poly_int: current_vector_size and TARGET_AUTOVECTORIZE_VECTOR_SIZES
[official-gcc.git] / gcc / ChangeLog
blob39f3a91d24d3edd2c47c4f0aad355bb4330e70b8
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         * target.h (vector_sizes, auto_vector_sizes): New typedefs.
6         * target.def (autovectorize_vector_sizes): Return the vector sizes
7         by pointer, using vector_sizes rather than a bitmask.
8         * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
9         * targhooks.c (default_autovectorize_vector_sizes): Likewise.
10         * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
11         Likewise.
12         * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
13         * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
14         * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
15         * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
16         * omp-general.c (omp_max_vf): Likewise.
17         * omp-low.c (omp_clause_aligned_alignment): Likewise.
18         * optabs-query.c (can_vec_mask_load_store_p): Likewise.
19         * tree-vect-loop.c (vect_analyze_loop): Likewise.
20         * tree-vect-slp.c (vect_slp_bb): Likewise.
21         * doc/tm.texi: Regenerate.
22         * tree-vectorizer.h (current_vector_size): Change from an unsigned int
23         to a poly_uint64.
24         * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
25         the vector size as a poly_uint64 rather than an unsigned int.
26         (current_vector_size): Change from an unsigned int to a poly_uint64.
27         (get_vectype_for_scalar_type): Update accordingly.
28         * tree.h (build_truth_vector_type): Take the size and number of
29         units as a poly_uint64 rather than an unsigned int.
30         (build_vector_type): Add a temporary overload that takes
31         the number of units as a poly_uint64 rather than an unsigned int.
32         * tree.c (make_vector_type): Likewise.
33         (build_truth_vector_type): Take the number of units as a poly_uint64
34         rather than an unsigned int.
36 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
37             Alan Hayward  <alan.hayward@arm.com>
38             David Sherwood  <david.sherwood@arm.com>
40         * target.def (get_mask_mode): Take the number of units and length
41         as poly_uint64s rather than unsigned ints.
42         * targhooks.h (default_get_mask_mode): Update accordingly.
43         * targhooks.c (default_get_mask_mode): Likewise.
44         * config/i386/i386.c (ix86_get_mask_mode): Likewise.
45         * doc/tm.texi: Regenerate.
47 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
48             Alan Hayward  <alan.hayward@arm.com>
49             David Sherwood  <david.sherwood@arm.com>
51         * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
52         * omp-general.c (omp_max_vf): Likewise.
53         * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
54         (expand_omp_simd): Handle polynomial safelen.
55         * omp-low.c (omplow_simd_context): Add a default constructor.
56         (omplow_simd_context::max_vf): Change from int to poly_uint64.
57         (lower_rec_simd_input_clauses): Update accordingly.
58         (lower_rec_input_clauses): Likewise.
60 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
61             Alan Hayward  <alan.hayward@arm.com>
62             David Sherwood  <david.sherwood@arm.com>
64         * tree-vectorizer.h (vect_nunits_for_cost): New function.
65         * tree-vect-loop.c (vect_model_reduction_cost): Use it.
66         * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
67         (vect_analyze_slp_cost): Likewise.
68         * tree-vect-stmts.c (vect_model_store_cost): Likewise.
69         (vect_model_load_cost): Likewise.
71 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
72             Alan Hayward  <alan.hayward@arm.com>
73             David Sherwood  <david.sherwood@arm.com>
75         * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
76         (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
77         from an unsigned int * to a poly_uint64_pod *.
78         (calculate_unrolling_factor): New function.
79         (vect_analyze_slp_instance): Use it.  Track polynomial max_nunits.
81 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
82             Alan Hayward  <alan.hayward@arm.com>
83             David Sherwood  <david.sherwood@arm.com>
85         * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
86         from an unsigned int to a poly_uint64.
87         (_loop_vec_info::slp_unrolling_factor): Likewise.
88         (_loop_vec_info::vectorization_factor): Change from an int
89         to a poly_uint64.
90         (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
91         (vect_get_num_vectors): New function.
92         (vect_update_max_nunits, vect_vf_for_cost): Likewise.
93         (vect_get_num_copies): Use vect_get_num_vectors.
94         (vect_analyze_data_ref_dependences): Change max_vf from an int *
95         to an unsigned int *.
96         (vect_analyze_data_refs): Change min_vf from an int * to a
97         poly_uint64 *.
98         (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
99         than an unsigned HOST_WIDE_INT.
100         * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
101         (vect_analyze_data_ref_dependence): Change max_vf from an int *
102         to an unsigned int *.
103         (vect_analyze_data_ref_dependences): Likewise.
104         (vect_compute_data_ref_alignment): Handle polynomial vf.
105         (vect_enhance_data_refs_alignment): Likewise.
106         (vect_prune_runtime_alias_test_list): Likewise.
107         (vect_shift_permute_load_chain): Likewise.
108         (vect_supportable_dr_alignment): Likewise.
109         (dependence_distance_ge_vf): Take the vectorization factor as a
110         poly_uint64 rather than an unsigned HOST_WIDE_INT.
111         (vect_analyze_data_refs): Change min_vf from an int * to a
112         poly_uint64 *.
113         * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
114         vfm1 as a poly_uint64 rather than an int.  Make the same change
115         for the returned bound_scalar.
116         (vect_gen_vector_loop_niters): Handle polynomial vf.
117         (vect_do_peeling): Likewise.  Update call to
118         vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
119         (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
120         be constant.
121         * tree-vect-loop.c (vect_determine_vectorization_factor)
122         (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
123         (vect_get_known_peeling_cost): Likewise.
124         (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
125         (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
126         (vect_transform_loop): Likewise.  Use the lowest possible VF when
127         updating the upper bounds of the loop.
128         (vect_min_worthwhile_factor): Make static.  Return an unsigned int
129         rather than an int.
130         * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
131         polynomial unroll factors.
132         (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
133         (vect_make_slp_decision): Likewise.
134         (vect_supported_load_permutation_p): Likewise, and polynomial
135         vf too.
136         (vect_analyze_slp_cost): Handle polynomial vf.
137         (vect_slp_analyze_node_operations): Likewise.
138         (vect_slp_analyze_bb_1): Likewise.
139         (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
140         than an unsigned HOST_WIDE_INT.
141         * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
142         (vectorizable_load): Handle polynomial vf.
143         * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
144         a poly_uint64.
145         (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
147 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
148             Alan Hayward  <alan.hayward@arm.com>
149             David Sherwood  <david.sherwood@arm.com>
151         * match.pd: Handle bit operations involving three constants
152         and try to fold one pair.
154 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
156         * tree-vect-loop-manip.c: Include gimple-fold.h.
157         (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
158         niters_maybe_zero parameters.  Handle other cases besides a step of 1.
159         (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
160         Add a path that uses a step of VF instead of 1, but disable it
161         for now.
162         (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
163         and niters_no_overflow parameters.  Update calls to
164         slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
165         Create a new SSA name if the latter choses to use a ste other
166         than zero, and return it via niters_vector_mult_vf_var.
167         * tree-vect-loop.c (vect_transform_loop): Update calls to
168         vect_do_peeling, vect_gen_vector_loop_niters and
169         slpeel_make_loop_iterate_ntimes.
170         * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
171         (vect_gen_vector_loop_niters): Update declarations after above changes.
173 2018-01-02  Michael Meissner  <meissner@linux.vnet.ibm.com>
175         * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
176         128-bit round to integer instructions.
177         (ceil<mode>2): Likewise.
178         (btrunc<mode>2): Likewise.
179         (round<mode>2): Likewise.
181 2018-01-02  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
183         * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
184         unaligned VSX load/store on P8/P9.
185         (expand_block_clear): Allow the use of unaligned VSX
186         load/store on P8/P9.
188 2018-01-02  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
190         * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
191         New function.
192         (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
193         swap associated with both a load and a store.
195 2018-01-02  Andrew Waterman  <andrew@sifive.com>
197         * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
198         * config/riscv/riscv.md (clear_cache): Use it.
200 2018-01-02  Artyom Skrobov  <tyomitch@gmail.com>
202         * web.c: Remove out-of-date comment.
204 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
206         * expr.c (fixup_args_size_notes): Check that any existing
207         REG_ARGS_SIZE notes are correct, and don't try to re-add them.
208         (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
209         (emit_single_push_insn): ...here.
211 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
213         * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
214         (const_vector_encoded_nelts): New function.
215         (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
216         (const_vector_int_elt, const_vector_elt): Declare.
217         * emit-rtl.c (const_vector_int_elt_1): New function.
218         (const_vector_elt): Likewise.
219         * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
220         of CONST_VECTOR_ELT.
222 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
224         * expr.c: Include rtx-vector-builder.h.
225         (const_vector_mask_from_tree): Use rtx_vector_builder and operate
226         directly on the tree encoding.
227         (const_vector_from_tree): Likewise.
228         * optabs.c: Include rtx-vector-builder.h.
229         (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
230         sequence of "u" values.
231         * vec-perm-indices.c: Include rtx-vector-builder.h.
232         (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
233         directly on the vec_perm_indices encoding.
235 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
237         * doc/rtl.texi (const_vector): Describe new encoding scheme.
238         * Makefile.in (OBJS): Add rtx-vector-builder.o.
239         * rtx-vector-builder.h: New file.
240         * rtx-vector-builder.c: Likewise.
241         * rtl.h (rtx_def::u2): Add a const_vector field.
242         (CONST_VECTOR_NPATTERNS): New macro.
243         (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
244         (CONST_VECTOR_DUPLICATE_P): Likewise.
245         (CONST_VECTOR_STEPPED_P): Likewise.
246         (CONST_VECTOR_ENCODED_ELT): Likewise.
247         (const_vec_duplicate_p): Check for a duplicated vector encoding.
248         (unwrap_const_vec_duplicate): Likewise.
249         (const_vec_series_p): Check for a non-duplicated vector encoding.
250         Say that the function only returns true for integer vectors.
251         * emit-rtl.c: Include rtx-vector-builder.h.
252         (gen_const_vec_duplicate_1): Delete.
253         (gen_const_vector): Call gen_const_vec_duplicate instead of
254         gen_const_vec_duplicate_1.
255         (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
256         (gen_const_vec_duplicate): Use rtx_vector_builder.
257         (gen_const_vec_series): Likewise.
258         (gen_rtx_CONST_VECTOR): Likewise.
259         * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
260         (swap_const_vector_halves): Take an rtx pointer rather than rtx.
261         Build a new vector rather than modifying a CONST_VECTOR in-place.
262         (handle_special_swappables): Update call accordingly.
263         * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
264         (swap_const_vector_halves): Take an rtx pointer rather than rtx.
265         Build a new vector rather than modifying a CONST_VECTOR in-place.
266         (handle_special_swappables): Update call accordingly.
268 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
270         * simplify-rtx.c (simplify_const_binary_operation): Use
271         CONST_VECTOR_ELT instead of XVECEXP.
273 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
275         * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
276         the selector elements to be different from the data elements
277         if the selector is a VECTOR_CST.
278         * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
279         ssizetype for the selector.
281 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
283         * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
284         before testing each element individually.
285         * tree-vect-generic.c (lower_vec_perm): Likewise.
287 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
289         * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
290         * selftest-run-tests.c (selftest::run_tests): Call it.
291         * vector-builder.h (vector_builder::operator ==): New function.
292         (vector_builder::operator !=): Likewise.
293         * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
294         (vec_perm_indices::all_from_input_p): New function.
295         * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
296         (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
297         * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
298         instead of reading the VECTOR_CST directly.  Detect whether both
299         vector inputs are the same before constructing the vec_perm_indices,
300         and update the number of inputs argument accordingly.  Use the
301         utility functions added above.  Only construct sel2 if we need to.
303 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
305         * optabs.c (expand_vec_perm_var): Use an explicit encoding for
306         the broadcast of the low byte.
307         (expand_mult_highpart): Use an explicit encoding for the permutes.
308         * optabs-query.c (can_mult_highpart_p): Likewise.
309         * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
310         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
311         (vectorizable_bswap): Likewise.
312         * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
313         explicit encoding for the power-of-2 permutes.
314         (vect_permute_store_chain): Likewise.
315         (vect_grouped_load_supported): Likewise.
316         (vect_permute_load_chain): Likewise.
318 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
320         * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
321         * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
322         * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
323         * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
324         * tree-vect-stmts.c (vectorizable_bswap): Likewise.
325         (vect_gen_perm_mask_any): Likewise.
327 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
329         * int-vector-builder.h: New file.
330         * vec-perm-indices.h: Include int-vector-builder.h.
331         (vec_perm_indices): Redefine as an int_vector_builder.
332         (auto_vec_perm_indices): Delete.
333         (vec_perm_builder): Redefine as a stand-alone class.
334         (vec_perm_indices::vec_perm_indices): New function.
335         (vec_perm_indices::clamp): Likewise.
336         * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
337         (vec_perm_indices::new_vector): New function.
338         (vec_perm_indices::new_expanded_vector): Update for new
339         vec_perm_indices class.
340         (vec_perm_indices::rotate_inputs): New function.
341         (vec_perm_indices::all_in_range_p): Operate directly on the
342         encoded form, without computing elided elements.
343         (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
344         encoding.  Update for new vec_perm_indices class.
345         * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
346         the given vec_perm_builder.
347         (expand_vec_perm_var): Update vec_perm_builder constructor.
348         (expand_mult_highpart): Use vec_perm_builder instead of
349         auto_vec_perm_indices.
350         * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
351         vec_perm_indices instead of auto_vec_perm_indices.  Use a single
352         or double series encoding as appropriate.
353         * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
354         vec_perm_indices instead of auto_vec_perm_indices.
355         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
356         * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
357         (vect_permute_store_chain): Likewise.
358         (vect_grouped_load_supported): Likewise.
359         (vect_permute_load_chain): Likewise.
360         (vect_shift_permute_load_chain): Likewise.
361         * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
362         (vect_transform_slp_perm_load): Likewise.
363         (vect_schedule_slp_instance): Likewise.
364         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
365         (vectorizable_mask_load_store): Likewise.
366         (vectorizable_bswap): Likewise.
367         (vectorizable_store): Likewise.
368         (vectorizable_load): Likewise.
369         * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
370         vec_perm_indices instead of auto_vec_perm_indices.  Use
371         tree_to_vec_perm_builder to read the vector from a tree.
372         * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
373         vec_perm_builder instead of a vec_perm_indices.
374         (have_whole_vector_shift): Use vec_perm_builder and
375         vec_perm_indices instead of auto_vec_perm_indices.  Leave the
376         truncation to calc_vec_perm_mask_for_shift.
377         (vect_create_epilog_for_reduction): Likewise.
378         * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
379         from auto_vec_perm_indices to vec_perm_indices.
380         (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
381         instead of changing individual elements.
382         (aarch64_vectorize_vec_perm_const): Use new_vector to install
383         the vector in d.perm.
384         * config/arm/arm.c (expand_vec_perm_d::perm): Change
385         from auto_vec_perm_indices to vec_perm_indices.
386         (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
387         instead of changing individual elements.
388         (arm_vectorize_vec_perm_const): Use new_vector to install
389         the vector in d.perm.
390         * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
391         Update vec_perm_builder constructor.
392         (rs6000_expand_interleave): Likewise.
393         * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
394         (rs6000_expand_interleave): Likewise.
396 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
398         * optabs-query.c (can_vec_perm_var_p): Check whether lowering
399         to qimode could truncate the indices.
400         * optabs.c (expand_vec_perm_var): Likewise.
402 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
404         * Makefile.in (OBJS): Add vec-perm-indices.o.
405         * vec-perm-indices.h: New file.
406         * vec-perm-indices.c: Likewise.
407         * target.h (vec_perm_indices): Replace with a forward class
408         declaration.
409         (auto_vec_perm_indices): Move to vec-perm-indices.h.
410         * optabs.h: Include vec-perm-indices.h.
411         (expand_vec_perm): Delete.
412         (selector_fits_mode_p, expand_vec_perm_var): Declare.
413         (expand_vec_perm_const): Declare.
414         * target.def (vec_perm_const_ok): Replace with...
415         (vec_perm_const): ...this new hook.
416         * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
417         (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
418         * doc/tm.texi: Regenerate.
419         * optabs.def (vec_perm_const): Delete.
420         * doc/md.texi (vec_perm_const): Likewise.
421         (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
422         * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
423         expand_vec_perm for constant permutation vectors.  Assert that
424         the mode of variable permutation vectors is the integer equivalent
425         of the mode that is being permuted.
426         * optabs-query.h (selector_fits_mode_p): Declare.
427         * optabs-query.c: Include vec-perm-indices.h.
428         (selector_fits_mode_p): New function.
429         (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
430         is defined, instead of checking whether the vec_perm_const_optab
431         exists.  Use targetm.vectorize.vec_perm_const instead of
432         targetm.vectorize.vec_perm_const_ok.  Check whether the indices
433         fit in the vector mode before using a variable permute.
434         * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
435         vec_perm_indices instead of an rtx.
436         (expand_vec_perm): Replace with...
437         (expand_vec_perm_const): ...this new function.  Take the selector
438         as a vec_perm_indices rather than an rtx.  Also take the mode of
439         the selector.  Update call to shift_amt_for_vec_perm_mask.
440         Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
441         Use vec_perm_indices::new_expanded_vector to expand the original
442         selector into bytes.  Check whether the indices fit in the vector
443         mode before using a variable permute.
444         (expand_vec_perm_var): Make global.
445         (expand_mult_highpart): Use expand_vec_perm_const.
446         * fold-const.c: Includes vec-perm-indices.h.
447         * tree-ssa-forwprop.c: Likewise.
448         * tree-vect-data-refs.c: Likewise.
449         * tree-vect-generic.c: Likewise.
450         * tree-vect-loop.c: Likewise.
451         * tree-vect-slp.c: Likewise.
452         * tree-vect-stmts.c: Likewise.
453         * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
454         Delete.
455         * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
456         * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
457         (aarch64_vectorize_vec_perm_const_ok): Fuse into...
458         (aarch64_vectorize_vec_perm_const): ...this new function.
459         (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
460         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
461         * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
462         * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
463         * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
464         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
465         (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
466         into...
467         (arm_vectorize_vec_perm_const): ...this new function.  Explicitly
468         check for NEON modes.
469         * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
470         * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
471         * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
472         (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
473         into...
474         (ix86_vectorize_vec_perm_const): ...this new function.  Incorporate
475         the old VEC_PERM_CONST conditions.
476         * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
477         * config/ia64/vect.md (vec_perm_const<mode>): Delete.
478         * config/ia64/ia64.c (ia64_expand_vec_perm_const)
479         (ia64_vectorize_vec_perm_const_ok): Merge into...
480         (ia64_vectorize_vec_perm_const): ...this new function.
481         * config/mips/loongson.md (vec_perm_const<mode>): Delete.
482         * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
483         * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
484         * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
485         * config/mips/mips.c (mips_expand_vec_perm_const)
486         (mips_vectorize_vec_perm_const_ok): Merge into...
487         (mips_vectorize_vec_perm_const): ...this new function.
488         * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
489         * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
490         * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
491         * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
492         * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
493         (rs6000_expand_vec_perm_const): Delete.
494         * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
495         Delete.
496         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
497         (altivec_expand_vec_perm_const_le): Take each operand individually.
498         Operate on constant selectors rather than rtxes.
499         (altivec_expand_vec_perm_const): Likewise.  Update call to
500         altivec_expand_vec_perm_const_le.
501         (rs6000_expand_vec_perm_const): Delete.
502         (rs6000_vectorize_vec_perm_const_ok): Delete.
503         (rs6000_vectorize_vec_perm_const): New function.
504         (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
505         an element count and rtx array.
506         (rs6000_expand_extract_even): Update call accordingly.
507         (rs6000_expand_interleave): Likewise.
508         * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
509         * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
510         * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
511         * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
512         (rs6000_expand_vec_perm_const): Delete.
513         * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
514         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
515         (altivec_expand_vec_perm_const_le): Take each operand individually.
516         Operate on constant selectors rather than rtxes.
517         (altivec_expand_vec_perm_const): Likewise.  Update call to
518         altivec_expand_vec_perm_const_le.
519         (rs6000_expand_vec_perm_const): Delete.
520         (rs6000_vectorize_vec_perm_const_ok): Delete.
521         (rs6000_vectorize_vec_perm_const): New function.  Remove stray
522         reference to the SPE evmerge intructions.
523         (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
524         an element count and rtx array.
525         (rs6000_expand_extract_even): Update call accordingly.
526         (rs6000_expand_interleave): Likewise.
527         * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
528         * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
529         new function.
530         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
532 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
534         * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
535         vector mode and that that mode matches the mode of the data
536         being permuted.
537         (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
538         out into expand_vec_perm_var.  Do all CONST_VECTOR handling here,
539         directly using expand_vec_perm_1 when forcing selectors into
540         registers.
541         (expand_vec_perm_var): New function, split out from expand_vec_perm.
543 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
545         * optabs-query.h (can_vec_perm_p): Delete.
546         (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
547         * optabs-query.c (can_vec_perm_p): Split into...
548         (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
549         (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
550         particular selector is valid.
551         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
552         * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
553         (vect_grouped_load_supported): Likewise.
554         (vect_shift_permute_load_chain): Likewise.
555         * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
556         (vect_transform_slp_perm_load): Likewise.
557         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
558         (vectorizable_bswap): Likewise.
559         (vect_gen_perm_mask_checked): Likewise.
560         * fold-const.c (fold_ternary_loc): Likewise.  Don't take
561         implementations of variable permutation vectors into account
562         when deciding which selector to use.
563         * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
564         vec_perm_const_optab is supported; instead use can_vec_perm_const_p
565         with a false third argument.
566         * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
567         to test whether the constant selector is valid and can_vec_perm_var_p
568         to test whether a variable selector is valid.
570 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
572         * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
573         * optabs-query.c (can_vec_perm_p): Likewise.
574         * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
575         instead of vec_perm_indices.
576         * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
577         (vect_gen_perm_mask_checked): Likewise,
578         * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
579         (vect_gen_perm_mask_checked): Likewise,
581 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
583         * optabs-query.h (qimode_for_vec_perm): Declare.
584         * optabs-query.c (can_vec_perm_p): Split out qimode search to...
585         (qimode_for_vec_perm): ...this new function.
586         * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
588 2018-01-02  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
590         * rtlanal.c (canonicalize_condition): Return 0 if final rtx
591         does not have a conditional at the top.
593 2018-01-02  Richard Biener  <rguenther@suse.de>
595         * ipa-inline.c (big_speedup_p): Fix expression.
597 2018-01-02  Jan Hubicka  <hubicka@ucw.cz>
599         PR target/81616
600         * config/i386/x86-tune-costs.h: Increase cost of integer load costs
601         for generic 4->6.
603 2018-01-02  Jan Hubicka  <hubicka@ucw.cz>
605         PR target/81616
606         Generic tuning.
607         * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
608         cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
609         and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
610         cond_taken_branch_cost 3->4.
612 2018-01-01  Jakub Jelinek  <jakub@redhat.com>
614         PR tree-optimization/83581
615         * tree-loop-distribution.c (pass_loop_distribution::execute): Return
616         TODO_cleanup_cfg if any changes have been made.
618         PR middle-end/83608
619         * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
620         convert_modes if target mode has the right side, but different mode
621         class.
623         PR middle-end/83609
624         * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
625         last argument when extracting from CONCAT.  If either from_real or
626         from_imag is NULL, use expansion through memory.  If result is not
627         a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
628         the parts directly to inner mode, if even that fails, use expansion
629         through memory.
631         PR middle-end/83623
632         * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
633         check for bswap in mode rather than HImode and use that in expand_unop
634         too.
636 2017-12-31  Jakub Jelinek  <jakub@redhat.com>
638         PR target/83536
639         * config/i386/i386.c (ix86_attribute_table): Remove excess
640         initializer for "shared" attribute.
642 2017-12-30  Tom de Vries  <tom@codesourcery.com>
644         PR libgomp/83046
645         * omp-expand.c (expand_omp_target): If in_lto_p, mark offload_funcs with
646         DECL_PRESERVE_P.
647         * lto-streamer-out.c (prune_offload_funcs): New function.  Remove
648         offload_funcs entries that no longer have a corresponding cgraph_node.
649         Mark the remaining ones as DECL_PRESERVE_P.
650         (output_lto): Call prune_offload_funcs.
652 2017-12-30  Jakub Jelinek  <jakub@redhat.com>
654         * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
655         vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>,
656         vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
657         vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
658         vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
659         vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1, vpdpbusd_<mode>,
660         vpdpbusd_<mode>_mask, vpdpbusd_<mode>_maskz, vpdpbusd_<mode>_maskz_1,
661         vpdpbusds_<mode>, vpdpbusds_<mode>_mask, vpdpbusds_<mode>_maskz,
662         vpdpbusds_<mode>_maskz_1, vpdpwssd_<mode>, vpdpwssd_<mode>_mask,
663         vpdpwssd_<mode>_maskz, vpdpwssd_<mode>_maskz_1, vpdpwssds_<mode>,
664         vpdpwssds_<mode>_mask, vpdpwssds_<mode>_maskz,
665         vpdpwssds_<mode>_maskz_1, vaesdec_<mode>, vaesdeclast_<mode>,
666         vaesenc_<mode>, vpclmulqdq_<mode>,
667         avx512vl_vpshufbitqmb<mode><mask_scalar_merge_name>): Formatting fixes.
669 2017-12-28  Michael Meissner  <meissner@linux.vnet.ibm.com>
671         * builtins.def: (_Float<N> and _Float<N>X BUILT_IN_CEIL): Add
672         _Float<N> and _Float<N>X variants for rounding built-in
673         functions.
674         (_Float<N> and _Float<N>X BUILT_IN_FLOOR): Likewise.
675         (_Float<N> and _Float<N>X BUILT_IN_NEARBYINT): Likewise.
676         (_Float<N> and _Float<N>X BUILT_IN_RINT): Likewise.
677         (_Float<N> and _Float<N>X BUILT_IN_ROUND): Likewise.
678         (_Float<N> and _Float<N>X BUILT_IN_TRUNC): Likewise.
679         * builtins.c (mathfn_built_in_2): Likewise.
680         * internal-fn.def (CEIL): Likewise.
681         (FLOOR): Likewise.
682         (NEARBYINT): Likewise.
683         (RINT): Likewise.
684         (ROUND): Likewise.
685         (TRUNC): Likewise.
686         * convert.c (convert_to_integer_1): Likewise.
687         * fold-const.c (tree_call_nonnegative_warnv_p): Likewise.
688         (integer_valued_real_call_p): Likewise.
689         * fold-const-call.c (fold_const_call_ss): Likewise.
690         * gencfn-macros.c (print_case_cfn): Change CFN and operator
691         printers to take a const char * suffix instead of a bool.
692         (print_define_operator_list): Likewise.
693         (fltall_suffixes): New list of suffixes, that include the
694         traditional suffixes as well as all of the _Float<N> and
695         _Float<N>X suffixes.
696         (main): For _Float<N> and _Float<N>X functions, emit both
697         <name>_FN and <name>_ALL variants.  The <macro>_FN variant only
698         has the _Float<N> and _Float<N>X case names or operators.  The
699         <name>_ALL variant has both the traditional and the
700         _Float<N>/_Float<N>X case names or operators.
701         * match.pd (COPYSIGN optimizations): Provide optimizations for
702         _Float<N> and _Float<N>X types where possible.
703         (MIN/MAX optimizations): Likewise.
704         (sqrt optimizations): Likewise.
705         (rounding optimizations): Likewise.
707 2017-12-28  Richard Sandiford  <richard.sandiford@linaro.org>
709         * config/powerpcspe/powerpcspe.md (ieee_128bit_negative_zero): Use
710         gen_int_mode rather than GEN_INT.
711         * config/rs6000/rs6000.md (ieee_128bit_negative_zero): Likewise.
713 2017-12-28  Richard Sandiford  <richard.sandiford@linaro.org>
715         * emit-rtl.c (gen_const_vec_series): Use valid_for_const_vector_p
716         instead of CONSTANT_P.
717         (gen_vec_series): Likewise.
718         * simplify-rtx.c (simplify_binary_operation_1): Likewise.
720 2017-12-28  Andreas Schwab  <schwab@linux-m68k.org>
722         * config/m68k/m68k.md (ashrdi3_const1, lshrdi3_const1): Add
723         CC_STATUS_INIT.
725 2017-12-27  Kugan Vivekanandarajah  <kuganv@linaro.org>
727         * config/aarch64/aarch64-simd.md (aarch64_ld1x2<VQ:mode>): New.
728         (aarch64_ld1x2<VDC:mode>): Likewise.
729         (aarch64_simd_ld1<mode>_x2): Likewise.
730         (aarch64_simd_ld1<mode>_x2): Likewise.
731         * config/aarch64/arm_neon.h (vld1_u8_x2): New.
732         (vld1_s8_x2): Likewise.
733         (vld1_u16_x2): Likewise.
734         (vld1_s16_x2): Likewise.
735         (vld1_u32_x2): Likewise.
736         (vld1_s32_x2): Likewise.
737         (vld1_u64_x2): Likewise.
738         (vld1_s64_x2): Likewise.
739         (vld1_f16_x2): Likewise.
740         (vld1_f32_x2): Likewise.
741         (vld1_f64_x2): Likewise.
742         (vld1_p8_x2): Likewise.
743         (vld1_p16_x2): Likewise.
744         (vld1_p64_x2): Likewise.
745         (vld1q_u8_x2): Likewise.
746         (vld1q_s8_x2): Likewise.
747         (vld1q_u16_x2): Likewise.
748         (vld1q_s16_x2): Likewise.
749         (vld1q_u32_x2): Likewise.
750         (vld1q_s32_x2): Likewise.
751         (vld1q_u64_x2): Likewise.
752         (vld1q_s64_x2): Likewise.
753         (vld1q_f16_x2): Likewise.
754         (vld1q_f32_x2): Likewise.
755         (vld1q_f64_x2): Likewise.
756         (vld1q_p8_x2): Likewise.
757         (vld1q_p16_x2): Likewise.
758         (vld1q_p64_x2): Likewise.
760 2017-12-27  Martin Liska  <mliska@suse.cz>
762         PR tree-optimization/83552
763         * tree-ssa-strlen.c (fold_strstr_to_strncmp): Assign result
764         of get_string_lenth to a SSA_NAME if not a GIMPLE value.
766 2017-12-27  Tom de Vries  <tom@codesourcery.com>
768         * config/nvptx/nvptx.c (nvptx_option_override): Disable
769         -gstatement-frontiers.
771 2017-12-26  Alexander Monakov  <amonakov@ispras.ru>
773         PR rtl-optimization/83513
774         * sel-sched.c (sel_rank_for_schedule): Order by non-zero usefulness
775         before priority comparison.
777 2017-12-25  Jakub Jelinek  <jakub@redhat.com>
779         PR target/83488
780         * config/i386/i386.opt (-mavx512vpopcntdq, -mmavx512bitalg): Move from
781         ix86_isa_flags2 to ix86_isa_flags.
782         * config/i386/i386-c.c (ix86_target_macros_internal): Test
783         OPTION_MASK_ISA_AVX512BITALG and OPTION_MASK_ISA_AVX512VPOPCNTDQ in
784         isa_flags rather than isa_flags2.
785         * config/i386/i386.c (ix86_target_string): Move -mavx512vpopcntdq
786         and -mavx512bitalg from isa2_opts to isa_opts.
787         (ix86_option_override_internal): Test OPTION_MASK_ISA_AVX512VPOPCNTDQ
788         in x_ix86_isa_flags_explicit rather than x_ix86_isa_flags2_explicit
789         and set it in x_ix86_isa_flags rather than x_ix86_isa_flags2.
790         Formatting fixes.
791         (def_builtin): Treat OPTION_MASK_ISA_AVX512BW or
792         OPTION_MASK_ISA_AVX512F ored with another option similarly to
793         OPTION_MASK_ISA_AVX512VL.  Even for OPTION_MASK_ISA_AVX512VL don't
794         clear it if mask is just OPTION_MASK_ISA_AVX512VL itself.
795         (ix86_expand_builtin): Don't handle OPTION_MASK_ISA_GFNI and
796         OPTION_MASK_ISA_VPCLMULQDQ specially, instead handle
797         OPTION_MASK_ISA_AVX512BW and OPTION_MASK_ISA_AVX512F that way.
798         * config/i386/i386-builtin.def: Move AVX512VPOPCNTDQ and AVX512BITALG
799         builtins from bdesc_args2 to bdesc_args section.
800         (__builtin_ia32_compressstoreuqi512_mask,
801         __builtin_ia32_compressstoreuhi512_mask,
802         __builtin_ia32_compressstoreuqi256_mask,
803         __builtin_ia32_expandloadqi512_mask,
804         __builtin_ia32_expandloadqi512_maskz,
805         __builtin_ia32_expandloadhi512_mask,
806         __builtin_ia32_expandloadhi512_maskz,
807         __builtin_ia32_compressqi512_mask, __builtin_ia32_compresshi512_mask,
808         __builtin_ia32_compressqi256_mask, __builtin_ia32_expandqi512_mask,
809         __builtin_ia32_expandqi512_maskz, __builtin_ia32_expandhi512_mask,
810         __builtin_ia32_expandhi512_maskz, __builtin_ia32_expandqi256_mask,
811         __builtin_ia32_expandqi256_maskz, __builtin_ia32_vpshrd_v32hi_mask,
812         __builtin_ia32_vpshld_v32hi_mask, __builtin_ia32_vpshrdv_v32hi_mask,
813         __builtin_ia32_vpshrdv_v32hi_maskz, __builtin_ia32_vpshldv_v32hi_mask,
814         __builtin_ia32_vpshldv_v32hi_maskz,
815         __builtin_ia32_vpopcountb_v64qi_mask,
816         __builtin_ia32_vpopcountw_v32hi_mask,
817         __builtin_ia32_vpshufbitqmb512_mask,
818         __builtin_ia32_vpshufbitqmb256_mask): Add
819         " | OPTION_MASK_ISA_AVX512BW".
820         (__builtin_ia32_expandloadqi256_mask,
821         __builtin_ia32_expandloadqi256_maskz,
822         __builtin_ia32_vpopcountb_v32qi_mask): Add
823         " | OPTION_MASK_ISA_AVX512VL | OPTION_MASK_ISA_AVX512BW".
824         (__builtin_ia32_expandloadhi256_mask,
825         __builtin_ia32_expandloadhi256_maskz,
826         __builtin_ia32_expandloadqi128_mask,
827         __builtin_ia32_expandloadqi128_maskz,
828         __builtin_ia32_expandloadhi128_mask,
829         __builtin_ia32_expandloadhi128_maskz,
830         __builtin_ia32_vpshrd_v16hi, __builtin_ia32_vpshrd_v16hi_mask,
831         __builtin_ia32_vpshrd_v8hi, __builtin_ia32_vpshrd_v8hi_mask,
832         __builtin_ia32_vpshrd_v8si, __builtin_ia32_vpshrd_v8si_mask,
833         __builtin_ia32_vpshrd_v4si, __builtin_ia32_vpshrd_v4si_mask,
834         __builtin_ia32_vpshrd_v4di, __builtin_ia32_vpshrd_v4di_mask,
835         __builtin_ia32_vpshrd_v2di, __builtin_ia32_vpshrd_v2di_mask,
836         __builtin_ia32_vpshld_v16hi, __builtin_ia32_vpshld_v16hi_mask,
837         __builtin_ia32_vpshld_v8hi, __builtin_ia32_vpshld_v8hi_mask,
838         __builtin_ia32_vpshld_v8si, __builtin_ia32_vpshld_v8si_mask,
839         __builtin_ia32_vpshld_v4si, __builtin_ia32_vpshld_v4si_mask,
840         __builtin_ia32_vpshld_v4di, __builtin_ia32_vpshld_v4di_mask,
841         __builtin_ia32_vpshld_v2di, __builtin_ia32_vpshld_v2di_mask,
842         __builtin_ia32_vpshrdv_v16hi, __builtin_ia32_vpshrdv_v16hi_mask,
843         __builtin_ia32_vpshrdv_v16hi_maskz, __builtin_ia32_vpshrdv_v8hi,
844         __builtin_ia32_vpshrdv_v8hi_mask, __builtin_ia32_vpshrdv_v8hi_maskz,
845         __builtin_ia32_vpshrdv_v8si, __builtin_ia32_vpshrdv_v8si_mask,
846         __builtin_ia32_vpshrdv_v8si_maskz, __builtin_ia32_vpshrdv_v4si,
847         __builtin_ia32_vpshrdv_v4si_mask, __builtin_ia32_vpshrdv_v4si_maskz,
848         __builtin_ia32_vpshrdv_v4di, __builtin_ia32_vpshrdv_v4di_mask,
849         __builtin_ia32_vpshrdv_v4di_maskz, __builtin_ia32_vpshrdv_v2di,
850         __builtin_ia32_vpshrdv_v2di_mask, __builtin_ia32_vpshrdv_v2di_maskz,
851         __builtin_ia32_vpshldv_v16hi, __builtin_ia32_vpshldv_v16hi_mask,
852         __builtin_ia32_vpshldv_v16hi_maskz, __builtin_ia32_vpshldv_v8hi,
853         __builtin_ia32_vpshldv_v8hi_mask, __builtin_ia32_vpshldv_v8hi_maskz,
854         __builtin_ia32_vpshldv_v8si, __builtin_ia32_vpshldv_v8si_mask,
855         __builtin_ia32_vpshldv_v8si_maskz, __builtin_ia32_vpshldv_v4si,
856         __builtin_ia32_vpshldv_v4si_mask, __builtin_ia32_vpshldv_v4si_maskz,
857         __builtin_ia32_vpshldv_v4di, __builtin_ia32_vpshldv_v4di_mask,
858         __builtin_ia32_vpshldv_v4di_maskz, __builtin_ia32_vpshldv_v2di,
859         __builtin_ia32_vpshldv_v2di_mask, __builtin_ia32_vpshldv_v2di_maskz,
860         __builtin_ia32_vpopcountb_v32qi, __builtin_ia32_vpopcountb_v16qi,
861         __builtin_ia32_vpopcountb_v16qi_mask, __builtin_ia32_vpopcountw_v16hi,
862         __builtin_ia32_vpopcountw_v16hi_mask, __builtin_ia32_vpopcountw_v8hi,
863         __builtin_ia32_vpopcountw_v8hi_mask): Add
864         " | OPTION_MASK_ISA_AVX512VL".
865         * config/i386/avx512vbmi2intrin.h (_mm512_shrdi_epi16,
866         _mm512_shrdi_epi32, _mm512_mask_shrdi_epi32, _mm512_maskz_shrdi_epi32,
867         _mm512_shrdi_epi64, _mm512_mask_shrdi_epi64, _mm512_maskz_shrdi_epi64,
868         _mm512_shldi_epi16, _mm512_shldi_epi32, _mm512_mask_shldi_epi32,
869         _mm512_maskz_shldi_epi32, _mm512_shldi_epi64, _mm512_mask_shldi_epi64,
870         _mm512_maskz_shldi_epi64, _mm512_shrdv_epi16, _mm512_shrdv_epi32,
871         _mm512_mask_shrdv_epi32, _mm512_maskz_shrdv_epi32, _mm512_shrdv_epi64,
872         _mm512_mask_shrdv_epi64, _mm512_maskz_shrdv_epi64, _mm512_shldv_epi16,
873         _mm512_shldv_epi32, _mm512_mask_shldv_epi32, _mm512_maskz_shldv_epi32,
874         _mm512_shldv_epi64, _mm512_mask_shldv_epi64,
875         _mm512_maskz_shldv_epi64): Don't require avx512bw for these intrinsics.
876         * config/i386/avx512bitalgintrin.h (_mm_bitshuffle_epi64_mask,
877         _mm_mask_bitshuffle_epi64_mask): Likewise.
878         * common/config/i386/i386-common.c
879         (OPTION_MASK_ISA_AVX512VPOPCNTDQ_SET,
880         OPTION_MASK_ISA_AVX512BITALG_SET): Or in OPTION_MASK_ISA_AVX512F_SET.
881         (OPTION_MASK_ISA_AVX512F_UNSET): Or in
882         OPTION_MASK_ISA_AVX512VPOPCNTDQ_UNSET and
883         OPTION_MASK_ISA_AVX512BITALG_UNSET.
884         (OPTION_MASK_ISA2_AVX512F_UNSET,
885         OPTION_MASK_ISA2_GENERAL_REGS_ONLY_UNSET): Define.
886         (ix86_handle_option): For -mno-general-regs-only, clear from
887         ix86_isa_flags2 OPTION_MASK_ISA2_GENERAL_REGS_ONLY_UNSET rather than
888         just OPTION_MASK_ISA_MPX.  For -mno-sse{,2,3,4,4.1,4.2,avx,avx2} and
889         -mno-ssse3 clear OPTION_MASK_ISA2_AVX512F_UNSET bits from
890         ix86_isa_flags2.  For -mno-avx512f likewise, instead of masking
891         individually listed ISAs.  For -m{,no-}avx512{vpopcntdq,bitalg} adjust
892         for moving from ix86_isa_flags2 to ix86_isa_flags.
894 2017-12-23  Jakub Jelinek  <jakub@redhat.com>
896         PR c++/83553
897         * fold-const.c (struct contains_label_data): New type.
898         (contains_label_1): Return non-NULL even for CASE_LABEL_EXPR, unless
899         inside of a SWITCH_BODY seen during the walk.
900         (contains_label_p): Use walk_tree instead of
901         walk_tree_without_duplicates, prepare data for contains_label_1 and
902         provide own pset.
904 2017-12-22  Jakub Jelinek  <jakub@redhat.com>
906         PR debug/83547
907         * tree-iterator.c (alloc_stmt_list): Start with cleared
908         TREE_SIDE_EFFECTS regardless whether a new STATEMENT_LIST is allocated
909         or old one reused.
911         PR target/83488
912         * config/i386/avx512vnniintrin.h: Don't check for __AVX512F__ nor
913         enable avx512f explicitly in #pragma GCC target.
914         * config/i386/i386-builtin.def (__builtin_ia32_vpdpbusd_v8si,
915         __builtin_ia32_vpdpbusd_v8si_mask, __builtin_ia32_vpdpbusd_v8si_maskz,
916         __builtin_ia32_vpdpbusd_v4si, __builtin_ia32_vpdpbusd_v4si_mask,
917         __builtin_ia32_vpdpbusd_v4si_maskz, __builtin_ia32_vpdpbusds_v8si,
918         __builtin_ia32_vpdpbusds_v8si_mask,
919         __builtin_ia32_vpdpbusds_v8si_maskz, __builtin_ia32_vpdpbusds_v4si,
920         __builtin_ia32_vpdpbusds_v4si_mask,
921         __builtin_ia32_vpdpbusds_v4si_maskz, __builtin_ia32_vpdpwssd_v8si,
922         __builtin_ia32_vpdpwssd_v8si_mask, __builtin_ia32_vpdpwssd_v8si_maskz,
923         __builtin_ia32_vpdpwssd_v4si, __builtin_ia32_vpdpwssd_v4si_mask,
924         __builtin_ia32_vpdpwssd_v4si_maskz, __builtin_ia32_vpdpwssds_v8si,
925         __builtin_ia32_vpdpwssds_v8si_mask,
926         __builtin_ia32_vpdpwssds_v8si_maskz, __builtin_ia32_vpdpwssds_v4si,
927         __builtin_ia32_vpdpwssds_v4si_mask,
928         __builtin_ia32_vpdpwssds_v4si_maskz): Use
929         OPTION_MASK_ISA_AVX512VNNI | OPTION_MASK_ISA_AVX512VL instead of
930         just OPTION_MASK_ISA_AVX512VNNI.
932 2017-12-22  Martin Jambor  <mjambor@suse.cz>
934         PR lto/82027
935         * lto-cgraph.c (output_cgraph_opt_summary_p): Also check former
936         clones.
938 2017-12-22  Julia Koval  <julia.koval@intel.com>
939             Sebastian Peryt  <sebastian.peryt@intel.com>
941         * common/config/i386/i386-common.c (OPTION_MASK_ISA_AVX512BITALG_SET,
942         OPTION_MASK_ISA_AVX512BITALG_UNSET): New.
943         (ix86_handle_option): Handle -mavx512bitalg, fix 4VNNIW formatting.
944         * config.gcc: Add avx512vpopcntdqvlintrin.h and avx512bitalgintrin.h.
945         * config/i386/avx512bitalgintrin.h (_mm512_popcnt_epi8, _mm512_popcnt_epi16,
946         _mm512_mask_popcnt_epi8, _mm512_maskz_popcnt_epi8, _mm512_mask_popcnt_epi16,
947         _mm512_maskz_popcnt_epi16, _mm512_bitshuffle_epi64_mask, _mm256_popcnt_epi8,
948         _mm512_mask_bitshuffle_epi64_mask, _mm256_mask_popcnt_epi8, _mm_popcnt_epi8,
949         _mm256_maskz_popcnt_epi8, _mm_bitshuffle_epi64_mask, _mm256_popcnt_epi16,
950         _mm_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
951         _mm256_mask_bitshuffle_epi64_mask, _mm_popcnt_epi16, _mm_maskz_popcnt_epi8,
952         _mm256_mask_popcnt_epi16, _mm256_maskz_popcnt_epi16, _mm_mask_popcnt_epi8,
953         _mm_mask_popcnt_epi16, _mm_maskz_popcnt_epi16): New intrinsics.
954         * config/i386/avx512vpopcntdqvlintrin.h (_mm_popcnt_epi32, _mm_popcnt_epi64,
955         _mm_mask_popcnt_epi32, _mm_maskz_popcnt_epi32, _mm256_popcnt_epi32,
956         _mm256_mask_popcnt_epi32, _mm256_maskz_popcnt_epi32, _mm_mask_popcnt_epi64,
957         _mm_maskz_popcnt_epi64, _mm256_popcnt_epi64, _mm256_mask_popcnt_epi64,
958         _mm256_maskz_popcnt_epi64): New intrinsics.
959         * config/i386/cpuid.h (bit_AVX512BITALG): New bit.
960         * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mavx512bitalg.
961         * config/i386/i386-builtin-types.def (V64QI_FTYPE_V64QI, V64QI_FTYPE_V64QI,
962         V4DI_FTYPE_V4DI, UHI_FTYPE_V2DI_V2DI_UHI, USI_FTYPE_V4DI_V4DI_USI,
963         V4SI_FTYPE_V4SI_V4SI_UHI, V8SI_FTYPE_V8SI_V8SI_UHI): New types.
964         * config/i386/i386-builtin.def (__builtin_ia32_vpopcountq_v4di,
965         __builtin_ia32_vpopcountq_v4di_mask, __builtin_ia32_vpopcountq_v2di,
966         __builtin_ia32_vpopcountq_v2di_mask, __builtin_ia32_vpopcountd_v4si,
967         __builtin_ia32_vpopcountd_v4si_mask, __builtin_ia32_vpopcountd_v8si,
968         __builtin_ia32_vpopcountd_v8si_mask, __builtin_ia32_vpopcountb_v64qi,
969         __builtin_ia32_vpopcountb_v64qi_mask, __builtin_ia32_vpopcountb_v32qi,
970         __builtin_ia32_vpopcountb_v32qi_mask, __builtin_ia32_vpopcountb_v16qi,
971         __builtin_ia32_vpopcountb_v16qi_mask, __builtin_ia32_vpopcountw_v32hi,
972         __builtin_ia32_vpopcountw_v32hi_mask, __builtin_ia32_vpopcountw_v16hi,
973         __builtin_ia32_vpopcountw_v16hi_mask, __builtin_ia32_vpopcountw_v8hi,
974         __builtin_ia32_vpopcountw_v8hi_mask, __builtin_ia32_vpshufbitqmb128_mask,
975         __builtin_ia32_vpshufbitqmb256_mask,
976         __builtin_ia32_vpshufbitqmb512_mask): New builtins.
977         * config/i386/i386-c.c (__AVX512BITALG__): New.
978         * config/i386/i386.c (isa2_opts): Add -mavx512bitalg.
979         (ix86_valid_target_attribute_inner_p): Ditto.
980         (ix86_expand_args_builtin): Handle new types.
981         * config/i386/i386.h (TARGET_AVX512BITALG, TARGET_AVX512BITALG_P): New.
982         * config/i386/i386.opt: Add -mavx512bitalg.
983         * config/i386/immintrin.h: Add avx512vpopcntdqvlintrin.h and
984         avx512bitalgintrin.h.
985         * config/i386/sse.md (VI48_AVX512VLBW): New iterator.
986         (vpopcount<mode><mask_name>): Add more types.
987         (avx512vl_vpshufbitqmb<mode><mask_scalar_merge_name>): New.
988         * doc/invoke.texi: Add -mavx512bitalg and -mavx512vpopcntdq.
990 2017-12-22  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
992         * common/config/i386/i386-common.c (OPTION_MASK_ISA_AVX512VNNI_SET):
993         Or in OPTION_MASK_ISA_AVX512F_SET.
994         (OPTION_MASK_ISA_AVX512F_UNSET): Or in
995         OPTION_MASK_ISA_AVX512VNNI_UNSET.
996         (ix86_handle_option): Adjust for
997         OPTION_MASK_ISA_AVX512VNNI_*SET being in ix86_isa_flags.
998         * config/i386/i386-builtin.def: Move VNNI builtins from ARGS2
999         section to ARGS.
1000         * config/i386/i386-c.c: Check for OPTION_MASK_ISA_AVX512VNNI in
1001         isa_flag instead of isa_flag2.
1002         * config/i386/i386.c (ix86_target_string): Move -mavx512vnni from
1003         isa_opts2 to isa_opts.
1004         * config/i386/i386.opt (mavx512vnni): Move from ix86_isa_flags2
1005         to ix86_isa_flags.
1007 2017-12-22  Mike Stump  <mikestump@comcast.net>
1008             Eric Botcazou  <ebotcazou@adacore.com>
1010         * doc/extend.texi (Loop-Specific Pragmas): Document pragma GCC unroll.
1012 2017-12-21  Jakub Jelinek  <jakub@redhat.com>
1014         PR middle-end/83487
1015         * config/i386/i386.c (ix86_function_arg_boundary): Return
1016         PARM_BOUNDARY for TYPE_EMPTY_P types.
1018         PR c/83448
1019         * gimple-ssa-sprintf.c (maybe_warn): Don't call set_caret_index
1020         if navail is >= dir.len.
1022 2017-12-21  Steve Ellcey  <sellcey@cavium.com>
1024         * config/aarch64/t-aarch64-linux (MULTILIB_OSDIRNAMES): Fix
1025         triplet for ilp32.
1027 2017-12-21  Jakub Jelinek  <jakub@redhat.com>
1029         PR rtl-optimization/80747
1030         PR rtl-optimization/83512
1031         * cfgrtl.c (force_nonfallthru_and_redirect): When splitting
1032         succ edge from ENTRY, copy partition from e->dest to the newly
1033         created bb.
1034         * bb-reorder.c (reorder_basic_blocks_simple): If last_tail is
1035         ENTRY, use BB_PARTITION of its successor block as current_partition.
1036         Don't copy partition when splitting succ edge from ENTRY.
1038         PR tree-optimization/83523
1039         * tree-ssa-math-opts.c (is_widening_mult_p): Return false if
1040         for INTEGER_TYPE TYPE_OVERFLOW_TRAPS.
1041         (convert_mult_to_fma): Likewise.
1043         PR tree-optimization/83521
1044         * tree-ssa-phiopt.c (factor_out_conditional_conversion): Use
1045         gimple_build_assign without code on result of
1046         fold_build1 (VIEW_CONVERT_EXPR, ...), as it might not create
1047         a VIEW_CONVERT_EXPR.
1049 2017-12-21  Andrew Pinski  <apinski@cavium.com>
1050             Steve Ellcey  <sellcey@cavium.com>
1052         * config/aarch64/t-aarch64-linux (MULTILIB_OSDIRNAMES): Handle
1053         multi-arch for ilp32.
1055 2017-12-21  Uros Bizjak  <ubizjak@gmail.com>
1057         PR target/83467
1058         * config/i386/i386.md (*ashl<mode>3_mask): Add operand
1059         constraints to operand 2.
1060         (*ashl<mode>3_mask_1): Ditto.
1061         (*<shift_insn><mode>3_mask): Ditto.
1062         (*<shift_insn><mode>3_mask_1): Ditto.
1063         (*<rotate_insn><mode>3_mask): Ditto.
1064         (*<rotate_insn><mode>3_mask_1): Ditto.
1066 2017-12-21  Alexandre Oliva  <aoliva@redhat.com>
1068         * reorg.c (make_return_insns): Reemit each insn with its own location.
1070 2017-12-21  James Greenhalgh  <james.greenhalgh@arm.com>
1072         * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
1073         generation for cases where splatting a value is not useful.
1074         * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
1075         across a vec_duplicate and a paradoxical subreg forming a vector
1076         mode to a vec_concat.
1078 2017-12-21  James Greenhalgh  <james.greenhalgh@arm.com>
1080         * combine.c (simplify_set): Do not transform subregs to zero_extends
1081         if the destination is not a scalar int mode.
1083 2017-12-21  Eric Botcazou  <ebotcazou@adacore.com>
1085         PR c++/82872
1086         * convert.c (convert_to_integer_1) <POINTER_TYPE>: Do not return the
1087         shared zero if the input has overflowed.
1089 2017-12-21  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
1091         * config/arm/driver-arm.c (arm_cpu_table): Specify dotprod
1092         support for Cortex-A55 and Cortex-A75.
1094 2017-12-21  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
1096         * common/config/arm/arm-common.c (compare_opt_names): Add function
1097         comment.  Use strcmp instead of manual loop.
1099 2017-12-21  Martin Liska  <mliska@suse.cz>
1101         PR gcov-profile/83509
1102         * gcov-dump.c (dump_gcov_file): Do not read info about
1103         support_unexecuted_blocks for gcda files.
1105 2017-12-21  Jakub Jelinek  <jakub@redhat.com>
1107         PR rtl-optimization/82973
1108         * emit-rtl.h (valid_for_const_vec_duplicate_p): Rename to ...
1109         (valid_for_const_vector_p): ... this.
1110         * emit-rtl.c (valid_for_const_vec_duplicate_p): Rename to ...
1111         (valid_for_const_vector_p): ... this.  Adjust function comment.
1112         (gen_vec_duplicate): Adjust caller.
1113         * optabs.c (expand_vector_broadcast): Likewise.
1114         * simplify-rtx.c (simplify_const_unary_operation): Don't optimize into
1115         CONST_VECTOR if some element isn't simplified valid_for_const_vector_p
1116         constant.
1117         (simplify_const_binary_operation): Likewise.  Use CONST_FIXED_P macro
1118         instead of GET_CODE == CONST_FIXED.
1119         (simplify_subreg): Use CONST_FIXED_P macro instead of
1120         GET_CODE == CONST_FIXED.
1122         PR target/83488
1123         * config/i386/i386.c (ix86_target_string): Move -mavx512vbmi2 and
1124         -mshstk entries from isa_opts2 to isa_opts and -mhle, -mmovbe,
1125         -mclzero and -mmwaitx entries from isa_opts to isa_opts2.
1126         (ix86_option_override_internal): Adjust for
1127         OPTION_MASK_ISA_{HLE,MOVBE,CLZERO,MWAITX} moving to ix86_isa_flags2
1128         and OPTION_MASK_ISA_SHSTK moving to ix86_isa_flags.
1129         (BDESC_VERIFYS): Remove SPECIAL_ARGS2 related checks.
1130         (ix86_init_mmx_sse_builtins): Remove bdesc_special_args2 handling.
1131         Use def_builtin2 instead of def_builtin for OPTION_MASK_ISA_MWAITX
1132         and OPTION_MASK_ISA_CLZERO builtins.  Use def_builtin instead of
1133         def_builtin2 for CET builtins.
1134         (ix86_expand_builtin): Remove bdesc_special_args2 handling.  Fix
1135         up formatting in IX86_BUILTIN_RDPID code.
1136         * config/i386/i386-builtin.def: Move VBMI2 builtins from SPECIAL_ARGS2
1137         section to SPECIAL_ARGS and from ARGS2 section to ARGS.
1138         * config/i386/i386.opt (mavx512vbmi2, mshstk): Move from
1139         ix86_isa_flags2 to ix86_isa_flags.
1140         (mhle, mmovbe, mclzero, mmwaitx): Move from ix86_isa_flags to
1141         ix86_isa_flags2.
1142         * config/i386/i386-c.c (ix86_target_macros_internal): Check for
1143         OPTION_MASK_ISA_{CLZERO,MWAITX} in isa_flag2 instead of isa_flag.
1144         Check for OPTION_MASK_ISA_{SHSTK,AVX512VBMI2} in isa_flag instead
1145         of isa_flag2.
1146         * common/config/i386/i386-common.c (OPTION_MASK_ISA_AVX512VBMI2_SET):
1147         Or in OPTION_MASK_ISA_AVX512F_SET.
1148         (OPTION_MASK_ISA_AVX512F_UNSET): Or in
1149         OPTION_MASK_ISA_AVX512VBMI2_UNSET.
1150         (ix86_handle_option): Adjust for
1151         OPTION_MASK_ISA_{SHSTK,AVX512VBMI2}_*SET being in ix86_isa_flags
1152         and OPTION_MASK_ISA_{MOVBE,MWAITX,CLZERO}_*SET in ix86_isa_flags2.
1154 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1155             Alan Hayward  <alan.hayward@arm.com>
1156             David Sherwood  <david.sherwood@arm.com>
1158         * tree-data-ref.h (prune_runtime_alias_test_list): Take the
1159         factor as a poly_uint64 rather than an unsigned HOST_WIDE_INT.
1160         * tree-data-ref.c (prune_runtime_alias_test_list): Likewise.
1161         Track polynomial offsets.
1163 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1164             Alan Hayward  <alan.hayward@arm.com>
1165             David Sherwood  <david.sherwood@arm.com>
1167         * tree-vect-data-refs.c (vect_compute_data_ref_alignment):
1168         Treat drb->init as a poly_int.  Fail if its misalignment wrt
1169         vector_alignment isn't known.
1171 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1172             Alan Hayward  <alan.hayward@arm.com>
1173             David Sherwood  <david.sherwood@arm.com>
1175         * tree-vectorizer.h (_loop_vec_info): Add a versioning_threshold
1176         field.
1177         (LOOP_VINFO_VERSIONING_THRESHOLD): New macro
1178         (vect_loop_versioning): Take the loop versioning threshold as a
1179         separate parameter.
1180         * tree-vect-loop-manip.c (vect_loop_versioning): Likewise.
1181         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
1182         versioning_threshold.
1183         (vect_analyze_loop_2): Compute the loop versioning threshold
1184         whenever loop versioning is needed, and store it in the new
1185         field rather than combining it with the cost model threshold.
1186         (vect_transform_loop): Update call to vect_loop_versioning.
1187         Try to combine the loop versioning and cost thresholds here.
1189 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1190             Alan Hayward  <alan.hayward@arm.com>
1191             David Sherwood  <david.sherwood@arm.com>
1193         * tree-ssa-loop-ivopts.h (strip_offset): Return the offset as
1194         poly_uint64_pod rather than an unsigned HOST_WIDE_INT.
1195         * tree-loop-distribution.c (classify_builtin_st): Update accordingly.
1196         * tree-ssa-loop-ivopts.c (iv_use::addr_offset): Change from
1197         an unsigned HOST_WIDE_INT to a poly_uint64_pod.
1198         (group_compare_offset): Update accordingly.
1199         (split_small_address_groups_p): Likewise.
1200         (record_use): Take addr_offset as a poly_uint64 rather than
1201         an unsigned HOST_WIDE_INT.
1202         (strip_offset): Return the offset as a poly_uint64 rather than
1203         an unsigned HOST_WIDE_INT.
1204         (record_group_use, split_address_groups): Track polynomial offsets.
1205         (add_iv_candidate_for_use): Likewise.
1206         (addr_offset_valid_p): Take the offset as a poly_int64 rather
1207         than a HOST_WIDE_INT.
1208         (strip_offset_1): Return the offset as a poly_int64 rather than
1209         a HOST_WIDE_INT.
1211 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1212             Alan Hayward  <alan.hayward@arm.com>
1213             David Sherwood  <david.sherwood@arm.com>
1215         * tree.h (get_binfo_at_offset): Take the offset as a poly_int64
1216         rather than a HOST_WIDE_INT.
1217         * tree.c (get_binfo_at_offset): Likewise.
1219 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1220             Alan Hayward  <alan.hayward@arm.com>
1221             David Sherwood  <david.sherwood@arm.com>
1223         * ipa-prop.h (build_ref_for_offset): Take the offset as a poly_int64
1224         rather than a HOST_WIDE_INT.
1225         * tree-sra.c (build_ref_for_offset): Likewise.
1227 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1228             Alan Hayward  <alan.hayward@arm.com>
1229             David Sherwood  <david.sherwood@arm.com>
1231         * fold-const.h (mem_ref_offset): Return a poly_offset_int rather
1232         than an offset_int.
1233         * tree.c (mem_ref_offset): Likewise.
1234         (build_simple_mem_ref_loc): Treat MEM_REF offsets as poly_ints.
1235         * builtins.c (get_object_alignment_2): Likewise.
1236         * expr.c (get_inner_reference, expand_expr_real_1): Likewise.
1237         * gimple-fold.c (get_base_constructor): Likewise.
1238         * gimple-ssa-strength-reduction.c (restructure_reference): Likewise.
1239         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref):
1240         Likewise.
1241         * ipa-polymorphic-call.c
1242         (ipa_polymorphic_call_context::ipa_polymorphic_call_context): Likewise.
1243         * ipa-prop.c (compute_complex_assign_jump_func): Likewise.
1244         (get_ancestor_addr_info): Likewise.
1245         * ipa-param-manipulation.c (ipa_get_adjustment_candidate): Likewise.
1246         * match.pd: Likewise.
1247         * tree-data-ref.c (dr_analyze_innermost): Likewise.
1248         * tree-dfa.c (get_addr_base_and_unit_offset_1): Likewise.
1249         * tree-eh.c (tree_could_trap_p): Likewise.
1250         * tree-object-size.c (addr_object_size): Likewise.
1251         * tree-ssa-address.c (copy_ref_info): Likewise.
1252         * tree-ssa-alias.c (indirect_ref_may_alias_decl_p): Likewise.
1253         (indirect_refs_may_alias_p): Likewise.
1254         * tree-ssa-sccvn.c (copy_reference_ops_from_ref): Likewise.
1255         * tree-ssa.c (maybe_rewrite_mem_ref_base): Likewise.
1256         (non_rewritable_mem_ref_base): Likewise.
1257         * tree-vect-data-refs.c (vect_check_gather_scatter): Likewise.
1258         * tree-vrp.c (vrp_prop::check_array_ref): Likewise.
1259         * varasm.c (decode_addr_const): Likewise.
1261 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1262             Alan Hayward  <alan.hayward@arm.com>
1263             David Sherwood  <david.sherwood@arm.com>
1265         * gimple-ssa-stor-merging.c (find_bswap_or_nop_load): Track polynomial
1266         offsets for MEM_REFs.
1268 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1269             Alan Hayward  <alan.hayward@arm.com>
1270             David Sherwood  <david.sherwood@arm.com>
1272         * tree-ssanames.h (adjust_ptr_info_misalignment): Take the increment
1273         as a poly_uint64 rather than an unsigned int.
1274         * tree-ssanames.c (adjust_ptr_info_misalignment): Likewise.
1276 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1277             Alan Hayward  <alan.hayward@arm.com>
1278             David Sherwood  <david.sherwood@arm.com>
1280         * varasm.c (addr_const::offset): Change from HOST_WIDE_INT
1281         to poly_int64.
1282         (decode_addr_const): Update accordingly.
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         * tree.h (bit_field_size, bit_field_offset): New functions.
1289         * hsa-gen.c (gen_hsa_addr): Use them.
1290         * tree-ssa-forwprop.c (simplify_bitfield_ref): Likewise.
1291         (simplify_vector_constructor): Likewise.
1292         * tree-ssa-sccvn.c (copy_reference_ops_from_ref): Likewise.
1293         * tree-cfg.c (verify_expr): Require the sizes and offsets of a
1294         BIT_FIELD_REF to be poly_uint64s rather than uhwis.
1295         * fold-const.c (fold_ternary_loc): Protect tree_to_uhwi with
1296         tree_fits_uhwi_p.
1298 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1299             Alan Hayward  <alan.hayward@arm.com>
1300             David Sherwood  <david.sherwood@arm.com>
1302         * expr.h (emit_group_load, emit_group_load_into_temps)
1303         (emit_group_store): Take the size as a poly_int64 rather than an int.
1304         * expr.c (emit_group_load_1, emit_group_load): Likewise.
1305         (emit_group_load_into_temp, emit_group_store): Likewise.
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         * ira-int.h (ira_spilled_reg_stack_slot::width): Change from
1312         an unsigned int to a poly_uint64.
1313         * ira.h (ira_reuse_stack_slot, ira_mark_new_stack_slot): Take the
1314         sizes as poly_uint64s rather than unsigned ints.
1315         * ira-color.c (ira_reuse_stack_slot, ira_mark_new_stack_slot):
1316         Likewise.
1318 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1319             Alan Hayward  <alan.hayward@arm.com>
1320             David Sherwood  <david.sherwood@arm.com>
1322         * lra-constraints.c (emit_inc): Change inc_amount from an int
1323         to a poly_int64.
1325 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1326             Alan Hayward  <alan.hayward@arm.com>
1327             David Sherwood  <david.sherwood@arm.com>
1329         * cfgexpand.c (stack_var::size): Change from a HOST_WIDE_INT
1330         to a poly_uint64.
1331         (add_stack_var, stack_var_cmp, partition_stack_vars)
1332         (dump_stack_var_partition): Update accordingly.
1333         (alloc_stack_frame_space): Take the size as a poly_int64 rather
1334         than a HOST_WIDE_INT.
1335         (expand_stack_vars, expand_one_stack_var_1): Handle polynomial sizes.
1336         (defer_stack_allocation, estimated_stack_frame_size): Likewise.
1337         (account_stack_vars, expand_one_var): Likewise.  Return a poly_uint64
1338         rather than a HOST_WIDE_INT.
1340 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1341             Alan Hayward  <alan.hayward@arm.com>
1342             David Sherwood  <david.sherwood@arm.com>
1344         * target.def (return_pops_args): Treat both the input and output
1345         sizes as poly_int64s rather than HOST_WIDE_INTS.
1346         * targhooks.h (default_return_pops_args): Update accordingly.
1347         * targhooks.c (default_return_pops_args): Likewise.
1348         * doc/tm.texi: Regenerate.
1349         * emit-rtl.h (incoming_args): Change pops_args, size and
1350         outgoing_args_size from int to poly_int64_pod.
1351         * function.h (expr_status): Change x_pending_stack_adjust and
1352         x_stack_pointer_delta from int to poly_int64.
1353         (args_size::constant): Change from HOST_WIDE_INT to poly_int64.
1354         (ARGS_SIZE_RTX): Update accordingly.
1355         * calls.c (highest_outgoing_arg_in_use): Change from int to
1356         unsigned int.
1357         (stack_usage_watermark, stored_args_watermark): New variables.
1358         (stack_region_maybe_used_p, mark_stack_region_used): New functions.
1359         (emit_call_1): Change the stack_size and rounded_stack_size
1360         parameters from HOST_WIDE_INT to poly_int64.  Track n_popped
1361         as a poly_int64.
1362         (save_fixed_argument_area): Check stack_usage_watermark.
1363         (initialize_argument_information): Change old_pending_adj from
1364         a HOST_WIDE_INT * to a poly_int64_pod *.
1365         (compute_argument_block_size): Return the size as a poly_int64
1366         rather than an int.
1367         (finalize_must_preallocate): Track polynomial argument sizes.
1368         (compute_argument_addresses): Likewise.
1369         (internal_arg_pointer_based_exp): Track polynomial offsets.
1370         (mem_overlaps_already_clobbered_arg_p): Rename to...
1371         (mem_might_overlap_already_clobbered_arg_p): ...this and take the
1372         size as a poly_uint64 rather than an unsigned HOST_WIDE_INT.
1373         Check stored_args_used_watermark.
1374         (load_register_parameters): Update accordingly.
1375         (check_sibcall_argument_overlap_1): Likewise.
1376         (combine_pending_stack_adjustment_and_call): Take the unadjusted
1377         args size as a poly_int64 rather than an int.  Return a bool
1378         indicating whether the optimization was possible and return
1379         the new adjustment by reference.
1380         (check_sibcall_argument_overlap): Track polynomail argument sizes.
1381         Update stored_args_watermark.
1382         (can_implement_as_sibling_call_p): Handle polynomial argument sizes.
1383         (expand_call): Likewise.  Maintain stack_usage_watermark and
1384         stored_args_watermark.  Update calls to
1385         combine_pending_stack_adjustment_and_call.
1386         (emit_library_call_value_1): Handle polynomial argument sizes.
1387         Call stack_region_maybe_used_p and mark_stack_region_used.
1388         Maintain stack_usage_watermark.
1389         (store_one_arg): Likewise.  Update call to
1390         mem_overlaps_already_clobbered_arg_p.
1391         * config/arm/arm.c (arm_output_function_prologue): Add a cast to
1392         HOST_WIDE_INT.
1393         * config/avr/avr.c (avr_outgoing_args_size): Likewise.
1394         * config/microblaze/microblaze.c (microblaze_function_prologue):
1395         Likewise.
1396         * config/cr16/cr16.c (cr16_return_pops_args): Update for new
1397         TARGET_RETURN_POPS_ARGS interface.
1398         (cr16_compute_frame, cr16_initial_elimination_offset): Add casts
1399         to HOST_WIDE_INT.
1400         * config/ft32/ft32.c (ft32_compute_frame): Likewise.
1401         * config/i386/i386.c (ix86_return_pops_args): Update for new
1402         TARGET_RETURN_POPS_ARGS interface.
1403         (ix86_expand_split_stack_prologue): Add a cast to HOST_WIDE_INT.
1404         * config/moxie/moxie.c (moxie_compute_frame): Likewise.
1405         * config/m68k/m68k.c (m68k_return_pops_args): Update for new
1406         TARGET_RETURN_POPS_ARGS interface.
1407         * config/vax/vax.c (vax_return_pops_args): Likewise.
1408         * config/pa/pa.h (STACK_POINTER_OFFSET): Add a cast to poly_int64.
1409         (EXIT_IGNORE_STACK): Update reference to crtl->outgoing_args_size.
1410         * config/arm/arm.h (CALLER_INTERWORKING_SLOT_SIZE): Likewise.
1411         * config/powerpcspe/aix.h (STACK_DYNAMIC_OFFSET): Likewise.
1412         * config/powerpcspe/darwin.h (STACK_DYNAMIC_OFFSET): Likewise.
1413         * config/powerpcspe/powerpcspe.h (STACK_DYNAMIC_OFFSET): Likewise.
1414         * config/rs6000/aix.h (STACK_DYNAMIC_OFFSET): Likewise.
1415         * config/rs6000/darwin.h (STACK_DYNAMIC_OFFSET): Likewise.
1416         * config/rs6000/rs6000.h (STACK_DYNAMIC_OFFSET): Likewise.
1417         * dojump.h (saved_pending_stack_adjust): Change x_pending_stack_adjust
1418         and x_stack_pointer_delta from int to poly_int64.
1419         * dojump.c (do_pending_stack_adjust): Update accordingly.
1420         * explow.c (allocate_dynamic_stack_space): Handle polynomial
1421         stack_pointer_deltas.
1422         * function.c (STACK_DYNAMIC_OFFSET): Add a cast to poly_int64.
1423         (pad_to_arg_alignment): Track polynomial offsets.
1424         (assign_parm_find_stack_rtl): Likewise.
1425         (assign_parms, locate_and_pad_parm): Handle polynomial argument sizes.
1426         * toplev.c (output_stack_usage): Update reference to
1427         current_function_pushed_stack_size.
1429 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1430             Alan Hayward  <alan.hayward@arm.com>
1431             David Sherwood  <david.sherwood@arm.com>
1433         * function.c (in_arg_offset, var_offset, dynamic_offset)
1434         (out_arg_offset, cfa_offset): Change from int to poly_int64.
1435         (instantiate_new_reg): Return the new offset as a poly_int64_pod
1436         rather than a HOST_WIDE_INT.
1437         (instantiate_virtual_regs_in_rtx): Track polynomial offsets.
1438         (instantiate_virtual_regs_in_insn): Likewise.
1440 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1441             Alan Hayward  <alan.hayward@arm.com>
1442             David Sherwood  <david.sherwood@arm.com>
1444         * rtl.h (get_args_size, add_args_size_note): New functions.
1445         (find_args_size_adjust): Return a poly_int64 rather than a
1446         HOST_WIDE_INT.
1447         (fixup_args_size_notes): Likewise.  Make the same change to the
1448         end_args_size parameter.
1449         * rtlanal.c (get_args_size, add_args_size_note): New functions.
1450         * builtins.c (expand_builtin_trap): Use add_args_size_note.
1451         * calls.c (emit_call_1): Likewise.
1452         * explow.c (adjust_stack_1): Likewise.
1453         * cfgcleanup.c (old_insns_match_p): Update use of
1454         find_args_size_adjust.
1455         * combine.c (distribute_notes): Track polynomial arg sizes.
1456         * dwarf2cfi.c (dw_trace_info): Change beg_true_args_size,
1457         end_true_args_size, beg_delay_args_size and end_delay_args_size
1458         from HOST_WIDE_INT to poly_int64.
1459         (add_cfi_args_size): Take the args_size as a poly_int64 rather
1460         than a HOST_WIDE_INT.
1461         (notice_args_size, notice_eh_throw, maybe_record_trace_start)
1462         (maybe_record_trace_start_abnormal, scan_trace, connect_traces): Track
1463         polynomial arg sizes.
1464         * emit-rtl.c (try_split): Use get_args_size.
1465         * recog.c (peep2_attempt): Likewise.
1466         * reload1.c (reload_as_needed): Likewise.
1467         * expr.c (find_args_size_adjust): Return the adjustment as a
1468         poly_int64 rather than a HOST_WIDE_INT.
1469         (fixup_args_size_notes): Change end_args_size from a HOST_WIDE_INT
1470         to a poly_int64 and change the return type in the same way.
1471         (emit_single_push_insn): Track polynomial arg sizes.
1473 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1474             Alan Hayward  <alan.hayward@arm.com>
1475             David Sherwood  <david.sherwood@arm.com>
1477         * expr.h (push_block, emit_push_insn): Change the "extra" parameter
1478         from HOST_WIDE_INT to poly_int64.
1479         * expr.c (push_block, emit_push_insn): Likewise.
1481 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1482             Alan Hayward  <alan.hayward@arm.com>
1483             David Sherwood  <david.sherwood@arm.com>
1485         * function.h (frame_space): Change start and length from HOST_WIDE_INT
1486         to poly_int64.
1487         (get_frame_size): Return the size as a poly_int64 rather than a
1488         HOST_WIDE_INT.
1489         (frame_offset_overflow): Take the offset as a poly_int64 rather
1490         than a HOST_WIDE_INT.
1491         (assign_stack_local_1, assign_stack_local, assign_stack_temp_for_type)
1492         (assign_stack_temp): Likewise for the size.
1493         * function.c (get_frame_size): Return a poly_int64 rather than
1494         a HOST_WIDE_INT.
1495         (frame_offset_overflow): Take the offset as a poly_int64 rather
1496         than a HOST_WIDE_INT.
1497         (try_fit_stack_local): Take the start, length and size as poly_int64s
1498         rather than HOST_WIDE_INTs.  Return the offset as a poly_int64_pod
1499         rather than a HOST_WIDE_INT.
1500         (add_frame_space): Take the start and end as poly_int64s rather than
1501         HOST_WIDE_INTs.
1502         (assign_stack_local_1, assign_stack_local, assign_stack_temp_for_type)
1503         (assign_stack_temp): Likewise for the size.
1504         (temp_slot): Change size, base_offset and full_size from HOST_WIDE_INT
1505         to poly_int64.
1506         (find_temp_slot_from_address): Handle polynomial offsets.
1507         (combine_temp_slots): Likewise.
1508         * emit-rtl.h (rtl_data::x_frame_offset): Change from HOST_WIDE_INT
1509         to poly_int64.
1510         * cfgexpand.c (alloc_stack_frame_space): Return the offset as a
1511         poly_int64 rather than a HOST_WIDE_INT.
1512         (expand_one_stack_var_at): Take the offset as a poly_int64 rather
1513         than a HOST_WIDE_INT.
1514         (expand_stack_vars, expand_one_stack_var_1, expand_used_vars): Handle
1515         polynomial frame offsets.
1516         * config/m32r/m32r-protos.h (m32r_compute_frame_size): Take the size
1517         as a poly_int64 rather than an int.
1518         * config/m32r/m32r.c (m32r_compute_frame_size): Likewise.
1519         * config/v850/v850-protos.h (compute_frame_size): Likewise.
1520         * config/v850/v850.c (compute_frame_size): Likewise.
1521         * config/xtensa/xtensa-protos.h (compute_frame_size): Likewise.
1522         * config/xtensa/xtensa.c (compute_frame_size): Likewise.
1523         * config/pa/pa-protos.h (pa_compute_frame_size): Likewise.
1524         * config/pa/pa.c (pa_compute_frame_size): Likewise.
1525         * explow.h (get_dynamic_stack_base): Take the offset as a poly_int64
1526         rather than a HOST_WIDE_INT.
1527         * explow.c (get_dynamic_stack_base): Likewise.
1528         * final.c (final_start_function): Use the constant lower bound
1529         of the frame size for -Wframe-larger-than.
1530         * ira.c (do_reload): Adjust for new get_frame_size return type.
1531         * lra.c (lra): Likewise.
1532         * reload1.c (reload): Likewise.
1533         * config/avr/avr.c (avr_asm_function_end_prologue): Likewise.
1534         * config/pa/pa.h (EXIT_IGNORE_STACK): Likewise.
1535         * rtlanal.c (get_initial_register_offset): Return the offset as
1536         a poly_int64 rather than a HOST_WIDE_INT.
1538 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1539             Alan Hayward  <alan.hayward@arm.com>
1540             David Sherwood  <david.sherwood@arm.com>
1542         * reload1.c (elim_table): Change initial_offset, offset and
1543         previous_offset from HOST_WIDE_INT to poly_int64_pod.
1544         (offsets_at): Change the target array's element type from
1545         HOST_WIDE_INT to poly_int64_pod.
1546         (set_label_offsets, eliminate_regs_1, eliminate_regs_in_insn)
1547         (elimination_costs_in_insn, update_eliminable_offsets)
1548         (verify_initial_elim_offsets, set_offsets_for_label)
1549         (init_eliminable_invariants): Update after above changes.
1551 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1552             Alan Hayward  <alan.hayward@arm.com>
1553             David Sherwood  <david.sherwood@arm.com>
1555         * reload.h (reload::inc): Change from an int to a poly_int64_pod.
1556         * reload.c (combine_reloads, debug_reload_to_stream): Likewise.
1557         (decomposition): Change start and end from HOST_WIDE_INT
1558         to poly_int64_pod.
1559         (decompose, immune_p): Update accordingly.
1560         (find_inc_amount): Return a poly_int64 rather than an int.
1561         * reload1.c (inc_for_reload): Take the inc_amount as a poly_int64
1562         rather than an int.
1564 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1565             Alan Hayward  <alan.hayward@arm.com>
1566             David Sherwood  <david.sherwood@arm.com>
1568         * tree.h (get_inner_reference): Return the bitsize and bitpos
1569         as poly_int64_pods rather than HOST_WIDE_INT.
1570         * fold-const.h (ptr_difference_const): Return the pointer difference
1571         as a poly_int64_pod rather than a HOST_WIDE_INT.
1572         * expr.c (get_inner_reference): Return the bitsize and bitpos
1573         as poly_int64_pods rather than HOST_WIDE_INT.
1574         (expand_expr_addr_expr_1, expand_expr_real_1): Track polynomial
1575         offsets and sizes.
1576         * fold-const.c (make_bit_field_ref): Take the bitpos as a poly_int64
1577         rather than a HOST_WIDE_INT.  Update call to get_inner_reference.
1578         (optimize_bit_field_compare): Update call to get_inner_reference.
1579         (decode_field_reference): Likewise.
1580         (fold_unary_loc): Track polynomial offsets and sizes.
1581         (split_address_to_core_and_offset): Return the bitpos as a
1582         poly_int64_pod rather than a HOST_WIDE_INT.
1583         (ptr_difference_const): Likewise for the pointer difference.
1584         * asan.c (instrument_derefs): Track polynomial offsets and sizes.
1585         * config/mips/mips.c (r10k_safe_mem_expr_p): Likewise.
1586         * dbxout.c (dbxout_expand_expr): Likewise.
1587         * dwarf2out.c (loc_list_for_address_of_addr_expr_of_indirect_ref)
1588         (loc_list_from_tree_1, fortran_common): Likewise.
1589         * gimple-laddress.c (pass_laddress::execute): Likewise.
1590         * gimple-ssa-store-merging.c (find_bswap_or_nop_load): Likewise.
1591         * gimplify.c (gimplify_scan_omp_clauses): Likewise.
1592         * simplify-rtx.c (delegitimize_mem_from_attrs): Likewise.
1593         * tree-affine.c (tree_to_aff_combination): Likewise.
1594         (get_inner_reference_aff): Likewise.
1595         * tree-data-ref.c (split_constant_offset_1): Likewise.
1596         (dr_analyze_innermost): Likewise.
1597         * tree-scalar-evolution.c (interpret_rhs_expr): Likewise.
1598         * tree-sra.c (ipa_sra_check_caller): Likewise.
1599         * tree-vect-data-refs.c (vect_check_gather_scatter): Likewise.
1600         * ubsan.c (maybe_instrument_pointer_overflow): Likewise.
1601         (instrument_bool_enum_load, instrument_object_size): Likewise.
1602         * gimple-ssa-strength-reduction.c (slsr_process_ref): Update call
1603         to get_inner_reference.
1604         * hsa-gen.c (gen_hsa_addr): Likewise.
1605         * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Likewise.
1606         * tsan.c (instrument_expr): Likewise.
1607         * match.pd: Update call to ptr_difference_const.
1609 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1610             Alan Hayward  <alan.hayward@arm.com>
1611             David Sherwood  <david.sherwood@arm.com>
1613         * fold-const.c (fold_comparison): Track sizes and offsets as
1614         poly_int64s rather than HOST_WIDE_INTs when folding address
1615         comparisons.
1617 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1618             Alan Hayward  <alan.hayward@arm.com>
1619             David Sherwood  <david.sherwood@arm.com>
1621         * expr.h (get_bit_range): Return the bitstart and bitend as
1622         poly_uint64s rather than unsigned HOST_WIDE_INTs.  Return the bitpos
1623         as a poly_int64 rather than a HOST_WIDE_INT.
1624         * expr.c (get_bit_range): Likewise.
1625         (expand_assignment): Update call accordingly.
1626         * fold-const.c (optimize_bit_field_compare): Likewise.
1628 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1629             Alan Hayward  <alan.hayward@arm.com>
1630             David Sherwood  <david.sherwood@arm.com>
1632         * config/aarch64/aarch64-protos.h (aarch64_addr_query_type): New enum.
1633         (aarch64_legitimate_address_p): Use it instead of an rtx code,
1634         as an optional final parameter.
1635         * config/aarch64/aarch64.c (aarch64_classify_address): Likewise.
1636         (aarch64_legitimate_address_p): Likewise.
1637         (aarch64_print_address_internal): Take an aarch64_addr_query_type
1638         instead of an rtx code.
1639         (aarch64_address_valid_for_prefetch_p): Update calls accordingly.
1640         (aarch64_legitimate_address_hook_p): Likewise.
1641         (aarch64_print_ldpstp_address): Likewise.
1642         (aarch64_print_operand_address): Likewise.
1643         (aarch64_address_cost): Likewise.
1644         * config/aarch64/constraints.md (Uml, Umq, Ump, Utq): Likewise.
1645         * config/aarch64/predicates.md (aarch64_mem_pair_operand): Likewise.
1646         (aarch64_mem_pair_lanes_operand): Likewise.
1648 2017-12-20  Richard Biener  <rguenther@suse.de>
1650         * tree-ssa-dom.c (dom_opt_dom_walker::optimize_stmt): Call
1651         update_stmt_if_modified.
1653 2017-12-20  Wilco Dijkstra  <wdijkstr@arm.com>
1655         PR tree-optimization/83491
1656         * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Check for SSA_NAME
1657         before walking uses.  Improve coding style and comments.
1659 2017-12-20  Tom de Vries  <tom@codesourcery.com>
1661         * gimple-fold.c (fold_internal_goacc_dim): Simplify.
1663 2017-12-20  Jakub Jelinek  <jakub@redhat.com>
1665         PR ipa/83506
1666         * ipa-fnsummary.c (pass_data_ipa_free_fn_summary): Use 0 for
1667         todo_flags_finish.
1668         (pass_ipa_free_fn_summary): Add small_p private data member,
1669         initialize to false in the ctor.
1670         (pass_ipa_free_fn_summary::clone,
1671         pass_ipa_free_fn_summary::set_pass_param,
1672         pass_ipa_free_fn_summary::gate): New methods.
1673         (pass_ipa_free_fn_summary::execute): Return TODO_remove_functions
1674         | TODO_dump_symtab if small_p.
1675         * passes.def: Add true parm for the existing pass_ipa_free_fn_summary
1676         entry and add another instance of the pass with false parm after
1677         ipa-pure-const.
1678         * ipa-pure-const.c (pass_ipa_pure_const): Don't call
1679         ipa_free_fn_summary here.
1681 2017-12-20  Paolo Carlini  <paolo.carlini@oracle.com>
1683         * gimplify.c (gimplify_return_expr): Remove dead error_mark_node check.
1685 2017-12-20  Martin Sebor  <msebor@redhat.com>
1687         PR testsuite/83131
1688         * builtins.c (expand_builtin_strlen): Use get_callee_fndecl.
1689         (expand_builtin_strcmp): Call maybe_warn_nonstring_arg.
1690         (expand_builtin_strncmp): Same.
1692 2017-12-20  Alexandre Oliva  <aoliva@redhat.com>
1694         PR bootstrap/83396
1695         * cfgexpand.c (label_rtx_for_bb): Revert SFN changes that
1696         allowed debug stmts before labels.
1697         (expand_gimple_basic_block): Likewise.
1698         * gimple-iterator.c (gimple_find_edge_insert_loc): Likewise.
1699         * gimple-iterator.h (gsi_after_labels): Likewise.
1700         * tree-cfgcleanup (remove_forwarder_block): Likewise, but
1701         rename reused variable, and simplify using gsi_move_before.
1702         * tree-ssa-tail-merge.c (find_duplicate): Likewise.
1703         * tree-cfg.c (make_edges, cleanup_dead_labels): Likewise.
1704         (gimple_can_merge_blocks_p, verify_gimple_in_cfg): Likewise.
1705         (gimple_verify_flow_info, gimple_block_label): Likewise.
1706         (make_blocks): Move debug markers after adjacent labels.
1707         * cfgrtl.c (skip_insns_after_block): Revert SFN changes that
1708         allowed debug insns outside blocks.
1709         * df-scan.c (df_insn_delete): Likewise.
1710         * lra-constraints.c (update_ebb_live_info): Likewise.
1711         * var-tracking.c (get_first_insn, vt_emit_notes): Likewise.
1712         (vt_initialize, delete_vta_debug_insns): Likewise.
1713         (reemit_marker_as_note): Drop BB parm.  Adjust callers.
1715 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
1716             Alan Hayward  <alan.hayward@arm.com>
1717             David Sherwood  <david.sherwood@arm.com>
1719         * poly-int-types.h (round_down_to_byte_boundary): New macro.
1720         (round_up_to_byte_boundary): Likewise.
1721         * expr.h (get_bit_range): Add temporary shim.
1722         * gimple-ssa-store-merging.c (store_operand_info): Change the
1723         bitsize, bitpos, bitregion_start and bitregion_end fields from
1724         unsigned HOST_WIDE_INT to poly_uint64.
1725         (merged_store_group): Likewise load_align_base.
1726         (compatible_load_p, compatible_load_p): Update accordingly.
1727         (imm_store_chain_info::coalesce_immediate_stores): Likewise.
1728         (split_group, imm_store_chain_info::output_merged_store): Likewise.
1729         (mem_valid_for_store_merging): Return the bitsize, bitpos,
1730         bitregion_start and bitregion_end as poly_uint64s rather than
1731         unsigned HOST_WIDE_INTs.  Track polynomial offsets internally.
1732         (handled_load): Take the bitsize, bitpos,
1733         bitregion_start and bitregion_end as poly_uint64s rather than
1734         unsigned HOST_WIDE_INTs.
1735         (pass_store_merging::process_store): Update call to
1736         mem_valid_for_store_merging.
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         * builtins.c (get_object_alignment_2): Track polynomial offsets
1743         and sizes.  Update the alignment handling.
1745 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
1746             Alan Hayward  <alan.hayward@arm.com>
1747             David Sherwood  <david.sherwood@arm.com>
1749         * tree.h (get_inner_reference): Add a version that returns the
1750         offset and size as poly_int64_pods rather than HOST_WIDE_INTs.
1751         * cfgexpand.c (expand_debug_expr): Track polynomial offsets.  Simply
1752         the case in which bitpos is not associated with the first byte.
1754 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
1755             Alan Hayward  <alan.hayward@arm.com>
1756             David Sherwood  <david.sherwood@arm.com>
1758         * tree-affine.h (get_inner_reference_aff): Return the size as a
1759         poly_widest_int.
1760         * tree-affine.c (get_inner_reference_aff): Likewise.
1761         * tree-data-ref.c (dr_may_alias_p): Update accordingly.
1762         * tree-ssa-loop-im.c (mem_refs_may_alias_p): Likewise.
1764 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
1765             Alan Hayward  <alan.hayward@arm.com>
1766             David Sherwood  <david.sherwood@arm.com>
1768         * fold-const.c (pointer_may_wrap_p): Take the offset as a
1769         HOST_WIDE_INT rather than a poly_int64.
1771 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
1772             Alan Hayward  <alan.hayward@arm.com>
1773             David Sherwood  <david.sherwood@arm.com>
1775         * gimple-ssa-store-merging.c (symbolic_number::bytepos): Change from
1776         HOST_WIDE_INT to poly_int64_pod.
1777         (perform_symbolic_merge): Update accordingly.
1778         (bswap_replace): Likewise.
1780 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
1781             Alan Hayward  <alan.hayward@arm.com>
1782             David Sherwood  <david.sherwood@arm.com>
1784         * tree-affine.h (aff_tree::offset): Change from widest_int
1785         to poly_widest_int.
1786         (wide_int_ext_for_comb): Delete.
1787         (aff_combination_const, aff_comb_cannot_overlap_p): Take the
1788         constants as poly_widest_int rather than widest_int.
1789         (aff_combination_constant_multiple_p): Return the multiplier
1790         as a poly_widest_int.
1791         (aff_combination_zero_p, aff_combination_singleton_var_p): Handle
1792         polynomial offsets.
1793         * tree-affine.c (wide_int_ext_for_comb): Make original widest_int
1794         version static and add an overload for poly_widest_int.
1795         (aff_combination_const, aff_combination_add_cst)
1796         (wide_int_constant_multiple_p, aff_comb_cannot_overlap_p): Take
1797         the constants as poly_widest_int rather than widest_int.
1798         (tree_to_aff_combination): Generalize INTEGER_CST case to
1799         poly_int_tree_p.
1800         (aff_combination_to_tree): Track offsets as poly_widest_ints.
1801         (aff_combination_add_product, aff_combination_mult): Handle
1802         polynomial offsets.
1803         (aff_combination_constant_multiple_p): Return the multiplier
1804         as a poly_widest_int.
1805         * tree-predcom.c (determine_offset): Return the offset as a
1806         poly_widest_int.
1807         (split_data_refs_to_components, suitable_component_p): Update
1808         accordingly.
1809         (valid_initializer_p): Update call to
1810         aff_combination_constant_multiple_p.
1811         * tree-ssa-address.c (addr_to_parts): Handle polynomial offsets.
1812         * tree-ssa-loop-ivopts.c (get_address_cost_ainc): Take the step
1813         as a poly_int64 rather than a HOST_WIDE_INT.
1814         (get_address_cost): Handle polynomial offsets.
1815         (iv_elimination_compare_lt): Likewise.
1816         (rewrite_use_nonlinear_expr): Likewise.
1818 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
1819             Alan Hayward  <alan.hayward@arm.com>
1820             David Sherwood  <david.sherwood@arm.com>
1822         * tree-dfa.h (get_addr_base_and_unit_offset_1): Return the offset
1823         as a poly_int64_pod rather than a HOST_WIDE_INT.
1824         (get_addr_base_and_unit_offset): Likewise.
1825         * tree-dfa.c (get_addr_base_and_unit_offset_1): Likewise.
1826         (get_addr_base_and_unit_offset): Likewise.
1827         * doc/match-and-simplify.texi: Change off from HOST_WIDE_INT
1828         to poly_int64 in example.
1829         * fold-const.c (fold_binary_loc): Update call to
1830         get_addr_base_and_unit_offset.
1831         * gimple-fold.c (gimple_fold_builtin_memory_op): Likewise.
1832         (maybe_canonicalize_mem_ref_addr): Likewise.
1833         (gimple_fold_stmt_to_constant_1): Likewise.
1834         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref):
1835         Likewise.
1836         * ipa-param-manipulation.c (ipa_modify_call_arguments): Likewise.
1837         * match.pd: Likewise.
1838         * omp-low.c (lower_omp_target): Likewise.
1839         * tree-sra.c (build_ref_for_offset): Likewise.
1840         (build_debug_ref_for_model): Likewise.
1841         * tree-ssa-address.c (maybe_fold_tmr): Likewise.
1842         * tree-ssa-alias.c (ao_ref_init_from_ptr_and_size): Likewise.
1843         * tree-ssa-ccp.c (optimize_memcpy): Likewise.
1844         * tree-ssa-forwprop.c (forward_propagate_addr_expr_1): Likewise.
1845         (constant_pointer_difference): Likewise.
1846         * tree-ssa-loop-niter.c (expand_simple_operations): Likewise.
1847         * tree-ssa-phiopt.c (jump_function_from_stmt): Likewise.
1848         * tree-ssa-pre.c (create_component_ref_by_pieces_1): Likewise.
1849         * tree-ssa-sccvn.c (vn_reference_fold_indirect): Likewise.
1850         (vn_reference_maybe_forwprop_address, vn_reference_lookup_3): Likewise.
1851         (set_ssa_val_to): Likewise.
1852         * tree-ssa-strlen.c (get_addr_stridx, addr_stridxptr)
1853         (maybe_diag_stxncpy_trunc): Likewise.
1854         * tree-vrp.c (vrp_prop::check_array_ref): Likewise.
1855         * tree.c (build_simple_mem_ref_loc): Likewise.
1856         (array_at_struct_end_p): Likewise.
1858 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
1859             Alan Hayward  <alan.hayward@arm.com>
1860             David Sherwood  <david.sherwood@arm.com>
1862         * tree-dfa.h (get_ref_base_and_extent): Return the base, size and
1863         max_size as poly_int64_pods rather than HOST_WIDE_INTs.
1864         (get_ref_base_and_extent_hwi): Declare.
1865         * tree-dfa.c (get_ref_base_and_extent): Return the base, size and
1866         max_size as poly_int64_pods rather than HOST_WIDE_INTs.
1867         (get_ref_base_and_extent_hwi): New function.
1868         * cfgexpand.c (expand_debug_expr): Update call to
1869         get_ref_base_and_extent.
1870         * dwarf2out.c (add_var_loc_to_decl): Likewise.
1871         * gimple-fold.c (get_base_constructor): Return the offset as a
1872         poly_int64_pod rather than a HOST_WIDE_INT.
1873         (fold_const_aggregate_ref_1): Track polynomial sizes and offsets.
1874         * ipa-polymorphic-call.c
1875         (ipa_polymorphic_call_context::set_by_invariant)
1876         (extr_type_from_vtbl_ptr_store): Track polynomial offsets.
1877         (ipa_polymorphic_call_context::ipa_polymorphic_call_context)
1878         (check_stmt_for_type_change): Use get_ref_base_and_extent_hwi
1879         rather than get_ref_base_and_extent.
1880         (ipa_polymorphic_call_context::get_dynamic_type): Likewise.
1881         * ipa-prop.c (ipa_load_from_parm_agg, compute_complex_assign_jump_func)
1882         (get_ancestor_addr_info, determine_locally_known_aggregate_parts):
1883         Likewise.
1884         * ipa-param-manipulation.c (ipa_get_adjustment_candidate): Update
1885         call to get_ref_base_and_extent.
1886         * tree-sra.c (create_access, get_access_for_expr): Likewise.
1887         * tree-ssa-alias.c (ao_ref_base, aliasing_component_refs_p)
1888         (stmt_kills_ref_p): Likewise.
1889         * tree-ssa-dce.c (mark_aliased_reaching_defs_necessary_1): Likewise.
1890         * tree-ssa-scopedtables.c (avail_expr_hash, equal_mem_array_ref_p):
1891         Likewise.
1892         * tree-ssa-sccvn.c (vn_reference_lookup_3): Likewise.
1893         Use get_ref_base_and_extent_hwi rather than get_ref_base_and_extent
1894         when calling native_encode_expr.
1895         * tree-ssa-structalias.c (get_constraint_for_component_ref): Update
1896         call to get_ref_base_and_extent.
1897         (do_structure_copy): Use get_ref_base_and_extent_hwi rather than
1898         get_ref_base_and_extent.
1899         * var-tracking.c (track_expr_p): Likewise.
1901 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
1902             Alan Hayward  <alan.hayward@arm.com>
1903             David Sherwood  <david.sherwood@arm.com>
1905         * ipa-param-manipulation.h (ipa_parm_adjustment::offset): Change from
1906         HOST_WIDE_INT to poly_int64_pod.
1907         * ipa-param-manipulation.c (ipa_modify_call_arguments): Track
1908         polynomail parameter offsets.
1910 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
1911             Alan Hayward  <alan.hayward@arm.com>
1912             David Sherwood  <david.sherwood@arm.com>
1914         * gengtype.c (main): Handle poly_int64_pod.
1915         * dwarf2out.h (dw_cfi_oprnd_cfa_loc): New dw_cfi_oprnd_type.
1916         (dw_cfi_oprnd::dw_cfi_cfa_loc): New field.
1917         (dw_cfa_location::offset, dw_cfa_location::base_offset): Change
1918         from HOST_WIDE_INT to poly_int64_pod.
1919         * dwarf2cfi.c (queued_reg_save::cfa_offset): Likewise.
1920         (copy_cfa): New function.
1921         (lookup_cfa_1): Use the cached dw_cfi_cfa_loc, if it exists.
1922         (cfi_oprnd_equal_p): Handle dw_cfi_oprnd_cfa_loc.
1923         (cfa_equal_p, dwarf2out_frame_debug_adjust_cfa)
1924         (dwarf2out_frame_debug_cfa_offset, dwarf2out_frame_debug_expr)
1925         (initial_return_save): Treat offsets as poly_ints.
1926         (def_cfa_0): Likewise.  Cache the CFA in dw_cfi_cfa_loc if either
1927         offset is nonconstant.
1928         (reg_save): Take the offset as a poly_int64.  Fall back to
1929         DW_CFA_expression for nonconstant offsets.
1930         (queue_reg_save): Take the offset as a poly_int64.
1931         * dwarf2out.c (dw_cfi_oprnd2_desc): Handle DW_CFA_def_cfa_expression.
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         * rtl.h (operand_subword, operand_subword_force): Take the offset
1938         as a poly_uint64 an unsigned int.
1939         * emit-rtl.c (operand_subword, operand_subword_force): Likewise.
1941 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
1942             Alan Hayward  <alan.hayward@arm.com>
1943             David Sherwood  <david.sherwood@arm.com>
1945         * doc/rtl.texi: Update documentation of SUBREG_BYTE.  Document the
1946         'p' format code.  Use INT_LIST rather than SUBREG as the example of
1947         a code with an XINT and an XEXP.  Remove the implication that
1948         accessing an rtx field using XINT is expected to work.
1949         * rtl.def (SUBREG): Change format from "ei" to "ep".
1950         * rtl.h (rtunion::rt_subreg): New field.
1951         (XCSUBREG): New macro.
1952         (SUBREG_BYTE): Use it.
1953         (subreg_shape): Change offset from an unsigned int to a poly_uint16.
1954         Update constructor accordingly.
1955         (subreg_shape::operator ==): Update accordingly.
1956         (subreg_shape::unique_id): Return an unsigned HOST_WIDE_INT rather
1957         than an unsigned int.
1958         (subreg_lsb, subreg_lowpart_offset, subreg_highpart_offset): Return
1959         a poly_uint64 rather than an unsigned int.
1960         (subreg_lsb_1): Likewise.  Take the offset as a poly_uint64 rather
1961         than an unsigned int.
1962         (subreg_size_offset_from_lsb, subreg_size_lowpart_offset)
1963         (subreg_size_highpart_offset): Return a poly_uint64 rather than
1964         an unsigned int.  Take the sizes as poly_uint64s.
1965         (subreg_offset_from_lsb): Return a poly_uint64 rather than
1966         an unsigned int.  Take the shift as a poly_uint64 rather than
1967         an unsigned int.
1968         (subreg_regno_offset, subreg_offset_representable_p): Take the offset
1969         as a poly_uint64 rather than an unsigned int.
1970         (simplify_subreg_regno): Likewise.
1971         (byte_lowpart_offset): Return the memory offset as a poly_int64
1972         rather than an int.
1973         (subreg_memory_offset): Likewise.  Take the subreg offset as a
1974         poly_uint64 rather than an unsigned int.
1975         (simplify_subreg, simplify_gen_subreg, subreg_get_info)
1976         (gen_rtx_SUBREG, validate_subreg): Take the subreg offset as a
1977         poly_uint64 rather than an unsigned int.
1978         * rtl.c (rtx_format): Describe 'p' in comment.
1979         (copy_rtx, rtx_equal_p_cb, rtx_equal_p): Handle 'p'.
1980         * emit-rtl.c (validate_subreg, gen_rtx_SUBREG): Take the subreg
1981         offset as a poly_uint64 rather than an unsigned int.
1982         (byte_lowpart_offset): Return the memory offset as a poly_int64
1983         rather than an int.
1984         (subreg_memory_offset): Likewise.  Take the subreg offset as a
1985         poly_uint64 rather than an unsigned int.
1986         (subreg_size_lowpart_offset, subreg_size_highpart_offset): Take the
1987         mode sizes as poly_uint64s rather than unsigned ints.  Return a
1988         poly_uint64 rather than an unsigned int.
1989         (subreg_lowpart_p): Treat subreg offsets as poly_ints.
1990         (copy_insn_1): Handle 'p'.
1991         * rtlanal.c (set_noop_p): Treat subregs offsets as poly_uint64s.
1992         (subreg_lsb_1): Take the subreg offset as a poly_uint64 rather than
1993         an unsigned int.  Return the shift in the same way.
1994         (subreg_lsb): Return the shift as a poly_uint64 rather than an
1995         unsigned int.
1996         (subreg_size_offset_from_lsb): Take the sizes and shift as
1997         poly_uint64s rather than unsigned ints.  Return the offset as
1998         a poly_uint64.
1999         (subreg_get_info, subreg_regno_offset, subreg_offset_representable_p)
2000         (simplify_subreg_regno): Take the offset as a poly_uint64 rather than
2001         an unsigned int.
2002         * rtlhash.c (add_rtx): Handle 'p'.
2003         * genemit.c (gen_exp): Likewise.
2004         * gengenrtl.c (type_from_format, gendef): Likewise.
2005         * gensupport.c (subst_pattern_match, get_alternatives_number)
2006         (collect_insn_data, alter_predicate_for_insn, alter_constraints)
2007         (subst_dup): Likewise.
2008         * gengtype.c (adjust_field_rtx_def): Likewise.
2009         * genrecog.c (find_operand, find_matching_operand, validate_pattern)
2010         (match_pattern_2): Likewise.
2011         (rtx_test::SUBREG_FIELD): New rtx_test::kind_enum.
2012         (rtx_test::subreg_field): New function.
2013         (operator ==, safe_to_hoist_p, transition_parameter_type)
2014         (print_nonbool_test, print_test): Handle SUBREG_FIELD.
2015         * genattrtab.c (attr_rtx_1): Say that 'p' is deliberately not handled.
2016         * genpeep.c (match_rtx): Likewise.
2017         * print-rtl.c (print_poly_int): Include if GENERATOR_FILE too.
2018         (rtx_writer::print_rtx_operand): Handle 'p'.
2019         (print_value): Handle SUBREG.
2020         * read-rtl.c (apply_int_iterator): Likewise.
2021         (rtx_reader::read_rtx_operand): Handle 'p'.
2022         * alias.c (rtx_equal_for_memref_p): Likewise.
2023         * cselib.c (rtx_equal_for_cselib_1, cselib_hash_rtx): Likewise.
2024         * caller-save.c (replace_reg_with_saved_mem): Treat subreg offsets
2025         as poly_ints.
2026         * calls.c (expand_call): Likewise.
2027         * combine.c (combine_simplify_rtx, expand_field_assignment): Likewise.
2028         (make_extraction, gen_lowpart_for_combine): Likewise.
2029         * loop-invariant.c (hash_invariant_expr_1, invariant_expr_equal_p):
2030         Likewise.
2031         * cse.c (remove_invalid_subreg_refs): Take the offset as a poly_uint64
2032         rather than an unsigned int.  Treat subreg offsets as poly_ints.
2033         (exp_equiv_p): Handle 'p'.
2034         (hash_rtx_cb): Likewise.  Treat subreg offsets as poly_ints.
2035         (equiv_constant, cse_insn): Treat subreg offsets as poly_ints.
2036         * dse.c (find_shift_sequence): Likewise.
2037         * dwarf2out.c (rtl_for_decl_location): Likewise.
2038         * expmed.c (extract_low_bits): Likewise.
2039         * expr.c (emit_group_store, undefined_operand_subword_p): Likewise.
2040         (expand_expr_real_2): Likewise.
2041         * final.c (alter_subreg): Likewise.
2042         (leaf_renumber_regs_insn): Handle 'p'.
2043         * function.c (assign_parm_find_stack_rtl, assign_parm_setup_stack):
2044         Treat subreg offsets as poly_ints.
2045         * fwprop.c (forward_propagate_and_simplify): Likewise.
2046         * ifcvt.c (noce_emit_move_insn, noce_emit_cmove): Likewise.
2047         * ira.c (get_subreg_tracking_sizes): Likewise.
2048         * ira-conflicts.c (go_through_subreg): Likewise.
2049         * ira-lives.c (process_single_reg_class_operands): Likewise.
2050         * jump.c (rtx_renumbered_equal_p): Likewise.  Handle 'p'.
2051         * lower-subreg.c (simplify_subreg_concatn): Take the subreg offset
2052         as a poly_uint64 rather than an unsigned int.
2053         (simplify_gen_subreg_concatn, resolve_simple_move): Treat
2054         subreg offsets as poly_ints.
2055         * lra-constraints.c (operands_match_p): Handle 'p'.
2056         (match_reload, curr_insn_transform): Treat subreg offsets as poly_ints.
2057         * lra-spills.c (assign_mem_slot): Likewise.
2058         * postreload.c (move2add_valid_value_p): Likewise.
2059         * recog.c (general_operand, indirect_operand): Likewise.
2060         * regcprop.c (copy_value, maybe_mode_change): Likewise.
2061         (copyprop_hardreg_forward_1): Likewise.
2062         * reginfo.c (simplifiable_subregs_hasher::hash, simplifiable_subregs)
2063         (record_subregs_of_mode): Likewise.
2064         * rtlhooks.c (gen_lowpart_general, gen_lowpart_if_possible): Likewise.
2065         * reload.c (operands_match_p): Handle 'p'.
2066         (find_reloads_subreg_address): Treat subreg offsets as poly_ints.
2067         * reload1.c (alter_reg, choose_reload_regs): Likewise.
2068         (compute_reload_subreg_offset): Likewise, and return an poly_int64.
2069         * simplify-rtx.c (simplify_truncation, simplify_binary_operation_1):
2070         (test_vector_ops_duplicate): Treat subreg offsets as poly_ints.
2071         (simplify_const_poly_int_tests<N>::run): Likewise.
2072         (simplify_subreg, simplify_gen_subreg): Take the subreg offset as
2073         a poly_uint64 rather than an unsigned int.
2074         * valtrack.c (debug_lowpart_subreg): Likewise.
2075         * var-tracking.c (var_lowpart): Likewise.
2076         (loc_cmp): Handle 'p'.
2078 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2079             Alan Hayward  <alan.hayward@arm.com>
2080             David Sherwood  <david.sherwood@arm.com>
2082         * ira.c (get_subreg_tracking_sizes): New function.
2083         (init_live_subregs): Take an integer size rather than a register.
2084         (build_insn_chain): Use get_subreg_tracking_sizes.  Update calls
2085         to init_live_subregs.
2087 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2088             Alan Hayward  <alan.hayward@arm.com>
2089             David Sherwood  <david.sherwood@arm.com>
2091         * expr.c (store_constructor_field): Change bitsize from a
2092         unsigned HOST_WIDE_INT to a poly_uint64 and bitpos from a
2093         HOST_WIDE_INT to a poly_int64.
2094         (store_constructor): Change size from a HOST_WIDE_INT to
2095         a poly_int64.
2096         (store_field): Likewise bitsize and bitpos.
2098 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2099             Alan Hayward  <alan.hayward@arm.com>
2100             David Sherwood  <david.sherwood@arm.com>
2102         * expmed.h (store_bit_field): Change bitregion_start and
2103         bitregion_end from unsigned HOST_WIDE_INT to poly_uint64.
2104         * expmed.c (adjust_bit_field_mem_for_reg, strict_volatile_bitfield_p)
2105         (store_bit_field_1, store_integral_bit_field, store_bit_field)
2106         (store_fixed_bit_field, store_split_bit_field): Likewise.
2107         * expr.c (store_constructor_field, store_field): Likewise.
2108         (optimize_bitfield_assignment_op): Likewise.  Make the same change
2109         to bitsize and bitpos.
2110         * machmode.h (bit_field_mode_iterator): Change m_bitregion_start
2111         and m_bitregion_end from HOST_WIDE_INT to poly_int64.  Make the
2112         same change in the constructor arguments.
2113         (get_best_mode): Change bitregion_start and bitregion_end from
2114         unsigned HOST_WIDE_INT to poly_uint64.
2115         * stor-layout.c (bit_field_mode_iterator::bit_field_mode_iterator):
2116         Change bitregion_start and bitregion_end from HOST_WIDE_INT to
2117         poly_int64.
2118         (bit_field_mode_iterator::next_mode): Update for new types
2119         of m_bitregion_start and m_bitregion_end.
2120         (get_best_mode): Change bitregion_start and bitregion_end from
2121         unsigned HOST_WIDE_INT to poly_uint64.
2123 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2124             Alan Hayward  <alan.hayward@arm.com>
2125             David Sherwood  <david.sherwood@arm.com>
2127         * rtl.h (simplify_gen_subreg): Add a temporary overload that
2128         accepts poly_uint64 offsets.
2129         * expmed.h (extract_bit_field): Take bitsize and bitnum as
2130         poly_uint64s rather than unsigned HOST_WIDE_INTs.
2131         * expmed.c (lowpart_bit_field_p): Likewise.
2132         (extract_bit_field_as_subreg): New function, split out from...
2133         (extract_bit_field_1): ...here.  Take bitsize and bitnum as
2134         poly_uint64s rather than unsigned HOST_WIDE_INTs.  For vector
2135         extractions, check that BITSIZE matches the size of the extracted
2136         value and that BITNUM is an exact multiple of that size.
2137         If all else fails, try forcing the value into memory if
2138         BITNUM is variable, and adjusting the address so that the
2139         offset is constant.  Split the part that can only handle constant
2140         bitsize and bitnum out into...
2141         (extract_integral_bit_field): ...this new function.
2142         (extract_bit_field): Take bitsize and bitnum as poly_uint64s
2143         rather than unsigned HOST_WIDE_INTs.
2145 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2146             Alan Hayward  <alan.hayward@arm.com>
2147             David Sherwood  <david.sherwood@arm.com>
2149         * expmed.h (store_bit_field): Take bitsize and bitnum as
2150         poly_uint64s rather than unsigned HOST_WIDE_INTs.
2151         * expmed.c (simple_mem_bitfield_p): Likewise.  Add a parameter
2152         that returns the byte size.
2153         (store_bit_field_1): Take bitsize and bitnum as
2154         poly_uint64s rather than unsigned HOST_WIDE_INTs.  Update call
2155         to simple_mem_bitfield_p.  Split the part that can only handle
2156         constant bitsize and bitnum out into...
2157         (store_integral_bit_field): ...this new function.
2158         (store_bit_field): Take bitsize and bitnum as poly_uint64s rather
2159         than unsigned HOST_WIDE_INTs.
2160         (extract_bit_field_1): Update call to simple_mem_bitfield_p.
2162 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2163             Alan Hayward  <alan.hayward@arm.com>
2164             David Sherwood  <david.sherwood@arm.com>
2166         * lra-int.h (lra_reg): Change offset from int to poly_int64.
2167         (lra_insn_recog_data): Change sp_offset from HOST_WIDE_INT
2168         to poly_int64.
2169         (lra_eliminate_regs_1, eliminate_regs_in_insn): Change
2170         update_sp_offset from a HOST_WIDE_INT to a poly_int64.
2171         (lra_update_reg_val_offset, lra_reg_val_equal_p): Take the
2172         offset as a poly_int64 rather than an int.
2173         * lra-assigns.c (find_hard_regno_for_1): Handle poly_int64 offsets.
2174         (setup_live_pseudos_and_spill_after_risky_transforms): Likewise.
2175         * lra-constraints.c (equiv_address_substitution): Track offsets
2176         as poly_int64s.
2177         (emit_inc): Check poly_int_rtx_p instead of CONST_INT_P.
2178         (curr_insn_transform): Handle the new form of sp_offset.
2179         * lra-eliminations.c (lra_elim_table): Change previous_offset
2180         and offset from HOST_WIDE_INT to poly_int64.
2181         (print_elim_table, update_reg_eliminate): Update accordingly.
2182         (self_elim_offsets): Change from HOST_WIDE_INT to poly_int64_pod.
2183         (get_elimination): Update accordingly.
2184         (form_sum): Check poly_int_rtx_p instead of CONST_INT_P.
2185         (lra_eliminate_regs_1, eliminate_regs_in_insn): Change
2186         update_sp_offset from a HOST_WIDE_INT to a poly_int64.  Handle
2187         poly_int64 offsets generally.
2188         (curr_sp_change): Change from HOST_WIDE_INT to poly_int64.
2189         (mark_not_eliminable, init_elimination): Update accordingly.
2190         (remove_reg_equal_offset_note): Return a bool and pass the new
2191         offset back by pointer as a poly_int64.
2192         * lra-remat.c (change_sp_offset): Take sp_offset as a poly_int64
2193         rather than a HOST_WIDE_INT.
2194         (do_remat): Track offsets poly_int64s.
2195         * lra.c (lra_update_insn_recog_data, setup_sp_offset): Likewise.
2197 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2198             Alan Hayward  <alan.hayward@arm.com>
2199             David Sherwood  <david.sherwood@arm.com>
2201         * rtl.h (mem_attrs): Add a default constructor.  Change size and
2202         offset from HOST_WIDE_INT to poly_int64.
2203         * emit-rtl.h (set_mem_offset, set_mem_size, adjust_address_1)
2204         (adjust_automodify_address_1, set_mem_attributes_minus_bitpos)
2205         (widen_memory_access): Take the sizes and offsets as poly_int64s
2206         rather than HOST_WIDE_INTs.
2207         * alias.c (ao_ref_from_mem): Handle the new form of MEM_OFFSET.
2208         (offset_overlap_p): Take poly_int64s rather than HOST_WIDE_INTs
2209         and ints.
2210         (adjust_offset_for_component_ref): Change the offset from a
2211         HOST_WIDE_INT to a poly_int64.
2212         (nonoverlapping_memrefs_p): Track polynomial offsets and sizes.
2213         * cfgcleanup.c (merge_memattrs): Update after mem_attrs changes.
2214         * dce.c (find_call_stack_args): Likewise.
2215         * dse.c (record_store): Likewise.
2216         * dwarf2out.c (tls_mem_loc_descriptor, dw_sra_loc_expr): Likewise.
2217         * print-rtl.c (rtx_writer::print_rtx): Likewise.
2218         * read-rtl-function.c (test_loading_mem): Likewise.
2219         * rtlanal.c (may_trap_p_1): Likewise.
2220         * simplify-rtx.c (delegitimize_mem_from_attrs): Likewise.
2221         * var-tracking.c (int_mem_offset, track_expr_p): Likewise.
2222         * emit-rtl.c (mem_attrs_eq_p, get_mem_align_offset): Likewise.
2223         (mem_attrs::mem_attrs): New function.
2224         (set_mem_attributes_minus_bitpos): Change bitpos from a
2225         HOST_WIDE_INT to poly_int64.
2226         (set_mem_alias_set, set_mem_addr_space, set_mem_align, set_mem_expr)
2227         (clear_mem_offset, clear_mem_size, change_address)
2228         (get_spill_slot_decl, set_mem_attrs_for_spill): Directly
2229         initialize mem_attrs.
2230         (set_mem_offset, set_mem_size, adjust_address_1)
2231         (adjust_automodify_address_1, offset_address, widen_memory_access):
2232         Likewise.  Take poly_int64s rather than HOST_WIDE_INT.
2234 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2235             Alan Hayward  <alan.hayward@arm.com>
2236             David Sherwood  <david.sherwood@arm.com>
2238         * rtlanal.c (rtx_addr_can_trap_p_1): Take the offset and size
2239         as poly_int64s rather than HOST_WIDE_INTs.  Use a size of -1
2240         rather than 0 to represent an unknown size.  Assert that the size
2241         is known when the mode isn't BLKmode.
2242         (may_trap_p_1): Use -1 for unknown sizes.
2243         (rtx_addr_can_trap_p): Likewise.  Pass BLKmode rather than VOIDmode.
2245 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2246             Alan Hayward  <alan.hayward@arm.com>
2247             David Sherwood  <david.sherwood@arm.com>
2249         * dse.c (store_info): Change offset and width from HOST_WIDE_INT
2250         to poly_int64.  Update commentary for positions_needed.large.
2251         (read_info_type): Change offset and width from HOST_WIDE_INT
2252         to poly_int64.
2253         (set_usage_bits): Likewise.
2254         (canon_address): Return the offset as a poly_int64 rather than
2255         a HOST_WIDE_INT.  Use strip_offset_and_add.
2256         (set_all_positions_unneeded, any_positions_needed_p): Use
2257         positions_needed.large to track stores with non-constant widths.
2258         (all_positions_needed_p): Likewise.  Take the offset and width
2259         as poly_int64s rather than ints.  Assert that rhs is nonnull.
2260         (record_store): Cope with non-constant offsets and widths.
2261         Nullify the rhs of an earlier store if we can't tell which bytes
2262         of it are needed.
2263         (find_shift_sequence): Take the access_size and shift as poly_int64s
2264         rather than ints.
2265         (get_stored_val): Take the read_offset and read_width as poly_int64s
2266         rather than HOST_WIDE_INTs.
2267         (check_mem_read_rtx, scan_stores, scan_reads, dse_step5): Handle
2268         non-constant offsets and widths.
2270 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2271             Alan Hayward  <alan.hayward@arm.com>
2272             David Sherwood  <david.sherwood@arm.com>
2274         * inchash.h (inchash::hash::add_poly_int): New function.
2275         * tree-ssa-alias.h (ao_ref::offset, ao_ref::size, ao_ref::max_size):
2276         Use poly_int64 rather than HOST_WIDE_INT.
2277         (ao_ref::max_size_known_p): New function.
2278         * tree-ssa-sccvn.h (vn_reference_op_struct::off): Use poly_int64_pod
2279         rather than HOST_WIDE_INT.
2280         * tree-ssa-alias.c (ao_ref_base): Apply get_ref_base_and_extent
2281         to temporaries until its interface is adjusted to match.
2282         (ao_ref_init_from_ptr_and_size): Handle polynomial offsets and sizes.
2283         (aliasing_component_refs_p, decl_refs_may_alias_p)
2284         (indirect_ref_may_alias_decl_p, indirect_refs_may_alias_p): Take
2285         the offsets and max_sizes as poly_int64s instead of HOST_WIDE_INTs.
2286         (refs_may_alias_p_1, stmt_kills_ref_p): Adjust for changes to
2287         ao_ref fields.
2288         * alias.c (ao_ref_from_mem): Likewise.
2289         * tree-ssa-dce.c (mark_aliased_reaching_defs_necessary_1): Likewise.
2290         * tree-ssa-dse.c (valid_ao_ref_for_dse, normalize_ref)
2291         (clear_bytes_written_by, setup_live_bytes_from_ref, compute_trims)
2292         (maybe_trim_complex_store, maybe_trim_constructor_store)
2293         (live_bytes_read, dse_classify_store): Likewise.
2294         * tree-ssa-sccvn.c (vn_reference_compute_hash, vn_reference_eq):
2295         (copy_reference_ops_from_ref, ao_ref_init_from_vn_reference)
2296         (fully_constant_vn_reference_p, valueize_refs_1): Likewise.
2297         (vn_reference_lookup_3): Likewise.
2298         * tree-ssa-uninit.c (warn_uninitialized_vars): Likewise.
2300 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2301             Alan Hayward  <alan.hayward@arm.com>
2302             David Sherwood  <david.sherwood@arm.com>
2304         * tree-ssa-alias.c (indirect_ref_may_alias_decl_p)
2305         (indirect_refs_may_alias_p): Use ranges_may_overlap_p
2306         instead of ranges_overlap_p.
2308 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2309             Alan Hayward  <alan.hayward@arm.com>
2310             David Sherwood  <david.sherwood@arm.com>
2312         * tree-ssa-alias.c (same_addr_size_stores_p): Take the offsets and
2313         sizes as poly_int64s rather than HOST_WIDE_INTs.
2315 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2316             Alan Hayward  <alan.hayward@arm.com>
2317             David Sherwood  <david.sherwood@arm.com>
2319         * gimple-fold.h (fold_ctor_reference): Take the offset and size
2320         as poly_uint64 rather than unsigned HOST_WIDE_INT.
2321         * gimple-fold.c (fold_ctor_reference): 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         * target.def (dwarf_poly_indeterminate_value): New hook.
2328         * targhooks.h (default_dwarf_poly_indeterminate_value): Declare.
2329         * targhooks.c (default_dwarf_poly_indeterminate_value): New function.
2330         * doc/tm.texi.in (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Document.
2331         * doc/tm.texi: Regenerate.
2332         * dwarf2out.h (build_cfa_loc, build_cfa_aligned_loc): Take the
2333         offset as a poly_int64.
2334         * dwarf2out.c (new_reg_loc_descr): Move later in file.  Take the
2335         offset as a poly_int64.
2336         (loc_descr_plus_const, loc_list_plus_const, build_cfa_aligned_loc):
2337         Take the offset as a poly_int64.
2338         (build_cfa_loc): Likewise.  Use loc_descr_plus_const.
2339         (frame_pointer_fb_offset): Change to a poly_int64.
2340         (int_loc_descriptor): Take the offset as a poly_int64.  Use
2341         targetm.dwarf_poly_indeterminate_value for polynomial offsets.
2342         (based_loc_descr): Take the offset as a poly_int64.
2343         Use strip_offset_and_add to handle (plus X (const)).
2344         Use new_reg_loc_descr instead of an open-coded version of the
2345         previous implementation.
2346         (mem_loc_descriptor): Handle CONST_POLY_INT.
2347         (compute_frame_pointer_to_fb_displacement): Take the offset as a
2348         poly_int64.  Use strip_offset_and_add to handle (plus X (const)).
2350 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2351             Alan Hayward  <alan.hayward@arm.com>
2352             David Sherwood  <david.sherwood@arm.com>
2354         * rtl.h (reg_attrs::offset): Change from HOST_WIDE_INT to poly_int64.
2355         (gen_rtx_REG_offset): Take the offset as a poly_int64.
2356         * inchash.h (inchash::hash::add_poly_hwi): New function.
2357         * gengtype.c (main): Register poly_int64.
2358         * emit-rtl.c (reg_attr_hasher::hash): Use inchash.  Treat the
2359         offset as a poly_int.
2360         (reg_attr_hasher::equal): Use must_eq to compare offsets.
2361         (get_reg_attrs, update_reg_offset, gen_rtx_REG_offset): Take the
2362         offset as a poly_int64.
2363         (set_reg_attrs_from_value): Treat the offset as a poly_int64.
2364         * print-rtl.c (print_poly_int): New function.
2365         (rtx_writer::print_rtx_operand_code_r): Treat REG_OFFSET as
2366         a poly_int.
2367         * var-tracking.c (track_offset_p, get_tracked_reg_offset): New
2368         functions.
2369         (var_reg_set, var_reg_delete_and_set, var_reg_delete): Use them.
2370         (same_variable_part_p, track_loc_p): Take the offset as a poly_int64.
2371         (vt_get_decl_and_offset): Return the offset as a poly_int64.
2372         Enforce track_offset_p for parts of a PARALLEL.
2373         (vt_add_function_parameter): Use const_offset for the final
2374         offset to track.  Use get_tracked_reg_offset for the parts
2375         of a PARALLEL.
2377 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2378             Alan Hayward  <alan.hayward@arm.com>
2379             David Sherwood  <david.sherwood@arm.com>
2381         * target.def (truly_noop_truncation): Take poly_uint64s instead of
2382         unsigned ints.  Change default to hook_bool_puint64_puint64_true.
2383         * doc/tm.texi: Regenerate.
2384         * hooks.h (hook_bool_uint_uint_true): Delete.
2385         (hook_bool_puint64_puint64_true): Declare.
2386         * hooks.c (hook_bool_uint_uint_true): Delete.
2387         (hook_bool_puint64_puint64_true): New function.
2388         * config/mips/mips.c (mips_truly_noop_truncation): Take poly_uint64s
2389         instead of unsigned ints.
2390         * config/spu/spu.c (spu_truly_noop_truncation): Likewise.
2391         * config/tilegx/tilegx.c (tilegx_truly_noop_truncation): Likewise.
2393 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2394             Alan Hayward  <alan.hayward@arm.com>
2395             David Sherwood  <david.sherwood@arm.com>
2397         * optabs.h (expand_operand): Add an int_value field.
2398         (create_expand_operand): Add an int_value parameter and use it
2399         to initialize the new expand_operand field.
2400         (create_integer_operand): Replace with a declaration of a function
2401         that accepts poly_int64s.  Move the implementation to...
2402         * optabs.c (create_integer_operand): ...here.
2403         (maybe_legitimize_operand): For EXPAND_INTEGER, check whether
2404         the mode preserves the value of int_value, instead of calling
2405         const_int_operand on the rtx.  Use gen_int_mode to generate
2406         the new rtx.
2408 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2409             Alan Hayward  <alan.hayward@arm.com>
2410             David Sherwood  <david.sherwood@arm.com>
2412         * dumpfile.h (dump_dec): Declare.
2413         * dumpfile.c (dump_dec): New function.
2414         * pretty-print.h (pp_wide_integer): Turn into a function and
2415         declare a poly_int version.
2416         * pretty-print.c (pp_wide_integer): New function for poly_ints.
2418 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2419             Alan Hayward  <alan.hayward@arm.com>
2420             David Sherwood  <david.sherwood@arm.com>
2422         * doc/generic.texi (POLY_INT_CST): Document.
2423         * tree.def (POLY_INT_CST): New tree code.
2424         * treestruct.def (TS_POLY_INT_CST): New tree layout.
2425         * tree-core.h (tree_poly_int_cst): New struct.
2426         (tree_node): Add a poly_int_cst field.
2427         * tree.h (POLY_INT_CST_P, POLY_INT_CST_COEFF): New macros.
2428         (wide_int_to_tree, force_fit_type): Take a poly_wide_int_ref
2429         instead of a wide_int_ref.
2430         (build_int_cst, build_int_cst_type): Take a poly_int64 instead
2431         of a HOST_WIDE_INT.
2432         (build_int_cstu, build_array_type_nelts): Take a poly_uint64
2433         instead of an unsigned HOST_WIDE_INT.
2434         (build_poly_int_cst, tree_fits_poly_int64_p, tree_fits_poly_uint64_p)
2435         (ptrdiff_tree_p): Declare.
2436         (tree_to_poly_int64, tree_to_poly_uint64): Likewise.  Provide
2437         extern inline implementations if the target doesn't use POLY_INT_CST.
2438         (poly_int_tree_p): New function.
2439         (wi::unextended_tree): New class.
2440         (wi::int_traits <unextended_tree>): New override.
2441         (wi::extended_tree): Add a default constructor.
2442         (wi::extended_tree::get_tree): New function.
2443         (wi::widest_extended_tree, wi::offset_extended_tree): New typedefs.
2444         (wi::tree_to_widest_ref, wi::tree_to_offset_ref): Use them.
2445         (wi::tree_to_poly_widest_ref, wi::tree_to_poly_offset_ref)
2446         (wi::tree_to_poly_wide_ref): New typedefs.
2447         (wi::ints_for): Provide overloads for extended_tree and
2448         unextended_tree.
2449         (poly_int_cst_value, wi::to_poly_widest, wi::to_poly_offset)
2450         (wi::to_wide): New functions.
2451         (wi::fits_to_boolean_p, wi::fits_to_tree_p): Handle poly_ints.
2452         * tree.c (poly_int_cst_hasher): New struct.
2453         (poly_int_cst_hash_table): New variable.
2454         (tree_node_structure_for_code, tree_code_size, simple_cst_equal)
2455         (valid_constant_size_p, add_expr, drop_tree_overflow): Handle
2456         POLY_INT_CST.
2457         (initialize_tree_contains_struct): Handle TS_POLY_INT_CST.
2458         (init_ttree): Initialize poly_int_cst_hash_table.
2459         (build_int_cst, build_int_cst_type, build_invariant_address): Take
2460         a poly_int64 instead of a HOST_WIDE_INT.
2461         (build_int_cstu, build_array_type_nelts): Take a poly_uint64
2462         instead of an unsigned HOST_WIDE_INT.
2463         (wide_int_to_tree): Rename to...
2464         (wide_int_to_tree_1): ...this.
2465         (build_new_poly_int_cst, build_poly_int_cst): New functions.
2466         (force_fit_type): Take a poly_wide_int_ref instead of a wide_int_ref.
2467         (wide_int_to_tree): New function that takes a poly_wide_int_ref.
2468         (ptrdiff_tree_p, tree_to_poly_int64, tree_to_poly_uint64)
2469         (tree_fits_poly_int64_p, tree_fits_poly_uint64_p): New functions.
2470         * lto-streamer-out.c (DFS::DFS_write_tree_body, hash_tree): Handle
2471         TS_POLY_INT_CST.
2472         * tree-streamer-in.c (lto_input_ts_poly_tree_pointers): Likewise.
2473         (streamer_read_tree_body): Likewise.
2474         * tree-streamer-out.c (write_ts_poly_tree_pointers): Likewise.
2475         (streamer_write_tree_body): Likewise.
2476         * tree-streamer.c (streamer_check_handled_ts_structures): Likewise.
2477         * asan.c (asan_protect_global): Require the size to be an INTEGER_CST.
2478         * cfgexpand.c (expand_debug_expr): Handle POLY_INT_CST.
2479         * expr.c (expand_expr_real_1, const_vector_from_tree): Likewise.
2480         * gimple-expr.h (is_gimple_constant): Likewise.
2481         * gimplify.c (maybe_with_size_expr): Likewise.
2482         * print-tree.c (print_node): Likewise.
2483         * tree-data-ref.c (data_ref_compare_tree): Likewise.
2484         * tree-pretty-print.c (dump_generic_node): Likewise.
2485         * tree-ssa-address.c (addr_for_mem_ref): Likewise.
2486         * tree-vect-data-refs.c (dr_group_sort_cmp): Likewise.
2487         * tree-vrp.c (compare_values_warnv): Likewise.
2488         * tree-ssa-loop-ivopts.c (determine_base_object, constant_multiple_of)
2489         (get_loop_invariant_expr, add_candidate_1, get_computation_aff_1)
2490         (force_expr_to_var_cost): Likewise.
2491         * tree-ssa-loop.c (for_each_index): Likewise.
2492         * fold-const.h (build_invariant_address, size_int_kind): Take a
2493         poly_int64 instead of a HOST_WIDE_INT.
2494         * fold-const.c (fold_negate_expr_1, const_binop, const_unop)
2495         (fold_convert_const, multiple_of_p, fold_negate_const): Handle
2496         POLY_INT_CST.
2497         (size_binop_loc): Likewise.  Allow int_const_binop_1 to fail.
2498         (int_const_binop_2): New function, split out from...
2499         (int_const_binop_1): ...here.  Handle POLY_INT_CST.
2500         (size_int_kind): Take a poly_int64 instead of a HOST_WIDE_INT.
2501         * expmed.c (make_tree): Handle CONST_POLY_INT_P.
2502         * gimple-ssa-strength-reduction.c (slsr_process_add)
2503         (slsr_process_mul): Check for INTEGER_CSTs before using them
2504         as candidates.
2505         * stor-layout.c (bits_from_bytes): New function.
2506         (bit_from_pos): Use it.
2507         (layout_type): Likewise.  For vectors, multiply the TYPE_SIZE_UNIT
2508         by BITS_PER_UNIT to get the TYPE_SIZE.
2509         * tree-cfg.c (verify_expr, verify_types_in_gimple_reference): Allow
2510         MEM_REF and TARGET_MEM_REF offsets to be a POLY_INT_CST.
2512 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2513             Alan Hayward  <alan.hayward@arm.com>
2514             David Sherwood  <david.sherwood@arm.com>
2516         * doc/rtl.texi (const_poly_int): Document.  Also document the
2517         rtl sharing behavior.
2518         * gengenrtl.c (excluded_rtx): Return true for CONST_POLY_INT.
2519         * rtl.h (const_poly_int_def): New struct.
2520         (rtx_def::u): Add a cpi field.
2521         (CASE_CONST_UNIQUE, CASE_CONST_ANY): Add CONST_POLY_INT.
2522         (CONST_POLY_INT_P, CONST_POLY_INT_COEFFS): New macros.
2523         (wi::rtx_to_poly_wide_ref): New typedef
2524         (const_poly_int_value, wi::to_poly_wide, rtx_to_poly_int64)
2525         (poly_int_rtx_p): New functions.
2526         (trunc_int_for_mode): Declare a poly_int64 version.
2527         (plus_constant): Take a poly_int64 instead of a HOST_WIDE_INT.
2528         (immed_wide_int_const): Take a poly_wide_int_ref rather than
2529         a wide_int_ref.
2530         (strip_offset): Declare.
2531         (strip_offset_and_add): New function.
2532         * rtl.def (CONST_POLY_INT): New rtx code.
2533         * rtl.c (rtx_size): Handle CONST_POLY_INT.
2534         (shared_const_p): Use poly_int_rtx_p.
2535         * emit-rtl.h (gen_int_mode): Take a poly_int64 instead of a
2536         HOST_WIDE_INT.
2537         (gen_int_shift_amount): Likewise.
2538         * emit-rtl.c (const_poly_int_hasher): New class.
2539         (const_poly_int_htab): New variable.
2540         (init_emit_once): Initialize it when NUM_POLY_INT_COEFFS > 1.
2541         (const_poly_int_hasher::hash): New function.
2542         (const_poly_int_hasher::equal): Likewise.
2543         (gen_int_mode): Take a poly_int64 instead of a HOST_WIDE_INT.
2544         (immed_wide_int_const): Rename to...
2545         (immed_wide_int_const_1): ...this and make static.
2546         (immed_wide_int_const): New function, taking a poly_wide_int_ref
2547         instead of a wide_int_ref.
2548         (gen_int_shift_amount): Take a poly_int64 instead of a HOST_WIDE_INT.
2549         (gen_lowpart_common): Handle CONST_POLY_INT.
2550         * cse.c (hash_rtx_cb, equiv_constant): Likewise.
2551         * cselib.c (cselib_hash_rtx): Likewise.
2552         * dwarf2out.c (const_ok_for_output_1): Likewise.
2553         * expr.c (convert_modes): Likewise.
2554         * print-rtl.c (rtx_writer::print_rtx, print_value): Likewise.
2555         * rtlhash.c (add_rtx): Likewise.
2556         * explow.c (trunc_int_for_mode): Add a poly_int64 version.
2557         (plus_constant): Take a poly_int64 instead of a HOST_WIDE_INT.
2558         Handle existing CONST_POLY_INT rtxes.
2559         * expmed.h (expand_shift): Take a poly_int64 instead of a
2560         HOST_WIDE_INT.
2561         * expmed.c (expand_shift): Likewise.
2562         * rtlanal.c (strip_offset): New function.
2563         (commutative_operand_precedence): Give CONST_POLY_INT the same
2564         precedence as CONST_DOUBLE and put CONST_WIDE_INT between that
2565         and CONST_INT.
2566         * rtl-tests.c (const_poly_int_tests): New struct.
2567         (rtl_tests_c_tests): Use it.
2568         * simplify-rtx.c (simplify_const_unary_operation): Handle
2569         CONST_POLY_INT.
2570         (simplify_const_binary_operation): Likewise.
2571         (simplify_binary_operation_1): Fold additions of symbolic constants
2572         and CONST_POLY_INTs.
2573         (simplify_subreg): Handle extensions and truncations of
2574         CONST_POLY_INTs.
2575         (simplify_const_poly_int_tests): New struct.
2576         (simplify_rtx_c_tests): Use it.
2577         * wide-int.h (storage_ref): Add default constructor.
2578         (wide_int_ref_storage): Likewise.
2579         (trailing_wide_ints): Use GTY((user)).
2580         (trailing_wide_ints::operator[]): Add a const version.
2581         (trailing_wide_ints::get_precision): New function.
2582         (trailing_wide_ints::extra_size): Likewise.
2584 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2585             Alan Hayward  <alan.hayward@arm.com>
2586             David Sherwood  <david.sherwood@arm.com>
2588         * emit-rtl.h (gen_int_shift_amount): Declare.
2589         * emit-rtl.c (gen_int_shift_amount): New function.
2590         * asan.c (asan_emit_stack_protection): Use gen_int_shift_amount
2591         instead of GEN_INT.
2592         * calls.c (shift_return_value): Likewise.
2593         * cse.c (fold_rtx): Likewise.
2594         * dse.c (find_shift_sequence): Likewise.
2595         * expmed.c (init_expmed_one_mode, store_bit_field_1, expand_shift_1)
2596         (expand_shift, expand_smod_pow2): Likewise.
2597         * lower-subreg.c (shift_cost): Likewise.
2598         * optabs.c (expand_superword_shift, expand_doubleword_mult)
2599         (expand_unop, expand_binop, shift_amt_for_vec_perm_mask)
2600         (expand_vec_perm_var): Likewise.
2601         * simplify-rtx.c (simplify_unary_operation_1): Likewise.
2602         (simplify_binary_operation_1): Likewise.
2603         * combine.c (try_combine, find_split_point, force_int_to_mode)
2604         (simplify_shift_const_1, simplify_shift_const): Likewise.
2605         (change_zero_ext): Likewise.  Use simplify_gen_binary.
2607 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2609         * poly-int.h (multiple_p): Fix handling of two non-poly_ints.
2611 2017-12-20  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
2613         * doc/invoke.texi (ARM Options): Document accepted extension options
2614         for -march=armv8.3-a.
2616 2017-12-20  Richard Earnshaw  <rearnsha@arm.com>
2618         PR target/83105
2619         * config.gcc (arm*-*-linux*): When configured with --with-float=hard
2620         or --with-float=softfp, set the default CPU to arm10e.
2622 2017-12-20  Eric Botcazou  <ebotcazou@adacore.com>
2624         * config/visium/constraints.md (J, K, L): Use IN_RANGE macro.
2625         * config/visium/predicates.md (const_shift_operand): Likewise.
2626         * config/visium/visium.c (visium_legitimize_address): Fix oversight.
2627         (visium_legitimize_reload_address): Likewise.
2629 2017-12-20  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
2631         PR target/82975
2632         * config/arm/arm.h (TEST_REGNO): Adjust comment as expected in r255830.
2634 2017-12-20  Jakub Jelinek  <jakub@redhat.com>
2636         PR c++/83490
2637         * calls.c (compute_argument_addresses): Ignore TYPE_EMPTY_P arguments.
2639 2017-12-20  Julia Koval  <julia.koval@intel.com>
2641         * common/config/i386/i386-common.c (OPTION_MASK_ISA_VPCLMULQDQ_SET,
2642         OPTION_MASK_ISA_VPCLMULQDQ_UNSET): New.
2643         (ix86_handle_option): Handle -mvpclmulqdq, move cx6 to flags2.
2644         * config.gcc: Include vpclmulqdqintrin.h.
2645         * config/i386/cpuid.h: Handle bit_VPCLMULQDQ.
2646         * config/i386/driver-i386.c (host_detect_local_cpu): Handle
2647         -mvpclmulqdq.
2648         * config/i386/i386-builtin.def (__builtin_ia32_vpclmulqdq_v2di,
2649         __builtin_ia32_vpclmulqdq_v4di, __builtin_ia32_vpclmulqdq_v8di): New.
2650         * config/i386/i386-c.c (__VPCLMULQDQ__): New.
2651         * config/i386/i386.c (isa2_opts): Add -mcx16.
2652         (isa_opts): Add -mpclmulqdq, remove -mcx16.
2653         (ix86_option_override_internal): Move mcx16 to flags2.
2654         (ix86_valid_target_attribute_inner_p): Add vpclmulqdq.
2655         (ix86_expand_builtin): Handle OPTION_MASK_ISA_VPCLMULQDQ.
2656         * config/i386/i386.h (TARGET_VPCLMULQDQ, TARGET_VPCLMULQDQ_P): New.
2657         * config/i386/i386.opt: Add mvpclmulqdq, move mcx16 to flags2.
2658         * config/i386/immintrin.h: Include vpclmulqdqintrin.h.
2659         * config/i386/sse.md (vpclmulqdq_<mode>): New pattern.
2660         * config/i386/vpclmulqdqintrin.h (_mm512_clmulepi64_epi128,
2661         _mm_clmulepi64_epi128, _mm256_clmulepi64_epi128): New intrinsics.
2662         * doc/invoke.texi: Add -mvpclmulqdq.
2664 2017-12-20  Tom de Vries  <tom@codesourcery.com>
2666         PR middle-end/83423
2667         * config/i386/i386.c (ix86_static_chain): Move
2668         DECL_STATIC_CHAIN test ...
2669         * calls.c (rtx_for_static_chain): ... here.  New function.
2670         * calls.h (rtx_for_static_chain): Declare.
2671         * builtins.c (expand_builtin_setjmp_receiver): Use rtx_for_static_chain
2672         instead of targetm.calls.static_chain.
2673         * df-scan.c (df_get_entry_block_def_set): Same.
2675 2017-12-19  Tom de Vries  <tom@codesourcery.com>
2677         PR tree-optimization/83493
2678         * graphite-isl-ast-to-gimple.c (translate_isl_ast_node_for): Unshare ub
2679         and lb.
2681 2017-12-19  Jakub Jelinek  <jakub@redhat.com>
2683         * gimple-ssa-sprintf.c (format_directive): Use inform_n instead of
2684         inform with hardcoded english plural handling.
2686 2017-12-18  Jeff Law  <law@redhat.com>
2688         PR tree-optimization/83477
2689         * tree-ssa-threadedge.c (record_temporary_equivalences_from_phis): For
2690         a non-virtual PHI, always push a new range.
2692 2017-12-19  Martin Sebor  <msebor@redhat.com>
2694         PR middle-end/77608
2695         * builtins.c (compute_objsize): Handle non-constant offsets.
2697 2017-12-19  Jakub Jelinek  <jakub@redhat.com>
2699         PR tree-optimization/83444
2700         * tree-ssa-strlen.c (strlen_check_and_optimize_stmt): For the
2701         character load case, if get_stridx on MEM_REF's operand doesn't
2702         look usable, retry with get_addr_stridx.
2704 2017-12-19  Alexandre Oliva  <aoliva@redhat.com>
2706         PR debug/83422
2707         * var-tracking.c (vt_debug_insns_local): Do not drop markers.
2708         (variable_tracking_main_1): Keep markers even when VTA fails.
2710         PR bootstrap/83396
2711         * cfgexpand.c (expand_gimple_basic_block): Expand label first,
2712         even if there are markers before it.
2713         * cfgrtl.c (rtl_verify_bb_layout): Reject DEBUG_INSNs outside BBs.
2715 2017-12-19  Jakub Jelinek  <jakub@redhat.com>
2717         * read-rtl.c (parse_reg_note_name): Replace Yoda conditions with
2718         typical order conditions.
2719         * sel-sched.c (extract_new_fences_from): Likewise.
2720         * config/visium/constraints.md (J, K, L): Likewise.
2721         * config/visium/predicates.md (const_shift_operand): Likewise.
2722         * config/visium/visium.c (visium_legitimize_address,
2723         visium_legitimize_reload_address): Likewise.
2724         * config/m68k/m68k.c (output_reg_adjust, emit_reg_adjust): Likewise.
2725         * config/arm/arm.c (arm_block_move_unaligned_straight): Likewise.
2726         * config/avr/constraints.md (Y01, Ym1, Y02, Ym2): Likewise.
2727         * config/avr/avr-log.c (avr_vdump, avr_log_set_avr_log,
2728         SET_DUMP_DETAIL): Likewise.
2729         * config/avr/predicates.md (const_8_16_24_operand): Likewise.
2730         * config/avr/avr.c (STR_PREFIX_P, avr_popcount_each_byte,
2731         avr_is_casesi_sequence, avr_casei_sequence_check_operands,
2732         avr_set_core_architecture, avr_set_current_function,
2733         avr_legitimize_reload_address, avr_asm_len, avr_print_operand,
2734         output_movqi, output_movsisf, avr_out_plus, avr_out_bitop,
2735         avr_out_fract, avr_adjust_insn_length, avr_encode_section_info,
2736         avr_2word_insn_p, output_reload_in_const, avr_has_nibble_0xf,
2737         avr_map_decompose, avr_fold_builtin): Likewise.
2738         * config/avr/driver-avr.c (avr_devicespecs_file): Likewise.
2739         * config/avr/gen-avr-mmcu-specs.c (str_prefix_p, print_mcu): Likewise.
2740         * config/i386/i386.c (ix86_parse_stringop_strategy_string): Likewise.
2741         * config/m32c/m32c-pragma.c (m32c_pragma_memregs): Likewise.
2742         * config/m32c/m32c.c (m32c_conditional_register_usage,
2743         m32c_address_cost): Likewise.
2744         * config/m32c/predicates.md (shiftcount_operand,
2745         longshiftcount_operand): Likewise.
2746         * config/iq2000/iq2000.c (iq2000_expand_prologue): Likewise.
2747         * config/nios2/nios2.c (nios2_handle_custom_fpu_insn_option,
2748         can_use_cdx_ldstw): Likewise.
2749         * config/nios2/nios2.h (CDX_REG_P): Likewise.
2750         * config/cr16/cr16.h (RETURN_ADDR_RTX, REGNO_MODE_OK_FOR_BASE_P):
2751         Likewise.
2752         * config/cr16/cr16.md (*mov<mode>_double): Likewise.
2753         * config/cr16/cr16.c (cr16_create_dwarf_for_multi_push): Likewise.
2754         * config/h8300/h8300.c (h8300_rtx_costs, get_shift_alg): Likewise.
2755         * config/vax/constraints.md (U06, U08, U16, CN6, S08, S16): Likewise.
2756         * config/vax/vax.c (adjacent_operands_p): Likewise.
2757         * config/ft32/constraints.md (L, b, KA): Likewise.
2758         * config/ft32/ft32.c (ft32_load_immediate, ft32_expand_prologue):
2759         Likewise.
2760         * cfgexpand.c (expand_stack_alignment): Likewise.
2761         * gcse.c (insert_expr_in_table): Likewise.
2762         * print-rtl.c (rtx_writer::print_rtx_operand_codes_E_and_V): Likewise.
2763         * cgraphunit.c (cgraph_node::expand): Likewise.
2764         * ira-build.c (setup_min_max_allocno_live_range_point): Likewise.
2765         * emit-rtl.c (add_insn): Likewise.
2766         * input.c (dump_location_info): Likewise.
2767         * passes.c (NEXT_PASS): Likewise.
2768         * read-rtl-function.c (parse_note_insn_name,
2769         function_reader::read_rtx_operand_r, function_reader::parse_mem_expr):
2770         Likewise.
2771         * sched-rgn.c (sched_rgn_init): Likewise.
2772         * diagnostic-show-locus.c (layout::show_ruler): Likewise.
2773         * combine.c (find_split_point, simplify_if_then_else, force_to_mode,
2774         if_then_else_cond, simplify_shift_const_1, simplify_comparison):
2775         Likewise.
2776         * explow.c (eliminate_constant_term): Likewise.
2777         * final.c (leaf_renumber_regs_insn): Likewise.
2778         * cfgrtl.c (print_rtl_with_bb): Likewise.
2779         * genhooks.c (emit_init_macros): Likewise.
2780         * poly-int.h (maybe_ne, maybe_le, maybe_lt): Likewise.
2781         * tree-data-ref.c (conflict_fn): Likewise.
2782         * selftest.c (assert_streq): Likewise.
2783         * expr.c (store_constructor_field, expand_expr_real_1): Likewise.
2784         * fold-const.c (fold_range_test, extract_muldiv_1, fold_truth_andor,
2785         fold_binary_loc, multiple_of_p): Likewise.
2786         * reload.c (push_reload, find_equiv_reg): Likewise.
2787         * et-forest.c (et_nca, et_below): Likewise.
2788         * dbxout.c (dbxout_symbol_location): Likewise.
2789         * reorg.c (relax_delay_slots): Likewise.
2790         * dojump.c (do_compare_rtx_and_jump): Likewise.
2791         * gengtype-parse.c (type): Likewise.
2792         * simplify-rtx.c (simplify_gen_ternary, simplify_gen_relational,
2793         simplify_const_relational_operation): Likewise.
2794         * reload1.c (do_output_reload): Likewise.
2795         * dumpfile.c (get_dump_file_info_by_switch): Likewise.
2796         * gengtype.c (type_for_name): Likewise.
2797         * gimple-ssa-sprintf.c (format_directive): Likewise.
2799 2017-12-19  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
2801         PR target/82975
2802         * config/arm/arm.h (TEST_REGNO): Check reg_renumber is set before
2803         accessing it.  Adjust comment.
2805 2017-12-19  Jakub Jelinek  <jakub@redhat.com>
2807         PR middle-end/81914
2808         * predict.c (zero_one_minusone): New function.
2809         (apply_return_prediction): Avoid return prediction for functions
2810         returning only -1, 0 and 1 values, unless they only return -1 and 0
2811         or 0 and 1.
2813 2017-12-19  Claudiu Zissulescu  <claziss@synopsys.com>
2815         * config/arc/arc.c (legitimate_scaled_address_p): Clean
2816         fall-through warning.
2817         (arc_compute_frame_size): Remove unused variables.
2818         (arc_print_operand): Fix fprintif format.
2819         (arc_can_follow_jump): Clean fall-through warning.
2821 2017-12-19  Marek Polacek  <polacek@redhat.com>
2823         PR c++/83489
2824         * config/i386/i386.c (init_cumulative_args): Don't check TYPE_EMPTY_P
2825         on an error node.
2827 2017-12-19  Claudiu Zissulescu  <claziss@synopsys.com>
2829         * config/arc/arc.c (overriderregs): New variable.
2830         (arc_override_options): Track fixed/call saved/call options.
2831         (arc_conditional_register_usage): Check against overrideregs
2832         variable whenever we change register properties.
2834 2017-12-19  Nathan Sidwell  <nathan@acm.org>
2836         * opts.c (finish_options): Don't prefix dump_base_name if it
2837         already contains directories.
2839 2017-12-19  Martin Liska  <mliska@suse.cz>
2841         PR rtl-optimization/82675
2842         * loop-unroll.c (unroll_loop_constant_iterations): Allocate one
2843         more element in sbitmap.
2845 2017-12-19  Paolo Carlini  <paolo.carlini@oracle.com>
2847         * gimplify.c (gimplify_expr): Use error_operand_p.
2849 2017-12-19  Sebastian Huber  <sebastian.huber@embedded-brains.de>
2851         PR target/83387
2852         * config/rs6000/t-rtems (MULTILIB_REQUIRED): Remove 64-bit soft-float
2853         multilib.
2855 2017-12-19  Daniel Cederman  <cederman@gaisler.com>
2857         * config/sparc/sparc.c (sparc_do_work_around_errata): Make sure
2858         the jump is to a label.
2860 2017-12-19  Jakub Jelinek  <jakub@redhat.com>
2862         PR tree-optimization/83444
2863         * tree-ssa-strlen.c (strlen_check_and_optimize_stmt): Optimize
2864         character loads.
2866         PR ipa/82801
2867         PR ipa/83346
2868         * ipa-inline.c (flatten_remove_node_hook): New function.
2869         (ipa_inline): Keep only nodes with flatten attribute at the end of
2870         the array in the order from ipa_reverse_postorder, only walk that
2871         portion of array for flattening, if there is more than one such
2872         node, temporarily register a removal hook and ignore removed nodes.
2874         PR tree-optimization/80631
2875         * tree-vect-loop.c (vect_create_epilog_for_reduction): Compare
2876         induc_code against MAX_EXPR or MIN_EXPR instead of reduc_fn against
2877         IFN_REDUC_MAX or IFN_REDUC_MIN.
2879 2017-12-18  Jeff Law  <law@redhat.com>
2881         * tree-ssa-dom.c (record_equivalences_from_phis): Do not
2882         record symbolic equivalences from backedges in the CFG.
2884         Revert
2885         2017-11-19  Jeff Law  <law@redhat.com>
2887         * tree-ssa-dom.c (record_equivalences_from_phis): Fix handling
2888         of degenerates resulting from ignoring an edge.
2890 2017-12-18  Martin Sebor  <msebor@redhat.com>
2892         PR middle-end/83373
2893         PR tree-optimization/78450
2894         * tree-ssa-strlen.c (maybe_set_strlen_range): New function.
2895         (handle_builtin_strlen): Call it.
2897 2017-12-18  Segher Boessenkool  <segher@kernel.crashing.org>
2899         PR rtl-optimization/83424
2900         * rtlanal.c (dead_or_set_regno_p): Handle CLOBBER just like SET.
2902 2017-12-18  Marek Polacek  <polacek@redhat.com>
2904         PR middle-end/83463
2905         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref):
2906         Check if TYPE is INTEGRAL_TYPE_P before accessing its min/max
2907         values.
2909 2017-12-18  Claudiu Zissulescu  <claziss@synopsys.com>
2911         * config/arc/arc.md (maddsidi4, maddsidi4_split): Update pattern.
2912         (umaddsidi4, umaddsidi_split): Likewise.
2914 2017-12-18  Claudiu Zissulescu  <claziss@synopsys.com>
2916         * config/arc/arc.c (arc_legitimate_constant_p): Always check all
2917         constants.
2919 2017-12-18  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
2921         PR target/83420
2922         * config/s390/s390.c (s390_option_override): Avoid strncpy.
2924 2017-12-18  Richard Biener  <rguenther@suse.de>
2926         PR tree-optimization/81877
2927         * tree-ssa-loop-im.c (ref_indep_loop_p): Remove safelen parameters.
2928         (outermost_indep_loop): Adjust.
2929         (ref_indep_loop_p_1): Likewise.  Remove safelen handling again.
2930         (can_sm_ref_p): Adjust.
2932 2017-12-18  Richard Biener  <rguenther@suse.de>
2934         PR middle-end/77291
2935         * tree.c (array_at_struct_end_p): Return true if the underlying
2936         object has space for at least one element in excess of what
2937         the array domain specifies.
2939 2017-12-17  Sandra Loosemore  <sandra@codesourcery.com>
2941         * doc/extend.texi (x86 Function Attributes): Reformat nocf_check
2942         example to avoid overfull hbox.
2943         * doc/invoke.texi (Option Summary): Add missing @gol.
2944         (C++ Dialect Options): Reformat -Wnoexcept-type example to avoid
2945         overfull hbox.
2947 2017-12-17  Sandra Loosemore  <sandra@codesourcery.com>
2948             Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
2950         * doc/invoke.texi (Option Summary): Add -mverbose-cost-dump
2951         to AArch64 and ARM lists, plus missing -mflip-thumb for ARM.
2952         (AArch64 Options): Document -mverbose-cost-dump.
2953         (ARM Options): Likewise, plus -mflip-thumb.
2955 2017-12-17  Martin Sebor  <msebor@redhat.com>
2957         PR bootstrap/83446
2958         * gimple-ssa-warn-restrict.c
2959         (builtin_memref::offset_out_of_bounds): Correct the handling of
2960         anti-ranges.
2962 2017-12-17  Sandra Loosemore  <sandra@codesourcery.com>
2964         * doc/invoke.texi (ARC Options): Add missing -mlra entry.
2966 2017-12-17  John David Anglin  <danglin@gcc.gnu.org>
2968         * config/pa/pa.c (pa_som_asm_init_sections): Fix comment.
2970 2017-12-17  Markus Trippelsdorf  <markus@trippelsdorf.de>
2972         * x86-tune-costs.h (skylake_cost, core_cost): Decrease r64 multiply
2973         latencies.
2975 2017-12-16  Sandra Loosemore  <sandra@codesourcery.com>
2977         * doc/invoke.texi: Fix some typos.
2979 2017-12-16  Martin Sebor  <msebor@redhat.com>
2981         PR tree-optimization/78918
2982         * Makefile.in (OBJS): Add gimple-ssa-warn-restrict.o.
2983         * builtins.c (check_sizes): Rename...
2984         (check_access): ...to this.  Rename function arguments for clarity.
2985         (check_memop_sizes): Adjust names.
2986         (expand_builtin_memchr, expand_builtin_memcpy): Same.
2987         (expand_builtin_memmove, expand_builtin_mempcpy): Same.
2988         (expand_builtin_strcat, expand_builtin_stpncpy): Same.
2989         (check_strncat_sizes, expand_builtin_strncat): Same.
2990         (expand_builtin_strncpy, expand_builtin_memset): Same.
2991         (expand_builtin_bzero, expand_builtin_memcmp): Same.
2992         (expand_builtin_memory_chk, maybe_emit_chk_warning): Same.
2993         (maybe_emit_sprintf_chk_warning): Same.
2994         (expand_builtin_strcpy): Adjust.
2995         (expand_builtin_stpcpy): Same.
2996         (expand_builtin_with_bounds): Detect out-of-bounds accesses
2997         in pointer-checking forms of memcpy, memmove, and mempcpy.
2998         (gcall_to_tree_minimal, max_object_size): Define new functions.
2999         * builtins.h (max_object_size): Declare.
3000         * calls.c (alloc_max_size): Call max_object_size instead of
3001         hardcoding ssizetype limit.
3002         (get_size_range): Handle new argument.
3003         * calls.h (get_size_range): Add a new argument.
3004         * cfgexpand.c (expand_call_stmt): Propagate no-warning bit.
3005         * doc/invoke.texi (-Wrestrict): Adjust, add example.
3006         * gimple-fold.c (gimple_fold_builtin_memory_op): Detect overlapping
3007         operations.
3008         (gimple_fold_builtin_memory_chk): Same.
3009         (gimple_fold_builtin_stxcpy_chk): New function.
3010         * gimple-ssa-warn-restrict.c: New source.
3011         * gimple-ssa-warn-restrict.h: New header.
3012         * gimple.c (gimple_build_call_from_tree): Propagate location.
3013         * passes.def (pass_warn_restrict): Add new pass.
3014         * tree-pass.h (make_pass_warn_restrict): Declare.
3015         * tree-ssa-strlen.c (handle_builtin_strcpy): Detect overlapping
3016         operations.
3017         (handle_builtin_strcat): Same.
3018         (strlen_optimize_stmt): Rename...
3019         (strlen_check_and_optimize_stmt): ...to this.  Handle strncat,
3020         stpncpy, strncpy, and their checking forms.
3022 2017-12-16  Jan Hubicka  <hubicka@ucw.cz>
3024         PR rtl-optimization/82849
3025         * modulo-sched.c (sms_schedule): Use get_estimated_loop_iterations_int
3026         and get_max_loop_iterations_int.
3028 2017-12-16  Richard Sandiford  <richard.sandiford@linaro.org>
3029             Alan Hayward  <alan.hayward@arm.com>
3030             David Sherwood  <david.sherwood@arm.com>
3032         * machmode.h (mode_for_size, int_mode_for_size, float_mode_for_size)
3033         (smallest_mode_for_size, smallest_int_mode_for_size): Take the mode
3034         size as a poly_uint64.
3035         (mode_for_vector, mode_for_int_vector): Take the number of vector
3036         elements as a poly_uint64.
3037         * stor-layout.c (mode_for_size, smallest_mode_for_size): Take the mode
3038         size as a poly_uint64.
3039         (mode_for_vector, mode_for_int_vector): Take the number of vector
3040         elements as a poly_uint64.
3042 2017-12-16  Richard Sandiford  <richard.sandiford@linaro.org>
3043             Alan Hayward  <alan.hayward@arm.com>
3044             David Sherwood  <david.sherwood@arm.com>
3046         * machmode.h (MACRO_MODE): New macro.
3047         * addresses.h (base_reg_class, ok_for_base_p_1): Use it.
3048         * targhooks.c (default_libcall_value, default_secondary_reload)
3049         (default_memory_move_cost, default_register_move_cost)
3050         (default_class_max_nregs): Likewise.
3052 2017-12-16  Richard Sandiford  <richard.sandiford@linaro.org>
3053             Alan Hayward  <alan.hayward@arm.com>
3054             David Sherwood  <david.sherwood@arm.com>
3056         * doc/sourcebuild.texi: Document IN_TARGET_CODE.
3057         * genattrtab.c (write_header): Define IN_TARGET_CODE to 1 in the
3058         target C file.
3059         * genautomata.c (main): Likewise.
3060         * genconditions.c (write_header): Likewise.
3061         * genemit.c (main): Likewise.
3062         * genextract.c (print_header): Likewise.
3063         * genopinit.c (main): Likewise.
3064         * genoutput.c (output_prologue): Likewise.
3065         * genpeep.c (main): Likewise.
3066         * genpreds.c (write_insn_preds_c): Likewise.
3067         * genrecog.c (writer_header): Likewise.
3068         * config/aarch64/aarch64-builtins.c (IN_TARGET_CODE): Define.
3069         * config/aarch64/aarch64-c.c (IN_TARGET_CODE): Likewise.
3070         * config/aarch64/aarch64.c (IN_TARGET_CODE): Likewise.
3071         * config/aarch64/cortex-a57-fma-steering.c (IN_TARGET_CODE): Likewise.
3072         * config/aarch64/driver-aarch64.c (IN_TARGET_CODE): Likewise.
3073         * config/alpha/alpha.c (IN_TARGET_CODE): Likewise.
3074         * config/alpha/driver-alpha.c (IN_TARGET_CODE): Likewise.
3075         * config/arc/arc-c.c (IN_TARGET_CODE): Likewise.
3076         * config/arc/arc.c (IN_TARGET_CODE): Likewise.
3077         * config/arc/driver-arc.c (IN_TARGET_CODE): Likewise.
3078         * config/arm/aarch-common.c (IN_TARGET_CODE): Likewise.
3079         * config/arm/arm-builtins.c (IN_TARGET_CODE): Likewise.
3080         * config/arm/arm-c.c (IN_TARGET_CODE): Likewise.
3081         * config/arm/arm.c (IN_TARGET_CODE): Likewise.
3082         * config/arm/driver-arm.c (IN_TARGET_CODE): Likewise.
3083         * config/avr/avr-c.c (IN_TARGET_CODE): Likewise.
3084         * config/avr/avr-devices.c (IN_TARGET_CODE): Likewise.
3085         * config/avr/avr-log.c (IN_TARGET_CODE): Likewise.
3086         * config/avr/avr.c (IN_TARGET_CODE): Likewise.
3087         * config/avr/driver-avr.c (IN_TARGET_CODE): Likewise.
3088         * config/avr/gen-avr-mmcu-specs.c (IN_TARGET_CODE): Likewise.
3089         * config/bfin/bfin.c (IN_TARGET_CODE): Likewise.
3090         * config/c6x/c6x.c (IN_TARGET_CODE): Likewise.
3091         * config/cr16/cr16.c (IN_TARGET_CODE): Likewise.
3092         * config/cris/cris.c (IN_TARGET_CODE): Likewise.
3093         * config/darwin.c (IN_TARGET_CODE): Likewise.
3094         * config/epiphany/epiphany.c (IN_TARGET_CODE): Likewise.
3095         * config/epiphany/mode-switch-use.c (IN_TARGET_CODE): Likewise.
3096         * config/epiphany/resolve-sw-modes.c (IN_TARGET_CODE): Likewise.
3097         * config/fr30/fr30.c (IN_TARGET_CODE): Likewise.
3098         * config/frv/frv.c (IN_TARGET_CODE): Likewise.
3099         * config/ft32/ft32.c (IN_TARGET_CODE): Likewise.
3100         * config/h8300/h8300.c (IN_TARGET_CODE): Likewise.
3101         * config/i386/djgpp.c (IN_TARGET_CODE): Likewise.
3102         * config/i386/driver-i386.c (IN_TARGET_CODE): Likewise.
3103         * config/i386/driver-mingw32.c (IN_TARGET_CODE): Likewise.
3104         * config/i386/host-cygwin.c (IN_TARGET_CODE): Likewise.
3105         * config/i386/host-i386-darwin.c (IN_TARGET_CODE): Likewise.
3106         * config/i386/host-mingw32.c (IN_TARGET_CODE): Likewise.
3107         * config/i386/i386-c.c (IN_TARGET_CODE): Likewise.
3108         * config/i386/i386.c (IN_TARGET_CODE): Likewise.
3109         * config/i386/intelmic-mkoffload.c (IN_TARGET_CODE): Likewise.
3110         * config/i386/msformat-c.c (IN_TARGET_CODE): Likewise.
3111         * config/i386/winnt-cxx.c (IN_TARGET_CODE): Likewise.
3112         * config/i386/winnt-stubs.c (IN_TARGET_CODE): Likewise.
3113         * config/i386/winnt.c (IN_TARGET_CODE): Likewise.
3114         * config/i386/x86-tune-sched-atom.c (IN_TARGET_CODE): Likewise.
3115         * config/i386/x86-tune-sched-bd.c (IN_TARGET_CODE): Likewise.
3116         * config/i386/x86-tune-sched-core.c (IN_TARGET_CODE): Likewise.
3117         * config/i386/x86-tune-sched.c (IN_TARGET_CODE): Likewise.
3118         * config/ia64/ia64-c.c (IN_TARGET_CODE): Likewise.
3119         * config/ia64/ia64.c (IN_TARGET_CODE): Likewise.
3120         * config/iq2000/iq2000.c (IN_TARGET_CODE): Likewise.
3121         * config/lm32/lm32.c (IN_TARGET_CODE): Likewise.
3122         * config/m32c/m32c-pragma.c (IN_TARGET_CODE): Likewise.
3123         * config/m32c/m32c.c (IN_TARGET_CODE): Likewise.
3124         * config/m32r/m32r.c (IN_TARGET_CODE): Likewise.
3125         * config/m68k/m68k.c (IN_TARGET_CODE): Likewise.
3126         * config/mcore/mcore.c (IN_TARGET_CODE): Likewise.
3127         * config/microblaze/microblaze-c.c (IN_TARGET_CODE): Likewise.
3128         * config/microblaze/microblaze.c (IN_TARGET_CODE): Likewise.
3129         * config/mips/driver-native.c (IN_TARGET_CODE): Likewise.
3130         * config/mips/frame-header-opt.c (IN_TARGET_CODE): Likewise.
3131         * config/mips/mips.c (IN_TARGET_CODE): Likewise.
3132         * config/mmix/mmix.c (IN_TARGET_CODE): Likewise.
3133         * config/mn10300/mn10300.c (IN_TARGET_CODE): Likewise.
3134         * config/moxie/moxie.c (IN_TARGET_CODE): Likewise.
3135         * config/msp430/driver-msp430.c (IN_TARGET_CODE): Likewise.
3136         * config/msp430/msp430-c.c (IN_TARGET_CODE): Likewise.
3137         * config/msp430/msp430.c (IN_TARGET_CODE): Likewise.
3138         * config/nds32/nds32-cost.c (IN_TARGET_CODE): Likewise.
3139         * config/nds32/nds32-fp-as-gp.c (IN_TARGET_CODE): Likewise.
3140         * config/nds32/nds32-intrinsic.c (IN_TARGET_CODE): Likewise.
3141         * config/nds32/nds32-isr.c (IN_TARGET_CODE): Likewise.
3142         * config/nds32/nds32-md-auxiliary.c (IN_TARGET_CODE): Likewise.
3143         * config/nds32/nds32-memory-manipulation.c (IN_TARGET_CODE): Likewise.
3144         * config/nds32/nds32-pipelines-auxiliary.c (IN_TARGET_CODE): Likewise.
3145         * config/nds32/nds32-predicates.c (IN_TARGET_CODE): Likewise.
3146         * config/nds32/nds32.c (IN_TARGET_CODE): Likewise.
3147         * config/nios2/nios2.c (IN_TARGET_CODE): Likewise.
3148         * config/nvptx/mkoffload.c (IN_TARGET_CODE): Likewise.
3149         * config/nvptx/nvptx.c (IN_TARGET_CODE): Likewise.
3150         * config/pa/pa.c (IN_TARGET_CODE): Likewise.
3151         * config/pdp11/pdp11.c (IN_TARGET_CODE): Likewise.
3152         * config/powerpcspe/driver-powerpcspe.c (IN_TARGET_CODE): Likewise.
3153         * config/powerpcspe/host-darwin.c (IN_TARGET_CODE): Likewise.
3154         * config/powerpcspe/host-ppc64-darwin.c (IN_TARGET_CODE): Likewise.
3155         * config/powerpcspe/powerpcspe-c.c (IN_TARGET_CODE): Likewise.
3156         * config/powerpcspe/powerpcspe-linux.c (IN_TARGET_CODE): Likewise.
3157         * config/powerpcspe/powerpcspe.c (IN_TARGET_CODE): Likewise.
3158         * config/riscv/riscv-builtins.c (IN_TARGET_CODE): Likewise.
3159         * config/riscv/riscv-c.c (IN_TARGET_CODE): Likewise.
3160         * config/riscv/riscv.c (IN_TARGET_CODE): Likewise.
3161         * config/rl78/rl78-c.c (IN_TARGET_CODE): Likewise.
3162         * config/rl78/rl78.c (IN_TARGET_CODE): Likewise.
3163         * config/rs6000/driver-rs6000.c (IN_TARGET_CODE): Likewise.
3164         * config/rs6000/host-darwin.c (IN_TARGET_CODE): Likewise.
3165         * config/rs6000/host-ppc64-darwin.c (IN_TARGET_CODE): Likewise.
3166         * config/rs6000/rs6000-c.c (IN_TARGET_CODE): Likewise.
3167         * config/rs6000/rs6000-linux.c (IN_TARGET_CODE): Likewise.
3168         * config/rs6000/rs6000-p8swap.c (IN_TARGET_CODE): Likewise.
3169         * config/rs6000/rs6000-string.c (IN_TARGET_CODE): Likewise.
3170         * config/rs6000/rs6000.c (IN_TARGET_CODE): Likewise.
3171         * config/rx/rx.c (IN_TARGET_CODE): Likewise.
3172         * config/s390/driver-native.c (IN_TARGET_CODE): Likewise.
3173         * config/s390/s390-c.c (IN_TARGET_CODE): Likewise.
3174         * config/s390/s390.c (IN_TARGET_CODE): Likewise.
3175         * config/sh/sh-c.c (IN_TARGET_CODE): Likewise.
3176         * config/sh/sh-mem.cc (IN_TARGET_CODE): Likewise.
3177         * config/sh/sh.c (IN_TARGET_CODE): Likewise.
3178         * config/sh/sh_optimize_sett_clrt.cc (IN_TARGET_CODE): Likewise.
3179         * config/sh/sh_treg_combine.cc (IN_TARGET_CODE): Likewise.
3180         * config/sparc/driver-sparc.c (IN_TARGET_CODE): Likewise.
3181         * config/sparc/sparc-c.c (IN_TARGET_CODE): Likewise.
3182         * config/sparc/sparc.c (IN_TARGET_CODE): Likewise.
3183         * config/spu/spu-c.c (IN_TARGET_CODE): Likewise.
3184         * config/spu/spu.c (IN_TARGET_CODE): Likewise.
3185         * config/stormy16/stormy16.c (IN_TARGET_CODE): Likewise.
3186         * config/tilegx/mul-tables.c (IN_TARGET_CODE): Likewise.
3187         * config/tilegx/tilegx-c.c (IN_TARGET_CODE): Likewise.
3188         * config/tilegx/tilegx.c (IN_TARGET_CODE): Likewise.
3189         * config/tilepro/mul-tables.c (IN_TARGET_CODE): Likewise.
3190         * config/tilepro/tilepro-c.c (IN_TARGET_CODE): Likewise.
3191         * config/tilepro/tilepro.c (IN_TARGET_CODE): Likewise.
3192         * config/v850/v850-c.c (IN_TARGET_CODE): Likewise.
3193         * config/v850/v850.c (IN_TARGET_CODE): Likewise.
3194         * config/vax/vax.c (IN_TARGET_CODE): Likewise.
3195         * config/visium/visium.c (IN_TARGET_CODE): Likewise.
3196         * config/vms/vms-c.c (IN_TARGET_CODE): Likewise.
3197         * config/vms/vms-f.c (IN_TARGET_CODE): Likewise.
3198         * config/vms/vms.c (IN_TARGET_CODE): Likewise.
3199         * config/xtensa/xtensa.c (IN_TARGET_CODE): Likewise.
3201 2017-12-16  Richard Sandiford  <richard.sandiford@linaro.org>
3202             Alan Hayward  <alan.hayward@arm.com>
3203             David Sherwood  <david.sherwood@arm.com>
3205         * tree-ssa-sccvn.c (vn_reference_lookup_3): Avoid repeated
3206         checks for MEM_REF.
3208 2017-12-16  Richard Sandiford  <richard.sandiford@linaro.org>
3209             Alan Hayward  <alan.hayward@arm.com>
3210             David Sherwood  <david.sherwood@arm.com>
3212         * doc/generic.texi (VEC_SERIES_EXPR): Document.
3213         * doc/md.texi (vec_series@var{m}): Document.
3214         * tree.def (VEC_SERIES_EXPR): New tree code.
3215         * tree.h (build_vec_series): Declare.
3216         * tree.c (build_vec_series): New function.
3217         * cfgexpand.c (expand_debug_expr): Handle VEC_SERIES_EXPR.
3218         * tree-pretty-print.c (dump_generic_node): Likewise.
3219         * gimple-pretty-print.c (dump_binary_rhs): Likewise.
3220         * tree-inline.c (estimate_operator_cost): Likewise.
3221         * expr.c (expand_expr_real_2): Likewise.
3222         * optabs-tree.c (optab_for_tree_code): Likewise.
3223         * tree-cfg.c (verify_gimple_assign_binary): Likewise.
3224         * fold-const.c (const_binop): Fold VEC_SERIES_EXPRs of constants.
3225         * expmed.c (make_tree): Handle VEC_SERIES.
3226         * optabs.def (vec_series_optab): New optab.
3227         * optabs.h (expand_vec_series_expr): Declare.
3228         * optabs.c (expand_vec_series_expr): New function.
3229         * tree-vect-generic.c (expand_vector_operations_1): Check that
3230         the operands also have vector type.
3232 2017-12-16  Richard Sandiford  <richard.sandiford@linaro.org>
3233             Alan Hayward  <alan.hawyard@arm.com>
3234             David Sherwood  <david.sherwood@arm.com>
3236         * doc/generic.texi (VEC_DUPLICATE_EXPR): Document.
3237         (VEC_COND_EXPR): Add missing @tindex.
3238         * doc/md.texi (vec_duplicate@var{m}): Document.
3239         * tree.def (VEC_DUPLICATE_EXPR): New tree codes.
3240         * tree.c (build_vector_from_val): Add stubbed-out handling of
3241         variable-length vectors, using VEC_DUPLICATE_EXPR.
3242         (uniform_vector_p): Handle VEC_DUPLICATE_EXPR.
3243         * cfgexpand.c (expand_debug_expr): Likewise.
3244         * tree-cfg.c (verify_gimple_assign_unary): Likewise.
3245         * tree-inline.c (estimate_operator_cost): Likewise.
3246         * tree-pretty-print.c (dump_generic_node): Likewise.
3247         * tree-vect-generic.c (ssa_uniform_vector_p): Likewise.
3248         * fold-const.c (const_unop): Fold VEC_DUPLICATE_EXPRs of a constant.
3249         (test_vec_duplicate_folding): New function.
3250         (fold_const_c_tests): Call it.
3251         * optabs.def (vec_duplicate_optab): New optab.
3252         * optabs-tree.c (optab_for_tree_code): Handle VEC_DUPLICATE_EXPR.
3253         * optabs.h (expand_vector_broadcast): Declare.
3254         * optabs.c (expand_vector_broadcast): Make non-static.  Try using
3255         vec_duplicate_optab.
3256         * expr.c (store_constructor): Try using vec_duplicate_optab for
3257         uniform vectors.
3258         (expand_expr_real_2): Handle VEC_DUPLICATE_EXPR.
3260 2017-12-15  Markus Trippelsdorf  <markus@trippelsdorf.de>
3262         PR target/83358
3263         * config/i386/x86-tune-costs.h (skylake_cost, core_cost): Increase
3264         div/mod latencies a bit.
3266 2017-12-15  Jeff Law  <law@redhat.com>
3268         PR tree-optimization/36550
3269         * tree-ssa-threadupdate.c (count_stmts_and_phis_in_block): New.
3270         (mark_threaded_blocks): Rewrite code to avoid block copying when
3271         optimizing for size.  Don't pessimize blocks which will be
3272         copied, but all the statements will be dead.
3274 2017-12-15  Alexandre Oliva  <aoliva@redhat.com>
3276         PR tree-optimization/81165
3277         * tree-ssa-threadupdate.c (uses_in_bb): New.
3278         (estimate_threading_killed_stmts): New.
3279         * tree-ssa-threadupdate.h (estimate_threading_killed_stmts): Prototype.
3280         * tree-ssa-threadedge.c
3281         (record_temporary_equivalences_from_stmts_at_dest): Expand limit
3282         when its hit.
3284 2017-12-15  Jeff Law  <law@redhat.com>
3286         PR tree-optimization/83410
3287         * tree-ssa-threadupdate.c (thread_block_1): Avoid certain jump
3288         threads when parallelizing loops.
3290 2017-12-15  Jakub Jelinek  <jakub@redhat.com>
3292         * tree-core.h (struct attribute_spec): Swap affects_type_identity and
3293         handler fields.
3294         * config/alpha/alpha.c (vms_attribute_table): Swap
3295         affects_type_identity and handler fields, adjust comments.
3296         * config/mips/mips.c (mips_attribute_table): Likewise.
3297         * config/visium/visium.c (visium_attribute_table): Likewise.
3298         * config/epiphany/epiphany.c (epiphany_attribute_table): Likewise.
3299         * config/microblaze/microblaze.c (microblaze_attribute_table):
3300         Likewise.
3301         * config/spu/spu.c (spu_attribute_table): Likewise.
3302         * config/mcore/mcore.c (mcore_attribute_table): Likewise.
3303         * config/arc/arc.c (arc_attribute_table): Likewise.
3304         * config/m68k/m68k.c (m68k_attribute_table): Likewise.
3305         * config/v850/v850.c (v850_handle_interrupt_attribute,
3306         v850_handle_data_area_attribute): Formatting fixes.
3307         (v850_attribute_table): Swap affects_type_identity and handler
3308         fields, adjust comments.
3309         * config/m32r/m32r.c (m32r_attribute_table): Likewise.
3310         * config/arm/arm.c (arm_attribute_table): Likewise.
3311         * config/avr/avr.c (avr_attribute_table): Likewise.
3312         * config/s390/s390.c (s390_attribute_table): Likewise.
3313         * config/sh/sh.c (sh_attribute_table): Likewise.
3314         * config/i386/i386.c (ix86_handle_cconv_attribute,
3315         ix86_handle_callee_pop_aggregate_return): Formatting fixes.
3316         (ix86_attribute_table): Swap affects_type_identity and handler
3317         fields, adjust comments.
3318         * config/i386/cygming.h (SUBTARGET_ATTRIBUTE_TABLE): Likewise.
3319         * config/sparc/sparc.c (sparc_attribute_table): Likewise.
3320         * config/m32c/m32c.c (m32c_attribute_table): Likewise.
3321         * config/sol2.h (SOLARIS_ATTRIBUTE_TABLE): Likewise.
3322         * config/ia64/ia64.c (ia64_attribute_table): Likewise.
3323         * config/msp430/msp430.c (msp430_attribute_table): Likewise.
3324         * config/rx/rx.c (rx_attribute_table): Likewise.
3325         * config/cr16/cr16.c (cr16_attribute_table): Likewise.
3326         * config/h8300/h8300.c (h8300_attribute_table): Likewise.
3327         * config/nvptx/nvptx.c (nvptx_attribute_table): Likewise.
3328         * config/powerpcspe/powerpcspe.c (rs6000_attribute_table): Likewise.
3329         * config/darwin.h (SUBTARGET_ATTRIBUTE_TABLE): Likewise.
3330         * config/stormy16/stormy16.c (xstormy16_attribute_table): Likewise.
3331         * config/bfin/bfin.c (bfin_attribute_table): Likewise.
3332         * config/rs6000/rs6000.c (rs6000_attribute_table): Likewise.
3333         * config/rl78/rl78.c (rl78_attribute_table): Likewise.
3334         * config/nds32/nds32.c (nds32_attribute_table): Likewise.
3335         * doc/plugins.texi (user_attr): Likewise.  Add NULL for
3336         exclude.
3337         * attribs.c (empty_attribute_table): Swap affects_type_identity and
3338         handler fields.
3339         (register_scoped_attributes, decl_attributes): Formatting fixes.
3341         PR tree-optimization/83269
3342         * fold-const.c (fold_binary_loc): Perform (-A) - B -> (-B) - A
3343         subtraction in arg0's type if type is signed and arg0 is unsigned.
3344         Formatting fix.
3346         PR sanitizer/81281
3347         * match.pd ((T)(P + A) - (T)P -> (T) A): Use @@0 instead of @0 and
3348         convert? on @0 instead of convert.  Check type of @1, not @0.
3349         ((T)P - (T)(P + A) -> -(T) A): Use @@0 instead of @0 and
3350         convert? on @0 instead of convert.  Check type of @1, not @0.
3351         ((T)(P + A) - (T)(P + B) -> (T)A - (T)B): Use @@0 instead of @0,
3352         only optimize if either both @1 and @2 types are narrower
3353         precision, or both are wider or equal precision, and in the former
3354         case only if both have undefined overflow.
3356 2017-12-15  Richard Biener  <rguenther@suse.de>
3358         PR lto/83388
3359         * internal-fn.def (IFN_NOP): Add.
3360         * internal-fn.c (expand_NOP): Do nothing.
3361         * lto-streamer-in.c (input_function): Instead of removing
3362         sanitizer calls replace them with IFN_NOP calls.
3364 2017-12-15  Richard Sandiford  <richard.sandiford@linaro.org>
3365             Alan Hayward  <alan.hayward@arm.com>
3366             David Sherwood  <david.sherwood@arm.com>
3368         * dse.c (store_info, read_info_type): Replace begin and end with
3369         offset and width.
3370         (print_range): New function.
3371         (set_all_positions_unneeded, any_positions_needed_p)
3372         (check_mem_read_rtx, scan_stores, scan_reads, dse_step5): Update
3373         accordingly.
3374         (record_store): Likewise.  Optimize the case in which all positions
3375         are unneeded.
3376         (get_stored_val): Replace read_begin and read_end with read_offset
3377         and read_width.
3378         (replace_read): Update call accordingly.
3380 2017-12-15  Bin Cheng  <bin.cheng@arm.com>
3382         * gimple-loop-interchange.cc (STMT_COST_RATIO): New macro.
3383         (loop_cand::m_num_stmts, loop_cand::m_const_init_reduc): New members.
3384         (loop_cand::loop_cand): Initialize above members.
3385         (loop_cand::supported_operations): Delete.
3386         (loop_cand::can_interchange_p): Inline above function.
3387         (loop_cand::classify_simple_reduction): Record number of constant
3388         initialized simple reductions.
3389         (should_interchange_loops): New parameters.  Check stmt cost of loops
3390         to be interchange.
3391         (tree_loop_interchange::interchange): Prepare stmt cost of outer loop.
3392         Update call to should_interchange_loops.
3393         (should_interchange_loop_nest): Update call to
3394         should_interchange_loops.
3396 2017-12-15  Eric Botcazou  <ebotcazou@adacore.com>
3398         PR target/66488
3399         * ggc-page.c (HOST_BITS_PER_PTR): Do not define here...
3400         * hwint.h (HOST_BITS_PER_PTR): ...but here instead.
3401         * config/i386/xm-mingw32.h (HOST_BITS_PER_PTR): Delete.
3403 2017-12-15  Julia Koval  <julia.koval@intel.com>
3405         * config/i386/i386-builtin.def (__builtin_ia32_vaesenclast_v16qi,
3406         __builtin_ia32_vaesenclast_v32qi, __builtin_ia32_vaesenclast_v64qi):
3407         New.
3408         * config/i386/sse.md (vaesenclast_<mode>): New pattern.
3409         * config/i386/vaesintrin.h (_mm256_aesenclast_epi128,
3410         _mm512_aesenclast_epi128, _mm_aesenclast_epi128): New intrinsics.
3412 2017-12-15  Julia Koval  <julia.koval@intel.com>
3414         * config/i386/i386-builtin.def (__builtin_ia32_vaesenc_v16qi,
3415         __builtin_ia32_vaesenc_v32qi, __builtin_ia32_vaesenc_v64qi): New.
3416         * config/i386/sse.md (vaesenc_<mode>): New pattern.
3417         * config/i386/vaesintrin.h (_mm256_aesenc_epi128, _mm512_aesenc_epi128,
3418         _mm_aesenc_epi128): New intrinsics.
3420 2017-12-15  Julia Koval  <julia.koval@intel.com>
3422         * config/i386/i386-builtin.def (__builtin_ia32_vaesdeclast_v16qi,
3423         __builtin_ia32_vaesdeclast_v32qi, __builtin_ia32_vaesdeclast_v64qi):
3424         New.
3425         * config/i386/sse.md (vaesdeclast_<mode>): New pattern.
3426         * config/i386/vaesintrin.h (_mm256_aesdeclast_epi128,
3427         _mm512_aesdeclast_epi128, _mm_aesdeclast_epi128): New intrinsics.
3429 2017-12-14  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
3431         * gimple-ssa-strength-reduction.c (analyze_increments):
3432         Distinguish replacement costs for constant strides from those for
3433         unknown strides.
3435 2017-12-14  Jakub Jelinek  <jakub@redhat.com>
3437         * var-tracking.c (variable_tracking_main_1): Formatting fix.
3439 2017-12-14  Bernd Edlinger  <bernd.edlinger@hotmail.de>
3441         * doc/invoke.texi: Document -Wcast-function-type.
3442         * recog.h (stored_funcptr): Change signature.
3443         * tree-dump.c (dump_node): Avoid warning.
3444         * typed-splay-tree.h (typed_splay_tree): Avoid warning.
3446 2017-12-14  Qing Zhao  <qing.zhao@oracle.com>
3448         PR middle_end/79538
3449         * gimple-fold.c (get_range_strlen): Add the handling of non-member 
3450         array.
3452 2017-12-14  David Malcolm  <dmalcolm@redhat.com>
3454         PR tree-optimization/83312
3455         * domwalk.h (dom_walker::dom_walker): Fix typo in comment.
3456         * tree-cfg.c (find_taken_edge): Update to handle NULL_TREE for
3457         "val" param, and to cope with arbitrary basic blocks.
3458         (find_taken_edge_cond_expr): Add "cond_stmt" param and use it to
3459         handle NULL_TREE for "val", dropping "bb" param.
3460         (find_taken_edge_switch_expr): Make "switch_stmt" param const and
3461         drop "bb" param.  Handle NULL_TREE for "val".
3462         (find_case_label_for_value): Make "switch_stmt" param const.
3463         * tree-vrp.c (class check_array_bounds_dom_walker): New subclass
3464         of dom_walker.
3465         (vrp_prop::check_all_array_refs): Reimplement as...
3466         (check_array_bounds_dom_walker::before_dom_children): ...this new
3467         vfunc.  Replace linear search through BB block list, excluding
3468         those with non-executable in-edges via dominator walk.
3470 2017-12-14  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
3472         * config/arm/arm.opt (mverbose-cost-dump): New option.
3473         * config/arm/arm.c (arm_rtx_costs): Use it.
3475 2017-12-14  Andreas Schwab  <schwab@linux-m68k.org>
3477         PR bootstrap/83396
3478         * reload1.c (emit_input_reload_insns): Skip debug markers.
3480 2017-12-14  Alexandre Oliva  <aoliva@redhat.com>
3482         * config/i386/i386.c (rest_of_insert_endbranch): Use call loc
3483         for its nop_endbr.
3485         PR bootstrap/83396
3486         * config/arc/arc.c (hwloop_optimize): Skip debug insns.
3487         * config/sh/sh-protos.h (sh_find_set_of_reg): Adjust.
3488         * config/sh/sh.c: Skip debug insns besides notes.
3489         * config/sh/sh.md: Likewise.
3490         * config/sh/sh_treg_combine.cc: Likewise.
3491         * config/sh/sync.md: Likewise.
3493 2017-12-14  Tom de Vries  <tom@codesourcery.com>
3495         * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
3496         Add item for weak_undefined.
3498 2017-12-14  Richard Biener  <rguenther@suse.de>
3500         PR tree-optimization/67842
3501         * sese.h (bb_in_region): Remove #if 0'ed code.
3503 2017-12-14  Richard Biener  <rguenther@suse.de>
3505         PR tree-optimization/83326
3506         * tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Add
3507         may_be_zero parameter and handle it by not marking the first
3508         peeled copy as not exiting the loop.
3509         (try_peel_loop): Likewise.
3510         (canonicalize_loop_induction_variables): Use number_of_iterations_exit
3511         to handle the case of constant or zero iterations and perform
3512         loop header copying on-the-fly.
3514 2017-12-14  Richard Biener  <rguenther@suse.de>
3516         PR tree-optimization/83418
3517         * vr-values.c (vr_values::extract_range_for_var_from_comparison_expr):
3518         Instead of asserting we don't get unfolded comparisons deal with them.
3520 2017-12-14  Jakub Jelinek  <jakub@redhat.com>
3522         PR bootstrap/83396
3523         * var-tracking.c (vt_initialize): Ignore non-DEBUG_INSNs outside of
3524         basic blocks.  Assert debug bind insns don't appear outside of bbs,
3525         don't reset them.  Assert insns without BLOCK_FOR_INSN are outside of
3526         bb.  Simplify.
3528         PR tree-optimization/83198
3529         * gimple-ssa-sprintf.c (format_floating): Set type solely based on
3530         dir.modifier, regardless of TREE_TYPE (arg).  Assume non-REAL_CST
3531         value if arg is a REAL_CST with incompatible type.
3533 2017-12-14  Sudakshina Das  <sudi.das@arm.com>
3534             Bin Cheng  <bin.cheng@arm.com>
3536         PR target/81228
3537         * config/aarch64/aarch64.c (aarch64_select_cc_mode): Move LTGT
3538         to CCFPEmode.
3539         * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Add
3540         LTGT.
3542 2017-12-14  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
3544         * config/aarch64/aarch64-cores.def (cortex-a55, cortex-a75,
3545         cortex-a75.cortex-a55): Specify AARCH64_FL_F16 in the arch features.
3547 2017-12-14  Richard Sandiford  <richard.sandiford@linaro.org>
3548             Alan Hayward  <alan.hayward@arm.com>
3549             David Sherwood  <david.sherwood@arm.com>
3551         * poly-int.h: New file.
3552         * poly-int-types.h: Likewise.
3553         * coretypes.h: Include them.
3554         (POLY_INT_CONVERSION): Define.
3555         * target.def (estimated_poly_value): New hook.
3556         * doc/tm.texi.in (TARGET_ESTIMATED_POLY_VALUE): New hook.
3557         * doc/tm.texi: Regenerate.
3558         * doc/poly-int.texi: New file.
3559         * doc/gccint.texi: Include it.
3560         * doc/rtl.texi: Describe restrictions on subreg modes.
3561         * Makefile.in (TEXI_GCCINT_FILES): Add poly-int.texi.
3562         * genmodes.c (NUM_POLY_INT_COEFFS): Provide a default definition.
3563         (emit_insn_modes_h): Emit a definition of NUM_POLY_INT_COEFFS.
3564         * targhooks.h (default_estimated_poly_value): Declare.
3565         * targhooks.c (default_estimated_poly_value): New function.
3566         * target.h (estimated_poly_value): Likewise.
3567         * wide-int.h (WI_UNARY_RESULT): Use wi::binary_traits.
3568         (wi::unary_traits): Delete.
3569         (wi::binary_traits::signed_shift_result_type): Define for
3570         offset_int << HOST_WIDE_INT, etc.
3571         (generic_wide_int::operator <<=): Define for all types and use
3572         wi::lshift instead of <<.
3573         (wi::hwi_with_prec): Add a default constructor.
3574         (wi::ints_for): New class.
3575         (operator <<): Define for all wide-int types.
3576         (operator /): New function.
3577         (operator %): Likewise.
3578         * selftest.h (ASSERT_KNOWN_EQ, ASSERT_KNOWN_EQ_AT, ASSERT_MAYBE_NE)
3579         (ASSERT_MAYBE_NE_AT): New macros.
3581 2017-12-13  Eric Botcazou  <ebotcazou@adacore.com>
3582             Dominik Vogt  <vogt@linux.vnet.ibm.com>
3584         PR middle-end/78468
3585         * emit-rtl.c (init_emit): Remove ??? comment.
3586         * explow.c (get_dynamic_stack_size): Take known alignment of stack
3587         pointer + STACK_DYNAMIC_OFFSET into account in lieu of STACK_BOUNDARY.
3588         * config/sparc/sparc.h (INIT_EXPANDERS): In 32-bit mode, lower the
3589         alignment of 3 virtual registers to BITS_PER_WORD.
3591         * config/sparc/sparc.c (sparc_compute_frame_size): Simplify.
3593 2017-12-13  Peter Bergner  <bergner@vnet.ibm.com>
3595         * config/rs6000/ppc-auxv.h (PPC_FEATURE2_HTM_NO_SUSPEND): New define.
3596         * config/rs6000/rs6000.c (cpu_supports_info): Use it.
3598 2017-12-13  Alexandre Oliva  <aoliva@redhat.com>
3600         PR bootstrap/83396
3601         * reload1.c (eliminate_regs_in_insn): Skip debug markers.
3603 2017-12-13  Jakub Jelinek  <jakub@redhat.com>
3605         * tree-cfg.c (verify_gimple_in_cfg): Verify no non-label stmts
3606         with the exception of debug begin stmt markers appear before
3607         labels.
3609         PR bootstrap/83396
3610         * final.c (rest_of_handle_final): Call variable_tracking_main only
3611         if !flag_var_tracking.
3613 2017-12-13  Alexandre Oliva  <aoliva@redhat.com>
3614             Jakub Jelinek  <jakub@redhat.com>
3616         PR bootstrap/83396
3617         PR debug/83391
3618         * tree-cfgcleanup.c (remove_forwarder_block): Keep after
3619         labels debug stmts that can only appear after labels.
3621 2017-12-13  Alexander Monakov  <amonakov@ispras.ru>
3623         PR rtl-optimization/82398
3624         * sel-sched.c (sel_rank_for_schedule): Fix check for zero
3625         EXPR_USEFULNESS in priority comparison.
3627 2017-12-13  Segher Boessenkool  <segher@kernel.crashing.org>
3629         PR rtl-optimization/83393
3630         * combine.c (move_deaths): If reg_stat points to a too new insn in
3631         last_death, do not use it: find the proper insn instead.
3633 2017-12-12  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
3635         PR target/82190
3636         * config/rs6000/rs6000-string.c (expand_block_compare,
3637         expand_strn_compare): Fix set_mem_size() calls.
3639 2017-12-12  Jeff Law  <law@redhat.com>
3641         PR tree-optimization/83298
3642         PR tree-optimization/83362
3643         PR tree-optimization/83383
3644         * gimple-ssa-evrp-analyze.h (class evrp_range_analyzer): Make
3645         push_value_range a public interface.  Add new argument to
3646         record_ranges_from_stmt.
3647         * gimple-ssa-evrp-analyze.c
3648         (evrp_range_analyzer::record_ranges_from_stmt): Add new argument.
3649         Update comments.  Handle recording temporary equivalences.
3650         * tree-ssa-dom.c (dom_opt_opt_walker::before_dom_children): Add
3651         new argument to call to evrp_range_analyzer::record_ranges_from_stmt.
3652         * gimple-ssa-evrp.c (evrp_dom_walker::before_dom_children): Likewise.
3653         * tree-ssa-threadedge.c: Include alloc-pool.h, vr-values.h and
3654         gimple-ssa-evrp-analyze.h.
3655         (record_temporary_equivalences_from_phis): Add new argument.  When
3656         the PHI arg is an SSA_NAME, set the result's range to the range
3657         of the PHI arg.
3658         (record_temporary_equivalences_from_stmts_at_dest): Record ranges
3659         from statements too.
3660         (thread_through_normal_block): Accept new argument, evrp_range_analyzer.
3661         Pass it down to children as needed.
3662         (thread_outgoing_edges): Likewise.
3663         (thread_across_edge): Likewise.   Push/pop range state as needed.
3664         * tree-ssa-threadedge.h (thread_outgoing_edges): Update prototype.
3666 2017-12-12  Julia Koval  <julia.koval@intel.com>
3668         * config/i386/i386.c (PTA_SKYLAKE_AVX512): Add PTA_CLWB.
3669         (PTA_CANNONLAKE): Remove PTA_CLWB.
3671 2017-12-12  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
3673         PR target/83332
3674         * config/rs6000/vector.md (vcondv2dfv2di): New define_expand.
3675         (vcondv2div2df): Likewise.
3676         (vconduv2dfv2di): Likewise.
3678 2017-12-12  Jakub Jelinek  <jakub@redhat.com>
3680         * builtins.def (BUILT_IN_NEXTAFTER, BUILT_IN_NEXTAFTERF,
3681         BUILT_IN_NEXTAFTERL, BUILT_IN_NEXTTOWARD, BUILT_IN_NEXTTOWARDF,
3682         BUILT_IN_NEXTTOWARDL): Use ATTR_MATHFN_ERRNO instead of
3683         ATTR_MATHFN_FPROUNDING_ERRNO.
3685 2017-12-12  Richard Biener  <rguenther@suse.de>
3687         PR tree-optimization/83385
3688         * graphite-scop-detection.c (get_order, order): Remove.
3689         (bb_to_rpo): New global.
3690         (cmp_pbbs): Adjust.
3691         (build_scops): Sort pbbs in RPO order.
3693 2017-12-12  James Greenhalgh  <james.greenhalgh@arm.com>
3695         * combine.c (simplify_set): Do not transform subregs to zero_extends
3696         if the destination mode is a vector mode.
3698 2017-12-12  Jakub Jelinek  <jakub@redhat.com>
3700         PR tree-optimization/83359
3701         * tree-cfg.h (fold_loop_internal_call): Declare.
3702         * tree-vectorizer.c (fold_loop_internal_call): Moved to ...
3703         * tree-cfg.c (fold_loop_internal_call): ... here.  No longer static.
3704         (find_loop_dist_alias): New function.
3705         (move_sese_region_to_fn): If any dloop->orig_loop_num value is
3706         updated, also adjust any corresponding LOOP_DIST_ALIAS internal
3707         calls.
3709         PR tree-optimization/80631
3710         * tree-vect-loop.c (get_initial_def_for_reduction): Fix comment typo.
3711         (vect_create_epilog_for_reduction): Add INDUC_VAL and INDUC_CODE
3712         arguments, for INTEGER_INDUC_COND_REDUCTION use INDUC_VAL instead of
3713         hardcoding zero as the value if COND_EXPR is never true.  For
3714         INTEGER_INDUC_COND_REDUCTION don't emit the final COND_EXPR if
3715         INDUC_VAL is equal to INITIAL_DEF, and use INDUC_CODE instead of
3716         hardcoding MAX_EXPR as the reduction operation.
3717         (is_nonwrapping_integer_induction): Allow negative step.
3718         (vectorizable_reduction): Compute INDUC_VAL and INDUC_CODE for
3719         vect_create_epilog_for_reduction, if no value is suitable, don't
3720         use INTEGER_INDUC_COND_REDUCTION for now.  Formatting fixes.
3722 2017-12-12  Richard Biener  <rguenther@suse.de>
3724         PR tree-optimization/81889
3725         * tree-ssa-loop-niter.c (infer_loop_bounds_from_signedness): Use
3726         range info from the non-wrapping IV instead of just the range
3727         of the type.
3729 2017-12-12  Julia Koval  <julia.koval@intel.com>
3731         * config.gcc: Add vaesintrin.h.
3732         * config/i386/i386-builtin-types.def (V64QI_FTYPE_V64QI_V64QI):
3733         New type.
3734         * config/i386/i386-builtin.def (__builtin_ia32_vaesdec_v16qi,
3735         __builtin_ia32_vaesdec_v32qi, __builtin_ia32_vaesdec_v64qi):
3736         New builtins.
3737         * config/i386/i386.c (ix86_expand_args_builtin): Handle new type.
3738         * config/i386/immintrin.h: Include vaesintrin.h.
3739         * config/i386/sse.md (vaesdec_<mode>): New pattern.
3740         * config/i386/vaesintrin.h (_mm256_aesdec_epi128, _mm512_aesdec_epi128,
3741         _mm_aesdec_epi128): New intrinsics.
3743 2017-12-12  Julia Koval  <julia.koval@intel.com>
3745         * common/config/i386/i386-common.c (OPTION_MASK_ISA_VAES_SET,
3746         OPTION_MASK_ISA_VAES_UNSET): New.
3747         (ix86_handle_option): Handle -mvaes.
3748         * config/i386/cpuid.h: Define bit_VAES.
3749         * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mvaes.
3750         * config/i386/i386-c.c (__VAES__): New.
3751         * config/i386/i386.c (ix86_target_string): Add -mvaes.
3752         (ix86_valid_target_attribute_inner_p): Ditto.
3753         * config/i386/i386.h (TARGET_VAES, TARGET_VAES_P): New.
3754         * config/i386/i386.opt: Add -mvaes.
3755         * doc/invoke.texi: Ditto.
3757 2017-12-12  Alexandre Oliva  <aoliva@redhat.com>
3759         * debug.h (gcc_debug_hooks): Add inline_entry.
3760         * dbxout.c (dbx_debug_hooks, xcoff_debug_hooks): Likewise.
3761         * debug.c (do_nothing_debug_hooks): Likewise.
3762         * vmsdbgout.c (vmsdbg_debug_hooks): Likewise.
3763         * dwarf2out.c (dwarf2_debug_hooks): Likewise.
3764         (dwarf2_lineno_debug_hooks): Likewise.
3766         * common.opt (gstatement-frontiers): New, setting
3767         debug_nonbind_markers_p.
3768         * rtl.h (MAY_HAVE_DEBUG_MARKER_INSNS): Activate.
3769         * toplev.c (process_options): Autodetect value for debug statement
3770         frontiers option.
3771         * tree.h (MAY_HAVE_DEBUG_MARKER_STMTS): Activate.
3772         * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers): New.
3774         * cfgexpand.c (expand_gimple_basic_block): Handle begin stmt
3775         markers.  Integrate source bind into debug stmt expand loop.
3776         (pass_expand::execute): Check debug marker limit.  Avoid deep
3777         TER and expand debug locations for debug bind insns only.
3778         * cse.c (insn_live_p): Keep nonbind markers and debug bindings
3779         followed by them.
3780         * df-scan.c (df_insn_delete): Accept out-of-block debug insn.
3781         * final.c (reemit_insn_block_notes): Take current block from
3782         nonbind markers.  Declare note where it's first set.
3783         (final_scan_insn): Handle begin stmt notes.  Emit is_stmt according to
3784         begin stmt markers if enabled.
3785         (notice_source_line): Handle nonbind markers.  Fail if their
3786         location is unknown or that of builtins.
3787         (rest_of_handle_final): Convert begin stmt markers to notes if
3788         var-tracking didn't run.
3789         (rest_of_clean_state): Skip begin stmt markers.
3790         * gimple-pretty-print.c (dump_gimple_debug): Handle begin stmt
3791         markers.
3792         * function.c (allocate_struct_function): Set begin_stmt_markers.
3793         * function.h (struct function): Add debug_marker_count counter
3794         and debug_nonbind_markers flag.
3795         * gimple-iterator.c (gsi_remove): Adjust debug_marker_count.
3796         * gimple-low.c (lower_function_body): Adjust
3797         debug_nonbind_markers.
3798         (lower_stmt): Drop or skip gimple debug stmts.
3799         (lower_try_catch): Skip debug stmts.
3800         * gimple.c (gimple_build_debug_begin_stmt): New.
3801         (gimple_copy): Increment debug_marker_count if copying one.
3802         * gimple.h (gimple_build_debug_begin_stmt): Declare.
3803         * gimplify.c (rexpr_location): New.
3804         (rexpr_has_location): New.
3805         (warn_switch_unreachable_r): Handle gimple debug stmts.
3806         (shortcut_cond_r): Call expr_location.
3807         (find_goto): New.
3808         (find_goto_label): New.
3809         (shortcut_cond_expr): Call expr_has_location, expr_location, and
3810         find_goto_label.
3811         (gimplify_cond_expr): Call find_goto_label, expr_has_location, and
3812         expr_location.
3813         (gimplify_expr): Handle begin stmt markers.  Reject debug expr decls.
3814         * langhooks-def.h (LANG_HOOKS_EMITS_BEGIN_STMT): New.  Add to...
3815         (LANG_HOOKS_INITIALIZER): ... this.
3816         * langhooks.h (struct lang_hooks): Add emits_begin_stmt.
3817         * lra-contraints.c (inherit_reload_reg): Tolerate between-blocks
3818         debug insns.
3819         (update_ebb_live_info): Skip debug insn markers.
3820         * lra.c (debug_insn_static_data): Rename to...
3821         (debug_bind_static_data): ... this.
3822         (debug_marker_static_data): New.
3823         (lra_set_insn_recog_data): Select one of the above depending
3824         on debug insn kind.
3825         (lra_update_isn_regno_info): Don't assume debug insns have
3826         freqs.
3827         (push_insns): Skip debug insns.
3828         * lto-streamer-in.c (input_function): Drop debug stmts
3829         depending on active options.  Adjust debug_nonbind_markers.
3830         * params.def (PARAM_MAX_DEBUG_MARKER_COUNT): New.
3831         * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
3832         begin stmt marker notes.
3833         (print_insn): Likewise.
3834         * recog.c (extract_insn): Recognize rtl for debug markers.
3835         * rtl.def (DEBUG_MARKER): New.
3836         * tree-inline.c: Include params.h.
3837         (remap_gimple_stmt): Handle nonbind markers.
3838         (maybe_move_debug_stmts_to_successors): Likewise.
3839         (copy_debug_stmt): Likewise.
3840         * tree-iterator.c (append_to_statement_list_1): Append begin stmt
3841         markers regardless of no side effects.
3842         (tsi_link_before): Don't update container's side effects when adding
3843         a begin stmt marker.
3844         (tsi_link_after): Likewise.
3845         (expr_first): Skip begin stmt markers.
3846         (expr_last): Likewise.
3847         * tree-pretty-print (dump_generic_node): Handle begin stmt markers.
3848         * tree-ssa-threadedge.c (propagate_threaded_block_debug_info):
3849         Disregard nonbind markers.
3850         * tree.c (make_node_stat): Don't set side effects for begin stmt
3851         markers.
3852         (build1_stat): Likewise.
3853         * tree.def (DEBUG_BEGIN_STMT): New.
3854         * tree.h (GOTO_DESTINATION): Require a GOTO_EXPR.
3855         * var-tracking.c (delete_debug_insns): Renamed to...
3856         (delete_vta_debug_insns): ... this.
3857         (reemit_marker_as_note): New.
3858         (vt_initialize): Reemit markers.
3859         (delete_vta_debug_insns): Likewise.
3860         (vt_debug_insns_local): Reemit or delete markers.
3861         (variable_tracking_main_1): Likewise.
3862         * doc/generic.texi (DEBUG_BEGIN_STMT): Document.
3863         * doc/gimple.texi (gimple_debug_begin_stmt_p): New.
3864         (gimple_debug_nonbind_marker_p): New.
3865         (gimple_build_debug_bind): Adjust.
3866         (gimple_build_debug_begin_stmt): New.
3867         * doc/invoke.texi (max-debug-marker-count): New param.
3868         * doc/rtl.texi (debug_implicit_ptr, entry_value): New.
3869         (debug_parameter_ref, debug_marker): New.
3870         (NOTE_INSN_BEGIN_STMT): New.
3871         (DEBUG_INSN): Describe begin stmt markers.
3873         * cfgbuild.c (find_bb_boundaries): Don't purge dead edges if,
3874         without debug insns, we wouldn't, but clean up debug insns
3875         after a control flow insn nevertheless.
3877         * cfgcleanup.c (delete_unreachable_blocks): Use alternate
3878         block removal order if MAY_HAVE_DEBUG_BIND_INSNS.
3879         * cfgexpand.c (label_rtx_for_bb): Skip debug insns.
3880         * cfgrtl.c (try_redirect_by_replacing_jump): Skip debug insns.
3881         (rtl_tidy_fallthru_edge): Likewise.
3882         (rtl_verify_fallthru): Likewise.
3883         (rtl_verify_bb_layout): Likewise.
3884         (skip_insns_after_block): Likewise.
3885         (duplicate_insn_chain): Use DEBUG_BIND_INSN_P.
3886         * dwarf2out.c: Include print-rtl.h.
3887         (dwarf2out_next_real_insn): New.
3888         (dwarf2out_var_location): Call it.  Disregard begin stmt markers.
3889         Dump debug binds in asm comments.
3890         * gimple-iterator.c (gimple_find_edge_insert_loc): Skip debug stmts.
3891         * gimple-iterator.h (gsi_start_bb_nondebug): Remove; adjust
3892         callers to use gsi_start_nondebug_bb instead.
3893         (gsi_after_labels): Skip gimple debug stmts.
3894         (gsi_start_nondebug): New.
3895         * gimple-loop-interchange.c (find_deps_in_bb_for_stmt): Adjust.
3896         (proper_loop_form_for_interchange): Adjust.
3897         * gimple-low.c (gimple_seq_may_fallthru): Take last nondebug stmt.
3898         * gimple.h (gimple_seq_last_nondebug_stmt): New.
3899         * gimplify.c (last_stmt_in_scope): Skip debug stmts.
3900         (collect_fallthrough_labels): Likewise.
3901         (should_warn_for_implicit_fallthrough): Likewise.
3902         (warn_implicit_fallthrough_r): Likewise.
3903         (expand_FALLTHROUGH_r): Likewise.
3904         * graphite-isl-ast-to-gimple.c (gsi_insert_earliest): Adjust.
3905         (graphite_copy_stmts_from_block): Skip nonbind markers.
3906         * haifa-sched.c (sched_extend_bb): Skip debug insns.
3907         * ipa-icf-gimple.c (func_checker::compare_bb): Adjust.
3908         * jump.c (clean_barriers): Skip debug insns.
3909         * omp-expand.c (expand_parallel_call): Skip debug insns.
3910         (expand_task_call): Likewise.
3911         (remove_exit_barrier): Likewise.
3912         (expand_omp_taskreg): Likewise.
3913         (expand_omp_for_init_counts): Likewise.
3914         (expand_omp_for_generic): Likewise.
3915         (expand_omp_for_static_nochunk): Likewise.
3916         (expand_omp_for_static_chunk): Likewise.
3917         (expand_omp_simd): Likewise.
3918         (expand_omp_taskloop_for_outer): Likewise.
3919         (expand_omp_taskloop_for_inner): Likewise.
3920         (expand_oacc_for): Likewise.
3921         (expand_omp_sections): Likewise.
3922         (expand_omp_single): Likewise.
3923         (expand_omp_synch): Likewise.
3924         (expand_omp_atomic_load): Likewise.
3925         (expand_omp_atomic_store): Likewise.
3926         (expand_omp_atomic_fetch_op): Likewise.
3927         (expand_omp_atomic_pipeline): Likewise.
3928         (expand_omp_atomic_mutex): Likewise.
3929         (expand_omp_target): Likewise.
3930         (grid_expand_omp_for_loop): Likewise.
3931         (grid_expand_target_grid_body): Likewise.
3932         (build_omp_regions_1): Likewise.
3933         * omp-low.c (check_combined_parallel): Skip debug stmts.
3934         * postreload.c (fixup_debug_insns): Skip nonbind debug insns.
3935         * regcprop.c (find_oldest_value_reg): Ensure REGNO is not a pseudo.
3936         * sese.c (sese_trivially_empty_bb_p): Call is_gimple_debug in
3937         test.
3938         * tree-cfg.c (make_blobs_1): Skip debug stmts.
3939         (make_edges): Likewise.
3940         (cleanup_dead_labels): Likewise.
3941         (gimple_can_merge_blocks_p): Likewise.
3942         (stmt_starts_bb_p): Likewise.
3943         (gimple_block_label): Likewise.
3944         (gimple_redirect_edge_and_branch): Likewise.
3945         * tree-cfgcleanup.c (remove_forwarder_block): Rearrange skipping
3946         of debug stmts.
3947         (execute_cleanup_cfg_post_optimizing): Dump enumerated decls with
3948         TDF_SLIM.
3949         * tree-pretty-print (print_declaration): Omit initializer in slim
3950         dumps.
3951         * tree-ssa-dce.c (mark_stmt_if_obviously_necessary): Mark begin stmt
3952         markers.
3953         (eliminate_unnecessary_stmts): Stabilize block removal order.
3954         * tree-ssa-tail-merge.c (find_duplicate): Skip debug stmts.
3955         * var-tracking.c (get_first_insn): New.
3956         (vt_emit_notes): Call it.
3957         (vt_initialize): Walk any insns before the first BB.
3958         (delete_debug_insns): Likewise.
3960         * gimple.h (enum gimple_debug_subcode): Add
3961         GIMPLE_DEBUG_BEGIN_STMT.
3962         (gimple_debug_begin_stmt_p): New.
3963         (gimple_debug_nonbind_marker_p): New.
3964         * tree.h (MAY_HAVE_DEBUG_MARKER_STMTS): New.
3965         (MAY_HAVE_DEBUG_BIND_STMTS): Renamed from....
3966         (MAY_HAVE_DEBUG_STMTS): ... this.  Check both.
3967         * insn-notes.def (BEGIN_STMT): New.
3968         * rtl.h (MAY_HAVE_DEBUG_MARKER_INSNS): New.
3969         (MAY_HAVE_DEBUG_BIND_INSNS): Renamed from....
3970         (MAY_HAVE_DEBUG_INSNS): ... this.  Check both.
3971         (NOTE_MARKER_LOCATION, NOTE_MARKER_P): New.
3972         (DEBUG_BIND_INSN_P, DEBUG_MARKER_INSN_P): New.
3973         (INSN_DEBUG_MARKER_KIND): New.
3974         (GEN_RTX_DEBUG_MARKER_BEGIN_STMT_PAT): New.
3975         (INSN_VAR_LOCATION): Check for VAR_LOCATION.
3976         (INSN_VAR_LOCATION_PTR): New.
3977         * cfgexpand.c (expand_debug_locations): Handle debug bind insns
3978         only.
3979         (expand_gimple_basic_block): Likewise.  Emit debug temps for TER
3980         deps only if debug bind insns are enabled.
3981         (pass_expand::execute): Avoid deep TER and expand
3982         debug locations for debug bind insns only.
3983         * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Narrow
3984         debug stmts special handling down to debug bind stmts.
3985         * combine.c (try_combine): Narrow debug insns special handling
3986         down to debug bind insns.
3987         * cse.c (delete_trivially_dead_insns): Handle debug bindings.
3988         Narrow debug insns preexisting special handling down to debug
3989         bind insns.
3990         * dce.c (rest_of_handle_ud_dce): Narrow debug insns special
3991         handling down to debug bind insns.
3992         * function.c (instantiate_virtual_regs): Skip debug markers,
3993         adjust handling of debug binds.
3994         * gimple-ssa-backprop.c (backprop::prepare_change): Try debug
3995         temp insertion iff MAY_HAVE_DEBUG_BIND_STMTS.
3996         * haifa-sched.c (schedule_insn): Narrow special handling of debug
3997         insns to debug bind insns.
3998         * ipa-param-manipulation.c (ipa_modify_call_arguments): Narrow
3999         special handling of debug stmts to debug bind stmts.
4000         * ipa-split.c (split_function): Likewise.
4001         * ira.c (combine_and_move_insns): Adjust debug bind insns only.
4002         * loop-unroll.c (apply_opt_in_copies): Adjust tests on bind
4003         debug insns.
4004         * reg-stack.c (convert_regs_1): Use DEBUG_BIND_INSN_P.
4005         * regrename.c (build_def_use): Likewise.
4006         * regcprop.c (copyprop_hardreg_forward_1): Likewise.
4007         (pass_cprop_hardreg): Narrow special casing of debug insns to
4008         debug bind insns.
4009         * regstat.c (regstat_init_n_sets_and_refs): Likewise.
4010         * reload1.c (reload): Likewise.
4011         * sese.c (sese_insert_phis_for_liveouts): Narrow special
4012         casing of debug stmts to debug bind stmts.
4013         * shrink-wrap.c (move_insn_for_shrink_wrap): Likewise.
4014         * ssa-iterators.h (num_imm_uses): Likewise.
4015         * tree-cfg.c (gimple_merge_blocks): Narrow special casing of
4016         debug stmts to debug bind stmts.
4017         * tree-inline.c (tree_function_versioning): Narrow special casing
4018         of debug stmts to debug bind stmts.
4019         * tree-loop-distribution.c (generate_loops_for_partition):
4020         Narrow special casing of debug stmts to debug bind stmts.
4021         * tree-sra.c (analyze_access_subtree): Narrow special casing
4022         of debug stmts to debug bind stmts.
4023         * tree-ssa-dce.c (remove_dead_stmt): Narrow special casing of debug
4024         stmts to debug bind stmts.
4025         * tree-ssa-loop-ivopt.c (remove_unused_ivs): Narrow special
4026         casing of debug stmts to debug bind stmts.
4027         * tree-ssa-reassoc.c (reassoc_remove_stmt): Likewise.
4028         * tree-ssa-tail-merge.c (tail_merge_optimize): Narrow special
4029         casing of debug stmts to debug bind stmts.
4030         * tree-ssa-threadedge.c (propagate_threaded_block_debug_info):
4031         Likewise.
4032         * tree-ssa.c (flush_pending_stmts): Narrow special casing of
4033         debug stmts to debug bind stmts.
4034         (gimple_replace_ssa_lhs): Likewise.
4035         (insert_debug_temp_for_var_def): Likewise.
4036         (insert_debug_temps_for_defs): Likewise.
4037         (reset_debug_uses): Likewise.
4038         * tree-ssanames.c (release_ssa_name_fn): Likewise.
4039         * tree-vect-loop-manip.c (adjust_debug_stmts_now): Likewise.
4040         (adjust_debug_stmts): Likewise.
4041         (adjust_phi_and_debug_stmts): Likewise.
4042         (vect_do_peeling): Likewise.
4043         * tree-vect-loop.c (vect_transform_loop): Likewise.
4044         * valtrack.c (propagate_for_debug): Use BIND_DEBUG_INSN_P.
4045         * var-tracking.c (adjust_mems): Narrow special casing of debug
4046         insns to debug bind insns.
4047         (dv_onepart_p, dataflow_set_clar_at_call, use_type): Likewise.
4048         (compute_bb_dataflow, vt_find_locations): Likewise.
4049         (vt_expand_loc, emit_notes_for_changes): Likewise.
4050         (vt_init_cfa_base): Likewise.
4051         (vt_emit_notes): Likewise.
4052         (vt_initialize): Likewise.
4053         (vt_finalize): Likewise.
4055         * emit-rtl.c (next_nondebug_insn, prev_nondebug_insn): Reorder.
4056         (next_nonnote_nondebug_insn, prev_nonnote_nondebug_insn): Reorder.
4057         (next_nonnote_nondebug_insn_bb): New.
4058         (prev_nonnote_nondebug_insn_bb): New.
4059         (prev_nonnote_insn_bb, next_nonnote_insn_bb): Remove.
4060         * rtl.h (prev_nonnote_insn_bb, next_nonnote_insn_bb): Remove decls.
4061         (prev_nonnote_nondebug_insn_bb): Declare.
4062         (next_nonnote_nondebug_insn_bb): Declare.
4063         * cfgbuild.c (find_bb_boundaries): Adjust to skip debug insns.
4064         * cfgrtl.c (get_last_bb_insn): Likewise.
4065         * lra.c (push_insns): Likewise.
4067 2017-12-11  David Malcolm  <dmalcolm@redhat.com>
4069         PR c/82050
4070         * selftest-run-tests.c (selftest::run_tests): Move start/finish code
4071         to...
4072         * selftest.c (selftest::test_runner::test_runner): New ctor.
4073         (selftest::test_runner::~test_runner): New dtor.
4074         * selftest.h (class selftest::test_runner): New class.
4076 2017-12-11  Carl Love  <cel@us.ibm.com>
4078         * config/rs6000/altivec.h (vec_extract_fp32_from_shorth,
4079         vec_extract_fp32_from_shortl]): Add #defines.
4080         * config/rs6000/rs6000-builtin.def (VSLDOI_2DI): Add macro expansion.
4081         * config/rs6000/rs6000-c.c (ALTIVEC_BUILTIN_VEC_UNPACKH,
4082         ALTIVEC_BUILTIN_VEC_UNPACKL, ALTIVEC_BUILTIN_VEC_AND,
4083         ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SRL,
4084         ALTIVEC_BUILTIN_VEC_SRO, ALTIVEC_BUILTIN_VEC_SLD,
4085         ALTIVEC_BUILTIN_VEC_SLL): Add expansions.
4086         * doc/extend.texi: Add documentation for the added builtins.
4088 2017-12-11  Alexandre Oliva  <aoliva@redhat.com>
4090         PR rtl-optimization/80693
4091         PR rtl-optimization/81019
4092         PR rtl-optimization/81020
4093         * combine.c (distribute_notes): Reset any REG_UNUSED REGs that
4094         are not mentioned in i3.  Place the REG_UNUSED note on i2,
4095         possibly modified to REG_DEAD, if it did not originate in i3.
4097 2017-12-11  Jakub Jelinek  <jakub@redhat.com>
4099         * recog.c (store_data_bypass_p_1): New function.
4100         (store_data_bypass_p): Handle USE in a PARALLEL like CLOBBER.  Use
4101         store_data_bypass_p_1 to avoid code duplication.  Formatting fixes.
4103 2017-12-11  Segher Boessenkool  <segher@kernel.crashing.org>
4105         PR rtl-optimization/83361
4106         * ifcvt.c (if_convert): Call fixup_partitions.
4108 2017-12-11  Will Schmidt  <will_schmidt@vnet.ibm.com>
4110         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add support for
4111         early folding of splat_u{8,16,32}.
4113 2017-12-11  Jakub Jelinek  <jakub@redhat.com>
4115         * config/aarch64/aarch64.c (aarch64_print_operand): Don't start
4116         output_operand_lossage first argument with capital letter.
4117         (aarch64_override_options): Don't start error and sorry first argument
4118         with capital letter.
4120 2017-12-11  Andi Kleen  <ak@linux.intel.com>
4122         PR gcov-profile/83355
4123         * auto-profile.c (string_table::get_index_by_decl): Don't
4124         recurse when abstract origin points to itself.
4126 2017-12-11  Bin Cheng  <bin.cheng@arm.com>
4128         PR tree-optimization/83320
4129         * gimple-loop-interchange.cc (free_data_refs_with_aux): Use delete.
4130         (prune_datarefs_not_in_loop): Ditto.
4132 2017-12-10  Gerald Pfeifer  <gerald@pfeifer.com>
4134         * doc/install.texi (Specific): Tweak link to mkssoftware.com.
4136 2017-12-10  Jakub Jelinek  <jakub@redhat.com>
4138         PR tree-optimization/83337
4139         * gimple-loop-interchange.cc (compute_access_stride): Handle
4140         bitfield DRs properly.
4142 2017-12-09  Jakub Jelinek  <jakub@redhat.com>
4144         PR tree-optimization/83338
4145         * tree-vect-stmts.c (vectorizable_operation): Handle POINTER_DIFF_EXPR
4146         vectorization as MINUS_EXPR with a subsequent VIEW_CONVERT_EXPR from
4147         vector of unsigned integers to vector of signed integers.
4149 2017-12-08  Vladimir Makarov  <vmakarov@redhat.com>
4151         PR rtl-optimization/83317
4152         * lra-constraints.c (process_address_1): Add insn code check.
4154 2017-12-08  Michael Matz  <matz@suse.de>
4156         Fix PR tree-optimization/83323
4157         * gimple-loop-jam (unroll_jam_possible_p): Correct test for
4158         head-controlled loops and loop BBs.
4159         * common.opt (funroll-and-jam): Remove, instead ...
4160         (floop-unroll-and-jam): ... reuse this option.
4161         * opts.c (default_options_table): Use OPT_floop_unroll_and_jam.
4162         * doc/invoke.texi (-funroll-and-jam): Move docu to ...
4163         (-floop-unroll-and-jam): ... this option.
4165 2017-12-08  Jakub Jelinek  <jakub@redhat.com>
4167         * ipa-polymorphic-call.c (noncall_stmt_may_be_vtbl_ptr_store): Fix
4168         a comment typo, get_base_ref_and_offset -> get_ref_base_and_extent.
4169         * ipa-prop.c (stmt_may_be_vtbl_ptr_store): Likewise.
4171 2017-12-08  Richard Biener  <rguenther@suse.de>
4173         PR middle-end/81782
4174         * tree-ssa-uninit.c (warn_uninitialized_vars): Properly
4175         handle accesses outside of zero-sized vars.
4177 2017-12-08  Martin Jambor  <mjambor@suse.cz>
4179         PR tree-optimization/83141
4180         * tree-sra.c (contains_vce_or_bfcref_p): Move up in the file, also
4181         test for MEM_REFs implicitely changing types with padding.  Remove
4182         inline keyword.
4183         (build_accesses_from_assign): Added contains_vce_or_bfcref_p checks.
4185 2017-12-08  Jakub Jelinek  <jakub@redhat.com>
4187         * config/arc/arc.c (arc_attribute_table): Add exclusions to
4188         the comment.
4189         * config/avr/avr.c (avr_attribute_table): Likewise.
4190         * config/msp430/msp430.c (msp430_attribute_table): Likewise.
4191         * config/rl78/rl78.c (rl78_attribute_table): Likewise.
4192         * config/nds32/nds32.c (nds32_attribute_table): Likewise.
4193         * config/darwin.h (SUBTARGET_ATTRIBUTE_TABLE): Initialize new member
4194         of struct attribute_spec.
4195         * config/i386/cygming.h (SUBTARGET_ATTRIBUTE_TABLE): Likewise.
4197 2017-12-08  Ulrich Weigand  <uweigand@de.ibm.com>
4199         PR target/82960
4200         * config/spu/spu.c (pad_bb): Only check INSN_CODE when INSN_P is true.
4202 2017-12-08  Jan Hubicka  <hubicka@ucw.cz>
4204         PR middle-end/83609
4205         * profile-count.c (profile_count::from_gcov_type): Move from
4206         profile-count.h; handle overflow.
4207         * profile-count.h (profile_count::from_gcov_type): Move offline.
4209 2017-12-08  Segher Boessenkool  <segher@kernel.crashing.org>
4211         PR rtl-optimization/83304
4212         * combine.c (move_deaths): If we do not know where a register died,
4213         search for it.
4215 2017-12-08  Richard Biener  <rguenther@suse.de>
4217         * gimple-loop-interchange.cc (tree_loop_interchange::interchange):
4218         Provide -fopt-info-loop feedback when we interchange in a nest.
4220 2017-06-08  Richard Earnshaw  <rearnsha@arm.com>
4222         * config/arm/driver-arm.c (arm_cpu_table): Use fp-capable product names
4223         for armv6 ARM CPU IDs.
4225 2017-06-08  Richard Earnshaw  <rearnsha@arm.com>
4227         * common/config/arm/arm-common.c: Include <algorithm>.
4228         (INCLUDE_VECTOR): Define.
4229         (compare_opt_names): New function.
4230         (arm_rewrite_selected_arch): Only strip out extensions that can be
4231         expressed through -mfpu.  Sort the remaining extensions
4232         alphabetically.
4234 2017-06-08  Richard Earnshaw  <rearnsha@arm.com>
4236         * config/arm/arm.h (arm_asm_auto_mfpu): Declare.
4237         (ASM_CPU_SPEC_FUNCTIONS): Add new rule asm_auto_mfpu.
4238         (ASM_CPU_SPEC): Use it if -mfpu is set to auto.
4239         * common/config/arm/arm-common.c (arm_asm_auto_mfpu): New function.
4241 2017-06-08  Tristan Gingold  <gindold@adacore.com>
4243         PR ada/81470
4244         * dwarf2out.c (dwarf2out_do_cfi_startproc): Only emit
4245         .cfi_personality or .cfi_lsda if the eh data format is dwarf2.
4247 2017-12-08  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
4249         * config/sol2.h (SOLARIS_ATTRIBUTE_TABLE): Initialize new member
4250         of struct attribute_spec.
4252 2017-12-08  Julia Koval  <julia.koval@intel.com>
4254         * config/i386/avx512vnniintrin.h (_mm512_dpwssds_epi32,
4255         _mm512_mask_dpwssds_epi32, _mm512_maskz_dpwssds_epi32): New intrinsics.
4256         * config/i386/avx512vnnivlintrin.h (_mm256_dpwssds_epi32,
4257         _mm256_mask_dpwssds_epi32, _mm256_maskz_dpwssds_epi32,
4258         _mm_dpwssds_epi32, _mm_mask_dpwssds_epi32,
4259         _mm_maskz_dpwssds_epi32): Ditto.
4261 2017-12-08  Richard Biener  <rguenther@suse.de>
4263         PR tree-optimization/81303
4264         * tree-vect-stmts.c (vect_is_simple_cond): For invariant
4265         conditions try to create a comparison vector type matching
4266         the data vector type.
4267         (vectorizable_condition): Adjust.
4268         * tree-vect-patterns.c (vect_recog_mask_conversion_pattern):
4269         Leave invariant conditions alone in case we can vectorize those.
4271 2017-12-08  Julia Koval  <julia.koval@intel.com>
4273         * config/i386/avx512vnniintrin.h (_mm512_dpwssd_epi32,
4274         _mm512_mask_dpwssd_epi32, _mm512_maskz_dpwssd_epi32): New intrinsics.
4275         * config/i386/avx512vnnivlintrin.h (_mm256_dpwssd_epi32,
4276         _mm256_mask_dpwssd_epi32, _mm256_maskz_dpwssd_epi32, _mm_dpwssd_epi32,
4277         _mm_mask_dpwssd_epi32, _mm_maskz_dpwssd_epi32): Ditto.
4279 2017-12-08  Julia Koval  <julia.koval@intel.com>
4281         * config/i386/avx512vnniintrin.h (_mm512_dpbusds_epi32,
4282         _mm512_mask_dpbusds_epi32, _mm512_maskz_dpbusds_epi32): New.
4283         * config/i386/avx512vnnivlintrin.h (_mm256_dpbusds_epi32,
4284         _mm256_mask_dpbusds_epi32, _mm256_maskz_dpbusds_epi32,
4285         _mm_dpbusds_epi32, _mm_mask_dpbusds_epi32,
4286         _mm_maskz_dpbusds_epi32): New intrinsics.
4288 2017-12-07  Sandra Loosemore  <sandra@codesourcery.com>
4290         * config/nios2/nios2.md (ld<bhw_uns>io): Add splitter for memory
4291         operand.
4292         (ld<bh>io_signed): Likewise.
4293         (st<bhw>io): Likewise.
4294         * config/nios2/predicates.md (ldstio_memory_operand): Allow
4295         SMALL_INT12 constant integer operand.
4297 2017-12-07  Sandra Loosemore  <sandra@codesourcery.com>
4299         * config/nios2/nios2.c (nios2_symbolic_constant_allowed):
4300         Rename to...
4301         (nios2_large_constant_allowed): ...this.  Adjust uses.
4302         (nios2_plus_symbolic_constant_p): Rename to...
4303         (nios2_plus_large_constant_p): ...this.  Adjust uses.
4304         (nios2_legitimate_address_p): Correct CONST_INT handling.
4305         (nios2_symbolic_memory_operand_p): Rename to...
4306         (nios2_large_constant_memory_operand_p): ...this.  Adjust uses.
4307         (nios2_large_constant_p): Check for large constant integers too.
4308         (nios2_split_large_constant): Handle constant integers.
4309         (nios2_split_symbolic_memory_operand): Rename to...
4310         (nios2_split_large_constant_memory_operand): ...this.  Adjust uses.
4311         (nios2_legitimize_constant_address): Handle constant integers.
4312         (r0rel_constant_p): Handle small constant integers.
4313         (nios2_print_operand_address): Handle r0-relative integer addresses.
4314         * config/nios2/nios2-protos.h: Adjust for renamed functions.
4315         * config/nios2/nios2.md: Adjust for renamed functions.
4317 2017-12-07  Andrew Waterman  <andrew@sifive.com>
4319         * config/riscv/riscv.c (TARGET_ASM_SELECT_SECTION): New define.
4320         (TARGET_HAVE_SRODATA_SECTION): New define.
4321         (riscv_select_section): New function.
4323 2017-12-08  Joseph Myers  <joseph@codesourcery.com>
4324             Alexander Monakov  <amonakov@ispras.ru>
4325             Jakub Jelinek  <jakub@redhat.com>
4327         PR target/81906
4328         * config/i386/i386.c (ix86_expand_rint): Handle flag_rounding_math.
4330 2017-12-07  Richard Sandiford  <richard.sandiford@linaro.org>
4332         * config/aarch64/aarch64.c (aarch64_print_address_internal): Return
4333         a bool success value.  Don't call output_operand_lossage here.
4334         (aarch64_print_ldpstp_address): Return a bool success value.
4335         (aarch64_print_operand_address): Call output_addr_const if
4336         aarch64_print_address_internal fails.
4337         (aarch64_print_operand): Don't assert that the mode is 16 bytes for
4338         'y'; call output_operand_lossage instead.  Call output_operand_lossage
4339         if aarch64_print_ldpstp_address fails.
4341 2017-12-07  Richard Sandiford  <richard.sandiford@linaro.org>
4343         * tree-vector-builder.h
4344         (tree_vector_builder::binary_encoded_nelts): Declare.
4345         * tree-vector-builder.c
4346         (tree_vector_builder::binary_encoded_nelts): New function.
4347         * fold-const.c (negate_expr_p): Likewise.
4348         (operand_equal_p, fold_checksum_tree): Likewise.
4349         * tree-loop-distribution.c (const_with_all_bytes_same): Likewise.
4350         * tree.c (integer_zerop, integer_onep, integer_all_onesp, real_zerop)
4351         (real_onep, real_minus_onep, add_expr, initializer_zerop): Likewise.
4352         (uniform_vector_p): Likewise.
4353         * varasm.c (const_hash_1, compare_constant): Likewise.
4354         * tree-ssa-ccp.c: Include tree-vector-builder.h.
4355         (valid_lattice_transition): Operate directly on the VECTOR_CST
4356         encoding.
4357         * ipa-icf.c: Include tree-vector-builder.h.
4358         (sem_variable::equals): Operate directly on the VECTOR_CST encoding.
4359         * print-tree.c (print_node): Print encoding of VECTOR_CSTs.
4361 2017-12-07  Richard Sandiford  <richard.sandiford@linaro.org>
4363         * tree.c (build_vector): Delete.
4364         * tree.h (build_vector): Make static and move into the self-testing
4365         block.
4367 2017-12-07  Richard Sandiford  <richard.sandiford@linaro.org>
4369         * vector-builder.h (vector_builder::derived): New const overload.
4370         (vector_builder::elt): New function.
4371         * tree-vector-builder.h (tree_vector_builder::type): New function.
4372         (tree_vector_builder::apply_step): Declare.
4373         * tree-vector-builder.c (tree_vector_builder::apply_step): New
4374         function.
4375         * gimple-fold.h (tree_vector_builder): Declare.
4376         (gimple_build_vector): Take a tree_vector_builder instead of a
4377         type and vector of elements.
4378         * gimple-fold.c (gimple_build_vector): Likewise.
4379         * tree-vect-loop.c (get_initial_def_for_reduction): Update call
4380         accordingly.
4381         (get_initial_defs_for_reduction): Likewise.
4382         (vectorizable_induction): Likewise.
4384 2017-12-07  Richard Sandiford  <richard.sandiford@linaro.org>
4386         * tree-vector-builder.h
4387         (tree_vector_builder::new_binary_operation): Declare.
4388         * tree-vector-builder.c
4389         (tree_vector_builder::new_binary_operation): New function.
4390         * fold-const.c (fold_relational_const): Use it.
4391         (const_binop): Likewise.  Check that both input vectors have
4392         the same number of elements, thus excluding things like WIDEN_SUM.
4393         Check whether it is possible to operate directly on the encodings
4394         of stepped inputs.
4396 2017-12-07  Richard Sandiford  <richard.sandiford@linaro.org>
4398         * fold-const.c (fold_negate_expr_1): Use tree_vector_builder and
4399         new_unary_operation, operating only on the encoded elements.
4400         (const_unop): Likewise.
4401         (exact_inverse): Likewise.
4402         (distributes_over_addition_p): New function.
4403         (const_binop): Use tree_vector_builder and new_unary_operation
4404         for combinations of VECTOR_CST and INTEGER_CST.  Operate only
4405         on the encoded elements unless the encoding is strided and the
4406         operation does not distribute over addition.
4407         (fold_convert_const):  Use tree_vector_builder and
4408         new_unary_operation.  Operate only on the encoded elements
4409         for truncating integer conversions, or for non-stepped encodings.
4411 2017-12-07  Richard Sandiford  <richard.sandiford@linaro.org>
4413         * config/sparc/sparc.c: Include tree-vector-builder.h.
4414         (sparc_fold_builtin): Use tree_vector_builder instead of build_vector.
4415         * expmed.c: Include tree-vector-builder.h.
4416         (make_tree): Use tree_vector_builder instead of build_vector.
4417         * fold-const.c: Include tree-vector-builder.h.
4418         (const_binop): Use tree_vector_builder instead of build_vector.
4419         (const_unop): Likewise.
4420         (native_interpret_vector): Likewise.
4421         (fold_vec_perm): Likewise.
4422         (fold_ternary_loc): Likewise.
4423         * gimple-fold.c: Include tree-vector-builder.h.
4424         (gimple_fold_stmt_to_constant_1): Use tree_vector_builder instead
4425         of build_vector.
4426         * tree-ssa-forwprop.c: Include tree-vector-builder.h.
4427         (simplify_vector_constructor): Use tree_vector_builder instead
4428         of build_vector.
4429         * tree-vect-generic.c: Include tree-vector-builder.h.
4430         (add_rshift): Use tree_vector_builder instead of build_vector.
4431         (expand_vector_divmod): Likewise.
4432         (optimize_vector_constructor): Likewise.
4433         * tree-vect-loop.c: Include tree-vector-builder.h.
4434         (vect_create_epilog_for_reduction): Use tree_vector_builder instead
4435         of build_vector.  Explicitly use a stepped encoding for
4436         { 1, 2, 3, ... }.
4437         * tree-vect-slp.c: Include tree-vector-builder.h.
4438         (vect_get_constant_vectors): Use tree_vector_builder instead
4439         of build_vector.
4440         (vect_transform_slp_perm_load): Likewise.
4441         (vect_schedule_slp_instance): Likewise.
4442         * tree-vect-stmts.c: Include tree-vector-builder.h.
4443         (vectorizable_bswap): Use tree_vector_builder instead of build_vector.
4444         (vect_gen_perm_mask_any): Likewise.
4445         (vectorizable_call): Likewise.  Explicitly use a stepped encoding.
4446         * tree.c: (build_vector_from_ctor): Use tree_vector_builder instead
4447         of build_vector.
4448         (build_vector_from_val): Likewise.  Explicitly use a duplicate
4449         encoding.
4451 2017-12-07  Richard Sandiford  <richard.sandiford@arm.com>
4453         * doc/generic.texi (VECTOR_CST): Describe new representation of
4454         vector constants.
4455         * vector-builder.h: New file.
4456         * tree-vector-builder.h: Likewise.
4457         * tree-vector-builder.c: Likewise.
4458         * Makefile.in (OBJS): Add tree-vector-builder.o.
4459         * tree.def (VECTOR_CST): Update comment to refer to generic.texi.
4460         * tree-core.h (tree_base): Add a vector_cst field to the u union.
4461         (tree_vector): Change the number of elements to
4462         vector_cst_encoded_nelts.
4463         * tree.h (VECTOR_CST_NELTS): Redefine using TYPE_VECTOR_SUBPARTS.
4464         (VECTOR_CST_ELTS): Delete.
4465         (VECTOR_CST_ELT): Redefine using vector_cst_elt.
4466         (VECTOR_CST_LOG2_NPATTERNS, VECTOR_CST_NPATTERNS): New macros.
4467         (VECTOR_CST_NELTS_PER_PATTERN, VECTOR_CST_DUPLICATE_P): Likewise.
4468         (VECTOR_CST_STEPPED_P, VECTOR_CST_ENCODED_ELTS): Likewise.
4469         (VECTOR_CST_ENCODED_ELT): Likewise.
4470         (vector_cst_encoded_nelts): New function.
4471         (make_vector): Take the values of VECTOR_CST_LOG2_NPATTERNS and
4472         VECTOR_CST_NELTS_PER_PATTERN as arguments.
4473         (vector_cst_int_elt, vector_cst_elt): Declare.
4474         * tree.c: Include tree-vector-builder.h.
4475         (tree_code_size): Abort if passed VECTOR_CST.
4476         (tree_size): Update for new VECTOR_CST layout.
4477         (make_vector): Take the values of VECTOR_CST_LOG2_NPATTERNS and
4478         VECTOR_CST_NELTS_PER_PATTERN as arguments.
4479         (build_vector): Use tree_vector_builder.
4480         (vector_cst_int_elt, vector_cst_elt): New functions.
4481         (drop_tree_overflow): For VECTOR_CST, drop the TREE_OVERFLOW from the
4482         encoded elements and then create the vector in the canonical form.
4483         (check_vector_cst, check_vector_cst_duplicate, check_vector_cst_fill)
4484         (check_vector_cst_stepped, test_vector_cst_patterns): New functions.
4485         (tree_c_tests): Call test_vector_cst_patterns.
4486         * lto-streamer-out.c (DFS::DFS_write_tree_body): Handle the new
4487         VECTOR_CST fields.
4488         (hash_tree): Likewise.
4489         * tree-streamer-out.c (write_ts_vector_tree_pointers): Likewise.
4490         (streamer_write_tree_header): Likewise.
4491         * tree-streamer-in.c (lto_input_ts_vector_tree_pointers): Likewise.
4492         (streamer_alloc_tree): Likewise.  Update call to make_vector.
4493         * fold-const.c (fold_ternary_loc): Avoid using VECTOR_CST_ELTS.
4495 2017-12-07  Richard Sandiford  <richard.sandiford@linaro.org>
4497         * selftest.h (ASSERT_TRUE_AT, ASSERT_FALSE_AT, ASSERT_EQ_AT)
4498         (ASSERT_NE, ASSERT_PRED1): Add underscores to local variable names
4499         * selftest-rtl.h (ASSERT_RTX_EQ, ASSERT_RTX_PTR_EQ): Likewise.
4501 2017-12-07  Bin Cheng  <bin.cheng@arm.com>
4502             Richard Biener  <rguenther@suse.de>
4504         PR tree-optimization/81303
4505         * Makefile.in (gimple-loop-interchange.o): New object file.
4506         * common.opt (floop-interchange): Reuse the option from graphite.
4507         * doc/invoke.texi (-floop-interchange): Ditto.  New document for
4508         -floop-interchange and mention it for -O3.
4509         * opts.c (default_options_table): Enable -floop-interchange at -O3.
4510         * gimple-loop-interchange.cc: New file.
4511         * params.def (PARAM_LOOP_INTERCHANGE_MAX_NUM_STMTS): New parameter.
4512         (PARAM_LOOP_INTERCHANGE_STRIDE_RATIO): New parameter.
4513         * passes.def (pass_linterchange): New pass.
4514         * timevar.def (TV_LINTERCHANGE): New time var.
4515         * tree-pass.h (make_pass_linterchange): New declaration.
4516         * tree-ssa-loop-ivcanon.c (create_canonical_iv): Change to external
4517         interchange.  Record IV before/after increment in new parameters.
4518         * tree-ssa-loop-ivopts.h (create_canonical_iv): New declaration.
4519         * tree-vect-loop.c (vect_is_simple_reduction): Factor out reduction
4520         path check into...
4521         (check_reduction_path): ...New function here.
4522         * tree-vectorizer.h (check_reduction_path): New declaration.
4524 2017-12-07  Vladimir Makarov  <vmakarov@redhat.com>
4526         PR target/83252
4527         PR rtl-optimization/80818
4528         * lra.c (add_regs_to_insn_regno_info): Make a hard reg in CLOBBER
4529         always early clobbered.
4530         * lra-lives.c (process_bb_lives): Check input hard regs for early
4531         clobbered non-operand hard reg.
4533 2017-12-07  Jakub Jelinek  <jakub@redhat.com>
4535         PR middle-end/83164
4536         * tree-cfg.c (verify_gimple_assign_binary): Don't require
4537         types_compatible_p, just that TYPE_MODE is the same.
4539 2017-12-07  Martin Sebor  <msebor@redhat.com>
4541         PR c/81544
4542         * attribs.c (empty_attribute_table): Initialize new member of
4543         struct attribute_spec.
4544         (decl_attributes): Add argument.  Handle mutually exclusive
4545         combinations of attributes.
4546         (selftests::test_attribute_exclusions): New function.
4547         (selftests::attribute_c_tests): Ditto.
4548         * attribs.h (decl_attributes): Add default argument.
4549         * selftest.h (attribute_c_tests): Declare.
4550         * selftest-run-tests.c (selftest::run_tests): Call attribute_c_tests.
4551         * tree-core.h (attribute_spec::exclusions, exclude): New type and
4552         member.
4553         * doc/extend.texi (Common Function Attributes): Update const and pure.
4554         * config/alpha/alpha.c (vms_attribute_table): Initialize new member
4555         of struct attribute_spec.
4556         * config/arc/arc.c (arc_attribute_table): Same.
4557         * config/arm/arm.c (arm_attribute_table): Same.
4558         * config/avr/avr.c ( avr_attribute_table): Same.
4559         * config/bfin/bfin.c (bfin_attribute_table): Same.
4560         * config/cr16/cr16.c (cr16_attribute_table): Same.
4561         * config/epiphany/epiphany.c (epiphany_attribute_table): Same.
4562         * config/h8300/h8300.c (h8300_attribute_table): Same.
4563         * config/i386/i386.c (ix86_attribute_table): Same.
4564         * config/ia64/ia64.c (ia64_attribute_table): Same.
4565         * config/m32c/m32c.c (m32c_attribute_table): Same.
4566         * config/m32r/m32r.c (m32r_attribute_table): Same.
4567         * config/m68k/m68k.c (m68k_attribute_table): Same.
4568         * config/mcore/mcore.c (mcore_attribute_table): Same.
4569         * config/microblaze/microblaze.c (microblaze_attribute_table): Same.
4570         * config/mips/mips.c (mips_attribute_table): Same.
4571         * config/msp430/msp430.c (msp430_attribute_table): Same.
4572         * config/nds32/nds32.c (nds32_attribute_table): Same.
4573         * config/nvptx/nvptx.c (nvptx_attribute_table): Same.
4574         * config/powerpcspe/powerpcspe.c (rs6000_attribute_table): Same.
4575         * config/rl78/rl78.c (rl78__attribute_table): Same.
4576         * config/rs6000/rs6000.c (rs6000_attribute_table): Same.
4577         * onfig/rx/rx.c (rx_attribute_table): Same.
4578         * config/s390/s390.c (s390_handle_vectorbool_attribute): Same.
4579         * config/sh/sh.c (sh_attribute_table): Same.
4580         * config/sparc/sparc.c (sparc_attribute_table): Same.
4581         * config/spu/spu.c (spu_attribute_table): Same.
4582         * config/stormy16/stormy16.c (xstormy16_attribute_table): Same.
4583         * config/v850/v850.c (v850_attribute_table): Same.
4584         * config/visium/visium.c (visium_attribute_table): Same.
4586 2017-12-07  Tamar Christina  <tamar.christina@arm.com>
4588         PR target/82641
4589         * config/arm/arm.c (INCLUDE_STRING): Define.
4590         (arm_last_printed_arch_string, arm_last_printed_fpu_string): New.
4591         (arm_declare_function_name): Conservatively emit .arch, .arch_extensions
4592         and .fpu.
4594 2017-12-07  Michael Matz  <matz@suse.de>
4596         Add unroll and jam pass
4598         * gimple-loop-jam.c: New file.
4599         * Makefile.in (OBJS): Add gimple-loop-jam.o.
4600         * common.opt (funroll-and-jam): New option.
4601         * opts.c (default_options_table): Add unroll-and-jam at -O3.
4602         * params.def (PARAM_UNROLL_JAM_MIN_PERCENT): New param.
4603         (PARAM_UNROLL_JAM_MAX_UNROLL): Ditto.
4604         * passes.def: Add pass_loop_jam.
4605         * timevar.def (TV_LOOP_JAM): Add.
4606         * tree-pass.h (make_pass_loop_jam): Declare.
4607         * cfgloop.c (flow_loop_tree_node_add): Add AT argument.
4608         * cfgloop.h (flow_loop_tree_node_add): Adjust declaration.
4609         * cfgloopmanip.c (duplicate_loop): Add AT argument, adjust call
4610         to flow_loop_tree_node_add.
4611         (duplicate_subloops, copy_loops_to): Append to sibling list.
4612         * cfgloopmanip.h: (duplicate_loop): Adjust declaration.
4613         * doc/invoke.texi (-funroll-and-jam): Document new option.
4614         (unroll-jam-min-percent, unroll-jam-max-unroll): Document new params.
4616 2017-12-07  Richard Biener  <rguenther@suse.de>
4618         PR tree-optimization/83296
4619         PR tree-optimization/67769
4620         * tree-ssa-phiopt.c (conditional_replacement): Do not reset
4621         flow sensitive info in an unrelated BB.
4622         (value_replacement): Use reset_flow_sensitive_info.
4623         (minmax_replacement): Reset flow sensitive info on the def
4624         we move.  Do not reset flow sensitive info in the whole BB
4625         we move the stmt to.
4626         (abs_replacement): Likewise.
4628 2017-12-07  Segher Boessenkool  <segher@kernel.crashing.org>
4630         PR target/43871
4631         * config/rs6000/rs6000.c (rs6000_option_override_internal): Set
4632         rs6000_cpu to the given -mcpu=, or to the default processor.
4634 2017-12-07  Segher Boessenkool  <segher@kernel.crashing.org>
4636         * config/rs6000/rs6000.h (rs6000_cpu_attr): Delete.
4637         * config/rs6000/rs6000.c (rs6000_variable_issue_1): Use rs6000_tune
4638         instead of rs6000_cpu_attr.
4639         (rs6000_adjust_cost): Ditto.
4640         (is_microcoded_insn): Ditto.
4641         (rs6000_adjust_priority): Ditto.
4642         (rs6000_issue_rate): Ditto.
4643         (rs6000_use_sched_lookahead): Ditto.
4644         (rs6000_use_sched_lookahead_guard): Ditto.
4645         (rs6000_sched_reorder): Ditto.
4646         (force_new_group): Ditto.
4647         * config/rs6000/rs6000.md (cpu attribute): Ditto.
4648         (group_ending_nop): Ditto.
4650 2017-12-07  Segher Boessenkool  <segher@kernel.crashing.org>
4652         * config/rs6000/rs6000.opt (rs6000_tune): New variable.
4653         * config/rs6000/rs6000.c (rs6000_option_override_internal): Also set
4654         rs6000_tune.  Use rs6000_tune instead of rs6000_cpu where appropriate.
4655         (rs6000_loop_align): Use rs6000_tune instead of rs6000_cpu where
4656         appropriate.
4657         (rs6000_reassociation_width): Ditto.
4658         (rs6000_emit_epilogue): Ditto.
4659         (rs6000_adjust_cost): Ditto.
4660         (is_microcoded_insn): Ditto.
4661         (is_cracked_insn): Ditto.
4662         (rs6000_adjust_priority): Ditto.
4663         (rs6000_sched_reorder): Ditto.
4664         (rs6000_sched_reorder2): Ditto.
4665         (insn_must_be_first_in_group): Ditto.
4666         (insn_must_be_last_in_group): Ditto.
4667         (rs6000_register_move_cost): Ditto.
4668         * config/rs6000/rs6000.h (rs6000_cpu_attr): Use rs6000_tune instead of
4669         rs6000_cpu.
4671 2017-12-07  Julia Koval  <julia.koval@intel.com>
4673         * config.gcc: Add vaesintrin.h.
4674         * config/i386/i386-builtin-types.def (V64QI_FTYPE_V64QI_V64QI):
4675         New type.
4676         * config/i386/i386-builtin.def (__builtin_ia32_vaesdec_v16qi,
4677         __builtin_ia32_vaesdec_v32qi, __builtin_ia32_vaesdec_v64qi):
4678         New builtins.
4679         * config/i386/i386.c (ix86_expand_args_builtin): Handle new type.
4680         * config/i386/immintrin.h: Include vaesintrin.h.
4681         * config/i386/sse.md (vaesdec_<mode>): New pattern.
4682         * config/i386/vaesintrin.h (_mm256_aesdec_epi128, _mm512_aesdec_epi128,
4683         _mm_aesdec_epi128): New intrinsics.
4685 2017-12-06  David Malcolm  <dmalcolm@redhat.com>
4687         * Makefile.in (C_COMMON_OBJS): Add c-family/c-spellcheck.o.
4688         * spellcheck-tree.c (find_closest_macro_cpp_cb): Move to
4689         c-family/c-spellcheck.cc.
4690         (best_macro_match::best_macro_match): Likewise.
4691         * spellcheck-tree.h
4692         (struct edit_distance_traits<cpp_hashnode *>): Move to
4693         c-family/c-spellcheck.h.
4694         (class best_macro_match): Likewise.
4696 2017-12-06  Jakub Jelinek  <jakub@redhat.com>
4698         PR tree-optimization/83293
4699         * gimple-ssa-strength-reduction.c (insert_initializers): Use
4700         GSI_NEW_STMT instead of GSI_SAME_STMT in gsi_insert_after that
4701         might insert into empty bb.
4703         PR sanitizer/81281
4704         * match.pd ((T)(P + A) - (T)P -> (T) A): Split into separate
4705         simplify for plus with :c added, and pointer_plus without that.
4706         ((T)P - (T)(P + A) -> -(T) A): Likewise.  If type is integral
4707         with undefined overflow and the conversion is not widening,
4708         perform negation in utype and only convert to type afterwards.
4709         ((T)(P + A) - (T)(P + B) -> (T)A - (T)B): Split into separate
4710         simplify for plus with :c added, and pointer_plus without that.
4711         If type is integral with undefined overflow and the conversion is
4712         not widening, perform minus in utype and only convert to type
4713         afterwards.  Move the last pointer_diff_expr simplify into the
4714         two outermost ifs.
4716 2017-12-06  Martin Sebor  <msebor@redhat.com>
4718         PR tree-optimization/82646
4719         * builtins.c (maybe_emit_chk_warning): Use size as the bound for
4720         strncpy, not maxlen.
4722 2017-12-06  Martin Sebor  <msebor@redhat.com>
4724         * doc/invoke.texi (-Wstringop-truncation): Mention attribute
4725         nonstring.
4727         PR tree-optimization/83075
4728         * tree-ssa-strlen.c (handle_builtin_stxncpy): Avoid assuming
4729         strncat/strncpy don't change length of source string.
4731 2017-12-06  Eric Botcazou  <ebotcazou@adacore.com>
4733         Revert
4734         2017-11-29  Martin Aberg  <maberg@gaisler.com>
4736         * config/sparc/sparc.md (divdf3_fix): Add NOP and adjust length
4737         to prevent b2bst errata sequence.
4738         (sqrtdf2_fix): Likewise.
4740 2017-12-06  Jakub Jelinek  <jakub@redhat.com>
4742         PR tree-optimization/81945
4743         * cfgloop.h (FOR_EACH_LOOP_FN): Use FN instead of hardcoding fn.
4744         * tree-cfg.c (move_sese_region_to_fn): If any of the loops moved
4745         to dest_cfun has orig_loop_num set, either remap it to the new
4746         loop number if the loop got moved too, or clear it.
4748 2017-12-05  Steve Ellcey  <sellcey@cavium.com>
4750         * config/aarch64/thunderx2-t99.md (thunderx2t99_branch): Add trap
4751         to reservation.
4752         (thunderx2t99_nothing): New insn reservation.
4753         (thunderx2t99_mrs): New insn reservation.
4754         (thunderx2t99_multiple): New insn reservation.
4755         (thunderx2t99_alu_basi): Add bfx to reservation.
4756         (thunderx2t99_fp_cmp): Add fccmps and fccmpd to reservation.
4758 2017-12-05  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
4760         PR target/82248
4761         * config/arm/arm.md (probe_stack) : Use the 'o' constraint.
4763 2017-12-05  Bin Cheng  <bin.cheng@arm.com>
4765         * tree-ssa-dce.c (simple_dce_from_worklist): Move and rename from
4766         tree-ssa-pre.c::remove_dead_inserted_code.
4767         * tree-ssa-dce.h: New file.
4768         * tree-ssa-pre.c (tree-ssa-dce.h): Include new header file.
4769         (remove_dead_inserted_code): Move and rename to function
4770         tree-ssa-dce.c::simple_dce_from_worklist.
4771         (pass_pre::execute): Update use.
4773 2017-12-05  Richard Biener  <rguenther@suse.de>
4775         PR tree-optimization/83277
4776         * graphite-isl-ast-to-gimple.c (should_copy_to_new_region): Make sure
4777         to code-gen liveout vars.
4779 2017-12-05  Richard Sandiford  <richard.sandiford@linaro.org>
4781         * config/aarch64/aarch64-simd.md (aarch64_simd_bsldi_internal)
4782         (aarch64_simd_bsldi_alt): Check REG_P before GP_REGNUM_P.
4783         (aarch64_cm<optab>di, aarch64_cmtstdi): Add leading "&&" to
4784         split condition.
4786 2017-12-05  Max Filippov  <jcmvbkbc@gmail.com>
4788         * config/xtensa/xtensa.c (xtensa_asan_shadow_offset): New
4789         function.
4790         (TARGET_ASAN_SHADOW_OFFSET): New macro definition.
4791         * config/xtensa/xtensa.h (FRAME_GROWS_DOWNWARD): Set to 1 if
4792         ASAN is enabled.
4794 2017-12-05  Richard Biener  <rguenther@suse.de>
4796         * timevar.def (TV_TREE_RECIP, TV_TREE_SINCOS, TV_TREE_WIDEN_MUL): Add.
4797         * tree-ssa-math-opts.c (pass_data_cse_reciprocal): Use TV_TREE_RECIP.
4798         (pass_data_cse_sincos): Use TV_TREE_SINCOS.
4799         (pass_data_optimize_widening_mul): Use TV_TREE_WIDEN_MUL.
4801 2017-12-05  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
4803         * dbxout.c (dbxout_block): Grow buf to 30 bytes.
4805 2017-12-05  Martin Liska  <mliska@suse.cz>
4806             Jakub Jelinek  <jakub@redhat.com>
4808         * doc/invoke.texi: Document the options.
4809         * flag-types.h (enum sanitize_code): Add
4810         SANITIZE_POINTER_COMPARE and SANITIZE_POINTER_SUBTRACT.
4811         * ipa-inline.c (sanitize_attrs_match_for_inline_p): Add handling
4812         of SANITIZE_POINTER_COMPARE and SANITIZE_POINTER_SUBTRACT.
4813         * opts.c: Define new sanitizer options.
4814         * sanitizer.def (BUILT_IN_ASAN_POINTER_COMPARE): Likewise.
4815         (BUILT_IN_ASAN_POINTER_SUBTRACT): Likewise.
4817 2017-12-05  Julia Koval  <julia.koval@intel.com>
4819         * common/config/i386/i386-common.c (OPTION_MASK_ISA_AVX512VNNI_SET,
4820         OPTION_MASK_ISA_AVX512VNNI_UNSET): New.
4821         (ix86_handle_option): Handle -mavx512vnni.
4822         * config/i386/cpuid.h (bit_AVX512VNNI): New bit.
4823         * config/i386/driver-i386.c (host_detect_local_cpu): Handle new bit.
4824         * config/i386/i386-c (__AVX512VNNI__): New.
4825         * config/i386/i386.c (ix86_target_string): Handle new option.
4826         (ix86_valid_target_attribute_inner_p): Handle new option.
4827         * config/i386/i386.h (TARGET_AVX512VNNI, TARGET_AVX512VNNI_P): New.
4828         * config/i386/i386.opt (mavx512vnni): New option.
4830 2017-12-01  Jan Hubicka  <hubicka@ucw.cz>
4832         PR target/81616
4833         * athlon.md: Disable for generic.
4834         * haswell.md: Enable for generic.
4835         * i386.c (ix86_sched_init_global): Add core hooks for generic.
4836         * x86-tune-sched.c (ix86_issue_rate): Increase issue rate for generic
4837         to 4.
4838         (ix86_adjust_cost): Move generic to haswell path.
4840 2017-12-04  Eric Botcazou  <ebotcazou@adacore.com>
4842         * config/sparc/sparc.c (sparc_do_work_around_errata): Use mem_ref
4843         instead of MEM_P in a couple more places.  Fix formatting issues.
4845 2017-12-04  Jim Wilson  <jimw@sifive.com>
4847         * config/riscv/riscv.c (riscv_for_each_saved_reg): Use GP_REG_LAST
4848         instead of GP_REG_LAST-1.
4849         (riscv_adjust_libcall_cfi_prologue): Likewise.
4850         (riscv_adjust_libcall_cri_epilogue): Likewise.
4851         * config/riscv/riscv.h (CALL_USED_REGISTERS): Change a6 to t6 in
4852         comment.
4854 2017-12-04  Luis Machado  <luis.machado@linaro.org>
4856         * ipa-pure-const.c (check_decl): Add missing newline.
4857         (state_from_flags): Likewise.
4859 2017-12-04  Jeff Law  <law@redhat.com>
4861         PR tree-optimizatin/78496
4862         * gimple-ssa-evrp-analyze.h
4863         (evrp_range_analyzer::get_vr_values): Simplify.
4864         * gimple-ssa-evrp-analyze.c: Corresponding changes.
4865         * tree-ssa-dom.c: Include alloc-pool.h, tree-vrp.h, vr-values.h
4866         and gimple-ssa-evrp-analyze.h.
4867         (dom_opt_dom_walker class): Add evrp_range_analyzer member.
4868         (simplify_stmt_for_jump_threading): Copy a blob of code from
4869         tree-vrp.c to use ranges to simplify statements.
4870         (dom_opt_dom_walker::before_dom_children): Call
4871         evrp_range_analyzer::{enter,record_ranges_from_stmt} methods.
4872         (dom_opt_dom_walker::after_dom_children): Similarly for
4873         evrp_range_analyzer::leave.
4874         (dom_opt_dom_walker::optimize_stmt): Use EVRP ranges to optimize
4875         conditionals.
4877         * gimple-ssa-evrp-analyze.c
4878         (evrp_range_analyzer::extract_range_from_stmt):  Always use
4879         vr_values::update_value_range so preexisting range info is
4880         medged with new range info, even if the new range is VR_VARYING.
4882 2017-12-04  Segher Boessenkool  <segher@kernel.crashing.org>
4884         * combine.c: Adjust comment.
4885         (use_crosses_set_p): Delete.
4886         (can_combine_p): Use modified_between_p instead of use_crosses_set_p.
4887         (try_combine): Ditto.
4889 2017-12-04  Richard Biener  <rguenther@suse.de>
4891         PR tree-optimization/83255
4892         * graphite-isl-ast-to-gimple.c (translate_isl_ast_node_for):
4893         Re-add zero-iteration check.
4895 2017-12-04  Segher Boessenkool  <segher@kernel.crashing.org>
4897         PR rtl-optimization/83245
4898         * lra.c (collect_non_operand_hard_regs): Treat clobbers of non-operand
4899         hard registers as earlyclobber, also if not in an asm.
4901 2017-12-04  Segher Boessenkool  <segher@kernel.crashing.org>
4903         PR bootstrap/83265
4904         Revert
4905         2017-12-01  Segher Boessenkool  <segher@kernel.crashing.org>
4907         PR target/43871
4908         * config/rs6000/rs6000.c (rs6000_option_override_internal): Set
4909         rs6000_cpu based on cpu_index, not tune_index.
4911 2017-12-04  Richard Biener  <rguenther@suse.de>
4913         PR tree-optimization/83238
4914         * graphite-scop-detection.c (scop_detection::merge_sese): Make
4915         code match comment, rejecting invalid SESE regions.
4917 2017-12-03  John David Anglin  <danglin@gcc.gnu.org>
4919         * config/pa/pa.c (pa_legitimate_address_p): For scaled indexing,
4920         require base operand is a REG_POINTER prior to reload on targets
4921         with non-equivalent space registers.
4923 2017-12-01  Jan Hubicka  <hubicka@ucw.cz>
4925         * ipa-cp.c (ipcp_lattice<valtype>::print): Update dumping.
4926         (update_specialized_profile): Fix updating of counts.
4927         (perhaps_add_new_callers): Likewise.
4929 2017-12-01  Jan Hubicka  <hubicka@ucw.cz>
4931         PR target/81616
4932         * x86-tune.def: Remove obsolette FIXMEs.
4933         (X86_TUNE_PARTIAL_FLAG_REG_STALL): Disable for generic
4934         (X86_TUNE_FUSE_CMP_AND_BRANCH_32, X86_TUNE_FUSE_CMP_AND_BRANCH_64,
4935         X86_TUNE_FUSE_CMP_AND_BRANCH_SOFLAGS, X86_TUNE_FUSE_ALU_AND_BRANCH):
4936         Enable for generic.
4937         (X86_TUNE_PAD_RETURNS): Disable for generic.
4939 2017-12-02  Jakub Jelinek  <jakub@redhat.com>
4941         PR tree-optimization/83170
4942         PR tree-optimization/83241
4943         * gimple-ssa-store-merging.c
4944         (imm_store_chain_info::try_coalesce_bswap): Update vuse field from
4945         gimple_vuse (ins_stmt) in case it has changed.
4946         (imm_store_chain_info::output_merged_store): Likewise.
4948         * tree-chkp.c (chkp_compute_bounds_for_assignment): Handle
4949         POINTER_DIFF_EXPR.
4951         PR c++/81212
4952         * tree-cfg.c (pass_warn_function_return::execute): Handle
4953         __builtin_ubsan_handle_missing_return like __builtin_unreachable
4954         with BUILTINS_LOCATION.
4956         PR target/78643
4957         PR target/80583
4958         * expr.c (get_inner_reference): If DECL_MODE of a non-bitfield
4959         is BLKmode for vector field with vector raw mode, use TYPE_MODE
4960         instead of DECL_MODE.
4962         * config/i386/i386-protos.h (standard_sse_constant_opcode): Change
4963         last argument to rtx pointer.
4964         * config/i386/i386.c (standard_sse_constant_opcode): Replace X argument
4965         with OPERANDS.  For AVX+ 128-bit VEX encoded instructions over 256-bit
4966         or 512-bit.  If setting EXT_REX_SSE_REG_P, use EVEX encoded insn
4967         depending on the chosen ISAs.
4968         * config/i386/i386.md (*movxi_internal_avx512f, *movoi_internal_avx,
4969         *movti_internal, *movdi_internal, *movsi_internal, *movtf_internal,
4970         *movdf_internal, *movsf_internal): Adjust standard_sse_constant_opcode
4971         callers.
4972         * config/i386/sse.md (mov<mode>_internal): Likewise.
4973         * config/i386/mmx.md (*mov<mode>_internal): Likewise.
4975 2017-12-01  Segher Boessenkool  <segher@kernel.crashing.org>
4977         * doc/invoke.texi (-dp): Say that instruction cost is printed as well.
4979 2017-12-01  Segher Boessenkool  <segher@kernel.crashing.org>
4981         * config/rs6000/rs6000-protos.h (emit_fusion_addis): Remove last two
4982         parameters from prototype.
4983         * config/rs6000/rs6000.c (emit_fusion_addis): Remove last two
4984         parameters.  Don't print a comment.
4985         (emit_fusion_gpr_load): Adjust.
4986         (emit_fusion_load_store): Adjust.
4987         * config/rs6000/rs6000.md (*fusion_p9_<mode>_constant): Adjust.
4988         * config/rs6000/vsx.md (two peepholes): Print the "vector load fusion"
4989         comment on the second line.
4991 2017-12-01  Segher Boessenkool  <segher@kernel.crashing.org>
4993         PR target/43871
4994         * config/rs6000/rs6000.c (rs6000_option_override_internal): Set
4995         rs6000_cpu based on cpu_index, not tune_index.
4997 2017-12-01  Segher Boessenkool  <segher@kernel.crashing.org>
4999         * final.c (output_asm_name): Print insn_cost.  Shorten output.  Print
5000         which_alternative instead of which_alternative + 1.
5001         (output_asm_insn): Print an extra tab if the template is short.
5003 2017-12-01  Jim Wilson  <jimw@sifive.com>
5005         * common.opt (use_gnu_debug_info_extensions): Delete DWARF_DEBUG from
5006         comment.
5007         * config/vx-common.h (DWARF_DEBUGGING_INFO): Delete undef.
5008         * doc/tm.texi.in (PREFERRED_DEBUGGING_TYPE): Delete DWARF_DEBUG
5009         reference.
5010         * doc/tm.texi: Regenerate.
5012 2017-12-01  Michael Meissner  <meissner@linux.vnet.ibm.com>
5014         PR target/81959
5015         * config/rs6000/rs6000.c (rs6000_address_for_fpconvert): Check for
5016         whether we can allocate pseudos before trying to fix an address.
5017         * config/rs6000/rs6000.md (float_<mode>si2_hw): Make sure the
5018         memory address is indexed or indirect.
5019         (floatuns_<mode>si2_hw2): Likewise.
5021 2017-12-01  Jason Merrill  <jason@redhat.com>
5023         * Makefile.in (TAGS): Add c-family/*.cc.
5025 2017-12-01  Wilco Dijkstra  <wdijkstr@arm.com>
5027         * config/aarch64/aarch64.md (call_insn): Use %c rather than %a.
5028         (call_value_insn): Likewise.
5029         (sibcall_insn): Likewise.
5030         (sibcall_value_insn): Likewise.
5031         (movsi_aarch64): Likewise.
5032         (movdi_aarch64): Likewise.
5033         (add_losym_): Likewise.
5034         (ldr_got_small_): Likewise.
5035         (ldr_got_small_sidi): Likewise.
5036         (ldr_got_small_28k_): Likewise.
5037         (ldr_got_small_28k_sidi): Likewise.
5038         * config/aarch64/aarch64.c (aarch64_print_address_internal):
5039         Move output_addr_const to symbolic case. Add error check.
5041 2017-12-01  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
5043         * config/s390/predicates.md (plus16_Q_operand): New predicate.
5044         * config/s390/s390.md: Disable MVC merging peephole if it would
5045         disable operand forwarding.
5046         (new peephole2): Split MVCs if it would turn them into up to 2
5047         forwardable MVCs.
5049 2017-12-01  Richard Biener  <rguenther@suse.de>
5051         PR tree-optimization/83232
5052         * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Fix
5053         detection of same access. Instead of breaking the group here
5054         do not consider the duplicate.  Add comment explaining real fix.
5056 2017-12-01  Jonathan Wakely  <jwakely@redhat.com>
5058         * doc/md.texi (Insn Splitting): Fix "central flowgraph" typo.
5060 2017-12-01  Sudakshina Das  <sudi.das@arm.com>
5062         * config/arm/vfp.md (*movhf_vfp_fp16): Add conds attribute.
5064 2017-12-01  Jakub Jelinek  <jakub@redhat.com>
5066         * function.h (struct function): Remove cilk_frame_decl,
5067         is_cilk_function and calls_cilk_spawn fields.
5068         * tree-inline.h (struct copy_body_data): Remove remap_var_for_cilk
5069         field.
5070         * omp-simd-clone.c (simd_clone_clauses_extract): Don't clear
5071         cilk_elemental field.
5072         * cgraph.h (struct cgraph_simd_clone): Remove cilk_elemental field.
5073         * target.def: Adjust comment.
5074         * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
5075         Don't test cilk_elemental.
5077         PR tree-optimization/83233
5078         * gimple-ssa-store-merging.c (nop_stats, bswap_stats): Use
5079         bswap_stat name for the struct.
5081         PR c/79153
5082         * tree.h (SWITCH_BREAK_LABEL_P): Define.
5083         * gimplify.c (collect_fallthrough_labels): Handle GIMPLE_BIND
5084         starting with a GIMPLE_SWITCH and ending with GIMPLE_LABEL with
5085         SWITCH_BREAK_LABEL_P set on the label.
5086         (gimplify_switch_expr): Set SWITCH_BREAK_LABEL_P on the label
5087         added for default case if it was missing and not all cases covered.
5088         Wrap GIMPLE_SWITCH and the switch_body_seq into a GIMPLE_BIND if
5089         switch_body_seq ends with a GIMPLE_LABEL with SWITCH_BREAK_LABEL_P
5090         set on the label.
5091         * tree-chrec.c (evolution_function_is_univariate_p): Add return true;
5092         to avoid -Wimplicit-fallthrough warning.
5093         * config/i386/i386.c (ix86_expand_special_args_builtin): Add
5094         FALLTHRU comment to avoid -Wimplicit-fallthrough warning.
5096         PR tree-optimization/83221
5097         * tree-ssa-reassoc.c (sort_by_operand_rank): Shift bb_rank
5098         down by 16.
5099         (init_reassoc): Formatting fix.
5101         PR sanitizer/81275
5102         * tree-cfg.c (group_case_labels_stmt): Don't optimize away
5103         C++ FE implicitly added __builtin_unreachable () until -Wreturn-type
5104         is diagnosed.
5106         PR sanitizer/83219
5107         * tree-cfg.c: Include asan.h.
5108         (gimple_seq_unreachable_p): Return false for -fsanitize=unreachable.
5110 2017-12-01  Sergey Shalnov  <Sergey.Shalnov@intel.com>
5112         * config/i386/i386.md: Fix AVX512 register width in AVX512 instruction.
5114 2017-12-01  Sergey Shalnov  <Sergey.Shalnov@intel.com>
5116         * config/i386/i386.c (standard_sse_constant_opcode): Fix wrong form for
5117         vpcmpeqd instruction.
5119 2017-12-01  Sergey Shalnov  <Sergey.Shalnov@intel.com>
5121         * config/i386/i386.c (standard_sse_constant_opcode): Fix
5122         registers type for 128bit mode.
5124 2017-11-30  Jakub Jelinek  <jakub@redhat.com>
5126         * spellcheck-tree.c (test_find_closest_identifier): Use ; instead
5127         of ;;.
5128         * gengtype-state.c (read_state_pair): Likewise.
5129         * gimple-fold.c (gimple_fold_builtin_string_compare): Likewise.
5130         * sel-sched-dump.c (dump_insn_rtx_1): Likewise.
5131         * ipa-cp.c (intersect_aggregates_with_edge): Likewise.
5132         * ifcvt.c (noce_try_store_flag_constants): Likewise.
5133         * tree-ssa-ccp.c (ccp_finalize): Likewise.
5134         * omp-grid.c (grid_process_kernel_body_copy): Likewise.
5135         * builtins.c (fold_builtin_3): Likewise.
5136         * graphite-scop-detection.c
5137         (scop_detection::stmt_has_simple_data_refs_p): Likewise.
5138         * hsa-gen.c (hsa_function_representation::hsa_function_representation):
5139         Likewise.
5141 2017-12-01  Maxim Ostapenko  <m.ostapenko@samsung.com>
5143         PR sanitizer/81697
5144         * asan.c (asan_protect_global): Add new ignore_decl_rtl_set_p
5145         parameter. Return true if ignore_decl_rtl_set_p is true and other
5146         conditions are satisfied.
5147         * asan.h (asan_protect_global): Add new parameter.
5148         * varasm.c (categorize_decl_for_section): Pass true as second parameter
5149         to asan_protect_global calls.
5151 2017-11-30  Jim Wilson  <jimw@sifive.com>
5153         * doc/invoke.texi (RISC-V Options): Delete nonexistent -mmemcpy and
5154         -mno-memcpy options.  For -mplt, -mfdiv, -mdiv, -msave-restore, and
5155         -mstrict-align, add info on default value.  Delete redundant lines for
5156         -mabi.  Add missing -mexplicit-relocs docs.
5158 2017-11-30  Claudiu Zissulescu  <claziss@synopsys.com>
5160         * config/arc/arc.md (trap): New pattern.
5162 2017-11-30  Claudiu Zissulescu  <claziss@synopsys.com>
5164         * config/arc/arc.c (hwloop_optimize): Prevent the last ZOL
5165         instruction to end into a delay slot.
5166         * config/arc/arc.md (cond_delay_insn): Check if the instruction
5167         can be placed into a delay slot against reg_note.
5169 2017-11-30  Claudiu Zissulescu  <claziss@synopsys.com>
5171         * config/arc/arc.c (hwloop_optimize): Update hw-loop's end/start
5172         labels number of usages.
5174 2017-11-30  Claudiu Zissulescu  <claziss@synopsys.com>
5176         * config/arc/arc.c (arc_cannot_substitue_mem_equiv_p): New
5177         function.
5178         (TARGET_CANNOT_SUBSTITUTE_MEM_EQUIV_P): Define.
5180 2017-11-30  Jakub Jelinek  <jakub@redhat.com>
5182         PR target/83210
5183         * internal-fn.c (expand_mul_overflow): Optimize unsigned
5184         multiplication by power of 2 constant into two shifts + comparison.
5186 2017-11-30  Jan Hubicka  <hubicka@ucw.cz>
5188         PR target/81616
5189         * config/i386/x86-tune-costs.h (generic_cost): Revise for modern CPUs.
5191 2017-11-30  Richard Biener  <rguenther@suse.de>
5193         PR tree-optimization/83202
5194         * tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Add
5195         allow_peel argument and guard peeling.
5196         (canonicalize_loop_induction_variables): Likewise.
5197         (canonicalize_induction_variables): Pass false.
5198         (tree_unroll_loops_completely_1): Pass unroll_outer to disallow
5199         peeling from cunrolli.
5201 2017-11-29  Segher Boessenkool  <segher@kernel.crashing.org>
5203         * combine.c (try_combine): Print a message to dump file whenever
5204         I0, I1, or I2 cannot be combined into I3.
5206 2017-11-29  Segher Boessenkool  <segher@kernel.crashing.org>
5208         PR rtl-optimization/83156
5209         PR rtl-optimization/82621
5210         * combine.c (try_combine): Don't split an I2 if one of the dests is
5211         set again before I3.  Allow unused dests.
5213 2017-11-29  Segher Boessenkool  <segher@kernel.crashing.org>
5215         * config/rs6000/rs6000.md (*add<mode>3_carry_in_internal2): New.
5217 2017-11-29  Vladimir Makarov  <vmakarov@redhat.com>
5219         PR rtl-optimization/80818
5220         * lra.c (collect_non_operand_hard_regs): New arg insn.  Pass it
5221         recursively.  Use insn code for clobber.
5222         (lra_set_insn_recog_data): Pass the new arg to
5223         collect_non_operand_hard_regs.
5224         (add_regs_to_insn_regno_info): Pass insn instead of uid.  Use insn
5225         code for clobber.
5226         (lra_update_insn_regno_info): Pass insn to
5227         add_regs_to_insn_regno_info.
5229 2017-11-29  Jim Wilson  <jimw@sifive.com>
5230             Andrew Waterman  <andrew@sifive.com>
5232         * config/riscv/riscv.c (SINGLE_SHIFT_COST): New.
5233         (riscv_rtx_costs): Case ZERO_EXTRACT, match new pattern, and return
5234         SINGLE_SHIFT_COST.  Case LT and ZERO_EXTEND, likewise.  Case ASHIFT,
5235         use SINGLE_SHIFT_COST.
5236         * config/riscv/riscv.md (lshrsi3_zero_extend_1): New.
5237         (lshrsi3_zero_extend_2, lshrsi3_zero_extend_3): New.
5239 2017-11-29  Julia Koval  <julia.koval@intel.com>
5241         * config/i386/avx512vbmi2intrin.h (_mm512_shldv_epi16,
5242         _mm512_mask_shldv_epi16, _mm512_maskz_shldv_epi16, _mm512_shldv_epi32,
5243         _mm512_mask_shldv_epi32, _mm512_maskz_shldv_epi32, _mm512_shldv_epi64,
5244         _mm512_mask_shldv_epi64, _mm512_maskz_shldv_epi64): New intrinsics.
5245         * config/i386/avx512vbmi2vlintrin.h (_mm256_shldv_epi16,
5246         _mm256_mask_shldv_epi16, _mm256_maskz_shldv_epi16, _mm256_shldv_epi32,
5247         _mm256_mask_shldv_epi32, _mm256_maskz_shldv_epi32, _mm256_shldv_epi64,
5248         _mm256_mask_shldv_epi64, _mm256_maskz_shldv_epi64, _mm_shldv_epi16,
5249         _mm_mask_shldv_epi16, _mm_maskz_shldv_epi16, _mm_shldv_epi32,
5250         _mm_mask_shldv_epi32, _mm_maskz_shldv_epi32, _mm_shldv_epi64,
5251         _mm_mask_shldv_epi64, _mm_maskz_shldv_epi64): Ditto.
5252         * config/i386/i386-builtin.def (__builtin_ia32_vpshldv_v32hi,
5253         __builtin_ia32_vpshldv_v32hi_mask, __builtin_ia32_vpshldv_v32hi_maskz,
5254         __builtin_ia32_vpshldv_v16hi, __builtin_ia32_vpshldv_v16hi_mask,
5255         __builtin_ia32_vpshldv_v16hi_maskz, __builtin_ia32_vpshldv_v8hi,
5256         __builtin_ia32_vpshldv_v8hi_mask, __builtin_ia32_vpshldv_v8hi_maskz,
5257         __builtin_ia32_vpshldv_v16si, __builtin_ia32_vpshldv_v16si_mask,
5258         __builtin_ia32_vpshldv_v16si_maskz, __builtin_ia32_vpshldv_v8si,
5259         __builtin_ia32_vpshldv_v8si_mask, __builtin_ia32_vpshldv_v8si_maskz,
5260         __builtin_ia32_vpshldv_v4si, __builtin_ia32_vpshldv_v4si_mask,
5261         __builtin_ia32_vpshldv_v4si_maskz, __builtin_ia32_vpshldv_v8di,
5262         __builtin_ia32_vpshldv_v8di_mask, __builtin_ia32_vpshldv_v8di_maskz,
5263         __builtin_ia32_vpshldv_v4di, __builtin_ia32_vpshldv_v4di_mask,
5264         __builtin_ia32_vpshldv_v4di_maskz, __builtin_ia32_vpshldv_v2di,
5265         __builtin_ia32_vpshldv_v2di_mask,
5266         __builtin_ia32_vpshldv_v2di_maskz): New builtins.
5267         * config/i386/sse.md (vpshldv_<mode>, vpshldv_<mode>_mask,
5268         vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): New patterns.
5270 2017-11-29  Julia Koval  <julia.koval@intel.com>
5272         * config/i386/avx512vbmi2intrin.h (_mm512_shrdv_epi16,
5273         _mm512_mask_shrdv_epi16, _mm512_maskz_shrdv_epi16, _mm512_shrdv_epi32,
5274         _mm512_mask_shrdv_epi32, _mm512_maskz_shrdv_epi32, _mm512_shrdv_epi64,
5275         _mm512_mask_shrdv_epi64, _mm512_maskz_shrdv_epi64): New intrinsics.
5276         * config/i386/avx512vbmi2vlintrin.h (_mm256_shrdv_epi16,
5277         _mm256_mask_shrdv_epi16, _mm256_maskz_shrdv_epi16, _mm256_shrdv_epi32,
5278         _mm256_mask_shrdv_epi32, _mm256_maskz_shrdv_epi32, _mm256_shrdv_epi64,
5279         _mm256_mask_shrdv_epi64, _mm256_maskz_shrdv_epi64, _mm_shrdv_epi16,
5280         _mm_mask_shrdv_epi16, _mm_maskz_shrdv_epi16, _mm_shrdv_epi32,
5281         _mm_mask_shrdv_epi32, _mm_maskz_shrdv_epi32, _mm_shrdv_epi64,
5282         _mm_mask_shrdv_epi64, _mm_maskz_shrdv_epi64): Ditto.
5283         * config/i386/i386-builtin-types.def (V32HI_FTYPE_V32HI_V32HI_V32HI,
5284         V32HI_FTYPE_V32HI_V32HI_V32HI_INT, V16HI_FTYPE_V16HI_V16HI_V16HI_INT,
5285         V8HI_FTYPE_V8HI_V8HI_V8HI_INT, V8SI_FTYPE_V8SI_V8SI_V8SI_INT,
5286         V4SI_FTYPE_V4SI_V4SI_V4SI_INT, V8DI_FTYPE_V8DI_V8DI_V8DI,
5287         V8DI_FTYPE_V8DI_V8DI_V8DI_INT, V4DI_FTYPE_V4DI_V4DI_V4DI_INT,
5288         V16SI_FTYPE_V16SI_V16SI_V16SI, V16SI_FTYPE_V16SI_V16SI_V16SI_INT,
5289         V2DI_FTYPE_V2DI_V2DI_V2DI_INT): New types.
5290         * config/i386/i386.c (ix86_expand_args_builtin): Handle new types.
5291         * config/i386/sse.md (vpshrdv_<mode>, vpshrdv_<mode>_mask,
5292         vpshrdv_<mode>_maskz, vpshrdv_<mode>_maskz_1): New pattern.
5294 2017-11-29  Daniel Cederman  <cederman@gaisler.com>
5296         * config/sparc/sparc.c (sparc_do_work_around_errata): Treat the
5297         movsi_pic_gotdata_op instruction as a load for the UT699 errata
5298         workaround.
5300 2017-11-29  Martin Aberg  <maberg@gaisler.com>
5302         * config/sparc/sparc.md (divdf3_fix): Add NOP and adjust length
5303         to prevent b2bst errata sequence.
5304         (sqrtdf2_fix): Likewise.
5306 2017-11-29  Daniel Cederman  <cederman@gaisler.com>
5308         * config/sparc/sparc.c (fpop_reg_depend_p): New function.
5309         (div_sqrt_insn_p): New function.
5310         (sparc_do_work_around_errata): Insert NOP instructions to
5311         prevent sequences that could trigger the TN-0013 errata for
5312         certain LEON3 processors.
5313         (pass_work_around_errata::gate): Also test sparc_fix_lost_divsqrt.
5314         (sparc_option_override): Set sparc_fix_lost_divsqrt appropriately.
5315         * config/sparc/sparc.md (fix_lost_divsqrt): New attribute.
5316         (in_branch_delay): Prevent div and sqrt in delay slot if
5317         fix_lost_divsqrt.
5318         * config/sparc/sparc.opt (sparc_fix_lost_divsqrt): New variable.
5320 2017-11-29  Daniel Cederman  <cederman@gaisler.com>
5322         * config/sparc/sparc.c (atomic_insn_p): New function.
5323         (sparc_do_work_around_errata): Insert NOP instructions to
5324         prevent sequences that could trigger the TN-0010 errata for
5325         UT700.
5326         * config/sparc/sync.md (atomic_compare_and_swap_leon3_1): Make
5327         instruction referable in atomic_insns_p.
5329 2017-11-29  Daniel Cederman  <cederman@gaisler.com>
5331         * config/sparc/sync.md (swapsi): 16-byte align if sparc_fix_gr712rc.
5332         (atomic_compare_and_swap_leon3_1): Likewise.
5333         (ldstub): Likewise.
5335 2017-11-29  Daniel Cederman  <cederman@gaisler.com>
5337         * config/sparc/sparc.c (fpop_insn_p): New function.
5338         (sparc_do_work_around_errata): Insert NOP instructions to
5339         prevent sequences that could trigger the TN-0012 errata for
5340         GR712RC.
5341         (pass_work_around_errata::gate): Also test sparc_fix_gr712rc.
5342         * config/sparc/sparc.md (fix_gr712rc): New attribute.
5343         (in_branch_annul_delay): Prevent floating-point instructions
5344         in delay slot of annulled integer branch.
5346 2017-11-29  Richard Biener  <rguenther@suse.de>
5348         PR tree-optimization/83202
5349         * tree-vect-slp.c (scalar_stmts_set_t): New typedef.
5350         (bst_fail): Use it.
5351         (vect_analyze_slp_cost_1): Add visited set, do not account SLP
5352         nodes vectorized to the same stmts multiple times.
5353         (vect_analyze_slp_cost): Allocate a visited set and pass it down.
5354         (vect_analyze_slp_instance): Adjust.
5355         (scalar_stmts_to_slp_tree_map_t): New typedef.
5356         (vect_schedule_slp_instance): Add a map recording the SLP node
5357         representing the vectorized stmts for a set of scalar stmts.
5358         Avoid code-generating redundancies.
5359         (vect_schedule_slp): Allocate map and pass it down.
5361 2017-11-29  Nathan Sidwell  <nathan@acm.org>
5363         PR c++/83187
5364         * tree.c (build_complex_type): Fix canonicalization.  Only fill in
5365         type if it is new.
5367 2017-11-29  Wilco Dijkstra  <wdijkstr@arm.com>
5369         * config/aarch64/aarch64.c (aarch64_print_operand): Add new
5370         cases for printing LDP/STP memory addresses.
5371         (aarch64_print_address_internal): Renamed from
5372         aarch64_print_operand_address, added parameter, add Pmode check.
5373         (aarch64_print_ldpstp_address): New function for LDP/STP addresses.
5374         (aarch64_print_operand_address): Indirect to
5375         aarch64_print_address_internal.
5376         * config/aarch64/aarch64-simd.md (store_pair_lanes): Use new
5377         'y' operand output specifier.
5379 2017-11-29  Jakub Jelinek  <jakub@redhat.com>
5381         PR middle-end/83185
5382         * tree.c (build_simple_mem_ref_loc): Handle
5383         get_addr_base_and_unit_offset returning a MEM_REF.
5385         PR middle-end/80929
5386         * rtlanal.c (seq_cost): For non-single_set insns try to use insn_cost.
5388         PR target/80819
5389         * config/i386/sse.md (vec_concatv2di): Remove * from (=Yr,0,*rm)
5390         alternative.
5392 2017-11-28  David Malcolm  <dmalcolm@redhat.com>
5394         * diagnostic-show-locus.c (layout::print_trailing_fixits): Handle
5395         m_x_offset.
5396         (layout::move_to_column): Likewise.
5398 2017-11-28  Jakub Jelinek  <jakub@redhat.com>
5400         PR sanitizer/81275
5401         * tree.c (block_may_fallthru): Return false if SWITCH_ALL_CASES_P
5402         is set on SWITCH_EXPR and !block_may_fallthru (SWITCH_BODY ()).
5404 2017-11-28  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
5405             Martin Jambor  <mjambor@suse.cz>
5407         PR ipa/82808
5408         * tree.h (expr_type_first_operand_type_p): Declare
5409         * tree.c (expr_type_first_operand_type_p): New function.
5410         * ipa-prop.h (ipa_get_type): Allow i to be out of bounds.
5411         (ipa_value_from_jfunc): Adjust declaration.
5412         * ipa-cp.c (ipa_get_jf_pass_through_result): New parameter RES_TYPE.
5413         Use it as result type for arithmetics, unless it is NULL in which case
5414         be more conservative.
5415         (ipa_value_from_jfunc): New parameter PARM_TYPE, pass it to
5416         ipa_get_jf_pass_through_result.
5417         (propagate_vals_across_pass_through): Likewise.
5418         (propagate_scalar_across_jump_function): New parameter PARM_TYPE, pass
5419         is to propagate_vals_across_pass_through.
5420         (propagate_constants_across_call): Pass PARM_TYPE to
5421         propagate_scalar_across_jump_function.
5422         (find_more_scalar_values_for_callers_subset): Pass parameter type to
5423         ipa_value_from_jfunc.
5424         (cgraph_edge_brings_all_scalars_for_node): Likewise.
5425         * ipa-fnsummary.c (evaluate_properties_for_edge): Renamed parms_info
5426         to caller_parms_info, pass parameter type to ipa_value_from_jfunc.
5427         * ipa-prop.c (try_make_edge_direct_simple_call): New parameter
5428         target_type, pass it to ipa_value_from_jfunc.
5429         (update_indirect_edges_after_inlining): Pass parameter type to
5430         try_make_edge_direct_simple_call.
5432 2017-11-28  Jeff Law  <law@redhat.com>
5434         * gimple-ssa-evrp-analyze.c
5435         (evrp_range_analyzer::record_ranges_from_phis): Only use SCEV to
5436         refine ranges if scev_initialized_p returns true.
5437         * vr-values.c (vr_values::extract_range_from_phi_node): Likewise.
5439 2017-11-28  Julia Koval  <julia.koval@intel.com>
5441         * config/i386/avx512vbmi2intrin.h (_mm512_shrdi_epi16,
5442         _mm512_mask_shrdi_epi16, _mm512_maskz_shrdi_epi16, _mm512_shrdi_epi32,
5443         _mm512_mask_shrdi_epi32, _mm512_maskz_shrdi_epi32, _mm512_shrdi_epi64,
5444         _mm512_mask_shrdi_epi64, _mm512_maskz_shrdi_epi64): New intrinsics.
5445         * config/i386/avx512vbmi2vlintrin.h (_mm256_shrdi_epi16,
5446         _mm256_mask_shrdi_epi16, _mm256_maskz_shrdi_epi16,
5447         _mm256_mask_shrdi_epi32, _mm256_maskz_shrdi_epi32, _mm256_shrdi_epi32,
5448         _mm256_mask_shrdi_epi64, _mm256_maskz_shrdi_epi64, _mm256_shrdi_epi64,
5449         _mm_mask_shrdi_epi16, _mm_maskz_shrdi_epi16, _mm_shrdi_epi16,
5450         _mm_mask_shrdi_epi32, _mm_maskz_shrdi_epi32, _mm_shrdi_epi32,
5451         _mm_mask_shrdi_epi64, _mm_maskz_shrdi_epi64, _mm_shrdi_epi64): Ditto.
5452         * config/i386/i386-builtin.def (__builtin_ia32_vpshrd_v32hi,
5453         __builtin_ia32_vpshrd_v32hi_mask, __builtin_ia32_vpshrd_v16hi,
5454         __builtin_ia32_vpshrd_v16hi_mask, __builtin_ia32_vpshrd_v8hi,
5455         __builtin_ia32_vpshrd_v8hi_mask, __builtin_ia32_vpshrd_v16si,
5456         __builtin_ia32_vpshrd_v16si_mask, __builtin_ia32_vpshrd_v8si,
5457         __builtin_ia32_vpshrd_v8si_mask, __builtin_ia32_vpshrd_v4si,
5458         __builtin_ia32_vpshrd_v4si_mask, __builtin_ia32_vpshrd_v8di,
5459         __builtin_ia32_vpshrd_v8di_mask, __builtin_ia32_vpshrd_v4di,
5460         __builtin_ia32_vpshrd_v4di_mask, __builtin_ia32_vpshrd_v2di,
5461         __builtin_ia32_vpshrd_v2di_mask): New builtins.
5462         * config/i386/sse.md (vpshrd_<mode><mask_name>): New pattern.
5464 2017-11-28  Julia Koval  <julia.koval@intel.com>
5466         * config/i386/avx512vbmi2intrin.h (_mm512_shldi_epi16,
5467         _mm512_mask_shldi_epi16, _mm512_maskz_shldi_epi16, _mm512_shldi_epi32,
5468         _mm512_mask_shldi_epi32, _mm512_maskz_shldi_epi32, _mm512_shldi_epi64,
5469         _mm512_mask_shldi_epi64, _mm512_maskz_shldi_epi64): New intrinsics.
5470         * config/i386/avx512vbmi2vlintrin.h (_mm256_shldi_epi16,
5471         _mm256_mask_shldi_epi16, _mm256_maskz_shldi_epi16,
5472         _mm256_mask_shldi_epi32, _mm256_maskz_shldi_epi32, _mm256_shldi_epi32,
5473         _mm256_mask_shldi_epi64, _mm256_maskz_shldi_epi64, _mm256_shldi_epi64,
5474         _mm_mask_shldi_epi16, _mm_maskz_shldi_epi16, _mm_shldi_epi16,
5475         _mm_mask_shldi_epi32, _mm_maskz_shldi_epi32, _mm_shldi_epi32,
5476         _mm_mask_shldi_epi64, _mm_maskz_shldi_epi64, _mm_shldi_epi64): Ditto.
5477         * config/i386/i386-builtin-types.def (V32HI_FTYPE_V32HI_V32HI_INT,
5478         V32HI_FTYPE_V32HI_V32HI_INT_V32HI_INT, V16SI_FTYPE_V16SI_V16SI_INT,
5479         V16SI_FTYPE_V16SI_V16SI_INT_V16SI_INT,
5480         V8DI_FTYPE_V8DI_V8DI_INT_V8DI_INT, V8SI_FTYPE_V8SI_V8SI_INT_V8SI_INT,
5481         V16HI_FTYPE_V16HI_V16HI_INT_V16HI_INT,
5482         V4DI_FTYPE_V4DI_V4DI_INT_V4DI_INT,
5483         V8HI_FTYPE_V8HI_V8HI_INT_V8HI_INT,
5484         V4SI_FTYPE_V4SI_V4SI_INT_V4SI_INT,
5485         V2DI_FTYPE_V2DI_V2DI_INT_V2DI_INT): New types.
5486         * config/i386/i386-builtin.def (__builtin_ia32_vpshld_v32hi,
5487         __builtin_ia32_vpshld_v32hi_mask, __builtin_ia32_vpshld_v16hi,
5488         __builtin_ia32_vpshld_v16hi_mask, __builtin_ia32_vpshld_v8hi,
5489         __builtin_ia32_vpshld_v8hi_mask, __builtin_ia32_vpshld_v16si,
5490         __builtin_ia32_vpshld_v16si_mask, __builtin_ia32_vpshld_v8si,
5491         __builtin_ia32_vpshld_v8si_mask, __builtin_ia32_vpshld_v4si,
5492         __builtin_ia32_vpshld_v4si_mask, __builtin_ia32_vpshld_v8di,
5493         __builtin_ia32_vpshld_v8di_mask, __builtin_ia32_vpshld_v4di,
5494         __builtin_ia32_vpshld_v4di_mask, __builtin_ia32_vpshld_v2di,
5495         __builtin_ia32_vpshld_v2di_mask): New builtins.
5496         * config/i386/i386.c (ix86_expand_args_builtin): Handle new types.
5497         * config/i386/sse.md (vpshld_<mode><mask_name>): New pattern.
5499 2017-11-28  Richard Biener  <rguenther@suse.de>
5501         PR tree-optimization/80776
5502         * gimple-ssa-evrp-analyze.h (evrp_range_analyzer::set_ssa_range_info):
5503         Declare.
5504         * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::set_ssa_range_info):
5505         New function.
5506         (evrp_range_analyzer::record_ranges_from_incoming_edges):
5507         If the incoming edge is an effective fallthru because the other
5508         edge only reaches a __builtin_unreachable () then record ranges
5509         derived from the controlling condition in SSA info.
5510         (evrp_range_analyzer::record_ranges_from_phis): Use set_ssa_range_info.
5511         (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
5513 2017-11-28  Olivier Hainque  <hainque@adacore.com>
5515         * Makefile.in (SELFTEST_FLAGS): Use nul instead of /dev/null
5516         on mingw build hosts.
5518 2017-11-28  Jakub Jelinek  <jakub@redhat.com>
5520         PR debug/81307
5521         * dbxout.c (dbx_block_with_cold_children): Fix function comment.
5523 2017-11-28  Richard Biener  <rguenther@suse.de>
5525         PR middle-end/83141
5526         * gimple-fold.c (gimple_fold_builtin_memory_op): For aggregate
5527         copies generated from memcpy use a character array as reference
5528         type.
5530 2017-11-28  Julia Koval  <julia.koval@intel.com>
5531             Sebastian Peryt  <sebastian.peryt@intel.com>
5533         * Makefile.in (cilkplus.def, cilk-builtins.def, c-family/cilk.o,
5534         c-family/c-cilkplus.o, c-family/array-notation-common.o,
5535         cilk-common.o, cilk.h, cilk-common.c): Remove.
5536         * builtin-types.def
5537         (BT_FN_INT_PTR_PTR_PTR_FTYPE_BT_INT_BT_PTR_BT_PTR_BT_PTR): Remove.
5538         * builtins.c (is_builtin_name): Remove cilkplus condition.
5539         (BUILT_IN_CILK_DETACH, BUILT_IN_CILK_POP_FRAME): Remove.
5540         * builtins.def (DEF_CILK_BUILTIN_STUB, DEF_CILKPLUS_BUILTIN,
5541         cilk-builtins.def, cilkplus.def): Remove.
5542         * cif-code.def (CILK_SPAWN): Remove.
5543         * cilk-builtins.def: Delete.
5544         * cilk-common.c: Ditto.
5545         * cilk.h: Ditto.
5546         * cilkplus.def: Ditto.
5547         * config/darwin.h (fcilkplus): Delete.
5548         * cppbuiltin.c: Ditto.
5549         * doc/extend.texi: Remove cilkplus doc.
5550         * doc/generic.texi: Ditto.
5551         * doc/invoke.texi: Ditto.
5552         * doc/passes.texi: Ditto.
5553         * gcc.c (fcilkplus): Remove.
5554         * gengtype.c (cilk.h): Remove.
5555         * gimple-pretty-print.c (dump_gimple_omp_for): Remove cilkplus
5556         support.
5557         * gimple.h (GF_OMP_FOR_KIND_CILKFOR, GF_OMP_FOR_KIND_CILKSIMD):
5558         Remove.
5559         * gimplify.c (gimplify_return_expr, maybe_fold_stmt,
5560         gimplify_call_expr,
5561         is_gimple_stmt, gimplify_modify_expr, gimplify_scan_omp_clauses,
5562         gimplify_adjust_omp_clauses, gimplify_omp_for, gimplify_expr): Remove
5563         cilkplus conditions.
5564         * ipa-fnsummary.c (ipa_dump_fn_summary, compute_fn_summary,
5565         inline_read_section): Ditto.
5566         * ipa-inline-analysis.c (cilk.h): Remove.
5567         * ira.c (ira_setup_eliminable_regset): Remove cilkplus support.
5568         * lto-wrapper.c (merge_and_complain, append_compiler_options,
5569         append_linker_options): Remove condition for fcilkplus.
5570         * lto/lto-lang.c (cilk.h): Remove.
5571         (lto_init): Remove condition for fcilkplus.
5572         * omp-expand.c (expand_cilk_for_call): Delete.
5573         (expand_omp_taskreg, expand_omp_for_static_chunk,
5574         expand_omp_for): Remove cilkplus
5575         conditions.
5576         (expand_cilk_for): Delete.
5577         * omp-general.c (omp_extract_for_data): Remove cilkplus support.
5578         * omp-low.c (scan_sharing_clauses, create_omp_child_function,
5579         execute_lower_omp, diagnose_sb_0): Ditto.
5580         * omp-simd-clone.c (simd_clone_clauses_extract): Ditto.
5581         * tree-core.h (OMP_CLAUSE__CILK_FOR_COUNT_): Delete.
5582         * tree-nested.c: Ditto.
5583         * tree-pretty-print.c (dump_omp_clause): Remove cilkplus support.
5584         (dump_generic_node): Ditto.
5585         * tree.c (OMP_CLAUSE__CILK_FOR_COUNT_): Delete.
5586         * tree.def (cilk_simd, cilk_for, cilk_spawn_stmt, cilk_sync_stmt):
5587         Delete.
5588         * tree.h (CILK_SPAWN_FN, EXPR_CILK_SPAWN): Delete.
5590 2017-11-28  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
5592         * config/aarch64/aarch64.md (div<mode>3): Change check to TARGET_FLOAT.
5593         * config/aarch64/aarch64.c (aarch64_emit_approx_div): Add early exit
5594         for vector mode and !TARGET_SIMD.
5596 2017-11-28  Jakub Jelinek  <jakub@redhat.com>
5598         * tree.def (SWITCH_EXPR): Change from 3 operand to 2 operand tree.
5599         Adjust comment.
5600         * tree.h (SWITCH_LABELS): Remove.
5601         * gimplify.c (gimplify_switch_expr): Don't test SWITCH_LABELS,
5602         assert SWITCH_BODY is non-NULL.
5603         * tree-pretty-print.c (dump_generic_node): Remove SWITCH_LABELS
5604         handling.
5605         * tree.c (block_may_fallthru): Always return true; for SWITCH_EXPR.
5607         PR tree-optimization/80788
5608         * match.pd (X +- C1 CMP C2 -> X CMP C2 -+ C1): If res
5609         has TREE_OVERFLOW set, call drop_tree_overflow.
5611 2017-11-28  Richard Biener  <rguenther@suse.de>
5613         PR tree-optimization/83158
5614         * tree-vrp.c (intersect_ranges): Prefer ~[0, 0] in a few more cases.
5616 2017-11-28  Segher Boessenkool  <segher@kernel.crashing.org>
5618         PR 81288/target
5619         * config/rs6000/rs6000.c (rs6000_rtx_costs): Do not handle
5620         TARGET_ISEL && !TARGET_MFCRF differently.  Simplify code.
5622 2017-11-27  Segher Boessenkool  <segher@kernel.crashing.org>
5624         * config/rs6000/rs6000.md (<code><GPR:mode><GPR2:mode>2_isel): Change
5625         LT/GT/LTU/GTU to LE/GE/LEU/GEU where possible.
5627 2017-11-27  Michael Meissner  <meissner@linux.vnet.ibm.com>
5629         PR middle_end/82333
5630         * varasm.c (compare_constant): Take the mode of the constants into
5631         account when comparing floating point constants.
5633 2017-11-27  Gerald Pfeifer  <gerald@pfeifer.com>
5635         * hash-set.h (DEFINE_DEBUG_HASH_SET): Remove static qualifier
5636         from explicit instantiation of debug_helper.
5637         * vec.h (DEFINE_DEBUG_VEC): Ditto.
5639 2017-11-27  Richard Biener  <rguenther@suse.de>
5641         * gimple-fold.c (gimple_fold_builtin_memory_op): Remove dead code,
5642         refactor a bit.
5644 2017-11-27  Richard Biener  <rguenther@suse.de>
5646         * tree.c (wide_int_to_tree): Free discarded INTEGER_CST.
5647         (type_hash_canon): Also clear int_cst_hash_table entry for
5648         TYPE_MIN/MAX_VALUE.
5649         (build_nonstandard_integer_type): Hash all TYPE_MAX_VALUEs.
5651 2017-11-27  Tamar Christina  <tamar.christina@arm.com>
5653         * doc/extend.texi: Add -A suffix (ARMv8*-A, ARMv7-A).
5654         * doc/invoke.texi: Add -A suffix (ARMv8*-A, ARMv7-A).
5655         * doc/sourcebuild.texi: Add -A suffix (ARMv8*-A, ARMv7-A).
5657 2017-11-27  Markus Trippelsdorf  <markus@trippelsdorf.de>
5659         * hash-map.h (gt_cleare_cache): Avoid UB.
5661 2017-11-27  Eric Botcazou  <ebotcazou@adacore.com>
5663         * cfgloop.h (struct loop): Document usage of USHRT_MAX for unroll.
5664         * loop-unroll.c (decide_unroll_constant_iterations): Implement it.
5665         (decide_unroll_runtime_iterations): Likewise.
5666         (decide_unroll_stupid): Likewise.
5668 2017-11-27  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
5670         PR target/83109
5671         * config/i386/i386.md: Add a loop with incssp.
5673 2017-11-27  Martin Jambor  <mjambor@suse.cz>
5675         PR tree-optimization/81248
5676         * tree-sra.c (splice_param_accesses): Remove size check.
5677         (decide_one_param_reduction): Fix size check.
5678         * gimple-pretty-print.c (dump_profile): Silence warning.
5679         * params.def (PARAM_IPA_SRA_PTR_GROWTH_FACTOR): Adjust description.
5681 2017-11-27  Jakub Jelinek  <jakub@redhat.com>
5683         PR debug/81307
5684         * dbxout.c (lastlineno): New variable.
5685         (dbx_debug_hooks): Use dbxout_switch_text_section as
5686         switch_text_section debug hook.
5687         (dbxout_function_end): Switch to current_function_section
5688         rather than function_section.  If crtl->has_bb_partition,
5689         output just one N_FUN, depending on in_cold_section_p.
5690         (dbxout_source_line): Remember last lineno in lastlineno.
5691         (dbxout_switch_text_section): New function.
5692         (dbxout_function_decl): Adjust dbxout_block caller.
5693         (dbx_block_with_cold_children): New function.
5694         (dbxout_block): Return true if any LBRAC/RBRAC have been
5695         emitted.  Use dbx_block_with_cold_children at depth == 0
5696         in second partition.  Add PARENT_BLOCKNUM argument, pass
5697         it optionally adjusted to children.  Output LBRAC/RBRAC
5698         around recursive call only if the block is in the current
5699         partition, if not and anything was output, emit empty
5700         range LBRAC/RBRAC.
5701         * final.c (final_scan_insn): Compute cold_function_name
5702         before calling switch_text_section debug hook.  Call
5703         that hook even if dwarf2out_do_frame if not emitting
5704         dwarf debug info.
5706         PR target/83100
5707         * varasm.c (bss_initializer_p): Return true for DECL_COMMON
5708         TREE_READONLY decls.
5710 2017-11-27  Markus Trippelsdorf  <markus@trippelsdorf.de>
5712         PR rtl-optimization/82488
5713         * expr.c (fixup_args_size_notes): Avoid signed integer overflow.
5715 2017-11-26  Julia Koval  <julia.koval@intel.com>
5717         * config/i386/i386.c (processor_target_table): Add skylake_cost for
5718         skylake-avx512.
5719         * config/i386/x86-tune-costs.h (skylake_memcpy, skylake_memset,
5720         skylake_cost): New.
5722 2017-11-26  Julia Koval  <julia.koval@intel.com>
5724         * config/i386/driver-i386.c (host_detect_local_cpu):
5725         Detect skylake-avx512.
5727 2017-11-26  Julia Koval  <julia.koval@intel.com>
5729         * config.gcc: Add -march=cannonlake.
5730         * config/i386/driver-i386.c (host_detect_local_cpu): Detect cannonlake.
5731         * config/i386/i386-c.c (ix86_target_macros_internal): Handle cannonlake.
5732         * config/i386/i386.c (processor_costs): Add m_CANNONLAKE.
5733         (PTA_CANNONLAKE): New.
5734         (processor_target_table): Add cannonlake.
5735         (ix86_option_override_internal): Ditto.
5736         (fold_builtin_cpu): Ditto.
5737         (get_builtin_code_for_version): Handle cannonlake.
5738         (M_INTEL_COREI7_CANNONLAKE): New.
5739         * config/i386/i386.h (TARGET_CANNONLAKE, PROCESSOR_CANNONLAKE): New.
5740         * doc/invoke.texi: Add -march=cannonlake.
5742 2017-11-14  Boris Kolpackov  <boris@codesynthesis.com>
5744         * plugin.c (add_new_plugin): Use platform-specific library extensions.
5745         (try_init_one_plugin): Alternative implementation for MinGW.
5746         * Makefile.in (plugin_implib): New.
5747         (gengtype-lex.c): Fix broken AIX workaround.
5748         * configure: Regenerate.
5749         * doc/plugins.texi: Document support for MinGW.
5751 2017-11-25  Jakub Jelinek  <jakub@redhat.com>
5753         PR rtl-optimization/81553
5754         * combine.c (simplify_if_then_else): In (if_then_else COND (OP Z C1) Z)
5755         to (OP Z (mult COND (C1 * STORE_FLAG_VALUE))) optimization, if OP
5756         is a shift where C1 has different mode than the whole shift, use C1's
5757         mode for MULT rather than the shift's mode.
5759         PR target/82848
5760         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Don't fold
5761         builtins not enabled in the currently selected ISA.
5763 2017-11-24  Jackson Woodruff  <jackson.woodruff@arm.com>
5765         PR tree-optimization/71026
5766         * tree-ssa-math-opts (is_division_by_square, is_square_of): New.
5767         (insert_reciprocals): Change to insert reciprocals before a division
5768         by a square and to insert the square of a reciprocal.
5769         (execute_cse_reciprocals_1): Change to consider division by a square.
5770         (register_division_in): Add importance parameter.
5772 2017-11-24  Richard Biener  <rguenther@suse.de>
5774         PR tree-optimization/82402
5775         * tree-vect-loop-manip.c (create_lcssa_for_virtual_phi): Properly
5776         set SSA_NAME_OCCURS_IN_ABNORMAL_PHI.
5778 2017-11-24  Marc Glisse  <marc.glisse@inria.fr>
5780         * match.pd (0-ptr): New transformation.
5782 2017-11-24  Jan Hubicka  <hubicka@ucw.cz>
5784         PR bootstrap/83015
5785         * ipa-inline.c (inline_small_functions): Set current badnes correctly
5786         when skipping checking.
5788 2017-11-24  Richard Biener  <rguenther@suse.de>
5790         PR tree-optimization/83128
5791         * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Handle STRING_CSTs.
5792         (vn_reference_lookup_3): Likewise.
5794 2017-11-24  Jakub Jelinek  <jakub@redhat.com>
5796         PR sanitizer/83014
5797         * ubsan.c (ubsan_type_descriptor): Use pp_unsigned_wide_integer
5798         instead of pp_printf with HOST_WIDE_INT_PRINT_DEC.  Avoid calling
5799         tree_to_uhwi twice.
5801         * tree-object-size.c (pass_through_call): Use gimple_call_return_flags
5802         ERF_RETURN*ARG* for builtins other than BUILT_IN_ASSUME_ALIGNED,
5803         check for the latter with gimple_call_builtin_p.  Do not handle
5804         BUILT_IN_STPNCPY_CHK which is not a pass through call.
5806 2017-11-24  Christophe Lyon  <christophe.lyon@linaro.org>
5808         * config/arm/arm_neon.h: Fix pragma GCC push_options before
5809         vdot_u32.
5811 2017-11-23  Julia Koval  <julia.koval@intel.com>
5813         * config/i386/avx512vbmi2intrin.h (_mm512_mask_expand_epi8,
5814         _mm512_maskz_expand_epi8, _mm512_mask_expandloadu_epi8,
5815         _mm512_maskz_expandloadu_epi8, _mm512_mask_expand_epi16,
5816         _mm512_maskz_expand_epi16, _mm512_mask_expandloadu_epi16,
5817         _mm512_maskz_expandloadu_epi16): New intrinsics.
5818         * config/i386/avx512vbmi2vlintrin.h (_mm_mask_expand_epi8,
5819         _mm_maskz_expand_epi8, _mm_mask_expandloadu_epi8,
5820         _mm_maskz_expandloadu_epi8, _mm_mask_expand_epi16,
5821         _mm_maskz_expand_epi16, _mm_mask_expandloadu_epi16,
5822         _mm_maskz_expandloadu_epi16, _mm256_mask_expand_epi16,
5823         _mm256_maskz_expand_epi16, _mm256_mask_expandloadu_epi16,
5824         _mm256_maskz_expandloadu_epi16, _mm256_mask_expand_epi8,
5825         _mm256_maskz_expand_epi8, _mm256_mask_expandloadu_epi8,
5826         _mm256_maskz_expandloadu_epi8): New intrinsics.
5827         * config/i386/i386-builtin-types.def (V64QI_FTYPE_PCV64QI_V64QI_UDI,
5828         V32HI_FTYPE_PCV32HI_V32HI_USI, V32QI_FTYPE_PCV32QI_V32QI_USI,
5829         V16HI_FTYPE_PCV16HI_V16HI_UHI, V16QI_FTYPE_PCV16QI_V16QI_UHI,
5830         V8HI_FTYPE_PCV8HI_V8HI_UQI): New types.
5831         * config/i386/i386.c (ix86_expand_special_args_builtin): Use new types.
5832         * config/i386/sse.md (VI248_VLBW): New iterator.
5833         (expand<mode>_mask, expand<mode>_maskz): New patterns.
5835 2017-11-23  Julia Koval  <julia.koval@intel.com>
5837         * config.gcc (avx512vbmi2intrin.h, avx512vbmi2vlintrin): New headers.
5838         * config/i386/avx512vbmi2intrin.h (_mm512_mask_compress_epi8,
5839         _mm512_maskz_compress_epi8, _mm512_mask_compressstoreu_epi8,
5840         _mm512_mask_compress_epi16, _mm512_maskz_compress_epi16,
5841         _mm512_mask_compressstoreu_epi16): New.
5842         * config/i386/avx512vbmi2vlintrin.h (_mm_mask_compress_epi8,
5843         _mm_maskz_compress_epi8, _mm256_mask_compressstoreu_epi16,
5844         _mm_mask_compress_epi16, _mm_maskz_compress_epi16,
5845         _mm256_mask_compress_epi16, _mm256_maskz_compress_epi16,
5846         _mm_mask_compressstoreu_epi8, _mm_mask_compressstoreu_epi16,
5847         _mm256_mask_compress_epi8, _mm256_maskz_compress_epi8,
5848         _mm256_mask_compressstoreu_epi8): New.
5849         * config/i386/i386-builtin-types.def (VOID_FTYPE_PV64QI_V64QI_UDI,
5850         VOID_FTYPE_PV32HI_V32HI_USI, VOID_FTYPE_PV32QI_V32QI_USI,
5851         VOID_FTYPE_PV16QI_V16QI_UHI, VOID_FTYPE_PV16HI_V16HI_UHI,
5852         VOID_FTYPE_PV8HI_V8HI_UQI): New types.
5853         * config/i386/i386-builtin.def (__builtin_ia32_compressqi512_mask,
5854         __builtin_ia32_compresshi512_mask, __builtin_ia32_compressqi256_mask,
5855         __builtin_ia32_compressqi128_mask, __builtin_ia32_compresshi256_mask,
5856         __builtin_ia32_compresshi128_mask,
5857         __builtin_ia32_compressstoreuqi512_mask,
5858         __builtin_ia32_compressstoreuhi512_mask,
5859         __builtin_ia32_compressstoreuqi256_mask,
5860         __builtin_ia32_compressstoreuqi128_mask,
5861         __builtin_ia32_compressstoreuhi256_mask,
5862         __builtin_ia32_compressstoreuhi128_mask): New builtins.
5863         * config/i386/i386.c (ix86_init_mmx_sse_builtins): Create special args
5864         array for flags2.
5865         (ix86_expand_special_args_builtin): Handle new types.
5866         (s4fma_expand): Handle new builtin array.
5867         * config/i386/immintrin.h: Include new headers.
5868         * config/i386/sse.md (VI12_AVX512VLBW): New iterator.
5869         (compress<mode>_mask, compressstore<mode>_mask): New patterns.
5871 2017-11-23  Jakub Jelinek  <jakub@redhat.com>
5873         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Formatting
5874         fixes.  Declare temp and g variables at the top in order to avoid
5875         {} in most of the cases.
5877 2017-11-23  Marc Glisse  <marc.glisse@inria.fr>
5879         * match.pd (ptr-0): New transformation.
5881 2017-11-23  Charles Baylis  <charles.baylis@linaro.org>
5883         * config/arm/arm-protos.h (enum arm_addr_mode_op): New.
5884         (struct addr_mode_cost_table): New.
5885         (struct tune_params): Add field addr_mode_costs.
5886         * config/arm/arm.c (generic_addr_mode_costs): New.
5887         (arm_slowmul_tune): Initialise addr_mode_costs field.
5888         (arm_fastmul_tune): Likewise.
5889         (arm_strongarm_tune): Likewise.
5890         (arm_xscale_tune): Likewise.
5891         (arm_9e_tune): Likewise.
5892         (arm_marvell_pj4_tune): Likewise.
5893         (arm_v6t2_tune): Likewise.
5894         (arm_cortex_tune): Likewise.
5895         (arm_cortex_a8_tune): Likewise.
5896         (arm_cortex_a7_tune): Likewise.
5897         (arm_cortex_a15_tune): Likewise.
5898         (arm_cortex_a35_tune): Likewise.
5899         (arm_cortex_a53_tune): Likewise.
5900         (arm_cortex_a57_tune): Likewise.
5901         (arm_exynosm1_tune): Likewise.
5902         (arm_xgene1_tune): Likewise.
5903         (arm_cortex_a5_tune): Likewise.
5904         (arm_cortex_a9_tune): Likewise.
5905         (arm_cortex_a12_tune): Likewise.
5906         (arm_cortex_a73_tune): Likewise.
5907         (arm_v7m_tune): Likewise.
5908         (arm_cortex_m7_tune): Likewise.
5909         (arm_v6m_tune): Likewise.
5910         (arm_fa726te_tune): Likewise.
5911         (arm_mem_costs): Use table lookup to calculate cost of addressing mode.
5913 2017-11-23  Charles Baylis  <charles.baylis@linaro.org>
5915         * config/arm/arm.c (arm_mem_costs): New function.
5916         (arm_rtx_costs_internal): Use arm_mem_costs.
5918 2017-11-23  Mark Wielaard  <mark@klomp.org>
5920         * dwarf2out.c (init_sections_and_labels): Use generation to create
5921         unique ranges_section_label and ranges_base_label. Return generation.
5922         (output_rnglists): Add generation argument. Use generation to create
5923         unique ranges labels.
5924         (dwarf2out_finish): Get generation from init_sections_and_labels
5925         and pass generation to output_rnglists.
5927 2017-11-23  Mike Stump  <mikestump@comcast.net>
5928             Eric Botcazou  <ebotcazou@adacore.com>
5930         * doc/generic.texi (ANNOTATE_EXPR): Document 3rd operand.
5931         * cfgloop.h (struct loop): Add unroll field.
5932         * function.h (struct function): Add has_unroll bitfield.
5933         * gimplify.c (gimple_boolify) <ANNOTATE_EXPR>: Deal with unroll kind.
5934         (gimplify_expr) <ANNOTATE_EXPR>: Propagate 3rd operand.
5935         * loop-init.c (pass_loop2::gate): Return true if cfun->has_unroll.
5936         (pass_rtl_unroll_loops::gate): Likewise.
5937         * loop-unroll.c (decide_unrolling): Tweak note message.  Skip loops
5938         for which loop->unroll==1.
5939         (decide_unroll_constant_iterations): Use note for consistency and
5940         take loop->unroll into account.  Return early if loop->unroll is set.
5941         Fix thinko in existing test.
5942         (decide_unroll_runtime_iterations): Use note for consistency and
5943         take loop->unroll into account.
5944         (decide_unroll_stupid): Likewise.
5945         * lto-streamer-in.c (input_cfg): Read loop->unroll.
5946         * lto-streamer-out.c (output_cfg): Write loop->unroll.
5947         * tree-cfg.c (replace_loop_annotate_in_block) <annot_expr_unroll_kind>:
5948         New case.
5949         (replace_loop_annotate) <annot_expr_unroll_kind>: Likewise.
5950         (print_loop): Print loop->unroll if set.
5951         * tree-core.h (enum annot_expr_kind): Add annot_expr_unroll_kind.
5952         * tree-inline.c (copy_loops): Copy unroll and set cfun->has_unroll.
5953         * tree-pretty-print.c (dump_generic_node) <annot_expr_unroll_kind>:
5954         New case.
5955         * tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Bail out if
5956         loop->unroll is set and smaller than the trip count.  Otherwise bypass
5957         entirely the heuristics if loop->unroll is set.  Remove dead note.
5958         Fix off-by-one bug in other note.
5959         (try_peel_loop): Bail out if loop->unroll is set.  Fix formatting.
5960         (tree_unroll_loops_completely_1): Force unrolling if loop->unroll
5961         is greater than 1.
5962         (tree_unroll_loops_completely): Make static.
5963         (pass_complete_unroll::execute): Use correct type for variable.
5964         (pass_complete_unrolli::execute): Fix formatting.
5965         * tree.def (ANNOTATE_EXPR): Add 3rd operand.
5967 2017-11-23  Sergey Shalnov  <Sergey.Shalnov@intel.com>
5969         * config/i386/i386.h (TARGET_PREFER_AVX256): Also
5970         enable when TARGET_PREFER_AVX128 is set.
5972 2017-11-23  Jan Hubicka  <hubicka@ucw.cz>
5974         * ipa-profile.c (ipa_propagate_frequency_1): Use count instead of
5975         frequency.
5976         * cgraph.c (cgraph_edge::dump_edge_flags): Dump sreal frequencies.
5977         (cgraph_edge::maybe_hot_p): Use sreal frequencies.
5979 2017-11-23  Jan Hubicka  <hubicka@ucw.cz>
5981         * ipa-fnsummary.c (record_modified_bb_info): Use sreal
5982         frequencies. Fix estimation of aggregate parameters.
5984 2017-11-23  Jan Hubicka  <hubicka@ucw.cz>
5986         * cgraphclones.c (cgraph_node::create_clone): Fix updating of profile
5987         when inlining.
5989 2017-11-23  Tom de Vries  <tom@codesourcery.com>
5991         * config/sh/sh.h (ASM_OUTPUT_ADDR_VEC_ELT): Wrap in "do {} while (0)".
5993 2017-11-23  Tom de Vries  <tom@codesourcery.com>
5995         * config/ft32/ft32.h (ASM_OUTPUT_ADDR_VEC_ELT): Remove semicolon after
5996         macro.
5998 2017-11-23  Oleg Endo  <olegendo@gcc.gnu.org>
6000         PR target/83111
6001         * config/sh/sh.md (udivsi3, divsi3, sibcall_value_pcrel,
6002         sibcall_value_pcrel_fdpic): Use local variable instead of
6003         operands[3].
6004         (calli_tbr_rel): Add missing operand 2.
6005         (call_valuei_tbr_rel): Add missing operand 3.
6007 2017-11-23  Jakub Jelinek  <jakub@redhat.com>
6009         PR middle-end/82253
6010         * expr.c (expand_assignment): For CONCAT to_rtx, complex type from and
6011         bitpos/bitsize covering the whole destination, use store_expr only if
6012         the complex mode is the same.  Otherwise, use expand_normal and if
6013         it returns CONCAT, subreg each part separately instead of trying to
6014         subreg the whole result.
6016 2017-11-23  Richard Biener  <rguenther@suse.de>
6018         PR tree-optimization/23094
6019         * tree-ssa-sccvn.c (vuse_ssa_val): Handle VN_TOP when we
6020         come here from walking over backedges in the first iteration.
6021         (vn_reference_lookup_3): Skip clobbers that store the same value.
6023 2017-11-23  Richard Biener  <rguenther@suse.de>
6025         PR tree-optimization/81403
6026         * tree-ssa-pre.c (get_representative_for): Add parameter specifying
6027         a block we need a leader relative to.
6028         (phi_translate_1): For nary processing require a leader from
6029         get_representative_for given we run expression simplification
6030         using match-and-simplify.  Remove previous fix.
6032 2017-11-22  Jeff Law  <law@redhat.com>
6034         * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::try_find_new_range):
6035         Use new method allocate_value_range rather than accessing the
6036         vrp_value_range_pool data member directly.
6037         * tree-vrp.c (simplify_stmt_for_jump_threading): Tweak slightly
6038         to use extract_range_from_stmt method to avoid need for
6039         extract_range_from_assignment method.
6040         (vrp_prop::vrp_finalize): Use set_lattice_propagation_complete
6041         method rather than setting values_propgated data member directly.
6042         * vr-values.h (class vr_values): Privatize vrp_value_range_pool,
6043         and values propagated data members and extract_range_from_assignment
6044         method.  Reorder private data members to conform to standards.
6045         Add new methods set_lattice_propagation_complete and
6046         allocate_value_range.
6048 2017-11-22  Eric Botcazou  <ebotcazou@adacore.com>
6050         PR rtl-optimization/83030
6051         * doc/rtl.texi (Flags in an RTL Expression): Alphabetize, add entry
6052         for CROSSING_JUMP_P and mention usage of 'jump' for JUMP_INSNs.
6053         (Insns): Delete entry for REG_CROSSING_JUMP in register notes.
6054         * bb-reorder.c (update_crossing_jump_flags): Do not test whether the
6055         CROSSING_JUMP_P flag is already set before setting it.
6056         * cfgrtl.c (fixup_partition_crossing): Likewise.
6057         * reorg.c (relax_delay_slots): Do not consider a CROSSING_JUMP_P
6058         insn as useless.
6060 2017-11-22  Jakub Jelinek  <jakub@redhat.com>
6062         * simplify-rtx.c (simplify_binary_operation_1) <case VEC_SERIES>:
6063         Handle the case where both arguments are using gen_const_vec_series.
6065 2017-11-22  David Malcolm  <dmalcolm@redhat.com>
6067         PR c++/62170
6068         * pretty-print.c (pp_format): Move quoting implementation to
6069         pp_begin_quote and pp_end_quote.  Update pp_format_decoder call
6070         to pass address of "quote" local.
6071         (pp_begin_quote): New function.
6072         (pp_end_quote): New function.
6073         * pretty-print.h (printer_fn): Convert penultimate param from bool
6074         to bool *.
6075         (pp_begin_quote): New decl.
6076         (pp_end_quote): New decl.
6077         * tree-diagnostic.c (default_tree_printer): Convert penultimate
6078         param from bool to bool *.
6079         * tree-diagnostic.h (default_tree_printer): Likewise.
6081 2017-11-22  Jeff Law  <law@redhat.com>
6083         * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::evrp_range_analyzer)
6084         Initialize vr_values.
6085         (evrp_range_analyzer::try_find_new_range): Call methods attached to
6086         vr_values via vr_values class instance rather than delegators.
6087         (evrp_range_analyzer::record_ranges_from_phis): Likewise.
6088         (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
6089         (evrp_range_analyzer::push_value_range): Likewise.
6090         (evrp_range_analyzer::pop_value_range): Likewise.
6091         * gimple-ssa-evrp-analyze.h (class evrp_range_analyzer): Remove
6092         most delegators.  Those remaining are exposed as public interfaces.
6093         Make vr_values a pointer and private.
6094         (evrp_range_analyzer::~evrp_range_analyzer): Delete the attached
6095         vr_values.
6096         (evrp_range_analyzer::get_vr_value): New method.
6097         * gimple-ssa-evrp.c (class evrp_folder): Use DISABLE_COPY_AND_ASSIGN.
6098         (evrp_folder::evrp_folder): New ctor to initialize vr_values.
6099         (class evrp_dom_walker): Attach evrp_folder class, initialize
6100         it in the ctor.  Remove temporary delegators.
6101         (evrp_dom_walker::before_dom_children): Call methods in attached
6102         evrp_range_analyzer class via class instance pointer.  Use
6103         free value_range_constant_singleton to remove need for
6104         op_with_constant_singleton_value delegator method.  Do not
6105         create a vrp_prop class instance for every call!  Narrow
6106         scope of a couple variables.
6107         (evrp_dom_walker::cleanup): Call methods in attached
6108         evrp_range_analyzer class via class instance pointer.
6109         * vr-values.h (class vr_values): Privatize many methods and
6110         data members.
6112 2017-11-22  Richard Sandiford  <richard.sandiford@linaro.org>
6114         * tree.def (REDUC_MAX_EXPR, REDUC_MIN_EXPR, REDUC_PLUS_EXPR): Delete.
6115         * cfgexpand.c (expand_debug_expr): Remove handling for them.
6116         * expr.c (expand_expr_real_2): Likewise.
6117         * fold-const.c (const_unop): Likewise.
6118         * optabs-tree.c (optab_for_tree_code): Likewise.
6119         * tree-cfg.c (verify_gimple_assign_unary): Likewise.
6120         * tree-inline.c (estimate_operator_cost): Likewise.
6121         * tree-pretty-print.c (dump_generic_node): Likewise.
6122         (op_code_prio): Likewise.
6123         (op_symbol_code): Likewise.
6124         * internal-fn.def (DEF_INTERNAL_SIGNED_OPTAB_FN): Define.
6125         (IFN_REDUC_PLUS, IFN_REDUC_MAX, IFN_REDUC_MIN): New internal functions.
6126         * internal-fn.c (direct_internal_fn_optab): New function.
6127         (direct_internal_fn_array, direct_internal_fn_supported_p
6128         (internal_fn_expanders): Handle DEF_INTERNAL_SIGNED_OPTAB_FN.
6129         * fold-const-call.c (fold_const_reduction): New function.
6130         (fold_const_call): Handle CFN_REDUC_PLUS, CFN_REDUC_MAX and
6131         CFN_REDUC_MIN.
6132         * tree-vect-loop.c: Include internal-fn.h.
6133         (reduction_code_for_scalar_code): Rename to...
6134         (reduction_fn_for_scalar_code): ...this and return an internal
6135         function.
6136         (vect_model_reduction_cost): Take an internal_fn rather than
6137         a tree_code.
6138         (vect_create_epilog_for_reduction): Likewise.  Build calls rather
6139         than assignments.
6140         (vectorizable_reduction): Use internal functions rather than tree
6141         codes for the reduction operation.  Update calls to the functions
6142         above.
6143         * config/aarch64/aarch64-builtins.c (aarch64_gimple_fold_builtin):
6144         Use calls to internal functions rather than REDUC tree codes.
6145         * config/aarch64/aarch64-simd.md: Update comment accordingly.
6147 2017-11-22  Olivier Hainque  <hainque@adacore.com>
6149         * config/vxworks.c (vxworks_override_options): Pick default
6150         dwarf version from macro value, VXWORKS_DWARF_VERSION_DEFAULT.
6151         * config/vxworks.h: Define VXWORKS_DWARF_VERSION_DEFAULT and
6152         DWARF_GNAT_ENCODINGS_DEFAULT.
6153         * config/vxworksae.h: Likewise.
6155 2017-11-22  Marc Glisse  <marc.glisse@inria.fr>
6157         PR tree-optimization/83104
6158         * vr-values.c (simplify_stmt_using_ranges): Check integral argument,
6159         not result.
6161 2017-11-22  Marek Polacek  <polacek@redhat.com>
6162             H.J. Lu  <hongjiu.lu@intel.com>
6163             Jason Merrill  <jason@redhat.com>
6165         PR c++/60336
6166         PR middle-end/67239
6167         PR target/68355
6168         * calls.c (initialize_argument_information): Call
6169         warn_parameter_passing_abi target hook.
6170         (store_one_arg): Use 0 for empty record size.  Don't push 0 size
6171         argument onto stack.
6172         (must_pass_in_stack_var_size_or_pad): Return false for empty types.
6173         * common.opt: Update -fabi-version description.
6174         * config/i386/i386.c (init_cumulative_args): Set cum->warn_empty.
6175         (ix86_gimplify_va_arg): Call arg_int_size_in_bytes instead of
6176         int_size_in_bytes.
6177         (ix86_is_empty_record): New function.
6178         (ix86_warn_parameter_passing_abi): New function.
6179         (TARGET_EMPTY_RECORD_P): Redefine.
6180         (TARGET_WARN_PARAMETER_PASSING_ABI): Redefine.
6181         * config/i386/i386.h (CUMULATIVE_ARGS): Add warn_empty.
6182         * doc/tm.texi: Regenerated.
6183         * doc/tm.texi.in (TARGET_EMPTY_RECORD_P,
6184         TARGET_WARN_PARAMETER_PASSING_ABI): Add.
6185         * dwarf2out.c (get_ultimate_context): Move to tree.c.
6186         * explow.c (hard_function_value): Call arg_int_size_in_bytes
6187         instead of int_size_in_bytes.
6188         * expr.c (copy_blkmode_to_reg): Likewise.
6189         * function.c (aggregate_value_p): Return 0 for empty types.
6190         (assign_parm_find_entry_rtl): Call warn_parameter_passing_abi
6191         target hook.
6192         (locate_and_pad_parm): Call arg size_in_bytes instead
6193         size_in_bytes.
6194         * lto-streamer-out.c (hash_tree): Hash TYPE_EMPTY_P and DECL_PADDING_P.
6195         * stor-layout.c (finalize_type_size): Set TYPE_EMPTY_P.
6196         * target.def (empty_record_p, warn_parameter_passing_abi): New target
6197         hooks.
6198         * targhooks.c (hook_void_CUMULATIVE_ARGS_tree): New hook.
6199         (std_gimplify_va_arg_expr): Skip empty records.  Call
6200         arg_size_in_bytes instead size_in_bytes.
6201         * targhooks.h (hook_void_CUMULATIVE_ARGS_tree): Declare.
6202         * tree-core.h (tree_type_common): Add empty_flag.
6203         (tree_decl_common): Update comments.
6204         * tree-streamer-in.c (unpack_ts_decl_common_value_fields): Stream
6205         DECL_PADDING_P.
6206         (unpack_ts_type_common_value_fields): Stream TYPE_EMPTY_P.
6207         * tree-streamer-out.c (pack_ts_decl_common_value_fields): Stream
6208         DECL_PADDING_P.
6209         (pack_ts_type_common_value_fields): Stream TYPE_EMPTY_P.
6210         * tree.c (default_is_empty_type): New function.
6211         (default_is_empty_record): New function.
6212         (arg_int_size_in_bytes): New function.
6213         (arg_size_in_bytes): New function.
6214         (get_ultimate_context): New function.
6215         * tree.h: Define TYPE_EMPTY_P, DECL_PADDING_P and
6216         TRANSLATION_UNIT_WARN_EMPTY_P.
6217         (default_is_empty_record, arg_int_size_in_bytes,
6218         arg_size_in_bytes, get_ultimate_context): Declare.
6220 2017-11-22  Thomas Preud'homme  <thomas.preudhomme@arm.com>
6222         * config/arm/arm.c (cmse_clear_registers): New function.
6223         (cmse_nonsecure_call_clear_caller_saved): Replace register clearing
6224         code by call to cmse_clear_registers.
6225         (cmse_nonsecure_entry_clear_before_return): Likewise.
6227 2017-11-22  Tamar Christina  <tamar.christina@arm.com>
6229         * config/arm/arm_neon.h (vdot_u32, vdotq_u32)
6230         (vdot_s32, vdotq_s32): New.
6231         (vdot_lane_u32, vdotq_lane_u32): New.
6232         (vdot_lane_s32, vdotq_lane_s32): New.
6235 2017-11-22  Richard Sandiford  <richard.sandiford@linaro.org>
6237         PR middle-end/82547
6238         * wide-int.cc (wi::add_large, wi::sub_large): Fix overflow detection
6239         for unsigned values with fewer HWIs than the precision.
6240         (test_overflow): New function.
6241         (wide_int_cc_tests): Call it.
6243 2017-11-22  Richard Sandiford  <richard.sandiford@linaro.org>
6245         * emit-rtl.c (init_derived_machine_modes): Make sure ptr_mode
6246         has the same mode class as Pmode.
6248 2017-11-22  Thomas Preud'homme  <thomas.preudhomme@arm.com>
6250         * config/arm/arm.c (cmse_nonsecure_call_clear_caller_saved): Get rid of
6251         padding_bits_to_clear_ptr.
6252         (cmse_nonsecure_entry_clear_before_return): Likewise.
6254 2017-11-22  Thomas Preud'homme  <thomas.preudhomme@arm.com>
6256         * config/arm/arm.c (cmse_nonsecure_call_clear_caller_saved): Use
6257         auto_sbitap instead of integer bitfield to control register needing
6258         clearing.
6260 2017-11-22  Jakub Jelinek  <jakub@redhat.com>
6262         PR tree-optimization/83044
6263         * tree-vrp.c (vrp_prop::check_array_ref): If eltsize is not
6264         INTEGER_CST or is 0, clear up_bound{,_p1} and later ignore tests
6265         that need the upper bound.  Subtract offset from
6266         get_addr_base_and_unit_offset only if positive and subtract it
6267         before division by eltsize rather than after it.
6269         PR debug/83084
6270         * valtrack.c (propagate_for_debug_subst, propagate_for_debug): Reset
6271         debug insns if they would contain UNSPEC_VOLATILE or volatile asm.
6272         (dead_debug_insert_temp): Likewise, but also ignore even non-volatile
6273         asm.
6275         PR middle-end/82875
6276         * optabs.c (expand_doubleword_mult, expand_binop): Before calling
6277         expand_binop with *mul_widen_optab, make sure at least one of the
6278         operands doesn't have VOIDmode.
6280         PR debug/83034
6281         * dwarf2out.c (mem_loc_descriptor): Handle VEC_SERIES.
6283         PR rtl-optimization/82044
6284         PR tree-optimization/82042
6285         * dse.c (record_store): Check for overflow.
6286         (check_mem_read_rtx): Properly check for overflow if width == -1, call
6287         add_wild_read instead of clear_rhs_from_active_local_stores on
6288         overflow and log it into dump_file.
6290 2017-11-22  Richard Biener  <rguenther@suse.de>
6292         * gimple-iterator.c (gimple_find_edge_insert_loc): Ignore
6293         fake edges to exit when looking for a place to insert.
6294         * tree-ssa-pre.c (clear_expression_ids): Inline into callers
6295         and remove.
6296         (insert_into_preds_of_block): Commit edge insertion immediately,
6297         assert that doesn't require new BBs.
6298         (fini_pre): Release expressions.
6299         (pass_pre::execute): Shuffle things around a bit, if the fn
6300         is too large do not compute AVAIL either as this is really the
6301         quadratic bit.
6303 2017-11-22  Richard Biener  <rguenther@suse.de>
6305         PR tree-optimization/83089
6306         * tree-if-conv.c (pass_if_conversion::execute): If anything
6307         changed reset SCEV and free the number of iteration estimates.
6309 2017-11-21  Martin Sebor  <msebor@redhat.com>
6311         PR tree-optimization/82945
6312         * calls.h (warn_nonstring_bound): Remove unused function.
6314 2017-11-21  Martin Sebor  <msebor@redhat.com>
6316         PR tree-optimization/82945
6317         * builtins.c (expand_builtin_strlen): Call maybe_warn_nonstring_arg.
6318         * calls.h (maybe_warn_nonstring_arg): Declare new function.
6319         * calls.c (get_attr_nonstring_decl, maybe_warn_nonstring_arg): New
6320         functions.
6321         (initialize_argument_information): Call maybe_warn_nonstring_arg.
6322         * calls.h (get_attr_nonstring_decl): Declare new function.
6323         * doc/extend.texi (attribute nonstring): Update.
6324         * gimple-fold.c (gimple_fold_builtin_strncpy): Call
6325         get_attr_nonstring_decl and handle it.
6326         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Same.  Improve
6327         detection of nul-termination.
6328         (strlen_to_stridx): Change to a pointer.
6329         (handle_builtin_strlen, handle_builtin_stxncpy): Adjust.
6330         (pass_strlen::execute): Same.
6332 2017-11-21  Sergey Shalnov  <Sergey.Shalnov@intel.com>
6334         * config/i386/i386-opts.h (enum prefer_vector_width): Added new enum
6335         for the new option -mprefer-vector-width=[none|128|256|512].
6336         * config/i386/i386.c (ix86_target_string): remove old style options
6337         -mprefer-avx256 and make -mprefer-avx128 as alias.
6338         (ix86_option_override_internal):  Apply defaults for the
6339         -mprefer-vector-width=[128|256] option.
6340         * config/i386/i386.h (TARGET_PREFER_AVX128, TARGET_PREFER_AVX256):
6341         Implement macros to work with -mprefer-vector-width=.
6342         * config/i386/i386.opt: Implemented option
6343         -mprefer-vector-width=[none|128|256|512].
6344         * doc/invoke.texi: Documentation for
6345         -mprefer-vector-width=[none|128|256|512].
6347 2017-11-21  Pat Haugen  <pthaugen@us.ibm.com>
6349         * config/rs6000/ppc-asm.h (f50, vs50): Fix values.
6351 2017-11-21  Jonathan Wakely  <jwakely@redhat.com>
6353         * doc/invoke.texi (-Wmaybe-uninitialized): Rephrase for clarity.
6355 2017-11-21  Marc Glisse  <marc.glisse@inria.fr>
6357         * doc/generic.texi: Document POINTER_DIFF_EXPR, update
6358         POINTER_PLUS_EXPR.
6359         * cfgexpand.c (expand_debug_expr): Handle POINTER_DIFF_EXPR.
6360         * expr.c (expand_expr_real_2): Likewise.
6361         * fold-const.c (const_binop, fold_addr_of_array_ref_difference,
6362         fold_binary_loc): Likewise.
6363         * match.pd (X-X, P+(Q-P), &D-P, (P+N)-P, P-(P+N), (P+M)-(P+N),
6364         P-Q==0, -(A-B), X-Z<Y-Z, (X-Z)-(Y-Z), Z-X<Z-Y, (Z-X)-(Z-Y),
6365         (A-B)+(C-A)): New transformations for POINTER_DIFF_EXPR, based on
6366         MINUS_EXPR transformations.
6367         * optabs-tree.c (optab_for_tree_code): Handle POINTER_DIFF_EXPR.
6368         * tree-cfg.c (verify_expr, verify_gimple_assign_binary): Likewise.
6369         * tree-inline.c (estimate_operator_cost): Likewise.
6370         * tree-pretty-print.c (dump_generic_node, op_code_prio,
6371         op_symbol_code): Likewise.
6372         * tree-vect-stmts.c (vectorizable_operation): Likewise.
6373         * vr-values.c (extract_range_from_binary_expr): Likewise.
6374         * varasm.c (initializer_constant_valid_p_1): Likewise.
6375         * tree.def: New tree code POINTER_DIFF_EXPR.
6377 2017-11-21  Uros Bizjak  <ubizjak@gmail.com>
6379         * config/i386/i386.md (*bswap<mode>2_movbe): Add
6380         integer suffix to movbe mnemonic.
6381         (*bswaphi2_movbe): Ditto.
6382         (bswaphi_lowpart): Merge with *bswaphi_lowpart_1.
6384 2017-11-21  Jakub Jelinek  <jakub@redhat.com>
6386         PR c++/83045
6387         * tree-cfg.c (pass_warn_function_return::execute): Formatting fix.
6388         Also warn if seen __builtin_unreachable () call with BUILTINS_LOCATION.
6389         Use LOCATION_LOCUS when comparing against UNKNOWN_LOCATION.
6391 2017-11-21  Martin Liska  <mliska@suse.cz>
6393         * tree-inline.c (expand_call_inline): Remove not needed
6394         xstrdup_for_dump.
6396 2017-11-21  James Cowgill  <James.Cowgill@imgtec.com>
6397             Jakub Jelinek  <jakub@redhat.com>
6399         PR target/82880
6400         * config/mips/frame-header-opt.c (mips_register_frame_header_opt):
6401         Remove static keyword from f variable.
6403 2017-11-21  Jakub Jelinek  <jakub@redhat.com>
6405         PR tree-optimization/83086
6406         * gimple-ssa-store-merging.c
6407         (imm_store_chain_info::try_coalesce_bswap): Test this_n.base_addr
6408         rather than n.base_addr.
6410 2017-11-21  Martin Liska  <mliska@suse.cz>
6412         PR rtl-optimization/82044
6413         PR tree-optimization/82042
6414         * dse.c (check_mem_read_rtx): Check for overflow.
6416 2017-11-21  Aldy Hernandez  <aldyh@redhat.com>
6418         * tree-ssa-threadbackward.c (find_jump_threads_backwards): Fix
6419         typo in comment.
6421 2017-11-21  Claudiu Zissulescu  <claziss@synopsys.com>
6423         * cfgrtl.c (force_nonfallthru_and_redirect): Don't split a call
6424         and its corresponding call arg location note.
6426 2017-11-21  Claudiu Zissulescu  <claziss@synopsys.com>
6427             Andrew Burgess  <andrew.burgess@embecosm.com>
6429         * config/arc/arc-protos.h (arc_compute_frame_size): Delete
6430         declaration.
6431         (arc_return_slot_offset): Likewise.
6432         (arc_eh_return_address_location): New declaration.
6433         * config/arc/arc.c (TARGET_BUILTIN_SETJMP_FRAME_VALUE): Define.
6434         (MUST_SAVE_REGISTER): Add exception handler case.
6435         (MUST_SAVE_RETURN_ADDR): Likewise.
6436         (arc_frame_pointer_required): Likewise.
6437         (arc_frame_pointer_needed): New function.
6438         (arc_compute_frame_size): Changed.
6439         (arc_expand_prologue): Likewise.
6440         (arc_expand_epilogue): Likewise.
6441         (arc_initial_elimination_offset): Likewise.
6442         (arc_return_slot_offset): Delete.
6443         (arc_eh_return_address_location): New function.
6444         (arc_builtin_setjmp_frame_value): Likewise.
6445         * config/arc/arc.h (EH_RETURN_DATA_REGNO): Use 2 registers.
6446         (EH_RETURN_STACKADJ_RTX): Define.
6447         (EH_RETURN_HANDLER_RTX): Likewise.
6448         * config/arc/arc.md (eh_return): Delete.
6450 2017-11-21  Aldy Hernandez  <aldyh@redhat.com>
6452         * print-tree.h (debug_vec_tree): Remove prototype.
6453         * gdbinit.in (pvt): Remove macro.
6455 2017-11-21  Jakub Jelinek  <jakub@redhat.com>
6457         PR tree-optimization/83047
6458         * gimple-ssa-store-merging.c
6459         (imm_store_chain_info::output_merged_store): If the loads with the
6460         same vuse are in different basic blocks, for load_gsi pick a load
6461         location that is dominated by the other loads.
6463         PR c++/83059
6464         * config/i386/i386.c (ix86_memmodel_check): Start
6465         -Winvalid-memory-model diagnostics with lowercase letter.
6467         PR debug/82718
6468         * dwarf2out.c (dw_loc_list): If crtl->has_bb_partition, temporarily
6469         set in_cold_section_p to the partition containing loc_list->first.
6470         When seeing loc_list->last_before_switch node, update secname and
6471         perform range_across_switch second partition handling only after that.
6473         PR debug/82933
6474         * run-rtl-passes.c: Include debug.h.
6475         (run_rtl_passes): Call debug_hooks->assembly_start.
6476         * dwarf2out.c (dwarf2out_assembly_start): Return early if invoked
6477         multiple times.
6479         PR target/82981
6480         * internal-fn.c (expand_mul_overflow): Use OPTAB_WIDEN instead of
6481         OPTAB_DIRECT in calls to expand_simple_binop.
6483 2017-11-20  David Malcolm  <dmalcolm@redhat.com>
6485         PR c/81404
6486         * Makefile.in (C_COMMON_OBJS): Add c-family/known-headers.o.
6488 2017-11-20  David Malcolm  <dmalcolm@redhat.com>
6490         PR c++/72786
6491         * spellcheck.h (best_match::blithely_get_best_candidate): New
6492         accessor.
6494 2017-11-20  Jakub Jelinek  <jakub@redhat.com>
6496         * config/i386/i386.c (parse_mtune_ctrl_str): Start diagnostics
6497         with lower case letter.
6499 2017-11-20  Uros Bizjak  <ubizjak@gmail.com>
6501         * config/i386/i386.md (bswaphi2): New expander.
6502         (*bswaphi2_movbe): New insn pattern.
6503         (bswaphi -> rorhi pepehole2): New peephole pattern.
6505 2017-11-20  Jeff Law  <law@redhat.com>
6507         * Makefile.in (OBJS): Add gimple-ssa-evrp-analyze.o.
6508         * gimple-ssa-evrp-analyze.c: New file pulled from gimple-ssa-evrp.c.
6509         * gimple-ssa-evrp-analyze.h: New file pulled from gimple-ssa-evrp.c.
6510         * gimple-ssa-evrp.c: Remove bits moved into new files.  Include
6511         gimple-ssa-evrp-analyze.h.
6513         * gimple-ssa-evrp.c (evrp_dom_walker::before_dom_children): Do not
6514         set BB_VISITED here.
6515         (evrp_range_analyzer::enter): Set BB_VISITED here instead.
6517 2017-11-20  Bin Cheng  <bin.cheng@arm.com>
6519         * tree-predcom.c: Add general comment on Store-Store chains.
6520         (split_data_refs_to_components): Postpone clearing eliminate_store_p
6521         flag in component.
6522         (get_chain_last_ref_at): Rename into...
6523         (get_chain_last_write_at): ...this.
6524         (get_chain_last_write_before_load): New function.
6525         (add_ref_to_chain): Promote type of chain from CT_STORE_LOAD to
6526         CT_STORE_STORE when write reference is added.
6527         (determine_roots_comp): Support load ref in CT_STORE_STORE chains.
6528         (is_inv_store_elimination_chain): Update get_chain_last_write_at call.
6529         (initialize_root_vars_store_elim_1): Ditto.
6530         (initialize_root_vars_store_elim_2): Ditto.  Replace rhs once default
6531         definition is created.
6532         (execute_pred_commoning_chain): Support load ref in CT_STORE_STORE
6533         chain by replacing it with dominant stored value.
6535 2017-11-20  Bin Cheng  <bin.cheng@arm.com>
6537         * tree-predcom.c (add_ref_to_chain): Remove check on distance.
6539 2017-11-20  Marc Glisse  <marc.glisse@inria.fr>
6541         * vr-values.c (extract_range_from_binary_expr): Use a full range
6542         for VR_VARYING.
6544 2017-11-20  Thomas Preud'homme  <thomas.preudhomme@arm.com>
6546         * config/arm/arm.md (R4_REGNUM): Define constant.
6547         (nonsecure_call_internal): Remove r4 clobber.
6548         (nonsecure_call_value_internal): Likewise.
6549         * config/arm/thumb1.md (nonsecure_call_reg_thumb1_v5): Remove second
6550         clobber and resequence match_operands.
6551         (nonsecure_call_value_reg_thumb1_v5): Likewise.
6552         * config/arm/thumb2.md (nonsecure_call_reg_thumb2): Likewise.
6553         (nonsecure_call_value_reg_thumb2): Likewise.
6555 2017-11-20  Jakub Jelinek  <jakub@redhat.com>
6557         PR tree-optimization/78821
6558         * gimple-ssa-store-merging.c (find_bswap_or_nop_load): Give up
6559         if base is TARGET_MEM_REF.  If base is not MEM_REF, set base_addr
6560         to the address of the base rather than the base itself.
6561         (find_bswap_or_nop_1): Just use pointer comparison for vuse check.
6562         (find_bswap_or_nop_finalize): New function.
6563         (find_bswap_or_nop): Use it.
6564         (bswap_replace): Return a tree rather than bool, change first
6565         argument from gimple * to gimple_stmt_iterator, allow inserting
6566         into an empty sequence, allow ins_stmt to be NULL - then emit
6567         all stmts into gsi.  Fix up MEM_REF address gimplification.
6568         (pass_optimize_bswap::execute): Adjust bswap_replace caller.
6569         (struct store_immediate_info): Add N and INS_STMT non-static
6570         data members.
6571         (store_immediate_info::store_immediate_info): Initialize them
6572         from newly added ctor args.
6573         (merged_store_group::apply_stores): Formatting fixes.  Sort by
6574         bitpos at the end.
6575         (stmts_may_clobber_ref_p): For stores call also
6576         refs_anti_dependent_p.
6577         (gather_bswap_load_refs): New function.
6578         (imm_store_chain_info::try_coalesce_bswap): New method.
6579         (imm_store_chain_info::coalesce_immediate_stores): Use it.
6580         (split_group): Handle LROTATE_EXPR and NOP_EXPR rhs_code specially.
6581         (imm_store_chain_info::output_merged_store): Fail if number of
6582         new estimated stmts is bigger or equal than old.  Handle LROTATE_EXPR
6583         and NOP_EXPR rhs_code.
6584         (pass_store_merging::process_store): Compute n and ins_stmt, if
6585         ins_stmt is non-NULL and the store rhs is otherwise invalid, use
6586         LROTATE_EXPR rhs_code.  Pass n and ins_stmt to store_immediate_info
6587         ctor.
6588         (pass_store_merging::execute): Calculate dominators.
6590         * tree-ssa-math-opts.c (nop_stats, bswap_stats, struct symbolic_number,
6591         BITS_PER_MARKER, MARKER_MASK, MARKER_BYTE_UNKNOWN, HEAD_MARKER, CMPNOP,
6592         CMPXCHG, do_shift_rotate, verify_symbolic_number_p,
6593         init_symbolic_number, find_bswap_or_nop_load, perform_symbolic_merge,
6594         find_bswap_or_nop_1, find_bswap_or_nop, pass_data_optimize_bswap,
6595         class pass_optimize_bswap, bswap_replace,
6596         pass_optimize_bswap::execute): Moved to ...
6597         * gimple-ssa-store-merging.c: ... this file.
6598         Include optabs-tree.h.
6599         (nop_stats, bswap_stats, do_shift_rotate, verify_symbolic_number_p,
6600         init_symbolic_number, find_bswap_or_nop_load, perform_symbolic_merge,
6601         find_bswap_or_nop_1, find_bswap_or_nop, bswap_replace): Put into
6602         anonymous namespace, remove static keywords.
6603         (pass_optimize_bswap::gate): Test BITS_PER_UNIT == 8 here...
6604         (pass_optimize_bswap::execute): ... rather than here.  Formatting fix.
6606 2017-11-20  Jan Hubicka  <hubicka@ucw.cz>
6608         PR bootstrap/83062
6609         * ipa-inline.c (can_inline_edge_p): Fix typo in previous patch.
6611 2017-11-20  Aldy Hernandez  <aldyh@redhat.com>
6613         * vec.h (debug_helper): New function.
6614         (DEFINE_DEBUG_VEC): New macro.
6615         * hash-set.h (debug_helper): New function.
6616         (DEFINE_DEBUG_HASH_SET): New macro.
6617         * cfg.c (debug_slim (edge)): New function.
6618         Call DEFINE_DEBUG_VEC for edges.
6619         Call DEFINE_DEBUG_HASH_SET for edges.
6620         * cfghooks.c (debug_slim (basic_block)): New function.
6621         Call DEFINE_DEBUG_VEC for basic blocks.
6622         Call DEFINE_DEBUG_HASH_SET for basic blocks.
6623         * print-tree.c (debug_slim): New function to handle trees.
6624         Call DEFINE_DEBUG_VEC for trees.
6625         Call DEFINE_DEBUG_HASH_SET for trees.
6626         (debug (vec<tree, va_gc>) &): Remove.
6627         (debug (<vec<tree, va_gc>) *): Remove.
6628         * print-rtl.c (debug_slim): New function to handle const_rtx.
6629         Call DEFINE_DEBUG_VEC for rtx_def.
6630         Call DEFINE_DEBUG_VEC for rtx_insn.
6631         Call DEFINE_DEBUG_HASH_SET for rtx_def.
6632         Call DEFINE_DEBUG_HASH_SET for rtx_insn.
6633         * sel-sched-dump.c (debug (vec<rtx_insn *> &): Remove.
6634         (debug (vec<rtx_insn *> *ptr): Remove.
6635         (debug_insn_vector): Remove.
6636         * stor-layout.c (debug_rli): Call debug() instead of debug_vec_tree.
6638 2017-11-20  Tom de Vries  <tom@codesourcery.com>
6640         PR rtl-optimization/82020
6641         * simplify-rtx.c (simplify_ternary_operation): Fix comparison mode of
6642         IF_THEN_ELSE condition.
6644 2017-11-19  Jeff Law  <law@redhat.com>
6646         * tree-ssa-dom.c (record_equivalences_from_phis): Fix handling
6647         of degenerates resulting from ignoring an edge.
6649 2017-11-19  Jan Hubicka  <hubicka@ucw.cz>
6651         PR ipa/81360
6652         * ipa-inline.c (can_inline_edge_p): Also check that caller is optimized
6654 2017-11-19  Jan Hubicka  <hubicka@ucw.cz>
6656         PR ipa/83001
6657         * profile-count.c (profile_count::to_sreal_scale): Fix return value
6658         for uninitialied counts.
6660 2017-11-19  Jan Hubicka  <hubicka@ucw.cz>
6662         PR ipa/60243
6663         * tree-inline.c (estimate_num_insns): Set to 1 at least.
6665 2017-11-19  Jan Hubicka  <hubicka@ucw.cz>
6667         PR target/82713
6668         * i386.c (ix86_builtin_vectorization_cost): Be ready for insane types.
6670 2017-11-19  Tom de Vries  <tom@codesourcery.com>
6672         * config/arc/arc.h (FUNCTION_PROFILER): Remove semicolon after
6673         "do while (0)".
6675 2017-11-19  Tom de Vries  <tom@codesourcery.com>
6677         * config/phoenix.h (TARGET_OS_CPP_BUILTINS): Remove semicolon after
6678         "do {} while (0)".
6680 2017-11-19  Tom de Vries  <tom@codesourcery.com>
6682         * config/visium/visium.h (ASM_OUTPUT_CASE_END): Remove semicolon after
6683         macro body.
6685 2017-11-19  Tom de Vries  <tom@codesourcery.com>
6687         * config/ft32/ft32.h (REGISTER_TARGET_PRAGMAS): Remove semicolon after
6688         "do {} while (0)".
6689         * config/spu/spu.h (REGISTER_TARGET_PRAGMAS): Same.
6691 2017-11-19  Tom de Vries  <tom@codesourcery.com>
6693         * config/mcore/mcore-elf.h (MCORE_EXPORT_NAME): Remove semicolon after
6694         "do {} while (0)".
6695         * config/mcore/mcore.h (ASM_OUTPUT_ALIGNED_COMMON): After missing
6696         semicolon after MCORE_EXPORT_NAME call.
6698 2017-11-19  Tom de Vries  <tom@codesourcery.com>
6700         PR target/82961
6701         * vmsdbgout.c (vmsdbgout_early_finish): New function.
6702         (vmsdbg_debug_hooks): Set early_finish field to vmsdbgout_early_finish.
6704 2017-11-18  Jan Hubicka  <hubicka@ucw.cz>
6706         * cgraphclones.c (cgraph_edge::clone): Rename gcov_count to prof_count.
6707         (cgraph_edge::clone): Cleanup updating of profile.
6708         * ipa-cp.c (update_profiling_info): Likewise.
6709         * ipa-inline-transform.c (inline_transform): Likewise.
6710         * ipa-inline.c (inline_small_functions): Add missing space to dump.
6711         * ipa-split.c (execute_split_functions): Do not split when function
6712         is cold.
6713         * predict.c (estimate_bb_frequencies): Cleanup updating of profile.
6714         * profile-count.c (profile_count::dump): Add global0.
6715         (profile_count::to_cgraph_frequency): Do not ICE when entry is
6716         undefined.
6717         (profile_count::to_sreal_scale): Likewise.
6718         (profile_count::adjust_for_ipa_scaling): Fix typo in comment.
6719         (profile_count::combine_with_ipa_count): New function.
6720         * profile-count.h (profile_guessed_global0adjusted): New.
6721         (profile_count::adjusted_zero): New.
6722         (profile_count::global0adjusted): New.
6723         (profile_count::combine_with_ipa_count): New.
6724         * tree-inline.c (copy_edges_for_bb): Add NUM/DEN arugment;
6725         correct profile of return block of split functions.
6726         (copy_cfg_body): Remove unused profile_count.
6727         (copy_body): Likewise.
6728         (expand_call_inline): Update.
6729         (tree_function_versioning): Update.
6731 2017-11-18  Aldy Hernandez  <aldyh@redhat.com>
6733         * hash-set.h (hash_set::empty): New.
6734         * tree-ssa-threadbackward.h: Delete.
6735         * tree-ssa-threadbackward.c (class thread_jumps): New.
6736         Move max_threaded_paths into class.
6737         (fsm_find_thread_path): Remove arguments that are now in class.
6738         (profitable_jump_thread_path): Rename to...
6739         (thread_jumps::profitable_jump_thread_path): ...this.
6740         (convert_and_register_jump_thread_path): Rename to...
6741         (thread_jumps::convert_and_register_current_path): ...this.
6742         (check_subpath_and_update_thread_path): Rename to...
6743         (thread_jumps::check_subpath_and_update_thread_path): ...this.
6744         (register_jump_thread_path_if_profitable): Rename to...
6745         (thread_jumps::register_jump_thread_path_if_profitable): ...this.
6746         (handle_phi): Rename to...
6747         (thread_jumps::handle_phi): ...this.
6748         (handle_assignment): Rename to...
6749         (thread_jumps::handle_assignment): ...this.
6750         (fsm_find_control_statement_thread_paths): Rename to...
6751         (thread_jumps::fsm_find_control_statement_thread_paths): ...this.
6752         (find_jump_threads_backwards): Rename to...
6753         (thread_jumps::find_jump_threads_backwards): ...this.
6754         Initialize path local data.
6755         (pass_thread_jumps::execute): Call find_jump_threads_backwards
6756         from within thread_jumps class.
6757         (pass_early_thread_jumps::execute): Same.
6759 2017-11-17  Jan Hubicka  <hubicka@ucw.cz>
6761         * cgraph.c (cgraph_node::dump): Do IPA sanity checking on IPA counts.
6763 2017-11-17  Steve Ellcey  <sellcey@cavium.com>
6765         * config/aarch64/aarch64-simd.md (fnma<mode>4): Move neg operator
6766         to canonical location.
6768 2017-11-17  Steve Ellcey  <sellcey@cavium.com>
6770         PR target/81356
6771         * config/aarch64/aarch64.c (aarch64_use_by_pieces_infrastructure_p):
6772         Remove.
6773         (TARGET_USE_BY_PIECES_INFRASTRUCTURE_P): Remove define.
6775 2017-11-17  H.J. Lu  <hongjiu.lu@intel.com>
6777         * config/gnu-user.h (GNU_USER_TARGET_STARTFILE_SPEC): Use
6778         rcrt1.o%s/grcrt1.o%s for -static-pie.
6780 2017-11-17  Jan Hubicka  <hubicka@ucw.cz>
6782         * i386.c (ix86_multiplication_cost, ix86_division_cost,
6783         ix86_shift_rotate_cost): Break out from ...
6784         (ix86_rtx_costs): ... here.
6785         (ix86_add_stmt_cost): Use rtx cost machinery to compute cost of
6786         vector operations.
6788 2017-11-17  Jan Hubicka  <hubicka@ucw.cz>
6790         * predict.c (determine_unlikely_bbs): Set cgraph node count to 0
6791         when entry block was promoted unlikely.
6792         (estimate_bb_frequencies): Increase frequency scale.
6793         * profile-count.h (profile_count): Export precision info.
6795 2017-11-17  Jan Hubicka  <hubicka@ucw.cz>
6797         * tree-tailcall.c (eliminate_tail_call): Be more careful about not
6798         disturbin profile of entry block.
6800 2017-11-17  Jan Hubicka  <hubicka@ucw.cz>
6802         * ipa-fnsummary.c (estimate_node_size_and_time): Be more tolerant for
6803         roundoff errors.
6805 2017-11-17  Jan Hubicka  <hubicka@ucw.cz>
6807         * ipa-cp.c (update_profiling_info): Handle conversion to local profile.
6808         * tree-cfg.c (execute_fixup_cfg): Do fixup same way as inliner does.
6810 2017-11-17  Jeff Law  <law@redhat.com>
6812         * gimple-ssa-evrp.c (class evrp_range_analyzer): New class extracted
6813         from evrp_dom_walker class.  Various methods moved into new class.
6814         (evrp_range_analyzer::evrp_range_analyzer): Constructor for new class.
6815         (evrp_range_analyzer::enter): New method.
6816         (evrp_range_analyzer::leave): New method.
6817         (evrp_dom_walker): Remove delegators no longer needed by this class.
6818         Replace vr_values data member with evrp_range_analyzer
6820         * gimple-ssa-evrp.c (evrp_dom_walker::record_ranges_from_phis): New
6821         method extracted from evrp_dom_walker::before_dom_children.
6822         (evrp_dom_walker::record_ranges_from_stmt): Likewise.
6823         (evrp_dom_walker::record_ranges_from_incoming_edge): Likewise.
6825         * gimple-ssa-evrp.c (evrp_dom_walker): Add cleanup method.
6826         Add private copy constructor and move assignment operators.
6827         Privatize methods and class data where trivially possible.
6828         (evrp_dom_walker::cleanup): New function, extracted from
6829         execute_early_vrp.  Simplify access to class data.
6831         * vr-values.h (get_output_for_vrp): Prototype.
6832         * vr-values.c (get_output_for_vrp): New function extracted from
6833         vrp_visit_assignment_or_call and extract_range_from_stmt.
6834         (vrp_visit_assignment_or_call): Use get_output_for_vrp.  Simplify.
6836 2017-11-17  Luis Machado  <luis.machado@linaro.org>
6838         * config/aarch64/aarch64.c
6839         (qdf24xx_prefetch_tune) <default_opt_level>: Set to -1.
6840         (qdf24xx_tunings) <autoprefetcher_model>: Set to
6841         tune_params::AUTOPREFETCHER_WEAK.
6843 2017-11-17  Tamar Christina  <tamar.christina@arm.com>
6845         PR target/82641
6846         * config/arm/arm.c (arm_valid_target_attribute_rec):
6847         Parse "arch=" and "+<ext>".
6848         (arm_valid_target_attribute_tree): Re-init global options.
6849         (arm_option_override): Make non-static.
6850         (arm_options_perform_arch_sanity_checks): Make errors fatal.
6851         * gcc/config/arm/arm-c.c (__ARM_FEATURE_CMSE): Support undef.
6852         (__ARM_FEATURE_CRC32): Support undef.
6853         * config/arm/arm_acle.h (__ARM_FEATURE_CRC32): Replace with pragma.
6854         * doc/extend.texi (ARM Function Attributes): Add pragma and target.
6856 2017-11-17  David Malcolm  <dmalcolm@redhat.com>
6858         * gdbinit.in (break-on-diagnostic): New command.
6860 2017-11-17  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
6862         * config/i386/i386.c (ix86_expand_epilogue): Change simple
6863         return to indirect jump for EH return if control-flow
6864         protection is enabled. Change explicit 'false' argument in
6865         pro_epilogue_adjust_stack with a value of
6866         flag_cf_protection.
6867         * config/i386/i386.md (simple_return_indirect_internal):
6868         Remove SImode restriction to support 64-bit.
6870 2017-11-17  Segher Boessenkool  <segher@kernel.crashing.org>
6872         * combine.c (added_notes_insn): New.
6873         (try_combine): Handle added_notes_insn like added_links_insn.
6874         Rewrite return value code.
6875         (distribute_notes): Set added_notes_insn to the earliest insn we added
6876         a note to.
6878 2017-11-17  Segher Boessenkool  <segher@kernel.crashing.org>
6880         PR rtl-optimization/82621
6881         * combine.c (try_combine): Do not split PARALLELs of two SETs if the
6882         dest of one of those SETs is unused.
6884 2017-11-17  Richard Biener  <rguenther@suse.de>
6886         PR fortran/83017
6887         * tree-core.h (enum annot_expr_kind): Add annot_expr_parallel_kind.
6888         * tree-pretty-print.c (dump_generic_node): Handle
6889         annot_expr_parallel_kind.
6890         * tree-cfg.c (replace_loop_annotate_in_block): Likewise.
6891         * gimplify.c (gimple_boolify): Likewise.
6893 2017-11-17  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
6895         * config.gcc (extra_headers): Add cet.h for x86 targets.
6896         * config/i386/cet.h: New file.
6897         * doc/install.texi: Add --enable-cet/--disable-cet.
6899 2017-11-17  Richard Biener  <rguenther@suse.de>
6901         PR tree-optimization/83017
6902         * tree-parloops.c (MIN_PER_THREAD): Use --param parloops-min-per-thread.
6903         (gen_parallel_loop): Properly count iterations.
6904         (parallelize_loops): Handle loop->can_be_parallel independent
6905         of flag_loop_parallelize_all.  Make static profitability test match
6906         the runtime one.
6907         * params.def (PARAM_PARLOOPS_MIN_PER_THREAD): New.
6908         * invoke.texi (parloops-min-per-thread): Document.
6910 2017-11-17  Vineet Gupta  <vgupta@synopsys.com>
6912         * config/arc/linux.h: GLIBC_DYNAMIC_LINKER update per glibc
6913         upstreaming review comments.
6915 2017-11-17  Tamar Christina  <tamar.christina@arm.com>
6917         * expr.c (copy_blkmode_to_reg): Fix bitsize for targets
6918         with fast unaligned access.
6919         * doc/sourcebuild.texi (word_mode_no_slow_unalign): New.
6921 2017-11-17  Thomas Preud'homme  <thomas.preudhomme@arm.com>
6923         * config/arm/arm.c (cmse_nonsecure_entry_clear_before_return): Allocate
6924         to_clear_arg_regs_bitmap to the same size as to_clear_bitmap.
6926 2017-11-17  Richard Biener  <rguenther@suse.de>
6928         * tree-ssa-pre.c (phi_translate_1): Remove redundant constant
6929         folding of references.
6931 2017-11-17  Qing Zhao  <qing.zhao@oracle.com>
6933         PR middle-end/78809
6934         * gimple-fold.c (gimple_fold_builtin_string_compare): Add handling
6935         of replacing call to strncmp with corresponding call to strcmp when
6936         meeting conditions.
6938 2017-11-17  Sergey Shalnov  <Sergey.Shalnov@intel.com>
6940         * config/i386/x86-tune.def (X86_TUNE_AVX256_OPTIMAL): Add tuning
6941         option prefer-avx256 for skylake-avx512 configuration.
6942         * config/i386/i386.c (ix86_option_override_internal): Ditto.
6943         (get_builtin_code_for_version): Ditto.
6945 2017-11-17  Chung-Ju Wu  <jasonwucj@gmail.com>
6946             Monk Chiang  <sh.chiang04@gmail.com>
6948         * config/nds32/nds32.h (FIRST_PSEUDO_REGISTER): Modify.
6949         (FIXED_REGISTERS): Reserve more register numbers.
6950         (CALL_USED_REGISTERS): Likewise.
6951         (REG_ALLOC_ORDER): Likewise.
6952         (REG_CLASS_CONTENTS): Likewise.
6953         (REGISTER_NAMES): Likewise.
6955 2017-11-17  Chung-Ju Wu  <jasonwucj@gmail.com>
6956             Kito Cheng  <kito.cheng@gmail.com>
6958         * config/nds32/nds32-modes.def: Add vector mode V4QI V2HI V8QI V4HI
6959         V2SI.
6960         * config/nds32/iterators.md: Add vector mode iterators and attributes.
6962 2017-11-16  Steven Munroe  <munroesj@gcc.gnu.org>
6964         * config/rs6000/mmintrin.h (_mm_add_pi32[_ARCH_PWR]): Correct
6965         parameter list for vec_splats.
6967 2017-11-16  Joseph Myers  <joseph@codesourcery.com>
6969         * doc/invoke.texi (-std=c17): Refer to 2018 expected publication
6970         date of C17.
6971         (-std=c18, -std=iso9899:2018, -std=gnu18): Document option aliases.
6973 2017-11-16  Thomas Koenig  <tkoenig@gcc.gnu.org>
6975         PR bootstrap/82856
6976         * doc/install.texi: Document incompatibility of Perl >=5.6.26
6977         with the required version of automake 1.11.6.
6979 2017-11-16  Pat Haugen  <pthaugen@us.ibm.com>
6981         * rs6000/power9.md (power9fpdiv): New automaton and cpu_unit defined
6982         for it.
6983         (DU_C2_3_power9): Correct reservation combinations.
6984         (FP_DIV_power9, VEC_DIV_power9): New.
6985         (power9-alu): Split out rotate/shift...
6986         (power9-rot): ...to here, correct dispatch resource.
6987         (power9-cracked-alu, power9-mul, power9-mul-compare): Correct dispatch
6988         resource.
6989         (power9-fp): Correct latency.
6990         (power9-sdiv): Add div/sqrt resource.
6991         (power9-ddiv): Correct latency, add div/sqrt resource.
6992         (power9-sqrt, power9-dsqrt): Add div/sqrt resource.
6993         (power9-vecfdiv, power9-vecdiv): Correct latency, add div/sqrt
6994         resource.
6995         (power9-qpdiv, power9-qpmul): Adjust resource usage.
6997 2017-11-15  Michael Meissner  <meissner@linux.vnet.ibm.com>
6999         * config/rs6000/rs6000.c (rs6000_expand_builtin): Do not do the
7000         switch statement mapping KF built-ins to TF built-ins if we don't
7001         have the proper ISA 3.0 assembler support.
7003 2017-11-16  Jan Hubicka  <hubicka@ucw.cz>
7005         * tree-emutls.c (lower_emutls_data): Remove unused bb_freq.
7006         (lower_emutls_function_body): Do not compute it.
7008 2017-11-16  Jan Hubicka  <hubicka@ucw.cz>
7010         * ipa-split.c (split_bb_info): Turn time to sreal.
7011         (split_point): Likewise.
7012         (dump_split_point): Likewise.
7013         (fine_split_points): Likewise.
7014         (execute_split_functions): Only zero split_bbs; turn time to sreals.
7016 2017-11-16  Jan Hubicka  <hubicka@ucw.cz>
7018         * ipa-fnsummary.c (analyze_function_body): Accumulate time consistently
7019         in sreal.
7021 2017-11-16  Jan Hubicka  <hubicka@ucw.cz>
7023         * predict.c (combine_predictions_for_bb): Preserve zero predicted
7024         edges.
7025         (expensive_function_p): Remove useless assert.
7026         (determine_unlikely_bbs): Propagate also forward; determine cold blocks
7028 2017-11-16  Martin Sebor  <msebor@redhat.com>
7030         PR tree-optimization/82588
7031         PR tree-optimization/82583
7032         * tree-vrp.c (check_array_ref): Handle flexible array members,
7033         string literals, and inner indices.
7034         (search_for_addr_array): Add detail to diagnostics.
7036 2017-11-16  Nathan Sidwell  <nathan@acm.org>
7038         PR c++/82836
7039         PR c++/82737
7040         * tree.h (COPY_DECL_RTL): Rename parms for clarity.
7041         (SET_DECL_ASSEMBLER_NAME): Forward to
7042         overwrite_decl_assembler_name.
7043         (COPY_DECL_ASSEMBLER_NAME): Rename parms for clarity.
7044         (overwrite_decl_assembler_name): Declare.
7045         * tree.c (overwrite_decl_assembler_name): New.
7046         * langhooks-def.h (lhd_overwrite_decl_assembler_name): Declare.
7047         (LANG_HOOKS_OVERWRITE_DECL_ASSEMBLER_NAME): Provide default.
7048         (LANG_HOOKS_INITIALIZER): Add it.
7049         * langhooks.h (struct lang_hooks): Add overwrite_decl_assembler_name.
7050         * langhooks.c (lhd_set_decl_assembler_name): Use
7051         SET_DECL_ASSEMBLER_NAME.
7052         (lhd_overwrite_decl_assembler_name): Default implementation.
7054 2017-11-16  Wilco Dijkstra  <wdijkstr@arm.com>
7055             Jackson Woodruff  <jackson.woodruff@arm.com>
7057         PR tree-optimization/71026
7058         * match.pd: Canonicalize constant multiplies in division.
7060 2017-11-16  Wilco Dijkstra  <wdijkstr@arm.com>
7062         * opts.c (default_options_table): Add OPT_fomit_frame_pointer entry.
7063         * common/config/alpha/alpha-common.c (TARGET_OPTION_OPTIMIZATION_TABLE):
7064         Remove OPT_fomit_frame_pointer entry.
7065         * common/config/arc/arc-common.c: Likewise.
7066         * common/config/arm/arm-common.c: Likewise.
7067         * common/config/avr/avr-common.c: Likewise.
7068         * common/config/c6x/c6x-common.c: Likewise.
7069         * common/config/cr16/cr16-common.c: Likewise.
7070         * common/config/cris/cris-common.c: Likewise.
7071         * common/config/epiphany/epiphany-common.c: Likewise.
7072         * common/config/fr30/fr30-common.c: Likewise.
7073         * common/config/frv/frv-common.c: Likewise.
7074         * common/config/ia64/ia64-common.c: Likewise.
7075         * common/config/iq2000/iq2000-common.c: Likewise.
7076         * common/config/lm32/lm32-common.c: Likewise.
7077         * common/config/m32r/m32r-common.c: Likewise.
7078         * common/config/mcore/mcore-common.c: Likewise.
7079         * common/config/microblaze/microblaze-common.c: Likewise.
7080         * common/config/mips/mips-common.c: Likewise.
7081         * common/config/mmix/mmix-common.c: Likewise.
7082         * common/config/mn10300/mn10300-common.c: Likewise.
7083         * common/config/nios2/nios2-common.c: Likewise.
7084         * common/config/pa/pa-common.c: Likewise.
7085         * common/config/pdp11/pdp11-common.c: Likewise.
7086         * common/config/powerpcspe/powerpcspe-common.c: Likewise.
7087         * common/config/riscv/riscv-common.c: Likewise.
7088         * common/config/rs6000/rs6000-common.c: Likewise.
7089         * common/config/rx/rx-common.c: Likewise.
7090         * common/config/s390/s390-common.c: Likewise.
7091         * common/config/sh/sh-common.c: Likewise.
7092         * common/config/sparc/sparc-common.c: Likewise.
7093         * common/config/tilegx/tilegx-common.c: Likewise.
7094         * common/config/tilepro/tilepro-common.c: Likewise.
7095         * common/config/v850/v850-common.c: Likewise.
7096         * common/config/visium/visium-common.c: Likewise.
7097         * common/config/xstormy16/xstormy16-common.c: Likewise.
7098         * common/config/xtensa/xtensa-common.c: Likewise.
7099         * invoke.texi (-fomit-frame-pointer): Update documentation.
7101 2017-11-16  Jan Hubicka  <hubicka@ucw.cz>
7103         * tree-cfg.c (gimple_find_sub_bbs): Do not compute freq.
7105 2017-11-14  Jan Hubicka  <hubicka@ucw.cz>
7107         * cfg.c (scale_bbs_frequencies_int,
7108         cale_bbs_frequencies_gcov_type): Remove.
7109         * cfg.h (scale_bbs_frequencies_int,
7110         cale_bbs_frequencies_gcov_type): Remove.
7112 2017-11-14  Jan Hubicka  <hubicka@ucw.cz>
7114         * tree-ssa-loop-manip.c
7115         (scale_dominated_blocks_in_loop): Update to profile counts.
7116         (tree_transform_and_unroll_loop): Likewise.
7118 2017-11-14  Jan Hubicka  <hubicka@ucw.cz>
7120         * tree-vect-loop-manip.c (vect_do_peeling): Do not use
7121         scale_bbs_frequencies_int.
7123 2017-11-14  Jan Hubicka  <hubicka@ucw.cz>
7125         * final.c (compute_alignments): Use counts rather than frequencies.
7127 2017-11-14  Jan Hubicka  <hubicka@ucw.cz>
7129         * cfgloopanal.c: Include sreal.h
7130         (average_num_loop_insns): Use counts and sreal for accounting.
7132 2017-11-14  Jan Hubicka  <hubicka@ucw.cz>
7134         * cfgloopmanip.c (duplicate_loop_to_header_edge): Cleanup profile
7135         manipulation.
7137 2017-11-16  Chung-Ju Wu  <jasonwucj@gmail.com>
7138             Kito Cheng  <kito.cheng@gmail.com>
7140         * config/nds32/constraints.md: Provide more constraints.
7141         * config/nds32/nds32.h (enum reg_class, REG_CLASS_NAMES,
7142         REG_CLASS_CONTENTS): Define R5_REG, R8_REG, and FRAME_POINTER_REG to
7143         support constraints usage.
7145 2017-11-16  Chung-Ju Wu  <jasonwucj@gmail.com>
7147         * config/nds32/constants.md (UNSPEC_VOLATILE_FUNC_RETURN): Remove.
7149 2017-11-16  Chung-Ju Wu  <jasonwucj@gmail.com>
7150             Kito Cheng  <kito.cheng@gmail.com>
7152         * config/nds32/nds32.opt: Add mext-perf, mext-perf2, mext-string.
7153         * config/nds32/nds32.opt: Refine the layout.
7154         * config/nds32/nds32.c (TARGET_EXT_PERF, TARGET_EXT_PERF2,
7155         TARGET_EXT_STRING): Support new options.
7156         * config/nds32/nds32.h: Likewise.
7157         * config/nds32/nds32.md: Likewise.
7158         * config/nds32/nds32-predicates.c: Likewise.
7159         * config/nds32/constraints.md: Likewise.
7160         * common/config/nds32/nds32-common.c: Likewise.
7162 2017-11-16  Julia Koval  <julia.koval@intel.com>
7164         PR target/82983
7165         * config/i386/gfniintrin.h: Add sse check.
7166         * config/i386/i386.c (ix86_expand_builtin): Fix gfni check.
7168 2017-11-16  Julia Koval  <julia.koval@intel.com>
7170         * common/config/i386/i386-common.c (OPTION_MASK_ISA_AVX512VBMI2_SET,
7171         OPTION_MASK_ISA_AVX512VBMI2_UNSET): New.
7172         (ix86_handle_option): Handle -mavx512vbmi2.
7173         * config/i386/cpuid.h: Add bit_AVX512VBMI2.
7174         * config/i386/driver-i386.c (host_detect_local_cpu): Handle new bit.
7175         * config/i386/i386-c.c (__AVX512VBMI2__): New.
7176         * config/i386/i386.c (ix86_target_string): Handle -mavx512vbmi2.
7177         (ix86_valid_target_attribute_inner_p): Ditto.
7178         * config/i386/i386.h (TARGET_AVX512VBMI2, TARGET_AVX512VBMI2_P): New.
7179         * config/i386/i386.opt (mavx512vbmi2): New option.
7180         * doc/invoke.texi: Add new option.
7182 2017-11-16  Julia Koval  <julia.koval@intel.com>
7184         * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8, _mm256_gf2p8mul_epi8,
7185         _mm_mask_gf2p8mul_epi8, _mm_maskz_gf2p8mul_epi8,
7186         _mm256_mask_gf2p8mul_epi8, _mm256_maskz_gf2p8mul_epi8,
7187         _mm512_mask_gf2p8mul_epi8, _mm512_maskz_gf2p8mul_epi8,
7188         _mm512_gf2p8mul_epi8): New intrinsics.
7189         * config/i386/i386-builtin-types.def
7190         (V64QI_FTYPE_V64QI_V64QI): New type.
7191         * config/i386/i386-builtin.def (__builtin_ia32_vgf2p8mulb_v64qi,
7192         __builtin_ia32_vgf2p8mulb_v64qi_mask, __builtin_ia32_vgf2p8mulb_v32qi,
7193         __builtin_ia32_vgf2p8mulb_v32qi_mask, __builtin_ia32_vgf2p8mulb_v16qi,
7194         __builtin_ia32_vgf2p8mulb_v16qi_mask): New builtins.
7195         * config/i386/sse.md (vgf2p8mulb_*): New pattern.
7196         * config/i386/i386.c (ix86_expand_args_builtin): Handle new type.
7198 2017-11-15  Uros Bizjak  <ubizjak@gmail.com>
7200         * config/i386/i386.c (x86_print_call_or_nop): Emit 5 byte nop
7201         explicitly as a stream of bytes.
7203 2017-11-15  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
7205         * config/rs6000/altivec.h (vec_xst_be): New #define.
7206         * config/rs6000/altivec.md (altivec_vperm_<mode>_direct): Rename
7207         and externalize from *altivec_vperm_<mode>_internal.
7208         * config/rs6000/rs6000-builtin.def (XL_BE_V16QI): Remove macro
7209         instantiation.
7210         (XL_BE_V8HI): Likewise.
7211         (XL_BE_V4SI): Likewise.
7212         (XL_BE_V4SI): Likewise.
7213         (XL_BE_V2DI): Likewise.
7214         (XL_BE_V4SF): Likewise.
7215         (XL_BE_V2DF): Likewise.
7216         (XST_BE): Add BU_VSX_OVERLOAD_X macro instantiation.
7217         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Correct
7218         all array entries with these keys: VSX_BUILTIN_VEC_XL,
7219         VSX_BUILTIN_VEC_XL_BE, VSX_BUILTIN_VEC_XST.  Add entries for key
7220         VSX_BUILTIN_VEC_XST_BE.
7221         * config/rs6000/rs6000.c (altivec_expand_xl_be_builtin): Remove.
7222         (altivec_expand_builtin): Remove handling for VSX_BUILTIN_XL_BE_*
7223         built-ins.
7224         (altivec_init_builtins): Replace conditional calls to def_builtin
7225         for __builtin_vsx_ld_elemrev_{v8hi,v16qi} and
7226         __builtin_vsx_st_elemrev_{v8hi,v16qi} based on TARGET_P9_VECTOR
7227         with unconditional calls.  Remove calls to def_builtin for
7228         __builtin_vsx_le_be_<mode>.  Add a call to def_builtin for
7229         __builtin_vec_xst_be.
7230         * config/rs6000/vsx.md (vsx_ld_elemrev_v8hi): Convert define_insn
7231         to define_expand, and add alternate RTL generation for P8.
7232         (*vsx_ld_elemrev_v8hi_internal): New define_insn based on
7233         vsx_ld_elemrev_v8hi.
7234         (vsx_ld_elemrev_v16qi): Convert define_insn to define_expand, and
7235         add alternate RTL generation for P8.
7236         (*vsx_ld_elemrev_v16qi_internal): New define_insn based on
7237         vsx_ld_elemrev_v16qi.
7238         (vsx_st_elemrev_v8hi): Convert define_insn
7239         to define_expand, and add alternate RTL generation for P8.
7240         (*vsx_st_elemrev_v8hi_internal): New define_insn based on
7241         vsx_st_elemrev_v8hi.
7242         (vsx_st_elemrev_v16qi): Convert define_insn to define_expand, and
7243         add alternate RTL generation for P8.
7244         (*vsx_st_elemrev_v16qi_internal): New define_insn based on
7245         vsx_st_elemrev_v16qi.
7247 2017-11-15  H.J. Lu  <hongjiu.lu@intel.com>
7249         PR target/82990
7250         * config/i386/i386.c (pass_insert_vzeroupper::gate): Remove
7251         TARGET_AVX512ER check.
7252         (ix86_option_override_internal): Set MASK_VZEROUPPER if
7253         neither -mzeroupper nor -mno-zeroupper is used and
7254         TARGET_EMIT_VZEROUPPER is set.
7255         * config/i386/i386.h (TARGET_EMIT_VZEROUPPER): New.
7256         * config/i386/x86-tune.def: Add X86_TUNE_EMIT_VZEROUPPER.
7258 2017-11-15  Will Schmidt  <will_schmidt@vnet.ibm.com>
7260         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add support for
7261         folding of vector compares.
7262         (fold_build_vec_cmp): New helper function.
7263         (fold_compare_helper): New helper function.
7264         (builtin_function_type): Add compare builtins to the list of functions
7265         having unsigned arguments.  Cosmetic updates to comment indentation.
7266         * config/rs6000/vsx.md (vcmpneb, vcmpneh, vcmpnew): Update to specify
7267         the not+eq combination.
7269 2017-11-15  Bin Cheng  <bin.cheng@arm.com>
7271         PR tree-optimization/82726
7272         PR tree-optimization/70754
7273         * tree-predcom.c (order_drefs_by_pos): New function.
7274         (combine_chains): Move code setting has_max_use_after to...
7275         (try_combine_chains): ...here.  New parameter.  Sort combined chains
7276         according to position information.
7277         (tree_predictive_commoning_loop): Update call to above function.
7278         (update_pos_for_combined_chains, pcom_stmt_dominates_stmt_p): New.
7280 2017-11-15  Bin Cheng  <bin.cheng@arm.com>
7282         PR tree-optimization/82726
7283         Revert
7284         2017-01-23  Bin Cheng  <bin.cheng@arm.com>
7286         PR tree-optimization/70754
7287         * tree-predcom.c (stmt_combining_refs): New parameter INSERT_BEFORE.
7288         (reassociate_to_the_same_stmt): New parameter INSERT_BEFORE.  Insert
7289         combined stmt before it if not NULL.
7290         (combine_chains): Process refs reversely and compute dominance point
7291         for root ref.
7293         Revert
7294         2017-02-23  Bin Cheng  <bin.cheng@arm.com>
7296         PR tree-optimization/79663
7297         * tree-predcom.c (combine_chains): Process refs in reverse order
7298         only for ZERO length chains, and add explaining comment.
7300 2017-11-15  Tamar Christina  <tamar.christina@arm.com>
7302         * config/arm/arm-cpus.in (armv8_3, ARMv8_3a, armv8.3-a): New
7303         * config/arm/arm-tables.opt (armv8.3-a): Regenerated.
7304         * doc/invoke.texi (ARM Options): Add armv8.3-a.
7306 2017-11-15  Tamar Christina  <tamar.christina@arm.com>
7308         * config/arm/arm.h (TARGET_DOTPROD): Add arm_arch8_2.
7310 2017-11-15  Martin Liska  <mliska@suse.cz>
7312         * tree-cfg.c (pass_warn_function_return::execute):
7313         Compare warn_return_type for greater than zero.
7315 2017-11-15  Sebastian Peryt  <sebastian.peryt@intel.com>
7317         PR target/82941
7318         PR target/82942
7319         * config/i386/i386.c (pass_insert_vzeroupper): Modify gate condition
7320         to return true on Xeon and not on Xeon Phi.
7321         (ix86_check_avx256_register): Changed to ...
7322         (ix86_check_avx_upper_register): ... this. Add extra check for
7323         VALID_AVX512F_REG_OR_XI_MODE.
7324         (ix86_avx_u128_mode_needed): Changed
7325         ix86_check_avx256_register to ix86_check_avx_upper_register.
7326         (ix86_check_avx256_stores): Changed to ...
7327         (ix86_check_avx_upper_stores): ... this. Changed
7328         ix86_check_avx256_register to ix86_check_avx_upper_register.
7329         (ix86_avx_u128_mode_after): Changed
7330         avx_reg256_found to avx_upper_reg_found. Changed
7331         ix86_check_avx256_stores to ix86_check_avx_upper_stores.
7332         (ix86_avx_u128_mode_entry): Changed
7333         ix86_check_avx256_register to ix86_check_avx_upper_register.
7334         (ix86_avx_u128_mode_exit): Ditto.
7335         * config/i386/i386.h: (host_detect_local_cpu): New define.
7337 2017-11-15  Dominik Infuehr  <dominik.infuehr@theobroma-systems.com>
7339         * config/arm/xgene1.md (xgene1): Split into automatons
7340         xgene1_main, xgene1_decoder, xgene1_div, xgene1_simd.
7341         (xgene1_f_load): Adjust reservations and/or types.
7342         (xgene1_f_store): Likewise.
7343         (xgene1_load_pair): Likewise.
7344         (xgene1_store_pair): Likewise.
7345         (xgene1_fp_load1): Likewise.
7346         (xgene1_load1): Likewise.
7347         (xgene1_store1): Likewise.
7348         (xgene1_move): Likewise.
7349         (xgene1_alu): Likewise.
7350         (xgene1_simd): Likewise.
7351         (xgene1_bfm): Likewise.
7352         (xgene1_neon_load1): Likewise.
7353         (xgene1_neon_store1): Likewise.
7354         (xgene1_neon_logic): Likewise.
7355         (xgene1_neon_st1): Likewise.
7356         (xgene1_neon_ld1r): Likewise.
7357         (xgene1_alu_cond): Added.
7358         (xgene1_shift_reg): Likwise.
7359         (xgene1_bfx): Likewise.
7360         (xgene1_mul): Split into xgene1_mul32, xgene1_mul64.
7362 2017-11-15  Jakub Jelinek  <jakub@redhat.com>
7364         PR target/82981
7365         * internal-fn.c: Include gimple-ssa.h, tree-phinodes.h and
7366         ssa-iterators.h.
7367         (can_widen_mult_without_libcall): New function.
7368         (expand_mul_overflow): If only checking unsigned mul overflow,
7369         not result, and can do efficiently MULT_HIGHPART_EXPR, emit that.
7370         Don't use WIDEN_MULT_EXPR if it would involve a libcall, unless
7371         no other way works.  Add MULT_HIGHPART_EXPR + MULT_EXPR support.
7372         (expand_DIVMOD): Formatting fix.
7373         * expmed.h (expand_mult): Add NO_LIBCALL argument.
7374         * expmed.c (expand_mult): Likewise.  Use OPTAB_WIDEN rather
7375         than OPTAB_LIB_WIDEN if NO_LIBCALL is true, and allow it to fail.
7377         PR tree-optimization/82977
7378         * tree-ssa-strlen.c (strlen_optimize_stmt): Pass a reference to a copy
7379         constructed temporary to strlen_to_stridx.put.
7381 2017-11-15  Martin Liska  <mliska@suse.cz>
7383         * configure.ac: Remove -fkeep-inline-functions from coverage_flags.
7384         * configure: Regenerate.
7386 2017-11-15  Martin Liska  <mliska@suse.cz>
7388         PR target/82927
7389         * config/sh/sh-mem.cc: Use proper probability for
7390         REG_BR_PROB_NOTE.
7392 2017-11-14  Jeff Law  <law@redhat.com>
7394         * explow.c (anti_adjust_stack_and_probe_stack_clash): Avoid probing
7395         the red zone for stack_clash_protection_final_dynamic_probe targets
7396         when the total dynamic stack size is zero bytes.
7398         * tree-ssa-threadupdate.c (thread_through_all_blocks): Thread
7399         blocks is post order.
7401 2017-11-15  Alexandre Oliva  <aoliva@redhat.com>
7403         * dumpfile.h (TDF_COMPARE_DEBUG): New.
7404         * final.c (rest_of_clean_state): Set it for the
7405         -fcompare-debug dump.
7406         * tree-pretty-print.c (dump_generic_node): Omit OBJ_TYPE_REF
7407         class when TDF_COMPARE_DEBUG is set.
7409         * dwarf2out.c (gen_producer_string): Discard
7410         OPT_fcompare_debug.
7412 2017-11-15  Joseph Myers  <joseph@codesourcery.com>
7414         PR c/81156
7415         * doc/extend.texi (Other Builtins): Document __builtin_tgmath.
7416         * ginclude/tgmath.h (__tg_cplx, __tg_ldbl, __tg_dbl, __tg_choose)
7417         (__tg_choose_2, __tg_choose_3, __TGMATH_REAL_1_2)
7418         (__TGMATH_REAL_2_3): Remove macros.
7419         (__TGMATH_CPLX, __TGMATH_CPLX_2, __TGMATH_REAL, __TGMATH_REAL_2)
7420         (__TGMATH_REAL_3, __TGMATH_CPLX_ONLY): Define using
7421         __builtin_tgmath.
7422         (frexp, ldexp, nexttoward, scalbn, scalbln): Define using
7423         __TGMATH_REAL_2.
7424         (remquo): Define using __TGMATH_REAL_3.
7426 2017-11-14  Jeff Law  <law@redhat.com>
7428         * vr-values.c: New file with contents extracted from tree-vrp.c.
7429         * Makefile.in (OBJS): Add vr-values.o
7430         * tree-vrp.h (set_value_range_to_nonnull): Prototype.
7431         (set_value_range, set_and_canonicalize_value_range): Likewise.
7432         (vrp_bitmap_equal_p, range_is_nonnull): Likewise.
7433         (value_range_constant_singleton, symbolic_range_p): Likewise.
7434         (compare_values, compare_values_warnv, vrp_val_is_min): Likewise.
7435         (vrp_val_is_max, copy_value_range, set_value_range_to_value): Likewise.
7436         (extract_range_from_binary_expr_1, vrp_val_min, vrp_val_max): Likewise.
7437         (set_value_range_to_null, range_int_cst_p, opreand_less_p): Likewise.
7438         (find_case_label_range, find_case_label_index): Likewise.
7439         (zero_nonzero_bits_from_vr, overflow_comparison_p): Likewise.
7440         (range_int_cst_singleton_p, value_inside_range): Likewise.
7441         (get_single_symbol): Likewise.
7442         (switch_update): Move structure definition here.
7443         (to_remove_edges, to_update_switch_stmts): Provide externs.
7444         * tree-vrp.c: Move all methods for vr-values class to vr-values.c
7445         (vrp_val_max, vrp_val_min, vrp_val_is_max): Make externally visible.
7446         (vrp_val_is_min, set_value_range): Likewise.
7447         (set_and_canonicalize_value_range, copy_value_range): Likewise.
7448         (set_value_range_to_value, set_value_range_to_nonnull): Likewise.
7449         (set_value_range_to_null, vrp_bitmap_equal_p): Likewise.
7450         (range_is_nonnull, range_int_cst_p): Likewwise.
7451         (range_int_cst_singleton_p, symbolic_range_p): Likewise.
7452         (get_single_symbol, operand_less_p): Likewise
7453         (compare_values_warnv, compare_values): Likewise.
7454         (value_inside_range, value_range_constant_singleton): Likewise.
7455         (zero_nonzero_bitgs_from_vr): Likewise.
7456         (extract_range_from_binary_expr_1): Likewise.
7457         (overflow_comparison_p): Likewise.
7458         (to_remove_edges, to_update_switch_stmts): Likewise.
7459         (find_case_label-index, find_case_label_range): Likewise.
7460         (switch_update, set_value_range_to_nonnegative): Remove.
7461         (set_value_range_to_truthvalue): Likewise.
7462         (symbolic_range_based_on_p, gimple_assign_nonzero_p): Likewise.
7463         (gimple_stmt_nonzero_p, compare_ranges): Likewise.
7464         (compare_range_with_value, vrp_valueize, vrp_valueize_1): Likewise.
7465         (find_case_label_ranges, test_for_singularity): Likewise.
7466         (range_fits_type_p, simplify_conversion_using_ranges): LIkewise.
7467         (x_vr_values): Move to its remaining use site.
7469 2017-11-10  Jeff Law  <law@redhat.com>
7471         * vr-values.h (VR_INITIALIZER): Move #define here.
7472         * gimple-ssa-evrp.c: New file with contents extracted from tree-vrp.c
7473         * Makefile.in (OBJS): Add tree-evrp.o
7474         * tree-vrp.h (assert_info): Move structure definition here.
7475         (set_value_range_to_varying): Prototype.
7476         (vrp_operand_equal_p, range_includes_zero_p): Likewise.
7477         (infer_value_range, register_edge_assert_for): Likewise.
7478         (stmt_interesting_for_vrp): Likewise.
7479         * tree-vrp.c: Move all methods for evrp class into gimple-ssa-evrp.c.
7480         (set_value_range_to_varying): No longer static.
7481         (vrp_operand_equal_p, range_includes_zero_p): Likewise.
7482         (infer_value_range, register_edge_assert_for): Likewise.
7484 2017-11-14  Michael Meissner  <meissner@linux.vnet.ibm.com>
7486         * config/rs6000/rs6000.md (bswapdi2): On 32-bit ISA 3.0, don't
7487         generate the XXBRD instruction.
7489         * config/rs6000/rs6000-c.c (is_float128_p): New helper function.
7490         (rs6000_builtin_type_compatible): Treat _Float128 and long double
7491         as being compatible if -mabi=ieeelongdouble.
7492         * config/rs6000/rs6000-builtin.def (BU_FLOAT128_HW_1): New macros
7493         to setup float128 built-ins with hardware support.
7494         (BU_FLOAT128_HW_2): Likewise.
7495         (BU_FLOAT128_HW_3): Likewise.
7496         (BU_FLOAT128_HW_VSX_1): Likewise.
7497         (BU_FLOAT128_HW_VSX_2): Likewise.
7498         (scalar_extract_expq): Change float128 built-in functions to
7499         accommodate having both KFmode and TFmode functions.  Use the
7500         KFmode variant as the default.
7501         (scalar_extract_sigq): Likewise.
7502         (scalar_test_neg_qp): Likewise.
7503         (scalar_insert_exp_q): Likewise.
7504         (scalar_insert_exp_qp): Likewise.
7505         (scalar_test_data_class_qp): Likewise.
7506         (sqrtf128_round_to_odd): Delete processing the round to odd
7507         built-in functions as special built-in functions, and define them
7508         as float128 built-ins.  Use the KFmode variant as the default.
7509         (truncf128_round_to_odd): Likewise.
7510         (addf128_round_to_odd): Likewise.
7511         (subf128_round_to_odd): Likewise.
7512         (mulf128_round_to_odd): Likewise.
7513         (divf128_round_to_odd): Likewise.
7514         (fmaf128_round_to_odd): Likewise.
7515         * config/rs6000/rs6000.c (rs6000_expand_binop_builtin): Add
7516         support for KFmode and TFmode xststdcqp calls.
7517         (rs6000_expand_builtin): If long double is IEEE 128-bit floating
7518         point, switch the built-in handlers for the get/set float128
7519         exponent, get float128 mantissa, float128 test built-ins, and the
7520         float128 round to odd built-in functions.  Eliminate creating the
7521         float128 round to odd built-in functions as special built-ins.
7522         (rs6000_init_builtins): Eliminate special creation of the float128
7523         round to odd built-in functions.
7524         * config/rs6000/vsx.md (xsxexpqp_<mode>): Change float128 built-in
7525         function insns to support both TFmode and KFmode variants.
7526         (xsxsigqp_<mode>): Likewise.
7527         (xsiexpqpf_<mode>): Likewise.
7528         (xsiexpqp_<mode>): Likewise.
7529         (xststdcqp_<mode>): Likewise.
7530         (xststdcnegqp_<mode>): Likewise.
7531         (xststdcqp_<mode>): Likewise.
7533 2017-11-14  Jan Hubicka  <hubicka@ucw.cz>
7535         * tree-ssa-threadupdate.c (compute_path_counts): Remove
7536         unused path_in_freq_ptr parameter.
7537         (ssa_fix_duplicate_block_edges): Do not pass around path_in_freq
7539 2017-11-14  Jan Hubicka  <hubicka@ucw.cz>
7541         * ipa-inline.c (edge_badness): Dump sreal frequency.
7542         (compute_inlined_call_time): Match natural implementaiton ...
7543         * ipa-fnsummary.c (estimate_edge_size_and_time): ... here; remove
7544         forgotten division by CGRAPH_FREQ_BASE.
7546 2017-11-14  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
7548         * config.gcc (*-*-solaris2*): Enable default_use_cxa_atexit since
7549         Solaris 11.  Update comment.
7550         * configure.ac (gcc_cv_ld_pid): Adapt comment for Solaris 12
7551         renaming.
7552         * config/sol2.h (STARTFILE_SPEC): Likewise.
7553         * configure: Regenerate.
7555 2017-11-14  Carl Love  <cel@us.ibm.com>
7557         * config/rs6000/rs6000.c (swap_endian_selector_for_mode): Remove
7558         le_ and be_ prefixes to swap* variables.  Remove
7559         if (VECTOR_ELT_ORDER_BIG) statement. Remove E_V16QImode case
7560         statements.
7562 2017-11-14  Jason Merrill  <jason@redhat.com>
7564         Support GTY((cache)) on hash_map.
7565         * hash-traits.h (ggc_remove): Add ggc_maybe_mx member function.
7566         (ggc_cache_remove): Override it instead of ggc_mx.
7567         * hash-table.h (gt_ggc_mx): Call it instead of ggc_mx.
7568         (gt_cleare_cache): Call ggc_mx instead of gt_ggc_mx.
7569         * hash-map-traits.h (simple_hashmap_traits): Add maybe_mx member.
7570         (simple_cache_map_traits): Override maybe_mx.
7571         * hash-map.h (hash_entry): Add ggc_maybe_mx and keep_cache_entry.
7572         (hash_map): Friend gt_cleare_cache.
7573         (gt_cleare_cache): New.
7574         * tree.h (tree_cache_traits): New hash_map traits class.
7575         (tree_cache_map): New typedef.
7577 2017-11-14  Richard Biener  <rguenther@suse.de>
7579         * tree-cfgcleanup.c (cleanup_control_expr_graph): Remove first_p
7580         paramter and handling.
7581         (cleanup_control_flow_bb): Likewise.
7582         (cleanup_control_flow_pre): New helper performing a DFS walk
7583         to call cleanup_control_flow_bb in PRE order.
7584         (cleanup_tree_cfg_1): Do the first phase of cleanup_control_flow_bb
7585         via cleanup_control_flow_pre.
7587 2017-11-14  James Greenhalgh  <james.greenhalgh@arm.com>
7589         * config/aarch64/aarch64-simd.md
7590         (aarch64_simd_bsl<mode>_internal): Remove DImode.
7591         (*aarch64_simd_bsl<mode>_alt): Likewise.
7592         (aarch64_simd_bsldi_internal): New.
7593         (aarch64_simd_bsldi_alt): Likewise.
7595 2017-11-13  Jan Hubicka  <hubicka@ucw.cz>
7597         * tracer.c (better_p): Do not compare frequencies.
7598         * reg-stack.c (better_edge): Likewise.
7599         * shrink-wrap.c (try_shrink_wrapping): Do not convert to gcov counts
7600         and back.
7602 2017-11-13  Jan Hubicka  <hubicka@ucw.cz>
7604         * auto-profile.c (afdo_annotate_cfg): Use update_max_bb_count.
7605         * cgraphunit.c (cgraph_node::expand_thunk): Use update_max_bb_count.
7606         * ipa-utils.c (ipa_merge_profiles): Use update_max_bb_count.
7607         * lto-streamer-in.c (input_function): Use update_max_bb_count.
7608         * omp-expand.c (expand_omp_taskreg): Use update_max_bb_count.
7609         * predict.c (maybe_hot_frequency_p): Inline to ...
7610         (maybe_hot_count_p): ... here; rewrite to counts.
7611         (counts_to_freqs): Rename to ...
7612         (update_max_bb_count): ... this one.
7613         (expensive_function_p): Use counts.
7614         (estimate_bb_frequencies): Update.
7615         (rebuild_frequencies): Update.
7616         * predict.h (counts_to_freqs): Rename to ...
7617         (update_max_bb_count): ... this one.
7618         * profile.c (compute_branch_probabilities): Add debug info
7619         * tree-inline.c (expand_call_inline): Update debug info.
7620         (optimize_inline_calls): Use update_max_bb_count..
7621         (tree_function_versioning): Use update_max_bb_count..
7622         * value-prof.c (gimple_value_profile_transformations):
7623         Do not use update_max_bb_count.
7625 2017-11-13  Jan Hubicka  <hubicka@ucw.cz>
7627         * ipa-inline.c (compute_uninlined_call_time, compute_inlined_call_time):
7628         always use frequencies.
7630 2017-11-13  Jan Hubicka  <hubicka@ucw.cz>
7632         * bb-reorder.c: Remove frequencies from comments.
7633         (better_edge_p): Use profile counts.
7634         (find_traces): Dump profile counts.
7635         (rotate_loop): Use profile counts.
7636         (find_traces_1_round): Likewise.
7637         (connect_better_edge_p): Use counts instead of probabilities for
7638         reverse walk.
7639         (copy_bb_p): Drop early check for non-0 frequency.
7640         (sanitize_hot_paths): Update comments.
7642 2017-11-13  Jan Hubicka  <hubicka@ucw.cz>
7644         * ipa-split.c (struct split_point): Add count.
7645         (consider_split): Do not compute incoming frequency; compute incoming
7646         count and store it to split_point.
7647         (split_function): Set count of the call to split part correctly.
7649 2017-11-13  Carl Love  <cel@us.ibm.com>
7651         * config/rs6000/altivec.md (altivec_vsumsws_be): Add define_expand.
7653 2017-11-13  Tom Tromey  <tom@tromey.com>
7655         * doc/cpp.texi (Variadic Macros): Document __VA_OPT__.
7657 2017-11-13  Carl Love  <cel@us.ibm.com>
7659         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins):
7660         Add support for builtins:
7661         unsigned int vec_first_{,miss}_match_{,or_eos}index,
7662         vector {un,}signed {char,int,short},
7663         vector {un,}signed {char,int,short}) arguments.
7664         * config/rs6000/rs6000-builtin.def (VFIRSTMATCHINDEX,
7665         VFIRSTMATCHOREOSINDEX, VFIRSTMISMATCHINDEX, VFIRSTMISMATCHOREOSINDEX):
7666         Add BU_P9V_AV_2 expansions for the builtins.
7667         * config/rs6000/altivec.h (vec_first_match_index,
7668         vec_first_mismatch_index, vec_first_match_or_eos_index,
7669         vec_first_mismatch_or_eos_index): Add #defines for the builtins.
7670         * config/rs6000/rs6000-protos.h (bytes_in_mode): Add
7671         new extern declaration.
7672         * config/rs6000/rs6000.c (bytes_in_mode): Add new function.
7673         * config/rs6000/vsx.md (first_match_index_<mode>,
7674         first_match_or_eos_index_<mode>, first_mismatch_index_<mode>,
7675         first_mismatch_or_eos_index_<mode>): Add define expand.
7676         (vctzlsbb_<mode>): Add mode field to define_insn for vctzlsbb.
7677         * doc/extend.texi: Update the built-in documenation file for the new
7678         built-in functions.
7680 2017-11-13  Michael Meissner  <meissner@linux.vnet.ibm.com>
7682         * match.pd: Convert fminf<N>, fminf<N>x, fmax<N>, and fmax<N>x
7683         into the min/max operations for _Float<N> and _Float<N>X types.
7685 2017-11-13  Eric Botcazou  <ebotcazou@adacore.com>
7687         PR lto/81351
7688         * dwarf2out.c (do_eh_frame): New static variable.
7689         (dwarf2out_begin_prologue): Set it.
7690         (dwarf2out_frame_finish): Test it instead of dwarf2out_do_eh_frame.
7692 2017-11-13  Jan Hubicka  <hubicka@ucw.cz>
7694         * tree-ssa-coalesce.c (coalesce_cost): Fix formating.
7696         * tree-ssa-sink.c (select_best_block): Do not use frequencies.
7698 2017-11-13  Eric Botcazou  <ebotcazou@adacore.com>
7700         PR lto/81351
7701         * debug.h (dwarf2out_do_eh_frame): Declare.
7702         * dwarf2cfi.c (dwarf2out_do_eh_frame): New predicate.
7703         (dwarf2out_do_frame): Use it.
7704         (dwarf2out_do_cfi_asm): Likewise.
7705         * dwarf2out.c (dwarf2out_frame_finish): Likewise.
7706         (dwarf2out_assembly_start): Likewise.
7707         (dwarf2out_begin_prologue): Fix comment.
7708         * toplev.c (compile_file): Always call dwarf2out_frame_finish
7709         if the target needs either debug or unwind DWARF2 info.
7710         * lto-opts.c (lto_write_options): Do not save -fexceptions,
7711         -fnon-call-exceptions, -ffp-contract, -fmath-errno, -fsigned-zeros,
7712         -ftrapping-math, -ftrapv and -fwrapv.
7714 2017-11-13  Jan Hubicka  <hubicka@ucw.cz>
7716         * cgraph.c (cgraph_edge::sreal_frequency): New function.
7717         * cgraph.h (cgraph_edge::sreal_frequency): Declare.
7718         * ipa-fnsummary.c (dump_ipa_call_summary): Use sreal_frequency.
7719         (estimate_edge_size_and_time): Likewise.
7720         (ipa_merge_fn_summary_after_inlining): Likewise.
7721         * ipa-inline.c (cgraph_freq_base_rec): Remove.
7722         (compute_uninlined_call_time): Use sreal_frequency.
7723         (compute_inlined_call_time): Likewise.
7724         (ipa_inline): Do not initialize cgraph_freq_base_rec.
7725         * profile-count.c: Include sreal.h.
7726         (profile_count::to_sreal_scale): New.
7727         * profile-count.h: Forward declare sreal.
7728         (profile_count::to_sreal_scale): Declare.
7730 2017-11-13  Nathan Sidwell  <nathan@acm.org>
7732         * diagnostic.c (maybe_line_and_column): New.
7733         (diagnostic_get_location_text): Use it.
7734         (diagnostic_report_current_module): Likewise.
7735         (test_diagnostic_get_location_text): Add tests.
7737 2017-11-13  Luis Machado  <luis.machado@linaro.org>
7739         * doc/md.texi (Specifying processor pipeline description): Fix
7740         incorrect latency for the div instruction example.
7742 2017-11-13  Jakub Jelinek  <jakub@redhat.com>
7744         PR tree-optimization/78821
7745         * gimple-ssa-store-merging.c (compatible_load_p): Don't require
7746         that bit_not_p is the same.
7747         (imm_store_chain_info::coalesce_immediate_stores): Likewise.
7748         (split_group): Count precisely bit_not_p bits in each statement.
7749         (invert_op): New function.
7750         (imm_store_chain_info::output_merged_store): Use invert_op to
7751         emit BIT_XOR_EXPR with a xor_mask instead of BIT_NOT_EXPR if some
7752         but not all orig_stores have BIT_NOT_EXPR in the corresponding spots.
7754 2017-11-13  Martin Liska  <mliska@suse.cz>
7756         * gcov.c (struct coverage_info): Remove typedef of coverage_t.
7757         (struct source_info): Likewise.
7758         (add_branch_counts): Likewise.
7759         (add_line_counts): Likewise.
7760         (function_summary): Likewise.
7761         (output_intermediate_line): Likewise.
7762         (generate_results): Likewise.
7764 2017-11-13  Martin Liska  <mliska@suse.cz>
7766         * gcov.c (struct block_info): Remove typedef for block_t.
7767         (struct line_info): Likewise.
7768         (line_info::has_block): Likewise.
7769         (EXIT_BLOCK): Likewise.
7770         (unblock): Likewise.
7771         (circuit): Likewise.
7772         (get_cycles_count): Likewise.
7773         (process_file): Likewise.
7774         (read_graph_file): Likewise.
7775         (solve_flow_graph): Likewise.
7776         (find_exception_blocks): Likewise.
7777         (add_line_counts): Likewise.
7778         (accumulate_line_info): Likewise.
7779         (output_line_details): Likewise.
7781 2017-11-13  Martin Liska  <mliska@suse.cz>
7783         * gcov.c (struct arc_info): Remove typedef for arc_t.
7784         (struct line_info): Likewise.
7785         (add_branch_counts): Likewise.
7786         (output_branch_count): Likewise.
7787         (function_info::~function_info): Likewise.
7788         (circuit): Likewise.
7789         (output_intermediate_line): Likewise.
7790         (read_graph_file): Likewise.
7791         (solve_flow_graph): Likewise.
7792         (find_exception_blocks): Likewise.
7793         (add_line_counts): Likewise.
7794         (accumulate_line_info): Likewise.
7795         (output_line_details): Likewise.
7796         (output_function_details): Likewise.
7798 2017-11-13  Martin Liska  <mliska@suse.cz>
7800         * gcov.c (struct function_info): Remove typedef for function_t.
7801         (struct source_info): Likewise.
7802         (source_info::get_functions_at_location): Likewise.
7803         (solve_flow_graph): Likewise.
7804         (find_exception_blocks): Likewise.
7805         (add_line_counts): Likewise.
7806         (output_intermediate_file): Likewise.
7807         (process_file): Likewise.
7808         (generate_results): Likewise.
7809         (release_structures): Likewise.
7810         (read_graph_file): Likewise.
7811         (read_count_file): Likewise.
7812         (accumulate_line_counts): Likewise.
7813         (output_lines): Likewise.
7815 2017-11-13  Martin Liska  <mliska@suse.cz>
7817         * gcov.c (function_info::function_info): Remove num_counts
7818         and add vector<gcov_type>.
7819         (function_info::~function_info): Use the vector.
7820         (process_file): Likewise.
7821         (read_graph_file): Likewise.
7822         (read_count_file): Likewise.
7823         (solve_flow_graph): Likewise.
7825 2017-11-13  Martin Liska  <mliska@suse.cz>
7827         * gcov.c (function_info::is_artificial): New function.
7828         (process_file): Erase all artificial early.
7829         (generate_results): Skip as all artificial are already
7830         removed.
7832 2017-11-13  Martin Liska  <mliska@suse.cz>
7834         * gcov.c (read_graph_file): Store to global vector of functions.
7835         (read_count_file): Iterate the vector.
7836         (process_file): Likewise.
7837         (generate_results): Likewise.
7838         (release_structures): Likewise.
7840 2017-11-13  Jakub Jelinek  <jakub@redhat.com>
7842         PR tree-optimization/82954
7843         * gimple-ssa-store-merging.c
7844         (imm_store_chain_info::coalesce_immediate_stores): If
7845         !infof->ops[N].base_addr, split group if info->ops[N].base_addr.
7847 2017-11-13  Richard Sandiford  <richard.sandiford@linaro.org>
7849         * config/aarch64/aarch64-simd.md (aarch64_store_lane0<mode>):
7850         Upddate call to ENDIAN_LANE_N.
7851         (aarch64_<sur>dot_lane<vsi2qi>): Use aarch64_endian_lane_rtx.
7852         (aarch64_<sur>dot_laneq<vsi2qi>): Likewise.
7853         (*aarch64_simd_vec_copy_lane<mode>): Update calls to ENDIAN_LANE_N
7854         and use aarch64_endian_lane_rtx.
7855         (*aarch64_simd_vec_copy_lane_<vswap_width_name><mode>): Likewise.
7857 2017-11-12  Tom de Vries  <tom@codesourcery.com>
7859         * config/riscv/riscv.h (ASM_OUTPUT_LABELREF): Wrap in do {} while (0).
7861 2017-11-12  Tom de Vries  <tom@codesourcery.com>
7863         * config/elfos.h (ASM_OUTPUT_ASCII): Remove semicolon after macro body.
7865 2017-11-12  Tom de Vries  <tom@codesourcery.com>
7867         * config/cr16/cr16.h (ASM_OUTPUT_LABELREF): Remove semicolon after macro
7868         body.
7869         * config/powerpcspe/xcoff.h (ASM_OUTPUT_LABELREF): Same.
7870         * config/rs6000/xcoff.h (ASM_OUTPUT_LABELREF): Same.
7871         * defaults.h (ASM_OUTPUT_LABELREF): Same.
7873 2017-11-11  Martin Sebor  <msebor@redhat.com>
7875         PR c/81117
7876         * doc/extend.texi (attribute nonstring): Remove spurious argument.
7878         PR bootstrap/82948
7879         * prefic.c (translate_name): Replace strncpy with memcpy to
7880         avoid -Wstringop-truncation.
7882 2017-11-10  Jan Hubicka  <hubicka@ucw.cz>
7884         * tree-ssa-loop-im.c (execute_sm_if_changed): Do not compute freq_sum.
7886 2017-11-10  Jan Hubicka  <hubicka@ucw.cz>
7888         * predict.c (maybe_hot_frequency_p): Do not use cfun.
7890 2017-11-10  Jan Hubicka  <hubicka@ucw.cz>
7892         * tree-ssa-tail-merge.c (replace_block_by): Fix and re-enable profile
7893         merging.
7895 2017-11-10  Michael Meissner  <meissner@linux.vnet.ibm.com>
7897         * config/rs6000/rs6000.md (bswaphi2_reg): On ISA 3.0 systems,
7898         enable generating XXBRH if the value is in a vector register.
7899         (bswapsi2_reg): On ISA 3.0 systems, enable generating XXBRW if the
7900         value is in a vector register.
7901         (bswapdi2_reg): On ISA 3.0 systems, always use XXBRD to do
7902         register to register bswap64's instead of doing the GPR sequence
7903         used on previous machines.
7904         (bswapdi2_xxbrd): New insn.
7905         (bswapdi2_reg): Disallow on ISA 3.0.
7906         (register to register bswap64 splitter): Do not split the insn on
7907         ISA 3.0 systems that use XXBRD.
7909 2017-11-10  Martin Sebor  <msebor@redhat.com>
7911         PR c/81117
7912         * config/darwin-c.c (framework_construct_pathname): Replace strncpy
7913         with memcpy.
7914         (find_subframework_file): Same.
7916 2017-11-10  Jan Hubicka  <hubicka@ucw.cz>
7918         * auto-profile.c (afdo_indirect_call): Drop frequency.
7919         * cgraph.c (symbol_table::create_edge): Drop frequency argument.
7920         (cgraph_node::create_edge): Drop frequency argument.
7921         (cgraph_node::create_indirect_edge): Drop frequency argument.
7922         (cgraph_edge::make_speculative): Drop frequency arguments.
7923         (cgraph_edge::resolve_speculation): Do not update frequencies
7924         (cgraph_edge::dump_edge_flags): Do not dump frequency.
7925         (cgraph_node::dump): Check consistency in IPA mode.
7926         (cgraph_edge::maybe_hot_p): Use IPA counter.
7927         (cgraph_edge::verify_count_and_frequency): Rename to ...
7928         (cgraph_edge::verify_count): ... this one; drop frequency checking.
7929         (cgraph_node::verify_node): Update.
7930         * cgraph.h (struct cgraph_edge): Drop frequency.
7931         (cgraph_edge::frequency): New function.
7932         * cgraphbuild.c (pass_build_cgraph_edges::execute): Donot pass
7933         frequencies.
7934         (cgraph_edge::rebuild_edges): Likewise.
7935         * cgraphclones.c (cgraph_edge::clone): Scale only counts.
7936         (duplicate_thunk_for_node): Do not pass frequency.
7937         (cgraph_node::create_clone): Scale only counts.
7938         (cgraph_node::create_virtual_clone): Do not pass frequency.
7939         (cgraph_node::create_edge_including_clones): Do not pass frequency.
7940         (cgraph_node::create_version_clone): Do not pass frequency.
7941         * cgraphunit.c (cgraph_node::analyze): Do not pass frequency.
7942         (cgraph_node::expand_thunk): Do not pass frequency.
7943         (cgraph_node::create_wrapper): Do not pass frequency.
7944         * gimple-iterator.c (update_call_edge_frequencies): Do not pass
7945         frequency.
7946         * gimple-streamer-in.c (input_bb): Scale only IPA counts.
7947         * ipa-chkp.c (chkp_produce_thunks): Do not pass frequency.
7948         * ipa-cp.c (ipcp_lattice::print): Use frequency function.
7949         (gather_caller_stats): Use frequency function.
7950         (ipcp_cloning_candidate_p): Use frequency function.
7951         (ipcp_propagate_stage): Use frequency function.
7952         (get_info_about_necessary_edges): Use frequency function.
7953         (update_profiling_info): Update only IPA profile.
7954         (update_specialized_profile): Use frequency functoin.
7955         (perhaps_add_new_callers): Update only IPA profile.
7956         * ipa-devirt.c (ipa_devirt): Use IPA profile.
7957         * ipa-fnsummary.c (redirect_to_unreachable): Do not set frequrency.
7958         (dump_ipa_call_summary): Use frequency function.
7959         (estimate_edge_size_and_time): Use frequency function.
7960         (ipa_merge_fn_summary_after_inlining): Use frequency function.
7961         * ipa-inline-analysis.c (do_estimate_edge_time): Use IPA profile.
7962         * ipa-inline-transform.c (update_noncloned_frequencies): Rename to ..
7963         (update_noncloned_counts): ... ths one; scale counts only.
7964         (clone_inlined_nodes): Do not scale frequency.
7965         (inline_call): Do not pass frequency.
7966         * ipa-inline.c (compute_uninlined_call_time): Use IPA profile.
7967         (compute_inlined_call_time): Use IPA profile.
7968         (want_inline_small_function_p): Use IPA profile.
7969         (want_inline_self_recursive_call_p): Use IPA profile.
7970         (edge_badness): Use IPA profile.
7971         (lookup_recursive_calls): Use IPA profile.
7972         (recursive_inlining): Do not pass frequency.
7973         (resolve_noninline_speculation): Do not update frequency.
7974         (inline_small_functions): Collect max of IPA profile.
7975         (dump_overall_stats): Dump IPA porfile.
7976         (dump_inline_stats): Dump IPA porfile.
7977         (ipa_inline): Collect IPA stats.
7978         * ipa-inline.h (clone_inlined_nodes): Update prototype.
7979         * ipa-profile.c (ipa_propagate_frequency_1): Use frequency function.
7980         (ipa_propagate_frequency): Use frequency function.
7981         (ipa_profile): Cleanup.
7982         * ipa-prop.c (ipa_make_edge_direct_to_target): Do not pass frequency
7983         * ipa-utils.c (ipa_merge_profiles): Merge all profiles.
7984         * lto-cgraph.c (lto_output_edge): Do not stream frequency.
7985         (input_node): Do not stream frequency.
7986         (input_edge): Do not stream frequency.
7987         (merge_profile_summaries): Scale only IPA profiles.
7988         * omp-simd-clone.c (simd_clone_adjust): Do not pass frequency.
7989         * predict.c (drop_profile): Do not recompute frequency.
7990         * trans-mem.c (ipa_tm_insert_irr_call): Do not pass frequency.
7991         (ipa_tm_insert_gettmclone_call): Do not pass frequency.
7992         * tree-cfg.c (execute_fixup_cfg): Drop profile to global0 if needed.
7993         * tree-chkp.c (chkp_copy_bounds_for_assign): Do not pass frequency.
7994         * tree-emutls.c (gen_emutls_addr): Do not pass frequency.
7995         * tree-inline.c (copy_bb): Do not scale frequency.
7996         (expand_call_inline): Do not scale frequency.
7997         (tree_function_versioning): Do not scale frequency.
7998         * ubsan.c (ubsan_create_edge): Do not pass frequency.
8000 2017-11-10  Julia Koval  <julia.koval@intel.com>
8002         * config/i386/gfniintrin.h (_mm_gf2p8affine_epi64_epi8)
8003         (_mm256_gf2p8affine_epi64_epi8, _mm_mask_gf2p8affine_epi64_epi8)
8004         (_mm_maskz_gf2p8affine_epi64_epi8, _mm256_mask_gf2p8affine_epi64_epi8)
8005         (_mm256_maskz_gf2p8affine_epi64_epi8)
8006         (_mm512_mask_gf2p8affine_epi64_epi8, _mm512_gf2p8affine_epi64_epi8)
8007         (_mm512_maskz_gf2p8affine_epi64_epi8): New intrinsics.
8008         * config/i386/i386-builtin.def (__builtin_ia32_vgf2p8affineqb_v64qi)
8009         (__builtin_ia32_vgf2p8affineqb_v32qi)
8010         (__builtin_ia32_vgf2p8affineqb_v16qi): New builtins.
8011         * config/i386/sse.md (vgf2p8affineqb_<mode><mask_name>): New pattern.
8013 2017-11-10  Tamar Christina  <tamar.christina@arm.com>
8015         PR target/82641
8016         * config/arm/arm.c
8017         (arm_option_override): Refactor.
8018         (arm_option_reconfigure_globals): New.
8019         (arm_options_perform_arch_sanity_checks): New.
8020         * config/arm/arm-protos.h (arm_option_reconfigure_globals):
8021         New prototype.
8022         (arm_options_perform_arch_sanity_checks): Likewise
8024 2017-11-10  Pat Haugen  <pthaugen@us.ibm.com>
8026         * rs6000/power9.md (power9-qpdiv): Correct DFU pipe usage.
8027         (power9-qpmul): New.
8028         * rs6000/rs6000.md ("type" attr): Add qmul.
8029         (mul<mode>3, fma<mode>4_hw, *fms<mode>4_hw, *nfma<mode>4_hw,
8030         *nfms<mode>4_hw, mul<mode>3_odd, fma<mode>4_odd, *fms<mode>4_odd,
8031         *nfma<mode>4_odd, *nfms<mode>4_odd): Change type to qmul.
8033 2017-11-10  Martin Sebor  <msebor@redhat.com>
8035         PR c/81117
8036         * builtins.c (compute_objsize): Handle arrays that
8037         compute_builtin_object_size likes to fail for.  Make extern.
8038         * builtins.h (compute_objsize): Declare.
8039         (check_strncpy_sizes): New function.
8040         (expand_builtin_strncpy): Call check_strncpy_sizes.
8041         * gimple-fold.c (gimple_fold_builtin_strncpy): Implement
8042         -Wstringop-truncation.
8043         (gimple_fold_builtin_strncat): Same.
8044         * gimple.c (gimple_build_call_from_tree): Set call location.
8045         * tree-ssa-strlen.c (strlen_to_stridx): New global variable.
8046         (maybe_diag_bound_equal_length, is_strlen_related_p): New functions.
8047         (handle_builtin_stxncpy, handle_builtin_strncat): Same.
8048         (handle_builtin_strlen): Use strlen_to_stridx.
8049         (strlen_optimize_stmt): Handle flavors of strncat, strncpy, and
8050         stpncpy.
8051         Use strlen_to_stridx.
8052         (pass_strlen::execute): Release strlen_to_stridx.
8053         * doc/invoke.texi (-Wsizeof-pointer-memaccess): Document enhancement.
8054         (-Wstringop-truncation): Document new option.
8056 2017-11-10  Martin Liska  <mliska@suse.cz>
8058         PR gcov-profile/82702
8059         * gcov.c (main): Handle intermediate files in a different
8060         way.
8061         (get_gcov_intermediate_filename): New function.
8062         (output_gcov_file): Remove support of intermediate files.
8063         (generate_results): Allocate intermediate file.
8064         (release_structures): Clean-up properly fn_end.
8065         (output_intermediate_file): Start iterating with line 1.
8067 2017-11-10  Jakub Jelinek  <jakub@redhat.com>
8069         PR tree-optimization/82929
8070         * gimple-ssa-store-merging.c (struct store_immediate_info): Add
8071         ops_swapped_p non-static data member.
8072         (store_immediate_info::store_immediate_info): Clear it.
8073         (imm_store_chain_info::coalesce_immediate_stores): If swapping
8074         ops set ops_swapped_p.
8075         (count_multiple_uses): Handle ops_swapped_p.
8077 2017-11-10  Martin Liska  <mliska@suse.cz>
8079         * coverage.c (coverage_init): Stream information about
8080         support of has_unexecuted_blocks.
8081         * doc/gcov.texi: Document that.
8082         * gcov-dump.c (dump_gcov_file): Support it in gcov_dump tool.
8083         * gcov.c (read_graph_file): Likewise.
8084         (output_line_beginning): Fix a small issue with
8085         color output.
8087 2017-11-10  Bin Cheng  <bin.cheng@arm.com>
8089         * tree-predcom.c (determine_roots_comp): Avoid memory leak by freeing
8090         reference of trivial component.
8092 2017-11-10  Jakub Jelinek  <jakub@redhat.com>
8094         PR bootstrap/82916
8095         * gimple-ssa-store-merging.c
8096         (pass_store_merging::terminate_all_aliasing_chains): For
8097         gimple_store_p stmts also call refs_output_dependent_p.
8099         PR rtl-optimization/82913
8100         * compare-elim.c (try_merge_compare): Punt if def_insn is not
8101         single set.
8103 2017-11-09  Jeff Law  <law@redhat.com>
8105         * vr-values.h: New file with vr_values class.
8106         * tree-vrp.c: Include vr-values.h
8107         (vrp_value_range_pool, vrp_equiv_obstack, num_vr_values): Move static
8108         data objects into the vr_values class.
8109         (vr_value, values_propagated, vr_phi_edge_counts): Likewise.
8110         (get_value_range): Make it a member function within vr_values class.
8111         (set_defs_to_varying, update_value_range, add_equivalence): Likewise.
8112         (vrp_stmt_computes_nonzero_p, op_with_boolean_value_range_p): Likewise.
8113         (op_with_constant_singleton_value_range): Likewise.
8114         (extract_range_for_var_from_comparison_expr): Likewise.
8115         (extract_range_from_assert, extract_range_from_ssa_name): Likewise.
8116         (extract_range_from_binary_expr): Likewise.
8117         (extract_range_from_unary_expr): Likewise.
8118         (extract_range_from_cond_expr, extrat_range_from_comparison): Likewise.
8119         (check_for_binary_op_overflow, extract_range_basic): Likewise.
8120         (extract_range_from_assignment, adjust_range_with_scev): Likewise.
8121         (dump_all_value_ranges, get_vr_for_comparison): Likewise.
8122         (compare_name_with_value, compare_names): Likewise.
8123         (vrp_evaluate_conditional_warnv_with_ops_using_ranges): Likewise.
8124         (vrp_evaluate_conditional_warnv_with_ops): Likewise.  Remove prototype.
8125         (vrp_evaluate_conditional, vrp_visit_cond_stmt): Likewise.
8126         (vrp_visit_switch_stmt, extract_range_from_stmt): Likewise.
8127         (extract_range_from_phi_node): Likewise.
8128         (simplify_truth_ops_using_ranges): Likewise.
8129         (simplify_div_or_mod_using_ranges): Likewise.
8130         (simplify_min_or_max_using_ranges, simplify_abs_using_ranges): Likewise.
8131         (simplify_bit_ops_using_ranges, simplify_cond_using_ranges_1): Likewise.
8132         (simplify_cond_using_ranges_2, simplify_switch_using_ranges): Likewise.
8133         (simplify_float_conversion_using_ranges): Likewise.
8134         (simplify_internal_call_using_ranges): Likewise.
8135         (two_valued_val_range_p, simplify_stmt_using_ranges): Likewise.
8136         (vrp_visit_assignment_or_call): Likewise.  Smuggle class instance
8137         poitner via x_vr_values for calls into gimple folder.
8138         (vrp_initialize_lattice): Make this the vr_values ctor.
8139         (vrp_free_lattice): Make this the vr_values dtor.
8140         (set_vr_value): New function.
8141         (class vrp_prop): Add vr_values data member.  Add various member
8142         functions as well as member functions that delegate to vr_values.
8143         (check_array_ref): Make a member function within vrp_prop class.
8144         (search_for_addr_array, vrp_initialize): Likewise.
8145         (vrp_finalize): Likewise.  Revamp to avoid direct access to
8146         vr_value, values_propagated, etc.
8147         (check_array_bounds): Extract vrp_prop class instance pointer from
8148         walk info structure.  Use it to call member functions.
8149         (check_all_array_refs): Make a member function within vrp_prop class.
8150         Smuggle class instance pointer via walk info structure.
8151         (x_vr_values): New local static.
8152         (vrp_valueize): Use x_vr_values to get class instance.
8153         (vr_valueize_1): Likewise.
8154         (class vrp_folder): Add vr_values data member.  Add various member
8155         functions as well as member functions that delegate to vr_values.
8156         (fold_predicate_in): Make a mber fucntion within vrp_folder class.
8157         (simplify_stmt_for_jump_threading): Extract smuggled vr_values
8158         class instance from vr_values.  Use it to call member functions.
8159         (vrp_dom_walker): Add vr_values data member.
8160         (vrp_dom_walker::after_dom_children): Smuggle vr_values class
8161         instance via x_vr_values.
8162         (identify_jump_threads): Accept vr_values as argument.  Store
8163         it into the walker structure.
8164         (evrp_dom_walker): Add vr_values class data member.  Add various
8165         delegators.
8166         (evrp_dom_walker::try_find_new_range): Use vr_values data
8167         member to access the memory allocator.
8168         (evrp_dom_walker::before_dom_children): Store vr_values class
8169         instance into the vrp_folder class.
8170         (evrp_dom_walker::push_value_range): Rework to avoid direct
8171         access to num_vr_values and vr_value.
8172         (evrp_dom_walker::pop_value_range): Likewise.
8173         (execute_early_vrp): Remove call to vrp_initialize_lattice.
8174         Use vr_values to get to dump_all_value_ranges member function.
8175         Remove call to vrp_free_lattice.  Call vrp_initialize, vrp_finalize,
8176         and simplify_cond_using_ranges_2 via vrp_prop class instance.
8177         Pass vr_values class instance down to identify_jump_threads.
8178         Remove call to vrp_free_lattice.
8179         (debug_all_value_ranges): Remove.
8181         * tree-vrp.c (vrp_prop): Move class to earlier point in the file.
8182         (vrp_folder): Likewise.
8184         * tree-vrp.c (set_value_range): Do not reference vrp_equiv_obstack.
8185         Get it from the existing bitmap instead.
8186         (vrp_intersect_ranges_1): Likewise.
8188 2017-11-09  Jakub Jelinek  <jakub@redhat.com>
8190         * gimple-ssa-store-merging.c (struct store_immediate_info): Add
8191         bit_not_p field.
8192         (store_immediate_info::store_immediate_info): Add bitnotp argument,
8193         set bit_not_p to it.
8194         (imm_store_chain_info::coalesce_immediate_stores): Break group
8195         if bit_not_p is different.
8196         (count_multiple_uses, split_group,
8197         imm_store_chain_info::output_merged_store): Handle info->bit_not_p.
8198         (handled_load): Avoid multiple chained BIT_NOT_EXPRs.
8199         (pass_store_merging::process_store): Handle BIT_{AND,IOR,XOR}_EXPR
8200         result inverted using BIT_NOT_EXPR, compute bit_not_p, pass it
8201         to store_immediate_info ctor.
8203 2017-11-09  Jim Wilson  <jimw@sifive.com>
8205         * collect2.c (OBJECT_FORMAT_COFF): Remove EXTENDED_COFF support.
8206         (scan_prog_file): Likewise.
8208 2017-11-09  Jan Hubicka  <hubicka@ucw.cz>
8210         * bb-reorder.c (max_entry_frequency): Remove.
8211         (find_traces, rotate_loop, mark_bb_visited, connect_better_edge_p,
8212         connect_traces, push_to_next_round_p): Remove prototypes.
8213         (find_traces_1_round): Use counts only.
8214         (push_to_next_round_p): Likewise.
8215         (find_traces): Likewise.
8216         (rotate_loop): Likewise.
8217         (find_traces_1_round): Likewise.
8218         (connect_traces): Likewise.
8219         (edge_order): Likewise.
8221 2017-11-09  Thomas Preud'homme  <thomas.preudhomme@arm.com>
8223         * config/arm/arm.c (output_return_instruction): Add comments to
8224         indicate requirement for cmse_nonsecure_entry return to account
8225         for the size of clearing instruction output here.
8226         (thumb_exit): Likewise.
8227         * config/arm/thumb2.md (thumb2_cmse_entry_return): Fix length for
8228         return in hardfloat mode.
8230 2017-11-09  Segher Boessenkool  <segher@kernel.crashing.org>
8232         * config/rs6000/rs6000.c (machine_function): Add a bool,
8233         "toc_is_wrapped_separately".
8234         (rs6000_option_override_internal): Enable OPTION_MASK_SAVE_TOC_INDIRECT
8235         if it wasn't explicitly set or unset, we are optimizing for speed, and
8236         doing separate shrink-wrapping.
8237         (rs6000_get_separate_components): Enable the TOC component if
8238         saving the TOC register in the prologue.
8239         (rs6000_components_for_bb): Handle the TOC component.
8240         (rs6000_emit_prologue_components): Store the TOC register where needed.
8241         (rs6000_set_handled_components): Mark TOC as handled, if handled.
8242         (rs6000_emit_prologue): Don't save the TOC if that is already done.
8244 2017-11-09  Martin Jambor  <mjambor@suse.cz>
8246         * ipa-param-manipulation.c: New file.
8247         * ipa-param-manipulation.h: Likewise.
8248         * Makefile.in (OBJS): Add ipa-param-manipulation.o.
8249         (PLUGIN_HEADERS): Addded ipa-param-manipulation.h
8250         * ipa-param.h (ipa_parm_op): Moved to ipa-param-manipulation.h.
8251         (ipa_parm_adjustment): Likewise.
8252         (ipa_parm_adjustment_vec): Likewise.
8253         (ipa_get_vector_of_formal_parms): Moved declaration to
8254         ipa-param-manipulation.h.
8255         (ipa_get_vector_of_formal_parm_types): Likewise.
8256         (ipa_modify_formal_parameters): Likewise.
8257         (ipa_modify_call_arguments): Likewise.
8258         (ipa_combine_adjustments): Likewise.
8259         (ipa_dump_param_adjustments): Likewise.
8260         (ipa_modify_expr): Likewise.
8261         (ipa_get_adjustment_candidate): Likewise.
8262         * ipa-prop.c (ipa_get_vector_of_formal_parms): Moved to
8263         ipa-param-manipulation.c.
8264         (ipa_get_vector_of_formal_parm_types): Likewise.
8265         (ipa_modify_formal_parameters): Likewise.
8266         (ipa_modify_call_arguments): Likewise.
8267         (ipa_modify_expr): Likewise.
8268         (get_ssa_base_param): Likewise.
8269         (ipa_get_adjustment_candidate): Likewise.
8270         (index_in_adjustments_multiple_times_p): Likewise.
8271         (ipa_combine_adjustments): Likewise.
8272         (ipa_dump_param_adjustments): Likewise.
8273         * tree-sra.c: Also include ipa-param-manipulation.h
8274         * omp-simd-clone.c: Include ipa-param-manipulation.h instead of
8275         ipa-param.h.
8277 2017-11-09  Richard Sandiford  <richard.sandiford@linaro.org>
8278             Alan Hayward  <alan.hayward@arm.com>
8279             David Sherwood  <david.sherwood@arm.com>
8281         * doc/sourcebuild.texi (vect_masked_store): Document.
8283 2017-11-09  Richard Sandiford  <richard.sandiford@linaro.org>
8284             Alan Hayward  <alan.hayward@arm.com>
8285             David Sherwood  <david.sherwood@arm.com>
8287         * doc/sourcebuild.texi (vect_align_stack_vars): Document.
8289 2017-11-09  Richard Sandiford  <richard.sandiford@linaro.org>
8290             Alan Hayward  <alan.hayward@arm.com>
8291             David Sherwood  <david.sherwood@arm.com>
8293         * doc/sourcebuild.texi (vect_variable_length): Document.
8295 2017-11-09  Richard Sandiford  <richard.sandiford@linaro.org>
8296             Alan Hayward  <alan.hayward@arm.com>
8297             David Sherwood  <david.sherwood@arm.com>
8299         * doc/sourcebuild.texi (vect_unaligned_possible): Document.
8301 2017-11-09  Richard Sandiford  <richard.sandiford@linaro.org>
8302             Alan Hayward  <alan.hayward@arm.com>
8303             David Sherwood  <david.sherwood@arm.com>
8305         * doc/sourcebuild.texi (vect_element_align_preferred): Document.
8307 2017-11-09  Richard Sandiford  <richard.sandiford@linaro.org>
8308             Alan Hayward  <alan.hayward@arm.com>
8309             David Sherwood  <david.sherwood@arm.com>
8311         * doc/sourcebuild.texi (vect_perm_short, vect_perm_byte): Document
8312         previously undocumented selectors.
8313         (vect_perm3_byte, vect_perm3_short, vect_perm3_int): Document.
8315 2017-11-09  Richard Sandiford  <richard.sandiford@linaro.org>
8317         * doc/rtl.texi (const_vector): Say that elements can be
8318         const_wide_ints too.
8319         * emit-rtl.h (valid_for_const_vec_duplicate_p): Declare.
8320         * emit-rtl.c (valid_for_const_vec_duplicate_p): New function.
8321         (gen_vec_duplicate): Use it instead of CONSTANT_P.
8322         * optabs.c (expand_vector_broadcast): Likewise.
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         * tree-ssa-loop-ivopts.c (get_address_cost): Try using a
8329         scaled index even if the unscaled address was invalid.
8330         Don't increase the complexity of using a scale in that case.
8332 2017-11-09  Richard Sandiford  <richard.sandiford@linaro.org>
8333             Alan Hayward  <alan.hayward@arm.com>
8334             David Sherwood  <david.sherwood@arm.com>
8336         * doc/rtl.texi: Rewrite the subreg rules so that they partition
8337         the inner register into REGMODE_NATURAL_SIZE bytes rather than
8338         UNITS_PER_WORD bytes.
8339         * emit-rtl.c (validate_subreg): Divide subregs into blocks
8340         based on REGMODE_NATURAL_SIZE of the inner mode.
8341         (gen_lowpart_common): Split the SCALAR_FLOAT_MODE_P and
8342         !SCALAR_FLOAT_MODE_P cases.  Use REGMODE_NATURAL_SIZE for the latter.
8343         * expmed.c (lowpart_bit_field_p): Divide the value up into
8344         chunks of REGMODE_NATURAL_SIZE rather than UNITS_PER_WORD.
8345         * expr.c (store_constructor): Use REGMODE_NATURAL_SIZE to test
8346         whether something is likely to occupy more than one register.
8348 2017-11-09  Jan Hubicka  <hubicka@ucw.cz>
8350         PR ipa/82879
8351         * ipa-inline-transform.c (update_noncloned_frequencies): Use
8352         profile_count::adjust_for_ipa_scaling.
8353         * tree-inline.c (copy_bb, copy_cfg_body): Likewise.
8354         * profile-count.c (profile_count::adjust_for_ipa_scaling): New member
8355         function.
8356         * profile-count.h (profile_count::adjust_for_ipa_scaling): Declare.
8358 2017-11-09  Jakub Jelinek  <jakub@redhat.com>
8360         * gimple-ssa-store-merging.c (count_multiple_uses): New function.
8361         (split_group): Add total_orig and total_new arguments, estimate the
8362         number of statements related to the store group without store merging
8363         and with store merging.
8364         (imm_store_chain_info::output_merged_store): Adjust split_group
8365         callers, punt if estimated number of statements with store merging
8366         is not smaller than estimated number of statements without it.
8367         Formatting fix.
8368         (handled_load): Remove has_single_use checks.
8369         (pass_store_merging::process_store): Likewise.
8371 2017-11-09  Richard Biener  <rguenther@suse.de>
8373         PR tree-optimization/82902
8374         * tree-ssa-phiprop.c (propagate_with_phi): Test proper type.
8376 2017-11-09  Martin Liska  <mliska@suse.cz>
8378         PR target/82863
8379         * emit-rtl.c (init_emit_regs): Initialize split_branch_probability to
8380         uninitialized.
8382 2017-11-09  Martin Liska  <mliska@suse.cz>
8384         PR tree-optimization/82669
8385         * sbitmap.h (bmp_iter_set_init): Remove non needed check.
8387 2017-11-09  Martin Liska  <mliska@suse.cz>
8389         PR gcov-profile/48463
8390         * coverage.c (coverage_begin_function): Output also end locus
8391         of a function and information whether the function is
8392         artificial.
8393         * gcov-dump.c (tag_function): Parse and print the information.
8394         * gcov.c (INCLUDE_MAP): Add include.
8395         (INCLUDE_SET): Likewise.
8396         (struct line_info): Move earlier in the source file because
8397         of vector<line_info> in function_info structure.
8398         (line_info::line_info): Likewise.
8399         (line_info::has_block): Likewise.
8400         (struct source_info): Add new member index.
8401         (source_info::get_functions_at_location): New function.
8402         (function_info::group_line_p): New function.
8403         (output_intermediate_line): New function.
8404         (output_intermediate_file): Use the mentioned function.
8405         (struct function_start): New.
8406         (struct function_start_pair_hash): Likewise.
8407         (process_file): Add code that identifies group functions.
8408         Assign lines either to global or function scope.
8409         (generate_results): Skip artificial functions.
8410         (find_source): Assign index for each source file.
8411         (read_graph_file): Read new flag artificial and end_line.
8412         (add_line_counts): Assign it either to global of function scope.
8413         (accumulate_line_counts): Isolate core of the function to
8414         accumulate_line_info and call it for both function and global
8415         scope lines.
8416         (accumulate_line_info): New function.
8417         (output_line_beginning): Fix GNU coding style.
8418         (print_source_line): New function.
8419         (output_line_details): Likewise.
8420         (output_function_details): Likewise.
8421         (output_lines): Iterate both source (global) scope and function
8422         scope.
8423         (struct function_line_start_cmp): New class.
8424         * doc/gcov.texi: Reflect changes in documentation.
8426 2017-11-09  Jakub Jelinek  <jakub@redhat.com>
8428         PR debug/82837
8429         * dwarf2out.c (const_ok_for_output_1): Reject NEG in addition to NOT.
8430         (mem_loc_descriptor): Handle (const (neg (...))) as (neg (const (...)))
8431         and similarly for not instead of neg.
8433 2017-11-08  Andi Kleen  <ak@linux.intel.com>
8435         * config/i386/i386.opt: Add -mforce-indirect-call.
8436         * config/i386/predicates.md: Check for flag_force_indirect_call.
8437         * doc/invoke.texi: Document -mforce-indirect-call
8439 2017-11-08  Kito Cheng  <kito.cheng@gmail.com>
8441         * config/riscv/riscv-protos.h (riscv_slow_unaligned_access_p):
8442         New extern.
8443         (MOVE_RATIO): Use riscv_slow_unaligned_access_p.
8444         config/riscv/riscv.c (predict.h): New include.
8445         (riscv_slow_unaligned_access_p): No longer static.
8446         (riscv_block_move_straight): Add require.
8447         * config/riscv/riscv-protos.h (riscv_hard_regno_nregs): Delete.
8449 2017-11-08  Jakub Jelinek  <jakub@redhat.com>
8451         PR target/82855
8452         * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>,
8453         <avx512>_eq<mode>3<mask_scalar_merge_name>_1): Use
8454         nonimmediate_operand predicate for operand 1 instead of
8455         register_operand.
8457 2017-11-08  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
8459         * config/aarch64/aarch64-simd.md (store_pair_lanes<mode>):
8460         New pattern.
8461         * config/aarch64/constraints.md (Uml): New constraint.
8462         * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): New
8463         predicate.
8465 2017-11-08  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
8467         * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
8468         of two vec_duplicates into a vec_concat.
8470 2017-11-08  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
8472         * simplify-rtx.c (simplify_ternary_operation, VEC_MERGE):
8473         Simplify vec_merge of vec_duplicate and vec_concat.
8474         * config/aarch64/constraints.md (Utq): New constraint.
8475         * config/aarch64/aarch64-simd.md (load_pair_lanes<mode>): New
8476         define_insn.
8478 2017-11-08  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
8480         * simplify-rtx.c (simplify_ternary_operation, VEC_MERGE):
8481         Simplify vec_merge of vec_duplicate and const_vector.
8482         * config/aarch64/predicates.md (aarch64_simd_or_scalar_imm_zero):
8483         New predicate.
8484         * config/aarch64/aarch64-simd.md (*aarch64_combinez<mode>): Use VDC
8485         mode iterator.  Update predicate on operand 1 to
8486         handle non-const_vec constants.  Delete constraints.
8487         (*aarch64_combinez_be<mode>): Likewise for operand 2.
8489 2017-11-08  Jakub Jelinek  <jakub@redhat.com>
8491         PR tree-optimization/78821
8492         * gimple-ssa-store-merging.c (struct store_operand_info): Add bit_not_p
8493         data member.
8494         (store_operand_info::store_operand_info): Initialize it to false.
8495         (pass_store_merging::terminate_all_aliasing_chains): Rewritten to use
8496         ref_maybe_used_by_stmt_p and stmt_may_clobber_ref_p on lhs of each
8497         store in the group, and if chain_info is non-NULL, to ignore altogether
8498         that chain.
8499         (compatible_load_p): Fail if bit_not_p does not match.
8500         (imm_store_chain_info::output_merged_store): Handle bit_not_p loads.
8501         (handled_load): Fill in bit_not_p.  Handle BIT_NOT_EXPR.
8502         (pass_store_merging::process_store): Adjust
8503         terminate_all_aliasing_chains calls to pass NULL in all current spots,
8504         call terminate_all_aliasing_chains newly when adding a store into
8505         a chain with non-NULL chain_info.
8507 2017-11-08  Wilco Dijkstra  <wdijkstr@arm.com>
8509         * config/aarch64/aarch64.c (aarch64_can_eliminate): Simplify logic.
8511 2017-11-08  Wilco Dijkstra  <wdijkstr@arm.com>
8513         * config/aarch64/aarch64.c (aarch64_frame_pointer_required)
8514         Remove.
8515         (aarch64_layout_frame): Initialise emit_frame_chain.
8516         (aarch64_can_eliminate): Remove omit leaf frame pointer code.
8517         (TARGET_FRAME_POINTER_REQUIRED): Remove define.
8519 2017-11-08  Martin Liska  <mliska@suse.cz>
8521         * gimplify.c (expand_FALLTHROUGH_r): Simplify usage
8522         of gimple_call_internal_p.
8524 2017-11-07  Tom de Vries  <tom@codesourcery.com>
8526         * config/mips/mips.h (ASM_OUTPUT_LABELREF): Wrap in "do {} while (0)".
8528 2017-11-07  Tom de Vries  <tom@codesourcery.com>
8530         * config/mips/mips.h (ASM_OUTPUT_CASE_END): Remove semicolon after
8531         "do {} while (0)".
8533 2017-11-08  Martin Liska  <mliska@suse.cz>
8535         PR sanitizer/82792
8536         * gimplify.c (expand_FALLTHROUGH_r): Skip IFN_ASAN_MARK.
8538 2017-11-07  Eric Botcazou  <ebotcazou@adacore.com>
8540         * gimple-pretty-print.c (dump_profile): Return "" instead of NULL.
8542 2017-11-07  Jakub Jelinek  <jakub@redhat.com>
8544         PR target/82855
8545         * config/i386/i386.md (SWI1248_AVX512BWDQ2_64): New mode iterator.
8546         (*cmp<mode>_ccz_1): New insn with $k alternative.
8548         PR target/82855
8549         * config/i386/i386.c (ix86_swap_binary_operands_p): Treat
8550         RTX_COMM_COMPARE as commutative as well.
8551         (ix86_binary_operator_ok): Formatting fix.
8552         * config/i386/sse.md (*mul<mode>3<mask_name><round_name>,
8553         *<code><mode>3<mask_name><round_saeonly_name>,
8554         *<code><mode>3<mask_name>, *<code>tf3, *mul<mode>3<mask_name>,
8555         *<s>mul<mode>3_highpart<mask_name>,
8556         *vec_widen_umult_even_v16si<mask_name>,
8557         *vec_widen_umult_even_v8si<mask_name>,
8558         *vec_widen_umult_even_v4si<mask_name>,
8559         *vec_widen_smult_even_v16si<mask_name>,
8560         *vec_widen_smult_even_v8si<mask_name>, *sse4_1_mulv2siv2di3<mask_name>,
8561         *avx2_pmaddwd, *sse2_pmaddwd, *<sse4_1_avx2>_mul<mode>3<mask_name>,
8562         *avx2_<code><mode>3, *avx512f_<code><mode>3<mask_name>,
8563         *sse4_1_<code><mode>3<mask_name>, *<code>v8hi3,
8564         *sse4_1_<code><mode>3<mask_name>, *<code>v16qi3, *avx2_eq<mode>3,
8565         <avx512>_eq<mode>3<mask_scalar_merge_name>_1, *sse4_1_eqv2di3,
8566         *sse2_eq<mode>3, <mask_codefor><code><mode>3<mask_name>,
8567         *<code><mode>3, *<sse2_avx2>_uavg<mode>3<mask_name>,
8568         *<ssse3_avx2>_pmulhrsw<mode>3<mask_name>, *ssse3_pmulhrswv4hi3): Use
8569         !(MEM_P (operands[1]) && MEM_P (operands[2])) condition instead of
8570         ix86_binary_operator_ok.  Formatting fixes.
8571         (*<plusminus_insn><mode>3<mask_name><round_name>,
8572         *<plusminus_insn><mode>3, *<plusminus_insn><mode>3_m): Formatting
8573         fixes.
8575 2017-11-07  Segher Boessenkool  <segher@kernel.crashing.org>
8577         * config/rs6000/rs6000.md (GPR2): New mode_iterator.
8578         ("cstore<mode>4"): Don't always expand with rs6000_emit_int_cmove for
8579         eq and ne if TARGET_ISEL.
8580         (cmp): New code_iterator.
8581         (UNS, UNSU_, UNSIK): New code_attrs.
8582         (<code><GPR:mode><GPR2:mode>2_isel): New define_insn_and_split.
8583         ("eq<mode>3"): New define_expand, rename the define_insn_and_split
8584         to...
8585         ("eq<mode>3"): ... this.
8586         ("ne<mode>3"): New define_expand, rename the define_insn_and_split
8587         to...
8588         ("ne<mode>3"): ... this.
8590 2017-11-07  Julia Koval  <julia.koval@intel.com>
8592         PR target/82812
8593         * common/config/i386/i386-common.c
8594         (OPTION_MASK_ISA_GENERAL_REGS_ONLY_UNSET): Remove MPX from flag.
8595         (ix86_handle_option): Move MPX to isa_flags2 and GFNI to isa_flags.
8596         * config/i386/i386-c.c (ix86_target_macros_internal): Ditto.
8597         * config/i386/i386.opt: Ditto.
8598         * config/i386/i386.c (ix86_target_string): Ditto.
8599         (ix86_option_override_internal): Ditto.
8600         (ix86_init_mpx_builtins): Move MPX to args2.
8601         (ix86_expand_builtin): Special handling for OPTION_MASK_ISA_GFNI.
8602         * config/i386/i386-builtin.def (__builtin_ia32_vgf2p8affineinvqb_v64qi,
8603         __builtin_ia32_vgf2p8affineinvqb_v64qi_mask,
8604         __builtin_ia32_vgf2p8affineinvqb_v32qi,
8605         __builtin_ia32_vgf2p8affineinvqb_v32qi_mask,
8606         __builtin_ia32_vgf2p8affineinvqb_v16qi,
8607         __builtin_ia32_vgf2p8affineinvqb_v16qi_mask): Move to ARGS array.
8609 2017-11-07  Uros Bizjak  <ubizjak@gmail.com>
8611         PR target/80425
8612         * config/i386.i386.md (*zero_extendsidi2): Change (?r,*Yj), (?*Yi,r)
8613         and (*x,m) to ($r,Yj), ($Yi,r) and ($x,m).
8614         (zero-extendsidi peephole2): Remove peephole.
8616 2017-11-07  Eric Botcazou  <ebotcazou@adacore.com>
8618         PR c/53037
8619         * stor-layout.c: Include attribs.h.
8620         (handle_warn_if_not_align): Replace test on TYPE_USER_ALIGN with
8621         explicit lookup of "aligned" attribute.
8623 2017-11-07  Andrew Waterman  <andrew@sifive.com>
8625         * config/riscv/riscv-protos.h (riscv_hard_regno_nregs): New prototype.
8626         (riscv_expand_block_move): Likewise.
8627         * config/riscv/riscv.h (MOVE_RATIO): Tune cost to movmemsi
8628         implementation.
8629         (RISCV_MAX_MOVE_BYTES_PER_LOOP_ITER): New define.
8630         (RISCV_MAX_MOVE_BYTES_STRAIGHT): New define.
8631         * config/riscv/riscv.c (riscv_block_move_straight): New function.
8632         (riscv_adjust_block_mem): Likewise.
8633         (riscv_block_move_loop): Likewise.
8634         (riscv_expand_block_move): Likewise.
8635         * config/riscv/riscv.md (movmemsi): New pattern.
8637 2017-11-07  Michael Clark  <michaeljclark@mac.com>
8639         * config/riscv/linux.h (MUSL_ABI_SUFFIX): New define.
8640         (MUSL_DYNAMIC_LINKER): Likewise.
8642 2017-11-07  Richard Sandiford  <richard.sandiford@linaro.org>
8644         * config/aarch64/aarch64.md (ashl<mode>3, ashr<mode>3, lshr<mode>3)
8645         (rotr<mode>3, rotl<mode>3): Use aarch64_reg_or_imm instead of
8646         nonmmory_operand.
8648 2017-11-07  Richard Biener  <rguenther@suse.de>
8650         * match.pd: Fix build.
8652 2017-11-07  Wilco Dijkstra  <wdijkstr@arm.com>
8653             Jackson Woodruff  <jackson.woodruff@arm.com>
8655         PR tree-optimization/71026
8656         * match.pd: Canonicalize negate in division.
8658 2017-11-07  Sudakshina Das  <sudi.das@arm.com>
8660         PR middle-end/80131
8661         * match.pd: Simplify 1 << (C - x) where C = precision (x) - 1.
8663 2017-11-07  Marc Glisse  <marc.glisse@inria.fr>
8665         * match.pd ((a&~b)|(a^b),(a&~b)^~a,(a|b)&~(a^b),a|~(a^b),
8666         (a|b)|(a&^b),(a&b)|~(a^b),~(~a&b),~X^Y): New transformations.
8668 2017-11-07  Marc Glisse  <marc.glisse@inria.fr>
8670         * fold-const.c (negate_expr_p) [PLUS_EXPR, MINUS_EXPR]: Handle
8671         non-scalar integral types.
8672         * match.pd (negate_expr_p): Handle MINUS_EXPR.
8673         (-(A-B), -(~A)): New transformations.
8675 2017-11-07  Tom de Vries  <tom@codesourcery.com>
8677         * config/powerpcspe/aix43.h (SUBTARGET_OVERRIDE_OPTIONS): Remove
8678         semicolon after "do {} while (0)".
8679         * config/powerpcspe/aix51.h (SUBTARGET_OVERRIDE_OPTIONS): Same.
8680         * config/powerpcspe/aix52.h (SUBTARGET_OVERRIDE_OPTIONS): Same.
8681         * config/powerpcspe/aix53.h (SUBTARGET_OVERRIDE_OPTIONS): Same.
8682         * config/powerpcspe/aix61.h (SUBTARGET_OVERRIDE_OPTIONS): Same.
8683         * config/powerpcspe/aix71.h (SUBTARGET_OVERRIDE_OPTIONS): Same.
8685 2017-11-07  Tom de Vries  <tom@codesourcery.com>
8687         * config/rs6000/aix43.h (SUBTARGET_OVERRIDE_OPTIONS): Remove semicolon
8688         after "do {} while (0)".
8689         * config/rs6000/aix51.h (SUBTARGET_OVERRIDE_OPTIONS): Same.
8690         * config/rs6000/aix52.h (SUBTARGET_OVERRIDE_OPTIONS): Same.
8691         * config/rs6000/aix53.h (SUBTARGET_OVERRIDE_OPTIONS): Same.
8692         * config/rs6000/aix61.h (SUBTARGET_OVERRIDE_OPTIONS): Same.
8693         * config/rs6000/aix71.h (SUBTARGET_OVERRIDE_OPTIONS): Same.
8695 2017-11-07  Tom de Vries  <tom@codesourcery.com>
8697         PR other/82784
8698         * config/arm/arm.c (HANDLE_NARROW_SHIFT_ARITH): Remove semicolon after
8699         "while {} do (0)".
8700         (arm_rtx_costs_internal): Add missing semicolon after
8701         HANDLE_NARROW_SHIFT_ARITH call.
8703 2017-11-06  Segher Boessenkool  <segher@kernel.crashing.org>
8705         * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
8706         disable isel if it was not set explicitly.
8708 2017-11-06  James Bowman  <james.bowman@ftdichip.com>
8710         * gcc/dwarf2out.c (modified_type_die): Retain ADDR_SPACE qualifiers.
8711         (add_type_attribute) likewise.
8713 2017-11-06  H.J. Lu  <hongjiu.lu@intel.com>
8715         * config/i386/i386.c (ix86_can_use_return_insn_p): Use reference
8716         of struct ix86_frame.
8717         (ix86_initial_elimination_offset): Likewise.
8718         (ix86_expand_split_stack_prologue): Likewise.
8720 2017-11-06  Marc Glisse  <marc.glisse@inria.fr>
8722         * tree-vrp.h (enum value_range_type): Update stale comment.
8724 2017-11-06  Richard Sandiford  <richard.sandiford@linaro.org>
8725             Alan Hayward  <alan.hayward@arm.com>
8726             David Sherwood  <david.sherwood@arm.com>
8728         * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm)
8729         (aarch64_expand_vec_perm_const): Take the number of units too.
8730         * config/aarch64/aarch64.c (aarch64_expand_vec_perm)
8731         (aarch64_expand_vec_perm_const): Likewise.
8732         * config/aarch64/aarch64-simd.md (vec_perm_const<mode>)
8733         (vec_perm<mode>): Update accordingly.
8735 2017-11-06  Richard Sandiford  <richard.sandiford@linaro.org>
8736             Alan Hayward  <alan.hayward@arm.com>
8737             David Sherwood  <david.sherwood@arm.com>
8739         * config/aarch64/aarch64-protos.h (aarch64_simd_vect_par_cnst_half):
8740         Take the number of units too.
8741         * config/aarch64/aarch64.c (aarch64_simd_vect_par_cnst_half): Likewise.
8742         (aarch64_simd_check_vect_par_cnst_half): Update call accordingly,
8743         but check for a vector mode before rather than after the call.
8744         * config/aarch64/aarch64-simd.md (aarch64_split_simd_mov<mode>)
8745         (move_hi_quad_<mode>, vec_unpack<su>_hi_<mode>)
8746         (vec_unpack<su>_lo_<mode, vec_widen_<su>mult_lo_<mode>)
8747         (vec_widen_<su>mult_hi_<mode>, vec_unpacks_lo_<mode>)
8748         (vec_unpacks_hi_<mode>, aarch64_saddl2<mode>, aarch64_uaddl2<mode>)
8749         (aarch64_ssubl2<mode>, aarch64_usubl2<mode>, widen_ssum<mode>3)
8750         (widen_usum<mode>3, aarch64_saddw2<mode>, aarch64_uaddw2<mode>)
8751         (aarch64_ssubw2<mode>, aarch64_usubw2<mode>, aarch64_sqdmlal2<mode>)
8752         (aarch64_sqdmlsl2<mode>, aarch64_sqdmlal2_lane<mode>)
8753         (aarch64_sqdmlal2_laneq<mode>, aarch64_sqdmlsl2_lane<mode>)
8754         (aarch64_sqdmlsl2_laneq<mode>, aarch64_sqdmlal2_n<mode>)
8755         (aarch64_sqdmlsl2_n<mode>, aarch64_sqdmull2<mode>)
8756         (aarch64_sqdmull2_lane<mode>, aarch64_sqdmull2_laneq<mode>)
8757         (aarch64_sqdmull2_n<mode>): Update accordingly.
8759 2017-11-06  Richard Sandiford  <richard.sandiford@linaro.org>
8760             Alan Hayward  <alan.hayward@arm.com>
8761             David Sherwood  <david.sherwood@arm.com>
8763         * config/aarch64/aarch64-protos.h (aarch64_reverse_mask): Take
8764         the number of units too.
8765         * config/aarch64/aarch64.c (aarch64_reverse_mask): Likewise.
8766         * config/aarch64/aarch64-simd.md (vec_load_lanesoi<mode>)
8767         (vec_store_lanesoi<mode>, vec_load_lanesci<mode>)
8768         (vec_store_lanesci<mode>, vec_load_lanesxi<mode>)
8769         (vec_store_lanesxi<mode>): Update accordingly.
8771 2017-11-06  Richard Sandiford  <richard.sandiford@linaro.org>
8772             Alan Hayward  <alan.hayward@arm.com>
8773             David Sherwood  <david.sherwood@arm.com>
8775         * config/aarch64/aarch64-protos.h (aarch64_endian_lane_rtx): Declare.
8776         * config/aarch64/aarch64.c (aarch64_endian_lane_rtx): New function.
8777         * config/aarch64/aarch64.h (ENDIAN_LANE_N): Take the number
8778         of units rather than the mode.
8779         * config/aarch64/iterators.md (nunits): New mode attribute.
8780         * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args):
8781         Use aarch64_endian_lane_rtx instead of GEN_INT (ENDIAN_LANE_N ...).
8782         * config/aarch64/aarch64-simd.md (aarch64_dup_lane<mode>)
8783         (aarch64_dup_lane_<vswap_width_name><mode>, *aarch64_mul3_elt<mode>)
8784         (*aarch64_mul3_elt_<vswap_width_name><mode>): Likewise.
8785         (*aarch64_mul3_elt_to_64v2df, *aarch64_mla_elt<mode>): Likewise.
8786         (*aarch64_mla_elt_<vswap_width_name><mode>, *aarch64_mls_elt<mode>)
8787         (*aarch64_mls_elt_<vswap_width_name><mode>, *aarch64_fma4_elt<mode>)
8788         (*aarch64_fma4_elt_<vswap_width_name><mode>):: Likewise.
8789         (*aarch64_fma4_elt_to_64v2df, *aarch64_fnma4_elt<mode>): Likewise.
8790         (*aarch64_fnma4_elt_<vswap_width_name><mode>): Likewise.
8791         (*aarch64_fnma4_elt_to_64v2df, reduc_plus_scal_<mode>): Likewise.
8792         (reduc_plus_scal_v4sf, reduc_<maxmin_uns>_scal_<mode>): Likewise.
8793         (reduc_<maxmin_uns>_scal_<mode>): Likewise.
8794         (*aarch64_get_lane_extend<GPI:mode><VDQQH:mode>): Likewise.
8795         (*aarch64_get_lane_zero_extendsi<mode>): Likewise.
8796         (aarch64_get_lane<mode>, *aarch64_mulx_elt_<vswap_width_name><mode>)
8797         (*aarch64_mulx_elt<mode>, *aarch64_vgetfmulx<mode>): Likewise.
8798         (aarch64_sq<r>dmulh_lane<mode>, aarch64_sq<r>dmulh_laneq<mode>)
8799         (aarch64_sqrdml<SQRDMLH_AS:rdma_as>h_lane<mode>): Likewise.
8800         (aarch64_sqrdml<SQRDMLH_AS:rdma_as>h_laneq<mode>): Likewise.
8801         (aarch64_sqdml<SBINQOPS:as>l_lane<mode>): Likewise.
8802         (aarch64_sqdml<SBINQOPS:as>l_laneq<mode>): Likewise.
8803         (aarch64_sqdml<SBINQOPS:as>l2_lane<mode>_internal): Likewise.
8804         (aarch64_sqdml<SBINQOPS:as>l2_laneq<mode>_internal): Likewise.
8805         (aarch64_sqdmull_lane<mode>, aarch64_sqdmull_laneq<mode>): Likewise.
8806         (aarch64_sqdmull2_lane<mode>_internal): Likewise.
8807         (aarch64_sqdmull2_laneq<mode>_internal): Likewise.
8808         (aarch64_vec_load_lanesoi_lane<mode>): Likewise.
8809         (aarch64_vec_store_lanesoi_lane<mode>): Likewise.
8810         (aarch64_vec_load_lanesci_lane<mode>): Likewise.
8811         (aarch64_vec_store_lanesci_lane<mode>): Likewise.
8812         (aarch64_vec_load_lanesxi_lane<mode>): Likewise.
8813         (aarch64_vec_store_lanesxi_lane<mode>): Likewise.
8814         (aarch64_simd_vec_set<mode>): Update use of ENDIAN_LANE_N.
8815         (aarch64_simd_vec_setv2di): Likewise.
8817 2017-11-06  Carl Love  <cel@us.ibm.com>
8819         * config/rs6000/rs6000-c.c (P8V_BUILTIN_VEC_REVB): Add power 8
8820         definitions.
8821         (P9V_BUILTIN_VEC_REVB): Remove the power 9 instance definitions.
8822         * config/rs6000/altivec.h (vec_revb): Change the #define from power 9
8823         to power 8.
8824         * config/rs6000/r6000-protos.h (swap_endian_selector_for_mode): Add new
8825         extern declaration.
8826         * config/rs6000/rs6000.c (swap_endian_selector_for_mode): Add function.
8827         * config/rs6000/rs6000-builtin.def (BU_P8V_VSX_1, BU_P8V_OVERLOAD_1):
8828         Add power 8 macro expansions.
8829         (BU_P9V_OVERLOAD_1): Remove power 9 overload expansion.
8830         * config/rs6000/vsx.md (revb_<mode>): Add define_expand to generate
8831         power 8 instructions.  (VSX_XXBR): Add iterator.
8833 2017-11-06  Wilco Dijkstra  <wdijkstr@arm.com>
8835         * config/arm/arm.md (predicable_short_it): Change default to "no",
8836         improve documentation, remove uses that are identical to the default.
8837         (enabled_for_depr_it): Rename to enabled_for_short_it.
8838         * gcc/config/arm/arm-fixed.md (predicable_short_it):
8839         Remove default uses.
8840         * gcc/config/arm/ldmstm.md (predicable_short_it): Likewise.
8841         * gcc/config/arm/sync.md (predicable_short_it): Likewise.
8842         * gcc/config/arm/thumb2.md (predicable_short_it): Likewise.
8843         * gcc/config/arm/vfp.md (predicable_short_it): Likewise.
8845 2017-11-06  Michael Meissner  <meissner@linux.vnet.ibm.com>
8847         PR target/82748
8848         * config/rs6000/rs6000-builtin.def (BU_FLOAT128_1): Delete
8849         float128 helper macros, which are no longer used after deleting
8850         the old 'q' built-in functions, and moving the round to odd
8851         built-in functions to being special built-in functions.
8852         (BU_FLOAT128_2): Likewise.
8853         (BU_FLOAT128_1_HW): Likewise.
8854         (BU_FLOAT128_2_HW): Likewise.
8855         (BU_FLOAT128_3_HW): Likewise.
8856         (FABSQ): Delete old 'q' built-in functions.
8857         (COPYSIGNQ): Likewise.
8858         (SQRTF128_ODD): Move round to odd built-in functions to be
8859         special built-in functions, so that we can handle
8860         -mabi=ieeelongdouble.
8861         (TRUNCF128_ODD): Likewise.
8862         (ADDF128_ODD): Likewise.
8863         (SUBF128_ODD): Likewise.
8864         (MULF128_ODD): Likewise.
8865         (DIVF128_ODD): Likewise.
8866         (FMAF128_ODD): Likewise.
8867         * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Map old 'q'
8868         built-in names to 'f128'.
8869         * config/rs6000/rs6000.c (rs6000_fold_builtin): Remove folding the
8870         old 'q' built-in functions, as the machine independent code for
8871         'f128' built-in functions handles this.
8872         (rs6000_expand_builtin): Add expansion for float128 round to odd
8873         functions, keying off on -mabi=ieeelongdouble of whether to use
8874         the KFmode or TFmode variant.
8875         (rs6000_init_builtins): Initialize the _Float128 round to odd
8876         built-in functions.
8877         * doc/extend.texi (PowerPC Built-in Functions): Document the old
8878         _Float128 'q' built-in functions are now mapped into the new
8879         'f128' built-in functions.
8881 2017-11-06  David Edelsohn  <dje.gcc@gmail.com>
8883         * collect2.c (add_lto_object): Compile for OBJECT_COFF.
8884         (scan_prog_file): Don't skip PASS_LTOINFO. Scan for LTO objects.
8886 2017-11-06  David Malcolm  <dmalcolm@redhat.com>
8888         PR jit/82826
8889         * ipa-fnsummary.c (ipa_fnsummary_c_finalize): New function.
8890         * ipa-fnsummary.h (ipa_fnsummary_c_finalize): New decl.
8891         * toplev.c: Include "ipa-fnsummary.h".
8892         (toplev::finalize): Call ipa_fnsummary_c_finalize.
8894 2017-11-06  Jakub Jelinek  <jakub@redhat.com>
8896         PR tree-optimization/82838
8897         * gimple-ssa-store-merging.c
8898         (imm_store_chain_info::output_merged_store): Call force_gimple_operand_1
8899         on a separate gimple_seq which is then appended to seq.
8901 2017-11-06  Jeff Law  <law@redhat.com>
8903         PR target/82788
8904         * config/i386/i386.c (PROBE_INTERVAL): Remove.
8905         (get_probe_interval): New functions.
8906         (ix86_adjust_stack_and_probe_stack_clash): Use get_probe_interval.
8907         (ix86_adjust_stack_and_probe): Likewise.
8908         (output_adjust_stack_and_probe): Likewise.
8909         (ix86_emit_probe_stack_range): Likewise.
8910         (ix86_expand_prologue): Likewise.
8912 2017-11-06  Richard Sandiford  <richard.sandiford@linaro.org>
8914         PR tree-optimization/82816
8915         * tree-ssa-math-opts.c (convert_mult_to_widen): Return false
8916         if the modes of the two types are the same.
8917         (convert_plusminus_to_widen): Likewise.
8919 2017-11-06  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
8921         * config/rs6000/altivec.md (*p9_vadu<mode>3) Rename to
8922         p9_vadu<mode>3.
8923         (usadv16qi): New define_expand.
8924         (usadv8hi): New define_expand.
8926 2017-11-06  Jan Hubicka  <hubicka@ucw.cz>
8928         PR bootstrap/82832
8929         * ipa-inline-transform.c (update_noncloned_frequencies): Always
8930         scale.
8931         (inline_transform): Likewise.
8932         * predict.c (counts_to_freqs): Remove useless conditional.
8933         * profile-count.h (profile_count::apply_scale): Move sanity check.
8934         * tree-inline.c (copy_bb): Always scale.
8935         (copy_cfg_body): Likewise.
8937 2017-11-06  Christophe Lyon  <christophe.lyon@linaro.org>
8939         PR target/67591
8940         * config/arm/arm.md (*sub_shiftsi): Add predicable_short_it
8941         attribute.
8942         (*cmp_ite0): Add enabled_for_depr_it attribute.
8943         (*cmp_ite1): Likewise.
8945 2017-11-06  Segher Boessenkool  <segher@kernel.crashing.org>
8947         * config/rs6000/rs6000.c (rs6000_insn_cost): Handle TYPE_MFCR and
8948         TYPE_MFCRF.
8950 2017-11-06  Richard Sandiford  <richard.sandiford@linaro.org>
8952         * tree-vrp.c (vrp_int_const_binop): Return true on success and
8953         return the value by pointer.
8954         (extract_range_from_multiplicative_op_1): Update accordingly.
8955         Return as soon as an operation fails.
8957 2017-11-05  Tom de Vries  <tom@codesourcery.com>
8959         PR other/82784
8960         * asan.c (DEF_SANITIZER_BUILTIN_1): Factor out of ...
8961         (DEF_SANITIZER_BUILTIN): ... here.
8962         (initialize_sanitizer_builtins): Use DEF_SANITIZER_BUILTIN_1 instead of
8963         DEF_SANITIZER_BUILTIN in if stmt.  Add missing semicolon.
8965 2017-11-05  Tom de Vries  <tom@codesourcery.com>
8967         PR other/82784
8968         * config/elfos.h (ASM_OUTPUT_BEFORE_CASE_LABEL): Remove semicolon after
8969         macro body.
8970         (ASM_OUTPUT_CASE_LABEL): Add semicolon after
8971         ASM_OUTPUT_BEFORE_CASE_LABEL call.
8972         * config/arc/arc.h (ASM_OUTPUT_BEFORE_CASE_LABEL): Remove semicolon
8973         after macro body.
8974         * config/m68k/m68kelf.h (ASM_OUTPUT_BEFORE_CASE_LABEL): Same.
8975         * config/mips/mips.h (ASM_OUTPUT_BEFORE_CASE_LABEL): Same.
8976         * config/v850/v850.h (ASM_OUTPUT_BEFORE_CASE_LABEL): Same.
8978 2017-11-05  Tom de Vries  <tom@codesourcery.com>
8980         PR other/82784
8981         * graphite-scop-detection.c (DEBUG_PRINT): Remove semicolon after
8982         "do {} while (0)".
8984 2017-11-04  Michael Clark  <michaeljclark@mac.com>
8986         * config/riscv/riscv.c (riscv_print_operand): Add a 'i' format.
8987         config/riscv/riscv.md (addsi3): Use 'i' for immediates.
8988         (adddi3): Likewise.
8989         (*addsi3_extended): Likewise.
8990         (*addsi3_extended2): Likewise.
8991         (<optab>si3): Likewise.
8992         (<optab>di3): Likewise.
8993         (<optab><mode>3): Likewise.
8994         (<*optabe>si3_internal): Likewise.
8995         (zero_extendqi<SUPERQI:mode>2): Likewise.
8996         (*add<mode>hi3): Likewise.
8997         (*xor<mode>hi3): Likewise.
8998         (<optab>di3): Likewise.
8999         (*<optab>si3_extend): Likewise.
9000         (*sge<u>_<X:mode><GPR:mode>): Likewise.
9001         (*slt<u>_<X:mode><GPR:mode>): Likewise.
9002         (*sle<u>_<X:mode><GPR:mode>): Likewise.
9004 2017-11-04  Andrew Waterman  <andrew@sifive.com>
9006         * config/riscv/riscv.c (riscv_option_override): Conditionally set
9007         TARGET_STRICT_ALIGN based upon -mtune argument.
9009 2017-11-04  Andrew Waterman  <andrew@sifive.com>
9011         * config/riscv/riscv.h (SLOW_BYTE_ACCESS): Change to 1.
9013 2017-11-04  Daniel Santos  <daniel.santos@pobox.com>
9015         * config/i386/i386.c (choose_basereg): Use optional scratch
9016         register and add assertion.
9017         (x86_emit_outlined_ms2sysv_save): Use scratch register when
9018         needed, and don't allocate stack.
9019         (ix86_expand_prologue): Rearrange where SSE saves/stub call is
9020         emitted, correct wrong allocation with -mcall-ms2sysv-xlogues.
9021         (ix86_emit_outlined_ms2sysv_restore): Fix non-immediate offsets.
9023 2017-11-03  Jeff Law  <law@redhat.com>
9025         * config/i386/i386.c (ix86_emit_restore_reg_using_pop): Prototype.
9026         (ix86_adjust_stack_and_probe_stack_clash): Use a push/pop sequence
9027         to probe at the start of a noreturn function.
9029 2017-11-03  Jakub Jelinek  <jakub@redhat.com>
9031         PR tree-optimization/78821
9032         * gimple-ssa-store-merging.c: Update the file comment.
9033         (MAX_STORE_ALIAS_CHECKS): Define.
9034         (struct store_operand_info): New type.
9035         (store_operand_info::store_operand_info): New constructor.
9036         (struct store_immediate_info): Add rhs_code and ops data members.
9037         (store_immediate_info::store_immediate_info): Add rhscode, op0r
9038         and op1r arguments to the ctor, initialize corresponding data members.
9039         (struct merged_store_group): Add load_align_base and load_align
9040         data members.
9041         (merged_store_group::merged_store_group): Initialize them.
9042         (merged_store_group::do_merge): Update them.
9043         (merged_store_group::apply_stores): Pick the constant for
9044         encode_tree_to_bitpos from one of the two operands, or skip
9045         encode_tree_to_bitpos if neither operand is a constant.
9046         (class pass_store_merging): Add process_store method decl.  Remove
9047         bool argument from terminate_all_aliasing_chains method decl.
9048         (pass_store_merging::terminate_all_aliasing_chains): Remove
9049         var_offset_p argument and corresponding handling.
9050         (stmts_may_clobber_ref_p): New function.
9051         (compatible_load_p): New function.
9052         (imm_store_chain_info::coalesce_immediate_stores): Terminate group
9053         if there is overlap and rhs_code is not INTEGER_CST.  For
9054         non-overlapping stores terminate group if rhs is not mergeable.
9055         (get_alias_type_for_stmts): Change first argument from
9056         auto_vec<gimple *> & to vec<gimple *> &.  Add IS_LOAD, CLIQUEP and
9057         BASEP arguments.  If IS_LOAD is true, look at rhs1 of the stmts
9058         instead of lhs.  Compute *CLIQUEP and *BASEP in addition to the
9059         alias type.
9060         (get_location_for_stmts): Change first argument from
9061         auto_vec<gimple *> & to vec<gimple *> &.
9062         (struct split_store): Remove orig_stmts data member, add orig_stores.
9063         (split_store::split_store): Create orig_stores rather than orig_stmts.
9064         (find_constituent_stmts): Renamed to ...
9065         (find_constituent_stores): ... this.  Change second argument from
9066         vec<gimple *> * to vec<store_immediate_info *> *, push pointers
9067         to info structures rather than the statements.
9068         (split_group): Rename ALLOW_UNALIGNED argument to
9069         ALLOW_UNALIGNED_STORE, add ALLOW_UNALIGNED_LOAD argument and handle
9070         it.  Adjust find_constituent_stores caller.
9071         (imm_store_chain_info::output_merged_store): Handle rhs_code other
9072         than INTEGER_CST, adjust split_group, get_alias_type_for_stmts and
9073         get_location_for_stmts callers.  Set MR_DEPENDENCE_CLIQUE and
9074         MR_DEPENDENCE_BASE on the MEM_REFs if they are the same in all stores.
9075         (mem_valid_for_store_merging): New function.
9076         (handled_load): New function.
9077         (pass_store_merging::process_store): New method.
9078         (pass_store_merging::execute): Use process_store method.  Adjust
9079         terminate_all_aliasing_chains caller.
9081 2017-11-03  Wilco Dijkstra  <wdijkstr@arm.com>
9083         * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
9084         Return true for more constants, symbols and label references.
9085         (aarch64_valid_floating_const): Remove unused function.
9087 2017-11-03  Wilco Dijkstra  <wdijkstr@arm.com>
9089         PR target/82786
9090         * config/aarch64/aarch64.c (aarch64_layout_frame):
9091         Undo forcing of LR at bottom of frame.
9093 2017-11-03  Jeff Law  <law@redhat.com>
9095         PR target/82823
9096         * config/i386/i386.c (ix86_expand_prologue): Tighten assert
9097         for int_registers_saved.
9099         * cfganal.c (single_pred_edge_ignoring_loop_edges): New function
9100         extracted from tree-ssa-dom.c.
9101         * cfganal.h (single_pred_edge_ignoring_loop_edges): Prototype.
9102         * tree-ssa-dom.c (single_incoming_edge_ignoring_loop_edges): Remove.
9103         (record_equivalences_from_incoming_edge): Add additional argument
9104         to single_pred_edge_ignoring_loop_edges call.
9105         * tree-ssa-uncprop.c (single_incoming_edge_ignoring_loop_edges): Remove.
9106         (uncprop_dom_walker::before_dom_children): Add additional argument
9107         to single_pred_edge_ignoring_loop_edges call.
9108         * tree-ssa-sccvn.c (sccvn_dom_walker::before_dom_children): Use
9109         single_pred_edge_ignoring_loop_edges rather than open coding.
9110         * tree-vrp.c (evrp_dom_walker::before_dom_children): Similarly.
9112 2017-11-03  Marc Glisse  <marc.glisse@inria.fr>
9114         * match.pd (-(-A)): Rewrite.
9116 2017-11-03  Segher Boessenkool  <segher@kernel.crashing.org>
9118         * config/rs6000/rs60000-protos.h (rs6000_emit_sISEL): Delete.
9119         (rs6000_emit_int_cmove): New declaration.
9120         * config/rs6000/rs6000.c (rs6000_emit_int_cmove): Delete declaration.
9121         (rs6000_emit_sISEL): Delete.
9122         (rs6000_emit_int_cmove): Make non-static.
9123         * config/rs6000/rs6000.md (cstore<mode>4): Use rs6000_emit_int_cmove
9124         instead of rs6000_emit_sISEL.
9126 2017-11-03  Jan Hubicka  <hubicka@ucw.cz>
9128         * asan.c (create_cond_insert_point): Maintain profile.
9129         * ipa-utils.c (ipa_merge_profiles): Be sure only IPA profiles are
9130         merged.
9131         * basic-block.h (struct basic_block_def): Remove frequency.
9132         (EDGE_FREQUENCY): Use to_frequency
9133         * bb-reorder.c (push_to_next_round_p): Use only IPA counts for global
9134         heuristics.
9135         (find_traces): Update to use to_frequency.
9136         (find_traces_1_round): Likewise; use only IPA counts.
9137         (bb_to_key): Likewise.
9138         (connect_traces): Use IPA counts only.
9139         (copy_bb_p): Update to use to_frequency.
9140         (fix_up_crossing_landing_pad): Likewise.
9141         (sanitize_hot_paths): Likewise.
9142         * bt-load.c (basic_block_freq): Likewise.
9143         * cfg.c (init_flow): Set count_max to uninitialized.
9144         (check_bb_profile): Remove frequencies; check counts.
9145         (dump_bb_info): Do not dump frequencies.
9146         (update_bb_profile_for_threading): Update counts only.
9147         (scale_bbs_frequencies_int): Likewise.
9148         (MAX_SAFE_MULTIPLIER): Remove.
9149         (scale_bbs_frequencies_gcov_type): Update counts only.
9150         (scale_bbs_frequencies_profile_count): Update counts only.
9151         (scale_bbs_frequencies): Update counts only.
9152         * cfg.h (struct control_flow_graph): Add count-max.
9153         (update_bb_profile_for_threading): Update prototype.
9154         * cfgbuild.c (find_bb_boundaries): Do not update frequencies.
9155         (find_many_sub_basic_blocks): Likewise.
9156         * cfgcleanup.c (try_forward_edges): Likewise.
9157         (try_crossjump_to_edge): Likewise.
9158         * cfgexpand.c (expand_gimple_cond): Likewise.
9159         (expand_gimple_tailcall): Likewise.
9160         (construct_init_block): Likewise.
9161         (construct_exit_block): Likewise.
9162         * cfghooks.c (verify_flow_info): Check consistency of counts.
9163         (dump_bb_for_graph): Do not dump frequencies.
9164         (split_block_1): Do not update frequencies.
9165         (split_edge): Do not update frequencies.
9166         (make_forwarder_block): Do not update frequencies.
9167         (duplicate_block): Do not update frequencies.
9168         (account_profile_record): Do not update frequencies.
9169         * cfgloop.c (find_subloop_latch_edge_by_profile): Use IPA counts
9170         for global heuristics.
9171         * cfgloopanal.c (average_num_loop_insns): Update to use to_frequency.
9172         (expected_loop_iterations_unbounded): Use counts only.
9173         * cfgloopmanip.c (scale_loop_profile): Simplify.
9174         (create_empty_loop_on_edge): Simplify
9175         (loopify): Simplify
9176         (duplicate_loop_to_header_edge): Simplify
9177         * cfgrtl.c (force_nonfallthru_and_redirect): Update profile.
9178         (update_br_prob_note): Take care of removing note when profile
9179         becomes undefined.
9180         (relink_block_chain): Do not dump frequency.
9181         (rtl_account_profile_record): Use to_frequency.
9182         * cgraph.c (symbol_table::create_edge): Convert count to ipa count.
9183         (cgraph_edge::redirect_call_stmt_to_calle): Conver tcount to ipa count.
9184         (cgraph_update_edges_for_call_stmt_node): Likewise.
9185         (cgraph_edge::verify_count_and_frequency): Update.
9186         (cgraph_node::verify_node): Temporarily disable frequency verification.
9187         * cgraphbuild.c (compute_call_stmt_bb_frequency): Use
9188         to_cgraph_frequency.
9189         (cgraph_edge::rebuild_edges): Convert to ipa counts.
9190         * cgraphunit.c (init_lowered_empty_function): Do not initialize
9191         frequencies.
9192         (cgraph_node::expand_thunk): Update profile.
9193         * except.c (dw2_build_landing_pads): Do not update frequency.
9194         * final.c (compute_alignments): Use to_frequency.
9195         (dump_basic_block_info): Do not dump frequency.
9196         * gimple-pretty-print.c (dump_profile): Do not dump frequency.
9197         (dump_gimple_bb_header): Do not dump frequency.
9198         * gimple-ssa-isolate-paths.c (isolate_path): Do not update frequency;
9199         do update count.
9200         * gimple-streamer-in.c (input_bb): Do not stream frequency.
9201         * gimple-streamer-out.c (output_bb): Do not stream frequency.
9202         * haifa-sched.c (sched_pressure_start_bb): Use to_freuqency.
9203         (init_before_recovery): Do not update frequency.
9204         (sched_create_recovery_edges): Do not update frequency.
9205         * hsa-gen.c (convert_switch_statements): Do not update frequency.
9206         * ipa-cp.c (ipcp_propagate_stage): Update search for max_count.
9207         (ipa_cp_c_finalize): Set max_count to uninitialized.
9208         * ipa-fnsummary.c (get_minimal_bb): Use counts.
9209         (param_change_prob): Use counts.
9210         * ipa-profile.c (ipa_profile_generate_summary): Do not summarize
9211         local profiles.
9212         * ipa-split.c (consider_split): Use to_frequency.
9213         (split_function): Use to_frequency.
9214         * ira-build.c (loop_compare_func): Likewise.
9215         (mark_loops_for_removal): Likewise.
9216         (mark_all_loops_for_removal): Likewise.
9217         * loop-doloop.c (doloop_modify): Do not update frequency.
9218         * loop-unroll.c (unroll_loop_runtime_iterations): Do not update
9219         frequency.
9220         * lto-streamer-in.c (input_function): Update count_max.
9221         * omp-expand.c (expand_omp_taskreg): Update count_max.
9222         * omp-simd-clone.c (simd_clone_adjust): Update profile.
9223         * predict.c (maybe_hot_frequency_p): Use to_frequency.
9224         (maybe_hot_count_p): Use ipa counts only.
9225         (maybe_hot_bb_p): Simplify.
9226         (maybe_hot_edge_p): Simplify.
9227         (probably_never_executed): Do not take frequency argument.
9228         (probably_never_executed_bb_p): Do not pass frequency.
9229         (probably_never_executed_edge_p): Likewise.
9230         (combine_predictions_for_bb): Check that profile is nonzero.
9231         (propagate_freq): Do not set frequency.
9232         (drop_profile): Simplify.
9233         (counts_to_freqs): Simplify.
9234         (expensive_function_p): Use to_frequency.
9235         (propagate_unlikely_bbs_forward): Simplify.
9236         (determine_unlikely_bbs): Simplify.
9237         (estimate_bb_frequencies): Add hack to silence graphite issues.
9238         (compute_function_frequency): Use ipa counts.
9239         (pass_profile::execute): Update.
9240         (rebuild_frequencies): Use counts only.
9241         (force_edge_cold): Use counts only.
9242         * profile-count.c (profile_count::dump): Dump new count types.
9243         (profile_count::differs_from_p): Check compatiblity.
9244         (profile_count::to_frequency): New function.
9245         (profile_count::to_cgraph_frequency): New function.
9246         * profile-count.h (struct function): Declare.
9247         (enum profile_quality): Add profile_guessed_local and
9248         profile_guessed_global0.
9249         (class profile_proability): Decrease number of bits to 29;
9250         update from_reg_br_prob_note and to_reg_br_prob_note.
9251         (class profile_count: Update comment; decrease number of bits
9252         to 61. Check compatibility.
9253         (profile_count::compatible_p): New private member function.
9254         (profile_count::ipa_p): New member function.
9255         (profile_count::operator<): Handle global zero correctly.
9256         (profile_count::operator>): Handle global zero correctly.
9257         (profile_count::operator<=): Handle global zero correctly.
9258         (profile_count::operator>=): Handle global zero correctly.
9259         (profile_count::nonzero_p): New member function.
9260         (profile_count::force_nonzero): New member function.
9261         (profile_count::max): New member function.
9262         (profile_count::apply_scale): Handle IPA scalling.
9263         (profile_count::guessed_local): New member function.
9264         (profile_count::global0): New member function.
9265         (profile_count::ipa): New member function.
9266         (profile_count::to_frequency): Declare.
9267         (profile_count::to_cgraph_frequency): Declare.
9268         * profile.c (OVERLAP_BASE): Delete.
9269         (compute_frequency_overlap): Delete.
9270         (compute_branch_probabilities): Do not use compute_frequency_overlap.
9271         * regs.h (REG_FREQ_FROM_BB): Use to_frequency.
9272         * sched-ebb.c (rank): Use counts only.
9273         * shrink-wrap.c (handle_simple_exit): Use counts only.
9274         (try_shrink_wrapping): Use counts only.
9275         (place_prologue_for_one_component): Use counts only.
9276         * tracer.c (find_best_predecessor): Use to_frequency.
9277         (find_trace): Use to_frequency.
9278         (tail_duplicate): Use to_frequency.
9279         * trans-mem.c (expand_transaction): Do not update frequency.
9280         * tree-call-cdce.c: Do not update frequency.
9281         * tree-cfg.c (gimple_find_sub_bbs): Likewise.
9282         (gimple_merge_blocks): Likewise.
9283         (gimple_split_edge): Likewise.
9284         (gimple_duplicate_sese_region): Likewise.
9285         (gimple_duplicate_sese_tail): Likewise.
9286         (move_sese_region_to_fn): Likewise.
9287         (gimple_account_profile_record): Likewise.
9288         (insert_cond_bb): Likewise.
9289         * tree-complex.c (expand_complex_div_wide): Likewise.
9290         * tree-eh.c (lower_resx): Update profile.
9291         * tree-inline.c (copy_bb): Simplify count scaling; do not scale
9292         frequencies.
9293         (initialize_cfun): Do not initialize frequencies
9294         (freqs_to_counts): Delete.
9295         (copy_cfg_body): Ignore count parameter.
9296         (copy_body): Update.
9297         (expand_call_inline): Update count_max.
9298         (optimize_inline_calls): Update count_max.
9299         (tree_function_versioning): Update count_max.
9300         * tree-ssa-coalesce.c (coalesce_cost_bb): Use to_frequency.
9301         * tree-ssa-ifcombine.c (update_profile_after_ifcombine): Do not update
9302         frequency.
9303         * tree-ssa-loop-im.c (execute_sm_if_changed): Use counts only.
9304         * tree-ssa-loop-ivcanon.c (unloop_loops): Do not update freuqency.
9305         (try_peel_loop): Likewise.
9306         * tree-ssa-loop-ivopts.c (get_scaled_computation_cost_at): Use
9307         to_frequency.
9308         * tree-ssa-loop-manip.c (niter_for_unrolled_loop): Pass -1.
9309         (tree_transform_and_unroll_loop): Do not use frequencies
9310         * tree-ssa-loop-niter.c (estimate_numbers_of_iterations):
9311         Use reliable prediction only.
9312         * tree-ssa-loop-unswitch.c (hoist_guard): Do not use frequencies.
9313         * tree-ssa-sink.c (select_best_block): Use to_frequency.
9314         * tree-ssa-tail-merge.c (replace_block_by): Temporarily disable
9315         probability scaling.
9316         * tree-ssa-threadupdate.c (create_block_for_threading): Do
9317         not update frequency
9318         (any_remaining_duplicated_blocks): Likewise.
9319         (update_profile): Likewise.
9320         (estimated_freqs_path): Delete.
9321         (freqs_to_counts_path): Delete.
9322         (clear_counts_path): Delete.
9323         (ssa_fix_duplicate_block_edges): Likewise.
9324         (duplicate_thread_path): Likewise.
9325         * tree-switch-conversion.c (gen_inbound_check): Use counts.
9326         * tree-tailcall.c (decrease_profile): Do not update frequency.
9327         (eliminate_tail_call): Likewise.
9328         * tree-vect-loop-manip.c (vect_do_peeling): Likewise.
9329         * tree-vect-loop.c (scale_profile_for_vect_loop): Likewise.
9330         (optimize_mask_stores): Likewise.
9331         * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
9332         * ubsan.c (ubsan_expand_null_ifn): Update profile.
9333         (ubsan_expand_ptr_ifn): Update profile.
9334         * value-prof.c (gimple_ic): Simplify.
9335         * value-prof.h (gimple_ic): Update prototype.
9336         * ipa-inline-transform.c (inline_transform): Fix scaling conditoins.
9337         * ipa-inline.c (compute_uninlined_call_time): Be sure that
9338         counts are nonzero.
9339         (want_inline_self_recursive_call_p): Likewise.
9340         (resolve_noninline_speculation): Only cummulate defined counts.
9341         (inline_small_functions): Use nonzero_p.
9342         (ipa_inline): Do not access freed node.
9344 2017-11-03  Wilco Dijkstra  <wdijkstr@arm.com>
9346         * config/aarch64/aarch64.c (aarch64_override_options_internal):
9347         Set PARAM_SCHED_PRESSURE_ALGORITHM to SCHED_PRESSURE_MODEL.
9349 2017-11-03  Kito Cheng  <kito.cheng@gmail.com>
9351         * config/riscv/riscv.c (riscv_legitimize_move): Handle
9352         non-legitimate address.
9354 2017-11-03  Segher Boessenkool  <segher@kernel.crashing.org>
9356         * config/rs6000/rs6000.md (*lt0_disi): Delete.
9357         (*lt0_<mode>di, *lt0_<mode>si): New.
9359 2017-11-03  Segher Boessenkool  <segher@kernel.crashing.org>
9361         * config/rs6000/rs6000.md (move_from_CR_ov_bit): Change condition to
9362         TARGET_PAIRED_FLOAT.
9364 2017-11-03  Siddhesh Poyarekar  <siddhesh.poyarekar@linaro.org>
9365             Jim Wilson  <jim.wilson@linaro.org>
9367         * config/aarch64/aarch64-cores.def (saphira): New CPU.
9368         * config/aarch64/aarch64-tune.md: Regenerated.
9369         * doc/invoke.texi (AArch64 Options/-mtune): Add "saphira".
9370         * gcc/config/aarch64/aarch64.c (saphira_tunings): New tuning table.
9372 2017-11-03  Cupertino Miranda  <cmiranda@synopsys.com>
9374         * config/arc/arc.c (arc_save_restore): Corrected CFA note.
9375         (arc_expand_prologue): Restore blink for millicode.
9376         * config/arc/linux.h (LINK_EH_SPEC): Defined.
9378 2017-11-03  Richard Sandiford  <richard.sandiford@linaro.org>
9380         PR target/82809
9381         * config/i386/i386.c (ix86_vector_duplicate_value): Use
9382         gen_vec_duplicate after forcing the scalar into a register.
9384 2017-11-02  Segher Boessenkool  <segher@kernel.crashing.org>
9386         * combine (try_combine): Print the insns input to try_combine to the
9387         dump file.
9389 2017-11-02  Steve Ellcey  <sellcey@cavium.com>
9391         PR target/79868
9392         * config/aarch64/aarch64-c.c (aarch64_pragma_target_parse):
9393         Remove second argument from aarch64_process_target_attr call.
9394         * config/aarch64/aarch64-protos.h (aarch64_process_target_attr):
9395         Ditto.
9396         * config/aarch64/aarch64.c (aarch64_attribute_info): Change
9397         field type.
9398         (aarch64_handle_attr_arch): Remove second argument.
9399         (aarch64_handle_attr_cpu): Ditto.
9400         (aarch64_handle_attr_tune): Ditto.
9401         (aarch64_handle_attr_isa_flags): Ditto.
9402         (aarch64_process_one_target_attr): Ditto.
9403         (aarch64_process_target_attr): Ditto.
9404         (aarch64_option_valid_attribute_p): Remove second argument.
9405         on aarch64_process_target_attr call.
9407 2017-11-02  David Malcolm  <dmalcolm@redhat.com>
9409         * diagnostic.c: Include "selftest-diagnostic.h".
9410         (selftest::assert_location_text): New function.
9411         (selftest::test_diagnostic_get_location_text): New function.
9412         (selftest::diagnostic_c_tests): Call it.
9414 2017-11-02  David Malcolm  <dmalcolm@redhat.com>
9416         * Makefile.in (OBJS-libcommon): Add selftest-diagnostic.o.
9417         * diagnostic-show-locus.c: Include "selftest-diagnostic.h".
9418         (class selftest::test_diagnostic_context): Move to...
9419         * selftest-diagnostic.c: New file.
9420         * selftest-diagnostic.h: New file.
9422 2017-11-02  James Bowman  <james.bowman@ftdichip.com>
9424         * config/ft32/ft32.c (ft32_addr_space_legitimate_address_p): increase
9425         offset range for FT32B.
9426         * config/ft32/ft32.h: option "mcompress" enables relaxation.
9427         * config/ft32/ft32.md: Add TARGET_NOPM.
9428         * config/ft32/ft32.opt: Add mft32b, mcompress, mnopm.
9429         * gcc/doc/invoke.texi: Add mft32b, mcompress, mnopm.
9431 2017-11-02  Wilco Dijkstra  <wdijkstr@arm.com>
9433         * config/aarch64/aarch64.h (MALLOC_ABI_ALIGNMENT): New define.
9435 2017-11-02  Jeff Law  <law@redhat.com>
9437         * gimple-ssa-sprintf.c (sprintf_dom_walker): Remove
9438         virtual keyword on FINAL OVERRIDE members.
9440         * tree-ssa-propagate.h (ssa_propagation_engine): Group
9441         virtuals together.  Add virtual destructor.
9442         (substitute_and_fold_engine): Similarly.
9444 2017-11-02  Jan Hubicka  <hubicka@ucw.cz>
9446         * x86-tune.def (X86_TUNE_USE_INCDEC): Enable for Haswell+.
9448 2017-11-02  Richard Biener  <rguenther@suse.de>
9450         PR tree-optimization/82795
9451         * tree-if-conv.c (predicate_mem_writes): Remove bogus assert.
9453 2017-11-02  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
9455         * acinclude.m4 (gcc_AC_INITFINI_ARRAY): Don't require
9456         gcc_SUN_LD_VERSION.
9457         (gcc_GAS_CHECK_FEATURE): Remove.
9458         * configure.ac (ld_vers) <*-*-solaris2*>: Move comments from
9459         gcc_AC_INITFINI_ARRAY here.  Update for Solaris 11.4 changes.
9460         * configure: Regenerate.
9462 2017-11-02  Claudiu Zissulescu <claziss@synopsys.com>
9464         * config/arc/arc.c (hwloop_optimize): Account for empty
9465         body loops.
9467 2017-11-02  Richard Biener  <rguenther@suse.de>
9469         PR middle-end/82765
9470         * varasm.c (decode_addr_const): Make offset HOST_WIDE_INT.
9471         Truncate ARRAY_REF index and element size.
9473 2017-11-01  Palmer Dabbelt  <palmer@dabbelt.com>
9475         * doc/invoke.texi (RISC-V Options): Use "@minus{}2 GB", not "-2 GB".
9477 2017-11-01  Jeff Law  <law@redhat.com>
9479         * tree-ssa-ccp.c (ccp_folder): New class derived from
9480         substitute_and_fold_engine.
9481         (ccp_folder::get_value): New member function.
9482         (ccp_folder::fold_stmt): Renamed from ccp_fold_stmt.
9483         (ccp_fold_stmt): Remove prototype.
9484         (ccp_finalize): Call substitute_and_fold from the ccp_class.
9485         * tree-ssa-copy.c (copy_folder): New class derived from
9486         substitute_and_fold_engine.
9487         (copy_folder::get_value): Renamed from get_value.
9488         (fini_copy_prop): Call substitute_and_fold from copy_folder class.
9489         * tree-vrp.c (vrp_folder): New class derived from
9490         substitute_and_fold_engine.
9491         (vrp_folder::fold_stmt): Renamed from vrp_fold_stmt.
9492         (vrp_folder::get_value): New member function.
9493         (vrp_finalize): Call substitute_and_fold from vrp_folder class.
9494         (evrp_dom_walker::before_dom_children): Similarly for replace_uses_in.
9495         * tree-ssa-propagate.h (substitute_and_fold_engine): New class to
9496         provide a class interface to folder/substitute routines.
9497         (ssa_prop_fold_stmt_fn): Remove typedef.
9498         (ssa_prop_get_value_fn): Likewise.
9499         (subsitute_and_fold): Remove prototype.
9500         (replace_uses_in): Likewise.
9501         * tree-ssa-propagate.c (substitute_and_fold_engine::replace_uses_in):
9502         Renamed from replace_uses_in.  Call the virtual member function
9503         (substitute_and_fold_engine::replace_phi_args_in): Similarly.
9504         (substitute_and_fold_dom_walker): Remove initialization of
9505         data member entries for calbacks.  Add substitute_and_fold_engine
9506         member and initialize it.
9507         (substitute_and_fold_dom_walker::before_dom_children0: Use the
9508         member functions for get_value, replace_phi_args_in c
9509         replace_uses_in, and fold_stmt calls.
9510         (substitute_and_fold_engine::substitute_and_fold): Renamed from
9511         substitute_and_fold.  Remove assert.   Update ctor call.
9513         * tree-ssa-propagate.h (ssa_prop_visit_stmt_fn): Remove typedef.
9514         (ssa_prop_visit_phi_fn): Likewise.
9515         (class ssa_propagation_engine): New class to provide an interface
9516         into ssa_propagate.
9517         * tree-ssa-propagate.c (ssa_prop_visit_stmt): Remove file scoped
9518         variable.
9519         (ssa_prop_visit_phi): Likewise.
9520         (ssa_propagation_engine::simulate_stmt): Moved into class.
9521         Call visit_phi/visit_stmt from the class rather than via
9522         file scoped static variables.
9523         (ssa_propagation_engine::simulate_block): Moved into class.
9524         (ssa_propagation_engine::process_ssa_edge_worklist): Similarly.
9525         (ssa_propagation_engine::ssa_propagate): Similarly.  No longer
9526         set file scoped statics for the visit_stmt/visit_phi callbacks.
9527         * tree-complex.c (complex_propagate): New class derived from
9528         ssa_propagation_engine.
9529         (complex_propagate::visit_stmt): Renamed from complex_visit_stmt.
9530         (complex_propagate::visit_phi): Renamed from complex_visit_phi.
9531         (tree_lower_complex): Call ssa_propagate via the complex_propagate
9532         class.
9533         * tree-ssa-ccp.c: (ccp_propagate): New class derived from
9534         ssa_propagation_engine.
9535         (ccp_propagate::visit_phi): Renamed from ccp_visit_phi_node.
9536         (ccp_propagate::visit_stmt): Renamed from ccp_visit_stmt.
9537         (do_ssa_ccp): Call ssa_propagate from the ccp_propagate class.
9538         * tree-ssa-copy.c (copy_prop): New class derived from
9539         ssa_propagation_engine.
9540         (copy_prop::visit_stmt): Renamed from copy_prop_visit_stmt.
9541         (copy_prop::visit_phi): Renamed from copy_prop_visit_phi_node.
9542         (execute_copy_prop): Call ssa_propagate from the copy_prop class.
9543         * tree-vrp.c (vrp_prop): New class derived from ssa_propagation_engine.
9544         (vrp_prop::visit_stmt): Renamed from vrp_visit_stmt.
9545         (vrp_prop::visit_phi): Renamed from vrp_visit_phi_node.
9546         (execute_vrp): Call ssa_propagate from the vrp_prop class.
9548 2017-11-01  Jakub Jelinek  <jakub@redhat.com>
9550         PR rtl-optimization/82778
9551         PR rtl-optimization/82597
9552         * compare-elim.c (struct comparison): Add in_a_setter field.
9553         (find_comparison_dom_walker::before_dom_children): Remove killed
9554         bitmap and df_simulate_find_defs call, instead walk the defs.
9555         Compute last_setter and initialize in_a_setter.  Merge definitions
9556         with first initialization for a few variables.
9557         (try_validate_parallel): Use insn_invalid_p instead of
9558         recog_memoized.  Return insn rather than just the pattern.
9559         (try_merge_compare): Fix up comment.  Don't uselessly test if
9560         in_a is a REG_P.  Use cmp->in_a_setter instead of walking UD
9561         chains.
9562         (execute_compare_elim_after_reload): Remove df_chain_add_problem
9563         call.
9565 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9566             Alan Hayward  <alan.hayward@arm.com>
9567             David Sherwood  <david.sherwood@arm.com>
9569         * config/aarch64/aarch64.c (aarch64_rtx_costs): Use
9570         aarch64_hard_regno_nregs to get the number of registers
9571         in a mode.
9573 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9574             Alan Hayward  <alan.hayward@arm.com>
9575             David Sherwood  <david.sherwood@arm.com>
9577         * config/aarch64/constraints.md (Upl): Rename to...
9578         (Uaa): ...this.
9579         * config/aarch64/aarch64.md
9580         (*zero_extend<SHORT:mode><GPI:mode>2_aarch64, *addsi3_aarch64_uxtw):
9581         Update accordingly.
9583 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9584             Alan Hayward  <alan.hayward@arm.com>
9585             David Sherwood  <david.sherwood@arm.com>
9587         * config/aarch64/aarch64.c (aarch64_add_constant_internal)
9588         (aarch64_add_constant, aarch64_add_sp, aarch64_sub_sp): Move
9589         earlier in file.
9591 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9592             Alan Hayward  <alan.hayward@arm.com>
9593             David Sherwood  <david.sherwood@arm.com>
9595         * config/aarch64/aarch64.c (aarch64_evpc_trn, aarch64_evpc_uzp)
9596         (aarch64_evpc_zip, aarch64_evpc_ext, aarch64_evpc_rev)
9597         (aarch64_evpc_dup): Generate rtl direcly, rather than using
9598         named expanders.
9599         (aarch64_expand_vec_perm_const_1): Explicitly check for permutes
9600         of a single element.
9601         * config/aarch64/iterators.md: Add a comment above the permute
9602         unspecs to say that they are generated directly by
9603         aarch64_expand_vec_perm_const.
9604         * config/aarch64/aarch64-simd.md: Likewise the permute instructions.
9606 2017-11-01  Nathan Sidwell  <nathan@acm.org>
9608         * tree-dump.c (dequeue_and_dump): Use HAS_DECL_ASSEMBLER_NAME_P.
9610 2017-11-01  Palmer Dabbelt  <palmer@dabbelt.com>
9612         * doc/invoke.texi (RISC-V Options): Explicitly name the medlow
9613         and medany code models, and describe what they do.
9615 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9617         Revert accidental duplicate:
9619         * combine.c (can_change_dest_mode): Reject changes in
9620         REGMODE_NATURAL_SIZE.
9622 2017-11-01  Segher Boessenkool  <segher@kernel.crashing.org>
9624         PR rtl-optimization/64682
9625         PR rtl-optimization/69567
9626         PR rtl-optimization/69737
9627         PR rtl-optimization/82683
9628         * combine.c (distribute_notes) <REG_DEAD>: If the new I2 sets the same
9629         register mentioned in the note, drop the note, unless it came from I3,
9630         in which case it should go to I3 again.
9632 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9634         * tree-ssa-dse.c (normalize_ref): Check whether the ranges overlap
9635         and return false if not.
9636         (clear_bytes_written_by, live_bytes_read): Update accordingly.
9638 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9640         * tree-ssa-alias.h (ranges_overlap_p): Return false if either
9641         range is known to be empty.
9643 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9644             Alan Hayward  <alan.hayward@arm.com>
9645             David Sherwood  <david.sherwood@arm.com>
9647         * simplify-rtx.c (simplify_const_unary_operation): Use GET_MODE_NUNITS
9648         and CONST_VECTOR_NUNITS instead of computing the number of units from
9649         the byte sizes of the vector and element.
9650         (simplify_binary_operation_1): Likewise.
9651         (simplify_const_binary_operation): Likewise.
9652         (simplify_ternary_operation): Likewise.
9654 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9655             Alan Hayward  <alan.hayward@arm.com>
9656             David Sherwood  <david.sherwood@arm.com>
9658         * var-tracking.c (INT_MEM_OFFSET): Replace with...
9659         (int_mem_offset): ...this new function.
9660         (var_mem_set, var_mem_delete_and_set, var_mem_delete)
9661         (find_mem_expr_in_1pdv, dataflow_set_preserve_mem_locs)
9662         (same_variable_part_p, use_type, add_stores, vt_get_decl_and_offset):
9663         Update accordingly.
9665 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9666             Alan Hayward  <alan.hayward@arm.com>
9667             David Sherwood  <david.sherwood@arm.com>
9669         * lower-subreg.c (interesting_mode_p): New function.
9670         (compute_costs, find_decomposable_subregs, decompose_register)
9671         (simplify_subreg_concatn, can_decompose_p, resolve_simple_move)
9672         (resolve_clobber, dump_choices): Use it.
9674 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9675             Alan Hayward  <alan.hayward@arm.com>
9676             David Sherwood  <david.sherwood@arm.com>
9678         * rtlhash.c (add_rtx): Use add_hwi for 'w' and add_int for 'i'.
9680 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9681             Alan Hayward  <alan.hayward@arm.com>
9682             David Sherwood  <david.sherwood@arm.com>
9684         * alias.c (find_base_value, find_base_term): Only process integer
9685         truncations.  Check the precision rather than the size.
9687 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9688             Alan Hayward  <alan.hayward@arm.com>
9689             David Sherwood  <david.sherwood@arm.com>
9691         * machmode.h (is_narrower_int_mode): New function
9692         * optabs.c (expand_float, expand_fix): Use it.
9693         * dwarf2out.c (rotate_loc_descriptor): Likewise.
9695 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9696             Alan Hayward  <alan.hayward@arm.com>
9697             David Sherwood  <david.sherwood@arm.com>
9699         * rtl.h (narrower_subreg_mode): New function.
9700         * ira-color.c (update_costs_from_allocno): Use it.
9702 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9703             Alan Hayward  <alan.hayward@arm.com>
9704             David Sherwood  <david.sherwood@arm.com>
9706         * optabs-query.h (convert_optab_p): New function, split out from...
9707         (convert_optab_handler): ...here.
9708         (widening_optab_handler): Delete.
9709         (find_widening_optab_handler): Remove permit_non_widening parameter.
9710         (find_widening_optab_handler_and_mode): Likewise.  Provide an
9711         override that operates on mode class wrappers.
9712         * optabs-query.c (widening_optab_handler): Delete.
9713         (find_widening_optab_handler_and_mode): Remove permit_non_widening
9714         parameter.  Assert that the two modes are the same class and that
9715         the "from" mode is narrower than the "to" mode.  Use
9716         convert_optab_handler instead of widening_optab_handler.
9717         * expmed.c (expmed_mult_highpart_optab): Use convert_optab_handler
9718         instead of widening_optab_handler.
9719         * expr.c (expand_expr_real_2): Update calls to
9720         find_widening_optab_handler.
9721         * optabs.c (expand_widen_pattern_expr): Likewise.
9722         (expand_binop_directly): Take the insn_code as a parameter.
9723         (expand_binop): Only call find_widening_optab_handler for
9724         conversion optabs; use optab_handler otherwise.  Update calls
9725         to find_widening_optab_handler and expand_binop_directly.
9726         Use convert_optab_handler instead of widening_optab_handler.
9727         * tree-ssa-math-opts.c (convert_mult_to_widen): Update calls to
9728         find_widening_optab_handler and use scalar_mode rather than
9729         machine_mode.
9730         (convert_plusminus_to_widen): Likewise.
9732 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9733             Alan Hayward  <alan.hayward@arm.com>
9734             David Sherwood  <david.sherwood@arm.com>
9736         * machmode.h (fixed_size_mode): New class.
9737         * rtl.h (get_pool_mode): Return fixed_size_mode.
9738         * gengtype.c (main): Add fixed_size_mode.
9739         * target.def (get_raw_result_mode): Return a fixed_size_mode.
9740         (get_raw_arg_mode): Likewise.
9741         * doc/tm.texi: Regenerate.
9742         * targhooks.h (default_get_reg_raw_mode): Return a fixed_size_mode.
9743         * targhooks.c (default_get_reg_raw_mode): Likewise.
9744         * config/ia64/ia64.c (ia64_get_reg_raw_mode): Likewise.
9745         * config/mips/mips.c (mips_get_reg_raw_mode): Likewise.
9746         * config/msp430/msp430.c (msp430_get_raw_arg_mode): Likewise.
9747         (msp430_get_raw_result_mode): Likewise.
9748         * config/avr/avr-protos.h (regmask): Use as_a <fixed_side_mode>
9749         * dbxout.c (dbxout_parms): Require fixed-size modes.
9750         * expr.c (copy_blkmode_from_reg, copy_blkmode_to_reg): Likewise.
9751         * gimple-ssa-store-merging.c (encode_tree_to_bitpos): Likewise.
9752         * omp-low.c (lower_oacc_reductions): Likewise.
9753         * simplify-rtx.c (simplify_immed_subreg): Take fixed_size_modes.
9754         (simplify_subreg): Update accordingly.
9755         * varasm.c (constant_descriptor_rtx::mode): Change to fixed_size_mode.
9756         (force_const_mem): Update accordingly.  Return NULL_RTX for modes
9757         that aren't fixed-size.
9758         (get_pool_mode): Return a fixed_size_mode.
9759         (output_constant_pool_2): Take a fixed_size_mode.
9761 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9762             Alan Hayward  <alan.hayward@arm.com>
9763             David Sherwood  <david.sherwood@arm.com>
9765         * doc/rtl.texi (vec_series): Document.
9766         (const): Say that the operand can be a vec_series.
9767         * rtl.def (VEC_SERIES): New rtx code.
9768         * rtl.h (const_vec_series_p_1): Declare.
9769         (const_vec_series_p): New function.
9770         * emit-rtl.h (gen_const_vec_series): Declare.
9771         (gen_vec_series): Likewise.
9772         * emit-rtl.c (const_vec_series_p_1, gen_const_vec_series)
9773         (gen_vec_series): Likewise.
9774         * optabs.c (expand_mult_highpart): Use gen_const_vec_series.
9775         * simplify-rtx.c (simplify_unary_operation): Handle negations
9776         of vector series.
9777         (simplify_binary_operation_series): New function.
9778         (simplify_binary_operation_1): Use it.  Handle VEC_SERIES.
9779         (test_vector_ops_series): New function.
9780         (test_vector_ops): Call it.
9781         * config/powerpcspe/altivec.md (altivec_lvsl): Use
9782         gen_const_vec_series.
9783         (altivec_lvsr): Likewise.
9784         * config/rs6000/altivec.md (altivec_lvsl, altivec_lvsr): Likewise.
9786 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9787             Alan Hayward  <alan.hayward@arm.com>
9788             David Sherwood  <david.sherwood@arm.com>
9790         * doc/rtl.texi (const): Update description of address constants.
9791         Say that vector constants are allowed too.
9792         * common.md (E, F): Use CONSTANT_P instead of checking for
9793         CONST_VECTOR.
9794         * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
9795         checking for CONST_VECTOR.
9796         * expmed.c (make_tree): Use build_vector_from_val for a CONST
9797         VEC_DUPLICATE.
9798         * expr.c (expand_expr_real_2): Check for vector modes instead
9799         of checking for CONST_VECTOR.
9800         * rtl.h (const_vec_p): New function.
9801         (const_vec_duplicate_p): Check for a CONST VEC_DUPLICATE.
9802         (unwrap_const_vec_duplicate): Handle them here too.
9804 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9805             David Malcolm  <dmalcolm@redhat.com>
9806             Alan Hayward  <alan.hayward@arm.com>
9807             David Sherwood  <david.sherwood@arm.com>
9809         * rtl.h (vec_duplicate_p): New function.
9810         * selftest-rtl.c (assert_rtx_eq_at): New function.
9811         * selftest-rtl.h (ASSERT_RTX_EQ): New macro.
9812         (assert_rtx_eq_at): Declare.
9813         * selftest.h (selftest::simplify_rtx_c_tests): Declare.
9814         * selftest-run-tests.c (selftest::run_tests): Call it.
9815         * simplify-rtx.c: Include selftest.h and selftest-rtl.h.
9816         (simplify_unary_operation_1): Recursively handle vector duplicates.
9817         (simplify_binary_operation_1): Likewise.  Handle VEC_SELECTs of
9818         vector duplicates.
9819         (simplify_subreg): Handle subregs of vector duplicates.
9820         (make_test_reg, test_vector_ops_duplicate, test_vector_ops)
9821         (selftest::simplify_rtx_c_tests): New functions.
9823 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9824             Alan Hayward  <alan.hayward@arm.com>
9825             David Sherwood  <david.sherwood@arm.com>
9827         * emit-rtl.h (gen_const_vec_duplicate): Declare.
9828         (gen_vec_duplicate): Likewise.
9829         * emit-rtl.c (gen_const_vec_duplicate_1): New function, split
9830         out from...
9831         (gen_const_vector): ...here.
9832         (gen_const_vec_duplicate, gen_vec_duplicate): New functions.
9833         (gen_rtx_CONST_VECTOR): Use gen_const_vec_duplicate for constants
9834         whose elements are all equal.
9835         * optabs.c (expand_vector_broadcast): Use gen_const_vec_duplicate.
9836         * simplify-rtx.c (simplify_const_unary_operation): Likewise.
9837         (simplify_relational_operation): Likewise.
9838         * config/aarch64/aarch64.c (aarch64_simd_gen_const_vector_dup):
9839         Likewise.
9840         (aarch64_simd_dup_constant): Use gen_vec_duplicate.
9841         (aarch64_expand_vector_init): Likewise.
9842         * config/arm/arm.c (neon_vdup_constant): Likewise.
9843         (neon_expand_vector_init): Likewise.
9844         (arm_expand_vec_perm): Use gen_const_vec_duplicate.
9845         (arm_block_set_unaligned_vect): Likewise.
9846         (arm_block_set_aligned_vect): Likewise.
9847         * config/arm/neon.md (neon_copysignf<mode>): Likewise.
9848         * config/i386/i386.c (ix86_expand_vec_perm): Likewise.
9849         (expand_vec_perm_even_odd_pack): Likewise.
9850         (ix86_vector_duplicate_value): Use gen_vec_duplicate.
9851         * config/i386/sse.md (one_cmpl<mode>2): Use CONSTM1_RTX.
9852         * config/ia64/ia64.c (ia64_expand_vecint_compare): Use
9853         gen_const_vec_duplicate.
9854         * config/ia64/vect.md (addv2sf3, subv2sf3): Use CONST1_RTX.
9855         * config/mips/mips.c (mips_gen_const_int_vector): Use
9856         gen_const_vec_duplicate.
9857         (mips_expand_vector_init): Use CONST0_RTX.
9858         * config/powerpcspe/altivec.md (abs<mode>2, nabs<mode>2): Likewise.
9859         (define_split): Use gen_const_vec_duplicate.
9860         * config/rs6000/altivec.md (abs<mode>2, nabs<mode>2): Use CONST0_RTX.
9861         (define_split): Use gen_const_vec_duplicate.
9862         * config/s390/vx-builtins.md (vec_genmask<mode>): Likewise.
9863         (vec_ctd_s64, vec_ctd_u64, vec_ctsl, vec_ctul): Likewise.
9864         * config/spu/spu.c (spu_const): Likewise.
9866 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9867             Alan Hayward  <alan.hayward@arm.com>
9868             David Sherwood  <david.sherwood@arm.com>
9870         * combine.c (can_change_dest_mode): Reject changes in
9871         REGMODE_NATURAL_SIZE.
9873 2017-10-31  Sandra Loosemore  <sandra@codesourcery.com>
9875         * configure.ac (--enable-libssp): New.
9876         (gcc_cv_libc_provides_ssp): Check for explicit setting before
9877         trying to determine target-specific default.  Adjust indentation.
9878         * configure: Regenerated.
9879         * doc/install.texi (Configuration): Expand --disable-libssp
9880         documentation.
9882 2017-10-31  Daniel Santos  <daniel.santos@pobox.com>
9884         config/i386/i386.c (ix86_expand_epilogue): Correct stack
9885         calculation.
9887 2017-10-31  Martin Jambor  <mjambor@suse.cz>
9889         PR c++/81702
9890         * gimple-fold.c (gimple_get_virt_method_for_vtable): Remove assert.
9892 2017-10-31  David Malcolm  <dmalcolm@redhat.com>
9894         * auto-profile.c (autofdo_source_profile::read): Use
9895         UNKNOWN_LOCATION rather than 0.
9896         * diagnostic-core.h (warning_at_rich_loc): Rename to...
9897         (warning_at): ...this overload.
9898         (warning_at_rich_loc_n): Rename to...
9899         (warning_n): ...this overload.
9900         (error_at_rich_loc): Rename to...
9901         (error_at): ...this overload.
9902         (pedwarn_at_rich_loc): Rename to...
9903         (pedwarn): ...this overload.
9904         (permerror_at_rich_loc): Rename to...
9905         (permerror): ...this overload.
9906         (inform_at_rich_loc): Rename to...
9907         (inform): ...this overload.
9908         * diagnostic.c: (diagnostic_n_impl): Delete location_t-based decl.
9909         (diagnostic_n_impl_richloc): Rename to...
9910         (diagnostic_n_impl): ...this rich_location *-based decl.
9911         (inform_at_rich_loc): Rename to...
9912         (inform): ...this, and add an assertion.
9913         (inform_n): Update for removal of location_t-based diagnostic_n_impl.
9914         (warning_at_rich_loc): Rename to...
9915         (warning_at): ...this, and add an assertion.
9916         (warning_at_rich_loc_n): Rename to...
9917         (warning_n): ...this, and add an assertion.
9918         (warning_n): Update location_t-based implementation for removal of
9919         location_t-based diagnostic_n_impl.
9920         (pedwarn_at_rich_loc): Rename to...
9921         (pedwarn): ...this, and add an assertion.
9922         (permerror_at_rich_loc): Rename to...
9923         (permerror): ...this, and add an assertion.
9924         (error_n): Update for removal of location_t-based diagnostic_n_impl.
9925         (error_at_rich_loc): Rename to...
9926         (error_at): ...this, and add an assertion.
9927         * gcc.c (do_spec_1): Use UNKNOWN_LOCATION rather than 0.
9928         (driver::do_spec_on_infiles): Likewise.
9929         * substring-locations.c (format_warning_va): Update for renaming
9930         of inform_at_rich_loc.
9932 2017-10-31  Michael Meissner  <meissner@linux.vnet.ibm.com>
9934         * builtins.def (DEF_FLOATN_BUILTIN): Change most _Float<N> and
9935         _Float<N>X built-in functions so that the variant without the
9936         "__builtin_" prefix is only enabled for the GNU C and Objective C
9937         languages when they are in non-strict ANSI/ISO mode.
9938         (DEF_EXT_LIB_FLOATN_NX_BUILTINS): Likewise.
9939         * target.def (floatn_builtin_p): Add a target hook to control
9940         whether _Float<N> and _Float<N>X built-in functions without the
9941         "__builtin_" prefix are enabled, and return true for C and
9942         Objective C in the default hook.  Include langhooks.h in
9943         targhooks.c.
9944         * targhooks.h (default_floatn_builtin_p): Likewise.
9945         * targhooks.c (default_floatn_builtin_p): Likewise.
9946         * doc/tm.texi.in (TARGET_FLOATN_BUILTIN_P): Document the
9947         floatn_builtin_p target hook.
9948         * doc/tm.texi (TARGET_FLOATN_BUILTIN_P): Likewise.
9950 2017-10-31  Matthew Fortune  <matthew.fortune@imgtec.com>
9951             Eric Botcazou  <ebotcazou@adacore.com>
9953         PR rtl-optimization/81803
9954         * lra-constraints.c (curr_insn_transform): Also reload the whole
9955         register for a strict subreg no wider than a word if this is for
9956         a WORD_REGISTER_OPERATIONS target.
9958 2017-10-31  Jason Merrill  <jason@redhat.com>
9960         * gdbinit.in: Skip over inlines from timevar.h.
9962 2017-10-31  Martin Liska  <mliska@suse.cz>
9964         * doc/gcov.texi: Document new option.
9965         * gcov.c (print_usage): Likewise print it.
9966         (process_args): Support the argument.
9967         (format_count): New function.
9968         (format_gcov): Use the function.
9970 2017-10-31  Martin Liska  <mliska@suse.cz>
9972         * gcov.c (struct name_map): do not use typedef.
9973         Define operator== and operator<.
9974         (name_search): Remove.
9975         (name_sort): Remove.
9976         (main): Do not allocate names.
9977         (process_file): Add vertical space.
9978         (generate_results): Use std::find.
9979         (release_structures): Do not release memory.
9980         (find_source): Use std::find.
9982 2017-10-31  Martin Liska  <mliska@suse.cz>
9984         * gcov.c (struct line_info): Remove it's typedef.
9985         (line_info::line_info): Add proper ctor.
9986         (line_info::has_block): Do not use a typedef.
9987         (struct source_info): Do not use typedef.
9988         (circuit): Likewise.
9989         (get_cycles_count): Likewise.
9990         (output_intermediate_file): Iterate via vector iterator.
9991         (add_line_counts): Use std::vector methods.
9992         (accumulate_line_counts): Likewise.
9993         (output_lines): Likewise.
9995 2017-10-31  Martin Liska  <mliska@suse.cz>
9997         * gcov.c (struct source_info): Remove typedef.
9998         (source_info::source_info): Add proper ctor.
9999         (accumulate_line_counts): Use struct, not it's typedef.
10000         (output_gcov_file): Likewise.
10001         (output_lines): Likewise.
10002         (main): Do not allocate an array.
10003         (output_intermediate_file): Use size of vector container.
10004         (process_file): Resize the vector.
10005         (generate_results): Do not preallocate, use newly added vector
10006         lines.
10007         (release_structures): Do not release sources.
10008         (find_source): Use vector methods.
10009         (add_line_counts): Do not use typedef.
10011 2017-10-31  Martin Liska  <mliska@suse.cz>
10013         * doc/gcov.texi: Document that.
10014         * gcov.c (add_line_counts): Mark lines with a non-executed
10015         statement.
10016         (output_line_beginning): Handle such lines.
10017         (output_lines): Pass new argument.
10018         (output_intermediate_file): Print it in intermediate format.
10020 2017-10-31  Martin Liska  <mliska@suse.cz>
10022         * color-macros.h: New file.
10023         * diagnostic-color.c: Factor out color related to macros to
10024         color-macros.h.
10025         * doc/gcov.texi: Document -k option.
10026         * gcov.c (INCLUDE_STRING): Include string.h.
10027         (print_usage): Add -k option.
10028         (process_args): Parse it.
10029         (pad_count_string): New function.
10030         (output_line_beginning): Likewise.
10031         (DEFAULT_LINE_START): New macro.
10032         (output_lines): Support color output.
10034 2017-10-31  Martin Liska  <mliska@suse.cz>
10036         PR gcov-profile/82633
10037         * doc/gcov.texi: Document -fkeep-{static,inline}-functions and
10038         their interaction with GCOV infrastructure.
10039         * configure.ac: Add -fkeep-{inline,static}-functions to
10040         coverage_flags.
10041         * configure: Regenerate.
10043 2017-10-31  Uros Bizjak  <ubizjak@gmail.com>
10045         PR target/82772
10046         * config/alpha/sync.md (fetchop_constr) <and>: Change to "rINM".
10048 2017-10-31  Segher Boessenkool  <segher@kernel.crashing.org>
10050         PR target/82674
10051         * config/rs6000/rs6000.md (allocate_stack): Force update interval
10052         into a register if it does not fit into an immediate offset field.
10054 2017-10-31  Olivier Hainque  <hainque@adacore.com>
10056         * gcc/Makefile.in (FLAGS_TO_PASS): Pass libsubdir as well.
10058 2017-10-31  Julia Koval  <julia.koval@intel.com>
10060         * config.gcc: Add gfniintrin.h.
10061         * config/i386/gfniintrin.h: New.
10062         * config/i386/i386-builtin-types.def
10063         (__builtin_ia32_vgf2p8affineinvqb_v64qi,
10064         __builtin_ia32_vgf2p8affineinvqb_v64qi_mask,
10065         __builtin_ia32_vgf2p8affineinvqb_v32qi,
10066         __builtin_ia32_vgf2p8affineinvqb_v32qi_mask,
10067         __builtin_ia32_vgf2p8affineinvqb_v16qi,
10068         __builtin_ia32_vgf2p8affineinvqb_v16qi_mask): New builtins.
10069         * config/i386/i386-builtin.def (V64QI_FTYPE_V64QI_V64QI_INT_V64QI_UDI,
10070         V32QI_FTYPE_V32QI_V32QI_INT_V32QI_USI,
10071         V16QI_FTYPE_V16QI_V16QI_INT_V16QI_UHI,
10072         V64QI_FTYPE_V64QI_V64QI_INT): New types.
10073         * config/i386/i386.c (ix86_expand_args_builtin): Handle new types.
10074         * config/i386/immintrin.h: Include gfniintrin.h.
10075         * config/i386/sse.md (vgf2p8affineinvqb_*) New pattern.
10077 2017-10-30  Eric Botcazou  <ebotcazou@adacore.com>
10079         * gcc.c (HAVE_TARGET_EXECUTABLE_SUFFIX): Remove old kludge.
10081 2017-10-30  Wilco Dijkstra  <wdijkstr@arm.com>
10083         * config/arm/arm.md (ashldi3): Remove shift by 1 expansion.
10084         (arm_ashldi3_1bit): Remove pattern.
10085         (ashrdi3): Remove shift by 1 expansion.
10086         (arm_ashrdi3_1bit): Remove pattern.
10087         (lshrdi3): Remove shift by 1 expansion.
10088         (arm_lshrdi3_1bit): Remove pattern.
10089         * config/arm/arm.c (arm_rtx_costs_internal): Slightly increase
10090         cost of ashldi3 by 1.
10091         * config/arm/neon.md (ashldi3_neon): Remove shift by 1 expansion.
10092         (<shift>di3_neon): Likewise.
10094 2017-10-30  Dominik Infuehr  <dominik.infuehr@theobroma-systems.com>
10096         * config/aarch64/aarch64-simd.md (*aarch64_simd_mov): Rename
10097         both identically named patterns to (*aarch64_simd_mov<VD:mode>)
10098         and (*aarch64_simd_mov<VQ:mode>).
10099         (*aarch64_simd_mov<VD:mode>): Change type attribute to match
10100         pattern alternative.
10101         (*aarch64_simd_mov<VQ:mode>): Re-order and change type
10102         attributes to match pattern alternative.
10104 2017-10-30  Steven Munroe  <munroesj@gcc.gnu.org>
10106         * config.gcc (powerpc*-*-*): Add emmintrin.h.
10107         * config/rs6000/emmintrin.h: New file.
10108         * config/rs6000/x86intrin.h [__ALTIVEC__]: Include emmintrin.h.
10110 2017-10-30  Wilco Dijkstra  <wdijkstr@arm.com>
10112         * config/arm/vfp.md (movdi_vfp): Merge changes from movdi_vfp_cortexa8.
10113         * (movdi_vfp_cortexa8): Remove pattern.
10115 2017-10-30  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
10117         * doc/install.texi (Specific, alpha*-*-*): Remove DEC OSF/1
10118         etc. reference.
10119         (Specific, alpha*-dec-osf5.1): Remove.
10120         (Specific, mips-sgi-irix5): Remove.
10121         (Specific, mips-sgi-irix6): Remove.
10123 2017-10-30  Jakub Jelinek  <jakub@redhat.com>
10125         PR middle-end/22141
10126         * gimple-ssa-store-merging.c (merged_store_group::apply_stores): Fix
10127         arguments to clear_bit_region_be.
10129 2017-10-30  Jim Wilson  <wilson@tuliptree.org>
10131         * gimplify.c: Include memmodel.h.
10133 2017-10-30  Martin Jambor  <mjambor@suse.cz>
10135         * omp-grid.c (grid_attempt_target_gridification): Also insert a
10136         condition whether loop should be executed at all.
10138 2017-10-30  Will Schmidt  <will_schmidt@vnet.ibm.com>
10140         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add support for
10141         gimple folding of vec_madd() intrinsics.
10142         * config/rs6000/altivec.md (mulv8hi3): Rename altivec_vmladduhm to
10143         fmav8hi4.  (altivec_vmladduhm): Rename to fmav8hi4.
10144         * config/rs6000/rs6000-builtin.def: Rename vmladduhm to fmav8hi4
10146 2017-10-30  Richard Biener  <rguenther@suse.de>
10148         PR tree-optimization/82762
10149         Revert
10150         2017-10-23  Richard Biener  <rguenther@suse.de>
10152         PR tree-optimization/82129
10153         Revert
10154         2017-08-01  Richard Biener  <rguenther@suse.de>
10156         PR tree-optimization/81181
10157         * tree-ssa-pre.c (compute_antic_aux): Defer clean() to ...
10158         (compute_antic): ... end of iteration here.
10160 2017-10-30  Joseph Myers  <joseph@codesourcery.com>
10162         * doc/invoke.texi (C Dialect Options): Document -std=c17,
10163         -std=iso9899:2017 and -std=gnu17.
10164         * doc/standards.texi (C Language): Document C17 support.
10165         * doc/cpp.texi (Overview): Mention -std=c17.
10166         (Standard Predefined Macros): Document C11 and C17 values of
10167         __STDC_VERSION__.  Do not refer to C99 support as incomplete.
10168         * doc/extend.texi (Inline): Do not list individual options for
10169         standards newer than C99.
10170         * dwarf2out.c (highest_c_language, gen_compile_unit_die): Handle
10171         "GNU C17".
10172         * config/rl78/rl78.c (rl78_option_override): Handle "GNU C17"
10173         language name.
10175 2017-10-30  Maxim Ostapenko  <m.ostapenko@samsung.com>
10177         * asan.c (asan_finish_file): Align asan globals array by shadow
10178         granularity.
10180 2017-10-30  Jakub Jelinek  <jakub@redhat.com>
10182         PR middle-end/22141
10183         * gimple-ssa-store-merging.c: Include rtl.h and expr.h.
10184         (struct store_immediate_info): Add bitregion_start and bitregion_end
10185         fields.
10186         (store_immediate_info::store_immediate_info): Add brs and bre
10187         arguments and initialize bitregion_{start,end} from those.
10188         (struct merged_store_group): Add bitregion_start, bitregion_end,
10189         align_base and mask fields.  Drop unnecessary struct keyword from
10190         struct store_immediate_info.  Add do_merge method.
10191         (clear_bit_region_be): Use memset instead of loop storing zeros.
10192         (merged_store_group::do_merge): New method.
10193         (merged_store_group::merge_into): Use do_merge.  Allow gaps in between
10194         stores as long as the surrounding bitregions have no gaps.
10195         (merged_store_group::merge_overlapping): Use do_merge.
10196         (merged_store_group::apply_stores): Test that bitregion_{start,end}
10197         is byte aligned, rather than requiring that start and width are
10198         byte aligned.  Drop unnecessary struct keyword from
10199         struct store_immediate_info.  Allocate and populate also mask array.
10200         Make start of the arrays relative to bitregion_start rather than
10201         start and size them according to bitregion_{end,start} difference.
10202         (struct imm_store_chain_info): Drop unnecessary struct keyword from
10203         struct store_immediate_info.
10204         (pass_store_merging::gate): Punt if BITS_PER_UNIT or CHAR_BIT is not 8.
10205         (pass_store_merging::terminate_all_aliasing_chains): Drop unnecessary
10206         struct keyword from struct store_immediate_info.
10207         (imm_store_chain_info::coalesce_immediate_stores): Allow gaps in
10208         between stores as long as the surrounding bitregions have no gaps.
10209         Formatting fixes.
10210         (struct split_store): Add orig non-static data member.
10211         (split_store::split_store): Initialize orig to false.
10212         (find_constituent_stmts): Return store_immediate_info *, non-NULL
10213         if there is exactly a single original stmt.  Change stmts argument
10214         to pointer from reference, if NULL, don't push anything to it.  Add
10215         first argument, use it to optimize skipping over orig stmts that
10216         are known to be before bitpos already.  Simplify.
10217         (split_group): Return unsigned int count how many stores are or
10218         would be needed rather than a bool.  Add allow_unaligned argument.
10219         Change split_stores argument from reference to pointer, if NULL,
10220         only do a dry run computing how many stores would be produced.
10221         Rewritten algorithm to use both alignment and misalign if
10222         !allow_unaligned and handle bitfield stores with gaps.
10223         (imm_store_chain_info::output_merged_store): Set start_byte_pos
10224         from bitregion_start instead of start.  Compute allow_unaligned
10225         here, if true, do 2 split_group dry runs to compute which one
10226         produces fewer stores and prefer aligned if equal.  Punt if
10227         new count is bigger or equal than original before emitting any
10228         statements, rather than during that.  Remove no longer needed
10229         new_ssa_names tracking.  Replace num_stmts with
10230         split_stores.length ().  Use 32-bit stack allocated entries
10231         in split_stores auto_vec.  Try to reuse original store lhs/rhs1
10232         if possible.  Handle bitfields with gaps.
10233         (pass_store_merging::execute): Ignore bitsize == 0 stores.
10234         Compute bitregion_{start,end} for the stores and construct
10235         store_immediate_info with that.  Formatting fixes.
10237 2017-10-30  Uros Bizjak  <ubizjak@gmail.com>
10239         PR target/82725
10240         * config/i386/i386.c (legitimate_pic_address_disp_p): Allow
10241         UNSPEC_DTPOFF and UNSPEC_NTPOFF with SImode immediate offset.
10243 2017-10-29  Jim Wilson  <wilson@tuliptree.org>
10245         * gimplify.c: Include tm_p.h.
10247         * common.opt (gcoff): Re-add as ignored option.
10248         (gcoff1, gcoff2, gcoff3): Likewise.
10250         * Makefile.in (OBJS): Delete sdbout.o.
10251         (GTFILES): Delete $(srcdir)/sdbout.c.
10252         * debug.h: Delete sdb_debug_hooks.
10253         * final.c: Delete sdbout.h include.
10254         (final_scan_insn): Delete SDB_DEBUG check.
10255         (rest_of_clean_state): Likewise.
10256         * output.h: Delete sdb_begin_function_line.
10257         * sdbout.c: Delete.
10258         * sdbout.h: Delete.
10259         * toplev.c: Delete sdbout.h include.
10260         (process_options): Delete SDB_DEBUG check.
10261         * tree-core.h (tree_type_common): Delete pointer field of
10262         tree_type_symtab.
10263         * tree.c (copy_node): Clear TYPE_SYMTAB_DIE instead of
10264         TYPE_SYMTAB_POINTER.
10265         * tree.h (TYPE_SYMTAB_POINTER): Delete.
10266         (TYPE_SYMTAB_IS_POINTER): Delete.
10267         (TYPE_SYMTAB_IS_DIE): Renumber.
10268         * xcoffout.c: Refer to former sdbout.c file.
10269         (xcoffout_begin_prologue): Use past tense for sdbout.c reference.
10271         * doc/install.texi (--with-stabs): Delete COFF and ECOFF info.
10272         * doc/invoke.texi (SEEALSO): Delete adb and sdb references.
10273         (Debugging Options): Delete -gcoff.
10274         (-gstabs): Delete SDB reference.
10275         (-gcoff): Delete.
10276         (-gcoff@var{level}): Delete.
10277         * doc/passes.texi (Debugging information output): Delete SDB and
10278         sdbout.c references.
10279         * doc/tm.texi: Regenerate.
10280         * doc/tm.texi.in (DWARF_CIE_DATA_ALIGNMENT): Delete SDB from xref.
10281         (SDB and DWARF): Change node name to DWARF and delete SDB and COFF
10282         references.
10283         (DEBUGGER_AUTO_OFFSET): Delete COFF and SDB references.
10284         (PREFERRED_DEBUGGING_TYPE): Delete SDB_DEBUG and -gcoff references.
10285         (SDB_DEBUGGING_INFO): Delete.
10286         (PUT_SDB_@dots{}, SDB_DELIM, SDB_ALLOW_UNKNOWN_REFERENCES)
10287         SDB_ALLOW_FORWARD_REFERENCES, SDB_OUTPUT_SOURCE_LINE): Delete.
10288         * target.def (output_source_filename): Delete COFF reference.
10290         * common.opt (gcoff): Delete.
10291         (gxcoff+): Update Negative chain.
10292         * defaults.h: Delete all references to SDB_DEBUGGING_INFO and
10293         SDB_DEBUG.
10294         * dwarf2out.c (gen_array_type_die): Change SDB to debuggers.
10295         * flag-types.h (enum debug_info_type): Delete SDB_DEBUG.
10296         * function.c (number_blocks): Delete SDB_DEBUGGING_INFO, SDB_DEBUG,
10297         and SDB references.
10298         (expand_function_start): Change sdb reference to past tense.
10299         (expand_function_end): Change sdb reference to past tense.
10300         * gcc.c (cpp_unique_options): Delete gcoff3 reference.
10301         * opts.c (debug_type_names): Delete coff entry.
10302         (common_handle_option): Delete OPT_gcoff case.
10303         * system.h (SDB_DEBUG, SDB_DEBUGGING_INFO): Poison.
10305         * config/dbxcoff.h (PREFERRED_DEBUGGING_TYPE): Set to DBX_DEBUG.
10306         * config/cris/cris.h: Delete SDB reference in comment.
10307         * config/i386/cygming.h: Don't define SDB_DEBUGGING_INFO.
10308         (ASM_DECLARE_FUNCTION_NAME): Delete SDB reference from comment.
10309         * config/i386/gas.h: Don't define SDB_DEBUGGING_INFO.
10310         * config/i386/i386.c (svr4_dbx_register_map): Change SDB references
10311         to past tense.
10312         (ix86_expand_prologue): Likewise.
10313         * config/i386/winnt.c (i386_pe_start_function): Don't check SDB_DEBUG.
10314         * config/ia64/ia64.h: Likewise.
10315         * config/m68k/m68kelf.h (DBX_REGISTER_NUMBER): Delete SDB reference.
10316         * config/mips/mips.h (SUBTARGET_ASM_DEBUGGING_SPEC): Delete gcoff*
10317         support.
10318         * config/mmix/mmix.h: Likewise.
10319         * config/nds32/nds32.c: Likewise.
10320         * config/stormy/storym16.h: Likewise.
10321         * config/visium/visium.h: Likewise.
10322         * config/vx-common.h (SDB_DEBUGGING_INFO): Delete undef.
10324 2017-10-28  Sandra Loosemore  <sandra@codesourcery.com>
10326         * config/nios2/nios2.h (FRAME_GROWS_DOWNWARD): Define to 1.
10327         * config/nios2/nios2.c (nios2_initial_elimination_offset):  Make
10328         FRAME_POINTER_REGNUM point at high end of local var area.
10330 2017-10-27  Eric Botcazou  <ebotcazou@adacore.com>
10332         * bb-reorder.c (find_traces_1_round): Fix off-by-one index.
10333         Move comment around.  Do not reset best_edge for a copiable
10334         destination if the copy would cause a partition change.
10335         (better_edge_p): Remove redundant check.
10337 2017-10-27  Uros Bizjak  <ubizjak@gmail.com>
10339         * config/i386/i386-protos.h (ix86_fp_compare_mode): Remove prototype.
10341 2017-10-27  Michael Meissner  <meissner@linux.vnet.ibm.com>
10343         * builtins.c (CASE_MATHFN_FLOATN): New helper macro to add cases
10344         for math functions that have _Float<N> and _Float<N>X variants.
10345         (mathfn_built_in_2): Add support for math functions that have
10346         _Float<N> and _Float<N>X variants.
10347         (DEF_INTERNAL_FLT_FLOATN_FN): New helper macro.
10348         (expand_builtin_mathfn_ternary): Add support for fma with
10349         _Float<N> and _Float<N>X variants.
10350         (expand_builtin): Likewise.
10351         (fold_builtin_3): Likewise.
10352         * builtins.def (DEF_EXT_LIB_FLOATN_NX_BUILTINS): New macro to
10353         create math function _Float<N> and _Float<N>X variants as external
10354         library builtins.
10355         (BUILT_IN_COPYSIGN _Float<N> and _Float<N>X variants) Use
10356         DEF_EXT_LIB_FLOATN_NX_BUILTINS to make built-in functions using
10357         the __builtin_ prefix and if not strict ansi, without the prefix.
10358         (BUILT_IN_FABS _Float<N> and _Float<N>X variants): Likewise.
10359         (BUILT_IN_FMA _Float<N> and _Float<N>X variants): Likewise.
10360         (BUILT_IN_FMAX _Float<N> and _Float<N>X variants): Likewise.
10361         (BUILT_IN_FMIN _Float<N> and _Float<N>X variants): Likewise.
10362         (BUILT_IN_NAN _Float<N> and _Float<N>X variants): Likewise.
10363         (BUILT_IN_SQRT _Float<N> and _Float<N>X variants): Likewise.
10364         * builtin-types.def (BT_FN_FLOAT16_FLOAT16_FLOAT16_FLOAT16): New
10365         function signatures for fma _Float<N> and _Float<N>X variants.
10366         (BT_FN_FLOAT32_FLOAT32_FLOAT32_FLOAT32): Likewise.
10367         (BT_FN_FLOAT64_FLOAT64_FLOAT64_FLOAT64): Likewise.
10368         (BT_FN_FLOAT128_FLOAT128_FLOAT128_FLOAT128): Likewise.
10369         (BT_FN_FLOAT32X_FLOAT32X_FLOAT32X_FLOAT32X): Likewise.
10370         (BT_FN_FLOAT64X_FLOAT64X_FLOAT64X_FLOAT64X): Likewise.
10371         (BT_FN_FLOAT128X_FLOAT128X_FLOAT128X_FLOAT128X): Likewise.
10372         * gencfn-macros.c (print_case_cfn): Add support for math functions
10373         that have _Float<N> and _Float<N>X variants.
10374         (print_define_operator_list): Likewise.
10375         (fltfn_suffixes): Likewise.
10376         (main): Likewise.
10377         * internal-fn.def (DEF_INTERNAL_FLT_FLOATN_FN): New helper macro
10378         for math functions that have _Float<N> and _Float<N>X variants.
10379         (SQRT): Add support for sqrt, copysign, fmin and fmax _Float<N>
10380         and _Float<N>X variants.
10381         (COPYSIGN): Likewise.
10382         (FMIN): Likewise.
10383         (FMAX): Likewise.
10384         * fold-const.c (tree_call_nonnegative_warnv_p): Add support for
10385         copysign, fma, fmax, fmin, and sqrt _Float<N> and _Float<N>X
10386         variants.
10387         (integer_valued_read_call_p): Likewise.
10388         * fold-const-call.c (fold_const_call_ss): Likewise.
10389         (fold_const_call_sss): Add support for copysign, fmin, and fmax
10390         _Float<N> and _Float<N>X variants.
10391         (fold_const_call_ssss): Add support for fma _Float<N> and
10392         _Float<N>X variants.
10393         * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Add
10394         support for copysign and fma _Float<N> and _Float<N>X variants.
10395         (backprop::process_builtin_call_use): Likewise.
10396         * tree-call-cdce.c (can_test_argument_range); Add support for
10397         sqrt _Float<N> and _Float<N>X variants.
10398         (edom_only_function): Likewise.
10399         (get_no_error_domain): Likewise.
10400         * tree-ssa-math-opts.c (internal_fn_reciprocal): Likewise.
10401         * tree-ssa-reassoc.c (attempt_builtin_copysign): Add support for
10402         copysign _Float<N> and _Float<N>X variants.
10403         * config/rs6000/rs6000-builtin.def (SQRTF128): Delete, this is now
10404         handled by machine independent code.
10405         (FMAF128): Likewise.
10406         * doc/cpp.texi (Common Predefined Macros): Document defining
10407         __FP_FAST_FMAF<N> and __FP_FAST_FMAF<N>X if the backend supports
10408         fma _Float<N> and _Float<N>X variants.
10410 2017-10-27  Uros Bizjak  <ubizjak@gmail.com>
10412         PR target/82692
10413         * config/i386/i386-modes.def (CCFPU): Remove definition.
10414         * config/i386/i386.c (put_condition_mode): Remove CCFPU mode handling.
10415         (ix86_cc_modes_compatible): Ditto.
10416         (ix86_expand_carry_flag_compare): Ditto.
10417         (ix86_expand_int_movcc): Ditto.
10418         (ix86_expand_int_addcc): Ditto.
10419         (ix86_reverse_condition): Ditto.
10420         (ix86_unordered_fp_compare): Rename from ix86_fp_compare_mode.
10421         Return true/false for unordered/ordered fp comparisons.
10422         (ix86_cc_mode): Always return CCFPmode for float mode comparisons.
10423         (ix86_prepare_fp_compare_args): Update for rename.
10424         (ix86_expand_fp_compare): Update for rename.  Generate unordered
10425         compare RTXes wrapped with UNSPEC_NOTRAP unspec.
10426         (ix86_expand_sse_compare_and_jump): Ditto.
10427         * config/i386/predicates.md (fcmov_comparison_operator):
10428         Remove CCFPU mode handling.
10429         (ix86_comparison_operator): Ditto.
10430         (ix86_carry_flag_operator): Ditto.
10431         * config/i386/i386.md (UNSPEC_NOTRAP): New unspec.
10432         (*cmpu<mode>_i387): Wrap compare RTX with UNSPEC_NOTRAP unspec.
10433         (*cmpu<mode>_cc_i387): Ditto.
10434         (FPCMP): Remove mode iterator.
10435         (unord): Remove mode attribute.
10436         (unord_subst): New define_subst transformation
10437         (unord): New define_subst attribute.
10438         (unordered): Ditto.
10439         (*cmpi<unord><MODEF:mode>): Rewrite using unord_subst transformation.
10440         (*cmpi<unord>xf_i387): Ditto.
10441         * config/i386/sse.md (<sse>_<unord>comi<round_saeonly_name>): Merge
10442         from <sse>_comi<round_saeonly_name> and <sse>_ucomi<round_saeonly_name>
10443         using unord_subst transformation.
10444         * config/i386/subst.md (SUBST_A): Remove CCFP and CCFPU modes.
10445         (round_saeonly): Also handle CCFP mode.
10446         * reg-stack.c (subst_stack_regs_pat): Handle UNSPEC_NOTRAP unspec.
10447         Remove UNSPEC_SAHF unspec handling.
10449 2017-10-27  Jan Hubicka  <hubicka@ucw.cz>
10451         * x86-tune.def (X86_TUNE_INTER_UNIT_MOVES_TO_VEC): Disable for Zen.
10453 2017-10-27  Jeff Law  <law@redhat.com>
10455         * gimple-ssa-sprintf.c: Include domwalk.h.
10456         (class sprintf_dom_walker): New class, derived from dom_walker.
10457         (sprintf_dom_walker::before_dom_children): New function.
10458         (struct call_info): Moved into sprintf_dom_walker class
10459         (compute_formath_length, handle_gimple_call): Likewise.
10460         (sprintf_length::execute): Call the dominator walker rather
10461         than walking the statements.
10463         * tree-vrp.c (check_all_array_refs): Do not use wi->info to smuggle
10464         gimple statement locations.
10465         (check_array_bounds): Corresponding changes.  Get the statement's
10466         location directly from wi->stmt.
10468 2017-10-27  Palmer Dabbelt  <palmer@dabbelt.com>
10470         PR target/82717
10471         * doc/invoke.texi (RISC-V) <-mabi>: Correct and improve.
10473 2017-10-27  Jan Hubicka  <hubicka@ucw.cz>
10475         * config/i386/x86-tune.def (X86_TUNE_PARTIAL_REG_DEPENDENCY,
10476         X86_TUNE_MOVX): Disable for Haswell and newer CPUs.
10478 2017-10-27  Jakub Jelinek  <jakub@redhat.com>
10480         PR target/82703
10481         * config/i386/i386-protos.h (maybe_get_pool_constant): Removed.
10482         * config/i386/i386.c (maybe_get_pool_constant): Removed.
10483         (ix86_split_to_parts): Use avoid_constant_pool_reference instead of
10484         maybe_get_pool_constant.
10485         * config/i386/predicates.md (zero_extended_scalar_load_operand):
10486         Likewise.
10488 2017-10-27  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
10490         * doc/install.texi (Specific, i?86-*-solaris2.10): Simplify gas
10491         2.26 caveat.  Update gas and gld versions.
10492         (Specific, *-*-solaris2*): Update binutils version.  Remove caveat
10493         reference.
10495 2017-10-27  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
10497         * cgraph.h (set_malloc_flag): Declare.
10498         * cgraph.c (set_malloc_flag_1): New function.
10499         (set_malloc_flag): Likewise.
10500         * ipa-fnsummary.h (ipa_call_summary): Add new field is_return_callee.
10501         * ipa-fnsummary.c (ipa_call_summary::reset): Set is_return_callee to
10502         false.
10503         (read_ipa_call_summary): Add support for reading is_return_callee.
10504         (write_ipa_call_summary): Stream is_return_callee.
10505         * ipa-inline.c (ipa_inline): Remove call to ipa_free_fn_summary.
10506         * ipa-pure-const.c: Add headers ssa.h, alloc-pool.h, symbol-summary.h,
10507         ipa-prop.h, ipa-fnsummary.h.
10508         (pure_const_names): Change to static.
10509         (malloc_state_e): Define.
10510         (malloc_state_names): Define.
10511         (funct_state_d): Add field malloc_state.
10512         (varying_state): Set malloc_state to STATE_MALLOC_BOTTOM.
10513         (check_retval_uses): New function.
10514         (malloc_candidate_p): Likewise.
10515         (analyze_function): Add support for malloc attribute.
10516         (pure_const_write_summary): Stream malloc_state.
10517         (pure_const_read_summary): Add support for reading malloc_state.
10518         (dump_malloc_lattice): New function.
10519         (propagate_malloc): New function.
10520         (warn_function_malloc): New function.
10521         (ipa_pure_const::execute): Call propagate_malloc and
10522         ipa_free_fn_summary.
10523         (pass_local_pure_const::execute): Add support for malloc attribute.
10524         * ssa-iterators.h (RETURN_FROM_IMM_USE_STMT): New macro.
10525         * doc/invoke.texi: Document Wsuggest-attribute=malloc.
10527 2017-10-27  Martin Liska  <mliska@suse.cz>
10529         PR gcov-profile/82457
10530         * doc/invoke.texi: Document that one needs a non-strict ISO mode
10531         for fork-like functions to be properly instrumented.
10533 2017-10-27  Richard Biener  <rguenther@suse.de>
10535         PR middle-end/81659
10536         * tree-eh.c (pass_lower_eh_dispatch::execute): Free dominator
10537         info when we redirected EH.
10539 2017-10-26  Michael Collison  <michael.collison@arm.com>
10541         * config/aarch64/aarch64.md(<optab>_trunc><vf><GPI:mode>2):
10542         New pattern.
10543         (<optab>_trunchf<GPI:mode>2: New pattern.
10544         (<optab>_trunc<vgp><GPI:mode>2: New pattern.
10545         * config/aarch64/iterators.md (wv): New mode attribute.
10546         (vf, VF): New mode attributes.
10547         (vgp, VGP): New mode attributes.
10548         (s): Update attribute with SImode and DImode prefixes.
10550 2017-10-26  Sandra Loosemore  <sandra@codesourcery.com>
10552         * config/nios2/constraints.md ("S"): Match r0rel_constant_p too.
10553         * config/nios2/nios2-protos.h (r0rel_constant_p): Declare.
10554         * config/nios2/nios2.c: (nios2_r0rel_sec_regex): New.
10555         (nios2_option_overide): Initialize it.  Don't allow R0-relative
10556         addressing with PIC.
10557         (nios2_rtx_costs): Handle r0rel_constant_p like gprel_constant_p.
10558         (nios2_symbolic_constant_p): Likewise.
10559         (nios2_legitimate_address_p): Likewise.
10560         (nios2_r0rel_section_name_p): New.
10561         (nios2_symbol_ref_in_r0rel_data_p): New.
10562         (nios2_emit_move_sequence): Handle r0rel_constant_p.
10563         (r0rel_constant_p): New.
10564         (nios2_print_operand_address): Handle r0rel_constant_p.
10565         (nios2_cdx_narrow_form_p): Likewise.
10566         * config/nios2/nios2.opt (mr0rel-sec=): New option.
10567         * doc/invoke.texi (Option Summary): Add -mr0rel-sec.
10568         (Nios II Options): Document -mr0rel-sec.
10570 2017-10-26  Sandra Loosemore  <sandra@codesourcery.com>
10572         * config/nios2/nios2.c: Include xregex.h.
10573         (nios2_gprel_sec_regex): New.
10574         (nios2_option_overide): Initialize it.  Don't allow GP-relative
10575         addressing with PIC.
10576         (nios2_small_section_name_p): Check for regex match.
10577         * config/nios2/nios2.opt (mgprel-sec=): New option.
10578         * doc/invoke.texi (Option Summary): Add -mgprel-sec.
10579         (Nios II Options): Document -mgprel-sec.
10581 2017-10-26  Jim Wilson  <wilson@tuliptree.org>
10583         * doc/invoke.texi (-fdebug-prefix-map): Expand documentation.
10585 2017-10-26  Tom de Vries  <tom@codesourcery.com>
10587         PR tree-optimization/82707
10588         * gimple.c (gimple_copy): Fix unsharing of
10589         GIMPLE_OMP_{SINGLE,TARGET,TEAMS}.
10591 2017-10-26  Olga Makhotina  <olga.makhotina@intel.com>
10593         * config/i386/avx512fintrin.h (_mm512_cmpeq_pd_mask,
10594         _mm512_cmple_pd_mask, _mm512_cmplt_pd_mask,
10595         _mm512_cmpneq_pd_mask, _mm512_cmpnle_pd_mask,
10596         _mm512_cmpnlt_pd_mask, _mm512_cmpord_pd_mask,
10597         _mm512_cmpunord_pd_mask, _mm512_mask_cmpeq_pd_mask,
10598         _mm512_mask_cmple_pd_mask, _mm512_mask_cmplt_pd_mask,
10599         _mm512_mask_cmpneq_pd_mask, _mm512_mask_cmpnle_pd_mask,
10600         _mm512_mask_cmpnlt_pd_mask, _mm512_mask_cmpord_pd_mask,
10601         _mm512_mask_cmpunord_pd_mask, _mm512_cmpeq_ps_mask,
10602         _mm512_cmple_ps_mask, _mm512_cmplt_ps_mask,
10603         _mm512_cmpneq_ps_mask, _mm512_cmpnle_ps_mask,
10604         _mm512_cmpnlt_ps_mask, _mm512_cmpord_ps_mask,
10605         _mm512_cmpunord_ps_mask, _mm512_mask_cmpeq_ps_mask,
10606         _mm512_mask_cmple_ps_mask, _mm512_mask_cmplt_ps_mask,
10607         _mm512_mask_cmpneq_ps_mask, _mm512_mask_cmpnle_ps_mask,
10608         _mm512_mask_cmpnlt_ps_mask, _mm512_mask_cmpord_ps_mask,
10609         _mm512_mask_cmpunord_ps_mask): New intrinsics.
10611 2017-10-26  Michael Meissner  <meissner@linux.vnet.ibm.com>
10613         * config/rs6000/aix.h (TARGET_IEEEQUAD_DEFAULT): Set long double
10614         default to IBM.
10615         * config/rs6000/darwin.h (TARGET_IEEEQUAD_DEFAULT): Likewise.
10616         * config/rs6000/rs6000.opt (-mabi=ieeelongdouble): Move the
10617         warning to rs6000.c.  Remove the Undocumented flag, since it has
10618         been documented.
10619         (-mabi=ibmlongdouble): Likewise.
10620         * config/rs6000/rs6000.c (TARGET_IEEEQUAD_DEFAULT): If it is not
10621         already set, set the default format for long double.
10622         (rs6000_debug_reg_global): Print whether long double is IBM or
10623         IEEE.
10624         (rs6000_option_override_internal): Rework setting long double
10625         format.  Only warn if the user is changing the long double default
10626         and they did not use -Wno-psabi.
10627         * doc/invoke.texi (PowerPC options): Update the documentation for
10628         -mabi=ieeelongdouble and -mabi=ibmlongdouble.
10630 2017-10-26  Richard Sandiford  <richard.sandiford@linaro.org>
10631             Alan Hayward  <alan.hayward@arm.com>
10632             David Sherwood  <david.sherwood@arm.com>
10634         * rtl.h (wider_subreg_mode): New function.
10635         * ira.h (ira_sort_regnos_for_alter_reg): Take a machine_mode *
10636         rather than an unsigned int *.
10637         * ira-color.c (regno_max_ref_width): Replace with...
10638         (regno_max_ref_mode): ...this new variable.
10639         (coalesced_pseudo_reg_slot_compare): Update accordingly.
10640         Use wider_subreg_mode.
10641         (ira_sort_regnos_for_alter_reg): Likewise.  Take a machine_mode *
10642         rather than an unsigned int *.
10643         * lra-constraints.c (uses_hard_regs_p): Use wider_subreg_mode.
10644         (process_alt_operands): Likewise.
10645         (invariant_p): Likewise.
10646         * lra-spills.c (assign_mem_slot): Likewise.
10647         (add_pseudo_to_slot): Likewise.
10648         * lra.c (collect_non_operand_hard_regs): Likewise.
10649         (add_regs_to_insn_regno_info): Likewise.
10650         * reload1.c (regno_max_ref_width): Replace with...
10651         (regno_max_ref_mode): ...this new variable.
10652         (reload): Update accordingly.  Update call to
10653         ira_sort_regnos_for_alter_reg.
10654         (alter_reg): Update to use regno_max_ref_mode.  Call wider_subreg_mode.
10655         (init_eliminable_invariants): Update to use regno_max_ref_mode.
10656         (scan_paradoxical_subregs): Likewise.
10658 2017-10-26  Wilco Dijkstra  <wdijkstr@arm.com>
10660         * config/aarch64/aarch64.h (EXIT_IGNORE_STACK): Set if alloca is used.
10661         (aarch64_frame): Add emit_frame_chain boolean.
10662         * config/aarch64/aarch64.c (aarch64_frame_pointer_required)
10663         Move eh_return case to aarch64_layout_frame.
10664         (aarch64_layout_frame): Initialize emit_frame_chain.
10665         (aarch64_expand_prologue): Use emit_frame_chain.
10667 2017-10-26  Wilco Dijkstra  <wdijkstr@arm.com>
10669         * config/aarch64/aarch64.c (aarch64_layout_frame):
10670         Ensure LR is always stored at the bottom of the callee-saves.
10671         Remove rarely used frame layout which saves callee-saves at top of
10672         frame, so the store of LR can be used as a valid probe in all cases.
10674 2017-10-26  Wilco Dijkstra  <wdijkstr@arm.com>
10676         * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
10677         Improve unaligned TImode/TFmode base/offset split.
10679 2017-10-26  Richard Sandiford  <richard.sandiford@linaro.org>
10680             Alan Hayward  <alan.hayward@arm.com>
10681             David Sherwood  <david.sherwood@arm.com>
10683         * caller-save.c (mark_referenced_regs):  Use read_modify_subreg_p.
10684         * combine.c (find_single_use_1): Likewise.
10685         (expand_field_assignment): Likewise.
10686         (move_deaths): Likewise.
10687         * lra-constraints.c (simplify_operand_subreg): Likewise.
10688         (curr_insn_transform): Likewise.
10689         * lra.c (collect_non_operand_hard_regs): Likewise.
10690         (add_regs_to_insn_regno_info): Likewise.
10691         * rtlanal.c (reg_referenced_p): Likewise.
10692         (covers_regno_no_parallel_p): Likewise.
10694 2017-10-26  Richard Sandiford  <richard.sandiford@linaro.org>
10696         * wide-int-print.cc (print_hex): Loop based on extract_uhwi.
10697         Don't print any bits outside the precision of the value.
10698         * wide-int.cc (test_printing): Add some new tests.
10700 2017-10-26  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
10702         * configure.ac (gcc_cv_as_ix86_xbrace_comment): Check if assembler
10703         supports -xbrace_comment option.
10704         * configure: Regenerate.
10705         * config.in: Regenerate.
10706         * config/i386/sol2.h (ASM_XBRACE_COMMENT_SPEC): Define.
10707         (ASM_CPU_SPEC): Use it.
10709 2017-10-26  Richard Sandiford  <richard.sandiford@linaro.org>
10711         * target.def (static_rtx_alignment): New hook.
10712         * targhooks.h (default_static_rtx_alignment): Declare.
10713         * targhooks.c (default_static_rtx_alignment): New function.
10714         * doc/tm.texi.in (TARGET_STATIC_RTX_ALIGNMENT): New hook.
10715         * doc/tm.texi: Regenerate.
10716         * varasm.c (force_const_mem): Use targetm.static_rtx_alignment
10717         instead of targetm.constant_alignment.  Remove call to
10718         set_mem_attributes.
10719         * config/cris/cris.c (TARGET_STATIC_RTX_ALIGNMENT): Redefine.
10720         (cris_preferred_mininum_alignment): New function, split out from...
10721         (cris_constant_alignment): ...here.
10722         (cris_static_rtx_alignment): New function.
10723         * config/i386/i386.c (ix86_static_rtx_alignment): New function,
10724         split out from...
10725         (ix86_constant_alignment): ...here.
10726         (TARGET_STATIC_RTX_ALIGNMENT): Redefine.
10727         * config/mmix/mmix.c (TARGET_STATIC_RTX_ALIGNMENT): Redefine.
10728         (mmix_static_rtx_alignment): New function.
10729         * config/spu/spu.c (spu_static_rtx_alignment): New function.
10730         (TARGET_STATIC_RTX_ALIGNMENT): Redefine.
10732 2017-10-26  Tamar Christina  <tamar.christina@arm.com>
10734         PR target/81800
10735         * config/aarch64/aarch64.md (lrint<GPF:mode><GPI:mode>2):
10736         Add flag_trapping_math and flag_fp_int_builtin_inexact.
10738 2017-10-25  Palmer Dabbelt  <palmer@dabbelt.com>
10740         * config/riscv/riscv.md (ZERO_EXTEND_LOAD): Define.
10741         * config/riscv/pic.md (local_pic_load): Rename to local_pic_load_s,
10742         mark as a sign-extending load.
10743         (local_pic_load_u): Define.
10745 2017-10-25  Eric Botcazou  <ebotcazou@adacore.com>
10747         PR middle-end/82062
10748         * fold-const.c (operand_equal_for_comparison_p): Also return true
10749         if ARG0 is a simple variant of ARG1 with narrower precision.
10750         (fold_ternary_loc): Always pass unstripped operands to the predicate.
10752 2017-10-25  Jan Hubicka  <hubicka@ucw.cz>
10754         * i386.c (ix86_builtin_vectorization_cost): Compute scatter/gather
10755         cost correctly.
10756         * i386.h (processor_costs): Add gather_static, gather_per_elt,
10757         scatter_static, scatter_per_elt.
10758         * x86-tune-costs.h: Add new cost entries.
10760 2017-10-25  Richard Biener  <rguenther@suse.de>
10762         * tree-ssa-sccvn.h (vn_eliminate): Declare.
10763         * tree-ssa-pre.c (class eliminate_dom_walker, eliminate,
10764         class pass_fre): Move to ...
10765         * tree-ssa-sccvn.c (class eliminate_dom_walker, vn_eliminate,
10766         class pass_fre): ... here and adjust for statistics.
10768 2017-10-25  Jakub Jelinek  <jakub@redhat.com>
10770         PR libstdc++/81706
10771         * attribs.c (attribute_value_equal): Use omp_declare_simd_clauses_equal
10772         for comparison of OMP_CLAUSEs regardless of flag_openmp{,_simd}.
10773         (duplicate_one_attribute, copy_attributes_to_builtin): New functions.
10774         * attribs.h (duplicate_one_attribute, copy_attributes_to_builtin): New
10775         declarations.
10777 2017-10-25  Richard Biener  <rguenther@suse.de>
10779         * tree-ssa-pre.c (need_eh_cleanup, need_ab_cleanup, el_to_remove,
10780         el_to_fixup, el_todo, el_avail, el_avail_stack, eliminate_avail,
10781         eliminate_push_avail, eliminate_insert): Move inside...
10782         (class eliminate_dom_walker): ... this class in preparation
10783         of move.
10784         (fini_eliminate): Remove by merging with ...
10785         (eliminate): ... this function.  Adjust for class changes.
10786         (pass_pre::execute): Remove fini_eliminate call.
10787         (pass_fre::execute): Likewise.
10789 2017-10-24  Jakub Jelinek  <jakub@redhat.com>
10791         PR target/82460
10792         * config/i386/sse.md (UNSPEC_VPERMI2, UNSPEC_VPERMI2_MASK): Remove.
10793         (VPERMI2, VPERMI2I): New mode iterators.
10794         (<avx512>_vpermi2var<mode>3_maskz): Remove 3 define_expand patterns.
10795         (<avx512>_vpermi2var<mode>3<sd_maskz_name>): Remove 3 define_insn
10796         patterns.
10797         (<avx512>_vpermi2var<mode>3_mask): New define_expand using VPERMI2
10798         mode iterator.  Remove 3 old define_insn patterns.
10799         (*<avx512>_vpermi2var<mode>3_mask): 2 new define_insn patterns.
10800         (<avx512>_vpermt2var<mode>3_maskz): Adjust 1 define_expand to use
10801         VPERMI2 mode iterator, remove the other two expanders.
10802         (<avx512>_vpermt2var<mode>3<sd_maskz_name>): Adjust 1 define_insn
10803         to use VPERMI2 mode iterator, add another alternative for vpermi2*
10804         instructions, remove the other two patterns.
10805         (<avx512>_vpermt2var<mode>3_mask): Adjust 1 define_insn to use VPERMI2
10806         mode iterator, remove the other two patterns.
10807         * config/i386/i386.c (ix86_expand_vec_perm_vpermi2): Renamed to ...
10808         (ix86_expand_vec_perm_vpermt2): ... this.  Swap mask and op0
10809         arguments, use gen_*vpermt2* expanders instead of gen_*vpermi2*
10810         and adjust argument order accordingly.
10811         (ix86_expand_vec_perm): Adjust caller.
10812         (expand_vec_perm_1): Likewise.
10813         (expand_vec_perm_vpermi2_vpshub2): Rename to ...
10814         (expand_vec_perm_vpermt2_vpshub2): ... this.
10815         (ix86_expand_vec_perm_const_1): Adjust caller.
10816         (ix86_vectorize_vec_perm_const_ok): Adjust comments.
10818         PR target/82370
10819         * config/i386/sse.md (VIMAX_AVX2): Remove V4TImode.
10820         (VIMAX_AVX2_AVX512BW, VIMAX_AVX512VL): New mode iterators.
10821         (vec_shl_<mode>): Remove unused expander.
10822         (avx512bw_<shift_insn><mode>3): New define_insn.
10823         (<sse2_avx2>_ashl<mode>3, <sse2_avx2>_lshr<mode>3): Replaced by ...
10824         (<sse2_avx2>_<shift_insn><mode>3): ... this.  New define_insn.
10826 2017-10-24  Paolo Carlini  <paolo.carlini@oracle.com>
10828         PR c++/82466
10829         * doc/invoke.texi ([Wbuiltin-declaration-mismatch]): Extend
10830         description.
10832 2017-10-24  Wilco Dijkstra  <wdijkstr@arm.com>
10834         PR rtl-optimization/82396
10835         * gcc/haifa-sched.c (ready_sort_real): Remove qsort workaround.
10836         (autopref_multipass_init): Simplify initialization.
10837         (autopref_rank_data): Simplify sort order.
10838         * gcc/sched-int.h (autopref_multipass_data_): Remove
10839         multi_mem_insn_p, min_offset and max_offset.
10841 2017-10-24  Wilco Dijkstra  <wdijkstr@arm.com>
10843         PR middle-end/60580
10844         * config/aarch64/aarch64.c (aarch64_frame_pointer_required)
10845         Check special value of flag_omit_frame_pointer.
10846         (aarch64_can_eliminate): Likewise.
10847         (aarch64_override_options_after_change_1): Simplify handling of
10848         -fomit-frame-pointer and -fomit-leaf-frame-pointer.
10850 2017-10-24  Richard Biener  <rguenther@suse.de>
10852         PR tree-optimization/82697
10853         * tree-ssa-phiopt.c (cond_store_replacement): Use alias-set
10854         zero for conditional load and unconditional store.
10856 2017-10-24  H.J. Lu  <hongjiu.lu@intel.com>
10858         * doc/install.texi: Document bootstrap-cet.
10860 2017-10-24  H.J. Lu  <hongjiu.lu@intel.com>
10862         PR target/82659
10863         * config/i386/i386.c (rest_of_insert_endbranch): Don't insert
10864         ENDBR instruction at function entrance if function is only
10865         called directly.
10867 2017-10-24  Jakub Jelinek  <jakub@redhat.com>
10869         PR target/82628
10870         * config/i386/i386.md (addcarry<mode>, subborrow<mode>): Change
10871         patterns to better describe from which operation the CF is computed.
10872         (addcarry<mode>_0, subborrow<mode>_0): New patterns.
10873         * config/i386/i386.c (ix86_expand_builtin) <case handlecarry>: Pass
10874         one LTU with [DT]Imode and another one with [SD]Imode.  If arg0
10875         is 0, use _0 suffixed expanders instead of emitting a comparison
10876         before it.
10878 2017-10-06  Sergey Shalnov  <Sergey.Shalnov@intel.com>
10880         * config/i386/i386.md(*movsf_internal, *movdf_internal):
10881         Avoid 512-bit AVX modes for TARGET_PREFER_AVX256.
10883 2017-10-24  Eric Botcazou  <ebotcazou@adacore.com>
10885         PR middle-end/82569
10886         * tree-outof-ssa.h (always_initialized_rtx_for_ssa_name_p): Delete.
10887         * expr.c (expand_expr_real_1) <expand_decl_rtl>: Revert latest change.
10888         * loop-iv.c (iv_get_reaching_def): Likewise.
10889         * cfgexpand.c (expand_one_ssa_partition): Initialize the RTX if the
10890         variable is promoted and the partition contains undefined values.
10892 2017-10-23  Sandra Loosemore  <sandra@codesourcery.com>
10894         * config/nios2/nios2.c (nios2_rtx_costs): Make costs better
10895         reflect reality.
10896         (nios2_address_cost): Define.
10897         (nios2_legitimize_address): Recognize (exp + constant) directly.
10898         (TARGET_ADDRESS_COST): Define.
10900 2017-10-23  Sandra Loosemore  <sandra@codesourcery.com>
10902         * config/nios2/nios2-protos.h (nios2_large_constant_p): Declare.
10903         (nios2_symbolic_memory_operand_p): Declare.
10904         (nios2_split_large_constant): Declare.
10905         (nios2_split_symbolic_memory_operand): Declare.
10906         * config/nios2/nios2.c: Adjust includes.
10907         (nios2_symbolic_constant_allowed): New.
10908         (nios2_symbolic_constant_p): New.
10909         (nios2_plus_symbolic_constant_p): New.
10910         (nios2_valid_addr_expr_p): Recognize addresses involving
10911         symbolic constants.
10912         (nios2_legitimate_address_p): Likewise, also LO_SUM.
10913         (nios2_symbolic_memory_operand_p): New.
10914         (nios2_large_constant_p): New.
10915         (nios2_split_large_constant): New.
10916         (nios2_split_plus_large_constant): New.
10917         (nios2_split_symbolic_memory_operand): New.
10918         (nios2_legitimize_address): Code refactoring.  Handle addresses
10919         involving symbolic constants.
10920         (nios2_emit_move_sequence): Likewise.
10921         (nios2_print_operand): Improve error output.
10922         (nios2_print_operand_address): Handle LO_SUM.
10923         (nios2_cdx_narrow_form_p): Likewise.
10924         * config/nios2/nios2.md (movqi_internal): Add splitter for memory
10925         operands involving symbolic constants.
10926         (movhi_internal, movsi_internal): Likewise.
10927         (zero_extendhisi2, zero_extendqi<mode>2): Likewise.
10928         (extendhisi2, extendqi<mode>2): Likewise.
10930 2017-10-23  Sandra Loosemore  <sandra@codesourcery.com>
10932         * tree-pass.h (PROP_rtl_split_insns): Define.
10933         * recog.c (pass_data_split_all_insns): Provide PROP_rtl_split_insns.
10935 2017-10-23  Sandra Loosemore  <sandra@codesourcery.com>
10937         * config/nios2/nios2.c (TARGET_LRA_P): Don't override.
10939 2017-10-23  Jakub Jelinek  <jakub@redhat.com>
10941         PR debug/82630
10942         * target.def (const_not_ok_for_debug_p): Default to
10943         default_const_not_ok_for_debug_p instead of hook_bool_rtx_false.
10944         * targhooks.h (default_const_not_ok_for_debug_p): New declaration.
10945         * targhooks.c (default_const_not_ok_for_debug_p): New function.
10946         * dwarf2out.c (const_ok_for_output_1): Only reject UNSPECs for
10947         which targetm.const_not_ok_for_debug_p returned true.
10948         * config/arm/arm.c (arm_const_not_ok_for_debug_p): Return true
10949         for UNSPECs.
10950         * config/powerpcspe/powerpcspe.c (rs6000_const_not_ok_for_debug_p):
10951         Likewise.
10952         * config/rs6000/rs6000.c (rs6000_const_not_ok_for_debug_p): Likewise.
10953         * config/i386/i386.c (ix86_delegitimize_address_1): Don't delegitimize
10954         UNSPEC_GOTOFF with addend into addend - _GLOBAL_OFFSET_TABLE_ + symbol
10955         if !base_term_p.
10956         (ix86_const_not_ok_for_debug_p): New function.
10957         (i386_asm_output_addr_const_extra): Handle UNSPEC_GOTOFF.
10958         (TARGET_CONST_NOT_OK_FOR_DEBUG_P): Redefine.
10960 2017-10-23  David Malcolm  <dmalcolm@redhat.com>
10962         PR bootstrap/82610
10963         * system.h: Conditionally include "unique-ptr.h" if
10964         INCLUDE_UNIQUE_PTR is defined.
10965         * unique-ptr-tests.cc: Remove include of "unique-ptr.h" in favor
10966         of defining INCLUDE_UNIQUE_PTR before including "system.h".
10968 2017-10-23  Sebastian Perta  <sebastian.perta@renesas.com>
10970         * config/rl78/rl78.md: New define_expand "subdi3".
10972 2017-10-23  H.J. Lu  <hongjiu.lu@intel.com>
10974         PR target/82673
10975         * config/i386/i386.c (ix86_finalize_stack_frame_flags): Skip
10976         DF_REF_INSN if DF_REF_INSN_INFO is false.
10978 2017-10-23  Jan Hubicka  <hubicka@ucw.cz>
10980         * i386.c (dimode_scalar_chain::compute_convert_gain): Use
10981         xmm_move instead of sse_move.
10982         (sse_store_index): New function.
10983         (ix86_register_move_cost): Be more sensible about mismatch stall;
10984         model AVX moves correctly; make difference between sse->integer and
10985         integer->sse.
10986         (ix86_builtin_vectorization_cost): Model correctly aligned and unaligned
10987         moves; make difference between SSE and AVX.
10988         * i386.h (processor_costs): Remove sse_move; add xmm_move, ymm_move
10989         and zmm_move. Increase size of sse load and store tables;
10990         add unaligned load and store tables; add ssemmx_to_integer.
10991         * x86-tune-costs.h: Update all entries according to real
10992         move latencies from Agner Fog's manual and chip documentation.
10994 2017-10-23  Jakub Jelinek  <jakub@redhat.com>
10996         PR target/82628
10997         * config/i386/predicates.md (x86_64_dwzext_immediate_operand): New.
10998         * config/i386/constraints.md (Wf): New constraint.
10999         * config/i386/i386.md (UNSPEC_SBB): New unspec.
11000         (cmp<dwi>_doubleword): Removed.
11001         (sub<mode>3_carry_ccc, *sub<mode>3_carry_ccc_1): New patterns.
11002         (sub<mode>3_carry_ccgz): Use unspec instead of compare.
11003         * config/i386/i386.c (ix86_expand_branch) <case E_TImode>: Don't
11004         expand with cmp<dwi>_doubleword.  For LTU and GEU use
11005         sub<mode>3_carry_ccc instead of sub<mode>3_carry_ccgz and use CCCmode.
11007         * common.opt (gcolumn-info): Enable by default.
11008         * doc/invoke.texi (gcolumn-info): Document new default.
11010 2017-10-23  Richard Biener  <rguenther@suse.de>
11012         PR tree-optimization/82672
11013         * graphite-isl-ast-to-gimple.c (graphite_copy_stmts_from_block):
11014         Fold the stmt if we propagated into it.
11016 2017-10-23  Richard Biener  <rguenther@suse.de>
11018         * tree-ssa-pre.c (bitmap_remove_from_set): Rename to...
11019         (bitmap_remove_expr_from_set): ... this.  All callers call this
11020         for non-constant values.
11021         (bitmap_set_subtract): Rename to...
11022         (bitmap_set_subtract_expressions): ... this.  Adjust and
11023         optimize.
11024         (bitmap_set_contains_value): Remove superfluous check.
11025         (bitmap_set_replace_value): Inline into single caller ...
11026         (bitmap_value_replace_in_set): ... here and simplify.
11027         (dependent_clean): Merge into ...
11028         (clean): ... this using an overload.  Adjust.
11029         (prune_clobbered_mems): Adjust.
11030         (compute_antic_aux): Likewise.
11031         (compute_partial_antic_aux): Likewise.
11033 2017-10-23  Richard Biener  <rguenther@suse.de>
11035         PR tree-optimization/82129
11036         Revert
11037         2017-08-01  Richard Biener  <rguenther@suse.de>
11039         PR tree-optimization/81181
11040         * tree-ssa-pre.c (compute_antic_aux): Defer clean() to ...
11041         (compute_antic): ... end of iteration here.
11043 2017-10-23  Richard Sandiford  <richard.sandiford@linaro.org>
11045         * target.def (starting_frame_offset): New hook.
11046         * doc/tm.texi (STARTING_FRAME_OFFSET): Remove in favor of...
11047         (TARGET_STARTING_FRAME_OFFSET): ...this new hook.
11048         * doc/tm.texi.in: Regenerate.
11049         * hooks.h (hook_hwi_void_0): Declare.
11050         * hooks.c (hook_hwi_void_0): New function.
11051         * doc/rtl.texi: Refer to TARGET_STARTING_FRAME_OFFSET instead of
11052         STARTING_FRAME_OFFSET.
11053         * builtins.c (expand_builtin_setjmp_receiver): Likewise.
11054         * reload1.c (reload): Likewise.
11055         * cfgexpand.c (expand_used_vars): Use targetm.starting_frame_offset
11056         instead of STARTING_FRAME_OFFSET.
11057         * function.c (try_fit_stack_local): Likewise.
11058         (assign_stack_local_1): Likewise
11059         (instantiate_virtual_regs): Likewise.
11060         * rtlanal.c (rtx_addr_can_trap_p_1): Likewise.
11061         * config/avr/avr.md (nonlocal_goto_receiver): Likewise.
11062         * config/aarch64/aarch64.h (STARTING_FRAME_OFFSET): Delete.
11063         * config/alpha/alpha.h (STARTING_FRAME_OFFSET): Likewise.
11064         * config/arc/arc.h (STARTING_FRAME_OFFSET): Likewise.
11065         * config/arm/arm.h (STARTING_FRAME_OFFSET): Likewise.
11066         * config/bfin/bfin.h (STARTING_FRAME_OFFSET): Likewise.
11067         * config/c6x/c6x.h (STARTING_FRAME_OFFSET): Likewise.
11068         * config/cr16/cr16.h (STARTING_FRAME_OFFSET): Likewise.
11069         * config/cris/cris.h (STARTING_FRAME_OFFSET): Likewise.
11070         * config/fr30/fr30.h (STARTING_FRAME_OFFSET): Likewise.
11071         * config/frv/frv.h (STARTING_FRAME_OFFSET): Likewise.
11072         * config/ft32/ft32.h (STARTING_FRAME_OFFSET): Likewise.
11073         * config/h8300/h8300.h (STARTING_FRAME_OFFSET): Likewise.
11074         * config/i386/i386.h (STARTING_FRAME_OFFSET): Likewise.
11075         * config/ia64/ia64.h (STARTING_FRAME_OFFSET): Likewise.
11076         * config/m32c/m32c.h (STARTING_FRAME_OFFSET): Likewise.
11077         * config/m68k/m68k.h (STARTING_FRAME_OFFSET): Likewise.
11078         * config/mcore/mcore.h (STARTING_FRAME_OFFSET): Likewise.
11079         * config/mn10300/mn10300.h (STARTING_FRAME_OFFSET): Likewise.
11080         * config/moxie/moxie.h (STARTING_FRAME_OFFSET): Likewise.
11081         * config/msp430/msp430.h (STARTING_FRAME_OFFSET): Likewise.
11082         * config/nds32/nds32.h (STARTING_FRAME_OFFSET): Likewise.
11083         * config/nios2/nios2.h (STARTING_FRAME_OFFSET): Likewise.
11084         * config/nvptx/nvptx.h (STARTING_FRAME_OFFSET): Likewise.
11085         * config/pdp11/pdp11.h (STARTING_FRAME_OFFSET): Likewise.
11086         * config/riscv/riscv.h (STARTING_FRAME_OFFSET): Likewise.
11087         * config/rl78/rl78.h (STARTING_FRAME_OFFSET): Likewise.
11088         * config/rx/rx.h (STARTING_FRAME_OFFSET): Likewise.
11089         * config/s390/s390.h (STARTING_FRAME_OFFSET): Likewise.
11090         * config/sh/sh.h (STARTING_FRAME_OFFSET): Likewise.
11091         * config/sparc/sparc.c (sparc_compute_frame_size): Likewise.
11092         * config/sparc/sparc.h (STARTING_FRAME_OFFSET): Likewise.
11093         * config/spu/spu.h (STARTING_FRAME_OFFSET): Likewise.
11094         * config/stormy16/stormy16.h (STARTING_FRAME_OFFSET): Likewise.
11095         * config/tilegx/tilegx.h (STARTING_FRAME_OFFSET): Likewise.
11096         * config/tilepro/tilepro.h (STARTING_FRAME_OFFSET): Likewise.
11097         * config/v850/v850.h (STARTING_FRAME_OFFSET): Likewise.
11098         * config/visium/visium.h (STARTING_FRAME_OFFSET): Likewise.
11099         * config/avr/avr.h (STARTING_FRAME_OFFSET): Likewise.
11100         * config/avr/avr-protos.h (avr_starting_frame_offset): Likewise.
11101         * config/avr/avr.c (avr_starting_frame_offset): Make static and
11102         return a HOST_WIDE_INT.
11103         (avr_builtin_setjmp_frame_value): Use it instead of
11104         STARTING_FRAME_OFFSET.
11105         (TARGET_STARTING_FRAME_OFFSET): Redefine.
11106         * config/epiphany/epiphany.h (STARTING_FRAME_OFFSET): Delete.
11107         * config/epiphany/epiphany.c (epiphany_starting_frame_offset):
11108         New function.
11109         (TARGET_STARTING_FRAME_OFFSET): Redefine.
11110         * config/iq2000/iq2000.h (STARTING_FRAME_OFFSET): Delete.
11111         * config/iq2000/iq2000.c (iq2000_starting_frame_offset): New function.
11112         (TARGET_CONSTANT_ALIGNMENT): Redefine.
11113         * config/lm32/lm32.h (STARTING_FRAME_OFFSET): Delete.
11114         * config/lm32/lm32.c (lm32_starting_frame_offset): New function.
11115         (TARGET_STARTING_FRAME_OFFSET): Redefine.
11116         * config/m32r/m32r.h (STARTING_FRAME_OFFSET): Delete.
11117         * config/m32r/m32r.c (m32r_starting_frame_offset): New function.
11118         (TARGET_STARTING_FRAME_OFFSET): Redefine.
11119         * config/microblaze/microblaze.h (STARTING_FRAME_OFFSET): Delete.
11120         * config/microblaze/microblaze.c (microblaze_starting_frame_offset):
11121         New function.
11122         (TARGET_STARTING_FRAME_OFFSET): Redefine.
11123         * config/mips/mips.h (STARTING_FRAME_OFFSET): Delete.
11124         * config/mips/mips.c (mips_compute_frame_info): Refer to
11125         TARGET_STARTING_FRAME_OFFSET instead of STARTING_FRAME_OFFSET.
11126         (mips_starting_frame_offset): New function.
11127         (TARGET_STARTING_FRAME_OFFSET): Redefine.
11128         * config/mmix/mmix.h (STARTING_FRAME_OFFSET): Delete.
11129         * config/mmix/mmix-protos.h (mmix_starting_frame_offset): Delete.
11130         * config/mmix/mmix.c (mmix_starting_frame_offset): Make static
11131         and return a HOST_WIDE_INT.
11132         (TARGET_STARTING_FRAME_OFFSET): Redefine.
11133         (mmix_initial_elimination_offset): Refer to
11134         TARGET_STARTING_FRAME_OFFSET instead of STARTING_FRAME_OFFSET.
11135         * config/pa/pa.h (STARTING_FRAME_OFFSET): Delete.
11136         * config/pa/pa.c (pa_starting_frame_offset): New function.
11137         (pa_compute_frame_size): Use it instead of STARTING_FRAME_OFFSET.
11138         (pa_expand_prologue): Likewise.
11139         (TARGET_STARTING_FRAME_OFFSET): Redefine.
11140         * config/powerpcspe/aix.h (STARTING_FRAME_OFFSET): Split out
11141         !FRAME_GROWS_DOWNWARD handling to...
11142         (RS6000_STARTING_FRAME_OFFSET): ...this new macro.
11143         * config/powerpcspe/darwin.h (STARTING_FRAME_OFFSET): Split out
11144         !FRAME_GROWS_DOWNWARD handling to...
11145         (RS6000_STARTING_FRAME_OFFSET): ...this new macro.
11146         * config/powerpcspe/powerpcspe.h (STARTING_FRAME_OFFSET): Split out
11147         !FRAME_GROWS_DOWNWARD handling to...
11148         (RS6000_STARTING_FRAME_OFFSET): ...this new macro.
11149         * config/powerpcspe/powerpcspe.c (TARGET_STARTING_FRAME_OFFSET):
11150         Redefine.
11151         (rs6000_starting_frame_offset): New function.
11152         * config/rs6000/aix.h (STARTING_FRAME_OFFSET): Split out
11153         !FRAME_GROWS_DOWNWARD handling to...
11154         (RS6000_STARTING_FRAME_OFFSET): ...this new macro.
11155         * config/rs6000/darwin.h (STARTING_FRAME_OFFSET): Split out
11156         !FRAME_GROWS_DOWNWARD handling to...
11157         (RS6000_STARTING_FRAME_OFFSET): ...this new macro.
11158         * config/rs6000/rs6000.h (STARTING_FRAME_OFFSET): Split out
11159         !FRAME_GROWS_DOWNWARD handling to...
11160         (RS6000_STARTING_FRAME_OFFSET): ...this new macro.
11161         * config/rs6000/rs6000.c (TARGET_STARTING_FRAME_OFFSET): Refine.
11162         (rs6000_starting_frame_offset): New function.
11163         * config/vax/elf.h (STARTING_FRAME_OFFSET): Delete.
11164         * config/vax/vax.h (STARTING_FRAME_OFFSET): Delete.
11165         * config/vax/vax.c (vax_starting_frame_offset): New function.
11166         (vax_expand_prologue): Use it instead of STARTING_FRAME_OFFSET.
11167         (TARGET_STARTING_FRAME_OFFSET): Redefine.
11168         * config/xtensa/xtensa.h (STARTING_FRAME_OFFSET): Delete.
11169         * config/xtensa/xtensa.c (xtensa_starting_frame_offset): New function.
11170         (TARGET_STARTING_FRAME_OFFSET): Redefine.
11171         * system.h (STARTING_FRAME_OFFSET): Poison.
11173 2017-10-23  Richard Sandiford  <richard.sandiford@linaro.org>
11175         * tree-vect-loop.c (vect_create_epilog_for_reduction): Use
11176         SCALAR_TYPE_MODE instead of TYPE_MODE.
11178 2017-10-23  Richard Sandiford  <richard.sandiford@linaro.org>
11179             Alan Hayward  <alan.hayward@arm.com>
11180             David Sherwood  <david.sherwood@arm.com>
11182         * dwarf2out.c (loc_list_from_tree_1): Use SCALAR_INT_TYPE_MODE
11184 2017-10-23  Richard Sandiford  <richard.sandiford@linaro.org>
11185             Alan Hayward  <alan.hayward@arm.com>
11186             David Sherwood  <david.sherwood@arm.com>
11188         * expmed.c (expand_shift_1): Use scalar_mode for scalar_mode.
11190 2017-10-23  Richard Biener  <rguenther@suse.de>
11192         PR tree-optimization/82129
11193         * tree-ssa-pre.c (bitmap_set_and): Remove.
11194         (compute_antic_aux): Compute ANTIC_OUT intersection in a way
11195         canonicalizing expressions in the set to those with lowest
11196         ID rather than taking that from the first edge.
11198 2017-10-23  Richard Sandiford  <richard.sandiford@linaro.org>
11200         * combine.c (rtx_equal_for_field_assignment_p): Use
11201         byte_lowpart_offset.
11203 2017-10-22  Richard Sandiford  <richard.sandiford@linaro.org>
11204             Alan Hayward  <alan.hayward@arm.com>
11205             David Sherwood  <david.sherwood@arm.com>
11207         * internal-fn.c (expand_direct_optab_fn): Don't assign directly
11208         to a SUBREG_PROMOTED_VAR.
11210 2017-10-22  Richard Sandiford  <richard.sandiford@linaro.org>
11211             Alan Hayward  <alan.hayward@arm.com>
11212             David Sherwood  <david.sherwood@arm.com>
11214         * cfgexpand.c (expand_debug_expr): Use GET_MODE_UNIT_PRECISION.
11215         (expand_debug_source_expr): Likewise.
11216         * combine.c (combine_simplify_rtx): Likewise.
11217         * cse.c (fold_rtx): Likewise.
11218         * optabs.c (expand_float): Likewise.
11219         * simplify-rtx.c (simplify_unary_operation_1): Likewise.
11220         (simplify_binary_operation_1): Likewise.
11222 2017-10-22  Richard Sandiford  <richard.sandiford@linaro.org>
11223             Alan Hayward  <alan.hayward@arm.com>
11224             David Sherwood  <david.sherwood@arm.com>
11226         * combine.c (simplify_comparison): Use HWI_COMPUTABLE_MODE_P.
11227         (record_promoted_value): Likewise.
11228         * expr.c (expand_expr_real_2): Likewise.
11229         * ree.c (update_reg_equal_equiv_notes): Likewise.
11230         (combine_set_extension): Likewise.
11231         * rtlanal.c (low_bitmask_len): Likewise.
11232         * simplify-rtx.c (neg_const_int): Likewise.
11233         (simplify_binary_operation_1): Likewise.
11235 2017-10-22  Richard Sandiford  <richard.sandiford@linaro.org>
11236             Alan Hayward  <alan.hayward@arm.com>
11237             David Sherwood  <david.sherwood@arm.com>
11239         * lra-spills.c (assign_mem_slot): Use subreg_size_lowpart_offset.
11240         * regcprop.c (maybe_mode_change): Likewise.
11241         * reload1.c (alter_reg): Likewise.
11243 2017-10-22  Richard Sandiford  <richard.sandiford@linaro.org>
11245         * inchash.h (inchash::hash::add_wide_int): New function.
11246         * lto-streamer-out.c (hash_tree): Use it.
11248 2017-10-22  Richard Sandiford  <richard.sandiford@linaro.org>
11250         * inchash.h (inchash::hash::add_wide_int): Rename to...
11251         (inchash::hash::add_hwi): ...this.
11252         * ipa-devirt.c (hash_odr_vtable): Update accordingly.
11253         (polymorphic_call_target_hasher::hash): Likewise.
11254         * ipa-icf.c (sem_function::get_hash, sem_function::init): Likewise.
11255         (sem_item::add_expr, sem_item::add_type, sem_variable::get_hash)
11256         (sem_item_optimizer::update_hash_by_addr_refs): Likewise.
11257         * lto-streamer-out.c (hash_tree): Likewise.
11258         * optc-save-gen.awk: Likewise.
11259         * tree.c (add_expr): Likewise.
11261 2017-10-22  Uros Bizjak  <ubizjak@gmail.com>
11263         PR target/52451
11264         * config/i386/i386.c (ix86_fp_compare_mode): Return CCFPmode
11265         for ordered inequality comparisons even with TARGET_IEEE_FP.
11267 2017-10-22  Uros Bizjak  <ubizjak@gmail.com>
11269         PR target/82628
11270         * config/i386/i386.md (cmp<dwi>_doubleword): New pattern.
11271         * config/i386/i386.c (ix86_expand_branch) <case E_TImode>:
11272         Expand with cmp<dwi>_doubleword.
11274 2017-10-21  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
11276         * extend.texi: Add x86 specific to 'nocf_check' attribute.
11277         List CET intrinsics.
11278         * invoke.texi: Add -mcet, -mibt, -mshstk options.  Add x86
11279         specific to -fcf-protection option.
11281 2017-10-21  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
11283         * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET): New.
11284         (OPTION_MASK_ISA_SHSTK_SET): Likewise.
11285         (OPTION_MASK_ISA_IBT_UNSET): Likewise.
11286         (OPTION_MASK_ISA_SHSTK_UNSET): Likewise.
11287         (ix86_handle_option): Add -mibt, -mshstk, -mcet handling.
11288         * config.gcc (extra_headers): Add cetintrin.h for x86 targets.
11289         (extra_objs): Add cet.o for Linux/x86 targets.
11290         (tmake_file): Add i386/t-cet for Linux/x86 targets.
11291         * config/i386/cet.c: New file.
11292         * config/i386/cetintrin.h: Likewise.
11293         * config/i386/t-cet: Likewise.
11294         * config/i386/cpuid.h (bit_SHSTK): New.
11295         (bit_IBT): Likewise.
11296         * config/i386/driver-i386.c (host_detect_local_cpu): Detect and
11297         pass IBT and SHSTK bits.
11298         * config/i386/i386-builtin-types.def
11299         (VOID_FTYPE_UNSIGNED_PVOID): New.
11300         (VOID_FTYPE_UINT64_PVOID): Likewise.
11301         * config/i386/i386-builtin.def: Add CET intrinsics.
11302         * config/i386/i386-c.c (ix86_target_macros_internal): Add
11303         OPTION_MASK_ISA_IBT, OPTION_MASK_ISA_SHSTK handling.
11304         * config/i386/i386-passes.def: Add pass_insert_endbranch pass.
11305         * config/i386/i386-protos.h (make_pass_insert_endbranch): New
11306         prototype.
11307         * config/i386/i386.c (rest_of_insert_endbranch): New.
11308         (pass_data_insert_endbranch): Likewise.
11309         (pass_insert_endbranch): Likewise.
11310         (make_pass_insert_endbranch): Likewise.
11311         (ix86_notrack_prefixed_insn_p): Likewise.
11312         (ix86_target_string): Add -mibt, -mshstk flags.
11313         (ix86_option_override_internal): Add flag_cf_protection
11314         processing.
11315         (ix86_valid_target_attribute_inner_p): Set OPT_mibt, OPT_mshstk.
11316         (ix86_print_operand): Add 'notrack' prefix output.
11317         (ix86_init_mmx_sse_builtins): Add CET intrinsics.
11318         (ix86_expand_builtin): Expand CET intrinsics.
11319         (x86_output_mi_thunk): Add 'endbranch' instruction.
11320         * config/i386/i386.h (TARGET_IBT): New.
11321         (TARGET_IBT_P): Likewise.
11322         (TARGET_SHSTK): Likewise.
11323         (TARGET_SHSTK_P): Likewise.
11324         * config/i386/i386.md (unspecv): Add UNSPECV_NOP_RDSSP,
11325         UNSPECV_INCSSP, UNSPECV_SAVEPREVSSP, UNSPECV_RSTORSSP,
11326         UNSPECV_WRSS, UNSPECV_WRUSS, UNSPECV_SETSSBSY, UNSPECV_CLRSSBSY.
11327         (builtin_setjmp_setup): New pattern.
11328         (builtin_longjmp): Likewise.
11329         (rdssp<mode>): Likewise.
11330         (incssp<mode>): Likewise.
11331         (saveprevssp): Likewise.
11332         (rstorssp): Likewise.
11333         (wrss<mode>): Likewise.
11334         (wruss<mode>): Likewise.
11335         (setssbsy): Likewise.
11336         (clrssbsy): Likewise.
11337         (nop_endbr): Likewise.
11338         * config/i386/i386.opt: Add -mcet, -mibt, -mshstk and -mcet-switch
11339         options.
11340         * config/i386/immintrin.h: Include <cetintrin.h>.
11341         * config/i386/linux-common.h
11342         (file_end_indicate_exec_stack_and_cet): New prototype.
11343         (TARGET_ASM_FILE_END): New.
11345 2017-10-20  Jan Hubicka  <hubicka@ucw.cz>
11347         * i386.c (ix86_builtin_vectorization_cost): Use existing rtx_cost
11348         latencies instead of having separate table; make difference between
11349         integer and float costs.
11350         * i386.h (processor_costs): Remove scalar_stmt_cost,
11351         scalar_load_cost, scalar_store_cost, vec_stmt_cost, vec_to_scalar_cost,
11352         scalar_to_vec_cost, vec_align_load_cost, vec_unalign_load_cost,
11353         vec_store_cost.
11354         * x86-tune-costs.h: Remove entries which has been removed in
11355         procesor_costs from all tables; make cond_taken_branch_cost
11356         and cond_not_taken_branch_cost COST_N_INSNS based.
11358 2017-10-20  Jan Hubicka  <hubicka@ucw.cz>
11360         * x86-tune-costs.h (intel_cost, generic_cost): Fix move costs.
11362 2017-10-20  Jakub Jelinek  <jakub@redhat.com>
11364         * config/i386/i386.md (isa): Remove fma_avx512f.
11365         * config/i386/sse.md (<avx512>_fmadd_<mode>_mask<round_name>,
11366         <avx512>_fmadd_<mode>_mask3<round_name>,
11367         <avx512>_fmsub_<mode>_mask<round_name>,
11368         <avx512>_fmsub_<mode>_mask3<round_name>,
11369         <avx512>_fnmadd_<mode>_mask<round_name>,
11370         <avx512>_fnmadd_<mode>_mask3<round_name>,
11371         <avx512>_fnmsub_<mode>_mask<round_name>,
11372         <avx512>_fnmsub_<mode>_mask3<round_name>,
11373         <avx512>_fmaddsub_<mode>_mask<round_name>,
11374         <avx512>_fmaddsub_<mode>_mask3<round_name>,
11375         <avx512>_fmsubadd_<mode>_mask<round_name>,
11376         <avx512>_fmsubadd_<mode>_mask3<round_name>): Remove isa attribute.
11377         (*vec_widen_umult_even_v16si<mask_name>,
11378         *vec_widen_smult_even_v16si<mask_name>): Likewise.
11379         (<mask_codefor>avx512bw_dbpsadbw<mode><mask_name>): Likewise.
11381 2017-10-20  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
11383         * extend.texi: Add 'nocf_check' documentation.
11384         * gimple.texi: Add second parameter to
11385         gimple_build_call_from_tree.
11386         * invoke.texi: Add -fcf-protection documentation.
11387         * rtl.texi: Add REG_CALL_NOTRACK documenation.
11389 2017-10-20  Richard Biener  <rguenther@suse.de>
11391         PR tree-optimization/82473
11392         * tree-vect-loop.c (vectorizable_reduction): Properly get at
11393         the largest input type.
11395 2017-10-20  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
11397         * c-attribs.c (handle_nocf_check_attribute): New function.
11398         (c_common_attribute_table): Add 'nocf_check' handling.
11399         * gimple-parser.c: Add second argument NULL to
11400         gimple_build_call_from_tree.
11401         * attrib.c (comp_type_attributes): Check nocf_check attribute.
11402         * cfgexpand.c (expand_call_stmt): Set REG_CALL_NOCF_CHECK for
11403         call insn.
11404         * combine.c (distribute_notes): Add REG_CALL_NOCF_CHECK handling.
11405         * common.opt: Add fcf-protection flag.
11406         * emit-rtl.c (try_split): Add REG_CALL_NOCF_CHECK handling.
11407         * flag-types.h: Add enum cf_protection_level.
11408         * gimple.c (gimple_build_call_from_tree): Add second parameter.
11409         Add 'nocf_check' attribute propagation to gimple call.
11410         * gimple.h (gf_mask): Add GF_CALL_NOCF_CHECK.
11411         (gimple_build_call_from_tree): Update prototype.
11412         (gimple_call_nocf_check_p): New function.
11413         (gimple_call_set_nocf_check): Likewise.
11414         * gimplify.c: Add second argument to gimple_build_call_from_tree.
11415         * ipa-icf.c: Add nocf_check attribute in statement hash.
11416         * recog.c (peep2_attempt): Add REG_CALL_NOCF_CHECK handling.
11417         * reg-notes.def: Add REG_NOTE (CALL_NOCF_CHECK).
11418         * toplev.c (process_options): Add flag_cf_protection handling.
11420 2017-10-19  Jan Hubicka  <hubicka@ucw.cz>
11422         * x86-tune-costs.h (core_cost): Fix div, move and sqrt latencies.
11424 2017-10-20  Richard Biener  <rguenther@suse.de>
11426         PR tree-optimization/82603
11427         * tree-if-conv.c (predicate_mem_writes): Make sure to only
11428         remove false predicated stores.
11430 2017-10-20  Richard Biener  <rguenther@suse.de>
11432         * graphite-isl-ast-to-gimple.c
11433         (translate_isl_ast_to_gimple::graphite_copy_stmts_from_block):
11434         Remove return value and simplify, dump copied stmt after lhs
11435         adjustment.
11436         (translate_isl_ast_to_gimple::translate_isl_ast_node_user):
11437         Reduce dump verbosity.
11438         (gsi_insert_earliest): Likewise.
11439         (translate_isl_ast_to_gimple::copy_bb_and_scalar_dependences): Adjust.
11440         * graphite.c (print_global_statistics): Adjust dumping.
11441         (print_graphite_scop_statistics): Likewise.
11442         (print_graphite_statistics): Do not dump loops here.
11443         (graphite_transform_loops): But here.
11445 2017-10-20  Nicolas Roche  <roche@adacore.com>
11447         * configure.ac (ACX_PROG_GNAT): Append "libgnat" to include search dir.
11448         * configure: Regenerate.
11450 2017-10-20  Jakub Jelinek  <jakub@redhat.com>
11452         PR target/82158
11453         * tree-cfg.c (pass_warn_function_return::execute): In noreturn
11454         functions when optimizing replace GIMPLE_RETURN stmts with
11455         calls to __builtin_unreachable ().
11457         PR sanitizer/82595
11458         * config/gnu-user.h (LIBTSAN_EARLY_SPEC): Add libtsan_preinit.o
11459         for -fsanitize=thread link of executables.
11460         (LIBLSAN_EARLY_SPEC): Add liblsan_preinit.o for -fsanitize=leak
11461         link of executables.
11463         PR target/82370
11464         * config/i386/sse.md (VI248_AVX2, VI248_AVX512BW, VI248_AVX512BW_2):
11465         New mode iterators.
11466         (<shift_insn><mode>3<mask_name>): Change the last of the 3
11467         define_insns for logical vector shifts to use VI248_AVX512BW
11468         iterator instead of VI48_AVX512, remove <mask_mode512bit_condition>
11469         condition, useless isa and prefix attributes.  Change the first
11470         2 of these define_insns to ...
11471         (<mask_codefor><shift_insn><mode>3<mask_name>): ... this, new
11472         define_insn for avx512vl.
11473         (<shift_insn><mode>3): ... and this, new define_insn without
11474         masking for non-avx512vl.
11476         PR target/82370
11477         * config/i386/sse.md (*andnot<mode>3,
11478         <mask_codefor><code><mode>3<mask_name>, *<code><mode>3): Split
11479         (=v,v,vm) alternative into (=x,x,xm) and (=v,v,vm), for 128-bit
11480         and 256-bit vectors, the (=x,x,xm) alternative and when mask is
11481         not applied use empty suffix even for TARGET_AVX512VL.
11482         * config/i386/subst.md (mask_prefix3, mask_prefix4): When mask
11483         is applied, supply evex,evex or evex,evex,evex instead of just
11484         evex.
11486 2017-10-20  Julia Koval  <julia.koval@intel.com>
11488         * common/config/i386/i386-common.c (OPTION_MASK_ISA_GFNI_SET,
11489         (OPTION_MASK_ISA_GFNI_UNSET): New.
11490         (ix86_handle_option): Handle OPT_mgfni.
11491         * config/i386/cpuid.h (bit_GFNI): New.
11492         * config/i386/driver-i386.c (host_detect_local_cpu): Detect gfni.
11493         * config/i386/i386-c.c (ix86_target_macros_internal): Define __GFNI__.
11494         * config/i386/i386.c (ix86_target_string): Add -mgfni.
11495         (ix86_valid_target_attribute_inner_p): Add OPT_mgfni.
11496         * config/i386/i386.h (TARGET_GFNI, TARGET_GFNI_P): New.
11497         * config/i386/i386.opt: Add mgfni.
11499 2017-10-20  Orlando Arias  <oarias@knights.ucf.edu>
11501         * config/msp430/msp430.c (msp430_option_override): Disable
11502         -fdelete-null-pointer-checks.
11503         * doc/invoke.text (-fdelete-null-pointer-checks): Document that.
11505 2017-10-19  Jan Hubicka  <hubicka@ucw.cz>
11507         * x86-tune-costs.h (generic_cost, core_cost): Correct costs
11508         of x87 and SSE instructions.
11510 2017-10-19  Jan Hubicka  <hubicka@ucw.cz>
11512         * asan.c (create_cond_insert_point): Do not update edge count.
11513         * auto-profile.c (afdo_propagate_edge): Update for edge count removal.
11514         (afdo_propagate_circuit): Likewise.
11515         (afdo_calculate_branch_prob): Likewise.
11516         (afdo_annotate_cfg): Likewise.
11517         * basic-block.h (struct edge_def): Remove count.
11518         (edge_def::count): New accessor.
11519         * bb-reorder.c (rotate_loop): Update.
11520         (find_traces_1_round): Update.
11521         (connect_traces): Update.
11522         (sanitize_hot_paths): Update.
11523         * cfg.c (unchecked_make_edge): Update.
11524         (make_single_succ_edge): Update.
11525         (check_bb_profile): Update.
11526         (dump_edge_info): Update.
11527         (update_bb_profile_for_threading): Update.
11528         (scale_bbs_frequencies_int): Update.
11529         (scale_bbs_frequencies_gcov_type): Update.
11530         (scale_bbs_frequencies_profile_count): Update.
11531         (scale_bbs_frequencies): Update.
11532         * cfganal.c (connect_infinite_loops_to_exit): Update.
11533         * cfgbuild.c (compute_outgoing_frequencies): Update.
11534         (find_many_sub_basic_blocks): Update.
11535         * cfgcleanup.c (try_forward_edges): Update.
11536         (try_crossjump_to_edge): Update
11537         * cfgexpand.c (expand_gimple_cond): Update
11538         (expand_gimple_tailcall): Update
11539         (construct_exit_block): Update
11540         * cfghooks.c (verify_flow_info): Update
11541         (redirect_edge_succ_nodup): Update
11542         (split_edge): Update
11543         (make_forwarder_block): Update
11544         (duplicate_block): Update
11545         (account_profile_record): Update
11546         * cfgloop.c (find_subloop_latch_edge_by_profile): Update.
11547         * cfgloopanal.c (expected_loop_iterations_unbounded): Update.
11548         * cfgloopmanip.c (scale_loop_profile): Update.
11549         (loopify): Update.
11550         (lv_adjust_loop_entry_edge): Update.
11551         * cfgrtl.c (try_redirect_by_replacing_jump): Update.
11552         (force_nonfallthru_and_redirect): Update.
11553         (purge_dead_edges): Update.
11554         (rtl_flow_call_edges_add): Update.
11555         * cgraphunit.c (init_lowered_empty_function): Update.
11556         (cgraph_node::expand_thunk): Update.
11557         * gimple-pretty-print.c (dump_probability): Update.
11558         (dump_edge_probability): Update.
11559         * gimple-ssa-isolate-paths.c (isolate_path): Update.
11560         * haifa-sched.c (sched_create_recovery_edges): Update.
11561         * hsa-gen.c (convert_switch_statements): Update.
11562         * ifcvt.c (dead_or_predicable): Update.
11563         * ipa-inline-transform.c (inline_transform): Update.
11564         * ipa-split.c (split_function): Update.
11565         * ipa-utils.c (ipa_merge_profiles): Update.
11566         * loop-doloop.c (add_test): Update.
11567         * loop-unroll.c (unroll_loop_runtime_iterations): Update.
11568         * lto-streamer-in.c (input_cfg): Update.
11569         (input_function): Update.
11570         * lto-streamer-out.c (output_cfg): Update.
11571         * modulo-sched.c (sms_schedule): Update.
11572         * postreload-gcse.c (eliminate_partially_redundant_load): Update.
11573         * predict.c (maybe_hot_edge_p): Update.
11574         (unlikely_executed_edge_p): Update.
11575         (probably_never_executed_edge_p): Update.
11576         (dump_prediction): Update.
11577         (drop_profile): Update.
11578         (propagate_unlikely_bbs_forward): Update.
11579         (determine_unlikely_bbs): Update.
11580         (force_edge_cold): Update.
11581         * profile.c (compute_branch_probabilities): Update.
11582         * reg-stack.c (better_edge): Update.
11583         * shrink-wrap.c (handle_simple_exit): Update.
11584         * tracer.c (better_p): Update.
11585         * trans-mem.c (expand_transaction): Update.
11586         (split_bb_make_tm_edge): Update.
11587         * tree-call-cdce.c: Update.
11588         * tree-cfg.c (gimple_find_sub_bbs): Update.
11589         (gimple_split_edge): Update.
11590         (gimple_duplicate_sese_region): Update.
11591         (gimple_duplicate_sese_tail): Update.
11592         (gimple_flow_call_edges_add): Update.
11593         (insert_cond_bb): Update.
11594         (execute_fixup_cfg): Update.
11595         * tree-cfgcleanup.c (cleanup_control_expr_graph): Update.
11596         * tree-complex.c (expand_complex_div_wide): Update.
11597         * tree-eh.c (lower_resx): Update.
11598         (unsplit_eh): Update.
11599         (cleanup_empty_eh_move_lp): Update.
11600         * tree-inline.c (copy_edges_for_bb): Update.
11601         (freqs_to_counts): Update.
11602         (copy_cfg_body): Update.
11603         * tree-ssa-dce.c (remove_dead_stmt): Update.
11604         * tree-ssa-ifcombine.c (update_profile_after_ifcombine): Update.
11605         * tree-ssa-loop-im.c (execute_sm_if_changed): Update.
11606         * tree-ssa-loop-ivcanon.c (remove_exits_and_undefined_stmts): Update.
11607         (unloop_loops): Update.
11608         * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Update.
11609         * tree-ssa-loop-split.c (connect_loops): Update.
11610         (split_loop): Update.
11611         * tree-ssa-loop-unswitch.c (hoist_guard): Update.
11612         * tree-ssa-phionlycprop.c (propagate_rhs_into_lhs): Update.
11613         * tree-ssa-phiopt.c (replace_phi_edge_with_variable): Update.
11614         * tree-ssa-reassoc.c (branch_fixup): Update.
11615         * tree-ssa-tail-merge.c (replace_block_by): Update.
11616         * tree-ssa-threadupdate.c (remove_ctrl_stmt_and_useless_edges): Update.
11617         (compute_path_counts): Update.
11618         (update_profile): Update.
11619         (recompute_probabilities): Update.
11620         (update_joiner_offpath_counts): Update.
11621         (estimated_freqs_path): Update.
11622         (freqs_to_counts_path): Update.
11623         (clear_counts_path): Update.
11624         (ssa_fix_duplicate_block_edges): Update.
11625         (duplicate_thread_path): Update.
11626         * tree-switch-conversion.c (hoist_edge_and_branch_if_true): Update.
11627         (case_bit_test_cmp): Update.
11628         (collect_switch_conv_info): Update.
11629         (gen_inbound_check): Update.
11630         (do_jump_if_equal): Update.
11631         (emit_cmp_and_jump_insns): Update.
11632         * tree-tailcall.c (decrease_profile): Update.
11633         (eliminate_tail_call): Update.
11634         * tree-vect-loop-manip.c (slpeel_add_loop_guard): Update.
11635         (vect_do_peeling): Update.
11636         * tree-vect-loop.c (scale_profile_for_vect_loop): Update.
11637         * ubsan.c (ubsan_expand_null_ifn): Update.
11638         (ubsan_expand_ptr_ifn): Update.
11639         * value-prof.c (gimple_divmod_fixed_value): Update.
11640         (gimple_mod_pow2): Update.
11641         (gimple_mod_subtract): Update.
11642         (gimple_ic): Update.
11643         (gimple_stringop_fixed_value): Update.
11645 2017-10-19  Uros Bizjak  <ubizjak@gmail.com>
11647         PR target/82618
11648         * config/i386/i386.md (sub to cmp): New peephole2 pattern.
11650 2017-10-19  Alexander Monakov  <amonakov@ispras.ru>
11652         PR rtl-optimization/82395
11653         * ira-color.c (allocno_priority_compare_func): Fix comparison step
11654         based on non_spilled_static_chain_regno_p.
11656 2017-10-19  Uros Bizjak  <ubizjak@gmail.com>
11658         * config/i386/i386.c (output_387_binary_op): Rewrite SSE part.
11659         (ix86_emit_mode_set): Rewrite insn mnemonic construction.
11660         (ix86_prepare_fp_compare_args): Redefine is_sse as bool.
11662 2017-10-19  Martin Sebor  <msebor@redhat.com>
11664         PR tree-optimization/82596
11665         * tree.c (array_at_struct_end_p): Handle STRING_CST.
11667 2017-10-19  Eric Botcazou  <ebotcazou@adacore.com>
11669         * asan.c (handle_builtin_alloca): Deal with all alloca variants.
11670         (get_mem_refs_of_builtin_call): Likewise.
11671         * builtins.c (expand_builtin_apply): Adjust call to
11672         allocate_dynamic_stack_space.
11673         (expand_builtin_alloca): For __builtin_alloca_with_align_and_max, pass
11674         the third argument to allocate_dynamic_stack_space, otherwise -1.
11675         (expand_builtin): Deal with all alloca variants.
11676         (is_inexpensive_builtin): Likewise.
11677         * builtins.def (BUILT_IN_ALLOCA_WITH_ALIGN_AND_MAX): New.
11678         * calls.c (special_function_p): Deal with all alloca variants.
11679         (initialize_argument_information): Adjust call to
11680         allocate_dynamic_stack_space.
11681         (expand_call): Likewise.
11682         * cfgexpand.c (expand_call_stmt): Deal with all alloca variants.
11683         * doc/extend.texi (Built-ins): Add __builtin_alloca_with_align_and_max
11684         * explow.c (allocate_dynamic_stack_space): Add MAX_SIZE parameter and
11685         use it for the stack usage computation.
11686         * explow.h (allocate_dynamic_stack_space): Adjust prototype.
11687         * function.c (gimplify_parameters): Call build_alloca_call_expr.
11688         * gimple-ssa-warn-alloca.c (alloca_call_type): Simplify control flow.
11689         Take into account 3rd argument of __builtin_alloca_with_align_and_max.
11690         (in_loop_p): Remove first argument and useless check.
11691         (pass_walloca::execute): Remove useless test and adjust call to above.
11692         * gimple.c (gimple_build_call_from_tree): Deal with all alloc variants
11693         * gimplify.c (gimplify_vla_decl): Call build_alloca_call_expr.
11694         (gimplify_call_expr): Deal with all alloca variants.
11695         * hsa-gen.c (gen_hsa_alloca): Likewise.
11696         (gen_hsa_insns_for_call): Likewise.
11697         * ipa-pure-const.c (special_builtin_state): Likewise.
11698         * tree-chkp.c (chkp_build_returned_bound): Likewise.
11699         * tree-object-size.c (alloc_object_size): Likewise.
11700         * tree-ssa-alias.c (ref_maybe_used_by_call_p_1): Likewise.
11701         (call_may_clobber_ref_p_1): Likewise.
11702         * tree-ssa-ccp.c (evaluate_stmt): Likewise.
11703         (ccp_fold_stmt): Likewise.
11704         (optimize_stack_restore): Likewise.
11705         * tree-ssa-dce.c (mark_stmt_if_obviously_necessary): Likewise.
11706         (mark_all_reaching_defs_necessary_1): Likewise.
11707         (propagate_necessity): Likewise.
11708         (eliminate_unnecessary_stmts): Likewise.
11709         * tree.c (build_common_builtin_nodes): Build
11710         BUILT_IN_ALLOCA_WITH_ALIGN_AND_MAX.
11711         (build_alloca_call_expr): New function.
11712         * tree.h (ALLOCA_FUNCTION_CODE_P): New macro.
11713         (CASE_BUILT_IN_ALLOCA): Likewise.
11714         (build_alloca_call_expr): Declare.
11715         * varasm.c (incorporeal_function_p): Deal with all alloca variants.
11717 2017-10-19  Eric Botcazou  <ebotcazou@adacore.com>
11719         PR debug/82509
11720         * dwarf2out.c (new_die_raw): New static inline function.
11721         (new_die): Use it to create the DIE.
11722         (add_AT_external_die_ref): Likewise.
11723         (clone_die): Likewise.
11724         (clone_as_declaration): Likewise.
11725         (dwarf2out_vms_debug_main_pointer): Likewise.
11726         (base_type_die): Likewise.  Remove early return for corner cases.
11727         Do not call add_pubtype on the DIE here.
11728         (is_base_type): Remove ERROR_MARK and return 0 for VOID_TYPE.
11729         (modified_type_die): Adjust the lookup for reverse order DIEs.  Skip
11730         typedefs for base types with DW_AT_endianity.  Make sure a DIE with
11731         native order exists for base types, attach the DIE manually and call
11732         add_pubtype on it.  Do not equate a reverse order DIE to the type.
11734 2017-10-19  Richard Earnshaw  <rearnsha@arm.com>
11736         * config/arm/arm.c (align_ok_ldrd_strd): New function.
11737         (mem_ok_for_ldrd_strd): New parameter align.  Extract the alignment of
11738         the mem into it.
11739         (gen_operands_ldrd_strd): Validate the alignment of the accesses.
11741 2017-10-19  Jakub Jelinek  <jakub@redhat.com>
11743         * flag-types.h (enum sanitize_code): Add SANITIZE_BUILTIN.  Or
11744         SANITIZE_BUILTIN into SANITIZE_UNDEFINED.
11745         * sanitizer.def (BUILT_IN_UBSAN_HANDLE_INVALID_BUILTIN,
11746         BUILT_IN_UBSAN_HANDLE_INVALID_BUILTIN_ABORT): New builtins.
11747         * opts.c (sanitizer_opts): Add builtin.
11748         * ubsan.c (instrument_builtin): New function.
11749         (pass_ubsan::execute): Call it.
11750         (pass_ubsan::gate): Enable even for SANITIZE_BUILTIN.
11751         * doc/invoke.texi: Document -fsanitize=builtin.
11753         * ubsan.c (ubsan_expand_null_ifn): Use _v1 suffixed type mismatch
11754         builtins, store max (log2 (align), 0) into uchar field instead of
11755         align into uptr field.
11756         (ubsan_expand_objsize_ifn): Use _v1 suffixed type mismatch builtins,
11757         store uchar 0 field instead of uptr 0 field.
11758         (instrument_nonnull_return): Use _v1 suffixed nonnull return builtin,
11759         instead of passing one address of struct with 2 locations pass
11760         two addresses of structs with 1 location each.
11761         * sanitizer.def (BUILT_IN_UBSAN_HANDLE_TYPE_MISMATCH,
11762         BUILT_IN_UBSAN_HANDLE_TYPE_MISMATCH_ABORT,
11763         BUILT_IN_UBSAN_HANDLE_NONNULL_RETURN,
11764         BUILT_IN_UBSAN_HANDLE_NONNULL_RETURN_ABORT): Removed.
11765         (BUILT_IN_UBSAN_HANDLE_TYPE_MISMATCH_V1,
11766         BUILT_IN_UBSAN_HANDLE_TYPE_MISMATCH_V1_ABORT,
11767         BUILT_IN_UBSAN_HANDLE_NONNULL_RETURN_V1,
11768         BUILT_IN_UBSAN_HANDLE_NONNULL_RETURN_V1_ABORT): New builtins.
11770 2017-10-19  Martin Liska  <mliska@suse.cz>
11772         PR driver/81829
11773         * file-find.c (remove_prefix): Remove.
11774         * file-find.h (remove_prefix): Likewise.
11775         * gcc-ar.c: Remove smartness of lookup.
11777 2017-10-19  Segher Boessenkool  <segher@kernel.crashing.org>
11779         * config/rs6000/rs6000.md (*call_indirect_aix<mode>,
11780         *call_value_indirect_aix<mode>, *call_indirect_elfv2<mode>,
11781         *call_value_indirect_elfv2<mode>): Add correct mode to the unspec.
11783 2017-10-19  Jakub Jelinek  <jakub@redhat.com>
11785         PR target/82580
11786         * config/i386/i386.md (setcc + movzbl to xor + setcc): New peephole2.
11787         (setcc + and to xor + setcc): New peephole2.
11789 2017-10-19  Tom de Vries  <tom@codesourcery.com>
11791         * doc/sourcebuild.texi (Test Directives, Variants of
11792         dg-require-support): Add dg-require-stack-size.
11794 2017-10-19  Martin Liska  <mliska@suse.cz>
11796         PR sanitizer/82517
11797         * gimplify.c (gimplify_decl_expr): Do not instrument variables
11798         that have a large alignment.
11799         (gimplify_target_expr): Likewise.
11801 2017-10-18  Segher Boessenkool  <segher@kernel.crashing.org>
11803         PR rtl-optimization/82602
11804         * ira.c (rtx_moveable_p): Return false for volatile asm.
11806 2017-10-18  Uros Bizjak  <ubizjak@gmail.com>
11808         PR target/82580
11809         * config/i386/i386-modes.def (CCGZ): New CC mode.
11810         * config/i386/i386.md (sub<mode>3_carry_ccgz): New insn pattern.
11811         * config/i386/predicates.md (ix86_comparison_operator):
11812         Handle CCGZmode.
11813         * config/i386/i386.c (ix86_expand_branch) <case E_TImode>:
11814         Emulate LE, LEU, GT, GTU, LT, LTU, GE and GEU double-word comparisons
11815         with double-word subtraction.
11816         (put_condition_code): Handle CCGZmode.
11818 2017-10-18  Aldy Hernandez  <aldyh@redhat.com>
11820         * wide-int.cc (debug (const wide_int &)): New.
11821         (debug (const wide_int *)): New.
11822         (debug (const widest_int &)): New.
11823         (debug (const widest_int *)): New.
11825 2017-10-18  Vladimir Makarov  <vmakarov@redhat.com>
11827         PR middle-end/82556
11828         * lra-constraints.c (curr_insn_transform): Use non-input operand
11829         instead of output one for matched reload.
11831 2017-10-18  Bin Cheng  <bin.cheng@arm.com>
11833         * tree-loop-distribution.c (INCLUDE_ALGORITHM): New header file.
11834         (tree-ssa-loop-ivopts.h): New header file.
11835         (struct builtin_info): New fields.
11836         (classify_builtin_1): Compute and record base and offset parts for
11837         memset builtin partition by calling strip_offset.
11838         (offset_cmp, fuse_memset_builtins): New functions.
11839         (finalize_partitions): Fuse adjacent memset partitions by calling
11840         above function.
11841         * tree-ssa-loop-ivopts.c (strip_offset): Delete static declaration.
11842         Expose the interface.
11843         * tree-ssa-loop-ivopts.h (strip_offset): New declaration.
11845 2017-10-18  Bin Cheng  <bin.cheng@arm.com>
11847         PR tree-optimization/82574
11848         * tree-loop-distribution.c (find_single_drs): New parameter.  Check
11849         that data reference must be executed exactly once per iteration
11850         against the outermost loop in nest.
11851         (classify_partition): Update call to above function.
11853 2017-10-18  Richard Biener  <rguenther@suse.de>
11855         PR tree-optimization/82591
11856         * graphite.c (graphite_transform_loops): Move code gen message
11857         printing ...
11858         * graphite-isl-ast-to-gimple.c (graphite_regenerate_ast_isl):
11859         Here.  Handle scop_to_isl_ast failing.
11860         (scop_to_isl_ast): Limit the number of ISL operations.
11862 2017-10-18  Richard Biener  <rguenther@suse.de>
11864         * graphite-isl-ast-to-gimple.c
11865         (translate_isl_ast_to_gimple::set_rename): Simplify.
11866         (translate_isl_ast_to_gimple::set_rename_for_each_def): Inline...
11867         (graphite_copy_stmts_from_block): ... here.
11868         (copy_bb_and_scalar_dependences): Simplify.
11869         (add_parameters_to_ivs_params): Canonicalize.
11870         (generate_entry_out_of_ssa_copies): Simplify.
11871         * graphite-sese-to-poly.c (extract_affine_name): Simplify
11872         by passing in ISL dimension.
11873         (parameter_index_in_region_1): Rename to ...
11874         (parameter_index_in_region): ... this.
11875         (extract_affine): Adjust assert, pass down parameter index.
11876         (add_param_constraints): Use range-info when available.
11877         (build_scop_context): Adjust.
11878         * sese.c (new_sese_info): Adjust.
11879         (free_sese_info): Likewise.
11880         * sese.h (bb_map_t, rename_map_t, phi_rename, init_back_edge_pair_t):
11881         Remove unused typedefs.
11882         (struct sese_info_t): Simplify rename_map, remove incomplete_phis.
11884 2017-10-18  Martin Liska  <mliska@suse.cz>
11886         * combine.c (simplify_compare_const): Add gcc_fallthrough.
11888 2017-10-18  Robin Dapp  <rdapp@linux.vnet.ibm.com>
11890         * config/s390/s390.c (s390_bb_fallthru_entry_likely): New function.
11891         (s390_sched_init): Do not reset s390_sched_state if we entered the
11892         current basic block via a fallthru edge and all others are unlikely.
11894 2017-10-18  Robin Dapp  <rdapp@linux.vnet.ibm.com>
11896         * config/s390/s390.c (NUM_SIDES): New variable.
11897         (LONGRUNNING_THRESHOLD): New variable.
11898         (LATENCY_FACTOR): New variable.
11899         (s390_sched_score): Decrease score for long-running instructions on
11900         wrong side.
11901         (s390_sched_variable_issue): Perform bookkeeping for long-running
11902         instructions.
11904 2017-10-18  Richard Biener  <rguenther@suse.de>
11906         * graphite-isl-ast-to-gimple.c (gcc_expression_from_isl_ast_expr_id):
11907         Simplify with removal of the parameter rename map.
11908         (set_rename): Likewise.
11909         (should_copy_to_new_region): Likewise.
11910         (graphite_copy_stmts_from_block): Likewise.
11911         (copy_bb_and_scalar_dependences): Remove initialization of
11912         unused copied_bb_map.
11913         (copy_def): Remove.
11914         (copy_internal_parameters): Likewise.
11915         (graphite_regenerate_ast_isl): Do not call copy_internal_parameters.
11916         * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
11917         Use INTEGRAL_TYPE_P.
11918         (parameter_index_in_region_1): Rename to ...
11919         (assign_parameter_index_in_region): ... this.  Assert we have
11920         a parameter we handle.
11921         (scan_tree_for_params): Adjust.
11922         * sese.h (parameter_rename_map_t): Remove.
11923         (struct sese_info_t): Remove unused parameter_rename_map and
11924         copied_bb_map members.
11925         * sese.c (new_sese_info): Adjust.
11926         (free_sese_info): Likewise.
11928 2017-10-18  Martin Liska  <mliska@suse.cz>
11930         PR sanitizer/82545
11931         * asan.c (asan_expand_poison_ifn): Do not put gimple stmt
11932         on an abnormal edge.
11934 2017-10-18  Sebastian Huber  <sebastian.huber@embedded-brains.de>
11936         * doc/invoke.texi (ffunction-sections and fdata-sections):
11937         Update.
11939 2017-10-17  Eric Botcazou  <ebotcazou@adacore.com>
11941         * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Bail out only if
11942         the use statement can throw internally.
11944 2017-10-17  Eric Botcazou  <ebotcazou@adacore.com>
11946         * config/visium/visium.c (visium_select_cc_mode): Return CCmode for
11947         any RTX present on the RHS of a SET.
11948         * compare-elim.c (try_eliminate_compare): Restore comment.
11950 2017-10-17  Jakub Jelinek  <jakub@redhat.com>
11952         * langhooks.h (struct lang_hooks): Document that tree_size langhook
11953         may be also called on tcc_type nodes.
11954         * langhooks.c (lhd_tree_size): Likewise.
11956 2017-10-17  David Malcolm  <dmalcolm@redhat.com>
11958         * gimple-ssa-sprintf.c (fmtwarn): Update for changed signature of
11959         format_warning_at_substring.
11960         (maybe_warn): Convert source_range * param to a location_t.  Pass
11961         UNKNOWN_LOCATION rather than NULL to fmtwarn.
11962         (format_directive): Remove code to extract source_ranges and
11963         source_range * in favor of just a location_t.
11964         (parse_directive): Pass UNKNOWN_LOCATION rather than NULL to
11965         fmtwarn.
11966         * substring-locations.c (format_warning_va): Convert
11967         source_range * param to a location_t.
11968         (format_warning_at_substring): Likewise.
11969         * substring-locations.h (format_warning_va): Likewise.
11970         (format_warning_at_substring): Likewise.
11972 2017-10-17  Jan Hubicka  <hubicka@ucw.cz>
11974         * target.h (enum vect_cost_for_stmt): Add vec_gather_load and
11975         vec_scatter_store
11976         * tree-vect-stmts.c (record_stmt_cost): Make difference between normal
11977         and scatter/gather ops.
11979         * aarch64/aarch64.c (aarch64_builtin_vectorization_cost): Add
11980         vec_gather_load and vec_scatter_store.
11981         * arm/arm.c (arm_builtin_vectorization_cost): Likewise.
11982         * powerpcspe/powerpcspe.c (rs6000_builtin_vectorization_cost): Likewise.
11983         * rs6000/rs6000.c (rs6000_builtin_vectorization_cost): Likewise.
11984         * s390/s390.c (s390_builtin_vectorization_cost): Likewise.
11985         * spu/spu.c (spu_builtin_vectorization_cost): Likewise.
11986         * i386/i386.c (x86_builtin_vectorization_cost): Likewise.
11988 2017-10-17  Uros Bizjak  <ubizjak@gmail.com>
11990         * reg-stack.c (compare_for_stack_reg): Add bool argument.
11991         Detect FTST instruction and handle its register pops.  Only pop
11992         second operand if can_pop_second_op is true.
11993         (subst_stack_regs_pat) <case COMPARE>: Detect FCOMI instruction to
11994         set can_pop_second_op to false in the compare_for_stack_reg call.
11996         * config/i386/i386.md (*cmpi<FPCMP:unord><MODEF:mode>): Only call
11997         output_fp_compare for stack register operands.
11998         * config/i386/i386.c (output_fp_compare): Do not output SSE compare
11999         instructions here.  Do not emit stack register pops here.  Assert
12000         that FCOMPP pops next to top stack register.  Rewrite function.
12002 2017-10-17  Nathan Sidwell  <nathan@acm.org>
12004         PR middle-end/82577
12005         * alias.c (compare_base_decls): Check HAS_DECL_ASSEMBLER_NAME_P,
12006         use DECL_ASSEMBLER_NAME_RAW.
12008         PR middle-end/82546
12009         * tree.c (tree_code_size): Reformat.  Punt to lang hook for unknown
12010         TYPE nodes.
12012 2017-10-17  Qing Zhao <qing.zhao@oracle.com>
12013             Wilco Dijkstra <wilco.dijkstra@arm.com>
12015         * builtins.c (expand_builtin_update_setjmp_buf): Add a
12016         converstion to Pmode from the buf_addr.
12018 2017-10-17  Richard Biener  <rguenther@suse.de>
12020         * graphite-dependences.c (scop_get_reads_and_writes): Change
12021         output parameters to references.
12023 2017-10-17  Jackson Woodruff  <jackson.woodruff@arm.com>
12025         PR 71026/tree-optimization
12026         * fold-const.c (distribute_real_division): Removed.
12027         (fold_binary_loc): Remove calls to distribute_real_divison.
12029 2017-10-17  Richard Biener  <rguenther@suse.de>
12031         * graphite-scop-detection.c
12032         (scop_detection::stmt_has_simple_data_refs_p): Always use
12033         the full nest as region.
12034         (try_generate_gimple_bb): Likewise.
12035         * sese.c (scalar_evolution_in_region): Simplify now that
12036         SCEV can handle instantiation in regions.
12037         * tree-scalar-evolution.c (instantiate_scev_name): Also instantiate
12038         in the non-loop part of a function if requested.
12040 2017-10-17  Richard Biener  <rguenther@suse.de>
12042         PR tree-optimization/82563
12043         * graphite-isl-ast-to-gimple.c (generate_entry_out_of_ssa_copies):
12044         New function.
12045         (graphite_regenerate_ast_isl): Call it.
12046         * graphite-scop-detection.c (build_scops): Remove entry edge split.
12048 2017-10-17  Jakub Jelinek  <jakub@redhat.com>
12050         PR tree-optimization/82549
12051         * fold-const.c (optimize_bit_field_compare, fold_truth_andor_1):
12052         Formatting fixes.  Instead of calling make_bit_field_ref with negative
12053         bitpos return 0.
12055 2017-10-17  Olga Makhotina  <olga.makhotina@intel.com>
12057         * config/i386/avx512dqintrin.h (_mm_mask_reduce_sd,
12058         _mm_maskz_reduce_sd, _mm_mask_reduce_ss,=20
12059         _mm_maskz_reduce_ss): New.
12060         * config/i386/i386-builtin.def (__builtin_ia32_reducesd_mask,
12061         __builtin_ia32_reducess_mask): Ditto..
12062         (__builtin_ia32_reducesd, __builtin_ia32_reducess): Remove.
12063         * config/i386/sse.md (reduces<mode>): Renamed to ...
12064         (reduces<mode><mask_scalar_name>): ... this.
12065         (vreduce<ssescalarmodesuffix>\t{%3, %2, %1, %0|%0, %1, %2, %3}):
12066         Changed to ...
12067         (vreduce<ssescalarmodesuffix>\t{%3, %2, %1, %0<mask_scalar_operand4>|
12068         %0<mask_scalar_operand4>, %1, %2, %3}): ... this.
12070 2017-10-16  David Malcolm  <dmalcolm@redhat.com>
12072         * Makefile.in (OBJS): Add unique-ptr-tests.o.
12073         * selftest-run-tests.c (selftest::run_tests): Call
12074         selftest::unique_ptr_tests_cc_tests.
12075         * selftest.h (selftest::unique_ptr_tests_cc_tests): New decl.
12076         * unique-ptr-tests.cc: New file.
12078 2017-10-16  Vladimir Makarov  <vmakarov@redhat.com>
12080         PR sanitizer/82353
12081         * lra.c (collect_non_operand_hard_regs): Don't ignore operator
12082         locations.
12083         * lra-lives.c (bb_killed_pseudos, bb_gen_pseudos): Move up.
12084         (make_hard_regno_born, make_hard_regno_dead): Update
12085         bb_killed_pseudos and bb_gen_pseudos for fixed regs.
12087 2017-10-16  Jeff Law  <law@redhat.com>
12089         * tree-ssa-dse.c (live_bytes_read): Fix thinko.
12091 2017-10-16  Jan Hubicka  <hubicka@ucw.cz>
12093         * x86-tune-costs.h (znver1_cost): Fix move cost tables.
12095 2017-10-16  Olivier Hainque  <hainque@adacore.com>
12097         * gcc/config.gcc (powerpc*-*-*spe*): Pick 8548 as the default
12098         with_cpu if we were configured for an e500v2 target cpu name.
12100 2017-10-16  Thomas Preud'homme  <thomas.preudhomme@arm.com>
12102         * config/arm/arm-cpus.in (cortex-m33): Add nodsp option.
12103         * doc/invoke.texi: Document +nodsp as a valid extension for
12104         -mcpu=cortex-m33.
12106 2017-10-16  Martin Liska  <mliska@suse.cz>
12108         * sbitmap.c (bitmap_bit_in_range_p_checking): New function.
12109         (test_set_range): Likewise.
12110         (test_range_functions): Rename to ...
12111         (test_bit_in_range): ... this.
12112         (sbitmap_c_tests): Add new test.
12114 2017-10-16  Tamar Christina  <tamar.christina@arm.com>
12116         * config/aarch64/arm_neon.h (vdot_u32, vdotq_u32, vdot_s32, vdotq_s32):
12117         New.
12118         (vdot_lane_u32, vdot_laneq_u32, vdotq_lane_u32, vdotq_laneq_u32): New.
12119         (vdot_lane_s32, vdot_laneq_s32, vdotq_lane_s32, vdotq_laneq_s32): New.
12121 2017-10-16  Tamar Christina  <tamar.christina@arm.com>
12123         * config/aarch64/aarch64-builtins.c
12124         (aarch64_types_quadopu_lane_qualifiers): New.
12125         (TYPES_QUADOPU_LANE): New.
12126         * config/aarch64/aarch64-simd.md (aarch64_<sur>dot<vsi2qi>): New.
12127         (<sur>dot_prod<vsi2qi>, aarch64_<sur>dot_lane<vsi2qi>): New.
12128         (aarch64_<sur>dot_laneq<vsi2qi>): New.
12129         * config/aarch64/aarch64-simd-builtins.def (sdot, udot): New.
12130         (sdot_lane, udot_lane, sdot_laneq, udot_laneq): New.
12131         * config/aarch64/iterators.md (sur): Add UNSPEC_SDOT, UNSPEC_UDOT.
12132         (Vdottype, DOTPROD): New.
12133         (sur): Add SDOT and UDOT.
12135 2017-10-16  Tamar Christina  <tamar.christina@arm.com>
12137         * config/aarch64/aarch64.h (AARCH64_FL_DOTPROD): New.
12138         (AARCH64_ISA_DOTPROD, TARGET_DOTPROD): New.
12139         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
12140         Add TARGET_DOTPROD.
12141         * config/aarch64/aarch64-option-extensions.def (dotprod): New.
12142         * config/aarch64/aarch64-cores.def (cortex-a55, cortex-a75):
12143         Enable TARGET_DOTPROD.
12144         (cortex-a75.cortex-a55): Likewise.
12145         * doc/invoke.texi (aarch64-feature-modifiers): Document dotprod.
12147 2017-10-16  Tamar Christina  <tamar.christina@arm.com>
12149         * config/arm/arm-builtins.c (arm_unsigned_uternop_qualifiers): New.
12150         (UTERNOP_QUALIFIERS, arm_umac_lane_qualifiers, UMAC_LANE_QUALIFIERS):
12151         New.
12152         * config/arm/arm_neon_builtins.def (sdot, udot, sdot_lane, udot_lane):
12153         New.
12154         * config/arm/iterators.md (DOTPROD, VSI2QI, vsi2qi): New.
12155         (UNSPEC_DOT_S, UNSPEC_DOT_U, opsuffix): New.
12156         * config/arm/neon.md (neon_<sup>dot<vsi2qi>): New.
12157         (neon_<sup>dot_lane<vsi2qi>, <sup>dot_prod<vsi2qi>): New.
12158         * config/arm/types.md (neon_dot, neon_dot_q): New.
12159         * config/arm/unspecs.md (sup): Add UNSPEC_DOT_S, UNSPEC_DOT_U.
12161 2017-10-16  Tamar Christina  <tamar.christina@arm.com>
12163         * config/arm/arm.h (TARGET_DOTPROD): New.
12164         * config/arm/arm.c (arm_arch_dotprod): New.
12165         (arm_option_reconfigure_globals): Add arm_arch_dotprod.
12166         * config/arm/arm-c.c (__ARM_FEATURE_DOTPROD): New.
12167         * config/arm/arm-cpus.in (armv8.2-a): Enabled +dotprod.
12168         (feature dotprod, group dotprod, ALL_SIMD_INTERNAL): New.
12169         (ALL_FPU_INTERNAL): Use ALL_SIMD_INTERNAL.
12170         * config/arm/t-multilib (v8_2_a_simd_variants): Add dotprod.
12171         * doc/invoke.texi (armv8.2-a): Document dotprod
12173 2017-10-14  Jan Hubicka  <hubicka@ucw.cz>
12175         * i386.c (ix86_vec_cost): New function.
12176         (ix86_rtx_costs): Handle vector operations better.
12177         * i386.h (struct processor_costs): Add sse_op, fmasd, fmass.
12178         * x86-tune-costs.h: Add new costs to all tables.
12180 2017-10-14  Jan Hubicka  <hubicka@ucw.cz>
12182         * i386.c (ix86_rtx_costs): Make difference between x87 and SSE
12183         operations.
12184         * i386.h (struct processor_costs): Add addss, mulss, mulsd, divss,
12185         divsd, sqrtss and sqrtsd
12186         * x86-tune-costs.h: Add new entries to all costs.
12187         (znver1_cost): Fix to match real instruction latencies.
12189 2017-10-14  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
12190             Michael Collison <michael.collison@arm.com>
12192         * compare-elim.c: Include emit-rtl.h.
12193         (can_merge_compare_into_arith): New function.
12194         (try_validate_parallel): Likewise.
12195         (try_merge_compare): Likewise.
12196         (try_eliminate_compare): Call the above when no previous clobber
12197         is available.
12198         (execute_compare_elim_after_reload): Add DF_UD_CHAIN and DF_DU_CHAIN
12199         dataflow problems.
12201 2017-10-14  Jakub Jelinek  <jakub@redhat.com>
12203         PR middle-end/62263
12204         PR middle-end/82498
12205         * tree-ssa-phiopt.c (value_replacement): Comment fix.  Handle
12206         up to 2 preparation statements for ASSIGN in MIDDLE_BB.
12208         PR middle-end/62263
12209         PR middle-end/82498
12210         * tree-ssa-forwprop.c (simplify_rotate): Allow def_arg1[N]
12211         to be any operand_equal_p operands.  For & (B - 1) require
12212         B to be power of 2.  Recognize
12213         (X << (Y & (B - 1))) | (X >> ((-Y) & (B - 1))) and similar patterns.
12215 2017-10-14  Uros Bizjak  <ubizjak@gmail.com>
12217         PR bootstrap/82553
12218         * optabs.c (expand_memory_blockage): Fix call of
12219         targetm.have_memory_blockage.
12221 2017-10-14  Jakub Jelinek  <jakub@redhat.com>
12223         PR bootstrap/82548
12224         * config.gcc (*-*-solaris2*, i[34567]86-*-cygwin*,
12225         x86_64-*-cygwin*, i[34567]86-*-mingw* | x86_64-*-mingw*): Append
12226         objects to extra_objs instead of overwriting it.
12228 2017-10-14  Uros Bizjak  <ubizjak@gmail.com>
12230         * config/i386/sync.md (FILD_ATOMIC/FIST_ATOMIC FP load peephole2):
12231         Use any_fp_register_operand as operand[3] predicate.  Simplify
12232         equality test for operands[2] and operands[4] memory location.
12233         (LDX_ATOMIC/STX_ATOMIC FP load peephole2): Ditto.
12234         (FILD_ATOMIC/FIST_ATOMIC FP load peephole2 with mem blockage): New.
12235         (LDX_ATOMIC/LDX_ATOMIC FP load peephole2 with mem blockage): Ditto.
12236         (FILD_ATOMIC/FIST_ATOMIC FP store peephole2): Use
12237         any_fp_register_operand as operand[1] predicate.  Simplify
12238         equality test for operands[0] and operands[3] memory location.
12239         (LDX_ATOMIC/STX_ATOMIC FP store peephole2): Ditto.
12240         (FILD_ATOMIC/FIST_ATOMIC FP store peephole2 with mem blockage): New.
12241         (LDX_ATOMIC/LDX_ATOMIC FP storepeephole2 with mem blockage): Ditto.
12243 2017-10-14  Uros Bizjak  <ubizjak@gmail.com>
12245         * target-insns.def: Add memory_blockage.
12246         * optabs.c (expand_memory_blockage): New function.
12247         (expand_asm_memory_barrier): Rename ...
12248         (expand_asm_memory_blockage): ... to this.
12249         (expand_mem_thread_fence): Call expand_memory_blockage
12250         instead of expand_asm_memory_barrier.
12251         (expand_mem_singnal_fence): Ditto.
12252         (expand_atomic_load): Ditto.
12253         (expand_atomic_store): Ditto.
12254         * doc/md.texi (Standard Pattern Names For Generation):
12255         Document memory_blockage instruction pattern.
12257 2017-10-13  Sebastian Perta  <sebastian.perta@renesas.com>
12259         * config/rl78/rl78.c (rl78_emit_libcall): New function.
12260         * config/rl78/rl78-protos.h (rl78_emit_libcall): New function.
12261         * config/rl78/rl78.md: New define_expand "adddi3".
12263 2017-10-13  Jan Hubicka  <hubicka@ucw.cz>
12265         * cfghooks.c (verify_flow_info): Disable check that all probabilities
12266         are set correctly.
12268 2017-10-13  Jeff Law  <law@redhat.com>
12270         * tree-ssa-reassoc.c (reassociate_bb): Clarify code slighly.
12272 2017-10-13  Jakub Jelinek  <jakub@redhat.com>
12274         PR target/82274
12275         * internal-fn.c (expand_mul_overflow): If both operands have
12276         the same highpart of -1 or 0 and the topmost bit of lowpart
12277         is different, overflow is if res <= 0 rather than res < 0.
12279 2017-10-13  Pat Haugen  <pthaugen@us.ibm.com>
12281         * config/rs6000/rs6000.c (rs6000_builtin_vectorization_cost): Remove
12282         TARGET_P9_VECTOR code for unaligned_load case.
12284 2017-10-13  Jan Hubicka  <hubicka@ucw.cz>
12286         * cfghooks.c (verify_flow_info): Check that edge probabilities are set.
12288 2017-10-13  Nathan Sidwell  <nathan@acm.org>
12290         * tree-core.h (tree_contains_struct): Make bool.
12291         * tree.c (tree_contains_struct): Likewise.
12292         * tree.h (MARK_TS_BASE): Remove do ... while (0) idiom.
12293         (MARK_TS_TYPED, MARK_TS_COMMON, MARK_TS_TYPE_COMMON,
12294         MARK_TS_TYPE_WITH_LANG_SPECIFIC, MARK_TS_DECL_MINIMAL,
12295         MARK_TS_DECL_COMMON, MARK_TS_DECL_WRTL, MARK_TS_DECL_WITH_VIS,
12296         MARK_TS_DECL_NON_COMMON): Likewise, use comma operator.
12298 2017-10-13  Richard Biener  <rguenther@suse.de>
12300         * graphite-isl-ast-to-gimple.c
12301         (translate_isl_ast_to_gimple::get_rename_from_scev): Remove unused
12302         parameters and dominance check.
12303         (translate_isl_ast_to_gimple::graphite_copy_stmts_from_block): Adjust.
12304         (translate_isl_ast_to_gimple::copy_bb_and_scalar_dependences): Likewise.
12305         (translate_isl_ast_to_gimple::graphite_regenerate_ast_isl):
12306         Do not update SSA form here or do intermediate IL verification.
12307         * graphite.c: Include tree-ssa.h and tree-into-ssa.h.
12308         (graphite_initialize): Remove check on the number of loops in
12309         the function and inline into graphite_transform_loops.
12310         (graphite_finalize): Inline into graphite_transform_loops.
12311         (graphite_transform_loops): Perform SSA update and IL verification
12312         here.
12313         * params.def (PARAM_GRAPHITE_MIN_LOOPS_PER_FUNCTION): Remove.
12315 2017-10-13  Richard Biener  <rguenther@suse.de>
12317         * graphite-isl-ast-to-gimple.c (max_mode_int_precision,
12318         graphite_expression_type_precision): Avoid global constructor
12319         by moving ...
12320         (translate_isl_ast_to_gimple::translate_isl_ast_to_gimple): Here.
12321         (translate_isl_ast_to_gimple::graphite_expr_type): Add type member.
12322         (translate_isl_ast_to_gimple::translate_isl_ast_node_for): Use it.
12323         (translate_isl_ast_to_gimple::build_iv_mapping): Likewise.
12324         (translate_isl_ast_to_gimple::graphite_create_new_guard): Likewise.
12325         * graphite-sese-to-poly.c (build_original_schedule): Return nothing.
12327 2017-10-13  H.J. Lu  <hongjiu.lu@intel.com>
12329         PR target/82499
12330         * config/i386/i386.h (ix86_red_zone_size): New.
12331         * config/i386/i386.md (push peephole2s): Replace
12332         "!ix86_using_red_zone ()" with "ix86_red_zone_size == 0".
12334 2017-10-13  Richard Sandiford  <richard.sandiford@linaro.org>
12335             Alan Hayward  <alan.hayward@arm.com>
12336             David Sherwood  <david.sherwood@arm.com>
12338         * combine.c (can_change_dest_mode): Reject changes in
12339         REGMODE_NATURAL_SIZE.
12341 2017-10-13  Richard Sandiford  <richard.sandiford@linaro.org>
12342             Alan Hayward  <alan.hayward@arm.com>
12343             David Sherwood  <david.sherwood@arm.com>
12345         * cfgexpand.c (expand_debug_expr): Use GET_MODE_UNIT_BITSIZE.
12346         (expand_debug_source_expr): Likewise.
12347         * combine.c (combine_simplify_rtx): Likewise.
12348         * cse.c (fold_rtx): Likewise.
12349         * fwprop.c (canonicalize_address): Likewise.
12350         * targhooks.c (default_shift_truncation_mask): Likewise.
12352 2017-10-13  Richard Sandiford  <richard.sandiford@linaro.org>
12353             Alan Hayward  <alan.hayward@arm.com>
12354             David Sherwood  <david.sherwood@arm.com>
12356         * optabs.c (add_equal_note): Use GET_MODE_UNIT_SIZE.
12357         (widened_mode): Likewise.
12358         (expand_unop): Likewise.
12359         * ree.c (transform_ifelse): Likewise.
12360         (merge_def_and_ext): Likewise.
12361         (combine_reaching_defs): Likewise.
12362         * simplify-rtx.c (simplify_unary_operation_1): Likewise.
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         * caller-save.c (replace_reg_with_saved_mem): Use byte_lowpart_offset.
12369         * combine.c (gen_lowpart_for_combine): Likewise.
12370         * dwarf2out.c (rtl_for_decl_location): Likewise.
12371         * final.c (alter_subreg): Likewise.
12372         * rtlhooks.c (gen_lowpart_general): Likewise.
12373         (gen_lowpart_if_possible): 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         * calls.c (expand_call): Use subreg_lowpart_offset.
12380         * cse.c (cse_insn): Likewise.
12381         * regcprop.c (copy_value): Likewise.
12382         (copyprop_hardreg_forward_1): Likewise.
12384 2017-10-13  Jakub Jelinek  <jakub@redhat.com>
12386         PR target/82524
12387         * config/i386/i386.md (addqi_ext_1, andqi_ext_1,
12388         *andqi_ext_1_cc, *<code>qi_ext_1, *xorqi_ext_1_cc): Change
12389         =Q constraints to +Q and into insn condition add check
12390         that operands[0] and operands[1] are equal.
12391         (*addqi_ext_2, *andqi_ext_2, *<code>qi_ext_2): Change
12392         =Q constraints to +Q and into insn condition add check
12393         that operands[0] is equal to either operands[1] or operands[2].
12395         PR target/82498
12396         * fold-const.c (fold_binary_loc) <bit_rotate>: Code cleanups,
12397         instead of handling MINUS_EXPR twice (once for each argument),
12398         canonicalize operand order and handle just once, use rtype where
12399         possible.  Handle (A << B) | (A >> (-B & (Z - 1))).
12401         PR target/82498
12402         * config/i386/ia32intrin.h (__rold, __rord, __rolq, __rorq): Allow
12403         any values of __C while still being pattern recognizable as a simple
12404         rotate instruction.
12406 2017-10-13  Richard Biener  <rguenther@suse.de>
12408         PR tree-optimization/82451
12409         Revert
12410         2017-10-02  Richard Biener  <rguenther@suse.de>
12412         PR tree-optimization/82355
12413         * graphite-isl-ast-to-gimple.c (build_iv_mapping): Also build
12414         a mapping for the enclosing loop but avoid generating one for
12415         the loop tree root.
12416         (copy_bb_and_scalar_dependences): Remove premature codegen
12417         error on PHIs in blocks duplicated into multiple places.
12418         * graphite-scop-detection.c
12419         (scop_detection::stmt_has_simple_data_refs_p): For a loop not
12420         in the region use it as loop and nest to analyze the DR in.
12421         (try_generate_gimple_bb): Likewise.
12422         * graphite-sese-to-poly.c (extract_affine_chrec): Adjust.
12423         (add_loop_constraints): For blocks in a loop not in the region
12424         create a dimension with a single iteration.
12425         * sese.h (gbb_loop_at_index): Remove assert.
12427         * cfgloop.c (loop_preheader_edge): For the loop tree root
12428         return the single successor of the entry block.
12429         * graphite-isl-ast-to-gimple.c (graphite_regenerate_ast_isl):
12430         Reset the SCEV hashtable and niters.
12431         * graphite-scop-detection.c
12432         (scop_detection::graphite_can_represent_scev): Add SCOP parameter,
12433         assert that we only have POLYNOMIAL_CHREC that vary in loops
12434         contained in the region.
12435         (scop_detection::graphite_can_represent_expr): Adjust.
12436         (scop_detection::stmt_has_simple_data_refs_p): For loops
12437         not in the region set loop to NULL.  The nest is now the
12438         entry edge to the region.
12439         (try_generate_gimple_bb): Likewise.
12440         * sese.c (scalar_evolution_in_region): Adjust for
12441         instantiate_scev change.
12442         * tree-data-ref.h (graphite_find_data_references_in_stmt):
12443         Make nest parameter the edge into the region.
12444         (create_data_ref): Likewise.
12445         * tree-data-ref.c (dr_analyze_indices): Make nest parameter an
12446         entry edge into a region and adjust instantiate_scev calls.
12447         (create_data_ref): Likewise.
12448         (graphite_find_data_references_in_stmt): Likewise.
12449         (find_data_references_in_stmt): Pass the loop preheader edge
12450         from the nest argument.
12451         * tree-scalar-evolution.h (instantiate_scev): Make instantiate_below
12452         parameter the edge into the region.
12453         (instantiate_parameters): Use the loop preheader edge as entry.
12454         * tree-scalar-evolution.c (analyze_scalar_evolution): Handle
12455         NULL loop.
12456         (get_instantiated_value_entry): Make instantiate_below parameter
12457         the edge into the region.
12458         (instantiate_scev_name): Likewise.  Adjust dominance checks,
12459         when we cannot use loop-based instantiation instantiate by
12460         walking use-def chains.
12461         (instantiate_scev_poly): Adjust.
12462         (instantiate_scev_binary): Likewise.
12463         (instantiate_scev_convert): Likewise.
12464         (instantiate_scev_not): Likewise.
12465         (instantiate_array_ref): Remove.
12466         (instantiate_scev_3): Likewise.
12467         (instantiate_scev_2): Likewise.
12468         (instantiate_scev_1): Likewise.
12469         (instantiate_scev_r): Do not blindly handle N-operand trees.
12470         Do not instantiate array-refs.  Handle all constants and invariants.
12471         (instantiate_scev): Make instantiate_below parameter
12472         the edge into the region.
12473         (resolve_mixers): Use the loop preheader edge for the region
12474         parameter to instantiate_scev_r.
12475         * tree-ssa-loop-prefetch.c (determine_loop_nest_reuse): Adjust.
12477 2017-10-13  Richard Biener  <rguenther@suse.de>
12479         PR tree-optimization/82525
12480         * graphite-isl-ast-to-gimple.c
12481         (translate_isl_ast_to_gimple::widest_int_from_isl_expr_int): Split
12482         out from ...
12483         (translate_isl_ast_to_gimple::gcc_expression_from_isl_expr_int): Here.
12484         Fail code generation when we cannot represent the isl integer.
12485         (binary_op_to_tree): Elide modulo operations that are no-ops
12486         in the type we code generate.  Remove now superfluous code
12487         generation errors.
12489 2017-10-13  Richard Biener  <rguenther@suse.de>
12491         * graphite-scop-detection.c (loop_ivs_can_be_represented): Remove.
12492         (scop_detection::harmful_loop_in_region): Remove premature
12493         IV type restriction.
12494         (scop_detection::graphite_can_represent_scev): We can handle
12495         pointer IVs just fine.
12497 2017-10-13  Alan Modra  <amodra@gmail.com>
12499         * doc/extend.texi (Extended Asm <Clobbers>): Rename to
12500         "Clobbers and Scratch Registers".  Add paragraph on
12501         alternative to clobbers for scratch registers and OpenBLAS
12502         example.
12504 2017-10-13  Alan Modra  <amodra@gmail.com>
12506         * doc/extend.texi (Clobbers): Correct vax example.  Delete old
12507         example of a memory input for a string of known length.  Move
12508         commentary out of table.  Add a number of new examples
12509         covering array memory inputs.
12511 2017-10-12  Martin Liska  <mliska@suse.cz>
12513         PR tree-optimization/82493
12514         * sbitmap.c (bitmap_bit_in_range_p): Fix the implementation.
12515         (test_range_functions): New function.
12516         (sbitmap_c_tests): Likewise.
12517         * selftest-run-tests.c (selftest::run_tests): Run new tests.
12518         * selftest.h (sbitmap_c_tests): New function.
12520         * tree-ssa-dse.c (live_bytes_read): Fix thinko.
12522 2017-10-12  Michael Meissner  <meissner@linux.vnet.ibm.com>
12524         * config/rs6000/amo.h: Fix spacing issue.
12526 2017-10-12  Jakub Jelinek  <jakub@redhat.com>
12528         PR target/82498
12529         * config/i386/i386.md (*ashl<mode>3_mask_1,
12530         *<shift_insn><mode>3_mask_1, *<rotate_insn><mode>3_mask_1,
12531         *<btsc><mode>_mask_1, *btr<mode>_mask_1): New define_insn_and_split
12532         patterns.
12534 2017-10-12  Jan Hubicka  <hubicka@ucw.cz>
12536         * profile-count.h (safe_scale_64bit): Fix GCC4.x path.
12537         (profile_probability): Set max_probability
12538         to (uint32_t) 1 << (n_bits - 2) and update accessors to avoid overlfows
12539         in temporaries.
12540         * profile-count.c (profile_probability::differs_from_p): Do not
12541         rely on max_probaiblity == 10000
12543 2017-10-12  Jeff Law  <law@redhat.com>
12545         * tree-ssa-dse.c (valid_ao_ref_for_dse): Reject ao_refs with
12546         negative offsets.
12548 2017-10-12  Martin Sebor  <msebor@redhat.com>
12550         PR other/82301
12551         PR c/82435
12552         * cgraphunit.c (maybe_diag_incompatible_alias): New function.
12553         (handle_alias_pairs): Call it.
12554         * common.opt (-Wattribute-alias): New option.
12555         * doc/extend.texi (ifunc attribute): Discuss C++ specifics.
12556         * doc/invoke.texi (-Wattribute-alias): Document.
12558 2017-10-12  Vladimir Makarov  <vmakarov@redhat.com>
12560         Revert
12561         2017-10-11  Vladimir Makarov  <vmakarov@redhat.com>
12562         PR sanitizer/82353
12563         * lra.c (collect_non_operand_hard_regs): Don't ignore operator
12564         locations.
12565         * lra-lives.c (bb_killed_pseudos, bb_gen_pseudos): Move up.
12566         (make_hard_regno_born, make_hard_regno_dead): Update
12567         bb_killed_pseudos and bb_gen_pseudos.
12569 2017-10-12  Jan Hubicka  <hubicka@ucw.cz>
12571         * config/i386/x86-tune-sched.c (ix86_adjust_cost): Fix Zen support.
12573 2017-10-12  Uros Bizjak  <ubizjak@gmail.com>
12575         * config/alpha/alpha.c (alpha_split_conditional_move):
12576         Use std::swap instead of manually swapping.
12577         (alpha_stdarg_optimize_hook): Ditto.
12578         (alpha_canonicalize_comparison): Ditto.
12580 2017-10-12  Bin Cheng  <bin.cheng@arm.com>
12582         * tree-loop-distribution.c (struct builtin_info): New struct.
12583         (struct partition): Refactor fields into struct builtin_info.
12584         (partition_free): Free struct builtin_info.
12585         (build_size_arg_loc, build_addr_arg_loc): Delete.
12586         (generate_memset_builtin, generate_memcpy_builtin): Get memory range
12587         information from struct builtin_info.
12588         (find_single_drs): New function refactored from classify_partition.
12589         Also moved builtin validity checks to this function.
12590         (compute_access_range, alloc_builtin): New functions.
12591         (classify_builtin_st, classify_builtin_ldst): New functions.
12592         (classify_partition): Refactor code into functions find_single_drs,
12593         classify_builtin_st and classify_builtin_ldst.
12594         (distribute_loop): Don't do runtime alias check when distributing
12595         loop nest.
12596         (find_seed_stmts_for_distribution): New function.
12597         (pass_loop_distribution::execute): Refactor code finding seed
12598         stmts into above function.  Support distribution for the innermost
12599         two-level loop nest.  Adjust dump information.
12601 2017-10-12  Bin Cheng  <bin.cheng@arm.com>
12603         * tree-loop-distribution.c: Adjust the general comment.
12604         (NUM_PARTITION_THRESHOLD): New macro.
12605         (ssa_name_has_uses_outside_loop_p): Support loop nest distribution.
12606         (classify_partition): Skip builtin pattern of loop nest's inner loop.
12607         (merge_dep_scc_partitions): New parameter ignore_alias_p and use it
12608         in call to build_partition_graph.
12609         (finalize_partitions): New parameter.  Make loop distribution more
12610         conservative by fusing more partitions.
12611         (distribute_loop): Don't do runtime alias check in case of loop nest
12612         distribution.
12613         (find_seed_stmts_for_distribution): New function.
12614         (prepare_perfect_loop_nest): New function.
12615         (pass_loop_distribution::execute): Refactor code finding seed stmts
12616         and loop nest into above functions.  Support loop nest distribution.
12617         Adjust dump information accordingly.
12619 2017-10-12  Bin Cheng  <bin.cheng@arm.com>
12621         * tree-loop-distribution.c (break_alias_scc_partitions): Add comment
12622         and set PTYPE_SEQUENTIAL for merged partition.
12624 2017-10-12  Richard Biener  <rguenther@suse.de>
12626         PR tree-optimization/69728
12627         Revert
12628         2017-09-19  Richard Biener  <rguenther@suse.de>
12630         PR tree-optimization/69728
12631         * graphite-sese-to-poly.c (schedule_error): New global.
12632         (add_loop_schedule): Handle empty domain by failing the
12633         schedule.
12634         (build_original_schedule): Handle schedule_error.
12636         * graphite-sese-to-poly.c (add_loop_schedule): Handle empty
12637         domain by returning an unchanged schedule.
12639 2017-10-12  Jakub Jelinek  <jakub@redhat.com>
12641         * genrecog.c (validate_pattern): For VEC_SELECT verify that
12642         CONST_INT selectors are 0 to GET_MODE_NUNITS (imode) - 1.
12644 2017-10-12  Aldy Hernandez  <aldyh@redhat.com>
12646         * Makefile.in (TAGS): Merge all the *.def files into one pattern.
12647         Handle params.def.
12649 2017-10-12  Jakub Jelinek  <jakub@redhat.com>
12651         PR c++/82159
12652         * expr.c (store_field): Don't optimize away bitsize == 0 store
12653         from CALL_EXPR with addressable return type.
12655 2017-10-11  Segher Boessenkool  <segher@kernel.crashing.org>
12657         * config/rs6000/rs6000.h (TARGET_ISEL64): Delete.
12658         * config/rs6000/rs6000.md (sel): Delete mode attribute.
12659         (mov<mode>cc, isel_signed_<mode>, isel_unsigned_<mode>,
12660         *isel_reversed_signed_<mode>, *isel_reversed_unsigned_<mode>): Use
12661         TARGET_ISEL instead of TARGET_ISEL<sel>.
12663 2017-10-11  David Edelsohn  <dje.gcc@gmail.com>
12665         * config/rs6000/rs6000.c
12666         (rs6000_xcoff_asm_output_aligned_decl_common): Test for NULL decl.
12668 2017-10-11  Segher Boessenkool  <segher@kernel.crashing.org>
12670         * config/rs6000/predicates.md (zero_constant, all_ones_constant):
12671         Move up in file.
12672         (reg_or_cint_operand): Fix comment.
12673         (reg_or_zero_operand): New predicate.
12674         * config/rs6000/rs6000-protos.h (output_isel): Delete.
12675         * config/rs6000/rs6000.c (output_isel): Delete.
12676         * config/rs6000/rs6000.md (isel_signed_<mode>): Use reg_or_zero_operand
12677         instead of reg_or_cint_operand.  Output instruction directly (not via
12678         output_isel).
12679         (isel_unsigned_<mode>): Ditto.
12680         (*isel_reversed_signed_<mode>): Use reg_or_zero_operand instead of
12681         gpc_reg_operand.  Add an instruction alternative for this.  Output
12682         instruction directly.
12683         (*isel_reversed_unsigned_<mode>): Ditto.
12685 2017-10-11  Uros Bizjak  <ubizjak@gmail.com>
12687         * config/i386/i386.c (ix86_canonicalize_comparison): New function.
12688         (TARGET_CANONICALIZE_COMPARISON): Define.
12690 2017-10-11  Qing Zhao  <qing.zhao@oracle.com>
12692         PR target/81422
12693         * config/aarch64/aarch64.c (aarch64_load_symref_appropriately):
12694         Check whether the dest is REG before adding REG_EQUIV note.
12696 2017-10-11  Vladimir Makarov  <vmakarov@redhat.com>
12698         PR sanitizer/82353
12699         * lra.c (collect_non_operand_hard_regs): Don't ignore operator
12700         locations.
12701         * lra-lives.c (bb_killed_pseudos, bb_gen_pseudos): Move up.
12702         (make_hard_regno_born, make_hard_regno_dead): Update
12703         bb_killed_pseudos and bb_gen_pseudos.
12705 2017-10-11  Nathan Sidwell  <nathan@acm.org>
12707         * incpath.h (enum incpath_kind): Name enum, prefix values.
12708         (add_path, add_cpp_dir_path, get_added_cpp_dirs): Use incpath_kind.
12709         * incpath.c (heads, tails): Use INC_MAX.
12710         (add_env_var_paths, add_standard_paths): Use incpath_kind.
12711         (merge_include_chains, split_quote_chain,
12712         register_include_chains): Update incpath_kind names.
12713         (add_cpp_dir_path, add_path, get_added_cpp_dirs): Use incpath_kind.
12714         * config/darwin-c.c (add_system_framework_path): Update incpath_kind
12715         names.
12716         (add_framework_path, darwin_register_objc_includes): Likewise.
12717         * config/vms/vms-c.c (vms_c_register_includes): Likewise.
12719 2017-10-11  Uros Bizjak  <ubizjak@gmail.com>
12721         * config/i386/i386.md (*cmp<X87MODEF:mode>_<SWI24:mode>_i387):
12722         Do not use float_operator operator predicate.
12723         (*cmp<X87MODEF:mode>_<SWI24:mode>_cc_i387): Ditto.
12724         * config/i386/predicates.md (float_operator): Remove predicate.
12726 2017-10-11  Uros Bizjak  <ubizjak@gmail.com>
12728         * config/i386/i386.md (*jcc<mode>_0_i387): Remove insn pattern.
12729         (*jccxf_i387): Ditto.
12730         (*jcc<mode>_i387): Ditto.
12731         (*jccu<mode>_i387): Ditto.
12732         (*jcc<X87MODEF:mode>_<SWI24:mode>_i387): Ditto.
12733         (*jcc_*_i387 splitters): Remove.
12734         * config/i386/i386-protos.h (ix86_split_fp_branch): Remove prototype.
12735         * config/i386/i386.c (ix86_split_fp_branch): Remove.
12736         * config/i386/predicates.md (ix86_swapped_fp_comparison_operator):
12737         Remove predicate.
12739 2017-10-11  Jan Hubicka  <hubicka@ucw.cz>
12741         * profile-count.h (slow_safe_scale_64bit): New function.
12742         (safe_scale_64bit): New inline.
12743         (profile_count::max_safe_multiplier): Remove; use safe_scale_64bit.
12744         * profile-count.c: Include wide-int.h
12745         (slow_safe_scale_64bit): New.
12747 2017-10-11  Nathan Sidwell  <nathan@acm.org>
12749         * tree.h (DECL_ASSEMBLER_NAME_SET_P): Don't check
12750         HAS_DECL_ASSEMBLER_NAME_P.
12751         * gimple-expr.c (gimple_decl_printable_name: Check
12752         HAS_DECL_ASSEMBLER_NAME_P too.
12753         * ipa-utils.h (type_in_anonymous_namespace_p): Check
12754         DECL_ASSEMBLER_NAME_SET_P of TYPE_NAME.
12755         (odr_type_p): No need to assert TYPE_NAME is a TYPE_DECL.
12756         * passes.c (rest_of_decl_compilation): Check
12757         HAS_DECL_ASSEMBLER_NAME_P too.
12758         * recog.c (verify_changes): Likewise.
12759         * tree-pretty-print.c (dump_decl_name): Likewise.
12760         * tree-ssa-structalias.c (alias_get_name): Likewise.  Reimplement.
12762         * tree.h (DECL_ASSEMBLER_NAME_RAW): New.
12763         (SET_DECL_ASSEMBLER_NAME): Use it.
12764         (DECL_ASSEMBLER_NAME_SET_P): Likewise.
12765         (COPY_DECL_ASSEMBLER_NAME): Likewise.
12766         * tree.c (decl_assembler_name): Use DECL_ASSEMBLER_NAME_RAW.
12768 2017-10-11  Jan Hubicka  <hubicka@ucw.cz>
12770         * config.gcc (i386, x86_64): Add extra objects.
12771         * config/i386/i386-protos.h (ix86_rip_relative_addr_p): Declare.
12772         (ix86_min_insn_size): Declare.
12773         (ix86_issue_rate): Declare.
12774         (ix86_adjust_cost): Declare.
12775         (ia32_multipass_dfa_lookahead): Declare.
12776         (ix86_macro_fusion_p): Declare.
12777         (ix86_macro_fusion_pair_p): Declare.
12778         (ix86_bd_has_dispatch): Declare.
12779         (ix86_bd_do_dispatch): Declare.
12780         (ix86_core2i7_init_hooks): Declare.
12781         (ix86_atom_sched_reorder): Declare.
12782         * config/i386/i386.c Move all CPU cost tables to x86-tune-costs.h.
12783         (COSTS_N_BYTES): Move to x86-tune-costs.h.
12784         (DUMMY_STRINGOP_ALGS):Move to x86-tune-costs.h.
12785         (rip_relative_addr_p): Rename to ...
12786         (ix86_rip_relative_addr_p): ... this one; export.
12787         (memory_address_length): Update.
12788         (ix86_issue_rate): Move to x86-tune-sched.c.
12789         (ix86_flags_dependent): Move to x86-tune-sched.c.
12790         (ix86_agi_dependent): Move to x86-tune-sched.c.
12791         (exact_dependency_1): Move to x86-tune-sched.c.
12792         (exact_store_load_dependency): Move to x86-tune-sched.c.
12793         (ix86_adjust_cost): Move to x86-tune-sched.c.
12794         (ia32_multipass_dfa_lookahead): Move to x86-tune-sched.c.
12795         (ix86_macro_fusion_p): Move to x86-tune-sched.c.
12796         (ix86_macro_fusion_pair_p): Move to x86-tune-sched.c.
12797         (do_reorder_for_imul): Move to x86-tune-sched-atom.c.
12798         (swap_top_of_ready_list): Move to x86-tune-sched-atom.c.
12799         (ix86_sched_reorder): Move to x86-tune-sched-atom.c.
12800         (core2i7_first_cycle_multipass_init): Move to x86-tune-sched-core.c.
12801         (core2i7_dfa_post_advance_cycle): Move to x86-tune-sched-core.c.
12802         (min_insn_size): Rename to ...
12803         (ix86_min_insn_size): ... this one; export.
12804         (core2i7_first_cycle_multipass_begin): Move to x86-tune-sched-core.c.
12805         (core2i7_first_cycle_multipass_issue): Move to x86-tune-sched-core.c.
12806         (core2i7_first_cycle_multipass_backtrack): Move to
12807         x86-tune-sched-core.c.
12808         (core2i7_first_cycle_multipass_end): Move to x86-tune-sched-core.c.
12809         (core2i7_first_cycle_multipass_fini): Move to x86-tune-sched-core.c.
12810         (ix86_sched_init_global): Break up logic to ix86_core2i7_init_hooks.
12811         (ix86_avoid_jump_mispredicts): Update.
12812         (TARGET_SCHED_DISPATCH): Move to ix86-tune-sched-bd.c.
12813         (TARGET_SCHED_DISPATCH_DO): Move to ix86-tune-sched-bd.c.
12814         (TARGET_SCHED_REORDER): Move to ix86-tune-sched-bd.c.
12815         (DISPATCH_WINDOW_SIZE): Move to ix86-tune-sched-bd.c.
12816         (MAX_DISPATCH_WINDOWS): Move to ix86-tune-sched-bd.c.
12817         (MAX_INSN): Move to ix86-tune-sched-bd.c.
12818         (MAX_IMM): Move to ix86-tune-sched-bd.c.
12819         (MAX_IMM_SIZE): Move to ix86-tune-sched-bd.c.
12820         (MAX_IMM_32): Move to ix86-tune-sched-bd.c.
12821         (MAX_IMM_64): Move to ix86-tune-sched-bd.c.
12822         (MAX_LOAD): Move to ix86-tune-sched-bd.c.
12823         (MAX_STORE): Move to ix86-tune-sched-bd.c.
12824         (BIG): Move to ix86-tune-sched-bd.c.
12825         (enum dispatch_group): Move to ix86-tune-sched-bd.c.
12826         (enum insn_path): Move to ix86-tune-sched-bd.c.
12827         (get_mem_group): Move to ix86-tune-sched-bd.c.
12828         (is_cmp): Move to ix86-tune-sched-bd.c.
12829         (dispatch_violation): Move to ix86-tune-sched-bd.c.
12830         (is_branch): Move to ix86-tune-sched-bd.c.
12831         (is_prefetch): Move to ix86-tune-sched-bd.c.
12832         (init_window): Move to ix86-tune-sched-bd.c.
12833         (allocate_window): Move to ix86-tune-sched-bd.c.
12834         (init_dispatch_sched): Move to ix86-tune-sched-bd.c.
12835         (is_end_basic_block): Move to ix86-tune-sched-bd.c.
12836         (process_end_window): Move to ix86-tune-sched-bd.c.
12837         (allocate_next_window): Move to ix86-tune-sched-bd.c.
12838         (find_constant): Move to ix86-tune-sched-bd.c.
12839         (get_num_immediates): Move to ix86-tune-sched-bd.c.
12840         (has_immediate): Move to ix86-tune-sched-bd.c.
12841         (get_insn_path): Move to ix86-tune-sched-bd.c.
12842         (get_insn_group): Move to ix86-tune-sched-bd.c.
12843         (count_num_restricted): Move to ix86-tune-sched-bd.c.
12844         (fits_dispatch_window): Move to ix86-tune-sched-bd.c.
12845         (add_insn_window): Move to ix86-tune-sched-bd.c.
12846         (add_to_dispatch_window): Move to ix86-tune-sched-bd.c.
12847         (debug_dispatch_window_file): Move to ix86-tune-sched-bd.c.
12848         (debug_dispatch_window): Move to ix86-tune-sched-bd.c.
12849         (debug_insn_dispatch_info_file): Move to ix86-tune-sched-bd.c.
12850         (debug_ready_dispatch): Move to ix86-tune-sched-bd.c.
12851         (do_dispatch): Move to ix86-tune-sched-bd.c.
12852         (has_dispatch): Move to ix86-tune-sched-bd.c.
12853         * config/i386/t-i386: Add new object files.
12854         * config/i386/x86-tune-costs.h: New file.
12855         * config/i386/x86-tune-sched-atom.c: New file.
12856         * config/i386/x86-tune-sched-bd.c: New file.
12857         * config/i386/x86-tune-sched-core.c: New file.
12858         * config/i386/x86-tune-sched.c: New file.
12860 2017-10-11  Liu Hao  <lh_mouse@126.com>
12862         * pretty-print.c [_WIN32] (colorize_init): Remove.  Use
12863         the generic version below instead.
12864         (should_colorize): Recognize Windows consoles as terminals
12865         for MinGW targets.
12866         * pretty-print.c [__MINGW32__] (write_all): New function.
12867         [__MINGW32__] (find_esc_head): Likewise.
12868         [__MINGW32__] (find_esc_terminator): Likewise.
12869         [__MINGW32__] (eat_esc_sequence): Likewise.
12870         [__MINGW32__] (mingw_ansi_fputs): New function that handles
12871         ANSI escape codes.
12872         (pp_write_text_to_stream): Use mingw_ansi_fputs instead of fputs
12873         for MinGW targets.
12875 2017-10-11  Richard Biener  <rguenther@suse.de>
12877         * tree-ssa-loop-niter.c (infer_loop_bounds_from_pointer_arith):
12878         Properly call analyze_scalar_evolution with the loop of the stmt.
12880 2017-10-11  Richard Biener  <rguenther@suse.de>
12882         * tree.def (POLYNOMIAL_CHREC): Remove CHREC_VARIABLE tree operand.
12883         * tree-core.h (tree_base): Add chrec_var union member.
12884         * tree.h (CHREC_VAR): Remove.
12885         (CHREC_LEFT, CHREC_RIGHT, CHREC_VARIABLE): Adjust.
12886         * tree-chrec.h (build_polynomial_chrec): Adjust.
12887         * tree-chrec.c (reset_evolution_in_loop): Use build_polynomial_chrec.
12888         * tree-pretty-print.c (dump_generic_node): Use CHREC_VARIABLE.
12890 2017-10-11  Marc Glisse  <marc.glisse@inria.fr>
12892         * fold-const.c (fold_binary_loc) [X +- Y CMP X]: Move ...
12893         * match.pd: ... here.
12894         ((T) X == (T) Y): Relax condition.
12896 2017-10-11  Bin Cheng  <bin.cheng@arm.com>
12898         PR tree-optimization/82472
12899         * tree-loop-distribution.c (sort_partitions_by_post_order): Refine
12900         comment.
12901         (break_alias_scc_partitions): Update postorder number.
12903 2017-10-11  Martin Liska  <mliska@suse.cz>
12905         PR sanitizer/82490
12906         * opts.c (parse_no_sanitize_attribute): Do not use error_value
12907         variable.
12908         * opts.h (parse_no_sanitize_attribute): Remove last argument.
12910 2017-10-11  Martin Liska  <mliska@suse.cz>
12912         * print-rtl.c (print_insn): Move declaration of idbuf
12913         to same scope as name.
12915 2017-10-11  Martin Liska  <mliska@suse.cz>
12917         Revert r253637:
12919         PR sanitizer/82484
12920         * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
12921         volatile arguments.
12923 2017-10-11  Martin Liska  <mliska@suse.cz>
12925         PR sanitizer/82484
12926         * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
12927         volatile arguments.
12929 2017-10-11  Adhemerval Zanella  <adhemerval.zanella@linaro.org>
12931         * config.gcc (default_gnu_indirect_function): Default to yes for
12932         arm*-*-linux* with glibc.
12934 2017-10-11  Richard Biener  <rguenther@suse.de>
12936         * tree-scalar-evolution.c (get_scalar_evolution): Handle
12937         default-defs and types we do not want to analyze.
12938         (interpret_loop_phi): Replace unreachable code with an assert.
12939         (compute_scalar_evolution_in_loop): Remove and inline ...
12940         (analyze_scalar_evolution_1): ... here, replacing condition with
12941         what makes the intent clearer.  Remove handling of cases
12942         get_scalar_evolution now handles.
12944 2017-10-10  Jim Wilson  <wilson@tuliptree.org>
12946         PR rtl-optimization/81434
12947         * haifa-sched.c (prune_ready_list): Init min_cost_group to 0.  Update
12948         comment for main loop.  In sched_group_found if, also add checks for
12949         pass and min_cost_group.
12951 2017-10-10  Segher Boessenkool  <segher@kernel.crashing.org>
12953         * config/rs6000/rs6000.c (TARGET_INSN_COST): New.
12954         (rs6000_insn_cost): New function.
12955         * config/rs6000/rs6000.md (cost): New attribute.
12957 2017-10-10  Jakub Jelinek  <jakub@redhat.com>
12958             H.J. Lu  <hongjiu.lu@intel.com>
12960         PR target/79565
12961         PR target/82483
12962         * config/i386/i386.c (ix86_init_mmx_sse_builtins): Add
12963         OPTION_MASK_ISA_MMX for __builtin_ia32_maskmovq,
12964         __builtin_ia32_vec_ext_v4hi and __builtin_ia32_vec_set_v4hi.
12965         (ix86_expand_builtin): Treat OPTION_MASK_ISA_MMX similarly
12966         to OPTION_MASK_ISA_AVX512VL - builtins that have both
12967         OPTION_MASK_ISA_MMX and some other bit set require both
12968         mmx and the ISAs without the mmx bit.
12969         * config/i386/i386-builtin.def (__builtin_ia32_cvtps2pi,
12970         __builtin_ia32_cvttps2pi, __builtin_ia32_cvtpi2ps,
12971         __builtin_ia32_pavgb, __builtin_ia32_pavgw, __builtin_ia32_pmulhuw,
12972         __builtin_ia32_pmaxub, __builtin_ia32_pmaxsw, __builtin_ia32_pminub,
12973         __builtin_ia32_pminsw, __builtin_ia32_psadbw, __builtin_ia32_pmovmskb,
12974         __builtin_ia32_pshufw, __builtin_ia32_cvtpd2pi,
12975         __builtin_ia32_cvttpd2pi, __builtin_ia32_cvtpi2pd,
12976         __builtin_ia32_pmuludq, __builtin_ia32_pabsb, __builtin_ia32_pabsw,
12977         __builtin_ia32_pabsd, __builtin_ia32_phaddw, __builtin_ia32_phaddd,
12978         __builtin_ia32_phaddsw, __builtin_ia32_phsubw, __builtin_ia32_phsubd,
12979         __builtin_ia32_phsubsw, __builtin_ia32_pmaddubsw,
12980         __builtin_ia32_pmulhrsw, __builtin_ia32_pshufb, __builtin_ia32_psignb,
12981         __builtin_ia32_psignw, __builtin_ia32_psignd, __builtin_ia32_movntq,
12982         __builtin_ia32_paddq, __builtin_ia32_psubq, __builtin_ia32_palignr):
12983         Add OPTION_MASK_ISA_MMX.
12985 2017-10-10  Andreas Tobler  <andreast@gcc.gnu.org>
12987         * config.gcc (armv7*-*-freebsd*): New target.
12988         (armv6*-*-freebsd*): Remove obsolete TARGET_FREEBSD_ARMv6 define.
12990 2017-10-10  Jan Hubicka  <hubicka@ucw.cz>
12992         * x86-tune.def (X86_TUNE_AVOID_FALSE_DEP_FOR_BMI,
12993         X86_TUNE_ADJUST_UNROLL, X86_TUNE_ONE_IF_CONV_INSN): Move to right
12994         spot in the file.
12996 2017-10-10  Richard Sandiford  <richard.sandiford@linaro.org>
12998         * wide-int.h (wide_int_ref_storage): Make host_dependent_precision
12999         a template parameter.
13000         (WIDE_INT_REF_FOR): Update accordingly.
13001         * tree.h (wi::int_traits <const_tree>): Delete.
13002         (wi::tree_to_widest_ref, wi::tree_to_offset_ref): New typedefs.
13003         (wi::to_widest, wi::to_offset): Use them.  Expand commentary.
13004         (wi::tree_to_wide_ref): New typedef.
13005         (wi::to_wide): New function.
13006         * calls.c (get_size_range): Use wi::to_wide when operating on
13007         trees as wide_ints.
13008         * cgraph.c (cgraph_node::create_thunk): Likewise.
13009         * config/i386/i386.c (ix86_data_alignment): Likewise.
13010         (ix86_local_alignment): Likewise.
13011         * dbxout.c (stabstr_O): Likewise.
13012         * dwarf2out.c (add_scalar_info, gen_enumeration_type_die): Likewise.
13013         * expr.c (const_vector_from_tree): Likewise.
13014         * fold-const-call.c (host_size_t_cst_p, fold_const_call_1): Likewise.
13015         * fold-const.c (may_negate_without_overflow_p, negate_expr_p)
13016         (fold_negate_expr_1, int_const_binop_1, const_binop)
13017         (fold_convert_const_int_from_real, optimize_bit_field_compare)
13018         (all_ones_mask_p, sign_bit_p, unextend, extract_muldiv_1)
13019         (fold_div_compare, fold_single_bit_test, fold_plusminus_mult_expr)
13020         (pointer_may_wrap_p, expr_not_equal_to, fold_binary_loc)
13021         (fold_ternary_loc, multiple_of_p, fold_negate_const, fold_abs_const)
13022         (fold_not_const, round_up_loc): Likewise.
13023         * gimple-fold.c (gimple_fold_indirect_ref): Likewise.
13024         * gimple-ssa-warn-alloca.c (alloca_call_type_by_arg): Likewise.
13025         (alloca_call_type): Likewise.
13026         * gimple.c (preprocess_case_label_vec_for_gimple): Likewise.
13027         * godump.c (go_output_typedef): Likewise.
13028         * graphite-sese-to-poly.c (tree_int_to_gmp): Likewise.
13029         * internal-fn.c (get_min_precision): Likewise.
13030         * ipa-cp.c (ipcp_store_vr_results): Likewise.
13031         * ipa-polymorphic-call.c
13032         (ipa_polymorphic_call_context::ipa_polymorphic_call_context): Likewise.
13033         * ipa-prop.c (ipa_print_node_jump_functions_for_edge): Likewise.
13034         (ipa_modify_call_arguments): Likewise.
13035         * match.pd: Likewise.
13036         * omp-low.c (scan_omp_1_op, lower_omp_ordered_clauses): Likewise.
13037         * print-tree.c (print_node_brief, print_node): Likewise.
13038         * stmt.c (expand_case): Likewise.
13039         * stor-layout.c (layout_type): Likewise.
13040         * tree-affine.c (tree_to_aff_combination): Likewise.
13041         * tree-cfg.c (group_case_labels_stmt): Likewise.
13042         * tree-data-ref.c (dr_analyze_indices): Likewise.
13043         (prune_runtime_alias_test_list): Likewise.
13044         * tree-dump.c (dequeue_and_dump): Likewise.
13045         * tree-inline.c (remap_gimple_op_r, copy_tree_body_r): Likewise.
13046         * tree-predcom.c (is_inv_store_elimination_chain): Likewise.
13047         * tree-pretty-print.c (dump_generic_node): Likewise.
13048         * tree-scalar-evolution.c (iv_can_overflow_p): Likewise.
13049         (simple_iv_with_niters): Likewise.
13050         * tree-ssa-address.c (addr_for_mem_ref): Likewise.
13051         * tree-ssa-ccp.c (ccp_finalize, evaluate_stmt): Likewise.
13052         * tree-ssa-loop-ivopts.c (constant_multiple_of): Likewise.
13053         * tree-ssa-loop-niter.c (split_to_var_and_offset)
13054         (refine_value_range_using_guard, number_of_iterations_ne_max)
13055         (number_of_iterations_lt_to_ne, number_of_iterations_lt)
13056         (get_cst_init_from_scev, record_nonwrapping_iv)
13057         (scev_var_range_cant_overflow): Likewise.
13058         * tree-ssa-phiopt.c (minmax_replacement): Likewise.
13059         * tree-ssa-pre.c (compute_avail): Likewise.
13060         * tree-ssa-sccvn.c (vn_reference_fold_indirect): Likewise.
13061         (vn_reference_maybe_forwprop_address, valueized_wider_op): Likewise.
13062         * tree-ssa-structalias.c (get_constraint_for_ptr_offset): Likewise.
13063         * tree-ssa-uninit.c (is_pred_expr_subset_of): Likewise.
13064         * tree-ssanames.c (set_nonzero_bits, get_nonzero_bits): Likewise.
13065         * tree-switch-conversion.c (collect_switch_conv_info, array_value_type)
13066         (dump_case_nodes, try_switch_expansion): Likewise.
13067         * tree-vect-loop-manip.c (vect_gen_vector_loop_niters): Likewise.
13068         (vect_do_peeling): Likewise.
13069         * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
13070         * tree-vect-stmts.c (vectorizable_load): Likewise.
13071         * tree-vrp.c (compare_values_warnv, vrp_int_const_binop): Likewise.
13072         (zero_nonzero_bits_from_vr, ranges_from_anti_range): Likewise.
13073         (extract_range_from_binary_expr_1, adjust_range_with_scev): Likewise.
13074         (overflow_comparison_p_1, register_edge_assert_for_2): Likewise.
13075         (is_masked_range_test, find_switch_asserts, maybe_set_nonzero_bits)
13076         (vrp_evaluate_conditional_warnv_with_ops, intersect_ranges): Likewise.
13077         (range_fits_type_p, two_valued_val_range_p, vrp_finalize): Likewise.
13078         (evrp_dom_walker::before_dom_children): Likewise.
13079         * tree.c (cache_integer_cst, real_value_from_int_cst, integer_zerop)
13080         (integer_all_onesp, integer_pow2p, integer_nonzerop, tree_log2)
13081         (tree_floor_log2, tree_ctz, mem_ref_offset, tree_int_cst_sign_bit)
13082         (tree_int_cst_sgn, get_unwidened, int_fits_type_p): Likewise.
13083         (get_type_static_bounds, num_ending_zeros, drop_tree_overflow)
13084         (get_range_pos_neg): Likewise.
13085         * ubsan.c (ubsan_expand_ptr_ifn): Likewise.
13086         * config/darwin.c (darwin_mergeable_constant_section): Likewise.
13087         * config/aarch64/aarch64.c (aapcs_vfp_sub_candidate): Likewise.
13088         * config/arm/arm.c (aapcs_vfp_sub_candidate): Likewise.
13089         * config/avr/avr.c (avr_fold_builtin): Likewise.
13090         * config/bfin/bfin.c (bfin_local_alignment): Likewise.
13091         * config/msp430/msp430.c (msp430_attr): Likewise.
13092         * config/nds32/nds32.c (nds32_insert_attributes): Likewise.
13093         * config/powerpcspe/powerpcspe-c.c
13094         (altivec_resolve_overloaded_builtin): Likewise.
13095         * config/powerpcspe/powerpcspe.c (rs6000_aggregate_candidate)
13096         (rs6000_expand_ternop_builtin): Likewise.
13097         * config/rs6000/rs6000-c.c
13098         (altivec_resolve_overloaded_builtin): Likewise.
13099         * config/rs6000/rs6000.c (rs6000_aggregate_candidate): Likewise.
13100         (rs6000_expand_ternop_builtin): Likewise.
13101         * config/s390/s390.c (s390_handle_hotpatch_attribute): Likewise.
13103 2017-10-10  Bin Cheng  <bin.cheng@arm.com>
13105         * tree-vect-loop-manip.c (rename_variables_in_bb): Rename PHI nodes
13106         when copying loop nest with only one inner loop.
13108 2017-10-10  Richard Biener  <rguenther@suse.de>
13110         * tree-cfgcleanup.c (cleanup_tree_cfg_noloop): Avoid compacting
13111         blocks if SCEV is active.
13112         * tree-scalar-evolution.c (analyze_scalar_evolution_1): Remove
13113         dead code.
13114         (analyze_scalar_evolution): Handle cached evolutions the obvious way.
13115         (scev_initialize): Assert we are not yet initialized.
13117 2017-10-10  Bin Cheng  <bin.cheng@arm.com>
13119         * tree-loop-distribution.c (generate_loops_for_partition): Remove
13120         inner loop's exit stmt by making it always exit the loop, otherwise
13121         we would generate an infinite empty loop.
13123 2017-10-10  Bin Cheng  <bin.cheng@arm.com>
13125         * tree-vect-loop-manip.c (slpeel_tree_duplicate_loop_to_edge_cfg): Skip
13126         renaming variables in new preheader if it's deleted.
13128 2017-10-10  Bin Cheng  <bin.cheng@arm.com>
13130         * tree-loop-distribution.c (struct partition): Remove unused field
13131         loops of the structure.
13132         (partition_alloc, partition_free): Ditto.
13133         (build_rdg_partition_for_vertex): Ditto.
13135 2017-10-09  Jeff Law  <law@redhat.com>
13137         * targhooks.c (default_stack_clash_protection_final_dynamic_probe): Fix
13138         return type to match prototype and documentation.
13140 2010-10-09  Segher Boessenkool  <segher@kernel.crashing.org>
13142         * config/rs6000/rs6000.c (processor_costs): Move to ...
13143         * config/rs6000/rs6000.h: ... here.
13144         (rs6000_cost): Declare.
13146 2017-10-09  Eric Botcazou  <ebotcazou@adacore.com>
13148         * except.c (setjmp_fn): New global variable.
13149         (init_eh): Initialize it if DONT_USE_BUILTIN_SETJMP is defined.
13150         (sjlj_emit_function_enter): Call it instead of BUILTIN_SETJMP
13151         if DONT_USE_BUILTIN_SETJMP is defined.
13153 2017-10-09  Segher Boessenkool  <segher@kernel.crashing.org>
13155         * target.def (insn_cost): New hook.
13156         * doc/tm.texi.in (TARGET_INSN_COST): New hook.
13157         * doc/tm.texi: Regenerate.
13158         * rtlanal.c (insn_cost): Use the new hook.
13160 2017-10-09  Segher Boessenkool  <segher@kernel.crashing.org>
13162         * combine.c (combine_validate_cost): Compute the new insn_cost,
13163         not just pattern_cost.
13164         (try_combine): Adjust comment.
13166 2017-10-09  Segher Boessenkool  <segher@kernel.crashing.org>
13168         * cfgrtl.c (rtl_account_profile_record): Replace insn_rtx_cost with
13169         insn_cost.
13170         * combine.c (uid_insn_cost): Adjust comment.
13171         (combine_validate_cost): Adjust comment.  Use pattern_cost instead
13172         of insn_rtx_cost
13173         (combine_instructions): Use insn_cost instead of insn_rtx_cost.
13174         * dse.c (find_shift_sequence): Ditto.
13175         * ifcvt.c (cheap_bb_rtx_cost_p): Ditto.
13176         (bb_valid_for_noce_process_p): Use pattern_cost.
13177         * rtl.h (insn_rtx_cost): Delete.
13178         (pattern_cost): New prototype.
13179         (insn_cost): New prototype.
13180         * rtlanal.c (insn_rtx_cost): Rename to...
13181         (pattern_cost): ... this.
13182         (insn_cost): New.
13184 2017-10-09  Uros Bizjak  <ubizjak@gmail.com>
13186         * config/i386/i386.md (*jcc_2): Remove insn pattern.
13187         (*jcc<mode>_0_r_i387): Ditto.
13188         (*jccxf_r_i387): Ditto.
13189         (*jcc<mode>_r_i387): Ditto.
13190         (*jccu<mode>_r_i387): Ditto.
13191         (*jcc<X87MODEF:mode>_<SWI24:mode>_r_i387): Ditto.
13192         (*jcc): Rename from *jcc_1.
13194 2017-10-09  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
13196         * config/rs6000/rs6000-p8swap.c (rs6000_analyze_swaps): Process
13197         deferred rescans after the lvx/stvx recombination pre-pass.
13199 2017-10-09  Michael Meissner  <meissner@linux.vnet.ibm.com>
13201         * config/rs6000/amo.h: New include file to provide ISA 3.0 atomic
13202         memory operation instruction support.
13203         * config.gcc (powerpc*-*-*): Include amo.h as an extra header.
13204         (rs6000-ibm-aix[789]*): Likewise.
13205         * doc/extend.texi (PowerPC Atomic Memory Operation Functions):
13206         Document new functions.
13208 2017-10-09  Richard Biener  <rguenther@suse.de>
13210         PR tree-optimization/82397
13211         * tree-data-ref.c (data_ref_compare_tree): Make sure to return
13212         equality only for semantically equal trees.
13214 2017-10-09  Richard Biener  <rguenther@suse.de>
13216         PR tree-optimization/82449
13217         * sese.c (scev_analyzable_p): Check whether the SCEV is linear.
13218         * tree-chrec.h (evolution_function_is_constant_p): Adjust to
13219         allow constant addresses.
13220         * tree-chrec.c (scev_is_linear_expression): Constant evolutions
13221         are linear.
13223 2017-10-09  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
13225         * config/s390/s390-builtins.def (vec_nabs, vec_vfi): Fix builtin
13226         flags.
13228 2017-10-09  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
13230         PR target/82463
13231         * config/s390/vecintrin.h (vec_madd, vec_msub): Fix macro
13232         definitions.
13234 2017-10-09  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
13236         PR target/82465
13237         * config/s390/s390-builtins.def (vec_sqrt): Fix builtin flags.
13239 2017-10-09  Jakub Jelinek  <jakub@redhat.com>
13241         PR target/82464
13242         * config/s390/s390-builtins.def (s390_vec_xor_flt_a,
13243         s390_vec_xor_flt_b, s390_vec_xor_flt_c): New.
13245 2017-10-09  Richard Sandiford  <richard.sandiford@linaro.org>
13247         * wide-int.h (WI_BINARY_OPERATOR_RESULT): New macro.
13248         (WI_BINARY_PREDICATE_RESULT): Likewise.
13249         (wi::binary_traits::operator_result): New type.
13250         (wi::binary_traits::predicate_result): Likewise.
13251         (generic_wide_int::operator~, unary generic_wide_int::operator-)
13252         (generic_wide_int::operator==, generic_wide_int::operator!=)
13253         (generic_wide_int::operator&, generic_wide_int::and_not)
13254         (generic_wide_int::operator|, generic_wide_int::or_not)
13255         (generic_wide_int::operator^, generic_wide_int::operator+
13256         (binary generic_wide_int::operator-, generic_wide_int::operator*):
13257         Delete.
13258         (operator~, unary operator-, operator==, operator!=, operator&)
13259         (operator|, operator^, operator+, binary operator-, operator*): New
13260         functions.
13261         * expr.c (get_inner_reference): Use wi::bit_and_not.
13262         * fold-const.c (fold_binary_loc): Likewise.
13263         * ipa-prop.c (ipa_compute_jump_functions_for_edge): Likewise.
13264         * tree-ssa-ccp.c (get_value_from_alignment): Likewise.
13265         (bit_value_binop): Likewise.
13266         * tree-ssa-math-opts.c (find_bswap_or_nop_load): Likewise.
13267         * tree-vrp.c (zero_nonzero_bits_from_vr): Likewise.
13268         (extract_range_from_binary_expr_1): Likewise.
13269         (masked_increment): Likewise.
13270         (simplify_bit_ops_using_ranges): Likewise.
13272 2017-10-09  Martin Jambor  <mjambor@suse.cz>
13274         PR hsa/82416
13275         * hsa-common.h (hsa_op_with_type): New method extend_int_to_32bit.
13276         * hsa-gen.c (hsa_extend_inttype_to_32bit): New function.
13277         (hsa_type_for_scalar_tree_type): Use it.  Always force min32int for
13278         COMPLEX types.
13279         (hsa_fixup_mov_insn_type): New function.
13280         (hsa_op_with_type::get_in_type): Use it.
13281         (hsa_build_append_simple_mov): Likewise.  Allow sub-32bit
13282         immediates in an assert.
13283         (hsa_op_with_type::extend_int_to_32bit): New method.
13284         (gen_hsa_insns_for_bitfield): Fixup instruction and intermediary
13285         types.  Convert to dest type if necessary.
13286         (gen_hsa_insns_for_bitfield_load): Fixup load type if necessary.
13287         (reg_for_gimple_ssa): Pass false as min32int to
13288         hsa_type_for_scalar_tree_type.
13289         (gen_hsa_addr): Fixup type when creating addresable temporary.
13290         (gen_hsa_cmp_insn_from_gimple): Extend operands if necessary.
13291         (gen_hsa_unary_operation): Extend operands and convert to dest type if
13292         necessary.  Call hsa_fixup_mov_insn_type.
13293         (gen_hsa_binary_operation): Changed operand types to hsa_op_with_type,
13294         extend operands and convert to dest type if necessary.
13295         (gen_hsa_insns_for_operation_assignment): Extend operands and convert
13296         to dest type if necessary.
13297         (set_output_in_type): Call hsa_fixup_mov_insn_type.  Just ude dest
13298         if conversion nt necessary and size matches.
13299         (gen_hsa_insns_for_load): Call hsa_fixup_mov_insn_type, convert
13300         to dest type if necessary.
13301         (gen_hsa_insns_for_store): Call hsa_fixup_mov_insn_type.
13302         (gen_hsa_insns_for_switch_stmt): Likewise. Also extend operands if
13303         necessary.
13304         (gen_hsa_clrsb): Likewise.
13305         (gen_hsa_ffs): Likewise.
13306         (gen_hsa_divmod): Extend operands and convert to dest type if
13307         necessary.
13308         (gen_hsa_atomic_for_builtin): Change type of op to hsa_op_with_type.
13310 2017-10-08  Segher Boessenkool  <segher@kernel.crashing.org>
13312         * config/rs6000/rs6000.md (conditional branch): Clean up formatting.
13313         Remove empty default arguments.  Use a brace block as output
13314         statement.
13315         (conditional return): Ditto.
13316         (jump): Ditto.
13317         (indirect_jump): Ditto.  Use b%T0 instead of bctr/blr.
13318         (tablejump, tablejumpsi, tablejumpdi, *tablejump<mode>_internal1):
13319         Ditto.
13320         (group_ending_nop): Ditto.
13321         (doloop_end): Ditto.
13322         (ctr<mode>, ctr<mode>_internal1, ctr<mode>_internal2): Ditto.
13323         (splitters for those): Ditto.
13325 2017-10-08  Segher Boessenkool  <segher@kernel.crashing.org>
13327         * config/rs6000/rs6000-string.c (expand_strncmp_align_check): Invert
13328         a conditional jump (and the compare for it) so that pc_rtx is the
13329         last operand.
13330         * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Adjust
13331         for the deleted and renamed ctr<mode>_internal[234] patterns.
13332         * config/rs6000/rs6000.md: Delete second conditional branch pattern.
13333         Delete second conditional return pattern.
13334         (ctr<mode>_internal2): Delete this second bdnz pattern.
13335         (ctr<mode>_internal3): Rename to ctr<mode>_internal2.
13336         (ctr<mode>_internal4): Delete this second bdz pattern.
13338 2017-10-08  Eric Botcazou  <ebotcazou@adacore.com>
13340         * tree-outof-ssa.h (ssaexpand): Add partitions_for_undefined_values.
13341         (always_initialized_rtx_for_ssa_name_p): New predicate.
13342         * tree-outof-ssa.c (remove_ssa_form): Initialize new field of SA.
13343         (finish_out_of_ssa): Free new field of SA.
13344         * tree-ssa-coalesce.h (get_undefined_value_partitions): Declare.
13345         * tree-ssa-coalesce.c: Include tree-ssa.h.
13346         (get_parm_default_def_partitions): Remove extern keyword.
13347         (get_undefined_value_partitions): New function.
13348         * expr.c (expand_expr_real_1) <expand_decl_rtl>: For a SSA_NAME, do
13349         not set SUBREG_PROMOTED_VAR_P on the sub-register if it may contain
13350         uninitialized bits.
13351         * loop-iv.c (iv_get_reaching_def): Disqualify all subregs.
13353 2017-10-08  Eric Botcazou  <ebotcazou@adacore.com>
13355         * builtins.def (BUILT_IN_SETJMP): Revert latest change.
13357 2017-10-08  Jan Hubicka  <hubicka@ucw.cz>
13359         * config/i386/i386.c (ix86_expand_set_or_movmem): Disable 512bit loops
13360         for targets that preffer 128bit.
13362 2017-10-08  Jan Hubicka  <hubicka@ucw.cz>
13364         * config/i386/i386.c (has_dispatch): Disable for Ryzen.
13366 2017-10-08  Olivier Hainque  <hainque@adacore.com>
13368         * config/arm/arm.c (arm_set_return_address): Use MEM_VOLATILE_P
13369         on the target mem instead of RTX_FRAME_RELATED_P on the insn to
13370         prevent DSE.
13371         (thumb_set_return_address): Likewise.
13373 2017-10-08  Olivier Hainque  <hainque@adacore.com>
13375         * common/config/arm/arm-common.c (arm_except_unwind_info):
13376         Handle DWARF2_UNWIND_INFO.
13378 2017-10-07  Michael Collison <michael.collison@arm.com>
13380         * config/aarch64/aarch64.md (*aarch64_reg_<optab>_minus<mode>3):
13381         New pattern.
13383 2017-10-07  Eric Botcazou  <ebotcazou@adacore.com>
13385         * builtins.def (BUILT_IN_SETJMP): Declare as library builtin instead
13386         of GCC builtin if DONT_USE_BUILTIN_SETJMP is defined.
13387         * except.c (sjlj_emit_function_enter): If DONT_USE_BUILTIN_SETJMP is
13388         defined, force the creation of a new block for a dispatch label.
13390 2017-10-07  Jan Hubicka  <hubicka@ucw.cz>
13392         * invoke.texi (Wsuggest-attribute=cold): Document.
13393         * common.opt (Wsuggest-attribute=cold): New
13394         * ipa-pure-const.c (warn_function_cold): New function.
13395         * predict.c (compute_function_frequency): Use it.
13396         * predict.h (warn_function_cold): Declare.
13398 2017-10-06  Jan Hubicka  <hubicka@ucw.cz>
13400         * tree-switch-conversion.c (do_jump_if_equal, emit_cmp_and_jump_insns):
13401         Update profile.
13403 2017-10-06  Martin Liska  <mliska@suse.cz>
13405         * sanopt.c (struct sanopt_tree_triplet_hash): Remove inline
13406         keyword for member functions.
13407         (struct sanopt_tree_couple): New struct.
13408         (struct sanopt_tree_couple_hash): New function.
13409         (struct sanopt_ctx): Add new hash_map.
13410         (has_dominating_ubsan_ptr_check): New function.
13411         (record_ubsan_ptr_check_stmt): Likewise.
13412         (maybe_optimize_ubsan_ptr_ifn): Likewise.
13413         (sanopt_optimize_walker): Handle IFN_UBSAN_PTR.
13414         (pass_sanopt::execute): Handle also SANITIZE_POINTER_OVERFLOW.
13416 2017-10-06  Sudakshina Das  <sudi.das@arm.com>
13418         PR target/82440
13419         * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Only call
13420         aarch64_simd_valid_immediate on CONST_VECTORs.
13421         (aarch64_reg_or_bic_imm): Likewise.
13423 2017-10-06  Wilco Dijkstra  <wdijkstr@arm.com>
13425         PR rtl-optimization/82396
13426         * haifa-sched.c (ready_sort_real): Disable qsort checking.
13428 2017-10-06  Sebastian Pop  <sebpop@gmail.com>
13430         * graphite-dependences.c (scop_get_reads): Move code to...
13431         (scop_get_must_writes): Move code to...
13432         (scop_get_may_writes): Move code to...
13433         (scop_get_reads_and_writes): ... here.
13434         (scop_get_dependences): Call scop_get_reads_and_writes.
13436 2017-10-06  Jakub Jelinek  <jakub@redhat.com>
13438         PR tree-optimization/82434
13439         * fold-const.h (can_native_encode_type_p,
13440         can_native_encode_string_p): Remove.
13441         * fold-const.c (native_encode_int): Formatting fixes.  If ptr is NULL,
13442         don't encode anything, just return what would be otherwise returned.
13443         (native_encode_fixed, native_encode_complex, native_encode_vector):
13444         Likewise.
13445         (native_encode_string): Likewise.  Inline by hand
13446         can_native_encode_string_p.
13447         (can_native_encode_type_p): Remove.
13448         (can_native_encode_string_p): Remove.
13449         * tree-vect-stmts.c (vectorizable_store): Instead of testing just
13450         STRING_CSTs using can_native_encode_string_p, test all
13451         CONSTANT_CLASS_P values using native_encode_expr with NULL ptr.
13452         * gimple-ssa-store-merging.c (encode_tree_to_bitpos): Remove last
13453         argument from native_encode_expr.
13454         (rhs_valid_for_store_merging_p): Use native_encode_expr with NULL ptr.
13455         (pass_store_merging::execute): Don't unnecessarily look for 3 stmts,
13456         but just 2.
13458 2017-10-06  Richard Biener  <rguenther@suse.de>
13460         PR tree-optimization/82397
13461         * tree-vect-data-refs.c (dr_group_sort_cmp): Do not use
13462         operand_equal_p but rely on data_ref_compare_tree for detecting
13463         equalities.
13464         (vect_analyze_data_ref_accesses): Use data_ref_compare_tree
13465         to match up with dr_group_sort_cmp.
13467 2017-10-06  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
13469         PR target/82322
13470         * config/s390/s390-builtins.def (s390_vfi): Define new overloaded
13471         builtin.
13472         * config/s390/s390-builtin-types.def: Regenerate.
13474 2017-10-06  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
13476         PR target/82317
13477         * config/s390/s390-builtin-types.def: Regenerate.
13478         * config/s390/s390-builtins.def (s390_vfmaxdb_4, s390_vfmindb_4):
13479         Change flag from B_VXE to B_VX.
13480         (s390_vec_min_dbl): Remove B_VXE flag.
13482 2017-10-06  Richard Biener  <rguenther@suse.de>
13484         * graphite-isl-ast-to-gimple.c: Include ssa.h and tree-ssa.h.
13485         (translate_isl_ast_to_gimple::translate_pending_phi_nodes,
13486         translate_isl_ast_to_gimple::is_valid_rename,
13487         translate_isl_ast_to_gimple::get_rename,
13488         translate_isl_ast_to_gimple::get_def_bb_for_const,
13489         translate_isl_ast_to_gimple::get_new_name,
13490         translate_isl_ast_to_gimple::collect_all_ssa_names,
13491         translate_isl_ast_to_gimple::copy_loop_phi_args,
13492         translate_isl_ast_to_gimple::collect_all_ssa_names,
13493         translate_isl_ast_to_gimple::copy_loop_phi_args,
13494         translate_isl_ast_to_gimple::copy_loop_phi_nodes,
13495         translate_isl_ast_to_gimple::add_close_phis_to_merge_points,
13496         translate_isl_ast_to_gimple::add_close_phis_to_outer_loops,
13497         translate_isl_ast_to_gimple::copy_loop_close_phi_args,
13498         translate_isl_ast_to_gimple::copy_loop_close_phi_nodes,
13499         translate_isl_ast_to_gimple::copy_cond_phi_args,
13500         translate_isl_ast_to_gimple::copy_cond_phi_nodes,
13501         translate_isl_ast_to_gimple::edge_for_new_close_phis,
13502         translate_isl_ast_to_gimple::add_phi_arg_for_new_expr,
13503         translate_isl_ast_to_gimple::rename_uses,
13504         translate_isl_ast_to_gimple::rename_all_uses): Remove.
13505         (translate_isl_ast_to_gimple::get_rename_from_scev): Simplify.
13506         (set_rename_for_each_def): Likewise.
13507         (graphite_copy_stmts_from_block): Handle debug stmt resetting
13508         here.  Handle rewriting SCEV analyzable uses here.
13509         (copy_bb_and_scalar_dependences): Generate code for PHI
13510         copy-in/outs.
13511         (graphite_regenerate_ast_isl): Adjust.
13512         * graphite-scop-detection.c (trivially_empty_bb_p): Move to sese.[ch].
13513         (add_write, add_read): New functions.
13514         (build_cross_bb_scalars_def): Use it and simplify.
13515         (build_cross_bb_scalars_use): Likewise.
13516         (graphite_find_cross_bb_scalar_vars): Inline into...
13517         (try_generate_gimple_bb): ...here.  Add dependences for PHIs,
13518         simulating out-of-SSA.  Compute liveout and add dependencies.
13519         (build_scops): Force an empty entry block.
13520         * sese.h (sese_info_t::liveout, sese_info_t::debug_liveout): New
13521         members.
13522         (sese_build_liveouts): Declare.
13523         (sese_trivially_empty_bb_p): Likewise.
13524         * sese.c (sese_build_liveouts_bb): Properly handle PHIs,
13525         compute liveout and debug_liveout.
13526         (sese_bad_liveouts_use): Remove.
13527         (sese_reset_debug_liveouts_bb): Likewise.
13528         (sese_reset_debug_liveouts): Rewrite in terms of debug_liveout.
13529         (sese_build_liveouts): Build liveout and debug_liveout and store
13530         it in region.
13531         (new_sese_info): Adjust.
13532         (free_sese_info): Likewise.
13533         (sese_insert_phis_for_liveouts): Reset debug stmts from here,
13534         do not build liveout here.
13535         (move_sese_in_condition): Adjust region entry.
13536         (scev_analyzable_p): Match up with chrec_apply requirements.
13537         (sese_trivially_empty_bb_p): New.
13538         * tree-into-ssa.c (get_reaching_def): Properly support generating
13539         default-defs for incremental rewrite of anonymous names.
13541 2017-10-06  Richard Biener  <rguenther@suse.de>
13543         * graphite-sese-to-poly.c (extract_affine): For casts increasing
13544         precision do not perform modulo reduction.
13546 2017-10-06  Richard Biener  <rguenther@suse.de>
13548         PR tree-optimization/82436
13549         * tree-vect-slp.c (vect_supported_load_permutation_p): More
13550         conservatively choose the vectorization factor when checking
13551         whether we can perform the required load permutation.
13552         (vect_transform_slp_perm_load): Assert when we may not fail.
13554 2017-10-05  Segher Boessenkool  <segher@kernel.crashing.org>
13556         * config/rs6000/sysv4.h (SUBTARGET_OVERRIDE_OPTIONS): Correct error
13557         message for incompatible -msdata=* and -mcall-* options.
13559 2017-10-05  Jan Hubicka <hubicka@ucw.cz>
13561         * config/i386/i386.c (ia32_multipass_dfa_lookahead): Default to issue
13562         rate for post-reload scheduling.
13564 2017-10-05  Tamar Christina  <tamar.christina@arm.com>
13566         * doc/sourcebuild.texi (vect_sizes_16B_8B, vect_sizes_32B_16B): New.
13568 2017-10-05  Jan Hubicka <hubicka@ucw.cz>
13570         * config/i386/i386.c (znver1_cost): Set branch_cost to 3 (instead of 2)
13571         to improve monte carlo in scimark.
13573 2017-10-05  Jan Hubicka <hubicka@ucw.cz>
13575         * config/i386/i386.c (ix86_size_cost, i386_cost, i486_cost,
13576         pentium_cost, lakemont_cost, pentiumpro_cost, geode_cost, k6_cost,
13577         athlon_cost, k8_cost, amdfam10_cost, btver1_cost, btver2_cost,
13578         pentium4_cost, nocona_cost): Set reassociation width to 1.
13579         (bdver1_cost, bdver2_cost, bdver3_cost, bdver4_cost): Set reassociation
13580         width to 2 for fp operations and 1 otherwise.
13581         (znver1_cost): Set scalar reassoc width to 4 and vector to 3 and 6
13582         for int and fp.
13583         (atom_cost): Set reassociation width to 2.
13584         (slm_cost, generic_cost): Set fp reassociation width
13585         to 2 and 1 otherwise.
13586         (intel_cost): Set fp reassociation width to 4 and 1 otherwise.
13587         (core_cost): Set fp reassociation width to 4 and vector to 2.
13588         (ix86_reassociation_width): Rewrite using cost table; special case
13589         plus/minus on Zen; honor X86_TUNE_SSE_SPLIT_REGS
13590         and TARGET_AVX128_OPTIMAL.
13591         * config/i386/i386.h (processor_costs): Add
13592         reassoc_int, reassoc_fp, reassoc_vec_int, reassoc_vec_fp.
13593         (TARGET_VECTOR_PARALLEL_EXECUTION, TARGET_REASSOC_INT_TO_PARALLEL,
13594         TARGET_REASSOC_FP_TO_PARALLEL): Remove.
13595         * x86-tune.def (X86_TUNE_REASSOC_INT_TO_PARALLEL): Remove.
13596         (X86_TUNE_REASSOC_FP_TO_PARALLEL): Remove.
13597         (X86_TUNE_VECTOR_PARALLEL_EXECUTION):  Remove.
13599 2017-10-05  Nathan Sidwell  <nathan@acm.org>
13601         * doc/invoke.texi (Wparentheses): Document C++ MVP behaviour.
13603 2017-10-05  Tamar Christina  <tamar.christina@arm.com>
13605         * config/arm/arm.c (arm_test_fpu_data): New.
13606         (arm_run_selftests): Call arm_test_fpu_data.
13608 2017-10-04  Nathan Sidwell  <nathan@acm.org>
13610         * toplev.c (toplev::main): Remove excess parens on pretty_printer
13611         decl.
13612         * caller-save.c (insert_save): Remove excess parens on TO_SAVE parm.
13614 2017-10-04  Sudakshina Das  <sudi.das@arm.com>
13616         * config/aarch64/aarch64-protos.h (enum simd_immediate_check): New
13617         check type for aarch64_simd_valid_immediate.
13618         (aarch64_output_simd_mov_immediate): Update prototype.
13619         (aarch64_simd_valid_immediate): Update prototype.
13620         * config/aarch64/aarch64-simd.md (orr<mode>3): modified pattern to add
13621         support for ORR-immediate.
13622         (and<mode>3): modified pattern to add support for BIC-immediate.
13623         * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Function
13624         now checks for valid immediate for BIC and ORR based on new enum
13625         argument.
13626         (aarch64_output_simd_mov_immediate): Function now used to output
13627         BIC/ORR imm as well based on new enum argument.
13628         * config/aarch64/constraints.md (Do): New vector immediate constraint.
13629         (Db) : Likewise.
13630         * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): New predicate.
13631         (aarch64_reg_or_bic_imm): Likewise.
13633 2017-10-04  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
13635         * config/s390/vx-builtins.md ("vec_mergeh<mode>")
13636         ("vec_mergel<mode>"): Change mode iterator to V_128_NOSINGLE.
13638 2017-10-04  Wilco Dijkstra  <wdijkstr@arm.com>
13640         Revert r253399:
13642         PR rtl-optimization/82396
13643         * haifa-sched.c (autopref_multipass_init): Simplify
13644         initialization.
13645         (autopref_rank_data): Simplify sort order.
13646         * sched-int.h (autopref_multipass_data_): Remove
13647         multi_mem_insn_p, min_offset and max_offset.
13649 2017-10-04  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
13651         * doc/sourcebuild.texi: Document vect_peeling_profitable.
13653 2017-10-04  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
13655         * doc/sourcebuild.texi: Document vect_intdouble_cvt and
13656         vect_doubleint_cvt.
13658 2017-10-04  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
13660         * doc/sourcebuild.texi: Document vect_long_mult.
13662 2017-10-04  Richard Sandiford  <richard.sandiford@linaro.org>
13664         PR tree-optimization/82413
13665         * fold-const.c (build_range_check): Use widest_int when comparing
13666         the maximum ETYPE value with HIGH.
13668 2017-10-04  Wilco Dijkstra  <wdijkstr@arm.com>
13670         PR rtl-optimization/82396
13671         * haifa-sched.c (autopref_multipass_init): Simplify
13672         initialization.
13673         (autopref_rank_data): Simplify sort order.
13674         * sched-int.h (autopref_multipass_data_): Remove
13675         multi_mem_insn_p, min_offset and max_offset.
13677 2017-10-04  Jakub Jelinek  <jakub@redhat.com>
13679         PR tree-optimization/82381
13680         * tree-ssa-reassoc.c (sort_by_operand_rank): Check for different
13681         oeN->rank first.  Return 1 or -1 if one op is SSA_NAME and the other
13682         is not.
13684         PR tree-optimization/82374
13685         * omp-low.c (create_omp_child_function): Copy DECL_ATTRIBUTES,
13686         DECL_FUNCTION_SPECIFIC_OPTIMIZATION,
13687         DECL_FUNCTION_SPECIFIC_TARGET and DECL_FUNCTION_VERSIONED from
13688         current_function_decl to the new decl.
13690 2017-10-03  Michael Meissner  <meissner@linux.vnet.ibm.com>
13692         * config/rs6000/rs6000-builtin.def (BU_FLOAT128_2_HW): Define new
13693         helper macro for IEEE float128 hardware built-in functions.
13694         (SQRTF128_ODD): Add built-in functions with the round-to-odd
13695         semantics.
13696         (TRUNCF128_ODD): Likewise.
13697         (ADDF128_ODD): Likewise.
13698         (SUBF128_ODD): Likewise.
13699         (MULF128_ODD): Likewise.
13700         (DIVF128_ODD): Likewise.
13701         (FMAF128_ODD): Likewise.
13702         * config/rs6000/rs6000.md (UNSPEC_ROUND_TO_ODD): Rename to
13703         UNSPEC_TRUNC_ROUND_TO_ODD.
13704         (UNSPEC_TRUNC_ROUND_TO_ODD): Likewise.
13705         (UNSPEC_ADD_ROUND_TO_ODD): New unspec codes for the IEEE 128-bit
13706         floating point round to odd instructions.
13707         (UNSPEC_SUB_ROUND_TO_ODD): Likewise.
13708         (UNSPEC_MUL_ROUND_TO_ODD): Likewise.
13709         (UNSPEC_DIV_ROUND_TO_ODD): Likewise.
13710         (UNSPEC_FMA_ROUND_TO_ODD): Likewise.
13711         (UNSPEC_SQRT_ROUND_TO_ODD): Likewise.
13712         (trunc<mode>sf2_hw): Change the truncate with round to odd
13713         expansion to use UNSPEC_TRUNC_ROUND_TO_ODD.
13714         (add<mode>3_odd): Add insns for IEEE 128-bit floating point round
13715         to odd hardware instructions.
13716         (sub<mode>3_odd): Likewise.
13717         (mul<mode>3_odd): Likewise.
13718         (div<mode>3_odd): Likewise.
13719         (sqrt<mode>2_odd): Likewise.
13720         (fma<mode>4_odd): Likewise.
13721         (fms<mode>4_odd): Likewise.
13722         (nfma<mode>4_odd): Likewise.
13723         (nfms<mode>4_odd): Likewise.
13724         (trunc<mode>df2_odd): Change the truncate with round to odd
13725         expansion to use UNSPEC_TRUNC_ROUND_TO_ODD.  Add a generator
13726         function.
13727         * doc/extend.texi (PowerPC built-in functions): Update documentation
13728         for existing IEEE float128-bit built-in functions.  Add built-in
13729         functions that generate the IEEE 128-bit floating point round to
13730         odd instructions.
13732 2017-10-03  Segher Boessenkool  <segher@kernel.crashing.org>
13734         PR rtl-optimization/77729
13735         * simplify-rtx.c (simplify_binary_operation_1): Delete the (X&C1)|C2
13736         to (X&(C1&~C2))|C2 transformations.
13738 2017-10-03  Martin Jambor  <mjambor@suse.cz>
13740         PR tree-optimization/82363
13741         * tree-sra.c (propagate_subaccesses_across_link): In unrecoverable
13742         mismatch, mark lacc written regardless of racc.
13744 2017-10-03  Jakub Jelinek  <jakub@redhat.com>
13746         PR tree-optimization/82381
13747         * tree-ssa-reassoc.c (sort_by_operand_rank): Don't check
13748         stmt_to_insert nor wheather SSA_NAMEs are default defs.
13749         Return 1 or -1 if one of bba and bbb is NULL. If bb_rank is equal,
13750         fallthrough into reassoc_stmt_dominates_stmt_p.
13752         PR target/82386
13753         * combine.c (combine_instructions): Don't combine in unreachable
13754         basic blocks.
13756 2017-10-02  Peter Bergner  <bergner@vnet.ibm.com>
13758         PR target/80210
13759         * config/rs6000/rs6000.c (rs6000_option_override_internal): Rewrite
13760         function to not use the have_cpu variable.  Do not set cpu_index,
13761         rs6000_cpu_index or rs6000_tune_index if we end up using TARGET_DEFAULT
13762         or the default cpu.
13763         (rs6000_valid_attribute_p): Remove duplicate initializations of
13764         old_optimize and func_optimize.
13765         (rs6000_pragma_target_parse): Call rs6000_activate_target_options ().
13766         (rs6000_activate_target_options): Make global.
13767         * config/rs6000/rs6000-protos.h (rs6000_activate_target_options): Add
13768         prototype.
13770 2017-10-02  Jakub Jelinek  <jakub@redhat.com>
13772         * tree-dfa.c (get_ref_base_and_extent): Set *pmax_size to -1
13773         if *poffset + *pmax_size overflows in HOST_WIDE_INT.
13774         Set *poffset to 0 and *psize and *pmax_size to -1 if
13775         *poffset + *psize overflows in HOST_WIDE_INT.
13777         PR tree-optimization/82387
13778         PR tree-optimization/82388
13779         PR tree-optimization/82389
13780         * tree-ssa-dse.c (dse_classify_store): Test byte_tracking_enabled
13781         instead of live_bytes non-NULL.
13783 2017-10-02  Georg-Johann Lay  <avr@gjlay.de>
13785         PR target/41076
13786         * confg/avr/avr.md (*iorhi3.ashift8-ext.zerox): Add "r,r,0"
13787         alternative.
13789 2017-10-02  Richard Biener  <rguenther@suse.de>
13791         * graphite-isl-ast-to-gimple.c (set_codegen_error): With
13792         -fchecking and --param graphite-allow-codegen-errors=0 ICE.
13793         * params.def (PARAM_GRAPHITE_ALLOW_CODEGEN_ERRORS): New param.
13795 2017-10-02  Richard Sandiford  <richard.sandiford@linaro.org>
13797         * tree.h (wi::int_traits <const_tree>::decompose): Assert that the
13798         requested precision matches the type's.
13799         * calls.c (alloc_max_size): Calculate the new candidate size as
13800         a widest_int and use wi::to_widest when comparing it with the
13801         current candidate size.
13802         * gimple-ssa-warn-alloca.c (pass_walloca::execute): Compare with
13803         zero rather than integer_zero_node.
13804         * match.pd: Check for a no-op conversion before using wi::add
13805         rather than after.  Use tree_to_uhwi when summing small shift
13806         counts into an unsigned int.
13808 2017-10-02  Richard Sandiford  <richard.sandiford@linaro.org>
13809             Alan Hayward  <alan.hayward@arm.com>
13810             David Sherwood  <david.sherwood@arm.com>
13812         PR target/71307
13813         * config/aarch64/aarch64.h (POINTER_AND_FP_REGS): New reg class.
13814         (REG_CLASS_NAMES, REG_CLASS_CONTENTS): Update accordingly.
13815         * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
13816         POINTER_AND_FP_REGS.
13818 2017-10-02  Richard Biener  <rguenther@suse.de>
13820         PR tree-optimization/82355
13821         * graphite-isl-ast-to-gimple.c (build_iv_mapping): Also build
13822         a mapping for the enclosing loop but avoid generating one for
13823         the loop tree root.
13824         (copy_bb_and_scalar_dependences): Remove premature codegen
13825         error on PHIs in blocks duplicated into multiple places.
13826         * graphite-scop-detection.c
13827         (scop_detection::stmt_has_simple_data_refs_p): For a loop not
13828         in the region use it as loop and nest to analyze the DR in.
13829         (try_generate_gimple_bb): Likewise.
13830         * graphite-sese-to-poly.c (extract_affine_chrec): Adjust.
13831         (add_loop_constraints): For blocks in a loop not in the region
13832         create a dimension with a single iteration.
13833         * sese.h (gbb_loop_at_index): Remove assert.
13835 2017-10-01  Kevin Buettner  <kevinb@redhat.com>
13837         * omp-expand.c (adjust_context_scope): New function.
13838         (expand_parallel_call): Call adjust_context_scope.
13840 2017-10-01  Jeff Law  <law@redhat.com>
13842         * tree-ssa-dom.c (optimize_stmt): Make this a method within the
13843         dom_opt_dom_walker class with direct access to private members.
13844         Add comments.  Call test_for_singularity.
13845         (dom_opt_dom_walker::before_dom_children): Corresponding changes.
13846         (dom_opt_dom_walker::after_dom_children): Do not lazily initialize
13847         m_dummy_cond anymore.
13848         (class dom_opt_dom_walker): Initialize m_dummy_cond member in the
13849         class ctor.
13850         (pass_dominator:execute): Build the dummy_cond here and pass it
13851         to the dom_opt_dom_walker ctor.
13852         (test_for_singularity): New function.
13854 2017-09-30  Krister Walfridsson  <krister.walfridsson@gmail.com>
13855             Maya Rashish  <coypu@sdf.org>
13857         * config.gcc (*-*-netbsd*): New variable nbsd_tm_file containing
13858         netbsd.h, netbsd-stdint.h, and netbsd-elf.h.
13859         (alpha*-*-netbsd*) Use nbsd_tm_file.
13860         (arm*-*-netbsdelf*) Likewise.
13861         (i[34567]86-*-netbsdelf*) Likewise.
13862         (x86_64-*-netbsd*) Likewise.
13863         (mips*-*-netbsd*) Likewise.
13864         (powerpc-*-netbsd*) Likewise.
13865         (sh*-*-netbsd*) Likewise.
13866         (sparc-*-netbsdelf*) Likewise.
13867         (sparc64-*-netbsd*) Likewise.
13868         (m68k*-*-netbsdelf*) Use nbsd_tm_file and add CHAR_FAST8/SHORT_FAST16
13869         to tm_defines.
13870         (vax-*-netbsdelf*) Likewise.
13871         * config/netbsd-stdint.h (INT_FAST8_TYPE): Check CHAR_FAST8.
13872         (UINT_FAST8_TYPE) Likewise.
13873         (INT_FAST16_TYPE) Check CHAR_FAST16.
13874         (UINT_FAST16_TYPE) Likewise.
13876 2017-09-30  Jakub Jelinek  <jakub@redhat.com>
13878         PR target/82361
13879         * config/i386/i386.md
13880         (TARGET_USE_8BIT_IDIV zext divmodsi4 splitter): New define_split.
13881         (divmodsi4_zext_1, divmodsi4_zext_2, *divmodsi4_zext_1,
13882         *divmodsi4_zext_2): New define_insn_and_split.
13883         (*divmodsi4_noext_zext_1, *divmodsi4_noext_zext_2): New define_insn.
13884         (TARGET_USE_8BIT_IDIV zext udivmodsi4 splitter): New define_split.
13885         (udivmodsi4_zext_1, udivmodsi4_zext_2, *udivmodsi4_zext_1,
13886         *udivmodsi4_zext_2, *udivmodsi4_pow2_zext_1, *udivmodsi4_pow2_zext_2):
13887         New define_insn_and_split.
13888         (*udivmodsi4_noext_zext_1, *udivmodsi4_noext_zext_2): New define_insn.
13889         * config/i386/i386.c (ix86_split_idivmod): Handle operands[0] or
13890         operands[1] having DImode when mode is SImode.
13892         * config/i386/i386.c (ix86_split_idivmod): Use mode instead of
13893         always SImode for DIV and MOD in REG_EQUAL notes.
13895 2017-09-29  Yury Gribov  <tetra2005@gmail.com>
13897         PR middle-end/82319
13898         * match.pd: Fix handling of NaNs in pattern.
13900 2017-09-29  Jeff Law  <law@redhat.com>
13902         * sbitmap.c (bitmap_bit_in_range_p): New function.
13903         * sbitmap.h (bitmap_bit_in_range_p): Prototype.
13904         * tree-ssa-dse.c (live_bytes_read): New function.
13905         (dse_classify_store): Ignore reads of dead bytes.
13907         * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Fix
13908         typos and whitespace errors.
13909         * config/i386/predicates.md (address_no_seg_operand): Likewise.
13910         * config/s390/s390.c (s390_emit_prologue): Likewise.
13912 2017-09-29  Vladimir Makarov  <vmakarov@redhat.com>
13914         PR target/81481
13915         * ira-costs.c (scan_one_insn): Don't take into account PIC equiv
13916         with a symbol for LRA.
13918 2017-09-29  Vladimir Makarov  <vmakarov@redhat.com>
13920         PR rtl-optimization/82338
13921         * lra-constraints.c (inherit_in_ebb): Check usage_insns check.
13923 2017-09-29  Alexander Monakov  <amonakov@ispras.ru>
13925         * genmodes.c (calc_wider_mode): Suppress qsort macro.
13926         * system.h [CHECKING_P] (qsort): Redirect to qsort_chk.
13927         (qsort_chk): Declare.
13928         * vec.c [CHECKING_P] (qsort_chk_error): New static function.
13929         (qsort_chk): New function.
13931 2017-09-29  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
13933         PR tree-optimization/82337
13934         * gimple-ssa-strength-reduction.c (find_phi_def): Don't record a
13935         phi definition if the PHI result appears in an abnormal PHI.
13936         (find_basis_for_base_expr): Don't record a basis if the LHS of the
13937         basis appears in an abnormal PHI.
13939 2017-09-29  Richard Biener  <rguenther@suse.de>
13941         * graphite-isl-ast-to-gimple.c
13942         (translate_isl_ast_to_gimple::set_codegen_error): New function.
13943         (binary_op_to_tree): Use it.
13944         (get_rename_from_scev): Likewise.
13945         (copy_loop_phi_nodes): Likewise.
13946         (copy_bb_and_scalar_dependences): Likewise.
13947         (translate_pending_phi_nodes): Likewise.
13949 2017-09-29  Jakub Jelinek  <jakub@redhat.com>
13951         PR target/82339
13952         * config/i386/i386.md (*movdi_internal peephole2): New -Os peephole
13953         for movabsq $(i32 << shift), r64.
13955 2017-09-28  Uros Bizjak  <ubizjak@gmail.com>
13957         * config/i386/i386.c (ix86_print_operand_address_as): Do not check
13958         index when encoding %esp as %rsp to avoid 0x67 prefix.
13960 2017-09-28  Sergey Shalnov  <Sergey.Shalnov@intel.com>
13962         * config/i386/i386.md (*movsf_internal, *movdf_internal):
13963         Return 256-bit AVX modes for TARGET_PREFER_AVX256.
13965 2017-09-28  Thomas Preud'homme  <thomas.preudhomme@arm.com>
13967         * config/arm/arm.c (arm_option_override): Forbid ARMv8-M Security
13968         Extensions with more than 16 double VFP registers.
13969         (cmse_nonsecure_entry_clear_before_return): Remove second entry of
13970         to_clear_mask and all code related to it.  Replace the remaining
13971         entry by a sbitmap and adapt code accordingly.
13973 2017-09-28  Henry Linjamäki  <henry.linjamaki@parmance.com>
13975         * brig-builtins.def: Change pure attributes to const.
13977 2017-09-28  Joseph Myers  <joseph@codesourcery.com>
13979         * config.gcc (default_gnu_indirect_function): Default to yes for
13980         sparc*-*-linux* with glibc.
13982 2017-09-28  Joseph Myers  <joseph@codesourcery.com>
13984         * config/aarch64/aarch64.c (aarch64_elf_asm_constructor)
13985         (aarch64_elf_asm_destructor): Pass SECTION_NOTYPE to get_section
13986         when creating .init_array and .fini_array sections with priority
13987         specified.
13989 2017-09-27  Christophe Lyon  <christophe.lyon@linaro.org>
13991         PR target/71727
13992         * config/aarch64/aarch64.c
13993         (aarch64_builtin_support_vector_misalignment): Always return false
13994         when misalignment is unknown.
13996 2017-09-27  Kelvin Nilsen  <kelvin@gcc.gnu.org>
13998         * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Revise
13999         this function to return false if the definition used by the swap
14000         instruction is artificial, or if the memory address from which the
14001         constant value is loaded is not represented by a base address held
14002         in a register or if the base address register is a frame or stack
14003         pointer.  Additionally, return false if the base address of the
14004         loaded constant is a SYMBOL_REF but is not considered to be a
14005         constant.
14006         (replace_swapped_load_constant): New function.
14007         (rs6000_analyze_swaps): Add a new pass to replace a swap of a
14008         loaded constant vector with a load of a swapped constant vector.
14010 2017-09-27  Carl Love  <cel@us.ibm.com>
14012         * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
14013         (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
14014         * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
14015         fctiw instruction.
14017 2017-09-27  Alexander Monakov  <amonakov@ispras.ru>
14019         * haifa-sched.c (autopref_rank_for_schedule): Order 'irrelevant' insns
14020         first, always call autopref_rank_data otherwise.
14022 2017-09-27  Richard Biener  <rguenther@suse.de>
14024         * graphite-scop-detection.c (find_scop_parameters): Move
14025         loop bound handling ...
14026         (gather_bbs::before_dom_children): ... here, avoiding the need
14027         to build scop_info->loop_nest.
14028         (record_loop_in_sese): Remove.
14029         * sese.h (sese_info_t::loop_nest): Remove.
14030         * sese.c (new_sese_info): Do not allocate loop_nest.
14031         (free_sese_info): Do not free loop_nest.
14033 2017-09-27  Jakub Jelinek  <jakub@redhat.com>
14035         PR c++/82159
14036         * gimplify.c (gimplify_modify_expr): Don't optimize away zero sized
14037         lhs from calls if the lhs has addressable type.
14039 2017-09-27  Richard Biener  <rguenther@suse.de>
14041         * graphite.h (scop::max_alias_set): New member.
14042         * graphite-scop-detection.c: Remove references to non-existing
14043         --param in comments.
14044         (build_alias_sets): Record the maximum alias set used for drs.
14045         (build_scops): Support zero as unlimited for
14046         --param graphite-max-arrays-per-scop.
14047         * graphite-sese-to-poly.c (add_scalar_version_numbers): Remove
14048         and inline into ...
14049         (build_poly_sr_1): ... here.  Compute alias set based on the
14050         maximum alias set used for drs rather than
14051         PARAM_GRAPHITE_MAX_ARRAYS_PER_SCOP
14053 2017-09-27  Richard Biener  <rguenther@suse.de>
14055         * graphite-optimize-isl.c (get_schedule_for_node_st): Allow
14056         --param loop-block-tile-size=0 to disable tiling.
14058 2017-09-27  Richard Biener  <rguenther@suse.de>
14060         * doc/invoke.texi (graphite-max-bbs-per-function): Remove.
14061         (graphite-max-nb-scop-params): Document special value zero.
14062         * domwalk.h (dom_walker::STOP): New symbolical constant.
14063         (dom_walker::dom_walker): Add optional parameter for bb to
14064         RPO mapping.
14065         (dom_walker::~dom_walker): Declare.
14066         (dom_walker::before_dom_children): Document STOP return value.
14067         (dom_walker::m_user_bb_to_rpo): New member.
14068         (dom_walker::m_bb_to_rpo): Likewise.
14069         * domwalk.c (dom_walker::dom_walker): Compute bb to RPO
14070         mapping here if not provided by the user.
14071         (dom_walker::~dom_walker): Free bb to RPO mapping if not
14072         provided by the user.
14073         (dom_walker::STOP): Define.
14074         (dom_walker::walk): Do not compute bb to RPO mapping here.
14075         Support STOP return value from before_dom_children to stop
14076         walking.
14077         * graphite-optimize-isl.c (optimize_isl): If the schedule
14078         is the same still generate code if -fgraphite-identity
14079         or -floop-parallelize-all are given.
14080         * graphite-scop-detection.c: Include cfganal.h.
14081         (gather_bbs::gather_bbs): Get and pass through bb to RPO
14082         mapping.
14083         (gather_bbs::before_dom_children): Return STOP for BBs
14084         not in the region.
14085         (build_scops): Compute bb to RPO mapping and pass it to
14086         the domwalk.  Treat --param graphite-max-nb-scop-params=0
14087         as not limiting the number of params.
14088         * graphite.c (graphite_initialize): Remove limit on the
14089         number of basic-blocks in a function.
14090         * params.def (PARAM_GRAPHITE_MAX_BBS_PER_FUNCTION): Remove.
14091         (PARAM_GRAPHITE_MAX_NB_SCOP_PARAMS): Adjust to documented
14092         default value of 10.
14094 2017-09-26  Michael Meissner  <meissner@linux.vnet.ibm.com>
14096         * config/rs6000/vsx.md (peephole for optimizing move SF to GPR):
14097         Adjust code to eliminate needing to do the shift right 32-bits
14098         operation after XSCVDPSPN.
14100 2017-09-26  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
14102         * match.pd ((X / Y) == 0 -> X < Y): New pattern.
14103         ((X / Y) != 0 -> X >= Y): Likewise.
14105 2017-09-26  Carl Love  <cel@us.ibm.com>
14107         * config/rs6000/rs6000-c.c (P9V_BUILTIN_VEC_XL_LEN_R,
14108         P9V_BUILTIN_VEC_XST_LEN_R): Add support for builtins
14109         vector unsigned char vec_xl_len_r (unsigned char *, size_t);
14110         void vec_xst_len_r (vector unsigned char, unsigned char *, size_t);
14111         * config/rs6000/altivec.h (vec_xl_len_r, vec_xst_len_r): Add defines.
14112         * config/rs6000/rs6000-builtin.def (XL_LEN_R, XST_LEN_R): Add
14113         definitions and overloading.
14114         * config/rs6000/rs6000.c (altivec_expand_builtin): Add case
14115         statement for P9V_BUILTIN_XST_LEN_R.
14116         (altivec_init_builtins): Add def_builtin for P9V_BUILTIN_STXVLL.
14117         * config/rs6000/vsx.md (lxvll, stxvll, xl_len_r, xst_len_r): Add
14118         define_expand and define_insn for the instructions and builtins.
14119         * doc/extend.texi: Update the built-in documentation file for the new
14120         built-in functions.
14121         * config/rs6000/altivec.md (altivec_lvsl_reg, altivec_lvsr_reg): Add
14122         define_insn for the instructions
14124 2017-09-26  Krister Walfridsson  <krister.walfridsson@gmail.com>
14126         PR target/39570
14127         * gcc/config/netbsd-protos.h: New file.
14128         * gcc/config/netbsd.c: New file.
14129         * gcc/config/netbsd.h (SUBTARGET_INIT_BUILTINS): Define.
14130         * gcc/config/t-netbsd: New file.
14131         * gcc/config.gcc (tm_p_file): Add netbsd-protos.h.
14132         (tmake_file) Add t-netbsd.
14133         (extra_objs) Add netbsd.o.
14135 2017-09-26  Janus Weil  <janus@gcc.gnu.org>
14137         PR fortran/82143
14138         PR fortran/82324
14139         * doc/sourcebuild.texi: Document fortran_real_10 and fortran_real_16.
14141 2017-09-26  Michael Meissner  <meissner@linux.vnet.ibm.com>
14143         * config/rs6000/rs6000.md (extendsi<mode>2): Add a splitter to do
14144         sign extension from a vector register to a GPR by doing a 32-bit
14145         direct move and then an EXTSW.
14146         (extendsi<mode>2 splitter): Likewise.
14147         (movsi_from_sf): Adjust code to eliminate doing a 32-bit shift
14148         right or vector extract after doing XSCVDPSPN.  Use
14149         zero_extendsidi2 instead of p8_mfvsrd_4_disf to move the value to
14150         the GPRs.
14151         (movdi_from_sf_zero_ext): Likewise.
14152         (reload_gpr_from_vsxsf): Likewise.
14153         (p8_mfvsrd_4_disf): Delete, no longer used.
14154         (movsi_from_df): Optimize converting a DFmode to a SFmode, and
14155         then needing to move the SFmode to a GPR to use the XSCVDPSP
14156         instruction instead of FRSP and XSCVDPSPN.
14157         * config/rs6000/vsx.md (vsx_xscvspdp_scalar2): Move insn so that
14158         it is adjacent to the other XSCVSPDP insns.
14159         (vsx_xscvdpsp_scalar): Use "ww" constraint instead of "f" to allow
14160         SFmode to be in traditional Altivec registers.
14161         (vsx_xscvdpspn): Eliminate useless alternative constraint.
14162         (vsx_xscvspdpn): Likewise.
14163         (vsx_xscvspdpn_scalar): Likewise.
14165 2017-09-26  Martin Jambor  <mjambor@suse.cz>
14167         * tree-sra.c (compare_access_positions): Put integral types first,
14168         stabilize sorting of integral types, remove conditions putting
14169         non-full-precision integers last.
14170         (sort_and_splice_var_accesses): Disable scalarization if a
14171         non-integert would be represented by a non-full-precision integer.
14173 2017-09-26  Joseph Myers  <joseph@codesourcery.com>
14175         * config/microblaze/linux.h (TARGET_ASM_FILE_END): Likewise.
14176         * config/pa/pa.h (NEED_INDICATE_EXEC_STACK): Likewise.
14177         * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Likewise.
14178         * config/pa/pa.c (pa_hpux_file_end): Rename to pa_file_end.
14179         Define unconditionally, with [ASM_OUTPUT_EXTERNAL_REAL]
14180         conditionals inside the function instead of around it.  Call
14181         file_end_indicate_exec_stack if NEED_INDICATE_EXEC_STACK.
14182         (TARGET_ASM_FILE_END): Define unconditionally to pa_file_end.
14184 2017-09-26  Richard Biener  <rguenther@suse.de>
14186         * graphite-scop-detection.c (scop_detection::build_scop_depth): Rewrite,
14187         fold in ...
14188         (scop_detection::build_scop_breadth): ... this.  Removed.
14189         (scop_detection::loop_is_valid_in_scop): Fold into single caller.
14190         (scop_detection::harmful_stmt_in_bb): Likewise.
14191         (scop_detection::graphite_can_represent_stmt): Likewise.
14192         (scop_detection::loop_body_is_valid_scop): Likewise.  Remove recursion.
14193         (scop_detection::can_represent_loop): Remove recursion, fold in ...
14194         (scop_detection::can_represent_loop_1): ... this.  Removed.
14195         (scop_detection::harmful_loop_in_region): Simplify after inlining
14196         the above and remove more quadraticness.
14197         (build_scops): Adjust.
14198         * tree-data-ref.c (loop_nest_has_data_refs): Remove pointless
14199         quadraticness.
14201 2017-09-26  Jakub Jelinek  <jakub@redhat.com>
14203         PR target/82267
14204         * config/i386/i386.c (ix86_print_operand_address_as): Only test
14205         REGNO (base) == SP_REG if base is a REG.
14207         PR middle-end/35691
14208         * tree-ssa-reassoc.c (update_range_test): Dump r->exp each time
14209         if it is different SSA_NAME.
14210         (optimize_range_tests_cmp_bitwise): New function.
14211         (optimize_range_tests): Call it.
14213 2017-09-26  Richard Biener  <rguenther@suse.de>
14215         PR tree-optimization/82321
14216         * graphite.c (canonicalize_loop_closed_ssa): Properly check
14217         for the def being inside the loop.
14219 2017-09-26  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
14221         * config/s390/vx-builtins.md ("vmslg"): Add missing operand in
14222         assembler output.
14223         * config/s390/s390-builtins.def: Fix constraint on op4.
14225 2017-09-26  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
14227         * config/s390/s390.c (s390_expand_vec_compare): Use the new mode
14228         independent expanders.
14229         * config/s390/vector.md ("vec_cmpuneq", "vec_cmpltgt")
14230         ("vec_ordered", "vec_unordered"): New expanders.
14232 2017-09-26  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
14234         * config/s390/s390.c (s390_preferred_simd_mode): Return V4SFmode
14235         for SFmode.
14237 2017-09-26  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
14239         * config/s390/vector.md ("vec_unpacks_low_v16qi"): Rename to
14240         vec_unpacks_lo_v16qi.
14241         ("vec_unpacku_low_v16qi"): Rename to vec_unpacku_lo_v16qi.
14243 2017-09-26  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
14245         * config/s390/vector.md ("vec_unpacks_lo_v4sf")
14246         ("vec_unpacks_hi_v4sf", "vec_unpacks_lo_v2df")
14247         ("vec_unpacks_hi_v2df", "vec_pack_trunc_v2df"): New expanders.
14249 2017-09-26  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
14251         * config/s390/predicates.md ("const_shift_by_byte_operand"): New
14252         predicate.
14253         * config/s390/vector.md ("*vec_srb<mode>"): Change modes to V_128
14254         and V16QI.
14255         ("*vec_slb<mode>"): New insn pattern.
14256         ("vec_shr_<mode>"): New expander.
14257         * config/s390/vx-builtins.md ("vec_slb<mode>"): Turn into expander
14258         and force the shift count operand to V16QImode.
14259         ("vec_srb<mode>"): Set shift count mode to V16QI.
14261 2017-09-26  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
14263         * config/s390/vector.md ("vec_widen_umult_lo_<mode>")
14264         ("vec_widen_umult_hi_<mode>", "vec_widen_smult_lo_<mode>")
14265         ("vec_widen_smult_hi_<mode>"): New expander definitions.
14267 2017-09-26  Richard Earnshaw  <rearnsha@arm.com>
14269         PR target/82175
14270         * config/arm/arm.h (DRIVER_SELF_SPECS): Separate sub-rules with commas.
14272 2017-09-26  Richard Biener  <rguenther@suse.de>
14274         PR tree-optimization/82320
14275         * tree-ssa-sccvn.c (set_ssa_val_to): Changing undef to undef
14276         isn't a change.
14278 2017-09-25  Jeff Law  <law@redhat.com>
14280         * config/rs6000/rs6000-protos.h (output_probe_stack_range): Update
14281         prototype for new argument.
14282         * config/rs6000/rs6000.c (rs6000_emit_allocate_stack_1): New function,
14283         mostly extracted from rs6000_emit_allocate_stack.
14284         (rs6000_emit_probe_stack_range_stack_clash): New function.
14285         (rs6000_emit_allocate_stack): Call
14286         rs6000_emit_probe_stack_range_stack_clash as needed.
14287         (rs6000_emit_probe_stack_range): Add additional argument
14288         to call to gen_probe_stack_range{si,di}.
14289         (output_probe_stack_range): New.
14290         (output_probe_stack_range_1): Renamed from output_probe_stack_range.
14291         (output_probe_stack_range_stack_clash): New.
14292         (rs6000_emit_prologue): Emit notes into dump file as requested.
14293         * rs6000.md (allocate_stack): Handle -fstack-clash-protection.
14294         (probe_stack_range<P:mode>): Operand 0 is now early-clobbered.
14295         Add additional operand and pass it to output_probe_stack_range.
14297 2017-09-25  Bin Cheng  <bin.cheng@arm.com>
14299         PR tree-optimization/82163
14300         * tree-ssa-loop-manip.h (verify_loop_closed_ssa): New parameter.
14301         (checking_verify_loop_closed_ssa): New parameter.
14302         * tree-ssa-loop-manip.c (check_loop_closed_ssa_use): Delete.
14303         (check_loop_closed_ssa_stmt): Delete.
14304         (check_loop_closed_ssa_def, check_loop_closed_ssa_bb): New functions.
14305         (verify_loop_closed_ssa): Check loop closed ssa form for LOOP.
14306         (tree_transform_and_unroll_loop): Check loop closed ssa form only for
14307         changed loops.
14309 2017-09-25  Pekka Jaaskelainen <pekka@parmance.com>
14311         * brig-builtins.def: Treat HSAIL barrier builtins as
14312         setjmp/longjump style functions.
14314 2017-09-25  Richard Sandiford  <richard.sandiford@linaro.org>
14316         * target.def (constant_alignment): New hook.
14317         * defaults.h (CONSTANT_ALIGNMENT): Delete.
14318         * doc/tm.texi.in (CONSTANT_ALIGNMENT): Replace with...
14319         (TARGET_CONSTANT_ALIGNMENT): ...this new hook.
14320         * doc/tm.texi: Regenerate.
14321         * targhooks.h (default_constant_alignment): Declare.
14322         (constant_alignment_word_strings): Likewise.
14323         * targhooks.c (default_constant_alignment): New function.
14324         (constant_alignment_word_strings): Likewise.
14325         * builtins.c (get_object_alignment_2): Use targetm.constant_alignment
14326         instead of CONSTANT_ALIGNMENT.
14327         * varasm.c (align_variable, get_variable_align, build_constant_desc)
14328         (force_const_mem): Likewise.
14329         * config/aarch64/aarch64.h (CONSTANT_ALIGNMENT): Delete.
14330         * config/aarch64/aarch64.c (aarch64_constant_alignment): New function.
14331         (aarch64_classify_address): Call it instead of CONSTANT_ALIGNMENT.
14332         (TARGET_CONSTANT_ALIGNMENT): Redefine.
14333         * config/alpha/alpha.h (CONSTANT_ALIGNMENT): Delete commented-out
14334         definition.
14335         * config/arc/arc.h (CONSTANT_ALIGNMENT): Delete.
14336         * config/arc/arc.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14337         constant_alignment_word_strings.
14338         * config/arm/arm.h (CONSTANT_ALIGNMENT_FACTOR): Delete.
14339         (CONSTANT_ALIGNMENT): Likewise.
14340         * config/arm/arm.c (TARGET_CONSTANT_ALIGNMENT): Redefine.
14341         (arm_constant_alignment): New function.
14342         * config/bfin/bfin.h (CONSTANT_ALIGNMENT): Delete.
14343         * config/bfin/bfin.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14344         constant_alignment_word_strings.
14345         * config/cr16/cr16.h (CONSTANT_ALIGNMENT): Delete.
14346         * config/cr16/cr16.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14347         constant_alignment_word_strings.
14348         * config/cris/cris.h (CONSTANT_ALIGNMENT): Delete.
14349         * config/cris/cris.c (TARGET_CONSTANT_ALIGNMENT): Redefine.
14350         (cris_constant_alignment): New function.
14351         * config/epiphany/epiphany.h (CONSTANT_ALIGNMENT): Delete.
14352         * config/epiphany/epiphany.c (TARGET_CONSTANT_ALIGNMENT): Redefine.
14353         (epiphany_constant_alignment): New function.
14354         * config/fr30/fr30.h (CONSTANT_ALIGNMENT): Delete.
14355         * config/fr30/fr30.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14356         constant_alignment_word_strings.
14357         * config/frv/frv.h (CONSTANT_ALIGNMENT): Delete.
14358         * config/frv/frv.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14359         constant_alignment_word_strings.
14360         * config/ft32/ft32.h (CONSTANT_ALIGNMENT): Delete.
14361         * config/ft32/ft32.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14362         constant_alignment_word_strings.
14363         * config/i386/i386.h (CONSTANT_ALIGNMENT): Delete.
14364         * config/i386/i386-protos.h (ix86_constant_alignment): Delete.
14365         * config/i386/i386.c (ix86_constant_alignment): Make static.
14366         Use the same interface as the target hook.
14367         (TARGET_CONSTANT_ALIGNMENT): Redefine.
14368         * config/ia64/ia64.h (CONSTANT_ALIGNMENT): Delete.
14369         * config/ia64/ia64.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14370         constant_alignment_word_strings.
14371         * config/iq2000/iq2000.h (CONSTANT_ALIGNMENT): Delete.
14372         * config/iq2000/iq2000.c (iq2000_constant_alignment): New function.
14373         (TARGET_CONSTANT_ALIGNMENT): Redefine.
14374         * config/lm32/lm32.h (CONSTANT_ALIGNMENT): Delete.
14375         * config/lm32/lm32.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14376         constant_alignment_word_strings.
14377         * config/m32r/m32r.h (CONSTANT_ALIGNMENT): Delete.
14378         * config/m32r/m32r.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14379         constant_alignment_word_strings.
14380         * config/mcore/mcore.h (CONSTANT_ALIGNMENT): Delete.
14381         * config/mcore/mcore.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14382         constant_alignment_word_strings.
14383         * config/microblaze/microblaze.h (CONSTANT_ALIGNMENT): Delete.
14384         * config/microblaze/microblaze.c (microblaze_constant_alignment):
14385         New function.
14386         (TARGET_CONSTANT_ALIGNMENT): Redefine.
14387         * config/mips/mips.h (CONSTANT_ALIGNMENT): Delete.
14388         * config/mips/mips.c (mips_constant_alignment): New function.
14389         (TARGET_CONSTANT_ALIGNMENT): Redefine.
14390         * config/mmix/mmix.h (CONSTANT_ALIGNMENT): Delete.
14391         * config/mmix/mmix-protos.h (mmix_constant_alignment): Delete.
14392         * config/mmix/mmix.c (TARGET_CONSTANT_ALIGNMENT): Redefine.
14393         (mmix_constant_alignment): Make static.  Use the same interface
14394         as the target hook.
14395         * config/moxie/moxie.h (CONSTANT_ALIGNMENT): Delete.
14396         * config/moxie/moxie.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14397         constant_alignment_word_strings.
14398         * config/nios2/nios2.h (CONSTANT_ALIGNMENT): Delete.
14399         * config/nios2/nios2.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14400         constant_alignment_word_strings.
14401         * config/pa/pa.h (CONSTANT_ALIGNMENT): Delete.
14402         * config/pa/pa.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14403         constant_alignment_word_strings.
14404         * config/powerpcspe/powerpcspe.h (CONSTANT_ALIGNMENT): Delete.
14405         * config/powerpcspe/powerpcspe.c (TARGET_CONSTANT_ALIGNMENT): Redefine.
14406         (rs6000_constant_alignment): New function.
14407         * config/riscv/riscv.h (CONSTANT_ALIGNMENT): Delete.
14408         * config/riscv/riscv.c (riscv_constant_alignment): New function.
14409         (TARGET_CONSTANT_ALIGNMENT): Redefine.
14410         * config/rs6000/rs6000.h (CONSTANT_ALIGNMENT): Delete.
14411         * config/rs6000/rs6000.c (TARGET_CONSTANT_ALIGNMENT): Redefine.
14412         (rs6000_constant_alignment): New function.
14413         * config/s390/s390.h (CONSTANT_ALIGNMENT): Delete.
14414         * config/s390/s390.c (s390_constant_alignment): New function.
14415         (TARGET_CONSTANT_ALIGNMENT): Redefine.
14416         * config/sh/sh.h (CONSTANT_ALIGNMENT): Delete.
14417         * config/sh/sh.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14418         constant_alignment_word_strings.
14419         * config/sparc/sparc.h (CONSTANT_ALIGNMENT): Delete.
14420         * config/sparc/sparc.c (TARGET_CONSTANT_ALIGNMENT): Redefine.
14421         (sparc_constant_alignment): New function.
14422         * config/spu/spu.h (CONSTANT_ALIGNMENT): Delete.
14423         * config/spu/spu.c (spu_constant_alignment): New function.
14424         (TARGET_CONSTANT_ALIGNMENT): Redefine.
14425         * config/stormy16/stormy16.h (CONSTANT_ALIGNMENT): Delete.
14426         * config/stormy16/stormy16.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14427         constant_alignment_word_strings.
14428         * config/tilegx/tilegx.h (CONSTANT_ALIGNMENT): Delete.
14429         * config/tilegx/tilegx.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14430         constant_alignment_word_strings.
14431         * config/tilepro/tilepro.h (CONSTANT_ALIGNMENT): Delete.
14432         * config/tilepro/tilepro.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14433         constant_alignment_word_strings.
14434         * config/visium/visium.h (CONSTANT_ALIGNMENT): Delete.
14435         * config/visium/visium.c (TARGET_CONSTANT_ALIGNMENT): Redefine.
14436         (visium_constant_alignment): New function.
14437         * config/xtensa/xtensa.h (CONSTANT_ALIGNMENT): Delete.
14438         * config/xtensa/xtensa.c (TARGET_CONSTANT_ALIGNMENT): Redefine.
14439         (xtensa_constant_alignment): New function.
14440         * system.h (CONSTANT_ALIGNMENT): Poison.
14442 2017-09-25  Will Schmidt  <will_schmidt@vnet.ibm.com>
14444         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add handling
14445         for early folding of vector stores (ALTIVEC_BUILTIN_ST_*).
14446         (rs6000_builtin_valid_without_lhs): New helper function.
14447         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
14448         Remove obsoleted code for handling ALTIVEC_BUILTIN_VEC_ST.
14450 2017-09-25  Richard Sandiford  <richard.sandiford@linaro.org>
14452         * target.h (vec_perm_indices): Use unsigned short rather than
14453         unsigned char.
14454         (auto_vec_perm_indices): Likewise.
14455         * config/aarch64/aarch64.c (aarch64_vectorize_vec_perm_const_ok):
14456         Use unsigned int rather than unsigned char.
14457         * config/arm/arm.c (arm_vectorize_vec_perm_const_ok): Likewise.
14459 2017-09-25  Richard Biener  <rguenther@suse.de>
14461         * cfgloop.h (sort_sibling_loops): Declare.
14462         * cfgloop.c (sort_sibling_loops_cmp): New helper.
14463         (sort_sibling_loops): New function sorting the sibling loop list
14464         in RPO order.
14465         * graphite.c (graphite_transform_loops): Sort sibling loops.
14467 2017-09-25  Richard Sandiford  <richard.sandifird@linaro.org>
14469         * target.def (vec_perm_const_ok): Change sel parameter to
14470         vec_perm_indices.
14471         * optabs-query.c (can_vec_perm_p): Update accordingly.
14472         * doc/tm.texi: Regenerate.
14473         * config/aarch64/aarch64.c (expand_vec_perm_d): Change perm to
14474         auto_vec_perm_indices and remove separate nelt field.
14475         (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
14476         (aarch64_evpc_ext, aarch64_evpc_rev, aarch64_evpc_dup)
14477         (aarch64_evpc_tbl, aarch64_expand_vec_perm_const_1)
14478         (aarch64_expand_vec_perm_const): Update accordingly.
14479         (aarch64_vectorize_vec_perm_const_ok): Likewise.  Change sel
14480         to vec_perm_indices.
14481         * config/arm/arm.c (expand_vec_perm_d): Change perm to
14482         auto_vec_perm_indices and remove separate nelt field.
14483         (arm_evpc_neon_vuzp, arm_evpc_neon_vzip, arm_evpc_neon_vrev)
14484         (arm_evpc_neon_vtrn, arm_evpc_neon_vext, arm_evpc_neon_vtbl)
14485         (arm_expand_vec_perm_const_1, arm_expand_vec_perm_const): Update
14486         accordingly.
14487         (arm_vectorize_vec_perm_const_ok): Likewise.  Change sel
14488         to vec_perm_indices.
14489         * config/i386/i386.c (ix86_vectorize_vec_perm_const_ok): Change
14490         sel to vec_perm_indices.
14491         * config/ia64/ia64.c (ia64_vectorize_vec_perm_const_ok): Likewise.
14492         * config/mips/mips.c (mips_vectorize_vec_perm_const_ok): Likewise.
14493         * config/powerpcspe/powerpcspe.c (rs6000_vectorize_vec_perm_const_ok):
14494         Likewise.
14495         * config/rs6000/rs6000.c (rs6000_vectorize_vec_perm_const_ok):
14496         Likewise.
14498 2017-09-25  Pierre-Marie de Rodat  <derodat@adacore.com>
14500         PR debug/82155
14501         * dwarf2out.c (dwarf2out_early_global_decl): Call dwarf2out_decl
14502         on the FUNCTION_DECL function context if it has a DIE that is a
14503         declaration.
14505 2017-09-25  Richard Biener  <rguenther@suse.de>
14507         PR tree-optimization/82285
14508         * tree-vect-patterns.c (vect_recog_bool_pattern): Also handle
14509         enumeral types.
14511 2017-09-25  Tom de Vries  <tom@codesourcery.com>
14513         PR target/80035
14514         PR target/81069
14515         * config/nvptx/nvptx.c (nvptx_output_call_insn): Add exit after call to
14516         noreturn function.
14518 2017-09-25  Richard Biener  <rguenther@suse.de>
14520         * graphite-optimize-isl.c (optimize_isl): Fail and dump if
14521         ISL errors other than isl_error_quota happen.  Dump if the
14522         schedule is the same.
14523         * graphite-sese-to-poly.c (build_poly_scop): Fail on ISL
14524         errors instead of aborting inside ISL.
14526 2017-09-25  Iain Sandoe  <iain@codesourcery.com>
14528         PR target/80556
14529         * config/i386/darwin.h (REAL_LIB_SPEC): New; put libSystem ahead
14530         of libgcc_eh for m64.
14531         * config/i386/darwin64.h: Likewise.
14533 2017-09-25  Richard Biener  <rguenther@suse.de>
14535         PR middle-end/82144
14536         * dwarf2out.c (gen_enumeration_type_die): Do not add alignment
14537         attribute for incomplete types nor twice for complete ones.
14539 2017-09-24  Uros Bizjak  <ubizjak@gmail.com>
14541         PR target/82267
14542         * config/i386/i386.c (ix86_print_operand_address_as): Encode
14543         %esp as %rsp to avoid 0x67 prefix if there is no index or base
14544         register.
14546 2017-09-23  Uros Bizjak  <ubizjak@gmail.com>
14548         PR bootstrap/82306
14549         * config/i386/i386.opt (mprefer-avx256): Use
14550         ix86_target_flags variable.
14551         * config/i386/i386.c (ix86_target_string): Move
14552         -mprefer-avx256 to flag2_opts.
14554 2017-09-22  Jakub Jelinek  <jakub@redhat.com>
14556         PR middle-end/35691
14557         * match.pd: Simplify x == -1 & y == -1 into (x & y) == -1
14558         and x != -1 | y != -1 into (x & y) != -1.
14560 2017-09-22  Steve Ellcey  <sellcey@cavium.com>
14562         * config.gcc: Add new case statement to set
14563         default_gnu_indirect_function.  Remove it from x86_64-*-linux*,
14564         i[34567]86-*, powerpc*-*-linux*spe*, powerpc*-*-linux*, s390-*-linux*,
14565         s390x-*-linux* case statements.   Added aarch64 to the list of
14566         supported architectures.
14568 2017-09-22  Richard Sandiford  <richard.sandiford@linaro.org>
14570         PR tree-optimization/82289
14571         * tree-vect-data-refs.c (vect_get_peeling_costs_all_drs): Check
14572         STMT_VINFO_RELEVANT_P.
14574 2017-09-22  Richard Sandiford  <richard.sandiford@linaro.org>
14575             Alan Hayward  <alan.hayward@arm.com>
14576             David Sherwood  <david.sherwood@arm.com>
14578         * tree-vrp.c (extract_range_from_multiplicative_op_1): Assert
14579         for VR_RANGE only; don't allow VR_ANTI_RANGE.
14580         (extract_range_from_binary_expr_1): Don't call
14581         extract_range_from_multiplicative_op_1 if !range_int_cst_p.
14583 2017-09-22  Richard Sandiford  <richard.sandiford@linaro.org>
14584             Alan Hayward  <alan.hayward@arm.com>
14585             David Sherwood  <david.sherwood@arm.com>
14587         * target.def (preferred_vector_alignment): New hook.
14588         * doc/tm.texi.in (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): New
14589         hook.
14590         * doc/tm.texi: Regenerate.
14591         * targhooks.h (default_preferred_vector_alignment): Declare.
14592         * targhooks.c (default_preferred_vector_alignment): New function.
14593         * tree-vectorizer.h (dataref_aux): Add a target_alignment field.
14594         Expand commentary.
14595         (DR_TARGET_ALIGNMENT): New macro.
14596         (aligned_access_p): Update commentary.
14597         (vect_known_alignment_in_bytes): New function.
14598         * tree-vect-data-refs.c (vect_calculate_required_alignment): New
14599         function.
14600         (vect_compute_data_ref_alignment): Set DR_TARGET_ALIGNMENT.
14601         Calculate the misalignment based on the target alignment rather than
14602         the vector size.
14603         (vect_update_misalignment_for_peel): Use DR_TARGET_ALIGMENT
14604         rather than TYPE_ALIGN / BITS_PER_UNIT to update the misalignment.
14605         (vect_enhance_data_refs_alignment): Mask the byte misalignment with
14606         the target alignment, rather than masking the element misalignment
14607         with the number of elements in a vector.  Also use the target
14608         alignment when calculating the maximum number of peels.
14609         (vect_find_same_alignment_drs): Use vect_calculate_required_alignment
14610         instead of TYPE_ALIGN_UNIT.
14611         (vect_duplicate_ssa_name_ptr_info): Remove stmt_info parameter.
14612         Measure DR_MISALIGNMENT relative to DR_TARGET_ALIGNMENT.
14613         (vect_create_addr_base_for_vector_ref): Update call accordingly.
14614         (vect_create_data_ref_ptr): Likewise.
14615         (vect_setup_realignment): Realign by ANDing with
14616         -DR_TARGET_MISALIGNMENT.
14617         * tree-vect-loop-manip.c (vect_gen_prolog_loop_niters): Calculate
14618         the number of peels based on DR_TARGET_ALIGNMENT.
14619         * tree-vect-stmts.c (get_group_load_store_type): Compare the gap
14620         with the guaranteed alignment boundary when deciding whether
14621         overrun is OK.
14622         (vectorizable_mask_load_store): Interpret DR_MISALIGNMENT
14623         relative to DR_TARGET_ALIGNMENT instead of TYPE_ALIGN_UNIT.
14624         (ensure_base_align): Remove stmt_info parameter.  Get the
14625         target base alignment from DR_TARGET_ALIGNMENT.
14626         (vectorizable_store): Update call accordingly.   Interpret
14627         DR_MISALIGNMENT relative to DR_TARGET_ALIGNMENT instead of
14628         TYPE_ALIGN_UNIT.
14629         (vectorizable_load): Likewise.
14631 2017-09-22  Richard Sandiford  <richard.sandiford@linaro.org>
14632             Alan Hayward  <alan.hayward@arm.com>
14633             David Sherwood  <david.sherwood@arm.com>
14635         * tree-vectorizer.h (vect_get_scalar_dr_size): New function.
14636         * tree-vect-data-refs.c (vect_update_misalignment_for_peel): Use it.
14637         (vect_enhance_data_refs_alignment): Likewise.
14639 2017-09-22  Richard Earnshaw  <richard.earnshaw@arm.com>
14641         * config/arm/parsecpu.awk (fatal): Note that we've encountered an
14642         error.  Only quit immediately if parsing is complete.
14643         (BEGIN): Initialize fatal_err and parse_done.
14644         (begin fpu, end fpu): Check number of arguments.
14645         (begin arch, end arch): Likewise.
14646         (begin cpu, end cpu): Likewise.
14647         (cname, tune for, tune flags, architecture, fpu, option): Likewise.
14648         (optalias): Likewise.
14650 2017-09-22  Richard Earnshaw  <richard.earnshaw@arm.com>
14652         * config.gcc (arm*-*-*): Don't add arm-isa.h to tm_p_file.
14653         * config/arm/arm-isa.h: Delete.  Move definitions to ...
14654         * arm-cpus.in: ... here.  Use new feature and fgroup values.
14655         * config/arm/arm.c (arm_option_override): Use lower case for feature
14656         bit names.
14657         * config/arm/arm.h (TARGET_HARD_FLOAT): Likewise.
14658         (TARGET_VFP3, TARGET_VFP5, TARGET_FMA): Likewise.
14659         * config/arm/parsecpu.awk (END): Add new command 'isa'.
14660         (isa_pfx): Delete.
14661         (print_isa_bits_for): New function.
14662         (gen_isa): New function.
14663         (gen_comm_data): Use print_isa_bits_for.
14664         (define feature): New keyword.
14665         (define fgroup): New keyword.
14666         * config/arm/t-arm (TM_H): Remove.
14667         (GTM_H): Add arm-isa.h.
14668         (arm-isa.h): Add rule to generate file.
14669         * common/config/arm/arm-common.c: (arm_canon_arch_option): Use lower
14670         case for feature bit names.
14672 2017-09-22  Richard Biener  <rguenther@suse.de>
14674         * graphite-isl-ast-to-gimple.c (graphite_verify): Inline into
14675         single caller.
14676         (graphite_regenerate_ast_isl): Do not reset SCEV.  Move debug
14677         print of no dependency loops ...
14678         * graphite.c (graphite_transform_loops): ... here.
14679         (canonicalize_loop_closed_ssa_form): Work from inner to outer
14680         loops.
14681         (same_close_phi_node, remove_duplicate_close_phi,
14682         make_close_phi_nodes_unique, defined_in_loop_p): Fold into ...
14683         (canonicalize_loop_closed_ssa): ... here and simplify.
14684         * graphite-optimize-isl.c: Include tree-vectorizer.h.
14685         (optimize_isl): Use dump_printf_loc to tell when we stopped
14686         optimizing because of an ISL timeout.
14688 2017-09-22  Richard Biener  <rguenther@suse.de>
14690         PR tree-optimization/82291
14691         * tree-if-conv.c (predicate_mem_writes): Make sure to
14692         remove writes in blocks predicated with false.
14694 2017-09-22  Richard Biener  <rguenther@suse.de>
14696         * sese.c: Include cfganal.h.
14697         (if_region_set_false_region): Remove.
14698         (create_if_region_on_edge): Likewise.
14699         (move_sese_in_condition): Re-implement without destroying
14700         dominators.
14702 2017-09-22  Richard Biener  <rguenther@suse.de>
14704         * graphite-isl-ast-to-gimple.c (translate_pending_phi_nodes):
14705         Verify both BBs contain loop PHI nodes before dispatching to
14706         copy_loop_phi_args.
14707         (graphite_regenerate_ast_isl): Do not recompute dominators,
14708         do not verify three times.  Restructure for clarity.
14709         * graphite-scop-detection.c (same_close_phi_node,
14710         remove_duplicate_close_phi, make_close_phi_nodes_unique,
14711         defined_in_loop_p, canonicalize_loop_closed_ssa,
14712         canonicalize_loop_closed_ssa_form): Simplify, remove excess
14713         checking and SSA rewrite, move to ...
14714         * graphite.c: ... here.  Include ssa.h and tree-ssa-loop-manip.h.
14715         (graphite_initialize): Do not pass in ctx, do not reset the
14716         SCEV cache, compute only dominators.
14717         (graphite_transform_loops): Allocate ISL ctx after
14718         graphite_initialize.  Call canonicalize_loop_closed_ssa_form.
14719         Maintain post-dominators only around build_scops.
14720         * sese.c (if_region_set_false_region): Make static.  Free
14721         and recompute dominators.
14722         (move_sese_in_condition): Assert we don't get called with
14723         post-dominators computed.
14724         * sese.h (if_region_set_false_region): Remove.
14726 2017-09-22  Sergey Shalnov  <sergey.shalnov@intel.com>
14728         * config/i386/sse.md ("mov<mode>_internal"): Use <sseinsnmode>
14729         mode attribute for TARGET_AVX512VL.
14731 2017-09-21  Sergey Shalnov  <sergey.shalnov@intel.com>
14733         * config/i386/i386.opt (mprefer-avx256): New option.
14734         * config/i386/i386.c (ix86_target_string): Add -mprefer-avx256
14735         to flag_opts.
14736         (ix86_preferred_simd_mode): Return 256-bit AVX modes
14737         for TARGET_PREFER_AVX256.
14738         * doc/invoke.texi (x86 Options): Document -mprefer-avx256.
14740 2017-09-21  Jeff Law  <law@redhat.com>
14742         * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash):
14743         Fix dump output if the only stack space is for pushed registers.
14745 2017-09-21  Richard Sandiford  <richard.sandiford@linaro.org>
14747         * config/spu/spu.c (spu_sched_adjust_cost): Update after renaming
14748         of insn_cost.
14750 2017-09-21  Martin Sebor  <msebor@redhat.com>
14752         PR c/81882
14753         * doc/extend.texi (attribute ifunc): Avoid relying on ill-formed
14754         code (in C++) or code that triggers warnings.
14756 2017-09-21  Eric Botcazou  <ebotcazou@adacore.com>
14758         * stor-layout.c (bit_from_pos): Do not distribute the conversion.
14760 2017-09-21  Segher Boessenkool  <segher@kernel.crashing.org>
14762         * haifa-sched.c: Rename insn_cost to insn_sched_cost.
14763         * sched-rgn.c: Ditto.
14764         * sel-sched-ir.c: Ditto.
14766 2017-09-21  Alexander Monakov  <amonakov@ispras.ru>
14768         * toplev.h (set_random_seed): Adjust return type.
14769         * toplev.c (init_local_tick): Move eager initialization of random_seed
14770         to get_random_seed.  Adjust comment.
14771         (init_random_seed): Inline to get_random_seed, delete.
14772         (get_random_seed): Initialize random_seed lazily.
14773         (set_random_seed): Do not return previous value.
14774         (print_switch_value): Do not call get_random_seed.
14776 2017-09-21  Evgeny Kudryashov  <kudryashov@ispras.ru>
14778         * cgraph.c (delete_function_version): New, broken out from...
14779         (cgraph_node::delete_function_version): ...here.  Rename to
14780         cgraph_node::delete_function_version_by_decl.  Update all uses.
14781         (cgraph_node::remove): Call delete_function_version.
14783 2017-09-21  Jakub Jelinek  <jakub@redhat.com>
14785         PR sanitizer/81715
14786         * tree-inline.c (expand_call_inline): Emit clobber stmts for
14787         VAR_DECLs to which addressable non-volatile parameters are mapped
14788         and for id->retvar after the return value assignment.  Clear
14789         id->retval and id->retbnd after inlining.
14791 2017-09-21  Richard Biener  <rguenther@suse.de>
14793         PR tree-optimization/82276
14794         PR tree-optimization/82244
14795         * tree-vrp.c (build_assert_expr_for): Set
14796         SSA_NAME_OCCURS_IN_ABNORMAL_PHI if the variable we assert on
14797         has it set.
14798         (remove_range_assertions): Revert earlier change.
14800 2017-09-21  Wilco Dijkstra  <wdijkstr@arm.com>
14802         PR target/71951
14803         * config/aarch64/aarch64.h (LIBGCC2_UNWIND_ATTRIBUTE): Define.
14805 2017-09-21  Richard Biener  <rguenther@suse.de>
14807         * graphite-isl-ast-to-gimple.c (graphite_regenerate_ast_isl):
14808         Restore valid IL after code generation errors.
14809         * graphite.c (graphite_transform_loops): Diagnose code
14810         generation issues as MSG_MISSED_OPTIMIZATION and continue
14811         with processing SCOPs.
14813 2017-09-21  Richard Sandiford  <richard.sandiford@linaro.org>
14814             Alan Hayward  <alan.hayward@arm.com>
14815             David Sherwood  <david.sherwood@arm.com>
14817         * calls.c (compute_argument_addresses): Use simplify_gen_binary
14818         rather than choosing between plus_constant and gen_rtx_<CODE>.
14819         * expr.c (emit_push_insn): Likewise.
14820         (expand_expr_real_2): Likewise.
14822 2017-09-21  Richard Sandiford  <richard.sandiford@linaro.org>
14823             Alan Hayward  <alan.hayward@arm.com>
14824             David Sherwood  <david.sherwood@arm.com>
14826         * loop-unroll.c (split_iv): Call copy_rtx on the step.
14828 2017-09-21  Richard Sandiford  <richard.sandiford@linaro.org>
14829             Alan Hayward  <alan.hayward@arm.com>
14830             David Sherwood  <david.sherwood@arm.com>
14832         * tree.c (find_atomic_core_type): Check tree_fits_uhwi_p before
14833         calling tree_to_uhwi.
14835 2017-09-21  Richard Sandiford  <richard.sandiford@linaro.org>
14836             Alan Hayward  <alan.hayward@arm.com>
14837             David Sherwood  <david.sherwood@arm.com>
14839         * tree-ssa-ccp.c (get_value_for_expr): Use a positive test for
14840         INTEGER_CST rather than a negative test for ADDR_EXPR.
14842 2017-09-21  Richard Sandiford  <richard.sandiford@linaro.org>
14843             Alan Hayward  <alan.hayward@arm.com>
14844             David Sherwood  <david.sherwood@arm.com>
14846         * tree-vrp.c (extract_range_from_binary_expr_1): Check
14847         int_cst_rangeN before calling value_range_constant_singleton (&vrN).
14849 2017-09-21  Richard Biener  <rguenther@suse.de>
14851         PR tree-optimization/71351
14852         * graphite-isl-ast-to-gimple.c (translate_isl_ast_to_gimple::
14853         graphite_create_new_loop_guard): Remove, fold remaining parts
14854         into caller ...
14855         (translate_isl_ast_node_for): ... here and simplify.
14857 2017-09-21  Jakub Jelinek  <jakub@redhat.com>
14859         PR target/82260
14860         * config/i386/i386.md (*movqi_internal): Replace (=q,q) alternative
14861         with (=Q,Q), (=R,R) and (=r,r) alternatives, only enable the
14862         latter two for 64-bit, renumber alternatives, for -Os imov (=q,n)
14863         alternative always use QI mode, for -Os imov (=R,R) alternative
14864         always use SI mode, for imov (=Q,Q) or (=r,r) alternatives
14865         ignore -Os.
14867 2017-09-20  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
14868             Jeff Law  <law@redhat.com>
14870         * config/s390/s390.c (MIN_UNROLL_PROBES): Define.
14871         (allocate_stack_space): New function, partially extracted from
14872         s390_emit_prologue.
14873         (s390_emit_prologue): Track offset to most recent stack probe.
14874         Code to allocate space moved into allocate_stack_space.
14875         Dump actions when no stack is allocated.
14876         (s390_prologue_plus_offset): New function.
14877         (s390_emit_stack_probe): Likewise.
14879 2017-09-20  Alexandre Oliva  <aoliva@redhat.com>
14881         * common.opt (Wa, Wl, Wp, g, gz=): Add
14882         RejectNegative.
14883         (gno-column-info): Remove.
14884         (gcolumn-info): Drop RejectNegative.
14885         (gno-): New prefix.
14886         (gno-record-gcc-switches): Remove.
14887         (grecord-gcc-switches): Drop RejectNegative.
14888         (gno-split-dwarf): Remove.
14889         (gsplit-dwarf): Drop RejectNegative.
14890         (gno-strict-dwarf): Remove.
14891         (gstrict-dwarf): Drop RejectNegative.
14892         * config/darwin.opt (gfull, gused): Add RejectNegative.
14893         * dwarf2out.c (gen_producer_string): Drop
14894         gno-record-gcc-switches handler.
14895         * optc-gen.awk: Add g to prefixes with negative forms.
14896         * opts-common.c (remapping_prefix_p): New.
14897         (find_opt): Check it.
14898         (generate_canonical_option): Test g prefix.
14899         (option_map): Add -gno- mapping.
14900         (add_misspelling_candidates): Check remapping_prefix_p.
14902 2017-09-20  Jeff Law  <law@redhat.com>
14904         * config/powerpcspe/powerpcspe.c (rs6000_expand_prologue): Fix
14905         thinko in stack clash protection support.
14907         * explow.c (compute_stack_clash_protection_loop_data): Use
14908         CONST_INT_P instead of explicit test.  Verify object is a
14909         CONST_INT_P before looking at INTVAL.
14910         (anti_adjust_stack_and_probe_stack_clash): Use CONST_INT_P
14911         instead of explicit test.
14913 2017-09-20  Segher Boessenkool  <segher@kernel.crashing.org>
14915         PR target/77687
14916         * config/rs6000/rs6000.md (stack_restore_tie): Store to a scratch
14917         address instead of to r1 and r11.
14919 2017-09-20  Sebastian Peryt  <sebastian.peryt@intel.com>
14921         * config.gcc: Support "knm".
14922         * config/i386/driver-i386.c (host_detect_local_cpu): Detect "knm".
14923         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
14924         PROCESSOR_KNM.
14925         * config/i386/i386.c (m_KNM): Define.
14926         (processor_target_table): Add "knm".
14927         (PTA_KNM): Define.
14928         (ix86_option_override_internal): Add "knm".
14929         (ix86_issue_rate): Add PROCESSOR_KNM.
14930         (ix86_adjust_cost): Ditto.
14931         (ia32_multipass_dfa_lookahead): Ditto.
14932         (get_builtin_code_for_version): Handle PROCESSOR_KNM.
14933         (fold_builtin_cpu): Add M_INTEL_KNM.
14934         * config/i386/i386.h (processor_costs): Define TARGET_KNM.
14935         (processor_type): Add PROCESSOR_KNM.
14936         * config/i386/x86-tune.def: Add m_KNM.
14937         * doc/invoke.texi: Add knm as x86 -march=/-mtune= CPU type.
14939 2017-09-20  Richard Biener  <rguenther@suse.de>
14941         PR tree-optimization/80213
14942         * graphite-scop-detection.c (trivially_empty_bb_p): Labels
14943         are allowed in empty BBs as well.
14944         (canonicalize_loop_closed_ssa): Also look for other complex
14945         edges.
14946         (scop_detection::get_sese): Include the loop-closed PHI block
14947         in loop SESEs.
14948         (scop_detection::merge_sese): Remove code adding extra blocks.
14949         (scop_detection::region_has_one_loop): Adjust for get_sese changes.
14950         (build_scops): Assert the final returned scop is invalid.
14952 2017-09-20  Richard Biener  <rguenther@suse.de>
14954         PR tree-optimization/82264
14955         * tree-ssa-sccvn.c (vn_phi_eq): Use safe_dyn_cast to check
14956         for GIMPLE_CONDs.
14957         (vn_phi_lookup): Likewise.
14958         (vn_phi_insert): Likewise.
14960 2017-09-20  Jakub Jelinek  <jakub@redhat.com>
14962         * dwarf2out.c (tree_add_const_value_attribute): For INTEGER_CST
14963         that fits into uhwi or shwi, add DW_AT_const_value regardless
14964         of early_dwarf without going through RTL, using add_AT_unsigned
14965         or add_AT_int.
14967         * dwarf2out.c (DEBUG_LTO_DWO_INFO_SECTION): Reorder defines.
14968         (DEBUG_LTO_ABBREV_SECTION): Likewise.
14969         (DEBUG_LTO_MACINFO_SECTION): Likewise.
14970         (DEBUG_MACRO_SECTION): Likewise.
14971         (DEBUG_LTO_MACRO_SECTION): Likewise.
14972         (DEBUG_STR_DWO_SECTION): Likewise.
14973         (DEBUG_LTO_STR_DWO_SECTION): Likewise.
14974         (DEBUG_LTO_LINE_SECTION): Drop .dwo suffix from the name.
14975         (DEBUG_LTO_DWO_LINE_SECTION): Define.
14976         (DEBUG_LTO_LINE_STR_SECTION): Define.
14977         (init_sections_and_labels): Initialize debug_line_str_section
14978         variable.  Initialize debug_loc_section for -gdwarf-5 to
14979         DEBUG_LOCLISTS_SECTION.  Formatting fixes.
14981 2017-09-20  Richard Biener  <rguenther@suse.de>
14983         * graphite-sese-to-poly.c (extract_affine): Properly handle
14984         POINTER_PLUS_EXPR, BIT_NOT_EXPR and conversion to signed.
14986 2017-09-20  Richard Biener  <rguenther@suse.de>
14988         PR tree-optimization/81373
14989         * graphite-scop-detection.c (build_cross_bb_scalars_def):
14990         Force SESE live-out defs to be handled even if they are
14991         scev_analyzable_p.
14993 2017-09-19  Jeff Law  <law@redhat.com>
14995         * combine-stack-adj.c (combine_stack_adjustments_for_block): Do
14996         nothing for stack adjustments with REG_STACK_CHECK.
14997         * sched-deps.c (parse_add_or_inc): Reject insns with
14998         REG_STACK_CHECK from dependency breaking.
14999         * config/i386/i386.c (pro_epilogue_adjust_stack): Return insn.
15000         (ix86_adjust_satck_and_probe_stack_clash): Add REG_STACK_NOTEs.
15001         * reg-notes.def (STACK_CHECK): New note.
15003         * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): New.
15004         (ix86_expand_prologue): Dump stack clash info as needed.
15005         Call ix86_adjust_stack_and_probe_stack_clash as needed.
15007         * function.c (dump_stack_clash_frame_info): New function.
15008         * function.h (dump_stack_clash_frame_info): Prototype.
15009         (enum stack_clash_probes): New enum.
15011         * config/alpha/alpha.c (alpha_expand_prologue): Also check
15012         flag_stack_clash_protection.
15013         * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Likewise.
15014         (arm_expand_prologue, thumb1_expand_prologue): Likewise.
15015         (arm_frame_pointer_required): Likewise.
15016         * config/ia64/ia64.c (ia64_compute_frame_size): Likewise.
15017         (ia64_expand_prologue): Likewise.
15018         * config/mips/mips.c (mips_expand_prologue): Likewise.
15019         * config/powerpcspe/powerpcspe.c (rs6000_expand_prologue): Likewise.
15020         * config/sparc/sparc.c (sparc_expand_prologue): Likewise.
15021         (sparc_flat_expand_prologue): Likewise.
15022         * config/spu/spu.c (spu_expand_prologue): Likewise.
15024         * explow.c: Include "params.h".
15025         (anti_adjust_stack_and_probe_stack_clash): New function.
15026         (get_stack_check_protect): Likewise.
15027         (compute_stack_clash_protection_loop_data): Likewise.
15028         (emit_stack_clash_protection_loop_start): Likewise.
15029         (emit_stack_clash_protection_loop_end): Likewise.
15030         (allocate_dynamic_stack_space): Use get_stack_check_protect.
15031         Use anti_adjust_stack_and_probe_stack_clash.
15032         * explow.h (compute_stack_clash_protection_loop_data): Prototype.
15033         (emit_stack_clash_protection_loop_start): Likewise.
15034         (emit_stack_clash_protection_loop_end): Likewise.
15035         * rtl.h (get_stack_check_protect): Prototype.
15036         * target.def (stack_clash_protection_final_dynamic_probe): New hook.
15037         * targhooks.c (default_stack_clash_protection_final_dynamic_probe): New.
15038         * targhooks.h (default_stack_clash_protection_final_dynamic_probe):
15039         Prototype.
15040         * doc/tm.texi.in (TARGET_STACK_CLASH_PROTECTION_FINAL_DYNAMIC_PROBE):
15041         Add @hook.
15042         * doc/tm.texi: Rebuilt.
15043         * config/aarch64/aarch64.c (aarch64_expand_prologue): Use
15044         get_stack_check_protect.
15045         * config/alpha/alpha.c (alpha_expand_prologue): Likewise.
15046         * config/arm/arm.c (arm_expand_prologue): Likewise.
15047         (arm_frame_pointer_required): Likewise.
15048         * config/i386/i386.c (ix86_expand_prologue): Likewise.
15049         * config/ia64/ia64.c (ia64_expand_prologue): Likewise.
15050         * config/mips/mips.c (mips_expand_prologue): Likewise.
15051         * config/powerpcspe/powerpcspe.c (rs6000_emit_prologue): Likewise.
15052         * config/rs6000/rs6000.c (rs6000_emit_prologue): Likewise.
15053         * config/sparc/sparc.c (sparc_expand_prologue): Likewise.
15054         (sparc_flat_expand_prologue): Likewise.
15056         * common.opt (-fstack-clash-protection): New option.
15057         * flag-types.h (enum stack_check_type): Note difference between
15058         -fstack-check= and -fstack-clash-protection.
15059         * params.def (PARAM_STACK_CLASH_PROTECTION_GUARD_SIZE): New PARAM.
15060         (PARAM_STACK_CLASH_PROTECTION_PROBE_INTERVAL): Likewise.
15061         * toplev.c (process_options): Issue warnings/errors for cases
15062         not handled with -fstack-clash-protection.
15063         * doc/invoke.texi (-fstack-clash-protection): Document new option.
15064         (-fstack-check): Note additional problem with -fstack-check=generic.
15065         Note that -fstack-check is primarily for Ada and refer users
15066         to -fstack-clash-protection for stack-clash-protection.
15067         Document new params for stack clash protection.
15069 2017-09-19  Uros Bizjak  <ubizjak@gmail.com>
15071         * config/i386/i386.c (ix86_split_long_move): Do not handle
15072         address used for LEA in a special way.
15074 2017-09-19  Segher Boessenkool  <segher@kernel.crashing.org>
15076         * simplify-rtx.c (simplify_binary_operation_1): Fix typo in comment.
15078 2017-09-19  Martin Sebor  <msebor@redhat.com>
15080         PR c/81854
15081         * cgraphunit.c (handle_alias_pairs): Reject aliases between functions
15082         of incompatible types.
15084 2017-09-19  Will Schmidt  <will_schmidt@vnet.ibm.com>
15086         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add handling
15087         for early folding of vector loads (ALTIVEC_BUILTIN_LVX_*).
15088         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
15089         Remove obsoleted code for handling ALTIVEC_BUILTIN_VEC_LD.
15091 2017-09-19  Richard Biener  <rguenther@suse.de>
15093         PR tree-optimization/82244
15094         * tree-vrp.c (remove_range_assertions): Do not propagate
15095         a constant to abnormals but replace the assert with a copy.
15097 2017-09-19  Alexander Monakov  <amonakov@ispras.ru>
15099         PR rtl-optimization/57878
15100         PR rtl-optimization/68988
15101         * lra-assigns.c (reload_pseudo_compare_func): Remove fragmentation
15102         avoidance test involving non_reload_pseudos.  Move frequency test
15103         below the general fragmentation avoidance test.
15105 2017-09-19  Richard Biener  <rguenther@suse.de>
15107         PR tree-optimization/69728
15108         * graphite-sese-to-poly.c (schedule_error): New global.
15109         (add_loop_schedule): Handle empty domain by failing the
15110         schedule.
15111         (build_original_schedule): Handle schedule_error.
15113 2017-09-19  Richard Biener  <rguenther@suse.de>
15115         * graphite-scop-detection.c (scop_detection::can_represent_loop):
15116         Do not iterate to sibling loops but only to siblings of inner
15117         loops.
15119 2017-09-18  Andreas Schwab  <schwab@linux-m68k.org>
15121         PR target/81613
15122         * config/m68k/m68k.md (moveq feeding equality comparison): Check
15123         that the registers are different.
15125 2017-09-18  Uros Bizjak  <ubizjak@gmail.com>
15127         * config/i386/i386.c (fold_builtin_cpu): Add M_AMDFAM17H
15128         to processor_model and "amdfam17h" to arch_names_table.
15129         * doc/extend.texi (__builtin_cpu_is): Document amdfam17h CPU name.
15131 2017-09-18  Jakub Jelinek  <jakub@redhat.com>
15133         PR c/82234
15134         * doc/extend.texi: Add @findex entry for __builtin_shuffle.
15136 2017-09-18  Richard Sandiford  <richard.sandiford@linaro.org>
15137             Alan Hayward  <alan.hayward@arm.com>
15138             David Sherwood  <david.sherwood@arm.com>
15140         * tree-vectorizer.h (vect_slp_analyze_operations): Replace parameters
15141         with a vec_info *.
15142         * tree-vect-loop.c (vect_analyze_loop_operations): Update call
15143         accordingly.
15144         * tree-vect-slp.c (vect_slp_analyze_node_operations): Add vec_info *
15145         parameter.  Set SLP_TREE_NUMBER_OF_VEC_STMTS here rather than in
15146         vect_schedule_slp_instance.
15147         (vect_slp_analyze_operations): Replace parameters with a vec_info *.
15148         Update call to vect_slp_analyze_node_operations.  Simplify return
15149         value.
15150         (vect_slp_analyze_bb_1): Update call accordingly.
15151         (vect_schedule_slp_instance): Remove vectorization_factor parameter.
15152         Don't calculate SLP_TREE_NUMBER_OF_VEC_STMTS here.
15153         (vect_schedule_slp): Update call accordingly.
15155 2017-09-18  Richard Sandiford  <richard.sandiford@linaro.org>
15156             Alan Hayward  <alan.hayward@arm.com>
15157             David Sherwood  <david.sherwood@arm.com>
15159         * tree-vect-data-refs.c (vect_get_smallest_scalar_type): Cope
15160         with types that aren't in fact scalar.
15162 2017-09-18  Richard Sandiford  <richard.sandiford@linaro.org>
15164         * tree-vect-slp.c (vect_record_max_nunits): New function,
15165         split out from...
15166         (vect_build_slp_tree_1): ...here.
15167         (vect_build_slp_tree_2): Call it for phis too.
15169 2017-09-18  Richard Sandiford  <richard.sandiford@linaro.org>
15171         * tree-vect-stmts.c (vectorizable_mask_load_store): Pass mask_vectype
15172         to vect_get_vec_def_for_operand when getting the mask operand.
15174 2017-09-18  Richard Sandiford  <richard.sandiford@linaro.org>
15175             Alan Hayward  <alan.hayward@arm.com>
15176             David Sherwood  <david.sherwood@arm.com>
15178         * tree-vect-loop.c (vectorizable_live_operation): Fix type of
15179         bitstart.
15181 2017-09-18  Richard Sandiford  <richard.sandiford@linaro.org>
15182             Alan Hayward  <alan.hayward@arm.com>
15183             David Sherwood  <david.sherwood@arm.com>
15185         * tree-vect-loop.c (vectorizable_live_operation): Fix element size
15186         calculation for vector booleans.
15188 2017-09-18  Richard Sandiford  <richard.sandiford@linaro.org>
15189             Alan Hayward  <alan.hayward@arm.com>
15190             David Sherwood  <david.sherwood@arm.com>
15192         * tree-vect-stmts.c (can_vectorize_live_stmts): New function,
15193         split out from...
15194         (vect_transform_stmt): ...here.
15195         (vect_analyze_stmt): Use it instead of calling
15196         vectorizable_live_operation directly.
15198 2017-09-18  Cesar Philippidis  <cesar@codesourcery.com>
15200         * omp-offload.c (oacc_xform_loop): Enable SIMD vectorization on
15201         non-SIMT targets in acc vector loops.
15203 2017-09-18  Claudiu Zissulescu  <claziss@synopsys.com>
15205         * configure.ac: Add arc and check if assembler supports gdwarf2.
15206         * configure: Regenerate.
15208 2017-09-18  Richard Biener  <rguenther@suse.de>
15210         PR tree-optimization/82220
15211         * tree-vect-loop.c (vect_estimate_min_profitable_iters): Exclude
15212         epilogue niters from the min_profitable_iters compute.
15214 2017-09-18  Jakub Jelinek  <jakub@redhat.com>
15216         PR target/82145
15217         * config/i386/i386.c (ix86_init_large_pic_reg): Revert 2017-09-01
15218         changes.  Turn CODE_LABEL into NOTE_INSN_DELETED_LABEL immediately.
15219         (ix86_init_pic_reg): Revert 2017-09-01 changes.
15221 2017-09-18  Eric Botcazou  <ebotcazou@adacore.com>
15223         PR target/81361
15224         * dwarf2cfi.c (add_cfis_to_fde): Do not generate DW_CFA_set_loc after
15225         switching to a new text section.
15227 2017-09-18  Richard Biener  <rguenther@suse.de>
15229         * graphite-scop-detection.c (scop_detection::stmt_has_simple_data_ref):
15230         Simplify.
15231         (build_alias_set): Reject aliases with no access function.
15233 2017-09-18  Richard Biener  <rguenther@suse.de>
15235         PR tree-optimization/79622
15236         * graphite-scop-detection.c (build_cross_bb_scalars_def): Properly
15237         handle PHIs.
15238         (build_cross_bb_scalars_use): Likewise.
15240 2017-09-18  Pierre-Marie de Rodat  <derodat@adacore.com>
15242         * cgraph.h (cgraph_thunk_info): Fix a typo in a comment.
15244 2017-09-18  Alan Modra  <amodra@gmail.com>
15246         PR target/81996
15247         * gcc/config/rs6000/rs6000.c (rs6000_return_addr): Use
15248         stack_pointer_rtx for count 0.  Update comments.  Break up
15249         large rtl expression.
15251 2017-09-17  Daniel Santos  <daniel.santos@pobox.com>
15253         * config/i386/i386.c (xlogue_layout::STUB_NAME_MAX_LEN):
15254         Increase to 20 bytes.
15255         (xlogue_layout::s_stub_names): Add an additional size-2 diminsion.
15256         (xlogue_layout::get_stub_name): Modify to select the appropairate sse
15257         or avx version of the stub.
15259 2017-09-17  H.J. Lu  <hongjiu.lu@intel.com>
15261         PR target/82166
15262         * config/i386/i386.c (ix86_finalize_stack_frame_flags): Properly
15263         compute the minimum stack alignment.  Also update preferred stack
15264         boundary for leaf functions.
15266 2017-09-16  Richard Sandiford  <richard.sandiford@linaro.org>
15268         PR tree-optimization/82228
15269         * tree-vect-loop.c (vectorizable_live_operation): Move initialization
15270         of ncopies.
15272 2017-09-16  Chung-Ju Wu  <jasonwucj@gmail.com>
15274         * common/config/nds32/nds32-common.c
15275         (nds32_option_optimization_table): Refine formatting.
15276         (nds32_option_optimization_table): Use -fsched-pressure and
15277         -fomit-frame-pointer for specific optimization level.
15279 2017-09-16  Chung-Ju Wu  <jasonwucj@gmail.com>
15281         * config/nds32/nds32.c: Refine formatting and comments.
15282         * config/nds32/nds32.h: Likewise.
15283         * config/nds32/nds32.md: Likewise.
15284         * config/nds32/nds32-cost.c: Likewise.
15285         * config/nds32/nds32-isr.c: Likewise.
15286         * config/nds32/nds32-md-auxiliary.c: Likewise.
15287         * config/nds32/nds32-multiple.md: Likewise.
15288         * config/nds32/nds32-predicates.c: Likewise.
15290 2017-09-15  Andrew Sutton  <andrew.n.sutton@gmail.com>
15291             Jakub Jelinek  <jakub@redhat.com>
15293         Add support for -std=c++2a.
15294         * doc/cpp.texi (__cplusplus): Document value for -std=c++2a
15295         or -std=gnu+2a.
15296         * doc/invoke.texi: Document -std=c++2a and -std=gnu++2a.
15298 2017-09-15  Steve Ellcey  <sellcey@cavium.com>
15300         PR target/82066
15301         * doc/extend.texi (Common Function Attributes): Add 
15302         references to ARM, AArch64, and S/390 specific attributes.
15303         (Function Specific Option Pragmas): Add AArch64 and S/390
15304         to list of back ends that support the target pragma.
15306 2017-09-15  Nathan Sidwell  <nathan@acm.org>
15308         * doc/standards.texi: Fix C++17 description.  Update URLs for
15309         C++11 & 14.
15311 2017-09-15  Bernd Edlinger  <bernd.edlinger@hotmail.de>
15313         * common.opt (Wcast-align=strict): New warning option.
15314         * doc/invoke.texi: Document -Wcast-align=strict. 
15316 2017-09-15  Pierre-Marie de Rodat  <derodat@adacore.com>
15318         * cgraph.h (cgraph_thunk_info): Add comments.
15319         * cgraph.c (cgraph_node::create_thunk): Adjust comment, make
15320         assert for VIRTUAL_* arguments stricter.
15322 2017-09-15  Jackson Woodruff  <jackson.woodruff@arm.com>
15324         PR tree-optimization/71026
15325         * match.pd: Move RDIV patterns from fold-const.c
15326         * fold-const.c (distribute_real_division): Removed.
15327         (fold_binary_loc): Remove calls to distribute_real_divison.
15329 2017-09-15  Jakub Jelinek  <jakub@redhat.com>
15331         * doc/invoke.texi: Document -std=c++17 and -std=gnu++17 and document
15332         c++1z and gnu++1z as deprecated.  Change other references to
15333         -std=c++1z to -std=c++17 and -std=gnu++1z to -std=gnu++17.
15334         Change -Wc++1z-compat to -Wc++17-compat.
15335         * doc/cpp.texi: Document -std=c++17 defines __cplusplus 201703L.
15336         * dwarf2out.c (highest_c_language): Handle C++17.
15337         (gen_compile_unit_die): Likewise.
15339 2017-09-15  Jakub Jelinek  <jakub@redhat.com>
15341         PR rtl-optimization/82192
15342         * combine.c (make_extraction): Don't look through non-paradoxical
15343         SUBREGs or TRUNCATE if pos + len is or might be bigger than
15344         inner's mode.
15346 2017-09-15  Richard Sandiford  <richard.sandiford@linaro.org>
15347             Alan Hayward  <alan.hayward@arm.com>
15348             David Sherwood  <david.sherwood@arm.com>
15350         * target.def (function_arg_offset): New hook.
15351         * targhooks.h (default_function_arg_offset): Declare.
15352         * targhooks.c (default_function_arg_offset): New function.
15353         * function.c (locate_and_pad_parm): Use
15354         targetm.calls.function_arg_offset instead of FUNCTION_ARG_OFFSET.
15355         * doc/tm.texi.in (FUNCTION_ARG_OFFSET): Replace with...
15356         (TARGET_FUNCTION_ARG_OFFSET): ...this.
15357         * doc/tm.texi: Regenerate.
15358         * config/spu/spu.h (FUNCTION_ARG_OFFSET): Delete.
15359         * config/spu/spu.c (spu_function_arg_offset): New function.
15360         (TARGET_FUNCTION_ARG_OFFSET): Redefine.
15361         * system.h (FUNCTION_ARG_OFFSET): Poison.
15363 2017-09-15  Richard Sandiford  <richard.sandiford@linaro.org>
15364             Alan Hayard  <alan.hayward@arm.com>
15365             David Sherwood  <david.sherwood@arm.com>
15367         * target.def (truly_noop_truncation): New hook.
15368         (mode_rep_extended): Refer to TARGET_TRULY_NOOP_TRUNCATION rather
15369         than TRULY_NOOP_TRUNCATION.
15370         * hooks.h (hook_bool_uint_uint_true): Declare.
15371         * hooks.c (hook_bool_uint_uint_true): New function.
15372         * doc/tm.texi.in (TRULY_NOOP_TRUNCATION): Replace with...
15373         (TARGET_TRULY_NOOP_TRUNCATION): ...this.
15374         * doc/tm.texi: Regenerate.
15375         * combine.c (make_extraction): Refer to TARGET_TRULY_NOOP_TRUNCATION
15376         rather than TRULY_NOOP_TRUNCATION in comments.
15377         (simplify_comparison): Likewise.
15378         (record_truncated_value): Likewise.
15379         * expmed.c (extract_bit_field_1): Likewise.
15380         (extract_split_bit_field): Likewise.
15381         * convert.c (convert_to_integer_1): Use targetm.truly_noop_truncation
15382         instead of TRULY_NOOP_TRUNCATION.
15383         * function.c (assign_parm_setup_block): Likewise.
15384         * machmode.h (TRULY_NOOP_TRUNCATION_MODES_P): Likewise.
15385         * rtlhooks.c: Include target.h.
15386         * config/aarch64/aarch64.h (TRULY_NOOP_TRUNCATION): Delete.
15387         * config/alpha/alpha.h (TRULY_NOOP_TRUNCATION): Delete.
15388         * config/arc/arc.h (TRULY_NOOP_TRUNCATION): Delete.
15389         * config/arm/arm.h (TRULY_NOOP_TRUNCATION): Delete.
15390         * config/avr/avr.h (TRULY_NOOP_TRUNCATION): Delete.
15391         * config/bfin/bfin.h (TRULY_NOOP_TRUNCATION): Delete.
15392         * config/c6x/c6x.h (TRULY_NOOP_TRUNCATION): Delete.
15393         * config/cr16/cr16.h (TRULY_NOOP_TRUNCATION): Delete.
15394         * config/cris/cris.h (TRULY_NOOP_TRUNCATION): Delete.
15395         * config/epiphany/epiphany.h (TRULY_NOOP_TRUNCATION): Delete.
15396         * config/fr30/fr30.h (TRULY_NOOP_TRUNCATION): Delete.
15397         * config/frv/frv.h (TRULY_NOOP_TRUNCATION): Delete.
15398         * config/ft32/ft32.h (TRULY_NOOP_TRUNCATION): Delete.
15399         * config/h8300/h8300.h (TRULY_NOOP_TRUNCATION): Delete.
15400         * config/i386/i386.h (TRULY_NOOP_TRUNCATION): Delete.
15401         * config/ia64/ia64.h (TRULY_NOOP_TRUNCATION): Delete.
15402         * config/iq2000/iq2000.h (TRULY_NOOP_TRUNCATION): Delete.
15403         * config/lm32/lm32.h (TRULY_NOOP_TRUNCATION): Delete.
15404         * config/m32c/m32c.h (TRULY_NOOP_TRUNCATION): Delete.
15405         * config/m32r/m32r.h (TRULY_NOOP_TRUNCATION): Delete.
15406         * config/m68k/m68k.h (TRULY_NOOP_TRUNCATION): Delete.
15407         * config/mcore/mcore.h (TRULY_NOOP_TRUNCATION): Delete.
15408         * config/microblaze/microblaze.h (TRULY_NOOP_TRUNCATION): Delete.
15409         * config/mips/mips.h (TRULY_NOOP_TRUNCATION): Delete.
15410         * config/mips/mips.c (mips_truly_noop_truncation): New function.
15411         (TARGET_TRULY_NOOP_TRUNCATION): Redefine.
15412         * config/mips/mips.md: Refer to TARGET_TRULY_NOOP_TRUNCATION
15413         rather than TRULY_NOOP_TRUNCATION in comments.
15414         * config/mmix/mmix.h (TRULY_NOOP_TRUNCATION): Delete.
15415         * config/mn10300/mn10300.h (TRULY_NOOP_TRUNCATION): Delete.
15416         * config/moxie/moxie.h (TRULY_NOOP_TRUNCATION): Delete.
15417         * config/msp430/msp430.h (TRULY_NOOP_TRUNCATION): Delete.
15418         * config/nds32/nds32.h (TRULY_NOOP_TRUNCATION): Delete.
15419         * config/nios2/nios2.h (TRULY_NOOP_TRUNCATION): Delete.
15420         * config/nvptx/nvptx.h (TRULY_NOOP_TRUNCATION): Delete.
15421         * config/pa/pa.h (TRULY_NOOP_TRUNCATION): Delete.
15422         * config/pdp11/pdp11.h (TRULY_NOOP_TRUNCATION): Delete.
15423         * config/powerpcspe/powerpcspe.h (TRULY_NOOP_TRUNCATION): Delete.
15424         * config/riscv/riscv.h (TRULY_NOOP_TRUNCATION): Delete.
15425         * config/riscv/riscv.md: Refer to TARGET_TRULY_NOOP_TRUNCATION
15426         rather than TRULY_NOOP_TRUNCATION in comments.
15427         * config/rl78/rl78.h (TRULY_NOOP_TRUNCATION): Delete.
15428         * config/rs6000/rs6000.h (TRULY_NOOP_TRUNCATION): Delete.
15429         * config/rx/rx.h (TRULY_NOOP_TRUNCATION): Delete.
15430         * config/s390/s390.h (TRULY_NOOP_TRUNCATION): Delete.
15431         * config/sh/sh.h (MAYBE_BASE_REGISTER_RTX_P): Remove
15432         TRULY_NOOP_TRUNCATION condition.
15433         (MAYBE_INDEX_REGISTER_RTX_P): Likewise.
15434         (TRULY_NOOP_TRUNCATION): Delete.
15435         * config/sparc/sparc.h (TRULY_NOOP_TRUNCATION): Delete.
15436         * config/spu/spu.h (TRULY_NOOP_TRUNCATION): Delete.
15437         * config/spu/spu.c (spu_truly_noop_truncation): New function.
15438         (TARGET_TRULY_NOOP_TRUNCATION): Redefine.
15439         * config/stormy16/stormy16.h (TRULY_NOOP_TRUNCATION): Delete.
15440         * config/tilegx/tilegx.h (TRULY_NOOP_TRUNCATION): Delete.
15441         * config/tilegx/tilegx.c (tilegx_truly_noop_truncation): New fuction.
15442         (TARGET_TRULY_NOOP_TRUNCATION): Redefine.
15443         * config/tilegx/tilegx.md: Refer to TARGET_TRULY_NOOP_TRUNCATION
15444         rather than TRULY_NOOP_TRUNCATION in comments.
15445         * config/tilepro/tilepro.h (TRULY_NOOP_TRUNCATION): Delete.
15446         * config/v850/v850.h (TRULY_NOOP_TRUNCATION): Delete.
15447         * config/vax/vax.h (TRULY_NOOP_TRUNCATION): Delete.
15448         * config/visium/visium.h (TRULY_NOOP_TRUNCATION): Delete.
15449         * config/xtensa/xtensa.h (TRULY_NOOP_TRUNCATION): Delete.
15450         * system.h (TRULY_NOOP_TRUNCATION): Poison.
15452 2017-09-15  Christophe Lyon  <christophe.lyon@linaro.org>
15454         PR target/67591
15455         * config/arm/arm.md (*cmp_and): Add enabled_for_depr_it attribute.
15456         (*cmp_ior): Likewise.
15457         (*ior_scc_scc): Add alternative for enabled_for_depr_it attribute.
15458         (*ior_scc_scc_cmp): Likewise.
15459         (*and_scc_scc): Likewise.
15460         (*and_scc_scc_cmp): Likewise.
15462 2017-09-15  Richard Sandiford  <richard.sandiford@linaro.org>
15463             Alan Hayard  <alan.hayward@arm.com>
15464             David Sherwood  <david.sherwood@arm.com>
15466         * target.def (can_change_mode_class): New hook.
15467         (mode_rep_extended): Refer to it instead of CANNOT_CHANGE_MODE_CLASS.
15468         (hard_regno_nregs): Likewise.
15469         * hooks.h (hook_bool_mode_mode_reg_class_t_true): Declare.
15470         * hooks.c (hook_bool_mode_mode_reg_class_t_true): New function.
15471         * doc/tm.texi.in (CANNOT_CHANGE_MODE_CLASS): Replace with...
15472         (TARGET_CAN_CHANGE_MODE_CLASS): ...this.
15473         (LOAD_EXTEND_OP): Update accordingly.
15474         * doc/tm.texi: Regenerate.
15475         * doc/rtl.texi: Refer to TARGET_CAN_CHANGE_MODE_CLASS instead of
15476         CANNOT_CHANGE_MODE_CLASS.
15477         * hard-reg-set.h (REG_CANNOT_CHANGE_MODE_P): Replace with...
15478         (REG_CAN_CHANGE_MODE_P): ...this new macro.
15479         * combine.c (simplify_set): Update accordingly.
15480         * emit-rtl.c (validate_subreg): Likewise.
15481         * recog.c (general_operand): Likewise.
15482         * regcprop.c (mode_change_ok): Likewise.
15483         * reload1.c (choose_reload_regs): Likewise.
15484         (inherit_piecemeal_p): Likewise.
15485         * rtlanal.c (simplify_subreg_regno): Likewise.
15486         * postreload.c (reload_cse_simplify_set): Use REG_CAN_CHANGE_MODE_P
15487         instead of CANNOT_CHANGE_MODE_CLASS.
15488         (reload_cse_simplify_operands): Likewise.
15489         * reload.c (push_reload): Use targetm.can_change_mode_class
15490         instead of CANNOT_CHANGE_MODE_CLASS.
15491         (push_reload): Likewise.  Also use REG_CAN_CHANGE_MODE_P instead of
15492         REG_CANNOT_CHANGE_MODE_P.
15493         * config/alpha/alpha.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15494         * config/alpha/alpha.c (alpha_can_change_mode_class): New function.
15495         (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15496         * config/arm/arm.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15497         * config/arm/arm.c (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15498         (arm_can_change_mode_class): New function.
15499         * config/arm/neon.md: Refer to TARGET_CAN_CHANGE_MODE_CLASS rather
15500         than CANNOT_CHANGE_MODE_CLASS in comments.
15501         * config/i386/i386.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15502         * config/i386/i386-protos.h (ix86_cannot_change_mode_class): Delete.
15503         * config/i386/i386.c (ix86_cannot_change_mode_class): Replace with...
15504         (ix86_can_change_mode_class): ...this new function, inverting the
15505         sense of the return value.
15506         (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15507         * config/ia64/ia64.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15508         * config/ia64/ia64.c (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15509         (ia64_can_change_mode_class): New function.
15510         * config/m32c/m32c.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15511         * config/m32c/m32c-protos.h (m32c_cannot_change_mode_class): Delete.
15512         * config/m32c/m32c.c (m32c_cannot_change_mode_class): Replace with...
15513         (m32c_can_change_mode_class): ...this new function, inverting the
15514         sense of the return value.
15515         (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15516         * config/mips/mips.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15517         * config/mips/mips-protos.h (mips_cannot_change_mode_class): Delete.
15518         * config/mips/mips.c (mips_cannot_change_mode_class): Replace with...
15519         (mips_can_change_mode_class): ...this new function, inverting the
15520         sense of the return value.
15521         (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15522         * config/msp430/msp430.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15523         * config/msp430/msp430.c (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15524         (msp430_can_change_mode_class): New function.
15525         * config/nvptx/nvptx.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15526         * config/nvptx/nvptx.c (nvptx_can_change_mode_class): New function.
15527         (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15528         * config/pa/pa32-regs.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15529         * config/pa/pa64-regs.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15530         * config/pa/pa-protos.h (pa_cannot_change_mode_class): Delete.
15531         * config/pa/pa.c (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15532         (pa_cannot_change_mode_class): Replace with...
15533         (pa_can_change_mode_class): ...this new function, inverting the
15534         sense of the return value.
15535         (pa_modes_tieable_p): Refer to TARGET_CAN_CHANGE_MODE_CLASS rather
15536         than CANNOT_CHANGE_MODE_CLASS in comments.
15537         * config/pdp11/pdp11.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15538         * config/pdp11/pdp11-protos.h (pdp11_cannot_change_mode_class): Delete.
15539         * config/pdp11/pdp11.c (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15540         (pdp11_cannot_change_mode_class): Replace with...
15541         (pdp11_can_change_mode_class): ...this new function, inverting the
15542         sense of the return value.
15543         * config/powerpcspe/powerpcspe.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15544         * config/powerpcspe/powerpcspe-protos.h
15545         (rs6000_cannot_change_mode_class_ptr): Delete.
15546         * config/powerpcspe/powerpcspe.c
15547         (rs6000_cannot_change_mode_class_ptr): Delete.
15548         (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15549         (rs6000_option_override_internal): Assign to
15550         targetm.can_change_mode_class instead of
15551         rs6000_cannot_change_mode_class_ptr.
15552         (rs6000_cannot_change_mode_class): Replace with...
15553         (rs6000_can_change_mode_class): ...this new function, inverting the
15554         sense of the return value.
15555         (rs6000_debug_cannot_change_mode_class): Replace with...
15556         (rs6000_debug_can_change_mode_class): ...this new function.
15557         * config/riscv/riscv.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15558         * config/riscv/riscv.c (riscv_can_change_mode_class): New function.
15559         (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15560         * config/rs6000/rs6000.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15561         * config/rs6000/rs6000-protos.h (rs6000_cannot_change_mode_class_ptr):
15562         Delete.
15563         * config/rs6000/rs6000.c (rs6000_cannot_change_mode_class_ptr): Delete.
15564         (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15565         (rs6000_option_override_internal): Assign to
15566         targetm.can_change_mode_class instead of
15567         rs6000_cannot_change_mode_class_ptr.
15568         (rs6000_cannot_change_mode_class): Replace with...
15569         (rs6000_can_change_mode_class): ...this new function, inverting the
15570         sense of the return value.
15571         (rs6000_debug_cannot_change_mode_class): Replace with...
15572         (rs6000_debug_can_change_mode_class): ...this new function.
15573         * config/s390/s390.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15574         * config/s390/s390-protos.h (s390_cannot_change_mode_class): Delete.
15575         * config/s390/s390.c (s390_cannot_change_mode_class): Replace with...
15576         (s390_can_change_mode_class): ...this new function, inverting the
15577         sense of the return value.
15578         (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15579         * config/sh/sh.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15580         * config/sh/sh-protos.h (sh_cannot_change_mode_class): Delete.
15581         * config/sh/sh.c (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15582         (sh_cannot_change_mode_class): Replace with...
15583         (sh_can_change_mode_class): ...this new function, inverting the
15584         sense of the return value.
15585         * config/sparc/sparc.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15586         * config/sparc/sparc.c (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15587         (sparc_can_change_mode_class): New function.
15588         * config/spu/spu.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15589         * config/spu/spu.c (spu_can_change_mode_class): New function.
15590         (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15591         * config/visium/visium.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15592         * config/visium/visium.c (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15593         (visium_can_change_mode_class): New function.
15594         * system.h (CANNOT_CHANGE_MODE_CLASS): Poison.
15596 2017-09-15  Richard Biener  <rguenther@suse.de>
15598         PR tree-optimization/82217
15599         * tree-ssa-sccvn.c (visit_phi): Properly handle all VN_TOP
15600         but not undefined case.
15602 2017-09-15  Jakub Jelinek  <jakub@redhat.com>
15604         PR target/82145
15605         * postreload.c (reload_cse_simplify_operands): Skip
15606         NOTE_INSN_DELETED_LABEL similarly to skipping CODE_LABEL.
15608 2017-09-15  Richard Biener  <rguenther@suse.de>
15610         PR tree-optimization/68823
15611         * graphite-scop-detection.c (build_alias_set): If we have a
15612         possible dependence check whether we can handle them by just
15613         looking at the DRs DR_ACCESS_FNs.
15614         (build_scops): If build_alias_set fails, fail the SCOP.
15616 2017-09-14  Michael Meissner  <meissner@linux.vnet.ibm.com>
15618         * config/rs6000/rs6000-builtin.def (BU_FLOAT128_1_HW): New macros
15619         to support float128 built-in functions that require the ISA 3.0
15620         hardware.
15621         (BU_FLOAT128_3_HW): Likewise.
15622         (SQRTF128): Add support for the IEEE 128-bit square root and fma
15623         built-in functions.
15624         (FMAF128): Likewise.
15625         (FMAQ): Likewise.
15626         * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add
15627         support for built-in functions that need the ISA 3.0 IEEE 128-bit
15628         floating point instructions.
15629         (rs6000_invalid_builtin): Likewise.
15630         (rs6000_builtin_mask_names): Likewise.
15631         * config/rs6000/rs6000.h (MASK_FLOAT128_HW): Likewise.
15632         (RS6000_BTM_FLOAT128_HW): Likewise.
15633         (RS6000_BTM_COMMON): Likewise.
15634         * config/rs6000/rs6000.md (fma<mode>4_hw): Add a generator
15635         function.
15636         * doc/extend.texi (RS/6000 built-in functions): Document the
15637         IEEE 128-bit floating point square root and fused multiply-add
15638         built-in functions.
15640 2017-09-14  Pat Haugen  <pthaugen@us.ibm.com>
15642         * config/rs6000/rs6000.c (rs6000_set_up_by_prologue): Make sure the TOC
15643         reg (r2) isn't in the set of registers defined in the prologue.
15645 2017-09-14  Richard Sandiford  <richard.sandiford@linaro.org>
15646             Alan Hayward  <alan.hayward@arm.com>
15647             David Sherwood  <david.sherwood@arm.com>
15649         * tree-vectorizer.h (_loop_vec_info): Add max_vectorization_factor.
15650         (LOOP_VINFO_MAX_VECT_FACTOR): New macro.
15651         (LOOP_VINFO_ORIG_VECT_FACTOR): Replace with...
15652         (LOOP_VINFO_ORIG_MAX_VECT_FACTOR): ...this new macro.
15653         * tree-vect-data-refs.c (vect_analyze_data_ref_dependences): Update
15654         accordingly.
15655         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
15656         max_vectorization_factor.
15657         (vect_analyze_loop_2): Set LOOP_VINFO_MAX_VECT_FACTOR.
15659 2017-09-14  Richard Sandiford  <richard.sandiford@linaro.org>
15660             Alan Hayward  <alan.hayward@arm.com>
15661             David Sherwood  <david.sherwood@arm.com>
15663         * tree-vectorizer.h (vect_min_worthwhile_factor): Delete.
15664         (vect_worthwhile_without_simd_p): Declare.
15665         * tree-vect-loop.c (vect_worthwhile_without_simd_p): New function.
15666         (vectorizable_reduction): Use it.
15667         * tree-vect-stmts.c (vectorizable_shift): Likewise.
15668         (vectorizable_operation): Likewise.
15670 2017-09-14  Richard Sandiford  <richard.sandiford@linaro.org>
15671             Alan Hayward  <alan.hayward@arm.com>
15672             David Sherwood  <david.sherwood@arm.com>
15674         * tree-vectorizer.h (vect_get_num_copies): New function.
15675         * tree-vect-data-refs.c (vect_get_data_access_cost): Use it.
15676         * tree-vect-loop.c (vectorizable_reduction): Likewise.
15677         (vectorizable_induction): Likewise.
15678         (vectorizable_live_operation): Likewise.
15679         * tree-vect-stmts.c (vectorizable_mask_load_store): Likewise.
15680         (vectorizable_bswap): Likewise.
15681         (vectorizable_call): Likewise.
15682         (vectorizable_conversion): Likewise.
15683         (vectorizable_assignment): Likewise.
15684         (vectorizable_shift): Likewise.
15685         (vectorizable_operation): Likewise.
15686         (vectorizable_store): Likewise.
15687         (vectorizable_load): Likewise.
15688         (vectorizable_condition): Likewise.
15689         (vectorizable_comparison): Likewise.
15690         (vect_analyze_stmt): Pass the slp node to vectorizable_live_operation.
15692 2017-09-14  Richard Sandiford  <richard.sandiford@linaro.org>
15693             Alan Hayward  <alan.hayward@arm.com>
15694             David Sherwood  <david.sherwood@arm.com>
15696         * tree-vect-loop.c (vectorizable_induction): Use gimple_build instead
15697         of vect_init_vector.
15699 2017-09-14  Richard Sandiford  <richard.sandiford@linaro.org>
15700             Alan Hayward  <alan.hayward@arm.com>
15701             David Sherwood  <david.sherwood@arm.com>
15703         * gimple-fold.h (gimple_build_vector_from_val): Declare, and provide
15704         an inline wrapper that provides a location.
15705         (gimple_build_vector): Likewise.
15706         * gimple-fold.c (gimple_build_vector_from_val): New function.
15707         (gimple_build_vector): Likewise.
15708         * tree-vect-loop.c (get_initial_def_for_reduction): Use the new
15709         functions to build the initial value.  Always return a gimple value.
15710         (get_initial_defs_for_reduction): Likewise.  Only compute
15711         neutral_vec once.
15712         (vect_create_epilog_for_reduction): Don't call force_gimple_operand or
15713         vect_init_vector on the results from get_initial_def(s)_for_reduction.
15714         (vectorizable_induction): Use gimple_build_vector rather than
15715         vect_init_vector.
15717 2017-09-14  Richard Sandiford  <richard.sandiford@linaro.org>
15718             Alan Hayward  <alan.hayward@arm.com>
15719             David Sherwood  <david.sherwood@arm.com>
15721         * target.h (vec_perm_indices): New typedef.
15722         (auto_vec_perm_indices): Likewise.
15723         * optabs-query.h: Include target.h
15724         (can_vec_perm_p): Take a vec_perm_indices *.
15725         * optabs-query.c (can_vec_perm_p): Likewise.
15726         (can_mult_highpart_p): Update accordingly.  Use auto_vec_perm_indices.
15727         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
15728         * tree-vect-generic.c (lower_vec_perm): Likewise.
15729         * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
15730         (vect_grouped_load_supported): Likewise.
15731         (vect_shift_permute_load_chain): Likewise.
15732         (vect_permute_store_chain): Use auto_vec_perm_indices.
15733         (vect_permute_load_chain): Likewise.
15734         * fold-const.c (fold_vec_perm): Take vec_perm_indices.
15735         (fold_ternary_loc): Update accordingly.  Use auto_vec_perm_indices.
15736         Update uses of can_vec_perm_p.
15737         * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Replace the
15738         mode with a number of elements.  Take a vec_perm_indices *.
15739         (vect_create_epilog_for_reduction): Update accordingly.
15740         Use auto_vec_perm_indices.
15741         (have_whole_vector_shift): Likewise.  Update call to can_vec_perm_p.
15742         * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
15743         (vect_transform_slp_perm_load): Likewise.
15744         (vect_schedule_slp_instance): Use auto_vec_perm_indices.
15745         * tree-vectorizer.h (vect_gen_perm_mask_any): Take a vec_perm_indices.
15746         (vect_gen_perm_mask_checked): Likewise.
15747         * tree-vect-stmts.c (vect_gen_perm_mask_any): Take a vec_perm_indices.
15748         (vect_gen_perm_mask_checked): Likewise.
15749         (vectorizable_mask_load_store): Use auto_vec_perm_indices.
15750         (vectorizable_store): Likewise.
15751         (vectorizable_load): Likewise.
15752         (perm_mask_for_reverse): Likewise.  Update call to can_vec_perm_p.
15753         (vectorizable_bswap): Likewise.
15755 2017-09-14  Richard Sandiford  <richard.sandiford@linaro.org>
15756             Alan Hayward  <alan.hayward@arm.com>
15757             David Sherwood  <david.sherwood@arm.com>
15759         * tree.h (build_vector): Take a vec<tree> instead of a tree *.
15760         * tree.c (build_vector): Likewise.
15761         (build_vector_from_ctor): Update accordingly.
15762         (build_vector_from_val): Likewise.
15763         * gimple-fold.c (gimple_fold_stmt_to_constant_1): Likewise.
15764         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
15765         * tree-vect-generic.c (add_rshift): Likewise.
15766         (expand_vector_divmod): Likewise.
15767         (optimize_vector_constructor): Likewise.
15768         * tree-vect-slp.c (vect_get_constant_vectors): Likewise.
15769         (vect_transform_slp_perm_load): Likewise.
15770         (vect_schedule_slp_instance): Likewise.
15771         * tree-vect-stmts.c (vectorizable_bswap): Likewise.
15772         (vectorizable_call): Likewise.
15773         (vect_gen_perm_mask_any): Likewise.  Add elements in order.
15774         * expmed.c (make_tree): Likewise.
15775         * fold-const.c (fold_negate_expr_1): Use auto_vec<tree> when building
15776         a vector passed to build_vector.
15777         (fold_convert_const): Likewise.
15778         (exact_inverse): Likewise.
15779         (fold_ternary_loc): Likewise.
15780         (fold_relational_const): Likewise.
15781         (const_binop): Likewise.  Use VECTOR_CST_ELT directly when operating
15782         on VECTOR_CSTs, rather than going through vec_cst_ctor_to_array.
15783         (const_unop): Likewise.  Store the reduction accumulator in a
15784         variable rather than an array.
15785         (vec_cst_ctor_to_array): Take the number of elements as a parameter.
15786         (fold_vec_perm): Update calls accordingly.  Use auto_vec<tree> for
15787         the new vector, rather than constructing it after the input arrays.
15788         (native_interpret_vector): Use auto_vec<tree> when building
15789         a vector passed to build_vector.  Add elements in order.
15790         * tree-vect-loop.c (get_initial_defs_for_reduction): Use
15791         auto_vec<tree> when building a vector passed to build_vector.
15792         (vect_create_epilog_for_reduction): Likewise.
15793         (vectorizable_induction): Likewise.
15794         (get_initial_def_for_reduction): Likewise.  Fix indentation of
15795         case statements.
15796         * config/sparc/sparc.c (sparc_handle_vis_mul8x16): Change n_elts
15797         to a vec<tree> *.
15798         (sparc_fold_builtin): Use auto_vec<tree> when building a vector
15799         passed to build_vector.
15801 2017-09-14  Richard Sandiford  <richard.sandiford@linaro.org>
15802             Alan Hayward  <alan.hayward@arm.com>
15803             David Sherwood  <david.sherwood@arm.com>
15805         * tree-core.h (tree_base::u): Add an "nelts" field.
15806         (tree_vector): Use VECTOR_CST_NELTS as the length.
15807         * tree.c (tree_size): Likewise.
15808         (make_vector): Initialize VECTOR_CST_NELTS.
15809         * tree.h (VECTOR_CST_NELTS): Use the u.nelts field.
15810         * cfgexpand.c (expand_debug_expr): Use VECTOR_CST_NELTS instead of
15811         TYPE_VECTOR_SUBPARTS.
15812         * expr.c (const_vector_mask_from_tree): Consistently use "units"
15813         as the number of units, setting it from VECTOR_CST_NELTS.
15814         (const_vector_from_tree): Likewise.
15815         * fold-const.c (negate_expr_p): Use VECTOR_CST_NELTS instead of
15816         TYPE_VECTOR_SUBPARTS for the number of elements in a VECTOR_CST.
15817         (fold_negate_expr_1): Likewise.
15818         (fold_convert_const): Likewise.
15819         (const_binop): Likewise.  Differentiate the number of output and
15820         input elements.
15821         (const_unop): Likewise.
15822         (fold_ternary_loc): Use VECTOR_CST_NELTS for the number of elements
15823         in a VECTOR_CST, asserting that it is the same as TYPE_VECTOR_SUBPARTS
15824         in cases that did the opposite.
15826 2017-09-14  Richard Biener  <rguenther@suse.de>
15828         * tree-ssa-sccvn.c (visit_phi): Merge undefined values similar
15829         to VN_TOP.
15831 2017-09-14  Eric Botcazou  <ebotcazou@adacore.com>
15833         * dwarf2out.c (dwarf2out_source_line): Remove superfluous test.
15835 2017-09-14  Jakub Jelinek  <jakub@redhat.com>
15837         PR target/81325
15838         * cfgbuild.c (find_bb_boundaries): Ignore debug insns in decisions
15839         if and where to split a bb, except for splitting before debug insn
15840         sequences followed by non-label real insn.  Delete debug insns
15841         in between basic blocks.
15843         * combine.c (make_compound_operation_int): Formatting fixes.
15845         * config/alpha/elf.h (LINK_EH_SPEC): Add -static-pie support.
15846         * config/alpha/linux.h (LINK_GCC_C_SEQUENCE_SPEC): Likewise.
15847         * config/netbsd.h (LINK_EH_SPEC): Likewise.
15848         * config/sol2.h (LINK_EH_SPEC): Likewise.
15849         * config/arm/uclinux-elf.h (LINK_GCC_C_SEQUENCE_SPEC): Likewise.
15850         * config/s390/linux.h (LINK_SPEC): Likewise.
15851         * config/freebsd.h (LINK_EH_SPEC): Likewise.
15852         * config/openbsd.h (LINK_EH_SPEC): Likewise.
15853         * config/lm32/uclinux-elf.h (LINK_GCC_C_SEQUENCE_SPEC): Likewise.
15854         * config/aarch64/aarch64-linux.h (LINUX_TARGET_LINK_SPEC): Likewise.
15855         * config/powerpcspe/sysv4.h (LINK_EH_SPEC): Likewise.
15856         * config/bfin/linux.h (LINK_GCC_C_SEQUENCE_SPEC): Likewise.
15857         * config/i386/gnu-user64.h (GNU_USER_TARGET_LINK_SPEC): Fix a typo.
15858         * config/i386/gnu-user.h (GNU_USER_TARGET_LINK_SPEC): Formatting fix.
15860 2017-09-13  Jakub Jelinek  <jakub@redhat.com>
15862         * config/rs6000/sysv4.h (STARTFILE_LINUX_SPEC): Add -static-pie
15863         support.
15864         (ENDFILE_LINUX_SPEC): Likewise.
15865         (LINK_EH_SPEC): Likewise.
15866         * config/rs6000/linux64.h (LINK_SHLIB_SPEC): Likewise.
15867         (LINK_OS_LINUX_SPEC32): Likewise.
15868         (LINK_OS_LINUX_SPEC64): Likewise.
15869         * config/rs6000/linux.h (LINK_SHLIB_SPEC): Likewise.
15870         (LINK_OS_LINUX_SPEC): Likewise.
15872 2017-09-13  Martin Liska  <mliska@suse.cz>
15874         PR middle-end/82154
15875         * stmt.c (expand_sjlj_dispatch_table): Use CASE_LOW when
15876         CASE_HIGH is NULL_TREE.
15878 2017-09-13  Richard Sandiford  <richard.sandiford@linaro.org>
15879             Alan Hayward  <alan.hayward@arm.com>
15880             David Sherwood  <david.sherwood@arm.com>
15882         * target.def (secondary_memory_needed): New hook.
15883         (secondary_reload): Refer to TARGET_SECONDARY_MEMORY_NEEDED
15884         instead of SECONDARY_MEMORY_NEEDED.
15885         (secondary_memory_needed_mode): Likewise.
15886         * hooks.h (hook_bool_mode_reg_class_t_reg_class_t_false): Declare.
15887         * hooks.c (hook_bool_mode_reg_class_t_reg_class_t_false): New function.
15888         * doc/tm.texi.in (SECONDARY_MEMORY_NEEDED): Replace with...
15889         (TARGET_SECONDARY_MEMORY_NEEDED): ...this.
15890         (SECONDARY_MEMORY_NEEDED_RTX): Update reference accordingly.
15891         * doc/tm.texi: Regenerate.
15892         * config/alpha/alpha.h (SECONDARY_MEMORY_NEEDED): Delete.
15893         * config/alpha/alpha.c (alpha_secondary_memory_needed): New function.
15894         (TARGET_SECONDARY_MEMORY_NEEDED): Redefine.
15895         * config/i386/i386.h (SECONDARY_MEMORY_NEEDED): Delete.
15896         * config/i386/i386-protos.h (ix86_secondary_memory_needed): Delete.
15897         * config/i386/i386.c (inline_secondary_memory_needed): Put the
15898         mode argument first and change the reg_class arguments to reg_class_t.
15899         (ix86_secondary_memory_needed): Likewise.  Remove the strict parameter.
15900         Make static.  Update the call to inline_secondary_memory_needed.
15901         (ix86_register_move_cost): Update the call to
15902         inline_secondary_memory_needed.
15903         (TARGET_SECONDARY_MEMORY_NEEDED): Redefine.
15904         * config/ia64/ia64.h (SECONDARY_MEMORY_NEEDED): Delete commented-out
15905         definition.
15906         * config/ia64/ia64.c (spill_xfmode_rfmode_operand): Refer to
15907         TARGET_SECONDARY_MEMORY_NEEDED rather than SECONDARY_MEMORY_NEEDED
15908         in comment.
15909         * config/mips/mips.h (SECONDARY_MEMORY_NEEDED): Delete.
15910         * config/mips/mips-protos.h (mips_secondary_memory_needed): Delete.
15911         * config/mips/mips.c (mips_secondary_memory_needed): Make static
15912         and match hook interface.  Add comment from mips.h.
15913         (TARGET_SECONDARY_MEMORY_NEEDED): Redefine.
15914         * config/mmix/mmix.md (truncdfsf2): Refer to
15915         TARGET_SECONDARY_MEMORY_NEEDED rather than SECONDARY_MEMORY_NEEDED
15916         in comment.
15917         * config/pa/pa-64.h (SECONDARY_MEMORY_NEEDED): Rename to...
15918         (PA_SECONDARY_MEMORY_NEEDED): ...this, and put the mode argument first.
15919         * config/pa/pa.c (TARGET_SECONDARY_MEMORY_NEEDED): Redefine.
15920         (pa_secondary_memory_needed): New function.
15921         * config/pdp11/pdp11.h (SECONDARY_MEMORY_NEEDED): Delete.
15922         * config/pdp11/pdp11-protos.h (pdp11_secondary_memory_needed): Delete.
15923         * config/pdp11/pdp11.c (TARGET_SECONDARY_MEMORY_NEEDED): Redefine.
15924         (pdp11_secondary_memory_needed): Make static and match hook interface.
15925         * config/powerpcspe/powerpcspe.h (SECONDARY_MEMORY_NEEDED): Delete.
15926         * config/powerpcspe/powerpcspe-protos.h
15927         (rs6000_secondary_memory_needed_ptr): Delete.
15928         * config/powerpcspe/powerpcspe.c (rs6000_secondary_memory_needed_ptr):
15929         Delete.
15930         (TARGET_SECONDARY_MEMORY_NEEDED): Redefine.
15931         (rs6000_option_override_internal): Assign to
15932         targetm.secondary_memory_needed rather than
15933         rs6000_secondary_memory_needed_ptr.
15934         (rs6000_secondary_memory_needed): Match hook interface.
15935         (rs6000_debug_secondary_memory_needed): Likewise.
15936         * config/riscv/riscv.h (SECONDARY_MEMORY_NEEDED): Delete.
15937         * config/riscv/riscv.c (riscv_secondary_memory_needed): New function.
15938         (riscv_register_move_cost): Use it instead of SECONDARY_MEMORY_NEEDED.
15939         (TARGET_SECONDARY_MEMORY_NEEDED): Redefine.
15940         * config/rs6000/rs6000.h (SECONDARY_MEMORY_NEEDED): Delete.
15941         * config/rs6000/rs6000-protos.h (rs6000_secondary_memory_needed_ptr):
15942         Delete.
15943         * config/rs6000/rs6000.c (rs6000_secondary_memory_needed_ptr): Delete.
15944         (TARGET_SECONDARY_MEMORY_NEEDED): Redefine.
15945         (rs6000_option_override_internal): Assign to
15946         targetm.secondary_memory_needed rather than
15947         rs6000_secondary_memory_needed_ptr.
15948         (rs6000_secondary_memory_needed): Match hook interface.
15949         (rs6000_debug_secondary_memory_needed): Likewise.
15950         * config/s390/s390.h (SECONDARY_MEMORY_NEEDED): Delete.
15951         * config/s390/s390.c (s390_secondary_memory_needed): New function.
15952         (TARGET_SECONDARY_MEMORY_NEEDED): Redefine.
15953         * config/sparc/sparc.h (SECONDARY_MEMORY_NEEDED): Delete.
15954         * config/sparc/sparc.c (TARGET_SECONDARY_MEMORY_NEEDED): Redefine.
15955         (sparc_secondary_memory_needed): New function.
15956         * lra-constraints.c (check_and_process_move): Refer to
15957         TARGET_SECONDARY_MEMORY_NEEDED rather than SECONDARY_MEMORY_NEEDED
15958         in comment.
15959         (curr_insn_transform): Likewise.
15960         (process_alt_operands): Use targetm.secondary_memory_needed
15961         instead of TARGET_SECONDARY_MEMORY_NEEDED.
15962         (check_secondary_memory_needed_p): Likewise.
15963         (choose_split_class): Likewise.
15964         * reload.c: Unconditionally include code that was previously
15965         conditional on SECONDARY_MEMORY_NEEDED.
15966         (push_secondary_reload): Use targetm.secondary_memory_needed
15967         instead of TARGET_SECONDARY_MEMORY_NEEDED.
15968         (push_reload): Likewise.
15969         * reload1.c: Unconditionally include code that was previously
15970         conditional on SECONDARY_MEMORY_NEEDED.
15971         (choose_reload_regs): Use targetm.secondary_memory_needed
15972         instead of TARGET_SECONDARY_MEMORY_NEEDED.
15973         (gen_reload): Likewise.
15974         * system.h (SECONDARY_MEMORY_NEEDED): Poison.
15976 2017-09-13  Richard Sandiford  <richard.sandiford@linaro.org>
15977             Alan Hayward  <alan.hayward@arm.com>
15978             David Sherwood  <david.sherwood@arm.com>
15980         * target.def (secondary_memory_needed_mode): New hook:
15981         * targhooks.c (default_secondary_memory_needed_mode): Declare.
15982         * targhooks.h (default_secondary_memory_needed_mode): New function.
15983         * doc/tm.texi.in (SECONDARY_MEMORY_NEEDED_MODE): Replace with...
15984         (TARGET_SECONDARY_MEMORY_NEEDED_MODE): ...this.
15985         * doc/tm.texi: Regenerate.
15986         * lra-constraints.c (check_and_process_move): Use
15987         targetm.secondary_memory_needed_mode instead of
15988         TARGET_SECONDARY_MEMORY_NEEDED_MODE.
15989         (curr_insn_transform): Likewise.
15990         * reload.c (get_secondary_mem): Likewise.
15991         * config/alpha/alpha.h (SECONDARY_MEMORY_NEEDED_MODE): Delete.
15992         * config/alpha/alpha.c (alpha_secondary_memory_needed_mode): New
15993         function.
15994         (TARGET_SECONDARY_MEMORY_NEEDED_MODE): Redefine.
15995         * config/i386/i386.h (SECONDARY_MEMORY_NEEDED_MODE): Delete.
15996         * config/i386/i386.c (ix86_secondary_memory_needed_mode): New function.
15997         (TARGET_SECONDARY_MEMORY_NEEDED_MODE): Redefine.
15998         * config/powerpcspe/powerpcspe.h (SECONDARY_MEMORY_NEEDED_MODE):
15999         Delete.
16000         * config/powerpcspe/powerpcspe-protos.h
16001         (rs6000_secondary_memory_needed_mode): Delete.
16002         * config/powerpcspe/powerpcspe.c
16003         (TARGET_SECONDARY_MEMORY_NEEDED_MODE): Redefine.
16004         (rs6000_secondary_memory_needed_mode): Make static.
16005         * config/rs6000/rs6000.h (SECONDARY_MEMORY_NEEDED_MODE): Delete.
16006         * config/rs6000/rs6000-protos.h (rs6000_secondary_memory_needed_mode):
16007         Delete.
16008         * config/rs6000/rs6000.c (TARGET_SECONDARY_MEMORY_NEEDED_MODE):
16009         Redefine.
16010         (rs6000_secondary_memory_needed_mode): Make static.
16011         * config/s390/s390.h (SECONDARY_MEMORY_NEEDED_MODE): Delete.
16012         * config/s390/s390.c (s390_secondary_memory_needed_mode): New function.
16013         (TARGET_SECONDARY_MEMORY_NEEDED_MODE): Redefine.
16014         * config/sparc/sparc.h (SECONDARY_MEMORY_NEEDED_MODE): Delete.
16015         * config/sparc/sparc.c (TARGET_SECONDARY_MEMORY_NEEDED_MODE):
16016         Redefine.
16017         (sparc_secondary_memory_needed_mode): New function.
16018         * system.h (TARGET_SECONDARY_MEMORY_NEEDED_MODE): Poison.
16020 2017-09-13  Jackson Woodruff  <jackson.woodruff@arm.com>
16022         * config/aarch64/constraints.md (Umq): New constraint.
16023         * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>):
16024         Change to use Umq.
16025         (mov<mode>): Update condition.
16027 2017-09-13  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
16029         * gimple-ssa-store-merging.c (sort_by_bitpos): Compare store order
16030         when bitposition is the same.
16032 2017-09-13  Richard Biener  <rguenther@suse.de>
16034         * dwarf2out.c (output_die_symbol): Remove.
16035         (output_die): Do not output a DIEs symbol.
16037 2017-09-13  Richard Biener  <rguenther@suse.de>
16039         PR middle-end/82128
16040         * gimple-fold.c (gimple_fold_call): Update SSA name in-place to
16041         default-def to avoid breaking iterator update with the weird
16042         interaction with cgraph_update_edges_for_call_stmt_node.
16044 2017-09-13  Richard Biener  <rguenther@suse.de>
16046         * tree-cfg.c (verify_gimple_assign_binary): Add verification
16047         for WIDEN_SUM_EXPR, VEC_WIDEN_MULT_{HI,LO,EVEN,ODD}_EXPR,
16048         VEC_PACK_{TRUNC,SAT,FIX_TRUNC}_EXPR.
16049         (verify_gimple_assign_ternary): Add verification for DOT_PROD_EXPR.
16051 2017-09-13  Kugan Vivekanandarajah  <kuganv@linaro.org>
16053         * config/aarch64/aarch64.c (aarch64_override_options_after_change_1):
16054         Disable pc relative literal load irrespective of
16055         TARGET_FIX_ERR_A53_84341 for default.
16057 2017-09-12  Eric Botcazou  <ebotcazou@adacore.com>
16059         * config/sparc/sparc.c (output_return): Output the source location of
16060         the insn in the delay slot, if any.
16061         (output_sibcall): Likewise.
16063 2017-09-12  H.J. Lu  <hongjiu.lu@intel.com>
16065         PR driver/81498
16066         * common.opt (-static-pie): New alias.
16067         (shared): Negate static-pie.
16068         (-no-pie): Update help text.
16069         (-pie): Likewise.
16070         (static-pie): New option.
16071         * config/gnu-user.h (GNU_USER_TARGET_STARTFILE_SPEC): Add
16072         -static-pie support.
16073         (GNU_USER_TARGET_ENDFILE_SPEC): Likewise.
16074         (LINK_EH_SPEC): Likewise.
16075         (LINK_GCC_C_SEQUENCE_SPEC): Likewise.
16076         * config/i386/gnu-user.h (GNU_USER_TARGET_LINK_SPEC): Likewise.
16077         * config/i386/gnu-user64.h (GNU_USER_TARGET_LINK_SPEC): Likewise.
16078         * gcc.c (LINK_COMMAND_SPEC): Likewise.
16079         (init_gcc_specs): Likewise.
16080         (init_spec): Likewise.
16081         (display_help): Update help message for -pie.
16082         * doc/invoke.texi: Update -pie, -no-pie and -static.  Document
16083         -static-pie.
16085 2017-09-12  Wilco Dijkstra  <wdijkstr@arm.com>
16087         * config/aarch64/aarch64.md (movsi_aarch64): Remove all '*'.
16088         (movdi_aarch64): Likewise.
16089         (movti_aarch64): Likewise.
16091 2017-09-12 Simon Wright <simon@pushface.org>
16093         PR target/80204
16094         * config/darwin-driver.c (darwin_find_version_from_kernel): Eliminate
16095         calculation of the minor version, always output as 0.
16097 2017-09-12  Jakub Jelinek  <jakub@redhat.com>
16099         PR target/82112
16100         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): For
16101         ALTIVEC_BUILTIN_VEC_LD if arg1 has array type call default_conversion
16102         on it early, rather than manual conversion late.  For
16103         ALTIVEC_BUILTIN_VEC_ST if arg2 has array type call default_conversion
16104         instead of performing manual conversion.
16106 2017-09-12  Carl Love  <cel@us.ibm.com>
16108         * config/rs6000/altivec.md (vec_widen_umult_even_v4si,
16109         vec_widen_smult_even_v4si): Add define expands for vmuleuw, vmulesw,
16110         vmulouw, vmulosw.
16111         * config/rs6000/rs6000-builtin.def (VMLEUW, VMULESW, VMULOUW,
16112         VMULOSW): Add definitions.
16113         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
16114         ALTIVEC_BUILTIN_VMULESW, ALTIVEC_BUILTIN_VMULEUW,
16115         ALTIVEC_BUILTIN_VMULOSW, ALTIVEC_BUILTIN_VMULOUW entries.
16116         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin,
16117         builtin_function_type): Add ALTIVEC_BUILTIN_* case statements.
16119 2017-09-12  James Greenhalgh  <james.greenhalgh@arm.com>
16121         * config/aarch64/aarch64.md (movdi_aarch64): Set load/store
16122         types correctly.
16123         (movti_aarch64): Likewise.
16124         (movdf_aarch64): Likewise.
16125         (movtf_aarch64): Likewise.
16126         (load_pairdi): Likewise.
16127         (store_pairdi): Likewise.
16128         (load_pairdf): Likewise.
16129         (store_pairdf): Likewise.
16130         (loadwb_pair<GPI:mode>_<P:mode>): Likewise.
16131         (storewb_pair<GPI:mode>_<P:mode>): Likewise.
16132         (ldr_got_small_<mode>): Likewise.
16133         (ldr_got_small_28k_<mode>): Likewise.
16134         (ldr_got_tiny): Likewise.
16135         * config/aarch64/iterators.md (ldst_sz): New.
16136         (ldpstp_sz): Likewise.
16137         * config/aarch64/thunderx.md (thunderx_storepair): Split store_8
16138         to store_16.
16139         (thunderx_load): Split load_8 to load_16.
16140         * config/aarch64/thunderx2t99.md (thunderx2t99_loadpair): Split
16141         load_8 to load_16.
16142         (thunderx2t99_storepair_basic): Split store_8 to store_16.
16143         * config/arm/xgene1.md (xgene1_load_pair): Split load_8 to load_16.
16144         (xgene1_store_pair): Split store_8 to store_16.
16145         * config/aarch64/falkor.md (falkor_ld_3_ld): Split load_8 to load_16.
16146         (falkor_st_0_st_sd): Split store_8 to store_16.
16148 2017-09-12  James Greenhalgh  <james.greenhalgh@arm.com>
16150         * config/arm/types.md (type): Rename load1/2/3/4 to load_4/8/12/16
16151         and store1/2/3/4 to store_4/8/12/16.
16152         * config/aarch64/aarch64.md: Update for rename.
16153         * config/arm/arm.md: Likewise.: Likewise.
16154         * config/arm/arm.c: Likewise.
16155         * config/arm/thumb1.md: Likewise.
16156         * config/arm/thumb2.md: Likewise.
16157         * config/arm/vfp.md: Likewise.
16158         * config/arm/arm-generic.md: Likewise.
16159         * config/arm/arm1020e.md: Likewise.
16160         * config/arm/arm1026ejs.md: Likewise.
16161         * config/arm/arm1136jfs.md: Likewise.
16162         * config/arm/arm926ejs.md: Likewise.
16163         * config/arm/cortex-a15.md: Likewise.
16164         * config/arm/cortex-a17.md: Likewise.
16165         * config/arm/cortex-a5.md: Likewise.
16166         * config/arm/cortex-a53.md: Likewise.
16167         * config/arm/cortex-a57.md: Likewise.
16168         * config/arm/cortex-a7.md: Likewise.
16169         * config/arm/cortex-a8.md: Likewise.
16170         * config/arm/cortex-a9.md: Likewise.
16171         * config/arm/cortex-m4.md: Likewise.
16172         * config/arm/cortex-m7.md: Likewise.
16173         * config/arm/cortex-r4.md: Likewise.
16174         * config/arm/exynos-m1.md: Likewise.
16175         * config/arm/fa526.md: Likewise.
16176         * config/arm/fa606te.md: Likewise.
16177         * config/arm/fa626te.md: Likewise.
16178         * config/arm/fa726te.md: Likewise.
16179         * config/arm/fmp626.md: Likewise.
16180         * config/arm/iwmmxt.md: Likewise.
16181         * config/arm/ldmstm.md: Likewise.
16182         * config/arm/marvell-pj4.md: Likewise.
16183         * config/arm/xgene1.md: Likewise.
16184         * config/aarch64/thunderx.md: Likewise.
16185         * config/aarch64/thunderx2t99.md: Likewise.
16186         * config/aarch64/falkor.md: Likewise.
16188 2017-09-12  Martin Liska  <mliska@suse.cz>
16190         * attribs.c (private_lookup_attribute): New function.
16191         * attribs.h (private_lookup_attribute): Declared here.
16192         (lookup_attribute): Called from this place.
16194 2017-09-12  Richard Biener  <rguenther@suse.de>
16196         PR tree-optimization/82157
16197         * tree-ssa-pre.c (remove_dead_inserted_code): Do not remove
16198         stmts with side-effects.
16200 2017-09-12  Richard Sandiford  <richard.sandiford@linaro.org>
16201             Alan Hayward  <alan.hayward@arm.com>
16202             David Sherwood <david.sherwood@arm.com>
16204         * target.def (hard_regno_nregs): New hook.
16205         (class_max_nregs): Refer to it instead of HARD_REGNO_NREGS.
16206         * targhooks.h (default_hard_regno_nregs): Declare.
16207         * targhooks.c (default_hard_regno_nregs): New function.
16208         * doc/tm.texi.in (HARD_REGNO_NREGS): Replace with...
16209         (TARGET_HARD_REGNO_NREGS): ...this hook.
16210         (HARD_REGNO_NREGS_HAS_PADDING): Update accordingly.
16211         (CLASS_MAX_NREGS): Likewise.
16212         * doc/tm.texi: Regenerate.
16213         * reginfo.c (init_reg_modes_target): Use targetm.hard_regno_nregs
16214         instead of HARD_REGNO_NREGS.
16215         * rtl.h (REG_NREGS): Refer to TARGET_HARD_REGNO_NREGS rather than
16216         HARD_REGNO_NREGS in the comment.
16217         * config/aarch64/aarch64.h (HARD_REGNO_NREGS): Delete.
16218         * config/aarch64/aarch64-protos.h (aarch64_hard_regno_nregs): Delete.
16219         * config/aarch64/aarch64.c (aarch64_hard_regno_nregs): Make static.
16220         Return an unsigned int.
16221         (TARGET_HARD_REGNO_NREGS): Redefine.
16222         * config/alpha/alpha.h (HARD_REGNO_NREGS): Delete.
16223         * config/arc/arc.h (HARD_REGNO_NREGS): Delete.
16224         * config/arc/arc.c (TARGET_HARD_REGNO_NREGS): Redefine.
16225         (arc_hard_regno_nregs): New function.
16226         * config/arm/arm.h (HARD_REGNO_NREGS): Delete.
16227         * config/arm/arm.c (TARGET_HARD_REGNO_NREGS): Redefine.
16228         (arm_hard_regno_nregs): New function.
16229         * config/avr/avr.h (HARD_REGNO_NREGS): Delete.
16230         * config/bfin/bfin.h (HARD_REGNO_NREGS): Delete.
16231         * config/bfin/bfin.c (bfin_hard_regno_nregs): New function.
16232         (TARGET_HARD_REGNO_NREGS): Redefine.
16233         * config/c6x/c6x.h (HARD_REGNO_NREGS): Delete.
16234         * config/cr16/cr16.h (LONG_REG_P): Use targetm.hard_regno_nregs.
16235         (HARD_REGNO_NREGS): Delete.
16236         * config/cr16/cr16.c (TARGET_HARD_REGNO_NREGS): Redefine.
16237         (cr16_hard_regno_nregs): New function.
16238         (cr16_memory_move_cost): Use it instead of HARD_REGNO_NREGS.
16239         * config/cris/cris.h (HARD_REGNO_NREGS): Delete.
16240         * config/cris/cris.c (TARGET_HARD_REGNO_NREGS): Redefine.
16241         (cris_hard_regno_nregs): New function.
16242         * config/epiphany/epiphany.h (HARD_REGNO_NREGS): Delete.
16243         * config/fr30/fr30.h (HARD_REGNO_NREGS): Delete.
16244         (CLASS_MAX_NREGS): Use targetm.hard_regno_nregs.
16245         * config/frv/frv.h (HARD_REGNO_NREGS): Delete.
16246         (CLASS_MAX_NREGS): Remove outdated copy of documentation.
16247         * config/frv/frv-protos.h (frv_hard_regno_nregs): Delete.
16248         * config/frv/frv.c (TARGET_HARD_REGNO_NREGS): Redefine.
16249         (frv_hard_regno_nregs): Make static.  Take and return an
16250         unsigned int.
16251         (frv_class_max_nregs): Remove outdated copy of documentation.
16252         * config/ft32/ft32.h (HARD_REGNO_NREGS): Delete.
16253         * config/h8300/h8300.h (HARD_REGNO_NREGS): Delete.
16254         * config/h8300/h8300-protos.h (h8300_hard_regno_nregs): Delete.
16255         * config/h8300/h8300.c (h8300_hard_regno_nregs): Delete.
16256         * config/i386/i386.h (HARD_REGNO_NREGS): Delete.
16257         * config/i386/i386.c (ix86_hard_regno_nregs): New function.
16258         (TARGET_HARD_REGNO_NREGS): Redefine.
16259         * config/ia64/ia64.h (HARD_REGNO_NREGS): Delete.
16260         (CLASS_MAX_NREGS): Update comment.
16261         * config/ia64/ia64.c (TARGET_HARD_REGNO_NREGS): Redefine.
16262         (ia64_hard_regno_nregs): New function.
16263         * config/iq2000/iq2000.h (HARD_REGNO_NREGS): Delete.
16264         * config/lm32/lm32.h (HARD_REGNO_NREGS): Delete.
16265         * config/m32c/m32c.h (HARD_REGNO_NREGS): Delete.
16266         * config/m32c/m32c-protos.h (m32c_hard_regno_nregs): Delete.
16267         * config/m32c/m32c.c (m32c_hard_regno_nregs_1): Take and return
16268         an unsigned int.
16269         (m32c_hard_regno_nregs): Likewise.  Make static.
16270         (TARGET_HARD_REGNO_NREGS): Redefine.
16271         * config/m32r/m32r.h (HARD_REGNO_NREGS): Delete.
16272         * config/m68k/m68k.h (HARD_REGNO_NREGS): Delete.
16273         * config/m68k/m68k.c (TARGET_HARD_REGNO_NREGS): Redefine.
16274         (m68k_hard_regno_nregs): New function.
16275         * config/mcore/mcore.h (HARD_REGNO_NREGS): Delete.
16276         * config/microblaze/microblaze.h (HARD_REGNO_NREGS): Delete.
16277         * config/mips/mips.h (HARD_REGNO_NREGS): Delete.
16278         * config/mips/mips-protos.h (mips_hard_regno_nregs): Delete.
16279         * config/mips/mips.c (mips_hard_regno_nregs): Make static.
16280         Take and return an unsigned int.
16281         (TARGET_HARD_REGNO_NREGS): Redefine.
16282         * config/mmix/mmix.h (HARD_REGNO_NREGS): Delete.
16283         (CLASS_MAX_NREGS): Use targetm.hard_regno_nregs.
16284         * config/mn10300/mn10300.h (HARD_REGNO_NREGS): Delete.
16285         * config/moxie/moxie.h (HARD_REGNO_NREGS): Delete.
16286         * config/msp430/msp430.h (HARD_REGNO_NREGS): Delete.
16287         * config/msp430/msp430-protos.h (msp430_hard_regno_nregs): Delete.
16288         * config/msp430/msp430.c (TARGET_HARD_REGNO_NREGS): Redefine.
16289         (msp430_hard_regno_nregs): Make static.  Take and return an
16290         unsigned int.
16291         * config/nds32/nds32.h (HARD_REGNO_NREGS): Delete.
16292         * config/nds32/nds32-protos.h (nds32_hard_regno_nregs): Delete.
16293         * config/nds32/nds32.c (nds32_hard_regno_nregs): Delete.
16294         (nds32_hard_regno_mode_ok): Use targetm.hard_regno_nregs.
16295         * config/nios2/nios2.h (HARD_REGNO_NREGS): Delete.
16296         * config/nvptx/nvptx.h (HARD_REGNO_NREGS): Delete.
16297         * config/nvptx/nvptx.c (nvptx_hard_regno_nregs): New function.
16298         (TARGET_HARD_REGNO_NREGS): Redefine.
16299         * config/pa/pa32-regs.h (HARD_REGNO_NREGS): Rename to...
16300         (PA_HARD_REGNO_NREGS): ...this.
16301         * config/pa/pa64-regs.h (HARD_REGNO_NREGS): Rename to...
16302         (PA_HARD_REGNO_NREGS): ...this.
16303         * config/pa/pa.c (TARGET_HARD_REGNO_NREGS): Redefine.
16304         (pa_hard_regno_nregs): New function.
16305         * config/pdp11/pdp11.h (HARD_REGNO_NREGS): Delete.
16306         * config/pdp11/pdp11.c (TARGET_HARD_REGNO_NREGS): Redefine.
16307         (pdp11_hard_regno_nregs): New function.
16308         * config/powerpcspe/powerpcspe.h (HARD_REGNO_NREGS): Delete.
16309         * config/powerpcspe/powerpcspe.c (TARGET_HARD_REGNO_NREGS): Redefine.
16310         (rs6000_hard_regno_nregs_hook): New function.
16311         * config/riscv/riscv.h (HARD_REGNO_NREGS): Delete.
16312         * config/riscv/riscv-protos.h (riscv_hard_regno_nregs): Delete.
16313         * config/riscv/riscv.c (riscv_hard_regno_nregs): Make static.
16314         Take and return an unsigned int.  Move earlier in file.
16315         (TARGET_HARD_REGNO_NREGS): Redefine.
16316         * config/rl78/rl78.h (HARD_REGNO_NREGS): Delete.
16317         * config/rl78/rl78-protos.h (rl78_hard_regno_nregs): Delete.
16318         * config/rl78/rl78.c (TARGET_HARD_REGNO_NREGS): Reefine.
16319         (rl78_hard_regno_nregs): Make static.  Take and return an
16320         unsigned int.
16321         * config/rs6000/rs6000.h (HARD_REGNO_NREGS): Delete.
16322         * config/rs6000/rs6000.c (TARGET_HARD_REGNO_NREGS): Redefine.
16323         (rs6000_hard_regno_nregs_hook): New function.
16324         * config/rx/rx.h (HARD_REGNO_NREGS): Delete.
16325         * config/rx/rx.c (rx_hard_regno_nregs): New function.
16326         (TARGET_HARD_REGNO_NREGS): Redefine.
16327         * config/s390/s390.h (HARD_REGNO_NREGS): Delete.
16328         * config/s390/s390.c (REGNO_PAIR_OK): Use s390_hard_regno_nregs
16329         instead of HARD_REGNO_NREGS.
16330         (s390_hard_regno_nregs): New function.
16331         (s390_hard_regno_mode_ok): Add comment from s390.h.
16332         (TARGET_HARD_REGNO_NREGS): Redefine.
16333         * config/sh/sh.h (HARD_REGNO_NREGS): Delete.
16334         * config/sh/sh.c (TARGET_HARD_REGNO_NREGS): Redefine.
16335         (sh_hard_regno_nregs): New function.
16336         (sh_pass_in_reg_p): Use it.
16337         * config/sparc/sparc.h (HARD_REGNO_NREGS): Delete.
16338         * config/sparc/sparc.c (TARGET_HARD_REGNO_NREGS): Redefine.
16339         (sparc_hard_regno_nregs): New function.
16340         * config/spu/spu.h (HARD_REGNO_NREGS): Delete.
16341         * config/spu/spu.c (spu_hard_regno_nregs): New function.
16342         (spu_function_arg_advance): Use it, supplying a valid register number.
16343         (TARGET_HARD_REGNO_NREGS): Redefine.
16344         * config/stormy16/stormy16.h (HARD_REGNO_NREGS): Delete.
16345         * config/tilegx/tilegx.h (HARD_REGNO_NREGS): Delete.
16346         * config/tilepro/tilepro.h (HARD_REGNO_NREGS): Delete.
16347         * config/v850/v850.h (HARD_REGNO_NREGS): Delete.
16348         * config/vax/vax.h (HARD_REGNO_NREGS): Delete.
16349         * config/visium/visium.h (HARD_REGNO_NREGS): Delete.
16350         (CLASS_MAX_NREGS): Remove copy of old documentation.
16351         * config/visium/visium.c (TARGET_HARD_REGNO_NREGS): Redefine.
16352         (visium_hard_regno_nregs): New function.
16353         (visium_hard_regno_mode_ok): Use it instead of HARD_REGNO_NREGS.
16354         * config/xtensa/xtensa.h (HARD_REGNO_NREGS): Delete.
16355         * config/xtensa/xtensa.c (TARGET_HARD_REGNO_NREGS): Redefine.
16356         xtensa_hard_regno_nregs): New function.
16357         * system.h (HARD_REGNO_NREGS): Poison.
16359 2017-09-12  Richard Sandiford  <richard.sandiford@linaro.org>
16361         * config/arm/arm.h (THUMB_SECONDARY_INPUT_RELOAD_CLASS): Use
16362         hard_regno_nregs instead of HARD_REGNO_NREGS.
16363         (THUMB_SECONDARY_OUTPUT_RELOAD_CLASS): Likewise.
16364         * config/c6x/c6x.c (c6x_expand_prologue): Likewise.
16365         (c6x_expand_epilogue): Likewise.
16366         * config/frv/frv.c (frv_alloc_temp_reg): Likewise.
16367         (frv_read_iacc_argument): Likewise.
16368         * config/sh/sh.c: Include regs.h.
16369         (sh_print_operand): Use hard_regno_nregs instead of HARD_REGNO_NREGS.
16370         (regs_used): Likewise.
16371         (output_stack_adjust): Likewise.
16372         * config/xtensa/xtensa.c (xtensa_copy_incoming_a7): Likewise.
16373         * expmed.c: Include regs.h.
16374         (store_bit_field_1): Use hard_regno_nregs instead of HARD_REGNO_NREGS.
16375         * ree.c: Include regs.h.
16376         (combine_reaching_defs): Use hard_regno_nregs instead of
16377         HARD_REGNO_NREGS.
16378         (add_removable_extension): Likewise.
16380 2017-09-12  Richard Sandiford  <richard.sandiford@linaro.org>
16382         * regs.h (hard_regno_nregs): Turn into a function.
16383         (end_hard_regno): Update accordingly.
16384         * caller-save.c (setup_save_areas): Likewise.
16385         (save_call_clobbered_regs): Likewise.
16386         (replace_reg_with_saved_mem): Likewise.
16387         (insert_restore): Likewise.
16388         (insert_save): Likewise.
16389         * combine.c (can_change_dest_mode): Likewise.
16390         (move_deaths): Likewise.
16391         (distribute_notes): Likewise.
16392         * config/mips/mips.c (mips_hard_regno_call_part_clobbered): Likewise.
16393         * config/powerpcspe/powerpcspe.c (rs6000_cannot_change_mode_class)
16394         (rs6000_split_multireg_move): Likewise.
16395         (rs6000_register_move_cost): Likewise.
16396         (rs6000_memory_move_cost): Likewise.
16397         * config/rs6000/rs6000.c (rs6000_cannot_change_mode_class): Likewise.
16398         (rs6000_split_multireg_move): Likewise.
16399         (rs6000_register_move_cost): Likewise.
16400         (rs6000_memory_move_cost): Likewise.
16401         * cselib.c (cselib_reset_table): Likewise.
16402         (cselib_lookup_1): Likewise.
16403         * emit-rtl.c (set_mode_and_regno): Likewise.
16404         * function.c (aggregate_value_p): Likewise.
16405         * ira-color.c (setup_profitable_hard_regs): Likewise.
16406         (check_hard_reg_p): Likewise.
16407         (calculate_saved_nregs): Likewise.
16408         (assign_hard_reg): Likewise.
16409         (improve_allocation): Likewise.
16410         (calculate_spill_cost): Likewise.
16411         * ira-emit.c (modify_move_list): Likewise.
16412         * ira-int.h (ira_hard_reg_set_intersection_p): Likewise.
16413         (ira_hard_reg_in_set_p): Likewise.
16414         * ira.c (setup_reg_mode_hard_regset): Likewise.
16415         (clarify_prohibited_class_mode_regs): Likewise.
16416         (check_allocation): Likewise.
16417         * lra-assigns.c (find_hard_regno_for_1): Likewise.
16418         (lra_setup_reg_renumber): Likewise.
16419         (setup_try_hard_regno_pseudos): Likewise.
16420         (spill_for): Likewise.
16421         (assign_hard_regno): Likewise.
16422         (setup_live_pseudos_and_spill_after_risky_transforms): Likewise.
16423         * lra-constraints.c (in_class_p): Likewise.
16424         (lra_constraint_offset): Likewise.
16425         (simplify_operand_subreg): Likewise.
16426         (lra_constraints): Likewise.
16427         (split_reg): Likewise.
16428         (split_if_necessary): Likewise.
16429         (invariant_p): Likewise.
16430         (inherit_in_ebb): Likewise.
16431         * lra-lives.c (process_bb_lives): Likewise.
16432         * lra-remat.c (reg_overlap_for_remat_p): Likewise.
16433         (get_hard_regs): Likewise.
16434         (do_remat): Likewise.
16435         * lra-spills.c (assign_spill_hard_regs): Likewise.
16436         * mode-switching.c (create_pre_exit): Likewise.
16437         * postreload.c (reload_combine_recognize_pattern): Likewise.
16438         * recog.c (peep2_find_free_register): Likewise.
16439         * regcprop.c (kill_value_regno): Likewise.
16440         (set_value_regno): Likewise.
16441         (copy_value): Likewise.
16442         (maybe_mode_change): Likewise.
16443         (find_oldest_value_reg): Likewise.
16444         (copyprop_hardreg_forward_1): Likewise.
16445         * regrename.c (check_new_reg_p): Likewise.
16446         (regrename_do_replace): Likewise.
16447         * reload.c (push_reload): Likewise.
16448         (combine_reloads): Likewise.
16449         (find_dummy_reload): Likewise.
16450         (operands_match_p): Likewise.
16451         (find_reloads): Likewise.
16452         (find_equiv_reg): Likewise.
16453         (reload_adjust_reg_for_mode): Likewise.
16454         * reload1.c (count_pseudo): Likewise.
16455         (count_spilled_pseudo): Likewise.
16456         (find_reg): Likewise.
16457         (clear_reload_reg_in_use): Likewise.
16458         (free_for_value_p): Likewise.
16459         (allocate_reload_reg): Likewise.
16460         (choose_reload_regs): Likewise.
16461         (reload_adjust_reg_for_temp): Likewise.
16462         (emit_reload_insns): Likewise.
16463         (delete_output_reload): Likewise.
16464         * rtlanal.c (subreg_get_info): Likewise.
16465         * sched-deps.c (sched_analyze_reg): Likewise.
16466         * sel-sched.c (init_regs_for_mode): Likewise.
16467         (mark_unavailable_hard_regs): Likewise.
16468         (choose_best_reg_1): Likewise.
16469         (verify_target_availability): Likewise.
16470         * valtrack.c (dead_debug_insert_temp): Likewise.
16471         * var-tracking.c (track_loc_p): Likewise.
16472         (emit_note_insn_var_location): Likewise.
16473         * varasm.c (make_decl_rtl): Likewise.
16474         * reginfo.c (choose_hard_reg_mode): Likewise.
16475         (init_reg_modes_target): Refer directly to
16476         this_target_regs->x_hard_regno_nregs.
16478 2017-09-12  Richard Sandiford  <richard.sandiford@linaro.org>
16480         * ira-costs.c (record_operand_costs): Use in_hard_reg_set_p
16481         instead of hard_regno_nregs.
16483 2017-09-12  Richard Sandiford  <richard.sandiford@linaro.org>
16485         * config/aarch64/aarch64.c (aarch64_hard_regno_mode_ok): Use
16486         end_hard_regno instead of hard_regno_nregs.
16487         * config/s390/s390.c (s390_reg_clobbered_rtx): Likewise.
16488         * config/sparc/sparc.h (ASM_DECLARE_REGISTER_GLOBAL): Likewise.
16489         * config/visium/visium.c (visium_hard_regno_mode_ok): Likewise.
16490         * ira-color.c (improve_allocation): Likewise.
16491         * lra-assigns.c (find_hard_regno_for_1): Likewise.
16492         * lra-lives.c (mark_regno_live): Likewise.
16493         (mark_regno_dead): Likewise.
16494         * lra-remat.c (operand_to_remat): Likewise.
16495         * lra.c (collect_non_operand_hard_regs): Likewise.
16496         * postreload.c (reload_combine_note_store): Likewise.
16497         (move2add_valid_value_p): Likewise.
16498         * reload.c (regno_clobbered_p): Likewise.
16500 2017-09-12  Richard Sandiford  <richard.sandiford@linaro.org>
16502         * config/frv/frv.c (FOR_EACH_REGNO): Use END_REGNO instead of
16503         hard_regno_nregs.
16504         * config/v850/v850.c (v850_reorg): Likewise.
16505         * reload.c (refers_to_regno_for_reload_p): Likewise.
16506         (find_equiv_reg): Likewise.
16507         * reload1.c (reload_reg_reaches_end_p): Likewise.
16509 2017-09-12  Richard Sandiford  <richard.sandiford@linaro.org>
16511         * caller-save.c (add_used_regs): Use REG_NREGS instead of
16512         hard_regno_nregs.
16513         * config/aarch64/aarch64.c (aarch64_split_combinev16qi): Likewise.
16514         * config/arm/arm.c (output_move_neon): Likewise.
16515         (arm_attr_length_move_neon): Likewise.
16516         (neon_split_vcombine): Likewise.
16517         * config/c6x/c6x.c (c6x_mark_reg_read): Likewise.
16518         (c6x_mark_reg_written): Likewise.
16519         (c6x_dwarf_register_span): Likewise.
16520         * config/i386/i386.c (ix86_save_reg): Likewise.
16521         * config/ia64/ia64.c (mark_reg_gr_used_mask): Likewise.
16522         (rws_access_reg): Likewise.
16523         * config/s390/s390.c (s390_call_saved_register_used): Likewise.
16524         * mode-switching.c (create_pre_exit): Likewise.
16525         * ree.c (combine_reaching_defs): Likewise.
16526         (add_removable_extension): Likewise.
16527         * regcprop.c (find_oldest_value_reg): Likewise.
16528         (copyprop_hardreg_forward_1): Likewise.
16529         * reload.c (reload_inner_reg_of_subreg): Likewise.
16530         (push_reload): Likewise.
16531         (combine_reloads): Likewise.
16532         (find_dummy_reload): Likewise.
16533         (reload_adjust_reg_for_mode): Likewise.
16534         * reload1.c (find_reload_regs): Likewise.
16535         (forget_old_reloads_1): Likewise.
16536         (reload_reg_free_for_value_p): Likewise.
16537         (reload_adjust_reg_for_temp): Likewise.
16538         (emit_reload_insns): Likewise.
16539         (delete_output_reload): Likewise.
16540         * sel-sched.c (choose_best_reg_1): Likewise.
16541         (choose_best_pseudo_reg): Likewise.
16543 2017-09-12  Richard Sandiford  <richard.sandiford@linaro.org>
16544             Alan Hayward  <alan.hayward@arm.com>
16545             David Sherwood <david.sherwood@arm.com>
16547         * defaults.h (SLOW_UNALIGNED_ACCESS): Delete.
16548         * target.def (slow_unaligned_access): New hook.
16549         * targhooks.h (default_slow_unaligned_access): Declare.
16550         * targhooks.c (default_slow_unaligned_access): New function.
16551         * doc/tm.texi.in (SLOW_UNALIGNED_ACCESS): Replace with...
16552         (TARGET_SLOW_UNALIGNED_ACCESS): ...this.
16553         * doc/tm.texi: Regenerate.
16554         * config/alpha/alpha.h (SLOW_UNALIGNED_ACCESS): Delete.
16555         * config/arm/arm.h (SLOW_UNALIGNED_ACCESS): Delete.
16556         * config/i386/i386.h (SLOW_UNALIGNED_ACCESS): Delete commented-out
16557         definition.
16558         * config/powerpcspe/powerpcspe.h (SLOW_UNALIGNED_ACCESS): Delete.
16559         * config/powerpcspe/powerpcspe.c (TARGET_SLOW_UNALIGNED_ACCESS):
16560         Redefine.
16561         (rs6000_slow_unaligned_access): New function.
16562         (rs6000_emit_move): Use it instead of SLOW_UNALIGNED_ACCESS.
16563         (expand_block_compare): Likewise.
16564         (expand_strn_compare): Likewise.
16565         (rs6000_rtx_costs): Likewise.
16566         * config/riscv/riscv.h (SLOW_UNALIGNED_ACCESS): Delete.
16567         (riscv_slow_unaligned_access): Likewise.
16568         * config/riscv/riscv.c (riscv_slow_unaligned_access): Rename to...
16569         (riscv_slow_unaligned_access_p): ...this and make static.
16570         (riscv_option_override): Update accordingly.
16571         (riscv_slow_unaligned_access): New function.
16572         (TARGET_SLOW_UNALIGNED_ACCESS): Redefine.
16573         * config/rs6000/rs6000.h (SLOW_UNALIGNED_ACCESS): Delete.
16574         * config/rs6000/rs6000.c (TARGET_SLOW_UNALIGNED_ACCESS): Redefine.
16575         (rs6000_slow_unaligned_access): New function.
16576         (rs6000_emit_move): Use it instead of SLOW_UNALIGNED_ACCESS.
16577         (rs6000_rtx_costs): Likewise.
16578         * config/rs6000/rs6000-string.c (expand_block_compare)
16579         (expand_strn_compare): Use targetm.slow_unaligned_access instead
16580         of SLOW_UNALIGNED_ACCESS.
16581         * config/tilegx/tilegx.h (SLOW_UNALIGNED_ACCESS): Delete.
16582         * config/tilepro/tilepro.h (SLOW_UNALIGNED_ACCESS): Delete.
16583         * calls.c (expand_call): Use targetm.slow_unaligned_access instead
16584         of SLOW_UNALIGNED_ACCESS.
16585         * expmed.c (simple_mem_bitfield_p): Likewise.
16586         * expr.c (alignment_for_piecewise_move): Likewise.
16587         (emit_group_load_1): Likewise.
16588         (emit_group_store): Likewise.
16589         (copy_blkmode_from_reg): Likewise.
16590         (emit_push_insn): Likewise.
16591         (expand_assignment): Likewise.
16592         (store_field): Likewise.
16593         (expand_expr_real_1): Likewise.
16594         * gimple-fold.c (gimple_fold_builtin_memory_op): Likewise.
16595         * lra-constraints.c (simplify_operand_subreg): Likewise.
16596         * stor-layout.c (bit_field_mode_iterator::next_mode): Likewise.
16597         * gimple-ssa-store-merging.c: Likewise in block comment at start
16598         of file.
16599         * tree-ssa-strlen.c: Include target.h.
16600         (handle_builtin_memcmp): Use targetm.slow_unaligned_access instead
16601         of SLOW_UNALIGNED_ACCESS.
16602         * system.h (SLOW_UNALIGNED_ACCESS): Poison.
16604 2017-09-12  Richard Sandiford  <richard.sandiford@linaro.org>
16606         PR rtl-optimization/82185
16607         * expmed.c (emit_store_flag_int): Only test tem if it has been
16608         initialized.
16610 2017-09-12  Richard Biener  <rguenther@suse.de>
16612         PR middle-end/82149
16613         * match.pd ((FTYPE) N CMP CST): Fix typo.
16615 2017-09-12  Simon Atanasyan  <simon.atanasyan@imgtec.com>
16617         * config/mips/mips.c (mips_attribute_table): Add 'short_call'
16618         attribute.
16619         (mips_near_type_p): Add 'short_call' attribute as a synonym
16620         for 'near'.
16621         * doc/extend.texi (short_call): Document new function attribute.
16623 2017-09-12  Jakub Jelinek  <jakub@redhat.com>
16625         PR target/82112
16626         * c-common.c (sync_resolve_size): Instead of c_dialect_cxx ()
16627         assertion check that in the condition.
16628         (get_atomic_generic_size): Likewise.  Before testing if parameter
16629         has pointer type, if it has array type, call for C++
16630         default_conversion to perform array-to-pointer conversion.
16632 2017-09-12  Richard Biener  <rguenther@suse.de>
16634         * tree-vect-generic.c (expand_vector_operations_1): Do nothing
16635         for operations we cannot scalarize.
16637 2017-09-12  Aldy Hernandez  <aldyh@redhat.com>
16639         * tree-ssa-threadbackward.c (fsm_find_thread_path): Make GC
16640         vectors heap vectors.  Clean up comments.
16641         Make visited_bbs a reference.
16642         (profitable_jump_thread_path): Make GC
16643         vectors heap vectors.  Clean up comments.
16644         Misc cleanups.
16645         (convert_and_register_jump_thread_path): Make GC vectors heap
16646         vectors.
16647         (check_subpath_and_update_thread_path): Same.  Clean up comments.
16648         Make visited_bbs a reference.
16649         (handle_phi): Abstract common code to to
16650         register_jump_thread_path_if_profitable.
16651         Rename VAR_BB to DEF_BB.
16652         Update comments.
16653         Make GC vectors heap vectors.
16654         Make visited_bbs a reference.
16655         (handle_assignment): Same.
16656         (register_jump_thread_path_if_profitable): New.
16657         (fsm_find_control_statement_thread_paths): Rename VAR_BB to
16658         DEF_BB.
16659         Make GC vectors heap vectors.  Clean up comments.
16660         Make visited_bbs a reference.
16661         (find_jump_threads_backwards): Make visited_bbs live in the stack.
16662         * tree-ssa-threadupdate.c (delete_jump_thread_path): Fix typo in
16663         comment.
16665 2017-09-11  Max Filippov  <jcmvbkbc@gmail.com>
16667         PR target/82181
16668         * config/xtensa/xtensa.c (xtensa_mem_offset): Check that both
16669         words of E_DImode object are reachable by xtensa_uimm8x4 access.
16671 2017-09-11  Vidya Praveen  <vidyapraveen@arm.com>
16673         Revert r251800 and r251799.
16675 2017-09-11  Martin Jambor  <mjambor@suse.cz>
16677         PR hsa/82119
16678         * hsa-gen.c (gen_hsa_phi_from_gimple_phi): Process ADDR_EXPRs in
16679         arguments in advance.
16680         * hsa-regalloc.c (naive_process_phi): New parameter predecessors,
16681         use it to find predecessor edges.
16682         (naive_outof_ssa): Collect vector of predecessors.
16684 2017-09-08  Jason Merrill  <jason@redhat.com>
16686         PR c++/70029 - ICE with ref-qualifier and -flto
16687         * langhooks.h (struct lang_hooks_for_types): Add
16688         copy_lang_qualifiers.
16689         * attribs.c (build_type_attribute_qual_variant): Use it.
16690         * langhooks-def.h (LANG_HOOKS_COPY_LANG_QUALIFIERS): Default to
16691         NULL.
16692         (LANG_HOOKS_FOR_TYPES_INITIALIZER): Use it.
16693         * tree.c (verify_type): Re-enable TYPE_CANONICAL main variant check.
16695 2017-09-08  Eric Botcazou  <ebotcazou@adacore.com>
16697         PR target/81988
16698         * config/sparc/sparc.md (mulsi3): Rename into *mulsi3_sp32.
16699         (*mulsi3_sp64): New instruction.
16700         (mulsi3): New expander.
16702 2017-09-08  Uros Bizjak  <ubizjak@gmail.com>
16704         * config/alpha/alpha.c (alpha_print_operand) <case 'S'>: Remove.
16706 2017-09-08  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
16708         * sancov.c: Include memmodel.h.
16710 2017-09-07  Eric Botcazou  <ebotcazou@adacore.com>
16712         PR target/80897
16713         * config/sparc/sparc.c (sparc_emit_set_symbolic_const64): Deal with too
16714         large offsets.
16716 2017-09-07  Carl Love  <cel@us.ibm.com>
16718         * config/rs6000/vsx.md (define_insn "*stxvl"): Add missing argument to
16719         the sldi instruction.
16721 2017-09-07  David Edelsohn  <dje.gcc@gmail.com>
16723         * sancov.c: Include tm_p.h.
16725 2017-09-07  Jakub Jelinek  <jakub@redhat.com>
16727         PR target/81979
16728         * output.h (switch_to_other_text_partition): New declaration.
16729         * varasm.c (switch_to_other_text_partition): New function.
16730         * config/rs6000/rs6000.c (uses_TOC): Return 2 if
16731         NOTE_INSN_SWITCH_TEXT_SECTIONS is seen before finding load_toc_* insn.
16732         (rs6000_elf_declare_function_name): If uses_TOC returned 2, switch
16733         to the other text partition before emitting LCL label and switch back
16734         after emitting the word after it.
16736 2017-09-07  Richard Biener  <rguenther@suse.de>
16738         * passes.def (pass_split_crit_edges): Remove instance before PRE.
16739         * tree-ssa-pre.c (pass_pre::execute): Instead manually split
16740         critical edges here, after loop init.
16741         (pass_data_pre): Remove PROP_no_crit_edges flags.
16742         * tree-ssa-sccvn.c (vn_reference_lookup_3): Use vn_valueize
16743         for valueization of call args to avoid leaking VN_TOP.
16744         (visit_use): Assert we do not visit default defs.
16745         (init_scc_vn): Use build_decl for VN_TOP to make name nicer.
16746         Use error_mark_node to more easily detect leaking VN_TOP.
16747         All default-defs are varying, not VN_TOP.  Mark them visited.
16748         (run_scc_vn): Make code match comment.
16750 2017-09-07  Michael Meissner  <meissner@linux.vnet.ibm.com>
16752         * config/rs6000/rs6000-cpus.def (OTHER_VSX_VECTOR_MASKS): Delete
16753         OPTION_MASK_FLOAT128_KEYWORD.
16754         (POWERPC_MASKS): Likewise.
16755         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Delete
16756         support for the -mfloat128-type option, and make -mfloat128
16757         default on PowerPC Linux systems.  Define or undefine
16758         __FLOAT128__ and  __FLOAT128_HARDWARE__ for the current options.
16759         Define __float128 to be __ieee128 if IEEE 128-bit support is
16760         enabled, or undefine it.
16761         (rs6000_cpu_cpp_builtins): Delete defining __FLOAT128__ here.
16762         Delete defining __FLOAT128_TYPE__.
16763         * config/rs6000/rs6000.opt (x_TARGET_FLOAT128_TYPE): Delete the
16764         -mfloat128-type option and make -mfloat128 default on PowerPC
16765         Linux systems.
16766         (TARGET_FLOAT128_TYPE): Likewise.
16767         (-mfloat128-type): Likewise.
16768         * config/rs6000/rs6000.c (rs6000_option_override_internal):
16769         Delete the -mfloat128-type option and make -mfloat128 default on
16770         PowerPC Linux systems.  Always use __ieee128 to be the keyword for
16771         the IEEE 128-bit type, and map __float128 to __ieee128 if IEEE
16772         128-bit floating point is enabled.  Change tests from using
16773         -mfloat128-type to -mfloat128.
16774         (rs6000_mangle_type): Use the correct mangling for the __float128
16775         type even if normal long double is restricted to 64-bits.
16776         (floatn_mode): Enable the _Float128 type by default on VSX Linux
16777         systems.
16778         * config/rs6000/rs6000.h (MASK_FLOAT128_TYPE): Delete.
16779         (MASK_FLOAT128_KEYWORD): Define new shortcut macro.
16780         (RS6000BTM_FLOAT128): Define in terms of -mfloat128, not
16781         -mfloat128-type.
16782         * doc/invoke.texi (RS/6000 and PowerPC Options): Update
16783         documentation for -mfloat128.
16785 2017-09-06  Olivier Hainque  <hainque@adacore.com>
16787         * config.gcc (powerpc-wrs-vxworksspe): Now match as vxworks*spe.
16789 2017-09-06  Wish Wu  <wishwu007@gmail.com>
16790             Jakub Jelinek  <jakub@redhat.com>
16792         * asan.c (initialize_sanitizer_builtins): Add
16793         BT_FN_VOID_UINT8_UINT8, BT_FN_VOID_UINT16_UINT16,
16794         BT_FN_VOID_UINT32_UINT32, BT_FN_VOID_UINT64_UINT64,
16795         BT_FN_VOID_FLOAT_FLOAT, BT_FN_VOID_DOUBLE_DOUBLE and
16796         BT_FN_VOID_UINT64_PTR variables.
16797         * builtin-types.def (BT_FN_VOID_UINT8_UINT8): New fn type.
16798         (BT_FN_VOID_UINT16_UINT16): Likewise.
16799         (BT_FN_VOID_UINT32_UINT32): Likewise.
16800         (BT_FN_VOID_FLOAT_FLOAT): Likewise.
16801         (BT_FN_VOID_DOUBLE_DOUBLE): Likewise.
16802         (BT_FN_VOID_UINT64_PTR): Likewise.
16803         * common.opt (flag_sanitize_coverage): New variable.
16804         (fsanitize-coverage=trace-pc): Remove.
16805         (fsanitize-coverage=): Add.
16806         * flag-types.h (enum sanitize_coverage_code): New enum.
16807         * fold-const.c (fold_range_test): Disable non-short-circuit
16808         optimization if flag_sanitize_coverage.
16809         (fold_truth_andor): Likewise.
16810         * tree-ssa-ifcombine.c (ifcombine_ifandif): Likewise.
16811         * opts.c (COVERAGE_SANITIZER_OPT): Define.
16812         (coverage_sanitizer_opts): New array.
16813         (get_closest_sanitizer_option): Add OPTS argument, handle also
16814         OPT_fsanitize_coverage_.
16815         (parse_sanitizer_options): Adjusted to also handle
16816         OPT_fsanitize_coverage_.
16817         (common_handle_option): Add OPT_fsanitize_coverage_.
16818         * sancov.c (instrument_comparison, instrument_switch): New function.
16819         (sancov_pass): Add trace-cmp support.
16820         * sanitizer.def (BUILT_IN_SANITIZER_COV_TRACE_CMP1,
16821         BUILT_IN_SANITIZER_COV_TRACE_CMP2, BUILT_IN_SANITIZER_COV_TRACE_CMP4,
16822         BUILT_IN_SANITIZER_COV_TRACE_CMP8,
16823         BUILT_IN_SANITIZER_COV_TRACE_CONST_CMP1,
16824         BUILT_IN_SANITIZER_COV_TRACE_CONST_CMP2,
16825         BUILT_IN_SANITIZER_COV_TRACE_CONST_CMP4,
16826         BUILT_IN_SANITIZER_COV_TRACE_CONST_CMP8,
16827         BUILT_IN_SANITIZER_COV_TRACE_CMPF, BUILT_IN_SANITIZER_COV_TRACE_CMPD,
16828         BUILT_IN_SANITIZER_COV_TRACE_SWITCH): New builtins.
16829         * doc/invoke.texi: Document -fsanitize-coverage=trace-cmp.
16831 2017-09-06  Richard Earnshaw  <rearnsha@arm.com>
16833         * config/arm/parsecpu.awk (fatal): Note that we've encountered an
16834         error.  Only quit immediately if parsing is complete.
16835         (BEGIN): Initialize fatal_err and parse_done.
16836         (begin fpu, end fpu): Check number of arguments.
16837         (begin arch, end arch): Likewise.
16838         (begin cpu, end cpu): Likewise.
16839         (cname, tune for, tune flags, architecture, fpu, option): Likewise.
16840         (optalias): Likewise.
16842 2017-09-06  Richard Earnshaw  <rearnsha@arm.com>
16844         * config.gcc (arm*-*-*): Don't add arm-isa.h to tm_p_file.
16845         * config/arm/arm-isa.h: Delete.  Move definitions to ...
16846         * arm-cpus.in: ... here.  Use new feature and fgroup values.
16847         * config/arm/arm.c (arm_option_override): Use lower case for feature
16848         bit names.
16849         * config/arm/arm.h (TARGET_HARD_FLOAT): Likewise.
16850         (TARGET_VFP3, TARGET_VFP5, TARGET_FMA): Likewise.
16851         * config/arm/parsecpu.awk (END): Add new command 'isa'.
16852         (isa_pfx): Delete.
16853         (print_isa_bits_for): New function.
16854         (gen_isa): New function.
16855         (gen_comm_data): Use print_isa_bits_for.
16856         (define feature): New keyword.
16857         (define fgroup): New keyword.
16858         * config/arm/t-arm (OPTIONS_H_EXTRA): Add arm-isa.h
16859         (arm-isa.h): Add rule to generate file.
16860         * common/config/arm/arm-common.c: (arm_canon_arch_option): Use lower
16861         case for feature bit names.
16863 2017-09-06  Richard Biener  <rguenther@suse.de>
16865         * tree-ssa-pre.c (NECESSARY): Remove.
16866         (create_expression_by_pieces): Do not touch pass-local flags.
16867         (insert_into_preds_of_block): Likewise.
16868         (do_pre_regular_insertion): Likewise.
16869         (eliminate_insert): Likewise.
16870         (eliminate_dom_walker::before_dom_children): Likewise.
16871         (fini_eliminate): Do not look at inserted_exprs.
16872         (mark_operand_necessary): Remove.
16873         (remove_dead_inserted_code): Replace with simple work-list
16874         algorithm based on inserted_exprs and SSA uses.
16875         (pass_pre::execute): Re-order fini_eliminate and
16876         remove_dead_inserted_code.
16878 2017-09-06  Olivier Hainque  <hainque@adacore.com>
16880         * config/powerpcspe/vxworks.h (VXCPU_FOR_8548): Correct definition
16881         for VxWorks 7.  Adjust surrounding comments.
16883 2017-09-06  Richard Biener  <rguenther@suse.de>
16885         * gimple-ssa-strength-reduction.c
16886         (find_candidates_dom_walker::before_dom_children): Also allow
16887         pointer types.
16889 2017-09-06  Richard Biener  <rguenther@suse.de>
16891         PR tree-optimization/82108
16892         * tree-vect-stmts.c (vectorizable_load): Fix pointer adjustment
16893         for gap in the non-permutation SLP case.
16895 2017-09-06  Martin Jambor  <mjambor@suse.cz>
16897         PR tree-optimization/82078
16898         * tree-sra.c (sort_and_splice_var_accesses): Move call to
16899         add_access_to_work_queue...
16900         (build_accesses_from_assign): ...here.
16901         (propagate_all_subaccesses): Make sure racc is the group
16902         representative, if there is one.
16904 2017-09-06  Jakub Jelinek  <jakub@redhat.com>
16906         PR middle-end/82095
16907         * varasm.c (categorize_decl_for_section): Use SECCAT_TBSS for TLS vars with
16908         NULL DECL_INITIAL.
16910 2017-09-06  Richard Biener  <rguenther@suse.de>
16912         * gimple-ssa-strength-reduction.c
16913         (find_candidates_dom_walker::before_doom_children): Use a
16914         type and not a mode check.
16916 2017-09-06  Bernd Edlinger  <bernd.edlinger@hotmail.de>
16918         PR target/77308
16919         * config/arm/predicates.md (arm_general_adddi_operand): Create new
16920         non-vfp predicate.
16921         * config/arm/arm.md (*arm_adddi3, *arm_subdi3): Use new predicates.
16923 2017-09-05  Jeff Law  <law@redhat.com>
16925         PR tree-optimization/64910
16926         * tree-ssa-reassoc.c (reassociate_bb): Restrict last change to
16927         cases where we have 3 or more operands.
16929 2017-09-05  Jakub Jelinek  <jakub@redhat.com>
16931         PR middle-end/81768
16932         * omp-low.c (lower_omp_for): Recompute tree invariant if
16933         gimple_omp_for_initial/final is ADDR_EXPR.
16935         PR middle-end/81768
16936         * omp-expand.c (expand_omp_simd): Force second operands of COND_EXPR
16937         into gimple val before gimplification fo the COND_EXPR.
16939 2017-09-05  Aldy Hernandez  <aldyh@redhat.com>
16941         * tree-ssa-threadupdate.c (duplicate_thread_path): Remove unused
16942         REGION_COPY argument.
16943         (thread_through_all_blocks): Remove unused argument to
16944         duplicate_thread_path.
16946 2017-09-05  Richard Sandiford  <richard.sandiford@linaro.org>
16947             Alan Hayward  <alan.hayward@arm.com>
16948             David Sherwood  <david.sherwood@arm.com>
16950         * config/aarch64/aarch64-protos.h (aarch64_gen_adjusted_ldpstp):
16951         Take a scalar_mode rather than a machine_mode.
16952         (aarch64_operands_adjust_ok_for_ldpstp): Likewise.
16953         * config/aarch64/aarch64.c (aarch64_simd_container_mode): Likewise.
16954         (aarch64_operands_adjust_ok_for_ldpstp): Likewise.
16955         (aarch64_gen_adjusted_ldpstp): Likewise.
16956         (aarch64_expand_vector_init): Use scalar_mode instead of machine_mode.
16958 2017-09-05  Richard Sandiford  <richard.sandiford@linaro.org>
16959             Alan Hayward  <alan.hayward@arm.com>
16960             David Sherwood  <david.sherwood@arm.com>
16962         * config/aarch64/aarch64-protos.h (aarch64_is_extend_from_extract):
16963         Take a scalar_int_mode instead of a machine_mode.
16964         (aarch64_mask_and_shift_for_ubfiz_p): Likewise.
16965         (aarch64_output_scalar_simd_mov_immediate): Likewise.
16966         (aarch64_simd_scalar_immediate_valid_for_move): Likewise.
16967         (aarch64_simd_attr_length_rglist): Delete.
16968         * config/aarch64/aarch64.c (aarch64_is_extend_from_extract): Take
16969         a scalar_int_mode instead of a machine_mode.
16970         (aarch64_add_offset): Likewise.
16971         (aarch64_internal_mov_immediate): Likewise
16972         (aarch64_add_constant_internal): Likewise.
16973         (aarch64_add_constant): Likewise.
16974         (aarch64_movw_imm): Likewise.
16975         (aarch64_rtx_arith_op_extract_p): Likewise.
16976         (aarch64_mask_and_shift_for_ubfiz_p): Likewise.
16977         (aarch64_simd_scalar_immediate_valid_for_move): Likewise.
16978         Remove assert that the mode isn't a vector.
16979         (aarch64_output_scalar_simd_mov_immediate): Likewise.
16980         (aarch64_expand_mov_immediate): Update calls after above changes.
16981         (aarch64_output_casesi): Use as_a <scalar_int_mode>.
16982         (aarch64_and_bitmask_imm): Check for scalar integer modes.
16983         (aarch64_move_imm): Likewise.
16984         (aarch64_can_const_movi_rtx_p): Likewise.
16985         (aarch64_strip_extend): Likewise.
16986         (aarch64_extr_rtx_p): Likewise.
16987         (aarch64_rtx_costs): Likewise, using wode_mode as the mode of
16988         a CONST_INT when the mode parameter is VOIDmode.
16989         (aarch64_float_const_rtx_p): Use scalar_int_mode for a temporary.
16991 2017-09-05  Richard Sandiford  <richard.sandiford@linaro.org>
16993         * machmode.h (bitwise_mode_for_mode): Return opt_mode.
16994         * stor-layout.c (bitwise_mode_for_mode): Likewise.
16995         (bitwise_type_for_mode): Update accordingly.
16997 2017-09-05  Richard Sandiford  <richard.sandiford@linaro.org>
16999         * stor-layout.h (mode_for_size_tree): Return an opt_mode.
17000         * stor-layout.c (mode_for_size_tree): Likewise.
17001         (mode_for_array): Update accordingly.
17002         (layout_decl): Likewise.
17003         (compute_record_mode): Likewise.  Only set the mode once.
17005 2017-09-05  Richard Sandiford  <richard.sandiford@linaro.org>
17007         * target.def (get_mask_mode): Change return type to opt_mode.
17008         Expand commentary.
17009         * doc/tm.texi: Regenerate.
17010         * targhooks.h (default_get_mask_mode): Return an opt_mode.
17011         * targhooks.c (default_get_mask_mode): Likewise.
17012         * config/i386/i386.c (ix86_get_mask_mode): Likewise.
17013         * optabs-query.c (can_vec_mask_load_store_p): Update use of
17014         targetm.get_mask_mode.
17015         * tree.c (build_truth_vector_type): Likewise.
17017 2017-09-05  Richard Sandiford  <richard.sandiford@linaro.org>
17019         * machmode.h (mode_for_vector): Return an opt_mode.
17020         * stor-layout.c (mode_for_vector): Likewise.
17021         (mode_for_int_vector): Update accordingly.
17022         (layout_type): Likewise.
17023         * config/i386/i386.c (emit_memmov): Likewise.
17024         (ix86_expand_set_or_movmem): Likewise.
17025         (ix86_expand_vector_init): Likewise.
17026         (ix86_get_mask_mode): Likewise.
17027         * config/powerpcspe/powerpcspe.c (rs6000_expand_vec_perm_const_1):
17028         Likewise.
17029         * config/rs6000/rs6000.c (rs6000_expand_vec_perm_const_1): Likewise.
17030         * expmed.c (extract_bit_field_1): Likewise.
17031         * expr.c (expand_expr_real_2): Likewise.
17032         * optabs-query.c (can_vec_perm_p): Likewise.
17033         (can_vec_mask_load_store_p): Likewise.
17034         * optabs.c (expand_vec_perm): Likewise.
17035         * targhooks.c (default_get_mask_mode): Likewise.
17036         * tree-vect-stmts.c (vectorizable_store): Likewise.
17037         (vectorizable_load): Likewise.
17038         (get_vectype_for_scalar_type_and_size): Likewise.
17040 2017-09-05  Richard Sandiford  <richard.sandiford@linaro.org>
17042         * machmode.h (mode_for_int_vector): New function.
17043         * stor-layout.c (mode_for_int_vector): Likewise.
17044         * config/aarch64/aarch64.c (aarch64_emit_approx_sqrt): Use it.
17045         * config/powerpcspe/powerpcspe.c (rs6000_do_expand_vec_perm): Likewise.
17046         * config/rs6000/rs6000.c (rs6000_do_expand_vec_perm): Likewise.
17047         * config/s390/s390.c (s390_expand_vec_compare_cc): Likewise.
17048         (s390_expand_vcond): Likewise.
17050 2017-09-05  Richard Sandiford  <richard.sandiford@linaro.org>
17052         * machmode.h (opt_machine_mode): New type.
17053         (opt_mode<T>): Allow construction from anything that can be
17054         converted to a T.
17055         (is_a, as_a, dyn_cast): Add overloads for opt_mode.
17056         (mode_for_size): Return an opt_machine_mode.
17057         * stor-layout.c (mode_for_size): Likewise.
17058         (mode_for_size_tree): Update call accordingly.
17059         (bitwise_mode_for_mode): Likewise.
17060         (make_fract_type): Likewise.
17061         (make_accum_type): Likewise.
17062         * caller-save.c (replace_reg_with_saved_mem): Update call
17063         accordingly.
17064         * config/alpha/alpha.h (SECONDARY_MEMORY_NEEDED_MODE): Likewise.
17065         * config/i386/i386.h (SECONDARY_MEMORY_NEEDED_MODE): Likewise.
17066         * config/s390/s390.h (SECONDARY_MEMORY_NEEDED_MODE): Likewise.
17067         * config/sparc/sparc.h (SECONDARY_MEMORY_NEEDED_MODE): Likewise.
17068         * expmed.c (extract_bit_field_1): Likewise.
17069         * reload.c (get_secondary_mem): Likewise.
17070         * varasm.c (assemble_integer): Likewise.
17071         * lower-subreg.c (simplify_subreg_concatn): Likewise.  Move
17072         early-out.
17074 2017-09-05  Richard Sandiford  <richard.sandiford@linaro.org>
17076         * machmode.h (decimal_float_mode_for_size): New function.
17077         * real.h (REAL_VALUE_TO_TARGET_LONG_DOUBLE): Use float_mode_for_size.
17078         (REAL_VALUE_TO_TARGET_DOUBLE): Likewise.
17079         (REAL_VALUE_TO_TARGET_SINGLE): Likewise.
17080         (REAL_VALUE_TO_TARGET_DECIMAL128): Use decimal_float_mode_for_size.
17081         (REAL_VALUE_TO_TARGET_DECIMAL64): Likewise.
17082         (REAL_VALUE_TO_TARGET_DECIMAL32): Likewise.
17084 2017-09-05  Richard Sandiford  <richard.sandiford@linaro.org>
17086         * builtins.c (expand_builtin_powi): Use int_mode_for_size.
17087         (get_builtin_sync_mode): Likewise.
17088         (expand_ifn_atomic_compare_exchange): Likewise.
17089         (expand_builtin_atomic_clear): Likewise.
17090         (expand_builtin_atomic_test_and_set): Likewise.
17091         (fold_builtin_atomic_always_lock_free): Likewise.
17092         * calls.c (compute_argument_addresses): Likewise.
17093         (emit_library_call_value_1): Likewise.
17094         (store_one_arg): Likewise.
17095         * combine.c (combine_instructions): Likewise.
17096         * config/aarch64/aarch64.c (aarch64_function_value): Likewise.
17097         * config/arm/arm.c (arm_function_value): Likewise.
17098         (aapcs_allocate_return_reg): Likewise.
17099         * config/c6x/c6x.c (c6x_expand_movmem): Likewise.
17100         * config/i386/i386.c (construct_container): Likewise.
17101         (ix86_gimplify_va_arg): Likewise.
17102         (ix86_expand_sse_cmp): Likewise.
17103         (emit_memmov): Likewise.
17104         (emit_memset): Likewise.
17105         (expand_small_movmem_or_setmem): Likewise.
17106         (ix86_expand_pextr): Likewise.
17107         (ix86_expand_pinsr): Likewise.
17108         * config/lm32/lm32.c (lm32_block_move_inline): Likewise.
17109         * config/microblaze/microblaze.c (microblaze_block_move_straight):
17110         Likewise.
17111         * config/mips/mips.c (mips_function_value_1) Likewise.
17112         (mips_block_move_straight): Likewise.
17113         (mips_expand_ins_as_unaligned_store): Likewise.
17114         * config/powerpcspe/powerpcspe.c
17115         (rs6000_darwin64_record_arg_advance_flush): Likewise.
17116         (rs6000_darwin64_record_arg_flush): Likewise.
17117         * config/rs6000/rs6000.c
17118         (rs6000_darwin64_record_arg_advance_flush): Likewise.
17119         (rs6000_darwin64_record_arg_flush): Likewise.
17120         * config/sparc/sparc.c (sparc_function_arg_1): Likewise.
17121         (sparc_function_value_1): Likewise.
17122         * config/spu/spu.c (adjust_operand): Likewise.
17123         (spu_emit_branch_or_set): Likewise.
17124         (arith_immediate_p): Likewise.
17125         * emit-rtl.c (gen_lowpart_common): Likewise.
17126         * expr.c (expand_expr_real_1): Likewise.
17127         * function.c (assign_parm_setup_block): Likewise.
17128         * gimple-ssa-store-merging.c (encode_tree_to_bitpos): Likewise.
17129         * reload1.c (alter_reg): Likewise.
17130         * stor-layout.c (mode_for_vector): Likewise.
17131         (layout_type): Likewise.
17133 2017-09-05  Richard Sandiford  <richard.sandiford@linaro.org>
17135         * config/spu/spu.c (exp2_immediate_p): Use int_mode_for_mode.
17136         (spu_convert_move): Likewise.
17137         * lower-subreg.c (resolve_simple_move): Likewise.
17139 2017-09-05  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
17141         PR target/81833
17142         * config/rs6000/altivec.md (altivec_vsum2sws): Convert from a
17143         define_insn to a define_expand.
17144         (altivec_vsum2sws_direct): New define_insn.
17145         (altivec_vsumsws): Convert from a define_insn to a define_expand.
17147 2017-09-05  Wilco Dijkstra  <wdijkstr@arm.com>
17149         * config/arm/arm.c (arm_option_params_internal): Improve setting of
17150         max_insns_skipped.
17152 2017-09-05  H.J. Lu  <hongjiu.lu@intel.com>
17154         PR target/59501
17155         PR target/81624
17156         PR target/81769
17157         * config/i386/i386.c (ix86_finalize_stack_frame_flags): Don't
17158         realign stack if stack alignment needed is less than incoming
17159         stack boundary.
17161 2017-09-05  Marek Polacek  <polacek@redhat.com>
17163         PR sanitizer/82072
17164         * convert.c (convert_to_integer_1) <case NEGATE_EXPR>: Move the ubsan
17165         check earlier.
17167 2017-09-05  Wilco Dijkstra  <wdijkstr@arm.com>
17169         * explow.c (get_dynamic_stack_size): Improve dynamic alignment.
17171 2017-09-05  Richard Biener  <rguenther@suse.de>
17173         PR tree-optimization/82084
17174         * fold-const.c (can_native_encode_string_p): Handle wide characters.
17176 2017-09-05  Richard Biener  <rguenther@suse.de>
17178         PR tree-optimization/82102
17179         * tree-ssa-pre.c (fini_eliminate): Check if lhs is NULL.
17181 2017-09-05  Martin Liska  <mliska@suse.cz>
17183         PR tree-optimization/82032
17184         * tree-cfg.c (generate_range_test): New function.
17185         * tree-cfg.h (generate_range_test): Declared here.
17186         * tree-cfgcleanup.c (convert_single_case_switch): New function.
17187         (cleanup_control_expr_graph): Use it.
17188         * tree-switch-conversion.c (try_switch_expansion): Remove
17189         assert.
17190         (emit_case_nodes): Use generate_range_test.
17192 2017-09-04  Uros Bizjak  <ubizjak@gmail.com>
17194         PR target/82098
17195         * config/i386/i386.md (*<btsc><mode>_mask): Add
17196         TARGET_USE_BT to insn constraint.
17197         (*btr<mode>_mask): Ditto.
17199 2017-09-04  Wilco Dijkstra  <wdijkstr@arm.com>
17201         * config/arm/arm.c (arm_legitimate_index_p): Add comment.
17202         (thumb2_legitimate_index_p): Use correct range for DI/DF mode.
17204 2017-09-04  Bernd Edlinger  <bernd.edlinger@hotmail.de>
17206         PR target/77308
17207         * config/arm/arm.md (*arm_adddi3, *arm_subdi3): Split early except for
17208         TARGET_NEON and TARGET_IWMMXT.
17209         (anddi3, iordi3, xordi3, one_cmpldi2): Split while expanding except for
17210         TARGET_NEON and TARGET_IWMMXT.
17211         (*one_cmpldi2_insn): Moved the body of one_cmpldi2 here.
17213 2017-09-04  Uros Bizjak  <ubizjak@gmail.com>
17215         * config/i386/i386-protos.h (ix86_tls_address_pattern_p) New prototype.
17216         (ix86_rewrite_tls_address): Ditto.
17217         * config/i386/i386.c (ix86_tls_address_pattern_p) New function.
17218         (ix86_rewrite_tls_address_1): Ditto.
17219         (ix86_rewrite_tls_address): Ditto.
17220         * config/i386/predicates.md (tls_address_pattern): New predicate.
17221         * config/i386/i386.md (TLS address splitter): New splitter.
17223 2017-09-04  Richard Biener  <rguenther@suse.de>
17225         PR tree-optimization/82084
17226         * fold-const.h (can_native_encode_string_p): Declare.
17227         * fold-const.c (can_native_encode_string_p): Factor out from ...
17228         (native_encode_string): ... here.
17229         * tree-vect-stmts.c (vectorizable_store): Call it to avoid
17230         vectorizing stores from constants we later cannot handle.
17232 2017-09-04  Marek Polacek  <polacek@redhat.com>
17234         PR c/81783
17235         * doc/invoke.texi: Update -Wtautological-compare documentation.
17237 2017-09-04  Jeff Law  <law@redhat.com>
17239         PR tree-optimization/64910
17240         * tree-ssa-reassoc.c (reassociate_bb): For bitwise binary ops,
17241         swap the first and last operand if the last is a constant.
17243 2017-09-04  Marek Polacek  <polacek@redhat.com>
17245         PR sanitizer/82072
17246         * convert.c (do_narrow): When sanitizing signed integer overflows,
17247         bail out for signed types.
17248         (convert_to_integer_1) <case NEGATE_EXPR>: Likewise.
17250 2017-09-04  Richard Biener  <rguenther@suse.de>
17252         PR tree-optimization/82060
17253         * tree-ssa-pre.c (eliminate_dom_walker::before_dom_children):
17254         Move devirtualization after stmt folding and before EH/AB/noreturn
17255         cleanup to get the stmt refs canonicalized.  Use a bool instead
17256         of gimple_modified_p since that doesn't work for NOPs.  Schedule
17257         NOPs generated by folding for removal.
17259 2017-09-04  Richard Sandiford  <richard.sandiford@linaro.org>
17260             Alan Hayward  <alan.hayward@arm.com>
17261             David Sherwood  <david.sherwood@arm.com>
17263         * coretypes.h (pad_direction): New enum.
17264         * defaults.h (DEFAULT_FUNCTION_ARG_PADDING): Delete.
17265         (FUNCTION_ARG_PADDING): Likewise.
17266         * target.def (function_arg_padding): New hook.
17267         * targhooks.h (default_function_arg_padding): Declare.
17268         * targhooks.c (default_function_arg_padding): New function.
17269         * doc/tm.texi.in (FUNCTION_ARG_PADDING): Replace with...
17270         (TARGET_FUNCTION_ARG_PADDING): ...this.
17271         * doc/tm.texi: Regenerate.
17272         * calls.c (store_unaligned_arguments_into_pseudos): Use pad_direction
17273         instead of direction.
17274         (compute_argument_addresses): Likewise.
17275         (load_register_parameters): Likewise.
17276         (emit_library_call_value_1): Likewise.
17277         (store_one_arg): Use targetm.calls.function_arg_padding instead
17278         of FUNCTION_ARG_PADDING.
17279         (must_pass_in_stack_var_size_or_pad): Likewise.
17280         * expr.c (emit_group_load_1): Use pad_direction instead of direction.
17281         (emit_group_store): Likewise.
17282         (emit_single_push_insn_1): Use targetm.calls.function_arg_padding
17283         instead of FUNCTION_ARG_PADDING.
17284         (emit_push_insn): Likewise, and propagate enum change throughout
17285         function.
17286         * function.h (direction): Delete.
17287         (locate_and_pad_arg_data::where_pad): Use pad_direction instead
17288         of direction.
17289         * function.c (assign_parm_find_stack_rtl): Likewise.
17290         (assign_parm_setup_block_p): Likewise.
17291         (assign_parm_setup_block): Likewise.
17292         (gimplify_parameters): Likewise.
17293         (locate_and_pad_parm): Use targetm.calls.function_arg_padding
17294         instead of FUNCTION_ARG_PADDING, and propagate enum change throughout
17295         function.
17296         * config/aarch64/aarch64.h (FUNCTION_ARG_PADDING): Delete.
17297         (BLOCK_REG_PADDING): Use pad_direction instead of direction.
17298         * config/aarch64/aarch64-protos.h (aarch64_pad_arg_upward): Delete.
17299         * config/aarch64/aarch64.c (aarch64_pad_arg_upward): Replace with...
17300         (aarch64_function_arg_padding): ...this new function.
17301         (aarch64_gimplify_va_arg_expr): Use pad_direction instead of direction.
17302         (TARGET_FUNCTION_ARG_PADDING): Redefine.
17303         * config/arm/arm.h (FUNCTION_ARG_PADDING): Delete.
17304         (BLOCK_REG_PADDING): Use pad_direction instead of direction.
17305         * config/arm/arm-protos.h (arm_pad_arg_upward): Delete.
17306         * config/arm/arm.c (TARGET_FUNCTION_ARG_PADDING): Redefine.
17307         (arm_pad_arg_upward): Replace with...
17308         (arm_function_arg_padding): ...this new function.
17309         * config/c6x/c6x.h (BLOCK_REG_PADDING): Use pad_direction instead
17310         of direction.
17311         * config/ia64/hpux.h (FUNCTION_ARG_PADDING): Delete.
17312         * config/ia64/ia64-protos.h (ia64_hpux_function_arg_padding): Delete.
17313         * config/ia64/ia64.c (TARGET_FUNCTION_ARG_PADDING): Redefine.
17314         (ia64_hpux_function_arg_padding): Replace with...
17315         (ia64_function_arg_padding): ...this new function.  Use pad_direction
17316         instead of direction.  Check for TARGET_HPUX.
17317         * config/iq2000/iq2000.h (FUNCTION_ARG_PADDING): Delete.
17318         * config/iq2000/iq2000.c (TARGET_FUNCTION_ARG_PADDING): Redefine.
17319         (iq2000_function_arg_padding): New function.
17320         * config/mips/mips-protos.h (mips_pad_arg_upward): Delete.
17321         * config/mips/mips.c (mips_pad_arg_upward): Replace with...
17322         (mips_function_arg_padding): ...this new function.
17323         (mips_pad_reg_upward): Update accordingly.
17324         (TARGET_FUNCTION_ARG_PADDING): Redefine.
17325         * config/mips/mips.h (PAD_VARARGS_DOWN): Use
17326         targetm.calls.function_arg_padding.
17327         (FUNCTION_ARG_PADDING): Delete.
17328         (BLOCK_REG_PADDING): Use pad_direction instead of direction.
17329         * config/nios2/nios2.h (FUNCTION_ARG_PADDING): Delete.
17330         (PAD_VARARGS_DOWN): Use targetm.calls.function_arg_padding.
17331         * config/nios2/nios2-protos.h (nios2_function_arg_padding): Delete.
17332         (nios2_block_reg_padding): Return pad_direction instead of direction.
17333         * config/nios2/nios2.c (nios2_block_reg_padding): Return pad_direction
17334         instead of direction.
17335         (nios2_function_arg_padding): Likewise.  Make static.
17336         (TARGET_FUNCTION_ARG_PADDING): Redefine.
17337         * config/pa/pa.h (FUNCTION_ARG_PADDING): Delete.
17338         (BLOCK_REG_PADDING): Use targetm.calls.function_arg_padding.
17339         * config/pa/pa-protos.h (pa_function_arg_padding): Delete.
17340         * config/pa/pa.c (TARGET_FUNCTION_ARG_PADDING): Redefine.
17341         (pa_function_arg_padding): Make static.  Return pad_direction instead
17342         of direction.
17343         * config/powerpcspe/powerpcspe.h (FUNCTION_ARG_PADDING): Delete.
17344         (PAD_VARARGS_DOWN): Use targetm.calls.function_arg_padding.
17345         * config/powerpcspe/aix.h (BLOCK_REG_PADDING): Use pad_direction
17346         instead of direction.  Use targetm.calls.function_arg_padding.
17347         * config/powerpcspe/darwin.h (BLOCK_REG_PADDING): Likewise.
17348         * config/powerpcspe/freebsd64.h (BLOCK_REG_PADDING): Likewise.
17349         * config/powerpcspe/linux64.h (BLOCK_REG_PADDING): Likewise.
17350         * config/powerpcspe/powerpcspe-protos.h (function_arg_padding): Delete.
17351         * config/powerpcspe/powerpcspe.c (TARGET_FUNCTION_ARG_PADDING):
17352         Redefine.
17353         (function_arg_padding): Rename to...
17354         (rs6000_function_arg_padding): ...this.  Make static.  Return
17355         pad_direction instead of direction.
17356         (rs6000_return_in_msb): Use rs6000_function_arg_padding.
17357         * config/rs6000/rs6000.h (FUNCTION_ARG_PADDING): Delete.
17358         (PAD_VARARGS_DOWN): Use targetm.calls.function_arg_padding.
17359         * config/rs6000/aix.h (BLOCK_REG_PADDING): Use pad_direction
17360         instead of direction.  Use targetm.calls.function_arg_padding.
17361         * config/rs6000/darwin.h (BLOCK_REG_PADDING): Likewise.
17362         * config/rs6000/freebsd64.h (BLOCK_REG_PADDING): Likewise.
17363         * config/rs6000/linux64.h (BLOCK_REG_PADDING): Likewise.
17364         * config/rs6000/rs6000-protos.h (function_arg_padding): Delete.
17365         * config/rs6000/rs6000.c (TARGET_FUNCTION_ARG_PADDING): Redefine.
17366         (function_arg_padding): Rename to...
17367         (rs6000_function_arg_padding): ...this.  Make static.  Return
17368         pad_direction instead of direction.
17369         (rs6000_return_in_msb): Use rs6000_function_arg_padding.
17370         * config/s390/s390.h (FUNCTION_ARG_PADDING): Delete.
17371         * config/s390/s390.c (s390_function_arg_padding): New function.
17372         (TARGET_FUNCTION_ARG_PADDING): Redefine.
17373         * config/sparc/sparc.h (FUNCTION_ARG_PADDING): Delete.
17374         * config/sparc/sparc-protos.h (function_arg_padding): Delete.
17375         * config/sparc/sparc.c (TARGET_FUNCTION_ARG_PADDING): Redefine.
17376         (function_arg_padding): Rename to...
17377         (sparc_function_arg_padding): ...this.  Make static.  Return
17378         pad_direction instead of direction.
17379         * config/spu/spu.h (FUNCTION_ARG_PADDING): Delete.
17380         * config/spu/spu.c (spu_function_arg_padding): New function.
17381         (TARGET_FUNCTION_ARG_PADDING): Redefine.
17382         * system.h (FUNCTION_ARG_PADDING): Poison.
17384 2017-09-04  Richard Sandiford  <richard.sandiford@linaro.org>
17385             Alan Hayward  <alan.hayward@arm.com>
17386             David Sherwood  <david.sherwood@arm.com>
17388         * target.def (modes_tieable_p): New hook.
17389         * doc/tm.texi (MODES_TIEABLE_P): Replace with...
17390         (TARGET_MODES_TIEABLE_P): ...this.
17391         * doc/tm.texi.in: Regenerate.
17392         * hooks.h (hook_bool_mode_mode_true): Declare.
17393         * hooks.c (hook_bool_mode_mode_true): New function.
17394         * combine.c (subst): Use targetm.modes_tieable_p instead of
17395         MODES_TIEABLE_P.
17396         * dse.c (find_shift_sequence): Likewise.
17397         * expmed.c (extract_low_bits): Likewise.
17398         * lower-subreg.c: Include target.h.
17399         (find_decomposable_subregs): Use targetm.modes_tieable_p instead of
17400         MODES_TIEABLE_P.
17401         * rtlanal.c (rtx_cost): Likewise.
17402         * config/aarch64/aarch64.h (MODES_TIEABLE_P): Delete.
17403         * config/aarch64/aarch64-protos.h (aarch64_modes_tieable_p): Delete.
17404         * config/aarch64/aarch64.c (aarch64_modes_tieable_p): Make static.
17405         (TARGET_MODES_TIEABLE_P): Redefine.
17406         * config/alpha/alpha.h (MODES_TIEABLE_P): Delete.
17407         * config/alpha/alpha.c (alpha_modes_tieable_p): New function.
17408         (TARGET_MODES_TIEABLE_P): Redefine.
17409         * config/arc/arc.h (MODES_TIEABLE_P): Delete.
17410         * config/arc/arc.c (TARGET_MODES_TIEABLE_P): Redefine.
17411         (arc_modes_tieable_p): New function.
17412         * config/arm/arm.h (MODES_TIEABLE_P): Delete.
17413         * config/arm/arm-protos.h (arm_modes_tieable_p): Delete.
17414         * config/arm/arm.c (TARGET_MODES_TIEABLE_P): Redefine.
17415         (arm_modes_tieable_p): Make static.
17416         * config/avr/avr.h (MODES_TIEABLE_P): Delete.
17417         * config/bfin/bfin.h (MODES_TIEABLE_P): Delete.
17418         * config/bfin/bfin.c (bfin_modes_tieable_p): New function.
17419         (TARGET_MODES_TIEABLE_P): Redefine.
17420         * config/c6x/c6x.h (MODES_TIEABLE_P): Delete.
17421         * config/c6x/c6x.c (c6x_modes_tieable_p): New function.
17422         (TARGET_MODES_TIEABLE_P): Redefine.
17423         * config/cr16/cr16.h (MODES_TIEABLE_P): Delete.
17424         * config/cr16/cr16.c (TARGET_MODES_TIEABLE_P): Redefine.
17425         (cr16_modes_tieable_p): New function.
17426         * config/cris/cris.h (MODES_TIEABLE_P): Delete.
17427         * config/epiphany/epiphany.h (MODES_TIEABLE_P): Delete.
17428         * config/fr30/fr30.h (MODES_TIEABLE_P): Delete.
17429         (TRULY_NOOP_TRUNCATION): Update comment.
17430         * config/frv/frv.h (MODES_TIEABLE_P): Delete.
17431         (TRULY_NOOP_TRUNCATION): Update comment.
17432         * config/frv/frv.c (TARGET_MODES_TIEABLE_P): Redefine.
17433         (frv_modes_tieable_p): New function.
17434         * config/ft32/ft32.h (MODES_TIEABLE_P): Delete.
17435         * config/h8300/h8300.h (MODES_TIEABLE_P): Delete.
17436         * config/h8300/h8300.c (h8300_modes_tieable_p): New function.
17437         (TARGET_MODES_TIEABLE_P): Redefine.
17438         * config/i386/i386.h (MODES_TIEABLE_P): Delete.
17439         * config/i386/i386-protos.h (ix86_modes_tieable_p): Delete.
17440         * config/i386/i386.c (ix86_modes_tieable_p): Make static.
17441         (TARGET_MODES_TIEABLE_P): Redefine.
17442         * config/ia64/ia64.h (MODES_TIEABLE_P): Delete.
17443         * config/ia64/ia64.c (TARGET_MODES_TIEABLE_P): Redefine.
17444         (ia64_modes_tieable_p): New function.
17445         * config/iq2000/iq2000.h (MODES_TIEABLE_P): Delete.
17446         * config/iq2000/iq2000.c (TARGET_MODES_TIEABLE_P): Redefine.
17447         (iq2000_modes_tieable_p): New function.
17448         * config/lm32/lm32.h (MODES_TIEABLE_P): Delete.
17449         * config/lm32/lm32.c (TARGET_MODES_TIEABLE_P): Redefine.
17450         (lm32_modes_tieable_p): New function.
17451         * config/m32c/m32c.h (MODES_TIEABLE_P): Delete.
17452         * config/m32c/m32c-protos.h (m32c_modes_tieable_p): Delete.
17453         * config/m32c/m32c.c (m32c_modes_tieable_p): Make static.
17454         (TARGET_MODES_TIEABLE_P): Redefine.
17455         * config/m32r/m32r.h (MODES_TIEABLE_P): Delete.
17456         * config/m32r/m32r.c (TARGET_MODES_TIEABLE_P): Redefine.
17457         (m32r_modes_tieable_p): New function.
17458         * config/m68k/m68k.h (MODES_TIEABLE_P): Delete.
17459         * config/m68k/m68k.c (TARGET_MODES_TIEABLE_P): Redefine.
17460         (m68k_modes_tieable_p): New function.
17461         * config/mcore/mcore.h (MODES_TIEABLE_P): Delete.
17462         * config/mcore/mcore.c (TARGET_MODES_TIEABLE_P): Redefine.
17463         (mcore_modes_tieable_p): New function.
17464         * config/microblaze/microblaze.h (MODES_TIEABLE_P): Delete.
17465         * config/microblaze/microblaze.c (microblaze_modes_tieable_p): New
17466         function.
17467         (TARGET_MODES_TIEABLE_P): Redefine.
17468         * config/mips/mips.h (MODES_TIEABLE_P): Delete.
17469         * config/mips/mips-protos.h (mips_modes_tieable_p): Delete.
17470         * config/mips/mips.c (mips_modes_tieable_p): Make static.
17471         (TARGET_MODES_TIEABLE_P): Redefine.
17472         * config/mmix/mmix.h (MODES_TIEABLE_P): Delete.
17473         * config/mn10300/mn10300.h (MODES_TIEABLE_P): Delete.
17474         * config/mn10300/mn10300-protos.h (mn10300_modes_tieable): Delete.
17475         * config/mn10300/mn10300.c (mn10300_modes_tieable): Rename to...
17476         (mn10300_modes_tieable_p): ...this and make static.
17477         (TARGET_MODES_TIEABLE_P): Redefine.
17478         * config/moxie/moxie.h (MODES_TIEABLE_P): Delete.
17479         * config/msp430/msp430.h (MODES_TIEABLE_P): Delete.
17480         * config/msp430/msp430-protos.h (msp430_modes_tieable_p): Delete.
17481         * config/msp430/msp430.c (TARGET_MODES_TIEABLE_P): Redefine.
17482         (msp430_modes_tieable_p): Make static.
17483         * config/nds32/nds32.h (MODES_TIEABLE_P): Delete.
17484         * config/nds32/nds32.c (nds32_modes_tieable_p): New function.
17485         (TARGET_MODES_TIEABLE_P): Redefine.
17486         * config/nios2/nios2.h (MODES_TIEABLE_P): Delete.
17487         * config/nvptx/nvptx.h (MODES_TIEABLE_P): Delete.
17488         * config/nvptx/nvptx.c (nvptx_modes_tieable_p): New function.
17489         (TARGET_MODES_TIEABLE_P): Redefine.
17490         * config/pa/pa.h (MODES_TIEABLE_P): Delete.
17491         * config/pa/pa-protos.h (pa_modes_tieable_p): Delete.
17492         * config/pa/pa.c (pa_modes_tieable_p): Make static.
17493         (TARGET_MODES_TIEABLE_P): Redefine.
17494         * config/pdp11/pdp11.h (MODES_TIEABLE_P): Delete.
17495         * config/pdp11/pdp11.c (TARGET_MODES_TIEABLE_P): Redefine.
17496         (pdp11_modes_tieable_p): New function.
17497         * config/powerpcspe/powerpcspe.h (MODES_TIEABLE_P): Delete.
17498         * config/powerpcspe/powerpcspe.c (TARGET_MODES_TIEABLE_P): Redefine.
17499         (rs6000_modes_tieable_p): New function.
17500         (rs6000_debug_reg_global): Use it instead of MODES_TIEABLE_P.
17501         * config/powerpcspe/powerpcspe.md: Update comment.
17502         * config/riscv/riscv.h (MODES_TIEABLE_P): Delete.
17503         * config/riscv/riscv.c (riscv_modes_tieable_p): New function.
17504         (TARGET_MODES_TIEABLE_P): Redefine.
17505         * config/rl78/rl78.h (MODES_TIEABLE_P): Delete.
17506         * config/rl78/rl78.c (TARGET_MODES_TIEABLE_P): Redefine.
17507         (rl78_modes_tieable_p): New function.
17508         * config/rs6000/rs6000.h (MODES_TIEABLE_P): Delete.
17509         * config/rs6000/rs6000.c (TARGET_MODES_TIEABLE_P): Redefine.
17510         (rs6000_modes_tieable_p): New function.
17511         (rs6000_debug_reg_global): Use it instead of MODES_TIEABLE_P.
17512         * config/rs6000/rs6000.md: Update comment.
17513         * config/rx/rx.h (MODES_TIEABLE_P): Delete.
17514         * config/rx/rx.c (rx_modes_tieable_p): New function.
17515         (TARGET_MODES_TIEABLE_P): Redefine.
17516         * config/s390/s390.h (MODES_TIEABLE_P): Delete.
17517         * config/s390/s390.c (s390_modes_tieable_p): New function.
17518         (TARGET_MODES_TIEABLE_P): Redefine.
17519         * config/sh/sh.h (MODES_TIEABLE_P): Delete.
17520         * config/sh/sh.c (TARGET_MODES_TIEABLE_P): Redefine.
17521         (sh_modes_tieable_p): New function.
17522         * config/sparc/sparc.h (MODES_TIEABLE_P): Delete.
17523         * config/sparc/sparc-protos.h (sparc_modes_tieable_p): Delete.
17524         * config/sparc/sparc.c (TARGET_MODES_TIEABLE_P): Redefine.
17525         (sparc_modes_tieable_p): Make static.
17526         * config/spu/spu.h (MODES_TIEABLE_P): Delete.
17527         * config/spu/spu.c (spu_modes_tieable_p): New function.
17528         (TARGET_MODES_TIEABLE_P): Redefine.
17529         * config/stormy16/stormy16.h (MODES_TIEABLE_P): Delete.
17530         * config/stormy16/stormy16.c (xstormy16_modes_tieable_p): New function.
17531         (TARGET_MODES_TIEABLE_P): Redefine.
17532         * config/tilegx/tilegx.h (MODES_TIEABLE_P): Delete.
17533         * config/tilepro/tilepro.h (MODES_TIEABLE_P): Delete.
17534         * config/v850/v850.h (MODES_TIEABLE_P): Delete.
17535         * config/v850/v850.c (v850_modes_tieable_p): New function.
17536         (TARGET_MODES_TIEABLE_P): Redefine.
17537         * config/vax/vax.h (MODES_TIEABLE_P): Delete.
17538         * config/visium/visium.h (MODES_TIEABLE_P): Delete.
17539         * config/visium/visium.c (TARGET_MODES_TIEABLE_P): Redefine.
17540         (visium_modes_tieable_p): New function.
17541         * config/xtensa/xtensa.h (MODES_TIEABLE_P): Delete.
17542         * config/xtensa/xtensa.c (TARGET_MODES_TIEABLE_P): Redefine.
17543         (xtensa_modes_tieable_p): New function.
17544         * system.h (MODES_TIEABLE_P): Poison.
17546 2017-09-04  Richard Sandiford  <richard.sandiford@linaro.org>
17547             Alan Hayward  <alan.hayward@arm.com>
17548             David Sherwood  <david.sherwood@arm.com>
17550         * target.def (hard_regno_mode_ok): New hook.
17551         * doc/tm.texi (HARD_REGNO_MODE_OK): Replace with...
17552         (TARGET_HARD_REGNO_MODE_OK): ...this.
17553         * doc/tm.texi.in: Regenerate.
17554         * hooks.h (hook_bool_uint_mode_true): Declare.
17555         * hooks.c (hook_bool_uint_mode_true): New function.
17556         * doc/md.texi: Refer to targetm.hard_regno_mode_ok instead of
17557         HARD_REGNO_MODE_OK.
17558         * genpreds.c (write_insn_preds_c): Add an include of target.h.
17559         * alias.c (init_alias_target): Use targetm.hard_regno_mode_ok
17560         instead of HARD_REGNO_MODE_OK.
17561         * caller-save.c: Include target.h.
17562         (reg_save_code): Use targetm.hard_regno_mode_ok instead of
17563         HARD_REGNO_MODE_OK.
17564         * combine.c (can_combine_p): Likewise.
17565         (combinable_i3pat): Likewise.
17566         (can_change_dest_mode): Likewise.
17567         * expr.c (init_expr_target): Likewise.
17568         (convert_move): Likewise.
17569         (convert_modes): Likewise.
17570         * ira.c (setup_prohibited_class_mode_regs): Likewise.
17571         (setup_prohibited_mode_move_regs): Likewise.
17572         * ira.h (target_ira): Likewise.
17573         * lra-assigns.c (find_hard_regno_for_1): Likewise.
17574         * lra-constraints.c (process_alt_operands): Likewise.
17575         (split_reg): Likewise.
17576         * recog.c (peep2_find_free_register): Likewise.
17577         * ree.c (combine_reaching_defs): Likewise.
17578         * regcprop.c (maybe_mode_change): Likewise.
17579         * reginfo.c (init_reg_sets_1): Likewise.
17580         (choose_hard_reg_mode): Likewise.
17581         (simplifiable_subregs): Likewise.
17582         * regrename.c (check_new_reg_p): Likewise.
17583         * reload.c (find_valid_class): Likewise.
17584         (find_valid_class_1): Likewise.
17585         (reload_inner_reg_of_subreg): Likewise.
17586         (push_reload): Likewise.
17587         (combine_reloads): Likewise.
17588         (find_dummy_reload): Likewise.
17589         (find_reloads): Likewise.
17590         * reload1.c (find_reg): Likewise.
17591         (set_reload_reg): Likewise.
17592         (allocate_reload_reg): Likewise.
17593         (choose_reload_regs): Likewise.
17594         (reload_adjust_reg_for_temp): Likewise.
17595         * rtlanal.c (subreg_size_offset_from_lsb): Likewise.
17596         (simplify_subreg_regno): Likewise.
17597         * sel-sched.c (init_regs_for_mode): Likewise.
17598         * varasm.c (make_decl_rtl): Likewise.
17599         * config/aarch64/aarch64.h (HARD_REGNO_MODE_OK): Delete.
17600         (MODES_TIEABLE_P): Use targetm.hard_regno_mode_ok instead of
17601         HARD_REGNO_MODE_OK.
17602         * config/aarch64/aarch64-protos.h (aarch64_hard_regno_mode_ok): Delete.
17603         * config/aarch64/aarch64.c (aarch64_hard_regno_mode_ok): Make static.
17604         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17605         * config/alpha/alpha.h (HARD_REGNO_MODE_OK): Delete.
17606         * config/alpha/alpha.c (alpha_hard_regno_mode_ok): New function.
17607         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17608         * config/arc/arc.h (arc_hard_regno_mode_ok): Delete.
17609         (arc_mode_class): Delete.
17610         (HARD_REGNO_MODE_OK): Delete.
17611         * config/arc/arc.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17612         (arc_hard_regno_mode_ok): Rename old array to...
17613         (arc_hard_regno_mode_ok_modes): ...this.
17614         (arc_conditional_register_usage): Update accordingly.
17615         (arc_mode_class): Make static.
17616         (arc_hard_regno_mode_ok): New function.
17617         * config/arm/arm.h (HARD_REGNO_MODE_OK): Delete.
17618         * config/arm/arm-protos.h (arm_hard_regno_mode_ok): Delete.
17619         * config/arm/arm.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17620         (arm_hard_regno_mode_ok): Make static.
17621         * config/arm/arm.md (movdi): Use targetm.hard_regno_mode_ok instead of
17622         HARD_REGNO_MODE_OK.
17623         * config/avr/avr-protos.h (avr_hard_regno_mode_ok): Delete.
17624         * config/avr/avr.h (HARD_REGNO_MODE_OK): Delete.
17625         * config/avr/avr.c (avr_hard_regno_mode_ok): Make static and
17626         return a bool.
17627         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17628         * config/bfin/bfin-protos.h (hard_regno_mode_ok): Delete.
17629         * config/bfin/bfin.h (HARD_REGNO_MODE_OK): Delete.
17630         * config/bfin/bfin.c (hard_regno_mode_ok): Rename to...
17631         (bfin_hard_regno_mode_ok): ...this.  Make static and return a bool.
17632         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17633         * config/bfin/predicates.md (valid_reg_operand): Use
17634         targetm.hard_regno_mode_ok instead of HARD_REGNO_MODE_OK.
17635         * config/c6x/c6x.h (HARD_REGNO_MODE_OK): Delete.
17636         * config/c6x/c6x.c (c6x_hard_regno_mode_ok): New function.
17637         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17638         * config/cr16/cr16.h (HARD_REGNO_MODE_OK): Delete.
17639         * config/cr16/cr16-protos.h (cr16_hard_regno_mode_ok): Delete.
17640         * config/cr16/cr16.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17641         (cr16_hard_regno_mode_ok): Make static and return a bool.
17642         * config/cris/cris.h (HARD_REGNO_MODE_OK): Delete.
17643         * config/cris/cris.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17644         (cris_hard_regno_mode_ok): New function.
17645         * config/epiphany/epiphany.h (epiphany_hard_regno_mode_ok): Delete.
17646         (epiphany_mode_class): Delete.
17647         (HARD_REGNO_MODE_OK): Delete.
17648         * config/epiphany/epiphany-protos.h (hard_regno_mode_ok): Delete.
17649         * config/epiphany/epiphany.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17650         (hard_regno_mode_ok): Rename to...
17651         (epiphany_hard_regno_mode_ok): ...this.  Make static and return a bool.
17652         * config/fr30/fr30.h (HARD_REGNO_MODE_OK): Delete.
17653         * config/fr30/fr30.md: Refer to targetm.hard_regno_mode_ok instead of
17654         HARD_REGNO_MODE_OK.
17655         * config/frv/frv.h (HARD_REGNO_MODE_OK): Delete.
17656         * config/frv/frv-protos.h (frv_hard_regno_mode_ok): Delete.
17657         * config/frv/frv.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17658         (frv_hard_regno_mode_ok): Make static and return a bool.
17659         * config/frv/frv.md: Refer to targetm.hard_regno_mode_ok instead of
17660         HARD_REGNO_MODE_OK.
17661         * config/ft32/ft32.h (HARD_REGNO_MODE_OK): Delete.
17662         * config/h8300/h8300.h (HARD_REGNO_MODE_OK): Delete.
17663         * config/h8300/h8300-protos.h (h8300_hard_regno_mode_ok): Delete.
17664         * config/h8300/h8300.c (h8300_hard_regno_mode_ok): Make static
17665         and return a bool.
17666         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17667         * config/i386/i386.h (HARD_REGNO_MODE_OK): Delete.
17668         * config/i386/i386-protos.h (ix86_hard_regno_mode_ok): Delete.
17669         * config/i386/i386.c (ix86_hard_regno_mode_ok): Make static and
17670         return a bool.
17671         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17672         * config/ia64/ia64.h (HARD_REGNO_MODE_OK): Delete.
17673         * config/ia64/ia64.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17674         (ia64_hard_regno_mode_ok): New function.
17675         * config/iq2000/iq2000.h (HARD_REGNO_MODE_OK): Delete.
17676         * config/iq2000/iq2000.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17677         (iq2000_hard_regno_mode_ok): New function.
17678         * config/lm32/lm32.h (HARD_REGNO_MODE_OK): Delete.
17679         * config/lm32/lm32.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17680         (lm32_hard_regno_mode_ok): New function.
17681         * config/m32c/m32c.h (HARD_REGNO_MODE_OK): Delete.
17682         * config/m32c/m32c-protos.h (m32c_hard_regno_ok): Delete.
17683         * config/m32c/m32c.c (class_can_hold_mode): Use m32c_hard_regno_mode_ok
17684         instead of HARD_REGNO_MODE_OK.
17685         (m32c_hard_regno_ok): Rename to...
17686         (m32c_hard_regno_mode_ok): ...this.  Make static and return a bool.
17687         (m32c_cannot_change_mode_class): Update accordingly.
17688         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17689         * config/m32r/m32r.h (m32r_hard_regno_mode_ok): Delete.
17690         (m32r_mode_class): Delete.
17691         (HARD_REGNO_MODE_OK): Delete.
17692         * config/m32r/m32r.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17693         (m32r_hard_regno_mode_ok): Rename to...
17694         (m32r_hard_regno_modes): ...this.
17695         (m32r_mode_class): Make static.
17696         (m32r_hard_regno_mode_ok): New function.
17697         * config/m68k/m68k.h (HARD_REGNO_MODE_OK): Delete.
17698         * config/m68k/m68k-protos.h (m68k_regno_mode_ok): Delete.
17699         * config/m68k/m68k.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17700         (m68k_hard_regno_mode_ok): Make static.
17701         * config/mcore/mcore.h (HARD_REGNO_MODE_OK): Delete.
17702         * config/mcore/mcore.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17703         (mcore_hard_regno_mode_ok): New function.
17704         * config/microblaze/microblaze.h (microblaze_hard_regno_mode_ok)
17705         (HARD_REGNO_MODE_OK): Delete.
17706         * config/microblaze/microblaze.c (microblaze_hard_regno_mode_ok):
17707         Rename to...
17708         (microblaze_hard_regno_mode_ok_p): ...this and make static.
17709         (microblaze_hard_regno_mode_ok): New function.
17710         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17711         * config/mips/mips.h (HARD_REGNO_MODE_OK): Delete.
17712         (mips_hard_regno_mode_ok): Delete.
17713         * config/mips/mips.c (mips_hard_regno_mode_ok): Rename to...
17714         (mips_hard_regno_mode_ok_p): ...this and make static.
17715         (mips_hard_regno_mode_ok_p): Rename to...
17716         (mips_hard_regno_mode_ok_uncached): ...this.
17717         (mips_hard_regno_mode_ok): New function.
17718         (mips_class_max_nregs): Use mips_hard_regno_mode_ok instead
17719         of HARD_REGNO_MODE_OK.
17720         (mips_option_override): Update after above name changes.
17721         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17722         * config/mmix/mmix.h (HARD_REGNO_MODE_OK): Delete.
17723         * config/mn10300/mn10300.h (HARD_REGNO_MODE_OK): Delete.
17724         * config/mn10300/mn10300-protos.h (mn10300_hard_regno_mode_ok): Delete.
17725         * config/mn10300/mn10300.c (mn10300_hard_regno_mode_ok): Make static.
17726         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17727         * config/moxie/moxie.h (HARD_REGNO_MODE_OK): Delete.
17728         * config/msp430/msp430.h (HARD_REGNO_MODE_OK): Delete.
17729         * config/msp430/msp430-protos.h (msp430_hard_regno_mode_ok): Delete.
17730         * config/msp430/msp430.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17731         (msp430_hard_regno_mode_ok): Make static and return a bool.
17732         * config/nds32/nds32.h (HARD_REGNO_MODE_OK): Delete.
17733         * config/nds32/nds32-protos.h (nds32_hard_regno_mode_ok): Delete.
17734         * config/nds32/nds32.c (nds32_hard_regno_mode_ok): Make static
17735         and return a bool.
17736         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17737         * config/nios2/nios2.h (HARD_REGNO_MODE_OK): Delete.
17738         * config/nvptx/nvptx.h (HARD_REGNO_MODE_OK): Delete.
17739         * config/pa/pa.h (MODES_TIEABLE_P): Update commentary.
17740         * config/pa/pa32-regs.h (HARD_REGNO_MODE_OK): Rename to...
17741         (PA_HARD_REGNO_MODE_OK): ...this
17742         * config/pa/pa64-regs.h (HARD_REGNO_MODE_OK): Rename to...
17743         (PA_HARD_REGNO_MODE_OK): ...this.
17744         * config/pa/pa.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17745         (pa_hard_regno_mode_ok): New function.
17746         * config/pdp11/pdp11.h (HARD_REGNO_MODE_OK): Delete.
17747         * config/pdp11/pdp11.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17748         (pdp11_hard_regno_mode_ok): New function.
17749         * config/powerpcspe/powerpcspe.h (HARD_REGNO_MODE_OK): Delete.
17750         * config/powerpcspe/powerpcspe-protos.h (rs6000_hard_regno_mode_ok_p):
17751         Delete.
17752         * config/powerpcspe/powerpcspe.c (rs6000_hard_regno_mode_ok_p):
17753         Make static.
17754         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17755         (rs6000_hard_regno_mode_ok): Rename to...
17756         (rs6000_hard_regno_mode_ok_uncached): ...this.
17757         (rs6000_init_hard_regno_mode_ok): Update accordingly.
17758         (rs6000_hard_regno_mode_ok): New function.
17759         * config/riscv/riscv.h (HARD_REGNO_MODE_OK): Delete.
17760         * config/riscv/riscv-protos.h (riscv_hard_regno_mode_ok_p): Delete.
17761         * config/riscv/riscv.c (riscv_hard_regno_mode_ok_p): Rename to...
17762         (riscv_hard_regno_mode_ok): ...this and make static.
17763         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17764         * config/rl78/rl78.h (HARD_REGNO_MODE_OK): Delete.
17765         * config/rl78/rl78-protos.h (rl78_hard_regno_mode_ok): Delete.
17766         * config/rl78/rl78.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17767         (rl78_hard_regno_mode_ok): Make static and return bool.
17768         * config/rs6000/rs6000.h (HARD_REGNO_MODE_OK): Delete.
17769         * config/rs6000/rs6000-protos.h (rs6000_hard_regno_mode_ok_p):
17770         Delete.
17771         * config/rs6000/rs6000.c (rs6000_hard_regno_mode_ok_p): Make static.
17772         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17773         (rs6000_hard_regno_mode_ok): Rename to...
17774         (rs6000_hard_regno_mode_ok_uncached): ...this.
17775         (rs6000_init_hard_regno_mode_ok): Update accordingly.
17776         (rs6000_hard_regno_mode_ok): New function.
17777         * config/rx/rx.h (HARD_REGNO_MODE_OK): Delete.
17778         * config/rx/rx.c (rx_hard_regno_mode_ok): New function.
17779         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17780         * config/s390/s390.h (HARD_REGNO_MODE_OK): Delete.
17781         * config/s390/s390-protos.h (s390_hard_regno_mode_ok): Delete.
17782         * config/s390/s390.c (s390_hard_regno_mode_ok): Make static.
17783         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17784         * config/sh/sh.h (HARD_REGNO_MODE_OK): Delete.
17785         * config/sh/sh-protos.h (sh_hard_regno_mode_ok): Delete.
17786         * config/sh/sh.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17787         (sh_hard_regno_mode_ok): Make static.
17788         * config/sparc/constraints.md: Refer to targetm.hard_regno_mode_ok
17789         instead of HARD_REGNO_MODE_OK.
17790         * config/sparc/sparc.h (hard_regno_mode_classes): Delete.
17791         (sparc_mode_class): Delete.
17792         (HARD_REGNO_MODE_OK): Delete.
17793         * config/sparc/sparc.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17794         (hard_regno_mode_classes): Make static.
17795         (sparc_mode_class): Likewise.
17796         (sparc_hard_regno_mode_ok): New function.
17797         * config/spu/spu.h (HARD_REGNO_MODE_OK): Delete.
17798         * config/stormy16/stormy16.h (HARD_REGNO_MODE_OK): Delete.
17799         * config/stormy16/stormy16.c (xstormy16_hard_regno_mode_ok): New
17800         function.
17801         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17802         * config/tilegx/tilegx.h (HARD_REGNO_MODE_OK): Delete.
17803         * config/tilepro/tilepro.h (HARD_REGNO_MODE_OK): Delete.
17804         * config/v850/v850.h (HARD_REGNO_MODE_OK): Delete.
17805         * config/v850/v850.c (v850_hard_regno_mode_ok): New function.
17806         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17807         * config/vax/vax.h (HARD_REGNO_MODE_OK): Delete.
17808         * config/visium/visium.h (HARD_REGNO_MODE_OK): Delete.
17809         * config/visium/visium.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17810         (visium_hard_regno_mode_ok): New function.
17811         * config/visium/visium.md: Refer to targetm.hard_regno_mode_ok
17812         instead of HARD_REGNO_MODE_OK.
17813         * config/xtensa/xtensa.h (xtensa_hard_regno_mode_ok): Delete.
17814         (HARD_REGNO_MODE_OK): Delete.
17815         * config/xtensa/xtensa.c (xtensa_hard_regno_mode_ok): Rename to...
17816         (xtensa_hard_regno_mode_ok_p): ...this and make static.
17817         (xtensa_option_override): Update accordingly.
17818         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17819         (xtensa_hard_regno_mode_ok): New function.
17820         * system.h (HARD_REGNO_MODE_OK): Poison.
17822 2017-09-04  Richard Sandiford  <richard.sandiford@linaro.org>
17823             Alan Hayward  <alan.hayward@arm.com>
17824             David Sherwood  <david.sherwood@arm.com>
17826         * target.def (hard_regno_call_part_clobbered): New hook.
17827         * doc/tm.texi.in (HARD_REGNO_CALL_PART_CLOBBERED): Replace with...
17828         (TARGET_HARD_REGNO_CALL_PART_CLOBBERED): ...this hook.
17829         * doc/tm.texi: Regenerate.
17830         * hooks.h (hook_bool_uint_mode_false): Declare.
17831         * hooks.c (hook_bool_uint_mode_false): New function.
17832         * regs.h (HARD_REGNO_CALL_PART_CLOBBERED): Delete.
17833         * cselib.c (cselib_process_insn): Use
17834         targetm.hard_regno_call_part_clobbered instead of
17835         HARD_REGNO_CALL_PART_CLOBBERED.
17836         * ira-conflicts.c (ira_build_conflicts): Likewise.
17837         * ira-costs.c (ira_tune_allocno_costs): Likewise.
17838         * lra-constraints.c (need_for_call_save_p): Likewise.
17839         * lra-lives.c: Include target.h.
17840         (check_pseudos_live_through_calls): Use
17841         targetm.hard_regno_call_part_clobbered instead of
17842         HARD_REGNO_CALL_PART_CLOBBERED.
17843         * regcprop.c: Include target.h.
17844         (copyprop_hardreg_forward_1): Use
17845         targetm.hard_regno_call_part_clobbered instead of
17846         HARD_REGNO_CALL_PART_CLOBBERED.
17847         * reginfo.c (choose_hard_reg_mode): Likewise.
17848         * regrename.c (check_new_reg_p): Likewise.
17849         * reload.c (find_equiv_reg): Likewise.
17850         * reload1.c (emit_reload_insns): Likewise.
17851         * sched-deps.c (deps_analyze_insn): Likewise.
17852         * sel-sched.c (init_regs_for_mode): Likewise.
17853         (mark_unavailable_hard_regs): Likewise.
17854         * targhooks.c (default_dwarf_frame_reg_mode): Likewise.
17855         * config/aarch64/aarch64.h (HARD_REGNO_CALL_PART_CLOBBERED): Delete.
17856         * config/aarch64/aarch64.c (aarch64_hard_regno_call_part_clobbered):
17857         New function.
17858         (TARGET_HARD_REGNO_CALL_PART_CLOBBERED): Redefine.
17859         * config/avr/avr.h (HARD_REGNO_CALL_PART_CLOBBERED): Delete.
17860         * config/avr/avr-protos.h (avr_hard_regno_call_part_clobbered):
17861         Delete.
17862         * config/avr/avr.c (avr_hard_regno_call_part_clobbered): Make static
17863         and return a bool.
17864         (TARGET_HARD_REGNO_CALL_PART_CLOBBERED): Redefine.
17865         * config/i386/i386.h (HARD_REGNO_CALL_PART_CLOBBERED): Delete.
17866         * config/i386/i386.c (ix86_hard_regno_call_part_clobbered): New
17867         function.
17868         (TARGET_HARD_REGNO_CALL_PART_CLOBBERED): Redefine.
17869         * config/mips/mips.h (HARD_REGNO_CALL_PART_CLOBBERED): Delete.
17870         * config/mips/mips.c (mips_hard_regno_call_part_clobbered): New
17871         function.
17872         (TARGET_HARD_REGNO_CALL_PART_CLOBBERED): Redefine.
17873         * config/powerpcspe/powerpcspe.h (HARD_REGNO_CALL_PART_CLOBBERED):
17874         Delete.
17875         * config/powerpcspe/powerpcspe.c
17876         (rs6000_hard_regno_call_part_clobbered): New function.
17877         (TARGET_HARD_REGNO_CALL_PART_CLOBBERED): Redefine.
17878         * config/rs6000/rs6000.h (HARD_REGNO_CALL_PART_CLOBBERED): Delete.
17879         * config/rs6000/rs6000.c (rs6000_hard_regno_call_part_clobbered):
17880         New function.
17881         (TARGET_HARD_REGNO_CALL_PART_CLOBBERED): Redefine.
17882         * config/s390/s390.h (HARD_REGNO_CALL_PART_CLOBBERED): Delete.
17883         * config/s390/s390.c (s390_hard_regno_call_part_clobbered): New
17884         function.
17885         (TARGET_HARD_REGNO_CALL_PART_CLOBBERED): Redefine.
17886         * config/sh/sh.h (HARD_REGNO_CALL_PART_CLOBBERED): Delete.
17887         * system.h (HARD_REGNO_CALL_PART_CLOBBERED): Poison.
17889 2017-09-04  Richard Sandiford  <richard.sandiford@linaro.org>
17890             Alan Hayward  <alan.hayward@arm.com>
17891             David Sherwood  <david.sherwood@arm.com>
17893         * rtl.h (subreg_memory_offset): Declare.
17894         * emit-rtl.c (subreg_memory_offset): New function.
17895         * expmed.c (store_bit_field_1): Use it.
17896         * expr.c (undefined_operand_subword_p): Likewise.
17897         * simplify-rtx.c (simplify_subreg): Likewise.
17899 2017-09-04  Alexander Monakov  <amonakov@ispras.ru>
17901         PR rtl-optimization/57448
17902         PR target/67458
17903         PR target/81316
17904         * optabs.c (expand_atomic_load): Place compiler memory barriers if
17905         using atomic_load pattern.
17906         (expand_atomic_store): Likewise.
17908 2017-09-04  Jakub Jelinek  <jakub@redhat.com>
17910         PR sanitizer/81981
17911         * gimple-fold.c (gimple_fold_call): Optimize away useless UBSAN_PTR
17912         and UBSAN_BOUNDS internal calls.  Clean up IFN_UBSAN_OBJECT_SIZE
17913         handling.  Use replace_call_with_value with NULL instead of
17914         gsi_replace, unlink_stmt_vdef and release_defs.
17916         * gdbhooks.py (OptMachineModePrinter.to_string): Use 8 spaces
17917         instead of tab.
17919         * lra-remat.c (reg_overlap_for_remat_p): Fix a pasto.
17921 2017-09-04  Richard Sandiford  <richard.sandiford@linaro.org>
17923         PR bootstrap/82045
17924         * rtl.h (emit_library_call_value_1): Declare.
17925         (emit_library_call): Replace declaration with a series of overloads.
17926         Remove the parameter count argument.
17927         (emit_library_call_value): Likewise.
17928         * calls.c (emit_library_call_value_1): Make global.  Replace varargs
17929         with an "rtx_mode_t *".
17930         (emit_library_call_value): Delete.
17931         (emit_library_call): Likewise.
17932         * asan.c (asan_emit_stack_protection): Update calls accordingly.
17933         (asan_emit_allocas_unpoison): Likewise.
17934         * builtins.c (expand_builtin_powi): Likewise.
17935         (expand_asan_emit_allocas_unpoison): Likewise.
17936         * cfgexpand.c (expand_main_function): Likewise.
17937         * config/aarch64/aarch64.c (aarch64_trampoline_init): Likewise.
17938         * config/aarch64/aarch64.h (PROFILE_HOOK): Likewise.
17939         * config/alpha/alpha.c (alpha_trampoline_init): Likewise.
17940         * config/arm/arm.c (arm_trampoline_init): Likewise.
17941         (arm_call_tls_get_addr): Likewise.
17942         (arm_expand_divmod_libfunc): Likewise.
17943         * config/bfin/bfin.md (umulsi3_highpart): Likewise.
17944         (smulsi3_highpart): Likewise.
17945         * config/c6x/c6x.c (c6x_initialize_trampoline): Likewise.
17946         (c6x_expand_compare): Likewise.
17947         (c6x_expand_movmem): Likewise.
17948         * config/frv/frv.c (frv_trampoline_init): Likewise.
17949         * config/i386/i386.c (ix86_trampoline_init): Likewise.
17950         (ix86_expand_divmod_libfunc): Likewise.
17951         * config/ia64/ia64.c (ia64_expand_tls_address): Likewise.
17952         (ia64_expand_compare): Likewise.
17953         (ia64_profile_hook): Likewise.
17954         * config/ia64/ia64.md (save_stack_nonlocal): Likewise.
17955         (nonlocal_goto): Likewise.
17956         (restore_stack_nonlocal): Likewise.
17957         * config/m32r/m32r.c (block_move_call): Likewise.
17958         (m32r_trampoline_init): Likewise.
17959         * config/m68k/linux.h (FINALIZE_TRAMPOLINE): Likewise.
17960         * config/m68k/m68k.c (m68k_call_tls_get_addr): Likewise.
17961         (m68k_call_m68k_read_tp): Likewise.
17962         * config/microblaze/microblaze.c (microblaze_call_tls_get_addr)
17963         (microblaze_expand_divide): Likewise.
17964         * config/mips/mips.h (mips_args): Likewise.
17965         * config/mips/sdemtk.h (mips_sync_icache): Likewise.
17966         (MIPS_ICACHE_SYNC): Likewise.
17967         * config/nios2/nios2.c (nios2_emit_expensive_div): Likewise.
17968         (nios2_trampoline_init): Likewise.
17969         * config/pa/pa.c (hppa_tls_call): Likewise.
17970         (pa_trampoline_init): Likewise.
17971         * config/pa/pa.md (canonicalize_funcptr_for_compare): Likewise.
17972         * config/powerpcspe/powerpcspe.c (rs6000_legitimize_tls_address)
17973         (expand_strn_compare): Likewise.
17974         (rs6000_generate_compare): Likewise.
17975         (rs6000_expand_float128_convert): Likewise.
17976         (output_profile_hook): Likewise.
17977         (rs6000_trampoline_init): Likewise.
17978         * config/powerpcspe/powerpcspe.md (neg<mode>2): Likewise.
17979         * config/riscv/riscv.h (PROFILE_HOOK): Likewise.
17980         * config/rs6000/rs6000-string.c (expand_strn_compare): Likewise.
17981         * config/rs6000/rs6000.c (rs6000_legitimize_tls_address): Likewise.
17982         (rs6000_generate_compare): Likewise.
17983         (rs6000_expand_float128_convert): Likewise.
17984         (output_profile_hook): Likewise.
17985         (rs6000_trampoline_init): Likewise.
17986         * config/rs6000/rs6000.md (neg<mode>2): Likewise.
17987         * config/sh/sh.c (sh_trampoline_init): Likewise.
17988         * config/sparc/sparc.c (emit_soft_tfmode_libcall): Likewise.
17989         (sparc_emit_float_lib_cmp): Likewise.
17990         (sparc32_initialize_trampoline): Likewise.
17991         (sparc64_initialize_trampoline): Likewise.
17992         (sparc_profile_hook): Likewise.
17993         * config/spu/spu.c (ea_load_store): Likewise.
17994         * config/spu/spu.md (floatunssidf2): Likewise.
17995         * config/tilegx/tilegx.c (tilegx_trampoline_init): Likewise.
17996         * config/tilepro/tilepro.c (tilepro_trampoline_init): Likewise.
17997         * config/visium/visium.c (expand_block_move_4): Likewise.
17998         (expand_block_move_2): Likewise.
17999         (expand_block_move_1): Likewise.
18000         (expand_block_set_4): Likewise.
18001         (expand_block_set_2): Likewise.
18002         (expand_block_set_1): Likewise.
18003         (visium_trampoline_init): Likewise.
18004         (visium_profile_hook): Likewise.
18005         * config/xtensa/xtensa.c (xtensa_expand_nonlocal_goto): Likewise.
18006         (xtensa_setup_frame_addresses): Likewise.
18007         (xtensa_trampoline_init): Likewise.
18008         * except.c (sjlj_emit_function_enter): Likewise.
18009         (sjlj_emit_function_exit): Likewise.
18010         * explow.c (allocate_dynamic_stack_space): Likewise.
18011         (probe_stack_range): Likewise.
18012         * expr.c (convert_mode_scalar): Likewise.
18013         * optabs.c (expand_binop): Likewise.
18014         (expand_twoval_binop_libfunc): Likewise.
18015         (expand_unop): Likewise.
18016         (prepare_cmp_insn): Likewise.
18017         (prepare_float_lib_cmp): Likewise.
18018         (expand_float): Likewise.
18019         (expand_fix): Likewise.
18020         (expand_fixed_convert): Likewise.
18021         (maybe_emit_sync_lock_test_and_set): Likewise.
18022         (expand_atomic_compare_and_swap): Likewise.
18023         (expand_mem_thread_fence): Likewise.
18024         (expand_atomic_fetch_op): Likewise.
18026 2017-09-03  Gerald Pfeifer  <gerald@pfeifer.com>
18028         * doc/generic.texi (OpenACC): Adjust URL.
18029         * doc/invoke.texi (C Dialect Options): Ditto.
18031 2017-09-03  Uros Bizjak  <ubizjak@gmail.com>
18033         * config/i386/i386.md (*bt<mode>): Use nonimmediate_operand
18034         predicate for operand 1.  Add (m,<S>) constraint.
18035         (*jcc_bt<mode>): Use nonimmediate_operand predicate for operand 1.
18036         Prevent memory operand 1 with register operand 2.
18038 2017-09-01  Segher Boessenkool  <segher@kernel.crashing.org>
18040         PR rtl-optimization/82024
18041         * combine.c (try_combine): If the combination result is a PARALLEL,
18042         and we only need to retain the SET in there that would be placed
18043         at I2, check that we can place that at I3 instead, before doing so.
18045 2017-09-01  Jakub Jelinek  <jakub@redhat.com>
18047         PR target/81766
18048         * config/i386/i386.c (ix86_init_large_pic_reg): Return label
18049         instead of void.
18050         (ix86_init_pic_reg): Remember label from ix86_init_large_pic_reg,
18051         if non-NULL and preceded by NOTE_INSN_BASIC_BLOCK, swap the note
18052         and label.
18054 2017-09-01  Joerg Sonnenberger  <joerg@bec.de>
18055             Jeff Law  <law@redhat.com>
18057         * varasm.c (bss_initializer_p): Do not put constants into .bss
18058         (categorize_decl_for_section): Handle bss_initializer_p returning
18059         false when DECL_INITIAL is NULL.
18061 2017-09-01  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
18063         PR target/82012
18064         * config/s390/s390.c (s390_can_inline_p): New function.
18066 2017-09-01  Jeff Law  <law@redhat.com>
18068         PR tree-optimization/82052
18069         * tree-ssa-scopedtables.c (avail_exprs_stack::lookup_avail_expr):
18070         Always initialize the returned slot after a hash table miss
18071         when INSERT is true.
18073 2017-09-01  Alexander Monakov  <amonakov@ispras.ru>
18075         * config/s390/s390.md (mem_signal_fence): Remove.
18076         * doc/md.texi (mem_signal_fence): Remove.
18077         * optabs.c (expand_mem_signal_fence): Remove uses of mem_signal_fence.
18078         Update comments.
18079         * target-insns.def (mem_signal_fence): Remove.
18081 2017-09-01  Jakub Jelinek  <jakub@redhat.com>
18083         PR sanitizer/81902
18084         * doc/invoke.texi: Document -fsanitize=pointer-overflow.
18086         PR sanitizer/81923
18087         * asan.c (create_odr_indicator): Strip name encoding from assembler
18088         name before appending it after __odr_asan_.
18090 2017-09-01  Martin Liska  <mliska@suse.cz>
18092         PR tree-optimization/82059
18093         * gimple-ssa-isolate-paths.c (isolate_path): Add profile and
18094         frequency only when an edge is redirected.
18096 2017-09-01  Claudiu Zissulescu  <claziss@synopsys.com>
18098         * config/arc/arc-c.c (__ARC_LPC_WIDTH__): Add builtin define.
18099         * config/arc/arc.c (ARC_MAX_LOOP_LENGTH): Define.
18100         (arc_conditional_register_usage): Remove ARC600 lp_count
18101         exception.
18102         (arc_file_start): Emit Tag_ARC_CPU_variation.
18103         (arc_can_use_doloop_p): New conditions to use ZOLs.
18104         (hwloop_fail): New function.
18105         (hwloop_optimize): Likewise.
18106         (hwloop_pattern_reg): Likewise.
18107         (arc_doloop_hooks): New struct, to be used with reorg_loops.
18108         (arc_reorg_loops): New function, calls reorg_loops.
18109         (arc_reorg): Call arc_reorg_loops.  Remove old ZOL handling.
18110         (arc600_corereg_hazard): Remove ZOL checking, case handled by
18111         hwloop_optimize.
18112         (arc_loop_hazard): Remove function, functionality moved into
18113         hwloop_optimize.
18114         (arc_hazard): Remove arc_loop_hazard call.
18115         (arc_adjust_insn_length): Remove ZOL handling, functionality moved
18116         into hwloop_optimize.
18117         (arc_label_align): Remove ZOL handling.
18118         * config/arc/arc.h (LOOP_ALIGN): Changed to 0.
18119         * config/arc/arc.md (doloop_begin): Remove pattern.
18120         (doloop_begin_i): Likewise.
18121         (doloop_end_i): Likewise.
18122         (doloop_fallback): Likewise.
18123         (doloop_fallback_m): Likewise.
18124         (doloop_end): Reimplement expand.
18125         (arc_lp): New pattern for LP instruction.
18126         (loop_end): New pattern.
18127         (loop_fail): Likewise.
18128         (decrement_and_branch_until_zero): Likewise.
18129         * config/arc/arc.opt (mlpc-width): New option.
18130         * doc/invoke.texi (mlpc-width): Document option.
18132 2017-09-01  Claudiu Zissulescu  <claziss@synopsys.com>
18134         * config/arc/arc.c (arc_ifcvt): Remove use of merge_blocks call.
18135         (arc_ccfsm_advance): Fix checking for delay slots.
18136         (arc_reorg): Add rtl dump after each call to arc_ifcvt.
18138 2017-09-01  Claudiu Zissulescu  <claziss@synopsys.com>
18140         * config/arc/arc.md (movqi_insn): Add stores to save constant long
18141         immediates.
18142         (movhi_insn): Update store instruction constraint which are saving
18143         6-bit short immediates.
18144         (movsi_insn): Consider also short scaled load operations.
18145         (zero_extendhisi2_i): Use Usd constraint instead of T.
18146         (extendhisi2_i): Add q constraint.
18147         (arc_clzsi2): Add type and length attributes.
18148         (arc_ctzsi2): Likewise.
18149         * config/arc/constraints.md (Usc): Update constraint, the
18150         assembler can parse two relocations for a single instruction.
18152 2017-09-01  Claudiu Zissulescu  <claziss@synopsys.com>
18154         * config/arc/arc.c (arc_use_anchors_for_symbol_p): New function.
18155         (TARGET_USE_ANCHORS_FOR_SYMBOL_P): Define.
18157 2017-08-31  Olivier Hainque  <hainque@adacore.com>
18159         * config.gcc (powerpc-wrs-vxworks|vxworksae|vxworksmils): Now
18160         match as powerpc-wrs-vxworks*.
18162 2017-08-31  James Greenhalgh  <james.greenhalgh@arm.com>
18164         * config/aarch64/aarch64-simd.md (aarch64_mla_elt_merge<mode>): Fix
18165         register constraint for by-element operand.
18166         (aarch64_mls_elt_merge<mode>): Likewise.
18168 2017-08-31  Claudiu Zissulescu  <claziss@synopsys.com>
18170         * config/arc/arc.c (arc_can_follow_jump): Check for short
18171         branches.
18173 2017-08-31  Claudiu Zissulescu  <claziss@synopsys.com>
18175         * config.gcc: Use g.opt for arc.
18176         * config/arc/arc.c (LEGITIMATE_SCALED_ADDRESS_P): Deleted,
18177         functionality moved to ...
18178         (legitimate_scaled_address_p): New function, ...here.
18179         (LEGITIMATE_SMALL_DATA_OFFSET_P): New define.
18180         (LEGITIMATE_SMALL_DATA_ADDRESS_P): Use the above define.
18181         (legitimate_offset_address_p): Delete TARGET_NO_SDATA_SET
18182         condition.
18183         (arc_override_options): Handle G option.
18184         (arc_output_pic_addr_const): Correct function definition.
18185         (arc_legitimate_address_p): Use legitimate_scaled_address_p.
18186         (arc_decl_anon_ns_mem_p): Delete.
18187         (arc_in_small_data_p): Overhaul this function to take into
18188         consideration the value given via G option.
18189         (arc_rewrite_small_data_1): Renamed and corrected old
18190         arc_rewrite_small_data function.
18191         (arc_rewrite_small_data): New function.
18192         (small_data_pattern): Don't use pic_offset_table_rtx.
18193         * config/arc/arc.h (CC1_SPEC): Recognize G option.
18194         * config/arc/simdext.md (movmisalignv2hi): Use
18195         prepare_move_operands function.
18196         (mov*): Likewise.
18197         (movmisalign*): Likewise.
18198         * doc/invoke.texi (ARC options): Document -G option.
18200 2017-08-31  Claudiu Zissulescu  <claziss@synopsys.com>
18202         * config/arc/arc-protos.h (compact_sda_memory_operand): Update
18203         prototype.
18204         * config/arc/arc.c (arc_print_operand): Output scalled address for
18205         sdata whenever is possible.
18206         (arc_in_small_data_p): Allow sdata for 64bit datum when double
18207         load/stores are available.
18208         (compact_sda_memory_operand): Check for the alignment required by
18209         code density instructions.
18210         * config/arc/arc.md (movsi_insn): Use newly introduced Us0
18211         constraint.
18212         * config/arc/constraints.md (Usd): Update constraint.
18213         (Us0): New constraint.
18214         (Usc): Update constraint.
18216 2017-08-31  Richard Biener  <rguenther@suse.de>
18218         PR middle-end/82054
18219         * dwarf2out.c (dwarf2out_early_global_decl): Process each
18220         function only once.
18222 2017-08-31  Tamar Christina  <tamar.christina@arm.com>
18224         * config/aarch64/aarch64-builtins.c (aarch64_init_simd_builtins):
18225         Resize type_signature.
18227 2017-08-31  Richard Sandiford  <richard.sandiford@linaro.org>
18228             Alan Hayward  <alan.hayward@arm.com>
18229             David Sherwood  <david.sherwood@arm.com>
18231         * config/aarch64/aarch64.c (aarch64_base_register_rtx_p): Only allow
18232         subregs whose inner modes can be stored in GPRs.
18233         (aarch64_classify_index): Likewise.
18235 2017-08-31  Richard Sandiford  <richard.sandiford@linaro.org>
18236             Alan Hayward  <alan.hayward@arm.com>
18237             David Sherwood  <david.sherwood@arm.com>
18239         * config/aarch64/iterators.md (V_cmp_result): Rename to...
18240         (V_INT_EQUIV): ...this.
18241         (v_cmp_result): Rename to...
18242         (v_int_equiv): ...this.
18243         * config/aarch64/aarch64.md (xorsign<mode>3): Update accordingly.
18244         * config/aarch64/aarch64-simd.md (xorsign<mode>3): Likewise.
18245         (copysign<mode>3): Likewise.
18246         (aarch64_simd_bsl<mode>_internal): Likewise.
18247         (aarch64_simd_bsl<mode>): Likewise.
18248         (vec_cmp<mode><mode>): Likewise.
18249         (vcond<mode><mode>): Likewise.
18250         (vcond<v_cmp_mixed><mode>): Likewise.
18251         (vcondu<mode><v_cmp_mixed>): Likewise.
18252         (aarch64_cm<optab><mode>): Likewise.
18253         (aarch64_cmtst<mode>): Likewise.
18254         (aarch64_fac<optab><mode>): Likewise.
18255         (vec_perm_const<mode>): Likewise.
18256         (vcond_mask_<mode><v_cmp_result>): Rename to...
18257         (vcond_mask_<mode><v_int_equiv>): ...this.
18258         (vec_cmp<mode><v_cmp_result>): Rename to...
18259         (vec_cmp<mode><v_int_equiv>): ...this.
18261 2017-08-31  Richard Sandiford  <richard.sandiford@linaro.org>
18262             Alan Hayward  <alan.hayward@arm.com>
18263             David Sherwood  <david.sherwood@arm.com>
18265         * config/aarch64/aarch64-modes.def: Remove 32-, 48- and 64-byte
18266         vector modes.
18267         * config/aarch64/iterators.md (VRL2, VRL3, VRL4): Delete.
18268         * config/aarch64/aarch64.md (UNSPEC_LD2_DREG, UNSPEC_LD3_DREG)
18269         (UNSPEC_LD4_DREG): New unspecs.
18270         * config/aarch64/aarch64-simd.md (aarch64_ld2<mode>_dreg_le)
18271         (aarch64_ld2<mode>_dreg_be): Replace with...
18272         (aarch64_ld2<mode>_dreg): ...this pattern and use the new DREG
18273         unspec.
18274         (aarch64_ld3<mode>_dreg_le)
18275         (aarch64_ld3<mode>_dreg_be): Replace with...
18276         (aarch64_ld3<mode>_dreg): ...this pattern and use the new DREG
18277         unspec.
18278         (aarch64_ld4<mode>_dreg_le)
18279         (aarch64_ld4<mode>_dreg_be): Replace with...
18280         (aarch64_ld4<mode>_dreg): ...this pattern and use the new DREG
18281         unspec.
18283 2017-08-30  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
18285         PR tree-optimization/81987
18286         * gimple-ssa-strength-reduction.c (insert_initializers): Don't
18287         insert an initializer in a location not dominated by the stride
18288         definition.
18290 2017-08-30  Eric Botcazou  <ebotcazou@adacore.com>
18292         * tree-eh.c (lower_try_finally_switch): Set the location of the finally
18293         on the entire header of the finally block in the fallthru case.
18295 2017-08-30  Eric Botcazou  <ebotcazou@adacore.com>
18297         * varasm.c (decode_addr_const): Deal with INDIRECT_REF <INTEGER_CST>.
18299 2017-08-30  Pat Haugen  <pthaugen@us.ibm.com>
18301         * config/rs6000/rs6000.c (rs6000_emit_prologue_move_from_cr): Rename from
18302         rs6000_emit_move_from_cr and call renamed function.
18303         (rs6000_emit_prologue): Call renamed functions.
18304         * config/rs6000/rs6000.md (prologue_movesi_from_cr): Rename from
18305         movesi_from_cr, remove volatile CRs.
18307 2017-08-30  Jon Beniston  <jon@beniston.com>
18308             Richard Biener  <rguenther@suse.de>
18310         * tree-vect-patterns.c (vect_pattern_recog_1): Use VECTOR_TYPE_P instead
18311         of VECTOR_MODE_P check.
18312         * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Allow single
18313         element vector types.
18315 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18317         * df.h (df_read_modify_subreg_p): Remove in favor of...
18318         * rtl.h (read_modify_subreg_p): ...this new function.  Take a
18319         const_rtx instead of an rtx.
18320         * cprop.c (local_cprop_find_used_regs): Update accordingly.
18321         * df-problems.c (df_word_lr_mark_ref): Likewise.
18322         * ira-lives.c (mark_pseudo_reg_live): Likewise.
18323         (mark_pseudo_reg_dead): Likewise.
18324         (mark_ref_dead): Likewise.
18325         * reginfo.c (init_subregs_of_mode): Likewise.
18326         * sched-deps.c (sched_analyze_1): Likewise.
18327         * df-scan.c (df_def_record_1): Likewise.
18328         (df_uses_record): Likewise.
18329         (df_read_modify_subreg_p): Remove in favor of...
18330         * rtlanal.c (read_modify_subreg_p): ...this new function.  Take a
18331         const_rtx instead of an rtx.
18333 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18334             Alan Hayward  <alan.hayward@arm.com>
18335             David Sherwood  <david.sherwood@arm.com>
18337         * rtl.h (partial_subreg_p): New function.
18338         * caller-save.c (save_call_clobbered_regs): Use it.
18339         * calls.c (expand_call): Likewise.
18340         * combine.c (combinable_i3pat): Likewise.
18341         (simplify_set): Likewise.
18342         (make_extraction): Likewise.
18343         (make_compound_operation_int): Likewise.
18344         (gen_lowpart_or_truncate): Likewise.
18345         (force_to_mode): Likewise.
18346         (make_field_assignment): Likewise.
18347         (reg_truncated_to_mode): Likewise.
18348         (record_truncated_value): Likewise.
18349         (move_deaths): Likewise.
18350         * cse.c (record_jump_cond): Likewise.
18351         (cse_insn): Likewise.
18352         * cselib.c (cselib_lookup_1): Likewise.
18353         * expmed.c (extract_bit_field_using_extv): Likewise.
18354         * function.c (assign_parm_setup_reg): Likewise.
18355         * ifcvt.c (noce_convert_multiple_sets): Likewise.
18356         * ira-build.c (create_insn_allocnos): Likewise.
18357         * lra-coalesce.c (merge_pseudos): Likewise.
18358         * lra-constraints.c (match_reload): Likewise.
18359         (simplify_operand_subreg): Likewise.
18360         (curr_insn_transform): Likewise.
18361         * lra-lives.c (process_bb_lives): Likewise.
18362         * lra.c (new_insn_reg): Likewise.
18363         (lra_substitute_pseudo): Likewise.
18364         * regcprop.c (mode_change_ok): Likewise.
18365         (maybe_mode_change): Likewise.
18366         (copyprop_hardreg_forward_1): Likewise.
18367         * reload.c (push_reload): Likewise.
18368         (find_reloads): Likewise.
18369         (find_reloads_subreg_address): Likewise.
18370         * reload1.c (alter_reg): Likewise.
18371         (eliminate_regs_1): Likewise.
18372         * simplify-rtx.c (simplify_unary_operation_1): Likewise.
18374 2017-08-30  David Edelsohn  <dje.gcc@gmail.com>
18376         * config/rs6000/rs6000.c (rs6000_expand_binop_builtin): Revert
18377         back to if statements, including unpack.
18379 2017-08-30  Martin Liska  <mliska@suse.cz>
18381         PR inline-asm/82001
18382         * ipa-icf-gimple.c (func_checker::compare_tree_list_operand):
18383         Rename to ...
18384         (func_checker::compare_asm_inputs_outputs): ... this function.
18385         (func_checker::compare_gimple_asm): Use the function to compare
18386         also ASM constrains.
18387         * ipa-icf-gimple.h: Rename the function.
18389 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18390             Alan Hayward  <alan.hayward@arm.com>
18391             David Sherwood  <david.sherwood@arm.com>
18393         * coretypes.h (complex_mode): New type.
18394         * gdbhooks.py (build_pretty_printer): Handle it.
18395         * machmode.h (complex_mode): New class.
18396         (complex_mode::includes_p): New function.
18397         (is_complex_int_mode): Likewise.
18398         (is_complex_float_mode): Likewise.
18399         * genmodes.c (get_mode_class): Handle complex mode classes.
18400         * function.c (expand_function_end): Use is_complex_int_mode.
18402 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18403             Alan Hayward  <alan.hayward@arm.com>
18404             David Sherwood  <david.sherwood@arm.com>
18406         * coretypes.h (scalar_mode_pod): New typedef.
18407         * gdbhooks.py (build_pretty_printer): Handle it.
18408         * machmode.h (gt_ggc_mx, gt_pch_nx): New functions.
18409         * fixed-value.h (fixed_value::mode): Change type to scalar_mode_pod.
18410         * fold-const.c (fold_convert_const_int_from_fixed): Use scalar_mode.
18411         * tree-streamer-in.c (unpack_ts_fixed_cst_value_fields): Use
18412         as_a <scalar_mode>.
18414 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18415             Alan Hayward  <alan.hayward@arm.com>
18416             David Sherwood  <david.sherwood@arm.com>
18418         * machmode.h (mode_for_vector): Take a scalar_mode instead
18419         of a machine_mode.
18420         * stor-layout.c (mode_for_vector): Likewise.
18421         * explow.c (promote_mode): Use as_a <scalar_mode>.
18422         * sdbout.c (sdbout_parms): Use is_a <scalar_mode>.
18424 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18425             Alan Hayward  <alan.hayward@arm.com>
18426             David Sherwood  <david.sherwood@arm.com>
18428         * target.def (preferred_simd_mode): Take a scalar_mode
18429         instead of a machine_mode.
18430         * targhooks.h (default_preferred_simd_mode): Likewise.
18431         * targhooks.c (default_preferred_simd_mode): Likewise.
18432         * config/arc/arc.c (arc_preferred_simd_mode): Likewise.
18433         * config/arm/arm.c (arm_preferred_simd_mode): Likewise.
18434         * config/c6x/c6x.c (c6x_preferred_simd_mode): Likewise.
18435         * config/epiphany/epiphany.c (epiphany_preferred_simd_mode): Likewise.
18436         * config/i386/i386.c (ix86_preferred_simd_mode): Likewise.
18437         * config/mips/mips.c (mips_preferred_simd_mode): Likewise.
18438         * config/nvptx/nvptx.c (nvptx_preferred_simd_mode): Likewise.
18439         * config/powerpcspe/powerpcspe.c (rs6000_preferred_simd_mode):
18440         Likewise.
18441         * config/rs6000/rs6000.c (rs6000_preferred_simd_mode): Likewise.
18442         * config/s390/s390.c (s390_preferred_simd_mode): Likewise.
18443         * config/sparc/sparc.c (sparc_preferred_simd_mode): Likewise.
18444         * config/aarch64/aarch64.c (aarch64_preferred_simd_mode): Likewise.
18445         (aarch64_simd_scalar_immediate_valid_for_move): Update accordingly.
18446         * doc/tm.texi: Regenerate.
18447         * optabs-query.c (can_vec_mask_load_store_p): Return false for
18448         non-scalar modes.
18450 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18451             Alan Hayward  <alan.hayward@arm.com>
18452             David Sherwood  <david.sherwood@arm.com>
18454         * target.def (scalar_mode_supported_p): Take a scalar_mode
18455         instead of a machine_mode.
18456         * targhooks.h (default_scalar_mode_supported_p): Likewise.
18457         * targhooks.c (default_scalar_mode_supported_p): Likewise.
18458         * config/aarch64/aarch64.c (aarch64_scalar_mode_supported_p): Likewise.
18459         * config/alpha/alpha.c (alpha_scalar_mode_supported_p): Likewise.
18460         * config/arm/arm.c (arm_scalar_mode_supported_p): Likewise.
18461         * config/avr/avr.c (avr_scalar_mode_supported_p): Likewise.
18462         * config/c6x/c6x.c (c6x_scalar_mode_supported_p): Likewise.
18463         * config/i386/i386.c (ix86_scalar_mode_supported_p): Likewise.
18464         * config/ia64/ia64.c (ia64_scalar_mode_supported_p): Likewise.
18465         * config/mips/mips.c (mips_scalar_mode_supported_p): Likewise.
18466         * config/msp430/msp430.c (msp430_scalar_mode_supported_p): Likewise.
18467         * config/pa/pa.c (pa_scalar_mode_supported_p): Likewise.
18468         * config/pdp11/pdp11.c (pdp11_scalar_mode_supported_p): Likewise.
18469         * config/powerpcspe/powerpcspe.c (rs6000_scalar_mode_supported_p):
18470         Likewise.
18471         * config/rs6000/rs6000.c (rs6000_scalar_mode_supported_p): Likewise.
18472         * config/s390/s390.c (s390_scalar_mode_supported_p): Likewise.
18473         * config/spu/spu.c (spu_scalar_mode_supported_p): Likewise.
18474         * config/tilegx/tilegx.c (tilegx_scalar_mode_supported_p): Likewise.
18475         * config/tilepro/tilepro.c (tilepro_scalar_mode_supported_p):
18476         Likewise.
18477         * doc/tm.texi: Regenerate.
18479 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18480             Alan Hayward  <alan.hayward@arm.com>
18481             David Sherwood  <david.sherwood@arm.com>
18483         * coretypes.h (opt_scalar_mode): New typedef.
18484         * gdbhooks.py (build_pretty_printers): Handle it.
18485         * machmode.h (mode_iterator::get_2xwider): Add overload for
18486         opt_mode<T>.
18487         * emit-rtl.c (init_emit_once): Use opt_scalar_mode when iterating
18488         over scalar modes.
18489         * expr.c (convert_mode_scalar): Likewise.
18490         * omp-low.c (omp_clause_aligned_alignment): Likewise.
18491         * optabs.c (expand_float): Likewise.
18492         (expand_fix): Likewise.
18493         * tree-vect-stmts.c (vectorizable_conversion): Likewise.
18495 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18496             Alan Hayward  <alan.hayward@arm.com>
18497             David Sherwood  <david.sherwood@arm.com>
18499         * optabs.c (expand_float): Explicitly check for scalars before
18500         using a branching expansion.
18501         (expand_fix): Likewise.
18503 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18504             Alan Hayward  <alan.hayward@arm.com>
18505             David Sherwood  <david.sherwood@arm.com>
18507         * expr.c (convert_mode): Split scalar handling out into...
18508         (convert_mode_scalar): ...this new function.  Treat the modes
18509         as scalar_modes.
18511 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18512             Alan Hayward  <alan.hayward@arm.com>
18513             David Sherwood  <david.sherwood@arm.com>
18515         * omp-expand.c (expand_omp_atomic): Use is_int_mode, is_float_mode
18516         and scalar_mode.
18517         * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Likewise.
18519 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18520             Alan Hayward  <alan.hayward@arm.com>
18521             David Sherwood  <david.sherwood@arm.com>
18523         * fixed-value.h (fixed_from_double_int): Take a scalar_mode
18524         rather than a machine_mode.
18525         (fixed_from_string): Likewise.
18526         (fixed_convert): Likewise.
18527         (fixed_convert_from_int): Likewise.
18528         (fixed_convert_from_real): Likewise.
18529         (real_convert_from_fixed): Likewise.
18530         * fixed-value.c (fixed_from_double_int): Likewise.
18531         (fixed_from_string): Likewise.
18532         (fixed_convert): Likewise.
18533         (fixed_convert_from_int): Likewise.
18534         (fixed_convert_from_real): Likewise.
18535         (real_convert_from_fixed): Likewise.
18536         * config/avr/avr.c (avr_out_round): Use as_a <scalar_mode>.
18538 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18539             Alan Hayward  <alan.hayward@arm.com>
18540             David Sherwood  <david.sherwood@arm.com>
18542         * emit-rtl.c (immed_double_const): Use is_a <scalar_mode> instead
18543         of separate mode class checks.  Do not allow vector modes here.
18544         (immed_wide_int_const): Use as_a <scalar_mode>.
18545         * explow.c (trunc_int_for_mode): Likewise.
18546         * rtl.h (wi::int_traits<rtx_mode_t>::get_precision): Likewise.
18547         (wi::shwi): Likewise.
18548         (wi::min_value): Likewise.
18549         (wi::max_value): Likewise.
18550         * dwarf2out.c (loc_descriptor): Likewise.
18551         * simplify-rtx.c (simplify_immed_subreg): Fix rtx_mode_t argument
18552         for CONST_WIDE_INT.
18554 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18555             Alan Hayward  <alan.hayward@arm.com>
18556             David Sherwood  <david.sherwood@arm.com>
18558         * tree.h (SCALAR_TYPE_MODE): New macro.
18559         * expr.c (expand_expr_addr_expr_1): Use it.
18560         (expand_expr_real_2): Likewise.
18561         * fold-const.c (fold_convert_const_fixed_from_fixed): Likeise.
18562         (fold_convert_const_fixed_from_int): Likewise.
18563         (fold_convert_const_fixed_from_real): Likewise.
18564         (native_encode_fixed): Likewise
18565         (native_encode_complex): Likewise
18566         (native_encode_vector): Likewise.
18567         (native_interpret_fixed): Likewise.
18568         (native_interpret_real): Likewise.
18569         (native_interpret_complex): Likewise.
18570         (native_interpret_vector): Likewise.
18571         * omp-simd-clone.c (simd_clone_adjust_return_type): Likewise.
18572         (simd_clone_adjust_argument_types): Likewise.
18573         (simd_clone_init_simd_arrays): Likewise.
18574         (simd_clone_adjust): Likewise.
18575         * stor-layout.c (layout_type): Likewise.
18576         * tree.c (build_minus_one_cst): Likewise.
18577         * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
18578         * tree-inline.c (estimate_move_cost): Likewise.
18579         * tree-ssa-math-opts.c (convert_plusminus_to_widen): Likewise.
18580         * tree-vect-loop.c (vect_create_epilog_for_reduction): Likewise.
18581         (vectorizable_reduction): Likewise.
18582         * tree-vect-patterns.c (vect_recog_widen_mult_pattern): Likewise.
18583         (vect_recog_mixed_size_cond_pattern): Likewise.
18584         (check_bool_pattern): Likewise.
18585         (adjust_bool_pattern): Likewise.
18586         (search_type_for_mask_1): Likewise.
18587         * tree-vect-slp.c (vect_schedule_slp_instance): Likewise.
18588         * tree-vect-stmts.c (vectorizable_conversion): Likewise.
18589         (vectorizable_load): Likewise.
18590         (vectorizable_store): Likewise.
18591         * ubsan.c (ubsan_encode_value): Likewise.
18592         * varasm.c (output_constant): Likewise.
18594 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18595             Alan Hayward  <alan.hayward@arm.com>
18596             David Sherwood  <david.sherwood@arm.com>
18598         * coretypes.h (scalar_mode): New class.
18599         * machmode.h (scalar_mode): Likewise.
18600         (scalar_mode::includes_p): New function.
18601         (mode_to_inner): Return a scalar_mode rather than a machine_mode.
18602         * gdbhooks.py (build_pretty_printers): Handle scalar_mode.
18603         * genmodes.c (get_mode_class): Handle remaining scalar modes.
18604         * cfgexpand.c (expand_debug_expr): Use scalar_mode.
18605         * expmed.c (store_bit_field_1): Likewise.
18606         (extract_bit_field_1): Likewise.
18607         * expr.c (write_complex_part): Likewise.
18608         (read_complex_part): Likewise.
18609         (emit_move_complex_push): Likewise.
18610         (expand_expr_real_2): Likewise.
18611         * function.c (assign_parm_setup_reg): Likewise.
18612         (assign_parms_unsplit_complex): Likewise.
18613         * optabs.c (expand_binop): Likewise.
18614         * rtlanal.c (subreg_get_info): Likewise.
18615         * simplify-rtx.c (simplify_immed_subreg): Likewise.
18616         * varasm.c (output_constant_pool_2): Likewise.
18618 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18619             Alan Hayward  <alan.hayward@arm.com>
18620             David Sherwood  <david.sherwood@arm.com>
18622         * expmed.c (extract_high_half): Use scalar_int_mode and remove
18623         assertion.
18624         (expmed_mult_highpart_optab): Likewise.
18625         (expmed_mult_highpart): Likewise.
18627 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18628             Alan Hayward  <alan.hayward@arm.com>
18629             David Sherwood  <david.sherwood@arm.com>
18631         * builtins.h (builtin_strncpy_read_str): Take a scalar_int_mode
18632         instead of a machine_mode.
18633         (builtin_memset_read_str): Likewise.
18634         * builtins.c (c_readstr): Likewise.
18635         (builtin_memcpy_read_str): Likewise.
18636         (builtin_strncpy_read_str): Likewise.
18637         (builtin_memset_read_str): Likewise.
18638         (builtin_memset_gen_str): Likewise.
18639         (expand_builtin_signbit): Use scalar_int_mode for local variables.
18640         * cfgexpand.c (convert_debug_memory_address): Take a scalar_int_mode
18641         instead of a machine_mode.
18642         * combine.c (simplify_if_then_else): Use scalar_int_mode for local
18643         variables.
18644         (make_extraction): Likewise.
18645         (try_widen_shift_mode): Take and return scalar_int_modes instead
18646         of machine_modes.
18647         * config/aarch64/aarch64.c (aarch64_libgcc_cmp_return_mode): Return
18648         a scalar_int_mode instead of a machine_mode.
18649         * config/avr/avr.c (avr_addr_space_address_mode): Likewise.
18650         (avr_addr_space_pointer_mode): Likewise.
18651         * config/cr16/cr16.c (cr16_unwind_word_mode): Likewise.
18652         * config/msp430/msp430.c (msp430_addr_space_pointer_mode): Likewise.
18653         (msp430_unwind_word_mode): Likewise.
18654         * config/spu/spu.c (spu_unwind_word_mode): Likewise.
18655         (spu_addr_space_pointer_mode): Likewise.
18656         (spu_addr_space_address_mode): Likewise.
18657         (spu_libgcc_cmp_return_mode): Likewise.
18658         (spu_libgcc_shift_count_mode): Likewise.
18659         * config/rl78/rl78.c (rl78_addr_space_address_mode): Likewise.
18660         (rl78_addr_space_pointer_mode): Likewise.
18661         (fl78_unwind_word_mode): Likewise.
18662         (rl78_valid_pointer_mode): Take a scalar_int_mode instead of a
18663         machine_mode.
18664         * config/alpha/alpha.c (vms_valid_pointer_mode): Likewise.
18665         * config/ia64/ia64.c (ia64_vms_valid_pointer_mode): Likewise.
18666         * config/mips/mips.c (mips_mode_rep_extended): Likewise.
18667         (mips_valid_pointer_mode): Likewise.
18668         * config/tilegx/tilegx.c (tilegx_mode_rep_extended): Likewise.
18669         * config/ft32/ft32.c (ft32_valid_pointer_mode): Likewise.
18670         (ft32_addr_space_pointer_mode): Return a scalar_int_mode instead
18671         of a machine_mode.
18672         (ft32_addr_space_address_mode): Likewise.
18673         * config/m32c/m32c.c (m32c_valid_pointer_mode): Take a
18674         scalar_int_mode instead of a machine_mode.
18675         (m32c_addr_space_pointer_mode): Return a scalar_int_mode instead
18676         of a machine_mode.
18677         (m32c_addr_space_address_mode): Likewise.
18678         * config/powerpcspe/powerpcspe.c (rs6000_abi_word_mode): Likewise.
18679         (rs6000_eh_return_filter_mode): Likewise.
18680         * config/rs6000/rs6000.c (rs6000_abi_word_mode): Likewise.
18681         (rs6000_eh_return_filter_mode): Likewise.
18682         * config/s390/s390.c (s390_libgcc_cmp_return_mode): Likewise.
18683         (s390_libgcc_shift_count_mode): Likewise.
18684         (s390_unwind_word_mode): Likewise.
18685         (s390_valid_pointer_mode): Take a scalar_int_mode rather than a
18686         machine_mode.
18687         * target.def (mode_rep_extended): Likewise.
18688         (valid_pointer_mode): Likewise.
18689         (addr_space.valid_pointer_mode): Likewise.
18690         (eh_return_filter_mode): Return a scalar_int_mode rather than
18691         a machine_mode.
18692         (libgcc_cmp_return_mode): Likewise.
18693         (libgcc_shift_count_mode): Likewise.
18694         (unwind_word_mode): Likewise.
18695         (addr_space.pointer_mode): Likewise.
18696         (addr_space.address_mode): Likewise.
18697         * doc/tm.texi: Regenerate.
18698         * dojump.c (prefer_and_bit_test): Take a scalar_int_mode rather than
18699         a machine_mode.
18700         (do_jump): Use scalar_int_mode for local variables.
18701         * dwarf2cfi.c (init_return_column_size): Take a scalar_int_mode
18702         rather than a machine_mode.
18703         * dwarf2out.c (convert_descriptor_to_mode): Likewise.
18704         (scompare_loc_descriptor_wide): Likewise.
18705         (scompare_loc_descriptor_narrow): Likewise.
18706         * emit-rtl.c (adjust_address_1): Use scalar_int_mode for local
18707         variables.
18708         * except.c (sjlj_emit_dispatch_table): Likewise.
18709         (expand_builtin_eh_copy_values): Likewise.
18710         * explow.c (convert_memory_address_addr_space_1): Likewise.
18711         Take a scalar_int_mode rather than a machine_mode.
18712         (convert_memory_address_addr_space): Take a scalar_int_mode rather
18713         than a machine_mode.
18714         (memory_address_addr_space): Use scalar_int_mode for local variables.
18715         * expmed.h (expand_mult_highpart_adjust): Take a scalar_int_mode
18716         rather than a machine_mode.
18717         * expmed.c (mask_rtx): Likewise.
18718         (init_expmed_one_conv): Likewise.
18719         (expand_mult_highpart_adjust): Likewise.
18720         (extract_high_half): Likewise.
18721         (expmed_mult_highpart_optab): Likewise.
18722         (expmed_mult_highpart): Likewise.
18723         (expand_smod_pow2): Likewise.
18724         (expand_sdiv_pow2): Likewise.
18725         (emit_store_flag_int): Likewise.
18726         (adjust_bit_field_mem_for_reg): Use scalar_int_mode for local
18727         variables.
18728         (extract_low_bits): Likewise.
18729         * expr.h (by_pieces_constfn): Take a scalar_int_mode rather than
18730         a machine_mode.
18731         * expr.c (pieces_addr::adjust):  Likewise.
18732         (can_store_by_pieces): Likewise.
18733         (store_by_pieces): Likewise.
18734         (clear_by_pieces_1): Likewise.
18735         (expand_expr_addr_expr_1): Likewise.
18736         (expand_expr_addr_expr): Use scalar_int_mode for local variables.
18737         (expand_expr_real_1): Likewise.
18738         (try_casesi): Likewise.
18739         * final.c (shorten_branches): Likewise.
18740         * fold-const.c (fold_convert_const_int_from_fixed): Change the
18741         type of "mode" to machine_mode.
18742         * internal-fn.c (expand_arith_overflow_result_store): Take a
18743         scalar_int_mode rather than a machine_mode.
18744         (expand_mul_overflow): Use scalar_int_mode for local variables.
18745         * loop-doloop.c (doloop_modify): Likewise.
18746         (doloop_optimize): Likewise.
18747         * optabs.c (expand_subword_shift): Take a scalar_int_mode rather
18748         than a machine_mode.
18749         (expand_doubleword_shift_condmove): Likewise.
18750         (expand_doubleword_shift): Likewise.
18751         (expand_doubleword_clz): Likewise.
18752         (expand_doubleword_popcount): Likewise.
18753         (expand_doubleword_parity): Likewise.
18754         (expand_absneg_bit): Use scalar_int_mode for local variables.
18755         (prepare_float_lib_cmp): Likewise.
18756         * rtl.h (convert_memory_address_addr_space_1): Take a scalar_int_mode
18757         rather than a machine_mode.
18758         (convert_memory_address_addr_space): Likewise.
18759         (get_mode_bounds): Likewise.
18760         (get_address_mode): Return a scalar_int_mode rather than a
18761         machine_mode.
18762         * rtlanal.c (get_address_mode): Likewise.
18763         * stor-layout.c (get_mode_bounds): Take a scalar_int_mode rather
18764         than a machine_mode.
18765         * targhooks.c (default_mode_rep_extended): Likewise.
18766         (default_valid_pointer_mode): Likewise.
18767         (default_addr_space_valid_pointer_mode): Likewise.
18768         (default_eh_return_filter_mode): Return a scalar_int_mode rather
18769         than a machine_mode.
18770         (default_libgcc_cmp_return_mode): Likewise.
18771         (default_libgcc_shift_count_mode): Likewise.
18772         (default_unwind_word_mode): Likewise.
18773         (default_addr_space_pointer_mode): Likewise.
18774         (default_addr_space_address_mode): Likewise.
18775         * targhooks.h (default_eh_return_filter_mode): Likewise.
18776         (default_libgcc_cmp_return_mode): Likewise.
18777         (default_libgcc_shift_count_mode): Likewise.
18778         (default_unwind_word_mode): Likewise.
18779         (default_addr_space_pointer_mode): Likewise.
18780         (default_addr_space_address_mode): Likewise.
18781         (default_mode_rep_extended): Take a scalar_int_mode rather than
18782         a machine_mode.
18783         (default_valid_pointer_mode): Likewise.
18784         (default_addr_space_valid_pointer_mode): Likewise.
18785         * tree-ssa-address.c (addr_for_mem_ref): Use scalar_int_mode for
18786         local variables.
18787         * tree-ssa-loop-ivopts.c (get_shiftadd_cost): Take a scalar_int_mode
18788         rather than a machine_mode.
18789         * tree-switch-conversion.c (array_value_type): Use scalar_int_mode
18790         for local variables.
18791         * tree-vrp.c (simplify_float_conversion_using_ranges): Likewise.
18792         * var-tracking.c (use_narrower_mode): Take a scalar_int_mode rather
18793         than a machine_mode.
18795 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18796             Alan Hayward  <alan.hayward@arm.com>
18797             David Sherwood  <david.sherwood@arm.com>
18799         * dojump.c (do_jump_by_parts_greater_rtx): Change the type of
18800         the mode argument to scalar_int_mode.
18801         (do_jump_by_parts_zero_rtx): Likewise.
18802         (do_jump_by_parts_equality_rtx): Likewise.
18803         (do_jump_by_parts_greater): Take a mode argument.
18804         (do_jump_by_parts_equality): Likewise.
18805         (do_jump_1): Update calls accordingly.
18807 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18808             Alan Hayward  <alan.hayward@arm.com>
18809             David Sherwood  <david.sherwood@arm.com>
18811         * is-a.h (safe_dyn_cast): New function.
18812         * rtl.h (rtx_jump_table_data::get_data_mode): New function.
18813         (jump_table_for_label): Likewise.
18814         * final.c (final_addr_vec_align): Take an rtx_jump_table_data *
18815         instead of an rtx_insn *.
18816         (shorten_branches): Use dyn_cast instead of LABEL_P and
18817         JUMP_TABLE_DATA_P.  Use jump_table_for_label and
18818         rtx_jump_table_data::get_data_mode.
18819         (final_scan_insn): Likewise.
18821 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18822             Alan Hayward  <alan.hayward@arm.com>
18823             David Sherwood  <david.sherwood@arm.com>
18825         * combine.c (try_combine): Use is_a <scalar_int_mode> when
18826         trying to combine a full-register integer set with a subreg
18827         integer set.
18829 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18830             Alan Hayward  <alan.hayward@arm.com>
18831             David Sherwood  <david.sherwood@arm.com>
18833         * expr.c (expand_expr_addr_expr): Add a new_tmode local variable
18834         that is always either address_mode or pointer_mode.
18836 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18837             Alan Hayward  <alan.hayward@arm.com>
18838             David Sherwood  <david.sherwood@arm.com>
18840         * expr.c (expand_expr_real_2): Use word_mode instead of innermode
18841         when the two are known to be equal.
18843 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18844             Alan Hayward  <alan.hayward@arm.com>
18845             David Sherwood  <david.sherwood@arm.com>
18847         * simplify-rtx.c (simplify_const_unary_operation): Use
18848         is_a <scalar_int_mode> instead of checking for a nonzero
18849         precision.  Forcibly convert op_mode to a scalar_int_mode
18850         in that case.  More clearly differentiate the operand and
18851         result modes and use the former when deciding what the value
18852         of a count-bits operation should be.  Use is_int_mode instead
18853         of checking for a MODE_INT.  Remove redundant check for whether
18854         this mode has a zero precision.
18856 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18857             Alan Hayward  <alan.hayward@arm.com>
18858             David Sherwood  <david.sherwood@arm.com>
18860         * optabs.c (widen_leading): Change the type of the mode argument
18861         to scalar_int_mode.  Use opt_scalar_int_mode for the mode iterator.
18862         (widen_bswap): Likewise.
18863         (expand_parity): Likewise.
18864         (expand_ctz): Change the type of the mode argument to scalar_int_mode.
18865         (expand_ffs): Likewise.
18866         (epand_unop): Check for scalar integer modes before calling the
18867         above routines.
18869 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18870             Alan Hayward  <alan.hayward@arm.com>
18871             David Sherwood  <david.sherwood@arm.com>
18873         * expr.c (const_scalar_mask_from_tree): Add a mode argument.
18874         Expand commentary.
18875         (expand_expr_real_1): Update call accordingly.
18877 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18878             Alan Hayward  <alan.hayward@arm.com>
18879             David Sherwood  <david.sherwood@arm.com>
18881         * expmed.c (store_bit_field_using_insv): Add op0_mode and
18882         value_mode arguments.  Use scalar_int_mode internally.
18883         (store_bit_field_1): Rename the new integer mode from imode
18884         to op0_mode and use it instead of GET_MODE (op0).  Update calls
18885         to store_split_bit_field, store_bit_field_using_insv and
18886         store_fixed_bit_field.
18887         (store_fixed_bit_field): Add op0_mode and value_mode arguments.
18888         Use scalar_int_mode internally.  Use a bit count rather than a mode
18889         when calculating the largest bit size for get_best_mode.
18890         Update calls to store_split_bit_field and store_fixed_bit_field_1.
18891         (store_fixed_bit_field_1): Add mode and value_mode arguments.
18892         Remove assertion that OP0 has a scalar integer mode.
18893         (store_split_bit_field): Add op0_mode and value_mode arguments.
18894         Update calls to extract_fixed_bit_field.
18895         (extract_bit_field_using_extv): Add an op0_mode argument.
18896         Use scalar_int_mode internally.
18897         (extract_bit_field_1): Rename the new integer mode from imode to
18898         op0_mode and use it instead of GET_MODE (op0).  Update calls to
18899         extract_split_bit_field, extract_bit_field_using_extv and
18900         extract_fixed_bit_field.
18901         (extract_fixed_bit_field): Add an op0_mode argument.  Update calls
18902         to extract_split_bit_field and extract_fixed_bit_field_1.
18903         (extract_fixed_bit_field_1): Add a mode argument.  Remove assertion
18904         that OP0 has a scalar integer mode.  Use as_a <scalar_int_mode>
18905         on the target mode.
18906         (extract_split_bit_field): Add an op0_mode argument.  Update call
18907         to extract_fixed_bit_field.
18909 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18910             Alan Hayward  <alan.hayward@arm.com>
18911             David Sherwood  <david.sherwood@arm.com>
18913         * cse.c (cse_insn): Use opt_scalar_int_mode for the mode iterator.
18914         * explow.c (hard_function_value): Likewise.
18915         * expmed.c (extract_fixed_bit_field_1): Likewise.  Move the
18916         convert_to_mode call outside the loop.
18917         * expr.c (alignment_for_piecewise_move): Use opt_scalar_int_mode
18918         for the mode iterator.  Require the mode specified by max_pieces
18919         to exist.
18920         (emit_block_move_via_movmem): Use opt_scalar_int_mode for the
18921         mode iterator.
18922         (copy_blkmode_to_reg): Likewise.
18923         (set_storage_via_setmem): Likewise.
18924         * optabs.c (prepare_cmp_insn): Likewise.
18925         * rtlanal.c (init_num_sign_bit_copies_in_rep): Likewise.
18926         * stor-layout.c (finish_bitfield_representative): Likewise.
18928 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18929             Alan Hayward  <alan.hayward@arm.com>
18930             David Sherwood  <david.sherwood@arm.com>
18932         * rtl.h (subreg_unpromoted_mode, subreg_promoted_mode): New functions.
18933         * expr.c (convert_move): Use them.
18934         (convert_modes): Likewise.
18935         (store_expr_with_bounds): Likewise.
18937 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18938             Alan Hayward  <alan.hayward@arm.com>
18939             David Sherwood  <david.sherwood@arm.com>
18941         * rtl.h (rtl_hooks::reg_nonzero_bits): Add a scalar_int_mode
18942         parameter for the mode of "x".  Remove the "known_x", "known_mode"
18943         and "known_ret" arguments.  Change the type of the mode argument
18944         to scalar_int_mode.
18945         (rtl_hooks:reg_num_sign_bit_copies): Likewise.
18946         * combine.c (reg_nonzero_bits_for_combine): Update accordingly.
18947         (reg_num_sign_bit_copies_for_combine): Likewise.
18948         * rtlanal.c (nonzero_bits1): Likewise.
18949         (num_sign_bit_copies1): Likewise.
18950         * rtlhooks-def.h (reg_nonzero_bits_general): Likewise.
18951         (reg_num_sign_bit_copies_general): Likewise.
18952         * rtlhooks.c (reg_num_sign_bit_copies_general): Likewise.
18953         (reg_nonzero_bits_general): Likewise.
18955 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18956             Alan Hayward  <alan.hayward@arm.com>
18957             David Sherwood  <david.sherwood@arm.com>
18959         * rtlanal.c (num_sign_bit_copies): Handle VOIDmode here rather
18960         than in subroutines.  Return 1 for non-integer modes.
18961         (cached_num_sign_bit_copies): Change the type of the mode parameter
18962         to scalar_int_mode.
18963         (num_sign_bit_copies1): Likewise.  Remove early exit for other mode
18964         classes.  Handle CONST_INT_P first and then check whether X also
18965         has a scalar integer mode.  Check the same thing for inner registers
18966         of a SUBREG and for values that are being extended or truncated.
18968 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18969             Alan Hayward  <alan.hayward@arm.com>
18970             David Sherwood  <david.sherwood@arm.com>
18972         * rtlanal.c (nonzero_bits): Handle VOIDmode here rather than
18973         in subroutines.  Return the mode mask for non-integer modes.
18974         (cached_nonzero_bits): Change the type of the mode parameter
18975         to scalar_int_mode.
18976         (nonzero_bits1): Likewise.  Remove early exit for other mode
18977         classes.  Handle CONST_INT_P first and then check whether X
18978         also has a scalar integer mode.
18980 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18981             Alan Hayward  <alan.hayward@arm.com>
18982             David Sherwood  <david.sherwood@arm.com>
18984         * expr.c (widest_int_mode_for_size): Make the comment match the code.
18985         Return a scalar_int_mode and assert that the size is greater than
18986         one byte.
18987         (by_pieces_ninsns): Update accordingly and remove VOIDmode handling.
18988         (op_by_pieces_d::op_by_pieces_d): Likewise.
18989         (op_by_pieces_d::run): Likewise.
18990         (can_store_by_pieces): Likewise.
18992 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18993             Alan Hayward  <alan.hayward@arm.com>
18994             David Sherwood  <david.sherwood@arm.com>
18996         * combine.c (extract_left_shift): Add a mode argument and update
18997         recursive calls.
18998         (make_compound_operation_int): Change the type of the mode parameter
18999         to scalar_int_mode and update the call to extract_left_shift.
19001 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19002             Alan Hayward  <alan.hayward@arm.com>
19003             David Sherwood  <david.sherwood@arm.com>
19005         * combine.c (simplify_and_const_int): Change the type of the mode
19006         parameter to scalar_int_mode.
19007         (simplify_and_const_int_1): Likewise.  Update recursive call.
19009 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19010             Alan Hayward  <alan.hayward@arm.com>
19011             David Sherwood  <david.sherwood@arm.com>
19013         * combine.c (simplify_compare_const): Check that the mode is a
19014         scalar_int_mode (rather than VOIDmode) before testing its
19015         precision.
19016         (simplify_comparison): Move COMPARISON_P handling out of the
19017         loop and restrict the latter part of the loop to scalar_int_modes.
19018         Check is_a <scalar_int_mode> before calling HWI_COMPUTABLE_MODE_P
19019         and when considering SUBREG_REGs.  Use is_int_mode instead of
19020         checking GET_MODE_CLASS against MODE_INT.
19022 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19023             Alan Hayward  <alan.hayward@arm.com>
19024             David Sherwood  <david.sherwood@arm.com>
19026         * combine.c (try_widen_shift_mode): Move check for equal modes to...
19027         (simplify_shift_const_1): ...here.  Use scalar_int_mode for
19028         shift_unit_mode and for modes involved in scalar shifts.
19030 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19031             Alan Hayward  <alan.hayward@arm.com>
19032             David Sherwood  <david.sherwood@arm.com>
19034         * combine.c (force_int_to_mode): New function, split out from...
19035         (force_to_mode): ...here.  Keep xmode up-to-date and use it
19036         instead of GET_MODE (x).
19038 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19039             Alan Hayward  <alan.hayward@arm.com>
19040             David Sherwood  <david.sherwood@arm.com>
19042         * optabs-query.h (extraction_insn::struct_mode): Change type to
19043         opt_scalar_int_mode and update comment.
19044         (extraction_insn::field_mode): Change type to scalar_int_mode.
19045         (extraction_insn::pos_mode): Likewise.
19046         * combine.c (make_extraction): Update accordingly.
19047         * optabs-query.c (get_traditional_extraction_insn): Likewise.
19048         (get_optab_extraction_insn): Likewise.
19049         * recog.c (simplify_while_replacing): Likewise.
19050         * expmed.c (narrow_bit_field_mem): Change the type of the mode
19051         parameter to opt_scalar_int_mode.
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         * machmode.h (bit_field_mode_iterator::next_mode): Take a pointer
19058         to a scalar_int_mode instead of a machine_mode.
19059         (bit_field_mode_iterator::m_mode): Change type to opt_scalar_int_mode.
19060         (get_best_mode): Return a boolean and use a pointer argument to store
19061         the selected mode.  Replace the limit mode parameter with a bit limit.
19062         * expmed.c (adjust_bit_field_mem_for_reg): Use scalar_int_mode
19063         for the values returned by bit_field_mode_iterator::next_mode.
19064         (store_bit_field): Update call to get_best_mode.
19065         (store_fixed_bit_field): Likewise.
19066         (extract_fixed_bit_field): Likewise.
19067         * expr.c (optimize_bitfield_assignment_op): Likewise.
19068         * fold-const.c (optimize_bit_field_compare): Likewise.
19069         (fold_truth_andor_1): Likewise.
19070         * stor-layout.c (bit_field_mode_iterator::next_mode): As above.
19071         Update for new type of m_mode.
19072         (get_best_mode): As above.
19074 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19075             Alan Hayward  <alan.hayward@arm.com>
19076             David Sherwood  <david.sherwood@arm.com>
19078         * expmed.c (strict_volatile_bitfield_p): Change the type of fieldmode
19079         to scalar_int_mode.  Remove check for SCALAR_INT_MODE_P.
19080         (store_bit_field): Check is_a <scalar_int_mode> before calling
19081         strict_volatile_bitfield_p.
19082         (extract_bit_field): Likewise.
19084 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19085             Alan Hayward  <alan.hayward@arm.com>
19086             David Sherwood  <david.sherwood@arm.com>
19088         * target.def (cstore_mode): Return a scalar_int_mode.
19089         * doc/tm.texi: Regenerate.
19090         * config/sparc/sparc.c (sparc_cstore_mode): Return a scalar_int_mode.
19091         * targhooks.h (default_cstore_mode): Likewise.
19092         * targhooks.c (default_cstore_mode): Likewise, using a forced
19093         conversion.
19094         * expmed.c (emit_cstore): Expect the target of the cstore to be
19095         a scalar_int_mode.
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         * cfgloop.h (rtx_iv): Change type of extend_mode and mode to
19102         scalar_int_mode.
19103         (niter_desc): Likewise mode.
19104         (iv_analyze): Add a mode parameter.
19105         (biv_p): Likewise.
19106         (iv_analyze_expr): Pass the mode paraeter before the rtx it describes
19107         and change its type to scalar_int_mode.
19108         * loop-iv.c: Update commentary at head of file.
19109         (iv_constant): Pass the mode paraeter before the rtx it describes
19110         and change its type to scalar_int_mode.  Remove VOIDmode handling.
19111         (iv_subreg): Change the type of the mode parameter to scalar_int_mode.
19112         (iv_extend): Likewise.
19113         (shorten_into_mode): Likewise.
19114         (iv_add): Use scalar_int_mode.
19115         (iv_mult): Likewise.
19116         (iv_shift): Likewise.
19117         (canonicalize_iv_subregs): Likewise.
19118         (get_biv_step_1): Pass the outer_mode parameter before the rtx
19119         it describes and change its mode to scalar_int_mode.   Also change
19120         the type of the returned inner_mode to scalar_int_mode.
19121         (get_biv_step): Likewise, turning outer_mode from a pointer
19122         into a direct parameter.  Update call to get_biv_step_1.
19123         (iv_analyze_biv): Add an outer_mode parameter.  Update calls to
19124         iv_constant and get_biv_step.
19125         (iv_analyze_expr): Pass the mode parameter before the rtx it describes
19126         and change its type to scalar_int_mode.  Don't initialise iv->mode
19127         to VOIDmode and remove later checks for its still being VOIDmode.
19128         Update calls to iv_analyze_op and iv_analyze_expr.  Check
19129         is_a <scalar_int_mode> when changing the mode under consideration.
19130         (iv_analyze_def): Ignore registers that don't have a scalar_int_mode.
19131         Update call to iv_analyze_expr.
19132         (iv_analyze_op): Add a mode parameter.  Reject subregs whose
19133         inner register is not also a scalar_int_mode.  Update call to
19134         iv_analyze_biv.
19135         (iv_analyze): Add a mode parameter.  Update call to iv_analyze_op.
19136         (biv_p): Add a mode parameter.  Update call to iv_analyze_biv.
19137         (iv_number_of_iterations): Use is_a <scalar_int_mode> instead of
19138         separate mode class checks.  Update calls to iv_analyze.  Remove
19139         fix-up of VOIDmodes after iv_analyze_biv.
19140         * loop-unroll.c (analyze_iv_to_split_insn): Reject registers that
19141         don't have a scalar_int_mode.  Update call to biv_p.
19143 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19144             Alan Hayward  <alan.hayward@arm.com>
19145             David Sherwood  <david.sherwood@arm.com>
19147         * cfgexpand.c (convert_debug_memory_address): Use
19148         as_a <scalar_int_mode>.
19149         * combine.c (expand_compound_operation): Likewise.
19150         (make_extraction): Likewise.
19151         (change_zero_ext): Likewise.
19152         (simplify_comparison): Likewise.
19153         * cse.c (cse_insn): Likewise.
19154         * dwarf2out.c (minmax_loc_descriptor): Likewise.
19155         (mem_loc_descriptor): Likewise.
19156         (loc_descriptor): Likewise.
19157         * expmed.c (init_expmed_one_mode): Likewise.
19158         (synth_mult): Likewise.
19159         (emit_store_flag_1): Likewise.
19160         (expand_divmod): Likewise.  Use HWI_COMPUTABLE_MODE_P instead
19161         of a comparison with size.
19162         * expr.c (expand_assignment): Use as_a <scalar_int_mode>.
19163         (reduce_to_bit_field_precision): Likewise.
19164         * function.c (expand_function_end): Likewise.
19165         * internal-fn.c (expand_arith_overflow_result_store): Likewise.
19166         * loop-doloop.c (doloop_modify): Likewise.
19167         * optabs.c (expand_binop): Likewise.
19168         (expand_unop): Likewise.
19169         (expand_copysign_absneg): Likewise.
19170         (prepare_cmp_insn): Likewise.
19171         (maybe_legitimize_operand): Likewise.
19172         * recog.c (const_scalar_int_operand): Likewise.
19173         * rtlanal.c (get_address_mode): Likewise.
19174         * simplify-rtx.c (simplify_unary_operation_1): Likewise.
19175         (simplify_cond_clz_ctz): Likewise.
19176         * tree-nested.c (get_nl_goto_field): Likewise.
19177         * tree.c (build_vector_type_for_mode): Likewise.
19178         * var-tracking.c (use_narrower_mode): Likewise.
19180 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19181             Alan Hayward  <alan.hayward@arm.com>
19182             David Sherwood  <david.sherwood@arm.com>
19184         * tree.h (SCALAR_INT_TYPE_MODE): New macro.
19185         * builtins.c (expand_builtin_signbit): Use it.
19186         * cfgexpand.c (expand_debug_expr): Likewise.
19187         * dojump.c (do_jump): Likewise.
19188         (do_compare_and_jump): Likewise.
19189         * dwarf2cfi.c (expand_builtin_init_dwarf_reg_sizes): Likewise.
19190         * expmed.c (make_tree): Likewise.
19191         * expr.c (expand_expr_real_2): Likewise.
19192         (expand_expr_real_1): Likewise.
19193         (try_casesi): Likewise.
19194         * fold-const-call.c (fold_const_call_ss): Likewise.
19195         * fold-const.c (unextend): Likewise.
19196         (extract_muldiv_1): Likewise.
19197         (fold_single_bit_test): Likewise.
19198         (native_encode_int): Likewise.
19199         (native_encode_string): Likewise.
19200         (native_interpret_int): Likewise.
19201         * gimple-fold.c (gimple_fold_builtin_memset): Likewise.
19202         * internal-fn.c (expand_addsub_overflow): Likewise.
19203         (expand_neg_overflow): Likewise.
19204         (expand_mul_overflow): Likewise.
19205         (expand_arith_overflow): Likewise.
19206         * match.pd: Likewise.
19207         * stor-layout.c (layout_type): Likewise.
19208         * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
19209         * tree-ssa-math-opts.c (convert_mult_to_widen): Likewise.
19210         * tree-ssanames.c (get_range_info): Likewise.
19211         * tree-switch-conversion.c (array_value_type) Likewise.
19212         * tree-vect-patterns.c (vect_recog_rotate_pattern): Likewise.
19213         (vect_recog_divmod_pattern): Likewise.
19214         (vect_recog_mixed_size_cond_pattern): Likewise.
19215         * tree-vrp.c (extract_range_basic): Likewise.
19216         (simplify_float_conversion_using_ranges): Likewise.
19217         * tree.c (int_fits_type_p): Likewise.
19218         * ubsan.c (instrument_bool_enum_load): Likewise.
19219         * varasm.c (mergeable_string_section): Likewise.
19220         (narrowing_initializer_constant_valid_p): Likewise.
19221         (output_constant): Likewise.
19223 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19224             Alan Hayward  <alan.hayward@arm.com>
19225             David Sherwood  <david.sherwood@arm.com>
19227         * machmode.h (NARROWEST_INT_MODE): New macro.
19228         * expr.c (alignment_for_piecewise_move): Use it instead of
19229         GET_CLASS_NARROWEST_MODE (MODE_INT).
19230         (push_block): Likewise.
19231         * stor-layout.c (bit_field_mode_iterator::bit_field_mode_iterator):
19232         Likewise.
19233         * tree-vrp.c (simplify_float_conversion_using_ranges): Likewise.
19235 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19236             Alan Hayward  <alan.hayward@arm.com>
19237             David Sherwood  <david.sherwood@arm.com>
19239         * postreload.c (move2add_valid_value_p): Change the type of the
19240         mode parameter to scalar_int_mode.
19241         (move2add_use_add2_insn): Add a mode parameter and use it instead
19242         of GET_MODE (reg).
19243         (move2add_use_add3_insn): Likewise.
19244         (reload_cse_move2add): Update accordingly.
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         * expr.c (expand_expr_real_2): Use scalar_int_mode for the
19251         double-word mode.
19252         * lower-subreg.c (resolve_shift_zext): Use is_a <scalar_int_mode>.
19253         * optabs.c (expand_unop): Likewise.
19255 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19256             Alan Hayward  <alan.hayward@arm.com>
19257             David Sherwood  <david.sherwood@arm.com>
19259         * dwarf2out.c (typed_binop): Change mode parameter to scalar_int_mode.
19260         (clz_loc_descriptor): Likewise.  Remove SCALAR_INT_MODE_P check.
19261         (popcount_loc_descriptor): Likewise.
19262         (bswap_loc_descriptor): Likewise.
19263         (rotate_loc_descriptor): Likewise.
19264         (mem_loc_descriptor): Add is_a <scalar_int_mode> checks before
19265         calling the functions above.
19267 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19268             Alan Hayward  <alan.hayward@arm.com>
19269             David Sherwood  <david.sherwood@arm.com>
19271         * combine.c (sign_extend_short_imm): Add is_a <scalar_int_mode>
19272         checks.
19273         (try_combine): Likewise.
19274         (simplify_if_then_else): Likewise.
19275         * cse.c (cse_insn): Likewise.
19276         * dwarf2out.c (mem_loc_descriptor): Likewise.
19277         * emit-rtl.c (gen_lowpart_common): Likewise.
19278         * simplify-rtx.c (simplify_truncation): Likewise.
19279         (simplify_binary_operation_1): Likewise.
19280         (simplify_const_relational_operation): Likewise.
19281         (simplify_ternary_operation): Likewise.
19282         * tree-ssa-loop-ivopts.c (force_expr_to_var_cost): Likewise.
19284 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19285             Alan Hayward  <alan.hayward@arm.com>
19286             David Sherwood  <david.sherwood@arm.com>
19288         * cse.c (cse_insn): Add is_a <scalar_int_mode> checks.
19289         * reload.c (push_reload): Likewise.
19290         (find_reloads): Likewise.
19292 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19293             Alan Hayward  <alan.hayward@arm.com>
19294             David Sherwood  <david.sherwood@arm.com>
19296         * combine.c (find_split_point): Add is_a <scalar_int_mode> checks.
19297         (make_compound_operation_int): Likewise.
19298         (change_zero_ext): Likewise.
19299         * expr.c (convert_move): Likewise.
19300         (convert_modes): Likewise.
19301         * fwprop.c (forward_propagate_subreg): Likewise.
19302         * loop-iv.c (get_biv_step_1): Likewise.
19303         * optabs.c (widen_operand): Likewise.
19304         * postreload.c (move2add_valid_value_p): Likewise.
19305         * recog.c (simplify_while_replacing): Likewise.
19306         * simplify-rtx.c (simplify_unary_operation_1): Likewise.
19307         (simplify_binary_operation_1): Likewise.  Remove redundant
19308         mode equality check.
19310 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19311             Alan Hayward  <alan.hayward@arm.com>
19312             David Sherwood  <david.sherwood@arm.com>
19314         * combine.c (combine_simplify_rtx): Add checks for
19315         is_a <scalar_int_mode>.
19316         (simplify_if_then_else): Likewise.
19317         (make_field_assignment): Likewise.
19318         (simplify_comparison): Likewise.
19319         * ifcvt.c (noce_try_bitop): Likewise.
19320         * loop-invariant.c (canonicalize_address_mult): Likewise.
19321         * simplify-rtx.c (simplify_unary_operation_1): Likewise.
19323 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19324             Alan Hayward  <alan.hayward@arm.com>
19325             David Sherwood  <david.sherwood@arm.com>
19327         * gimple-fold.c (gimple_fold_builtin_memory_op): Use
19328         is_a <scalar_int_mode> instead of != BLKmode.
19330 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19331             Alan Hayward  <alan.hayward@arm.com>
19332             David Sherwood  <david.sherwood@arm.com>
19334         * cfgexpand.c (expand_debug_expr): Use is_a <scalar_int_mode>
19335         instead of != VOIDmode.
19336         * combine.c (if_then_else_cond): Likewise.
19337         (change_zero_ext): Likewise.
19338         * dwarf2out.c (mem_loc_descriptor): Likewise.
19339         (loc_descriptor): Likewise.
19340         * rtlanal.c (canonicalize_condition): Likewise.
19341         * simplify-rtx.c (simplify_relational_operation_1): Likewise.
19343 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19344             Alan Hayward  <alan.hayward@arm.com>
19345             David Sherwood  <david.sherwood@arm.com>
19347         * simplify-rtx.c (simplify_binary_operation_1): Use
19348         is_a <scalar_int_mode> instead of !VECTOR_MODE_P.
19350 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19351             Alan Hayward  <alan.hayward@arm.com>
19352             David Sherwood  <david.sherwood@arm.com>
19354         * wide-int.h (int_traits<unsigned char>) New class.
19355         (int_traits<unsigned short>) Likewise.
19356         * cfgexpand.c (expand_debug_expr): Use is_a <scalar_int_mode>.
19357         Use GET_MODE_UNIT_PRECISION and remove redundant test for
19358         SCALAR_INT_MODE_P.
19359         * combine.c (set_nonzero_bits_and_sign_copies): Use
19360         is_a <scalar_int_mode>.
19361         (find_split_point): Likewise.
19362         (combine_simplify_rtx): Likewise.
19363         (simplify_logical): Likewise.
19364         (expand_compound_operation): Likewise.
19365         (expand_field_assignment): Likewise.
19366         (make_compound_operation): Likewise.
19367         (extended_count): Likewise.
19368         (change_zero_ext): Likewise.
19369         (simplify_comparison): Likewise.
19370         * dwarf2out.c (scompare_loc_descriptor): Likewise.
19371         (ucompare_loc_descriptor): Likewise.
19372         (minmax_loc_descriptor): Likewise.
19373         (mem_loc_descriptor): Likewise.
19374         (loc_descriptor): Likewise.
19375         * expmed.c (init_expmed_one_mode): Likewise.
19376         * lra-constraints.c (lra_constraint_offset): Likewise.
19377         * optabs.c (prepare_libcall_arg): Likewise.
19378         * postreload.c (move2add_note_store): Likewise.
19379         * reload.c (operands_match_p): Likewise.
19380         * rtl.h (load_extend_op): Likewise.
19381         * rtlhooks.c (gen_lowpart_general): Likewise.
19382         * simplify-rtx.c (simplify_truncation): Likewise.
19383         (simplify_unary_operation_1): Likewise.
19384         (simplify_binary_operation_1): Likewise.
19385         (simplify_const_binary_operation): Likewise.
19386         (simplify_const_relational_operation): Likewise.
19387         (simplify_subreg): Likewise.
19388         * stor-layout.c (bitwise_mode_for_mode): Likewise.
19389         * var-tracking.c (adjust_mems): Likewise.
19390         (prepare_call_arguments): Likewise.
19392 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19393             Alan Hayward  <alan.hayward@arm.com>
19394             David Sherwood  <david.sherwood@arm.com>
19396         * machmode.h (is_int_mode): New fuction.
19397         * combine.c (find_split_point): Use it.
19398         (combine_simplify_rtx): Likewise.
19399         (simplify_if_then_else): Likewise.
19400         (simplify_set): Likewise.
19401         (simplify_shift_const_1): Likewise.
19402         (simplify_comparison): Likewise.
19403         * config/aarch64/aarch64.c (aarch64_rtx_costs): Likewise.
19404         * cse.c (notreg_cost): Likewise.
19405         (cse_insn): Likewise.
19406         * cselib.c (cselib_lookup_1): Likewise.
19407         * dojump.c (do_jump_1): Likewise.
19408         (do_compare_rtx_and_jump): Likewise.
19409         * dse.c (get_call_args): Likewise.
19410         * dwarf2out.c (rtl_for_decl_init): Likewise.
19411         (native_encode_initializer): Likewise.
19412         * expmed.c (emit_store_flag_1): Likewise.
19413         (emit_store_flag): Likewise.
19414         * expr.c (convert_modes): Likewise.
19415         (store_field): Likewise.
19416         (expand_expr_real_1): Likewise.
19417         * fold-const.c (fold_read_from_constant_string): Likewise.
19418         * gimple-ssa-sprintf.c (get_format_string): Likewise.
19419         * optabs-libfuncs.c (gen_int_libfunc): Likewise.
19420         * optabs.c (expand_binop): Likewise.
19421         (expand_unop): Likewise.
19422         (expand_abs_nojump): Likewise.
19423         (expand_one_cmpl_abs_nojump): Likewise.
19424         * simplify-rtx.c (mode_signbit_p): Likewise.
19425         (val_signbit_p): Likewise.
19426         (val_signbit_known_set_p): Likewise.
19427         (val_signbit_known_clear_p): Likewise.
19428         (simplify_relational_operation_1): Likewise.
19429         * tree.c (vector_type_mode): Likewise.
19431 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19432             Alan Hayward  <alan.hayward@arm.com>
19433             David Sherwood  <david.sherwood@arm.com>
19435         * machmode.h (smallest_mode_for_size): Fix formatting.
19436         (smallest_int_mode_for_size): New function.
19437         * cfgexpand.c (expand_debug_expr): Use smallest_int_mode_for_size
19438         instead of smallest_mode_for_size.
19439         * combine.c (make_extraction): Likewise.
19440         * config/arc/arc.c (arc_expand_movmem): Likewise.
19441         * config/arm/arm.c (arm_expand_divmod_libfunc): Likewise.
19442         * config/i386/i386.c (ix86_get_mask_mode): Likewise.
19443         * config/s390/s390.c (s390_expand_insv): Likewise.
19444         * config/sparc/sparc.c (assign_int_registers): Likewise.
19445         * config/spu/spu.c (spu_function_value): Likewise.
19446         (spu_function_arg): Likewise.
19447         * coverage.c (get_gcov_type): Likewise.
19448         (get_gcov_unsigned_t): Likewise.
19449         * dse.c (find_shift_sequence): Likewise.
19450         * expmed.c (store_bit_field_1): Likewise.
19451         * expr.c (convert_move): Likewise.
19452         (store_field): Likewise.
19453         * internal-fn.c (expand_arith_overflow): Likewise.
19454         * optabs-query.c (get_best_extraction_insn): Likewise.
19455         * optabs.c (expand_twoval_binop_libfunc): Likewise.
19456         * stor-layout.c (layout_type): Likewise.
19457         (initialize_sizetypes): Likewise.
19458         * targhooks.c (default_get_mask_mode): Likewise.
19459         * tree-ssa-loop-manip.c (canonicalize_loop_ivs): Likewise.
19461 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19462             Alan Hayward  <alan.hayward@arm.com>
19463             David Sherwood  <david.sherwood@arm.com>
19465         * machmode.h (opt_mode::else_blk): New function.
19466         (int_mode_for_mode): Declare.
19467         * stor-layout.c (int_mode_for_mode): Return an opt_scalar_int_mode.
19468         * builtins.c (expand_builtin_signbit): Adjust for new int_mode_for_mode
19469         return type.
19470         * cfgexpand.c (expand_debug_expr): Likewise.
19471         * combine.c (gen_lowpart_or_truncate): Likewise.
19472         (gen_lowpart_for_combine): Likewise.
19473         * config/aarch64/aarch64.c (aarch64_emit_approx_sqrt): Likewise.
19474         * config/avr/avr.c (avr_to_int_mode): Likewise.
19475         (avr_out_plus_1): Likewise.
19476         (avr_out_plus): Likewise.
19477         (avr_out_round): Likewise.
19478         * config/i386/i386.c (ix86_split_to_parts): Likewise.
19479         * config/s390/s390.c (s390_expand_vec_compare_cc): Likewise.
19480         (s390_expand_vcond): Likewise.
19481         * config/spu/spu.c (spu_split_immediate): Likewise.
19482         (spu_expand_mov): Likewise.
19483         * dse.c (get_stored_val): Likewise.
19484         * expmed.c (store_bit_field_1): Likewise.
19485         (convert_extracted_bit_field): Use int_mode_for_mode instead of
19486         int_mode_for_size.
19487         (extract_bit_field_1): Adjust for new int_mode_for_mode return type.
19488         (extract_low_bits): Likewise.
19489         * expr.c (emit_group_load_1): Likewise.  Separate out the BLKmode
19490         handling rather than repeating the check.
19491         (emit_group_store): Likewise.
19492         (emit_move_via_integer): Adjust for new int_mode_for_mode return type.
19493         * optabs.c (expand_absneg_bit): Likewise.
19494         (expand_copysign_absneg): Likewise.
19495         (expand_copysign_bit): Likewise.
19496         * tree-if-conv.c (ifcvt_can_use_mask_load_store): Likewise.
19497         * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
19498         * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
19499         * var-tracking.c (prepare_call_arguments):  Likewise.
19500         * config/powerpcspe/powerpcspe.c (rs6000_do_expand_vec_perm): Use
19501         int_mode_for_mode instead of mode_for_size.
19502         * config/rs6000/rs6000.c (rs6000_do_expand_vec_perm): Likewise.
19504 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19505             Alan Hayward  <alan.hayward@arm.com>
19506             David Sherwood  <david.sherwood@arm.com>
19508         * machmode.h (int_mode_for_size): New function.
19509         * builtins.c (set_builtin_user_assembler_name): Use int_mode_for_size
19510         instead of mode_for_size.
19511         * calls.c (save_fixed_argument_area): Likewise.  Make use of BLKmode
19512         explicit.
19513         * combine.c (expand_field_assignment): Use int_mode_for_size
19514         instead of mode_for_size.
19515         (make_extraction): Likewise.
19516         (simplify_shift_const_1): Likewise.
19517         (simplify_comparison): Likewise.
19518         * dojump.c (do_jump): Likewise.
19519         * dwarf2out.c (mem_loc_descriptor): Likewise.
19520         * emit-rtl.c (init_derived_machine_modes): Likewise.
19521         * expmed.c (flip_storage_order): Likewise.
19522         (convert_extracted_bit_field): Likewise.
19523         * expr.c (copy_blkmode_from_reg): Likewise.
19524         * graphite-isl-ast-to-gimple.c (max_mode_int_precision): Likewise.
19525         * internal-fn.c (expand_mul_overflow): Likewise.
19526         * lower-subreg.c (simple_move): Likewise.
19527         * optabs-libfuncs.c (init_optabs): Likewise.
19528         * simplify-rtx.c (simplify_unary_operation_1): Likewise.
19529         * tree.c (vector_type_mode): Likewise.
19530         * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
19531         * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Likewise.
19532         * tree-vect-generic.c (expand_vector_parallel): Likewise.
19533         * tree-vect-stmts.c (vectorizable_load): Likewise.
19534         (vectorizable_store): Likewise.
19536 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19537             Alan Hayward  <alan.hayward@arm.com>
19538             David Sherwood  <david.sherwood@arm.com>
19540         * coretypes.h (pod_mode): New type.
19541         (scalar_int_mode_pod): New typedef.
19542         * machmode.h (pod_mode): New class.
19543         (int_n_data_t::m): Change type to scalar_int_mode_pod.
19544         * genmodes.c (emit_mode_int_n): Update accordingly.
19545         * lower-subreg.h (target_lower_subreg): Change type to
19546         scalar_int_mode_pod.
19547         * gdbhooks.py (build_pretty_printer): Handle pod_mode and
19548         scalar_int_mode_pod.
19550 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19551             Alan Hayward  <alan.hayward@arm.com>
19552             David Sherwood  <david.sherwood@arm.com>
19554         * config/powerpcspe/powerpcspe.h (rs6000_pmode): Change type from
19555         machine_mode to scalar_int_mode.
19556         * config/powerpcspe/powerpcspe.c (rs6000_pmode): Likewise.
19557         (rs6000_option_override_internal): Remove cast to int.
19558         * config/rs6000/rs6000.h (rs6000_pmode): Change type from
19559         machine_mode to scalar_int_mode.
19560         * config/rs6000/rs6000.c (rs6000_pmode): Likewise.
19561         (rs6000_option_override_internal): Remove cast to int.
19562         * config/s390/s390.h (Pmode): Remove cast to machine_mode.
19563         * config/epiphany/epiphany.h (RTX_OK_FOR_OFFSET_P): Add cast
19564         to machine_mode.
19565         * config/s390/s390.c (s390_expand_builtin): Likewise.
19566         * coretypes.h (scalar_int_mode): New type.
19567         (opt_scalar_int_mode): New typedef.
19568         * machmode.h (scalar_int_mode): New class.
19569         (scalar_int_mode::includes_p): New function.
19570         (byte_mode): Change type to scalar_int_mode.
19571         (word_mode): Likewise.
19572         (ptr_mode): Likewise.
19573         * emit-rtl.c (byte_mode): Likewise.
19574         (word_mode): Likewise.
19575         (ptr_mode): Likewise.
19576         (init_derived_machine_modes): Update accordingly.
19577         * genmodes.c (get_mode_class): Return scalar_int_mode for MODE_INT
19578         and MODE_PARTIAL_INT.
19579         * gdbhooks.py (build_pretty_printer): Handle scalar_int_mode and
19580         opt_scalar_int_mode.
19582 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19583             Alan Hayward  <alan.hayward@arm.com>
19584             David Sherwood  <david.sherwood@arm.com>
19586         * target.def (libgcc_floating_mode_supported_p): Take a
19587         scalar_float_mode.
19588         * doc/tm.texi: Regenerate.
19589         * targhooks.h (default_libgcc_floating_mode_supported_p): Take a
19590         scalar_float_mode.
19591         * targhooks.c (default_libgcc_floating_mode_supported_p): Likewise.
19592         * config/aarch64/aarch64.c (aarch64_libgcc_floating_mode_supported_p):
19593         Likewise.
19595 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19596             Alan Hayward  <alan.hayward@arm.com>
19597             David Sherwood  <david.sherwood@arm.com>
19599         * target.def (default_floatn_mode): Return an opt_scalar_float_mode.
19600         * doc/tm.texi: Regenerate.
19601         * config/arm/arm.c (arm_floatn_mode): Return an opt_scalar_float_mode.
19602         * config/powerpcspe/powerpcspe.c (rs6000_floatn_mode): Likewise.
19603         * config/rs6000/rs6000.c (rs6000_floatn_mode): Likewise.
19604         * targhooks.h (default_floatn_mode): Likewise.
19605         * targhooks.c (default_floatn_mode): Likewise.
19606         * tree.c (build_common_tree_nodes): Update accordingly.
19608 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19609             Alan Hayward  <alan.hayward@arm.com>
19610             David Sherwood  <david.sherwood@arm.com>
19612         * machmode.h (mode_iterator::start): Provide overload for opt_modes.
19613         (mode_iterator::iterate_p): Likewise.
19614         (mode_iterator::get_wider): Likewise.
19615         * expr.c (init_expr_target): Use opt_scalar_float_mode.
19617 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19618             Alan Hayward  <alan.hayward@arm.com>
19619             David Sherwood  <david.sherwood@arm.com>
19621         * coretypes.h (opt_scalar_float_mode): New typedef.
19622         * machmode.h (float_mode_for_size): New function.
19623         * emit-rtl.c (double_mode): Delete.
19624         (init_emit_once): Use float_mode_for_size.
19625         * stor-layout.c (layout_type): Likewise.
19626         * gdbhooks.py (build_pretty_printer): Handle opt_scalar_float_mode.
19628 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19629             Alan Hayward  <alan.hayward@arm.com>
19630             David Sherwood  <david.sherwood@arm.com>
19632         * output.h (assemble_real): Take a scalar_float_mode.
19633         * config/arm/arm.c (arm_assemble_integer): Update accordingly.
19634         * config/arm/arm.md (consttable_4): Likewise.
19635         (consttable_8): Likewise.
19636         (consttable_16): Likewise.
19637         * config/mips/mips.md (consttable_float): Likewise.
19638         * config/s390/s390.c (s390_output_pool_entry): Likewise.
19639         * varasm.c (assemble_real): Take a scalar_float_mode.
19640         (output_constant_pool_2): Update accordingly.
19641         (output_constant): Likewise.
19643 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19644             Alan Hayward  <alan.hayward@arm.com>
19645             David Sherwood  <david.sherwood@arm.com>
19647         * tree.h (SCALAR_FLOAT_TYPE_MODE): New macro.
19648         * builtins.c (expand_builtin_signbit): Use it instead of TYPE_MODE.
19649         * fold-const.c (fold_convert_const_real_from_fixed): Likewise.
19650         (native_encode_real): Likewise.
19651         (native_interpret_real): Likewise.
19652         * hsa-brig.c (emit_immediate_scalar_to_buffer): Likewise.
19653         * tree-vrp.c (simplify_float_conversion_using_ranges): Likewise.
19655 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19656             Alan Hayward  <alan.hayward@arm.com>
19657             David Sherwood  <david.sherwood@arm.com>
19659         * optabs-libfuncs.c (gen_trunc_conv_libfunc): Use is_a
19660         <scalar_float_mode>.  Simplify.
19661         (gen_extend_conv_libfunc): Likewise.
19663 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19664             Alan Hayward  <alan.hayward@arm.com>
19665             David Sherwood  <david.sherwood@arm.com>
19667         * coretypes.h (scalar_float_mode): New type.
19668         * machmode.h (mode_traits::from_int): Use machine_mode if
19669         USE_ENUM_MODES is defined.
19670         (is_a): New function.
19671         (as_a): Likewise.
19672         (dyn_cast): Likewise.
19673         (scalar_float_mode): New class.
19674         (scalar_float_mode::includes_p): New function.
19675         (is_float_mode): Likewise.
19676         * gdbhooks.py (MachineModePrinter): New class.
19677         (build_pretty_printer): Use it for scalar_float_mode.
19678         * real.h (FLOAT_MODE_FORMAT): Use as_a <scalar_float_mode>.
19679         (format_helper::format_helper): Turn into a template.
19680         * genmodes.c (get_mode_class): New function.
19681         (emit_insn_modes_h): Give modes the class returned by get_mode_class,
19682         or machine_mode if none.
19683         * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Use
19684         as_a <scalar_float_mode>.
19685         * dwarf2out.c (mem_loc_descriptor): Likewise.
19686         (insert_float): Likewise.
19687         (add_const_value_attribute): Likewise.
19688         * simplify-rtx.c (simplify_immed_subreg): Likewise.
19689         * optabs.c (expand_absneg_bit): Take a scalar_float_mode.
19690         (expand_unop): Update accordingly.
19691         (expand_abs_nojump): Likewise.
19692         (expand_copysign_absneg): Take a scalar_float_mode.
19693         (expand_copysign_bit): Likewise.
19694         (expand_copysign): Update accordingly.
19696 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19697             Alan Hayward  <alan.hayward@arm.com>
19698             David Sherwood  <david.sherwood@arm.com>
19700         * coretypes.h (opt_mode): New class.
19701         * machmode.h (opt_mode): Likewise.
19702         (opt_mode::else_void): New function.
19703         (opt_mode::require): Likewise.
19704         (opt_mode::exists): Likewise.
19705         (GET_MODE_WIDER_MODE): Turn into a function and return an opt_mode.
19706         (GET_MODE_2XWIDER_MODE): Likewise.
19707         (mode_iterator::get_wider): Update accordingly.
19708         (mode_iterator::get_2xwider): Likewise.
19709         (mode_iterator::get_known_wider): Likewise, turning into a template.
19710         * combine.c (make_extraction): Update use of GET_MODE_WIDER_MODE,
19711         forcing a wider mode to exist.
19712         * config/cr16/cr16.h (LONG_REG_P): Likewise.
19713         * rtlanal.c (init_num_sign_bit_copies_in_rep): Likewise.
19714         * config/c6x/c6x.c (c6x_rtx_costs): Update use of
19715         GET_MODE_2XWIDER_MODE, forcing a wider mode to exist.
19716         * lower-subreg.c (init_lower_subreg): Likewise.
19717         * optabs-libfuncs.c (init_sync_libfuncs_1): Likewise, but not
19718         on the final iteration.
19719         * config/i386/i386.c (ix86_expand_set_or_movmem): Check whether
19720         a wider mode exists before asking for a move pattern.
19721         (get_mode_wider_vector): Update use of GET_MODE_WIDER_MODE,
19722         forcing a wider mode to exist.
19723         (expand_vselect_vconcat): Update use of GET_MODE_2XWIDER_MODE,
19724         returning false if no such mode exists.
19725         * config/ia64/ia64.c (expand_vselect_vconcat): Likewise.
19726         * config/mips/mips.c (mips_expand_vselect_vconcat): Likewise.
19727         * expmed.c (init_expmed_one_mode): Update use of GET_MODE_WIDER_MODE.
19728         Avoid checking for a MODE_INT if we already know the mode is not a
19729         SCALAR_INT_MODE_P.
19730         (extract_high_half): Update use of GET_MODE_WIDER_MODE,
19731         forcing a wider mode to exist.
19732         (expmed_mult_highpart_optab): Likewise.
19733         (expmed_mult_highpart): Likewise.
19734         * expr.c (expand_expr_real_2): Update use of GET_MODE_WIDER_MODE,
19735         using else_void.
19736         * lto-streamer-in.c (lto_input_mode_table): Likewise.
19737         * optabs-query.c (find_widening_optab_handler_and_mode): Likewise.
19738         * stor-layout.c (bit_field_mode_iterator::next_mode): Likewise.
19739         * internal-fn.c (expand_mul_overflow): Update use of
19740         GET_MODE_2XWIDER_MODE.
19741         * omp-low.c (omp_clause_aligned_alignment): Likewise.
19742         * tree-ssa-math-opts.c (convert_mult_to_widen): Update use of
19743         GET_MODE_WIDER_MODE.
19744         (convert_plusminus_to_widen): Likewise.
19745         * tree-switch-conversion.c (array_value_type): Likewise.
19746         * var-tracking.c (emit_note_insn_var_location): Likewise.
19747         * tree-vrp.c (simplify_float_conversion_using_ranges): Likewise.
19748         Return false inside rather than outside the loop if no wider mode
19749         exists
19750         * optabs.c (expand_binop): Update use of GET_MODE_WIDER_MODE
19751         and GET_MODE_2XWIDER_MODE
19752         (can_compare_p): Use else_void.
19753         * gdbhooks.py (OptMachineModePrinter): New class.
19754         (build_pretty_printer): Use it for opt_mode.
19756 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19757             Alan Hayward  <alan.hayward@arm.com>
19758             David Sherwood  <david.sherwood@arm.com>
19760         * tree-switch-conversion.c (array_value_type): Only read TYPE_MODE
19761         once.  Use get_narrowest_mode instead of GET_CLASS_NARROWEST_MODE.
19763 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19764             Alan Hayward  <alan.hayward@arm.com>
19765             David Sherwood  <david.sherwood@arm.com>
19767         * machmode.h (mode_traits): New structure.
19768         (get_narrowest_mode): New function.
19769         (mode_iterator::start): Likewise.
19770         (mode_iterator::iterate_p): Likewise.
19771         (mode_iterator::get_wider): Likewise.
19772         (mode_iterator::get_known_wider): Likewise.
19773         (mode_iterator::get_2xwider): Likewise.
19774         (FOR_EACH_MODE_IN_CLASS): New mode iterator.
19775         (FOR_EACH_MODE): Likewise.
19776         (FOR_EACH_MODE_FROM): Likewise.
19777         (FOR_EACH_MODE_UNTIL): Likewise.
19778         (FOR_EACH_WIDER_MODE): Likewise.
19779         (FOR_EACH_2XWIDER_MODE): Likewise.
19780         * builtins.c (expand_builtin_strlen): Use new mode iterators.
19781         * combine.c (simplify_comparison): Likewise
19782         * config/i386/i386.c (type_natural_mode): Likewise.
19783         * cse.c (cse_insn): Likewise.
19784         * dse.c (find_shift_sequence): Likewise.
19785         * emit-rtl.c (init_derived_machine_modes): Likewise.
19786         (init_emit_once): Likewise.
19787         * explow.c (hard_function_value): Likewise.
19788         * expmed.c (extract_fixed_bit_field_1): Likewise.
19789         (extract_bit_field_1): Likewise.
19790         (expand_divmod): Likewise.
19791         (emit_store_flag_1): Likewise.
19792         * expr.c (init_expr_target): Likewise.
19793         (convert_move): Likewise.
19794         (alignment_for_piecewise_move): Likewise.
19795         (widest_int_mode_for_size): Likewise.
19796         (emit_block_move_via_movmem): Likewise.
19797         (copy_blkmode_to_reg): Likewise.
19798         (set_storage_via_setmem): Likewise.
19799         (compress_float_constant): Likewise.
19800         * omp-low.c (omp_clause_aligned_alignment): Likewise.
19801         * optabs-query.c (get_best_extraction_insn): Likewise.
19802         * optabs.c (expand_binop): Likewise.
19803         (expand_twoval_unop): Likewise.
19804         (expand_twoval_binop): Likewise.
19805         (widen_leading): Likewise.
19806         (widen_bswap): Likewise.
19807         (expand_parity): Likewise.
19808         (expand_unop): Likewise.
19809         (prepare_cmp_insn): Likewise.
19810         (prepare_float_lib_cmp): Likewise.
19811         (expand_float): Likewise.
19812         (expand_fix): Likewise.
19813         (expand_sfix_optab): Likewise.
19814         * postreload.c (move2add_use_add2_insn): Likewise.
19815         * reg-stack.c (reg_to_stack): Likewise.
19816         * reginfo.c (choose_hard_reg_mode): Likewise.
19817         * rtlanal.c (init_num_sign_bit_copies_in_rep): Likewise.
19818         * stor-layout.c (mode_for_size): Likewise.
19819         (smallest_mode_for_size): Likewise.
19820         (mode_for_vector): Likewise.
19821         (finish_bitfield_representative): Likewise.
19822         * tree-ssa-math-opts.c (target_supports_divmod_p): Likewise.
19823         * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
19824         * tree-vect-stmts.c (vectorizable_conversion): Likewise.
19825         * var-tracking.c (prepare_call_arguments): Likewise.
19827 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19828             Alan Hayward  <alan.hayward@arm.com>
19829             David Sherwood  <david.sherwood@arm.com>
19831         * genconditions.c (write_header): Add a "#define USE_ENUM_MODES".
19832         * genmodes.c (emit_insn_modes_h): Define FOOmode to E_FOOmode if
19833         USE_ENUM_MODES is defined and to ((void) 0, E_FOOmode) otherwise.
19834         * machmode.h (mode_size): Move earlier in file.
19835         (mode_precision): Likewise.
19836         (mode_inner): Likewise.
19837         (mode_nunits): Likewise.
19838         (mode_unit_size): Likewise.
19839         (unit_unit_precision): Likewise.
19840         (mode_wider): Likewise.
19841         (mode_2xwider): Likewise.
19842         (machine_mode): New class.
19843         (mode_to_bytes): New function.
19844         (mode_to_bits): Likewise.
19845         (mode_to_precision): Likewise.
19846         (mode_to_inner): Likewise.
19847         (mode_to_unit_size): Likewise.
19848         (mode_to_unit_precision): Likewise.
19849         (mode_to_nunits): Likewise.
19850         (GET_MODE_SIZE): Use mode_to_bytes.
19851         (GET_MODE_BITSIZE): Use mode_to_bits.
19852         (GET_MODE_PRECISION): Use mode_to_precision.
19853         (GET_MODE_INNER): Use mode_to_inner.
19854         (GET_MODE_UNIT_SIZE): Use mode_to_unit_size.
19855         (GET_MODE_UNIT_PRECISION): Use mode_to_unit_precision.
19856         (GET_MODE_NUNITS): Use mode_to_nunits.
19857         * system.h (ALWAYS_INLINE): New macro.
19858         * config/powerpcspe/powerpcspe-c.c
19859         (altivec_resolve_overloaded_builtin): Use machine_mode instead of
19860         int for arg1_mode and arg2_mode.
19862 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19863             Alan Hayward  <alan.hayward@arm.com>
19864             David Sherwood  <david.sherwood@arm.com>
19866         * config/aarch64/aarch64-builtins.c (aarch64_simd_builtin_std_type):
19867         Prefix mode names with E_ in case statements.
19868         * config/aarch64/aarch64-elf.h (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
19869         * config/aarch64/aarch64.c (aarch64_split_simd_combine): Likewise.
19870         (aarch64_split_simd_move): Likewise.
19871         (aarch64_gen_storewb_pair): Likewise.
19872         (aarch64_gen_loadwb_pair): Likewise.
19873         (aarch64_gen_store_pair): Likewise.
19874         (aarch64_gen_load_pair): Likewise.
19875         (aarch64_get_condition_code_1): Likewise.
19876         (aarch64_constant_pool_reload_icode): Likewise.
19877         (get_rsqrte_type): Likewise.
19878         (get_rsqrts_type): Likewise.
19879         (get_recpe_type): Likewise.
19880         (get_recps_type): Likewise.
19881         (aarch64_gimplify_va_arg_expr): Likewise.
19882         (aarch64_simd_container_mode): Likewise.
19883         (aarch64_emit_load_exclusive): Likewise.
19884         (aarch64_emit_store_exclusive): Likewise.
19885         (aarch64_expand_compare_and_swap): Likewise.
19886         (aarch64_gen_atomic_cas): Likewise.
19887         (aarch64_emit_bic): Likewise.
19888         (aarch64_emit_atomic_swap): Likewise.
19889         (aarch64_emit_atomic_load_op): Likewise.
19890         (aarch64_evpc_trn): Likewise.
19891         (aarch64_evpc_uzp): Likewise.
19892         (aarch64_evpc_zip): Likewise.
19893         (aarch64_evpc_ext): Likewise.
19894         (aarch64_evpc_rev): Likewise.
19895         (aarch64_evpc_dup): Likewise.
19896         (aarch64_gen_ccmp_first): Likewise.
19897         (aarch64_gen_ccmp_next): Likewise.
19898         * config/alpha/alpha.c (alpha_scalar_mode_supported_p): Likewise.
19899         (alpha_emit_xfloating_libcall): Likewise.
19900         (emit_insxl): Likewise.
19901         (alpha_arg_type): Likewise.
19902         * config/arc/arc.c (arc_vector_mode_supported_p): Likewise.
19903         (arc_preferred_simd_mode): Likewise.
19904         (arc_secondary_reload): Likewise.
19905         (get_arc_condition_code): Likewise.
19906         (arc_print_operand): Likewise.
19907         (arc_legitimate_constant_p): Likewise.
19908         * config/arc/arc.h (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
19909         * config/arc/arc.md (casesi_load): Likewise.
19910         (casesi_compact_jump): Likewise.
19911         * config/arc/predicates.md (proper_comparison_operator): Likewise.
19912         (cc_use_register): Likewise.
19913         * config/arm/aout.h (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
19914         * config/arm/arm-builtins.c (arm_simd_builtin_std_type): Likewise.
19915         (arm_init_iwmmxt_builtins): Likewise.
19916         * config/arm/arm.c (thumb1_size_rtx_costs): Likewise.
19917         (neon_expand_vector_init): Likewise.
19918         (arm_attr_length_move_neon): Likewise.
19919         (maybe_get_arm_condition_code): Likewise.
19920         (arm_emit_vector_const): Likewise.
19921         (arm_preferred_simd_mode): Likewise.
19922         (arm_output_iwmmxt_tinsr): Likewise.
19923         (thumb1_output_casesi): Likewise.
19924         (thumb2_output_casesi): Likewise.
19925         (arm_emit_load_exclusive): Likewise.
19926         (arm_emit_store_exclusive): Likewise.
19927         (arm_expand_compare_and_swap): Likewise.
19928         (arm_evpc_neon_vuzp): Likewise.
19929         (arm_evpc_neon_vzip): Likewise.
19930         (arm_evpc_neon_vrev): Likewise.
19931         (arm_evpc_neon_vtrn): Likewise.
19932         (arm_evpc_neon_vext): Likewise.
19933         (arm_validize_comparison): Likewise.
19934         * config/arm/neon.md (neon_vc<cmp_op><mode>): Likewise.
19935         * config/avr/avr-c.c (avr_resolve_overloaded_builtin): Likewise.
19936         * config/avr/avr.c (avr_rtx_costs_1): Likewise.
19937         * config/c6x/c6x.c (c6x_vector_mode_supported_p): Likewise.
19938         (c6x_preferred_simd_mode): Likewise.
19939         * config/epiphany/epiphany.c (get_epiphany_condition_code): Likewise.
19940         (epiphany_rtx_costs): Likewise.
19941         * config/epiphany/predicates.md (proper_comparison_operator):
19942         Likewise.
19943         * config/frv/frv.c (condexec_memory_operand): Likewise.
19944         (frv_emit_move): Likewise.
19945         (output_move_single): Likewise.
19946         (output_condmove_single): Likewise.
19947         (frv_hard_regno_mode_ok): Likewise.
19948         (frv_matching_accg_mode): Likewise.
19949         * config/h8300/h8300.c (split_adds_subs): Likewise.
19950         (h8300_rtx_costs): Likewise.
19951         (h8300_print_operand): Likewise.
19952         (compute_mov_length): Likewise.
19953         (output_logical_op): Likewise.
19954         (compute_logical_op_length): Likewise.
19955         (compute_logical_op_cc): Likewise.
19956         (h8300_shift_needs_scratch_p): Likewise.
19957         (output_a_shift): Likewise.
19958         (compute_a_shift_length): Likewise.
19959         (compute_a_shift_cc): Likewise.
19960         (expand_a_rotate): Likewise.
19961         (output_a_rotate): Likewise.
19962         * config/i386/i386.c (classify_argument): Likewise.
19963         (function_arg_advance_32): Likewise.
19964         (function_arg_32): Likewise.
19965         (function_arg_64): Likewise.
19966         (function_value_64): Likewise.
19967         (ix86_gimplify_va_arg): Likewise.
19968         (ix86_legitimate_constant_p): Likewise.
19969         (put_condition_code): Likewise.
19970         (split_double_mode): Likewise.
19971         (ix86_avx256_split_vector_move_misalign): Likewise.
19972         (ix86_expand_vector_logical_operator): Likewise.
19973         (ix86_split_idivmod): Likewise.
19974         (ix86_expand_adjust_ufix_to_sfix_si): Likewise.
19975         (ix86_build_const_vector): Likewise.
19976         (ix86_build_signbit_mask): Likewise.
19977         (ix86_match_ccmode): Likewise.
19978         (ix86_cc_modes_compatible): Likewise.
19979         (ix86_expand_branch): Likewise.
19980         (ix86_expand_sse_cmp): Likewise.
19981         (ix86_expand_sse_movcc): Likewise.
19982         (ix86_expand_int_sse_cmp): Likewise.
19983         (ix86_expand_vec_perm_vpermi2): Likewise.
19984         (ix86_expand_vec_perm): Likewise.
19985         (ix86_expand_sse_unpack): Likewise.
19986         (ix86_expand_int_addcc): Likewise.
19987         (ix86_split_to_parts): Likewise.
19988         (ix86_vectorize_builtin_gather): Likewise.
19989         (ix86_vectorize_builtin_scatter): Likewise.
19990         (avx_vpermilp_parallel): Likewise.
19991         (inline_memory_move_cost): Likewise.
19992         (ix86_tieable_integer_mode_p): Likewise.
19993         (x86_maybe_negate_const_int): Likewise.
19994         (ix86_expand_vector_init_duplicate): Likewise.
19995         (ix86_expand_vector_init_one_nonzero): Likewise.
19996         (ix86_expand_vector_init_one_var): Likewise.
19997         (ix86_expand_vector_init_concat): Likewise.
19998         (ix86_expand_vector_init_interleave): Likewise.
19999         (ix86_expand_vector_init_general): Likewise.
20000         (ix86_expand_vector_set): Likewise.
20001         (ix86_expand_vector_extract): Likewise.
20002         (emit_reduc_half): Likewise.
20003         (ix86_emit_i387_round): Likewise.
20004         (ix86_mangle_type): Likewise.
20005         (ix86_expand_round_sse4): Likewise.
20006         (expand_vec_perm_blend): Likewise.
20007         (canonicalize_vector_int_perm): Likewise.
20008         (ix86_expand_vec_one_operand_perm_avx512): Likewise.
20009         (expand_vec_perm_1): Likewise.
20010         (expand_vec_perm_interleave3): Likewise.
20011         (expand_vec_perm_even_odd_pack): Likewise.
20012         (expand_vec_perm_even_odd_1): Likewise.
20013         (expand_vec_perm_broadcast_1): Likewise.
20014         (ix86_vectorize_vec_perm_const_ok): Likewise.
20015         (ix86_expand_vecop_qihi): Likewise.
20016         (ix86_expand_mul_widen_hilo): Likewise.
20017         (ix86_expand_sse2_abs): Likewise.
20018         (ix86_expand_pextr): Likewise.
20019         (ix86_expand_pinsr): Likewise.
20020         (ix86_preferred_simd_mode): Likewise.
20021         (ix86_simd_clone_compute_vecsize_and_simdlen): Likewise.
20022         * config/i386/sse.md (*andnot<mode>3): Likewise.
20023         (<mask_codefor><code><mode>3<mask_name>): Likewise.
20024         (*<code><mode>3): Likewise.
20025         * config/ia64/ia64.c (ia64_expand_vecint_compare): Likewise.
20026         (ia64_expand_atomic_op): Likewise.
20027         (ia64_arg_type): Likewise.
20028         (ia64_mode_to_int): Likewise.
20029         (ia64_scalar_mode_supported_p): Likewise.
20030         (ia64_vector_mode_supported_p): Likewise.
20031         (expand_vec_perm_broadcast): Likewise.
20032         * config/iq2000/iq2000.c (iq2000_move_1word): Likewise.
20033         (iq2000_function_arg_advance): Likewise.
20034         (iq2000_function_arg): Likewise.
20035         * config/m32c/m32c.c (m32c_preferred_reload_class): Likewise.
20036         * config/m68k/m68k.c (output_dbcc_and_branch): Likewise.
20037         (m68k_libcall_value): Likewise.
20038         (m68k_function_value): Likewise.
20039         (sched_attr_op_type): Likewise.
20040         * config/mcore/mcore.c (mcore_output_move): Likewise.
20041         * config/microblaze/microblaze.c (microblaze_function_arg_advance):
20042         Likewise.
20043         (microblaze_function_arg): Likewise.
20044         * config/mips/mips.c (mips16_build_call_stub): Likewise.
20045         (mips_print_operand): Likewise.
20046         (mips_mode_ok_for_mov_fmt_p): Likewise.
20047         (mips_vector_mode_supported_p): Likewise.
20048         (mips_preferred_simd_mode): Likewise.
20049         (mips_expand_vpc_loongson_even_odd): Likewise.
20050         (mips_expand_vec_unpack): Likewise.
20051         (mips_expand_vi_broadcast): Likewise.
20052         (mips_expand_vector_init): Likewise.
20053         (mips_expand_vec_reduc): Likewise.
20054         (mips_expand_msa_cmp): Likewise.
20055         * config/mips/mips.md (casesi_internal_mips16_<mode>): Likewise.
20056         * config/mn10300/mn10300.c (mn10300_print_operand): Likewise.
20057         (cc_flags_for_mode): Likewise.
20058         * config/msp430/msp430.c (msp430_print_operand): Likewise.
20059         * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Likewise.
20060         (nds32_output_casesi_pc_relative): Likewise.
20061         * config/nds32/nds32.h (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
20062         * config/nvptx/nvptx.c (nvptx_ptx_type_from_mode): Likewise.
20063         (nvptx_gen_unpack): Likewise.
20064         (nvptx_gen_pack): Likewise.
20065         (nvptx_gen_shuffle): Likewise.
20066         (nvptx_gen_wcast): Likewise.
20067         (nvptx_preferred_simd_mode): Likewise.
20068         * config/pa/pa.c (pa_secondary_reload): Likewise.
20069         * config/pa/predicates.md (base14_operand): Likewise.
20070         * config/powerpcspe/powerpcspe-c.c
20071         (altivec_resolve_overloaded_builtin): Likewise.
20072         * config/powerpcspe/powerpcspe.c (rs6000_setup_reg_addr_masks):
20073         Likewise.
20074         (rs6000_preferred_simd_mode): Likewise.
20075         (output_vec_const_move): Likewise.
20076         (rs6000_expand_vector_extract): Likewise.
20077         (rs6000_split_vec_extract_var): Likewise.
20078         (reg_offset_addressing_ok_p): Likewise.
20079         (rs6000_legitimate_offset_address_p): Likewise.
20080         (rs6000_legitimize_address): Likewise.
20081         (rs6000_emit_set_const): Likewise.
20082         (rs6000_const_vec): Likewise.
20083         (rs6000_emit_move): Likewise.
20084         (spe_build_register_parallel): Likewise.
20085         (rs6000_darwin64_record_arg_recurse): Likewise.
20086         (swap_selector_for_mode): Likewise.
20087         (spe_init_builtins): Likewise.
20088         (paired_init_builtins): Likewise.
20089         (altivec_init_builtins): Likewise.
20090         (do_load_for_compare): Likewise.
20091         (rs6000_generate_compare): Likewise.
20092         (rs6000_expand_float128_convert): Likewise.
20093         (emit_load_locked): Likewise.
20094         (emit_store_conditional): Likewise.
20095         (rs6000_output_function_epilogue): Likewise.
20096         (rs6000_handle_altivec_attribute): Likewise.
20097         (rs6000_function_value): Likewise.
20098         (emit_fusion_gpr_load): Likewise.
20099         (emit_fusion_p9_load): Likewise.
20100         (emit_fusion_p9_store): Likewise.
20101         * config/powerpcspe/predicates.md (easy_fp_constant): Likewise.
20102         (fusion_gpr_mem_load): Likewise.
20103         (fusion_addis_mem_combo_load): Likewise.
20104         (fusion_addis_mem_combo_store): Likewise.
20105         * config/rs6000/predicates.md (easy_fp_constant): Likewise.
20106         (fusion_gpr_mem_load): Likewise.
20107         (fusion_addis_mem_combo_load): Likewise.
20108         (fusion_addis_mem_combo_store): Likewise.
20109         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
20110         Likewise.
20111         * config/rs6000/rs6000-string.c (do_load_for_compare): Likewise.
20112         * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): Likewise.
20113         (rs6000_preferred_simd_mode): Likewise.
20114         (output_vec_const_move): Likewise.
20115         (rs6000_expand_vector_extract): Likewise.
20116         (rs6000_split_vec_extract_var): Likewise.
20117         (reg_offset_addressing_ok_p): Likewise.
20118         (rs6000_legitimate_offset_address_p): Likewise.
20119         (rs6000_legitimize_address): Likewise.
20120         (rs6000_emit_set_const): Likewise.
20121         (rs6000_const_vec): Likewise.
20122         (rs6000_emit_move): Likewise.
20123         (rs6000_darwin64_record_arg_recurse): Likewise.
20124         (swap_selector_for_mode): Likewise.
20125         (paired_init_builtins): Likewise.
20126         (altivec_init_builtins): Likewise.
20127         (rs6000_expand_float128_convert): Likewise.
20128         (emit_load_locked): Likewise.
20129         (emit_store_conditional): Likewise.
20130         (rs6000_output_function_epilogue): Likewise.
20131         (rs6000_handle_altivec_attribute): Likewise.
20132         (rs6000_function_value): Likewise.
20133         (emit_fusion_gpr_load): Likewise.
20134         (emit_fusion_p9_load): Likewise.
20135         (emit_fusion_p9_store): Likewise.
20136         * config/rx/rx.c (rx_gen_move_template): Likewise.
20137         (flags_from_mode): Likewise.
20138         * config/s390/predicates.md (s390_alc_comparison): Likewise.
20139         (s390_slb_comparison): Likewise.
20140         * config/s390/s390.c (s390_handle_vectorbool_attribute): Likewise.
20141         (s390_vector_mode_supported_p): Likewise.
20142         (s390_cc_modes_compatible): Likewise.
20143         (s390_match_ccmode_set): Likewise.
20144         (s390_canonicalize_comparison): Likewise.
20145         (s390_emit_compare_and_swap): Likewise.
20146         (s390_branch_condition_mask): Likewise.
20147         (s390_rtx_costs): Likewise.
20148         (s390_secondary_reload): Likewise.
20149         (__SECONDARY_RELOAD_CASE): Likewise.
20150         (s390_expand_cs): Likewise.
20151         (s390_preferred_simd_mode): Likewise.
20152         * config/s390/vx-builtins.md (vec_packsu_u<mode>): Likewise.
20153         * config/sh/sh.c (sh_print_operand): Likewise.
20154         (dump_table): Likewise.
20155         (sh_secondary_reload): Likewise.
20156         * config/sh/sh.h (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
20157         * config/sh/sh.md (casesi_worker_1): Likewise.
20158         (casesi_worker_2): Likewise.
20159         * config/sparc/predicates.md (icc_comparison_operator): Likewise.
20160         (fcc_comparison_operator): Likewise.
20161         * config/sparc/sparc.c (sparc_expand_move): Likewise.
20162         (emit_soft_tfmode_cvt): Likewise.
20163         (sparc_preferred_simd_mode): Likewise.
20164         (output_cbranch): Likewise.
20165         (sparc_print_operand): Likewise.
20166         (sparc_expand_vec_perm_bmask): Likewise.
20167         (vector_init_bshuffle): Likewise.
20168         * config/spu/spu.c (spu_scalar_mode_supported_p): Likewise.
20169         (spu_vector_mode_supported_p): Likewise.
20170         (spu_expand_insv): Likewise.
20171         (spu_emit_branch_or_set): Likewise.
20172         (spu_handle_vector_attribute): Likewise.
20173         (spu_builtin_splats): Likewise.
20174         (spu_builtin_extract): Likewise.
20175         (spu_builtin_promote): Likewise.
20176         (spu_expand_sign_extend): Likewise.
20177         * config/tilegx/tilegx.c (tilegx_scalar_mode_supported_p): Likewise.
20178         (tilegx_simd_int): Likewise.
20179         * config/tilepro/tilepro.c (tilepro_scalar_mode_supported_p): Likewise.
20180         (tilepro_simd_int): Likewise.
20181         * config/v850/v850.c (const_double_split): Likewise.
20182         (v850_print_operand): Likewise.
20183         (ep_memory_offset): Likewise.
20184         * config/vax/vax.c (vax_rtx_costs): Likewise.
20185         (vax_output_int_move): Likewise.
20186         (vax_output_int_add): Likewise.
20187         (vax_output_int_subtract): Likewise.
20188         * config/visium/predicates.md (visium_branch_operator): Likewise.
20189         * config/visium/visium.c (rtx_ok_for_offset_p): Likewise.
20190         (visium_print_operand_address): Likewise.
20191         * config/visium/visium.h (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
20192         * config/xtensa/xtensa.c (xtensa_mem_offset): Likewise.
20193         (xtensa_expand_conditional_branch): Likewise.
20194         (xtensa_copy_incoming_a7): Likewise.
20195         (xtensa_output_literal): Likewise.
20196         * dfp.c (decimal_real_maxval): Likewise.
20197         * targhooks.c (default_libgcc_floating_mode_supported_p): Likewise.
20199 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
20200             Alan Hayward  <alan.hayward@arm.com>
20201             David Sherwood  <david.sherwood@arm.com>
20203         * genmodes.c (mode_size_inline): Add an E_ prefix to mode names.
20204         (mode_nunits_inline): Likewise.
20205         (mode_inner_inline): Likewise.
20206         (mode_unit_size_inline): Likewise.
20207         (mode_unit_precision_inline): Likewise.
20208         (emit_insn_modes_h): Likewise.  Also emit a #define of the
20209         unprefixed name.
20210         (emit_mode_wider): Add an E_ prefix to mode names.
20211         (emit_mode_complex): Likewise.
20212         (emit_mode_inner): Likewise.
20213         (emit_mode_adjustments): Likewise.
20214         (emit_mode_int_n): Likewise.
20215         * config/aarch64/aarch64-builtins.c (v8qi_UP, v4hi_UP, v4hf_UP)
20216         (v2si_UP, v2sf_UP, v1df_UP, di_UP, df_UP, v16qi_UP, v8hi_UP, v8hf_UP)
20217         (v4si_UP, v4sf_UP, v2di_UP, v2df_UP, ti_UP, oi_UP, ci_UP, xi_UP)
20218         (si_UP, sf_UP, hi_UP, hf_UP, qi_UP): Likewise.
20219         (CRC32_BUILTIN, ENTRY): Likewise.
20220         * config/aarch64/aarch64.c (aarch64_push_regs): Likewise.
20221         (aarch64_pop_regs): Likewise.
20222         (aarch64_process_components): Likewise.
20223         * config/alpha/alpha.c (alpha_emit_conditional_move): Likewise.
20224         * config/arm/arm-builtins.c (v8qi_UP, v4hi_UP, v4hf_UP, v2si_UP)
20225         (v2sf_UP, di_UP, v16qi_UP, v8hi_UP, v8hf_UP, v4si_UP, v4sf_UP)
20226         (v2di_UP, ti_UP, ei_UP, oi_UP, hf_UP, si_UP, void_UP): Likewise.
20227         * config/arm/arm.c (arm_init_libfuncs): Likewise.
20228         * config/i386/i386-builtin-types.awk (ix86_builtin_type_vect_mode):
20229         Likewise.
20230         * config/i386/i386-builtin.def (pcmpestr): Likewise.
20231         (pcmpistr): Likewise.
20232         * config/microblaze/microblaze.c (double_memory_operand): Likewise.
20233         * config/mmix/mmix.c (mmix_output_condition): Likewise.
20234         * config/powerpcspe/powerpcspe.c (rs6000_init_hard_regno_mode_ok):
20235         Likewise.
20236         * config/rl78/rl78.c (mduc_regs): Likewise.
20237         * config/rs6000/rs6000.c (rs6000_init_hard_regno_mode_ok): Likewise.
20238         (htm_expand_builtin): Likewise.
20239         * config/sh/sh.h (REGISTER_NATURAL_MODE): Likewise.
20240         * config/sparc/sparc.c (emit_save_or_restore_regs): Likewise.
20241         * config/xtensa/xtensa.c (print_operand): Likewise.
20242         * expmed.h (NUM_MODE_PARTIAL_INT): Likewise.
20243         (NUM_MODE_VECTOR_INT): Likewise.
20244         * genoutput.c (null_operand): Likewise.
20245         (output_operand_data): Likewise.
20246         * genrecog.c (print_parameter_value): Likewise.
20247         * lra.c (debug_operand_data): Likewise.
20249 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
20251         * dwarf2out.c (scompare_loc_descriptor_wide)
20252         (scompare_loc_descriptor_narrow): New functions, split out from...
20253         (scompare_loc_descriptor): ...here.
20254         * expmed.c (emit_store_flag_int): New function, split out from...
20255         (emit_store_flag): ...here.
20257 2017-08-30  Richard Biener  <rguenther@suse.de>
20259         * dwarf2out.c (dwarf2out_finish): Remove setting AT_pubnames.
20260         (dwarf2out_early_finish): Move setting of AT_pubnames from
20261         early debug output to early finish.
20263 2017-08-30  Jozef Lawrynowicz  <jozef.l@somniumtech.com>
20265         * gcc/config/msp430/msp430.h: Pass -mcode/data-region to the linker
20266         and -mdata-region to the assembler.
20268 2017-08-30  Richard Biener  <rguenther@suse.de>
20270         * dwarf2out.c (add_dwarf_attr): Check we don't add duplicate
20271         attributes.
20272         (gen_subprogram_die): Add DW_AT_object_pointer only early.
20273         (dwarf2out_early_global_decl): Only generate a DIE for the
20274         abstract origin if it doesn't already exist or is a declaration DIE.
20275         (resolve_addr): Do not add the linkage name twice when
20276         generating a stub DIE for the DW_TAG_GNU_call_site target.
20278 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
20280         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
20281         Use machine_mode rather than int for arg1_mode.
20283 2017-08-29  Michael Meissner  <meissner@linux.vnet.ibm.com>
20285         PR target/82015
20286         * config/rs6000/rs6000.c (rs6000_expand_binop_builtin): Insure
20287         that the second argument of the built-in functions to unpack
20288         128-bit scalar types to 64-bit values is 0 or 1.  Change to use a
20289         switch statement instead a lot of if statements.
20290         * config/rs6000/rs6000.md (unpack<mode>, FMOVE128_VSX iterator):
20291         Allow 64-bit values to be in Altivec registers as well as
20292         traditional floating point registers.
20293         (pack<mode>, FMOVE128_VSX iterator): Likewise.
20295 2017-08-29  Alexander Monakov  <amonakov@ispras.ru>
20297         * ira-costs.c (record_address_regs): Handle both operands of PLUS for
20298         MAX_REGS_PER_ADDRESS == 1.
20300 2017-08-29  Uros Bizjak  <ubizjak@gmail.com>
20302         * config/i386/i386.opt (flag_fentry): Do not init to -1.
20303         * config/i386/i386.c (ix86_option_override_internal): Simplify
20304         setting of opts->x_flag_entry.
20306 2017-08-29  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
20307             Jakub Jelinek  <jakub@redhat.com>
20308             Richard Biener  <rguenther@suse.de>
20310         PR tree-optimization/81503
20311         * gimple-ssa-strength-reduction.c (replace_mult_candidate): Ensure
20312         folded constant fits in the target type; reorder tests for clarity.
20314 2017-08-29  Martin Liska  <mliska@suse.cz>
20316         * passes.def: Include pass_lower_switch.
20317         * stmt.c (dump_case_nodes): Remove and move to
20318         tree-switch-conversion.
20319         (case_values_threshold): Likewise.
20320         (expand_switch_as_decision_tree_p): Likewise.
20321         (emit_case_decision_tree): Likewise.
20322         (expand_case): Likewise.
20323         (balance_case_nodes): Likewise.
20324         (node_has_low_bound): Likewise.
20325         (node_has_high_bound): Likewise.
20326         (node_is_bounded): Likewise.
20327         (emit_case_nodes): Likewise.
20328         (struct simple_case_node): New struct.
20329         (add_case_node): Remove.
20330         (emit_case_dispatch_table): Use vector instead of case_list.
20331         (reset_out_edges_aux): Remove.
20332         (compute_cases_per_edge): Likewise.
20333         (expand_case): Build list of simple_case_node.
20334         (expand_sjlj_dispatch_table): Use it.
20335         * tree-switch-conversion.c (struct case_node): Moved from
20336         stmt.c and adjusted.
20337         (emit_case_nodes): Likewise.
20338         (node_has_low_bound): Likewise.
20339         (node_has_high_bound): Likewise.
20340         (node_is_bounded): Likewise.
20341         (case_values_threshold): Likewise.
20342         (reset_out_edges_aux): Likewise.
20343         (compute_cases_per_edge): Likewise.
20344         (add_case_node): Likewise.
20345         (dump_case_nodes): Likewise.
20346         (balance_case_nodes): Likewise.
20347         (expand_switch_as_decision_tree_p): Likewise.
20348         (emit_jump): Likewise.
20349         (emit_case_decision_tree): Likewise.
20350         (try_switch_expansion): Likewise.
20351         (do_jump_if_equal): Likewise.
20352         (emit_cmp_and_jump_insns): Likewise.
20353         (fix_phi_operands_for_edge): New function.
20354         (record_phi_operand_mapping): Likewise.
20355         (class pass_lower_switch): New pass.
20356         (pass_lower_switch::execute): New function.
20357         (make_pass_lower_switch): Likewise.
20358         (conditional_probability):
20359         * timevar.def: Add TV_TREE_SWITCH_LOWERING.
20360         * tree-pass.h: Add make_pass_lower_switch.
20362 2017-08-29  Jozef Lawrynowicz  <jozef.l@somniumtech.com>
20364         PR target/80993
20365         * gcc/config/msp430/msp430.c (msp430_attr): Mark interrupt
20366         handlers as used.
20368 2017-08-29  Richard Biener  <rguenther@suse.de>
20370         * dwarf2out.c (add_dwarf_attr): When checking is enabled verify
20371         we do not add a DW_AT_inline attribute twice.
20372         (gen_subprogram_die): Remove code setting DW_AT_inline on
20373         DECL_ABSTRACT_P nodes.
20375 2017-08-29  Richard Sandiford  <richard.sandiford@linaro.org>
20377         * gimplify.c (gimplify_call_expr): Copy the nothrow flag to
20378         calls to internal functions.
20379         (gimplify_modify_expr): Likewise.
20380         * tree-call-cdce.c (use_internal_fn): Likewise.
20381         * tree-ssa-math-opts.c (pass_cse_reciprocals::execute): Likewise.
20382         (convert_to_divmod): Set the nothrow flag.
20383         * tree-if-conv.c (predicate_mem_writes):  Likewise.
20384         * tree-vect-stmts.c (vectorizable_mask_load_store): Likewise.
20385         (vectorizable_call): Likewise.
20386         (vectorizable_store): Likewise.
20387         (vectorizable_load): Likewise.
20388         * tree-vect-patterns.c (vect_recog_pow_pattern): Likewise.
20389         (vect_recog_mask_conversion_pattern): Likewise.
20391 2017-08-29  Martin Liska  <mliska@suse.cz>
20393         PR other/39851
20394         * gcc.c (driver_handle_option): Add new argument.
20395         * opts-common.c (handle_option): Pass
20396         target_option_override_hook.
20397         * opts-global.c (lang_handle_option): Add new option.
20398         (set_default_handlers):  Add new argument.
20399         (decode_options): Likewise.
20400         * opts.c (target_handle_option): Likewise.
20401         (common_handle_option): Call target_option_override_hook.
20402         * opts.h (struct cl_option_handler_func): Add hook for
20403         target option override.
20404         (struct cl_option_handlers): Likewise.
20405         (set_default_handlers): Add new argument.
20406         (decode_options): Likewise.
20407         (common_handle_option): Likewise.
20408         (target_handle_option): Likewise.
20409         * toplev.c (toplev::main): Pass targetm.target_option.override
20410         hook.
20412 2017-08-29  Richard Biener  <rguenther@suse.de>
20413         Dominik Infuehr <dominik.infuehr@theobroma-systems.com>
20415         * tree-vect-slp.c (vect_bb_slp_scalar_cost): Properly confine
20416         life to the active subtree.
20418 2017-08-28  Jeff Law  <law@redhat.com>
20420         * tree-ssa-dom.c (edge_info::record_simple_equiv): Call
20421         derive_equivalences.
20422         (derive_equivalences_from_bit_ior, record_temporary_equivalences):
20423         Code moved into....
20424         (edge_info::derive_equivalences): New private member function
20426         * tree-ssa-dom.c (class edge_info): Changed from a struct
20427         to a class.  Add ctor/dtor, methods and data members.
20428         (edge_info::edge_info): Renamed from allocate_edge_info.
20429         Initialize additional members.
20430         (edge_info::~edge_info): New.
20431         (free_dom_edge_info): Delete the edge info.
20432         (record_edge_info): Use new class & associated member functions.
20433         Tighten forms for testing for edge equivalences.
20434         (record_temporary_equivalences): Iterate over the simple
20435         equivalences rather than assuming there's only one per edge.
20436         (cprop_into_successor_phis): Iterate over the simple
20437         equivalences rather than assuming there's only one per edge.
20438         (optimize_stmt): Use operand_equal_p rather than pointer
20439         equality for mini-DSE code.
20441 2017-08-28  Nathan Sidwell  <nathan@acm.org>
20443         * gcc.c (execute): Fold SIGPIPE handling into switch
20444         statement.  Adjust internal error message.
20446 2017-08-28  Richard Biener  <rguenther@suse.de>
20448         PR debug/81993
20449         * dwarf2out.c (gen_remaining_tmpl_value_param_die_attributes):
20450         Do nothing for removed DIEs.
20452 2017-08-28  Richard Biener  <rguenther@suse.de>
20454         PR tree-optimization/81977
20455         * tree-ssa-sccvn.c (vn_reference_lookup_3): Fix look through
20456         memcpy.
20458 2017-08-28  Alexander Monakov  <amonakov@ispras.ru>
20460         PR target/80640
20461         * doc/md.texi (mem_thread_fence): Remove mention of mode.  Rewrite.
20462         * optabs.c (expand_mem_thread_fence): Emit a compiler barrier when
20463         using targetm.gen_mem_thread_fence.
20465 2017-08-27  Uros Bizjak  <ubizjak@gmail.com>
20467         PR target/81995
20468         * config/i386/i386.md (*<btsc><mode>): Change operand 2
20469         predicate to register_operand.  Reorder operands.
20470         (*btr<mode>): Ditto.
20471         (*<btsc><mode>_mask): Change operand 3 predicate to register_operand.
20472         (*btr<mode>_mask): Ditto.
20474 2017-08-25  Steven Munroe  <munroesj@gcc.gnu.org>
20476         * config.gcc (powerpc*-*-*): Add xmmintrin.h and mm_malloc.h.
20477         * config/rs6000/xmmintrin.h: New file.
20478         * config/rs6000/x86intrin.h [__ALTIVEC__]: Include xmmintrin.h.
20480 2017-08-25  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
20482         PR target/81504
20483         * config/rs6000/rs6000-p8swap.c (find_alignment_op): Add reference
20484         parameter and_insn and return it.
20485         (recombine_lvx_pattern): Insert a copy to ensure availability of
20486         the base register of the copied masking operation at the point of
20487         the instruction replacement.
20488         (recombine_stvx_pattern): Likewise.
20490 2017-08-25  Michael Meissner  <meissner@linux.vnet.ibm.com>
20492         * config/rs6000/rs6000.opt (-mpower9-dform-scalar): Delete
20493         undocumented switches.
20494         (-mpower9-dform-vector): Likewise.
20495         (-mpower9-dform): Likewise.
20496         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
20497         comments to delete references to -mpower9-dform* switches.
20498         * config/rs6000/predicates.md (vsx_quad_dform_memory_operand):
20499         Delete reference to -mpower9-dform* switches, test for
20500         -mpower9-vector instead.
20501         * config/rs6000/rs6000-cpus.def (ISA_3_0_MASKS_SERVER): Likewise.
20502         (OTHER_P9_VECTOR_MASKS): Likewise.
20503         (POWERPC_MASKS): Likewise.
20504         * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): Change
20505         tests against -mpower9-dform* to -mpower9-vector.  Delete code
20506         that checked for -mpower9-dform* consistancy with other options.
20507         Add test for -mpower9-misc to enable other power9 switches.
20508         (rs6000_init_hard_regno_mode_ok): Likewise.
20509         (rs6000_option_override_internal): Likewise.
20510         (rs6000_emit_prologue): Likewise.
20511         (rs6000_emit_epilogue): Likewise.
20512         (rs6000_opt_masks): Delete -mpower9-dform-{scalar,vector}.
20513         (rs6000_disable_incompatiable_switches): Delete -mpower9-dform.
20514         (emit_fusion_p9_load): Change tests for -mpower9-dform-scalar
20515         -mpower9-vector.
20516         (emit_fusion_p9_store): Likewise.
20517         * config/rs6000/rs6000.h (TARGET_P9_DFORM_SCALAR): Delete
20518         resetting these macros if the assembler does not support ISA 3.0
20519         instructions.
20520         (TARGET_P9_DFORM_VECTOR): Likewise.
20521         * config/rs6000/rs6000.md (peepholes to optimize altivec memory):
20522         Change to use -mpower9-vector instead of -mpower9-dform-scalar.
20524 2017-08-25  Alan Modra  <amodra@gmail.com>
20526         PR rtl-optimization/81747
20527         * cse.c (cse_extended_basic_block): Don't attempt to record
20528         equivalences for degenerate conditional branches that branch
20529         to their fall-through.
20531 2017-08-24  Martin Sebor  <msebor@redhat.com>
20533         PR middle-end/81908
20534         * gimple-fold.c (size_must_be_zero_p): New function.
20535         (gimple_fold_builtin_memory_op): Call it.
20537 2017-08-24  Steven Munroe  <munroesj@gcc.gnu.org>
20539         * config/rs6000/mm_malloc.h: New file.
20541 2017-08-24  Bin Cheng  <bin.cheng@arm.com>
20543         PR tree-optimization/81913
20544         * tree-ssa-loop-niter.c (number_of_iterations_cond): Skip niter
20545         analysis when either IVs in condition can wrap.
20547 2017-08-24  Uros Bizjak  <ubizjak@gmail.com>
20549         * dwarf2out.c (MAX_ARTIFICIAL_LABEL_BYTES): Increase to 40.
20550         * dwarf2cfi.c (MAX_ARTIFICIAL_LABEL_BYTES): Remove.
20552 2017-08-24  Richard Biener  <rguenther@suse.de>
20554         PR target/81921
20555         * targhooks.c (default_target_can_inline_p): Properly
20556         use target_option_default_node when no DECL_FUNCTION_SPECIFIC_TARGET
20557         is present and always compare.
20558         * config/i386/i386.c (ix86_valid_target_attribute_tree): Do not
20559         infer -mfpmath=sse from TARGET_SSE_P.
20560         (ix86_can_inline_p): Properly use target_option_default_node when
20561         no DECL_FUNCTION_SPECIFIC_TARGET is present and always compare.
20563 2017-08-24  Richard Biener  <rguenther@suse.de>
20565         PR debug/81936
20566         * dwarf2out.c (output_die): Handle flag_generate_offload like
20567         flag_generate_lto.
20568         (output_comp_unit): Likewise.
20569         (gen_array_type_die): Likewise.
20570         (dwarf2out_early_finish): Likewise.
20571         (note_variable_value_in_expr): Likewise.
20572         (dwarf2out_finish): Likewise.  Adjust assert.
20573         * cgraphunit.c (symbol_table::compile): Move setting of
20574         flag_generate_offload earlier ...
20575         (symbol_table::finalize_compilation_unit): ... here, before
20576         early debug finalization.
20578 2017-08-24  Richard Biener  <rguenther@suse.de>
20580         * config/i386/i386.c: Include symbol-summary.h, ipa-prop.h
20581         and ipa-fnsummary.h.
20582         (ix86_can_inline_p): When ix86_fpmath flags do not match
20583         check whether the callee uses FP math at all.
20585 2017-08-24  Aldy Hernandez  <aldyh@redhat.com>
20587         PR middle-end/81931
20588         * tree-ssanames.c (get_nonzero_bits): Use element_precision
20589         instead of TYPE_PRECISION.
20591 2017-08-24  Richard Sandiford  <richard.sandiford@linaro.org>
20592             Alan Hayward  <alan.hayward@arm.com>
20593             David Sherwood  <david.sherwood@arm.com>
20595         * combine.c (make_extraction): Use subreg_offset_from_lsb.
20597 2017-08-23  Daniel Santos  <daniel.santos@pobox.com>
20599         * config/i386/i386.h (ix86_frame::stack_realign_allocate_offset):
20600         Remove field.
20601         (ix86_frame::stack_realign_allocate): New field.
20602         (struct machine_frame_state): Modify comments.
20603         (machine_frame_state::sp_realigned_fp_end): New field.
20604         * config/i386/i386.c (ix86_compute_frame_layout): Rework stack frame
20605         layout calculation.
20606         (sp_valid_at): Add assertion to assure no attempt to access invalid
20607         offset of a realigned stack.
20608         (fp_valid_at): Likewise.
20609         (choose_baseaddr): Modify comments.
20610         (ix86_emit_outlined_ms2sysv_save): Adjust to changes in
20611         ix86_expand_prologue.
20612         (ix86_expand_prologue): Modify stack realignment and allocation.
20613         (ix86_expand_epilogue): Modify comments.
20614         * doc/sourcebuild.texi: Add documentation for target selectors avx2,
20615         avx2_runtime, avx512f, and avx512f_runtime.
20617 2017-08-23  Uros Bizjak  <ubizjak@gmail.com>
20619         * config/i386/i386.opt: Remove unneeded Init(0) initializations.
20620         (mstackrealign): Do not init to -1.
20621         * config/i386/i386.c (ix86_option_override_internal):
20622         Check opts_set, not opts when setting default value of
20623         opts->x_ix86_force_align_arg_pointer.
20625 2017-08-23  Richard Biener  <rguenther@suse.de>
20627         * function.c (fndecl_name): Use verbosity 1 (no arguments) for
20628         lang_hooks.decl_printable_name.
20629         * print-rtl-function.c (print_rtx_function): Likewise.
20630         * tree-pretty-print.c (dump_function_header): Likewise.
20632 2017-08-23  Richard Biener  <rguenther@suse.de>
20634         PR lto/81940
20635         * dwarf2out.c (dwarf2out_abstract_function): Handle LTO with
20636         -g0 at compile-time.
20638 2017-08-23  Tamar Christina  <tamar.christina@arm.com>
20640         PR middle-end/19706
20641         * doc/sourcebuild.texi (Other hardware attributes):
20642         Document xorsign.
20644 2017-08-23  Tamar Christina  <tamar.christina@arm.com>
20646         PR middle-end/19706
20647         * tree-ssa-math-opts.c (convert_expand_mult_copysign):
20648         Fix single-use check.
20650 2017-08-23  Thomas Preud'homme  <thomas.preudhomme@arm.com>
20652         * gcc.c (execute): Only test for SIGKILL and SIGQUIT if available.
20654 2017-08-22  Daniel Santos  <daniel.santos@pobox.com>
20656         * doc/install.texi: Modify to add more details on running selected
20657         tests.
20659 2017-08-22  Daniel Santos  <daniel.santos@pobox.com>
20661         * config/i386/i386.c (ix86_option_override_internal): Error when -mx32
20662         is combined with -mabi=ms.
20663         (ix86_function_type_abi): Limit errors for mixing -mx32 with attribute
20664         ms_abi.
20666 2017-08-22  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
20668         PR tree-optimization/81488
20669         * gimple-ssa-strength-reduction (struct slsr_cand_d): Add visited
20670         and cached_basis fields.
20671         (MAX_SPREAD): New constant.
20672         (alloc_cand_and_find_basis): Initialize new fields.
20673         (clear_visited): New function.
20674         (create_phi_basis_1): Rename from create_phi_basis, set visited
20675         and cached_basis fields.
20676         (create_phi_basis): New wrapper function.
20677         (phi_add_costs_1): Rename from phi_add_costs, add spread
20678         parameter, set visited field, short-circuit when limits reached.
20679         (phi_add_costs): New wrapper function.
20680         (record_phi_increments_1): Rename from record_phi_increments, set
20681         visited field.
20682         (record_phi_increments): New wrapper function.
20683         (phi_incr_cost_1): Rename from phi_incr_cost, set visited field.
20684         (phi_incr_cost): New wrapper function.
20685         (all_phi_incrs_profitable_1): Rename from
20686         all_phi_incrs_profitable, set visited field.
20687         (all_phi_incrs_profitable): New wrapper function.
20689 2017-08-22  Richard Sandiford  <richard.sandiford@linaro.org>
20690             Alan Hayward  <alan.hayward@arm.com>
20691             David Sherwood  <david.sherwood@arm.com>
20693         * rtl.h (paradoxical_subreg_p): Define inline, and add a version
20694         that takes the outer and inner modes.
20695         * doc/rtl.texi: Use paradoxical_subreg_p instead of a GET_MODE_SIZE
20696         comparison as the canonical test for a paradoxical subreg.
20697         * combine.c (simplify_set): Use paradoxical_subreg_p.
20698         (make_extraction): Likewise.
20699         (force_to_mode): Likewise.
20700         (rtx_equal_for_field_assignment_p): Likewise.
20701         (gen_lowpart_for_combine): Likewise.
20702         (simplify_comparison): Likewise.
20703         * cse.c (equiv_constant): Likewise.
20704         * expmed.c (store_bit_field_1): Likewise.
20705         * final.c (alter_subreg): Likewise.
20706         * fwprop.c (propagate_rtx): Likewise.
20707         (forward_propagate_subreg): Likewise.
20708         * ira-conflicts.c (ira_build_conflicts): Likewise.
20709         * lower-subreg.c (simplify_gen_subreg_concatn): Likewise.
20710         * lra-constraints.c (curr_insn_transform): Likewise.
20711         (split_reg): Likewise.
20712         * lra-eliminations.c (move_plus_up): Likewise.
20713         (lra_eliminate_regs_1): Likewise.
20714         * recog.c (general_operand): Likewise.
20715         * ree.c (combine_reaching_defs): Likewise.
20716         * reload.c (push_reload): Likewise.
20717         (find_reloads): Likewise.
20718         * reload1.c (elimination_effects): Likewise.
20719         (compute_reload_subreg_offset): Likewise.
20720         (choose_reload_regs): Likewise.
20721         * rtlanal.c (subreg_lsb_1): Likewise.
20722         * simplify-rtx.c (simplify_unary_operation_1): Likewise.
20723         (simplify_subreg): Likewise.
20724         * var-tracking.c (track_loc_p): Likewise.
20725         * emit-rtl.c (byte_lowpart_offset): Likewise.
20726         (paradoxical_subreg_p): Delete out-of-line definition.
20728 2017-08-22  Jeff Law  <law@redhat.com>
20730         PR tree-optimization/81741
20731         PR tree-optimization/71947
20732         * tree-ssa-dom.c: Include tree-inline.h.
20733         (record_temporary_equivalences): Only record SSA_NAME = SSA_NAME
20734         equivalences if one is more expensive to compute than the other.
20735         * tree-ssa-scopedtables.h (class const_or_copies): Make
20736         record_const_or_copy_raw method private.
20737         (class avail_exprs_stack): New method simplify_binary_operation.
20738         * tree-ssa-scopedtables.c (avail_exprs_stack::lookup_avail_expr): Call
20739         avail_exprs_stack::simplify_binary_operation as needed.
20740         (avail_exprs_stack::simplify_binary_operation): New function.
20742 2017-08-22  Sebastian Huber  <sebastian.huber@embedded-brains.de>
20744         * config.gcc (powerpc-*-rtems*): Add rs6000/linux64.opt.
20745         * config/rs6000/rtems.h (ASM_PREFERRED_EH_DATA_FORMAT): New define.
20746         (DOT_SYMBOLS): Likewise.
20747         (MINIMAL_TOC_SECTION_ASM_OP): Likewise.
20748         (RELOCATABLE_NEEDS_FIXUP): Likewise.
20749         (RS6000_ABI_NAME): Likewise.
20750         (TARGET_CMODEL): Likewise.
20751         (TOC_SECTION_ASM_OP): Likewise.
20752         (SET_CMODEL): New macro.
20753         (SUBSUBTARGET_OVERRIDE_OPTIONS): Evaluate cmodel options.
20755 2017-08-22  Richard Biener  <rguenther@suse.de>
20757         * tree-inline.c (remap_type_1): Change asserts on TYPE_SIZE[_UNIT]
20758         to allow for free-lang-data replacements similar to verify_type_variant.
20760 2017-08-22  Richard Sandiford  <richard.sandiford@linaro.org>
20761             Alan Hayward  <alan.hayward@arm.com>
20762             David Sherwood  <david.sherwood@arm.com>
20764         * config/aarch64/aarch64.md (casesi): Use DImode rather than
20765         VOIDmode for the LABEL_REF.
20767 2017-08-22  Richard Biener  <rguenther@suse.de>
20769         * tree-cfg.c (gimple_split_edge): Avoid reallocating target PHI nodes.
20771 2017-08-22  Richard Biener  <rguenther@suse.de>
20773         * common.opt (feliminate-dwarf2-dups): Ignore.
20774         * doc/invoke.texi (feliminate-dwarf2-dups): Remove documentation.
20775         * dwarf2out.c (push_new_compile_unit, pop_compile_unit,
20776         same_die_p_wrap, compute_section_prefix,
20777         is_symbol_die, assign_symbol_names, break_out_includes): Remove.
20778         (comdat_symbol_id, comdat_symbol_number): Likewise.
20779         (cu_hash_table_entry, cu_hash_table_entry_hasher, cu_hash_type):
20780         Likewise.
20781         (check_duplicate_cu, record_comdat_symbol_number): Likewise.
20782         (output_die): Mark unreachable path unreachable.
20783         (dwarf2out_start_source_file): Do not create DW_TAG_GNU_BINCL.
20784         (dwarf2out_end_source_file): Do not create DW_TAG_GNU_EINCL.
20785         (dwarf2out_init): Remove code handling flag_eliminate_dwarf2_dups.
20786         (dwarf2out_early_finish): Likewise.
20788 2017-08-22  Aldy Hernandez  <aldyh@redhat.com>
20790         * wide-int.h (hwi_with_prec::hwi_with_prec): Sign extend.
20792 2017-08-22  Georg-Johann Lay  <avr@gjlay.de>
20794         PR target/81910
20795         * config/avr/avr.c (avr_handle_addr_attribute): Early return if
20796         not VAR_P. Filter attribute warnings with OPT_Wattributes.
20797         (avr_attribute_table) <io, io_low, address>: Initialize
20798         .decl_required with true.
20800 2017-08-21  Michael Meissner  <meissner@linux.vnet.ibm.com>
20802         * config/rs6000/rs6000-cpus.def (-mvsx-scalar-float): Delete
20803         undocumented debugging options.
20804         (-mvsx-scalar-double): Likewise.
20805         (-mallow-df-permute): Likewise.
20806         (-mvectorize-builtins): Likewise.
20807         * config/rs6000/rs6000.c (rs6000_init_hard_regno_mode_ok): Likewise.
20808         (rs6000_builtin_vectorized_function): Likewise.
20809         (rs6000_builtin_md_vectorized_function): Likewise.
20810         (rs6000_opt_vars): Likewise.
20812 2017-08-21  Uros Bizjak  <ubizjak@gmail.com>
20814         PR target/46091
20815         * config/i386/i386.md (*btsq_imm): Rename from *btsq.
20816         (*btrq_imm): Rename from *btrq.
20817         (*btcq_imm): Rename from *btcq.
20818         (btsc): New code attribute.
20819         (*<btsc><mode>): New insn pattern.
20820         (*btr<mode>): Ditto.
20821         (*<btsc><mode>_mask): New insn_and_split pattern.
20822         (*btr<mode>_mask): Ditto.
20824 2017-08-21  Richard Sandiford  <richard.sandiford@linaro.org>
20825             Alan Hayward  <alan.hayward@arm.com>
20826             David Sherwood  <david.sherwood@arm.com>
20828         * function.c (pad_below): Simplify padding calculation.
20830 2017-08-21  Richard Sandiford  <richard.sandiford@linaro.org>
20831             Alan Hayward  <alan.hayward@arm.com>
20832             David Sherwood  <david.sherwood@arm.com>
20834         * target.def (function_prologue): Remove frame size argument.
20835         (function_epilogue): Likewise.
20836         * doc/tm.texi: Regenerate.
20837         * final.c (final_start_function): Update call to function_prologue.
20838         (final_end_function): Update call to function_epilogue.
20839         (default_function_pro_epilogue): Remove frame size argument.
20840         * output.h (default_function_pro_epilogue): Likewise.
20841         * config/arm/arm.c (arm_output_function_epilogue): Likewise.
20842         (arm_output_function_prologue): Likewise.
20843         * config/frv/frv.c (frv_function_prologue): Likewise.
20844         (frv_function_epilogue): Likewise.
20845         * config/i386/i386.c (ix86_output_function_epilogue): Likewise.
20846         * config/ia64/ia64.c (ia64_output_function_prologue): Likewise.
20847         (ia64_output_function_epilogue): Likewise.
20848         * config/m32r/m32r.c (m32r_output_function_prologue): Likewise.
20849         (m32r_output_function_epilogue): Likewise.
20850         * config/microblaze/microblaze.c (microblaze_function_prologue)
20851         (microblaze_function_epilogue): Likewise.
20852         * config/mips/mips.c (mips_output_function_prologue): Likewise.
20853         (mips_output_function_epilogue): Likewise.
20854         * config/mmix/mmix.c (mmix_target_asm_function_prologue): Likewise.
20855         (mmix_target_asm_function_epilogue): Likewise.
20856         * config/msp430/msp430.c (msp430_start_function): Likewise.
20857         * config/nds32/nds32.c (nds32_asm_function_prologue): Likewise.
20858         (nds32_asm_function_epilogue): Likewise.
20859         * config/nios2/nios2.c (nios2_asm_function_prologue): Likewise.
20860         * config/pa/pa.c (pa_output_function_prologue): Likewise.
20861         (pa_output_function_epilogue): Likewise.
20862         * config/powerpcspe/powerpcspe.c (rs6000_output_function_prologue)
20863         (rs6000_output_function_epilogue): Likewise.
20864         * config/rl78/rl78.c (rl78_start_function): Likewise.
20865         * config/rs6000/rs6000.c (rs6000_output_function_prologue): Likewise.
20866         (rs6000_output_function_epilogue): Likewise.
20867         * config/rx/rx.c (rx_output_function_prologue): Likewise.
20868         * config/sh/sh.c (sh_output_function_epilogue): Likewise.
20869         * config/sparc/sparc.c (sparc_asm_function_prologue): Likewise.
20870         (sparc_asm_function_epilogue): Likewise.
20872 2017-08-21  Richard Sandiford  <richard.sandiford@linaro.org>
20874         * tree.h (type_has_mode_precision_p): New function.
20875         * convert.c (convert_to_integer_1): Use it.
20876         * expr.c (expand_expr_real_2): Likewise.
20877         (expand_expr_real_1): Likewise.
20878         * fold-const.c (fold_single_bit_test_into_sign_test): Likewise.
20879         * match.pd: Likewise.
20880         * tree-ssa-forwprop.c (simplify_rotate): Likewise.
20881         * tree-ssa-math-opts.c (convert_mult_to_fma): Likewise.
20882         * tree-tailcall.c (process_assignment): Likewise.
20883         * tree-vect-loop.c (vectorizable_reduction): Likewise.
20884         * tree-vect-patterns.c (vect_recog_vector_vector_shift_pattern)
20885         (vect_recog_mult_pattern, vect_recog_divmod_pattern): Likewise.
20886         * tree-vect-stmts.c (vectorizable_conversion): Likewise.
20887         (vectorizable_assignment): Likewise.
20888         (vectorizable_shift): Likewise.
20889         (vectorizable_operation): Likewise.
20890         * tree-vrp.c (register_edge_assert_for_2): Likewise.
20892 2017-08-21  Wilco Dijkstra  <wdijkstr@arm.com>
20894         * match.pd: Add pow (C, x) simplification.
20896 2017-08-21  Richard Biener  <rguenther@suse.de>
20898         PR tree-optimization/81900
20899         * tree-ssa-pre.c (compute_antic_aux): Properly compute changed
20900         for blocks with abnormal predecessors.
20901         (compute_antic): Do not set visited flag prematurely.
20903 2017-08-21  Georg-Johann Lay  <avr@gjlay.de>
20905         PR target/79883
20906         * config/avr/avr.c (avr_set_current_function): Typo in diagnostic.
20908 2017-08-21  Richard Sandiford  <richard.sandiford@linaro.org>
20910         * stor-layout.h (vector_type_mode): Move to...
20911         * tree.h (vector_type_mode): ...here.
20912         * stor-layout.c (vector_type_mode): Move to...
20913         * tree.c (vector_type_mode): ...here.  Include rtl.h and regs.h.
20915 2017-08-21  Richard Biener  <rguenther@suse.de>
20917         * debug.h (struct gcc_debug_hooks): Add die_ref_for_decl and
20918         register_external_die hooks.
20919         (debug_false_tree_charstarstar_uhwistar): Declare.
20920         (debug_nothing_tree_charstar_uhwi): Likewise.
20921         * debug.c (do_nothing_debug_hooks): Adjust.
20922         (debug_false_tree_charstarstar_uhwistar): New do nothing.
20923         (debug_nothing_tree_charstar_uhwi): Likewise.
20924         * dbxout.c (dbx_debug_hooks): Adjust.
20925         (xcoff_debug_hooks): Likewise.
20926         * sdbout.c (sdb_debug_hooks): Likewise.
20927         * vmsdbgout.c (vmsdbg_debug_hooks): Likewise.
20928         * dwarf2out.c (macinfo_label_base): New global.
20929         (dwarf2out_register_external_die): New function for the
20930         register_external_die hook.
20931         (dwarf2out_die_ref_for_decl): Likewise for die_ref_for_decl.
20932         (dwarf2_debug_hooks): Use them.
20933         (dwarf2_lineno_debug_hooks): Adjust.
20934         (struct die_struct): Add with_offset flag.
20935         (DEBUG_LTO_DWO_INFO_SECTION, DEBUG_LTO_INFO_SECTION,
20936         DEBUG_LTO_DWO_ABBREV_SECTION, DEBUG_LTO_ABBREV_SECTION,
20937         DEBUG_LTO_DWO_MACINFO_SECTION, DEBUG_LTO_MACINFO_SECTION,
20938         DEBUG_LTO_DWO_MACRO_SECTION, DEBUG_LTO_MACRO_SECTION,
20939         DEBUG_LTO_LINE_SECTION, DEBUG_LTO_DWO_STR_OFFSETS_SECTION,
20940         DEBUG_LTO_STR_DWO_SECTION, DEBUG_STR_LTO_SECTION): New macros
20941         defining section names for the early LTO debug variants.
20942         (reset_indirect_string): New helper.
20943         (add_AT_external_die_ref): Helper for dwarf2out_register_external_die.
20944         (print_dw_val): Add support for offsetted symbol references.
20945         (get_ultimate_context): Split out from is_cxx.
20946         (is_cxx): Use get_ultimate_context.
20947         (is_fortran): Add decl overload.
20948         (compute_comp_unit_symbol): Split out worker from
20949         compute_section_prefix.
20950         (compute_section_prefix): Call compute_comp_unit_symbol and
20951         set comdat_type_p here.
20952         (output_die): Skip DIE symbol output for the LTO added one.
20953         Handle DIE symbol references with offset.
20954         (output_comp_unit): Guard section name mangling properly.
20955         For LTO debug sections emit a symbol at the section beginning
20956         which we use to refer to its DIEs.
20957         (add_abstract_origin_attribute): For DIEs registered via
20958         dwarf2out_register_external_die directly refer to the early
20959         DIE rather than indirectly through the shadow one we created.
20960         Remove obsolete call to dwarf2out_abstract_function for
20961         non-function/block origins.
20962         (gen_array_type_die): When generating early LTO debug do
20963         not emit DW_AT_string_length.
20964         (gen_formal_parameter_die): Do not re-create DIEs for PARM_DECLs
20965         late when in LTO.  As suggested place a gcc_unreachable for
20966         the DECL_ABSTRACT_P case.
20967         (gen_subprogram_die): Avoid another specification DIE
20968         for early built declarations/definitions for the late LTO case.
20969         (gen_variable_die): Add type references for late duplicated VLA dies
20970         when in late LTO.
20971         (gen_inlined_subroutine_die): Do not call dwarf2out_abstract_function,
20972         we have the abstract instance already.
20973         (process_scope_var): Adjust decl DIE contexts in LTO which
20974         first puts them in limbo.
20975         (gen_decl_die): Do not generate type DIEs late apart from
20976         types for VLAs or for decls we do not yet have a DIE.  Do not
20977         call dwarf2out_abstract_function late.
20978         (dwarf2out_early_global_decl): Make sure to create DIEs
20979         for abstract instances of a decl first.
20980         (dwarf2out_late_global_decl): Adjust comment.
20981         (output_macinfo_op): With multiple macro sections use
20982         macinfo_label_base to distinguish labels.
20983         (output_macinfo): Likewise.  Update macinfo_label_base.
20984         Pass in the line info label.
20985         (note_variable_value_in_expr): When generating LTO resolve
20986         all variable values here by generating DIEs as needed.
20987         (init_sections_and_labels): Add early LTO debug flag parameter
20988         and generate different sections and names if set.  Add generation
20989         counter for the labels so we can have multiple of them.
20990         (reset_dies): Helper to allow DIEs to be output multiple times.
20991         (dwarf2out_finish): When outputting DIEs to the fat part of an
20992         LTO object first reset DIEs.
20993         (dwarf2out_early_finish): Output early DIEs when generating LTO.
20994         (modified_type_die): Check for decl_ultimate_origin being self
20995         before recursing.
20996         (gen_type_die_with_usage): Likewise.
20997         (gen_typedef_die): Allow decl_ultimate_origin being self.
20998         (set_decl_abstract_flags): Remove.
20999         (set_block_abstract_flags): Likewise.
21000         (dwarf2out_abstract_function): Treat the early generated DIEs
21001         as the abstract copy and only add DW_AT_inline and
21002         DW_AT_artificial here and call set_decl_origin_self.
21003         If the DIE has an abstract origin don't do anything.
21004         * tree.c (free_lang_data): Build a dummy TRANSLATION_UNIT_DECL
21005         if we have none yet (Go fails to build one, PR78628).
21006         (variably_modified_type_p): Prevent endless recursion for Ada
21007         cyclic pointer types.
21008         * lto-streamer-in.c: Include debug.h.
21009         (dref_queue): New global.
21010         (lto_read_tree_1): Stream in DIE references.
21011         (lto_input_tree): Register DIE references.
21012         (input_function): Stream DECL_DEBUG_ARGS.
21013         * lto-streamer-out.c: Include debug.h.
21014         (lto_write_tree_1): Output DIE references.
21015         (DFS::DFS_write_tree_body): Follow DECL_ABSTRACT_ORIGIN.
21016         Force a TRANSLATION_UNIT_DECL DECL_CONTEXT for file-scope decls.
21017         (output_function): Stream DECL_DEBUG_ARGS.
21018         * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
21019         Stream DECL_ABSTRACT_ORIGIN.
21020         * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
21021         (write_ts_decl_minimal_tree_pointers): Force a TRANSLATION_UNIT_DECL
21022         DECL_CONTEXT for file-scope decls.
21023         * lto-streamer.h (struct dref_entry): Declare.
21024         (dref_queue): Likewise.
21025         * cfgexpand.c (pass_expand::execute): Do not call the
21026         outlining_inline_function hook here.
21027         * lto-wrapper.c (debug_obj): New global.
21028         (tool_cleanup): Unlink it if required.
21029         (debug_objcopy): New function.
21030         (run_gcc): Handle early debug sections in the IL files by
21031         extracting them to separate files, partially linkin them and
21032         feeding the result back as result to the linker.
21033         * config/darwin.h (DEBUG_LTO_INFO_SECTION, DEBUG_LTO_ABBREV_SECTION,
21034         DEBUG_LTO_MACINFO_SECTION, DEBUG_LTO_LINE_SECTION,
21035         DEBUG_STR_LTO_SECTION, DEBUG_LTO_MACRO_SECTION): Put early debug
21036         sections into a separate segment.
21037         * config/darwin.c (darwin_asm_named_section): Handle __GNU_DWARF_LTO
21038         segments.
21039         (darwin_asm_dwarf_section): Likewise.
21040         (darwin_asm_output_dwarf_offset): Likewise.
21041         * config/i386/i386.c (make_resolver_func): Set DECL_IGNORED_P.
21043 2017-08-21  Richard Sandiford  <richard.sandiford@linaro.org>
21044             Alan Hayward  <alan.hayward@arm.com>
21045             David Sherwood  <david.sherwood@arm.com>
21047         * read-md.h (md_reader::record_potential_iterator_use): Replace
21048         pointer argument with an rtx and an index.
21049         * read-rtl.c (iterator_group::apply_iterator): Likewise.
21050         (apply_mode_iterator): Likewise.
21051         (apply_code_iterator): Likewise.
21052         (apply_int_iterator): Likewise.
21053         (apply_subst_iterator): Likewise.
21054         (record_iterator_use): Likewise.
21055         (record_attribute_use): Likewise.
21056         (md_reader::record_potential_iterator_use): Likewise.  Update calls
21057         to record_iterator_use and apply_iterator.
21058         (iterator_use): Replace ptr with x and index.
21059         (attribute_use): Likewise.
21060         (apply_attribute_uses): Update calls to apply_iterator.
21061         (apply_iterators): Likewise.  Update initialization of iterator_use.
21062         (rtx_reader::read_rtx_code): Update calls to record_iterator_use
21063         and record_potential_iterator_use.
21064         (rtx_reader::read_rtx_operand): Likewise.
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         * varasm.c (const_rtx_hash_1): Don't hash in the mode of a
21071         CONST_WIDE_INT.
21073 2017-08-21  Richard Biener  <rguenther@suse.de>
21075         PR middle-end/81884
21076         * tree-ssa-alias.c (stmt_kills_ref_p): Handle array accesses
21077         at struct end conservatively when comparing common bases.
21079 2017-08-21  Richard Biener  <rguenther@suse.de>
21081         * tree-ssa-loop-im.c (struct lim_aux_data): Add ref index member.
21082         (mem_ref_in_stmt): Remove.
21083         (determine_max_movement): Use ref index to get at the reference.
21084         (invariantness_dom_walker::before_dom_children): Deal with
21085         lim data already initialized.
21086         (gather_mem_refs_stmt): Initialize lim data and record ref index.
21088 2017-08-19  Uros Bizjak  <ubizjak@gmail.com>
21090         * config/i386/i386.h (OPTION_MASK_ISA_ROUND): Remove.
21091         (TARGET_ISA_ROUND): Ditto.
21092         (TARGET_ROUND): Ditto.
21093         * config/i386/i386.c: Substitute TARGET_ROUND with TARGET_SSE4_1.
21094         * config/i386/i386.md: Ditto.
21095         * config/i386/sse.md: Ditto.
21096         * config/i386/i386-builtin.def: Substitute OPTION_MASK_ISA_ROUND
21097         with OPTION_MASK_ISA_SSE4_1.
21099 2017-08-19  Uros Bizjak  <ubizjak@gmail.com>
21101         PR target/81894
21102         * doc/extend.texi (x86 Built-in Functions): Correct the name of
21103         __builtin_ia32_lzcnt_u16.
21105 2017-08-18  Peter Bergner  <bergner@vnet.ibm.com>
21107         PR target/80210
21108         * config/rs6000/rs6000.c (rs6000_activate_target_options): New function.
21109         (rs6000_set_current_function): Rewrite function to use it.
21111 2017-08-18  H.J. Lu  <hongjiu.lu@intel.com>
21113         PR c/53037
21114         * print-tree.c (print_node): Support DECL_WARN_IF_NOT_ALIGN
21115         and TYPE_WARN_IF_NOT_ALIGN.
21116         * stor-layout.c (do_type_align): Merge DECL_WARN_IF_NOT_ALIGN.
21117         (handle_warn_if_not_align): New.
21118         (place_union_field): Call handle_warn_if_not_align.
21119         (place_field): Call handle_warn_if_not_align.
21120         Copy TYPE_WARN_IF_NOT_ALIGN.
21121         (finish_builtin_struct): Copy TYPE_WARN_IF_NOT_ALIGN.
21122         (layout_type): Likewise.
21123         * tree-core.h (tree_type_common): Add warn_if_not_align.  Set
21124         spare to 18.
21125         (tree_decl_common): Add warn_if_not_align.
21126         * tree.c (build_range_type_1): Copy TYPE_WARN_IF_NOT_ALIGN.
21127         * tree.h (TYPE_WARN_IF_NOT_ALIGN): New.
21128         (SET_TYPE_WARN_IF_NOT_ALIGN): Likewise.
21129         (DECL_WARN_IF_NOT_ALIGN): Likewise.
21130         (SET_DECL_WARN_IF_NOT_ALIGN): Likewise.
21131         * doc/extend.texi: Document warn_if_not_aligned attribute.
21132         * doc/invoke.texi: Document -Wif-not-aligned and -Wpacked-not-aligned.
21134 2017-08-17  Martin Liska  <mliska@suse.cz>
21136         PR bootstrap/81864
21137         * tree-loop-distribution.c (ddrs_table): Change type to pointer type.
21138         (get_data_dependence): Use it as pointer type.
21139         (distribute_loop): Likewise.
21141 2017-08-17  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
21143         * config/rs6000/altivec.md (UNSPEC_VMRGOW_DIRECT): New constant.
21144         (p8_vmrgew_v4sf_direct): Generalize to p8_vmrgew_<mode>_direct.
21145         (p8_vmrgow_<mode>_direct): New define_insn.
21146         * config/rs6000/rs6000.c (altivec_expand_vec_perm_const): Properly
21147         handle endianness for vmrgew and vmrgow permute patterns.
21149 2017-08-17  Peter Bergner  <bergner@vnet.ibm.com>
21151         * config/rs6000/altivec.md (VParity): Remove TARGET_VSX_TIMODE.
21152         * config/rs6000/rs6000-cpus.def: Remove comment.
21153         (ISA_2_7_MASKS_SERVER): Delete OPTION_MASK_VSX_TIMODE;
21154         (POWERPC_MASKS): Likewise.
21155         * config/rs6000/rs6000.c (rs6000_hard_regno_mode_ok): Remove unneeded
21156         use of TARGET_VSX_TIMODE.
21157         (rs6000_setup_reg_addr_masks): Change TARGET_VSX_TIMODE to TARGET_VSX.
21158         (rs6000_init_hard_regno_mode_ok): Remove unneeded uses of
21159         TARGET_VSX_TIMODE.  Change use of TARGET_VSX_TIMODE to TARGET_VSX.
21160         (rs6000_option_override_internal): Remove dead code.
21161         (rs6000_legitimize_address): Change TARGET_VSX_TIMODE to TARGET_VSX.
21162         (rs6000_legitimize_reload_address): Likewise.
21163         (rs6000_legitimate_address_p): Likewise.
21164         (rs6000_opt_masks): Delete "vsx-timode".
21165         (rs6000_disable_incompatible_switches): Remove mention of -mvsx-timode
21166         from function comment.
21167         * config/rs6000/rs6000.h (MASK_VSX_TIMODE): Delete.
21168         * config/rs6000/rs6000.md (FMOVE128_GPR): Remove TARGET_VSX_TIMODE.
21169         (V16QI, V8HI, V4SI, V4SF, V2DI, V2DF, V1TI): Remove useless empty
21170         condition.
21171         * config/rs6000/rs6000.opt (mvsx-timode): Replace with stub.
21172         * config/rs6000/vector.md (VEC_IP): Remove TARGET_VSX_TIMODE.
21173         * config/rs6000/vsx.md (VSX_LE_128): Likewise.
21174         (VSX_TI): Likewise.
21175         (VSX_M): Likewise.
21176         (define_peephole2): Likewise.
21178 2017-08-17  Martin Sebor  <msebor@redhat.com>
21180         PR c/81859
21181         * pretty-print.c (pp_format): Use strnlen in %.*s to avoid reading
21182         past the end of an array.
21183         (test_pp_format): Add test cases.
21185 2017-08-17  Richard Sandiford  <richard.sandiford@linaro.org>
21187         * internal-fn.def (CLRSB, CLZ, CTZ, FFS, PARITY, POPCOUNT): Add
21188         missing ECF_NOTHROW flags.
21190 2017-08-17  Peter Bergner  <bergner@vnet.ibm.com>
21192         PR target/72804
21193         * config/rs6000/vsx.md (*vsx_le_permute_<mode>): Add support for
21194         operands residing in integer registers.
21195         (*vsx_le_perm_load_<mode>): Likewise.
21196         (*vsx_le_perm_store_<mode>): Likewise.
21197         (define_peephole2): Add peepholes to optimize the above.
21199 2017-08-17  Marek Polacek  <polacek@redhat.com>
21201         PR middle-end/81814
21202         * fold-const.c (operand_equal_for_comparison_p): Remove code that used
21203         to mimic what shorten_compare did.  Change the return type to bool.
21204         (fold_cond_expr_with_comparison): Update call to
21205         operand_equal_for_comparison_p.
21206         (fold_ternary_loc): Likewise.
21208 2017-08-17  Jackson Woodruff  <jackson.woodruff@arm.com>
21210         * aarch64-simd.md (mov<mode>): No longer force zero immediate into
21211         register.
21212         (*aarch64_simd_mov<mode>): Add new case for stp using zero immediate.
21214 2017-08-17  Richard Biener  <rguenther@suse.de>
21216         * tree-ssa-structalias.c (solve_graph): When propagating
21217         to successors update the graphs succ edges and avoid duplicate work.
21219 2017-08-17  Maxim Ostapenko  <m.ostapenko@samsung.com>
21221         PR target/81861
21222         * config/i386/i386.c (ix86_option_override_internal): Save target
21223         specific options after ix86_stack_protector_guard_reg was changed.
21225 2017-08-17  Richard Biener  <rguenther@suse.de>
21227         PR tree-optimization/81827
21228         * tree-ssa-structalias.c (struct variable_info): Add is_reg_var flag.
21229         (new_var_info): Initialize it conservatively.
21230         (get_call_vi): Mark register vars.
21231         (new_scalar_tmp_constraint_exp): Likewise.
21232         (handle_rhs_call): Likewise.
21233         (handle_const_call): Likewise.
21234         (create_function_info_for): Likewise.
21235         (solve_constraints): Sort varinfos to separate register from
21236         non-register vars to pack points-to solution bitmaps during
21237         iteration.
21239 2017-08-17  Marek Polacek  <polacek@redhat.com>
21241         * gimplify.c (gimplify_adjust_omp_clauses): Compare with 0 instead of 1.
21243 2017-08-17  Richard Biener  <rguenther@suse.de>
21245         * tree-vrp.c (vrp_int_const_binop): Do not set *overflow_p
21246         to true when overflow is undefined and we saturated the result.
21248 2017-08-17  Alan Modra  <amodra@gmail.com>
21250         PR target/80938
21251         * config/rs6000/rs6000.c (rs6000_savres_strategy): Revert 2017-08-09.
21252         Don't use store multiple if only one reg needs saving.
21253         (interesting_frame_related_regno): New function.
21254         (rs6000_frame_related): Don't emit frame info for regs that
21255         don't need saving.
21256         (rs6000_emit_epilogue): Likewise.
21258 2017-08-16  Nathan Sidwell  <nathan@acm.org>
21260         * tree-core.h (tree_type_non_common): Rename binfo to lang_1.
21261         * tree.h (TYPE_BINFO): Use type_non_common.maxval.
21262         (TYPE_LANG_SLOT_1): Use type_non_common.lang_1, for any type.
21263         * tree.c (free_lang_data_in_type): Use else-if chain.  Always
21264         clear TYPE_LANG_1.  Remove obsolete member-function stripping.
21265         (find_decls_types_r): Comment about TYPE_MAX_VALUES_RAW.
21266         (verify_type): Adjust for TYPE_BINFO move.
21267         * lto-streamer-out.c (DFS::DFS_write_tree_body): No need to
21268         process TYPE_BINFO directly.
21269         (hash_tree): Likewise.
21270         * tree-streamer-in.c (lto_input_ts_type_non_common_tree_pointers):
21271         Likewise.
21272         * tree-streamer-out.c (write_ts_type_non_common_tree_pointers):
21273         Likewise.
21275 2017-08-16  David Malcolm  <dmalcolm@redhat.com>
21277         * diagnostic-show-locus.c (colorizer::m_caret): Remove unused field.
21279 2017-08-16  Uros Bizjak  <ubizjak@gmail.com>
21281         PR target/46091
21282         * config/i386/i386.md (*anddi_1_btr): Change predicates of
21283         operand 0 and operand 1 to nomimmediate_operand. Add "m" constraint.
21284         Add ix86_binary_operator_ok to insn constraint.
21285         (*iordi_1_bts): Ditto.
21286         (*xordi_1_btc): Ditto.
21287         (*btsq): Change predicate of operand 0 to nonimmediate_operand.
21288         Update corresponding peephole2 pattern.
21289         (*btrq): Ditto.
21290         (*btcq): Ditto.
21292 2017-08-16  Bin Cheng  <bin.cheng@arm.com>
21294         PR tree-optimization/81832
21295         * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): Don't
21296         copy loop header which has IFN_LOOP_DIST_ALIAS call.
21298 2017-08-16  Marek Polacek  <polacek@redhat.com>
21300         PR middle/81695
21301         * fold-const.c (fold_indirect_ref_1): Restore original behavior
21302         regarding size_zero_node.
21304 2017-08-16  Martin Liska  <mliska@suse.cz>
21306         PR target/81753
21307         * config.gcc: Respect previously set extra_objs in case
21308         of darwin target.
21310 2017-08-16  Richard Sandiford  <richard.sandiford@linaro.org>
21312         PR tree-optimization/81835
21313         * tree-vect-loop.c (vect_is_simple_reduction): Simply checks for
21314         the phi SSA_NAME.  Check that the condition in a COND_EXPR does
21315         not depend on the phi.
21317 2017-08-16  Alan Modra  <amodra@gmail.com>
21319         * config/rs6000/rs6000.c (rs6000_init_hard_regno_mode_ok): Delete
21320         dead code.
21322 2017-08-16  Alan Modra  <amodra@gmail.com>
21324         * config/rs6000/rs6000.c (rs6000_reg_live_or_pic_offset_p): Merge..
21325         (save_reg_p): ..into this.  Update all callers.
21326         (first_reg_to_save): Simplify.
21328 2017-08-16  Alan Modra  <amodra@gmail.com>
21330         * config/rs6000/rs6000.c (rs6000_savres_strategy): Don't restore
21331         fixed regs.
21333 2017-08-15  Joseph Myers  <joseph@codesourcery.com>
21335         PR target/78460
21336         PR target/67712
21337         * config/sh/sh-mem.cc (sh_expand_cmpnstr): Only unroll for
21338         constant count if that count is less than 32.
21340 2017-08-15  Nathan Sidwell  <nathan@acm.org>
21342         * gcc.c (execute): Emit friendlier message if inferior is killed
21343         by an external cause.
21345 2017-08-15  Richard Biener  <rguenther@suse.de>
21347         PR tree-optimization/81790
21348         * tree-ssa-sccvn.c (vn_lookup_simplify_result): Handle both
21349         CONSTRUCTORs from simplifying and VN.
21351 2017-08-14  Martin Sebor  <msebor@redhat.com>
21353         * builtin-attrs.def: Add comments.
21355 2017-08-14  Martin Sebor  <msebor@redhat.com>
21357         PR c/81117
21358         * doc/extend.texi (attribute nonstring): Document new attribute.
21360 2017-08-14  Martin Sebor  <msebor@redhat.com>
21362         PR c/81117
21363         * tree-diagnostic.c (default_tree_printer): Handle %G.
21364         * gimple-pretty-print.h (percent_G_format): Declare new function.
21365         * gimple-pretty-print.c (percent_G_format): Define.
21366         * tree-pretty-print.c (percent_K_format): Add argument.
21368 2017-08-14  Martin Sebor  <msebor@redhat.com>
21370         PR translation/79998
21371         * gimple-ssa-sprintf.c (pass_sprintf_length::handle_gimple_call):
21372         Remove a stray space.
21374 2017-08-14  Uros Bizjak  <ubizjak@gmail.com>
21376         PR target/46091
21377         * config/i386/i386.md (*anddi_1_btr): New insn_and_split pattern.
21378         (*iordi_1_bts): Ditto.
21379         (*xordi_1_btc): Ditto.
21381 2017-08-14  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
21383         PR target/79845
21384         * config/rs6000/linux64.h (INVALID_64BIT): Use quoted strings.
21385         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
21386         Likewise.
21387         * config/rs6000/rs6000.c (rs6000_init_hard_regno_mode_ok): Use
21388         quoted strings, and make more translator-friendly.
21389         (darwin_rs6000_override_options): Likewise.
21390         (rs6000_option_override_internal): Likewise.
21391         (rs6000_return_in_memory): Fix overlong line.
21392         (init_cmulative_args): Use quoted strings, and make more
21393         translator-friendly.
21394         (rs6000_pass_by_reference): Fix overlong line.
21395         (def_builtin): Use quoted strings.
21396         (altivec_expand_predicate_builtin): Use quoted strings, and make
21397         more translator-friendly.
21398         (htm_expand_builtin): Use quoted strings.
21399         (cpu_expand_builtin): Use quoted strings, and make more
21400         translator-friendly.
21401         (altivec_expand_builtin): Likewise.
21402         (paired_expand_predicate_builtin): Likewise.
21403         (rs6000_invalid_builtin): Likewise.
21404         (builtin_function_type): Use quoted strings.
21405         (rs6000_expand_split_stack_prologue): Use quoted strings, and make
21406         more translator-friendly.
21407         (rs6000_trampoline_init): Likewise.
21408         (rs6000_handle_altivec_attribute): Likewise.
21409         (rs6000_inner_target_options): Use quoted strings.
21410         (rs6000_disable_incompatible_switches): Likewise.
21411         * config/rs6000/sysv4.h (SUBTARGET_OVERRIDE_OPTIONS): Use quoted
21412         strings, and make more translator-friendly.
21413         (SUBSUBTARGET_OVERRIDE_OPTIONS): Use quoted strings.
21415 2017-08-14  Bin Cheng  <bin.cheng@arm.com>
21417         PR tree-optimization/81799
21418         * tree-loop-distribution.c (version_loop_by_alias_check): Force
21419         cond_expr to simple gimple operand.
21421 2017-08-14  Wilco Dijkstra  <wdijkstr@arm.com>
21423         PR middle-end/46932
21424         * doc/sourcebuild.texi (autoincdec): Add autoincdec description.
21426 2017-08-14  Georg-Johann Lay  <avr@gjlay.de>
21428         PR target/81754
21429         PR target/81268
21430         * config/avr/avr.opt (mgas-isr-prologues): New Var avr_gasisr_prologues.
21431         * config/avr/avr.md (gasisr, *gasisr): Use it instead of
21432         TARGET_GASISR_PROLOGUES.
21433         * config/avr/avr.c (avr_option_override): Same.
21434         (avr_pass_pre_proep::execute): Same.
21436 2017-08-13  H.J. Lu  <hongjiu.lu@intel.com>
21438         PR target/81820
21439         * config/i386/i386.c (ix86_finalize_stack_frame_flags): Replace
21440         frame pointer with stack pointer - UNITS_PER_WORD in debug insns.
21442 2017-08-13  Uros Bizjak  <ubizjak@gmail.com>
21444         * config/i386/i386.md (*load_tp_<mode>): Redefine as
21445         define_insn_and_split.  Split to a memory load from 0 in
21446         DEFAULT_TLS_SEG_REG address space.  Merge with *load_tp_x32
21447         using PTR mode iterator.
21448         (*load_tp_x32_zext"): Redefine as define_insn_and_split.
21449         Split to a memory load from 0 in DEFAULT_TLS_SEG_REG address space.
21450         (*add_tp_<mode>): Redefine as define_insn_and_split.
21451         Split to an add with a memory load from 0 in DEFAULT_TLS_SEG_REG
21452         address space.  Merge with *add_tp_x32 using PTR mode iterator.
21453         (*add_tp_x32_zext"): Redefine as define_insn_and_split.
21454         Split to an add with a  memory load from 0 in
21455         DEFAULT_TLS_SEG_REG address space.
21457 2017-08-12  Andrew Pinski  <apinski@cavium.com>
21459         * config/aarch64/aarch64-option-extensions.def (rdma):
21460         Fix feature string to what Linux prints out in /proc/cpuinfo.
21462 2017-08-12  Pierre-Marie de Rodat  <derodat@adacore.com>
21464         PR ada/79542
21465         * dwarf2out.c (modified_type_die): For C typedef types that have
21466         an ultimate origin, process the ultimate origin instead of the
21467         input type.
21468         (gen_typedef_die): Assert that input DECLs have no ultimate
21469         origin.
21470         (gen_type_die_with_usage): For typedef variants that have an
21471         ultimate origin, just call gen_decl_die on the original DECL.
21472         (process_scope_var): Avoid creating DIEs for local typedefs and
21473         concrete static variables.
21475 2017-08-12  Alan Modra  <amodra@gmail.com>
21477         PR target/81170
21478         PR target/81295
21479         * config/rs6000/sysv4.h (STARTFILE_LINUX_SPEC): Upgrade to
21480         match gnu-user.h startfile.
21481         (ENDFILE_LINUX_SPEC): Similarly.
21483 2017-08-11  Thomas Schwinge  <thomas@codesourcery.com>
21485         PR lto/81430
21486         * config/nvptx/nvptx.c (nvptx_override_options_after_change):
21487         Remove function.
21488         (TARGET_OVERRIDE_OPTIONS_AFTER_CHANGE): Remove definition.
21490 2017-08-11  Tamar Christina  <tamar.christina@arm.com>
21491         * config/aarch64/aarch64.md (mov<mode>): Change.
21492         (*movhf_aarch64, *movsf_aarch64, *movdf_aarch64):
21493         aarch64_reg_or_fp_float into aarch64_reg_or_fp_zero.
21494         * config/aarch64/predicates.md (aarch64_reg_or_fp_float): Removed.
21496 2017-08-11  Eric Botcazou  <ebotcazou@adacore.com>
21498         * tree-sra.c (build_access_from_expr_1): Use more precise diagnostics
21499         for storage order barriers.
21501 2017-08-11  Martin Liska  <mliska@suse.cz>
21503         PR tree-opt/79987
21504         * tree-chkp.c (chkp_get_bounds_for_decl_addr): Do not instrument
21505         variables of void type.
21507 2017-08-11  Martin Liska  <mliska@suse.cz>
21509         * asan.c (asan_protect_global): Replace ASM_OUTPUT_DEF with
21510         TARGET_SUPPORTS_ALIASES.
21511         * cgraph.c (cgraph_node::create_same_body_alias): Likewise.
21512         * ipa-visibility.c (can_replace_by_local_alias): Likewise.
21513         (optimize_weakref): Likewise.
21514         * symtab.c (symtab_node::noninterposable_alias): Likewise.
21515         * varpool.c (varpool_node::create_extra_name_alias): Likewise.
21516         * defaults.h: Introduce TARGET_SUPPORTS_ALIASES.
21518 2017-08-11  Martin Liska  <mliska@suse.cz>
21520         PR ipa/81213
21521         * config/i386/i386.c (make_resolver_func): Do complete
21522         refactoring of the function.
21524 2017-08-10  Uros Bizjak  <ubizjak@gmail.com>
21526         PR target/81708
21527         * config/i386/i386.opt (mstack-protector-guard-symbol=): New option
21528         * config/i386/i386.c (ix86_stack_protect_guard): Use
21529         ix86_stack_protect_guard_symbol_str to generate varible declaration.
21530         * doc/invoke.texi (x86 Options): Document
21531         -mstack-protector-guard-symbol= option.
21533 2017-08-10  Uros Bizjak  <ubizjak@gmail.com>
21535         * config/i386/i386-protos.h (ix86_split_stack_guard): New prototype.
21536         * config/i386/i386.c (ix86_split_stack_guard): New function.
21537         (ix86_xpand_split_stack_prologue): Call ix86_split_stack_guard.
21538         (ix86_legitimate_address_p) <case UNSPEC_STACK_CHECK>: Remove.
21539         (i386_asm_output_addr_const_extra) <case UNSPEC_STACK_CHECK>: Ditto.
21540         (optput_pic_addr_const): Remove UNSPEC_STACK_CHECK handling.
21541         * config/i386/i386.md (unspec): Remove UNSPEC_STACK_CHECK.
21542         (split_stack_space_check): Call ix86_split_stack_guard.
21544 2017-08-10  Martin Sebor  <msebor@redhat.com>
21546         * print-tree.c (print_node): Print location using the established
21547         format %s:%i%i.
21548         Replace spaces with colons.
21549         (debug_raw, debug): Ditto.
21551 2017-08-10  Martin Sebor  <msebor@redhat.com>
21553         PR c++/81586
21554         * pretty-print.c (pp_format): Correct the handling of %s precision.
21556 2017-08-10  H.J. Lu  <hongjiu.lu@intel.com>
21558         PR target/81736
21559         * config/i386/i386.c (ix86_finalize_stack_realign_flags): Renamed
21560         to ...
21561         (ix86_finalize_stack_frame_flags): This.  Also clear
21562         frame_pointer_needed if -fno-omit-frame-pointer is used without
21563         stack access.
21564         (ix86_expand_prologue): Replace ix86_finalize_stack_realign_flags
21565         with ix86_finalize_stack_frame_flags.
21566         (ix86_expand_epilogue): Likewise.
21567         (ix86_expand_split_stack_prologue): Likewise.
21568         * doc/invoke.texi: Add a note for -fno-omit-frame-pointer.
21570 2017-08-10  Martin Liska  <mliska@suse.cz>
21572         PR c++/81355
21573         * c-attribs.c (handle_target_attribute):
21574         Report warning for an empty string argument of target attribute.
21576 2017-08-09  Jakub Jelinek  <jakub@redhat.com>
21578         PR c/81687
21579         * omp-low.c (omp_copy_decl): Don't remap FORCED_LABEL or DECL_NONLOCAL
21580         LABEL_DECLs.
21581         * tree-cfg.c (move_stmt_op): Don't adjust DECL_CONTEXT of FORCED_LABEL
21582         or DECL_NONLOCAL labels.
21583         (move_stmt_r) <case GIMPLE_LABEL>: Adjust DECL_CONTEXT of FORCED_LABEL
21584         or DECL_NONLOCAL labels here.
21586 2017-08-09  Will Schmidt  <will_schmidt@vnet.ibm.com>
21588         * config/rs6000/rs6000.c (rs6000_option_override_internal): Add blurb
21589         to indicate when early gimple folding has been disabled.
21590         (rs6000_gimple_fold_builtin): Add debug content.
21591         (rs6000_invalid_builtin): Fix whitespace.
21592         (rs6000_expand_builtin): Fix whitespace.
21593         * config/rs6000/rs6000.opt: Add option for -mfold-gimple.
21595 2017-08-09  Segher Boessenkool  <segher@kernel.crashing.org>
21597         PR target/80938
21598         * config/rs6000/rs6000.c (rs6000_savres_strategy): Don't use
21599         SAVE_MULTIPLE if not all the registers that saves, should be saved.
21601 2017-08-09  Jim Wilson  <jim.wilson@linaro.org>
21603         * config/aarch64/aarch64-cores.def (falkor): Use falkor pipeline.
21604         (qdf24xx): Likewise.
21605         * config/aarch64/aarch64.md: Include falkor.md.
21606         * config/aarch64/falkor.md: New.
21608 2017-08-09  Marek Polacek  <polacek@redhat.com>
21610         PR c/81233
21611         * diagnostic-core.h (emit_diagnostic_valist): Add declaration.
21612         * diagnostic.c (emit_diagnostic): Add a comment.
21613         (emit_diagnostic_valist): New function.
21615 2017-08-09  Marek Polacek  <polacek@redhat.com>
21617         PR c/81417
21618         * input.c (make_location): New overload.
21619         * input.h (make_location): Declare.
21621 2017-08-08  Alan Modra  <amodra@gmail.com>
21622             H.J. Lu  <hongjiu.lu@intel.com>
21624         PR driver/81523
21625         * gcc.c (NO_PIE_SPEC): Delete.
21626         (PIE_SPEC): Define as !no-pie/pie.  Move static|shared|r
21627         exclusion..
21628         (LINK_PIE_SPEC): ..to here.
21629         (LINK_COMMAND_SPEC): Support -no-pie.
21630         * config/gnu-user.h (GNU_USER_TARGET_STARTFILE_SPEC): Correct
21631         chain of crtbegin*.o selection, update for PIE_SPEC changes and
21632         format.
21633         (GNU_USER_TARGET_ENDFILE_SPEC): Similarly.
21634         * config/sol2.h (STARTFILE_CRTBEGIN_SPEC): Similarly.
21635         (ENDFILE_CRTEND_SPEC): Similarly.
21637 2017-08-08  Uros Bizjak  <ubizjak@gmail.com>
21639         PR target/81708
21640         * config/i386/i386.opt (mstack-protector-guard-reg=): New option
21641         (mstack-protector-guard-offset=): Ditto.
21642         * config/i386/i386.c (ix86_option_override): Handle
21643         -mstack-protector-guard-reg= and -mstack-protector-guard-offset=
21644         options.
21645         (ix86_stack_protect_guard): Use ix86_stack_protect_guard_reg and
21646         ix86_stack_protect_guard_offset variables.
21647         (TARGET_STACK_PROTECT_GUARD): Always define.
21648         * doc/invoke.texi (x86 Options): Document -mstack-protector-guard-reg=
21649         and -mstack-protector-guard-offset= options.
21651 2017-08-08  Bin Cheng  <bin.cheng@arm.com>
21653         * tree-ssa-loop-ivopts.c (relate_compare_use_with_all_cands): Handle
21654         boundary case for the last candidate.
21656 2017-08-08  Bin Cheng  <bin.cheng@arm.com>
21658         * doc/invoke.texi: Document -ftree-loop-distribution for O3.
21659         * opts.c (default_options_table): Add OPT_ftree_loop_distribution.
21661 2017-08-08  Tamar Christina  <tamar.christina@arm.com>
21663         PR middle-end/19706
21664         * config/aarch64/aarch64.md (xorsign<mode>3): New optabs.
21665         * config/aarch64/aarch64-builtins.c
21666         (aarch64_builtin_vectorized_function): Added CASE_CFN_XORSIGN.
21667         * config/aarch64/aarch64-simd-builtins.def: Added xorsign BINOP.
21668         * config/aarch64/aarch64-simd.md: Added xorsign<mode>3
21670 2017-08-08  Tamar Christina  <tamar.christina@arm.com>
21671             Andrew Pinski <pinskia@gmail.com>
21673         PR middle-end/19706
21674         * internal-fn.def (XORSIGN): New.
21675         * optabs.def (xorsign_optab): New.
21676         * tree-ssa-math-opts.c (is_copysign_call_with_1): New.
21677         (convert_expand_mult_copysign): New.
21678         (pass_optimize_widening_mul::execute): Call
21679         convert_expand_mult_copysign.
21681 2017-08-08  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
21683         PR tree-optimization/81354
21684         * gimple-ssa-strength-reduction.c (create_add_on_incoming_edge):
21685         Insert on edges rather than explicitly creating landing pads.
21686         (analyze_candidates_and_replace): Commit edge inserts.
21688 2017-08-08  Richard Biener  <rguenther@suse.de>
21690         PR middle-end/81719
21691         * tree-ssa-loop-niter.c: Include tree-dfa.h.
21692         (expand_simple_operations): Also look through ADDR_EXPRs with
21693         MEM_REF bases treating them as POINTER_PLUS_EXPR.
21695 2017-08-08  Richard Biener  <rguenther@suse.de>
21697         PR tree-optimization/81723
21698         * tree-vect-slp.c (struct bst_traits): New hash traits.
21699         (bst_fail): New global.
21700         (vect_build_slp_tree_2): New worker, split out from ...
21701         (vect_build_slp_tree): ... this now wrapping it with using
21702         bst_fail set to cache SLP tree build fails.  Properly handle
21703         max_tree_size.
21704         (vect_analyze_slp_instance): Allocate and free bst_fail.
21706 2017-08-08  Martin Liska  <mliska@suse.cz>
21708         PR tree-opt/81696
21709         * ipa-icf-gimple.c (func_checker::compare_cst_or_decl): Consider
21710         LABEL_DECLs that can be from a different function.
21712 2017-08-08  Bin Cheng  <bin.cheng@arm.com>
21714         PR tree-optimization/81744
21715         * tree-predcom.c (prepare_finalizers_chain): Deep copy expr of
21716         loop's number of iterations.
21718 2017-08-08  Martin Liska  <mliska@suse.cz>
21720         * asan.c: Include header files.
21721         * attribs.c (build_decl_attribute_variant): New function moved
21722         from tree.[ch].
21723         (build_type_attribute_qual_variant): Likewise.
21724         (cmp_attrib_identifiers): Likewise.
21725         (simple_cst_list_equal): Likewise.
21726         (omp_declare_simd_clauses_equal): Likewise.
21727         (attribute_value_equal): Likewise.
21728         (comp_type_attributes): Likewise.
21729         (build_type_attribute_variant): Likewise.
21730         (lookup_ident_attribute): Likewise.
21731         (remove_attribute): Likewise.
21732         (merge_attributes): Likewise.
21733         (merge_type_attributes): Likewise.
21734         (merge_decl_attributes): Likewise.
21735         (merge_dllimport_decl_attributes): Likewise.
21736         (handle_dll_attribute): Likewise.
21737         (attribute_list_equal): Likewise.
21738         (attribute_list_contained): Likewise.
21739         * attribs.h (lookup_attribute): New function moved from tree.[ch].
21740         (lookup_attribute_by_prefix): Likewise.
21741         * bb-reorder.c: Include header files.
21742         * builtins.c: Likewise.
21743         * calls.c: Likewise.
21744         * cfgexpand.c: Likewise.
21745         * cgraph.c: Likewise.
21746         * cgraphunit.c: Likewise.
21747         * convert.c: Likewise.
21748         * dwarf2out.c: Likewise.
21749         * final.c: Likewise.
21750         * fold-const.c: Likewise.
21751         * function.c: Likewise.
21752         * gimple-expr.c: Likewise.
21753         * gimple-fold.c: Likewise.
21754         * gimple-pretty-print.c: Likewise.
21755         * gimple.c: Likewise.
21756         * gimplify.c: Likewise.
21757         * hsa-common.c: Likewise.
21758         * hsa-gen.c: Likewise.
21759         * internal-fn.c: Likewise.
21760         * ipa-chkp.c: Likewise.
21761         * ipa-cp.c: Likewise.
21762         * ipa-devirt.c: Likewise.
21763         * ipa-fnsummary.c: Likewise.
21764         * ipa-inline.c: Likewise.
21765         * ipa-visibility.c: Likewise.
21766         * ipa.c: Likewise.
21767         * lto-cgraph.c: Likewise.
21768         * omp-expand.c: Likewise.
21769         * omp-general.c: Likewise.
21770         * omp-low.c: Likewise.
21771         * omp-offload.c: Likewise.
21772         * omp-simd-clone.c: Likewise.
21773         * opts-global.c: Likewise.
21774         * passes.c: Likewise.
21775         * predict.c: Likewise.
21776         * sancov.c: Likewise.
21777         * sanopt.c: Likewise.
21778         * symtab.c: Likewise.
21779         * toplev.c: Likewise.
21780         * trans-mem.c: Likewise.
21781         * tree-chkp.c: Likewise.
21782         * tree-eh.c: Likewise.
21783         * tree-into-ssa.c: Likewise.
21784         * tree-object-size.c: Likewise.
21785         * tree-parloops.c: Likewise.
21786         * tree-profile.c: Likewise.
21787         * tree-ssa-ccp.c: Likewise.
21788         * tree-ssa-live.c: Likewise.
21789         * tree-ssa-loop.c: Likewise.
21790         * tree-ssa-sccvn.c: Likewise.
21791         * tree-ssa-structalias.c: Likewise.
21792         * tree-ssa.c: Likewise.
21793         * tree-streamer-in.c: Likewise.
21794         * tree-vectorizer.c: Likewise.
21795         * tree-vrp.c: Likewise.
21796         * tsan.c: Likewise.
21797         * ubsan.c: Likewise.
21798         * varasm.c: Likewise.
21799         * varpool.c: Likewise.
21800         * tree.c: Remove functions moved to attribs.[ch].
21801         * tree.h: Likewise.
21802         * config/aarch64/aarch64.c: Add attrs.h header file.
21803         * config/alpha/alpha.c: Likewise.
21804         * config/arc/arc.c: Likewise.
21805         * config/arm/arm.c: Likewise.
21806         * config/avr/avr.c: Likewise.
21807         * config/bfin/bfin.c: Likewise.
21808         * config/c6x/c6x.c: Likewise.
21809         * config/cr16/cr16.c: Likewise.
21810         * config/cris/cris.c: Likewise.
21811         * config/darwin.c: Likewise.
21812         * config/epiphany/epiphany.c: Likewise.
21813         * config/fr30/fr30.c: Likewise.
21814         * config/frv/frv.c: Likewise.
21815         * config/ft32/ft32.c: Likewise.
21816         * config/h8300/h8300.c: Likewise.
21817         * config/i386/winnt.c: Likewise.
21818         * config/ia64/ia64.c: Likewise.
21819         * config/iq2000/iq2000.c: Likewise.
21820         * config/lm32/lm32.c: Likewise.
21821         * config/m32c/m32c.c: Likewise.
21822         * config/m32r/m32r.c: Likewise.
21823         * config/m68k/m68k.c: Likewise.
21824         * config/mcore/mcore.c: Likewise.
21825         * config/microblaze/microblaze.c: Likewise.
21826         * config/mips/mips.c: Likewise.
21827         * config/mmix/mmix.c: Likewise.
21828         * config/mn10300/mn10300.c: Likewise.
21829         * config/moxie/moxie.c: Likewise.
21830         * config/msp430/msp430.c: Likewise.
21831         * config/nds32/nds32-isr.c: Likewise.
21832         * config/nds32/nds32.c: Likewise.
21833         * config/nios2/nios2.c: Likewise.
21834         * config/nvptx/nvptx.c: Likewise.
21835         * config/pa/pa.c: Likewise.
21836         * config/pdp11/pdp11.c: Likewise.
21837         * config/powerpcspe/powerpcspe.c: Likewise.
21838         * config/riscv/riscv.c: Likewise.
21839         * config/rl78/rl78.c: Likewise.
21840         * config/rx/rx.c: Likewise.
21841         * config/s390/s390.c: Likewise.
21842         * config/sh/sh.c: Likewise.
21843         * config/sol2.c: Likewise.
21844         * config/sparc/sparc.c: Likewise.
21845         * config/spu/spu.c: Likewise.
21846         * config/stormy16/stormy16.c: Likewise.
21847         * config/tilegx/tilegx.c: Likewise.
21848         * config/tilepro/tilepro.c: Likewise.
21849         * config/v850/v850.c: Likewise.
21850         * config/vax/vax.c: Likewise.
21851         * config/visium/visium.c: Likewise.
21852         * config/xtensa/xtensa.c: Likewise.
21854 2017-08-07  Michael Meissner  <meissner@linux.vnet.ibm.com>
21856         PR target/81593
21857         * config/rs6000/vsx.md (vsx_concat_<mode>, VSX_D): Cleanup
21858         constraints since the -mupper-regs-* switches have been
21859         eliminated.
21860         (vsx_concat_<mode>_1): New combiner insns to recognize inserting
21861         into a vector from a double word element that was extracted from
21862         another vector, and eliminate extra XXPERMDI instructions.
21863         (vsx_concat_<mode>_2): Likewise.
21864         (vsx_concat_<mode>_3): Likewise.
21865         (vsx_set_<mode>, VSX_D): Rewrite vector set in terms of vector
21866         concat to allow optimizing inserts from previous extracts.
21868 2017-08-07  Uros Bizjak  <ubizjak@gmail.com>
21870         * config/i386/i386.c (ix86_stack_protect_guard): Generate
21871         memory reference to a SSP offset in TLS address space.
21872         (ix86_print_operand) <case '@'>: Remove.
21873         (ix86_print_operand_punct_valid_p): Remove '@' code.
21874         * config/i386/i386.md (unspec): Remove UNSPEC_SP_TLS_SET and
21875         UNSPEC_SP_TLS_TEST.
21876         (stack_tls_protect_set_<mode>): Remove.
21877         (stack_protect_set): Do not call gen_stack_tls_protect_set_<mode>.
21878         (stack_tls_protect_test_<mode>): Remove.
21879         (stack_protect_test): Do not call gen_stack_tls_protect_test_<mode>.
21881 2017-08-07  Olivier Hainque  <hainque@adacore.com>
21883         PR target/81755
21884         * config/vxworksae.h (VXWORKS_HAVE_TLS): Define.
21886 2017-08-07  Douglas Rupp  <rupp@adacore.com>
21888         * Makefile.in (install-mkheaders): Fix typo, where the multi_dir
21889         variable was referenced as multidir in command.
21891 2017-08-07  Jakub Jelinek  <jakub@redhat.com>
21893         PR c/69389
21894         * gimplify.c (goa_stabilize_expr): Handle BIT_INSERT_EXPR and
21895         BIT_FIELD_REF.
21897 2017-08-07  Martin Liska  <mliska@suse.cz>
21899         * config/m32c/m32c.c: Add include of stringpool.h and attribs.h.
21900         * config/rl78/rl78.c: Add include of attribs.h.
21901         * config/sh/sh.c: Likewise.
21902         * config/v850/v850.c: Likewise.
21904 2017-08-07  Tom de Vries  <tom@codesourcery.com>
21906         PR middle-end/78266
21907         * omp-expand.c (expand_oacc_for): Ensure diff_type is large enough.
21909 2017-08-07  Martin Liska  <mliska@suse.cz>
21911         * config/mips/mips.c: Include attribs.h.
21913 2017-08-07  Thomas Koenig  <tkoenig@gcc.gnu.org>
21915         PR fortran/68829
21916         * doc/invoke.texi: Document change in behvaior for -Ofast for
21917         Fortran.
21919 2017-08-07  Wilco Dijkstra  <wdijkstr@arm.com>
21921         * config/aarch64/aarch64.c (aarch64_pushwb_single_reg):
21922         Use gen_frame_mem.
21923         (aarch64_pop_regs): Likewise.
21924         (aarch64_gen_load_pair): Likewise.
21925         (aarch64_save_callee_saves): Likewise.
21926         (aarch64_restore_callee_saves): Likewise.
21928 2017-08-07  H.J. Lu  <hongjiu.lu@intel.com>
21930         * config/i386/i386.c: Revert the last change.
21932 2017-08-07  H.J. Lu  <hongjiu.lu@intel.com>
21934         PR target/81736
21935         * config/i386/i386.c (ix86_finalize_stack_realign_flags): Renamed
21936         to ...
21937         (ix86_finalize_stack_frame_flags): This.  Also clear
21938         frame_pointer_needed if -fno-omit-frame-pointer is used without
21939         stack access.
21940         (ix86_expand_prologue): Replace ix86_finalize_stack_realign_flags
21941         with ix86_finalize_stack_frame_flags.
21942         (ix86_expand_epilogue): Likewise.
21943         (ix86_expand_split_stack_prologue): Likewise.
21945 2017-08-07  H.J. Lu  <hongjiu.lu@intel.com>
21947         PR target/81743
21948         * config/i386/i386.c (get_builtin_code_for_version): Set priority
21949         to P_AES for Westmere.
21951 2017-08-07  Jonathan Yong  <10walls@gmail.com>
21953         * config/i386/mingw.opt (fset-stack-executable): Removed.
21954         * config/i386/cygming.opt (fset-stack-executable): Moved
21955         from mingw.opt.
21956         * config/i386/cygwin.h: Define CHECK_EXECUTE_STACK_ENABLED.
21958 2017-08-07  Segher Boessenkool  <segher@kernel.crashing.org>
21960         * print-rtl.c (print_exp): Print NOT as "~" instead of as "!".
21962 2017-08-07  Marek Polacek  <polacek@redhat.com>
21964         PR middle-end/81737
21965         * fold-const.c (fold_indirect_ref_1): Check type_domain.
21967 2017-08-07  Martin Liska  <mliska@suse.cz>
21969         * attribs.h (canonicalize_attr_name): New function.
21970         (cmp_attribs): Move from c-format.c and adjusted.
21971         (is_attribute_p): Moved from tree.h.
21972         * tree-inline.c: Add new includes.
21973         * tree.c (cmp_attrib_identifiers): Use cmp_attribs.
21974         (private_is_attribute_p): Remove.
21975         (private_lookup_attribute): Likewise.
21976         (private_lookup_attribute_by_prefix): Simplify.
21977         (remove_attribute): Use is_attribute_p.
21978         * tree.h: Remove removed declarations.
21980 2017-08-07  Jakub Jelinek  <jakub@redhat.com>
21982         PR middle-end/81698
21983         * stmt.c (emit_case_dispatch_table): Add DEFAULT_EDGE argument,
21984         instead of computing it in the function.  Formatting fix.
21985         (expand_case): Don't rely on default_edge being the first edge,
21986         clear it if removing it, pass default_edge to
21987         emit_case_dispatch_table.
21988         (expand_sjlj_dispatch_table): Pass NULL as DEFAULT_EDGE, formatting
21989         fix.
21991 2017-08-06  Uros Bizjak  <ubizjak@gmail.com>
21993         * config/alpha/alpha.c (alpha_reorg): If trap is the last active
21994         insn in the function, emit NOP after the insn.
21996 2017-08-06  Tom de Vries  <tom@codesourcery.com>
21998         * omp-expand.c (expand_oacc_for): Add missing edge probability for tile
21999         and element loops.
22001 2017-08-06  Tom de Vries  <tom@codesourcery.com>
22003         * omp-expand.c (expand_oacc_for): Add missing edge probability for chunk
22004         loop.
22006 2017-08-04  Yury Gribov  <tetra2005@gmail.com>
22008         PR tree-optimization/57371
22009         * match.pd: New pattern.
22011 2017-08-04  Marek Polacek  <polacek@redhat.com>
22013         PR middle-end/81695
22014         * fold-const.c (fold_indirect_ref_1): For ((int *)&a + 4 -> a[1],
22015         perform the computation in offset_int.
22017 2017-08-04  Richard Sandiford  <richard.sandiford@linaro.org>
22019         PR tree-optimization/81136
22020         * tree-vectorizer.h: Include tree-hash-traits.h.
22021         (vec_base_alignments): New typedef.
22022         (vec_info): Add a base_alignments field.
22023         (vect_record_base_alignments): Declare.
22024         * tree-data-ref.h (data_reference): Add an is_conditional_in_stmt
22025         field.
22026         (DR_IS_CONDITIONAL_IN_STMT): New macro.
22027         (create_data_ref): Add an is_conditional_in_stmt argument.
22028         * tree-data-ref.c (create_data_ref): Likewise.  Use it to initialize
22029         the is_conditional_in_stmt field.
22030         (data_ref_loc): Add an is_conditional_in_stmt field.
22031         (get_references_in_stmt): Set the is_conditional_in_stmt field.
22032         (find_data_references_in_stmt): Update call to create_data_ref.
22033         (graphite_find_data_references_in_stmt): Likewise.
22034         * tree-ssa-loop-prefetch.c (determine_loop_nest_reuse): Likewise.
22035         * tree-vect-data-refs.c (vect_analyze_data_refs): Likewise.
22036         (vect_record_base_alignment): New function.
22037         (vect_record_base_alignments): Likewise.
22038         (vect_compute_data_ref_alignment): Adjust base_addr and aligned_to
22039         for nested statements even if we fail to compute a misalignment.
22040         Use pooled base alignments for unconditional references.
22041         (vect_find_same_alignment_drs): Compare base addresses instead
22042         of base objects.
22043         (vect_analyze_data_refs_alignment): Call vect_record_base_alignments.
22044         * tree-vect-slp.c (vect_slp_analyze_bb_1): Likewise.
22046 2017-08-04  Richard Sandiford  <richard.sandiford@linaro.org>
22048         * tree-vectorizer.h (vec_info): Add a constructor and destructor.
22049         Add an explicit name for the enum.  Use auto_vec for slp_instances
22050         and grouped_stores.
22051         (_loop_vec_info): Add a constructor and destructor.  Use auto_vec
22052         for all vectors.
22053         (_bb_vec_info): Add a constructor and destructor.
22054         (vinfo_for_stmt): Return NULL for uids of -1 as well.
22055         (destroy_loop_vec_info): Delete.
22056         (vect_destroy_datarefs): Likewise.
22057         * tree-vectorizer.c (vect_destroy_datarefs): Delete.
22058         (vec_info::vec_info): New function.
22059         (vec_info::~vec_info): Likewise.
22060         (vectorize_loops): Use delete instead of destroy_loop_vec_info.
22061         * tree-parloops.c (gather_scalar_reductions): Use delete instead of
22062         destroy_loop_vec_info.
22063         * tree-vect-loop.c (new_loop_vec_info): Replace with...
22064         (_loop_vec_info::_loop_vec_info): ...this.
22065         (destroy_loop_vec_info): Replace with...
22066         (_loop_vec_info::~_loop_vec_info): ...this.  Unconditionally delete
22067         the stmt_vec_infos.  Leave handling of vec_info information to its
22068         destructor.  Remove explicit vector releases.
22069         (vect_analyze_loop_form): Use new instead of new_loop_vec_info.
22070         (vect_analyze_loop): Use delete instead of destroy_loop_vec_info.
22071         * tree-vect-slp.c (new_bb_vec_info): Replace with...
22072         (_bb_vec_info::_bb_vec_info): ...this.  Don't reserve space in
22073         BB_VINFO_GROUPED_STORES or BB_VINFO_SLP_INSTANCES.
22074         (destroy_bb_vec_info): Replace with...
22075         (_bb_vec_info::~_bb_vec_info): ...this.  Leave handling of vec_info
22076         information to its destructor.
22077         (vect_slp_analyze_bb_1): Use new and delete instead of
22078         new_bb_vec_info and destroy_bb_vec_info.
22079         (vect_slp_bb): Replace 2 calls to destroy_bb_vec_info with a
22080         single delete.
22082 2017-08-04  Richard Sandiford  <richard.sandiford@linaro.org>
22084         * tree-data-ref.h (subscript): Add access_fn field.
22085         (data_dependence_relation): Add could_be_independent_p.
22086         (SUB_ACCESS_FN, DDR_COULD_BE_INDEPENDENT_P): New macros.
22087         (same_access_functions): Move to tree-data-ref.c.
22088         * tree-data-ref.c (ref_contains_union_access_p): New function.
22089         (access_fn_component_p): Likewise.
22090         (access_fn_components_comparable_p): Likewise.
22091         (dr_analyze_indices): Add a reference to access_fn_component_p.
22092         (dump_data_dependence_relation): Use SUB_ACCESS_FN instead of
22093         DR_ACCESS_FN.
22094         (constant_access_functions): Likewise.
22095         (add_other_self_distances): Likewise.
22096         (same_access_functions): Likewise.  (Moved from tree-data-ref.h.)
22097         (initialize_data_dependence_relation): Use XCNEW and remove
22098         explicit zeroing of DDR_REVERSED_P.  Look for a subsequence
22099         of access functions that have the same type.  Allow the
22100         subsequence to end with different bases in some circumstances.
22101         Record the chosen access functions in SUB_ACCESS_FN.
22102         (build_classic_dist_vector_1): Replace ddr_a and ddr_b with
22103         a_index and b_index.  Use SUB_ACCESS_FN instead of DR_ACCESS_FN.
22104         (subscript_dependence_tester_1): Likewise dra and drb.
22105         (build_classic_dist_vector): Update calls accordingly.
22106         (subscript_dependence_tester): Likewise.
22107         * tree-ssa-loop-prefetch.c (determine_loop_nest_reuse): Check
22108         DDR_COULD_BE_INDEPENDENT_P.
22109         * tree-vectorizer.h (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Test
22110         comp_alias_ddrs instead of may_alias_ddrs.
22111         * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr):
22112         New function.
22113         (vect_analyze_data_ref_dependence): Use it if
22114         DDR_COULD_BE_INDEPENDENT_P, but fall back to using the recorded
22115         distance vectors if that fails.
22116         (dependence_distance_ge_vf): New function.
22117         (vect_prune_runtime_alias_test_list): Use it.  Don't clear
22118         LOOP_VINFO_MAY_ALIAS_DDRS.
22120 2017-08-04  Richard Biener  <rguenther@suse.de>
22122         PR middle-end/81705
22123         * fold-const.c (fold_binary_loc): Properly restrict
22124         minus_var0 && minus_var1 case when associating undefined overflow
22125         entities.
22127 2017-08-04  Tom de Vries  <tom@codesourcery.com>
22129         * omp-simd-clone.c (simd_clone_adjust): Add missing edge probability.
22131 2017-08-03  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
22133         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
22134         Don't start diagnostic messages with a capital letter.
22135         * config/rs6000/rs6000.c (rs6000_option_override_internal):
22136         Likewise.
22137         (rs6000_invalid_builtin): Likewise.
22138         (rs6000_trampoline_init): Likewise.
22140 2017-08-03  Jakub Jelinek  <jakub@redhat.com>
22142         PR target/81621
22143         * bb-reorder.c (pass_partition_blocks::execute): Return TODO_df_finish
22144         after setting changeable df flags.
22146 2017-08-03  Richard Biener  <rguenther@suse.de>
22148         * tree-ssa-reassoc.c (should_break_up_subtract): Also break
22149         up if the use is in USE - X.
22151 2017-08-03  Alexander Monakov  <amonakov@ispras.ru>
22153         * toplev.c (dumpfile.h): New include.
22154         (internal_error_reentered): New static function.  Use it...
22155         (internal_error_function): ...here to handle reentered internal_error.
22157 2017-08-03  Richard Biener  <rguenther@suse.de>
22159         PR middle-end/81148
22160         * fold-const.c (split_tree): Add minus_var and minus_con
22161         arguments, remove unused loc arg.  Never generate NEGATE_EXPRs
22162         here but always use minus_*.
22163         (associate_trees): Assert we never associate with MINUS_EXPR
22164         and NULL first operand.  Do not recurse for PLUS_EXPR operands
22165         when associating as MINUS_EXPR either.
22166         (fold_binary_loc): Track minus_var and minus_con.
22168 2017-08-03  Tom de Vries  <tom@codesourcery.com>
22170         PR lto/81430
22171         * tree-streamer-in.c (lto_input_ts_function_decl_tree_pointers): If
22172         ACCEL_COMPILER, apply finish_options on
22173         DECL_FUNCTION_SPECIFIC_OPTIMIZATION.
22175 2017-08-03  Tom de Vries  <tom@codesourcery.com>
22177         PR target/81662
22178         * config/nvptx/nvptx.c (nvptx_option_override): Emit sorry if
22179         function_entry_patch_area_size > 0.
22181 2017-08-03  Jakub Jelinek  <jakub@redhat.com>
22183         PR driver/81650
22184         * calls.c (alloc_max_size): Use HOST_WIDE_INT_UC (10??)
22185         instead of 10??LU, perform unit multiplication in wide_int,
22186         don't change alloc_object_size_limit if the limit is larger
22187         than SSIZE_MAX.
22189         PR tree-optimization/81655
22190         PR tree-optimization/81588
22191         * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Handle also
22192         the case when ranges[i].low and high are 1 for unsigned type with
22193         precision 1.
22195         PR middle-end/81052
22196         * omp-low.c (diagnose_sb_0): Handle flag_openmp_simd like flag_openmp.
22197         (pass_diagnose_omp_blocks::gate): Enable also for flag_openmp_simd.
22199 2017-08-03  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
22201         * tree-vrp.h: Add include guard.
22203 2017-08-02  Uros Bizjak  <ubizjak@gmail.com>
22205         PR target/81644
22206         * config/i386/i386.md (unspecv): Add UNSPECV_UD2.
22207         (ud2): New insn pattern.
22208         * config/i386/i386.c (ix86_expand_epilogue):
22209         For naked functions, generate ud2 instead of trap insn.
22211 2017-08-02  Marek Polacek  <polacek@redhat.com>
22213         PR other/81667
22214         * alloc-pool.h (base_pool_allocator): Initialize m_elt_size.
22216 2017-08-02  Tom de Vries  <tom@codesourcery.com>
22217             Cesar Philippidis  <cesar@codesourcery.com>
22219         * config/nvptx/nvptx.c (nvptx_lockless_update, nvptx_lockfull_update):
22220         Add missing edge probabilities.
22222 2017-08-02  Tamar Christina  <tamar.christina@arm.com>
22224         * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
22225         Correct endianness.
22227 2017-08-02  Jakub Jelinek  <jakub@redhat.com>
22229         PR middle-end/79499
22230         * function.c (thread_prologue_and_epilogue_insns): Determine blocks
22231         for find_many_sub_basic_blocks bitmap by looking up BLOCK_FOR_INSN
22232         of first NONDEBUG_INSN_P in each of the split_prologue_seq and
22233         prologue_seq sequences - if any.
22235 2017-08-02  Richard Biener  <rguenther@suse.de>
22237         * tree-vect-stmts.c (vectorizable_store): Perform vector extracts
22238         via vectors if supported, integer extracts via punning if supported
22239         or otherwise vector extracts.
22241 2017-08-02  Richard Biener  <rguenther@suse.de>
22243         * tree-ssa-pre.c (bitmap_insert_into_set_1): Remove and inline
22244         into ...
22245         (bitmap_insert_into_set): ... this.
22247 2017-08-02  Richard Biener  <rguenther@suse.de>
22249         PR tree-optimization/81633
22250         Revert
22251         2015-08-17  Alan Hayward  <alan.hayward@arm.com>
22253         PR tree-optimization/71752
22254         * tree-vect-slp.c (vect_get_slp_defs): Handle null operands.
22256 2017-08-01  Daniel Santos  <daniel.santos@pobox.com>
22258         * config/i386/i386.h (ix86_frame::outlined_save_offset): Remove field.
22259         (machine_function::call_ms2sysv_pad_out): Remove field.
22260         * config/i386/i386.c (xlogue_layout::get_stack_space_used): Modify.
22261         (ix86_compute_frame_layout): Likewise.
22263 2017-08-01  H.J. Lu  <hongjiu.lu@intel.com>
22265         PR target/81654
22266         * config/i386/i386.c (ix86_set_func_type): Disallow naked
22267         attribute with interrupt attribute.
22269 2017-08-01  Andrew Pinski  <apinski@cavium.com>
22271         * tree-ssa-scopedtables.c (hashable_expr_equal_p): Check
22272         BIT_INSERT_EXPR's operand 1
22273         to see if the types precision matches.
22275 2017-08-01  Martin Liska  <mliska@suse.cz>
22277         PR middle-end/70140
22278         * builtins.c (expand_builtin_memcpy_args): Remove.
22279         (expand_builtin_memcpy): Call newly added function
22280         expand_builtin_memory_copy_args.
22281         (expand_builtin_memcpy_with_bounds): Likewise.
22282         (expand_builtin_mempcpy): Remove last argument.
22283         (expand_builtin_mempcpy_with_bounds): Likewise.
22284         (expand_builtin_memory_copy_args): New function created from
22285         expand_builtin_mempcpy_args with small modifications.
22286         (expand_builtin_mempcpy_args): Remove.
22287         (expand_builtin_stpcpy): Remove unused argument.
22288         (expand_builtin): Likewise.
22289         (expand_builtin_with_bounds): Likewise.
22291 2017-08-01  Martin Liska  <mliska@suse.cz>
22293         Revert r250771
22294         Make mempcpy more optimal (PR middle-end/70140).
22296 2017-08-01  Jakub Jelinek  <jakub@redhat.com>
22298         PR target/81622
22299         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): For
22300         __builtin_vec_cmpne verify both arguments are compatible vectors
22301         before looking at TYPE_MODE on the element type.  For __builtin_vec_ld
22302         verify arg1_type is a pointer or array type.  For __builtin_vec_st,
22303         move computation of aligned to after checking the argument types.
22304         Formatting fixes.
22306         PR target/80846
22307         * config/rs6000/vsx.md (vextract_fp_from_shorth,
22308         vextract_fp_from_shortl): Add element mode after mode in gen_vec_init*
22309         calls.
22311 2017-08-01  Jerome Lambourg  <lambourg@adacore.com>
22312             Doug Rupp  <rupp@adacore.com>
22313             Olivier Hainque  <hainque@adacore.com>
22315         * config.gcc (arm-wrs-vxworks*): Rework to handle arm-wrs-vxworks7 as
22316         well as arm-wrs-vxworks. Update target_cpu_name from arm6 (arch v3) to
22317         arm8 (arch v4).
22318         * config/arm/vxworks.h (MAYBE_TARGET_BPABI_CPP_BUILTINS): New, helper
22319         for TARGET_OS_CPP_BUILTIN.
22320         (TARGET_OS_CPP_BUILTIN): Invoke MAYBE_TARGET_BPABI_CPP_BUILTINS(),
22321         refine CPU definitions for arm_arch5 and add those for arm_arch6 and
22322         arm_arch7.
22323         (MAYBE_ASM_ABI_SPEC): New, helper for SUBTARGET_EXTRA_ASM_SPEC,
22324         passing required abi options to the assembler for EABI configurations.
22325         (EXTRA_CC1_SPEC): New macro, to help prevent the implicit production
22326         of .text.hot and .text.unlikely sections for kernel modules when
22327         using ARM style exceptions.
22328         (CC1_SPEC): Remove obsolete attempt at mimicking Diab toolchain
22329         options. Add EXTRA_CC1_SPEC.
22330         (VXWORKS_ENDIAN_SPEC): Adjust comment and remove handling of Diab
22331         toolchain options.
22332         (DWARF2_UNWIND_INFO): Redefine to handle the pre/post VxWorks 7
22333         transition.
22334         (ARM_TARGET2_DWARF_FORMAT): Define.
22335         * config/arm/t-vxworks: Adjust multilib control to removal of the
22336         Diab command line options.
22338 2017-08-01  Martin Liska  <mliska@suse.cz>
22340         PR gcov-profile/81561
22341         * gcov.c (unblock): Make unblocking safe as we need to preserve
22342         index correspondence of blocks and block_lists.
22344 2017-08-01  Richard Biener  <rguenther@suse.de>
22346         PR tree-optimization/81181
22347         * tree-ssa-pre.c (compute_antic_aux): Defer clean() to ...
22348         (compute_antic): ... end of iteration here.
22350 2017-08-01  James Greenhalgh  <james.greenhalgh@arm.com>
22352         * common.opt (ftree-vectorize): No longer set flag_tree_vectorize.
22353         (ftree-loop-vectorize): Set as EnabledBy ftree-vectorize.
22354         (ftree-slp-vectorize): Likewise.
22355         * omp-expand (expand_omp_simd): Remove flag_tree_vectorize, as it
22356         can no longer be set independent of flag_tree_loop_vectorize.
22357         * omp-general.c (emp_max_vf): Likewise.
22358         * opts.c (enable_fdo_optimizations): Remove references to
22359         flag_tree_vectorize, these are now implicit.
22360         (common_handle_option): Remove handling for OPT_ftree_vectorize,
22361         and leave it for the options machinery.
22363 2017-08-01  Martin Liska  <mliska@suse.cz>
22365         PR middle-end/70140
22366         * builtins.c (expand_builtin_memcpy_args): Remove.
22367         (expand_builtin_memcpy): Call newly added function
22368         expand_builtin_memory_copy_args.
22369         (expand_builtin_memcpy_with_bounds): Likewise.
22370         (expand_builtin_mempcpy): Remove last argument.
22371         (expand_builtin_mempcpy_with_bounds): Likewise.
22372         (expand_builtin_memory_copy_args): New function created from
22373         expand_builtin_mempcpy_args with small modifications.
22374         (expand_builtin_mempcpy_args): Remove.
22375         (expand_builtin_stpcpy): Remove unused argument.
22376         (expand_builtin): Likewise.
22377         (expand_builtin_with_bounds): Likewise.
22379 2017-08-01  Uros Bizjak  <ubizjak@gmail.com>
22381         PR target/81641
22382         * config/i386/i386.c (ix86_print_operand_address_as): For -masm=intel
22383         print "ds:" only for immediates in generic address space.
22385 2017-08-01  Uros Bizjak  <ubizjak@gmail.com>
22387         PR target/81639
22388         * config/i386/i386.c (ix86_funciton_naked): New prototype.
22389         (ix86_function_ok_for_sibcall): Return false for naked functions.
22391 2017-08-01  Richard Biener  <rguenther@suse.de>
22393         * tree-ssa-pre.c (print_pre_expr): Handle NULL expr.
22394         (compute_antic): Seed worklist with exit block predecessors.
22395         * cfganal.c (dfs_find_deadend): For a cycle return the source
22396         of the edge closing it.
22398 2017-08-01  Tamar Christina  <tamar.christina@arm.com>
22400         * config/aarch64/aarch64.c
22401         (aarch64_can_const_movi_rtx_p): Move 0 check.
22403 2017-08-01  Bin Cheng  <bin.cheng@arm.com>
22405         * tree.h (POINTER_TYPE_OVERFLOW_UNDEFINED): Delete.
22406         * fold-const.c (fold_comparison, fold_binary_loc): Delete use of
22407         above macro.
22408         * match.pd: Ditto in address comparison pattern.
22410 2017-08-01  Bin Cheng  <bin.cheng@arm.com>
22412         PR tree-optimization/81627
22413         * tree-predcom.c (prepare_finalizers): Always rewrite into loop
22414         closed ssa form for store-store chain.
22416 2017-08-01  Bin Cheng  <bin.cheng@arm.com>
22418         PR tree-optimization/81620
22419         * tree-predcom.c (add_ref_to_chain): Don't set has_max_use_after
22420         for store-store chain.
22422 2017-08-01  Jakub Jelinek  <jakub@redhat.com>
22424         PR tree-optimization/81588
22425         * tree-ssa-reassoc.c (optimize_range_tests_var_bound): If
22426         ranges[i].in_p, invert comparison code ccode.  For >/>=,
22427         swap rhs1 and rhs2 and comparison code unconditionally,
22428         for </<= don't do that.  Don't swap rhs1/rhs2 again if
22429         ranges[i].in_p, instead invert comparison code ccode if
22430         opcode or oe->rank is BIT_IOR_EXPR.
22432         PR target/80846
22433         * optabs.def (vec_extract_optab, vec_init_optab): Change from
22434         a direct optab to conversion optab.
22435         * optabs.c (expand_vector_broadcast): Use convert_optab_handler
22436         with GET_MODE_INNER as last argument instead of optab_handler.
22437         * expmed.c (extract_bit_field_1): Likewise.  Use vector from
22438         vector extraction if possible and optab is available.
22439         * expr.c (store_constructor): Use convert_optab_handler instead
22440         of optab_handler.  Use vector initialization from smaller
22441         vectors if possible and optab is available.
22442         * tree-vect-stmts.c (vectorizable_load): Likewise.
22443         * doc/md.texi (vec_extract, vec_init): Document that the optabs
22444         now have two modes.
22445         * config/i386/i386.c (ix86_expand_vector_init): Handle expansion
22446         of vec_init from half-sized vectors with the same element mode.
22447         * config/i386/sse.md (ssehalfvecmode): Add V4TI case.
22448         (ssehalfvecmodelower, ssescalarmodelower): New mode attributes.
22449         (reduc_plus_scal_v8df, reduc_plus_scal_v4df, reduc_plus_scal_v2df,
22450         reduc_plus_scal_v16sf, reduc_plus_scal_v8sf, reduc_plus_scal_v4sf,
22451         reduc_<code>_scal_<mode>, reduc_umin_scal_v8hi): Add element mode
22452         after mode in gen_vec_extract* calls.
22453         (vec_extract<mode>): Renamed to ...
22454         (vec_extract<mode><ssescalarmodelower>): ... this.
22455         (vec_extract<mode><ssehalfvecmodelower>): New expander.
22456         (rotl<mode>3, rotr<mode>3, <shift_insn><mode>3, ashrv2di3): Add
22457         element mode after mode in gen_vec_init* calls.
22458         (VEC_INIT_HALF_MODE): New mode iterator.
22459         (vec_init<mode>): Renamed to ...
22460         (vec_init<mode><ssescalarmodelower>): ... this.
22461         (vec_init<mode><ssehalfvecmodelower>): New expander.
22462         * config/i386/mmx.md (vec_extractv2sf): Renamed to ...
22463         (vec_extractv2sfsf): ... this.
22464         (vec_initv2sf): Renamed to ...
22465         (vec_initv2sfsf): ... this.
22466         (vec_extractv2si): Renamed to ...
22467         (vec_extractv2sisi): ... this.
22468         (vec_initv2si): Renamed to ...
22469         (vec_initv2sisi): ... this.
22470         (vec_extractv4hi): Renamed to ...
22471         (vec_extractv4hihi): ... this.
22472         (vec_initv4hi): Renamed to ...
22473         (vec_initv4hihi): ... this.
22474         (vec_extractv8qi): Renamed to ...
22475         (vec_extractv8qiqi): ... this.
22476         (vec_initv8qi): Renamed to ...
22477         (vec_initv8qiqi): ... this.
22478         * config/rs6000/vector.md (VEC_base_l): New mode attribute.
22479         (vec_init<mode>): Renamed to ...
22480         (vec_init<mode><VEC_base_l>): ... this.
22481         (vec_extract<mode>): Renamed to ...
22482         (vec_extract<mode><VEC_base_l>): ... this.
22483         * config/rs6000/paired.md (vec_initv2sf): Renamed to ...
22484         (vec_initv2sfsf): ... this.
22485         * config/rs6000/altivec.md (splitter, altivec_copysign_v4sf3,
22486         vec_unpacku_hi_v16qi, vec_unpacku_hi_v8hi, vec_unpacku_lo_v16qi,
22487         vec_unpacku_lo_v8hi, mulv16qi3, altivec_vreve<mode>2): Add
22488         element mode after mode in gen_vec_init* calls.
22489         * config/aarch64/aarch64-simd.md (vec_init<mode>): Renamed to ...
22490         (vec_init<mode><Vel>): ... this.
22491         (vec_extract<mode>): Renamed to ...
22492         (vec_extract<mode><Vel>): ... this.
22493         * config/aarch64/iterators.md (Vel): New mode attribute.
22494         * config/s390/s390.c (s390_expand_vec_strlen, s390_expand_vec_movstr):
22495         Add element mode after mode in gen_vec_extract* calls.
22496         * config/s390/vector.md (non_vec_l): New mode attribute.
22497         (vec_extract<mode>): Renamed to ...
22498         (vec_extract<mode><non_vec_l>): ... this.
22499         (vec_init<mode>): Renamed to ...
22500         (vec_init<mode><non_vec_l>): ... this.
22501         * config/s390/s390-builtins.def (s390_vlgvb, s390_vlgvh, s390_vlgvf,
22502         s390_vlgvf_flt, s390_vlgvg, s390_vlgvg_dbl): Add element mode after
22503         vec_extract mode.
22504         * config/arm/iterators.md (V_elem_l): New mode attribute.
22505         * config/arm/neon.md (vec_extract<mode>): Renamed to ...
22506         (vec_extract<mode><V_elem_l>): ... this.
22507         (vec_extractv2di): Renamed to ...
22508         (vec_extractv2didi): ... this.
22509         (vec_init<mode>): Renamed to ...
22510         (vec_init<mode><V_elem_l>): ... this.
22511         (reduc_plus_scal_<mode>, reduc_plus_scal_v2di, reduc_smin_scal_<mode>,
22512         reduc_smax_scal_<mode>, reduc_umin_scal_<mode>,
22513         reduc_umax_scal_<mode>, neon_vget_lane<mode>, neon_vget_laneu<mode>):
22514         Add element mode after gen_vec_extract* calls.
22515         * config/mips/mips-msa.md (vec_init<mode>): Renamed to ...
22516         (vec_init<mode><unitmode>): ... this.
22517         (vec_extract<mode>): Renamed to ...
22518         (vec_extract<mode><unitmode>): ... this.
22519         * config/mips/loongson.md (vec_init<mode>): Renamed to ...
22520         (vec_init<mode><unitmode>): ... this.
22521         * config/mips/mips-ps-3d.md (vec_initv2sf): Renamed to ...
22522         (vec_initv2sfsf): ... this.
22523         (vec_extractv2sf): Renamed to ...
22524         (vec_extractv2sfsf): ... this.
22525         (reduc_plus_scal_v2sf, reduc_smin_scal_v2sf, reduc_smax_scal_v2sf):
22526         Add element mode after gen_vec_extract* calls.
22527         * config/mips/mips.md (unitmode): New mode iterator.
22528         * config/spu/spu.c (spu_expand_prologue, spu_allocate_stack,
22529         spu_builtin_extract): Add element mode after gen_vec_extract* calls.
22530         * config/spu/spu.md (inner_l): New mode attribute.
22531         (vec_init<mode>): Renamed to ...
22532         (vec_init<mode><inner_l>): ... this.
22533         (vec_extract<mode>): Renamed to ...
22534         (vec_extract<mode><inner_l>): ... this.
22535         * config/sparc/sparc.md (veltmode): New mode iterator.
22536         (vec_init<VMALL:mode>): Renamed to ...
22537         (vec_init<VMALL:mode><VMALL:veltmode>): ... this.
22538         * config/ia64/vect.md (vec_initv2si): Renamed to ...
22539         (vec_initv2sisi): ... this.
22540         (vec_initv2sf): Renamed to ...
22541         (vec_initv2sfsf): ... this.
22542         (vec_extractv2sf): Renamed to ...
22543         (vec_extractv2sfsf): ... this.
22544         * config/powerpcspe/vector.md (VEC_base_l): New mode attribute.
22545         (vec_init<mode>): Renamed to ...
22546         (vec_init<mode><VEC_base_l>): ... this.
22547         (vec_extract<mode>): Renamed to ...
22548         (vec_extract<mode><VEC_base_l>): ... this.
22549         * config/powerpcspe/paired.md (vec_initv2sf): Renamed to ...
22550         (vec_initv2sfsf): ... this.
22551         * config/powerpcspe/altivec.md (splitter, altivec_copysign_v4sf3,
22552         vec_unpacku_hi_v16qi, vec_unpacku_hi_v8hi, vec_unpacku_lo_v16qi,
22553         vec_unpacku_lo_v8hi, mulv16qi3): Add element mode after mode in
22554         gen_vec_init* calls.
22556 2017-08-01  Richard Biener  <rguenther@suse.de>
22558         PR tree-optimization/81297
22559         * tree-vrp.c (get_single_symbol): Remove assert, instead drop
22560         TREE_OVERFLOW from INTEGER_CSTs.
22562 2017-07-31  Segher Boessenkool  <segher@kernel.crashing.org>
22564         * config/rs6000/rs6000.c (enum rs6000_reg_type): Delete trailing comma.
22566 2017-07-31  Carl Love  <cel@us.ibm.com>
22568         * config/rs6000/rs6000-c: Add support for built-in functions
22569         vector signed char vec_xl_be (signed long long, signed char *);
22570         vector unsigned char vec_xl_be (signed long long, unsigned char *);
22571         vector signed int vec_xl_be (signed long long, signed int *);
22572         vector unsigned int vec_xl_be (signed long long, unsigned int *);
22573         vector signed long long vec_xl_be (signed long long, signed long long *);
22574         vector unsigned long long vec_xl_be (signed long long, unsigned long long *);
22575         vector signed short vec_xl_be (signed long long, signed short *);
22576         vector unsigned short vec_xl_be (signed long long, unsigned short *);
22577         vector double vec_xl_be (signed long long, double *);
22578         vector float vec_xl_be (signed long long, float *);
22579         * config/rs6000/altivec.h (vec_xl_be): Add #define.
22580         * config/rs6000/rs6000-builtin.def (XL_BE_V16QI, XL_BE_V8HI, XL_BE_V4SI)
22581         XL_BE_V2DI, XL_BE_V4SF, XL_BE_V2DF, XL_BE): Add definitions
22582         for the builtins.
22583         * config/rs6000/rs6000.c (altivec_expand_xl_be_builtin): Add function.
22584         (altivec_expand_builtin): Add switch statement to call
22585         altivec_expand_xl_be for each builtin.
22586         (altivec_init_builtins): Add def_builtin for _builtin_vsx_le_be_v8hi,
22587         __builtin_vsx_le_be_v4si, __builtin_vsx_le_be_v2di,
22588         __builtin_vsx_le_be_v4sf, __builtin_vsx_le_be_v2df,
22589         __builtin_vsx_le_be_v16qi.
22590         * doc/extend.texi: Update the built-in documentation file for the
22591         new built-in functions.
22593 2017-07-31  Uros Bizjak  <ubizjak@gmail.com>
22595         PR target/25967
22596         * config/i386/i386.c (ix86_allocate_stack_slots_for_args):
22597         New function.
22598         (TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS): Define.
22600 2017-07-31  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
22602         * config.gcc: Add z14.
22603         * config/s390/driver-native.c (s390_host_detect_local_cpu): Add
22604         CPU model numbers for z13s and z14.
22605         * config/s390/s390-c.c (s390_resolve_overloaded_builtin): Replace
22606         arch12 with z14.
22607         * config/s390/s390-opts.h (enum processor_type): Rename
22608         PROCESSOR_ARCH12 to PROCESSOR_3906_Z14.
22609         * config/s390/s390.c (processor_table): Add field for CPU name to
22610         be passed to Binutils.
22611         (s390_asm_output_machine_for_arch): Use the new field in
22612         processor_table for Binutils.
22613         (s390_expand_builtin): Replace arch12 with z14.
22614         (s390_issue_rate): Rename PROCESSOR_ARCH12 to PROCESSOR_3906_Z14.
22615         (s390_get_sched_attrmask): Likewise.
22616         (s390_get_unit_mask): Likewise.
22617         * config/s390/s390.opt: Add z14 to processor_type enum.
22619 2017-07-31  Martin Jambor  <mjambor@suse.cz>
22621         PR hsa/81477
22622         * ipa-fnsummary.c (ipa_fn_summary_generate): Set versionable
22623         regardless of optimization level.
22625 2017-07-31  Jan Hubicka <hubicka@ucw.cz>
22626             Martin Liska  <mliska@suse.cz>
22628         * predict.def: Remove old comment and adjust probability.
22629         * gimplify.c (should_warn_for_implicit_fallthrough): Ignore
22630         PREDICT statements.
22632 2017-07-31  Uros Bizjak  <ubizjak@gmail.com>
22634         PR target/25967
22635         * config/i386/i386.c (ix86_function_naked): New function.
22636         (ix86_can_use_return_insn_p): Return false for naked functions.
22637         (ix86_expand_prologue): Skip prologue for naked functions.
22638         (ix86_expand_epilogue): Skip epilogue for naked functions
22639         and emit trap instruction.
22640         (ix86_warn_func_return): New function.
22641         (ix86_attribute_table): Add "naked" attribute specification.
22642         (TARGET_WARN_FUNC_RETURN): Define.
22643         * doc/extend.texi (x86 Function Attributes) <naked>: Document it.
22645 2017-07-31  Martin Liska  <mliska@suse.cz>
22647         * gimple-pretty-print.c (dump_gimple_label): Never dump BB info.
22648         (dump_gimple_bb_header): Always dump BB info.
22649         (pp_cfg_jump): Do not append info about BB when dumping a jump.
22651 2017-07-31  Martin Liska  <mliska@suse.cz>
22653         PR sanitize/81530
22654         * convert.c (convert_to_integer_1): Guard condition with flag_sanitize_p
22655         also with current_function_decl non-null equality.
22657 2017-07-31  Jakub Jelinek  <jakub@redhat.com>
22659         PR sanitizer/81604
22660         * ubsan.c (ubsan_type_descriptor): For UBSAN_PRINT_ARRAY don't
22661         change type to the element type, instead add eltype variable and
22662         use it where we are interested in the element type.
22664         PR tree-optimization/81603
22665         * ipa-polymorphic-call.c
22666         (ipa_polymorphic_call_context::ipa_polymorphic_call_context): Perform
22667         offset arithmetic in offset_int, bail out if the resulting bit offset
22668         doesn't fit into shwi.
22670 2017-07-31  Martin Liska  <mliska@suse.cz>
22672         * gimplify.c (mostly_copy_tree_r): Remove Java specific hunk.
22673         (gimplify_save_expr): Fix comment.
22675 2017-07-30  H.J. Lu  <hongjiu.lu@intel.com>
22677         PR target/79793
22678         * config/i386/i386.c (ix86_function_arg): Update arguments for
22679         exception handler.
22680         (ix86_compute_frame_layout): Set the initial stack offset to
22681         INCOMING_FRAME_SP_OFFSET.  Update red-zone offset with
22682         INCOMING_FRAME_SP_OFFSET.
22683         (ix86_expand_epilogue): Don't pop the 'ERROR_CODE' off the
22684         stack before exception handler returns.
22685         * config/i386/i386.h (INCOMING_FRAME_SP_OFFSET): Add the
22686         the 'ERROR_CODE' for exception handler.
22688 2017-07-30  Uros Bizjak  <ubizjak@gmail.com>
22690         * config/i386/i386.h (ASM_PRINTF_EXTENSIONS): New macro.
22691         (ASM_OUTPUT_REG_PUSH): Rewrite with new operand modifiers.
22692         (ASM_OUTPUT_REG_POP): Ditto.
22693         * config/i386/i386.c (ix86_asm_output_function_label): Use fputs
22694         instead of asm_fprintf to output pure string.
22696 2017-07-29  Jakub Jelinek  <jakub@redhat.com>
22698         * debug.h (struct gcc_debug_hooks): Add IMPLICIT argument
22699         to imported_module_or_decl hook.
22700         (debug_nothing_tree_tree_tree_bool): Remove.
22701         (debug_nothing_tree_tree_tree_bool_bool): New declaration.
22702         * debug.c (do_nothing_debug_hooks): Use
22703         debug_nothing_tree_tree_tree_bool_bool instead of
22704         debug_nothing_tree_tree_tree_bool.
22705         * vmsdbgout.c (vmsdbg_debug_hooks): Likewise.
22706         * dbxout.c (dbx_debug_hooks, xcoff_debug_hooks): Likewise.
22707         * sdbout.c (sdb_debug_hooks): Likewise.
22708         * dwarf2out.c (dwarf2_lineno_debug_hooks): Likewise.
22709         (gen_namespace_die): Add DW_AT_export_symbols attribute if
22710         langhook wants it.
22711         (dwarf2out_imported_module_or_decl): Add IMPLICIT argument,
22712         if true, -gdwarf-5 and decl will have DW_AT_export_symbols
22713         attribute, don't add anything.
22715 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22717         * fold-const.c (fold_build1_stat_loc): Remove _stat from name.
22718         (fold_build2_stat_loc): Likewise.
22719         (fold_build3_stat_loc): Likewise.
22720         * fold-const.h (fold_build1, fold_build2, fold_build3): Adjust.
22721         (fold_build1_loc): Remove macro.
22722         (fold_build2_loc): Likewise.
22723         (fold_build3_loc): Likewise.
22725 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22727         * gimple.c (gimple_build_debug_bind_stat): Remove _stat from name.
22728         (gimple_build_debug_bind_source_stat): Likewise.
22729         * gimple.h (gimple_build_debug_bind): Remove macro.
22730         (gimple_build_debug_bind_source): Likewise.
22732 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22734         * bitmap.c (bitmap_alloc): Adjust.
22735         (bitmap_gc_alloc): Likewise.
22736         * bitmap.h (bitmap_initialize_stat): Remove _stat from name.
22738 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22740         * bitmap.c (bitmap_obstack_alloc_stat): Rename to bitmap_alloc.
22741         (bitmap_gc_alloc_stat): Rename to bitmap_gc_alloc.
22742         * bitmap.h (bitmap_obstack_alloc_stat): Adjust prototype.
22743         (bitmap_gc_alloc_stat): Likewise.
22744         (BITMAP_ALLOC, BITMAP_GGC_ALLOC): Adjust.
22746 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22748         * rtl.c (shallow_copy_rtx_stat): Remove _stat from name.
22749         * rtl.h (shallow_copy_rtx): Remove macro.
22751 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22753         * emit-rtl.c (gen_raw_REG): Adjust.
22754         * gengenrtl.c (gendef): Likewise.
22755         * rtl.c (rtx_alloc_stat): Remove _stat from name.
22756         * rtl.h (rtx_alloc): Remove macro.
22758 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22760         * tree.c (build_tree_list_vec_stat): Remove _stat from name.
22761         (build_tree_list_stat): Likewise.
22762         * tree.h (build_tree_list): Remove macro.
22763         (build_tree_list_vec): Likewise.
22765 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22767         * tree.c (make_vector_stat): Remove _stat from name.
22768         (build_vector_stat): Likewise.
22769         * tree.h (make_vector_stat): Remove macro.
22770         (build_vector_stat): Likewise.
22772 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22774         * tree.h (build_var_debug_value): Remove prototype.
22776 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22778         * tree.c (tree_cons_stat): Remove _stat from name.
22779         * tree.h (tree_cons): Remove macro.
22781 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22783         * tree.c (build_vl_exp_stat): Remove _stat from name.
22784         * tree.h (build_vl_exp): Remove macro.
22786 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22788         * tree.c (build_decl_stat): Remove _stat from name.
22789         * tree.h (build_decl): Remove macro.
22791 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22793         * gimple.c (gimple_build_with_ops_stat): Adjust.
22794         (gimple_alloc_stat): Remove _stat from name.
22795         * gimple.h (gimple_alloc): Remove macro.
22797 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22799         * tree.c (make_tree_vec_stat): Remove _stat from name.
22800         (grow_tree_vec_stat): Likewise.
22801         * tree.h (make_tree_vec_stat): Adjust prototype.
22802         (grow_tree_vec_stat): Likewise.
22803         (make_tree_vec): Remove macro.
22804         (grow_tree_vec): Likewise.
22806 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22808         * fold-const.c (fold_build1_stat_loc): Adjust.
22809         (fold_build2_stat_loc): Likewise.
22810         (fold_build3_stat_loc): Likewise.
22811         * tree.c (build0_stat): Remove _stat from name.
22812         (build1_stat): Likewise.
22813         (build2_stat): Likewise.
22814         (build3_stat): Likewise.
22815         (build4_stat): Likewise.
22816         (build5_stat): Likewise.
22817         * tree.h (build1_loc): Remove macro, and rename _stat function
22818         to this.
22819         (build2_loc): Likewise.
22820         (build3_loc): Likewise.
22821         (build4_loc): Likewise.
22822         (build5_loc): Likewise.
22824 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22826         * tree.c (make_int_cst_stat): Remove _stat from name.
22827         * tree.h (make_int_cst_stat): Adjust prototype.
22828         (make_int_cst): Remove macro.
22830 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22832         * tree.c (make_tre_binfo_stat): Remove _stat from name.
22833         * tree.h (make_tree_binfo_stat): Adjust prototype.
22834         (make_tree_binfo): Remove.
22836 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22838         * tree.c (copy_node_stat): Rename to copy_node.
22839         (build_distinct_type_copy): Adjust.
22840         * tree.h (copy_node_stat): Adjust prototype.
22841         (copy_node): Remove macro.
22843 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22845         * tree.c (make_node_stat): rename to make_node.
22846         (build_tree_list_stat): Adjust.
22847         (build0_stat): Likewise.
22848         (build2_stat): Likewise.
22849         (build3_stat): Likewise.
22850         (build4_stat): Likewise.
22851         (build5_stat): Likewise.
22852         (build_decl_stat): Likewise.
22853         * tree.h (make_node_stat): Adjust prototype.
22854         (make_node): remove macro.
22856 2017-07-28  Peter Bergner  <bergner@vnet.ibm.com>
22858         * config/rs6000/ppc-auxv.h (PPC_FEATURE2_DARN): New define.
22859         (PPC_FEATURE2_SCV): Likewise.
22860         * config/rs6000/rs6000.c (cpu_supports_info): Use them.
22862 2017-07-28  Tamar Christina  <tamar.christina@arm.com>
22864         * config/aarch64/aarch64.c
22865         (aarch64_internal_mov_immediate): Add new special pattern.
22866         * config/aarch64/aarch64.md (*movdi_aarch64):
22867         Add reg/32bit const mov case.
22869 2017-07-28  Tamar Christina  <tamar.christina@arm.com>
22870             Richard Sandiford <richard.sandiford@linaro.org>
22872         * config/aarch64/aarch64.md (mov<mode>): Generalize.
22873         (*movhf_aarch64, *movsf_aarch64, *movdf_aarch64):
22874         Add integer and movi cases.
22875         (movi-split-hf-df-sf split, fp16): New.
22876         (enabled): Added TARGET_FP_F16INST.
22877         * config/aarch64/iterators.md (GPF_HF): New.
22878         * config/aarch64/predicates.md (aarch64_reg_or_fp_float): New.
22880 2017-07-28  Tamar Christina  <tamar.christina@arm.com>
22882         * config/aarch64/aarch64.c
22883         (aarch64_simd_container_mode): Add prototype.
22884         (aarch64_expand_mov_immediate): Add HI support.
22885         (aarch64_reinterpret_float_as_int, aarch64_float_const_rtx_p: New.
22886         (aarch64_can_const_movi_rtx_p): New.
22887         (aarch64_preferred_reload_class):
22888         Remove restrictions of using FP registers for certain SIMD operations.
22889         (aarch64_rtx_costs): Added new cost for CONST_DOUBLE moves.
22890         (aarch64_valid_floating_const): Add integer move validation.
22891         (aarch64_simd_imm_scalar_p): Remove.
22892         (aarch64_output_scalar_simd_mov_immediate): Generalize function.
22893         (aarch64_legitimate_constant_p): Expand list of supported cases.
22894         * config/aarch64/aarch64-protos.h
22895         (aarch64_float_const_rtx_p, aarch64_can_const_movi_rtx_p): New.
22896         (aarch64_reinterpret_float_as_int): New.
22897         (aarch64_simd_imm_scalar_p): Remove.
22898         * config/aarch64/constraints.md (Uvi): New.
22899         (Dd): Split into Ds and new Dd.
22900         * config/aarch64/aarch64.md (*movsi_aarch64):
22901         Add SIMD mov case.
22902         (*movdi_aarch64): Add SIMD mov case.
22904 2017-07-28  Bin Cheng  <bin.cheng@arm.com>
22906         * tree-predcom.c: (struct chain): Handle store-store chain in which
22907         stores for elimination only store loop invariant values.
22908         (execute_pred_commoning_chain): Ditto.
22909         (prepare_initializers_chain_store_elim): Ditto.
22910         (prepare_finalizers): Ditto.
22911         (is_inv_store_elimination_chain): New function.
22912         (initialize_root_vars_store_elim_1): New function.
22914 2017-07-28  Bin Cheng  <bin.cheng@arm.com>
22916         * tree-predcom.c: Revise general description of the pass.
22917         (enum chain_type): New enum type for store elimination.
22918         (struct chain): New field supporting store elimination.
22919         (struct component): Ditto.
22920         (dump_chain): Dump store-stores chain.
22921         (release_chain): Release resources.
22922         (split_data_refs_to_components): Compute and create component
22923         contains only stores for elimination.
22924         (get_chain_last_ref_at): New function.
22925         (make_invariant_chain): Initialization.
22926         (make_rooted_chain): Specify chain type in parameter and record it.
22927         (add_looparound_copies): Skip for store-stores chain.
22928         (determine_roots_comp): Compute type of chain and pass it to
22929         make_rooted_chain.
22930         (initialize_root_vars_store_elim_2): New function.
22931         (finalize_eliminated_stores): New function.
22932         (remove_stmt): Handle store for elimination.
22933         (execute_pred_commoning_chain): Execute predictive commoning on
22934         store-store chains.
22935         (determine_unroll_factor): Skip unroll for store-stores chain.
22936         (prepare_initializers_chain_store_elim): New function.
22937         (prepare_initializers_chain): Hanlde store-store chain.
22938         (prepare_finalizers_chain, prepare_finalizers): New function.
22939         (tree_predictive_commoning_loop): Return integer value indicating
22940         if loop is unrolled or lcssa form is corrupted.
22941         (tree_predictive_commoning): Rewrite for lcssa form if necessary.
22943 2017-07-28  Bin Cheng  <bin.cheng@arm.com>
22945         * tree-predcom.c (initialize_root): Delete.
22946         (execute_pred_commoning_chain): Initialize root vars and replace
22947         reference of non-combined chain directly, rather than call above
22948         function.
22950 2017-07-28  Bin Cheng  <bin.cheng@arm.com>
22952         * tree-predcom.c (ref_at_iteration): Add parameter NITERS.  Compute
22953         memory reference to DR at (NITERS + ITERS)-th iteration of loop.
22955 2017-07-28  Bin Cheng  <bin.cheng@arm.com>
22957         * tree-predcom.c (struct chain): New field init_seq.
22958         (release_chain): Release init_seq.
22959         (prepare_initializers_chain): Record intialization stmts in above
22960         field.
22961         (insert_init_seqs): New function.
22962         (tree_predictive_commoning_loop): Call insert_init_seqs.
22964 2017-07-28  Bin Cheng  <bin.cheng@arm.com>
22966         * tree-predcom.c (determine_roots_comp): Skip trivial components.
22968 2017-07-28  Richard Biener  <rguenther@suse.de>
22970         * match.pd: Remove superfluous :c.
22971         * genmatch.c (simplify::id): Add member.
22972         (lower_commutative, lower_opt_convert, lower_cond, lower_for):
22973         Copy id.
22974         (current_id): New global.
22975         (dt_node::parent): Move from ...
22976         (dt_operand::parent): ... here.  Add for_id member.
22977         (is_a_helper <dt_operand *>::test): DT_TRUE is also a dt_operand.
22978         (decision_tree::find_node): Relax order requirement when
22979         merging DT_TRUE nodes to ones inbetween the current simplify
22980         and the one we try to merge with.  Add diagnostic whenever
22981         we need to enforce pattern order by not merging.
22982         (decision_tree::insert): Set current_id.
22983         (decision_tree::print_node): Dump parent node and for_id.
22984         (parser::last_id): Add member.
22985         (parser::push_simplify): Assign unique id.
22986         (parser::parser): Initialize last_id.
22988 2017-07-28  Martin Liska  <mliska@suse.cz>
22990         PR sanitizer/81340
22991         * sanopt.c (sanitize_rewrite_addressable_params): Set VALUE_EXPR after
22992         gimple_build_debug_bind.
22994 2017-07-28  Richard Biener  <rguenther@suse.de>
22996         PR tree-optimization/81502
22997         * match.pd: Add pattern combining BIT_INSERT_EXPR with
22998         BIT_FIELD_REF.
22999         * tree-cfg.c (verify_expr): Verify types of BIT_FIELD_REF
23000         size/pos operands.
23001         (verify_gimple_assign_ternary): Likewise for BIT_INSERT_EXPR pos.
23002         * gimple-fold.c (maybe_canonicalize_mem_ref_addr): Use bitsizetype
23003         for BIT_FIELD_REF args.
23004         * fold-const.c (make_bit_field_ref): Likewise.
23005         * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
23007 2017-07-28  Jakub Jelinek  <jakub@redhat.com>
23009         PR sanitizer/80998
23010         * sanopt.c (pass_sanopt::execute): Handle IFN_UBSAN_PTR.
23011         * tree-ssa-alias.c (call_may_clobber_ref_p_1): Likewise.
23012         * flag-types.h (enum sanitize_code): Add SANITIZER_POINTER_OVERFLOW.
23013         Or it into SANITIZER_UNDEFINED.
23014         * ubsan.c: Include gimple-fold.h and varasm.h.
23015         (ubsan_expand_ptr_ifn): New function.
23016         (instrument_pointer_overflow): New function.
23017         (maybe_instrument_pointer_overflow): New function.
23018         (instrument_object_size): Formatting fix.
23019         (pass_ubsan::execute): Call instrument_pointer_overflow
23020         and maybe_instrument_pointer_overflow.
23021         * internal-fn.c (expand_UBSAN_PTR): New function.
23022         * ubsan.h (ubsan_expand_ptr_ifn): Declare.
23023         * sanitizer.def (__ubsan_handle_pointer_overflow,
23024         __ubsan_handle_pointer_overflow_abort): New builtins.
23025         * tree-ssa-tail-merge.c (merge_stmts_p): Handle IFN_UBSAN_PTR.
23026         * internal-fn.def (UBSAN_PTR): New internal function.
23027         * opts.c (sanitizer_opts): Add pointer-overflow.
23028         * lto-streamer-in.c (input_function): Handle IFN_UBSAN_PTR.
23029         * fold-const.c (build_range_check): Compute pointer range check in
23030         integral type if pointer arithmetics would be needed.  Formatting
23031         fixes.
23033 2017-07-28  Martin Liska  <mliska@suse.cz>
23035         PR sanitizer/81460
23036         * sanopt.c (sanitize_rewrite_addressable_params): Do not rewrite
23037         parameters that are of a variable-length.
23039 2017-07-28  Sebastian Huber  <sebastian.huber@embedded-brains.de>
23041         * config.gcc (powerpc-*-rtems*): Remove rs6000/eabi.h.  Add
23042         rs6000/biarch64.h.
23043         * config/rs6000/rtems.h (ASM_DECLARE_FUNCTION_SIZE): New macro.
23044         (ASM_OUTPUT_SPECIAL_POOL_ENTRY_P): Likewise.
23045         (CRT_CALL_STATIC_FUNCTION): Likewise.
23046         (ASM_DEFAULT_SPEC): New define.
23047         (ASM_SPEC32): Likewise.
23048         (ASM_SPEC64): Likewise.
23049         (ASM_SPEC_COMMON): Likewise.
23050         (ASM_SPEC): Likewise.
23051         (INVALID_64BIT): Likewise.
23052         (LINK_OS_DEFAULT_SPEC): Likewise.
23053         (LINK_OS_SPEC32): Likewise.
23054         (LINK_OS_SPEC64): Likewise.
23055         (POWERPC_LINUX): Likewise.
23056         (PTRDIFF_TYPE): Likewise.
23057         (RESTORE_FP_PREFIX): Likewise.
23058         (RESTORE_FP_SUFFIX): Likewise.
23059         (SAVE_FP_PREFIX): Likewise.
23060         (SAVE_FP_SUFFIX): Likewise.
23061         (SIZE_TYPE): Likewise.
23062         (SUBSUBTARGET_OVERRIDE_OPTIONS): Likewise.
23063         (TARGET_64BIT): Likewise.
23064         (TARGET_64BIT): Likewise.
23065         (TARGET_AIX): Likewise.
23066         (WCHAR_TYPE_SIZE): Likewise.
23067         (WCHAR_TYPE): Undefine.
23068         (TARGET_OS_CPP_BUILTINS): Add 64-bit PowerPC defines.
23069         (CPP_OS_DEFAULT_SPEC): Use previous CPP_OS_RTEMS_SPEC.
23070         (CPP_OS_RTEMS_SPEC): Delete.
23071         (SUBSUBTARGET_EXTRA_SPECS): Remove cpp_os_rtems.  Add
23072         asm_spec_common, asm_spec32, asm_spec64, link_os_spec32, and
23073         link_os_spec64.
23074         * config/rs6000/t-rtems: Add mcpu=e6500/m64 multilibs.
23076 2017-07-28  Jakub Jelinek  <jakub@redhat.com>
23078         PR tree-optimization/81578
23079         * tree-parloops.c (build_new_reduction): Bail out if
23080         reduction_code isn't one of the standard OpenMP reductions.
23081         Move the details printing after that decision.
23083 2017-07-27  Peter Bergner  <bergner@vnet.ibm.com>
23085         * config/rs6000/predicates.md (volatile_mem_operand): Remove code
23086         related to reload_in_progress.
23087         (splat_input_operand): Likewise.
23088         * config/rs6000/rs6000-protos.h (rs6000_secondary_memory_needed_rtx):
23089         Delete prototype.
23090         * config/rs6000/rs6000.c (machine_function): Remove sdmode_stack_slot
23091         field.
23092         (TARGET_EXPAND_TO_RTL_HOOK): Delete.
23093         (TARGET_INSTANTIATE_DECLS): Likewise.
23094         (legitimate_indexed_address_p): Delete reload_in_progress code.
23095         (rs6000_debug_legitimate_address_p): Likewise.
23096         (rs6000_eliminate_indexed_memrefs): Likewise.
23097         (rs6000_emit_le_vsx_store): Likewise.
23098         (rs6000_emit_move_si_sf_subreg): Likewise.
23099         (rs6000_emit_move): Likewise.
23100         (register_to_reg_type): Likewise.
23101         (rs6000_pre_atomic_barrier): Likewise.
23102         (rs6000_machopic_legitimize_pic_address): Likewise.
23103         (rs6000_allocate_stack_temp): Likewise.
23104         (rs6000_address_for_fpconvert): Likewise.
23105         (rs6000_address_for_altivec): Likewise.
23106         (rs6000_secondary_memory_needed_rtx): Delete function.
23107         (rs6000_check_sdmode): Likewise.
23108         (rs6000_alloc_sdmode_stack_slot): Likewise.
23109         (rs6000_instantiate_decls): Likewise.
23110         * config/rs6000/rs6000.h (SECONDARY_MEMORY_NEEDED_RTX): Delete.
23111         * config/rs6000/rs6000.md (splitter for *movsi_got_internal):
23112         Delete reload_in_progress.
23113         (*vec_reload_and_plus_<mptrsize>): Likewise.
23114         * config/rs6000/vsx.md (vsx_mul_v2di): Likewise.
23115         (vsx_div_v2di): Likewise.
23116         (vsx_udiv_v2di): Likewise.
23118 2017-07-27  Peter Bergner  <bergner@vnet.ibm.com>
23120         * config/rs6000/rs6000.opt (mlra): Replace with stub.
23121         * config/rs6000/rs6000-cpus.def (POWERPC_MASKS): Delete OPTION_MASK_LRA.
23122         * config/rs6000/rs6000.c (TARGET_LRA_P): Delete.
23123         (rs6000_debug_reg_global): Delete print of LRA status.
23124         (rs6000_option_override_internal): Delete dead LRA related code.
23125         (rs6000_lra_p): Delete function.
23126         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mlra.
23128 2017-07-27  Sebastian Huber  <sebastian.huber@embedded-brains.de>
23130         * config.gcc (riscv*-*-elf*): Add (riscv*-*-rtems*).
23131         * config/riscv/rtems.h: New file.
23133 2017-07-27  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
23134             Sudakshina Das  <sudi.das@arm.com>
23136         * config/aarch64/aarch64.md
23137         (define_split for and<mode>3nr_compare): Move
23138         non aarch64_logical_operand to a register.
23139         (define_split for and_<SHIFT:optab><mode>3nr_compare0): Move non
23140         register immediate operand to a register.
23141         * config/aarch64/predicates.md (aarch64_mov_imm_operand): New.
23143 2017-07-27  Peter Bergner  <bergner@vnet.ibm.com>
23145         PR middle-end/81564
23146         * tree-cfg.c (group_case_labels_stmt): Handle already deleted blocks.
23148 2017-07-27  Richard Biener  <rguenther@suse.de>
23150         PR tree-optimization/81573
23151         PR tree-optimization/81494
23152         * tree-vect-loop.c (vect_create_epilog_for_reduction): Handle
23153         multi defuse cycle case.
23155 2017-07-27  Richard Biener  <rguenther@suse.de>
23157         PR tree-optimization/81571
23158         * tree-vect-slp.c (vect_build_slp_tree): Properly verify reduction
23159         PHIs.
23161 2017-07-27  Eric Botcazou  <ebotcazou@adacore.com>
23163         * config/sparc/sparc.c (sparc_option_override): Set MASK_FSMULD flag
23164         earlier and only if MASK_FPU is set.  Adjust formatting.
23166 2017-07-27  Martin Liska  <mliska@suse.cz>
23168         * opt-functions.awk: Add validation of value of Init.
23169         * optc-gen.awk: Pass new argument.
23171 2017-07-27  Martin Liska  <mliska@suse.cz>
23173         * auto-profile.c (autofdo_source_profile::update_inlined_ind_target):
23174         Fix wrong condition.
23176 2017-07-27  Martin Liska  <mliska@suse.cz>
23178         * auto-profile.c (afdo_annotate_cfg): Assign zero counts to
23179         BBs and edges seen by autoFDO.
23181 2017-07-27  Richard Biener  <rguenther@suse.de>
23183         PR tree-optimization/81502
23184         * tree-ssa.c (non_rewritable_lvalue_p): Handle BIT_INSERT_EXPR
23185         with incompatible but same sized type.
23186         (execute_update_addresses_taken): Likewise.
23188 2017-07-27  James Greenhalgh  <james.greenhalgh@arm.com>
23190         * tree-ssa-loop-ch.c (pass_ch::process_loop_p): Guard on
23191         flag_tree_loop_vectorize rather than flag_tree_vectorize.
23193 2017-07-27  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
23195         PR target/81534
23196         * config/s390/s390.md ("*atomic_compare_and_swap<mode>_1")
23197         ("*atomic_compare_and_swapdi_2", "*atomic_compare_and_swapsi_3"):
23198         Change s_operand to memory_operand.
23200 2017-07-27  Richard Sandiford  <richard.sandiford@linaro.org>
23202         * config/rs6000/rs6000-protos.h (rs6000_emit_le_vsx_permute): Declare.
23203         * config/rs6000/rs6000.c (rs6000_gen_le_vsx_permute): Replace with...
23204         (rs6000_emit_le_vsx_permute): ...this.  Take the destination as input.
23205         Emit instructions rather than returning an expression.  Handle TFmode
23206         and KFmode by casting to TImode.
23207         (rs6000_emit_le_vsx_load): Update to use rs6000_emit_le_vsx_permute.
23208         (rs6000_emit_le_vsx_store): Likewise.
23209         * config/rs6000/vsx.md (VSX_TI): New iterator.
23210         (*vsx_le_permute_<mode>): Use it instead of VSX_LE_128.
23211         (*vsx_le_undo_permute_<mode>): Likewise.
23212         (*vsx_le_perm_load_<mode>): Use rs6000_emit_le_vsx_permute to
23213         emit the split sequence.
23214         (*vsx_le_perm_store_<mode>): Likewise.
23216 2017-07-27  Jakub Jelinek  <jakub@redhat.com>
23218         PR tree-optimization/81555
23219         PR tree-optimization/81556
23220         * tree-ssa-reassoc.c (rewrite_expr_tree): Add NEXT_CHANGED argument,
23221         if true, force CHANGED for the recursive invocation.
23222         (reassociate_bb): Remember original length of ops array, pass
23223         len != orig_len as NEXT_CHANGED in rewrite_expr_tree call.
23225         * attribs.c (decl_attributes): Imply noinline, noclone and no_icf
23226         attributes for noipa attribute.  For naked attribute use
23227         lookup_attribute first before lookup_attribute_spec.
23228         * final.c (rest_of_handle_final): Disable IPA RA for functions with
23229         noipa attribute.
23230         * ipa-visibility.c (non_local_p): Fix comment typos.  Return true
23231         for functions with noipa attribute.
23232         (cgraph_externally_visible_p): Return true for functions with noipa
23233         attribute.
23234         * cgraph.c (cgraph_node::get_availability): Return AVAIL_INTERPOSABLE
23235         for functions with noipa attribute.
23236         * doc/extend.texi: Document noipa function attribute.
23237         * tree-ssa-structalias.c (refered_from_nonlocal_fn): Set *nonlocal_p
23238         also for functions with noipa attribute.
23239         (ipa_pta_execute): Set nonlocal_p also for nodes with noipa attribute.
23241 2017-07-26  Andrew Pinski  <apinski@cavium.com>
23243         * config/aarch64/aarch64.c (thunderx_vector_cost): Decrease cost of
23244         vec_unalign_load_cost and vec_unalign_store_cost.
23246 2017-07-26  Michael Meissner  <meissner@linux.vnet.ibm.com>
23248         * config/rs6000/rs6000-cpus.def (ISA_2_7_MASKS_SERVER): Delete
23249         -mvsx-small-integer option.
23250         (ISA_3_0_MASKS_IEEE): Likewise.
23251         (OTHER_VSX_VECTOR_MASKS): Likewise.
23252         (POWERPC_MASKS): Likewise.
23253         * config/rs6000/rs6000.opt (-mvsx-small-integer): Likewise.
23254         * config/rs6000/rs6000.c (rs6000_hard_regno_mode_ok): Simplify
23255         code, only testing for DImode being allowed in non-VSX floating
23256         point registers.
23257         (rs6000_init_hard_regno_mode_ok): Change TARGET_VSX_SMALL_INTEGER
23258         to TARGET_P8_VECTOR test.  Remove redundant VSX test inside of
23259         another VSX test.
23260         (rs6000_option_override_internal): Delete -mvsx-small-integer.
23261         (rs6000_expand_vector_set): Change TARGET_VSX_SMALL_INTEGER to
23262         TARGET_P8_VECTOR test.
23263         (rs6000_secondary_reload_simple_move): Likewise.
23264         (rs6000_preferred_reload_class): Delete TARGET_VSX_SMALL_INTEGER,
23265         since TARGET_P9_VECTOR was already tested.
23266         (rs6000_opt_masks): Remove -mvsx-small-integer.
23267         * config/rs6000/vsx.md (vsx_extract_<mode>): Delete
23268         TARGET_VSX_SMALL_INTEGER, since a test for TARGET_P9_VECTOR was
23269         used.
23270         (vsx_extract_<mode>_p9): Delete TARGET_VSX_SMALL_INTEGER, since a
23271         test for TARGET_VEXTRACTUB was used, and that uses
23272         TARGET_P9_VECTOR.
23273         (p9 extract splitter): Likewise.
23274         (vsx_extract_<mode>_di_p9): Likewise.
23275         (vsx_extract_<mode>_store_p9): Likewise.
23276         (vsx_extract_si): Delete TARGET_VSX_SMALL_INTEGER, since a test
23277         for TARGET_P9_VECTOR was used.  Delete code that is now dead with
23278         the elimination of TARGET_VSX_SMALL_INTEGER.
23279         (vsx_extract_<mode>_p8): Likewise.
23280         (vsx_ext_<VSX_EXTRACT_I:VS_scalar>_fl_<FL_CONV:mode>): Likewise.
23281         (vsx_ext_<VSX_EXTRACT_I:VS_scalar>_ufl_<FL_CONV:mode>): Likewise.
23282         (vsx_set_<mode>_p9): Likewise.
23283         (vsx_set_v4sf_p9): Likewise.
23284         (vsx_set_v4sf_p9_zero): Likewise.
23285         (vsx_insert_extract_v4sf_p9): Likewise.
23286         (vsx_insert_extract_v4sf_p9_2): Likewise.
23287         * config/rs6000/rs6000.md (sign extend splitter): Change
23288         TARGET_VSX_SMALL_INTEGER to TARGET_P8_VECTOR test.
23289         (floatsi<mode>2_lfiwax_mem): Likewise.
23290         (floatunssi<mode>2_lfiwzx_mem): Likewise.
23291         (float<QHI:mode><FP_ISA3:mode>2): Delete TARGET_VSX_SMALL_INTEGER,
23292         since a test for TARGET_P9_VECTOR was used.
23293         (float<QHI:mode><FP_ISA3:mode>2_internal): Likewise.
23294         (floatuns<QHI:mode><FP_ISA3:mode>2): Likewise.
23295         (floatuns<QHI:mode><FP_ISA3:mode>2_internal): Likewise.
23296         (fix_trunc<mode>si2): Change TARGET_VSX_SMALL_INTEGER to
23297         TARGET_P8_VECTOR test.
23298         (fix_trunc<mode>si2_stfiwx): Likewise.
23299         (fix_trunc<mode>si2_internal): Likewise.
23300         (fix_trunc<SFDF:mode><QHI:mode>2): Delete
23301         TARGET_VSX_SMALL_INTEGER, since a test for TARGET_P9_VECTOR was
23302         used.
23303         (fix_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
23304         (fixuns_trunc<mode>si2): Change TARGET_VSX_SMALL_INTEGER to
23305         TARGET_P8_VECTOR test.
23306         (fixuns_trunc<mode>si2_stfiwx): Likewise.
23307         (fixuns_trunc<SFDF:mode><QHI:mode>2): Delete
23308         TARGET_VSX_SMALL_INTEGER, since a test for TARGET_P9_VECTOR was
23309         used.
23310         (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
23311         (fctiw<u>z_<mode>_smallint): Delete TARGET_VSX_SMALL_INTEGER,
23312         since a test for TARGET_P9_VECTOR was used.
23313         (splitter for loading small constants): Likewise.
23315 2017-07-26  Andrew Pinski  <apinski@cavium.com>
23317         * config/aarch64/aarch64.c (thunderx_vector_cost): Fix
23318         vec_fp_stmt_cost.
23320 2017-07-26  H.J. Lu  <hongjiu.lu@intel.com>
23322         PR target/81563
23323         * config/i386/i386.c (sp_valid_at): Properly check CFA offset.
23324         (fp_valid_at): Likewise.
23326 2017-07-26  James Greenhalgh  <james.greenhalgh@arm.com>
23328         * config/aarch64/aarch64.c (cortexa57_addrcost_table): Remove.
23329         (qdf24xx_addrcost_table): Likewise.
23330         (cortexa57_tunings): Update to use generic_branch_cost.
23331         (cortexa72_tunings): Likewise.
23332         (cortexa73_tunings): Likewise.
23333         (qdf24xx_tunings): Likewise.
23335 2017-07-26  James Greenhalgh  <james.greenhalgh@arm.com>
23337         * config/aarch64/aarch64.c (cortexa57_branch_cost): Remove.
23338         (thunderx2t99_branch_cost): Likewise.
23339         (cortexa35_tunings): Update to use generic_branch_cost.
23340         (cortexa53_tunings): Likewise.
23341         (cortexa57_tunings): Likewise.
23342         (cortexa72_tunings): Likewise.
23343         (cortexa73_tunings): Likewise.
23344         (thunderx2t99_tunings): Likewise.
23346 2017-07-26  Sebastian Huber  <sebastian.huber@embedded-brains.de>
23348         * config/sparc/sparc.c (dump_target_flag_bits): Dump MASK_FSMULD.
23349         (sparc_option_override): Honour MASK_FSMULD.
23350         * config/sparc/sparc.h (MASK_FEATURES): Add MASK_FSMULD.
23351         * config/sparc/sparc.md (muldf3_extend): Use TARGET_FSMULD.
23352         * config/sparc/sparc.opt (mfsmuld): New option.
23353         * doc/invoke.texi (mfsmuld): Document option.
23355 2017-07-26  Marek Polacek  <polacek@redhat.com>
23357         PR middle-end/70992
23358         * tree.c (build2_stat): Don't set TREE_CONSTANT on divisions by zero.
23360 2017-07-26  Richard Biener  <rguenther@suse.de>
23362         * gimple-match-head.c (do_valueize): Return OP if valueize
23363         returns NULL_TREE.
23364         (get_def): New helper to get at the def stmt of a SSA name
23365         if valueize allows.
23366         * genmatch.c (dt_node::gen_kids_1): Use get_def instead of
23367         do_valueize to get at the def stmt.
23368         (dt_operand::gen_gimple_expr): Simplify do_valueize calls.
23370 2017-07-26  Wilco Dijkstra  <wdijkstr@arm.com>
23372         PR middle-end/46932
23373         * auto-inc-dec.c (parse_add_or_inc): Block autoinc on sfp.
23375 2017-07-26  Martin Liska  <mliska@suse.cz>
23377         PR sanitize/81186
23378         * function.c (expand_function_start): Make expansion of
23379         nonlocal_goto_save_area after parm_birth_insn.
23381 2017-07-26  Sebastian Huber  <sebastian.huber@embedded-brains.de>
23383         * config/sparc/sparc.c (sparc_option_override): Remove MASK_FPU
23384         from all CPU target flags enable members.
23386 2017-07-26  Richard Biener  <rguenther@suse.de>
23388         * genmatch.c (dt_simplify::gen): Make iterator vars const.
23389         (decision_tree::gen): Make 'type' const.
23390         (write_predicate): Likewise.
23392 2017-07-24  Michael Meissner  <meissner@linux.vnet.ibm.com>
23394         * config/rs6000/rs6000.c (rs6000_init_hard_regno_mode_ok):
23395         Eliminate TARGET_UPPER_REGS_{DF,DI,SF} usage.
23396         (rs6000_option_override_internal): Likewise.
23397         (rs6000_expand_vector_set): Likewise.
23398         * config/rs6000/rs6000.h (TARGET_UPPER_REGS_DF): Delete.
23399         (TARGET_UPPER_REGS_SF): Likewise.
23400         (TARGET_UPPER_REGS_DI): Likewise.
23401         (TARGET_VEXTRACTUB): Eliminate TARGET_UPPER_REGS_{DF,DI,SF}.
23402         (TARGET_DIRECT_MOVE_64BIT): Likewise.
23403         * config/rs6000/rs6000.md (ALTIVEC_DFORM): Likewise.
23404         (float<QHI:mode><FP_ISA3:mode>2_internal): Likewise.
23405         (Splitters for DI constants in Altivec registers): Likewise.
23406         * config/rs6000/vsx.md (vsx_set_<mode>_p9): Likewise.
23407         (vsx_set_v4sf_p9): Likewise.
23408         (vsx_set_v4sf_p9_zero): Likewise.
23409         (vsx_insert_extract_v4sf_p9): Likewise.
23410         (vsx_insert_extract_v4sf_p9_2): Likewise.
23412 2017-07-25  Carl Love  <cel@us.ibm.com>
23414         * doc/extend.texi: Update the built-in documentation file for the
23415         existing built-in functions
23416         vector signed char vec_cnttz (vector signed char);
23417         vector unsigned char vec_cnttz (vector unsigned char);
23418         vector signed short vec_cnttz (vector signed short);
23419         vector unsigned short vec_cnttz (vector unsigned short);
23420         vector signed int vec_cnttz (vector signed int);
23421         vector unsigned int vec_cnttz (vector unsigned int);
23422         vector signed long long vec_cnttz (vector signed long long);
23423         vector unsigned long long vec_cnttz (vector unsigned long long);
23425 2017-07-25  Andrew Pinski  <apinski@cavium.com>
23427         * tree-ssa-uninit.c (warn_uninitialized_vars): Don't warn about memory
23428         accesses where the use is for the first operand of a BIT_INSERT.
23430 2017-07-25  Jim Wilson  <jim.wilson@linaro.org>
23432         PR bootstrap/81521
23433         * config/i386/winnt-cxx.c (i386_pe_adjust_class_at_definition): Look
23434         for FUNCTION_DECLs in TYPE_FIELDS rather than TYPE_METHODS.
23436 2017-07-25  Jim Wilson  <jim.wilson@linaro.org>
23438         * config/i386/gstabs.h: Delete.
23439         * config/i386/openbsd.h, config/i386/t-openbsd: Likewise.
23441 2017-07-25  Uros Bizjak  <ubizjak@gmail.com>
23443         * config/i386/i386.c (ix86_decompose_address): Do not check for
23444         register RTX when looking at index_reg or base_reg.
23445         * config/i386/i386.h (INCOMING_RETURN_ADDR_RTX): Use stack_pointer_rtx.
23447 2017-07-25  Eric Botcazou  <ebotcazou@adacore.com>
23449         * gimple.c (gimple_assign_set_rhs_with_ops): Do not ask gsi_replace
23450         to update EH info here.
23452 2017-07-25  Alexander Monakov  <amonakov@ispras.ru>
23454         * match.pd ((X * CST1) * CST2): Simplify to X * (CST1 * CST2).
23456 2017-07-25  Alexander Monakov  <amonakov@ispras.ru>
23458         * match.pd ((X * CST) * Y): Reassociate to (X * Y) * CST.
23460 2017-07-25  Torsten Duwe  <duwe@suse.de>
23462         * common.opt: Introduce -fpatchable-function-entry
23463         command line option, and its variables function_entry_patch_area_size
23464         and function_entry_patch_area_start.
23465         * opts.c (common_handle_option): Add -fpatchable_function_entry_ case,
23466         including a two-value parser.
23467         * target.def (print_patchable_function_entry): New target hook.
23468         * targhooks.h (default_print_patchable_function_entry): New function.
23469         * targhooks.c (default_print_patchable_function_entry): Likewise.
23470         * toplev.c (process_options): Switch off IPA-RA if
23471         patchable function entries are being generated.
23472         * varasm.c (assemble_start_function): Look at the
23473         patchable-function-entry command line switch and current
23474         function attributes and maybe generate NOP instructions by
23475         calling the print_patchable_function_entry hook.
23476         * doc/extend.texi: Document patchable_function_entry attribute.
23477         * doc/invoke.texi: Document -fpatchable_function_entry
23478         command line option.
23479         * doc/tm.texi.in (TARGET_ASM_PRINT_PATCHABLE_FUNCTION_ENTRY):
23480         New target hook.
23481         * doc/tm.texi: Re-generate.
23483 2017-07-25  Jakub Jelinek  <jakub@redhat.com>
23485         PR target/81532
23486         * config/i386/constraints.md (Yd, Ye): Use ALL_SSE_REGS for
23487         TARGET_AVX512DQ rather than TARGET_AVX512BW.
23489 2017-07-25  Tamar Christina  <tamar.christina@arm.com>
23491         * config/arm/parsecpu.awk (all_cores): Remove duplicates.
23493 2017-07-25  Richard Biener  <rguenther@suse.de>
23495         PR tree-optimization/81455
23496         * tree-ssa-loop-unswitch.c (find_loop_guard): Make sure to
23497         not walk in cycles when looking for guards.
23499 2017-07-25  Richard Biener  <rguenther@suse.de>
23501         PR tree-optimization/81529
23502         * tree-vect-stmts.c (process_use): Disregard live induction PHIs
23503         when optimizing backedge uses.
23505 2017-07-25  David Edelsohn  <dje.gcc@gmail.com>
23507         * dwarf2asm.c (dw2_asm_output_nstring): Encode double quote
23508         character for AIX.
23509         * dwarf2out.c (output_macinfo): Copy debug_line_section_label
23510         to dl_section_ref.  On AIX, append an expression to subtract
23511         the size of the section length to dl_section_ref.
23513 2017-07-25  Segher Boessenkool  <segher@kernel.crashing.org>
23515         * configure.ac: If any of the config.* scripts fail, exit 1.
23516         * configure: Regenerate.
23518 2017-07-25  Richard Biener  <rguenther@suse.de>
23520         PR middle-end/81546
23521         * tree-ssa-operands.c (verify_imm_links): Remove cap on number
23522         of immediate uses, be more verbose on errors.
23524 2017-07-25  Richard Biener  <rguenther@suse.de>
23526         PR tree-optimization/81510
23527         * tree-vect-loop.c (vect_is_simple_reduction): When the
23528         reduction stmt is not inside the loop bail out.
23530 2017-07-25  Richard Biener  <rguenther@suse.de>
23532         PR tree-optimization/81303
23533         * tree-vect-loop-manip.c (vect_loop_versioning): Build
23534         profitability check against LOOP_VINFO_NITERSM1.
23536 2017-07-25  Alexander Monakov  <amonakov@ispras.ru>
23538         * domwalk.c (cmp_bb_postorder): Simplify.
23539         (sort_bbs_postorder): New function.  Use it...
23540         (dom_walker::walk): ...here to optimize common cases.
23542 2017-07-25  Martin Liska  <mliska@suse.cz>
23544         PR ipa/81520
23545         * ipa-visibility.c (function_and_variable_visibility): Make the
23546         redirection just on target that supports aliasing.
23547         Fix GNU coding style.
23549 2017-07-25  Sebastian Huber  <sebastian.huber@embedded-brains.de>
23551         PR libgcc/61152
23552         * config/aarch64/rtems.h: Add GCC Runtime Library Exception.
23553         Format changes.
23554         * config/arm/rtems.h: Likewise.
23555         * config/bfin/rtems.h: Likewise.
23556         * config/i386/rtemself.h: Likewise.
23557         * config/lm32/rtems.h: Likewise.
23558         * config/m32c/rtems.h: Likewise.
23559         * config/m68k/rtemself.h: Likewise.
23560         * config/microblaze/rtems.h: Likewise.
23561         * config/mips/rtems.h: Likewise.
23562         * config/moxie/rtems.h: Likewise.
23563         * config/nios2/rtems.h: Likewise.
23564         * config/powerpcspe/rtems.h: Likewise.
23565         * config/rs6000/rtems.h: Likewise.
23566         * config/rtems.h: Likewise.
23567         * config/sh/rtems.h: Likewise.
23568         * config/sh/rtemself.h: Likewise.
23569         * config/sparc/rtemself.h: Likewise.
23571 2017-07-25  Georg-Johann Lay  <avr@gjlay.de>
23573         PR 81487
23574         * hsa-brig.c (brig_init): Use xasprintf instead of asprintf.
23575         * gimple-pretty-print.c (dump_profile, dump_probability): Same.
23576         * tree-ssa-structalias.c (alias_get_name): Same.
23578 2017-07-25  Bin Cheng  <bin.cheng@arm.com>
23580         PR target/81414
23581         * config/aarch64/cortex-a57-fma-steering.c (analyze): Skip fmul/fmac
23582         instructions if no du chain is found.
23584 2017-07-25  Georg-Johann Lay  <avr@gjlay.de>
23586         * config/avr/avr-log.c (avr_log_vadump) ['T']: Print NULL-TREE.
23588 2017-07-25  Richard Biener  <rguenther@suse.de>
23590         PR middle-end/81505
23591         * fold-const.c (fold_negate_const): TREE_OVERFLOW should be
23592         sticky.
23594 2017-07-24  Michael Meissner  <meissner@linux.vnet.ibm.com>
23596         * config/rs6000/rs6000-cpus.def (ISA_2_6_MASKS_SERVER): Delete
23597         upper-regs options.
23598         (ISA_2_7_MASKS_SERVER): Likewise.
23599         (ISA_3_0_MASKS_IEEE): Likewise.
23600         (OTHER_P8_VECTOR_MASKS): Likewise.
23601         (OTHER_VSX_VECTOR_MASKS): Likewise.
23602         (POWERPC_MASKS): Likewise.
23603         (power7 cpu): Use ISA_2_6_MASKS_SERVER instead of using a
23604         duplicate list of options.
23605         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove
23606         explicit -mupper-regs options.
23607         * config/rs6000/rs6000.opt (-mvsx-scalar-memory): Delete
23608         -mupper-regs* options.  Delete -mvsx-scalar-memory, which was an
23609         alias for -mupper-regs-df.
23610         * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): Likewise.
23611         (rs6000_init_hard_regno_mode_ok): Likewise.
23612         (rs6000_option_override_internal): Likewise.
23613         (rs6000_opt_masks): Likewise.
23614         * config/rs6000/rs6000.h (TARGET_UPPER_REGS_DF): Define upper regs
23615         options in terms of whether -mvsx or -mpower8-vector was used.
23616         (TARGET_UPPER_REGS_DI): Likewise.
23617         (TARGET_UPPER_REGS_SF): Likewise.
23618         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete the
23619         -mupper-regs-* options.
23621 2017-07-24  Segher Boessenkool  <segher@kernel.crashing.org>
23623         * passes.c (emergency_dump_function): Print some empty lines and a
23624         header before the RTL dump.
23626 2017-07-24  Segher Boessenkool  <segher@kernel.crashing.org>
23628         * cfgrtl.c (rtl_dump_bb): Don't call NEXT_INSN on NULL.
23630 2017-07-24  Wilco Dijkstra  <wdijkstr@arm.com>
23632         PR target/79041
23633         * config/aarch64/aarch64.c (aarch64_classify_symbol):
23634         Avoid SYMBOL_SMALL_ABSOLUTE for literals with pc-relative literals.
23636 2017-07-24  Carl Love  <cel@us.ibm.com>
23638         * config/rs6000/rs6000-c.c: Add support for built-in functions
23639         vector float vec_extract_fp32_from_shorth (vector unsigned short);
23640         vector float vec_extract_fp32_from_shortl (vector unsigned short);
23641         * config/rs6000/altivec.h (vec_extract_fp_from_shorth,
23642         vec_extract_fp_from_shortl): Add defines for the two builtins.
23643         * config/rs6000/rs6000-builtin.def (VEXTRACT_FP_FROM_SHORTH,
23644         VEXTRACT_FP_FROM_SHORTL): Add BU_P9V_OVERLOAD_1 and BU_P9V_VSX_1
23645         new builtins.
23646         * config/rs6000/vsx.md vsx_xvcvhpsp): Add define_insn.
23647         (vextract_fp_from_shorth, vextract_fp_from_shortl): Add define_expands.
23648         * doc/extend.texi: Update the built-in documentation file for the
23649         new built-in function.
23651 2017-07-24  Jakub Jelinek  <jakub@redhat.com>
23653         PR bootstrap/81521
23654         * tree.def: Remove TYPE_METHODS documentation, adjust TYPE_FIELDS
23655         documentation.
23656         * doc/generic.texi: Likewise.
23657         * config/i386/winnt-cxx.c (i386_pe_adjust_class_at_definition): Look
23658         for FUNCTION_DECLs in TYPE_FIELDS rather than TYPE_METHODS.
23660 2017-07-24  Jackson Woodruff  <jackson.woodruff@arm.com>
23662         * config/aarch64/aarch64-simd.md (aarch64_mla_elt_merge<mode>): New.
23663         (aarch64_mls_elt_merge<mode>): Likewise.
23665 2017-07-23  Krister Walfridsson  <krister.walfridsson@gmail.com>
23667         * config.gcc (*-*-netbsd*): Remove check for NetBSD versions not
23668         having __cxa_atexit.
23670 2017-07-23  Michael Collison  <michael.collison@arm.com>
23672         * config/arm/arm.c (arm_option_override): Deprecate
23673         use of -mstructure-size-boundary.
23674         * config/arm/arm.opt: Deprecate -mstructure-size-boundary.
23675         * doc/invoke.texi: Deprecate -mstructure-size-boundary.
23677 2017-07-23  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
23679         PR target/80695
23680         * config/rs6000/rs6000.c (rs6000_builtin_vectorization_cost):
23681         Reduce cost estimate for direct moves.
23683 2017-07-23  Uros Bizjak  <ubizjak@gmail.com>
23685         PR target/80569
23686         * config/i386/i386.c (ix86_option_override_internal): Disable
23687         BMI, BMI2 and TBM instructions for -m16.
23689 2017-07-21  Carl Love  <cel@us.ibm.com>
23691         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
23692         ALTIVEC_BUILTIN_VMULESW, ALTIVEC_BUILTIN_VMULEUW,
23693         ALTIVEC_BUILTIN_VMULOSW, ALTIVEC_BUILTIN_VMULOUW entries.
23694         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin,
23695         builtin_function_type): Add ALTIVEC_BUILTIN_* case statements.
23696         * config/rs6000/altivec.md (MVULEUW, VMULESW, VMULOUW,
23697         VMULOSW): New enum "unspec" values.
23698         (altivec_vmuleuw, altivec_vmulesw, altivec_vmulouw,
23699         altivec_vmulosw): New patterns.
23700         * config/rs6000/rs6000-builtin.def (VMLEUW, VMULESW, VMULOUW,
23701         VMULOSW): Add definitions.
23703 2017-07-21  Jim Wilson  <jim.wilson@linaro.org>
23705         * config/aarch64/aarch64-cores.def (falkor): Add AARCH64_FL_RDMA.
23706         (qdf24xx): Likewise.
23707         * config/aarch64/aarch64-options-extensions.def (rdma); New.
23708         * config/aarch64/aarch64.h (AARCH64_FL_RDMA): New.
23709         (AARCH64_FL_V8_1): Renumber.
23710         (AARCH64_FL_FOR_ARCH8_1): Add AARCH64_FL_RDMA.
23711         (AARCH64_ISA_RDMA): Use AARCH64_FL_RDMA.
23712         * config/aarch64/arm_neon.h: Use +rdma instead of arch=armv8.1-a.
23713         * doc/invoke.texi (AArch64 Options): Mention +rmda in -march docs.  Add
23714         rdma to feature modifiers list.
23716 2017-07-21  Yury Gribov  <tetra2005@gmail.com>
23718         PR middle-end/56727
23719         * ipa-visibility (function_and_variable_visibility): Convert
23720         recursive PLT call to direct call if appropriate.
23722 2017-07-21  Andrew Pinski  <apinski@cavium.com>
23724         * tree-ssa-sccvn.c (vn_nary_op_eq): Check BIT_INSERT_EXPR's
23725         operand 1 to see if the types precision matches.
23726         * fold-const.c (operand_equal_p): Likewise.
23728 2017-07-21  Richard Biener  <rguenther@suse.de>
23730         PR tree-optimization/81303
23731         * tree-vect-data-refs.c (vect_get_peeling_costs_all_drs): Pass
23732         in datarefs vector.  Allow NULL dr0 for no peeling cost estimate.
23733         (vect_peeling_hash_get_lowest_cost): Adjust.
23734         (vect_enhance_data_refs_alignment): Likewise.  Use
23735         vect_get_peeling_costs_all_drs to compute the penalty for no
23736         peeling to match up costs.
23738 2017-07-21  Richard Biener  <rguenther@suse.de>
23740         PR tree-optimization/81500
23741         * tree-vect-loop.c (vect_is_simple_reduction): Properly fail if
23742         we didn't identify a reduction path.
23744 2017-07-21  Tom de Vries  <tom@codesourcery.com>
23745             Cesar Philippidis  <cesar@codesourcery.com>
23747         PR gcov-profile/81442
23748         * config/nvptx/nvptx.c (nvptx_goacc_reduction_init): Add missing edge
23749         probabilities.
23751 2017-07-21  Tom de Vries  <tom@codesourcery.com>
23753         PR lto/81430
23754         * config/nvptx/nvptx.c (nvptx_override_options_after_change): New
23755         function.
23756         (TARGET_OVERRIDE_OPTIONS_AFTER_CHANGE): Define to
23757         nvptx_override_options_after_change.
23759 2017-07-21  Ulrich Drepper  <drepper@redhat.com>
23761         * dwarf2out.c (output_file_names): Avoid double testing for
23762         dwarf_version >= 5.
23764 2017-07-21  Georg-Johann Lay  <avr@gjlay.de>
23766         * doc/invoke.texi (AVR Built-in Functions): Re-layout section.
23768 2016-07-21  Jan Hubicka  <hubicka@ucw.cz>
23770         * cfgcleanup.c (flow_find_cross_jump): Do not crossjump across
23771         hot/cold regions.
23772         (try_crossjump_to_edge): Do not punt on partitioned functions.
23774 2016-07-21  Jan Hubicka  <hubicka@ucw.cz>
23776         * bb-reorder.c (find_rarely_executed_basic_blocks_and_crossing_edges):
23777         Put all BBs reachable only via paths crossing cold region to cold
23778         region.
23779         * cfgrtl.c (find_bbs_reachable_by_hot_paths): New function.
23781 2016-07-21  Richard Biener  <rguenther@suse.de>
23783         PR tree-optimization/81303
23784         * tree-vect-loop.c (vect_estimate_min_profitable_iters): Take
23785         into account prologue and epilogue iterations when raising
23786         min_profitable_iters to sth at least covering one vector iteration.
23788 2017-07-21  Tamar Christina  <tamar.christina@arm.com>
23790         * config/arm/arm.c (arm_test_cpu_arch_dat):
23791         Check for overlap.
23793 2017-07-20  Nathan Sidwell  <nathan@acm.org>
23795         Remove TYPE_METHODS.
23796         * tree.h (TYPE_METHODS): Delete.
23797         * dwarf2out.c (gen_member_die): Member fns are on TYPE_FIELDS.
23798         * dbxout.c (dbxout_type_fields): Ignore FUNCTION_DECLs.
23799         (dbxout_type_methods): Scan TYPE_FIELDS.
23800         (dbxout_type): Don't check TYPE_METHODS here.
23801         * function.c (use_register_for_decl): Always ignore register for
23802         class types when not optimizing.
23803         * ipa-devirt.c (odr_types_equivalent_p): Delete TYPE_METHODS scan.
23804         * tree.c (free_lang_data_in_type): Stitch out member functions and
23805         templates from TYPE_FIELDS.
23806         (build_distinct_type_copy, verify_type_variant,
23807         verify_type): Member fns are on TYPE_FIELDS.
23808         * tree-dump.c (dequeue_and_dump): No TYPE_METHODS.
23809         * tree-pretty-print.c (dump_generic_node): Likewise.
23811 2017-07-20  Jakub Jelinek  <jakub@redhat.com>
23813         PR target/80846
23814         * config/i386/i386.c (ix86_expand_vector_init_general): Handle
23815         V2TImode and V4TImode.
23816         (ix86_expand_vector_extract): Likewise.
23817         * config/i386/sse.md (VMOVE): Enable V4TImode even for just
23818         TARGET_AVX512F, instead of only for TARGET_AVX512BW.
23819         (ssescalarmode): Handle V4TImode and V2TImode.
23820         (VEC_EXTRACT_MODE): Add V4TImode and V2TImode.
23821         (*vec_extractv2ti, *vec_extractv4ti): New insns.
23822         (VEXTRACTI128_MODE): New mode iterator.
23823         (splitter for *vec_extractv?ti first element): New.
23824         (VEC_INIT_MODE): New mode iterator.
23825         (vec_init<mode>): Consolidate 3 expanders into one using
23826         VEC_INIT_MODE mode iterator.
23828 2017-07-20  Alexander Monakov  <amonakov@ispras.ru>
23830         * lra-assigns.c (pseudo_compare_func): Fix comparison step based on
23831         non_spilled_static_chain_regno_p.
23833 2017-07-20  Alexander Monakov  <amonakov@ispras.ru>
23835         * gimple-ssa-store-merging.c (sort_by_bitpos): Return 0 on equal bitpos.
23837 2017-07-20  Jan Hubicka  <hubicka@ucw.cz>
23839         * bb-reorder.c (connect_traces): Allow copying of blocks within
23840         single partition.
23842 2017-07-20  Richard Biener  <rguenther@suse.de>
23844         * gimple.h (gimple_phi_result): Add gphi * overload.
23845         (gimple_phi_result_ptr): Likewise.
23846         (gimple_phi_arg): Likewise.  Adjust index assert to only
23847         allow actual argument accesses rather than all slots available
23848         by capacity.
23849         (gimple_phi_arg_def): Add gphi * overload.
23850         * tree-phinodes.c (make_phi_node): Initialize only actual
23851         arguments.
23852         (resize_phi_node): Clear memory not covered by old node,
23853         do not initialize excess argument slots.
23854         (reserve_phi_args_for_new_edge): Initialize new argument slot
23855         completely.
23857 2017-07-20  Bin Cheng  <bin.cheng@arm.com>
23859         PR tree-optimization/81388
23860         Revert r238585:
23861         2016-07-21  Bin Cheng  <bin.cheng@arm.com>
23863         * tree-ssa-loop-niter.c (number_of_iterations_lt_to_ne): Clean up
23864         by removing computation of may_be_zero.
23866 2017-07-18  Jan Hubicka  <hubicka@ucw.cz>
23867             Tom de Vries  <tom@codesourcery.com>
23869         PR middle-end/81030
23870         * cfgbuild.c (find_many_sub_basic_blocks): Update REG_BR_PROB note
23871         when gimple level profile disagrees with what RTL expander did.
23873 2017-07-20  Richard Biener  <rguenther@suse.de>
23875         PR tree-optimization/61171
23876         * tree-vectorizer.h (slp_instance): Add reduc_phis member.
23877         (vect_analyze_stmt): Add slp instance parameter.
23878         (vectorizable_reduction): Likewise.
23879         * tree-vect-loop.c (vect_analyze_loop_operations): Adjust.
23880         (vect_is_simple_reduction): Deal with chains not detected
23881         as SLP reduction chain, specifically not properly associated
23882         chains containing a mix of plus/minus.
23883         (get_reduction_op): Remove.
23884         (get_initial_defs_for_reduction): Simplify, pass in whether
23885         this is a reduction chain, pass in the SLP node for the PHIs.
23886         (vect_create_epilog_for_reduction): Get the SLP instance as
23887         arg and adjust.
23888         (vectorizable_reduction): Get the SLP instance as arg.
23889         During analysis remember the SLP node with the PHIs in the
23890         instance.  Simplify getting at the vectorized reduction PHIs.
23891         * tree-vect-slp.c (vect_slp_analyze_node_operations): Pass
23892         through SLP instance.
23893         (vect_slp_analyze_operations): Likewise.
23894         * tree-vect-stms.c (vect_analyze_stmt): Likewise.
23895         (vect_transform_stmt): Likewise.
23897 2017-07-20  Tom de Vries  <tom@codesourcery.com>
23899         PR tree-optimization/81489
23900         * gimple-ssa-isolate-paths.c (find_implicit_erroneous_behavior): Move
23901         read of phi arg location to before loop that modifies phi.
23903 2017-07-20  Naveen H.S  <Naveen.Hurugalawadi@cavium.com>
23905         * match.pd (((m1 >/</>=/<= m2) * d -> (m1 >/</>=/<= m2) ? d : 0):
23906         New pattern.
23908 2017-07-19  Jan Hubicka  <hubicka@ucw.cz>
23910         PR middle-end/81331
23911         * except.c (execute): Fix ordering issue.
23913 2018-07-19  Segher Boessenkool  <segher@kernel.crashing.org>
23915         PR rtl-optimization/81423
23916         * combine.c (make_compound_operation_int): Don't try to optimize
23917         the AND of a SUBREG of an LSHIFTRT if that SUBREG is paradoxical.
23919 2017-07-19  Segher Boessenkool  <segher@kernel.crashing.org>
23921         PR rtl-optimization/81423
23922         * simplify-rtx.c (simplify_truncation): Handle truncating an IOR
23923         with a constant that is -1 in the truncated to mode.
23925 2017-07-19  Jan Hubicka  <hubicka@ucw.cz>
23927         * predict.c (propagate_unlikely_bbs_forward): Break out from ...
23928         (determine_unlikely_bbs): ... here.
23929         * predict.h (propagate_unlikely_bbs_forward): Declare.
23930         * cfgexpand.c (pass_expand::execute): Use it.
23931         * bb-reorder.c (sanitize_hot_paths): Do not consider known to be
23932         unlikely edges.
23933         (find_rarely_executed_basic_blocks_and_crossing_edges): Use
23934         propagate_unlikely_bbs_forward.
23936 2017-07-19  Jan Hubicka  <hubicka@ucw.cz>
23938         PR middle-end/81331
23939         * except.c (maybe_add_nop_after_section_switch): New function.
23940         (execute): Use it.
23942 2017-07-19  Tom de Vries  <tom@codesourcery.com>
23944         * gimple.h (gimple_phi_set_arg): Make assert more strict.
23946 2017-07-19  Tom de Vries  <tom@codesourcery.com>
23948         * gimple.h (gimple_phi_arg): Make assert more strict.
23950 2017-07-19  Steven Munroe  <munroesj@gcc.gnu.org>
23952         * config.gcc (powerpc*-*-*): Add mmintrin.h.
23953         * config/rs6000/mmintrin.h: New file.
23954         * config/rs6000/x86intrin.h [__ALTIVEC__]: Include mmintrin.h.
23956 2017-07-19  Jakub Jelinek  <jakub@redhat.com>
23958         PR tree-optimization/81346
23959         * match.pd: Optimize (X - 1U) <= INT_MAX-1U into (int) X > 0.
23961 2017-07-19  Tom de Vries  <tom@codesourcery.com>
23963         * config/nvptx/nvptx.md (VECIM): Add V2DI.
23965 2017-07-19  Tom de Vries  <tom@codesourcery.com>
23967         * config/nvptx/nvptx-modes.def: Add V2DImode.
23968         * config/nvptx/nvptx-protos.h (nvptx_data_alignment): Declare.
23969         * config/nvptx/nvptx.c (nvptx_ptx_type_from_mode): Handle V2DImode.
23970         (nvptx_output_mov_insn): Handle lack of mov.b128.
23971         (nvptx_print_operand): Handle 'H' and 'L' codes.
23972         (nvptx_vector_mode_supported): Allow V2DImode.
23973         (nvptx_preferred_simd_mode): New function.
23974         (nvptx_data_alignment): New function.
23975         (TARGET_VECTORIZE_PREFERRED_SIMD_MODE): Redefine to
23976         nvptx_preferred_simd_mode.
23977         * config/nvptx/nvptx.h (STACK_BOUNDARY, BIGGEST_ALIGNMENT): Change from
23978         64 to 128 bits.
23979         (DATA_ALIGNMENT): Define.  Set to nvptx_data_alignment.
23981 2017-07-19  Tom de Vries  <tom@codesourcery.com>
23983         * config/nvptx/nvptx-modes.def: New file.  Add V2SImode.
23984         * config/nvptx/nvptx.c (nvptx_ptx_type_from_mode): Handle V2SImode.
23985         (nvptx_vector_mode_supported): New function.  Allow V2SImode.
23986         (TARGET_VECTOR_MODE_SUPPORTED_P): Redefine to nvptx_vector_mode_supported.
23987         * config/nvptx/nvptx.md (VECIM): New mode iterator. Add V2SI.
23988         (mov<VECIM>_insn): New define_insn.
23989         (define_expand "mov<VECIM>): New define_expand.
23991 2017-07-19  Tom de Vries  <tom@codesourcery.com>
23993         * config/nvptx/nvptx.c (nvptx_print_operand): Handle v2 vector mode.
23995 2017-07-19  Jakub Jelinek  <jakub@redhat.com>
23997         PR tree-optimization/81346
23998         * fold-const.h (fold_div_compare, range_check_type): Declare.
23999         * fold-const.c (range_check_type): New function.
24000         (build_range_check): Use range_check_type.
24001         (fold_div_compare): No longer static, rewritten into
24002         a match.pd helper function.
24003         (fold_comparison): Don't call fold_div_compare here.
24004         * match.pd (X / C1 op C2): New optimization using fold_div_compare
24005         as helper function.
24007 2017-07-19  Nathan Sidwell  <nathan@acm.org>
24009         * tree.h (TYPE_MINVAL, TYPE_MAXVAL): Rename to ...
24010         (TYPE_MIN_VALUE_RAW, TYPE_MAX_VALUE_RAW): ... these.
24011         * tree.c (find_decls_types_r, verify_type): Use
24012         TYPE_{MIN,MAX}_VALUE_RAW.
24013         * lto-streamer-out.c (DFS::DFS_write_tree_body): Likewise.
24014         (hash_tree): Likewise.
24015         * tree-streamer-in.c (lto_input_ts_type_non_common_tree_pointers):
24016         Likewise.
24017         * tree-streamer-out.c (write_ts_type_non_common_tree_pointers):
24018         Likewise.
24020 2017-07-18  Tom de Vries  <tom@codesourcery.com>
24022         PR middle-end/81464
24023         * omp-expand.c (expand_omp_for_static_chunk): Handle
24024         equal-argument loop exit phi.
24026 2017-07-18  Uros Bizjak  <ubizjak@gmail.com>
24028         PR target/81471
24029         * config/i386/i386.md (rorx_immediate_operand): New mode attribute.
24030         (*bmi2_rorx<mode>3_1): Use rorx_immediate_operand as
24031         operand 2 predicate.
24032         (*bmi2_rorxsi3_1_zext): Use const_0_to_31_operand as
24033         operand 2 predicate.
24034         (ror,rol -> rorx splitters): Use const_int_operand as
24035         operand 2 predicate.
24037 2017-06-18  Richard Biener  <rguenther@suse.de>
24039         PR tree-optimization/81410
24040         * tree-vect-stmts.c (vectorizable_load): Properly adjust for
24041         the gap in the ! slp_perm SLP case after each group.
24043 2017-07-18  Jan Hubicka  <hubicka@ucw.cz>
24045         PR middle-end/81463
24046         * cfgloopmanip.c (scale_loop_profile): Watch out for zero frequency
24047         again.
24049 2017-07-18  Jan Hubicka  <hubicka@ucw.cz>
24051         PR middle-end/81462
24052         * predict.c (set_even_probabilities): Cleanup; do not affect
24053         probabilities that are already known.
24054         (combine_predictions_for_bb): Call even when count is set.
24056 2017-07-18  Nathan Sidwell  <nathan@acm.org>
24058         * tree-parloops.c (try_transform_to_exit_first_loop_alt): Use
24059         TYPE_MAX_VALUE.
24061 2017-07-18  Bin Cheng  <bin.cheng@arm.com>
24063         PR target/81408
24064         * tree-ssa-loop-niter.c (number_of_iterations_exit): Dump missed
24065         optimization for loop niter analysis.
24067 2017-07-18  Georg-Johann Lay  <avr@gjlay.de>
24069         PR target/81473
24070         * config/avr/avr.c (avr_optimize_casesi): Don't use
24071         INT8_MIN, INT8_MAX, UINT8_MAX, INT16_MIN, INT16_MAX, UINT16_MAX.
24073 2017-07-18  Robin Dapp  <rdapp@linux.vnet.ibm.com>
24075         * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Remove
24076         body_cost_vec from _vect_peel_extended_info.
24077         (vect_peeling_hash_get_lowest_cost): Do not set body_cost_vec.
24078         (vect_peeling_hash_choose_best_peeling): Remove body_cost_vec and
24079         npeel.
24081 2017-07-18  Bin Cheng  <bin.cheng@arm.com>
24083         * config/arm/arm.c (emit_unlikely_jump): Remove unused var.
24085 2017-07-18  Richard Biener  <rguenther@suse.de>
24087         PR tree-optimization/80620
24088         PR tree-optimization/81403
24089         * tree-ssa-pre.c (phi_translate_1): Clear range and points-to
24090         info when re-using a VN table entry.
24092 2017-07-18  Richard Biener  <rguenther@suse.de>
24094         PR tree-optimization/81418
24095         * tree-vect-loop.c (vectorizable_reduction): Properly compute
24096         vectype_in.  Verify that with lane-reducing reduction operations
24097         we have a single def-use cycle.
24099 2017-07-17  Carl Love  <cel@us.ibm.com>
24101         Revert commit r249424   2017-06-20  Carl Love  <cel@us.ibm.com>
24103         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
24104         ALTIVEC_BUILTIN_VMULESW, ALTIVEC_BUILTIN_VMULEUW,
24105         ALTIVEC_BUILTIN_VMULOSW, ALTIVEC_BUILTIN_VMULOUW entries.
24106         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin,
24107         builtin_function_type): Add ALTIVEC_BUILTIN_* case statements.
24108         * config/rs6000/altivec.md (MVULEUW, VMULESW, VMULOUW,
24109         VMULOSW): New enum "unspec" values.
24110         (vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
24111         vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si,
24112         altivec_vmuleuw, altivec_vmulesw, altivec_vmulouw,
24113         altivec_vmulosw): New patterns.
24114         * config/rs6000/rs6000-builtin.def (VMLEUW, VMULESW, VMULOUW,
24115         VMULOSW): Add definitions.
24117 2017-07-17  Uros Bizjak  <ubizjak@gmail.com>
24119         * config/alpha/alpha.c: Include predict.h.
24121 2017-07-17  Yury Gribov  <tetra2005@gmail.com>
24123         * tree-vrp.c (compare_assert_loc): Fix comparison function
24124         to return predictable results.
24126 2017-07-17  Claudiu Zissulescu  <claziss@synopsys.com>
24128         * config/arc/arc.md (adddi3): Remove support for mexpand-adddi
24129         option.
24130         (subdi3): Likewise.
24131         * config/arc/arc.opt (mexpand-adddi): Deprecate it.
24132         * doc/invoke.texi (mexpand-adddi): Update text.
24134 2017-07-17  Claudiu Zissulescu  <claziss@synopsys.com>
24136         * config/arc/arc.md (clzsi2): Expand to an arc_clzsi2 instruction
24137         that also clobbers the CC register. The old expand code is moved
24138         to ...
24139         (*arc_clzsi2): ... here.
24140         (ctzsi2): Expand to an arc_ctzsi2 instruction that also clobbers
24141         the CC register. The old expand code is moved to ...
24142         (arc_ctzsi2): ... here.
24144 2017-07-17  Claudiu Zissulescu  <claziss@synopsys.com>
24146         * config/arc/arc.opt (mindexed-loads): Use initial value
24147         TARGET_INDEXED_LOADS_DEFAULT.
24148         (mauto-modify-reg): Use initial value
24149         TARGET_AUTO_MODIFY_REG_DEFAULT.
24150         * config/arc/elf.h (TARGET_INDEXED_LOADS_DEFAULT): Define.
24151         (TARGET_AUTO_MODIFY_REG_DEFAULT): Likewise.
24152         * config/arc/linux.h (TARGET_INDEXED_LOADS_DEFAULT): Define.
24153         (TARGET_AUTO_MODIFY_REG_DEFAULT): Likewise.
24155 2017-07-17  Martin Liska  <mliska@suse.cz>
24157         PR sanitizer/81302
24158         * opts.c (finish_options): Do not allow -fgnu-tm
24159         w/ -fsanitize={kernel-,}address.  Say sorry.
24161 2017-07-17  Bin Cheng  <bin.cheng@arm.com>
24163         PR target/81369
24164         * tree-loop-distribution.c (classify_partition): Only assert on
24165         numer of iterations.
24166         (merge_dep_scc_partitions): Delete prameter.  Update function call.
24167         (distribute_loop): Remove code handling loop with unknown niters.
24168         (pass_loop_distribution::execute): Skip loop with unknown niters.
24170 2017-07-17  Bin Cheng  <bin.cheng@arm.com>
24172         PR target/81369
24173         * tree-loop-distribution.c (merge_dep_scc_partitions): Sink call to
24174         function sort_partitions_by_post_order.
24176 2017-07-17  Bin Cheng  <bin.cheng@arm.com>
24178         PR tree-optimization/81374
24179         * tree-loop-distribution.c (pass_loop_distribution::execute): Record
24180         the max index of basic blocks, rather than number of basic blocks.
24182 2017-07-17  Claudiu Zissulescu  <claziss@synopsys.com>
24184         * config/arc/arc-protos.h (arc_legitimate_pc_offset_p): Remove
24185         proto.
24186         (arc_legitimate_pic_operand_p): Likewise.
24187         * config/arc/arc.c (arc_legitimate_pic_operand_p): Remove
24188         function.
24189         (arc_needs_pcl_p): Likewise.
24190         (arc_legitimate_pc_offset_p): Likewise.
24191         (arc_legitimate_pic_addr_p): Remove LABEL_REF case, as this
24192         function is also used in constrains.md.
24193         (arc_legitimate_constant_p): Use arc_legitimate_pic_addr_p to
24194         validate pic constants. Handle CONST_INT, CONST_DOUBLE, MINUS and
24195         PLUS.  Only return true/false in known cases, otherwise assert.
24196         (arc_legitimate_address_p): Remove arc_legitimate_pic_addr_p as it
24197         is already called in arc_legitimate_constant_p.
24198         * config/arc/arc.h (CONSTANT_ADDRESS_P): Consider also LABEL for
24199         pic addresses.
24200         (LEGITIMATE_PIC_OPERAND_P): Use
24201         arc_raw_symbolic_reference_mentioned_p function.
24202         * config/arc/constraints.md (Cpc): Use arc_legitimate_pic_addr_p
24203         function.
24204         (Cal): Likewise.
24205         (C32): Likewise.
24207 2017-07-17  Claudiu Zissulescu  <claziss@synopsys.com>
24208         Andrew Burgess  <andrew.burgess@embecosm.com>
24210         * config/arc/arc-protos.h (arc_compute_function_type): Change prototype.
24211         (arc_return_address_register): New function.
24212         * config/arc/arc.c (arc_handle_fndecl_attribute): New function.
24213         (arc_handle_fndecl_attribute): Add naked attribute.
24214         (TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS): Define.
24215         (TARGET_WARN_FUNC_RETURN): Likewise.
24216         (arc_allocate_stack_slots_for_args): New function.
24217         (arc_warn_func_return): Likewise.
24218         (machine_function): Change type fn_type.
24219         (arc_compute_function_type): Consider new naked function type,
24220         change function return type.
24221         (arc_must_save_register): Adapt to handle new
24222         arc_compute_function_type's return type.
24223         (arc_expand_prologue): Likewise.
24224         (arc_expand_epilogue): Likewise.
24225         (arc_return_address_regs): Delete.
24226         (arc_return_address_register): New function.
24227         (arc_epilogue_uses): Use above function.
24228         * config/arc/arc.h (arc_return_address_regs): Delete prototype.
24229         (arc_function_type): Change encoding, add naked type.
24230         (ARC_INTERRUPT_P): Change to handle the new encoding.
24231         (ARC_FAST_INTERRUPT_P): Likewise.
24232         (ARC_NORMAL_P): Define.
24233         (ARC_NAKED_P): Likewise.
24234         (arc_compute_function_type): Delete prototype.
24235         * config/arc/arc.md (in_ret_delay_slot): Use
24236         arc_return_address_register function.
24237         (simple_return): Likewise.
24238         (p_return_i): Likewise.
24240 2017-07-17  Jakub Jelinek  <jakub@redhat.com>
24242         PR tree-optimization/81428
24243         * match.pd (X / X -> one): Don't optimize _Fract divisions, as 1
24244         can't be built for those types.
24246 2017-07-17  Georg-Johann Lay  <avr@gjlay.de>
24248         Remove stuff dead since r239246.
24250         * config/avr/avr-arch.h (avr_inform_devices): Remove dead proto.
24251         * config/avr/avr-devices.c (mcu_name, comparator, avr_mcus_str)
24252         (avr_inform_devices): Remove dead stuff.
24254 2017-07-17  Tamar Christina  <tamar.christina@arm.com>
24256         * config/arm/arm_neon.h: Fix softp typo.
24258 2017-07-17  Jakub Jelinek  <jakub@redhat.com>
24260         PR tree-optimization/81365
24261         * tree-ssa-phiprop.c (propagate_with_phi): When considering hoisting
24262         aggregate moves onto bb predecessor edges, make sure there are no
24263         loads that could alias the lhs in between the start of bb and the
24264         loads from *phi.
24266 2017-07-17  Georg-Johann Lay  <avr@gjlay.de>
24268         PR 80929
24269         * config/avr/avr.c (avr_mul_highpart_cost): New static function.
24270         (avr_rtx_costs_1) [TRUNCATE]: Use it to compute mul_highpart cost.
24271         [LSHIFTRT, outer_code = TRUNCATE]: Same.
24273 2017-07-17  Jakub Jelinek  <jakub@redhat.com>
24275         PR tree-optimization/81396
24276         * tree-ssa-math-opts.c (struct symbolic_number): Add n_ops field.
24277         (init_symbolic_number): Initialize it to 1.
24278         (perform_symbolic_merge): Add n_ops from both operands into the new
24279         n_ops.
24280         (find_bswap_or_nop): Don't consider n->n == cmpnop computations
24281         without base_addr as useless if they need more than one operation.
24282         (bswap_replace): Handle !bswap case for NULL base_addr.
24284 2017-07-17  Tom de Vries  <tom@codesourcery.com>
24286         PR target/81069
24287         * config/nvptx/nvptx.c (nvptx_single): Insert diverging branch as late
24288         as possible.
24290 2017-07-17  Sebastian Huber  <sebastian.huber@embedded-brains.de>
24292         * config/sparc/rtemself.h (TARGET_OS_CPP_BUILTINS): Add
24293         conditional builtin define __FIX_LEON3FT_B2BST.
24295 2017-07-17  Daniel Cederman  <cederman@gaisler.com>
24297         * config/sparc/t-rtems: Add mfix-gr712rc multilibs. Replace
24298         MULTILIB_EXCEPTIONS with MULTILIB_REQUIRED. Match -mfix-gr712rc
24299         with -mfix-ut700.
24301 2017-07-16  Eric Botcazou  <ebotcazou@adacore.com>
24303         PR rtl-optimization/81424
24304         * optabs.c (prepare_cmp_insn): Use copy_to_reg instead of force_reg
24305         to remove potential trapping from operands if -fnon-call-exceptions.
24307 2017-07-16  Jan Hubicka  <hubicka@ucw.cz>
24309         * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Use
24310         profile_proability for scalling.
24311         * scale_profile_for_vect_loop.c (scale_profile_for_vect_loop): Likewise.
24313 2017-07-16  Jan Hubicka  <hubicka@ucw.cz>
24315         * cgraph.c (cgraph_edge::redirect_call_stmt_to_caller): Cleanup.
24317 2017-07-16  Jan Hubicka  <hubicka@ucw.cz>
24319         * cfgloopmanip.c (scale_loop_profile): Avoid use of REG_BR_PROB_BASE
24320         fixpoint arithmetics.
24322 2017-07-16  Jan Hubicka  <hubicka@ucw.cz>
24324         * tree-ssa-loop-unswitch.c (hoist_guard): Avoid use of REG_BR_PROB_BASE
24325         fixpoint arithmetics.
24327 2017-07-16  Jan Hubicka  <hubicka@ucw.cz>
24329         * asan.c (create_cond_insert_point): Avoid use of REG_BR_PROB_BASE
24330         fixpoint arithmetics.
24332 2017-07-16  Jan Hubicka  <hubicka@ucw.cz>
24334         * profile-count.h (profile_probability::from_reg_br_prob_note,
24335         profile_probability::to_reg_br_prob_note): New functions.
24336         * doc/rtl.texi (REG_BR_PROB_NOTE): Update documentation.
24337         * reg-notes.h (REG_BR_PROB, REG_BR_PRED): Update docs.
24338         * predict.c (probability_reliable_p): Update.
24339         (edge_probability_reliable_p): Update.
24340         (br_prob_note_reliable_p): Update.
24341         (invert_br_probabilities): Update.
24342         (add_reg_br_prob_note): New function.
24343         (combine_predictions_for_insn): Update.
24344         * asan.c (asan_clear_shadow): Update.
24345         * cfgbuild.c (compute_outgoing_frequencies): Update.
24346         * cfgrtl.c (force_nonfallthru_and_redirect): Update.
24347         (update_br_prob_note): Update.
24348         (rtl_verify_edges): Update.
24349         (purge_dead_edges): Update.
24350         (fixup_reorder_chain): Update.
24351         * emit-rtl.c (try_split): Update.
24352         * ifcvt.c (cond_exec_process_insns): Update.
24353         (cond_exec_process_if_block): Update.
24354         (dead_or_predicable): Update.
24355         * internal-fn.c (expand_addsub_overflow): Update.
24356         (expand_neg_overflow): Update.
24357         (expand_mul_overflow): Update.
24358         * loop-doloop.c (doloop_modify): Update.
24359         * loop-unroll.c (compare_and_jump_seq): Update.
24360         * optabs.c (emit_cmp_and_jump_insn_1): Update.
24361         * predict.h: Update.
24362         * reorg.c (mostly_true_jump): Update.
24363         * rtl.h: Update.
24364         * config/aarch64/aarch64.c (aarch64_emit_unlikely_jump): Update.
24365         * config/alpha/alpha.c (emit_unlikely_jump): Update.
24366         * config/arc/arc.c: (emit_unlikely_jump): Update.
24367         * config/arm/arm.c: (emit_unlikely_jump): Update.
24368         * config/bfin/bfin.c (cbranch_predicted_taken_p): Update.
24369         * config/frv/frv.c (frv_print_operand_jump_hint): Update.
24370         * config/i386/i386.c (ix86_expand_split_stack_prologue): Update.
24371         (ix86_print_operand): Update.
24372         (ix86_split_fp_branch): Update.
24373         (predict_jump): Update.
24374         * config/ia64/ia64.c (ia64_print_operand): Update.
24375         * config/mmix/mmix.c (mmix_print_operand): Update.
24376         * config/powerpcspe/powerpcspe.c (output_cbranch): Update.
24377         (rs6000_expand_split_stack_prologue): Update.
24378         * config/rs6000/rs6000.c: Update.
24379         * config/s390/s390.c (s390_expand_vec_strlen): Update.
24380         (s390_expand_vec_movstr): Update.
24381         (s390_expand_cs_tdsi): Update.
24382         (s390_expand_split_stack_prologue): Update.
24383         * config/sh/sh.c (sh_print_operand): Update.
24384         (expand_cbranchsi4): Update.
24385         (expand_cbranchdi4): Update.
24386         * config/sparc/sparc.c (output_v9branch): Update.
24387         * config/spu/spu.c (get_branch_target): Update.
24388         (ea_load_store_inline): Update.
24389         * config/tilegx/tilegx.c (cbranch_predicted_p): Update.
24390         * config/tilepro/tilepro.c: Update.
24392 2017-07-16  Eric Botcazou  <ebotcazou@adacore.com>
24394         * gimplify.c (mostly_copy_tree_r): Revert latest change.
24395         (gimplify_save_expr): Likewise.
24397 2017-07-07  Jan Hubicka  <hubicka@ucw.cz>
24399         * ipa-visibility.c (function_and_variable_visibility): Fix pasto.
24401 2017-07-07  Jan Hubicka  <hubicka@ucw.cz>
24403         * ipa-fnsummary.c (pass_data_ipa_fn_summary): Use
24404         TV_IPA_FNSUMMARY.
24405         * timevar.def (TV_IPA_FNSUMMARY): Define.
24407 2017-07-16  Daniel Cederman  <cederman@gaisler.com>
24409         * config/sparc/sparc.md (divdf3_fix): Add NOP to prevent back
24410         to back store errata sensitive sequence from being generated.
24411         (sqrtdf2_fix): Likewise.
24413 2017-07-07  Jan Hubicka  <hubicka@ucw.cz>
24415         * tree-ssa-threadupdate.c (compute_path_counts,
24416         update_joiner_offpath_counts): Use profile_probability.
24418 2017-07-15  Thomas Preud'homme  <thomas.preudhomme@arm.com>
24420         Revert:
24421         2017-07-14  Thomas Preud'homme  <thomas.preudhomme@arm.com>
24423         * config/arm/arm-c.c (arm_cpu_builtins): Define
24424         __ARM_FEATURE_NUMERIC_MAXMIN solely based on TARGET_VFP5.
24426 2017-07-14  Kelvin Nilsen  <kelvin@gcc.gnu.org>
24428         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
24429         array entries to represent __ieee128 versions of the
24430         scalar_test_data_class, scalar_test_neg, scalar_extract_exp,
24431         scalar_extract_sig, and scalar_insert_exp built-in functions.
24432         (altivec_resolve_overloaded_builtin): Add special case handling
24433         for the __builtin_scalar_insert_exp function, as represented by
24434         the P9V_BUILTIN_VEC_VSIEDP constant.
24435         * config/rs6000/rs6000-builtin.def (VSEEQP): Add scalar extract
24436         exponent support for __ieee128 argument.
24437         (VSESQP): Add scalar extract signature support for __ieee128
24438         argument.
24439         (VSTDCNQP): Add scalar test negative support for __ieee128
24440         argument.
24441         (VSIEQP): Add scalar insert exponent support for __int128 argument
24442         with __ieee128 result.
24443         (VSIEQPF): Add scalar insert exponent support for __ieee128
24444         argument with __ieee128 result.
24445         (VSTDCQP): Add scalar test data class support for __ieee128
24446         argument.
24447         (VSTDCNQP): Add overload support for scalar test negative with
24448         __ieee128 argument.
24449         (VSTDCQP): Add overload support for scalar test data class
24450         __ieee128 argument.
24451         * config/rs6000/vsx.md (UNSPEC_VSX_SXSIG) Replace
24452         UNSPEC_VSX_SXSIGDP.
24453         (UNSPEC_VSX_SIEXPQP): New constant.
24454         (xsxexpqp): New insn for VSX scalar extract exponent quad
24455         precision.
24456         (xsxsigqp): New insn for VSX scalar extract significand quad
24457         precision.
24458         (xsiexpqpf): New insn for VSX scalar insert exponent quad
24459         precision with floating point argument.
24460         (xststdcqp): New expand for VSX scalar test data class quad
24461         precision.
24462         (xststdcnegqp): New expand for VSX scalar test negative quad
24463         precision.
24464         (xststdcqp): New insn to match expansions for VSX scalar test data
24465         class quad precision and VSX scalar test negative quad precision.
24466         * config/rs6000/rs6000.c (rs6000_expand_binop_builtin): Add
24467         special case operand checking to enforce that second operand of
24468         VSX scalar test data class with quad precision argument is a 7-bit
24469         unsigned literal.
24470         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Add
24471         prototypes and descriptions of __ieee128 versions of
24472         scalar_extract_exp, scalar_extract_sig, scalar_insert_exp,
24473         scalar_test_data_class, and scalar_test_neg built-in functions.
24475 2016-07-14  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
24477         PR tree-optimization/81162
24478         * gimple-ssa-strength-reduction.c (replace_mult_candidate): Don't
24479         replace a negate with an add.
24481 2017-07-14  James Greenhalgh  <james.greenhalgh@arm.com>
24483         * doc/invoke.texi (arm/-mcpu): Document +crypto.
24485 2017-07-14  Thomas Preud'homme  <thomas.preudhomme@arm.com>
24487         * config/arm/arm-c.c (arm_cpu_builtins): Define
24488         __ARM_FEATURE_NUMERIC_MAXMIN solely based on TARGET_VFP5.
24490 2017-07-14  Thomas Preud'homme  <thomas.preudhomme@arm.com>
24492         * config/arm/arm-cpus.in (cortex-r52): Add new entry.
24493         (armv8-r): Set ARM Cortex-R52 as default CPU.
24494         * config/arm/arm-tables.opt: Regenerate.
24495         * config/arm/arm-tune.md: Regenerate.
24496         * config/arm/driver-arm.c (arm_cpu_table): Add entry for ARM
24497         Cortex-R52.
24498         * doc/invoke.texi: Mention -mtune=cortex-r52 and availability of fp.dp
24499         extension for -mcpu=cortex-r52.
24501 2017-07-14  Thomas Preud'homme  <thomas.preudhomme@arm.com>
24503         * config/arm/arm-isa.h (isa_bit_FP_ARMv8): Delete enumerator.
24504         (ISA_FP_ARMv8): Define as ISA_FPv5 and ISA_FP_D32.
24505         * config/arm/arm-cpus.in (armv8-r): Define fp.sp as enabling FPv5.
24506         (fp-armv8): Define it as FP_ARMv8 only.
24507         config/arm/arm.h (TARGET_FPU_ARMV8): Delete.
24508         (TARGET_VFP_FP16INST): Define using TARGET_VFP5 rather than
24509         TARGET_FPU_ARMV8.
24510         config/arm/arm.c (arm_rtx_costs_internal): Replace checks against
24511         TARGET_FPU_ARMV8 by checks against TARGET_VFP5.
24512         * config/arm/arm-builtins.c (arm_builtin_vectorized_function): Define
24513         first ARM_CHECK_BUILTIN_MODE definition using TARGET_VFP5 rather
24514         than TARGET_FPU_ARMV8.
24515         * config/arm/arm-c.c (arm_cpu_builtins): Likewise for
24516         __ARM_FEATURE_NUMERIC_MAXMIN macro definition.
24517         * config/arm/arm.md (cmov<mode>): Condition on TARGET_VFP5 rather than
24518         TARGET_FPU_ARMV8.
24519         * config/arm/neon.md (neon_vrint): Likewise.
24520         (neon_vcvt): Likewise.
24521         (neon_<fmaxmin_op><mode>): Likewise.
24522         (<fmaxmin><mode>3): Likewise.
24523         * config/arm/vfp.md (l<vrint_pattern><su_optab><mode>si2): Likewise.
24524         * config/arm/predicates.md (arm_cond_move_operator): Check against
24525         TARGET_VFP5 rather than TARGET_FPU_ARMV8 and fix spacing.
24527 2017-07-14  Jackson Woodruff  <jackson.woodruff@arm.com>
24529         * config/aarch64/aarch64.c (aarch64_print_operand): Move comments
24530         to top of function.
24532 2017-07-14  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
24534         * gimple-ssa-store-merging.c (clear_bit_region): Replace reference to
24535         loop in comment with memset.
24537 2017-07-14  Martin Liska  <mliska@suse.cz>
24539         * cfgexpand.c (expand_gimple_basic_block): Remove dead comment.
24540         * dwarf2out.c (is_java): Remove the function.
24541         (output_pubname): Remove usage of the function.
24542         (lower_bound_default): Remove usage of DW_LANG_Java.
24543         (gen_compile_unit_die): Likewise.
24544         * gcc.c: Remove compiler defaults for .java and .zip files.
24545         * gimple-expr.c (remove_suffix): Change as there's no longer
24546         extension than 4-letter one.
24547         * gimplify.c (mostly_copy_tree_r): Remove Java-special part.
24548         (gimplify_save_expr): Likewise.
24549         * ipa-utils.h (polymorphic_type_binfo_p): Remove the comment
24550         as it's possible even for other languages than Java.
24551         * langhooks.h (struct lang_hooks): Remove Java from a comment.
24552         * lto-opts.c (lto_write_options): Remove reference to Java.
24553         * opts.c (strip_off_ending): Update file extension handling.
24554         * tree-cfg.c (verify_gimple_call): Remove comment with Java.
24555         * tree-eh.c (lower_resx): Likewise.
24556         * tree.c (free_lang_data_in_type): Remove dead code.
24557         (find_decls_types_r): Likewise.
24558         (build_common_builtin_nodes): Remove Java from a comment.
24559         (verify_type): Remove dead code.
24560         * varasm.c (assemble_external): Remove Java from a comment.
24562 2017-07-14  Martin Liska  <mliska@suse.cz>
24564         * opts.c (finish_options): Add quotes.
24565         (common_handle_option): Likewise.
24567 2017-07-14  Martin Liska  <mliska@suse.cz>
24569         * dbxout.c (get_lang_number): Do not handle GNU Pascal.
24570         * dbxout.h (extern void dbxout_stab_value_internal_label_diff):
24571         Remove N_SO_PASCAL.
24572         * dwarf2out.c (lower_bound_default): Do not handle
24573         DW_LANG_Pascal83.
24574         (gen_compile_unit_die): Likewise.
24575         * gcc.c: Remove default extension binding for GNU Pascal.
24576         * stmt.c: Remove Pascal language from a comment.
24577         * xcoffout.c: Likewise.
24579 2017-07-13  David Malcolm  <dmalcolm@redhat.com>
24581         PR c/81405
24582         * diagnostic-show-locus.c (fixit_cmp): New function.
24583         (layout::layout): Sort m_fixit_hints.
24584         (column_range::column_range): Assert that the values are valid.
24585         (struct char_span): New struct.
24586         (correction::overwrite): New method.
24587         (struct source_line): New struct.
24588         (line_corrections::add_hint): Add assertions.  Reimplement memcpy
24589         calls in terms of classes source_line and char_span, and
24590         correction::overwrite.
24591         (selftest::test_overlapped_fixit_printing_2): New function.
24592         (selftest::diagnostic_show_locus_c_tests): Call it.
24594 2017-07-13  Will Schmidt  <will_schmidt@vnet.ibm.com>
24596         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Return
24597         early if there is no lhs.
24599 2017-07-13  Martin Liska  <mliska@suse.cz>
24601         * dwarf2out.c (gen_pointer_type_die): Remove dead code.
24602         (gen_reference_type_die): Likewise.
24603         * stor-layout.c: Remove Pascal-related comment.
24605 2017-07-13  Martin Liska  <mliska@suse.cz>
24607         * opts.c (finish_options): Add quotes to error messages.
24608         (parse_sanitizer_options): Likewise.
24610 2017-07-13  Thomas Preud'homme  <thomas.preudhomme@arm.com>
24612         * doc/invoke.texi (armv8-r): Document +fp.sp ARMv8-R extension.
24614 2017-07-13  Richard Earnshaw  <rearnsha@arm.com>
24616         * config/arm/vxworks.h (TARGET_ENDIAN_DEFAULT): Define.
24618 2017-07-13  Maxim Ostapenko  <m.ostapenko@samsung.com>
24620         * asan.c (asan_emit_allocas_unpoison): Use ptr_mode for arguments
24621         during expansion.
24622         * builtins.c (expand_asan_emit_allocas_unpoison): Likewise.
24624 2017-07-12  Michael Meissner  <meissner@linux.vnet.ibm.com>
24626         PR target/81193
24627         * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If GLIBC
24628         provides the hardware capability bits, define the macro
24629         __BUILTIN_CPU_SUPPORTS__.
24630         * config/rs6000/rs6000.c (cpu_expand_builtin): Generate a warning
24631         if GLIBC does not provide the hardware capability bits.  Add a
24632         gcc_unreachable call if the built-in cpu function is neither
24633         __builtin_cpu_is nor __builtin_cpu_supports.
24634         (rs6000_get_function_versions_dispatcher): Change the warning
24635         that an old GLIBC is used which does not export the capability
24636         bits to be an error.
24637         * doc/extend.texi (target_clones attribute): Document the
24638         restriction that GLIBC 2.23 or newer is needed on the PowerPC.
24639         (PowerPC built-in functions): Document that GLIBC 2.23 or newer is
24640         needed by __builtin_cpu_is and __builtin_cpu_supports.  Document
24641         the macros defined by GCC if the newer GLIBC is available.
24643 2017-07-12  Jeff Law  <law@redhat.com>
24645         * config/riscv/riscv.c: Remove unnecessary includes.  Reorder
24646         remaining includes slightly.
24647         * config/riscv/riscv-builtins.c: Include profile-count.h.
24649 2017-07-12  Georg-Johann Lay  <avr@gjlay.de>
24651         PR target/79883
24652         * config/avr/avr.c (avr_set_current_function): In diagnostic
24653         messages: Quote keywords and (parts of) identifiers.
24654         [WITH_AVRLIBC]: Warn for functions named "ISR", "SIGNAL" or
24655         "INTERUPT".
24657 2017-07-12  Carl Love  <cel@us.ibm.com>
24659         * config/rs6000/rs6000-c.c: Add support for built-in functions
24660         vector bool char vec_revb (vector bool char);
24661         vector bool short vec_revb (vector short char);
24662         vector bool int vec_revb (vector bool int);
24663         vector bool long long vec_revb (vector bool long long);
24664         * doc/extend.texi: Update the built-in documentation file for the
24665         new built-in functions.
24667 2017-07-12  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
24669         * config/s390/s390.md: Remove movcc splitter.
24671 2017-07-12  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
24673         * config/s390/s390.c (s390_rtx_costs): Return proper costs for
24674         load/store on condition.
24676 2017-07-12  Georg-Johann Lay  <avr@gjlay.de>
24678         PR target/81407
24679         * config/avr/avr.c (avr_encode_section_info)
24680         [progmem && !TREE_READONLY]: Error if progmem object needs
24681         constructing.
24683 2017-07-11  Michael Collison  <michael.collison@arm.com>
24685         * config/aarch64/aarch64-simd.md (aarch64_sub<mode>_compare0):
24686         New pattern.
24688 2017-07-11  Carl Love  <cel@us.ibm.com>
24690         * config/rs6000/rs6000-c.c: Add support for builtins
24691         vector unsigned int vec_parity_lsbb (vector signed int);
24692         vector unsigned int vec_parity_lsbb (vector unsigned int);
24693         vector unsigned __int128 vec_parity_lsbb (vector signed __int128);
24694         vector unsigned __int128 vec_parity_lsbb (vector unsigned __int128);
24695         vector unsigned long long vec_parity_lsbb (vector signed long long);
24696         vector unsigned long long vec_parity_lsbb (vector unsigned long long);
24697         * config/rs6000/rs6000-builtin.def (VPARITY_LSBB): Add BU_P9V_OVERLOAD1.
24698         * config/rs6000/altivec.h (vec_parity_lsbb): Add define.
24699         * doc/extend.texi: Update the built-in documentation file for the
24700         new built-in functions.
24702 2017-07-11  David Malcolm  <dmalcolm@redhat.com>
24704         * diagnostic-show-locus.c: Include "gcc-rich-location.h".
24705         (layout::m_primary_loc): New field.
24706         (layout::layout): Initialize new field.  Move location filtering
24707         logic from here to...
24708         (layout::maybe_add_location_range): ...this new method.  Add
24709         support for filtering to just the lines already specified by other
24710         locations.
24711         (layout::will_show_line_p): New method.
24712         (gcc_rich_location::add_location_if_nearby): New method.
24713         (selftest::test_add_location_if_nearby): New test function.
24714         (selftest::diagnostic_show_locus_c_tests): Call it.
24715         * gcc-rich-location.h (gcc_rich_location::add_location_if_nearby):
24716         New method.
24718 2017-07-11  Tom de Vries  <tom@codesourcery.com>
24720         * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG): New macro.
24721         (bb_first_real_insn): New function.
24722         (nvptx_single): Add extra initialization of broadcasted condition
24723         variables.
24725 2017-07-11  Nathan Sidwell  <nathan@acm.org>
24727         * dwarf2out.c (gen_member_die): Remove useless check for anon ctors.
24729 2017-07-11  Georg-Johann Lay  <avr@gjlay.de>
24731         * doc/extend.texi (AVR Function Attributes): Remove weblink to
24732         Binutils doc as TEXI will mess them up.
24733         * doc/invoke.texi (AVR Options): Same here.
24735 2017-07-11  Daniel Cederman  <cederman@gaisler.com>
24737         * config/sparc/sparc.opt (mfix-ut700): New option.
24738         (mfix-gr712rc): Likewise.
24739         (sparc_fix_b2bst): New variable.
24740         * doc/invoke.texi (SPARC options): Document them.
24741         (ARM options): Fix warnings.
24742         * config/sparc/sparc.c (sparc_do_work_around_errata): Insert NOP
24743         instructions to prevent sequences that can trigger the store-store
24744         errata for certain LEON3FT processors.
24745         (pass_work_around_errata::gate): Also test sparc_fix_b2bst.
24746         (sparc_option_override): Set sparc_fix_b2bst appropriately.
24747         * config/sparc/sparc.md (fix_b2bst): New attribute.
24748         (in_branch_delay): Prevent stores in delay slot if fix_b2bst.
24750 2017-07-10  Uros Bizjak  <ubizjak@gmail.com>
24752         PR target/81375
24753         * config/i386/i386.md (divsf3): Add TARGET_SSE to TARGET_SSE_MATH.
24754         (rcpps): Ditto.
24755         (*rsqrtsf2_sse): Ditto.
24756         (rsqrtsf2): Ditto.
24757         (div<mode>3): Macroize insn from divdf3 and divsf3
24758         using MODEF mode iterator.
24760 2017-07-10  Martin Sebor  <msebor@redhat.com>
24762         PR tree-optimization/80397
24763         * gimple-ssa-sprintf.c (format_integer): Use INTEGRAL_TYPE_P()
24764         instead of testing for equality to INTEGER_TYPE.
24766 2017-07-10  Vineet Gupta <vgupta@synopsys.com>
24768         * config.gcc: Remove uclibc from arc target spec.
24770 2017-07-10  Claudiu Zissulescu  <claziss@synopsys.com>
24772         * config/arc/arc.h (ADDITIONAL_REGISTER_NAMES): Define.
24774 2017-07-07  Jan Hubicka  <hubicka@ucw.cz>
24776         PR lto/80838
24777         * lto-wrapper.c (remove_option): New function.
24778         (merge_and_complain): Merge PIC/PIE options more realistically.
24780 2017-07-10  Georg-Johann Lay  <avr@gjlay.de>
24782         Better ISR prologues by supporting GASes __gcc_isr pseudo insn.
24784         PR target/20296
24785         PR target/81268
24786         * configure.ac [target=avr]: Add GAS check for -mgcc-isr.
24787         (HAVE_AS_AVR_MGCCISR_OPTION):  If so, AC_DEFINE it.
24788         * config.in: Regenerate.
24789         * configure: Regenerate.
24790         * doc/extend.texi (AVR Function Attributes) <no_gccisr>: Document it.
24791         * doc/invoke.texi (AVR Options) <-mgas-isr-prologues>: Document it.
24792         * config/avr/avr.opt (-mgas-isr-prologues): New option and...
24793         (TARGET_GASISR_PROLOGUES): ...target mask.
24794         * common/config/avr/avr-common.c
24795         (avr_option_optimization_table) [OPT_LEVELS_1_PLUS_NOT_DEBUG]:
24796         Set -mgas-isr-prologues.
24797         * config/avr/avr-passes.def (avr_pass_pre_proep): Add
24798         INSERT_PASS_BEFORE for it.
24799         * config/avr/avr-protos.h (make_avr_pass_pre_proep): New proto.
24800         * config/avr/avr.c (avr_option_override)
24801         [!HAVE_AS_AVR_MGCCISR_OPTION]: Unset TARGET_GASISR_PROLOGUES.
24802         (avr_no_gccisr_function_p, avr_hregs_split_reg): New static functions.
24803         (avr_attribute_table) <no_gccisr>: Add new function attribute.
24804         (avr_set_current_function) <is_no_gccisr>: Init machine field.
24805         (avr_pass_data_pre_proep, avr_pass_pre_proep): New pass data
24806         and rtl_opt_pass.
24807         (make_avr_pass_pre_proep): New function.
24808         (emit_push_sfr) <treg>: Add argument to function and use it
24809         instead of TMP_REG.
24810         (avr_expand_prologue) [machine->gasisr.maybe]: Emit gasisr insn
24811         and set machine->gasisr.yes.
24812         (avr_expand_epilogue) [machine->gasisr.yes]: Similar.
24813         (avr_asm_function_end_prologue) [machine->gasisr.yes]: Add
24814         __gcc_isr.n_pushed to .L__stack_usage.
24815         (TARGET_ASM_FINAL_POSTSCAN_INSN): Define to...
24816         (avr_asm_final_postscan_insn): ...this new static function.
24817         * config/avr/avr.h (machine_function)
24818         <is_no_gccisr, use_L__stack_usage>: New fields.
24819         <gasisr, gasisr.yes, gasisr.maybe, gasisr.regno>: New fields.
24820         * config/avr/avr.md (UNSPECV_GASISR): Add unspecv enum.
24821         (GASISR_Prologue, GASISR_Epilogue, GASISR_Done): New define_constants.
24822         (gasisr, *gasisr): New expander and insn.
24823         * config/avr/gen-avr-mmcu-specs.c (print_mcu)
24824         [HAVE_AS_AVR_MGCCISR_OPTION]: Print asm_gccisr spec.
24825         * config/avr/specs.h (ASM_SPEC) <asm_gccisr>: Add sub spec.
24827 2017-07-10  Richard Earnshaw  <rearnsha@arm.com>
24829         * config/arm/parsecpu.awk (gen_comm_data): Do not escape single quotes
24830         in quoted strings.
24832 2017-07-10  Georg-Johann Lay  <avr@gjlay.de>
24834         Move jump-tables out of .text again.
24836         PR target/81075
24837         * config/avr/avr.c (ASM_OUTPUT_ADDR_VEC_ELT): Remove function.
24838         (ASM_OUTPUT_ADDR_VEC): New function.
24839         (avr_adjust_insn_length) [JUMP_TABLE_DATA_P]: Return 0.
24840         (avr_final_prescan_insn) [avr_log.insn_addresses]: Dump
24841         INSN_ADDRESSes as asm comment.
24842         * config/avr/avr.h (JUMP_TABLES_IN_TEXT_SECTION): Adjust comment.
24843         (ASM_OUTPUT_ADDR_VEC_ELT): Remove define.
24844         (ASM_OUTPUT_ADDR_VEC): Define to avr_output_addr_vec.
24845         * config/avr/avr.md (*tablejump): Adjust comment.
24846         * config/avr/elf.h (ASM_OUTPUT_BEFORE_CASE_LABEL): Remove.
24847         * config/avr/avr-log.c (avr_log_set_avr_log) <insn_addresses>:
24848         New detail.
24849         * config/avr/avr-protos.h (avr_output_addr_vec_elt): Remove proto.
24850         (avr_output_addr_vec): New proto.
24851         (avr_log_t) <insn_addresses>: New field.
24853 2017-07-09  H.J. Lu  <hongjiu.lu@intel.com>
24855         PR target/81313
24856         * config/i386/i386.c (ix86_function_arg_advance): Set
24857         outgoing_args_on_stack to true if there are outgoing arguments
24858         on stack.
24859         (ix86_function_arg): Likewise.
24860         (ix86_get_drap_rtx): Use DRAP only if there are outgoing
24861         arguments on stack and ACCUMULATE_OUTGOING_ARGS is false.
24862         * config/i386/i386.h (machine_function): Add
24863         outgoing_args_on_stack.
24865 2017-07-09  Krister Walfridsson  <krister.walfridsson@gmail.com>
24867         * config.gcc (*-*-netbsd*): Remove check for NetBSD versions not
24868         supporting pthreds.
24869         * config/netbsd.h (NETBSD_LIBGCC_SPEC): Always enable pthreads.
24871 2017-07-08  Richard Sandiford  <richard.sandiford@linaro.org>
24873         * Makefile.in (HOOKS_H, RTL_BASE_H, FUNCTION_H, EXPR_H, REGS_H)
24874         (REAL_H): Remove $(MACHMODE_H).
24875         (FIXED_VALUE_H, TREE_CORE_H, CFGLOOP_H): Remove $(MACHMODE_H) and
24876         double-int.h.
24877         (CORETYPES_H): Add signop.h, wide-int.h, wide-int-print.h,
24878         $(MACHMODE_H) and double-int.h.
24879         (build/min-insn-modes.o): Depend on $(CORETYPES_H) rather than
24880         $(MACHMODE_H).
24881         (gengtype-state.o, gengtype.o, build/gengtype.o): Don't depend on
24882         double-int.h.
24884 2017-07-07  Andrew Pinski  <apinski@cavium.com>
24886         * config/aarch64/aarch64.c (aarch_macro_fusion_pair_p): Check
24887         prev_set and curr_set for AARCH64_FUSE_ALU_BRANCH.
24889 2017-07-07  Michael Meissner  <meissner@linux.vnet.ibm.com>
24891         * config/rs6000/rs6000.c (rs6000_get_function_versions_dispatcher):
24892         Add warning if GCC was not configured to link against a GLIBC that
24893         exports the hardware capability bits.
24894         (make_resolver_func): Make resolver function private and not a
24895         COMDAT function.  Create the name with clone_function_name instead
24896         of make_unique_name.
24898         PR target/81348
24899         * config/rs6000/rs6000.md (HI sign_extend splitter): Use the
24900         correct operand in doing the split.
24902 2017-07-07 Carl Love  <cel@us.ibm.com>
24904         * config/rs6000/rs6000-c: Add support for built-in function
24905         vector unsigned short vec_pack_to_short_fp32 (vector float,
24906                                                       vector float).
24907         * config/rs6000/rs6000-builtin.def (CONVERT_4F32_8I16): Add
24908         BU_P9V_AV_2 and BU_P9V_OVERLOAD_2 definitions.
24909         * config/rs6000/altivec.h (vec_pack_to_short_fp32): Add define.
24910         * config/rs6000/altivec.md(UNSPEC_CONVERT_4F32_8I16): Add UNSPEC.
24911         (convert_4f32_8i16): Add define_expand.
24912         * doc/extend.texi: Update the built-in documentation file for the
24913         new built-in function.
24915 2017-07-07  Jose E. Marchesi  <jose.marchesi@oracle.com>
24917         * config/sparc/m8.md: New file.
24918         * config/sparc/sparc.md: Include m8.md.
24920 2017-07-07  Jose E. Marchesi  <jose.marchesi@oracle.com>
24922         * config/sparc/sparc.opt: New option -mvis4b.
24923         * config/sparc/sparc.c (dump_target_flag_bits): Handle MASK_VIS4B.
24924         (sparc_option_override): Handle VIS4B.
24925         (enum sparc_builtins): Define
24926         SPARC_BUILTIN_DICTUNPACK{8,16,32},
24927         SPARC_BUILTIN_FPCMP{LE,GT,EQ,NE}{8,16,32}SHL,
24928         SPARC_BUILTIN_FPCMPU{LE,GT}{8,16,32}SHL,
24929         SPARC_BUILTIN_FPCMPDE{8,16,32}SHL and
24930         SPARC_BUILTIN_FPCMPUR{8,16,32}SHL.
24931         (check_constant_argument): New function.
24932         (sparc_vis_init_builtins): Define builtins
24933         __builtin_vis_dictunpack{8,16,32},
24934         __builtin_vis_fpcmp{le,gt,eq,ne}{8,16,32}shl,
24935         __builtin_vis_fpcmpu{le,gt}{8,16,32}shl,
24936         __builtin_vis_fpcmpde{8,16,32}shl and
24937         __builtin_vis_fpcmpur{8,16,32}shl.
24938         (sparc_expand_builtin): Check that the constant operands to
24939         __builtin_vis_fpcmp*shl and _builtin_vis_dictunpack* are indeed
24940         constant and in range.
24941         * config/sparc/sparc-c.c (sparc_target_macros): Handle
24942         TARGET_VIS4B.
24943         * config/sparc/sparc.h (SPARC_IMM2_P): Define.
24944         (SPARC_IMM5_P): Likewise.
24945         * config/sparc/sparc.md (cpu_feature): Add new feagure "vis4b".
24946         (enabled): Handle vis4b.
24947         (UNSPEC_DICTUNPACK): New unspec.
24948         (UNSPEC_FPCMPSHL): Likewise.
24949         (UNSPEC_FPUCMPSHL): Likewise.
24950         (UNSPEC_FPCMPDESHL): Likewise.
24951         (UNSPEC_FPCMPURSHL): Likewise.
24952         (cpu_feature): New CPU feature `vis4b'.
24953         (dictunpack{8,16,32}): New insns.
24954         (FPCSMODE): New mode iterator.
24955         (fpcscond): New code iterator.
24956         (fpcsucond): Likewise.
24957         (fpcmp{le,gt,eq,ne}{8,16,32}{si,di}shl): New insns.
24958         (fpcmpu{le,gt}{8,16,32}{si,di}shl): Likewise.
24959         (fpcmpde{8,16,32}{si,di}shl): Likewise.
24960         (fpcmpur{8,16,32}{si,di}shl): Likewise.
24961         * config/sparc/constraints.md: Define constraints `q' for unsigned
24962         2-bit integer constants and `t' for unsigned 5-bit integer
24963         constants.
24964         * config/sparc/predicates.md (imm5_operand_dictunpack8): New
24965         predicate.
24966         (imm5_operand_dictunpack16): Likewise.
24967         (imm5_operand_dictunpack32): Likewise.
24968         (imm2_operand): Likewise.
24969         * doc/invoke.texi (SPARC Options): Document -mvis4b.
24970         * doc/extend.texi (SPARC VIS Built-in Functions): Document the
24971         ditunpack* and fpcmp*shl builtins.
24973 2017-07-07  Jose E. Marchesi  <jose.marchesi@oracle.com>
24975         * config.gcc: Handle m8 in --with-{cpu,tune} options.
24976         * config.in: Add HAVE_AS_SPARC6 define.
24977         * config/sparc/driver-sparc.c (cpu_names): Add entry for the SPARC
24978         M8.
24979         * config/sparc/sol2.h (CPP_CPU64_DEFAULT_SPEC): Define for
24980         TARGET_CPU_m8.
24981         (ASM_CPU32_DEFAUILT_SPEC): Likewise.
24982         (CPP_CPU_SPEC): Handle m8.
24983         (ASM_CPU_SPEC): Likewise.
24984         * config/sparc/sparc-opts.h (enum processor_type): Add
24985         PROCESSOR_M8.
24986         * config/sparc/sparc.c (m8_costs): New struct.
24987         (sparc_option_override): Handle TARGET_CPU_m8.
24988         (sparc32_initialize_trampoline): Likewise.
24989         (sparc64_initialize_trampoline): Likewise.
24990         (sparc_issue_rate): Likewise.
24991         (sparc_register_move_cost): Likewise.
24992         * config/sparc/sparc.h (TARGET_CPU_m8): Define.
24993         (CPP_CPU64_DEFAULT_SPEC): Define for M8.
24994         (ASM_CPU64_DEFAULT_SPEC): Likewise.
24995         (CPP_CPU_SPEC): Handle M8.
24996         (ASM_CPU_SPEC): Likewise.
24997         (AS_M8_FLAG): Define.
24998         * config/sparc/sparc.md: Add m8 to the cpu attribute.
24999         * config/sparc/sparc.opt: New option -mcpu=m8 for sparc targets.
25000         * configure.ac (HAVE_AS_SPARC6): Check for assembler support for
25001         M8 instructions.
25002         * configure: Regenerate.
25003         * doc/invoke.texi (SPARC Options): Document -mcpu=m8 and
25004         -mtune=m8.
25006 2017-07-07  Jose E. Marchesi  <jose.marchesi@oracle.com>
25008         * config/sparc/niagara7.md: Rework the DFA scheduler to use insn
25009         subtypes.
25010         * config/sparc/sparc.md: Remove the `v3pipe' insn attribute.
25011         ("*movdi_insn_sp32"): Do not set v3pipe.
25012         ("*movsi_insn"): Likewise.
25013         ("*movdi_insn_sp64"): Likewise.
25014         ("*movsf_insn"): Likewise.
25015         ("*movdf_insn_sp32"): Likewise.
25016         ("*movdf_insn_sp64"): Likewise.
25017         ("*zero_extendsidi2_insn_sp64"): Likewise.
25018         ("*sign_extendsidi2_insn"): Likewise.
25019         ("*mov<VM32:mode>_insn"): Likewise.
25020         ("*mov<VM64:mode>_insn_sp64"): Likewise.
25021         ("*mov<VM64:mode>_insn_sp32"): Likewise.
25022         ("<plusminus_insn><VADDSUB:mode>3"): Likewise.
25023         ("<vlop:code><VL:mode>3"): Likewise.
25024         ("*not_<vlop:code><VL:mode>3"): Likewise.
25025         ("*nand<VL:mode>_vis"): Likewise.
25026         ("*<vlnotop:code>_not1<VL:mode>_vis"): Likewise.
25027         ("*<vlnotop:code>_not2<VL:mode>_vis"): Likewise.
25028         ("one_cmpl<VL:mode>2"): Likewise.
25029         ("faligndata<VM64:mode>_vis"): Likewise.
25030         ("alignaddrsi_vis"): Likewise.
25031         ("alignaddrdi_vis"): Likweise.
25032         ("alignaddrlsi_vis"): Likewise.
25033         ("alignaddrldi_vis"): Likewise.
25034         ("fcmp<gcond:code><GCM:gcm_name><P:mode>_vis"): Likewise.
25035         ("bmaskdi_vis"): Likewise.
25036         ("bmasksi_vis"): Likewise.
25037         ("bshuffle<VM64:mode>_vis"): Likewise.
25038         ("cmask8<P:mode>_vis"): Likewise.
25039         ("cmask16<P:mode>_vis"): Likewise.
25040         ("cmask32<P:mode>_vis"): Likewise.
25041         ("pdistn<P:mode>_vis"): Likewise.
25042         ("<vis3_addsub_ss_patname><VASS:mode>3"): Likewise.
25044 2017-07-07  Jose E. Marchesi  <jose.marchesi@oracle.com>
25046         * config/sparc/sparc.md ("subtype"): New insn attribute.
25047         ("*wrgsr_sp64"): Set insn subtype.
25048         ("*rdgsr_sp64"): Likewise.
25049         ("alignaddrsi_vis"): Likewise.
25050         ("alignaddrdi_vis"): Likewise.
25051         ("alignaddrlsi_vis"): Likewise.
25052         ("alignaddrldi_vis"): Likewise.
25053         ("<plusminus_insn><VADDSUB:mode>3"): Likewise.
25054         ("fexpand_vis"): Likewise.
25055         ("fpmerge_vis"): Likewise.
25056         ("faligndata<VM64:mode>_vis"): Likewise.
25057         ("bshuffle<VM64:mode>_vis"): Likewise.
25058         ("cmask8<P:mode>_vis"): Likewise.
25059         ("cmask16<P:mode>_vis"): Likewise.
25060         ("cmask32<P:mode>_vis"): Likewise.
25061         ("fchksm16_vis"): Likewise.
25062         ("v<vis3_shift_patname><GCM:mode>3"): Likewise.
25063         ("fmean16_vis"): Likewise.
25064         ("fp<plusminus_insn>64_vis"): Likewise.
25065         ("<plusminus_insn>v8qi3"): Likewise.
25066         ("<vis3_addsub_ss_patname><VASS:mode>3"): Likewise.
25067         ("<vis4_minmax_patname><VMMAX:mode>3"): Likewise.
25068         ("<vis4_uminmax_patname><VMMAX:mode>3"): Likewise.
25069         ("<vis3_addsub_ss_patname>v8qi3"): Likewise.
25070         ("<vis4_addsub_us_patname><VAUS:mode>3"): Likewise.
25071         ("*movqi_insn"): Likewise.
25072         ("*movhi_insn"): Likewise.
25073         ("*movsi_insn"): Likewise.
25074         ("movsi_pic_gotdata_op"): Likewise.
25075         ("*movdi_insn_sp32"): Likewise.
25076         ("*movdi_insn_sp64"): Likewise.
25077         ("movdi_pic_gotdata_op"): Likewise.
25078         ("*movsf_insn"): Likewise.
25079         ("*movdf_insn_sp32"): Likewise.
25080         ("*movdf_insn_sp64"): Likewise.
25081         ("*zero_extendhisi2_insn"): Likewise.
25082         ("*zero_extendqihi2_insn"): Likewise.
25083         ("*zero_extendqisi2_insn"): Likewise.
25084         ("*zero_extendqidi2_insn"): Likewise.
25085         ("*zero_extendhidi2_insn"): Likewise.
25086         ("*zero_extendsidi2_insn_sp64"): Likewise.
25087         ("ldfsr"): Likewise.
25088         ("prefetch_64"): Likewise.
25089         ("prefetch_32"): Likewise.
25090         ("tie_ld32"): Likewise.
25091         ("tie_ld64"): Likewise.
25092         ("*tldo_ldub_sp32"): Likewise.
25093         ("*tldo_ldub1_sp32"): Likewise.
25094         ("*tldo_ldub2_sp32"): Likewise.
25095         ("*tldo_ldub_sp64"): Likewise.
25096         ("*tldo_ldub1_sp64"): Likewise.
25097         ("*tldo_ldub2_sp64"): Likewise.
25098         ("*tldo_ldub3_sp64"): Likewise.
25099         ("*tldo_lduh_sp32"): Likewise.
25100         ("*tldo_lduh1_sp32"): Likewise.
25101         ("*tldo_lduh_sp64"): Likewise.
25102         ("*tldo_lduh1_sp64"): Likewise.
25103         ("*tldo_lduh2_sp64"): Likewise.
25104         ("*tldo_lduw_sp32"): Likewise.
25105         ("*tldo_lduw_sp64"): Likewise.
25106         ("*tldo_lduw1_sp64"): Likewise.
25107         ("*tldo_ldx_sp64"): Likewise.
25108         ("*mov<VM32:mode>_insn"): Likewise.
25109         ("*mov<VM64:mode>_insn_sp64"): Likewise.
25110         ("*mov<VM64:mode>_insn_sp32"): Likewise.
25112 2017-07-07  Jose E. Marchesi  <jose.marchesi@oracle.com>
25114         * config/sparc/sparc.md ("type"): New insn type viscmp.
25115         ("fcmp<gcond:code><GCM:gcm_name><P:mode>_vis"): Set insn type to
25116         viscmp.
25117         ("fpcmp<gcond:code>8<P:mode>_vis"): Likewise.
25118         ("fucmp<gcond:code>8<P:mode>_vis"): Likewise.
25119         ("fpcmpu<gcond:code><GCM:gcm_name><P:mode>_vis"): Likewise.
25120         * config/sparc/niagara7.md ("n7_vis_logical_v3pipe"): Handle
25121         viscmp.
25122         ("n7_vis_logical_11cycle"): Likewise.
25123         * config/sparc/niagara4.md ("n4_vis_logical"): Likewise.
25124         * config/sparc/niagara2.md ("niag3_vis": Likewise.
25125         * config/sparc/niagara.md ("niag_vis"): Likewise.
25126         * config/sparc/ultra3.md ("us3_fga"): Likewise.
25127         * config/sparc/ultra1_2.md ("us1_fga_double"): Likewise.
25129 2017-07-07  Jose E. Marchesi  <jose.marchesi@oracle.com>
25131         * config/sparc/sparc.md: New instruction type `bmask'.
25132         (bmaskdi_vis): Use the `bmask' type.
25133         (bmasksi_vis): Likewise.
25134         * config/sparc/ultra3.md (us3_array): Likewise.
25135         * config/sparc/niagara7.md (n7_array): Likewise.
25136         * config/sparc/niagara4.md (n4_array): Likewise.
25137         * config/sparc/niagara2.md (niag2_vis): Likewise.
25138         (niag3_vis): Likewise.
25139         * config/sparc/niagara.md (niag_vis): Likewise.
25141 2017-07-06  Jan Hubicka  <hubicka@ucw.cz>
25143         * ipa-comdats.c: Remove optimize check from gate.
25144         * ipa-fnsummary.c (ipa_fn_summary_generate): do not generate summary
25145         for functions not optimized.
25146         (ipa_fn_summary_read): Skip optimize check.
25147         (ipa_fn_summary_write): Likewise.
25148         * ipa-inline-analysis.c (do_estimate_growth_1): Check that caller
25149         is optimized.
25150         * ipa-inline.c (can_inline_edge_p): Not optimized functions are
25151         uninlinable.
25152         (can_inline_edge_p): Check flag_pcc_struct_return for match.
25153         (check_callers): Give up on caller which is not optimized.
25154         (inline_small_functions): Likewise.
25155         (ipa_inline): Do not give up when not optimizing.
25156         * ipa-visbility.c (function_and_variable_visibility): Do not optimize
25157         away unoptimizes cdtors.
25158         (whole_program_function_and_variable_visibility): Do
25159         ipa_discover_readonly_nonaddressable_vars in LTO mode.
25160         * ipa.c (process_references): Do not check optimize.
25161         (symbol_table::remove_unreachable_nodes): Update optimize check.
25162         (set_writeonly_bit): Update optimize check.
25163         (pass_ipa_cdtor_merge::gate): Do not check optimize.
25164         (pass_ipa_single_use::gate): Remove.
25166 2017-07-06  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
25168         * config/rs6000/rs6000.c (union_defs, union_uses, insn_is_load_p,
25169         insn_is_store_p, insn_is_swap_p, const_load_sequence_p, v2df_reduction_p,
25170         rtx_is_swappable_p, insn_is_swappable_p, chain_contains_only_swaps,
25171         mark_swaps_for_removal, swap_const_vector_halves, adjust_subreg_index,
25172         permute_load, permute_store, adjust_extract, adjust_splat,
25173         adjust_xxpermdi, adjust_concat, adjust_vperm, handle_special_swappables,
25174         replace_swap_with_copy, dump_swap_insn_table,
25175         alignment_with_canonical_addr, alignment_mask, find_alignment_op,
25176         recombine_lvx_pattern, recombine_stvx_pattern,
25177         recombine_lvx_stvx_patterns, rs6000_analyze_swaps,
25178         make_pass_analyze_swaps): Move all code related to p8 swap optimizations
25179         to file rs6000-p8swap.c.
25180         * config/rs6000/rs6000-p8swap.c: New file.
25181         * config/rs6000/t-rs6000: Add rule to build rs6000-p8swap.o.
25182         * config.gcc: Add rs6000-p8swap.o to extra_objs for powerpc*-*-*
25183         and rs6000*-*-* targets.
25185 2017-07-06  David Malcolm  <dmalcolm@redhat.com>
25187         * Makefile.in (selftest): Remove dependency on s-selftest-c++.
25189 2017-07-06  Jan Hubicka  <hubicka@ucw.cz>
25191         * lto-wrapper.c (merge_and_complain): Do not merge
25192         fexceptions, fnon_call_exceptions, ftrapv, ffp_contract_, fmath_errno,
25193         fsigned_zeros, ftrapping_math, fwrapv.
25194         (append_compiler_options): Do not track these options.
25195         (append_linker_options): Likewie
25197 2017-07-06  Jan Hubicka  <hubicka@ucw.cz>
25199         * cgraphunit.c (cgraph_node::finalize_function): When
25200         !flag_toplevel_reorde set no_reorder flag.
25201         (varpool_node::finalize_decl): Likewise.
25202         (symbol_table::compile): Drop no toplevel reorder path.
25204 2017-07-06  Jan Hubicka  <hubicka@ucw.cz>
25206         * bb-reorder.c (better_edge_p): Do not build traces across abnormal/eh
25207         edges; zero probability is not better than uninitialized.
25209 2017-07-06  Maxim Ostapenko  <m.ostapenko@samsung.com>
25211         * asan.h (asan_sanitize_allocas_p): Declare.
25212         * asan.c (asan_sanitize_allocas_p): New function.
25213         (handle_builtin_stack_restore): Bail out if !asan_sanitize_allocas_p.
25214         (handle_builtin_alloca): Likewise.
25215         * cfgexpand.c (expand_used_vars): Do not add allocas unpoisoning stuff
25216         if !asan_sanitize_allocas_p.
25217         * params.def (asan-instrument-allocas): Add new option.
25218         * params.h (ASAN_PROTECT_ALLOCAS): Define.
25219         * opts.c (common_handle_option): Disable allocas sanitization for
25220         KASan by default.
25222 2017-07-06  Maxim Ostapenko  <m.ostapenko@samsung.com>
25224         * asan.c: Include gimple-fold.h.
25225         (get_last_alloca_addr): New function.
25226         (handle_builtin_stackrestore): Likewise.
25227         (handle_builtin_alloca): Likewise.
25228         (asan_emit_allocas_unpoison): Likewise.
25229         (get_mem_refs_of_builtin_call): Add new parameter, remove const
25230         quallifier from first paramerer. Handle BUILT_IN_ALLOCA,
25231         BUILT_IN_ALLOCA_WITH_ALIGN and BUILT_IN_STACK_RESTORE builtins.
25232         (instrument_builtin_call): Pass gimple iterator to
25233         get_mem_refs_of_builtin_call.
25234         (last_alloca_addr): New global.
25235         * asan.h (asan_emit_allocas_unpoison): Declare.
25236         * builtins.c (expand_asan_emit_allocas_unpoison): New function.
25237         (expand_builtin): Handle BUILT_IN_ASAN_ALLOCAS_UNPOISON.
25238         * cfgexpand.c (expand_used_vars): Call asan_emit_allocas_unpoison
25239         if function calls alloca.
25240         * gimple-fold.c (replace_call_with_value): Remove static keyword.
25241         * gimple-fold.h (replace_call_with_value): Declare.
25242         * internal-fn.c: Include asan.h.
25243         * sanitizer.def (BUILT_IN_ASAN_ALLOCA_POISON,
25244         BUILT_IN_ASAN_ALLOCAS_UNPOISON): New builtins.
25246 2017-07-06  David Malcolm  <dmalcolm@redhat.com>
25248         * Makefile.in (SELFTEST_FLAGS): Drop "-x c", moving it to...
25249         (C_SELFTEST_FLAGS): New.
25250         (CPP_SELFTEST_FLAGS): New.
25251         (SELFTEST_DEPS): New, from deps of s-selftest.
25252         (C_SELFTEST_DEPS): New, from deps of s-selftest.
25253         (CPP_SELFTEST_DEPS): New.
25254         (selftest): Add dependency on s-selftest-c++.
25255         (s-selftest): Rename to...
25256         (s-selftest-c): ...this, moving deps to SELFTEST_DEPS
25257         and C_SELFTEST_DEPS, and using C_SELFTEST_FLAGS rather
25258         than SELFTEST_FLAGS.
25259         (selftest-gdb): Rename to...
25260         (selftest-c-gdb): ...this, using C_SELFTEST_DEPS and
25261         C_SELFTEST_FLAGS.
25262         (selftest-gdb): Reintroduce as an alias for selftest-c-gdb.
25263         (selftest-valgrind): Rename to...
25264         (selftest-c-valgrind): ...this, using C_SELFTEST_DEPS and
25265         C_SELFTEST_FLAGS.
25266         (selftest-valgrind): Reintroduce as an alias for
25267         selftest-c-valgrind.
25268         (s-selftest-c++): New.
25269         (selftest-c++-gdb): New.
25270         (selftest-c++-valgrind): New.
25272 2017-07-06  Olivier Hainque  <hainque@adacore.com>
25274         * gcc.c (process_command): When deciding if undefined variables
25275         should be ignored when processing specs, accept "gcc -v" as well.
25277 2017-07-06  Jan Hubicka  <hubicka@ucw.cz>
25279         * auto-profile.c (afdo_set_bb_count, afdo_propagate_edge,
25280         afdo_annotate_cfg): Set counts/probabilities as determined by afdo.
25282 2017-07-06  Thomas Preud'homme  <thomas.preudhomme@arm.com>
25284         * config/arm/arm-cpus.in (armv8-r): Add new entry.
25285         * config/arm/arm-isa.h (ISA_ARMv8r): Define macro.
25286         * config/arm/arm-tables.opt: Regenerate.
25287         * config/arm/arm.h (enum base_architecture): Add BASE_ARCH_8R
25288         enumerator.
25289         * doc/invoke.texi: Mention -march=armv8-r and its extensions.
25291 2017-07-06  Carl Love  <cel@us.ibm.com>
25293         * ChangeLog: Clean up from mid air collision
25295 2017-07-06  Carl Love  <cel@us.ibm.com>
25297         * config/rs6000/rs6000-c.c: Add support for built-in functions
25298         vector signed int vec_subc (vector signed int, vector signed int);
25299         vector signed __int128 vec_subc (vector signed __int128,
25300                                          vector signed __int128);
25301         vector unsigned __int128 vec_subc (vector unsigned __int128,
25302                                            vector unsigned __int128);
25303         vector signed int vec_sube (vector signed int, vector signed int,
25304                                     vector signed int);
25305         vector unsigned int vec_sube (vector unsigned int,
25306                                       vector unsigned int,
25307                                       vector unsigned int);
25308         vector signed __int128 vec_sube (vector signed __int128,
25309                                          vector signed __int128,
25310                                          vector signed__int128);
25311         vector unsigned __int128 vec_sube (vector unsigned __int128,
25312                                            vector unsigned __int128,
25313                                            vector unsigned __int128);
25314         vector signed int vec_subec (vector signed int, vector signed int,
25315                                      vector signed int);
25316         vector unsigned int vec_subec (vector unsigned int,
25317                                        vector unsigned int,
25318                                        vector unsigned int);
25319         vector signed __int128 vec_subec (vector signed __int128,
25320                                           vector signed __int128,
25321                                           vector signed__int128);
25322         vector unsigned __int128 vec_subec (vector unsigned __int128,
25323                                             vector unsigned __int128,
25324                                             vector unsigned __int128);
25325         * config/rs6000/rs6000.c (ALTIVEC_BUILTIN_VEC_SUBE,
25326         ALTIVEC_BUILTIN_VEC_SUBEC): Add ef_builtins.
25327         * config/rs6000/rs6000-builtin.def (SUBE, SUBEC): Add
25328         BU_ALTIVEC_OVERLOAD_X definitions.
25329         * config/rs6000/altivec.h (vec_sube, vec_subec): Add builtin defines.
25330         * doc/extend.texi: Update the built-in documentation file for the new
25331         built-in functions.
25333 2017-07-06  David Malcolm  <dmalcolm@redhat.com>
25335         PR c++/79300
25336         * diagnostic-show-locus.c (layout::layout): Use start and finish
25337         spelling location for the start and finish of each range.
25338         * genmatch.c (linemap_client_expand_location_to_spelling_point):
25339         Add unused aspect param.
25340         * input.c (expand_location_1): Add "aspect" param, and use it
25341         to access the correct part of the location.
25342         (expand_location): Pass LOCATION_ASPECT_CARET to new param of
25343         expand_location_1.
25344         (expand_location_to_spelling_point): Likewise.
25345         (linemap_client_expand_location_to_spelling_point): Add "aspect"
25346         param, and pass it to expand_location_1.
25348 2017-07-06  Sebastian Peryt  <sebastian.peryt@intel.com>
25350         * config/i386/avx512fintrin.h (_mm_mask_getexp_round_ss,
25351         _mm_maskz_getexp_round_ss, _mm_mask_getexp_round_sd,
25352         _mm_maskz_getexp_round_sd, _mm_mask_getmant_round_sd,
25353         _mm_maskz_getmant_round_sd, _mm_mask_getmant_round_ss,
25354         _mm_maskz_getmant_round_ss, _mm_mask_getexp_ss, _mm_maskz_getexp_ss,
25355         _mm_mask_getexp_sd, _mm_maskz_getexp_sd, _mm_mask_getmant_sd,
25356         _mm_maskz_getmant_sd, _mm_mask_getmant_ss,
25357         _mm_maskz_getmant_ss): New intrinsics.
25358         (__builtin_ia32_getexpss128_mask): Changed to ...
25359         __builtin_ia32_getexpss128_round ... this.
25360         (__builtin_ia32_getexpsd128_mask): Changed to ...
25361         __builtin_ia32_getexpsd128_round ... this.
25362         * config/i386/i386-builtin-types.def
25363         ((V2DF, V2DF, V2DF, INT, V2DF, UQI, INT),
25364         (V4SF, V4SF, V4SF, INT, V4SF, UQI, INT)): New function type aliases.
25365         * config/i386/i386-builtin.def (__builtin_ia32_getexpsd_mask_round,
25366         __builtin_ia32_getexpss_mask_round,     __builtin_ia32_getmantsd_mask_round,
25367         __builtin_ia32_getmantss_mask_round): New builtins.
25368         * config/i386/i386.c (V2DF_FTYPE_V2DF_V2DF_INT_V2DF_UQI_INT,
25369         V4SF_FTYPE_V4SF_V4SF_INT_V4SF_UQI_INT): Handle new types.
25370         (CODE_FOR_avx512f_vgetmantv2df_mask_round,
25371         CODE_FOR_avx512f_vgetmantv4sf_mask_round): New cases.
25372         * config/i386/sse.md
25373         (avx512f_sgetexp<mode><round_saeonly_name>): Changed to ...
25374         avx512f_sgetexp<mode><mask_scalar_name>
25375         <round_saeonly_scalar_name> ... this.
25376         (vgetexp<ssescalarmodesuffix>\t{<round_saeonly_op3>%2, %1, %0|
25377         %0, %1, %2<round_saeonly_op3>}): Changed to ...
25378         vgetexp<ssescalarmodesuffix>
25379         \t{<round_saeonly_scalar_mask_op3>%2, %1, %0<mask_scalar_operand3>|
25380         %0<mask_scalar_operand3>, %1, %2<round_saeonly_scalar_mask_op3>} ... this.
25381         (avx512f_vgetmant<mode><round_saeonly_name>): Changed to ...
25382         avx512f_vgetmant<mode><mask_scalar_name>
25383         <round_saeonly_scalar_name> ... this.
25384         (vgetmant<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
25385         %0, %1, %2<round_saeonly_op4>, %3}): Changed to ...
25386         vgetmant<ssescalarmodesuffix>
25387         \t{%3, <round_saeonly_scalar_mask_op4>%2, %1, %0<mask_scalar_operand4>|
25388         %0<mask_scalar_operand4>, %1, %2
25389         <round_saeonly_scalar_mask_op4>, %3} ... this.
25390         * config/i386/subst.md (mask_scalar_operand4,
25391         round_saeonly_scalar_mask_operand4,     round_saeonly_scalar_mask_op4,
25392         round_saeonly_scalar_nimm_predicate): New subst attributes.
25394 2017-07-06  Julia Koval  <julia.koval@intel.com>
25396         * config/i386/i386.c (ix86_erase_embedded_rounding):
25397         Remove code for old rounding pattern.
25399 2017-07-06  Richard Earnshaw  <rearnsha@arm.com>
25401         * config/arm/t-arm (GTM_H): Add arm-cpu.h.
25403 2017-07-06  Christophe Lyon  <christophe.lyon@linaro.org>
25405         * doc/sourcebuild.texi (Test Directives, Variants of
25406         dg-require-support): Add documentation for dg-require-stack-check.
25408 2017-07-05  Sebastian Peryt  <sebastian.peryt@intel.com>
25410         * config/i386/subst.md (mask_scalar, round_scalar,
25411         round_saeonly_scalar): New meta-templates.
25412         (mask_scalar_name, mask_scalar_operand3, round_scalar_name,
25413         round_scalar_mask_operand3, round_scalar_mask_op3,
25414         round_scalar_constraint, round_scalar_prefix, round_saeonly_scalar_name,
25415         round_saeonly_scalar_mask_operand3, round_saeonly_scalar_mask_op3,
25416         round_saeonly_scalar_constraint,
25417         round_saeonly_scalar_prefix): New subst attribute.
25418         * config/i386/sse.md
25419         (<sse>_vm<plusminus_insn><mode>3<mask_name><round_name>): Renamed to ...
25420         <sse>_vm<plusminus_insn><mode>3<mask_scalar_name>
25421         <round_scalar_name> ... this.
25422         (<sse>_vm<multdiv_mnemonic><mode>3<mask_name><round_name>): Renamed to ...
25423         <sse>_vm<multdiv_mnemonic><mode>3<mask_scalar_name>
25424         <round_scalar_name> ... this.
25425         (<sse>_vm<code><mode>3<mask_name><round_saeonly_name>): Renamed to ...
25426         <sse>_vm<code><mode>3<mask_scalar_name>
25427         <round_saeonly_scalar_name> ... this.
25428         (v<plusminus_mnemonic><ssescalarmodesuffix>
25429         \t{<round_mask_op3>%2, %1, %0<mask_operand3>|
25430         %0<mask_operand3>, %1, %<iptr>2<round_mask_op3>}): Changed to ...
25431         v<plusminus_mnemonic><ssescalarmodesuffix>
25432         \t{<round_scalar_mask_op3>%2, %1, %0<mask_scalar_operand3>|
25433         %0<mask_scalar_operand3>, %1, %<iptr>2<round_scalar_mask_op3>} ... this.
25434         (v<multdiv_mnemonic><ssescalarmodesuffix>
25435         \t{<round_mask_op3>%2, %1, %0<mask_operand3>|
25436         %0<mask_operand3>, %1, %<iptr>2<round_mask_op3>}): Changed to ...
25437         v<multdiv_mnemonic><ssescalarmodesuffix>
25438         \t{<round_scalar_mask_op3>%2, %1, %0<mask_scalar_operand3>|
25439         %0<mask_scalar_operand3>, %1, %<iptr>2<round_scalar_mask_op3>} ... this.
25440         (v<maxmin_float><ssescalarmodesuffix>
25441         \t{<round_saeonly_mask_op3>%2, %1, %0<mask_operand3>|
25442         %0<mask_operand3>, %1, %<iptr>2<round_saeonly_mask_op3>}): Changed to ...
25443         v<maxmin_float><ssescalarmodesuffix>
25444         \t{<round_saeonly_scalar_mask_op3>%2, %1, %0<mask_scalar_operand3>|
25445         %0<mask_scalar_operand3>, %1, %<iptr>2
25446         <round_saeonly_scalar_mask_op3>} ... this.
25448 2017-07-05  Richard Earnshaw  <rearnsha@arm.com>
25450         * config/arm/arm.c (arm_fixed_condition_code_regs): New function.
25451         (TARGET_FIXED_CONDITION_CODE_REGS): Redefine.
25453 2017-07-05  Richard Sandiford  <richard.sandiford@linaro.org>
25454             Alan Hayward  <alan.hayward@arm.com>
25455             David Sherwood  <david.sherwood@arm.com>
25457         * combine.c (simplify_if_then_else): Remove "enum" before
25458         "machine_mode".
25459         * compare-elim.c (can_eliminate_compare): Likewise.
25460         * config/aarch64/aarch64-builtins.c (aarch64_simd_builtin_std_type):
25461         Likewise.
25462         (aarch64_lookup_simd_builtin_type): Likewise.
25463         (aarch64_simd_builtin_type): Likewise.
25464         (aarch64_init_simd_builtin_types): Likewise.
25465         (aarch64_simd_expand_args): Likewise.
25466         * config/aarch64/aarch64-protos.h (aarch64_simd_attr_length_rglist):
25467         Likewise.
25468         (aarch64_reverse_mask): Likewise.
25469         (aarch64_simd_emit_reg_reg_move): Likewise.
25470         (aarch64_gen_adjusted_ldpstp): Likewise.
25471         (aarch64_ccmp_mode_to_code): Likewise.
25472         (aarch64_operands_ok_for_ldpstp): Likewise.
25473         (aarch64_operands_adjust_ok_for_ldpstp): Likewise.
25474         * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
25475         Likewise.
25476         (aarch64_min_divisions_for_recip_mul): Likewise.
25477         (aarch64_reassociation_width): Likewise.
25478         (aarch64_get_condition_code_1): Likewise.
25479         (aarch64_simd_emit_reg_reg_move): Likewise.
25480         (aarch64_simd_attr_length_rglist): Likewise.
25481         (aarch64_reverse_mask): Likewise.
25482         (aarch64_operands_ok_for_ldpstp): Likewise.
25483         (aarch64_operands_adjust_ok_for_ldpstp): Likewise.
25484         (aarch64_gen_adjusted_ldpstp): Likewise.
25485         * config/aarch64/cortex-a57-fma-steering.c (fma_node::rename):
25486         Likewise.
25487         * config/arc/arc.c (legitimate_offset_address_p): Likewise.
25488         * config/arm/arm-builtins.c (arm_simd_builtin_std_type): Likewise.
25489         (arm_lookup_simd_builtin_type): Likewise.
25490         (arm_simd_builtin_type): Likewise.
25491         (arm_init_simd_builtin_types): Likewise.
25492         (arm_expand_builtin_args): Likewise.
25493         * config/arm/arm-protos.h (arm_expand_builtin): Likewise.
25494         * config/ft32/ft32.c (ft32_libcall_value): Likewise.
25495         (ft32_setup_incoming_varargs): Likewise.
25496         (ft32_function_arg): Likewise.
25497         (ft32_function_arg_advance): Likewise.
25498         (ft32_pass_by_reference): Likewise.
25499         (ft32_arg_partial_bytes): Likewise.
25500         (ft32_valid_pointer_mode): Likewise.
25501         (ft32_addr_space_pointer_mode): Likewise.
25502         (ft32_addr_space_legitimate_address_p): Likewise.
25503         * config/i386/i386-protos.h (ix86_operands_ok_for_move_multiple):
25504         Likewise.
25505         * config/i386/i386.c (ix86_setup_incoming_vararg_bounds): Likewise.
25506         (ix86_emit_outlined_ms2sysv_restore): Likewise.
25507         (iamcu_alignment): Likewise.
25508         (canonicalize_vector_int_perm): Likewise.
25509         (ix86_noce_conversion_profitable_p): Likewise.
25510         (ix86_mpx_bound_mode): Likewise.
25511         (ix86_operands_ok_for_move_multiple): Likewise.
25512         * config/microblaze/microblaze-protos.h
25513         (microblaze_expand_conditional_branch_reg): Likewise.
25514         * config/microblaze/microblaze.c
25515         (microblaze_expand_conditional_branch_reg): Likewise.
25516         * config/powerpcspe/powerpcspe.c (rs6000_init_hard_regno_mode_ok):
25517         Likewise.
25518         (rs6000_reassociation_width): Likewise.
25519         (rs6000_invalid_binary_op): Likewise.
25520         (fusion_p9_p): Likewise.
25521         (emit_fusion_p9_load): Likewise.
25522         (emit_fusion_p9_store): Likewise.
25523         * config/riscv/riscv-protos.h (riscv_regno_mode_ok_for_base_p):
25524         Likewise.
25525         (riscv_hard_regno_mode_ok_p): Likewise.
25526         (riscv_address_insns): Likewise.
25527         (riscv_split_symbol): Likewise.
25528         (riscv_legitimize_move): Likewise.
25529         (riscv_function_value): Likewise.
25530         (riscv_hard_regno_nregs): Likewise.
25531         (riscv_expand_builtin): Likewise.
25532         * config/riscv/riscv.c (riscv_build_integer_1): Likewise.
25533         (riscv_build_integer): Likewise.
25534         (riscv_split_integer): Likewise.
25535         (riscv_legitimate_constant_p): Likewise.
25536         (riscv_cannot_force_const_mem): Likewise.
25537         (riscv_regno_mode_ok_for_base_p): Likewise.
25538         (riscv_valid_base_register_p): Likewise.
25539         (riscv_valid_offset_p): Likewise.
25540         (riscv_valid_lo_sum_p): Likewise.
25541         (riscv_classify_address): Likewise.
25542         (riscv_legitimate_address_p): Likewise.
25543         (riscv_address_insns): Likewise.
25544         (riscv_load_store_insns): Likewise.
25545         (riscv_force_binary): Likewise.
25546         (riscv_split_symbol): Likewise.
25547         (riscv_force_address): Likewise.
25548         (riscv_legitimize_address): Likewise.
25549         (riscv_move_integer): Likewise.
25550         (riscv_legitimize_const_move): Likewise.
25551         (riscv_legitimize_move): Likewise.
25552         (riscv_address_cost): Likewise.
25553         (riscv_subword): Likewise.
25554         (riscv_output_move): Likewise.
25555         (riscv_canonicalize_int_order_test): Likewise.
25556         (riscv_emit_int_order_test): Likewise.
25557         (riscv_function_arg_boundary): Likewise.
25558         (riscv_pass_mode_in_fpr_p): Likewise.
25559         (riscv_pass_fpr_single): Likewise.
25560         (riscv_pass_fpr_pair): Likewise.
25561         (riscv_get_arg_info): Likewise.
25562         (riscv_function_arg): Likewise.
25563         (riscv_function_arg_advance): Likewise.
25564         (riscv_arg_partial_bytes): Likewise.
25565         (riscv_function_value): Likewise.
25566         (riscv_pass_by_reference): Likewise.
25567         (riscv_setup_incoming_varargs): Likewise.
25568         (riscv_print_operand): Likewise.
25569         (riscv_elf_select_rtx_section): Likewise.
25570         (riscv_save_restore_reg): Likewise.
25571         (riscv_for_each_saved_reg): Likewise.
25572         (riscv_register_move_cost): Likewise.
25573         (riscv_hard_regno_mode_ok_p): Likewise.
25574         (riscv_hard_regno_nregs): Likewise.
25575         (riscv_class_max_nregs): Likewise.
25576         (riscv_memory_move_cost): Likewise.
25577         * config/rl78/rl78-protos.h (rl78_split_movsi): Likewise.
25578         * config/rl78/rl78.c (rl78_split_movsi): Likewise.
25579         (rl78_addr_space_address_mode): Likewise.
25580         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
25581         Likewise.
25582         * config/rs6000/rs6000.c (rs6000_init_hard_regno_mode_ok): Likewise.
25583         (rs6000_reassociation_width): Likewise.
25584         (rs6000_invalid_binary_op): Likewise.
25585         (fusion_p9_p): Likewise.
25586         (emit_fusion_p9_load): Likewise.
25587         (emit_fusion_p9_store): Likewise.
25588         * config/visium/visium-protos.h (prepare_move_operands): Likewise.
25589         (ok_for_simple_move_operands): Likewise.
25590         (ok_for_simple_move_strict_operands): Likewise.
25591         (ok_for_simple_arith_logic_operands): Likewise.
25592         (visium_legitimize_reload_address): Likewise.
25593         (visium_select_cc_mode): Likewise.
25594         (output_cbranch): Likewise.
25595         (visium_split_double_move): Likewise.
25596         (visium_expand_copysign): Likewise.
25597         (visium_expand_int_cstore): Likewise.
25598         (visium_expand_fp_cstore): Likewise.
25599         * config/visium/visium.c (visium_pass_by_reference): Likewise.
25600         (visium_function_arg): Likewise.
25601         (visium_function_arg_advance): Likewise.
25602         (visium_libcall_value): Likewise.
25603         (visium_setup_incoming_varargs): Likewise.
25604         (visium_legitimate_constant_p): Likewise.
25605         (visium_legitimate_address_p): Likewise.
25606         (visium_legitimize_address): Likewise.
25607         (visium_secondary_reload): Likewise.
25608         (visium_register_move_cost): Likewise.
25609         (visium_memory_move_cost): Likewise.
25610         (prepare_move_operands): Likewise.
25611         (ok_for_simple_move_operands): Likewise.
25612         (ok_for_simple_move_strict_operands): Likewise.
25613         (ok_for_simple_arith_logic_operands): Likewise.
25614         (visium_function_value_1): Likewise.
25615         (rtx_ok_for_offset_p): Likewise.
25616         (visium_legitimize_reload_address): Likewise.
25617         (visium_split_double_move): Likewise.
25618         (visium_expand_copysign): Likewise.
25619         (visium_expand_int_cstore): Likewise.
25620         (visium_expand_fp_cstore): Likewise.
25621         (visium_split_cstore): Likewise.
25622         (visium_select_cc_mode): Likewise.
25623         (visium_split_cbranch): Likewise.
25624         (output_cbranch): Likewise.
25625         (visium_print_operand_address): Likewise.
25626         * expmed.c (flip_storage_order): Likewise.
25627         * expmed.h (emit_cstore): Likewise.
25628         (flip_storage_order): Likewise.
25629         * genrecog.c (validate_pattern): Likewise.
25630         * hsa-gen.c (gen_hsa_addr): Likewise.
25631         * internal-fn.c (expand_arith_overflow): Likewise.
25632         * ira-color.c (allocno_copy_cost_saving): Likewise.
25633         * lra-assigns.c (find_hard_regno_for_1): Likewise.
25634         * lra-constraints.c (prohibited_class_reg_set_mode_p): Likewise.
25635         (process_invariant_for_inheritance): Likewise.
25636         * lra-eliminations.c (move_plus_up): Likewise.
25637         * omp-low.c (lower_oacc_reductions): Likewise.
25638         * simplify-rtx.c (simplify_subreg): Likewise.
25639         * target.def (TARGET_SETUP_INCOMING_VARARG_BOUNDS): Likewise.
25640         (TARGET_CHKP_BOUND_MODE): Likewise..
25641         * targhooks.c (default_chkp_bound_mode): Likewise.
25642         (default_setup_incoming_vararg_bounds): Likewise.
25643         * targhooks.h (default_chkp_bound_mode): Likewise.
25644         (default_setup_incoming_vararg_bounds): Likewise.
25645         * tree-ssa-math-opts.c (divmod_candidate_p): Likewise.
25646         * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
25647         (have_whole_vector_shift): Likewise.
25648         * tree-vect-stmts.c (vectorizable_load): Likewise.
25649         * doc/tm.texi: Regenerate.
25651 2017-07-05  Georg-Johann Lay  <avr@gjlay.de>
25653         Graceful degrade if Binutils PR21472 is not available.
25655         PR target/81072
25656         * configure.ac [target=avr]: WARN instead of ERROR if avrxmega3
25657         .rodata in flash test fails.
25658         (HAVE_LD_AVR_AVRXMEGA3_RODATA_IN_FLASH): Define it if test passes.
25659         * confgure: Regenerate.
25660         * config.in: Regenerate.
25661         * config/avr/avr.c (avr_asm_named_section)
25662         [HAVE_LD_AVR_AVRXMEGA3_RODATA_IN_FLASH]: Only trigger
25663         __do_copy_data for stuff in .rodata if flash_pm_offset = 0.
25664         (avr_asm_init_sections): Same.
25666 2017-07-05  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
25668         * config/arm/neon.md (fma<VCVTF:mode>4): Remove %?.
25669         (fma<VH:mode>4_intrinsic): Likewise.
25670         (*fmsub<VCVTF:mode>4): Likewise.
25671         (*fmsub<VH:mode>4_intrinsic): Likewise.
25673 2017-07-05  Georg-Johann Lay  <avr@gjlay.de>
25675         PR target/81305
25676         * config/avr/avr.c (avr_out_movhi_mr_r_xmega) [CONSTANT_ADDRESS_P]:
25677         Don't depend on "optimize > 0".
25678         (out_movhi_r_mr, out_movqi_mr_r): Same.
25679         (out_movhi_mr_r, out_movqi_r_mr): Same.
25680         (avr_address_cost) [CONSTANT_ADDRESS_P]: Don't depend cost for
25681         io_address_operand on "optimize > 0".
25683 2017-07-05  Bin Cheng  <bin.cheng@arm.com>
25685         * tree-loop-distribution.c: Add general explanantion on the pass.
25686         (generate_loops_for_partition): Mark distributed loop.
25687         (pg_add_dependence_edges): New parameter.  Handle alias data
25688         dependence specially and record it in the parameter if asked.
25689         (struct pg_vdata, pg_edata, pg_edge_callback_data): New structs.
25690         (init_partition_graph_vertices, add_partition_graph_edge): New.
25691         (pg_skip_alias_edge, free_partition_graph_edata_cb): New.
25692         (free_partition_graph_vdata, build_partition_graph): New.
25693         (sort_partitions_by_post_order, merge_dep_scc_partitions): New.
25694         (pg_collect_alias_ddrs, break_alias_scc_partitions): New.
25695         (data_ref_segment_size, latch_dominated_by_data_ref): New.
25696         (compute_alias_check_pairs, version_loop_by_alias_check): New.
25697         (version_for_distribution_p, finalize_partitions): New.
25698         (distribute_loop): Handle alias data dependence specially.  Factor
25699         out loop fusion code as functions and call these functions.
25701 2017-07-05  Bin Cheng  <bin.cheng@arm.com>
25703         * tree-loop-distribution.c (classify_partition): New parameter and
25704         better handle reduction statement.
25705         (rdg_build_partitions): Revise comment.
25706         (distribute_loop): Compute statements in all partitions and pass it
25707         to classify_partition.
25709 2017-07-05  Bin Cheng  <bin.cheng@arm.com>
25711         * tree-loop-distribution.c (enum partition_type): New.
25712         (struct partition): New field type.
25713         (partition_merge_into): Add parameter.  Update partition type.
25714         (data_dep_in_cycle_p, update_type_for_merge): New functions.
25715         (build_rdg_partition_for_vertex): Compute partition type.
25716         (rdg_build_partitions): Dump partition type.
25717         (distribute_loop): Update calls to partition_merge_into.
25719 2017-07-05  Bin Cheng  <bin.cheng@arm.com>
25721         * tree-loop-distribution.c (struct ddr_hasher): New.
25722         (ddr_hasher::hash, ::equal, get_data_dependence): New function.
25723         (ddrs_table): New.
25724         (classify_partition): Call get_data_dependence.
25725         (pg_add_dependence_edges): Ditto.
25726         (distribute_loop): Release data dependence hash table.
25728 2017-07-05  Bin Cheng  <bin.cheng@arm.com>
25730         * tree-loop-distribution.c (ref_base_address): Delete.
25731         (similar_memory_accesses): Rename ...
25732         (share_memory_accesses): ... to this.  Check if partitions access
25733         the same memory reference.
25734         (distribute_loop): Call share_memory_accesses.
25736 2017-07-05  Bin Cheng  <bin.cheng@arm.com>
25738         * tree-loop-distribution.c (struct partition): New field recording
25739         its data reference.
25740         (partition_alloc, partition_free): Init and release data refs.
25741         (partition_merge_into): Merge data refs.
25742         (build_rdg_partition_for_vertex): Collect data refs for partition.
25743         (pg_add_dependence_edges): Change parameters from vector to bitmap.
25744         Update uses.
25745         (distribute_loop): Remve data refs from vertice data of partition
25746         graph.
25748 2017-07-05  Bin Cheng  <bin.cheng@arm.com>
25750         * tree-loop-distribution.c (params.h): Include header file.
25751         (MAX_DATAREFS_NUM, DR_INDEX): New macro.
25752         (datarefs_vec): New global var.
25753         (create_rdg_vertices): Use datarefs_vec directly.
25754         (free_rdg): Don't free data references.
25755         (build_rdg): Update use.  Don't free data references.
25756         (distribute_loop): Compute global variable for data references.
25757         Bail out if there are too many data references.
25759 2017-07-05  Bin Cheng  <bin.cheng@arm.com>
25761         * tree-loop-distribution.c (loop_nest): New global var.
25762         (build_rdg): Use loop directly, rather than loop nest.
25763         (pg_add_dependence_edges): Remove loop nest parameter.  Use global
25764         variable directly.
25765         (distribute_loop): Compute global variable loop nest.  Update use.
25767 2017-07-05  Bin Cheng  <bin.cheng@arm.com>
25769         * tree-loop-distribution.c (enum fuse_type, fuse_message): New.
25770         (partition_merge_into): New parameter.  Dump reason for fusion.
25771         (distribute_loop): Update use of partition_merge_into.
25773 2017-07-05  Bin Cheng  <bin.cheng@arm.com>
25775         * tree-loop-distribution.c (bb_top_order_index): New.
25776         (bb_top_order_index_size, bb_top_order_cmp): New.
25777         (stmts_from_loop): Use topological order.
25778         (pass_loop_distribution::execute): Compute and release topological
25779         order for basic blocks.
25781 2017-07-05  Bin Cheng  <bin.cheng@arm.com>
25783         * tree-loop-distribution.c (pass_loop_distribution::execute): Skip
25784         if no loops.
25786 2017-07-05  Bin Cheng  <bin.cheng@arm.com>
25788         * cfgloop.h (struct loop): Add comment.  New field orig_loop_num.
25789         * cfgloopmanip.c (lv_adjust_loop_entry_edge): Comment change.
25790         * internal-fn.c (expand_LOOP_DIST_ALIAS): New function.
25791         * internal-fn.def (LOOP_DIST_ALIAS): New.
25792         * tree-vectorizer.c (fold_loop_vectorized_call): Rename to ...
25793         (fold_loop_internal_call): ... this.
25794         (vect_loop_dist_alias_call): New function.
25795         (set_uid_loop_bbs): Call fold_loop_internal_call.
25796         (vectorize_loops): Fold IFN_LOOP_VECTORIZED and IFN_LOOP_DIST_ALIAS
25797         internal calls.
25799 2017-07-04  Uros Bizjak  <ubizjak@gmail.com>
25801         PR target/81300
25802         * config/i386/i386.md (setcc + movzbl/and to xor + setcc peepholes):
25803         Require dead FLAGS_REG at the beginning of a peephole.
25805 2017-07-04  Uros Bizjak  <ubizjak@gmail.com>
25807         PR target/81294
25808         * config/i386/adxintrin.h (_subborrow_u32): Swap _X and _Y
25809         arguments in the call to __builtin_ia32_sbb_u32.
25810         (_subborrow_u64): Swap _X and _Y arguments in the call to
25811         __builtin_ia32_sbb_u64.
25813 2017-07-04  Jakub Jelinek  <jakub@redhat.com>
25815         PR debug/81278
25816         * tree-vrp.c (compare_assert_loc): Turn into a function template
25817         with stable template parameter.  Only test if a->e is NULL,
25818         !a->e == !b->e has been verified already.  Use e == NULL or
25819         e != NULL instead of e or ! e tests.  If stable is true, don't use
25820         iterative_hash_expr, on the other side allow a or b or both NULL
25821         and sort the NULLs last.
25822         (process_assert_insertions): Sort using compare_assert_loc<false>
25823         instead of compare_assert_loc, later sort using
25824         compare_assert_loc<true> before calling process_assert_insertions_for
25825         in a loop.  Use break instead of continue once seen NULL pointer.
25827 2017-07-04  Thomas Preud'homme  <thomas.preudhomme@arm.com>
25829         * config/arm/driver-arm.c (arm_cpu_table): Add entry for ARM
25830         Cortex-R7 and Cortex-R8 processors.
25832 2017-07-04  Jan Hubicka  <hubicka@ucw.cz>
25834         * ipa-utils.c (ipa_merge_profiles): Fix merging when dst is
25835         uninitialized while src is not.
25837 2017-07-04  Richard Earnshaw  <rearnsha@arm.com>
25839         * common/config/arm/arm-common.c: Adjust include path for
25840         arm-cpu-cdata.h
25841         * config/arm/t-arm (TM_H): Adjust path for arm-cpu.h.
25842         (arm-cpu.h): Create in build directory.  Adjust dependency rules.
25843         (arm-cpu-data.h): Likewise.
25844         (arm-cpu-cdata.h): Likewise.
25845         * config/arm/arm-cpu.h: Delete.
25846         * config/arm/arm-cpu-cdata.h: Delete.
25847         * config/arm/arm-cpu-data.h: Delete.
25849 2017-07-04  James Greenhalgh  <james.greenhalgh@arm.com>
25851         * config/arm/arm-cpus.in (cortex-a55): New.
25852         (cortex-a75): Likewise.
25853         (cortex-a75.cortex-a55): Likewise.
25854         * config/arm/driver-arm.c (arm_cpu_table): Add cortex-a55 and
25855         cortex-a75.
25856         * doc/invoke.texi (-mcpu): Document cortex-a55 and cortex-a75.
25857         * config/arm/arm-cpu-cdata.h: Regenerate.
25858         * config/arm/arm-cpu-data.h: Regenerate.
25859         * config/arm/arm-cpu.h: Regenerate.
25860         * config/arm/arm-tables.opt: Regenerate.
25861         * config/arm/arm-tune.md: Regenerate.
25863 2017-07-04  Jan Hubicka  <hubicka@ucw.cz>
25865         * haifa-sched.c (sched_create_recovery_edges): Update profile.
25867 2017-07-04  Jan Hubicka  <hubicka@ucw.cz>
25869         * bb-reorder.c (better_edge_p): Fix handling of uninitialized
25870         probability.
25872 2017-07-04  Richard Sandiford  <richard.sandiford@linaro.org>
25874         PR tree-optimization/81292
25875         * tree-ssa-strlen.c (handle_builtin_strlen): When setting
25876         full_string_p, also call adjust_related_strinfos if the adjustment
25877         is simple, otherwise invalidate related strinfos.
25879 2017-07-04  Martin Liska  <mliska@suse.cz>
25881         PR sanitizer/81040
25882         * sanopt.c (sanitize_rewrite_addressable_params): Mark the
25883         newly created variable as DECL_IGNORED_P.
25885 2017-07-04  Martin Liska  <mliska@suse.cz>
25887         PR ipa/81293
25888         * ipa-inline.c (inline_small_functions):
25889         Use xstrdup_for_dump.
25891 2017-07-04  Tom de Vries  <tom@codesourcery.com>
25893         * graph.c (draw_cfg_edges): Save and restore EDGE_DFS_BACK.
25895 2017-07-03  Dominique d'Humieres  <dominiq@lps.ens.fr>
25897         PR target/81033
25898         * config/darwin.c (darwin_function_switched_text_sections):
25899         Fix spaces.
25901 2017-07-03  Jan Hubicka  <hubicka@ucw.cz>
25903         * tree-vect-loop-manip.c (vect_do_peeling): Fix scaling up.
25905 2017-07-03  Richard Earnshaw  <rearnsha@arm.com>
25907         * doc/invoke.texi (ARM Options): Add -mbe8 and -mbe32 to option summary.
25909 2017-07-03  Richard Sandiford  <richard.sandiford@linaro.org>
25911         * tree-vect-loop.c (vect_analyze_loop_2): Treat min_scalar_loop_bound,
25912         min_profitable_iters, and th as inclusive lower bounds.
25913         Fix LOOP_VINFO_PEELING_FOR_GAPS condition.
25914         (vect_estimate_min_profitable_iters): Return inclusive lower bounds
25915         for min_profitable_iters and min_profitable_estimate.
25916         (vect_transform_loop): Treat th as an inclusive lower bound.
25917         * tree-vect-loop-manip.c (vect_loop_versioning): Likewise.
25919 2017-07-03  Dominique d'Humieres  <dominiq@lps.ens.fr>
25921         PR target/81033
25922         * config/darwin.c (darwin_function_switched_text_sections):
25923         Replace DECL_NAME with DECL_ASSEMBLER_NAME, split assemble_name_raw
25924         in two pieces, and suppress the use of buf.
25926 2017-07-03  Nathan Sidwell  <nathan@acm.org>
25928         * hash-table.h (hash_table_mod1): Fix indentation.
25930 2017-07-02  Jan Hubicka  <hubicka@ucw.cz>
25932         PR middle-end/81290
25933         * predict.c (force_edge_cold): Be more careful about propagation
25934         backward.
25935         * profile-count.h (profile_probability::guessed,
25936         profile_probability::fdo, profile_count::guessed, profile_count::fdo):
25937         New.
25938         * tree-ssa-threadupdate.c (recompute_probabilities): Result is guessed.
25940 2017-07-03  James Greenhalgh  <james.greenhalgh@arm.com>
25942         * doc/invoke.texi (rcpc architecture extension): Document it.
25944 2017-07-03  Richard Biener  <rguenther@suse.de>
25946         PR tree-optimization/60510
25947         * tree-vect-loop.c (vect_create_epilog_for_reduction): Pass in
25948         the scalar reduction PHI and use it.
25949         (vectorizable_reduction): Properly guard the single_defuse_cycle
25950         path for non-SLP reduction chains where we cannot use it.
25951         Rework reduc_def/index and vector type deduction.  Rework
25952         vector operand gathering during reduction op code-gen.
25953         * tree-vect-slp.c (vect_analyze_slp): For failed SLP reduction
25954         chains dissolve the chain and leave it to non-SLP reduction
25955         handling.
25957 2017-07-03  Richard Sandiford  <richard.sandiford@linaro.org>
25959         * tree-data-ref.h (dr_alignment): Declare.
25960         * tree-data-ref.c (dr_alignment): New function.
25961         * tree-vectorizer.h (dataref_aux): Remove base_element_aligned.
25962         * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
25963         set it.
25964         * tree-vect-stmts.c (vectorizable_store): Use dr_alignment.
25966 2017-07-03  Richard Sandiford  <richard.sandiford@linaro.org>
25968         * tree-data-ref.h (innermost_loop_behavior): Add base_alignment
25969         and base_misalignment fields.
25970         (DR_BASE_ALIGNMENT, DR_BASE_MISALIGNMENT): New macros.
25971         * tree-data-ref.c: Include builtins.h.
25972         (dr_analyze_innermost): Set up the new innmost_loop_behavior fields.
25973         * tree-vectorizer.h (STMT_VINFO_DR_BASE_ALIGNMENT): New macro.
25974         (STMT_VINFO_DR_BASE_MISALIGNMENT): Likewise.
25975         * tree-vect-data-refs.c: Include tree-cfg.h.
25976         (vect_compute_data_ref_alignment): Use the new innermost_loop_behavior
25977         fields instead of calculating an alignment here.
25978         (vect_analyze_data_refs): Use dr_analyze_innermost.  Dump the new
25979         innermost_loop_behavior fields.
25981 2017-07-03  Richard Sandiford  <richard.sandiford@linaro.org>
25983         * tree-data-ref.h (innermost_loop_behavior): Add a step_alignment
25984         field.
25985         (DR_STEP_ALIGNMENT): New macro.
25986         * tree-vectorizer.h (STMT_VINFO_DR_STEP_ALIGNMENT): Likewise.
25987         * tree-data-ref.c (dr_analyze_innermost): Initalize step_alignment.
25988         (create_data_ref): Print it.
25989         * tree-vect-stmts.c (vectorizable_load): Use the step alignment
25990         to tell whether the step preserves vector (mis)alignment.
25991         * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Likewise.
25992         Move the check for an integer step and generalise to all INTEGER_CST.
25993         (vect_analyze_data_refs): Set DR_STEP_ALIGNMENT when setting DR_STEP.
25994         Print the outer step alignment.
25996 2017-07-03  Richard Sandiford  <richard.sandiford@linaro.org>
25998         * tree-data-ref.h (innermost_loop_behavior): Replace aligned_to
25999         with offset_alignment.
26000         (DR_ALIGNED_TO): Delete.
26001         (DR_OFFSET_ALIGNMENT): New macro.
26002         * tree-vectorizer.h (STMT_VINFO_DR_ALIGNED_TO): Delete.
26003         (STMT_VINFO_DR_OFFSET_ALIGNMENT): New macro.
26004         * tree-data-ref.c (dr_analyze_innermost): Update after above changes.
26005         (create_data_ref): Likewise.
26006         * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Likewise.
26007         (vect_analyze_data_refs): Likewise.
26008         * tree-if-conv.c (if_convertible_loop_p_1): Use memset before
26009         creating dummy innermost behavior.
26011 2017-07-03  Richard Sandiford  <richard.sandiford@linaro.org>
26013         * tree-data-ref.h (dr_analyze_innermost): Replace the dr argument
26014         with a "innermost_loop_behavior *" and refeence tree.
26015         * tree-data-ref.c (dr_analyze_innermost): Likewise.
26016         (create_data_ref): Update call accordingly.
26017         * tree-predcom.c (find_looparound_phi): Likewise.
26019 2017-07-03  Richard Sandiford  <richard.sandiford@linaro.org>
26021         * tree-vectorizer.h (_stmt_vec_info): Replace individual dr_*
26022         fields with dr_wrt_vec_loop.
26023         (STMT_VINFO_DR_BASE_ADDRESS, STMT_VINFO_DR_INIT, STMT_VINFO_DR_OFFSET)
26024         (STMT_VINFO_DR_STEP, STMT_VINFO_DR_ALIGNED_TO): Update accordingly.
26025         (STMT_VINFO_DR_WRT_VEC_LOOP): New macro.
26026         (vect_dr_behavior): New function.
26027         (vect_create_addr_base_for_vector_ref): Remove loop parameter.
26028         * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
26029         vect_dr_behavior.  Use a step_preserves_misalignment_p boolean to
26030         track whether the step preserves the misalignment.
26031         (vect_create_addr_base_for_vector_ref): Remove loop parameter.
26032         Use vect_dr_behavior.
26033         (vect_setup_realignment): Update call accordingly.
26034         (vect_create_data_ref_ptr): Likewise.  Use vect_dr_behavior.
26035         * tree-vect-loop-manip.c (vect_gen_prolog_loop_niters): Update
26036         call to vect_create_addr_base_for_vector_ref.
26037         (vect_create_cond_for_align_checks): Likewise.
26038         * tree-vect-patterns.c (vect_recog_bool_pattern): Copy
26039         STMT_VINFO_DR_WRT_VEC_LOOP as a block.
26040         (vect_recog_mask_conversion_pattern): Likewise.
26041         * tree-vect-stmts.c (compare_step_with_zero): Use vect_dr_behavior.
26042         (new_stmt_vec_info): Remove redundant zeroing.
26044 2017-07-03  Richard Earnshaw  <rearnsha@arm.com>
26046         * common/config/arm/arm-common.c (arm_be8_option): New function.
26047         * config/arm/arm-isa.h (isa_feature): Add new feature bit isa_bit_be8.
26048         (ISA_ARMv6): Add isa_bit_be8.
26049         * config/arm/arm.h (arm_be8_option): Add prototype.
26050         (BE8_SPEC_FUNCTION): New define.
26051         (EXTRA_SPEC_FUNCTIONS): Add BE8_SPEC_FUNCTION.
26052         * config/arm/arm.opt (mbig-endian): Mark as Negative of mlittle-endian.
26053         (mlittle-endian): Similarly.
26054         (mbe8, mbe32): New options.
26055         * config/arm/bpabi.h (BE8_LINK_SPEC): Call arm_be8_option.
26056         * doc/invoke.texi (ARM Options): Document -mbe8 and -mbe32.
26058 2017-07-02  Jan Hubicka  <hubicka@ucw.cz>
26060         * tree-cfg.c (gimple_find_sub_bbs): Fix profile updating.
26062 2017-07-02  Jan Hubicka  <hubicka@ucw.cz>
26064         * tree-cfgcleanup.c (want_merge_blocks_p): New function.
26065         (cleanup_tree_cfg_bb): Use it.
26066         * profile-count.h (profile_count::of_for_merging, profile_count::merge):
26067         New functions.
26068         * tree-cfg.c (gimple_merge_blocks): Use profile_count::merge.
26070 2017-07-02  Jan Hubicka  <hubicka@ucw.cz>
26072         PR bootstrap/81285
26073         * loop-doloop.c (add_test): Update profile.
26075 2017-07-03  Martin Liska  <mliska@suse.cz>
26077         PR sanitize/81040
26078         * sanopt.c (rewrite_usage_of_param): New function.
26079         (sanitize_rewrite_addressable_params): Likewise.
26080         (pass_sanopt::execute): Call rewrite_usage_of_param.
26082 2017-07-03  Richard Biener  <rguenther@suse.de>
26084         * tree-vect-loop.c (vect_create_epilog_for_reduction): Revert
26085         back to using VIEW_CONVERT_EXPR.
26087 2017-07-03  Martin Liska  <mliska@suse.cz>
26089         PR other/78366
26090         * doc/extend.texi: Document when a resolver function is
26091         generated for target_clones.
26093 2017-07-03  Martin Liska  <mliska@suse.cz>
26095         * asan.c (asan_emit_stack_protection): Unpoison just red zones
26096         and shadow memory of auto variables which are subject of
26097         use-after-scope sanitization.
26098         (asan_expand_mark_ifn): Add do set only when is_poison.
26100 2016-07-03  Richard Biener  <rguenther@suse.de>
26102         * tree-vect-loop.c (vect_analyze_loop_operations): Also analyze
26103         reduction PHIs.
26104         (vect_force_simple_reduction): Record reduction def -> phi mapping.
26105         (vectorizable_reduction): Perform reduction PHI creation when
26106         visiting a reduction PHI and adjust and simplify code generation
26107         phase of the reduction op.  Cache dts, use fold_binary, not fold_build2.
26108         (vect_transform_loop): Visit reduction PHIs.
26109         * tree-vect-slp.c (vect_get_and_check_slp_defs): Record reduction
26110         defs into the SLP tree.
26111         (vect_build_slp_tree): Reduction defs terminate the recursion.
26112         * tree-vect-stmts.c (vect_get_vec_def_for_operand_1): Allow lookup
26113         of reduction defs.
26114         (vect_get_vec_defs_for_stmt_copy): Export.
26115         (vect_get_vec_defs): Likewise.
26116         * tree-vectorizer.h (struct _stmt_vec_info): Amend reduc_def
26117         purpose.
26118         (vect_get_vec_defs_for_stmt_copy): Declare.
26119         (vect_get_vec_defs): Likewise.
26121 2017-07-03  Richard Sandiford  <richard.sandiford@linaro.org>
26123         * tree-data-ref.c (dr_analyze_innermost): Replace the "nest"
26124         parameter with a "loop" parameter and use it instead of the
26125         loop containing DR_STMT.  Don't check simple_iv when doing
26126         BB analysis.  Describe the two analysis modes in the comment.
26128 2017-07-03  Tom de Vries  <tom@codesourcery.com>
26130         PR tree-optimization/69468
26131         * tree-ssa-tail-merge.c (ignore_edge_flags): New constant.
26132         (find_same_succ_bb): Handle ignore_edge_flags.
26134 2017-07-03  Tom de Vries  <tom@codesourcery.com>
26136         PR tree-optimization/81192
26137         * tree-ssa-tail-merge.c (same_succ_hash): Use bb->loop_father->num in
26138         hash.
26139         (same_succ::equal): Don't find bbs to be equal if bb->loop_father
26140         differs.
26141         (find_same_succ_bb): Remove obsolete test on bb->loop_father->latch.
26143 2017-07-03  Tom de Vries  <tom@codesourcery.com>
26145         PR tree-optimization/81192
26146         * tree-ssa-tail-merge.c (same_succ_flush_bb): Handle
26147         BB_SAME_SUCC (bb) == NULL.
26149 2017-07-02  Jan Hubicka  <hubicka@ucw.cz>
26151         * cfgrtl.c (rtl_verify_edges): Enable checking of profile_probability
26152         consistency.
26154 2017-07-02  Jan Hubicka  <hubicka@ucw.cz>
26156         * dumpfile.c: Include profile-count.h
26157         * tree-cfg.c (gimple_duplicate_sese_tail): Drop UNUSED attributes;
26158         update profile.
26159         (insert_cond_bb): Update profile.
26160         * tree-cfg.h (insert_cond_bb): Update prototype.
26161         * tree-chkp-opt.c (chkp_optimize_string_function_calls): Update.
26162         * tree-dump.c: Do not include tree-cfg.
26164 2017-07-02  Jan Hubicka  <hubicka@ucw.cz>
26166         * bb-reorder.c (fix_up_crossing_landing_pad): Update profile.
26168 2017-07-02  Jan Hubicka  <hubicka@ucw.cz>
26170         * expect.c (dw2_build_landing_pads): Update profile of the landing pad
26171         bb.
26173 2017-07-02  Jan Hubicka  <hubicka@ucw.cz>
26175         * tree-complex.c (expand_complex_div_wide): update profile.
26177 2017-07-02  Richard Sandiford  <richard.sandiford@linaro.org>
26178             Alan Hayward  <alan.hayward@arm.com>
26179             David Sherwood  <david.sherwood@arm.com>
26181         * Makefile.in (MACHMODE_H): Remove insn-modes.h
26182         (CORETYPES_H): New define.
26183         (MOSTLYCLEANFILES): Add insn-modes-inline.h.
26184         (insn-modes-inline.h, s-modes-inline-h): New rules.
26185         (generated_files): Add insn-modes-inline.h.
26186         (RTL_BASE_H, TREE_CORE_H): Use CORETYPES_H instead of coretypes.h.
26187         (build/gensupport.o, build/ggc-none.o, build/print-rtl.o): Likewise.
26188         (build/read-md.o, build/read-rtl.o, build/rtl.o): Likewise.
26189         (build/vec.o, build/hash-table.o, build/inchash.o): Likewise.
26190         (build/gencondmd.o, build/genattr.o, build/genattr-common.o): Likewise.
26191         (build/genattrtab.o, build/genautomata.o, build/gencheck.o): Likewise.
26192         (build/gencodes.o, build/genconditions.o): Likewise.
26193         (build/genconfig.o, build/genconstants.o, build/genemit.o): Likewise.
26194         (build/genenums.o, build/genextract.o, build/genflags.o): Likewise.
26195         (build/gentarget-def.o, build/genmddeps.o, build/genopinit.o)
26196         (build/genoutput.o, build/genpeep.o, build/genpreds.o): Likewise.
26197         (build/genrecog.o, build/genmddump.o, build/genmatch.o): Likewise.
26198         (build/gencfn-macros.o, build/gcov-iov.o): Likewise.
26199         * coretypes.h: Include everything up to real.h for generators.
26200         Include insn-modes.h first.  Include wide-int-print.h after
26201         wide-int.h.  Include insn-modes-inline.h and then machmode.h.
26202         * machmode.h: Don't include insn-modes.h here.
26203         * function-tests.c: Remove includes of signop.h, machmode.h,
26204         double-int.h and wide-int.h.
26205         * rtl.h: Likewise.
26206         * gcc-rich-location.c: Remove includes of machmode.h, double-int.h
26207         and wide-int.h.
26208         * optc-save-gen.awk: Likewise.
26209         * gencheck.c (BITS_PER_UNIT): Delete dummy definition.
26210         * godump.c: Remove include of wide-int-print.h.
26211         * pretty-print.h: Likewise.
26212         * wide-int-print.cc: Likewise.
26213         * wide-int.cc: Likewise.
26214         * hash-map-tests.c: Remove include of signop.h.
26215         * hash-set-tests.c: Likewise.
26216         * rtl-tests.c: Likewise.
26217         * mkconfig.sh: Remove include of machmode.h.
26218         * genmodes.c (emit_insn_modes_h): Split emission of inline functions
26219         into...
26220         (emit_insn_modes_inline_h): ...this new function.  Emit the code
26221         into an insn-modes-inline.h header file, adding appropriate
26222         include guards and end comments.
26223         (emit_insn_modes_c_header): Remove include of machmode.h.
26224         (emit_min_insn_modes_c_header): Include coretypes.h rather than
26225         machmode.h.
26226         (main): Handle -i flag and call emit_insn_modes_inline_h when
26227         it is passed.
26229 2017-07-02  Richard Sandiford  <richard.sandiford@linaro.org>
26231         * tree-ssa-strlen.c (strinfo): Rename the length field to
26232         nonzero_chars.  Add a full_string_p field.
26233         (compare_nonzero_chars, zero_length_string_p): New functions.
26234         (get_addr_stridx): Add an offset_out parameter.
26235         Use compare_nonzero_chars.
26236         (get_stridx): Update accordingly.  Use compare_nonzero_chars.
26237         (new_strinfo): Update after above changes to strinfo.
26238         (set_endptr_and_length): Set full_string_p.
26239         (get_string_length): Update after above changes to strinfo.
26240         (unshare_strinfo): Update call to new_strinfo.
26241         (maybe_invalidate): Likewise.
26242         (get_stridx_plus_constant): Change off to unsigned HOST_WIDE_INT.
26243         Use compare_nonzero_chars and zero_string_p.  Treat nonzero_chars
26244         as a uhwi instead of an shwi.  Update after above changes to
26245         strinfo and new_strinfo.
26246         (zero_length_string): Assert that chainsi contains full strings.
26247         Use zero_length_string_p.  Update call to new_strinfo.
26248         (adjust_related_strinfos): Update after above changes to strinfo.
26249         Copy full_string_p from origsi.
26250         (adjust_last_stmt): Use zero_length_string_p.
26251         (handle_builtin_strlen): Update after above changes to strinfo and
26252         new_strinfo.  Install the lhs as the string length if the previous
26253         entry didn't describe a full string.
26254         (handle_builtin_strchr): Update after above changes to strinfo
26255         and new_strinfo.
26256         (handle_builtin_strcpy): Likewise.
26257         (handle_builtin_strcat): Likewise.
26258         (handle_builtin_malloc): Likewise.
26259         (handle_pointer_plus): Likewise.
26260         (handle_builtin_memcpy): Likewise.  Track nonzero characters
26261         that aren't necessarily followed by a nul terminator.
26262         (handle_char_store): Likewise.
26264 2017-07-02  Richard Sandiford  <richard.sandiford@linaro.org>
26266         PR tree-optimization/80769
26267         * tree-ssa-strlen.c (strinfo): Document that "stmt" is also used
26268         for malloc and calloc.  Document the new invariant that all related
26269         strinfos have delayed lengths or none do.
26270         (verify_related_strinfos): Move earlier in file.
26271         (set_endptr_and_length): New function, split out from...
26272         (get_string_length): ...here.  Also set the lengths of related
26273         strinfos.
26274         (zero_length_string): Assert that chainsi has known (rather than
26275         delayed) lengths.
26276         (adjust_related_strinfos): Likewise.
26278 2017-07-02  Richard Sandiford  <richard.sandiford@linaro.org>
26280         PR tree-optimization/81136
26281         * tree-vect-data-refs.c (vect_update_misalignment_for_peel): Only
26282         assert that two references with the same misalignment have the same
26283         compile-time misalignment if those compile-time misalignments
26284         are known.
26286 2017-07-01  Andi Kleen  <ak@linux.intel.com>
26288         * print-tree.c (print_node): Print all attributes.
26290 2017-07-01  Jan Hubicka  <hubicka@ucw.cz>
26292         * cfg.c (scale_bbs_frequencies): New function.
26293         * cfg.h (scale_bbs_frequencies): Declare it.
26294         * cfgloopanal.c (single_likely_exit): Cleanup.
26295         * cfgloopmanip.c (scale_loop_frequencies): Take profile_probability
26296         as parameter.
26297         (scale_loop_profile): Likewise.
26298         (loop_version): Likewise.
26299         (create_empty_loop_on_edge): Update.
26300         * cfgloopmanip.h (scale_loop_frequencies, scale_loop_profile,
26301         scale_loop_frequencies, scale_loop_profile, loopify,
26302         loop_version): Update prototypes.
26303         * modulo-sched.c (sms_schedule): Update.
26304         * predict.c (unlikely_executed_edge_p): Also check probability.
26305         (probably_never_executed_edge_p): Fix typo.
26306         * tree-if-conv.c (version_loop_for_if_conversion): Update.
26307         * tree-parloops.c (gen_parallel_loop): Update.
26308         * tree-ssa-loop-ivcanon.c (try_peel_loop): Update.
26309         * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Update.
26310         * tree-ssa-loop-split.c (split_loop): Update.
26311         * tree-ssa-loop-unswitch.c (tree_unswitch_loop): Update.
26312         * tree-vect-loop-manip.c (vect_do_peeling): Update.
26313         (vect_loop_versioning): Update.
26314         * tree-vect-loop.c (scale_profile_for_vect_loop): Update.
26316 2017-07-01  Jan Hubicka  <hubicka@ucw.cz>
26318         * trans-mem.c (split_bb_make_tm_edge): Update profile.
26320 2017-07-01  Jan Hubicka  <hubicka@ucw.cz>
26322         * tree-if-conv.c (combine_blocks): Use make_single_succ_edge
26323         to keep profile consistent.
26325 2017-07-01  Jan Hubicka  <hubicka@ucw.cz>
26327         * cfgrtl.c (rtl_flow_call_edges_add): Update profile.
26328         * tree-cfg.c (gimple_flow_call_edges_add): Likewise.
26329         * profile-count.h (max_safe_multiplier): Make unsigned.
26330         (profile_count::guessed_zero): New.
26332 2017-07-01  Jan Hubicka  <hubicka@ucw.cz>
26334         * bb-reorder.c (fix_up_crossing_landing_pad,
26335         fix_crossing_conditional_branches): Use make_single_succ_edge
26336         to keep profile consistent.
26338 2017-07-01  Jan Hubicka  <hubicka@ucw.cz>
26340         * tree-vect-loop.c (optimize_mask_stores): Use make_single_succ_edge
26341         to update profile.
26343 2017-07-01  Jakub Jelinek  <jakub@redhat.com>
26345         PR sanitizer/81262
26346         * bb-reorder.c (fix_up_fall_thru_edges): Move variable declarations to
26347         the right scopes, make sure cond_jump isn't preserved between multiple
26348         iterations.  Search for fallthru edge whenever there are 3+ edges and
26349         use find_fallthru_edge for it.
26351 2017-06-29  Jan Hubicka  <hubicka@ucw.cz>
26353         Patch by Alexander Monakov <amonakov@ispras.ru>
26354         * sel-sched-ir.c (compute_succs_info): Handle uninitialized
26355         probabilities consistently.
26357 2017-06-29  Jan Hubicka  <hubicka@ucw.cz>
26359         * pa.c (pa_expand_compare_and_swap_loop): Update call of
26360         emit_cmp_and_jump_insns.
26362 2017-06-29  Jan Hubicka  <hubicka@ucw.cz>
26364         PR ipa/81261
26365         * tree-inline.c (expand_call_inline): Combine profile statuses.
26367 2017-06-30  Andrew Pinski  <apinski@cavium.com>
26369         * tree-if-conv.c (predicate_scalar_phi): Update new_stmt if
26370         fold_stmt returned true.
26372 2017-06-30  Nathan Sidwell  <nathan@acm.org>
26374         * ggc.h (empty_string): Delete.
26375         * cfgexpand.c (expand_asm_stmt): Use plain "".
26376         * optabs.c (expand_asm_memory_barrier): Likewise.
26377         * stringpool.c (empty_string): Delete.
26378         (digit_vector, digit_string): Delete.
26379         (ggc_alloc_string): Use plain "", don't optimize single digit
26380         strings.  Use ggc_alloc_atomic.
26382 2017-06-30  Richard Earnshaw  <rearnsha@arm.com>
26384         * rtlanal.c (insn_rtx_cost): If a parallel contains exactly one
26385         comparison set and one other set, use the cost of the non-comparison
26386         set.
26388 2017-06-30  Nathan Sidwell  <nathan@acm.org>
26390         * ggc.h: Replace all 'static inline' with plain 'inline'.  Fix
26391         some formatting.
26393 2017-06-30  Peter Bergner  <bergner@vnet.ibm.com>
26395         * tree-cfg.c (group_case_labels_stmt): Merge scanning and compressing
26396         loops.  Remove now unneeded calls to gimple_switch_set_label() that
26397         just set removed labels to NULL_TREE.
26399 2017-06-30  Aldy Hernandez  <aldyh@redhat.com>
26401         * tree-ssanames.c (set_range_info_raw): Abstract from ...
26402         (set_range_info): ...here.  Only call set_range_info_raw if domain
26403         is useful.
26404         (set_nonzero_bits): Call set_range_info_raw.
26405         * tree-ssanames.h (set_range_info_raw): New.
26407 2017-06-30  Jakub Jelinek  <jakub@redhat.com>
26409         PR target/81225
26410         * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): For V8FI,
26411         V16FI and VI8F_256 iterators, use <store_mask_predicate> instead
26412         of nonimmediate_operand and <store_mask_constraint> instead of m
26413         for the input operand.  For V8FI iterator, always split if input
26414         is a MEM.  For V16FI and V8SF_256 iterators, don't test if both
26415         operands are MEM if <mask_applied>.  For VI4F_256 iterator, use
26416         <store_mask_predicate> instead of register_operand and
26417         <store_mask_constraint> instead of v for the input operand.  Make
26418         sure both operands aren't MEMs for if not <mask_applied>.
26420 2017-06-30  Sylvestre Ledru  <sylvestre@debian.org>
26422         * lto-wrapper.c (copy_file) Close both file descriptors before
26423         exiting normally.
26425 2017-06-30  Martin Liska  <mliska@suse.cz>
26427         PR ipa/81214
26428         * multiple_target.c (create_dispatcher_calls): Make ifunc
26429         also for function that don't have calls or are not referenced.
26431 2017-06-30  Richard Biener  <rguenther@suse.de>
26433         * tree-vect-slp.c (vect_slp_analyze_node_operations): Only
26434         analyze the first scalar stmt.  Move vector type computation
26435         for the BB case here from ...
26436         * tree-vect-stmts.c (vect_analyze_stmt): ... here.  Guard
26437         live operation processing in the SLP case properly.
26439 2017-06-30  Richard Biener  <rguenther@suse.de>
26441         * graph.c (draw_cfg_node_succ_edges): Fix broken dot syntax.
26443 2017-06-30  Martin Liska  <mliska@suse.cz>
26445         PR sanitizer/81021
26446         * tree-eh.c (lower_resx): Call BUILT_IN_ASAN_HANDLE_NO_RETURN
26447         before BUILT_IN_UNWIND_RESUME when ASAN is used.
26449 2017-06-30  Yvan Roux  <yvan.roux@linaro.org>
26451         * doc/invoke.texi (AArch64): Add missing options and remove redundant
26452         ones.
26454 2017-06-30  Richard Biener  <rguenther@suse.de>
26456         PR tree-optimization/81249
26457         * tree-vect-loop.c (vect_create_epilog_for_reduction): Convert
26458         condition reduction result to original scalar type.
26460 2017-06-30  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
26462         * profile-count.h (enum profile_quality): Fix typos and whitespace
26463         issues.
26465 2017-06-30  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
26467         * config/s390/s390.c (s390_expand_setmem): Adjust to the new data
26468         type for branch probabilities.
26470 2017-06-29  Julian Brown  <julian@codesourcery.com>
26471             Naveen H.S  <Naveen.Hurugalawadi@cavium.com>
26473         * config/aarch64/aarch64-fusion-pairs.def: Add ALU_BRANCH entry.
26474         * config/aarch64/aarch64.c (AARCH64_FUSE_ALU_BRANCH): New fusion type.
26475         (thunderx2t99_tunings): Set AARCH64_FUSE_ALU_BRANCH flag.
26476         (aarch_macro_fusion_pair_p): Add support for AARCH64_FUSE_ALU_BRANCH.
26478 2017-06-29  Naveen H.S  <Naveen.Hurugalawadi@cavium.com>
26480         * config/aarch64/aarch64.c (aarch_macro_fusion_pair_p): Push the
26481         check for CC usage into AARCH64_FUSE_CMP_BRANCH.
26482         * config/i386/i386.c (ix86_macro_fusion_pair_p): Push the check for
26483         CC usage from generic code to here.
26484         * sched-deps.c (sched_macro_fuse_insns): Move the condition for
26485         CC usage into the target macros.
26487 2017-06-29  Maya Rashish  <coypu@sdf.org>
26489         * config/netbsd.h (NETBSD_LIB_SPEC): Add -lc when creating shared
26490         objects.
26492 2017-06-29  Jan Hubicka  <hubicka@ucw.cz>
26494         * arm/arm-builtins.c: Include profile-count.h
26495         * except.c (sjlj_emit_function_enter): Use
26496         profile_probability::unlikely.
26498 2017-06-29  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
26500         * config/rs6000/rs6000.c (toc_relative_expr_p): Make tocrel_base
26501         and tocrel_offset be pointer args rather than implicitly using
26502         static versions.
26503         (legitimate_constant_pool_address_p, rs6000_emit_move,
26504         const_load_sequence_p, adjust_vperm): Add local tocrel_base and
26505         tocrel_offset and use in toc_relative_expr_p call.
26506         (print_operand, print_operand_address): Use static tocrel_base_oac
26507         and tocrel_offset_oac.
26508         (rs6000_output_addr_const_extra): Use static tocrel_base_oac and
26509         tocrel_offset_oac.
26511 2017-06-29  Maya Rashish  <coypu@sdf.org>
26513         * config/vax/builtins.md (ffssi2_internal): Correct constraint.
26515 2017-06-29  Eric Botcazou  <ebotcazou@adacore.com>
26517         * expr.c (expand_expr) <normal_inner_ref>: When testing for unaligned
26518         objects, take into account only the alignment of 'op0' and 'mode1' if
26519         'op0' is a MEM.
26521 2017-06-29  Steve Ellcey  <sellcey@cavium.com>
26523         * ccmp.c (ccmp_tree_comparison_p): New function.
26524         (ccmp_candidate_p): Update to use above function.
26525         (get_compare_parts): New function.
26526         (expand_ccmp_next): Update to use new functions.
26527         (expand_ccmp_expr_1): Take tree arg instead of gimple, update to use
26528         new functions.
26529         (expand_ccmp_expr): Pass tree instead of gimple to expand_ccmp_expr_1,
26530         take mode as argument.
26531         * ccmp.h (expand_ccmp_expr): Add mode as argument.
26532         * expr.c (expand_expr_real_1): Pass mode as argument.
26534 2017-06-29  Segher Boessenkool  <segher@kernel.crashing.org>
26536         * combine.c (combine_instructions): Print insns to dump_file, together
26537         with their costs.
26539 2017-06-29  Jan Hubicka  <hubicka@ucw.cz>
26541         * asan.c (asan_emit_stack_protection): Update.
26542         (create_cond_insert_point): Update.
26543         * auto-profile.c (afdo_propagate_circuit): Update.
26544         * basic-block.h (struct edge_def): Turn probability to
26545         profile_probability.
26546         (EDGE_FREQUENCY): Update.
26547         * bb-reorder.c (find_traces_1_round): Update.
26548         (better_edge_p): Update.
26549         (sanitize_hot_paths): Update.
26550         * cfg.c (unchecked_make_edge): Initialize probability to uninitialized.
26551         (make_single_succ_edge): Update.
26552         (check_bb_profile): Update.
26553         (dump_edge_info): Update.
26554         (update_bb_profile_for_threading): Update.
26555         * cfganal.c (connect_infinite_loops_to_exit): Initialize new edge
26556         probabilitycount to 0.
26557         * cfgbuild.c (compute_outgoing_frequencies): Update.
26558         * cfgcleanup.c (try_forward_edges): Update.
26559         (outgoing_edges_match): Update.
26560         (try_crossjump_to_edge): Update.
26561         * cfgexpand.c (expand_gimple_cond): Update make_single_succ_edge.
26562         (expand_gimple_tailcall): Update.
26563         (construct_init_block): Use make_single_succ_edge.
26564         (construct_exit_block): Use make_single_succ_edge.
26565         * cfghooks.c (verify_flow_info): Update.
26566         (redirect_edge_succ_nodup): Update.
26567         (split_edge): Update.
26568         (account_profile_record): Update.
26569         * cfgloopanal.c (single_likely_exit): Update.
26570         * cfgloopmanip.c (scale_loop_profile): Update.
26571         (set_zero_probability): Remove.
26572         (duplicate_loop_to_header_edge): Update.
26573         * cfgloopmanip.h (loop_version): Update prototype.
26574         * cfgrtl.c (try_redirect_by_replacing_jump): Update.
26575         (force_nonfallthru_and_redirect): Update.
26576         (update_br_prob_note): Update.
26577         (rtl_verify_edges): Update.
26578         (purge_dead_edges): Update.
26579         (rtl_lv_add_condition_to_bb): Update.
26580         * cgraph.c: (cgraph_edge::redirect_call_stmt_to_calle): Update.
26581         * cgraphunit.c (init_lowered_empty_function): Update.
26582         (cgraph_node::expand_thunk): Update.
26583         * cilk-common.c: Include profile-count.h
26584         * dojump.c (inv): Remove.
26585         (jumpifnot): Update.
26586         (jumpifnot_1): Update.
26587         (do_jump_1): Update.
26588         (do_jump): Update.
26589         (do_jump_by_parts_greater_rtx): Update.
26590         (do_compare_rtx_and_jump): Update.
26591         * dojump.h (jumpifnot, jumpifnot_1, jumpif_1, jumpif, do_jump,
26592         do_jump_1. do_compare_rtx_and_jump): Update prototype.
26593         * dwarf2cfi.c: Include profile-count.h
26594         * except.c (dw2_build_landing_pads): Use make_single_succ_edge.
26595         (sjlj_emit_dispatch_table): Likewise.
26596         * explow.c: Include profile-count.h
26597         * expmed.c (emit_store_flag_force): Update.
26598         (do_cmp_and_jump): Update.
26599         * expr.c (compare_by_pieces_d::generate): Update.
26600         (compare_by_pieces_d::finish_mode): Update.
26601         (emit_block_move_via_loop): Update.
26602         (store_expr_with_bounds): Update.
26603         (store_constructor): Update.
26604         (expand_expr_real_2): Update.
26605         (expand_expr_real_1): Update.
26606         * expr.h (try_casesi, try_tablejump): Update prototypes.
26607         * gimple-pretty-print.c (dump_probability): Update.
26608         (dump_profile): New.
26609         (dump_gimple_label): Update.
26610         (dump_gimple_bb_header): Update.
26611         * graph.c (draw_cfg_node_succ_edges): Update.
26612         * hsa-gen.c (convert_switch_statements): Update.
26613         * ifcvt.c (cheap_bb_rtx_cost_p): Update.
26614         (find_if_case_1): Update.
26615         (find_if_case_2): Update.
26616         * internal-fn.c (expand_arith_overflow_result_store): Update.
26617         (expand_addsub_overflow): Update.
26618         (expand_neg_overflow): Update.
26619         (expand_mul_overflow): Update.
26620         (expand_vector_ubsan_overflow): Update.
26621         * ipa-cp.c (good_cloning_opportunity_p): Update.
26622         * ipa-split.c (split_function): Use make_single_succ_edge.
26623         * ipa-utils.c (ipa_merge_profiles): Update.
26624         * loop-doloop.c (add_test): Update.
26625         (doloop_modify): Update.
26626         * loop-unroll.c (compare_and_jump_seq): Update.
26627         (unroll_loop_runtime_iterations): Update.
26628         * lra-constraints.c (lra_inheritance): Update.
26629         * lto-streamer-in.c (input_cfg): Update.
26630         * lto-streamer-out.c (output_cfg): Update.
26631         * mcf.c (adjust_cfg_counts): Update.
26632         * modulo-sched.c (sms_schedule): Update.
26633         * omp-expand.c (expand_omp_for_init_counts): Update.
26634         (extract_omp_for_update_vars): Update.
26635         (expand_omp_ordered_sink): Update.
26636         (expand_omp_for_ordered_loops): Update.
26637         (expand_omp_for_generic): Update.
26638         (expand_omp_for_static_nochunk): Update.
26639         (expand_omp_for_static_chunk): Update.
26640         (expand_cilk_for): Update.
26641         (expand_omp_simd): Update.
26642         (expand_omp_taskloop_for_outer): Update.
26643         (expand_omp_taskloop_for_inner): Update.
26644         * omp-simd-clone.c (simd_clone_adjust): Update.
26645         * optabs.c (expand_doubleword_shift): Update.
26646         (expand_abs): Update.
26647         (emit_cmp_and_jump_insn_1): Update.
26648         (expand_compare_and_swap_loop): Update.
26649         * optabs.h (emit_cmp_and_jump_insns): Update prototype.
26650         * predict.c (predictable_edge_p): Update.
26651         (edge_probability_reliable_p): Update.
26652         (set_even_probabilities): Update.
26653         (combine_predictions_for_insn): Update.
26654         (combine_predictions_for_bb): Update.
26655         (propagate_freq): Update.
26656         (estimate_bb_frequencies): Update.
26657         (force_edge_cold): Update.
26658         * profile-count.c (profile_count::dump): Add missing space into dump.
26659         (profile_count::debug): Add newline.
26660         (profile_count::differs_from_p): Explicitly convert to unsigned.
26661         (profile_count::stream_in): Update.
26662         (profile_probability::dump): New member function.
26663         (profile_probability::debug): New member function.
26664         (profile_probability::differs_from_p): New member function.
26665         (profile_probability::differs_lot_from_p): New member function.
26666         (profile_probability::stream_in): New member function.
26667         (profile_probability::stream_out): New member function.
26668         * profile-count.h (profile_count_quality): Rename to ...
26669         (profile_quality): ... this one.
26670         (profile_probability): New.
26671         (profile_count): Update.
26672         * profile.c (compute_branch_probabilities): Update.
26673         * recog.c (peep2_attempt): Update.
26674         * sched-ebb.c (schedule_ebbs): Update.
26675         * sched-rgn.c (find_single_block_region): Update.
26676         (compute_dom_prob_ps): Update.
26677         (schedule_region): Update.
26678         * sel-sched-ir.c (compute_succs_info): Update.
26679         * stmt.c (struct case_node): Update.
26680         (do_jump_if_equal): Update.
26681         (get_outgoing_edge_probs): Update.
26682         (conditional_probability): Update.
26683         (emit_case_dispatch_table): Update.
26684         (expand_case): Update.
26685         (expand_sjlj_dispatch_table): Update.
26686         (emit_case_nodes): Update.
26687         * targhooks.c: Update.
26688         * tracer.c (better_p): Update.
26689         (find_best_successor): Update.
26690         * trans-mem.c (expand_transaction): Update.
26691         * tree-call-cdce.c: Update.
26692         * tree-cfg.c (gimple_split_edge): Upate.
26693         (move_sese_region_to_fn): Upate.
26694         * tree-cfgcleanup.c (cleanup_control_expr_graph): Upate.
26695         * tree-eh.c (lower_resx): Upate.
26696         (cleanup_empty_eh_move_lp): Upate.
26697         * tree-if-conv.c (version_loop_for_if_conversion): Update.
26698         * tree-inline.c (copy_edges_for_bb): Update.
26699         (copy_cfg_body): Update.
26700         * tree-parloops.c (gen_parallel_loop): Update.
26701         * tree-profile.c (gimple_gen_ic_func_profiler): Update.
26702         (gimple_gen_time_profiler): Update.
26703         * tree-ssa-dce.c (remove_dead_stmt): Update.
26704         * tree-ssa-ifcombine.c (update_profile_after_ifcombine): Update.
26705         * tree-ssa-loop-im.c (execute_sm_if_changed): Update.
26706         * tree-ssa-loop-ivcanon.c (remove_exits_and_undefined_stmts): Update.
26707         (unloop_loops): Update.
26708         (try_peel_loop): Update.
26709         * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Update.
26710         * tree-ssa-loop-split.c (connect_loops): Update.
26711         (split_loop): Update.
26712         * tree-ssa-loop-unswitch.c (tree_unswitch_loop): Update.
26713         (hoist_guard): Update.
26714         * tree-ssa-phionlycprop.c (propagate_rhs_into_lhs): Update.
26715         * tree-ssa-phiopt.c (replace_phi_edge_with_variable): Update.
26716         (value_replacement): Update.
26717         * tree-ssa-reassoc.c (branch_fixup): Update.
26718         * tree-ssa-tail-merge.c (replace_block_by): Update.
26719         * tree-ssa-threadupdate.c (remove_ctrl_stmt_and_useless_edges): Update.
26720         (create_edge_and_update_destination_phis): Update.
26721         (compute_path_counts): Update.
26722         (recompute_probabilities): Update.
26723         (update_joiner_offpath_counts): Update.
26724         (freqs_to_counts_path): Update.
26725         (duplicate_thread_path): Update.
26726         * tree-switch-conversion.c (hoist_edge_and_branch_if_true): Update.
26727         (struct switch_conv_info): Update.
26728         (gen_inbound_check): Update.
26729         * tree-vect-loop-manip.c (slpeel_add_loop_guard): Update.
26730         (vect_do_peeling): Update.
26731         (vect_loop_versioning): Update.
26732         * tree-vect-loop.c (scale_profile_for_vect_loop): Update.
26733         (optimize_mask_stores): Update.
26734         * ubsan.c (ubsan_expand_null_ifn): Update.
26735         * value-prof.c (gimple_divmod_fixed_value): Update.
26736         (gimple_divmod_fixed_value_transform): Update.
26737         (gimple_mod_pow2): Update.
26738         (gimple_mod_pow2_value_transform): Update.
26739         (gimple_mod_subtract): Update.
26740         (gimple_mod_subtract_transform): Update.
26741         (gimple_ic): Update.
26742         (gimple_stringop_fixed_value): Update.
26743         (gimple_stringops_transform): Update.
26744         * value-prof.h: Update.
26746 2017-06-29  Carl Love  <cel@us.ibm.com>
26748         * config/rs6000/rs6000-c.c: Add support for built-in functions
26749         vector signed int vec_signed (vector float);
26750         vector signed long long vec_signed (vector double);
26751         vector signed int vec_signed2 (vector double, vector double);
26752         vector signed int vec_signede (vector double);
26753         vector signed int vec_signedo (vector double);
26754         * config/rs6000/rs6000.c (rs6000_generate_vsigned2_code): Add
26755         instruction generator.
26756         * config/rs6000/vsx.md (UNSPEC_VSX_XVCVSPSXWS, UNSPEC_VSX_XVCVSPSXDS,
26757         UNSPEC_VSX_VSIGNED2): Add UNSPECS.
26758         (vsx_xvcvspsxws, vsx_xvcvdpuxds_scale, vsx_xvcvspuxws, vsigned2_v2df):
26759         Add define_insn.
26760         (vsignedo_v2df, vsignede_v2df, vunsigned2_v2df, vunsignedo_v2df,
26761         vunsignede_v2df): Add define_expands.
26762         * config/rs6000/rs6000-builtin.def (VEC_SIGNED, VEC_UNSIGNED,
26763         VEC_SIGNED2, VEC_UNSIGNED2, VEC_SIGNEDE, VEC_UNSIGNEDE, VEC_SIGNEDO,
26764         VEC_UNSIGNEDO): Add definitions.
26765         * config/vsx.md (UNSPEC_VSX_XVCVSPSXWS, UNSPEC_VSX_XVCVSPSXDS,
26766         UNSPEC_VSX_VSIGNED2): Add UNSPECs.
26767         (vsx_xvcvspsxws, vsx_xvcvspuxws): Add define_insn.
26768         (vsigned2_v2df, vsigendo_v2df, vsignede_v2df,
26769         vunsigned2_v2df, vunsignedo_v2df, vunsignede_v2df): Add define_expands.
26770         * config/rs6000/altivec.h (vec_signed, vec_signed2,
26771         vec_signede and vec_signedo, vec_unsigned, vec_unsigned2,
26772         vec_unsignede, vec_unsignedo): Add builtin defines.
26773         * config/rs6000-protos.h (rs6000_generate_vsigned2_code): Add extern
26774         declaration.
26775         * doc/extend.texi: Update the built-in documentation file for the
26776         new built-in functions.
26778 2017-06-29  Richard Biener  <rguenther@suse.de>
26780         * tree-vect-loop.c (vect_analyze_scalar_cycles_1): Do not add
26781         reduction chains to LOOP_VINFO_REDUCTIONS.
26782         * tree-vect-slp.c (vect_analyze_slp): Continue looking for
26783         SLP reductions after processing reduction chains.
26785 2017-06-29  Nathan Sidwell  <nathan@acm.org>
26787         * builtins.c (fold_builtin_FUNCTION): Use
26788         lang_hooks.decl_printable_name.
26790 2017-06-29  Peter Bergner  <bergner@vnet.ibm.com>
26792         PR middle-end/81194
26793         * cfgexpand.c (expand_gimple_stmt_1): Handle switch statements
26794         with only one label.
26795         * stmt.c (expand_case): Assert NCASES is greater than one.
26797 2017-06-29  Richard Biener  <rguenther@suse.de>
26799         * tree-cfg.c (group_case_labels_stmt): Return whether we changed
26800         anything.
26801         (group_case_labels): Likewise.
26802         (find_taken_edge): Push sanity checking on val to workers...
26803         (find_taken_edge_cond_expr): ... here
26804         (find_taken_edge_switch_expr): ... and here, handle cases
26805         with just a default label.
26806         * tree-cfg.h (group_case_labels_stmt): Adjust prototype.
26807         (group_case_labels): Likewise.
26808         * tree-cfgcleanup.c (execute_cleanup_cfg_post_optimizing): When
26809         group_case_labels does anything cleanup the CFG again.
26811 2017-06-29  Bin Cheng  <bin.cheng@arm.com>
26813         PR tree-optimization/81196
26814         * tree-ssa-loop-niter.c (number_of_iterations_cond): Handle loop
26815         exit condition comparing two IVs.
26817 2017-06-29  Richard Earnshaw  <rearnsha@arm.com>
26819         * config/arm/parsecpu.awk (gen_comm_data): Add initializer for
26820         profile to the dummy entry at the end of the list of architectures.
26821         * config/arm/arm-cpu-cdata.h: Regenerated.
26823 2017-06-29  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
26824             Michael Collison <michael.collison@arm.com>
26826         PR target/70119
26827         * config/aarch64/aarch64.md (*aarch64_<optab>_reg_<mode>3_mask1):
26828         New pattern.
26829         (*aarch64_reg_<mode>3_neg_mask2): New pattern.
26830         (*aarch64_reg_<mode>3_minus_mask): New pattern.
26831         (*aarch64_<optab>_reg_di3_mask2): New pattern.
26832         * config/aarch64/aarch64.c (aarch64_rtx_costs): Account for cost
26833         of shift when the shift amount is masked with constant equal to
26834         the size of the mode.
26835         * config/aarch64/predicates.md (subreg_lowpart_operator): New
26836         predicate.
26838 2017-06-29  Martin Liska  <mliska@suse.cz>
26840         * config/i386/i386.opt: Change range from [1,5] to [0,5].
26842 2017-06-29  Yury Gribov  <tetra2005@gmail.com>
26844         PR bootstrap/80565
26845         * ipa-cp.c (allocate_and_init_ipcp_value): Add initialization
26846         code.
26847         * ipa-inline.h
26848         (edge_growth_cache_entry::edge_growth_cache_entry): New
26849         function.
26850         (reset_edge_growth_cache): Update to use constructor.
26852 2017-06-28  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
26854         * config/aarch64/aarch64.h (AARCH64_EXPAND_ALIGNMENT): New.
26855         (DATA_ALIGNMENT): Update to use AARCH64_EXPAND_ALIGNMENT.
26856         (LOCAL_ALIGNMENT): Update to use AARCH64_EXPAND_ALIGNMENT.
26858 2017-06-28  Sebastian Peryt  <sebastian.peryt@intel.com>
26860         * config/i386/avx512vlintrin.h (_mm256_permutexvar_epi64)
26861         (_mm256_permutexvar_epi32, _mm256_permutex_epi64): New intrinsics.
26863 2017-06-28  Szabolcs Nagy  <szabolcs.nagy@arm.com>
26865         * config.gcc (*-linux-musl*): Add t-musl tmake_file.
26866         (*-linux-uclibc*): Add t-uclibc tmake_file.
26867         * config/t-musl: New.
26868         * config/t-uclibc: New.
26870 2017-06-28  Richard Earnshaw  <rearnsha@arm.com>
26872         * config/arm/parsecpu.awk (profile): Parse new keyword in an arch
26873         context.
26874         (gen_comm_data): Emit architectural setting of arch_prof.
26875         * config/arm/arm-cpus.in (armv6-m, armv6s-m, armv7-a, armv7ve): Set the
26876         profile.
26877         (armv7-r, armv7-m, armv7e-m, armv8-a, armv8.1-a, armv8.2-a): Likewise.
26878         (armv8-m.base, armv8-m.main): Likewise.
26879         * arm-protos.h (arm_build_target): Add profile field.
26880         (arch_option): Likewise.
26881         * config/arm/arm.c (arm_configure_build_target): Copy the profile to
26882         the active target.
26883         * config/arm/arm.h (TARGET_ARM_ARCH_PROFILE): Use
26884         arm_active_target.profile.
26886 2017-06-28  Richard Biener  <rguenther@suse.de>
26888         PR middle-end/81227
26889         * fold-const.c (negate_expr_p): Use TYPE_UNSIGNED, not
26890         TYPE_OVERFLOW_WRAPS.
26891         * match.pd (negate_expr_p): Likewise.
26892         * tree-ssa-reassoc.c (optimize_range_tests_diff): Use
26893         fold_build2, not fold_binary.
26895 2017-06-28  Wilco Dijkstra  <wdijkstr@arm.com>
26897         * config/aarch64/aarch64 (aarch64_expand_mov_immediate):
26898         Convert memory address to Pmode.
26899         (aarch64_print_operand): Assert MEM operands are always Pmode.
26901 2017-06-28  Wilco Dijkstra  <wdijkstr@arm.com>
26903         PR target/79665
26904         * config/arm/aarch-common.c (arm_no_early_alu_shift_dep):
26905         Remove redundant if.
26906         (aarch_forward_to_shift_is_not_shifted_reg): Remove.
26907         * config/arm/aarch-common-protos.h
26908         (aarch_forward_to_shift_is_not_shifted_re): Remove.
26909         * config/arm/cortex-a53.md: Use arm_no_early_alu_shift_dep in bypass.
26911 2017-06-28  Michael Meissner  <meissner@linux.vnet.ibm.com>
26913         PR ipa/81238
26914         * multiple_target.c (create_dispatcher_calls): Set the default
26915         clone to be static, not public.
26917 2017-06-28  Richard Biener  <rguenther@suse.de>
26919         * tree-vect-loop.c (vectorizable_reduction): Move special
26920         cond reduction IV var creation ...
26921         (vect_create_epilog_for_reduction): ... here.  Remove induction_index
26922         parameter.  Use STMT_VINFO_VECTYPE.
26923         * tree-vect-slp.c (vect_get_constant_vectors): Properly reset
26924         constant_p.
26926 2017-06-28  Martin Liska  <mliska@suse.cz>
26928         PR ipa/81128
26929         * ipa-visibility.c (non_local_p): Handle visibility.
26931 2017-06-28  Martin Liska  <mliska@suse.cz>
26933         PR driver/79659
26934         * common.opt: Add IntegerRange to various options.
26935         * opt-functions.awk (integer_range_info): New function.
26936         * optc-gen.awk: Add integer_range_info to cl_options struct.
26937         * opts-common.c (decode_cmdline_option): Handle
26938         CL_ERR_INT_RANGE_ARG.
26939         (cmdline_handle_error): Likewise.
26940         * opts.c (print_filtered_help): Show valid interval in
26941         when --help is provided.
26942         * opts.h (struct cl_option): Add range_min and range_max fields.
26943         * config/i386/i386.opt: Add IntegerRange for -mbranch-cost.
26945 2017-06-28  Marc Glisse  <marc.glisse@inria.fr>
26947         * match.pd ((X & ~Y) | (~X & Y)): Generalize to + and ^.
26948         (x * C EQ/NE y * C): New transformation.
26950 2017-06-28  Christophe Lyon  <christophe.lyon@linaro.org>
26952         * genmultilib (combination_space): Accept '+' in option names.
26954 2017-06-28  Martin Liska  <mliska@suse.cz>
26956         PR sanitizer/81224
26957         * asan.c (instrument_derefs): Bail out inner references
26958         that are hard register variables.
26960 2017-06-28  Jakub Jelinek  <jakub@redhat.com>
26962         PR target/81175
26963         * config/i386/i386.c (ix86_init_mmx_sse_builtins): Use def_builtin
26964         rather than def_builtin_pure for __builtin_ia32_gatherpf*.
26966 2017-06-28  Richard Biener  <rguenther@suse.de>
26968         * tree-vectorizer.h (vect_get_vec_defs): Remove.
26969         (vect_get_slp_defs): Adjust.
26970         * tree-vect-loop.c (get_initial_defs_for_reduction): Split
26971         out from ...
26972         * tree-vect-slp.c (vect_get_constant_vectors): ... here and
26973         simplify.
26974         * tree-vect-loop.c (vect_create_epilog_for_reduction): Use
26975         get_initial_defs_for_reduction instead of vect_get_vec_defs.
26976         (vectorizable_reduction): Adjust.
26977         * tree-vect-slp.c (vect_get_constant_vectors): Remove reduction
26978         handling.
26979         (vect_get_slp_defs): Likewise.
26980         * tree-vect-stmts.c (vect_get_vec_defs): Make static and adjust.
26981         (vectorizable_bswap): Adjust.
26982         (vectorizable_call): Likewise.
26983         (vectorizable_conversion): Likewise.
26984         (vectorizable_assignment): Likewise.
26985         (vectorizable_shift): Likewise.
26986         (vectorizable_operation): Likewise.
26987         (vectorizable_store): Likewise.
26988         (vectorizable_condition): Likewise.
26989         (vectorizable_comparison): Likewise.
26991 2017-06-28  Michael Collison  <michael.collison@arm.com>
26993         PR target/68535
26994         * config/arm/arm.c (gen_ldm_seq): Remove last unnecessary
26995         set of base_reg
26996         (arm_gen_movmemqi): Removed unused variable 'i'.
26997         Convert 'for' loop into 'while' loop.
26998         (arm_expand_prologue): Remove last unnecessary set of insn.
26999         (thumb_pop): Remove unused variable 'pushed_words'.
27000         (thumb_exit): Remove last unnecessary set of regs_to_pop.
27002 2017-06-28  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
27004         * config/s390/predicates.md: Use s390_rel_address_ok_p.
27005         * config/s390/s390-protos.h: Add prototype of
27006         s390_rel_address_ok_p.
27007         * config/s390/s390.c (s390_got_symbol): New function.
27008         (s390_rel_address_ok_p): New function.
27009         (legitimize_pic_address): Use s390_rel_address_ok_p.
27010         (s390_load_got): Use s390_got_symbol.
27011         (s390_option_override): Issue error if
27012         -mno-pic-data-is-text-relative is used without -fpic/-fPIC.
27013         * config/s390/s390.h (TARGET_DEFAULT_PIC_DATA_IS_TEXT_RELATIVE):
27014         New macro.
27015         * config/s390/s390.opt: New option mpic-data-is-text-relative.
27017 2017-06-27  Andrew Pinski  <apinski@cavium.com>
27019         * match.pd (X >/>=/</<= 0 ? 1.0 : -1.0): New patterns.
27020         (X * copysign (1.0, X)): New pattern.
27021         (X * copysign (1.0, -X)): New pattern.
27022         (copysign (-1.0, CST)): New pattern.
27024 2017-06-27  Joseph Myers  <joseph@codesourcery.com>
27026         * genmultilib (combination_space): Remove variable.
27027         Validate reuse rules against regular expression for any sequence
27028         of multilib options in any order.
27030 2017-06-27  Michael Collison  <michael.collison@arm.com>
27032         * config/aarch64/aarch64-simd.md (aarch64_combine<mode>): Directly
27033         call aarch64_split_simd_combine.
27034         * (aarch64_combine_internal<mode>): Delete pattern.
27035         * config/aarch64/aarch64.c (aarch64_split_simd_combine):
27036         Allow register and subreg operands.
27038 2017-06-27  Jerome Lambourg  <lambourg@adacore.com>
27040         * config/i386/vxworks.h (ASM_SPEC): Remove definition. No target
27041         specific need, just fallback on defaults.
27042         (ASM_OUTPUT_ALIGNED_BSS): Add #undef before #define.
27044 2017-06-27  Jerome Lambourg  <lambourg@adacore.com>
27045             Olivier Hainque  <hainque@adacore.com>
27047         * config/i386/vxworks.h (DBX_REGISTER_NUMBER): Pick distinct
27048         map for 64bits.
27049         (TARGET_OS_CPP_BUILTINS): builtin_define CPU to X86_64 for 64bit
27050         targets. Pick a default if no particular attempt applied.
27051         (STACK_CHECK_PROTECT): Double for 64bit targets, which have
27052         larger contexts.
27054 2017-06-27  Jerome Lambourg  <lambourg@adacore.com>
27056         * config.gcc (i*86-wrs-vxworks7): Handle new acceptable triplet.
27057         (x86_64-wrs-vxworks7): Likewise.
27059 2017-06-27  Marek Polacek  <polacek@redhat.com>
27061         PR sanitizer/81223
27062         * ubsan.c (instrument_null): Check get_base_address's result for null.
27064 2017-06-27  Marc Glisse  <marc.glisse@inria.fr>
27066         * match.pd ((A+-B)+(C-A), (A+B)-(A-C)): New transformations.
27068 2017-06-27  Marc Glisse  <marc.glisse@inria.fr>
27070         * builtin-types.def (BT_FENV_T_PTR, BT_CONST_FENV_T_PTR,
27071         BT_FEXCEPT_T_PTR, BT_CONST_FEXCEPT_T_PTR): New primitive types.
27072         (BT_FN_INT_FENV_T_PTR, BT_FN_INT_CONST_FENV_T_PTR,
27073         BT_FN_INT_FEXCEPT_T_PTR_INT, BT_FN_INT_CONST_FEXCEPT_T_PTR_INT):
27074         New function types.
27075         * builtins.def (BUILT_IN_FECLEAREXCEPT, BUILT_IN_FEGETENV,
27076         BUILT_IN_FEGETEXCEPTFLAG, BUILT_IN_FEGETROUND,
27077         BUILT_IN_FEHOLDEXCEPT, BUILT_IN_FERAISEEXCEPT,
27078         BUILT_IN_FESETENV, BUILT_IN_FESETEXCEPTFLAG,
27079         BUILT_IN_FESETROUND, BUILT_IN_FETESTEXCEPT,
27080         BUILT_IN_FEUPDATEENV): New builtins.
27081         * tree-core.h (TI_FENV_T_PTR_TYPE, TI_CONST_FENV_T_PTR_TYPE,
27082         TI_FEXCEPT_T_PTR_TYPE, TI_CONST_FEXCEPT_T_PTR_TYPE): New entries.
27083         * tree.h (fenv_t_ptr_type_node, const_fenv_t_ptr_type_node,
27084         fexcept_t_ptr_type_node, const_fexcept_t_ptr_type_node): New
27085         macros.
27086         (builtin_structptr_types): Adjust size.
27087         * tree.c (builtin_structptr_types): Add four entries.
27089 2017-06-27  Jerome Lambourg  <lambourg@adacore.com>
27090             Olivier Hainque  <hainque@adacore.com>
27092         * config/vxworks.h (VXWORKS_LIB_SPEC): Incorporate ...
27093         (TLS_SYM): New local macro, forcing reference to __tls__ on
27094         link command lines for VxWorks 7 RTPs, triggering initialization
27095         of tlsLib.
27096         (VXWORKS_HAVE_TLS): New macro. State whether the target VxWorks
27097         OS features TLS support, true for RTPs on VxWorks 7.
27098         * config/vxworks.c (vxworks_override_options): Setup emutls
27099         accordingly.
27101 2017-06-27  Jakub Jelinek  <jakub@redhat.com>
27103         * predict.c (test_prediction_value_range): Use -1U instead of -1
27104         to avoid narrowing conversion warning.
27105         * dumpfile.c (dump_options): Wrap all value into dump_flags_t cast
27106         to avoid narrowing conversion warning.
27107         * opt-functions.awk (var_ref): Return (unsigned short) -1 instead of
27108         -1.
27109         * optc-gen.awk (END): Expect (unsigned short) -1 instead of -1.
27111 2017-06-27  Jerome Lambourg  <lambourg@adacore.com>
27113         * config/vxworks.h (VXWORKS_LIBS_RTP): Alternative definition for
27114         64bit configurations.
27115         (PTR_DIFF_TYPE): Alternative definition for TARGET_LP64.
27116         (SIZE_TYPE): Likewise.
27117         * config/vxworks.c (vxworks_emutls_var_fields): Use
27118         long_unsigned_type_node instead of unsigned_type_node as the offset
27119         field type, which is "pointer" mode in emutls.c.
27121 2017-06-27  Jakub Jelinek  <jakub@redhat.com>
27123         PR sanitizer/81209
27124         * ubsan.c (ubsan_encode_value): Initialize DECL_CONTEXT on var.
27126         PR middle-end/81207
27127         * gimple-fold.c (replace_call_with_call_and_fold): Handle
27128         gimple_vuse copying separately from gimple_vdef copying.
27130 2017-06-27  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
27132         * value-prof.c (free_hist): Remove call to memset and the enclosing if
27133         condition.
27135 2017-06-26  Jerome Lambourg  <lambourg@adacore.com>
27136             Olivier Hainque  <hainque@adacore.com>
27138         * config.gcc (*-*-vxworks*): Add TARGET_VXWORKS7=1 to tm_defines
27139         for all vxworks7 targets.
27140         * config/vxworks.h (TARGET_VXWORKS7): If not defined, define to 0.
27141         (VXWORKS_ADDITIONAL_CPP_SPEC): Alternative definition for VXWORKS7.
27142         (VXWORKS_LIBS_RTP, VXWORKS_LIBS_RTP_DIR): New macros, allowing
27143         variations for VX6/VX7 and 32/64bits later on in ...
27144         (VXWORKS_LIB_SPEC): Leverage new macros.
27145         (VXWORKS_OS_CPP_BUILTINS): Define _VSB_CONFIG_FILE for VXWORKS7,
27146         as well as _ALLOW_KEYWORD_MACROS when "inline" is not a keyword.
27148 2017-06-26  Jerome Lambourg  <lambourg@adacore.com>
27150         * config/vxworks.h (VXWORKS_OS_CPP_BUILTINS): builtin_define
27151         _VX_TOOL_FAMILY and _VX_TOOL to gnu.
27153 2017-06-26  Carl Love  <cel@us.ibm.com>
27155         * config/rs6000/rs6000-c.c: Add support for built-in functions
27156         vector bool char vec_reve (vector bool char);
27157         vector signed char vec_reve (vector signed char);
27158         vector unsigned char vec_reve (vector unsigned char);
27159         vector bool int vec_reve (vector bool int);
27160         vector signed int vec_reve (vector signed int);
27161         vector unsigned int vec_reve (vector unsigned int);
27162         vector bool long long vec_reve (vector bool long long);
27163         vector signed long long vec_reve (vector signed long long);
27164         vector unsigned long long vec_reve (vector unsigned long long);
27165         vector bool short vec_reve (vector bool short);
27166         vector signed short vec_reve (vector signed short);
27167         vector double vec_reve (vector double);
27168         vector float vec_reve (vector float);
27169         * config/rs6000/rs6000-builtin.def (VREVE_V2DI, VREVE_V4SI,
27170         VREVE_V8HI, VREVE_V16QI, VREVE_V2DF, VREVE_V4SF, VREVE): New builtin.
27171         * config/rs6000/altivec.md (UNSPEC_VREVEV): New UNSPEC.
27172         (altivec_vreve): New pattern.
27173         * config/rs6000/altivec.h (vec_reve): New define.
27174         * doc/extend.texi (vec_rev): Update the built-in documentation file
27175         for the new built-in functions.
27177 2016-06-26  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
27179         PR tree-optimization/71815
27180         * gimple-ssa-strength-reduction.c (uses_consumed_by_stmt): New
27181         function.
27182         (find_basis_for_candidate): Call uses_consumed_by_stmt rather than
27183         has_single_use.
27184         (slsr_process_phi): Likewise.
27185         (replace_uncond_cands_and_profitable_phis): Don't replace a
27186         multiply candidate with a stride of 1 (copy or cast).
27187         (phi_incr_cost): Call uses_consumed_by_stmt rather than
27188         has_single_use.
27189         (lowest_cost_path): Likewise.
27190         (total_savings): Likewise.
27192 2017-06-26  Richard Biener  <rguenther@suse.de>
27194         PR target/81175
27195         * config/i386/i386.c (ix86_init_mmx_sse_builtins):
27196         Use def_builtin_pure for all gather builtins.
27198 2017-06-26  Richard Biener  <rguenther@suse.de>
27200         PR tree-optimization/81203
27201         * tree-tailcall.c (find_tail_calls): Do not move stmts into
27202         non-dominating BBs.
27204 2017-06-26  Marek Polacek  <polacek@redhat.com>
27206         PR c/80116
27207         * doc/invoke.texi: Document -Wmultistatement-macros.
27209 2017-06-26  Christophe Lyon  <christophe.lyon@linaro.org>
27211         * doc/sourcebuild.texi (ARM-specific attributes): Document new
27212         arm_neon_ok_no_float_abi effective target.
27214 2017-06-26  Richard Biener  <rguenther@suse.de>
27216         PR tree-optimization/80928
27217         * cfghooks.c (duplicate_block): Do not copy BB_DUPLICATED flag.
27218         (copy_bbs): Set BB_DUPLICATED flag early.
27219         (execute_on_growing_pred): Do not execute for BB_DUPLICATED
27220         marked blocks.
27221         (execute_on_shrinking_pred): Likewise.
27222         * tree-ssa.c (ssa_redirect_edge): Do not look for PHI args in
27223         BB_DUPLICATED blocks.
27224         * tree-ssa-phionlycoprop.c (eliminate_degenerate_phis_1): Properly
27225         iterate over all PHIs considering removal of *gsi.
27227 2017-06-23  Jim Wilson  <jim.wilson@linaro.org>
27229         * doc/invoke.texi (AArch64 Options, -mtune): Re-add falkor and
27230         qdf24xx.
27232 2017-06-23  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
27234         * config/rs6000/rs6000-string.c: (expand_block_clear,
27235         do_load_for_compare, select_block_compare_mode,
27236         compute_current_alignment, expand_block_compare,
27237         expand_strncmp_align_check, expand_strn_compare,
27238         expand_block_move, rs6000_output_load_multiple)
27239         Move functions related to string/block move/compare
27240         to a separate file.
27241         * config/rs6000/rs6000.c: Move above functions to rs6000-string.c.
27242         * config/rs6000/rs6000-protos.h (rs6000_emit_dot_insn): Add prototype
27243         for this function which is now used in two files.
27244         * config/rs6000/t-rs6000: Add rule to compile rs6000-string.o.
27245         * config.gcc: Add rs6000-string.o to extra_objs for
27246         targets powerpc*-*-* and rs6000*-*-*.
27248 2017-06-23  Michael Meissner  <meissner@linux.vnet.ibm.com>
27250         PR target/80510
27251         * config/rs6000/rs6000.md (ALTIVEC_DFORM): Do not allow DImode in
27252         32-bit, since indexed is not valid for DImode.
27253         (mov<mode>_hardfloat32): Reorder ISA 2.07 load/stores before ISA
27254         3.0 d-form load/stores to be the same as mov<mode>_hardfloat64.
27255         (define_peephole2 for Altivec d-form load): Add 32-bit support.
27256         (define_peephole2 for Altivec d-form store): Likewise.
27258         PR ipa/81185
27259         * multiple_target.c (create_dispatcher_calls): Only create the
27260         dispatcher call if the function is the default clone of a
27261         versioned function.
27263 2017-06-23  Segher Boessenkool  <segher@kernel.crashing.org>
27265         PR middle-end/80902
27266         * builtins.c (expand_builtin_atomic_fetch_op): If emitting code after
27267         a call, force the call to not be a tail call.
27269 2017-06-23  Jeff Law  <law@redhat.com>
27271         * doc/contrib.texi: Add entry for Steven Pemberton's work on
27272         enquire.
27274 2017-06-23  Will Schmidt  <will_schmidt@vnet.ibm.com>
27276         * config/rs6000/rs6000.c: Add include of ssa-propagate.h for
27277         update_call_from_tree().  (rs6000_gimple_fold_builtin): Add
27278         handling for early expansion of vector shifts (sl,sr,sra,rl).
27279         (builtin_function_type): Add vector shift right instructions
27280         to the unsigned argument list.
27282 2017-06-23  Bernd Edlinger  <bernd.edlinger@hotmail.de>
27284         rtl-optimizatoin/79286
27285         * ira.c (update_equiv_regs): Revert to using may_trap_or_fault_p again.
27286         * rtlanal.c (rtx_addr_can_trap_p_1): SYMBOL_REF_FUNCTION_P can never
27287         trap.  PIC register plus a const unspec without offset can never trap.
27289 2017-06-23  Marc Glisse  <marc.glisse@inria.fr>
27291         * tree.h (builtin_structptr_type): New type.
27292         (builtin_structptr_types): Declare new array.
27293         * tree.c (builtin_structptr_types): New array.
27294         (free_lang_data, build_common_tree_nodes): Use it.
27296 2017-06-23  Jonathan Wakely  <jwakely@redhat.com>
27298         PR c++/81187
27299         * doc/invoke.texi (-Wnoexcept-type): Fix name of option, from
27300         -Wnoexcept.
27302 2017-06-22  Matt Turner  <mattst88@gmail.com>
27304         * config/i386/driver-i386.c (host_detect_local_cpu): Add Kaby
27305         Lake models to skylake case.  Assume skylake for unknown
27306         models with clflushopt.
27308 2017-06-22  Jeff Law  <law@redhat.com>
27310         * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Handle
27311         frame sizes that do not satisfy aarch64_uimm12_shift.
27313 2017-06-22  Jan Hubicka <hubicka@ucw.cz>
27315         * profile-count.h (apply_probability,
27316         apply_scale, probability_in): Fix checks for zero.
27318 2017-06-22  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
27320         * incpath.c (add_sysroot_to_chain): Allow for $SYSROOT prefix.
27321         * doc/cppdiropts.texi (-I @var{dir}): Document it.
27323 2016-06-22  Richard Biener  <rguenther@suse.de>
27325         * tree-vect-loop.c (vect_model_reduction_cost): Handle
27326         COND_REDUCTION and INTEGER_INDUC_COND_REDUCTION without
27327         REDUC_MAX_EXPR support.
27328         (vectorizable_reduction): Likewise.
27329         (vect_create_epilog_for_reduction): Likewise.
27331 2017-06-22  James Greenhalgh  <james.greenhalgh@arm.com>
27333         * match.pd (A / (1 << B) -> A >> B): New.
27334         * generic-match-head.c: Include optabs-tree.h.
27335         * gimple-match-head.c: Likewise.
27336         * optabs-tree.h (target_supports_op_p): New.
27337         * optabs-tree.c (target_supports_op_p): New.
27339 2017-06-22  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
27341         * configure.ac (gcc_cv_ld_static_dynamic): Also check stderr for
27342         $gcc_cv_ld --help output.
27343         (gcc_cv_ld_demangle): Likewise.
27344         (gcc_cv_ld_eh_frame_hdr): Likewise.
27345         (gcc_cv_ld_pie): Likewise.
27346         (gcc_cv_ld_as_needed): Likewise.  Prefer native forms unless $gnu_ld.
27347         (gcc_cv_ld_buildid): Likewise.
27348         (gcc_cv_ld_sysroot): Likewise.
27349         (ld_bndplt_support): Likewise.
27350         (ld_pushpopstate_support): Likewise.
27351         * configure: Regenerate.
27352         * config/sol2.h [!USE_GLD] (SYSROOT_SPEC): Define.
27354 2017-06-21  Jakub Jelinek  <jakub@redhat.com>
27356         PR target/81151
27357         * config/i386/sse.md (round<mode>2): Renumber match_dup and
27358         operands indexes to avoid gap between operands and match_dups.
27360 2017-06-21  Andrew Pinski  <apinski@cavium.com>
27362         * config/aarch64/aarch64-cost-tables.h (thunderx_extra_costs):
27363         Increment Arith_shift and Arith_shift_reg by 1.
27364         * config/aarch64/aarch64-tuning-flags.def (cheap_shift_extend):
27365         New tuning flag.
27366         * config/aarch64/aarch64.c (thunderx_tunings): Enable
27367         AARCH64_EXTRA_TUNE_CHEAP_SHIFT_EXTEND.
27368         (aarch64_strip_extend): Add new argument and test for it.
27369         (aarch64_cheap_mult_shift_p): New function.
27370         (aarch64_rtx_mult_cost): Call aarch64_cheap_mult_shift_p and don't
27371         add a cost if it is true.
27372         Update calls to aarch64_strip_extend.
27373         (aarch64_rtx_costs): Update calls to aarch64_strip_extend.
27375 2017-06-21  Andrew Pinski  <apinski@cavium.com>
27377         * config/aarch64/aarch64-cores.def (thunderxt88p1): Use thunderxt88
27378         tunings.
27379         (thunderxt88): Likewise.
27380         * config/aarch64/aarch64.c (thunderxt88_prefetch_tune): New variable.
27381         (thunderx_prefetch_tune): New variable.
27382         (thunderx2t99_prefetch_tune): Update for the correct values.
27383         (thunderxt88_tunings): New variable.
27384         (thunderx_tunings): Use thunderx_prefetch_tune instead of
27385         generic_prefetch_tune.
27386         (thunderx2t99_tunings): Use AUTOPREFETCHER_WEAK.
27388 2017-06-21  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
27390         * config/aarch64/atomics.md (aarch64_compare_and_swap<mode>_lse,
27391         SHORT): Relax operand 3 to aarch64_reg_or_zero and constraint to Z.
27392         (aarch64_compare_and_swap<mode>_lse, GPI): Likewise.
27393         (aarch64_atomic_cas<mode>, SHORT): Likewise for operand 2.
27394         (aarch64_atomic_cas<mode>, GPI): Likewise.
27396 2017-06-21  Martin Liska  <mliska@suse.cz>
27398         * gimplify.c (gimplify_label_expr): Insert GIMPLE_PREDICT
27399         statements on cold and hot labels.
27400         * predict.c (tree_estimate_probability_bb): Remove the
27401         prediction from this place.
27403 2017-06-21  Martin Liska  <mliska@suse.cz>
27405         PR tree-optimization/79489
27406         * gimplify.c (maybe_add_early_return_predict_stmt): New
27407         function.
27408         (gimplify_return_expr): Call the function.
27409         * predict.c (tree_estimate_probability_bb): Remove handling
27410         of early return.
27411         * predict.def: Update comment about early return predictor.
27412         * gimple-predict.h (is_gimple_predict): New function.
27413         * predict.def: Change default value of early return to 66.
27414         * tree-tailcall.c (find_tail_calls): Skip GIMPLE_PREDICT
27415         statements.
27416         * passes.def: Put pass_strip_predict_hints to the beginning of
27417         IPA passes.
27419 2017-06-21  Pierre-Marie de Rodat  <derodat@adacore.com>
27421         * dwarf2out.c (gen_decl_die): Remove the guard to skip file-scope
27422         FUNCTION_DECL declarations.
27423         (dwarf2out_early_global_decl): Remove the guard to skip FUNCTION_DECL
27424         declarations.
27425         (dwaf2out_decl): Likewise.
27426         * godump.c (go_early_global_decl): Skip call to the real debug hook
27427         for FUNCTION_DECL declarations.
27428         * passes.c (rest_of_decl_compilation): Skip call to the
27429         early_global_decl debug hook for FUNCTION_DECL declarations, unless
27430         -fdump-go-spec is passed.
27432 2017-06-21  Marc Glisse  <marc.glisse@inria.fr>
27434         * config/i386/i386.c (struct builtin_isa): New field pure_p.
27435         Reorder for compactness.
27436         (def_builtin, def_builtin2, ix86_add_new_builtins): Handle pure_p.
27437         (def_builtin_pure, def_builtin_pure2): New functions.
27438         (ix86_init_mmx_sse_builtins) [__builtin_ia32_stmxcsr]: Mark as pure.
27440 2017-06-21  Marc Glisse  <marc.glisse@inria.fr>
27442         * match.pd (nop_convert): New predicate.
27443         ((A +- CST1) +- CST2): Allow some NOP conversions.
27445 2017-06-21  Jakub Jelinek  <jakub@redhat.com>
27447         PR c++/81130
27448         * gimplify.c (omp_add_variable): Don't force GOVD_SEEN for types
27449         with ctors/dtors if GOVD_SHARED is set.
27451 2017-06-21  Wilco Dijkstra  <wdijkstr@arm.com>
27453         * config/aarch64/aarch64.md (movti_aarch64):
27454         Emit mov rather than orr.
27455         (movtf_aarch64): Likewise.
27456         * config/aarch64/aarch64-simd.md (aarch64_simd_mov):
27457         Emit mov rather than orr.
27459 2017-06-21  Wilco Dijkstra  <wdijkstr@arm.com>
27461         * config/aarch64/aarch64-simd.md (aarch64_simd_dup):
27462         Swap alternatives, make integer dup more expensive.
27464 2017-06-21  Wilco Dijkstra  <wdijkstr@arm.com>
27466         * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
27467         Return true for non-tls symbols.
27469 2017-06-21  James Greenhalgh  <james.greenhalgh@arm.com>
27471         * config/aarch64/aarch64-cores.def (cortex-a55): New.
27472         (cortex-a75): Likewise.
27473         (cortex-a75.cortex-a55): Likewise.
27474         * config/aarch64/aarch64-tune.md: Regenerate.
27475         * doc/invoke.texi (-mtune): Document new values for -mtune.
27477 2017-06-21  Tom de Vries  <tom@codesourcery.com>
27479         * doc/sourcebuild.texi (Add Options, Features for dg-add-options): Add
27480         stack_size feature.
27481         (Effective-Target Keywords, Other attributes): Suggest using
27482         dg-add-options stack_size feature to get stack limit in stack_size
27483         effective target documentation.
27485 2017-06-21  Julian Brown  <julian@codesourcery.com>
27486             Naveen H.S  <Naveen.Hurugalawadi@cavium.com>
27488         * config/aarch64/aarch64-simd.md (aarch64_crypto_pmulldi)
27489         (aarch64_crypto_pmullv2di): Change type attribute to crypto_pmull.
27490         * config/aarch64/thunderx2t99.md (thunderx2t99_pmull): New
27491         reservation.
27492         * config/arm/cortex-a53.md (cortex_a53_advsimd_type): Add crypto_pmull to
27493         attribute type list for neon_multiply.
27494         * config/arm/cortex-a57.md (cortex_a57_neon_type): Add crypto_pmull to
27495         attribute type list for neon_multiply.
27496         * config/arm/crypto.md (crypto_vmullp64): Change type to crypto_pmull.
27497         * config/arm/exynos-m1.md (exynos_m1_neon_type): Add crypto_pmull to
27498         attribute type list for neon_multiply.
27499         * config/arm/types.md (crypto_pmull): Add.
27500         * config/arm/xgene1.md (xgene1_neon_pmull): Add crypto_pmull to
27501         attribute type list.
27503 2017-06-20  Andreas Tobler  <andreast@gcc.gnu.org>
27505         * config.gcc (armv6*-*-freebsd*): Change the target_cpu_cname to
27506         arm1176jzf-s.
27508 2017-06-20  Jakub Jelinek  <jakub@redhat.com>
27510         * ira-costs.c (find_costs_and_classes): Initialize cost_classes later
27511         to make sure not to dereference a NULL cost_classes_ptr pointer.
27513 2017-06-20  Carl Love  <cel@us.ibm.com>
27515         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
27516         ALTIVEC_BUILTIN_VMULESW, ALTIVEC_BUILTIN_VMULEUW,
27517         ALTIVEC_BUILTIN_VMULOSW, ALTIVEC_BUILTIN_VMULOUW entries.
27518         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin,
27519         builtin_function_type): Add ALTIVEC_BUILTIN_* case statements.
27520         * config/rs6000/altivec.md (MVULEUW, VMULESW, VMULOUW,
27521         VMULOSW): New enum "unspec" values.
27522         (vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
27523         vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si,
27524         altivec_vmuleuw, altivec_vmulesw, altivec_vmulouw,
27525         altivec_vmulosw): New patterns.
27526         * config/rs6000/rs6000-builtin.def (VMLEUW, VMULESW, VMULOUW,
27527         VMULOSW): Add definitions.
27529 2017-06-20  Julia Koval  <julia.koval@intel.com>
27531         * config/i386/i386.c: Fix rounding expand for new pattern.
27532         * config/i386/subst.md: Fix pattern (parallel -> unspec).
27534 2017-06-20  James Greenhalgh  <james.greenhalgh@arm.com>
27536         * config/aarch64/aarch64-option-extensions.def (rcpc): New.
27537         * config/aarch64/aarch64.h (AARCH64_FL_RCPC): New.
27539 2017-06-20  James Greenhalgh  <james.greenhalgh@arm.com>
27541         * config/aarch64/aarch64-option-extensions.def (fp16): Fix expected
27542         feature string.
27544 2017-06-20  James Greenhalgh  <james.greenhalgh@arm.com>
27546         * config/aarch64/aarch64-cores.def: Rearrange to sort by
27547         architecture, then by implementer ID.
27548         * config/aarch64/aarch64-tune.md: Regenerate.
27550 2017-06-20  Richard Biener  <rguenther@suse.de>
27552         PR middle-end/81097
27553         * fold-const.c (split_tree): Fold to type before negating.
27555 2017-06-20  David Malcolm  <dmalcolm@redhat.com>
27557         * diagnostic-show-locus.c
27558         (selftest::test_fixit_deletion_affecting_newline): New function.
27559         (selftest::diagnostic_show_locus_c_tests): Call it.
27561 2017-06-20  Andreas Schwab  <schwab@suse.de>
27563         PR target/80970
27564         * config/m68k/m68k.md (bsetdreg, bchgdreg, bclrdreg): Use "=d"
27565         instead of "+d".
27567 2017-06-20  Prakhar Bahuguna  <prakhar.bahuguna@arm.com>
27569         * config/arm/arm-c.c (arm_cpu_builtins): New block to define
27570         __ARM_FEATURE_COPROC according to support.
27572 2017-06-20  Jakub Jelinek  <jakub@redhat.com>
27574         * tree-chkp.c (chkp_get_hard_register_var_fake_base_address):
27575         Rewritten to avoid overflow for > 32-bit pointers.
27577         PR sanitizer/81125
27578         * ubsan.h (ubsan_encode_value): Workaround buggy clang++ parser
27579         by removing enum keyword.
27580         (ubsan_type_descriptor): Likewise.  Formatting fix.
27582         PR target/81121
27583         * config/i386/i386.md (TARGET_USE_VECTOR_CONVERTS float si->{sf,df}
27584         splitter): Require TARGET_SSE2 in the condition.
27586 2017-06-20  Michael Meissner  <meissner@linux.vnet.ibm.com>
27588         PR target/79799
27589         * config/rs6000/rs6000.c (rs6000_expand_vector_init): Add support
27590         for doing vector set of SFmode on ISA 3.0.
27591         * config/rs6000/vsx.md (vsx_set_v4sf_p9): Likewise.
27592         (vsx_set_v4sf_p9_zero): Special case setting 0.0f to a V4SF
27593         element.
27594         (vsx_insert_extract_v4sf_p9): Add an optimization for inserting a
27595         SFmode value into a V4SF variable that was extracted from another
27596         V4SF variable without converting the element to double precision
27597         and back to single precision vector format.
27598         (vsx_insert_extract_v4sf_p9_2): Likewise.
27600 2017-06-19  Jakub Jelinek  <jakub@redhat.com>
27602         * tree-ssa-structalias.c (get_constraint_for_ptr_offset): Multiply
27603         in UWHI to avoid undefined overflow.
27605         PR sanitizer/81125
27606         * ubsan.h (enum ubsan_encode_value_phase): New.
27607         (ubsan_encode_value): Change second argument to
27608         enum ubsan_encode_value_phase with default value of
27609         UBSAN_ENCODE_VALUE_GENERIC.
27610         * ubsan.c (ubsan_encode_value): Change second argument to
27611         enum ubsan_encode_value_phase PHASE from bool IN_EXPAND_P,
27612         adjust uses, for UBSAN_ENCODE_VALUE_GENERIC use just
27613         create_tmp_var_raw instead of create_tmp_var and use a
27614         TARGET_EXPR.
27615         (ubsan_expand_bounds_ifn, ubsan_build_overflow_builtin,
27616         instrument_bool_enum_load, ubsan_instrument_float_cast): Adjust
27617         ubsan_encode_value callers.
27619         PR sanitizer/81111
27620         * ubsan.c (ubsan_encode_value): If current_function_decl is NULL,
27621         use create_tmp_var_raw instead of create_tmp_var, mark it addressable
27622         just by setting TREE_ADDRESSABLE on the result and use a TARGET_EXPR.
27624 2017-06-19  Richard Biener  <rguenther@suse.de>
27626         PR middle-end/81118
27627         * tree-cfgcleanup.c (cleanup_tree_cfg_noloop): Clear niter
27628         estimates if we changed anything.
27630 2017-06-19  Richard Biener  <rguenther@suse.de>
27632         PR tree-optimization/80887
27633         * tree-ssa-sccvn.c (mprts_hook_cnt): New global.
27634         (vn_lookup_simplify_result): Allow only mprts_hook_cnt succesful
27635         simplified lookups, then reset mprts_hook.
27636         (vn_nary_build_or_lookup_1): Set mprts_hook_cnt to 9 before
27637         simplifying.
27638         (try_to_simplify): Likewise.
27640 2017-06-19  Martin Liska  <mliska@suse.cz>
27642         PR sanitizer/80879
27643         * gimplify.c (gimplify_switch_expr):
27644         Initialize live_switch_vars for SWITCH_BODY == STATEMENT_LIST.
27646 2017-06-19  Martin Liska  <mliska@suse.cz>
27648         * doc/install.texi: Document that PGO runs in 4 stages.
27650 2017-06-19  Martin Liska  <mliska@suse.cz>
27652         PR ipa/80732
27653         * attribs.c (make_dispatcher_decl): Do not append '.ifunc'
27654         to dispatcher function name.
27655         * multiple_target.c (replace_function_decl): New function.
27656         (create_dispatcher_calls): Redirect both edges and references.
27658 2017-06-19  Jan Hubicka <hubicka@ucw.cz>
27660         * profile-count.c (profile_count::dump): Dump quality.
27661         (profile_count::differs_from_p): Update for unsigned val.
27662         * profile-count.h (profile_count_quality): New enum.
27663         (profile_count): Turn m_val to 62bit unsigned, add quality tracking.
27665 2017-06-19  Richard Biener  <rguenther@suse.de>
27667         * tree-ssa-loop-niter.h (estimate_numbers_of_iterations): Take
27668         struct function as arg.
27669         (estimate_numbers_of_iterations): Export overload with loop arg.
27670         (free_numbers_of_iterations_estimates_loop): Use an overload of
27671         free_numbers_of_iterations_estimates instead.
27672         * tree-cfg.c (remove_bb): Adjust.
27673         * tree-cfgcleanup.c (remove_forwarder_block_with_phi): Likewise.
27674         * tree-parloops.c (gen_parallel_loop): Likewise.
27675         * tree-ssa-loop-ivcanon.c (canonicalize_induction_variables):
27676         Likewise.
27677         (tree_unroll_loops_completely): Likewise.
27678         * tree-ssa-loop-niter.c (estimate_numbers_of_iterations_loop):
27679         Use an overload instead and export.
27680         (estimated_loop_iterations): Adjust.
27681         (max_loop_iterations): Likewise.
27682         (likely_max_loop_iterations): Likewise.
27683         (estimate_numbers_of_iterations): Take struct function as arg
27684         and adjust.
27685         (loop_exits_before_overflow): Adjust.
27686         (free_numbers_of_iterations_estimates_loop): Use an overload.
27687         * tree-vect-loop.c (vect_analyze_loop_form): Adjust.
27688         * tree-vectorizer.c (vect_free_loop_info_assumptions): Likewise.
27690 2017-06-19  Richard Biener  <rguenther@suse.de>
27692         PR ipa/81112
27693         * ipa-prop.c (find_constructor_constant_at_offset): Handle
27694         RANGE_EXPR conservatively.
27696 2017-06-16  Carl Love  <cel@us.ibm.com>
27698         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
27699         definitions for vec_float, vec_float2, vec_floato,
27700         vec_floate built-ins.
27701         * config/rs6000/vsx.md (define_c_enum "unspec"): Add RTL code
27702         for instructions vsx_xvcvsxws vsx_xvcvuxwsp, float2, floato and
27703         floate.
27704         * config/rs6000/rs6000-builtin.def (FLOAT2_V2DI, FLOATE_V2DF,
27705         FLOATE_2DI, FLOATO_V2DF, FLOATEE_V2DI, XVCVSXWSP_V4SF,
27706         UNS_FLOATO_V2DI, UNS_FLOATE_V2DI): Add definitions.
27707         * config/altivec.md (define_insn "p8_vmrgew_<mode>",
27708         define_mode_attr VF_sxddp): Add V4SF type to p8_vmrgew.
27709         * config/rs6000/altivec.h (vec_float, vec_float2, vec_floate,
27710         vec_floato): Add builtin defines.
27711         * doc/extend.texi (vec_float, vec_float2, vec_floate, vec_floato):
27712         Update the built-in documentation file for the new built-in
27713         functions.
27715 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27717         * config/arm/arm.opt (marm): Mark as the negative of of -mthumb.
27718         (mthumb): Mark as the negative of -marm.
27720 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27722         * doc/invoke.texi (ARM Options, -mcpu): Document supported
27723         extension options.
27724         (ARM Options, -mtune): Document that this accepts the same
27725         extension options as -mcpu.
27726         (ARM Options, -mfpu): Document addition of -mfpu=auto.
27728 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27730         * doc/invoke.texi (ARM Options, -march=): Document new syntax and
27731         permitted extensions.
27733 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27735         * config/arm/arm-cpus.in (armv7): Add extension +nofp.
27736         (armv7-r): Add aliases vfpv3xd and vfpv3-d16.
27737         (armv8-m.main): Add option +nodsp.
27738         * config/arm/arm-cpu-cdata.h: Regenerated.
27740 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27742         * config/arm/t-fuchsia: New file.
27743         * config.gcc (arm*-*-fuchsia*): Use it.
27745 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27747         * config/arm/t-symbian: Rewrite for new option infrastructure.
27749 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27751         * config/arm/t-phoenix (MULTILIB_REUSE): Clear variable.
27752         (MULTILIB_REQUIRED): Likewise.
27754 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27756         * config/arm/t-linux-eabi (MULTILIB_EXCEPTIONS): Set to empty.
27757         (MULTILIB_RESUE): Likewise.
27758         (MULTILIB_MATCHES): Likewise.
27759         (MULTLIB_REQUIRED): Likewise.
27761 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27763         * config/arm/t-rtems: Rewrite for new option framework.
27765 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27767         * config/arm/t-aprofile (v7_a_nosimd_variants, v7_a_simd_variants)
27768         (v7ve_nosimd_variatns, v7ve_vfpv3_simd_variants)
27769         (v7ve_vfpv4_simd_variants, v8_a_nosimd_variants, v8_a_simd_variants)
27770         (v8_1_a_simd_variants, v8_2_a_simd_variants): Move to ...
27771         * config/arm/t-multilib: ... here.
27772         (MULTILIB_OPTIONS): Add armv7 and armv7+fp architectures.
27773         (MULTILIB_MATCHES): Use armv7 libraries for armv7-r.  Also use for
27774         armv7-a and armv8*-a when A-profile libraries have not been built.
27775         * config/arm/t-rmprofile: Rewrite.
27777 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27779         * genmultilib (multilib_reuse): Allow an explicit period to be escaped
27780         with a backslash.  Remove the backslash after substituting unescaped
27781         periods.
27782         * doc/fragments.texi (MULTILIB_REUSE): Document it.
27784 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27786         * config.gcc: (arm*-*-*): When building a-profile libraries, force
27787         the driver to pass through the default setting of -mfloat-abi.
27788         * common/config/arm/arm-common.c (arm_target_thumb_only): Return -marm
27789         rather than NULL.
27790         * config/arm/t-multilib (MULTILIB_REUSE): Initialize to empty.
27791         (all_feat_combs): New rule.
27792         (MULTILIB_OPTIONS): Use explicit ARM and Thumb directories.  Rework
27793         default libraries.
27794         * config/arm/t-aprofile: Rewrite.
27796 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27798         * config/arm/arm.h (FPUTYPE_AUTO): Define.
27799         * config/arm/arm.c (arm_option_override): Use FPUTYPE_AUTO if the
27800         fpu is not specified by the user/command-line.
27801         * config/arm/bpabi.h (FPUTYPE_DEFAULT): Delete.
27802         * config/arm/netbsd-elf.h (FPUTYPE_DEFAULT): Delete.
27803         * config/arm/linux-elf.h (FPUTYPE_DEFAULT): Delete.
27804         * config/arm/vxworks.h (FPUTYPE_DEFAULT): Delete.
27805         * common/config/arm/arm-common.c (arm_canon_arch_option): Use
27806         FPUTYPE_AUTO insted of FPUTYPE_DEFAULT.
27808 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27810         * config/arm/elf.h (MULTILIB_DEFAULTS): Delete.
27811         * config/arm/t-arm-elf: Rewritten.
27813 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27815         * config/arm/arm.h (TARGET_HARD_FLOAT): Also check that we
27816         have some floating-point instructions.
27817         (TARGET_SOFT_FLOAT): Define as inverse of TARGET_HARD_FLOAT.
27818         (TARGET_MAYBE_HARD_FLOAT): New macro.
27819         * config/arm/arm-builtins.c (arm_init_builtins): Use
27820         TARGET_MAYBE_HARD_FLOAT.
27821         * config/arm/arm.c (arm_option_override): Use TARGET_HARD_FLOAT_ABI.
27823 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27825         * common/config/arm/arm-common.c: Define INCLUDE_LIST.
27826         (configargs.h): Include it.
27827         (arm_print_hint_for_fpu_option): New function.
27828         (arm_parse_fpu_option): New function.
27829         (candidate_extension): New class.
27830         (arm_canon_for_multilib): New function.
27831         * config/arm/arm.h (CANON_ARCH_SPEC_FUNCTION): New macro.
27832         (EXTRA_SPEC_FUNCTIONS): Add CANON_ARCH_SPEC_FUNCTION.
27833         (ARCH_CANONICAL_SPECS): New macro.
27834         (DRIVER_SELF_SPECS): Add ARCH_CANONICAL_SPECS.
27836 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27838         * config.gcc (arm*-*-*): Ensure both target_cpu_cname and with_cpu
27839         are set after handling multilib fragments.  Set target_cpu_default2
27840         from with_cpu.
27842 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27844         * config.gcc (arm*-*-fucshia*): Set target_cpu_cname to the real
27845         cpu name.
27846         (arm*-*-*): Set target_cpu_default2 to a quoted string.
27847         * config/arm/parsecpu.awk (check_cpu): Validate any extension
27848         options.
27849         (check_arch): Likewise.
27850         * config/arm/arm.c (arm_configure_build_target): Handle
27851         TARGET_CPU_DEFAULT being a string constant.  Scan any feature
27852         options in the default.
27854 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27856         * config/arm/arm-protos.h (cpu_arch_extension): Add field to record
27857         when an option is an alias of another.
27858         * config/arm/parsecpu.awk (optalias): New parser token.
27859         (gen_comm_data): Mark non-alias options as such.  Emit entries
27860         for extension aliases.
27861         * config/arm/arm-cpus.in (armv5e): Make vfpv2 an alias.
27862         (armv5te, armv5tej, armv6, armv6j, armv6k, armv6z): Likewise.
27863         (armv6kz, armv6zk, armv6t2): Likewise.
27864         (armv7): Make vfpv3-d16 an alias.
27865         (armv7-a): Make vfpv3-d16, neon and neon-vfpv3 aliases.  Sort in
27866         canonical order.
27867         (armv7ve): Make vfpv4-d16, neon-vfpv3 and neon-vfpv4 aliases.
27868         Sort in canonical order.
27869         (armv8-a): Sort in canonical order.
27870         (armv8.1-a, armv8.2-a):  Likewise.
27871         (generic-armv7-a): Make neon and neon-vfpv3 aliases.  Sort in
27872         canonical order.
27873         (cortex-a9): Sort in canonical order.
27874         * config/arm/arm.c (selftests.h): Include it.
27875         (arm_test_cpu_arch_data): New function.
27876         (arm_run_self_tests): New function.
27877         (TARGET_RUN_TARGET_SELFTESTS): Redefine.
27878         (targetm): Move declaration to the end of the file.
27879         * arm-cpu-cdata.h: Regenerated.
27881 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27883         * config/arm/arm.h (TARGET_MODE_SPECS): Add additional parameter to
27884         call to target_mode_check describing the type of option passed.
27885         * common/config/arm/arm-common.c (arm_arch_core_flag): Delete.
27886         (arm_target_thumb_only): Use arm_parse_arch_option_name or
27887         arm_parse_cpu_option_name to match parameters against list of
27888         available targets.
27889         * config/arm/parsecpu.awk (gen_comm_data): Don't generate
27890         arm_arch_core_flags data structure.
27891         * config/arm/arm-cpu_cdata.h: Regenerated.
27893 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27895         * common/config/arm/arm-common.c (arm_initialize_isa): Moved here from
27896         config/arm/arm.c.
27897         (arm_print_hint_for_cpu_option): Likewise.
27898         (arm_print_hint_for_arch_option): Likewise.
27899         (arm_parse_cpu_option_name): Likewise.
27900         (arm_parse_arch_option_name): Likewise.
27901         * config/arm/arm.c (arm_identify_fpu_from_isa): Use the computed number
27902         of entries in the all_fpus list.
27903         * config/arm/arm-protos.h (all_architectures, all_cores): Declare.
27904         (arm_parse_cpu_option_name): Declare.
27905         (arm_parse_arch_option_name): Declare.
27906         (arm_parse_option_features): Declare.
27907         (arm_intialize_isa): Declare.
27908         * config/arm/parsecpu.awk (gen_data): Move CPU and architecture
27909         data tables to ...
27910         (gen_comm_data): ... here.  Make definitions non-static.
27911         * config/arm/arm-cpu-data.h: Regenerated.
27912         * config/arm/arm-cpu-cdata.h: Regenerated.
27914 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27916         * config/arm/arm-protos.h (arm_build_target): Remove arch_core.
27917         (cpu_arch_extension): New structure.
27918         (cpu_arch_option, arch_option, cpu_option): New structures.
27919         * config/arm/parsecpu.awk (gen_headers): Build an enumeration of
27920         architecture types.
27921         (gen_data): Generate new format data tables.
27922         * config/arm/arm.c (cpu_tune): New structure.
27923         (cpu_option, processors): Delete.
27924         (arm_print_hint_for_core_or_arch): Delete.  Replace with ...
27925         (arm_print_hint_for_cpu_option): ... this and ...
27926         (arm_print_hint_for_arch_option): ... this.
27927         (arm_parse_arch_cpu_name): Delete.  Replace with ...
27928         (arm_parse_cpu_option_name): ... this and ...
27929         (arm_parse_arch_option_name): ... this.
27930         (arm_unrecognized_feature): Change type of target parameter to
27931         cpu_arch_option.
27932         (arm_parse_arch_cpu_features): Delete.  Replace with ...
27933         (arm_parse_option_features): ... this.
27934         (arm_configure_build_target): Rework to use new configuration data
27935         tables.
27936         (arm_print_tune_info): Rework for new configuration data tables.
27937         * config/arm/arm-cpu-data.h: Regenerated.
27938         * config/arm/arm-cpu.h: Regenerated.
27940 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27942         * Makefile.in (OBJS): Move sbitmap.o from here ...
27943         (OBJS-libcommon): ... to here.
27945 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27947         * config/arm/arm-isa.h (ISA_ALL_FPU_INTERNAL): Renamed from ISA_ALL_FPU.
27948         (ISA_ALL_CRYPTO): New macro.
27949         (ISA_ALL_SIMD): New macro
27950         (ISA_ALL_FP): New macro.
27951         * config/arm/arm.c (fpu_bitlist): Update initializer.
27952         * config/arm/arm-cpus.in: Use new ISA_ALL macros to disable crypto,
27953         simd or fp.
27954         (arm9e): Add fpu.  Add option for nofp
27955         (arm946e-s, arm966e-s, arm968e-s, arm10e, arm1020e, arm1022e): Likewise.
27956         (arm926ej-s, arm1026ej-s): Likewise.
27957         (generic-armv7-a): Add fpu.  Add options for simd, vfpv3, vfpv3-d16,
27958         vfpv3-fp16, vfpv3-d16-fp16, vfpv4, vfpv4-d16, neon, neon-vfp3,
27959         neon-fp16, neon-vfpv4, nofp and nosimd.
27960         (cortex-a5, cortex-a7): Add fpu.  Add options for nosimd and nofp.
27961         (cortex-a8): Add fpu.  Add option for nofp.
27962         (cortex-a9): Add fpu.  Add options for nosimd and nofp.
27963         (cortex-a12, cortex-a15, cortex-a17): Add fpu.  Add option for nofp.
27964         (cortex-r4f): Add fpu.
27965         (cortex-r5): Add fpu.  Add options for nofp.dp and nofp.
27966         (cortex-r7): Use idiv option from architecture.  Add fpu.  Add option
27967         for nofp.
27968         (cortex-r8): Likewise.
27969         (cortex-m4): Add fpu.  Add option for nofp.
27970         (cortex-a15.cortex-a7): Add fpu.  Add option for nofp.
27971         (cortex-a17.cortex-a7): Likewise.
27972         (cortex-a32): Add fpu.  Add options for crypto and nofp.
27973         (cortex-a35, cortex-a53): Likewise.
27974         (cortex-a57): Add fpu.  Add option for crypto.
27975         (cortex-a72, cortex-a73): Likewise.
27976         (exynos-m1): Likewise.
27977         (cortex-a57.cortex-a53, cortex-a72.cortex-a53): Likewise.
27978         (cortex-a73.cortex-a35, cortex-a73.cortex-a53): Likewise.
27979         (cortex-m33): Add fpu.  Add option for nofp.
27980         * config/arm/arm-cpu-cdata.h: Regenerated
27981         * config/arm/arm-cpu-data.h: Regenerated.
27983 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27985         * arm-cpus.in (armv5e): Add options fp, vfpv2 and nofp.
27986         (armv5te, armv5tej): Likewise.
27987         (armv6, armv6j, armv6k, armv6z, armv6kz, armv6zk, armv6t2): Likewise.
27988         (armv7): Add options fp and vfpv3-d16.
27989         (armv7-a): Add options fp, simd, vfpv3, vfpv3-d16, vfpv3-d16-fp16,
27990         vfpv3-fp16, vfpv4, vfpv4-d16, neon, neon-vfpv3, neon-fp16, neon-vfpv4,
27991         nofp and nosimd.
27992         (armv7ve): Likewise.
27993         (armv7-r): Add options fp, fp.sp, idiv, nofp and noidiv.
27994         (armv7e-m): Add options fp, fpv5, fp.dp and nofp.
27995         (armv8-a): Add nocrypto option.
27996         (armv8.1-a, armv8.2-a): Likewise.
27997         (armv8-m.main): add options fp, fp.dp and nofp.
27999 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
28001         * config/arm/arm-cpus.in (armv8-a): Add options crc, simd crypto and
28002         nofp.
28003         (armv8-a+crc): Delete.
28004         (armv8.1-a): Add options simd, crypto and nofp.
28005         (armv8.2-a): Add options fp16, simd, crypto and nofp.
28006         (armv8.2-a+fp16): Delete.
28007         (armv8-m.main): Add option dsp.
28008         (armv8-m.main+dsp): Delete.
28009         (cortex-a8): Add fpu.  Add nofp option.
28010         (cortex-a9): Add fpu.  Add nofp and nosimd options.
28011         * config/arm/parsecpu.awk (gen_data): Generate option tables and
28012         link to main cpu and architecture data structures.
28013         (gen_comm_data): Only put isa attributes from the main architecture
28014         in common tables.
28015         (option): New statement for architecture and CPU entries.
28016         * arm.c (struct cpu_option): New structure.
28017         (struct processors): Add entry for options.
28018         (arm_unrecognized_feature): New function.
28019         (arm_parse_arch_cpu_name): Ignore any characters after the first
28020         '+' character.
28021         (arm_parse_arch_cpu_feature): New function.
28022         (arm_configure_build_target): Separate out any CPU and architecture
28023         features and parse separately.  Don't error out if -mfpu=auto is
28024         used with only an architecture string.
28025         (arm_print_asm_arch_directives): New function.
28026         (arm_file_start): Call it.
28027         * config/arm/arm-cpu-cdata.h: Regenerated.
28028         * config/arm/arm-cpu-data.h: Likewise.
28029         * config/arm/arm-tables.opt: Likewise.
28031 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
28033         * config/arm/elf.h (ASM_SPEC): Only pass -mfpu through to the
28034         assembler when it is not -mfpu=auto.
28036 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
28038         * config/arm/arm.h (BIG_LITTLE_SPEC): Delete macro.
28039         (ASM_REWRITE_SPEC_FUNCTIONS): New macro.
28040         (BIG_LITTLE_CPU_SPEC_FUNCTIONS): Delete macro.
28041         (ASM_CPU_SPEC): Rewrite.
28042         (MCPU_MTUNE_NATIVE_FUNCTIONS): New macro.
28043         (EXTRA_SPEC_FUNCTIONS): Move outside of ifdef.  Use
28044         MCPU_MTUNE_NATIVE_FUNCTIONS and ASM_REWRITE_SPEC_FUNCTIONS.  Remove
28045         reference to BIG_LITTLE_CPU_SPEC_FUNCTIONS.
28046         * common/config/arm/arm-common.c (arm_rewrite_selected_cpu): Ensure
28047         copied string is NUL-terminated.  Also strip any characters prefixed
28048         by '+'.
28049         (arm_rewrite_selected_arch): New function.
28050         (arm_rewrite_march): New function.
28052 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
28054         * config/arm/arm.opt (x_arm_arch_string): New TargetSave option.
28055         (x_arm_cpu_string, x_arm_tune_string): Likewise.
28056         (march, mcpu, mtune): Convert to string-based options.
28057         * config/arm/arm.c (arm_print_hint_for_core_or_arch): New function.
28058         (arm_parse_arch_cpu_name): New function.
28059         (arm_configure_build_target): Use arm_parse_arch_cpu_name to
28060         identify selected architecture or CPU.
28061         (arm_option_save): New function.
28062         (TARGET_OPTION_SAVE): Redefine.
28063         (arm_option_restore): Restore string options.
28064         (arm_option_print): Print string options.
28066 2017-06-16  Martin Sebor  <msebor@redhat.com>
28068         PR tree-optimization/80933
28069         PR tree-optimization/80934
28070         * builtins.c (fold_builtin_3): Do not handle bcmp here.
28071         * gimple-fold.c (gimple_fold_builtin_bcmp): New function.
28072         (gimple_fold_builtin_bcopy, gimple_fold_builtin_bzero): Likewise.
28073         (gimple_fold_builtin): Call them.
28075 2017-06-16  Jan Hubicka  <hubicka@ucw.cz>
28077         * gimple-ssa-isolate-paths.c (isolate_path): Set edge leading to path
28078         as unlikely; update profile.
28080 2017-06-16  Jan Hubicka  <hubicka@ucw.cz>
28082         * predict.c (force_edge_cold): Handle declaring edges impossible
28083         more aggresively.
28085 2017-06-16  Jan Hubicka  <hubicka@ucw.cz>
28087         * tree-ssa-loop-ivcanon.c (remove_exits_and_undefined_stmts): Update
28088         profile.
28089         (try_unroll_loop_completely): Fix reporting.
28091 2017-06-16  Jan Hubicka  <hubicka@ucw.cz>
28093         * tree-ssa-tail-merge.c (replace_block_by): Fix profile updating.
28095 2017-06-16  James Greenhalgh  <james.greenhalgh@arm.com>
28097         PR target/71778
28098         * config/arm/arm-builtins.c (arm_expand_builtin_args): Return TARGET
28099         if given a non-constant argument for an intrinsic which requires a
28100         constant.
28102 2017-06-16  Jan Hubicka  <hubicka@ucw.cz>
28104         * profile.c (compare_freqs): New function.
28105         (branch_prob): Sort edge list.
28106         (find_spanning_tree): Assume that the list is priority sorted.
28108 2017-06-16  Richard Biener  <rguenther@suse.de>
28110         PR tree-optimization/81090
28111         * passes.def (pass_record_bounds): Remove.
28112         * tree-pass.h (make_pass_record_bounds): Likewise.
28113         * tree-ssa-loop.c (pass_data_record_bounds, pass_record_bounds,
28114         make_pass_record_bounds): Likewise.
28115         * tree-ssa-loop-ivcanon.c (canonicalize_induction_variables): Do
28116         not free niter estimates at the beginning but at the end.
28117         * tree-scalar-evolution.c (scev_finalize): Free niter estimates.
28119 2017-06-16  Richard Biener  <rguenther@suse.de>
28121         * tree-switch-conversion.c (emit_case_bit_tests): Adjust
28122         initializer to workaround ICE in host GCC 4.8.
28124 2017-06-16  Jan Hubicka  <hubicka@ucw.cz>
28126         * ipa-inline-transform.c (update_noncloned_frequencies): Update also
28127         counts.
28128         (clone_inlined_nodes): Update.
28130 2017-06-16  Maxim Kuvyrkov  <maxim.kuvyrkov@linaro.org>
28132         * config/aarch64/aarch64.c (qdf24xx_prefetch_tune): Update
28133         prefetch settings, and enable prefetching by default at -O3.
28135 2017-06-16  Maxim Kuvyrkov  <maxim.kuvyrkov@linaro.org>
28137         * config/aarch64/aarch64.c (aarch64_override_options_internal):
28138         Set flag_prefetch_loop_arrays according to tuning data.
28140 2017-06-16  Maxim Kuvyrkov  <maxim.kuvyrkov@linaro.org>
28142         * config/aarch64/aarch64-protos.h (struct cpu_prefetch_tune):
28143         New tune structure.
28144         (struct tune_params): Use cpu_prefetch_tune instead of cache_line_size.
28145         [Unrelated to main purpose of the patch] Place the pointer field last
28146         to enable type checking errors when tune structure are wrongly merged.
28147         * config/aarch64/aarch64.c (generic_prefetch_tune,)
28148         (exynosm1_prefetch_tune, qdf24xx_prefetch_tune,)
28149         (thunderx2t99_prefetch_tune): New tune constants.
28150         (tune_params *_tunings): Update all tunings (no functional change).
28151         (aarch64_override_options_internal): Set PARAM_SIMULTANEOUS_PREFETCHES,
28152         PARAM_L1_CACHE_SIZE, PARAM_L1_CACHE_LINE_SIZE, and PARAM_L2_CACHE_SIZE
28153         from tunings structures.
28155 2017-06-16  Jakub Jelinek  <jakub@redhat.com>
28157         PR sanitizer/81094
28158         * ubsan.c (instrument_null): Add T argument, use it instead
28159         of computing it based on IS_LHS.
28160         (instrument_object_size): Likewise.
28161         (pass_ubsan::execute): Adjust instrument_null and
28162         instrument_object_size callers to pass gimple_get_lhs or
28163         gimple_assign_rhs1 result to it.  Use instrument_null instead of
28164         calling get_base_address and instrument_mem_ref.  Handle
28165         aggregate call arguments for object-size sanitization.
28167 2017-06-16  Yury Gribov  <tetra2005@gmail.com>
28169         PR tree-optimization/81089
28170         * tree-vrp.c (is_masked_range_test): Validate operands of
28171         subexpression.
28173 2017-06-15  Martin Sebor  <msebor@redhat.com>
28175         PR c++/80560
28176         * dumpfile.c (dump_register): Avoid calling memset to initialize
28177         a class with a default ctor.
28178         * gcc.c (struct compiler): Remove const qualification.
28179         * genattrtab.c (gen_insn_reserv): Replace memset with initialization.
28180         * hash-table.h: Ditto.
28181         * ipa-cp.c (allocate_and_init_ipcp_value): Replace memset with
28182           assignment.
28183         * ipa-prop.c (ipa_free_edge_args_substructures): Ditto.
28184         * omp-low.c (lower_omp_ordered_clauses): Replace memset with
28185         default ctor.
28186         * params.h (struct param_info): Make struct members non-const.
28187         * tree-switch-conversion.c (emit_case_bit_tests): Replace memset
28188         with default initialization.
28189         * vec.h (vec_copy_construct, vec_default_construct): New helper
28190         functions.
28191         (vec<T>::copy, vec<T>::splice, vec<T>::reserve): Replace memcpy
28192         with vec_copy_construct.
28193         (vect<T>::quick_grow_cleared): Replace memset with default ctor.
28194         (vect<T>::vec_safe_grow_cleared, vec_safe_grow_cleared): Same.
28195         * doc/invoke.texi (-Wclass-memaccess): Document.
28197 2017-06-15  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
28199         * emit-rtl.h (is_leaf): Update comment about local
28200         register allocator.
28202 2017-06-15  Jozef Lawrynowicz   <jozef.l@somniumtech.com>
28204         PR target/78818
28205         * config/msp430/msp430.c (msp430_data_attr): Check that it's possible
28206         for a variable to have a section before checking if the section has a
28207         name.
28208         Set section to.persistent if persistent attribute is set.
28209         Warn if .persistent attribute is used on an automatic variable.
28211 2017-06-15  Eric Botcazou  <ebotcazou@adacore.com>
28213         PR rtl-optimization/80474
28214         * reorg.c (update_block): Do not ignore instructions in a delay slot.
28216 2017-06-15  Segher Boessenkool  <segher@kernel.crashing.org>
28218         * config/rs6000/rs6000.md (add<mode>3): Use reg_or_subregno instead
28219         of REGNO.
28221 2017-06-14  Maciej W. Rozycki  <macro@imgtec.com>
28223         * config/mips/mips.md (MIPS16_T_REGNUM): Remove constant.
28224         (casesi): Emit bounds checking as RTL.
28225         (casesi_internal_mips16_<mode>): Remove bounds checking.
28227 2017-06-14  Max Filippov  <jcmvbkbc@gmail.com>
28229         * config/xtensa/xtensa.c (xtensa_option_override): Append
28230         MASK_CONST16 to target_flags in the absence of TARGET_L32R.
28231         (hwloop_optimize, hwloop_fail, hwloop_pattern_reg,
28232          xtensa_doloop_hooks): Define unconditionally.
28233         (xtensa_reorg_loops): Only call reorg_loops in the presence of
28234         TARGET_LOOPS.
28235         * config/xtensa/xtensa.h (TARGET_L32R): New definition.
28236         (TARGET_DEFAULT): Remove XCHAL_HAVE_L32R condition and account
28237         for it in xtensa_option_override.
28238         (HARD_FRAME_POINTER_IS_FRAME_POINTER,
28239          HARD_FRAME_POINTER_IS_ARG_POINTER): New definitions.
28241 2017-06-14  Boris Kolpackov  <boris@codesynthesis.com>
28243         * doc/cppopts.texi: Document '-' special value to -MF.
28245 2017-06-14  Wilco Dijkstra  <wdijkstr@arm.com>
28247         * config/arm/cortex-a53.md (cortex_a53_fpalu) Adjust latency.
28248         (cortex_a53_fconst): Likewise.
28249         (cortex_a53_fpmul): Likewise.
28250         (cortex_a53_f_load_64): Likewise.
28251         (cortex_a53_f_load_many): Likewise.
28252         (cortex_a53_advsimd_alu): Likewise.
28253         (cortex_a53_advsimd_alu_q): Likewise.
28254         (cortex_a53_advsimd_mul): Likewise.
28255         (cortex_a53_advsimd_mul_q): Likewise.
28256         (fpmac bypass): Add new bypass for fpmac-fpmac case.
28257         Add missing fmul, r2f_cvt and fconst cases.
28259 2017-06-14  Richard Biener  <rguenther@suse.de>
28261         PR middle-end/81088
28262         * fold-const.c (split_tree): Drop TREE_OVERFLOW flag from
28263         literal constants.
28264         (fold_binary_loc): When associating do not treat pre-existing
28265         TREE_OVERFLOW on literal constants as a reason to allow
28266         TREE_OVERFLOW on associated literal constants.
28268 2017-06-14  Eric Botcazou  <ebotcazou@adacore.com>
28270         * config/sparc/sparc.h (MASK_ISA): Add MASK_LEON and MASK_LEON3.
28271         (MASK_FEATURES): New macro.
28272         * config/sparc/sparc.c (sparc_option_override): Remove the special
28273         handling of -mfpu and generalize it to all MASK_FEATURES switches.
28275 2017-06-14  Eric Botcazou  <ebotcazou@adacore.com>
28277         * simplify-rtx.c (simplify_binary_operation_1) <UDIV>: Do not simplify
28278         a division of 0 if non-call exceptions are enabled.
28280 2017-06-14  Andrew Pinski  <apinski@cavium.com>
28281             Naveen H.S  <Naveen.Hurugalawadi@cavium.com>
28283         PR target/71663
28284         * config/aarch64/aarch64.c (aarch64_expand_vector_init):
28285         Improve vector initialization code gen for only variable case.
28287 2017-06-14  Eric Botcazou  <ebotcazou@adacore.com>
28289         * config/sparc/driver-sparc.c (cpu_names): Add SPARC-T5 entry.
28291 2017-06-14  Richard Biener  <rguenther@suse.de>
28293         PR tree-optimization/81083
28294         * tree-ssa-sccvn.c (vn_reference_lookup_3): Do not use abnormals
28295         as values.
28297 2017-06-13  Segher Boessenkool  <segher@kernel.crashing.org>
28299         * config/rs6000/rs6000.c: Update all comments that mentioned SPE.
28300         (rs6000_expand_builtin): Remove RS6000_BTC_EVSEL.
28301         * config/rs6000/rs6000.h (RS6000_BTC_EVSEL): Delete.
28302         * config/rs6000/vxworks.h (VXCPU_FOR_8548): Delete.  Adjust former use.
28303         * config/rs6000/vxworksae.h (VXCPU_FOR_8548): Delete.
28304         * config/rs6000/vxworksmils.h (VXCPU_FOR_8548): Delete.
28306 2017-06-13  Segher Boessenkool  <segher@kernel.crashing.org>
28308         * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete VECTOR_SPE.
28309         * config/rs6000/rs6000.c (rs6000_debug_vector_unit): Delete VECTOR_SPE.
28311 2017-06-13  Segher Boessenkool  <segher@kernel.crashing.org>
28313         * config/rs6000/rs6000.h (FIXED_SCRATCH): Delete.
28315 2017-06-13  Segher Boessenkool  <segher@kernel.crashing.org>
28317         * config/rs6000/t-rtems: Don't handle SPE.
28319 2017-06-13  Segher Boessenkool  <segher@kernel.crashing.org>
28321         * config/rs6000/t-linux: Don't handle SPE.
28323 2017-06-13  Segher Boessenkool  <segher@kernel.crashing.org>
28325         * config/rs6000/eabispe.h: Delete file.
28327 2017-06-13  Segher Boessenkool  <segher@kernel.crashing.org>
28329         * config/rs6000/t-spe: Delete file.
28331 2017-06-13  Segher Boessenkool  <segher@kernel.crashing.org>
28333         * config/rs6000/rs6000.c (SPE_CONST_OFFSET_OK): Delete.
28334         (rs6000_legitimate_offset_address_p): Return false for anything in
28335         V2SImode or V2SFmode.
28337 2017-06-13  Segher Boessenkool  <segher@kernel.crashing.org>
28339         * config/rs6000/rs6000-modes.def: Remove all 8-byte vector modes
28340         except V2SF and V2SI.  Rearrange the vector modes, and add comments.
28341         * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove V8QImode
28342         and V4HImode.
28343         (reg_offset_addressing_ok_p): Remove V4HImode and V1DImode.
28344         (rs6000_legitimate_offset_address_p): Ditto.
28345         (rs6000_emit_move): Ditto.
28346         (rs6000_init_builtins): Remove V4HI_type_node.
28348 2017-06-13  Martin Liska  <mliska@suse.cz>
28350         PR sanitize/78204
28351         * asan.c (asan_sanitize_stack_p): Use sanitize_flags_p.
28352         (gate_asan): Likewise.
28353         * asan.h (asan_no_sanitize_address_p): Remove the function.
28354         (sanitize_flags_p): New function.
28355         * builtins.def: Fix coding style.
28356         * common.opt: Use renamed enum value.
28357         * convert.c (convert_to_integer_1): Use sanitize_flags_p.
28358         * doc/extend.texi: Document no_sanitize attribute.
28359         * flag-types.h (enum sanitize_code): Rename SANITIZE_NONDEFAULT
28360         to SANITIZE_UNDEFINED_NONDEFAULT.
28361         * gcc.c (sanitize_spec_function): Use the renamed enum value.
28362         * gimple-fold.c (optimize_atomic_compare_exchange_p):
28363         Use sanitize_flags_p.
28364         * gimplify.c (gimplify_function_tree): Likewise.
28365         * ipa-inline.c (sanitize_attrs_match_for_inline_p): Likewise.
28366         * opts.c (parse_no_sanitize_attribute): New function.
28367         (common_handle_option): Use renamed enum value.
28368         * opts.h (parse_no_sanitize_attribute): Declare.
28369         * tree.c (sanitize_flags_p): New function.
28370         * tree.h: Declared here.
28371         * tsan.c: Use sanitize_flags_p.
28372         * ubsan.c (ubsan_expand_null_ifn): Likewise.
28373         (instrument_mem_ref): Likewise.
28374         (instrument_bool_enum_load): Likewise.
28375         (do_ubsan_in_current_function): Remove the function.
28376         (pass_ubsan::execute): Use sanitize_flags_p.
28377         * ubsan.h: Remove do_ubsan_in_current_function
28378         * tree-cfg.c (print_no_sanitize_attr_value): New function.
28379         (dump_function_to_file): Use it here.
28381 2017-06-13  Martin Jambor  <mjambor@suse.cz>
28383         PR tree-optimization/80803
28384         PR tree-optimization/81063
28385         * tree-sra.c (subtree_mark_written_and_enqueue): Move up in the file.
28386         (propagate_subaccesses_across_link): Enqueue subtree whenever
28387         necessary instead of relying on the caller.
28389 2017-06-13  Martin Jambor  <mjambor@suse.cz>
28391         * tree-sra.c (add_access_to_work_queue): Only enqueue accesses
28392         that have a first_link.
28393         (sort_and_splice_var_accesses): Do not check first_link before
28394         enquing.
28395         (subtree_mark_written_and_enqueue): Likewise.
28396         (propagate_all_subaccesses): Likewise and do not stop at first
28397         parent with a first_link.
28399 2017-06-13  Martin Jambor  <mjambor@suse.cz>
28401         * tree-sra.c (dump_access_tree_1): Fix accidental dumping to stderr
28402         instead of f.
28404 2017-06-13  Yury Gribov  <tetra2005@gmail.com>
28406         * match.pd: New pattern.
28408 2017-06-13  Yury Gribov  <tetra2005@gmail.com>
28410         * tree-vrp.c (is_masked_range_test): New function.
28411         (register_edge_assert_for): Determine ranges for
28412         some bit tests.
28414 2017-06-13  Yury Gribov  <tetra2005@gmail.com>
28416         PR tree-optimization/67328
28417         * fold-const.c (maskable_range_p): New function.
28418         (build_range_check): Generate bittests if possible.
28420 2017-06-13  Martin Liska  <mliska@suse.cz>
28422         * gimple-pretty-print.c (dump_probability): Add new argument.
28423         (dump_edge_probability): Dump both probability and count.
28424         (dump_gimple_label): Likewise.
28425         (dump_gimple_bb_header): Likewise.
28427 2017-06-13  Georg-Johann Lay  <avr@gjlay.de>
28429         PR target/81072
28430         * config/avr/avr-devices.c: Fix indentation.
28431         * config/avr/gen-avr-mmcu-specs.c: Dito.
28433 2017-06-13  Richard Biener  <rguenther@suse.de>
28435         * tree-vect-loop.c (vect_model_reduction_cost): Do not fail,
28436         instead get vector type from stmt_info.
28437         (vectorizable_reduction): Adjust.  Remove dead code.
28439 2017-06-13  Richard Biener  <rguenther@suse.de>
28441         PR middle-end/81065
28442         * fold-const.c (extract_muldiv_1): Remove bogus distribution
28443         case of C * (x * C2 + C3).
28444         (fold_addr_of_array_ref_difference): Properly fold index difference.
28446 2017-06-12  David S. Miller  <davem@davemloft.net>
28448         PR target/80968
28449         * config/sparc/sparc.md (return expander): Emit frame blockage if
28450         function uses alloca.
28452 2017-06-12  Richard Sandiford  <richard.sandiford@linaro.org>
28454         * combine.c (make_field_assignment): Check len rather than the mode
28455         precision when calling force_to_mode.
28457 2017-06-12  Georg-Johann Lay  <avr@gjlay.de>
28459         Support multilibs and devices that see flash in RAM address range.
28461         PR target/81072
28462         * config/avr/avr-arch.h (avr_arch_id) <ARCH_AVRXMEGA3>: New enum.
28463         (avr_mcu_t) <flash_pm_offset>: New field.
28464         (avr_device_specific_features) <AVR_ISA_RCALL>: New enum.
28465         * config/avr/avr.h (AVR_SHORT_CALLS): New define.
28466         (AVR_HAVE_JMP_CALL): Don't set if AVR_SHORT_CALLS.
28467         (AVR_TINY_PM_OFFSET): Remove macro.
28468         * config/avr/avr.opt (-mshort-calls): New option.
28469         * config/avr/gen-avr-mmcu-specs.c (print_mcu)
28470         [*self_spec]: Add / remove -mshort-calls depending on AVR_ISA_RCALL.
28471         * config/avr/avr-c.c (avr_cpu_cpp_builtins)
28472         <__AVR_SHORT_CALLS__>: Built-in define if AVR_SHORT_CALLS.
28473         <__AVR_HAVE_JMP_CALL__>: Use AVR_HAVE_JMP_CALL as condition
28474         instead of avr_arch->have_jmp_call.
28475         <__AVR_PM_BASE_ADDRESS__>: Built-in define if avr_arch->flash_pm_offset.
28476         [AVR_TINY] <__AVR_TINY_PM_BASE_ADDRESS__>: Use
28477         avr_arch->flash_pm_offset to define.
28478         * config/avr/avr-devices.c (avr_arch_types): Add initializers for
28479         new field flash_pm_offset.  Add entry for avrxmega3.
28480         (avr_texinfo): Add entry for avrxmega3.
28481         * config/avr/avr-mcus.def: Add entries for: avrxmega3,
28482         attiny212, attiny214,
28483         attiny412, attiny414, attiny416, attiny417,
28484         attiny814, attiny816, attiny817,
28485         attiny1614, attiny1616, attiny1617,
28486         attiny3214, attiny3216, attiny3217.
28487         * config/avr/avr.c (avr_assemble_integer)[AVR_TINY]: Use
28488         avr_arch->flash_pm_offset instead of AVR_TINY_PM_OFFSET.
28489         (avr_print_operand_address) [AVR_TINY]: Same.
28490         (avr_asm_init_sections) <readonly_data_section>: Only patch
28491         callback if avr_arch->flash_pm_offset = 0.
28492         (avr_asm_named_section) <avr_need_copy_data_p>: Skip setting it
28493         for rodata if avr_arch->flash_pm_offset != 0.
28494         (avr_encode_section_info) [AVR_TINY]: Adjust comment.
28495         * config/avr/genmultilib.awk (dir_rcall, opt_rcall): New vars.
28496         (opts) [AVR_ISA_RCALL]: Append opt_rcall.
28497         (m_options): Append opt_rcall.
28498         (m_dirnames): Append dir_rcall.
28499         * config/avr/t-multilib: Regenerate.
28501         * configure.ac [target=avr]: Check whether avrxmega3 default
28502         linker description file works as needed.
28503         * configure: Regenerate.
28504         * doc/avr-mmcu.texi: Regenerate.
28505         * doc/invoke.texi (AVR Options) <-mshort-calls>: Document it.
28506         <__AVR_ARCH__>: Document avrxmega3 and 103.
28507         <__AVR_HAVE_JMP_CALL__>: Adjust documentation.
28508         <__AVR_SHORT_CALLS__>: Document it.
28509         <__AVR_PM_BASE_ADDRESS__>: Document it.
28510         * doc/extend.texi (AVR Options) <-mshort-calls>: Document it.
28511         (AVR Variable Attributes) <progmem>: Document this is
28512         not needed for avrxmega3.
28513         (AVR Named Address Spaces) <__flash>: Dito.
28515 2017-06-12  Jan Hubicka  <hubicka@ucw.cz>
28517         * cgraph.c (cgraph_node::dump): Complain about profile insanities.
28519 2017-06-12  Doug Rupp  <rupp@adacore.com>
28521         * config.gcc (*-*-vxworks*): Set use_gcc_stdint to "provide".
28522         Append vxworks-stdint.h to the tm_file list.
28523         * config/vxworks-stdint.h: New file.
28525 2017-06-12  Martin Liska  <mliska@suse.cz>
28527         PR tree-optimization/81041
28528         * tree-profile.c (gimple_gen_ic_func_profiler):
28529         Create an extra BB in profile-generate
28530         (gimple_gen_time_profiler): Likewise.
28532 2017-06-12  Jakub Jelinek  <jakub@redhat.com>
28534         PR tree-optimization/81003
28535         * tree-ssa-reassoc.c (force_into_ssa_name): New function.
28536         (update_range_test): Use it instead of force_gimple_operand_gsi.
28538 2017-06-12  Richard Biener  <rguenther@suse.de>
28540         PR tree-optimization/81053
28541         * tree-vect-loop.c (vect_is_simple_reduction): Handle PHI
28542         with backedge value not defined in loop.  Simplify def stmt
28543         compute.
28545 2017-06-11  Tom de Vries  <tom@codesourcery.com>
28547         PR target/79939
28548         * config/nvptx/nvptx.c (nvptx_cannot_force_const_mem): New function.
28549         Return true.
28550         (TARGET_CANNOT_FORCE_CONST_MEM): Redefine to
28551         nvptx_cannot_force_const_mem.
28553 2017-06-10  Jan Hubicka  <hubicka@ucw.cz>
28555         * opts.c (finish_options): Move test for flag_split_stack after
28556         it has been initialized.
28558 2017-06-11  Jason Merrill  <jason@redhat.com>
28560         * tree.h (id_equal): New.
28561         * dwarf2out.c, hsa-gen.c, ipa-devirt.c, omp-expand.c,
28562         omp-simd-clone.c, read-rtl-function.c, tree-chkp.c, tree.c: Use it
28563         instead of strcmp of IDENTIFIER_POINTER.
28565 2017-06-10  Jan Hubicka  <hubicka@ucw.cz>
28567         * ipa-inline-transform.c: Include function.h, cfg.h and basic-block.h
28568         (mark_all_inlined_calls_cdtor): Fix formating.
28569         (inline_transform): Rescale profile before inlining.
28571 2017-06-10  Jan Hubicka  <hubicka@ucw.cz>
28573         * cgraph.h (cgraph_edge::clone): Update prototype.
28574         * cgraphclones.c (cgraph_edge::clone): Update profile scaling.
28575         (cgraph_node::create_clone): Update.
28576         (cgraph_node::create_version_clone): Update.
28577         * tree-inline.c (copy_bb): Update.
28578         (expand_call_inline): Update.
28580 2017-06-10  Segher Boessenkool  <segher@kernel.crashing.org>
28582         * config/rs6000/rs6000.c (emit_vrsave_prologue): New function,
28583         factored out from ...
28584         (rs6000_emit_prologue): ... here.
28586 2017-06-10  Segher Boessenkool  <segher@kernel.crashing.org>
28588         * config/rs6000/rs6000.c (emit_split_stack_prologue): New function,
28589         factored out from ...
28590         (rs6000_emit_prologue): ... here.
28592 2017-06-10  Jan Hubicka  <hubicka@ucw.cz>
28594         * predict.c (drop_profile): Also drop individual bb/edge and cgraph
28595         edge counts.
28596         (handle_missing_profiles): Fix computation of tp_first_run.
28597         (counts_to_freqs): Do not touch freqs when count is 0.
28599 2017-06-10  Jan Hubicka  <hubicka@ucw.cz>
28601         * cgraphbuild.c (cgraph_edge::rebuild_references): Do not touch
28602         profile.
28604 2017-06-10  Tom de Vries  <tom@codesourcery.com>
28606         * doc/sourcebuild.texi (Effective-Target Keywords, Environment
28607         attributes): Document signal effective target.
28609 2017-06-10  Tom de Vries  <tom@codesourcery.com>
28611         * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
28612         Document effective target stack_size.
28614 2017-06-09  David Malcolm  <dmalcolm@redhat.com>
28616         * diagnostic.c (diagnostic_report_diagnostic): Only add fixits
28617         to the edit_context if they can be auto-applied.
28619 2017-06-9  Ian Lance Taylor  <iant@golang.org>
28621         * opts.c (finish_options): If -fsplit-stack, disable implicit
28622         -forder-blocks-and-partition.
28623         * doc/invoke.texi (Optimize Options): Document that when using
28624         -fsplit-stack -forder-blocks-and-partition is not implicitly
28625         enabled.
28627 2017-06-09  Jan Hubicka  <hubicka@ucw.cz>
28629         * builtin-attrs.def (ATTR_NORETURN_NOTHROW_LEAF_COLD_LIST,
28630         ATTR_CONST_NORETURN_NOTHROW_LEAF_COLD_LIST,
28631         ATTR_TMPURE_NORETURN_NOTHROW_LEAF_COLD_LIST): New.
28632         * builtins.def (abort, trap, unreachable): Declare cold.
28633         * calls.c (flags_from_decl_or_type): Lookup ECF_COLD.
28634         * tree-core.h (ECF_COLD): New.
28635         * tree.c (set_call_expr_flags): Handle ECF_COLD.
28636         (build_common_builtin_nodes): Mark unreachable and abort as cold.
28638 2017-06-09  Jan Hubicka  <hubicka@ucw.cz>
28640         * predict.c (unlikely_executed_stmt_p): Cleanup.
28642 2017-06-09  Richard Biener  <rguenther@suse.de>
28644         * tree-ssa-loop-im.c (execute_sm): Do not force multi-threaded
28645         model if the ref is always written to.
28647 2017-06-09  Tamar Christina  <tamar.christina@arm.com>
28649         * config/aarch64/aarch64.md (lrint<GPF:mode><GPI:mode>2): New.
28651 2017-06-09  Tamar Christina  <tamar.christina@arm.com>
28653         * config/arm/arm.c (arm_rtx_costs_internal): Make sdiv more expensive
28654         than udiv.
28656 2017-06-09  Tom de Vries  <tom@codesourcery.com>
28658         PR target/80855
28659         * config/nvptx/nvptx.md (define_expand "mov<QHSDISDFM>"): Error out with
28660         "target cannot support label values" when encountering LABEL_REF.
28662 2017-06-09  Martin Liska  <mliska@suse.cz>
28664         * tree-profile.c (gimple_gen_ic_profiler): Update comment.
28665         (gimple_gen_ic_func_profiler): Emit direct comparison
28666         of __gcov_indirect_call_callee with NULL.
28667         (gimple_gen_time_profiler): Change probability from
28668         PROB_VERY_UNLIKELY to PROB_UNLIKELY.
28670 2017-06-09  Jan Hubicka  <hubicka@ucw.cz>
28672         * profile.c (edge_gcov_counts): Turn to pointer.
28673         (compute_branch_probabilities, compute_branch_probabilities): Update.
28674         (branch_prob): Do not clear edge_gcov_count.
28675         * profile.h (edge_gcov_counts): Turn to pointer.
28676         (edge_gcov_count): Update.
28678 2017-06-09  Jan Hubicka  <hubicka@ucw.cz>
28680         * gimple.h (gimple_check_failed): Mark cold.
28682 2017-06-09  Richard Biener  <rguenther@suse.de>
28684         PR tree-optimization/66623
28685         * tree-vect-loop.c (vect_is_simple_reduction): Cleanup,
28686         refactor check_reduction into two parts, properly computing
28687         whether we have to check reduction validity for outer loop
28688         vectorization.
28690 2017-06-09  Richard Biener  <rguenther@suse.de>
28692         PR tree-optimization/79483
28693         * graphite-scop-detection.c (order): New global.
28694         (get_order): Compute bb to order mapping that satisfies code
28695         generation constraints.
28696         (cmp_pbbs): New helper.
28697         (build_scops): Start domwalk at entry block, sort generated
28698         pbbs.
28700 2017-06-09  Richard Biener  <rguenther@suse.de>
28702         PR middle-end/81007
28703         * ipa-polymorphic-call.c
28704         (ipa_polymorphic_call_context::restrict_to_inner_class):
28705         Skip FIELD_DECLs with error_mark_node type.
28706         * passes.def (all_lowering_passes): Run pass_build_cgraph_edges
28707         last again.
28709 2017-06-09  Martin Liska  <mliska@suse.cz>
28711         * predict.c (struct branch_predictor): New struct.
28712         (test_prediction_value_range): New test.
28713         (predict_c_tests): New function.
28714         * selftest-run-tests.c (selftest::run_tests): Run the function.
28715         * selftest.h: Declare new tests.
28717 2017-06-09  Segher Boessenkool  <segher@kernel.crashing.org>
28719         PR target/80966
28720         * config/rs6000/rs6000.c (rs6000_emit_allocate_stack): Assert that
28721         gen_add3_insn did not fail.
28722         * config/rs6000/rs6000.md (add<mode>3): If asked to add a constant to
28723         r0, construct that number in a temporary reg and add that reg to r0.
28724         If asked to put the result in r0 as well, fail.
28726 2017-06-08  Will Schmidt  <will_schmidt@vnet.ibm.com>
28728         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add handling
28729         for early expansion of vec_eqv.
28731 2017-06-08  Jakub Jelinek  <jakub@redhat.com>
28733         PR middle-end/81005
28734         * ubsan.c (instrument_null): Avoid pointless code temporary.
28735         (pass_ubsan::execute): Instrument aggregate arguments of calls.
28737 2017-06-08  Uros Bizjak  <ubizjak@gmail.com>
28739         PR target/81015
28740         Revert:
28741         2016-12-14  Uros Bizjak  <ubizjak@gmail.com>
28743         PR target/59874
28744         * config/i386/i386.md (*ctzhi2): New insn_and_split pattern.
28745         (*clzhi2): Ditto.
28747 2017-06-08  Jan Hubicka  <hubicka@ucw.cz>
28749         * predict.c (unlikely_executed_edge_p): Move ahead.
28750         (probably_never_executed_edge_p): Use it.
28752 2017-06-08  Alexander Ivchenko  <alexander.ivchenko@intel.com>
28754         PR middle-end/79988
28755         * tree-chkp.c (chkp_gimple_call_builtin_p): Remove
28756         gimple_call_builtin_p call.
28758 2017-06-08  Jan Hubicka  <hubicka@ucw.cz>
28760         * system.h (fancy_abort): Annotate by ATTRIBUTE_COLD.
28761         * rtl.h (rtl_check_failed_bounds, rtl_check_failed_type1,
28762         rtl_check_failed_type2, rtl_check_failed_code1,
28763         rtl_check_failed_code2, rtl_check_failed_code_mode,
28764         rtl_check_failed_block_symbol, cwi_check_failed_bounds,
28765         rtvec_check_failed_bounds, rtl_check_failed_flag,
28766         _fatal_insn_not_found, _fatal_insn): Likewise.
28767         * tree.h (tree_contains_struct_check_failed,
28768         tree_check_failed, tree_not_check_failed,
28769         tree_class_check_failed, tree_range_check_failed,
28770         tree_not_class_check_failed, tree_int_cst_elt_check_failed,
28771         tree_vec_elt_check_failed, phi_node_elt_check_failed,
28772         tree_operand_check_failed, omp_clause_check_failed,
28773         omp_clause_operand_check_failed, omp_clause_range_check_failed):
28774         Likewise.
28776 2017-06-08  Jan Hubicka  <hubicka@ucw.cz>
28778         * cgraph.c (cgraph_edge::maybe_hot_p): Do not check
28779         flag_branch_probabilities.
28780         * ipa-inline.c (edge_badness): Likewise.
28781         * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
28782         * postreload-gcse.c (eliminate_partially_redundant_load): Likewise.
28783         * predict.c (maybe_hot_frequency_p): Likewise.
28784         (probably_never_executed): Likewise.
28785         * sched-ebb.c (schedule_ebbs): Likewise.
28786         * sched-rgn.c (find_single_block_region): Likewise.
28787         * tracer.c (tail_duplicate): Likewise.
28789 2017-06-08  Jan Hubicka  <hubicka@ucw.cz>
28791         * opts.c (finish_options): x_flag_reorder_blocks_and_partition no
28792         longer requires x_flag_profile_use.
28794 2017-06-08  Jan Hubicka  <hubicka@ucw.cz>
28796         * cfgrtl.c (cfg_layout_initialize): Check crtl->has_bb_partition
28797         instead of flag_reorder_blocks_and_partition.
28798         * dbxout.c (dbxout_function_end): Likewise.
28799         * dwarf2out.c (gen_subprogram_die): Likewise.
28800         * haifa-sched.c (sched_create_recovery_edges): Likewise.
28801         * hw-doloop.c (reorg_loops): Likewise.
28802         * varasm.c (assemble_start_function,
28803         assemble_end_function): Likewise.
28804         (decide_function_section): Do not check for
28805         flag_reorder_blocks_and_partition.
28807 2017-06-08  Alexander Ivchenko  <alexander.ivchenko@intel.com>
28809         * tree-chkp.c (chkp_get_hard_register_var_fake_base_address):
28810         New function.
28811         (chkp_get_hard_register_fake_addr_expr): Ditto.
28812         (chkp_build_addr_expr): Add check for hard reg case.
28813         (chkp_parse_array_and_component_ref): Ditto.
28814         (chkp_find_bounds_1): Ditto.
28815         (chkp_process_stmt): Don't generate bounds store for
28816         hard reg case.
28818 2017-06-08  Jan Hubicka  <hubicka@ucw.cz>
28820         * predict.c (maybe_hot_bb_p): Do not check profile status.
28821         (maybe_hot_edge_p): Likewise.
28822         (probably_never_executed): Check for zero counts even if profile
28823         is not read.
28824         (unlikely_executed_edge_p): New function.
28825         (unlikely_executed_stmt_p): New function.
28826         (unlikely_executed_bb_p): New function.
28827         (set_even_probabilities): Use unlikely predicates.
28828         (combine_predictions_for_bb): Likewise.
28829         (predict_paths_for_bb): Likewise.
28830         (predict_paths_leading_to_edge): Likewise.
28831         (determine_unlikely_bbs): New function.
28832         (estimate_bb_frequencies): Use it.
28833         (compute_function_frequency): Use zero counts even if profile is
28834         not read.
28835         * profile-count.h: Fix typo.
28837 2017-08-08  Julia Koval  <julia.koval@intel.com>
28839         * config/i386/avx512bwintrin.h (_mm512_mask_cvtepi16_storeu_epi8,
28840         _mm512_mask_cvtsepi16_storeu_epi8,
28841         _mm512_mask_cvtusepi16_storeu_epi8): New intrinsics.
28842         * config/i386/avx512vlbwintrin.h (_mm256_mask_cvtepi16_storeu_epi8,
28843         _mm_mask_cvtsepi16_storeu_epi8, _mm256_mask_cvtsepi16_storeu_epi8,
28844         _mm_mask_cvtusepi16_storeu_epi8, _mm256_mask_cvtusepi16_storeu_epi8,
28845         _mm_mask_cvtepi16_storeu_epi8): New intrinsics.
28846         * config/i386/i386-builtin-types.def (PV8Q, V8QI): New pointer type.
28847         (VOID_FTYPE_PV32QI_V32HI_USI, VOID_FTYPE_PV8QI_V8HI_UQI,
28848         VOID_FTYPE_PV16QI_V16HI_UHI): New function types.
28849         * config/i386/i386-builtin.def (__builtin_ia32_pmovwb128mem_mask,
28850         __builtin_ia32_pmovwb256mem_mask, __builtin_ia32_pmovswb128mem_mask,
28851         __builtin_ia32_pmovswb256mem_mask, __builtin_ia32_pmovuswb128mem_mask,
28852         __builtin_ia32_pmovuswb256mem_mask,
28853         __builtin_ia32_pmovuswb512mem_mask, __builtin_ia32_pmovswb512mem_mask)
28854         __builtin_ia32_pmovwb512mem_mask): New builtins.
28856 2017-08-08  Julia Koval  <julia.koval@intel.com>
28858         PR target/73350,80862
28859         * config/i386/subst.md (round): Fix round pattern.
28860         * config/i386/i386.c (ix86_erase_embedded_rounding):
28861         Fix erasing rounding for the fixed pattern.
28863 2017-06-08  Jan Hubicka  <hubicka@ucw.cz>
28865         * cfgbuild.c (find_many_sub_basic_blocks): Fix thinko.
28867 2017-06-08  Martin Liska  <mliska@suse.cz>
28869         PR gcov-profile/80911
28870         * gcov.c (block_info::block_info): New constructor.
28872 2017-06-07  Carl Love  <cel@us.ibm.com>
28874         * config/rs6000/rs6000-c: The return type of the following
28875         built-in functions was implemented as int not long long.  Fix sign
28876         of return value for the unsigned version of vec_mulo and vec_mule.
28877         vector unsigned long long vec_bperm (vector unsigned long long,
28878                                              vector unsigned char)
28879         vector signed long long vec_mule (vector signed int,
28880                                           vector signed int)
28881         vector unsigned long long vec_mule (vector unsigned int,
28882                                             vector unsigned int)
28883         vector signed long long vec_mulo (vector signed int,
28884                                           vector signed int)
28885         vector unsigned long long vec_mulo (vector unsigned int,
28886                                             vector unsigned int)
28887         * doc/extend.texi: Fix the documentation for the built-in
28888         functions.
28890 2017-06-07  Carl Love  <cel@us.ibm.com>
28892         PR target/80982
28893         * config/rs6000/altivec.md (double<mode>2): Fix the implementation of
28894         for BE.
28896 2017-06-07  Carl Love  <cel@us.ibm.com>
28898         * config/rs6000/altivec.md: Fix argument swizzle in vec_doublel
28899         support, Generate       doublehv for signed int/float for BE case only.
28901 2017-06-07  Alexander Monakov  <amonakov@ispras.ru>
28903         * doc/invoke.texi (mcx16): Rewrite.
28905 2017-06-07  Segher Boessenkool  <segher@kernel.crashing.org>
28907         * config/rs6000/predicates.md (rs6000_nonimmediate_operand): Delete.
28908         * config/rs6000/rs6000.md (*movsi_internal1, movsi_from_sf,
28909         *mov<mode>_softfloat, and an anonymous splitter): Use
28910         nonimmediate_operand instead of rs6000_nonimmediate_operand.
28912 2017-06-07  Segher Boessenkool  <segher@kernel.crashing.org>
28914         * config/rs6000/darwin.h (REGISTER_NAMES): Delete the SPE_ACC and
28915         SPEFSCR registers.
28916         * config/rs6000/rs6000.c (rs6000_reg_names, alt_reg_names): Ditto.
28917         (enum rs6000_reg_type): Delete SPE_ACC_TYPE and SPEFSCR_REG_TYPE.
28918         (rs6000_debug_reg_global): Adjust.
28919         (rs6000_init_hard_regno_mode_ok): Adjust.
28920         (rs6000_dbx_register_number): Adjust.
28921         * config/rs6000/rs6000.h (FIRST_PSEUDO_REGISTER): Change to 115.
28922         (FIXED_REGISTERS, CALL_USED_REGISTERS, CALL_REALLY_USED_REGISTERS):
28923         Remove SPE_ACC and SPEFSCR.
28924         (REG_ALLOC_ORDER): Ditto.
28925         (FRAME_POINTER_REGNUM): Change to 111.
28926         (enum reg_class): Remove the SPE_ACC and SPEFSCR registers.
28927         (REG_CLASS_NAMES): Ditto.
28928         (REG_CLASS_CONTENTS): Delete the SPE_ACC and SPEFSCR registers.
28929         (REGISTER_NAMES): Ditto.
28930         (ADDITIONAL_REG_NAMES): Ditto.
28931         (rs6000_reg_names): Ditto.
28932         * config/rs6000/rs6000.md: Renumber some register number
28933         define_constants.
28935 2017-06-07  Segher Boessenkool  <segher@kernel.crashing.org>
28937         * config/rs6000/darwin.h (REGISTER_NAMES): Delete the SPE high
28938         registers.
28939         * config/rs6000/rs6000.c (rs6000_reg_names, alt_reg_names): Ditto.
28940         * config/rs6000/rs6000.h (FIRST_PSEUDO_REGISTER): Change from 149
28941         to 117.
28942         (DWARF_REG_TO_UNWIND_COLUMN): Do not define.
28943         (FIXED_REGISTERS, CALL_USED_REGISTERS, CALL_REALLY_USED_REGISTERS):
28944         Delete the SPE high registers.
28945         (REG_ALLOC_ORDER): Ditto.
28946         (enum reg_class): Remove SPE_HIGH_REGS.
28947         (REG_CLASS_NAMES): Ditto.
28948         (REG_CLASS_CONTENTS): Delete the SPE high registers.
28949         (REGISTER_NAMES): Ditto.
28950         (rs6000_reg_names): Ditto.
28951         * doc/tm.texi.in: Remove SPE as example.
28952         * doc/tm.texi: Regenerate.
28954 2017-06-07  Segher Boessenkool  <segher@kernel.crashing.org>
28956         * config/rs6000/8540.md (ppc8540_brinc): Delete.
28957         * config/rs6000/e500mc.md (e500mc_brinc): Delete.
28958         * config/rs6000/e500mc64.md (e500mc64_brinc): Delete.
28959         * config/rs6000/rs6000.md (type): Remove "brinc".
28961 2017-06-07  Segher Boessenkool  <segher@kernel.crashing.org>
28963         * config.gcc (powerpc*-*-*): Don't add spe.h to extra_headers.
28964         (powerpc*-linux*spe*): Use ${cpu_type} instead of rs6000.
28965         * config/rs6000/linuxspe.h: Delete file.
28966         * config/rs6000/rs6000.md: Don't include spe.md.
28967         * config/rs6000/spe.h: Delete file.
28968         * config/rs6000/spe.md: Delete file.
28969         * config/rs6000/t-rs6000: Remove spe.md.
28971 2017-06-07  Segher Boessenkool  <segher@kernel.crashing.org>
28973         * config/rs6000/predicates.md (reg_or_mem_operand): Reformat.
28974         (reg_or_none500mem_operand): Delete.
28975         * config/rs6000/rs6000.md (extendsfdf2): Use reg_or_mem_operand
28976         instead of reg_or_none500mem_operand.
28978 2017-06-07  Segher Boessenkool  <segher@kernel.crashing.org>
28980         * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
28981         handling of SPE flags.
28982         * config/rs6000/rs6000.opt (-mspe, -mspe=no, -mspe=yes): Delete.
28984 2017-06-07  Segher Boessenkool  <segher@kernel.crashing.org>
28986         * config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
28987         SPE ABI handling.
28988         * config/rs6000/paired.md (paired_negv2sf2): Rename to negv2sf2.
28989         (paired_absv2sf2, paired_addv2sf3, paired_subv2sf3, paired_mulv2sf3,
28990         paired_divv2sf3): Similar.
28991         * config/rs6000/predicates.md: Replace TARGET_SPE, TARGET_SPE_ABI,
28992         SPE_VECTOR_MODE and SPE_HIGH_REGNO_P by 0; simplify.
28993         * config/rs6000/rs6000-builtin.def: Delete RS6000_BUILTIN_E and
28994         RS6000_BUILTIN_S.
28995         Delete BU_SPE_1, BU_SPE_2, BU_SPE_3, BU_SPE_E, BU_SPE_P, and BU_SPE_X.
28996         Rename the paired_* instruction patterns.
28997         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Do not
28998         define __SPE__.
28999         * config/rs6000/rs6000-protos.h (invalid_e500_subreg): Delete.
29000         * config/rs6000/rs6000.c: Delete RS6000_BUILTIN_E and RS6000_BUILTIN_S.
29001         (struct rs6000_stack): Delete fields spe_gp_save_offset, spe_gp_size,
29002         spe_padding_size, and spe_64bit_regs_used.  Replace TARGET_SPE and
29003         TARGET_SPE_ABI with 0, simplify.  Replace SPE_VECTOR_MODE with
29004         PAIRED_VECTOR_MODE.
29005         (struct machine_function): Delete field spe_insn_chain_scanned_p.
29006         (spe_func_has_64bit_regs_p): Delete.
29007         (spe_expand_predicate_builtin): Delete.
29008         (spe_expand_evsel_builtin): Delete.
29009         (TARGET_DWARF_REGISTER_SPAN): Do not define.
29010         (TARGET_MEMBER_TYPE_FORCES_BLK): Do not define.
29011         (invalid_e500_subreg): Delete.
29012         (rs6000_legitimize_address): Always force_reg op2 as well, for
29013         paired single memory accesses.
29014         (rs6000_member_type_forces_blk): Delete.
29015         (rs6000_spe_function_arg): Delete.
29016         (rs6000_expand_unop_builtin): Delete SPE handling.
29017         (rs6000_expand_binop_builtin): Ditto.
29018         (spe_expand_stv_builtin): Delete.
29019         (bdesc_2arg_spe): Delete.
29020         (spe_expand_builtin): Delete.
29021         (spe_expand_predicate_builtin): Delete.
29022         (spe_expand_evsel_builtin): Delete.
29023         (rs6000_invalid_builtin): Remove RS6000_BTM_SPE handling.
29024         (spe_init_builtins): Delete.
29025         (spe_func_has_64bit_regs_p): Delete.
29026         (savres_routine_name): Delete "info" parameter.  Adjust callers.
29027         (rs6000_emit_stack_reset): Ditto.
29028         (rs6000_dwarf_register_span): Delete.
29029         * config/rs6000/rs6000.h (TARGET_SPE_ABI, TARGET_SPE,
29030         UNITS_PER_SPE_WORD, SPE_HIGH_REGNO_P, SPE_SIMD_REGNO_P,
29031         SPE_VECTOR_MODE, RS6000_BTM_SPE, RS6000_BUILTIN_E, RS6000_BUILTIN_S):
29032         Delete.
29033         * config/rs6000/rs6000.md (FIRST_SPE_HIGH_REGNO, LAST_SPE_HIGH_REGNO):
29034         Delete.
29035         * config/rs6000/rs6000.opt (-mabi=spe, -mabi=no-spe): Delete.
29036         * config/rs6000/spe.md: Delete every pattern that uses TARGET_SPE.
29037         * config/rs6000/vector.md (absv2sf2, negv2sf2, addv2sf3, subv2sf3,
29038         mulv2sf3, divv2sf3): Delete expanders.
29040 2017-06-07  Segher Boessenkool  <segher@kernel.crashing.org>
29042         config/rs6000/rs6000.md (UNSPEC_MV_CR_GT): Delete.
29044 2017-06-07  Segher Boessenkool  <segher@kernel.crashing.org>
29046         * config/rs6000/rs6000-protos.h (output_e500_flip_gt_bit): Delete.
29047         * config/rs6000/rs6000.c: Ditto.
29049 2017-06-07  Segher Boessenkool  <segher@kernel.crashing.org>
29051         * config/rs6000/predicated.md (rs6000_cbranch_operator): Delete.
29052         * config/rs6000/rs6000.md: Replace rs6000_cbranch_operator by
29053         comparison_operator.
29055 2017-06-07  Segher Boessenkool  <segher@kernel.crashing.org>
29057         * config/rs6000/rs6000.c: Remove everything related to -mfloat-gprs.
29058         * config/rs6000/rs6000.opt: Ditto.
29059         * config/rs6000/t-rtems: Ditto.
29061 2017-06-07  Segher Boessenkool  <segher@kernel.crashing.org>
29063         * config/rs6000/predicates.md: Replace TARGET_E500_DOUBLE and
29064         TARGET_E500_SINGLE by 0, simplify.
29065         * config/rs6000/rs6000.c: Ditto.
29066         (rs6000_option_override_internal): Delete CHECK_E500_OPTIONS.
29067         (spe_build_register_parallel): Delete.
29068         * config/rs6000/rs6000.h: Delete TARGET_E500_SINGLE,
29069         TARGET_E500_DOUBLE, and CHECK_E500_OPTIONS.
29070         * config/rs6000/rs6000.md: Replace TARGET_E500_DOUBLE,
29071         TARGET_E500_SINGLE, and <E500_CONVERT> by 0, simplify.
29072         (E500_CONVERT): Delete.
29073         * config/rs6000/spe.md: Remove many patterns and all define_constants.
29075 2017-06-07  Segher Boessenkool  <segher@kernel.crashing.org>
29077         * config/rs6000/darwin.md: Replace TARGET_FPRS by 1 and simplify.
29078         * config/rs6000/dfp.md: Ditto.
29079         (negdd2, *negdd2_fpr): Merge.
29080         (absdd2, *absdd2_fpr): Merge.
29081         (negtd2, *negtd2_fpr): Merge.
29082         (abstd2, *abstd2_fpr): Merge.
29083         * config/rs6000/e500.h: Delete file.
29084         * config/rs6000/predicates.md (rs6000_cbranch_operator): Replace
29085         TARGET_FPRS by 1 and simplify.
29086         * config/rs6000/rs6000-c.c: Ditto.
29087         * config/rs6000/rs6000.c: Ditto.  Also replace TARGET_SF_SPE and
29088         TARGET_DF_SPE by 0.
29089         * config/rs6000/rs6000.h: Ditto.  Delete TARGET_SF_SPE and
29090         TARGET_DF_SPE.
29091         * config/rs6000/rs6000.md: Ditto.
29092         (floatdidf2, *floatdidf2_fpr): Merge.
29093         (move_from_CR_gt_bit): Delete.
29094         * config/rs6000/spe.md: Replace TARGET_FPRS by 1 and simplify.
29095         (E500_CR_IOR_COMPARE): Delete.
29096         (All patterns that require !TARGET_FPRS): Delete.
29097         * config/rs6000/vsx.md: Replace TARGET_FPRS by 1 and simplify.
29099 2017-06-07  Bin Cheng  <bin.cheng@arm.com>
29101         * passes.def (pass_iv_canon): Move before pass_loop_distribution.
29103 2017-06-07  Bin Cheng  <bin.cheng@arm.com>
29105         * graphds.c (add_edge): Intitialize edge's attached data.
29106         (foll_in_subgraph, dfs_fst_edge, dfs_next_edge): New function
29107         pointer parameter.  Call pointed function on each edge during
29108         graph traversing.  Skip traversing the edge when the function
29109         returns true.
29110         (graphds_dfs, graphds_scc): Ditto.
29111         (for_each_edge): New parameter.  Pass the new parameter to callback
29112         function.
29113         * graphds.h (skip_edge_callback): New function pointer type.
29114         (graphds_dfs, graphds_scc): New function pointer parameter.
29115         (graphds_edge_callback, for_each_edge): New parameter.
29117 2017-06-07  Bin Cheng  <bin.cheng@arm.com>
29119         * tree-vect-data-refs.c (vect_mark_for_runtime_alias_test): Factor
29120         out code checking if runtime alias check is possible to below ...
29121         Call the new function.
29122         * tree-data-ref.c (runtime_alias_check_p): ... to new function.
29123         * tree-data-ref.h (runtime_alias_check_p): New decalaration.
29125 2017-06-07  Marek Polacek  <polacek@redhat.com>
29127         PR sanitizer/80932
29128         * fold-const.c (extract_muldiv_1) <case MINUS_EXPR>: Add
29129         TYPE_OVERFLOW_WRAPS check.
29131 2017-06-07  Bin Cheng  <bin.cheng@arm.com>
29133         * tree-vect-loop-manip.c (vect_do_peeling): Don't skip vector loop
29134         if versioning is required.
29135         * tree-vect-loop.c (vect_analyze_loop_2): Merge niter check for loop
29136         peeling with the check for versioning.
29138 2017-06-07  Bin Cheng  <bin.cheng@arm.com>
29140         * tree-vectorizer.h (vect_build_loop_niters): New parameter.
29141         * tree-vect-loop-manip.c (vect_build_loop_niters): New parameter.
29142         Set true to new parameter if new ssa variable is defined.
29143         (vect_gen_vector_loop_niters): Refactor.  Set range information
29144         for the new vector loop bound variable.
29145         (vect_do_peeling): Ditto.
29147 2017-06-07  Bin Cheng  <bin.cheng@arm.com>
29149         * tree-affine.c (ssa.h): Include header file.
29150         (tree_to_aff_combination): Handle (T1)(X - CST) when inner type
29151         has wrapping overflow behavior.
29153 2017-06-07  Bin Cheng  <bin.cheng@arm.com>
29155         * tree-affine.c (tree_to_aff_combination): Handle (T1)(X + X).
29157 2017-06-07  Bin Cheng  <bin.cheng@arm.com>
29159         (aff_combination_expand): Move (T1)(X *+- CST) simplification to ...
29160         (tree_to_aff_combination): ... here.
29162 2017-06-07  Bin Cheng  <bin.cheng@arm.com>
29164         * tree-ssa-loop-ivopts.c (ivopts_estimate_reg_pressure): New
29165         reg_pressure model function.
29166         (ivopts_global_cost_for_size): Delete.
29167         (determine_set_costs, iv_ca_recount_cost): Call new model function
29168         ivopts_estimate_reg_pressure.
29170 2017-06-07  Tamar Christina  <tamar.christina@arm.com>
29172         * config/aarch64/aarch64.c (aarch64_rtx_costs): Make sdiv more
29173         expensive than udiv.  Remove floating point cases from mod.
29175 2017-06-07  Tamar Christina  <tamar.christina@arm.com>
29177         * config/arm/aarch-cost-tables.h (cortexa53_extra_cost):
29178         Increase idiv cost.
29180 2017-06-07  Tamar Christina  <tamar.christina@arm.com>
29182         * config/aarch64/aarch64.md
29183         (copysignsf3): Fix mask generation.
29185 2017-06-07  Jakub Jelinek  <jakub@redhat.com>
29187         * dumpfile.h (enum tree_dump_index): Rename TDI_generic to
29188         TDI_gimple.
29189         (class dump_manager): Add register_dumps method.
29190         * dumpfile.c: Include langhooks.h.
29191         (dump_files): Use 0 instead of 3/4/5 for TDI_{original,gimple,nested}.
29192         (FIRST_AUTO_NUMBERED_DUMP): Decrease to 1.
29193         (FIRST_ME_AUTO_NUMBERED_DUMP): Define.
29194         (dump_manager::dump_register): Start with 512 entries instead of 32.
29195         (dump_manager::register_dumps): New method.
29196         * toplev.c (general_init): Instead of invoking register_dumps
29197         langhook, invoke register_dumps method on the dump manager.
29198         * gimplify.c (gimplify_function_tree): Use TDI_gimple instead of
29199         TDI_generic.
29201 2017-06-07  Richard Sandiford  <richard.sandiford@linaro.org>
29203         * doc/md.texi: Clarify the restrictions on a define_insn condition.
29204         Say that # requires an associated define_split to exist, and that
29205         the define_split must be suitable for use after register allocation.
29207 2017-06-06  Jan Hubicka  <hubicka@ucw.cz>
29209         * cfgbuild.c (find_bb_boundaries): Initialize profile of split blocks.
29210         (compute_outgoing_frequencies): Also initialize zero counts.
29211         (find_many_sub_basic_blocks): Do not produce uninitialized profile
29212         around loops; preserve more of profile when nothing changes.
29214 2017-06-06  Jim Wilson  <jim.wilson@linaro.org>
29216         * config/aarch64/aarch64-cost-tables.h (qdf24xx_extra_costs): Move to
29217         here.
29218         * config/arm/aarch-cost-tables.h (qdf24xx_extra_costs): From here.
29219         * config/arm/arm-cpu-cdata.h: Regenerate.
29220         * config/arm/arm-cpu-data.h, config/arm/arm-cpu.h: Likewise.
29221         * config/arm/arm-tables.opt, config/arm/arm-tune.md: Likewise.
29222         * config/arm/arm-cpus.in: Delete falkor and qdf24xx entries.
29223         * config/arm/arm.c (arm_qdf24xx_tune): Delete.
29224         * config/arm/bpabi.h (BE8_LINK_SPEC): Delete falkor and qdf24xx
29225         support.
29226         * config/arm/t-aprofile (MULTILIB_MATCHES): Delete falkor and qdf24xx
29227         support.
29228         * config/arm/t-rmprofile: Likewise.
29229         * doc/invoke.texi (ARM Options): Drop falkor and qdf24xx support.
29231 2017-06-06  David S. Miller  <davem@davemloft.net>
29233         PR target/80968
29234         * config/sparc/sparc.c (sparc_expand_prologue): Emit frame
29235         blockage if function uses alloca.
29237 2017-06-06  Maxim Kuvyrkov  <maxim.kuvyrkov@linaro.org>
29239         * tree-ssa-loop-prefetch.c (struct mem_ref_group, struct mem_ref):
29240         New "uid" fields to hold pretty-print IDs of group and ref.
29241         Memory references are now identified as <group_id>:<ref_id>
29242         instead of using [random] addresses.
29243         (dump_mem_details): Simplify, no functional change.
29244         (dump_mem_ref): Simplify and make output more concise.
29245         Replace couple of fprintf's throughout code with calls to dump_mem_ref.
29246         (find_or_create_group): Initialize group uid.
29247         (record_ref): Initialize ref uid.  Improve debug output.
29248         (prune_group_by_reuse, should_issue_prefetch_p,)
29249         (should_issue_prefetch_p, schedule_prefetches, issue_prefetch_ref,)
29250         (mark_nontemporal_store, determine_loop_nest_reuse):
29251         Improve debug output.
29253 2017-06-06  Maxim Kuvyrkov  <maxim.kuvyrkov@linaro.org>
29255         * dbgcnt.def (prefetch): New debug counter.
29256         * tree-ssa-loop-prefetch.c (dbgcnt.h): New include.
29257         (schedule_prefetches): Stop issueing prefetches if debug counter
29258         tripped.
29260 2017-06-06  Tom de Vries  <tom@codesourcery.com>
29262         * doc/sourcebuild.texi (Testsuites, C Language Testsuites,
29263         gcc.c-torture/compile): Remove mention of NO_LABEL_VALUES in fixme.
29265 2017-06-06  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
29267         * config/aarch64/atomics.md (atomic_compare_and_swap<mode> expander):
29268         Use aarch64_reg_or_zero predicate for operand 4.
29269         (aarch64_compare_and_swap<mode> define_insn_and_split):
29270         Use aarch64_reg_or_zero predicate for operand 3.  Add 'Z' constraint.
29271         (aarch64_store_exclusive<mode>): Likewise for operand 2.
29273 2017-06-06  Thomas Preud'homme  <thomas.preudhomme@arm.com>
29275         * config/arm/arm.c (arm_compute_save_reg_mask): Rename into ...
29276         (arm_compute_save_core_reg_mask): This.
29277         (thumb1_compute_save_reg_mask): Rename into ...
29278         (thumb1_compute_save_core_reg_mask): This.
29279         (arm_compute_save_reg0_reg12_mask): Adapt comment.
29280         (arm_compute_frame_layout): Likewise.
29282 2017-06-06  Richard Biener  <rguenther@suse.de>
29284         PR tree-optimization/80974
29285         * tree-ssa-sccvn.c (set_ssa_val_to): Do not change but only
29286         keep or clear leaders SSA info.
29288 2017-06-06  Tom de Vries  <tom@codesourcery.com>
29290         * config/nvptx/nvptx.c (split_mode_p): New function.
29291         (nvptx_declare_function_name, nvptx_print_operand): Use split_mode_p.
29293 2017-06-06  Tom de Vries  <tom@codesourcery.com>
29295         * config/nvptx/nvptx.c (nvptx_print_operand): Use maybe_split_mode.
29297 2017-06-06  Jan Hubicka  <hubicka@ucw.cz>
29299         PR bootstrap/80978
29300         * tree-cfg.c (execute_fixup_cfg): Fix condition on when to rescale
29301         profile.
29303 2017-06-06  Jan Hubicka  <hubicka@ucw.cz>
29305         * shrink-wrap.c (handle_simple_exit): Update profile.
29306         (try_shrink_wrapping): Upate profile.
29308 2017-06-06  Jan Hubicka  <hubicka@ucw.cz>
29310         * predict.c (tree_estimate_probability_bb): Add LOCAL_ONLY.
29311         (tree_guess_outgoing_edge_probabilities): New.
29312         * predict.h (tree_guess_outgoing_edge_probabilities): Declare.
29313         * tree-cfg.c (gimple_find_sub_bbs): Propagate profile.
29315 2017-06-06  Jan Hubicka  <hubicka@ucw.cz>
29317         * ipa-split.c (split_function): Initialize return bb profile.
29319 2017-06-06  Jan Hubicka  <hubicka@ucw.cz>
29321         * profile.c (compute_branch_probabilities): Also initialize
29322         EXIT_BLOCK profile.
29324 2017-06-06  Richard Biener  <rguenther@suse.de>
29326         PR tree-optimization/80928
29327         * tree-vect-loop.c (vect_update_vf_for_slp): Amend dumps.
29328         (vect_analyze_loop_operations): Properly guard analysis for
29329         pure SLP case.
29330         (vect_transform_loop): Likewise.
29331         (vect_analyze_loop_2): Also reset SLP type on PHIs.
29332         (vect_model_induction_cost): Do not cost for pure SLP.
29333         (vectorizable_induction): Pass in SLP node, implement SLP vectorization
29334         of induction in inner loop vectorization.
29335         * tree-vect-slp.c (vect_create_new_slp_node): Handle PHIs.
29336         (vect_get_and_check_slp_defs): Handle vect_induction_def.
29337         (vect_build_slp_tree): Likewise.  Handle PHIs as terminating the
29338         recursion.
29339         (vect_analyze_slp_cost_1): Cost induction.
29340         (vect_detect_hybrid_slp_stmts): Handle PHIs.
29341         (vect_get_slp_vect_defs): Likewise.
29342         * tree-vect-stmts.c (vect_analyze_stmt): Handle induction.
29343         (vect_transform_stmt): Handle SLP reductions.
29344         * tree-vectorizer.h (vectorizable_induction): Adjust.
29346 2017-06-05  Michael Meissner  <meissner@linux.vnet.ibm.com>
29348         * config/rs6000/rs6000.c (make_resolver_func): Update
29349         init_lowered_empty_function call.
29351 2017-06-05  Bernd Edlinger  <bernd.edlinger@hotmail.de>
29353         * doc/invoke.texi: Document the -fprofile-abs-path option.
29354         * common.opt (fprofile-abs-path): New option.
29355         * gcov-io.h (gcov_write_filename): Declare.
29356         * gcov-io.c (gcov_write_filename): New function.
29357         * coverage.c (coverage_begin_function): Use gcov_write_filename.
29358         * profile.c (output_location): Likewise.
29360 2017-06-05  Jan Hubicka  <hubicka@ucw.cz>
29362         * shring-wrap.c: Revert accidental commit.
29364 2017-06-05  Volker Reichelt  <v.reichelt@netcologne.de>
29366         * doc/invoke.texi (-Wduplicated-branches): Add to warning list.
29368 2017-06-05  Jan Hubicka  <hubicka@ucw.cz>
29370         * cfgexpand.c (expand_gimple_tailcall): Initialize profile of
29371         new edge.
29372         * ipa-inline.c (want_inline_self_recursive_call_p): Watch for missing
29373         profile in callgraph edge.
29374         * profile-count.h (apply_probability): If THIS is 0, then result is 0
29375         (apply_scale): Likewise.
29376         * tree-inline.c (copy_bb, copy_edges_for_bb, copy_cfg_body):
29377         Also scale profile when inlining function with zero profile.
29378         (initialize_cfun): Update exit block profile even when it is zero.
29379         * tree-ssa-threadupdate.c (clear_counts_path): Handle correctly case
29380         when profile is read.
29382 2017-06-05  Michael Meissner  <meissner@linux.vnet.ibm.com>
29384         * config/rs6000/rs6000.c (toplevel): Include attribs.h.
29385         (CLONE_*): New constants to define the processors we can generate
29386         code for with the target_clone attribute.
29387         (rs6000_clone_map): New array to identify which clone processors
29388         the current program is running on.
29389         (TARGET_COMPARE_VERSION_PRIORITY): Define to enable the
29390         target_clone attribute.
29391         (TARGET_GENERATE_VERSION_DISPATCHER_BODY): Likewise.
29392         (TARGET_GET_FUNCTION_VERSIONS_DISPATCHER): Likewise.
29393         (TARGET_OPTION_FUNCTION_VERSIONS): Likewise.
29394         (cpu_expand_builtin): Add support for target_clone attribute.
29395         (rs6000_valid_attribute_p): Allow "default" attribute.
29396         (get_decl_name): New debug function to simplify printing the
29397         current function name in debugging statements.
29398         (rs6000_clone_priority): New functions to support the target_clone
29399         attribute, and be able to generate code to switch between ISA 2.05
29400         through ISA 3.0 (power6 through power9).
29401         (rs6000_compare_version_priority): Likewise.
29402         (rs6000_get_function_versions_dispatcher): Likewise.
29403         (make_resolver_func): Likewise.
29404         (add_condition_to_bb): Likewise.
29405         (dispatch_function_versions): Likewise.
29406         (rs6000_generate_version_dispatcher_body): Likewise.
29407         (rs6000_can_inline_p): Call get_decl_name for debugging usage.
29408         (fusion_gpr_load_p): Fix a spacing issue.
29409         * doc/extend.texi (Common Function Attributes): Document that the
29410         PowerPC supports the target_clone attribute.
29412 2017-06-05  Thomas Preud'homme  <thomas.preudhomme@arm.com>
29414         * config/arm/arm.h: explain F symbol found in description of ARM
29415         register allocation in its legend.
29417 2017-06-05  Jan Hubicka  <hubicka@ucw.cz>
29419         * config/mips/frame-header-opt.c: Include profile-count.h.
29420         * config/riscv/riscv.c: Include profile-count.h
29422 2017-06-05  Jan Hubicka  <hubicka@ucw.cz>
29424         * tree-ssa-loop-im.c (execute_sm_if_changed): Add FLAG_BBS parameter;
29425         update profile.
29426         (sm_set_flag_if_changed): Add bbs field.
29427         (execute_sm_if_changed_flag_set): Pass BBS.
29428         (execute_sm): Update.
29430 2017-06-05  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
29432         * config/aarch64/aarch64-simd.md (aarch64_store_lane0<mode>):
29433         New pattern.
29435 2017-06-05  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
29437         * config/aarch64/aarch64.md (sub<mode>3_compare1_imm): New define_insn.
29438         (peephole2): New peephole2 to emit the above.
29439         * config/aarch64/predicates.md (aarch64_sub_immediate): New predicate.
29441 2017-06-05  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
29443         * config/aarch64/aarch64.c (define_peephole2 above
29444         *sub_<shift>_<mode>): New peephole.
29446 2017-05-23  Jan Hubicka  <hubicka@ucw.cz>
29448         * config/i386/i386.c (make_resolver_func): Update.
29449         * Makefile.in: Add profile-count.h and profile-count.o
29450         * auto-profile.c (afdo_indirect_call): Update to new API.
29451         (afdo_set_bb_count): Update.
29452         (afdo_propagate_edge): Update.
29453         (afdo_propagate_circuit): Update.
29454         (afdo_calculate_branch_prob): Update.
29455         (afdo_annotate_cfg): Update.
29456         * basic-block.h: Include profile-count.h
29457         (struct edge_def): Turn count to profile_count.
29458         (struct basic_block_def): Likewie.
29459         (REG_BR_PROB_BASE): Move to profile-count.h
29460         (RDIV): Move to profile-count.h
29461         * bb-reorder.c (max_entry_count): Turn to profile_count.
29462         (find_traces): Update.
29463         (rotate_loop):Update.
29464         (connect_traces):Update.
29465         (sanitize_hot_paths):Update.
29466         * bt-load.c (migrate_btr_defs): Update.
29467         * cfg.c (RDIV): Remove.
29468         (init_flow): Use alloc_block.
29469         (alloc_block): Uninitialize count.
29470         (unchecked_make_edge): Uninitialize count.
29471         (check_bb_profile): Update.
29472         (dump_edge_info): Update.
29473         (dump_bb_info): Update.
29474         (update_bb_profile_for_threading): Update.
29475         (scale_bbs_frequencies_int): Update.
29476         (scale_bbs_frequencies_gcov_type): Update.
29477         (scale_bbs_frequencies_profile_count): New.
29478         * cfg.h (update_bb_profile_for_threading): Update.
29479         (scale_bbs_frequencies_profile_count): Declare.
29480         * cfgbuild.c (compute_outgoing_frequencies): Update.
29481         (find_many_sub_basic_blocks): Update.
29482         * cfgcleanup.c (try_forward_edges): Update.
29483         (try_crossjump_to_edge): Update.
29484         * cfgexpand.c (expand_gimple_tailcall): Update.
29485         (construct_exit_block): Update.
29486         * cfghooks.c (verify_flow_info): Update.
29487         (dump_bb_for_graph): Update.
29488         (split_edge): Update.
29489         (make_forwarder_block): Update.
29490         (duplicate_block): Update.
29491         (account_profile_record): Update.
29492         * cfgloop.c (find_subloop_latch_edge_by_profile): Update.
29493         (get_estimated_loop_iterations): Update.
29494         * cfgloopanal.c (expected_loop_iterations_unbounded): Update.
29495         (single_likely_exit): Update.
29496         * cfgloopmanip.c (scale_loop_profile): Update.
29497         (loopify): Update.
29498         (set_zero_probability): Update.
29499         (lv_adjust_loop_entry_edge): Update.
29500         * cfgrtl.c (force_nonfallthru_and_redirect): Update.
29501         (purge_dead_edges): Update.
29502         (rtl_account_profile_record): Update.
29503         * cgraph.c (cgraph_node::create): Uninitialize count.
29504         (symbol_table::create_edge): Uninitialize count.
29505         (cgraph_update_edges_for_call_stmt_node): Update.
29506         (cgraph_edge::dump_edge_flags): Update.
29507         (cgraph_node::dump): Update.
29508         (cgraph_edge::maybe_hot_p): Update.
29509         * cgraph.h: Include profile-count.h
29510         (create_clone), create_edge, create_indirect_edge): Update.
29511         (cgraph_node): Turn count to profile_count.
29512         (cgraph_edge0: Likewise.
29513         (make_speculative, clone): Update.
29514         (create_edge): Update.
29515         (init_lowered_empty_function): Update.
29516         * cgraphclones.c (cgraph_edge::clone): Update.
29517         (duplicate_thunk_for_node): Update.
29518         (cgraph_node::create_clone): Update.
29519         * cgraphunit.c (cgraph_node::analyze): Update.
29520         (cgraph_node::expand_thunk): Update.
29521         * final.c (dump_basic_block_info): Update.
29522         * gimple-streamer-in.c (input_bb): Update.
29523         * gimple-streamer-out.c (output_bb): Update.
29524         * graphite.c (print_global_statistics): Update.
29525         (print_graphite_scop_statistics): Update.
29526         * hsa-brig.c: Include basic-block.h.
29527         * hsa-dump.c: Include basic-block.h.
29528         * hsa-gen.c (T sum_slice): Update.
29529         (convert_switch_statements):Update.
29530         * hsa-regalloc.c: Include basic-block.h.
29531         * ipa-chkp.c (chkp_produce_thunks): Update.
29532         * ipa-cp.c (struct caller_statistics): Update.
29533         (init_caller_stats): Update.
29534         (gather_caller_stats): Update.
29535         (ipcp_cloning_candidate_p): Update.
29536         (good_cloning_opportunity_p): Update.
29537         (get_info_about_necessary_edges): Update.
29538         (dump_profile_updates): Update.
29539         (update_profiling_info): Update.
29540         (update_specialized_profile): Update.
29541         (perhaps_add_new_callers): Update.
29542         (decide_about_value): Update.
29543         (ipa_cp_c_finalize): Update.
29544         * ipa-devirt.c (struct odr_type_warn_count): Update.
29545         (struct decl_warn_count): Update.
29546         (struct final_warning_record): Update.
29547         (possible_polymorphic_call_targets): Update.
29548         (ipa_devirt): Update.
29549         * ipa-fnsummary.c (redirect_to_unreachable): Update.
29550         * ipa-icf.c (sem_function::merge): Update.
29551         * ipa-inline-analysis.c (do_estimate_edge_time): Update.
29552         * ipa-inline.c (compute_uninlined_call_time): Update.
29553         (compute_inlined_call_time): Update.
29554         (want_inline_small_function_p): Update.
29555         (want_inline_self_recursive_call_p): Update.
29556         (edge_badness): Update.
29557         (lookup_recursive_calls): Update.
29558         (recursive_inlining): Update.
29559         (inline_small_functions): Update.
29560         (dump_overall_stats): Update.
29561         (dump_inline_stats): Update.
29562         * ipa-profile.c (ipa_profile_generate_summary): Update.
29563         (ipa_propagate_frequency): Update.
29564         (ipa_profile): Update.
29565         * ipa-prop.c (ipa_make_edge_direct_to_target): Update.
29566         * ipa-utils.c (ipa_merge_profiles): Update.
29567         * loop-doloop.c (doloop_modify): Update.
29568         * loop-unroll.c (report_unroll): Update.
29569         (unroll_loop_runtime_iterations): Update.
29570         * lto-cgraph.c (lto_output_edge): Update.
29571         (lto_output_node): Update.
29572         (input_node): Update.
29573         (input_edge): Update.
29574         (merge_profile_summaries): Update.
29575         * lto-streamer-in.c (input_cfg): Update.
29576         * lto-streamer-out.c (output_cfg): Update.
29577         * mcf.c (create_fixup_graph): Update.
29578         (adjust_cfg_counts): Update.
29579         (sum_edge_counts): Update.
29580         * modulo-sched.c (sms_schedule): Update.
29581         * postreload-gcse.c (eliminate_partially_redundant_load): Update.
29582         * predict.c (maybe_hot_count_p): Update.
29583         (probably_never_executed): Update.
29584         (dump_prediction): Update.
29585         (combine_predictions_for_bb): Update.
29586         (propagate_freq): Update.
29587         (handle_missing_profiles): Update.
29588         (counts_to_freqs): Update.
29589         (rebuild_frequencies): Update.
29590         (force_edge_cold): Update.
29591         * predict.h: Include profile-count.h
29592         (maybe_hot_count_p, counts_to_freqs): UPdate.
29593         * print-rtl-function.c: Do not include cfg.h
29594         * print-rtl.c: Include basic-block.h
29595         * profile-count.c: New file.
29596         * profile-count.h: New file.
29597         * profile.c (is_edge_inconsistent): Update.
29598         (correct_negative_edge_counts): Update.
29599         (is_inconsistent): Update.
29600         (set_bb_counts): Update.
29601         (read_profile_edge_counts): Update.
29602         (compute_frequency_overlap): Update.
29603         (compute_branch_probabilities): Update; Initialize and deinitialize
29604         gcov_count tables.
29605         (branch_prob): Update.
29606         * profile.h (bb_gcov_counts, edge_gcov_counts): New.
29607         (edge_gcov_count): New.
29608         (bb_gcov_count): New.
29609         * shrink-wrap.c (try_shrink_wrapping): Update.
29610         * tracer.c (better_p): Update.
29611         * trans-mem.c (expand_transaction): Update.
29612         (ipa_tm_insert_irr_call): Update.
29613         (ipa_tm_insert_gettmclone_call): Update.
29614         * tree-call-cdce.c: Update.
29615         * tree-cfg.c (gimple_duplicate_sese_region): Update.
29616         (gimple_duplicate_sese_tail): Update.
29617         (gimple_account_profile_record): Update.
29618         (execute_fixup_cfg): Update.
29619         * tree-inline.c (copy_bb): Update.
29620         (copy_edges_for_bb): Update.
29621         (initialize_cfun): Update.
29622         (freqs_to_counts): Update.
29623         (copy_cfg_body): Update.
29624         (expand_call_inline): Update.
29625         * tree-ssa-ifcombine.c (update_profile_after_ifcombine): Update.
29626         * tree-ssa-loop-ivcanon.c (unloop_loops): Update.
29627         (try_unroll_loop_completely): Update.
29628         (try_peel_loop): Update.
29629         * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Update.
29630         * tree-ssa-loop-niter.c (estimate_numbers_of_iterations_loop): Update.
29631         * tree-ssa-loop-split.c (connect_loops): Update.
29632         * tree-ssa-loop-unswitch.c (hoist_guard): Update.
29633         * tree-ssa-reassoc.c (branch_fixup): Update.
29634         * tree-ssa-tail-merge.c (replace_block_by): Update.
29635         * tree-ssa-threadupdate.c (create_block_for_threading): Update.
29636         (compute_path_counts): Update.
29637         (update_profile): Update.
29638         (recompute_probabilities): Update.
29639         (update_joiner_offpath_counts): Update.
29640         (estimated_freqs_path): Update.
29641         (freqs_to_counts_path): Update.
29642         (clear_counts_path): Update.
29643         (ssa_fix_duplicate_block_edges): Update.
29644         (duplicate_thread_path): Update.
29645         * tree-switch-conversion.c (case_bit_test_cmp): Update.
29646         (struct switch_conv_info): Update.
29647         * tree-tailcall.c (decrease_profile): Update.
29648         * tree-vect-loop-manip.c (slpeel_add_loop_guard): Update.
29649         * tree-vect-loop.c (scale_profile_for_vect_loop): Update.
29650         * value-prof.c (check_counter): Update.
29651         (gimple_divmod_fixed_value): Update.
29652         (gimple_mod_pow2): Update.
29653         (gimple_mod_subtract): Update.
29654         (gimple_ic_transform): Update.
29655         (gimple_stringop_fixed_value): Update.
29656         * value-prof.h (gimple_ic): Update.
29658 2017-06-02  Carl Love  <cel@us.ibm.com>
29660         * config/rs6000/rs6000-c: Add support for built-in functions
29661         vector double vec_doublee (vector signed int);
29662         vector double vec_doublee (vector unsigned int);
29663         vector double vec_doublee (vector float);
29664         vector double vec_doubleh (vector signed int);
29665         vector double vec_doubleh (vector unsigned int);
29666         vector double vec_doubleh (vector float);
29667         vector double vec_doublel (vector signed int);
29668         vector double vec_doublel (vector unsigned int);
29669         vector double vec_doublel (vector float);
29670         vector double vec_doubleo (vector signed int);
29671         vector double vec_doubleo (vector unsigned int);
29672         vector double vec_doubleo (vector float);.
29673         * config/rs6000/rs6000-builtin.def: Add definitions for DOUBLEE,
29674         DOUBLEO, DOUBLEH, DOUBLEL, UNS_DOUBLEO, UNS_DOUBLEE, UNS_DOUBLEH,
29675         UNS_DOUBLEL.
29676         * config/rs6000/altivec.md: Add code generator for doublee<mode>2,
29677         unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2, doubleh<mode>2,
29678         unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2, add mode attribute
29679         VS_sxwsp.
29680         * config/rs6000/altivec.h: Add define for vec_doublee, vec_doubleo,
29681         vec_doublel, vec_doubleh.
29682         * doc/extend.texi: Update the built-in documentation file for the
29683         new built-in functions.
29685 2017-06-02  David Malcolm  <dmalcolm@redhat.com>
29687         PR jit/80954
29688         * ipa-inline-analysis.c (free_growth_caches): Set
29689         edge_removal_hook_holder to NULL after removing it.
29691 2017-06-02  Sudakshina Das  <sudi.das@arm.com>
29693         * config/aarch64/aarch64.c (aarch64_select_cc_mode): Return CC_SWP for
29694         comparision with zero.
29696 2017-06-02  Will Schmidt  <will_schmidt@vnet.ibm.com>
29697         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add handling
29698         for early expansion of vec_min and vec_max builtins.
29699         (builtin_function_type): Add min/max unsigned variants to those
29700         identified as having unsigned arguments.
29702 2017-06-02  Olivier Hainque  <hainque@adacore.com>
29704         * config/vx-common.h (DWARF_UNWIND_INFO): Switch #define to 1.
29706 2017-06-02  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
29708         * config/aarch64/aarch64-simd.md (*aarch64_simd_vec_copy_lane<mode>):
29709         Use VALL_F16 iterator rather than VALL.
29711 2017-06-02  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
29713         * config/aarch64/aarch64.c (aarch64_split_compare_and_swap):
29714         Emit CBNZ inside loop when doing a strong exchange and comparing
29715         against zero.  Generate the CC flags after the loop.
29717 2017-06-02  David Edelsohn  <dje.gcc@gmail.com>
29719         * dwarf2out.c (DWARF_INITIAL_LENGTH_SIZE_STR): New.
29720         (dl_section_ref): New.
29721         (dwarf2out_finish): Copy debug_line_section_label to dl_section_ref.
29722         On AIX, append an expression to subtract the size of the
29723         section length to dl_section_ref.
29725 2017-06-02  Will Schmidt  <will_schmidt@vnet.ibm.com>
29727         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add handling
29728         for early expansion of vector absolute builtins.
29730 2017-06-02  Richard Biener  <rguenther@suse.de>
29732         * tree-vect-slp.c (vect_detect_hybrid_slp_2): Match up
29733         what we consider a relevant use stmt with vect_detect_hybrid_slp_stmts.
29735 2017-06-02  Richard Biener  <rguenther@suse.de>
29737         PR tree-optimization/80948
29738         * tree-tailcall.c (find_tail_calls): Track stmts to move in
29739         stmt order as well.
29741 2017-06-02  Richard Biener  <rguenther@suse.de>
29743         * tree-vect-loop.c (vect_analyze_loop_operations): Not relevant
29744         PHIs are ok.
29745         * tree-vect-stmts.c (process_use): Do not mark backedge defs
29746         for inductions as relevant.
29748 2017-06-02  Richard Biener  <rguenther@suse.de>
29750         * tree-vect-loop.c (get_initial_def_for_induction): Inline into ...
29751         (vectorizable_induction): ... this.  Remove dead code.
29753 2017-06-02  Eric Botcazou  <ebotcazou@adacore.com>
29755         * builtins. (expand_builtin_alloca): Remove second parameter and
29756         infer its value from the first parameter instead.
29757         (expand_builtin) <BUILT_IN_ALLOCA>: Adjust call to above.
29759 2017-06-02  Jakub Jelinek  <jakub@redhat.com>
29761         PR rtl-optimization/80903
29762         * loop-doloop.c (add_test): Unshare sequence.
29764 2017-06-02  Bernd Edlinger  <bernd.edlinger@hotmail.de>
29766         * doc/invoke.texi: Document the -Wsizeof-pointer-div warning.
29768 2017-06-01  Bernd Edlinger  <bernd.edlinger@hotmail.de>
29770         * config/i386/i386.c (x86_64_ms_sysv_extra_clobbered_registers): Make
29771         static.
29772         (xlogue_layout::get_stack_space_used, xlogue_layout::s_instances,
29773         xlogue_layout::get_instance, logue_layout::xlogue_layout,
29774         sp_valid_at, fp_valid_at, choose_basereg): Formatting.
29775         (xlogue_layout::get_stub_rtx): Make static.
29776         (xlogue_layout::get_stub_name): Avoid const-cast, make static.
29777         (xlogue_layout::compute_stub_managed_regs): Rename to...
29778         (xlogue_layout::count_stub_managed_regs): ...this.
29779         (xlogue_layout::is_stub_managed_reg): New function.
29780         (xlogue_layout::m_stub_names): Rename to...
29781         (xlogue_layout::s_stub_names): ...this, make static.
29782         (xlogue_layout::STUB_INDEX_OFFSET, xlogue_layout::MIN_REGS,
29783         xlogue_layout::MAX_REGS, xlogue_layout::MAX_EXTRA_REGS,
29784         xlogue_layout::VARIANT_COUNT, xlogue_layout::STUB_NAME_MAX_LEN,
29785         xlogue_layout::s_stub_names): Instantiate statics.
29786         (stub_managed_regs): Remove.
29787         (ix86_save_reg): Use xlogue_layout::compute_stub_managed_regs.
29788         (disable_call_ms2sysv_xlogues): Rename to...
29789         (warn_once_call_ms2sysv_xlogues): ...this, and warn only once.
29790         (ix86_initial_elimination_offset, ix86_expand_call): Fix call_ms2sysv
29791         warning logic.
29792         (ix86_static_chain): Make sure that ix86_static_chain_on_stack can't
29793         change after reload_completed.
29794         (ix86_can_use_return_insn_p): Use the ix86_frame data structure
29795         directly.
29796         (ix86_expand_prologue): Likewise.
29797         (ix86_expand_epilogue): Likewise.
29798         (ix86_expand_split_stack_prologue): Likewise.
29799         (ix86_compute_frame_layout): Remove frame parameter ...
29800         (TARGET_COMPUTE_FRAME_LAYOUT): ... and export it as a target hook.
29801         (ix86_finalize_stack_realign_flags): Call ix86_compute_frame_layout
29802         only if necessary.
29803         (ix86_init_machine_status): Don't set use_fast_prologue_epilogue_nregs.
29804         (ix86_frame): Move from here ...
29805         * config/i386/i386.h (ix86_frame): ... to here.
29806         (machine_function): Remove use_fast_prologue_epilogue_nregs, cache the
29807         complete ix86_frame data structure instead.  Remove some_ld_name.
29809 2017-06-01  Pierre-Marie de Rodat  <derodat@adacore.com>
29811         * dwarf2out.c (dwarf2out_late_global_decl): Add locations for
29812         symbols that hold a DECL_VALUE_EXPR.
29814 2017-06-01  Martin Jambor  <mjambor@suse.cz>
29816         PR tree-optimization/80898
29817         * tree-sra.c (process_subtree_disqualification): Removed.
29818         (disqualify_candidate): Do not acll
29819         process_subtree_disqualification.
29820         (subtree_mark_written_and_enqueue): New function.
29821         (propagate_all_subaccesses): Set grp_write of LHS subtree if the
29822         RHS has been disqualified and re-queue LHS if necessary.  Apart
29823         from that, ignore disqualified RHS.
29825 2017-06-01  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
29827         * config/s390/s390.c (s390_emit_epilogue): Disable early return
29828         address fetch for z10 or later.
29830 2017-06-01  Claudiu Zissulescu  <claziss@synopsys.com>
29832         * config/arc/arc.md (tst_movb): Add guard when splitting.
29834 2017-06-01  Claudiu Zissulescu  <claziss@synopsys.com>
29836         * config/arc/arc.c (arc_can_eliminate): Test against
29837         arc_frame_pointer_needed.
29839 2017-06-01  Claudiu Zissulescu  <claziss@synopsys.com>
29841         * config/arc/arc.c (arc_expand_prologue): Emit a special barrier
29842         to prevent store reordering.
29843         * config/arc/arc.md (UNSPEC_ARC_STKTIE): Define.
29844         (type): Add block type.
29845         (stack_tie): Define special instruction to be used in
29846         expand_prologue.
29848 2017-06-01  Claudiu Zissulescu  <claziss@synopsys.com>
29850         * config/arc/arc.md (commutative_binary_comparison): Remove 'I'
29851         constraint. It is not valid for the pattern.
29852         (noncommutative_binary_comparison): Likewise.
29854 2017-06-01  Claudiu Zissulescu  <claziss@synopsys.com>
29856         * config/arc/simdext.md (movv2hi_insn): Change predicate to avoid
29857         scaled addresses.
29859 2017-06-01  Claudiu Zissulescu  <claziss@synopsys.com>
29861         * config/arc/arc.c (arc_conditional_register_usage): Allow r30 to
29862         be used by the reg-alloc.
29864 2017-06-01  Claudiu Zissulescu  <claziss@synopsys.com>
29866         * config/arc/arc.md (mulsi3): Avoid use of hard registers before
29867         reg-alloc when having mul64 or mul32x16 instructions.
29868         (mulsidi3): Likewise.
29869         (umulsidi3): Likewise.
29870         (mulsi32x16): New pattern.
29871         (mulsi64): Likewise.
29872         (mulsidi64): Likewise.
29873         (umulsidi64): Likewise.
29874         (MUL32x16_REG): Define.
29875         (mul64_600): Use MUL32x16_REG.
29876         (mac64_600): Likewise.
29877         (umul64_600): Likewise.
29878         (umac64_600): Likewise.
29880 2017-06-01  Claudiu Zissulescu  <claziss@synopsys.com>
29882         * config/arc/arc.md (mulsi3_700): Make it commutative.
29884 2017-06-01  Jose E. Marchesi  <jose.marchesi@oracle.com>
29886         * config/sparc/sparc.md (*zero_extendsidi2_insn_sp64): Set insn
29887         type for movstouw.
29888         (*sign_extendsidi2_insn): Likewise for movstosw.
29890 2017-06-01  Pierre-Marie de Rodat  <derodat@adacore.com>
29892         * dwarf2out.c (get_discr_value): Call the get_debug_type hook on
29893         the type of the input discriminant value.  Convert the
29894         discriminant value of signedness vary.
29896 2017-06-01  Volker Reichelt  <v.reichelt@netcologne.de>
29898         * doc/invoke.texi (-Wcatch-value): Document new shortcut.
29899         Add to -Wall section.
29901 2017-06-01  Richard Biener  <rguenther@suse.de>
29903         PR middle-end/66313
29904         * fold-const.c (fold_plusminus_mult_expr): If the factored
29905         factor may be zero use a wrapping type for the inner operation.
29906         * tree-tailcall.c (independent_of_stmt_p): Pass in to_move bitmap
29907         and handle moved defs.
29908         (process_assignment): Properly guard the unary op case.  Return a
29909         tri-state indicating that moving the stmt before the call may allow
29910         to continue.  Pass through to_move.
29911         (find_tail_calls): Handle moving unrelated defs before
29912         the call.
29914 2017-05-31  Segher Boessenkool  <segher@kernel.crashing.org>
29916         PR target/80618
29917         * config/rs6000/vector.md (*vector_uneq<mode>): Write the nor in the
29918         splitter result in the canonical way.
29920 2017-05-31  Uros Bizjak  <ubizjak@gmail.com>
29922         * config/i386/i386.md (*zero_extendsidi2): Enable alternative (?r, *Yj)
29923         also for 32bit target.  Update insn attributes.
29924         (zero-extendsidi2 splitter): Allow all registers for operand 1.
29926 2017-05-31  Sebastian Peryt  <sebastian.peryt@intel.com>
29928         * config/i386/avx512fintrin.h (_mm_mask_max_sd)
29929         (_mm_maskz_max_sd, _mm_mask_max_ss, _mm_maskz_max_ss)
29930         (_mm_mask_min_sd, _mm_maskz_min_sd, _mm_mask_min_ss)
29931         (_mm_maskz_min_ss): New intrinsics.
29933 2017-05-31  Martin Liska  <mliska@suse.cz>
29935         * tree-vect-loop.c (vect_create_epilog_for_reduction):
29936         Change comment style to one we normally use.
29937         (vectorizable_reduction): Likewise.
29938         (vectorizable_induction): Likewise.
29939         * tree-vect-stmts.c (vectorizable_mask_load_store): Likewise.
29940         (vectorizable_call): Likewise.
29941         (vectorizable_simd_clone_call): Likewise.
29942         (vectorizable_conversion): Likewise.
29943         (vectorizable_assignment): Likewise.
29944         (vectorizable_shift): Likewise.
29945         (vectorizable_operation): Likewise.
29946         (vectorizable_store): Likewise.
29947         (vectorizable_load): Likewise.
29948         * tree-vectorizer.h: Likewise.
29950 2017-05-31  Alexander Monakov  <amonakov@ispras.ru>
29952         * passes.c (emergency_dump_function): New.
29953         * tree-pass.h (emergency_dump_function): Declare.
29954         * plugin.c (plugins_internal_error_function): Remove.
29955         * plugin.h (plugins_internal_error_function): Remove declaration.
29956         * toplev.c (internal_error_function): New static function.  Use it...
29957         (general_init): ...here.
29959 2017-05-31  Graham Markall  <graham.markall@embecosm.com>
29961         * config/arc/arc.c (arc_print_operand): Handle constant operands.
29962         (arc_rtx_costs): Add costs for new patterns.
29963         * config/arc/arc.md: Additional *add_n and *sub_n patterns.
29964         * config/arc/predicates.md: Add _1_2_3_operand predicate.
29966 2017-05-31  Richard Sandiford  <richard.sandiford@linaro.org>
29968         * tree-ssa-strlen.c (get_next_strinfo): New function.
29969         (get_stridx_plus_constant): Use it.
29970         (zero_length_string): Likewise.
29971         (adjust_related_strinfos): Likewise.
29972         (adjust_last_stmt): Likewise.
29974 2017-05-31  Richard Biener  <rguenther@suse.de>
29976         PR target/80880
29977         * config/i386/i386.c (ix86_expand_builtin): Remove assert
29978         for arg being an SSA name when expanding IX86_BUILTIN_BNDRET.
29980 2017-05-31  Richard Sandiford  <richard.sandiford@linaro.org>
29982         * tree-vect-data-refs.c (vect_find_same_alignment_drs): Remove
29983         loop_vinfo argument and use of dependence distance vectors.
29984         Check instead whether the two references differ only in their
29985         initial value and assume that they have the same alignment if the
29986         difference is a multiple of the vector alignment.
29987         (vect_analyze_data_refs_alignment): Update call accordingly.
29989 2017-05-31  Martin Liska  <mliska@suse.cz>
29991         PR target/79155
29992         * config/i386/cpuid.h: Fix typo in a comment in cpuid.h.
29994 2017-05-31  Bin Cheng  <bin.cheng@arm.com>
29996         * tree-vect-loop-manip.c (create_intersect_range_checks_index)
29997         (create_intersect_range_checks): Move from ...
29998         * tree-data-ref.c (create_intersect_range_checks_index)
29999         (create_intersect_range_checks): ... to here.
30000         (create_runtime_alias_checks): New function factored from ...
30001         * tree-vect-loop-manip.c (vect_create_cond_for_alias_checks): ...
30002         here.  Call above function.
30003         * tree-data-ref.h (create_runtime_alias_checks): New function.
30005 2017-05-31  Bin Cheng  <bin.cheng@arm.com>
30007         * tree-data-ref.c (prune_runtime_alias_test_list): Relax minimal
30008         segment length for dr_b and compute it in wide_int.
30010 2017-05-31  Richard Biener  <rguenther@suse.de>
30012         PR tree-optimization/80906
30013         * graphite-isl-ast-to-gimple.c (copy_loop_close_phi_nodes): Get
30014         and pass through iv_map.
30015         (copy_bb_and_scalar_dependences): Adjust.
30016         (translate_pending_phi_nodes): Likewise.
30017         (copy_loop_close_phi_args): Handle code-generating IVs instead
30018         of ICEing.
30020 2017-05-30  David Malcolm  <dmalcolm@redhat.com>
30022         * diagnostic-color.c (color_dict): Add "type-diff".
30023         (parse_gcc_colors): Update comment.
30024         * doc/invoke.texi (Diagnostic Message Formatting Options): Add
30025         -fdiagnostics-show-template-tree and -fno-elide-type.
30026         (GCC_COLORS): Add type-diff to example.
30027         (type-diff=): New.
30028         (-fdiagnostics-show-template-tree): New.
30029         (-fno-elide-type): New.
30030         * pretty-print.c (pp_format): Pass quote and formatters[argno] to
30031         the pp_format_decoder callback.  Call any m_format_postprocessor's
30032         "handle" method.
30033         (pretty_printer::pretty_printer): Initialize
30034         m_format_postprocessor.
30035         (pretty_printer::~pretty_printer): Delete any
30036         m_format_postprocessor.
30037         * pretty-print.h (printer_fn): Add bool and const char ** parameters.
30038         (class format_postprocessor): New class.
30039         (struct pretty_printer::format_decoder): Document the new parameters.
30040         (struct pretty_printer::m_format_postprocessor): New field.
30041         * tree-diagnostic.c (default_tree_printer): Update for new
30042         bool and const char ** params.
30043         * tree-diagnostic.h (default_tree_printer): Likewise.
30045 2017-05-30  Segher Boessenkool  <segher@kernel.crashing.org>
30047         * config/rs6000/predicates.md (cc_reg_not_micro_cr0_operand): Delete.
30048         (lwa_operand): Delete rs6000_gen_cell_microcode test.
30049         * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
30050         rs6000_gen_cell_microcode code.
30051         (rs6000_final_prescan_insn): Delete.
30052         (rs6000_opt_vars): Delete the "gen-cell-microcode" and
30053         "warn-cell-microcode" entries.
30054         * config/rs6000/rs6000.h (FINAL_PRESCAN_INSN): Delete.
30055         * config/rs6000/rs6000.md: Delete rs6000_gen_cell_microcode tests
30056         throughout.  Change cc_reg_not_micro_cr0_operand to
30057         cc_reg_not_cr0_operand throughout.
30058         (*extendhi<mode>2_noload): Delete.
30059         * config/rs6000/rs6000.opt (mgen-cell-microcode): Replace by stub.
30060         (mwarn-cell-microcode): Delete.
30061         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete
30062         -mgen-cell-microcode and -mwarn-cell-microcode.
30064 2017-05-30  Uros Bizjak  <ubizjak@gmail.com>
30066         PR target/80833
30067         * config/i386/constraints.md (Yd): New constraint.
30068         (Ye): Ditto.
30069         * config/i386/i386.md (*movti_internal): Add (?r, Ye)
30070         and (?Yd, r) alternatives.  Update insn attributes.
30071         * config/i386/i386.md (*movti_internal): Add (?r, *Ye)
30072         and (?*Yd, r) alternatives.  Update insn attributes.
30073         (double-mode inter-unit splitters): Add new GR<->XMM splitters.
30075 2017-05-30  Pierre-Marie de Rodat  <derodat@adacore.com>
30077         * gimplify.c (gimplify_modify_expr): Don't create a
30078         DECL_DEBUG_EXPR link if *FROM_P does not belong to the current
30079         function.
30081 2017-05-30  Wilco Dijkstra  <wdijkstr@arm.com>
30083         * config/arm/arm-builtins.c (arm_expand_builtin): Remove const.
30085 2017-05-30  Richard Biener  <rguenther@suse.de>
30087         * tree-vectorizer.h (struct _stmt_vec_info): Add reduc_type
30088         and reduc_def fields.
30089         (STMT_VINFO_REDUC_TYPE): New define.
30090         (STMT_VINFO_REDUC_DEF): Likewise.
30091         (vect_force_simple_reduction): Adjust prototype.
30092         * tree-vect-loop.c (vect_analyze_scalar_cycles_1): Adjust.
30093         (vect_is_simple_reduction): Remove check_reduction argument.
30094         (vect_force_simple_reduction): Adjust and set
30095         STMT_VINFO_REDUC_TYPE and STMT_VINFO_REDUC_DEF.
30096         (vectorizable_reduction): Do not re-do reduction analysis
30097         but use STMT_VINFO_REDUC_TYPE and STMT_VINFO_REDUC_DEF.
30098         * tree-parloops.c (gather_scalar_reductions): Adjust.
30100 2017-05-30  Richard Biener  <rguenther@suse.de>
30102         PR middle-end/80901
30103         * cfgexpand.c (expand_gimple_cond): Match up loop fixup with
30104         split_edge code.
30106 2017-05-24  Robin Dapp  <rdapp@linux.vnet.ibm.com>
30108         * tree-vect-data-refs.c (vect_get_peeling_costs_all_drs):
30109         Introduce unknown_misalignment parameter and remove vf.
30110         (vect_peeling_hash_get_lowest_cost):
30111         Pass unknown_misalignment parameter.
30112         (vect_enhance_data_refs_alignment):
30113         Fix unsupportable data ref treatment.
30115 2017-05-30  Robin Dapp  <rdapp@linux.vnet.ibm.com>
30117         * tree-vect-data-refs.c (vect_get_data_access_cost):
30118         Workaround for SLP handling.
30119         (vect_enhance_data_refs_alignment):
30120         Compute costs for doing no peeling at all, compare to the best
30121         peeling costs so far and avoid peeling if cheaper.
30123 2017-05-30  Robin Dapp  <rdapp@linux.vnet.ibm.com>
30125         * tree-vect-data-refs.c (vect_peeling_hash_choose_best_peeling):
30126         Return peeling info and set costs to zero for unlimited cost
30127         model.
30128         (vect_enhance_data_refs_alignment): Also inspect all datarefs
30129         with unknown misalignment. Compute and costs for unknown
30130         misalignment, compare them to the costs for known misalignment
30131         and choose the cheapest for peeling.
30133 2017-05-30  Robin Dapp  <rdapp@linux.vnet.ibm.com>
30135         * tree-vect-data-refs.c (vect_update_misalignment_for_peel): Rename.
30136         (vect_get_peeling_costs_all_drs): Create function.
30137         (vect_peeling_hash_get_lowest_cost):
30138         Use vect_get_peeling_costs_all_drs.
30139         (vect_peeling_supportable): Create function.
30140         (vect_enhance_data_refs_alignment): Use vect_peeling_supportable.
30142 2017-05-30  Robin Dapp  <rdapp@linux.vnet.ibm.com>
30144         * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Create
30145         DR_HAS_NEGATIVE_STEP.
30146         (vect_update_misalignment_for_peel): Define DR_MISALIGNMENT.
30147         (vect_enhance_data_refs_alignment): Use.
30148         (vect_duplicate_ssa_name_ptr_info): Use.
30149         * tree-vectorizer.h (dr_misalignment): Use.
30150         (known_alignment_for_access_p): Use.
30152 2017-05-30  Jozef Lawrynowicz  <jozef.l@somniumtech.com>
30154         PR target/78838
30155         * config/msp430/msp430.c (gen_prefix): Return NULL when section name is
30156         .lowtext.
30157         (has_section_name): New function.
30159 2017-05-30  Martin Liska  <mliska@suse.cz>
30161         PR other/80909
30162         * auto-profile.c (get_function_decl_from_block): Fix
30163         parenthesis.
30165 2017-05-30  Richard Biener  <rguenther@suse.de>
30167         PR middle-end/80876
30168         * cfgexpand.c (expand_gimple_cond): Fixup preserving loops again.
30170 2017-05-30  Martin Liska  <mliska@suse.cz>
30172         * dumpfile.c: Use newly added macro DUMP_FILE_INFO.
30173         * dumpfile.h (struct dump_file_info): Remove ctors.
30175 2017-05-30  Martin Liska  <mliska@suse.cz>
30177         * predict.def: Fix GNU coding style.
30179 2017-05-29  Max Filippov  <jcmvbkbc@gmail.com>
30181         * config/xtensa/xtensa.c (xtensa_initial_elimination_offset):
30182         Mark 'to' argument with ATTRIBUTE_UNUSED.
30184 2017-05-29  Max Filippov  <jcmvbkbc@gmail.com>
30186         * config/xtensa/xtensa.c (xtensa_emit_call): Use
30187         HOST_WIDE_INT_PRINT_HEX instead of 0x%lx format string.
30188         (print_operand): Use HOST_WIDE_INT_PRINT_DEC instead of %ld
30189         format string.
30191 2017-05-29  Eric Botcazou  <ebotcazou@adacore.com>
30193         * doc/install.texi (Options specification): Restore entry of
30194         --enable-sjlj-exceptions.
30196 2017-05-27  Michael Eager  <eager@eagercon.com>
30198         Revert:
30199         2016-01-21  Ajit Agarwal  <ajitkum@xilinx.com>
30201         See https://gcc.gnu.org/ml/gcc/2017-05/msg00221.html.
30203         * config/microblaze/microblaze.h
30204         (FIXED_REGISTERS): Update in macro.
30205         (CALL_USED_REGISTERS): Update in macro.
30207 2017-05-27  François-Xavier Coudett  <fxcoudert@gcc.gnu.org>
30209         * doc/install.texi: Add links to macOS binary distributions.
30211 2017-05-27  Jakub Jelinek  <jakub@redhat.com>
30213         PR bootstrap/80887
30214         Revert:
30215         2017-05-25  Marc Glisse  <marc.glisse@inria.fr>
30217         * match.pd ((A +- CST1) +- CST2): Allow some conversions.
30219 2017-05-26  Martin Liska  <mliska@suse.cz>
30221         * dumpfile.h (enum dump_kind): Renumber TDF_* flags to be contiguous.
30223 2017-05-26  Martin Liska  <mliska@suse.cz>
30225         * cfg.c (check_bb_profile): Do not use TDF_COMMENT and print
30226         always leading ';; '.
30227         (dump_bb_info): Likewise.
30228         (brief_dump_cfg): Likewise.
30229         * cfgrtl.c (print_rtl_with_bb): Do not use TDF_COMMENT.
30230         * dumpfile.c: Remove usage of TDF_VERBOSE.
30231         * dumpfile.h (enum dump_kind): Likewise.
30232         (dump_gimple_bb_header): Do not use TDF_COMMENT.
30233         * print-tree.c (debug_verbose): Remove.
30234         * tree-cfg.c (gimple_dump_cfg): Do not use TDF_COMMENT.
30235         (dump_function_to_file): Remove dumps guarded with TDF_VERBOSE.
30236         * tree-diagnostic.c (default_tree_printer): Replace
30237         TDF_DIAGNOSTIC with TDF_SLIM.
30239 2017-05-26  Bin Cheng  <bin.cheng@arm.com>
30241         * tree-vect-loop-manip.c (create_intersect_range_checks_index): Pass
30242         in parameter loop, rather than loop_vinfo.
30243         (create_intersect_range_checks): Ditto.
30244         (vect_create_cond_for_alias_checks): Update call to above functions.
30246 2017-05-26  Bin Cheng  <bin.cheng@arm.com>
30248         PR tree-optimization/80815
30249         * tree-data-ref.c (prune_runtime_alias_test_list): Simplify condition
30250         for merging runtime alias checks.  Handle negative DR_STEPs.
30252 2017-05-26  Bin Cheng  <bin.cheng@arm.com>
30254         * tree-vect-data-refs.c (Operator==, comp_dr_with_seg_len_pair):
30255         Move from ...
30256         * tree-data-ref.c (Operator==, comp_dr_with_seg_len_pair): To here.
30257         * tree-vect-data-refs.c (vect_prune_runtime_alias_test_list): Factor
30258         out code pruning runtime alias checks.
30259         * tree-data-ref.c (prune_runtime_alias_test_list): New function
30260         factored out from above.
30261         * tree-vectorizer.h (struct dr_with_seg_len, dr_with_seg_len_pair_t):
30262         Move from ...
30263         * tree-data-ref.h (struct dr_with_seg_len, dr_with_seg_len_pair_t):
30264         ... to here.
30265         (prune_runtime_alias_test_list): New decalaration.
30267 2017-05-26  Bin Cheng  <bin.cheng@arm.com>
30269         * tree-vect-data-refs.c (compare_tree): Rename and move ...
30270         * tree-data-ref.c (data_ref_compare_tree): ... to here.
30271         * tree-data-ref.h (data_ref_compare_tree): New decalaration.
30272         * tree-vect-data-refs.c (dr_group_sort_cmp): Update uses.
30273         (operator==, comp_dr_with_seg_len_pair): Ditto.
30274         (vect_prune_runtime_alias_test_list): Ditto.
30276 2017-05-26  Martin Liska  <mliska@suse.cz>
30278         PR ipa/80663
30279         * params.def: Bound partial-inlining-entry-probability param.
30281 2017-05-26  Marek Polacek  <polacek@redhat.com>
30283         PR sanitizer/80875
30284         * fold-const.c (fold_binary_loc) <case MULT_EXPR>: Check if OP1
30285         can be negated.
30287 2017-05-26  Richard Biener  <rguenther@suse.de>
30289         PR tree-optimization/80842
30290         * tree-ssa-ccp.c (set_lattice_value): Always meet with the old
30291         value.
30293 2017-05-26  Richard Biener  <rguenther@suse.de>
30295         PR tree-optimization/80844
30296         * tree-vectorizer.c (adjust_simduid_builtins): Propagate results.
30298 2017-05-25  Sebastian Peryt  <sebastian.peryt@intel.com>
30300         * doc/md.texi (Machine Constraints): Update x86 family
30301         machine constraints section to match 'config/i386/constraints.md'.
30303 2017-05-25  Volker Reichelt  <v.reichelt@netcologne.de>
30305         * doc/invoke.texi (-Wcatch-value=): Document new warning option.
30307 2017-05-25  Nathan Sidwell  <nathan@acm.org>
30309         * doc/invoke.texi (--enable-languages): Update documentation.
30311 2017-05-25  Martin Liska  <mliska@suse.cz>
30313         * dumpfile.c: Add TDF_FOLDING.
30314         * dumpfile.h (enum dump_kind): Likewise.
30315         * genmatch.c (dt_simplify::gen_1): Use it.
30317 2017-05-25  Marc Glisse  <marc.glisse@inria.fr>
30319         * match.pd (view_convert (convert@0 @1)): Handle zero-extension.
30321 2017-05-25  Marc Glisse  <marc.glisse@inria.fr>
30323         * match.pd ((A +- CST1) +- CST2): Allow some conversions.
30324         * tree.c (drop_tree_overflow): Handle COMPLEX_CST and VECTOR_CST.
30326 2017-05-25  Marc Glisse  <marc.glisse@inria.fr>
30328         * fold-const.c (fold_binary_loc) [(A & C) == D]: Remove transformation.
30329         * match.pd (X == C): Rewrite it here.
30330         (with_possible_nonzero_bits, with_possible_nonzero_bits2,
30331         with_certain_nonzero_bits2): New predicates.
30332         * tree-ssanames.c (get_nonzero_bits): Handle INTEGER_CST.
30334 2017-05-24  Nathan Sidwell  <nathan@acm.org>
30336         * lto-streamer-in.c (lto_input_data_block): Adjust T const cast to
30337         avoid warning.
30339         * auto-profile.c (afdo_propagate): Adjust T const cast to avoid
30340         warning.
30342 2017-05-24  Segher Boessenkool  <segher@kernel.crashing.org>
30344         * config/powerpcspe: New port.  Files are copied from the rs6000
30345         port, with "rs6000" in filenames replaced by "powerpcspe".
30347 2017-05-24  Wilco Dijkstra  <wdijkstr@arm.com>
30349         PR rtl-optimization/80754
30350         * lra-remat.c (do_remat): Add overlap checks for dst_regno.
30352 2017-05-24  Sheldon Lobo  <smlobo@sheldon.us.oracle.com>
30354         * config/sparc/sparc.md (length): Return the correct value for -mflat
30355         sibcalls to match output_sibcall.
30357 2017-05-24  Segher Boessenkool  <segher@kernel.crashing.org>
30359         PR bootstrap/80860
30360         PR bootstrap/80843
30361         * config/rs6000/rs6000.c (struct machine_function): Add new field
30362         n_components.
30363         (rs6000_get_separate_components): Init that field, use it.
30364         (rs6000_components_for_bb): Use the field.
30366 2017-05-24  Thomas Preud'homme  <thomas.preudhomme@arm.com>
30368         * config/arm/arm.c (arm_expand_prologue): Fix typo in comment.
30370 2017-05-24  Peter Bergner  <bergner@vnet.ibm.com>
30372         PR middle-end/80823
30373         * tree-cfg.c (group_case_labels_stmt): Delete increment of "i";
30375 2017-05-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
30377         PR target/80725
30378         * config/s390/s390.c (s390_check_qrst_address): Check incoming
30379         address against address_operand predicate.
30380         * config/s390/s390.md ("*indirect_jump"): Swap alternatives.
30382 2017-05-24  Eric Botcazou  <ebotcazou@adacore.com>
30384         * var-tracking.c (track_expr_p): Do not return 0 for tracked record
30385         parameters passed indirectly.
30387 2017-05-23  Uros Bizjak  <ubizjak@gmail.com>
30389         * config/i386/i386.md (*movdi_internal): Remove SSE4
30390         alternative 18 (?r, *v).  Update insn attributes.
30391         (*movsi_internal): Remove SSE4 alternative 13 (?r, *v).
30392         Update insn attributes.
30393         (*zero_extendsidi2): Remove SSE4 alternative (?r, *x).
30394         Update insn attributes.
30395         * config/i386/sse.md (vec_extract<ssevecmodelower>_0): Remove SSE4
30396         alternative 1 (r, v). Remove isa attribute.
30397         * config/i386/i386.c (dimode_scalar_chain::make_vector_copies):
30398         Always move value through stack for !TARGET_INTER_UNIT_MOVES_TO_VEC
30399         and !TARGET_INTER_UNIT_MOVES_TO_VEC targets.
30401 2017-05-23  Tom de Vries  <tom@codesourcery.com>
30403         * doc/sourcebuild.texi (Directives, Verify compiler message): Document
30404         dg-line directive.
30406 2017-05-23  Jan Hubicka  <hubicka@ucw.cz>
30408         * cgraphunit.c (symbol_table::process_new_functions): Update.
30409         * ipa-fnsummary.c (pass_data_inline_parameters): Remove.
30410         (inline_generate_summary): Rename to ...
30411         (ipa_fn_summary_generate): ... this one.
30412         (inline_read_summary): Rename to ...
30413         (ipa_fn_summary_read): ... this one.
30414         (inline_write_summary): Rename to ...
30415         (ipa_fn_summary_write): ... this one.
30416         (inline_free_summary): Rename to ...
30417         (ipa_free_fn_summary): ... this one.
30418         (pass_data_local_fn_summary, pass_local_fn_summary,
30419         make_pass_local_fn_summary, pass_data_ipa_free_fn_summary,
30420         pass_ipa_free_fn_summary, make_pass_ipa_free_fn_summary,
30421         pass_data_ipa_fn_summary, pass_ipa_fn_summary,
30422         make_pass_ipa_fn_summary): New.
30423         * ipa-fnsummary.h (inline_generate_summary, inline_read_summary,
30424         inline_write_summary, inline_free_summary): Remove.
30425         (ipa_free_fn_summary) : New.
30426         * ipa-inline.c (ipa_inline): Update.
30427         (pass_ipa_inline): Do not generate summaries.
30428         * ipa.c (pass_data_ipa_free_fn_summary, pass_ipa_free_fn_summary):
30429         Remove.
30430         * passes.def: Replace pass_inline_parameters by pass_local_fn_summary
30431         and add pass_ipa_fn_summary.
30432         * tree-pass.h (make_pass_ipa_fn_summary, make_pass_local_fn_summary):
30433         New.
30434         (make_pass_inline_parameters): Remove.
30436 2017-05-23  Thomas Schwinge  <thomas@codesourcery.com>
30438         * omp-low.c (struct omp_context): Remove "default_kind" member.
30439         Adjust all users.
30441         * omp-offload.c (execute_oacc_device_lower): Remove the
30442         parallelism dimensions function attributes for unparallelized
30443         OpenACC kernels constructs.
30445 2017-05-23  Martin Liska  <mliska@suse.cz>
30447         * cgraph.c (cgraph_node::get_create): Use symtab_node::dump_{asm_,}name
30448         functions.
30449         (cgraph_edge::make_speculative): Likewise.
30450         (cgraph_edge::resolve_speculation): Likewise.
30451         (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
30452         (cgraph_node::dump): Likewise.
30453         * cgraph.h: Likewise.
30454         * cgraphunit.c (analyze_functions): Likewise.
30455         (symbol_table::compile): Likewise.
30456         * ipa-cp.c (print_all_lattices): Likewise.
30457         (determine_versionability): Likewise.
30458         (initialize_node_lattices): Likewise.
30459         (ipcp_verify_propagated_values): Likewise.
30460         (estimate_local_effects): Likewise.
30461         (update_profiling_info): Likewise.
30462         (create_specialized_node): Likewise.
30463         (perhaps_add_new_callers): Likewise.
30464         (decide_about_value): Likewise.
30465         (decide_whether_version_node): Likewise.
30466         (identify_dead_nodes): Likewise.
30467         (ipcp_store_bits_results): Likewise.
30468         * ipa-devirt.c (dump_targets): Likewise.
30469         (ipa_devirt): Likewise.
30470         * ipa-icf.c (sem_item::dump): Likewise.
30471         (sem_function::equals): Likewise.
30472         (sem_variable::equals): Likewise.
30473         (sem_item_optimizer::read_section): Likewise.
30474         (sem_item_optimizer::execute): Likewise.
30475         (congruence_class::dump): Likewise.
30476         * ipa-inline-analysis.c (dump_ipa_call_summary): Likewise.
30477         (dump_inline_summary): Likewise.
30478         (estimate_node_size_and_time): Likewise.
30479         (inline_analyze_function): Likewise.
30480         * ipa-inline-transform.c (inline_call): Likewise.
30481         * ipa-inline.c (report_inline_failed_reason): Likewise.
30482         (want_early_inline_function_p): Likewise.
30483         (edge_badness): Likewise.
30484         (update_edge_key): Likewise.
30485         (inline_small_functions): Likewise.
30486         * ipa-profile.c (ipa_profile): Likewise.
30487         * ipa-prop.c (ipa_print_node_jump_functions): Likewise.
30488         (ipa_make_edge_direct_to_target): Likewise.
30489         (remove_described_reference): Likewise.
30490         (ipa_impossible_devirt_target): Likewise.
30491         (propagate_controlled_uses): Likewise.
30492         (ipa_print_node_params): Likewise.
30493         (ipcp_transform_function): Likewise.
30494         * ipa-pure-const.c (pure_const_read_summary): Likewise.
30495         (propagate_pure_const): Likewise.
30496         * ipa-reference.c (generate_summary): Likewise.
30497         (read_write_all_from_decl): Likewise.
30498         (propagate): Likewise.
30499         (ipa_reference_read_optimization_summary): Likewise.
30500         * ipa-utils.c (ipa_merge_profiles): Likewise.
30501         * ipa.c (walk_polymorphic_call_targets): Likewise.
30502         (symbol_table::remove_unreachable_nodes): Likewise.
30503         (ipa_single_use): Likewise.
30504         * passes.c (execute_todo): Likewise.
30505         * predict.c (drop_profile): Likewise.
30506         * symtab.c (symtab_node::get_dump_name): New function.
30507         (symtab_node::dump_name): Likewise.
30508         (symtab_node::dump_asm_name): Likewise.
30509         (symtab_node::dump_references): Likewise.
30510         (symtab_node::dump_referring): Likewise.
30511         (symtab_node::dump_base): Likewise.
30512         (symtab_node::debug_symtab): Likewise.
30513         * tree-sra.c (convert_callers_for_node): Likewise.
30514         * tree-ssa-structalias.c (ipa_pta_execute): Likewise.
30515         * value-prof.c (init_node_map): Likewise.
30517 2017-05-23  Martin Liska  <mliska@suse.cz>
30519         * cgraph.h: Move symtab_node::dump_table to symbol_table::dump
30520         and symtab_node::debug_symtab to symbol_table::debug.
30521         * cgraphunit.c (analyze_functions): Use the renamed function.
30522         (symbol_table::compile): Likewise.
30523         * ipa-cp.c (ipcp_verify_propagated_values): Likewise.
30524         * ipa-icf.c (sem_item_optimizer::execute): Likewise.
30525         * passes.c (execute_todo): Likewise.
30526         * symtab.c (symbol_table::dump): New function.
30527         * tree-ssa-structalias.c (ipa_pta_execute): Use the renamed function.
30529 2017-05-23  Jan Hubicka  <hubicka@ucw.cz>
30531         * ipa-fnsummary.c (estimate_node_size_and_time): Do not sanity check
30532         that nonconst implies exec.
30534 2017-05-23  Jan Hubicka  <hubicka@ucw.cz>
30536         * ipa-inline-analysis.c (cgraph_2edge_hook_list, cgraph_edge_hook_list,
30537         inline_edge_removal_hook, inline_edge_duplication_hook): Remove.
30538         (inline_edge_summary_vec): Turn into ...
30539         (ipa_call_summaries): ... this one.
30540         (redirect_to_unreachable, edge_set_predicate,
30541         evaluate_properties_for_edge, inline_summary_alloc,
30542         reset_ipa_call_summary, reset_inline_summary,
30543         inline_summary_t::duplicate): Update.
30544         (inline_edge_duplication_hook): Turn to ...
30545         (ipa_call_summary_t::duplicate): ... this one.
30546         (inline_edge_removal_hook): Turn to ...
30547         (ipa_call_summary_t::remove): ... this one.
30548         (dump_inline_edge_summary): Turn to ...
30549         (dump_ipa_call_summary): ... this one.
30550         (estimate_function_body_sizes): Update.
30551         (inline_update_callee_summaries): Update.
30552         (remap_edge_change_prob): Update.
30553         (remap_edge_summaries): Update.
30554         (inline_merge_summary): Update.
30555         (do_estimate_edge_time): Update.
30556         (inline_generate_summary): Update.
30557         (inline_read_section): Update.
30558         (inline_read_summary): Update.
30559         (inline_free_summary): Update.
30560         * ipa-inline.c (can_inline_edge_p): Update.
30561         (compute_inlined_call_time): Update.
30562         (want_inline_small_function_p): Update.
30563         (edge_badness): Update.
30564         (early_inliner): Update.
30565         * ipa-inline.h (inline_edge_summary): Turn to ...
30566         (ipa_call_summary): ... this one.
30567         (ipa_call_summary_t): New class.
30568         (inline_edge_summary_t, inline_edge_summary_vec): Remove.
30569         (ipa_call_summaries): New.
30570         (inline_edge_summary): Remove.
30571         (estimate_edge_growth): Update.
30572         * ipa-profile.c (ipa_propagate_frequency_1): Update.
30573         * ipa-prop.c (ipa_make_edge_direct_to_target): Update.
30574         * ipa-split.c (execute_split_functions): Update.
30575         * ipa.c (symbol_table::remove_unreachable_nodes): Update.
30577 2017-05-23  Tom de Vries  <tom@codesourcery.com>
30579         * doc/sourcebuild.texi (Effective-Target Keywords, Other hardware
30580         attributes): Document rdrand effective target.
30582 2017-05-23  Tom de Vries  <tom@codesourcery.com>
30584         * doc/sourcebuild.texi (Effective-Target Keywords, Other hardware
30585         attributes): Sort alphabetically.
30587 2017-05-23  Georg-Johann Lay  <avr@gjlay.de>
30589         * config/avr/genmultilib.awk: Use gsub instead of gensub.
30591 2017-05-22  Michael Meissner  <meissner@linux.vnet.ibm.com>
30593         PR target/80718
30594         * config/rs6000/vsx.md (vsx_splat_<mode>, VSX_D iterator): Split
30595         V2DF/V2DI splat into two separate patterns, one that handles
30596         registers, and the other that only handles memory.  Drop support
30597         for splatting from a GPR on ISA 2.07 and then splitting the
30598         splat into direct move and splat.
30599         (vsx_splat_<mode>_reg): Likewise.
30600         (vsx_splat_<mode>_mem): Likewise.
30602 2017-05-22  Segher Boessenkool  <segher@kernel.crashing.org>
30604         * cfgcleanup.c (bb_is_just_return): Allow CLOBBERs.
30606 2017-05-22  Jakub Jelinek  <jakub@redhat.com>
30608         PR middle-end/80809
30609         * omp-low.c (finish_taskreg_remap): New function.
30610         (finish_taskreg_scan): If unit size of ctx->record_type
30611         is non-constant, unshare the size expression and replace
30612         decls in it with possible outer var refs.
30614         PR middle-end/80809
30615         * gimplify.c (omp_add_variable): For GOVD_DEBUG_PRIVATE use
30616         GOVD_SHARED rather than GOVD_PRIVATE with it.
30617         (gimplify_adjust_omp_clauses_1, gimplify_adjust_omp_clauses): Expect
30618         GOVD_SHARED rather than GOVD_PRIVATE with GOVD_DEBUG_PRIVATE.
30620         PR middle-end/80853
30621         * omp-low.c (lower_reduction_clauses): Pass OMP_CLAUSE_PRIVATE
30622         as last argument to build_outer_var_ref for pointer bases of array
30623         section reductions.
30625 2017-05-19  Martin Sebor  <msebor@redhat.com>
30627         * print-tree.c (print_node): Print DECL_READ_P flag.
30629 2017-05-22  Jan Hubicka  <hubicka@ucw.cz>
30631         * Makefile.in: Add ipa-fnsummary.o and ipa-fnsummary.h
30632         * auto-profile.c: Replace ipa-inline.h by ipa-fnsummary.h
30633         * cgraph.c: Likewise.
30634         * cgraphunit.c: Likewise.
30635         * gengtype.c: Likewise.
30636         * ipa-cp.c: Likewise.
30637         * ipa-devirt.c: Likewise.
30638         * ipa-icf.c: Likewise.
30639         * ipa-predicate.c: Likewise.
30640         * ipa-profile.c: Likewise.
30641         * ipa-prop.c: Likewise.
30642         * ipa-split.c: Likewise.
30643         * ipa.c: Likewise.
30644         * ipa-inline-analysis.c (inline_summaries, ipa_call_summaries,
30645         edge_predicate_pool, dump_inline_hints,
30646         inline_summary::account_size_time, redirect_to_unreachable,
30647         edge_set_predicate, set_hint_predicate,
30648         evaluate_conditions_for_known_args, evaluate_properties_for_edge,
30649         inline_summary_alloc, ipa_call_summary::reset, inline_summary::reset,
30650         inline_summary_t::remove, remap_hint_predicate_after_duplication,
30651         inline_summary_t::duplicate, ipa_call_summary_t::duplicate,
30652         ipa_call_summary_t::remove, initialize_growth_caches,
30653         free_growth_caches, dump_ipa_call_summary, dump_inline_summary,
30654         debug_inline_summary, dump_inline_summaries, initialize_inline_failed,
30655         mark_modified, unmodified_parm_1, unmodified_parm,
30656         unmodified_parm_or_parm_agg_item, eliminated_by_inlining_prob,
30657         set_cond_stmt_execution_predicate, set_switch_stmt_execution_predicate,
30658         compute_bb_predicates, will_be_nonconstant_expr_predicate,
30659         will_be_nonconstant_predicate, record_modified_bb_info,
30660         get_minimal_bb, record_modified, param_change_prob,
30661         phi_result_unknown_predicate, predicate_for_phi_result,
30662         array_index_predicate, clobber_only_eh_bb_p, fp_expression_p,
30663         estimate_function_body_sizes, compute_inline_parameters,
30664         compute_inline_parameters_for_curren, pass_data_inline_parameters,
30665         estimate_node_size_and_time, estimate_ipcp_clone_size_and_time,
30666         inline_update_callee_summaries, remap_edge_change_prob,
30667         remap_edge_summaries, remap_hint_predicate, inline_merge_summary,
30668         inline_update_overall_summary, inline_indirect_intraprocedural_analysis,
30669         inline_analyze_function, inline_summary_t::insert,
30670         inline_generate_summary, read_ipa_call_summary, inline_read_section,
30671         inline_read_summary, write_ipa_call_summary, inline_write_summary,
30672         inline_free_summary): Move to ipa-fnsummary.h
30673         (predicate_t): Remove.
30674         * ipa-fnsummary.c: New file.
30675         * ipa-inline.h:  Do not include sreal.h and ipa-predicate.h
30676         (enum inline_hints_vals, inline_hints, agg_position_info,
30677         INLINE_SIZE_SCALE, size_time_entry, inline_summary, inline_summary_t,
30678         inline_summaries, ipa_call_summary, ipa_call_summary_t,
30679         ipa_call_summaries, debug_inline_summary, dump_inline_summaries,
30680         dump_inline_summary, dump_inline_hints, inline_generate_summary,
30681         inline_read_summary, inline_write_summary, inline_free_summary,
30682         inline_analyze_function, initialize_inline_failed,
30683         inline_merge_summary, inline_update_overall_summary,
30684         compute_inline_parameters): Move to ipa-fnsummary.h
30685         * ipa-fnsummary.h: New file.
30686         * ipa-inline-transform.h: Include ipa-inline.h.
30687         * ipa-inline.c: LIkewise.
30689 2017-05-22  Jan Hubicka  <hubicka@ucw.cz>
30691         * ipa-inline.c (edge_badness): Use inlined_time instead of
30692         inline_summaries->get.
30694 2017-05-22  Jan Hubicka  <hubicka@ucw.cz>
30696         * ipa-inline.c (edge_badness): Use estimate_size_after_inlining.
30698 2017-05-22  Nathan Sidwell  <nathan@acm.org>
30700         * doc/invoke.texi (fdump-translation-unit): Delete documentation.
30701         (fdump-lang): Document 'raw' option.
30702         * dumpfile.h (TDI_tu): Delete.
30703         * dumpfile.c (dump_files): Remove translation-unit.
30704         (FIRST_AUTO_NUMBERED_DUMP): Decrement.
30706 2017-05-22  Georg-Johann Lay  <avr@gjlay.de>
30708         * config/avr/t-avr (AWK) [t-multilib]: Remove "-v FORMAT=Makefile"
30709         command option from $(AWK) call.
30710         * config/avr/genmultilib.awk: Simplify and rewrite so that it
30711         generates MULTILIB_REQUIRED instead of MULTILIB_EXCEPTIONS.
30712         [FORMAT]: Remove handling of variable.
30713         * config/avr/t-multilib: Regenerate.
30715 2017-05-22  Jan Hubicka  <hubicka@ucw.cz>
30717         * ipa-inline-analysis.c (inline_summary::reset): Do not reset
30718         self_time.
30719         (dump_inline_summary): Do not print self_time.
30720         (estimate_function_body_sizes): Do not set self_time.
30721         (compute_inline_parameters): Likewise.
30722         (inline_read_section, inline_write_summary): Do not stream self_time.
30723         * ipa-inline.h (inline_summary): Drop self_time.
30725 2017-05-22  Jan Hubicka  <hubicka@ucw.cz>
30727         * ipa-inline-analysis.c (account_size_time): Rename to ...
30728         (inline_summary::account_size_time): ... this one.
30729         (reset_ipa_call_summary): Turn to ...
30730         (ipa_call_summary::reset): ... this one.
30731         (reset_inline_summary): Turn to ...
30732         (inline_summary::reset): ... this one.
30733         (inline_summary_t::remove): Update.
30734         (inline_summary_t::duplicate): Update.
30735         (ipa_call_summary_t::remove): Update.
30736         (dump_inline_summary): Update.
30737         (estimate_function_body_sizes): Update.
30738         (compute_inline_parameters): Update.
30739         (estimate_node_size_and_time): Update.
30740         (inline_merge_summary): Update.
30741         (inline_update_overall_summary): Update.
30742         (inline_read_section): Update.
30743         (inline_write_summary): Update.
30744         * ipa-inline.h (inline_summary): Rename entry to size_time_table;
30745         add account_size_time and reset member functions.
30746         (ipa_call_summary): Add reset function.
30747         * ipa-predicate.h (predicate::operator &): Constify.
30749 2017-05-22  Richard Biener  <rguenther@suse.de>
30751         * df-scan.c (df_insn_refs_verify): Speedup when not verifying.
30753 2017-05-19  Jason Merrill  <jason@redhat.com>
30755         * tree.c (make_tree_vec_stat, grow_tree_vec_stat): Use size_t.
30757 2017-05-19  Marek Polacek  <polacek@redhat.com>
30759         PR sanitizer/80800
30760         * fold-const.c (extract_muldiv_1) <case TRUNC_DIV_EXPR>: Add
30761         TYPE_OVERFLOW_WRAPS checks.
30763 2017-05-19  Thomas Schwinge  <thomas@codesourcery.com>
30765         * tree-core.h (enum omp_clause_default_kind): Add
30766         "OMP_CLAUSE_DEFAULT_PRESENT".
30767         * tree-pretty-print.c (dump_omp_clause): Handle it.
30768         * gimplify.c (enum gimplify_omp_var_data): Add
30769         "GOVD_MAP_FORCE_PRESENT".
30770         (gimplify_adjust_omp_clauses_1): Map it to
30771         "GOMP_MAP_FORCE_PRESENT".
30772         (oacc_default_clause): Handle "OMP_CLAUSE_DEFAULT_PRESENT".
30774         * gimplify.c (oacc_default_clause): Clarify.
30776 2017-05-19  Nathan Sidwell  <nathan@acm.org>
30778         LANG_HOOK_REGISTER_DUMPS
30779         * toplev.c (general_init): Call register dump lang hook.
30780         * doc/invoke.texi: Document -fdump-lang option family.
30781         * dumpfile.c (dump_files): Remove class dump here.
30782         (FIRST_AUTO_NUMBERED_DUMP): Adjust.
30783         * dumpfile.h (tree_dump_index): Remove TDI_class.
30784         * langhooks-def.h (lhd_register_dumps): Declare.
30785         (LANG_HOOKS_REGISTER_DUMPS): Define.
30786         (LANG_HOOKS_INITIALIZER): Add it.
30787         * langhooks.c (lhd_register_dumps): Define.
30788         * langhooks.h (struct lang_hooks): Add register_dumps.
30790 2017-05-19  Nathan Sidwell  <nathan@acm.org>
30792         * context.h (context::set_passes): New.
30793         * context.c (context::context): Do not create pass manager.
30794         * toplev.c (general_init): Create pass manager here.
30796 2017-05-19  Segher Boessenkool  <segher@kernel.crashing.org>
30798         * config/rs6000/rs6000.md (splitter to load of -1 and mask): Don't
30799         use this splitter if two add or or instructions would also work for
30800         the constant we want to generate.
30802 2017-05-19  Richard Biener  <rguenther@suse.de>
30804         PR build/80821
30805         * genmatch.c (dt_node::gen_kids_1): Add missing scope around
30806         predicate evaluation.
30808 2017-05-19  Jan Hubicka  <hubicka@ucw.cz>
30810         * ipa-inline.h (ipa_call_summary): Turn sizes into signed;
30811         add ctor.
30812         * ipa-inline.c (want_inline_small_function_p): Do not cast to
30813         unsigned.
30815 2017-05-19  Jan Hubicka  <hubicka@ucw.cz>
30817         * ipa-inline-analysis.c (cgraph_2edge_hook_list, cgraph_edge_hook_list,
30818         inline_edge_removal_hook, inline_edge_duplication_hook): Remove.
30819         (inline_edge_summary_vec): Turn into ...
30820         (ipa_call_summaries): ... this one.
30821         (redirect_to_unreachable, edge_set_predicate,
30822         evaluate_properties_for_edge, inline_summary_alloc,
30823         reset_ipa_call_summary, reset_inline_summary,
30824         inline_summary_t::duplicate): Update.
30825         (inline_edge_duplication_hook): Turn to ...
30826         (ipa_call_summary_t::duplicate): ... this one.
30827         (inline_edge_removal_hook): Turn to ...
30828         (ipa_call_summary_t::remove): ... this one.
30829         (dump_inline_edge_summary): Turn to ...
30830         (dump_ipa_call_summary): ... this one.
30831         (estimate_function_body_sizes): Update.
30832         (inline_update_callee_summaries): Update.
30833         (remap_edge_change_prob): Update.
30834         (remap_edge_summaries): Update.
30835         (inline_merge_summary): Update.
30836         (do_estimate_edge_time): Update.
30837         (inline_generate_summary): Update.
30838         (inline_read_section): Update.
30839         (inline_read_summary): Update.
30840         (inline_free_summary): Update.
30841         * ipa-inline.c (can_inline_edge_p): Update.
30842         (compute_inlined_call_time): Update.
30843         (want_inline_small_function_p): Update.
30844         (edge_badness): Update.
30845         (early_inliner): Update.
30846         * ipa-inline.h (inline_edge_summary): Turn to ...
30847         (ipa_call_summary): ... this one.
30848         (ipa_call_summary_t): New class.
30849         (inline_edge_summary_t, inline_edge_summary_vec): Remove.
30850         (ipa_call_summaries): New.
30851         (inline_edge_summary): Remove.
30852         (estimate_edge_growth): Update.
30853         * ipa-profile.c (ipa_propagate_frequency_1): Update.
30854         * ipa-prop.c (ipa_make_edge_direct_to_target): Update.
30855         * ipa-split.c (execute_split_functions): Update.
30856         * ipa.c (symbol_table::remove_unreachable_nodes): Update.
30858 2017-05-19  Richard Biener  <rguenther@suse.de>
30860         PR middle-end/80764
30861         * cfgexpand.c (expand_gimple_cond): Fix loop fixup.
30863 2017-05-18  Segher Boessenkool  <segher@kernel.crashing.org>
30865         * config/rs6000/rs6000.c (struct machine_function): Add field
30866         fpr_is_wrapped_separately.
30867         (rs6000_get_separate_components): Use 64 components.  Handle the
30868         new FPR components.
30869         (rs6000_components_for_bb): Handle the FPR components.
30870         (rs6000_emit_prologue_components): Handle the FPR components.
30871         (rs6000_emit_epilogue_components): Handle the FPR components.
30872         (rs6000_set_handled_components): Handle the FPR components.
30873         (rs6000_emit_prologue): Don't output prologue code for those FPRs
30874         that are already separately shrink-wrapped.
30875         (rs6000_emit_epilogue): Don't output epilogue code for those FPRs
30876         that are already separately shrink-wrapped.
30878 2017-05-18  Michael Meissner  <meissner@linux.vnet.ibm.com>
30880         PR target/80510
30881         * config/rs6000/predicates.md (simple_offsettable_mem_operand):
30882         New predicate.
30884         * config/rs6000/rs6000.md (ALTIVEC_DFORM): New iterator.
30885         (define_peephole2 for Altivec d-form load): Add peepholes to catch
30886         cases where the register allocator uses a move and an offsettable
30887         memory operation to/from a FPR register on ISA 2.06/2.07.
30888         (define_peephole2 for Altivec d-form store): Likewise.
30890 2017-05-18  Uros Bizjak  <ubizjak@gmail.com>
30892         PR target/80799
30893         * config/i386/mmx.md (*mov<mode>_internal): Enable
30894         alternatives 11, 12, 13 and 14 also for 32bit targets.
30895         Remove alternatives 15, 16, 17 and 18.
30896         * config/i386/sse.md (vec_concatv2di): Change
30897         alternative (!x, *y) to (x, ?!*Yn).
30899 2017-05-18  Paolo Carlini  <paolo.carlini@oracle.com>
30901         * dumpfile.h (enum dump_kind): Remove stray comma.
30903 2017-05-18  Jan Hubicka  <hubicka@ucw.cz>
30905         * Makefile.in: Add ipa-predicate.o and ipa-predicate.h
30906         * ipa-inline-analysis.c (NUM_CONDITIONS): turn into
30907         predicate::num_conditions
30908         (IS_NOT_CONSTANT): turn into predicate::is_not_constant.
30909         (CHANGED): turn into predicate::changed.
30910         (agg_position_info): Move to ipa-predicate.h
30911         (add_condition, predicate::add_clause, predicate::operator &=,
30912         predicate::or_with, predicate::evaluate, predicate::probability,
30913         dump_condition, dump_clause, predicate::dump,
30914         predicate::remap_after_duplication, predicate::remap_after_inlining,
30915         predicate::stream_in, predicate::stream_out): Move to ipa-predicate.c
30916         (evaluate_conditions_for_known_args): Update.
30917         (set_cond_stmt_execution_predicate): Update.
30918         * ipa-inline.h: Include ipa-predicate.h
30919         (condition, inline_param_summary, conditions, agg_position_info,
30920         predicate): Move to ipa-predicate.h
30921         * ipa-predicate.c: New file.
30922         * ipa-predicate.h: New file.
30924 2017-05-18  Wilco Dijkstra  <wdijkstr@arm.com>
30926         * final.c (leaf_function_p): Check we are not in a sequence.
30928 2017-05-18  Martin Liska  <mliska@suse.cz>
30930         * cfgrtl.c (rtl_verify_edges): Remove usage of TDF_RTL.
30931         * dumpfile.c (dump_register): Use new enum dump_kind.
30932         (get_dump_file_name): Likewise.
30933         (dump_enable_all): Likewise.
30934         (dump_switch_p_1): Likewise.
30935         (enable_rtl_dump_file): Remove usage of TDF_RTL.
30936         * dumpfile.h (enum dump_kind): New enum type.
30937         (struct dump_file_info): Create constructor and
30938         format fields and comments.
30939         * passes.c (pass_manager::register_one_dump_file):
30940         Use num dump_kind.
30941         * statistics.c (statistics_early_init): Likewise.
30942         * tree-ssa-loop-prefetch.c (dump_mem_details): Replace
30943         TDF_TREE with TDF_SLIM.
30944         (gather_memory_references_ref): Likewise.
30946 2017-05-18  Martin Liska  <mliska@suse.cz>
30948         * vec.h (struct vnull): Use it.
30950 2017-05-18  Jan Hubicka  <hubicka@ucw.cz>
30952         * ipa-inline-analysis.c (predicate_conditions): Move to ipa-inline.h
30953         (true_predicate, false_predicate, true_predicate_p,
30954         false_predicate_p): Remove.
30955         (single_cond_predicate, not_inlined_predicate): Turn to member function
30956         in ipa-inline.h
30957         (add_condition): Update.
30958         (add_clause): Turn to...
30959         (predicate::add_clause): ... this one; update; allow passing NULL
30960         as parameter.
30961         (and_predicates): Turn to ...
30962         (predicate::operator &=): ... this one.
30963         (predicates_equal_p): Move to predicate::operator == in ipa-inline.h
30964         (or_predicates): Turn to ...
30965         (predicate::or_with): ... this one.
30966         (evaluate_predicate): Turn to ...
30967         (predicate::evaluate): ... this one.
30968         (predicate_probability): Turn to ...
30969         (predicate::probability): ... this one.
30970         (dump_condition): Update.
30971         (dump_predicate): Turn to ...
30972         (predicate::dump): ... this one.
30973         (account_size_time): Update.
30974         (edge_set_predicate): Update.
30975         (set_hint_predicate): UPdate.
30976         (evaluate_conditions_for_known_args): Update.
30977         (evaluate_properties_for_edge): Update.
30978         (remap_predicate_after_duplication): Turn to...
30979         (predicate::remap_after_duplication): ... this one.
30980         (remap_hint_predicate_after_duplication): Update.
30981         (inline_summary_t::duplicate): UPdate.
30982         (dump_inline_edge_summary): Update.
30983         (dump_inline_summary): Update.
30984         (set_cond_stmt_execution_predicate): Update.
30985         (set_switch_stmt_execution_predicate): Update.
30986         (compute_bb_predicates): Update.
30987         (will_be_nonconstant_expr_predicate): Update.
30988         (will_be_nonconstant_predicate): Update.
30989         (phi_result_unknown_predicate): Update.
30990         (predicate_for_phi_result): Update.
30991         (array_index_predicate): Update.
30992         (estimate_function_body_sizes): Update.
30993         (estimate_node_size_and_time): Update.
30994         (estimate_ipcp_clone_size_and_time): Update.
30995         (remap_predicate): Rename to ...
30996         (predicate::remap_after_inlining): ... this one.
30997         (remap_hint_predicate): Update.
30998         (inline_merge_summary): Update.
30999         (inline_update_overall_summary): Update.
31000         (estimate_size_after_inlining): Update.
31001         (read_predicate): Rename to ...
31002         (predicate::stream_in): ... this one.
31003         (read_inline_edge_summary): Update.
31004         (write_predicate): Rename to ...
31005         (predicate::stream_out): ... this one.
31006         (write_inline_edge_summary): Update.
31007         * ipa-inline.h (MAX_CLAUSES): Turn to predicate::max_clauses.
31008         (clause_t): Turn to uint32_t
31009         (predicate): Turn to class; implement constructor and operators
31010         ==, !=, &
31011         (size_time_entry): Update.
31012         (inline_summary): Update.
31013         (inline_edge_summary): Update.
31015 2017-05-18  Marc Glisse  <marc.glisse@inria.fr>
31017         * fold-const.c (fold_binary_loc): Move transformation...
31018         * match.pd (C - X CMP X): ... here.
31020 2017-05-18  Sheldon Lobo  <sheldon.lobo@oracle.com>
31022         * config/sparc/sparc.c (sparc_option_override): Set function
31023         alignment for -mcpu=niagara7 to 64 to match the I$ line.
31024         * config/sparc/sparc.h (BRANCH_COST): Set the SPARC M7 branch
31025         latency to 1.
31026         * config/sparc/sparc.h (BRANCH_COST): Set the SPARC T4 branch
31027         latency to 2.
31028         * config/sparc/sol2.h: Fix a ASM_CPU32_DEFAULT_SPEC typo.
31030 2017-05-18  Marek Polacek  <polacek@redhat.com>
31032         PR sanitizer/80797
31033         * ubsan.c (instrument_null): Unwrap ADDR_EXPRs.
31034         (pass_ubsan::execute): Call gimple_assign_single_p instead of
31035         gimple_assign_load_p.
31037 2017-05-17  Segher Boessenkool  <segher@kernel.crashing.org>
31039         PR middle-end/80692
31040         * real.c (do_compare): Give decimal_do_compare preference over
31041         comparing just the signs.
31043 2017-05-17  Uros Bizjak  <ubizjak@gmail.com>
31045         * doc/md.texi (Canonicalization of Instructions): Describe the
31046         canonical form of instructions that inherently set a condition
31047         code register.
31049 2017-05-17  Peter Bergner  <bergner@vnet.ibm.com>
31051         PR middle-end/80775
31052         * tree-cfg.c: Move deletion of unreachable case statements to after
31053         the merging of consecutive case labels.
31055 2017-05-17  Thomas Preud'homme  <thomas.preudhomme@arm.com>
31057         * config/arm/arm.c (cmse_nonsecure_call_clear_caller_saved): Refer
31058         readers to __gnu_cmse_nonsecure_call libcall for saving, clearing and
31059         restoring of callee-saved registers.
31061 2017-05-17  Eric Botcazou  <ebotcazou@adacore.com>
31063         * compare-elim.c (equivalent_reg_at_start): Adjust outdated comment.
31064         * config/visium/visium.c (single_set_and_flags): Likewise.
31065         * config/visium/visium.md (Substitutions): Likewise.
31067 2017-05-17  Martin Liska  <mliska@suse.cz>
31069         * cfg.c: Introduce dump_flags_t type and
31070         use it instead of int type.
31071         * cfg.h: Likewise.
31072         * cfghooks.c: Likewise.
31073         * cfghooks.h (struct cfg_hooks): Likewise.
31074         * cfgrtl.c: Likewise.
31075         * cfgrtl.h: Likewise.
31076         * cgraph.c (cgraph_node::get_body): Likewise.
31077         * coretypes.h: Likewise.
31078         * domwalk.c: Likewise.
31079         * domwalk.h: Likewise.
31080         * dumpfile.c (struct dump_option_value_info): Likewise.
31081         (dump_enable_all): Likewise.
31082         (dump_switch_p_1): Likewise.
31083         (opt_info_switch_p): Likewise.
31084         * dumpfile.h (enum tree_dump_index): Likewise.
31085         (struct dump_file_info): Likewise.
31086         * genemit.c: Likewise.
31087         * generic-match-head.c: Likewise.
31088         * gengtype.c (open_base_files): Likewise.
31089         * gimple-pretty-print.c: Likewise.
31090         * gimple-pretty-print.h: Likewise.
31091         * graph.c (print_graph_cfg): Likewise.
31092         * graphite-scop-detection.c (dot_all_sese): Likewise.
31093         * ipa-devirt.c (build_type_inheritance_graph): Likewise.
31094         * loop-unroll.c (report_unroll): Likewise.
31095         * passes.c (pass_manager::register_one_dump_file): Likewise.
31096         * print-tree.c: Likewise.
31097         * statistics.c: Likewise.
31098         * tree-cfg.c: Likewise.
31099         * tree-cfg.h: Likewise.
31100         * tree-dfa.c: Likewise.
31101         * tree-dfa.h: Likewise.
31102         * tree-dump.c (dump_function): Likewise.
31103         * tree-dump.h (struct dump_info): Likewise.
31104         * tree-pretty-print.c: Likewise.
31105         * tree-pretty-print.h: Likewise.
31106         * tree-ssa-live.c: Likewise.
31107         * tree-ssa-live.h: Likewise.
31108         * tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Likewise.
31109         * tree-vect-loop.c: Likewise.
31110         * tree-vect-slp.c: Likewise.
31112 2017-05-16  James Greenhalgh  <james.greenhalgh@arm.com>
31113             Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
31115         PR tree-optimization/80457
31116         * tree-vect-stmts.c (vect_model_simple_cost): Model the cost
31117         of all arguments to a statement as scalar_to_vec operations.
31118         (vectorizable_call): Adjust call to vect_model_simple_cost for
31119         new parameter.
31120         (vectorizable_conversion): Likewise.
31121         (vectorizable_assignment): Likewise.
31122         (vectorizable_shift): Likewise.
31123         (vectorizable_operation): Likewise.
31124         (vectorizable_comparison): Likewise.
31125         (vect_is_simple_cond): Record the def types for operands.
31126         (vectorizable_condition): Likewise, call vect_model_simple_cost.
31127         * tree-vectorizer.h (vect_model_simple_cost): Add new parameter
31128         for statement argument count.
31130 2017-05-16  Carl Love  <cel@us.ibm.com>
31132         * config/rs6000/rs6000-c: Add support for built-in functions
31133         vector unsigned long long vec_bperm (vector unsigned long long,
31134                                              vector unsigned char)
31135         vector signed long long vec_mule (vector signed int,
31136                                           vector signed int)
31137         vector unsigned long long vec_mule (vector unsigned int,
31138                                             vector unsigned int)
31139         vector signed long long vec_mulo (vector signed int,
31140                                           vector signed int)
31141         vector unsigned long long vec_mulo (vector unsigned int,
31142                                             vector unsigned int)
31143         vector signed char vec_sldw (vector signed char,
31144                                      vector signed char,
31145                                      const int)
31146         vector unsigned char vec_sldw (vector unsigned char,
31147                                        vector unsigned char,
31148                                        const int)
31149         vector signed short vec_sldw (vector signed short,
31150                                       vector signed short,
31151                                       const int)
31152         vector unsigned short vec_sldw (vector unsigned short,
31153                                         vector unsigned short,
31154                                         const int)
31155         vector signed int vec_sldw (vector signed int,
31156                                     vector signed int,
31157                                     const int)
31158         vector unsigned int vec_sldw (vector unsigned int,
31159                                       vector unsigned int,
31160                                       const int)
31161         vector signed long long vec_sldw (vector signed long long,
31162                                           vector signed long long,
31163                                           const int)
31164         vector unsigned long long vec_sldw (vector unsigned long long,
31165                                             vector unsigned long long,
31166                                             const int)
31167         * config/rs6000/rs6000-c: Add support for built-in functions
31168         * config/rs6000/rs6000-builtin.def: Add definition for SLDW.
31169         * config/rs6000/altivec.h: Add defintion for vec_sldw.
31170         * doc/extend.texi: Update the built-in documentation for the
31171         new built-in functions.
31173 2017-05-16  Marek Polacek  <polacek@redhat.com>
31175         PR sanitizer/80536
31176         PR sanitizer/80386
31177         * tree.c (save_expr): Don't fold the expression.
31179 2017-05-16  Uros Bizjak  <ubizjak@gmail.com>
31181         * config/i386/i386.md (*movsi_internal): Split (?rm,*y) alternative
31182         to (?r,*Yn) and (?m,*y) alternatives, and (?*y,rm) to (?*Ym,r)
31183         and (?*y,m).  Update insn attributes.
31185 2017-05-16  Martin Liska  <mliska@suse.cz>
31187         * cgraph.c (cgraph_edge::resolve_speculation): Add default value for
31188         flags argument of print_gimple_stmt, print_gimple_expr,
31189         print_generic_stmt and print_generic_expr.
31190         * cgraphclones.c (symbol_table::materialize_all_clones): Likewise.
31191         * coretypes.h: Likewise.
31192         * except.c (dump_eh_tree): Likewise.
31193         * gimple-fold.c (gimple_fold_stmt_to_constant_1): Likewise.
31194         * gimple-pretty-print.h: Likewise.
31195         * gimple-ssa-backprop.c (dump_usage_prefix): Likewise.
31196         (backprop::push_to_worklist): Likewise.
31197         (backprop::pop_from_worklist): Likewise.
31198         (backprop::process_use): Likewise.
31199         (backprop::intersect_uses): Likewise.
31200         (note_replacement): Likewise.
31201         * gimple-ssa-store-merging.c
31202         (pass_store_merging::terminate_all_aliasing_chains): Likewise.
31203         (imm_store_chain_info::coalesce_immediate_stores): Likewise.
31204         (pass_store_merging::execute): Likewise.
31205         * gimple-ssa-strength-reduction.c (dump_candidate): Likewise.
31206         (ssa_base_cand_dump_callback): Likewise.
31207         (dump_incr_vec): Likewise.
31208         (replace_refs): Likewise.
31209         (replace_mult_candidate): Likewise.
31210         (create_add_on_incoming_edge): Likewise.
31211         (create_phi_basis): Likewise.
31212         (insert_initializers): Likewise.
31213         (all_phi_incrs_profitable): Likewise.
31214         (introduce_cast_before_cand): Likewise.
31215         (replace_one_candidate): Likewise.
31216         * gimplify.c (gimplify_expr): Likewise.
31217         * graphite-isl-ast-to-gimple.c (is_valid_rename): Likewise.
31218         (set_rename): Likewise.
31219         (rename_uses): Likewise.
31220         (copy_loop_phi_nodes): Likewise.
31221         (add_close_phis_to_merge_points): Likewise.
31222         (copy_loop_close_phi_args): Likewise.
31223         (copy_cond_phi_args): Likewise.
31224         (graphite_copy_stmts_from_block): Likewise.
31225         (translate_pending_phi_nodes): Likewise.
31226         * graphite-poly.c (print_pdr): Likewise.
31227         (dump_gbb_cases): Likewise.
31228         (dump_gbb_conditions): Likewise.
31229         (print_scop_params): Likewise.
31230         * graphite-scop-detection.c (build_cross_bb_scalars_def): Likewise.
31231         (build_cross_bb_scalars_use): Likewise.
31232         (gather_bbs::before_dom_children): Likewise.
31233         * hsa-dump.c (dump_hsa_immed): Likewise.
31234         * ipa-cp.c (print_ipcp_constant_value): Likewise.
31235         (get_replacement_map): Likewise.
31236         * ipa-inline-analysis.c (dump_condition): Likewise.
31237         (estimate_function_body_sizes): Likewise.
31238         * ipa-polymorphic-call.c (check_stmt_for_type_change): Likewise.
31239         (ipa_polymorphic_call_context::get_dynamic_type): Likewise.
31240         * ipa-prop.c (ipa_dump_param): Likewise.
31241         (ipa_print_node_jump_functions_for_edge): Likewise.
31242         (ipa_modify_call_arguments): Likewise.
31243         (ipa_modify_expr): Likewise.
31244         (ipa_dump_param_adjustments): Likewise.
31245         (ipa_dump_agg_replacement_values): Likewise.
31246         (ipcp_modif_dom_walker::before_dom_children): Likewise.
31247         * ipa-pure-const.c (check_stmt): Likewise.
31248         (pass_nothrow::execute): Likewise.
31249         * ipa-split.c (execute_split_functions): Likewise.
31250         * omp-offload.c (dump_oacc_loop_part): Likewise.
31251         (dump_oacc_loop): Likewise.
31252         * trans-mem.c (tm_log_emit): Likewise.
31253         (tm_memopt_accumulate_memops): Likewise.
31254         (dump_tm_memopt_set): Likewise.
31255         (dump_tm_memopt_transform): Likewise.
31256         * tree-cfg.c (gimple_verify_flow_info): Likewise.
31257         (print_loop): Likewise.
31258         * tree-chkp-opt.c (chkp_print_addr): Likewise.
31259         (chkp_gather_checks_info): Likewise.
31260         (chkp_get_check_result): Likewise.
31261         (chkp_remove_check_if_pass): Likewise.
31262         (chkp_use_outer_bounds_if_possible): Likewise.
31263         (chkp_reduce_bounds_lifetime): Likewise.
31264         * tree-chkp.c (chkp_register_addr_bounds): Likewise.
31265         (chkp_mark_completed_bounds): Likewise.
31266         (chkp_register_incomplete_bounds): Likewise.
31267         (chkp_mark_invalid_bounds): Likewise.
31268         (chkp_maybe_copy_and_register_bounds): Likewise.
31269         (chkp_build_returned_bound): Likewise.
31270         (chkp_get_bound_for_parm): Likewise.
31271         (chkp_build_bndldx): Likewise.
31272         (chkp_get_bounds_by_definition): Likewise.
31273         (chkp_generate_extern_var_bounds): Likewise.
31274         (chkp_get_bounds_for_decl_addr): Likewise.
31275         * tree-chrec.c (chrec_apply): Likewise.
31276         * tree-data-ref.c (dump_data_reference): Likewise.
31277         (dump_subscript): Likewise.
31278         (dump_data_dependence_relation): Likewise.
31279         (analyze_overlapping_iterations): Likewise.
31280         * tree-inline.c (expand_call_inline): Likewise.
31281         (tree_function_versioning): Likewise.
31282         * tree-into-ssa.c (dump_defs_stack): Likewise.
31283         (dump_currdefs): Likewise.
31284         (dump_names_replaced_by): Likewise.
31285         (dump_update_ssa): Likewise.
31286         (update_ssa): Likewise.
31287         * tree-object-size.c (pass_object_sizes::execute): Likewise.
31288         * tree-parloops.c (build_new_reduction): Likewise.
31289         (try_create_reduction_list): Likewise.
31290         (ref_conflicts_with_region): Likewise.
31291         (oacc_entry_exit_ok_1): Likewise.
31292         (oacc_entry_exit_single_gang): Likewise.
31293         * tree-pretty-print.h: Likewise.
31294         * tree-scalar-evolution.c (set_scalar_evolution): Likewise.
31295         (get_scalar_evolution): Likewise.
31296         (add_to_evolution): Likewise.
31297         (get_loop_exit_condition): Likewise.
31298         (analyze_evolution_in_loop): Likewise.
31299         (analyze_initial_condition): Likewise.
31300         (analyze_scalar_evolution): Likewise.
31301         (instantiate_scev): Likewise.
31302         (number_of_latch_executions): Likewise.
31303         (gather_chrec_stats): Likewise.
31304         (final_value_replacement_loop): Likewise.
31305         (scev_const_prop): Likewise.
31306         * tree-sra.c (dump_access): Likewise.
31307         (disqualify_candidate): Likewise.
31308         (create_access): Likewise.
31309         (reject): Likewise.
31310         (maybe_add_sra_candidate): Likewise.
31311         (create_access_replacement): Likewise.
31312         (analyze_access_subtree): Likewise.
31313         (analyze_all_variable_accesses): Likewise.
31314         (sra_modify_assign): Likewise.
31315         (initialize_constant_pool_replacements): Likewise.
31316         (find_param_candidates): Likewise.
31317         (decide_one_param_reduction): Likewise.
31318         (replace_removed_params_ssa_names): Likewise.
31319         * tree-ssa-ccp.c (ccp_fold_stmt): Likewise.
31320         * tree-ssa-copy.c (dump_copy_of): Likewise.
31321         (copy_prop_visit_cond_stmt): Likewise.
31322         * tree-ssa-dce.c (mark_operand_necessary): Likewise.
31323         * tree-ssa-dom.c (pass_dominator::execute): Likewise.
31324         (record_equivalences_from_stmt): Likewise.
31325         * tree-ssa-dse.c (compute_trims): Likewise.
31326         (delete_dead_call): Likewise.
31327         (delete_dead_assignment): Likewise.
31328         * tree-ssa-forwprop.c (forward_propagate_into_gimple_cond): Likewise.
31329         (forward_propagate_into_cond): Likewise.
31330         (pass_forwprop::execute): Likewise.
31331         * tree-ssa-ifcombine.c (ifcombine_ifandif): Likewise.
31332         * tree-ssa-loop-im.c (invariantness_dom_walker::before_dom_children):
31333         Likewise.
31334         (move_computations_worker): Likewise.
31335         (execute_sm): Likewise.
31336         * tree-ssa-loop-ivcanon.c (tree_estimate_loop_size): Likewise.
31337         (remove_exits_and_undefined_stmts): Likewise.
31338         (remove_redundant_iv_tests): Likewise.
31339         * tree-ssa-loop-ivopts.c (dump_use): Likewise.
31340         (adjust_iv_update_pos): Likewise.
31341         * tree-ssa-math-opts.c (bswap_replace): Likewise.
31342         * tree-ssa-phiopt.c (factor_out_conditional_conversion): Likewise.
31343         (value_replacement): Likewise.
31344         * tree-ssa-phiprop.c (phiprop_insert_phi): Likewise.
31345         * tree-ssa-pre.c (print_pre_expr): Likewise.
31346         (get_representative_for): Likewise.
31347         (create_expression_by_pieces): Likewise.
31348         (insert_into_preds_of_block): Likewise.
31349         (eliminate_insert): Likewise.
31350         (eliminate_dom_walker::before_dom_children): Likewise.
31351         (eliminate): Likewise.
31352         (remove_dead_inserted_code): Likewise.
31353         * tree-ssa-propagate.c (substitute_and_fold): Likewise.
31354         * tree-ssa-reassoc.c (get_rank): Likewise.
31355         (eliminate_duplicate_pair): Likewise.
31356         (eliminate_plus_minus_pair): Likewise.
31357         (eliminate_not_pairs): Likewise.
31358         (undistribute_ops_list): Likewise.
31359         (eliminate_redundant_comparison): Likewise.
31360         (update_range_test): Likewise.
31361         (optimize_range_tests_var_bound): Likewise.
31362         (optimize_vec_cond_expr): Likewise.
31363         (rewrite_expr_tree): Likewise.
31364         (rewrite_expr_tree_parallel): Likewise.
31365         (linearize_expr): Likewise.
31366         (break_up_subtract): Likewise.
31367         (linearize_expr_tree): Likewise.
31368         (attempt_builtin_powi): Likewise.
31369         (attempt_builtin_copysign): Likewise.
31370         (transform_stmt_to_copy): Likewise.
31371         (transform_stmt_to_multiply): Likewise.
31372         (dump_ops_vector): Likewise.
31373         * tree-ssa-sccvn.c (vn_nary_build_or_lookup_1): Likewise.
31374         (print_scc): Likewise.
31375         (set_ssa_val_to): Likewise.
31376         (visit_reference_op_store): Likewise.
31377         (visit_use): Likewise.
31378         (sccvn_dom_walker::before_dom_children): Likewise.
31379         (run_scc_vn): Likewise.
31380         * tree-ssa-scopedtables.c (avail_exprs_stack::lookup_avail_expr):
31381         Likewise.
31382         (expr_hash_elt::print): Likewise.
31383         (const_and_copies::pop_to_marker): Likewise.
31384         (const_and_copies::record_const_or_copy_raw): Likewise.
31385         * tree-ssa-structalias.c (compute_dependence_clique): Likewise.
31386         * tree-ssa-uninit.c (collect_phi_def_edges): Likewise.
31387         (dump_predicates): Likewise.
31388         (find_uninit_use): Likewise.
31389         (warn_uninitialized_phi): Likewise.
31390         (pass_late_warn_uninitialized::execute): Likewise.
31391         * tree-ssa.c (verify_vssa): Likewise.
31392         (verify_ssa): Likewise.
31393         (maybe_optimize_var): Likewise.
31394         * tree-vrp.c (dump_value_range): Likewise.
31395         (dump_all_value_ranges): Likewise.
31396         (dump_asserts_for): Likewise.
31397         (register_edge_assert_for_2): Likewise.
31398         (vrp_visit_cond_stmt): Likewise.
31399         (vrp_visit_switch_stmt): Likewise.
31400         (vrp_visit_stmt): Likewise.
31401         (vrp_visit_phi_node): Likewise.
31402         (simplify_cond_using_ranges_1): Likewise.
31403         (fold_predicate_in): Likewise.
31404         (evrp_dom_walker::before_dom_children): Likewise.
31405         (evrp_dom_walker::push_value_range): Likewise.
31406         (evrp_dom_walker::pop_value_range): Likewise.
31407         (execute_early_vrp): Likewise.
31409 2017-05-16  Richard Biener  <rguenther@suse.de>
31411         * dwarf2out.c (loc_list_from_tree_1): Do not create
31412         DW_OP_GNU_variable_value for DECL_IGNORED_P decls.
31414 2017-05-16  Richard Biener  <rguenther@suse.de>
31416         * dwarf2out.c (resolve_variable_value_in_expr): Lookup DIE
31417         just generated.
31418         (note_variable_value_in_expr): If we resolved the decl ref
31419         do not push to the stack.
31421 2017-05-16  Matthew Wahab  <matthew.wahab@arm.com>
31423         * config/arm/arm_neon.h (vadd_f16): Use standard arithmetic
31424         operations in fast-math mode.
31425         (vaddq_f16): Likewise.
31426         (vmul_f16): Likewise.
31427         (vmulq_f16): Likewise.
31428         (vsub_f16): Likewise.
31429         (vsubq_f16): Likewise.
31430         * config/arm/neon.md (add<mode>3): New.
31431         (sub<mode>3): New.
31432         (fma:<VH:mode>3): New.  Also remove outdated comment.
31433         (mul<mode>3): New.
31435 2017-05-16  Martin Liska  <mliska@suse.cz>
31437         PR ipa/79849.
31438         PR ipa/79850.
31439         * ipa-devirt.c (warn_types_mismatch): Fix typo.
31440         (odr_types_equivalent_p): Likewise.
31442 2017-05-15  Sylvestre Ledru  <sylvestre@debian.org>
31444         * plugin.c (try_init_one_plugin): Fix ressource leaks (CID 726637).
31446 2017-05-15  Uros Bizjak  <ubizjak@gmail.com>
31448         PR target/80425
31449         * config/i386.i386.md (*zero_extendsidi2): Do not penalize
31450         non-interunit SSE move alternatives with '?'.
31451         (zero-extendsidi peephole2): New peephole to skip intermediate
31452         general register in SSE zero-extend sequence.
31454 2017-05-15  Jeff Law  <law@redhat.com>
31456         * reorg.c (relax_delay_slots): Create a new variable to hold
31457         the temporary target rather than clobbering TARGET_LABEL.
31459         * config/tilegx/tilegx.c (tilegx_expand_unaligned_load): Add
31460         missing argument to extract_bit_field call.
31461         * config/tilepro/tilepro.c (tilepro_expand_unaligned_load): Likewise.
31463 2017-05-15  Martin Liska  <mliska@suse.cz>
31465         PR driver/31468
31466         * gcc.c (process_command): Do not allow empty argument of -o option.
31468 2017-05-15  Renlin Li  <renlin.li@arm.com>
31470         * config/aarch64/aarch64-protos.h (aarch64_expand_call): Declare.
31471         * config/aarch64/aarch64.c (aarch64_expand_call): Define.
31472         * config/aarch64/constraints.md (Usf): Add long call check.
31473         * config/aarch64/aarch64.md (call): Use aarch64_expand_call.
31474         (call_value): Likewise.
31475         (sibcall): Likewise.
31476         (sibcall_value): Likewise.
31477         (call_insn): New.
31478         (call_value_insn): New.
31479         (sibcall_insn): Update rtx pattern.
31480         (sibcall_value_insn): Likewise.
31481         (call_internal): Remove.
31482         (call_value_internal): Likewise.
31483         (sibcall_internal): Likewise.
31484         (sibcall_value_internal): Likewise.
31485         (call_reg): Likewise.
31486         (call_symbol): Likewise.
31487         (call_value_reg): Likewise.
31488         (call_value_symbol): Likewise.
31490 2017-05-14  Krister Walfridsson  <krister.walfridsson@gmail.com>
31492         PR target/80600
31493         * config/netbsd.h (NETBSD_LIBGCC_SPEC): Always add -lgcc.
31495 2017-05-14  Uros Bizjak  <ubizjak@gmail.com>
31497         * config/i386.i386.c (ix86_cc_modes_compatible): CCNOmode is
31498         compatible with CCGOCmode and with CCZmode.
31500 2017-05-14  Martin Sebor  <msebor@redhat.com>
31502         PR middle-end/77671
31503         * gimple-fold.c (gimple_fold_builtin_sprintf): Make extern.
31504         (gimple_fold_builtin_snprintf): Same.
31505         * gimple-fold.h (gimple_fold_builtin_sprintf): Declare.
31506         (gimple_fold_builtin_snprintf): Same.
31507         * gimple-ssa-sprintf.c (get_format_string): Correct the detection
31508         of character types.
31509         (is_call_safe): New function.
31510         (try_substitute_return_value): Call it.
31511         (try_simplify_call): New function.
31512         (pass_sprintf_length::handle_gimple_call): Call it.
31514 2017-05-14  Martin Sebor  <msebor@redhat.com>
31516         PR middle-end/80669
31517         * builtins.c (expand_builtin_stpncpy): Simplify.
31519 2017-05-14  Daniel Santos  <daniel.santos@pobox.com>
31521         * config/i386/i386.opt: Add option -mcall-ms2sysv-xlogues.
31522         * config/i386/i386.h
31523         (x86_64_ms_sysv_extra_clobbered_registers): Change type to unsigned.
31524         (NUM_X86_64_MS_CLOBBERED_REGS): New macro.
31525         (struct machine_function): Add new members call_ms2sysv,
31526         call_ms2sysv_pad_in, call_ms2sysv_pad_out and call_ms2sysv_extra_regs.
31527         (struct machine_frame_state): New fields sp_realigned and
31528         sp_realigned_offset.
31529         * config/i386/i386.c
31530         (enum xlogue_stub): New enum.
31531         (enum xlogue_stub_sets): New enum.
31532         (class xlogue_layout): New class.
31533         (struct ix86_frame): New fields stack_realign_allocate_offset,
31534         stack_realign_offset and outlined_save_offset.  Modify comments to
31535         detail stack layout when using out-of-line stubs.
31536         (ix86_target_string): Add -mcall-ms2sysv-xlogues option.
31537         (ix86_option_override_internal): Add sorry() for TARGET_SEH and
31538         -mcall-ms2sysv-xlogues.
31539         (stub_managed_regs): New static variable.
31540         (ix86_save_reg): Add new parameter ignore_outlined to optionally omit
31541         registers managed by out-of-line stub.
31542         (disable_call_ms2sysv_xlogues): New function.
31543         (ix86_compute_frame_layout): Modify re-alignment calculations, disable
31544         m->call_ms2sysv when appropriate and compute frame layout for
31545         out-of-line stubs.
31546         (sp_valid_at, fp_valid_at): New inline functions.
31547         (choose_basereg): New function.
31548         (choose_baseaddr): Add align parameter, use choose_basereg and modify
31549         all callers.
31550         (ix86_emit_save_reg_using_mov, ix86_emit_restore_sse_regs_using_mov):
31551         Use align parameter of choose_baseaddr to generated aligned SSE movs
31552         when possible.
31553         (pro_epilogue_adjust_stack): Modify to track
31554         machine_frame_state::sp_realigned.
31555         (ix86_nsaved_regs): Modify to accommodate changes to ix86_save_reg.
31556         (ix86_nsaved_sseregs): Likewise.
31557         (ix86_emit_save_regs): Likewise.
31558         (ix86_emit_save_regs_using_mov): Likewise.
31559         (ix86_emit_save_sse_regs_using_mov): Likewise.
31560         (get_scratch_register_on_entry): Likewise.
31561         (gen_frame_set): New function.
31562         (gen_frame_load): Likewise.
31563         (gen_frame_store): Likewise.
31564         (emit_outlined_ms2sysv_save): Likewise.
31565         (emit_outlined_ms2sysv_restore): Likewise.
31566         (ix86_expand_prologue): Modify stack re-alignment code and call
31567         emit_outlined_ms2sysv_save when appropriate.
31568         (ix86_emit_leave): Clear machine_frame_state::sp_realigned.  Add
31569         parameter rtx_insn *insn, which allows the function to be used to only
31570         generate the notes.
31571         (ix86_expand_epilogue): Modify validity checks of frame and stack
31572         pointers, and call emit_outlined_ms2sysv_restore when appropriate.
31573         (ix86_expand_call): Modify to enable m->call_ms2sysv when appropriate.
31574         * config/i386/predicates.md
31575         (save_multiple): New predicate.
31576         (restore_multiple): Likewise.
31577         * config/i386/sse.md
31578         (save_multiple<mode>): New pattern.
31579         (save_multiple_realign<mode>): Likewise.
31580         (restore_multiple<mode>): Likewise.
31581         (restore_multiple_and_return<mode>): Likewise.
31582         (restore_multiple_leave_return<mode>): Likewise.
31583         * Makefile.in: Export HOSTCXX and HOSTCXXFLAGS to site.exp
31585 2017-05-14  Julia Koval  <julia.koval@intel.com>
31587         * config/i386/i386-builtin-types.def (VOID_FTYPE_INT_INT64): New type.
31588         * config/i386/i386-builtin.def (__builtin_ia32_xgetbv)
31589         (__builtin_ia32_xsetbv): New builtins.
31590         * config/i386/i386.c (ix86_expand_special_args_builtin):
31591         Process new types.
31592         (ix86_expand_builtin): Special expand for new intrinsics.
31593         * config/i386/i386.md (UNSPECV_XGETBV, UNSPECV_XSETBV): New.
31594         (xsetbv, xsetbv_rex64, xgetbv, xgetbv_rex64): New insn patterns.
31595         * config/i386/xsaveintrin.h (_xsetbv, _getbv): New intrinsics.
31597 2017-05-13  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
31599         * cfganal.c (inverted_post_order_compute): Change argument type
31600         to vec *.
31601         * cfganal.h (inverted_post_order_compute): Adjust prototype.
31602         * df-core.c (rest_of_handle_df_initialize): Adjust.
31603         (rest_of_handle_df_finish): Likewise.
31604         (df_analyze_1): Likewise.
31605         (df_analyze): Likewise.
31606         (loop_inverted_post_order_compute): Change argument to be a vec *.
31607         (df_analyze_loop): Adjust.
31608         (df_get_n_blocks): Likewise.
31609         (df_get_postorder): Likewise.
31610         * df.h (struct df_d): Change field to be a vec.
31611         * lcm.c (compute_laterin): Adjust.
31612         (compute_available): Likewise.
31613         * lra-lives.c (lra_create_live_ranges_1): Likewise.
31614         * tree-ssa-dce.c (remove_dead_stmt): Likewise.
31615         * tree-ssa-pre.c (compute_antic): Likewise.
31617 2017-05-13  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
31619         * cfganal.c (connect_infinite_loops_to_exit): Adjust.
31620         (depth_first_search::depth_first_search): Change structure init
31621         function to this constructor.
31622         (depth_first_search::add_bb): Rename function to this member.
31623         (depth_first_search::execute): Likewise.
31624         (flow_dfs_compute_reverse_finish): Adjust.
31626 2017-05-13  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
31628         * ddg.c (find_nodes_on_paths): Use auto_sbitmap.
31629         (longest_simple_path): Likewise.
31630         * shrink-wrap.c (spread_components): Likewise.
31631         (disqualify_problematic_components): Likewise.
31632         (emit_common_heads_for_components): Likewise.
31633         (emit_common_tails_for_components): Likewise.
31634         (insert_prologue_epilogue_for_components): Likewise.
31636 2017-05-13  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
31638         * tree-ssa-dse.c (dse_dom_walker): Make m_live_byes a
31639         auto_sbitmap.
31641 2017-05-13  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
31643         * df-core.c (df_set_blocks): Start using auto_bitmap.
31644         (df_compact_blocks): Likewise.
31645         * df-problems.c (df_rd_confluence_n): Likewise.
31646         * df-scan.c (df_insn_rescan_all): Likewise.
31647         (df_process_deferred_rescans): Likewise.
31648         (df_update_entry_block_defs): Likewise.
31649         (df_update_exit_block_uses): Likewise.
31650         (df_entry_block_bitmap_verify): Likewise.
31651         (df_exit_block_bitmap_verify): Likewise.
31652         (df_scan_verify): Likewise.
31653         * lra-constraints.c (lra_constraints): Likewise.
31654         (undo_optional_reloads): Likewise.
31655         (lra_undo_inheritance): Likewise.
31656         * lra-remat.c (calculate_gen_cands): Likewise.
31657         (do_remat): Likewise.
31658         * lra-spills.c (assign_spill_hard_regs): Likewise.
31659         (spill_pseudos): Likewise.
31660         * tree-ssa-pre.c (bitmap_set_and): Likewise.
31661         (bitmap_set_subtract_values): Likewise.
31663 2017-05-13  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
31665         * haifa-sched.c (estimate_shadow_tick): Replace manual bitmap
31666         management with auto_bitmap.
31667         (fix_inter_tick): Likewise.
31668         (fix_recovery_deps): Likewise.
31669         * ira.c (add_store_equivs): Likewise.
31670         (find_moveable_pseudos): Likewise.
31671         (split_live_ranges_for_shrink_wrap): Likewise.
31672         * print-rtl.c (rtx_reuse_manager::rtx_reuse_manager): Likewise.
31673         (rtx_reuse_manager::seen_def_p): Likewise.
31674         (rtx_reuse_manager::set_seen_def): Likewise.
31675         * print-rtl.h (class rtx_reuse_manager): Likewise.
31677 2017-05-13  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
31679         * bt-load.c (combine_btr_defs): Use auto_bitmap to manage bitmap
31680         lifetime.
31681         (migrate_btr_def): Likewise.
31682         * cfgloop.c (get_loop_body_in_bfs_order): Likewise.
31683         * df-core.c (loop_post_order_compute): Likewise.
31684         (loop_inverted_post_order_compute): Likewise.
31685         * hsa-common.h: Likewise.
31686         * hsa-gen.c (hsa_bb::~hsa_bb): Likewise.
31687         * init-regs.c (initialize_uninitialized_regs): Likewise.
31688         * ipa-inline.c (resolve_noninline_speculation): Likewise.
31689         (inline_small_functions): Likewise.
31690         * ipa-reference.c (ipa_reference_write_optimization_summary): Likewise.
31691         * ira.c (combine_and_move_insns): Likewise.
31692         (build_insn_chain): Likewise.
31693         * loop-invariant.c (find_invariants): Likewise.
31694         * lower-subreg.c (propagate_pseudo_copies): Likewise.
31695         * predict.c (tree_predict_by_opcode): Likewise.
31696         (predict_paths_leading_to): Likewise.
31697         (predict_paths_leading_to_edge): Likewise.
31698         (estimate_loops_at_level): Likewise.
31699         (estimate_loops): Likewise.
31700         * shrink-wrap.c (try_shrink_wrapping): Likewise.
31701         (spread_components): Likewise.
31702         * tree-cfg.c (remove_edge_and_dominated_blocks): Likewise.
31703         * tree-loop-distribution.c (rdg_build_partitions): Likewise.
31704         * tree-predcom.c (tree_predictive_commoning_loop): Likewise.
31705         * tree-ssa-coalesce.c (coalesce_ssa_name): Likewise.
31706         * tree-ssa-phionlycprop.c (pass_phi_only_cprop::execute): Likewise.
31707         * tree-ssa-pre.c (remove_dead_inserted_code): Likewise.
31708         * tree-ssa-sink.c (nearest_common_dominator_of_uses): Likewise.
31709         * tree-ssa-threadupdate.c (compute_path_counts): Likewise.
31710         (mark_threaded_blocks): Likewise.
31711         (thread_through_all_blocks): Likewise.
31712         * tree-ssa.c (verify_ssa): Likewise.
31713         (execute_update_addresses_taken): Likewise.
31714         * tree-ssanames.c (verify_ssaname_freelists): Likewise.
31716 2017-05-13  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
31718         * cfganal.c (mark_dfs_back_edges): Replace manual stack with
31719         auto_vec.
31720         (post_order_compute): Likewise.
31721         (inverted_post_order_compute): Likewise.
31722         (pre_and_rev_post_order_compute_fn): Likewise.
31724 2017-05-13  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
31726         * genrecog.c (int_set::int_set): Explicitly construct our
31727         auto_vec base class.
31728         * vec.h (auto_vec::auto_vec): New constructor.
31730 2017-05-13  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
31732         * bitmap.h (class auto_bitmap): New constructor taking
31733         bitmap_obstack * argument.
31735 2017-05-13  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
31737         * bitmap.h (class auto_bitmap): Change type of m_bits to
31738         bitmap_head, and adjust ctor / dtor and member operators.
31740 2017-05-13  Uros Bizjak  <ubizjak@gmail.com>
31742         * compare-elim.c (equivalent_reg_at_start): Return NULL_RTX
31743         when returned register mode doesn't match original mode.
31745 2017-05-12  Jeff Law  <law@redhat.com>
31746             Jakub Jelinek  <jakub@redhat.com>
31748         * config/mn10300/mn10300.c (mn10300_match_ccmode): Fix where
31749         we look for cc setter after the compare-elim changes.
31750         * config/mn10300/mn10300.md (addsi3_flags): Fix order of patterns
31751         within the vector to match what compare-elim now expects.
31752         (subsi3_flags, andsi3_flags, iorsi3_flags): Likewise.
31753         (xorsi3_flags, one_cmplsi2_flags): Likewise.
31755         * config/rx/rx.c (rx_match_ccmode): Fix where we look cc setter
31756         after the compare-elim changes.
31757         * config/rx/rx.md (abssi2_flags): Fix order of patterns within
31758         the vector to match what compare-elim now expects.
31759         (addsi3_flags, adc_flags, addsi3_flags peepholes): Likewise.
31760         (andsi3_flags, negsi2_flags, one_cmplsi2_flags): Likewise.
31761         (iorsi3_flags, rotlsi3_flags, rotrsi3_flags): Likewise.
31762         (ashrsi3_flags, lshrsi3_flags, ashlsi3_flags): Likewise.
31763         (ssaddsi3, subsi3_flags, sbb_flags, xorsi3_flags): Likewise.
31765         * config/visium/visium.c (single_set_and_flags): Fix where
31766         we look for cc setter after the compare-elim changes.
31767         * config/visium/visium.md (flags_subst_logic): Fix order of patterns
31768         with the vector to match what compare-elim now expects.
31769         (flags_subst_arith, add<mode>3_insn_set_carry): Likewise.
31770         (add<mode>3_insn_set_overflow, addsi3_insn_set_carry): Likewise.
31771         (addsi3_insn_set_overflow, sub<mode>3_insn_set_carry): Likewise.
31772         (sub<mode>3_insn_set_overflow, subsi3_insn_set_carry): Likewise.
31773         (subsi3_insn_set_overflow, negsi2_insn_set_carry): Likewise.
31774         (neg<mode>2_insn_set_overflow): Likewise.
31776 2017-05-12  Jim Wilson  <jim.wilson@linaro.org>
31778         PR middle-end/79794
31779         * expmed.c (extract_bit_field_1): Add alt_rtl argument.  Before
31780         maybe_expand_insn call, set ops[0].target.  If still set after call,
31781         set alt_rtl.  Add extra arg to recursive calls.
31782         (extract_bit_field): Add alt_rtl argument.  Pass to
31783         extract_bit_field.
31784         * expmed.h (extract_bit_field): Fix prototype.
31785         * expr.c (emit_group_load_1, copy_blkmode_from_reg)
31786         (copy_blkmode_to_reg, read_complex_part, store_field): Pass extra NULL
31787         to extract_bit_field_calls.
31788         (expand_expr_real_1): Pass alt_rtl to expand_expr_real instead of 0.
31789         Pass alt_rtl to extract_bit_field calls.
31790         * calls.c (store_unaligned_arguments_into_psuedos)
31791         load_register_parameters): Pass extra NULL to extract_bit_field calls.
31792         * optabs.c (maybe_legitimize_operand): Clear op->target when call
31793         gen_reg_rtx.
31794         * optabs.h (struct expand_operand): Add target bitfield.
31796 2017-05-12  Uros Bizjak  <ubizjak@gmail.com>
31798         * compare-elim.c (try_eliminate_compare): Canonicalize
31799         operation with embedded compare to
31800         [(set (reg:CCM) (compare:CCM (operation) (immediate)))
31801          (set (reg) (operation)].
31803         * config/i386/i386.c (TARGET_FLAGS_REGNUM): New define.
31805 2017-05-12  Uros Bizjak  <ubizjak@gmail.com>
31807         PR target/80723
31808         * config/i386/i386.c (ix86_rtx_cost) [case PLUS]: Ignore the
31809         cost of adding a carry flag for ADC instruction.
31810         [case MINUS]: Ignore the cost of subtracting a carry flag
31811         for SBB instruction.
31813 2017-05-12  Steven Munroe  <munroesj@gcc.gnu.org>
31815         * config.gcc (powerpc*-*-*): Add bmi2intrin.h, bmiintrin.h,
31816         and x86intrin.h
31817         * config/rs6000/bmiintrin.h: New file.
31818         * config/rs6000/bmi2intrin.h: New file.
31819         * config/rs6000/x86intrin.h: New file.
31821 2017-05-12  Jeff Law  <law@redhat.com>
31823         * tree-vrp.c (vrp_dom_walker::before_dom_children): Push unwinding
31824         markers.
31826 2017-05-12  Peter Bergner  <bergner@vnet.ibm.com>
31828         PR middle-end/80707
31829         * tree-cfg.c: Remove cfg edges of unreachable case statements.
31831 2017-05-12  Will Schmidt  <will_schmidt@vnet.ibm.com>
31833         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add handling for
31834         early expansion of vector divide builtins.
31835         (builtin_function_type): Add VSX_BUILTIN_UDIV_V2DI to the list of
31836         builtins identified as having unsigned arguments.
31838 2017-05-12  Will Schmidt  <will_schmidt@vnet.ibm.com>
31840         * config/rs6000/rs6000.c (gimple-fold.h): New #include.
31841         (rs6000_gimple_fold_builtin): Add handling for early GIMPLE
31842         expansion of vector logical operations (and, andc, or, xor,
31843         nor, orc, nand).
31845 2017-05-12  Will Schmidt  <will_schmidt@vnet.ibm.com>
31847         * gimple-fold.c (create_tmp_reg_or_ssa_name): Remove static declaration.
31848         * gimple-fold.h (create_tmp_reg_or_ssa_name): New prototype.
31850 2017-05-12  Will Schmidt  <will_schmidt@vnet.ibm.com>
31852         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add handling for
31853         early GIMPLE expansion of vector multiplies.
31855 2017-05-12  Prakhar Bahuguna  <prakhar.bahuguna@arm.com>
31857         * config/arm/arm.md (movsi): Add TARGET_32BIT in addition to the
31858         TARGET_HAVE_MOVT conditional.
31859         (movt splitter): Likewise.
31861 2017-05-12  Richard Biener  <rguenther@suse.de>
31863         * tree-ssa-sccvn.h (has_VN_INFO): Declare.
31864         * tree-ssa-pre.c (eliminate_dom_walker::before_dom_children):
31865         Fold all stmts not inplace.
31867 2017-05-12  Richard Biener  <rguenther@suse.de>
31869         PR tree-optimization/80713
31870         * tree-ssa-pre.c (remove_dead_inserted_code): Clear
31871         inserted_exprs bit for not removed stmts.
31873 2017-05-12  Thomas Schwinge  <thomas@codesourcery.com>
31875         PR middle-end/69921
31876         * tree-parloops.c (create_parallel_loop): Set "oacc kernels
31877         parallelized" attribute for parallelized OpenACC kernels.
31878         * omp-offload.c (execute_oacc_device_lower): Use it.
31880         * omp-expand.c (expand_omp_target) <GF_OMP_TARGET_KIND_OACC_KERNELS>:
31881         Set "oacc kernels" attribute.
31882         * omp-general.c (oacc_set_fn_attrib): Remove is_kernel formal
31883         parameter.  Adjust all users.
31884         (oacc_fn_attrib_kernels_p): Remove function.
31885         * omp-offload.c (execute_oacc_device_lower): Look for "oacc
31886         kernels" attribute instead of calling oacc_fn_attrib_kernels_p.
31887         * tree-ssa-loop.c (gate_oacc_kernels): Likewise.
31888         * tree-parloops.c (create_parallel_loop): If oacc_kernels_p,
31889         assert "oacc kernels" attribute is set.
31891 2017-05-11  Carl Love  <cel@us.ibm.com>
31893         * config/rs6000/rs6000-c: Add support for built-in functions
31894         vector unsigned char vec_popcnt (vector signed char)
31895         vector unsigned char vec_popcnt (vector unsigned char)
31896         vector unsigned short vec_popcnt (vector signed short)
31897         vector unsigned short vec_popcnt (vector unsigned short)
31898         vector unsigned int vec_popcnt (vector signed int)
31899         vector unsigned int vec_popcnt (vector unsigned int)
31900         vector unsigned long long vec_popcnt (vector signed long long)
31901         vector unsigned long long vec_popcnt (vector unsigned long long)
31902         vector signed long long vec_slo (vector signed long long,
31903                                          vector signed char)
31904         vector signed long long vec_slo (vector signed long long,
31905                                          vector unsigned char)
31906         vector unsigned long long vec_slo (vector unsigned long long,
31907                                            vector signed char)
31908         vector unsigned long long vec_slo (vector unsigned long long,
31909                                            vector unsigned char)
31910         * config/rs6000/rs6000-builtin.def: Add definitions for VPOPCNTUB,
31911         VPOPCNTUH, VPOPCNTUW, and VPOPCNTUD overloads.
31912         * config/rs6000/altivec.h: Add define for vec_popcnt, vec_popcntb,
31913         vec_popcnth, vec_popcntw and vec_popcntd built-in functions.
31914         * doc/extend.texi: Update the built-in documentation file for the
31915         new built-in functions.
31917 2017-05-11  Michael Meissner  <meissner@linux.vnet.ibm.com>
31919         * attribs.h (sorted_attr_string): Move machine independent
31920         functions for target clone support from the i386 port to common
31921         code.  Rename ix86_function_versions to common_function_versions.
31922         Rename make_name to make_unique_name.
31923         (common_function_versions): Likewise.
31924         (make_unique_name): Likewise.
31925         (make_dispatcher_decl): Likewise.
31926         (is_function_default_version): Likewise.
31927         * attribs.c (attr_strcmp): Likewise.
31928         (sorted_attr_string): Likewise.
31929         (common_function_versions): Likewise.
31930         (make_unique_name): Likewise.
31931         (make_dispatcher_decl): Likewise.
31932         (is_function_default_version): Likewise.
31933         * config/i386/i386.c (attr_strcmp): Likewise.
31934         (sorted_attr_string): Likewise.
31935         (ix86_function_versions): Likewise.
31936         (make_name): Likewise.
31937         (make_dispatcher_decl): Likewise.
31938         (is_function_default_version): Likewise.
31939         (TARGET_OPTION_FUNCTION_VERSIONS): Update target function hook.
31941 2017-05-11  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
31943         PR target/80695
31944         * config/rs6000/rs6000.c (rs6000_builtin_vectorization_cost):
31945         Account for direct move costs for vec_construct of integer
31946         vectors.
31948 2017-05-11  Uros Bizjak  <ubizjak@gmail.com>
31950         PR target/80706
31951         * config/i386/sync.md (UNSPEC_LDX_ATOMIC): New unspec.
31952         (UNSPEC_STX_ATOMIC): Ditto.
31953         (loaddi_via_sse): New insn.
31954         (storedi_via_sse): Ditto.
31955         (atomic_loaddi_fpu): Emit loaddi_via_sse and storedi_via_sse.
31956         Update corresponding peephole2 patterns.
31957         (atomic_storedi_fpu): Ditto.
31959 2017-05-11  Julia Koval  <julia.koval@intel.com>
31961         * config/i386/avx512fintrin.h (_mm_mask_rsqrt14_sd)
31962         (_mm_maskz_rsqrt14_sd, _mm_mask_rsqrt14_ss, _mm_maskz_rsqrt14_ss):
31963         New intrinsics.
31964         * config/i386/i386-builtin.def (__builtin_ia32_rsqrt14sd_mask)
31965         (__builtin_ia32_rsqrt14ss_mask): New builtins.
31966         * config/i386/sse.md (rsqrt14_<mode>_mask): New pattern.
31968 2017-05-11  Nathan Sidwell  <nathan@acm.org>
31970         * graphite-poly.c: Include dumpfile.h.
31972         * dumpfle.h (dump_function): Declare here ...
31973         * tree-dump.h (dump_function): ... not here.
31974         * dumpfile.c: #include tree-cfg.h.
31975         (dump_function): Move here from ...
31976         * tree-dump.c (dump_function): ... here.
31977         * gimplify.c: #include splay-tree.h, not tree-dump.h.
31978         * graphite-poly.c: Don't include tree-dump.h.
31979         * cgraphclones.c: Include dumpfile.h not tree-dump.h.
31980         * print-tree.c: Likewise.
31981         * stor-layout.c: Likewise.
31982         * tree-nested.c: Likewise.
31984         * dumpfile.c (dump_start): Use TDF_FLAGS.
31985         (dump_enable_all): Fix TDF_KIND check thinko.
31987 2017-05-11  Kelvin Nilsen  <kelvin@gcc.gnu.org>
31989         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
31990         array entries to represent two legal parameterizations of the
31991         overloaded __builtin_cmpb function, as represented by the
31992         P6_OV_BUILTIN_CMPB constant.
31993         (altivec_resolve_overloaded_builtin): Add special case handling
31994         for the __builtin_cmpb function, as represented by the
31995         P6_OV_BUILTIN_CMPB constant.
31996         * config/rs6000/rs6000-builtin.def (BU_P6_2): New macro.
31997         (BU_P6_64BIT_2): New macro.
31998         (BU_P6_OVERLOAD_2): New macro
31999         (CMPB_32): Add 32-bit compare-bytes support for 32-bit only targets.
32000         (CMPB): Add 64-bit compare-bytes support for 32-bit and 64-bit targets.
32001         (CMPB): Add overload support to represent both 32-bit and 64-bit
32002         compare-bytes function.
32003         * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add
32004         support for TARGET_CMPB.
32005         * config/rs6000/rs6000.h: Add support for RS6000_BTM_CMPB.
32006         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Add
32007         documentation of the __builtin_cmpb overloaded built-in function.
32009 2017-05-11  Richard Biener  <rguenther@suse.de>
32011         PR tree-optimization/80705
32012         * tree-vect-data-refs.c (vect_analyze_data_refs): DECL_NONALIASED
32013         bases are not vectorizable.
32015 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32017         * tree-ssa-loop-ivopts.c (determine_set_costs): Skip non-interger
32018         when counting register pressure.
32020 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32022         * tree-ssa-loop-ivopts.c (compare_cost_pair): New.
32023         (iv_ca_more_deps): Renamed to ...
32024         (iv_ca_compare_deps): ... this.
32025         (iv_ca_extend): Extend iv_ca if NEW_CP is cheaper than OLD_CP.
32027 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32029         * tree-ssa-loop-ivopts.c (find_interesting_uses): Move inv vars dump
32030         to ...
32031         (determine_group_iv_costs): ... here.
32032         (find_inv_vars_cb): Record inv var if it's not recorded before.
32034 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32036         * tree-ssa-loop-ivopts.c (allow_ip_end_pos_p): Refine comments.
32037         (get_shiftadd_cost): Ditto.
32039 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32041         * tree-ssa-address.c: Include header file.
32042         (move_hint_to_base): Return TRUE if BASE_HINT is moved to memory
32043         address.
32044         (add_to_parts): Refactor.
32045         (addr_to_parts): New parameter.  Update use of move_hint_to_base.
32046         (create_mem_ref): Update use of addr_to_parts.  Re-associate addr
32047         in new order.
32049 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32051         PR tree-optimization/53090
32052         * tree-ssa-loop-ivopts.c (enum comp_iv_rewrite): New enum value
32053         COMP_IV_EXPR_2.
32054         (extract_cond_operands): Detect condition with IV on both sides
32055         and return COMP_IV_EXPR_2.
32056         (find_interesting_uses_cond): Add iv_use for both IVs in condition.
32057         (rewrite_use_compare): Simplify by removing call to function
32058         extract_cond_operands.
32060 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32062         * tree-ssa-loop-ivopts.c (enum comp_iv_rewrite): New.
32063         (extract_cond_operands): Detect condition comparing against non-
32064         invariant bound and return appropriate enum value.
32065         (find_interesting_uses_cond): Update use of extract_cond_operands.
32066         Handle its return value accordingly.
32067         (determine_group_iv_cost_cond, rewrite_use_compare): Ditto.
32069 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32071         * tree-ssa-loop-ivopts.c (rewrite_use_nonlinear_expr): Re-associate
32072         nonlinear iv_use computation in loop invariant sensitive way.
32074 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32076         * tree-ssa-loop-ivopts.c (relate_compare_use_with_all_cands): New.
32077         (find_iv_candidates): Call relate_compare_use_with_all_cands.
32079 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32081         * tree-ssa-loop-ivopts.c (struct iv_cand): New field inv_exprs.
32082         (dump_cand): Support iv_cand.inv_exprs.
32083         (add_candidate_1): Record invariant exprs in iv_cand.inv_exprs
32084         for candidates.
32085         (iv_ca_set_no_cp, iv_ca_set_cp, free_loop_data): Support
32086         iv_cand.inv_exprs.
32088 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32090         * tree-ssa-loop-ivopts.c (multiplier_allowed_in_address_p): Move
32091         from ...
32092         * tree-ssa-address.c (multiplier_allowed_in_address_p): ... to here
32093         as local function.  Include necessary header files.
32094         * tree-ssa-loop-ivopts.h (multiplier_allowed_in_address_p): Delete.
32096 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32098         * tree-ssa-loop-ivopts.c (autoinc_possible_for_pair): Simplify.
32100 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32102         * tree-ssa-loop-ivopts.c (force_expr_to_var_cost): Handle more
32103         operators: TRUNC_DIV_EXPR, BIT_AND_EXPR, BIT_IOR_EXPR, LSHIFT_EXPR,
32104         RSHIFT_EXPR and BIT_NOT_EXPR.
32106 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32108         * tree-ssa-loop-ivopts.c (get_loop_invariant_expr): Simplify.
32109         (adjust_setup_cost): New parameter supporting round up adjustment.
32110         (struct address_cost_data): Delete.
32111         (force_expr_to_var_cost): Don't bound cost with spill_cost.
32112         (split_address_cost, ptr_difference_cost): Delete.
32113         (difference_cost, compare_aff_trees, record_inv_expr): Delete.
32114         (struct ainc_cost_data): New struct.
32115         (get_address_cost_ainc): New function.
32116         (get_address_cost, get_computation_cost): Reimplement.
32117         (determine_group_iv_cost_address): Record inv_expr for all uses of
32118         a group.
32119         (determine_group_iv_cost_cond): Call get_loop_invariant_expr.
32120         (iv_ca_has_deps): Reimplemented to ...
32121         (iv_ca_more_deps): ... this.  Check if NEW_CP introduces more deps
32122         than OLD_CP.
32123         (iv_ca_extend): Call iv_ca_more_deps.
32125 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32127         * tree-ssa-address.c (struct mem_address): Move to header file.
32128         (valid_mem_ref_p, move_fixed_address_to_symbol): Make it global.
32129         * tree-ssa-address.h (struct mem_address): Move from C file.
32130         (valid_mem_ref_p, move_fixed_address_to_symbol): Declare.
32132 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32134         * tree-affine.h (aff_combination_type): New interface.
32135         (aff_combination_zero_p): Remove static.
32136         (aff_combination_const_p): New interface.
32137         (aff_combination_singleton_var_p): New interfaces.
32139 2017-05-11  Richard Biener  <rguenther@suse.de>
32141         * tree-ssa-pre.c (eliminate_dom_walker::before_dom_children):
32142         Skip unreachable blocks and destinations.
32143         (eliminate): Move stmt removal and fixup ...
32144         (fini_eliminate): ... here.  Skip inserted exprs.
32145         (pass_pre::execute): Move fini_pre after fini_eliminate.
32146         * tree-ssa-tailmerge.c: Include tree-cfgcleanup.h.
32147         (tail_merge_optimize): Run cleanup_tree_cfg if requested by
32148         PRE to get rid of dead code that has invalid SSA form and
32149         split critical edges again.
32151 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32153         * rtlanal.c (rtx_cost): Handle TRUNCATE between tieable modes.
32155 2017-05-11  Richard Biener  <rguenther@suse.de>
32157         * passes.c (execute_function_todo): Verify loops if they are
32158         said to be up-to-date.
32159         * cfgexpand.c (pass_expand::execute): Discard loops for -dx.
32160         * trans-mem.c (pass_tm_edges::execute): Mark loops for fixup.
32162 2017-05-10  John David Anglin  <danglin@gcc.gnu.org>
32164         PR target/80090
32165         * config/pa/pa.c (pa_assemble_integer): When outputting a SYMBOL_REF,
32166         handle calling assemble_external ourself.
32168         PR target/79027
32169         * config/pa/pa.c (pa_cannot_change_mode_class): Reject changes to/from
32170         modes with zero size.  Enhance comment.
32172 2017-05-10  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
32174         * config/rs6000/rs6000.c (altivec_init_builtins): Define POWER8
32175         built-ins for vec_xl and vec_xst with short and char pointer
32176         arguments.
32178 2017-05-10  Sebastian Peryt  <sebastian.peryt@intel.com>
32180         * config/i386/avx512fintrin.h (_mm_mask_max_round_sd)
32181         (_mm_maskz_max_round_sd, _mm_mask_max_round_ss)
32182         (_mm_maskz_max_round_ss, _mm_mask_min_round_sd)
32183         (_mm_maskz_min_round_sd, _mm_mask_min_round_ss)
32184         (_mm_maskz_min_round_ss): New intrinsics.
32185         * config/i386/i386-builtin-types.def (V2DF, V2DF, V2DF, V2DF, UQI, INT)
32186         (V4SF, V4SF, V4SF, V4SF, UQI, INT): New function type aliases.
32187         * config/i386/i386-builtin.def (__builtin_ia32_maxsd_mask_round)
32188         (__builtin_ia32_maxss_mask_round, __builtin_ia32_minsd_mask_round)
32189         (__builtin_ia32_minss_mask_round): New builtins.
32190         * config/i386/i386.c (V2DF_FTYPE_V2DF_V2DF_V2DF_UQI_INT)
32191         (V4SF_FTYPE_V4SF_V4SF_V4SF_UQI_INT): Handle new types.
32192         * config/i386/sse.md (<sse>_vm<code><mode>3<round_saeonly_name>):
32193         Rename to ...
32194         (<sse>_vm<code><mode>3<mask_name><round_saeonly_name>): ... this.
32195         (v<maxmin_float><ssescalarmodesuffix>\t{<round_saeonly_op3>%2, %1, %0|%0, %1, %<iptr>2<round_saeonly_op3>}):
32196         Change to ...
32197         (v<maxmin_float><ssescalarmodesuffix>\t{<round_saeonly_mask_op3>%2, %1, %0<mask_operand3>|%0<mask_operand3>, %1, %<iptr>2<round_saeonly_mask_op3>}):
32198         ... this.
32200 2017-05-10  Sebastian Peryt  <sebastian.peryt@intel.com>
32202         * config/i386/avx512fintrin.h (_mm_mask_mul_round_sd)
32203         (_mm_maskz_mul_round_sd, _mm_mask_mul_round_ss)
32204         (_mm_maskz_mul_round_ss, _mm_mask_div_round_sd)
32205         (_mm_maskz_div_round_sd, _mm_mask_div_round_ss)
32206         (_mm_maskz_div_round_ss, _mm_mask_mul_sd, _mm_maskz_mul_sd)
32207         (_mm_mask_mul_ss, _mm_maskz_mul_ss, _mm_mask_div_sd)
32208         (_mm_maskz_div_sd, _mm_mask_div_ss, _mm_maskz_div_ss): New intrinsics.
32209         * config/i386/i386-builtin-types.def (V2DF_FTYPE_V2DF_V2DF_V2DF_UQI_INT)
32210         (V4SF_FTYPE_V4SF_V4SF_V4SF_UQI_INT): New function type aliases.
32211         * config/i386/i386-builtin.def (__builtin_ia32_divsd_mask_round)
32212         (__builtin_ia32_divss_mask_round, __builtin_ia32_mulsd_mask_round)
32213         (__builtin_ia32_mulss_mask_round): New builtins.
32214         * config/i386/i386.c (V2DF_FTYPE_V2DF_V2DF_V2DF_UQI_INT)
32215         (V4SF_FTYPE_V4SF_V4SF_V4SF_UQI_INT): Handle new types.
32216         * config/i386/sse.md (<sse>_vm<multdiv_mnemonic><mode>3<round_name>):
32217         Rename to ...
32218         (<sse>_vm<multdiv_mnemonic><mode>3<mask_name><round_name>): ... this.
32219         (v<multdiv_mnemonic><ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|%0, %1, %<iptr>2<round_op3>}):
32220         Change to ...
32221         (v<multdiv_mnemonic><ssescalarmodesuffix>\t{<round_mask_op3>%2, %1, %0<mask_operand3>|%0<mask_operand3>, %1, %<iptr>2<round_mask_op3>}):
32222         ... this.
32224 2017-05-10  Julia Koval  <julia.koval@intel.com>
32226         * config/i386/avxintrin.h (_mm256_set_m128, _mm256_set_m128d)
32227         (_mm256_set_m128i, _mm256_setr_m128, _mm256_setr_m128d)
32228         (_mm256_setr_m128i): New intrinsics.
32230 2017-05-10  Julia Koval  <julia.koval@intel.com>
32232         * config/i386/avx512fintrin.h (_mm_mask_rcp14_sd)
32233         (_mm_maskz_rcp14_sd, _mm_mask_rcp14_ss)
32234         (_mm_maskz_rcp14_ss): New intrinsics.
32235         * config/i386/i386-builtin.def (__builtin_ia32_rcp14sd_mask)
32236         (__builtin_ia32_rcp14ss_mask): New builtins.
32237         * config/i386/sse.md (srcp14<mode>_mask): New pattern.
32239 2017-05-10  Peter Bergner  <bergner@vnet.ibm.com>
32241         PR tree-optimization/51513
32242         * tree-cfg.c (gimple_seq_unreachable_p): New function.
32243         (assert_unreachable_fallthru_edge_p): Use it.
32244         (group_case_labels_stmt): Likewise.
32245         * tree-cfg.h: Prototype it.
32246         * stmt.c: Include cfghooks.h and tree-cfg.h.
32247         (emit_case_dispatch_table) <gap_label>: New local variable.
32248         Use it to fill dispatch table gaps.
32249         Test for default_label before updating probabilities.
32250         (expand_case) <default_label>: Remove unneeded initialization.
32251         Test for unreachable default case statement and remove its edge.
32252         Set default_label accordingly.
32253         * tree-ssa-ccp.c (optimize_unreachable): Update comment.
32255 2017-05-10  Carl Love  <cel@us.ibm.com>
32257         * config/rs6000/rs6000-c: Add support for built-in functions
32258         vector signed char      vec_neg (vector signed char)
32259         vector signed short int vec_neg (vector short int)
32260         vector signed int       vec_neg (vector signed int)
32261         vector signed long long vec_neg (vector signed long long)
32262         vector float            vec_neg (vector float)
32263         vector double           vec_neg (vector double)
32264         * config/rs6000/rs6000-builtin.def: Add definitions for NEG function
32265         overload.
32266         * config/rs6000/altivec.h: Add define for vec_neg
32267         * doc/extend.texi: Update the built-in documentation for the
32268         new built-in functions.
32270 2017-05-10  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
32272         PR tree-optimization/77644
32273         * match.pd (sqrt(x) cmp sqrt(y) -> x cmp y): New pattern.
32275 2017-05-10  Nathan Sidwell  <nathan@acm.org>
32277         * dumpfile.h (TDI_lang_all): New.
32278         (TDF_KIND): New. Renumber others
32279         (TDF_LANG, TDF_TREE, TDF_RTL, TDF_IPA): Enumerate value, rather
32280         than bits.
32281         * dumpfile.c (dump_files): Mark language dumps as TDF_LANG.  add
32282         lang-all.
32283         (get_dump_file_name): Adjust suffix generation.
32284         (dump_enable_all): Use TDF_KIND.
32285         * doc/invoke.texi (-fdump-lang-all): Document.
32287         * dumpfile.h: Tabify.
32289 2017-05-10  Wilco Dijkstra  <wdijkstr@arm.com>
32291         PR target/80671
32292         * config/aarch64/cortex-a57-fma-steering.c (merge_forest):
32293         Move member access before delete.
32295 2017-05-10  Alexandre Oliva  <aoliva@redhat.com>
32297         * tree-inline.c (expand_call_inline): Split block at stmt
32298         before the call.
32300 2017-05-09  Michael Meissner  <meissner@linux.vnet.ibm.com>
32302         PR target/68163
32303         * config/rs6000/rs6000.md (f32_lr): Delete mode attributes that
32304         are now unused after splitting mov{sf,sd}_hardfloat.
32305         (f32_lr2): Likewise.
32306         (f32_lm): Likewise.
32307         (f32_lm2): Likewise.
32308         (f32_li): Likewise.
32309         (f32_li2): Likewise.
32310         (f32_lv): Likewise.
32311         (f32_sr): Likewise.
32312         (f32_sr2): Likewise.
32313         (f32_sm): Likewise.
32314         (f32_sm2): Likewise.
32315         (f32_si): Likewise.
32316         (f32_si2): Likewise.
32317         (f32_sv): Likewise.
32318         (f32_dm): Likewise.
32319         (f32_vsx): Likewise.
32320         (f32_av): Likewise.
32321         (mov<mode>_hardfloat): Split into separate movsf and movsd pieces.
32322         For movsf, order stores so the VSX stores occur before the GPR
32323         store which encourages the register allocator to use a traditional
32324         FPR instead of a GPR.  For movsd, order the stores so that the GPR
32325         store comes before the VSX stores to allow the power6 to work.
32326         This is due to the power6 not having a 32-bit integer store
32327         instruction from a FPR.
32328         (movsf_hardfloat): Likewise.
32329         (movsd_hardfloat): Likewise.
32331 2017-05-09  Martin Sebor  <msebor@redhat.com>
32333         PR translation/80280
32334         * config/sol2-c.c (cmn_err_flag_specs): Initialize new data member
32335         added in r247778.
32337         PR translation/80280
32338         * config/i386/msformat-c.c (ms_printf_flag_specs): Initialize new
32339         data member added in r247778.
32340         (ms_scanf_flag_specs, ms_strftime_flag_specs): Same.
32342 2017-05-09  Nathan Sidwell  <nathan@acm.org>
32344         * tree.h (tree_fits_shwi_p, tree_fits_uhwi_p): Unconditionally pure.
32346         * ipa-devirt.c (default_hash_traits<type_pair>): Skip struct-scope
32347         typedefs.
32349 2017-05-09  Marek Polacek  <polacek@redhat.com>
32351         * doc/invoke.texi: Fix typo.
32353 2017-05-09  Richard Biener  <rguenther@suse.de>
32355         * tree-vrp.c (vrp_val_is_max): Adjust comment.
32356         (vrp_val_is_min): Likewise.
32357         (set_value_range_to_value): Likewise.
32358         (set_value_range_to_nonnegative): Likewise.
32359         (gimple_assign_nonzero_p): Likewise.
32360         (gimple_stmt_nonzero_p): Likewise.
32361         (vrp_int_const_binop): Likewise.  Remove unreachable case.
32362         (adjust_range_with_scev): Adjust comments.
32363         (compare_range_with_value): Likewise.
32364         (extract_range_from_phi_node): Likewise.
32365         (test_for_singularity): Likewise.
32367 2017-05-09  Richard Biener  <rguenther@suse.de>
32369         * tree-vrp.c (get_single_symbol): Add assert that we don't
32370         get overflowed constants as invariant part.
32371         (compare_values_warnv): Add comment before the TREE_NO_WARNING
32372         checks.  Use wi::cmp instead of recursing for integer constants.
32373         (compare_values): Just ignore whether we assumed undefined
32374         overflow instead of failing the compare.
32375         (extract_range_for_var_from_comparison_expr): Add comment before the
32376         TREE_NO_WARNING sets.
32377         (test_for_singularity): Likewise.
32378         (extract_range_from_comparison): Do not disable optimization
32379         when we assumed undefined overflow.
32380         (extract_range_basic): Remove init of unused var.
32382 2017-05-09  Richard Biener  <rguenther@suse.de>
32384         * tree-vrp.c (vrp_int_const_binop): Use wide-ints and simplify.
32385         (extract_range_from_multiplicative_op_1): Adjust.
32386         (extract_range_from_binary_expr_1): Use int_const_binop.
32388 2017-05-08  Kelvin Nilsen  <kelvin@gcc.gnu.org>
32390         PR target/80101
32391         * config/rs6000/power6.md: Replace store_data_bypass_p calls with
32392         rs6000_store_data_bypass_p in seven define_bypass directives and
32393         in several comments.
32394         * config/rs6000/rs6000-protos.h: Add prototype for
32395         rs6000_store_data_bypass_p function.
32396         * config/rs6000/rs6000.c (rs6000_store_data_bypass_p): New
32397         function implements slightly different (rs6000-specific) semantics
32398         than store_data_bypass_p, returning false rather than aborting
32399         with assertion error when arguments do not satisfy the
32400         requirements of store data bypass.
32401         (rs6000_adjust_cost): Replace six calls of store_data_bypass_p with
32402         rs6000_store_data_bypass_p.
32404 2017-05-08  Max Filippov  <jcmvbkbc@gmail.com>
32406         * config/xtensa/xtensa-protos.h
32407         (xtensa_initial_elimination_offset): New declaration.
32408         * config/xtensa/xtensa.c (xtensa_initial_elimination_offset):
32409         New function. Move its body from the INITIAL_ELIMINATION_OFFSET
32410         macro definition, add case for FRAME_POINTER_REGNUM when
32411         FRAME_GROWS_DOWNWARD.
32412         * config/xtensa/xtensa.h (FRAME_GROWS_DOWNWARD): New macro definition.
32413         (INITIAL_ELIMINATION_OFFSET): Replace body with call to
32414         xtensa_initial_elimination_offset.
32416 2017-05-08  Nathan Sidwell  <nathan@acm.org>
32418         * doc/invoke.texi: Alphabetize -fdump options.
32420 2017-05-08  Martin Sebor  <msebor@redhat.com>
32422         PR translation/80280
32423         * config/sol2-c.c (solaris_pragma_align): Correct quoting.
32425 2017-05-08  Bernd Edlinger  <bernd.edlinger@hotmail.de>
32427         * target.def (compute_frame_layout): New optional target hook.
32428         * doc/tm.texi.in (TARGET_COMPUTE_FRAME_LAYOUT): Add hook.
32429         * doc/tm.texi (TARGET_COMPUTE_FRAME_LAYOUT): Add documentation.
32430         * lra-eliminations.c (update_reg_eliminate): Call compute_frame_layout
32431         target hook.
32432         * reload1.c (verify_initial_elim_offsets): Likewise.
32433         * config/arm/arm.c (TARGET_COMPUTE_FRAME_LAYOUT): Define.
32434         (use_simple_return_p): Call arm_compute_frame_layout if needed.
32435         (arm_get_frame_offsets): Split up into this ...
32436         (arm_compute_frame_layout): ... and this function.
32438 2017-05-08  Richard Sandiford  <richard.sandiford@arm.com>
32440         * config/aarch64/constraints.md (Usa): New constraint.
32441         * config/aarch64/aarch64.md (*movsi_aarch64, *movdi_aarch64): Use it.
32443 2017-05-08  Thomas Preud'homme  <thomas.preudhomme@arm.com>
32445         * config.gcc (arm*-*-*): Set TM_MULTILIB_CONFIG from
32446         with_multilib_list after it has been checked.
32448 2017-05-08  Richard Biener  <rguenther@suse.de>
32450         * tree-ssa-pre.c (bitmap_set_and): Avoid bitmap copy.
32451         (bitmap_set_subtract_values): Likewise.
32453 2017-05-08  Richard Biener  <rguenther@suse.de>
32455         * tree-vrp.c (gimple_assign_nonzero_warnv_p): Rename to ...
32456         (gimple_assign_nonzero): ... this and remove strict_overflow_p
32457         argument.
32458         (gimple_stmt_nonzero_warnv_p): Rename to ...
32459         (gimple_stmt_nonzero_p): ... this and remove strict_overflow_p
32460         argument.
32461         (vrp_stmt_computes_nonzero): Remove strict_overflow_p argument.
32462         (extract_range_basic): Adjust, do not disable propagation on
32463         strict overflow sensitive simplification.
32464         (vrp_visit_cond_stmt): Likewise.
32466 2017-05-05  Jan Hubicka  <hubicka@ucw.cz>
32468         * ipa-inline-analysis.c (estimate_function_body_sizes): Recompute
32469         body size unconditionally.
32471 2017-05-07  Jeff Law  <law@redhat.com>
32473         Revert:
32474         2017-05-06  Jeff Law  <law@redhat.com>
32475         PR tree-optimization/78496
32476         * tree-vrp.c (simplify_assert_expr_using_ranges): Remove debugging
32477         code.
32479         PR tree-optimization/78496
32480         * tree-vrp.c (simplify_assert_expr_using_ranges): New function.
32481         (simplify_stmt_using_ranges): Call it.
32482         (vrp_dom_walker::before_dom_children): Extract equivalences
32483         from an ASSERT_EXPR with an equality comparison against a
32484         constant.
32486 2017-05-06  Jeff Law  <law@redhat.com>
32488         PR tree-optimization/78496
32489         * tree-vrp.c (simplify_assert_expr_using_ranges): Remove debugging
32490         code.
32492         PR tree-optimization/78496
32493         * tree-vrp.c (simplify_assert_expr_using_ranges): New function.
32494         (simplify_stmt_using_ranges): Call it.
32495         (vrp_dom_walker::before_dom_children): Extract equivalences
32496         from an ASSERT_EXPR with an equality comparison against a
32497         constant.
32499 2017-05-06  Richard Sandiford  <richard.sandiford@linaro.org>
32501         * lra-constraints.c (lra_copy_reg_equiv): New function.
32502         (split_reg): Use it to copy equivalence information from the
32503         original register to the spill register.
32505 2017-05-06  Richard Sandiford  <richard.sandiford@linaro.org>
32507         PR rtl-optimization/75964
32508         * simplify-rtx.c (simplify_const_relational_operation): Remove
32509         invalid handling of comparisons of integer ABS.
32511 2017-05-06  Uros Bizjak  <ubizjak@gmail.com>
32513         * config/i386/i386.c (ext_80387_constant_init): Do not explicitly
32514         initialize to zero.
32515         (init_regs): Remove declaration.
32516         (function_arg_advance_32): Initialize error_p as boolean variable.
32518 2017-05-05  Nathan Sidwell  <nathan@acm.org>
32520         * store-motion.c (remove_reachable_equiv_notes): Reformat long
32521         lines.  Use for (;;).
32523 2017-05-05  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
32525         * config/rs6000/rs6000.c (rs6000_vect_nonmem): New static var.
32526         (rs6000_init_cost): Initialize rs6000_vect_nonmem.
32527         (rs6000_add_stmt_cost): Update rs6000_vect_nonmem.
32528         (rs6000_finish_cost): Avoid vectorizing simple copy loops with
32529         VF=2 that require versioning.
32531 2017-05-05  David Malcolm  <dmalcolm@redhat.com>
32533         * diagnostic.h (CARET_LINE_MARGIN): Convert from macro to const
32534         int.
32536 2017-05-05  David Malcolm  <dmalcolm@redhat.com>
32538         * diagnostic.h (diagnostic_override_option_index): Convert from
32539         macro to inline function.
32541 2017-05-05  David Malcolm  <dmalcolm@redhat.com>
32543         * diagnostic.c (last_module_changed_p): New function.
32544         (set_last_module): New function.
32545         (diagnostic_report_current_module): Convert macro usage to
32546         the above functions.
32547         * diagnostic.h (diagnostic_context::last_module): Strengthen
32548         from const line_map * to const line_map_ordinary *.
32549         (diagnostic_last_module_changed): Delete macro.
32550         (diagnostic_set_last_module): Delete macro.
32552 2017-05-05  David Malcolm  <dmalcolm@redhat.com>
32554         * diagnostic.c (diagnostic_impl): Replace report_diagnostic
32555         with diagnostic_report_diagnostic.
32556         (diagnostic_n_impl_richloc): Likewise.
32557         * diagnostic.h (report_diagnostic): Delete macro.
32558         * rtl-error.c (diagnostic_for_asm): Replace report_diagnostic
32559         with diagnostic_report_diagnostic.
32560         * substring-locations.c (format_warning_va): Likewise.
32562 2017-05-05  David Malcolm  <dmalcolm@redhat.com>
32564         * diagnostic.c (diagnostic_report_diagnostic): Eliminate
32565         save/restor of format_spec.  Move option-printing code to...
32566         (print_option_information): ...this new function, and
32567         reimplement by simply printing to the pretty_printer,
32568         rather than appending to the format string.
32570 2017-05-05  David Malcolm  <dmalcolm@redhat.com>
32572         * diagnostic.c (diagnostic_report_diagnostic): Split out pragma
32573         handling logic into...
32574         (update_effective_level_from_pragmas): ...this new function.
32576 2017-05-04  Andrew Waterman  <andrew@sifive.com>
32578         * config/riscv/riscv.opt (mstrict-align): New option.
32579         * config/riscv/riscv.h (STRICT_ALIGNMENT): Use it.  Update comment.
32580         (SLOW_UNALIGNED_ACCESS): Define.
32581         (riscv_slow_unaligned_access): Declare.
32582         * config/riscv/riscv.c (riscv_tune_info): Add slow_unaligned_access
32583         field.
32584         (riscv_slow_unaligned_access): New variable.
32585         (rocket_tune_info): Set slow_unaligned_access to true.
32586         (optimize_size_tune_info): Set slow_unaligned_access to false.
32587         (riscv_cpu_info_table): Add entry for optimize_size_tune_info.
32588         (riscv_valid_lo_sum_p): Use TARGET_STRICT_ALIGN.
32589         (riscv_option_override): Set riscv_slow_unaligned_access.
32590         * doc/invoke.texi: Add -mstrict-align to RISC-V.
32592 2017-05-04  Kito Cheng  <kito.cheng@gmail.com>
32594         * config/riscv/riscv.md: Unify indentation.
32596 2017-05-05  Michael Meissner  <meissner@linux.vnet.ibm.com>
32598         PR target/79038
32599         PR target/79202
32600         PR target/79203
32601         * config/rs6000/rs6000.md (u code attribute): Add FIX and
32602         UNSIGNED_FIX.
32603         (extendsi<mode>2): Add support for doing sign extension via
32604         VUPKHSW and XXPERMDI if the value is in Altivec registers and we
32605         don't have ISA 3.0 instructions.
32606         (extendsi<mode>2 splitter): Likewise.
32607         (fix_trunc<mode>si2): If we are at ISA 2.07 (VSX small integer),
32608         generate the normal insns since SImode can now go in vector
32609         registers.  Disallow the special UNSPECs needed for previous
32610         machines to hide SImode being used.  Add new insns
32611         fctiw{,w}_<mode>_smallint if SImode can go in vector registers.
32612         (fix_trunc<mode>si2_stfiwx): Likewise.
32613         (fix_trunc<mode>si2_internal): Likewise.
32614         (fixuns_trunc<mode>si2): Likewise.
32615         (fixuns_trunc<mode>si2_stfiwx): Likewise.
32616         (fctiw<u>z_<mode>_smallint): Likewise.
32617         (fctiw<u>z_<mode>_mem): New combiner pattern to prevent conversion
32618         of floating point to 32-bit integer from doing a direct move to
32619         the GPR registers to do a store.
32620         (fctiwz_<mode>): Break long line.
32622 2017-05-05  Bin Cheng  <bin.cheng@arm.com>
32624         * Makefile.in (GTFILES): Add tree-ssa-loop-ivopts.c.
32625         * tree-ssa-loop-ivopts.c (compute_max_addr_offset): Delete.
32626         (addr_list, addr_offset_valid_p): New.
32627         (split_address_groups): Check offset validity with above function.
32628         (gt-tree-ssa-loop-ivopts.h): Include header file.
32630 2017-05-05  Nathan Sidwell  <nathan@acm.org>
32632         * config.gcc (arm*-*-*): Add missing 'fi'.
32634 2017-05-05  Steve Ellcey  <sellcey@cavium.com>
32636         * doc/invoke.texi (-fopt-info): Explicitly say order of options
32637         included in -fopt-info does not matter.
32638         * doc/optinfo.texi (-fopt-info): Fix description of default
32639         behavour. Explicitly say order of options included in -fopt-info
32640         does not matter.
32642 2017-05-05  Thomas Preud'homme  <thomas.preudhomme@arm.com>
32644         * config.gcc: Allow combinations of aprofile and rmprofile values for
32645         --with-multilib-list.
32646         * config/arm/t-multilib: New file.
32647         * config/arm/t-aprofile: Remove initialization of MULTILIB_*
32648         variables.  Remove setting of ISA and floating-point ABI in
32649         MULTILIB_OPTIONS and MULTILIB_DIRNAMES.  Set architecture and FPU in
32650         MULTI_ARCH_OPTS_A and MULTI_ARCH_DIRS_A rather than MULTILIB_OPTIONS
32651         and MULTILIB_DIRNAMES respectively.  Add comment to introduce all
32652         matches.  Add architecture matches for marvel-pj4 and generic-armv7-a
32653         CPU options.
32654         * config/arm/t-rmprofile: Likewise except for the matches changes.
32655         * doc/install.texi (--with-multilib-list): Document the combination of
32656         aprofile and rmprofile values and warn about pitfalls in doing that.
32658 2017-05-05  Wilco Dijkstra  <wdijkstr@arm.com>
32660         * config/aarch64/aarch64.md (movsi_aarch64): Remove '*' from r=w.
32661         (movdi_aarch64): Likewise.
32663 2017-05-05  Jakub Jelinek  <jakub@redhat.com>
32665         PR tree-optimization/80632
32666         * tree-switch-conversion.c (struct switch_conv_info): Add target_vop
32667         field.
32668         (build_arrays): Initialize it for virtual phis.
32669         (fix_phi_nodes): Use it for virtual phis.
32671         PR tree-optimization/80558
32672         * tree-vrp.c (extract_range_from_binary_expr_1): Optimize
32673         [x, y] op z into [x op, y op z] for op & or | if conditions
32674         are met.
32676 2017-05-05  Andre Vieira  <andre.simoesdiasvieira@arm.com>
32677             Prakhar Bahuguna  <prakhar.bahuguna@arm.com>
32679         PR target/71607
32680         * config/arm/arm.md (use_literal_pool): Remove.
32681         (64-bit immediate split): No longer takes cost into consideration
32682         if arm_disable_literal_pool is enabled.
32683         * config/arm/arm.c (arm_tls_referenced_p): Add diagnostic if TLS is
32684         used when arm_disable_literal_pool is enabled.
32685         (arm_max_const_double_inline_cost): Remove use of
32686         arm_disable_literal_pool.
32687         (push_minipool_fix): Add assert.
32688         (arm_reorg): Add return if arm_disable_literal_pool is enabled.
32689         * config/arm/vfp.md (no_literal_pool_df_immediate): New.
32690         (no_literal_pool_sf_immediate): New.
32692 2017-05-05  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
32694         PR tree-optimization/80613
32695         * tree-ssa-dce.c (propagate_necessity): Remove cases for
32696         BUILT_IN_STRDUP and BUILT_IN_STRNDUP.
32698 2017-05-05  Richard Biener  <rguenther@suse.de>
32700         * tree-ssa-pre.c (get_or_alloc_expr_for): Simplify.
32702 2017-05-05  Georg-Johann Lay  <avr@gjlay.de>
32704         * config/avr/avr.md [flag_strict_overflow]: Remove any occurence
32705         of this flag from insn conditions due to removal from r247495.
32707 2017-05-05  Wilco Dijkstra  <wdijkstr@arm.com>
32709         * config/arm/aarch-common.c (arm_early_load_addr_dep_ptr):
32710         New function.
32711         (arm_early_store_addr_dep_ptr): Likewise.
32712         * config/arm/aarch-common-protos.h
32713         (arm_early_load_addr_dep_ptr): Add prototype.
32714         (arm_early_store_addr_dep_ptr): Likewise.
32715         * config/arm/cortex-a53.md: Add new bypasses.
32717 2017-05-05  Jakub Jelinek  <jakub@redhat.com>
32719         * tree.c (next_type_uid): Change type to unsigned.
32720         (type_hash_canon): Decrement back next_type_uid if
32721         freeing a type node with the highest TYPE_UID.  For INTEGER_TYPEs
32722         also ggc_free TYPE_MIN_VALUE, TYPE_MAX_VALUE and TYPE_CACHED_VALUES
32723         if possible.
32725 2017-05-04  Martin Sebor  <msebor@redhat.com>
32727         * builtins.c: Fix a trivial typo in a comment.
32729         PR middle-end/79234
32730         * builtins.c (check_sizes): Adjust to handle reading past the end.
32731         Avoid printing excessive upper bound of ranges.  Use %E to print
32732         tree nodes instead of converting them to %wu.
32733         (expand_builtin_memchr): New function.
32734         (compute_dest_size): Rename...
32735         (compute_objsize): ...to this.
32736         (expand_builtin_memcpy): Adjust.
32737         (expand_builtin_mempcpy): Adjust.
32738         (expand_builtin_strcat): Adjust.
32739         (expand_builtin_strcpy): Adjust.
32740         (check_strncat_sizes): Adjust.
32741         (expand_builtin_strncat): Adjust.
32742         (expand_builtin_strncpy): Adjust and simplify.
32743         (expand_builtin_memset): Adjust.
32744         (expand_builtin_bzero): Adjust.
32745         (expand_builtin_memcmp): Adjust.
32746         (expand_builtin): Handle memcmp.
32747         (maybe_emit_chk_warning): Check strncat just once.
32749 2017-05-04  Martin Sebor  <msebor@redhat.com>
32751         PR preprocessor/79214
32752         PR middle-end/79222
32753         PR middle-end/79223
32754         * builtins.c (check_sizes): Add inlining context and issue
32755         warnings even when -Wno-system-headers is set.
32756         (check_strncat_sizes): Same.
32757         (expand_builtin_strncat): Same.
32758         (expand_builtin_memmove): New function.
32759         (expand_builtin_stpncpy): Same.
32760         (expand_builtin): Handle memmove and stpncpy.
32762 2017-05-04  Bin Cheng  <bin.cheng@arm.com>
32764         * tree-ssa-loop-ivopts.c (struct cost_pair): Remove field inv_expr
32765         which is not used any more.
32767 2017-05-04  Wilco Dijkstra  <wdijkstr@arm.com>
32769         * config/aarch64/aarch64.c (generic_tunings): Update prefetch model.
32771 2017-05-04  Wilco Dijkstra  <wdijkstr@arm.com>
32773         * config/aarch64/aarch64.c (cortexa35_tunings): Set jump alignment to 4.
32774         (cortexa53_tunings): Likewise.
32775         (cortexa57_tunings): Likewise.
32776         (cortexa72_tunings): Likewise.
32777         (cortexa73_tunings): Likewise.
32779 2017-05-04  Wilco Dijkstra  <wdijkstr@arm.com>
32781         * config/aarch64/aarch64.c (generic_tunings): Set jump alignment to 4.
32782         Set loop alignment to 8.
32784 2017-05-04  Martin Sebor  <msebor@redhat.com>
32786         PR translation/80280
32787         * builtins.c (expand_builtin_object_size): Add missing quoting to
32788         %D and like directives.
32789         * hsa-gen.c (hsa_type_for_scalar_tree_type): Same.
32790         (hsa_type_for_tree_type): Same.
32791         (verify_function_arguments): Same.
32792         * symtab.c (symbol_table::change_decl_assembler_name): Same.
32793         * varasm.c (get_section): Same.
32794         (mark_weak): Same.
32796 2017-05-04  Martin Sebor  <msebor@redhat.com>
32798         PR translation/80280
32799         * config/i386/i386.c (ix86_function_versions): Quote a %D directive.
32801 2017-05-04  Wilco Dijkstra  <wdijkstr@arm.com>
32803         * config/aarch64/aarch64.c (generic_addrcost_table):
32804         Change HI/TI mode setting.
32806 2017-05-04  Martin Jambor  <mjambor@suse.cz>
32808         PR tree-optimization/80622
32809         * tree-sra.c (comes_initialized_p): New function.
32810         (build_accesses_from_assign): Only set write lazily when
32811         comes_initialized_p is false.
32812         (analyze_access_subtree): Use comes_initialized_p.
32813         (propagate_subaccesses_across_link): Assert !comes_initialized_p
32814         instead of testing for PARM_DECL.
32816 2017-05-04  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
32818         * config/aarch64/aarch64.md (prefetch); Adjust predicate and
32819         constraint on operand 0 to allow more general addressing modes.
32820         Adjust output template.
32821         * config/aarch64/aarch64.c (aarch64_address_valid_for_prefetch_p):
32822         New function.
32823         * config/aarch64/aarch64-protos.h
32824         (aarch64_address_valid_for_prefetch_p): Declare prototype.
32825         * config/aarch64/constraints.md (Dp): New address constraint.
32826         * config/aarch64/predicates.md (aarch64_prefetch_operand): New
32827         predicate.
32829 2017-05-04  Jan Hubicka  <hubicka@ucw.cz>
32831         * ipa-cp.c (perform_estimation_of_a_value): Drop base_time parameter;
32832         update use of estimate_ipcp_clone_size_and_time.
32833         (estimate_local_effects): Update use of
32834         estimate_ipcp_clone_size_and_time and perform_estimation_of_a_value.
32835         * ipa-inline.h (estimate_ipcp_clone_size_and_time): Update prototype.
32836         * ipa-inline-analysis.c (estimate_ipcp_clone_size_and_time):
32837         Return nonspecialized time.
32839 2017-05-04  Richard Biener  <rguenther@suse.de>
32841         * tree-ssa-alias.c (get_continuation_for_phi): Improve looking
32842         for the last VUSE which def dominates the PHI.  Directly call
32843         maybe_skip_until.
32844         (get_continuation_for_phi_1): Remove.
32846 2017-05-04  Richard Sandiford  <richard.sandiford@linaro.org>
32848         * tree-ssa-loop-manip.c (niter_for_unrolled_loop): Add commentary
32849         to explain the use of truncating division.  Cap the number of
32850         iterations to the maximum given by nb_iterations_upper_bound,
32851         if defined.
32853 2017-05-04  Thomas Preud'homme  <thomas.preudhomme@arm.com>
32855         * configure.ac (--enable-mingw-wildcard): Add new configurable feature.
32856         * configure: Regenerate.
32857         * config.in: Regenerate.
32858         * config/i386/driver-mingw32.c: new file.
32859         * config/i386/x-mingw32: Add rule to build driver-mingw32.o.
32860         * config.host: Link driver-mingw32.o on MinGW host.
32861         * doc/install.texi: Document new --enable-mingw-wildcard configure
32862         option.
32864 2017-05-04  Marek Polacek  <polacek@redhat.com>
32866         PR tree-optimization/80612
32867         * calls.c (get_size_range): Check for INTEGRAL_TYPE_P.
32869 2017-05-04  Prakhar Bahuguna  <prakhar.bahuguna@arm.com>
32870             Andre Simoes Dias Vieira  <andre.simoesdiasvieira@arm.com>
32872         * config/arm/arm.md (movsi): Change TARGET_32BIT to TARGET_HAVE_MOVT.
32873         (movt splitter): Likewise.
32874         * config/arm/arm.c (arm_option_check_internal): Change arm_arch_thumb2
32875         to TARGET_HAVE_MOVT, and merge with -mslow-flash-data check.
32876         (const_ok_for_arm): Change else to else if (TARGET_THUMB2) and add else
32877         block for Thumb-1 with MOVT.
32878         (thumb2_legitimate_address_p): Move code block ...
32879         (can_avoid_literal_pool_for_label_p): ... into this new function.
32880         (thumb1_legitimate_address_p): Add check for TARGET_HAVE_MOVT and
32881         literal pool.
32882         (thumb_legitimate_constant_p): Add conditional on TARGET_HAVE_MOVT
32883         * doc/invoke.texi (-mpure-code): Change "ARMv7-M targets" for
32884         "M-profile targets with the MOVT instruction".
32886 2017-05-04  Prakhar Bahuguna  <prakhar.bahuguna@arm.com>
32888         * config/arm/arm-builtins.c (arm_init_builtins): Rename
32889         __builtin_arm_ldfscr to __builtin_arm_get_fpscr, and rename
32890         __builtin_arm_stfscr to __builtin_arm_set_fpscr.
32892 2017-05-04  Martin Liska  <mliska@suse.cz>
32894         * tree-vrp.c (simplify_cond_using_ranges_2): Remove unused
32895         variable cond_code.
32897 2017-05-04  Richard Biener  <rguenther@suse.de>
32899         * tree.c (array_at_struct_end_p): Handle arrays at struct
32900         end with flexarrays more conservatively.  Refactor and treat
32901         arrays of arrays or aggregates more strict.  Fix
32902         VIEW_CONVERT_EXPR handling.  Remove allow_compref argument.
32903         * tree.c (array_at_struct_end_p): Adjust prototype.
32904         * emit-rtl.c (set_mem_attributes_minus_bitpos): Adjust.
32905         * gimple-fold.c (get_range_strlen): Likewise.
32906         * tree-chkp.c (chkp_may_narrow_to_field): Likewise.
32908 2017-05-04  Richard Biener  <rguenther@suse.de>
32910         PR tree-optimization/31130
32911         * tree-vrp.c (needs_overflow_infinity): Remove as always returning
32912         false.
32913         (supports_overflow_infinity): Likewise.
32914         (is_negative_overflow_infinity): Likewise.
32915         (is_positive_overflow_infinity): Likewise.
32916         (is_overflow_infinity): Likewise.
32917         (stmt_overflow_infinity): Likewise.
32918         (overflow_infinity_range_p): Likewise.
32919         (usable_range_p): Remove as always returning true.
32920         (make_overflow_infinity): Remove.
32921         (negative_overflow_infinity): Likewise.
32922         (positive_overflow_infinity): Likewise.
32923         (avoid_overflow_infinity): Likewise.
32924         (set_value_range): Adjust accordingly.
32925         (set_value_range_to_nonnegative): Likewise, remove now unused
32926         overflow_infinity arg.
32927         (vrp_operand_equal_p): Adjust.
32928         (update_value_range): Likewise.
32929         (range_int_cst_singleton_p): Likewise.
32930         (operand_less_p): Likewise.
32931         (compare_values_warnv): Likewise.
32932         (extract_range_for_var_from_comparison_expr): Likewise.
32933         (vrp_int_const_binop): Likewise.
32934         (zero_nonzero_bits_from_vr): Likewise.
32935         (extract_range_from_multiplicative_op_1): Likewise.
32936         (extract_range_from_binary_expr_1): Likewise.
32937         (extract_range_from_unary_expr): Likewise.
32938         (extract_range_from_comparison): Likewise.
32939         (extract_range_basic): Likewise.
32940         (adjust_range_with_scev): Likewise.
32941         (compare_ranges): Likewise.
32942         (compare_range_with_value): Likewise.
32943         (dump_value_range): Likewise.
32944         (test_for_singularity): Likewise, remove strict_overflow_p parameter
32945         never used.
32946         (simplify_cond_using_ranges): Adjust.
32948 2017-05-04  Pekka Jääskeläinen  <pekka.jaaskelainen@parmance.com>
32950         * brig-builtins.def: Added a builtin for class_f64.
32951         * builtin-types.def: Added a builtin type needed by class_f64.
32953 2017-05-03  Jason Merrill  <jason@redhat.com>
32955         * timevar.def: Add TV_CONSTEXPR.
32957 2017-05-03  David Malcolm  <dmalcolm@redhat.com>
32959         * common.opt (fdiagnostics-parseable-fixits): Fix typo.
32961 2017-05-03  Martin Jambor  <mjambor@suse.cz>
32963         * ipa-prop.c (ipa_update_after_lto_read): Removed.
32964         * ipa-prop.h (ipa_update_after_lto_read): Remove declaration.
32965         * ipa-cp.c (ipcp_propagate_stage): Do not call
32966         ipa_update_after_lto_read.
32967         * ipa-inline.c (ipa_inline): Likewise.
32969 2017-05-03  Martin Jambor  <mjambor@suse.cz>
32971         * ipa-prop.h (ipa_edge_args): Make a class.  Mark with for_user GTY
32972         tag.  Added a default constructor and a destructor.
32973         (ipa_edge_args_sum_t): New class;
32974         (ipa_edge_args_sum): Declare.
32975         (ipa_edge_args_vector): Remove declaration.
32976         (IPA_EDGE_REF): Use ipa_edge_args_sum.
32977         (ipa_free_edge_args_substructures): Remove declaration.
32978         (ipa_check_create_edge_args): Use ipa_edge_args_sum.
32979         (ipa_edge_args_info_available_for_edge_p): Likewise.
32980         * ipa-prop.c (ipa_edge_args_vector): Removed.
32981         (edge_removal_hook_holder): Likewise.
32982         (edge_duplication_hook_holder): Likewise.
32983         (ipa_edge_args_sum): New variable.
32984         (ipa_propagate_indirect_call_infos): Test ipa_edge_args_sum instead of
32985         ipa_edge_args_vector.
32986         (ipa_free_edge_args_substructures): Likewise.
32987         (ipa_free_all_edge_args): Free ipa_edge_args_sum instead of
32988         ipa_edge_args_vector.
32989         (ipa_edge_removal_hook): Turned into method
32990         ipa_edge_args_sum_t::remove.
32991         (ipa_edge_duplication_hook): Turned into method
32992         ipa_edge_args_sum_t::duplicate.
32993         (ipa_register_cgraph_hooks): Create ipa_edge_args_sum instead of
32994         registering edge hooks.
32995         (ipa_unregister_cgraph_hooks): Do not unregister edge hooks.
32996         * ipa-inline-analysis.c (estimate_function_body_sizes): Test
32997         ipa_edge_args_sum instead of ipa_edge_args_vector.
32998         * ipa-profile.c (ipa_profile): Likewise.
33000 2017-05-03  Martin Jambor  <mjambor@suse.cz>
33002         * symbol-summary.h (function_summary): New method exists.
33003         (function_summary::symtab_removal): Deallocate through release.
33004         (call_summary): New class.
33005         (gt_ggc_mx): New overload.
33006         (gt_pch_nx): Likewise.
33007         (gt_pch_nx): Likewise.
33009 2017-05-03  Jeff Law  <law@redhat.com>
33011         PR tree-optimization/78496
33012         * tree-vrp.c (simplify_cond_using_ranges_1): Renamed
33013         from simplify_cond_using_ranges.  Split off code to walk
33014         backwards through casts into ...
33015         (simplify_cond_using_ranges_2): New function.
33016         (simplify_stmt_using_ranges): Call simplify_cond_using_ranges_1.
33017         (execute_vrp): After identifying jump threads, call
33018         simplify_cond_using_ranges_2.
33020 2017-05-03  Jan Hubicka  <hubicka@ucw.cz>
33022         PR bootstrap/80609
33023         * ipa-inline.h (inline_summary): Add ctor.
33024         (create_ggc): Do not use ggc_cleared_alloc.
33026 2017-05-03  Jeff Downs  <heydowns@somuchpressure.net>
33027             Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
33029         * gcc.c (handle_braces): Support escaping in switch matching
33030         text.
33031         * doc/invoke.texi (Spec Files): Document it.
33032         Remove superfluous @code markup in items.
33034 2017-05-03  David Malcolm  <dmalcolm@redhat.com>
33036         * diagnostic-show-locus.c (struct column_range): New struct.
33037         (get_affected_columns): New function.
33038         (get_printed_columns): New function.
33039         (struct correction): New struct.
33040         (correction::ensure_capacity): New function.
33041         (correction::ensure_terminated): New function.
33042         (struct line_corrections): New struct.
33043         (line_corrections::~line_corrections): New dtor.
33044         (line_corrections::add_hint): New function.
33045         (layout::print_trailing_fixits): Reimplement in terms of the new
33046         classes.
33047         (selftest::test_overlapped_fixit_printing): New function.
33048         (selftest::diagnostic_show_locus_c_tests): Call it.
33050 2017-05-03  Nathan Sidwell  <nathan@acm.org>
33052         Canonicalize canonical type hashing
33053         * tree.h (type_hash_canon_hash): Declare.
33054         * tree.c (type_hash_list, attribute_hash_list): Move into
33055         type_hash_canon_hash.
33056         (build_type_attribute_qual_variant): Break out hash code calc into
33057         type_hash_canon_hash.
33058         (type_hash_canon_hash): New.  Generic type hash computation.
33059         (build_range_type_1, build_array_type_1, build_function_type,
33060         build_method_type_directly, build_offset_type, build_complex_type,
33061         make_vector_type): Call it.
33063 2017-05-03  Richard Biener  <rguenther@suse.de>
33065         * tree-vect-data-refs.c (vect_enhance_data_refs_alignment):
33066         When all DRs have unknown misaligned do not always peel
33067         when there is a store but apply the same costing model as if
33068         there were only loads.
33070 2017-05-03  Richard Biener  <rguenther@suse.de>
33072         Revert
33073         PR tree-optimization/80492
33074         * tree-ssa-alias.c (decl_refs_may_alias_p): Handle
33075         compare_base_decls returning dont-know properly.
33077 2017-05-03  Thomas Preud'homme  <thomas.preudhomme@arm.com>
33079         * config/arm/iterators.md (CCSI): New mode iterator.
33080         (arch): New mode attribute.
33081         * config/arm/sync.md (atomic_compare_and_swap<mode>_1): Rename into ...
33082         (atomic_compare_and_swap<CCSI:arch><NARROW:mode>_1): This and ...
33083         (atomic_compare_and_swap<CCSI:arch><SIDI:mode>_1): This.  Use CCSI
33084         code iterator for success result mode.
33085         * config/arm/arm.c (arm_expand_compare_and_swap): Adapt code to use
33086         the corresponding new insn generators.
33088 2017-05-03  Bin Cheng  <bin.cheng@arm.com>
33090         Revert r247509
33091         2017-05-02  Bin Cheng  <bin.cheng@arm.com>
33092         * rtlanal.c (rtx_cost): Handle TRUNCATE between tieable modes.
33094 2017-05-03  Richard Sandiford  <richard.sandiford@linaro.org>
33096         * tree-data-ref.h (SUB_CONFLICTS_IN_A): Wrap SUB argument in brackets.
33097         (SUB_CONFLICTS_IN_B, SUB_LAST_CONFLICT, SUB_DISTANCE): Likewise.
33098         (DDR_A): Wrap DDR argument in brackets.
33099         (DDR_B, DDR_AFFINE_P, DDR_ARE_DEPENDENT, DDR_SUBSCRIPTS): Likewise.
33100         (DDR_LOOP_NEST, DDR_INNER_LOOP, DDR_SELF_REFERENCE): Likewise.
33101         (DDR_REVERSED_P): Likewise.
33103 2017-05-03  Jakub Jelinek  <jakub@redhat.com>
33105         PR tree-optimization/79472
33106         * tree-switch-conversion.c (struct switch_conv_info): Add
33107         contiguous_range and default_case_nonstandard fields.
33108         (collect_switch_conv_info): Compute contiguous_range and
33109         default_case_nonstandard fields, don't clear final_bb if
33110         contiguous_range and only the default case doesn't have the required
33111         structure.
33112         (check_all_empty_except_final): Set default_case_nonstandard instead
33113         of failing if contiguous_range and the default case doesn't have empty
33114         block.
33115         (check_final_bb): Add SWTCH argument, don't fail if contiguous_range
33116         and only the default case doesn't have the required constants.  Skip
33117         virtual phis.
33118         (gather_default_values): Skip virtual phis.  Allow non-NULL CASE_LOW
33119         if default_case_nonstandard.
33120         (build_constructors): Build constant 1 just once.  Assert that default
33121         values aren't inserted in between cases if contiguous_range.  Skip
33122         virtual phis.
33123         (build_arrays): Skip virtual phis.
33124         (prune_bbs): Add DEFAULT_BB argument, don't remove that bb.
33125         (fix_phi_nodes): Don't add e2f phi arg if default_case_nonstandard.
33126         Handle virtual phis.
33127         (gen_inbound_check): Handle default_case_nonstandard case.
33128         (process_switch): Adjust check_final_bb caller.  Call
33129         gather_default_values with the first non-default case instead of
33130         default case if default_case_nonstandard.
33132 2017-05-02  Nathan Sidwell  <nathan@acm.org>
33134         * ggc-page.c (move_ptes_to_front): Replace unsigned >0 with i--
33135         check.  Fix formatting.
33137 2017-05-02  Jan Hubicka  <hubicka@ucw.cz>
33139         * ipa-inline-analysis.c (estimate_node_size_and_time): Allow roundoff
33140         errors when comparing specialized and unspecialized times.
33142 2017-05-02  David Malcolm  <dmalcolm@redhat.com>
33144         * diagnostic-show-locus.c
33145         (layout::should_print_annotation_line_p): Make private.
33146         (layout::print_annotation_line): Make private.
33147         (layout::annotation_line_showed_range_p): Make private.
33148         (layout::show_ruler): Make private.
33149         (layout::print_source_line): Make private.  Pass in line and
33150         line_width, rather than calling location_get_source_line.  Drop
33151         returned value.
33152         (layout::print_leading_fixits): New method.
33153         (layout::print_any_fixits): Rename to...
33154         (layout::print_trailing_fixits): ...this, and make private.
33155         Don't print newline fixits.
33156         (diagnostic_show_locus): Move logic for printing one row into...
33157         (layout::print_line): ...this new function.  Move the
33158         location_get_source_line call and error-handling from
33159         print_source_line to here.  Call print_leading_fixits, and rename
33160         print_any_fixits to print_trailing_fixits.
33161         (selftest::test_fixit_insert_containing_newline): Update now that
33162         newlines are partially supported.
33163         (selftest::test_fixit_insert_containing_newline_2): New test.
33164         (selftest::test_fixit_replace_containing_newline): Update comments.
33165         (selftest::diagnostic_show_locus_c_tests): Call the new test.
33166         * edit-context.c (class added_line): New class.
33167         (class edited_line): Describe newline handling in comment.
33168         (edited_line::actually_edited_p): New method.
33169         (edited_line::print_content): Delete redundant decl.
33170         (edited_line::m_predecessors): New field.
33171         (edited_file::print_content): Call edited_line::print_content.
33172         (edited_file::print_diff): Update to support newlines.
33173         (edited_file::print_diff_hunk): Likewise.
33174         (edited_file::print_run_of_changed_lines): New function.
33175         (edited_file::print_diff_line): Convert to...
33176         (print_diff_line): ...this.
33177         (edited_file::get_effective_line_count): New function.
33178         (edited_line::edited_line): Initialize new field m_predecessors.
33179         (edited_line::~edited_line): Clean up m_predecessors.
33180         (edited_line::apply_fixit): Handle newlines.
33181         (edited_line::get_effective_line_count): New function.
33182         (edited_line::print_content): New function.
33183         (edited_line::print_diff_lines): New function.
33184         (selftest::test_applying_fixits_insert_containing_newline): New
33185         test.
33186         (selftest::test_applying_fixits_replace_containing_newline): New
33187         test.
33188         (selftest::insert_line): New function.
33189         (selftest::test_applying_fixits_multiple_lines): Add example of
33190         inserting a line.
33191         (selftest::edit_context_c_tests): Call the new tests.
33193 2017-05-02  Bin Cheng  <bin.cheng@arm.com>
33195         * tree-ssa-loop-ivopts.c (get_scaled_computation_cost_at): Delete
33196         parameter cand.  Update dump information.
33197         (get_computation_cost): Update uses.
33199 2017-05-02  Bin Cheng  <bin.cheng@arm.com>
33201         * tree-ssa-loop-ivopts.c (get_computation_aff_1): New.
33202         (get_computation_aff): Reorder parameters.  Use get_computation_aff_1.
33203         (get_computation_at, rewrite_use_address): Update use of
33204         get_computation_aff.
33206 2017-05-02  Bin Cheng  <bin.cheng@arm.com>
33208         * tree-ssa-loop-ivopts.c (get_computation_at): Reorder parameters.
33209         (get_computation): Delete.
33210         (get_computation_cost): Implement like get_computation_cost_at.
33211         Use get_computation_at.
33212         (get_computation_cost_at): Delete.
33213         (rewrite_use_nonlinear_expr): Use get_computation_at.
33214         (rewrite_use_compare, remove_unused_ivs): Ditto.
33216 2017-05-02  Bin Cheng  <bin.cheng@arm.com>
33218         * tree-ssa-loop-ivopts.c (rewrite_use_address): Simple refactor.
33220 2017-05-02  Bin Cheng  <bin.cheng@arm.com>
33222         * tree-ssa-loop-ivopts.c (struct iv_ca): Rename n_regs to n_invs.
33223         (ivopts_global_cost_for_size): Rename parameter and update uses.
33224         (iv_ca_recount_cost): Update uses.
33225         (iv_ca_set_remove_invs, iv_ca_set_no_cp): Record invariants and
33226         candidates seperately in n_invs and n_cands.
33227         (iv_ca_set_add_invs, iv_ca_set_cp, iv_ca_new): Ditto.
33229 2017-05-02  Bin Cheng  <bin.cheng@arm.com>
33231         * tree-ssa-loop-ivopts.c (struct walk_tree_data): New.
33232         (find_inv_vars_cb): New.
33233         (find_depends): Renamed to ...
33234         (find_inv_vars): ... this.
33235         (add_candidate_1, force_var_cost): Call find_inv_vars.
33236         (split_address_cost, determine_group_iv_cost_cond): Ditto.
33238 2017-05-02  Bin Cheng  <bin.cheng@arm.com>
33240         * tree-ssa-loop-ivopts.c (struct cost_pair): Rename depends_on to
33241         inv_vars.  Add inv_exprs.
33242         (struct iv_cand): Rename depends_on to inv_vars.
33243         (struct ivopts_data): Rename max_inv_id/n_invariant_uses to
33244         max_inv_var_id/n_inv_var_uses.  Move max_inv_expr_id around.
33245         Refactor field used_inv_exprs from has_map to array n_inv_expr_uses.
33246         (dump_cand): Dump inv_vars.
33247         (tree_ssa_iv_optimize_init): Support inv_vars and inv_exprs.
33248         (record_invariant, find_depends, add_candidate_1): Ditto.
33249         (set_group_iv_cost, force_var_cost): Ditto.
33250         (split_address_cost, ptr_difference_cost, difference_cost): Ditto.
33251         (get_computation_cost_at, get_computation_cost): Ditto.
33252         (determine_group_iv_cost_generic): Ditto.
33253         (determine_group_iv_cost_address): Ditto.
33254         (determine_group_iv_cost_cond, autoinc_possible_for_pair): Ditto.
33255         (determine_group_iv_costs): Ditto.
33256         (iv_ca_recount_cost): Update call to ivopts_global_cost_for_size.
33257         (iv_ca_set_remove_invariants): Renamed to ...
33258         (iv_ca_set_remove_invs): ... this.  Support inv_vars and inv_exprs.
33259         (iv_ca_set_no_cp): Use iv_ca_set_remove_invs.
33260         (iv_ca_set_add_invariants):  Renamed to ...
33261         (iv_ca_set_add_invs): ... this.  Support inv_vars and inv_exprs.
33262         (iv_ca_set_cp): Use iv_ca_set_add_invs.
33263         (iv_ca_has_deps): Support inv_vars and inv_exprs.
33264         (iv_ca_new, iv_ca_free, iv_ca_dump, free_loop_data): Ditto.
33265         (create_new_ivs): Remove useless dump.
33267 2017-05-02  Bin Cheng  <bin.cheng@arm.com>
33269         * tree-ssa-loop-ivopts.c (get_computation_cost_at): Remove pseudo
33270         iv_cand code.
33271         (determine_group_iv_cost_cond, determine_iv_cost): Ditto.
33272         (iv_ca_set_no_cp, create_new_iv): Ditto.
33274 2017-05-02  Bin Cheng  <bin.cheng@arm.com>
33276         * rtlanal.c (rtx_cost): Handle TRUNCATE between tieable modes.
33278 2017-05-02  Xi Ruoyao  <ryxi@stu.xidian.edu.cn>
33280         * tree.h (EXPR_CILK_SPAWN): Use macro TREE_CHECK2 instead of
33281         function tree_check2.
33283 2017-05-02  Martin Liska  <mliska@suse.cz>
33285         * doc/gcov.texi: Add missing preposition.
33286         * gcov.c (function_info::function_info): Properly fill up
33287         all member variables.
33289 2017-05-02  Tamar Christina  <tamar.christina@arm.com>
33291         * expr.c (expand_expr_real_2): Re-cost if previous costs are the same.
33293 2017-05-02  Tamar Christina  <tamar.christina@arm.com>
33295         * simplify-rtx.c (simplify_binary_operation_1): Add LSHIFTRT case.
33297 2017-05-02  Martin Liska  <mliska@suse.cz>
33299         PR lto/77954.
33300         * lto-streamer-in.c (lto_read_tree_1): Remove
33301         LTO_STREAMER_DEBUG.
33302         * lto-streamer.c (struct tree_hash_entry): Likewise.
33303         (struct tree_entry_hasher): Likewise.
33304         (tree_entry_hasher::hash): Likewise.
33305         (tree_entry_hasher::equal): Likewise.
33306         (lto_streamer_init): Likewise.
33307         (lto_orig_address_map): Likewise.
33308         (lto_orig_address_get): Likewise.
33309         (lto_orig_address_remove): Likewise.
33310         * lto-streamer.h: Likewise.
33311         * tree-streamer-in.c (streamer_alloc_tree): Likewise.
33312         * tree-streamer-out.c (streamer_write_tree_header): Likewise.
33314 2017-05-02  Sebastian Peryt  <sebastian.peryt@intel.com>
33316         * config/i386/avx512fintrin.h (_mm_mask_add_round_sd)
33317         (_mm_maskz_add_round_sd, _mm_mask_add_round_ss)
33318         (mm_maskz_add_round_ss, _mm_mask_sub_round_sd)
33319         (mm_maskz_sub_round_sd, _mm_mask_sub_round_ss)
33320         (mm_maskz_sub_round_ss, _mm_mask_add_sd)
33321         (mm_maskz_add_sd, _mm_mask_add_ss, _mm_maskz_add_ss)
33322         (mm_mask_sub_sd, _mm_maskz_sub_sd, _mm_mask_sub_ss)
33323         (mm_maskz_sub_ss): New intrinsics.
33324         * config/i386/i386-builtin-types.def (V2DF_FTYPE_V2DF_V2DF_V2DF_UQI_INT)
33325         (V4SF_FTYPE_V4SF_V4SF_V4SF_UQI_INT): New function type aliases.
33326         * config/i386/i386-builtin.def (__builtin_ia32_addsd_mask_round)
33327         (__builtin_ia32_addss_mask_round, __builtin_ia32_subsd_mask_round)
33328         (__builtin_ia32_subss_mask_round): New builtins.
33329         * config/i386/i386.c (V2DF_FTYPE_V2DF_V2DF_V2DF_UQI_INT)
33330         (V4SF_FTYPE_V4SF_V4SF_V4SF_UQI_INT): Handle new types.
33331         * config/i386/sse.md (<sse>_vm<plusminus_insn><mode>3<round_name>):
33332         Renamed to ...
33333         (<sse>_vm<plusminus_insn><mode>3<mask_name><round_name>): ... this.
33334         (v<plusminus_mnemonic><ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|%0, %1, %<iptr>2<round_op3>}):
33335         Changed to ...
33336         (v<plusminus_mnemonic><ssescalarmodesuffix>\t{<round_mask_op3>%2, %1, %0<mask_operand3>|%0<mask_operand3>, %1, %<iptr>2<round_mask_op3>}):
33337         ... this.
33339 2017-05-02  Martin Jambor  <mjambor@suse.cz>
33341         PR tree-optimization/78687
33342         * tree-sra.c (access): New field parent.
33343         (process_subtree_disqualification): New function.
33344         (disqualify_candidate): Call it.
33345         (build_accesses_from_assign): Reset write flag if creating an
33346         assighnment link.
33347         (build_access_subtree): Fill in parent field and also prpagate
33348         down grp_write flag.
33349         (create_artificial_child_access): New parameter set_grp_write, set
33350         grp_write to its value.
33351         (propagate_subaccesses_across_link): Also propagate grp_write flag
33352         values.
33353         (propagate_all_subaccesses): Push the closest parent back to work
33354         queue if add_access_to_work_queue returned true.
33356 2017-05-02  Richard Biener  <rguenther@suse.de>
33358         * common.opt (fstrict-overflow): Alias negative to fwrapv.
33359         * doc/invoke.texi (fstrict-overflow): Remove all traces of
33360         -fstrict-overflow documentation.
33361         * tree.h (TYPE_OVERFLOW_UNDEFINED): Do not test flag_strict_overflow.
33362         (POINTER_TYPE_OVERFLOW_UNDEFINED): Test !flag_wrapv instead of
33363         flag_strict_overflow.
33364         * ipa-inline.c (can_inline_edge_p): Do not test flag_strict_overflow.
33365         * lto-opts.c (lto_write_options): Do not stream it.
33366         * lto-wrapper.c (merge_and_complain): Do not handle it.
33367         * opts.c (default_options_table): Do not set -fstrict-overflow.
33368         (finish_options): Likewise do not clear it when sanitizing.
33369         * simplify-rtx.c (simplify_const_relational_operation): Do not
33370         test flag_strict_overflow.
33372 2017-05-02  Uros Bizjak  <ubizjak@gmail.com>
33374         * config/alpha/alpha.md (*add<mode>3_ieee): Merge to add<mode>3
33375         using enabled attribute.
33376         (*sub<mode>3_ieee): Merge to sub<mode>3 using enabled attribute.
33377         (*mul<mode>3_ieee): Merge to mul<mode>3 using enabled attribute.
33378         (*div<mode>3_ieee): Merge to div<mode>3 using enabled attribute.
33379         (*sqrt<mode>2_ieee): Merge to sqrt<mode>2 using enabled attribute.
33380         (*fix_truncdfdi_ieee): Merge to *fix_truncdfdi2 using enabled attribute.
33381         (*fix_truncsfdi_ieee): Merge to *fix_truncsfdi2 using enabled attribute.
33382         (*floatdisf_ieee): Merge to floatdisf2 using enabled attribute.
33383         (*floatdidf_ieee): Merge to floatdidf2 using enabled attribute.
33384         (*truncdfsf2_ieee): Merge to truncdfsf2 using enabled attribute.
33385         (*cmpdf_ieee): Merge to *cmpdf_internal using enabled attribute.
33387 2017-05-02  Uros Bizjak  <ubizjak@gmail.com>
33389         * config/i386/i386.c (ix86_code_end): Use {FIRST,LAST}_INT_REG.
33391 2017-05-02  Richard Biener  <rguenther@suse.de>
33393         PR tree-optimization/80591
33394         Revert
33395         2017-04-10  Richard Biener  <rguenther@suse.de>
33397         * tree-ssa-structalias.c (find_func_aliases): Properly handle
33398         asm inputs.
33400 2017-05-02  Richard Biener  <rguenther@suse.de>
33402         PR tree-optimization/80549
33403         * tree-cfgcleanup.c (mfb_keep_latches): New helper.
33404         (cleanup_tree_cfg_noloop): Create forwarders to known loop
33405         headers if they do not have a preheader.
33407 2017-05-02  Martin Liska  <mliska@suse.cz>
33409         PR other/80589
33410         * common.opt: Fix typo.
33411         * doc/invoke.texi: Likewise.
33413 2017-05-01  Jan Beulich  <jbeulich@suse.com>
33415         * config/i386/sse.md (xop_vpermil2<mode>3): Do not allow operand
33416         swapping, add (x,x,m,x,n) alternative.
33418 2017-05-01  Nathan Sidwell  <nathan@acm.org>
33420         * calls.c (combine_pending_stack_adjustment_and_call): Remove
33421         unnecessary unadjusted_alignment check.
33423 2017-05-01  Xi Ruoyao  <ryxi@stu.xidian.edu.cn>
33425         PR c++/80038
33426         * cilk_common.c (expand_builtin_cilk_detach): Move pedigree
33427         operations here.
33428         * gimplify.c (gimplify_cilk_detach): New function.
33429         (gimplify_call_expr, gimplify_modify_expr): Call it as needed.
33430         * tree-core.h: Document EXPR_CILK_SPAWN.
33431         * tree.h (EXPR_CILK_SPAWN): Define.
33433 2017-05-01  David Malcolm  <dmalcolm@redhat.com>
33435         * diagnostic-show-locus.c (layout::get_expanded_location): Rewrite
33436         to use new fixit_hint representation, using the "replace" logic.
33437         (get_line_span_for_fixit_hint): Likewise.
33438         (layout::print_any_fixits): Likewise.
33439         (selftest::test_one_liner_many_fixits): Rename to...
33440         (selftest::test_one_liner_many_fixits_1): ...this, and update
33441         comment and expected output to reflect that the multiple fix-it
33442         hints are now consolidated into one insertion.
33443         (selftest::test_one_liner_many_fixits_2): New test.
33444         (selftest::test_diagnostic_show_locus_one_liner): Update for
33445         above.
33446         (selftest::test_fixit_consolidation): Update for fix-it API
33447         change.
33448         * diagnostic.c (print_parseable_fixits): Likewise.
33449         * edit-context.c (edited_line::m_line_events): Convert from
33450         auto_vec <line_event *> to auto_vec <line_event>.
33451         (class line_event): Convert from abstract base class to a concrete
33452         class, taking over the role of replace_event.
33453         (class insert_event): Delete.
33454         (class replace_event): Rename to class line_event.  Convert to
33455         half-open range.
33456         (edit_context::add_fixits): Reimplement.
33457         (edit_context::apply_insert): Delete.
33458         (edit_context::apply_replace): Rename to...
33459         (edit_context::apply_fixit): ...this.  Convert to half-open range.
33460         (edited_file::apply_insert): Delete.
33461         (edited_file::apply_replace): Rename to...
33462         (edited_file::apply_fixit): ...this.
33463         (edited_line::~edited_line): Drop deletion of events.
33464         (edited_line::apply_insert): Delete.
33465         (edited_line::apply_replace): Rename to...
33466         (edited_line::apply_fixit): ...this.  Convert to half-open range.
33467         Update for change to type of m_line_events.
33468         * edit-context.h (edit_context::apply_insert): Delete.
33469         (edit_context::apply_replace): Rename to...
33470         (edit_context::apply_fixit): ...this.
33472 2017-05-01  Martin Sebor  <msebor@redhat.com>
33474         * gimple-ssa-sprintf.c (format_integer): Set knownrange when it's
33475         known.
33477 2017-05-01  Uros Bizjak  <ubizjak@gmail.com>
33479         PR target/68491
33480         * config/i386/cpuid.h (__get_cpuid): Always return 0 when
33481         __get_cpuid_max returns 0.
33482         (__get_cpuid_count): Ditto.
33484 2017-05-01  Eric Botcazou  <ebotcazou@adacore.com>
33486         * tree.c (substitute_in_expr) <tcc_vl_exp>: Also inline a call if the
33487         replacement expression is another instance of one of its arguments.
33489 2017-05-01  Jakub Jelinek  <jakub@redhat.com>
33491         PR target/79430
33492         * rtlanal.c (reg_set_p): If reg is a stack_pointer_rtx, also
33493         check for stack push/pop autoinc.
33494         * config/i386/i386.c (ix86_agi_dependent): Return false
33495         if the only reason why modified_in_p returned true is that
33496         addr is SP based and set_insn is a push or pop.
33498 2017-04-29  Jan Hubicka  <hubicka@ucw.cz>
33500         * ipa-inline.c (compute_inlined_call_time): Remove now unnecesary
33501         overflow check.
33503 2017-04-29  Jan Hubicka  <hubicka@ucw.cz>
33505         PR ipa/79224
33506         * ipa-inline-analysis.c (dump_predicate): Add optional parameter NL.
33507         (account_size_time): Use two predicates - exec_pred and
33508         nonconst_pred_ptr.
33509         (evaluate_conditions_for_known_args): Compute both clause and
33510         nonspec_clause.
33511         (evaluate_properties_for_edge): Evaulate both clause and nonspec_clause.
33512         (inline_summary_t::duplicate): Update.
33513         (estimate_function_body_sizes): Caluculate exec and nonconst predicates
33514         separately.
33515         (compute_inline_parameters): Likewise.
33516         (estimate_edge_size_and_time): Update caluclation of time.
33517         (estimate_node_size_and_time): Compute both time and nonspecialized
33518         time.
33519         (estimate_ipcp_clone_size_and_time): Update.
33520         (inline_merge_summary): Update.
33521         (do_estimate_edge_time): Update.
33522         (do_estimate_edge_size): Update.
33523         (do_estimate_edge_hints): Update.
33524         (inline_read_section, inline_write_summary): Stream both new predicates.
33525         * ipa-inline.c (compute_uninlined_call_time): Take uninlined_call_time
33526         as argument.
33527         (compute_inlined_call_time): Cleanup.
33528         (big_speedup_p): Update.
33529         (edge_badness): Update.
33530         * ipa-inline.h (INLINE_TIME_SCALE): Remove.
33531         (size_time_entry): Replace predicate by exec_predicate and
33532         nonconst_predicate.
33533         (edge_growth_cache_entry): Cache both time nad nonspecialized time.
33534         (estimate_edge_time): Return also nonspec_time.
33535         (reset_edge_growth_cache): Update.
33537 2017-04-29  Jakub Jelinek  <jakub@redhat.com>
33539         PR rtl-optimization/80491
33540         * ifcvt.c (noce_process_if_block): When looking for x setter
33541         with missing else_bb, don't check only the insn right before
33542         cond_earliest, but look for the last insn that x is modified in
33543         within the same bb.
33545         PR rtl-optimization/80491
33546         * alias.c (memory_modified_in_insn_p): Return true for CALL_INSNs.
33548 2017-04-29  Marc Glisse  <marc.glisse@inria.fr>
33550         PR tree-optimization/80487
33551         * tree-ssa-alias.c (stmt_kills_ref_p): Handle stpncpy and strncpy.
33553 2017-04-29  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
33555         PR tree-optimization/79697
33556         * tree-ssa-dce.c (mark_stmt_if_obviously_necessary): Check if callee
33557         is BUILT_IN_STRDUP, BUILT_IN_STRNDUP, BUILT_IN_REALLOC.
33558         (propagate_necessity): Check if def_callee is BUILT_IN_STRDUP or
33559         BUILT_IN_STRNDUP.
33560         * gimple-fold.c (gimple_fold_builtin_realloc): New function.
33561         (gimple_fold_builtin): Call gimple_fold_builtin_realloc.
33563 2017-04-28  Martin Sebor  <msebor@redhat.com>
33565         PR tree-optimization/80523
33566         * gimple-ssa-sprintf.c (target_to_host_charmap): New global variable.
33567         (init_target_to_host_charmap, target_to_host, target_strtol10): New
33568         functions.
33569         (maybe_warn, format_directive, parse_directive): Use new functions.
33570         (pass_sprintf_length::execute): Call init_target_to_host_charmap.
33572 2017-04-28  Marc Glisse  <marc.glisse@inria.fr>
33574         * match.pd (X+Z OP Y+Z, X-Z OP Y-Z, Z-X OP Z-Y): New transformations.
33576 2017-04-28  Bernd Edlinger  <bernd.edlinger@hotmail.de>
33578         * configure.ac (SYSTEM_HEADER_DIR, BUILD_SYSTEM_HEADER_DIR,
33579         target_header_dir): Set correctly.
33580         * configure: Regenerated.
33581         * Makefile.in (BUILD_SYSTEM_HEADER_DIR): New make variabe.
33582         (LIMITS_H_TEST, if_multiarch, stmp-fixinc): Use BUILD_SYSTEM_HEADER_DIR
33583         instead of SYSTEM_HEADER_DIR.
33585 2017-04-28  Jan Hubicka  <hubicka@ucw.cz>
33587         * ipa-cp.c (perform_estimation_of_a_value): Turn time to sreal.
33588         (estimate_local_effects): Likewise.
33589         * ipa-inline.c (compute_inlined_call_time, want_inline_small_function_p,
33590         edge_badness, inline_small_functions, dump_overall_stats): LIkewise.
33591         * ipa-inline.h (edge_growth_cache_entry, estimate_time_after_inlining,
33592         estimate_ipcp_clone_size_and_time, do_estimate_edge_time,
33593         do_estimate_edge_time, estimate_edge_time): Likewise.
33594         * ipa-inline-analysis.c (estimate_node_size_and_time,
33595         estimate_ipcp_clone_size_and_time, do_estimate_edge_time): Likewise.
33596         (estimate_time_after_inlining): Remove.
33598 2017-04-28  Martin Liska  <mliska@suse.cz>
33600         * doc/gcov.texi: Enhance documentation of gcov.
33602 2017-04-28  Martin Liska  <mliska@suse.cz>
33604         * doc/gcov.texi: Sort options in alphabetic order.
33605         * doc/gcov-dump.texi: Likewise.
33606         * doc/gcov-tool.texi: Likewise.
33607         * gcov.c (print_usage): Likewise.
33608         * gcov-dump.c (print_usage): Likewise.
33609         * gcov-tool.c (print_merge_usage_message): Likewise.
33610         (print_rewrite_usage_message): Likewise.
33611         (print_overlap_usage_message): Likewise.
33613 2017-04-28  Martin Liska  <mliska@suse.cz>
33615         PR gcov-profile/53915
33616         * gcov.c (format_gcov): Print 'NAN %' when top > bottom.
33618 2017-04-28  Martin Liska  <mliska@suse.cz>
33620         PR gcov-profile/79891
33621         * gcov.c (add_line_counts): Assign BBs to lines just if the BB
33622         is marked by compiler as living on a line.
33623         (get_cycles_count): Remove usage of the union.
33624         (output_intermediate_file): Likewise.
33625         (find_source): Fix GNU coding style.
33626         (accumulate_line_counts): Remove old non-all block mode.
33627         (output_lines): Remove usage of the union.
33628         * profile.c (output_location): Include all BBs, even if
33629         belonging to a same line (and file) as a previous BB.
33631 2017-04-28  Martin Liska  <mliska@suse.cz>
33633         * gcov.c (process_args): Handle new argument 'w'.
33634         (read_graph_file): Assign ID to BBs.
33635         (output_branch_count): Display BB # if verbose flag is set.
33636         (output_lines): Likewise for arcs.
33637         (print_usage): Add '--verbose' option help.
33638         * doc/gcov.texi: Document --verbose (-w) option.
33640 2017-04-28  Martin Liska  <mliska@suse.cz>
33642         * gcov.c (struct block_location_info): New struct.
33643         (process_file): Fill up the new structure.
33644         (read_graph_file): Replace usage of encoding by the newly added
33645         struct.
33646         (add_line_counts): Likewise.
33647         (accumulate_line_counts): Remove usage of the union.
33648         (function_info::function_info): New function.
33649         (function_info::~function_info): Likewise.
33650         (process_file): Call delete instead of release_function.
33651         (release_function): Release the function.
33652         (release_structures): Call delete instead of release_function.
33653         (solve_flow_graph): Replace usage of num_blocks.
33654         (find_exception_blocks): Likewise.
33655         (output_lines): Fix GNU coding style.
33657 2017-04-28  Martin Liska  <mliska@suse.cz>
33659         PR driver/56469
33660         * coverage.c (coverage_remove_note_file): New function.
33661         * coverage.h: Declare the function.
33662         * toplev.c (finalize): Clean if an error has been seen.
33664 2017-04-28  Martin Liska  <mliska@suse.cz>
33666         PR gcov-profile/80031
33667         * gcov-dump.c (tag_blocks): Just print number of basic blocks.
33668         * gcov-io.h (GCOV_TAG_BLOCKS_NUM): Remove unused macro.
33669         * gcov.c (read_graph_file): Read just number of blocks.
33670         * profile.c (branch_prob): Do not stream 0 flags per a basic
33671         block.
33673 2017-04-28  Martin Liska  <mliska@suse.cz>
33675         * gcov-dump.c (tag_*): Add new argument to declarations.
33676         (dump_gcov_file): Likewise.
33677         (tag_blocks): Add and use new argument depth.
33678         (tag_arcs): Likewise.
33679         (tag_lines): Likewise.
33680         (tag_counters): Likewise.
33681         (tag_summary): Likewise.
33682         (dump_working_sets): Use depth to do a proper indentation.
33684 2017-04-28  Jakub Jelinek  <jakub@redhat.com>
33686         PR bootstrap/80531
33687         * cgraph.h (symtab_node::debug_symtab): No longer inline.
33688         * symtab.c (symtab_node::debug_symtab): Move definition here.
33690 2017-04-28  Richard Biener  <rguenther@suse.de>
33692         * lto-streamer.h (LTO_major_version): Bump to 7.
33694 2017-04-28  Richard Biener  <rguenther@suse.de>
33696         * tree-vrp.c (assert_info): New struct.
33697         (add_assert_info): New helper.
33698         (register_edge_assert_for_2): Refactor to add asserts to a vector
33699         of assert_info.
33700         (register_edge_assert_for_1): Likewise.
33701         (register_edge_assert_for): Likewise.
33702         (finish_register_edge_assert_for): New helper actually registering
33703         asserts where live on edge.
33704         (find_conditional_asserts): Adjust.
33705         (find_switch_asserts): Likewise.
33706         (evrp_dom_walker::try_find_new_range): Generalize.
33707         (evrp_dom_walker::before_dom_children): Use register_edge_assert_for.
33709 2017-04-27  Marek Polacek  <polacek@redhat.com>
33711         PR sanitizer/80349
33712         * fold-const.c (fold_binary_loc) <case EQ_EXPR, NE_EXPR>: Convert
33713         arg10 and arg11 to itype.
33715 2017-04-27  Jonathan Wakely  <jwakely@redhat.com>
33717         * doc/extend.texi (Object Size Checking): Improve grammar.
33719 2017-04-27  Richard Earnshaw  <rearnsha@arm.com>
33721         PR target/80530
33722         * config/aarch64/aarch64.c (aarch64_emit_approx_sqrt): Ensure
33723         that the logic for permitting reciprocal estimates matches that
33724         in use_rsqrt_p.
33726 2017-04-27  Jakub Jelinek  <jakub@redhat.com>
33728         PR c++/80534
33729         * tree.c (type_cache_hasher::equal): Only compare
33730         TYPE_TYPELESS_STORAGE flag on non-aggregate element types.
33731         (build_array_type_1): Only hash TYPE_TYPELESS_STORAGE flag on
33732         non-aggregate element types.
33733         * tree.h (TYPE_TYPELESS_STORAGE): Fix comment typo, add more details
33734         about the flag on ARRAY_TYPEs in the comment, formatting fix.
33736 2017-04-27  Richard Biener  <rguenther@suse.de>
33738         PR middle-end/80533
33739         * emit-rtl.c (set_mem_attributes_minus_bitpos): When
33740         stripping ARRAY_REFs from MEM_EXPR make sure we're not
33741         keeping a reference to a trailing array.
33743 2017-04-27  Richard Biener  <rguenther@suse.de>
33745         PR middle-end/80539
33746         * tree-chrec.c (chrec_fold_plus_poly_poly): Deal with not
33747         being in loop-closed SSA form conservatively.
33748         (chrec_fold_multiply_poly_poly): Likewise.
33750 2017-04-27  Tamar Christina  <tamar.christina@arm.com>
33752         PR middle-end/79665
33753         * expr.c (expand_expr_real_2): Move TRUNC_MOD_EXPR, FLOOR_MOD_EXPR,
33754         CEIL_MOD_EXPR, ROUND_MOD_EXPR cases.
33756 2017-04-27  Jakub Jelinek  <jakub@redhat.com>
33758         PR target/77728
33759         * config/aarch64/aarch64.c (struct aarch64_fn_arg_alignment): Remove.
33760         (aarch64_function_arg_alignment): Return unsigned int again, but still
33761         ignore TYPE_FIELDS chain decls other than FIELD_DECLs.
33762         (aarch64_layout_arg): Adjust aarch64_function_arg_alignment caller.
33763         Don't emit -Wpsabi note.
33764         (aarch64_function_arg_boundary): Likewise.
33765         (aarch64_gimplify_va_arg_expr): Adjust aarch64_function_arg_alignment
33766         caller.
33768 2017-04-26  Nathan Sidwell  <nathan@acm.org>
33770         * tree.h (crc32_unsigned_n): Declare.
33771         (crc32_unsigned, crc32_unsigned): Make inline.
33772         * tree.c (crc32_unsigned_bits): Replace with ...
33773         (crc32_unsigned_n): ... this.
33774         (crc32_unsigned, crc32_byte): Remove.
33775         (crc32_string): Remove unnecessary braces.
33777 2017-04-25  Jan Hubicka  <hubicka@ucw.cz>
33779         * ipa-cp.c (estimate_local_effects): Convert sreal to int.
33780         * ipa-inline-analysis.c (MAX_TIME): Remove.
33781         (account_size_time): Use sreal for time.
33782         (dump_inline_summary): Update.
33783         (estimate_function_body_sizes): Update.
33784         (estimate_edge_size_and_time): Update.
33785         (estimate_calls_size_and_time): Update.
33786         (estimate_node_size_and_time): Update.
33787         (inline_merge_summary): Update.
33788         (inline_update_overall_summary): Update.
33789         (estimate_time_after_inlining): Update.
33790         (inline_read_section): Update.
33791         (inline_write_summary): Update.
33792         * ipa-inline.c (compute_uninlined_call_time): Update.
33793         (compute_inlined_call_time): Update.
33794         (recursive_inlining): Update.
33795         (inline_small_functions): Update.
33796         (dump_overall_stats): Update.
33797         * ipa-inline.h: Include sreal.h.
33798         (size_time_entry): Turn time to sreal.
33799         (inline_summary): Turn self_time nad time to sreal.
33801 2017-04-25  Jan Hubicka  <hubicka@ucw.cz>
33803         * sreal.c: Include backend.h, tree.h, gimple.h, cgraph.h and
33804         data-streamer.h
33805         (sreal::stream_out, sreal::stream_in): New.
33806         * sreal.h (sreal::stream_out, sreal::stream_in): Declare.
33808 2017-04-25  Jakub Jelinek  <jakub@redhat.com>
33810         * Makefile.in (s-options): Invoke opt-gather.awk with LC_ALL=C in the
33811         environment.
33813 2017-04-25  Uros Bizjak  <ubizjak@gmail.com>
33815         PR target/70799
33816         * config/i386/i386.c (dimode_scalar_to_vector_candidate_p):
33817         Handle ASHIFTRT.
33818         (dimode_scalar_chain::compute_convert_gain): Ditto.
33819         (dimode_scalar_chain::make_vector_copies): Ditto.
33820         (dimode_scalar_chain::convert_reg): Ditto.
33821         (dimode_scalar_chain::convert_insn): Ditto.
33822         * config/i386/sse.md (VI24_AVX512BW_1): Remove mode iterator.
33823         (VI248_AVX512BW_1): New mode iterator.
33824         (<mask_codefor>ashr<mode>3<mask_name>): Merge insn pattern with
33825         <mask_codefor>ashrv2di3<mask_name> insn using VI248_AVX512BW_1
33826         mode iterator.
33828 2017-04-25  Martin Sebor  <msebor@redhat.com>
33830         PR tree-optimization/80497
33831         * gimple-ssa-sprintf.c (get_int_range): Avoid assuming all integer
33832         constants are representable in HOST_WIDE_INT.
33833         (parse_directive): Ditto.
33835 2017-04-25  Martin Sebor  <msebor@redhat.com>
33837         PR bootstrap/80486
33838         * dominance.c (dom_info::m_n_basic_blocks): Change type to unsigned.
33839         (new_zero_array): Adjust signature.
33840         (dom_info::dom_init): Used unsigned rather that size_t.
33841         (dom_info::dom_info): Same.
33843 2017-04-25  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
33844             Jakub Jelinek  <jakub@redhat.com>
33846         PR target/77728
33847         * config/arm/arm.c: Include gimple.h.
33848         (aapcs_layout_arg): Emit -Wpsabi note if arm_needs_doubleword_align
33849         returns negative, increment ncrn only if it returned positive.
33850         (arm_needs_doubleword_align): Return int instead of bool,
33851         ignore DECL_ALIGN of non-FIELD_DECL TYPE_FIELDS chain
33852         members, but if there is any such non-FIELD_DECL
33853         > PARM_BOUNDARY aligned decl, return -1 instead of false.
33854         (arm_function_arg): Emit -Wpsabi note if arm_needs_doubleword_align
33855         returns negative, increment nregs only if it returned positive.
33856         (arm_setup_incoming_varargs): Likewise.
33857         (arm_function_arg_boundary): Emit -Wpsabi note if
33858         arm_needs_doubleword_align returns negative, return
33859         DOUBLEWORD_ALIGNMENT only if it returned positive.
33861 2017-04-25  Marek Polacek  <polacek@redhat.com>
33863         PR sanitizer/80349
33864         * fold-const.c (fold_binary_loc) <case BIT_IOR_EXPR>: Convert arg0's
33865         first argument to type.
33867 2017-04-25  Bill Seurer  <seurer@linux.vnet.ibm.com>
33869         PR target/80482
33870         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Change
33871         type checks to test for compatibility instead of equality.
33873 2017-04-25  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
33874             Jakub Jelinek  <jakub@redhat.com>
33876         PR target/77728
33877         * config/aarch64/aarch64.c (struct aarch64_fn_arg_alignment): New
33878         type.
33879         (aarch64_function_arg_alignment): Return aarch64_fn_arg_alignment
33880         struct.  Ignore DECL_ALIGN of decls other than FIELD_DECL for
33881         the alignment computation, but return their maximum in warn_alignment.
33882         (aarch64_layout_arg): Adjust aarch64_function_arg_alignment caller.
33883         Emit a -Wpsabi note if warn_alignment is 16 bytes, but alignment
33884         is smaller.
33885         (aarch64_function_arg_boundary): Likewise.  Simplify using MIN/MAX.
33886         (aarch64_gimplify_va_arg_expr): Adjust aarch64_function_arg_alignment
33887         caller.
33889 2017-04-25  Claudiu Zissulescu  <claziss@synopsys.com>
33891         * config/arc/simdext.md (dmpyh): Fix typo.
33893 2017-04-25  Richard Biener  <rguenther@suse.de>
33895         PR tree-optimization/80492
33896         * alias.c (compare_base_decls): Handle registers with asm
33897         specification conservatively.
33898         * tree-ssa-alias.c (decl_refs_may_alias_p): Handle
33899         compare_base_decls returning dont-know properly.
33901 2017-04-25  Claudiu Zissulescu  <claziss@synopsys.com>
33903         * config/arc/arc.c (LEGITIMATE_OFFSET_ADDRESS_P): Delete macro.
33904         (legitimate_offset_address_p): New function.
33905         (arc_legitimate_address_p): Use above function.
33907 2017-04-25  Claudiu Zissulescu  <claziss@synopsys.com>
33909         * config/arc/arc.c (arc_output_mi_thunk): Emit PIC calls.
33911 2017-04-25  Claudiu Zissulescu  <claziss@synopsys.com>
33913         * config/arc/arc.c (arc_conditional_register_usage): Use ACCL,
33914         ACCH registers whenever they are available.
33916 2017-04-25  Claudiu Zissulescu  <claziss@synopsys.com>
33918         * config/arc/arc.c (arc_conditional_register_usage): Make D0, D1
33919         double regs fix when not used.
33921 2017-04-25  Claudiu Zissulescu  <claziss@synopsys.com>
33923         * config/arc/arc.h (REGNO_OK_FOR_BASE_P): Consider also extension
33924         core registers.
33925         (REG_OK_FOR_INDEX_P_NONSTRICT): Likewise.
33926         (REG_OK_FOR_BASE_P_NONSTRICT): Likewise.
33928 2017-04-25  Claudiu Zissulescu  <claziss@synopsys.com>
33930         * config/arc/arc.c (arc_output_addsi): Check for h-register class
33931         when emitting short ADD instructions.
33933 2017-04-25  Claudiu Zissulescu  <claziss@synopsys.com>
33935         * config/arc/arc.md (cmpsi_cc_insn_mixed): Use 'h' register
33936         constraint.
33937         (cmpsi_cc_c_insn): Likewise.
33938         (cbranchsi4_scratch): Compute proper instruction length using
33939         compact_hreg_operand.
33940         * config/arc/predicates.md (compact_hreg_operand): New predicate.
33942 2017-04-25  Richard Biener  <rguenther@suse.de>
33944         PR middle-end/80509
33945         * passes.c (pass_manager::pass_manager): Initialize
33946         m_name_to_pass_map.
33948 2017-04-25  Richard Biener  <rguenther@suse.de>
33950         PR tree-optimization/79201
33951         * tree-ssa-sink.c (statement_sink_location): Handle calls.
33953 2017-04-25  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
33955         PR target/80464
33956         * config/s390/vector.md: Split MEM->GPR vector moves for
33957         non-s_operand addresses.
33959 2017-04-25  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
33961         PR target/79895
33962         * config/s390/predicates.md (reload_const_wide_int_operand): New
33963         predicate.
33964         * config/s390/s390.md ("movti"): Remove d/P alternative.
33965         ("movti_bigconst"): New pattern definition.
33967 2017-04-25  Dominik Vogt  <vogt@linux.vnet.ibm.com>
33969         PR target/80080
33970         * s390-protos.h (s390_expand_cs_hqi): Removed.
33971         (s390_expand_cs, s390_expand_atomic_exchange_tdsi): New prototypes.
33972         * config/s390/s390.c (s390_emit_compare_and_swap): Handle all integer
33973         modes as well as CCZ1mode and CCZmode.
33974         (s390_expand_atomic_exchange_tdsi, s390_expand_atomic): Adapt to new
33975         signature of s390_emit_compare_and_swap.
33976         (s390_expand_cs_hqi): Likewise, make static.
33977         (s390_expand_cs_tdsi): Generate an explicit compare before trying
33978         compare-and-swap, in some cases.
33979         (s390_expand_cs): Wrapper function.
33980         (s390_expand_atomic_exchange_tdsi): New backend specific expander for
33981         atomic_exchange.
33982         (s390_match_ccmode_set): Allow CCZmode <-> CCZ1 mode.
33983         * config/s390/s390.md ("atomic_compare_and_swap<mode>"): Merge the
33984         patterns for small and large integers.  Forbid symref memory operands.
33985         Move expander to s390.c.  Require cc register.
33986         ("atomic_compare_and_swap<DGPR:mode><CCZZ1:mode>_internal")
33987         ("*atomic_compare_and_swap<TDI:mode><CCZZ1:mode>_1")
33988         ("*atomic_compare_and_swapdi<CCZZ1:mode>_2")
33989         ("*atomic_compare_and_swapsi<CCZZ1:mode>_3"): Use s_operand to forbid
33990         symref memory operands.  Remove CC mode and call s390_match_ccmode
33991         instead.
33992         ("atomic_exchange<mode>"): Allow and implement all integer modes.
33994 2017-04-25  Dominik Vogt  <vogt@linux.vnet.ibm.com>
33996         * config/s390/s390.md (define_peephole2): New peephole to help
33997         combining the load-and-test pattern with volatile memory.
33999 2017-04-25  Dominik Vogt  <vogt@linux.vnet.ibm.com>
34001         * config/s390/s390.md ("cstorecc4"): Use load-on-condition and deal
34002         with CCZmode for TARGET_Z196.
34004 2017-04-25  Jakub Jelinek  <jakub@redhat.com>
34006         PR rtl-optimization/80501
34007         * combine.c (make_compound_operation_int): Set subreg_code to SET
34008         even for AND with mask of the sign bit of mode.
34010         PR rtl-optimization/80500
34011         * loop-unroll.c (combine_var_copies_in_loop_exit): Call copy_rtx on
34012         sum's initial value.
34014 2017-04-25  Julian Brown  <julian@codesourcery.com>
34015             Naveen H.S  <Naveen.Hurugalawadi@cavium.com>
34017         * config/aarch64/thunderx2t99.md (thunderx2t99_crc): New Reservation.
34019 2017-04-25  Marc Glisse  <marc.glisse@inria.fr>
34021         * fold-const.c (tree_single_nonzero_warnv_p): Handle SSA_NAME.
34023 2017-04-25  Julian Brown  <julian@codesourcery.com>
34024             Naveen H.S  <Naveen.Hurugalawadi@cavium.com>
34026         * config/aarch64/thunderx2t99.md (thunderx2t99_aes): New Reservation.
34027         (thunderx2t99_sha): New Reservation.
34029 2017-04-25  Julian Brown  <julian@codesourcery.com>
34030             Naveen H.S  <Naveen.Hurugalawadi@cavium.com>
34032         * config/aarch64/aarch64-simd.md (aarch64_simd_vec_set<mode>): Fix
34033         type for 1-element load.
34035 2017-04-24  Marc Glisse  <marc.glisse@inria.fr>
34037         * match.pd (X/[ex]C CMP Y/[ex]C): New transformation.
34039 2017-04-24  Martin Jambor  <mjambor@suse.cz>
34041         PR tree-optimization/80293
34042         * tree-sra.c (scalarizable_type_p): New parameter const_decl, make
34043         char arrays not totally scalarizable if it is false.
34044         (analyze_all_variable_accesses): Pass correct value in the new
34045         parameter.  Add a statistics counter.
34047 2017-04-24  Jan Hubicka  <hubicka@ucw.cz>
34049         PR middle-end/79931
34050         * ipa-devirt.c (dump_possible_polymorphic_call_targets): Fix ICE.
34052 2017-04-24  Richard Biener  <rguenther@suse.de>
34054         PR tree-optimization/80494
34055         * tree-scalar-evolution.c (analyze_scalar_evolution_1): Bail
34056         out for complex types.
34058 2017-04-24  Richard Biener  <rguenther@suse.de>
34060         * tree-ssa-sccvn.h (run_scc_vn): Adjust prototype.
34061         * tree-ssa-sccvn.c (print_scc): Print SCC size.
34062         (extract_and_process_scc_for_name): Never fail but drop SCC to varying.
34063         (DFS): Adjust and never fail.
34064         (sccvn_dom_walker::fail): Remove.
34065         (sccvn_dom_walker::before_dom_children): Adjust.
34066         (run_scc_vn): Likewise and never fail.
34067         * tree-ssa-pre.c (pass_pre::execute): Adjust.
34068         (pass_fre::execute): Likewise.
34070 2017-04-24  Richard Biener  <rguenther@suse.de>
34072         PR tree-optimization/79725
34073         * tree-ssa-sink.c (statement_sink_location): Return whether
34074         failure reason was zero uses.  Move that check later.
34075         (sink_code_in_bb): Deal with zero uses by removing the stmt
34076         if possible.
34078 2017-04-24  Richard Biener  <rguenther@suse.de>
34080         PR c++/2972
34081         * tree-ssa-uninit.c (warn_uninitialized_vars): Handle some
34082         pointer-based references.
34084 2017-04-24  Richard Biener  <rguenther@suse.de>
34086         PR bootstrap/79814
34087         * pass_manager.h (pass_manager::operator new): Remove.
34088         (pass_manager::operator delete): Likewise.
34089         * passes.c (pass_manager::operator new): Remove.
34090         (pass_manager::operator delete): Likewise.
34091         (pass_manager::pass_manager): Zero individual pass members.
34093 2017-04-23  Uros Bizjak  <ubizjak@gmail.com>
34095         PR target/70799
34096         * config/i386/i386.c (dimode_scalar_to_vector_candidate_p)
34097         <case ASHIFT, case LSHIFTRT>: Also consider variable shifts.
34098         Check "XEXP (src, 1)" operand here.
34099         <case PLUS, case MINUS, case IOR, case XOR, case AND>:
34100         Check "XEXP (src, 1)" operand here.
34101         (dimode_scalar_chain::make_vector_copies): Detect count register
34102         of a shift instruction.  Zero extend count register from QImode
34103         to DImode to satisfy vector shift pattern count operand predicate.
34104         Substitute vector shift count operand with a DImode copy.
34105         (dimode_scalar_chain::convert_reg): Ditto, zero-extend from
34106         vector register.
34108 2017-04-21  Uros Bizjak  <ubizjak@gmail.com>
34110         * config/i386/i386.md (*extzvqi_mem_rex64): Move above *extzv<mode>.
34111         Remove UNSPEC_NOREX_MEM tag.  Update corresponding peephole2 pattern.
34112         (*insvqi_1_mem_rex64): Move above insv<mode>_1.  Remove
34113         UNSPEC_NOREX_MEM tag.  Update corresponding peephole2 pattern.
34114         (UNSPEC_NOREX_MEM): Remove definition.
34116 2017-04-21  Richard Biener  <rguenther@suse.de>
34118         PR tree-optimization/79547
34119         * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
34120         Handle strlen, strcmp, strncmp, strcasecmp, strncasecmp, memcmp,
34121         bcmp, strspn, strcspn, __builtin_object_size and __builtin_constant_p
34122         without any constraints.
34124 2017-04-21  Richard Biener  <rguenther@suse.de>
34126         PR tree-optimization/78847
34127         * fold-const.c (split_tree): Handle POINTER_PLUS_EXPR.
34129 2017-04-21  Richard Biener  <rguenther@suse.de>
34131         * tree.h (build_qualified_type): Annotate with CXX_MEM_STAT_INFO.
34132         (build_distinct_type_copy): Likewise.
34133         (build_variant_type_copy): Likewise.
34134         * tree.c (build_qualified_type): Pass down mem-stat info.
34135         (build_distinct_type_copy): Likewise.
34136         (build_variant_type_copy): Likewise.
34138 2017-04-21  Richard Biener  <rguenther@suse.de>
34140         PR tree-optimization/80237
34141         * tree-ssa-pre.c (find_leader_in_sets): Add third set argument,
34142         defaulted to NULL.
34143         (phi_translate_1): Also allow a leader in AVAIL_OUT of pred
34144         for a simplified result.
34146 2016-04-21  Richard Biener  <rguenther@suse.de>
34148         * tree-ssa-loop-ivcanon.c (constant_after_peeling): Do not require
34149         sth as strict as a simple_iv but a chrec without symbols and an
34150         operand defined in the loop we are peeling (and not some subloop).
34151         (propagate_constants_for_unrolling): Propagate all constants.
34153 2017-04-20  Uros Bizjak  <ubizjak@gmail.com>
34155         PR target/79804
34156         * config/i386/i386.c (print_reg): Remove assert for disalowed
34157         regno values, call output_operand_lossage instead.
34159 2017-04-20  Uros Bizjak  <ubizjak@gmail.com>
34161         PR target/78090
34162         * config/i386/constraints.md (Yc): New register constraint.
34163         * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
34164         Use Yc constraint for alternative 2 of operand 0.  Remove
34165         preferred_for_speed attribute.
34167 2017-04-20  Alexander Monakov  <amonakov@ispras.ru>
34169         * omp-low.c (lower_lastprivate_clauses): Correct handling of linear and
34170         lastprivate clauses in SIMT case.
34172 2017-04-20  Volker Reichelt  <v.reichelt@netcologne.de>
34174         * doc/invoke.texi (-Wextra-semi): Document new warning option.
34176 2017-04-20  Richard Biener  <rguenther@suse.de>
34178         PR tree-optimization/57796
34179         * tree-vect-stmts.c (vect_model_store_cost): Cost scatters
34180         as N scalar stores.
34181         (vect_model_load_cost): Cost gathers as N scalar loads.
34183 2017-04-20  Richard Biener  <rguenther@suse.de>
34185         * ggc-page.c (ggc_allocated_p): Rename to ...
34186         (safe_lookup_page_table_entry): ... this and return the lookup
34187         result.
34188         (gt_ggc_m_S): Use safe_lookup_page_table_entry.
34190 2017-04-20  Richard Biener  <rguenther@suse.de>
34192         PR tree-optimization/80453
34193         * tree-ssa-sccvn.h (struct vn_phi_s): Add cclhs and ccrhs members.
34194         * tree-ssa-sccvn.c (cond_stmts_equal_p): Use recorded lhs and rhs
34195         from the conditions.
34196         (vn_phi_eq): Pass them down.
34197         (vn_phi_lookup): Record them.
34198         (vn_phi_insert): Likewise.
34200 2017-04-20  Matthew Fortune  <matthew.fortune@imgtec.com>
34202         * config/mips/mips.c (mips_expand_vec_perm_const): Re-fix
34203         uninitialized variable warning to avoid buffer overrun.
34205 2017-04-20  Alexander Monakov  <amonakov@ispras.ru>
34207         PR other/71250
34208         * doc/invoke.texi (-Wmissing-field-initializers): Mention that warning
34209         is suppressed for '{ 0 }' in C.
34211 2017-04-20  Jakub Jelinek  <jakub@redhat.com>
34213         * BASE-VER: Set to 8.0.0.
34215 2017-04-20  Thomas Preud'homme  <thomas.preudhomme@arm.com>
34217         * config/arm/arm.c (arm_elf_asm_cdtor): Create non-default
34218         priority .init_array and .fini_array section with SECTION_NOTYPE
34219         flag.
34221 2017-04-20  Jakub Jelinek  <jakub@redhat.com>
34223         PR middle-end/80423
34224         * tree.h (build_array_type): Add typeless_storage default argument.
34225         * tree.c (type_cache_hasher::equal): Also compare
34226         TYPE_TYPELESS_STORAGE flag for ARRAY_TYPEs.
34227         (build_array_type): Add typeless_storage argument, set
34228         TYPE_TYPELESS_STORAGE to it, if shared also hash it, and pass to
34229         recursive call.
34230         (build_nonshared_array_type): Adjust build_array_type_1 caller.
34231         (build_array_type): Likewise.  Add typeless_storage argument.
34233 2017-04-19  Eric Botcazou  <ebotcazou@adacore.com>
34234             Jakub Jelinek  <jakub@redhat.com>
34236         PR tree-optimization/80426
34237         * tree-vrp.c (extract_range_from_binary_expr_1): For an additive
34238         operation on symbolic operands, also compute the overflow for the
34239         invariant part when the operation degenerates into a negation.
34241 2017-04-19  Jakub Jelinek  <jakub@redhat.com>
34243         PR debug/80461
34244         * dwarf2out.c (modified_type_die, gen_type_die_with_usage):
34245         Check for t with zero TYPE_QUALS_NO_ADDR_SPACE.
34247         PR debug/80436
34248         * tree-ssa-loop-manip.c (find_uses_to_rename_def): Ignore debug uses.
34250 2017-04-19  Georg-Johann Lay  <avr@gjlay.de>
34252         PR target/80462
34253         * config/avr/avr.c (tree.h): Include it.
34254         (cgraph.h): Include it.
34255         (avr_encode_section_info): Don't warn for uninitialized progmem
34256         variable if it's just an alias.
34258 2017-04-19  Richard Biener  <rguenther@suse.de>
34260         PR ipa/65972
34261         * auto-profile.c (afdo_vpt_for_early_inline): Update SSA
34262         when needed by AutoPGO.
34264 2017-04-19  Paulo J. Matos  <paulo@matos-sorge.com>
34266         PR lto/50345
34267         * doc/lto.texi: Remove an extra 'that'.
34269 2017-04-19  Segher Boessenkool  <segher@kernel.crashing.org>
34271         PR rtl-optimization/80429
34272         * ira.c (split_live_ranges_for_shrink_wrap): Don't split regs that
34273         are only used in debug insns.
34275 2017-04-19  Eric Botcazou  <ebotcazou@adacore.com>
34276             Vladimir Makarov  <vmakarov@redhat.com>
34278         * config/sparc/predicates.md (input_operand): Add comment.  Return
34279         true for any memory operand when LRA is in progress.
34280         * config/sparc/sparc.c (sparc_expand_move): Minor formatting fix.
34282 2017-04-18  Jeff Law  <law@redhat.com>
34284         PR target/74563
34285         * mips.md ({return,simple_return}_internal): Do not overwrite
34286         operands[0].
34288 2017-04-18  Jakub Jelinek  <jakub@redhat.com>
34290         PR tree-optimization/80443
34291         * tree-vrp.c (intersect_ranges): For signed 1-bit precision type,
34292         instead of adding 1, subtract -1 and similarly instead of subtracting
34293         1 add -1.
34295 2017-04-18  Richard Sandiford  <richard.sandiford@arm.com>
34297         PR rtl-optimization/80357
34298         * haifa-sched.c (tmp_bitmap): New variable.
34299         (model_recompute): Handle duplicate use records.
34300         (alloc_global_sched_pressure_data): Initialize tmp_bitmap.
34301         (free_global_sched_pressure_data): Free it.
34303 2017-04-18  Bernd Edlinger  <bernd.edlinger@hotmail.de>
34305         Revert:
34306         2017-02-20  Bernd Edlinger  <bernd.edlinger@hotmail.de>
34307         * Makefile.in (BUILD_SYSTEM_HEADER_DIR): New make variabe.
34308         (LIMITS_H_TEST, if_multiarch, stmp-fixinc): Use BUILD_SYSTEM_HEADER_DIR
34309         instead of SYSTEM_HEADER_DIR.
34311 2017-04-18  Jeff Law  <law@redhat.com>
34313         PR middle-end/80422
34314         * cfgcleanup.c (try_crossjump_to_edge): Verify SRC1 and SRC2 have
34315         predecessors after walking up the insn chain.
34317 2017-04-18  Jakub Jelinek  <jakub@redhat.com>
34319         PR debug/80263
34320         * dwarf2out.c (modified_type_die): Try harder not to emit internal
34321         sizetype type into debug info.
34323 2017-04-18  Michael Meissner  <meissner@linux.vnet.ibm.com>
34325         PR target/80099
34326         * config/rs6000/rs6000.c (rs6000_expand_vector_extract): Eliminate
34327         unneeded test for TARGET_UPPER_REGS_SF.
34328         * config/rs6000/vsx.md (vsx_extract_v4sf_var): Likewise.
34330 2017-04-18  Jakub Jelinek  <jakub@redhat.com>
34332         PR sanitizer/80444
34333         * sancov.c (sancov_pass): Use gsi_start_nondebug_after_labels_bb
34334         instead of gsi_after_labels.
34336 2017-04-18  Jeff Law  <law@redhat.com>
34338         * regcprop.c (maybe_mode_change): Avoid creating copies of the
34339         stack pointer.
34341         Revert:
34342         2017-04-13  Jeff Law  <law@redhat.com>
34343         * config/mips.mips.md (zero_extendsidi2): Do not allow SP to appear
34344         in operands[1] if it is a MEM and TARGET_MIPS16 is active.
34346 2017-04-18  Georg-Johann Lay  <avr@gjlay.de>
34348         PR target/79453
34349         * config/avr/avr.c (intl.h): Include it.
34350         (avr_pgm_check_var_decl) [reason]: Wrap diagnostic snippets into _().
34352 2017-04-18  Martin Liska  <mliska@suse.cz>
34354         PR gcov-profile/78783
34355         * gcov-tool.c (gcov_output_files): Validate that destination
34356         file is either removed by the tool or by a user.
34358 2017-04-14  Andrew Burgess  <andrew.burgess@embecosm.com>
34359             Guy Benyei  <guybe@mellanox.com>
34361         * config/arc/arc.c (arc_reorg): Move loop_end_id into a more local
34362         block, and do not negate it, the stored id is already negative.
34364 2017-04-14  Andrew Burgess  <andrew.burgess@embecosm.com>
34366         * config/arc/arc.md (doloop_begin_i): Use @pcl assembler syntax.
34368 2017-04-14  Michael Meissner  <meissner@linux.vnet.ibm.com>
34370         PR target/80098
34371         * config/rs6000/rs6000-cpus.def (OTHER_P9_VECTOR_MASKS): Define
34372         masks of options that should be turned off if the VSX vector
34373         options are turned off.
34374         (OTHER_P8_VECTOR_MASKS): Likewise.
34375         (OTHER_VSX_VECTOR_MASKS): Likewise.
34376         * config/rs6000/rs6000.c (rs6000_option_override_internal): Call
34377         rs6000_disable_incompatible_switches to validate no type switches
34378         like -mvsx.
34379         (rs6000_incompatible_switch): New function to disallow turning on
34380         other vector options if -mno-vsx, -mno-power8-vector, or
34381         -mno-power9-vector are specified.
34383 2017-04-14  Claudiu Zissulescu  <claziss@synopsys.com>
34385         * config/arc/arc.h (CRT_CALL_STATIC_FUNCTION): Use long calls.
34387 2017-04-14  Claudiu Zissulescu  <claziss@synopsys.com>
34389         * config/arc/arc-protos.h (arc_decl_pretend_args): Remove.
34390         * config/arc/arc.c (arc_decl_pretend_args): Likewise.
34391         * config/arc/arc.h (CFA_FRAME_BASE_OFFSET): Likewise.
34392         (ARG_POINTER_CFA_OFFSET): Likewise.
34394 2017-04-14  Claudiu Zissulescu  <claziss@synopsys.com>
34396         * config/arc/arc.c (arc_mode_dependent_address_p): Relax
34397         conditions to take advantage of various optimizations.
34399 2017-04-13  Jeff Law  <law@redhat.com>
34401         * config/mips.mips.md (zero_extendsidi2): Do not allow SP to appear
34402         in operands[1] if it is a MEM and TARGET_MIPS16 is active.
34403         (zero_extendsidi2_dext): Likewise.
34405 2017-04-13  Jakub Jelinek  <jakub@redhat.com>
34407         PR sanitizer/80403
34408         * fold-const.c (fold_ternary_loc): Revert
34409         use op0 instead of fold_convert_loc (loc, type, arg0) part of
34410         2017-04-12 change.
34412 2017-04-13  Vladimir Makarov  <vmakarov@redhat.com>
34414         PR rtl-optimization/80343
34415         * lra-remat.c (update_scratch_ops): Assign original hard reg to
34416         new scratch pseudo.
34418 2017-04-13  Denis Khalikov <d.khalikov@partner.samsung.com>
34420         PR sanitizer/80414
34421         * ubsan.c (ubsan_expand_bounds_ifn): Pass original index
34422         to ubsan_encode_value.
34424 2017-04-13  Jeff Law  <law@redhat.com>
34426         * reload1.c (eliminate_regs_1): Call gen_rtx_raw_SUBREG for SUBREGs
34427         appearing in DEBUG_INSNs.
34429 2017-04-13  Martin Liska  <mliska@suse.cz>
34431         PR gcov-profile/80413
34432         * gcov-io.c (gcov_write_string): Copy to buffer just when
34433         allocated size is greater than zero.
34435 2017-04-13  Jakub Jelinek  <jakub@redhat.com>
34437         PR debug/80321
34438         * dwarf2out.c (decls_for_scope): Ignore declarations of
34439         current_function_decl in BLOCK_NONLOCALIZED_VARS.
34441 2017-04-12  Jan Hubicka  <hubicka@ucw.cz>
34443         PR lto/69953
34444         * ipa-visibility.c (non_local_p): Fix typos.
34445         (localize_node): When localizing symbol in same comdat group,
34446         dissolve the group only when we know external symbols are going
34447         to be privatized.
34448         (function_and_variable_visibility): Do not localize DECL_EXTERNAL.
34450 2017-04-12  Jakub Jelinek  <jakub@redhat.com>
34452         PR tree-optimization/79390
34453         * optabs.c (emit_conditional_move): If the preferred op2/op3 operand
34454         order does not result in usable sequence, retry with reversed operand
34455         order.
34457         PR sanitizer/80403
34458         PR sanitizer/80404
34459         PR sanitizer/80405
34460         * fold-const.c (fold_ternary_loc): Use op1 instead of arg1 as argument
34461         to fold_build2_loc.  Convert TREE_OPERAND (tem, 0) to type.  Use
34462         op0 instead of fold_convert_loc (loc, type, arg0).
34464 2017-04-12  Jeff Law  <law@redhat.com>
34466         * genattrtab.c (write_eligible_delay): Verify DELAY_INSN still
34467         has a delay slot in the generated code.
34469         * config/cris/cris.md (cris_preferred_reload_class): Return
34470         GENNONACR_REGS rather than GENERAL_REGS.
34472 2017-04-12  Jakub Jelinek  <jakub@redhat.com>
34474         PR c/80163
34475         * expr.c <CASE_CONVERT>: For EXPAND_INITIALIZER determine SIGN_EXTEND
34476         vs. ZERO_EXTEND based on signedness of treeop0's type rather than
34477         signedness of the result type.
34479 2017-04-12  Richard Biener  <rguenther@suse.de>
34480             Jeff Law  <law@redhat.com>
34482         PR tree-optimization/80359
34483         * tree-ssa-dse.c (maybe_trim_partially_dead_store): Do not
34484         trim stores to TARGET_MEM_REFs.
34486 2017-04-12  Richard Biener  <rguenther@suse.de>
34488         PR tree-optimization/79390
34489         * gimple-ssa-split-paths.c (is_feasible_trace): Restrict
34490         threading case even more.
34492 2017-04-12  Segher Boessenkool  <segher@kernel.crashing.org>
34494         PR target/80382
34495         * config/rs6000/sync.md (atomic_load<mode>, atomic_store<mode): Test
34496         for quad_address_p for TImode, instead of just not indexed_address.
34498 2017-04-12  Richard Biener  <rguenther@suse.de>
34499             Bernd Edlinger  <bernd.edlinger@hotmail.de>
34501         PR middle-end/79671
34502         * alias.c (component_uses_parent_alias_set_from): Handle
34503         TYPE_TYPELESS_STORAGE.
34504         (get_alias_set): Likewise.
34505         * tree-core.h (tree_type_common): Add typeless_storage flag.
34506         * tree.h (TYPE_TYPELESS_STORAGE): New macro.
34507         * stor-layout.c (place_union_field): Set TYPE_TYPELESS_STORAGE
34508         for types containing members with TYPE_TYPELESS_STORAGE.
34509         (place_field): Likewise.
34510         (layout_type): Likewise for ARRAY_TYPE.
34511         * lto-streamer-out.c (hash_tree): Hash TYPE_TYPELESS_STORAGE.
34512         * tree-streamer-in.c (unpack_ts_type_common_value_fields): Stream
34513         TYPE_TYPELESS_STORAGE.
34514         * tree-streamer-out.c (pack_ts_type_common_value_fields): Likewise.
34516 2017-04-12  Jakub Jelinek  <jakub@redhat.com>
34518         PR sanitizer/80349
34519         * fold-const.c (fold_binary_loc) <case BIT_IOR_EXPR>: Convert arg0's
34520         first argument to type.
34522 2017-04-11  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
34524         PR target/80376
34525         PR target/80315
34526         * config/rs6000/rs6000.c (rs6000_expand_unop_builtin): Return
34527         CONST0_RTX (mode) rather than const0_rtx where appropriate.
34528         (rs6000_expand_binop_builtin): Likewise.
34529         (rs6000_expand_ternop_builtin): Likewise; also add missing
34530         vsx_xxpermdi_* variants; also fix typo (arg1 => arg2) for
34531         vshasigma built-ins.
34532         * doc/extend.texi: Document that vec_xxpermdi's third argument
34533         must be a constant.
34535 2017-04-11  Uros Bizjak  <ubizjak@gmail.com>
34537         * config/i386/i386.c (dimode_scalar_chain::compute_convert_gain):
34538         Use shift_const cost parameter when calculating gain of STV shifts.
34540 2017-04-11  Vladimir Makarov  <vmakarov@redhat.com>
34542         PR rtl-optimization/70478
34543         * lra-constraints.c (process_alt_operands): Check memory for
34544         disfavoring memory insn operand.
34546 2017-04-11  Jakub Jelinek  <jakub@redhat.com>
34548         PR middle-end/80100
34549         * simplify-rtx.c (simplify_binary_operation_1) <case IOR>: Perform
34550         left shift in unsigned HOST_WIDE_INT type.
34552         PR rtl-optimization/80385
34553         * simplify-rtx.c (simplify_unary_operation_1): Don't transform
34554         (not (neg X)) into (plus X -1) for complex or non-integral modes.
34556         PR libgomp/80394
34557         * omp-low.c (scan_omp_task): Don't optimize away empty tasks
34558         if they have any depend clauses.
34560 2017-04-11  Martin Liska  <mliska@suse.cz>
34562         PR ipa/80212
34563         * cgraph.c (cgraph_node::dump): Dump calls_comdat_local.
34564         * ipa-split.c (split_function): Create a local comdat symbol
34565         if caller is in a comdat group.
34567 2017-04-11  Martin Liska  <mliska@suse.cz>
34569         PR ipa/80212
34570         * ipa-cp.c (determine_versionability): Handle calls_comdat_local
34571         flags.
34573 2017-04-11  Martin Sebor  <msebor@redhat.com>
34575         PR middle-end/80364
34576         * gimple-ssa-sprintf.c (get_int_range): Remove second argument and
34577         always use the int type.  Use INTEGRAL_TYPE_P() rather than testing
34578         for INTEGER_TYPE.
34579         (directive::set_width, directive::set_precision, format_character):
34580         Adjust.
34581         (parse_directive): Use INTEGRAL_TYPE_P() rather than testing for
34582         INTEGER_TYPE.
34584 2017-04-11  Richard Earnshaw  <rearnsha@arm.com>
34586         PR target/80389
34587         * config/arm/arm.c (arm_configure_build_target): When -mcpu and -arch
34588         conflict, set target->arch_name instead of target->cpu_name.
34590 2017-04-11  Richard Biener  <rguenther@suse.de>
34592         PR tree-optimization/80374
34593         * tree-ssa-dom.c (derive_equivalences_from_bit_ior): Use
34594         build_zero_cst, remove fold_convertible_p check again.
34596 2017-04-11  Martin Liska  <mliska@suse.cz>
34598         PR sanitizer/70878
34599         * ubsan.c (instrument_object_size): Do not instrument register
34600         variables.
34602 2017-04-11  Jakub Jelinek  <jakub@redhat.com>
34604         PR target/80381
34605         * config/i386/i386-builtin-types.def
34606         (V16HI_FTYPE_V16HI_INT_V16HI_UHI_COUNT,
34607         V16HI_FTYPE_V16HI_V8HI_V16HI_UHI_COUNT,
34608         V16SI_FTYPE_V16SI_INT_V16SI_UHI_COUNT,
34609         V16SI_FTYPE_V16SI_V4SI_V16SI_UHI_COUNT,
34610         V2DI_FTYPE_V2DI_INT_V2DI_UQI_COUNT,
34611         V2DI_FTYPE_V2DI_V2DI_V2DI_UQI_COUNT,
34612         V32HI_FTYPE_V32HI_INT_V32HI_USI_COUNT,
34613         V32HI_FTYPE_V32HI_V8HI_V32HI_USI_COUNT,
34614         V4DI_FTYPE_V4DI_INT_V4DI_UQI_COUNT,
34615         V4DI_FTYPE_V4DI_V2DI_V4DI_UQI_COUNT,
34616         V4SI_FTYPE_V4SI_INT_V4SI_UQI_COUNT,
34617         V4SI_FTYPE_V4SI_V4SI_V4SI_UQI_COUNT,
34618         V8DI_FTYPE_V8DI_INT_V8DI_UQI_COUNT,
34619         V8DI_FTYPE_V8DI_V2DI_V8DI_UQI_COUNT,
34620         V8HI_FTYPE_V8HI_INT_V8HI_UQI_COUNT,
34621         V8HI_FTYPE_V8HI_V8HI_V8HI_UQI_COUNT,
34622         V8SI_FTYPE_V8SI_INT_V8SI_UQI_COUNT,
34623         V8SI_FTYPE_V8SI_V4SI_V8SI_UQI_COUNT): New function type aliases.
34624         * config/i386/i386-builtin.def (__builtin_ia32_pslld512_mask,
34625         __builtin_ia32_pslldi512_mask, __builtin_ia32_psllq512_mask,
34626         __builtin_ia32_psllqi512_mask, __builtin_ia32_psrad512_mask,
34627         __builtin_ia32_psradi512_mask, __builtin_ia32_psraq512_mask,
34628         __builtin_ia32_psraqi512_mask, __builtin_ia32_psrld512_mask,
34629         __builtin_ia32_psrldi512_mask, __builtin_ia32_psrlq512_mask,
34630         __builtin_ia32_psrlqi512_mask, __builtin_ia32_psllwi128_mask,
34631         __builtin_ia32_pslldi128_mask, __builtin_ia32_psllqi128_mask,
34632         __builtin_ia32_psllw128_mask, __builtin_ia32_pslld128_mask,
34633         __builtin_ia32_psllq128_mask, __builtin_ia32_psllwi256_mask,
34634         __builtin_ia32_psllw256_mask, __builtin_ia32_pslldi256_mask,
34635         __builtin_ia32_pslld256_mask, __builtin_ia32_psllqi256_mask,
34636         __builtin_ia32_psllq256_mask, __builtin_ia32_psradi128_mask,
34637         __builtin_ia32_psrad128_mask, __builtin_ia32_psradi256_mask,
34638         __builtin_ia32_psrad256_mask, __builtin_ia32_psraqi128_mask,
34639         __builtin_ia32_psraq128_mask, __builtin_ia32_psraqi256_mask,
34640         __builtin_ia32_psraq256_mask, __builtin_ia32_psrldi128_mask,
34641         __builtin_ia32_psrld128_mask, __builtin_ia32_psrldi256_mask,
34642         __builtin_ia32_psrld256_mask, __builtin_ia32_psrlqi128_mask,
34643         __builtin_ia32_psrlq128_mask, __builtin_ia32_psrlqi256_mask,
34644         __builtin_ia32_psrlq256_mask, __builtin_ia32_psrawi256_mask,
34645         __builtin_ia32_psraw256_mask, __builtin_ia32_psrawi128_mask,
34646         __builtin_ia32_psraw128_mask, __builtin_ia32_psrlwi256_mask,
34647         __builtin_ia32_psrlw256_mask, __builtin_ia32_psrlwi128_mask,
34648         __builtin_ia32_psrlw128_mask, __builtin_ia32_psllwi512_mask,
34649         __builtin_ia32_psllw512_mask, __builtin_ia32_psrawi512_mask,
34650         __builtin_ia32_psraw512_mask, __builtin_ia32_psrlwi512_mask,
34651         __builtin_ia32_psrlw512_mask): Use _COUNT suffixed function type
34652         aliases.
34653         * config/i386/i386.c (ix86_expand_args_builtin): Rename last_arg_count
34654         flag to second_arg_count, handle 4 argument function type _COUNT
34655         aliases, handle second_arg_count on second argument rather than last.
34657 2017-04-10  Jeff Law  <law@redhat.com>
34659         PR tree-optimization/80374
34660         * tree-ssa-dom.c (derive_equivalences_from_bit_ior): Do not try to
34661         record anything if we can not convert integer_zero_node to the
34662         desired type.
34664 2017-04-10  Kelvin Nilsen  <kelvin@gcc.gnu.org>
34666         PR target/80108
34667         * config/rs6000/rs6000.c (rs6000_option_override_internal):
34668         Enhance special handling given to the TARGET_P9_MINMAX option in
34669         relation to certain other options.
34671 2017-04-10  Bin Cheng  <bin.cheng@arm.com>
34673         PR tree-optimization/80153
34674         * tree-ssa-loop-ivopts.c (add_iv_candidate_for_use): Check and
34675         remove POINTER_PLUS_EXPR's base part directly, rather than through
34676         aff_tree.
34678 2017-04-10  Richard Biener  <rguenther@suse.de>
34679             Bin Cheng  <bin.cheng@arm.com>
34681         PR tree-optimization/80153
34682         * tree-affine.c (aff_combination_to_tree): Get base pointer from
34683         the first element of pointer type aff_tree.  Build result expr in
34684         aff_tree's type.
34685         (add_elt_to_tree): Convert to type unconditionally.  Remove other
34686         fold_convert calls.
34687         * tree-ssa-loop-ivopts.c (alloc_iv): Pass in consistent types.
34688         (rewrite_use_nonlinear_expr): Check invariant using iv information.
34690 2017-04-10  Richard Biener  <rguenther@suse.de>
34692         * tree-ssa-structalias.c (find_func_aliases): Properly handle
34693         asm inputs.
34695 2017-04-10  Vladimir Makarov  <vmakarov@redhat.com>
34697         PR rtl-optimization/70478
34698         * lra-constraints.c (curr_small_class_check): New.
34699         (update_and_check_small_class_inputs): New.
34700         (process_alt_operands): Update curr_small_class_check.  Disfavor
34701         alternative insn memory operands.  Check available regs for small
34702         class operands.
34704 2017-03-31  Matthew Fortune  <matthew.fortune@imgtec.com>
34706         PR target/80057
34707         * config/mips/mips.opt (-mvirt): Update description.
34708         * doc/invoke.texi (-mvirt): Likewise.
34710 2017-04-10  Richard Biener  <rguenther@suse.de>
34712         PR middle-end/80362
34713         * fold-const.c (fold_binary_loc): Look at unstripped ops when
34714         looking for NEGATE_EXPR in -A / -B to A / B folding.
34716 2017-04-10  Martin Liska  <mliska@suse.cz>
34718         PR gcov-profile/80224
34719         * gcov.c (print_usage): Fix usage string.
34720         (get_gcov_intermediate_filename): Remove.
34721         (output_gcov_file): Use both for normal and intermediate format.
34722         (generate_results): Do not initialize special file for
34723         intermediate format.
34725 2017-04-10  Richard Biener  <rguenther@suse.de>
34727         PR tree-optimization/80304
34728         * tree-ssa-loop-im.c (ref_indep_loop_p_1): Also recurse
34729         for safelen.
34731 2017-04-10  Nathan Sidwell  <nathan@acm.org>
34733         PR target/79905
34734         * config/rs6000/rs6000.c (rs6000_vector_type): New.
34735         (rs6000_init_builtins): Use it.
34737 2016-04-10  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
34739         * config/arm/arm.md (<mrc>): Add mode to SET source.
34740         (<mrrc>): Likewise.
34742 2017-04-10  Richard Biener  <rguenther@suse.de>
34744         PR middle-end/80344
34745         * gimplify.c (is_gimple_mem_rhs_or_call): Allow CLOBBERs.
34747 2017-04-10  Jakub Jelinek  <jakub@redhat.com>
34749         PR target/80324
34750         * config/i386/avx512fintrin.h (_mm512_reduce_add_epi32,
34751         _mm512_reduce_mul_epi32, _mm512_reduce_and_epi32,
34752         _mm512_reduce_or_epi32, _mm512_mask_reduce_add_epi32,
34753         _mm512_mask_reduce_mul_epi32, _mm512_mask_reduce_and_epi32,
34754         _mm512_mask_reduce_or_epi32, _mm512_reduce_min_epi32,
34755         _mm512_reduce_max_epi32, _mm512_reduce_min_epu32,
34756         _mm512_reduce_max_epu32, _mm512_mask_reduce_min_epi32,
34757         _mm512_mask_reduce_max_epi32, _mm512_mask_reduce_min_epu32,
34758         _mm512_mask_reduce_max_epu32, _mm512_reduce_add_ps,
34759         _mm512_reduce_mul_ps, _mm512_mask_reduce_add_ps,
34760         _mm512_mask_reduce_mul_ps, _mm512_reduce_min_ps, _mm512_reduce_max_ps,
34761         _mm512_mask_reduce_min_ps, _mm512_mask_reduce_max_ps,
34762         _mm512_reduce_add_epi64, _mm512_reduce_mul_epi64,
34763         _mm512_reduce_and_epi64, _mm512_reduce_or_epi64,
34764         _mm512_mask_reduce_add_epi64, _mm512_mask_reduce_mul_epi64,
34765         _mm512_mask_reduce_and_epi64, _mm512_mask_reduce_or_epi64,
34766         _mm512_reduce_min_epi64, _mm512_reduce_max_epi64,
34767         _mm512_mask_reduce_min_epi64, _mm512_mask_reduce_max_epi64,
34768         _mm512_reduce_min_epu64, _mm512_reduce_max_epu64,
34769         _mm512_mask_reduce_min_epu64, _mm512_mask_reduce_max_epu64,
34770         _mm512_reduce_add_pd, _mm512_reduce_mul_pd, _mm512_mask_reduce_add_pd,
34771         _mm512_mask_reduce_mul_pd, _mm512_reduce_min_pd, _mm512_reduce_max_pd,
34772         _mm512_mask_reduce_min_pd, _mm512_mask_reduce_max_pd): New intrinsics.
34774 2017-04-08  Vladimir Makarov  <vmakarov@redhat.com>
34776         PR rtl-optimization/70478
34777         * lra-constraints.c: Reverse the last patch.
34779 2017-04-08  Andreas Tobler  <andreast@gcc.gnu.org>
34781         * config/aarch64/aarch64-freebsd.h: Define MCOUNT_NAME.
34782         Add comment for WCHAR_T.
34784 2017-04-08  Martin Liska  <mliska@suse.cz>
34786         Revert:
34787         2017-04-07  Martin Liska  <mliska@suse.cz>
34789         PR ipa/80212
34790         * ipa-split.c (split_function): Add function part to a same comdat
34791         group.
34793 2017-04-08  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
34795         PR target/80358
34796         * config/rs6000/rs6000.c (expand_block_compare): Fix boundary check.
34798 2017-04-07  Pat Haugen  <pthaugen@us.ibm.com>
34800         * rs6000/rs6000.c (vec_load_pendulum): Rename...
34801         (vec_pairing): ...to this.
34802         (power9_sched_reorder2): Rewrite code for pairing vector/vecload insns.
34803         (rs6000_sched_init): Adjust for name change.
34804         (struct rs6000_sched_context): Likewise.
34805         (rs6000_init_sched_context): Likewise.
34806         (rs6000_set_sched_context): Likewise.
34808 2017-04-07  Jakub Jelinek  <jakub@redhat.com>
34810         PR target/80322
34811         PR target/80323
34812         PR target/80325
34813         PR target/80326
34814         * config/i386/avxintrin.h (_mm256_cvtsd_f64, _mm256_cvtss_f32): New
34815         intrinsics.
34816         * config/i386/avx512fintrin.h (_mm512_int2mask, _mm512_mask2int,
34817         _mm512_abs_ps, _mm512_mask_abs_ps, _mm512_abs_pd, _mm512_mask_abs_pd,
34818         _mm512_cvtsd_f64, _mm512_cvtss_f32): Likewise.
34820 2017-04-07  Andreas Tobler  <andreast@gcc.gnu.org>
34822         * config/aarch64/aarch64-freebsd.h: Define WCHAR_TYPE.
34824 2017-04-07  Vladimir Makarov  <vmakarov@redhat.com>
34826         PR rtl-optimization/70703
34827         * ira-color.c (update_conflict_hard_regno_costs): Use
34828         int64_t instead of HOST_WIDE_INT.
34830 2017-04-07  Vladimir Makarov  <vmakarov@redhat.com>
34832         PR rtl-optimization/70478
34833         * lra-constraints.c (process_alt_operands): Disfavor alternative
34834         insn memory operands.
34836 2017-04-07  Jeff Law  <law@redhat.com>
34838         * config/iq2000/iq2000.c (final_prescan_insn): Do not separate a
34839         CALL and NOTE_INSN_CALL_ARG_LOCATION.
34841 2017-04-07  Martin Liska  <mliska@suse.cz>
34843         PR target/79889
34844         * config/aarch64/aarch64.c (aarch64_process_target_attr):
34845         Show error message instead of an ICE.
34847 2017-04-07  Martin Liska  <mliska@suse.cz>
34849         PR ipa/80212
34850         * ipa-split.c (split_function): Add function part to a same comdat
34851         group.
34853 2017-04-07  Richard Biener  <rguenther@suse.de>
34855         PR middle-end/80341
34856         * tree.c (get_unwidened): Also handle ! for_type case for
34857         INTEGER_CSTs.
34858         * convert.c (do_narrow): Split out from ...
34859         (convert_to_integer_1): ... here.  Do not pass final truncation
34860         type to get_unwidened for TRUNC_DIV_EXPR.
34862 2017-04-07  Richard Biener  <rguenther@suse.de>
34864         * tree-affine.c (wide_int_ext_for_comb): Take type rather
34865         than aff_tree.
34866         (aff_combination_const): Adjust.
34867         (aff_combination_scale): Likewise.
34868         (aff_combination_add_elt): Likewise.
34869         (aff_combination_add_cst): Likewise.
34870         (aff_combination_convert): Likewise.
34871         (add_elt_to_tree): Likewise.  Remove unused argument.
34872         (aff_combination_to_tree): Adjust calls to add_elt_to_tree.
34874 2017-04-07  Sebastian Huber  <sebastian.huber@embedded-brains.de>
34876         * config/arm/arm.h (ARM_DEFAULT_SHORT_ENUMS): Provide default
34877         definition.
34878         * config/arm/arm.c (arm_default_short_enums): Use
34879         ARM_DEFAULT_SHORT_ENUMS.
34880         * config/arm/rtems.h (ARM_DEFAULT_SHORT_ENUMS): Define.
34882 2017-04-06  Jakub Jelinek  <jakub@redhat.com>
34884         PR debug/80234
34885         * dwarf2out.c (gen_member_die): Handle C++17 inline static data
34886         members with redundant out-of-class redeclaration.
34888 2017-04-06  Uros Bizjak  <ubizjak@gmail.com>
34890         PR target/80286
34891         * config/i386/sse.md (*vec_extractv4si_0_zext_sse4): New pattern.
34892         * config/i386/i386.md (*zero_extendsidi2):
34893         Add (?*x,*x) and (?*v,*v) alternatives.
34895 2017-04-06  Uros Bizjak  <ubizjak@gmail.com>
34897         PR target/79733
34898         * config/i386/i386.c (ix86_expand_builtin)
34899         <case IX86_BUILTIN_K{,OR}TEST{C,Z}{8,16,32,64}>: Determine insn operand
34900         mode from insn data. Convert operands to insn operand mode.
34901         Copy operands that don't satisfy insn predicate to a register.
34903 2017-04-06  Sam Thursfield  <sam.thursfield@codethink.co.uk>
34905         * config/rs6000/x-aix: Increase memory limit for genautomata on AIX.
34906         Update comments.
34908 2017-04-06  Richard Biener  <rguenther@suse.de>
34910         PR tree-optimization/80334
34911         * tree-ssa-loop-ivopts.c (rewrite_use_address): Properly
34912         preserve alignment of accesses.
34914 2017-04-06  Richard Biener  <rguenther@suse.de>
34916         PR tree-optimization/80262
34917         * tree-sra.c (build_ref_for_offset): Preserve address-space
34918         information.
34919         * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address):
34920         Drop useless address-space information on MEM_REF offsets.
34922 2017-04-05  Andreas Schwab  <schwab@linux-m68k.org>
34924         * builtins.def (BUILT_IN_UPDATE_SETJMP_BUF): Fix type.
34926 2017-04-05  Vladimir Makarov  <vmakarov@redhat.com>
34928         PR rtl-optimization/70703
34929         * ira-color.c (update_conflict_hard_regno_costs): Use
34930         HOST_WIDE_INT instead of long.
34932 2017-04-05  Uros Bizjak  <ubizjak@gmail.com>
34934         PR target/80298
34935         * config/i386/mmintrin.h: Add -msse target option when __SSE__ is
34936         not defined for x86_64 target.  Add -mmmx target option when __SSE2__
34937         is not defined.
34938         * config/i386/mm3dnow.h: Add -msse target when __SSE__ is not defined
34939         for x86_64 target.  Handle -m3dnowa option.
34941 2017-04-05  Vladimir Makarov  <vmakarov@redhat.com>
34943         PR rtl-optimization/70703
34944         * ira-color.c (update_costs_from_allocno): Use the smallest mode.
34945         (update_conflict_hard_regno_costs): Use long instead of unsigned
34946         arithmetic for cost calculation.
34948 2017-04-05  Jakub Jelinek  <jakub@redhat.com>
34949             Bernd Edlinger  <bernd.edlinger@hotmail.de>
34951         PR sanitizer/80308
34952         * asan.c (asan_store_shadow_bytes): Fix location of last_chunk_value
34953         for big endian.
34955 2017-04-05  Eric Botcazou  <ebotcazou@adacore.com>
34957         PR target/78002
34958         * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Replace
34959         ptr_mode with Pmode throughout.
34960         * config/aarch64/aarch64.md (probe_stack_range_<PTR:mode): Rename
34961         into probe_stack_range and use DImode.
34963 2017-04-05  Dominik Vogt  <vogt@linux.vnet.ibm.com>
34965         PR target/79890
34966         * config/s390/s390.c (s390_register_info_gprtofpr): Return if
34967         call_eh_return is true.
34969 2017-04-05  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
34971         * config/s390/s390-c.c (s390_resolve_overloaded_builtin):
34972         Initialize last_match_fntype_index.
34974 2017-04-05  Jakub Jelinek  <jakub@redhat.com>
34976         PR target/80310
34977         * tree-nvr.c: Include internal-fn.h.
34978         (pass_return_slot::execute): Ignore internal calls without
34979         direct optab.
34981 2017-04-04  Jakub Jelinek  <jakub@redhat.com>
34982             Richard Biener  <rguenther@suse.de>
34984         PR c++/80297
34985         * genmatch.c (capture::gen_transform): For GENERIC unshare_expr
34986         captures used multiple times, except for the last use.
34987         * generic-match-head.c: Include gimplify.h.
34989 2017-04-04  Jakub Jelinek  <jakub@redhat.com>
34991         PR tree-optimization/79390
34992         * target.h (struct noce_if_info): Declare.
34993         * targhooks.h (default_noce_conversion_profitable_p): Declare.
34994         * target.def (noce_conversion_profitable_p): New target hook.
34995         * ifcvt.h (struct noce_if_info): New type, moved from ...
34996         * ifcvt.c (struct noce_if_info): ... here.
34997         (noce_conversion_profitable_p): Renamed to ...
34998         (default_noce_conversion_profitable_p): ... this.  No longer
34999         static nor inline.
35000         (noce_try_store_flag_constants, noce_try_addcc,
35001         noce_try_store_flag_mask, noce_try_cmove, noce_try_cmove_arith,
35002         noce_convert_multiple_sets): Use targetm.noce_conversion_profitable_p
35003         instead of noce_conversion_profitable_p.
35004         * config/i386/i386.c: Include ifcvt.h.
35005         (ix86_option_override_internal): Don't override
35006         PARAM_MAX_RTL_IF_CONVERSION_INSNS default.
35007         (ix86_noce_conversion_profitable_p): New function.
35008         (TARGET_NOCE_CONVERSION_PROFITABLE_P): Redefine.
35009         * config/i386/x86-tune.def (X86_TUNE_ONE_IF_CONV_INSN): Adjust comment.
35010         * doc/tm.texi.in (TARGET_NOCE_CONVERSION_PROFITABLE_P): Add.
35011         * doc/tm.texi: Regenerated.
35013 2017-04-04  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
35015         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Grammar
35016         correction.
35018 2017-04-04  Thomas Preud'homme  <thomas.preudhomme@arm.com>
35020         PR target/80307
35021         * config/arm/arm.c (thumb1_rtx_costs): Give a cost of 32
35022         instructions for small multiply cores.
35024 2017-04-04  Jeff Law  <law@redhat.com>
35026         * config/mips/mips.c (mips_multi_add): Zero initialize the newly
35027         added member.
35028         (mips_expand_vec_perm_const): Initialize elements in orig_perm
35029         that are not set by the loop over the elements.
35031 2017-04-04  Jakub Jelinek  <jakub@redhat.com>
35033         PR target/80286
35034         * config/i386/i386.c (ix86_expand_args_builtin): If op has scalar
35035         int mode, convert_modes it to mode as unsigned, otherwise use
35036         lowpart_subreg to mode rather than SImode.
35037         * config/i386/sse.md (<mask_codefor>ashr<mode>3<mask_name>,
35038         ashr<mode>3, ashr<mode>3<mask_name>, <shift_insn><mode>3<mask_name>):
35039         Use DImode instead of SImode for the shift count operand.
35040         * config/i386/mmx.md (mmx_ashr<mode>3, mmx_<shift_insn><mode>3):
35041         Likewise.
35043 2017-04-04  Richard Biener  <rguenther@suse.de>
35045         PR middle-end/80281
35046         * match.pd (A + (-B) -> A - B): Make sure to preserve unsigned
35047         arithmetic done for the negate or the plus.  Simplify.
35048         (A - (-B) -> A + B): Likewise.
35049         * fold-const.c (split_tree): Make sure to not negate pointers.
35051 2017-04-04  Segher Boessenkool  <segher@kernel.crashing.org>
35053         PR rtl-optimization/60818
35054         * simplify-rtx.c (simplify_binary_operation_1): Do not replace
35055         a compare of comparisons with the thing compared if this results
35056         in a different machine mode.
35058 2017-04-03  Jonathan Wakely  <jwakely@redhat.com>
35060         * alias.c (base_alias_check): Fix typo in comment.
35061         * cgraph.h (class ipa_polymorphic_call_context): Likewise.
35062         * cgraphunit.c (symbol_table::compile): Likewise.
35063         * collect2.c (maybe_run_lto_and_relink): Likewise.
35064         * config/arm/arm.c (arm_thumb1_mi_thunk): Likewise.
35065         * config/avr/avr-arch.h (avr_arch_info_t): Likewise.
35066         * config/avr/avr.c (avr_map_op_t): Likewise.
35067         * config/cr16/cr16.h (DATA_ALIGNMENT): Likewise.
35068         * config/epiphany/epiphany.c (TARGET_ARG_PARTIAL_BYTES): Likewise.
35069         * config/epiphany/epiphany.md (movcc): Likewise.
35070         * config/i386/i386.c (legitimize_pe_coff_extern_decl): Likewise.
35071         * config/m68k/m68k.c (struct _sched_ib, m68k_sched_variable_issue):
35072         Likewise.
35073         * config/mips/mips.c (mips_save_restore_reg): Likewise.
35074         * config/rx/rx.c (rx_is_restricted_memory_address): Likewise.
35075         * config/s390/s390.c (Z10_EARLYLOAD_DISTANCE): Likewise.
35076         * config/sh/sh.c (sh_rtx_costs): Likewise.
35077         * fold-const.c (fold_truth_andor): Likewise.
35078         * genautomata.c (collapse_flag): Likewise.
35079         * gengtype.h (struct type::u::s): Likewise.
35080         * gensupport.c (has_subst_attribute, add_mnemonic_string): Likewise.
35081         * input.c (FORMAT_AMOUNT): Likewise.
35082         * ipa-cp.c (class ipcp_lattice, agg_replacements_to_vector)
35083         (known_aggs_to_agg_replacement_list): Likewise.
35084         * ipa-inline-analysis.c: Likewise.
35085         * ipa-inline.h (estimate_edge_time, estimate_edge_hints): Likewise.
35086         * ipa-polymorphic-call.c
35087         (ipa_polymorphic_call_context::restrict_to_inner_class): Likewise.
35088         * loop-unroll.c (analyze_insn_to_expand_var): Likewise.
35089         * lra.c (lra_optional_reload_pseudos, lra_subreg_reload_pseudos):
35090         Likewise.
35091         * modulo-sched.c (apply_reg_moves): Likewise.
35092         * omp-expand.c (build_omp_regions_1): Likewise.
35093         * trans-mem.c (struct tm_wrapper_hasher): Likewise.
35094         * tree-ssa-loop-ivopts.c (may_eliminate_iv): Likewise.
35095         * tree-ssa-loop-niter.c (maybe_lower_iteration_bound): Likewise.
35096         * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
35097         * value-prof.c: Likewise.
35098         * var-tracking.c (val_reset): Likewise.
35100 2017-04-03  Richard Biener  <rguenther@suse.de>
35102         PR tree-optimization/80275
35103         * fold-const.c (split_address_to_core_and_offset): Handle
35104         POINTER_PLUS_EXPR.
35106 2017-04-03  Eric Botcazou  <ebotcazou@adacore.com>
35108         * tree-nested.c (get_descriptor_type): Make sure that the alignment of
35109         descriptors is at least equal to that of functions.
35111 2017-04-02  Uros Bizjak  <ubizjak@gmail.com>
35113         * config/i386/sse.md (movdi_to_sse): Add missing DONE.
35115 2017-04-02  Uros Bizjak  <ubizjak@gmail.com>
35117         PR target/80250
35118         * config/i386/sse.md (mov<IMOD4:mode>): Remove insn pattern.
35119         (mov<IMOD4:mode>): New expander.
35120         (*mov<IMOD4:mode>_internal): New insn and split pattern.
35122 2017-03-31  Segher Boessenkool  <segher@kernel.crashing.org>
35124         PR rtl-optimization/79405
35125         * fwprop.c (propagations_left): New variable.
35126         (forward_propagate_into): Decrement it.
35127         (fwprop_init): Initialize it.
35128         (fw_prop): If the variable has reached zero, stop propagating.
35129         (fwprop_addr): Ditto.
35131 2017-03-31  Jakub Jelinek  <jakub@redhat.com>
35133         PR debug/79255
35134         * dwarf2out.c (decls_for_scope): If BLOCK_NONLOCALIZED_VAR is
35135         a FUNCTION_DECL, pass it as decl instead of origin to
35136         process_scope_var.
35138 2017-03-31  Alexander Monakov  <amonakov@ispras.ru>
35140         * config/nvptx/nvptx.c (nvptx_output_softstack_switch): Correct format
35141         string.
35143 2017-03-31  Pat Haugen  <pthaugen@us.ibm.com>
35145         PR target/80107
35146         * config/rs6000/rs6000.md (extendhi<mode>2): Add test for
35147         TARGET_VSX_SMALL_INTEGER.
35149 2017-03-31  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
35151         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Add
35152         reference to the OpenPOWER 64-Bit ELF V2 ABI Specification.
35154 2017-03-31  Matthew Fortune  <matthew.fortune@imgtec.com>
35156         * config/mips/mips-msa.md (msa_vec_extract_<msafmt_f>): Update
35157         extraction from odd-numbered MSA register.
35159 2017-03-31  Jakub Jelinek  <jakub@redhat.com>
35161         PR middle-end/80173
35162         * expmed.c (store_bit_field_1): Don't attempt to create
35163         a word subreg out of hard registers wider than word if they
35164         have HARD_REGNO_NREGS of 1 for their mode.
35166         PR middle-end/80163
35167         * varasm.c (initializer_constant_valid_p_1): Disallow sign-extending
35168         conversions to integer types wider than word and pointer.
35170         PR debug/80025
35171         * cselib.h (rtx_equal_for_cselib_1): Add depth argument.
35172         (rtx_equal_for_cselib_p): Pass 0 to it.
35173         * cselib.c (cselib_hasher::equal): Likewise.
35174         (rtx_equal_for_cselib_1): Add depth argument.  If depth
35175         is 128, don't look up VALUE locs and punt.  Increment
35176         depth in recursive calls when walking VALUE locs.
35178 2017-03-31  Bernd Edlinger  <bernd.edlinger@hotmail.de>
35180         * gcov.c (md5sum_to_hex): Fix output of MD5 hex bytes.
35181         (make_gcov_file_name): Use the canonical path name for generating
35182         the MD5 value.
35183         (read_line): Fix handling of files with ascii null bytes.
35185 2017-03-30  Matthew Fortune  <matthew.fortune@imgtec.com>
35187         * config/mips/mips.c (mips_expand_vector_init): Create a const_vector
35188         to initialise a vector register instead
35189         of using a const_int.
35191 2017-03-30  Jakub Jelinek  <jakub@redhat.com>
35193         PR translation/80189
35194         * gimplify.c (omp_default_clause): Use %qs instead of %s in
35195         diagnostic messages.
35197 2017-03-30  Peter Bergner  <bergner@vnet.ibm.com>
35199         PR target/80246
35200         * config/rs6000/dfp.md (dfp_dxex_<mode>): Update mode of operand 0.
35201         (dfp_diex_<mode>): Update mode of operand 1.
35202         * doc/extend.texi (dxex, dxexq): Document change to return type.
35203         (diex, diexq): Document change to argument type.
35205 2017-03-30  Martin Jambor  <mjambor@suse.cz>
35207         PR ipa/77333
35208         * cgraph.h (cgraph_build_function_type_skip_args): Declare.
35209         * cgraph.c (redirect_call_stmt_to_callee): Set gimple fntype so that
35210         it reflects the signature changes performed at the callee side.
35211         * cgraphclones.c (build_function_type_skip_args): Make public, renamed
35212         to cgraph_build_function_type_skip_args.
35213         (build_function_decl_skip_args): Adjust call to the above function.
35215 2017-03-30  Jakub Jelinek  <jakub@redhat.com>
35217         PR target/80206
35218         * config/i386/sse.md
35219         (<extract_type>_vextract<shuffletype><extract_suf>_mask): Use
35220         register as dest whenever it is a MEM not rtx_equal_p to the
35221         corresponding dup operand, and when forcing into reg move the
35222         reg into the memory afterwards.
35223         (<extract_type_2>_vextract<shuffletype><extract_suf_2>_mask):
35224         Likewise.  Use <ssehalfvecmode> instead of <ssequartermode>
35225         for the force_reg mode.
35226         (avx512vl_vextractf128<mode>): Use register as dest either
35227         always when a MEM, or when it is a MEM not rtx_equal_p to the
35228         corresponding dup operand, or even not when it is a CONST_VECTOR
35229         depending on the mode and lo vs. hi.
35230         (avx512dq_vextract<shuffletype>64x2_1_maskm): Remove extraneous
35231         parens.
35232         (avx512f_vextract<shuffletype>32x4_1_maskm): Likewise.
35233         (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name>):
35234         Likewise.  Require that operands[2] is even.
35235         (<mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name>):
35236         Remove extraneous parens.  Require that operands[2] is a multiple
35237         of 4.
35238         (vec_extract_lo_<mode><mask_name>): Don't bother testing if
35239         operands[0] is a MEM if <mask_applied>, the predicates/constraints
35240         disallow memory then.
35242 2017-03-30  Richard Biener  <rguenther@suse.de>
35244         PR tree-optimization/77498
35245         * tree-ssa-pre.c (phi_translate_1): Do not allow simplifications
35246         to non-constants over backedges.
35248 2017-03-29  Segher Boessenkool  <segher@kernel.crashing.org>
35250         PR rtl-optimization/80233
35251         * combine.c (combine_instructions): Only take NONDEBUG_INSN_P insns
35252         as last_combined_insn.  Do not test for BARRIER_P separately.
35254 2017-03-29  Andreas Schwab  <schwab@suse.de>
35256         PR ada/80146
35257         * calls.c (prepare_call_address): Convert funexp to Pmode before
35258         copying to temp reg.
35260 2017-03-29  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
35262         PR tree-optimization/80158
35263         * gimple-ssa-strength-reduction.c (replace_mult_candidate):
35264         Handle possible future case of more than one alternate
35265         interpretation.
35266         (replace_rhs_if_not_dup): Likewise.
35267         (replace_one_candidate): Likewise.
35269 2017-03-28  Vladimir Makarov  <vmakarov@redhat.com>
35271         PR rtl-optimization/80193
35272         * ira.c (ira): Do not check allocation for LRA.
35274 2017-03-28  Alexander Monakov  <amonakov@ispras.ru>
35276         * config/nvptx/nvptx-protos.h (nvptx_output_simt_enter): Declare.
35277         (nvptx_output_simt_exit): Declare.
35278         * config/nvptx/nvptx.c (nvptx_init_unisimt_predicate): Use
35279         cfun->machine->unisimt_location.  Handle NULL unisimt_predicate.
35280         (init_softstack_frame): Move initialization of crtl->is_leaf to...
35281         (nvptx_declare_function_name): ...here.  Emit declaration of local
35282         memory space buffer for omp_simt_enter insn.
35283         (nvptx_output_unisimt_switch): New.
35284         (nvptx_output_softstack_switch): New.
35285         (nvptx_output_simt_enter): New.
35286         (nvptx_output_simt_exit): New.
35287         * config/nvptx/nvptx.h (struct machine_function): New fields
35288         has_simtreg, unisimt_location, simt_stack_size, simt_stack_align.
35289         * config/nvptx/nvptx.md (UNSPECV_SIMT_ENTER): New unspec.
35290         (UNSPECV_SIMT_EXIT): Ditto.
35291         (omp_simt_enter_insn): New insn.
35292         (omp_simt_enter): New expansion.
35293         (omp_simt_exit): New insn.
35294         * config/nvptx/nvptx.opt (msoft-stack-reserve-local): New option.
35296         * internal-fn.c (expand_GOMP_SIMT_ENTER): New.
35297         (expand_GOMP_SIMT_ENTER_ALLOC): New.
35298         (expand_GOMP_SIMT_EXIT): New.
35299         * internal-fn.def (GOMP_SIMT_ENTER): New internal function.
35300         (GOMP_SIMT_ENTER_ALLOC): Ditto.
35301         (GOMP_SIMT_EXIT): Ditto.
35302         * target-insns.def (omp_simt_enter): New insn.
35303         (omp_simt_exit): Ditto.
35304         * omp-low.c (struct omplow_simd_context): New fields simt_eargs,
35305         simt_dlist.
35306         (lower_rec_simd_input_clauses): Implement SIMT privatization.
35307         (lower_rec_input_clauses): Likewise.
35308         (lower_lastprivate_clauses): Handle SIMT privatization.
35310         * omp-offload.c: Include langhooks.h, tree-nested.h, stor-layout.h.
35311         (ompdevlow_adjust_simt_enter): New.
35312         (find_simtpriv_var_op): New.
35313         (execute_omp_device_lower): Handle IFN_GOMP_SIMT_ENTER,
35314         IFN_GOMP_SIMT_ENTER_ALLOC, IFN_GOMP_SIMT_EXIT.
35316         * tree-inline.h (struct copy_body_data): New field dst_simt_vars.
35317         * tree-inline.c (expand_call_inline): Handle SIMT privatization.
35318         (copy_decl_for_dup_finish): Ditto.
35320         * tree-ssa.c (execute_update_addresses_taken): Handle GOMP_SIMT_ENTER.
35322 2017-03-28  Uros Bizjak  <ubizjak@gmail.com>
35324         PR target/53383
35325         * config/i386/i386.c (ix86_option_override_internal): Always
35326         allow -mpreferred-stack-boundary=3 for 64-bit targets.
35328 2017-03-28  Bin Cheng  <bin.cheng@arm.com>
35330         * tree-vect-loop.c (optimize_mask_stores): Add bb to the right loop.
35332 2017-03-28  Bin Cheng  <bin.cheng@arm.com>
35334         * tree-vect-loop-manip.c (slpeel_add_loop_guard): New param and
35335         mark new edge's irreducible flag accordign to it.
35336         (vect_do_peeling): Check loop preheader edge's irreducible flag
35337         and pass it to function slpeel_add_loop_guard.
35339 2017-03-28  Richard Sandiford  <richard.sandiford@arm.com>
35341         PR tree-optimization/80218
35342         * tree-call-cdce.c (shrink_wrap_one_built_in_call_with_conds):
35343         Update block frequencies and counts.
35345 2017-03-28  Richard Biener  <rguenther@suse.de>
35347         PR tree-optimization/78644
35348         * tree-ssa-ccp.c (evaluate_stmt): When we may not use the value
35349         of a simplification result we may not use it at all.
35351 2017-03-28  Richard Biener  <rguenther@suse.de>
35353         PR ipa/80205
35354         * tree-inline.c (copy_phis_for_bb): Do not create PHI node
35355         without arguments, generate default definition of a SSA name.
35357 2017-03-28  Richard Biener  <rguenther@suse.de>
35359         PR middle-end/80222
35360         * gimple-fold.c (gimple_fold_indirect_ref): Do not touch
35361         TYPE_REF_CAN_ALIAS_ALL references.
35362         * fold-const.c (fold_indirect_ref_1): Likewise.
35364 2017-03-28  Martin Liska  <mliska@suse.cz>
35366         PR ipa/80104
35367         * cgraphunit.c (cgraph_node::expand_thunk): Mark argument of a
35368         thunk call as DECL_GIMPLE_REG_P when vector or complex type.
35370 2017-03-28  Claudiu Zissulescu  <claziss@synopsys.com>
35371             Thomas Petazzoni  <thomas.petazzoni@free-electrons.com>
35373         * config/arc/arc.h (CPP_SPEC): Add subtarget_cpp_spec.
35374         (EXTRA_SPECS): Define.
35375         (SUBTARGET_EXTRA_SPECS): Likewise.
35376         (SUBTARGET_CPP_SPEC): Likewise.
35377         * config/arc/elf.h (EXTRA_SPECS): Renamed to
35378         SUBTARGET_EXTRA_SPECS.
35379         * config/arc/linux.h (SUBTARGET_CPP_SPEC): Define.
35381 2017-03-28  Claudiu Zissulescu  <claziss@synopsys.com>
35383         * config/arc/simdext.md (vst64_insn): Update pattern.
35384         (vld32wh_insn): Likewise.
35385         (vld32wl_insn): Likewise.
35386         (vld64_insn): Likewise.
35387         (vld32_insn): Likewise.
35389 2017-03-28  Marek Polacek  <polacek@redhat.com>
35391         PR sanitizer/80067
35392         * fold-const.c (fold_comparison): Use protected_set_expr_location
35393         instead of SET_EXPR_LOCATION.
35395 2017-03-28  Markus Trippelsdorf  <markus@trippelsdorf.de>
35397         * tree.c (add_expr): Avoid name lookup warning.
35399 2017-03-27  Jeff Law  <law@redhat.com>
35401         PR tree-optimization/80216
35402         * tree-ssa-dom.c (derive_equivalences_from_bit_ior): Fix typo in
35403         function name.  Limit recursion depth.
35404         (record_temporary_equivalences): Corresponding changes.
35406 2017-03-27  Jonathan Wakely  <jwakely@redhat.com>
35408         * doc/invoke.texi (-Wno-narrowing): Reorder so default behavior is
35409         covered first.
35411 2017-03-27  Jakub Jelinek  <jakub@redhat.com>
35413         PR target/80102
35414         * reg-notes.def (REG_CFA_NOTE): Define.  Use it for CFA related
35415         notes.
35416         * cfgcleanup.c (reg_note_cfa_p): New array.
35417         (insns_have_identical_cfa_notes): New function.
35418         (old_insns_match_p): Don't cross-jump in between /f
35419         and non-/f instructions.  If both i1 and i2 are frame related,
35420         verify all CFA notes, their order and content.
35422 2017-03-27  Michael Meissner  <meissner@linux.vnet.ibm.com>
35424         PR target/78543
35425         * config/rs6000/rs6000.md (bswaphi2_extenddi): Combine bswap
35426         HImode and SImode with zero extend to DImode to one insn.
35427         (bswap<mode>2_extenddi): Likewise.
35428         (bswapsi2_extenddi): Likewise.
35429         (bswaphi2_extendsi): Likewise.
35430         (bswaphi2): Combine bswap HImode and SImode into one insn.
35431         Separate memory insns from swapping register.
35432         (bswapsi2): Likewise.
35433         (bswap<mode>2): Likewise.
35434         (bswaphi2_internal): Delete, no longer used.
35435         (bswapsi2_internal): Likewise.
35436         (bswap<mode>2_load): Split bswap HImode/SImode into separate load,
35437         store, and gpr<-gpr swap insns.
35438         (bswap<mode>2_store): Likewise.
35439         (bswaphi2_reg): Register only splitter, combine with the splitter.
35440         (bswaphi2 splitter): Likewise.
35441         (bswapsi2_reg): Likewise.
35442         (bswapsi2 splitter): Likewise.
35443         (bswapdi2): If we have the LDBRX and STDBRX instructions, split
35444         the insns into load, store, and register/register insns.
35445         (bswapdi2_ldbrx): Likewise.
35446         (bswapdi2_load): Likewise.
35447         (bswapdi2_store): Likewise.
35448         (bswapdi2_reg): Likewise.
35450 2017-03-27  Gunther Nikl  <gnikl@users.sourceforge.net>
35452         * system.h (HAVE_DESIGNATED_INITIALIZERS): Fix non C++ case.
35453         (HAVE_DESIGNATED_UNION_INITIALIZERS): Likewise.
35455 2017-03-27  Kelvin Nilsen  <kelvin@gcc.gnu.org>
35457         PR target/80103
35458         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Edit and
35459         add comments.
35460         * config/rs6000/rs6000.c (rs6000_option_override_internal): Add
35461         special handling for target option conflicts between dform
35462         options (-mpower9-dform, -mpower9-dform-vector,
35463         -mpower9-dform-scalar) and -mno-direct-move.
35465 2017-03-27  Richard Biener  <rguenther@suse.de>
35467         PR tree-optimization/80181
35468         * tree-ssa-ccp.c (likely_value): UNDEFINED ^ X is UNDEFINED.
35470 2017-03-27  Claudiu Zissulescu  <claziss@synopsys.com>
35472         * config/arc/predicates.md (move_double_src_operand): Replace the
35473         call to move_double_src_operand with a call to address_operand.
35475 2017-03-27  Claudiu Zissulescu  <claziss@synopsys.com>
35477         * config/arc/elf.h (ARGET_ARC_TP_REGNO_DEFAULT): Define.
35478         * config/arc/linux.h (ARGET_ARC_TP_REGNO_DEFAULT): Likewise.
35479         * config/arc/arc.opt (mtp-regno): Use ARGET_ARC_TP_REGNO_DEFAULT.
35481 2017-03-27  Claudiu Zissulescu  <claziss@synopsys.com>
35483         * config/arc/predicates.md (long_immediate_loadstore_operand):
35484         Consider scaled addresses cases.
35486 2017-03-27  Claudiu Zissulescu  <claziss@synopsys.com>
35488         * config/arc/arc.c (arc_epilogue_uses): BLINK should be also
35489         restored when in interrupt.
35490         * config/arc/arc.md (simple_return): ARCv2 rtie instruction
35491         doesn't have delay slot.
35493 2017-03-27  Richard Biener  <rguenther@suse.de>
35495         PR ipa/79776
35496         * tree-ssa-structalias.c (associate_varinfo_to_alias): Skip
35497         inlined thunk clones.
35499 2017-03-27  Jakub Jelinek  <jakub@redhat.com>
35501         PR sanitizer/80168
35502         * asan.c (instrument_derefs): Copy over last operand from
35503         original COMPONENT_REF to the new COMPONENT_REF with
35504         DECL_BIT_FIELD_REPRESENTATIVE.
35505         * ubsan.c (instrument_object_size): Likewise.
35507 2017-03-27  Richard Biener  <rguenther@suse.de>
35509         PR tree-optimization/80170
35510         * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Make
35511         sure DR/SCEV didnt fold in constants we do not see when looking
35512         at the reference base alignment.
35514 2017-03-27  Richard Biener  <rguenther@suse.de>
35516         PR middle-end/80171
35517         * gimple-fold.c (fold_ctor_reference): Properly guard against
35518         NULL return value from canonicalize_constructor_val.
35520 2017-03-25  Uros Bizjak  <ubizjak@gmail.com>
35522         PR target/80180
35523         * config/i386/i386.c (ix86_expand_builtin)
35524         <IX86_BUILTIN_RDSEED{16,32,64}_STEP>: Do not expand arg0 between
35525         flags reg setting and flags reg using instructions.
35526         <IX86_BUILTIN_RDRAND{16,32,64}_STEP>: Ditto.  Use non-flags reg
35527         clobbering instructions to zero extend op2.
35529 2017-03-25  Gerald Pfeifer  <gerald@pfeifer.com>
35531         * doc/install.texi (Configuration) <--with-aix-soname>:
35532         Update link to AIX ld.
35534 2017-03-25  Bernd Schmidt  <bschmidt@redhat.com>
35536         PR rtl-optimization/80160
35537         PR rtl-optimization/80159
35538         * lra-assigns.c (must_not_spill_p): Tighten new test to also take
35539         reg_alternate_class into account.
35541 2017-03-24  Vladimir Makarov  <vmakarov@redhat.com>
35543         PR target/80148
35544         * lra-assigns.c (assign_by_spills): Add spilled non-reload pseudos
35545         to consider in curr_insn_transform.
35547 2017-03-24  Jakub Jelinek  <jakub@redhat.com>
35549         * genrecog.c (validate_pattern): Add VEC_SELECT validation.
35550         * genmodes.c (emit_min_insn_modes_c): Call emit_mode_nunits
35551         and emit_mode_inner.
35553 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35555         * config/s390/s390-builtins.def: Add VXE builtins.  Add a flags
35556         argument to the overloaded builtin variants.  Use the new flag to
35557         deprecate certain builtin variants.
35558         * config/s390/s390-builtin-types.def: Add new builtin types.
35559         * config/s390/s390-builtins.h: Support new flags field for
35560         overloaded builtins.
35561         * config/s390/s390-c.c (OB_DEF_VAR): New flags field.
35562         (s390_macro_to_expand): Enable vector float data type.
35563         (s390_cpu_cpp_builtins_internal): Indicate support of the new
35564         builtins by incrementing the __VEC__ version number.
35565         (s390_expand_overloaded_builtin): Support expansion of vec_xl and
35566         vec_xst.
35567         (s390_resolve_overloaded_builtin): Emit error messages depending
35568         on the builtin flags.
35569         * config/s390/s390.c (s390_expand_builtin): Support additional
35570         flags argument.  Change error message to match the messages
35571         emitted in s390-c.c.
35572         * config/s390/s390.md: New UNSPEC_* constants.
35573         (op_type): Add new instruction types.
35574         * config/s390/vecintrin.h: Add new builtins and test data class
35575         constants.
35576         * config/s390/vx-builtins.md (V_HW_32_64): Add V4SF.
35577         (V_HW_4, VEC_HW, VECF_HW): New mode iterators.
35578         (VEC_INEXACT, VEC_NOINEXACT): New constants.
35579         ("vec_splats<mode>", "vec_insert<mode>", "vec_promote<mode>")
35580         ("vec_insert_and_zero<mode>", "vec_mergeh<mode>")
35581         ("vec_mergel<mode>"): V_HW -> VEC_HW.
35583         ("vlrlrv16qi", "vstrlrv16qi", "vbpermv16qi", "vec_msumv2di")
35584         ("vmslg", "*vftci<mode>_cconly", "vftci<mode>_intcconly")
35585         ("*vftci<mode>", "vftci<mode>_intcc", "vec_double_s64")
35586         ("vec_double_u64", "vfmin<mode>", "vfmax<mode>"): New definition.
35588         ("and_av2df3", "and_cv2df3", "vec_andc_av2df3")
35589         ("vec_andc_cv2df3", "xor_av2df3", "xor_cv2df3", "vec_nor_av2df3")
35590         ("vec_nor_cv2df3", "ior_av2df3", "ior_cv2df3", "vec_nabs")
35591         ("*vftcidb", "*vftcidb_cconly", "vftcidb"): Remove definition.
35593         ("vec_all_<fpcmpcc:code>v2df", "vec_any_<fpcmpcc:code>v2df")
35594         ("vec_scatter_elementv4si_DI", "vec_cmp<fpcmp:code>v2df")
35595         ("vec_di_to_df_s64", "vec_di_to_df_u64", "vec_df_to_di_u64")
35596         ("vfidb", "*vldeb", "*vledb", "*vec_cmp<insn_cmp>v2df_cconly")
35597         ("vec_cmpeqv2df_cc", "vec_cmpeqv2df_cc", "vec_cmphv2df_cc")
35598         ("vec_cmphev2df_cc", "*vec_cmpeqv2df_cc")
35599         ("*vec_cmphv2df_cc", "*vec_cmphev2df_cc"): Enable new modes as ...
35601         ("vec_all_<fpcmpcc:code><mode>", "vec_any_<fpcmpcc:code><mode>")
35602         ("vec_scatter_element<V_HW_4:mode>_DI")
35603         ("vec_cmp<fpcmp:code><mode>", "vcdgb", "vcdlgb", "vclgdb")
35604         ("vec_fpint<mode>", "vflls")
35605         ("vflrd", "*vec_cmp<insn_cmp><mode>_cconly", "vec_cmpeq<mode>_cc")
35606         ("vec_cmpeq<mode>_cc", "vec_cmph<mode>_cc", "vec_cmphe<mode>_cc")
35607         ("*vec_cmpeq<mode>_cc", "*vec_cmph<mode>_cc")
35608         ("*vec_cmphe<mode>_cc"): ... these.
35610         ("vec_ctd_s64", "vec_ctsl", "vec_ctul", "vec_st2f"): Use rounding
35611         mode constant instead of magic value.
35613 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35615         * config/s390/s390.c (s390_expand_vec_compare): Support other
35616         vector floating point modes than just V2DF.
35617         (s390_expand_vcond): Likewise.
35618         (s390_hard_regno_mode_ok): Allow SFmode values in VRs.
35619         (s390_cannot_change_mode_class): Prevent mode changes between TF
35620         and V1TF in vector registers.
35621         * config/s390/s390.md (DF, SF): New mode attributes.
35622         ("*cmp<mode>_ccs", "add<mode>3", "sub<mode>3", "mul<mode>3")
35623         ("fma<mode>4", "fms<mode>4", "div<mode>3", "*neg<mode>2"): Add
35624         SFmode support for VRs.
35625         * config/s390/vector.md (V_HW, V_HW2, VT_HW, ti*, nonvec): Add new
35626         vector fp modes.
35627         (VFT, VF_HW): New mode iterators.
35628         (vw, sdx): New mode attributes.
35629         ("addv2df3", "subv2df3", "mulv2df3", "divv2df3", "sqrtv2df2")
35630         ("fmav2df4","fmsv2df4", "negv2df2", "absv2df2", "*negabsv2df2")
35631         ("smaxv2df3", "sminv2df3", "*vec_cmp<VFCMP_HW_OP:code>v2df_nocc")
35632         ("vec_cmpuneqv2df", "vec_cmpltgtv2df", "vec_orderedv2df")
35633         ("vec_unorderedv2df"): Adjust the v2df only patterns to support
35634         also the new vector floating point modes.  Renaming to ...
35636         ("add<mode>3", "sub<mode>3", "mul<mode>3", "div<mode>3")
35637         ("sqrt<mode>2", "fma<mode>4", "fms<mode>4", "neg<mode>2")
35638         ("abs<mode>2", "negabs<mode>2", "smax<mode>3")
35639         ("smin<mode>3", "*vec_cmp<VFCMP_HW_OP:code><mode>_nocc")
35640         ("vec_cmpuneq<mode>", "vec_cmpltgt<mode>", "vec_ordered<mode>")
35641         ("vec_unordered<mode>"): ... these.
35643         ("neg_fma<mode>4", "neg_fms<mode>4", "*smax<mode>3_vxe")
35644         ("*smin<mode>3_vxe", "*sminv2df3_vx", "*vec_extendv4sf")
35645         ("*vec_extendv2df"): New insn definitions.
35647 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35649         * config/s390/s390.md ("*adddi3_sign", "*subdi3_sign", "mulditi3")
35650         ("mulditi3_2", "*muldi3_sign"): New patterns.
35651         ("muldi3", "*muldi3", "mulsi3", "*mulsi3"): Add an expander and
35652         rename the pattern definition.
35654 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35656         * config/s390/s390.md ("indirect_jump"): Turn insn definition into
35657         expander.
35658         ("*indirect_jump", "*indirect2_jump"): New pattern definitions.
35660 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35662         * config/s390/s390.c (s390_expand_vec_init): Use vllezl
35663         instruction if possible.
35664         * config/s390/vector.md (vec_halfnumelts): New mode
35665         attribute.
35666         ("*vec_vllezlf<mode>"): New pattern.
35668 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35670         * config/s390/vector.md ("popcountv16qi2", "popcountv8hi2")
35671         ("popcountv4si2", "popcountv2di2"): Rename to ...
35672         ("popcount<mode>2", "popcountv8hi2_vx", "popcountv4si2_vx")
35673         ("popcountv2di2_vx"): ... these and add !TARGET_VXE to the
35674         condition.
35675         ("popcount<mode>2_vxe"): New pattern.
35677 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35679         * common/config/s390/s390-common.c (processor_flags_table): Add
35680         arch12.
35681         * config.gcc: Add arch12.
35682         * config/s390/driver-native.c (s390_host_detect_local_cpu):
35683         Default to arch12 for unknown CPU model numbers.
35684         * config/s390/s390-builtins.def: Add B_VXE builtin flag.
35685         * config/s390/s390-c.c (s390_cpu_cpp_builtins_internal): Adjust
35686         PROCESSOR_max sanity check.
35687         * config/s390/s390-opts.h (enum processor_type): Add
35688         PROCESSOR_ARCH12.
35689         * config/s390/s390.c (processor_table): Add arch12.
35690         (s390_expand_builtin): Add check for B_VXE flag.
35691         (s390_issue_rate): Add PROCESSOR_ARCH12.
35692         (s390_get_sched_attrmask): Likewise.
35693         (s390_get_unit_mask): Likewise.
35694         (s390_sched_score): Enable z13 scheduling for arch12.
35695         (s390_sched_reorder): Likewise.
35696         (s390_sched_variable_issue): Likewise.
35697         * config/s390/s390.h (enum processor_flags): Add PF_ARCH12 and
35698         PF_VXE.
35699         (s390_tune_attr): Use z13 scheduling also for arch12.
35700         (TARGET_CPU_ARCH12, TARGET_CPU_ARCH12_P, TARGET_CPU_VXE)
35701         (TARGET_CPU_VXE_P, TARGET_ARCH12, TARGET_ARCH12_P, TARGET_VXE)
35702         (TARGET_VXE_P): New macros.
35703         * config/s390/s390.md: Add arch12 to cpu attribute.  Add arch12
35704         and vxe to cpu_facility.  Add arch12 and vxe to enabled attribute.
35705         * config/s390/s390.opt: Add arch12 as processor_type.
35707 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35709         * config/s390/s390.md
35710         ("fixuns_truncdddi2", "fixuns_trunctddi2")
35711         ("fixuns_trunc<BFP:mode><GPR:mode>2"): Merge into ...
35712         ("fixuns_trunc<FP:mode><GPR:mode>2"): New expander.
35714         ("fixuns_trunc<BFP:mode><GPR:mode>2", "fixuns_trunc<mode>si2"):
35715         Rename expanders to ...
35717         ("fixuns_trunc<BFP:mode><GPR:mode>2_emu")
35718         ("fixuns_truncdddi2_emu"): ... these.
35720         ("fixuns_trunc<mode>si2_emu"): New expander.
35722         ("*fixuns_truncdfdi2_z13"): Rename to ...
35723         ("*fixuns_truncdfdi2_vx"): ... this.
35725 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35727         * config/s390/2964.md: Remove the single element vector compare
35728         instructions which are no longer used.
35729         * config/s390/s390.c (s390_select_ccmode): Remove handling of
35730         vector CCmodes.
35731         (s390_canonicalize_comparison): Remove handling of DFmode
35732         compares.
35733         (s390_expand_vec_compare_scalar): Remove function.
35734         (s390_emit_compare): Don't call s390_expand_vec_compare_scalar.
35735         * config/s390/s390.md ("*vec_cmp<insn_cmp>df_cconly"): Remove
35736         pattern.
35737         ("*cmp<mode>_ccs"): Add wfcdb instruction.
35739 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35741         * config/s390/s390.md ("mov<mode>_64dfp" DD_DF): Use vleig for loading a
35742         FP zero.
35743         ("*mov<mode>_64" DD_DF): Remove the vector instructions. These
35744         will anyway by matched by mov<mode>_64dfp.
35746 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35748         * config/s390/s390.md ("mov<mode>" SD_SF): Change vleg/vsteg to
35749         vlef/vstef.  Add missing operand to vleif.
35751 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35753         * config/s390/s390.c (s390_expand_vec_init): Enable vector load
35754         pair for all vector types with 64 bit elements.
35755         * config/s390/vx-builtins.md (V_HW_64): Move mode iterator to ...
35756         * config/s390/vector.md (V_HW_64): ... here.
35757         (V_128_NOSINGLE): New mode iterator.
35758         ("vec_init<V_HW:mode>"): Use V_128 as mode iterator.
35759         ("*vec_splat<mode>"): Use V_128_NOSINGLE mode iterator.
35760         ("*vec_tf_to_v1tf", "*vec_ti_to_v1ti"): New pattern definitions.
35761         ("*vec_load_pairv2di"): Change to ...
35762         ("*vec_load_pair<mode>"): ... this one.
35764 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35766         * config/s390/constraints.md: Add comments.
35767         (jKK): Reject element sizes > 8 bytes.
35768         * config/s390/s390.c (s390_split_ok_p): Enable splitting also for
35769         s_operands.
35770         * config/s390/s390.md: Add the s_operand checks formerly in
35771         s390_split_ok_p to various splitters where they are still
35772         required.
35773         * config/s390/vector.md ("mov<mode>" V_128): Add GPR alternatives
35774         for 128 bit vectors.  Plus two splitters.
35776 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35778         * config/s390/s390.md: Rename the cpu facilty vec to vx throughout
35779         the file.
35781 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35783         PR target/79893
35784         * config/s390/s390-c.c (s390_adjust_builtin_arglist): Issue an
35785         error if the boundary argument is not constant.
35787 2017-03-24  Jakub Jelinek  <jakub@redhat.com>
35789         PR rtl-optimization/80112
35790         * loop-doloop.c (doloop_condition_get): Don't check condition
35791         if cmp isn't SET with IF_THEN_ELSE src.
35793 2017-03-24  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
35795         PR tree-optimization/80158
35796         * gimple-ssa-strength-reduction.c (replace_mult_candidate): When
35797         replacing a candidate statement, also replace it for the
35798         candidate's alternate interpretation.
35799         (replace_rhs_if_not_dup): Likewise.
35800         (replace_one_candidate): Likewise.
35802 2017-03-24  Richard Biener  <rguenther@suse.de>
35804         PR tree-optimization/80167
35805         * graphite-isl-ast-to-gimple.c
35806         (translate_isl_ast_to_gimple::is_valid_rename): Handle default-defs
35807         properly.
35808         (translate_isl_ast_to_gimple::get_rename): Likewise.
35810 2017-03-23  Kelvin Nilsen  <kelvin@gcc.gnu.org>
35812         * config/rs6000/rs6000.c (rs6000_option_override_internal): Change
35813         handling of certain combinations of target options, including the
35814         combinations -mpower8-vector vs. -mno-vsx, -mpower9-vector vs.
35815         -mno-power8-vector, and -mpower9_dform vs. -mno-power9-vector.
35817 2017-03-23  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
35819         PR target/71436
35820         * config/arm/arm.md (*load_multiple): Add reload_completed to
35821         matching condition.
35823 2017-03-23  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
35824             Richard Biener  <rguenth@suse.de>
35826         PR tree-optimization/79908
35827         PR tree-optimization/80136
35828         * tree-stdarg.c (expand_ifn_va_arg_1): For a VA_ARG whose LHS has
35829         been cast away, gimplify_and_add suffices.
35831 2017-03-23  Markus Trippelsdorf  <markus@trippelsdorf.de>
35833         * tree-vrp.c (identify_jump_threads): Delete avail_exprs.
35835 2017-03-23  Richard Biener  <rguenther@suse.de>
35837         PR tree-optimization/80032
35838         * gimplify.c (gimple_push_cleanup): Forced unconditional
35839         cleanups still have to go to the conditional_cleanups
35840         sequence.
35842 2017-03-22  Jakub Jelinek  <jakub@redhat.com>
35844         PR tree-optimization/80072
35845         * tree-ssa-reassoc.c (struct operand_entry): Change id field type
35846         to unsigned int.
35847         (next_operand_entry_id): Change type to unsigned int.
35848         (sort_by_operand_rank): Make sure to return the right return value
35849         even if unsigned fields are bigger than INT_MAX.
35850         (struct oecount): Change cnt and id type to unsigned int.
35851         (oecount_hasher::equal): Formatting fix.
35852         (oecount_cmp): Make sure to return the right return value
35853         even if unsigned fields are bigger than INT_MAX.
35854         (undistribute_ops_list): Change next_oecount_id type to unsigned int.
35856         PR c++/80129
35857         * gimplify.c (gimplify_modify_expr_rhs) <case COND_EXPR>: Clear
35858         TREE_READONLY on result if writing it more than once.
35860         PR sanitizer/80110
35861         * doc/invoke.texi (-fsanitize=thread): Document that with
35862         -fnon-call-exceptions atomics are not able to throw
35863         exceptions.
35865         PR sanitizer/80110
35866         * tsan.c: Include tree-eh.h.
35867         (instrument_builtin_call): Call maybe_clean_eh_stmt or
35868         maybe_clean_or_replace_eh_stmt where needed.
35869         (instrument_memory_accesses): Add cfg_changed argument.
35870         Call gimple_purge_dead_eh_edges on each block and set *cfg_changed
35871         if it returned true.
35872         (tsan_pass): Adjust caller.  Return TODO_cleanup_cfg if cfg_changed.
35874         PR rtl-optimization/63191
35875         * config/i386/i386.c (ix86_delegitimize_address): Turn into small
35876         wrapper function, moved the whole old content into ...
35877         (ix86_delegitimize_address_1): ... this.  New inline function.
35878         (ix86_find_base_term): Use ix86_delegitimize_address_1 with
35879         true as last argument instead of ix86_delegitimize_address.
35881 2017-03-22  Wilco Dijkstra  <wdijkstr@arm.com>
35883         * config/aarch64/aarch64.c (generic_branch_cost): Copy
35884         cortexa57_branch_cost.
35886 2017-03-22  Wilco Dijkstra  <wdijkstr@arm.com>
35888         * config/aarch64/aarch64.c (generic_tunings): Add AES fusion.
35890 2017-03-21  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
35892         PR target/80123
35893         * doc/md.texi (Constraints): Document wA constraint.
35894         * config/rs6000/constraints.md (wA): New.
35895         * config/rs6000/rs6000.c (rs6000_debug_reg_global): Add wA reg_class.
35896         (rs6000_init_hard_regno_mode_ok): Init wA constraint.
35897         * config/rs6000/rs6000.h (RS6000_CONSTRAINT_wA): New.
35898         * config/rs6000/vsx.md (vsx_splat_<mode>): Use wA constraint.
35900 2017-03-22  Cesar Philippidis  <cesar@codesourcery.com>
35902         PR c++/80029
35903         * gimplify.c (is_oacc_declared): New function.
35904         (oacc_default_clause): Use it to set default flags for acc declared
35905         variables inside parallel regions.
35906         (gimplify_scan_omp_clauses): Strip firstprivate pointers for acc
35907         declared variables.
35908         (gimplify_oacc_declare): Gimplify the declare clauses.  Add the
35909         declare attribute to any decl as necessary.
35911 2017-03-22  Thomas Preud'homme  <thomas.preudhomme@arm.com>
35913         PR target/80082
35914         * config/arm/arm-isa.h (isa_bit_lpae): New feature bit.
35915         (ISA_ARMv7ve): Add isa_bit_lpae to the definition.
35916         * config/arm/arm-protos.h (arm_arch7ve): Rename into ...
35917         (arm_arch_lpae): This.
35918         * config/arm/arm.c (arm_arch7ve): Rename into ...
35919         (arm_arch_lpae): This.  Define it in term of isa_bit_lpae.
35920         * config/arm/arm.h (TARGET_HAVE_LPAE): Redefine in term of
35921         arm_arch_lpae.
35923 2017-03-22  Martin Liska  <mliska@suse.cz>
35925         PR target/79906
35926         * config/rs6000/rs6000.c (rs6000_inner_target_options): Show
35927         error message instead of an ICE.
35929 2017-03-21  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
35931         * doc/extend.texi (6.11 Additional Floating Types): Revise.
35933 2017-03-21  Kelvin Nilsen  <kelvin@gcc.gnu.org>
35935         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Add
35936         comments.
35937         * config/rs6000/rs6000.c (rs6000_option_override_internal): Add
35938         comments.
35940 2017-03-21  Martin Sebor  <msebor@redhat.com>
35942         * doc/extend.texi: Use "cannot" instead of "can't."
35943         * doc/hostconfig.texi: Same.
35944         * doc/install.texi: Same.
35945         * doc/invoke.texi: Same.
35946         * doc/loop.texi: Same.
35947         * doc/md.texi: Same.
35948         * doc/objc.texi: Same.
35949         * doc/rtl.texi: Same.
35950         * doc/tm.texi: Same.
35951         * doc/tm.texi.in: Same.
35952         * doc/trouble.texi: Same.
35954 2017-03-21  Alexandre Oliva  <aoliva@redhat.com>
35956         PR debug/63238
35957         * dwarf2out.c (struct checksum_attributes): Add at_alignment.
35958         (collect_checksum_attributes): Set it.
35959         (die_checksum_ordered): Use it.
35961 2017-03-21  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
35963         PR tree-optimization/79908
35964         * tree-stdarg.c (expand_ifn_va_arg_1): Revert the following
35965         change: For a VA_ARG whose LHS has been cast away, use
35966         force_gimple_operand to construct the side effects.
35968 2017-03-21  David Malcolm  <dmalcolm@redhat.com>
35970         PR translation/80001
35971         * omp-offload.c (oacc_loop_fixed_partitions): Make diagnostics
35972         more amenable to translation.
35973         (oacc_loop_auto_partitions): Likewise.
35975 2017-03-21  Marek Polacek  <polacek@redhat.com>
35976             Martin Sebor  <msebor@redhat.com>
35978         PR tree-optimization/80109
35979         * gimple-ssa-warn-alloca.c (alloca_call_type): Only call get_range_info
35980         on INTEGRAL_TYPE_P.
35982 2017-03-21  Jakub Jelinek  <jakub@redhat.com>
35983             Segher Boessenkool  <segher@kernel.crashing.org>
35985         PR target/80125
35986         * combine.c (can_combine_p): Revert the 2017-03-20 change, only
35987         check reg_used_between_p between insn and one of succ or succ2
35988         depending on if succ is artificial insn not inserted into insn
35989         stream.
35991 2017-03-21  Martin Liska  <mliska@suse.cz>
35993         PR gcov-profile/80081
35994         * Makefile.in: Add gcov-dump and fix installation of gcov-tool.
35995         * doc/gcc.texi: Include gcov-dump stuff.
35996         * doc/gcov-dump.texi: New file.
35998 2017-03-21  Toma Tabacu  <toma.tabacu@imgtec.com>
36000         PR rtl-optimization/79150
36001         * config/mips/mips.c (mips_block_move_loop): Emit a NOP after the
36002         conditional jump, if the jump is the last insn of the loop.
36004 2017-03-21  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
36005             Richard Biener  <rguenth@suse.de>
36007         PR tree-optimization/79908
36008         * tree-stdarg.c (expand_ifn_va_arg_1): For a VA_ARG whose LHS has
36009         been cast away, use force_gimple_operand to construct the side
36010         effects.
36012 2017-03-21  Martin Liska  <mliska@suse.cz>
36014         PR libfortran/79956
36015         * simplify-rtx.c (simplify_immed_subreg): Initialize a variable
36016         to NULL.
36018 2017-03-21  Brad Spengler <spender@grsecurity.net>
36020         PR plugins/80094
36021         * plugin.c (htab_hash_plugin): New function.
36022         (add_new_plugin): Use it and adjust.
36023         (parse_plugin_arg_opt): Adjust.
36024         (init_one_plugin): Likewise.
36026 2017-03-21  Richard Biener  <rguenther@suse.de>
36028         PR tree-optimization/80032
36029         * gimplify.c (gimple_push_cleanup): Add force_uncond parameter,
36030         if set force the cleanup to happen unconditionally.
36031         (gimplify_target_expr): Push inserted clobbers with force_uncond
36032         to avoid them being removed by control-dependent DCE.
36034 2017-03-21  Richard Biener  <rguenther@suse.de>
36036         PR tree-optimization/80122
36037         * tree-inline.c (copy_bb): Do not expans va-arg packs or
36038         va_arg_pack_len when the inlined call stmt requires pack
36039         expansion itself.
36040         * tree-inline.h (struct copy_body_data): Make call_stmt a gcall *.
36042 2017-03-21  Jakub Jelinek  <jakub@redhat.com>
36044         PR sanitizer/78158
36045         * tsan.c (instrument_builtin_call): If the memory model argument
36046         is not a constant, assume it is valid.
36048         PR c/67338
36049         * fold-const.c (round_up_loc): Negate divisor in unsigned type to
36050         avoid UB.
36052 2017-03-20  Segher Boessenkool  <segher@kernel.crashing.org>
36054         PR rtl-optimization/79910
36055         * combine.c (can_combine_p): Do not allow combining an I0 or I1
36056         if its dest is used by an insn before I2 (other than the combined
36057         insns themselves, which are properly handled already).
36059 2017-03-20  Segher Boessenkool  <segher@kernel.crashing.org>
36061         Revert:
36062         2017-03-17  Bernd Schmidt  <bschmidt@redhat.com>
36064         * combine.c (record_used_regs): New static function.
36065         (try_combine): Handle situations where there is an additional
36066         instruction between I2 and I3 which needs to have a LOG_LINK
36067         updated.
36069         Revert:
36070         2017-03-17  Jim Wilson  <jim.wilson@linaro.org>
36072         * combine.c (try_combine): Delete redundant i1 test.  Call
36073         prev_nonnote_nondebug_insn instead of prev_nonnote_insn.
36075 2017-03-20  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
36077         PR target/80083
36078         * config/rs6000/rs6000.md (*movsi_internal1): Fix constraints for
36079         alternatives 13/14.
36081 2017-03-20  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
36083         PR tree-optimization/80054
36084         * gimple-ssa-strength-reduction.c (all_phi_incrs_profitable): Fail
36085         the optimization if a PHI or any of its arguments is not dominated
36086         by the candidate's basis.  Use gphi* rather than gimple* as
36087         appropriate.
36088         (replace_profitable_candidates): Clean up a gimple* variable that
36089         should be a gphi* variable.
36091 2017-03-20  Martin Sebor  <msebor@redhat.com>
36093         PR c++/52477
36094         * doc/extend.texi (attribute constructor): Document present limitation.
36096 2017-03-20  Kelvin Nilsen  <kelvin@gcc.gnu.org>
36098         PR target/79963
36099         * config/rs6000/altivec.h (vec_all_ne): Under __cplusplus__ and
36100         __POWER9_VECTOR__ #ifdef control, change template definition to
36101         use Power9-specific built-in function.
36102         (vec_any_eq): Likewise.
36103         * config/rs6000/vector.md (vector_ae_v2di_p): Change the flag used
36104         to control outcomes from this test.
36105         (vector_ae_<mode>p): For VEC_F modes, likewise.
36107 2017-03-20  Ian Lance Taylor  <iant@google.com>
36109         * config/i386/i386.c (ix86_function_regparm): Save an extra
36110         register for -fsplit-stack with DECL_STATIC_CHAIN.
36112 2017-03-17  Palmer Dabbelt  <palmer@dabbelt.com>
36114         PR target/79912
36115         * config/riscv/riscv.c (riscv_preferred_reload_class): Remove.
36116         (TARGET_PREFERRED_RELOAD_CLASS): Likewise.
36118 2017-03-17  Palmer Dabbelt  <palmer@dabbelt.com>
36120         * config/riscv/riscv.c (riscv_print_operand): Use "fence
36121         iorw,ow".
36122         * config/riscv/sync.mc (mem_thread_fence_1): Use "fence
36123         iorw,iorw".
36125 2017-03-20  Marek Polacek  <polacek@redhat.com>
36127         PR sanitizer/80063
36128         * asan.c (DEF_SANITIZER_BUILTIN): Use do { } while (0).
36130 2017-03-20  Richard Biener  <rguenther@suse.de>
36132         PR tree-optimization/80113
36133         * graphite-isl-ast-to-gimple.c (copy_loop_phi_nodes): Do not
36134         allocate extra SSA name for PHI def.
36135         (add_close_phis_to_outer_loops): Likewise.
36136         (add_close_phis_to_merge_points): Likewise.
36137         (copy_loop_close_phi_args): Likewise.
36138         (copy_cond_phi_nodes): Likewise.
36140 2017-03-20  Martin Liska  <mliska@suse.cz>
36142         PR middle-end/79753
36143         * tree-chkp.c (chkp_build_returned_bound): Do not build
36144         returned bounds for a LHS that's not a BOUNDED_P type.
36146 2017-03-20  Martin Liska  <mliska@suse.cz>
36148         PR target/79769
36149         PR target/79770
36150         * tree-chkp.c (chkp_find_bounds_1): Handle REAL_CST,
36151         COMPLEX_CST and VECTOR_CST.
36153 2017-03-20  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
36155         PR target/78857
36156         * config/s390/s390.md ("cmp<mode>_ccs_0"): Add a clobber of the
36157         target operand.  A new splitter adds the clobber statement in case
36158         the target operand is dead anyway.
36160 2017-03-19  Gerald Pfeifer  <gerald@pfeifer.com>
36162         * doc/install.texi (Specific) <sparc-*-linux*>: No longer refer
36163         to age-old versions of binutils and glibc.
36165 2017-03-18  Segher Boessenkool  <segher@kernel.crashing.org>
36167         * doc/contrib.texi (Contributors): Remove duplicate entry for myself.
36169 2017-03-18  Gerald Pfeifer  <gerald@pfeifer.com>
36171         * doc/contrib.texi (Contributors): Add Segher Boessenkool.
36173 2017-03-18  Gerald Pfeifer  <gerald@pfeifer.com>
36175         * doc/install.texi (Specific) <arm-*-eabi>: Remove old
36176         requirement for binutils 2.13.
36178 2017-03-17  Jim Wilson  <jim.wilson@linaro.org>
36180         * combine.c (try_combine): Delete redundant i1 test.  Call
36181         prev_nonnote_nondebug_insn instead of prev_nonnote_insn.
36183 2017-03-17  Palmer Dabbelt  <palmer@dabbelt.com
36185         * doc/install.texi (Specific) <riscv32-*-elf>: Add riscv32-*-elf,
36186         riscv32-*-linux, riscv64-*-elf, riscv64-*-linux to the table of
36187         contents.
36188         <riscv64-*-elf>: Re-arrange section
36189         <riscv32-*-elf>: Add a note about requiring binutils 2.28.
36190         <riscv32-*-linux>: Likewise.
36191         <riscv64-*-elf>: Likewise
36192         <riscv64-*-linux>: Likewise.
36194 2017-03-17  Richard Earnshaw  <rearnsha@arm.com>
36196         PR target/80052
36197         * aarch64.opt(verbose-cost-dump): Fix typo.
36199 2017-03-17  Pat Haugen  <pthaugen@us.ibm.com>
36201         PR target/79951
36202         * config/rs6000/rs6000.md (copysign<mode>3_fcpsgn): Test
36203         for VECTOR_UNIT_VSX_P (<MODE>mode) too.
36205 2017-03-17  Bernd Schmidt  <bschmidt@redhat.com>
36207         * reload.c (find_reloads): When reloading a nonoffsettable address,
36208         use RELOAD_OTHER for it and its address reloads.
36210         PR rtl-optimization/79910
36211         * combine.c (record_used_regs): New static function.
36212         (try_combine): Handle situations where there is an additional
36213         instruction between I2 and I3 which needs to have a LOG_LINK
36214         updated.
36216 2017-03-17  Jeff Law  <law@redhat.com>
36218         PR tree-optimization/71437
36219         * tree-vrp.c (simplify_stmt_for_jump_threading): Lookup the
36220         conditional in the hash table first.
36221         (vrp_dom_walker::before_dom_children): Extract condition from
36222         ASSERT_EXPR.  Record condition, its inverion and any implied
36223         conditions as well.
36225 2017-03-17  Marek Polacek  <polacek@redhat.com>
36226             Markus Trippelsdorf  <markus@trippelsdorf.de>
36228         PR tree-optimization/80079
36229         * gimple-ssa-store-merging.c (class pass_store_merging): Initialize
36230         m_stores_head.
36232 2017-03-17  Richard Biener  <rguenther@suse.de>
36234         PR middle-end/80075
36235         * tree-eh.c (stmt_could_throw_1_p): Only handle gimple assigns.
36236         Properly verify the LHS before the RHS possibly claims to be
36237         handled.
36238         (stmt_could_throw_p): Hande gimple conds fully here.  Clobbers
36239         do not throw.
36241 2017-03-17  Martin Jambor  <mjambor@suse.cz>
36243         * doc/invoke.texi (Option Options): Include -fipa-vrp in the list.
36244         (List of -O2 options): Likewise.
36245         (-fipa-bit-cp): Replace "ipa" with "interprocedural."
36246         (-fipa-vrp) New.
36248 2017-03-17  Tom de Vries  <tom@codesourcery.com>
36250         * gcov-dump.c (print_usage): Print bug_report_url.
36252 2017-03-17  Richard Biener  <rguenther@suse.de>
36254         PR middle-end/80050
36255         * genmatch.c (parser::next): Remove pointless check for CPP_EOF.
36256         (parser::peek): Likewise.
36258 2017-03-17  Richard Biener  <rguenther@suse.de>
36260         PR tree-optimization/80048
36261         * sese.c (free_sese_info): Properly release rename_map and
36262         copied_bb_map elements.
36264 2017-03-16  Alexandre Oliva  <aoliva@redhat.com>
36266         * gimple-ssa-store-merging.c (struct imm_store_chain_info):
36267         Add linked-list forward and backlinks.  Insert on
36268         construction, remove on destruction.
36269         (class pass_store_merging): Add m_stores_head field.
36270         (pass_store_merging::terminate_and_process_all_chains):
36271         Iterate over m_stores_head list.
36272         (pass_store_merging::terminate_all_aliasing_chains):
36273         Likewise.
36274         (pass_store_merging::execute): Check for debug stmts first.
36275         Push new chains onto the m_stores_head stack.
36277 2017-03-16  Michael Meissner  <meissner@linux.vnet.ibm.com>
36279         PR target/71294
36280         * config/rs6000/vsx.md (vsx_splat_<mode>, VSX_D iterator): Allow a
36281         SPLAT operation on ISA 2.07 64-bit systems that have direct move,
36282         but no MTVSRDD support, by doing MTVSRD and XXPERMDI.
36284 2017-03-16  Jeff Law  <law@redhat.com>
36286         PR tree-optimization/71437
36287         * tree-ssa-dom.c (dom_opt_dom_walker): Remove thread_across_edge
36288         member function.  Implementation moved into after_dom_children
36289         member function and into the threader's thread_outgoing_edges
36290         function.
36291         (dom_opt_dom_walker::after_dom_children): Simplify by moving
36292         some code into new thread_outgoing_edges.
36293         * tree-ssa-threadedge.c (thread_across_edge): Make static and simplify
36294         definition.  Simplify marker handling (do it here).   Assume we always
36295         have the available expression and the const/copies tables.
36296         (thread_outgoing_edges): New function extracted from tree-ssa-dom.c
36297         and tree-vrp.c
36298         * tree-ssa-threadedge.h (thread_outgoing_edges): Declare.
36299         * tree-vrp.c (equiv_stack): No longer file scoped.
36300         (vrp_dom_walker): New class.
36301         (vrp_dom_walker::before_dom_children): New member function.
36302         (vrp_dom_walker::after_dom_children): Likewise.
36303         (identify_jump_threads):  Setup domwalker.  Use it rather than
36304         walking edges in a random order by hand.  Simplify setup/finalization.
36305         (finalize_jump_threads): Remove.
36306         (vrp_finalize): Do not call identify_jump_threads here.
36307         (execute_vrp): Do it here instead and call thread_through_all_blocks
36308         here too.
36310         PR tree-optimization/71437
36311         * tree-ssa-dom.c (pfn_simplify): Add basic_block argument.  All
36312         callers changed.
36313         (simplify_stmt_for_jump_threading): Add basic_block argument.  All
36314         callers changed.
36315         (lhs_of_dominating_assert): Moved from here into tree-vrp.c.
36316         (dom_opt_dom_walker::thread_across_edge): Remove
36317         handle_dominating_asserts argument.  All callers changed.
36318         (record_temporary_equivalences_from_stmts_at_dest): Corresponding
36319         changes.  Remove calls to lhs_of_dominating_assert.  Other
36320         uses of handle_dominating_asserts turn into unconditional code
36321         (simplify_control_stmt_condition_1): Likewise.
36322         (simplify_control_stmt_condition): Likewise.
36323         (thread_through_normal_block, thread_across_edge): Likewise.
36324         * tree-ssa-threadedge.h (thread_across_edge): Corresponding changes.
36325         * tree-vrp.c (lhs_of_dominating_assert): Move here.  Return original
36326         object if it is not an SSA_NAME.
36327         (simplify_stmt_for_jump_threading): Call lhs_of_dominating_assert
36328         before calling into the VRP specific simplifiers.
36329         (identify_jump_threads): Remove handle_dominating_asserts
36330         argument.
36332 2017-03-16  Jakub Jelinek  <jakub@redhat.com>
36334         PR fortran/79886
36335         * tree-diagnostic.c (default_tree_printer): No longer static.
36336         * tree-diagnostic.h (default_tree_printer): New prototype.
36338 2017-03-16  Tamar Christina  <tamar.christina@arm.com>
36340         * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>)
36341         Change ins into fmov.
36343 2017-03-16  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
36345         * config/aarch64/iterators.md (h_con): Return "x" for V4HF and V8HF.
36346         * config/aarch64/aarch64-simd.md (*aarch64_fma4_elt_from_dup<mode>):
36347         Use h_con constraint for operand 1.
36348         (*aarch64_fnma4_elt_from_dup<mode>): Likewise.
36349         (*aarch64_mulx_elt_from_dup<mode>): Likewise for operand 2.
36351 2017-03-15  Jeff Law  <law@redhat.com>
36353         PR tree-optimization/71437
36354         * tree-ssa-dom.c (derive_equivalences_from_bit_ior): New function.
36355         (record_temporary_equivalences): Use it.
36357         PR tree-optimization/71437
36358         * tree-ssa-dom.c (struct cond_equivalence): Moved from here into
36359         tree-ssa-scopedtables.
36360         (lookup_avail_expr, build_and_record_new_cond): Likewise.
36361         (record_conditions, record_cond, vuse_eq): Likewise.
36362         (record_edge_info): Adjust to API tweak of record_conditions.
36363         (simplify_stmt_for_jump_threading): Similarly for lookup_avail_expr.
36364         (record_temporary_equivalences, optimize_stmt): Likewise.
36365         (eliminate_redundant_computations): Likewise.
36366         (record_equivalences_from_stmt): Likewise.
36367         * tree-ssa-scopedtables.c: Include options.h and params.h.
36368         (vuse_eq): New function, moved from tree-ssa-dom.c
36369         (build_and_record_new_cond): Likewise.
36370         (record_conditions): Likewise.  Accept vector of conditions rather
36371         than edge_equivalence structure for first argument.
36372         for the first argument.
36373         (avail_exprs_stack::lookup_avail_expr): New member function, moved
36374         from tree-ssa-dom.c.
36375         (avail_exprs_stack::record_cond): Likewise.
36376         * tree-ssa-scopedtables.h (struct cond_equivalence): Moved here
36377         from tree-ssa-dom.c.
36378         (avail_exprs_stack): Add new member functions lookup_avail_expr
36379         and record_cond.
36380         (record_conditions): Declare.
36382 2017-03-15  Vladimir Makarov  <vmakarov@redhat.com>
36384         PR target/80017
36385         * lra-constraints.c (process_alt_operands): Increase reject for
36386         reloading an input/output operand.
36388 2017-03-15  Michael Meissner  <meissner@linux.vnet.ibm.com>
36390         PR target/79038
36391         * config/rs6000/rs6000.md (float<QHI:mode><IEEE128:mode>2): Define
36392         insns to convert from signed/unsigned char/short to IEEE 128-bit
36393         floating point.
36394         (floatuns<QHI:mode><IEEE128:mode>2): Likewise.
36396 2017-03-15  Uros Bizjak  <ubizjak@gmail.com>
36398         PR target/80019
36399         * config/i386/i386.c (ix86_vector_duplicate_value): Create
36400         subreg of inner mode for values already in registers.
36402 2017-03-15  Bernd Schmidt  <bschmidt@redhat.com>
36404         * config/c6x/c6x.c (hwloop_optimize): Handle case where the old
36405         iteration reg is used after the loop.
36407 2017-03-14  Martin Sebor  <msebor@redhat.com>
36409         PR tree-optimization/79800
36410         * gimple-ssa-sprintf.c (format_floating: Add argument.  Handle
36411         precision in negative-positive range.
36412         (format_floating): Call non-const overload with adjusted precision.
36414 2017-03-14  Michael Meissner  <meissner@linux.vnet.ibm.com>
36416         PR target/79947
36417         * config/rs6000/rs6000.h (TARGET_FRSQRTES): Add check for
36418         -mpowerpc-gfxopt.
36420 2017-03-14  Martin Sebor  <msebor@redhat.com>
36422         PR middle-end/80020
36423         * builtin-attrs.def (ATTR_ALLOC_SIZE_2_NOTHROW_LIST): New macro.
36424         * builtins.def (aligned_alloc): Use it.
36426         PR c/79936
36427         * Makefile.in (GTFILES): Add calls.c.
36428         * calls.c: Include "gt-calls.h".
36430 2017-03-14  Bernd Schmidt  <bschmidt@redhat.com>
36432         PR rtl-optimization/79728
36433         * regs.h (struct target_regs): New field
36434         x_contains_allocatable_regs_of_mode.
36435         (contains_allocatable_regs_of_mode): New macro.
36436         * reginfo.c (init_reg_sets_1): Initialize it, and change
36437         contains_reg_of_mode so it includes global regs as well.
36438         * reload.c (push_reload): Use contains_allocatable_regs_of_mode
36439         rather than contains_regs_of_mode.
36441 2017-03-14  Martin Liska  <mliska@suse.cz>
36443         * doc/invoke.texi: Document options that can't be combined with
36444         -fcheck-pointer-bounds.
36446 2017-03-14  Martin Liska  <mliska@suse.cz>
36448         PR middle-end/79831
36449         * doc/invoke.texi (-Wchkp): Document the option.
36451 2017-03-14  Martin Liska  <mliska@suse.cz>
36453         * Makefile.in: Install gcov-dump.
36455 2017-03-14  Martin Liska  <mliska@suse.cz>
36457         * multiple_target.c (expand_target_clones): Bail out for
36458         an invalid attribute.
36460 2017-03-14  Richard Biener  <rguenther@suse.de>
36462         * alias.c (struct alias_set_entry): Pack properly.
36463         * cfgloop.h (struct loop): Likewise.
36464         * cse.c (struct set): Likewise.
36465         * ipa-utils.c (struct searchc_env): Likewise.
36466         * loop-invariant.c (struct invariant): Likewise.
36467         * lra-remat.c (struct cand): Likewise.
36468         * recog.c (struct change_t): Likewise.
36469         * rtl.h (struct address_info): Likewise.
36470         * symbol-summary.h (function_summary): Likewise.
36471         * tree-loop-distribution.c (struct partition): Likewise.
36472         * tree-object-size.c (struct object_size_info): Likewise.
36473         * tree-ssa-loop-ivopts.c (struct cost_pair): Likewise.
36474         * tree-ssa-threadupdate.c (struct ssa_local_info_t): Likewise.
36475         * tree-vect-data-refs.c (struct _vect_peel_info): Likewise.
36476         * tree-vect-slp.c (struct _slp_oprnd_info): Likewise.
36477         * tree-vect-stmts.c (struct simd_call_arg_info): Likewise.
36478         * tree-vectorizer.h (struct _loop_vec_info): Likewise.
36479         (struct _stmt_vec_info): Likewise.
36481 2017-03-14  Martin Liska  <mliska@suse.cz>
36483         PR target/79892
36484         * multiple_target.c (create_dispatcher_calls): Check that
36485         a target can create a function dispatcher.
36487 2017-03-14  Martin Liska  <mliska@suse.cz>
36489         PR lto/66295
36490         * multiple_target.c (expand_target_clones): Drop local.local
36491         flag for default implementation.
36493 2017-03-14  Richard Biener  <rguenther@suse.de>
36495         PR tree-optimization/80030
36496         * tree-vect-stmts.c (vectorizable_store): Plug memleak.
36498 2017-03-13  Kito Cheng  <kito.cheng@gmail.com>
36500         * config/riscv/riscv.c (riscv_emit_float_compare>: Use
36501         gcc_fallthrough() instead of __attribute__((fallthrough));
36503 2017-03-13  Gerald Pfeifer  <gerald@pfeifer.com>
36505         * doc/gcc.texi: Remove "up" link to (DIR).
36506         * doc/gccint.texi: Ditto.
36508 2017-03-13  Gerald Pfeifer  <gerald@pfeifer.com>
36510         * doc/install.texi (Specific) <avr>: Remove reference to
36511         binutils 2.13.
36513 2017-03-13  Jeff Law  <law@redhat.com>
36515         * config/riscv/riscv.c (riscv_emit_float_compare): Use fallthru
36516         attribute rather than comments.
36518         * config/pdp11/pdp11.md (movmemhi): Adjust operand numbers to
36519         match_scratch operand is highest.
36521 2017-03-13  Martin Liska  <mliska@suse.cz>
36523         PR middle-end/78339
36524         * ipa-pure-const.c (warn_function_noreturn): If the declarations
36525         is a CHKP clone, use original declaration.
36527 2017-03-13  Claudiu Zissulescu  <claziss@synopsys.com>
36529         * config/arc/arc.c (arc_init): Use multiplier whenever we have it.
36530         (arc_conditional_register_usage): Use a different allocation order
36531         when optimizing for size.
36532         * common/config/arc/arc-common.c (arc_option_optimization_table):
36533         Section anchors default on when optimizing for size.
36535 2017-03-13  Claudiu Zissulescu  <claziss@synopsys.com>
36537         * config/arc/arc.md (*tst_bitfield_tst): Fix pattern.
36539 2017-03-13  Claudiu Zissulescu  <claziss@synopsys.com>
36541         * config/arc/arc.c (arc_output_addsi): Emit code density adds.
36542         * config/arc/arc.md (cpu_facility): Add cd variant.
36543         (*movqi_insn): Add code density variant.
36544         (*movhi_insn): Likewise.
36545         (*movqi_insn): Likewise.
36546         (*addsi3_mixed): Likewise.
36547         (subsi3_insn): Likewise.
36549 2017-03-13  Claudiu Zissulescu  <claziss@synopsys.com>
36551         * config/arc/arc.md (movsi_cond_exec): Update constraint.
36553 2017-03-13  Claudiu Zissulescu  <claziss@synopsys.com>
36555         * config/arc/arc.c (arc_legitimize_pic_address): Handle PIC
36556         expressions with MINUS and UNARY ops.
36558 2017-03-13  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
36560         PR target/79911
36561         * config/arm/neon.md (vec_sel_widen_ssum_lo<VQI:mode><VW:mode>3):
36562         Rename to...
36563         (vec_sel_widen_ssum_lo<mode><V_half>3): ... This. Avoid mismatch
36564         between vec_select and vector argument.
36565         (vec_sel_widen_ssum_hi<VQI:mode><VW:mode>3): Rename to...
36566         (vec_sel_widen_ssum_hi<mode><V_half>3): ... This. Likewise.
36567         (vec_sel_widen_usum_lo<VQI:mode><VW:mode>3): Rename to...
36568         (vec_sel_widen_usum_lo<mode><V_half>3): ... This.
36569         (vec_sel_widen_usum_hi<VQI:mode><VW:mode>3): Rename to...
36570         (vec_sel_widen_usum_hi<mode><V_half>3): ... This.
36572 2017-03-13  Richard Biener  <rguenther@suse.de>
36574         PR other/79991
36575         * params.def (vect-max-peeling-for-alignment): Fix typo.
36577 2017-03-12  Gerald Pfeifer  <gerald@pfeifer.com>
36579         * doc/install.texi (Specific) <mips-*-*>: Remove description of
36580         issue that only occurred with binutils below 2.18.
36582 2017-03-12  Gerald Pfeifer  <gerald@pfeifer.com>
36584         * doc/install.texi (Specific) <cris-axis-elf>: No longer
36585         refer to binutils 2.11/2.12 minimum.
36587 2017-03-12  Gerald Pfeifer  <gerald@pfeifer.com>
36589         * doc/install.texi (Specific) <powerpc-*-*>: Remove link to
36590         ftp.kernel.org and simplify binutils requirement.
36592 2017-03-11  Gerald Pfeifer  <gerald@pfeifer.com>
36594         * doc/invoke.texi (Warning Options): Fix spelling of link-time
36595         optimization.
36596         (Optimize Options): Ditto.  Also remove redundancy.
36598 2017-03-10  David Malcolm  <dmalcolm@redhat.com>
36600         PR translation/79848
36601         * ipa-devirt.c (warn_types_mismatch): Simplify uses of "%<%s%>" to
36602         "%qs".
36603         * ipa-pure-const.c (suggest_attribute): Likewise.  Convert _
36604         to G_ to avoid double translation.
36606 2017-03-10  David Malcolm  <dmalcolm@redhat.com>
36608         PR translation/79923
36609         * auto-profile.c (get_combined_location): Convert leading
36610         character of diagnostics to lower case and remove trailing period.
36611         (read_profile): Likewise for various diagnostics.
36612         * config/arm/arm.c (arm_option_override): Remove trailing period
36613         from various diagnostics.
36614         * config/msp430/msp430.c (msp430_expand_delay_cycles): Likewise.
36615         (msp430_expand_delay_cycles): Likewise.
36617 2017-03-10  David Malcolm  <dmalcolm@redhat.com>
36619         PR target/79925
36620         * config/aarch64/aarch64.c (aarch64_validate_mcpu): Quote the
36621         full command-line argument, rather than just "str".
36622         (aarch64_validate_march): Likewise.
36623         (aarch64_validate_mtune): Likewise.
36625 2017-03-10  Bernd Schmidt  <bschmidt@redhat.com>
36627         PR rtl-optimization/78911
36628         * lra-assigns.c (must_not_spill_p): New function.
36629         (spill_for): Use it.
36631 2017-03-10  Jakub Jelinek  <jakub@redhat.com>
36633         PR tree-optimization/79981
36634         * tree-vrp.c (extract_range_basic): Handle IMAGPART_EXPR of
36635         ATOMIC_COMPARE_EXCHANGE ifn result.
36636         (stmt_interesting_for_vrp, vrp_visit_stmt): Handle
36637         IFN_ATOMIC_COMPARE_EXCHANGE.
36639 2017-03-10  David Malcolm  <dmalcolm@redhat.com>
36641         PR driver/79875
36642         * opts.c (parse_sanitizer_options): Add missing question mark to
36643         "did you mean" message.
36645 2017-03-10  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
36647         * config/rs6000/rs6000-builtin.def (VMULEUB_UNS): Remove orphaned
36648         built-in.
36649         (VMULEUH_UNS): Likewise.
36650         (VMULOUB_UNS): Likewise.
36651         (VMULOUH_UNS): Likewise.
36652         * config/rs6000/rs6000.c (builtin_function_type): Remove
36653         references to ALTIVEC_BUILTIN_VMUL[EO]U[BH]_UNS.
36655 2017-03-10  David Malcolm  <dmalcolm@redhat.com>
36657         PR bootstrap/79952
36658         * read-rtl-function.c (function_reader::read_rtx_operand): Update
36659         x with result of extra_parsing_for_operand_code_0.
36660         (function_reader::extra_parsing_for_operand_code_0): Convert
36661         return type from void to rtx, returning x.  When reading
36662         SYMBOL_REF with SYMBOL_FLAG_HAS_BLOCK_INFO, reallocate x to the
36663         larger size containing struct block_symbol.
36665 2017-03-10  Segher Boessenkool  <segher@kernel.crashing.org>
36667         * config/rs6000/rs6000.c (rs6000_option_override_internal): Disallow
36668         -mfloat128-hardware without -m64.
36670 2017-03-10  Will Schmidt <will_schmidt@vnet.ibm.com>
36672         PR target/79941
36673         * config/rs6000/rs6000.c (builtin_function_type): Add VMUL*U[HB]
36674         entries to the case statement that marks unsigned arguments to
36675         overloaded functions.
36677 2017-03-10  Kelvin Nilsen  <kelvin@gcc.gnu.org>
36679         * config/rs6000/rs6000.c (rs6000_option_override_internal): Fix
36680         two typographic errors in the handling of TARGET_UPPER_REGS_DI.
36682 2017-03-10  Pat Haugen  <pthaugen@us.ibm.com>
36684         PR target/79907
36685         * config/rs6000/rs6000.c (rs6000_init_hard_regno_mode_ok): Test
36686         TARGET_UPPER_REGS_DI when setting 'wi' constraint regclass.
36688 2017-03-10  Martin Liska  <mliska@suse.cz>
36690         PR target/65705
36691         PR target/69804
36692         * toplev.c (process_options): Enable MPX with LSAN and UBSAN.
36693         * tree-chkp.c (chkp_walk_pointer_assignments): Verify that
36694         FIELD != NULL.
36696 2017-03-10  Olivier Hainque  <hainque@adacore.com>
36698         * tree-switch-conversion (array_value_type): Start by resetting
36699         candidate type to it's main variant.
36701 2017-03-10  Jakub Jelinek  <jakub@redhat.com>
36703         PR rtl-optimization/79909
36704         * combine.c (try_combine): Use simplify_replace_rtx on individual
36705         CALL_INSN_FUNCTION_USAGE elements instead of replace_rtx on copy_rtx
36706         of the whole CALL_INSN_FUNCTION_USAGE.
36708         PR tree-optimization/79972
36709         * gimple-ssa-warn-alloca.c (alloca_call_type): Only call
36710         get_range_info on SSA_NAMEs.  Formatting fixes.
36712 2017-03-10  Richard Biener  <rguenther@suse.de>
36713             Jakub Jelinek  <jakub@redhat.com>
36715         PR tree-optimization/77975
36716         * tree-ssa-loop-niter.c (get_base_for): Allow phi argument from latch
36717         edge to be constant.
36718         (get_val_for): For constant x return it.  Formatting fix.
36719         (loop_niter_by_eval): Avoid pointless looping if the next iteration
36720         would use the same bases as the current one.
36722 2017-03-09  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
36724         * config/rs6000/rs6000.c (rs6000_gen_le_vsx_permute): Use rotate
36725         instead of vec_select for V1TImode.
36726         * conifg/rs6000/vsx.md (VSX_LE): Remove mode iterator that is no
36727         longer needed.
36728         (VSX_LE_128): Add V1TI to this mode iterator.
36729         (*vsx_le_perm_load_<mode>): Change to use VSX_D mode iterator.
36730         (*vsx_le_perm_store_<mode>): Likewise.
36731         (pre-reload splitter for VSX stores): Likewise.
36732         (post-reload splitter for VSX stores): Likewise.
36733         (*vsx_xxpermdi2_le_<mode>): Likewise.
36734         (*vsx_lxvd2x2_le_<mode>): Likewise.
36735         (*vsx_stxvd2x2_le_<mode>): Likewise.
36737 2017-03-09  Michael Eager  <eager@eagercon.com>
36739         Correct failures with --enable-checking=yes,rtl.
36741         * config/microblaze/microblaze.c (microblaze_expand_shift):
36742         Replace GET_CODE test with CONST_INT_P and INTVAL test with
36743         test for const0_rtx.
36744         * config/microblaze/microblaze.md (ashlsi3_byone, ashrsi3_byone,
36745         lshrsi3_byone): Replace INTVAL with test for const1_rtx.
36747 2017-03-09  Richard Biener  <rguenther@suse.de>
36749         PR tree-optimization/79977
36750         * graphite-scop-detection.c (scop_detection::merge_sese):
36751         Handle the case of extra exits to blocks dominating the entry.
36753 2017-03-09  Toma Tabacu  <toma.tabacu@imgtec.com>
36755         * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
36756         Document rdynamic.
36758 2017-03-09  Vladimir Makarov  <vmakarov@redhat.com>
36760         PR rtl-optimization/79949
36761         * lra-constraints.c (process_alt_operands): Check memory when
36762         trying to predict a cycle.  Print about the overall increase.
36764 2017-03-09  Richard Biener  <rguenther@suse.de>
36766         PR middle-end/79971
36767         * gimple-expr.c (useless_type_conversion_p): Preserve
36768         TYPE_SATURATING for fixed-point types.
36770 2017-03-09  Richard Biener  <rguenther@suse.de>
36772         PR ipa/79970
36773         * ipa-prop.c (ipa_modify_formal_parameters): Avoid changing
36774         alignment of BLKmode params.
36776 2017-03-09  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
36778         PR target/79913
36779         * config/aarch64/iterators.md (VALL_F16_NO_V2Q): New mode iterator.
36780         (VALL_NO_V2Q): Likewise.
36781         (VDQF_DF): Delete.
36782         * config/aarch64/aarch64-simd.md
36783         (aarch64_dup_lane_<vswap_width_name><mode>): Use VALL_F16_NO_V2Q
36784         iterator.
36785         (*aarch64_simd_vec_copy_lane_<vswap_width_name><mode>): Use
36786         VALL_NO_V2Q mode iterator.
36787         (*aarch64_vgetfmulx<mode>): Use VDQF iterator.
36789 2017-03-09  Martin Liska  <mliska@suse.cz>
36791         PR tree-optimization/79631
36792         * tree-chkp-opt.c (chkp_is_constant_addr): Call
36793         tree_int_cst_sign_bit just for INTEGER constants.
36795 2017-03-09  Martin Liska  <mliska@suse.cz>
36797         PR target/65705
36798         PR target/69804
36799         * toplev.c (process_options): Disable -fcheck-pointer-bounds with
36800         sanitizers.
36802 2017-03-09  Marek Polacek  <polacek@redhat.com>
36804         PR c++/79672
36805         * tree.c (inchash::add_expr): Handle TREE_VEC.
36807 2017-03-09  Martin Liska  <mliska@suse.cz>
36809         PR ipa/79764
36810         (chkp_narrow_size_and_offset): New function.
36811         (chkp_parse_array_and_component_ref): Support BIT_FIELD_REF.
36812         (void chkp_parse_bit_field_ref): New function.
36813         (chkp_make_addressed_object_bounds): Add case for BIT_FIELD_REF.
36814         (chkp_process_stmt): Use chkp_parse_bit_field_ref.
36816 2017-03-09  Martin Liska  <mliska@suse.cz>
36818         PR ipa/79761
36819         * tree-chkp.c (chkp_get_bound_for_parm): Get bounds for a param.
36820         (chkp_find_bounds_1): Remove gcc_unreachable.
36822 2017-03-09  Jakub Jelinek  <jakub@redhat.com>
36824         PR sanitizer/79944
36825         * asan.c (get_mem_refs_of_builtin_call): For BUILT_IN_ATOMIC* and
36826         BUILT_IN_SYNC*, determine the access type from the size suffix and
36827         always build a MEM_REF with that type.  Handle forgotten
36828         BUILT_IN_SYNC_FETCH_AND_NAND_16 and BUILT_IN_SYNC_NAND_AND_FETCH_16.
36830         PR target/79932
36831         * config/i386/avx512vlintrin.h (_mm256_cmpge_epi32_mask,
36832         _mm256_cmpge_epi64_mask, _mm256_cmpge_epu32_mask,
36833         _mm256_cmpge_epu64_mask, _mm256_cmple_epi32_mask,
36834         _mm256_cmple_epi64_mask, _mm256_cmple_epu32_mask,
36835         _mm256_cmple_epu64_mask, _mm256_cmplt_epi32_mask,
36836         _mm256_cmplt_epi64_mask, _mm256_cmplt_epu32_mask,
36837         _mm256_cmplt_epu64_mask, _mm256_cmpneq_epi32_mask,
36838         _mm256_cmpneq_epi64_mask, _mm256_cmpneq_epu32_mask,
36839         _mm256_cmpneq_epu64_mask, _mm256_mask_cmpge_epi32_mask,
36840         _mm256_mask_cmpge_epi64_mask, _mm256_mask_cmpge_epu32_mask,
36841         _mm256_mask_cmpge_epu64_mask, _mm256_mask_cmple_epi32_mask,
36842         _mm256_mask_cmple_epi64_mask, _mm256_mask_cmple_epu32_mask,
36843         _mm256_mask_cmple_epu64_mask, _mm256_mask_cmplt_epi32_mask,
36844         _mm256_mask_cmplt_epi64_mask, _mm256_mask_cmplt_epu32_mask,
36845         _mm256_mask_cmplt_epu64_mask, _mm256_mask_cmpneq_epi32_mask,
36846         _mm256_mask_cmpneq_epi64_mask, _mm256_mask_cmpneq_epu32_mask,
36847         _mm256_mask_cmpneq_epu64_mask, _mm_cmpge_epi32_mask,
36848         _mm_cmpge_epi64_mask, _mm_cmpge_epu32_mask, _mm_cmpge_epu64_mask,
36849         _mm_cmple_epi32_mask, _mm_cmple_epi64_mask, _mm_cmple_epu32_mask,
36850         _mm_cmple_epu64_mask, _mm_cmplt_epi32_mask, _mm_cmplt_epi64_mask,
36851         _mm_cmplt_epu32_mask, _mm_cmplt_epu64_mask, _mm_cmpneq_epi32_mask,
36852         _mm_cmpneq_epi64_mask, _mm_cmpneq_epu32_mask, _mm_cmpneq_epu64_mask,
36853         _mm_mask_cmpge_epi32_mask, _mm_mask_cmpge_epi64_mask,
36854         _mm_mask_cmpge_epu32_mask, _mm_mask_cmpge_epu64_mask,
36855         _mm_mask_cmple_epi32_mask, _mm_mask_cmple_epi64_mask,
36856         _mm_mask_cmple_epu32_mask, _mm_mask_cmple_epu64_mask,
36857         _mm_mask_cmplt_epi32_mask, _mm_mask_cmplt_epi64_mask,
36858         _mm_mask_cmplt_epu32_mask, _mm_mask_cmplt_epu64_mask,
36859         _mm_mask_cmpneq_epi32_mask, _mm_mask_cmpneq_epi64_mask,
36860         _mm_mask_cmpneq_epu32_mask, _mm_mask_cmpneq_epu64_mask): Move
36861         definitions outside of __OPTIMIZE__ guarded section.
36863         PR target/79932
36864         * config/i386/avx512bwintrin.h (_mm512_packs_epi32,
36865         _mm512_maskz_packs_epi32, _mm512_mask_packs_epi32,
36866         _mm512_packus_epi32, _mm512_maskz_packus_epi32,
36867         _mm512_mask_packus_epi32): Move definitions outside of __OPTIMIZE__
36868         guarded section.
36870 2017-03-09  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
36872         * config/s390/vx-builtins.md ("vfee<mode>", "vfeez<mode>")
36873         ("vfenez<mode>"): Add missing constraints.
36875 2017-03-08  Martin Sebor  <msebor@redhat.com>
36877         PR target/79928
36878         * config/nds32/nds32.c (nds32_option_override):
36879         Fix misspelled diagnostic.
36881 2017-03-08  Jakub Jelinek  <jakub@redhat.com>
36883         PR c/79940
36884         * gimplify.c (gimplify_omp_for): Replace index var in outer
36885         taskloop statement with an artificial variable and add
36886         OMP_CLAUSE_PRIVATE clause for it.
36888 2017-03-08  Richard Biener  <rguenther@suse.de>
36890         PR tree-optimization/79955
36891         * tree-ssa-uninit.c (warn_uninitialized_vars): Do not warn
36892         for accesses that are completely outside of the variable.
36894 2017-03-08  Andrew Haley  <aph@redhat.com>
36896         PR tree-optimization/79943
36897         * tree-ssa-loop-split.c (compute_new_first_bound): When
36898         calculating the new upper bound, (END-BEG) should be added, not
36899         subtracted.
36901 2017-03-08  Jakub Jelinek  <jakub@redhat.com>
36903         * config/avr/avr.md (setmemhi): Make sure match_dup
36904         operand number comes before match_scratch.
36906 2017-03-08  Richard Biener  <rguenther@suse.de>
36908         PR tree-optimization/79920
36909         * tree-vect-slp.c (vect_create_mask_and_perm): Remove and inline
36910         with ncopies == 1 to ...
36911         (vect_transform_slp_perm_load): ... here.  Properly compute
36912         all element loads by iterating VF times over the group.  Do
36913         not handle ncopies (computed in a broken way) in
36914         vect_create_mask_and_perm.
36916 2017-03-08  Jakub Jelinek  <jakub@redhat.com>
36918         PR sanitizer/79904
36919         * internal-fn.c (expand_vector_ubsan_overflow): If arg0 or arg1
36920         is a uniform vector, use uniform_vector_p return value instead of
36921         building ARRAY_REF on folded VIEW_CONVERT_EXPR to array type.
36923 2017-03-07  Marek Polacek  <polacek@redhat.com>
36925         PR middle-end/79809
36926         * gimple-ssa-warn-alloca.c (pass_walloca::gate): Use HOST_WIDE_INT.
36927         (alloca_call_type): Likewise.
36929 2017-03-07  Martin Liska  <mliska@suse.cz>
36931         * gcov.c (process_args): Put comment to correct location.
36933 2017-03-07  Martin Liska  <mliska@suse.cz>
36935         PR middle-end/68270
36936         * tree-chkp.c (chkp_may_narrow_to_field): Add new argument ref.
36937         Use array_at_struct_end_p instead of DECL_CHAIN (field).
36938         (chkp_narrow_bounds_for_field): Likewise.
36939         (chkp_parse_array_and_component_ref): Pass one more argument to
36940         call.
36942 2017-03-07  Richard Biener  <rguenther@suse.de>
36944         * tree-vect-loop-manip.c (slpeel_add_loop_guard): Preserve
36945         preheaders.
36947 2017-03-07  Segher Boessenkool  <segher@kernel.crashing.org>
36949         * config/i386/i386.c (ix86_local_alignment): Align most aggregates
36950         of 16 bytes and more to 16 bytes, not those of 16 bits and more.
36952 2017-03-07  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
36954         PR c/79855
36955         * params.def (PARAM_STORE_MERGING_ALLOW_UNALIGNED): Add full stop
36956         to end of description.
36957         (PARAM_MAX_STORES_TO_MERGE): Likewise.
36959 2017-03-07  Jakub Jelinek  <jakub@redhat.com>
36961         PR rtl-optimization/79901
36962         * config/i386/sse.md (*avx512bw_<code><mode>3<mask_name>): Renamed to
36963         ...
36964         (*avx512f_<code><mode>3<mask_name>): ... this.
36965         (<code><mode>3 with maxmin code iterator): Use VI8_AVX2_AVX512F
36966         iterator instead of VI8_AVX2_AVX512BW.
36968         PR rtl-optimization/79901
36969         * expr.c (expand_expr_real_2): For vector MIN/MAX, if there is no
36970         min/max expander, expand it using expand_vec_cond_expr.
36972         PR sanitizer/79897
36973         * ubsan.c (ubsan_encode_value): Call mark_addressable on the
36974         temporary.
36976 2017-03-06  Jakub Jelinek  <jakub@redhat.com>
36978         PR c++/79821
36979         * dwarf2out.h (dw_vec_const): Change array type from unsigned char *
36980         to void * for PCH reasons.
36981         * dwarf2out.c (output_loc_operands, output_die): Cast
36982         v.val_vec.array to unsigned char *.
36984 2017-03-06  John David Anglin  <danglin@gcc.gnu.org>
36986         PR target/77850
36987         * config/pa/pa-64.h (PAD_VARARGS_DOWN): Don't pad down complex and
36988         vector types.
36990 2017-03-06  Vladimir Makarov  <vmakarov@redhat.com>
36992         PR rtl-optimization/79571
36993         * lra-constraints.c (process_alt_operands): Calculate static
36994         reject and subtract it from overall when only addresses will be
36995         reloaded.
36997 2017-03-06  Julia Koval  <julia.koval@intel.com>
36999         PR target/79793
37000         * config/i386/i386.c (ix86_minimum_incoming_stack_boundary): Set
37001         incoming stack boundary to 128 for 64-bit targets.
37003 2017-03-06  Richard Biener  <rguenther@suse.de>
37005         PR tree-optimization/79894
37006         * tree-vectorizer.c (vectorize_loops): Set loop_vectorized_call
37007         to NULL after folding it.
37009 2017-03-06  Richard Biener  <rguenther@suse.de>
37011         PR tree-optimization/79824
37012         * tree-vect-stmts.c (get_group_load_store_type): Fix alignment
37013         check disabling peeling for gaps.
37015 2017-03-06  Toma Tabacu  <toma.tabacu@imgtec.com>
37017         * doc/sourcebuild.texi (Effective-Target Keywords, Environment
37018         attributes): Document gettimeofday.
37020 2017-03-06  Robin Dapp  <rdapp@linux.vnet.ibm.com>
37022         * config/s390/s390.c (s390_option_override_internal): Set
37023         PARAM_MIN_VECT_LOOP_BOUND
37025 2017-03-06  Robin Dapp  <rdapp@linux.vnet.ibm.com>
37027         * config/s390/s390.c (s390_asm_output_function_label): Use nopr %r0.
37028         * config/s390/s390.md: Likewise.
37030 2017-03-06  Jakub Jelinek  <jakub@redhat.com>
37032         PR target/79812
37033         * config/i386/sse.md (VI8F_256_512): Remove mode iterator.
37034         (<avx2_avx512>_perm<mode>): Rename to ...
37035         (avx2_perm<mode>): ... this.  Use VI8F_256 iterator instead
37036         of VI8F_256_512.
37037         (<avx512>_perm<mode>_mask): Rename to ...
37038         (avx512vl_perm<mode>_mask): ... this.  Use VI8F_256 iterator instead
37039         of VI8F_256_512.
37040         (<avx2_avx512>_perm<mode>_1<mask_name>): Rename to ...
37041         (avx2_perm<mode>_1<mask_name): ... this.  Use VI8F_256 iterator
37042         instead of VI8F_256_512.
37043         (avx512f_perm<mode>): New define_expand.
37044         (avx512f_perm<mode>_mask): Likewise.
37045         (avx512f_perm<mode>_1<mask_name>): New define_insn.
37046         (<avx512>_vec_dup<mode>_1): Fix up vec_select mode.
37048 2017-03-06  Prachi Godbole  <prachi.godbole@imgtec.com>
37050         * config/mips/mips-msa.md (msa_fmax_a_<msafmt>, msa_fmin_a_<msafmt>,
37051         msa_max_a_<msafmt>, msa_min_a_<msafmt>): Introduce mode interator for
37052         if_then_else.
37053         (smin<mode>3, smax<mode>3): Change operand print code from 'B' to 'E'.
37055 2017-03-06  Martin Liska  <mliska@suse.cz>
37057         PR sanitize/79783
37058         * asan.c (asan_expand_poison_ifn): Do not expand ASAN_POISON
37059         when having a SSA NAME w/o VAR_DECL assigned to it.
37061 2017-03-06  Prachi Godbole  <prachi.godbole@imgtec.com>
37063         * config/mips/mips-msa.md (msa_dotp_<su>_d, msa_dpadd_<su>_d,
37064         msa_dpsub_<su>_d): Fix MODE for vec_select.
37066 2017-03-06  Prachi Godbole  <prachi.godbole@imgtec.com>
37068         * config/mips/mips.c (mips_gen_const_int_vector): Change type of last
37069         argument.
37070         * config/mips/mips-protos.h (mips_gen_const_int_vector): Likewise.
37072 2017-03-06  Richard Biener  <rguenther@suse.de>
37074         * lto-streamer.c (lto_check_version): Use %qs in diagnostics.
37075         * plugin.c (register_plugin_info): Likewise.
37076         * tree-chkp.c (chkp_make_static_const_bounds): Likewise.
37078 2017-03-05  Jakub Jelinek  <jakub@redhat.com>
37080         * config/i386/sse.md (sse_storehps, sse_storelps,
37081         avx_<castmode><avxsizesuffix>_<castmode>,
37082         avx512f_<castmode><avxsizesuffix>_<castmode>,
37083         avx512f_<castmode><avxsizesuffix>_256<castmode>): Require
37084         in condition that at least one operand is not a MEM.
37086 2017-03-03  Jakub Jelinek  <jakub@redhat.com>
37088         PR middle-end/79805
37089         * internal-fn.def (ATOMIC_BIT_TEST_AND_SET, ATOMIC_BIT_TEST_AND_RESET,
37090         ATOMIC_BIT_TEST_AND_COMPLEMENT, ATOMIC_COMPARE_EXCHANGE): Remove
37091         ECF_NOTHROW.
37092         * gimple-fold.c (fold_builtin_atomic_compare_exchange): Set
37093         gimple_call_nothrow_p flag based on whether original builtin can throw.
37094         If it can, emit following stmts on the fallthrough edge.
37095         * tree-ssa-ccp.c (optimize_atomic_bit_test_and): Similarly, except
37096         don't create new bb if inserting just debug stmts on the edge, try to
37097         insert them on the fallthru bb or just reset debug stmts.
37099 2017-03-03  Segher Boesssenkool  <segher@kernel.crashing.org>
37101         PR target/43763
37102         * config/rs6000/rs6000.c (rs6000_final_prescan_insn): Save and
37103         restore recog_data (including the operand rtxes inside it) around
37104         the call to get_insn_template.
37106 2017-03-03  Martin Sebor  <msebor@redhat.com>
37108         PR tree-optimization/79699
37109         * context.c (context::~context): Free MPFR caches to avoid
37110         a memory leak on program exit.
37112 2017-03-03  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
37114         * config/aarch64/aarch64.c (aarch64_float_const_representable_p):
37115         Use wide_int::ulow () instead of .elt (0).
37117 2017-03-03  Uros Bizjak  <ubizjak@gmail.com>
37119         * config/i386/i386.md (*pushtf): Change *roF constraint to *roC.
37120         (*pushxf): Limit oF constraint to 32bit targets and add oC
37121         constraint for 64bit targets.
37122         (pushxf splitter): Use PUSH_ROUNDING to calculate stack adjustment.
37123         (*pushdf): Change rmF constraint to rmC.
37125 2017-03-03  Martin Liska  <mliska@suse.cz>
37127         * tree-ssa-loop-prefetch.c (pass_loop_prefetch::execute):
37128         Remove unused variable.
37130 2017-03-03  Jakub Jelinek  <jakub@redhat.com>
37132         PR target/79807
37133         * config/i386/i386.c (ix86_expand_multi_arg_builtin): If target
37134         is a memory operand, increase num_memory.
37135         (ix86_expand_args_builtin): Likewise.
37137 2017-03-03  Jan Hubicka  <jh@suse.cz>
37139         PR lto/79760
37140         * ipa-devirt.c (maybe_record_node): Properly handle
37141         __cxa_pure_virtual visibility.
37143 2017-03-03  Martin Liska  <mliska@suse.cz>
37145         PR tree-optimization/79803
37146         * tree-ssa-loop-prefetch.c (tree_ssa_prefetch_arrays): Remove
37147         assert.
37148         (pass_loop_prefetch::execute): Disabled optimization if an
37149         assumption about L1 cache size is not met.
37151 2017-03-03  Martin Liska  <mliska@suse.cz>
37153         PR rtl-optimization/79574
37154         * gcse.c (struct gcse_expr): Use HOST_WIDE_INT instead of int.
37155         (hash_scan_set): Likewise.
37156         (dump_hash_table): Likewise.
37157         (hoist_code): Likewise.
37159 2017-03-03  Richard Biener  <rguenther@suse.de>
37161         * fixed-value.c (fixed_from_string): Restore use of elt (1)
37162         in place of uhigh ().
37163         (fixed_convert_from_real): Likewise.
37165 2017-03-03  Uros Bizjak  <ubizjak@gmail.com>
37167         PR target/79514
37168         * config/i386/i386.md (*pushxf_rounded): Use Pmode instead of DImode.
37170 2017-03-03  Richard Biener  <rguenther@suse.de>
37172         PR middle-end/79818
37173         * match.pd ( X +- C1 CMP C2 -> X CMP C2 -+ C1): Add missing
37174         TYPE_OVERFLOW_UNDEFINED check.
37176 2017-03-02  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
37178         * config/rs6000/vector.md (vector_ne_<mode>_p): Correct operand
37179         numbers.
37180         (vector_ae_<mode>_p): Likewise.
37181         (vector_nez_<mode>_p): Likewise.
37182         (vector_ne_v2di_p): Likewise.
37183         (vector_ae_v2di_p): Likewise.
37184         (vector_ne_<mode>_p): Likewise.
37185         * config/rs6000/vsx.md (vsx_tsqrt<mode>2_fg): Correct operand
37186         numbers.
37187         (vsx_tsqrt<mode>2_fe): Likewise.
37189 2017-03-02  Uros Bizjak  <ubizjak@gmail.com>
37191         PR target/79514
37192         * config/i386/i386.md (*pushxf_rounded): New insn_and_split pattern.
37194 2017-03-02  Jakub Jelinek  <jakub@redhat.com>
37196         PR rtl-optimization/79780
37197         * cprop.c (one_cprop_pass): When second and further conditional trap
37198         in a single basic block is turned into an unconditional trap, turn it
37199         into a deleted note to avoid RTL verification failures.
37201 2017-03-02  Richard Biener  <rguenther@suse.de>
37203         * fold-const.c (const_binop): Use ulow () instead of elt (0).
37205 2017-03-02  Richard Biener  <rguenther@suse.de>
37207         PR tree-optimization/79345
37208         PR c++/42000
37209         * tree-ssa-alias.c (walk_aliased_vdefs_1): Take a limit
37210         param and abort the walk, returning -1 if it is hit.
37211         (walk_aliased_vdefs): Take a limit param and pass it on.
37212         * tree-ssa-alias.h (walk_aliased_vdefs): Add a limit param,
37213         defaulting to 0 and return a signed int.
37214         * tree-ssa-uninit.c (struct check_defs_data): New struct.
37215         (check_defs): New helper.
37216         (warn_uninitialized_vars): Use walk_aliased_vdefs to warn
37217         about uninitialized memory.
37218         * fixed-value.c (fixed_from_string): Use ulow/uhigh to avoid
37219         bogus uninitialized warning.
37220         (fixed_convert_from_real): Likewise.
37222 2017-03-02  Bin Cheng  <bin.cheng@arm.com>
37224         PR tree-optimization/66768
37225         * tree-ssa-loop-ivopts.c (find_interesting_uses_address): Skip addr
37226         iv_use if base object can't be determined.
37228 2017-03-02  Jakub Jelinek  <jakub@redhat.com>
37230         PR tree-optimization/79345
37231         * gensupport.h (struct pattern_stats): Add min_scratch_opno field.
37232         * gensupport.c (get_pattern_stats_1) <case MATCH_SCRATCH>: Update it.
37233         (get_pattern_stats): Initialize it.
37234         * genemit.c (gen_expand): Verify match_scratch numbers come after
37235         match_operand/match_dup numbers.
37236         * config/i386/i386.md (<s>mul<mode>3_highpart): Swap match_dup and
37237         match_scratch numbers.
37238         * config/i386/sse.md (avx2_gathersi<mode>, avx2_gatherdi<mode>):
37239         Likewise.
37240         * config/s390/s390.md (trunctdsd2): Likewise.
37242 2017-03-02  Richard Biener  <rguenther@suse.de>
37244         * wide-int.h (wide_int_storage::operator=): Implement in terms
37245         of wi::copy.
37247 2017-03-02  Richard Biener  <rguenther@suse.de>
37249         PR tree-optimization/79777
37250         * tree-ssa-pre.c (eliminate_insert): Give up if we simplify
37251         the to insert expression to sth existing.
37253 2017-03-01  Martin Sebor  <msebor@redhat.com>
37255         PR middle-end/79692
37256         * gimple-ssa-sprintf.c
37257         (directive::known_width_and_precision): New function.
37258         (format_integer): Use it.
37259         (get_mpfr_format_length): Consider the full range of precision
37260         when computing %g output with the # flag.  Set the likely byte
37261         count to 3 rather than 1 when precision is indeterminate.
37262         (format_floating): Correct the lower bound of precision.
37264 2017-03-01  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
37266         * doc/invoke.texi: Document default code model for 64-bit Linux.
37268 2017-03-01  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
37270         PR target/79752
37271         * config/rs6000/rs6000.md (peephole2 for udiv/umod): Should emit
37272         udiv rather than div since input pattern is unsigned.
37274 2017-03-01  Uros Bizjak  <ubizjak@gmail.com>
37276         * config/i386/i386.c (print_reg): Warn for values of
37277         unsupported size in integer register.
37279 2017-03-01  Michael Meissner  <meissner@linux.vnet.ibm.com>
37281         PR target/79439
37282         * config/rs6000/predicates.md (current_file_function_operand): Do
37283         not allow self calls to be local if the function is replaceable.
37285 2017-03-01  Kelvin Nilsen  <kelvin@gcc.gnu.org>
37287         PR target/79395
37288         * config/rs6000/altivec.h (vec_ctz and others): Change the
37289         preprocessor macro that controls conditional compilation from
37290         _ARCH_PWR9 to __POWER9_VECTOR__.
37291         (vec_all_ne): Change parameterization of __altivec_scalar_pred
37292         macro expansion under preprocessor #ifdef __POWER9_VECTOR__
37293         control (instead of _ARCH_PWR9 control) so that template
37294         definition uses power9-specific function.
37295         (vec_any_eq): Likewise.
37296         (vec_all_ne): Change macro definition to use a power9-specific
37297         expansion under #ifdef __POWER9_VECTOR__ control (instead of
37298         _ARCH_PWR9 control).
37299         (vec_any_eq) Likewise.
37300         * config/rs6000/rs6000-builtin.def (CMPNEF): Remove BU_P9V_AV_2
37301         expansion for CMPNEF to remove support for xvcmpnesp instruction.
37302         (CMPNED): Remove BU_P9V_AV2 expansion for CMPNED to remove
37303         support for xvcmpnedp instruction.
37304         (VCMPNEB_P): Replace BU_P9V_AV_P macro expansion with BU_P9V_AV_2
37305         macro expansion so that Power9 implementation of vec_all_ne does
37306         not use the AltiVec predicate framework.
37307         (VCMPNEH_P): Likewise.
37308         (VCMPNEW_P): Likewise.
37309         (VCMPNED_P): Likewise.
37310         (VCMPNEFP_P): Likewise.
37311         (VCMPNEDP_P): Likewise.
37312         (VCMPAEB_P): Add BU_P9V_AV_2 macro expansion to change
37313         implementation of vec_any_eq to not use AltiVec predicate
37314         framework.
37315         (VCMPAEH_P): Likewise.
37316         (VCMPAEW_P): Likewise.
37317         (VCMPAED_P): Likewise.
37318         (VCMPAEFP_P): Likewise.
37319         (VCMPAEDP_P): Likewise.
37320         (VCMPNE_P): Replace BU_P9V_OVERLOAD_P macro expansion with
37321         BU_P9V_OVERLOAD_2 so that Power9 implementation of vec_all_ne does
37322         not use the AltiVec predicate framework.
37323         (VCMPAE_P): Add BU_P9V_OVERLOAD_2 macro to change implementation
37324         of vec_any_eq to not use AltiVec predicate framework.
37325         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Add
37326         support for predefined __POWER9_VECTOR__ macro to indicate that
37327         Power9 instruction selection is enabled.
37328         (altivec_overloaded_builtins): Remove extraneous
37329         ALTIVEC_BUILTIN_VEC_CMPNE entry for overloaded
37330         function argument types RS6000_BTI_bool_V16QI and
37331         RS6000_BTI_bool_V16QI.  Remove erroneous ALTIVEC_BUILTIN_VEC_CMPNE
37332         entry for overloaded function argument types RS6000_BTI_bool_V4SI
37333         andRS6000_BTI_bool_V4SI, mapping to P9V_BUILTIN_CMPNEB.  Remove
37334         two entries mapping to P9V_BUITIN_CMPNED and one entry mapping to
37335         P9V_BUILTIN_CMPNEF to force use of instructions not specific to
37336         Power9 for implementations of vec_cmpne.  Change the signature for
37337         all definitions of the overloaded P9V_BUILTIN_VEC_CMPNE_P function
37338         (representing vec_all_ne) to remove the previously described first
37339         argument of type RS6000_BTI_INTSI, as this was an artifact of
37340         reliance on the AltiVec predicate framework, which is no longer
37341         used in the implementation of these functions.  Add
37342         P9V_BUILTIN_VEC_VCMPAE_P entries (representing the vec_anyeq
37343         function) to match all of the P9V_BUILTIN_VEC_VCMNE_P entries
37344         since, unlike the AltiVec predicate framework implementation, we
37345         do not share function descriptors between vec_alle and vec_anyeq.
37346         (altivec_resolve_overloaded_builtin): Add SFmode and DFmode to the
37347         set of modes that receive special treatment even when
37348         TARGET_P9_VECTOR is true.  The special treatment emits code that
37349         does not depend on Power9 instructions.
37350         * config/rs6000/vector.md (vector_ne_<mode>_p): Change this
37351         define_expand to not rely on AltiVec predicate framework.
37352         (vector_ae_<mode>p): New define_expand to represent vec_any_eq
37353         function.
37354         (vector_ne_v2di_p): Change this define_expand to not rely on
37355         AltiVec predicate framework.
37356         (vector_ae_v2di_p): New define_expand to represent vec_any_eq
37357         function.
37358         (vector_ne_<mode>_p): Change this define_expand to not rely on
37359         AltiVec predicate framework.
37360         (vector_ae_<mode>p): New define_expand to represent vec_any_eq
37361         function.
37362         * config/rs6000/vsx.md (*vsx_ne_<mode>_p): For modes VSX_EXTRACT_I
37363         (V16QI, V8HI, V4SI), correct a typo in the code emitted for this
37364         define_insn pattern.
37365         (*vsx_ne_<mode>_p): For modes VSX_F (V4SF and V2DF), remove this
37366         define_insn pattern because the xvcmpne<VSs>. instruction is not
37367         supported.
37368         (vcmpne<VSs>): Remove this define_insn because xvcmpne<VSs>
37369         instruction is not supported.
37371 2017-03-01  Jakub Jelinek  <jakub@redhat.com>
37373         * config/nvptx/nvptx.c: Include intl.h.
37375 2017-03-01  Martin Jambor  <mjambor@suse.cz>
37377         PR lto/78140
37378         * ipa-prop.h (ipa_bits): Removed field known.
37379         (ipa_jump_func): Removed field vr_known.  Changed fields bits and m_vr
37380         to pointers.  Adjusted their comments to warn about their sharing.
37381         (ipcp_transformation_summary): Change bits to a vector of pointers.
37382         (ipa_check_create_edge_args): Moved to ipa-prop.c, declare.
37383         (ipa_get_ipa_bits_for_value): Declare.
37384         * tree-vrp.h (value_range): Mark as GTY((for_user)).
37385         * ipa-prop.c (ipa_bit_ggc_hash_traits): New.
37386         (ipa_bits_hash_table): Likewise.
37387         (ipa_vr_ggc_hash_traits): Likewise.
37388         (ipa_vr_hash_table): Likewise.
37389         (ipa_print_node_jump_functions_for_edge): Adjust for bits and m_vr
37390         being pointers and vr_known being removed.
37391         (ipa_set_jf_unknown): Likewise.
37392         (ipa_get_ipa_bits_for_value): New function.
37393         (ipa_set_jfunc_bits): Likewise.
37394         (ipa_get_value_range): New overloaded functions.
37395         (ipa_set_jfunc_vr): Likewise.
37396         (ipa_compute_jump_functions_for_edge): Use the above functions to
37397         construct bits and vr parts of jump functions.
37398         (ipa_check_create_edge_args): Move here from ipa-prop.h, also allocate
37399         ipa_bits_hash_table and ipa_vr_hash_table if they do not already
37400         exist.
37401         (ipcp_grow_transformations_if_necessary): Also allocate
37402         ipa_bits_hash_table and ipa_vr_hash_table if they do not already
37403         exist.
37404         (ipa_node_params_t::duplicate): Do not copy bits, just pointers to
37405         them.  Fix too long lines.
37406         (ipa_write_jump_function): Adjust for bits and m_vr being pointers and
37407         vr_known being removed.
37408         (ipa_read_jump_function): Use new setter functions to construct bits
37409         and vr parts of jump functions or set them to NULL.
37410         (write_ipcp_transformation_info): Adjust for bits being pointers.
37411         (read_ipcp_transformation_info): Likewise.
37412         (ipcp_update_bits): Likewise.  Fix excessively long lines a trailing
37413         space.
37414         Include gt-ipa-prop.h.
37415         * ipa-cp.c (propagate_bits_across_jump_function): Adjust for bits
37416         being pointers.
37417         (ipcp_store_bits_results): Likewise.
37418         (propagate_vr_across_jump_function): Adjust for m_vr being a pointer.
37419         Do not write to existing jump functions but use a temporary instead.
37421 2017-03-01  Jakub Jelinek  <jakub@redhat.com>
37423         PR c++/79681
37424         * fold-const.c (make_bit_field_ref): If orig_inner is COMPONENT_REF,
37425         attempt to use its first operand as BIT_FIELD_REF base.
37427 2017-03-01  Richard Biener  <rguenther@suse.de>
37429         PR middle-end/79721
37430         * tree-chrec.c (chrec_evaluate): Perform computation of Newtons
37431         interpolating formula in wrapping arithmetic.
37432         (chrec_apply): Convert chrec_evaluate return value to wanted type.
37434 2017-03-01  Jakub Jelinek  <jakub@redhat.com>
37436         PR tree-optimization/79734
37437         * tree-vect-generic.c (expand_vector_condition): Optimize
37438         AVX512 vector boolean VEC_COND_EXPRs into bitwise operations.
37439         Handle VEC_COND_EXPR where comparison has different inner width from
37440         type's inner width.
37442 2017-02-28  Sandra Loosemore  <sandra@codesourcery.com>
37444         * doc/invoke.texi (ARC Options): Copy-edit to fix punctuation,
37445         markup, and similar issues.  Remove @opindex entries for things
37446         that aren't options.  Add missing -mmpy-option entries.
37448 2017-02-28  Jakub Jelinek  <jakub@redhat.com>
37450         PR tree-optimization/79737
37451         * gimple-ssa-store-merging.c (encode_tree_to_bitpos): If bitlen is
37452         a multiple of BITS_PER_UNIT and !BYTES_BIG_ENDIAN, clear
37453         tmpbuf[byte_size - 1].  Call natice_encode_expr with byte_size - 1
37454         instead of byte_size.  Formatting fix.
37455         (shift_bytes_in_array_right): Formatting fix.
37457 2017-02-28  Eric Botcazou  <ebotcazou@adacore.com>
37459         PR target/79749
37460         * config/sparc/sparc.c (sparc_frame_pointer_required): Add missing
37461         condition on optimize for the leaf function test.
37463 2017-02-28  Martin Liska  <mliska@suse.cz>
37465         PR lto/79625
37466         * read-rtl-function.c (function_reader::handle_unknown_directive):
37467         Bail out when one uses -flto.
37469 2017-02-28  Martin Liska  <mliska@suse.cz>
37471         * common.opt: Replace space with tabular for options of <number>
37472         type.
37473         * config/i386/i386.opt: Show <number> value for
37474         -mlarge-data-threshold.
37475         * opts.c (print_filtered_help): Do not display number in hexadecimal
37476         format.
37478 2017-02-28  Martin Liska  <mliska@suse.cz>
37480         * common.opt: Fix --help=option -Q for options which are of
37481         an enum type.
37483 2017-02-28  Uros Bizjak  <ubizjak@gmail.com>
37485         * config/i386/i386.c (print_reg): Error out for values
37486         of 8-bit size in invalid integer register.
37488 2017-02-28  Martin Sebor  <msebor@redhat.com>
37490         PR tree-optimization/79691
37491         * passes.def (pass_all_optimizations_g): Enable pass_sprintf_length.
37493 2017-02-28  Jakub Jelinek  <jakub@redhat.com>
37495         PR target/79729
37496         * config/i386/i386.c (ix86_print_operand) <case 'R'>: Replace
37497         gcc_unreachable with output_operand_lossage.
37499 2017-02-28  Richard Biener  <rguenther@suse.de>
37501         PR tree-optimization/79740
37502         * tree-ssa-sccvn.c (vn_nary_op_insert_into): Allow redundant
37503         inserts.
37504         (visit_nary_op): Insert the nary into the hashtable if we
37505         pattern-matched sth.
37506         * tree-ssa-pre.c (eliminate_insert): Robustify.
37508 2017-02-28  Richard Biener  <rguenther@suse.de>
37510         PR middle-end/79731
37511         * fold-const.c (decode_field_reference): Reject out-of-bound
37512         accesses.
37514 2017-02-28  Jakub Jelinek  <jakub@redhat.com>
37516         * config/i386/i386.c: Include intl.h.
37517         (ix86_option_override_internal): Use cond ? G_("...") : G_("...")
37518         instead of just cond ? "..." : "...".
37519         * config/nvptx/nvptx.c (nvptx_goacc_validate_dims): Likewise.
37520         * coverage.c (read_counts_file): Likewise.
37521         * omp-offload.c: Include intl.h.
37522         (oacc_loop_fixed_partitions): Use cond ? G_("...") : G_("...") instead
37523         of just cond ? "..." : "...".
37524         * gcov.c (read_count_file): Use cond ? N_("...") : N_("...") instead
37525         of just cond ? "..." : "...".
37527 2017-02-28  Richard Earnshaw  <rearnsha@arm.com>
37529         PR target/79742
37530         * config/arm/parsecpu.awk (gen_data): Set tuning target to 'tune for'
37531         entry, if present.
37532         * config/arm/arm-cpus.in (cortex-m0plus.small-multiply): Correct
37533         'tune for' CPU name.
37534         * config/arm/arm-cpu-data.h: Regenerated.
37536 2017-02-28  Richard Biener  <rguenther@suse.de>
37538         PR tree-optimization/79732
37539         * tree-inline.c (expand_call_inline): Do not shadow var.
37541 2017-02-28  Richard Biener  <rguenther@suse.de>
37543         PR tree-optimization/79723
37544         * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Preserve
37545         address-space properly.
37547 2017-02-28  Thomas Schwinge  <thomas@codesourcery.com>
37549         * doc/optinfo.texi (Optimization groups): Fix option used for
37550         OPTGROUP_ALL.
37551         * doc/invoke.texi (-fopt-info): Document "omp".
37552         * dumpfile.h: Sort OPTGROUP_OMP before OPTGROUP_VEC.
37553         (OPTGROUP_ALL): Add OPTGROUP_OMP.
37554         * hsa-gen.c (pass_data_gen_hsail): Use OPTGROUP_OMP.
37555         * ipa-hsa.c (pass_data_ipa_hsa): Likewise.
37556         * omp-simd-clone.c (pass_data_omp_simd_clone): Likewise.
37558         * dumpfile.h (OPTGROUP_OPENMP): Rename to OPTGROUP_OMP.  Adjust
37559         all users.
37560         * dumpfile.c (optgroup_options): Instead of "openmp", associate
37561         OPTGROUP_OMP with "omp".
37563 2017-02-27  Pat Haugen  <pthaugen@us.ibm.com>
37565         PR target/79544
37566         * config/rs6000/rs6000-c.c (struct altivec_builtin_types): Use VSRAD
37567         for arithmetic shift of unsigned V2DI.
37569 2017-02-27  Claudiu Zissulescu  <claziss@synopsys.com>
37571         * config.gcc (arc*-): Clean up, use arc/big.h, arc/elf.h, and
37572         arc/linux.h headers.
37573         * config/arc/arc.h (TARGET_OS_CPP_BUILTINS): Remove.
37574         (LINK_SPEC): Likewise.
37575         (ARC_TLS_EXTRA_START_SPEC): Likewise.
37576         (EXTRA_SPECS): Likewise.
37577         (STARTFILE_SPEC): Likewise.
37578         (ENDFILE_SPEC): Likewise.
37579         (LIB_SPEC): Likewise.
37580         (TARGET_SDATA_DEFAULT): Likewise.
37581         (TARGET_MMEDIUM_CALLS_DEFAULT): Likewise.
37582         (MULTILIB_DEFAULTS): Likewise.
37583         (DWARF2_UNWIND_INFO): Likewise.
37584         * config/arc/big.h: New file.
37585         * config/arc/elf.h: Likewise.
37586         * config/arc/linux.h: Likewise.
37587         * config/arc/t-uClibc: Remove.
37589 2017-02-27  Bin Cheng  <bin.cheng@arm.com>
37591         PR tree-optimization/77536
37592         * tree-ssa-loop-manip.c (niter_for_unrolled_loop): New function.
37593         (tree_transform_and_unroll_loop): Use above function to compute the
37594         estimated niter of unrolled loop and use it when scaling profile.
37595         Also use count info rather than frequency if it's non-zero.
37596         * tree-ssa-loop-manip.h niter_for_unrolled_loop(): New declaration.
37597         * tree-vect-loop.c (scale_profile_for_vect_loop): New function.
37598         (vect_transform_loop): Call above function.
37600 2017-02-27  Richard Biener  <rguenther@suse.de>
37602         PR tree-optimization/45397
37603         * tree-ssa-pre.c (eliminate_insert): Handle BIT_AND_EXPR.
37604         * tree-ssa-sccvn.c (valueized_wider_op): New helper.
37605         (visit_nary_op): Add pattern matching for CSEing sign-changed
37606         or truncated operations with wider ones.
37608 2017-02-27  Richard Biener  <rguenther@suse.de>
37610         PR tree-optimization/79690
37611         * tree-vect-stmts.c (vectorizable_store): Use vector type
37612         built from the DR with address-space.
37614 2017-02-26  Gerald Pfeifer  <gerald@pfeifer.com>
37616         * doc/invoke.texi (Optimize Options): Refine the description
37617         of asan-use-after-return.
37619 2017-02-25  Alan Modra  <amodra@gmail.com>
37621         PR rtl-optimization/79584
37622         * lra-constraints.c (base_to_reg): Reload ad->base, the entire
37623         base, not ad->base_term, the reg within base.  Remove assertion
37624         that ad->base == ad->base_term.  Replace gen_int_mode using
37625         bogus mode with const0_rtx.
37627 2017-02-25  Jakub Jelinek  <jakub@redhat.com>
37629         PR middle-end/79396
37630         * tree-eh.c (operation_could_trap_p, stmt_could_throw_1_p): Handle
37631         FMA_EXPR like tcc_binary or tcc_unary.
37633         * tree-ssa-loop-niter.c (number_of_iterations_exit): Simplify warning.
37635         PR debug/77589
37636         * dwarf2out.c (struct dw_loc_list_struct): Add noted_variable_value
37637         bitfield.
37638         (size_of_loc_descr): Handle DW_OP_GNU_variable_value.
37639         (output_loc_operands): Handle DW_OP_call_ref and
37640         DW_OP_GNU_variable_value.
37641         (struct variable_value_struct): New type.
37642         (struct variable_value_hasher): Likewise.
37643         (variable_value_hash): New variable.
37644         (string_types): Remove.
37645         (copy_loc_descr): New function.
37646         (add_loc_descr_to_each): Clarify comment.  Use copy_loc_descr.
37647         (prepend_loc_descr_to_each): New function.
37648         (add_loc_list): Fix comment typo.  Use prepend_loc_descr_to_each
37649         instead of add_loc_descr_to_each if the first argument is single
37650         location list and the second has multiple.
37651         (resolve_args_picking_1): Handle DW_OP_GNU_variable_value.
37652         (loc_list_from_tree_1): For early_dwarf, emit DW_OP_GNU_variable_value
37653         when looking for variable value which doesn't have other location info.
37654         (loc_list_from_tree): Formatting fix.
37655         (gen_array_type_die): Simplify DW_AT_string_length handling.
37656         (adjust_string_types): Remove.
37657         (gen_subprogram_die): Don't call adjust_string_types nor test/set
37658         string_types.  Call resolve_variable_values.
37659         (prune_unused_types_walk_loc_descr): Handle DW_OP_GNU_variable_value.
37660         (resolve_addr_in_expr): Likewise.  Add A argument.
37661         (copy_deref_exprloc): Remove deref argument.  Adjust for the
37662         original expression being DW_OP_GNU_variable_value with optionally
37663         DW_OP_stack_value after it instead of DW_OP_call4 with DW_OP_deref
37664         optionally after it.
37665         (optimize_string_length): Rework for DW_OP_GNU_variable_value.
37666         (resolve_addr): Adjust optimize_string_length and resolve_addr_in_expr
37667         callers.  Set remove_AT_byte_size if removing DW_AT_string_length.
37668         (variable_value_hasher::hash, variable_value_hasher::equal): New
37669         methods.
37670         (resolve_variable_value_in_expr, resolve_variable_value,
37671         resolve_variable_values, note_variable_value_in_expr,
37672         note_variable_value): New functions.
37673         (dwarf2out_early_finish): Call note_variable_value on all toplevel
37674         DIEs.
37676 2017-02-24  Jakub Jelinek  <jakub@redhat.com>
37678         PR c/79677
37679         * opts.h (handle_generated_option): Add GENERATED_P argument.
37680         * opts-common.c (handle_option): Adjust function comment.
37681         (handle_generated_option): Add GENERATED_P argument, pass it to
37682         handle_option.
37683         (control_warning_option): Pass false to handle_generated_option
37684         GENERATED_P.
37685         * opts.c (maybe_default_option): Pass true to handle_generated_option
37686         GENERATED_P.
37687         * optc-gen.awk: Likewise.
37689 2017-02-24  Segher Boessenkool  <segher@kernel.crashing.org>
37691         * config/sh/sh.md (tstsi_t): If operands[0] is a SUBREG instead of
37692         a REG, look at the REG it is a SUBREG of.
37693         (splitter for cmpeqsi_t): Ditto.
37695 2017-02-24  Segher Boessenkool  <segher@kernel.crashing.org>
37697         * config/pa/pa.c (pa_combine_instructions): Do not share RTL.  Make
37698         the special USEs with the pattern of the insn, not the insn itself.
37700 2017-02-24  Matthew Fortune  <matthew.fortune@imgtec.com>
37702         PR target/79473
37703         * doc/invoke.texi: Document -mload-store-pairs.
37705 2017-02-24  Segher Boessenkool  <segher@kernel.crashing.org>
37706             Sandra Loosemore  <sandra@codesourcery.com>
37708         * config/nios2/nios2.c (nios2_simple_const_p): Returns false if the
37709         argument isn't a CONST_INT.
37710         (nios2_alternate_compare_const): Assert op is a CONST_INT.
37711         (nios2_valid_compare_const_p): Assert op is a CONST_INT.
37712         (nios2_validate_compare): Bypass alternate compare logic if *op2
37713         is not a CONST_INT.
37714         (ldstwm_operation_p): Return false if first_base is not a REG or
37715         if first_offset is not a CONST_INT.
37717 2017-02-24  Segher Boessenkool  <segher@kernel.crashing.org>
37719         * config/cris/cris.md: Use correct operand in a define_peephole2.
37721 2017-02-24  Segher Boessenkool  <segher@kernel.crashing.org>
37723         * config/c6x/c6x.c (predicate_insn): Do not incorrectly share RTL.
37725 2017-02-24  Segher Boessenkool  <segher@kernel.crashing.org>
37727         * config/arc/arc.c (arc_ccfsm_advance): Only take the PATTERN of
37728         this_insn if it is an INSN or JUMP_INSN.
37729         (force_offsettable): Look at base, not at addr.
37730         * config/arc/predicates.md (brcc_nolimm_operator): Don't call INTVAL
37731         on things that aren't necessarily CONST_INTs.
37733 2017-02-24  Uros Bizjak  <ubizjak@gmail.com>
37735         * doc/invoke.texi (x86 Options, -mfpmath=sse): Mention that
37736         -mfpmath=sse is the default also for x86-32 targets with SSE2
37737         instruction set when @option{-ffast-math} is enabled
37739 2017-02-24  Jeff Law  <law@redhat.com>
37741         PR rtl-optimizatoin/79286
37742         * ira.c (update_equiv_regs): Drop may_trap_p exception to
37743         dominance test.
37745 2017-02-24  Richard Biener  <rguenther@suse.de>
37747         PR tree-optimization/79389
37748         * gimple-ssa-split-paths.c (is_feasible_trace): Properly skip
37749         debug insns.
37751 2017-02-24  Aldy Hernandez  <aldyh@redhat.com>
37753         * tree-ssa-loop-niter.c (number_of_iterations_exit): Update
37754         function comment to reflect reality.
37755         (loop_exits_before_overflow): Fix typo in function description.
37757 2017-02-24  Richard Biener  <rguenther@suse.de>
37759         PR tree-optimization/79389
37760         * gimple-ssa-split-paths.c (is_feasible_trace): Verify more
37761         properly that a threading opportunity exists.  Detect conditional
37762         copy/constant propagation opportunities.
37764 2017-02-23  Eric Botcazou  <ebotcazou@adacore.com>
37766         * config/visium/visium.md (type): Add trap.
37767         (b): New mode attribute.
37768         (*btst): Rename into...
37769         (*btst<mode>): ...this and adjust.
37770         (*cbranchsi4_btst_insn): Rename into...
37771         (*cbranch<mode>4_btst_insn): ...this and adjust.
37772         (trap): New define_insn.
37774 2017-02-23  Jakub Jelinek  <jakub@redhat.com>
37776         PR tree-optimization/79389
37777         * ifcvt.c (struct noce_if_info): Add rev_cond field.
37778         (noce_reversed_cond_code): New function.
37779         (noce_emit_store_flag): Use rev_cond if non-NULL instead of
37780         reversed_comparison_code.  Formatting fix.
37781         (noce_try_store_flag): Test rev_cond != NULL in addition to
37782         reversed_comparison_code.
37783         (noce_try_store_flag_constants): Likewise.
37784         (noce_try_store_flag_mask): Likewise.
37785         (noce_try_addcc): Use rev_cond if non-NULL instead of
37786         reversed_comparison_code.
37787         (noce_try_cmove_arith): Likewise.  Formatting fixes.
37788         (noce_try_minmax, noce_try_abs): Clear rev_cond.
37789         (noce_find_if_block): Initialize rev_cond.
37790         (find_cond_trap): Call noce_get_condition with then_bb == trap_bb
37791         instead of false as last argument never attempt to reverse it
37792         afterwards.
37794 2017-02-23  Bin Cheng  <bin.cheng@arm.com>
37796         PR tree-optimization/79663
37797         * tree-predcom.c (combine_chains): Process refs in reverse order
37798         only for ZERO length chains, and add explaining comment.
37800 2017-02-23  Jeff Law  <law@redhat.com>
37802         PR tree-optimization/79578
37803         * tree-ssa-dse.c (clear_bytes_written_by): Use OEP_ADDRESS_OF
37804         in call to operand_equal_p.
37806 2017-01-23  Dominique d'Humieres  <dominiq@lps.ens.fr>
37808         PR target/71017
37809         * config/i386/cpuid.h: Fix another undefined behavior.
37811 2017-02-23  Richard Biener  <rguenther@suse.de>
37813         PR tree-optimization/79683
37814         * tree-vect-stmts.c (vect_analyze_stmt): Do not overwrite
37815         vector types for data-refs.
37817 2017-02-23  Martin Liska  <mliska@suse.cz>
37819         * params.def (PARAM_MIN_NONDEBUG_INSN_UID): Change default to 0.
37821 2017-02-23  Jakub Jelinek  <jakub@redhat.com>
37823         PR middle-end/79665
37824         * internal-fn.c (get_range_pos_neg): Moved to ...
37825         * tree.c (get_range_pos_neg): ... here.  No longer static.
37826         * tree.h (get_range_pos_neg): New prototype.
37827         * expr.c (expand_expr_real_2) <case TRUNC_DIV_EXPR>: If both arguments
37828         are known to be in between 0 and signed maximum inclusive, try to
37829         expand both unsigned and signed divmod and use the cheaper one from
37830         those.
37832 2017-02-22  Jeff Law  <law@redhat.com>
37834         PR tree-optimization/79578
37835         * tree-ssa-dse.c (clear_bytes_written_by): Use operand_equal_p
37836         to compare base operands.
37838 2017-02-22  Segher Boessenkool  <segher@kernel.crashing.org>
37840         PR target/79211
37841         * config/rs6000/rs6000.md (*fsel<SFDF:mode><SFDF2:mode>4): Use
37842         gpc_reg_operand instead of fpr_reg_operand.
37844 2017-02-22  Sameera Deshpande  <sameera.deshpande@imgtec.com>
37846         * config/mips/mips.c (mips_return_in_memory): Force FP
37847         vector types to be returned in memory for o32 ABI.
37849 2017-02-22  Jakub Jelinek  <jakub@redhat.com>
37851         * dwarf2out.c (gen_variable_die): For -gdwarf-5, use DW_TAG_variable
37852         instead of DW_TAG_member for static data member declarations and don't
37853         set no_linkage_name for static inline data members.
37854         (gen_member_die): For -gdwarf-5 don't change DW_TAG_variable
37855         to DW_TAG_member.
37857 2017-02-22  Martin Liska  <mliska@suse.cz>
37859         * doc/invoke.texi: Replace inequality signs with square brackets
37860         for -Wnormalized.
37862 2017-02-22  Matthew Fortune  <matthew.fortune@imgtec.com>
37864         PR target/78660
37865         * lra-constraints.c (simplify_operand_subreg): Handle
37866         WORD_REGISTER_OPERATIONS targets.
37868 2017-02-22  Jakub Jelinek  <jakub@redhat.com>
37870         PR target/70465
37871         * reg-stack.c (emit_swap_insn): Treat (float_extend:?F (mem:?F))
37872         and (const_double:?F) like (mem:?F) for the purpose of fxch %st(1)
37873         elimination by swapping fld*.
37875 2017-02-22  Richard Biener  <rguenther@suse.de>
37877         PR tree-optimization/79673
37878         * tree-ssa-pre.c (compute_avail): Use wide_int_to_tree to
37879         convert the [TARGET_]MEM_REF offset INTEGER_CST, scrapping off
37880         irrelevant address-space qualifiers and avoiding a
37881         ADDR_SPACE_CONVERT_EXPR from fold_convert.
37883 2017-02-22  Richard Biener  <rguenther@suse.de>
37885         PR tree-optimization/79666
37886         * tree-vrp.c (extract_range_from_binary_expr_1): Make sure
37887         to not symbolically negate if that may introduce undefined
37888         overflow.
37890 2017-02-22  Martin Liska  <mliska@suse.cz>
37892         PR lto/79587
37893         * data-streamer-in.c (streamer_read_gcov_count): Remove assert.
37894         * data-streamer-out.c (streamer_write_gcov_count_stream):
37895         Likewise.
37896         * value-prof.c (stream_out_histogram_value): Make assert more
37897         precise based on type of counter.
37899 2017-02-21  Uros Bizjak  <ubizjak@gmail.com>
37901         PR target/79593
37902         * config/i386/i386.md (standard_x87sse_constant_load splitter):
37903         Use nonimmediate_operand instead of memory_operand for operand 1.
37904         (float-extend standard_x87sse_constant_load splitter): Ditto.
37906 2017-02-21  Jeff Law  <law@redhat.com>
37908         PR tree-optimization/79621
37909         * gimple-ssa-isolate-paths.c (find_implicit_erroneous_behavior): Ignore
37910         blocks with edges to themselves.
37912 2017-02-21  Jakub Jelinek  <jakub@redhat.com>
37914         PR target/79633
37915         * tree-chkp-opt.c (chkp_optimize_string_function_calls): Use
37916         is_gimple_call instead of comparing gimple_code with GIMPLE_CALL.
37917         Use gimple_call_builtin_p.
37919         PR target/79570
37920         * sel-sched.c (moveup_expr_cached): Don't call sel_bb_head
37921         on temporarily removed DEBUG_INSNs.
37923         PR tree-optimization/79649
37924         * tree-loop-distribution.c (classify_partition): Give up on
37925         non-generic address space loads/stores.
37927 2017-02-21  Aldy Hernandez  <aldyh@redhat.com>
37929         * doc/loop.texi (Loop manipulation): Remove nonexistent
37930         tree_ssa_loop_version from the documentation.
37931         * cfgloopmanip.c (loop_version): Document CONDITION_BB argument.
37933 2017-02-21  Jakub Jelinek  <jakub@redhat.com>
37935         PR target/79494
37936         * config/i386/i386.c (ix86_expand_split_stack_prologue): Call
37937         make_reg_eh_region_note_nothrow_nononlocal on call_insn.
37938         * config/rs6000/rs6000.c: Include except.h.
37939         (rs6000_expand_split_stack_prologue): Call
37940         make_reg_eh_region_note_nothrow_nononlocal on the call insn.
37942 2017-02-21  Martin Jambor  <mjambor@suse.cz>
37944         PR lto/79579
37945         * ipa-prop.c (ipa_prop_write_jump_functions): Bail out if no edges
37946         have been analyzed.
37948 2017-02-21  Martin Jambor  <mjambor@suse.cz>
37950         * common.opt (-fipa-cp-alignment): Mark as ignored and preserved
37951         for backward compatibility only.
37952         * doc/invoke.texi (Option Summary): Remove all references to
37953         -fipa-cp-alignment.
37955 2017-02-21  Matthew Fortune  <matthew.fortune@imgtec.com>
37957         PR target/78660
37958         Revert:
37959         2017-02-20  Matthew Fortune  <matthew.fortune@imgtec.com>
37961         * lra-constraints.c (curr_insn_transform): Handle
37962         WORD_REGISTER_OPERATIONS requirements when reloading SUBREGs.
37964 2017-02-21  Martin Liska  <mliska@suse.cz>
37966         * config/i386/i386.opt: Replace -masm-dialect with -masm.
37968 2017-02-21  Thomas Schwinge  <thomas@codesourcery.com>
37970         PR translation/79638
37971         * config/nvptx/nvptx.c (ENTRY_TEMPLATE): Single out "%ntid.y".
37973 2017-02-21  Eric Botcazou  <ebotcazou@adacore.com>
37975         PR ada/67205
37976         * config/arm/arm.c (TARGET_CUSTOM_FUNCTION_DESCRIPTORS): Define.
37977         (arm_function_ok_for_sibcall): Return false for an indirect call by
37978         descriptor if all the argument registers are used.
37979         (arm_relayout_function): Use FUNCTION_ALIGNMENT macro to adjust the
37980         alignment of the function.
37982 2017-02-21  Jakub Jelinek  <jakub@redhat.com>
37984         PR tree-optimization/61441
37985         * simplify-rtx.c (simplify_const_unary_operation): For
37986         -fsignaling-nans and sNaN operand, return NULL_RTX rather than
37987         the sNaN unmodified.
37989 2017-02-20  Bernd Edlinger  <bernd.edlinger@hotmail.de>
37991         * Makefile.in (BUILD_SYSTEM_HEADER_DIR): New make variabe.
37992         (LIMITS_H_TEST, if_multiarch, stmp-fixinc): Use BUILD_SYSTEM_HEADER_DIR
37993         instead of SYSTEM_HEADER_DIR.
37995 2017-02-20  Gerald Pfeifer  <gerald@pfeifer.com>
37996             Martin LiÅ¡ka  <mliska@suse.cz>
37998         * doc/invoke.texi (use-after-scope-direct-emission-threshold):
37999         Fix typos and grammar, use active voice, and clarify.
38001 2017-02-20  Marek Polacek  <polacek@redhat.com>
38003         PR middle-end/79537
38004         * gimplify.c (gimplify_expr): Handle unused *&&L;.
38006         PR sanitizer/79558
38007         * ubsan.c (ubsan_type_descriptor): Check if TYPE_MAX_VALUE is null.
38009 2017-02-20  Jakub Jelinek  <jakub@redhat.com>
38011         PR target/79568
38012         * config/i386/i386.c (ix86_expand_builtin): Handle
38013         OPTION_MASK_ISA_AVX512VL and OPTION_MASK_ISA_64BIT in
38014         ix86_builtins_isa[fcode].isa as a requirement of those
38015         flags and any other flag in the bitmask.
38016         (ix86_init_mmx_sse_builtins): Use 0 instead of
38017         ~OPTION_MASK_ISA_64BIT as mask.
38018         * config/i386/i386-builtin.def (__builtin_ia32_rdtsc,
38019         __builtin_ia32_rdtscp, __builtin_ia32_pause, __builtin_ia32_bsrsi,
38020         __builtin_ia32_rdpmc, __builtin_ia32_rolqi, __builtin_ia32_rolhi,
38021         __builtin_ia32_rorqi, __builtin_ia32_rorhi): Likewise.
38023 2017-02-20  Matthew Fortune  <matthew.fortune@imgtec.com>
38025         PR target/78012
38026         * lra-constraints.c (split_reg): Check requested split mode
38027         is supported by the register.
38029 2017-02-20  Matthew Fortune  <matthew.fortune@imgtec.com>
38031         * lra-constraints.c (simplify_operand_subreg): Remove early
38032         return false.
38034 2017-02-20  Matthew Fortune  <matthew.fortune@imgtec.com>
38036         PR target/78660
38037         * lra-constraints.c (curr_insn_transform): Tighten condition
38038         for converting SUBREG reloads from OP_OUT to OP_INOUT.
38040 2017-02-20  Matthew Fortune  <matthew.fortune@imgtec.com>
38042         PR target/78660
38043         * lra-constraints.c (curr_insn_transform): Handle
38044         WORD_REGISTER_OPERATIONS requirements when reloading SUBREGs.
38046 2017-02-19  Uros Bizjak  <ubizjak@gmail.com>
38048         Revert:
38049         2016-05-30  Uros Bizjak  <ubizjak@gmail.com>
38051         * config/i386/sync.md (mfence_nosse): Use "lock orl $0, -4(%esp)".
38053 2017-02-19  Jonathan Wakely  <jwakely@redhat.com>
38055         PR c++/69523
38056         * doc/invoke.texi (C++ Dialect Options) [-Wliteral-suffix]: Update
38057         description.
38059 2017-02-19  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
38061         * gimple-pretty-print.c (dump_ternary_rhs): Adjust gimple dump format
38062         for FMA_EXPR.
38064 2017-02-18  Jakub Jelinek  <jakub@redhat.com>
38066         * final.c (last_columnnum, override_columnnum): New variables.
38067         (final_start_function): Set last_columnnum, pass it to begin_prologue
38068         hook and pass 0 to dwarf2out_begin_prologue.
38069         (final_scan_insn): Update override_columnnum.  Pass last_columnnum
38070         to source_line debug hook.
38071         (notice_source_line): Compute last_columnnum and for debug_column_info
38072         return true on column changes.
38073         * debug.h (struct gcc_debug_hooks): Add column argument to
38074         source_line and begin_prologue hooks.
38075         (debug_nothing_int_charstar_int_bool): Remove prototype.
38076         (debug_nothing_int_int_charstar,
38077         debug_nothing_int_int_charstar_int_bool): New prototypes.
38078         (dwarf2out_begin_prologue): Add column argument.
38079         * debug.c (do_nothing_debug_hooks): Adjust source_line and
38080         begin_prologue hooks.
38081         (debug_nothing_int_charstar_int_bool): Remove.
38082         (debug_nothing_int_int_charstar,
38083         debug_nothing_int_int_charstar_int_bool): New functions.
38084         * dwarf2out.c (dwarf2out_begin_prologue): Add column argument, pass it
38085         through to dwarf2out_source_line.
38086         (dwarf2_lineno_debug_hooks): Adjust begin_prologue hook.
38087         (dwarf2out_source_line): Add column argument, emit it if requested.
38088         * sdbout.c (sdbout_source_line, sdbout_begin_prologue): Add column
38089         arguments.
38090         * xcoffout.h (xcoffout_begin_prologue, xcoffout_source_line): Likewise.
38091         * xcoffout.c (xcoffout_begin_prologue, xcoffout_source_line): Likewise.
38092         * vmsdbgout.c (vmsdbgout_begin_prologue): Add column argument, pass it
38093         through to dwarf2out_begin_prologue.
38094         (vmsdbgout_source_line): Add column argument, pass it through to
38095         dwarf2out_source_line.
38096         * dbxout.c (dbxout_begin_prologue): Add column argument, adjust
38097         dbxout_source_line caller.
38098         (dbxout_source_line): Add column argument.
38100         * common.opt (gno-column-info, gcolumn-info): New options.
38101         * dwarf2out.c (dwarf2_lineno_debug_hooks): Formatting fix.
38102         (check_die): Also test for multiple DW_AT_decl_column attributes.
38103         (add_src_coords_attributes, dwarf2out_imported_module_or_decl_1): Add
38104         DW_AT_decl_column if requested.
38105         (gen_subprogram_die): Compare and/or add also DW_AT_decl_column
38106         if requested.
38107         (gen_variable_die): Likewise.
38108         (add_call_src_coords_attributes): Add DW_AT_call_column if requested.
38109         * doc/invoke.texi (-gcolumn-info, -gno-column-info): Document.
38111         PR target/79569
38112         * config/i386/i386.opt (m3dnowa): Replace Undocumented with Report.
38113         * common/config/i386/i386-common.c (OPTION_MASK_ISA_3DNOW_A_SET): Define.
38114         (ix86_handle_option): Handle OPT_m3dnowa.
38115         * doc/invoke.texi (-m3dnowa): Document.
38116         * doc/extend.texi (__builtin_ia32_pmulhuw, __builtin_ia32_pf2iw): Use
38117         -m3dnowa instead of -m3dnow -march=athlon.
38119         PR target/79559
38120         * config/i386/i386.c (ix86_print_operand): Use output_operand_lossage
38121         instead of gcc_assert for K, r and R code checks.  Formatting fixes.
38123 2017-02-17  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
38125         PR target/79261
38126         * config/rs6000/rs6000.c (rs6000_expand_ternop_builtin): Add
38127         support for CODE_FOR_vsx_xxpermdi_v2d[fi]_be.
38128         * config/rs6000/rs6000.md (reload_gpr_from_vsx<mode>): Call
38129         generator for vsx_xxpermdi_<mode>_be.
38130         * config/rs6000/vsx.md (vsx_xxpermdi_<mode>): Remove logic to
38131         force big-endian semantics.
38132         (vsx_xxpermdi_<mode>_be): New define_expand with same
38133         implementation as previous version of vsx_xxpermdi_<mode>.
38135 2017-02-17  Jakub Jelinek  <jakub@redhat.com>
38137         PR tree-optimization/79327
38138         * gimple-ssa-sprintf.c (format_integer): Remove likely_adjust
38139         variable, its initialization and use.
38141 2017-02-17  Julia Koval  <julia.koval@intel.com>
38143         * common/config/i386/i386-common.c (OPTION_MASK_ISA_RDPID_SET): New.
38144         (OPTION_MASK_ISA_PKU_UNSET): New.
38145         (ix86_handle_option): Handle -mrdpid.
38146         * config/i386/cpuid.h (bit_RDPID): New.
38147         * config/i386/driver-i386.c (host_detect_local_cpu):
38148         Detect RDPID feature.
38149         * config/i386/i386-builtin.def (__builtin_ia32_rdpid): New.
38150         * config/i386/i386-c.c (ix86_target_macros_internal):
38151         Handle RDPID flag.
38152         * config/i386/i386.c (ix86_target_string): Add -mrdpid to isa2_opts.
38153         (ix86_valid_target_attribute_inner_p): Add "rdpid".
38154         (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
38155         * config/i386/i386.h (TARGET_RDPID, TARGET_RDPID_P): New.
38156         * config/i386/i386.md (define_insn "rdpid"): New.
38157         * config/i386/i386.opt Add -mrdpid.
38158         * config/i386/immintrin.h (_rdpid_u32): New.
38160 2017-02-17  Vladimir Makarov  <vmakarov@redhat.com>
38162         PR rtl-optimization/79541
38163         * lra-constraints.c (curr_insn_transform): Remove wrong asm insn
38164         instead of transforming it into USE.
38166 2017-02-17  Segher Boessenkool  <segher@kernel.crashing.org>
38168         * config/rs6000/rs6000.md (extendsfdf2): Remove default arguments.
38169         If HONOR_SNANS (SFmode) force the input to a register.
38170         (*extendsfdf2_fpr): Add !HONOR_SNANS (SFmode) condition.
38171         (*extendsfdf2_snan): New pattern, used when using SNaNs; it generates
38172         an frsp or similar insn.
38174 2017-02-17  Martin Liska  <mliska@suse.cz>
38176         PR rtl-optimization/79577
38177         * params.def (selsched-max-sched-times): Increase minimum to 1.
38179 2017-02-17  Martin Liska  <mliska@suse.cz>
38181         PR rtl-optimization/79574
38182         * gcse.c (want_to_gcse_p): Prevent integer overflow.
38184 2017-02-17  Martin Liska  <mliska@suse.cz>
38186         PR tree-optimization/79529
38187         * tree-ssa-loop-unswitch.c (is_maybe_undefined): Use
38188         ssa_defined_default_def_p to handle cases which are implicitly
38189         defined.
38190         * tree-ssa.c (ssa_defined_default_def_p): New function.
38191         (ssa_undefined_value_p): Use ssa_defined_default_def_p to handle cases
38192         which are implicitly defined.
38193         * tree-ssa.h (ssa_defined_default_def_p): Declare.
38195 2017-02-17  Richard Biener  <rguenther@suse.de>
38197         PR middle-end/79576
38198         * params.def (max-ssa-name-query-depth): Limit to 10.
38200 2017-02-17  Richard Biener  <rguenther@suse.de>
38202         PR tree-optimization/79552
38203         * tree-ssa-structalias.c (visit_loadstore): Properly verify
38204         default defs.
38206 2017-02-17  Richard Biener  <rguenther@suse.de>
38208         PR bootstrap/79567
38209         * genmatch.c (output_line_directive): Handle DIR_SEPARATOR_2.
38211 2017-02-17  Marek Polacek  <polacek@redhat.com>
38213         PR middle-end/79536
38214         * fold-const.c (fold_negate_expr_1): Renamed from fold_negate_expr.
38215         (fold_negate_expr): New wrapper.
38217 2017-02-16  Sandra Loosemore  <sandra@codesourcery.com>
38219         * doc/invoke.texi (C++ Dialect Options) [-Wno-non-template-friend]: 
38220         Correct terminology and de-emphasize pre-standard behavior.
38222 2017-02-16  Alan Modra  <amodra@gmail.com>
38224         PR rtl-optimization/79286
38225         * ira.c (def_dominates_uses): New function.
38226         (update_equiv_regs): Don't create an equivalence for insns that
38227         may trap where the register def does not dominate the use.
38229 2017-02-16  Vladimir Makarov  <vmakarov@redhat.com>
38231         PR rtl-optimization/78127
38232         * lra.c (lra): Call lra_eliminate before finish the loop after
38233         lra_constraint.
38235 2017-02-16  Richard Biener  <rguenther@suse.de>
38237         * graphite.h: Do not include isl/isl_val_gmp.h, instead include
38238         isl/isl_val.h.
38239         * graphite-isl-ast-to-gimple.c (gmp_cst_to_tree): Remove.
38240         (gcc_expression_from_isl_expr_int): Use generic isl_val interface.
38241         * graphite-sese-to-poly.c: Do not include isl/isl_val_gmp.h.
38242         (isl_val_int_from_wi): New function.
38243         (extract_affine_gmp): Rename to ...
38244         (extract_affine_wi): ... this, take a widest_int.
38245         (extract_affine_int): Just wrap extract_affine_wi.
38246         (add_param_constraints): Use isl_val_int_from_wi.
38247         (add_loop_constraints): Likewise, and extract_affine_wi.
38249 2017-02-15  Jeff Law  <law@redhat.com>
38251         PR middle-end/79521
38252         * ira-costs.c (scan_one_insn): Check have_regs_of_mode before calling
38253         ira_init_register_move_cost_if_necessary.
38255 2017-02-15  Martin Sebor  <msebor@redhat.com>
38257         PR middle-end/32003
38258         * doc/invoke.texi (-fdump-final-insns): Replace option accidentally
38259         removed in a prior commit.
38261 2017-02-15  Bin Cheng  <bin.cheng@arm.com>
38263         PR tree-optimization/79347
38264         * tree-vect-loop-manip.c (vect_do_peeling): Maintain profile
38265         counters during peeling.
38267 2017-02-15  Thomas Schwinge  <thomas@codesourcery.com>
38269         * Makefile.in (site.exp): Remove "set ISLVER".
38271 2017-02-15  Jakub Jelinek  <jakub@redhat.com>
38273         PR target/79487
38274         * real.c (real_from_integer): Call real_convert even for decimal.
38276 2017-02-15  Dominik Vogt  <vogt@linux.vnet.ibm.com>
38278         PR target/79421
38279         * config/s390/s390.c: define TARGET_CUSTOM_FUNCTION_DESCRIPTORS.
38281 2017-02-14  Andrew Pinski  <apinski@cavium.com>
38283         * config/aarch64/aarch64-cores.def (thunderx2t99): Move to under 'C"
38284         cores and change the partno/implementer to be correct.
38285         (thunderx2t99p1): New core which replaces thunderx2t99 and still has
38286         the 'B" as the implementer.
38287         * config/aarch64/aarch64-tune.md: Regenerate.
38289 2017-02-14  Carl Love  <cel@us.ibm.com>
38291         * config/rs6000/rs6000.c: Add case statement entry to make the
38292         xvcvuxdsp built-in argument unsigned.
38293         * config/rs6000/vsx.md: Fix the source and return operand types so they
38294         match the instruction definitions from the ISA document.  Fix typo
38295         in the instruction generation for the (define_insn "vsx_xvcvuxdsp"
38296         statement.
38298 2017-02-14  Vladimir Makarov  <vmakarov@redhat.com>
38300         PR target/79282
38301         * lra-int.h (struct lra_operand_data, struct lra_insn_reg): Add
38302         member early_clobber_alts.
38303         * lra-lives.c (reg_early_clobber_p): New.
38304         (process_bb_lives): Use it.
38305         * lra.c (new_insn_reg): New arg early_clobber_alts.  Use it.
38306         (debug_operand_data): Initialize early_clobber_alts.
38307         (setup_operand_alternative): Set up early_clobber_alts.
38308         (collect_non_operand_hard_regs): Ditto.  Pass early clobber
38309         alternatives to new_insn_reg.
38310         (add_regs_to_insn_regno_info): Add arg early_clobber_alts.  Use
38311         it.
38312         (lra_update_insn_regno_info): Pass the new arg.
38314 2017-02-14  Jakub Jelinek  <jakub@redhat.com>
38316         PR middle-end/79505
38317         * omp-offload.c (free_oacc_loop): Release loop->ifns vector.
38318         (new_oacc_loop_raw): Don't clear already cleared fields.
38320         PR target/79481
38321         * config/i386/avx512pfintrin.h (_mm512_prefetch_i32gather_pd,
38322         _mm512_prefetch_i32gather_ps, _mm512_prefetch_i64gather_pd,
38323         _mm512_prefetch_i64gather_ps): New inline functions and macros.
38325 2017-02-14  Uros Bizjak  <ubizjak@gmail.com>
38327         PR target/79495
38328         * config/i386/i386.md (*movxf_internal): Add (o,rC) alternative.
38330 2017-02-14  H.J. Lu  <hongjiu.lu@intel.com>
38332         PR target/79498
38333         * config/i386/i386.c (timode_scalar_chain::convert_insn): Insert
38334         the extra instruction to the right place to store 128-bit constant
38335         when needed.
38337 2017-02-14  Martin Sebor  <msebor@redhat.com>
38339         PR middle-end/79448
38340         * gimple-ssa-sprintf.c (format_directive): Avoid issuing INT_MAX
38341           warning for strings of unknown length.
38343 2017-02-13  Segher Boessenkool  <segher@kernel.crashing.org>
38345         * config.gcc (supported_defaults) [powerpc*-*-*]: Update.
38347 2017-02-14  Jeff Law  <law@redhat.com>
38349         PR target/79404
38350         * ira-costs.c (scan_one_insn): Initialize register move costs
38351         for pseudos seen in USE/CLOBBER insns.
38353         PR tree-optimization/79095
38354         * tree-vrp.c (extract_range_from_binary_expr_1): For EXACT_DIV_EXPR,
38355         if the numerator has the range ~[0,0] make the resultant range ~[0,0].
38356         (extract_range_from_binary_expr): For MINUS_EXPR with no derived range,
38357         if the operands are known to be not equal, then the resulting range
38358         is ~[0,0].
38359         (intersect_ranges): If the new range is ~[0,0] and the old range is
38360         wide, then prefer ~[0,0].
38361         * tree-vrp.c (overflow_comparison_p_1): New function.
38362         (overflow_comparison_p): New function.
38363         * tree-vrp.c (register_edge_assert_for_2): Register additional asserts
38364         if NAME is used in an overflow test.
38365         (vrp_evaluate_conditional_warnv_with_ops): If the ops represent an
38366         overflow check that can be expressed as an equality test, then adjust
38367         ops to be that equality test.
38369 2017-02-14  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
38371         * config/s390/s390-builtin-types.def: Remove flags argument.
38372         * config/s390/s390.c (s390_init_builtins): Likewise.
38374 2017-02-14  Martin Liska  <mliska@suse.cz>
38376         * tree-ssa-loop-unswitch.c (hoist_guard): Release get_loop_body
38377         vector.  Fix trailing white spaces.
38379 2017-02-14  James Greenhalgh  <james.greenhalgh@arm.com>
38381         * config/aarch64/aarch64.c (aarch64_simd_container_mode): Handle
38382         HFmode.
38384 2017-02-14  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
38386         PR rtl-optimization/68664
38387         * config/arm/arm.c (arm_sched_can_speculate_insn):
38388         New function.  Declare prototype.
38389         (TARGET_SCHED_CAN_SPECULATE_INSN): Define.
38391 2017-02-14  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
38393         PR rtl-optimization/68664
38394         * config/aarch64/aarch64.c (aarch64_sched_can_speculate_insn):
38395         New function.
38396         (TARGET_SCHED_CAN_SPECULATE_INSN): Define.
38398 2017-02-14  Amit Pawar  <amit.pawar@amd.com>
38400         * config/i386/i386.c (znver1_cost): Fix the alignment for function and
38401         max skip bytes for function, loop and jump.
38403 2017-02-14  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
38405         * gimple-pretty-print.c (dump_unary_rhs): Adjust dump format for
38406         ABS_EXPR for gimple dump.
38408 2017-02-14  Jakub Jelinek  <jakub@redhat.com>
38410         PR target/79462
38411         * config/sh/sh.c (expand_cbranchdi4): Don't clear operands[4].
38413         PR tree-optimization/79408
38414         * tree-vrp.c (simplify_div_or_mod_using_ranges): Handle also the
38415         case when on TRUNC_MOD_EXPR op0 is INTEGER_CST.
38416         (simplify_stmt_using_ranges): Call simplify_div_or_mod_using_ranges
38417         also if rhs1 is INTEGER_CST.
38419 2017-02-14  Richard Biener  <rguenther@suse.de>
38421         PR middle-end/79432
38422         * tree-into-ssa.c (insert_phi_nodes): When the function can
38423         have abnormal edges rewrite SSA names with broken use-def
38424         dominance out of SSA and register them for PHI insertion.
38426 2017-02-13  Martin Sebor  <msebor@redhat.com>
38428         PR middle-end/79496
38429         * gimple-ssa-sprintf.c (pass_sprintf_length::handle_gimple_call): Avoid
38430         clearing info.nowrite flag when snprintf size argument is a range.
38432 2017-02-13  Jakub Jelinek  <jakub@redhat.com>
38434         * cprop.c (cprop_jump): Add missing space in string literal.
38435         * tree-ssa-structalias.c (rewrite_constraints): Likewise.
38436         (get_constraint_for_component_ref): Likewise.
38437         * df-core.c (df_worklist_dataflow_doublequeue): Likewise.
38438         * tree-outof-ssa.c (insert_partition_copy_on_edge): Likewise.
38439         * lra-constraints.c (process_alt_operands): Likewise.
38440         * ipa-inline.c (inline_small_functions): Likewise.
38441         * tree-ssa-sccvn.c (visit_reference_op_store): Likewise.
38442         * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
38443         * trans-mem.c (diagnose_tm_1_op): Likewise.
38444         * omp-grid.c (grid_find_single_omp_among_assignments): Likewise.
38445         (grid_parallel_clauses_gridifiable): Likewise.
38447         * config/nvptx/mkoffload.c (process): Add space in between
38448         , and %d.
38450         * config/i386/i386.h (REG_CLASS_NAMES): Add , in between
38451         "MOD4_SSE_REGS" and "ALL_REGS".
38453         * spellcheck.c (test_data): Add , in between "foo" and "food".
38455 2017-02-13  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
38457         PR target/79449
38458         * config/rs6000/rs6000.c (expand_block_compare): Make sure runtime
38459         boundary crossing check and subsequent code generation agree.
38461 2017-02-13  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
38463         * config/aarch64/aarch64.c (has_memory_op): Delete.
38464         (aarch64_madd_needs_nop): Use contains_mem_rtx_p instead of
38465         has_memory_op.
38467 2017-02-13  Jakub Jelinek  <jakub@redhat.com>
38469         PR rtl-optimization/79388
38470         PR rtl-optimization/79450
38471         * combine.c (distribute_notes): When removing TEM_INSN for which
38472         corresponding dest has last value recorded, invalidate that last
38473         value.
38475 2017-02-13  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
38477         * config/arm/arm.c (arm_print_tune_info): Use ASM_COMMENT_START instead
38478         of explicit '@'.  Add missing assembly comment marker on branch costs
38479         printout.
38481 2017-02-13  Nathan Sidwell  <nathan@acm.org>
38483         * gengtype-lex.l (<in_struct>): Add '/'.
38485 2017-02-13  Martin Liska  <mliska@suse.cz>
38487         PR c/79471
38488         * calls.c (expand_call): Replace XALLOCAVEC with XCNEWVEC.
38490 2017-02-13  Richard Biener  <rguenther@suse.de>
38492         * configure.ac (HAVE_ISL_OPTIONS_SET_SCHEDULE_SERIALIZE_SCCS):
38493         Remove.
38494         * configure: Re-generate.
38495         * config.in: Likewise.
38496         * graphite-dependences.c: Simplify as if
38497         HAVE_ISL_OPTIONS_SET_SCHEDULE_SERIALIZE_SCCS was defined.
38498         * graphite-isl-ast-to-gimple.c: Likewise.
38499         * graphite-optimize-isl.c: Likewise.
38500         * graphite-poly.c: Likewise.
38501         * graphite-sese-to-poly.c: Likewise.
38502         * graphite.h: Likewise.
38503         * toplev.c: Include isl/version.h and use isl_version () for
38504         printing the ISL version.
38505         * doc/install.texi: Update ISL requirement.
38507 2017-02-12  Gerald Pfeifer  <gerald@pfeifer.com>
38509         * doc/standards.texi (Standards): Update reference to
38510         Objective-C 2.0.
38512 2017-02-12  Gerald Pfeifer  <gerald@pfeifer.com>
38514         * doc/extend.texi (Named Address Spaces): sourceware.org now
38515         defaults to https.
38516         * doc/install.texi (Binaries): Ditto.
38517         (Specific): Ditto.
38519 2017-02-11  Sandra Loosemore  <sandra@codesourcery.com>
38521         * doc/cpp.texi: Replace "stringify"/"stringification" with C
38522         standard terminology "stringize"/"stringizing" throughout.
38523         * doc/cppinternals.texi: Likewise.
38525 2017-02-11  Sandra Loosemore  <sandra@codesourcery.com>
38527         * doc/extend.texi: Fix some spelling mistakes and typos.
38528         * doc/invoke.texi: Likewise.
38530 2017-02-11  Jan Hubicka  <hubicka@ucw.cz>
38532         PR ipa/79224
38533         * params.def (inline-min-speedup) Change from 10 to 8.
38535 2017-02-11  Jakub Jelinek  <jakub@redhat.com>
38537         * doc/invoke.texi (fopenmp): Bump OpenMP version from 4.0 to
38538         4.5.
38540 2017-02-11  Jan Hubicka  <hubicka@ucw.cz>
38542         PR ipa/79224
38543         * ipa-inline-analysis.c (get_minimal_bb): New function.
38544         (record_modified): Use it.
38545         (remap_edge_change_prob): Handle also ancestor functions.
38547 2017-02-11  Gerald Pfeifer  <gerald@pfeifer.com>
38549         * doc/contrib.texi (Contributors): Remove broken link into
38550         the Mauve CVS repository.
38552 2017-02-11  Jakub Jelinek  <jakub@redhat.com>
38554         PR middle-end/79454
38555         * internal-fn.c (expand_vector_ubsan_overflow): Use piece-wise
38556         result computation whenever lhs doesn't have vector mode, not
38557         just when it has BLKmode.
38559 2017-02-10  Gerald Pfeifer  <gerald@pfeifer.com>
38561         * doc/makefile.texi (profiledbootstrap): Refer to the
38562         installation instructions only in textual form.
38564 2017-02-10  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
38566         PR target/79295
38567         * config/rs6000/altivec.md (bcd<bcd_add_sub>): Fix constraints.
38569 2017-02-10  Gerald Pfeifer  <gerald@pfeifer.com>
38571         * doc/install.texi (Specific): Use https for blackfin.uclinux.org.
38572         (Specific): Update mingw-w64 reference.
38573         (Binaries): Ditto.
38574         (Specific): Remove broken link to Renesas RX processor.
38576 2017-02-10  Richard Biener  <rguenther@suse.de>
38578         * toplev.c (process_options): Do not mention obsolete graphite
38579         options when printing sorry message about missing graphite support.
38580         Mention -floop-nest-optimize.
38582 2017-02-10  Christophe Lyon  <christophe.lyon@linaro.org>
38584         * config/aarch64/arm_neon.h (vtst_p8): Rewrite without asm.
38585         (vtst_p16): Likewise.
38586         (vtstq_p8): Likewise.
38587         (vtstq_p16): Likewise.
38588         (vtst_p64): New.
38589         (vtstq_p64): Likewise.
38590         * config/arm/arm_neon.h (vgetq_lane_p64): New.
38591         (vset_lane_p64): New.
38592         (vsetq_lane_p64): New.
38594 2017-02-10  Jakub Jelinek  <jakub@redhat.com>
38596         PR tree-optimization/79411
38597         * tree-ssa-reassoc.c (is_reassociable_op): Return false if
38598         stmt operands are SSA_NAMEs used in abnormal phis.
38599         (can_reassociate_p): Return false if op is SSA_NAME used in abnormal
38600         phis.
38602 2017-02-09  Jan Hubicka  <hubicka@ucw.cz>
38604         PR ipa/70795
38605         * cgraphunit.c (cgraph_node::add_new_function): Set externally_visible
38606         flag if needed.
38608 2017-02-09  Jan Hubicka  <hubicka@ucw.cz>
38610         * tree-ssa-loop-unswitch.c (hoist_guard): Update profile.
38612 2017-02-09  Jakub Jelinek  <jakub@redhat.com>
38614         * omp-offload.c (oacc_loop_auto_partitions): Use || instead of |
38615         to avoid warning.
38617         PR c/79413
38618         * gimplify.h (is_gimple_sizepos): Only test for INTEGER_CST constants,
38619         not arbitrary TREE_CONSTANT.
38621         PR c/79431
38622         * gimplify.c (gimplify_adjust_omp_clauses): Ignore
38623         "omp declare target link" attribute unless is_global_var.
38624         * omp-offload.c (find_link_var_op): Likewise.
38626 2017-02-09  Nathan Sidwell  <nathan@codesourcery.com>
38627             Chung-Lin Tang  <cltang@codesourcery.com>
38629         * gimplify.c (gimplify_scan_omp_clauses): No special handling for
38630         OMP_CLAUSE_TILE.
38631         (gimplify_adjust_omp_clauses): Don't delete TILE.
38632         (gimplify_omp_for): Deal with TILE.
38633         * internal-fn.c (expand_GOACC_TILE): New function.
38634         * internal-fn.def (GOACC_DIM_POS): Comment may be overly conservative.
38635         (GOACC_TILE): New.
38636         * omp-expand.c (struct oacc_collapse): Add tile and outer fields.
38637         (expand_oacc_collapse_init): Add LOC paramter.  Initialize tile
38638         element fields.
38639         (expand_oacc_collapse_vars): Add INNER parm, adjust for tiling,
38640         avoid DIV for outermost collapse var.
38641         (expand_oacc_for): Insert tile element loop as needed.  Adjust.
38642         Remove out of date comments, fix whitespace.
38643         * omp-general.c (omp_extract_for_data): Deal with tiling.
38644         * omp-general.h (enum oacc_loop_flags): Add OLF_TILE flag,
38645         adjust OLF_DIM_BASE value.
38646         (struct omp_for_data): Add tiling field.
38647         * omp-low.c (scan_sharing_clauses): Allow OMP_CLAUSE_TILE.
38648         (lower_oacc_head_mark): Add OLF_TILE as appropriate.  Ensure 2 levels
38649         for auto loops.  Remove default auto determining, moved to
38650         oacc_loop_fixed_partitions.
38651         * omp-offload.c (struct oacc_loop): Change 'ifns' to vector of call
38652         stmts, add e_mask field.
38653         (oacc_dim_call): New function, abstracted out from oacc_thread_numbers.
38654         (oacc_thread_numbers): Use oacc_dim_call.
38655         (oacc_xform_tile): New.
38656         (new_oacc_loop_raw): Initialize e_mask, adjust for ifns vector.
38657         (finish_oacc_loop): Adjust for ifns vector.
38658         (oacc_loop_discover_walk): Append loop abstraction sites to list,
38659         add case for GOACC_TILE fns.
38660         (oacc_loop_xform_loop): Delete.
38661         (oacc_loop_process): Iterate over call list directly, and add
38662         handling for GOACC_TILE fns.
38663         (oacc_loop_fixed_partitions): Determine default auto, deal with TILE,
38664         dump partitioning.
38665         (oacc_loop_auto_partitions): Add outer_assign parm. Assign all but
38666         vector partitioning to outer loops.  Assign 2 partitions to loops
38667         when available. Add TILE handling.
38668         (oacc_loop_partition): Adjust oacc_loop_auto_partitions call.
38669         (execite_oacc_device_lower): Process GOACC_TILE fns,
38670         ignore unknown specs.
38671         * tree-nested.c (convert_nonlocal_omp_clauses): Allow OMP_CLAUSE_TILE.
38672         * tree.c (omp_clause_num_ops): Adjust TILE ops.
38673         * tree.h (OMP_CLAUSE_TILE_ITERVAR, OMP_CLAUSE_TILE_COUNT): New.
38675 2017-02-09  Gerald Pfeifer  <gerald@pfeifer.com>
38677         * configure.ac (ACX_BUGURL): Update.
38678         * configure: Regenerate.
38680 2017-02-09  Richard Biener  <rguenther@suse.de>
38682         PR tree-optimization/69823
38683         * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
38684         Properly enumerate all BBs in the region.  Use auto_vec/auto_bitmap.
38686 2017-02-09  Andrew Burgess  <andrew.burgess@embecosm.com>
38688         * config/arc/arc-c.def: Add __NPS400__ definition.
38689         * config/arc/arc.h (CPP_SPEC): Don't define __NPS400__ here.
38690         (TARGET_NPS400): Define.
38692 2017-02-09  Andrew Burgess  <andrew.burgess@embecosm.com>
38694         * config/arc/arc-arch.h (arc_arch_t): Move unchanged to earlier in
38695         file.
38696         (arc_cpu_t): Change base_architecture field, arch, to a arc_arc_t
38697         pointer, arch_info.
38698         (arc_cpu_types): Fill the arch_info field with a pointer into the
38699         arc_arch_types table.
38700         (arc_selected_cpu): Declare.
38701         * config/arc/arc.c (arc_selected_cpu): Make global.
38702         (arc_selected_arch): Delete.
38703         (arc_base_cpu): Delete.
38704         (arc_override_options): Remove references to deleted variables,
38705         update access to arch information.
38706         (ARC_OPT): Update access to arch information.
38707         (ARC_OPTX): Likewise.
38708         * config/arc/arc.h (arc_base_cpu): Remove declaration.
38709         (TARGET_ARC600): Update access to arch information.
38710         (TARGET_ARC601): Likewise.
38711         (TARGET_ARC700): Likewise.
38712         (TARGET_EM): Likewise.
38713         (TARGET_HS): Likewise.
38714         * config/arc/driver-arc.c (arc_cpu_to_as): Update access to arch
38715         information.
38717 2017-02-08  Pat Haugen  <pthaugen@us.ibm.com>
38719         PR target/78604
38720         * config/rs6000/rs6000.c (rs6000_emit_vector_cond_expr): Invert
38721         condition/operands for integer GE/LE/GEU/LEU operations.
38723 2017-02-08  Segher Boessenkool  <segher@kernel.crashing.org>
38725         PR translation/79397
38726         * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Fix spelling
38727         of AltiVec.
38729 2017-02-08  Martin Jambor  <mjambor@suse.cz>
38731         PR ipa/79375
38732         * ipa-prop.c (ipa_alloc_node_params): Make static, return bool
38733         whether allocation happened.
38734         (ipa_initialize_node_params): Do not call ipa_alloc_node_params if
38735         nothing was allocated.
38737 2017-02-08  Jakub Jelinek  <jakub@redhat.com>
38739         PR tree-optimization/79408
38740         * tree-vrp.c (simplify_div_or_mod_using_ranges): If op1 is not
38741         constant, but SSA_NAME with a known integer range, use the minimum
38742         of that range instead of op1 to determine if modulo can be replaced
38743         with its first operand.
38745 2016-02-08  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
38747         * config/riscv/riscv.c (riscv_build_integer_1): Avoid use of INT16_MAX.
38749 2017-02-08  Richard Biener  <rguenther@suse.de>
38751         PR tree-optimization/71824
38752         * graphite-scop-detection.c (scop_detection::build_scop_breadth):
38753         Check all loops contained in the merged region.
38755 2017-02-07  Andrew Pinski  <apinski@cavium.com>
38757         * config/aarch64/aarch64.md (popcount<mode>2): New pattern.
38759 2017-02-07  Andrew Pinski  <apinski@cavium.com>
38761         * config/aarch64/aarch64-cores.def (thunderx): Disable LSE.
38762         (thunderxt88): Likewise.
38763         (thunderxt81): Disable LSE and change v8.1 to v8.
38764         (thunderxt83): Likewise.
38766 2017-02-07  Jakub Jelinek  <jakub@redhat.com>
38767             Richard Biener  <rguenther@suse.de>
38769         PR middle-end/79399
38770         * ira-int.h (struct target_ira_int): Change x_max_struct_costs_size
38771         type from int to size_t.
38772         * ira-costs.c (struct_costs_size): Change type from int to size_t.
38774 2017-02-07  Jakub Jelinek  <jakub@redhat.com>
38776         PR rtl-optimization/79386
38777         * cprop.c (bypass_conditional_jumps): Initialize
38778         bypass_last_basic_block already before splitting bbs after
38779         unconditional traps...
38780         (bypass_conditional_jumps): ... rather than here.
38782         PR target/79299
38783         * config/i386/sse.md (xtg_mode, gatherq_mode): New mode attrs.
38784         (*avx512f_gathersi<mode>, *avx512f_gathersi<mode>_2,
38785         *avx512f_gatherdi<mode>, *avx512f_gatherdi<mode>_2): Use them,
38786         fix -masm=intel patterns.
38788 2017-02-07  Richard Biener  <rguenther@suse.de>
38790         PR tree-optimization/79256
38791         PR middle-end/79278
38792         * builtins.c (get_object_alignment_2): Use min_align_of_type
38793         to extract alignment for MEM_REFs to honor BIGGEST_FIELD_ALIGNMENT
38794         and ADJUST_FIELD_ALIGN.
38796         * doc/tm.texi.in (ADJUST_FIELD_ALIGN): Adjust to take additional
38797         type parameter.
38798         * doc/tm.texi: Regenerate.
38799         * stor-layout.c (layout_decl): Adjust.
38800         (update_alignment_for_field): Likewise.
38801         (place_field): Likewise.
38802         (min_align_of_type): Likewise.
38803         * config/arc/arc.h (ADJUST_FIELD_ALIGN): Adjust.
38804         * config/epiphany/epiphany.h (ADJUST_FIELD_ALIGN): Likewise.
38805         * config/epiphany/epiphany.c (epiphany_adjust_field_align): Likewise.
38806         * config/frv/frv.h (ADJUST_FIELD_ALIGN): Likewise.
38807         * config/frv/frv.c (frv_adjust_field_align): Likewise.
38808         * config/i386/i386.h (ADJUST_FIELD_ALIGN): Likewise.
38809         * config/i386/i386.c (x86_field_alignment): Likewise.
38810         * config/rs6000/aix.h (ADJUST_FIELD_ALIGN): Likewise.
38811         * config/rs6000/darwin.h (ADJUST_FIELD_ALIGN): Likewise.
38812         * config/rs6000/freebsd64.h (ADJUST_FIELD_ALIGN): Likewise.
38813         * config/rs6000/linux64.h (ADJUST_FIELD_ALIGN): Likewise.
38814         * config/rs6000/sysv4.h (ADJUST_FIELD_ALIGN): Likewise.
38815         * config/rs6000/rs6000.c (rs6000_special_adjust_field_align_p):
38816          Likewise.
38818         Revert
38819         2017-01-30  Richard Biener  <rguenther@suse.de>
38821         PR tree-optimization/79256
38822         * targhooks.c (default_builtin_vector_alignment_reachable): Honor
38823         BIGGEST_FIELD_ALIGNMENT and ADJUST_FIELD_ALIGN to fix up bogus
38824         alignment on TYPE.
38826 2017-02-07  Toma Tabacu  <toma.tabacu@imgtec.com>
38828         * config/mips/mips.c (mips_expand_builtin_insn): Convert the QImode
38829         argument of the pshufh, psllh, psllw, psrah, psraw, psrlh, psrlw
38830         builtins to SImode and emit a zero-extend, if necessary.
38832 2017-02-06  Palmer Dabbelt  <palmer@dabbelt.com>
38834         * docs/invoke.texi (RISC-V Options): Alphabetize.
38836 2017-02-06  Palmer Dabbelt  <palmer@dabbelt.com>
38838         * doc/invoke.texi (RISC-V Options): Use two spaces to separate
38839         options.
38841 2017-02-06  Palmer Dabbelt  <palmer@dabbelt.com>
38843         * config/riscv/riscv.c: New file.
38844         * common/config/riscv/riscv-common.c: Likewise.
38845         * config.gcc: Likewise.
38846         * config/riscv/constraints.md: Likewise.
38847         * config/riscv/elf.h: Likewise.
38848         * config/riscv/generic.md: Likewise.
38849         * config/riscv/linux.h: Likewise.
38850         * config/riscv/multilib-generator: Likewise.
38851         * config/riscv/peephole.md: Likewise.
38852         * config/riscv/pic.md: Likewise.
38853         * config/riscv/predicates.md: Likewise.
38854         * config/riscv/riscv-builtins.c: Likewise.
38855         * config/riscv/riscv-c.c: Likewise.
38856         * config/riscv/riscv-ftypes.def: Likewise.
38857         * config/riscv/riscv-modes.def: Likewise.
38858         * config/riscv/riscv-opts.h: Likewise.
38859         * config/riscv/riscv-protos.h: Likewise.
38860         * config/riscv/riscv.h: Likewise.
38861         * config/riscv/riscv.md: Likewise.
38862         * config/riscv/riscv.opt: Likewise.
38863         * config/riscv/sync.md: Likewise.
38864         * config/riscv/t-elf-multilib: Likewise.
38865         * config/riscv/t-linux: Likewise.
38866         * config/riscv/t-linux-multilib: Likewise.
38867         * config/riscv/t-riscv: Likewise.
38868         * configure.ac: Likewise.
38869         * doc/contrib.texi: Add Kito Cheng, Palmer Dabbelt, and Andrew
38870         Waterman as RISC-V maintainers.
38871         * doc/install.texi: Add RISC-V entries.
38872         * doc/invoke.texi: Add RISC-V options section.
38873         * doc/md.texi: Add RISC-V constraints section.
38874         * configure: Regenerated.
38876 2017-02-06  Michael Meissner  <meissner@linux.vnet.ibm.com>
38878         PR target/66144
38879         * config/rs6000/vector.md (vcond<mode><mode>): Allow the true and
38880         false values to be constant vectors with all 0 or all 1 bits set.
38881         (vcondu<mode><mode>): Likewise.
38882         * config/rs6000/predicates.md (vector_int_reg_or_same_bit): New
38883         predicate.
38884         (fpmask_comparison_operator): Update comment.
38885         (vecint_comparison_operator): New predicate.
38886         * config/rs6000/rs6000.c (rs6000_emit_vector_cond_expr): Optimize
38887         vector conditionals when the true and false values are constant
38888         vectors with all 0 bits or all 1 bits set.
38890 2017-02-06  Martin Sebor  <msebor@redhat.com>
38892         PR  tree-optimization/79376
38893         * gimple-fold.c (get_range_strlen): Set the minimum length to zero.
38895 2017-02-06  Uros Bizjak  <ubizjak@gmail.com>
38897         * config/i386/sse.md (vector modes -> vec_extract* splitter): Use
38898         explicit subreg RTX with operand 1.  Use VECTOR_MODE_P predicate
38899         to simplify split condition.
38901 2017-02-06  Jakub Jelinek  <jakub@redhat.com>
38903         * omp-expand.c (oxpand_omp_atomic_fetch_op,
38904         expand_omp_atomic_pipeline): Return false if can_atomic_load_p is
38905         false.
38907 2017-02-06  Segher Boessenkool  <segher@kernel.crashing.org>
38909         PR rtl-optimization/68664
38910         * target.def (can_speculate_insn): New hook.
38911         * doc/tm.texi.in (TARGET_SCHED_CAN_SPECULATE_INSN): New hook.
38912         * doc/tm.texi: Regenerate.
38913         * sched-rgn.c (can_schedule_ready_p): Use the new hook.
38914         * config/rs6000/rs6000.c (TARGET_SCHED_CAN_SPECULATE_INSN): New macro.
38915         (rs6000_sched_can_speculate_insn): New function.
38917 2017-02-06  Jakub Jelinek  <jakub@redhat.com>
38919         PR tree-optimization/79284
38920         * tree-vectorizer.h (VECT_SCALAR_BOOLEAN_TYPE_P): Define.
38921         * tree-vect-stmts.c (vect_get_vec_def_for_operand,
38922         vectorizable_mask_load_store, vectorizable_operation,
38923         vect_is_simple_cond, get_same_sized_vectype): Use it instead
38924         of comparing TREE_CODE of a type against BOOLEAN_TYPE.
38925         * tree-vect-patterns.c (check_bool_pattern, search_type_for_mask_1,
38926         vect_recog_bool_pattern, vect_recog_mask_conversion_pattern): Likewise.
38927         * tree-vect-slp.c (vect_get_constant_vectors): Likewise.
38928         * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
38929         Remove redundant gimple_code (stmt) == GIMPLE_ASSIGN test after
38930         is_gimple_assign (stmt).  Replace another such test with
38931         is_gimple_assign (stmt).
38933 2017-02-06  Georg-Johann Lay  <avr@gjlay.de>
38935         PR target/78883
38936         * config/avr/avr.c (rtl-iter.h): Include it.
38937         (TARGET_LEGITIMATE_COMBINED_INSN): New hook define...
38938         (avr_legitimate_combined_insn): ...and implementation.
38940 2017-02-06  Dominik Vogt  <vogt@linux.vnet.ibm.com>
38942         * config/s390/predicates.md ("larl_operand"): Use macros from hwint.h.
38943         * config/s390/s390.c (s390_const_operand_ok)
38944         (s390_canonicalize_comparison, s390_extract_part)
38945         (s390_single_part, s390_contiguous_bitmask_nowrap_p)
38946         (s390_contiguous_bitmask_p, s390_rtx_costs)
38947         (legitimize_pic_address): Likewise.
38948         * config/s390/s390.md ("clzdi2", "clztidi2"): Likewise.
38949         * config/s390/vx-builtins.md ("vec_genbytemaskv16qi")
38950         ("vec_permi<mode>", "vfae<mode>", "*vfaes<mode>", "vstrc<mode>")
38951         ("*vstrcs<mode>"): Use UINTVAL() to set unsigned HOST_WIDE_INT.
38952         * config/s390/vector.md ("vec_vfenes<mode>"): Likewise.
38954 2017-02-06  Georg-Johann Lay  <avr@gjlay.de>
38956         * config/avr/avr.md (*addhi3_zero_extend): Add alternative where
38957         REGNO($0) == REGNO($1).
38959 2017-02-06  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
38961         * config/s390/linux.h(SIZE_TYPE): Add comment.
38963 2017-02-06  Julian Brown  <julian@codesourcery.com>
38964             Naveen H.S  <Naveen.Hurugalawadi@cavium.com>
38965             Virendra Pathak  <virendra.pathak@broadcom.com>
38967         * config/aarch64/aarch64-cores.def: Change the scheduler
38968         to Thunderx2t99.
38969         * config/aarch64/aarch64.md: Include thunderx2t99.md.
38970         * config/aarch64/thunderx2t99.md: New file.
38972 2017-02-05  Gerald Pfeifer  <gerald@pfeifer.com>
38974         * doc/standards.texi (Go Language): Update link to language
38975         standard.
38977 2017-02-05  Jan Hubicka  <hubicka@ucw.cz>
38979         * tree-eh.c (lower_resx): Sanitize profile.
38980         (cleanup_empty_eh_move_lp): Likewise.
38982 2017-02-05  Jan Hubicka  <hubicka@ucw.cz>
38984         PR tree-ssa/79347
38985         * cfgloopmanip.c (lv_adjust_loop_entry_edge, loop_version): Add
38986         ELSE_PROB.
38987         * cfgloopmanip.h (loop_version): Update prototype.
38988         * modulo-sched.c (sms_schedule): Update call of loop_version.
38989         * tree-if-conv.c(version_loop_for_if_conversion): Likewise.
38990         * tree-parloops.c (gen_parallel_loop): Likewise.
38991         * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Likewise.
38992         * tree-ssa-loop-split.c (split_loop): Likewise.
38993         * tree-ssa-loop-unswitch.c (tree_unswitch_loop): Likewise.
38994         * tree-vect-loop-manip.c (vect_loop_versioning): Likewise.
38996 2017-02-05  Martin Liska  <mliska@suse.cz>
38998         PR bootstrap/78985
38999         * config/s390/s390.c (s390_gimplify_va_arg): Initialize local
39000         variable to NULL.
39001         (print_operand_address): Initialize a struct to zero.
39003 2017-02-05  Gerald Pfeifer  <gerald@pfeifer.com>
39005         * doc/contrib.texi (Contributors): Refer to Hans Boehm's
39006         garbage collector only in textual form.
39008 2017-02-05  Gerald Pfeifer  <gerald@pfeifer.com>
39010         * doc/extend.texi (x86 specific memory model extensions for
39011         transactional memory): Simplify a phrase.
39013 2017-02-05  Eric Botcazou  <ebotcazou@adacore.com>
39015         PR target/79353
39016         * config/sparc/sync.md (atomic_loaddi_1): Replace 'U' constraint with
39017         'r', 'm' constraint with 'T' and !TARGET_ARCH64 with TARGET_ARCH32.
39018         (atomic_storedi_1): Likewise.
39020 2017-02-04  Jakub Jelinek  <jakub@redhat.com>
39022         PR tree-optimization/79338
39023         * tree-parloops.c (gather_scalar_reductions): Don't call
39024         vect_analyze_loop_form for loop->inner before destroying loop's
39025         loop_vinfo.
39027 2017-02-03  Martin Sebor  <msebor@redhat.com>
39029         PR tree-optimization/79327
39030         * gimple-ssa-sprintf.c (tree_digits): Avoid adding the base prefix
39031         when precision has resulted in leading zeros.
39032         (format_integer): Adjust the likely counter to assume an unknown
39033         argument that may be zero is non-zero.
39035 2017-02-03  Jason Merrill  <jason@redhat.com>
39037         PR c++/78689
39038         * tree-inline.c (copy_tree_body_r) [COND_EXPR]: Revert change to
39039         avoid copying non-taken branch.
39041 2017-02-03  Jakub Jelinek  <jakub@redhat.com>
39043         PR tree-optimization/79340
39044         * tree-vect-loop.c (vectorizable_reduction): Release
39045         vec_defs elements after safe_splicing them into other vectors.
39046         Formatting fixes.
39048         PR tree-optimization/79327
39049         * gimple-ssa-sprintf.c (adjust_range_for_overflow): If returning
39050         true, always set *argmin and *argmax to TYPE_{MIN,MAX}_VALUE of
39051         dirtype.
39052         (format_integer): Use wide_int_to_tree instead of build_int_cst
39053         + to_?hwi.  If argmin is NULL, just set argmin and argmax to
39054         TYPE_{MIN,MAX}_VALUE of argtype.  Simplify and fix computation
39055         of shortest and longest sequence.
39057 2017-02-03  Uros Bizjak  <ubizjak@gmail.com>
39059         * config/i386/i386.c (dimode_scalar_chain::convert_reg):
39060         Use pextrd for TARGET_SSE4_1 when creating scalar copy.
39062 2017-02-03  Walter Lee  <walt@tilera.com>
39064         PR target/78862
39065         * config/tilegx/tilegx.md (tilegx_expand_prologue): Add blockage
39066         after initial stackframe link reg save.
39067         * config/tilepro/tilepro.md (tilepro_expand_prologue): Likewise.
39069 2017-02-03  Jakub Jelinek  <jakub@redhat.com>
39071         PR target/79354
39072         * config/rs6000/rs6000.md (movsi_from_sf): Use wb constraint instead of
39073         wu for stxssp alternative.
39075 2017-02-03  Martin Sebor  <msebor@redhat.com>
39077         PR tree-optimization/79352
39078         * gimple-fold.c (get_range_strlen): Add argument.
39079         (get_range_strlen): Change return type to bool.
39080         (get_maxval_strlen): Pass in a dummy argument.
39081         * gimple-fold.h (get_range_strlen): Change return type to bool.
39082         * gimple-ssa-sprintf.c (get_string_length): Set unlikely counter.
39083         * tree.h (array_at_struct_end_p): Add argument.
39084         * tree.c (array_at_struct_end_p): Handle it.
39086 2017-02-03  Martin Liska  <mliska@suse.cz>
39088         PR lto/66295
39089         * multiple_target.c (create_dispatcher_calls): Redirect edge
39090         from a caller of a dispatcher.
39091         (expand_target_clones): Make the clones local.
39092         (ipa_target_clone): Do both target clones and resolvers.
39093         (ipa_dispatcher_calls): Remove the pass.
39094         (pass_dispatcher_calls::gate): Likewise.
39095         (make_pass_dispatcher_calls): Likewise.
39096         * passes.def (pass_target_clone): Put as very first IPA early
39097         pass.
39099 2017-02-03  Martin Liska  <mliska@suse.cz>
39101         * symtab.c (symtab_node::binds_to_current_def_p): Bail out
39102         in case of a function with ifunc attribute.
39104 2017-02-03  Martin Liska  <mliska@suse.cz>
39106         * cgraph.c (cgraph_node::dump): Dump function version info.
39107         * symtab.c (symtab_node::dump_base): Add missing new line.
39109 2017-02-02  Jan Hubicka  <hubicka@ucw.cz>
39111         * tree-ssa-ifcombine.c (update_profile_after_ifcombine): New function.
39112         (ifcombine_ifandif): Use it.
39114 2017-02-03  Martin Liska  <mliska@suse.cz>
39116         * doc/invoke.texi: Document default value for
39117         use-after-scope-direct-emission-threshold.
39119 2017-02-03  Martin Liska  <mliska@suse.cz>
39121         PR tree-optimization/79339
39122         * gimple-ssa-sprintf.c (format_floating_max): Call mpfr_clear.
39123         (format_floating): Likewise.
39125 2017-02-03  Martin Liska  <mliska@suse.cz>
39127         PR ipa/79337
39128         * ipa-prop.c (ipa_node_params_t::insert): Remove current
39129         implementation.
39130         (ipa_node_params_t::remove): Likewise.
39131         * ipa-prop.h (ipa_node_params::ipa_node_params): Make default
39132         initialization from removed ipa_node_params_t::insert.
39133         (ipa_node_params::~ipa_node_params): Move from removed
39134         ipa_node_params_t::release.
39135         * symbol-summary.h (symbol_summary::m_released): New member.
39136         Do not release a summary twice.  Do not allow to call finalizer
39137         for types of a summary that live in GGC memory.
39139 2017-02-02  Naveen H.S  <Naveen.Hurugalawadi@cavium.com>
39141         * config/aarch64/aarch64.c (thunderx2t99_tunings): Enable AES and
39142         cmp_branch fusion.
39144 2017-02-02  Martin Sebor  <msebor@redhat.com>
39146         PR middle-end/79275
39147         * gimple-ssa-sprintf.c (get_string_length): Set lower bound to zero.
39148         (format_string): Tighten up the range of output for non-constant
39149         strings and correct the expected range for wide non-constant strings.
39151 2017-02-02  Martin Sebor  <msebor@redhat.com>
39153         * doc/invoke.texi (-maccumulate-args): Fix bad grammar.
39155         PR middle-end/32003
39156         * doc/invoke.texi (-fdump-tree-): Remove pass-specific options from
39157         index.
39158         (-fdump-tree-@var): Add to index and document how to come up
39159         with pass-specific option and dump file names.
39160         (-fdump-passes): Clarify where to look for output.
39162 2017-02-02  Jan Hubicka  <hubicka@ucw.cz>
39164         PR middle-end/77445
39165         * tree-ssa-threadbackward.c (profitable_jump_thread_path): Dump
39166         statistics of the analyzed path; allow threading for speed when
39167         any of BBs along the path are optimized for speed.
39169 2017-02-02  Eric Botcazou  <ebotcazou@adacore.com>
39171         PR middle-end/78468
39172         * emit-rtl.c (init_emit): Add ??? comment for problematic alignment
39173         settings of the virtual registers.
39175         Revert again
39176         2016-08-23  Dominik Vogt  <vogt@linux.vnet.ibm.com>
39178         * explow.c (get_dynamic_stack_size): Take known alignment of stack
39179         pointer + STACK_DYNAMIC_OFFSET into account when calculating the size
39180         needed.
39182 2017-02-02  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
39184         * config/s390/vx-builtins.md ("vec_ceil", "vec_floor")
39185         ("vec_trunc", "vec_roundc", "vec_round"): Remove expanders.
39187 2017-02-02  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
39189         * config/s390/s390.md: Add missing comments with the expanded
39190         mnemonics.
39191         * config/s390/vector.md: Likewise.
39192         * config/s390/vx-builtins.md: Likewise.
39194 2017-02-02  Jakub Jelinek  <jakub@redhat.com>
39196         PR target/79197
39197         * config/rs6000/rs6000.md (*fixuns_trunc<mode>di2_fctiduz): Rename to ...
39198         (fixuns_trunc<mode>di2): ... this, remove previous expander.  Put all
39199         conditions on a single line.
39201 2017-02-02  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
39203         * config/s390/s390-c.c (s390_cpu_cpp_builtins_internal): Rename
39204         __S390_VX__ to __VX__.
39206 2017-02-01  Andrew Pinski  <apinski@cavium.com>
39208         * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost): Pass
39209         stmt_info to record_stmt_cost.
39210         (vect_get_known_peeling_cost): Pass stmt_info if known to
39211         record_stmt_cost.
39212         * config/aarch64/aarch64-protos.h (cpu_vector_cost): Split
39213         cpu_vector_cost field into
39214         scalar_int_stmt_cost and scalar_fp_stmt_cost.  Split vec_stmt_cost
39215         field into vec_int_stmt_cost and vec_fp_stmt_cost.
39216         * config/aarch64/aarch64.c (generic_vector_cost): Update for the
39217         splitting of scalar_stmt_cost and vec_stmt_cost.
39218         (thunderx_vector_cost): Likewise.
39219         (cortexa57_vector_cost): LIkewise.
39220         (exynosm1_vector_cost): Likewise.
39221         (xgene1_vector_cost): Likewise.
39222         (thunderx2t99_vector_cost): Improve after the splitting of the two
39223         fields.
39224         (aarch64_builtin_vectorization_cost): Update for the splitting of
39225         scalar_stmt_cost and vec_stmt_cost.
39227 2017-02-01  Torvald Riegel  <triegel@redhat.com>
39228             Richard Henderson  <rth@redhat.com>
39230         * builtins.c (fold_builtin_atomic_always_lock_free): Make "lock-free"
39231         conditional on existance of a fast atomic load.
39232         * optabs-query.c (can_atomic_load_p): New function.
39233         * optabs-query.h (can_atomic_load_p): Declare it.
39234         * optabs.c (expand_atomic_exchange): Always delegate to libatomic if
39235         no fast atomic load is available for the particular size of access.
39236         (expand_atomic_compare_and_swap): Likewise.
39237         (expand_atomic_load): Likewise.
39238         (expand_atomic_store): Likewise.
39239         (expand_atomic_fetch_op): Likewise.
39240         * testsuite/lib/target-supports.exp
39241         (check_effective_target_sync_int_128): Remove x86 because it provides
39242         no fast atomic load.
39243         (check_effective_target_sync_int_128_runtime): Likewise.
39245 2017-02-01  Richard Biener  <rguenther@suse.de>
39247         * graphite.c: Include tree-vectorizer.h for find_loop_location.
39248         (graphite_transform_loops): Provide opt-info for optimized nests.
39249         * tree-parloop.c (parallelize_loops): Provide opt-info for
39250         parallelized loops.
39252 2017-02-01  Richard Biener  <rguenther@suse.de>
39254         PR middle-end/79315
39255         * tree-cfg.c (move_stmt_op): Never set TREE_BLOCK when it
39256         was not set before.
39258 2017-02-01  Richard Biener  <rguenther@suse.de>
39260         PR tree-optimization/71824
39261         * graphite-scop-detection.c (scop_detection::build_scop_breadth):
39262         Verify the loops are valid in the merged SESE region.
39263         (scop_detection::can_represent_loop_1): Check analyzing the
39264         evolution of the number of iterations in the region succeeds.
39266 2017-01-31  Ian Lance Taylor  <iant@golang.org>
39268         * config/i386/i386.c (ix86_expand_split_stack_prologue): Add
39269         REG_ARGS_SIZE note to 32-bit push insns and call insn.
39271 2017-01-31  David Malcolm  <dmalcolm@redhat.com>
39273         PR preprocessor/79210
39274         * input.c (get_substring_ranges_for_loc): Replace line_width
39275         assertion with error-handling.
39277 2017-01-31  Richard Biener  <rguenther@suse.de>
39279         PR tree-optimization/77318
39280         * graphite-sese-to-poly.c (extract_affine): Fix assert.
39281         (create_pw_aff_from_tree): Take loop parameter.
39282         (add_condition_to_pbb): Pass loop of the condition to
39283         create_pw_aff_from_tree.
39285 2017-01-31  Jakub Jelinek  <jakub@redhat.com>
39287         * config/s390/s390.c (s390_asan_shadow_offset): New function.
39288         (TARGET_ASAN_SHADOW_OFFSET): Redefine.
39290 2017-01-31  Michael Meissner  <meissner@linux.vnet.ibm.com>
39292         PR target/78597
39293         PR target/79038
39294         * config/rs6000/rs6000-protos.h (convert_float128_to_int): Delete,
39295         no longer used.
39296         (convert_int_to_float128): Likewise.
39297         * config/rs6000/rs6000.c (convert_float128_to_int): Likewise.
39298         (convert_int_to_float128): Likewise.
39299         * config/rs6000/rs6000.md (UNSPEC_IEEE128_MOVE): Likewise.
39300         (UNSPEC_IEEE128_CONVERT): Likewise.
39301         (floatsi<mode>2, FLOAT128 iterator): Bypass calling
39302         rs6000_expand_float128_convert if we have IEEE 128-bit hardware.
39303         Use local variables for IBM extended format.
39304         (fix_trunc<mode>si2, FLOAT128 iterator): Likewise.
39305         (fix_trunc<mode>si2_fprs): Likewise.
39306         (fixuns_trunc<IEEE128:mode><SDI:mode>2): Likewise.
39307         (floatuns<IEEE128:mode>2, IEEE128 iterator): Likewise.
39308         (fix<uns>_<mode>si2_hw): Rework the IEEE 128-bt hardware support
39309         to know that we can now have integers of all sizes in vector
39310         registers.
39311         (fix<uns>_<mode>di2_hw): Likewise.
39312         (float<uns>_<mode>si2_hw): Likewise.
39313         (fix_<mode>si2_hw): Likewise.
39314         (fixuns_<mode>si2_hw): Likewise.
39315         (float<uns>_<mode>di2_hw): Likewise.
39316         (float_<mode>di2_hw): Likewise.
39317         (float_<mode>si2_hw): Likewise.
39318         (floatuns_<mode>di2_hw): Likewise.
39319         (floatuns_<mode>si2_hw): Likewise.
39320         (xscvqp<su>wz_<mode>): Delete, no longer used.
39321         (xscvqp<su>dz_<mode>): Likewise.
39322         (xscv<su>dqp_<mode>): Likewise.
39323         (ieee128_mfvsrd_64bit): Likewise.
39324         (ieee128_mfvsrd_32bit): Likewise.
39325         (ieee128_mfvsrwz): Likewise.
39326         (ieee128_mtvsrw): Likewise.
39327         (ieee128_mtvsrd_64bit): Likewise.
39328         (ieee128_mtvsrd_32bit): Likewise.
39330 2017-01-31  Martin Liska  <mliska@suse.cz>
39332         PR ipa/79285
39333         * ipa-prop.c (ipa_free_all_node_params): Call release method
39334         instead of ~sumbol_summary to not to trigger double times
39335         dtor of hash_map.
39337 2017-01-31  Aldy Hernandez  <aldyh@redhat.com>
39339         PR tree-optimization/71691
39340         * bitmap.h (class auto_bitmap): New.
39341         * tree-ssa-loop-unswitch.c (tree_may_unswitch_on): Call
39342         is_maybe_undefined instead of ssa_undefined_value_p.
39344 2017-01-31  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
39346         * config/s390/s390-c.c (s390_cpu_cpp_builtins_internal): Rename
39347         __S390_ARCH_LEVEL__ to __ARCH__.
39349 2017-01-31  Jakub Jelinek  <jakub@redhat.com>
39351         PR tree-optimization/79267
39352         * value-prof.c (gimple_ic): Only drop lhs for noreturn calls
39353         if should_remove_lhs_p is true.
39355 2017-01-30  Alexandre Oliva  <aoliva@redhat.com>
39357         PR debug/63238
39358         * dwarf2out.c (clone_as_declaration): Drop DW_AT_alignment.
39359         (add_alignment_attribute): New.
39360         (base_type_die): Add alignment attribute.
39361         (subrange_type_die): Likewise.
39362         (modified_type_die): Likewise.
39363         (gen_array_type_die): Likewise.
39364         (gen_descr_array_type_die: Likewise.
39365         (gen_enumeration_type_die): Likewise.
39366         (gen_subprogram_die): Likewise.
39367         (gen_variable_die): Likewise.
39368         (gen_field_die): Likewise.
39369         (gen_ptr_to_mbr_type_die): Likewise.
39370         (gen_struct_or_union_type_die): Likewise.
39371         (gen_subroutine_type_die): Likewise.
39372         (gen_typedef_die): Likewise.
39373         (base_type_cmp): Compare alignment attribute.
39375 2017-01-30  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
39377         PR target/79170
39378         * config/rs6000/altivec.md (*setb_internal): Rename to setb_signed.
39379         (setb_unsigned) New pattern for setb with CCUNS.
39380         * config/rs6000/rs6000.c (expand_block_compare): Use a different
39381         subfc./subfe sequence to avoid overflow problems.  Generate a
39382         shorter sequence with cmpld/setb for power9.
39383         * config/rs6000/rs6000.md (subf<mode>3_carry_dot2): Add a new pattern
39384         for generating subfc. instruction.
39385         (cmpstrsi): Add TARGET_POPCNTD predicate as the generate sequence
39386         now uses this instruction.
39388 2017-01-30  Ian Lance Taylor  <iant@google.com>
39390         PR debug/79289
39391         * dwarf2out.c (gen_type_die_with_usage): When picking a variant
39392         for FUNCTION_TYPE/METHOD_TYPE, use the first matching one.
39394 2017-01-30  Martin Sebor  <msebor@redhat.com>
39396         * gimple-ssa-sprintf.c (fmtresult::adjust_for_width_or_precision):
39397         Move constant to the right of a relational operator.
39398         (get_mpfr_format_length, format_character, format_string): Ditto.
39399         (should_warn_p, maybe_warn): Same.
39401         * doc/invoke.texi (-Wformat-truncation=1): Fix typo.
39403 2017-01-30  Maxim Ostapenko  <m.ostapenko@samsung.com>
39405         PR lto/79061
39406         * asan.c (get_translation_unit_decl): Remove function.
39407         (asan_add_global): Force has_dynamic_init to zero in LTO mode.
39409 2017-01-30  Martin Liska  <mliska@suse.cz>
39411         PR gcov-profile/79259
39412         * opts.c (common_handle_option): Enable flag_ipa_bit_cp w/
39413         -fprofile-generate.
39415 2017-01-30  Martin Liska  <mliska@suse.cz>
39417         PR bootstrap/78985
39418         * config/aarch64/cortex-a57-fma-steering.c (func_fma_steering::analyze):
39419         Initialize variables with NULL value.
39421 2017-01-30  Richard Earnshaw  <rearnsh@arm.com>
39423         PR target/79260
39424         * config.gcc (arm*-*-*): Add arm/arm-flags.h and arm/arm-isa.h to
39425         tm_p_file.
39426         * arm/arm-protos.h: Don't directly include arm-flags.h and arm-isa.h.
39428 2017-01-30  Richard Biener  <rguenther@suse.de>
39430         PR tree-optimization/79276
39431         * tree-vrp.c (process_assert_insertions): Properly adjust common
39432         when removing a duplicate.
39434 2017-01-30  Richard Biener  <rguenther@suse.de>
39436         PR tree-optimization/79256
39437         * targhooks.c (default_builtin_vector_alignment_reachable): Honor
39438         BIGGEST_FIELD_ALIGNMENT and ADJUST_FIELD_ALIGN to fix up bogus
39439         alignment on TYPE.
39440         * tree.c (build_aligned_type): Set TYPE_USER_ALIGN.
39442 2017-01-30  Dominik Vogt  <vogt@linux.vnet.ibm.com>
39444         PR target/79240
39445         * config/s390/s390.md ("*r<noxa>sbg_<mode>_srl_bitmask")
39446         ("*r<noxa>sbg_<mode>_sll_bitmask")
39447         ("*extzv_<mode>_srl<clobbercc_or_nocc>")
39448         ("*extzv_<mode>_sll<clobbercc_or_nocc>"):
39449         Use contiguous_bitmask_nowrap_operand.
39451 2017-01-29  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
39453         PR target/79268
39454         * config/rs6000/altivec.h (vec_xl): Revise #define.
39455         (vec_xst): Likewise.
39457 2017-01-27  Uros Bizjak  <ubizjak@gmail.com>
39459         * config/i386/i386.c (print_reg): Use REGNO instead of true_regnum.
39461 2017-01-27  Bernd Schmidt  <bschmidt@redhat.com>
39463         PR rtl-optimization/79194
39464         * cprop.c (one_cprop_pass): Move deletion of code after unconditional
39465         traps before call to bypass_conditional_jumps.
39467 2017-01-27  Vladimir Makarov  <vmakarov@redhat.com>
39469         PR tree-optimization/71374
39470         * lra-constraints.c (check_conflict_input_operands): New.
39471         (match_reload): Use it.
39473 2017-01-27  Vladimir Makarov  <vmakarov@redhat.com>
39475         PR target/79131
39476         * lra-assigns.c (find_hard_regno_for_1): Take endianess for into
39477         account to calculate conflict_set.
39479 2017-01-27  Bin Cheng  <bin.cheng@arm.com>
39481         PR rtl-optimization/78559
39482         * combine.c (try_combine): Discard REG_EQUAL and REG_EQUIV for
39483         other_insn in combine.
39485 2017-01-27  Pekka Jääskeläinen  <pekka.jaaskelainen@parmance.com>
39487         * builtin-types.def: Use unsigned_char_type_node for BT_UINT8.  Use
39488         uint16_type_node for BT_UINT16.
39490 2017-01-27  David Malcolm  <dmalcolm@redhat.com>
39492         * doc/sourcebuild.texi (Testsuites): Add "GIMPLE Tests" and
39493         "RTL Tests" to menu.
39494         (GIMPLE Tests): New node.
39495         (RTL Tests): New node.
39497 2017-01-27  Richard Biener  <rguenther@suse.de>
39499         PR tree-optimization/79245
39500         * tree-loop-distribution.c (distribute_loop): Apply cost
39501         modeling also to detected patterns.
39503 2017-01-27  Richard Biener  <rguenther@suse.de>
39505         PR tree-optimization/71433
39506         * tree-vrp.c (register_new_assert_for): Revert earlier changes.
39507         (compare_assert_loc): New function.
39508         (process_assert_insertions): Sort and optimize assert locations
39509         to remove duplicates and push down identical assertions on
39510         edges to their destination block.
39512 2017-01-27  Richard Biener  <rguenther@suse.de>
39514         PR tree-optimization/79244
39515         * tree-vrp.c (remove_range_assertions): Forcefully propagate
39516         out SSA names even if abnormal.
39518 2017-01-27  Jakub Jelinek  <jakub@redhat.com>
39520         * realmpfr.h: Poison MPFR_RND{N,Z,U,D}.
39521         * gimple-ssa-sprintf.c (format_floating_max): Use GMP_RNDN
39522         instead of MPFR_RNDN.
39524 2017-01-27  Richard Earnshaw  <rearnsha@arm.com>
39526         PR target/79239
39527         * arm.c (arm_option_override): Don't call build_target_option_node
39528         until after doing all option overrides.
39529         (arm_valid_target_attribute_tree): Likewise.
39531 2017-01-27  Martin Liska  <mliska@suse.cz>
39533         * doc/invoke.texi (-fprofile-arcs): Document profiling support
39534         for {cd}tors and C++ {cd}tors.
39536 2017-01-27  Dominik Vogt  <vogt@linux.vnet.ibm.com>
39538         * config/s390/s390.md ("*setmem_long_and")
39539         ("*setmem_long_and_31z"): Use zero_extend instead of and.
39541 2017-01-26  Martin Sebor  <msebor@redhat.com>
39543         * gimple-ssa-sprintf.c (format_floating): Simplify the computation
39544         of precision.
39546 2017-01-26  Martin Sebor  <msebor@redhat.com>
39548         * gimple-ssa-sprintf.c (format_floating): Test HAVE_XFmode and
39549         HAVE_DFmode before using XFmode or DFmode.
39550         (parse_directive): Avoid using the z length modifier to avoid
39551         the ISO C++98 does not support the â€˜z’ gnu_printf length modifier.
39553         PR middle-end/78703
39554         * gimple-ssa-sprintf.c (adjust_for_width_or_precision): Change
39555         to accept adjustment as an array.
39556         (get_int_range): New function.
39557         (struct directive): Make width and prec arrays.
39558         (directive::set_width, directive::set_precision): Call get_int_range.
39559         (format_integer, format_floating): Handle width and precision ranges.
39560         (format_string, parse_directive): Same.
39562 2017-01-26  Jakub Jelinek  <jakub@redhat.com>
39564         PR debug/79129
39565         * dwarf2out.c (generate_skeleton_bottom_up): For children with
39566         comdat_type_p set, just clone them, but keep the children in the
39567         original DIE.
39569         PR debug/78835
39570         * dwarf2out.c (prune_unused_types): Mark all functions with DIEs
39571         which have direct callers with -fvar-tracking-assignments enabled
39572         in the current TU.
39573         (resolve_addr): Avoid adding skeleton DIEs for DW_AT_call_origin
39574         inside of type units.
39576 2017-01-26  Martin Sebor  <msebor@redhat.com>
39578         PR middle-end/78703
39579         * gimple-ssa-sprintf.c (struct result_range): Add likely and
39580         unlikely counters.
39581         (struct format_result): Replace number_chars, number_chars_min,
39582         and number_chars_max with a single member of struct result_range.
39583         Remove bounded.
39584         (format_result::operator+=): Adjust.
39585         (struct fmtresult): Remove bounded.  Handle likely and unlikely
39586         counters.
39587         (fmtresult::adjust_for_width_or_precision): New function.
39588         (fmtresult:type_max_digits): New function.
39589         (bytes_remaining): Handle likely and unlikely counters.
39590         (min_bytes_remaining): Remove.
39591         (format_percent): Simplify.
39592         (format_integer, format_floating): Set likely and unlikely counters.
39593         (get_string_length, format_character, format_string): Same.
39594         (format_plain, should_warn_p): New function.
39595         (maybe_warn): Call should_warn_p.  Update diagnostic messages
39596         and handle those for all directives, including plain strings.
39597         (format_directive): Handle likely and unlikely counters.
39598         Remove unnecessary quoting from diagnostics.  Add an informational
39599         note.
39600         (add_bytes): Remove.
39601         (pass_sprintf_length::compute_format_length): Simplify.
39602         (try_substitute_return_value): Handle likely and unlikely counters.
39604 2017-01-26  Carl Love  <cel@us.ibm.com>
39606         * config/rs6000/rs6000-c (altivec_overloaded_builtins): Remove
39607         bogus entries for the P8V_BUILTIN_VEC_VGBBD built-ins
39609 2017-01-26  Vladimir Makarov  <vmakarov@redhat.com>
39611         PR target/79131
39612         * lra-assigns.c (setup_live_pseudos_and_spill_after_risky): Take
39613         endianess for subregs into account.
39614         * lra-constraints.c (lra_constraints): Do risky transformations
39615         always on the first iteration.
39616         * lra-lives.c (check_pseudos_live_through_calls): Add arg
39617         last_call_used_reg_set.
39618         (process_bb_lives): Define and use last_call_used_reg_set.
39619         * lra.c (lra): Always continue after lra_constraints on the first
39620         iteration.
39622 2017-01-26  Kirill Yukhin  <kirill.yukhin@gmail.com>
39624         * gcc.target/i386/avx512bw-kshiftlq-2.c: Use unsigned long long
39625         constant.
39626         * gcc.target/i386/avx512bw-kshiftrq-2.c: Ditto.
39628 2017-01-26  Jakub Jelinek  <jakub@redhat.com>
39630         * config/i386/avx512fintrin.h (_ktest_mask16_u8,
39631         _ktestz_mask16_u8, _ktestc_mask16_u8, _kadd_mask16): Move to ...
39632         * config/i386/avx512dqintrin.h (_ktest_mask16_u8,
39633         _ktestz_mask16_u8, _ktestc_mask16_u8, _kadd_mask16): ... here.
39634         * config/i386/i386-builtin.def (__builtin_ia32_ktestchi,
39635         __builtin_ia32_ktestzhi, __builtin_ia32_kaddhi): Use
39636         OPTION_MASK_ISA_AVX512DQ instead of OPTION_MASK_ISA_AVX512F.
39637         * config/i386/sse.md (SWI1248_AVX512BWDQ2): New mode iterator.
39638         (kadd<mode>, ktest<mode>): Use it instead of SWI1248_AVX512BWDQ.
39640 2017-01-26  Marek Polacek  <polacek@redhat.com>
39642         PR c/79199
39643         * fold-const.c (operand_equal_p) [COND_EXPR]: Use OP_SAME_WITH_NULL
39644         for the third operand.
39646 2017-01-26  Jakub Jelinek  <jakub@redhat.com>
39648         PR middle-end/79236
39649         * omp-low.c (struct omp_context): Add simt_stmt field.
39650         (scan_omp_for): Return omp_context *.
39651         (scan_omp_simd): Set simt_stmt on the non-_simt_ SIMD
39652         context to the _simt_ SIMD stmt.
39653         (lower_omp_for): For combined SIMD with sibling _simt_
39654         SIMD, make sure to use the same decls in _looptemp_
39655         clauses as in the sibling.
39657 2017-01-26  David Sherwood  <david.sherwood@arm.com>
39659         PR middle-end/79212
39660         * gimplify.c (omp_notice_variable): Add GOVD_SEEN flag to variables in
39661         all contexts.
39663 2017-01-26  Jakub Jelinek  <jakub@redhat.com>
39665         PR target/70465
39666         * reg-stack.c (emit_swap_insn): Instead of fld a; fld b; fxchg %st(1);
39667         emit fld b; fld a; if possible.
39669         * brig-builtins.def: Update copyright years.
39670         * config/arm/arm_acle_builtins.def: Update copyright years.
39672 2017-01-25  Michael Meissner  <meissner@linux.vnet.ibm.com>
39674         PR target/79179
39675         * config/rs6000/vsx.md (vsx_extract_<mode>_store): Use wY
39676         constraint instead of o for the stxsd instruction.
39678 2017-01-25  Carl Love  <cel@us.ibm.com>
39680         * config/rs6000/rs6000-c (altivec_overloaded_builtins): Fix order
39681         of entries for ALTIVEC_BUILTIN_VEC_PACKS and P8V_BUILTIN_VEC_VGBBD.
39683 2017-01-25  Jonathan Wakely  <jwakely@redhat.com>
39685         * doc/invoke.texi (C++ Dialect Options): Fix typo.
39687 2017-01-25  Richard Biener  <rguenther@suse.de>
39689         PR tree-optimization/69264
39690         * target.def (vector_alignment_reachable): Improve documentation.
39691         * doc/tm.texi: Regenerate.
39692         * targhooks.c (default_builtin_vector_alignment_reachable): Simplify
39693         and add a comment.
39694         * tree-vect-data-refs.c (vect_supportable_dr_alignment): Revert
39695         earlier changes with respect to TYPE_USER_ALIGN.
39696         (vector_alignment_reachable_p): Likewise.  Improve dumping.
39698 2016-01-25  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
39700         PR target/79145
39701         * config/arm/arm.md (xordi3): Force constant operand into a register
39702         for TARGET_IWMMXT.
39704 2016-01-25  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
39706         * doc/invoke.texi (-fstore-merging): Correct default optimization
39707         levels at which it is enabled.
39708         (-O): Move -fstore-merging from list to...
39709         (-O2): ... Here.
39711 2017-01-25  Richard Biener  <rguenther@suse.de>
39713         PR debug/78363
39714         * omp-expand.c: Include debug.h.
39715         (expand_omp_taskreg): Make sure to generate early debug before
39716         outlining anything from a function.
39717         (expand_omp_target): Likewise.
39718         (grid_expand_target_grid_body): Likewise.
39720 2017-01-25  Maxim Ostapenko  <m.ostapenko@samsung.com>
39722         PR lto/79061
39723         * asan.c (get_translation_unit_decl): New function.
39724         (asan_add_global): Extract modules file name from globals
39725         TRANSLATION_UNIT_DECL name.
39727 2017-01-24  Eric Botcazou  <ebotcazou@adacore.com>
39729         PR target/77439
39730         * config/arm/arm.c (arm_function_ok_for_sibcall): Add back restriction
39731         for long calls with APCS frame and VFP.
39733 2017-01-24  David Malcolm  <dmalcolm@redhat.com>
39735         * cfg.c (original_copy_tables_initialized_p): New function.
39736         * cfg.h (original_copy_tables_initialized_p): New decl.
39737         * cfgrtl.c (relink_block_chain): Guard the call to
39738         free_original_copy_tables with a call to
39739         original_copy_tables_initialized_p.
39740         * cgraph.h (symtab_node::native_rtl_p): New decl.
39741         * cgraphunit.c (symtab_node::native_rtl_p): New function.
39742         (symtab_node::needed_p): Don't assert for early assembly output
39743         for __RTL functions.
39744         (cgraph_node::finalize_function): Set "force_output" for __RTL
39745         functions.
39746         (cgraph_node::analyze): Bail out early for __RTL functions.
39747         (analyze_functions): Update assertion to support __RTL functions.
39748         (cgraph_node::expand): Bail out early for __RTL functions.
39749         * final.c (rest_of_clean_state): Don't call delete_tree_ssa for
39750         __RTL functions.
39751         * function.h (struct function): Update comment for field
39752         "pass_startwith".
39753         * gimple-expr.c: Include "tree-pass.h".
39754         (gimple_has_body_p): Return false for __RTL functions.
39755         * Makefile.in (OBJS): Add run-rtl-passes.o.
39756         * pass_manager.h (gcc::pass_manager::get_rest_of_compilation): New
39757         accessor.
39758         (gcc::pass_manager::get_clean_slate): New accessor.
39759         * passes.c: Include "insn-addr.h".
39760         (should_skip_pass_p): Add logging.  Update logic for running
39761         "expand" to be compatible with both __GIMPLE and __RTL.  Guard
39762         property-provider override so it is only done for gimple passes.
39763         Don't skip dfinit.
39764         (skip_pass): New function.
39765         (execute_one_pass): Call skip_pass when skipping passes.
39766         * read-md.c (md_reader::read_char): Support filtering
39767         the input to a subset of line numbers.
39768         (md_reader::md_reader): Initialize fields
39769         m_first_line and m_last_line.
39770         (md_reader::read_file_fragment): New function.
39771         * read-md.h (md_reader::read_file_fragment): New decl.
39772         (md_reader::m_first_line): New field.
39773         (md_reader::m_last_line): New field.
39774         * read-rtl-function.c (function_reader::create_function): Only
39775         create cfun if it doesn't already exist.  Set PROP_rtl on cfun's
39776         curr_properties.  Set DECL_INITIAL to a dummy block.
39777         (read_rtl_function_body_from_file_range): New function.
39778         * read-rtl-function.h (read_rtl_function_body_from_file_range):
39779         New decl.
39780         * run-rtl-passes.c: New file.
39781         * run-rtl-passes.h: New file.
39783 2017-01-24  Jeff Law  <law@redhat.com>
39785         * config/microblaze/microblaze.h (ASM_FORMAT_PRIVATE_NAME): Increase
39786         buffer size.
39788 2017-01-24  Bin Cheng  <bin.cheng@arm.com>
39790         PR tree-optimization/79159
39791         * tree-ssa-loop-niter.c (get_cst_init_from_scev): New function.
39792         (record_nonwrapping_iv): Improve boundary using above function if no
39793         value range information.
39795 2017-01-24  Pekka Jääskeläinen  <pekka@parmance.com>
39796             Martin Jambor  <mjambor@suse.cz>
39798         * brig-builtins.def: New file.
39799         * builtins.def (DEF_HSAIL_BUILTIN): New macro.
39800         (DEF_HSAIL_ATOMIC_BUILTIN): Likewise.
39801         (DEF_HSAIL_SAT_BUILTIN): Likewise.
39802         (DEF_HSAIL_INTR_BUILTIN): Likewise.
39803         (DEF_HSAIL_CVT_ZEROI_SAT_BUILTIN): Likewise.
39804         * builtin-types.def (BT_INT8): New.
39805         (BT_INT16): Likewise.
39806         (BT_UINT8): Likewise.
39807         (BT_UINT16): Likewise.
39808         (BT_FN_ULONG): Likewise.
39809         (BT_FN_UINT_INT): Likewise.
39810         (BT_FN_UINT_ULONG): Likewise.
39811         (BT_FN_UINT_LONG): Likewise.
39812         (BT_FN_UINT_PTR): Likewise.
39813         (BT_FN_ULONG_PTR): Likewise.
39814         (BT_FN_INT8_FLOAT): Likewise.
39815         (BT_FN_INT16_FLOAT): Likewise.
39816         (BT_FN_UINT32_FLOAT): Likewise.
39817         (BT_FN_UINT16_FLOAT): Likewise.
39818         (BT_FN_UINT8_FLOAT): Likewise.
39819         (BT_FN_UINT64_FLOAT): Likewise.
39820         (BT_FN_UINT16_UINT32): Likewise.
39821         (BT_FN_UINT32_UINT16): Likewise.
39822         (BT_FN_UINT16_UINT16_UINT16): Likewise.
39823         (BT_FN_INT_PTR_INT): Likewise.
39824         (BT_FN_UINT_PTR_UINT): Likewise.
39825         (BT_FN_LONG_PTR_LONG): Likewise.
39826         (BT_FN_ULONG_PTR_ULONG): Likewise.
39827         (BT_FN_VOID_UINT64_UINT64): Likewise.
39828         (BT_FN_UINT8_UINT8_UINT8): Likewise.
39829         (BT_FN_INT8_INT8_INT8): Likewise.
39830         (BT_FN_INT16_INT16_INT16): Likewise.
39831         (BT_FN_INT_INT_INT): Likewise.
39832         (BT_FN_UINT_FLOAT_UINT): Likewise.
39833         (BT_FN_FLOAT_UINT_UINT): Likewise.
39834         (BT_FN_ULONG_UINT_UINT): Likewise.
39835         (BT_FN_ULONG_UINT_PTR): Likewise.
39836         (BT_FN_ULONG_ULONG_ULONG): Likewise.
39837         (BT_FN_UINT_UINT_UINT): Likewise.
39838         (BT_FN_VOID_UINT_PTR): Likewise.
39839         (BT_FN_UINT_UINT_PTR: Likewise.
39840         (BT_FN_UINT32_UINT64_PTR): Likewise.
39841         (BT_FN_INT_INT_UINT_UINT): Likewise.
39842         (BT_FN_UINT_UINT_UINT_UINT): Likewise.
39843         (BT_FN_UINT_UINT_UINT_PTR): Likewise.
39844         (BT_FN_UINT_ULONG_ULONG_UINT): Likewise.
39845         (BT_FN_ULONG_ULONG_ULONG_ULONG): Likewise.
39846         (BT_FN_LONG_LONG_UINT_UINT): Likewise.
39847         (BT_FN_ULONG_ULONG_UINT_UINT): Likewise.
39848         (BT_FN_VOID_UINT32_UINT64_PTR): Likewise.
39849         (BT_FN_VOID_UINT32_UINT32_PTR): Likewise.
39850         (BT_FN_UINT_UINT_UINT_UINT_UINT): Likewise.
39851         (BT_FN_UINT_FLOAT_FLOAT_FLOAT_FLOAT): Likewise.
39852         (BT_FN_ULONG_ULONG_ULONG_UINT_UINT): Likewise.
39853         * doc/frontends.texi: List BRIG FE.
39854         * doc/install.texi (Testing): Add BRIG tesring requirements.
39855         * doc/invoke.texi (Overall Options): Mention BRIG.
39856         * doc/standards.texi (Standards): Doucment BRIG HSA version.
39858 2017-01-24  Richard Biener  <rguenther@suse.de>
39860         PR translation/79208
39861         * ipa-devirt.c (odr_types_equivalent_p): Fix typo in diagnostic.
39863 2017-01-24  Martin Jambor  <mjambor@suse.cz>
39865         PR bootstrap/79198
39866         * ipa-prop.c (ipa_free_all_node_params): Call summary destructor.
39867         * ipa-prop.c (ipa_node_params_t::insert): Initialize fields known_csts
39868         and known_contexts.
39870 2017-01-24  Aldy Hernandez  <aldyh@redhat.com>
39872         PR middle-end/79123
39873         * gimple-ssa-warn-alloca.c (alloca_call_type): Make sure
39874         casts from signed to unsigned really don't have a range.
39876 2017-01-24  Markus Trippelsdorf  <markus@trippelsdorf.de>
39878         * gimple-ssa-sprintf.c (format_floating): Change MPFR_RNDx to
39879         GMP_RNDx for compatiblity.
39881 2017-01-24  Martin Liska  <mliska@suse.cz>
39883         PR bootstrap/79132
39884         * tree-ssa-reassoc.c (rewrite_expr_tree_parallel): Insert assert
39885         that would prevent us to call alloca with -1 as argument.
39887 2017-01-24  Jakub Jelinek  <jakub@redhat.com>
39889         * dwarf2out.c (output_compilation_unit_header, output_file_names):
39890         Avoid -Wformat-security warning.
39892 2017-01-23  Andrew Pinski  <apinski@cavium.com>
39894         * config/aarch64/aarch64.c (thunderx2t99_addrcost_table): Improve
39895         cost table.
39897 2017-01-23  Martin Sebor  <msebor@redhat.com>
39899         PR middle-end/78703
39900         * gimple-ssa-sprintf.c (warn_level): New global.
39901         (format_integer): Use it here and throughout the rest of the file.
39902         Use the same switch to compute sign as base.
39903         (maybe_warn): New function.
39904         (format_directive): Factor out warnings into maybe_warn.
39905         Add debugging output.  Use warn_level.
39906         (add_bytes): Use warn_level.
39907         (pass_sprintf_length::compute_format_length): Add debugging output.
39908         (try_substitute_return_value): Same.
39909         (pass_sprintf_length::handle_gimple_call): Set and use warn_level.
39911         PR middle-end/78703
39912         * gimple-ssa-sprintf.c (struct format_result): Remove constant member.
39913         (struct fmtresult, format_integer, format_floating): Adjust.
39914         (fmtresult::fmtresult): Set max correctly in two argument ctor.
39915         (get_string_length, format_string,format_directive): Same.
39916         (pass_sprintf_length::compute_format_length): Same.
39917         (try_substitute_return_value): Simplify slightly.
39919         PR middle-end/78703
39920         * gimple-ssa-sprintf.c (pass_sprintf_length::gate): Adjust formatting.
39921         (fmtresult::operator+=): Outlined.
39922         (struct fmtresult): Add ctors.
39923         (struct conversion_spec): Rename...
39924         (struct directive): ...to this.  Add and remove data members.
39925         (directive::set_width, directive::set_precision): New functions.
39926         (format_percent): Use fmtresult ctor.
39927         (get_width_and_precision): Remove.
39928         (format_integer): Make naming changes.  Avoid computing width and
39929         precision.
39930         (format_floating): Same.  Adjust indentation.
39931         (format_character, format_none): New functions.
39932         (format_string): Moved character handling to format_character.
39933         (format_directive): Remove arguments, change return type.
39934         (parse_directive): New function.
39935         (pass_sprintf_length::compute_format_length): Move directive
39936         parsing to parse_directive.
39938 2017-01-23  Jakub Jelinek  <jakub@redhat.com>
39940         * tree.h (assign_assembler_name_if_neeeded): Rename to ...
39941         (assign_assembler_name_if_needed): ... this.
39942         * tree.c (assign_assembler_name_if_neeeded): Rename to ...
39943         (assign_assembler_name_if_needed): ... this.
39944         (free_lang_data_in_cgraph): Adjust callers.
39945         * cgraphunit.c (cgraph_node::analyze): Likewise.
39946         * omp-expand.c (expand_omp_taskreg, expand_omp_target):
39947         Likewise.
39949 2017-01-23  Richard Biener  <rguenther@suse.de>
39951         PR tree-optimization/79088
39952         PR tree-optimization/79188
39953         * tree-ssa-threadupdate.c (mark_threaded_blocks): Move code
39954         resetting loop bounds after last path deletion.  Reset loop
39955         bounds of the target loop, make code match the comments.
39956         * tree-ssa-threadbackwards.c (pass_early_thread_jumps::execute):
39957         Make sure loops need no fixups.
39959 2017-01-23  Kelvin Nilsen  <kelvin@gcc.gnu.org>
39961         * config/rs6000/rs6000-builtin.def (VSIEDPF): Add scalar insert
39962         exponent support with double type for first argument.
39963         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Changed
39964         type returned by __builtin_vec_extract_sig,
39965         __builtin_vec_extract_sig_sp, and __builtin_vec_extract_sig_dp
39966         functions from "vector int" to "vector unsigned int" or from
39967         "vector long long int" to "vector unsigned long long int".
39968         Changed type returned by __builtin_vec_extract_exp,
39969         __builtin_vec_extract_exp_sp, and __builtin_vec_extract_exp_dp
39970         functions from "vector int" to "vector unsigned int" or from
39971         "vector long long int" to "vector unsigned long long int".
39972         Changed return type of __builtin_vec_test_data_class,
39973         __builtin_vec_test_data_class_sp, and
39974         __builtin_vec_test_data_class_dp from "vector int" to
39975         "vector bool int" or from "vector long long int" to "vector bool
39976         long long int" and changed second argument type from "unsigned
39977         int" to "int".  Added new overloaded function forms "vector float
39978         __builtin_vec_insert_exp (vector float, vector unsigned int)" and
39979         "vector float __builtin_vec_insert_exp_sp (vector float, vector
39980         unsigned int)" and "vector double __builtin_vec_insert_exp (vector
39981         double, vector unsigned long long int)" and "vector double
39982         __builtin_vec_insert_exp_dp (vector double, vector unsigned long
39983         long int)".  Changed return type of
39984         __builtin_scalar_test_data_class and
39985         __builtin_scalar_test_data_class_sp and
39986         __builtin_scalar_test_data_class_dp from "unsigned int" to "bool
39987         int" and changed second argument from "unsigned int" to "int".
39988         Changed type returned by __builtin_scalar_test_neg,
39989         __builtin_scalar_test_neg_sp, and __builtin_scalar_test_neg_dp
39990         from "int" to "bool int".  Added new overloaded function form
39991         "double __builtin_scalar_insert_exp (double, unsigned long long int)".
39992         * config/rs6000/vsx.md (xsiexpdpf): New insn for scalar insert
39993         exponent double-precision with floating point first argument.
39994         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Adjust
39995         documentation of scalar_test_data_class, scalar_test_neg,
39996         scalar_extract_sig, scalar_extract_exp, scalar_insert_exp,
39997         vector_extract_exp, vec_extract_sig, vec_insert_exp, and
39998         vec_test_data_class built-in functions to reflect refinements in
39999         their type signatures.
40001 2017-01-23  Andreas Tobler  <andreast@gcc.gnu.org>
40003         * config/aarch64/aarch64.c (aarch64_elf_asm_constructor): Increase
40004         size of buf.
40005         (aarch64_elf_asm_destructor): Likewise.
40007 2017-01-23  Bernd Schmidt  <bschmidt@redhat.com>
40009         PR rtl-optimization/78634
40010         * config/i386/i386.c (ix86_max_noce_ifcvt_seq_cost): New function.
40011         (TARGET_MAX_NOCE_IFCVT_SEQ_COST): Define.
40012         * ifcvt.c (noce_try_cmove): Add missing cost check.
40014         PR rtl-optimization/71724
40015         * combine.c (if_then_else_cond): Look for situations where it is
40016         beneficial to undo the work of one of the recursive calls.
40018 2017-01-23  Bin Cheng  <bin.cheng@arm.com>
40020         PR tree-optimization/70754
40021         * tree-predcom.c (stmt_combining_refs): New parameter INSERT_BEFORE.
40022         (reassociate_to_the_same_stmt): New parameter INSERT_BEFORE.  Insert
40023         combined stmt before it if not NULL.
40024         (combine_chains): Process refs reversely and compute dominance point
40025         for root ref.
40027 2017-01-23  Martin Liska  <mliska@suse.cz>
40029         PR tree-optimization/79196
40030         * tree-ssa-strlen.c (fold_strstr_to_memcmp): Rename to ...
40031         (fold_strstr_to_strncmp): ... this.  Fold the pattern to strncmp
40032         instead of memcmp.
40033         (strlen_optimize_stmt): Call the renamed function.
40035 2017-01-23  Michael Matz  <matz@suse.de>
40037         PR tree-optimization/78384
40038         * tree-ssa-loop-split.c (patch_loop_exit): Use correct edge.
40040 2017-01-23  Richard Biener  <rguenther@suse.de>
40042         PR tree-optimization/79186
40043         * tree-vrp.c (register_new_assert_for): Make sure we've seen
40044         both incoming edges before moving an assert.
40046 2017-01-23  Martin Jambor  <mjambor@suse.cz>
40048         * ipa-prop.c (load_from_param_1): Removed.
40049         (load_from_unmodified_param): Bits from load_from_param_1 put back
40050         here.
40051         (load_from_param): Removed.
40052         (compute_complex_assign_jump_func): Removed stmt2 and just replaced it
40053         with stmt.  Reverted back to use of load_from_unmodified_param.
40055 2017-01-23  Martin Jambor  <mjambor@suse.cz>
40057         PR ipa/79108
40058         * ipa-prop.h (ipa_param_descriptor): Anotate with with GTY(()).
40059         (ipa_node_params): Annotate with GTY((for_user)).  Make descriptors
40060         field a pointer to garbage collected vector, mark lattices and
40061         ipcp_orig_node with GTY((skip)).
40062         (ipa_get_param_count): Adjust to descriptors being a pointer.
40063         (ipa_get_param): Likewise.
40064         (ipa_get_type): Likewise.
40065         (ipa_get_param_move_cost): Likewise.
40066         (ipa_set_param_used): Likewise.
40067         (ipa_get_controlled_uses): Likewise.
40068         (ipa_set_controlled_uses): Likewise.
40069         (ipa_is_param_used): Likewise.
40070         (ipa_node_params_t): Move into garbage collector.  New methods insert
40071         and remove.
40072         (ipa_node_params_sum): Annotate wth GTY(()).
40073         (ipa_check_create_node_params): Adjust to ipa_node_params_sum being
40074         garbage collected.
40075         (ipa_load_from_parm_agg): Adjust declaration.
40076         * ipa-icf.c (param_used_p): Adjust to descriptors being a pointer.
40077         * ipa-profile.c (ipa_profile): Likewise.
40078         * ipa-prop.c (ipa_get_param_decl_index_1): Likewise.
40079         (ipa_populate_param_decls): Make descriptors parameter garbage
40080         collected.
40081         (ipa_dump_param): Adjust to descriptors being a pointer.
40082         (ipa_alloc_node_params): Likewise.
40083         (ipa_initialize_node_params): Likewise.
40084         (load_from_param_1): Make descriptors parameter garbage collected.
40085         (load_from_unmodified_param): Likewise.
40086         (load_from_param): Likewise.
40087         (ipa_load_from_parm_agg): Likewise.
40088         (ipa_node_params::~ipa_node_params): Removed.
40089         (ipa_free_all_node_params): Remove call to delete operator.
40090         (ipa_node_params_t::insert): New.
40091         (ipa_node_params_t::remove): Likewise.
40092         (ipa_node_params_t::duplicate): Adjust to descriptors being a pointer,
40093         copy known_csts and known_contexts vectors.
40094         (ipa_read_node_info): Adjust to descriptors being a pointer.
40095         (ipcp_modif_dom_walker): Make m_descriptors field garbage
40096         collected.
40097         (ipcp_transform_function): Make descriptors variable garbage
40098         collected.
40100 2017-01-23  Andrew Senkevich  <andrew.senkevich@intel.com>
40102         * config/i386/avx512bwintrin.h: Add k-mask test, kortest intrinsics.
40103         * config/i386/avx512dqintrin.h: Ditto.
40104         * config/i386/avx512fintrin.h: Ditto.
40105         * config/i386/i386.c: Handle new builtins.
40106         * config/i386/i386-builtin.def: Add new builtins.
40107         * config/i386/sse.md (ktest<mode>, kortest<mode>): New.
40108         (UNSPEC_KORTEST, UNSPEC_KTEST): New.
40110 2017-01-23  Jakub Jelinek  <jakub@redhat.com>
40111             Martin Liska  <mliska@suse.cz>
40113         * asan.h: Define ASAN_USE_AFTER_SCOPE_ATTRIBUTE.
40114         * asan.c (asan_expand_poison_ifn): Support stores and use
40115         appropriate ASAN report function.
40116         * internal-fn.c (expand_ASAN_POISON_USE): New function.
40117         * internal-fn.def (ASAN_POISON_USE): Declare.
40118         * tree-into-ssa.c (maybe_add_asan_poison_write): New function.
40119         (maybe_register_def): Create ASAN_POISON_USE when sanitizing.
40120         * tree-ssa-dce.c (eliminate_unnecessary_stmts): Remove
40121         ASAN_POISON calls w/o LHS.
40122         * tree-ssa.c (execute_update_addresses_taken): Create clobber
40123         for ASAN_MARK (UNPOISON, &x, ...) in order to prevent usage of a LHS
40124         from ASAN_MARK (POISON, &x, ...) coming to a PHI node.
40125         * gimplify.c (asan_poison_variables): Add attribute
40126         use_after_scope_memory to variables that really needs to live
40127         in memory.
40128         * tree-ssa.c (is_asan_mark_p): Do not rewrite into SSA when
40129         having the attribute.
40131 2017-01-23  Martin Liska  <mliska@suse.cz>
40133         * asan.c (create_asan_shadow_var): New function.
40134         (asan_expand_poison_ifn): Likewise.
40135         * asan.h (asan_expand_poison_ifn): New declaration.
40136         * internal-fn.c (expand_ASAN_POISON): Likewise.
40137         * internal-fn.def (ASAN_POISON): New builtin.
40138         * sanopt.c (pass_sanopt::execute): Expand
40139         asan_expand_poison_ifn.
40140         * tree-inline.c (copy_decl_for_dup_finish): Make function
40141         external.
40142         * tree-inline.h (copy_decl_for_dup_finish): Likewise.
40143         * tree-ssa.c (is_asan_mark_p): New function.
40144         (execute_update_addresses_taken): Rewrite local variables
40145         (identified just by use-after-scope as addressable) into SSA.
40147 2017-01-22  Gerald Pfeifer  <gerald@pfeifer.com>
40149         * doc/install.texi (Specific): opensource.apple.com uses https
40150         now. Remove trailing slash.
40152 2017-01-22  Gerald Pfeifer  <gerald@pfeifer.com>
40154         * README.Portability: Remove note on an Irix compatibility issue.
40156 2017-01-22  Dimitry Andric <dim@FreeBSD.org>
40158         * gcov.c (INCLUDE_ALGORITHM): Define.
40159         (INCLUDE_VECTOR): Define.
40160         No longer include <vector> and <algorithm> directly.
40162 2017-01-21  Gerald Pfeifer  <gerald@pfeifer.com>
40164         * doc/extend.texi (Thread-Local): Change www.akkadia.org reference
40165         to https.
40166         * doc/invoke.texi (Code Gen Options): Ditto.
40168 2017-01-21  Jan Hubicka  <hubicka@ucw.cz>
40170         PR lto/78407
40171         * cfg.c (update_bb_profile_for_threading): Fix updating of probablity.
40173 2017-01-21  Bernd Schmidt  <bschmidt@redhat.com>
40175         rtl-optimization/79125
40176         * cprop.c (local_cprop_pass): Handle cases where we make an
40177         unconditional trap.
40179 2017-01-20  Segher Boessenkool  <segher@kernel.crashing.org>
40181         PR target/61729
40182         PR target/77850
40183         * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Adjust address to
40184         read from, for big endian.
40186 2017-01-20  Jiong Wang  <jiong.wang@arm.com>
40188         * config/aarch64/aarch64-builtins.c (aarch64_init_builtins): Register
40189         register pauth builtins for LP64 only.
40191 2017-01-20  Marek Polacek  <polacek@redhat.com>
40193         PR c/79152
40194         * gimplify.c (should_warn_for_implicit_fallthrough): Handle consecutive
40195         non-case labels.
40197 2017-01-20  Alexander Monakov  <amonakov@ispras.ru>
40199         * omp-expand.c (expand_omp_simd): Clear PROP_gimple_lomp_dev regardless
40200         of safelen status.
40201         * omp-offload.c (pass_omp_device_lower::gate): Use PROP_gimple_lomp_dev.
40202         * passes.c (dump_properties): Handle PROP_gimple_lomp_dev.
40203         * tree-inline.c (expand_call_inline): Propagate PROP_gimple_lomp_dev.
40205 2017-01-20  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
40207         PR target/71270
40208         * config/arm/arm.c (neon_valid_immediate): Reject vector constants
40209         in big-endian mode when they are not a single duplicated value.
40211 2017-01-20  Richard Biener  <rguenther@suse.de>
40213         * BASE-VER: Bump to 7.0.1.
40215 2017-01-20  Alexander Monakov  <amonakov@ispras.ru>
40217         * omp-low.c (omplow_simd_context): New struct.  Use it...
40218         (lower_rec_simd_input_clauses): ...here and...
40219         (lower_rec_input_clauses): ...here to hold common data.  Adjust all
40220         references to idx, lane, max_vf, is_simt.
40222 2017-01-20  Graham Markall  <graham.markall@embecosm.com>
40224         * config/arc/arc.h (LINK_SPEC): Use arclinux_nps emulation when
40225         mcpu=nps400.
40227 2017-01-20  Martin Jambor  <mjambor@suse.cz>
40229         * hsa.h: Renaed to hsa-common.h.  Adjusted a comment.
40230         * hsa.c: Renaed to hsa-common.c.  Change include of gt-hsa.h to
40231         gt-hsa-common.h.
40232         * Makefile.in (OBJS): Rename hsa.o to hsa-common.o.
40233         (GTFILES): Rename hsa.c to hsa-common.c.
40234         * hsa-brig.c: Change include of hsa.h to hsa-common.h.
40235         * hsa-dump.c: Likewise.
40236         * hsa-gen.c: Likewise.
40237         * hsa-regalloc.c: Likewise.
40238         * ipa-hsa.c: Likewise.
40239         * omp-expand.c: Likewise.
40240         * omp-low.c: Likewise.
40241         * toplev.c: Likewise.
40243 2017-01-20  Marek Polacek  <polacek@redhat.com>
40245         PR c/64279
40246         * doc/invoke.texi: Document -Wduplicated-branches.
40247         * fold-const.c (operand_equal_p): Handle MODIFY_EXPR, INIT_EXPR,
40248         COMPOUND_EXPR, PREDECREMENT_EXPR, PREINCREMENT_EXPR,
40249         POSTDECREMENT_EXPR, POSTINCREMENT_EXPR, CLEANUP_POINT_EXPR, EXPR_STMT,
40250         STATEMENT_LIST, and RETURN_EXPR.  For non-pure non-const functions
40251         return 0 only when not OEP_LEXICOGRAPHIC.
40252         (fold_build_cleanup_point_expr): Use the expression
40253         location when building CLEANUP_POINT_EXPR.
40254         * tree-core.h (enum operand_equal_flag): Add OEP_LEXICOGRAPHIC.
40255         * tree.c (add_expr): Handle error_mark_node.
40257 2017-01-20  Martin Liska  <mliska@suse.cz>
40259         PR lto/69188
40260         * tree-profile.c (init_ic_make_global_vars): Do not call
40261         finalize_decl.
40262         (gimple_init_gcov_profiler): Likewise.
40264 2017-01-20  Martin Liska  <mliska@suse.cz>
40266         PR ipa/71190
40267         * cgraph.h (maybe_create_reference): Remove argument and
40268         update comment.
40269         * cgraphclones.c (cgraph_node::create_virtual_clone): Remove one
40270         argument.
40271         * ipa-cp.c (create_specialized_node): Likewise.
40272         * symtab.c (symtab_node::maybe_create_reference): Handle
40273         VAR_DECLs and ADDR_EXPRs and select ipa_ref_use type.
40275 2017-01-20  Martin Liska  <mliska@suse.cz>
40277         * read-rtl-function.c (function_reader::create_function): Use
40278         build_decl instread of build_decl_stat.
40280 2017-01-20  Andrew Senkevich  <andrew.senkevich@intel.com>
40282         * config/i386/avx512bwintrin.h: Add k-mask registers shift intrinsics.
40283         * config/i386/avx512dqintrin.h: Ditto.
40284         * config/i386/avx512fintrin.h: Ditto.
40285         * config/i386/i386-builtin-types.def: Add new types.
40286         * config/i386/i386.c: Handle new types.
40287         * config/i386/i386-builtin.def (__builtin_ia32_kshiftliqi)
40288         (__builtin_ia32_kshiftlihi, __builtin_ia32_kshiftlisi)
40289         (__builtin_ia32_kshiftlidi, __builtin_ia32_kshiftriqi)
40290         (__builtin_ia32_kshiftrihi, __builtin_ia32_kshiftrisi)
40291         (__builtin_ia32_kshiftridi): New.
40292         * config/i386/sse.md (k<code><mode>): Rename *k<code><mode>.
40294 2017-01-19  Segher Boessenkool  <segher@kernel.crashing.org>
40296         PR target/78875
40297         PR target/79140
40298         * config/rs6000/rs6000.c (TARGET_STACK_PROTECT_GUARD): Unconditionally
40299         define to rs6000_init_stack_protect_guard.
40300         (rs6000_init_stack_protect_guard): New function.
40302 2017-01-19  Matthew Fortune  <matthew.fortune@imgtec.com>
40303             Yunqiang Su  <yunqiang.su@imgtec.com>
40305         * config.gcc (supported_defaults): Add madd4.
40306         (with_madd4): Add validation.
40307         (all_defaults): Add madd4.
40308         * config/mips/mips.opt (mmadd4): New option.
40309         * config/mips/mips.h (OPTION_DEFAULT_SPECS): Add a default for
40310         mmadd4.
40311         (TARGET_CPU_CPP_BUILTINS): Add builtin_define for
40312         __mips_no_madd4.
40313         (ISA_HAS_UNFUSED_MADD4): Gate with mips_madd4.
40314         (ISA_HAS_FUSED_MADD4): Likewise.
40315         * doc/invoke.texi (-mmadd4): Document the new option.
40316         * doc/install.texi (--with-madd4): Document the new option.
40318 2017-01-19  Jiong Wang  <jiong.wang@arm.com>
40320         * config/aarch64/aarch64-builtins.c (enum aarch64_builtins): New
40321         entries for AARCH64_PAUTH_BUILTIN_XPACLRI,
40322         AARCH64_PAUTH_BUILTIN_PACIA1716, AARCH64_PAUTH_BUILTIN_AUTIA1716.
40323         (aarch64_init_pauth_hint_builtins): New.
40324         (aarch64_init_builtins): Call aarch64_init_pauth_hint_builtins.
40325         (aarch64_expand_builtin): Expand new builtins.
40327 2017-01-19  Jiong Wang  <jiong.wang@arm.com>
40329         * reg-notes.def (CFA_TOGGLE_RA_MANGLE): New reg-note.
40330         * combine-stack-adj.c (no_unhandled_cfa): Handle
40331         REG_CFA_TOGGLE_RA_MANGLE.
40332         * dwarf2cfi.c (dwarf2out_frame_debug): Handle REG_CFA_TOGGLE_RA_MANGLE.
40333         * config/aarch64/aarch64.c (aarch64_expand_prologue): Generates DWARF
40334         info for return address signing.
40335         (aarch64_expand_epilogue): Likewise.
40337 2017-01-19  Jiong Wang  <jiong.wang@arm.com>
40339         * config/aarch64/aarch64-opts.h (aarch64_function_type): New enum.
40340         * config/aarch64/aarch64-protos.h
40341         (aarch64_return_address_signing_enabled): New declaration.
40342         * config/aarch64/aarch64.c (aarch64_return_address_signing_enabled):
40343         New function.
40344         (aarch64_expand_prologue): Sign return address before it's pushed onto
40345         stack.
40346         (aarch64_expand_epilogue): Authenticate return address fetched from
40347         stack.
40348         (aarch64_override_options): Sanity check for ILP32 and ISA level.
40349         (aarch64_attributes): New function attributes for "sign-return-address".
40350         * config/aarch64/aarch64.md (UNSPEC_AUTI1716, UNSPEC_AUTISP,
40351         UNSPEC_PACI1716, UNSPEC_PACISP, UNSPEC_XPACLRI): New unspecs.
40352         ("*do_return"): Generate combined instructions according to key index.
40353         ("<pauth_mnem_prefix>sp", "<pauth_mnem_prefix1716", "xpaclri"): New.
40354         * config/aarch64/iterators.md (PAUTH_LR_SP, PAUTH_17_16): New integer
40355         iterators.
40356         (pauth_mnem_prefix, pauth_hint_num_a): New integer attributes.
40357         * config/aarch64/aarch64.opt (msign-return-address=): New.
40358         * doc/extend.texi (AArch64 Function Attributes): Documents
40359         "sign-return-address=".
40360         * doc/invoke.texi (AArch64 Options): Documents "-msign-return-address=".
40362 2017-01-19  Matthew Fortune  <matthew.fortune@imgtec.com>
40364         * doc/invoke.texi: Add missing -mlxc1-sxc1 options to
40365         overall option summary.
40367 2017-01-19  Jiong Wang  <jiong.wang@arm.com>
40369         * config/aarch64/aarch64-arches.def: New entry for "armv8.3-a".
40370         * config/aarch64/aarch64.h (AARCH64_FL_V8_3, AARCH64_FL_FOR_ARCH8_3,
40371         AARCH64_ISA_V8_3, TARGET_ARMV8_3): New.
40372         * doc/invoke.texi (AArch64 Options): Document "armv8.3-a".
40374 2017-01-19  Michael Meissner  <meissner@linux.vnet.ibm.com>
40376         * config/rs6000/rs6000-cpus.def (ISA_3_0_MASKS_SERVER): Enable
40377         -mpower9-minmax by default for -mcpu=power9.
40378         (ISA_3_MASKS_IEEE): Require -mvsx-small-integer to enable IEEE
40379         128-bit floating point.
40381 2017-01-20  Alan Modra  <amodra@gmail.com>
40383         * config/rs6000/rs6000.md (cmpstrnsi, cmpstrsi): Fail if
40384         optimizing for size.
40386 2017-01-20  Alan Modra  <amodra@gmail.com>
40388         PR target/79144
40389         * config/rs6000/rs6000.c (expand_strn_compare): Get the asm name
40390         for strcmp and strncmp from corresponding builtin decl.
40392 2017-01-19  Uros Bizjak  <ubizjak@gmail.com>
40394         * config.gcc (x86_64-*-rtems*): Use i386/rtemself.h
40395         instead of i386/rtems-64.h.
40396         * config/i386/rtems-64.h: Remove.
40398 2017-01-19  Uros Bizjak  <ubizjak@gmail.com>
40400         PR target/78478
40401         Revert:
40402         2013-11-05  Uros Bizjak  <ubizjak@gmail.com>
40404         * config/i386/rtemself.h (LONG_DOUBLE_TYPE_SIZE): New define.
40406 2017-01-19  Tamar Christina  <tamar.christina@arm.com>
40408         * config/aarch64/aarch64.c (aarch64_simd_gen_const_vector_dup):
40409         Change int to HOST_WIDE_INT.
40410         * config/aarch64/aarch64-protos.h
40411         (aarch64_simd_gen_const_vector_dup): Likewise.
40412         * config/aarch64/aarch64-simd.md: Add copysign<mode>3.
40414 2017-01-19  David Malcolm  <dmalcolm@redhat.com>
40416         * langhooks-def.h (lhd_type_for_size): New decl.
40417         (LANG_HOOKS_TYPE_FOR_SIZE): Define as lhd_type_for_size.
40418         * langhooks.c (lhd_type_for_size): New function, taken from
40419         lto_type_for_size.
40421 2017-01-19  Pat Haugen  <pthaugen@us.ibm.com>
40423         * config/rs6000/power9.md (power9-alu): Remove 'cmp' type and add
40424         define_bypass for CR latency.
40425         (power9-cracked-alu): Update bypass latency and remove power9-branch.
40426         (power9-alu2): Add define_bypass for CR latency.
40427         (power9-cmp): New.
40428         (power9-mul): Update insn latency.
40429         (power9-mul-compare): Update insn latency, bypass latency and remove
40430         power9-branch.
40432 2016-01-19  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
40434         * config/aarch64/aarch64-protos.h (aarch64_nopcrelative_literal_loads):
40435         Delete.
40436         * config/aarch64/aarch64.md
40437         (aarch64_reload_movcp<GPF_TF:mode><P:mode>): Delete reference to
40438         aarch64_nopcrelative_literal_loads.
40439         (aarch64_reload_movcp<VALL:mode><P:mode>): Likewise.
40441 2017-01-19  Chenghua Xu  <paul.hua.gm@gmail.com>
40443         * config/mips/mips.h (ISA_HAS_FUSED_MADD4): Enable for
40444         TARGET_LOONGSON_3A.
40445         (ISA_HAS_UNFUSED_MADD4): Exclude TARGET_LOONGSON_3A.
40447 2017-01-19  Doug Gilmore  <doug.gilmore@imgtec.com>
40449         PR target/78176
40450         * config.gcc (supported_defaults): Add lxc1-sxc1.
40451         (with_lxc1_sxc1): Add validation.
40452         (all_defaults): Add lxc1-sxc1.
40453         * config/mips/mips.opt (mlxc1-sxc1): New option.
40454         * config/mips/mips.h (OPTION_DEFAULT_SPECS): Add a default for
40455         mlxc1-sxc1.
40456         (TARGET_CPU_CPP_BUILTINS): Add builtin_define for
40457         __mips_no_lxc1_sxc1.
40458         (ISA_HAS_LXC1_SXC1): Gate with mips_lxc1_sxc1.
40459         * doc/invoke.texi (-mlxc1-sxc1): Document the new option.
40460         * doc/install.texi (--with-lxc1-sxc1): Document the new option.
40462 2017-01-19  Richard Biener  <rguenther@suse.de>
40464         PR tree-optimization/72488
40465         * tree-ssa-sccvn.c (run_scc_vn): When we abort the VN make
40466         sure to restore SSA info.
40467         * tree-ssa.c (verify_ssa): Verify SSA info is not shared.
40469 2017-01-19  Richard Earnshaw  <rearnsha@arm.com>
40471         PR rtl-optimization/79121
40472         * expr.c (expand_expr_real_2, case LSHIFT_EXPR): Look at the signedness
40473         of the inner type when shifting an extended value.
40475 2017-01-17  Jan Hubicka  <hubicka@ucw.cz>
40477         PR lto/78407
40478         * symtab.c (symtab_node::equal_address_to): Fix comparing of
40479         interposable aliases.
40481 2017-01-18  Peter Bergner  <bergner@vnet.ibm.com>
40483         PR target/78516
40484         * config/rs6000/spe.md (mov_si<mode>_e500_subreg0): Fix constraints.
40485         Use the evmergelohi instruction.
40486         (mov_si<mode>_e500_subreg4_2_le): Likewise.
40487         (mov_sitf_e500_subreg8_2_be): Likewise.
40488         (mov_sitf_e500_subreg12_2_le): Likewise.
40489         (mov_si<mode>_e500_subreg0_2_le): Fix constraints.
40490         (mov_si<mode>_e500_subreg4_2_be): Likewise.
40491         (mov_sitf_e500_subreg8_2_le): Likewise.
40492         (mov_sitf_e500_subreg12_2_be): Likewise.
40494 2017-01-18  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
40496         * config/rs6000/altivec.md (altivec_vbpermq): Change "type"
40497         attribute from vecsimple to vecperm.
40498         (altivec_vbpermq2): Likewise.
40500 2017-01-18  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
40502         PR target/79040
40503         * config/rs6000/altivec.h: Fix typo of vec_cntlz to vec_cnttz.
40505 2017-01-18  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
40506         * config/rs6000/rs6000-protos.h (expand_strn_compare): Add arg.
40507         * config/rs6000/rs6000.c (expand_strn_compare): Add ability to expand
40508         strcmp. Fix bug where comparison didn't stop with zero byte. Fix
40509         case where N arg is SIZE_MAX.
40510         * config/rs6000/rs6000.md (cmpstrnsi): Args to expand_strn_compare.
40511         (cmpstrsi): Add pattern.
40513 2017-01-18  Michael Meissner  <meissner@linux.vnet.ibm.com>
40515         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
40516         __builtin_vec_revb builtins.
40517         * config/rs6000/rs6000-builtins.def (P9V_BUILTIN_XXBRQ_V16QI): Add
40518         built-in functions to support generation of the ISA 3.0 XXBR<x>
40519         vector byte reverse instructions.
40520         (P9V_BUILTIN_XXBRQ_V1TI): Likewise.
40521         (P9V_BUILTIN_XXBRD_V2DI): Likewise.
40522         (P9V_BUILTIN_XXBRD_V2DF): Likewise.
40523         (P9V_BUILTIN_XXBGW_V4SI): Likewise.
40524         (P9V_BUILTIN_XXBGW_V4SF): Likewise.
40525         (P9V_BUILTIN_XXBGH_V8HI): Likewise.
40526         (P9V_BUILTIN_VEC_REVB): Likewise.
40527         * config/rs6000/vsx.md (p9_xxbrq_v1ti): New insns/expanders to
40528         generate the ISA 3.0 XXBR<x> vector byte reverse instructions.
40529         (p9_xxbrq_v16qi): Likewise.
40530         (p9_xxbrd_<mode>, VSX_D iterator): Likewise.
40531         (p9_xxbrw_<mode>, VSX_W iterator): Likewise.
40532         (p9_xxbrh_v8hi): Likewise.
40533         * config/rs6000/altivec.h (vec_revb): Define if ISA 3.0.
40534         * doc/extend.texi (RS/6000 Altivec Built-ins): Document the
40535         vec_revb built-in functions.
40537 2017-01-18  Uros Bizjak  <ubizjak@gmail.com>
40539         PR rtl-optimization/78952
40540         * config/i386/i386.md (any_extract): New code iterator.
40541         (*insvqi_2): Use any_extract for source operand.
40542         (*insvqi_3): Use any_shiftrt for source operand.
40544 2017-01-18  Wilco Dijkstra  <wdijkstr@arm.com>
40546         * config/aarch64/aarch64.c (aarch64_sched_adjust_priority)
40547         New function.
40548         (TARGET_SCHED_ADJUST_PRIORITY): Define target hook.
40550 2017-01-18  Matthias Klose  <doko@ubuntu.com>
40552         * doc/install.texi: Allow default for --with-target-bdw-gc-include.
40554 2016-01-18  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
40556         * config/rs6000/altivec.h (vec_bperm): Change #define.
40557         * config/rs6000/altivec.md (UNSPEC_VBPERMD): New enum constant.
40558         (altivec_vbpermq2): New define_insn.
40559         (altivec_vbpermd): Likewise.
40560         * config/rs6000/rs6000-builtin.def (VBPERMQ2): New monomorphic
40561         function interface.
40562         (VBPERMD): Likewise.
40563         (VBPERM): New polymorphic function interface.
40564         * config/rs6000/r6000-c.c (altivec_overloaded_builtins_table):
40565         Add entries for P9V_BUILTIN_VEC_VBPERM.
40566         * doc/extend.texi: Add interfaces for vec_bperm.
40568 2017-01-18  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
40570         * config/s390/s390-c.c (s390_expand_overloaded_builtin): Downcase
40571         first letter of error messages.
40572         (s390_resolve_overloaded_builtin): Likewise.
40573         * config/s390/s390.c (s390_expand_builtin): Likewise.
40574         (s390_invalid_arg_for_unprototyped_fn): Likewise.
40575         (s390_valid_target_attribute_inner_p): Likewise.
40576         * config/s390/s390.md ("tabort"): Likewise.
40578 2017-01-18  Toma Tabacu  <toma.tabacu@imgtec.com>
40580         * config/mips/mips.h (ISA_HAS_DIV3): Remove unused macro.
40581         (ISA_AVOID_DIV_HILO): New macro.
40582         (ISA_HAS_DIV): Use new ISA_AVOID_DIV_HILO macro.
40583         (ISA_HAS_DDIV): Likewise.
40585 2017-01-18  Markus Trippelsdorf  <markus@trippelsdorf.de>
40587         * doc/invoke.texi (fabi-version): Correct number of occurrences.
40589 2017-01-18  Markus Trippelsdorf  <markus@trippelsdorf.de>
40591         * doc/invoke.texi (fabi-version): Spelling fix.
40593 2017-01-18  Markus Trippelsdorf  <markus@trippelsdorf.de>
40595         PR c++/70182
40596         * doc/invoke.texi (fabi-version): Mention mangling fix for
40597         operator names.
40599 2017-01-18  Markus Trippelsdorf  <markus@trippelsdorf.de>
40601         PR c++/77489
40602         * doc/invoke.texi (fabi-version): Document discriminator mangling.
40604 2017-01-17  Segher Boessenkool  <segher@kernel.crashing.org>
40606         PR target/78875
40607         * config/rs6000/rs6000-opts.h (stack_protector_guard): New enum.
40608         * config/rs6000/rs6000.c (rs6000_option_override_internal): Handle
40609         the new options.
40610         * config/rs6000/rs6000.md (stack_protect_set): Handle the new more
40611         flexible settings.
40612         (stack_protect_test): Ditto.
40613         * config/rs6000/rs6000.opt (mstack-protector-guard=,
40614         mstack-protector-guard-reg=, mstack-protector-guard-offset=): New
40615         options.
40616         * doc/invoke.texi (Option Summary) [RS/6000 and PowerPC Options]:
40617         Add -mstack-protector-guard=, -mstack-protector-guard-reg=, and
40618         -mstack-protector-guard-offset=.
40619         (RS/6000 and PowerPC Options): Ditto.
40621 2017-01-17  Uros Bizjak  <ubizjak@gmail.com>
40623         * config/i386/i386.h (MASK_CLASS_P): New define.
40624         * config/i386/i386.c (inline_secondary_memory_needed): Ensure that
40625         there are no registers from different register sets also when
40626         mask registers are used.  Update function comment.
40627         * config/i386/i386.md (*movsi_internal): Split (*k/*krm) alternative
40628         to (*k/*r) and (*k/*km) alternatives.
40630 2017-01-17  Wilco Dijkstra  <wdijkstr@arm.com>
40632         * config/aarch64/aarch64.md (eh_return): Remove pattern and splitter.
40633         * config/aarch64/aarch64.h (AARCH64_EH_STACKADJ_REGNUM): Remove.
40634         (EH_RETURN_HANDLER_RTX): New define.
40635         * config/aarch64/aarch64.c (aarch64_frame_pointer_required):
40636         Force frame pointer in EH return functions.
40637         (aarch64_expand_epilogue): Add barrier for eh_return.
40638         (aarch64_final_eh_return_addr): Remove.
40639         (aarch64_eh_return_handler_rtx): New function.
40640         * config/aarch64/aarch64-protos.h (aarch64_final_eh_return_addr):
40641         Remove.
40642         (aarch64_eh_return_handler_rtx): New prototype.
40644 2017-01-17  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
40646         * config/rs6000/altivec.h (vec_rlmi): New #define.
40647         (vec_vrlnm): Likewise.
40648         (vec_rlnm): Likewise.
40649         * config/rs6000/altivec.md (UNSPEC_VRLMI): New UNSPEC enum value.
40650         (UNSPEC_VRLNM): Likewise.
40651         (VIlong): New mode iterator.
40652         (altivec_vrl<VI_char>mi): New define_insn.
40653         (altivec_vrl<VI_char>nm): Likewise.
40654         * config/rs6000/rs6000-builtin.def (VRLWNM): New monomorphic
40655         function entry.
40656         (VRLDNM): Likewise.
40657         (RLNM): New polymorphic function entry.
40658         (VRLWMI): New monomorphic function entry.
40659         (VRLDMI): Likewise.
40660         (RLMI): New polymorphic function entry.
40661         * config/rs6000/r6000-c.c (altivec_overloaded_builtin_table): Add
40662         new entries for P9V_BUILTIN_VEC_RLMI and P9V_BUILTIN_VEC_RLNM.
40663         * doc/extend.texi: Add description of vec_rlmi, vec_rlnm, and
40664         vec_vrlnm.
40666 2017-01-17  Jakub Jelinek  <jakub@redhat.com>
40668         PR debug/78839
40669         * dwarf2out.c (field_byte_offset): Restore the
40670         PCC_BITFIELD_TYPE_MATTERS behavior for INTEGER_CST DECL_FIELD_OFFSET
40671         and DECL_FIELD_BIT_OFFSET.  Use fold_build2 instead of build2 + fold.
40672         (analyze_variants_discr, gen_variant_part): Use fold_build2 instead
40673         of build2 + fold.
40675 2017-01-17  Eric Botcazou  <ebotcazou@adacore.com>
40677         PR ada/67205
40678         * config/aarch64/aarch64.c (TARGET_CUSTOM_FUNCTION_DESCRIPTORS): Define
40680 2017-01-17  Jakub Jelinek  <jakub@redhat.com>
40682         PR debug/71669
40683         * dwarf2out.c (add_data_member_location_attribute): For constant
40684         offset bitfield emit for -gdwarf-5 DW_AT_data_bit_offset attribute
40685         instead of DW_AT_data_member_location, DW_AT_bit_offset and
40686         DW_AT_byte_size attributes.
40688 2017-01-17  Eric Botcazou  <ebotcazou@adacore.com>
40690         * config/rs6000/rs6000.c (rs6000_emit_move): Also use a TOC reference
40691         after forcing to constant memory when the code model is medium.
40693 2017-01-17  Julia Koval  <julia.koval@intel.com>
40695         PR target/76731
40696         * config/i386/avx512fintrin.h
40697         (_mm512_i32gather_ps): Change __addr type to void const*.
40698         (_mm512_mask_i32gather_ps): Ditto.
40699         (_mm512_i32gather_pd): Ditto.
40700         (_mm512_mask_i32gather_pd): Ditto.
40701         (_mm512_i64gather_ps): Ditto.
40702         (_mm512_mask_i64gather_ps): Ditto.
40703         (_mm512_i64gather_pd): Ditto.
40704         (_mm512_mask_i64gather_pd): Ditto.
40705         (_mm512_i32gather_epi32): Ditto.
40706         (_mm512_mask_i32gather_epi32): Ditto.
40707         (_mm512_i32gather_epi64): Ditto.
40708         (_mm512_mask_i32gather_epi64): Ditto.
40709         (_mm512_i64gather_epi32): Ditto.
40710         (_mm512_mask_i64gather_epi32): Ditto.
40711         (_mm512_i64gather_epi64): Ditto.
40712         (_mm512_mask_i64gather_epi64): Ditto.
40713         (_mm512_i32scatter_ps): Change __addr type to void*.
40714         (_mm512_mask_i32scatter_ps): Ditto.
40715         (_mm512_i32scatter_pd): Ditto.
40716         (_mm512_mask_i32scatter_pd): Ditto.
40717         (_mm512_i64scatter_ps): Ditto.
40718         (_mm512_mask_i64scatter_ps): Ditto.
40719         (_mm512_i64scatter_pd): Ditto.
40720         (_mm512_mask_i64scatter_pd): Ditto.
40721         (_mm512_i32scatter_epi32): Ditto.
40722         (_mm512_mask_i32scatter_epi32): Ditto.
40723         (_mm512_i32scatter_epi64): Ditto.
40724         (_mm512_mask_i32scatter_epi64): Ditto.
40725         (_mm512_i64scatter_epi32): Ditto.
40726         (_mm512_mask_i64scatter_epi32): Ditto.
40727         (_mm512_i64scatter_epi64): Ditto.
40728         (_mm512_mask_i64scatter_epi64): Ditto.
40729         * config/i386/avx512pfintrin.h
40730         (_mm512_mask_prefetch_i32gather_pd): Change __addr type to void const*.
40731         (_mm512_mask_prefetch_i32gather_ps): Ditto.
40732         (_mm512_mask_prefetch_i64gather_pd): Ditto.
40733         (_mm512_mask_prefetch_i64gather_ps): Ditto.
40734         (_mm512_prefetch_i32scatter_pd): Change __addr type to void*.
40735         (_mm512_prefetch_i32scatter_ps): Ditto.
40736         (_mm512_mask_prefetch_i32scatter_pd): Ditto.
40737         (_mm512_mask_prefetch_i32scatter_ps): Ditto.
40738         (_mm512_prefetch_i64scatter_pd): Ditto.
40739         (_mm512_prefetch_i64scatter_ps): Ditto.
40740         (_mm512_mask_prefetch_i64scatter_pd): Ditto.
40741         (_mm512_mask_prefetch_i64scatter_ps): Ditto.
40742         * config/i386/avx512vlintrin.h
40743         (_mm256_mmask_i32gather_ps): Change __addr type to void const*.
40744         (_mm_mmask_i32gather_ps): Ditto.
40745         (_mm256_mmask_i32gather_pd): Ditto.
40746         (_mm_mmask_i32gather_pd): Ditto.
40747         (_mm256_mmask_i64gather_ps): Ditto.
40748         (_mm_mmask_i64gather_ps): Ditto.
40749         (_mm256_mmask_i64gather_pd): Ditto.
40750         (_mm_mmask_i64gather_pd): Ditto.
40751         (_mm256_mmask_i32gather_epi32): Ditto.
40752         (_mm_mmask_i32gather_epi32): Ditto.
40753         (_mm256_mmask_i32gather_epi64): Ditto.
40754         (_mm_mmask_i32gather_epi64): Ditto.
40755         (_mm256_mmask_i64gather_epi32): Ditto.
40756         (_mm_mmask_i64gather_epi32): Ditto.
40757         (_mm256_mmask_i64gather_epi64): Ditto.
40758         (_mm_mmask_i64gather_epi64): Ditto.
40759         (_mm256_i32scatter_ps): Change __addr type to void*.
40760         (_mm256_mask_i32scatter_ps): Ditto.
40761         (_mm_i32scatter_ps): Ditto.
40762         (_mm_mask_i32scatter_ps): Ditto.
40763         (_mm256_i32scatter_pd): Ditto.
40764         (_mm256_mask_i32scatter_pd): Ditto.
40765         (_mm_i32scatter_pd): Ditto.
40766         (_mm_mask_i32scatter_pd): Ditto.
40767         (_mm256_i64scatter_ps): Ditto.
40768         (_mm256_mask_i64scatter_ps): Ditto.
40769         (_mm_i64scatter_ps): Ditto.
40770         (_mm_mask_i64scatter_ps): Ditto.
40771         (_mm256_i64scatter_pd): Ditto.
40772         (_mm256_mask_i64scatter_pd): Ditto.
40773         (_mm_i64scatter_pd): Ditto.
40774         (_mm_mask_i64scatter_pd): Ditto.
40775         (_mm256_i32scatter_epi32): Ditto.
40776         (_mm256_mask_i32scatter_epi32): Ditto.
40777         (_mm_i32scatter_epi32): Ditto.
40778         (_mm_mask_i32scatter_epi32): Ditto.
40779         (_mm256_i32scatter_epi64): Ditto.
40780         (_mm256_mask_i32scatter_epi64): Ditto.
40781         (_mm_i32scatter_epi64): Ditto.
40782         (_mm_mask_i32scatter_epi64): Ditto.
40783         (_mm256_i64scatter_epi32): Ditto.
40784         (_mm256_mask_i64scatter_epi32): Ditto.
40785         (_mm_i64scatter_epi32): Ditto.
40786         (_mm_mask_i64scatter_epi32): Ditto.
40787         (_mm256_i64scatter_epi64): Ditto.
40788         (_mm256_mask_i64scatter_epi64): Ditto.
40789         (_mm_i64scatter_epi64): Ditto.
40790         (_mm_mask_i64scatter_epi64): Ditto.
40791         * config/i386/i386-builtin-types.def (V16SF_V16SF_PCFLOAT_V16SI_HI_INT)
40792         (V8DF_V8DF_PCDOUBLE_V8SI_QI_INT, V8SF_V8SF_PCFLOAT_V8DI_QI_INT)
40793         (V8DF_V8DF_PCDOUBLE_V8DI_QI_INT, V16SI_V16SI_PCINT_V16SI_HI_INT)
40794         (V8DI_V8DI_PCINT64_V8SI_QI_INT, V8SI_V8SI_PCINT_V8DI_QI_INT)
40795         (V8DI_V8DI_PCINT64_V8DI_QI_INT, V2DF_V2DF_PCDOUBLE_V4SI_QI_INT)
40796         (V4DF_V4DF_PCDOUBLE_V4SI_QI_INT, V2DF_V2DF_PCDOUBLE_V2DI_QI_INT)
40797         (V4DF_V4DF_PCDOUBLE_V4DI_QI_INT, V4SF_V4SF_PCFLOAT_V4SI_QI_INT)
40798         (V8SF_V8SF_PCFLOAT_V8SI_QI_INT, V4SF_V4SF_PCFLOAT_V2DI_QI_INT)
40799         (V4SF_V4SF_PCFLOAT_V4DI_QI_INT, V2DI_V2DI_PCINT64_V4SI_QI_INT)
40800         (V4DI_V4DI_PCINT64_V4SI_QI_INT, V2DI_V2DI_PCINT64_V2DI_QI_INT)
40801         (V4DI_V4DI_PCINT64_V4DI_QI_INT, V4SI_V4SI_PCINT_V4SI_QI_INT)
40802         (V8SI_V8SI_PCINT_V8SI_QI_INT, V4SI_V4SI_PCINT_V2DI_QI_INT)
40803         (V4SI_V4SI_PCINT_V4DI_QI_INT, VOID_PFLOAT_HI_V16SI_V16SF_INT)
40804         (VOID_PFLOAT_QI_V8SI_V8SF_INT, VOID_PFLOAT_QI_V4SI_V4SF_INT)
40805         (VOID_PDOUBLE_QI_V8SI_V8DF_INT, VOID_PDOUBLE_QI_V4SI_V4DF_INT)
40806         (VOID_PDOUBLE_QI_V4SI_V2DF_INT, VOID_PFLOAT_QI_V8DI_V8SF_INT)
40807         (VOID_PFLOAT_QI_V4DI_V4SF_INT, VOID_PFLOAT_QI_V2DI_V4SF_INT)
40808         (VOID_PDOUBLE_QI_V8DI_V8DF_INT, VOID_PDOUBLE_QI_V4DI_V4DF_INT)
40809         (VOID_PDOUBLE_QI_V2DI_V2DF_INT, VOID_PINT_HI_V16SI_V16SI_INT)
40810         (VOID_PINT_QI_V8SI_V8SI_INT, VOID_PINT_QI_V4SI_V4SI_INT)
40811         (VOID_PLONGLONG_QI_V8SI_V8DI_INT, VOID_PLONGLONG_QI_V4SI_V4DI_INT)
40812         (VOID_PLONGLONG_QI_V4SI_V2DI_INT, VOID_PINT_QI_V8DI_V8SI_INT)
40813         (VOID_PINT_QI_V4DI_V4SI_INT, VOID_PINT_QI_V2DI_V4SI_INT)
40814         (VOID_PLONGLONG_QI_V8DI_V8DI_INT, VOID_QI_V8SI_PCINT64_INT_INT)
40815         (VOID_PLONGLONG_QI_V4DI_V4DI_INT, VOID_PLONGLONG_QI_V2DI_V2DI_INT)
40816         (VOID_HI_V16SI_PCINT_INT_INT, VOID_QI_V8DI_PCINT64_INT_INT)
40817         (VOID_QI_V8DI_PCINT_INT_INT): Remove.
40818         (V16SF_V16SF_PCVOID_V16SI_HI_INT,  V8DF_V8DF_PCVOID_V8SI_QI_INT)
40819         (V8SF_V8SF_PCVOID_V8DI_QI_INT, V8DF_V8DF_PCVOID_V8DI_QI_INT)
40820         (V16SI_V16SI_PCVOID_V16SI_HI_INT, V8DI_V8DI_PCVOID_V8SI_QI_INT)
40821         (V8SI_V8SI_PCVOID_V8DI_QI_INT, V8DI_V8DI_PCVOID_V8DI_QI_INT)
40822         (VOID_PVOID_HI_V16SI_V16SF_INT, VOID_PVOID_QI_V8SI_V8DF_INT)
40823         (VOID_PVOID_QI_V8DI_V8SF_INT, VOID_PVOID_QI_V8DI_V8DF_INT)
40824         (VOID_PVOID_HI_V16SI_V16SI_INT, VOID_PVOID_QI_V8SI_V8DI_INT)
40825         (VOID_PVOID_QI_V8DI_V8SI_INT, VOID_PVOID_QI_V8DI_V8DI_INT)
40826         (V2DF_V2DF_PCVOID_V4SI_QI_INT, V4DF_V4DF_PCVOID_V4SI_QI_INT)
40827         (V2DF_V2DF_PCVOID_V2DI_QI_INT, V4DF_V4DF_PCVOID_V4DI_QI_INT
40828         (V4SF_V4SF_PCVOID_V4SI_QI_INT, V8SF_V8SF_PCVOID_V8SI_QI_INT)
40829         (V4SF_V4SF_PCVOID_V2DI_QI_INT, V4SF_V4SF_PCVOID_V4DI_QI_INT)
40830         (V2DI_V2DI_PCVOID_V4SI_QI_INT, V4DI_V4DI_PCVOID_V4SI_QI_INT)
40831         (V2DI_V2DI_PCVOID_V2DI_QI_INT, V4DI_V4DI_PCVOID_V4DI_QI_INT)
40832         (V4SI_V4SI_PCVOID_V4SI_QI_INT, V8SI_V8SI_PCVOID_V8SI_QI_INT)
40833         (V4SI_V4SI_PCVOID_V2DI_QI_INT, V4SI_V4SI_PCVOID_V4DI_QI_INT)
40834         (VOID_PVOID_QI_V8SI_V8SF_INT, VOID_PVOID_QI_V4SI_V4SF_INT)
40835         (VOID_PVOID_QI_V4SI_V4DF_INT, VOID_PVOID_QI_V4SI_V2DF_INT)
40836         (VOID_PVOID_QI_V4DI_V4SF_INT, VOID_PVOID_QI_V2DI_V4SF_INT)
40837         (VOID_PVOID_QI_V4DI_V4DF_INT, VOID_PVOID_QI_V2DI_V2DF_INT)
40838         (VOID_PVOID_QI_V8SI_V8SI_INT, VOID_PVOID_QI_V4SI_V4SI_INT)
40839         (VOID_PVOID_QI_V4SI_V4DI_INT, VOID_PVOID_QI_V4SI_V2DI_INT)
40840         (VOID_PVOID_QI_V4DI_V4SI_INT, VOID_PVOID_QI_V2DI_V4SI_INT)
40841         (VOID_PVOID_QI_V4DI_V4DI_INT, VOID_PVOID_QI_V2DI_V2DI_INT)
40842         (VOID_QI_V8SI_PCVOID_INT_INT, VOID_HI_V16SI_PCVOID_INT_INT)
40843         (VOID_QI_V8DI_PCVOID_INT_INT): Add.
40844         * config/i386/i386.c (ix86_init_mmx_sse_builtins): Adjust builtin
40845         definitions accordingly.
40847 2017-01-17  Kito Cheng  <kito.cheng@gmail.com>
40848             Kuan-Lin Chen  <kuanlinchentw@gmail.com>
40850         PR target/79079
40851         * internal-fn.c (expand_mul_overflow): Use convert_modes instead of
40852         gen_lowpart.
40854 2017-01-17  Vladimir Makarov  <vmakarov@redhat.com>
40856         PR target/79058
40857         * ira-conflicts.c (ira_build_conflicts): Update total conflict
40858         hard regs for inner regno.
40860 2017-01-17  Martin Liska  <mliska@suse.cz>
40862         PR ipa/71207
40863         * ipa-polymorphic-call.c (contains_type_p): Fix wrong
40864         assumption and add comment.
40866 2017-01-17  Nathan Sidwell  <nathan@acm.org>
40868         * ipa-visibility.c (localize_node): New function, broken out of ...
40869         (function_and_variable_visibility): ... here. Call it.
40871 2017-01-17  Jan Hubicka  <hubicka@ucw.cz>
40873         PR middle-end/77445
40874         * tree-ssa-threadupdate.c (remove_ctrl_stmt_and_useless_edges):
40875         correctly set frequency of oudgoing edge.
40876         (duplicate_thread_path): Fix profile updating.
40878 2017-01-17  Jakub Jelinek  <jakub@redhat.com>
40880         PR other/79046
40881         * configure.ac: Add GCC_BASE_VER.
40882         * Makefile.in (version): Use @get_gcc_base_ver@ instead of cat to get
40883         version from BASE-VER file.
40884         (CFLAGS-gcc.o): Add -DBASEVER=$(BASEVER_s).
40885         (gcc.o): Depend on $(BASEVER).
40886         * common.opt (dumpfullversion): New option.
40887         * gcc.c (driver_handle_option): Handle OPT_dumpfullversion.
40888         * doc/invoke.texi: Document -dumpfullversion.
40889         * doc/install.texi: Document --with-gcc-major-version-only.
40890         * configure: Regenerated.
40892 2017-01-17  Richard Biener  <rguenther@suse.de>
40894         PR tree-optimization/71433
40895         * tree-vrp.c (register_new_assert_for): Merge same asserts
40896         on all incoming edges.
40897         (process_assert_insertions_for): Handle insertions at the
40898         beginning of BBs.
40900 2017-01-17  Gerald Pfeifer  <gerald@pfeifer.com>
40902         * config/i386/cygwin.h (LIBGCJ_SONAME): Remove.
40903         * config/i386/mingw32.h (LIBGCJ_SONAME): Remove.
40905 2017-01-17  Kaz Kojima  <kkojima@gcc.gnu.org>
40907         PR target/78633
40908         * config/sh/sh.md (cmpeqsi_t+1): Call copy_rtx to avoid invalid
40909         RTL sharing.
40911 2017-01-17  Alan Modra  <amodra@gmail.com>
40913         PR target/79066
40914         * config/rs6000/rs6000.md (elf_high, elf_low): Disable when pic.
40915         * config/rs6000/rs6000.c (rs6000_emit_allocate_stack): Don't allow
40916         symbolic stack limit when pic.
40918 2017-01-16  Martin Sebor  <msebor@redhat.com>
40920         PR tree-optimization/78608
40921         * gimple-ssa-sprintf.c (tree_digits): Avoid negating TYPE_MIN.
40923 2017-01-16  Jeff Law  <law@redhat.com>
40925         Revert:
40926         2016-12-02  Tadek Kijkowski  <tkijkowski@gmail.com>
40927         * Makefile.in (PREPROCESSOR_DEFINES): Add a level of indirection
40928         for several include directories that may be relative to sysroot.
40929         * config/i386/x-mingw32 (gplus_includedir): Define.
40930         (gplus_tool_includedir, gplus_backward_include_dir): Likewise.
40931         (native_system_includedir): Likewise.
40932         * config/i386/mingw32.h (STANDARD_STARTFILE_PREFIX_1): Do not
40933         override if TARGET_SYSTEM_ROOT is defined.
40934         (NATIVE_SYSTEM_HEADER_DIR): Likewise.
40936         PR tree-optimization/79090
40937         PR tree-optimization/33562
40938         PR tree-optimization/61912
40939         PR tree-optimization/77485
40940         * tree-ssa-dse.c (compute_trims): Accept STMT argument.  Dump STMT
40941         and computed trims into the dump file.
40943 2017-01-17  Uros Bizjak  <ubizjak@gmail.com>
40945         * config/i386/i386.h (LIMIT_RELOAD_CLASS): Remove.
40947 2017-01-16  Jakub Jelinek  <jakub@redhat.com>
40949         PR c/79089
40950         * gimplify.c (gimplify_init_constructor): If want_value and
40951         object == lhs, unshare lhs to avoid invalid tree sharing.  Formatting
40952         fix.
40954         PR target/79080
40955         * loop-doloop.c (doloop_modify): Call unshare_all_rtl_in_chain on
40956         sequence.  Formatting fixes.
40957         (doloop_optimize): Formatting fixes.
40959         PR driver/49726
40960         * gcc.c (debug_level_greater_than_spec_func): New function.
40961         (static_spec_functions): Add debug-level-gt spec function.
40962         (ASM_DEBUG_SPEC, cpp_options): Use %:debug-level-gt(0) instead of
40963         !g0.
40964         * config/darwin.h (DSYMUTIL_SPEC, ASM_DEBUG_SPEC): Likewise.
40965         * config/darwin9.h (DSYMUTIL_SPEC, ASM_DEBUG_SPEC): Likewise.
40966         * common.opt (g, gcoff, gdwarf, gdwarf-, ggdb, gno-pubnames,
40967         gpubnames, ggnu-pubnames, gno-record-gcc-switches,
40968         grecord-gcc-switches, gno-strict-dwarf, gstrict-dwarf, gstabs,
40969         gstabs+, gtoggle, gvms, gxcoff, gxcoff+): Add Driver flag.
40971 2017-01-16  Uros Bizjak  <ubizjak@gmail.com>
40973         * config/i386/i386.h (HARD_REGNO_CALLER_SAVE_MODE): Apply HImode and
40974         QImode fixups to general and mask registers only.
40976 2017-01-16  Carl Love  <cel@us.ibm.com>
40978         * config/rs6000/rs6000-c (altivec_overloaded_builtins): Add support
40979         for built-in functions
40980         vector signed char vec_nabs (vector signed char)
40981         vector signed short vec_nabs (vector signed short)
40982         vector signed int vec_nabs (vector signed int)
40983         vector signed long long vec_nabs (vector signed long long)
40984         vector float vec_nabs (vector float)
40985         vector double vec_nabs (vector double)
40986         * config/rs6000/rs6000-builtin.def: Add definitions for NABS functions
40987         and NABS overload.
40988         * config/rs6000/altivec.md: New define_expand nabs<mode>2 types
40989         * config/rs6000/altivec.h: New define for vec_nabs built-in function.
40990         * doc/extend.texi: Update the documentation file for the new built-in
40991         functions.
40993 2017-01-16  Martin Sebor  <msebor@redhat.com>
40995         * gimple-ssa-sprintf.c (format_directive): Correct a typo in a warning
40996         message.
40998 2017-01-16  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
41000         * config/rs6000/rs6000.c (rtx_is_swappable_p): Change
41001         UNSPEC_VSX__XXSPLTD to require special splat handling.
41003 2017-01-16  David Malcolm  <dmalcolm@redhat.com>
41005         PR bootstrap/78616
41006         * system.h: Poison strndup.
41008 2017-01-16  Alan Modra  <amodra@gmail.com>
41010         PR target/79098
41011         * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Don't
41012         use a switch.
41014 2017-01-16  Georg-Johann Lay  <avr@gjlay.de>
41016         * config/avr/avr.h (BRANCH_COST) [reload_completed]: Increase by 4.
41018 2017-01-15  Uros Bizjak  <ubizjak@gmail.com>
41020         * config/i386/i386.c (ix86_legitimate_combined_insn): Do not
41021         call recog here.  Assert that INSN_CODE (insn) is non-negative.
41023 2017-01-15  Segher Boessenkool  <segher@kernel.crashing.org>
41025         PR target/72749
41026         * cfgrtl.c (rtl_split_edge): Also patch jump insns that jump to the
41027         fallthrough.
41028         * haifa-sched.c (dump_insn_stream): Don't crash if there is a label
41029         in the currently scheduled RTL fragment.
41031 2017-01-15  Segher Boessenkool  <segher@kernel.crashing.org>
41033         PR rtl-optimization/78751
41034         * ifcvt.c (find_cond_trap): If we generated a non-existing insn,
41035         give up.
41037 2017-01-14  Jeff Law  <law@redhat.com>
41039         PR tree-optimization/79090
41040         * tree-ssa-dse.c (valid_ao_ref_for_dse): Reject zero length and
41041         variable length stores.
41042         (compute_trims): Delete dead assignment to *trim_tail.
41043         (dse_dom_walker::dse_optimize_stmt): Optimize mem* calls with
41044         zero length.
41046 2017-01-14  Bernd Schmidt  <bschmidt@redhat.com>
41048         PR rtl-optimization/78626
41049         PR rtl-optimization/78727
41050         * cprop.c (one_cprop_pass): Collect unconditional traps in the middle
41051         of a block, and split such blocks after everything else is finished.
41053 2017-01-14  Alan Modra  <amodra@gmail.com>
41055         PR target/72749
41056         * combine.c (recog_for_combine_1): Set INSN_CODE before calling
41057         target legitimate_combined_insn.
41058         * config/rs6000/rs6000.c (TARGET_LEGITIMATE_COMBINED_INSN): Define.
41059         (rs6000_legitimate_combined_insn): New function.
41060         * config/rs6000/rs6000.md (UNSPEC_DOLOOP): Delete, and remove
41061         all uses.
41062         (ctr<mode>_internal3): Rename from *ctr<mode>_internal5.
41063         (ctr<mode>_internal4): Rename from *ctr<mode>_internal6.
41064         (ctr<mode>_internal1, ctr<mode>_internal2): Remove '*' from name.
41066 2017-01-14  Gerald Pfeifer  <gerald@pfeifer.com>
41068         * doc/frontends.texi (G++ and GCC): Remove references to Java.
41070 2017-01-13  Jeff Law  <law@redhat.com>
41072         PR tree-optimization/33562
41073         PR tree-optimization/61912
41074         PR tree-optimization/77485
41075         * tree-ssa-dse.c (delete_dead_call): Accept gsi rather than
41076         a statement.
41077         (delete_dead_assignment): Likewise.
41078         (dse_dom_walker::dse_optimize_stmt): Pass in the gsi rather than
41079         statement to delete_dead_call and delete_dead_assignment.
41081 2017-01-13  David Malcolm  <dmalcolm@redhat.com>
41083         PR c/78304
41084         * substring-locations.c (format_warning_va): Strengthen case 1 so
41085         that both endpoints of the substring must be within the format
41086         range for just the substring to be printed.
41088 2017-01-13  Uros Bizjak  <ubizjak@gmail.com>
41090         * config/i386/i386.opt (msgx): Use ix86_isa_flags2 variable.
41091         * config/i386/i386.c (ix86_target_string): Add missing options
41092         to isa_opts and reorder options by implied ISAs.  Rename isa_opts2 to
41093         isa2_opts, ix86_flag_opts to flag2_opts, ix86_target_other to
41094         flags_other and ix86_target_other to flags2_other.  Display unknown
41095         isa2 options.
41096         (ix86_valid_target_attribute_inner_p): Add missing options and
41097         reorder options by implied ISAs, as in ix86_target_string.
41099 2017-01-13  Richard Sandiford  <richard.sandiford@arm.com>
41101         * hash-table.h (hash_table::too_empty_p): New function.
41102         (hash_table::expand): Use it.
41103         (hash_table::traverse): Likewise.
41104         (hash_table::empty_slot): Use sizeof (value_type) instead of
41105         sizeof (PTR) to convert bytes to elements.  Shrink the table
41106         if the current size is excessive for the current number of
41107         elements.
41109 2017-01-13  Richard Sandiford  <richard.sandiford@arm.com>
41111         * ira-costs.c (record_reg_classes): Break from the inner loop
41112         early once alt_fail is known to be true.  Update outer loop
41113         handling accordingly.
41115 2017-01-13  Jeff Law  <law@redhat.com>
41117         * tree-ssa-dse.c (decrement_count): New function.
41118         (increment_start_addr, maybe_trim_memstar_call): Likewise.
41119         (dse_dom_walker::optimize_stmt): Call maybe_trim_memstar_call directly
41120         when we know the partially dead statement is a mem* function.
41122         PR tree-optimization/61912
41123         PR tree-optimization/77485
41124         * tree-ssa-dse.c: Include expr.h.
41125         (maybe_trim_constructor_store): New function.
41126         (maybe_trim_partially_dead_store): Call maybe_trim_constructor_store.
41128         PR tree-optimization/33562
41129         PR tree-optimization/61912
41130         PR tree-optimization/77485
41131         * doc/invoke.texi: Document new dse-max-object-size param.
41132         * params.def (PARM_DSE_MAX_OBJECT_SIZE): New PARAM.
41133         * tree-ssa-dse.c: Include params.h.
41134         (dse_store_status): New enum.
41135         (initialize_ao_ref_for_dse): New, partially extracted from
41136         dse_optimize_stmt.
41137         (valid_ao_ref_for_dse, normalize_ref): New.
41138         (setup_live_bytes_from_ref, compute_trims): Likewise.
41139         (clear_bytes_written_by, maybe_trim_complex_store): Likewise.
41140         (maybe_trim_partially_dead_store): Likewise.
41141         (maybe_trim_complex_store): Likewise.
41142         (dse_classify_store): Renamed from dse_possibly_dead_store_p.
41143         Track what bytes live from the original store.  Return tri-state
41144         for dead, partially dead or live.
41145         (dse_dom_walker): Add constructor, destructor and new private members.
41146         (delete_dead_call, delete_dead_assignment): New extracted from
41147         dse_optimize_stmt.
41148         (dse_optimize_stmt): Make a member of dse_dom_walker.
41149         Use initialize_ao_ref_for_dse.
41151         PR tree-optimization/33562
41152         PR tree-optimization/61912
41153         PR tree-optimization/77485
41154         * sbitmap.h (bitmap_count_bits): Prototype.
41155         (bitmap_clear_range, bitmap_set_range): Likewise.
41156         * sbitmap.c (bitmap_clear_range): New function.
41157         (bitmap_set_range, sbitmap_popcount, bitmap_count_bits): Likewise.
41159 2017-01-13  Martin Liska  <mliska@suse.cz>
41161         PR ipa/79043
41162         * function.c (set_cfun): Add new argument force.
41163         * function.h (set_cfun): Likewise.
41164         * ipa-inline-transform.c (inline_call): Use the function when
41165         strict alising from is dropped for function we inline to.
41167 2017-01-13  Richard Biener  <rguenther@suse.de>
41169         * tree-pretty-print.c (dump_generic_node): Fix inverted condition
41170         for dumping GIMPLE INTEGER_CSTs.
41172 2017-01-13  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
41174         * config/sol2.h (TARGET_OS_CPP_BUILTINS): Define __STDC_VERSION__
41175         to 201112L since C++17.
41177 2017-01-13  Maxim Ostapenko  <m.ostapenko@samsung.com>
41179         PR sanitizer/78887
41180         * asan.c (asan_needs_odr_indicator_p): Don't emit ODR indicators
41181         if -fsanitize=kernel-address is present.
41183 2017-01-13  Richard Biener  <rguenther@suse.de>
41185         * tree-pretty-print.c (dump_generic_node): Dump INTEGER_CSTs
41186         as _Literal ( type ) number in case usual suffixes do not
41187         preserve all information.
41189 2017-01-13  Richard Biener  <rguenther@suse.de>
41191         PR tree-optimization/77283
41192         * gimple-ssa-split-paths.c: Include gimple-ssa.h, tree-phinodes.h
41193         and ssa-iterators.h.
41194         (is_feasible_trace): Implement a cost model based on joiner
41195         PHI node uses.
41197 2017-01-12  Michael Meissner  <meissner@linux.vnet.ibm.com>
41199         PR target/79004
41200         * config/rs6000/rs6000.md (FP_ISA3): Do not optimize converting
41201         char or short to __float128/_Float128 directly.
41203 2017-01-12  Martin Sebor  <msebor@redhat.com>
41205         to -Wformat-overflow.
41206         * gimple-ssa-sprintf.c (pass_sprintf_length::gate): Adjust.
41207         (min_bytes_remaining): Same.
41208         (get_string_length): Same.
41209         (format_string): Same.
41210         (format_directive): Same.
41211         (add_bytes): Same.
41212         (pass_sprintf_length::handle_gimple_call): Same.
41214 2017-01-12  Jakub Jelinek  <jakub@redhat.com>
41216         * gimple-ssa-sprintf.c (try_substitute_return_value): Remove
41217         info.nowrite calls with no lhs that can't throw.  Return bool
41218         whether gsi_remove has been called or not.
41219         (pass_sprintf_length::handle_gimple_call): Return bool whether
41220         try_substitute_return_value called gsi_remove.  Formatting fix.
41221         (pass_sprintf_length::execute): Don't use gsi_remove if
41222         handle_gimple_call returned true.
41224         PR bootstrap/79069
41225         * cfgrtl.c (rtl_tidy_fallthru_edge): For any_uncondjump_p that can't
41226         be removed due to side-effects, don't remove following barrier nor
41227         turn the successor edge into fallthru edge.
41229 2017-01-12  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
41231         PR target/79044
41232         * config/rs6000/rs6000.c (insn_is_swappable_p): Mark
41233         element-reversing loads and stores as not swappable.
41235 2017-01-12  Nathan Sidwell  <nathan@acm.org>
41236             Nicolai Stange  <nicstange@gmail.com>
41238         * combine.c (try_combine): Don't ignore result of overlap checking
41239         loop.  Combine overlap & asm check into single loop.
41241 2017-01-12  Richard Biener  <rguenther@suse.de>
41243         * tree-pretty-print.c (dump_generic_node): Provide -gimple
41244         variant for MEM_REF.  Sanitize INTEGER_CST for -gimple.
41246 2017-01-12  Richard Biener  <rguenther@suse.de>
41248         * tree.c (initialize_tree_contains_struct): Make TS_OPTIMIZATION
41249         and TS_TARGET_OPTION directly derive from TS_BASE.
41250         * tree-core.h (tree_optimization_option): Derive from tree_base.
41251         (tree_target_option): Likewise.
41253 2017-01-11  Uros Bizjak  <ubizjak@gmail.com>
41255         * config/i386/i386.c (memory_address_length): Increase len
41256         only when rip_relative_addr_p returns false.
41258 2017-01-11  Julia Koval  <julia.koval@intel.com>
41260         * common/config/i386/i386-common.c (OPTION_MASK_ISA_SGX_UNSET): New.
41261         (OPTION_MASK_ISA_SGX_SET): New.
41262         (ix86_handle_option): Handle OPT_msgx.
41263         * config.gcc: Added sgxintrin.h.
41264         * config/i386/driver-i386.c (host_detect_local_cpu): Detect sgx.
41265         * config/i386/i386-c.c (ix86_target_macros_internal): Define __SGX__.
41266         * config/i386/i386.c (ix86_target_string): Add -msgx.
41267         (PTA_SGX): New.
41268         (ix86_option_override_internal): Handle new options.
41269         (ix86_valid_target_attribute_inner_p): Add sgx.
41270         * config/i386/i386.h (TARGET_SGX, TARGET_SGX_P): New.
41271         * config/i386/i386.opt: Add msgx.
41272         * config/i386/sgxintrin.h: New file.
41273         * config/i386/x86intrin.h: Add sgxintrin.h.
41275 2017-01-11  Jakub Jelinek  <jakub@redhat.com>
41277         PR c++/71537
41278         * fold-const.c (maybe_nonzero_address): Return 1 for function
41279         local objects.
41280         (tree_single_nonzero_warnv_p): Don't handle function local objects
41281         here.
41283         PR c++/72813
41284         * gcc.c (default_compilers): Don't add -o %g.s for -S -save-temps
41285         of c-header.
41287 2017-01-11  David Malcolm  <dmalcolm@redhat.com>
41289         PR driver/78877
41290         * opts.c: Include "spellcheck.h"
41291         (struct string_fragment): New struct.
41292         (struct edit_distance_traits<const string_fragment &>): New
41293         struct.
41294         (get_closest_sanitizer_option): New function.
41295         (parse_sanitizer_options): Offer suggestions for unrecognized arguments.
41297 2017-01-11  Jakub Jelinek  <jakub@redhat.com>
41299         * dwarf2out.c (DWARF_COMPILE_UNIT_HEADER_SIZE): For DWARF5 decrease
41300         by 12.
41301         (DWARF_COMDAT_TYPE_UNIT_HEADER_SIZE): Always
41302         DWARF_COMPILE_UNIT_HEADER_SIZE plus 12.
41303         (DWARF_COMPILE_UNIT_SKELETON_HEADER_SIZE): Define.
41304         (calc_base_type_die_sizes): Use DWARF_COMPILE_UNIT_SKELETON_HEADER_SIZE
41305         for initial die_offset if dwarf_split_debug_info.
41306         (output_comp_unit): Use DWARF_COMPILE_UNIT_SKELETON_HEADER_SIZE for
41307         initial next_die_offset if dwo_id is non-NULL.  Don't emit padding
41308         fields.
41309         (output_skeleton_debug_sections): Formatting fix.  Use
41310         DWARF_COMPILE_UNIT_SKELETON_HEADER_SIZE instead of
41311         DWARF_COMPILE_UNIT_HEADER_SIZE.  Don't emit padding.
41313 2017-01-11  Wilco Dijkstra  <wdijkstr@arm.com>
41315         * config/arm/cortex-a53.md: Add bypasses for
41316         cortex_a53_r2f_cvt.
41317         (cortex_a53_r2f): Only use for transfers.
41318         (cortex_a53_f2r): Likewise.
41319         (cortex_a53_r2f_cvt): Add reservation for conversions.
41320         (cortex_a53_f2r_cvt): Likewise.
41322 2017-01-11  Tamar Christina  <tamar.christina@arm.com>
41324         * config/arm/arm_neon.h: Add __artificial__ and gnu_inline
41325         to all inlined functions, change static to extern.
41327 2017-01-11  Christophe Lyon  <christophe.lyon@linaro.org>
41329         PR target/78253
41330         * config/arm/arm.c (legitimize_pic_address): Handle reference to
41331         weak symbol.
41332         (arm_assemble_integer): Likewise.
41334 2017-01-11  Richard Earnshaw  <rearnsha@arm.com>
41336         * config.gcc: Use new awk script to check CPU, FPU and architecture
41337         parameters for --with-... options.
41338         * config/arm/parsecpu.awk: New file
41339         * config/arm/arm-cpus.in: New file.
41340         * config/arm/arm-opts.h: Include arm-cpu.h instead of processing .def
41341         files.
41342         * config/arm/arm.c: Include arm-cpu-data.h instead of processing .def
41343         files.
41344         * config/arm/t-arm: Update dependency rules.
41345         * common/config/arm/arm-common.c: Include arm-cpu-cdata.h instead
41346         of processing .def files.
41347         * config/arm/genopt.sh: Deleted.
41348         * config/arm/gentune.sh: Deleted.
41349         * config/arm/arm-cores.def: Deleted.
41350         * config/arm/arm-arches.def: Deleted.
41351         * config/arm/arm-fpus.def: Deleted.
41352         * config/arm/arm-tune.md: Regenerated.
41353         * config/arm/arm-tables.opt: Regenerated.
41354         * config/arm/arm-cpu.h: New generated file.
41355         * config/arm/arm-cpu-data.h: New generated file.
41356         * config/arm/arm-cpu-cdata.h: New generated file.
41358 2017-01-11  Maxim Ostapenko  <m.ostapenko@samsung.com>
41360         PR lto/79042
41361         * lto-cgraph.c (lto_output_varpool_node): Pack dynamically_initialized
41362         bit.
41363         (input_varpool_node): Unpack dynamically_initialized bit.
41365 2017-01-11  Eric Botcazou  <ebotcazou@adacore.com>
41367         PR rtl-optimization/79032
41368         * lra-constraints.c (simplify_operand_subreg): In the MEM case, test
41369         the alignment of the adjusted memory reference against that of MODE,
41370         instead of the alignment of the original memory reference.
41372 2017-01-11  Martin Jambor  <mjambor@suse.cz>
41374         * hsa.c (hsa_callable_function_p): Revert addition of DECL_ARTIFICIAL
41375         test.
41376         * ipa-hsa.c (process_hsa_functions): Only duplicate non-artificial
41377         decorated functions.
41379 2017-01-11  Richard Biener  <rguenther@suse.de>
41381         * tree-vrp.c (evrp_dom_walker::before_dom_children): Also
41382         set range/nonnull info for PHI results.  Do not set it on
41383         stmts marked for removal.
41385 2017-01-10  Eric Botcazou  <ebotcazou@adacore.com>
41387         * expr.c (store_field): In the bitfield case, fetch the return value
41388         from the registers before applying a single big-endian adjustment.
41389         Always do a final load for a BLKmode value not larger than a word.
41391 2017-01-10  David Malcolm  <dmalcolm@redhat.com>
41393         PR c++/77949
41394         * input.c (selftest::test_accessing_ordinary_linemaps): Verify
41395         that we correctly handle column numbers greater than
41396         LINE_MAP_MAX_COLUMN_NUMBER.
41398 2017-01-10  Martin Sebor  <msebor@redhat.com>
41400         PR middle-end/78245
41401         * gimple-ssa-sprintf.c (get_destination_size): Call
41402         {init,fini}object_sizes.
41403         * tree-object-size.c (addr_object_size): Adjust.
41404         (pass_through_call): Adjust.
41405         (pass_object_sizes::execute): Adjust.
41406         * tree-object-size.h (fini_object_sizes): Declare.
41408 2017-01-10  Martin Sebor  <msebor@redhat.com>
41410         PR tree-optimization/78775
41411         * builtins.c (get_size_range): Move...
41412         * calls.c: ...to here.
41413         (alloc_max_size): Accept zero argument.
41414         (operand_signed_p): Remove.
41415         (maybe_warn_alloc_args_overflow): Call get_size_range.
41416         * calls.h (get_size_range): Declare.
41418 2017-01-10  Joe Seymour  <joe.s@somniumtech.com>
41420         * config/msp430/driver-msp430.c (msp430_mcu_data): Sync with data
41421         from TI's devices.csv file as of September 2016.
41422         * config/msp430/msp430.c (msp430_mcu_data): Likewise.
41424 2017-01-10  Sandra Loosemore  <sandra@codesourcery.com>
41426         * doc/extend.texi: Tweak formatting to fix overfull hbox warnings.
41427         * doc/invoke.texi: Likewise.
41428         * doc/md.texi: Likewise.
41429         * doc/objc.texi: Likewise.
41431 2017-01-10  Joshua Conner  <joshconner@google.com>
41433         * config/arm/fuchsia-elf.h: New file.
41434         * config/fuchsia.h: New file.
41435         * config.gcc (*-*-fuchsia*): Set native_system_header_dir.
41436         (aarch64*-*-fuchsia*, arm*-*-fuchsia*, x86_64-*-fuchsia*): Add to
41437         targets.
41438         * config.host: (aarch64*-*-fuchsia*, arm*-*-fuchsia*): Add to hosts.
41440 2016-01-10  Richard Biener  <rguenther@suse.de>
41442         PR tree-optimization/79034
41443         * tree-call-cdce.c (shrink_wrap_one_built_in_call_with_conds):
41444         Propagate out degenerate PHIs in the joiner.
41446 2017-01-10  Martin Liska  <mliska@suse.cz>
41448         * ipa-icf.c (sort_sem_items_by_decl_uid): New function.
41449         (sort_congruence_classes_by_decl_uid): Likewise.
41450         (sort_congruence_class_groups_by_decl_uid): Likewise.
41451         (sem_item_optimizer::merge_classes): Sort class, groups in these
41452         classes and members in the groups by DECL_UID of declarations.
41453         This would make merge operations stable.
41455 2017-01-10  Martin Liska  <mliska@suse.cz>
41457         * ipa-icf.c (sem_item_optimizer::sem_item_optimizer): Remove
41458         usage of m_classes_vec.
41459         (sem_item_optimizer::~sem_item_optimizer):  Likewise.
41460         (sem_item_optimizer::get_group_by_hash): Likewise.
41461         (sem_item_optimizer::subdivide_classes_by_equality): Likewise.
41462         (sem_item_optimizer::subdivide_classes_by_sensitive_refs): Likewise.
41463         (sem_item_optimizer::verify_classes): Likewise.
41464         (sem_item_optimizer::process_cong_reduction): Likewise.
41465         (sem_item_optimizer::dump_cong_classes): Likewise.
41466         (sem_item_optimizer::merge_classes): Likewise.
41467         * ipa-icf.h (congruence_class_hash): Rename from
41468         congruence_class_group_hash.  Remove declaration of m_classes_vec.
41470 2017-01-10  Andrew Senkevich  <andrew.senkevich@intel.com>
41472         * common/config/i386/i386-common.c (OPTION_MASK_ISA_AVX512VPOPCNTDQ_SET,
41473         OPTION_MASK_ISA_AVX512VPOPCNTDQ_UNSET): New.
41474         * config.gcc: Add avx512vpopcntdqintrin.h.
41475         * config/i386/avx512vpopcntdqintrin.h: New.
41476         * config/i386/cpuid.h (bit_AVX512VPOPCNTDQ): New.
41477         * config/i386/i386-builtin-types.def: Add new types.
41478         * config/i386/i386-builtin.def (__builtin_ia32_vpopcountd_v16si,
41479         __builtin_ia32_vpopcountd_v16si_mask, __builtin_ia32_vpopcountq_v8di,
41480         __builtin_ia32_vpopcountq_v8di_mask): New.
41481         * config/i386/i386-c.c (ix86_target_macros_internal): Define
41482         __AVX512VPOPCNTDQ__.
41483         * config/i386/i386.c (ix86_target_string): Add -mavx512vpopcntdq.
41484         (PTA_AVX512VPOPCNTDQ): Define.
41485         * config/i386/i386.h (TARGET_AVX512VPOPCNTDQ,
41486         TARGET_AVX512VPOPCNTDQ_P): Define.
41487         * config/i386/i386.opt: Add mavx512vpopcntdq.
41488         * config/i386/immintrin.h: Include avx512vpopcntdqintrin.h.
41489         * config/i386/sse.md (define_insn "vpopcount<mode><mask_name>"): New.
41491 2017-01-01  Jan Hubicka  <hubicka@ucw.cz>
41493         PR middle-end/77484
41494         * predict.def (PRED_CALL): Set to 67.
41496 2017-01-09  Eric Botcazou  <ebotcazou@adacore.com>
41498         * expr.c (store_field): In the bitfield case, if the value comes from
41499         a function call and is of an aggregate type returned in registers, do
41500         not modify the field mode; extract the value in all cases if the mode
41501         is BLKmode and the size is not larger than a word.
41503 2017-01-09  Dominique d'Humieres  <dominiq@lps.ens.fr>
41505         PR target/71017
41506         * config/i386/cpuid.h: Fix undefined behavior.
41508 2017-01-04  Jeff Law  <law@redhat.com>
41510         PR tree-optimization/79007
41511         PR tree-optimization/67955
41512         * tree-ssa-alias.c (same_addr_size_stores_p): Only need to be
41513         conservative for pt.null when flag_non_call_exceptions is on.
41515 2017-01-09  Jakub Jelinek  <jakub@redhat.com>
41517         PR translation/79019
41518         PR translation/79020
41519         * params.def (PARAM_INLINE_MIN_SPEEDUP,
41520         PARAM_IPA_CP_SINGLE_CALL_PENALTY,
41521         PARAM_USE_AFTER_SCOPE_DIRECT_EMISSION_THRESHOLD): Fix typos
41522         in descriptions.
41523         * config/avr/avr.opt (maccumulate-args): Likewise.
41524         * config/msp430/msp430.opt (mwarn-mcu): Likewise.
41525         * common.opt (freport-bug): Likewise.
41526         * cif-code.def (CIF_FINAL_ERROR): Likewise.
41527         * doc/invoke.texi (ipa-cp-single-call-penalty): Likewise.
41528         * config/s390/s390.c (s390_invalid_binary_op): Fix spelling in
41529         translatable string.
41530         * config/i386/i386.c (function_value_32): Likewise.
41531         * config/nios2/nios2.c (nios2_valid_target_attribute_rec): Likewise.
41532         * config/msp430/msp430.c (msp430_option_override, msp430_attr):
41533         Likewise.
41534         * config/msp430/driver-msp430.c (msp430_select_hwmult_lib): Likewise.
41535         * common/config/msp430/msp430-common.c (msp430_handle_option):
41536         Likewise.
41537         * symtab.c (symtab_node::verify_base): Likewise.
41538         * opts.c (set_debug_level): Likewise.
41539         * tree.c (verify_type_variant): Likewise.  Fix typo in comment.
41540         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Add
41541         missing whitespace to translatable strings.
41542         * config/avr/avr.md (bswapsi2): Fix typo in comment.
41543         * config/sh/superh.h: Likewise.
41544         * config/i386/xopintrin.h: Likewise.
41545         * config/i386/znver1.md: Likewise.
41546         * config/rs6000/rs6000.c (struct rs6000_opt_mask): Likewise.
41547         * ipa-inline-analysis.c (compute_inline_parameters): Likewise.
41548         * double-int.h (struct double_int): Likewise.
41549         * double-int.c (div_and_round_double): Likewise.
41550         * wide-int.cc: Likewise.
41551         * tree-ssa.c (non_rewritable_mem_ref_base): Likewise.
41552         * tree-ssa-sccvn.c (vn_reference_lookup_3): Likewise.
41553         * cfgcleanup.c (crossjumps_occured): Renamed to ...
41554         (crossjumps_occurred): ... this.
41555         (try_crossjump_bb, try_head_merge_bb, try_optimize_cfg, cleanup_cfg):
41556         Adjust all uses.
41558         PR tree-optimization/78899
41559         * tree-if-conv.c (version_loop_for_if_conversion): Instead of
41560         returning bool return struct loop *, NULL for failure and the new
41561         loop on success.
41562         (versionable_outer_loop_p): Don't version outer loop if it has
41563         dont_vectorized bit set.
41564         (tree_if_conversion): When versioning outer loop, ensure
41565         tree_if_conversion is performed also on the inner loop of the
41566         non-vectorizable outer loop copy.
41567         * tree-vectorizer.c (set_uid_loop_bbs): Formatting fix.  Fold
41568         LOOP_VECTORIZED in inner loop of the scalar outer loop and
41569         prevent vectorization of it.
41570         (vectorize_loops): For outer + inner LOOP_VECTORIZED, ensure
41571         the outer loop vectorization of the non-scalar version is attempted
41572         before vectorization of the inner loop in scalar version.  If
41573         outer LOOP_VECTORIZED guarded loop is not vectorized, prevent
41574         vectorization of its inner loop.
41575         * tree-vect-loop-manip.c (rename_variables_in_bb): If outer_loop
41576         has 2 inner loops, rename also on edges from bb whose single pred
41577         is outer_loop->header.  Fix typo in function comment.
41579 2017-01-09  Martin Sebor  <msebor@redhat.com>
41581         PR bootstrap/79033
41582         * asan.c (asan_emit_stack_protection): Increase local buffer size
41583         to avoid snprintf truncation warning.
41585 2017-01-09  Andrew Pinski  <apinski@cavium.com>
41587         * config/aarch64/aarch64-cores.def: Add thunderx2t99.  Change vulcan
41588         to reference thunderx2t99 for the tuning structure
41589         * config/aarch64/aarch64-cost-tables.h (vulcan_extra_costs):
41590         Rename to ...
41591         (thunderx2t99_extra_costs): This.
41592         * config/aarch64/aarch64-tune.md: Regenerate.
41593         * config/aarch64/aarch64.c (vulcan_addrcost_table): Rename to ...
41594         (vulcan_addrcost_table): This.
41595         (vulcan_regmove_cost): Rename to ...
41596         (thunderx2t99_regmove_cost): This.
41597         (vulcan_vector_cost): Rename to ...
41598         (thunderx2t99_vector_cost): this.
41599         (vulcan_branch_cost): Rename to ...
41600         (thunderx2t99_branch_cost): This.
41601         (vulcan_tunings): Rename to ...
41602         (thunderx2t99_tunings): This and s/vulcan/thunderx2t99 .
41603         * doc/invoke.texi (AARCH64/mtune): Add thunderx2t99.
41605 2017-01-09  Martin Jambor  <mjambor@suse.cz>
41607         PR ipa/78365
41608         PR ipa/78599
41609         * ipa-prop.h (ipa_jump_func): Swap positions of vr_known and m_vr.
41610         * ipa-cp.c (ipa_vr_operation_and_type_effects): New function.
41611         (propagate_vr_accross_jump_function): Use the above function for all
41612         value range computations for pass-through jump functions and type
41613         converasion from explicit value range values.
41614         (ipcp_propagate_stage): Do not attempt to deduce types of formal
41615         parameters from TYPE_ARG_TYPES.
41616         * ipa-prop.c (ipa_write_jump_function): Remove trailing whitespace.
41617         (ipa_write_node_info): Stream type of the actual argument.
41618         (ipa_read_node_info): Likewise. Also remove trailing whitespace.
41620 2017-01-09  Martin Liska  <mliska@suse.cz>
41622         PR pch/78970
41623         * gcc.c (driver_handle_option): Handle OPT_E and set have_E.
41624         (lookup_compiler): Do not show error message with have_E.
41626 2017-01-09  Jakub Jelinek  <jakub@redhat.com>
41628         PR tree-optimization/78938
41629         * tree-vect-stmts.c (vectorizable_condition): For non-masked COND_EXPR
41630         where comp_vectype is VECTOR_BOOLEAN_TYPE_P, use
41631         BIT_{NOT,XOR,AND,IOR}_EXPR on the comparison operands instead of
41632         {EQ,NE,GE,GT,LE,LT}_EXPR directly inside of VEC_COND_EXPR.  Formatting
41633         fixes.
41635 2017-01-09  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
41637         * tree-ssa-address.c (gen_addr_rtx): Don't handle index if it
41638         is const0_rtx.
41640 2017-01-09  Richard Biener  <rguenther@suse.de>
41642         PR tree-optimization/78997
41643         * tree-vect-slp.c (vect_mask_constant_operand_p): Handle SSA
41644         name condition properly.
41646 2017-01-09  Richard Biener  <rguenther@suse.de>
41648         PR debug/79000
41649         * dwarf2out.c (is_cxx): New overload with context.
41650         (is_naming_typedef_decl): Use it.
41652 2017-01-08  Sandra Loosemore  <sandra@codesourcery.com>
41654         * invoke.texi (Option Summary): Correct spacing in option lists
41655         and add line breaks to fix over-long lines.
41657 2017-01-08  Sandra Loosemore  <sandra@codesourcery.com>
41659         PR middle-end/17660
41661         * extend.texi (Common Variable Attributes): Add xref to GCC
41662         Internals manual to explain mode attribute keywords.
41664 2017-01-08  Sandra Loosemore  <sandra@codesourcery.com>
41666         PR other/16519
41667         * doc/invoke.texi (Option Summary): Move -pthread to Linker Options
41668         and Preprocessor Options.
41669         (Options for Linking): Document -pthread here....
41670         (RS/6000 and PowerPC Options): ...not here.
41671         (Solaris 2 Options): ...or here.
41672         * doc/cppopts.texi: Document -pthread.
41674 2017-01-08  Martin Sebor  <msebor@redhat.com>
41676         PR middle-end/77708
41677         * doc/invoke.texi (Warning Options): Document -Wformat-truncation.
41678         * gimple-ssa-sprintf.c (call_info::reval_used, call_info::warnopt):
41679         New member functions.
41680         (format_directive): Used them.
41681         (add_bytes): Same.
41682         (pass_sprintf_length::handle_gimple_call): Same.
41683         * graphite-sese-to-poly.c (tree_int_to_gmp): Increase buffer size
41684         to avoid truncation for any argument.
41685         (extract_affine_mul): Same.
41686         * tree.c (get_file_function_name): Same.
41688 2017-01-01  Jan Hubicka  <hubicka@ucw.cz>
41690         PR middle-end/77484
41691         * predict.def (PRED_INDIR_CALL): Set to 86.
41693 2017-01-07  Sandra Loosemore  <sandra@codesourcery.com>
41695         PR preprocessor/54124
41696         * doc/cppopts.texi: Reformat -d subtable to list the full name
41697         of the options.  Add cross-reference to the docs for the general
41698         compiler -d options.
41699         * doc/invoke.texi (Developer Options): Add cross-reference to the
41700         preprocessor-specific -d option documentation.
41702 2017-01-07  Sandra Loosemore  <sandra@codesourcery.com>
41704         PR preprocessor/13498
41705         * doc/cpp.texi (Search Path): Rewrite to remove obsolete and
41706         redudant material, and reflect new command-line options.
41707         (System Headers): Likewise.
41709 2017-01-07  Sandra Loosemore  <sandra@codesourcery.com>
41711         * doc/cppdiropts.texi: Merge documentation of -I, -iquote,
41712         -isystem, and -idirafter.  Copy-edit.
41713         * doc/cppopts.texi: Copy-edit.  Remove contradiction about
41714         default for -ftrack-macro-expansion.  Delete obsolete and
41715         badly-formatted implementation details about -fdebug-cpp output.
41716         * doc/cppwarnopts.texi: Copy-edit.
41718 2017-01-07  David Malcolm  <dmalcolm@redhat.com>
41720         PR c++/72803
41721         * input.c (selftest::test_accessing_ordinary_linemaps): Verify
41722         that the transition from a max line width >= 1<<10 to narrower
41723         lines works correctly.
41725 2017-01-07  Alexandre Oliva  <aoliva@redhat.com>
41727         * doc/options.texi (PerFunction): New.
41728         * opt-functions.awk (switch_flags): Map both Optimization and
41729         PerFunction to CL_OPTIMIZATION.
41730         * opth-gen.awk: Test for PerFunction flag along with
41731         Optimization.
41732         * optc-save-gen.awk: Likewise.  Introduce var_opt_hash and set
41733         it only when the latter is present.  Skip those that don't in
41734         the hash function generator.
41735         * common.opt (fvar-tracking): Mark as PerFunction instead of
41736         Optimization.
41737         (fvar-tracking-assignments): Likewise.
41738         (fvar-tracking-assignments-toggle): Likewise.
41739         (fvar-tracking-uninit): Likewise.
41741 2017-01-07  Jakub Jelinek  <jakub@redhat.com>
41743         PR translation/79018
41744         * params.def (PARAM_MAX_STORES_TO_MERGE): Add missing space between
41745         the and store.
41747 2017-01-06  Mikael Pettersson  <mikpelinux@gmail.com>
41749         PR target/57583
41750         * config/m68k/m68k.opt (LONG_JUMP_TABLE_OFFSETS): New option.
41751         * config/m68k/linux.h (ASM_RETURN_CASE_JUMP): Handle
41752         TARGET_LONG_JUMP_TABLE_OFFSETS.
41753         * config/m68k/m68kelf.h (ASM_RETURN_CASE_JUMP): Likewise.
41754         * config/m68k/netbsd-elf.h (ASM_RETURN_CASE_JUMP): Likewise.
41755         * config/m68k/m68k.h (CASE_VECTOR_MODE): Likewise.
41756         (ASM_OUTPUT_ADDR_DIFF_ELF): Likewise.
41757         * config/m68k/m68k.md (tablejump expander): Likewise.
41758         (*tablejump_pcrel_hi): Renamed from unnamed insn, reject
41759         TARGET_LONG_JUMP_TABLE_OFFSETS.
41760         (*tablejump_pcrel_si): New insn, handle TARGET_LONG_JUMP_TABLE_OFFSETS.
41761         * doc/invoke.texi (M68K options): Add -mlong-jump-table-offsets.
41763 2017-01-06  Edgar E. Iglesias <edgar.iglesias@xilinx.com>
41764             David Holsgrove <david.holsgrove@xilinx.com>
41766         * common/config/microblaze/microblaze-common.c
41767         (TARGET_EXCEPT_UNWIND_INFO): Remove.
41768         * config/microblaze/microblaze-protos.h (microblaze_eh_return):
41769         New prototype.
41770         * config/microblaze/microblaze.c (microblaze_must_save_register)
41771         (microblaze_expand_epilogue, microblaze_return_addr): Handle
41772         calls_eh_return.
41773         (microblaze_eh_return): New function.
41774         * config/microblaze/microblaze.h (RETURN_ADDR_OFFSET)
41775         (EH_RETURN_DATA_REGNO, MB_EH_STACKADJ_REGNUM)
41776         (EH_RETURN_STACKADJ_RTX, ASM_PREFERRED_EH_DATA_FORMAT): New macros.
41777         * config/microblaze/microblaze.md (eh_return): New pattern.
41779 2017-01-06  Jakub Jelinek  <jakub@redhat.com>
41781         * system.h (GCC_DIAGNOSTIC_PUSH_IGNORED, GCC_DIAGNOSTIC_POP,
41782         GCC_DIAGNOSTIC_STRINGIFY): Define.
41784         * read-rtl.c (rtx_reader::read_rtx_code): Avoid -Wsign-compare warning.
41786 2017-01-06  Andre Vieira  <andre.simoesdiasvieira@arm.com>
41788         * config/arm/arm.md (<mcrr>): New.
41789         (<mrrc>): New.
41790         * config/arm/arm.c (arm_arch5te): New.
41791         (arm_option_override): Set arm_arch5te.
41792         (arm_coproc_builtin_available): Add support for mcrr, mcrr2, mrrc
41793         and mrrc2.
41794         * config/arm/arm-builtins.c (MCRR_QUALIFIERS): Define to...
41795         (arm_mcrr_qualifiers): ... this. New.
41796         (MRRC_QUALIFIERS): Define to...
41797         (arm_mrrc_qualifiers): ... this. New.
41798         * config/arm/arm_acle.h (__arm_mcrr, __arm_mcrr2, __arm_mrrc,
41799         __arm_mrrc2): New.
41800         * config/arm/arm_acle_builtins.def (mcrr, mcrr2, mrrc, mrrc2): New.
41801         * config/arm/iterators.md (MCRRI, mcrr, MCRR): New.
41802         (MRRCI, mrrc, MRRC): New.
41803         * config/arm/unspecs.md (VUNSPEC_MCRR, VUNSPEC_MCRR2, VUNSPEC_MRRC,
41804         VUNSPEC_MRRC2): New.
41806 2017-01-06  Andre Vieira  <andre.simoesdiasvieira@arm.com>
41808         * config/arm/arm.md (<mcr>): New.
41809         (<mrc>): New.
41810         * config/arm/arm.c (arm_coproc_builtin_available): Add
41811         support for mcr, mrc, mcr2 and mrc2.
41812         * config/arm/arm-builtins.c (MCR_QUALIFIERS): Define to...
41813         (arm_mcr_qualifiers): ... this. New.
41814         (MRC_QUALIFIERS): Define to ...
41815         (arm_mrc_qualifiers): ... this. New.
41816         (MCR_QUALIFIERS): Define to ...
41817         (arm_mcr_qualifiers): ... this. New.
41818         * config/arm/arm_acle.h (__arm_mcr, __arm_mrc, __arm_mcr2,
41819         __arm_mrc2): New.
41820         * config/arm/arm_acle_builtins.def (mcr, mcr2, mrc, mrc2): New.
41821         * config/arm/iterators.md (MCRI, mcr, MCR, MRCI, mrc, MRC): New.
41822         * config/arm/unspecs.md (VUNSPEC_MCR, VUNSPEC_MCR2, VUNSPEC_MRC,
41823         VUNSPEC_MRC2): New.
41825 2017-01-06  Andre Vieira  <andre.simoesdiasvieira@arm.com>
41827         * config/arm/arm.md (*ldc): New.
41828         (*stc): New.
41829         (<ldc>): New.
41830         (<stc>): New.
41831         * config/arm/arm.c (arm_coproc_builtin_available): Add
41832         support for ldc,ldcl,stc,stcl,ldc2,ldc2l,stc2 and stc2l.
41833         (arm_coproc_ldc_stc_legitimate_address): New.
41834         * config/arm/arm-builtins.c (arm_type_qualifiers): Add
41835         'qualifier_const_pointer'.
41836         (LDC_QUALIFIERS): Define to...
41837         (arm_ldc_qualifiers): ... this. New.
41838         (STC_QUALIFIERS): Define to...
41839         (arm_stc_qualifiers): ... this. New.
41840         * config/arm/arm-protos.h
41841         (arm_coproc_ldc_stc_legitimate_address): New.
41842         * config/arm/arm_acle.h (__arm_ldc, __arm_ldcl, __arm_stc,
41843         __arm_stcl, __arm_ldc2, __arm_ldc2l, __arm_stc2, __arm_stc2l): New.
41844         * config/arm/arm_acle_builtins.def (ldc, ldc2, ldcl, ldc2l, stc,
41845         stc2, stcl, stc2l): New.
41846         * config/arm/constraints.md (Uz): New.
41847         * config/arm/iterators.md (LDCI, STCI, ldc, stc, LDC STC): New.
41848         * config/arm/unspecs.md (VUNSPEC_LDC, VUNSPEC_LDC2, VUNSPEC_LDCL,
41849         VUNSPEC_LDC2L, VUNSPEC_STC, VUNSPEC_STC2, VUNSPEC_STCL,
41850         VUNSPEC_STC2L): New.
41852 2017-01-06  Andre Vieira  <andre.simoesdiasvieira@arm.com>
41854         * config/arm/arm.md (<cdp>): New.
41855         * config/arm/arm.c (neon_const_bounds): Rename this ...
41856         (arm_const_bounds): ... this.
41857         (arm_coproc_builtin_available): New.
41858         * config/arm/arm-builtins.c (SIMD_MAX_BUILTIN_ARGS): Increase.
41859         (arm_type_qualifiers): Add 'qualifier_unsigned_immediate'.
41860         (CDP_QUALIFIERS): Define to...
41861         (arm_cdp_qualifiers): ... this. New.
41862         (void_UP): Define.
41863         (arm_expand_builtin_args): Add case for 6 arguments.
41864         * config/arm/arm-protos.h (neon_const_bounds): Rename this ...
41865         (arm_const_bounds): ... this.
41866         (arm_coproc_builtin_available): New.
41867         * config/arm/arm_acle.h (__arm_cdp): New.
41868         (__arm_cdp2): New.
41869         * config/arm/arm_acle_builtins.def (cdp): New.
41870         (cdp2): New.
41871         * config/arm/iterators.md (CDPI,CDP,cdp): New.
41872         * config/arm/neon.md: Rename all 'neon_const_bounds' to
41873         'arm_const_bounds'.
41874         * config/arm/types.md (coproc): New.
41875         * config/arm/unspecs.md (VUNSPEC_CDP, VUNSPEC_CDP2): New.
41876         * doc/extend.texi (ACLE): Add a mention of Coprocessor intrinsics.
41877         * doc/sourcebuild.texi (arm_coproc1_ok, arm_coproc2_ok,
41878         arm_coproc3_ok, arm_coproc4_ok): Document new effective targets.
41880 2017-01-06  Andre Vieira  <andre.simoesdiasvieira@arm.com>
41882         * config/arm/arm-builtins.c (arm_unsigned_binop_qualifiers): New.
41883         (UBINOP_QUALIFIERS): New.
41884         (si_UP): Define.
41885         (acle_builtin_data): New. Change comment.
41886         (arm_builtins): Remove ARM_BUILTIN_CRC32B, ARM_BUILTIN_CRC32H,
41887         ARM_BUILTIN_CRC32W, ARM_BUILTIN_CRC32CB, ARM_BUILTIN_CRC32CH,
41888         ARM_BUILTIN_CRC32CW. Add ARM_BUILTIN_ACLE_BASE and include
41889         arm_acle_builtins.def.
41890         (ARM_BUILTIN_ACLE_PATTERN_START): Define.
41891         (arm_init_acle_builtins): New.
41892         (CRC32_BUILTIN): Remove.
41893         (bdesc_2arg): Remove entries for crc32b, crc32h, crc32w,
41894         crc32cb, crc32ch and crc32cw.
41895         (arm_init_crc32_builtins): Remove.
41896         (arm_init_builtins): Use arm_init_acle_builtins rather
41897         than arm_init_crc32_builtins.
41898         (arm_expand_acle_builtin): New.
41899         (arm_expand_builtin): Use 'arm_expand_acle_builtin'.
41900         * config/arm/arm_acle_builtins.def: New.
41902 2017-01-06  Andre Vieira  <andre.simoesdiasvieira@arm.com>
41904         * config/arm/arm-builtins.c (neon_builtin_datum): Rename to ..
41905         (arm_builtin_datum): ... this.
41906         (arm_init_neon_builtin): Rename to ...
41907         (arm_init_builtin): ... this. Add a new parameters PREFIX
41908         and USE_SIG_IN_NAME.
41909         (arm_init_neon_builtins): Replace 'arm_init_neon_builtin' with
41910         'arm_init_builtin'. Replace type 'neon_builtin_datum' with
41911         'arm_builtin_datum'.
41912         (arm_init_vfp_builtins): Likewise.
41913         (builtin_arg): Rename enum's replacing 'NEON_ARG' with
41914         'ARG_BUILTIN' and add a 'ARG_BUILTIN_NEON_MEMORY.
41915         (arm_expand_neon_args): Rename to ...
41916         (arm_expand_builtin_args): ... this. Rename builtin_arg
41917         enum values and differentiate between ARG_BUILTIN_MEMORY
41918         and ARG_BUILTIN_NEON_MEMORY.
41919         (arm_expand_neon_builtin_1): Rename to ...
41920         (arm_expand_builtin_1): ... this. Rename builtin_arg enum
41921         values, arm_expand_builtin_args and add bool parameter NEON.
41922         (arm_expand_neon_builtin): Use arm_expand_builtin_1.
41923         (arm_expand_vfp_builtin): Likewise.
41924         (NEON_MAX_BUILTIN_ARGS): Remove, it was unused.
41926 2017-01-01  Jan Hubicka  <hubicka@ucw.cz>
41928         PR middle-end/77484
41929         * predict.def (PRED_POLYMORPHIC_CALL): Set to 59.
41930         * predict.c (tree_estimate_probability_bb): Reverse direction of
41931         polymorphic call predictor.
41933 2017-01-06  David Malcolm  <dmalcolm@redhat.com>
41935         * passes.c (execute_one_pass): Split out pass-skipping logic into...
41936         (determine_pass_name_match): ...this new function and...
41937         (should_skip_pass_p): ...this new function.
41939 2017-01-06  Nathan Sidwell  <nathan@acm.org>
41941         * ipa-visibility.c (function_and_variable_visibility): Reformat
41942         comments and long lines.  Remove extrneous if.
41943         * symtab.c (symtab_node::make_decl_local): Fix code format.
41944         (symtab_node::set_section_for_node): Fix comment typo.
41946 2017-01-06  Martin Liska  <mliska@suse.cz>
41948         PR bootstrap/79003
41949         * lra-constraints.c: Rename invariant to lra_invariant.
41950         * predict.c (set_even_probabilities): Initialize e to NULL.
41952 2017-01-05  Martin Sebor  <msebor@redhat.com>
41954         PR tree-optimization/78910
41955         * gimple-ssa-sprintf.c (tree_digits): Add an argument.
41956         (format_integer): Correct off-by-one error in the handling
41957         of precision with negative numbers in signed conversions..
41959 2017-01-05  Eric Botcazou  <ebotcazou@adacore.com>
41961         * doc/invoke.texi (C Dialect Options): Adjust -fsso-struct entry.
41963 2017-01-05  Jakub Jelinek  <jakub@redhat.com>
41965         PR tree-optimization/71016
41966         * tree-ssa-phiopt.c (tree_ssa_phiopt_worker): Pass cond_stmt to
41967         factor_out_conditional_conversion.  Formatting fix.
41968         (factor_out_conditional_conversion): Add cond_stmt argument.
41969         If arg1 is INTEGER_CST, punt if new_arg0 is not any operand of
41970         cond_stmt and if arg0_def_stmt is not the only stmt in its bb.
41971         Formatting fix.
41973 2017-01-05  David Malcolm  <dmalcolm@redhat.com>
41975         * Makefile.in (OBJS): Add read-md.o, read-rtl.o,
41976         read-rtl-function.o, and selftest-rtl.o.
41977         * config/aarch64/aarch64.c: Include selftest.h and selftest-rtl.h.
41978         (selftest::aarch64_test_loading_full_dump): New function.
41979         (selftest::aarch64_run_selftests): New function.
41980         (TARGET_RUN_TARGET_SELFTESTS): Wire it up to
41981         selftest::aarch64_run_selftests.
41982         * config/i386/i386.c
41983         (selftest::ix86_test_loading_dump_fragment_1): New function.
41984         (selftest::ix86_test_loading_call_insn): New function.
41985         (selftest::ix86_test_loading_full_dump): New function.
41986         (selftest::ix86_test_loading_unspec): New function.
41987         (selftest::ix86_run_selftests): Call the new functions.
41988         * emit-rtl.c (maybe_set_max_label_num): New function.
41989         * emit-rtl.h (maybe_set_max_label_num): New decl.
41990         * function.c (instantiate_decls): Guard call to
41991         instantiate_decls_1 with if (DECL_INITIAL (fndecl)).
41992         * function-tests.c (selftest::verify_three_block_rtl_cfg): Remove
41993         "static".
41994         * gensupport.c (gen_reader::gen_reader): Pass "false"
41995         for new "compact" param of rtx_reader.
41996         * print-rtl.c (rtx_writer::print_rtx_operand): Print "(nil)"
41997         rather than an empty string for NULL strings.
41998         * read-md.c: Potentially include config.h rather than bconfig.h.
41999         Wrap include of errors.h with #ifdef GENERATOR_FILE.
42000         (have_error): New global, copied from errors.c.
42001         (md_reader::read_name): Rename to...
42002         (md_reader::read_name_1): ...this, adding "out_loc" param,
42003         and converting "missing name or number" to returning false, rather
42004         than failing.
42005         (md_reader::read_name): Reimplement in terms of read_name_1.
42006         (md_reader::read_name_or_nil): New function.
42007         (md_reader::read_string): Handle "(nil)" by returning NULL.
42008         (md_reader::md_reader): Add new param "compact".
42009         (md_reader::read_md_files): Wrap with #ifdef GENERATOR_FILE.
42010         (md_reader::read_file): New method.
42011         * read-md.h (md_reader::md_reader): Add new param "compact".
42012         (md_reader::read_file): New method.
42013         (md_reader::is_compact): New accessor.
42014         (md_reader::read_name): Convert return type from void to file_location.
42015         (md_reader::read_name_or_nil): New decl.
42016         (md_reader::read_name_1): New decl.
42017         (md_reader::m_compact): New field.
42018         (noop_reader::noop_reader): Pass "false" for new "compact" param
42019         of rtx_reader.
42020         (rtx_reader::rtx_reader): Add new "compact" param.
42021         (rtx_reader::read_rtx_operand): Make virtual and convert return
42022         type from void to rtx.
42023         (rtx_reader::read_until): New decl.
42024         (rtx_reader::handle_any_trailing_information): New virtual function.
42025         (rtx_reader::postprocess): New virtual function.
42026         (rtx_reader::finalize_string): New virtual function.
42027         (rtx_reader::m_in_call_function_usage): New field.
42028         (rtx_reader::m_reuse_rtx_by_id): New field.
42029         * read-rtl-function.c: New file.
42030         * selftest-rtl.c (selftest::assert_rtx_ptr_eq_at): New function.
42031         * selftest-rtl.h (ASSERT_RTX_PTR_EQ): New macro.
42032         (selftest::verify_three_block_rtl_cfg): New decl.
42033         * read-rtl-function.h: New file.
42034         * read-rtl.c: Potentially include config.h rather than bconfig.h.
42035         For host, include function.h, memmodel.h, and emit-rtl.h.
42036         (one_time_initialization): New function.
42037         (struct compact_insn_name): New struct.
42038         (compact_insn_names): New array.
42039         (find_code): Handle insn codes in compact dumps.
42040         (apply_subst_iterator): Wrap with #ifdef GENERATOR_FILE.
42041         (bind_subst_iter_and_attr): Likewise.
42042         (add_condition_to_string): Likewise.
42043         (add_condition_to_rtx): Likewise.
42044         (apply_attribute_uses): Likewise.
42045         (add_current_iterators): Likewise.
42046         (apply_iterators): Likewise.
42047         (initialize_iterators): Guard usage of apply_subst_iterator with
42048         #ifdef GENERATOR_FILE.
42049         (read_conditions): Wrap with #ifdef GENERATOR_FILE.
42050         (md_reader::read_mapping): Likewise.
42051         (add_define_attr_for_define_subst): Likewise.
42052         (add_define_subst_attr): Likewise.
42053         (read_subst_mapping): Likewise.
42054         (check_code_iterator): Likewise.
42055         (rtx_reader::read_rtx): Likewise.  Move one-time initialization
42056         logic to...
42057         (one_time_initialization): New function.
42058         (rtx_reader::read_until): New method.
42059         (read_flags): New function.
42060         (parse_reg_note_name): New function.
42061         (rtx_reader::read_rtx_code): Initialize "iterator" to NULL.
42062         Handle reuse_rtx ids.
42063         Wrap iterator lookup within #ifdef GENERATOR_FILE.
42064         Add parsing support for RTL dumps, mirroring the special-cases in
42065         print_rtx, by calling read_flags, reading REG_NOTE names, INSN_UID
42066         values, and calling handle_any_trailing_information.
42067         (rtx_reader::read_rtx_operand): Convert return type from void
42068         to rtx, returning return_rtx.  Handle case 'e'.  Call
42069         finalize_string on XSTR and XTMPL fields.
42070         (rtx_reader::read_nested_rtx):  Handle dumps in which trailing
42071          "(nil)" values were omitted.  Call the postprocess vfunc on the
42072         return_rtx.
42073         (rtx_reader::rtx_reader): Add new "compact" param and pass to base
42074         class ctor.  Initialize m_in_call_function_usage.  Call
42075         one_time_initialization.
42076         * rtl-tests.c (selftest::test_uncond_jump): Call
42077         set_new_first_and_last_insn.
42078         * rtl.h (read_rtx): Wrap decl with #ifdef GENERATOR_FILE.
42079         * selftest-rtl.c: New file.
42080         * selftest-rtl.h (class selftest::rtl_dump_test): New class.
42081         (selftest::get_insn_by_uid): New decl.
42082         * selftest-run-tests.c (selftest::run_tests): Call
42083         read_rtl_function_c_tests.
42084         * selftest.h  (selftest::read_rtl_function_c_tests): New decl.
42085         * tree-dfa.c (ssa_default_def): Return NULL_TREE for rtl function
42086         dumps.
42088 2017-01-05  Uros Bizjak  <ubizjak@gmail.com>
42090         * config/i386/i386.md (*testqi_ext_3): No need to handle memory
42091         operands in a special way.  Assert that pos+len <= mode precision.
42093 2017-01-05  Jakub Jelinek  <jakub@redhat.com>
42095         * common.opt (fvect-cost-model): Remove RejectNegative flag, use
42096         3 argument Alias with unlimited for the negative form.
42097         (fno-vect-cost-model): Removed.
42099 2017-01-05  Martin Liska  <mliska@suse.cz>
42101         * hsa-gen.c (gen_hsa_divmod): New function.
42102         (gen_hsa_insn_for_internal_fn_call): Use the function for IFN_DIVMOD.
42104 2017-01-05  Martin Liska  <mliska@suse.cz>
42106         PR pch/78970
42107         * gcc.c (lookup_compiler): Reject '-' filename for a precompiled
42108         header.
42110 2017-01-05  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
42112         * config/s390/s390.c (s390_expand_setmem): Unroll the loop for
42113         small constant length operands.
42115 2017-01-05  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
42117         * config/s390/s390.c (s390_expand_setmem): Avoid overlapping bytes
42118         between loop iterations.
42120 2017-01-05  Martin Liska  <mliska@suse.cz>
42122         PR sanitizer/78815
42123         * gimplify.c (gimplify_decl_expr): Compare to
42124         asan_poisoned_variables instread of checking flags.
42125         (gimplify_target_expr): Likewise.
42126         (gimplify_expr): Likewise.
42127         (gimplify_function_tree): Conditionally initialize
42128         asan_poisoned_variables.
42130 2017-01-04  Jeff Law  <law@redhat.com>
42132         PR tree-optimizatin/78812
42133         * rtl.h (contains_mem_rtx_p): Prototype.
42134         * ifcvt.c (containts_mem_rtx_p): Move from here to...
42135         * rtlanal.c (contains_mem_rtx_p): Here and remove static linkage.
42136         * gcse.c (prune_expressions): Use contains_mem_rtx_p to discover
42137         and prune MEMs that are not at the toplevel of a SET_SRC rtx.  Look
42138         through ZERO_EXTEND and SIGN_EXTEND when trying to avoid pruning MEMs.
42140 2017-01-04  Alexandre Oliva  <aoliva@redhat.com>
42142         * input.c (assert_char_at_range): Default-initialize actual_range.
42144 2017-01-04  Alexandre Oliva  <aoliva@redhat.com>
42146         * df-scan.c (df_ref_create_structure): Make regno unsigned,
42147         to match the caller.
42149 2017-01-04  Alexandre Oliva  <aoliva@redhat.com>
42151         * cfgexpand.c (expand_gimple_basic_block): Disregard debug
42152         insns after final jump in test to emit dummy move.
42154 2017-01-04  Alexandre Oliva  <aoliva@redhat.com>
42156         * gimple-iterator.h (gsi_one_nondebug_before_end_p): New.
42157         * tree-eh.c (cleanup_empty_eh): Skip more debug stmts.
42159 2017-01-04  Alexandre Oliva  <aoliva@redhat.com>
42161         * multiple_target.c (create_dispatcher_calls): Init e_next.
42162         * tree-ssa-loop-split.c (split_loop): Init border.
42163         * tree-vect-loop.c (vect_determine_vectorization_factor): Init
42164         scalar_type.
42166 2017-01-04  Michael Meissner  <meissner@linux.vnet.ibm.com>
42168         PR target/71977
42169         PR target/70568
42170         PR target/78823
42171         * config/rs6000/predicates.md (sf_subreg_operand): New predicate.
42172         (altivec_register_operand): Do not return true if the operand
42173         contains a SUBREG mixing SImode and SFmode.
42174         (vsx_register_operand): Likewise.
42175         (vsx_reg_sfsubreg_ok): New predicate.
42176         (vfloat_operand): Do not return true if the operand contains a
42177         SUBREG mixing SImode and SFmode.
42178         (vint_operand): Likewise.
42179         (vlogical_operand): Likewise.
42180         (gpc_reg_operand): Likewise.
42181         (int_reg_operand): Likewise.
42182         * config/rs6000/rs6000-protos.h (valid_sf_si_move): Add declaration.
42183         * config/rs6000/rs6000.c (valid_sf_si_move): New function to
42184         determine if a MOVSI or MOVSF operation contains SUBREGs that mix
42185         SImode and SFmode.
42186         (rs6000_emit_move_si_sf_subreg): New helper function.
42187         (rs6000_emit_move): Call rs6000_emit_move_si_sf_subreg to possbily
42188         fixup SUBREGs involving SImode and SFmode.
42189         * config/rs6000/vsx.md (SFBOOL_*): New constants that are operand
42190         numbers for the new peephole2 optimization.
42191         (peephole2 for SFmode unions): New peephole2 to optimize cases in
42192         the GLIBC math library that do AND/IOR/XOR operations on single
42193         precision floating point.
42194         * config/rs6000/rs6000.h (TARGET_NO_SF_SUBREG): New internal
42195         target macros to say whether we need to avoid SUBREGs mixing
42196         SImode and SFmode.
42197         (TARGET_ALLOW_SF_SUBREG): Likewise.
42198         * config/rs6000/rs6000.md (UNSPEC_SF_FROM_SI): New unspecs.
42199         (UNSPEC_SI_FROM_SF): Likewise.
42200         (iorxor): Change spacing.
42201         (and_ior_xor): New iterator for AND, IOR, and XOR.
42202         (movsi_from_sf): New insns for SImode/SFmode SUBREG support.
42203         (movdi_from_sf_zero_ext): Likewise.
42204         (mov<mode>_hardfloat, FMOVE32 iterator): Use register_operand
42205         instead of gpc_reg_operand.  Add SImode/SFmode SUBREG support.
42206         (movsf_from_si): New insn for SImode/SFmode SUBREG support.
42207         (fma<mode>4): Use gpc_reg_operand instead of register_operand.
42208         (fms<mode>4): Likewise.
42209         (fnma<mode>4): Likewise.
42210         (fnms<mode>4): Likewise.
42211         (nfma<mode>4): Likewise.
42212         (nfms<mode>4): Likewise.
42214 2017-01-04  Marek Polacek  <polacek@redhat.com>
42216         PR c++/64767
42217         * doc/invoke.texi: Document -Wpointer-compare.
42219 2017-01-04  Jakub Jelinek  <jakub@redhat.com>
42221         * optc-gen.awk: Emit #error for -W*/-f*/-m* Enum without
42222         RejectNegative.
42224         * dwarf2out.c (output_loc_list): Don't throw away 64K+ location
42225         descriptions for -gdwarf-5 and emit them as uleb128 instead of
42226         2-byte data.
42228 2017-01-04  Kelvin Nilsen  <kelvin@gcc.gnu.org>
42230         PR target/78056
42231         * doc/sourcebuild.texi (PowerPC-specific attributes): Add
42232         documentation of the powerpc_popcntb_ok attribute.
42233         * config/rs6000/rs6000.c (rs6000_option_override_internal): Add
42234         code to issue warning messages if a requested CPU configuration is
42235         not supported by the binary (assembler and loader) toolchain.
42236         (spe_init_builtins): Add two assertions to prevent ICE if attempt is
42237         made to define a built-in function that has been disabled.
42238         (paired_init_builtins): Add assertion to prevent ICE if attempt is
42239         made to define a built-in function that has been disabled.
42240         (altivec_init_builtins): Add comment explaining why definition
42241         of the DST built-in functions is not preceded by an assertion
42242         check.  Add assertions to prevent ICE if attempts are made to
42243         define an altivec predicate or an abs* built-in function that has
42244         been disabled.
42245         (htm_init_builtins): Add comment explaining why definition of the
42246         htm built-in functions is not preceded by an assertion check.
42248 2017-01-04  Jeff Law  <law@redhat.com>
42250         PR tree-optimizatin/67955
42251         * tree-ssa-alias.c (same_addr_size_stores_p): Check offsets first.
42252         Allow any SSA_VAR_P as the base objects.  Use integer_zerop.  Verify
42253         the points-to solution does not include pt_null.  Use DECL_PT_UID
42254         unconditionally.
42256 2017-01-04  Uros Bizjak  <ubizjak@gmail.com>
42258         * config/i386/i386.md (HI/SImode test with imm to QImode splitters):
42259         Use gen_int_mode instead of gen_lopwart for const_int operands.
42261 2017-01-04  Jakub Jelinek  <jakub@redhat.com>
42263         PR tree-optimization/71563
42264         * match.pd: Simplify X << Y into X if Y is known to be 0 or
42265         out of range value - has low bits known to be zero.
42267 2017-01-04  Alan Modra  <amodra@gmail.com>
42269         * Makefile.in (aclocal_deps): Update and order as per aclocal.m4.
42270         * configure: Regenerate.
42271         * config.in: Regenerate.
42273 2017-01-04  Jakub Jelinek  <jakub@redhat.com>
42275         PR bootstrap/77569
42276         * input.c (ebcdic_execution_charset::on_error): Don't use strstr for
42277         a substring of the message, but strcmp with the whole message.  Ifdef
42278         ENABLE_NLS, translate the message first using dgettext.
42280 2017-01-03  Jeff Law  <law@redhat.com>
42282         PR tree-optimizatin/78856
42283         * tree-ssa-threadupdate.c: Include tree-vectorizer.h.
42284         (mark_threaded_blocks): Remove code to truncate thread paths that
42285         cross multiple loop headers.  Instead invalidate the cached loop
42286         iteration information and handle case of a thread path walking
42287         into an irreducible region.
42289 2017-01-03  Michael Meissner  <meissner@linux.vnet.ibm.com>
42291         PR target/78900
42292         * config/rs6000/rs6000.c (rs6000_split_signbit): Change some
42293         assertions.  Add support for doing the signbit if the IEEE 128-bit
42294         floating point value is in a GPR.
42295         * config/rs6000/rs6000.md (Fsignbit): Delete.
42296         (signbit<mode>2_dm): Delete using <Fsignbit> and just use "wa".
42297         Update the length attribute if the value is in a GPR.
42298         (signbit<mode>2_dm_<su>ext): Add combiner pattern to eliminate
42299         the sign or zero extension instruction, since the value is always 0/1.
42300         (signbit<mode>2_dm2): Delete using <Fsignbit>.
42302         PR target/78953
42303         * config/rs6000/vsx.md (vsx_extract_<mode>_store_p9): If we are
42304         extracting SImode to a GPR register so that we can generate a
42305         store, limit the vector to be in a traditional Altivec register
42306         for the vextuwrx instruction.
42308 2017-01-03  Ian Lance Taylor  <iant@google.com>
42310         * godump.c (go_format_type): Treat ENUMERAL_TYPE like INTEGER_TYPE.
42312 2017-01-03  Martin Sebor  <msebor@redhat.com>
42314         PR tree-optimization/78696
42315         * gimple-ssa-sprintf.c (format_floating): Correct handling of
42316         precision.  Use MPFR for %f for greater fidelity.  Correct handling
42317         of %g.
42318         (pass_sprintf_length::compute_format_length): Set width and precision
42319         specified by asrerisk to void_node for vararg functions.
42320         (try_substitute_return_value): Adjust dump output.
42322 2017-01-03  David Edelsohn  <dje.gcc@gmail.com>
42324         * doc/invoke.texi (RS6000 options): LRA is enabled by default.
42326 2017-01-03  Eric Botcazou  <ebotcazou@adacore.com>
42328         * doc/invoke.texi (SPARC options): Document -mlra as the default.
42329         * config/sparc/sparc.c (sparc_option_override): Force LRA unless
42330         -mlra/-mno-lra was passed to the compiler.
42332 2017-01-03  James Cowgill  <James.Cowgill@imgtec.com>
42334         PR rtl-optimization/65618
42335         * emit-rtl.c (try_split): Move initialization of "before" and
42336         "after" to just before the call to emit_insn_after_setloc.
42338 2017-01-03  Gerald Pfeifer  <gerald@pfeifer.com>
42340         * doc/md.texi (Standard Names): Remove reference to Java frontend.
42342 2017-01-03  Pierre-Marie de Rodat  <derodat@adacore.com>
42344         * dwarf2out.c (gen_enumeration_type_die): When
42345         -gno-strict-dwarf, add a DW_AT_encoding attribute.
42347 2017-01-03  Jakub Jelinek  <jakub@redhat.com>
42349         PR tree-optimization/78965
42350         * gimple-ssa-sprintf.c (pass_sprintf_length::compute_format_length):
42351         Change first argument from const call_info & to call_info &.  For %n
42352         set info.nowrite to false.
42354         PR middle-end/78901
42355         * gimple-ssa-sprintf.c (try_substitute_return_value): Don't change
42356         possibly throwing calls.
42358         * genmatch.c (dt_node::gen_kids_1): If generic_exprs include SSA_NAME
42359         and exprs_len || fns_len, emit the code for SSA_NAME next to the exprs
42360         and fns handling, rather than in a separate case SSA_NAME.
42362 2017-01-02  Jeff Law  <law@redhat.com>
42364         * config/darwin-driver.c (darwin_driver_init): Const-correctness
42365         fixes for first_period and second_period variables.
42367 2017-01-02  Uros Bizjak  <ubizjak@gmail.com>
42369         PR target/78967
42370         * config/i386/i386.md (UNSPEC_NOREX_MEM): New unspec.
42371         (*insvqi_1): New insn pattern.
42372         (*insvqi_1_mem_rex64): Ditto.
42373         (*insvqi_2): Ditto.
42374         (*insvqi_3): Rename from *insvqi.
42376         (*extzvqi_mem_rex64): Add UNSPEC_NOREX_MEM tag.
42378 2017-01-02  Gerald Pfeifer  <gerald@pfeifer.com>
42380         * doc/cfg.texi (Edges): Remove reference to Java.
42381         (Maintaining the CFG): Ditto.
42383 2017-01-01  Jan Hubicka  <hubicka@ucw.cz>
42385         PR middle-end/77674
42386         * symtab.c (symtab_node::binds_to_current_def_p): Fix handling of
42387         transparent aliases.
42389 2017-01-01  Jan Hubicka  <hubicka@ucw.cz>
42391         PR middle-end/77484
42392         * predict.def (PRED_CALL): Update hitrate.
42393         (PRED_INDIR_CALL, PRED_POLYMORPHIC_CALL): New predictors.
42394         * predict.c (tree_estimate_probability_bb): Split CALL predictor
42395         into direct/indirect/polymorphic variants.
42397 2017-01-01  Jakub Jelinek  <jakub@redhat.com>
42399         Update copyright years.
42401         * gcc.c (process_command): Update copyright notice dates.
42402         * gcov-dump.c (print_version): Ditto.
42403         * gcov.c (print_version): Ditto.
42404         * gcov-tool.c (print_version): Ditto.
42405         * gengtype.c (create_file): Ditto.
42406         * doc/cpp.texi: Bump @copying's copyright year.
42407         * doc/cppinternals.texi: Ditto.
42408         * doc/gcc.texi: Ditto.
42409         * doc/gccint.texi: Ditto.
42410         * doc/gcov.texi: Ditto.
42411         * doc/install.texi: Ditto.
42412         * doc/invoke.texi: Ditto.
42414 Copyright (C) 2017 Free Software Foundation, Inc.
42416 Copying and distribution of this file, with or without modification,
42417 are permitted in any medium without royalty provided the copyright
42418 notice and this notice are preserved.