end of the list of architectures.
26844         * config/arm/arm-cpu-cdata.h: Regenerated.
26846 2017-06-29  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
26847             Michael Collison <michael.collison@arm.com>
26849         PR target/70119
26850         * config/aarch64/aarch64.md (*aarch64_<optab>_reg_<mode>3_mask1):
26851         New pattern.
26852         (*aarch64_reg_<mode>3_neg_mask2): New pattern.
26853         (*aarch64_reg_<mode>3_minus_mask): New pattern.
26854         (*aarch64_<optab>_reg_di3_mask2): New pattern.
26855         * config/aarch64/aarch64.c (aarch64_rtx_costs): Account for cost
26856         of shift when the shift amount is masked with constant equal to
26857         the size of the mode.
26858         * config/aarch64/predicates.md (subreg_lowpart_operator): New
26859         predicate.
26861 2017-06-29  Martin Liska  <mliska@suse.cz>
26863         * config/i386/i386.opt: Change range from [1,5] to [0,5].
26865 2017-06-29  Yury Gribov  <tetra2005@gmail.com>
26867         PR bootstrap/80565
26868         * ipa-cp.c (allocate_and_init_ipcp_value): Add initialization
26869         code.
26870         * ipa-inline.h
26871         (edge_growth_cache_entry::edge_growth_cache_entry): New
26872         function.
26873         (reset_edge_growth_cache): Update to use constructor.
26875 2017-06-28  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
26877         * config/aarch64/aarch64.h (AARCH64_EXPAND_ALIGNMENT): New.
26878         (DATA_ALIGNMENT): Update to use AARCH64_EXPAND_ALIGNMENT.
26879         (LOCAL_ALIGNMENT): Update to use AARCH64_EXPAND_ALIGNMENT.
26881 2017-06-28  Sebastian Peryt  <sebastian.peryt@intel.com>
26883         * config/i386/avx512vlintrin.h (_mm256_permutexvar_epi64)
26884         (_mm256_permutexvar_epi32, _mm256_permutex_epi64): New intrinsics.
26886 2017-06-28  Szabolcs Nagy  <szabolcs.nagy@arm.com>
26888         * config.gcc (*-linux-musl*): Add t-musl tmake_file.
26889         (*-linux-uclibc*): Add t-uclibc tmake_file.
26890         * config/t-musl: New.
26891         * config/t-uclibc: New.
26893 2017-06-28  Richard Earnshaw  <rearnsha@arm.com>
26895         * config/arm/parsecpu.awk (profile): Parse new keyword in an arch
26896         context.
26897         (gen_comm_data): Emit architectural setting of arch_prof.
26898         * config/arm/arm-cpus.in (armv6-m, armv6s-m, armv7-a, armv7ve): Set the
26899         profile.
26900         (armv7-r, armv7-m, armv7e-m, armv8-a, armv8.1-a, armv8.2-a): Likewise.
26901         (armv8-m.base, armv8-m.main): Likewise.
26902         * arm-protos.h (arm_build_target): Add profile field.
26903         (arch_option): Likewise.
26904         * config/arm/arm.c (arm_configure_build_target): Copy the profile to
26905         the active target.
26906         * config/arm/arm.h (TARGET_ARM_ARCH_PROFILE): Use
26907         arm_active_target.profile.
26909 2017-06-28  Richard Biener  <rguenther@suse.de>
26911         PR middle-end/81227
26912         * fold-const.c (negate_expr_p): Use TYPE_UNSIGNED, not
26913         TYPE_OVERFLOW_WRAPS.
26914         * match.pd (negate_expr_p): Likewise.
26915         * tree-ssa-reassoc.c (optimize_range_tests_diff): Use
26916         fold_build2, not fold_binary.
26918 2017-06-28  Wilco Dijkstra  <wdijkstr@arm.com>
26920         * config/aarch64/aarch64 (aarch64_expand_mov_immediate):
26921         Convert memory address to Pmode.
26922         (aarch64_print_operand): Assert MEM operands are always Pmode.
26924 2017-06-28  Wilco Dijkstra  <wdijkstr@arm.com>
26926         PR target/79665
26927         * config/arm/aarch-common.c (arm_no_early_alu_shift_dep):
26928         Remove redundant if.
26929         (aarch_forward_to_shift_is_not_shifted_reg): Remove.
26930         * config/arm/aarch-common-protos.h
26931         (aarch_forward_to_shift_is_not_shifted_re): Remove.
26932         * config/arm/cortex-a53.md: Use arm_no_early_alu_shift_dep in bypass.
26934 2017-06-28  Michael Meissner  <meissner@linux.vnet.ibm.com>
26936         PR ipa/81238
26937         * multiple_target.c (create_dispatcher_calls): Set the default
26938         clone to be static, not public.
26940 2017-06-28  Richard Biener  <rguenther@suse.de>
26942         * tree-vect-loop.c (vectorizable_reduction): Move special
26943         cond reduction IV var creation ...
26944         (vect_create_epilog_for_reduction): ... here.  Remove induction_index
26945         parameter.  Use STMT_VINFO_VECTYPE.
26946         * tree-vect-slp.c (vect_get_constant_vectors): Properly reset
26947         constant_p.
26949 2017-06-28  Martin Liska  <mliska@suse.cz>
26951         PR ipa/81128
26952         * ipa-visibility.c (non_local_p): Handle visibility.
26954 2017-06-28  Martin Liska  <mliska@suse.cz>
26956         PR driver/79659
26957         * common.opt: Add IntegerRange to various options.
26958         * opt-functions.awk (integer_range_info): New function.
26959         * optc-gen.awk: Add integer_range_info to cl_options struct.
26960         * opts-common.c (decode_cmdline_option): Handle
26961         CL_ERR_INT_RANGE_ARG.
26962         (cmdline_handle_error): Likewise.
26963         * opts.c (print_filtered_help): Show valid interval in
26964         when --help is provided.
26965         * opts.h (struct cl_option): Add range_min and range_max fields.
26966         * config/i386/i386.opt: Add IntegerRange for -mbranch-cost.
26968 2017-06-28  Marc Glisse  <marc.glisse@inria.fr>
26970         * match.pd ((X & ~Y) | (~X & Y)): Generalize to + and ^.
26971         (x * C EQ/NE y * C): New transformation.
26973 2017-06-28  Christophe Lyon  <christophe.lyon@linaro.org>
26975         * genmultilib (combination_space): Accept '+' in option names.
26977 2017-06-28  Martin Liska  <mliska@suse.cz>
26979         PR sanitizer/81224
26980         * asan.c (instrument_derefs): Bail out inner references
26981         that are hard register variables.
26983 2017-06-28  Jakub Jelinek  <jakub@redhat.com>
26985         PR target/81175
26986         * config/i386/i386.c (ix86_init_mmx_sse_builtins): Use def_builtin
26987         rather than def_builtin_pure for __builtin_ia32_gatherpf*.
26989 2017-06-28  Richard Biener  <rguenther@suse.de>
26991         * tree-vectorizer.h (vect_get_vec_defs): Remove.
26992         (vect_get_slp_defs): Adjust.
26993         * tree-vect-loop.c (get_initial_defs_for_reduction): Split
26994         out from ...
26995         * tree-vect-slp.c (vect_get_constant_vectors): ... here and
26996         simplify.
26997         * tree-vect-loop.c (vect_create_epilog_for_reduction): Use
26998         get_initial_defs_for_reduction instead of vect_get_vec_defs.
26999         (vectorizable_reduction): Adjust.
27000         * tree-vect-slp.c (vect_get_constant_vectors): Remove reduction
27001         handling.
27002         (vect_get_slp_defs): Likewise.
27003         * tree-vect-stmts.c (vect_get_vec_defs): Make static and adjust.
27004         (vectorizable_bswap): Adjust.
27005         (vectorizable_call): Likewise.
27006         (vectorizable_conversion): Likewise.
27007         (vectorizable_assignment): Likewise.
27008         (vectorizable_shift): Likewise.
27009         (vectorizable_operation): Likewise.
27010         (vectorizable_store): Likewise.
27011         (vectorizable_condition): Likewise.
27012         (vectorizable_comparison): Likewise.
27014 2017-06-28  Michael Collison  <michael.collison@arm.com>
27016         PR target/68535
27017         * config/arm/arm.c (gen_ldm_seq): Remove last unnecessary
27018         set of base_reg
27019         (arm_gen_movmemqi): Removed unused variable 'i'.
27020         Convert 'for' loop into 'while' loop.
27021         (arm_expand_prologue): Remove last unnecessary set of insn.
27022         (thumb_pop): Remove unused variable 'pushed_words'.
27023         (thumb_exit): Remove last unnecessary set of regs_to_pop.
27025 2017-06-28  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
27027         * config/s390/predicates.md: Use s390_rel_address_ok_p.
27028         * config/s390/s390-protos.h: Add prototype of
27029         s390_rel_address_ok_p.
27030         * config/s390/s390.c (s390_got_symbol): New function.
27031         (s390_rel_address_ok_p): New function.
27032         (legitimize_pic_address): Use s390_rel_address_ok_p.
27033         (s390_load_got): Use s390_got_symbol.
27034         (s390_option_override): Issue error if
27035         -mno-pic-data-is-text-relative is used without -fpic/-fPIC.
27036         * config/s390/s390.h (TARGET_DEFAULT_PIC_DATA_IS_TEXT_RELATIVE):
27037         New macro.
27038         * config/s390/s390.opt: New option mpic-data-is-text-relative.
27040 2017-06-27  Andrew Pinski  <apinski@cavium.com>
27042         * match.pd (X >/>=/</<= 0 ? 1.0 : -1.0): New patterns.
27043         (X * copysign (1.0, X)): New pattern.
27044         (X * copysign (1.0, -X)): New pattern.
27045         (copysign (-1.0, CST)): New pattern.
27047 2017-06-27  Joseph Myers  <joseph@codesourcery.com>
27049         * genmultilib (combination_space): Remove variable.
27050         Validate reuse rules against regular expression for any sequence
27051         of multilib options in any order.
27053 2017-06-27  Michael Collison  <michael.collison@arm.com>
27055         * config/aarch64/aarch64-simd.md (aarch64_combine<mode>): Directly
27056         call aarch64_split_simd_combine.
27057         * (aarch64_combine_internal<mode>): Delete pattern.
27058         * config/aarch64/aarch64.c (aarch64_split_simd_combine):
27059         Allow register and subreg operands.
27061 2017-06-27  Jerome Lambourg  <lambourg@adacore.com>
27063         * config/i386/vxworks.h (ASM_SPEC): Remove definition. No target
27064         specific need, just fallback on defaults.
27065         (ASM_OUTPUT_ALIGNED_BSS): Add #undef before #define.
27067 2017-06-27  Jerome Lambourg  <lambourg@adacore.com>
27068             Olivier Hainque  <hainque@adacore.com>
27070         * config/i386/vxworks.h (DBX_REGISTER_NUMBER): Pick distinct
27071         map for 64bits.
27072         (TARGET_OS_CPP_BUILTINS): builtin_define CPU to X86_64 for 64bit
27073         targets. Pick a default if no particular attempt applied.
27074         (STACK_CHECK_PROTECT): Double for 64bit targets, which have
27075         larger contexts.
27077 2017-06-27  Jerome Lambourg  <lambourg@adacore.com>
27079         * config.gcc (i*86-wrs-vxworks7): Handle new acceptable triplet.
27080         (x86_64-wrs-vxworks7): Likewise.
27082 2017-06-27  Marek Polacek  <polacek@redhat.com>
27084         PR sanitizer/81223
27085         * ubsan.c (instrument_null): Check get_base_address's result for null.
27087 2017-06-27  Marc Glisse  <marc.glisse@inria.fr>
27089         * match.pd ((A+-B)+(C-A), (A+B)-(A-C)): New transformations.
27091 2017-06-27  Marc Glisse  <marc.glisse@inria.fr>
27093         * builtin-types.def (BT_FENV_T_PTR, BT_CONST_FENV_T_PTR,
27094         BT_FEXCEPT_T_PTR, BT_CONST_FEXCEPT_T_PTR): New primitive types.
27095         (BT_FN_INT_FENV_T_PTR, BT_FN_INT_CONST_FENV_T_PTR,
27096         BT_FN_INT_FEXCEPT_T_PTR_INT, BT_FN_INT_CONST_FEXCEPT_T_PTR_INT):
27097         New function types.
27098         * builtins.def (BUILT_IN_FECLEAREXCEPT, BUILT_IN_FEGETENV,
27099         BUILT_IN_FEGETEXCEPTFLAG, BUILT_IN_FEGETROUND,
27100         BUILT_IN_FEHOLDEXCEPT, BUILT_IN_FERAISEEXCEPT,
27101         BUILT_IN_FESETENV, BUILT_IN_FESETEXCEPTFLAG,
27102         BUILT_IN_FESETROUND, BUILT_IN_FETESTEXCEPT,
27103         BUILT_IN_FEUPDATEENV): New builtins.
27104         * tree-core.h (TI_FENV_T_PTR_TYPE, TI_CONST_FENV_T_PTR_TYPE,
27105         TI_FEXCEPT_T_PTR_TYPE, TI_CONST_FEXCEPT_T_PTR_TYPE): New entries.
27106         * tree.h (fenv_t_ptr_type_node, const_fenv_t_ptr_type_node,
27107         fexcept_t_ptr_type_node, const_fexcept_t_ptr_type_node): New
27108         macros.
27109         (builtin_structptr_types): Adjust size.
27110         * tree.c (builtin_structptr_types): Add four entries.
27112 2017-06-27  Jerome Lambourg  <lambourg@adacore.com>
27113             Olivier Hainque  <hainque@adacore.com>
27115         * config/vxworks.h (VXWORKS_LIB_SPEC): Incorporate ...
27116         (TLS_SYM): New local macro, forcing reference to __tls__ on
27117         link command lines for VxWorks 7 RTPs, triggering initialization
27118         of tlsLib.
27119         (VXWORKS_HAVE_TLS): New macro. State whether the target VxWorks
27120         OS features TLS support, true for RTPs on VxWorks 7.
27121         * config/vxworks.c (vxworks_override_options): Setup emutls
27122         accordingly.
27124 2017-06-27  Jakub Jelinek  <jakub@redhat.com>
27126         * predict.c (test_prediction_value_range): Use -1U instead of -1
27127         to avoid narrowing conversion warning.
27128         * dumpfile.c (dump_options): Wrap all value into dump_flags_t cast
27129         to avoid narrowing conversion warning.
27130         * opt-functions.awk (var_ref): Return (unsigned short) -1 instead of
27131         -1.
27132         * optc-gen.awk (END): Expect (unsigned short) -1 instead of -1.
27134 2017-06-27  Jerome Lambourg  <lambourg@adacore.com>
27136         * config/vxworks.h (VXWORKS_LIBS_RTP): Alternative definition for
27137         64bit configurations.
27138         (PTR_DIFF_TYPE): Alternative definition for TARGET_LP64.
27139         (SIZE_TYPE): Likewise.
27140         * config/vxworks.c (vxworks_emutls_var_fields): Use
27141         long_unsigned_type_node instead of unsigned_type_node as the offset
27142         field type, which is "pointer" mode in emutls.c.
27144 2017-06-27  Jakub Jelinek  <jakub@redhat.com>
27146         PR sanitizer/81209
27147         * ubsan.c (ubsan_encode_value): Initialize DECL_CONTEXT on var.
27149         PR middle-end/81207
27150         * gimple-fold.c (replace_call_with_call_and_fold): Handle
27151         gimple_vuse copying separately from gimple_vdef copying.
27153 2017-06-27  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
27155         * value-prof.c (free_hist): Remove call to memset and the enclosing if
27156         condition.
27158 2017-06-26  Jerome Lambourg  <lambourg@adacore.com>
27159             Olivier Hainque  <hainque@adacore.com>
27161         * config.gcc (*-*-vxworks*): Add TARGET_VXWORKS7=1 to tm_defines
27162         for all vxworks7 targets.
27163         * config/vxworks.h (TARGET_VXWORKS7): If not defined, define to 0.
27164         (VXWORKS_ADDITIONAL_CPP_SPEC): Alternative definition for VXWORKS7.
27165         (VXWORKS_LIBS_RTP, VXWORKS_LIBS_RTP_DIR): New macros, allowing
27166         variations for VX6/VX7 and 32/64bits later on in ...
27167         (VXWORKS_LIB_SPEC): Leverage new macros.
27168         (VXWORKS_OS_CPP_BUILTINS): Define _VSB_CONFIG_FILE for VXWORKS7,
27169         as well as _ALLOW_KEYWORD_MACROS when "inline" is not a keyword.
27171 2017-06-26  Jerome Lambourg  <lambourg@adacore.com>
27173         * config/vxworks.h (VXWORKS_OS_CPP_BUILTINS): builtin_define
27174         _VX_TOOL_FAMILY and _VX_TOOL to gnu.
27176 2017-06-26  Carl Love  <cel@us.ibm.com>
27178         * config/rs6000/rs6000-c.c: Add support for built-in functions
27179         vector bool char vec_reve (vector bool char);
27180         vector signed char vec_reve (vector signed char);
27181         vector unsigned char vec_reve (vector unsigned char);
27182         vector bool int vec_reve (vector bool int);
27183         vector signed int vec_reve (vector signed int);
27184         vector unsigned int vec_reve (vector unsigned int);
27185         vector bool long long vec_reve (vector bool long long);
27186         vector signed long long vec_reve (vector signed long long);
27187         vector unsigned long long vec_reve (vector unsigned long long);
27188         vector bool short vec_reve (vector bool short);
27189         vector signed short vec_reve (vector signed short);
27190         vector double vec_reve (vector double);
27191         vector float vec_reve (vector float);
27192         * config/rs6000/rs6000-builtin.def (VREVE_V2DI, VREVE_V4SI,
27193         VREVE_V8HI, VREVE_V16QI, VREVE_V2DF, VREVE_V4SF, VREVE): New builtin.
27194         * config/rs6000/altivec.md (UNSPEC_VREVEV): New UNSPEC.
27195         (altivec_vreve): New pattern.
27196         * config/rs6000/altivec.h (vec_reve): New define.
27197         * doc/extend.texi (vec_rev): Update the built-in documentation file
27198         for the new built-in functions.
27200 2016-06-26  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
27202         PR tree-optimization/71815
27203         * gimple-ssa-strength-reduction.c (uses_consumed_by_stmt): New
27204         function.
27205         (find_basis_for_candidate): Call uses_consumed_by_stmt rather than
27206         has_single_use.
27207         (slsr_process_phi): Likewise.
27208         (replace_uncond_cands_and_profitable_phis): Don't replace a
27209         multiply candidate with a stride of 1 (copy or cast).
27210         (phi_incr_cost): Call uses_consumed_by_stmt rather than
27211         has_single_use.
27212         (lowest_cost_path): Likewise.
27213         (total_savings): Likewise.
27215 2017-06-26  Richard Biener  <rguenther@suse.de>
27217         PR target/81175
27218         * config/i386/i386.c (ix86_init_mmx_sse_builtins):
27219         Use def_builtin_pure for all gather builtins.
27221 2017-06-26  Richard Biener  <rguenther@suse.de>
27223         PR tree-optimization/81203
27224         * tree-tailcall.c (find_tail_calls): Do not move stmts into
27225         non-dominating BBs.
27227 2017-06-26  Marek Polacek  <polacek@redhat.com>
27229         PR c/80116
27230         * doc/invoke.texi: Document -Wmultistatement-macros.
27232 2017-06-26  Christophe Lyon  <christophe.lyon@linaro.org>
27234         * doc/sourcebuild.texi (ARM-specific attributes): Document new
27235         arm_neon_ok_no_float_abi effective target.
27237 2017-06-26  Richard Biener  <rguenther@suse.de>
27239         PR tree-optimization/80928
27240         * cfghooks.c (duplicate_block): Do not copy BB_DUPLICATED flag.
27241         (copy_bbs): Set BB_DUPLICATED flag early.
27242         (execute_on_growing_pred): Do not execute for BB_DUPLICATED
27243         marked blocks.
27244         (execute_on_shrinking_pred): Likewise.
27245         * tree-ssa.c (ssa_redirect_edge): Do not look for PHI args in
27246         BB_DUPLICATED blocks.
27247         * tree-ssa-phionlycoprop.c (eliminate_degenerate_phis_1): Properly
27248         iterate over all PHIs considering removal of *gsi.
27250 2017-06-23  Jim Wilson  <jim.wilson@linaro.org>
27252         * doc/invoke.texi (AArch64 Options, -mtune): Re-add falkor and
27253         qdf24xx.
27255 2017-06-23  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
27257         * config/rs6000/rs6000-string.c: (expand_block_clear,
27258         do_load_for_compare, select_block_compare_mode,
27259         compute_current_alignment, expand_block_compare,
27260         expand_strncmp_align_check, expand_strn_compare,
27261         expand_block_move, rs6000_output_load_multiple)
27262         Move functions related to string/block move/compare
27263         to a separate file.
27264         * config/rs6000/rs6000.c: Move above functions to rs6000-string.c.
27265         * config/rs6000/rs6000-protos.h (rs6000_emit_dot_insn): Add prototype
27266         for this function which is now used in two files.
27267         * config/rs6000/t-rs6000: Add rule to compile rs6000-string.o.
27268         * config.gcc: Add rs6000-string.o to extra_objs for
27269         targets powerpc*-*-* and rs6000*-*-*.
27271 2017-06-23  Michael Meissner  <meissner@linux.vnet.ibm.com>
27273         PR target/80510
27274         * config/rs6000/rs6000.md (ALTIVEC_DFORM): Do not allow DImode in
27275         32-bit, since indexed is not valid for DImode.
27276         (mov<mode>_hardfloat32): Reorder ISA 2.07 load/stores before ISA
27277         3.0 d-form load/stores to be the same as mov<mode>_hardfloat64.
27278         (define_peephole2 for Altivec d-form load): Add 32-bit support.
27279         (define_peephole2 for Altivec d-form store): Likewise.
27281         PR ipa/81185
27282         * multiple_target.c (create_dispatcher_calls): Only create the
27283         dispatcher call if the function is the default clone of a
27284         versioned function.
27286 2017-06-23  Segher Boessenkool  <segher@kernel.crashing.org>
27288         PR middle-end/80902
27289         * builtins.c (expand_builtin_atomic_fetch_op): If emitting code after
27290         a call, force the call to not be a tail call.
27292 2017-06-23  Jeff Law  <law@redhat.com>
27294         * doc/contrib.texi: Add entry for Steven Pemberton's work on
27295         enquire.
27297 2017-06-23  Will Schmidt  <will_schmidt@vnet.ibm.com>
27299         * config/rs6000/rs6000.c: Add include of ssa-propagate.h for
27300         update_call_from_tree().  (rs6000_gimple_fold_builtin): Add
27301         handling for early expansion of vector shifts (sl,sr,sra,rl).
27302         (builtin_function_type): Add vector shift right instructions
27303         to the unsigned argument list.
27305 2017-06-23  Bernd Edlinger  <bernd.edlinger@hotmail.de>
27307         rtl-optimizatoin/79286
27308         * ira.c (update_equiv_regs): Revert to using may_trap_or_fault_p again.
27309         * rtlanal.c (rtx_addr_can_trap_p_1): SYMBOL_REF_FUNCTION_P can never
27310         trap.  PIC register plus a const unspec without offset can never trap.
27312 2017-06-23  Marc Glisse  <marc.glisse@inria.fr>
27314         * tree.h (builtin_structptr_type): New type.
27315         (builtin_structptr_types): Declare new array.
27316         * tree.c (builtin_structptr_types): New array.
27317         (free_lang_data, build_common_tree_nodes): Use it.
27319 2017-06-23  Jonathan Wakely  <jwakely@redhat.com>
27321         PR c++/81187
27322         * doc/invoke.texi (-Wnoexcept-type): Fix name of option, from
27323         -Wnoexcept.
27325 2017-06-22  Matt Turner  <mattst88@gmail.com>
27327         * config/i386/driver-i386.c (host_detect_local_cpu): Add Kaby
27328         Lake models to skylake case.  Assume skylake for unknown
27329         models with clflushopt.
27331 2017-06-22  Jeff Law  <law@redhat.com>
27333         * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Handle
27334         frame sizes that do not satisfy aarch64_uimm12_shift.
27336 2017-06-22  Jan Hubicka <hubicka@ucw.cz>
27338         * profile-count.h (apply_probability,
27339         apply_scale, probability_in): Fix checks for zero.
27341 2017-06-22  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
27343         * incpath.c (add_sysroot_to_chain): Allow for $SYSROOT prefix.
27344         * doc/cppdiropts.texi (-I @var{dir}): Document it.
27346 2016-06-22  Richard Biener  <rguenther@suse.de>
27348         * tree-vect-loop.c (vect_model_reduction_cost): Handle
27349         COND_REDUCTION and INTEGER_INDUC_COND_REDUCTION without
27350         REDUC_MAX_EXPR support.
27351         (vectorizable_reduction): Likewise.
27352         (vect_create_epilog_for_reduction): Likewise.
27354 2017-06-22  James Greenhalgh  <james.greenhalgh@arm.com>
27356         * match.pd (A / (1 << B) -> A >> B): New.
27357         * generic-match-head.c: Include optabs-tree.h.
27358         * gimple-match-head.c: Likewise.
27359         * optabs-tree.h (target_supports_op_p): New.
27360         * optabs-tree.c (target_supports_op_p): New.
27362 2017-06-22  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
27364         * configure.ac (gcc_cv_ld_static_dynamic): Also check stderr for
27365         $gcc_cv_ld --help output.
27366         (gcc_cv_ld_demangle): Likewise.
27367         (gcc_cv_ld_eh_frame_hdr): Likewise.
27368         (gcc_cv_ld_pie): Likewise.
27369         (gcc_cv_ld_as_needed): Likewise.  Prefer native forms unless $gnu_ld.
27370         (gcc_cv_ld_buildid): Likewise.
27371         (gcc_cv_ld_sysroot): Likewise.
27372         (ld_bndplt_support): Likewise.
27373         (ld_pushpopstate_support): Likewise.
27374         * configure: Regenerate.
27375         * config/sol2.h [!USE_GLD] (SYSROOT_SPEC): Define.
27377 2017-06-21  Jakub Jelinek  <jakub@redhat.com>
27379         PR target/81151
27380         * config/i386/sse.md (round<mode>2): Renumber match_dup and
27381         operands indexes to avoid gap between operands and match_dups.
27383 2017-06-21  Andrew Pinski  <apinski@cavium.com>
27385         * config/aarch64/aarch64-cost-tables.h (thunderx_extra_costs):
27386         Increment Arith_shift and Arith_shift_reg by 1.
27387         * config/aarch64/aarch64-tuning-flags.def (cheap_shift_extend):
27388         New tuning flag.
27389         * config/aarch64/aarch64.c (thunderx_tunings): Enable
27390         AARCH64_EXTRA_TUNE_CHEAP_SHIFT_EXTEND.
27391         (aarch64_strip_extend): Add new argument and test for it.
27392         (aarch64_cheap_mult_shift_p): New function.
27393         (aarch64_rtx_mult_cost): Call aarch64_cheap_mult_shift_p and don't
27394         add a cost if it is true.
27395         Update calls to aarch64_strip_extend.
27396         (aarch64_rtx_costs): Update calls to aarch64_strip_extend.
27398 2017-06-21  Andrew Pinski  <apinski@cavium.com>
27400         * config/aarch64/aarch64-cores.def (thunderxt88p1): Use thunderxt88
27401         tunings.
27402         (thunderxt88): Likewise.
27403         * config/aarch64/aarch64.c (thunderxt88_prefetch_tune): New variable.
27404         (thunderx_prefetch_tune): New variable.
27405         (thunderx2t99_prefetch_tune): Update for the correct values.
27406         (thunderxt88_tunings): New variable.
27407         (thunderx_tunings): Use thunderx_prefetch_tune instead of
27408         generic_prefetch_tune.
27409         (thunderx2t99_tunings): Use AUTOPREFETCHER_WEAK.
27411 2017-06-21  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
27413         * config/aarch64/atomics.md (aarch64_compare_and_swap<mode>_lse,
27414         SHORT): Relax operand 3 to aarch64_reg_or_zero and constraint to Z.
27415         (aarch64_compare_and_swap<mode>_lse, GPI): Likewise.
27416         (aarch64_atomic_cas<mode>, SHORT): Likewise for operand 2.
27417         (aarch64_atomic_cas<mode>, GPI): Likewise.
27419 2017-06-21  Martin Liska  <mliska@suse.cz>
27421         * gimplify.c (gimplify_label_expr): Insert GIMPLE_PREDICT
27422         statements on cold and hot labels.
27423         * predict.c (tree_estimate_probability_bb): Remove the
27424         prediction from this place.
27426 2017-06-21  Martin Liska  <mliska@suse.cz>
27428         PR tree-optimization/79489
27429         * gimplify.c (maybe_add_early_return_predict_stmt): New
27430         function.
27431         (gimplify_return_expr): Call the function.
27432         * predict.c (tree_estimate_probability_bb): Remove handling
27433         of early return.
27434         * predict.def: Update comment about early return predictor.
27435         * gimple-predict.h (is_gimple_predict): New function.
27436         * predict.def: Change default value of early return to 66.
27437         * tree-tailcall.c (find_tail_calls): Skip GIMPLE_PREDICT
27438         statements.
27439         * passes.def: Put pass_strip_predict_hints to the beginning of
27440         IPA passes.
27442 2017-06-21  Pierre-Marie de Rodat  <derodat@adacore.com>
27444         * dwarf2out.c (gen_decl_die): Remove the guard to skip file-scope
27445         FUNCTION_DECL declarations.
27446         (dwarf2out_early_global_decl): Remove the guard to skip FUNCTION_DECL
27447         declarations.
27448         (dwaf2out_decl): Likewise.
27449         * godump.c (go_early_global_decl): Skip call to the real debug hook
27450         for FUNCTION_DECL declarations.
27451         * passes.c (rest_of_decl_compilation): Skip call to the
27452         early_global_decl debug hook for FUNCTION_DECL declarations, unless
27453         -fdump-go-spec is passed.
27455 2017-06-21  Marc Glisse  <marc.glisse@inria.fr>
27457         * config/i386/i386.c (struct builtin_isa): New field pure_p.
27458         Reorder for compactness.
27459         (def_builtin, def_builtin2, ix86_add_new_builtins): Handle pure_p.
27460         (def_builtin_pure, def_builtin_pure2): New functions.
27461         (ix86_init_mmx_sse_builtins) [__builtin_ia32_stmxcsr]: Mark as pure.
27463 2017-06-21  Marc Glisse  <marc.glisse@inria.fr>
27465         * match.pd (nop_convert): New predicate.
27466         ((A +- CST1) +- CST2): Allow some NOP conversions.
27468 2017-06-21  Jakub Jelinek  <jakub@redhat.com>
27470         PR c++/81130
27471         * gimplify.c (omp_add_variable): Don't force GOVD_SEEN for types
27472         with ctors/dtors if GOVD_SHARED is set.
27474 2017-06-21  Wilco Dijkstra  <wdijkstr@arm.com>
27476         * config/aarch64/aarch64.md (movti_aarch64):
27477         Emit mov rather than orr.
27478         (movtf_aarch64): Likewise.
27479         * config/aarch64/aarch64-simd.md (aarch64_simd_mov):
27480         Emit mov rather than orr.
27482 2017-06-21  Wilco Dijkstra  <wdijkstr@arm.com>
27484         * config/aarch64/aarch64-simd.md (aarch64_simd_dup):
27485         Swap alternatives, make integer dup more expensive.
27487 2017-06-21  Wilco Dijkstra  <wdijkstr@arm.com>
27489         * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
27490         Return true for non-tls symbols.
27492 2017-06-21  James Greenhalgh  <james.greenhalgh@arm.com>
27494         * config/aarch64/aarch64-cores.def (cortex-a55): New.
27495         (cortex-a75): Likewise.
27496         (cortex-a75.cortex-a55): Likewise.
27497         * config/aarch64/aarch64-tune.md: Regenerate.
27498         * doc/invoke.texi (-mtune): Document new values for -mtune.
27500 2017-06-21  Tom de Vries  <tom@codesourcery.com>
27502         * doc/sourcebuild.texi (Add Options, Features for dg-add-options): Add
27503         stack_size feature.
27504         (Effective-Target Keywords, Other attributes): Suggest using
27505         dg-add-options stack_size feature to get stack limit in stack_size
27506         effective target documentation.
27508 2017-06-21  Julian Brown  <julian@codesourcery.com>
27509             Naveen H.S  <Naveen.Hurugalawadi@cavium.com>
27511         * config/aarch64/aarch64-simd.md (aarch64_crypto_pmulldi)
27512         (aarch64_crypto_pmullv2di): Change type attribute to crypto_pmull.
27513         * config/aarch64/thunderx2t99.md (thunderx2t99_pmull): New
27514         reservation.
27515         * config/arm/cortex-a53.md (cortex_a53_advsimd_type): Add crypto_pmull to
27516         attribute type list for neon_multiply.
27517         * config/arm/cortex-a57.md (cortex_a57_neon_type): Add crypto_pmull to
27518         attribute type list for neon_multiply.
27519         * config/arm/crypto.md (crypto_vmullp64): Change type to crypto_pmull.
27520         * config/arm/exynos-m1.md (exynos_m1_neon_type): Add crypto_pmull to
27521         attribute type list for neon_multiply.
27522         * config/arm/types.md (crypto_pmull): Add.
27523         * config/arm/xgene1.md (xgene1_neon_pmull): Add crypto_pmull to
27524         attribute type list.
27526 2017-06-20  Andreas Tobler  <andreast@gcc.gnu.org>
27528         * config.gcc (armv6*-*-freebsd*): Change the target_cpu_cname to
27529         arm1176jzf-s.
27531 2017-06-20  Jakub Jelinek  <jakub@redhat.com>
27533         * ira-costs.c (find_costs_and_classes): Initialize cost_classes later
27534         to make sure not to dereference a NULL cost_classes_ptr pointer.
27536 2017-06-20  Carl Love  <cel@us.ibm.com>
27538         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
27539         ALTIVEC_BUILTIN_VMULESW, ALTIVEC_BUILTIN_VMULEUW,
27540         ALTIVEC_BUILTIN_VMULOSW, ALTIVEC_BUILTIN_VMULOUW entries.
27541         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin,
27542         builtin_function_type): Add ALTIVEC_BUILTIN_* case statements.
27543         * config/rs6000/altivec.md (MVULEUW, VMULESW, VMULOUW,
27544         VMULOSW): New enum "unspec" values.
27545         (vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
27546         vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si,
27547         altivec_vmuleuw, altivec_vmulesw, altivec_vmulouw,
27548         altivec_vmulosw): New patterns.
27549         * config/rs6000/rs6000-builtin.def (VMLEUW, VMULESW, VMULOUW,
27550         VMULOSW): Add definitions.
27552 2017-06-20  Julia Koval  <julia.koval@intel.com>
27554         * config/i386/i386.c: Fix rounding expand for new pattern.
27555         * config/i386/subst.md: Fix pattern (parallel -> unspec).
27557 2017-06-20  James Greenhalgh  <james.greenhalgh@arm.com>
27559         * config/aarch64/aarch64-option-extensions.def (rcpc): New.
27560         * config/aarch64/aarch64.h (AARCH64_FL_RCPC): New.
27562 2017-06-20  James Greenhalgh  <james.greenhalgh@arm.com>
27564         * config/aarch64/aarch64-option-extensions.def (fp16): Fix expected
27565         feature string.
27567 2017-06-20  James Greenhalgh  <james.greenhalgh@arm.com>
27569         * config/aarch64/aarch64-cores.def: Rearrange to sort by
27570         architecture, then by implementer ID.
27571         * config/aarch64/aarch64-tune.md: Regenerate.
27573 2017-06-20  Richard Biener  <rguenther@suse.de>
27575         PR middle-end/81097
27576         * fold-const.c (split_tree): Fold to type before negating.
27578 2017-06-20  David Malcolm  <dmalcolm@redhat.com>
27580         * diagnostic-show-locus.c
27581         (selftest::test_fixit_deletion_affecting_newline): New function.
27582         (selftest::diagnostic_show_locus_c_tests): Call it.
27584 2017-06-20  Andreas Schwab  <schwab@suse.de>
27586         PR target/80970
27587         * config/m68k/m68k.md (bsetdreg, bchgdreg, bclrdreg): Use "=d"
27588         instead of "+d".
27590 2017-06-20  Prakhar Bahuguna  <prakhar.bahuguna@arm.com>
27592         * config/arm/arm-c.c (arm_cpu_builtins): New block to define
27593         __ARM_FEATURE_COPROC according to support.
27595 2017-06-20  Jakub Jelinek  <jakub@redhat.com>
27597         * tree-chkp.c (chkp_get_hard_register_var_fake_base_address):
27598         Rewritten to avoid overflow for > 32-bit pointers.
27600         PR sanitizer/81125
27601         * ubsan.h (ubsan_encode_value): Workaround buggy clang++ parser
27602         by removing enum keyword.
27603         (ubsan_type_descriptor): Likewise.  Formatting fix.
27605         PR target/81121
27606         * config/i386/i386.md (TARGET_USE_VECTOR_CONVERTS float si->{sf,df}
27607         splitter): Require TARGET_SSE2 in the condition.
27609 2017-06-20  Michael Meissner  <meissner@linux.vnet.ibm.com>
27611         PR target/79799
27612         * config/rs6000/rs6000.c (rs6000_expand_vector_init): Add support
27613         for doing vector set of SFmode on ISA 3.0.
27614         * config/rs6000/vsx.md (vsx_set_v4sf_p9): Likewise.
27615         (vsx_set_v4sf_p9_zero): Special case setting 0.0f to a V4SF
27616         element.
27617         (vsx_insert_extract_v4sf_p9): Add an optimization for inserting a
27618         SFmode value into a V4SF variable that was extracted from another
27619         V4SF variable without converting the element to double precision
27620         and back to single precision vector format.
27621         (vsx_insert_extract_v4sf_p9_2): Likewise.
27623 2017-06-19  Jakub Jelinek  <jakub@redhat.com>
27625         * tree-ssa-structalias.c (get_constraint_for_ptr_offset): Multiply
27626         in UWHI to avoid undefined overflow.
27628         PR sanitizer/81125
27629         * ubsan.h (enum ubsan_encode_value_phase): New.
27630         (ubsan_encode_value): Change second argument to
27631         enum ubsan_encode_value_phase with default value of
27632         UBSAN_ENCODE_VALUE_GENERIC.
27633         * ubsan.c (ubsan_encode_value): Change second argument to
27634         enum ubsan_encode_value_phase PHASE from bool IN_EXPAND_P,
27635         adjust uses, for UBSAN_ENCODE_VALUE_GENERIC use just
27636         create_tmp_var_raw instead of create_tmp_var and use a
27637         TARGET_EXPR.
27638         (ubsan_expand_bounds_ifn, ubsan_build_overflow_builtin,
27639         instrument_bool_enum_load, ubsan_instrument_float_cast): Adjust
27640         ubsan_encode_value callers.
27642         PR sanitizer/81111
27643         * ubsan.c (ubsan_encode_value): If current_function_decl is NULL,
27644         use create_tmp_var_raw instead of create_tmp_var, mark it addressable
27645         just by setting TREE_ADDRESSABLE on the result and use a TARGET_EXPR.
27647 2017-06-19  Richard Biener  <rguenther@suse.de>
27649         PR middle-end/81118
27650         * tree-cfgcleanup.c (cleanup_tree_cfg_noloop): Clear niter
27651         estimates if we changed anything.
27653 2017-06-19  Richard Biener  <rguenther@suse.de>
27655         PR tree-optimization/80887
27656         * tree-ssa-sccvn.c (mprts_hook_cnt): New global.
27657         (vn_lookup_simplify_result): Allow only mprts_hook_cnt succesful
27658         simplified lookups, then reset mprts_hook.
27659         (vn_nary_build_or_lookup_1): Set mprts_hook_cnt to 9 before
27660         simplifying.
27661         (try_to_simplify): Likewise.
27663 2017-06-19  Martin Liska  <mliska@suse.cz>
27665         PR sanitizer/80879
27666         * gimplify.c (gimplify_switch_expr):
27667         Initialize live_switch_vars for SWITCH_BODY == STATEMENT_LIST.
27669 2017-06-19  Martin Liska  <mliska@suse.cz>
27671         * doc/install.texi: Document that PGO runs in 4 stages.
27673 2017-06-19  Martin Liska  <mliska@suse.cz>
27675         PR ipa/80732
27676         * attribs.c (make_dispatcher_decl): Do not append '.ifunc'
27677         to dispatcher function name.
27678         * multiple_target.c (replace_function_decl): New function.
27679         (create_dispatcher_calls): Redirect both edges and references.
27681 2017-06-19  Jan Hubicka <hubicka@ucw.cz>
27683         * profile-count.c (profile_count::dump): Dump quality.
27684         (profile_count::differs_from_p): Update for unsigned val.
27685         * profile-count.h (profile_count_quality): New enum.
27686         (profile_count): Turn m_val to 62bit unsigned, add quality tracking.
27688 2017-06-19  Richard Biener  <rguenther@suse.de>
27690         * tree-ssa-loop-niter.h (estimate_numbers_of_iterations): Take
27691         struct function as arg.
27692         (estimate_numbers_of_iterations): Export overload with loop arg.
27693         (free_numbers_of_iterations_estimates_loop): Use an overload of
27694         free_numbers_of_iterations_estimates instead.
27695         * tree-cfg.c (remove_bb): Adjust.
27696         * tree-cfgcleanup.c (remove_forwarder_block_with_phi): Likewise.
27697         * tree-parloops.c (gen_parallel_loop): Likewise.
27698         * tree-ssa-loop-ivcanon.c (canonicalize_induction_variables):
27699         Likewise.
27700         (tree_unroll_loops_completely): Likewise.
27701         * tree-ssa-loop-niter.c (estimate_numbers_of_iterations_loop):
27702         Use an overload instead and export.
27703         (estimated_loop_iterations): Adjust.
27704         (max_loop_iterations): Likewise.
27705         (likely_max_loop_iterations): Likewise.
27706         (estimate_numbers_of_iterations): Take struct function as arg
27707         and adjust.
27708         (loop_exits_before_overflow): Adjust.
27709         (free_numbers_of_iterations_estimates_loop): Use an overload.
27710         * tree-vect-loop.c (vect_analyze_loop_form): Adjust.
27711         * tree-vectorizer.c (vect_free_loop_info_assumptions): Likewise.
27713 2017-06-19  Richard Biener  <rguenther@suse.de>
27715         PR ipa/81112
27716         * ipa-prop.c (find_constructor_constant_at_offset): Handle
27717         RANGE_EXPR conservatively.
27719 2017-06-16  Carl Love  <cel@us.ibm.com>
27721         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
27722         definitions for vec_float, vec_float2, vec_floato,
27723         vec_floate built-ins.
27724         * config/rs6000/vsx.md (define_c_enum "unspec"): Add RTL code
27725         for instructions vsx_xvcvsxws vsx_xvcvuxwsp, float2, floato and
27726         floate.
27727         * config/rs6000/rs6000-builtin.def (FLOAT2_V2DI, FLOATE_V2DF,
27728         FLOATE_2DI, FLOATO_V2DF, FLOATEE_V2DI, XVCVSXWSP_V4SF,
27729         UNS_FLOATO_V2DI, UNS_FLOATE_V2DI): Add definitions.
27730         * config/altivec.md (define_insn "p8_vmrgew_<mode>",
27731         define_mode_attr VF_sxddp): Add V4SF type to p8_vmrgew.
27732         * config/rs6000/altivec.h (vec_float, vec_float2, vec_floate,
27733         vec_floato): Add builtin defines.
27734         * doc/extend.texi (vec_float, vec_float2, vec_floate, vec_floato):
27735         Update the built-in documentation file for the new built-in
27736         functions.
27738 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27740         * config/arm/arm.opt (marm): Mark as the negative of of -mthumb.
27741         (mthumb): Mark as the negative of -marm.
27743 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27745         * doc/invoke.texi (ARM Options, -mcpu): Document supported
27746         extension options.
27747         (ARM Options, -mtune): Document that this accepts the same
27748         extension options as -mcpu.
27749         (ARM Options, -mfpu): Document addition of -mfpu=auto.
27751 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27753         * doc/invoke.texi (ARM Options, -march=): Document new syntax and
27754         permitted extensions.
27756 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27758         * config/arm/arm-cpus.in (armv7): Add extension +nofp.
27759         (armv7-r): Add aliases vfpv3xd and vfpv3-d16.
27760         (armv8-m.main): Add option +nodsp.
27761         * config/arm/arm-cpu-cdata.h: Regenerated.
27763 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27765         * config/arm/t-fuchsia: New file.
27766         * config.gcc (arm*-*-fuchsia*): Use it.
27768 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27770         * config/arm/t-symbian: Rewrite for new option infrastructure.
27772 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27774         * config/arm/t-phoenix (MULTILIB_REUSE): Clear variable.
27775         (MULTILIB_REQUIRED): Likewise.
27777 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27779         * config/arm/t-linux-eabi (MULTILIB_EXCEPTIONS): Set to empty.
27780         (MULTILIB_RESUE): Likewise.
27781         (MULTILIB_MATCHES): Likewise.
27782         (MULTLIB_REQUIRED): Likewise.
27784 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27786         * config/arm/t-rtems: Rewrite for new option framework.
27788 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27790         * config/arm/t-aprofile (v7_a_nosimd_variants, v7_a_simd_variants)
27791         (v7ve_nosimd_variatns, v7ve_vfpv3_simd_variants)
27792         (v7ve_vfpv4_simd_variants, v8_a_nosimd_variants, v8_a_simd_variants)
27793         (v8_1_a_simd_variants, v8_2_a_simd_variants): Move to ...
27794         * config/arm/t-multilib: ... here.
27795         (MULTILIB_OPTIONS): Add armv7 and armv7+fp architectures.
27796         (MULTILIB_MATCHES): Use armv7 libraries for armv7-r.  Also use for
27797         armv7-a and armv8*-a when A-profile libraries have not been built.
27798         * config/arm/t-rmprofile: Rewrite.
27800 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27802         * genmultilib (multilib_reuse): Allow an explicit period to be escaped
27803         with a backslash.  Remove the backslash after substituting unescaped
27804         periods.
27805         * doc/fragments.texi (MULTILIB_REUSE): Document it.
27807 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27809         * config.gcc: (arm*-*-*): When building a-profile libraries, force
27810         the driver to pass through the default setting of -mfloat-abi.
27811         * common/config/arm/arm-common.c (arm_target_thumb_only): Return -marm
27812         rather than NULL.
27813         * config/arm/t-multilib (MULTILIB_REUSE): Initialize to empty.
27814         (all_feat_combs): New rule.
27815         (MULTILIB_OPTIONS): Use explicit ARM and Thumb directories.  Rework
27816         default libraries.
27817         * config/arm/t-aprofile: Rewrite.
27819 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27821         * config/arm/arm.h (FPUTYPE_AUTO): Define.
27822         * config/arm/arm.c (arm_option_override): Use FPUTYPE_AUTO if the
27823         fpu is not specified by the user/command-line.
27824         * config/arm/bpabi.h (FPUTYPE_DEFAULT): Delete.
27825         * config/arm/netbsd-elf.h (FPUTYPE_DEFAULT): Delete.
27826         * config/arm/linux-elf.h (FPUTYPE_DEFAULT): Delete.
27827         * config/arm/vxworks.h (FPUTYPE_DEFAULT): Delete.
27828         * common/config/arm/arm-common.c (arm_canon_arch_option): Use
27829         FPUTYPE_AUTO insted of FPUTYPE_DEFAULT.
27831 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27833         * config/arm/elf.h (MULTILIB_DEFAULTS): Delete.
27834         * config/arm/t-arm-elf: Rewritten.
27836 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27838         * config/arm/arm.h (TARGET_HARD_FLOAT): Also check that we
27839         have some floating-point instructions.
27840         (TARGET_SOFT_FLOAT): Define as inverse of TARGET_HARD_FLOAT.
27841         (TARGET_MAYBE_HARD_FLOAT): New macro.
27842         * config/arm/arm-builtins.c (arm_init_builtins): Use
27843         TARGET_MAYBE_HARD_FLOAT.
27844         * config/arm/arm.c (arm_option_override): Use TARGET_HARD_FLOAT_ABI.
27846 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27848         * common/config/arm/arm-common.c: Define INCLUDE_LIST.
27849         (configargs.h): Include it.
27850         (arm_print_hint_for_fpu_option): New function.
27851         (arm_parse_fpu_option): New function.
27852         (candidate_extension): New class.
27853         (arm_canon_for_multilib): New function.
27854         * config/arm/arm.h (CANON_ARCH_SPEC_FUNCTION): New macro.
27855         (EXTRA_SPEC_FUNCTIONS): Add CANON_ARCH_SPEC_FUNCTION.
27856         (ARCH_CANONICAL_SPECS): New macro.
27857         (DRIVER_SELF_SPECS): Add ARCH_CANONICAL_SPECS.
27859 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27861         * config.gcc (arm*-*-*): Ensure both target_cpu_cname and with_cpu
27862         are set after handling multilib fragments.  Set target_cpu_default2
27863         from with_cpu.
27865 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27867         * config.gcc (arm*-*-fucshia*): Set target_cpu_cname to the real
27868         cpu name.
27869         (arm*-*-*): Set target_cpu_default2 to a quoted string.
27870         * config/arm/parsecpu.awk (check_cpu): Validate any extension
27871         options.
27872         (check_arch): Likewise.
27873         * config/arm/arm.c (arm_configure_build_target): Handle
27874         TARGET_CPU_DEFAULT being a string constant.  Scan any feature
27875         options in the default.
27877 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27879         * config/arm/arm-protos.h (cpu_arch_extension): Add field to record
27880         when an option is an alias of another.
27881         * config/arm/parsecpu.awk (optalias): New parser token.
27882         (gen_comm_data): Mark non-alias options as such.  Emit entries
27883         for extension aliases.
27884         * config/arm/arm-cpus.in (armv5e): Make vfpv2 an alias.
27885         (armv5te, armv5tej, armv6, armv6j, armv6k, armv6z): Likewise.
27886         (armv6kz, armv6zk, armv6t2): Likewise.
27887         (armv7): Make vfpv3-d16 an alias.
27888         (armv7-a): Make vfpv3-d16, neon and neon-vfpv3 aliases.  Sort in
27889         canonical order.
27890         (armv7ve): Make vfpv4-d16, neon-vfpv3 and neon-vfpv4 aliases.
27891         Sort in canonical order.
27892         (armv8-a): Sort in canonical order.
27893         (armv8.1-a, armv8.2-a):  Likewise.
27894         (generic-armv7-a): Make neon and neon-vfpv3 aliases.  Sort in
27895         canonical order.
27896         (cortex-a9): Sort in canonical order.
27897         * config/arm/arm.c (selftests.h): Include it.
27898         (arm_test_cpu_arch_data): New function.
27899         (arm_run_self_tests): New function.
27900         (TARGET_RUN_TARGET_SELFTESTS): Redefine.
27901         (targetm): Move declaration to the end of the file.
27902         * arm-cpu-cdata.h: Regenerated.
27904 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27906         * config/arm/arm.h (TARGET_MODE_SPECS): Add additional parameter to
27907         call to target_mode_check describing the type of option passed.
27908         * common/config/arm/arm-common.c (arm_arch_core_flag): Delete.
27909         (arm_target_thumb_only): Use arm_parse_arch_option_name or
27910         arm_parse_cpu_option_name to match parameters against list of
27911         available targets.
27912         * config/arm/parsecpu.awk (gen_comm_data): Don't generate
27913         arm_arch_core_flags data structure.
27914         * config/arm/arm-cpu_cdata.h: Regenerated.
27916 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27918         * common/config/arm/arm-common.c (arm_initialize_isa): Moved here from
27919         config/arm/arm.c.
27920         (arm_print_hint_for_cpu_option): Likewise.
27921         (arm_print_hint_for_arch_option): Likewise.
27922         (arm_parse_cpu_option_name): Likewise.
27923         (arm_parse_arch_option_name): Likewise.
27924         * config/arm/arm.c (arm_identify_fpu_from_isa): Use the computed number
27925         of entries in the all_fpus list.
27926         * config/arm/arm-protos.h (all_architectures, all_cores): Declare.
27927         (arm_parse_cpu_option_name): Declare.
27928         (arm_parse_arch_option_name): Declare.
27929         (arm_parse_option_features): Declare.
27930         (arm_intialize_isa): Declare.
27931         * config/arm/parsecpu.awk (gen_data): Move CPU and architecture
27932         data tables to ...
27933         (gen_comm_data): ... here.  Make definitions non-static.
27934         * config/arm/arm-cpu-data.h: Regenerated.
27935         * config/arm/arm-cpu-cdata.h: Regenerated.
27937 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27939         * config/arm/arm-protos.h (arm_build_target): Remove arch_core.
27940         (cpu_arch_extension): New structure.
27941         (cpu_arch_option, arch_option, cpu_option): New structures.
27942         * config/arm/parsecpu.awk (gen_headers): Build an enumeration of
27943         architecture types.
27944         (gen_data): Generate new format data tables.
27945         * config/arm/arm.c (cpu_tune): New structure.
27946         (cpu_option, processors): Delete.
27947         (arm_print_hint_for_core_or_arch): Delete.  Replace with ...
27948         (arm_print_hint_for_cpu_option): ... this and ...
27949         (arm_print_hint_for_arch_option): ... this.
27950         (arm_parse_arch_cpu_name): Delete.  Replace with ...
27951         (arm_parse_cpu_option_name): ... this and ...
27952         (arm_parse_arch_option_name): ... this.
27953         (arm_unrecognized_feature): Change type of target parameter to
27954         cpu_arch_option.
27955         (arm_parse_arch_cpu_features): Delete.  Replace with ...
27956         (arm_parse_option_features): ... this.
27957         (arm_configure_build_target): Rework to use new configuration data
27958         tables.
27959         (arm_print_tune_info): Rework for new configuration data tables.
27960         * config/arm/arm-cpu-data.h: Regenerated.
27961         * config/arm/arm-cpu.h: Regenerated.
27963 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27965         * Makefile.in (OBJS): Move sbitmap.o from here ...
27966         (OBJS-libcommon): ... to here.
27968 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27970         * config/arm/arm-isa.h (ISA_ALL_FPU_INTERNAL): Renamed from ISA_ALL_FPU.
27971         (ISA_ALL_CRYPTO): New macro.
27972         (ISA_ALL_SIMD): New macro
27973         (ISA_ALL_FP): New macro.
27974         * config/arm/arm.c (fpu_bitlist): Update initializer.
27975         * config/arm/arm-cpus.in: Use new ISA_ALL macros to disable crypto,
27976         simd or fp.
27977         (arm9e): Add fpu.  Add option for nofp
27978         (arm946e-s, arm966e-s, arm968e-s, arm10e, arm1020e, arm1022e): Likewise.
27979         (arm926ej-s, arm1026ej-s): Likewise.
27980         (generic-armv7-a): Add fpu.  Add options for simd, vfpv3, vfpv3-d16,
27981         vfpv3-fp16, vfpv3-d16-fp16, vfpv4, vfpv4-d16, neon, neon-vfp3,
27982         neon-fp16, neon-vfpv4, nofp and nosimd.
27983         (cortex-a5, cortex-a7): Add fpu.  Add options for nosimd and nofp.
27984         (cortex-a8): Add fpu.  Add option for nofp.
27985         (cortex-a9): Add fpu.  Add options for nosimd and nofp.
27986         (cortex-a12, cortex-a15, cortex-a17): Add fpu.  Add option for nofp.
27987         (cortex-r4f): Add fpu.
27988         (cortex-r5): Add fpu.  Add options for nofp.dp and nofp.
27989         (cortex-r7): Use idiv option from architecture.  Add fpu.  Add option
27990         for nofp.
27991         (cortex-r8): Likewise.
27992         (cortex-m4): Add fpu.  Add option for nofp.
27993         (cortex-a15.cortex-a7): Add fpu.  Add option for nofp.
27994         (cortex-a17.cortex-a7): Likewise.
27995         (cortex-a32): Add fpu.  Add options for crypto and nofp.
27996         (cortex-a35, cortex-a53): Likewise.
27997         (cortex-a57): Add fpu.  Add option for crypto.
27998         (cortex-a72, cortex-a73): Likewise.
27999         (exynos-m1): Likewise.
28000         (cortex-a57.cortex-a53, cortex-a72.cortex-a53): Likewise.
28001         (cortex-a73.cortex-a35, cortex-a73.cortex-a53): Likewise.
28002         (cortex-m33): Add fpu.  Add option for nofp.
28003         * config/arm/arm-cpu-cdata.h: Regenerated
28004         * config/arm/arm-cpu-data.h: Regenerated.
28006 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
28008         * arm-cpus.in (armv5e): Add options fp, vfpv2 and nofp.
28009         (armv5te, armv5tej): Likewise.
28010         (armv6, armv6j, armv6k, armv6z, armv6kz, armv6zk, armv6t2): Likewise.
28011         (armv7): Add options fp and vfpv3-d16.
28012         (armv7-a): Add options fp, simd, vfpv3, vfpv3-d16, vfpv3-d16-fp16,
28013         vfpv3-fp16, vfpv4, vfpv4-d16, neon, neon-vfpv3, neon-fp16, neon-vfpv4,
28014         nofp and nosimd.
28015         (armv7ve): Likewise.
28016         (armv7-r): Add options fp, fp.sp, idiv, nofp and noidiv.
28017         (armv7e-m): Add options fp, fpv5, fp.dp and nofp.
28018         (armv8-a): Add nocrypto option.
28019         (armv8.1-a, armv8.2-a): Likewise.
28020         (armv8-m.main): add options fp, fp.dp and nofp.
28022 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
28024         * config/arm/arm-cpus.in (armv8-a): Add options crc, simd crypto and
28025         nofp.
28026         (armv8-a+crc): Delete.
28027         (armv8.1-a): Add options simd, crypto and nofp.
28028         (armv8.2-a): Add options fp16, simd, crypto and nofp.
28029         (armv8.2-a+fp16): Delete.
28030         (armv8-m.main): Add option dsp.
28031         (armv8-m.main+dsp): Delete.
28032         (cortex-a8): Add fpu.  Add nofp option.
28033         (cortex-a9): Add fpu.  Add nofp and nosimd options.
28034         * config/arm/parsecpu.awk (gen_data): Generate option tables and
28035         link to main cpu and architecture data structures.
28036         (gen_comm_data): Only put isa attributes from the main architecture
28037         in common tables.
28038         (option): New statement for architecture and CPU entries.
28039         * arm.c (struct cpu_option): New structure.
28040         (struct processors): Add entry for options.
28041         (arm_unrecognized_feature): New function.
28042         (arm_parse_arch_cpu_name): Ignore any characters after the first
28043         '+' character.
28044         (arm_parse_arch_cpu_feature): New function.
28045         (arm_configure_build_target): Separate out any CPU and architecture
28046         features and parse separately.  Don't error out if -mfpu=auto is
28047         used with only an architecture string.
28048         (arm_print_asm_arch_directives): New function.
28049         (arm_file_start): Call it.
28050         * config/arm/arm-cpu-cdata.h: Regenerated.
28051         * config/arm/arm-cpu-data.h: Likewise.
28052         * config/arm/arm-tables.opt: Likewise.
28054 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
28056         * config/arm/elf.h (ASM_SPEC): Only pass -mfpu through to the
28057         assembler when it is not -mfpu=auto.
28059 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
28061         * config/arm/arm.h (BIG_LITTLE_SPEC): Delete macro.
28062         (ASM_REWRITE_SPEC_FUNCTIONS): New macro.
28063         (BIG_LITTLE_CPU_SPEC_FUNCTIONS): Delete macro.
28064         (ASM_CPU_SPEC): Rewrite.
28065         (MCPU_MTUNE_NATIVE_FUNCTIONS): New macro.
28066         (EXTRA_SPEC_FUNCTIONS): Move outside of ifdef.  Use
28067         MCPU_MTUNE_NATIVE_FUNCTIONS and ASM_REWRITE_SPEC_FUNCTIONS.  Remove
28068         reference to BIG_LITTLE_CPU_SPEC_FUNCTIONS.
28069         * common/config/arm/arm-common.c (arm_rewrite_selected_cpu): Ensure
28070         copied string is NUL-terminated.  Also strip any characters prefixed
28071         by '+'.
28072         (arm_rewrite_selected_arch): New function.
28073         (arm_rewrite_march): New function.
28075 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
28077         * config/arm/arm.opt (x_arm_arch_string): New TargetSave option.
28078         (x_arm_cpu_string, x_arm_tune_string): Likewise.
28079         (march, mcpu, mtune): Convert to string-based options.
28080         * config/arm/arm.c (arm_print_hint_for_core_or_arch): New function.
28081         (arm_parse_arch_cpu_name): New function.
28082         (arm_configure_build_target): Use arm_parse_arch_cpu_name to
28083         identify selected architecture or CPU.
28084         (arm_option_save): New function.
28085         (TARGET_OPTION_SAVE): Redefine.
28086         (arm_option_restore): Restore string options.
28087         (arm_option_print): Print string options.
28089 2017-06-16  Martin Sebor  <msebor@redhat.com>
28091         PR tree-optimization/80933
28092         PR tree-optimization/80934
28093         * builtins.c (fold_builtin_3): Do not handle bcmp here.
28094         * gimple-fold.c (gimple_fold_builtin_bcmp): New function.
28095         (gimple_fold_builtin_bcopy, gimple_fold_builtin_bzero): Likewise.
28096         (gimple_fold_builtin): Call them.
28098 2017-06-16  Jan Hubicka  <hubicka@ucw.cz>
28100         * gimple-ssa-isolate-paths.c (isolate_path): Set edge leading to path
28101         as unlikely; update profile.
28103 2017-06-16  Jan Hubicka  <hubicka@ucw.cz>
28105         * predict.c (force_edge_cold): Handle declaring edges impossible
28106         more aggresively.
28108 2017-06-16  Jan Hubicka  <hubicka@ucw.cz>
28110         * tree-ssa-loop-ivcanon.c (remove_exits_and_undefined_stmts): Update
28111         profile.
28112         (try_unroll_loop_completely): Fix reporting.
28114 2017-06-16  Jan Hubicka  <hubicka@ucw.cz>
28116         * tree-ssa-tail-merge.c (replace_block_by): Fix profile updating.
28118 2017-06-16  James Greenhalgh  <james.greenhalgh@arm.com>
28120         PR target/71778
28121         * config/arm/arm-builtins.c (arm_expand_builtin_args): Return TARGET
28122         if given a non-constant argument for an intrinsic which requires a
28123         constant.
28125 2017-06-16  Jan Hubicka  <hubicka@ucw.cz>
28127         * profile.c (compare_freqs): New function.
28128         (branch_prob): Sort edge list.
28129         (find_spanning_tree): Assume that the list is priority sorted.
28131 2017-06-16  Richard Biener  <rguenther@suse.de>
28133         PR tree-optimization/81090
28134         * passes.def (pass_record_bounds): Remove.
28135         * tree-pass.h (make_pass_record_bounds): Likewise.
28136         * tree-ssa-loop.c (pass_data_record_bounds, pass_record_bounds,
28137         make_pass_record_bounds): Likewise.
28138         * tree-ssa-loop-ivcanon.c (canonicalize_induction_variables): Do
28139         not free niter estimates at the beginning but at the end.
28140         * tree-scalar-evolution.c (scev_finalize): Free niter estimates.
28142 2017-06-16  Richard Biener  <rguenther@suse.de>
28144         * tree-switch-conversion.c (emit_case_bit_tests): Adjust
28145         initializer to workaround ICE in host GCC 4.8.
28147 2017-06-16  Jan Hubicka  <hubicka@ucw.cz>
28149         * ipa-inline-transform.c (update_noncloned_frequencies): Update also
28150         counts.
28151         (clone_inlined_nodes): Update.
28153 2017-06-16  Maxim Kuvyrkov  <maxim.kuvyrkov@linaro.org>
28155         * config/aarch64/aarch64.c (qdf24xx_prefetch_tune): Update
28156         prefetch settings, and enable prefetching by default at -O3.
28158 2017-06-16  Maxim Kuvyrkov  <maxim.kuvyrkov@linaro.org>
28160         * config/aarch64/aarch64.c (aarch64_override_options_internal):
28161         Set flag_prefetch_loop_arrays according to tuning data.
28163 2017-06-16  Maxim Kuvyrkov  <maxim.kuvyrkov@linaro.org>
28165         * config/aarch64/aarch64-protos.h (struct cpu_prefetch_tune):
28166         New tune structure.
28167         (struct tune_params): Use cpu_prefetch_tune instead of cache_line_size.
28168         [Unrelated to main purpose of the patch] Place the pointer field last
28169         to enable type checking errors when tune structure are wrongly merged.
28170         * config/aarch64/aarch64.c (generic_prefetch_tune,)
28171         (exynosm1_prefetch_tune, qdf24xx_prefetch_tune,)
28172         (thunderx2t99_prefetch_tune): New tune constants.
28173         (tune_params *_tunings): Update all tunings (no functional change).
28174         (aarch64_override_options_internal): Set PARAM_SIMULTANEOUS_PREFETCHES,
28175         PARAM_L1_CACHE_SIZE, PARAM_L1_CACHE_LINE_SIZE, and PARAM_L2_CACHE_SIZE
28176         from tunings structures.
28178 2017-06-16  Jakub Jelinek  <jakub@redhat.com>
28180         PR sanitizer/81094
28181         * ubsan.c (instrument_null): Add T argument, use it instead
28182         of computing it based on IS_LHS.
28183         (instrument_object_size): Likewise.
28184         (pass_ubsan::execute): Adjust instrument_null and
28185         instrument_object_size callers to pass gimple_get_lhs or
28186         gimple_assign_rhs1 result to it.  Use instrument_null instead of
28187         calling get_base_address and instrument_mem_ref.  Handle
28188         aggregate call arguments for object-size sanitization.
28190 2017-06-16  Yury Gribov  <tetra2005@gmail.com>
28192         PR tree-optimization/81089
28193         * tree-vrp.c (is_masked_range_test): Validate operands of
28194         subexpression.
28196 2017-06-15  Martin Sebor  <msebor@redhat.com>
28198         PR c++/80560
28199         * dumpfile.c (dump_register): Avoid calling memset to initialize
28200         a class with a default ctor.
28201         * gcc.c (struct compiler): Remove const qualification.
28202         * genattrtab.c (gen_insn_reserv): Replace memset with initialization.
28203         * hash-table.h: Ditto.
28204         * ipa-cp.c (allocate_and_init_ipcp_value): Replace memset with
28205           assignment.
28206         * ipa-prop.c (ipa_free_edge_args_substructures): Ditto.
28207         * omp-low.c (lower_omp_ordered_clauses): Replace memset with
28208         default ctor.
28209         * params.h (struct param_info): Make struct members non-const.
28210         * tree-switch-conversion.c (emit_case_bit_tests): Replace memset
28211         with default initialization.
28212         * vec.h (vec_copy_construct, vec_default_construct): New helper
28213         functions.
28214         (vec<T>::copy, vec<T>::splice, vec<T>::reserve): Replace memcpy
28215         with vec_copy_construct.
28216         (vect<T>::quick_grow_cleared): Replace memset with default ctor.
28217         (vect<T>::vec_safe_grow_cleared, vec_safe_grow_cleared): Same.
28218         * doc/invoke.texi (-Wclass-memaccess): Document.
28220 2017-06-15  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
28222         * emit-rtl.h (is_leaf): Update comment about local
28223         register allocator.
28225 2017-06-15  Jozef Lawrynowicz   <jozef.l@somniumtech.com>
28227         PR target/78818
28228         * config/msp430/msp430.c (msp430_data_attr): Check that it's possible
28229         for a variable to have a section before checking if the section has a
28230         name.
28231         Set section to.persistent if persistent attribute is set.
28232         Warn if .persistent attribute is used on an automatic variable.
28234 2017-06-15  Eric Botcazou  <ebotcazou@adacore.com>
28236         PR rtl-optimization/80474
28237         * reorg.c (update_block): Do not ignore instructions in a delay slot.
28239 2017-06-15  Segher Boessenkool  <segher@kernel.crashing.org>
28241         * config/rs6000/rs6000.md (add<mode>3): Use reg_or_subregno instead
28242         of REGNO.
28244 2017-06-14  Maciej W. Rozycki  <macro@imgtec.com>
28246         * config/mips/mips.md (MIPS16_T_REGNUM): Remove constant.
28247         (casesi): Emit bounds checking as RTL.
28248         (casesi_internal_mips16_<mode>): Remove bounds checking.
28250 2017-06-14  Max Filippov  <jcmvbkbc@gmail.com>
28252         * config/xtensa/xtensa.c (xtensa_option_override): Append
28253         MASK_CONST16 to target_flags in the absence of TARGET_L32R.
28254         (hwloop_optimize, hwloop_fail, hwloop_pattern_reg,
28255          xtensa_doloop_hooks): Define unconditionally.
28256         (xtensa_reorg_loops): Only call reorg_loops in the presence of
28257         TARGET_LOOPS.
28258         * config/xtensa/xtensa.h (TARGET_L32R): New definition.
28259         (TARGET_DEFAULT): Remove XCHAL_HAVE_L32R condition and account
28260         for it in xtensa_option_override.
28261         (HARD_FRAME_POINTER_IS_FRAME_POINTER,
28262          HARD_FRAME_POINTER_IS_ARG_POINTER): New definitions.
28264 2017-06-14  Boris Kolpackov  <boris@codesynthesis.com>
28266         * doc/cppopts.texi: Document '-' special value to -MF.
28268 2017-06-14  Wilco Dijkstra  <wdijkstr@arm.com>
28270         * config/arm/cortex-a53.md (cortex_a53_fpalu) Adjust latency.
28271         (cortex_a53_fconst): Likewise.
28272         (cortex_a53_fpmul): Likewise.
28273         (cortex_a53_f_load_64): Likewise.
28274         (cortex_a53_f_load_many): Likewise.
28275         (cortex_a53_advsimd_alu): Likewise.
28276         (cortex_a53_advsimd_alu_q): Likewise.
28277         (cortex_a53_advsimd_mul): Likewise.
28278         (cortex_a53_advsimd_mul_q): Likewise.
28279         (fpmac bypass): Add new bypass for fpmac-fpmac case.
28280         Add missing fmul, r2f_cvt and fconst cases.
28282 2017-06-14  Richard Biener  <rguenther@suse.de>
28284         PR middle-end/81088
28285         * fold-const.c (split_tree): Drop TREE_OVERFLOW flag from
28286         literal constants.
28287         (fold_binary_loc): When associating do not treat pre-existing
28288         TREE_OVERFLOW on literal constants as a reason to allow
28289         TREE_OVERFLOW on associated literal constants.
28291 2017-06-14  Eric Botcazou  <ebotcazou@adacore.com>
28293         * config/sparc/sparc.h (MASK_ISA): Add MASK_LEON and MASK_LEON3.
28294         (MASK_FEATURES): New macro.
28295         * config/sparc/sparc.c (sparc_option_override): Remove the special
28296         handling of -mfpu and generalize it to all MASK_FEATURES switches.
28298 2017-06-14  Eric Botcazou  <ebotcazou@adacore.com>
28300         * simplify-rtx.c (simplify_binary_operation_1) <UDIV>: Do not simplify
28301         a division of 0 if non-call exceptions are enabled.
28303 2017-06-14  Andrew Pinski  <apinski@cavium.com>
28304             Naveen H.S  <Naveen.Hurugalawadi@cavium.com>
28306         PR target/71663
28307         * config/aarch64/aarch64.c (aarch64_expand_vector_init):
28308         Improve vector initialization code gen for only variable case.
28310 2017-06-14  Eric Botcazou  <ebotcazou@adacore.com>
28312         * config/sparc/driver-sparc.c (cpu_names): Add SPARC-T5 entry.
28314 2017-06-14  Richard Biener  <rguenther@suse.de>
28316         PR tree-optimization/81083
28317         * tree-ssa-sccvn.c (vn_reference_lookup_3): Do not use abnormals
28318         as values.
28320 2017-06-13  Segher Boessenkool  <segher@kernel.crashing.org>
28322         * config/rs6000/rs6000.c: Update all comments that mentioned SPE.
28323         (rs6000_expand_builtin): Remove RS6000_BTC_EVSEL.
28324         * config/rs6000/rs6000.h (RS6000_BTC_EVSEL): Delete.
28325         * config/rs6000/vxworks.h (VXCPU_FOR_8548): Delete.  Adjust former use.
28326         * config/rs6000/vxworksae.h (VXCPU_FOR_8548): Delete.
28327         * config/rs6000/vxworksmils.h (VXCPU_FOR_8548): Delete.
28329 2017-06-13  Segher Boessenkool  <segher@kernel.crashing.org>
28331         * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete VECTOR_SPE.
28332         * config/rs6000/rs6000.c (rs6000_debug_vector_unit): Delete VECTOR_SPE.
28334 2017-06-13  Segher Boessenkool  <segher@kernel.crashing.org>
28336         * config/rs6000/rs6000.h (FIXED_SCRATCH): Delete.
28338 2017-06-13  Segher Boessenkool  <segher@kernel.crashing.org>
28340         * config/rs6000/t-rtems: Don't handle SPE.
28342 2017-06-13  Segher Boessenkool  <segher@kernel.crashing.org>
28344         * config/rs6000/t-linux: Don't handle SPE.
28346 2017-06-13  Segher Boessenkool  <segher@kernel.crashing.org>
28348         * config/rs6000/eabispe.h: Delete file.
28350 2017-06-13  Segher Boessenkool  <segher@kernel.crashing.org>
28352         * config/rs6000/t-spe: Delete file.
28354 2017-06-13  Segher Boessenkool  <segher@kernel.crashing.org>
28356         * config/rs6000/rs6000.c (SPE_CONST_OFFSET_OK): Delete.
28357         (rs6000_legitimate_offset_address_p): Return false for anything in
28358         V2SImode or V2SFmode.
28360 2017-06-13  Segher Boessenkool  <segher@kernel.crashing.org>
28362         * config/rs6000/rs6000-modes.def: Remove all 8-byte vector modes
28363         except V2SF and V2SI.  Rearrange the vector modes, and add comments.
28364         * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove V8QImode
28365         and V4HImode.
28366         (reg_offset_addressing_ok_p): Remove V4HImode and V1DImode.
28367         (rs6000_legitimate_offset_address_p): Ditto.
28368         (rs6000_emit_move): Ditto.
28369         (rs6000_init_builtins): Remove V4HI_type_node.
28371 2017-06-13  Martin Liska  <mliska@suse.cz>
28373         PR sanitize/78204
28374         * asan.c (asan_sanitize_stack_p): Use sanitize_flags_p.
28375         (gate_asan): Likewise.
28376         * asan.h (asan_no_sanitize_address_p): Remove the function.
28377         (sanitize_flags_p): New function.
28378         * builtins.def: Fix coding style.
28379         * common.opt: Use renamed enum value.
28380         * convert.c (convert_to_integer_1): Use sanitize_flags_p.
28381         * doc/extend.texi: Document no_sanitize attribute.
28382         * flag-types.h (enum sanitize_code): Rename SANITIZE_NONDEFAULT
28383         to SANITIZE_UNDEFINED_NONDEFAULT.
28384         * gcc.c (sanitize_spec_function): Use the renamed enum value.
28385         * gimple-fold.c (optimize_atomic_compare_exchange_p):
28386         Use sanitize_flags_p.
28387         * gimplify.c (gimplify_function_tree): Likewise.
28388         * ipa-inline.c (sanitize_attrs_match_for_inline_p): Likewise.
28389         * opts.c (parse_no_sanitize_attribute): New function.
28390         (common_handle_option): Use renamed enum value.
28391         * opts.h (parse_no_sanitize_attribute): Declare.
28392         * tree.c (sanitize_flags_p): New function.
28393         * tree.h: Declared here.
28394         * tsan.c: Use sanitize_flags_p.
28395         * ubsan.c (ubsan_expand_null_ifn): Likewise.
28396         (instrument_mem_ref): Likewise.
28397         (instrument_bool_enum_load): Likewise.
28398         (do_ubsan_in_current_function): Remove the function.
28399         (pass_ubsan::execute): Use sanitize_flags_p.
28400         * ubsan.h: Remove do_ubsan_in_current_function
28401         * tree-cfg.c (print_no_sanitize_attr_value): New function.
28402         (dump_function_to_file): Use it here.
28404 2017-06-13  Martin Jambor  <mjambor@suse.cz>
28406         PR tree-optimization/80803
28407         PR tree-optimization/81063
28408         * tree-sra.c (subtree_mark_written_and_enqueue): Move up in the file.
28409         (propagate_subaccesses_across_link): Enqueue subtree whenever
28410         necessary instead of relying on the caller.
28412 2017-06-13  Martin Jambor  <mjambor@suse.cz>
28414         * tree-sra.c (add_access_to_work_queue): Only enqueue accesses
28415         that have a first_link.
28416         (sort_and_splice_var_accesses): Do not check first_link before
28417         enquing.
28418         (subtree_mark_written_and_enqueue): Likewise.
28419         (propagate_all_subaccesses): Likewise and do not stop at first
28420         parent with a first_link.
28422 2017-06-13  Martin Jambor  <mjambor@suse.cz>
28424         * tree-sra.c (dump_access_tree_1): Fix accidental dumping to stderr
28425         instead of f.
28427 2017-06-13  Yury Gribov  <tetra2005@gmail.com>
28429         * match.pd: New pattern.
28431 2017-06-13  Yury Gribov  <tetra2005@gmail.com>
28433         * tree-vrp.c (is_masked_range_test): New function.
28434         (register_edge_assert_for): Determine ranges for
28435         some bit tests.
28437 2017-06-13  Yury Gribov  <tetra2005@gmail.com>
28439         PR tree-optimization/67328
28440         * fold-const.c (maskable_range_p): New function.
28441         (build_range_check): Generate bittests if possible.
28443 2017-06-13  Martin Liska  <mliska@suse.cz>
28445         * gimple-pretty-print.c (dump_probability): Add new argument.
28446         (dump_edge_probability): Dump both probability and count.
28447         (dump_gimple_label): Likewise.
28448         (dump_gimple_bb_header): Likewise.
28450 2017-06-13  Georg-Johann Lay  <avr@gjlay.de>
28452         PR target/81072
28453         * config/avr/avr-devices.c: Fix indentation.
28454         * config/avr/gen-avr-mmcu-specs.c: Dito.
28456 2017-06-13  Richard Biener  <rguenther@suse.de>
28458         * tree-vect-loop.c (vect_model_reduction_cost): Do not fail,
28459         instead get vector type from stmt_info.
28460         (vectorizable_reduction): Adjust.  Remove dead code.
28462 2017-06-13  Richard Biener  <rguenther@suse.de>
28464         PR middle-end/81065
28465         * fold-const.c (extract_muldiv_1): Remove bogus distribution
28466         case of C * (x * C2 + C3).
28467         (fold_addr_of_array_ref_difference): Properly fold index difference.
28469 2017-06-12  David S. Miller  <davem@davemloft.net>
28471         PR target/80968
28472         * config/sparc/sparc.md (return expander): Emit frame blockage if
28473         function uses alloca.
28475 2017-06-12  Richard Sandiford  <richard.sandiford@linaro.org>
28477         * combine.c (make_field_assignment): Check len rather than the mode
28478         precision when calling force_to_mode.
28480 2017-06-12  Georg-Johann Lay  <avr@gjlay.de>
28482         Support multilibs and devices that see flash in RAM address range.
28484         PR target/81072
28485         * config/avr/avr-arch.h (avr_arch_id) <ARCH_AVRXMEGA3>: New enum.
28486         (avr_mcu_t) <flash_pm_offset>: New field.
28487         (avr_device_specific_features) <AVR_ISA_RCALL>: New enum.
28488         * config/avr/avr.h (AVR_SHORT_CALLS): New define.
28489         (AVR_HAVE_JMP_CALL): Don't set if AVR_SHORT_CALLS.
28490         (AVR_TINY_PM_OFFSET): Remove macro.
28491         * config/avr/avr.opt (-mshort-calls): New option.
28492         * config/avr/gen-avr-mmcu-specs.c (print_mcu)
28493         [*self_spec]: Add / remove -mshort-calls depending on AVR_ISA_RCALL.
28494         * config/avr/avr-c.c (avr_cpu_cpp_builtins)
28495         <__AVR_SHORT_CALLS__>: Built-in define if AVR_SHORT_CALLS.
28496         <__AVR_HAVE_JMP_CALL__>: Use AVR_HAVE_JMP_CALL as condition
28497         instead of avr_arch->have_jmp_call.
28498         <__AVR_PM_BASE_ADDRESS__>: Built-in define if avr_arch->flash_pm_offset.
28499         [AVR_TINY] <__AVR_TINY_PM_BASE_ADDRESS__>: Use
28500         avr_arch->flash_pm_offset to define.
28501         * config/avr/avr-devices.c (avr_arch_types): Add initializers for
28502         new field flash_pm_offset.  Add entry for avrxmega3.
28503         (avr_texinfo): Add entry for avrxmega3.
28504         * config/avr/avr-mcus.def: Add entries for: avrxmega3,
28505         attiny212, attiny214,
28506         attiny412, attiny414, attiny416, attiny417,
28507         attiny814, attiny816, attiny817,
28508         attiny1614, attiny1616, attiny1617,
28509         attiny3214, attiny3216, attiny3217.
28510         * config/avr/avr.c (avr_assemble_integer)[AVR_TINY]: Use
28511         avr_arch->flash_pm_offset instead of AVR_TINY_PM_OFFSET.
28512         (avr_print_operand_address) [AVR_TINY]: Same.
28513         (avr_asm_init_sections) <readonly_data_section>: Only patch
28514         callback if avr_arch->flash_pm_offset = 0.
28515         (avr_asm_named_section) <avr_need_copy_data_p>: Skip setting it
28516         for rodata if avr_arch->flash_pm_offset != 0.
28517         (avr_encode_section_info) [AVR_TINY]: Adjust comment.
28518         * config/avr/genmultilib.awk (dir_rcall, opt_rcall): New vars.
28519         (opts) [AVR_ISA_RCALL]: Append opt_rcall.
28520         (m_options): Append opt_rcall.
28521         (m_dirnames): Append dir_rcall.
28522         * config/avr/t-multilib: Regenerate.
28524         * configure.ac [target=avr]: Check whether avrxmega3 default
28525         linker description file works as needed.
28526         * configure: Regenerate.
28527         * doc/avr-mmcu.texi: Regenerate.
28528         * doc/invoke.texi (AVR Options) <-mshort-calls>: Document it.
28529         <__AVR_ARCH__>: Document avrxmega3 and 103.
28530         <__AVR_HAVE_JMP_CALL__>: Adjust documentation.
28531         <__AVR_SHORT_CALLS__>: Document it.
28532         <__AVR_PM_BASE_ADDRESS__>: Document it.
28533         * doc/extend.texi (AVR Options) <-mshort-calls>: Document it.
28534         (AVR Variable Attributes) <progmem>: Document this is
28535         not needed for avrxmega3.
28536         (AVR Named Address Spaces) <__flash>: Dito.
28538 2017-06-12  Jan Hubicka  <hubicka@ucw.cz>
28540         * cgraph.c (cgraph_node::dump): Complain about profile insanities.
28542 2017-06-12  Doug Rupp  <rupp@adacore.com>
28544         * config.gcc (*-*-vxworks*): Set use_gcc_stdint to "provide".
28545         Append vxworks-stdint.h to the tm_file list.
28546         * config/vxworks-stdint.h: New file.
28548 2017-06-12  Martin Liska  <mliska@suse.cz>
28550         PR tree-optimization/81041
28551         * tree-profile.c (gimple_gen_ic_func_profiler):
28552         Create an extra BB in profile-generate
28553         (gimple_gen_time_profiler): Likewise.
28555 2017-06-12  Jakub Jelinek  <jakub@redhat.com>
28557         PR tree-optimization/81003
28558         * tree-ssa-reassoc.c (force_into_ssa_name): New function.
28559         (update_range_test): Use it instead of force_gimple_operand_gsi.
28561 2017-06-12  Richard Biener  <rguenther@suse.de>
28563         PR tree-optimization/81053
28564         * tree-vect-loop.c (vect_is_simple_reduction): Handle PHI
28565         with backedge value not defined in loop.  Simplify def stmt
28566         compute.
28568 2017-06-11  Tom de Vries  <tom@codesourcery.com>
28570         PR target/79939
28571         * config/nvptx/nvptx.c (nvptx_cannot_force_const_mem): New function.
28572         Return true.
28573         (TARGET_CANNOT_FORCE_CONST_MEM): Redefine to
28574         nvptx_cannot_force_const_mem.
28576 2017-06-10  Jan Hubicka  <hubicka@ucw.cz>
28578         * opts.c (finish_options): Move test for flag_split_stack after
28579         it has been initialized.
28581 2017-06-11  Jason Merrill  <jason@redhat.com>
28583         * tree.h (id_equal): New.
28584         * dwarf2out.c, hsa-gen.c, ipa-devirt.c, omp-expand.c,
28585         omp-simd-clone.c, read-rtl-function.c, tree-chkp.c, tree.c: Use it
28586         instead of strcmp of IDENTIFIER_POINTER.
28588 2017-06-10  Jan Hubicka  <hubicka@ucw.cz>
28590         * ipa-inline-transform.c: Include function.h, cfg.h and basic-block.h
28591         (mark_all_inlined_calls_cdtor): Fix formating.
28592         (inline_transform): Rescale profile before inlining.
28594 2017-06-10  Jan Hubicka  <hubicka@ucw.cz>
28596         * cgraph.h (cgraph_edge::clone): Update prototype.
28597         * cgraphclones.c (cgraph_edge::clone): Update profile scaling.
28598         (cgraph_node::create_clone): Update.
28599         (cgraph_node::create_version_clone): Update.
28600         * tree-inline.c (copy_bb): Update.
28601         (expand_call_inline): Update.
28603 2017-06-10  Segher Boessenkool  <segher@kernel.crashing.org>
28605         * config/rs6000/rs6000.c (emit_vrsave_prologue): New function,
28606         factored out from ...
28607         (rs6000_emit_prologue): ... here.
28609 2017-06-10  Segher Boessenkool  <segher@kernel.crashing.org>
28611         * config/rs6000/rs6000.c (emit_split_stack_prologue): New function,
28612         factored out from ...
28613         (rs6000_emit_prologue): ... here.
28615 2017-06-10  Jan Hubicka  <hubicka@ucw.cz>
28617         * predict.c (drop_profile): Also drop individual bb/edge and cgraph
28618         edge counts.
28619         (handle_missing_profiles): Fix computation of tp_first_run.
28620         (counts_to_freqs): Do not touch freqs when count is 0.
28622 2017-06-10  Jan Hubicka  <hubicka@ucw.cz>
28624         * cgraphbuild.c (cgraph_edge::rebuild_references): Do not touch
28625         profile.
28627 2017-06-10  Tom de Vries  <tom@codesourcery.com>
28629         * doc/sourcebuild.texi (Effective-Target Keywords, Environment
28630         attributes): Document signal effective target.
28632 2017-06-10  Tom de Vries  <tom@codesourcery.com>
28634         * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
28635         Document effective target stack_size.
28637 2017-06-09  David Malcolm  <dmalcolm@redhat.com>
28639         * diagnostic.c (diagnostic_report_diagnostic): Only add fixits
28640         to the edit_context if they can be auto-applied.
28642 2017-06-9  Ian Lance Taylor  <iant@golang.org>
28644         * opts.c (finish_options): If -fsplit-stack, disable implicit
28645         -forder-blocks-and-partition.
28646         * doc/invoke.texi (Optimize Options): Document that when using
28647         -fsplit-stack -forder-blocks-and-partition is not implicitly
28648         enabled.
28650 2017-06-09  Jan Hubicka  <hubicka@ucw.cz>
28652         * builtin-attrs.def (ATTR_NORETURN_NOTHROW_LEAF_COLD_LIST,
28653         ATTR_CONST_NORETURN_NOTHROW_LEAF_COLD_LIST,
28654         ATTR_TMPURE_NORETURN_NOTHROW_LEAF_COLD_LIST): New.
28655         * builtins.def (abort, trap, unreachable): Declare cold.
28656         * calls.c (flags_from_decl_or_type): Lookup ECF_COLD.
28657         * tree-core.h (ECF_COLD): New.
28658         * tree.c (set_call_expr_flags): Handle ECF_COLD.
28659         (build_common_builtin_nodes): Mark unreachable and abort as cold.
28661 2017-06-09  Jan Hubicka  <hubicka@ucw.cz>
28663         * predict.c (unlikely_executed_stmt_p): Cleanup.
28665 2017-06-09  Richard Biener  <rguenther@suse.de>
28667         * tree-ssa-loop-im.c (execute_sm): Do not force multi-threaded
28668         model if the ref is always written to.
28670 2017-06-09  Tamar Christina  <tamar.christina@arm.com>
28672         * config/aarch64/aarch64.md (lrint<GPF:mode><GPI:mode>2): New.
28674 2017-06-09  Tamar Christina  <tamar.christina@arm.com>
28676         * config/arm/arm.c (arm_rtx_costs_internal): Make sdiv more expensive
28677         than udiv.
28679 2017-06-09  Tom de Vries  <tom@codesourcery.com>
28681         PR target/80855
28682         * config/nvptx/nvptx.md (define_expand "mov<QHSDISDFM>"): Error out with
28683         "target cannot support label values" when encountering LABEL_REF.
28685 2017-06-09  Martin Liska  <mliska@suse.cz>
28687         * tree-profile.c (gimple_gen_ic_profiler): Update comment.
28688         (gimple_gen_ic_func_profiler): Emit direct comparison
28689         of __gcov_indirect_call_callee with NULL.
28690         (gimple_gen_time_profiler): Change probability from
28691         PROB_VERY_UNLIKELY to PROB_UNLIKELY.
28693 2017-06-09  Jan Hubicka  <hubicka@ucw.cz>
28695         * profile.c (edge_gcov_counts): Turn to pointer.
28696         (compute_branch_probabilities, compute_branch_probabilities): Update.
28697         (branch_prob): Do not clear edge_gcov_count.
28698         * profile.h (edge_gcov_counts): Turn to pointer.
28699         (edge_gcov_count): Update.
28701 2017-06-09  Jan Hubicka  <hubicka@ucw.cz>
28703         * gimple.h (gimple_check_failed): Mark cold.
28705 2017-06-09  Richard Biener  <rguenther@suse.de>
28707         PR tree-optimization/66623
28708         * tree-vect-loop.c (vect_is_simple_reduction): Cleanup,
28709         refactor check_reduction into two parts, properly computing
28710         whether we have to check reduction validity for outer loop
28711         vectorization.
28713 2017-06-09  Richard Biener  <rguenther@suse.de>
28715         PR tree-optimization/79483
28716         * graphite-scop-detection.c (order): New global.
28717         (get_order): Compute bb to order mapping that satisfies code
28718         generation constraints.
28719         (cmp_pbbs): New helper.
28720         (build_scops): Start domwalk at entry block, sort generated
28721         pbbs.
28723 2017-06-09  Richard Biener  <rguenther@suse.de>
28725         PR middle-end/81007
28726         * ipa-polymorphic-call.c
28727         (ipa_polymorphic_call_context::restrict_to_inner_class):
28728         Skip FIELD_DECLs with error_mark_node type.
28729         * passes.def (all_lowering_passes): Run pass_build_cgraph_edges
28730         last again.
28732 2017-06-09  Martin Liska  <mliska@suse.cz>
28734         * predict.c (struct branch_predictor): New struct.
28735         (test_prediction_value_range): New test.
28736         (predict_c_tests): New function.
28737         * selftest-run-tests.c (selftest::run_tests): Run the function.
28738         * selftest.h: Declare new tests.
28740 2017-06-09  Segher Boessenkool  <segher@kernel.crashing.org>
28742         PR target/80966
28743         * config/rs6000/rs6000.c (rs6000_emit_allocate_stack): Assert that
28744         gen_add3_insn did not fail.
28745         * config/rs6000/rs6000.md (add<mode>3): If asked to add a constant to
28746         r0, construct that number in a temporary reg and add that reg to r0.
28747         If asked to put the result in r0 as well, fail.
28749 2017-06-08  Will Schmidt  <will_schmidt@vnet.ibm.com>
28751         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add handling
28752         for early expansion of vec_eqv.
28754 2017-06-08  Jakub Jelinek  <jakub@redhat.com>
28756         PR middle-end/81005
28757         * ubsan.c (instrument_null): Avoid pointless code temporary.
28758         (pass_ubsan::execute): Instrument aggregate arguments of calls.
28760 2017-06-08  Uros Bizjak  <ubizjak@gmail.com>
28762         PR target/81015
28763         Revert:
28764         2016-12-14  Uros Bizjak  <ubizjak@gmail.com>
28766         PR target/59874
28767         * config/i386/i386.md (*ctzhi2): New insn_and_split pattern.
28768         (*clzhi2): Ditto.
28770 2017-06-08  Jan Hubicka  <hubicka@ucw.cz>
28772         * predict.c (unlikely_executed_edge_p): Move ahead.
28773         (probably_never_executed_edge_p): Use it.
28775 2017-06-08  Alexander Ivchenko  <alexander.ivchenko@intel.com>
28777         PR middle-end/79988
28778         * tree-chkp.c (chkp_gimple_call_builtin_p): Remove
28779         gimple_call_builtin_p call.
28781 2017-06-08  Jan Hubicka  <hubicka@ucw.cz>
28783         * system.h (fancy_abort): Annotate by ATTRIBUTE_COLD.
28784         * rtl.h (rtl_check_failed_bounds, rtl_check_failed_type1,
28785         rtl_check_failed_type2, rtl_check_failed_code1,
28786         rtl_check_failed_code2, rtl_check_failed_code_mode,
28787         rtl_check_failed_block_symbol, cwi_check_failed_bounds,
28788         rtvec_check_failed_bounds, rtl_check_failed_flag,
28789         _fatal_insn_not_found, _fatal_insn): Likewise.
28790         * tree.h (tree_contains_struct_check_failed,
28791         tree_check_failed, tree_not_check_failed,
28792         tree_class_check_failed, tree_range_check_failed,
28793         tree_not_class_check_failed, tree_int_cst_elt_check_failed,
28794         tree_vec_elt_check_failed, phi_node_elt_check_failed,
28795         tree_operand_check_failed, omp_clause_check_failed,
28796         omp_clause_operand_check_failed, omp_clause_range_check_failed):
28797         Likewise.
28799 2017-06-08  Jan Hubicka  <hubicka@ucw.cz>
28801         * cgraph.c (cgraph_edge::maybe_hot_p): Do not check
28802         flag_branch_probabilities.
28803         * ipa-inline.c (edge_badness): Likewise.
28804         * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
28805         * postreload-gcse.c (eliminate_partially_redundant_load): Likewise.
28806         * predict.c (maybe_hot_frequency_p): Likewise.
28807         (probably_never_executed): Likewise.
28808         * sched-ebb.c (schedule_ebbs): Likewise.
28809         * sched-rgn.c (find_single_block_region): Likewise.
28810         * tracer.c (tail_duplicate): Likewise.
28812 2017-06-08  Jan Hubicka  <hubicka@ucw.cz>
28814         * opts.c (finish_options): x_flag_reorder_blocks_and_partition no
28815         longer requires x_flag_profile_use.
28817 2017-06-08  Jan Hubicka  <hubicka@ucw.cz>
28819         * cfgrtl.c (cfg_layout_initialize): Check crtl->has_bb_partition
28820         instead of flag_reorder_blocks_and_partition.
28821         * dbxout.c (dbxout_function_end): Likewise.
28822         * dwarf2out.c (gen_subprogram_die): Likewise.
28823         * haifa-sched.c (sched_create_recovery_edges): Likewise.
28824         * hw-doloop.c (reorg_loops): Likewise.
28825         * varasm.c (assemble_start_function,
28826         assemble_end_function): Likewise.
28827         (decide_function_section): Do not check for
28828         flag_reorder_blocks_and_partition.
28830 2017-06-08  Alexander Ivchenko  <alexander.ivchenko@intel.com>
28832         * tree-chkp.c (chkp_get_hard_register_var_fake_base_address):
28833         New function.
28834         (chkp_get_hard_register_fake_addr_expr): Ditto.
28835         (chkp_build_addr_expr): Add check for hard reg case.
28836         (chkp_parse_array_and_component_ref): Ditto.
28837         (chkp_find_bounds_1): Ditto.
28838         (chkp_process_stmt): Don't generate bounds store for
28839         hard reg case.
28841 2017-06-08  Jan Hubicka  <hubicka@ucw.cz>
28843         * predict.c (maybe_hot_bb_p): Do not check profile status.
28844         (maybe_hot_edge_p): Likewise.
28845         (probably_never_executed): Check for zero counts even if profile
28846         is not read.
28847         (unlikely_executed_edge_p): New function.
28848         (unlikely_executed_stmt_p): New function.
28849         (unlikely_executed_bb_p): New function.
28850         (set_even_probabilities): Use unlikely predicates.
28851         (combine_predictions_for_bb): Likewise.
28852         (predict_paths_for_bb): Likewise.
28853         (predict_paths_leading_to_edge): Likewise.
28854         (determine_unlikely_bbs): New function.
28855         (estimate_bb_frequencies): Use it.
28856         (compute_function_frequency): Use zero counts even if profile is
28857         not read.
28858         * profile-count.h: Fix typo.
28860 2017-08-08  Julia Koval  <julia.koval@intel.com>
28862         * config/i386/avx512bwintrin.h (_mm512_mask_cvtepi16_storeu_epi8,
28863         _mm512_mask_cvtsepi16_storeu_epi8,
28864         _mm512_mask_cvtusepi16_storeu_epi8): New intrinsics.
28865         * config/i386/avx512vlbwintrin.h (_mm256_mask_cvtepi16_storeu_epi8,
28866         _mm_mask_cvtsepi16_storeu_epi8, _mm256_mask_cvtsepi16_storeu_epi8,
28867         _mm_mask_cvtusepi16_storeu_epi8, _mm256_mask_cvtusepi16_storeu_epi8,
28868         _mm_mask_cvtepi16_storeu_epi8): New intrinsics.
28869         * config/i386/i386-builtin-types.def (PV8Q, V8QI): New pointer type.
28870         (VOID_FTYPE_PV32QI_V32HI_USI, VOID_FTYPE_PV8QI_V8HI_UQI,
28871         VOID_FTYPE_PV16QI_V16HI_UHI): New function types.
28872         * config/i386/i386-builtin.def (__builtin_ia32_pmovwb128mem_mask,
28873         __builtin_ia32_pmovwb256mem_mask, __builtin_ia32_pmovswb128mem_mask,
28874         __builtin_ia32_pmovswb256mem_mask, __builtin_ia32_pmovuswb128mem_mask,
28875         __builtin_ia32_pmovuswb256mem_mask,
28876         __builtin_ia32_pmovuswb512mem_mask, __builtin_ia32_pmovswb512mem_mask)
28877         __builtin_ia32_pmovwb512mem_mask): New builtins.
28879 2017-08-08  Julia Koval  <julia.koval@intel.com>
28881         PR target/73350,80862
28882         * config/i386/subst.md (round): Fix round pattern.
28883         * config/i386/i386.c (ix86_erase_embedded_rounding):
28884         Fix erasing rounding for the fixed pattern.
28886 2017-06-08  Jan Hubicka  <hubicka@ucw.cz>
28888         * cfgbuild.c (find_many_sub_basic_blocks): Fix thinko.
28890 2017-06-08  Martin Liska  <mliska@suse.cz>
28892         PR gcov-profile/80911
28893         * gcov.c (block_info::block_info): New constructor.
28895 2017-06-07  Carl Love  <cel@us.ibm.com>
28897         * config/rs6000/rs6000-c: The return type of the following
28898         built-in functions was implemented as int not long long.  Fix sign
28899         of return value for the unsigned version of vec_mulo and vec_mule.
28900         vector unsigned long long vec_bperm (vector unsigned long long,
28901                                              vector unsigned char)
28902         vector signed long long vec_mule (vector signed int,
28903                                           vector signed int)
28904         vector unsigned long long vec_mule (vector unsigned int,
28905                                             vector unsigned int)
28906         vector signed long long vec_mulo (vector signed int,
28907                                           vector signed int)
28908         vector unsigned long long vec_mulo (vector unsigned int,
28909                                             vector unsigned int)
28910         * doc/extend.texi: Fix the documentation for the built-in
28911         functions.
28913 2017-06-07  Carl Love  <cel@us.ibm.com>
28915         PR target/80982
28916         * config/rs6000/altivec.md (double<mode>2): Fix the implementation of
28917         for BE.
28919 2017-06-07  Carl Love  <cel@us.ibm.com>
28921         * config/rs6000/altivec.md: Fix argument swizzle in vec_doublel
28922         support, Generate       doublehv for signed int/float for BE case only.
28924 2017-06-07  Alexander Monakov  <amonakov@ispras.ru>
28926         * doc/invoke.texi (mcx16): Rewrite.
28928 2017-06-07  Segher Boessenkool  <segher@kernel.crashing.org>
28930         * config/rs6000/predicates.md (rs6000_nonimmediate_operand): Delete.
28931         * config/rs6000/rs6000.md (*movsi_internal1, movsi_from_sf,
28932         *mov<mode>_softfloat, and an anonymous splitter): Use
28933         nonimmediate_operand instead of rs6000_nonimmediate_operand.
28935 2017-06-07  Segher Boessenkool  <segher@kernel.crashing.org>
28937         * config/rs6000/darwin.h (REGISTER_NAMES): Delete the SPE_ACC and
28938         SPEFSCR registers.
28939         * config/rs6000/rs6000.c (rs6000_reg_names, alt_reg_names): Ditto.
28940         (enum rs6000_reg_type): Delete SPE_ACC_TYPE and SPEFSCR_REG_TYPE.
28941         (rs6000_debug_reg_global): Adjust.
28942         (rs6000_init_hard_regno_mode_ok): Adjust.
28943         (rs6000_dbx_register_number): Adjust.
28944         * config/rs6000/rs6000.h (FIRST_PSEUDO_REGISTER): Change to 115.
28945         (FIXED_REGISTERS, CALL_USED_REGISTERS, CALL_REALLY_USED_REGISTERS):
28946         Remove SPE_ACC and SPEFSCR.
28947         (REG_ALLOC_ORDER): Ditto.
28948         (FRAME_POINTER_REGNUM): Change to 111.
28949         (enum reg_class): Remove the SPE_ACC and SPEFSCR registers.
28950         (REG_CLASS_NAMES): Ditto.
28951         (REG_CLASS_CONTENTS): Delete the SPE_ACC and SPEFSCR registers.
28952         (REGISTER_NAMES): Ditto.
28953         (ADDITIONAL_REG_NAMES): Ditto.
28954         (rs6000_reg_names): Ditto.
28955         * config/rs6000/rs6000.md: Renumber some register number
28956         define_constants.
28958 2017-06-07  Segher Boessenkool  <segher@kernel.crashing.org>
28960         * config/rs6000/darwin.h (REGISTER_NAMES): Delete the SPE high
28961         registers.
28962         * config/rs6000/rs6000.c (rs6000_reg_names, alt_reg_names): Ditto.
28963         * config/rs6000/rs6000.h (FIRST_PSEUDO_REGISTER): Change from 149
28964         to 117.
28965         (DWARF_REG_TO_UNWIND_COLUMN): Do not define.
28966         (FIXED_REGISTERS, CALL_USED_REGISTERS, CALL_REALLY_USED_REGISTERS):
28967         Delete the SPE high registers.
28968         (REG_ALLOC_ORDER): Ditto.
28969         (enum reg_class): Remove SPE_HIGH_REGS.
28970         (REG_CLASS_NAMES): Ditto.
28971         (REG_CLASS_CONTENTS): Delete the SPE high registers.
28972         (REGISTER_NAMES): Ditto.
28973         (rs6000_reg_names): Ditto.
28974         * doc/tm.texi.in: Remove SPE as example.
28975         * doc/tm.texi: Regenerate.
28977 2017-06-07  Segher Boessenkool  <segher@kernel.crashing.org>
28979         * config/rs6000/8540.md (ppc8540_brinc): Delete.
28980         * config/rs6000/e500mc.md (e500mc_brinc): Delete.
28981         * config/rs6000/e500mc64.md (e500mc64_brinc): Delete.
28982         * config/rs6000/rs6000.md (type): Remove "brinc".
28984 2017-06-07  Segher Boessenkool  <segher@kernel.crashing.org>
28986         * config.gcc (powerpc*-*-*): Don't add spe.h to extra_headers.
28987         (powerpc*-linux*spe*): Use ${cpu_type} instead of rs6000.
28988         * config/rs6000/linuxspe.h: Delete file.
28989         * config/rs6000/rs6000.md: Don't include spe.md.
28990         * config/rs6000/spe.h: Delete file.
28991         * config/rs6000/spe.md: Delete file.
28992         * config/rs6000/t-rs6000: Remove spe.md.
28994 2017-06-07  Segher Boessenkool  <segher@kernel.crashing.org>
28996         * config/rs6000/predicates.md (reg_or_mem_operand): Reformat.
28997         (reg_or_none500mem_operand): Delete.
28998         * config/rs6000/rs6000.md (extendsfdf2): Use reg_or_mem_operand
28999         instead of reg_or_none500mem_operand.
29001 2017-06-07  Segher Boessenkool  <segher@kernel.crashing.org>
29003         * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
29004         handling of SPE flags.
29005         * config/rs6000/rs6000.opt (-mspe, -mspe=no, -mspe=yes): Delete.
29007 2017-06-07  Segher Boessenkool  <segher@kernel.crashing.org>
29009         * config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
29010         SPE ABI handling.
29011         * config/rs6000/paired.md (paired_negv2sf2): Rename to negv2sf2.
29012         (paired_absv2sf2, paired_addv2sf3, paired_subv2sf3, paired_mulv2sf3,
29013         paired_divv2sf3): Similar.
29014         * config/rs6000/predicates.md: Replace TARGET_SPE, TARGET_SPE_ABI,
29015         SPE_VECTOR_MODE and SPE_HIGH_REGNO_P by 0; simplify.
29016         * config/rs6000/rs6000-builtin.def: Delete RS6000_BUILTIN_E and
29017         RS6000_BUILTIN_S.
29018         Delete BU_SPE_1, BU_SPE_2, BU_SPE_3, BU_SPE_E, BU_SPE_P, and BU_SPE_X.
29019         Rename the paired_* instruction patterns.
29020         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Do not
29021         define __SPE__.
29022         * config/rs6000/rs6000-protos.h (invalid_e500_subreg): Delete.
29023         * config/rs6000/rs6000.c: Delete RS6000_BUILTIN_E and RS6000_BUILTIN_S.
29024         (struct rs6000_stack): Delete fields spe_gp_save_offset, spe_gp_size,
29025         spe_padding_size, and spe_64bit_regs_used.  Replace TARGET_SPE and
29026         TARGET_SPE_ABI with 0, simplify.  Replace SPE_VECTOR_MODE with
29027         PAIRED_VECTOR_MODE.
29028         (struct machine_function): Delete field spe_insn_chain_scanned_p.
29029         (spe_func_has_64bit_regs_p): Delete.
29030         (spe_expand_predicate_builtin): Delete.
29031         (spe_expand_evsel_builtin): Delete.
29032         (TARGET_DWARF_REGISTER_SPAN): Do not define.
29033         (TARGET_MEMBER_TYPE_FORCES_BLK): Do not define.
29034         (invalid_e500_subreg): Delete.
29035         (rs6000_legitimize_address): Always force_reg op2 as well, for
29036         paired single memory accesses.
29037         (rs6000_member_type_forces_blk): Delete.
29038         (rs6000_spe_function_arg): Delete.
29039         (rs6000_expand_unop_builtin): Delete SPE handling.
29040         (rs6000_expand_binop_builtin): Ditto.
29041         (spe_expand_stv_builtin): Delete.
29042         (bdesc_2arg_spe): Delete.
29043         (spe_expand_builtin): Delete.
29044         (spe_expand_predicate_builtin): Delete.
29045         (spe_expand_evsel_builtin): Delete.
29046         (rs6000_invalid_builtin): Remove RS6000_BTM_SPE handling.
29047         (spe_init_builtins): Delete.
29048         (spe_func_has_64bit_regs_p): Delete.
29049         (savres_routine_name): Delete "info" parameter.  Adjust callers.
29050         (rs6000_emit_stack_reset): Ditto.
29051         (rs6000_dwarf_register_span): Delete.
29052         * config/rs6000/rs6000.h (TARGET_SPE_ABI, TARGET_SPE,
29053         UNITS_PER_SPE_WORD, SPE_HIGH_REGNO_P, SPE_SIMD_REGNO_P,
29054         SPE_VECTOR_MODE, RS6000_BTM_SPE, RS6000_BUILTIN_E, RS6000_BUILTIN_S):
29055         Delete.
29056         * config/rs6000/rs6000.md (FIRST_SPE_HIGH_REGNO, LAST_SPE_HIGH_REGNO):
29057         Delete.
29058         * config/rs6000/rs6000.opt (-mabi=spe, -mabi=no-spe): Delete.
29059         * config/rs6000/spe.md: Delete every pattern that uses TARGET_SPE.
29060         * config/rs6000/vector.md (absv2sf2, negv2sf2, addv2sf3, subv2sf3,
29061         mulv2sf3, divv2sf3): Delete expanders.
29063 2017-06-07  Segher Boessenkool  <segher@kernel.crashing.org>
29065         config/rs6000/rs6000.md (UNSPEC_MV_CR_GT): Delete.
29067 2017-06-07  Segher Boessenkool  <segher@kernel.crashing.org>
29069         * config/rs6000/rs6000-protos.h (output_e500_flip_gt_bit): Delete.
29070         * config/rs6000/rs6000.c: Ditto.
29072 2017-06-07  Segher Boessenkool  <segher@kernel.crashing.org>
29074         * config/rs6000/predicated.md (rs6000_cbranch_operator): Delete.
29075         * config/rs6000/rs6000.md: Replace rs6000_cbranch_operator by
29076         comparison_operator.
29078 2017-06-07  Segher Boessenkool  <segher@kernel.crashing.org>
29080         * config/rs6000/rs6000.c: Remove everything related to -mfloat-gprs.
29081         * config/rs6000/rs6000.opt: Ditto.
29082         * config/rs6000/t-rtems: Ditto.
29084 2017-06-07  Segher Boessenkool  <segher@kernel.crashing.org>
29086         * config/rs6000/predicates.md: Replace TARGET_E500_DOUBLE and
29087         TARGET_E500_SINGLE by 0, simplify.
29088         * config/rs6000/rs6000.c: Ditto.
29089         (rs6000_option_override_internal): Delete CHECK_E500_OPTIONS.
29090         (spe_build_register_parallel): Delete.
29091         * config/rs6000/rs6000.h: Delete TARGET_E500_SINGLE,
29092         TARGET_E500_DOUBLE, and CHECK_E500_OPTIONS.
29093         * config/rs6000/rs6000.md: Replace TARGET_E500_DOUBLE,
29094         TARGET_E500_SINGLE, and <E500_CONVERT> by 0, simplify.
29095         (E500_CONVERT): Delete.
29096         * config/rs6000/spe.md: Remove many patterns and all define_constants.
29098 2017-06-07  Segher Boessenkool  <segher@kernel.crashing.org>
29100         * config/rs6000/darwin.md: Replace TARGET_FPRS by 1 and simplify.
29101         * config/rs6000/dfp.md: Ditto.
29102         (negdd2, *negdd2_fpr): Merge.
29103         (absdd2, *absdd2_fpr): Merge.
29104         (negtd2, *negtd2_fpr): Merge.
29105         (abstd2, *abstd2_fpr): Merge.
29106         * config/rs6000/e500.h: Delete file.
29107         * config/rs6000/predicates.md (rs6000_cbranch_operator): Replace
29108         TARGET_FPRS by 1 and simplify.
29109         * config/rs6000/rs6000-c.c: Ditto.
29110         * config/rs6000/rs6000.c: Ditto.  Also replace TARGET_SF_SPE and
29111         TARGET_DF_SPE by 0.
29112         * config/rs6000/rs6000.h: Ditto.  Delete TARGET_SF_SPE and
29113         TARGET_DF_SPE.
29114         * config/rs6000/rs6000.md: Ditto.
29115         (floatdidf2, *floatdidf2_fpr): Merge.
29116         (move_from_CR_gt_bit): Delete.
29117         * config/rs6000/spe.md: Replace TARGET_FPRS by 1 and simplify.
29118         (E500_CR_IOR_COMPARE): Delete.
29119         (All patterns that require !TARGET_FPRS): Delete.
29120         * config/rs6000/vsx.md: Replace TARGET_FPRS by 1 and simplify.
29122 2017-06-07  Bin Cheng  <bin.cheng@arm.com>
29124         * passes.def (pass_iv_canon): Move before pass_loop_distribution.
29126 2017-06-07  Bin Cheng  <bin.cheng@arm.com>
29128         * graphds.c (add_edge): Intitialize edge's attached data.
29129         (foll_in_subgraph, dfs_fst_edge, dfs_next_edge): New function
29130         pointer parameter.  Call pointed function on each edge during
29131         graph traversing.  Skip traversing the edge when the function
29132         returns true.
29133         (graphds_dfs, graphds_scc): Ditto.
29134         (for_each_edge): New parameter.  Pass the new parameter to callback
29135         function.
29136         * graphds.h (skip_edge_callback): New function pointer type.
29137         (graphds_dfs, graphds_scc): New function pointer parameter.
29138         (graphds_edge_callback, for_each_edge): New parameter.
29140 2017-06-07  Bin Cheng  <bin.cheng@arm.com>
29142         * tree-vect-data-refs.c (vect_mark_for_runtime_alias_test): Factor
29143         out code checking if runtime alias check is possible to below ...
29144         Call the new function.
29145         * tree-data-ref.c (runtime_alias_check_p): ... to new function.
29146         * tree-data-ref.h (runtime_alias_check_p): New decalaration.
29148 2017-06-07  Marek Polacek  <polacek@redhat.com>
29150         PR sanitizer/80932
29151         * fold-const.c (extract_muldiv_1) <case MINUS_EXPR>: Add
29152         TYPE_OVERFLOW_WRAPS check.
29154 2017-06-07  Bin Cheng  <bin.cheng@arm.com>
29156         * tree-vect-loop-manip.c (vect_do_peeling): Don't skip vector loop
29157         if versioning is required.
29158         * tree-vect-loop.c (vect_analyze_loop_2): Merge niter check for loop
29159         peeling with the check for versioning.
29161 2017-06-07  Bin Cheng  <bin.cheng@arm.com>
29163         * tree-vectorizer.h (vect_build_loop_niters): New parameter.
29164         * tree-vect-loop-manip.c (vect_build_loop_niters): New parameter.
29165         Set true to new parameter if new ssa variable is defined.
29166         (vect_gen_vector_loop_niters): Refactor.  Set range information
29167         for the new vector loop bound variable.
29168         (vect_do_peeling): Ditto.
29170 2017-06-07  Bin Cheng  <bin.cheng@arm.com>
29172         * tree-affine.c (ssa.h): Include header file.
29173         (tree_to_aff_combination): Handle (T1)(X - CST) when inner type
29174         has wrapping overflow behavior.
29176 2017-06-07  Bin Cheng  <bin.cheng@arm.com>
29178         * tree-affine.c (tree_to_aff_combination): Handle (T1)(X + X).
29180 2017-06-07  Bin Cheng  <bin.cheng@arm.com>
29182         (aff_combination_expand): Move (T1)(X *+- CST) simplification to ...
29183         (tree_to_aff_combination): ... here.
29185 2017-06-07  Bin Cheng  <bin.cheng@arm.com>
29187         * tree-ssa-loop-ivopts.c (ivopts_estimate_reg_pressure): New
29188         reg_pressure model function.
29189         (ivopts_global_cost_for_size): Delete.
29190         (determine_set_costs, iv_ca_recount_cost): Call new model function
29191         ivopts_estimate_reg_pressure.
29193 2017-06-07  Tamar Christina  <tamar.christina@arm.com>
29195         * config/aarch64/aarch64.c (aarch64_rtx_costs): Make sdiv more
29196         expensive than udiv.  Remove floating point cases from mod.
29198 2017-06-07  Tamar Christina  <tamar.christina@arm.com>
29200         * config/arm/aarch-cost-tables.h (cortexa53_extra_cost):
29201         Increase idiv cost.
29203 2017-06-07  Tamar Christina  <tamar.christina@arm.com>
29205         * config/aarch64/aarch64.md
29206         (copysignsf3): Fix mask generation.
29208 2017-06-07  Jakub Jelinek  <jakub@redhat.com>
29210         * dumpfile.h (enum tree_dump_index): Rename TDI_generic to
29211         TDI_gimple.
29212         (class dump_manager): Add register_dumps method.
29213         * dumpfile.c: Include langhooks.h.
29214         (dump_files): Use 0 instead of 3/4/5 for TDI_{original,gimple,nested}.
29215         (FIRST_AUTO_NUMBERED_DUMP): Decrease to 1.
29216         (FIRST_ME_AUTO_NUMBERED_DUMP): Define.
29217         (dump_manager::dump_register): Start with 512 entries instead of 32.
29218         (dump_manager::register_dumps): New method.
29219         * toplev.c (general_init): Instead of invoking register_dumps
29220         langhook, invoke register_dumps method on the dump manager.
29221         * gimplify.c (gimplify_function_tree): Use TDI_gimple instead of
29222         TDI_generic.
29224 2017-06-07  Richard Sandiford  <richard.sandiford@linaro.org>
29226         * doc/md.texi: Clarify the restrictions on a define_insn condition.
29227         Say that # requires an associated define_split to exist, and that
29228         the define_split must be suitable for use after register allocation.
29230 2017-06-06  Jan Hubicka  <hubicka@ucw.cz>
29232         * cfgbuild.c (find_bb_boundaries): Initialize profile of split blocks.
29233         (compute_outgoing_frequencies): Also initialize zero counts.
29234         (find_many_sub_basic_blocks): Do not produce uninitialized profile
29235         around loops; preserve more of profile when nothing changes.
29237 2017-06-06  Jim Wilson  <jim.wilson@linaro.org>
29239         * config/aarch64/aarch64-cost-tables.h (qdf24xx_extra_costs): Move to
29240         here.
29241         * config/arm/aarch-cost-tables.h (qdf24xx_extra_costs): From here.
29242         * config/arm/arm-cpu-cdata.h: Regenerate.
29243         * config/arm/arm-cpu-data.h, config/arm/arm-cpu.h: Likewise.
29244         * config/arm/arm-tables.opt, config/arm/arm-tune.md: Likewise.
29245         * config/arm/arm-cpus.in: Delete falkor and qdf24xx entries.
29246         * config/arm/arm.c (arm_qdf24xx_tune): Delete.
29247         * config/arm/bpabi.h (BE8_LINK_SPEC): Delete falkor and qdf24xx
29248         support.
29249         * config/arm/t-aprofile (MULTILIB_MATCHES): Delete falkor and qdf24xx
29250         support.
29251         * config/arm/t-rmprofile: Likewise.
29252         * doc/invoke.texi (ARM Options): Drop falkor and qdf24xx support.
29254 2017-06-06  David S. Miller  <davem@davemloft.net>
29256         PR target/80968
29257         * config/sparc/sparc.c (sparc_expand_prologue): Emit frame
29258         blockage if function uses alloca.
29260 2017-06-06  Maxim Kuvyrkov  <maxim.kuvyrkov@linaro.org>
29262         * tree-ssa-loop-prefetch.c (struct mem_ref_group, struct mem_ref):
29263         New "uid" fields to hold pretty-print IDs of group and ref.
29264         Memory references are now identified as <group_id>:<ref_id>
29265         instead of using [random] addresses.
29266         (dump_mem_details): Simplify, no functional change.
29267         (dump_mem_ref): Simplify and make output more concise.
29268         Replace couple of fprintf's throughout code with calls to dump_mem_ref.
29269         (find_or_create_group): Initialize group uid.
29270         (record_ref): Initialize ref uid.  Improve debug output.
29271         (prune_group_by_reuse, should_issue_prefetch_p,)
29272         (should_issue_prefetch_p, schedule_prefetches, issue_prefetch_ref,)
29273         (mark_nontemporal_store, determine_loop_nest_reuse):
29274         Improve debug output.
29276 2017-06-06  Maxim Kuvyrkov  <maxim.kuvyrkov@linaro.org>
29278         * dbgcnt.def (prefetch): New debug counter.
29279         * tree-ssa-loop-prefetch.c (dbgcnt.h): New include.
29280         (schedule_prefetches): Stop issueing prefetches if debug counter
29281         tripped.
29283 2017-06-06  Tom de Vries  <tom@codesourcery.com>
29285         * doc/sourcebuild.texi (Testsuites, C Language Testsuites,
29286         gcc.c-torture/compile): Remove mention of NO_LABEL_VALUES in fixme.
29288 2017-06-06  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
29290         * config/aarch64/atomics.md (atomic_compare_and_swap<mode> expander):
29291         Use aarch64_reg_or_zero predicate for operand 4.
29292         (aarch64_compare_and_swap<mode> define_insn_and_split):
29293         Use aarch64_reg_or_zero predicate for operand 3.  Add 'Z' constraint.
29294         (aarch64_store_exclusive<mode>): Likewise for operand 2.
29296 2017-06-06  Thomas Preud'homme  <thomas.preudhomme@arm.com>
29298         * config/arm/arm.c (arm_compute_save_reg_mask): Rename into ...
29299         (arm_compute_save_core_reg_mask): This.
29300         (thumb1_compute_save_reg_mask): Rename into ...
29301         (thumb1_compute_save_core_reg_mask): This.
29302         (arm_compute_save_reg0_reg12_mask): Adapt comment.
29303         (arm_compute_frame_layout): Likewise.
29305 2017-06-06  Richard Biener  <rguenther@suse.de>
29307         PR tree-optimization/80974
29308         * tree-ssa-sccvn.c (set_ssa_val_to): Do not change but only
29309         keep or clear leaders SSA info.
29311 2017-06-06  Tom de Vries  <tom@codesourcery.com>
29313         * config/nvptx/nvptx.c (split_mode_p): New function.
29314         (nvptx_declare_function_name, nvptx_print_operand): Use split_mode_p.
29316 2017-06-06  Tom de Vries  <tom@codesourcery.com>
29318         * config/nvptx/nvptx.c (nvptx_print_operand): Use maybe_split_mode.
29320 2017-06-06  Jan Hubicka  <hubicka@ucw.cz>
29322         PR bootstrap/80978
29323         * tree-cfg.c (execute_fixup_cfg): Fix condition on when to rescale
29324         profile.
29326 2017-06-06  Jan Hubicka  <hubicka@ucw.cz>
29328         * shrink-wrap.c (handle_simple_exit): Update profile.
29329         (try_shrink_wrapping): Upate profile.
29331 2017-06-06  Jan Hubicka  <hubicka@ucw.cz>
29333         * predict.c (tree_estimate_probability_bb): Add LOCAL_ONLY.
29334         (tree_guess_outgoing_edge_probabilities): New.
29335         * predict.h (tree_guess_outgoing_edge_probabilities): Declare.
29336         * tree-cfg.c (gimple_find_sub_bbs): Propagate profile.
29338 2017-06-06  Jan Hubicka  <hubicka@ucw.cz>
29340         * ipa-split.c (split_function): Initialize return bb profile.
29342 2017-06-06  Jan Hubicka  <hubicka@ucw.cz>
29344         * profile.c (compute_branch_probabilities): Also initialize
29345         EXIT_BLOCK profile.
29347 2017-06-06  Richard Biener  <rguenther@suse.de>
29349         PR tree-optimization/80928
29350         * tree-vect-loop.c (vect_update_vf_for_slp): Amend dumps.
29351         (vect_analyze_loop_operations): Properly guard analysis for
29352         pure SLP case.
29353         (vect_transform_loop): Likewise.
29354         (vect_analyze_loop_2): Also reset SLP type on PHIs.
29355         (vect_model_induction_cost): Do not cost for pure SLP.
29356         (vectorizable_induction): Pass in SLP node, implement SLP vectorization
29357         of induction in inner loop vectorization.
29358         * tree-vect-slp.c (vect_create_new_slp_node): Handle PHIs.
29359         (vect_get_and_check_slp_defs): Handle vect_induction_def.
29360         (vect_build_slp_tree): Likewise.  Handle PHIs as terminating the
29361         recursion.
29362         (vect_analyze_slp_cost_1): Cost induction.
29363         (vect_detect_hybrid_slp_stmts): Handle PHIs.
29364         (vect_get_slp_vect_defs): Likewise.
29365         * tree-vect-stmts.c (vect_analyze_stmt): Handle induction.
29366         (vect_transform_stmt): Handle SLP reductions.
29367         * tree-vectorizer.h (vectorizable_induction): Adjust.
29369 2017-06-05  Michael Meissner  <meissner@linux.vnet.ibm.com>
29371         * config/rs6000/rs6000.c (make_resolver_func): Update
29372         init_lowered_empty_function call.
29374 2017-06-05  Bernd Edlinger  <bernd.edlinger@hotmail.de>
29376         * doc/invoke.texi: Document the -fprofile-abs-path option.
29377         * common.opt (fprofile-abs-path): New option.
29378         * gcov-io.h (gcov_write_filename): Declare.
29379         * gcov-io.c (gcov_write_filename): New function.
29380         * coverage.c (coverage_begin_function): Use gcov_write_filename.
29381         * profile.c (output_location): Likewise.
29383 2017-06-05  Jan Hubicka  <hubicka@ucw.cz>
29385         * shring-wrap.c: Revert accidental commit.
29387 2017-06-05  Volker Reichelt  <v.reichelt@netcologne.de>
29389         * doc/invoke.texi (-Wduplicated-branches): Add to warning list.
29391 2017-06-05  Jan Hubicka  <hubicka@ucw.cz>
29393         * cfgexpand.c (expand_gimple_tailcall): Initialize profile of
29394         new edge.
29395         * ipa-inline.c (want_inline_self_recursive_call_p): Watch for missing
29396         profile in callgraph edge.
29397         * profile-count.h (apply_probability): If THIS is 0, then result is 0
29398         (apply_scale): Likewise.
29399         * tree-inline.c (copy_bb, copy_edges_for_bb, copy_cfg_body):
29400         Also scale profile when inlining function with zero profile.
29401         (initialize_cfun): Update exit block profile even when it is zero.
29402         * tree-ssa-threadupdate.c (clear_counts_path): Handle correctly case
29403         when profile is read.
29405 2017-06-05  Michael Meissner  <meissner@linux.vnet.ibm.com>
29407         * config/rs6000/rs6000.c (toplevel): Include attribs.h.
29408         (CLONE_*): New constants to define the processors we can generate
29409         code for with the target_clone attribute.
29410         (rs6000_clone_map): New array to identify which clone processors
29411         the current program is running on.
29412         (TARGET_COMPARE_VERSION_PRIORITY): Define to enable the
29413         target_clone attribute.
29414         (TARGET_GENERATE_VERSION_DISPATCHER_BODY): Likewise.
29415         (TARGET_GET_FUNCTION_VERSIONS_DISPATCHER): Likewise.
29416         (TARGET_OPTION_FUNCTION_VERSIONS): Likewise.
29417         (cpu_expand_builtin): Add support for target_clone attribute.
29418         (rs6000_valid_attribute_p): Allow "default" attribute.
29419         (get_decl_name): New debug function to simplify printing the
29420         current function name in debugging statements.
29421         (rs6000_clone_priority): New functions to support the target_clone
29422         attribute, and be able to generate code to switch between ISA 2.05
29423         through ISA 3.0 (power6 through power9).
29424         (rs6000_compare_version_priority): Likewise.
29425         (rs6000_get_function_versions_dispatcher): Likewise.
29426         (make_resolver_func): Likewise.
29427         (add_condition_to_bb): Likewise.
29428         (dispatch_function_versions): Likewise.
29429         (rs6000_generate_version_dispatcher_body): Likewise.
29430         (rs6000_can_inline_p): Call get_decl_name for debugging usage.
29431         (fusion_gpr_load_p): Fix a spacing issue.
29432         * doc/extend.texi (Common Function Attributes): Document that the
29433         PowerPC supports the target_clone attribute.
29435 2017-06-05  Thomas Preud'homme  <thomas.preudhomme@arm.com>
29437         * config/arm/arm.h: explain F symbol found in description of ARM
29438         register allocation in its legend.
29440 2017-06-05  Jan Hubicka  <hubicka@ucw.cz>
29442         * config/mips/frame-header-opt.c: Include profile-count.h.
29443         * config/riscv/riscv.c: Include profile-count.h
29445 2017-06-05  Jan Hubicka  <hubicka@ucw.cz>
29447         * tree-ssa-loop-im.c (execute_sm_if_changed): Add FLAG_BBS parameter;
29448         update profile.
29449         (sm_set_flag_if_changed): Add bbs field.
29450         (execute_sm_if_changed_flag_set): Pass BBS.
29451         (execute_sm): Update.
29453 2017-06-05  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
29455         * config/aarch64/aarch64-simd.md (aarch64_store_lane0<mode>):
29456         New pattern.
29458 2017-06-05  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
29460         * config/aarch64/aarch64.md (sub<mode>3_compare1_imm): New define_insn.
29461         (peephole2): New peephole2 to emit the above.
29462         * config/aarch64/predicates.md (aarch64_sub_immediate): New predicate.
29464 2017-06-05  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
29466         * config/aarch64/aarch64.c (define_peephole2 above
29467         *sub_<shift>_<mode>): New peephole.
29469 2017-05-23  Jan Hubicka  <hubicka@ucw.cz>
29471         * config/i386/i386.c (make_resolver_func): Update.
29472         * Makefile.in: Add profile-count.h and profile-count.o
29473         * auto-profile.c (afdo_indirect_call): Update to new API.
29474         (afdo_set_bb_count): Update.
29475         (afdo_propagate_edge): Update.
29476         (afdo_propagate_circuit): Update.
29477         (afdo_calculate_branch_prob): Update.
29478         (afdo_annotate_cfg): Update.
29479         * basic-block.h: Include profile-count.h
29480         (struct edge_def): Turn count to profile_count.
29481         (struct basic_block_def): Likewie.
29482         (REG_BR_PROB_BASE): Move to profile-count.h
29483         (RDIV): Move to profile-count.h
29484         * bb-reorder.c (max_entry_count): Turn to profile_count.
29485         (find_traces): Update.
29486         (rotate_loop):Update.
29487         (connect_traces):Update.
29488         (sanitize_hot_paths):Update.
29489         * bt-load.c (migrate_btr_defs): Update.
29490         * cfg.c (RDIV): Remove.
29491         (init_flow): Use alloc_block.
29492         (alloc_block): Uninitialize count.
29493         (unchecked_make_edge): Uninitialize count.
29494         (check_bb_profile): Update.
29495         (dump_edge_info): Update.
29496         (dump_bb_info): Update.
29497         (update_bb_profile_for_threading): Update.
29498         (scale_bbs_frequencies_int): Update.
29499         (scale_bbs_frequencies_gcov_type): Update.
29500         (scale_bbs_frequencies_profile_count): New.
29501         * cfg.h (update_bb_profile_for_threading): Update.
29502         (scale_bbs_frequencies_profile_count): Declare.
29503         * cfgbuild.c (compute_outgoing_frequencies): Update.
29504         (find_many_sub_basic_blocks): Update.
29505         * cfgcleanup.c (try_forward_edges): Update.
29506         (try_crossjump_to_edge): Update.
29507         * cfgexpand.c (expand_gimple_tailcall): Update.
29508         (construct_exit_block): Update.
29509         * cfghooks.c (verify_flow_info): Update.
29510         (dump_bb_for_graph): Update.
29511         (split_edge): Update.
29512         (make_forwarder_block): Update.
29513         (duplicate_block): Update.
29514         (account_profile_record): Update.
29515         * cfgloop.c (find_subloop_latch_edge_by_profile): Update.
29516         (get_estimated_loop_iterations): Update.
29517         * cfgloopanal.c (expected_loop_iterations_unbounded): Update.
29518         (single_likely_exit): Update.
29519         * cfgloopmanip.c (scale_loop_profile): Update.
29520         (loopify): Update.
29521         (set_zero_probability): Update.
29522         (lv_adjust_loop_entry_edge): Update.
29523         * cfgrtl.c (force_nonfallthru_and_redirect): Update.
29524         (purge_dead_edges): Update.
29525         (rtl_account_profile_record): Update.
29526         * cgraph.c (cgraph_node::create): Uninitialize count.
29527         (symbol_table::create_edge): Uninitialize count.
29528         (cgraph_update_edges_for_call_stmt_node): Update.
29529         (cgraph_edge::dump_edge_flags): Update.
29530         (cgraph_node::dump): Update.
29531         (cgraph_edge::maybe_hot_p): Update.
29532         * cgraph.h: Include profile-count.h
29533         (create_clone), create_edge, create_indirect_edge): Update.
29534         (cgraph_node): Turn count to profile_count.
29535         (cgraph_edge0: Likewise.
29536         (make_speculative, clone): Update.
29537         (create_edge): Update.
29538         (init_lowered_empty_function): Update.
29539         * cgraphclones.c (cgraph_edge::clone): Update.
29540         (duplicate_thunk_for_node): Update.
29541         (cgraph_node::create_clone): Update.
29542         * cgraphunit.c (cgraph_node::analyze): Update.
29543         (cgraph_node::expand_thunk): Update.
29544         * final.c (dump_basic_block_info): Update.
29545         * gimple-streamer-in.c (input_bb): Update.
29546         * gimple-streamer-out.c (output_bb): Update.
29547         * graphite.c (print_global_statistics): Update.
29548         (print_graphite_scop_statistics): Update.
29549         * hsa-brig.c: Include basic-block.h.
29550         * hsa-dump.c: Include basic-block.h.
29551         * hsa-gen.c (T sum_slice): Update.
29552         (convert_switch_statements):Update.
29553         * hsa-regalloc.c: Include basic-block.h.
29554         * ipa-chkp.c (chkp_produce_thunks): Update.
29555         * ipa-cp.c (struct caller_statistics): Update.
29556         (init_caller_stats): Update.
29557         (gather_caller_stats): Update.
29558         (ipcp_cloning_candidate_p): Update.
29559         (good_cloning_opportunity_p): Update.
29560         (get_info_about_necessary_edges): Update.
29561         (dump_profile_updates): Update.
29562         (update_profiling_info): Update.
29563         (update_specialized_profile): Update.
29564         (perhaps_add_new_callers): Update.
29565         (decide_about_value): Update.
29566         (ipa_cp_c_finalize): Update.
29567         * ipa-devirt.c (struct odr_type_warn_count): Update.
29568         (struct decl_warn_count): Update.
29569         (struct final_warning_record): Update.
29570         (possible_polymorphic_call_targets): Update.
29571         (ipa_devirt): Update.
29572         * ipa-fnsummary.c (redirect_to_unreachable): Update.
29573         * ipa-icf.c (sem_function::merge): Update.
29574         * ipa-inline-analysis.c (do_estimate_edge_time): Update.
29575         * ipa-inline.c (compute_uninlined_call_time): Update.
29576         (compute_inlined_call_time): Update.
29577         (want_inline_small_function_p): Update.
29578         (want_inline_self_recursive_call_p): Update.
29579         (edge_badness): Update.
29580         (lookup_recursive_calls): Update.
29581         (recursive_inlining): Update.
29582         (inline_small_functions): Update.
29583         (dump_overall_stats): Update.
29584         (dump_inline_stats): Update.
29585         * ipa-profile.c (ipa_profile_generate_summary): Update.
29586         (ipa_propagate_frequency): Update.
29587         (ipa_profile): Update.
29588         * ipa-prop.c (ipa_make_edge_direct_to_target): Update.
29589         * ipa-utils.c (ipa_merge_profiles): Update.
29590         * loop-doloop.c (doloop_modify): Update.
29591         * loop-unroll.c (report_unroll): Update.
29592         (unroll_loop_runtime_iterations): Update.
29593         * lto-cgraph.c (lto_output_edge): Update.
29594         (lto_output_node): Update.
29595         (input_node): Update.
29596         (input_edge): Update.
29597         (merge_profile_summaries): Update.
29598         * lto-streamer-in.c (input_cfg): Update.
29599         * lto-streamer-out.c (output_cfg): Update.
29600         * mcf.c (create_fixup_graph): Update.
29601         (adjust_cfg_counts): Update.
29602         (sum_edge_counts): Update.
29603         * modulo-sched.c (sms_schedule): Update.
29604         * postreload-gcse.c (eliminate_partially_redundant_load): Update.
29605         * predict.c (maybe_hot_count_p): Update.
29606         (probably_never_executed): Update.
29607         (dump_prediction): Update.
29608         (combine_predictions_for_bb): Update.
29609         (propagate_freq): Update.
29610         (handle_missing_profiles): Update.
29611         (counts_to_freqs): Update.
29612         (rebuild_frequencies): Update.
29613         (force_edge_cold): Update.
29614         * predict.h: Include profile-count.h
29615         (maybe_hot_count_p, counts_to_freqs): UPdate.
29616         * print-rtl-function.c: Do not include cfg.h
29617         * print-rtl.c: Include basic-block.h
29618         * profile-count.c: New file.
29619         * profile-count.h: New file.
29620         * profile.c (is_edge_inconsistent): Update.
29621         (correct_negative_edge_counts): Update.
29622         (is_inconsistent): Update.
29623         (set_bb_counts): Update.
29624         (read_profile_edge_counts): Update.
29625         (compute_frequency_overlap): Update.
29626         (compute_branch_probabilities): Update; Initialize and deinitialize
29627         gcov_count tables.
29628         (branch_prob): Update.
29629         * profile.h (bb_gcov_counts, edge_gcov_counts): New.
29630         (edge_gcov_count): New.
29631         (bb_gcov_count): New.
29632         * shrink-wrap.c (try_shrink_wrapping): Update.
29633         * tracer.c (better_p): Update.
29634         * trans-mem.c (expand_transaction): Update.
29635         (ipa_tm_insert_irr_call): Update.
29636         (ipa_tm_insert_gettmclone_call): Update.
29637         * tree-call-cdce.c: Update.
29638         * tree-cfg.c (gimple_duplicate_sese_region): Update.
29639         (gimple_duplicate_sese_tail): Update.
29640         (gimple_account_profile_record): Update.
29641         (execute_fixup_cfg): Update.
29642         * tree-inline.c (copy_bb): Update.
29643         (copy_edges_for_bb): Update.
29644         (initialize_cfun): Update.
29645         (freqs_to_counts): Update.
29646         (copy_cfg_body): Update.
29647         (expand_call_inline): Update.
29648         * tree-ssa-ifcombine.c (update_profile_after_ifcombine): Update.
29649         * tree-ssa-loop-ivcanon.c (unloop_loops): Update.
29650         (try_unroll_loop_completely): Update.
29651         (try_peel_loop): Update.
29652         * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Update.
29653         * tree-ssa-loop-niter.c (estimate_numbers_of_iterations_loop): Update.
29654         * tree-ssa-loop-split.c (connect_loops): Update.
29655         * tree-ssa-loop-unswitch.c (hoist_guard): Update.
29656         * tree-ssa-reassoc.c (branch_fixup): Update.
29657         * tree-ssa-tail-merge.c (replace_block_by): Update.
29658         * tree-ssa-threadupdate.c (create_block_for_threading): Update.
29659         (compute_path_counts): Update.
29660         (update_profile): Update.
29661         (recompute_probabilities): Update.
29662         (update_joiner_offpath_counts): Update.
29663         (estimated_freqs_path): Update.
29664         (freqs_to_counts_path): Update.
29665         (clear_counts_path): Update.
29666         (ssa_fix_duplicate_block_edges): Update.
29667         (duplicate_thread_path): Update.
29668         * tree-switch-conversion.c (case_bit_test_cmp): Update.
29669         (struct switch_conv_info): Update.
29670         * tree-tailcall.c (decrease_profile): Update.
29671         * tree-vect-loop-manip.c (slpeel_add_loop_guard): Update.
29672         * tree-vect-loop.c (scale_profile_for_vect_loop): Update.
29673         * value-prof.c (check_counter): Update.
29674         (gimple_divmod_fixed_value): Update.
29675         (gimple_mod_pow2): Update.
29676         (gimple_mod_subtract): Update.
29677         (gimple_ic_transform): Update.
29678         (gimple_stringop_fixed_value): Update.
29679         * value-prof.h (gimple_ic): Update.
29681 2017-06-02  Carl Love  <cel@us.ibm.com>
29683         * config/rs6000/rs6000-c: Add support for built-in functions
29684         vector double vec_doublee (vector signed int);
29685         vector double vec_doublee (vector unsigned int);
29686         vector double vec_doublee (vector float);
29687         vector double vec_doubleh (vector signed int);
29688         vector double vec_doubleh (vector unsigned int);
29689         vector double vec_doubleh (vector float);
29690         vector double vec_doublel (vector signed int);
29691         vector double vec_doublel (vector unsigned int);
29692         vector double vec_doublel (vector float);
29693         vector double vec_doubleo (vector signed int);
29694         vector double vec_doubleo (vector unsigned int);
29695         vector double vec_doubleo (vector float);.
29696         * config/rs6000/rs6000-builtin.def: Add definitions for DOUBLEE,
29697         DOUBLEO, DOUBLEH, DOUBLEL, UNS_DOUBLEO, UNS_DOUBLEE, UNS_DOUBLEH,
29698         UNS_DOUBLEL.
29699         * config/rs6000/altivec.md: Add code generator for doublee<mode>2,
29700         unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2, doubleh<mode>2,
29701         unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2, add mode attribute
29702         VS_sxwsp.
29703         * config/rs6000/altivec.h: Add define for vec_doublee, vec_doubleo,
29704         vec_doublel, vec_doubleh.
29705         * doc/extend.texi: Update the built-in documentation file for the
29706         new built-in functions.
29708 2017-06-02  David Malcolm  <dmalcolm@redhat.com>
29710         PR jit/80954
29711         * ipa-inline-analysis.c (free_growth_caches): Set
29712         edge_removal_hook_holder to NULL after removing it.
29714 2017-06-02  Sudakshina Das  <sudi.das@arm.com>
29716         * config/aarch64/aarch64.c (aarch64_select_cc_mode): Return CC_SWP for
29717         comparision with zero.
29719 2017-06-02  Will Schmidt  <will_schmidt@vnet.ibm.com>
29720         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add handling
29721         for early expansion of vec_min and vec_max builtins.
29722         (builtin_function_type): Add min/max unsigned variants to those
29723         identified as having unsigned arguments.
29725 2017-06-02  Olivier Hainque  <hainque@adacore.com>
29727         * config/vx-common.h (DWARF_UNWIND_INFO): Switch #define to 1.
29729 2017-06-02  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
29731         * config/aarch64/aarch64-simd.md (*aarch64_simd_vec_copy_lane<mode>):
29732         Use VALL_F16 iterator rather than VALL.
29734 2017-06-02  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
29736         * config/aarch64/aarch64.c (aarch64_split_compare_and_swap):
29737         Emit CBNZ inside loop when doing a strong exchange and comparing
29738         against zero.  Generate the CC flags after the loop.
29740 2017-06-02  David Edelsohn  <dje.gcc@gmail.com>
29742         * dwarf2out.c (DWARF_INITIAL_LENGTH_SIZE_STR): New.
29743         (dl_section_ref): New.
29744         (dwarf2out_finish): Copy debug_line_section_label to dl_section_ref.
29745         On AIX, append an expression to subtract the size of the
29746         section length to dl_section_ref.
29748 2017-06-02  Will Schmidt  <will_schmidt@vnet.ibm.com>
29750         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add handling
29751         for early expansion of vector absolute builtins.
29753 2017-06-02  Richard Biener  <rguenther@suse.de>
29755         * tree-vect-slp.c (vect_detect_hybrid_slp_2): Match up
29756         what we consider a relevant use stmt with vect_detect_hybrid_slp_stmts.
29758 2017-06-02  Richard Biener  <rguenther@suse.de>
29760         PR tree-optimization/80948
29761         * tree-tailcall.c (find_tail_calls): Track stmts to move in
29762         stmt order as well.
29764 2017-06-02  Richard Biener  <rguenther@suse.de>
29766         * tree-vect-loop.c (vect_analyze_loop_operations): Not relevant
29767         PHIs are ok.
29768         * tree-vect-stmts.c (process_use): Do not mark backedge defs
29769         for inductions as relevant.
29771 2017-06-02  Richard Biener  <rguenther@suse.de>
29773         * tree-vect-loop.c (get_initial_def_for_induction): Inline into ...
29774         (vectorizable_induction): ... this.  Remove dead code.
29776 2017-06-02  Eric Botcazou  <ebotcazou@adacore.com>
29778         * builtins. (expand_builtin_alloca): Remove second parameter and
29779         infer its value from the first parameter instead.
29780         (expand_builtin) <BUILT_IN_ALLOCA>: Adjust call to above.
29782 2017-06-02  Jakub Jelinek  <jakub@redhat.com>
29784         PR rtl-optimization/80903
29785         * loop-doloop.c (add_test): Unshare sequence.
29787 2017-06-02  Bernd Edlinger  <bernd.edlinger@hotmail.de>
29789         * doc/invoke.texi: Document the -Wsizeof-pointer-div warning.
29791 2017-06-01  Bernd Edlinger  <bernd.edlinger@hotmail.de>
29793         * config/i386/i386.c (x86_64_ms_sysv_extra_clobbered_registers): Make
29794         static.
29795         (xlogue_layout::get_stack_space_used, xlogue_layout::s_instances,
29796         xlogue_layout::get_instance, logue_layout::xlogue_layout,
29797         sp_valid_at, fp_valid_at, choose_basereg): Formatting.
29798         (xlogue_layout::get_stub_rtx): Make static.
29799         (xlogue_layout::get_stub_name): Avoid const-cast, make static.
29800         (xlogue_layout::compute_stub_managed_regs): Rename to...
29801         (xlogue_layout::count_stub_managed_regs): ...this.
29802         (xlogue_layout::is_stub_managed_reg): New function.
29803         (xlogue_layout::m_stub_names): Rename to...
29804         (xlogue_layout::s_stub_names): ...this, make static.
29805         (xlogue_layout::STUB_INDEX_OFFSET, xlogue_layout::MIN_REGS,
29806         xlogue_layout::MAX_REGS, xlogue_layout::MAX_EXTRA_REGS,
29807         xlogue_layout::VARIANT_COUNT, xlogue_layout::STUB_NAME_MAX_LEN,
29808         xlogue_layout::s_stub_names): Instantiate statics.
29809         (stub_managed_regs): Remove.
29810         (ix86_save_reg): Use xlogue_layout::compute_stub_managed_regs.
29811         (disable_call_ms2sysv_xlogues): Rename to...
29812         (warn_once_call_ms2sysv_xlogues): ...this, and warn only once.
29813         (ix86_initial_elimination_offset, ix86_expand_call): Fix call_ms2sysv
29814         warning logic.
29815         (ix86_static_chain): Make sure that ix86_static_chain_on_stack can't
29816         change after reload_completed.
29817         (ix86_can_use_return_insn_p): Use the ix86_frame data structure
29818         directly.
29819         (ix86_expand_prologue): Likewise.
29820         (ix86_expand_epilogue): Likewise.
29821         (ix86_expand_split_stack_prologue): Likewise.
29822         (ix86_compute_frame_layout): Remove frame parameter ...
29823         (TARGET_COMPUTE_FRAME_LAYOUT): ... and export it as a target hook.
29824         (ix86_finalize_stack_realign_flags): Call ix86_compute_frame_layout
29825         only if necessary.
29826         (ix86_init_machine_status): Don't set use_fast_prologue_epilogue_nregs.
29827         (ix86_frame): Move from here ...
29828         * config/i386/i386.h (ix86_frame): ... to here.
29829         (machine_function): Remove use_fast_prologue_epilogue_nregs, cache the
29830         complete ix86_frame data structure instead.  Remove some_ld_name.
29832 2017-06-01  Pierre-Marie de Rodat  <derodat@adacore.com>
29834         * dwarf2out.c (dwarf2out_late_global_decl): Add locations for
29835         symbols that hold a DECL_VALUE_EXPR.
29837 2017-06-01  Martin Jambor  <mjambor@suse.cz>
29839         PR tree-optimization/80898
29840         * tree-sra.c (process_subtree_disqualification): Removed.
29841         (disqualify_candidate): Do not acll
29842         process_subtree_disqualification.
29843         (subtree_mark_written_and_enqueue): New function.
29844         (propagate_all_subaccesses): Set grp_write of LHS subtree if the
29845         RHS has been disqualified and re-queue LHS if necessary.  Apart
29846         from that, ignore disqualified RHS.
29848 2017-06-01  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
29850         * config/s390/s390.c (s390_emit_epilogue): Disable early return
29851         address fetch for z10 or later.
29853 2017-06-01  Claudiu Zissulescu  <claziss@synopsys.com>
29855         * config/arc/arc.md (tst_movb): Add guard when splitting.
29857 2017-06-01  Claudiu Zissulescu  <claziss@synopsys.com>
29859         * config/arc/arc.c (arc_can_eliminate): Test against
29860         arc_frame_pointer_needed.
29862 2017-06-01  Claudiu Zissulescu  <claziss@synopsys.com>
29864         * config/arc/arc.c (arc_expand_prologue): Emit a special barrier
29865         to prevent store reordering.
29866         * config/arc/arc.md (UNSPEC_ARC_STKTIE): Define.
29867         (type): Add block type.
29868         (stack_tie): Define special instruction to be used in
29869         expand_prologue.
29871 2017-06-01  Claudiu Zissulescu  <claziss@synopsys.com>
29873         * config/arc/arc.md (commutative_binary_comparison): Remove 'I'
29874         constraint. It is not valid for the pattern.
29875         (noncommutative_binary_comparison): Likewise.
29877 2017-06-01  Claudiu Zissulescu  <claziss@synopsys.com>
29879         * config/arc/simdext.md (movv2hi_insn): Change predicate to avoid
29880         scaled addresses.
29882 2017-06-01  Claudiu Zissulescu  <claziss@synopsys.com>
29884         * config/arc/arc.c (arc_conditional_register_usage): Allow r30 to
29885         be used by the reg-alloc.
29887 2017-06-01  Claudiu Zissulescu  <claziss@synopsys.com>
29889         * config/arc/arc.md (mulsi3): Avoid use of hard registers before
29890         reg-alloc when having mul64 or mul32x16 instructions.
29891         (mulsidi3): Likewise.
29892         (umulsidi3): Likewise.
29893         (mulsi32x16): New pattern.
29894         (mulsi64): Likewise.
29895         (mulsidi64): Likewise.
29896         (umulsidi64): Likewise.
29897         (MUL32x16_REG): Define.
29898         (mul64_600): Use MUL32x16_REG.
29899         (mac64_600): Likewise.
29900         (umul64_600): Likewise.
29901         (umac64_600): Likewise.
29903 2017-06-01  Claudiu Zissulescu  <claziss@synopsys.com>
29905         * config/arc/arc.md (mulsi3_700): Make it commutative.
29907 2017-06-01  Jose E. Marchesi  <jose.marchesi@oracle.com>
29909         * config/sparc/sparc.md (*zero_extendsidi2_insn_sp64): Set insn
29910         type for movstouw.
29911         (*sign_extendsidi2_insn): Likewise for movstosw.
29913 2017-06-01  Pierre-Marie de Rodat  <derodat@adacore.com>
29915         * dwarf2out.c (get_discr_value): Call the get_debug_type hook on
29916         the type of the input discriminant value.  Convert the
29917         discriminant value of signedness vary.
29919 2017-06-01  Volker Reichelt  <v.reichelt@netcologne.de>
29921         * doc/invoke.texi (-Wcatch-value): Document new shortcut.
29922         Add to -Wall section.
29924 2017-06-01  Richard Biener  <rguenther@suse.de>
29926         PR middle-end/66313
29927         * fold-const.c (fold_plusminus_mult_expr): If the factored
29928         factor may be zero use a wrapping type for the inner operation.
29929         * tree-tailcall.c (independent_of_stmt_p): Pass in to_move bitmap
29930         and handle moved defs.
29931         (process_assignment): Properly guard the unary op case.  Return a
29932         tri-state indicating that moving the stmt before the call may allow
29933         to continue.  Pass through to_move.
29934         (find_tail_calls): Handle moving unrelated defs before
29935         the call.
29937 2017-05-31  Segher Boessenkool  <segher@kernel.crashing.org>
29939         PR target/80618
29940         * config/rs6000/vector.md (*vector_uneq<mode>): Write the nor in the
29941         splitter result in the canonical way.
29943 2017-05-31  Uros Bizjak  <ubizjak@gmail.com>
29945         * config/i386/i386.md (*zero_extendsidi2): Enable alternative (?r, *Yj)
29946         also for 32bit target.  Update insn attributes.
29947         (zero-extendsidi2 splitter): Allow all registers for operand 1.
29949 2017-05-31  Sebastian Peryt  <sebastian.peryt@intel.com>
29951         * config/i386/avx512fintrin.h (_mm_mask_max_sd)
29952         (_mm_maskz_max_sd, _mm_mask_max_ss, _mm_maskz_max_ss)
29953         (_mm_mask_min_sd, _mm_maskz_min_sd, _mm_mask_min_ss)
29954         (_mm_maskz_min_ss): New intrinsics.
29956 2017-05-31  Martin Liska  <mliska@suse.cz>
29958         * tree-vect-loop.c (vect_create_epilog_for_reduction):
29959         Change comment style to one we normally use.
29960         (vectorizable_reduction): Likewise.
29961         (vectorizable_induction): Likewise.
29962         * tree-vect-stmts.c (vectorizable_mask_load_store): Likewise.
29963         (vectorizable_call): Likewise.
29964         (vectorizable_simd_clone_call): Likewise.
29965         (vectorizable_conversion): Likewise.
29966         (vectorizable_assignment): Likewise.
29967         (vectorizable_shift): Likewise.
29968         (vectorizable_operation): Likewise.
29969         (vectorizable_store): Likewise.
29970         (vectorizable_load): Likewise.
29971         * tree-vectorizer.h: Likewise.
29973 2017-05-31  Alexander Monakov  <amonakov@ispras.ru>
29975         * passes.c (emergency_dump_function): New.
29976         * tree-pass.h (emergency_dump_function): Declare.
29977         * plugin.c (plugins_internal_error_function): Remove.
29978         * plugin.h (plugins_internal_error_function): Remove declaration.
29979         * toplev.c (internal_error_function): New static function.  Use it...
29980         (general_init): ...here.
29982 2017-05-31  Graham Markall  <graham.markall@embecosm.com>
29984         * config/arc/arc.c (arc_print_operand): Handle constant operands.
29985         (arc_rtx_costs): Add costs for new patterns.
29986         * config/arc/arc.md: Additional *add_n and *sub_n patterns.
29987         * config/arc/predicates.md: Add _1_2_3_operand predicate.
29989 2017-05-31  Richard Sandiford  <richard.sandiford@linaro.org>
29991         * tree-ssa-strlen.c (get_next_strinfo): New function.
29992         (get_stridx_plus_constant): Use it.
29993         (zero_length_string): Likewise.
29994         (adjust_related_strinfos): Likewise.
29995         (adjust_last_stmt): Likewise.
29997 2017-05-31  Richard Biener  <rguenther@suse.de>
29999         PR target/80880
30000         * config/i386/i386.c (ix86_expand_builtin): Remove assert
30001         for arg being an SSA name when expanding IX86_BUILTIN_BNDRET.
30003 2017-05-31  Richard Sandiford  <richard.sandiford@linaro.org>
30005         * tree-vect-data-refs.c (vect_find_same_alignment_drs): Remove
30006         loop_vinfo argument and use of dependence distance vectors.
30007         Check instead whether the two references differ only in their
30008         initial value and assume that they have the same alignment if the
30009         difference is a multiple of the vector alignment.
30010         (vect_analyze_data_refs_alignment): Update call accordingly.
30012 2017-05-31  Martin Liska  <mliska@suse.cz>
30014         PR target/79155
30015         * config/i386/cpuid.h: Fix typo in a comment in cpuid.h.
30017 2017-05-31  Bin Cheng  <bin.cheng@arm.com>
30019         * tree-vect-loop-manip.c (create_intersect_range_checks_index)
30020         (create_intersect_range_checks): Move from ...
30021         * tree-data-ref.c (create_intersect_range_checks_index)
30022         (create_intersect_range_checks): ... to here.
30023         (create_runtime_alias_checks): New function factored from ...
30024         * tree-vect-loop-manip.c (vect_create_cond_for_alias_checks): ...
30025         here.  Call above function.
30026         * tree-data-ref.h (create_runtime_alias_checks): New function.
30028 2017-05-31  Bin Cheng  <bin.cheng@arm.com>
30030         * tree-data-ref.c (prune_runtime_alias_test_list): Relax minimal
30031         segment length for dr_b and compute it in wide_int.
30033 2017-05-31  Richard Biener  <rguenther@suse.de>
30035         PR tree-optimization/80906
30036         * graphite-isl-ast-to-gimple.c (copy_loop_close_phi_nodes): Get
30037         and pass through iv_map.
30038         (copy_bb_and_scalar_dependences): Adjust.
30039         (translate_pending_phi_nodes): Likewise.
30040         (copy_loop_close_phi_args): Handle code-generating IVs instead
30041         of ICEing.
30043 2017-05-30  David Malcolm  <dmalcolm@redhat.com>
30045         * diagnostic-color.c (color_dict): Add "type-diff".
30046         (parse_gcc_colors): Update comment.
30047         * doc/invoke.texi (Diagnostic Message Formatting Options): Add
30048         -fdiagnostics-show-template-tree and -fno-elide-type.
30049         (GCC_COLORS): Add type-diff to example.
30050         (type-diff=): New.
30051         (-fdiagnostics-show-template-tree): New.
30052         (-fno-elide-type): New.
30053         * pretty-print.c (pp_format): Pass quote and formatters[argno] to
30054         the pp_format_decoder callback.  Call any m_format_postprocessor's
30055         "handle" method.
30056         (pretty_printer::pretty_printer): Initialize
30057         m_format_postprocessor.
30058         (pretty_printer::~pretty_printer): Delete any
30059         m_format_postprocessor.
30060         * pretty-print.h (printer_fn): Add bool and const char ** parameters.
30061         (class format_postprocessor): New class.
30062         (struct pretty_printer::format_decoder): Document the new parameters.
30063         (struct pretty_printer::m_format_postprocessor): New field.
30064         * tree-diagnostic.c (default_tree_printer): Update for new
30065         bool and const char ** params.
30066         * tree-diagnostic.h (default_tree_printer): Likewise.
30068 2017-05-30  Segher Boessenkool  <segher@kernel.crashing.org>
30070         * config/rs6000/predicates.md (cc_reg_not_micro_cr0_operand): Delete.
30071         (lwa_operand): Delete rs6000_gen_cell_microcode test.
30072         * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
30073         rs6000_gen_cell_microcode code.
30074         (rs6000_final_prescan_insn): Delete.
30075         (rs6000_opt_vars): Delete the "gen-cell-microcode" and
30076         "warn-cell-microcode" entries.
30077         * config/rs6000/rs6000.h (FINAL_PRESCAN_INSN): Delete.
30078         * config/rs6000/rs6000.md: Delete rs6000_gen_cell_microcode tests
30079         throughout.  Change cc_reg_not_micro_cr0_operand to
30080         cc_reg_not_cr0_operand throughout.
30081         (*extendhi<mode>2_noload): Delete.
30082         * config/rs6000/rs6000.opt (mgen-cell-microcode): Replace by stub.
30083         (mwarn-cell-microcode): Delete.
30084         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete
30085         -mgen-cell-microcode and -mwarn-cell-microcode.
30087 2017-05-30  Uros Bizjak  <ubizjak@gmail.com>
30089         PR target/80833
30090         * config/i386/constraints.md (Yd): New constraint.
30091         (Ye): Ditto.
30092         * config/i386/i386.md (*movti_internal): Add (?r, Ye)
30093         and (?Yd, r) alternatives.  Update insn attributes.
30094         * config/i386/i386.md (*movti_internal): Add (?r, *Ye)
30095         and (?*Yd, r) alternatives.  Update insn attributes.
30096         (double-mode inter-unit splitters): Add new GR<->XMM splitters.
30098 2017-05-30  Pierre-Marie de Rodat  <derodat@adacore.com>
30100         * gimplify.c (gimplify_modify_expr): Don't create a
30101         DECL_DEBUG_EXPR link if *FROM_P does not belong to the current
30102         function.
30104 2017-05-30  Wilco Dijkstra  <wdijkstr@arm.com>
30106         * config/arm/arm-builtins.c (arm_expand_builtin): Remove const.
30108 2017-05-30  Richard Biener  <rguenther@suse.de>
30110         * tree-vectorizer.h (struct _stmt_vec_info): Add reduc_type
30111         and reduc_def fields.
30112         (STMT_VINFO_REDUC_TYPE): New define.
30113         (STMT_VINFO_REDUC_DEF): Likewise.
30114         (vect_force_simple_reduction): Adjust prototype.
30115         * tree-vect-loop.c (vect_analyze_scalar_cycles_1): Adjust.
30116         (vect_is_simple_reduction): Remove check_reduction argument.
30117         (vect_force_simple_reduction): Adjust and set
30118         STMT_VINFO_REDUC_TYPE and STMT_VINFO_REDUC_DEF.
30119         (vectorizable_reduction): Do not re-do reduction analysis
30120         but use STMT_VINFO_REDUC_TYPE and STMT_VINFO_REDUC_DEF.
30121         * tree-parloops.c (gather_scalar_reductions): Adjust.
30123 2017-05-30  Richard Biener  <rguenther@suse.de>
30125         PR middle-end/80901
30126         * cfgexpand.c (expand_gimple_cond): Match up loop fixup with
30127         split_edge code.
30129 2017-05-24  Robin Dapp  <rdapp@linux.vnet.ibm.com>
30131         * tree-vect-data-refs.c (vect_get_peeling_costs_all_drs):
30132         Introduce unknown_misalignment parameter and remove vf.
30133         (vect_peeling_hash_get_lowest_cost):
30134         Pass unknown_misalignment parameter.
30135         (vect_enhance_data_refs_alignment):
30136         Fix unsupportable data ref treatment.
30138 2017-05-30  Robin Dapp  <rdapp@linux.vnet.ibm.com>
30140         * tree-vect-data-refs.c (vect_get_data_access_cost):
30141         Workaround for SLP handling.
30142         (vect_enhance_data_refs_alignment):
30143         Compute costs for doing no peeling at all, compare to the best
30144         peeling costs so far and avoid peeling if cheaper.
30146 2017-05-30  Robin Dapp  <rdapp@linux.vnet.ibm.com>
30148         * tree-vect-data-refs.c (vect_peeling_hash_choose_best_peeling):
30149         Return peeling info and set costs to zero for unlimited cost
30150         model.
30151         (vect_enhance_data_refs_alignment): Also inspect all datarefs
30152         with unknown misalignment. Compute and costs for unknown
30153         misalignment, compare them to the costs for known misalignment
30154         and choose the cheapest for peeling.
30156 2017-05-30  Robin Dapp  <rdapp@linux.vnet.ibm.com>
30158         * tree-vect-data-refs.c (vect_update_misalignment_for_peel): Rename.
30159         (vect_get_peeling_costs_all_drs): Create function.
30160         (vect_peeling_hash_get_lowest_cost):
30161         Use vect_get_peeling_costs_all_drs.
30162         (vect_peeling_supportable): Create function.
30163         (vect_enhance_data_refs_alignment): Use vect_peeling_supportable.
30165 2017-05-30  Robin Dapp  <rdapp@linux.vnet.ibm.com>
30167         * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Create
30168         DR_HAS_NEGATIVE_STEP.
30169         (vect_update_misalignment_for_peel): Define DR_MISALIGNMENT.
30170         (vect_enhance_data_refs_alignment): Use.
30171         (vect_duplicate_ssa_name_ptr_info): Use.
30172         * tree-vectorizer.h (dr_misalignment): Use.
30173         (known_alignment_for_access_p): Use.
30175 2017-05-30  Jozef Lawrynowicz  <jozef.l@somniumtech.com>
30177         PR target/78838
30178         * config/msp430/msp430.c (gen_prefix): Return NULL when section name is
30179         .lowtext.
30180         (has_section_name): New function.
30182 2017-05-30  Martin Liska  <mliska@suse.cz>
30184         PR other/80909
30185         * auto-profile.c (get_function_decl_from_block): Fix
30186         parenthesis.
30188 2017-05-30  Richard Biener  <rguenther@suse.de>
30190         PR middle-end/80876
30191         * cfgexpand.c (expand_gimple_cond): Fixup preserving loops again.
30193 2017-05-30  Martin Liska  <mliska@suse.cz>
30195         * dumpfile.c: Use newly added macro DUMP_FILE_INFO.
30196         * dumpfile.h (struct dump_file_info): Remove ctors.
30198 2017-05-30  Martin Liska  <mliska@suse.cz>
30200         * predict.def: Fix GNU coding style.
30202 2017-05-29  Max Filippov  <jcmvbkbc@gmail.com>
30204         * config/xtensa/xtensa.c (xtensa_initial_elimination_offset):
30205         Mark 'to' argument with ATTRIBUTE_UNUSED.
30207 2017-05-29  Max Filippov  <jcmvbkbc@gmail.com>
30209         * config/xtensa/xtensa.c (xtensa_emit_call): Use
30210         HOST_WIDE_INT_PRINT_HEX instead of 0x%lx format string.
30211         (print_operand): Use HOST_WIDE_INT_PRINT_DEC instead of %ld
30212         format string.
30214 2017-05-29  Eric Botcazou  <ebotcazou@adacore.com>
30216         * doc/install.texi (Options specification): Restore entry of
30217         --enable-sjlj-exceptions.
30219 2017-05-27  Michael Eager  <eager@eagercon.com>
30221         Revert:
30222         2016-01-21  Ajit Agarwal  <ajitkum@xilinx.com>
30224         See https://gcc.gnu.org/ml/gcc/2017-05/msg00221.html.
30226         * config/microblaze/microblaze.h
30227         (FIXED_REGISTERS): Update in macro.
30228         (CALL_USED_REGISTERS): Update in macro.
30230 2017-05-27  François-Xavier Coudett  <fxcoudert@gcc.gnu.org>
30232         * doc/install.texi: Add links to macOS binary distributions.
30234 2017-05-27  Jakub Jelinek  <jakub@redhat.com>
30236         PR bootstrap/80887
30237         Revert:
30238         2017-05-25  Marc Glisse  <marc.glisse@inria.fr>
30240         * match.pd ((A +- CST1) +- CST2): Allow some conversions.
30242 2017-05-26  Martin Liska  <mliska@suse.cz>
30244         * dumpfile.h (enum dump_kind): Renumber TDF_* flags to be contiguous.
30246 2017-05-26  Martin Liska  <mliska@suse.cz>
30248         * cfg.c (check_bb_profile): Do not use TDF_COMMENT and print
30249         always leading ';; '.
30250         (dump_bb_info): Likewise.
30251         (brief_dump_cfg): Likewise.
30252         * cfgrtl.c (print_rtl_with_bb): Do not use TDF_COMMENT.
30253         * dumpfile.c: Remove usage of TDF_VERBOSE.
30254         * dumpfile.h (enum dump_kind): Likewise.
30255         (dump_gimple_bb_header): Do not use TDF_COMMENT.
30256         * print-tree.c (debug_verbose): Remove.
30257         * tree-cfg.c (gimple_dump_cfg): Do not use TDF_COMMENT.
30258         (dump_function_to_file): Remove dumps guarded with TDF_VERBOSE.
30259         * tree-diagnostic.c (default_tree_printer): Replace
30260         TDF_DIAGNOSTIC with TDF_SLIM.
30262 2017-05-26  Bin Cheng  <bin.cheng@arm.com>
30264         * tree-vect-loop-manip.c (create_intersect_range_checks_index): Pass
30265         in parameter loop, rather than loop_vinfo.
30266         (create_intersect_range_checks): Ditto.
30267         (vect_create_cond_for_alias_checks): Update call to above functions.
30269 2017-05-26  Bin Cheng  <bin.cheng@arm.com>
30271         PR tree-optimization/80815
30272         * tree-data-ref.c (prune_runtime_alias_test_list): Simplify condition
30273         for merging runtime alias checks.  Handle negative DR_STEPs.
30275 2017-05-26  Bin Cheng  <bin.cheng@arm.com>
30277         * tree-vect-data-refs.c (Operator==, comp_dr_with_seg_len_pair):
30278         Move from ...
30279         * tree-data-ref.c (Operator==, comp_dr_with_seg_len_pair): To here.
30280         * tree-vect-data-refs.c (vect_prune_runtime_alias_test_list): Factor
30281         out code pruning runtime alias checks.
30282         * tree-data-ref.c (prune_runtime_alias_test_list): New function
30283         factored out from above.
30284         * tree-vectorizer.h (struct dr_with_seg_len, dr_with_seg_len_pair_t):
30285         Move from ...
30286         * tree-data-ref.h (struct dr_with_seg_len, dr_with_seg_len_pair_t):
30287         ... to here.
30288         (prune_runtime_alias_test_list): New decalaration.
30290 2017-05-26  Bin Cheng  <bin.cheng@arm.com>
30292         * tree-vect-data-refs.c (compare_tree): Rename and move ...
30293         * tree-data-ref.c (data_ref_compare_tree): ... to here.
30294         * tree-data-ref.h (data_ref_compare_tree): New decalaration.
30295         * tree-vect-data-refs.c (dr_group_sort_cmp): Update uses.
30296         (operator==, comp_dr_with_seg_len_pair): Ditto.
30297         (vect_prune_runtime_alias_test_list): Ditto.
30299 2017-05-26  Martin Liska  <mliska@suse.cz>
30301         PR ipa/80663
30302         * params.def: Bound partial-inlining-entry-probability param.
30304 2017-05-26  Marek Polacek  <polacek@redhat.com>
30306         PR sanitizer/80875
30307         * fold-const.c (fold_binary_loc) <case MULT_EXPR>: Check if OP1
30308         can be negated.
30310 2017-05-26  Richard Biener  <rguenther@suse.de>
30312         PR tree-optimization/80842
30313         * tree-ssa-ccp.c (set_lattice_value): Always meet with the old
30314         value.
30316 2017-05-26  Richard Biener  <rguenther@suse.de>
30318         PR tree-optimization/80844
30319         * tree-vectorizer.c (adjust_simduid_builtins): Propagate results.
30321 2017-05-25  Sebastian Peryt  <sebastian.peryt@intel.com>
30323         * doc/md.texi (Machine Constraints): Update x86 family
30324         machine constraints section to match 'config/i386/constraints.md'.
30326 2017-05-25  Volker Reichelt  <v.reichelt@netcologne.de>
30328         * doc/invoke.texi (-Wcatch-value=): Document new warning option.
30330 2017-05-25  Nathan Sidwell  <nathan@acm.org>
30332         * doc/invoke.texi (--enable-languages): Update documentation.
30334 2017-05-25  Martin Liska  <mliska@suse.cz>
30336         * dumpfile.c: Add TDF_FOLDING.
30337         * dumpfile.h (enum dump_kind): Likewise.
30338         * genmatch.c (dt_simplify::gen_1): Use it.
30340 2017-05-25  Marc Glisse  <marc.glisse@inria.fr>
30342         * match.pd (view_convert (convert@0 @1)): Handle zero-extension.
30344 2017-05-25  Marc Glisse  <marc.glisse@inria.fr>
30346         * match.pd ((A +- CST1) +- CST2): Allow some conversions.
30347         * tree.c (drop_tree_overflow): Handle COMPLEX_CST and VECTOR_CST.
30349 2017-05-25  Marc Glisse  <marc.glisse@inria.fr>
30351         * fold-const.c (fold_binary_loc) [(A & C) == D]: Remove transformation.
30352         * match.pd (X == C): Rewrite it here.
30353         (with_possible_nonzero_bits, with_possible_nonzero_bits2,
30354         with_certain_nonzero_bits2): New predicates.
30355         * tree-ssanames.c (get_nonzero_bits): Handle INTEGER_CST.
30357 2017-05-24  Nathan Sidwell  <nathan@acm.org>
30359         * lto-streamer-in.c (lto_input_data_block): Adjust T const cast to
30360         avoid warning.
30362         * auto-profile.c (afdo_propagate): Adjust T const cast to avoid
30363         warning.
30365 2017-05-24  Segher Boessenkool  <segher@kernel.crashing.org>
30367         * config/powerpcspe: New port.  Files are copied from the rs6000
30368         port, with "rs6000" in filenames replaced by "powerpcspe".
30370 2017-05-24  Wilco Dijkstra  <wdijkstr@arm.com>
30372         PR rtl-optimization/80754
30373         * lra-remat.c (do_remat): Add overlap checks for dst_regno.
30375 2017-05-24  Sheldon Lobo  <smlobo@sheldon.us.oracle.com>
30377         * config/sparc/sparc.md (length): Return the correct value for -mflat
30378         sibcalls to match output_sibcall.
30380 2017-05-24  Segher Boessenkool  <segher@kernel.crashing.org>
30382         PR bootstrap/80860
30383         PR bootstrap/80843
30384         * config/rs6000/rs6000.c (struct machine_function): Add new field
30385         n_components.
30386         (rs6000_get_separate_components): Init that field, use it.
30387         (rs6000_components_for_bb): Use the field.
30389 2017-05-24  Thomas Preud'homme  <thomas.preudhomme@arm.com>
30391         * config/arm/arm.c (arm_expand_prologue): Fix typo in comment.
30393 2017-05-24  Peter Bergner  <bergner@vnet.ibm.com>
30395         PR middle-end/80823
30396         * tree-cfg.c (group_case_labels_stmt): Delete increment of "i";
30398 2017-05-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
30400         PR target/80725
30401         * config/s390/s390.c (s390_check_qrst_address): Check incoming
30402         address against address_operand predicate.
30403         * config/s390/s390.md ("*indirect_jump"): Swap alternatives.
30405 2017-05-24  Eric Botcazou  <ebotcazou@adacore.com>
30407         * var-tracking.c (track_expr_p): Do not return 0 for tracked record
30408         parameters passed indirectly.
30410 2017-05-23  Uros Bizjak  <ubizjak@gmail.com>
30412         * config/i386/i386.md (*movdi_internal): Remove SSE4
30413         alternative 18 (?r, *v).  Update insn attributes.
30414         (*movsi_internal): Remove SSE4 alternative 13 (?r, *v).
30415         Update insn attributes.
30416         (*zero_extendsidi2): Remove SSE4 alternative (?r, *x).
30417         Update insn attributes.
30418         * config/i386/sse.md (vec_extract<ssevecmodelower>_0): Remove SSE4
30419         alternative 1 (r, v). Remove isa attribute.
30420         * config/i386/i386.c (dimode_scalar_chain::make_vector_copies):
30421         Always move value through stack for !TARGET_INTER_UNIT_MOVES_TO_VEC
30422         and !TARGET_INTER_UNIT_MOVES_TO_VEC targets.
30424 2017-05-23  Tom de Vries  <tom@codesourcery.com>
30426         * doc/sourcebuild.texi (Directives, Verify compiler message): Document
30427         dg-line directive.
30429 2017-05-23  Jan Hubicka  <hubicka@ucw.cz>
30431         * cgraphunit.c (symbol_table::process_new_functions): Update.
30432         * ipa-fnsummary.c (pass_data_inline_parameters): Remove.
30433         (inline_generate_summary): Rename to ...
30434         (ipa_fn_summary_generate): ... this one.
30435         (inline_read_summary): Rename to ...
30436         (ipa_fn_summary_read): ... this one.
30437         (inline_write_summary): Rename to ...
30438         (ipa_fn_summary_write): ... this one.
30439         (inline_free_summary): Rename to ...
30440         (ipa_free_fn_summary): ... this one.
30441         (pass_data_local_fn_summary, pass_local_fn_summary,
30442         make_pass_local_fn_summary, pass_data_ipa_free_fn_summary,
30443         pass_ipa_free_fn_summary, make_pass_ipa_free_fn_summary,
30444         pass_data_ipa_fn_summary, pass_ipa_fn_summary,
30445         make_pass_ipa_fn_summary): New.
30446         * ipa-fnsummary.h (inline_generate_summary, inline_read_summary,
30447         inline_write_summary, inline_free_summary): Remove.
30448         (ipa_free_fn_summary) : New.
30449         * ipa-inline.c (ipa_inline): Update.
30450         (pass_ipa_inline): Do not generate summaries.
30451         * ipa.c (pass_data_ipa_free_fn_summary, pass_ipa_free_fn_summary):
30452         Remove.
30453         * passes.def: Replace pass_inline_parameters by pass_local_fn_summary
30454         and add pass_ipa_fn_summary.
30455         * tree-pass.h (make_pass_ipa_fn_summary, make_pass_local_fn_summary):
30456         New.
30457         (make_pass_inline_parameters): Remove.
30459 2017-05-23  Thomas Schwinge  <thomas@codesourcery.com>
30461         * omp-low.c (struct omp_context): Remove "default_kind" member.
30462         Adjust all users.
30464         * omp-offload.c (execute_oacc_device_lower): Remove the
30465         parallelism dimensions function attributes for unparallelized
30466         OpenACC kernels constructs.
30468 2017-05-23  Martin Liska  <mliska@suse.cz>
30470         * cgraph.c (cgraph_node::get_create): Use symtab_node::dump_{asm_,}name
30471         functions.
30472         (cgraph_edge::make_speculative): Likewise.
30473         (cgraph_edge::resolve_speculation): Likewise.
30474         (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
30475         (cgraph_node::dump): Likewise.
30476         * cgraph.h: Likewise.
30477         * cgraphunit.c (analyze_functions): Likewise.
30478         (symbol_table::compile): Likewise.
30479         * ipa-cp.c (print_all_lattices): Likewise.
30480         (determine_versionability): Likewise.
30481         (initialize_node_lattices): Likewise.
30482         (ipcp_verify_propagated_values): Likewise.
30483         (estimate_local_effects): Likewise.
30484         (update_profiling_info): Likewise.
30485         (create_specialized_node): Likewise.
30486         (perhaps_add_new_callers): Likewise.
30487         (decide_about_value): Likewise.
30488         (decide_whether_version_node): Likewise.
30489         (identify_dead_nodes): Likewise.
30490         (ipcp_store_bits_results): Likewise.
30491         * ipa-devirt.c (dump_targets): Likewise.
30492         (ipa_devirt): Likewise.
30493         * ipa-icf.c (sem_item::dump): Likewise.
30494         (sem_function::equals): Likewise.
30495         (sem_variable::equals): Likewise.
30496         (sem_item_optimizer::read_section): Likewise.
30497         (sem_item_optimizer::execute): Likewise.
30498         (congruence_class::dump): Likewise.
30499         * ipa-inline-analysis.c (dump_ipa_call_summary): Likewise.
30500         (dump_inline_summary): Likewise.
30501         (estimate_node_size_and_time): Likewise.
30502         (inline_analyze_function): Likewise.
30503         * ipa-inline-transform.c (inline_call): Likewise.
30504         * ipa-inline.c (report_inline_failed_reason): Likewise.
30505         (want_early_inline_function_p): Likewise.
30506         (edge_badness): Likewise.
30507         (update_edge_key): Likewise.
30508         (inline_small_functions): Likewise.
30509         * ipa-profile.c (ipa_profile): Likewise.
30510         * ipa-prop.c (ipa_print_node_jump_functions): Likewise.
30511         (ipa_make_edge_direct_to_target): Likewise.
30512         (remove_described_reference): Likewise.
30513         (ipa_impossible_devirt_target): Likewise.
30514         (propagate_controlled_uses): Likewise.
30515         (ipa_print_node_params): Likewise.
30516         (ipcp_transform_function): Likewise.
30517         * ipa-pure-const.c (pure_const_read_summary): Likewise.
30518         (propagate_pure_const): Likewise.
30519         * ipa-reference.c (generate_summary): Likewise.
30520         (read_write_all_from_decl): Likewise.
30521         (propagate): Likewise.
30522         (ipa_reference_read_optimization_summary): Likewise.
30523         * ipa-utils.c (ipa_merge_profiles): Likewise.
30524         * ipa.c (walk_polymorphic_call_targets): Likewise.
30525         (symbol_table::remove_unreachable_nodes): Likewise.
30526         (ipa_single_use): Likewise.
30527         * passes.c (execute_todo): Likewise.
30528         * predict.c (drop_profile): Likewise.
30529         * symtab.c (symtab_node::get_dump_name): New function.
30530         (symtab_node::dump_name): Likewise.
30531         (symtab_node::dump_asm_name): Likewise.
30532         (symtab_node::dump_references): Likewise.
30533         (symtab_node::dump_referring): Likewise.
30534         (symtab_node::dump_base): Likewise.
30535         (symtab_node::debug_symtab): Likewise.
30536         * tree-sra.c (convert_callers_for_node): Likewise.
30537         * tree-ssa-structalias.c (ipa_pta_execute): Likewise.
30538         * value-prof.c (init_node_map): Likewise.
30540 2017-05-23  Martin Liska  <mliska@suse.cz>
30542         * cgraph.h: Move symtab_node::dump_table to symbol_table::dump
30543         and symtab_node::debug_symtab to symbol_table::debug.
30544         * cgraphunit.c (analyze_functions): Use the renamed function.
30545         (symbol_table::compile): Likewise.
30546         * ipa-cp.c (ipcp_verify_propagated_values): Likewise.
30547         * ipa-icf.c (sem_item_optimizer::execute): Likewise.
30548         * passes.c (execute_todo): Likewise.
30549         * symtab.c (symbol_table::dump): New function.
30550         * tree-ssa-structalias.c (ipa_pta_execute): Use the renamed function.
30552 2017-05-23  Jan Hubicka  <hubicka@ucw.cz>
30554         * ipa-fnsummary.c (estimate_node_size_and_time): Do not sanity check
30555         that nonconst implies exec.
30557 2017-05-23  Jan Hubicka  <hubicka@ucw.cz>
30559         * ipa-inline-analysis.c (cgraph_2edge_hook_list, cgraph_edge_hook_list,
30560         inline_edge_removal_hook, inline_edge_duplication_hook): Remove.
30561         (inline_edge_summary_vec): Turn into ...
30562         (ipa_call_summaries): ... this one.
30563         (redirect_to_unreachable, edge_set_predicate,
30564         evaluate_properties_for_edge, inline_summary_alloc,
30565         reset_ipa_call_summary, reset_inline_summary,
30566         inline_summary_t::duplicate): Update.
30567         (inline_edge_duplication_hook): Turn to ...
30568         (ipa_call_summary_t::duplicate): ... this one.
30569         (inline_edge_removal_hook): Turn to ...
30570         (ipa_call_summary_t::remove): ... this one.
30571         (dump_inline_edge_summary): Turn to ...
30572         (dump_ipa_call_summary): ... this one.
30573         (estimate_function_body_sizes): Update.
30574         (inline_update_callee_summaries): Update.
30575         (remap_edge_change_prob): Update.
30576         (remap_edge_summaries): Update.
30577         (inline_merge_summary): Update.
30578         (do_estimate_edge_time): Update.
30579         (inline_generate_summary): Update.
30580         (inline_read_section): Update.
30581         (inline_read_summary): Update.
30582         (inline_free_summary): Update.
30583         * ipa-inline.c (can_inline_edge_p): Update.
30584         (compute_inlined_call_time): Update.
30585         (want_inline_small_function_p): Update.
30586         (edge_badness): Update.
30587         (early_inliner): Update.
30588         * ipa-inline.h (inline_edge_summary): Turn to ...
30589         (ipa_call_summary): ... this one.
30590         (ipa_call_summary_t): New class.
30591         (inline_edge_summary_t, inline_edge_summary_vec): Remove.
30592         (ipa_call_summaries): New.
30593         (inline_edge_summary): Remove.
30594         (estimate_edge_growth): Update.
30595         * ipa-profile.c (ipa_propagate_frequency_1): Update.
30596         * ipa-prop.c (ipa_make_edge_direct_to_target): Update.
30597         * ipa-split.c (execute_split_functions): Update.
30598         * ipa.c (symbol_table::remove_unreachable_nodes): Update.
30600 2017-05-23  Tom de Vries  <tom@codesourcery.com>
30602         * doc/sourcebuild.texi (Effective-Target Keywords, Other hardware
30603         attributes): Document rdrand effective target.
30605 2017-05-23  Tom de Vries  <tom@codesourcery.com>
30607         * doc/sourcebuild.texi (Effective-Target Keywords, Other hardware
30608         attributes): Sort alphabetically.
30610 2017-05-23  Georg-Johann Lay  <avr@gjlay.de>
30612         * config/avr/genmultilib.awk: Use gsub instead of gensub.
30614 2017-05-22  Michael Meissner  <meissner@linux.vnet.ibm.com>
30616         PR target/80718
30617         * config/rs6000/vsx.md (vsx_splat_<mode>, VSX_D iterator): Split
30618         V2DF/V2DI splat into two separate patterns, one that handles
30619         registers, and the other that only handles memory.  Drop support
30620         for splatting from a GPR on ISA 2.07 and then splitting the
30621         splat into direct move and splat.
30622         (vsx_splat_<mode>_reg): Likewise.
30623         (vsx_splat_<mode>_mem): Likewise.
30625 2017-05-22  Segher Boessenkool  <segher@kernel.crashing.org>
30627         * cfgcleanup.c (bb_is_just_return): Allow CLOBBERs.
30629 2017-05-22  Jakub Jelinek  <jakub@redhat.com>
30631         PR middle-end/80809
30632         * omp-low.c (finish_taskreg_remap): New function.
30633         (finish_taskreg_scan): If unit size of ctx->record_type
30634         is non-constant, unshare the size expression and replace
30635         decls in it with possible outer var refs.
30637         PR middle-end/80809
30638         * gimplify.c (omp_add_variable): For GOVD_DEBUG_PRIVATE use
30639         GOVD_SHARED rather than GOVD_PRIVATE with it.
30640         (gimplify_adjust_omp_clauses_1, gimplify_adjust_omp_clauses): Expect
30641         GOVD_SHARED rather than GOVD_PRIVATE with GOVD_DEBUG_PRIVATE.
30643         PR middle-end/80853
30644         * omp-low.c (lower_reduction_clauses): Pass OMP_CLAUSE_PRIVATE
30645         as last argument to build_outer_var_ref for pointer bases of array
30646         section reductions.
30648 2017-05-19  Martin Sebor  <msebor@redhat.com>
30650         * print-tree.c (print_node): Print DECL_READ_P flag.
30652 2017-05-22  Jan Hubicka  <hubicka@ucw.cz>
30654         * Makefile.in: Add ipa-fnsummary.o and ipa-fnsummary.h
30655         * auto-profile.c: Replace ipa-inline.h by ipa-fnsummary.h
30656         * cgraph.c: Likewise.
30657         * cgraphunit.c: Likewise.
30658         * gengtype.c: Likewise.
30659         * ipa-cp.c: Likewise.
30660         * ipa-devirt.c: Likewise.
30661         * ipa-icf.c: Likewise.
30662         * ipa-predicate.c: Likewise.
30663         * ipa-profile.c: Likewise.
30664         * ipa-prop.c: Likewise.
30665         * ipa-split.c: Likewise.
30666         * ipa.c: Likewise.
30667         * ipa-inline-analysis.c (inline_summaries, ipa_call_summaries,
30668         edge_predicate_pool, dump_inline_hints,
30669         inline_summary::account_size_time, redirect_to_unreachable,
30670         edge_set_predicate, set_hint_predicate,
30671         evaluate_conditions_for_known_args, evaluate_properties_for_edge,
30672         inline_summary_alloc, ipa_call_summary::reset, inline_summary::reset,
30673         inline_summary_t::remove, remap_hint_predicate_after_duplication,
30674         inline_summary_t::duplicate, ipa_call_summary_t::duplicate,
30675         ipa_call_summary_t::remove, initialize_growth_caches,
30676         free_growth_caches, dump_ipa_call_summary, dump_inline_summary,
30677         debug_inline_summary, dump_inline_summaries, initialize_inline_failed,
30678         mark_modified, unmodified_parm_1, unmodified_parm,
30679         unmodified_parm_or_parm_agg_item, eliminated_by_inlining_prob,
30680         set_cond_stmt_execution_predicate, set_switch_stmt_execution_predicate,
30681         compute_bb_predicates, will_be_nonconstant_expr_predicate,
30682         will_be_nonconstant_predicate, record_modified_bb_info,
30683         get_minimal_bb, record_modified, param_change_prob,
30684         phi_result_unknown_predicate, predicate_for_phi_result,
30685         array_index_predicate, clobber_only_eh_bb_p, fp_expression_p,
30686         estimate_function_body_sizes, compute_inline_parameters,
30687         compute_inline_parameters_for_curren, pass_data_inline_parameters,
30688         estimate_node_size_and_time, estimate_ipcp_clone_size_and_time,
30689         inline_update_callee_summaries, remap_edge_change_prob,
30690         remap_edge_summaries, remap_hint_predicate, inline_merge_summary,
30691         inline_update_overall_summary, inline_indirect_intraprocedural_analysis,
30692         inline_analyze_function, inline_summary_t::insert,
30693         inline_generate_summary, read_ipa_call_summary, inline_read_section,
30694         inline_read_summary, write_ipa_call_summary, inline_write_summary,
30695         inline_free_summary): Move to ipa-fnsummary.h
30696         (predicate_t): Remove.
30697         * ipa-fnsummary.c: New file.
30698         * ipa-inline.h:  Do not include sreal.h and ipa-predicate.h
30699         (enum inline_hints_vals, inline_hints, agg_position_info,
30700         INLINE_SIZE_SCALE, size_time_entry, inline_summary, inline_summary_t,
30701         inline_summaries, ipa_call_summary, ipa_call_summary_t,
30702         ipa_call_summaries, debug_inline_summary, dump_inline_summaries,
30703         dump_inline_summary, dump_inline_hints, inline_generate_summary,
30704         inline_read_summary, inline_write_summary, inline_free_summary,
30705         inline_analyze_function, initialize_inline_failed,
30706         inline_merge_summary, inline_update_overall_summary,
30707         compute_inline_parameters): Move to ipa-fnsummary.h
30708         * ipa-fnsummary.h: New file.
30709         * ipa-inline-transform.h: Include ipa-inline.h.
30710         * ipa-inline.c: LIkewise.
30712 2017-05-22  Jan Hubicka  <hubicka@ucw.cz>
30714         * ipa-inline.c (edge_badness): Use inlined_time instead of
30715         inline_summaries->get.
30717 2017-05-22  Jan Hubicka  <hubicka@ucw.cz>
30719         * ipa-inline.c (edge_badness): Use estimate_size_after_inlining.
30721 2017-05-22  Nathan Sidwell  <nathan@acm.org>
30723         * doc/invoke.texi (fdump-translation-unit): Delete documentation.
30724         (fdump-lang): Document 'raw' option.
30725         * dumpfile.h (TDI_tu): Delete.
30726         * dumpfile.c (dump_files): Remove translation-unit.
30727         (FIRST_AUTO_NUMBERED_DUMP): Decrement.
30729 2017-05-22  Georg-Johann Lay  <avr@gjlay.de>
30731         * config/avr/t-avr (AWK) [t-multilib]: Remove "-v FORMAT=Makefile"
30732         command option from $(AWK) call.
30733         * config/avr/genmultilib.awk: Simplify and rewrite so that it
30734         generates MULTILIB_REQUIRED instead of MULTILIB_EXCEPTIONS.
30735         [FORMAT]: Remove handling of variable.
30736         * config/avr/t-multilib: Regenerate.
30738 2017-05-22  Jan Hubicka  <hubicka@ucw.cz>
30740         * ipa-inline-analysis.c (inline_summary::reset): Do not reset
30741         self_time.
30742         (dump_inline_summary): Do not print self_time.
30743         (estimate_function_body_sizes): Do not set self_time.
30744         (compute_inline_parameters): Likewise.
30745         (inline_read_section, inline_write_summary): Do not stream self_time.
30746         * ipa-inline.h (inline_summary): Drop self_time.
30748 2017-05-22  Jan Hubicka  <hubicka@ucw.cz>
30750         * ipa-inline-analysis.c (account_size_time): Rename to ...
30751         (inline_summary::account_size_time): ... this one.
30752         (reset_ipa_call_summary): Turn to ...
30753         (ipa_call_summary::reset): ... this one.
30754         (reset_inline_summary): Turn to ...
30755         (inline_summary::reset): ... this one.
30756         (inline_summary_t::remove): Update.
30757         (inline_summary_t::duplicate): Update.
30758         (ipa_call_summary_t::remove): Update.
30759         (dump_inline_summary): Update.
30760         (estimate_function_body_sizes): Update.
30761         (compute_inline_parameters): Update.
30762         (estimate_node_size_and_time): Update.
30763         (inline_merge_summary): Update.
30764         (inline_update_overall_summary): Update.
30765         (inline_read_section): Update.
30766         (inline_write_summary): Update.
30767         * ipa-inline.h (inline_summary): Rename entry to size_time_table;
30768         add account_size_time and reset member functions.
30769         (ipa_call_summary): Add reset function.
30770         * ipa-predicate.h (predicate::operator &): Constify.
30772 2017-05-22  Richard Biener  <rguenther@suse.de>
30774         * df-scan.c (df_insn_refs_verify): Speedup when not verifying.
30776 2017-05-19  Jason Merrill  <jason@redhat.com>
30778         * tree.c (make_tree_vec_stat, grow_tree_vec_stat): Use size_t.
30780 2017-05-19  Marek Polacek  <polacek@redhat.com>
30782         PR sanitizer/80800
30783         * fold-const.c (extract_muldiv_1) <case TRUNC_DIV_EXPR>: Add
30784         TYPE_OVERFLOW_WRAPS checks.
30786 2017-05-19  Thomas Schwinge  <thomas@codesourcery.com>
30788         * tree-core.h (enum omp_clause_default_kind): Add
30789         "OMP_CLAUSE_DEFAULT_PRESENT".
30790         * tree-pretty-print.c (dump_omp_clause): Handle it.
30791         * gimplify.c (enum gimplify_omp_var_data): Add
30792         "GOVD_MAP_FORCE_PRESENT".
30793         (gimplify_adjust_omp_clauses_1): Map it to
30794         "GOMP_MAP_FORCE_PRESENT".
30795         (oacc_default_clause): Handle "OMP_CLAUSE_DEFAULT_PRESENT".
30797         * gimplify.c (oacc_default_clause): Clarify.
30799 2017-05-19  Nathan Sidwell  <nathan@acm.org>
30801         LANG_HOOK_REGISTER_DUMPS
30802         * toplev.c (general_init): Call register dump lang hook.
30803         * doc/invoke.texi: Document -fdump-lang option family.
30804         * dumpfile.c (dump_files): Remove class dump here.
30805         (FIRST_AUTO_NUMBERED_DUMP): Adjust.
30806         * dumpfile.h (tree_dump_index): Remove TDI_class.
30807         * langhooks-def.h (lhd_register_dumps): Declare.
30808         (LANG_HOOKS_REGISTER_DUMPS): Define.
30809         (LANG_HOOKS_INITIALIZER): Add it.
30810         * langhooks.c (lhd_register_dumps): Define.
30811         * langhooks.h (struct lang_hooks): Add register_dumps.
30813 2017-05-19  Nathan Sidwell  <nathan@acm.org>
30815         * context.h (context::set_passes): New.
30816         * context.c (context::context): Do not create pass manager.
30817         * toplev.c (general_init): Create pass manager here.
30819 2017-05-19  Segher Boessenkool  <segher@kernel.crashing.org>
30821         * config/rs6000/rs6000.md (splitter to load of -1 and mask): Don't
30822         use this splitter if two add or or instructions would also work for
30823         the constant we want to generate.
30825 2017-05-19  Richard Biener  <rguenther@suse.de>
30827         PR build/80821
30828         * genmatch.c (dt_node::gen_kids_1): Add missing scope around
30829         predicate evaluation.
30831 2017-05-19  Jan Hubicka  <hubicka@ucw.cz>
30833         * ipa-inline.h (ipa_call_summary): Turn sizes into signed;
30834         add ctor.
30835         * ipa-inline.c (want_inline_small_function_p): Do not cast to
30836         unsigned.
30838 2017-05-19  Jan Hubicka  <hubicka@ucw.cz>
30840         * ipa-inline-analysis.c (cgraph_2edge_hook_list, cgraph_edge_hook_list,
30841         inline_edge_removal_hook, inline_edge_duplication_hook): Remove.
30842         (inline_edge_summary_vec): Turn into ...
30843         (ipa_call_summaries): ... this one.
30844         (redirect_to_unreachable, edge_set_predicate,
30845         evaluate_properties_for_edge, inline_summary_alloc,
30846         reset_ipa_call_summary, reset_inline_summary,
30847         inline_summary_t::duplicate): Update.
30848         (inline_edge_duplication_hook): Turn to ...
30849         (ipa_call_summary_t::duplicate): ... this one.
30850         (inline_edge_removal_hook): Turn to ...
30851         (ipa_call_summary_t::remove): ... this one.
30852         (dump_inline_edge_summary): Turn to ...
30853         (dump_ipa_call_summary): ... this one.
30854         (estimate_function_body_sizes): Update.
30855         (inline_update_callee_summaries): Update.
30856         (remap_edge_change_prob): Update.
30857         (remap_edge_summaries): Update.
30858         (inline_merge_summary): Update.
30859         (do_estimate_edge_time): Update.
30860         (inline_generate_summary): Update.
30861         (inline_read_section): Update.
30862         (inline_read_summary): Update.
30863         (inline_free_summary): Update.
30864         * ipa-inline.c (can_inline_edge_p): Update.
30865         (compute_inlined_call_time): Update.
30866         (want_inline_small_function_p): Update.
30867         (edge_badness): Update.
30868         (early_inliner): Update.
30869         * ipa-inline.h (inline_edge_summary): Turn to ...
30870         (ipa_call_summary): ... this one.
30871         (ipa_call_summary_t): New class.
30872         (inline_edge_summary_t, inline_edge_summary_vec): Remove.
30873         (ipa_call_summaries): New.
30874         (inline_edge_summary): Remove.
30875         (estimate_edge_growth): Update.
30876         * ipa-profile.c (ipa_propagate_frequency_1): Update.
30877         * ipa-prop.c (ipa_make_edge_direct_to_target): Update.
30878         * ipa-split.c (execute_split_functions): Update.
30879         * ipa.c (symbol_table::remove_unreachable_nodes): Update.
30881 2017-05-19  Richard Biener  <rguenther@suse.de>
30883         PR middle-end/80764
30884         * cfgexpand.c (expand_gimple_cond): Fix loop fixup.
30886 2017-05-18  Segher Boessenkool  <segher@kernel.crashing.org>
30888         * config/rs6000/rs6000.c (struct machine_function): Add field
30889         fpr_is_wrapped_separately.
30890         (rs6000_get_separate_components): Use 64 components.  Handle the
30891         new FPR components.
30892         (rs6000_components_for_bb): Handle the FPR components.
30893         (rs6000_emit_prologue_components): Handle the FPR components.
30894         (rs6000_emit_epilogue_components): Handle the FPR components.
30895         (rs6000_set_handled_components): Handle the FPR components.
30896         (rs6000_emit_prologue): Don't output prologue code for those FPRs
30897         that are already separately shrink-wrapped.
30898         (rs6000_emit_epilogue): Don't output epilogue code for those FPRs
30899         that are already separately shrink-wrapped.
30901 2017-05-18  Michael Meissner  <meissner@linux.vnet.ibm.com>
30903         PR target/80510
30904         * config/rs6000/predicates.md (simple_offsettable_mem_operand):
30905         New predicate.
30907         * config/rs6000/rs6000.md (ALTIVEC_DFORM): New iterator.
30908         (define_peephole2 for Altivec d-form load): Add peepholes to catch
30909         cases where the register allocator uses a move and an offsettable
30910         memory operation to/from a FPR register on ISA 2.06/2.07.
30911         (define_peephole2 for Altivec d-form store): Likewise.
30913 2017-05-18  Uros Bizjak  <ubizjak@gmail.com>
30915         PR target/80799
30916         * config/i386/mmx.md (*mov<mode>_internal): Enable
30917         alternatives 11, 12, 13 and 14 also for 32bit targets.
30918         Remove alternatives 15, 16, 17 and 18.
30919         * config/i386/sse.md (vec_concatv2di): Change
30920         alternative (!x, *y) to (x, ?!*Yn).
30922 2017-05-18  Paolo Carlini  <paolo.carlini@oracle.com>
30924         * dumpfile.h (enum dump_kind): Remove stray comma.
30926 2017-05-18  Jan Hubicka  <hubicka@ucw.cz>
30928         * Makefile.in: Add ipa-predicate.o and ipa-predicate.h
30929         * ipa-inline-analysis.c (NUM_CONDITIONS): turn into
30930         predicate::num_conditions
30931         (IS_NOT_CONSTANT): turn into predicate::is_not_constant.
30932         (CHANGED): turn into predicate::changed.
30933         (agg_position_info): Move to ipa-predicate.h
30934         (add_condition, predicate::add_clause, predicate::operator &=,
30935         predicate::or_with, predicate::evaluate, predicate::probability,
30936         dump_condition, dump_clause, predicate::dump,
30937         predicate::remap_after_duplication, predicate::remap_after_inlining,
30938         predicate::stream_in, predicate::stream_out): Move to ipa-predicate.c
30939         (evaluate_conditions_for_known_args): Update.
30940         (set_cond_stmt_execution_predicate): Update.
30941         * ipa-inline.h: Include ipa-predicate.h
30942         (condition, inline_param_summary, conditions, agg_position_info,
30943         predicate): Move to ipa-predicate.h
30944         * ipa-predicate.c: New file.
30945         * ipa-predicate.h: New file.
30947 2017-05-18  Wilco Dijkstra  <wdijkstr@arm.com>
30949         * final.c (leaf_function_p): Check we are not in a sequence.
30951 2017-05-18  Martin Liska  <mliska@suse.cz>
30953         * cfgrtl.c (rtl_verify_edges): Remove usage of TDF_RTL.
30954         * dumpfile.c (dump_register): Use new enum dump_kind.
30955         (get_dump_file_name): Likewise.
30956         (dump_enable_all): Likewise.
30957         (dump_switch_p_1): Likewise.
30958         (enable_rtl_dump_file): Remove usage of TDF_RTL.
30959         * dumpfile.h (enum dump_kind): New enum type.
30960         (struct dump_file_info): Create constructor and
30961         format fields and comments.
30962         * passes.c (pass_manager::register_one_dump_file):
30963         Use num dump_kind.
30964         * statistics.c (statistics_early_init): Likewise.
30965         * tree-ssa-loop-prefetch.c (dump_mem_details): Replace
30966         TDF_TREE with TDF_SLIM.
30967         (gather_memory_references_ref): Likewise.
30969 2017-05-18  Martin Liska  <mliska@suse.cz>
30971         * vec.h (struct vnull): Use it.
30973 2017-05-18  Jan Hubicka  <hubicka@ucw.cz>
30975         * ipa-inline-analysis.c (predicate_conditions): Move to ipa-inline.h
30976         (true_predicate, false_predicate, true_predicate_p,
30977         false_predicate_p): Remove.
30978         (single_cond_predicate, not_inlined_predicate): Turn to member function
30979         in ipa-inline.h
30980         (add_condition): Update.
30981         (add_clause): Turn to...
30982         (predicate::add_clause): ... this one; update; allow passing NULL
30983         as parameter.
30984         (and_predicates): Turn to ...
30985         (predicate::operator &=): ... this one.
30986         (predicates_equal_p): Move to predicate::operator == in ipa-inline.h
30987         (or_predicates): Turn to ...
30988         (predicate::or_with): ... this one.
30989         (evaluate_predicate): Turn to ...
30990         (predicate::evaluate): ... this one.
30991         (predicate_probability): Turn to ...
30992         (predicate::probability): ... this one.
30993         (dump_condition): Update.
30994         (dump_predicate): Turn to ...
30995         (predicate::dump): ... this one.
30996         (account_size_time): Update.
30997         (edge_set_predicate): Update.
30998         (set_hint_predicate): UPdate.
30999         (evaluate_conditions_for_known_args): Update.
31000         (evaluate_properties_for_edge): Update.
31001         (remap_predicate_after_duplication): Turn to...
31002         (predicate::remap_after_duplication): ... this one.
31003         (remap_hint_predicate_after_duplication): Update.
31004         (inline_summary_t::duplicate): UPdate.
31005         (dump_inline_edge_summary): Update.
31006         (dump_inline_summary): Update.
31007         (set_cond_stmt_execution_predicate): Update.
31008         (set_switch_stmt_execution_predicate): Update.
31009         (compute_bb_predicates): Update.
31010         (will_be_nonconstant_expr_predicate): Update.
31011         (will_be_nonconstant_predicate): Update.
31012         (phi_result_unknown_predicate): Update.
31013         (predicate_for_phi_result): Update.
31014         (array_index_predicate): Update.
31015         (estimate_function_body_sizes): Update.
31016         (estimate_node_size_and_time): Update.
31017         (estimate_ipcp_clone_size_and_time): Update.
31018         (remap_predicate): Rename to ...
31019         (predicate::remap_after_inlining): ... this one.
31020         (remap_hint_predicate): Update.
31021         (inline_merge_summary): Update.
31022         (inline_update_overall_summary): Update.
31023         (estimate_size_after_inlining): Update.
31024         (read_predicate): Rename to ...
31025         (predicate::stream_in): ... this one.
31026         (read_inline_edge_summary): Update.
31027         (write_predicate): Rename to ...
31028         (predicate::stream_out): ... this one.
31029         (write_inline_edge_summary): Update.
31030         * ipa-inline.h (MAX_CLAUSES): Turn to predicate::max_clauses.
31031         (clause_t): Turn to uint32_t
31032         (predicate): Turn to class; implement constructor and operators
31033         ==, !=, &
31034         (size_time_entry): Update.
31035         (inline_summary): Update.
31036         (inline_edge_summary): Update.
31038 2017-05-18  Marc Glisse  <marc.glisse@inria.fr>
31040         * fold-const.c (fold_binary_loc): Move transformation...
31041         * match.pd (C - X CMP X): ... here.
31043 2017-05-18  Sheldon Lobo  <sheldon.lobo@oracle.com>
31045         * config/sparc/sparc.c (sparc_option_override): Set function
31046         alignment for -mcpu=niagara7 to 64 to match the I$ line.
31047         * config/sparc/sparc.h (BRANCH_COST): Set the SPARC M7 branch
31048         latency to 1.
31049         * config/sparc/sparc.h (BRANCH_COST): Set the SPARC T4 branch
31050         latency to 2.
31051         * config/sparc/sol2.h: Fix a ASM_CPU32_DEFAULT_SPEC typo.
31053 2017-05-18  Marek Polacek  <polacek@redhat.com>
31055         PR sanitizer/80797
31056         * ubsan.c (instrument_null): Unwrap ADDR_EXPRs.
31057         (pass_ubsan::execute): Call gimple_assign_single_p instead of
31058         gimple_assign_load_p.
31060 2017-05-17  Segher Boessenkool  <segher@kernel.crashing.org>
31062         PR middle-end/80692
31063         * real.c (do_compare): Give decimal_do_compare preference over
31064         comparing just the signs.
31066 2017-05-17  Uros Bizjak  <ubizjak@gmail.com>
31068         * doc/md.texi (Canonicalization of Instructions): Describe the
31069         canonical form of instructions that inherently set a condition
31070         code register.
31072 2017-05-17  Peter Bergner  <bergner@vnet.ibm.com>
31074         PR middle-end/80775
31075         * tree-cfg.c: Move deletion of unreachable case statements to after
31076         the merging of consecutive case labels.
31078 2017-05-17  Thomas Preud'homme  <thomas.preudhomme@arm.com>
31080         * config/arm/arm.c (cmse_nonsecure_call_clear_caller_saved): Refer
31081         readers to __gnu_cmse_nonsecure_call libcall for saving, clearing and
31082         restoring of callee-saved registers.
31084 2017-05-17  Eric Botcazou  <ebotcazou@adacore.com>
31086         * compare-elim.c (equivalent_reg_at_start): Adjust outdated comment.
31087         * config/visium/visium.c (single_set_and_flags): Likewise.
31088         * config/visium/visium.md (Substitutions): Likewise.
31090 2017-05-17  Martin Liska  <mliska@suse.cz>
31092         * cfg.c: Introduce dump_flags_t type and
31093         use it instead of int type.
31094         * cfg.h: Likewise.
31095         * cfghooks.c: Likewise.
31096         * cfghooks.h (struct cfg_hooks): Likewise.
31097         * cfgrtl.c: Likewise.
31098         * cfgrtl.h: Likewise.
31099         * cgraph.c (cgraph_node::get_body): Likewise.
31100         * coretypes.h: Likewise.
31101         * domwalk.c: Likewise.
31102         * domwalk.h: Likewise.
31103         * dumpfile.c (struct dump_option_value_info): Likewise.
31104         (dump_enable_all): Likewise.
31105         (dump_switch_p_1): Likewise.
31106         (opt_info_switch_p): Likewise.
31107         * dumpfile.h (enum tree_dump_index): Likewise.
31108         (struct dump_file_info): Likewise.
31109         * genemit.c: Likewise.
31110         * generic-match-head.c: Likewise.
31111         * gengtype.c (open_base_files): Likewise.
31112         * gimple-pretty-print.c: Likewise.
31113         * gimple-pretty-print.h: Likewise.
31114         * graph.c (print_graph_cfg): Likewise.
31115         * graphite-scop-detection.c (dot_all_sese): Likewise.
31116         * ipa-devirt.c (build_type_inheritance_graph): Likewise.
31117         * loop-unroll.c (report_unroll): Likewise.
31118         * passes.c (pass_manager::register_one_dump_file): Likewise.
31119         * print-tree.c: Likewise.
31120         * statistics.c: Likewise.
31121         * tree-cfg.c: Likewise.
31122         * tree-cfg.h: Likewise.
31123         * tree-dfa.c: Likewise.
31124         * tree-dfa.h: Likewise.
31125         * tree-dump.c (dump_function): Likewise.
31126         * tree-dump.h (struct dump_info): Likewise.
31127         * tree-pretty-print.c: Likewise.
31128         * tree-pretty-print.h: Likewise.
31129         * tree-ssa-live.c: Likewise.
31130         * tree-ssa-live.h: Likewise.
31131         * tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Likewise.
31132         * tree-vect-loop.c: Likewise.
31133         * tree-vect-slp.c: Likewise.
31135 2017-05-16  James Greenhalgh  <james.greenhalgh@arm.com>
31136             Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
31138         PR tree-optimization/80457
31139         * tree-vect-stmts.c (vect_model_simple_cost): Model the cost
31140         of all arguments to a statement as scalar_to_vec operations.
31141         (vectorizable_call): Adjust call to vect_model_simple_cost for
31142         new parameter.
31143         (vectorizable_conversion): Likewise.
31144         (vectorizable_assignment): Likewise.
31145         (vectorizable_shift): Likewise.
31146         (vectorizable_operation): Likewise.
31147         (vectorizable_comparison): Likewise.
31148         (vect_is_simple_cond): Record the def types for operands.
31149         (vectorizable_condition): Likewise, call vect_model_simple_cost.
31150         * tree-vectorizer.h (vect_model_simple_cost): Add new parameter
31151         for statement argument count.
31153 2017-05-16  Carl Love  <cel@us.ibm.com>
31155         * config/rs6000/rs6000-c: Add support for built-in functions
31156         vector unsigned long long vec_bperm (vector unsigned long long,
31157                                              vector unsigned char)
31158         vector signed long long vec_mule (vector signed int,
31159                                           vector signed int)
31160         vector unsigned long long vec_mule (vector unsigned int,
31161                                             vector unsigned int)
31162         vector signed long long vec_mulo (vector signed int,
31163                                           vector signed int)
31164         vector unsigned long long vec_mulo (vector unsigned int,
31165                                             vector unsigned int)
31166         vector signed char vec_sldw (vector signed char,
31167                                      vector signed char,
31168                                      const int)
31169         vector unsigned char vec_sldw (vector unsigned char,
31170                                        vector unsigned char,
31171                                        const int)
31172         vector signed short vec_sldw (vector signed short,
31173                                       vector signed short,
31174                                       const int)
31175         vector unsigned short vec_sldw (vector unsigned short,
31176                                         vector unsigned short,
31177                                         const int)
31178         vector signed int vec_sldw (vector signed int,
31179                                     vector signed int,
31180                                     const int)
31181         vector unsigned int vec_sldw (vector unsigned int,
31182                                       vector unsigned int,
31183                                       const int)
31184         vector signed long long vec_sldw (vector signed long long,
31185                                           vector signed long long,
31186                                           const int)
31187         vector unsigned long long vec_sldw (vector unsigned long long,
31188                                             vector unsigned long long,
31189                                             const int)
31190         * config/rs6000/rs6000-c: Add support for built-in functions
31191         * config/rs6000/rs6000-builtin.def: Add definition for SLDW.
31192         * config/rs6000/altivec.h: Add defintion for vec_sldw.
31193         * doc/extend.texi: Update the built-in documentation for the
31194         new built-in functions.
31196 2017-05-16  Marek Polacek  <polacek@redhat.com>
31198         PR sanitizer/80536
31199         PR sanitizer/80386
31200         * tree.c (save_expr): Don't fold the expression.
31202 2017-05-16  Uros Bizjak  <ubizjak@gmail.com>
31204         * config/i386/i386.md (*movsi_internal): Split (?rm,*y) alternative
31205         to (?r,*Yn) and (?m,*y) alternatives, and (?*y,rm) to (?*Ym,r)
31206         and (?*y,m).  Update insn attributes.
31208 2017-05-16  Martin Liska  <mliska@suse.cz>
31210         * cgraph.c (cgraph_edge::resolve_speculation): Add default value for
31211         flags argument of print_gimple_stmt, print_gimple_expr,
31212         print_generic_stmt and print_generic_expr.
31213         * cgraphclones.c (symbol_table::materialize_all_clones): Likewise.
31214         * coretypes.h: Likewise.
31215         * except.c (dump_eh_tree): Likewise.
31216         * gimple-fold.c (gimple_fold_stmt_to_constant_1): Likewise.
31217         * gimple-pretty-print.h: Likewise.
31218         * gimple-ssa-backprop.c (dump_usage_prefix): Likewise.
31219         (backprop::push_to_worklist): Likewise.
31220         (backprop::pop_from_worklist): Likewise.
31221         (backprop::process_use): Likewise.
31222         (backprop::intersect_uses): Likewise.
31223         (note_replacement): Likewise.
31224         * gimple-ssa-store-merging.c
31225         (pass_store_merging::terminate_all_aliasing_chains): Likewise.
31226         (imm_store_chain_info::coalesce_immediate_stores): Likewise.
31227         (pass_store_merging::execute): Likewise.
31228         * gimple-ssa-strength-reduction.c (dump_candidate): Likewise.
31229         (ssa_base_cand_dump_callback): Likewise.
31230         (dump_incr_vec): Likewise.
31231         (replace_refs): Likewise.
31232         (replace_mult_candidate): Likewise.
31233         (create_add_on_incoming_edge): Likewise.
31234         (create_phi_basis): Likewise.
31235         (insert_initializers): Likewise.
31236         (all_phi_incrs_profitable): Likewise.
31237         (introduce_cast_before_cand): Likewise.
31238         (replace_one_candidate): Likewise.
31239         * gimplify.c (gimplify_expr): Likewise.
31240         * graphite-isl-ast-to-gimple.c (is_valid_rename): Likewise.
31241         (set_rename): Likewise.
31242         (rename_uses): Likewise.
31243         (copy_loop_phi_nodes): Likewise.
31244         (add_close_phis_to_merge_points): Likewise.
31245         (copy_loop_close_phi_args): Likewise.
31246         (copy_cond_phi_args): Likewise.
31247         (graphite_copy_stmts_from_block): Likewise.
31248         (translate_pending_phi_nodes): Likewise.
31249         * graphite-poly.c (print_pdr): Likewise.
31250         (dump_gbb_cases): Likewise.
31251         (dump_gbb_conditions): Likewise.
31252         (print_scop_params): Likewise.
31253         * graphite-scop-detection.c (build_cross_bb_scalars_def): Likewise.
31254         (build_cross_bb_scalars_use): Likewise.
31255         (gather_bbs::before_dom_children): Likewise.
31256         * hsa-dump.c (dump_hsa_immed): Likewise.
31257         * ipa-cp.c (print_ipcp_constant_value): Likewise.
31258         (get_replacement_map): Likewise.
31259         * ipa-inline-analysis.c (dump_condition): Likewise.
31260         (estimate_function_body_sizes): Likewise.
31261         * ipa-polymorphic-call.c (check_stmt_for_type_change): Likewise.
31262         (ipa_polymorphic_call_context::get_dynamic_type): Likewise.
31263         * ipa-prop.c (ipa_dump_param): Likewise.
31264         (ipa_print_node_jump_functions_for_edge): Likewise.
31265         (ipa_modify_call_arguments): Likewise.
31266         (ipa_modify_expr): Likewise.
31267         (ipa_dump_param_adjustments): Likewise.
31268         (ipa_dump_agg_replacement_values): Likewise.
31269         (ipcp_modif_dom_walker::before_dom_children): Likewise.
31270         * ipa-pure-const.c (check_stmt): Likewise.
31271         (pass_nothrow::execute): Likewise.
31272         * ipa-split.c (execute_split_functions): Likewise.
31273         * omp-offload.c (dump_oacc_loop_part): Likewise.
31274         (dump_oacc_loop): Likewise.
31275         * trans-mem.c (tm_log_emit): Likewise.
31276         (tm_memopt_accumulate_memops): Likewise.
31277         (dump_tm_memopt_set): Likewise.
31278         (dump_tm_memopt_transform): Likewise.
31279         * tree-cfg.c (gimple_verify_flow_info): Likewise.
31280         (print_loop): Likewise.
31281         * tree-chkp-opt.c (chkp_print_addr): Likewise.
31282         (chkp_gather_checks_info): Likewise.
31283         (chkp_get_check_result): Likewise.
31284         (chkp_remove_check_if_pass): Likewise.
31285         (chkp_use_outer_bounds_if_possible): Likewise.
31286         (chkp_reduce_bounds_lifetime): Likewise.
31287         * tree-chkp.c (chkp_register_addr_bounds): Likewise.
31288         (chkp_mark_completed_bounds): Likewise.
31289         (chkp_register_incomplete_bounds): Likewise.
31290         (chkp_mark_invalid_bounds): Likewise.
31291         (chkp_maybe_copy_and_register_bounds): Likewise.
31292         (chkp_build_returned_bound): Likewise.
31293         (chkp_get_bound_for_parm): Likewise.
31294         (chkp_build_bndldx): Likewise.
31295         (chkp_get_bounds_by_definition): Likewise.
31296         (chkp_generate_extern_var_bounds): Likewise.
31297         (chkp_get_bounds_for_decl_addr): Likewise.
31298         * tree-chrec.c (chrec_apply): Likewise.
31299         * tree-data-ref.c (dump_data_reference): Likewise.
31300         (dump_subscript): Likewise.
31301         (dump_data_dependence_relation): Likewise.
31302         (analyze_overlapping_iterations): Likewise.
31303         * tree-inline.c (expand_call_inline): Likewise.
31304         (tree_function_versioning): Likewise.
31305         * tree-into-ssa.c (dump_defs_stack): Likewise.
31306         (dump_currdefs): Likewise.
31307         (dump_names_replaced_by): Likewise.
31308         (dump_update_ssa): Likewise.
31309         (update_ssa): Likewise.
31310         * tree-object-size.c (pass_object_sizes::execute): Likewise.
31311         * tree-parloops.c (build_new_reduction): Likewise.
31312         (try_create_reduction_list): Likewise.
31313         (ref_conflicts_with_region): Likewise.
31314         (oacc_entry_exit_ok_1): Likewise.
31315         (oacc_entry_exit_single_gang): Likewise.
31316         * tree-pretty-print.h: Likewise.
31317         * tree-scalar-evolution.c (set_scalar_evolution): Likewise.
31318         (get_scalar_evolution): Likewise.
31319         (add_to_evolution): Likewise.
31320         (get_loop_exit_condition): Likewise.
31321         (analyze_evolution_in_loop): Likewise.
31322         (analyze_initial_condition): Likewise.
31323         (analyze_scalar_evolution): Likewise.
31324         (instantiate_scev): Likewise.
31325         (number_of_latch_executions): Likewise.
31326         (gather_chrec_stats): Likewise.
31327         (final_value_replacement_loop): Likewise.
31328         (scev_const_prop): Likewise.
31329         * tree-sra.c (dump_access): Likewise.
31330         (disqualify_candidate): Likewise.
31331         (create_access): Likewise.
31332         (reject): Likewise.
31333         (maybe_add_sra_candidate): Likewise.
31334         (create_access_replacement): Likewise.
31335         (analyze_access_subtree): Likewise.
31336         (analyze_all_variable_accesses): Likewise.
31337         (sra_modify_assign): Likewise.
31338         (initialize_constant_pool_replacements): Likewise.
31339         (find_param_candidates): Likewise.
31340         (decide_one_param_reduction): Likewise.
31341         (replace_removed_params_ssa_names): Likewise.
31342         * tree-ssa-ccp.c (ccp_fold_stmt): Likewise.
31343         * tree-ssa-copy.c (dump_copy_of): Likewise.
31344         (copy_prop_visit_cond_stmt): Likewise.
31345         * tree-ssa-dce.c (mark_operand_necessary): Likewise.
31346         * tree-ssa-dom.c (pass_dominator::execute): Likewise.
31347         (record_equivalences_from_stmt): Likewise.
31348         * tree-ssa-dse.c (compute_trims): Likewise.
31349         (delete_dead_call): Likewise.
31350         (delete_dead_assignment): Likewise.
31351         * tree-ssa-forwprop.c (forward_propagate_into_gimple_cond): Likewise.
31352         (forward_propagate_into_cond): Likewise.
31353         (pass_forwprop::execute): Likewise.
31354         * tree-ssa-ifcombine.c (ifcombine_ifandif): Likewise.
31355         * tree-ssa-loop-im.c (invariantness_dom_walker::before_dom_children):
31356         Likewise.
31357         (move_computations_worker): Likewise.
31358         (execute_sm): Likewise.
31359         * tree-ssa-loop-ivcanon.c (tree_estimate_loop_size): Likewise.
31360         (remove_exits_and_undefined_stmts): Likewise.
31361         (remove_redundant_iv_tests): Likewise.
31362         * tree-ssa-loop-ivopts.c (dump_use): Likewise.
31363         (adjust_iv_update_pos): Likewise.
31364         * tree-ssa-math-opts.c (bswap_replace): Likewise.
31365         * tree-ssa-phiopt.c (factor_out_conditional_conversion): Likewise.
31366         (value_replacement): Likewise.
31367         * tree-ssa-phiprop.c (phiprop_insert_phi): Likewise.
31368         * tree-ssa-pre.c (print_pre_expr): Likewise.
31369         (get_representative_for): Likewise.
31370         (create_expression_by_pieces): Likewise.
31371         (insert_into_preds_of_block): Likewise.
31372         (eliminate_insert): Likewise.
31373         (eliminate_dom_walker::before_dom_children): Likewise.
31374         (eliminate): Likewise.
31375         (remove_dead_inserted_code): Likewise.
31376         * tree-ssa-propagate.c (substitute_and_fold): Likewise.
31377         * tree-ssa-reassoc.c (get_rank): Likewise.
31378         (eliminate_duplicate_pair): Likewise.
31379         (eliminate_plus_minus_pair): Likewise.
31380         (eliminate_not_pairs): Likewise.
31381         (undistribute_ops_list): Likewise.
31382         (eliminate_redundant_comparison): Likewise.
31383         (update_range_test): Likewise.
31384         (optimize_range_tests_var_bound): Likewise.
31385         (optimize_vec_cond_expr): Likewise.
31386         (rewrite_expr_tree): Likewise.
31387         (rewrite_expr_tree_parallel): Likewise.
31388         (linearize_expr): Likewise.
31389         (break_up_subtract): Likewise.
31390         (linearize_expr_tree): Likewise.
31391         (attempt_builtin_powi): Likewise.
31392         (attempt_builtin_copysign): Likewise.
31393         (transform_stmt_to_copy): Likewise.
31394         (transform_stmt_to_multiply): Likewise.
31395         (dump_ops_vector): Likewise.
31396         * tree-ssa-sccvn.c (vn_nary_build_or_lookup_1): Likewise.
31397         (print_scc): Likewise.
31398         (set_ssa_val_to): Likewise.
31399         (visit_reference_op_store): Likewise.
31400         (visit_use): Likewise.
31401         (sccvn_dom_walker::before_dom_children): Likewise.
31402         (run_scc_vn): Likewise.
31403         * tree-ssa-scopedtables.c (avail_exprs_stack::lookup_avail_expr):
31404         Likewise.
31405         (expr_hash_elt::print): Likewise.
31406         (const_and_copies::pop_to_marker): Likewise.
31407         (const_and_copies::record_const_or_copy_raw): Likewise.
31408         * tree-ssa-structalias.c (compute_dependence_clique): Likewise.
31409         * tree-ssa-uninit.c (collect_phi_def_edges): Likewise.
31410         (dump_predicates): Likewise.
31411         (find_uninit_use): Likewise.
31412         (warn_uninitialized_phi): Likewise.
31413         (pass_late_warn_uninitialized::execute): Likewise.
31414         * tree-ssa.c (verify_vssa): Likewise.
31415         (verify_ssa): Likewise.
31416         (maybe_optimize_var): Likewise.
31417         * tree-vrp.c (dump_value_range): Likewise.
31418         (dump_all_value_ranges): Likewise.
31419         (dump_asserts_for): Likewise.
31420         (register_edge_assert_for_2): Likewise.
31421         (vrp_visit_cond_stmt): Likewise.
31422         (vrp_visit_switch_stmt): Likewise.
31423         (vrp_visit_stmt): Likewise.
31424         (vrp_visit_phi_node): Likewise.
31425         (simplify_cond_using_ranges_1): Likewise.
31426         (fold_predicate_in): Likewise.
31427         (evrp_dom_walker::before_dom_children): Likewise.
31428         (evrp_dom_walker::push_value_range): Likewise.
31429         (evrp_dom_walker::pop_value_range): Likewise.
31430         (execute_early_vrp): Likewise.
31432 2017-05-16  Richard Biener  <rguenther@suse.de>
31434         * dwarf2out.c (loc_list_from_tree_1): Do not create
31435         DW_OP_GNU_variable_value for DECL_IGNORED_P decls.
31437 2017-05-16  Richard Biener  <rguenther@suse.de>
31439         * dwarf2out.c (resolve_variable_value_in_expr): Lookup DIE
31440         just generated.
31441         (note_variable_value_in_expr): If we resolved the decl ref
31442         do not push to the stack.
31444 2017-05-16  Matthew Wahab  <matthew.wahab@arm.com>
31446         * config/arm/arm_neon.h (vadd_f16): Use standard arithmetic
31447         operations in fast-math mode.
31448         (vaddq_f16): Likewise.
31449         (vmul_f16): Likewise.
31450         (vmulq_f16): Likewise.
31451         (vsub_f16): Likewise.
31452         (vsubq_f16): Likewise.
31453         * config/arm/neon.md (add<mode>3): New.
31454         (sub<mode>3): New.
31455         (fma:<VH:mode>3): New.  Also remove outdated comment.
31456         (mul<mode>3): New.
31458 2017-05-16  Martin Liska  <mliska@suse.cz>
31460         PR ipa/79849.
31461         PR ipa/79850.
31462         * ipa-devirt.c (warn_types_mismatch): Fix typo.
31463         (odr_types_equivalent_p): Likewise.
31465 2017-05-15  Sylvestre Ledru  <sylvestre@debian.org>
31467         * plugin.c (try_init_one_plugin): Fix ressource leaks (CID 726637).
31469 2017-05-15  Uros Bizjak  <ubizjak@gmail.com>
31471         PR target/80425
31472         * config/i386.i386.md (*zero_extendsidi2): Do not penalize
31473         non-interunit SSE move alternatives with '?'.
31474         (zero-extendsidi peephole2): New peephole to skip intermediate
31475         general register in SSE zero-extend sequence.
31477 2017-05-15  Jeff Law  <law@redhat.com>
31479         * reorg.c (relax_delay_slots): Create a new variable to hold
31480         the temporary target rather than clobbering TARGET_LABEL.
31482         * config/tilegx/tilegx.c (tilegx_expand_unaligned_load): Add
31483         missing argument to extract_bit_field call.
31484         * config/tilepro/tilepro.c (tilepro_expand_unaligned_load): Likewise.
31486 2017-05-15  Martin Liska  <mliska@suse.cz>
31488         PR driver/31468
31489         * gcc.c (process_command): Do not allow empty argument of -o option.
31491 2017-05-15  Renlin Li  <renlin.li@arm.com>
31493         * config/aarch64/aarch64-protos.h (aarch64_expand_call): Declare.
31494         * config/aarch64/aarch64.c (aarch64_expand_call): Define.
31495         * config/aarch64/constraints.md (Usf): Add long call check.
31496         * config/aarch64/aarch64.md (call): Use aarch64_expand_call.
31497         (call_value): Likewise.
31498         (sibcall): Likewise.
31499         (sibcall_value): Likewise.
31500         (call_insn): New.
31501         (call_value_insn): New.
31502         (sibcall_insn): Update rtx pattern.
31503         (sibcall_value_insn): Likewise.
31504         (call_internal): Remove.
31505         (call_value_internal): Likewise.
31506         (sibcall_internal): Likewise.
31507         (sibcall_value_internal): Likewise.
31508         (call_reg): Likewise.
31509         (call_symbol): Likewise.
31510         (call_value_reg): Likewise.
31511         (call_value_symbol): Likewise.
31513 2017-05-14  Krister Walfridsson  <krister.walfridsson@gmail.com>
31515         PR target/80600
31516         * config/netbsd.h (NETBSD_LIBGCC_SPEC): Always add -lgcc.
31518 2017-05-14  Uros Bizjak  <ubizjak@gmail.com>
31520         * config/i386.i386.c (ix86_cc_modes_compatible): CCNOmode is
31521         compatible with CCGOCmode and with CCZmode.
31523 2017-05-14  Martin Sebor  <msebor@redhat.com>
31525         PR middle-end/77671
31526         * gimple-fold.c (gimple_fold_builtin_sprintf): Make extern.
31527         (gimple_fold_builtin_snprintf): Same.
31528         * gimple-fold.h (gimple_fold_builtin_sprintf): Declare.
31529         (gimple_fold_builtin_snprintf): Same.
31530         * gimple-ssa-sprintf.c (get_format_string): Correct the detection
31531         of character types.
31532         (is_call_safe): New function.
31533         (try_substitute_return_value): Call it.
31534         (try_simplify_call): New function.
31535         (pass_sprintf_length::handle_gimple_call): Call it.
31537 2017-05-14  Martin Sebor  <msebor@redhat.com>
31539         PR middle-end/80669
31540         * builtins.c (expand_builtin_stpncpy): Simplify.
31542 2017-05-14  Daniel Santos  <daniel.santos@pobox.com>
31544         * config/i386/i386.opt: Add option -mcall-ms2sysv-xlogues.
31545         * config/i386/i386.h
31546         (x86_64_ms_sysv_extra_clobbered_registers): Change type to unsigned.
31547         (NUM_X86_64_MS_CLOBBERED_REGS): New macro.
31548         (struct machine_function): Add new members call_ms2sysv,
31549         call_ms2sysv_pad_in, call_ms2sysv_pad_out and call_ms2sysv_extra_regs.
31550         (struct machine_frame_state): New fields sp_realigned and
31551         sp_realigned_offset.
31552         * config/i386/i386.c
31553         (enum xlogue_stub): New enum.
31554         (enum xlogue_stub_sets): New enum.
31555         (class xlogue_layout): New class.
31556         (struct ix86_frame): New fields stack_realign_allocate_offset,
31557         stack_realign_offset and outlined_save_offset.  Modify comments to
31558         detail stack layout when using out-of-line stubs.
31559         (ix86_target_string): Add -mcall-ms2sysv-xlogues option.
31560         (ix86_option_override_internal): Add sorry() for TARGET_SEH and
31561         -mcall-ms2sysv-xlogues.
31562         (stub_managed_regs): New static variable.
31563         (ix86_save_reg): Add new parameter ignore_outlined to optionally omit
31564         registers managed by out-of-line stub.
31565         (disable_call_ms2sysv_xlogues): New function.
31566         (ix86_compute_frame_layout): Modify re-alignment calculations, disable
31567         m->call_ms2sysv when appropriate and compute frame layout for
31568         out-of-line stubs.
31569         (sp_valid_at, fp_valid_at): New inline functions.
31570         (choose_basereg): New function.
31571         (choose_baseaddr): Add align parameter, use choose_basereg and modify
31572         all callers.
31573         (ix86_emit_save_reg_using_mov, ix86_emit_restore_sse_regs_using_mov):
31574         Use align parameter of choose_baseaddr to generated aligned SSE movs
31575         when possible.
31576         (pro_epilogue_adjust_stack): Modify to track
31577         machine_frame_state::sp_realigned.
31578         (ix86_nsaved_regs): Modify to accommodate changes to ix86_save_reg.
31579         (ix86_nsaved_sseregs): Likewise.
31580         (ix86_emit_save_regs): Likewise.
31581         (ix86_emit_save_regs_using_mov): Likewise.
31582         (ix86_emit_save_sse_regs_using_mov): Likewise.
31583         (get_scratch_register_on_entry): Likewise.
31584         (gen_frame_set): New function.
31585         (gen_frame_load): Likewise.
31586         (gen_frame_store): Likewise.
31587         (emit_outlined_ms2sysv_save): Likewise.
31588         (emit_outlined_ms2sysv_restore): Likewise.
31589         (ix86_expand_prologue): Modify stack re-alignment code and call
31590         emit_outlined_ms2sysv_save when appropriate.
31591         (ix86_emit_leave): Clear machine_frame_state::sp_realigned.  Add
31592         parameter rtx_insn *insn, which allows the function to be used to only
31593         generate the notes.
31594         (ix86_expand_epilogue): Modify validity checks of frame and stack
31595         pointers, and call emit_outlined_ms2sysv_restore when appropriate.
31596         (ix86_expand_call): Modify to enable m->call_ms2sysv when appropriate.
31597         * config/i386/predicates.md
31598         (save_multiple): New predicate.
31599         (restore_multiple): Likewise.
31600         * config/i386/sse.md
31601         (save_multiple<mode>): New pattern.
31602         (save_multiple_realign<mode>): Likewise.
31603         (restore_multiple<mode>): Likewise.
31604         (restore_multiple_and_return<mode>): Likewise.
31605         (restore_multiple_leave_return<mode>): Likewise.
31606         * Makefile.in: Export HOSTCXX and HOSTCXXFLAGS to site.exp
31608 2017-05-14  Julia Koval  <julia.koval@intel.com>
31610         * config/i386/i386-builtin-types.def (VOID_FTYPE_INT_INT64): New type.
31611         * config/i386/i386-builtin.def (__builtin_ia32_xgetbv)
31612         (__builtin_ia32_xsetbv): New builtins.
31613         * config/i386/i386.c (ix86_expand_special_args_builtin):
31614         Process new types.
31615         (ix86_expand_builtin): Special expand for new intrinsics.
31616         * config/i386/i386.md (UNSPECV_XGETBV, UNSPECV_XSETBV): New.
31617         (xsetbv, xsetbv_rex64, xgetbv, xgetbv_rex64): New insn patterns.
31618         * config/i386/xsaveintrin.h (_xsetbv, _getbv): New intrinsics.
31620 2017-05-13  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
31622         * cfganal.c (inverted_post_order_compute): Change argument type
31623         to vec *.
31624         * cfganal.h (inverted_post_order_compute): Adjust prototype.
31625         * df-core.c (rest_of_handle_df_initialize): Adjust.
31626         (rest_of_handle_df_finish): Likewise.
31627         (df_analyze_1): Likewise.
31628         (df_analyze): Likewise.
31629         (loop_inverted_post_order_compute): Change argument to be a vec *.
31630         (df_analyze_loop): Adjust.
31631         (df_get_n_blocks): Likewise.
31632         (df_get_postorder): Likewise.
31633         * df.h (struct df_d): Change field to be a vec.
31634         * lcm.c (compute_laterin): Adjust.
31635         (compute_available): Likewise.
31636         * lra-lives.c (lra_create_live_ranges_1): Likewise.
31637         * tree-ssa-dce.c (remove_dead_stmt): Likewise.
31638         * tree-ssa-pre.c (compute_antic): Likewise.
31640 2017-05-13  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
31642         * cfganal.c (connect_infinite_loops_to_exit): Adjust.
31643         (depth_first_search::depth_first_search): Change structure init
31644         function to this constructor.
31645         (depth_first_search::add_bb): Rename function to this member.
31646         (depth_first_search::execute): Likewise.
31647         (flow_dfs_compute_reverse_finish): Adjust.
31649 2017-05-13  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
31651         * ddg.c (find_nodes_on_paths): Use auto_sbitmap.
31652         (longest_simple_path): Likewise.
31653         * shrink-wrap.c (spread_components): Likewise.
31654         (disqualify_problematic_components): Likewise.
31655         (emit_common_heads_for_components): Likewise.
31656         (emit_common_tails_for_components): Likewise.
31657         (insert_prologue_epilogue_for_components): Likewise.
31659 2017-05-13  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
31661         * tree-ssa-dse.c (dse_dom_walker): Make m_live_byes a
31662         auto_sbitmap.
31664 2017-05-13  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
31666         * df-core.c (df_set_blocks): Start using auto_bitmap.
31667         (df_compact_blocks): Likewise.
31668         * df-problems.c (df_rd_confluence_n): Likewise.
31669         * df-scan.c (df_insn_rescan_all): Likewise.
31670         (df_process_deferred_rescans): Likewise.
31671         (df_update_entry_block_defs): Likewise.
31672         (df_update_exit_block_uses): Likewise.
31673         (df_entry_block_bitmap_verify): Likewise.
31674         (df_exit_block_bitmap_verify): Likewise.
31675         (df_scan_verify): Likewise.
31676         * lra-constraints.c (lra_constraints): Likewise.
31677         (undo_optional_reloads): Likewise.
31678         (lra_undo_inheritance): Likewise.
31679         * lra-remat.c (calculate_gen_cands): Likewise.
31680         (do_remat): Likewise.
31681         * lra-spills.c (assign_spill_hard_regs): Likewise.
31682         (spill_pseudos): Likewise.
31683         * tree-ssa-pre.c (bitmap_set_and): Likewise.
31684         (bitmap_set_subtract_values): Likewise.
31686 2017-05-13  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
31688         * haifa-sched.c (estimate_shadow_tick): Replace manual bitmap
31689         management with auto_bitmap.
31690         (fix_inter_tick): Likewise.
31691         (fix_recovery_deps): Likewise.
31692         * ira.c (add_store_equivs): Likewise.
31693         (find_moveable_pseudos): Likewise.
31694         (split_live_ranges_for_shrink_wrap): Likewise.
31695         * print-rtl.c (rtx_reuse_manager::rtx_reuse_manager): Likewise.
31696         (rtx_reuse_manager::seen_def_p): Likewise.
31697         (rtx_reuse_manager::set_seen_def): Likewise.
31698         * print-rtl.h (class rtx_reuse_manager): Likewise.
31700 2017-05-13  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
31702         * bt-load.c (combine_btr_defs): Use auto_bitmap to manage bitmap
31703         lifetime.
31704         (migrate_btr_def): Likewise.
31705         * cfgloop.c (get_loop_body_in_bfs_order): Likewise.
31706         * df-core.c (loop_post_order_compute): Likewise.
31707         (loop_inverted_post_order_compute): Likewise.
31708         * hsa-common.h: Likewise.
31709         * hsa-gen.c (hsa_bb::~hsa_bb): Likewise.
31710         * init-regs.c (initialize_uninitialized_regs): Likewise.
31711         * ipa-inline.c (resolve_noninline_speculation): Likewise.
31712         (inline_small_functions): Likewise.
31713         * ipa-reference.c (ipa_reference_write_optimization_summary): Likewise.
31714         * ira.c (combine_and_move_insns): Likewise.
31715         (build_insn_chain): Likewise.
31716         * loop-invariant.c (find_invariants): Likewise.
31717         * lower-subreg.c (propagate_pseudo_copies): Likewise.
31718         * predict.c (tree_predict_by_opcode): Likewise.
31719         (predict_paths_leading_to): Likewise.
31720         (predict_paths_leading_to_edge): Likewise.
31721         (estimate_loops_at_level): Likewise.
31722         (estimate_loops): Likewise.
31723         * shrink-wrap.c (try_shrink_wrapping): Likewise.
31724         (spread_components): Likewise.
31725         * tree-cfg.c (remove_edge_and_dominated_blocks): Likewise.
31726         * tree-loop-distribution.c (rdg_build_partitions): Likewise.
31727         * tree-predcom.c (tree_predictive_commoning_loop): Likewise.
31728         * tree-ssa-coalesce.c (coalesce_ssa_name): Likewise.
31729         * tree-ssa-phionlycprop.c (pass_phi_only_cprop::execute): Likewise.
31730         * tree-ssa-pre.c (remove_dead_inserted_code): Likewise.
31731         * tree-ssa-sink.c (nearest_common_dominator_of_uses): Likewise.
31732         * tree-ssa-threadupdate.c (compute_path_counts): Likewise.
31733         (mark_threaded_blocks): Likewise.
31734         (thread_through_all_blocks): Likewise.
31735         * tree-ssa.c (verify_ssa): Likewise.
31736         (execute_update_addresses_taken): Likewise.
31737         * tree-ssanames.c (verify_ssaname_freelists): Likewise.
31739 2017-05-13  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
31741         * cfganal.c (mark_dfs_back_edges): Replace manual stack with
31742         auto_vec.
31743         (post_order_compute): Likewise.
31744         (inverted_post_order_compute): Likewise.
31745         (pre_and_rev_post_order_compute_fn): Likewise.
31747 2017-05-13  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
31749         * genrecog.c (int_set::int_set): Explicitly construct our
31750         auto_vec base class.
31751         * vec.h (auto_vec::auto_vec): New constructor.
31753 2017-05-13  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
31755         * bitmap.h (class auto_bitmap): New constructor taking
31756         bitmap_obstack * argument.
31758 2017-05-13  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
31760         * bitmap.h (class auto_bitmap): Change type of m_bits to
31761         bitmap_head, and adjust ctor / dtor and member operators.
31763 2017-05-13  Uros Bizjak  <ubizjak@gmail.com>
31765         * compare-elim.c (equivalent_reg_at_start): Return NULL_RTX
31766         when returned register mode doesn't match original mode.
31768 2017-05-12  Jeff Law  <law@redhat.com>
31769             Jakub Jelinek  <jakub@redhat.com>
31771         * config/mn10300/mn10300.c (mn10300_match_ccmode): Fix where
31772         we look for cc setter after the compare-elim changes.
31773         * config/mn10300/mn10300.md (addsi3_flags): Fix order of patterns
31774         within the vector to match what compare-elim now expects.
31775         (subsi3_flags, andsi3_flags, iorsi3_flags): Likewise.
31776         (xorsi3_flags, one_cmplsi2_flags): Likewise.
31778         * config/rx/rx.c (rx_match_ccmode): Fix where we look cc setter
31779         after the compare-elim changes.
31780         * config/rx/rx.md (abssi2_flags): Fix order of patterns within
31781         the vector to match what compare-elim now expects.
31782         (addsi3_flags, adc_flags, addsi3_flags peepholes): Likewise.
31783         (andsi3_flags, negsi2_flags, one_cmplsi2_flags): Likewise.
31784         (iorsi3_flags, rotlsi3_flags, rotrsi3_flags): Likewise.
31785         (ashrsi3_flags, lshrsi3_flags, ashlsi3_flags): Likewise.
31786         (ssaddsi3, subsi3_flags, sbb_flags, xorsi3_flags): Likewise.
31788         * config/visium/visium.c (single_set_and_flags): Fix where
31789         we look for cc setter after the compare-elim changes.
31790         * config/visium/visium.md (flags_subst_logic): Fix order of patterns
31791         with the vector to match what compare-elim now expects.
31792         (flags_subst_arith, add<mode>3_insn_set_carry): Likewise.
31793         (add<mode>3_insn_set_overflow, addsi3_insn_set_carry): Likewise.
31794         (addsi3_insn_set_overflow, sub<mode>3_insn_set_carry): Likewise.
31795         (sub<mode>3_insn_set_overflow, subsi3_insn_set_carry): Likewise.
31796         (subsi3_insn_set_overflow, negsi2_insn_set_carry): Likewise.
31797         (neg<mode>2_insn_set_overflow): Likewise.
31799 2017-05-12  Jim Wilson  <jim.wilson@linaro.org>
31801         PR middle-end/79794
31802         * expmed.c (extract_bit_field_1): Add alt_rtl argument.  Before
31803         maybe_expand_insn call, set ops[0].target.  If still set after call,
31804         set alt_rtl.  Add extra arg to recursive calls.
31805         (extract_bit_field): Add alt_rtl argument.  Pass to
31806         extract_bit_field.
31807         * expmed.h (extract_bit_field): Fix prototype.
31808         * expr.c (emit_group_load_1, copy_blkmode_from_reg)
31809         (copy_blkmode_to_reg, read_complex_part, store_field): Pass extra NULL
31810         to extract_bit_field_calls.
31811         (expand_expr_real_1): Pass alt_rtl to expand_expr_real instead of 0.
31812         Pass alt_rtl to extract_bit_field calls.
31813         * calls.c (store_unaligned_arguments_into_psuedos)
31814         load_register_parameters): Pass extra NULL to extract_bit_field calls.
31815         * optabs.c (maybe_legitimize_operand): Clear op->target when call
31816         gen_reg_rtx.
31817         * optabs.h (struct expand_operand): Add target bitfield.
31819 2017-05-12  Uros Bizjak  <ubizjak@gmail.com>
31821         * compare-elim.c (try_eliminate_compare): Canonicalize
31822         operation with embedded compare to
31823         [(set (reg:CCM) (compare:CCM (operation) (immediate)))
31824          (set (reg) (operation)].
31826         * config/i386/i386.c (TARGET_FLAGS_REGNUM): New define.
31828 2017-05-12  Uros Bizjak  <ubizjak@gmail.com>
31830         PR target/80723
31831         * config/i386/i386.c (ix86_rtx_cost) [case PLUS]: Ignore the
31832         cost of adding a carry flag for ADC instruction.
31833         [case MINUS]: Ignore the cost of subtracting a carry flag
31834         for SBB instruction.
31836 2017-05-12  Steven Munroe  <munroesj@gcc.gnu.org>
31838         * config.gcc (powerpc*-*-*): Add bmi2intrin.h, bmiintrin.h,
31839         and x86intrin.h
31840         * config/rs6000/bmiintrin.h: New file.
31841         * config/rs6000/bmi2intrin.h: New file.
31842         * config/rs6000/x86intrin.h: New file.
31844 2017-05-12  Jeff Law  <law@redhat.com>
31846         * tree-vrp.c (vrp_dom_walker::before_dom_children): Push unwinding
31847         markers.
31849 2017-05-12  Peter Bergner  <bergner@vnet.ibm.com>
31851         PR middle-end/80707
31852         * tree-cfg.c: Remove cfg edges of unreachable case statements.
31854 2017-05-12  Will Schmidt  <will_schmidt@vnet.ibm.com>
31856         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add handling for
31857         early expansion of vector divide builtins.
31858         (builtin_function_type): Add VSX_BUILTIN_UDIV_V2DI to the list of
31859         builtins identified as having unsigned arguments.
31861 2017-05-12  Will Schmidt  <will_schmidt@vnet.ibm.com>
31863         * config/rs6000/rs6000.c (gimple-fold.h): New #include.
31864         (rs6000_gimple_fold_builtin): Add handling for early GIMPLE
31865         expansion of vector logical operations (and, andc, or, xor,
31866         nor, orc, nand).
31868 2017-05-12  Will Schmidt  <will_schmidt@vnet.ibm.com>
31870         * gimple-fold.c (create_tmp_reg_or_ssa_name): Remove static declaration.
31871         * gimple-fold.h (create_tmp_reg_or_ssa_name): New prototype.
31873 2017-05-12  Will Schmidt  <will_schmidt@vnet.ibm.com>
31875         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add handling for
31876         early GIMPLE expansion of vector multiplies.
31878 2017-05-12  Prakhar Bahuguna  <prakhar.bahuguna@arm.com>
31880         * config/arm/arm.md (movsi): Add TARGET_32BIT in addition to the
31881         TARGET_HAVE_MOVT conditional.
31882         (movt splitter): Likewise.
31884 2017-05-12  Richard Biener  <rguenther@suse.de>
31886         * tree-ssa-sccvn.h (has_VN_INFO): Declare.
31887         * tree-ssa-pre.c (eliminate_dom_walker::before_dom_children):
31888         Fold all stmts not inplace.
31890 2017-05-12  Richard Biener  <rguenther@suse.de>
31892         PR tree-optimization/80713
31893         * tree-ssa-pre.c (remove_dead_inserted_code): Clear
31894         inserted_exprs bit for not removed stmts.
31896 2017-05-12  Thomas Schwinge  <thomas@codesourcery.com>
31898         PR middle-end/69921
31899         * tree-parloops.c (create_parallel_loop): Set "oacc kernels
31900         parallelized" attribute for parallelized OpenACC kernels.
31901         * omp-offload.c (execute_oacc_device_lower): Use it.
31903         * omp-expand.c (expand_omp_target) <GF_OMP_TARGET_KIND_OACC_KERNELS>:
31904         Set "oacc kernels" attribute.
31905         * omp-general.c (oacc_set_fn_attrib): Remove is_kernel formal
31906         parameter.  Adjust all users.
31907         (oacc_fn_attrib_kernels_p): Remove function.
31908         * omp-offload.c (execute_oacc_device_lower): Look for "oacc
31909         kernels" attribute instead of calling oacc_fn_attrib_kernels_p.
31910         * tree-ssa-loop.c (gate_oacc_kernels): Likewise.
31911         * tree-parloops.c (create_parallel_loop): If oacc_kernels_p,
31912         assert "oacc kernels" attribute is set.
31914 2017-05-11  Carl Love  <cel@us.ibm.com>
31916         * config/rs6000/rs6000-c: Add support for built-in functions
31917         vector unsigned char vec_popcnt (vector signed char)
31918         vector unsigned char vec_popcnt (vector unsigned char)
31919         vector unsigned short vec_popcnt (vector signed short)
31920         vector unsigned short vec_popcnt (vector unsigned short)
31921         vector unsigned int vec_popcnt (vector signed int)
31922         vector unsigned int vec_popcnt (vector unsigned int)
31923         vector unsigned long long vec_popcnt (vector signed long long)
31924         vector unsigned long long vec_popcnt (vector unsigned long long)
31925         vector signed long long vec_slo (vector signed long long,
31926                                          vector signed char)
31927         vector signed long long vec_slo (vector signed long long,
31928                                          vector unsigned char)
31929         vector unsigned long long vec_slo (vector unsigned long long,
31930                                            vector signed char)
31931         vector unsigned long long vec_slo (vector unsigned long long,
31932                                            vector unsigned char)
31933         * config/rs6000/rs6000-builtin.def: Add definitions for VPOPCNTUB,
31934         VPOPCNTUH, VPOPCNTUW, and VPOPCNTUD overloads.
31935         * config/rs6000/altivec.h: Add define for vec_popcnt, vec_popcntb,
31936         vec_popcnth, vec_popcntw and vec_popcntd built-in functions.
31937         * doc/extend.texi: Update the built-in documentation file for the
31938         new built-in functions.
31940 2017-05-11  Michael Meissner  <meissner@linux.vnet.ibm.com>
31942         * attribs.h (sorted_attr_string): Move machine independent
31943         functions for target clone support from the i386 port to common
31944         code.  Rename ix86_function_versions to common_function_versions.
31945         Rename make_name to make_unique_name.
31946         (common_function_versions): Likewise.
31947         (make_unique_name): Likewise.
31948         (make_dispatcher_decl): Likewise.
31949         (is_function_default_version): Likewise.
31950         * attribs.c (attr_strcmp): Likewise.
31951         (sorted_attr_string): Likewise.
31952         (common_function_versions): Likewise.
31953         (make_unique_name): Likewise.
31954         (make_dispatcher_decl): Likewise.
31955         (is_function_default_version): Likewise.
31956         * config/i386/i386.c (attr_strcmp): Likewise.
31957         (sorted_attr_string): Likewise.
31958         (ix86_function_versions): Likewise.
31959         (make_name): Likewise.
31960         (make_dispatcher_decl): Likewise.
31961         (is_function_default_version): Likewise.
31962         (TARGET_OPTION_FUNCTION_VERSIONS): Update target function hook.
31964 2017-05-11  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
31966         PR target/80695
31967         * config/rs6000/rs6000.c (rs6000_builtin_vectorization_cost):
31968         Account for direct move costs for vec_construct of integer
31969         vectors.
31971 2017-05-11  Uros Bizjak  <ubizjak@gmail.com>
31973         PR target/80706
31974         * config/i386/sync.md (UNSPEC_LDX_ATOMIC): New unspec.
31975         (UNSPEC_STX_ATOMIC): Ditto.
31976         (loaddi_via_sse): New insn.
31977         (storedi_via_sse): Ditto.
31978         (atomic_loaddi_fpu): Emit loaddi_via_sse and storedi_via_sse.
31979         Update corresponding peephole2 patterns.
31980         (atomic_storedi_fpu): Ditto.
31982 2017-05-11  Julia Koval  <julia.koval@intel.com>
31984         * config/i386/avx512fintrin.h (_mm_mask_rsqrt14_sd)
31985         (_mm_maskz_rsqrt14_sd, _mm_mask_rsqrt14_ss, _mm_maskz_rsqrt14_ss):
31986         New intrinsics.
31987         * config/i386/i386-builtin.def (__builtin_ia32_rsqrt14sd_mask)
31988         (__builtin_ia32_rsqrt14ss_mask): New builtins.
31989         * config/i386/sse.md (rsqrt14_<mode>_mask): New pattern.
31991 2017-05-11  Nathan Sidwell  <nathan@acm.org>
31993         * graphite-poly.c: Include dumpfile.h.
31995         * dumpfle.h (dump_function): Declare here ...
31996         * tree-dump.h (dump_function): ... not here.
31997         * dumpfile.c: #include tree-cfg.h.
31998         (dump_function): Move here from ...
31999         * tree-dump.c (dump_function): ... here.
32000         * gimplify.c: #include splay-tree.h, not tree-dump.h.
32001         * graphite-poly.c: Don't include tree-dump.h.
32002         * cgraphclones.c: Include dumpfile.h not tree-dump.h.
32003         * print-tree.c: Likewise.
32004         * stor-layout.c: Likewise.
32005         * tree-nested.c: Likewise.
32007         * dumpfile.c (dump_start): Use TDF_FLAGS.
32008         (dump_enable_all): Fix TDF_KIND check thinko.
32010 2017-05-11  Kelvin Nilsen  <kelvin@gcc.gnu.org>
32012         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
32013         array entries to represent two legal parameterizations of the
32014         overloaded __builtin_cmpb function, as represented by the
32015         P6_OV_BUILTIN_CMPB constant.
32016         (altivec_resolve_overloaded_builtin): Add special case handling
32017         for the __builtin_cmpb function, as represented by the
32018         P6_OV_BUILTIN_CMPB constant.
32019         * config/rs6000/rs6000-builtin.def (BU_P6_2): New macro.
32020         (BU_P6_64BIT_2): New macro.
32021         (BU_P6_OVERLOAD_2): New macro
32022         (CMPB_32): Add 32-bit compare-bytes support for 32-bit only targets.
32023         (CMPB): Add 64-bit compare-bytes support for 32-bit and 64-bit targets.
32024         (CMPB): Add overload support to represent both 32-bit and 64-bit
32025         compare-bytes function.
32026         * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add
32027         support for TARGET_CMPB.
32028         * config/rs6000/rs6000.h: Add support for RS6000_BTM_CMPB.
32029         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Add
32030         documentation of the __builtin_cmpb overloaded built-in function.
32032 2017-05-11  Richard Biener  <rguenther@suse.de>
32034         PR tree-optimization/80705
32035         * tree-vect-data-refs.c (vect_analyze_data_refs): DECL_NONALIASED
32036         bases are not vectorizable.
32038 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32040         * tree-ssa-loop-ivopts.c (determine_set_costs): Skip non-interger
32041         when counting register pressure.
32043 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32045         * tree-ssa-loop-ivopts.c (compare_cost_pair): New.
32046         (iv_ca_more_deps): Renamed to ...
32047         (iv_ca_compare_deps): ... this.
32048         (iv_ca_extend): Extend iv_ca if NEW_CP is cheaper than OLD_CP.
32050 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32052         * tree-ssa-loop-ivopts.c (find_interesting_uses): Move inv vars dump
32053         to ...
32054         (determine_group_iv_costs): ... here.
32055         (find_inv_vars_cb): Record inv var if it's not recorded before.
32057 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32059         * tree-ssa-loop-ivopts.c (allow_ip_end_pos_p): Refine comments.
32060         (get_shiftadd_cost): Ditto.
32062 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32064         * tree-ssa-address.c: Include header file.
32065         (move_hint_to_base): Return TRUE if BASE_HINT is moved to memory
32066         address.
32067         (add_to_parts): Refactor.
32068         (addr_to_parts): New parameter.  Update use of move_hint_to_base.
32069         (create_mem_ref): Update use of addr_to_parts.  Re-associate addr
32070         in new order.
32072 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32074         PR tree-optimization/53090
32075         * tree-ssa-loop-ivopts.c (enum comp_iv_rewrite): New enum value
32076         COMP_IV_EXPR_2.
32077         (extract_cond_operands): Detect condition with IV on both sides
32078         and return COMP_IV_EXPR_2.
32079         (find_interesting_uses_cond): Add iv_use for both IVs in condition.
32080         (rewrite_use_compare): Simplify by removing call to function
32081         extract_cond_operands.
32083 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32085         * tree-ssa-loop-ivopts.c (enum comp_iv_rewrite): New.
32086         (extract_cond_operands): Detect condition comparing against non-
32087         invariant bound and return appropriate enum value.
32088         (find_interesting_uses_cond): Update use of extract_cond_operands.
32089         Handle its return value accordingly.
32090         (determine_group_iv_cost_cond, rewrite_use_compare): Ditto.
32092 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32094         * tree-ssa-loop-ivopts.c (rewrite_use_nonlinear_expr): Re-associate
32095         nonlinear iv_use computation in loop invariant sensitive way.
32097 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32099         * tree-ssa-loop-ivopts.c (relate_compare_use_with_all_cands): New.
32100         (find_iv_candidates): Call relate_compare_use_with_all_cands.
32102 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32104         * tree-ssa-loop-ivopts.c (struct iv_cand): New field inv_exprs.
32105         (dump_cand): Support iv_cand.inv_exprs.
32106         (add_candidate_1): Record invariant exprs in iv_cand.inv_exprs
32107         for candidates.
32108         (iv_ca_set_no_cp, iv_ca_set_cp, free_loop_data): Support
32109         iv_cand.inv_exprs.
32111 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32113         * tree-ssa-loop-ivopts.c (multiplier_allowed_in_address_p): Move
32114         from ...
32115         * tree-ssa-address.c (multiplier_allowed_in_address_p): ... to here
32116         as local function.  Include necessary header files.
32117         * tree-ssa-loop-ivopts.h (multiplier_allowed_in_address_p): Delete.
32119 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32121         * tree-ssa-loop-ivopts.c (autoinc_possible_for_pair): Simplify.
32123 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32125         * tree-ssa-loop-ivopts.c (force_expr_to_var_cost): Handle more
32126         operators: TRUNC_DIV_EXPR, BIT_AND_EXPR, BIT_IOR_EXPR, LSHIFT_EXPR,
32127         RSHIFT_EXPR and BIT_NOT_EXPR.
32129 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32131         * tree-ssa-loop-ivopts.c (get_loop_invariant_expr): Simplify.
32132         (adjust_setup_cost): New parameter supporting round up adjustment.
32133         (struct address_cost_data): Delete.
32134         (force_expr_to_var_cost): Don't bound cost with spill_cost.
32135         (split_address_cost, ptr_difference_cost): Delete.
32136         (difference_cost, compare_aff_trees, record_inv_expr): Delete.
32137         (struct ainc_cost_data): New struct.
32138         (get_address_cost_ainc): New function.
32139         (get_address_cost, get_computation_cost): Reimplement.
32140         (determine_group_iv_cost_address): Record inv_expr for all uses of
32141         a group.
32142         (determine_group_iv_cost_cond): Call get_loop_invariant_expr.
32143         (iv_ca_has_deps): Reimplemented to ...
32144         (iv_ca_more_deps): ... this.  Check if NEW_CP introduces more deps
32145         than OLD_CP.
32146         (iv_ca_extend): Call iv_ca_more_deps.
32148 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32150         * tree-ssa-address.c (struct mem_address): Move to header file.
32151         (valid_mem_ref_p, move_fixed_address_to_symbol): Make it global.
32152         * tree-ssa-address.h (struct mem_address): Move from C file.
32153         (valid_mem_ref_p, move_fixed_address_to_symbol): Declare.
32155 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32157         * tree-affine.h (aff_combination_type): New interface.
32158         (aff_combination_zero_p): Remove static.
32159         (aff_combination_const_p): New interface.
32160         (aff_combination_singleton_var_p): New interfaces.
32162 2017-05-11  Richard Biener  <rguenther@suse.de>
32164         * tree-ssa-pre.c (eliminate_dom_walker::before_dom_children):
32165         Skip unreachable blocks and destinations.
32166         (eliminate): Move stmt removal and fixup ...
32167         (fini_eliminate): ... here.  Skip inserted exprs.
32168         (pass_pre::execute): Move fini_pre after fini_eliminate.
32169         * tree-ssa-tailmerge.c: Include tree-cfgcleanup.h.
32170         (tail_merge_optimize): Run cleanup_tree_cfg if requested by
32171         PRE to get rid of dead code that has invalid SSA form and
32172         split critical edges again.
32174 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32176         * rtlanal.c (rtx_cost): Handle TRUNCATE between tieable modes.
32178 2017-05-11  Richard Biener  <rguenther@suse.de>
32180         * passes.c (execute_function_todo): Verify loops if they are
32181         said to be up-to-date.
32182         * cfgexpand.c (pass_expand::execute): Discard loops for -dx.
32183         * trans-mem.c (pass_tm_edges::execute): Mark loops for fixup.
32185 2017-05-10  John David Anglin  <danglin@gcc.gnu.org>
32187         PR target/80090
32188         * config/pa/pa.c (pa_assemble_integer): When outputting a SYMBOL_REF,
32189         handle calling assemble_external ourself.
32191         PR target/79027
32192         * config/pa/pa.c (pa_cannot_change_mode_class): Reject changes to/from
32193         modes with zero size.  Enhance comment.
32195 2017-05-10  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
32197         * config/rs6000/rs6000.c (altivec_init_builtins): Define POWER8
32198         built-ins for vec_xl and vec_xst with short and char pointer
32199         arguments.
32201 2017-05-10  Sebastian Peryt  <sebastian.peryt@intel.com>
32203         * config/i386/avx512fintrin.h (_mm_mask_max_round_sd)
32204         (_mm_maskz_max_round_sd, _mm_mask_max_round_ss)
32205         (_mm_maskz_max_round_ss, _mm_mask_min_round_sd)
32206         (_mm_maskz_min_round_sd, _mm_mask_min_round_ss)
32207         (_mm_maskz_min_round_ss): New intrinsics.
32208         * config/i386/i386-builtin-types.def (V2DF, V2DF, V2DF, V2DF, UQI, INT)
32209         (V4SF, V4SF, V4SF, V4SF, UQI, INT): New function type aliases.
32210         * config/i386/i386-builtin.def (__builtin_ia32_maxsd_mask_round)
32211         (__builtin_ia32_maxss_mask_round, __builtin_ia32_minsd_mask_round)
32212         (__builtin_ia32_minss_mask_round): New builtins.
32213         * config/i386/i386.c (V2DF_FTYPE_V2DF_V2DF_V2DF_UQI_INT)
32214         (V4SF_FTYPE_V4SF_V4SF_V4SF_UQI_INT): Handle new types.
32215         * config/i386/sse.md (<sse>_vm<code><mode>3<round_saeonly_name>):
32216         Rename to ...
32217         (<sse>_vm<code><mode>3<mask_name><round_saeonly_name>): ... this.
32218         (v<maxmin_float><ssescalarmodesuffix>\t{<round_saeonly_op3>%2, %1, %0|%0, %1, %<iptr>2<round_saeonly_op3>}):
32219         Change to ...
32220         (v<maxmin_float><ssescalarmodesuffix>\t{<round_saeonly_mask_op3>%2, %1, %0<mask_operand3>|%0<mask_operand3>, %1, %<iptr>2<round_saeonly_mask_op3>}):
32221         ... this.
32223 2017-05-10  Sebastian Peryt  <sebastian.peryt@intel.com>
32225         * config/i386/avx512fintrin.h (_mm_mask_mul_round_sd)
32226         (_mm_maskz_mul_round_sd, _mm_mask_mul_round_ss)
32227         (_mm_maskz_mul_round_ss, _mm_mask_div_round_sd)
32228         (_mm_maskz_div_round_sd, _mm_mask_div_round_ss)
32229         (_mm_maskz_div_round_ss, _mm_mask_mul_sd, _mm_maskz_mul_sd)
32230         (_mm_mask_mul_ss, _mm_maskz_mul_ss, _mm_mask_div_sd)
32231         (_mm_maskz_div_sd, _mm_mask_div_ss, _mm_maskz_div_ss): New intrinsics.
32232         * config/i386/i386-builtin-types.def (V2DF_FTYPE_V2DF_V2DF_V2DF_UQI_INT)
32233         (V4SF_FTYPE_V4SF_V4SF_V4SF_UQI_INT): New function type aliases.
32234         * config/i386/i386-builtin.def (__builtin_ia32_divsd_mask_round)
32235         (__builtin_ia32_divss_mask_round, __builtin_ia32_mulsd_mask_round)
32236         (__builtin_ia32_mulss_mask_round): New builtins.
32237         * config/i386/i386.c (V2DF_FTYPE_V2DF_V2DF_V2DF_UQI_INT)
32238         (V4SF_FTYPE_V4SF_V4SF_V4SF_UQI_INT): Handle new types.
32239         * config/i386/sse.md (<sse>_vm<multdiv_mnemonic><mode>3<round_name>):
32240         Rename to ...
32241         (<sse>_vm<multdiv_mnemonic><mode>3<mask_name><round_name>): ... this.
32242         (v<multdiv_mnemonic><ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|%0, %1, %<iptr>2<round_op3>}):
32243         Change to ...
32244         (v<multdiv_mnemonic><ssescalarmodesuffix>\t{<round_mask_op3>%2, %1, %0<mask_operand3>|%0<mask_operand3>, %1, %<iptr>2<round_mask_op3>}):
32245         ... this.
32247 2017-05-10  Julia Koval  <julia.koval@intel.com>
32249         * config/i386/avxintrin.h (_mm256_set_m128, _mm256_set_m128d)
32250         (_mm256_set_m128i, _mm256_setr_m128, _mm256_setr_m128d)
32251         (_mm256_setr_m128i): New intrinsics.
32253 2017-05-10  Julia Koval  <julia.koval@intel.com>
32255         * config/i386/avx512fintrin.h (_mm_mask_rcp14_sd)
32256         (_mm_maskz_rcp14_sd, _mm_mask_rcp14_ss)
32257         (_mm_maskz_rcp14_ss): New intrinsics.
32258         * config/i386/i386-builtin.def (__builtin_ia32_rcp14sd_mask)
32259         (__builtin_ia32_rcp14ss_mask): New builtins.
32260         * config/i386/sse.md (srcp14<mode>_mask): New pattern.
32262 2017-05-10  Peter Bergner  <bergner@vnet.ibm.com>
32264         PR tree-optimization/51513
32265         * tree-cfg.c (gimple_seq_unreachable_p): New function.
32266         (assert_unreachable_fallthru_edge_p): Use it.
32267         (group_case_labels_stmt): Likewise.
32268         * tree-cfg.h: Prototype it.
32269         * stmt.c: Include cfghooks.h and tree-cfg.h.
32270         (emit_case_dispatch_table) <gap_label>: New local variable.
32271         Use it to fill dispatch table gaps.
32272         Test for default_label before updating probabilities.
32273         (expand_case) <default_label>: Remove unneeded initialization.
32274         Test for unreachable default case statement and remove its edge.
32275         Set default_label accordingly.
32276         * tree-ssa-ccp.c (optimize_unreachable): Update comment.
32278 2017-05-10  Carl Love  <cel@us.ibm.com>
32280         * config/rs6000/rs6000-c: Add support for built-in functions
32281         vector signed char      vec_neg (vector signed char)
32282         vector signed short int vec_neg (vector short int)
32283         vector signed int       vec_neg (vector signed int)
32284         vector signed long long vec_neg (vector signed long long)
32285         vector float            vec_neg (vector float)
32286         vector double           vec_neg (vector double)
32287         * config/rs6000/rs6000-builtin.def: Add definitions for NEG function
32288         overload.
32289         * config/rs6000/altivec.h: Add define for vec_neg
32290         * doc/extend.texi: Update the built-in documentation for the
32291         new built-in functions.
32293 2017-05-10  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
32295         PR tree-optimization/77644
32296         * match.pd (sqrt(x) cmp sqrt(y) -> x cmp y): New pattern.
32298 2017-05-10  Nathan Sidwell  <nathan@acm.org>
32300         * dumpfile.h (TDI_lang_all): New.
32301         (TDF_KIND): New. Renumber others
32302         (TDF_LANG, TDF_TREE, TDF_RTL, TDF_IPA): Enumerate value, rather
32303         than bits.
32304         * dumpfile.c (dump_files): Mark language dumps as TDF_LANG.  add
32305         lang-all.
32306         (get_dump_file_name): Adjust suffix generation.
32307         (dump_enable_all): Use TDF_KIND.
32308         * doc/invoke.texi (-fdump-lang-all): Document.
32310         * dumpfile.h: Tabify.
32312 2017-05-10  Wilco Dijkstra  <wdijkstr@arm.com>
32314         PR target/80671
32315         * config/aarch64/cortex-a57-fma-steering.c (merge_forest):
32316         Move member access before delete.
32318 2017-05-10  Alexandre Oliva  <aoliva@redhat.com>
32320         * tree-inline.c (expand_call_inline): Split block at stmt
32321         before the call.
32323 2017-05-09  Michael Meissner  <meissner@linux.vnet.ibm.com>
32325         PR target/68163
32326         * config/rs6000/rs6000.md (f32_lr): Delete mode attributes that
32327         are now unused after splitting mov{sf,sd}_hardfloat.
32328         (f32_lr2): Likewise.
32329         (f32_lm): Likewise.
32330         (f32_lm2): Likewise.
32331         (f32_li): Likewise.
32332         (f32_li2): Likewise.
32333         (f32_lv): Likewise.
32334         (f32_sr): Likewise.
32335         (f32_sr2): Likewise.
32336         (f32_sm): Likewise.
32337         (f32_sm2): Likewise.
32338         (f32_si): Likewise.
32339         (f32_si2): Likewise.
32340         (f32_sv): Likewise.
32341         (f32_dm): Likewise.
32342         (f32_vsx): Likewise.
32343         (f32_av): Likewise.
32344         (mov<mode>_hardfloat): Split into separate movsf and movsd pieces.
32345         For movsf, order stores so the VSX stores occur before the GPR
32346         store which encourages the register allocator to use a traditional
32347         FPR instead of a GPR.  For movsd, order the stores so that the GPR
32348         store comes before the VSX stores to allow the power6 to work.
32349         This is due to the power6 not having a 32-bit integer store
32350         instruction from a FPR.
32351         (movsf_hardfloat): Likewise.
32352         (movsd_hardfloat): Likewise.
32354 2017-05-09  Martin Sebor  <msebor@redhat.com>
32356         PR translation/80280
32357         * config/sol2-c.c (cmn_err_flag_specs): Initialize new data member
32358         added in r247778.
32360         PR translation/80280
32361         * config/i386/msformat-c.c (ms_printf_flag_specs): Initialize new
32362         data member added in r247778.
32363         (ms_scanf_flag_specs, ms_strftime_flag_specs): Same.
32365 2017-05-09  Nathan Sidwell  <nathan@acm.org>
32367         * tree.h (tree_fits_shwi_p, tree_fits_uhwi_p): Unconditionally pure.
32369         * ipa-devirt.c (default_hash_traits<type_pair>): Skip struct-scope
32370         typedefs.
32372 2017-05-09  Marek Polacek  <polacek@redhat.com>
32374         * doc/invoke.texi: Fix typo.
32376 2017-05-09  Richard Biener  <rguenther@suse.de>
32378         * tree-vrp.c (vrp_val_is_max): Adjust comment.
32379         (vrp_val_is_min): Likewise.
32380         (set_value_range_to_value): Likewise.
32381         (set_value_range_to_nonnegative): Likewise.
32382         (gimple_assign_nonzero_p): Likewise.
32383         (gimple_stmt_nonzero_p): Likewise.
32384         (vrp_int_const_binop): Likewise.  Remove unreachable case.
32385         (adjust_range_with_scev): Adjust comments.
32386         (compare_range_with_value): Likewise.
32387         (extract_range_from_phi_node): Likewise.
32388         (test_for_singularity): Likewise.
32390 2017-05-09  Richard Biener  <rguenther@suse.de>
32392         * tree-vrp.c (get_single_symbol): Add assert that we don't
32393         get overflowed constants as invariant part.
32394         (compare_values_warnv): Add comment before the TREE_NO_WARNING
32395         checks.  Use wi::cmp instead of recursing for integer constants.
32396         (compare_values): Just ignore whether we assumed undefined
32397         overflow instead of failing the compare.
32398         (extract_range_for_var_from_comparison_expr): Add comment before the
32399         TREE_NO_WARNING sets.
32400         (test_for_singularity): Likewise.
32401         (extract_range_from_comparison): Do not disable optimization
32402         when we assumed undefined overflow.
32403         (extract_range_basic): Remove init of unused var.
32405 2017-05-09  Richard Biener  <rguenther@suse.de>
32407         * tree-vrp.c (vrp_int_const_binop): Use wide-ints and simplify.
32408         (extract_range_from_multiplicative_op_1): Adjust.
32409         (extract_range_from_binary_expr_1): Use int_const_binop.
32411 2017-05-08  Kelvin Nilsen  <kelvin@gcc.gnu.org>
32413         PR target/80101
32414         * config/rs6000/power6.md: Replace store_data_bypass_p calls with
32415         rs6000_store_data_bypass_p in seven define_bypass directives and
32416         in several comments.
32417         * config/rs6000/rs6000-protos.h: Add prototype for
32418         rs6000_store_data_bypass_p function.
32419         * config/rs6000/rs6000.c (rs6000_store_data_bypass_p): New
32420         function implements slightly different (rs6000-specific) semantics
32421         than store_data_bypass_p, returning false rather than aborting
32422         with assertion error when arguments do not satisfy the
32423         requirements of store data bypass.
32424         (rs6000_adjust_cost): Replace six calls of store_data_bypass_p with
32425         rs6000_store_data_bypass_p.
32427 2017-05-08  Max Filippov  <jcmvbkbc@gmail.com>
32429         * config/xtensa/xtensa-protos.h
32430         (xtensa_initial_elimination_offset): New declaration.
32431         * config/xtensa/xtensa.c (xtensa_initial_elimination_offset):
32432         New function. Move its body from the INITIAL_ELIMINATION_OFFSET
32433         macro definition, add case for FRAME_POINTER_REGNUM when
32434         FRAME_GROWS_DOWNWARD.
32435         * config/xtensa/xtensa.h (FRAME_GROWS_DOWNWARD): New macro definition.
32436         (INITIAL_ELIMINATION_OFFSET): Replace body with call to
32437         xtensa_initial_elimination_offset.
32439 2017-05-08  Nathan Sidwell  <nathan@acm.org>
32441         * doc/invoke.texi: Alphabetize -fdump options.
32443 2017-05-08  Martin Sebor  <msebor@redhat.com>
32445         PR translation/80280
32446         * config/sol2-c.c (solaris_pragma_align): Correct quoting.
32448 2017-05-08  Bernd Edlinger  <bernd.edlinger@hotmail.de>
32450         * target.def (compute_frame_layout): New optional target hook.
32451         * doc/tm.texi.in (TARGET_COMPUTE_FRAME_LAYOUT): Add hook.
32452         * doc/tm.texi (TARGET_COMPUTE_FRAME_LAYOUT): Add documentation.
32453         * lra-eliminations.c (update_reg_eliminate): Call compute_frame_layout
32454         target hook.
32455         * reload1.c (verify_initial_elim_offsets): Likewise.
32456         * config/arm/arm.c (TARGET_COMPUTE_FRAME_LAYOUT): Define.
32457         (use_simple_return_p): Call arm_compute_frame_layout if needed.
32458         (arm_get_frame_offsets): Split up into this ...
32459         (arm_compute_frame_layout): ... and this function.
32461 2017-05-08  Richard Sandiford  <richard.sandiford@arm.com>
32463         * config/aarch64/constraints.md (Usa): New constraint.
32464         * config/aarch64/aarch64.md (*movsi_aarch64, *movdi_aarch64): Use it.
32466 2017-05-08  Thomas Preud'homme  <thomas.preudhomme@arm.com>
32468         * config.gcc (arm*-*-*): Set TM_MULTILIB_CONFIG from
32469         with_multilib_list after it has been checked.
32471 2017-05-08  Richard Biener  <rguenther@suse.de>
32473         * tree-ssa-pre.c (bitmap_set_and): Avoid bitmap copy.
32474         (bitmap_set_subtract_values): Likewise.
32476 2017-05-08  Richard Biener  <rguenther@suse.de>
32478         * tree-vrp.c (gimple_assign_nonzero_warnv_p): Rename to ...
32479         (gimple_assign_nonzero): ... this and remove strict_overflow_p
32480         argument.
32481         (gimple_stmt_nonzero_warnv_p): Rename to ...
32482         (gimple_stmt_nonzero_p): ... this and remove strict_overflow_p
32483         argument.
32484         (vrp_stmt_computes_nonzero): Remove strict_overflow_p argument.
32485         (extract_range_basic): Adjust, do not disable propagation on
32486         strict overflow sensitive simplification.
32487         (vrp_visit_cond_stmt): Likewise.
32489 2017-05-05  Jan Hubicka  <hubicka@ucw.cz>
32491         * ipa-inline-analysis.c (estimate_function_body_sizes): Recompute
32492         body size unconditionally.
32494 2017-05-07  Jeff Law  <law@redhat.com>
32496         Revert:
32497         2017-05-06  Jeff Law  <law@redhat.com>
32498         PR tree-optimization/78496
32499         * tree-vrp.c (simplify_assert_expr_using_ranges): Remove debugging
32500         code.
32502         PR tree-optimization/78496
32503         * tree-vrp.c (simplify_assert_expr_using_ranges): New function.
32504         (simplify_stmt_using_ranges): Call it.
32505         (vrp_dom_walker::before_dom_children): Extract equivalences
32506         from an ASSERT_EXPR with an equality comparison against a
32507         constant.
32509 2017-05-06  Jeff Law  <law@redhat.com>
32511         PR tree-optimization/78496
32512         * tree-vrp.c (simplify_assert_expr_using_ranges): Remove debugging
32513         code.
32515         PR tree-optimization/78496
32516         * tree-vrp.c (simplify_assert_expr_using_ranges): New function.
32517         (simplify_stmt_using_ranges): Call it.
32518         (vrp_dom_walker::before_dom_children): Extract equivalences
32519         from an ASSERT_EXPR with an equality comparison against a
32520         constant.
32522 2017-05-06  Richard Sandiford  <richard.sandiford@linaro.org>
32524         * lra-constraints.c (lra_copy_reg_equiv): New function.
32525         (split_reg): Use it to copy equivalence information from the
32526         original register to the spill register.
32528 2017-05-06  Richard Sandiford  <richard.sandiford@linaro.org>
32530         PR rtl-optimization/75964
32531         * simplify-rtx.c (simplify_const_relational_operation): Remove
32532         invalid handling of comparisons of integer ABS.
32534 2017-05-06  Uros Bizjak  <ubizjak@gmail.com>
32536         * config/i386/i386.c (ext_80387_constant_init): Do not explicitly
32537         initialize to zero.
32538         (init_regs): Remove declaration.
32539         (function_arg_advance_32): Initialize error_p as boolean variable.
32541 2017-05-05  Nathan Sidwell  <nathan@acm.org>
32543         * store-motion.c (remove_reachable_equiv_notes): Reformat long
32544         lines.  Use for (;;).
32546 2017-05-05  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
32548         * config/rs6000/rs6000.c (rs6000_vect_nonmem): New static var.
32549         (rs6000_init_cost): Initialize rs6000_vect_nonmem.
32550         (rs6000_add_stmt_cost): Update rs6000_vect_nonmem.
32551         (rs6000_finish_cost): Avoid vectorizing simple copy loops with
32552         VF=2 that require versioning.
32554 2017-05-05  David Malcolm  <dmalcolm@redhat.com>
32556         * diagnostic.h (CARET_LINE_MARGIN): Convert from macro to const
32557         int.
32559 2017-05-05  David Malcolm  <dmalcolm@redhat.com>
32561         * diagnostic.h (diagnostic_override_option_index): Convert from
32562         macro to inline function.
32564 2017-05-05  David Malcolm  <dmalcolm@redhat.com>
32566         * diagnostic.c (last_module_changed_p): New function.
32567         (set_last_module): New function.
32568         (diagnostic_report_current_module): Convert macro usage to
32569         the above functions.
32570         * diagnostic.h (diagnostic_context::last_module): Strengthen
32571         from const line_map * to const line_map_ordinary *.
32572         (diagnostic_last_module_changed): Delete macro.
32573         (diagnostic_set_last_module): Delete macro.
32575 2017-05-05  David Malcolm  <dmalcolm@redhat.com>
32577         * diagnostic.c (diagnostic_impl): Replace report_diagnostic
32578         with diagnostic_report_diagnostic.
32579         (diagnostic_n_impl_richloc): Likewise.
32580         * diagnostic.h (report_diagnostic): Delete macro.
32581         * rtl-error.c (diagnostic_for_asm): Replace report_diagnostic
32582         with diagnostic_report_diagnostic.
32583         * substring-locations.c (format_warning_va): Likewise.
32585 2017-05-05  David Malcolm  <dmalcolm@redhat.com>
32587         * diagnostic.c (diagnostic_report_diagnostic): Eliminate
32588         save/restor of format_spec.  Move option-printing code to...
32589         (print_option_information): ...this new function, and
32590         reimplement by simply printing to the pretty_printer,
32591         rather than appending to the format string.
32593 2017-05-05  David Malcolm  <dmalcolm@redhat.com>
32595         * diagnostic.c (diagnostic_report_diagnostic): Split out pragma
32596         handling logic into...
32597         (update_effective_level_from_pragmas): ...this new function.
32599 2017-05-04  Andrew Waterman  <andrew@sifive.com>
32601         * config/riscv/riscv.opt (mstrict-align): New option.
32602         * config/riscv/riscv.h (STRICT_ALIGNMENT): Use it.  Update comment.
32603         (SLOW_UNALIGNED_ACCESS): Define.
32604         (riscv_slow_unaligned_access): Declare.
32605         * config/riscv/riscv.c (riscv_tune_info): Add slow_unaligned_access
32606         field.
32607         (riscv_slow_unaligned_access): New variable.
32608         (rocket_tune_info): Set slow_unaligned_access to true.
32609         (optimize_size_tune_info): Set slow_unaligned_access to false.
32610         (riscv_cpu_info_table): Add entry for optimize_size_tune_info.
32611         (riscv_valid_lo_sum_p): Use TARGET_STRICT_ALIGN.
32612         (riscv_option_override): Set riscv_slow_unaligned_access.
32613         * doc/invoke.texi: Add -mstrict-align to RISC-V.
32615 2017-05-04  Kito Cheng  <kito.cheng@gmail.com>
32617         * config/riscv/riscv.md: Unify indentation.
32619 2017-05-05  Michael Meissner  <meissner@linux.vnet.ibm.com>
32621         PR target/79038
32622         PR target/79202
32623         PR target/79203
32624         * config/rs6000/rs6000.md (u code attribute): Add FIX and
32625         UNSIGNED_FIX.
32626         (extendsi<mode>2): Add support for doing sign extension via
32627         VUPKHSW and XXPERMDI if the value is in Altivec registers and we
32628         don't have ISA 3.0 instructions.
32629         (extendsi<mode>2 splitter): Likewise.
32630         (fix_trunc<mode>si2): If we are at ISA 2.07 (VSX small integer),
32631         generate the normal insns since SImode can now go in vector
32632         registers.  Disallow the special UNSPECs needed for previous
32633         machines to hide SImode being used.  Add new insns
32634         fctiw{,w}_<mode>_smallint if SImode can go in vector registers.
32635         (fix_trunc<mode>si2_stfiwx): Likewise.
32636         (fix_trunc<mode>si2_internal): Likewise.
32637         (fixuns_trunc<mode>si2): Likewise.
32638         (fixuns_trunc<mode>si2_stfiwx): Likewise.
32639         (fctiw<u>z_<mode>_smallint): Likewise.
32640         (fctiw<u>z_<mode>_mem): New combiner pattern to prevent conversion
32641         of floating point to 32-bit integer from doing a direct move to
32642         the GPR registers to do a store.
32643         (fctiwz_<mode>): Break long line.
32645 2017-05-05  Bin Cheng  <bin.cheng@arm.com>
32647         * Makefile.in (GTFILES): Add tree-ssa-loop-ivopts.c.
32648         * tree-ssa-loop-ivopts.c (compute_max_addr_offset): Delete.
32649         (addr_list, addr_offset_valid_p): New.
32650         (split_address_groups): Check offset validity with above function.
32651         (gt-tree-ssa-loop-ivopts.h): Include header file.
32653 2017-05-05  Nathan Sidwell  <nathan@acm.org>
32655         * config.gcc (arm*-*-*): Add missing 'fi'.
32657 2017-05-05  Steve Ellcey  <sellcey@cavium.com>
32659         * doc/invoke.texi (-fopt-info): Explicitly say order of options
32660         included in -fopt-info does not matter.
32661         * doc/optinfo.texi (-fopt-info): Fix description of default
32662         behavour. Explicitly say order of options included in -fopt-info
32663         does not matter.
32665 2017-05-05  Thomas Preud'homme  <thomas.preudhomme@arm.com>
32667         * config.gcc: Allow combinations of aprofile and rmprofile values for
32668         --with-multilib-list.
32669         * config/arm/t-multilib: New file.
32670         * config/arm/t-aprofile: Remove initialization of MULTILIB_*
32671         variables.  Remove setting of ISA and floating-point ABI in
32672         MULTILIB_OPTIONS and MULTILIB_DIRNAMES.  Set architecture and FPU in
32673         MULTI_ARCH_OPTS_A and MULTI_ARCH_DIRS_A rather than MULTILIB_OPTIONS
32674         and MULTILIB_DIRNAMES respectively.  Add comment to introduce all
32675         matches.  Add architecture matches for marvel-pj4 and generic-armv7-a
32676         CPU options.
32677         * config/arm/t-rmprofile: Likewise except for the matches changes.
32678         * doc/install.texi (--with-multilib-list): Document the combination of
32679         aprofile and rmprofile values and warn about pitfalls in doing that.
32681 2017-05-05  Wilco Dijkstra  <wdijkstr@arm.com>
32683         * config/aarch64/aarch64.md (movsi_aarch64): Remove '*' from r=w.
32684         (movdi_aarch64): Likewise.
32686 2017-05-05  Jakub Jelinek  <jakub@redhat.com>
32688         PR tree-optimization/80632
32689         * tree-switch-conversion.c (struct switch_conv_info): Add target_vop
32690         field.
32691         (build_arrays): Initialize it for virtual phis.
32692         (fix_phi_nodes): Use it for virtual phis.
32694         PR tree-optimization/80558
32695         * tree-vrp.c (extract_range_from_binary_expr_1): Optimize
32696         [x, y] op z into [x op, y op z] for op & or | if conditions
32697         are met.
32699 2017-05-05  Andre Vieira  <andre.simoesdiasvieira@arm.com>
32700             Prakhar Bahuguna  <prakhar.bahuguna@arm.com>
32702         PR target/71607
32703         * config/arm/arm.md (use_literal_pool): Remove.
32704         (64-bit immediate split): No longer takes cost into consideration
32705         if arm_disable_literal_pool is enabled.
32706         * config/arm/arm.c (arm_tls_referenced_p): Add diagnostic if TLS is
32707         used when arm_disable_literal_pool is enabled.
32708         (arm_max_const_double_inline_cost): Remove use of
32709         arm_disable_literal_pool.
32710         (push_minipool_fix): Add assert.
32711         (arm_reorg): Add return if arm_disable_literal_pool is enabled.
32712         * config/arm/vfp.md (no_literal_pool_df_immediate): New.
32713         (no_literal_pool_sf_immediate): New.
32715 2017-05-05  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
32717         PR tree-optimization/80613
32718         * tree-ssa-dce.c (propagate_necessity): Remove cases for
32719         BUILT_IN_STRDUP and BUILT_IN_STRNDUP.
32721 2017-05-05  Richard Biener  <rguenther@suse.de>
32723         * tree-ssa-pre.c (get_or_alloc_expr_for): Simplify.
32725 2017-05-05  Georg-Johann Lay  <avr@gjlay.de>
32727         * config/avr/avr.md [flag_strict_overflow]: Remove any occurence
32728         of this flag from insn conditions due to removal from r247495.
32730 2017-05-05  Wilco Dijkstra  <wdijkstr@arm.com>
32732         * config/arm/aarch-common.c (arm_early_load_addr_dep_ptr):
32733         New function.
32734         (arm_early_store_addr_dep_ptr): Likewise.
32735         * config/arm/aarch-common-protos.h
32736         (arm_early_load_addr_dep_ptr): Add prototype.
32737         (arm_early_store_addr_dep_ptr): Likewise.
32738         * config/arm/cortex-a53.md: Add new bypasses.
32740 2017-05-05  Jakub Jelinek  <jakub@redhat.com>
32742         * tree.c (next_type_uid): Change type to unsigned.
32743         (type_hash_canon): Decrement back next_type_uid if
32744         freeing a type node with the highest TYPE_UID.  For INTEGER_TYPEs
32745         also ggc_free TYPE_MIN_VALUE, TYPE_MAX_VALUE and TYPE_CACHED_VALUES
32746         if possible.
32748 2017-05-04  Martin Sebor  <msebor@redhat.com>
32750         * builtins.c: Fix a trivial typo in a comment.
32752         PR middle-end/79234
32753         * builtins.c (check_sizes): Adjust to handle reading past the end.
32754         Avoid printing excessive upper bound of ranges.  Use %E to print
32755         tree nodes instead of converting them to %wu.
32756         (expand_builtin_memchr): New function.
32757         (compute_dest_size): Rename...
32758         (compute_objsize): ...to this.
32759         (expand_builtin_memcpy): Adjust.
32760         (expand_builtin_mempcpy): Adjust.
32761         (expand_builtin_strcat): Adjust.
32762         (expand_builtin_strcpy): Adjust.
32763         (check_strncat_sizes): Adjust.
32764         (expand_builtin_strncat): Adjust.
32765         (expand_builtin_strncpy): Adjust and simplify.
32766         (expand_builtin_memset): Adjust.
32767         (expand_builtin_bzero): Adjust.
32768         (expand_builtin_memcmp): Adjust.
32769         (expand_builtin): Handle memcmp.
32770         (maybe_emit_chk_warning): Check strncat just once.
32772 2017-05-04  Martin Sebor  <msebor@redhat.com>
32774         PR preprocessor/79214
32775         PR middle-end/79222
32776         PR middle-end/79223
32777         * builtins.c (check_sizes): Add inlining context and issue
32778         warnings even when -Wno-system-headers is set.
32779         (check_strncat_sizes): Same.
32780         (expand_builtin_strncat): Same.
32781         (expand_builtin_memmove): New function.
32782         (expand_builtin_stpncpy): Same.
32783         (expand_builtin): Handle memmove and stpncpy.
32785 2017-05-04  Bin Cheng  <bin.cheng@arm.com>
32787         * tree-ssa-loop-ivopts.c (struct cost_pair): Remove field inv_expr
32788         which is not used any more.
32790 2017-05-04  Wilco Dijkstra  <wdijkstr@arm.com>
32792         * config/aarch64/aarch64.c (generic_tunings): Update prefetch model.
32794 2017-05-04  Wilco Dijkstra  <wdijkstr@arm.com>
32796         * config/aarch64/aarch64.c (cortexa35_tunings): Set jump alignment to 4.
32797         (cortexa53_tunings): Likewise.
32798         (cortexa57_tunings): Likewise.
32799         (cortexa72_tunings): Likewise.
32800         (cortexa73_tunings): Likewise.
32802 2017-05-04  Wilco Dijkstra  <wdijkstr@arm.com>
32804         * config/aarch64/aarch64.c (generic_tunings): Set jump alignment to 4.
32805         Set loop alignment to 8.
32807 2017-05-04  Martin Sebor  <msebor@redhat.com>
32809         PR translation/80280
32810         * builtins.c (expand_builtin_object_size): Add missing quoting to
32811         %D and like directives.
32812         * hsa-gen.c (hsa_type_for_scalar_tree_type): Same.
32813         (hsa_type_for_tree_type): Same.
32814         (verify_function_arguments): Same.
32815         * symtab.c (symbol_table::change_decl_assembler_name): Same.
32816         * varasm.c (get_section): Same.
32817         (mark_weak): Same.
32819 2017-05-04  Martin Sebor  <msebor@redhat.com>
32821         PR translation/80280
32822         * config/i386/i386.c (ix86_function_versions): Quote a %D directive.
32824 2017-05-04  Wilco Dijkstra  <wdijkstr@arm.com>
32826         * config/aarch64/aarch64.c (generic_addrcost_table):
32827         Change HI/TI mode setting.
32829 2017-05-04  Martin Jambor  <mjambor@suse.cz>
32831         PR tree-optimization/80622
32832         * tree-sra.c (comes_initialized_p): New function.
32833         (build_accesses_from_assign): Only set write lazily when
32834         comes_initialized_p is false.
32835         (analyze_access_subtree): Use comes_initialized_p.
32836         (propagate_subaccesses_across_link): Assert !comes_initialized_p
32837         instead of testing for PARM_DECL.
32839 2017-05-04  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
32841         * config/aarch64/aarch64.md (prefetch); Adjust predicate and
32842         constraint on operand 0 to allow more general addressing modes.
32843         Adjust output template.
32844         * config/aarch64/aarch64.c (aarch64_address_valid_for_prefetch_p):
32845         New function.
32846         * config/aarch64/aarch64-protos.h
32847         (aarch64_address_valid_for_prefetch_p): Declare prototype.
32848         * config/aarch64/constraints.md (Dp): New address constraint.
32849         * config/aarch64/predicates.md (aarch64_prefetch_operand): New
32850         predicate.
32852 2017-05-04  Jan Hubicka  <hubicka@ucw.cz>
32854         * ipa-cp.c (perform_estimation_of_a_value): Drop base_time parameter;
32855         update use of estimate_ipcp_clone_size_and_time.
32856         (estimate_local_effects): Update use of
32857         estimate_ipcp_clone_size_and_time and perform_estimation_of_a_value.
32858         * ipa-inline.h (estimate_ipcp_clone_size_and_time): Update prototype.
32859         * ipa-inline-analysis.c (estimate_ipcp_clone_size_and_time):
32860         Return nonspecialized time.
32862 2017-05-04  Richard Biener  <rguenther@suse.de>
32864         * tree-ssa-alias.c (get_continuation_for_phi): Improve looking
32865         for the last VUSE which def dominates the PHI.  Directly call
32866         maybe_skip_until.
32867         (get_continuation_for_phi_1): Remove.
32869 2017-05-04  Richard Sandiford  <richard.sandiford@linaro.org>
32871         * tree-ssa-loop-manip.c (niter_for_unrolled_loop): Add commentary
32872         to explain the use of truncating division.  Cap the number of
32873         iterations to the maximum given by nb_iterations_upper_bound,
32874         if defined.
32876 2017-05-04  Thomas Preud'homme  <thomas.preudhomme@arm.com>
32878         * configure.ac (--enable-mingw-wildcard): Add new configurable feature.
32879         * configure: Regenerate.
32880         * config.in: Regenerate.
32881         * config/i386/driver-mingw32.c: new file.
32882         * config/i386/x-mingw32: Add rule to build driver-mingw32.o.
32883         * config.host: Link driver-mingw32.o on MinGW host.
32884         * doc/install.texi: Document new --enable-mingw-wildcard configure
32885         option.
32887 2017-05-04  Marek Polacek  <polacek@redhat.com>
32889         PR tree-optimization/80612
32890         * calls.c (get_size_range): Check for INTEGRAL_TYPE_P.
32892 2017-05-04  Prakhar Bahuguna  <prakhar.bahuguna@arm.com>
32893             Andre Simoes Dias Vieira  <andre.simoesdiasvieira@arm.com>
32895         * config/arm/arm.md (movsi): Change TARGET_32BIT to TARGET_HAVE_MOVT.
32896         (movt splitter): Likewise.
32897         * config/arm/arm.c (arm_option_check_internal): Change arm_arch_thumb2
32898         to TARGET_HAVE_MOVT, and merge with -mslow-flash-data check.
32899         (const_ok_for_arm): Change else to else if (TARGET_THUMB2) and add else
32900         block for Thumb-1 with MOVT.
32901         (thumb2_legitimate_address_p): Move code block ...
32902         (can_avoid_literal_pool_for_label_p): ... into this new function.
32903         (thumb1_legitimate_address_p): Add check for TARGET_HAVE_MOVT and
32904         literal pool.
32905         (thumb_legitimate_constant_p): Add conditional on TARGET_HAVE_MOVT
32906         * doc/invoke.texi (-mpure-code): Change "ARMv7-M targets" for
32907         "M-profile targets with the MOVT instruction".
32909 2017-05-04  Prakhar Bahuguna  <prakhar.bahuguna@arm.com>
32911         * config/arm/arm-builtins.c (arm_init_builtins): Rename
32912         __builtin_arm_ldfscr to __builtin_arm_get_fpscr, and rename
32913         __builtin_arm_stfscr to __builtin_arm_set_fpscr.
32915 2017-05-04  Martin Liska  <mliska@suse.cz>
32917         * tree-vrp.c (simplify_cond_using_ranges_2): Remove unused
32918         variable cond_code.
32920 2017-05-04  Richard Biener  <rguenther@suse.de>
32922         * tree.c (array_at_struct_end_p): Handle arrays at struct
32923         end with flexarrays more conservatively.  Refactor and treat
32924         arrays of arrays or aggregates more strict.  Fix
32925         VIEW_CONVERT_EXPR handling.  Remove allow_compref argument.
32926         * tree.c (array_at_struct_end_p): Adjust prototype.
32927         * emit-rtl.c (set_mem_attributes_minus_bitpos): Adjust.
32928         * gimple-fold.c (get_range_strlen): Likewise.
32929         * tree-chkp.c (chkp_may_narrow_to_field): Likewise.
32931 2017-05-04  Richard Biener  <rguenther@suse.de>
32933         PR tree-optimization/31130
32934         * tree-vrp.c (needs_overflow_infinity): Remove as always returning
32935         false.
32936         (supports_overflow_infinity): Likewise.
32937         (is_negative_overflow_infinity): Likewise.
32938         (is_positive_overflow_infinity): Likewise.
32939         (is_overflow_infinity): Likewise.
32940         (stmt_overflow_infinity): Likewise.
32941         (overflow_infinity_range_p): Likewise.
32942         (usable_range_p): Remove as always returning true.
32943         (make_overflow_infinity): Remove.
32944         (negative_overflow_infinity): Likewise.
32945         (positive_overflow_infinity): Likewise.
32946         (avoid_overflow_infinity): Likewise.
32947         (set_value_range): Adjust accordingly.
32948         (set_value_range_to_nonnegative): Likewise, remove now unused
32949         overflow_infinity arg.
32950         (vrp_operand_equal_p): Adjust.
32951         (update_value_range): Likewise.
32952         (range_int_cst_singleton_p): Likewise.
32953         (operand_less_p): Likewise.
32954         (compare_values_warnv): Likewise.
32955         (extract_range_for_var_from_comparison_expr): Likewise.
32956         (vrp_int_const_binop): Likewise.
32957         (zero_nonzero_bits_from_vr): Likewise.
32958         (extract_range_from_multiplicative_op_1): Likewise.
32959         (extract_range_from_binary_expr_1): Likewise.
32960         (extract_range_from_unary_expr): Likewise.
32961         (extract_range_from_comparison): Likewise.
32962         (extract_range_basic): Likewise.
32963         (adjust_range_with_scev): Likewise.
32964         (compare_ranges): Likewise.
32965         (compare_range_with_value): Likewise.
32966         (dump_value_range): Likewise.
32967         (test_for_singularity): Likewise, remove strict_overflow_p parameter
32968         never used.
32969         (simplify_cond_using_ranges): Adjust.
32971 2017-05-04  Pekka Jääskeläinen  <pekka.jaaskelainen@parmance.com>
32973         * brig-builtins.def: Added a builtin for class_f64.
32974         * builtin-types.def: Added a builtin type needed by class_f64.
32976 2017-05-03  Jason Merrill  <jason@redhat.com>
32978         * timevar.def: Add TV_CONSTEXPR.
32980 2017-05-03  David Malcolm  <dmalcolm@redhat.com>
32982         * common.opt (fdiagnostics-parseable-fixits): Fix typo.
32984 2017-05-03  Martin Jambor  <mjambor@suse.cz>
32986         * ipa-prop.c (ipa_update_after_lto_read): Removed.
32987         * ipa-prop.h (ipa_update_after_lto_read): Remove declaration.
32988         * ipa-cp.c (ipcp_propagate_stage): Do not call
32989         ipa_update_after_lto_read.
32990         * ipa-inline.c (ipa_inline): Likewise.
32992 2017-05-03  Martin Jambor  <mjambor@suse.cz>
32994         * ipa-prop.h (ipa_edge_args): Make a class.  Mark with for_user GTY
32995         tag.  Added a default constructor and a destructor.
32996         (ipa_edge_args_sum_t): New class;
32997         (ipa_edge_args_sum): Declare.
32998         (ipa_edge_args_vector): Remove declaration.
32999         (IPA_EDGE_REF): Use ipa_edge_args_sum.
33000         (ipa_free_edge_args_substructures): Remove declaration.
33001         (ipa_check_create_edge_args): Use ipa_edge_args_sum.
33002         (ipa_edge_args_info_available_for_edge_p): Likewise.
33003         * ipa-prop.c (ipa_edge_args_vector): Removed.
33004         (edge_removal_hook_holder): Likewise.
33005         (edge_duplication_hook_holder): Likewise.
33006         (ipa_edge_args_sum): New variable.
33007         (ipa_propagate_indirect_call_infos): Test ipa_edge_args_sum instead of
33008         ipa_edge_args_vector.
33009         (ipa_free_edge_args_substructures): Likewise.
33010         (ipa_free_all_edge_args): Free ipa_edge_args_sum instead of
33011         ipa_edge_args_vector.
33012         (ipa_edge_removal_hook): Turned into method
33013         ipa_edge_args_sum_t::remove.
33014         (ipa_edge_duplication_hook): Turned into method
33015         ipa_edge_args_sum_t::duplicate.
33016         (ipa_register_cgraph_hooks): Create ipa_edge_args_sum instead of
33017         registering edge hooks.
33018         (ipa_unregister_cgraph_hooks): Do not unregister edge hooks.
33019         * ipa-inline-analysis.c (estimate_function_body_sizes): Test
33020         ipa_edge_args_sum instead of ipa_edge_args_vector.
33021         * ipa-profile.c (ipa_profile): Likewise.
33023 2017-05-03  Martin Jambor  <mjambor@suse.cz>
33025         * symbol-summary.h (function_summary): New method exists.
33026         (function_summary::symtab_removal): Deallocate through release.
33027         (call_summary): New class.
33028         (gt_ggc_mx): New overload.
33029         (gt_pch_nx): Likewise.
33030         (gt_pch_nx): Likewise.
33032 2017-05-03  Jeff Law  <law@redhat.com>
33034         PR tree-optimization/78496
33035         * tree-vrp.c (simplify_cond_using_ranges_1): Renamed
33036         from simplify_cond_using_ranges.  Split off code to walk
33037         backwards through casts into ...
33038         (simplify_cond_using_ranges_2): New function.
33039         (simplify_stmt_using_ranges): Call simplify_cond_using_ranges_1.
33040         (execute_vrp): After identifying jump threads, call
33041         simplify_cond_using_ranges_2.
33043 2017-05-03  Jan Hubicka  <hubicka@ucw.cz>
33045         PR bootstrap/80609
33046         * ipa-inline.h (inline_summary): Add ctor.
33047         (create_ggc): Do not use ggc_cleared_alloc.
33049 2017-05-03  Jeff Downs  <heydowns@somuchpressure.net>
33050             Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
33052         * gcc.c (handle_braces): Support escaping in switch matching
33053         text.
33054         * doc/invoke.texi (Spec Files): Document it.
33055         Remove superfluous @code markup in items.
33057 2017-05-03  David Malcolm  <dmalcolm@redhat.com>
33059         * diagnostic-show-locus.c (struct column_range): New struct.
33060         (get_affected_columns): New function.
33061         (get_printed_columns): New function.
33062         (struct correction): New struct.
33063         (correction::ensure_capacity): New function.
33064         (correction::ensure_terminated): New function.
33065         (struct line_corrections): New struct.
33066         (line_corrections::~line_corrections): New dtor.
33067         (line_corrections::add_hint): New function.
33068         (layout::print_trailing_fixits): Reimplement in terms of the new
33069         classes.
33070         (selftest::test_overlapped_fixit_printing): New function.
33071         (selftest::diagnostic_show_locus_c_tests): Call it.
33073 2017-05-03  Nathan Sidwell  <nathan@acm.org>
33075         Canonicalize canonical type hashing
33076         * tree.h (type_hash_canon_hash): Declare.
33077         * tree.c (type_hash_list, attribute_hash_list): Move into
33078         type_hash_canon_hash.
33079         (build_type_attribute_qual_variant): Break out hash code calc into
33080         type_hash_canon_hash.
33081         (type_hash_canon_hash): New.  Generic type hash computation.
33082         (build_range_type_1, build_array_type_1, build_function_type,
33083         build_method_type_directly, build_offset_type, build_complex_type,
33084         make_vector_type): Call it.
33086 2017-05-03  Richard Biener  <rguenther@suse.de>
33088         * tree-vect-data-refs.c (vect_enhance_data_refs_alignment):
33089         When all DRs have unknown misaligned do not always peel
33090         when there is a store but apply the same costing model as if
33091         there were only loads.
33093 2017-05-03  Richard Biener  <rguenther@suse.de>
33095         Revert
33096         PR tree-optimization/80492
33097         * tree-ssa-alias.c (decl_refs_may_alias_p): Handle
33098         compare_base_decls returning dont-know properly.
33100 2017-05-03  Thomas Preud'homme  <thomas.preudhomme@arm.com>
33102         * config/arm/iterators.md (CCSI): New mode iterator.
33103         (arch): New mode attribute.
33104         * config/arm/sync.md (atomic_compare_and_swap<mode>_1): Rename into ...
33105         (atomic_compare_and_swap<CCSI:arch><NARROW:mode>_1): This and ...
33106         (atomic_compare_and_swap<CCSI:arch><SIDI:mode>_1): This.  Use CCSI
33107         code iterator for success result mode.
33108         * config/arm/arm.c (arm_expand_compare_and_swap): Adapt code to use
33109         the corresponding new insn generators.
33111 2017-05-03  Bin Cheng  <bin.cheng@arm.com>
33113         Revert r247509
33114         2017-05-02  Bin Cheng  <bin.cheng@arm.com>
33115         * rtlanal.c (rtx_cost): Handle TRUNCATE between tieable modes.
33117 2017-05-03  Richard Sandiford  <richard.sandiford@linaro.org>
33119         * tree-data-ref.h (SUB_CONFLICTS_IN_A): Wrap SUB argument in brackets.
33120         (SUB_CONFLICTS_IN_B, SUB_LAST_CONFLICT, SUB_DISTANCE): Likewise.
33121         (DDR_A): Wrap DDR argument in brackets.
33122         (DDR_B, DDR_AFFINE_P, DDR_ARE_DEPENDENT, DDR_SUBSCRIPTS): Likewise.
33123         (DDR_LOOP_NEST, DDR_INNER_LOOP, DDR_SELF_REFERENCE): Likewise.
33124         (DDR_REVERSED_P): Likewise.
33126 2017-05-03  Jakub Jelinek  <jakub@redhat.com>
33128         PR tree-optimization/79472
33129         * tree-switch-conversion.c (struct switch_conv_info): Add
33130         contiguous_range and default_case_nonstandard fields.
33131         (collect_switch_conv_info): Compute contiguous_range and
33132         default_case_nonstandard fields, don't clear final_bb if
33133         contiguous_range and only the default case doesn't have the required
33134         structure.
33135         (check_all_empty_except_final): Set default_case_nonstandard instead
33136         of failing if contiguous_range and the default case doesn't have empty
33137         block.
33138         (check_final_bb): Add SWTCH argument, don't fail if contiguous_range
33139         and only the default case doesn't have the required constants.  Skip
33140         virtual phis.
33141         (gather_default_values): Skip virtual phis.  Allow non-NULL CASE_LOW
33142         if default_case_nonstandard.
33143         (build_constructors): Build constant 1 just once.  Assert that default
33144         values aren't inserted in between cases if contiguous_range.  Skip
33145         virtual phis.
33146         (build_arrays): Skip virtual phis.
33147         (prune_bbs): Add DEFAULT_BB argument, don't remove that bb.
33148         (fix_phi_nodes): Don't add e2f phi arg if default_case_nonstandard.
33149         Handle virtual phis.
33150         (gen_inbound_check): Handle default_case_nonstandard case.
33151         (process_switch): Adjust check_final_bb caller.  Call
33152         gather_default_values with the first non-default case instead of
33153         default case if default_case_nonstandard.
33155 2017-05-02  Nathan Sidwell  <nathan@acm.org>
33157         * ggc-page.c (move_ptes_to_front): Replace unsigned >0 with i--
33158         check.  Fix formatting.
33160 2017-05-02  Jan Hubicka  <hubicka@ucw.cz>
33162         * ipa-inline-analysis.c (estimate_node_size_and_time): Allow roundoff
33163         errors when comparing specialized and unspecialized times.
33165 2017-05-02  David Malcolm  <dmalcolm@redhat.com>
33167         * diagnostic-show-locus.c
33168         (layout::should_print_annotation_line_p): Make private.
33169         (layout::print_annotation_line): Make private.
33170         (layout::annotation_line_showed_range_p): Make private.
33171         (layout::show_ruler): Make private.
33172         (layout::print_source_line): Make private.  Pass in line and
33173         line_width, rather than calling location_get_source_line.  Drop
33174         returned value.
33175         (layout::print_leading_fixits): New method.
33176         (layout::print_any_fixits): Rename to...
33177         (layout::print_trailing_fixits): ...this, and make private.
33178         Don't print newline fixits.
33179         (diagnostic_show_locus): Move logic for printing one row into...
33180         (layout::print_line): ...this new function.  Move the
33181         location_get_source_line call and error-handling from
33182         print_source_line to here.  Call print_leading_fixits, and rename
33183         print_any_fixits to print_trailing_fixits.
33184         (selftest::test_fixit_insert_containing_newline): Update now that
33185         newlines are partially supported.
33186         (selftest::test_fixit_insert_containing_newline_2): New test.
33187         (selftest::test_fixit_replace_containing_newline): Update comments.
33188         (selftest::diagnostic_show_locus_c_tests): Call the new test.
33189         * edit-context.c (class added_line): New class.
33190         (class edited_line): Describe newline handling in comment.
33191         (edited_line::actually_edited_p): New method.
33192         (edited_line::print_content): Delete redundant decl.
33193         (edited_line::m_predecessors): New field.
33194         (edited_file::print_content): Call edited_line::print_content.
33195         (edited_file::print_diff): Update to support newlines.
33196         (edited_file::print_diff_hunk): Likewise.
33197         (edited_file::print_run_of_changed_lines): New function.
33198         (edited_file::print_diff_line): Convert to...
33199         (print_diff_line): ...this.
33200         (edited_file::get_effective_line_count): New function.
33201         (edited_line::edited_line): Initialize new field m_predecessors.
33202         (edited_line::~edited_line): Clean up m_predecessors.
33203         (edited_line::apply_fixit): Handle newlines.
33204         (edited_line::get_effective_line_count): New function.
33205         (edited_line::print_content): New function.
33206         (edited_line::print_diff_lines): New function.
33207         (selftest::test_applying_fixits_insert_containing_newline): New
33208         test.
33209         (selftest::test_applying_fixits_replace_containing_newline): New
33210         test.
33211         (selftest::insert_line): New function.
33212         (selftest::test_applying_fixits_multiple_lines): Add example of
33213         inserting a line.
33214         (selftest::edit_context_c_tests): Call the new tests.
33216 2017-05-02  Bin Cheng  <bin.cheng@arm.com>
33218         * tree-ssa-loop-ivopts.c (get_scaled_computation_cost_at): Delete
33219         parameter cand.  Update dump information.
33220         (get_computation_cost): Update uses.
33222 2017-05-02  Bin Cheng  <bin.cheng@arm.com>
33224         * tree-ssa-loop-ivopts.c (get_computation_aff_1): New.
33225         (get_computation_aff): Reorder parameters.  Use get_computation_aff_1.
33226         (get_computation_at, rewrite_use_address): Update use of
33227         get_computation_aff.
33229 2017-05-02  Bin Cheng  <bin.cheng@arm.com>
33231         * tree-ssa-loop-ivopts.c (get_computation_at): Reorder parameters.
33232         (get_computation): Delete.
33233         (get_computation_cost): Implement like get_computation_cost_at.
33234         Use get_computation_at.
33235         (get_computation_cost_at): Delete.
33236         (rewrite_use_nonlinear_expr): Use get_computation_at.
33237         (rewrite_use_compare, remove_unused_ivs): Ditto.
33239 2017-05-02  Bin Cheng  <bin.cheng@arm.com>
33241         * tree-ssa-loop-ivopts.c (rewrite_use_address): Simple refactor.
33243 2017-05-02  Bin Cheng  <bin.cheng@arm.com>
33245         * tree-ssa-loop-ivopts.c (struct iv_ca): Rename n_regs to n_invs.
33246         (ivopts_global_cost_for_size): Rename parameter and update uses.
33247         (iv_ca_recount_cost): Update uses.
33248         (iv_ca_set_remove_invs, iv_ca_set_no_cp): Record invariants and
33249         candidates seperately in n_invs and n_cands.
33250         (iv_ca_set_add_invs, iv_ca_set_cp, iv_ca_new): Ditto.
33252 2017-05-02  Bin Cheng  <bin.cheng@arm.com>
33254         * tree-ssa-loop-ivopts.c (struct walk_tree_data): New.
33255         (find_inv_vars_cb): New.
33256         (find_depends): Renamed to ...
33257         (find_inv_vars): ... this.
33258         (add_candidate_1, force_var_cost): Call find_inv_vars.
33259         (split_address_cost, determine_group_iv_cost_cond): Ditto.
33261 2017-05-02  Bin Cheng  <bin.cheng@arm.com>
33263         * tree-ssa-loop-ivopts.c (struct cost_pair): Rename depends_on to
33264         inv_vars.  Add inv_exprs.
33265         (struct iv_cand): Rename depends_on to inv_vars.
33266         (struct ivopts_data): Rename max_inv_id/n_invariant_uses to
33267         max_inv_var_id/n_inv_var_uses.  Move max_inv_expr_id around.
33268         Refactor field used_inv_exprs from has_map to array n_inv_expr_uses.
33269         (dump_cand): Dump inv_vars.
33270         (tree_ssa_iv_optimize_init): Support inv_vars and inv_exprs.
33271         (record_invariant, find_depends, add_candidate_1): Ditto.
33272         (set_group_iv_cost, force_var_cost): Ditto.
33273         (split_address_cost, ptr_difference_cost, difference_cost): Ditto.
33274         (get_computation_cost_at, get_computation_cost): Ditto.
33275         (determine_group_iv_cost_generic): Ditto.
33276         (determine_group_iv_cost_address): Ditto.
33277         (determine_group_iv_cost_cond, autoinc_possible_for_pair): Ditto.
33278         (determine_group_iv_costs): Ditto.
33279         (iv_ca_recount_cost): Update call to ivopts_global_cost_for_size.
33280         (iv_ca_set_remove_invariants): Renamed to ...
33281         (iv_ca_set_remove_invs): ... this.  Support inv_vars and inv_exprs.
33282         (iv_ca_set_no_cp): Use iv_ca_set_remove_invs.
33283         (iv_ca_set_add_invariants):  Renamed to ...
33284         (iv_ca_set_add_invs): ... this.  Support inv_vars and inv_exprs.
33285         (iv_ca_set_cp): Use iv_ca_set_add_invs.
33286         (iv_ca_has_deps): Support inv_vars and inv_exprs.
33287         (iv_ca_new, iv_ca_free, iv_ca_dump, free_loop_data): Ditto.
33288         (create_new_ivs): Remove useless dump.
33290 2017-05-02  Bin Cheng  <bin.cheng@arm.com>
33292         * tree-ssa-loop-ivopts.c (get_computation_cost_at): Remove pseudo
33293         iv_cand code.
33294         (determine_group_iv_cost_cond, determine_iv_cost): Ditto.
33295         (iv_ca_set_no_cp, create_new_iv): Ditto.
33297 2017-05-02  Bin Cheng  <bin.cheng@arm.com>
33299         * rtlanal.c (rtx_cost): Handle TRUNCATE between tieable modes.
33301 2017-05-02  Xi Ruoyao  <ryxi@stu.xidian.edu.cn>
33303         * tree.h (EXPR_CILK_SPAWN): Use macro TREE_CHECK2 instead of
33304         function tree_check2.
33306 2017-05-02  Martin Liska  <mliska@suse.cz>
33308         * doc/gcov.texi: Add missing preposition.
33309         * gcov.c (function_info::function_info): Properly fill up
33310         all member variables.
33312 2017-05-02  Tamar Christina  <tamar.christina@arm.com>
33314         * expr.c (expand_expr_real_2): Re-cost if previous costs are the same.
33316 2017-05-02  Tamar Christina  <tamar.christina@arm.com>
33318         * simplify-rtx.c (simplify_binary_operation_1): Add LSHIFTRT case.
33320 2017-05-02  Martin Liska  <mliska@suse.cz>
33322         PR lto/77954.
33323         * lto-streamer-in.c (lto_read_tree_1): Remove
33324         LTO_STREAMER_DEBUG.
33325         * lto-streamer.c (struct tree_hash_entry): Likewise.
33326         (struct tree_entry_hasher): Likewise.
33327         (tree_entry_hasher::hash): Likewise.
33328         (tree_entry_hasher::equal): Likewise.
33329         (lto_streamer_init): Likewise.
33330         (lto_orig_address_map): Likewise.
33331         (lto_orig_address_get): Likewise.
33332         (lto_orig_address_remove): Likewise.
33333         * lto-streamer.h: Likewise.
33334         * tree-streamer-in.c (streamer_alloc_tree): Likewise.
33335         * tree-streamer-out.c (streamer_write_tree_header): Likewise.
33337 2017-05-02  Sebastian Peryt  <sebastian.peryt@intel.com>
33339         * config/i386/avx512fintrin.h (_mm_mask_add_round_sd)
33340         (_mm_maskz_add_round_sd, _mm_mask_add_round_ss)
33341         (mm_maskz_add_round_ss, _mm_mask_sub_round_sd)
33342         (mm_maskz_sub_round_sd, _mm_mask_sub_round_ss)
33343         (mm_maskz_sub_round_ss, _mm_mask_add_sd)
33344         (mm_maskz_add_sd, _mm_mask_add_ss, _mm_maskz_add_ss)
33345         (mm_mask_sub_sd, _mm_maskz_sub_sd, _mm_mask_sub_ss)
33346         (mm_maskz_sub_ss): New intrinsics.
33347         * config/i386/i386-builtin-types.def (V2DF_FTYPE_V2DF_V2DF_V2DF_UQI_INT)
33348         (V4SF_FTYPE_V4SF_V4SF_V4SF_UQI_INT): New function type aliases.
33349         * config/i386/i386-builtin.def (__builtin_ia32_addsd_mask_round)
33350         (__builtin_ia32_addss_mask_round, __builtin_ia32_subsd_mask_round)
33351         (__builtin_ia32_subss_mask_round): New builtins.
33352         * config/i386/i386.c (V2DF_FTYPE_V2DF_V2DF_V2DF_UQI_INT)
33353         (V4SF_FTYPE_V4SF_V4SF_V4SF_UQI_INT): Handle new types.
33354         * config/i386/sse.md (<sse>_vm<plusminus_insn><mode>3<round_name>):
33355         Renamed to ...
33356         (<sse>_vm<plusminus_insn><mode>3<mask_name><round_name>): ... this.
33357         (v<plusminus_mnemonic><ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|%0, %1, %<iptr>2<round_op3>}):
33358         Changed to ...
33359         (v<plusminus_mnemonic><ssescalarmodesuffix>\t{<round_mask_op3>%2, %1, %0<mask_operand3>|%0<mask_operand3>, %1, %<iptr>2<round_mask_op3>}):
33360         ... this.
33362 2017-05-02  Martin Jambor  <mjambor@suse.cz>
33364         PR tree-optimization/78687
33365         * tree-sra.c (access): New field parent.
33366         (process_subtree_disqualification): New function.
33367         (disqualify_candidate): Call it.
33368         (build_accesses_from_assign): Reset write flag if creating an
33369         assighnment link.
33370         (build_access_subtree): Fill in parent field and also prpagate
33371         down grp_write flag.
33372         (create_artificial_child_access): New parameter set_grp_write, set
33373         grp_write to its value.
33374         (propagate_subaccesses_across_link): Also propagate grp_write flag
33375         values.
33376         (propagate_all_subaccesses): Push the closest parent back to work
33377         queue if add_access_to_work_queue returned true.
33379 2017-05-02  Richard Biener  <rguenther@suse.de>
33381         * common.opt (fstrict-overflow): Alias negative to fwrapv.
33382         * doc/invoke.texi (fstrict-overflow): Remove all traces of
33383         -fstrict-overflow documentation.
33384         * tree.h (TYPE_OVERFLOW_UNDEFINED): Do not test flag_strict_overflow.
33385         (POINTER_TYPE_OVERFLOW_UNDEFINED): Test !flag_wrapv instead of
33386         flag_strict_overflow.
33387         * ipa-inline.c (can_inline_edge_p): Do not test flag_strict_overflow.
33388         * lto-opts.c (lto_write_options): Do not stream it.
33389         * lto-wrapper.c (merge_and_complain): Do not handle it.
33390         * opts.c (default_options_table): Do not set -fstrict-overflow.
33391         (finish_options): Likewise do not clear it when sanitizing.
33392         * simplify-rtx.c (simplify_const_relational_operation): Do not
33393         test flag_strict_overflow.
33395 2017-05-02  Uros Bizjak  <ubizjak@gmail.com>
33397         * config/alpha/alpha.md (*add<mode>3_ieee): Merge to add<mode>3
33398         using enabled attribute.
33399         (*sub<mode>3_ieee): Merge to sub<mode>3 using enabled attribute.
33400         (*mul<mode>3_ieee): Merge to mul<mode>3 using enabled attribute.
33401         (*div<mode>3_ieee): Merge to div<mode>3 using enabled attribute.
33402         (*sqrt<mode>2_ieee): Merge to sqrt<mode>2 using enabled attribute.
33403         (*fix_truncdfdi_ieee): Merge to *fix_truncdfdi2 using enabled attribute.
33404         (*fix_truncsfdi_ieee): Merge to *fix_truncsfdi2 using enabled attribute.
33405         (*floatdisf_ieee): Merge to floatdisf2 using enabled attribute.
33406         (*floatdidf_ieee): Merge to floatdidf2 using enabled attribute.
33407         (*truncdfsf2_ieee): Merge to truncdfsf2 using enabled attribute.
33408         (*cmpdf_ieee): Merge to *cmpdf_internal using enabled attribute.
33410 2017-05-02  Uros Bizjak  <ubizjak@gmail.com>
33412         * config/i386/i386.c (ix86_code_end): Use {FIRST,LAST}_INT_REG.
33414 2017-05-02  Richard Biener  <rguenther@suse.de>
33416         PR tree-optimization/80591
33417         Revert
33418         2017-04-10  Richard Biener  <rguenther@suse.de>
33420         * tree-ssa-structalias.c (find_func_aliases): Properly handle
33421         asm inputs.
33423 2017-05-02  Richard Biener  <rguenther@suse.de>
33425         PR tree-optimization/80549
33426         * tree-cfgcleanup.c (mfb_keep_latches): New helper.
33427         (cleanup_tree_cfg_noloop): Create forwarders to known loop
33428         headers if they do not have a preheader.
33430 2017-05-02  Martin Liska  <mliska@suse.cz>
33432         PR other/80589
33433         * common.opt: Fix typo.
33434         * doc/invoke.texi: Likewise.
33436 2017-05-01  Jan Beulich  <jbeulich@suse.com>
33438         * config/i386/sse.md (xop_vpermil2<mode>3): Do not allow operand
33439         swapping, add (x,x,m,x,n) alternative.
33441 2017-05-01  Nathan Sidwell  <nathan@acm.org>
33443         * calls.c (combine_pending_stack_adjustment_and_call): Remove
33444         unnecessary unadjusted_alignment check.
33446 2017-05-01  Xi Ruoyao  <ryxi@stu.xidian.edu.cn>
33448         PR c++/80038
33449         * cilk_common.c (expand_builtin_cilk_detach): Move pedigree
33450         operations here.
33451         * gimplify.c (gimplify_cilk_detach): New function.
33452         (gimplify_call_expr, gimplify_modify_expr): Call it as needed.
33453         * tree-core.h: Document EXPR_CILK_SPAWN.
33454         * tree.h (EXPR_CILK_SPAWN): Define.
33456 2017-05-01  David Malcolm  <dmalcolm@redhat.com>
33458         * diagnostic-show-locus.c (layout::get_expanded_location): Rewrite
33459         to use new fixit_hint representation, using the "replace" logic.
33460         (get_line_span_for_fixit_hint): Likewise.
33461         (layout::print_any_fixits): Likewise.
33462         (selftest::test_one_liner_many_fixits): Rename to...
33463         (selftest::test_one_liner_many_fixits_1): ...this, and update
33464         comment and expected output to reflect that the multiple fix-it
33465         hints are now consolidated into one insertion.
33466         (selftest::test_one_liner_many_fixits_2): New test.
33467         (selftest::test_diagnostic_show_locus_one_liner): Update for
33468         above.
33469         (selftest::test_fixit_consolidation): Update for fix-it API
33470         change.
33471         * diagnostic.c (print_parseable_fixits): Likewise.
33472         * edit-context.c (edited_line::m_line_events): Convert from
33473         auto_vec <line_event *> to auto_vec <line_event>.
33474         (class line_event): Convert from abstract base class to a concrete
33475         class, taking over the role of replace_event.
33476         (class insert_event): Delete.
33477         (class replace_event): Rename to class line_event.  Convert to
33478         half-open range.
33479         (edit_context::add_fixits): Reimplement.
33480         (edit_context::apply_insert): Delete.
33481         (edit_context::apply_replace): Rename to...
33482         (edit_context::apply_fixit): ...this.  Convert to half-open range.
33483         (edited_file::apply_insert): Delete.
33484         (edited_file::apply_replace): Rename to...
33485         (edited_file::apply_fixit): ...this.
33486         (edited_line::~edited_line): Drop deletion of events.
33487         (edited_line::apply_insert): Delete.
33488         (edited_line::apply_replace): Rename to...
33489         (edited_line::apply_fixit): ...this.  Convert to half-open range.
33490         Update for change to type of m_line_events.
33491         * edit-context.h (edit_context::apply_insert): Delete.
33492         (edit_context::apply_replace): Rename to...
33493         (edit_context::apply_fixit): ...this.
33495 2017-05-01  Martin Sebor  <msebor@redhat.com>
33497         * gimple-ssa-sprintf.c (format_integer): Set knownrange when it's
33498         known.
33500 2017-05-01  Uros Bizjak  <ubizjak@gmail.com>
33502         PR target/68491
33503         * config/i386/cpuid.h (__get_cpuid): Always return 0 when
33504         __get_cpuid_max returns 0.
33505         (__get_cpuid_count): Ditto.
33507 2017-05-01  Eric Botcazou  <ebotcazou@adacore.com>
33509         * tree.c (substitute_in_expr) <tcc_vl_exp>: Also inline a call if the
33510         replacement expression is another instance of one of its arguments.
33512 2017-05-01  Jakub Jelinek  <jakub@redhat.com>
33514         PR target/79430
33515         * rtlanal.c (reg_set_p): If reg is a stack_pointer_rtx, also
33516         check for stack push/pop autoinc.
33517         * config/i386/i386.c (ix86_agi_dependent): Return false
33518         if the only reason why modified_in_p returned true is that
33519         addr is SP based and set_insn is a push or pop.
33521 2017-04-29  Jan Hubicka  <hubicka@ucw.cz>
33523         * ipa-inline.c (compute_inlined_call_time): Remove now unnecesary
33524         overflow check.
33526 2017-04-29  Jan Hubicka  <hubicka@ucw.cz>
33528         PR ipa/79224
33529         * ipa-inline-analysis.c (dump_predicate): Add optional parameter NL.
33530         (account_size_time): Use two predicates - exec_pred and
33531         nonconst_pred_ptr.
33532         (evaluate_conditions_for_known_args): Compute both clause and
33533         nonspec_clause.
33534         (evaluate_properties_for_edge): Evaulate both clause and nonspec_clause.
33535         (inline_summary_t::duplicate): Update.
33536         (estimate_function_body_sizes): Caluculate exec and nonconst predicates
33537         separately.
33538         (compute_inline_parameters): Likewise.
33539         (estimate_edge_size_and_time): Update caluclation of time.
33540         (estimate_node_size_and_time): Compute both time and nonspecialized
33541         time.
33542         (estimate_ipcp_clone_size_and_time): Update.
33543         (inline_merge_summary): Update.
33544         (do_estimate_edge_time): Update.
33545         (do_estimate_edge_size): Update.
33546         (do_estimate_edge_hints): Update.
33547         (inline_read_section, inline_write_summary): Stream both new predicates.
33548         * ipa-inline.c (compute_uninlined_call_time): Take uninlined_call_time
33549         as argument.
33550         (compute_inlined_call_time): Cleanup.
33551         (big_speedup_p): Update.
33552         (edge_badness): Update.
33553         * ipa-inline.h (INLINE_TIME_SCALE): Remove.
33554         (size_time_entry): Replace predicate by exec_predicate and
33555         nonconst_predicate.
33556         (edge_growth_cache_entry): Cache both time nad nonspecialized time.
33557         (estimate_edge_time): Return also nonspec_time.
33558         (reset_edge_growth_cache): Update.
33560 2017-04-29  Jakub Jelinek  <jakub@redhat.com>
33562         PR rtl-optimization/80491
33563         * ifcvt.c (noce_process_if_block): When looking for x setter
33564         with missing else_bb, don't check only the insn right before
33565         cond_earliest, but look for the last insn that x is modified in
33566         within the same bb.
33568         PR rtl-optimization/80491
33569         * alias.c (memory_modified_in_insn_p): Return true for CALL_INSNs.
33571 2017-04-29  Marc Glisse  <marc.glisse@inria.fr>
33573         PR tree-optimization/80487
33574         * tree-ssa-alias.c (stmt_kills_ref_p): Handle stpncpy and strncpy.
33576 2017-04-29  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
33578         PR tree-optimization/79697
33579         * tree-ssa-dce.c (mark_stmt_if_obviously_necessary): Check if callee
33580         is BUILT_IN_STRDUP, BUILT_IN_STRNDUP, BUILT_IN_REALLOC.
33581         (propagate_necessity): Check if def_callee is BUILT_IN_STRDUP or
33582         BUILT_IN_STRNDUP.
33583         * gimple-fold.c (gimple_fold_builtin_realloc): New function.
33584         (gimple_fold_builtin): Call gimple_fold_builtin_realloc.
33586 2017-04-28  Martin Sebor  <msebor@redhat.com>
33588         PR tree-optimization/80523
33589         * gimple-ssa-sprintf.c (target_to_host_charmap): New global variable.
33590         (init_target_to_host_charmap, target_to_host, target_strtol10): New
33591         functions.
33592         (maybe_warn, format_directive, parse_directive): Use new functions.
33593         (pass_sprintf_length::execute): Call init_target_to_host_charmap.
33595 2017-04-28  Marc Glisse  <marc.glisse@inria.fr>
33597         * match.pd (X+Z OP Y+Z, X-Z OP Y-Z, Z-X OP Z-Y): New transformations.
33599 2017-04-28  Bernd Edlinger  <bernd.edlinger@hotmail.de>
33601         * configure.ac (SYSTEM_HEADER_DIR, BUILD_SYSTEM_HEADER_DIR,
33602         target_header_dir): Set correctly.
33603         * configure: Regenerated.
33604         * Makefile.in (BUILD_SYSTEM_HEADER_DIR): New make variabe.
33605         (LIMITS_H_TEST, if_multiarch, stmp-fixinc): Use BUILD_SYSTEM_HEADER_DIR
33606         instead of SYSTEM_HEADER_DIR.
33608 2017-04-28  Jan Hubicka  <hubicka@ucw.cz>
33610         * ipa-cp.c (perform_estimation_of_a_value): Turn time to sreal.
33611         (estimate_local_effects): Likewise.
33612         * ipa-inline.c (compute_inlined_call_time, want_inline_small_function_p,
33613         edge_badness, inline_small_functions, dump_overall_stats): LIkewise.
33614         * ipa-inline.h (edge_growth_cache_entry, estimate_time_after_inlining,
33615         estimate_ipcp_clone_size_and_time, do_estimate_edge_time,
33616         do_estimate_edge_time, estimate_edge_time): Likewise.
33617         * ipa-inline-analysis.c (estimate_node_size_and_time,
33618         estimate_ipcp_clone_size_and_time, do_estimate_edge_time): Likewise.
33619         (estimate_time_after_inlining): Remove.
33621 2017-04-28  Martin Liska  <mliska@suse.cz>
33623         * doc/gcov.texi: Enhance documentation of gcov.
33625 2017-04-28  Martin Liska  <mliska@suse.cz>
33627         * doc/gcov.texi: Sort options in alphabetic order.
33628         * doc/gcov-dump.texi: Likewise.
33629         * doc/gcov-tool.texi: Likewise.
33630         * gcov.c (print_usage): Likewise.
33631         * gcov-dump.c (print_usage): Likewise.
33632         * gcov-tool.c (print_merge_usage_message): Likewise.
33633         (print_rewrite_usage_message): Likewise.
33634         (print_overlap_usage_message): Likewise.
33636 2017-04-28  Martin Liska  <mliska@suse.cz>
33638         PR gcov-profile/53915
33639         * gcov.c (format_gcov): Print 'NAN %' when top > bottom.
33641 2017-04-28  Martin Liska  <mliska@suse.cz>
33643         PR gcov-profile/79891
33644         * gcov.c (add_line_counts): Assign BBs to lines just if the BB
33645         is marked by compiler as living on a line.
33646         (get_cycles_count): Remove usage of the union.
33647         (output_intermediate_file): Likewise.
33648         (find_source): Fix GNU coding style.
33649         (accumulate_line_counts): Remove old non-all block mode.
33650         (output_lines): Remove usage of the union.
33651         * profile.c (output_location): Include all BBs, even if
33652         belonging to a same line (and file) as a previous BB.
33654 2017-04-28  Martin Liska  <mliska@suse.cz>
33656         * gcov.c (process_args): Handle new argument 'w'.
33657         (read_graph_file): Assign ID to BBs.
33658         (output_branch_count): Display BB # if verbose flag is set.
33659         (output_lines): Likewise for arcs.
33660         (print_usage): Add '--verbose' option help.
33661         * doc/gcov.texi: Document --verbose (-w) option.
33663 2017-04-28  Martin Liska  <mliska@suse.cz>
33665         * gcov.c (struct block_location_info): New struct.
33666         (process_file): Fill up the new structure.
33667         (read_graph_file): Replace usage of encoding by the newly added
33668         struct.
33669         (add_line_counts): Likewise.
33670         (accumulate_line_counts): Remove usage of the union.
33671         (function_info::function_info): New function.
33672         (function_info::~function_info): Likewise.
33673         (process_file): Call delete instead of release_function.
33674         (release_function): Release the function.
33675         (release_structures): Call delete instead of release_function.
33676         (solve_flow_graph): Replace usage of num_blocks.
33677         (find_exception_blocks): Likewise.
33678         (output_lines): Fix GNU coding style.
33680 2017-04-28  Martin Liska  <mliska@suse.cz>
33682         PR driver/56469
33683         * coverage.c (coverage_remove_note_file): New function.
33684         * coverage.h: Declare the function.
33685         * toplev.c (finalize): Clean if an error has been seen.
33687 2017-04-28  Martin Liska  <mliska@suse.cz>
33689         PR gcov-profile/80031
33690         * gcov-dump.c (tag_blocks): Just print number of basic blocks.
33691         * gcov-io.h (GCOV_TAG_BLOCKS_NUM): Remove unused macro.
33692         * gcov.c (read_graph_file): Read just number of blocks.
33693         * profile.c (branch_prob): Do not stream 0 flags per a basic
33694         block.
33696 2017-04-28  Martin Liska  <mliska@suse.cz>
33698         * gcov-dump.c (tag_*): Add new argument to declarations.
33699         (dump_gcov_file): Likewise.
33700         (tag_blocks): Add and use new argument depth.
33701         (tag_arcs): Likewise.
33702         (tag_lines): Likewise.
33703         (tag_counters): Likewise.
33704         (tag_summary): Likewise.
33705         (dump_working_sets): Use depth to do a proper indentation.
33707 2017-04-28  Jakub Jelinek  <jakub@redhat.com>
33709         PR bootstrap/80531
33710         * cgraph.h (symtab_node::debug_symtab): No longer inline.
33711         * symtab.c (symtab_node::debug_symtab): Move definition here.
33713 2017-04-28  Richard Biener  <rguenther@suse.de>
33715         * lto-streamer.h (LTO_major_version): Bump to 7.
33717 2017-04-28  Richard Biener  <rguenther@suse.de>
33719         * tree-vrp.c (assert_info): New struct.
33720         (add_assert_info): New helper.
33721         (register_edge_assert_for_2): Refactor to add asserts to a vector
33722         of assert_info.
33723         (register_edge_assert_for_1): Likewise.
33724         (register_edge_assert_for): Likewise.
33725         (finish_register_edge_assert_for): New helper actually registering
33726         asserts where live on edge.
33727         (find_conditional_asserts): Adjust.
33728         (find_switch_asserts): Likewise.
33729         (evrp_dom_walker::try_find_new_range): Generalize.
33730         (evrp_dom_walker::before_dom_children): Use register_edge_assert_for.
33732 2017-04-27  Marek Polacek  <polacek@redhat.com>
33734         PR sanitizer/80349
33735         * fold-const.c (fold_binary_loc) <case EQ_EXPR, NE_EXPR>: Convert
33736         arg10 and arg11 to itype.
33738 2017-04-27  Jonathan Wakely  <jwakely@redhat.com>
33740         * doc/extend.texi (Object Size Checking): Improve grammar.
33742 2017-04-27  Richard Earnshaw  <rearnsha@arm.com>
33744         PR target/80530
33745         * config/aarch64/aarch64.c (aarch64_emit_approx_sqrt): Ensure
33746         that the logic for permitting reciprocal estimates matches that
33747         in use_rsqrt_p.
33749 2017-04-27  Jakub Jelinek  <jakub@redhat.com>
33751         PR c++/80534
33752         * tree.c (type_cache_hasher::equal): Only compare
33753         TYPE_TYPELESS_STORAGE flag on non-aggregate element types.
33754         (build_array_type_1): Only hash TYPE_TYPELESS_STORAGE flag on
33755         non-aggregate element types.
33756         * tree.h (TYPE_TYPELESS_STORAGE): Fix comment typo, add more details
33757         about the flag on ARRAY_TYPEs in the comment, formatting fix.
33759 2017-04-27  Richard Biener  <rguenther@suse.de>
33761         PR middle-end/80533
33762         * emit-rtl.c (set_mem_attributes_minus_bitpos): When
33763         stripping ARRAY_REFs from MEM_EXPR make sure we're not
33764         keeping a reference to a trailing array.
33766 2017-04-27  Richard Biener  <rguenther@suse.de>
33768         PR middle-end/80539
33769         * tree-chrec.c (chrec_fold_plus_poly_poly): Deal with not
33770         being in loop-closed SSA form conservatively.
33771         (chrec_fold_multiply_poly_poly): Likewise.
33773 2017-04-27  Tamar Christina  <tamar.christina@arm.com>
33775         PR middle-end/79665
33776         * expr.c (expand_expr_real_2): Move TRUNC_MOD_EXPR, FLOOR_MOD_EXPR,
33777         CEIL_MOD_EXPR, ROUND_MOD_EXPR cases.
33779 2017-04-27  Jakub Jelinek  <jakub@redhat.com>
33781         PR target/77728
33782         * config/aarch64/aarch64.c (struct aarch64_fn_arg_alignment): Remove.
33783         (aarch64_function_arg_alignment): Return unsigned int again, but still
33784         ignore TYPE_FIELDS chain decls other than FIELD_DECLs.
33785         (aarch64_layout_arg): Adjust aarch64_function_arg_alignment caller.
33786         Don't emit -Wpsabi note.
33787         (aarch64_function_arg_boundary): Likewise.
33788         (aarch64_gimplify_va_arg_expr): Adjust aarch64_function_arg_alignment
33789         caller.
33791 2017-04-26  Nathan Sidwell  <nathan@acm.org>
33793         * tree.h (crc32_unsigned_n): Declare.
33794         (crc32_unsigned, crc32_unsigned): Make inline.
33795         * tree.c (crc32_unsigned_bits): Replace with ...
33796         (crc32_unsigned_n): ... this.
33797         (crc32_unsigned, crc32_byte): Remove.
33798         (crc32_string): Remove unnecessary braces.
33800 2017-04-25  Jan Hubicka  <hubicka@ucw.cz>
33802         * ipa-cp.c (estimate_local_effects): Convert sreal to int.
33803         * ipa-inline-analysis.c (MAX_TIME): Remove.
33804         (account_size_time): Use sreal for time.
33805         (dump_inline_summary): Update.
33806         (estimate_function_body_sizes): Update.
33807         (estimate_edge_size_and_time): Update.
33808         (estimate_calls_size_and_time): Update.
33809         (estimate_node_size_and_time): Update.
33810         (inline_merge_summary): Update.
33811         (inline_update_overall_summary): Update.
33812         (estimate_time_after_inlining): Update.
33813         (inline_read_section): Update.
33814         (inline_write_summary): Update.
33815         * ipa-inline.c (compute_uninlined_call_time): Update.
33816         (compute_inlined_call_time): Update.
33817         (recursive_inlining): Update.
33818         (inline_small_functions): Update.
33819         (dump_overall_stats): Update.
33820         * ipa-inline.h: Include sreal.h.
33821         (size_time_entry): Turn time to sreal.
33822         (inline_summary): Turn self_time nad time to sreal.
33824 2017-04-25  Jan Hubicka  <hubicka@ucw.cz>
33826         * sreal.c: Include backend.h, tree.h, gimple.h, cgraph.h and
33827         data-streamer.h
33828         (sreal::stream_out, sreal::stream_in): New.
33829         * sreal.h (sreal::stream_out, sreal::stream_in): Declare.
33831 2017-04-25  Jakub Jelinek  <jakub@redhat.com>
33833         * Makefile.in (s-options): Invoke opt-gather.awk with LC_ALL=C in the
33834         environment.
33836 2017-04-25  Uros Bizjak  <ubizjak@gmail.com>
33838         PR target/70799
33839         * config/i386/i386.c (dimode_scalar_to_vector_candidate_p):
33840         Handle ASHIFTRT.
33841         (dimode_scalar_chain::compute_convert_gain): Ditto.
33842         (dimode_scalar_chain::make_vector_copies): Ditto.
33843         (dimode_scalar_chain::convert_reg): Ditto.
33844         (dimode_scalar_chain::convert_insn): Ditto.
33845         * config/i386/sse.md (VI24_AVX512BW_1): Remove mode iterator.
33846         (VI248_AVX512BW_1): New mode iterator.
33847         (<mask_codefor>ashr<mode>3<mask_name>): Merge insn pattern with
33848         <mask_codefor>ashrv2di3<mask_name> insn using VI248_AVX512BW_1
33849         mode iterator.
33851 2017-04-25  Martin Sebor  <msebor@redhat.com>
33853         PR tree-optimization/80497
33854         * gimple-ssa-sprintf.c (get_int_range): Avoid assuming all integer
33855         constants are representable in HOST_WIDE_INT.
33856         (parse_directive): Ditto.
33858 2017-04-25  Martin Sebor  <msebor@redhat.com>
33860         PR bootstrap/80486
33861         * dominance.c (dom_info::m_n_basic_blocks): Change type to unsigned.
33862         (new_zero_array): Adjust signature.
33863         (dom_info::dom_init): Used unsigned rather that size_t.
33864         (dom_info::dom_info): Same.
33866 2017-04-25  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
33867             Jakub Jelinek  <jakub@redhat.com>
33869         PR target/77728
33870         * config/arm/arm.c: Include gimple.h.
33871         (aapcs_layout_arg): Emit -Wpsabi note if arm_needs_doubleword_align
33872         returns negative, increment ncrn only if it returned positive.
33873         (arm_needs_doubleword_align): Return int instead of bool,
33874         ignore DECL_ALIGN of non-FIELD_DECL TYPE_FIELDS chain
33875         members, but if there is any such non-FIELD_DECL
33876         > PARM_BOUNDARY aligned decl, return -1 instead of false.
33877         (arm_function_arg): Emit -Wpsabi note if arm_needs_doubleword_align
33878         returns negative, increment nregs only if it returned positive.
33879         (arm_setup_incoming_varargs): Likewise.
33880         (arm_function_arg_boundary): Emit -Wpsabi note if
33881         arm_needs_doubleword_align returns negative, return
33882         DOUBLEWORD_ALIGNMENT only if it returned positive.
33884 2017-04-25  Marek Polacek  <polacek@redhat.com>
33886         PR sanitizer/80349
33887         * fold-const.c (fold_binary_loc) <case BIT_IOR_EXPR>: Convert arg0's
33888         first argument to type.
33890 2017-04-25  Bill Seurer  <seurer@linux.vnet.ibm.com>
33892         PR target/80482
33893         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Change
33894         type checks to test for compatibility instead of equality.
33896 2017-04-25  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
33897             Jakub Jelinek  <jakub@redhat.com>
33899         PR target/77728
33900         * config/aarch64/aarch64.c (struct aarch64_fn_arg_alignment): New
33901         type.
33902         (aarch64_function_arg_alignment): Return aarch64_fn_arg_alignment
33903         struct.  Ignore DECL_ALIGN of decls other than FIELD_DECL for
33904         the alignment computation, but return their maximum in warn_alignment.
33905         (aarch64_layout_arg): Adjust aarch64_function_arg_alignment caller.
33906         Emit a -Wpsabi note if warn_alignment is 16 bytes, but alignment
33907         is smaller.
33908         (aarch64_function_arg_boundary): Likewise.  Simplify using MIN/MAX.
33909         (aarch64_gimplify_va_arg_expr): Adjust aarch64_function_arg_alignment
33910         caller.
33912 2017-04-25  Claudiu Zissulescu  <claziss@synopsys.com>
33914         * config/arc/simdext.md (dmpyh): Fix typo.
33916 2017-04-25  Richard Biener  <rguenther@suse.de>
33918         PR tree-optimization/80492
33919         * alias.c (compare_base_decls): Handle registers with asm
33920         specification conservatively.
33921         * tree-ssa-alias.c (decl_refs_may_alias_p): Handle
33922         compare_base_decls returning dont-know properly.
33924 2017-04-25  Claudiu Zissulescu  <claziss@synopsys.com>
33926         * config/arc/arc.c (LEGITIMATE_OFFSET_ADDRESS_P): Delete macro.
33927         (legitimate_offset_address_p): New function.
33928         (arc_legitimate_address_p): Use above function.
33930 2017-04-25  Claudiu Zissulescu  <claziss@synopsys.com>
33932         * config/arc/arc.c (arc_output_mi_thunk): Emit PIC calls.
33934 2017-04-25  Claudiu Zissulescu  <claziss@synopsys.com>
33936         * config/arc/arc.c (arc_conditional_register_usage): Use ACCL,
33937         ACCH registers whenever they are available.
33939 2017-04-25  Claudiu Zissulescu  <claziss@synopsys.com>
33941         * config/arc/arc.c (arc_conditional_register_usage): Make D0, D1
33942         double regs fix when not used.
33944 2017-04-25  Claudiu Zissulescu  <claziss@synopsys.com>
33946         * config/arc/arc.h (REGNO_OK_FOR_BASE_P): Consider also extension
33947         core registers.
33948         (REG_OK_FOR_INDEX_P_NONSTRICT): Likewise.
33949         (REG_OK_FOR_BASE_P_NONSTRICT): Likewise.
33951 2017-04-25  Claudiu Zissulescu  <claziss@synopsys.com>
33953         * config/arc/arc.c (arc_output_addsi): Check for h-register class
33954         when emitting short ADD instructions.
33956 2017-04-25  Claudiu Zissulescu  <claziss@synopsys.com>
33958         * config/arc/arc.md (cmpsi_cc_insn_mixed): Use 'h' register
33959         constraint.
33960         (cmpsi_cc_c_insn): Likewise.
33961         (cbranchsi4_scratch): Compute proper instruction length using
33962         compact_hreg_operand.
33963         * config/arc/predicates.md (compact_hreg_operand): New predicate.
33965 2017-04-25  Richard Biener  <rguenther@suse.de>
33967         PR middle-end/80509
33968         * passes.c (pass_manager::pass_manager): Initialize
33969         m_name_to_pass_map.
33971 2017-04-25  Richard Biener  <rguenther@suse.de>
33973         PR tree-optimization/79201
33974         * tree-ssa-sink.c (statement_sink_location): Handle calls.
33976 2017-04-25  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
33978         PR target/80464
33979         * config/s390/vector.md: Split MEM->GPR vector moves for
33980         non-s_operand addresses.
33982 2017-04-25  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
33984         PR target/79895
33985         * config/s390/predicates.md (reload_const_wide_int_operand): New
33986         predicate.
33987         * config/s390/s390.md ("movti"): Remove d/P alternative.
33988         ("movti_bigconst"): New pattern definition.
33990 2017-04-25  Dominik Vogt  <vogt@linux.vnet.ibm.com>
33992         PR target/80080
33993         * s390-protos.h (s390_expand_cs_hqi): Removed.
33994         (s390_expand_cs, s390_expand_atomic_exchange_tdsi): New prototypes.
33995         * config/s390/s390.c (s390_emit_compare_and_swap): Handle all integer
33996         modes as well as CCZ1mode and CCZmode.
33997         (s390_expand_atomic_exchange_tdsi, s390_expand_atomic): Adapt to new
33998         signature of s390_emit_compare_and_swap.
33999         (s390_expand_cs_hqi): Likewise, make static.
34000         (s390_expand_cs_tdsi): Generate an explicit compare before trying
34001         compare-and-swap, in some cases.
34002         (s390_expand_cs): Wrapper function.
34003         (s390_expand_atomic_exchange_tdsi): New backend specific expander for
34004         atomic_exchange.
34005         (s390_match_ccmode_set): Allow CCZmode <-> CCZ1 mode.
34006         * config/s390/s390.md ("atomic_compare_and_swap<mode>"): Merge the
34007         patterns for small and large integers.  Forbid symref memory operands.
34008         Move expander to s390.c.  Require cc register.
34009         ("atomic_compare_and_swap<DGPR:mode><CCZZ1:mode>_internal")
34010         ("*atomic_compare_and_swap<TDI:mode><CCZZ1:mode>_1")
34011         ("*atomic_compare_and_swapdi<CCZZ1:mode>_2")
34012         ("*atomic_compare_and_swapsi<CCZZ1:mode>_3"): Use s_operand to forbid
34013         symref memory operands.  Remove CC mode and call s390_match_ccmode
34014         instead.
34015         ("atomic_exchange<mode>"): Allow and implement all integer modes.
34017 2017-04-25  Dominik Vogt  <vogt@linux.vnet.ibm.com>
34019         * config/s390/s390.md (define_peephole2): New peephole to help
34020         combining the load-and-test pattern with volatile memory.
34022 2017-04-25  Dominik Vogt  <vogt@linux.vnet.ibm.com>
34024         * config/s390/s390.md ("cstorecc4"): Use load-on-condition and deal
34025         with CCZmode for TARGET_Z196.
34027 2017-04-25  Jakub Jelinek  <jakub@redhat.com>
34029         PR rtl-optimization/80501
34030         * combine.c (make_compound_operation_int): Set subreg_code to SET
34031         even for AND with mask of the sign bit of mode.
34033         PR rtl-optimization/80500
34034         * loop-unroll.c (combine_var_copies_in_loop_exit): Call copy_rtx on
34035         sum's initial value.
34037 2017-04-25  Julian Brown  <julian@codesourcery.com>
34038             Naveen H.S  <Naveen.Hurugalawadi@cavium.com>
34040         * config/aarch64/thunderx2t99.md (thunderx2t99_crc): New Reservation.
34042 2017-04-25  Marc Glisse  <marc.glisse@inria.fr>
34044         * fold-const.c (tree_single_nonzero_warnv_p): Handle SSA_NAME.
34046 2017-04-25  Julian Brown  <julian@codesourcery.com>
34047             Naveen H.S  <Naveen.Hurugalawadi@cavium.com>
34049         * config/aarch64/thunderx2t99.md (thunderx2t99_aes): New Reservation.
34050         (thunderx2t99_sha): New Reservation.
34052 2017-04-25  Julian Brown  <julian@codesourcery.com>
34053             Naveen H.S  <Naveen.Hurugalawadi@cavium.com>
34055         * config/aarch64/aarch64-simd.md (aarch64_simd_vec_set<mode>): Fix
34056         type for 1-element load.
34058 2017-04-24  Marc Glisse  <marc.glisse@inria.fr>
34060         * match.pd (X/[ex]C CMP Y/[ex]C): New transformation.
34062 2017-04-24  Martin Jambor  <mjambor@suse.cz>
34064         PR tree-optimization/80293
34065         * tree-sra.c (scalarizable_type_p): New parameter const_decl, make
34066         char arrays not totally scalarizable if it is false.
34067         (analyze_all_variable_accesses): Pass correct value in the new
34068         parameter.  Add a statistics counter.
34070 2017-04-24  Jan Hubicka  <hubicka@ucw.cz>
34072         PR middle-end/79931
34073         * ipa-devirt.c (dump_possible_polymorphic_call_targets): Fix ICE.
34075 2017-04-24  Richard Biener  <rguenther@suse.de>
34077         PR tree-optimization/80494
34078         * tree-scalar-evolution.c (analyze_scalar_evolution_1): Bail
34079         out for complex types.
34081 2017-04-24  Richard Biener  <rguenther@suse.de>
34083         * tree-ssa-sccvn.h (run_scc_vn): Adjust prototype.
34084         * tree-ssa-sccvn.c (print_scc): Print SCC size.
34085         (extract_and_process_scc_for_name): Never fail but drop SCC to varying.
34086         (DFS): Adjust and never fail.
34087         (sccvn_dom_walker::fail): Remove.
34088         (sccvn_dom_walker::before_dom_children): Adjust.
34089         (run_scc_vn): Likewise and never fail.
34090         * tree-ssa-pre.c (pass_pre::execute): Adjust.
34091         (pass_fre::execute): Likewise.
34093 2017-04-24  Richard Biener  <rguenther@suse.de>
34095         PR tree-optimization/79725
34096         * tree-ssa-sink.c (statement_sink_location): Return whether
34097         failure reason was zero uses.  Move that check later.
34098         (sink_code_in_bb): Deal with zero uses by removing the stmt
34099         if possible.
34101 2017-04-24  Richard Biener  <rguenther@suse.de>
34103         PR c++/2972
34104         * tree-ssa-uninit.c (warn_uninitialized_vars): Handle some
34105         pointer-based references.
34107 2017-04-24  Richard Biener  <rguenther@suse.de>
34109         PR bootstrap/79814
34110         * pass_manager.h (pass_manager::operator new): Remove.
34111         (pass_manager::operator delete): Likewise.
34112         * passes.c (pass_manager::operator new): Remove.
34113         (pass_manager::operator delete): Likewise.
34114         (pass_manager::pass_manager): Zero individual pass members.
34116 2017-04-23  Uros Bizjak  <ubizjak@gmail.com>
34118         PR target/70799
34119         * config/i386/i386.c (dimode_scalar_to_vector_candidate_p)
34120         <case ASHIFT, case LSHIFTRT>: Also consider variable shifts.
34121         Check "XEXP (src, 1)" operand here.
34122         <case PLUS, case MINUS, case IOR, case XOR, case AND>:
34123         Check "XEXP (src, 1)" operand here.
34124         (dimode_scalar_chain::make_vector_copies): Detect count register
34125         of a shift instruction.  Zero extend count register from QImode
34126         to DImode to satisfy vector shift pattern count operand predicate.
34127         Substitute vector shift count operand with a DImode copy.
34128         (dimode_scalar_chain::convert_reg): Ditto, zero-extend from
34129         vector register.
34131 2017-04-21  Uros Bizjak  <ubizjak@gmail.com>
34133         * config/i386/i386.md (*extzvqi_mem_rex64): Move above *extzv<mode>.
34134         Remove UNSPEC_NOREX_MEM tag.  Update corresponding peephole2 pattern.
34135         (*insvqi_1_mem_rex64): Move above insv<mode>_1.  Remove
34136         UNSPEC_NOREX_MEM tag.  Update corresponding peephole2 pattern.
34137         (UNSPEC_NOREX_MEM): Remove definition.
34139 2017-04-21  Richard Biener  <rguenther@suse.de>
34141         PR tree-optimization/79547
34142         * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
34143         Handle strlen, strcmp, strncmp, strcasecmp, strncasecmp, memcmp,
34144         bcmp, strspn, strcspn, __builtin_object_size and __builtin_constant_p
34145         without any constraints.
34147 2017-04-21  Richard Biener  <rguenther@suse.de>
34149         PR tree-optimization/78847
34150         * fold-const.c (split_tree): Handle POINTER_PLUS_EXPR.
34152 2017-04-21  Richard Biener  <rguenther@suse.de>
34154         * tree.h (build_qualified_type): Annotate with CXX_MEM_STAT_INFO.
34155         (build_distinct_type_copy): Likewise.
34156         (build_variant_type_copy): Likewise.
34157         * tree.c (build_qualified_type): Pass down mem-stat info.
34158         (build_distinct_type_copy): Likewise.
34159         (build_variant_type_copy): Likewise.
34161 2017-04-21  Richard Biener  <rguenther@suse.de>
34163         PR tree-optimization/80237
34164         * tree-ssa-pre.c (find_leader_in_sets): Add third set argument,
34165         defaulted to NULL.
34166         (phi_translate_1): Also allow a leader in AVAIL_OUT of pred
34167         for a simplified result.
34169 2016-04-21  Richard Biener  <rguenther@suse.de>
34171         * tree-ssa-loop-ivcanon.c (constant_after_peeling): Do not require
34172         sth as strict as a simple_iv but a chrec without symbols and an
34173         operand defined in the loop we are peeling (and not some subloop).
34174         (propagate_constants_for_unrolling): Propagate all constants.
34176 2017-04-20  Uros Bizjak  <ubizjak@gmail.com>
34178         PR target/79804
34179         * config/i386/i386.c (print_reg): Remove assert for disalowed
34180         regno values, call output_operand_lossage instead.
34182 2017-04-20  Uros Bizjak  <ubizjak@gmail.com>
34184         PR target/78090
34185         * config/i386/constraints.md (Yc): New register constraint.
34186         * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
34187         Use Yc constraint for alternative 2 of operand 0.  Remove
34188         preferred_for_speed attribute.
34190 2017-04-20  Alexander Monakov  <amonakov@ispras.ru>
34192         * omp-low.c (lower_lastprivate_clauses): Correct handling of linear and
34193         lastprivate clauses in SIMT case.
34195 2017-04-20  Volker Reichelt  <v.reichelt@netcologne.de>
34197         * doc/invoke.texi (-Wextra-semi): Document new warning option.
34199 2017-04-20  Richard Biener  <rguenther@suse.de>
34201         PR tree-optimization/57796
34202         * tree-vect-stmts.c (vect_model_store_cost): Cost scatters
34203         as N scalar stores.
34204         (vect_model_load_cost): Cost gathers as N scalar loads.
34206 2017-04-20  Richard Biener  <rguenther@suse.de>
34208         * ggc-page.c (ggc_allocated_p): Rename to ...
34209         (safe_lookup_page_table_entry): ... this and return the lookup
34210         result.
34211         (gt_ggc_m_S): Use safe_lookup_page_table_entry.
34213 2017-04-20  Richard Biener  <rguenther@suse.de>
34215         PR tree-optimization/80453
34216         * tree-ssa-sccvn.h (struct vn_phi_s): Add cclhs and ccrhs members.
34217         * tree-ssa-sccvn.c (cond_stmts_equal_p): Use recorded lhs and rhs
34218         from the conditions.
34219         (vn_phi_eq): Pass them down.
34220         (vn_phi_lookup): Record them.
34221         (vn_phi_insert): Likewise.
34223 2017-04-20  Matthew Fortune  <matthew.fortune@imgtec.com>
34225         * config/mips/mips.c (mips_expand_vec_perm_const): Re-fix
34226         uninitialized variable warning to avoid buffer overrun.
34228 2017-04-20  Alexander Monakov  <amonakov@ispras.ru>
34230         PR other/71250
34231         * doc/invoke.texi (-Wmissing-field-initializers): Mention that warning
34232         is suppressed for '{ 0 }' in C.
34234 2017-04-20  Jakub Jelinek  <jakub@redhat.com>
34236         * BASE-VER: Set to 8.0.0.
34238 2017-04-20  Thomas Preud'homme  <thomas.preudhomme@arm.com>
34240         * config/arm/arm.c (arm_elf_asm_cdtor): Create non-default
34241         priority .init_array and .fini_array section with SECTION_NOTYPE
34242         flag.
34244 2017-04-20  Jakub Jelinek  <jakub@redhat.com>
34246         PR middle-end/80423
34247         * tree.h (build_array_type): Add typeless_storage default argument.
34248         * tree.c (type_cache_hasher::equal): Also compare
34249         TYPE_TYPELESS_STORAGE flag for ARRAY_TYPEs.
34250         (build_array_type): Add typeless_storage argument, set
34251         TYPE_TYPELESS_STORAGE to it, if shared also hash it, and pass to
34252         recursive call.
34253         (build_nonshared_array_type): Adjust build_array_type_1 caller.
34254         (build_array_type): Likewise.  Add typeless_storage argument.
34256 2017-04-19  Eric Botcazou  <ebotcazou@adacore.com>
34257             Jakub Jelinek  <jakub@redhat.com>
34259         PR tree-optimization/80426
34260         * tree-vrp.c (extract_range_from_binary_expr_1): For an additive
34261         operation on symbolic operands, also compute the overflow for the
34262         invariant part when the operation degenerates into a negation.
34264 2017-04-19  Jakub Jelinek  <jakub@redhat.com>
34266         PR debug/80461
34267         * dwarf2out.c (modified_type_die, gen_type_die_with_usage):
34268         Check for t with zero TYPE_QUALS_NO_ADDR_SPACE.
34270         PR debug/80436
34271         * tree-ssa-loop-manip.c (find_uses_to_rename_def): Ignore debug uses.
34273 2017-04-19  Georg-Johann Lay  <avr@gjlay.de>
34275         PR target/80462
34276         * config/avr/avr.c (tree.h): Include it.
34277         (cgraph.h): Include it.
34278         (avr_encode_section_info): Don't warn for uninitialized progmem
34279         variable if it's just an alias.
34281 2017-04-19  Richard Biener  <rguenther@suse.de>
34283         PR ipa/65972
34284         * auto-profile.c (afdo_vpt_for_early_inline): Update SSA
34285         when needed by AutoPGO.
34287 2017-04-19  Paulo J. Matos  <paulo@matos-sorge.com>
34289         PR lto/50345
34290         * doc/lto.texi: Remove an extra 'that'.
34292 2017-04-19  Segher Boessenkool  <segher@kernel.crashing.org>
34294         PR rtl-optimization/80429
34295         * ira.c (split_live_ranges_for_shrink_wrap): Don't split regs that
34296         are only used in debug insns.
34298 2017-04-19  Eric Botcazou  <ebotcazou@adacore.com>
34299             Vladimir Makarov  <vmakarov@redhat.com>
34301         * config/sparc/predicates.md (input_operand): Add comment.  Return
34302         true for any memory operand when LRA is in progress.
34303         * config/sparc/sparc.c (sparc_expand_move): Minor formatting fix.
34305 2017-04-18  Jeff Law  <law@redhat.com>
34307         PR target/74563
34308         * mips.md ({return,simple_return}_internal): Do not overwrite
34309         operands[0].
34311 2017-04-18  Jakub Jelinek  <jakub@redhat.com>
34313         PR tree-optimization/80443
34314         * tree-vrp.c (intersect_ranges): For signed 1-bit precision type,
34315         instead of adding 1, subtract -1 and similarly instead of subtracting
34316         1 add -1.
34318 2017-04-18  Richard Sandiford  <richard.sandiford@arm.com>
34320         PR rtl-optimization/80357
34321         * haifa-sched.c (tmp_bitmap): New variable.
34322         (model_recompute): Handle duplicate use records.
34323         (alloc_global_sched_pressure_data): Initialize tmp_bitmap.
34324         (free_global_sched_pressure_data): Free it.
34326 2017-04-18  Bernd Edlinger  <bernd.edlinger@hotmail.de>
34328         Revert:
34329         2017-02-20  Bernd Edlinger  <bernd.edlinger@hotmail.de>
34330         * Makefile.in (BUILD_SYSTEM_HEADER_DIR): New make variabe.
34331         (LIMITS_H_TEST, if_multiarch, stmp-fixinc): Use BUILD_SYSTEM_HEADER_DIR
34332         instead of SYSTEM_HEADER_DIR.
34334 2017-04-18  Jeff Law  <law@redhat.com>
34336         PR middle-end/80422
34337         * cfgcleanup.c (try_crossjump_to_edge): Verify SRC1 and SRC2 have
34338         predecessors after walking up the insn chain.
34340 2017-04-18  Jakub Jelinek  <jakub@redhat.com>
34342         PR debug/80263
34343         * dwarf2out.c (modified_type_die): Try harder not to emit internal
34344         sizetype type into debug info.
34346 2017-04-18  Michael Meissner  <meissner@linux.vnet.ibm.com>
34348         PR target/80099
34349         * config/rs6000/rs6000.c (rs6000_expand_vector_extract): Eliminate
34350         unneeded test for TARGET_UPPER_REGS_SF.
34351         * config/rs6000/vsx.md (vsx_extract_v4sf_var): Likewise.
34353 2017-04-18  Jakub Jelinek  <jakub@redhat.com>
34355         PR sanitizer/80444
34356         * sancov.c (sancov_pass): Use gsi_start_nondebug_after_labels_bb
34357         instead of gsi_after_labels.
34359 2017-04-18  Jeff Law  <law@redhat.com>
34361         * regcprop.c (maybe_mode_change): Avoid creating copies of the
34362         stack pointer.
34364         Revert:
34365         2017-04-13  Jeff Law  <law@redhat.com>
34366         * config/mips.mips.md (zero_extendsidi2): Do not allow SP to appear
34367         in operands[1] if it is a MEM and TARGET_MIPS16 is active.
34369 2017-04-18  Georg-Johann Lay  <avr@gjlay.de>
34371         PR target/79453
34372         * config/avr/avr.c (intl.h): Include it.
34373         (avr_pgm_check_var_decl) [reason]: Wrap diagnostic snippets into _().
34375 2017-04-18  Martin Liska  <mliska@suse.cz>
34377         PR gcov-profile/78783
34378         * gcov-tool.c (gcov_output_files): Validate that destination
34379         file is either removed by the tool or by a user.
34381 2017-04-14  Andrew Burgess  <andrew.burgess@embecosm.com>
34382             Guy Benyei  <guybe@mellanox.com>
34384         * config/arc/arc.c (arc_reorg): Move loop_end_id into a more local
34385         block, and do not negate it, the stored id is already negative.
34387 2017-04-14  Andrew Burgess  <andrew.burgess@embecosm.com>
34389         * config/arc/arc.md (doloop_begin_i): Use @pcl assembler syntax.
34391 2017-04-14  Michael Meissner  <meissner@linux.vnet.ibm.com>
34393         PR target/80098
34394         * config/rs6000/rs6000-cpus.def (OTHER_P9_VECTOR_MASKS): Define
34395         masks of options that should be turned off if the VSX vector
34396         options are turned off.
34397         (OTHER_P8_VECTOR_MASKS): Likewise.
34398         (OTHER_VSX_VECTOR_MASKS): Likewise.
34399         * config/rs6000/rs6000.c (rs6000_option_override_internal): Call
34400         rs6000_disable_incompatible_switches to validate no type switches
34401         like -mvsx.
34402         (rs6000_incompatible_switch): New function to disallow turning on
34403         other vector options if -mno-vsx, -mno-power8-vector, or
34404         -mno-power9-vector are specified.
34406 2017-04-14  Claudiu Zissulescu  <claziss@synopsys.com>
34408         * config/arc/arc.h (CRT_CALL_STATIC_FUNCTION): Use long calls.
34410 2017-04-14  Claudiu Zissulescu  <claziss@synopsys.com>
34412         * config/arc/arc-protos.h (arc_decl_pretend_args): Remove.
34413         * config/arc/arc.c (arc_decl_pretend_args): Likewise.
34414         * config/arc/arc.h (CFA_FRAME_BASE_OFFSET): Likewise.
34415         (ARG_POINTER_CFA_OFFSET): Likewise.
34417 2017-04-14  Claudiu Zissulescu  <claziss@synopsys.com>
34419         * config/arc/arc.c (arc_mode_dependent_address_p): Relax
34420         conditions to take advantage of various optimizations.
34422 2017-04-13  Jeff Law  <law@redhat.com>
34424         * config/mips.mips.md (zero_extendsidi2): Do not allow SP to appear
34425         in operands[1] if it is a MEM and TARGET_MIPS16 is active.
34426         (zero_extendsidi2_dext): Likewise.
34428 2017-04-13  Jakub Jelinek  <jakub@redhat.com>
34430         PR sanitizer/80403
34431         * fold-const.c (fold_ternary_loc): Revert
34432         use op0 instead of fold_convert_loc (loc, type, arg0) part of
34433         2017-04-12 change.
34435 2017-04-13  Vladimir Makarov  <vmakarov@redhat.com>
34437         PR rtl-optimization/80343
34438         * lra-remat.c (update_scratch_ops): Assign original hard reg to
34439         new scratch pseudo.
34441 2017-04-13  Denis Khalikov <d.khalikov@partner.samsung.com>
34443         PR sanitizer/80414
34444         * ubsan.c (ubsan_expand_bounds_ifn): Pass original index
34445         to ubsan_encode_value.
34447 2017-04-13  Jeff Law  <law@redhat.com>
34449         * reload1.c (eliminate_regs_1): Call gen_rtx_raw_SUBREG for SUBREGs
34450         appearing in DEBUG_INSNs.
34452 2017-04-13  Martin Liska  <mliska@suse.cz>
34454         PR gcov-profile/80413
34455         * gcov-io.c (gcov_write_string): Copy to buffer just when
34456         allocated size is greater than zero.
34458 2017-04-13  Jakub Jelinek  <jakub@redhat.com>
34460         PR debug/80321
34461         * dwarf2out.c (decls_for_scope): Ignore declarations of
34462         current_function_decl in BLOCK_NONLOCALIZED_VARS.
34464 2017-04-12  Jan Hubicka  <hubicka@ucw.cz>
34466         PR lto/69953
34467         * ipa-visibility.c (non_local_p): Fix typos.
34468         (localize_node): When localizing symbol in same comdat group,
34469         dissolve the group only when we know external symbols are going
34470         to be privatized.
34471         (function_and_variable_visibility): Do not localize DECL_EXTERNAL.
34473 2017-04-12  Jakub Jelinek  <jakub@redhat.com>
34475         PR tree-optimization/79390
34476         * optabs.c (emit_conditional_move): If the preferred op2/op3 operand
34477         order does not result in usable sequence, retry with reversed operand
34478         order.
34480         PR sanitizer/80403
34481         PR sanitizer/80404
34482         PR sanitizer/80405
34483         * fold-const.c (fold_ternary_loc): Use op1 instead of arg1 as argument
34484         to fold_build2_loc.  Convert TREE_OPERAND (tem, 0) to type.  Use
34485         op0 instead of fold_convert_loc (loc, type, arg0).
34487 2017-04-12  Jeff Law  <law@redhat.com>
34489         * genattrtab.c (write_eligible_delay): Verify DELAY_INSN still
34490         has a delay slot in the generated code.
34492         * config/cris/cris.md (cris_preferred_reload_class): Return
34493         GENNONACR_REGS rather than GENERAL_REGS.
34495 2017-04-12  Jakub Jelinek  <jakub@redhat.com>
34497         PR c/80163
34498         * expr.c <CASE_CONVERT>: For EXPAND_INITIALIZER determine SIGN_EXTEND
34499         vs. ZERO_EXTEND based on signedness of treeop0's type rather than
34500         signedness of the result type.
34502 2017-04-12  Richard Biener  <rguenther@suse.de>
34503             Jeff Law  <law@redhat.com>
34505         PR tree-optimization/80359
34506         * tree-ssa-dse.c (maybe_trim_partially_dead_store): Do not
34507         trim stores to TARGET_MEM_REFs.
34509 2017-04-12  Richard Biener  <rguenther@suse.de>
34511         PR tree-optimization/79390
34512         * gimple-ssa-split-paths.c (is_feasible_trace): Restrict
34513         threading case even more.
34515 2017-04-12  Segher Boessenkool  <segher@kernel.crashing.org>
34517         PR target/80382
34518         * config/rs6000/sync.md (atomic_load<mode>, atomic_store<mode): Test
34519         for quad_address_p for TImode, instead of just not indexed_address.
34521 2017-04-12  Richard Biener  <rguenther@suse.de>
34522             Bernd Edlinger  <bernd.edlinger@hotmail.de>
34524         PR middle-end/79671
34525         * alias.c (component_uses_parent_alias_set_from): Handle
34526         TYPE_TYPELESS_STORAGE.
34527         (get_alias_set): Likewise.
34528         * tree-core.h (tree_type_common): Add typeless_storage flag.
34529         * tree.h (TYPE_TYPELESS_STORAGE): New macro.
34530         * stor-layout.c (place_union_field): Set TYPE_TYPELESS_STORAGE
34531         for types containing members with TYPE_TYPELESS_STORAGE.
34532         (place_field): Likewise.
34533         (layout_type): Likewise for ARRAY_TYPE.
34534         * lto-streamer-out.c (hash_tree): Hash TYPE_TYPELESS_STORAGE.
34535         * tree-streamer-in.c (unpack_ts_type_common_value_fields): Stream
34536         TYPE_TYPELESS_STORAGE.
34537         * tree-streamer-out.c (pack_ts_type_common_value_fields): Likewise.
34539 2017-04-12  Jakub Jelinek  <jakub@redhat.com>
34541         PR sanitizer/80349
34542         * fold-const.c (fold_binary_loc) <case BIT_IOR_EXPR>: Convert arg0's
34543         first argument to type.
34545 2017-04-11  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
34547         PR target/80376
34548         PR target/80315
34549         * config/rs6000/rs6000.c (rs6000_expand_unop_builtin): Return
34550         CONST0_RTX (mode) rather than const0_rtx where appropriate.
34551         (rs6000_expand_binop_builtin): Likewise.
34552         (rs6000_expand_ternop_builtin): Likewise; also add missing
34553         vsx_xxpermdi_* variants; also fix typo (arg1 => arg2) for
34554         vshasigma built-ins.
34555         * doc/extend.texi: Document that vec_xxpermdi's third argument
34556         must be a constant.
34558 2017-04-11  Uros Bizjak  <ubizjak@gmail.com>
34560         * config/i386/i386.c (dimode_scalar_chain::compute_convert_gain):
34561         Use shift_const cost parameter when calculating gain of STV shifts.
34563 2017-04-11  Vladimir Makarov  <vmakarov@redhat.com>
34565         PR rtl-optimization/70478
34566         * lra-constraints.c (process_alt_operands): Check memory for
34567         disfavoring memory insn operand.
34569 2017-04-11  Jakub Jelinek  <jakub@redhat.com>
34571         PR middle-end/80100
34572         * simplify-rtx.c (simplify_binary_operation_1) <case IOR>: Perform
34573         left shift in unsigned HOST_WIDE_INT type.
34575         PR rtl-optimization/80385
34576         * simplify-rtx.c (simplify_unary_operation_1): Don't transform
34577         (not (neg X)) into (plus X -1) for complex or non-integral modes.
34579         PR libgomp/80394
34580         * omp-low.c (scan_omp_task): Don't optimize away empty tasks
34581         if they have any depend clauses.
34583 2017-04-11  Martin Liska  <mliska@suse.cz>
34585         PR ipa/80212
34586         * cgraph.c (cgraph_node::dump): Dump calls_comdat_local.
34587         * ipa-split.c (split_function): Create a local comdat symbol
34588         if caller is in a comdat group.
34590 2017-04-11  Martin Liska  <mliska@suse.cz>
34592         PR ipa/80212
34593         * ipa-cp.c (determine_versionability): Handle calls_comdat_local
34594         flags.
34596 2017-04-11  Martin Sebor  <msebor@redhat.com>
34598         PR middle-end/80364
34599         * gimple-ssa-sprintf.c (get_int_range): Remove second argument and
34600         always use the int type.  Use INTEGRAL_TYPE_P() rather than testing
34601         for INTEGER_TYPE.
34602         (directive::set_width, directive::set_precision, format_character):
34603         Adjust.
34604         (parse_directive): Use INTEGRAL_TYPE_P() rather than testing for
34605         INTEGER_TYPE.
34607 2017-04-11  Richard Earnshaw  <rearnsha@arm.com>
34609         PR target/80389
34610         * config/arm/arm.c (arm_configure_build_target): When -mcpu and -arch
34611         conflict, set target->arch_name instead of target->cpu_name.
34613 2017-04-11  Richard Biener  <rguenther@suse.de>
34615         PR tree-optimization/80374
34616         * tree-ssa-dom.c (derive_equivalences_from_bit_ior): Use
34617         build_zero_cst, remove fold_convertible_p check again.
34619 2017-04-11  Martin Liska  <mliska@suse.cz>
34621         PR sanitizer/70878
34622         * ubsan.c (instrument_object_size): Do not instrument register
34623         variables.
34625 2017-04-11  Jakub Jelinek  <jakub@redhat.com>
34627         PR target/80381
34628         * config/i386/i386-builtin-types.def
34629         (V16HI_FTYPE_V16HI_INT_V16HI_UHI_COUNT,
34630         V16HI_FTYPE_V16HI_V8HI_V16HI_UHI_COUNT,
34631         V16SI_FTYPE_V16SI_INT_V16SI_UHI_COUNT,
34632         V16SI_FTYPE_V16SI_V4SI_V16SI_UHI_COUNT,
34633         V2DI_FTYPE_V2DI_INT_V2DI_UQI_COUNT,
34634         V2DI_FTYPE_V2DI_V2DI_V2DI_UQI_COUNT,
34635         V32HI_FTYPE_V32HI_INT_V32HI_USI_COUNT,
34636         V32HI_FTYPE_V32HI_V8HI_V32HI_USI_COUNT,
34637         V4DI_FTYPE_V4DI_INT_V4DI_UQI_COUNT,
34638         V4DI_FTYPE_V4DI_V2DI_V4DI_UQI_COUNT,
34639         V4SI_FTYPE_V4SI_INT_V4SI_UQI_COUNT,
34640         V4SI_FTYPE_V4SI_V4SI_V4SI_UQI_COUNT,
34641         V8DI_FTYPE_V8DI_INT_V8DI_UQI_COUNT,
34642         V8DI_FTYPE_V8DI_V2DI_V8DI_UQI_COUNT,
34643         V8HI_FTYPE_V8HI_INT_V8HI_UQI_COUNT,
34644         V8HI_FTYPE_V8HI_V8HI_V8HI_UQI_COUNT,
34645         V8SI_FTYPE_V8SI_INT_V8SI_UQI_COUNT,
34646         V8SI_FTYPE_V8SI_V4SI_V8SI_UQI_COUNT): New function type aliases.
34647         * config/i386/i386-builtin.def (__builtin_ia32_pslld512_mask,
34648         __builtin_ia32_pslldi512_mask, __builtin_ia32_psllq512_mask,
34649         __builtin_ia32_psllqi512_mask, __builtin_ia32_psrad512_mask,
34650         __builtin_ia32_psradi512_mask, __builtin_ia32_psraq512_mask,
34651         __builtin_ia32_psraqi512_mask, __builtin_ia32_psrld512_mask,
34652         __builtin_ia32_psrldi512_mask, __builtin_ia32_psrlq512_mask,
34653         __builtin_ia32_psrlqi512_mask, __builtin_ia32_psllwi128_mask,
34654         __builtin_ia32_pslldi128_mask, __builtin_ia32_psllqi128_mask,
34655         __builtin_ia32_psllw128_mask, __builtin_ia32_pslld128_mask,
34656         __builtin_ia32_psllq128_mask, __builtin_ia32_psllwi256_mask,
34657         __builtin_ia32_psllw256_mask, __builtin_ia32_pslldi256_mask,
34658         __builtin_ia32_pslld256_mask, __builtin_ia32_psllqi256_mask,
34659         __builtin_ia32_psllq256_mask, __builtin_ia32_psradi128_mask,
34660         __builtin_ia32_psrad128_mask, __builtin_ia32_psradi256_mask,
34661         __builtin_ia32_psrad256_mask, __builtin_ia32_psraqi128_mask,
34662         __builtin_ia32_psraq128_mask, __builtin_ia32_psraqi256_mask,
34663         __builtin_ia32_psraq256_mask, __builtin_ia32_psrldi128_mask,
34664         __builtin_ia32_psrld128_mask, __builtin_ia32_psrldi256_mask,
34665         __builtin_ia32_psrld256_mask, __builtin_ia32_psrlqi128_mask,
34666         __builtin_ia32_psrlq128_mask, __builtin_ia32_psrlqi256_mask,
34667         __builtin_ia32_psrlq256_mask, __builtin_ia32_psrawi256_mask,
34668         __builtin_ia32_psraw256_mask, __builtin_ia32_psrawi128_mask,
34669         __builtin_ia32_psraw128_mask, __builtin_ia32_psrlwi256_mask,
34670         __builtin_ia32_psrlw256_mask, __builtin_ia32_psrlwi128_mask,
34671         __builtin_ia32_psrlw128_mask, __builtin_ia32_psllwi512_mask,
34672         __builtin_ia32_psllw512_mask, __builtin_ia32_psrawi512_mask,
34673         __builtin_ia32_psraw512_mask, __builtin_ia32_psrlwi512_mask,
34674         __builtin_ia32_psrlw512_mask): Use _COUNT suffixed function type
34675         aliases.
34676         * config/i386/i386.c (ix86_expand_args_builtin): Rename last_arg_count
34677         flag to second_arg_count, handle 4 argument function type _COUNT
34678         aliases, handle second_arg_count on second argument rather than last.
34680 2017-04-10  Jeff Law  <law@redhat.com>
34682         PR tree-optimization/80374
34683         * tree-ssa-dom.c (derive_equivalences_from_bit_ior): Do not try to
34684         record anything if we can not convert integer_zero_node to the
34685         desired type.
34687 2017-04-10  Kelvin Nilsen  <kelvin@gcc.gnu.org>
34689         PR target/80108
34690         * config/rs6000/rs6000.c (rs6000_option_override_internal):
34691         Enhance special handling given to the TARGET_P9_MINMAX option in
34692         relation to certain other options.
34694 2017-04-10  Bin Cheng  <bin.cheng@arm.com>
34696         PR tree-optimization/80153
34697         * tree-ssa-loop-ivopts.c (add_iv_candidate_for_use): Check and
34698         remove POINTER_PLUS_EXPR's base part directly, rather than through
34699         aff_tree.
34701 2017-04-10  Richard Biener  <rguenther@suse.de>
34702             Bin Cheng  <bin.cheng@arm.com>
34704         PR tree-optimization/80153
34705         * tree-affine.c (aff_combination_to_tree): Get base pointer from
34706         the first element of pointer type aff_tree.  Build result expr in
34707         aff_tree's type.
34708         (add_elt_to_tree): Convert to type unconditionally.  Remove other
34709         fold_convert calls.
34710         * tree-ssa-loop-ivopts.c (alloc_iv): Pass in consistent types.
34711         (rewrite_use_nonlinear_expr): Check invariant using iv information.
34713 2017-04-10  Richard Biener  <rguenther@suse.de>
34715         * tree-ssa-structalias.c (find_func_aliases): Properly handle
34716         asm inputs.
34718 2017-04-10  Vladimir Makarov  <vmakarov@redhat.com>
34720         PR rtl-optimization/70478
34721         * lra-constraints.c (curr_small_class_check): New.
34722         (update_and_check_small_class_inputs): New.
34723         (process_alt_operands): Update curr_small_class_check.  Disfavor
34724         alternative insn memory operands.  Check available regs for small
34725         class operands.
34727 2017-03-31  Matthew Fortune  <matthew.fortune@imgtec.com>
34729         PR target/80057
34730         * config/mips/mips.opt (-mvirt): Update description.
34731         * doc/invoke.texi (-mvirt): Likewise.
34733 2017-04-10  Richard Biener  <rguenther@suse.de>
34735         PR middle-end/80362
34736         * fold-const.c (fold_binary_loc): Look at unstripped ops when
34737         looking for NEGATE_EXPR in -A / -B to A / B folding.
34739 2017-04-10  Martin Liska  <mliska@suse.cz>
34741         PR gcov-profile/80224
34742         * gcov.c (print_usage): Fix usage string.
34743         (get_gcov_intermediate_filename): Remove.
34744         (output_gcov_file): Use both for normal and intermediate format.
34745         (generate_results): Do not initialize special file for
34746         intermediate format.
34748 2017-04-10  Richard Biener  <rguenther@suse.de>
34750         PR tree-optimization/80304
34751         * tree-ssa-loop-im.c (ref_indep_loop_p_1): Also recurse
34752         for safelen.
34754 2017-04-10  Nathan Sidwell  <nathan@acm.org>
34756         PR target/79905
34757         * config/rs6000/rs6000.c (rs6000_vector_type): New.
34758         (rs6000_init_builtins): Use it.
34760 2016-04-10  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
34762         * config/arm/arm.md (<mrc>): Add mode to SET source.
34763         (<mrrc>): Likewise.
34765 2017-04-10  Richard Biener  <rguenther@suse.de>
34767         PR middle-end/80344
34768         * gimplify.c (is_gimple_mem_rhs_or_call): Allow CLOBBERs.
34770 2017-04-10  Jakub Jelinek  <jakub@redhat.com>
34772         PR target/80324
34773         * config/i386/avx512fintrin.h (_mm512_reduce_add_epi32,
34774         _mm512_reduce_mul_epi32, _mm512_reduce_and_epi32,
34775         _mm512_reduce_or_epi32, _mm512_mask_reduce_add_epi32,
34776         _mm512_mask_reduce_mul_epi32, _mm512_mask_reduce_and_epi32,
34777         _mm512_mask_reduce_or_epi32, _mm512_reduce_min_epi32,
34778         _mm512_reduce_max_epi32, _mm512_reduce_min_epu32,
34779         _mm512_reduce_max_epu32, _mm512_mask_reduce_min_epi32,
34780         _mm512_mask_reduce_max_epi32, _mm512_mask_reduce_min_epu32,
34781         _mm512_mask_reduce_max_epu32, _mm512_reduce_add_ps,
34782         _mm512_reduce_mul_ps, _mm512_mask_reduce_add_ps,
34783         _mm512_mask_reduce_mul_ps, _mm512_reduce_min_ps, _mm512_reduce_max_ps,
34784         _mm512_mask_reduce_min_ps, _mm512_mask_reduce_max_ps,
34785         _mm512_reduce_add_epi64, _mm512_reduce_mul_epi64,
34786         _mm512_reduce_and_epi64, _mm512_reduce_or_epi64,
34787         _mm512_mask_reduce_add_epi64, _mm512_mask_reduce_mul_epi64,
34788         _mm512_mask_reduce_and_epi64, _mm512_mask_reduce_or_epi64,
34789         _mm512_reduce_min_epi64, _mm512_reduce_max_epi64,
34790         _mm512_mask_reduce_min_epi64, _mm512_mask_reduce_max_epi64,
34791         _mm512_reduce_min_epu64, _mm512_reduce_max_epu64,
34792         _mm512_mask_reduce_min_epu64, _mm512_mask_reduce_max_epu64,
34793         _mm512_reduce_add_pd, _mm512_reduce_mul_pd, _mm512_mask_reduce_add_pd,
34794         _mm512_mask_reduce_mul_pd, _mm512_reduce_min_pd, _mm512_reduce_max_pd,
34795         _mm512_mask_reduce_min_pd, _mm512_mask_reduce_max_pd): New intrinsics.
34797 2017-04-08  Vladimir Makarov  <vmakarov@redhat.com>
34799         PR rtl-optimization/70478
34800         * lra-constraints.c: Reverse the last patch.
34802 2017-04-08  Andreas Tobler  <andreast@gcc.gnu.org>
34804         * config/aarch64/aarch64-freebsd.h: Define MCOUNT_NAME.
34805         Add comment for WCHAR_T.
34807 2017-04-08  Martin Liska  <mliska@suse.cz>
34809         Revert:
34810         2017-04-07  Martin Liska  <mliska@suse.cz>
34812         PR ipa/80212
34813         * ipa-split.c (split_function): Add function part to a same comdat
34814         group.
34816 2017-04-08  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
34818         PR target/80358
34819         * config/rs6000/rs6000.c (expand_block_compare): Fix boundary check.
34821 2017-04-07  Pat Haugen  <pthaugen@us.ibm.com>
34823         * rs6000/rs6000.c (vec_load_pendulum): Rename...
34824         (vec_pairing): ...to this.
34825         (power9_sched_reorder2): Rewrite code for pairing vector/vecload insns.
34826         (rs6000_sched_init): Adjust for name change.
34827         (struct rs6000_sched_context): Likewise.
34828         (rs6000_init_sched_context): Likewise.
34829         (rs6000_set_sched_context): Likewise.
34831 2017-04-07  Jakub Jelinek  <jakub@redhat.com>
34833         PR target/80322
34834         PR target/80323
34835         PR target/80325
34836         PR target/80326
34837         * config/i386/avxintrin.h (_mm256_cvtsd_f64, _mm256_cvtss_f32): New
34838         intrinsics.
34839         * config/i386/avx512fintrin.h (_mm512_int2mask, _mm512_mask2int,
34840         _mm512_abs_ps, _mm512_mask_abs_ps, _mm512_abs_pd, _mm512_mask_abs_pd,
34841         _mm512_cvtsd_f64, _mm512_cvtss_f32): Likewise.
34843 2017-04-07  Andreas Tobler  <andreast@gcc.gnu.org>
34845         * config/aarch64/aarch64-freebsd.h: Define WCHAR_TYPE.
34847 2017-04-07  Vladimir Makarov  <vmakarov@redhat.com>
34849         PR rtl-optimization/70703
34850         * ira-color.c (update_conflict_hard_regno_costs): Use
34851         int64_t instead of HOST_WIDE_INT.
34853 2017-04-07  Vladimir Makarov  <vmakarov@redhat.com>
34855         PR rtl-optimization/70478
34856         * lra-constraints.c (process_alt_operands): Disfavor alternative
34857         insn memory operands.
34859 2017-04-07  Jeff Law  <law@redhat.com>
34861         * config/iq2000/iq2000.c (final_prescan_insn): Do not separate a
34862         CALL and NOTE_INSN_CALL_ARG_LOCATION.
34864 2017-04-07  Martin Liska  <mliska@suse.cz>
34866         PR target/79889
34867         * config/aarch64/aarch64.c (aarch64_process_target_attr):
34868         Show error message instead of an ICE.
34870 2017-04-07  Martin Liska  <mliska@suse.cz>
34872         PR ipa/80212
34873         * ipa-split.c (split_function): Add function part to a same comdat
34874         group.
34876 2017-04-07  Richard Biener  <rguenther@suse.de>
34878         PR middle-end/80341
34879         * tree.c (get_unwidened): Also handle ! for_type case for
34880         INTEGER_CSTs.
34881         * convert.c (do_narrow): Split out from ...
34882         (convert_to_integer_1): ... here.  Do not pass final truncation
34883         type to get_unwidened for TRUNC_DIV_EXPR.
34885 2017-04-07  Richard Biener  <rguenther@suse.de>
34887         * tree-affine.c (wide_int_ext_for_comb): Take type rather
34888         than aff_tree.
34889         (aff_combination_const): Adjust.
34890         (aff_combination_scale): Likewise.
34891         (aff_combination_add_elt): Likewise.
34892         (aff_combination_add_cst): Likewise.
34893         (aff_combination_convert): Likewise.
34894         (add_elt_to_tree): Likewise.  Remove unused argument.
34895         (aff_combination_to_tree): Adjust calls to add_elt_to_tree.
34897 2017-04-07  Sebastian Huber  <sebastian.huber@embedded-brains.de>
34899         * config/arm/arm.h (ARM_DEFAULT_SHORT_ENUMS): Provide default
34900         definition.
34901         * config/arm/arm.c (arm_default_short_enums): Use
34902         ARM_DEFAULT_SHORT_ENUMS.
34903         * config/arm/rtems.h (ARM_DEFAULT_SHORT_ENUMS): Define.
34905 2017-04-06  Jakub Jelinek  <jakub@redhat.com>
34907         PR debug/80234
34908         * dwarf2out.c (gen_member_die): Handle C++17 inline static data
34909         members with redundant out-of-class redeclaration.
34911 2017-04-06  Uros Bizjak  <ubizjak@gmail.com>
34913         PR target/80286
34914         * config/i386/sse.md (*vec_extractv4si_0_zext_sse4): New pattern.
34915         * config/i386/i386.md (*zero_extendsidi2):
34916         Add (?*x,*x) and (?*v,*v) alternatives.
34918 2017-04-06  Uros Bizjak  <ubizjak@gmail.com>
34920         PR target/79733
34921         * config/i386/i386.c (ix86_expand_builtin)
34922         <case IX86_BUILTIN_K{,OR}TEST{C,Z}{8,16,32,64}>: Determine insn operand
34923         mode from insn data. Convert operands to insn operand mode.
34924         Copy operands that don't satisfy insn predicate to a register.
34926 2017-04-06  Sam Thursfield  <sam.thursfield@codethink.co.uk>
34928         * config/rs6000/x-aix: Increase memory limit for genautomata on AIX.
34929         Update comments.
34931 2017-04-06  Richard Biener  <rguenther@suse.de>
34933         PR tree-optimization/80334
34934         * tree-ssa-loop-ivopts.c (rewrite_use_address): Properly
34935         preserve alignment of accesses.
34937 2017-04-06  Richard Biener  <rguenther@suse.de>
34939         PR tree-optimization/80262
34940         * tree-sra.c (build_ref_for_offset): Preserve address-space
34941         information.
34942         * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address):
34943         Drop useless address-space information on MEM_REF offsets.
34945 2017-04-05  Andreas Schwab  <schwab@linux-m68k.org>
34947         * builtins.def (BUILT_IN_UPDATE_SETJMP_BUF): Fix type.
34949 2017-04-05  Vladimir Makarov  <vmakarov@redhat.com>
34951         PR rtl-optimization/70703
34952         * ira-color.c (update_conflict_hard_regno_costs): Use
34953         HOST_WIDE_INT instead of long.
34955 2017-04-05  Uros Bizjak  <ubizjak@gmail.com>
34957         PR target/80298
34958         * config/i386/mmintrin.h: Add -msse target option when __SSE__ is
34959         not defined for x86_64 target.  Add -mmmx target option when __SSE2__
34960         is not defined.
34961         * config/i386/mm3dnow.h: Add -msse target when __SSE__ is not defined
34962         for x86_64 target.  Handle -m3dnowa option.
34964 2017-04-05  Vladimir Makarov  <vmakarov@redhat.com>
34966         PR rtl-optimization/70703
34967         * ira-color.c (update_costs_from_allocno): Use the smallest mode.
34968         (update_conflict_hard_regno_costs): Use long instead of unsigned
34969         arithmetic for cost calculation.
34971 2017-04-05  Jakub Jelinek  <jakub@redhat.com>
34972             Bernd Edlinger  <bernd.edlinger@hotmail.de>
34974         PR sanitizer/80308
34975         * asan.c (asan_store_shadow_bytes): Fix location of last_chunk_value
34976         for big endian.
34978 2017-04-05  Eric Botcazou  <ebotcazou@adacore.com>
34980         PR target/78002
34981         * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Replace
34982         ptr_mode with Pmode throughout.
34983         * config/aarch64/aarch64.md (probe_stack_range_<PTR:mode): Rename
34984         into probe_stack_range and use DImode.
34986 2017-04-05  Dominik Vogt  <vogt@linux.vnet.ibm.com>
34988         PR target/79890
34989         * config/s390/s390.c (s390_register_info_gprtofpr): Return if
34990         call_eh_return is true.
34992 2017-04-05  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
34994         * config/s390/s390-c.c (s390_resolve_overloaded_builtin):
34995         Initialize last_match_fntype_index.
34997 2017-04-05  Jakub Jelinek  <jakub@redhat.com>
34999         PR target/80310
35000         * tree-nvr.c: Include internal-fn.h.
35001         (pass_return_slot::execute): Ignore internal calls without
35002         direct optab.
35004 2017-04-04  Jakub Jelinek  <jakub@redhat.com>
35005             Richard Biener  <rguenther@suse.de>
35007         PR c++/80297
35008         * genmatch.c (capture::gen_transform): For GENERIC unshare_expr
35009         captures used multiple times, except for the last use.
35010         * generic-match-head.c: Include gimplify.h.
35012 2017-04-04  Jakub Jelinek  <jakub@redhat.com>
35014         PR tree-optimization/79390
35015         * target.h (struct noce_if_info): Declare.
35016         * targhooks.h (default_noce_conversion_profitable_p): Declare.
35017         * target.def (noce_conversion_profitable_p): New target hook.
35018         * ifcvt.h (struct noce_if_info): New type, moved from ...
35019         * ifcvt.c (struct noce_if_info): ... here.
35020         (noce_conversion_profitable_p): Renamed to ...
35021         (default_noce_conversion_profitable_p): ... this.  No longer
35022         static nor inline.
35023         (noce_try_store_flag_constants, noce_try_addcc,
35024         noce_try_store_flag_mask, noce_try_cmove, noce_try_cmove_arith,
35025         noce_convert_multiple_sets): Use targetm.noce_conversion_profitable_p
35026         instead of noce_conversion_profitable_p.
35027         * config/i386/i386.c: Include ifcvt.h.
35028         (ix86_option_override_internal): Don't override
35029         PARAM_MAX_RTL_IF_CONVERSION_INSNS default.
35030         (ix86_noce_conversion_profitable_p): New function.
35031         (TARGET_NOCE_CONVERSION_PROFITABLE_P): Redefine.
35032         * config/i386/x86-tune.def (X86_TUNE_ONE_IF_CONV_INSN): Adjust comment.
35033         * doc/tm.texi.in (TARGET_NOCE_CONVERSION_PROFITABLE_P): Add.
35034         * doc/tm.texi: Regenerated.
35036 2017-04-04  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
35038         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Grammar
35039         correction.
35041 2017-04-04  Thomas Preud'homme  <thomas.preudhomme@arm.com>
35043         PR target/80307
35044         * config/arm/arm.c (thumb1_rtx_costs): Give a cost of 32
35045         instructions for small multiply cores.
35047 2017-04-04  Jeff Law  <law@redhat.com>
35049         * config/mips/mips.c (mips_multi_add): Zero initialize the newly
35050         added member.
35051         (mips_expand_vec_perm_const): Initialize elements in orig_perm
35052         that are not set by the loop over the elements.
35054 2017-04-04  Jakub Jelinek  <jakub@redhat.com>
35056         PR target/80286
35057         * config/i386/i386.c (ix86_expand_args_builtin): If op has scalar
35058         int mode, convert_modes it to mode as unsigned, otherwise use
35059         lowpart_subreg to mode rather than SImode.
35060         * config/i386/sse.md (<mask_codefor>ashr<mode>3<mask_name>,
35061         ashr<mode>3, ashr<mode>3<mask_name>, <shift_insn><mode>3<mask_name>):
35062         Use DImode instead of SImode for the shift count operand.
35063         * config/i386/mmx.md (mmx_ashr<mode>3, mmx_<shift_insn><mode>3):
35064         Likewise.
35066 2017-04-04  Richard Biener  <rguenther@suse.de>
35068         PR middle-end/80281
35069         * match.pd (A + (-B) -> A - B): Make sure to preserve unsigned
35070         arithmetic done for the negate or the plus.  Simplify.
35071         (A - (-B) -> A + B): Likewise.
35072         * fold-const.c (split_tree): Make sure to not negate pointers.
35074 2017-04-04  Segher Boessenkool  <segher@kernel.crashing.org>
35076         PR rtl-optimization/60818
35077         * simplify-rtx.c (simplify_binary_operation_1): Do not replace
35078         a compare of comparisons with the thing compared if this results
35079         in a different machine mode.
35081 2017-04-03  Jonathan Wakely  <jwakely@redhat.com>
35083         * alias.c (base_alias_check): Fix typo in comment.
35084         * cgraph.h (class ipa_polymorphic_call_context): Likewise.
35085         * cgraphunit.c (symbol_table::compile): Likewise.
35086         * collect2.c (maybe_run_lto_and_relink): Likewise.
35087         * config/arm/arm.c (arm_thumb1_mi_thunk): Likewise.
35088         * config/avr/avr-arch.h (avr_arch_info_t): Likewise.
35089         * config/avr/avr.c (avr_map_op_t): Likewise.
35090         * config/cr16/cr16.h (DATA_ALIGNMENT): Likewise.
35091         * config/epiphany/epiphany.c (TARGET_ARG_PARTIAL_BYTES): Likewise.
35092         * config/epiphany/epiphany.md (movcc): Likewise.
35093         * config/i386/i386.c (legitimize_pe_coff_extern_decl): Likewise.
35094         * config/m68k/m68k.c (struct _sched_ib, m68k_sched_variable_issue):
35095         Likewise.
35096         * config/mips/mips.c (mips_save_restore_reg): Likewise.
35097         * config/rx/rx.c (rx_is_restricted_memory_address): Likewise.
35098         * config/s390/s390.c (Z10_EARLYLOAD_DISTANCE): Likewise.
35099         * config/sh/sh.c (sh_rtx_costs): Likewise.
35100         * fold-const.c (fold_truth_andor): Likewise.
35101         * genautomata.c (collapse_flag): Likewise.
35102         * gengtype.h (struct type::u::s): Likewise.
35103         * gensupport.c (has_subst_attribute, add_mnemonic_string): Likewise.
35104         * input.c (FORMAT_AMOUNT): Likewise.
35105         * ipa-cp.c (class ipcp_lattice, agg_replacements_to_vector)
35106         (known_aggs_to_agg_replacement_list): Likewise.
35107         * ipa-inline-analysis.c: Likewise.
35108         * ipa-inline.h (estimate_edge_time, estimate_edge_hints): Likewise.
35109         * ipa-polymorphic-call.c
35110         (ipa_polymorphic_call_context::restrict_to_inner_class): Likewise.
35111         * loop-unroll.c (analyze_insn_to_expand_var): Likewise.
35112         * lra.c (lra_optional_reload_pseudos, lra_subreg_reload_pseudos):
35113         Likewise.
35114         * modulo-sched.c (apply_reg_moves): Likewise.
35115         * omp-expand.c (build_omp_regions_1): Likewise.
35116         * trans-mem.c (struct tm_wrapper_hasher): Likewise.
35117         * tree-ssa-loop-ivopts.c (may_eliminate_iv): Likewise.
35118         * tree-ssa-loop-niter.c (maybe_lower_iteration_bound): Likewise.
35119         * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
35120         * value-prof.c: Likewise.
35121         * var-tracking.c (val_reset): Likewise.
35123 2017-04-03  Richard Biener  <rguenther@suse.de>
35125         PR tree-optimization/80275
35126         * fold-const.c (split_address_to_core_and_offset): Handle
35127         POINTER_PLUS_EXPR.
35129 2017-04-03  Eric Botcazou  <ebotcazou@adacore.com>
35131         * tree-nested.c (get_descriptor_type): Make sure that the alignment of
35132         descriptors is at least equal to that of functions.
35134 2017-04-02  Uros Bizjak  <ubizjak@gmail.com>
35136         * config/i386/sse.md (movdi_to_sse): Add missing DONE.
35138 2017-04-02  Uros Bizjak  <ubizjak@gmail.com>
35140         PR target/80250
35141         * config/i386/sse.md (mov<IMOD4:mode>): Remove insn pattern.
35142         (mov<IMOD4:mode>): New expander.
35143         (*mov<IMOD4:mode>_internal): New insn and split pattern.
35145 2017-03-31  Segher Boessenkool  <segher@kernel.crashing.org>
35147         PR rtl-optimization/79405
35148         * fwprop.c (propagations_left): New variable.
35149         (forward_propagate_into): Decrement it.
35150         (fwprop_init): Initialize it.
35151         (fw_prop): If the variable has reached zero, stop propagating.
35152         (fwprop_addr): Ditto.
35154 2017-03-31  Jakub Jelinek  <jakub@redhat.com>
35156         PR debug/79255
35157         * dwarf2out.c (decls_for_scope): If BLOCK_NONLOCALIZED_VAR is
35158         a FUNCTION_DECL, pass it as decl instead of origin to
35159         process_scope_var.
35161 2017-03-31  Alexander Monakov  <amonakov@ispras.ru>
35163         * config/nvptx/nvptx.c (nvptx_output_softstack_switch): Correct format
35164         string.
35166 2017-03-31  Pat Haugen  <pthaugen@us.ibm.com>
35168         PR target/80107
35169         * config/rs6000/rs6000.md (extendhi<mode>2): Add test for
35170         TARGET_VSX_SMALL_INTEGER.
35172 2017-03-31  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
35174         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Add
35175         reference to the OpenPOWER 64-Bit ELF V2 ABI Specification.
35177 2017-03-31  Matthew Fortune  <matthew.fortune@imgtec.com>
35179         * config/mips/mips-msa.md (msa_vec_extract_<msafmt_f>): Update
35180         extraction from odd-numbered MSA register.
35182 2017-03-31  Jakub Jelinek  <jakub@redhat.com>
35184         PR middle-end/80173
35185         * expmed.c (store_bit_field_1): Don't attempt to create
35186         a word subreg out of hard registers wider than word if they
35187         have HARD_REGNO_NREGS of 1 for their mode.
35189         PR middle-end/80163
35190         * varasm.c (initializer_constant_valid_p_1): Disallow sign-extending
35191         conversions to integer types wider than word and pointer.
35193         PR debug/80025
35194         * cselib.h (rtx_equal_for_cselib_1): Add depth argument.
35195         (rtx_equal_for_cselib_p): Pass 0 to it.
35196         * cselib.c (cselib_hasher::equal): Likewise.
35197         (rtx_equal_for_cselib_1): Add depth argument.  If depth
35198         is 128, don't look up VALUE locs and punt.  Increment
35199         depth in recursive calls when walking VALUE locs.
35201 2017-03-31  Bernd Edlinger  <bernd.edlinger@hotmail.de>
35203         * gcov.c (md5sum_to_hex): Fix output of MD5 hex bytes.
35204         (make_gcov_file_name): Use the canonical path name for generating
35205         the MD5 value.
35206         (read_line): Fix handling of files with ascii null bytes.
35208 2017-03-30  Matthew Fortune  <matthew.fortune@imgtec.com>
35210         * config/mips/mips.c (mips_expand_vector_init): Create a const_vector
35211         to initialise a vector register instead
35212         of using a const_int.
35214 2017-03-30  Jakub Jelinek  <jakub@redhat.com>
35216         PR translation/80189
35217         * gimplify.c (omp_default_clause): Use %qs instead of %s in
35218         diagnostic messages.
35220 2017-03-30  Peter Bergner  <bergner@vnet.ibm.com>
35222         PR target/80246
35223         * config/rs6000/dfp.md (dfp_dxex_<mode>): Update mode of operand 0.
35224         (dfp_diex_<mode>): Update mode of operand 1.
35225         * doc/extend.texi (dxex, dxexq): Document change to return type.
35226         (diex, diexq): Document change to argument type.
35228 2017-03-30  Martin Jambor  <mjambor@suse.cz>
35230         PR ipa/77333
35231         * cgraph.h (cgraph_build_function_type_skip_args): Declare.
35232         * cgraph.c (redirect_call_stmt_to_callee): Set gimple fntype so that
35233         it reflects the signature changes performed at the callee side.
35234         * cgraphclones.c (build_function_type_skip_args): Make public, renamed
35235         to cgraph_build_function_type_skip_args.
35236         (build_function_decl_skip_args): Adjust call to the above function.
35238 2017-03-30  Jakub Jelinek  <jakub@redhat.com>
35240         PR target/80206
35241         * config/i386/sse.md
35242         (<extract_type>_vextract<shuffletype><extract_suf>_mask): Use
35243         register as dest whenever it is a MEM not rtx_equal_p to the
35244         corresponding dup operand, and when forcing into reg move the
35245         reg into the memory afterwards.
35246         (<extract_type_2>_vextract<shuffletype><extract_suf_2>_mask):
35247         Likewise.  Use <ssehalfvecmode> instead of <ssequartermode>
35248         for the force_reg mode.
35249         (avx512vl_vextractf128<mode>): Use register as dest either
35250         always when a MEM, or when it is a MEM not rtx_equal_p to the
35251         corresponding dup operand, or even not when it is a CONST_VECTOR
35252         depending on the mode and lo vs. hi.
35253         (avx512dq_vextract<shuffletype>64x2_1_maskm): Remove extraneous
35254         parens.
35255         (avx512f_vextract<shuffletype>32x4_1_maskm): Likewise.
35256         (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name>):
35257         Likewise.  Require that operands[2] is even.
35258         (<mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name>):
35259         Remove extraneous parens.  Require that operands[2] is a multiple
35260         of 4.
35261         (vec_extract_lo_<mode><mask_name>): Don't bother testing if
35262         operands[0] is a MEM if <mask_applied>, the predicates/constraints
35263         disallow memory then.
35265 2017-03-30  Richard Biener  <rguenther@suse.de>
35267         PR tree-optimization/77498
35268         * tree-ssa-pre.c (phi_translate_1): Do not allow simplifications
35269         to non-constants over backedges.
35271 2017-03-29  Segher Boessenkool  <segher@kernel.crashing.org>
35273         PR rtl-optimization/80233
35274         * combine.c (combine_instructions): Only take NONDEBUG_INSN_P insns
35275         as last_combined_insn.  Do not test for BARRIER_P separately.
35277 2017-03-29  Andreas Schwab  <schwab@suse.de>
35279         PR ada/80146
35280         * calls.c (prepare_call_address): Convert funexp to Pmode before
35281         copying to temp reg.
35283 2017-03-29  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
35285         PR tree-optimization/80158
35286         * gimple-ssa-strength-reduction.c (replace_mult_candidate):
35287         Handle possible future case of more than one alternate
35288         interpretation.
35289         (replace_rhs_if_not_dup): Likewise.
35290         (replace_one_candidate): Likewise.
35292 2017-03-28  Vladimir Makarov  <vmakarov@redhat.com>
35294         PR rtl-optimization/80193
35295         * ira.c (ira): Do not check allocation for LRA.
35297 2017-03-28  Alexander Monakov  <amonakov@ispras.ru>
35299         * config/nvptx/nvptx-protos.h (nvptx_output_simt_enter): Declare.
35300         (nvptx_output_simt_exit): Declare.
35301         * config/nvptx/nvptx.c (nvptx_init_unisimt_predicate): Use
35302         cfun->machine->unisimt_location.  Handle NULL unisimt_predicate.
35303         (init_softstack_frame): Move initialization of crtl->is_leaf to...
35304         (nvptx_declare_function_name): ...here.  Emit declaration of local
35305         memory space buffer for omp_simt_enter insn.
35306         (nvptx_output_unisimt_switch): New.
35307         (nvptx_output_softstack_switch): New.
35308         (nvptx_output_simt_enter): New.
35309         (nvptx_output_simt_exit): New.
35310         * config/nvptx/nvptx.h (struct machine_function): New fields
35311         has_simtreg, unisimt_location, simt_stack_size, simt_stack_align.
35312         * config/nvptx/nvptx.md (UNSPECV_SIMT_ENTER): New unspec.
35313         (UNSPECV_SIMT_EXIT): Ditto.
35314         (omp_simt_enter_insn): New insn.
35315         (omp_simt_enter): New expansion.
35316         (omp_simt_exit): New insn.
35317         * config/nvptx/nvptx.opt (msoft-stack-reserve-local): New option.
35319         * internal-fn.c (expand_GOMP_SIMT_ENTER): New.
35320         (expand_GOMP_SIMT_ENTER_ALLOC): New.
35321         (expand_GOMP_SIMT_EXIT): New.
35322         * internal-fn.def (GOMP_SIMT_ENTER): New internal function.
35323         (GOMP_SIMT_ENTER_ALLOC): Ditto.
35324         (GOMP_SIMT_EXIT): Ditto.
35325         * target-insns.def (omp_simt_enter): New insn.
35326         (omp_simt_exit): Ditto.
35327         * omp-low.c (struct omplow_simd_context): New fields simt_eargs,
35328         simt_dlist.
35329         (lower_rec_simd_input_clauses): Implement SIMT privatization.
35330         (lower_rec_input_clauses): Likewise.
35331         (lower_lastprivate_clauses): Handle SIMT privatization.
35333         * omp-offload.c: Include langhooks.h, tree-nested.h, stor-layout.h.
35334         (ompdevlow_adjust_simt_enter): New.
35335         (find_simtpriv_var_op): New.
35336         (execute_omp_device_lower): Handle IFN_GOMP_SIMT_ENTER,
35337         IFN_GOMP_SIMT_ENTER_ALLOC, IFN_GOMP_SIMT_EXIT.
35339         * tree-inline.h (struct copy_body_data): New field dst_simt_vars.
35340         * tree-inline.c (expand_call_inline): Handle SIMT privatization.
35341         (copy_decl_for_dup_finish): Ditto.
35343         * tree-ssa.c (execute_update_addresses_taken): Handle GOMP_SIMT_ENTER.
35345 2017-03-28  Uros Bizjak  <ubizjak@gmail.com>
35347         PR target/53383
35348         * config/i386/i386.c (ix86_option_override_internal): Always
35349         allow -mpreferred-stack-boundary=3 for 64-bit targets.
35351 2017-03-28  Bin Cheng  <bin.cheng@arm.com>
35353         * tree-vect-loop.c (optimize_mask_stores): Add bb to the right loop.
35355 2017-03-28  Bin Cheng  <bin.cheng@arm.com>
35357         * tree-vect-loop-manip.c (slpeel_add_loop_guard): New param and
35358         mark new edge's irreducible flag accordign to it.
35359         (vect_do_peeling): Check loop preheader edge's irreducible flag
35360         and pass it to function slpeel_add_loop_guard.
35362 2017-03-28  Richard Sandiford  <richard.sandiford@arm.com>
35364         PR tree-optimization/80218
35365         * tree-call-cdce.c (shrink_wrap_one_built_in_call_with_conds):
35366         Update block frequencies and counts.
35368 2017-03-28  Richard Biener  <rguenther@suse.de>
35370         PR tree-optimization/78644
35371         * tree-ssa-ccp.c (evaluate_stmt): When we may not use the value
35372         of a simplification result we may not use it at all.
35374 2017-03-28  Richard Biener  <rguenther@suse.de>
35376         PR ipa/80205
35377         * tree-inline.c (copy_phis_for_bb): Do not create PHI node
35378         without arguments, generate default definition of a SSA name.
35380 2017-03-28  Richard Biener  <rguenther@suse.de>
35382         PR middle-end/80222
35383         * gimple-fold.c (gimple_fold_indirect_ref): Do not touch
35384         TYPE_REF_CAN_ALIAS_ALL references.
35385         * fold-const.c (fold_indirect_ref_1): Likewise.
35387 2017-03-28  Martin Liska  <mliska@suse.cz>
35389         PR ipa/80104
35390         * cgraphunit.c (cgraph_node::expand_thunk): Mark argument of a
35391         thunk call as DECL_GIMPLE_REG_P when vector or complex type.
35393 2017-03-28  Claudiu Zissulescu  <claziss@synopsys.com>
35394             Thomas Petazzoni  <thomas.petazzoni@free-electrons.com>
35396         * config/arc/arc.h (CPP_SPEC): Add subtarget_cpp_spec.
35397         (EXTRA_SPECS): Define.
35398         (SUBTARGET_EXTRA_SPECS): Likewise.
35399         (SUBTARGET_CPP_SPEC): Likewise.
35400         * config/arc/elf.h (EXTRA_SPECS): Renamed to
35401         SUBTARGET_EXTRA_SPECS.
35402         * config/arc/linux.h (SUBTARGET_CPP_SPEC): Define.
35404 2017-03-28  Claudiu Zissulescu  <claziss@synopsys.com>
35406         * config/arc/simdext.md (vst64_insn): Update pattern.
35407         (vld32wh_insn): Likewise.
35408         (vld32wl_insn): Likewise.
35409         (vld64_insn): Likewise.
35410         (vld32_insn): Likewise.
35412 2017-03-28  Marek Polacek  <polacek@redhat.com>
35414         PR sanitizer/80067
35415         * fold-const.c (fold_comparison): Use protected_set_expr_location
35416         instead of SET_EXPR_LOCATION.
35418 2017-03-28  Markus Trippelsdorf  <markus@trippelsdorf.de>
35420         * tree.c (add_expr): Avoid name lookup warning.
35422 2017-03-27  Jeff Law  <law@redhat.com>
35424         PR tree-optimization/80216
35425         * tree-ssa-dom.c (derive_equivalences_from_bit_ior): Fix typo in
35426         function name.  Limit recursion depth.
35427         (record_temporary_equivalences): Corresponding changes.
35429 2017-03-27  Jonathan Wakely  <jwakely@redhat.com>
35431         * doc/invoke.texi (-Wno-narrowing): Reorder so default behavior is
35432         covered first.
35434 2017-03-27  Jakub Jelinek  <jakub@redhat.com>
35436         PR target/80102
35437         * reg-notes.def (REG_CFA_NOTE): Define.  Use it for CFA related
35438         notes.
35439         * cfgcleanup.c (reg_note_cfa_p): New array.
35440         (insns_have_identical_cfa_notes): New function.
35441         (old_insns_match_p): Don't cross-jump in between /f
35442         and non-/f instructions.  If both i1 and i2 are frame related,
35443         verify all CFA notes, their order and content.
35445 2017-03-27  Michael Meissner  <meissner@linux.vnet.ibm.com>
35447         PR target/78543
35448         * config/rs6000/rs6000.md (bswaphi2_extenddi): Combine bswap
35449         HImode and SImode with zero extend to DImode to one insn.
35450         (bswap<mode>2_extenddi): Likewise.
35451         (bswapsi2_extenddi): Likewise.
35452         (bswaphi2_extendsi): Likewise.
35453         (bswaphi2): Combine bswap HImode and SImode into one insn.
35454         Separate memory insns from swapping register.
35455         (bswapsi2): Likewise.
35456         (bswap<mode>2): Likewise.
35457         (bswaphi2_internal): Delete, no longer used.
35458         (bswapsi2_internal): Likewise.
35459         (bswap<mode>2_load): Split bswap HImode/SImode into separate load,
35460         store, and gpr<-gpr swap insns.
35461         (bswap<mode>2_store): Likewise.
35462         (bswaphi2_reg): Register only splitter, combine with the splitter.
35463         (bswaphi2 splitter): Likewise.
35464         (bswapsi2_reg): Likewise.
35465         (bswapsi2 splitter): Likewise.
35466         (bswapdi2): If we have the LDBRX and STDBRX instructions, split
35467         the insns into load, store, and register/register insns.
35468         (bswapdi2_ldbrx): Likewise.
35469         (bswapdi2_load): Likewise.
35470         (bswapdi2_store): Likewise.
35471         (bswapdi2_reg): Likewise.
35473 2017-03-27  Gunther Nikl  <gnikl@users.sourceforge.net>
35475         * system.h (HAVE_DESIGNATED_INITIALIZERS): Fix non C++ case.
35476         (HAVE_DESIGNATED_UNION_INITIALIZERS): Likewise.
35478 2017-03-27  Kelvin Nilsen  <kelvin@gcc.gnu.org>
35480         PR target/80103
35481         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Edit and
35482         add comments.
35483         * config/rs6000/rs6000.c (rs6000_option_override_internal): Add
35484         special handling for target option conflicts between dform
35485         options (-mpower9-dform, -mpower9-dform-vector,
35486         -mpower9-dform-scalar) and -mno-direct-move.
35488 2017-03-27  Richard Biener  <rguenther@suse.de>
35490         PR tree-optimization/80181
35491         * tree-ssa-ccp.c (likely_value): UNDEFINED ^ X is UNDEFINED.
35493 2017-03-27  Claudiu Zissulescu  <claziss@synopsys.com>
35495         * config/arc/predicates.md (move_double_src_operand): Replace the
35496         call to move_double_src_operand with a call to address_operand.
35498 2017-03-27  Claudiu Zissulescu  <claziss@synopsys.com>
35500         * config/arc/elf.h (ARGET_ARC_TP_REGNO_DEFAULT): Define.
35501         * config/arc/linux.h (ARGET_ARC_TP_REGNO_DEFAULT): Likewise.
35502         * config/arc/arc.opt (mtp-regno): Use ARGET_ARC_TP_REGNO_DEFAULT.
35504 2017-03-27  Claudiu Zissulescu  <claziss@synopsys.com>
35506         * config/arc/predicates.md (long_immediate_loadstore_operand):
35507         Consider scaled addresses cases.
35509 2017-03-27  Claudiu Zissulescu  <claziss@synopsys.com>
35511         * config/arc/arc.c (arc_epilogue_uses): BLINK should be also
35512         restored when in interrupt.
35513         * config/arc/arc.md (simple_return): ARCv2 rtie instruction
35514         doesn't have delay slot.
35516 2017-03-27  Richard Biener  <rguenther@suse.de>
35518         PR ipa/79776
35519         * tree-ssa-structalias.c (associate_varinfo_to_alias): Skip
35520         inlined thunk clones.
35522 2017-03-27  Jakub Jelinek  <jakub@redhat.com>
35524         PR sanitizer/80168
35525         * asan.c (instrument_derefs): Copy over last operand from
35526         original COMPONENT_REF to the new COMPONENT_REF with
35527         DECL_BIT_FIELD_REPRESENTATIVE.
35528         * ubsan.c (instrument_object_size): Likewise.
35530 2017-03-27  Richard Biener  <rguenther@suse.de>
35532         PR tree-optimization/80170
35533         * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Make
35534         sure DR/SCEV didnt fold in constants we do not see when looking
35535         at the reference base alignment.
35537 2017-03-27  Richard Biener  <rguenther@suse.de>
35539         PR middle-end/80171
35540         * gimple-fold.c (fold_ctor_reference): Properly guard against
35541         NULL return value from canonicalize_constructor_val.
35543 2017-03-25  Uros Bizjak  <ubizjak@gmail.com>
35545         PR target/80180
35546         * config/i386/i386.c (ix86_expand_builtin)
35547         <IX86_BUILTIN_RDSEED{16,32,64}_STEP>: Do not expand arg0 between
35548         flags reg setting and flags reg using instructions.
35549         <IX86_BUILTIN_RDRAND{16,32,64}_STEP>: Ditto.  Use non-flags reg
35550         clobbering instructions to zero extend op2.
35552 2017-03-25  Gerald Pfeifer  <gerald@pfeifer.com>
35554         * doc/install.texi (Configuration) <--with-aix-soname>:
35555         Update link to AIX ld.
35557 2017-03-25  Bernd Schmidt  <bschmidt@redhat.com>
35559         PR rtl-optimization/80160
35560         PR rtl-optimization/80159
35561         * lra-assigns.c (must_not_spill_p): Tighten new test to also take
35562         reg_alternate_class into account.
35564 2017-03-24  Vladimir Makarov  <vmakarov@redhat.com>
35566         PR target/80148
35567         * lra-assigns.c (assign_by_spills): Add spilled non-reload pseudos
35568         to consider in curr_insn_transform.
35570 2017-03-24  Jakub Jelinek  <jakub@redhat.com>
35572         * genrecog.c (validate_pattern): Add VEC_SELECT validation.
35573         * genmodes.c (emit_min_insn_modes_c): Call emit_mode_nunits
35574         and emit_mode_inner.
35576 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35578         * config/s390/s390-builtins.def: Add VXE builtins.  Add a flags
35579         argument to the overloaded builtin variants.  Use the new flag to
35580         deprecate certain builtin variants.
35581         * config/s390/s390-builtin-types.def: Add new builtin types.
35582         * config/s390/s390-builtins.h: Support new flags field for
35583         overloaded builtins.
35584         * config/s390/s390-c.c (OB_DEF_VAR): New flags field.
35585         (s390_macro_to_expand): Enable vector float data type.
35586         (s390_cpu_cpp_builtins_internal): Indicate support of the new
35587         builtins by incrementing the __VEC__ version number.
35588         (s390_expand_overloaded_builtin): Support expansion of vec_xl and
35589         vec_xst.
35590         (s390_resolve_overloaded_builtin): Emit error messages depending
35591         on the builtin flags.
35592         * config/s390/s390.c (s390_expand_builtin): Support additional
35593         flags argument.  Change error message to match the messages
35594         emitted in s390-c.c.
35595         * config/s390/s390.md: New UNSPEC_* constants.
35596         (op_type): Add new instruction types.
35597         * config/s390/vecintrin.h: Add new builtins and test data class
35598         constants.
35599         * config/s390/vx-builtins.md (V_HW_32_64): Add V4SF.
35600         (V_HW_4, VEC_HW, VECF_HW): New mode iterators.
35601         (VEC_INEXACT, VEC_NOINEXACT): New constants.
35602         ("vec_splats<mode>", "vec_insert<mode>", "vec_promote<mode>")
35603         ("vec_insert_and_zero<mode>", "vec_mergeh<mode>")
35604         ("vec_mergel<mode>"): V_HW -> VEC_HW.
35606         ("vlrlrv16qi", "vstrlrv16qi", "vbpermv16qi", "vec_msumv2di")
35607         ("vmslg", "*vftci<mode>_cconly", "vftci<mode>_intcconly")
35608         ("*vftci<mode>", "vftci<mode>_intcc", "vec_double_s64")
35609         ("vec_double_u64", "vfmin<mode>", "vfmax<mode>"): New definition.
35611         ("and_av2df3", "and_cv2df3", "vec_andc_av2df3")
35612         ("vec_andc_cv2df3", "xor_av2df3", "xor_cv2df3", "vec_nor_av2df3")
35613         ("vec_nor_cv2df3", "ior_av2df3", "ior_cv2df3", "vec_nabs")
35614         ("*vftcidb", "*vftcidb_cconly", "vftcidb"): Remove definition.
35616         ("vec_all_<fpcmpcc:code>v2df", "vec_any_<fpcmpcc:code>v2df")
35617         ("vec_scatter_elementv4si_DI", "vec_cmp<fpcmp:code>v2df")
35618         ("vec_di_to_df_s64", "vec_di_to_df_u64", "vec_df_to_di_u64")
35619         ("vfidb", "*vldeb", "*vledb", "*vec_cmp<insn_cmp>v2df_cconly")
35620         ("vec_cmpeqv2df_cc", "vec_cmpeqv2df_cc", "vec_cmphv2df_cc")
35621         ("vec_cmphev2df_cc", "*vec_cmpeqv2df_cc")
35622         ("*vec_cmphv2df_cc", "*vec_cmphev2df_cc"): Enable new modes as ...
35624         ("vec_all_<fpcmpcc:code><mode>", "vec_any_<fpcmpcc:code><mode>")
35625         ("vec_scatter_element<V_HW_4:mode>_DI")
35626         ("vec_cmp<fpcmp:code><mode>", "vcdgb", "vcdlgb", "vclgdb")
35627         ("vec_fpint<mode>", "vflls")
35628         ("vflrd", "*vec_cmp<insn_cmp><mode>_cconly", "vec_cmpeq<mode>_cc")
35629         ("vec_cmpeq<mode>_cc", "vec_cmph<mode>_cc", "vec_cmphe<mode>_cc")
35630         ("*vec_cmpeq<mode>_cc", "*vec_cmph<mode>_cc")
35631         ("*vec_cmphe<mode>_cc"): ... these.
35633         ("vec_ctd_s64", "vec_ctsl", "vec_ctul", "vec_st2f"): Use rounding
35634         mode constant instead of magic value.
35636 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35638         * config/s390/s390.c (s390_expand_vec_compare): Support other
35639         vector floating point modes than just V2DF.
35640         (s390_expand_vcond): Likewise.
35641         (s390_hard_regno_mode_ok): Allow SFmode values in VRs.
35642         (s390_cannot_change_mode_class): Prevent mode changes between TF
35643         and V1TF in vector registers.
35644         * config/s390/s390.md (DF, SF): New mode attributes.
35645         ("*cmp<mode>_ccs", "add<mode>3", "sub<mode>3", "mul<mode>3")
35646         ("fma<mode>4", "fms<mode>4", "div<mode>3", "*neg<mode>2"): Add
35647         SFmode support for VRs.
35648         * config/s390/vector.md (V_HW, V_HW2, VT_HW, ti*, nonvec): Add new
35649         vector fp modes.
35650         (VFT, VF_HW): New mode iterators.
35651         (vw, sdx): New mode attributes.
35652         ("addv2df3", "subv2df3", "mulv2df3", "divv2df3", "sqrtv2df2")
35653         ("fmav2df4","fmsv2df4", "negv2df2", "absv2df2", "*negabsv2df2")
35654         ("smaxv2df3", "sminv2df3", "*vec_cmp<VFCMP_HW_OP:code>v2df_nocc")
35655         ("vec_cmpuneqv2df", "vec_cmpltgtv2df", "vec_orderedv2df")
35656         ("vec_unorderedv2df"): Adjust the v2df only patterns to support
35657         also the new vector floating point modes.  Renaming to ...
35659         ("add<mode>3", "sub<mode>3", "mul<mode>3", "div<mode>3")
35660         ("sqrt<mode>2", "fma<mode>4", "fms<mode>4", "neg<mode>2")
35661         ("abs<mode>2", "negabs<mode>2", "smax<mode>3")
35662         ("smin<mode>3", "*vec_cmp<VFCMP_HW_OP:code><mode>_nocc")
35663         ("vec_cmpuneq<mode>", "vec_cmpltgt<mode>", "vec_ordered<mode>")
35664         ("vec_unordered<mode>"): ... these.
35666         ("neg_fma<mode>4", "neg_fms<mode>4", "*smax<mode>3_vxe")
35667         ("*smin<mode>3_vxe", "*sminv2df3_vx", "*vec_extendv4sf")
35668         ("*vec_extendv2df"): New insn definitions.
35670 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35672         * config/s390/s390.md ("*adddi3_sign", "*subdi3_sign", "mulditi3")
35673         ("mulditi3_2", "*muldi3_sign"): New patterns.
35674         ("muldi3", "*muldi3", "mulsi3", "*mulsi3"): Add an expander and
35675         rename the pattern definition.
35677 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35679         * config/s390/s390.md ("indirect_jump"): Turn insn definition into
35680         expander.
35681         ("*indirect_jump", "*indirect2_jump"): New pattern definitions.
35683 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35685         * config/s390/s390.c (s390_expand_vec_init): Use vllezl
35686         instruction if possible.
35687         * config/s390/vector.md (vec_halfnumelts): New mode
35688         attribute.
35689         ("*vec_vllezlf<mode>"): New pattern.
35691 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35693         * config/s390/vector.md ("popcountv16qi2", "popcountv8hi2")
35694         ("popcountv4si2", "popcountv2di2"): Rename to ...
35695         ("popcount<mode>2", "popcountv8hi2_vx", "popcountv4si2_vx")
35696         ("popcountv2di2_vx"): ... these and add !TARGET_VXE to the
35697         condition.
35698         ("popcount<mode>2_vxe"): New pattern.
35700 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35702         * common/config/s390/s390-common.c (processor_flags_table): Add
35703         arch12.
35704         * config.gcc: Add arch12.
35705         * config/s390/driver-native.c (s390_host_detect_local_cpu):
35706         Default to arch12 for unknown CPU model numbers.
35707         * config/s390/s390-builtins.def: Add B_VXE builtin flag.
35708         * config/s390/s390-c.c (s390_cpu_cpp_builtins_internal): Adjust
35709         PROCESSOR_max sanity check.
35710         * config/s390/s390-opts.h (enum processor_type): Add
35711         PROCESSOR_ARCH12.
35712         * config/s390/s390.c (processor_table): Add arch12.
35713         (s390_expand_builtin): Add check for B_VXE flag.
35714         (s390_issue_rate): Add PROCESSOR_ARCH12.
35715         (s390_get_sched_attrmask): Likewise.
35716         (s390_get_unit_mask): Likewise.
35717         (s390_sched_score): Enable z13 scheduling for arch12.
35718         (s390_sched_reorder): Likewise.
35719         (s390_sched_variable_issue): Likewise.
35720         * config/s390/s390.h (enum processor_flags): Add PF_ARCH12 and
35721         PF_VXE.
35722         (s390_tune_attr): Use z13 scheduling also for arch12.
35723         (TARGET_CPU_ARCH12, TARGET_CPU_ARCH12_P, TARGET_CPU_VXE)
35724         (TARGET_CPU_VXE_P, TARGET_ARCH12, TARGET_ARCH12_P, TARGET_VXE)
35725         (TARGET_VXE_P): New macros.
35726         * config/s390/s390.md: Add arch12 to cpu attribute.  Add arch12
35727         and vxe to cpu_facility.  Add arch12 and vxe to enabled attribute.
35728         * config/s390/s390.opt: Add arch12 as processor_type.
35730 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35732         * config/s390/s390.md
35733         ("fixuns_truncdddi2", "fixuns_trunctddi2")
35734         ("fixuns_trunc<BFP:mode><GPR:mode>2"): Merge into ...
35735         ("fixuns_trunc<FP:mode><GPR:mode>2"): New expander.
35737         ("fixuns_trunc<BFP:mode><GPR:mode>2", "fixuns_trunc<mode>si2"):
35738         Rename expanders to ...
35740         ("fixuns_trunc<BFP:mode><GPR:mode>2_emu")
35741         ("fixuns_truncdddi2_emu"): ... these.
35743         ("fixuns_trunc<mode>si2_emu"): New expander.
35745         ("*fixuns_truncdfdi2_z13"): Rename to ...
35746         ("*fixuns_truncdfdi2_vx"): ... this.
35748 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35750         * config/s390/2964.md: Remove the single element vector compare
35751         instructions which are no longer used.
35752         * config/s390/s390.c (s390_select_ccmode): Remove handling of
35753         vector CCmodes.
35754         (s390_canonicalize_comparison): Remove handling of DFmode
35755         compares.
35756         (s390_expand_vec_compare_scalar): Remove function.
35757         (s390_emit_compare): Don't call s390_expand_vec_compare_scalar.
35758         * config/s390/s390.md ("*vec_cmp<insn_cmp>df_cconly"): Remove
35759         pattern.
35760         ("*cmp<mode>_ccs"): Add wfcdb instruction.
35762 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35764         * config/s390/s390.md ("mov<mode>_64dfp" DD_DF): Use vleig for loading a
35765         FP zero.
35766         ("*mov<mode>_64" DD_DF): Remove the vector instructions. These
35767         will anyway by matched by mov<mode>_64dfp.
35769 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35771         * config/s390/s390.md ("mov<mode>" SD_SF): Change vleg/vsteg to
35772         vlef/vstef.  Add missing operand to vleif.
35774 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35776         * config/s390/s390.c (s390_expand_vec_init): Enable vector load
35777         pair for all vector types with 64 bit elements.
35778         * config/s390/vx-builtins.md (V_HW_64): Move mode iterator to ...
35779         * config/s390/vector.md (V_HW_64): ... here.
35780         (V_128_NOSINGLE): New mode iterator.
35781         ("vec_init<V_HW:mode>"): Use V_128 as mode iterator.
35782         ("*vec_splat<mode>"): Use V_128_NOSINGLE mode iterator.
35783         ("*vec_tf_to_v1tf", "*vec_ti_to_v1ti"): New pattern definitions.
35784         ("*vec_load_pairv2di"): Change to ...
35785         ("*vec_load_pair<mode>"): ... this one.
35787 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35789         * config/s390/constraints.md: Add comments.
35790         (jKK): Reject element sizes > 8 bytes.
35791         * config/s390/s390.c (s390_split_ok_p): Enable splitting also for
35792         s_operands.
35793         * config/s390/s390.md: Add the s_operand checks formerly in
35794         s390_split_ok_p to various splitters where they are still
35795         required.
35796         * config/s390/vector.md ("mov<mode>" V_128): Add GPR alternatives
35797         for 128 bit vectors.  Plus two splitters.
35799 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35801         * config/s390/s390.md: Rename the cpu facilty vec to vx throughout
35802         the file.
35804 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35806         PR target/79893
35807         * config/s390/s390-c.c (s390_adjust_builtin_arglist): Issue an
35808         error if the boundary argument is not constant.
35810 2017-03-24  Jakub Jelinek  <jakub@redhat.com>
35812         PR rtl-optimization/80112
35813         * loop-doloop.c (doloop_condition_get): Don't check condition
35814         if cmp isn't SET with IF_THEN_ELSE src.
35816 2017-03-24  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
35818         PR tree-optimization/80158
35819         * gimple-ssa-strength-reduction.c (replace_mult_candidate): When
35820         replacing a candidate statement, also replace it for the
35821         candidate's alternate interpretation.
35822         (replace_rhs_if_not_dup): Likewise.
35823         (replace_one_candidate): Likewise.
35825 2017-03-24  Richard Biener  <rguenther@suse.de>
35827         PR tree-optimization/80167
35828         * graphite-isl-ast-to-gimple.c
35829         (translate_isl_ast_to_gimple::is_valid_rename): Handle default-defs
35830         properly.
35831         (translate_isl_ast_to_gimple::get_rename): Likewise.
35833 2017-03-23  Kelvin Nilsen  <kelvin@gcc.gnu.org>
35835         * config/rs6000/rs6000.c (rs6000_option_override_internal): Change
35836         handling of certain combinations of target options, including the
35837         combinations -mpower8-vector vs. -mno-vsx, -mpower9-vector vs.
35838         -mno-power8-vector, and -mpower9_dform vs. -mno-power9-vector.
35840 2017-03-23  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
35842         PR target/71436
35843         * config/arm/arm.md (*load_multiple): Add reload_completed to
35844         matching condition.
35846 2017-03-23  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
35847             Richard Biener  <rguenth@suse.de>
35849         PR tree-optimization/79908
35850         PR tree-optimization/80136
35851         * tree-stdarg.c (expand_ifn_va_arg_1): For a VA_ARG whose LHS has
35852         been cast away, gimplify_and_add suffices.
35854 2017-03-23  Markus Trippelsdorf  <markus@trippelsdorf.de>
35856         * tree-vrp.c (identify_jump_threads): Delete avail_exprs.
35858 2017-03-23  Richard Biener  <rguenther@suse.de>
35860         PR tree-optimization/80032
35861         * gimplify.c (gimple_push_cleanup): Forced unconditional
35862         cleanups still have to go to the conditional_cleanups
35863         sequence.
35865 2017-03-22  Jakub Jelinek  <jakub@redhat.com>
35867         PR tree-optimization/80072
35868         * tree-ssa-reassoc.c (struct operand_entry): Change id field type
35869         to unsigned int.
35870         (next_operand_entry_id): Change type to unsigned int.
35871         (sort_by_operand_rank): Make sure to return the right return value
35872         even if unsigned fields are bigger than INT_MAX.
35873         (struct oecount): Change cnt and id type to unsigned int.
35874         (oecount_hasher::equal): Formatting fix.
35875         (oecount_cmp): Make sure to return the right return value
35876         even if unsigned fields are bigger than INT_MAX.
35877         (undistribute_ops_list): Change next_oecount_id type to unsigned int.
35879         PR c++/80129
35880         * gimplify.c (gimplify_modify_expr_rhs) <case COND_EXPR>: Clear
35881         TREE_READONLY on result if writing it more than once.
35883         PR sanitizer/80110
35884         * doc/invoke.texi (-fsanitize=thread): Document that with
35885         -fnon-call-exceptions atomics are not able to throw
35886         exceptions.
35888         PR sanitizer/80110
35889         * tsan.c: Include tree-eh.h.
35890         (instrument_builtin_call): Call maybe_clean_eh_stmt or
35891         maybe_clean_or_replace_eh_stmt where needed.
35892         (instrument_memory_accesses): Add cfg_changed argument.
35893         Call gimple_purge_dead_eh_edges on each block and set *cfg_changed
35894         if it returned true.
35895         (tsan_pass): Adjust caller.  Return TODO_cleanup_cfg if cfg_changed.
35897         PR rtl-optimization/63191
35898         * config/i386/i386.c (ix86_delegitimize_address): Turn into small
35899         wrapper function, moved the whole old content into ...
35900         (ix86_delegitimize_address_1): ... this.  New inline function.
35901         (ix86_find_base_term): Use ix86_delegitimize_address_1 with
35902         true as last argument instead of ix86_delegitimize_address.
35904 2017-03-22  Wilco Dijkstra  <wdijkstr@arm.com>
35906         * config/aarch64/aarch64.c (generic_branch_cost): Copy
35907         cortexa57_branch_cost.
35909 2017-03-22  Wilco Dijkstra  <wdijkstr@arm.com>
35911         * config/aarch64/aarch64.c (generic_tunings): Add AES fusion.
35913 2017-03-21  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
35915         PR target/80123
35916         * doc/md.texi (Constraints): Document wA constraint.
35917         * config/rs6000/constraints.md (wA): New.
35918         * config/rs6000/rs6000.c (rs6000_debug_reg_global): Add wA reg_class.
35919         (rs6000_init_hard_regno_mode_ok): Init wA constraint.
35920         * config/rs6000/rs6000.h (RS6000_CONSTRAINT_wA): New.
35921         * config/rs6000/vsx.md (vsx_splat_<mode>): Use wA constraint.
35923 2017-03-22  Cesar Philippidis  <cesar@codesourcery.com>
35925         PR c++/80029
35926         * gimplify.c (is_oacc_declared): New function.
35927         (oacc_default_clause): Use it to set default flags for acc declared
35928         variables inside parallel regions.
35929         (gimplify_scan_omp_clauses): Strip firstprivate pointers for acc
35930         declared variables.
35931         (gimplify_oacc_declare): Gimplify the declare clauses.  Add the
35932         declare attribute to any decl as necessary.
35934 2017-03-22  Thomas Preud'homme  <thomas.preudhomme@arm.com>
35936         PR target/80082
35937         * config/arm/arm-isa.h (isa_bit_lpae): New feature bit.
35938         (ISA_ARMv7ve): Add isa_bit_lpae to the definition.
35939         * config/arm/arm-protos.h (arm_arch7ve): Rename into ...
35940         (arm_arch_lpae): This.
35941         * config/arm/arm.c (arm_arch7ve): Rename into ...
35942         (arm_arch_lpae): This.  Define it in term of isa_bit_lpae.
35943         * config/arm/arm.h (TARGET_HAVE_LPAE): Redefine in term of
35944         arm_arch_lpae.
35946 2017-03-22  Martin Liska  <mliska@suse.cz>
35948         PR target/79906
35949         * config/rs6000/rs6000.c (rs6000_inner_target_options): Show
35950         error message instead of an ICE.
35952 2017-03-21  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
35954         * doc/extend.texi (6.11 Additional Floating Types): Revise.
35956 2017-03-21  Kelvin Nilsen  <kelvin@gcc.gnu.org>
35958         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Add
35959         comments.
35960         * config/rs6000/rs6000.c (rs6000_option_override_internal): Add
35961         comments.
35963 2017-03-21  Martin Sebor  <msebor@redhat.com>
35965         * doc/extend.texi: Use "cannot" instead of "can't."
35966         * doc/hostconfig.texi: Same.
35967         * doc/install.texi: Same.
35968         * doc/invoke.texi: Same.
35969         * doc/loop.texi: Same.
35970         * doc/md.texi: Same.
35971         * doc/objc.texi: Same.
35972         * doc/rtl.texi: Same.
35973         * doc/tm.texi: Same.
35974         * doc/tm.texi.in: Same.
35975         * doc/trouble.texi: Same.
35977 2017-03-21  Alexandre Oliva  <aoliva@redhat.com>
35979         PR debug/63238
35980         * dwarf2out.c (struct checksum_attributes): Add at_alignment.
35981         (collect_checksum_attributes): Set it.
35982         (die_checksum_ordered): Use it.
35984 2017-03-21  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
35986         PR tree-optimization/79908
35987         * tree-stdarg.c (expand_ifn_va_arg_1): Revert the following
35988         change: For a VA_ARG whose LHS has been cast away, use
35989         force_gimple_operand to construct the side effects.
35991 2017-03-21  David Malcolm  <dmalcolm@redhat.com>
35993         PR translation/80001
35994         * omp-offload.c (oacc_loop_fixed_partitions): Make diagnostics
35995         more amenable to translation.
35996         (oacc_loop_auto_partitions): Likewise.
35998 2017-03-21  Marek Polacek  <polacek@redhat.com>
35999             Martin Sebor  <msebor@redhat.com>
36001         PR tree-optimization/80109
36002         * gimple-ssa-warn-alloca.c (alloca_call_type): Only call get_range_info
36003         on INTEGRAL_TYPE_P.
36005 2017-03-21  Jakub Jelinek  <jakub@redhat.com>
36006             Segher Boessenkool  <segher@kernel.crashing.org>
36008         PR target/80125
36009         * combine.c (can_combine_p): Revert the 2017-03-20 change, only
36010         check reg_used_between_p between insn and one of succ or succ2
36011         depending on if succ is artificial insn not inserted into insn
36012         stream.
36014 2017-03-21  Martin Liska  <mliska@suse.cz>
36016         PR gcov-profile/80081
36017         * Makefile.in: Add gcov-dump and fix installation of gcov-tool.
36018         * doc/gcc.texi: Include gcov-dump stuff.
36019         * doc/gcov-dump.texi: New file.
36021 2017-03-21  Toma Tabacu  <toma.tabacu@imgtec.com>
36023         PR rtl-optimization/79150
36024         * config/mips/mips.c (mips_block_move_loop): Emit a NOP after the
36025         conditional jump, if the jump is the last insn of the loop.
36027 2017-03-21  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
36028             Richard Biener  <rguenth@suse.de>
36030         PR tree-optimization/79908
36031         * tree-stdarg.c (expand_ifn_va_arg_1): For a VA_ARG whose LHS has
36032         been cast away, use force_gimple_operand to construct the side
36033         effects.
36035 2017-03-21  Martin Liska  <mliska@suse.cz>
36037         PR libfortran/79956
36038         * simplify-rtx.c (simplify_immed_subreg): Initialize a variable
36039         to NULL.
36041 2017-03-21  Brad Spengler <spender@grsecurity.net>
36043         PR plugins/80094
36044         * plugin.c (htab_hash_plugin): New function.
36045         (add_new_plugin): Use it and adjust.
36046         (parse_plugin_arg_opt): Adjust.
36047         (init_one_plugin): Likewise.
36049 2017-03-21  Richard Biener  <rguenther@suse.de>
36051         PR tree-optimization/80032
36052         * gimplify.c (gimple_push_cleanup): Add force_uncond parameter,
36053         if set force the cleanup to happen unconditionally.
36054         (gimplify_target_expr): Push inserted clobbers with force_uncond
36055         to avoid them being removed by control-dependent DCE.
36057 2017-03-21  Richard Biener  <rguenther@suse.de>
36059         PR tree-optimization/80122
36060         * tree-inline.c (copy_bb): Do not expans va-arg packs or
36061         va_arg_pack_len when the inlined call stmt requires pack
36062         expansion itself.
36063         * tree-inline.h (struct copy_body_data): Make call_stmt a gcall *.
36065 2017-03-21  Jakub Jelinek  <jakub@redhat.com>
36067         PR sanitizer/78158
36068         * tsan.c (instrument_builtin_call): If the memory model argument
36069         is not a constant, assume it is valid.
36071         PR c/67338
36072         * fold-const.c (round_up_loc): Negate divisor in unsigned type to
36073         avoid UB.
36075 2017-03-20  Segher Boessenkool  <segher@kernel.crashing.org>
36077         PR rtl-optimization/79910
36078         * combine.c (can_combine_p): Do not allow combining an I0 or I1
36079         if its dest is used by an insn before I2 (other than the combined
36080         insns themselves, which are properly handled already).
36082 2017-03-20  Segher Boessenkool  <segher@kernel.crashing.org>
36084         Revert:
36085         2017-03-17  Bernd Schmidt  <bschmidt@redhat.com>
36087         * combine.c (record_used_regs): New static function.
36088         (try_combine): Handle situations where there is an additional
36089         instruction between I2 and I3 which needs to have a LOG_LINK
36090         updated.
36092         Revert:
36093         2017-03-17  Jim Wilson  <jim.wilson@linaro.org>
36095         * combine.c (try_combine): Delete redundant i1 test.  Call
36096         prev_nonnote_nondebug_insn instead of prev_nonnote_insn.
36098 2017-03-20  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
36100         PR target/80083
36101         * config/rs6000/rs6000.md (*movsi_internal1): Fix constraints for
36102         alternatives 13/14.
36104 2017-03-20  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
36106         PR tree-optimization/80054
36107         * gimple-ssa-strength-reduction.c (all_phi_incrs_profitable): Fail
36108         the optimization if a PHI or any of its arguments is not dominated
36109         by the candidate's basis.  Use gphi* rather than gimple* as
36110         appropriate.
36111         (replace_profitable_candidates): Clean up a gimple* variable that
36112         should be a gphi* variable.
36114 2017-03-20  Martin Sebor  <msebor@redhat.com>
36116         PR c++/52477
36117         * doc/extend.texi (attribute constructor): Document present limitation.
36119 2017-03-20  Kelvin Nilsen  <kelvin@gcc.gnu.org>
36121         PR target/79963
36122         * config/rs6000/altivec.h (vec_all_ne): Under __cplusplus__ and
36123         __POWER9_VECTOR__ #ifdef control, change template definition to
36124         use Power9-specific built-in function.
36125         (vec_any_eq): Likewise.
36126         * config/rs6000/vector.md (vector_ae_v2di_p): Change the flag used
36127         to control outcomes from this test.
36128         (vector_ae_<mode>p): For VEC_F modes, likewise.
36130 2017-03-20  Ian Lance Taylor  <iant@google.com>
36132         * config/i386/i386.c (ix86_function_regparm): Save an extra
36133         register for -fsplit-stack with DECL_STATIC_CHAIN.
36135 2017-03-17  Palmer Dabbelt  <palmer@dabbelt.com>
36137         PR target/79912
36138         * config/riscv/riscv.c (riscv_preferred_reload_class): Remove.
36139         (TARGET_PREFERRED_RELOAD_CLASS): Likewise.
36141 2017-03-17  Palmer Dabbelt  <palmer@dabbelt.com>
36143         * config/riscv/riscv.c (riscv_print_operand): Use "fence
36144         iorw,ow".
36145         * config/riscv/sync.mc (mem_thread_fence_1): Use "fence
36146         iorw,iorw".
36148 2017-03-20  Marek Polacek  <polacek@redhat.com>
36150         PR sanitizer/80063
36151         * asan.c (DEF_SANITIZER_BUILTIN): Use do { } while (0).
36153 2017-03-20  Richard Biener  <rguenther@suse.de>
36155         PR tree-optimization/80113
36156         * graphite-isl-ast-to-gimple.c (copy_loop_phi_nodes): Do not
36157         allocate extra SSA name for PHI def.
36158         (add_close_phis_to_outer_loops): Likewise.
36159         (add_close_phis_to_merge_points): Likewise.
36160         (copy_loop_close_phi_args): Likewise.
36161         (copy_cond_phi_nodes): Likewise.
36163 2017-03-20  Martin Liska  <mliska@suse.cz>
36165         PR middle-end/79753
36166         * tree-chkp.c (chkp_build_returned_bound): Do not build
36167         returned bounds for a LHS that's not a BOUNDED_P type.
36169 2017-03-20  Martin Liska  <mliska@suse.cz>
36171         PR target/79769
36172         PR target/79770
36173         * tree-chkp.c (chkp_find_bounds_1): Handle REAL_CST,
36174         COMPLEX_CST and VECTOR_CST.
36176 2017-03-20  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
36178         PR target/78857
36179         * config/s390/s390.md ("cmp<mode>_ccs_0"): Add a clobber of the
36180         target operand.  A new splitter adds the clobber statement in case
36181         the target operand is dead anyway.
36183 2017-03-19  Gerald Pfeifer  <gerald@pfeifer.com>
36185         * doc/install.texi (Specific) <sparc-*-linux*>: No longer refer
36186         to age-old versions of binutils and glibc.
36188 2017-03-18  Segher Boessenkool  <segher@kernel.crashing.org>
36190         * doc/contrib.texi (Contributors): Remove duplicate entry for myself.
36192 2017-03-18  Gerald Pfeifer  <gerald@pfeifer.com>
36194         * doc/contrib.texi (Contributors): Add Segher Boessenkool.
36196 2017-03-18  Gerald Pfeifer  <gerald@pfeifer.com>
36198         * doc/install.texi (Specific) <arm-*-eabi>: Remove old
36199         requirement for binutils 2.13.
36201 2017-03-17  Jim Wilson  <jim.wilson@linaro.org>
36203         * combine.c (try_combine): Delete redundant i1 test.  Call
36204         prev_nonnote_nondebug_insn instead of prev_nonnote_insn.
36206 2017-03-17  Palmer Dabbelt  <palmer@dabbelt.com
36208         * doc/install.texi (Specific) <riscv32-*-elf>: Add riscv32-*-elf,
36209         riscv32-*-linux, riscv64-*-elf, riscv64-*-linux to the table of
36210         contents.
36211         <riscv64-*-elf>: Re-arrange section
36212         <riscv32-*-elf>: Add a note about requiring binutils 2.28.
36213         <riscv32-*-linux>: Likewise.
36214         <riscv64-*-elf>: Likewise
36215         <riscv64-*-linux>: Likewise.
36217 2017-03-17  Richard Earnshaw  <rearnsha@arm.com>
36219         PR target/80052
36220         * aarch64.opt(verbose-cost-dump): Fix typo.
36222 2017-03-17  Pat Haugen  <pthaugen@us.ibm.com>
36224         PR target/79951
36225         * config/rs6000/rs6000.md (copysign<mode>3_fcpsgn): Test
36226         for VECTOR_UNIT_VSX_P (<MODE>mode) too.
36228 2017-03-17  Bernd Schmidt  <bschmidt@redhat.com>
36230         * reload.c (find_reloads): When reloading a nonoffsettable address,
36231         use RELOAD_OTHER for it and its address reloads.
36233         PR rtl-optimization/79910
36234         * combine.c (record_used_regs): New static function.
36235         (try_combine): Handle situations where there is an additional
36236         instruction between I2 and I3 which needs to have a LOG_LINK
36237         updated.
36239 2017-03-17  Jeff Law  <law@redhat.com>
36241         PR tree-optimization/71437
36242         * tree-vrp.c (simplify_stmt_for_jump_threading): Lookup the
36243         conditional in the hash table first.
36244         (vrp_dom_walker::before_dom_children): Extract condition from
36245         ASSERT_EXPR.  Record condition, its inverion and any implied
36246         conditions as well.
36248 2017-03-17  Marek Polacek  <polacek@redhat.com>
36249             Markus Trippelsdorf  <markus@trippelsdorf.de>
36251         PR tree-optimization/80079
36252         * gimple-ssa-store-merging.c (class pass_store_merging): Initialize
36253         m_stores_head.
36255 2017-03-17  Richard Biener  <rguenther@suse.de>
36257         PR middle-end/80075
36258         * tree-eh.c (stmt_could_throw_1_p): Only handle gimple assigns.
36259         Properly verify the LHS before the RHS possibly claims to be
36260         handled.
36261         (stmt_could_throw_p): Hande gimple conds fully here.  Clobbers
36262         do not throw.
36264 2017-03-17  Martin Jambor  <mjambor@suse.cz>
36266         * doc/invoke.texi (Option Options): Include -fipa-vrp in the list.
36267         (List of -O2 options): Likewise.
36268         (-fipa-bit-cp): Replace "ipa" with "interprocedural."
36269         (-fipa-vrp) New.
36271 2017-03-17  Tom de Vries  <tom@codesourcery.com>
36273         * gcov-dump.c (print_usage): Print bug_report_url.
36275 2017-03-17  Richard Biener  <rguenther@suse.de>
36277         PR middle-end/80050
36278         * genmatch.c (parser::next): Remove pointless check for CPP_EOF.
36279         (parser::peek): Likewise.
36281 2017-03-17  Richard Biener  <rguenther@suse.de>
36283         PR tree-optimization/80048
36284         * sese.c (free_sese_info): Properly release rename_map and
36285         copied_bb_map elements.
36287 2017-03-16  Alexandre Oliva  <aoliva@redhat.com>
36289         * gimple-ssa-store-merging.c (struct imm_store_chain_info):
36290         Add linked-list forward and backlinks.  Insert on
36291         construction, remove on destruction.
36292         (class pass_store_merging): Add m_stores_head field.
36293         (pass_store_merging::terminate_and_process_all_chains):
36294         Iterate over m_stores_head list.
36295         (pass_store_merging::terminate_all_aliasing_chains):
36296         Likewise.
36297         (pass_store_merging::execute): Check for debug stmts first.
36298         Push new chains onto the m_stores_head stack.
36300 2017-03-16  Michael Meissner  <meissner@linux.vnet.ibm.com>
36302         PR target/71294
36303         * config/rs6000/vsx.md (vsx_splat_<mode>, VSX_D iterator): Allow a
36304         SPLAT operation on ISA 2.07 64-bit systems that have direct move,
36305         but no MTVSRDD support, by doing MTVSRD and XXPERMDI.
36307 2017-03-16  Jeff Law  <law@redhat.com>
36309         PR tree-optimization/71437
36310         * tree-ssa-dom.c (dom_opt_dom_walker): Remove thread_across_edge
36311         member function.  Implementation moved into after_dom_children
36312         member function and into the threader's thread_outgoing_edges
36313         function.
36314         (dom_opt_dom_walker::after_dom_children): Simplify by moving
36315         some code into new thread_outgoing_edges.
36316         * tree-ssa-threadedge.c (thread_across_edge): Make static and simplify
36317         definition.  Simplify marker handling (do it here).   Assume we always
36318         have the available expression and the const/copies tables.
36319         (thread_outgoing_edges): New function extracted from tree-ssa-dom.c
36320         and tree-vrp.c
36321         * tree-ssa-threadedge.h (thread_outgoing_edges): Declare.
36322         * tree-vrp.c (equiv_stack): No longer file scoped.
36323         (vrp_dom_walker): New class.
36324         (vrp_dom_walker::before_dom_children): New member function.
36325         (vrp_dom_walker::after_dom_children): Likewise.
36326         (identify_jump_threads):  Setup domwalker.  Use it rather than
36327         walking edges in a random order by hand.  Simplify setup/finalization.
36328         (finalize_jump_threads): Remove.
36329         (vrp_finalize): Do not call identify_jump_threads here.
36330         (execute_vrp): Do it here instead and call thread_through_all_blocks
36331         here too.
36333         PR tree-optimization/71437
36334         * tree-ssa-dom.c (pfn_simplify): Add basic_block argument.  All
36335         callers changed.
36336         (simplify_stmt_for_jump_threading): Add basic_block argument.  All
36337         callers changed.
36338         (lhs_of_dominating_assert): Moved from here into tree-vrp.c.
36339         (dom_opt_dom_walker::thread_across_edge): Remove
36340         handle_dominating_asserts argument.  All callers changed.
36341         (record_temporary_equivalences_from_stmts_at_dest): Corresponding
36342         changes.  Remove calls to lhs_of_dominating_assert.  Other
36343         uses of handle_dominating_asserts turn into unconditional code
36344         (simplify_control_stmt_condition_1): Likewise.
36345         (simplify_control_stmt_condition): Likewise.
36346         (thread_through_normal_block, thread_across_edge): Likewise.
36347         * tree-ssa-threadedge.h (thread_across_edge): Corresponding changes.
36348         * tree-vrp.c (lhs_of_dominating_assert): Move here.  Return original
36349         object if it is not an SSA_NAME.
36350         (simplify_stmt_for_jump_threading): Call lhs_of_dominating_assert
36351         before calling into the VRP specific simplifiers.
36352         (identify_jump_threads): Remove handle_dominating_asserts
36353         argument.
36355 2017-03-16  Jakub Jelinek  <jakub@redhat.com>
36357         PR fortran/79886
36358         * tree-diagnostic.c (default_tree_printer): No longer static.
36359         * tree-diagnostic.h (default_tree_printer): New prototype.
36361 2017-03-16  Tamar Christina  <tamar.christina@arm.com>
36363         * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>)
36364         Change ins into fmov.
36366 2017-03-16  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
36368         * config/aarch64/iterators.md (h_con): Return "x" for V4HF and V8HF.
36369         * config/aarch64/aarch64-simd.md (*aarch64_fma4_elt_from_dup<mode>):
36370         Use h_con constraint for operand 1.
36371         (*aarch64_fnma4_elt_from_dup<mode>): Likewise.
36372         (*aarch64_mulx_elt_from_dup<mode>): Likewise for operand 2.
36374 2017-03-15  Jeff Law  <law@redhat.com>
36376         PR tree-optimization/71437
36377         * tree-ssa-dom.c (derive_equivalences_from_bit_ior): New function.
36378         (record_temporary_equivalences): Use it.
36380         PR tree-optimization/71437
36381         * tree-ssa-dom.c (struct cond_equivalence): Moved from here into
36382         tree-ssa-scopedtables.
36383         (lookup_avail_expr, build_and_record_new_cond): Likewise.
36384         (record_conditions, record_cond, vuse_eq): Likewise.
36385         (record_edge_info): Adjust to API tweak of record_conditions.
36386         (simplify_stmt_for_jump_threading): Similarly for lookup_avail_expr.
36387         (record_temporary_equivalences, optimize_stmt): Likewise.
36388         (eliminate_redundant_computations): Likewise.
36389         (record_equivalences_from_stmt): Likewise.
36390         * tree-ssa-scopedtables.c: Include options.h and params.h.
36391         (vuse_eq): New function, moved from tree-ssa-dom.c
36392         (build_and_record_new_cond): Likewise.
36393         (record_conditions): Likewise.  Accept vector of conditions rather
36394         than edge_equivalence structure for first argument.
36395         for the first argument.
36396         (avail_exprs_stack::lookup_avail_expr): New member function, moved
36397         from tree-ssa-dom.c.
36398         (avail_exprs_stack::record_cond): Likewise.
36399         * tree-ssa-scopedtables.h (struct cond_equivalence): Moved here
36400         from tree-ssa-dom.c.
36401         (avail_exprs_stack): Add new member functions lookup_avail_expr
36402         and record_cond.
36403         (record_conditions): Declare.
36405 2017-03-15  Vladimir Makarov  <vmakarov@redhat.com>
36407         PR target/80017
36408         * lra-constraints.c (process_alt_operands): Increase reject for
36409         reloading an input/output operand.
36411 2017-03-15  Michael Meissner  <meissner@linux.vnet.ibm.com>
36413         PR target/79038
36414         * config/rs6000/rs6000.md (float<QHI:mode><IEEE128:mode>2): Define
36415         insns to convert from signed/unsigned char/short to IEEE 128-bit
36416         floating point.
36417         (floatuns<QHI:mode><IEEE128:mode>2): Likewise.
36419 2017-03-15  Uros Bizjak  <ubizjak@gmail.com>
36421         PR target/80019
36422         * config/i386/i386.c (ix86_vector_duplicate_value): Create
36423         subreg of inner mode for values already in registers.
36425 2017-03-15  Bernd Schmidt  <bschmidt@redhat.com>
36427         * config/c6x/c6x.c (hwloop_optimize): Handle case where the old
36428         iteration reg is used after the loop.
36430 2017-03-14  Martin Sebor  <msebor@redhat.com>
36432         PR tree-optimization/79800
36433         * gimple-ssa-sprintf.c (format_floating: Add argument.  Handle
36434         precision in negative-positive range.
36435         (format_floating): Call non-const overload with adjusted precision.
36437 2017-03-14  Michael Meissner  <meissner@linux.vnet.ibm.com>
36439         PR target/79947
36440         * config/rs6000/rs6000.h (TARGET_FRSQRTES): Add check for
36441         -mpowerpc-gfxopt.
36443 2017-03-14  Martin Sebor  <msebor@redhat.com>
36445         PR middle-end/80020
36446         * builtin-attrs.def (ATTR_ALLOC_SIZE_2_NOTHROW_LIST): New macro.
36447         * builtins.def (aligned_alloc): Use it.
36449         PR c/79936
36450         * Makefile.in (GTFILES): Add calls.c.
36451         * calls.c: Include "gt-calls.h".
36453 2017-03-14  Bernd Schmidt  <bschmidt@redhat.com>
36455         PR rtl-optimization/79728
36456         * regs.h (struct target_regs): New field
36457         x_contains_allocatable_regs_of_mode.
36458         (contains_allocatable_regs_of_mode): New macro.
36459         * reginfo.c (init_reg_sets_1): Initialize it, and change
36460         contains_reg_of_mode so it includes global regs as well.
36461         * reload.c (push_reload): Use contains_allocatable_regs_of_mode
36462         rather than contains_regs_of_mode.
36464 2017-03-14  Martin Liska  <mliska@suse.cz>
36466         * doc/invoke.texi: Document options that can't be combined with
36467         -fcheck-pointer-bounds.
36469 2017-03-14  Martin Liska  <mliska@suse.cz>
36471         PR middle-end/79831
36472         * doc/invoke.texi (-Wchkp): Document the option.
36474 2017-03-14  Martin Liska  <mliska@suse.cz>
36476         * Makefile.in: Install gcov-dump.
36478 2017-03-14  Martin Liska  <mliska@suse.cz>
36480         * multiple_target.c (expand_target_clones): Bail out for
36481         an invalid attribute.
36483 2017-03-14  Richard Biener  <rguenther@suse.de>
36485         * alias.c (struct alias_set_entry): Pack properly.
36486         * cfgloop.h (struct loop): Likewise.
36487         * cse.c (struct set): Likewise.
36488         * ipa-utils.c (struct searchc_env): Likewise.
36489         * loop-invariant.c (struct invariant): Likewise.
36490         * lra-remat.c (struct cand): Likewise.
36491         * recog.c (struct change_t): Likewise.
36492         * rtl.h (struct address_info): Likewise.
36493         * symbol-summary.h (function_summary): Likewise.
36494         * tree-loop-distribution.c (struct partition): Likewise.
36495         * tree-object-size.c (struct object_size_info): Likewise.
36496         * tree-ssa-loop-ivopts.c (struct cost_pair): Likewise.
36497         * tree-ssa-threadupdate.c (struct ssa_local_info_t): Likewise.
36498         * tree-vect-data-refs.c (struct _vect_peel_info): Likewise.
36499         * tree-vect-slp.c (struct _slp_oprnd_info): Likewise.
36500         * tree-vect-stmts.c (struct simd_call_arg_info): Likewise.
36501         * tree-vectorizer.h (struct _loop_vec_info): Likewise.
36502         (struct _stmt_vec_info): Likewise.
36504 2017-03-14  Martin Liska  <mliska@suse.cz>
36506         PR target/79892
36507         * multiple_target.c (create_dispatcher_calls): Check that
36508         a target can create a function dispatcher.
36510 2017-03-14  Martin Liska  <mliska@suse.cz>
36512         PR lto/66295
36513         * multiple_target.c (expand_target_clones): Drop local.local
36514         flag for default implementation.
36516 2017-03-14  Richard Biener  <rguenther@suse.de>
36518         PR tree-optimization/80030
36519         * tree-vect-stmts.c (vectorizable_store): Plug memleak.
36521 2017-03-13  Kito Cheng  <kito.cheng@gmail.com>
36523         * config/riscv/riscv.c (riscv_emit_float_compare>: Use
36524         gcc_fallthrough() instead of __attribute__((fallthrough));
36526 2017-03-13  Gerald Pfeifer  <gerald@pfeifer.com>
36528         * doc/gcc.texi: Remove "up" link to (DIR).
36529         * doc/gccint.texi: Ditto.
36531 2017-03-13  Gerald Pfeifer  <gerald@pfeifer.com>
36533         * doc/install.texi (Specific) <avr>: Remove reference to
36534         binutils 2.13.
36536 2017-03-13  Jeff Law  <law@redhat.com>
36538         * config/riscv/riscv.c (riscv_emit_float_compare): Use fallthru
36539         attribute rather than comments.
36541         * config/pdp11/pdp11.md (movmemhi): Adjust operand numbers to
36542         match_scratch operand is highest.
36544 2017-03-13  Martin Liska  <mliska@suse.cz>
36546         PR middle-end/78339
36547         * ipa-pure-const.c (warn_function_noreturn): If the declarations
36548         is a CHKP clone, use original declaration.
36550 2017-03-13  Claudiu Zissulescu  <claziss@synopsys.com>
36552         * config/arc/arc.c (arc_init): Use multiplier whenever we have it.
36553         (arc_conditional_register_usage): Use a different allocation order
36554         when optimizing for size.
36555         * common/config/arc/arc-common.c (arc_option_optimization_table):
36556         Section anchors default on when optimizing for size.
36558 2017-03-13  Claudiu Zissulescu  <claziss@synopsys.com>
36560         * config/arc/arc.md (*tst_bitfield_tst): Fix pattern.
36562 2017-03-13  Claudiu Zissulescu  <claziss@synopsys.com>
36564         * config/arc/arc.c (arc_output_addsi): Emit code density adds.
36565         * config/arc/arc.md (cpu_facility): Add cd variant.
36566         (*movqi_insn): Add code density variant.
36567         (*movhi_insn): Likewise.
36568         (*movqi_insn): Likewise.
36569         (*addsi3_mixed): Likewise.
36570         (subsi3_insn): Likewise.
36572 2017-03-13  Claudiu Zissulescu  <claziss@synopsys.com>
36574         * config/arc/arc.md (movsi_cond_exec): Update constraint.
36576 2017-03-13  Claudiu Zissulescu  <claziss@synopsys.com>
36578         * config/arc/arc.c (arc_legitimize_pic_address): Handle PIC
36579         expressions with MINUS and UNARY ops.
36581 2017-03-13  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
36583         PR target/79911
36584         * config/arm/neon.md (vec_sel_widen_ssum_lo<VQI:mode><VW:mode>3):
36585         Rename to...
36586         (vec_sel_widen_ssum_lo<mode><V_half>3): ... This. Avoid mismatch
36587         between vec_select and vector argument.
36588         (vec_sel_widen_ssum_hi<VQI:mode><VW:mode>3): Rename to...
36589         (vec_sel_widen_ssum_hi<mode><V_half>3): ... This. Likewise.
36590         (vec_sel_widen_usum_lo<VQI:mode><VW:mode>3): Rename to...
36591         (vec_sel_widen_usum_lo<mode><V_half>3): ... This.
36592         (vec_sel_widen_usum_hi<VQI:mode><VW:mode>3): Rename to...
36593         (vec_sel_widen_usum_hi<mode><V_half>3): ... This.
36595 2017-03-13  Richard Biener  <rguenther@suse.de>
36597         PR other/79991
36598         * params.def (vect-max-peeling-for-alignment): Fix typo.
36600 2017-03-12  Gerald Pfeifer  <gerald@pfeifer.com>
36602         * doc/install.texi (Specific) <mips-*-*>: Remove description of
36603         issue that only occurred with binutils below 2.18.
36605 2017-03-12  Gerald Pfeifer  <gerald@pfeifer.com>
36607         * doc/install.texi (Specific) <cris-axis-elf>: No longer
36608         refer to binutils 2.11/2.12 minimum.
36610 2017-03-12  Gerald Pfeifer  <gerald@pfeifer.com>
36612         * doc/install.texi (Specific) <powerpc-*-*>: Remove link to
36613         ftp.kernel.org and simplify binutils requirement.
36615 2017-03-11  Gerald Pfeifer  <gerald@pfeifer.com>
36617         * doc/invoke.texi (Warning Options): Fix spelling of link-time
36618         optimization.
36619         (Optimize Options): Ditto.  Also remove redundancy.
36621 2017-03-10  David Malcolm  <dmalcolm@redhat.com>
36623         PR translation/79848
36624         * ipa-devirt.c (warn_types_mismatch): Simplify uses of "%<%s%>" to
36625         "%qs".
36626         * ipa-pure-const.c (suggest_attribute): Likewise.  Convert _
36627         to G_ to avoid double translation.
36629 2017-03-10  David Malcolm  <dmalcolm@redhat.com>
36631         PR translation/79923
36632         * auto-profile.c (get_combined_location): Convert leading
36633         character of diagnostics to lower case and remove trailing period.
36634         (read_profile): Likewise for various diagnostics.
36635         * config/arm/arm.c (arm_option_override): Remove trailing period
36636         from various diagnostics.
36637         * config/msp430/msp430.c (msp430_expand_delay_cycles): Likewise.
36638         (msp430_expand_delay_cycles): Likewise.
36640 2017-03-10  David Malcolm  <dmalcolm@redhat.com>
36642         PR target/79925
36643         * config/aarch64/aarch64.c (aarch64_validate_mcpu): Quote the
36644         full command-line argument, rather than just "str".
36645         (aarch64_validate_march): Likewise.
36646         (aarch64_validate_mtune): Likewise.
36648 2017-03-10  Bernd Schmidt  <bschmidt@redhat.com>
36650         PR rtl-optimization/78911
36651         * lra-assigns.c (must_not_spill_p): New function.
36652         (spill_for): Use it.
36654 2017-03-10  Jakub Jelinek  <jakub@redhat.com>
36656         PR tree-optimization/79981
36657         * tree-vrp.c (extract_range_basic): Handle IMAGPART_EXPR of
36658         ATOMIC_COMPARE_EXCHANGE ifn result.
36659         (stmt_interesting_for_vrp, vrp_visit_stmt): Handle
36660         IFN_ATOMIC_COMPARE_EXCHANGE.
36662 2017-03-10  David Malcolm  <dmalcolm@redhat.com>
36664         PR driver/79875
36665         * opts.c (parse_sanitizer_options): Add missing question mark to
36666         "did you mean" message.
36668 2017-03-10  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
36670         * config/rs6000/rs6000-builtin.def (VMULEUB_UNS): Remove orphaned
36671         built-in.
36672         (VMULEUH_UNS): Likewise.
36673         (VMULOUB_UNS): Likewise.
36674         (VMULOUH_UNS): Likewise.
36675         * config/rs6000/rs6000.c (builtin_function_type): Remove
36676         references to ALTIVEC_BUILTIN_VMUL[EO]U[BH]_UNS.
36678 2017-03-10  David Malcolm  <dmalcolm@redhat.com>
36680         PR bootstrap/79952
36681         * read-rtl-function.c (function_reader::read_rtx_operand): Update
36682         x with result of extra_parsing_for_operand_code_0.
36683         (function_reader::extra_parsing_for_operand_code_0): Convert
36684         return type from void to rtx, returning x.  When reading
36685         SYMBOL_REF with SYMBOL_FLAG_HAS_BLOCK_INFO, reallocate x to the
36686         larger size containing struct block_symbol.
36688 2017-03-10  Segher Boessenkool  <segher@kernel.crashing.org>
36690         * config/rs6000/rs6000.c (rs6000_option_override_internal): Disallow
36691         -mfloat128-hardware without -m64.
36693 2017-03-10  Will Schmidt <will_schmidt@vnet.ibm.com>
36695         PR target/79941
36696         * config/rs6000/rs6000.c (builtin_function_type): Add VMUL*U[HB]
36697         entries to the case statement that marks unsigned arguments to
36698         overloaded functions.
36700 2017-03-10  Kelvin Nilsen  <kelvin@gcc.gnu.org>
36702         * config/rs6000/rs6000.c (rs6000_option_override_internal): Fix
36703         two typographic errors in the handling of TARGET_UPPER_REGS_DI.
36705 2017-03-10  Pat Haugen  <pthaugen@us.ibm.com>
36707         PR target/79907
36708         * config/rs6000/rs6000.c (rs6000_init_hard_regno_mode_ok): Test
36709         TARGET_UPPER_REGS_DI when setting 'wi' constraint regclass.
36711 2017-03-10  Martin Liska  <mliska@suse.cz>
36713         PR target/65705
36714         PR target/69804
36715         * toplev.c (process_options): Enable MPX with LSAN and UBSAN.
36716         * tree-chkp.c (chkp_walk_pointer_assignments): Verify that
36717         FIELD != NULL.
36719 2017-03-10  Olivier Hainque  <hainque@adacore.com>
36721         * tree-switch-conversion (array_value_type): Start by resetting
36722         candidate type to it's main variant.
36724 2017-03-10  Jakub Jelinek  <jakub@redhat.com>
36726         PR rtl-optimization/79909
36727         * combine.c (try_combine): Use simplify_replace_rtx on individual
36728         CALL_INSN_FUNCTION_USAGE elements instead of replace_rtx on copy_rtx
36729         of the whole CALL_INSN_FUNCTION_USAGE.
36731         PR tree-optimization/79972
36732         * gimple-ssa-warn-alloca.c (alloca_call_type): Only call
36733         get_range_info on SSA_NAMEs.  Formatting fixes.
36735 2017-03-10  Richard Biener  <rguenther@suse.de>
36736             Jakub Jelinek  <jakub@redhat.com>
36738         PR tree-optimization/77975
36739         * tree-ssa-loop-niter.c (get_base_for): Allow phi argument from latch
36740         edge to be constant.
36741         (get_val_for): For constant x return it.  Formatting fix.
36742         (loop_niter_by_eval): Avoid pointless looping if the next iteration
36743         would use the same bases as the current one.
36745 2017-03-09  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
36747         * config/rs6000/rs6000.c (rs6000_gen_le_vsx_permute): Use rotate
36748         instead of vec_select for V1TImode.
36749         * conifg/rs6000/vsx.md (VSX_LE): Remove mode iterator that is no
36750         longer needed.
36751         (VSX_LE_128): Add V1TI to this mode iterator.
36752         (*vsx_le_perm_load_<mode>): Change to use VSX_D mode iterator.
36753         (*vsx_le_perm_store_<mode>): Likewise.
36754         (pre-reload splitter for VSX stores): Likewise.
36755         (post-reload splitter for VSX stores): Likewise.
36756         (*vsx_xxpermdi2_le_<mode>): Likewise.
36757         (*vsx_lxvd2x2_le_<mode>): Likewise.
36758         (*vsx_stxvd2x2_le_<mode>): Likewise.
36760 2017-03-09  Michael Eager  <eager@eagercon.com>
36762         Correct failures with --enable-checking=yes,rtl.
36764         * config/microblaze/microblaze.c (microblaze_expand_shift):
36765         Replace GET_CODE test with CONST_INT_P and INTVAL test with
36766         test for const0_rtx.
36767         * config/microblaze/microblaze.md (ashlsi3_byone, ashrsi3_byone,
36768         lshrsi3_byone): Replace INTVAL with test for const1_rtx.
36770 2017-03-09  Richard Biener  <rguenther@suse.de>
36772         PR tree-optimization/79977
36773         * graphite-scop-detection.c (scop_detection::merge_sese):
36774         Handle the case of extra exits to blocks dominating the entry.
36776 2017-03-09  Toma Tabacu  <toma.tabacu@imgtec.com>
36778         * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
36779         Document rdynamic.
36781 2017-03-09  Vladimir Makarov  <vmakarov@redhat.com>
36783         PR rtl-optimization/79949
36784         * lra-constraints.c (process_alt_operands): Check memory when
36785         trying to predict a cycle.  Print about the overall increase.
36787 2017-03-09  Richard Biener  <rguenther@suse.de>
36789         PR middle-end/79971
36790         * gimple-expr.c (useless_type_conversion_p): Preserve
36791         TYPE_SATURATING for fixed-point types.
36793 2017-03-09  Richard Biener  <rguenther@suse.de>
36795         PR ipa/79970
36796         * ipa-prop.c (ipa_modify_formal_parameters): Avoid changing
36797         alignment of BLKmode params.
36799 2017-03-09  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
36801         PR target/79913
36802         * config/aarch64/iterators.md (VALL_F16_NO_V2Q): New mode iterator.
36803         (VALL_NO_V2Q): Likewise.
36804         (VDQF_DF): Delete.
36805         * config/aarch64/aarch64-simd.md
36806         (aarch64_dup_lane_<vswap_width_name><mode>): Use VALL_F16_NO_V2Q
36807         iterator.
36808         (*aarch64_simd_vec_copy_lane_<vswap_width_name><mode>): Use
36809         VALL_NO_V2Q mode iterator.
36810         (*aarch64_vgetfmulx<mode>): Use VDQF iterator.
36812 2017-03-09  Martin Liska  <mliska@suse.cz>
36814         PR tree-optimization/79631
36815         * tree-chkp-opt.c (chkp_is_constant_addr): Call
36816         tree_int_cst_sign_bit just for INTEGER constants.
36818 2017-03-09  Martin Liska  <mliska@suse.cz>
36820         PR target/65705
36821         PR target/69804
36822         * toplev.c (process_options): Disable -fcheck-pointer-bounds with
36823         sanitizers.
36825 2017-03-09  Marek Polacek  <polacek@redhat.com>
36827         PR c++/79672
36828         * tree.c (inchash::add_expr): Handle TREE_VEC.
36830 2017-03-09  Martin Liska  <mliska@suse.cz>
36832         PR ipa/79764
36833         (chkp_narrow_size_and_offset): New function.
36834         (chkp_parse_array_and_component_ref): Support BIT_FIELD_REF.
36835         (void chkp_parse_bit_field_ref): New function.
36836         (chkp_make_addressed_object_bounds): Add case for BIT_FIELD_REF.
36837         (chkp_process_stmt): Use chkp_parse_bit_field_ref.
36839 2017-03-09  Martin Liska  <mliska@suse.cz>
36841         PR ipa/79761
36842         * tree-chkp.c (chkp_get_bound_for_parm): Get bounds for a param.
36843         (chkp_find_bounds_1): Remove gcc_unreachable.
36845 2017-03-09  Jakub Jelinek  <jakub@redhat.com>
36847         PR sanitizer/79944
36848         * asan.c (get_mem_refs_of_builtin_call): For BUILT_IN_ATOMIC* and
36849         BUILT_IN_SYNC*, determine the access type from the size suffix and
36850         always build a MEM_REF with that type.  Handle forgotten
36851         BUILT_IN_SYNC_FETCH_AND_NAND_16 and BUILT_IN_SYNC_NAND_AND_FETCH_16.
36853         PR target/79932
36854         * config/i386/avx512vlintrin.h (_mm256_cmpge_epi32_mask,
36855         _mm256_cmpge_epi64_mask, _mm256_cmpge_epu32_mask,
36856         _mm256_cmpge_epu64_mask, _mm256_cmple_epi32_mask,
36857         _mm256_cmple_epi64_mask, _mm256_cmple_epu32_mask,
36858         _mm256_cmple_epu64_mask, _mm256_cmplt_epi32_mask,
36859         _mm256_cmplt_epi64_mask, _mm256_cmplt_epu32_mask,
36860         _mm256_cmplt_epu64_mask, _mm256_cmpneq_epi32_mask,
36861         _mm256_cmpneq_epi64_mask, _mm256_cmpneq_epu32_mask,
36862         _mm256_cmpneq_epu64_mask, _mm256_mask_cmpge_epi32_mask,
36863         _mm256_mask_cmpge_epi64_mask, _mm256_mask_cmpge_epu32_mask,
36864         _mm256_mask_cmpge_epu64_mask, _mm256_mask_cmple_epi32_mask,
36865         _mm256_mask_cmple_epi64_mask, _mm256_mask_cmple_epu32_mask,
36866         _mm256_mask_cmple_epu64_mask, _mm256_mask_cmplt_epi32_mask,
36867         _mm256_mask_cmplt_epi64_mask, _mm256_mask_cmplt_epu32_mask,
36868         _mm256_mask_cmplt_epu64_mask, _mm256_mask_cmpneq_epi32_mask,
36869         _mm256_mask_cmpneq_epi64_mask, _mm256_mask_cmpneq_epu32_mask,
36870         _mm256_mask_cmpneq_epu64_mask, _mm_cmpge_epi32_mask,
36871         _mm_cmpge_epi64_mask, _mm_cmpge_epu32_mask, _mm_cmpge_epu64_mask,
36872         _mm_cmple_epi32_mask, _mm_cmple_epi64_mask, _mm_cmple_epu32_mask,
36873         _mm_cmple_epu64_mask, _mm_cmplt_epi32_mask, _mm_cmplt_epi64_mask,
36874         _mm_cmplt_epu32_mask, _mm_cmplt_epu64_mask, _mm_cmpneq_epi32_mask,
36875         _mm_cmpneq_epi64_mask, _mm_cmpneq_epu32_mask, _mm_cmpneq_epu64_mask,
36876         _mm_mask_cmpge_epi32_mask, _mm_mask_cmpge_epi64_mask,
36877         _mm_mask_cmpge_epu32_mask, _mm_mask_cmpge_epu64_mask,
36878         _mm_mask_cmple_epi32_mask, _mm_mask_cmple_epi64_mask,
36879         _mm_mask_cmple_epu32_mask, _mm_mask_cmple_epu64_mask,
36880         _mm_mask_cmplt_epi32_mask, _mm_mask_cmplt_epi64_mask,
36881         _mm_mask_cmplt_epu32_mask, _mm_mask_cmplt_epu64_mask,
36882         _mm_mask_cmpneq_epi32_mask, _mm_mask_cmpneq_epi64_mask,
36883         _mm_mask_cmpneq_epu32_mask, _mm_mask_cmpneq_epu64_mask): Move
36884         definitions outside of __OPTIMIZE__ guarded section.
36886         PR target/79932
36887         * config/i386/avx512bwintrin.h (_mm512_packs_epi32,
36888         _mm512_maskz_packs_epi32, _mm512_mask_packs_epi32,
36889         _mm512_packus_epi32, _mm512_maskz_packus_epi32,
36890         _mm512_mask_packus_epi32): Move definitions outside of __OPTIMIZE__
36891         guarded section.
36893 2017-03-09  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
36895         * config/s390/vx-builtins.md ("vfee<mode>", "vfeez<mode>")
36896         ("vfenez<mode>"): Add missing constraints.
36898 2017-03-08  Martin Sebor  <msebor@redhat.com>
36900         PR target/79928
36901         * config/nds32/nds32.c (nds32_option_override):
36902         Fix misspelled diagnostic.
36904 2017-03-08  Jakub Jelinek  <jakub@redhat.com>
36906         PR c/79940
36907         * gimplify.c (gimplify_omp_for): Replace index var in outer
36908         taskloop statement with an artificial variable and add
36909         OMP_CLAUSE_PRIVATE clause for it.
36911 2017-03-08  Richard Biener  <rguenther@suse.de>
36913         PR tree-optimization/79955
36914         * tree-ssa-uninit.c (warn_uninitialized_vars): Do not warn
36915         for accesses that are completely outside of the variable.
36917 2017-03-08  Andrew Haley  <aph@redhat.com>
36919         PR tree-optimization/79943
36920         * tree-ssa-loop-split.c (compute_new_first_bound): When
36921         calculating the new upper bound, (END-BEG) should be added, not
36922         subtracted.
36924 2017-03-08  Jakub Jelinek  <jakub@redhat.com>
36926         * config/avr/avr.md (setmemhi): Make sure match_dup
36927         operand number comes before match_scratch.
36929 2017-03-08  Richard Biener  <rguenther@suse.de>
36931         PR tree-optimization/79920
36932         * tree-vect-slp.c (vect_create_mask_and_perm): Remove and inline
36933         with ncopies == 1 to ...
36934         (vect_transform_slp_perm_load): ... here.  Properly compute
36935         all element loads by iterating VF times over the group.  Do
36936         not handle ncopies (computed in a broken way) in
36937         vect_create_mask_and_perm.
36939 2017-03-08  Jakub Jelinek  <jakub@redhat.com>
36941         PR sanitizer/79904
36942         * internal-fn.c (expand_vector_ubsan_overflow): If arg0 or arg1
36943         is a uniform vector, use uniform_vector_p return value instead of
36944         building ARRAY_REF on folded VIEW_CONVERT_EXPR to array type.
36946 2017-03-07  Marek Polacek  <polacek@redhat.com>
36948         PR middle-end/79809
36949         * gimple-ssa-warn-alloca.c (pass_walloca::gate): Use HOST_WIDE_INT.
36950         (alloca_call_type): Likewise.
36952 2017-03-07  Martin Liska  <mliska@suse.cz>
36954         * gcov.c (process_args): Put comment to correct location.
36956 2017-03-07  Martin Liska  <mliska@suse.cz>
36958         PR middle-end/68270
36959         * tree-chkp.c (chkp_may_narrow_to_field): Add new argument ref.
36960         Use array_at_struct_end_p instead of DECL_CHAIN (field).
36961         (chkp_narrow_bounds_for_field): Likewise.
36962         (chkp_parse_array_and_component_ref): Pass one more argument to
36963         call.
36965 2017-03-07  Richard Biener  <rguenther@suse.de>
36967         * tree-vect-loop-manip.c (slpeel_add_loop_guard): Preserve
36968         preheaders.
36970 2017-03-07  Segher Boessenkool  <segher@kernel.crashing.org>
36972         * config/i386/i386.c (ix86_local_alignment): Align most aggregates
36973         of 16 bytes and more to 16 bytes, not those of 16 bits and more.
36975 2017-03-07  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
36977         PR c/79855
36978         * params.def (PARAM_STORE_MERGING_ALLOW_UNALIGNED): Add full stop
36979         to end of description.
36980         (PARAM_MAX_STORES_TO_MERGE): Likewise.
36982 2017-03-07  Jakub Jelinek  <jakub@redhat.com>
36984         PR rtl-optimization/79901
36985         * config/i386/sse.md (*avx512bw_<code><mode>3<mask_name>): Renamed to
36986         ...
36987         (*avx512f_<code><mode>3<mask_name>): ... this.
36988         (<code><mode>3 with maxmin code iterator): Use VI8_AVX2_AVX512F
36989         iterator instead of VI8_AVX2_AVX512BW.
36991         PR rtl-optimization/79901
36992         * expr.c (expand_expr_real_2): For vector MIN/MAX, if there is no
36993         min/max expander, expand it using expand_vec_cond_expr.
36995         PR sanitizer/79897
36996         * ubsan.c (ubsan_encode_value): Call mark_addressable on the
36997         temporary.
36999 2017-03-06  Jakub Jelinek  <jakub@redhat.com>
37001         PR c++/79821
37002         * dwarf2out.h (dw_vec_const): Change array type from unsigned char *
37003         to void * for PCH reasons.
37004         * dwarf2out.c (output_loc_operands, output_die): Cast
37005         v.val_vec.array to unsigned char *.
37007 2017-03-06  John David Anglin  <danglin@gcc.gnu.org>
37009         PR target/77850
37010         * config/pa/pa-64.h (PAD_VARARGS_DOWN): Don't pad down complex and
37011         vector types.
37013 2017-03-06  Vladimir Makarov  <vmakarov@redhat.com>
37015         PR rtl-optimization/79571
37016         * lra-constraints.c (process_alt_operands): Calculate static
37017         reject and subtract it from overall when only addresses will be
37018         reloaded.
37020 2017-03-06  Julia Koval  <julia.koval@intel.com>
37022         PR target/79793
37023         * config/i386/i386.c (ix86_minimum_incoming_stack_boundary): Set
37024         incoming stack boundary to 128 for 64-bit targets.
37026 2017-03-06  Richard Biener  <rguenther@suse.de>
37028         PR tree-optimization/79894
37029         * tree-vectorizer.c (vectorize_loops): Set loop_vectorized_call
37030         to NULL after folding it.
37032 2017-03-06  Richard Biener  <rguenther@suse.de>
37034         PR tree-optimization/79824
37035         * tree-vect-stmts.c (get_group_load_store_type): Fix alignment
37036         check disabling peeling for gaps.
37038 2017-03-06  Toma Tabacu  <toma.tabacu@imgtec.com>
37040         * doc/sourcebuild.texi (Effective-Target Keywords, Environment
37041         attributes): Document gettimeofday.
37043 2017-03-06  Robin Dapp  <rdapp@linux.vnet.ibm.com>
37045         * config/s390/s390.c (s390_option_override_internal): Set
37046         PARAM_MIN_VECT_LOOP_BOUND
37048 2017-03-06  Robin Dapp  <rdapp@linux.vnet.ibm.com>
37050         * config/s390/s390.c (s390_asm_output_function_label): Use nopr %r0.
37051         * config/s390/s390.md: Likewise.
37053 2017-03-06  Jakub Jelinek  <jakub@redhat.com>
37055         PR target/79812
37056         * config/i386/sse.md (VI8F_256_512): Remove mode iterator.
37057         (<avx2_avx512>_perm<mode>): Rename to ...
37058         (avx2_perm<mode>): ... this.  Use VI8F_256 iterator instead
37059         of VI8F_256_512.
37060         (<avx512>_perm<mode>_mask): Rename to ...
37061         (avx512vl_perm<mode>_mask): ... this.  Use VI8F_256 iterator instead
37062         of VI8F_256_512.
37063         (<avx2_avx512>_perm<mode>_1<mask_name>): Rename to ...
37064         (avx2_perm<mode>_1<mask_name): ... this.  Use VI8F_256 iterator
37065         instead of VI8F_256_512.
37066         (avx512f_perm<mode>): New define_expand.
37067         (avx512f_perm<mode>_mask): Likewise.
37068         (avx512f_perm<mode>_1<mask_name>): New define_insn.
37069         (<avx512>_vec_dup<mode>_1): Fix up vec_select mode.
37071 2017-03-06  Prachi Godbole  <prachi.godbole@imgtec.com>
37073         * config/mips/mips-msa.md (msa_fmax_a_<msafmt>, msa_fmin_a_<msafmt>,
37074         msa_max_a_<msafmt>, msa_min_a_<msafmt>): Introduce mode interator for
37075         if_then_else.
37076         (smin<mode>3, smax<mode>3): Change operand print code from 'B' to 'E'.
37078 2017-03-06  Martin Liska  <mliska@suse.cz>
37080         PR sanitize/79783
37081         * asan.c (asan_expand_poison_ifn): Do not expand ASAN_POISON
37082         when having a SSA NAME w/o VAR_DECL assigned to it.
37084 2017-03-06  Prachi Godbole  <prachi.godbole@imgtec.com>
37086         * config/mips/mips-msa.md (msa_dotp_<su>_d, msa_dpadd_<su>_d,
37087         msa_dpsub_<su>_d): Fix MODE for vec_select.
37089 2017-03-06  Prachi Godbole  <prachi.godbole@imgtec.com>
37091         * config/mips/mips.c (mips_gen_const_int_vector): Change type of last
37092         argument.
37093         * config/mips/mips-protos.h (mips_gen_const_int_vector): Likewise.
37095 2017-03-06  Richard Biener  <rguenther@suse.de>
37097         * lto-streamer.c (lto_check_version): Use %qs in diagnostics.
37098         * plugin.c (register_plugin_info): Likewise.
37099         * tree-chkp.c (chkp_make_static_const_bounds): Likewise.
37101 2017-03-05  Jakub Jelinek  <jakub@redhat.com>
37103         * config/i386/sse.md (sse_storehps, sse_storelps,
37104         avx_<castmode><avxsizesuffix>_<castmode>,
37105         avx512f_<castmode><avxsizesuffix>_<castmode>,
37106         avx512f_<castmode><avxsizesuffix>_256<castmode>): Require
37107         in condition that at least one operand is not a MEM.
37109 2017-03-03  Jakub Jelinek  <jakub@redhat.com>
37111         PR middle-end/79805
37112         * internal-fn.def (ATOMIC_BIT_TEST_AND_SET, ATOMIC_BIT_TEST_AND_RESET,
37113         ATOMIC_BIT_TEST_AND_COMPLEMENT, ATOMIC_COMPARE_EXCHANGE): Remove
37114         ECF_NOTHROW.
37115         * gimple-fold.c (fold_builtin_atomic_compare_exchange): Set
37116         gimple_call_nothrow_p flag based on whether original builtin can throw.
37117         If it can, emit following stmts on the fallthrough edge.
37118         * tree-ssa-ccp.c (optimize_atomic_bit_test_and): Similarly, except
37119         don't create new bb if inserting just debug stmts on the edge, try to
37120         insert them on the fallthru bb or just reset debug stmts.
37122 2017-03-03  Segher Boesssenkool  <segher@kernel.crashing.org>
37124         PR target/43763
37125         * config/rs6000/rs6000.c (rs6000_final_prescan_insn): Save and
37126         restore recog_data (including the operand rtxes inside it) around
37127         the call to get_insn_template.
37129 2017-03-03  Martin Sebor  <msebor@redhat.com>
37131         PR tree-optimization/79699
37132         * context.c (context::~context): Free MPFR caches to avoid
37133         a memory leak on program exit.
37135 2017-03-03  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
37137         * config/aarch64/aarch64.c (aarch64_float_const_representable_p):
37138         Use wide_int::ulow () instead of .elt (0).
37140 2017-03-03  Uros Bizjak  <ubizjak@gmail.com>
37142         * config/i386/i386.md (*pushtf): Change *roF constraint to *roC.
37143         (*pushxf): Limit oF constraint to 32bit targets and add oC
37144         constraint for 64bit targets.
37145         (pushxf splitter): Use PUSH_ROUNDING to calculate stack adjustment.
37146         (*pushdf): Change rmF constraint to rmC.
37148 2017-03-03  Martin Liska  <mliska@suse.cz>
37150         * tree-ssa-loop-prefetch.c (pass_loop_prefetch::execute):
37151         Remove unused variable.
37153 2017-03-03  Jakub Jelinek  <jakub@redhat.com>
37155         PR target/79807
37156         * config/i386/i386.c (ix86_expand_multi_arg_builtin): If target
37157         is a memory operand, increase num_memory.
37158         (ix86_expand_args_builtin): Likewise.
37160 2017-03-03  Jan Hubicka  <jh@suse.cz>
37162         PR lto/79760
37163         * ipa-devirt.c (maybe_record_node): Properly handle
37164         __cxa_pure_virtual visibility.
37166 2017-03-03  Martin Liska  <mliska@suse.cz>
37168         PR tree-optimization/79803
37169         * tree-ssa-loop-prefetch.c (tree_ssa_prefetch_arrays): Remove
37170         assert.
37171         (pass_loop_prefetch::execute): Disabled optimization if an
37172         assumption about L1 cache size is not met.
37174 2017-03-03  Martin Liska  <mliska@suse.cz>
37176         PR rtl-optimization/79574
37177         * gcse.c (struct gcse_expr): Use HOST_WIDE_INT instead of int.
37178         (hash_scan_set): Likewise.
37179         (dump_hash_table): Likewise.
37180         (hoist_code): Likewise.
37182 2017-03-03  Richard Biener  <rguenther@suse.de>
37184         * fixed-value.c (fixed_from_string): Restore use of elt (1)
37185         in place of uhigh ().
37186         (fixed_convert_from_real): Likewise.
37188 2017-03-03  Uros Bizjak  <ubizjak@gmail.com>
37190         PR target/79514
37191         * config/i386/i386.md (*pushxf_rounded): Use Pmode instead of DImode.
37193 2017-03-03  Richard Biener  <rguenther@suse.de>
37195         PR middle-end/79818
37196         * match.pd ( X +- C1 CMP C2 -> X CMP C2 -+ C1): Add missing
37197         TYPE_OVERFLOW_UNDEFINED check.
37199 2017-03-02  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
37201         * config/rs6000/vector.md (vector_ne_<mode>_p): Correct operand
37202         numbers.
37203         (vector_ae_<mode>_p): Likewise.
37204         (vector_nez_<mode>_p): Likewise.
37205         (vector_ne_v2di_p): Likewise.
37206         (vector_ae_v2di_p): Likewise.
37207         (vector_ne_<mode>_p): Likewise.
37208         * config/rs6000/vsx.md (vsx_tsqrt<mode>2_fg): Correct operand
37209         numbers.
37210         (vsx_tsqrt<mode>2_fe): Likewise.
37212 2017-03-02  Uros Bizjak  <ubizjak@gmail.com>
37214         PR target/79514
37215         * config/i386/i386.md (*pushxf_rounded): New insn_and_split pattern.
37217 2017-03-02  Jakub Jelinek  <jakub@redhat.com>
37219         PR rtl-optimization/79780
37220         * cprop.c (one_cprop_pass): When second and further conditional trap
37221         in a single basic block is turned into an unconditional trap, turn it
37222         into a deleted note to avoid RTL verification failures.
37224 2017-03-02  Richard Biener  <rguenther@suse.de>
37226         * fold-const.c (const_binop): Use ulow () instead of elt (0).
37228 2017-03-02  Richard Biener  <rguenther@suse.de>
37230         PR tree-optimization/79345
37231         PR c++/42000
37232         * tree-ssa-alias.c (walk_aliased_vdefs_1): Take a limit
37233         param and abort the walk, returning -1 if it is hit.
37234         (walk_aliased_vdefs): Take a limit param and pass it on.
37235         * tree-ssa-alias.h (walk_aliased_vdefs): Add a limit param,
37236         defaulting to 0 and return a signed int.
37237         * tree-ssa-uninit.c (struct check_defs_data): New struct.
37238         (check_defs): New helper.
37239         (warn_uninitialized_vars): Use walk_aliased_vdefs to warn
37240         about uninitialized memory.
37241         * fixed-value.c (fixed_from_string): Use ulow/uhigh to avoid
37242         bogus uninitialized warning.
37243         (fixed_convert_from_real): Likewise.
37245 2017-03-02  Bin Cheng  <bin.cheng@arm.com>
37247         PR tree-optimization/66768
37248         * tree-ssa-loop-ivopts.c (find_interesting_uses_address): Skip addr
37249         iv_use if base object can't be determined.
37251 2017-03-02  Jakub Jelinek  <jakub@redhat.com>
37253         PR tree-optimization/79345
37254         * gensupport.h (struct pattern_stats): Add min_scratch_opno field.
37255         * gensupport.c (get_pattern_stats_1) <case MATCH_SCRATCH>: Update it.
37256         (get_pattern_stats): Initialize it.
37257         * genemit.c (gen_expand): Verify match_scratch numbers come after
37258         match_operand/match_dup numbers.
37259         * config/i386/i386.md (<s>mul<mode>3_highpart): Swap match_dup and
37260         match_scratch numbers.
37261         * config/i386/sse.md (avx2_gathersi<mode>, avx2_gatherdi<mode>):
37262         Likewise.
37263         * config/s390/s390.md (trunctdsd2): Likewise.
37265 2017-03-02  Richard Biener  <rguenther@suse.de>
37267         * wide-int.h (wide_int_storage::operator=): Implement in terms
37268         of wi::copy.
37270 2017-03-02  Richard Biener  <rguenther@suse.de>
37272         PR tree-optimization/79777
37273         * tree-ssa-pre.c (eliminate_insert): Give up if we simplify
37274         the to insert expression to sth existing.
37276 2017-03-01  Martin Sebor  <msebor@redhat.com>
37278         PR middle-end/79692
37279         * gimple-ssa-sprintf.c
37280         (directive::known_width_and_precision): New function.
37281         (format_integer): Use it.
37282         (get_mpfr_format_length): Consider the full range of precision
37283         when computing %g output with the # flag.  Set the likely byte
37284         count to 3 rather than 1 when precision is indeterminate.
37285         (format_floating): Correct the lower bound of precision.
37287 2017-03-01  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
37289         * doc/invoke.texi: Document default code model for 64-bit Linux.
37291 2017-03-01  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
37293         PR target/79752
37294         * config/rs6000/rs6000.md (peephole2 for udiv/umod): Should emit
37295         udiv rather than div since input pattern is unsigned.
37297 2017-03-01  Uros Bizjak  <ubizjak@gmail.com>
37299         * config/i386/i386.c (print_reg): Warn for values of
37300         unsupported size in integer register.
37302 2017-03-01  Michael Meissner  <meissner@linux.vnet.ibm.com>
37304         PR target/79439
37305         * config/rs6000/predicates.md (current_file_function_operand): Do
37306         not allow self calls to be local if the function is replaceable.
37308 2017-03-01  Kelvin Nilsen  <kelvin@gcc.gnu.org>
37310         PR target/79395
37311         * config/rs6000/altivec.h (vec_ctz and others): Change the
37312         preprocessor macro that controls conditional compilation from
37313         _ARCH_PWR9 to __POWER9_VECTOR__.
37314         (vec_all_ne): Change parameterization of __altivec_scalar_pred
37315         macro expansion under preprocessor #ifdef __POWER9_VECTOR__
37316         control (instead of _ARCH_PWR9 control) so that template
37317         definition uses power9-specific function.
37318         (vec_any_eq): Likewise.
37319         (vec_all_ne): Change macro definition to use a power9-specific
37320         expansion under #ifdef __POWER9_VECTOR__ control (instead of
37321         _ARCH_PWR9 control).
37322         (vec_any_eq) Likewise.
37323         * config/rs6000/rs6000-builtin.def (CMPNEF): Remove BU_P9V_AV_2
37324         expansion for CMPNEF to remove support for xvcmpnesp instruction.
37325         (CMPNED): Remove BU_P9V_AV2 expansion for CMPNED to remove
37326         support for xvcmpnedp instruction.
37327         (VCMPNEB_P): Replace BU_P9V_AV_P macro expansion with BU_P9V_AV_2
37328         macro expansion so that Power9 implementation of vec_all_ne does
37329         not use the AltiVec predicate framework.
37330         (VCMPNEH_P): Likewise.
37331         (VCMPNEW_P): Likewise.
37332         (VCMPNED_P): Likewise.
37333         (VCMPNEFP_P): Likewise.
37334         (VCMPNEDP_P): Likewise.
37335         (VCMPAEB_P): Add BU_P9V_AV_2 macro expansion to change
37336         implementation of vec_any_eq to not use AltiVec predicate
37337         framework.
37338         (VCMPAEH_P): Likewise.
37339         (VCMPAEW_P): Likewise.
37340         (VCMPAED_P): Likewise.
37341         (VCMPAEFP_P): Likewise.
37342         (VCMPAEDP_P): Likewise.
37343         (VCMPNE_P): Replace BU_P9V_OVERLOAD_P macro expansion with
37344         BU_P9V_OVERLOAD_2 so that Power9 implementation of vec_all_ne does
37345         not use the AltiVec predicate framework.
37346         (VCMPAE_P): Add BU_P9V_OVERLOAD_2 macro to change implementation
37347         of vec_any_eq to not use AltiVec predicate framework.
37348         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Add
37349         support for predefined __POWER9_VECTOR__ macro to indicate that
37350         Power9 instruction selection is enabled.
37351         (altivec_overloaded_builtins): Remove extraneous
37352         ALTIVEC_BUILTIN_VEC_CMPNE entry for overloaded
37353         function argument types RS6000_BTI_bool_V16QI and
37354         RS6000_BTI_bool_V16QI.  Remove erroneous ALTIVEC_BUILTIN_VEC_CMPNE
37355         entry for overloaded function argument types RS6000_BTI_bool_V4SI
37356         andRS6000_BTI_bool_V4SI, mapping to P9V_BUILTIN_CMPNEB.  Remove
37357         two entries mapping to P9V_BUITIN_CMPNED and one entry mapping to
37358         P9V_BUILTIN_CMPNEF to force use of instructions not specific to
37359         Power9 for implementations of vec_cmpne.  Change the signature for
37360         all definitions of the overloaded P9V_BUILTIN_VEC_CMPNE_P function
37361         (representing vec_all_ne) to remove the previously described first
37362         argument of type RS6000_BTI_INTSI, as this was an artifact of
37363         reliance on the AltiVec predicate framework, which is no longer
37364         used in the implementation of these functions.  Add
37365         P9V_BUILTIN_VEC_VCMPAE_P entries (representing the vec_anyeq
37366         function) to match all of the P9V_BUILTIN_VEC_VCMNE_P entries
37367         since, unlike the AltiVec predicate framework implementation, we
37368         do not share function descriptors between vec_alle and vec_anyeq.
37369         (altivec_resolve_overloaded_builtin): Add SFmode and DFmode to the
37370         set of modes that receive special treatment even when
37371         TARGET_P9_VECTOR is true.  The special treatment emits code that
37372         does not depend on Power9 instructions.
37373         * config/rs6000/vector.md (vector_ne_<mode>_p): Change this
37374         define_expand to not rely on AltiVec predicate framework.
37375         (vector_ae_<mode>p): New define_expand to represent vec_any_eq
37376         function.
37377         (vector_ne_v2di_p): Change this define_expand to not rely on
37378         AltiVec predicate framework.
37379         (vector_ae_v2di_p): New define_expand to represent vec_any_eq
37380         function.
37381         (vector_ne_<mode>_p): Change this define_expand to not rely on
37382         AltiVec predicate framework.
37383         (vector_ae_<mode>p): New define_expand to represent vec_any_eq
37384         function.
37385         * config/rs6000/vsx.md (*vsx_ne_<mode>_p): For modes VSX_EXTRACT_I
37386         (V16QI, V8HI, V4SI), correct a typo in the code emitted for this
37387         define_insn pattern.
37388         (*vsx_ne_<mode>_p): For modes VSX_F (V4SF and V2DF), remove this
37389         define_insn pattern because the xvcmpne<VSs>. instruction is not
37390         supported.
37391         (vcmpne<VSs>): Remove this define_insn because xvcmpne<VSs>
37392         instruction is not supported.
37394 2017-03-01  Jakub Jelinek  <jakub@redhat.com>
37396         * config/nvptx/nvptx.c: Include intl.h.
37398 2017-03-01  Martin Jambor  <mjambor@suse.cz>
37400         PR lto/78140
37401         * ipa-prop.h (ipa_bits): Removed field known.
37402         (ipa_jump_func): Removed field vr_known.  Changed fields bits and m_vr
37403         to pointers.  Adjusted their comments to warn about their sharing.
37404         (ipcp_transformation_summary): Change bits to a vector of pointers.
37405         (ipa_check_create_edge_args): Moved to ipa-prop.c, declare.
37406         (ipa_get_ipa_bits_for_value): Declare.
37407         * tree-vrp.h (value_range): Mark as GTY((for_user)).
37408         * ipa-prop.c (ipa_bit_ggc_hash_traits): New.
37409         (ipa_bits_hash_table): Likewise.
37410         (ipa_vr_ggc_hash_traits): Likewise.
37411         (ipa_vr_hash_table): Likewise.
37412         (ipa_print_node_jump_functions_for_edge): Adjust for bits and m_vr
37413         being pointers and vr_known being removed.
37414         (ipa_set_jf_unknown): Likewise.
37415         (ipa_get_ipa_bits_for_value): New function.
37416         (ipa_set_jfunc_bits): Likewise.
37417         (ipa_get_value_range): New overloaded functions.
37418         (ipa_set_jfunc_vr): Likewise.
37419         (ipa_compute_jump_functions_for_edge): Use the above functions to
37420         construct bits and vr parts of jump functions.
37421         (ipa_check_create_edge_args): Move here from ipa-prop.h, also allocate
37422         ipa_bits_hash_table and ipa_vr_hash_table if they do not already
37423         exist.
37424         (ipcp_grow_transformations_if_necessary): Also allocate
37425         ipa_bits_hash_table and ipa_vr_hash_table if they do not already
37426         exist.
37427         (ipa_node_params_t::duplicate): Do not copy bits, just pointers to
37428         them.  Fix too long lines.
37429         (ipa_write_jump_function): Adjust for bits and m_vr being pointers and
37430         vr_known being removed.
37431         (ipa_read_jump_function): Use new setter functions to construct bits
37432         and vr parts of jump functions or set them to NULL.
37433         (write_ipcp_transformation_info): Adjust for bits being pointers.
37434         (read_ipcp_transformation_info): Likewise.
37435         (ipcp_update_bits): Likewise.  Fix excessively long lines a trailing
37436         space.
37437         Include gt-ipa-prop.h.
37438         * ipa-cp.c (propagate_bits_across_jump_function): Adjust for bits
37439         being pointers.
37440         (ipcp_store_bits_results): Likewise.
37441         (propagate_vr_across_jump_function): Adjust for m_vr being a pointer.
37442         Do not write to existing jump functions but use a temporary instead.
37444 2017-03-01  Jakub Jelinek  <jakub@redhat.com>
37446         PR c++/79681
37447         * fold-const.c (make_bit_field_ref): If orig_inner is COMPONENT_REF,
37448         attempt to use its first operand as BIT_FIELD_REF base.
37450 2017-03-01  Richard Biener  <rguenther@suse.de>
37452         PR middle-end/79721
37453         * tree-chrec.c (chrec_evaluate): Perform computation of Newtons
37454         interpolating formula in wrapping arithmetic.
37455         (chrec_apply): Convert chrec_evaluate return value to wanted type.
37457 2017-03-01  Jakub Jelinek  <jakub@redhat.com>
37459         PR tree-optimization/79734
37460         * tree-vect-generic.c (expand_vector_condition): Optimize
37461         AVX512 vector boolean VEC_COND_EXPRs into bitwise operations.
37462         Handle VEC_COND_EXPR where comparison has different inner width from
37463         type's inner width.
37465 2017-02-28  Sandra Loosemore  <sandra@codesourcery.com>
37467         * doc/invoke.texi (ARC Options): Copy-edit to fix punctuation,
37468         markup, and similar issues.  Remove @opindex entries for things
37469         that aren't options.  Add missing -mmpy-option entries.
37471 2017-02-28  Jakub Jelinek  <jakub@redhat.com>
37473         PR tree-optimization/79737
37474         * gimple-ssa-store-merging.c (encode_tree_to_bitpos): If bitlen is
37475         a multiple of BITS_PER_UNIT and !BYTES_BIG_ENDIAN, clear
37476         tmpbuf[byte_size - 1].  Call natice_encode_expr with byte_size - 1
37477         instead of byte_size.  Formatting fix.
37478         (shift_bytes_in_array_right): Formatting fix.
37480 2017-02-28  Eric Botcazou  <ebotcazou@adacore.com>
37482         PR target/79749
37483         * config/sparc/sparc.c (sparc_frame_pointer_required): Add missing
37484         condition on optimize for the leaf function test.
37486 2017-02-28  Martin Liska  <mliska@suse.cz>
37488         PR lto/79625
37489         * read-rtl-function.c (function_reader::handle_unknown_directive):
37490         Bail out when one uses -flto.
37492 2017-02-28  Martin Liska  <mliska@suse.cz>
37494         * common.opt: Replace space with tabular for options of <number>
37495         type.
37496         * config/i386/i386.opt: Show <number> value for
37497         -mlarge-data-threshold.
37498         * opts.c (print_filtered_help): Do not display number in hexadecimal
37499         format.
37501 2017-02-28  Martin Liska  <mliska@suse.cz>
37503         * common.opt: Fix --help=option -Q for options which are of
37504         an enum type.
37506 2017-02-28  Uros Bizjak  <ubizjak@gmail.com>
37508         * config/i386/i386.c (print_reg): Error out for values
37509         of 8-bit size in invalid integer register.
37511 2017-02-28  Martin Sebor  <msebor@redhat.com>
37513         PR tree-optimization/79691
37514         * passes.def (pass_all_optimizations_g): Enable pass_sprintf_length.
37516 2017-02-28  Jakub Jelinek  <jakub@redhat.com>
37518         PR target/79729
37519         * config/i386/i386.c (ix86_print_operand) <case 'R'>: Replace
37520         gcc_unreachable with output_operand_lossage.
37522 2017-02-28  Richard Biener  <rguenther@suse.de>
37524         PR tree-optimization/79740
37525         * tree-ssa-sccvn.c (vn_nary_op_insert_into): Allow redundant
37526         inserts.
37527         (visit_nary_op): Insert the nary into the hashtable if we
37528         pattern-matched sth.
37529         * tree-ssa-pre.c (eliminate_insert): Robustify.
37531 2017-02-28  Richard Biener  <rguenther@suse.de>
37533         PR middle-end/79731
37534         * fold-const.c (decode_field_reference): Reject out-of-bound
37535         accesses.
37537 2017-02-28  Jakub Jelinek  <jakub@redhat.com>
37539         * config/i386/i386.c: Include intl.h.
37540         (ix86_option_override_internal): Use cond ? G_("...") : G_("...")
37541         instead of just cond ? "..." : "...".
37542         * config/nvptx/nvptx.c (nvptx_goacc_validate_dims): Likewise.
37543         * coverage.c (read_counts_file): Likewise.
37544         * omp-offload.c: Include intl.h.
37545         (oacc_loop_fixed_partitions): Use cond ? G_("...") : G_("...") instead
37546         of just cond ? "..." : "...".
37547         * gcov.c (read_count_file): Use cond ? N_("...") : N_("...") instead
37548         of just cond ? "..." : "...".
37550 2017-02-28  Richard Earnshaw  <rearnsha@arm.com>
37552         PR target/79742
37553         * config/arm/parsecpu.awk (gen_data): Set tuning target to 'tune for'
37554         entry, if present.
37555         * config/arm/arm-cpus.in (cortex-m0plus.small-multiply): Correct
37556         'tune for' CPU name.
37557         * config/arm/arm-cpu-data.h: Regenerated.
37559 2017-02-28  Richard Biener  <rguenther@suse.de>
37561         PR tree-optimization/79732
37562         * tree-inline.c (expand_call_inline): Do not shadow var.
37564 2017-02-28  Richard Biener  <rguenther@suse.de>
37566         PR tree-optimization/79723
37567         * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Preserve
37568         address-space properly.
37570 2017-02-28  Thomas Schwinge  <thomas@codesourcery.com>
37572         * doc/optinfo.texi (Optimization groups): Fix option used for
37573         OPTGROUP_ALL.
37574         * doc/invoke.texi (-fopt-info): Document "omp".
37575         * dumpfile.h: Sort OPTGROUP_OMP before OPTGROUP_VEC.
37576         (OPTGROUP_ALL): Add OPTGROUP_OMP.
37577         * hsa-gen.c (pass_data_gen_hsail): Use OPTGROUP_OMP.
37578         * ipa-hsa.c (pass_data_ipa_hsa): Likewise.
37579         * omp-simd-clone.c (pass_data_omp_simd_clone): Likewise.
37581         * dumpfile.h (OPTGROUP_OPENMP): Rename to OPTGROUP_OMP.  Adjust
37582         all users.
37583         * dumpfile.c (optgroup_options): Instead of "openmp", associate
37584         OPTGROUP_OMP with "omp".
37586 2017-02-27  Pat Haugen  <pthaugen@us.ibm.com>
37588         PR target/79544
37589         * config/rs6000/rs6000-c.c (struct altivec_builtin_types): Use VSRAD
37590         for arithmetic shift of unsigned V2DI.
37592 2017-02-27  Claudiu Zissulescu  <claziss@synopsys.com>
37594         * config.gcc (arc*-): Clean up, use arc/big.h, arc/elf.h, and
37595         arc/linux.h headers.
37596         * config/arc/arc.h (TARGET_OS_CPP_BUILTINS): Remove.
37597         (LINK_SPEC): Likewise.
37598         (ARC_TLS_EXTRA_START_SPEC): Likewise.
37599         (EXTRA_SPECS): Likewise.
37600         (STARTFILE_SPEC): Likewise.
37601         (ENDFILE_SPEC): Likewise.
37602         (LIB_SPEC): Likewise.
37603         (TARGET_SDATA_DEFAULT): Likewise.
37604         (TARGET_MMEDIUM_CALLS_DEFAULT): Likewise.
37605         (MULTILIB_DEFAULTS): Likewise.
37606         (DWARF2_UNWIND_INFO): Likewise.
37607         * config/arc/big.h: New file.
37608         * config/arc/elf.h: Likewise.
37609         * config/arc/linux.h: Likewise.
37610         * config/arc/t-uClibc: Remove.
37612 2017-02-27  Bin Cheng  <bin.cheng@arm.com>
37614         PR tree-optimization/77536
37615         * tree-ssa-loop-manip.c (niter_for_unrolled_loop): New function.
37616         (tree_transform_and_unroll_loop): Use above function to compute the
37617         estimated niter of unrolled loop and use it when scaling profile.
37618         Also use count info rather than frequency if it's non-zero.
37619         * tree-ssa-loop-manip.h niter_for_unrolled_loop(): New declaration.
37620         * tree-vect-loop.c (scale_profile_for_vect_loop): New function.
37621         (vect_transform_loop): Call above function.
37623 2017-02-27  Richard Biener  <rguenther@suse.de>
37625         PR tree-optimization/45397
37626         * tree-ssa-pre.c (eliminate_insert): Handle BIT_AND_EXPR.
37627         * tree-ssa-sccvn.c (valueized_wider_op): New helper.
37628         (visit_nary_op): Add pattern matching for CSEing sign-changed
37629         or truncated operations with wider ones.
37631 2017-02-27  Richard Biener  <rguenther@suse.de>
37633         PR tree-optimization/79690
37634         * tree-vect-stmts.c (vectorizable_store): Use vector type
37635         built from the DR with address-space.
37637 2017-02-26  Gerald Pfeifer  <gerald@pfeifer.com>
37639         * doc/invoke.texi (Optimize Options): Refine the description
37640         of asan-use-after-return.
37642 2017-02-25  Alan Modra  <amodra@gmail.com>
37644         PR rtl-optimization/79584
37645         * lra-constraints.c (base_to_reg): Reload ad->base, the entire
37646         base, not ad->base_term, the reg within base.  Remove assertion
37647         that ad->base == ad->base_term.  Replace gen_int_mode using
37648         bogus mode with const0_rtx.
37650 2017-02-25  Jakub Jelinek  <jakub@redhat.com>
37652         PR middle-end/79396
37653         * tree-eh.c (operation_could_trap_p, stmt_could_throw_1_p): Handle
37654         FMA_EXPR like tcc_binary or tcc_unary.
37656         * tree-ssa-loop-niter.c (number_of_iterations_exit): Simplify warning.
37658         PR debug/77589
37659         * dwarf2out.c (struct dw_loc_list_struct): Add noted_variable_value
37660         bitfield.
37661         (size_of_loc_descr): Handle DW_OP_GNU_variable_value.
37662         (output_loc_operands): Handle DW_OP_call_ref and
37663         DW_OP_GNU_variable_value.
37664         (struct variable_value_struct): New type.
37665         (struct variable_value_hasher): Likewise.
37666         (variable_value_hash): New variable.
37667         (string_types): Remove.
37668         (copy_loc_descr): New function.
37669         (add_loc_descr_to_each): Clarify comment.  Use copy_loc_descr.
37670         (prepend_loc_descr_to_each): New function.
37671         (add_loc_list): Fix comment typo.  Use prepend_loc_descr_to_each
37672         instead of add_loc_descr_to_each if the first argument is single
37673         location list and the second has multiple.
37674         (resolve_args_picking_1): Handle DW_OP_GNU_variable_value.
37675         (loc_list_from_tree_1): For early_dwarf, emit DW_OP_GNU_variable_value
37676         when looking for variable value which doesn't have other location info.
37677         (loc_list_from_tree): Formatting fix.
37678         (gen_array_type_die): Simplify DW_AT_string_length handling.
37679         (adjust_string_types): Remove.
37680         (gen_subprogram_die): Don't call adjust_string_types nor test/set
37681         string_types.  Call resolve_variable_values.
37682         (prune_unused_types_walk_loc_descr): Handle DW_OP_GNU_variable_value.
37683         (resolve_addr_in_expr): Likewise.  Add A argument.
37684         (copy_deref_exprloc): Remove deref argument.  Adjust for the
37685         original expression being DW_OP_GNU_variable_value with optionally
37686         DW_OP_stack_value after it instead of DW_OP_call4 with DW_OP_deref
37687         optionally after it.
37688         (optimize_string_length): Rework for DW_OP_GNU_variable_value.
37689         (resolve_addr): Adjust optimize_string_length and resolve_addr_in_expr
37690         callers.  Set remove_AT_byte_size if removing DW_AT_string_length.
37691         (variable_value_hasher::hash, variable_value_hasher::equal): New
37692         methods.
37693         (resolve_variable_value_in_expr, resolve_variable_value,
37694         resolve_variable_values, note_variable_value_in_expr,
37695         note_variable_value): New functions.
37696         (dwarf2out_early_finish): Call note_variable_value on all toplevel
37697         DIEs.
37699 2017-02-24  Jakub Jelinek  <jakub@redhat.com>
37701         PR c/79677
37702         * opts.h (handle_generated_option): Add GENERATED_P argument.
37703         * opts-common.c (handle_option): Adjust function comment.
37704         (handle_generated_option): Add GENERATED_P argument, pass it to
37705         handle_option.
37706         (control_warning_option): Pass false to handle_generated_option
37707         GENERATED_P.
37708         * opts.c (maybe_default_option): Pass true to handle_generated_option
37709         GENERATED_P.
37710         * optc-gen.awk: Likewise.
37712 2017-02-24  Segher Boessenkool  <segher@kernel.crashing.org>
37714         * config/sh/sh.md (tstsi_t): If operands[0] is a SUBREG instead of
37715         a REG, look at the REG it is a SUBREG of.
37716         (splitter for cmpeqsi_t): Ditto.
37718 2017-02-24  Segher Boessenkool  <segher@kernel.crashing.org>
37720         * config/pa/pa.c (pa_combine_instructions): Do not share RTL.  Make
37721         the special USEs with the pattern of the insn, not the insn itself.
37723 2017-02-24  Matthew Fortune  <matthew.fortune@imgtec.com>
37725         PR target/79473
37726         * doc/invoke.texi: Document -mload-store-pairs.
37728 2017-02-24  Segher Boessenkool  <segher@kernel.crashing.org>
37729             Sandra Loosemore  <sandra@codesourcery.com>
37731         * config/nios2/nios2.c (nios2_simple_const_p): Returns false if the
37732         argument isn't a CONST_INT.
37733         (nios2_alternate_compare_const): Assert op is a CONST_INT.
37734         (nios2_valid_compare_const_p): Assert op is a CONST_INT.
37735         (nios2_validate_compare): Bypass alternate compare logic if *op2
37736         is not a CONST_INT.
37737         (ldstwm_operation_p): Return false if first_base is not a REG or
37738         if first_offset is not a CONST_INT.
37740 2017-02-24  Segher Boessenkool  <segher@kernel.crashing.org>
37742         * config/cris/cris.md: Use correct operand in a define_peephole2.
37744 2017-02-24  Segher Boessenkool  <segher@kernel.crashing.org>
37746         * config/c6x/c6x.c (predicate_insn): Do not incorrectly share RTL.
37748 2017-02-24  Segher Boessenkool  <segher@kernel.crashing.org>
37750         * config/arc/arc.c (arc_ccfsm_advance): Only take the PATTERN of
37751         this_insn if it is an INSN or JUMP_INSN.
37752         (force_offsettable): Look at base, not at addr.
37753         * config/arc/predicates.md (brcc_nolimm_operator): Don't call INTVAL
37754         on things that aren't necessarily CONST_INTs.
37756 2017-02-24  Uros Bizjak  <ubizjak@gmail.com>
37758         * doc/invoke.texi (x86 Options, -mfpmath=sse): Mention that
37759         -mfpmath=sse is the default also for x86-32 targets with SSE2
37760         instruction set when @option{-ffast-math} is enabled
37762 2017-02-24  Jeff Law  <law@redhat.com>
37764         PR rtl-optimizatoin/79286
37765         * ira.c (update_equiv_regs): Drop may_trap_p exception to
37766         dominance test.
37768 2017-02-24  Richard Biener  <rguenther@suse.de>
37770         PR tree-optimization/79389
37771         * gimple-ssa-split-paths.c (is_feasible_trace): Properly skip
37772         debug insns.
37774 2017-02-24  Aldy Hernandez  <aldyh@redhat.com>
37776         * tree-ssa-loop-niter.c (number_of_iterations_exit): Update
37777         function comment to reflect reality.
37778         (loop_exits_before_overflow): Fix typo in function description.
37780 2017-02-24  Richard Biener  <rguenther@suse.de>
37782         PR tree-optimization/79389
37783         * gimple-ssa-split-paths.c (is_feasible_trace): Verify more
37784         properly that a threading opportunity exists.  Detect conditional
37785         copy/constant propagation opportunities.
37787 2017-02-23  Eric Botcazou  <ebotcazou@adacore.com>
37789         * config/visium/visium.md (type): Add trap.
37790         (b): New mode attribute.
37791         (*btst): Rename into...
37792         (*btst<mode>): ...this and adjust.
37793         (*cbranchsi4_btst_insn): Rename into...
37794         (*cbranch<mode>4_btst_insn): ...this and adjust.
37795         (trap): New define_insn.
37797 2017-02-23  Jakub Jelinek  <jakub@redhat.com>
37799         PR tree-optimization/79389
37800         * ifcvt.c (struct noce_if_info): Add rev_cond field.
37801         (noce_reversed_cond_code): New function.
37802         (noce_emit_store_flag): Use rev_cond if non-NULL instead of
37803         reversed_comparison_code.  Formatting fix.
37804         (noce_try_store_flag): Test rev_cond != NULL in addition to
37805         reversed_comparison_code.
37806         (noce_try_store_flag_constants): Likewise.
37807         (noce_try_store_flag_mask): Likewise.
37808         (noce_try_addcc): Use rev_cond if non-NULL instead of
37809         reversed_comparison_code.
37810         (noce_try_cmove_arith): Likewise.  Formatting fixes.
37811         (noce_try_minmax, noce_try_abs): Clear rev_cond.
37812         (noce_find_if_block): Initialize rev_cond.
37813         (find_cond_trap): Call noce_get_condition with then_bb == trap_bb
37814         instead of false as last argument never attempt to reverse it
37815         afterwards.
37817 2017-02-23  Bin Cheng  <bin.cheng@arm.com>
37819         PR tree-optimization/79663
37820         * tree-predcom.c (combine_chains): Process refs in reverse order
37821         only for ZERO length chains, and add explaining comment.
37823 2017-02-23  Jeff Law  <law@redhat.com>
37825         PR tree-optimization/79578
37826         * tree-ssa-dse.c (clear_bytes_written_by): Use OEP_ADDRESS_OF
37827         in call to operand_equal_p.
37829 2017-01-23  Dominique d'Humieres  <dominiq@lps.ens.fr>
37831         PR target/71017
37832         * config/i386/cpuid.h: Fix another undefined behavior.
37834 2017-02-23  Richard Biener  <rguenther@suse.de>
37836         PR tree-optimization/79683
37837         * tree-vect-stmts.c (vect_analyze_stmt): Do not overwrite
37838         vector types for data-refs.
37840 2017-02-23  Martin Liska  <mliska@suse.cz>
37842         * params.def (PARAM_MIN_NONDEBUG_INSN_UID): Change default to 0.
37844 2017-02-23  Jakub Jelinek  <jakub@redhat.com>
37846         PR middle-end/79665
37847         * internal-fn.c (get_range_pos_neg): Moved to ...
37848         * tree.c (get_range_pos_neg): ... here.  No longer static.
37849         * tree.h (get_range_pos_neg): New prototype.
37850         * expr.c (expand_expr_real_2) <case TRUNC_DIV_EXPR>: If both arguments
37851         are known to be in between 0 and signed maximum inclusive, try to
37852         expand both unsigned and signed divmod and use the cheaper one from
37853         those.
37855 2017-02-22  Jeff Law  <law@redhat.com>
37857         PR tree-optimization/79578
37858         * tree-ssa-dse.c (clear_bytes_written_by): Use operand_equal_p
37859         to compare base operands.
37861 2017-02-22  Segher Boessenkool  <segher@kernel.crashing.org>
37863         PR target/79211
37864         * config/rs6000/rs6000.md (*fsel<SFDF:mode><SFDF2:mode>4): Use
37865         gpc_reg_operand instead of fpr_reg_operand.
37867 2017-02-22  Sameera Deshpande  <sameera.deshpande@imgtec.com>
37869         * config/mips/mips.c (mips_return_in_memory): Force FP
37870         vector types to be returned in memory for o32 ABI.
37872 2017-02-22  Jakub Jelinek  <jakub@redhat.com>
37874         * dwarf2out.c (gen_variable_die): For -gdwarf-5, use DW_TAG_variable
37875         instead of DW_TAG_member for static data member declarations and don't
37876         set no_linkage_name for static inline data members.
37877         (gen_member_die): For -gdwarf-5 don't change DW_TAG_variable
37878         to DW_TAG_member.
37880 2017-02-22  Martin Liska  <mliska@suse.cz>
37882         * doc/invoke.texi: Replace inequality signs with square brackets
37883         for -Wnormalized.
37885 2017-02-22  Matthew Fortune  <matthew.fortune@imgtec.com>
37887         PR target/78660
37888         * lra-constraints.c (simplify_operand_subreg): Handle
37889         WORD_REGISTER_OPERATIONS targets.
37891 2017-02-22  Jakub Jelinek  <jakub@redhat.com>
37893         PR target/70465
37894         * reg-stack.c (emit_swap_insn): Treat (float_extend:?F (mem:?F))
37895         and (const_double:?F) like (mem:?F) for the purpose of fxch %st(1)
37896         elimination by swapping fld*.
37898 2017-02-22  Richard Biener  <rguenther@suse.de>
37900         PR tree-optimization/79673
37901         * tree-ssa-pre.c (compute_avail): Use wide_int_to_tree to
37902         convert the [TARGET_]MEM_REF offset INTEGER_CST, scrapping off
37903         irrelevant address-space qualifiers and avoiding a
37904         ADDR_SPACE_CONVERT_EXPR from fold_convert.
37906 2017-02-22  Richard Biener  <rguenther@suse.de>
37908         PR tree-optimization/79666
37909         * tree-vrp.c (extract_range_from_binary_expr_1): Make sure
37910         to not symbolically negate if that may introduce undefined
37911         overflow.
37913 2017-02-22  Martin Liska  <mliska@suse.cz>
37915         PR lto/79587
37916         * data-streamer-in.c (streamer_read_gcov_count): Remove assert.
37917         * data-streamer-out.c (streamer_write_gcov_count_stream):
37918         Likewise.
37919         * value-prof.c (stream_out_histogram_value): Make assert more
37920         precise based on type of counter.
37922 2017-02-21  Uros Bizjak  <ubizjak@gmail.com>
37924         PR target/79593
37925         * config/i386/i386.md (standard_x87sse_constant_load splitter):
37926         Use nonimmediate_operand instead of memory_operand for operand 1.
37927         (float-extend standard_x87sse_constant_load splitter): Ditto.
37929 2017-02-21  Jeff Law  <law@redhat.com>
37931         PR tree-optimization/79621
37932         * gimple-ssa-isolate-paths.c (find_implicit_erroneous_behavior): Ignore
37933         blocks with edges to themselves.
37935 2017-02-21  Jakub Jelinek  <jakub@redhat.com>
37937         PR target/79633
37938         * tree-chkp-opt.c (chkp_optimize_string_function_calls): Use
37939         is_gimple_call instead of comparing gimple_code with GIMPLE_CALL.
37940         Use gimple_call_builtin_p.
37942         PR target/79570
37943         * sel-sched.c (moveup_expr_cached): Don't call sel_bb_head
37944         on temporarily removed DEBUG_INSNs.
37946         PR tree-optimization/79649
37947         * tree-loop-distribution.c (classify_partition): Give up on
37948         non-generic address space loads/stores.
37950 2017-02-21  Aldy Hernandez  <aldyh@redhat.com>
37952         * doc/loop.texi (Loop manipulation): Remove nonexistent
37953         tree_ssa_loop_version from the documentation.
37954         * cfgloopmanip.c (loop_version): Document CONDITION_BB argument.
37956 2017-02-21  Jakub Jelinek  <jakub@redhat.com>
37958         PR target/79494
37959         * config/i386/i386.c (ix86_expand_split_stack_prologue): Call
37960         make_reg_eh_region_note_nothrow_nononlocal on call_insn.
37961         * config/rs6000/rs6000.c: Include except.h.
37962         (rs6000_expand_split_stack_prologue): Call
37963         make_reg_eh_region_note_nothrow_nononlocal on the call insn.
37965 2017-02-21  Martin Jambor  <mjambor@suse.cz>
37967         PR lto/79579
37968         * ipa-prop.c (ipa_prop_write_jump_functions): Bail out if no edges
37969         have been analyzed.
37971 2017-02-21  Martin Jambor  <mjambor@suse.cz>
37973         * common.opt (-fipa-cp-alignment): Mark as ignored and preserved
37974         for backward compatibility only.
37975         * doc/invoke.texi (Option Summary): Remove all references to
37976         -fipa-cp-alignment.
37978 2017-02-21  Matthew Fortune  <matthew.fortune@imgtec.com>
37980         PR target/78660
37981         Revert:
37982         2017-02-20  Matthew Fortune  <matthew.fortune@imgtec.com>
37984         * lra-constraints.c (curr_insn_transform): Handle
37985         WORD_REGISTER_OPERATIONS requirements when reloading SUBREGs.
37987 2017-02-21  Martin Liska  <mliska@suse.cz>
37989         * config/i386/i386.opt: Replace -masm-dialect with -masm.
37991 2017-02-21  Thomas Schwinge  <thomas@codesourcery.com>
37993         PR translation/79638
37994         * config/nvptx/nvptx.c (ENTRY_TEMPLATE): Single out "%ntid.y".
37996 2017-02-21  Eric Botcazou  <ebotcazou@adacore.com>
37998         PR ada/67205
37999         * config/arm/arm.c (TARGET_CUSTOM_FUNCTION_DESCRIPTORS): Define.
38000         (arm_function_ok_for_sibcall): Return false for an indirect call by
38001         descriptor if all the argument registers are used.
38002         (arm_relayout_function): Use FUNCTION_ALIGNMENT macro to adjust the
38003         alignment of the function.
38005 2017-02-21  Jakub Jelinek  <jakub@redhat.com>
38007         PR tree-optimization/61441
38008         * simplify-rtx.c (simplify_const_unary_operation): For
38009         -fsignaling-nans and sNaN operand, return NULL_RTX rather than
38010         the sNaN unmodified.
38012 2017-02-20  Bernd Edlinger  <bernd.edlinger@hotmail.de>
38014         * Makefile.in (BUILD_SYSTEM_HEADER_DIR): New make variabe.
38015         (LIMITS_H_TEST, if_multiarch, stmp-fixinc): Use BUILD_SYSTEM_HEADER_DIR
38016         instead of SYSTEM_HEADER_DIR.
38018 2017-02-20  Gerald Pfeifer  <gerald@pfeifer.com>
38019             Martin LiÅ¡ka  <mliska@suse.cz>
38021         * doc/invoke.texi (use-after-scope-direct-emission-threshold):
38022         Fix typos and grammar, use active voice, and clarify.
38024 2017-02-20  Marek Polacek  <polacek@redhat.com>
38026         PR middle-end/79537
38027         * gimplify.c (gimplify_expr): Handle unused *&&L;.
38029         PR sanitizer/79558
38030         * ubsan.c (ubsan_type_descriptor): Check if TYPE_MAX_VALUE is null.
38032 2017-02-20  Jakub Jelinek  <jakub@redhat.com>
38034         PR target/79568
38035         * config/i386/i386.c (ix86_expand_builtin): Handle
38036         OPTION_MASK_ISA_AVX512VL and OPTION_MASK_ISA_64BIT in
38037         ix86_builtins_isa[fcode].isa as a requirement of those
38038         flags and any other flag in the bitmask.
38039         (ix86_init_mmx_sse_builtins): Use 0 instead of
38040         ~OPTION_MASK_ISA_64BIT as mask.
38041         * config/i386/i386-builtin.def (__builtin_ia32_rdtsc,
38042         __builtin_ia32_rdtscp, __builtin_ia32_pause, __builtin_ia32_bsrsi,
38043         __builtin_ia32_rdpmc, __builtin_ia32_rolqi, __builtin_ia32_rolhi,
38044         __builtin_ia32_rorqi, __builtin_ia32_rorhi): Likewise.
38046 2017-02-20  Matthew Fortune  <matthew.fortune@imgtec.com>
38048         PR target/78012
38049         * lra-constraints.c (split_reg): Check requested split mode
38050         is supported by the register.
38052 2017-02-20  Matthew Fortune  <matthew.fortune@imgtec.com>
38054         * lra-constraints.c (simplify_operand_subreg): Remove early
38055         return false.
38057 2017-02-20  Matthew Fortune  <matthew.fortune@imgtec.com>
38059         PR target/78660
38060         * lra-constraints.c (curr_insn_transform): Tighten condition
38061         for converting SUBREG reloads from OP_OUT to OP_INOUT.
38063 2017-02-20  Matthew Fortune  <matthew.fortune@imgtec.com>
38065         PR target/78660
38066         * lra-constraints.c (curr_insn_transform): Handle
38067         WORD_REGISTER_OPERATIONS requirements when reloading SUBREGs.
38069 2017-02-19  Uros Bizjak  <ubizjak@gmail.com>
38071         Revert:
38072         2016-05-30  Uros Bizjak  <ubizjak@gmail.com>
38074         * config/i386/sync.md (mfence_nosse): Use "lock orl $0, -4(%esp)".
38076 2017-02-19  Jonathan Wakely  <jwakely@redhat.com>
38078         PR c++/69523
38079         * doc/invoke.texi (C++ Dialect Options) [-Wliteral-suffix]: Update
38080         description.
38082 2017-02-19  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
38084         * gimple-pretty-print.c (dump_ternary_rhs): Adjust gimple dump format
38085         for FMA_EXPR.
38087 2017-02-18  Jakub Jelinek  <jakub@redhat.com>
38089         * final.c (last_columnnum, override_columnnum): New variables.
38090         (final_start_function): Set last_columnnum, pass it to begin_prologue
38091         hook and pass 0 to dwarf2out_begin_prologue.
38092         (final_scan_insn): Update override_columnnum.  Pass last_columnnum
38093         to source_line debug hook.
38094         (notice_source_line): Compute last_columnnum and for debug_column_info
38095         return true on column changes.
38096         * debug.h (struct gcc_debug_hooks): Add column argument to
38097         source_line and begin_prologue hooks.
38098         (debug_nothing_int_charstar_int_bool): Remove prototype.
38099         (debug_nothing_int_int_charstar,
38100         debug_nothing_int_int_charstar_int_bool): New prototypes.
38101         (dwarf2out_begin_prologue): Add column argument.
38102         * debug.c (do_nothing_debug_hooks): Adjust source_line and
38103         begin_prologue hooks.
38104         (debug_nothing_int_charstar_int_bool): Remove.
38105         (debug_nothing_int_int_charstar,
38106         debug_nothing_int_int_charstar_int_bool): New functions.
38107         * dwarf2out.c (dwarf2out_begin_prologue): Add column argument, pass it
38108         through to dwarf2out_source_line.
38109         (dwarf2_lineno_debug_hooks): Adjust begin_prologue hook.
38110         (dwarf2out_source_line): Add column argument, emit it if requested.
38111         * sdbout.c (sdbout_source_line, sdbout_begin_prologue): Add column
38112         arguments.
38113         * xcoffout.h (xcoffout_begin_prologue, xcoffout_source_line): Likewise.
38114         * xcoffout.c (xcoffout_begin_prologue, xcoffout_source_line): Likewise.
38115         * vmsdbgout.c (vmsdbgout_begin_prologue): Add column argument, pass it
38116         through to dwarf2out_begin_prologue.
38117         (vmsdbgout_source_line): Add column argument, pass it through to
38118         dwarf2out_source_line.
38119         * dbxout.c (dbxout_begin_prologue): Add column argument, adjust
38120         dbxout_source_line caller.
38121         (dbxout_source_line): Add column argument.
38123         * common.opt (gno-column-info, gcolumn-info): New options.
38124         * dwarf2out.c (dwarf2_lineno_debug_hooks): Formatting fix.
38125         (check_die): Also test for multiple DW_AT_decl_column attributes.
38126         (add_src_coords_attributes, dwarf2out_imported_module_or_decl_1): Add
38127         DW_AT_decl_column if requested.
38128         (gen_subprogram_die): Compare and/or add also DW_AT_decl_column
38129         if requested.
38130         (gen_variable_die): Likewise.
38131         (add_call_src_coords_attributes): Add DW_AT_call_column if requested.
38132         * doc/invoke.texi (-gcolumn-info, -gno-column-info): Document.
38134         PR target/79569
38135         * config/i386/i386.opt (m3dnowa): Replace Undocumented with Report.
38136         * common/config/i386/i386-common.c (OPTION_MASK_ISA_3DNOW_A_SET): Define.
38137         (ix86_handle_option): Handle OPT_m3dnowa.
38138         * doc/invoke.texi (-m3dnowa): Document.
38139         * doc/extend.texi (__builtin_ia32_pmulhuw, __builtin_ia32_pf2iw): Use
38140         -m3dnowa instead of -m3dnow -march=athlon.
38142         PR target/79559
38143         * config/i386/i386.c (ix86_print_operand): Use output_operand_lossage
38144         instead of gcc_assert for K, r and R code checks.  Formatting fixes.
38146 2017-02-17  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
38148         PR target/79261
38149         * config/rs6000/rs6000.c (rs6000_expand_ternop_builtin): Add
38150         support for CODE_FOR_vsx_xxpermdi_v2d[fi]_be.
38151         * config/rs6000/rs6000.md (reload_gpr_from_vsx<mode>): Call
38152         generator for vsx_xxpermdi_<mode>_be.
38153         * config/rs6000/vsx.md (vsx_xxpermdi_<mode>): Remove logic to
38154         force big-endian semantics.
38155         (vsx_xxpermdi_<mode>_be): New define_expand with same
38156         implementation as previous version of vsx_xxpermdi_<mode>.
38158 2017-02-17  Jakub Jelinek  <jakub@redhat.com>
38160         PR tree-optimization/79327
38161         * gimple-ssa-sprintf.c (format_integer): Remove likely_adjust
38162         variable, its initialization and use.
38164 2017-02-17  Julia Koval  <julia.koval@intel.com>
38166         * common/config/i386/i386-common.c (OPTION_MASK_ISA_RDPID_SET): New.
38167         (OPTION_MASK_ISA_PKU_UNSET): New.
38168         (ix86_handle_option): Handle -mrdpid.
38169         * config/i386/cpuid.h (bit_RDPID): New.
38170         * config/i386/driver-i386.c (host_detect_local_cpu):
38171         Detect RDPID feature.
38172         * config/i386/i386-builtin.def (__builtin_ia32_rdpid): New.
38173         * config/i386/i386-c.c (ix86_target_macros_internal):
38174         Handle RDPID flag.
38175         * config/i386/i386.c (ix86_target_string): Add -mrdpid to isa2_opts.
38176         (ix86_valid_target_attribute_inner_p): Add "rdpid".
38177         (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
38178         * config/i386/i386.h (TARGET_RDPID, TARGET_RDPID_P): New.
38179         * config/i386/i386.md (define_insn "rdpid"): New.
38180         * config/i386/i386.opt Add -mrdpid.
38181         * config/i386/immintrin.h (_rdpid_u32): New.
38183 2017-02-17  Vladimir Makarov  <vmakarov@redhat.com>
38185         PR rtl-optimization/79541
38186         * lra-constraints.c (curr_insn_transform): Remove wrong asm insn
38187         instead of transforming it into USE.
38189 2017-02-17  Segher Boessenkool  <segher@kernel.crashing.org>
38191         * config/rs6000/rs6000.md (extendsfdf2): Remove default arguments.
38192         If HONOR_SNANS (SFmode) force the input to a register.
38193         (*extendsfdf2_fpr): Add !HONOR_SNANS (SFmode) condition.
38194         (*extendsfdf2_snan): New pattern, used when using SNaNs; it generates
38195         an frsp or similar insn.
38197 2017-02-17  Martin Liska  <mliska@suse.cz>
38199         PR rtl-optimization/79577
38200         * params.def (selsched-max-sched-times): Increase minimum to 1.
38202 2017-02-17  Martin Liska  <mliska@suse.cz>
38204         PR rtl-optimization/79574
38205         * gcse.c (want_to_gcse_p): Prevent integer overflow.
38207 2017-02-17  Martin Liska  <mliska@suse.cz>
38209         PR tree-optimization/79529
38210         * tree-ssa-loop-unswitch.c (is_maybe_undefined): Use
38211         ssa_defined_default_def_p to handle cases which are implicitly
38212         defined.
38213         * tree-ssa.c (ssa_defined_default_def_p): New function.
38214         (ssa_undefined_value_p): Use ssa_defined_default_def_p to handle cases
38215         which are implicitly defined.
38216         * tree-ssa.h (ssa_defined_default_def_p): Declare.
38218 2017-02-17  Richard Biener  <rguenther@suse.de>
38220         PR middle-end/79576
38221         * params.def (max-ssa-name-query-depth): Limit to 10.
38223 2017-02-17  Richard Biener  <rguenther@suse.de>
38225         PR tree-optimization/79552
38226         * tree-ssa-structalias.c (visit_loadstore): Properly verify
38227         default defs.
38229 2017-02-17  Richard Biener  <rguenther@suse.de>
38231         PR bootstrap/79567
38232         * genmatch.c (output_line_directive): Handle DIR_SEPARATOR_2.
38234 2017-02-17  Marek Polacek  <polacek@redhat.com>
38236         PR middle-end/79536
38237         * fold-const.c (fold_negate_expr_1): Renamed from fold_negate_expr.
38238         (fold_negate_expr): New wrapper.
38240 2017-02-16  Sandra Loosemore  <sandra@codesourcery.com>
38242         * doc/invoke.texi (C++ Dialect Options) [-Wno-non-template-friend]: 
38243         Correct terminology and de-emphasize pre-standard behavior.
38245 2017-02-16  Alan Modra  <amodra@gmail.com>
38247         PR rtl-optimization/79286
38248         * ira.c (def_dominates_uses): New function.
38249         (update_equiv_regs): Don't create an equivalence for insns that
38250         may trap where the register def does not dominate the use.
38252 2017-02-16  Vladimir Makarov  <vmakarov@redhat.com>
38254         PR rtl-optimization/78127
38255         * lra.c (lra): Call lra_eliminate before finish the loop after
38256         lra_constraint.
38258 2017-02-16  Richard Biener  <rguenther@suse.de>
38260         * graphite.h: Do not include isl/isl_val_gmp.h, instead include
38261         isl/isl_val.h.
38262         * graphite-isl-ast-to-gimple.c (gmp_cst_to_tree): Remove.
38263         (gcc_expression_from_isl_expr_int): Use generic isl_val interface.
38264         * graphite-sese-to-poly.c: Do not include isl/isl_val_gmp.h.
38265         (isl_val_int_from_wi): New function.
38266         (extract_affine_gmp): Rename to ...
38267         (extract_affine_wi): ... this, take a widest_int.
38268         (extract_affine_int): Just wrap extract_affine_wi.
38269         (add_param_constraints): Use isl_val_int_from_wi.
38270         (add_loop_constraints): Likewise, and extract_affine_wi.
38272 2017-02-15  Jeff Law  <law@redhat.com>
38274         PR middle-end/79521
38275         * ira-costs.c (scan_one_insn): Check have_regs_of_mode before calling
38276         ira_init_register_move_cost_if_necessary.
38278 2017-02-15  Martin Sebor  <msebor@redhat.com>
38280         PR middle-end/32003
38281         * doc/invoke.texi (-fdump-final-insns): Replace option accidentally
38282         removed in a prior commit.
38284 2017-02-15  Bin Cheng  <bin.cheng@arm.com>
38286         PR tree-optimization/79347
38287         * tree-vect-loop-manip.c (vect_do_peeling): Maintain profile
38288         counters during peeling.
38290 2017-02-15  Thomas Schwinge  <thomas@codesourcery.com>
38292         * Makefile.in (site.exp): Remove "set ISLVER".
38294 2017-02-15  Jakub Jelinek  <jakub@redhat.com>
38296         PR target/79487
38297         * real.c (real_from_integer): Call real_convert even for decimal.
38299 2017-02-15  Dominik Vogt  <vogt@linux.vnet.ibm.com>
38301         PR target/79421
38302         * config/s390/s390.c: define TARGET_CUSTOM_FUNCTION_DESCRIPTORS.
38304 2017-02-14  Andrew Pinski  <apinski@cavium.com>
38306         * config/aarch64/aarch64-cores.def (thunderx2t99): Move to under 'C"
38307         cores and change the partno/implementer to be correct.
38308         (thunderx2t99p1): New core which replaces thunderx2t99 and still has
38309         the 'B" as the implementer.
38310         * config/aarch64/aarch64-tune.md: Regenerate.
38312 2017-02-14  Carl Love  <cel@us.ibm.com>
38314         * config/rs6000/rs6000.c: Add case statement entry to make the
38315         xvcvuxdsp built-in argument unsigned.
38316         * config/rs6000/vsx.md: Fix the source and return operand types so they
38317         match the instruction definitions from the ISA document.  Fix typo
38318         in the instruction generation for the (define_insn "vsx_xvcvuxdsp"
38319         statement.
38321 2017-02-14  Vladimir Makarov  <vmakarov@redhat.com>
38323         PR target/79282
38324         * lra-int.h (struct lra_operand_data, struct lra_insn_reg): Add
38325         member early_clobber_alts.
38326         * lra-lives.c (reg_early_clobber_p): New.
38327         (process_bb_lives): Use it.
38328         * lra.c (new_insn_reg): New arg early_clobber_alts.  Use it.
38329         (debug_operand_data): Initialize early_clobber_alts.
38330         (setup_operand_alternative): Set up early_clobber_alts.
38331         (collect_non_operand_hard_regs): Ditto.  Pass early clobber
38332         alternatives to new_insn_reg.
38333         (add_regs_to_insn_regno_info): Add arg early_clobber_alts.  Use
38334         it.
38335         (lra_update_insn_regno_info): Pass the new arg.
38337 2017-02-14  Jakub Jelinek  <jakub@redhat.com>
38339         PR middle-end/79505
38340         * omp-offload.c (free_oacc_loop): Release loop->ifns vector.
38341         (new_oacc_loop_raw): Don't clear already cleared fields.
38343         PR target/79481
38344         * config/i386/avx512pfintrin.h (_mm512_prefetch_i32gather_pd,
38345         _mm512_prefetch_i32gather_ps, _mm512_prefetch_i64gather_pd,
38346         _mm512_prefetch_i64gather_ps): New inline functions and macros.
38348 2017-02-14  Uros Bizjak  <ubizjak@gmail.com>
38350         PR target/79495
38351         * config/i386/i386.md (*movxf_internal): Add (o,rC) alternative.
38353 2017-02-14  H.J. Lu  <hongjiu.lu@intel.com>
38355         PR target/79498
38356         * config/i386/i386.c (timode_scalar_chain::convert_insn): Insert
38357         the extra instruction to the right place to store 128-bit constant
38358         when needed.
38360 2017-02-14  Martin Sebor  <msebor@redhat.com>
38362         PR middle-end/79448
38363         * gimple-ssa-sprintf.c (format_directive): Avoid issuing INT_MAX
38364           warning for strings of unknown length.
38366 2017-02-13  Segher Boessenkool  <segher@kernel.crashing.org>
38368         * config.gcc (supported_defaults) [powerpc*-*-*]: Update.
38370 2017-02-14  Jeff Law  <law@redhat.com>
38372         PR target/79404
38373         * ira-costs.c (scan_one_insn): Initialize register move costs
38374         for pseudos seen in USE/CLOBBER insns.
38376         PR tree-optimization/79095
38377         * tree-vrp.c (extract_range_from_binary_expr_1): For EXACT_DIV_EXPR,
38378         if the numerator has the range ~[0,0] make the resultant range ~[0,0].
38379         (extract_range_from_binary_expr): For MINUS_EXPR with no derived range,
38380         if the operands are known to be not equal, then the resulting range
38381         is ~[0,0].
38382         (intersect_ranges): If the new range is ~[0,0] and the old range is
38383         wide, then prefer ~[0,0].
38384         * tree-vrp.c (overflow_comparison_p_1): New function.
38385         (overflow_comparison_p): New function.
38386         * tree-vrp.c (register_edge_assert_for_2): Register additional asserts
38387         if NAME is used in an overflow test.
38388         (vrp_evaluate_conditional_warnv_with_ops): If the ops represent an
38389         overflow check that can be expressed as an equality test, then adjust
38390         ops to be that equality test.
38392 2017-02-14  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
38394         * config/s390/s390-builtin-types.def: Remove flags argument.
38395         * config/s390/s390.c (s390_init_builtins): Likewise.
38397 2017-02-14  Martin Liska  <mliska@suse.cz>
38399         * tree-ssa-loop-unswitch.c (hoist_guard): Release get_loop_body
38400         vector.  Fix trailing white spaces.
38402 2017-02-14  James Greenhalgh  <james.greenhalgh@arm.com>
38404         * config/aarch64/aarch64.c (aarch64_simd_container_mode): Handle
38405         HFmode.
38407 2017-02-14  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
38409         PR rtl-optimization/68664
38410         * config/arm/arm.c (arm_sched_can_speculate_insn):
38411         New function.  Declare prototype.
38412         (TARGET_SCHED_CAN_SPECULATE_INSN): Define.
38414 2017-02-14  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
38416         PR rtl-optimization/68664
38417         * config/aarch64/aarch64.c (aarch64_sched_can_speculate_insn):
38418         New function.
38419         (TARGET_SCHED_CAN_SPECULATE_INSN): Define.
38421 2017-02-14  Amit Pawar  <amit.pawar@amd.com>
38423         * config/i386/i386.c (znver1_cost): Fix the alignment for function and
38424         max skip bytes for function, loop and jump.
38426 2017-02-14  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
38428         * gimple-pretty-print.c (dump_unary_rhs): Adjust dump format for
38429         ABS_EXPR for gimple dump.
38431 2017-02-14  Jakub Jelinek  <jakub@redhat.com>
38433         PR target/79462
38434         * config/sh/sh.c (expand_cbranchdi4): Don't clear operands[4].
38436         PR tree-optimization/79408
38437         * tree-vrp.c (simplify_div_or_mod_using_ranges): Handle also the
38438         case when on TRUNC_MOD_EXPR op0 is INTEGER_CST.
38439         (simplify_stmt_using_ranges): Call simplify_div_or_mod_using_ranges
38440         also if rhs1 is INTEGER_CST.
38442 2017-02-14  Richard Biener  <rguenther@suse.de>
38444         PR middle-end/79432
38445         * tree-into-ssa.c (insert_phi_nodes): When the function can
38446         have abnormal edges rewrite SSA names with broken use-def
38447         dominance out of SSA and register them for PHI insertion.
38449 2017-02-13  Martin Sebor  <msebor@redhat.com>
38451         PR middle-end/79496
38452         * gimple-ssa-sprintf.c (pass_sprintf_length::handle_gimple_call): Avoid
38453         clearing info.nowrite flag when snprintf size argument is a range.
38455 2017-02-13  Jakub Jelinek  <jakub@redhat.com>
38457         * cprop.c (cprop_jump): Add missing space in string literal.
38458         * tree-ssa-structalias.c (rewrite_constraints): Likewise.
38459         (get_constraint_for_component_ref): Likewise.
38460         * df-core.c (df_worklist_dataflow_doublequeue): Likewise.
38461         * tree-outof-ssa.c (insert_partition_copy_on_edge): Likewise.
38462         * lra-constraints.c (process_alt_operands): Likewise.
38463         * ipa-inline.c (inline_small_functions): Likewise.
38464         * tree-ssa-sccvn.c (visit_reference_op_store): Likewise.
38465         * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
38466         * trans-mem.c (diagnose_tm_1_op): Likewise.
38467         * omp-grid.c (grid_find_single_omp_among_assignments): Likewise.
38468         (grid_parallel_clauses_gridifiable): Likewise.
38470         * config/nvptx/mkoffload.c (process): Add space in between
38471         , and %d.
38473         * config/i386/i386.h (REG_CLASS_NAMES): Add , in between
38474         "MOD4_SSE_REGS" and "ALL_REGS".
38476         * spellcheck.c (test_data): Add , in between "foo" and "food".
38478 2017-02-13  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
38480         PR target/79449
38481         * config/rs6000/rs6000.c (expand_block_compare): Make sure runtime
38482         boundary crossing check and subsequent code generation agree.
38484 2017-02-13  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
38486         * config/aarch64/aarch64.c (has_memory_op): Delete.
38487         (aarch64_madd_needs_nop): Use contains_mem_rtx_p instead of
38488         has_memory_op.
38490 2017-02-13  Jakub Jelinek  <jakub@redhat.com>
38492         PR rtl-optimization/79388
38493         PR rtl-optimization/79450
38494         * combine.c (distribute_notes): When removing TEM_INSN for which
38495         corresponding dest has last value recorded, invalidate that last
38496         value.
38498 2017-02-13  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
38500         * config/arm/arm.c (arm_print_tune_info): Use ASM_COMMENT_START instead
38501         of explicit '@'.  Add missing assembly comment marker on branch costs
38502         printout.
38504 2017-02-13  Nathan Sidwell  <nathan@acm.org>
38506         * gengtype-lex.l (<in_struct>): Add '/'.
38508 2017-02-13  Martin Liska  <mliska@suse.cz>
38510         PR c/79471
38511         * calls.c (expand_call): Replace XALLOCAVEC with XCNEWVEC.
38513 2017-02-13  Richard Biener  <rguenther@suse.de>
38515         * configure.ac (HAVE_ISL_OPTIONS_SET_SCHEDULE_SERIALIZE_SCCS):
38516         Remove.
38517         * configure: Re-generate.
38518         * config.in: Likewise.
38519         * graphite-dependences.c: Simplify as if
38520         HAVE_ISL_OPTIONS_SET_SCHEDULE_SERIALIZE_SCCS was defined.
38521         * graphite-isl-ast-to-gimple.c: Likewise.
38522         * graphite-optimize-isl.c: Likewise.
38523         * graphite-poly.c: Likewise.
38524         * graphite-sese-to-poly.c: Likewise.
38525         * graphite.h: Likewise.
38526         * toplev.c: Include isl/version.h and use isl_version () for
38527         printing the ISL version.
38528         * doc/install.texi: Update ISL requirement.
38530 2017-02-12  Gerald Pfeifer  <gerald@pfeifer.com>
38532         * doc/standards.texi (Standards): Update reference to
38533         Objective-C 2.0.
38535 2017-02-12  Gerald Pfeifer  <gerald@pfeifer.com>
38537         * doc/extend.texi (Named Address Spaces): sourceware.org now
38538         defaults to https.
38539         * doc/install.texi (Binaries): Ditto.
38540         (Specific): Ditto.
38542 2017-02-11  Sandra Loosemore  <sandra@codesourcery.com>
38544         * doc/cpp.texi: Replace "stringify"/"stringification" with C
38545         standard terminology "stringize"/"stringizing" throughout.
38546         * doc/cppinternals.texi: Likewise.
38548 2017-02-11  Sandra Loosemore  <sandra@codesourcery.com>
38550         * doc/extend.texi: Fix some spelling mistakes and typos.
38551         * doc/invoke.texi: Likewise.
38553 2017-02-11  Jan Hubicka  <hubicka@ucw.cz>
38555         PR ipa/79224
38556         * params.def (inline-min-speedup) Change from 10 to 8.
38558 2017-02-11  Jakub Jelinek  <jakub@redhat.com>
38560         * doc/invoke.texi (fopenmp): Bump OpenMP version from 4.0 to
38561         4.5.
38563 2017-02-11  Jan Hubicka  <hubicka@ucw.cz>
38565         PR ipa/79224
38566         * ipa-inline-analysis.c (get_minimal_bb): New function.
38567         (record_modified): Use it.
38568         (remap_edge_change_prob): Handle also ancestor functions.
38570 2017-02-11  Gerald Pfeifer  <gerald@pfeifer.com>
38572         * doc/contrib.texi (Contributors): Remove broken link into
38573         the Mauve CVS repository.
38575 2017-02-11  Jakub Jelinek  <jakub@redhat.com>
38577         PR middle-end/79454
38578         * internal-fn.c (expand_vector_ubsan_overflow): Use piece-wise
38579         result computation whenever lhs doesn't have vector mode, not
38580         just when it has BLKmode.
38582 2017-02-10  Gerald Pfeifer  <gerald@pfeifer.com>
38584         * doc/makefile.texi (profiledbootstrap): Refer to the
38585         installation instructions only in textual form.
38587 2017-02-10  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
38589         PR target/79295
38590         * config/rs6000/altivec.md (bcd<bcd_add_sub>): Fix constraints.
38592 2017-02-10  Gerald Pfeifer  <gerald@pfeifer.com>
38594         * doc/install.texi (Specific): Use https for blackfin.uclinux.org.
38595         (Specific): Update mingw-w64 reference.
38596         (Binaries): Ditto.
38597         (Specific): Remove broken link to Renesas RX processor.
38599 2017-02-10  Richard Biener  <rguenther@suse.de>
38601         * toplev.c (process_options): Do not mention obsolete graphite
38602         options when printing sorry message about missing graphite support.
38603         Mention -floop-nest-optimize.
38605 2017-02-10  Christophe Lyon  <christophe.lyon@linaro.org>
38607         * config/aarch64/arm_neon.h (vtst_p8): Rewrite without asm.
38608         (vtst_p16): Likewise.
38609         (vtstq_p8): Likewise.
38610         (vtstq_p16): Likewise.
38611         (vtst_p64): New.
38612         (vtstq_p64): Likewise.
38613         * config/arm/arm_neon.h (vgetq_lane_p64): New.
38614         (vset_lane_p64): New.
38615         (vsetq_lane_p64): New.
38617 2017-02-10  Jakub Jelinek  <jakub@redhat.com>
38619         PR tree-optimization/79411
38620         * tree-ssa-reassoc.c (is_reassociable_op): Return false if
38621         stmt operands are SSA_NAMEs used in abnormal phis.
38622         (can_reassociate_p): Return false if op is SSA_NAME used in abnormal
38623         phis.
38625 2017-02-09  Jan Hubicka  <hubicka@ucw.cz>
38627         PR ipa/70795
38628         * cgraphunit.c (cgraph_node::add_new_function): Set externally_visible
38629         flag if needed.
38631 2017-02-09  Jan Hubicka  <hubicka@ucw.cz>
38633         * tree-ssa-loop-unswitch.c (hoist_guard): Update profile.
38635 2017-02-09  Jakub Jelinek  <jakub@redhat.com>
38637         * omp-offload.c (oacc_loop_auto_partitions): Use || instead of |
38638         to avoid warning.
38640         PR c/79413
38641         * gimplify.h (is_gimple_sizepos): Only test for INTEGER_CST constants,
38642         not arbitrary TREE_CONSTANT.
38644         PR c/79431
38645         * gimplify.c (gimplify_adjust_omp_clauses): Ignore
38646         "omp declare target link" attribute unless is_global_var.
38647         * omp-offload.c (find_link_var_op): Likewise.
38649 2017-02-09  Nathan Sidwell  <nathan@codesourcery.com>
38650             Chung-Lin Tang  <cltang@codesourcery.com>
38652         * gimplify.c (gimplify_scan_omp_clauses): No special handling for
38653         OMP_CLAUSE_TILE.
38654         (gimplify_adjust_omp_clauses): Don't delete TILE.
38655         (gimplify_omp_for): Deal with TILE.
38656         * internal-fn.c (expand_GOACC_TILE): New function.
38657         * internal-fn.def (GOACC_DIM_POS): Comment may be overly conservative.
38658         (GOACC_TILE): New.
38659         * omp-expand.c (struct oacc_collapse): Add tile and outer fields.
38660         (expand_oacc_collapse_init): Add LOC paramter.  Initialize tile
38661         element fields.
38662         (expand_oacc_collapse_vars): Add INNER parm, adjust for tiling,
38663         avoid DIV for outermost collapse var.
38664         (expand_oacc_for): Insert tile element loop as needed.  Adjust.
38665         Remove out of date comments, fix whitespace.
38666         * omp-general.c (omp_extract_for_data): Deal with tiling.
38667         * omp-general.h (enum oacc_loop_flags): Add OLF_TILE flag,
38668         adjust OLF_DIM_BASE value.
38669         (struct omp_for_data): Add tiling field.
38670         * omp-low.c (scan_sharing_clauses): Allow OMP_CLAUSE_TILE.
38671         (lower_oacc_head_mark): Add OLF_TILE as appropriate.  Ensure 2 levels
38672         for auto loops.  Remove default auto determining, moved to
38673         oacc_loop_fixed_partitions.
38674         * omp-offload.c (struct oacc_loop): Change 'ifns' to vector of call
38675         stmts, add e_mask field.
38676         (oacc_dim_call): New function, abstracted out from oacc_thread_numbers.
38677         (oacc_thread_numbers): Use oacc_dim_call.
38678         (oacc_xform_tile): New.
38679         (new_oacc_loop_raw): Initialize e_mask, adjust for ifns vector.
38680         (finish_oacc_loop): Adjust for ifns vector.
38681         (oacc_loop_discover_walk): Append loop abstraction sites to list,
38682         add case for GOACC_TILE fns.
38683         (oacc_loop_xform_loop): Delete.
38684         (oacc_loop_process): Iterate over call list directly, and add
38685         handling for GOACC_TILE fns.
38686         (oacc_loop_fixed_partitions): Determine default auto, deal with TILE,
38687         dump partitioning.
38688         (oacc_loop_auto_partitions): Add outer_assign parm. Assign all but
38689         vector partitioning to outer loops.  Assign 2 partitions to loops
38690         when available. Add TILE handling.
38691         (oacc_loop_partition): Adjust oacc_loop_auto_partitions call.
38692         (execite_oacc_device_lower): Process GOACC_TILE fns,
38693         ignore unknown specs.
38694         * tree-nested.c (convert_nonlocal_omp_clauses): Allow OMP_CLAUSE_TILE.
38695         * tree.c (omp_clause_num_ops): Adjust TILE ops.
38696         * tree.h (OMP_CLAUSE_TILE_ITERVAR, OMP_CLAUSE_TILE_COUNT): New.
38698 2017-02-09  Gerald Pfeifer  <gerald@pfeifer.com>
38700         * configure.ac (ACX_BUGURL): Update.
38701         * configure: Regenerate.
38703 2017-02-09  Richard Biener  <rguenther@suse.de>
38705         PR tree-optimization/69823
38706         * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
38707         Properly enumerate all BBs in the region.  Use auto_vec/auto_bitmap.
38709 2017-02-09  Andrew Burgess  <andrew.burgess@embecosm.com>
38711         * config/arc/arc-c.def: Add __NPS400__ definition.
38712         * config/arc/arc.h (CPP_SPEC): Don't define __NPS400__ here.
38713         (TARGET_NPS400): Define.
38715 2017-02-09  Andrew Burgess  <andrew.burgess@embecosm.com>
38717         * config/arc/arc-arch.h (arc_arch_t): Move unchanged to earlier in
38718         file.
38719         (arc_cpu_t): Change base_architecture field, arch, to a arc_arc_t
38720         pointer, arch_info.
38721         (arc_cpu_types): Fill the arch_info field with a pointer into the
38722         arc_arch_types table.
38723         (arc_selected_cpu): Declare.
38724         * config/arc/arc.c (arc_selected_cpu): Make global.
38725         (arc_selected_arch): Delete.
38726         (arc_base_cpu): Delete.
38727         (arc_override_options): Remove references to deleted variables,
38728         update access to arch information.
38729         (ARC_OPT): Update access to arch information.
38730         (ARC_OPTX): Likewise.
38731         * config/arc/arc.h (arc_base_cpu): Remove declaration.
38732         (TARGET_ARC600): Update access to arch information.
38733         (TARGET_ARC601): Likewise.
38734         (TARGET_ARC700): Likewise.
38735         (TARGET_EM): Likewise.
38736         (TARGET_HS): Likewise.
38737         * config/arc/driver-arc.c (arc_cpu_to_as): Update access to arch
38738         information.
38740 2017-02-08  Pat Haugen  <pthaugen@us.ibm.com>
38742         PR target/78604
38743         * config/rs6000/rs6000.c (rs6000_emit_vector_cond_expr): Invert
38744         condition/operands for integer GE/LE/GEU/LEU operations.
38746 2017-02-08  Segher Boessenkool  <segher@kernel.crashing.org>
38748         PR translation/79397
38749         * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Fix spelling
38750         of AltiVec.
38752 2017-02-08  Martin Jambor  <mjambor@suse.cz>
38754         PR ipa/79375
38755         * ipa-prop.c (ipa_alloc_node_params): Make static, return bool
38756         whether allocation happened.
38757         (ipa_initialize_node_params): Do not call ipa_alloc_node_params if
38758         nothing was allocated.
38760 2017-02-08  Jakub Jelinek  <jakub@redhat.com>
38762         PR tree-optimization/79408
38763         * tree-vrp.c (simplify_div_or_mod_using_ranges): If op1 is not
38764         constant, but SSA_NAME with a known integer range, use the minimum
38765         of that range instead of op1 to determine if modulo can be replaced
38766         with its first operand.
38768 2016-02-08  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
38770         * config/riscv/riscv.c (riscv_build_integer_1): Avoid use of INT16_MAX.
38772 2017-02-08  Richard Biener  <rguenther@suse.de>
38774         PR tree-optimization/71824
38775         * graphite-scop-detection.c (scop_detection::build_scop_breadth):
38776         Check all loops contained in the merged region.
38778 2017-02-07  Andrew Pinski  <apinski@cavium.com>
38780         * config/aarch64/aarch64.md (popcount<mode>2): New pattern.
38782 2017-02-07  Andrew Pinski  <apinski@cavium.com>
38784         * config/aarch64/aarch64-cores.def (thunderx): Disable LSE.
38785         (thunderxt88): Likewise.
38786         (thunderxt81): Disable LSE and change v8.1 to v8.
38787         (thunderxt83): Likewise.
38789 2017-02-07  Jakub Jelinek  <jakub@redhat.com>
38790             Richard Biener  <rguenther@suse.de>
38792         PR middle-end/79399
38793         * ira-int.h (struct target_ira_int): Change x_max_struct_costs_size
38794         type from int to size_t.
38795         * ira-costs.c (struct_costs_size): Change type from int to size_t.
38797 2017-02-07  Jakub Jelinek  <jakub@redhat.com>
38799         PR rtl-optimization/79386
38800         * cprop.c (bypass_conditional_jumps): Initialize
38801         bypass_last_basic_block already before splitting bbs after
38802         unconditional traps...
38803         (bypass_conditional_jumps): ... rather than here.
38805         PR target/79299
38806         * config/i386/sse.md (xtg_mode, gatherq_mode): New mode attrs.
38807         (*avx512f_gathersi<mode>, *avx512f_gathersi<mode>_2,
38808         *avx512f_gatherdi<mode>, *avx512f_gatherdi<mode>_2): Use them,
38809         fix -masm=intel patterns.
38811 2017-02-07  Richard Biener  <rguenther@suse.de>
38813         PR tree-optimization/79256
38814         PR middle-end/79278
38815         * builtins.c (get_object_alignment_2): Use min_align_of_type
38816         to extract alignment for MEM_REFs to honor BIGGEST_FIELD_ALIGNMENT
38817         and ADJUST_FIELD_ALIGN.
38819         * doc/tm.texi.in (ADJUST_FIELD_ALIGN): Adjust to take additional
38820         type parameter.
38821         * doc/tm.texi: Regenerate.
38822         * stor-layout.c (layout_decl): Adjust.
38823         (update_alignment_for_field): Likewise.
38824         (place_field): Likewise.
38825         (min_align_of_type): Likewise.
38826         * config/arc/arc.h (ADJUST_FIELD_ALIGN): Adjust.
38827         * config/epiphany/epiphany.h (ADJUST_FIELD_ALIGN): Likewise.
38828         * config/epiphany/epiphany.c (epiphany_adjust_field_align): Likewise.
38829         * config/frv/frv.h (ADJUST_FIELD_ALIGN): Likewise.
38830         * config/frv/frv.c (frv_adjust_field_align): Likewise.
38831         * config/i386/i386.h (ADJUST_FIELD_ALIGN): Likewise.
38832         * config/i386/i386.c (x86_field_alignment): Likewise.
38833         * config/rs6000/aix.h (ADJUST_FIELD_ALIGN): Likewise.
38834         * config/rs6000/darwin.h (ADJUST_FIELD_ALIGN): Likewise.
38835         * config/rs6000/freebsd64.h (ADJUST_FIELD_ALIGN): Likewise.
38836         * config/rs6000/linux64.h (ADJUST_FIELD_ALIGN): Likewise.
38837         * config/rs6000/sysv4.h (ADJUST_FIELD_ALIGN): Likewise.
38838         * config/rs6000/rs6000.c (rs6000_special_adjust_field_align_p):
38839          Likewise.
38841         Revert
38842         2017-01-30  Richard Biener  <rguenther@suse.de>
38844         PR tree-optimization/79256
38845         * targhooks.c (default_builtin_vector_alignment_reachable): Honor
38846         BIGGEST_FIELD_ALIGNMENT and ADJUST_FIELD_ALIGN to fix up bogus
38847         alignment on TYPE.
38849 2017-02-07  Toma Tabacu  <toma.tabacu@imgtec.com>
38851         * config/mips/mips.c (mips_expand_builtin_insn): Convert the QImode
38852         argument of the pshufh, psllh, psllw, psrah, psraw, psrlh, psrlw
38853         builtins to SImode and emit a zero-extend, if necessary.
38855 2017-02-06  Palmer Dabbelt  <palmer@dabbelt.com>
38857         * docs/invoke.texi (RISC-V Options): Alphabetize.
38859 2017-02-06  Palmer Dabbelt  <palmer@dabbelt.com>
38861         * doc/invoke.texi (RISC-V Options): Use two spaces to separate
38862         options.
38864 2017-02-06  Palmer Dabbelt  <palmer@dabbelt.com>
38866         * config/riscv/riscv.c: New file.
38867         * common/config/riscv/riscv-common.c: Likewise.
38868         * config.gcc: Likewise.
38869         * config/riscv/constraints.md: Likewise.
38870         * config/riscv/elf.h: Likewise.
38871         * config/riscv/generic.md: Likewise.
38872         * config/riscv/linux.h: Likewise.
38873         * config/riscv/multilib-generator: Likewise.
38874         * config/riscv/peephole.md: Likewise.
38875         * config/riscv/pic.md: Likewise.
38876         * config/riscv/predicates.md: Likewise.
38877         * config/riscv/riscv-builtins.c: Likewise.
38878         * config/riscv/riscv-c.c: Likewise.
38879         * config/riscv/riscv-ftypes.def: Likewise.
38880         * config/riscv/riscv-modes.def: Likewise.
38881         * config/riscv/riscv-opts.h: Likewise.
38882         * config/riscv/riscv-protos.h: Likewise.
38883         * config/riscv/riscv.h: Likewise.
38884         * config/riscv/riscv.md: Likewise.
38885         * config/riscv/riscv.opt: Likewise.
38886         * config/riscv/sync.md: Likewise.
38887         * config/riscv/t-elf-multilib: Likewise.
38888         * config/riscv/t-linux: Likewise.
38889         * config/riscv/t-linux-multilib: Likewise.
38890         * config/riscv/t-riscv: Likewise.
38891         * configure.ac: Likewise.
38892         * doc/contrib.texi: Add Kito Cheng, Palmer Dabbelt, and Andrew
38893         Waterman as RISC-V maintainers.
38894         * doc/install.texi: Add RISC-V entries.
38895         * doc/invoke.texi: Add RISC-V options section.
38896         * doc/md.texi: Add RISC-V constraints section.
38897         * configure: Regenerated.
38899 2017-02-06  Michael Meissner  <meissner@linux.vnet.ibm.com>
38901         PR target/66144
38902         * config/rs6000/vector.md (vcond<mode><mode>): Allow the true and
38903         false values to be constant vectors with all 0 or all 1 bits set.
38904         (vcondu<mode><mode>): Likewise.
38905         * config/rs6000/predicates.md (vector_int_reg_or_same_bit): New
38906         predicate.
38907         (fpmask_comparison_operator): Update comment.
38908         (vecint_comparison_operator): New predicate.
38909         * config/rs6000/rs6000.c (rs6000_emit_vector_cond_expr): Optimize
38910         vector conditionals when the true and false values are constant
38911         vectors with all 0 bits or all 1 bits set.
38913 2017-02-06  Martin Sebor  <msebor@redhat.com>
38915         PR  tree-optimization/79376
38916         * gimple-fold.c (get_range_strlen): Set the minimum length to zero.
38918 2017-02-06  Uros Bizjak  <ubizjak@gmail.com>
38920         * config/i386/sse.md (vector modes -> vec_extract* splitter): Use
38921         explicit subreg RTX with operand 1.  Use VECTOR_MODE_P predicate
38922         to simplify split condition.
38924 2017-02-06  Jakub Jelinek  <jakub@redhat.com>
38926         * omp-expand.c (oxpand_omp_atomic_fetch_op,
38927         expand_omp_atomic_pipeline): Return false if can_atomic_load_p is
38928         false.
38930 2017-02-06  Segher Boessenkool  <segher@kernel.crashing.org>
38932         PR rtl-optimization/68664
38933         * target.def (can_speculate_insn): New hook.
38934         * doc/tm.texi.in (TARGET_SCHED_CAN_SPECULATE_INSN): New hook.
38935         * doc/tm.texi: Regenerate.
38936         * sched-rgn.c (can_schedule_ready_p): Use the new hook.
38937         * config/rs6000/rs6000.c (TARGET_SCHED_CAN_SPECULATE_INSN): New macro.
38938         (rs6000_sched_can_speculate_insn): New function.
38940 2017-02-06  Jakub Jelinek  <jakub@redhat.com>
38942         PR tree-optimization/79284
38943         * tree-vectorizer.h (VECT_SCALAR_BOOLEAN_TYPE_P): Define.
38944         * tree-vect-stmts.c (vect_get_vec_def_for_operand,
38945         vectorizable_mask_load_store, vectorizable_operation,
38946         vect_is_simple_cond, get_same_sized_vectype): Use it instead
38947         of comparing TREE_CODE of a type against BOOLEAN_TYPE.
38948         * tree-vect-patterns.c (check_bool_pattern, search_type_for_mask_1,
38949         vect_recog_bool_pattern, vect_recog_mask_conversion_pattern): Likewise.
38950         * tree-vect-slp.c (vect_get_constant_vectors): Likewise.
38951         * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
38952         Remove redundant gimple_code (stmt) == GIMPLE_ASSIGN test after
38953         is_gimple_assign (stmt).  Replace another such test with
38954         is_gimple_assign (stmt).
38956 2017-02-06  Georg-Johann Lay  <avr@gjlay.de>
38958         PR target/78883
38959         * config/avr/avr.c (rtl-iter.h): Include it.
38960         (TARGET_LEGITIMATE_COMBINED_INSN): New hook define...
38961         (avr_legitimate_combined_insn): ...and implementation.
38963 2017-02-06  Dominik Vogt  <vogt@linux.vnet.ibm.com>
38965         * config/s390/predicates.md ("larl_operand"): Use macros from hwint.h.
38966         * config/s390/s390.c (s390_const_operand_ok)
38967         (s390_canonicalize_comparison, s390_extract_part)
38968         (s390_single_part, s390_contiguous_bitmask_nowrap_p)
38969         (s390_contiguous_bitmask_p, s390_rtx_costs)
38970         (legitimize_pic_address): Likewise.
38971         * config/s390/s390.md ("clzdi2", "clztidi2"): Likewise.
38972         * config/s390/vx-builtins.md ("vec_genbytemaskv16qi")
38973         ("vec_permi<mode>", "vfae<mode>", "*vfaes<mode>", "vstrc<mode>")
38974         ("*vstrcs<mode>"): Use UINTVAL() to set unsigned HOST_WIDE_INT.
38975         * config/s390/vector.md ("vec_vfenes<mode>"): Likewise.
38977 2017-02-06  Georg-Johann Lay  <avr@gjlay.de>
38979         * config/avr/avr.md (*addhi3_zero_extend): Add alternative where
38980         REGNO($0) == REGNO($1).
38982 2017-02-06  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
38984         * config/s390/linux.h(SIZE_TYPE): Add comment.
38986 2017-02-06  Julian Brown  <julian@codesourcery.com>
38987             Naveen H.S  <Naveen.Hurugalawadi@cavium.com>
38988             Virendra Pathak  <virendra.pathak@broadcom.com>
38990         * config/aarch64/aarch64-cores.def: Change the scheduler
38991         to Thunderx2t99.
38992         * config/aarch64/aarch64.md: Include thunderx2t99.md.
38993         * config/aarch64/thunderx2t99.md: New file.
38995 2017-02-05  Gerald Pfeifer  <gerald@pfeifer.com>
38997         * doc/standards.texi (Go Language): Update link to language
38998         standard.
39000 2017-02-05  Jan Hubicka  <hubicka@ucw.cz>
39002         * tree-eh.c (lower_resx): Sanitize profile.
39003         (cleanup_empty_eh_move_lp): Likewise.
39005 2017-02-05  Jan Hubicka  <hubicka@ucw.cz>
39007         PR tree-ssa/79347
39008         * cfgloopmanip.c (lv_adjust_loop_entry_edge, loop_version): Add
39009         ELSE_PROB.
39010         * cfgloopmanip.h (loop_version): Update prototype.
39011         * modulo-sched.c (sms_schedule): Update call of loop_version.
39012         * tree-if-conv.c(version_loop_for_if_conversion): Likewise.
39013         * tree-parloops.c (gen_parallel_loop): Likewise.
39014         * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Likewise.
39015         * tree-ssa-loop-split.c (split_loop): Likewise.
39016         * tree-ssa-loop-unswitch.c (tree_unswitch_loop): Likewise.
39017         * tree-vect-loop-manip.c (vect_loop_versioning): Likewise.
39019 2017-02-05  Martin Liska  <mliska@suse.cz>
39021         PR bootstrap/78985
39022         * config/s390/s390.c (s390_gimplify_va_arg): Initialize local
39023         variable to NULL.
39024         (print_operand_address): Initialize a struct to zero.
39026 2017-02-05  Gerald Pfeifer  <gerald@pfeifer.com>
39028         * doc/contrib.texi (Contributors): Refer to Hans Boehm's
39029         garbage collector only in textual form.
39031 2017-02-05  Gerald Pfeifer  <gerald@pfeifer.com>
39033         * doc/extend.texi (x86 specific memory model extensions for
39034         transactional memory): Simplify a phrase.
39036 2017-02-05  Eric Botcazou  <ebotcazou@adacore.com>
39038         PR target/79353
39039         * config/sparc/sync.md (atomic_loaddi_1): Replace 'U' constraint with
39040         'r', 'm' constraint with 'T' and !TARGET_ARCH64 with TARGET_ARCH32.
39041         (atomic_storedi_1): Likewise.
39043 2017-02-04  Jakub Jelinek  <jakub@redhat.com>
39045         PR tree-optimization/79338
39046         * tree-parloops.c (gather_scalar_reductions): Don't call
39047         vect_analyze_loop_form for loop->inner before destroying loop's
39048         loop_vinfo.
39050 2017-02-03  Martin Sebor  <msebor@redhat.com>
39052         PR tree-optimization/79327
39053         * gimple-ssa-sprintf.c (tree_digits): Avoid adding the base prefix
39054         when precision has resulted in leading zeros.
39055         (format_integer): Adjust the likely counter to assume an unknown
39056         argument that may be zero is non-zero.
39058 2017-02-03  Jason Merrill  <jason@redhat.com>
39060         PR c++/78689
39061         * tree-inline.c (copy_tree_body_r) [COND_EXPR]: Revert change to
39062         avoid copying non-taken branch.
39064 2017-02-03  Jakub Jelinek  <jakub@redhat.com>
39066         PR tree-optimization/79340
39067         * tree-vect-loop.c (vectorizable_reduction): Release
39068         vec_defs elements after safe_splicing them into other vectors.
39069         Formatting fixes.
39071         PR tree-optimization/79327
39072         * gimple-ssa-sprintf.c (adjust_range_for_overflow): If returning
39073         true, always set *argmin and *argmax to TYPE_{MIN,MAX}_VALUE of
39074         dirtype.
39075         (format_integer): Use wide_int_to_tree instead of build_int_cst
39076         + to_?hwi.  If argmin is NULL, just set argmin and argmax to
39077         TYPE_{MIN,MAX}_VALUE of argtype.  Simplify and fix computation
39078         of shortest and longest sequence.
39080 2017-02-03  Uros Bizjak  <ubizjak@gmail.com>
39082         * config/i386/i386.c (dimode_scalar_chain::convert_reg):
39083         Use pextrd for TARGET_SSE4_1 when creating scalar copy.
39085 2017-02-03  Walter Lee  <walt@tilera.com>
39087         PR target/78862
39088         * config/tilegx/tilegx.md (tilegx_expand_prologue): Add blockage
39089         after initial stackframe link reg save.
39090         * config/tilepro/tilepro.md (tilepro_expand_prologue): Likewise.
39092 2017-02-03  Jakub Jelinek  <jakub@redhat.com>
39094         PR target/79354
39095         * config/rs6000/rs6000.md (movsi_from_sf): Use wb constraint instead of
39096         wu for stxssp alternative.
39098 2017-02-03  Martin Sebor  <msebor@redhat.com>
39100         PR tree-optimization/79352
39101         * gimple-fold.c (get_range_strlen): Add argument.
39102         (get_range_strlen): Change return type to bool.
39103         (get_maxval_strlen): Pass in a dummy argument.
39104         * gimple-fold.h (get_range_strlen): Change return type to bool.
39105         * gimple-ssa-sprintf.c (get_string_length): Set unlikely counter.
39106         * tree.h (array_at_struct_end_p): Add argument.
39107         * tree.c (array_at_struct_end_p): Handle it.
39109 2017-02-03  Martin Liska  <mliska@suse.cz>
39111         PR lto/66295
39112         * multiple_target.c (create_dispatcher_calls): Redirect edge
39113         from a caller of a dispatcher.
39114         (expand_target_clones): Make the clones local.
39115         (ipa_target_clone): Do both target clones and resolvers.
39116         (ipa_dispatcher_calls): Remove the pass.
39117         (pass_dispatcher_calls::gate): Likewise.
39118         (make_pass_dispatcher_calls): Likewise.
39119         * passes.def (pass_target_clone): Put as very first IPA early
39120         pass.
39122 2017-02-03  Martin Liska  <mliska@suse.cz>
39124         * symtab.c (symtab_node::binds_to_current_def_p): Bail out
39125         in case of a function with ifunc attribute.
39127 2017-02-03  Martin Liska  <mliska@suse.cz>
39129         * cgraph.c (cgraph_node::dump): Dump function version info.
39130         * symtab.c (symtab_node::dump_base): Add missing new line.
39132 2017-02-02  Jan Hubicka  <hubicka@ucw.cz>
39134         * tree-ssa-ifcombine.c (update_profile_after_ifcombine): New function.
39135         (ifcombine_ifandif): Use it.
39137 2017-02-03  Martin Liska  <mliska@suse.cz>
39139         * doc/invoke.texi: Document default value for
39140         use-after-scope-direct-emission-threshold.
39142 2017-02-03  Martin Liska  <mliska@suse.cz>
39144         PR tree-optimization/79339
39145         * gimple-ssa-sprintf.c (format_floating_max): Call mpfr_clear.
39146         (format_floating): Likewise.
39148 2017-02-03  Martin Liska  <mliska@suse.cz>
39150         PR ipa/79337
39151         * ipa-prop.c (ipa_node_params_t::insert): Remove current
39152         implementation.
39153         (ipa_node_params_t::remove): Likewise.
39154         * ipa-prop.h (ipa_node_params::ipa_node_params): Make default
39155         initialization from removed ipa_node_params_t::insert.
39156         (ipa_node_params::~ipa_node_params): Move from removed
39157         ipa_node_params_t::release.
39158         * symbol-summary.h (symbol_summary::m_released): New member.
39159         Do not release a summary twice.  Do not allow to call finalizer
39160         for types of a summary that live in GGC memory.
39162 2017-02-02  Naveen H.S  <Naveen.Hurugalawadi@cavium.com>
39164         * config/aarch64/aarch64.c (thunderx2t99_tunings): Enable AES and
39165         cmp_branch fusion.
39167 2017-02-02  Martin Sebor  <msebor@redhat.com>
39169         PR middle-end/79275
39170         * gimple-ssa-sprintf.c (get_string_length): Set lower bound to zero.
39171         (format_string): Tighten up the range of output for non-constant
39172         strings and correct the expected range for wide non-constant strings.
39174 2017-02-02  Martin Sebor  <msebor@redhat.com>
39176         * doc/invoke.texi (-maccumulate-args): Fix bad grammar.
39178         PR middle-end/32003
39179         * doc/invoke.texi (-fdump-tree-): Remove pass-specific options from
39180         index.
39181         (-fdump-tree-@var): Add to index and document how to come up
39182         with pass-specific option and dump file names.
39183         (-fdump-passes): Clarify where to look for output.
39185 2017-02-02  Jan Hubicka  <hubicka@ucw.cz>
39187         PR middle-end/77445
39188         * tree-ssa-threadbackward.c (profitable_jump_thread_path): Dump
39189         statistics of the analyzed path; allow threading for speed when
39190         any of BBs along the path are optimized for speed.
39192 2017-02-02  Eric Botcazou  <ebotcazou@adacore.com>
39194         PR middle-end/78468
39195         * emit-rtl.c (init_emit): Add ??? comment for problematic alignment
39196         settings of the virtual registers.
39198         Revert again
39199         2016-08-23  Dominik Vogt  <vogt@linux.vnet.ibm.com>
39201         * explow.c (get_dynamic_stack_size): Take known alignment of stack
39202         pointer + STACK_DYNAMIC_OFFSET into account when calculating the size
39203         needed.
39205 2017-02-02  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
39207         * config/s390/vx-builtins.md ("vec_ceil", "vec_floor")
39208         ("vec_trunc", "vec_roundc", "vec_round"): Remove expanders.
39210 2017-02-02  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
39212         * config/s390/s390.md: Add missing comments with the expanded
39213         mnemonics.
39214         * config/s390/vector.md: Likewise.
39215         * config/s390/vx-builtins.md: Likewise.
39217 2017-02-02  Jakub Jelinek  <jakub@redhat.com>
39219         PR target/79197
39220         * config/rs6000/rs6000.md (*fixuns_trunc<mode>di2_fctiduz): Rename to ...
39221         (fixuns_trunc<mode>di2): ... this, remove previous expander.  Put all
39222         conditions on a single line.
39224 2017-02-02  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
39226         * config/s390/s390-c.c (s390_cpu_cpp_builtins_internal): Rename
39227         __S390_VX__ to __VX__.
39229 2017-02-01  Andrew Pinski  <apinski@cavium.com>
39231         * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost): Pass
39232         stmt_info to record_stmt_cost.
39233         (vect_get_known_peeling_cost): Pass stmt_info if known to
39234         record_stmt_cost.
39235         * config/aarch64/aarch64-protos.h (cpu_vector_cost): Split
39236         cpu_vector_cost field into
39237         scalar_int_stmt_cost and scalar_fp_stmt_cost.  Split vec_stmt_cost
39238         field into vec_int_stmt_cost and vec_fp_stmt_cost.
39239         * config/aarch64/aarch64.c (generic_vector_cost): Update for the
39240         splitting of scalar_stmt_cost and vec_stmt_cost.
39241         (thunderx_vector_cost): Likewise.
39242         (cortexa57_vector_cost): LIkewise.
39243         (exynosm1_vector_cost): Likewise.
39244         (xgene1_vector_cost): Likewise.
39245         (thunderx2t99_vector_cost): Improve after the splitting of the two
39246         fields.
39247         (aarch64_builtin_vectorization_cost): Update for the splitting of
39248         scalar_stmt_cost and vec_stmt_cost.
39250 2017-02-01  Torvald Riegel  <triegel@redhat.com>
39251             Richard Henderson  <rth@redhat.com>
39253         * builtins.c (fold_builtin_atomic_always_lock_free): Make "lock-free"
39254         conditional on existance of a fast atomic load.
39255         * optabs-query.c (can_atomic_load_p): New function.
39256         * optabs-query.h (can_atomic_load_p): Declare it.
39257         * optabs.c (expand_atomic_exchange): Always delegate to libatomic if
39258         no fast atomic load is available for the particular size of access.
39259         (expand_atomic_compare_and_swap): Likewise.
39260         (expand_atomic_load): Likewise.
39261         (expand_atomic_store): Likewise.
39262         (expand_atomic_fetch_op): Likewise.
39263         * testsuite/lib/target-supports.exp
39264         (check_effective_target_sync_int_128): Remove x86 because it provides
39265         no fast atomic load.
39266         (check_effective_target_sync_int_128_runtime): Likewise.
39268 2017-02-01  Richard Biener  <rguenther@suse.de>
39270         * graphite.c: Include tree-vectorizer.h for find_loop_location.
39271         (graphite_transform_loops): Provide opt-info for optimized nests.
39272         * tree-parloop.c (parallelize_loops): Provide opt-info for
39273         parallelized loops.
39275 2017-02-01  Richard Biener  <rguenther@suse.de>
39277         PR middle-end/79315
39278         * tree-cfg.c (move_stmt_op): Never set TREE_BLOCK when it
39279         was not set before.
39281 2017-02-01  Richard Biener  <rguenther@suse.de>
39283         PR tree-optimization/71824
39284         * graphite-scop-detection.c (scop_detection::build_scop_breadth):
39285         Verify the loops are valid in the merged SESE region.
39286         (scop_detection::can_represent_loop_1): Check analyzing the
39287         evolution of the number of iterations in the region succeeds.
39289 2017-01-31  Ian Lance Taylor  <iant@golang.org>
39291         * config/i386/i386.c (ix86_expand_split_stack_prologue): Add
39292         REG_ARGS_SIZE note to 32-bit push insns and call insn.
39294 2017-01-31  David Malcolm  <dmalcolm@redhat.com>
39296         PR preprocessor/79210
39297         * input.c (get_substring_ranges_for_loc): Replace line_width
39298         assertion with error-handling.
39300 2017-01-31  Richard Biener  <rguenther@suse.de>
39302         PR tree-optimization/77318
39303         * graphite-sese-to-poly.c (extract_affine): Fix assert.
39304         (create_pw_aff_from_tree): Take loop parameter.
39305         (add_condition_to_pbb): Pass loop of the condition to
39306         create_pw_aff_from_tree.
39308 2017-01-31  Jakub Jelinek  <jakub@redhat.com>
39310         * config/s390/s390.c (s390_asan_shadow_offset): New function.
39311         (TARGET_ASAN_SHADOW_OFFSET): Redefine.
39313 2017-01-31  Michael Meissner  <meissner@linux.vnet.ibm.com>
39315         PR target/78597
39316         PR target/79038
39317         * config/rs6000/rs6000-protos.h (convert_float128_to_int): Delete,
39318         no longer used.
39319         (convert_int_to_float128): Likewise.
39320         * config/rs6000/rs6000.c (convert_float128_to_int): Likewise.
39321         (convert_int_to_float128): Likewise.
39322         * config/rs6000/rs6000.md (UNSPEC_IEEE128_MOVE): Likewise.
39323         (UNSPEC_IEEE128_CONVERT): Likewise.
39324         (floatsi<mode>2, FLOAT128 iterator): Bypass calling
39325         rs6000_expand_float128_convert if we have IEEE 128-bit hardware.
39326         Use local variables for IBM extended format.
39327         (fix_trunc<mode>si2, FLOAT128 iterator): Likewise.
39328         (fix_trunc<mode>si2_fprs): Likewise.
39329         (fixuns_trunc<IEEE128:mode><SDI:mode>2): Likewise.
39330         (floatuns<IEEE128:mode>2, IEEE128 iterator): Likewise.
39331         (fix<uns>_<mode>si2_hw): Rework the IEEE 128-bt hardware support
39332         to know that we can now have integers of all sizes in vector
39333         registers.
39334         (fix<uns>_<mode>di2_hw): Likewise.
39335         (float<uns>_<mode>si2_hw): Likewise.
39336         (fix_<mode>si2_hw): Likewise.
39337         (fixuns_<mode>si2_hw): Likewise.
39338         (float<uns>_<mode>di2_hw): Likewise.
39339         (float_<mode>di2_hw): Likewise.
39340         (float_<mode>si2_hw): Likewise.
39341         (floatuns_<mode>di2_hw): Likewise.
39342         (floatuns_<mode>si2_hw): Likewise.
39343         (xscvqp<su>wz_<mode>): Delete, no longer used.
39344         (xscvqp<su>dz_<mode>): Likewise.
39345         (xscv<su>dqp_<mode>): Likewise.
39346         (ieee128_mfvsrd_64bit): Likewise.
39347         (ieee128_mfvsrd_32bit): Likewise.
39348         (ieee128_mfvsrwz): Likewise.
39349         (ieee128_mtvsrw): Likewise.
39350         (ieee128_mtvsrd_64bit): Likewise.
39351         (ieee128_mtvsrd_32bit): Likewise.
39353 2017-01-31  Martin Liska  <mliska@suse.cz>
39355         PR ipa/79285
39356         * ipa-prop.c (ipa_free_all_node_params): Call release method
39357         instead of ~sumbol_summary to not to trigger double times
39358         dtor of hash_map.
39360 2017-01-31  Aldy Hernandez  <aldyh@redhat.com>
39362         PR tree-optimization/71691
39363         * bitmap.h (class auto_bitmap): New.
39364         * tree-ssa-loop-unswitch.c (tree_may_unswitch_on): Call
39365         is_maybe_undefined instead of ssa_undefined_value_p.
39367 2017-01-31  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
39369         * config/s390/s390-c.c (s390_cpu_cpp_builtins_internal): Rename
39370         __S390_ARCH_LEVEL__ to __ARCH__.
39372 2017-01-31  Jakub Jelinek  <jakub@redhat.com>
39374         PR tree-optimization/79267
39375         * value-prof.c (gimple_ic): Only drop lhs for noreturn calls
39376         if should_remove_lhs_p is true.
39378 2017-01-30  Alexandre Oliva  <aoliva@redhat.com>
39380         PR debug/63238
39381         * dwarf2out.c (clone_as_declaration): Drop DW_AT_alignment.
39382         (add_alignment_attribute): New.
39383         (base_type_die): Add alignment attribute.
39384         (subrange_type_die): Likewise.
39385         (modified_type_die): Likewise.
39386         (gen_array_type_die): Likewise.
39387         (gen_descr_array_type_die: Likewise.
39388         (gen_enumeration_type_die): Likewise.
39389         (gen_subprogram_die): Likewise.
39390         (gen_variable_die): Likewise.
39391         (gen_field_die): Likewise.
39392         (gen_ptr_to_mbr_type_die): Likewise.
39393         (gen_struct_or_union_type_die): Likewise.
39394         (gen_subroutine_type_die): Likewise.
39395         (gen_typedef_die): Likewise.
39396         (base_type_cmp): Compare alignment attribute.
39398 2017-01-30  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
39400         PR target/79170
39401         * config/rs6000/altivec.md (*setb_internal): Rename to setb_signed.
39402         (setb_unsigned) New pattern for setb with CCUNS.
39403         * config/rs6000/rs6000.c (expand_block_compare): Use a different
39404         subfc./subfe sequence to avoid overflow problems.  Generate a
39405         shorter sequence with cmpld/setb for power9.
39406         * config/rs6000/rs6000.md (subf<mode>3_carry_dot2): Add a new pattern
39407         for generating subfc. instruction.
39408         (cmpstrsi): Add TARGET_POPCNTD predicate as the generate sequence
39409         now uses this instruction.
39411 2017-01-30  Ian Lance Taylor  <iant@google.com>
39413         PR debug/79289
39414         * dwarf2out.c (gen_type_die_with_usage): When picking a variant
39415         for FUNCTION_TYPE/METHOD_TYPE, use the first matching one.
39417 2017-01-30  Martin Sebor  <msebor@redhat.com>
39419         * gimple-ssa-sprintf.c (fmtresult::adjust_for_width_or_precision):
39420         Move constant to the right of a relational operator.
39421         (get_mpfr_format_length, format_character, format_string): Ditto.
39422         (should_warn_p, maybe_warn): Same.
39424         * doc/invoke.texi (-Wformat-truncation=1): Fix typo.
39426 2017-01-30  Maxim Ostapenko  <m.ostapenko@samsung.com>
39428         PR lto/79061
39429         * asan.c (get_translation_unit_decl): Remove function.
39430         (asan_add_global): Force has_dynamic_init to zero in LTO mode.
39432 2017-01-30  Martin Liska  <mliska@suse.cz>
39434         PR gcov-profile/79259
39435         * opts.c (common_handle_option): Enable flag_ipa_bit_cp w/
39436         -fprofile-generate.
39438 2017-01-30  Martin Liska  <mliska@suse.cz>
39440         PR bootstrap/78985
39441         * config/aarch64/cortex-a57-fma-steering.c (func_fma_steering::analyze):
39442         Initialize variables with NULL value.
39444 2017-01-30  Richard Earnshaw  <rearnsh@arm.com>
39446         PR target/79260
39447         * config.gcc (arm*-*-*): Add arm/arm-flags.h and arm/arm-isa.h to
39448         tm_p_file.
39449         * arm/arm-protos.h: Don't directly include arm-flags.h and arm-isa.h.
39451 2017-01-30  Richard Biener  <rguenther@suse.de>
39453         PR tree-optimization/79276
39454         * tree-vrp.c (process_assert_insertions): Properly adjust common
39455         when removing a duplicate.
39457 2017-01-30  Richard Biener  <rguenther@suse.de>
39459         PR tree-optimization/79256
39460         * targhooks.c (default_builtin_vector_alignment_reachable): Honor
39461         BIGGEST_FIELD_ALIGNMENT and ADJUST_FIELD_ALIGN to fix up bogus
39462         alignment on TYPE.
39463         * tree.c (build_aligned_type): Set TYPE_USER_ALIGN.
39465 2017-01-30  Dominik Vogt  <vogt@linux.vnet.ibm.com>
39467         PR target/79240
39468         * config/s390/s390.md ("*r<noxa>sbg_<mode>_srl_bitmask")
39469         ("*r<noxa>sbg_<mode>_sll_bitmask")
39470         ("*extzv_<mode>_srl<clobbercc_or_nocc>")
39471         ("*extzv_<mode>_sll<clobbercc_or_nocc>"):
39472         Use contiguous_bitmask_nowrap_operand.
39474 2017-01-29  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
39476         PR target/79268
39477         * config/rs6000/altivec.h (vec_xl): Revise #define.
39478         (vec_xst): Likewise.
39480 2017-01-27  Uros Bizjak  <ubizjak@gmail.com>
39482         * config/i386/i386.c (print_reg): Use REGNO instead of true_regnum.
39484 2017-01-27  Bernd Schmidt  <bschmidt@redhat.com>
39486         PR rtl-optimization/79194
39487         * cprop.c (one_cprop_pass): Move deletion of code after unconditional
39488         traps before call to bypass_conditional_jumps.
39490 2017-01-27  Vladimir Makarov  <vmakarov@redhat.com>
39492         PR tree-optimization/71374
39493         * lra-constraints.c (check_conflict_input_operands): New.
39494         (match_reload): Use it.
39496 2017-01-27  Vladimir Makarov  <vmakarov@redhat.com>
39498         PR target/79131
39499         * lra-assigns.c (find_hard_regno_for_1): Take endianess for into
39500         account to calculate conflict_set.
39502 2017-01-27  Bin Cheng  <bin.cheng@arm.com>
39504         PR rtl-optimization/78559
39505         * combine.c (try_combine): Discard REG_EQUAL and REG_EQUIV for
39506         other_insn in combine.
39508 2017-01-27  Pekka Jääskeläinen  <pekka.jaaskelainen@parmance.com>
39510         * builtin-types.def: Use unsigned_char_type_node for BT_UINT8.  Use
39511         uint16_type_node for BT_UINT16.
39513 2017-01-27  David Malcolm  <dmalcolm@redhat.com>
39515         * doc/sourcebuild.texi (Testsuites): Add "GIMPLE Tests" and
39516         "RTL Tests" to menu.
39517         (GIMPLE Tests): New node.
39518         (RTL Tests): New node.
39520 2017-01-27  Richard Biener  <rguenther@suse.de>
39522         PR tree-optimization/79245
39523         * tree-loop-distribution.c (distribute_loop): Apply cost
39524         modeling also to detected patterns.
39526 2017-01-27  Richard Biener  <rguenther@suse.de>
39528         PR tree-optimization/71433
39529         * tree-vrp.c (register_new_assert_for): Revert earlier changes.
39530         (compare_assert_loc): New function.
39531         (process_assert_insertions): Sort and optimize assert locations
39532         to remove duplicates and push down identical assertions on
39533         edges to their destination block.
39535 2017-01-27  Richard Biener  <rguenther@suse.de>
39537         PR tree-optimization/79244
39538         * tree-vrp.c (remove_range_assertions): Forcefully propagate
39539         out SSA names even if abnormal.
39541 2017-01-27  Jakub Jelinek  <jakub@redhat.com>
39543         * realmpfr.h: Poison MPFR_RND{N,Z,U,D}.
39544         * gimple-ssa-sprintf.c (format_floating_max): Use GMP_RNDN
39545         instead of MPFR_RNDN.
39547 2017-01-27  Richard Earnshaw  <rearnsha@arm.com>
39549         PR target/79239
39550         * arm.c (arm_option_override): Don't call build_target_option_node
39551         until after doing all option overrides.
39552         (arm_valid_target_attribute_tree): Likewise.
39554 2017-01-27  Martin Liska  <mliska@suse.cz>
39556         * doc/invoke.texi (-fprofile-arcs): Document profiling support
39557         for {cd}tors and C++ {cd}tors.
39559 2017-01-27  Dominik Vogt  <vogt@linux.vnet.ibm.com>
39561         * config/s390/s390.md ("*setmem_long_and")
39562         ("*setmem_long_and_31z"): Use zero_extend instead of and.
39564 2017-01-26  Martin Sebor  <msebor@redhat.com>
39566         * gimple-ssa-sprintf.c (format_floating): Simplify the computation
39567         of precision.
39569 2017-01-26  Martin Sebor  <msebor@redhat.com>
39571         * gimple-ssa-sprintf.c (format_floating): Test HAVE_XFmode and
39572         HAVE_DFmode before using XFmode or DFmode.
39573         (parse_directive): Avoid using the z length modifier to avoid
39574         the ISO C++98 does not support the â€˜z’ gnu_printf length modifier.
39576         PR middle-end/78703
39577         * gimple-ssa-sprintf.c (adjust_for_width_or_precision): Change
39578         to accept adjustment as an array.
39579         (get_int_range): New function.
39580         (struct directive): Make width and prec arrays.
39581         (directive::set_width, directive::set_precision): Call get_int_range.
39582         (format_integer, format_floating): Handle width and precision ranges.
39583         (format_string, parse_directive): Same.
39585 2017-01-26  Jakub Jelinek  <jakub@redhat.com>
39587         PR debug/79129
39588         * dwarf2out.c (generate_skeleton_bottom_up): For children with
39589         comdat_type_p set, just clone them, but keep the children in the
39590         original DIE.
39592         PR debug/78835
39593         * dwarf2out.c (prune_unused_types): Mark all functions with DIEs
39594         which have direct callers with -fvar-tracking-assignments enabled
39595         in the current TU.
39596         (resolve_addr): Avoid adding skeleton DIEs for DW_AT_call_origin
39597         inside of type units.
39599 2017-01-26  Martin Sebor  <msebor@redhat.com>
39601         PR middle-end/78703
39602         * gimple-ssa-sprintf.c (struct result_range): Add likely and
39603         unlikely counters.
39604         (struct format_result): Replace number_chars, number_chars_min,
39605         and number_chars_max with a single member of struct result_range.
39606         Remove bounded.
39607         (format_result::operator+=): Adjust.
39608         (struct fmtresult): Remove bounded.  Handle likely and unlikely
39609         counters.
39610         (fmtresult::adjust_for_width_or_precision): New function.
39611         (fmtresult:type_max_digits): New function.
39612         (bytes_remaining): Handle likely and unlikely counters.
39613         (min_bytes_remaining): Remove.
39614         (format_percent): Simplify.
39615         (format_integer, format_floating): Set likely and unlikely counters.
39616         (get_string_length, format_character, format_string): Same.
39617         (format_plain, should_warn_p): New function.
39618         (maybe_warn): Call should_warn_p.  Update diagnostic messages
39619         and handle those for all directives, including plain strings.
39620         (format_directive): Handle likely and unlikely counters.
39621         Remove unnecessary quoting from diagnostics.  Add an informational
39622         note.
39623         (add_bytes): Remove.
39624         (pass_sprintf_length::compute_format_length): Simplify.
39625         (try_substitute_return_value): Handle likely and unlikely counters.
39627 2017-01-26  Carl Love  <cel@us.ibm.com>
39629         * config/rs6000/rs6000-c (altivec_overloaded_builtins): Remove
39630         bogus entries for the P8V_BUILTIN_VEC_VGBBD built-ins
39632 2017-01-26  Vladimir Makarov  <vmakarov@redhat.com>
39634         PR target/79131
39635         * lra-assigns.c (setup_live_pseudos_and_spill_after_risky): Take
39636         endianess for subregs into account.
39637         * lra-constraints.c (lra_constraints): Do risky transformations
39638         always on the first iteration.
39639         * lra-lives.c (check_pseudos_live_through_calls): Add arg
39640         last_call_used_reg_set.
39641         (process_bb_lives): Define and use last_call_used_reg_set.
39642         * lra.c (lra): Always continue after lra_constraints on the first
39643         iteration.
39645 2017-01-26  Kirill Yukhin  <kirill.yukhin@gmail.com>
39647         * gcc.target/i386/avx512bw-kshiftlq-2.c: Use unsigned long long
39648         constant.
39649         * gcc.target/i386/avx512bw-kshiftrq-2.c: Ditto.
39651 2017-01-26  Jakub Jelinek  <jakub@redhat.com>
39653         * config/i386/avx512fintrin.h (_ktest_mask16_u8,
39654         _ktestz_mask16_u8, _ktestc_mask16_u8, _kadd_mask16): Move to ...
39655         * config/i386/avx512dqintrin.h (_ktest_mask16_u8,
39656         _ktestz_mask16_u8, _ktestc_mask16_u8, _kadd_mask16): ... here.
39657         * config/i386/i386-builtin.def (__builtin_ia32_ktestchi,
39658         __builtin_ia32_ktestzhi, __builtin_ia32_kaddhi): Use
39659         OPTION_MASK_ISA_AVX512DQ instead of OPTION_MASK_ISA_AVX512F.
39660         * config/i386/sse.md (SWI1248_AVX512BWDQ2): New mode iterator.
39661         (kadd<mode>, ktest<mode>): Use it instead of SWI1248_AVX512BWDQ.
39663 2017-01-26  Marek Polacek  <polacek@redhat.com>
39665         PR c/79199
39666         * fold-const.c (operand_equal_p) [COND_EXPR]: Use OP_SAME_WITH_NULL
39667         for the third operand.
39669 2017-01-26  Jakub Jelinek  <jakub@redhat.com>
39671         PR middle-end/79236
39672         * omp-low.c (struct omp_context): Add simt_stmt field.
39673         (scan_omp_for): Return omp_context *.
39674         (scan_omp_simd): Set simt_stmt on the non-_simt_ SIMD
39675         context to the _simt_ SIMD stmt.
39676         (lower_omp_for): For combined SIMD with sibling _simt_
39677         SIMD, make sure to use the same decls in _looptemp_
39678         clauses as in the sibling.
39680 2017-01-26  David Sherwood  <david.sherwood@arm.com>
39682         PR middle-end/79212
39683         * gimplify.c (omp_notice_variable): Add GOVD_SEEN flag to variables in
39684         all contexts.
39686 2017-01-26  Jakub Jelinek  <jakub@redhat.com>
39688         PR target/70465
39689         * reg-stack.c (emit_swap_insn): Instead of fld a; fld b; fxchg %st(1);
39690         emit fld b; fld a; if possible.
39692         * brig-builtins.def: Update copyright years.
39693         * config/arm/arm_acle_builtins.def: Update copyright years.
39695 2017-01-25  Michael Meissner  <meissner@linux.vnet.ibm.com>
39697         PR target/79179
39698         * config/rs6000/vsx.md (vsx_extract_<mode>_store): Use wY
39699         constraint instead of o for the stxsd instruction.
39701 2017-01-25  Carl Love  <cel@us.ibm.com>
39703         * config/rs6000/rs6000-c (altivec_overloaded_builtins): Fix order
39704         of entries for ALTIVEC_BUILTIN_VEC_PACKS and P8V_BUILTIN_VEC_VGBBD.
39706 2017-01-25  Jonathan Wakely  <jwakely@redhat.com>
39708         * doc/invoke.texi (C++ Dialect Options): Fix typo.
39710 2017-01-25  Richard Biener  <rguenther@suse.de>
39712         PR tree-optimization/69264
39713         * target.def (vector_alignment_reachable): Improve documentation.
39714         * doc/tm.texi: Regenerate.
39715         * targhooks.c (default_builtin_vector_alignment_reachable): Simplify
39716         and add a comment.
39717         * tree-vect-data-refs.c (vect_supportable_dr_alignment): Revert
39718         earlier changes with respect to TYPE_USER_ALIGN.
39719         (vector_alignment_reachable_p): Likewise.  Improve dumping.
39721 2016-01-25  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
39723         PR target/79145
39724         * config/arm/arm.md (xordi3): Force constant operand into a register
39725         for TARGET_IWMMXT.
39727 2016-01-25  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
39729         * doc/invoke.texi (-fstore-merging): Correct default optimization
39730         levels at which it is enabled.
39731         (-O): Move -fstore-merging from list to...
39732         (-O2): ... Here.
39734 2017-01-25  Richard Biener  <rguenther@suse.de>
39736         PR debug/78363
39737         * omp-expand.c: Include debug.h.
39738         (expand_omp_taskreg): Make sure to generate early debug before
39739         outlining anything from a function.
39740         (expand_omp_target): Likewise.
39741         (grid_expand_target_grid_body): Likewise.
39743 2017-01-25  Maxim Ostapenko  <m.ostapenko@samsung.com>
39745         PR lto/79061
39746         * asan.c (get_translation_unit_decl): New function.
39747         (asan_add_global): Extract modules file name from globals
39748         TRANSLATION_UNIT_DECL name.
39750 2017-01-24  Eric Botcazou  <ebotcazou@adacore.com>
39752         PR target/77439
39753         * config/arm/arm.c (arm_function_ok_for_sibcall): Add back restriction
39754         for long calls with APCS frame and VFP.
39756 2017-01-24  David Malcolm  <dmalcolm@redhat.com>
39758         * cfg.c (original_copy_tables_initialized_p): New function.
39759         * cfg.h (original_copy_tables_initialized_p): New decl.
39760         * cfgrtl.c (relink_block_chain): Guard the call to
39761         free_original_copy_tables with a call to
39762         original_copy_tables_initialized_p.
39763         * cgraph.h (symtab_node::native_rtl_p): New decl.
39764         * cgraphunit.c (symtab_node::native_rtl_p): New function.
39765         (symtab_node::needed_p): Don't assert for early assembly output
39766         for __RTL functions.
39767         (cgraph_node::finalize_function): Set "force_output" for __RTL
39768         functions.
39769         (cgraph_node::analyze): Bail out early for __RTL functions.
39770         (analyze_functions): Update assertion to support __RTL functions.
39771         (cgraph_node::expand): Bail out early for __RTL functions.
39772         * final.c (rest_of_clean_state): Don't call delete_tree_ssa for
39773         __RTL functions.
39774         * function.h (struct function): Update comment for field
39775         "pass_startwith".
39776         * gimple-expr.c: Include "tree-pass.h".
39777         (gimple_has_body_p): Return false for __RTL functions.
39778         * Makefile.in (OBJS): Add run-rtl-passes.o.
39779         * pass_manager.h (gcc::pass_manager::get_rest_of_compilation): New
39780         accessor.
39781         (gcc::pass_manager::get_clean_slate): New accessor.
39782         * passes.c: Include "insn-addr.h".
39783         (should_skip_pass_p): Add logging.  Update logic for running
39784         "expand" to be compatible with both __GIMPLE and __RTL.  Guard
39785         property-provider override so it is only done for gimple passes.
39786         Don't skip dfinit.
39787         (skip_pass): New function.
39788         (execute_one_pass): Call skip_pass when skipping passes.
39789         * read-md.c (md_reader::read_char): Support filtering
39790         the input to a subset of line numbers.
39791         (md_reader::md_reader): Initialize fields
39792         m_first_line and m_last_line.
39793         (md_reader::read_file_fragment): New function.
39794         * read-md.h (md_reader::read_file_fragment): New decl.
39795         (md_reader::m_first_line): New field.
39796         (md_reader::m_last_line): New field.
39797         * read-rtl-function.c (function_reader::create_function): Only
39798         create cfun if it doesn't already exist.  Set PROP_rtl on cfun's
39799         curr_properties.  Set DECL_INITIAL to a dummy block.
39800         (read_rtl_function_body_from_file_range): New function.
39801         * read-rtl-function.h (read_rtl_function_body_from_file_range):
39802         New decl.
39803         * run-rtl-passes.c: New file.
39804         * run-rtl-passes.h: New file.
39806 2017-01-24  Jeff Law  <law@redhat.com>
39808         * config/microblaze/microblaze.h (ASM_FORMAT_PRIVATE_NAME): Increase
39809         buffer size.
39811 2017-01-24  Bin Cheng  <bin.cheng@arm.com>
39813         PR tree-optimization/79159
39814         * tree-ssa-loop-niter.c (get_cst_init_from_scev): New function.
39815         (record_nonwrapping_iv): Improve boundary using above function if no
39816         value range information.
39818 2017-01-24  Pekka Jääskeläinen  <pekka@parmance.com>
39819             Martin Jambor  <mjambor@suse.cz>
39821         * brig-builtins.def: New file.
39822         * builtins.def (DEF_HSAIL_BUILTIN): New macro.
39823         (DEF_HSAIL_ATOMIC_BUILTIN): Likewise.
39824         (DEF_HSAIL_SAT_BUILTIN): Likewise.
39825         (DEF_HSAIL_INTR_BUILTIN): Likewise.
39826         (DEF_HSAIL_CVT_ZEROI_SAT_BUILTIN): Likewise.
39827         * builtin-types.def (BT_INT8): New.
39828         (BT_INT16): Likewise.
39829         (BT_UINT8): Likewise.
39830         (BT_UINT16): Likewise.
39831         (BT_FN_ULONG): Likewise.
39832         (BT_FN_UINT_INT): Likewise.
39833         (BT_FN_UINT_ULONG): Likewise.
39834         (BT_FN_UINT_LONG): Likewise.
39835         (BT_FN_UINT_PTR): Likewise.
39836         (BT_FN_ULONG_PTR): Likewise.
39837         (BT_FN_INT8_FLOAT): Likewise.
39838         (BT_FN_INT16_FLOAT): Likewise.
39839         (BT_FN_UINT32_FLOAT): Likewise.
39840         (BT_FN_UINT16_FLOAT): Likewise.
39841         (BT_FN_UINT8_FLOAT): Likewise.
39842         (BT_FN_UINT64_FLOAT): Likewise.
39843         (BT_FN_UINT16_UINT32): Likewise.
39844         (BT_FN_UINT32_UINT16): Likewise.
39845         (BT_FN_UINT16_UINT16_UINT16): Likewise.
39846         (BT_FN_INT_PTR_INT): Likewise.
39847         (BT_FN_UINT_PTR_UINT): Likewise.
39848         (BT_FN_LONG_PTR_LONG): Likewise.
39849         (BT_FN_ULONG_PTR_ULONG): Likewise.
39850         (BT_FN_VOID_UINT64_UINT64): Likewise.
39851         (BT_FN_UINT8_UINT8_UINT8): Likewise.
39852         (BT_FN_INT8_INT8_INT8): Likewise.
39853         (BT_FN_INT16_INT16_INT16): Likewise.
39854         (BT_FN_INT_INT_INT): Likewise.
39855         (BT_FN_UINT_FLOAT_UINT): Likewise.
39856         (BT_FN_FLOAT_UINT_UINT): Likewise.
39857         (BT_FN_ULONG_UINT_UINT): Likewise.
39858         (BT_FN_ULONG_UINT_PTR): Likewise.
39859         (BT_FN_ULONG_ULONG_ULONG): Likewise.
39860         (BT_FN_UINT_UINT_UINT): Likewise.
39861         (BT_FN_VOID_UINT_PTR): Likewise.
39862         (BT_FN_UINT_UINT_PTR: Likewise.
39863         (BT_FN_UINT32_UINT64_PTR): Likewise.
39864         (BT_FN_INT_INT_UINT_UINT): Likewise.
39865         (BT_FN_UINT_UINT_UINT_UINT): Likewise.
39866         (BT_FN_UINT_UINT_UINT_PTR): Likewise.
39867         (BT_FN_UINT_ULONG_ULONG_UINT): Likewise.
39868         (BT_FN_ULONG_ULONG_ULONG_ULONG): Likewise.
39869         (BT_FN_LONG_LONG_UINT_UINT): Likewise.
39870         (BT_FN_ULONG_ULONG_UINT_UINT): Likewise.
39871         (BT_FN_VOID_UINT32_UINT64_PTR): Likewise.
39872         (BT_FN_VOID_UINT32_UINT32_PTR): Likewise.
39873         (BT_FN_UINT_UINT_UINT_UINT_UINT): Likewise.
39874         (BT_FN_UINT_FLOAT_FLOAT_FLOAT_FLOAT): Likewise.
39875         (BT_FN_ULONG_ULONG_ULONG_UINT_UINT): Likewise.
39876         * doc/frontends.texi: List BRIG FE.
39877         * doc/install.texi (Testing): Add BRIG tesring requirements.
39878         * doc/invoke.texi (Overall Options): Mention BRIG.
39879         * doc/standards.texi (Standards): Doucment BRIG HSA version.
39881 2017-01-24  Richard Biener  <rguenther@suse.de>
39883         PR translation/79208
39884         * ipa-devirt.c (odr_types_equivalent_p): Fix typo in diagnostic.
39886 2017-01-24  Martin Jambor  <mjambor@suse.cz>
39888         PR bootstrap/79198
39889         * ipa-prop.c (ipa_free_all_node_params): Call summary destructor.
39890         * ipa-prop.c (ipa_node_params_t::insert): Initialize fields known_csts
39891         and known_contexts.
39893 2017-01-24  Aldy Hernandez  <aldyh@redhat.com>
39895         PR middle-end/79123
39896         * gimple-ssa-warn-alloca.c (alloca_call_type): Make sure
39897         casts from signed to unsigned really don't have a range.
39899 2017-01-24  Markus Trippelsdorf  <markus@trippelsdorf.de>
39901         * gimple-ssa-sprintf.c (format_floating): Change MPFR_RNDx to
39902         GMP_RNDx for compatiblity.
39904 2017-01-24  Martin Liska  <mliska@suse.cz>
39906         PR bootstrap/79132
39907         * tree-ssa-reassoc.c (rewrite_expr_tree_parallel): Insert assert
39908         that would prevent us to call alloca with -1 as argument.
39910 2017-01-24  Jakub Jelinek  <jakub@redhat.com>
39912         * dwarf2out.c (output_compilation_unit_header, output_file_names):
39913         Avoid -Wformat-security warning.
39915 2017-01-23  Andrew Pinski  <apinski@cavium.com>
39917         * config/aarch64/aarch64.c (thunderx2t99_addrcost_table): Improve
39918         cost table.
39920 2017-01-23  Martin Sebor  <msebor@redhat.com>
39922         PR middle-end/78703
39923         * gimple-ssa-sprintf.c (warn_level): New global.
39924         (format_integer): Use it here and throughout the rest of the file.
39925         Use the same switch to compute sign as base.
39926         (maybe_warn): New function.
39927         (format_directive): Factor out warnings into maybe_warn.
39928         Add debugging output.  Use warn_level.
39929         (add_bytes): Use warn_level.
39930         (pass_sprintf_length::compute_format_length): Add debugging output.
39931         (try_substitute_return_value): Same.
39932         (pass_sprintf_length::handle_gimple_call): Set and use warn_level.
39934         PR middle-end/78703
39935         * gimple-ssa-sprintf.c (struct format_result): Remove constant member.
39936         (struct fmtresult, format_integer, format_floating): Adjust.
39937         (fmtresult::fmtresult): Set max correctly in two argument ctor.
39938         (get_string_length, format_string,format_directive): Same.
39939         (pass_sprintf_length::compute_format_length): Same.
39940         (try_substitute_return_value): Simplify slightly.
39942         PR middle-end/78703
39943         * gimple-ssa-sprintf.c (pass_sprintf_length::gate): Adjust formatting.
39944         (fmtresult::operator+=): Outlined.
39945         (struct fmtresult): Add ctors.
39946         (struct conversion_spec): Rename...
39947         (struct directive): ...to this.  Add and remove data members.
39948         (directive::set_width, directive::set_precision): New functions.
39949         (format_percent): Use fmtresult ctor.
39950         (get_width_and_precision): Remove.
39951         (format_integer): Make naming changes.  Avoid computing width and
39952         precision.
39953         (format_floating): Same.  Adjust indentation.
39954         (format_character, format_none): New functions.
39955         (format_string): Moved character handling to format_character.
39956         (format_directive): Remove arguments, change return type.
39957         (parse_directive): New function.
39958         (pass_sprintf_length::compute_format_length): Move directive
39959         parsing to parse_directive.
39961 2017-01-23  Jakub Jelinek  <jakub@redhat.com>
39963         * tree.h (assign_assembler_name_if_neeeded): Rename to ...
39964         (assign_assembler_name_if_needed): ... this.
39965         * tree.c (assign_assembler_name_if_neeeded): Rename to ...
39966         (assign_assembler_name_if_needed): ... this.
39967         (free_lang_data_in_cgraph): Adjust callers.
39968         * cgraphunit.c (cgraph_node::analyze): Likewise.
39969         * omp-expand.c (expand_omp_taskreg, expand_omp_target):
39970         Likewise.
39972 2017-01-23  Richard Biener  <rguenther@suse.de>
39974         PR tree-optimization/79088
39975         PR tree-optimization/79188
39976         * tree-ssa-threadupdate.c (mark_threaded_blocks): Move code
39977         resetting loop bounds after last path deletion.  Reset loop
39978         bounds of the target loop, make code match the comments.
39979         * tree-ssa-threadbackwards.c (pass_early_thread_jumps::execute):
39980         Make sure loops need no fixups.
39982 2017-01-23  Kelvin Nilsen  <kelvin@gcc.gnu.org>
39984         * config/rs6000/rs6000-builtin.def (VSIEDPF): Add scalar insert
39985         exponent support with double type for first argument.
39986         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Changed
39987         type returned by __builtin_vec_extract_sig,
39988         __builtin_vec_extract_sig_sp, and __builtin_vec_extract_sig_dp
39989         functions from "vector int" to "vector unsigned int" or from
39990         "vector long long int" to "vector unsigned long long int".
39991         Changed type returned by __builtin_vec_extract_exp,
39992         __builtin_vec_extract_exp_sp, and __builtin_vec_extract_exp_dp
39993         functions from "vector int" to "vector unsigned int" or from
39994         "vector long long int" to "vector unsigned long long int".
39995         Changed return type of __builtin_vec_test_data_class,
39996         __builtin_vec_test_data_class_sp, and
39997         __builtin_vec_test_data_class_dp from "vector int" to
39998         "vector bool int" or from "vector long long int" to "vector bool
39999         long long int" and changed second argument type from "unsigned
40000         int" to "int".  Added new overloaded function forms "vector float
40001         __builtin_vec_insert_exp (vector float, vector unsigned int)" and
40002         "vector float __builtin_vec_insert_exp_sp (vector float, vector
40003         unsigned int)" and "vector double __builtin_vec_insert_exp (vector
40004         double, vector unsigned long long int)" and "vector double
40005         __builtin_vec_insert_exp_dp (vector double, vector unsigned long
40006         long int)".  Changed return type of
40007         __builtin_scalar_test_data_class and
40008         __builtin_scalar_test_data_class_sp and
40009         __builtin_scalar_test_data_class_dp from "unsigned int" to "bool
40010         int" and changed second argument from "unsigned int" to "int".
40011         Changed type returned by __builtin_scalar_test_neg,
40012         __builtin_scalar_test_neg_sp, and __builtin_scalar_test_neg_dp
40013         from "int" to "bool int".  Added new overloaded function form
40014         "double __builtin_scalar_insert_exp (double, unsigned long long int)".
40015         * config/rs6000/vsx.md (xsiexpdpf): New insn for scalar insert
40016         exponent double-precision with floating point first argument.
40017         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Adjust
40018         documentation of scalar_test_data_class, scalar_test_neg,
40019         scalar_extract_sig, scalar_extract_exp, scalar_insert_exp,
40020         vector_extract_exp, vec_extract_sig, vec_insert_exp, and
40021         vec_test_data_class built-in functions to reflect refinements in
40022         their type signatures.
40024 2017-01-23  Andreas Tobler  <andreast@gcc.gnu.org>
40026         * config/aarch64/aarch64.c (aarch64_elf_asm_constructor): Increase
40027         size of buf.
40028         (aarch64_elf_asm_destructor): Likewise.
40030 2017-01-23  Bernd Schmidt  <bschmidt@redhat.com>
40032         PR rtl-optimization/78634
40033         * config/i386/i386.c (ix86_max_noce_ifcvt_seq_cost): New function.
40034         (TARGET_MAX_NOCE_IFCVT_SEQ_COST): Define.
40035         * ifcvt.c (noce_try_cmove): Add missing cost check.
40037         PR rtl-optimization/71724
40038         * combine.c (if_then_else_cond): Look for situations where it is
40039         beneficial to undo the work of one of the recursive calls.
40041 2017-01-23  Bin Cheng  <bin.cheng@arm.com>
40043         PR tree-optimization/70754
40044         * tree-predcom.c (stmt_combining_refs): New parameter INSERT_BEFORE.
40045         (reassociate_to_the_same_stmt): New parameter INSERT_BEFORE.  Insert
40046         combined stmt before it if not NULL.
40047         (combine_chains): Process refs reversely and compute dominance point
40048         for root ref.
40050 2017-01-23  Martin Liska  <mliska@suse.cz>
40052         PR tree-optimization/79196
40053         * tree-ssa-strlen.c (fold_strstr_to_memcmp): Rename to ...
40054         (fold_strstr_to_strncmp): ... this.  Fold the pattern to strncmp
40055         instead of memcmp.
40056         (strlen_optimize_stmt): Call the renamed function.
40058 2017-01-23  Michael Matz  <matz@suse.de>
40060         PR tree-optimization/78384
40061         * tree-ssa-loop-split.c (patch_loop_exit): Use correct edge.
40063 2017-01-23  Richard Biener  <rguenther@suse.de>
40065         PR tree-optimization/79186
40066         * tree-vrp.c (register_new_assert_for): Make sure we've seen
40067         both incoming edges before moving an assert.
40069 2017-01-23  Martin Jambor  <mjambor@suse.cz>
40071         * ipa-prop.c (load_from_param_1): Removed.
40072         (load_from_unmodified_param): Bits from load_from_param_1 put back
40073         here.
40074         (load_from_param): Removed.
40075         (compute_complex_assign_jump_func): Removed stmt2 and just replaced it
40076         with stmt.  Reverted back to use of load_from_unmodified_param.
40078 2017-01-23  Martin Jambor  <mjambor@suse.cz>
40080         PR ipa/79108
40081         * ipa-prop.h (ipa_param_descriptor): Anotate with with GTY(()).
40082         (ipa_node_params): Annotate with GTY((for_user)).  Make descriptors
40083         field a pointer to garbage collected vector, mark lattices and
40084         ipcp_orig_node with GTY((skip)).
40085         (ipa_get_param_count): Adjust to descriptors being a pointer.
40086         (ipa_get_param): Likewise.
40087         (ipa_get_type): Likewise.
40088         (ipa_get_param_move_cost): Likewise.
40089         (ipa_set_param_used): Likewise.
40090         (ipa_get_controlled_uses): Likewise.
40091         (ipa_set_controlled_uses): Likewise.
40092         (ipa_is_param_used): Likewise.
40093         (ipa_node_params_t): Move into garbage collector.  New methods insert
40094         and remove.
40095         (ipa_node_params_sum): Annotate wth GTY(()).
40096         (ipa_check_create_node_params): Adjust to ipa_node_params_sum being
40097         garbage collected.
40098         (ipa_load_from_parm_agg): Adjust declaration.
40099         * ipa-icf.c (param_used_p): Adjust to descriptors being a pointer.
40100         * ipa-profile.c (ipa_profile): Likewise.
40101         * ipa-prop.c (ipa_get_param_decl_index_1): Likewise.
40102         (ipa_populate_param_decls): Make descriptors parameter garbage
40103         collected.
40104         (ipa_dump_param): Adjust to descriptors being a pointer.
40105         (ipa_alloc_node_params): Likewise.
40106         (ipa_initialize_node_params): Likewise.
40107         (load_from_param_1): Make descriptors parameter garbage collected.
40108         (load_from_unmodified_param): Likewise.
40109         (load_from_param): Likewise.
40110         (ipa_load_from_parm_agg): Likewise.
40111         (ipa_node_params::~ipa_node_params): Removed.
40112         (ipa_free_all_node_params): Remove call to delete operator.
40113         (ipa_node_params_t::insert): New.
40114         (ipa_node_params_t::remove): Likewise.
40115         (ipa_node_params_t::duplicate): Adjust to descriptors being a pointer,
40116         copy known_csts and known_contexts vectors.
40117         (ipa_read_node_info): Adjust to descriptors being a pointer.
40118         (ipcp_modif_dom_walker): Make m_descriptors field garbage
40119         collected.
40120         (ipcp_transform_function): Make descriptors variable garbage
40121         collected.
40123 2017-01-23  Andrew Senkevich  <andrew.senkevich@intel.com>
40125         * config/i386/avx512bwintrin.h: Add k-mask test, kortest intrinsics.
40126         * config/i386/avx512dqintrin.h: Ditto.
40127         * config/i386/avx512fintrin.h: Ditto.
40128         * config/i386/i386.c: Handle new builtins.
40129         * config/i386/i386-builtin.def: Add new builtins.
40130         * config/i386/sse.md (ktest<mode>, kortest<mode>): New.
40131         (UNSPEC_KORTEST, UNSPEC_KTEST): New.
40133 2017-01-23  Jakub Jelinek  <jakub@redhat.com>
40134             Martin Liska  <mliska@suse.cz>
40136         * asan.h: Define ASAN_USE_AFTER_SCOPE_ATTRIBUTE.
40137         * asan.c (asan_expand_poison_ifn): Support stores and use
40138         appropriate ASAN report function.
40139         * internal-fn.c (expand_ASAN_POISON_USE): New function.
40140         * internal-fn.def (ASAN_POISON_USE): Declare.
40141         * tree-into-ssa.c (maybe_add_asan_poison_write): New function.
40142         (maybe_register_def): Create ASAN_POISON_USE when sanitizing.
40143         * tree-ssa-dce.c (eliminate_unnecessary_stmts): Remove
40144         ASAN_POISON calls w/o LHS.
40145         * tree-ssa.c (execute_update_addresses_taken): Create clobber
40146         for ASAN_MARK (UNPOISON, &x, ...) in order to prevent usage of a LHS
40147         from ASAN_MARK (POISON, &x, ...) coming to a PHI node.
40148         * gimplify.c (asan_poison_variables): Add attribute
40149         use_after_scope_memory to variables that really needs to live
40150         in memory.
40151         * tree-ssa.c (is_asan_mark_p): Do not rewrite into SSA when
40152         having the attribute.
40154 2017-01-23  Martin Liska  <mliska@suse.cz>
40156         * asan.c (create_asan_shadow_var): New function.
40157         (asan_expand_poison_ifn): Likewise.
40158         * asan.h (asan_expand_poison_ifn): New declaration.
40159         * internal-fn.c (expand_ASAN_POISON): Likewise.
40160         * internal-fn.def (ASAN_POISON): New builtin.
40161         * sanopt.c (pass_sanopt::execute): Expand
40162         asan_expand_poison_ifn.
40163         * tree-inline.c (copy_decl_for_dup_finish): Make function
40164         external.
40165         * tree-inline.h (copy_decl_for_dup_finish): Likewise.
40166         * tree-ssa.c (is_asan_mark_p): New function.
40167         (execute_update_addresses_taken): Rewrite local variables
40168         (identified just by use-after-scope as addressable) into SSA.
40170 2017-01-22  Gerald Pfeifer  <gerald@pfeifer.com>
40172         * doc/install.texi (Specific): opensource.apple.com uses https
40173         now. Remove trailing slash.
40175 2017-01-22  Gerald Pfeifer  <gerald@pfeifer.com>
40177         * README.Portability: Remove note on an Irix compatibility issue.
40179 2017-01-22  Dimitry Andric <dim@FreeBSD.org>
40181         * gcov.c (INCLUDE_ALGORITHM): Define.
40182         (INCLUDE_VECTOR): Define.
40183         No longer include <vector> and <algorithm> directly.
40185 2017-01-21  Gerald Pfeifer  <gerald@pfeifer.com>
40187         * doc/extend.texi (Thread-Local): Change www.akkadia.org reference
40188         to https.
40189         * doc/invoke.texi (Code Gen Options): Ditto.
40191 2017-01-21  Jan Hubicka  <hubicka@ucw.cz>
40193         PR lto/78407
40194         * cfg.c (update_bb_profile_for_threading): Fix updating of probablity.
40196 2017-01-21  Bernd Schmidt  <bschmidt@redhat.com>
40198         rtl-optimization/79125
40199         * cprop.c (local_cprop_pass): Handle cases where we make an
40200         unconditional trap.
40202 2017-01-20  Segher Boessenkool  <segher@kernel.crashing.org>
40204         PR target/61729
40205         PR target/77850
40206         * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Adjust address to
40207         read from, for big endian.
40209 2017-01-20  Jiong Wang  <jiong.wang@arm.com>
40211         * config/aarch64/aarch64-builtins.c (aarch64_init_builtins): Register
40212         register pauth builtins for LP64 only.
40214 2017-01-20  Marek Polacek  <polacek@redhat.com>
40216         PR c/79152
40217         * gimplify.c (should_warn_for_implicit_fallthrough): Handle consecutive
40218         non-case labels.
40220 2017-01-20  Alexander Monakov  <amonakov@ispras.ru>
40222         * omp-expand.c (expand_omp_simd): Clear PROP_gimple_lomp_dev regardless
40223         of safelen status.
40224         * omp-offload.c (pass_omp_device_lower::gate): Use PROP_gimple_lomp_dev.
40225         * passes.c (dump_properties): Handle PROP_gimple_lomp_dev.
40226         * tree-inline.c (expand_call_inline): Propagate PROP_gimple_lomp_dev.
40228 2017-01-20  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
40230         PR target/71270
40231         * config/arm/arm.c (neon_valid_immediate): Reject vector constants
40232         in big-endian mode when they are not a single duplicated value.
40234 2017-01-20  Richard Biener  <rguenther@suse.de>
40236         * BASE-VER: Bump to 7.0.1.
40238 2017-01-20  Alexander Monakov  <amonakov@ispras.ru>
40240         * omp-low.c (omplow_simd_context): New struct.  Use it...
40241         (lower_rec_simd_input_clauses): ...here and...
40242         (lower_rec_input_clauses): ...here to hold common data.  Adjust all
40243         references to idx, lane, max_vf, is_simt.
40245 2017-01-20  Graham Markall  <graham.markall@embecosm.com>
40247         * config/arc/arc.h (LINK_SPEC): Use arclinux_nps emulation when
40248         mcpu=nps400.
40250 2017-01-20  Martin Jambor  <mjambor@suse.cz>
40252         * hsa.h: Renaed to hsa-common.h.  Adjusted a comment.
40253         * hsa.c: Renaed to hsa-common.c.  Change include of gt-hsa.h to
40254         gt-hsa-common.h.
40255         * Makefile.in (OBJS): Rename hsa.o to hsa-common.o.
40256         (GTFILES): Rename hsa.c to hsa-common.c.
40257         * hsa-brig.c: Change include of hsa.h to hsa-common.h.
40258         * hsa-dump.c: Likewise.
40259         * hsa-gen.c: Likewise.
40260         * hsa-regalloc.c: Likewise.
40261         * ipa-hsa.c: Likewise.
40262         * omp-expand.c: Likewise.
40263         * omp-low.c: Likewise.
40264         * toplev.c: Likewise.
40266 2017-01-20  Marek Polacek  <polacek@redhat.com>
40268         PR c/64279
40269         * doc/invoke.texi: Document -Wduplicated-branches.
40270         * fold-const.c (operand_equal_p): Handle MODIFY_EXPR, INIT_EXPR,
40271         COMPOUND_EXPR, PREDECREMENT_EXPR, PREINCREMENT_EXPR,
40272         POSTDECREMENT_EXPR, POSTINCREMENT_EXPR, CLEANUP_POINT_EXPR, EXPR_STMT,
40273         STATEMENT_LIST, and RETURN_EXPR.  For non-pure non-const functions
40274         return 0 only when not OEP_LEXICOGRAPHIC.
40275         (fold_build_cleanup_point_expr): Use the expression
40276         location when building CLEANUP_POINT_EXPR.
40277         * tree-core.h (enum operand_equal_flag): Add OEP_LEXICOGRAPHIC.
40278         * tree.c (add_expr): Handle error_mark_node.
40280 2017-01-20  Martin Liska  <mliska@suse.cz>
40282         PR lto/69188
40283         * tree-profile.c (init_ic_make_global_vars): Do not call
40284         finalize_decl.
40285         (gimple_init_gcov_profiler): Likewise.
40287 2017-01-20  Martin Liska  <mliska@suse.cz>
40289         PR ipa/71190
40290         * cgraph.h (maybe_create_reference): Remove argument and
40291         update comment.
40292         * cgraphclones.c (cgraph_node::create_virtual_clone): Remove one
40293         argument.
40294         * ipa-cp.c (create_specialized_node): Likewise.
40295         * symtab.c (symtab_node::maybe_create_reference): Handle
40296         VAR_DECLs and ADDR_EXPRs and select ipa_ref_use type.
40298 2017-01-20  Martin Liska  <mliska@suse.cz>
40300         * read-rtl-function.c (function_reader::create_function): Use
40301         build_decl instread of build_decl_stat.
40303 2017-01-20  Andrew Senkevich  <andrew.senkevich@intel.com>
40305         * config/i386/avx512bwintrin.h: Add k-mask registers shift intrinsics.
40306         * config/i386/avx512dqintrin.h: Ditto.
40307         * config/i386/avx512fintrin.h: Ditto.
40308         * config/i386/i386-builtin-types.def: Add new types.
40309         * config/i386/i386.c: Handle new types.
40310         * config/i386/i386-builtin.def (__builtin_ia32_kshiftliqi)
40311         (__builtin_ia32_kshiftlihi, __builtin_ia32_kshiftlisi)
40312         (__builtin_ia32_kshiftlidi, __builtin_ia32_kshiftriqi)
40313         (__builtin_ia32_kshiftrihi, __builtin_ia32_kshiftrisi)
40314         (__builtin_ia32_kshiftridi): New.
40315         * config/i386/sse.md (k<code><mode>): Rename *k<code><mode>.
40317 2017-01-19  Segher Boessenkool  <segher@kernel.crashing.org>
40319         PR target/78875
40320         PR target/79140
40321         * config/rs6000/rs6000.c (TARGET_STACK_PROTECT_GUARD): Unconditionally
40322         define to rs6000_init_stack_protect_guard.
40323         (rs6000_init_stack_protect_guard): New function.
40325 2017-01-19  Matthew Fortune  <matthew.fortune@imgtec.com>
40326             Yunqiang Su  <yunqiang.su@imgtec.com>
40328         * config.gcc (supported_defaults): Add madd4.
40329         (with_madd4): Add validation.
40330         (all_defaults): Add madd4.
40331         * config/mips/mips.opt (mmadd4): New option.
40332         * config/mips/mips.h (OPTION_DEFAULT_SPECS): Add a default for
40333         mmadd4.
40334         (TARGET_CPU_CPP_BUILTINS): Add builtin_define for
40335         __mips_no_madd4.
40336         (ISA_HAS_UNFUSED_MADD4): Gate with mips_madd4.
40337         (ISA_HAS_FUSED_MADD4): Likewise.
40338         * doc/invoke.texi (-mmadd4): Document the new option.
40339         * doc/install.texi (--with-madd4): Document the new option.
40341 2017-01-19  Jiong Wang  <jiong.wang@arm.com>
40343         * config/aarch64/aarch64-builtins.c (enum aarch64_builtins): New
40344         entries for AARCH64_PAUTH_BUILTIN_XPACLRI,
40345         AARCH64_PAUTH_BUILTIN_PACIA1716, AARCH64_PAUTH_BUILTIN_AUTIA1716.
40346         (aarch64_init_pauth_hint_builtins): New.
40347         (aarch64_init_builtins): Call aarch64_init_pauth_hint_builtins.
40348         (aarch64_expand_builtin): Expand new builtins.
40350 2017-01-19  Jiong Wang  <jiong.wang@arm.com>
40352         * reg-notes.def (CFA_TOGGLE_RA_MANGLE): New reg-note.
40353         * combine-stack-adj.c (no_unhandled_cfa): Handle
40354         REG_CFA_TOGGLE_RA_MANGLE.
40355         * dwarf2cfi.c (dwarf2out_frame_debug): Handle REG_CFA_TOGGLE_RA_MANGLE.
40356         * config/aarch64/aarch64.c (aarch64_expand_prologue): Generates DWARF
40357         info for return address signing.
40358         (aarch64_expand_epilogue): Likewise.
40360 2017-01-19  Jiong Wang  <jiong.wang@arm.com>
40362         * config/aarch64/aarch64-opts.h (aarch64_function_type): New enum.
40363         * config/aarch64/aarch64-protos.h
40364         (aarch64_return_address_signing_enabled): New declaration.
40365         * config/aarch64/aarch64.c (aarch64_return_address_signing_enabled):
40366         New function.
40367         (aarch64_expand_prologue): Sign return address before it's pushed onto
40368         stack.
40369         (aarch64_expand_epilogue): Authenticate return address fetched from
40370         stack.
40371         (aarch64_override_options): Sanity check for ILP32 and ISA level.
40372         (aarch64_attributes): New function attributes for "sign-return-address".
40373         * config/aarch64/aarch64.md (UNSPEC_AUTI1716, UNSPEC_AUTISP,
40374         UNSPEC_PACI1716, UNSPEC_PACISP, UNSPEC_XPACLRI): New unspecs.
40375         ("*do_return"): Generate combined instructions according to key index.
40376         ("<pauth_mnem_prefix>sp", "<pauth_mnem_prefix1716", "xpaclri"): New.
40377         * config/aarch64/iterators.md (PAUTH_LR_SP, PAUTH_17_16): New integer
40378         iterators.
40379         (pauth_mnem_prefix, pauth_hint_num_a): New integer attributes.
40380         * config/aarch64/aarch64.opt (msign-return-address=): New.
40381         * doc/extend.texi (AArch64 Function Attributes): Documents
40382         "sign-return-address=".
40383         * doc/invoke.texi (AArch64 Options): Documents "-msign-return-address=".
40385 2017-01-19  Matthew Fortune  <matthew.fortune@imgtec.com>
40387         * doc/invoke.texi: Add missing -mlxc1-sxc1 options to
40388         overall option summary.
40390 2017-01-19  Jiong Wang  <jiong.wang@arm.com>
40392         * config/aarch64/aarch64-arches.def: New entry for "armv8.3-a".
40393         * config/aarch64/aarch64.h (AARCH64_FL_V8_3, AARCH64_FL_FOR_ARCH8_3,
40394         AARCH64_ISA_V8_3, TARGET_ARMV8_3): New.
40395         * doc/invoke.texi (AArch64 Options): Document "armv8.3-a".
40397 2017-01-19  Michael Meissner  <meissner@linux.vnet.ibm.com>
40399         * config/rs6000/rs6000-cpus.def (ISA_3_0_MASKS_SERVER): Enable
40400         -mpower9-minmax by default for -mcpu=power9.
40401         (ISA_3_MASKS_IEEE): Require -mvsx-small-integer to enable IEEE
40402         128-bit floating point.
40404 2017-01-20  Alan Modra  <amodra@gmail.com>
40406         * config/rs6000/rs6000.md (cmpstrnsi, cmpstrsi): Fail if
40407         optimizing for size.
40409 2017-01-20  Alan Modra  <amodra@gmail.com>
40411         PR target/79144
40412         * config/rs6000/rs6000.c (expand_strn_compare): Get the asm name
40413         for strcmp and strncmp from corresponding builtin decl.
40415 2017-01-19  Uros Bizjak  <ubizjak@gmail.com>
40417         * config.gcc (x86_64-*-rtems*): Use i386/rtemself.h
40418         instead of i386/rtems-64.h.
40419         * config/i386/rtems-64.h: Remove.
40421 2017-01-19  Uros Bizjak  <ubizjak@gmail.com>
40423         PR target/78478
40424         Revert:
40425         2013-11-05  Uros Bizjak  <ubizjak@gmail.com>
40427         * config/i386/rtemself.h (LONG_DOUBLE_TYPE_SIZE): New define.
40429 2017-01-19  Tamar Christina  <tamar.christina@arm.com>
40431         * config/aarch64/aarch64.c (aarch64_simd_gen_const_vector_dup):
40432         Change int to HOST_WIDE_INT.
40433         * config/aarch64/aarch64-protos.h
40434         (aarch64_simd_gen_const_vector_dup): Likewise.
40435         * config/aarch64/aarch64-simd.md: Add copysign<mode>3.
40437 2017-01-19  David Malcolm  <dmalcolm@redhat.com>
40439         * langhooks-def.h (lhd_type_for_size): New decl.
40440         (LANG_HOOKS_TYPE_FOR_SIZE): Define as lhd_type_for_size.
40441         * langhooks.c (lhd_type_for_size): New function, taken from
40442         lto_type_for_size.
40444 2017-01-19  Pat Haugen  <pthaugen@us.ibm.com>
40446         * config/rs6000/power9.md (power9-alu): Remove 'cmp' type and add
40447         define_bypass for CR latency.
40448         (power9-cracked-alu): Update bypass latency and remove power9-branch.
40449         (power9-alu2): Add define_bypass for CR latency.
40450         (power9-cmp): New.
40451         (power9-mul): Update insn latency.
40452         (power9-mul-compare): Update insn latency, bypass latency and remove
40453         power9-branch.
40455 2016-01-19  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
40457         * config/aarch64/aarch64-protos.h (aarch64_nopcrelative_literal_loads):
40458         Delete.
40459         * config/aarch64/aarch64.md
40460         (aarch64_reload_movcp<GPF_TF:mode><P:mode>): Delete reference to
40461         aarch64_nopcrelative_literal_loads.
40462         (aarch64_reload_movcp<VALL:mode><P:mode>): Likewise.
40464 2017-01-19  Chenghua Xu  <paul.hua.gm@gmail.com>
40466         * config/mips/mips.h (ISA_HAS_FUSED_MADD4): Enable for
40467         TARGET_LOONGSON_3A.
40468         (ISA_HAS_UNFUSED_MADD4): Exclude TARGET_LOONGSON_3A.
40470 2017-01-19  Doug Gilmore  <doug.gilmore@imgtec.com>
40472         PR target/78176
40473         * config.gcc (supported_defaults): Add lxc1-sxc1.
40474         (with_lxc1_sxc1): Add validation.
40475         (all_defaults): Add lxc1-sxc1.
40476         * config/mips/mips.opt (mlxc1-sxc1): New option.
40477         * config/mips/mips.h (OPTION_DEFAULT_SPECS): Add a default for
40478         mlxc1-sxc1.
40479         (TARGET_CPU_CPP_BUILTINS): Add builtin_define for
40480         __mips_no_lxc1_sxc1.
40481         (ISA_HAS_LXC1_SXC1): Gate with mips_lxc1_sxc1.
40482         * doc/invoke.texi (-mlxc1-sxc1): Document the new option.
40483         * doc/install.texi (--with-lxc1-sxc1): Document the new option.
40485 2017-01-19  Richard Biener  <rguenther@suse.de>
40487         PR tree-optimization/72488
40488         * tree-ssa-sccvn.c (run_scc_vn): When we abort the VN make
40489         sure to restore SSA info.
40490         * tree-ssa.c (verify_ssa): Verify SSA info is not shared.
40492 2017-01-19  Richard Earnshaw  <rearnsha@arm.com>
40494         PR rtl-optimization/79121
40495         * expr.c (expand_expr_real_2, case LSHIFT_EXPR): Look at the signedness
40496         of the inner type when shifting an extended value.
40498 2017-01-17  Jan Hubicka  <hubicka@ucw.cz>
40500         PR lto/78407
40501         * symtab.c (symtab_node::equal_address_to): Fix comparing of
40502         interposable aliases.
40504 2017-01-18  Peter Bergner  <bergner@vnet.ibm.com>
40506         PR target/78516
40507         * config/rs6000/spe.md (mov_si<mode>_e500_subreg0): Fix constraints.
40508         Use the evmergelohi instruction.
40509         (mov_si<mode>_e500_subreg4_2_le): Likewise.
40510         (mov_sitf_e500_subreg8_2_be): Likewise.
40511         (mov_sitf_e500_subreg12_2_le): Likewise.
40512         (mov_si<mode>_e500_subreg0_2_le): Fix constraints.
40513         (mov_si<mode>_e500_subreg4_2_be): Likewise.
40514         (mov_sitf_e500_subreg8_2_le): Likewise.
40515         (mov_sitf_e500_subreg12_2_be): Likewise.
40517 2017-01-18  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
40519         * config/rs6000/altivec.md (altivec_vbpermq): Change "type"
40520         attribute from vecsimple to vecperm.
40521         (altivec_vbpermq2): Likewise.
40523 2017-01-18  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
40525         PR target/79040
40526         * config/rs6000/altivec.h: Fix typo of vec_cntlz to vec_cnttz.
40528 2017-01-18  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
40529         * config/rs6000/rs6000-protos.h (expand_strn_compare): Add arg.
40530         * config/rs6000/rs6000.c (expand_strn_compare): Add ability to expand
40531         strcmp. Fix bug where comparison didn't stop with zero byte. Fix
40532         case where N arg is SIZE_MAX.
40533         * config/rs6000/rs6000.md (cmpstrnsi): Args to expand_strn_compare.
40534         (cmpstrsi): Add pattern.
40536 2017-01-18  Michael Meissner  <meissner@linux.vnet.ibm.com>
40538         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
40539         __builtin_vec_revb builtins.
40540         * config/rs6000/rs6000-builtins.def (P9V_BUILTIN_XXBRQ_V16QI): Add
40541         built-in functions to support generation of the ISA 3.0 XXBR<x>
40542         vector byte reverse instructions.
40543         (P9V_BUILTIN_XXBRQ_V1TI): Likewise.
40544         (P9V_BUILTIN_XXBRD_V2DI): Likewise.
40545         (P9V_BUILTIN_XXBRD_V2DF): Likewise.
40546         (P9V_BUILTIN_XXBGW_V4SI): Likewise.
40547         (P9V_BUILTIN_XXBGW_V4SF): Likewise.
40548         (P9V_BUILTIN_XXBGH_V8HI): Likewise.
40549         (P9V_BUILTIN_VEC_REVB): Likewise.
40550         * config/rs6000/vsx.md (p9_xxbrq_v1ti): New insns/expanders to
40551         generate the ISA 3.0 XXBR<x> vector byte reverse instructions.
40552         (p9_xxbrq_v16qi): Likewise.
40553         (p9_xxbrd_<mode>, VSX_D iterator): Likewise.
40554         (p9_xxbrw_<mode>, VSX_W iterator): Likewise.
40555         (p9_xxbrh_v8hi): Likewise.
40556         * config/rs6000/altivec.h (vec_revb): Define if ISA 3.0.
40557         * doc/extend.texi (RS/6000 Altivec Built-ins): Document the
40558         vec_revb built-in functions.
40560 2017-01-18  Uros Bizjak  <ubizjak@gmail.com>
40562         PR rtl-optimization/78952
40563         * config/i386/i386.md (any_extract): New code iterator.
40564         (*insvqi_2): Use any_extract for source operand.
40565         (*insvqi_3): Use any_shiftrt for source operand.
40567 2017-01-18  Wilco Dijkstra  <wdijkstr@arm.com>
40569         * config/aarch64/aarch64.c (aarch64_sched_adjust_priority)
40570         New function.
40571         (TARGET_SCHED_ADJUST_PRIORITY): Define target hook.
40573 2017-01-18  Matthias Klose  <doko@ubuntu.com>
40575         * doc/install.texi: Allow default for --with-target-bdw-gc-include.
40577 2016-01-18  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
40579         * config/rs6000/altivec.h (vec_bperm): Change #define.
40580         * config/rs6000/altivec.md (UNSPEC_VBPERMD): New enum constant.
40581         (altivec_vbpermq2): New define_insn.
40582         (altivec_vbpermd): Likewise.
40583         * config/rs6000/rs6000-builtin.def (VBPERMQ2): New monomorphic
40584         function interface.
40585         (VBPERMD): Likewise.
40586         (VBPERM): New polymorphic function interface.
40587         * config/rs6000/r6000-c.c (altivec_overloaded_builtins_table):
40588         Add entries for P9V_BUILTIN_VEC_VBPERM.
40589         * doc/extend.texi: Add interfaces for vec_bperm.
40591 2017-01-18  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
40593         * config/s390/s390-c.c (s390_expand_overloaded_builtin): Downcase
40594         first letter of error messages.
40595         (s390_resolve_overloaded_builtin): Likewise.
40596         * config/s390/s390.c (s390_expand_builtin): Likewise.
40597         (s390_invalid_arg_for_unprototyped_fn): Likewise.
40598         (s390_valid_target_attribute_inner_p): Likewise.
40599         * config/s390/s390.md ("tabort"): Likewise.
40601 2017-01-18  Toma Tabacu  <toma.tabacu@imgtec.com>
40603         * config/mips/mips.h (ISA_HAS_DIV3): Remove unused macro.
40604         (ISA_AVOID_DIV_HILO): New macro.
40605         (ISA_HAS_DIV): Use new ISA_AVOID_DIV_HILO macro.
40606         (ISA_HAS_DDIV): Likewise.
40608 2017-01-18  Markus Trippelsdorf  <markus@trippelsdorf.de>
40610         * doc/invoke.texi (fabi-version): Correct number of occurrences.
40612 2017-01-18  Markus Trippelsdorf  <markus@trippelsdorf.de>
40614         * doc/invoke.texi (fabi-version): Spelling fix.
40616 2017-01-18  Markus Trippelsdorf  <markus@trippelsdorf.de>
40618         PR c++/70182
40619         * doc/invoke.texi (fabi-version): Mention mangling fix for
40620         operator names.
40622 2017-01-18  Markus Trippelsdorf  <markus@trippelsdorf.de>
40624         PR c++/77489
40625         * doc/invoke.texi (fabi-version): Document discriminator mangling.
40627 2017-01-17  Segher Boessenkool  <segher@kernel.crashing.org>
40629         PR target/78875
40630         * config/rs6000/rs6000-opts.h (stack_protector_guard): New enum.
40631         * config/rs6000/rs6000.c (rs6000_option_override_internal): Handle
40632         the new options.
40633         * config/rs6000/rs6000.md (stack_protect_set): Handle the new more
40634         flexible settings.
40635         (stack_protect_test): Ditto.
40636         * config/rs6000/rs6000.opt (mstack-protector-guard=,
40637         mstack-protector-guard-reg=, mstack-protector-guard-offset=): New
40638         options.
40639         * doc/invoke.texi (Option Summary) [RS/6000 and PowerPC Options]:
40640         Add -mstack-protector-guard=, -mstack-protector-guard-reg=, and
40641         -mstack-protector-guard-offset=.
40642         (RS/6000 and PowerPC Options): Ditto.
40644 2017-01-17  Uros Bizjak  <ubizjak@gmail.com>
40646         * config/i386/i386.h (MASK_CLASS_P): New define.
40647         * config/i386/i386.c (inline_secondary_memory_needed): Ensure that
40648         there are no registers from different register sets also when
40649         mask registers are used.  Update function comment.
40650         * config/i386/i386.md (*movsi_internal): Split (*k/*krm) alternative
40651         to (*k/*r) and (*k/*km) alternatives.
40653 2017-01-17  Wilco Dijkstra  <wdijkstr@arm.com>
40655         * config/aarch64/aarch64.md (eh_return): Remove pattern and splitter.
40656         * config/aarch64/aarch64.h (AARCH64_EH_STACKADJ_REGNUM): Remove.
40657         (EH_RETURN_HANDLER_RTX): New define.
40658         * config/aarch64/aarch64.c (aarch64_frame_pointer_required):
40659         Force frame pointer in EH return functions.
40660         (aarch64_expand_epilogue): Add barrier for eh_return.
40661         (aarch64_final_eh_return_addr): Remove.
40662         (aarch64_eh_return_handler_rtx): New function.
40663         * config/aarch64/aarch64-protos.h (aarch64_final_eh_return_addr):
40664         Remove.
40665         (aarch64_eh_return_handler_rtx): New prototype.
40667 2017-01-17  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
40669         * config/rs6000/altivec.h (vec_rlmi): New #define.
40670         (vec_vrlnm): Likewise.
40671         (vec_rlnm): Likewise.
40672         * config/rs6000/altivec.md (UNSPEC_VRLMI): New UNSPEC enum value.
40673         (UNSPEC_VRLNM): Likewise.
40674         (VIlong): New mode iterator.
40675         (altivec_vrl<VI_char>mi): New define_insn.
40676         (altivec_vrl<VI_char>nm): Likewise.
40677         * config/rs6000/rs6000-builtin.def (VRLWNM): New monomorphic
40678         function entry.
40679         (VRLDNM): Likewise.
40680         (RLNM): New polymorphic function entry.
40681         (VRLWMI): New monomorphic function entry.
40682         (VRLDMI): Likewise.
40683         (RLMI): New polymorphic function entry.
40684         * config/rs6000/r6000-c.c (altivec_overloaded_builtin_table): Add
40685         new entries for P9V_BUILTIN_VEC_RLMI and P9V_BUILTIN_VEC_RLNM.
40686         * doc/extend.texi: Add description of vec_rlmi, vec_rlnm, and
40687         vec_vrlnm.
40689 2017-01-17  Jakub Jelinek  <jakub@redhat.com>
40691         PR debug/78839
40692         * dwarf2out.c (field_byte_offset): Restore the
40693         PCC_BITFIELD_TYPE_MATTERS behavior for INTEGER_CST DECL_FIELD_OFFSET
40694         and DECL_FIELD_BIT_OFFSET.  Use fold_build2 instead of build2 + fold.
40695         (analyze_variants_discr, gen_variant_part): Use fold_build2 instead
40696         of build2 + fold.
40698 2017-01-17  Eric Botcazou  <ebotcazou@adacore.com>
40700         PR ada/67205
40701         * config/aarch64/aarch64.c (TARGET_CUSTOM_FUNCTION_DESCRIPTORS): Define
40703 2017-01-17  Jakub Jelinek  <jakub@redhat.com>
40705         PR debug/71669
40706         * dwarf2out.c (add_data_member_location_attribute): For constant
40707         offset bitfield emit for -gdwarf-5 DW_AT_data_bit_offset attribute
40708         instead of DW_AT_data_member_location, DW_AT_bit_offset and
40709         DW_AT_byte_size attributes.
40711 2017-01-17  Eric Botcazou  <ebotcazou@adacore.com>
40713         * config/rs6000/rs6000.c (rs6000_emit_move): Also use a TOC reference
40714         after forcing to constant memory when the code model is medium.
40716 2017-01-17  Julia Koval  <julia.koval@intel.com>
40718         PR target/76731
40719         * config/i386/avx512fintrin.h
40720         (_mm512_i32gather_ps): Change __addr type to void const*.
40721         (_mm512_mask_i32gather_ps): Ditto.
40722         (_mm512_i32gather_pd): Ditto.
40723         (_mm512_mask_i32gather_pd): Ditto.
40724         (_mm512_i64gather_ps): Ditto.
40725         (_mm512_mask_i64gather_ps): Ditto.
40726         (_mm512_i64gather_pd): Ditto.
40727         (_mm512_mask_i64gather_pd): Ditto.
40728         (_mm512_i32gather_epi32): Ditto.
40729         (_mm512_mask_i32gather_epi32): Ditto.
40730         (_mm512_i32gather_epi64): Ditto.
40731         (_mm512_mask_i32gather_epi64): Ditto.
40732         (_mm512_i64gather_epi32): Ditto.
40733         (_mm512_mask_i64gather_epi32): Ditto.
40734         (_mm512_i64gather_epi64): Ditto.
40735         (_mm512_mask_i64gather_epi64): Ditto.
40736         (_mm512_i32scatter_ps): Change __addr type to void*.
40737         (_mm512_mask_i32scatter_ps): Ditto.
40738         (_mm512_i32scatter_pd): Ditto.
40739         (_mm512_mask_i32scatter_pd): Ditto.
40740         (_mm512_i64scatter_ps): Ditto.
40741         (_mm512_mask_i64scatter_ps): Ditto.
40742         (_mm512_i64scatter_pd): Ditto.
40743         (_mm512_mask_i64scatter_pd): Ditto.
40744         (_mm512_i32scatter_epi32): Ditto.
40745         (_mm512_mask_i32scatter_epi32): Ditto.
40746         (_mm512_i32scatter_epi64): Ditto.
40747         (_mm512_mask_i32scatter_epi64): Ditto.
40748         (_mm512_i64scatter_epi32): Ditto.
40749         (_mm512_mask_i64scatter_epi32): Ditto.
40750         (_mm512_i64scatter_epi64): Ditto.
40751         (_mm512_mask_i64scatter_epi64): Ditto.
40752         * config/i386/avx512pfintrin.h
40753         (_mm512_mask_prefetch_i32gather_pd): Change __addr type to void const*.
40754         (_mm512_mask_prefetch_i32gather_ps): Ditto.
40755         (_mm512_mask_prefetch_i64gather_pd): Ditto.
40756         (_mm512_mask_prefetch_i64gather_ps): Ditto.
40757         (_mm512_prefetch_i32scatter_pd): Change __addr type to void*.
40758         (_mm512_prefetch_i32scatter_ps): Ditto.
40759         (_mm512_mask_prefetch_i32scatter_pd): Ditto.
40760         (_mm512_mask_prefetch_i32scatter_ps): Ditto.
40761         (_mm512_prefetch_i64scatter_pd): Ditto.
40762         (_mm512_prefetch_i64scatter_ps): Ditto.
40763         (_mm512_mask_prefetch_i64scatter_pd): Ditto.
40764         (_mm512_mask_prefetch_i64scatter_ps): Ditto.
40765         * config/i386/avx512vlintrin.h
40766         (_mm256_mmask_i32gather_ps): Change __addr type to void const*.
40767         (_mm_mmask_i32gather_ps): Ditto.
40768         (_mm256_mmask_i32gather_pd): Ditto.
40769         (_mm_mmask_i32gather_pd): Ditto.
40770         (_mm256_mmask_i64gather_ps): Ditto.
40771         (_mm_mmask_i64gather_ps): Ditto.
40772         (_mm256_mmask_i64gather_pd): Ditto.
40773         (_mm_mmask_i64gather_pd): Ditto.
40774         (_mm256_mmask_i32gather_epi32): Ditto.
40775         (_mm_mmask_i32gather_epi32): Ditto.
40776         (_mm256_mmask_i32gather_epi64): Ditto.
40777         (_mm_mmask_i32gather_epi64): Ditto.
40778         (_mm256_mmask_i64gather_epi32): Ditto.
40779         (_mm_mmask_i64gather_epi32): Ditto.
40780         (_mm256_mmask_i64gather_epi64): Ditto.
40781         (_mm_mmask_i64gather_epi64): Ditto.
40782         (_mm256_i32scatter_ps): Change __addr type to void*.
40783         (_mm256_mask_i32scatter_ps): Ditto.
40784         (_mm_i32scatter_ps): Ditto.
40785         (_mm_mask_i32scatter_ps): Ditto.
40786         (_mm256_i32scatter_pd): Ditto.
40787         (_mm256_mask_i32scatter_pd): Ditto.
40788         (_mm_i32scatter_pd): Ditto.
40789         (_mm_mask_i32scatter_pd): Ditto.
40790         (_mm256_i64scatter_ps): Ditto.
40791         (_mm256_mask_i64scatter_ps): Ditto.
40792         (_mm_i64scatter_ps): Ditto.
40793         (_mm_mask_i64scatter_ps): Ditto.
40794         (_mm256_i64scatter_pd): Ditto.
40795         (_mm256_mask_i64scatter_pd): Ditto.
40796         (_mm_i64scatter_pd): Ditto.
40797         (_mm_mask_i64scatter_pd): Ditto.
40798         (_mm256_i32scatter_epi32): Ditto.
40799         (_mm256_mask_i32scatter_epi32): Ditto.
40800         (_mm_i32scatter_epi32): Ditto.
40801         (_mm_mask_i32scatter_epi32): Ditto.
40802         (_mm256_i32scatter_epi64): Ditto.
40803         (_mm256_mask_i32scatter_epi64): Ditto.
40804         (_mm_i32scatter_epi64): Ditto.
40805         (_mm_mask_i32scatter_epi64): Ditto.
40806         (_mm256_i64scatter_epi32): Ditto.
40807         (_mm256_mask_i64scatter_epi32): Ditto.
40808         (_mm_i64scatter_epi32): Ditto.
40809         (_mm_mask_i64scatter_epi32): Ditto.
40810         (_mm256_i64scatter_epi64): Ditto.
40811         (_mm256_mask_i64scatter_epi64): Ditto.
40812         (_mm_i64scatter_epi64): Ditto.
40813         (_mm_mask_i64scatter_epi64): Ditto.
40814         * config/i386/i386-builtin-types.def (V16SF_V16SF_PCFLOAT_V16SI_HI_INT)
40815         (V8DF_V8DF_PCDOUBLE_V8SI_QI_INT, V8SF_V8SF_PCFLOAT_V8DI_QI_INT)
40816         (V8DF_V8DF_PCDOUBLE_V8DI_QI_INT, V16SI_V16SI_PCINT_V16SI_HI_INT)
40817         (V8DI_V8DI_PCINT64_V8SI_QI_INT, V8SI_V8SI_PCINT_V8DI_QI_INT)
40818         (V8DI_V8DI_PCINT64_V8DI_QI_INT, V2DF_V2DF_PCDOUBLE_V4SI_QI_INT)
40819         (V4DF_V4DF_PCDOUBLE_V4SI_QI_INT, V2DF_V2DF_PCDOUBLE_V2DI_QI_INT)
40820         (V4DF_V4DF_PCDOUBLE_V4DI_QI_INT, V4SF_V4SF_PCFLOAT_V4SI_QI_INT)
40821         (V8SF_V8SF_PCFLOAT_V8SI_QI_INT, V4SF_V4SF_PCFLOAT_V2DI_QI_INT)
40822         (V4SF_V4SF_PCFLOAT_V4DI_QI_INT, V2DI_V2DI_PCINT64_V4SI_QI_INT)
40823         (V4DI_V4DI_PCINT64_V4SI_QI_INT, V2DI_V2DI_PCINT64_V2DI_QI_INT)
40824         (V4DI_V4DI_PCINT64_V4DI_QI_INT, V4SI_V4SI_PCINT_V4SI_QI_INT)
40825         (V8SI_V8SI_PCINT_V8SI_QI_INT, V4SI_V4SI_PCINT_V2DI_QI_INT)
40826         (V4SI_V4SI_PCINT_V4DI_QI_INT, VOID_PFLOAT_HI_V16SI_V16SF_INT)
40827         (VOID_PFLOAT_QI_V8SI_V8SF_INT, VOID_PFLOAT_QI_V4SI_V4SF_INT)
40828         (VOID_PDOUBLE_QI_V8SI_V8DF_INT, VOID_PDOUBLE_QI_V4SI_V4DF_INT)
40829         (VOID_PDOUBLE_QI_V4SI_V2DF_INT, VOID_PFLOAT_QI_V8DI_V8SF_INT)
40830         (VOID_PFLOAT_QI_V4DI_V4SF_INT, VOID_PFLOAT_QI_V2DI_V4SF_INT)
40831         (VOID_PDOUBLE_QI_V8DI_V8DF_INT, VOID_PDOUBLE_QI_V4DI_V4DF_INT)
40832         (VOID_PDOUBLE_QI_V2DI_V2DF_INT, VOID_PINT_HI_V16SI_V16SI_INT)
40833         (VOID_PINT_QI_V8SI_V8SI_INT, VOID_PINT_QI_V4SI_V4SI_INT)
40834         (VOID_PLONGLONG_QI_V8SI_V8DI_INT, VOID_PLONGLONG_QI_V4SI_V4DI_INT)
40835         (VOID_PLONGLONG_QI_V4SI_V2DI_INT, VOID_PINT_QI_V8DI_V8SI_INT)
40836         (VOID_PINT_QI_V4DI_V4SI_INT, VOID_PINT_QI_V2DI_V4SI_INT)
40837         (VOID_PLONGLONG_QI_V8DI_V8DI_INT, VOID_QI_V8SI_PCINT64_INT_INT)
40838         (VOID_PLONGLONG_QI_V4DI_V4DI_INT, VOID_PLONGLONG_QI_V2DI_V2DI_INT)
40839         (VOID_HI_V16SI_PCINT_INT_INT, VOID_QI_V8DI_PCINT64_INT_INT)
40840         (VOID_QI_V8DI_PCINT_INT_INT): Remove.
40841         (V16SF_V16SF_PCVOID_V16SI_HI_INT,  V8DF_V8DF_PCVOID_V8SI_QI_INT)
40842         (V8SF_V8SF_PCVOID_V8DI_QI_INT, V8DF_V8DF_PCVOID_V8DI_QI_INT)
40843         (V16SI_V16SI_PCVOID_V16SI_HI_INT, V8DI_V8DI_PCVOID_V8SI_QI_INT)
40844         (V8SI_V8SI_PCVOID_V8DI_QI_INT, V8DI_V8DI_PCVOID_V8DI_QI_INT)
40845         (VOID_PVOID_HI_V16SI_V16SF_INT, VOID_PVOID_QI_V8SI_V8DF_INT)
40846         (VOID_PVOID_QI_V8DI_V8SF_INT, VOID_PVOID_QI_V8DI_V8DF_INT)
40847         (VOID_PVOID_HI_V16SI_V16SI_INT, VOID_PVOID_QI_V8SI_V8DI_INT)
40848         (VOID_PVOID_QI_V8DI_V8SI_INT, VOID_PVOID_QI_V8DI_V8DI_INT)
40849         (V2DF_V2DF_PCVOID_V4SI_QI_INT, V4DF_V4DF_PCVOID_V4SI_QI_INT)
40850         (V2DF_V2DF_PCVOID_V2DI_QI_INT, V4DF_V4DF_PCVOID_V4DI_QI_INT
40851         (V4SF_V4SF_PCVOID_V4SI_QI_INT, V8SF_V8SF_PCVOID_V8SI_QI_INT)
40852         (V4SF_V4SF_PCVOID_V2DI_QI_INT, V4SF_V4SF_PCVOID_V4DI_QI_INT)
40853         (V2DI_V2DI_PCVOID_V4SI_QI_INT, V4DI_V4DI_PCVOID_V4SI_QI_INT)
40854         (V2DI_V2DI_PCVOID_V2DI_QI_INT, V4DI_V4DI_PCVOID_V4DI_QI_INT)
40855         (V4SI_V4SI_PCVOID_V4SI_QI_INT, V8SI_V8SI_PCVOID_V8SI_QI_INT)
40856         (V4SI_V4SI_PCVOID_V2DI_QI_INT, V4SI_V4SI_PCVOID_V4DI_QI_INT)
40857         (VOID_PVOID_QI_V8SI_V8SF_INT, VOID_PVOID_QI_V4SI_V4SF_INT)
40858         (VOID_PVOID_QI_V4SI_V4DF_INT, VOID_PVOID_QI_V4SI_V2DF_INT)
40859         (VOID_PVOID_QI_V4DI_V4SF_INT, VOID_PVOID_QI_V2DI_V4SF_INT)
40860         (VOID_PVOID_QI_V4DI_V4DF_INT, VOID_PVOID_QI_V2DI_V2DF_INT)
40861         (VOID_PVOID_QI_V8SI_V8SI_INT, VOID_PVOID_QI_V4SI_V4SI_INT)
40862         (VOID_PVOID_QI_V4SI_V4DI_INT, VOID_PVOID_QI_V4SI_V2DI_INT)
40863         (VOID_PVOID_QI_V4DI_V4SI_INT, VOID_PVOID_QI_V2DI_V4SI_INT)
40864         (VOID_PVOID_QI_V4DI_V4DI_INT, VOID_PVOID_QI_V2DI_V2DI_INT)
40865         (VOID_QI_V8SI_PCVOID_INT_INT, VOID_HI_V16SI_PCVOID_INT_INT)
40866         (VOID_QI_V8DI_PCVOID_INT_INT): Add.
40867         * config/i386/i386.c (ix86_init_mmx_sse_builtins): Adjust builtin
40868         definitions accordingly.
40870 2017-01-17  Kito Cheng  <kito.cheng@gmail.com>
40871             Kuan-Lin Chen  <kuanlinchentw@gmail.com>
40873         PR target/79079
40874         * internal-fn.c (expand_mul_overflow): Use convert_modes instead of
40875         gen_lowpart.
40877 2017-01-17  Vladimir Makarov  <vmakarov@redhat.com>
40879         PR target/79058
40880         * ira-conflicts.c (ira_build_conflicts): Update total conflict
40881         hard regs for inner regno.
40883 2017-01-17  Martin Liska  <mliska@suse.cz>
40885         PR ipa/71207
40886         * ipa-polymorphic-call.c (contains_type_p): Fix wrong
40887         assumption and add comment.
40889 2017-01-17  Nathan Sidwell  <nathan@acm.org>
40891         * ipa-visibility.c (localize_node): New function, broken out of ...
40892         (function_and_variable_visibility): ... here. Call it.
40894 2017-01-17  Jan Hubicka  <hubicka@ucw.cz>
40896         PR middle-end/77445
40897         * tree-ssa-threadupdate.c (remove_ctrl_stmt_and_useless_edges):
40898         correctly set frequency of oudgoing edge.
40899         (duplicate_thread_path): Fix profile updating.
40901 2017-01-17  Jakub Jelinek  <jakub@redhat.com>
40903         PR other/79046
40904         * configure.ac: Add GCC_BASE_VER.
40905         * Makefile.in (version): Use @get_gcc_base_ver@ instead of cat to get
40906         version from BASE-VER file.
40907         (CFLAGS-gcc.o): Add -DBASEVER=$(BASEVER_s).
40908         (gcc.o): Depend on $(BASEVER).
40909         * common.opt (dumpfullversion): New option.
40910         * gcc.c (driver_handle_option): Handle OPT_dumpfullversion.
40911         * doc/invoke.texi: Document -dumpfullversion.
40912         * doc/install.texi: Document --with-gcc-major-version-only.
40913         * configure: Regenerated.
40915 2017-01-17  Richard Biener  <rguenther@suse.de>
40917         PR tree-optimization/71433
40918         * tree-vrp.c (register_new_assert_for): Merge same asserts
40919         on all incoming edges.
40920         (process_assert_insertions_for): Handle insertions at the
40921         beginning of BBs.
40923 2017-01-17  Gerald Pfeifer  <gerald@pfeifer.com>
40925         * config/i386/cygwin.h (LIBGCJ_SONAME): Remove.
40926         * config/i386/mingw32.h (LIBGCJ_SONAME): Remove.
40928 2017-01-17  Kaz Kojima  <kkojima@gcc.gnu.org>
40930         PR target/78633
40931         * config/sh/sh.md (cmpeqsi_t+1): Call copy_rtx to avoid invalid
40932         RTL sharing.
40934 2017-01-17  Alan Modra  <amodra@gmail.com>
40936         PR target/79066
40937         * config/rs6000/rs6000.md (elf_high, elf_low): Disable when pic.
40938         * config/rs6000/rs6000.c (rs6000_emit_allocate_stack): Don't allow
40939         symbolic stack limit when pic.
40941 2017-01-16  Martin Sebor  <msebor@redhat.com>
40943         PR tree-optimization/78608
40944         * gimple-ssa-sprintf.c (tree_digits): Avoid negating TYPE_MIN.
40946 2017-01-16  Jeff Law  <law@redhat.com>
40948         Revert:
40949         2016-12-02  Tadek Kijkowski  <tkijkowski@gmail.com>
40950         * Makefile.in (PREPROCESSOR_DEFINES): Add a level of indirection
40951         for several include directories that may be relative to sysroot.
40952         * config/i386/x-mingw32 (gplus_includedir): Define.
40953         (gplus_tool_includedir, gplus_backward_include_dir): Likewise.
40954         (native_system_includedir): Likewise.
40955         * config/i386/mingw32.h (STANDARD_STARTFILE_PREFIX_1): Do not
40956         override if TARGET_SYSTEM_ROOT is defined.
40957         (NATIVE_SYSTEM_HEADER_DIR): Likewise.
40959         PR tree-optimization/79090
40960         PR tree-optimization/33562
40961         PR tree-optimization/61912
40962         PR tree-optimization/77485
40963         * tree-ssa-dse.c (compute_trims): Accept STMT argument.  Dump STMT
40964         and computed trims into the dump file.
40966 2017-01-17  Uros Bizjak  <ubizjak@gmail.com>
40968         * config/i386/i386.h (LIMIT_RELOAD_CLASS): Remove.
40970 2017-01-16  Jakub Jelinek  <jakub@redhat.com>
40972         PR c/79089
40973         * gimplify.c (gimplify_init_constructor): If want_value and
40974         object == lhs, unshare lhs to avoid invalid tree sharing.  Formatting
40975         fix.
40977         PR target/79080
40978         * loop-doloop.c (doloop_modify): Call unshare_all_rtl_in_chain on
40979         sequence.  Formatting fixes.
40980         (doloop_optimize): Formatting fixes.
40982         PR driver/49726
40983         * gcc.c (debug_level_greater_than_spec_func): New function.
40984         (static_spec_functions): Add debug-level-gt spec function.
40985         (ASM_DEBUG_SPEC, cpp_options): Use %:debug-level-gt(0) instead of
40986         !g0.
40987         * config/darwin.h (DSYMUTIL_SPEC, ASM_DEBUG_SPEC): Likewise.
40988         * config/darwin9.h (DSYMUTIL_SPEC, ASM_DEBUG_SPEC): Likewise.
40989         * common.opt (g, gcoff, gdwarf, gdwarf-, ggdb, gno-pubnames,
40990         gpubnames, ggnu-pubnames, gno-record-gcc-switches,
40991         grecord-gcc-switches, gno-strict-dwarf, gstrict-dwarf, gstabs,
40992         gstabs+, gtoggle, gvms, gxcoff, gxcoff+): Add Driver flag.
40994 2017-01-16  Uros Bizjak  <ubizjak@gmail.com>
40996         * config/i386/i386.h (HARD_REGNO_CALLER_SAVE_MODE): Apply HImode and
40997         QImode fixups to general and mask registers only.
40999 2017-01-16  Carl Love  <cel@us.ibm.com>
41001         * config/rs6000/rs6000-c (altivec_overloaded_builtins): Add support
41002         for built-in functions
41003         vector signed char vec_nabs (vector signed char)
41004         vector signed short vec_nabs (vector signed short)
41005         vector signed int vec_nabs (vector signed int)
41006         vector signed long long vec_nabs (vector signed long long)
41007         vector float vec_nabs (vector float)
41008         vector double vec_nabs (vector double)
41009         * config/rs6000/rs6000-builtin.def: Add definitions for NABS functions
41010         and NABS overload.
41011         * config/rs6000/altivec.md: New define_expand nabs<mode>2 types
41012         * config/rs6000/altivec.h: New define for vec_nabs built-in function.
41013         * doc/extend.texi: Update the documentation file for the new built-in
41014         functions.
41016 2017-01-16  Martin Sebor  <msebor@redhat.com>
41018         * gimple-ssa-sprintf.c (format_directive): Correct a typo in a warning
41019         message.
41021 2017-01-16  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
41023         * config/rs6000/rs6000.c (rtx_is_swappable_p): Change
41024         UNSPEC_VSX__XXSPLTD to require special splat handling.
41026 2017-01-16  David Malcolm  <dmalcolm@redhat.com>
41028         PR bootstrap/78616
41029         * system.h: Poison strndup.
41031 2017-01-16  Alan Modra  <amodra@gmail.com>
41033         PR target/79098
41034         * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Don't
41035         use a switch.
41037 2017-01-16  Georg-Johann Lay  <avr@gjlay.de>
41039         * config/avr/avr.h (BRANCH_COST) [reload_completed]: Increase by 4.
41041 2017-01-15  Uros Bizjak  <ubizjak@gmail.com>
41043         * config/i386/i386.c (ix86_legitimate_combined_insn): Do not
41044         call recog here.  Assert that INSN_CODE (insn) is non-negative.
41046 2017-01-15  Segher Boessenkool  <segher@kernel.crashing.org>
41048         PR target/72749
41049         * cfgrtl.c (rtl_split_edge): Also patch jump insns that jump to the
41050         fallthrough.
41051         * haifa-sched.c (dump_insn_stream): Don't crash if there is a label
41052         in the currently scheduled RTL fragment.
41054 2017-01-15  Segher Boessenkool  <segher@kernel.crashing.org>
41056         PR rtl-optimization/78751
41057         * ifcvt.c (find_cond_trap): If we generated a non-existing insn,
41058         give up.
41060 2017-01-14  Jeff Law  <law@redhat.com>
41062         PR tree-optimization/79090
41063         * tree-ssa-dse.c (valid_ao_ref_for_dse): Reject zero length and
41064         variable length stores.
41065         (compute_trims): Delete dead assignment to *trim_tail.
41066         (dse_dom_walker::dse_optimize_stmt): Optimize mem* calls with
41067         zero length.
41069 2017-01-14  Bernd Schmidt  <bschmidt@redhat.com>
41071         PR rtl-optimization/78626
41072         PR rtl-optimization/78727
41073         * cprop.c (one_cprop_pass): Collect unconditional traps in the middle
41074         of a block, and split such blocks after everything else is finished.
41076 2017-01-14  Alan Modra  <amodra@gmail.com>
41078         PR target/72749
41079         * combine.c (recog_for_combine_1): Set INSN_CODE before calling
41080         target legitimate_combined_insn.
41081         * config/rs6000/rs6000.c (TARGET_LEGITIMATE_COMBINED_INSN): Define.
41082         (rs6000_legitimate_combined_insn): New function.
41083         * config/rs6000/rs6000.md (UNSPEC_DOLOOP): Delete, and remove
41084         all uses.
41085         (ctr<mode>_internal3): Rename from *ctr<mode>_internal5.
41086         (ctr<mode>_internal4): Rename from *ctr<mode>_internal6.
41087         (ctr<mode>_internal1, ctr<mode>_internal2): Remove '*' from name.
41089 2017-01-14  Gerald Pfeifer  <gerald@pfeifer.com>
41091         * doc/frontends.texi (G++ and GCC): Remove references to Java.
41093 2017-01-13  Jeff Law  <law@redhat.com>
41095         PR tree-optimization/33562
41096         PR tree-optimization/61912
41097         PR tree-optimization/77485
41098         * tree-ssa-dse.c (delete_dead_call): Accept gsi rather than
41099         a statement.
41100         (delete_dead_assignment): Likewise.
41101         (dse_dom_walker::dse_optimize_stmt): Pass in the gsi rather than
41102         statement to delete_dead_call and delete_dead_assignment.
41104 2017-01-13  David Malcolm  <dmalcolm@redhat.com>
41106         PR c/78304
41107         * substring-locations.c (format_warning_va): Strengthen case 1 so
41108         that both endpoints of the substring must be within the format
41109         range for just the substring to be printed.
41111 2017-01-13  Uros Bizjak  <ubizjak@gmail.com>
41113         * config/i386/i386.opt (msgx): Use ix86_isa_flags2 variable.
41114         * config/i386/i386.c (ix86_target_string): Add missing options
41115         to isa_opts and reorder options by implied ISAs.  Rename isa_opts2 to
41116         isa2_opts, ix86_flag_opts to flag2_opts, ix86_target_other to
41117         flags_other and ix86_target_other to flags2_other.  Display unknown
41118         isa2 options.
41119         (ix86_valid_target_attribute_inner_p): Add missing options and
41120         reorder options by implied ISAs, as in ix86_target_string.
41122 2017-01-13  Richard Sandiford  <richard.sandiford@arm.com>
41124         * hash-table.h (hash_table::too_empty_p): New function.
41125         (hash_table::expand): Use it.
41126         (hash_table::traverse): Likewise.
41127         (hash_table::empty_slot): Use sizeof (value_type) instead of
41128         sizeof (PTR) to convert bytes to elements.  Shrink the table
41129         if the current size is excessive for the current number of
41130         elements.
41132 2017-01-13  Richard Sandiford  <richard.sandiford@arm.com>
41134         * ira-costs.c (record_reg_classes): Break from the inner loop
41135         early once alt_fail is known to be true.  Update outer loop
41136         handling accordingly.
41138 2017-01-13  Jeff Law  <law@redhat.com>
41140         * tree-ssa-dse.c (decrement_count): New function.
41141         (increment_start_addr, maybe_trim_memstar_call): Likewise.
41142         (dse_dom_walker::optimize_stmt): Call maybe_trim_memstar_call directly
41143         when we know the partially dead statement is a mem* function.
41145         PR tree-optimization/61912
41146         PR tree-optimization/77485
41147         * tree-ssa-dse.c: Include expr.h.
41148         (maybe_trim_constructor_store): New function.
41149         (maybe_trim_partially_dead_store): Call maybe_trim_constructor_store.
41151         PR tree-optimization/33562
41152         PR tree-optimization/61912
41153         PR tree-optimization/77485
41154         * doc/invoke.texi: Document new dse-max-object-size param.
41155         * params.def (PARM_DSE_MAX_OBJECT_SIZE): New PARAM.
41156         * tree-ssa-dse.c: Include params.h.
41157         (dse_store_status): New enum.
41158         (initialize_ao_ref_for_dse): New, partially extracted from
41159         dse_optimize_stmt.
41160         (valid_ao_ref_for_dse, normalize_ref): New.
41161         (setup_live_bytes_from_ref, compute_trims): Likewise.
41162         (clear_bytes_written_by, maybe_trim_complex_store): Likewise.
41163         (maybe_trim_partially_dead_store): Likewise.
41164         (maybe_trim_complex_store): Likewise.
41165         (dse_classify_store): Renamed from dse_possibly_dead_store_p.
41166         Track what bytes live from the original store.  Return tri-state
41167         for dead, partially dead or live.
41168         (dse_dom_walker): Add constructor, destructor and new private members.
41169         (delete_dead_call, delete_dead_assignment): New extracted from
41170         dse_optimize_stmt.
41171         (dse_optimize_stmt): Make a member of dse_dom_walker.
41172         Use initialize_ao_ref_for_dse.
41174         PR tree-optimization/33562
41175         PR tree-optimization/61912
41176         PR tree-optimization/77485
41177         * sbitmap.h (bitmap_count_bits): Prototype.
41178         (bitmap_clear_range, bitmap_set_range): Likewise.
41179         * sbitmap.c (bitmap_clear_range): New function.
41180         (bitmap_set_range, sbitmap_popcount, bitmap_count_bits): Likewise.
41182 2017-01-13  Martin Liska  <mliska@suse.cz>
41184         PR ipa/79043
41185         * function.c (set_cfun): Add new argument force.
41186         * function.h (set_cfun): Likewise.
41187         * ipa-inline-transform.c (inline_call): Use the function when
41188         strict alising from is dropped for function we inline to.
41190 2017-01-13  Richard Biener  <rguenther@suse.de>
41192         * tree-pretty-print.c (dump_generic_node): Fix inverted condition
41193         for dumping GIMPLE INTEGER_CSTs.
41195 2017-01-13  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
41197         * config/sol2.h (TARGET_OS_CPP_BUILTINS): Define __STDC_VERSION__
41198         to 201112L since C++17.
41200 2017-01-13  Maxim Ostapenko  <m.ostapenko@samsung.com>
41202         PR sanitizer/78887
41203         * asan.c (asan_needs_odr_indicator_p): Don't emit ODR indicators
41204         if -fsanitize=kernel-address is present.
41206 2017-01-13  Richard Biener  <rguenther@suse.de>
41208         * tree-pretty-print.c (dump_generic_node): Dump INTEGER_CSTs
41209         as _Literal ( type ) number in case usual suffixes do not
41210         preserve all information.
41212 2017-01-13  Richard Biener  <rguenther@suse.de>
41214         PR tree-optimization/77283
41215         * gimple-ssa-split-paths.c: Include gimple-ssa.h, tree-phinodes.h
41216         and ssa-iterators.h.
41217         (is_feasible_trace): Implement a cost model based on joiner
41218         PHI node uses.
41220 2017-01-12  Michael Meissner  <meissner@linux.vnet.ibm.com>
41222         PR target/79004
41223         * config/rs6000/rs6000.md (FP_ISA3): Do not optimize converting
41224         char or short to __float128/_Float128 directly.
41226 2017-01-12  Martin Sebor  <msebor@redhat.com>
41228         to -Wformat-overflow.
41229         * gimple-ssa-sprintf.c (pass_sprintf_length::gate): Adjust.
41230         (min_bytes_remaining): Same.
41231         (get_string_length): Same.
41232         (format_string): Same.
41233         (format_directive): Same.
41234         (add_bytes): Same.
41235         (pass_sprintf_length::handle_gimple_call): Same.
41237 2017-01-12  Jakub Jelinek  <jakub@redhat.com>
41239         * gimple-ssa-sprintf.c (try_substitute_return_value): Remove
41240         info.nowrite calls with no lhs that can't throw.  Return bool
41241         whether gsi_remove has been called or not.
41242         (pass_sprintf_length::handle_gimple_call): Return bool whether
41243         try_substitute_return_value called gsi_remove.  Formatting fix.
41244         (pass_sprintf_length::execute): Don't use gsi_remove if
41245         handle_gimple_call returned true.
41247         PR bootstrap/79069
41248         * cfgrtl.c (rtl_tidy_fallthru_edge): For any_uncondjump_p that can't
41249         be removed due to side-effects, don't remove following barrier nor
41250         turn the successor edge into fallthru edge.
41252 2017-01-12  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
41254         PR target/79044
41255         * config/rs6000/rs6000.c (insn_is_swappable_p): Mark
41256         element-reversing loads and stores as not swappable.
41258 2017-01-12  Nathan Sidwell  <nathan@acm.org>
41259             Nicolai Stange  <nicstange@gmail.com>
41261         * combine.c (try_combine): Don't ignore result of overlap checking
41262         loop.  Combine overlap & asm check into single loop.
41264 2017-01-12  Richard Biener  <rguenther@suse.de>
41266         * tree-pretty-print.c (dump_generic_node): Provide -gimple
41267         variant for MEM_REF.  Sanitize INTEGER_CST for -gimple.
41269 2017-01-12  Richard Biener  <rguenther@suse.de>
41271         * tree.c (initialize_tree_contains_struct): Make TS_OPTIMIZATION
41272         and TS_TARGET_OPTION directly derive from TS_BASE.
41273         * tree-core.h (tree_optimization_option): Derive from tree_base.
41274         (tree_target_option): Likewise.
41276 2017-01-11  Uros Bizjak  <ubizjak@gmail.com>
41278         * config/i386/i386.c (memory_address_length): Increase len
41279         only when rip_relative_addr_p returns false.
41281 2017-01-11  Julia Koval  <julia.koval@intel.com>
41283         * common/config/i386/i386-common.c (OPTION_MASK_ISA_SGX_UNSET): New.
41284         (OPTION_MASK_ISA_SGX_SET): New.
41285         (ix86_handle_option): Handle OPT_msgx.
41286         * config.gcc: Added sgxintrin.h.
41287         * config/i386/driver-i386.c (host_detect_local_cpu): Detect sgx.
41288         * config/i386/i386-c.c (ix86_target_macros_internal): Define __SGX__.
41289         * config/i386/i386.c (ix86_target_string): Add -msgx.
41290         (PTA_SGX): New.
41291         (ix86_option_override_internal): Handle new options.
41292         (ix86_valid_target_attribute_inner_p): Add sgx.
41293         * config/i386/i386.h (TARGET_SGX, TARGET_SGX_P): New.
41294         * config/i386/i386.opt: Add msgx.
41295         * config/i386/sgxintrin.h: New file.
41296         * config/i386/x86intrin.h: Add sgxintrin.h.
41298 2017-01-11  Jakub Jelinek  <jakub@redhat.com>
41300         PR c++/71537
41301         * fold-const.c (maybe_nonzero_address): Return 1 for function
41302         local objects.
41303         (tree_single_nonzero_warnv_p): Don't handle function local objects
41304         here.
41306         PR c++/72813
41307         * gcc.c (default_compilers): Don't add -o %g.s for -S -save-temps
41308         of c-header.
41310 2017-01-11  David Malcolm  <dmalcolm@redhat.com>
41312         PR driver/78877
41313         * opts.c: Include "spellcheck.h"
41314         (struct string_fragment): New struct.
41315         (struct edit_distance_traits<const string_fragment &>): New
41316         struct.
41317         (get_closest_sanitizer_option): New function.
41318         (parse_sanitizer_options): Offer suggestions for unrecognized arguments.
41320 2017-01-11  Jakub Jelinek  <jakub@redhat.com>
41322         * dwarf2out.c (DWARF_COMPILE_UNIT_HEADER_SIZE): For DWARF5 decrease
41323         by 12.
41324         (DWARF_COMDAT_TYPE_UNIT_HEADER_SIZE): Always
41325         DWARF_COMPILE_UNIT_HEADER_SIZE plus 12.
41326         (DWARF_COMPILE_UNIT_SKELETON_HEADER_SIZE): Define.
41327         (calc_base_type_die_sizes): Use DWARF_COMPILE_UNIT_SKELETON_HEADER_SIZE
41328         for initial die_offset if dwarf_split_debug_info.
41329         (output_comp_unit): Use DWARF_COMPILE_UNIT_SKELETON_HEADER_SIZE for
41330         initial next_die_offset if dwo_id is non-NULL.  Don't emit padding
41331         fields.
41332         (output_skeleton_debug_sections): Formatting fix.  Use
41333         DWARF_COMPILE_UNIT_SKELETON_HEADER_SIZE instead of
41334         DWARF_COMPILE_UNIT_HEADER_SIZE.  Don't emit padding.
41336 2017-01-11  Wilco Dijkstra  <wdijkstr@arm.com>
41338         * config/arm/cortex-a53.md: Add bypasses for
41339         cortex_a53_r2f_cvt.
41340         (cortex_a53_r2f): Only use for transfers.
41341         (cortex_a53_f2r): Likewise.
41342         (cortex_a53_r2f_cvt): Add reservation for conversions.
41343         (cortex_a53_f2r_cvt): Likewise.
41345 2017-01-11  Tamar Christina  <tamar.christina@arm.com>
41347         * config/arm/arm_neon.h: Add __artificial__ and gnu_inline
41348         to all inlined functions, change static to extern.
41350 2017-01-11  Christophe Lyon  <christophe.lyon@linaro.org>
41352         PR target/78253
41353         * config/arm/arm.c (legitimize_pic_address): Handle reference to
41354         weak symbol.
41355         (arm_assemble_integer): Likewise.
41357 2017-01-11  Richard Earnshaw  <rearnsha@arm.com>
41359         * config.gcc: Use new awk script to check CPU, FPU and architecture
41360         parameters for --with-... options.
41361         * config/arm/parsecpu.awk: New file
41362         * config/arm/arm-cpus.in: New file.
41363         * config/arm/arm-opts.h: Include arm-cpu.h instead of processing .def
41364         files.
41365         * config/arm/arm.c: Include arm-cpu-data.h instead of processing .def
41366         files.
41367         * config/arm/t-arm: Update dependency rules.
41368         * common/config/arm/arm-common.c: Include arm-cpu-cdata.h instead
41369         of processing .def files.
41370         * config/arm/genopt.sh: Deleted.
41371         * config/arm/gentune.sh: Deleted.
41372         * config/arm/arm-cores.def: Deleted.
41373         * config/arm/arm-arches.def: Deleted.
41374         * config/arm/arm-fpus.def: Deleted.
41375         * config/arm/arm-tune.md: Regenerated.
41376         * config/arm/arm-tables.opt: Regenerated.
41377         * config/arm/arm-cpu.h: New generated file.
41378         * config/arm/arm-cpu-data.h: New generated file.
41379         * config/arm/arm-cpu-cdata.h: New generated file.
41381 2017-01-11  Maxim Ostapenko  <m.ostapenko@samsung.com>
41383         PR lto/79042
41384         * lto-cgraph.c (lto_output_varpool_node): Pack dynamically_initialized
41385         bit.
41386         (input_varpool_node): Unpack dynamically_initialized bit.
41388 2017-01-11  Eric Botcazou  <ebotcazou@adacore.com>
41390         PR rtl-optimization/79032
41391         * lra-constraints.c (simplify_operand_subreg): In the MEM case, test
41392         the alignment of the adjusted memory reference against that of MODE,
41393         instead of the alignment of the original memory reference.
41395 2017-01-11  Martin Jambor  <mjambor@suse.cz>
41397         * hsa.c (hsa_callable_function_p): Revert addition of DECL_ARTIFICIAL
41398         test.
41399         * ipa-hsa.c (process_hsa_functions): Only duplicate non-artificial
41400         decorated functions.
41402 2017-01-11  Richard Biener  <rguenther@suse.de>
41404         * tree-vrp.c (evrp_dom_walker::before_dom_children): Also
41405         set range/nonnull info for PHI results.  Do not set it on
41406         stmts marked for removal.
41408 2017-01-10  Eric Botcazou  <ebotcazou@adacore.com>
41410         * expr.c (store_field): In the bitfield case, fetch the return value
41411         from the registers before applying a single big-endian adjustment.
41412         Always do a final load for a BLKmode value not larger than a word.
41414 2017-01-10  David Malcolm  <dmalcolm@redhat.com>
41416         PR c++/77949
41417         * input.c (selftest::test_accessing_ordinary_linemaps): Verify
41418         that we correctly handle column numbers greater than
41419         LINE_MAP_MAX_COLUMN_NUMBER.
41421 2017-01-10  Martin Sebor  <msebor@redhat.com>
41423         PR middle-end/78245
41424         * gimple-ssa-sprintf.c (get_destination_size): Call
41425         {init,fini}object_sizes.
41426         * tree-object-size.c (addr_object_size): Adjust.
41427         (pass_through_call): Adjust.
41428         (pass_object_sizes::execute): Adjust.
41429         * tree-object-size.h (fini_object_sizes): Declare.
41431 2017-01-10  Martin Sebor  <msebor@redhat.com>
41433         PR tree-optimization/78775
41434         * builtins.c (get_size_range): Move...
41435         * calls.c: ...to here.
41436         (alloc_max_size): Accept zero argument.
41437         (operand_signed_p): Remove.
41438         (maybe_warn_alloc_args_overflow): Call get_size_range.
41439         * calls.h (get_size_range): Declare.
41441 2017-01-10  Joe Seymour  <joe.s@somniumtech.com>
41443         * config/msp430/driver-msp430.c (msp430_mcu_data): Sync with data
41444         from TI's devices.csv file as of September 2016.
41445         * config/msp430/msp430.c (msp430_mcu_data): Likewise.
41447 2017-01-10  Sandra Loosemore  <sandra@codesourcery.com>
41449         * doc/extend.texi: Tweak formatting to fix overfull hbox warnings.
41450         * doc/invoke.texi: Likewise.
41451         * doc/md.texi: Likewise.
41452         * doc/objc.texi: Likewise.
41454 2017-01-10  Joshua Conner  <joshconner@google.com>
41456         * config/arm/fuchsia-elf.h: New file.
41457         * config/fuchsia.h: New file.
41458         * config.gcc (*-*-fuchsia*): Set native_system_header_dir.
41459         (aarch64*-*-fuchsia*, arm*-*-fuchsia*, x86_64-*-fuchsia*): Add to
41460         targets.
41461         * config.host: (aarch64*-*-fuchsia*, arm*-*-fuchsia*): Add to hosts.
41463 2016-01-10  Richard Biener  <rguenther@suse.de>
41465         PR tree-optimization/79034
41466         * tree-call-cdce.c (shrink_wrap_one_built_in_call_with_conds):
41467         Propagate out degenerate PHIs in the joiner.
41469 2017-01-10  Martin Liska  <mliska@suse.cz>
41471         * ipa-icf.c (sort_sem_items_by_decl_uid): New function.
41472         (sort_congruence_classes_by_decl_uid): Likewise.
41473         (sort_congruence_class_groups_by_decl_uid): Likewise.
41474         (sem_item_optimizer::merge_classes): Sort class, groups in these
41475         classes and members in the groups by DECL_UID of declarations.
41476         This would make merge operations stable.
41478 2017-01-10  Martin Liska  <mliska@suse.cz>
41480         * ipa-icf.c (sem_item_optimizer::sem_item_optimizer): Remove
41481         usage of m_classes_vec.
41482         (sem_item_optimizer::~sem_item_optimizer):  Likewise.
41483         (sem_item_optimizer::get_group_by_hash): Likewise.
41484         (sem_item_optimizer::subdivide_classes_by_equality): Likewise.
41485         (sem_item_optimizer::subdivide_classes_by_sensitive_refs): Likewise.
41486         (sem_item_optimizer::verify_classes): Likewise.
41487         (sem_item_optimizer::process_cong_reduction): Likewise.
41488         (sem_item_optimizer::dump_cong_classes): Likewise.
41489         (sem_item_optimizer::merge_classes): Likewise.
41490         * ipa-icf.h (congruence_class_hash): Rename from
41491         congruence_class_group_hash.  Remove declaration of m_classes_vec.
41493 2017-01-10  Andrew Senkevich  <andrew.senkevich@intel.com>
41495         * common/config/i386/i386-common.c (OPTION_MASK_ISA_AVX512VPOPCNTDQ_SET,
41496         OPTION_MASK_ISA_AVX512VPOPCNTDQ_UNSET): New.
41497         * config.gcc: Add avx512vpopcntdqintrin.h.
41498         * config/i386/avx512vpopcntdqintrin.h: New.
41499         * config/i386/cpuid.h (bit_AVX512VPOPCNTDQ): New.
41500         * config/i386/i386-builtin-types.def: Add new types.
41501         * config/i386/i386-builtin.def (__builtin_ia32_vpopcountd_v16si,
41502         __builtin_ia32_vpopcountd_v16si_mask, __builtin_ia32_vpopcountq_v8di,
41503         __builtin_ia32_vpopcountq_v8di_mask): New.
41504         * config/i386/i386-c.c (ix86_target_macros_internal): Define
41505         __AVX512VPOPCNTDQ__.
41506         * config/i386/i386.c (ix86_target_string): Add -mavx512vpopcntdq.
41507         (PTA_AVX512VPOPCNTDQ): Define.
41508         * config/i386/i386.h (TARGET_AVX512VPOPCNTDQ,
41509         TARGET_AVX512VPOPCNTDQ_P): Define.
41510         * config/i386/i386.opt: Add mavx512vpopcntdq.
41511         * config/i386/immintrin.h: Include avx512vpopcntdqintrin.h.
41512         * config/i386/sse.md (define_insn "vpopcount<mode><mask_name>"): New.
41514 2017-01-01  Jan Hubicka  <hubicka@ucw.cz>
41516         PR middle-end/77484
41517         * predict.def (PRED_CALL): Set to 67.
41519 2017-01-09  Eric Botcazou  <ebotcazou@adacore.com>
41521         * expr.c (store_field): In the bitfield case, if the value comes from
41522         a function call and is of an aggregate type returned in registers, do
41523         not modify the field mode; extract the value in all cases if the mode
41524         is BLKmode and the size is not larger than a word.
41526 2017-01-09  Dominique d'Humieres  <dominiq@lps.ens.fr>
41528         PR target/71017
41529         * config/i386/cpuid.h: Fix undefined behavior.
41531 2017-01-04  Jeff Law  <law@redhat.com>
41533         PR tree-optimization/79007
41534         PR tree-optimization/67955
41535         * tree-ssa-alias.c (same_addr_size_stores_p): Only need to be
41536         conservative for pt.null when flag_non_call_exceptions is on.
41538 2017-01-09  Jakub Jelinek  <jakub@redhat.com>
41540         PR translation/79019
41541         PR translation/79020
41542         * params.def (PARAM_INLINE_MIN_SPEEDUP,
41543         PARAM_IPA_CP_SINGLE_CALL_PENALTY,
41544         PARAM_USE_AFTER_SCOPE_DIRECT_EMISSION_THRESHOLD): Fix typos
41545         in descriptions.
41546         * config/avr/avr.opt (maccumulate-args): Likewise.
41547         * config/msp430/msp430.opt (mwarn-mcu): Likewise.
41548         * common.opt (freport-bug): Likewise.
41549         * cif-code.def (CIF_FINAL_ERROR): Likewise.
41550         * doc/invoke.texi (ipa-cp-single-call-penalty): Likewise.
41551         * config/s390/s390.c (s390_invalid_binary_op): Fix spelling in
41552         translatable string.
41553         * config/i386/i386.c (function_value_32): Likewise.
41554         * config/nios2/nios2.c (nios2_valid_target_attribute_rec): Likewise.
41555         * config/msp430/msp430.c (msp430_option_override, msp430_attr):
41556         Likewise.
41557         * config/msp430/driver-msp430.c (msp430_select_hwmult_lib): Likewise.
41558         * common/config/msp430/msp430-common.c (msp430_handle_option):
41559         Likewise.
41560         * symtab.c (symtab_node::verify_base): Likewise.
41561         * opts.c (set_debug_level): Likewise.
41562         * tree.c (verify_type_variant): Likewise.  Fix typo in comment.
41563         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Add
41564         missing whitespace to translatable strings.
41565         * config/avr/avr.md (bswapsi2): Fix typo in comment.
41566         * config/sh/superh.h: Likewise.
41567         * config/i386/xopintrin.h: Likewise.
41568         * config/i386/znver1.md: Likewise.
41569         * config/rs6000/rs6000.c (struct rs6000_opt_mask): Likewise.
41570         * ipa-inline-analysis.c (compute_inline_parameters): Likewise.
41571         * double-int.h (struct double_int): Likewise.
41572         * double-int.c (div_and_round_double): Likewise.
41573         * wide-int.cc: Likewise.
41574         * tree-ssa.c (non_rewritable_mem_ref_base): Likewise.
41575         * tree-ssa-sccvn.c (vn_reference_lookup_3): Likewise.
41576         * cfgcleanup.c (crossjumps_occured): Renamed to ...
41577         (crossjumps_occurred): ... this.
41578         (try_crossjump_bb, try_head_merge_bb, try_optimize_cfg, cleanup_cfg):
41579         Adjust all uses.
41581         PR tree-optimization/78899
41582         * tree-if-conv.c (version_loop_for_if_conversion): Instead of
41583         returning bool return struct loop *, NULL for failure and the new
41584         loop on success.
41585         (versionable_outer_loop_p): Don't version outer loop if it has
41586         dont_vectorized bit set.
41587         (tree_if_conversion): When versioning outer loop, ensure
41588         tree_if_conversion is performed also on the inner loop of the
41589         non-vectorizable outer loop copy.
41590         * tree-vectorizer.c (set_uid_loop_bbs): Formatting fix.  Fold
41591         LOOP_VECTORIZED in inner loop of the scalar outer loop and
41592         prevent vectorization of it.
41593         (vectorize_loops): For outer + inner LOOP_VECTORIZED, ensure
41594         the outer loop vectorization of the non-scalar version is attempted
41595         before vectorization of the inner loop in scalar version.  If
41596         outer LOOP_VECTORIZED guarded loop is not vectorized, prevent
41597         vectorization of its inner loop.
41598         * tree-vect-loop-manip.c (rename_variables_in_bb): If outer_loop
41599         has 2 inner loops, rename also on edges from bb whose single pred
41600         is outer_loop->header.  Fix typo in function comment.
41602 2017-01-09  Martin Sebor  <msebor@redhat.com>
41604         PR bootstrap/79033
41605         * asan.c (asan_emit_stack_protection): Increase local buffer size
41606         to avoid snprintf truncation warning.
41608 2017-01-09  Andrew Pinski  <apinski@cavium.com>
41610         * config/aarch64/aarch64-cores.def: Add thunderx2t99.  Change vulcan
41611         to reference thunderx2t99 for the tuning structure
41612         * config/aarch64/aarch64-cost-tables.h (vulcan_extra_costs):
41613         Rename to ...
41614         (thunderx2t99_extra_costs): This.
41615         * config/aarch64/aarch64-tune.md: Regenerate.
41616         * config/aarch64/aarch64.c (vulcan_addrcost_table): Rename to ...
41617         (vulcan_addrcost_table): This.
41618         (vulcan_regmove_cost): Rename to ...
41619         (thunderx2t99_regmove_cost): This.
41620         (vulcan_vector_cost): Rename to ...
41621         (thunderx2t99_vector_cost): this.
41622         (vulcan_branch_cost): Rename to ...
41623         (thunderx2t99_branch_cost): This.
41624         (vulcan_tunings): Rename to ...
41625         (thunderx2t99_tunings): This and s/vulcan/thunderx2t99 .
41626         * doc/invoke.texi (AARCH64/mtune): Add thunderx2t99.
41628 2017-01-09  Martin Jambor  <mjambor@suse.cz>
41630         PR ipa/78365
41631         PR ipa/78599
41632         * ipa-prop.h (ipa_jump_func): Swap positions of vr_known and m_vr.
41633         * ipa-cp.c (ipa_vr_operation_and_type_effects): New function.
41634         (propagate_vr_accross_jump_function): Use the above function for all
41635         value range computations for pass-through jump functions and type
41636         converasion from explicit value range values.
41637         (ipcp_propagate_stage): Do not attempt to deduce types of formal
41638         parameters from TYPE_ARG_TYPES.
41639         * ipa-prop.c (ipa_write_jump_function): Remove trailing whitespace.
41640         (ipa_write_node_info): Stream type of the actual argument.
41641         (ipa_read_node_info): Likewise. Also remove trailing whitespace.
41643 2017-01-09  Martin Liska  <mliska@suse.cz>
41645         PR pch/78970
41646         * gcc.c (driver_handle_option): Handle OPT_E and set have_E.
41647         (lookup_compiler): Do not show error message with have_E.
41649 2017-01-09  Jakub Jelinek  <jakub@redhat.com>
41651         PR tree-optimization/78938
41652         * tree-vect-stmts.c (vectorizable_condition): For non-masked COND_EXPR
41653         where comp_vectype is VECTOR_BOOLEAN_TYPE_P, use
41654         BIT_{NOT,XOR,AND,IOR}_EXPR on the comparison operands instead of
41655         {EQ,NE,GE,GT,LE,LT}_EXPR directly inside of VEC_COND_EXPR.  Formatting
41656         fixes.
41658 2017-01-09  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
41660         * tree-ssa-address.c (gen_addr_rtx): Don't handle index if it
41661         is const0_rtx.
41663 2017-01-09  Richard Biener  <rguenther@suse.de>
41665         PR tree-optimization/78997
41666         * tree-vect-slp.c (vect_mask_constant_operand_p): Handle SSA
41667         name condition properly.
41669 2017-01-09  Richard Biener  <rguenther@suse.de>
41671         PR debug/79000
41672         * dwarf2out.c (is_cxx): New overload with context.
41673         (is_naming_typedef_decl): Use it.
41675 2017-01-08  Sandra Loosemore  <sandra@codesourcery.com>
41677         * invoke.texi (Option Summary): Correct spacing in option lists
41678         and add line breaks to fix over-long lines.
41680 2017-01-08  Sandra Loosemore  <sandra@codesourcery.com>
41682         PR middle-end/17660
41684         * extend.texi (Common Variable Attributes): Add xref to GCC
41685         Internals manual to explain mode attribute keywords.
41687 2017-01-08  Sandra Loosemore  <sandra@codesourcery.com>
41689         PR other/16519
41690         * doc/invoke.texi (Option Summary): Move -pthread to Linker Options
41691         and Preprocessor Options.
41692         (Options for Linking): Document -pthread here....
41693         (RS/6000 and PowerPC Options): ...not here.
41694         (Solaris 2 Options): ...or here.
41695         * doc/cppopts.texi: Document -pthread.
41697 2017-01-08  Martin Sebor  <msebor@redhat.com>
41699         PR middle-end/77708
41700         * doc/invoke.texi (Warning Options): Document -Wformat-truncation.
41701         * gimple-ssa-sprintf.c (call_info::reval_used, call_info::warnopt):
41702         New member functions.
41703         (format_directive): Used them.
41704         (add_bytes): Same.
41705         (pass_sprintf_length::handle_gimple_call): Same.
41706         * graphite-sese-to-poly.c (tree_int_to_gmp): Increase buffer size
41707         to avoid truncation for any argument.
41708         (extract_affine_mul): Same.
41709         * tree.c (get_file_function_name): Same.
41711 2017-01-01  Jan Hubicka  <hubicka@ucw.cz>
41713         PR middle-end/77484
41714         * predict.def (PRED_INDIR_CALL): Set to 86.
41716 2017-01-07  Sandra Loosemore  <sandra@codesourcery.com>
41718         PR preprocessor/54124
41719         * doc/cppopts.texi: Reformat -d subtable to list the full name
41720         of the options.  Add cross-reference to the docs for the general
41721         compiler -d options.
41722         * doc/invoke.texi (Developer Options): Add cross-reference to the
41723         preprocessor-specific -d option documentation.
41725 2017-01-07  Sandra Loosemore  <sandra@codesourcery.com>
41727         PR preprocessor/13498
41728         * doc/cpp.texi (Search Path): Rewrite to remove obsolete and
41729         redudant material, and reflect new command-line options.
41730         (System Headers): Likewise.
41732 2017-01-07  Sandra Loosemore  <sandra@codesourcery.com>
41734         * doc/cppdiropts.texi: Merge documentation of -I, -iquote,
41735         -isystem, and -idirafter.  Copy-edit.
41736         * doc/cppopts.texi: Copy-edit.  Remove contradiction about
41737         default for -ftrack-macro-expansion.  Delete obsolete and
41738         badly-formatted implementation details about -fdebug-cpp output.
41739         * doc/cppwarnopts.texi: Copy-edit.
41741 2017-01-07  David Malcolm  <dmalcolm@redhat.com>
41743         PR c++/72803
41744         * input.c (selftest::test_accessing_ordinary_linemaps): Verify
41745         that the transition from a max line width >= 1<<10 to narrower
41746         lines works correctly.
41748 2017-01-07  Alexandre Oliva  <aoliva@redhat.com>
41750         * doc/options.texi (PerFunction): New.
41751         * opt-functions.awk (switch_flags): Map both Optimization and
41752         PerFunction to CL_OPTIMIZATION.
41753         * opth-gen.awk: Test for PerFunction flag along with
41754         Optimization.
41755         * optc-save-gen.awk: Likewise.  Introduce var_opt_hash and set
41756         it only when the latter is present.  Skip those that don't in
41757         the hash function generator.
41758         * common.opt (fvar-tracking): Mark as PerFunction instead of
41759         Optimization.
41760         (fvar-tracking-assignments): Likewise.
41761         (fvar-tracking-assignments-toggle): Likewise.
41762         (fvar-tracking-uninit): Likewise.
41764 2017-01-07  Jakub Jelinek  <jakub@redhat.com>
41766         PR translation/79018
41767         * params.def (PARAM_MAX_STORES_TO_MERGE): Add missing space between
41768         the and store.
41770 2017-01-06  Mikael Pettersson  <mikpelinux@gmail.com>
41772         PR target/57583
41773         * config/m68k/m68k.opt (LONG_JUMP_TABLE_OFFSETS): New option.
41774         * config/m68k/linux.h (ASM_RETURN_CASE_JUMP): Handle
41775         TARGET_LONG_JUMP_TABLE_OFFSETS.
41776         * config/m68k/m68kelf.h (ASM_RETURN_CASE_JUMP): Likewise.
41777         * config/m68k/netbsd-elf.h (ASM_RETURN_CASE_JUMP): Likewise.
41778         * config/m68k/m68k.h (CASE_VECTOR_MODE): Likewise.
41779         (ASM_OUTPUT_ADDR_DIFF_ELF): Likewise.
41780         * config/m68k/m68k.md (tablejump expander): Likewise.
41781         (*tablejump_pcrel_hi): Renamed from unnamed insn, reject
41782         TARGET_LONG_JUMP_TABLE_OFFSETS.
41783         (*tablejump_pcrel_si): New insn, handle TARGET_LONG_JUMP_TABLE_OFFSETS.
41784         * doc/invoke.texi (M68K options): Add -mlong-jump-table-offsets.
41786 2017-01-06  Edgar E. Iglesias <edgar.iglesias@xilinx.com>
41787             David Holsgrove <david.holsgrove@xilinx.com>
41789         * common/config/microblaze/microblaze-common.c
41790         (TARGET_EXCEPT_UNWIND_INFO): Remove.
41791         * config/microblaze/microblaze-protos.h (microblaze_eh_return):
41792         New prototype.
41793         * config/microblaze/microblaze.c (microblaze_must_save_register)
41794         (microblaze_expand_epilogue, microblaze_return_addr): Handle
41795         calls_eh_return.
41796         (microblaze_eh_return): New function.
41797         * config/microblaze/microblaze.h (RETURN_ADDR_OFFSET)
41798         (EH_RETURN_DATA_REGNO, MB_EH_STACKADJ_REGNUM)
41799         (EH_RETURN_STACKADJ_RTX, ASM_PREFERRED_EH_DATA_FORMAT): New macros.
41800         * config/microblaze/microblaze.md (eh_return): New pattern.
41802 2017-01-06  Jakub Jelinek  <jakub@redhat.com>
41804         * system.h (GCC_DIAGNOSTIC_PUSH_IGNORED, GCC_DIAGNOSTIC_POP,
41805         GCC_DIAGNOSTIC_STRINGIFY): Define.
41807         * read-rtl.c (rtx_reader::read_rtx_code): Avoid -Wsign-compare warning.
41809 2017-01-06  Andre Vieira  <andre.simoesdiasvieira@arm.com>
41811         * config/arm/arm.md (<mcrr>): New.
41812         (<mrrc>): New.
41813         * config/arm/arm.c (arm_arch5te): New.
41814         (arm_option_override): Set arm_arch5te.
41815         (arm_coproc_builtin_available): Add support for mcrr, mcrr2, mrrc
41816         and mrrc2.
41817         * config/arm/arm-builtins.c (MCRR_QUALIFIERS): Define to...
41818         (arm_mcrr_qualifiers): ... this. New.
41819         (MRRC_QUALIFIERS): Define to...
41820         (arm_mrrc_qualifiers): ... this. New.
41821         * config/arm/arm_acle.h (__arm_mcrr, __arm_mcrr2, __arm_mrrc,
41822         __arm_mrrc2): New.
41823         * config/arm/arm_acle_builtins.def (mcrr, mcrr2, mrrc, mrrc2): New.
41824         * config/arm/iterators.md (MCRRI, mcrr, MCRR): New.
41825         (MRRCI, mrrc, MRRC): New.
41826         * config/arm/unspecs.md (VUNSPEC_MCRR, VUNSPEC_MCRR2, VUNSPEC_MRRC,
41827         VUNSPEC_MRRC2): New.
41829 2017-01-06  Andre Vieira  <andre.simoesdiasvieira@arm.com>
41831         * config/arm/arm.md (<mcr>): New.
41832         (<mrc>): New.
41833         * config/arm/arm.c (arm_coproc_builtin_available): Add
41834         support for mcr, mrc, mcr2 and mrc2.
41835         * config/arm/arm-builtins.c (MCR_QUALIFIERS): Define to...
41836         (arm_mcr_qualifiers): ... this. New.
41837         (MRC_QUALIFIERS): Define to ...
41838         (arm_mrc_qualifiers): ... this. New.
41839         (MCR_QUALIFIERS): Define to ...
41840         (arm_mcr_qualifiers): ... this. New.
41841         * config/arm/arm_acle.h (__arm_mcr, __arm_mrc, __arm_mcr2,
41842         __arm_mrc2): New.
41843         * config/arm/arm_acle_builtins.def (mcr, mcr2, mrc, mrc2): New.
41844         * config/arm/iterators.md (MCRI, mcr, MCR, MRCI, mrc, MRC): New.
41845         * config/arm/unspecs.md (VUNSPEC_MCR, VUNSPEC_MCR2, VUNSPEC_MRC,
41846         VUNSPEC_MRC2): New.
41848 2017-01-06  Andre Vieira  <andre.simoesdiasvieira@arm.com>
41850         * config/arm/arm.md (*ldc): New.
41851         (*stc): New.
41852         (<ldc>): New.
41853         (<stc>): New.
41854         * config/arm/arm.c (arm_coproc_builtin_available): Add
41855         support for ldc,ldcl,stc,stcl,ldc2,ldc2l,stc2 and stc2l.
41856         (arm_coproc_ldc_stc_legitimate_address): New.
41857         * config/arm/arm-builtins.c (arm_type_qualifiers): Add
41858         'qualifier_const_pointer'.
41859         (LDC_QUALIFIERS): Define to...
41860         (arm_ldc_qualifiers): ... this. New.
41861         (STC_QUALIFIERS): Define to...
41862         (arm_stc_qualifiers): ... this. New.
41863         * config/arm/arm-protos.h
41864         (arm_coproc_ldc_stc_legitimate_address): New.
41865         * config/arm/arm_acle.h (__arm_ldc, __arm_ldcl, __arm_stc,
41866         __arm_stcl, __arm_ldc2, __arm_ldc2l, __arm_stc2, __arm_stc2l): New.
41867         * config/arm/arm_acle_builtins.def (ldc, ldc2, ldcl, ldc2l, stc,
41868         stc2, stcl, stc2l): New.
41869         * config/arm/constraints.md (Uz): New.
41870         * config/arm/iterators.md (LDCI, STCI, ldc, stc, LDC STC): New.
41871         * config/arm/unspecs.md (VUNSPEC_LDC, VUNSPEC_LDC2, VUNSPEC_LDCL,
41872         VUNSPEC_LDC2L, VUNSPEC_STC, VUNSPEC_STC2, VUNSPEC_STCL,
41873         VUNSPEC_STC2L): New.
41875 2017-01-06  Andre Vieira  <andre.simoesdiasvieira@arm.com>
41877         * config/arm/arm.md (<cdp>): New.
41878         * config/arm/arm.c (neon_const_bounds): Rename this ...
41879         (arm_const_bounds): ... this.
41880         (arm_coproc_builtin_available): New.
41881         * config/arm/arm-builtins.c (SIMD_MAX_BUILTIN_ARGS): Increase.
41882         (arm_type_qualifiers): Add 'qualifier_unsigned_immediate'.
41883         (CDP_QUALIFIERS): Define to...
41884         (arm_cdp_qualifiers): ... this. New.
41885         (void_UP): Define.
41886         (arm_expand_builtin_args): Add case for 6 arguments.
41887         * config/arm/arm-protos.h (neon_const_bounds): Rename this ...
41888         (arm_const_bounds): ... this.
41889         (arm_coproc_builtin_available): New.
41890         * config/arm/arm_acle.h (__arm_cdp): New.
41891         (__arm_cdp2): New.
41892         * config/arm/arm_acle_builtins.def (cdp): New.
41893         (cdp2): New.
41894         * config/arm/iterators.md (CDPI,CDP,cdp): New.
41895         * config/arm/neon.md: Rename all 'neon_const_bounds' to
41896         'arm_const_bounds'.
41897         * config/arm/types.md (coproc): New.
41898         * config/arm/unspecs.md (VUNSPEC_CDP, VUNSPEC_CDP2): New.
41899         * doc/extend.texi (ACLE): Add a mention of Coprocessor intrinsics.
41900         * doc/sourcebuild.texi (arm_coproc1_ok, arm_coproc2_ok,
41901         arm_coproc3_ok, arm_coproc4_ok): Document new effective targets.
41903 2017-01-06  Andre Vieira  <andre.simoesdiasvieira@arm.com>
41905         * config/arm/arm-builtins.c (arm_unsigned_binop_qualifiers): New.
41906         (UBINOP_QUALIFIERS): New.
41907         (si_UP): Define.
41908         (acle_builtin_data): New. Change comment.
41909         (arm_builtins): Remove ARM_BUILTIN_CRC32B, ARM_BUILTIN_CRC32H,
41910         ARM_BUILTIN_CRC32W, ARM_BUILTIN_CRC32CB, ARM_BUILTIN_CRC32CH,
41911         ARM_BUILTIN_CRC32CW. Add ARM_BUILTIN_ACLE_BASE and include
41912         arm_acle_builtins.def.
41913         (ARM_BUILTIN_ACLE_PATTERN_START): Define.
41914         (arm_init_acle_builtins): New.
41915         (CRC32_BUILTIN): Remove.
41916         (bdesc_2arg): Remove entries for crc32b, crc32h, crc32w,
41917         crc32cb, crc32ch and crc32cw.
41918         (arm_init_crc32_builtins): Remove.
41919         (arm_init_builtins): Use arm_init_acle_builtins rather
41920         than arm_init_crc32_builtins.
41921         (arm_expand_acle_builtin): New.
41922         (arm_expand_builtin): Use 'arm_expand_acle_builtin'.
41923         * config/arm/arm_acle_builtins.def: New.
41925 2017-01-06  Andre Vieira  <andre.simoesdiasvieira@arm.com>
41927         * config/arm/arm-builtins.c (neon_builtin_datum): Rename to ..
41928         (arm_builtin_datum): ... this.
41929         (arm_init_neon_builtin): Rename to ...
41930         (arm_init_builtin): ... this. Add a new parameters PREFIX
41931         and USE_SIG_IN_NAME.
41932         (arm_init_neon_builtins): Replace 'arm_init_neon_builtin' with
41933         'arm_init_builtin'. Replace type 'neon_builtin_datum' with
41934         'arm_builtin_datum'.
41935         (arm_init_vfp_builtins): Likewise.
41936         (builtin_arg): Rename enum's replacing 'NEON_ARG' with
41937         'ARG_BUILTIN' and add a 'ARG_BUILTIN_NEON_MEMORY.
41938         (arm_expand_neon_args): Rename to ...
41939         (arm_expand_builtin_args): ... this. Rename builtin_arg
41940         enum values and differentiate between ARG_BUILTIN_MEMORY
41941         and ARG_BUILTIN_NEON_MEMORY.
41942         (arm_expand_neon_builtin_1): Rename to ...
41943         (arm_expand_builtin_1): ... this. Rename builtin_arg enum
41944         values, arm_expand_builtin_args and add bool parameter NEON.
41945         (arm_expand_neon_builtin): Use arm_expand_builtin_1.
41946         (arm_expand_vfp_builtin): Likewise.
41947         (NEON_MAX_BUILTIN_ARGS): Remove, it was unused.
41949 2017-01-01  Jan Hubicka  <hubicka@ucw.cz>
41951         PR middle-end/77484
41952         * predict.def (PRED_POLYMORPHIC_CALL): Set to 59.
41953         * predict.c (tree_estimate_probability_bb): Reverse direction of
41954         polymorphic call predictor.
41956 2017-01-06  David Malcolm  <dmalcolm@redhat.com>
41958         * passes.c (execute_one_pass): Split out pass-skipping logic into...
41959         (determine_pass_name_match): ...this new function and...
41960         (should_skip_pass_p): ...this new function.
41962 2017-01-06  Nathan Sidwell  <nathan@acm.org>
41964         * ipa-visibility.c (function_and_variable_visibility): Reformat
41965         comments and long lines.  Remove extrneous if.
41966         * symtab.c (symtab_node::make_decl_local): Fix code format.
41967         (symtab_node::set_section_for_node): Fix comment typo.
41969 2017-01-06  Martin Liska  <mliska@suse.cz>
41971         PR bootstrap/79003
41972         * lra-constraints.c: Rename invariant to lra_invariant.
41973         * predict.c (set_even_probabilities): Initialize e to NULL.
41975 2017-01-05  Martin Sebor  <msebor@redhat.com>
41977         PR tree-optimization/78910
41978         * gimple-ssa-sprintf.c (tree_digits): Add an argument.
41979         (format_integer): Correct off-by-one error in the handling
41980         of precision with negative numbers in signed conversions..
41982 2017-01-05  Eric Botcazou  <ebotcazou@adacore.com>
41984         * doc/invoke.texi (C Dialect Options): Adjust -fsso-struct entry.
41986 2017-01-05  Jakub Jelinek  <jakub@redhat.com>
41988         PR tree-optimization/71016
41989         * tree-ssa-phiopt.c (tree_ssa_phiopt_worker): Pass cond_stmt to
41990         factor_out_conditional_conversion.  Formatting fix.
41991         (factor_out_conditional_conversion): Add cond_stmt argument.
41992         If arg1 is INTEGER_CST, punt if new_arg0 is not any operand of
41993         cond_stmt and if arg0_def_stmt is not the only stmt in its bb.
41994         Formatting fix.
41996 2017-01-05  David Malcolm  <dmalcolm@redhat.com>
41998         * Makefile.in (OBJS): Add read-md.o, read-rtl.o,
41999         read-rtl-function.o, and selftest-rtl.o.
42000         * config/aarch64/aarch64.c: Include selftest.h and selftest-rtl.h.
42001         (selftest::aarch64_test_loading_full_dump): New function.
42002         (selftest::aarch64_run_selftests): New function.
42003         (TARGET_RUN_TARGET_SELFTESTS): Wire it up to
42004         selftest::aarch64_run_selftests.
42005         * config/i386/i386.c
42006         (selftest::ix86_test_loading_dump_fragment_1): New function.
42007         (selftest::ix86_test_loading_call_insn): New function.
42008         (selftest::ix86_test_loading_full_dump): New function.
42009         (selftest::ix86_test_loading_unspec): New function.
42010         (selftest::ix86_run_selftests): Call the new functions.
42011         * emit-rtl.c (maybe_set_max_label_num): New function.
42012         * emit-rtl.h (maybe_set_max_label_num): New decl.
42013         * function.c (instantiate_decls): Guard call to
42014         instantiate_decls_1 with if (DECL_INITIAL (fndecl)).
42015         * function-tests.c (selftest::verify_three_block_rtl_cfg): Remove
42016         "static".
42017         * gensupport.c (gen_reader::gen_reader): Pass "false"
42018         for new "compact" param of rtx_reader.
42019         * print-rtl.c (rtx_writer::print_rtx_operand): Print "(nil)"
42020         rather than an empty string for NULL strings.
42021         * read-md.c: Potentially include config.h rather than bconfig.h.
42022         Wrap include of errors.h with #ifdef GENERATOR_FILE.
42023         (have_error): New global, copied from errors.c.
42024         (md_reader::read_name): Rename to...
42025         (md_reader::read_name_1): ...this, adding "out_loc" param,
42026         and converting "missing name or number" to returning false, rather
42027         than failing.
42028         (md_reader::read_name): Reimplement in terms of read_name_1.
42029         (md_reader::read_name_or_nil): New function.
42030         (md_reader::read_string): Handle "(nil)" by returning NULL.
42031         (md_reader::md_reader): Add new param "compact".
42032         (md_reader::read_md_files): Wrap with #ifdef GENERATOR_FILE.
42033         (md_reader::read_file): New method.
42034         * read-md.h (md_reader::md_reader): Add new param "compact".
42035         (md_reader::read_file): New method.
42036         (md_reader::is_compact): New accessor.
42037         (md_reader::read_name): Convert return type from void to file_location.
42038         (md_reader::read_name_or_nil): New decl.
42039         (md_reader::read_name_1): New decl.
42040         (md_reader::m_compact): New field.
42041         (noop_reader::noop_reader): Pass "false" for new "compact" param
42042         of rtx_reader.
42043         (rtx_reader::rtx_reader): Add new "compact" param.
42044         (rtx_reader::read_rtx_operand): Make virtual and convert return
42045         type from void to rtx.
42046         (rtx_reader::read_until): New decl.
42047         (rtx_reader::handle_any_trailing_information): New virtual function.
42048         (rtx_reader::postprocess): New virtual function.
42049         (rtx_reader::finalize_string): New virtual function.
42050         (rtx_reader::m_in_call_function_usage): New field.
42051         (rtx_reader::m_reuse_rtx_by_id): New field.
42052         * read-rtl-function.c: New file.
42053         * selftest-rtl.c (selftest::assert_rtx_ptr_eq_at): New function.
42054         * selftest-rtl.h (ASSERT_RTX_PTR_EQ): New macro.
42055         (selftest::verify_three_block_rtl_cfg): New decl.
42056         * read-rtl-function.h: New file.
42057         * read-rtl.c: Potentially include config.h rather than bconfig.h.
42058         For host, include function.h, memmodel.h, and emit-rtl.h.
42059         (one_time_initialization): New function.
42060         (struct compact_insn_name): New struct.
42061         (compact_insn_names): New array.
42062         (find_code): Handle insn codes in compact dumps.
42063         (apply_subst_iterator): Wrap with #ifdef GENERATOR_FILE.
42064         (bind_subst_iter_and_attr): Likewise.
42065         (add_condition_to_string): Likewise.
42066         (add_condition_to_rtx): Likewise.
42067         (apply_attribute_uses): Likewise.
42068         (add_current_iterators): Likewise.
42069         (apply_iterators): Likewise.
42070         (initialize_iterators): Guard usage of apply_subst_iterator with
42071         #ifdef GENERATOR_FILE.
42072         (read_conditions): Wrap with #ifdef GENERATOR_FILE.
42073         (md_reader::read_mapping): Likewise.
42074         (add_define_attr_for_define_subst): Likewise.
42075         (add_define_subst_attr): Likewise.
42076         (read_subst_mapping): Likewise.
42077         (check_code_iterator): Likewise.
42078         (rtx_reader::read_rtx): Likewise.  Move one-time initialization
42079         logic to...
42080         (one_time_initialization): New function.
42081         (rtx_reader::read_until): New method.
42082         (read_flags): New function.
42083         (parse_reg_note_name): New function.
42084         (rtx_reader::read_rtx_code): Initialize "iterator" to NULL.
42085         Handle reuse_rtx ids.
42086         Wrap iterator lookup within #ifdef GENERATOR_FILE.
42087         Add parsing support for RTL dumps, mirroring the special-cases in
42088         print_rtx, by calling read_flags, reading REG_NOTE names, INSN_UID
42089         values, and calling handle_any_trailing_information.
42090         (rtx_reader::read_rtx_operand): Convert return type from void
42091         to rtx, returning return_rtx.  Handle case 'e'.  Call
42092         finalize_string on XSTR and XTMPL fields.
42093         (rtx_reader::read_nested_rtx):  Handle dumps in which trailing
42094          "(nil)" values were omitted.  Call the postprocess vfunc on the
42095         return_rtx.
42096         (rtx_reader::rtx_reader): Add new "compact" param and pass to base
42097         class ctor.  Initialize m_in_call_function_usage.  Call
42098         one_time_initialization.
42099         * rtl-tests.c (selftest::test_uncond_jump): Call
42100         set_new_first_and_last_insn.
42101         * rtl.h (read_rtx): Wrap decl with #ifdef GENERATOR_FILE.
42102         * selftest-rtl.c: New file.
42103         * selftest-rtl.h (class selftest::rtl_dump_test): New class.
42104         (selftest::get_insn_by_uid): New decl.
42105         * selftest-run-tests.c (selftest::run_tests): Call
42106         read_rtl_function_c_tests.
42107         * selftest.h  (selftest::read_rtl_function_c_tests): New decl.
42108         * tree-dfa.c (ssa_default_def): Return NULL_TREE for rtl function
42109         dumps.
42111 2017-01-05  Uros Bizjak  <ubizjak@gmail.com>
42113         * config/i386/i386.md (*testqi_ext_3): No need to handle memory
42114         operands in a special way.  Assert that pos+len <= mode precision.
42116 2017-01-05  Jakub Jelinek  <jakub@redhat.com>
42118         * common.opt (fvect-cost-model): Remove RejectNegative flag, use
42119         3 argument Alias with unlimited for the negative form.
42120         (fno-vect-cost-model): Removed.
42122 2017-01-05  Martin Liska  <mliska@suse.cz>
42124         * hsa-gen.c (gen_hsa_divmod): New function.
42125         (gen_hsa_insn_for_internal_fn_call): Use the function for IFN_DIVMOD.
42127 2017-01-05  Martin Liska  <mliska@suse.cz>
42129         PR pch/78970
42130         * gcc.c (lookup_compiler): Reject '-' filename for a precompiled
42131         header.
42133 2017-01-05  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
42135         * config/s390/s390.c (s390_expand_setmem): Unroll the loop for
42136         small constant length operands.
42138 2017-01-05  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
42140         * config/s390/s390.c (s390_expand_setmem): Avoid overlapping bytes
42141         between loop iterations.
42143 2017-01-05  Martin Liska  <mliska@suse.cz>
42145         PR sanitizer/78815
42146         * gimplify.c (gimplify_decl_expr): Compare to
42147         asan_poisoned_variables instread of checking flags.
42148         (gimplify_target_expr): Likewise.
42149         (gimplify_expr): Likewise.
42150         (gimplify_function_tree): Conditionally initialize
42151         asan_poisoned_variables.
42153 2017-01-04  Jeff Law  <law@redhat.com>
42155         PR tree-optimizatin/78812
42156         * rtl.h (contains_mem_rtx_p): Prototype.
42157         * ifcvt.c (containts_mem_rtx_p): Move from here to...
42158         * rtlanal.c (contains_mem_rtx_p): Here and remove static linkage.
42159         * gcse.c (prune_expressions): Use contains_mem_rtx_p to discover
42160         and prune MEMs that are not at the toplevel of a SET_SRC rtx.  Look
42161         through ZERO_EXTEND and SIGN_EXTEND when trying to avoid pruning MEMs.
42163 2017-01-04  Alexandre Oliva  <aoliva@redhat.com>
42165         * input.c (assert_char_at_range): Default-initialize actual_range.
42167 2017-01-04  Alexandre Oliva  <aoliva@redhat.com>
42169         * df-scan.c (df_ref_create_structure): Make regno unsigned,
42170         to match the caller.
42172 2017-01-04  Alexandre Oliva  <aoliva@redhat.com>
42174         * cfgexpand.c (expand_gimple_basic_block): Disregard debug
42175         insns after final jump in test to emit dummy move.
42177 2017-01-04  Alexandre Oliva  <aoliva@redhat.com>
42179         * gimple-iterator.h (gsi_one_nondebug_before_end_p): New.
42180         * tree-eh.c (cleanup_empty_eh): Skip more debug stmts.
42182 2017-01-04  Alexandre Oliva  <aoliva@redhat.com>
42184         * multiple_target.c (create_dispatcher_calls): Init e_next.
42185         * tree-ssa-loop-split.c (split_loop): Init border.
42186         * tree-vect-loop.c (vect_determine_vectorization_factor): Init
42187         scalar_type.
42189 2017-01-04  Michael Meissner  <meissner@linux.vnet.ibm.com>
42191         PR target/71977
42192         PR target/70568
42193         PR target/78823
42194         * config/rs6000/predicates.md (sf_subreg_operand): New predicate.
42195         (altivec_register_operand): Do not return true if the operand
42196         contains a SUBREG mixing SImode and SFmode.
42197         (vsx_register_operand): Likewise.
42198         (vsx_reg_sfsubreg_ok): New predicate.
42199         (vfloat_operand): Do not return true if the operand contains a
42200         SUBREG mixing SImode and SFmode.
42201         (vint_operand): Likewise.
42202         (vlogical_operand): Likewise.
42203         (gpc_reg_operand): Likewise.
42204         (int_reg_operand): Likewise.
42205         * config/rs6000/rs6000-protos.h (valid_sf_si_move): Add declaration.
42206         * config/rs6000/rs6000.c (valid_sf_si_move): New function to
42207         determine if a MOVSI or MOVSF operation contains SUBREGs that mix
42208         SImode and SFmode.
42209         (rs6000_emit_move_si_sf_subreg): New helper function.
42210         (rs6000_emit_move): Call rs6000_emit_move_si_sf_subreg to possbily
42211         fixup SUBREGs involving SImode and SFmode.
42212         * config/rs6000/vsx.md (SFBOOL_*): New constants that are operand
42213         numbers for the new peephole2 optimization.
42214         (peephole2 for SFmode unions): New peephole2 to optimize cases in
42215         the GLIBC math library that do AND/IOR/XOR operations on single
42216         precision floating point.
42217         * config/rs6000/rs6000.h (TARGET_NO_SF_SUBREG): New internal
42218         target macros to say whether we need to avoid SUBREGs mixing
42219         SImode and SFmode.
42220         (TARGET_ALLOW_SF_SUBREG): Likewise.
42221         * config/rs6000/rs6000.md (UNSPEC_SF_FROM_SI): New unspecs.
42222         (UNSPEC_SI_FROM_SF): Likewise.
42223         (iorxor): Change spacing.
42224         (and_ior_xor): New iterator for AND, IOR, and XOR.
42225         (movsi_from_sf): New insns for SImode/SFmode SUBREG support.
42226         (movdi_from_sf_zero_ext): Likewise.
42227         (mov<mode>_hardfloat, FMOVE32 iterator): Use register_operand
42228         instead of gpc_reg_operand.  Add SImode/SFmode SUBREG support.
42229         (movsf_from_si): New insn for SImode/SFmode SUBREG support.
42230         (fma<mode>4): Use gpc_reg_operand instead of register_operand.
42231         (fms<mode>4): Likewise.
42232         (fnma<mode>4): Likewise.
42233         (fnms<mode>4): Likewise.
42234         (nfma<mode>4): Likewise.
42235         (nfms<mode>4): Likewise.
42237 2017-01-04  Marek Polacek  <polacek@redhat.com>
42239         PR c++/64767
42240         * doc/invoke.texi: Document -Wpointer-compare.
42242 2017-01-04  Jakub Jelinek  <jakub@redhat.com>
42244         * optc-gen.awk: Emit #error for -W*/-f*/-m* Enum without
42245         RejectNegative.
42247         * dwarf2out.c (output_loc_list): Don't throw away 64K+ location
42248         descriptions for -gdwarf-5 and emit them as uleb128 instead of
42249         2-byte data.
42251 2017-01-04  Kelvin Nilsen  <kelvin@gcc.gnu.org>
42253         PR target/78056
42254         * doc/sourcebuild.texi (PowerPC-specific attributes): Add
42255         documentation of the powerpc_popcntb_ok attribute.
42256         * config/rs6000/rs6000.c (rs6000_option_override_internal): Add
42257         code to issue warning messages if a requested CPU configuration is
42258         not supported by the binary (assembler and loader) toolchain.
42259         (spe_init_builtins): Add two assertions to prevent ICE if attempt is
42260         made to define a built-in function that has been disabled.
42261         (paired_init_builtins): Add assertion to prevent ICE if attempt is
42262         made to define a built-in function that has been disabled.
42263         (altivec_init_builtins): Add comment explaining why definition
42264         of the DST built-in functions is not preceded by an assertion
42265         check.  Add assertions to prevent ICE if attempts are made to
42266         define an altivec predicate or an abs* built-in function that has
42267         been disabled.
42268         (htm_init_builtins): Add comment explaining why definition of the
42269         htm built-in functions is not preceded by an assertion check.
42271 2017-01-04  Jeff Law  <law@redhat.com>
42273         PR tree-optimizatin/67955
42274         * tree-ssa-alias.c (same_addr_size_stores_p): Check offsets first.
42275         Allow any SSA_VAR_P as the base objects.  Use integer_zerop.  Verify
42276         the points-to solution does not include pt_null.  Use DECL_PT_UID
42277         unconditionally.
42279 2017-01-04  Uros Bizjak  <ubizjak@gmail.com>
42281         * config/i386/i386.md (HI/SImode test with imm to QImode splitters):
42282         Use gen_int_mode instead of gen_lopwart for const_int operands.
42284 2017-01-04  Jakub Jelinek  <jakub@redhat.com>
42286         PR tree-optimization/71563
42287         * match.pd: Simplify X << Y into X if Y is known to be 0 or
42288         out of range value - has low bits known to be zero.
42290 2017-01-04  Alan Modra  <amodra@gmail.com>
42292         * Makefile.in (aclocal_deps): Update and order as per aclocal.m4.
42293         * configure: Regenerate.
42294         * config.in: Regenerate.
42296 2017-01-04  Jakub Jelinek  <jakub@redhat.com>
42298         PR bootstrap/77569
42299         * input.c (ebcdic_execution_charset::on_error): Don't use strstr for
42300         a substring of the message, but strcmp with the whole message.  Ifdef
42301         ENABLE_NLS, translate the message first using dgettext.
42303 2017-01-03  Jeff Law  <law@redhat.com>
42305         PR tree-optimizatin/78856
42306         * tree-ssa-threadupdate.c: Include tree-vectorizer.h.
42307         (mark_threaded_blocks): Remove code to truncate thread paths that
42308         cross multiple loop headers.  Instead invalidate the cached loop
42309         iteration information and handle case of a thread path walking
42310         into an irreducible region.
42312 2017-01-03  Michael Meissner  <meissner@linux.vnet.ibm.com>
42314         PR target/78900
42315         * config/rs6000/rs6000.c (rs6000_split_signbit): Change some
42316         assertions.  Add support for doing the signbit if the IEEE 128-bit
42317         floating point value is in a GPR.
42318         * config/rs6000/rs6000.md (Fsignbit): Delete.
42319         (signbit<mode>2_dm): Delete using <Fsignbit> and just use "wa".
42320         Update the length attribute if the value is in a GPR.
42321         (signbit<mode>2_dm_<su>ext): Add combiner pattern to eliminate
42322         the sign or zero extension instruction, since the value is always 0/1.
42323         (signbit<mode>2_dm2): Delete using <Fsignbit>.
42325         PR target/78953
42326         * config/rs6000/vsx.md (vsx_extract_<mode>_store_p9): If we are
42327         extracting SImode to a GPR register so that we can generate a
42328         store, limit the vector to be in a traditional Altivec register
42329         for the vextuwrx instruction.
42331 2017-01-03  Ian Lance Taylor  <iant@google.com>
42333         * godump.c (go_format_type): Treat ENUMERAL_TYPE like INTEGER_TYPE.
42335 2017-01-03  Martin Sebor  <msebor@redhat.com>
42337         PR tree-optimization/78696
42338         * gimple-ssa-sprintf.c (format_floating): Correct handling of
42339         precision.  Use MPFR for %f for greater fidelity.  Correct handling
42340         of %g.
42341         (pass_sprintf_length::compute_format_length): Set width and precision
42342         specified by asrerisk to void_node for vararg functions.
42343         (try_substitute_return_value): Adjust dump output.
42345 2017-01-03  David Edelsohn  <dje.gcc@gmail.com>
42347         * doc/invoke.texi (RS6000 options): LRA is enabled by default.
42349 2017-01-03  Eric Botcazou  <ebotcazou@adacore.com>
42351         * doc/invoke.texi (SPARC options): Document -mlra as the default.
42352         * config/sparc/sparc.c (sparc_option_override): Force LRA unless
42353         -mlra/-mno-lra was passed to the compiler.
42355 2017-01-03  James Cowgill  <James.Cowgill@imgtec.com>
42357         PR rtl-optimization/65618
42358         * emit-rtl.c (try_split): Move initialization of "before" and
42359         "after" to just before the call to emit_insn_after_setloc.
42361 2017-01-03  Gerald Pfeifer  <gerald@pfeifer.com>
42363         * doc/md.texi (Standard Names): Remove reference to Java frontend.
42365 2017-01-03  Pierre-Marie de Rodat  <derodat@adacore.com>
42367         * dwarf2out.c (gen_enumeration_type_die): When
42368         -gno-strict-dwarf, add a DW_AT_encoding attribute.
42370 2017-01-03  Jakub Jelinek  <jakub@redhat.com>
42372         PR tree-optimization/78965
42373         * gimple-ssa-sprintf.c (pass_sprintf_length::compute_format_length):
42374         Change first argument from const call_info & to call_info &.  For %n
42375         set info.nowrite to false.
42377         PR middle-end/78901
42378         * gimple-ssa-sprintf.c (try_substitute_return_value): Don't change
42379         possibly throwing calls.
42381         * genmatch.c (dt_node::gen_kids_1): If generic_exprs include SSA_NAME
42382         and exprs_len || fns_len, emit the code for SSA_NAME next to the exprs
42383         and fns handling, rather than in a separate case SSA_NAME.
42385 2017-01-02  Jeff Law  <law@redhat.com>
42387         * config/darwin-driver.c (darwin_driver_init): Const-correctness
42388         fixes for first_period and second_period variables.
42390 2017-01-02  Uros Bizjak  <ubizjak@gmail.com>
42392         PR target/78967
42393         * config/i386/i386.md (UNSPEC_NOREX_MEM): New unspec.
42394         (*insvqi_1): New insn pattern.
42395         (*insvqi_1_mem_rex64): Ditto.
42396         (*insvqi_2): Ditto.
42397         (*insvqi_3): Rename from *insvqi.
42399         (*extzvqi_mem_rex64): Add UNSPEC_NOREX_MEM tag.
42401 2017-01-02  Gerald Pfeifer  <gerald@pfeifer.com>
42403         * doc/cfg.texi (Edges): Remove reference to Java.
42404         (Maintaining the CFG): Ditto.
42406 2017-01-01  Jan Hubicka  <hubicka@ucw.cz>
42408         PR middle-end/77674
42409         * symtab.c (symtab_node::binds_to_current_def_p): Fix handling of
42410         transparent aliases.
42412 2017-01-01  Jan Hubicka  <hubicka@ucw.cz>
42414         PR middle-end/77484
42415         * predict.def (PRED_CALL): Update hitrate.
42416         (PRED_INDIR_CALL, PRED_POLYMORPHIC_CALL): New predictors.
42417         * predict.c (tree_estimate_probability_bb): Split CALL predictor
42418         into direct/indirect/polymorphic variants.
42420 2017-01-01  Jakub Jelinek  <jakub@redhat.com>
42422         Update copyright years.
42424         * gcc.c (process_command): Update copyright notice dates.
42425         * gcov-dump.c (print_version): Ditto.
42426         * gcov.c (print_version): Ditto.
42427         * gcov-tool.c (print_version): Ditto.
42428         * gengtype.c (create_file): Ditto.
42429         * doc/cpp.texi: Bump @copying's copyright year.
42430         * doc/cppinternals.texi: Ditto.
42431         * doc/gcc.texi: Ditto.
42432         * doc/gccint.texi: Ditto.
42433         * doc/gcov.texi: Ditto.
42434         * doc/install.texi: Ditto.
42435         * doc/invoke.texi: Ditto.
42437 Copyright (C) 2017 Free Software Foundation, Inc.
42439 Copying and distribution of this file, with or without modification,
42440 are permitted in any medium without royalty provided the copyright
42441 notice and this notice are preserved.